mcklogd: fixed termination method of mcklogd

Conflicts:
	configure
	kernel/process.c

Makefile.in

@@ -1,5 +1,6 @@
 TARGET = @TARGET@
 SBINDIR = @SBINDIR@
+ETCDIR = @ETCDIR@
 MANDIR = @MANDIR@
 
 all::
@@ -48,6 +49,10 @@ install::
 		mkdir -p -m 755 $(SBINDIR); \
 		install -m 755 arch/x86/tools/mcreboot-smp-x86.sh $(SBINDIR)/mcreboot.sh; \
 		install -m 755 arch/x86/tools/mcstop+release-smp-x86.sh $(SBINDIR)/mcstop+release.sh; \
+		install -m 755 arch/x86/tools/eclair-dump-backtrace.exp $(SBINDIR)/eclair-dump-backtrace.exp;\
+		mkdir -p -m 755 $(ETCDIR); \
+		install -m 644 arch/x86/tools/irqbalance_mck.service $(ETCDIR)/irqbalance_mck.service; \
+		install -m 644 arch/x86/tools/irqbalance_mck.in $(ETCDIR)/irqbalance_mck.in; \
 		mkdir -p -m 755 $(MANDIR)/man1; \
 		install -m 644 arch/x86/tools/mcreboot.1 $(MANDIR)/man1/mcreboot.1; \
 		;; \

arch/x86/kernel/cpu.c

@@ -148,7 +148,7 @@ extern char page_fault[], general_protection_exception[];
 extern char debug_exception[], int3_exception[];
 
 uint64_t boot_pat_state = 0;
-int no_turbo = 0; /* May be updated by early parsing of kargs */
+int no_turbo = 1; /* May be updated by early parsing of kargs */
 
 extern int num_processors; /* kernel/ap.c */
 struct pvclock_vsyscall_time_info *pvti = NULL;
@@ -181,6 +181,8 @@ static void init_idt(void)
 }
 
 static int xsave_available = 0;
+static int xsave_size = 0;
+static uint64_t xsave_mask = 0x0;
 
 void init_fpu(void)
 {
@@ -224,6 +226,26 @@ void init_fpu(void)
 		xsetbv(0, reg);
 		dkprintf("init_fpu(): AVX init: XCR0 = 0x%016lX\n", reg);
 	}
+	if(xsave_available){
+		unsigned long eax;
+		unsigned long ebx;
+		unsigned long ecx;
+		unsigned long edx;
+		asm volatile("cpuid" : "=a"(eax),"=b"(ebx),"=c"(ecx),"=d"(edx) : "a" (0x0d), "c" (0x00));
+		xsave_size = ecx;
+		dkprintf("init_fpu(): xsave_size = %d\n", xsave_size);
+
+		if ((eax & (1 << 5)) && (eax & (1 << 6)) && (eax & (1 << 7))) {
+			/* Set xcr0[7:5] to enable avx-512 ops */
+			reg = xgetbv(0);
+			reg |= 0xe6;
+			xsetbv(0, reg);
+			dkprintf("init_fpu(): AVX-512 init: XCR0 = 0x%016lX\n", reg);
+		}
+	}
+
+	xsave_mask = xgetbv(0);
+	dkprintf("init_fpu(): xsave_mask = 0x%016lX\n", xsave_mask);
 
 	/* TODO: set MSR_IA32_XSS to enable xsaves/xrstors */
 
@@ -234,6 +256,17 @@ void init_fpu(void)
 	asm volatile("finit");
 }
 
+int
+get_xsave_size()
+{
+	return xsave_size;
+}
+
+uint64_t get_xsave_mask()
+{
+	return xsave_mask;
+}
+
 void reload_gdt(struct x86_desc_ptr *gdt_ptr)
 {
 	asm volatile("pushq %1\n"
@@ -811,6 +844,25 @@ void set_signal(int sig, void *regs, struct siginfo *info);
 void check_signal(unsigned long, void *, int);
 extern void tlb_flush_handler(int vector);
 
+void __show_stack(uintptr_t *sp) {
+	while (((uintptr_t)sp >= 0xffff800000000000)
+			&& ((uintptr_t)sp <  0xffffffff80000000)) {
+		uintptr_t fp;
+		uintptr_t ip;
+
+		fp = sp[0];
+		ip = sp[1];
+		kprintf("IP: %016lx, SP: %016lx, FP: %016lx\n", ip, (uintptr_t)sp, fp);
+		sp = (void *)fp;
+	}
+	return;
+}
+
+void show_context_stack(uintptr_t *rbp) {
+	__show_stack(rbp);
+	return;
+}
+
 void handle_interrupt(int vector, struct x86_user_context *regs)
 {
 	struct ihk_mc_interrupt_handler *h;
@@ -883,19 +935,45 @@ void handle_interrupt(int vector, struct x86_user_context *regs)
 		dkprintf("timer[%lu]: CPU_FLAG_NEED_RESCHED \n", rdtsc());
 	}
 	else if (vector == LOCAL_PERF_VECTOR) {
+		struct siginfo info;
 		unsigned long value;
+		struct thread *thread = cpu_local_var(current);
+        	struct process *proc = thread->proc;
+		long irqstate;
+		struct mckfd *fdp;
+
+		lapic_write(LAPIC_LVTPC, LOCAL_PERF_VECTOR);
 
 		value = rdmsr(MSR_PERF_GLOBAL_STATUS);
 		wrmsr(MSR_PERF_GLOBAL_OVF_CTRL, value);
 		wrmsr(MSR_PERF_GLOBAL_OVF_CTRL, 0);
-		//TODO: counter overflow signal
-		//set_signal(0x1d, regs, NULL); // SIGIO
+
+		irqstate = ihk_mc_spinlock_lock(&proc->mckfd_lock);
+	        for(fdp = proc->mckfd; fdp; fdp = fdp->next) {
+			if(fdp->sig_no > 0)
+                	        break;
+		}
+	        ihk_mc_spinlock_unlock(&proc->mckfd_lock, irqstate);
+
+		if(fdp) {
+			memset(&info, '\0', sizeof info);
+			info.si_signo = fdp->sig_no;
+			info._sifields._sigfault.si_addr = (void *)regs->gpr.rip;
+			info._sifields._sigpoll.si_fd = fdp->fd;
+			set_signal(fdp->sig_no, regs, &info); 
+		}
+		else {
+			set_signal(SIGIO, regs, NULL);
+		}
 	}
 	else if (vector >= IHK_TLB_FLUSH_IRQ_VECTOR_START && 
 	         vector < IHK_TLB_FLUSH_IRQ_VECTOR_END) {
 
 			tlb_flush_handler(vector);
 	} 
+	else if (vector == 133) {
+		show_context_stack((uintptr_t *)regs->gpr.rbp);
+	}
 	else {
 		list_for_each_entry(h, &handlers[vector - 32], list) {
 			if (h->func) {
@@ -998,9 +1076,8 @@ unhandled_page_fault(struct thread *thread, void *fault_addr, void *regs)
 	unsigned long error = ((struct x86_user_context *)regs)->gpr.error;
 
 	irqflags = kprintf_lock();
-	dkprintf("[%d] Page fault for 0x%lX\n",
-			ihk_mc_get_processor_id(), address);
-	dkprintf("%s for %s access in %s mode (reserved bit %s set), "
+	__kprintf("Page fault for 0x%lx\n", address);
+	__kprintf("%s for %s access in %s mode (reserved bit %s set), "
 			"it %s an instruction fetch\n",
 			(error & PF_PROT ? "protection fault" : "no page found"),
 			(error & PF_WRITE ? "write" : "read"),
@@ -1012,18 +1089,22 @@ unhandled_page_fault(struct thread *thread, void *fault_addr, void *regs)
 	list_for_each_entry(range, &vm->vm_range_list, list) {
 		if (range->start <= address && range->end > address) {
 			found = 1;
-			dkprintf("address is in range, flag: 0x%X! \n",
+			__kprintf("address is in range, flag: 0x%lx\n",
 					range->flag);
 			ihk_mc_pt_print_pte(vm->address_space->page_table, (void*)address);
 			break;
 		}
 	}
 	if (!found) {
-		dkprintf("address is out of range! \n");
+		__kprintf("address is out of range! \n");
 	}
 
 	kprintf_unlock(irqflags);
 
+	if (!(error & PF_USER)) {
+		panic("panic: kernel mode PF");
+	}
+
 	/* TODO */
 	ihk_mc_debug_show_interrupt_context(regs);
 
@@ -1494,7 +1575,8 @@ release_fp_regs(struct thread *thread)
 	if (thread && !thread->fp_regs)
 		return;
 
-	pages = (sizeof(fp_regs_struct) + 4095) >> 12;
+	pages = (xsave_size + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	dkprintf("release_fp_regs: pages=%d\n", pages);
 	ihk_mc_free_pages(thread->fp_regs, pages);
 	thread->fp_regs = NULL;
 }
@@ -1508,7 +1590,8 @@ save_fp_regs(struct thread *thread)
 	int	pages;
 
 	if (!thread->fp_regs) {
-		pages = (sizeof(fp_regs_struct) + 4095) >> 12;
+		pages = (xsave_size + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+		dkprintf("save_fp_regs: pages=%d\n", pages);
 		thread->fp_regs = ihk_mc_alloc_pages(pages, IHK_MC_AP_NOWAIT);
 
 		if (!thread->fp_regs) {
@@ -1517,14 +1600,15 @@ save_fp_regs(struct thread *thread)
 		}
 
 		memset(thread->fp_regs, 0, sizeof(fp_regs_struct));
+		memset(thread->fp_regs, 0, pages * PAGE_SIZE);
 	}
 
 	if (xsave_available) {
 		unsigned int low, high;
 
-		/* Request full save of x87, SSE and AVX states */
-		low = 0x7;
-		high = 0;
+		/* Request full save of x87, SSE, AVX and AVX-512 states */
+		low = (unsigned int)xsave_mask;
+		high = (unsigned int)(xsave_mask >> 32);
 
 		asm volatile("xsave %0" : : "m" (*thread->fp_regs), "a" (low), "d" (high) 
 			: "memory");
@@ -1546,9 +1630,9 @@ restore_fp_regs(struct thread *thread)
 	if (xsave_available) {
 		unsigned int low, high;
 
-		/* Request full restore of x87, SSE and AVX states */
-		low = 0x7;
-		high = 0;
+		/* Request full restore of x87, SSE, AVX and AVX-512 states */
+		low = (unsigned int)xsave_mask;
+		high = (unsigned int)(xsave_mask >> 32);
 
 		asm volatile("xrstor %0" : : "m" (*thread->fp_regs), 
 				"a" (low), "d" (high));
@@ -1678,7 +1762,7 @@ int arch_setup_pvclock(void)
 	npages = (size + PAGE_SIZE - 1) / PAGE_SIZE;
 	pvti_npages = npages;
 
-	pvti = allocate_pages(npages, IHK_MC_AP_NOWAIT);
+	pvti = ihk_mc_alloc_pages(npages, IHK_MC_AP_NOWAIT);
 	if (!pvti) {
 		ekprintf("arch_setup_pvclock: allocate_pages failed.\n");
 		return -ENOMEM;
@@ -1708,44 +1792,6 @@ void arch_start_pvclock(void)
 	return;
 } /* arch_start_pvclock() */
 
-static struct cpu_mapping *cpu_mapping = NULL;
-
-int arch_get_cpu_mapping(struct cpu_mapping **buf, int *nelemsp)
-{
-	int error;
-	size_t size;
-	int npages;
-	struct cpu_mapping *mapping;
-	int cpu;
-	struct x86_cpu_local_variables *v;
-
-	if (!cpu_mapping) {
-		size = sizeof(*mapping) * num_processors;
-		npages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
-		mapping = allocate_pages(npages, IHK_MC_AP_NOWAIT);
-		if (!mapping) {
-			error = -ENOMEM;
-			ekprintf("arch_get_cpu_mapping:allocate_pages failed. %d\n", error);
-			goto out;
-		}
-
-		for (cpu = 0; cpu < num_processors; ++cpu) {
-			v = get_x86_cpu_local_variable(cpu);
-			mapping[cpu].cpu_number = cpu;
-			mapping[cpu].hw_id = v->apic_id;
-		}
-
-		cpu_mapping = mapping;
-	}
-
-	error = 0;
-	*buf = cpu_mapping;
-	*nelemsp = num_processors;
-
-out:
-	return error;
-} /* arch_get_cpu_mapping() */
-
 #define KVM_CPUID_SIGNATURE	0x40000000
 
 int running_on_kvm(void) {
@@ -1767,4 +1813,43 @@ int running_on_kvm(void) {
 	return 0;
 }
 
+void
+mod_nmi_ctx(void *nmi_ctx, void (*func)())
+{
+	unsigned long *l = nmi_ctx;
+	int i;
+	unsigned long flags;
+
+//struct x86_cpu_local_variables *v;
+
+//if(!ihk_mc_get_processor_id()) {
+//v = get_x86_this_cpu_local();
+//}
+	asm volatile("pushf; pop %0" : "=r"(flags) : : "memory", "cc");
+	for (i = 0; i < 22; i++)
+		l[i] = l[i + 5];
+	l[i++] = (unsigned long)func;		// return address
+	l[i++] = 0x20;				// KERNEL CS
+	l[i++] = flags & ~RFLAGS_IF;		// rflags (disable interrupt)
+	l[i++] = (unsigned long)(l + 27);	// ols rsp
+	l[i++] = 0x28;				// KERNEL DS
+}
+
+int arch_cpu_read_write_register(
+		struct mcctrl_os_cpu_register *desc,
+		enum mcctrl_os_cpu_operation op)
+{
+	if (op == MCCTRL_OS_CPU_READ_REGISTER) {
+		desc->val = rdmsr(desc->addr);
+	}
+	else if (op == MCCTRL_OS_CPU_WRITE_REGISTER) {
+		wrmsr(desc->addr, desc->val);
+	}
+	else {
+		return -1;
+	}
+
+	return 0;
+}
+
 /*** end of file ***/

arch/x86/kernel/include/arch-bitops.h

@@ -13,6 +13,8 @@
 #ifndef HEADER_X86_COMMON_ARCH_BITOPS_H
 #define HEADER_X86_COMMON_ARCH_BITOPS_H
 
+#define ARCH_HAS_FAST_MULTIPLIER 1
+
 static inline int fls(int x)
 {
 	int r;

arch/x86/kernel/include/arch-lock.h

@@ -131,6 +131,7 @@ static void __ihk_mc_spinlock_unlock(ihk_spinlock_t *lock, unsigned long flags)
 typedef struct mcs_lock_node {
 	unsigned long locked;
 	struct mcs_lock_node *next;
+	unsigned long irqsave;
 } __attribute__((aligned(64))) mcs_lock_node_t;
 
 static void mcs_lock_init(struct mcs_lock_node *node)
@@ -139,7 +140,7 @@ static void mcs_lock_init(struct mcs_lock_node *node)
 	node->next = NULL;
 }
 
-static void mcs_lock_lock(struct mcs_lock_node *lock,
+static void __mcs_lock_lock(struct mcs_lock_node *lock,
 		struct mcs_lock_node *node)
 {
 	struct mcs_lock_node *pred;
@@ -158,7 +159,7 @@ static void mcs_lock_lock(struct mcs_lock_node *lock,
 	}
 }
 
-static void mcs_lock_unlock(struct mcs_lock_node *lock,
+static void __mcs_lock_unlock(struct mcs_lock_node *lock,
 		struct mcs_lock_node *node)
 {
 	if (node->next == NULL) {
@@ -178,6 +179,35 @@ static void mcs_lock_unlock(struct mcs_lock_node *lock,
 	node->next->locked = 0;
 }
 
+static void mcs_lock_lock_noirq(struct mcs_lock_node *lock,
+		struct mcs_lock_node *node)
+{
+	preempt_disable();
+	__mcs_lock_lock(lock, node);
+}
+
+static void mcs_lock_unlock_noirq(struct mcs_lock_node *lock,
+		struct mcs_lock_node *node)
+{
+	__mcs_lock_unlock(lock, node);
+	preempt_enable();
+}
+
+static void mcs_lock_lock(struct mcs_lock_node *lock,
+		struct mcs_lock_node *node)
+{
+	node->irqsave = cpu_disable_interrupt_save();
+	mcs_lock_lock_noirq(lock, node);
+}
+
+static void mcs_lock_unlock(struct mcs_lock_node *lock,
+		struct mcs_lock_node *node)
+{
+	mcs_lock_unlock_noirq(lock, node);
+	cpu_restore_interrupt(node->irqsave);
+}
+
+
 // reader/writer lock
 typedef struct mcs_rwlock_node {
 	ihk_atomic_t count;	// num of readers (use only common reader)

arch/x86/kernel/include/arch-memory.h

@@ -306,7 +306,7 @@ struct page_table;
 void set_pte(pte_t *ppte, unsigned long phys, enum ihk_mc_pt_attribute attr);
 pte_t *get_pte(struct page_table *pt, void *virt, enum ihk_mc_pt_attribute attr);
 
-void *early_alloc_page(void);
+void *early_alloc_pages(int nr_pages);
 void *get_last_early_heap(void);
 void flush_tlb(void);
 void flush_tlb_single(unsigned long addr);
@@ -318,5 +318,5 @@ extern unsigned long ap_trampoline;
 #define AP_TRAMPOLINE_SIZE  0x2000
 
 /* Local is cachable */
-#define IHK_IKC_QUEUE_PT_ATTR (PTATTR_NO_EXECUTE | PTATTR_WRITABLE | PTATTR_UNCACHABLE)
+#define IHK_IKC_QUEUE_PT_ATTR (PTATTR_NO_EXECUTE | PTATTR_WRITABLE)
 #endif

arch/x86/kernel/include/arch-string.h

@@ -0,0 +1,23 @@
+#ifndef _ASM_X86_STRING_H
+#define _ASM_X86_STRING_H
+
+#define ARCH_FAST_MEMCPY
+
+static inline void *__inline_memcpy(void *to, const void *from, size_t n)
+{
+	unsigned long d0, d1, d2;
+	asm volatile("rep ; movsl\n\t"
+		     "testb $2,%b4\n\t"
+		     "je 1f\n\t"
+		     "movsw\n"
+		     "1:\ttestb $1,%b4\n\t"
+		     "je 2f\n\t"
+		     "movsb\n"
+		     "2:"
+		     : "=&c" (d0), "=&D" (d1), "=&S" (d2)
+		     : "0" (n / 4), "q" (n), "1" ((long)to), "2" ((long)from)
+		     : "memory");
+	return to;
+}
+
+#endif

arch/x86/kernel/include/ihk/atomic.h

@@ -215,4 +215,25 @@ static inline unsigned long atomic_cmpxchg4(unsigned int *addr,
 	return oldval;
 }
 
+static inline void ihk_atomic_add_long(long i, long *v) {
+	asm volatile("lock addq %1,%0"
+					: "+m" (*v)
+					: "ir" (i));
+}
+static inline void ihk_atomic_add_ulong(long i, unsigned long *v) {
+	asm volatile("lock addq %1,%0"
+					: "+m" (*v)
+					: "ir" (i));
+}
+
+static inline unsigned long ihk_atomic_add_long_return(long i, long *v) {
+        long __i;
+
+        __i = i;
+        asm volatile("lock xaddq %0, %1"
+                     : "+r" (i), "+m" (*v)
+                     : : "memory");
+        return i + __i;
+}
+
 #endif

arch/x86/kernel/include/syscall_list.h

@@ -22,7 +22,7 @@
 
 SYSCALL_HANDLED(0, read)
 SYSCALL_DELEGATED(1, write)
-SYSCALL_DELEGATED(2, open)
+SYSCALL_HANDLED(2, open)
 SYSCALL_HANDLED(3, close)
 SYSCALL_DELEGATED(4, stat)
 SYSCALL_DELEGATED(5, fstat)
@@ -66,7 +66,7 @@ SYSCALL_DELEGATED(65, semop)
 SYSCALL_HANDLED(67, shmdt)
 SYSCALL_DELEGATED(69, msgsnd)
 SYSCALL_DELEGATED(70, msgrcv)
-SYSCALL_DELEGATED(72, fcntl)
+SYSCALL_HANDLED(72, fcntl)
 SYSCALL_DELEGATED(79, getcwd)
 SYSCALL_DELEGATED(89, readlink)
 SYSCALL_HANDLED(96, gettimeofday)
@@ -150,5 +150,8 @@ SYSCALL_HANDLED(602, pmc_start)
 SYSCALL_HANDLED(603, pmc_stop)
 SYSCALL_HANDLED(604, pmc_reset)
 SYSCALL_HANDLED(700, get_cpu_id)
+#ifdef TRACK_SYSCALLS
+SYSCALL_HANDLED(__NR_track_syscalls, track_syscalls)
+#endif // TRACK_SYSCALLS
 
 /**** End of File ****/

arch/x86/kernel/interrupt.S

@@ -130,11 +130,40 @@ general_protection_exception:
 		addq	$8, %rsp
 		iretq
 
+.global __freeze
+__freeze:
+		PUSH_ALL_REGS
+		callq	freeze
+		POP_ALL_REGS
+		iretq
+
 .globl nmi
 nmi:
 #define PANICED 232
 #define PANIC_REGS 240
 		movq	%rax,%gs:PANIC_REGS+0x00
+		movq	%rsp,%gs:PANIC_REGS+0x08
+
+		movl	nmi_mode(%rip),%eax
+		cmp	$1,%rax
+		je	1f
+		cmp	$2,%rax
+		jne	3f
+1:
+		cld
+		movq	%gs:PANIC_REGS+0x00,%rax
+		PUSH_ALL_REGS
+		subq	$40, %rsp
+		movq	%rsp,%gs:PANIC_REGS+0x10
+		movq	%rsp, %rdi
+		call	freeze_thaw
+		cmpq	$0, %rax
+		jnz	2f
+		addq	$40, %rsp
+2:
+		POP_ALL_REGS
+		iretq
+3:
 		movq	%rbx,%gs:PANIC_REGS+0x08
 		movq	%rcx,%gs:PANIC_REGS+0x10
 		movq	%rdx,%gs:PANIC_REGS+0x18

arch/x86/kernel/memory.c

@@ -23,6 +23,7 @@
 #include <process.h>
 #include <page.h>
 #include <cls.h>
+#include <kmalloc.h>
 
 #define	dkprintf(...)	do { if (0) kprintf(__VA_ARGS__); } while (0)
 #define	ekprintf(...)	kprintf(__VA_ARGS__)
@@ -30,11 +31,10 @@
 static char *last_page;
 extern char _head[], _end[];
 
-static struct ihk_mc_pa_ops *pa_ops;
-
 extern unsigned long x86_kernel_phys_base;
 
-void *early_alloc_page(void)
+/* Arch specific early allocation routine */
+void *early_alloc_pages(int nr_pages)
 {
 	void *p;
 
@@ -45,59 +45,38 @@ void *early_alloc_page(void)
 		last_page = phys_to_virt(virt_to_phys(last_page));
 	} else if (last_page == (void *)-1) {
 		panic("Early allocator is already finalized. Do not use it.\n");
-	}
+	} else {
+        if(virt_to_phys(last_page) >= bootstrap_mem_end) {
+            panic("Early allocator: Out of memory\n");
+        }
+    }
 	p = last_page;
-	last_page += PAGE_SIZE;
+	last_page += (nr_pages * PAGE_SIZE);
 
 	return p;
 }
 
-void *arch_alloc_page(enum ihk_mc_ap_flag flag)
+void early_alloc_invalidate(void)
 {
-	if (pa_ops)
-		return pa_ops->alloc_page(1, PAGE_P2ALIGN, flag);
-	else
-		return early_alloc_page();
-}
-void arch_free_page(void *ptr)
-{
-	if (pa_ops)
-		pa_ops->free_page(ptr, 1);
+	last_page = (void *)-1;
 }
 
-void *ihk_mc_alloc_aligned_pages(int npages, int p2align, enum ihk_mc_ap_flag flag)
+void *ihk_mc_allocate(int size, int flag)
 {
-	if (pa_ops)
-		return pa_ops->alloc_page(npages, p2align, flag);
-	else
+	if (!cpu_local_var(kmalloc_initialized)) {
+		kprintf("%s: error, kmalloc not yet initialized\n", __FUNCTION__);
 		return NULL;
-}
-
-void *ihk_mc_alloc_pages(int npages, enum ihk_mc_ap_flag flag)
-{
-	return ihk_mc_alloc_aligned_pages(npages, PAGE_P2ALIGN, flag);
-}
-
-void ihk_mc_free_pages(void *p, int npages)
-{
-	if (pa_ops)
-		pa_ops->free_page(p, npages);
-}
-
-void *ihk_mc_allocate(int size, enum ihk_mc_ap_flag flag)
-{
-	if (pa_ops && pa_ops->alloc)
-		return pa_ops->alloc(size, flag);
-	else
-		return ihk_mc_alloc_pages(1, flag);
+	}
+	return kmalloc(size, IHK_MC_AP_NOWAIT);
 }
 
 void ihk_mc_free(void *p)
 {
-	if (pa_ops && pa_ops->free)
-		return pa_ops->free(p);
-	else
-		return ihk_mc_free_pages(p, 1);
+	if (!cpu_local_var(kmalloc_initialized)) {
+		kprintf("%s: error, kmalloc not yet initialized\n", __FUNCTION__);
+		return;
+	}
+	kfree(p);
 }
 
 void *get_last_early_heap(void)
@@ -172,7 +151,7 @@ static unsigned long setup_l3(struct page_table *pt,
 			pt->entry[i] = 0;
 			continue;
 		}
-		pt_phys = setup_l2(arch_alloc_page(IHK_MC_AP_CRITICAL), phys, start, end);
+		pt_phys = setup_l2(ihk_mc_alloc_pages(1, IHK_MC_AP_CRITICAL), phys, start, end);
 
 		pt->entry[i] = pt_phys | PFL3_PDIR_ATTR;
 	}
@@ -196,7 +175,7 @@ static void init_normal_area(struct page_table *pt)
 
 	for (phys = (map_start & ~(PTL4_SIZE - 1)); phys < map_end;
 	     phys += PTL4_SIZE) {
-		pt_phys = setup_l3(arch_alloc_page(IHK_MC_AP_CRITICAL), phys,
+		pt_phys = setup_l3(ihk_mc_alloc_pages(1, IHK_MC_AP_CRITICAL), phys,
 		                   map_start, map_end);
 
 		pt->entry[ident_index++] = pt_phys | PFL4_PDIR_ATTR;
@@ -204,9 +183,9 @@ static void init_normal_area(struct page_table *pt)
 	}
 }
 
-static struct page_table *__alloc_new_pt(enum ihk_mc_ap_flag ap_flag)
+static struct page_table *__alloc_new_pt(ihk_mc_ap_flag ap_flag)
 {
-	struct page_table *newpt = arch_alloc_page(ap_flag);
+	struct page_table *newpt = ihk_mc_alloc_pages(1, ap_flag);
 
 	if(newpt)
 		memset(newpt, 0, sizeof(struct page_table));
@@ -303,7 +282,7 @@ void set_pte(pte_t *ppte, unsigned long phys, enum ihk_mc_pt_attribute attr)
  *             and returns a pointer to the PTE corresponding to the
  *             virtual address.
  */
-pte_t *get_pte(struct page_table *pt, void *virt, enum ihk_mc_pt_attribute attr, enum ihk_mc_ap_flag ap_flag)
+pte_t *get_pte(struct page_table *pt, void *virt, enum ihk_mc_pt_attribute attr, ihk_mc_ap_flag ap_flag)
 {
 	int l4idx, l3idx, l2idx, l1idx;
 	unsigned long v = (unsigned long)virt;
@@ -364,7 +343,7 @@ static int __set_pt_page(struct page_table *pt, void *virt, unsigned long phys,
 	int l4idx, l3idx, l2idx, l1idx;
 	unsigned long v = (unsigned long)virt;
 	struct page_table *newpt;
-	enum ihk_mc_ap_flag ap_flag;
+	ihk_mc_ap_flag ap_flag;
 	int in_kernel =
 		(((unsigned long long)virt) >= 0xffff000000000000ULL);
 	unsigned long init_pt_lock_flags;
@@ -515,8 +494,10 @@ uint64_t ihk_mc_pt_virt_to_pagemap(struct page_table *pt, unsigned long virt)
 	return pagemap;
 }
 
-int ihk_mc_pt_virt_to_phys(struct page_table *pt,
-                           const void *virt, unsigned long *phys)
+int ihk_mc_pt_virt_to_phys_size(struct page_table *pt,
+                           const void *virt,
+						   unsigned long *phys,
+						   unsigned long *size)
 {
 	int l4idx, l3idx, l2idx, l1idx;
 	unsigned long v = (unsigned long)virt;
@@ -538,6 +519,7 @@ int ihk_mc_pt_virt_to_phys(struct page_table *pt,
 	if ((pt->entry[l3idx] & PFL3_SIZE)) {
 		*phys = pte_get_phys(&pt->entry[l3idx])
 			| (v & (PTL3_SIZE - 1));
+		if (size) *size = PTL3_SIZE;
 		return 0;
 	}
 	pt = phys_to_virt(pte_get_phys(&pt->entry[l3idx]));
@@ -548,6 +530,7 @@ int ihk_mc_pt_virt_to_phys(struct page_table *pt,
 	if ((pt->entry[l2idx] & PFL2_SIZE)) {
 		*phys = pte_get_phys(&pt->entry[l2idx])
 			| (v & (PTL2_SIZE - 1));
+		if (size) *size = PTL2_SIZE;
 		return 0;
 	}
 	pt = phys_to_virt(pte_get_phys(&pt->entry[l2idx]));
@@ -557,9 +540,17 @@ int ihk_mc_pt_virt_to_phys(struct page_table *pt,
 	}
 
 	*phys = pte_get_phys(&pt->entry[l1idx]) | (v & (PTL1_SIZE - 1));
+	if (size) *size = PTL1_SIZE;
 	return 0;
 }
 
+int ihk_mc_pt_virt_to_phys(struct page_table *pt,
+                           const void *virt, unsigned long *phys)
+{
+	return ihk_mc_pt_virt_to_phys_size(pt, virt, phys, NULL);
+}
+
+
 int ihk_mc_pt_print_pte(struct page_table *pt, void *virt)
 {
 	int l4idx, l3idx, l2idx, l1idx;
@@ -571,28 +562,34 @@ int ihk_mc_pt_print_pte(struct page_table *pt, void *virt)
 
 	GET_VIRT_INDICES(v, l4idx, l3idx, l2idx, l1idx);
 
+	__kprintf("l4 table: 0x%lX l4idx: %d \n", virt_to_phys(pt), l4idx);
 	if (!(pt->entry[l4idx] & PFL4_PRESENT)) {
 		__kprintf("0x%lX l4idx not present! \n", (unsigned long)virt);
-		__kprintf("l4 entry: 0x%lX\n", pt->entry[l4idx]);
 		return -EFAULT;
 	}
+	__kprintf("l4 entry: 0x%lX\n", pt->entry[l4idx]);
 	pt = phys_to_virt(pt->entry[l4idx] & PAGE_MASK);
 
 	__kprintf("l3 table: 0x%lX l3idx: %d \n", virt_to_phys(pt), l3idx);
 	if (!(pt->entry[l3idx] & PFL3_PRESENT)) {
 		__kprintf("0x%lX l3idx not present! \n", (unsigned long)virt);
-		__kprintf("l3 entry: 0x%lX\n", pt->entry[l3idx]);
 		return -EFAULT;
 	}
+	__kprintf("l3 entry: 0x%lX\n", pt->entry[l3idx]);
+	if ((pt->entry[l3idx] & PFL3_SIZE)) {
+		__kprintf("l3 entry is 1G page\n");
+		return 0;
+	}
 	pt = phys_to_virt(pt->entry[l3idx] & PAGE_MASK);
 	
 	__kprintf("l2 table: 0x%lX l2idx: %d \n", virt_to_phys(pt), l2idx);
 	if (!(pt->entry[l2idx] & PFL2_PRESENT)) {
 		__kprintf("0x%lX l2idx not present! \n", (unsigned long)virt);
-		__kprintf("l2 entry: 0x%lX\n", pt->entry[l2idx]);
 		return -EFAULT;
 	}
+	__kprintf("l2 entry: 0x%lX\n", pt->entry[l2idx]);
 	if ((pt->entry[l2idx] & PFL2_SIZE)) {
+		__kprintf("l2 entry is 2M page\n");
 		return 0;
 	}
 	pt = phys_to_virt(pt->entry[l2idx] & PAGE_MASK);
@@ -671,7 +668,7 @@ int ihk_mc_pt_prepare_map(page_table_t p, void *virt, unsigned long size,
 	return ret;
 }
 
-struct page_table *ihk_mc_pt_create(enum ihk_mc_ap_flag ap_flag)
+struct page_table *ihk_mc_pt_create(ihk_mc_ap_flag ap_flag)
 {
 	struct page_table *pt = ihk_mc_alloc_pages(1, ap_flag);
 
@@ -715,7 +712,7 @@ static void destroy_page_table(int level, struct page_table *pt)
 		}
 	}
 
-	arch_free_page(pt);
+	ihk_mc_free_pages(pt, 1);
 	return;
 }
 
@@ -910,11 +907,17 @@ static int split_large_page(pte_t *ptep, size_t pgsize)
 
 	*ptep = (virt_to_phys(pt) & PT_PHYSMASK) | PFL2_PDIR_ATTR;
 
-	if (phys_base != NOPHYS) {
-		page = phys_to_page(phys_base);
-		if (page && page_unmap(page)) {
-			kprintf("split_large_page:page_unmap:%p\n", page);
-			panic("split_large_page:page_unmap\n");
+	/* Do not do this check for large pages as they don't come from the zeroobj
+	 * and are not actually mapped.
+	 * TODO: clean up zeroobj as we don't really need it, anonymous mappings
+	 * should be allocated for real */
+	if (pgsize != PTL2_SIZE) {
+		if (phys_base != NOPHYS) {
+			page = phys_to_page(phys_base);
+			if (pgsize != PTL2_SIZE && page && page_unmap(page)) {
+				kprintf("split_large_page:page_unmap:%p\n", page);
+				panic("split_large_page:page_unmap\n");
+			}
 		}
 	}
 	return 0;
@@ -1098,13 +1101,15 @@ static int clear_range_l1(void *args0, pte_t *ptep, uint64_t base,
 		page = phys_to_page(phys);
 	}
 
-	if (page && page_is_in_memobj(page) && (old & PFL1_DIRTY)) {
+	if (page && page_is_in_memobj(page) && (old & PFL1_DIRTY) && (args->memobj) &&
+			!(args->memobj->flags & MF_ZEROFILL)) {
 		memobj_flush_page(args->memobj, phys, PTL1_SIZE);
 	}
 
 	if (!(old & PFL1_FILEOFF) && args->free_physical) {
-		if (page && page_unmap(page)) {
+		if (!page || (page && page_unmap(page))) {
 			ihk_mc_free_pages(phys_to_virt(phys), 1);
+			dkprintf("%s: freeing regular page at 0x%lx\n", __FUNCTION__, base);
 		}
 		args->vm->currss -= PTL1_SIZE;
 	}
@@ -1151,8 +1156,9 @@ static int clear_range_l2(void *args0, pte_t *ptep, uint64_t base,
 		}
 
 		if (!(old & PFL2_FILEOFF) && args->free_physical) {
-			if (page && page_unmap(page)) {
+			if (!page || (page && page_unmap(page))) {
 				ihk_mc_free_pages(phys_to_virt(phys), PTL2_SIZE/PTL1_SIZE);
+				dkprintf("%s: freeing large page at 0x%lx\n", __FUNCTION__, base);
 			}
 			args->vm->currss -= PTL2_SIZE;
 		}
@@ -1170,7 +1176,7 @@ static int clear_range_l2(void *args0, pte_t *ptep, uint64_t base,
 		*ptep = PTE_NULL;
 		remote_flush_tlb_cpumask(args->vm, base,
 				ihk_mc_get_processor_id());
-		arch_free_page(pt);
+		ihk_mc_free_pages(pt, 1);
 	}
 
 	return 0;
@@ -1215,7 +1221,7 @@ static int clear_range_l3(void *args0, pte_t *ptep, uint64_t base,
 		}
 
 		if (!(old & PFL3_FILEOFF) && args->free_physical) {
-			if (page && page_unmap(page)) {
+			if (!page || (page && page_unmap(page))) {
 				ihk_mc_free_pages(phys_to_virt(phys), PTL3_SIZE/PTL1_SIZE);
 			}
 			args->vm->currss -= PTL3_SIZE;
@@ -1234,7 +1240,7 @@ static int clear_range_l3(void *args0, pte_t *ptep, uint64_t base,
 		*ptep = PTE_NULL;
 		remote_flush_tlb_cpumask(args->vm, base,
 				ihk_mc_get_processor_id());
-		arch_free_page(pt);
+		ihk_mc_free_pages(pt, 1);
 	}
 
 	return 0;
@@ -1270,6 +1276,9 @@ static int clear_range(struct page_table *pt, struct process_vm *vm,
 	}
 
 	args.free_physical = free_physical;
+	if (memobj && (memobj->flags & MF_DEV_FILE)) {
+		args.free_physical = 0;
+	}
 	args.memobj = memobj;
 	args.vm = vm;
 
@@ -1585,7 +1594,7 @@ retry:
 	error = 0;
 out:
 	if (newpt) {
-		arch_free_page(newpt);
+		ihk_mc_free_pages(newpt, 1);
 	}
 	dkprintf("set_range_l2(%lx,%lx,%lx): %d %lx\n",
 			base, start, end, error, *ptep);
@@ -1668,7 +1677,7 @@ retry:
 	error = 0;
 out:
 	if (newpt) {
-		arch_free_page(newpt);
+		ihk_mc_free_pages(newpt, 1);
 	}
 	dkprintf("set_range_l3(%lx,%lx,%lx): %d\n",
 			base, start, end, error, *ptep);
@@ -1726,7 +1735,7 @@ retry:
 	error = 0;
 out:
 	if (newpt) {
-		arch_free_page(newpt);
+		ihk_mc_free_pages(newpt, 1);
 	}
 	dkprintf("set_range_l4(%lx,%lx,%lx): %d %lx\n",
 			base, start, end, error, *ptep);
@@ -1778,9 +1787,19 @@ int ihk_mc_pt_set_pte(page_table_t pt, pte_t *ptep, size_t pgsize,
 		*ptep = phys | attr_to_l1attr(attr);
 	}
 	else if (pgsize == PTL2_SIZE) {
+		if (phys & (PTL2_SIZE - 1)) {
+			kprintf("%s: error: phys needs to be PTL2_SIZE aligned\n", __FUNCTION__);
+			error = -1;
+			goto out;
+		}
 		*ptep = phys | attr_to_l2attr(attr | PTATTR_LARGEPAGE);
 	}
 	else if ((pgsize == PTL3_SIZE) && (use_1gb_page)) {
+		if (phys & (PTL3_SIZE - 1)) {
+			kprintf("%s: error: phys needs to be PTL3_SIZE aligned\n", __FUNCTION__);
+			error = -1;
+			goto out;
+		}
 		*ptep = phys | attr_to_l3attr(attr | PTATTR_LARGEPAGE);
 	}
 	else {
@@ -2083,7 +2102,7 @@ static void init_vsyscall_area(struct page_table *pt)
 void init_page_table(void)
 {
 	check_available_page_size();
-	init_pt = arch_alloc_page(IHK_MC_AP_CRITICAL);
+	init_pt = ihk_mc_alloc_pages(1, IHK_MC_AP_CRITICAL);
 	ihk_mc_spinlock_init(&init_pt_lock);
 	
 	memset(init_pt, 0, sizeof(PAGE_SIZE));
@@ -2100,27 +2119,27 @@ void init_page_table(void)
 }
 
 extern void __reserve_arch_pages(unsigned long, unsigned long,
-                                 void (*)(unsigned long, unsigned long, int));
+		void (*)(struct ihk_page_allocator_desc *, 
+			unsigned long, unsigned long, int));
 
-void ihk_mc_reserve_arch_pages(unsigned long start, unsigned long end,
-                               void (*cb)(unsigned long, unsigned long, int))
+void ihk_mc_reserve_arch_pages(struct ihk_page_allocator_desc *pa_allocator,
+		unsigned long start, unsigned long end,
+		void (*cb)(struct ihk_page_allocator_desc *, 
+			unsigned long, unsigned long, int))
 {
 	/* Reserve Text + temporal heap */
-	cb(virt_to_phys(_head), virt_to_phys(get_last_early_heap()), 0);
+	cb(pa_allocator, virt_to_phys(_head), virt_to_phys(get_last_early_heap()), 0);
 	/* Reserve trampoline area to boot the second ap */
-	cb(ap_trampoline, ap_trampoline + AP_TRAMPOLINE_SIZE, 0);
+	cb(pa_allocator, ap_trampoline, ap_trampoline + AP_TRAMPOLINE_SIZE, 0);
 	/* Reserve the null page */
-	cb(0, PAGE_SIZE, 0);
-	/* Micro-arch specific */
+	cb(pa_allocator, 0, PAGE_SIZE, 0);
+	/* 
+	 * Micro-arch specific 
+	 * TODO: this does nothing in SMP mode, update it for KNC if necessary 
+	 */
 	__reserve_arch_pages(start, end, cb);
 }
 
-void ihk_mc_set_page_allocator(struct ihk_mc_pa_ops *ops)
-{
-	last_page = (void *)-1;
-	pa_ops = ops;
-}
-
 unsigned long virt_to_phys(void *v)
 {
 	unsigned long va = (unsigned long)v;
@@ -2218,30 +2237,28 @@ int strcpy_from_user(char *dst, const char *src)
 	return err;
 }
 
-long getlong_user(const long *p)
+long getlong_user(long *dest, const long *p)
 {
 	int error;
-	long l;
 
-	error = copy_from_user(&l, p, sizeof(l));
+	error = copy_from_user(dest, p, sizeof(long));
 	if (error) {
 		return error;
 	}
 
-	return l;
+	return 0;
 }
 
-int getint_user(const int *p)
+int getint_user(int *dest, const int *p)
 {
 	int error;
-	int i;
 
-	error = copy_from_user(&i, p, sizeof(i));
+	error = copy_from_user(dest, p, sizeof(int));
 	if (error) {
 		return error;
 	}
 
-	return i;
+	return 0;
 }
 
 int read_process_vm(struct process_vm *vm, void *kdst, const void *usrc, size_t siz)
@@ -2261,13 +2278,18 @@ int read_process_vm(struct process_vm *vm, void *kdst, const void *usrc, size_t
 	if ((ustart < vm->region.user_start)
 			|| (vm->region.user_end <= ustart)
 			|| ((vm->region.user_end - ustart) < siz)) {
+		kprintf("%s: error: out of user range\n", __FUNCTION__);
 		return -EFAULT;
 	}
 
 	reason = PF_USER;	/* page not present */
 	for (addr = ustart & PAGE_MASK; addr < uend; addr += PAGE_SIZE) {
+		if (!addr)
+			return -EINVAL;
+
 		error = page_fault_process_vm(vm, (void *)addr, reason);
 		if (error) {
+			kprintf("%s: error: PF for %p failed\n", __FUNCTION__, addr);
 			return error;
 		}
 	}
@@ -2283,11 +2305,22 @@ int read_process_vm(struct process_vm *vm, void *kdst, const void *usrc, size_t
 
 		error = ihk_mc_pt_virt_to_phys(vm->address_space->page_table, from, &pa);
 		if (error) {
+			kprintf("%s: error: resolving physical address or %p\n", __FUNCTION__, from);
 			return error;
 		}
 
-		va = phys_to_virt(pa);
-		memcpy(to, va, cpsize);
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+			dkprintf("%s: pa is outside of LWK memory, to: %p, pa: %p,"
+				"cpsize: %d\n", __FUNCTION__, to, pa, cpsize);
+			va = ihk_mc_map_virtual(pa, 1, PTATTR_ACTIVE);
+			memcpy(to, va, cpsize);
+			ihk_mc_unmap_virtual(va, 1, 1);
+		}
+		else {
+			va = phys_to_virt(pa);
+			memcpy(to, va, cpsize);
+		}
 
 		from += cpsize;
 		to += cpsize;
@@ -2356,8 +2389,18 @@ int write_process_vm(struct process_vm *vm, void *udst, const void *ksrc, size_t
 			return error;
 		}
 
-		va = phys_to_virt(pa);
-		memcpy(va, from, cpsize);
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+			dkprintf("%s: pa is outside of LWK memory, from: %p,"
+				"pa: %p, cpsize: %d\n", __FUNCTION__, from, pa, cpsize);
+			va = ihk_mc_map_virtual(pa, 1, PTATTR_ACTIVE);
+			memcpy(va, from, cpsize);
+			ihk_mc_unmap_virtual(va, 1, 1);
+		}
+		else {
+			va = phys_to_virt(pa);
+			memcpy(va, from, cpsize);
+		}
 
 		from += cpsize;
 		to += cpsize;
@@ -2381,7 +2424,7 @@ int patch_process_vm(struct process_vm *vm, void *udst, const void *ksrc, size_t
 	unsigned long pa;
 	void *va;
 
-	kprintf("patch_process_vm(%p,%p,%p,%lx)\n", vm, udst, ksrc, siz);
+	dkprintf("patch_process_vm(%p,%p,%p,%lx)\n", vm, udst, ksrc, siz);
 	if ((ustart < vm->region.user_start)
 			|| (vm->region.user_end <= ustart)
 			|| ((vm->region.user_end - ustart) < siz)) {
@@ -2413,14 +2456,24 @@ int patch_process_vm(struct process_vm *vm, void *udst, const void *ksrc, size_t
 			return error;
 		}
 
-		va = phys_to_virt(pa);
-		memcpy(va, from, cpsize);
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+			dkprintf("%s: pa is outside of LWK memory, from: %p,"
+				"pa: %p, cpsize: %d\n", __FUNCTION__, from, pa, cpsize);
+			va = ihk_mc_map_virtual(pa, 1, PTATTR_ACTIVE);
+			memcpy(va, from, cpsize);
+			ihk_mc_unmap_virtual(va, 1, 1);
+		}
+		else {
+			va = phys_to_virt(pa);
+			memcpy(va, from, cpsize);
+		}
 
 		from += cpsize;
 		to += cpsize;
 		remain -= cpsize;
 	}
 
-	kprintf("patch_process_vm(%p,%p,%p,%lx):%d\n", vm, udst, ksrc, siz, 0);
+	dkprintf("patch_process_vm(%p,%p,%p,%lx):%d\n", vm, udst, ksrc, siz, 0);
 	return 0;
 } /* patch_process_vm() */

arch/x86/kernel/mikc.c

@@ -16,6 +16,7 @@
 #include <memory.h>
 #include <string.h>
 
+extern int num_processors;
 extern void arch_set_mikc_queue(void *r, void *w);
 ihk_ikc_ph_t arch_master_channel_packet_handler;
 
@@ -23,22 +24,28 @@ int ihk_mc_ikc_init_first_local(struct ihk_ikc_channel_desc *channel,
                                 ihk_ikc_ph_t packet_handler)
 {
 	struct ihk_ikc_queue_head *rq, *wq;
+	size_t mikc_queue_pages;
 
 	ihk_ikc_system_init(NULL);
 
 	memset(channel, 0, sizeof(struct ihk_ikc_channel_desc));
 
-	/* Place both sides in this side */
-	rq = arch_alloc_page(IHK_MC_AP_CRITICAL);
-	wq = arch_alloc_page(IHK_MC_AP_CRITICAL);
+	mikc_queue_pages = ((2 * num_processors * MASTER_IKCQ_PKTSIZE)
+			+ (PAGE_SIZE - 1)) / PAGE_SIZE;
 
-	ihk_ikc_init_queue(rq, 0, 0, PAGE_SIZE, MASTER_IKCQ_PKTSIZE);
-	ihk_ikc_init_queue(wq, 0, 0, PAGE_SIZE, MASTER_IKCQ_PKTSIZE);
+	/* Place both sides in this side */
+	rq = ihk_mc_alloc_pages(mikc_queue_pages, IHK_MC_AP_CRITICAL);
+	wq = ihk_mc_alloc_pages(mikc_queue_pages, IHK_MC_AP_CRITICAL);
+
+	ihk_ikc_init_queue(rq, 0, 0,
+			mikc_queue_pages * PAGE_SIZE, MASTER_IKCQ_PKTSIZE);
+	ihk_ikc_init_queue(wq, 0, 0,
+			mikc_queue_pages * PAGE_SIZE, MASTER_IKCQ_PKTSIZE);
 
 	arch_master_channel_packet_handler = packet_handler;
 
 	ihk_ikc_init_desc(channel, IKC_OS_HOST, 0, rq, wq,
-	                  ihk_ikc_master_channel_packet_handler);
+	                  ihk_ikc_master_channel_packet_handler, channel);
 	ihk_ikc_enable_channel(channel);
 
 	/* Set boot parameter */

arch/x86/kernel/perfctr.c

@@ -12,16 +12,29 @@
 #include <errno.h>
 #include <ihk/debug.h>
 #include <registers.h>
+#include <mc_perf_event.h>
 
 extern unsigned int *x86_march_perfmap;
 extern int running_on_kvm(void);
 
 #define X86_CR4_PCE     0x00000100
 
+int perf_counters_discovered = 0;
+int X86_IA32_NUM_PERF_COUNTERS = 0;
+unsigned long X86_IA32_PERF_COUNTERS_MASK = 0;
+int X86_IA32_NUM_FIXED_PERF_COUNTERS = 0;
+unsigned long X86_IA32_FIXED_PERF_COUNTERS_MASK = 0;
+
 void x86_init_perfctr(void)
 {
+	int i = 0;
 	unsigned long reg;
 	unsigned long value = 0;
+	uint64_t op;
+	uint64_t eax;
+	uint64_t ebx;
+	uint64_t ecx;
+	uint64_t edx;
 
 	/* Do not do it on KVM */
 	if (running_on_kvm()) return;
@@ -30,12 +43,41 @@ void x86_init_perfctr(void)
 	asm volatile("movq %%cr4, %0" : "=r"(reg));
 	reg |= X86_CR4_PCE;
 	asm volatile("movq %0, %%cr4" : : "r"(reg));
+	
+	/* Detect number of supported performance counters */
+	if (!perf_counters_discovered) {
+		/* See Table 35.2 - Architectural MSRs in Vol 3C */
+		op = 0x0a;
+		asm volatile("cpuid" : "=a"(eax),"=b"(ebx),"=c"(ecx),"=d"(edx):"a"(op));
 
+		X86_IA32_NUM_PERF_COUNTERS = ((eax & 0xFF00) >> 8);
+		X86_IA32_PERF_COUNTERS_MASK = (1 << X86_IA32_NUM_PERF_COUNTERS) - 1;
+		
+		X86_IA32_NUM_FIXED_PERF_COUNTERS = (edx & 0x0F);
+		X86_IA32_FIXED_PERF_COUNTERS_MASK = 
+			((1UL << X86_IA32_NUM_FIXED_PERF_COUNTERS) - 1) <<
+			X86_IA32_BASE_FIXED_PERF_COUNTERS;
+	
+		perf_counters_discovered = 1;
+		kprintf("X86_IA32_NUM_PERF_COUNTERS: %d, X86_IA32_NUM_FIXED_PERF_COUNTERS: %d\n",
+				X86_IA32_NUM_PERF_COUNTERS, X86_IA32_NUM_FIXED_PERF_COUNTERS);
+	}
+
+	/* Clear Fixed Counter Control */
+	value = rdmsr(MSR_PERF_FIXED_CTRL);
+	value &= 0xfffffffffffff000L;
+	wrmsr(MSR_PERF_FIXED_CTRL, value);
+
+	/* Clear Generic Counter Control */
+	for(i = 0; i < X86_IA32_NUM_PERF_COUNTERS; i++) {
+		wrmsr(MSR_IA32_PERFEVTSEL0 + i, 0);
+	}
+		
 	/* Enable PMC Control */
-        value = rdmsr(MSR_PERF_GLOBAL_CTRL);
-        value |= X86_IA32_PERF_COUNTERS_MASK;
-        value |= X86_IA32_FIXED_PERF_COUNTERS_MASK;
-        wrmsr(MSR_PERF_GLOBAL_CTRL, value);
+	value = rdmsr(MSR_PERF_GLOBAL_CTRL);
+	value |= X86_IA32_PERF_COUNTERS_MASK;
+	value |= X86_IA32_FIXED_PERF_COUNTERS_MASK;
+	wrmsr(MSR_PERF_GLOBAL_CTRL, value);
 }
 
 static int set_perfctr_x86_direct(int counter, int mode, unsigned int value)
@@ -63,12 +105,12 @@ static int set_perfctr_x86_direct(int counter, int mode, unsigned int value)
 	wrmsr(MSR_IA32_PERFEVTSEL0 + counter, value);
 
 	//kprintf("wrmsr: %d <= %x\n", MSR_PERF_GLOBAL_CTRL, 0);
-	kprintf("wrmsr: %d <= %x\n", MSR_IA32_PERFEVTSEL0 + counter, value);
+	//kprintf("wrmsr: %d <= %x\n", MSR_IA32_PERFEVTSEL0 + counter, value);
 
 	return 0;
 }
 
-static int set_pmc_x86_direct(int counter, unsigned long val)
+static int set_pmc_x86_direct(int counter, long val)
 {
 	unsigned long cnt_bit = 0;
 
@@ -76,6 +118,8 @@ static int set_pmc_x86_direct(int counter, unsigned long val)
 		return -EINVAL;
 	}
 
+	val &= 0x000000ffffffffff; // 40bit Mask
+
 	cnt_bit = 1UL << counter;
 	if ( cnt_bit & X86_IA32_PERF_COUNTERS_MASK ) {
 		// set generic pmc
@@ -102,7 +146,7 @@ static int set_perfctr_x86(int counter, int event, int mask, int inv, int count,
 static int set_fixed_counter(int counter, int mode)
 {
 	unsigned long value = 0;
-	unsigned int  ctr_mask = 0x7;
+	unsigned int  ctr_mask = 0xf;
 	int counter_idx = counter - X86_IA32_BASE_FIXED_PERF_COUNTERS ;
 	unsigned int  set_val = 0;
 
@@ -183,6 +227,24 @@ int ihk_mc_perfctr_stop(unsigned long counter_mask)
 	value &= ~counter_mask;
 	wrmsr(MSR_PERF_GLOBAL_CTRL, value);
 
+	if(counter_mask >> 32 & 0x1) {
+		value = rdmsr(MSR_PERF_FIXED_CTRL);
+		value &= ~(0xf);
+		wrmsr(MSR_PERF_FIXED_CTRL, value);
+	}
+		
+	if(counter_mask >> 32 & 0x2) {
+		value = rdmsr(MSR_PERF_FIXED_CTRL);
+		value &= ~(0xf << 4);
+		wrmsr(MSR_PERF_FIXED_CTRL, value);
+	}
+
+	if(counter_mask >> 32 & 0x4) {
+		value = rdmsr(MSR_PERF_FIXED_CTRL);
+		value &= ~(0xf << 8);
+		wrmsr(MSR_PERF_FIXED_CTRL, value);
+	}
+
 	return 0;
 }
 
@@ -190,7 +252,7 @@ int ihk_mc_perfctr_stop(unsigned long counter_mask)
 int ihk_mc_perfctr_fixed_init(int counter, int mode)
 {
 	unsigned long value = 0;
-	unsigned int  ctr_mask = 0x7;
+	unsigned int  ctr_mask = 0xf;
 	int counter_idx = counter - X86_IA32_BASE_FIXED_PERF_COUNTERS ;
 	unsigned int  set_val = 0;
 
@@ -210,6 +272,9 @@ int ihk_mc_perfctr_fixed_init(int counter, int mode)
 		set_val |= 1;
 	}
 
+	// enable PMI on overflow
+	set_val |= 1 << 3;
+
 	set_val <<= counter_idx * 4;
 	value |= set_val;
 
@@ -223,7 +288,7 @@ int ihk_mc_perfctr_reset(int counter)
 	return set_pmc_x86_direct(counter, 0);
 }
 
-int ihk_mc_perfctr_set(int counter, unsigned long val)
+int ihk_mc_perfctr_set(int counter, long val)
 {
 	return set_pmc_x86_direct(counter, val);
 }
@@ -297,23 +362,33 @@ unsigned long ihk_mc_perfctr_read_msr(int counter)
 	return retval;
 }
 
-int ihk_mc_perfctr_alloc_counter(unsigned long pmc_status)
+int ihk_mc_perfctr_alloc_counter(unsigned int *type, unsigned long *config, unsigned long pmc_status)
 {
+	int ret = -1;
 	int i = 0;
-        int ret = -1;
 
-        // find avail generic counter
-        for(i = 0; i < X86_IA32_NUM_PERF_COUNTERS; i++) {
+	if(*type == PERF_TYPE_HARDWARE) {
+		switch(*config){
+		case PERF_COUNT_HW_INSTRUCTIONS :
+			*type = PERF_TYPE_RAW;
+			*config = 0x5300c0;
+			break;
+		default :
+			// Unexpected config
+			return -1;
+		}
+	}
+	else if(*type != PERF_TYPE_RAW) {
+		return -1;
+	}
+
+	// find avail generic counter
+       	for(i = 0; i < X86_IA32_NUM_PERF_COUNTERS; i++) {
 		if(!(pmc_status & (1 << i))) {
 			ret = i;
-			pmc_status |= (1 << i);
 			break;
 		}
 	}
 
-	if(ret < 0){
-		return ret;
-	}
-
-        return ret;
+	return ret;
 }

arch/x86/kernel/syscall.c

@@ -38,6 +38,8 @@ void set_signal(int sig, void *regs0, siginfo_t *info);
 void check_signal(unsigned long rc, void *regs0, int num);
 extern unsigned long do_fork(int, unsigned long, unsigned long, unsigned long,
 	unsigned long, unsigned long, unsigned long);
+extern int get_xsave_size();
+extern uint64_t get_xsave_mask();
 
 //#define DEBUG_PRINT_SC
 
@@ -54,6 +56,7 @@ uintptr_t debug_constants[] = {
 	offsetof(struct cpu_local_var, current),
 	offsetof(struct cpu_local_var, runq),
 	offsetof(struct cpu_local_var, status),
+	offsetof(struct cpu_local_var, idle),
 	offsetof(struct thread, ctx),
 	offsetof(struct thread, sched_list),
 	offsetof(struct thread, proc),
@@ -67,71 +70,37 @@ static struct vdso vdso;
 static size_t container_size = 0;
 static ptrdiff_t vdso_offset;
 
-/*
-See dkprintf("BSP HW ID = %d, ", bsp_hw_id); (in ./mcos/kernel/ap.c)
+extern int num_processors;
 
-Core with BSP HW ID 224 is 1st logical core of last physical core.
-It                      boots first and is given SW-ID of 0
+int obtain_clone_cpuid(cpu_set_t *cpu_set) {
+	int min_queue_len = -1;
+	int cpu, min_cpu = -1;
 
-Core with BSP HW ID 0 is 1st logical core of 1st physical core. 
-It                      boots next and is given  SW-ID of 1.
-Core with BSP HW ID 1   boots next and is given  SW-ID of 2.
-Core with BSP HW ID 2   boots next and is given  SW-ID of 3.
-Core with BSP HW ID 3   boots next and is given  SW-ID of 4.
-...
-Core with BSP HW ID 220 is 1st logical core of 56-th physical core.
-It                      boots next and is given  SW-ID of 221.
-Core with BSP HW ID 221 boots next and is given  SW-ID of 222.
-Core with BSP HW ID 222 boots next and is given  SW-ID of 223.
-Core with BSP HW ID 223 boots next and is given  SW-ID of 224.
+	/* Find the first allowed core with the shortest run queue */
+	for (cpu = 0; cpu < num_processors; ++cpu) {
+		struct cpu_local_var *v;
+		unsigned long irqstate;
 
-Core with BSP HW ID 225 is 2nd logical core of last physical core.
-It                      boots next and is given  SW-ID of 225.
-Core with BSP HW ID 226 boots next and is given  SW-ID of 226.
-Core with BSP HW ID 227 boots next and is given  SW-ID of 227.
-*/
-ihk_spinlock_t cpuid_head_lock = 0;
-static int cpuid_head = 0;
+		if (!CPU_ISSET(cpu, cpu_set)) continue;
 
-/* archtecture-depended syscall handlers */
-int obtain_clone_cpuid() {
-    /* see above on BSP HW ID */
-	struct ihk_mc_cpu_info *cpu_info = ihk_mc_get_cpu_info();
-    int cpuid, nretry = 0;
-    ihk_mc_spinlock_lock_noirq(&cpuid_head_lock);
-	
-	/* Always start from 0 to fill in LWK cores linearily */
-	cpuid_head = 0;
- retry:
-    /* Try to obtain next physical core */
-    cpuid = cpuid_head;
+		v = get_cpu_local_var(cpu);
+		irqstate = ihk_mc_spinlock_lock(&v->runq_lock);
+		if (min_queue_len == -1 || v->runq_len < min_queue_len) {
+			min_queue_len = v->runq_len;
+			min_cpu = cpu;
+		}
+		ihk_mc_spinlock_unlock(&v->runq_lock, irqstate);
 
-    /* A hyper-threading core on the same physical core as
-       the parent process might be chosen. Use sched_setaffinity
-       if you want to skip that kind of busy physical core for
-       performance reason. */
-    cpuid_head += 1;
-    if(cpuid_head >= cpu_info->ncpus) {
-        cpuid_head = 0;
-    }
+		if (min_queue_len == 0)
+			break;
+	}
 
-    /* A hyper-threading core whose parent physical core has a
-       process on one of its hyper-threading core might
-       be chosen. Use sched_setaffinity if you want to skip that
-       kind of busy physical core for performance reason. */
-    if(get_cpu_local_var(cpuid)->status != CPU_STATUS_IDLE) {
-        nretry++;
-        if(nretry >= cpu_info->ncpus) {
-            cpuid = -1;
-            ihk_mc_spinlock_unlock_noirq(&cpuid_head_lock);
-            goto out;
-        }
-        goto retry; 
-    }
-	get_cpu_local_var(cpuid)->status = CPU_STATUS_RESERVED;
-    ihk_mc_spinlock_unlock_noirq(&cpuid_head_lock);
- out:
-    return cpuid;
+	if (min_cpu != -1) {
+		if (get_cpu_local_var(min_cpu)->status != CPU_STATUS_RESERVED)
+			get_cpu_local_var(min_cpu)->status = CPU_STATUS_RESERVED;
+	}
+
+    return min_cpu;
 }
 
 int
@@ -219,6 +188,7 @@ SYSCALL_DECLARE(rt_sigreturn)
 	struct x86_user_context *regs;
 	struct sigsp ksigsp;
 	struct sigsp *sigsp;
+	int xsavesize = get_xsave_size();
 
 	asm ("movq %%gs:(%1),%0"
 			: "=r"(regs)
@@ -265,12 +235,31 @@ SYSCALL_DECLARE(rt_sigreturn)
 		check_signal(0, regs, 0);
 		check_need_resched();
 	}
+
+	if(ksigsp.fpregs && xsavesize){
+		void *fpregs = kmalloc(xsavesize + 64, IHK_MC_AP_NOWAIT);
+
+		if(fpregs){
+			uint64_t xsave_mask = get_xsave_mask();
+			unsigned int low = (unsigned int)xsave_mask;
+			unsigned int high = (unsigned int)(xsave_mask >> 32);
+			struct xsave_struct *kfpregs;
+
+			kfpregs = (void *)((((unsigned long)fpregs) + 63) & ~63);
+
+			if(copy_from_user(kfpregs, ksigsp.fpregs, xsavesize))
+				return -EFAULT;
+			asm volatile("xrstor %0" : : "m"(*kfpregs), "a"(low), "d"(high) : "memory");
+			kfree(fpregs);
+		}
+	}
+
 	return sigsp->sigrc;
 }
 
 extern struct cpu_local_var *clv;
 extern unsigned long do_kill(struct thread *thread, int pid, int tid, int sig, struct siginfo *info, int ptracecont);
-extern void interrupt_syscall(int all, int pid);
+extern void interrupt_syscall(int pid, int tid);
 extern int num_processors;
 
 #define RFLAGS_MASK (RFLAGS_CF | RFLAGS_PF | RFLAGS_AF | RFLAGS_ZF | \
@@ -521,14 +510,14 @@ void ptrace_report_signal(struct thread *thread, int sig)
 	int parent_pid;
 	struct siginfo info;
 
-	dkprintf("ptrace_report_signal,pid=%d\n", thread->proc->pid);
+	dkprintf("ptrace_report_signal, tid=%d, pid=%d\n", thread->tid, thread->proc->pid);
 
 	mcs_rwlock_writer_lock(&proc->update_lock, &lock);	
 	if(!(proc->ptrace & PT_TRACED)){
 		mcs_rwlock_writer_unlock(&proc->update_lock, &lock);
 		return;
 	}
-	proc->exit_status = sig;
+	thread->exit_status = sig;
 	/* Transition thread state */
 	proc->status = PS_TRACED;
 	thread->status = PS_TRACED;
@@ -546,8 +535,8 @@ void ptrace_report_signal(struct thread *thread, int sig)
 	memset(&info, '\0', sizeof info);
 	info.si_signo = SIGCHLD;
 	info.si_code = CLD_TRAPPED;
-	info._sifields._sigchld.si_pid = thread->proc->pid;
-	info._sifields._sigchld.si_status = thread->proc->exit_status;
+	info._sifields._sigchld.si_pid = thread->tid;
+	info._sifields._sigchld.si_status = thread->exit_status;
 	do_kill(cpu_local_var(current), parent_pid, -1, SIGCHLD, &info, 0);
 	/* Wake parent (if sleeping in wait4()) */
 	waitq_wakeup(&proc->parent->waitpid_q);
@@ -672,10 +661,10 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 	int	orgsig;
 	int	ptraceflag = 0;
 	struct mcs_rwlock_node_irqsave lock;
-	unsigned long irqstate;
+	struct mcs_rwlock_node_irqsave mcs_rw_node;
 
 	for(w = pending->sigmask.__val[0], sig = 0; w; sig++, w >>= 1);
-	dkprintf("do_signal,pid=%d,sig=%d\n", proc->pid, sig);
+	dkprintf("do_signal(): tid=%d, pid=%d, sig=%d\n", thread->tid, proc->pid, sig);
 	orgsig = sig;
 
 	if((proc->ptrace & PT_TRACED) &&
@@ -695,18 +684,20 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		rc = regs->gpr.rax;
 	}
 
-	irqstate = ihk_mc_spinlock_lock(&thread->sigcommon->lock);
+	mcs_rwlock_writer_lock(&thread->sigcommon->lock, &mcs_rw_node);
 	k = thread->sigcommon->action + sig - 1;
 
 	if(k->sa.sa_handler == SIG_IGN){
 		kfree(pending);
-		ihk_mc_spinlock_unlock(&thread->sigcommon->lock, irqstate);
+		mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 		return;
 	}
 	else if(k->sa.sa_handler){
 		unsigned long *usp; /* user stack */
 		struct sigsp ksigsp;
 		struct sigsp *sigsp;
+		int xsavesize = get_xsave_size();
+		unsigned long fpregs;
 
 		if((k->sa.sa_flags & SA_ONSTACK) &&
 		   !(thread->sigstack.ss_flags & SS_DISABLE) &&
@@ -719,7 +710,8 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		else{
 			usp = (unsigned long *)regs->gpr.rsp;
 		}
-		sigsp = ((struct sigsp *)usp) - 1;
+		fpregs = (unsigned long)usp - xsavesize;
+		sigsp = ((struct sigsp *)fpregs) - 1;
 		sigsp = (struct sigsp *)((unsigned long)sigsp & 0xfffffffffffffff0UL);
 		memset(&ksigsp, '\0', sizeof ksigsp);
 
@@ -751,19 +743,43 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		ksigsp.restart = isrestart(num, rc, sig, k->sa.sa_flags & SA_RESTART);
 		if(num != 0 && rc == -EINTR && sig == SIGCHLD)
 			ksigsp.restart = 1;
+		if(xsavesize){
+			uint64_t xsave_mask = get_xsave_mask();
+			unsigned int low = (unsigned int)xsave_mask;
+			unsigned int high = (unsigned int)(xsave_mask >> 32);
+			void *_kfpregs = kmalloc(xsavesize + 64, IHK_MC_AP_NOWAIT);
+			struct xsave_struct *kfpregs;
+
+			if(!_kfpregs){
+				kfree(pending);
+				kfree(_kfpregs);
+				kprintf("do_signal,no space available\n");
+				terminate(0, sig);
+				return;
+			}
+			kfpregs = (void *)((((unsigned long)_kfpregs) + 63) & ~63);
+			memset(kfpregs, '\0', xsavesize);
+			asm volatile("xsave %0" : : "m"(*kfpregs), "a"(low), "d"(high) : "memory");
+			if(copy_to_user((void *)fpregs, kfpregs, xsavesize)){
+				kfree(pending);
+				kfree(_kfpregs);
+				kprintf("do_signal,write_process_vm failed\n");
+				terminate(0, sig);
+				return;
+			}
+			ksigsp.fpregs = (void *)fpregs;
+			kfree(_kfpregs);
+		}
 		memcpy(&ksigsp.info, &pending->info, sizeof(siginfo_t));
 
 		if(copy_to_user(sigsp, &ksigsp, sizeof ksigsp)){
 			kfree(pending);
-			ihk_mc_spinlock_unlock(&thread->sigcommon->lock, irqstate);
+			mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 			kprintf("do_signal,write_process_vm failed\n");
 			terminate(0, sig);
 			return;
 		}
 
-
-
-
 		usp = (unsigned long *)sigsp;
 		usp--;
 		*usp = (unsigned long)k->sa.sa_restorer;
@@ -777,7 +793,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		if(!(k->sa.sa_flags & SA_NODEFER))
 			thread->sigmask.__val[0] |= pending->sigmask.__val[0];
 		kfree(pending);
-		ihk_mc_spinlock_unlock(&thread->sigcommon->lock, irqstate);
+		mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 		if(regs->gpr.rflags & RFLAGS_TF){
 			struct siginfo info;
 
@@ -803,7 +819,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		}
 		else
 			kfree(pending);
-		ihk_mc_spinlock_unlock(&thread->sigcommon->lock, irqstate);
+		mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 		switch (sig) {
 		case SIGSTOP:
 		case SIGTSTP:
@@ -835,7 +851,8 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 				/* Wake up the parent who tried wait4 and sleeping */
 				waitq_wakeup(&proc->parent->waitpid_q);
 
-				dkprintf("do_signal,SIGSTOP,sleeping\n");
+				dkprintf("do_signal(): pid: %d, tid: %d SIGSTOP, sleeping\n", 
+					proc->pid, thread->tid);
 				/* Sleep */
 				schedule();
 				dkprintf("SIGSTOP(): woken up\n");
@@ -849,7 +866,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 
 			/* Update thread state in fork tree */
 			mcs_rwlock_writer_lock(&proc->update_lock, &lock);	
-			proc->exit_status = SIGTRAP;
+			thread->exit_status = SIGTRAP;
 			proc->status = PS_TRACED;
 			thread->status = PS_TRACED;
 			mcs_rwlock_writer_unlock(&proc->update_lock, &lock);	
@@ -903,11 +920,11 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 static struct sig_pending *
 getsigpending(struct thread *thread, int delflag){
 	struct list_head *head;
-	ihk_spinlock_t *lock;
+	mcs_rwlock_lock_t *lock;
+	struct mcs_rwlock_node_irqsave mcs_rw_node;
 	struct sig_pending *next;
 	struct sig_pending *pending;
 	__sigset_t w;
-	int	irqstate;
 	__sigset_t x;
 	int sig;
 	struct k_sigaction *k;
@@ -916,8 +933,12 @@ getsigpending(struct thread *thread, int delflag){
 
 	lock = &thread->sigcommon->lock;
 	head = &thread->sigcommon->sigpending;
-	for(;;){
-		irqstate = ihk_mc_spinlock_lock(lock);
+	for(;;) {
+		if (delflag)
+			mcs_rwlock_writer_lock(lock, &mcs_rw_node);
+		else 
+			mcs_rwlock_reader_lock(lock, &mcs_rw_node);
+
 		list_for_each_entry_safe(pending, next, head, list){
 			for(x = pending->sigmask.__val[0], sig = 0; x; sig++, x >>= 1);
 			k = thread->sigcommon->action + sig - 1;
@@ -926,17 +947,26 @@ getsigpending(struct thread *thread, int delflag){
 			   (k->sa.sa_handler != (void *)1 &&
 			    k->sa.sa_handler != NULL)){
 				if(!(pending->sigmask.__val[0] & w)){
-					if(delflag)
+					if(delflag) 
 						list_del(&pending->list);
-					ihk_mc_spinlock_unlock(lock, irqstate);
+
+					if (delflag)
+						mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
+					else 
+						mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
 					return pending;
 				}
 			}
 		}
-		ihk_mc_spinlock_unlock(lock, irqstate);
+
+		if (delflag)
+			mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
+		else 
+			mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
 
 		if(lock == &thread->sigpendinglock)
 			return NULL;
+
 		lock = &thread->sigpendinglock;
 		head = &thread->sigpending;
 	}
@@ -984,22 +1014,25 @@ check_signal(unsigned long rc, void *regs0, int num)
 			}
 		}
 		ihk_mc_spinlock_unlock(&(cpu_local_var(runq_lock)), irqstate);
-		return;
+		goto out;
 	}
 
 	if(regs != NULL && !interrupt_from_user(regs)) {
-		return;
+		goto out;
 	}
 
 	for(;;){
 		pending = getsigpending(thread, 1);
 		if(!pending) {
 			dkprintf("check_signal,queue is empty\n");
-			return;
+			goto out;
 		}
 
 		do_signal(rc, regs, thread, pending, num);
 	}
+
+out:
+	return;
 }
 
 unsigned long
@@ -1013,7 +1046,8 @@ do_kill(struct thread *thread, int pid, int tid, int sig, siginfo_t *info,
 	struct thread *tthread = NULL;
 	int i;
 	__sigset_t mask;
-	ihk_spinlock_t *savelock = NULL;
+	mcs_rwlock_lock_t *savelock = NULL;
+	struct mcs_rwlock_node mcs_rw_node;
 	struct list_head *head = NULL;
 	int rc;
 	unsigned long irqstate = 0;
@@ -1145,7 +1179,8 @@ done:
 			if(pid != -1 && tthread->proc->pid != pid){
 				continue;
 			}
-			if(tthread->tid == tid){
+			if (tthread->tid == tid &&
+			    tthread->status != PS_EXITED) {
 				found = 1;
 				break;
 			}
@@ -1197,7 +1232,7 @@ done:
 
 	doint = 0;
 
-	ihk_mc_spinlock_lock_noirq(savelock);
+	mcs_rwlock_writer_lock_noirq(savelock, &mcs_rw_node);
 
 	/* Put signal event even when handler is SIG_IGN or SIG_DFL
 	   because target ptraced thread must call ptrace_report_signal 
@@ -1236,11 +1271,11 @@ done:
 			}
 		}
 	}
-	ihk_mc_spinlock_unlock_noirq(savelock);
+	mcs_rwlock_writer_unlock_noirq(savelock, &mcs_rw_node);
 	cpu_restore_interrupt(irqstate);
 
 	if (doint && !(mask & tthread->sigmask.__val[0])) {
-		int cpuid = tthread->cpu_id;
+		int tid = tthread->tid;
 		int pid = tproc->pid;
 		int status = tthread->status;
 
@@ -1251,18 +1286,21 @@ done:
 		}
 
 		if(!tthread->proc->nohost)
-			interrupt_syscall(pid, cpuid);
+			interrupt_syscall(pid, tid);
 
 		if (status != PS_RUNNING) {
 			if(sig == SIGKILL){
 				/* Wake up the target only when stopped by ptrace-reporting */
-				sched_wakeup_thread(tthread, PS_TRACED | PS_STOPPED);
+				sched_wakeup_thread(tthread, PS_TRACED | PS_STOPPED | PS_INTERRUPTIBLE);
 			}
 			else if(sig == SIGCONT || ptracecont == 1){
 				/* Wake up the target only when stopped by SIGSTOP */
 				sched_wakeup_thread(tthread, PS_STOPPED);
 				tthread->proc->status = PS_RUNNING;
 			}
+			else {
+				sched_wakeup_thread(tthread, PS_INTERRUPTIBLE);
+			}
 		}
 	}
 	release_thread(tthread);
@@ -1387,9 +1425,8 @@ SYSCALL_DECLARE(mmap)
 		goto out;
 	}
 
-	if ((addr < region->user_start)
-			|| (region->user_end <= addr)
-			|| ((region->user_end - addr) < len)) {
+	if ((flags & MAP_FIXED) && ((addr < region->user_start)
+			|| (region->user_end <= addr))) {
 		ekprintf("sys_mmap(%lx,%lx,%x,%x,%x,%lx):ENOMEM\n",
 				addr0, len0, prot, flags0, fd, off0);
 		error = -ENOMEM;
@@ -1513,6 +1550,7 @@ static int vdso_get_vdso_info(void)
 	struct ihk_ikc_channel_desc *ch = cpu_local_var(syscall_channel);
 
 	dkprintf("vdso_get_vdso_info()\n");
+	memset(&vdso, '\0', sizeof vdso);
 	vdso.busy = 1;
 	vdso.vdso_npages = 0;
 
@@ -1707,7 +1745,8 @@ int arch_map_vdso(struct process_vm *vm)
 	vrflags = VR_REMOTE;
 	vrflags |= VR_PROT_READ | VR_PROT_EXEC;
 	vrflags |= VRFLAG_PROT_TO_MAXPROT(vrflags);
-	error = add_process_memory_range(vm, (intptr_t)s, (intptr_t)e, NOPHYS, vrflags, NULL, 0, PAGE_SHIFT);
+	error = add_process_memory_range(vm, (intptr_t)s, (intptr_t)e,
+			NOPHYS, vrflags, NULL, 0, PAGE_SHIFT, NULL);
 	if (error) {
 		ekprintf("ERROR: adding memory range for vdso. %d\n", error);
 		goto out;
@@ -1738,7 +1777,8 @@ int arch_map_vdso(struct process_vm *vm)
 		vrflags = VR_REMOTE;
 		vrflags |= VR_PROT_READ;
 		vrflags |= VRFLAG_PROT_TO_MAXPROT(vrflags);
-		error = add_process_memory_range(vm, (intptr_t)s, (intptr_t)e, NOPHYS, vrflags, NULL, 0, PAGE_SHIFT);
+		error = add_process_memory_range(vm, (intptr_t)s, (intptr_t)e,
+				NOPHYS, vrflags, NULL, 0, PAGE_SHIFT, NULL);
 		if (error) {
 			ekprintf("ERROR: adding memory range for vvar. %d\n", error);
 			goto out;

arch/x86/kernel/vsyscall.c

@@ -17,6 +17,7 @@
  * make sure that these are position-independent codes.
  */
 
+#include <cls.h>
 #include <syscall.h>
 #include <ihk/atomic.h>
 #include <arch/cpu.h>

arch/x86/tools/eclair-dump-backtrace.exp.in

@@ -0,0 +1,67 @@
+#!/usr/bin/expect
+
+set INST_DIR "@prefix@"
+
+spawn $INST_DIR/bin/eclair -d /tmp/mckernel.dump -k $INST_DIR/smp-x86/kernel/mckernel.img -i
+
+set state "init"
+set thread_id 0
+
+expect {
+	"in ?? ()" {
+		switch -- $state {
+			"thread_chosen" {
+				set state "thread_skip"
+			}
+			"thread_bt" {
+				set state "thread_skip"
+			}
+		}
+
+		exp_continue
+	}
+	"(eclair) " {
+		switch -- $state {
+			"init" {
+				set state "threads_list"
+				send "info threads\r"
+			}
+			"threads_list" {
+				incr thread_id
+				set state "thread_chosen"
+				send "thread $thread_id\r"
+			}
+			"thread_skip" {
+				incr thread_id
+				set state "thread_chosen"
+				send "thread $thread_id\r"
+			}
+			"thread_chosen" {
+				set state "thread_bt"
+				send "bt\r"
+			}
+		}
+
+		exp_continue
+	}
+	"Type <return> to continue, or q <return> to quit" { 
+		switch -- $state {
+			"threads_list" {
+				send "\r"
+			}
+			"thread_bt" {
+				send "\r"
+			}
+			"thread_skip" {
+				send "q\r"
+			}
+		}
+		exp_continue 
+	}
+	" not known." { 
+		expect "(eclair) " { send "quit\r" }
+		expect "Quit anyway? (y or n) " { send "y\r" }
+		exit 0
+	}
+}
+

arch/x86/tools/irqbalance_mck.in.in

@@ -0,0 +1,28 @@
+# irqbalance is a daemon process that distributes interrupts across
+# CPUS on SMP systems. The default is to rebalance once every 10
+# seconds. This is the environment file that is specified to systemd via the
+# EnvironmentFile key in the service unit file (or via whatever method the init
+# system you're using has.
+#
+# ONESHOT=yes
+# after starting, wait for a minute, then look at the interrupt
+# load and balance it once; after balancing exit and do not change
+# it again.
+#IRQBALANCE_ONESHOT=
+
+#
+# IRQBALANCE_BANNED_CPUS
+# 64 bit bitmask which allows you to indicate which cpu's should
+# be skipped when reblancing irqs. Cpu numbers which have their
+# corresponding bits set to one in this mask will not have any
+# irq's assigned to them on rebalance
+#
+IRQBALANCE_BANNED_CPUS=%mask%
+
+#
+# IRQBALANCE_ARGS
+# append any args here to the irqbalance daemon as documented in the man page
+#
+IRQBALANCE_ARGS=--banirq=%banirq%
+
+

arch/x86/tools/irqbalance_mck.service.in

@@ -0,0 +1,10 @@
+[Unit]
+Description=irqbalance daemon
+After=syslog.target
+
+[Service]
+EnvironmentFile=/tmp/irqbalance_mck
+ExecStart=/usr/sbin/irqbalance --foreground $IRQBALANCE_ARGS
+
+[Install]
+WantedBy=multi-user.target

arch/x86/tools/mcreboot-smp-x86.sh.in

@@ -3,37 +3,61 @@
 # IHK SMP-x86 example boot script.
 # author: Balazs Gerofi <bgerofi@riken.jp>
 #      Copyright (C) 2014  RIKEN AICS
-# 
-# This is an example script for loading IHK, configuring a partition and 
-# booting McKernel on it. 
-# The script reserves half of the CPU cores and 512MB of RAM from NUMA node 0 
-# when IHK is loaded for the first time, otherwise it destroys the current 
-# McKernel instance and reboots it using the same set of resources as it used
-# previously. 
+#
+# This is an example script for loading IHK, configuring a partition and
+# booting McKernel on it.  Unless specific CPUs and memory are requested,
+# the script reserves half of the CPU cores and 512MB of RAM from
+# NUMA node 0 when IHK is loaded for the first time.
+# Otherwise, it destroys the current McKernel instance and reboots it using
+# the same set of resources as it used previously.
 # Note that the script does not output anything unless an error occurs.
 
 prefix="@prefix@"
-BINDIR="@BINDIR@"
-SBINDIR="@SBINDIR@"
-KMODDIR="@KMODDIR@"
-KERNDIR="@KERNDIR@"
+BINDIR="${prefix}/bin"
+SBINDIR="${prefix}/sbin"
+ETCDIR=@ETCDIR@
+KMODDIR="${prefix}/kmod"
+KERNDIR="${prefix}/@TARGET@/kernel"
 ENABLE_MCOVERLAYFS="@ENABLE_MCOVERLAYFS@"
 
+mem="512M@0"
+cpus=""
+
+if [ "${BASH_VERSINFO[0]}" -lt 4 ]; then
+	echo "You need at least bash-4.0 to run this script." >&2
+	exit 1
+fi
+
 INTERVAL=1
 LOGMODE=0
-while getopts :i:k: OPT
+facility="LOG_LOCAL6"
+chown_option=`logname 2> /dev/null`
+
+if [ "`systemctl status irqbalance_mck.service 2> /dev/null |grep -E 'Active: active'`" != "" -o "`systemctl status irqbalance.service 2> /dev/null |grep -E 'Active: active'`" != "" ]; then
+	irqbalance_used="yes"
+else
+	irqbalance_used="no"
+fi
+
+turbo=""
+
+while getopts :ti:k:c:m:o:f: OPT
 do
 	case ${OPT} in
+	f)	facility=${OPTARG}
+		;;
+	o)	chown_option=${OPTARG}
+		;;
 	i)	INTERVAL=${OPTARG}
 		expr "${INTERVAL}" + 1 > /dev/null 2>&1
 		if [ $? -ge 2 ]
 		then
-			echo "invalid -i value"
+			echo "invalid -i value" >&2
 			exit 1
 		fi
 		if [ ${INTERVAL} -le 0 ]
 		then
-			echo "invalid -i value"
+			echo "invalid -i value" >&2
 			exit 1
 		fi
 		;;
@@ -41,22 +65,128 @@ do
 		expr "${LOGMODE}" + 1 > /dev/null 2>&1
 		if [ $? -ge 2 ]
 		then
-			echo "invalid -k value"
+			echo "invalid -k value" >&2
 			exit 1
 		fi
 		if [ ${LOGMODE} -lt 0 -o ${LOGMODE} -gt 2 ]
 		then
-			echo "invalid -k value"
+			echo "invalid -k value" >&2
 			exit 1
 		fi
 		;;
-	*)  echo "invalid option -${OPT}"
+	c) cpus=${OPTARG}
+		;;
+	m) mem=${OPTARG}
+		;;
+	t) turbo="turbo"
+		;;
+	*)  echo "invalid option -${OPT}" >&2
 		exit 1
 	esac
 done
 
-mem="512M@0"
-cpus=""
+#
+# Revert any state that has been initialized before the error occured.
+#
+error_exit() {
+	local status=$1
+
+	case $status in
+	mcos_sys_mounted)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			umount /tmp/mcos/mcos0_sys
+		fi
+		;&
+	mcos_proc_mounted)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			umount /tmp/mcos/mcos0_proc
+		fi
+		;&
+	mcoverlayfs_loaded)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			rmmod mcoverlay 2>/dev/null
+		fi
+		;&
+	linux_proc_bind_mounted)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			umount /tmp/mcos/linux_proc
+		fi
+		;&
+	tmp_mcos_mounted)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			umount /tmp/mcos
+		fi
+		;&
+	tmp_mcos_created)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			rm -rf /tmp/mcos
+		fi
+		;&
+	os_created)
+		# Destroy all LWK instances
+		if ls /dev/mcos* 1>/dev/null 2>&1; then
+			for i in /dev/mcos*; do
+				ind=`echo $i|cut -c10-`;
+				if ! ${SBINDIR}/ihkconfig 0 destroy $ind; then
+					echo "warning: failed to destroy LWK instance $ind" >&2
+				fi
+			done
+		fi
+		;&
+	mcctrl_loaded)
+		rmmod mcctrl 2>/dev/null || echo "warning: failed to remove mcctrl" >&2
+		;&
+	cpus_reserved)
+		cpus=`${SBINDIR}/ihkconfig 0 query cpu`
+		if [ "${cpus}" != "" ]; then
+			if ! ${SBINDIR}/ihkconfig 0 release cpu $cpus > /dev/null; then
+				echo "warning: failed to release CPUs" >&2
+			fi
+		fi
+		;&
+	mem_reserved)
+		mem=`${SBINDIR}/ihkconfig 0 query mem`
+		if [ "${mem}" != "" ]; then
+			if ! ${SBINDIR}/ihkconfig 0 release mem $mem > /dev/null; then
+				echo "warning: failed to release memory" >&2
+			fi
+		fi
+		;&
+	ihk_smp_loaded)
+		rmmod ihk_smp_x86 2>/dev/null || echo "warning: failed to remove ihk_smp_x86" >&2
+		;&
+	ihk_loaded)
+		rmmod ihk 2>/dev/null || echo "warning: failed to remove ihk" >&2
+		;&
+	irqbalance_stopped)
+		if [ "`systemctl status irqbalance_mck.service 2> /dev/null |grep -E 'Active: active'`" != "" ]; then
+			if ! systemctl stop irqbalance_mck.service 2>/dev/null; then
+				echo "warning: failed to stop irqbalance_mck" >&2
+			fi
+			if ! systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null; then
+				echo "warning: failed to disable irqbalance_mck" >&2
+			fi
+			if ! etcdir=@ETCDIR@ perl -e '$etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "$etcdir/proc/irq/*/smp_affinity"; foreach $file (@files) { $dest = substr($file, length($etcdir)); if(0) {print "cp $file $dest\n";} system("cp $file $dest 2>/dev/null"); }'; then
+				echo "warning: failed to restore /proc/irq/*/smp_affinity" >&2
+			fi
+			if ! systemctl start irqbalance.service; then
+				echo "warning: failed to start irqbalance" >&2;
+			fi
+		fi
+		;&
+	aslr_disabled)
+		if [ -f /tmp/mckernel_randomize_va_space ]; then
+			cat /tmp/mckernel_randomize_va_space > /proc/sys/kernel/randomize_va_space
+		fi
+		;&
+	initial)
+		# Nothing more to revert
+		;;
+	esac
+
+	exit 1
+}
+
 ihk_ikc_irq_core=0
 
 release=`uname -r`
@@ -65,13 +195,28 @@ minor=`echo ${release} | sed -e 's/^[0-9]*.\([0-9]*\).*/\1/'`
 patch=`echo ${release} | sed -e 's/^[0-9]*.[0-9]*.\([0-9]*\).*/\1/'`
 linux_version_code=`expr \( ${major} \* 65536 \) + \( ${minor} \* 256 \) + ${patch}`
 rhel_release=`echo ${release} | sed -e 's/^[0-9]*.[0-9]*.[0-9]*-\([0-9]*\).*/\1/'`
-if [ "${release}" == "${rhel_release}" ]; then rhel_release=""; fi
-if [ "${ENABLE_MCOVERLAYFS}" == "yes" ]; then
-	enable_mcoverlay=`if ( [ ${linux_version_code} -ge 262144 ] && [ ${linux_version_code} -lt 262400 ] ); then echo "yes"; else echo "no"; fi`
-else
-	enable_mcoverlay=no
+if [ "${release}" == "${rhel_release}" ]; then
+	rhel_release="";
 fi
 
+enable_mcoverlay="no"
+
+if [ "${ENABLE_MCOVERLAYFS}" == "yes" ]; then
+	if [ "${rhel_release}" == "" ]; then
+		if [ ${linux_version_code} -ge 262144 -a ${linux_version_code} -lt 262400 ]; then
+			enable_mcoverlay="yes"
+		fi
+		if [ ${linux_version_code} -ge 263680 -a ${linux_version_code} -lt 263936 ]; then
+			enable_mcoverlay="yes"
+		fi
+	else
+		if [ ${linux_version_code} -eq 199168 -a ${rhel_release} -ge 327 ]; then
+			enable_mcoverlay="yes"
+		fi
+	fi
+fi
+
+# Figure out CPUs if not requested by user
 if [ "$cpus" == "" ]; then
 	# Get the number of CPUs on NUMA node 0
 	nr_cpus=`lscpu --parse | awk -F"," '{if ($4 == 0) print $4}' | wc -l`
@@ -79,121 +224,300 @@ if [ "$cpus" == "" ]; then
 	# Use the second half of the cores
 	let nr_cpus="$nr_cpus / 2"
 	cpus=`lscpu --parse | awk -F"," '{if ($4 == 0) print $1}' | tail -n $nr_cpus | xargs echo -n | sed 's/ /,/g'`
-	if [ "$cpus" == "" ]; then echo "error: no available CPUs on NUMA node 0?"; exit; fi
+	if [ "$cpus" == "" ]; then
+		echo "error: no available CPUs on NUMA node 0?" >&2
+		exit 1
+	fi
 fi
 
-# Remove delegator if loaded
-if [ "`lsmod | grep mcctrl`" != "" ]; then
-        if ! rmmod mcctrl; then echo "error: removing mcctrl"; exit; fi
+# Disable address space layout randomization
+if [ -f /proc/sys/kernel/randomize_va_space ] && [ "`cat /proc/sys/kernel/randomize_va_space`" != "0" ]; then
+	cat /proc/sys/kernel/randomize_va_space > /tmp/mckernel_randomize_va_space
+	echo "0" > /proc/sys/kernel/randomize_va_space
 fi
 
 # Remove mcoverlay if loaded
 if [ "$enable_mcoverlay" == "yes" ]; then
-	if [ "`lsmod | grep mcoverlay`" != "" ]; then
+	if grep mcoverlay /proc/modules &>/dev/null; then
 		if [ "`cat /proc/mounts | grep /tmp/mcos/mcos0_sys`" != "" ]; then umount -l /tmp/mcos/mcos0_sys; fi
 		if [ "`cat /proc/mounts | grep /tmp/mcos/mcos0_proc`" != "" ]; then umount -l /tmp/mcos/mcos0_proc; fi
 		if [ "`cat /proc/mounts | grep /tmp/mcos/linux_proc`" != "" ]; then umount -l /tmp/mcos/linux_proc; fi
 		if [ "`cat /proc/mounts | grep /tmp/mcos`" != "" ]; then umount -l /tmp/mcos; fi
 		if [ -e /tmp/mcos ]; then rm -rf /tmp/mcos; fi
-		if ! rmmod mcoverlay; then echo "error: removing mcoverlay"; exit; fi
+		if ! rmmod mcoverlay 2>/dev/null; then
+			echo "error: removing mcoverlay" >&2
+			error_exit "aslr_disabled"
+		fi
 	fi
 fi
 
-# Load IHK if not loaded
-if [ "`lsmod | grep ihk`" == "" ]; then
-        if ! insmod ${KMODDIR}/ihk.ko; then echo "error: loading ihk"; exit; fi;
+# Stop irqbalance
+if [ "${irqbalance_used}" == "yes" ]; then
+    systemctl stop irqbalance_mck.service 2>/dev/null
+    if ! systemctl stop irqbalance.service 2>/dev/null ; then
+		echo "error: stopping irqbalance" >&2
+		error_exit "aslr_disabled"
+	fi;
 fi
 
+# Load IHK if not loaded
+if ! grep -E 'ihk\s' /proc/modules &>/dev/null; then
+	if ! insmod ${KMODDIR}/ihk.ko 2>/dev/null; then
+		echo "error: loading ihk" >&2
+		error_exit "irqbalance_stopped"
+	fi
+fi
+
+# Increase swappiness so that we have better chance to allocate memory for IHK
+echo 100 > /proc/sys/vm/swappiness
+
+# Drop Linux caches to free memory
+sync && echo 3 > /proc/sys/vm/drop_caches
+
+# Merge free memory areas into large, physically contigous ones
+echo 1 > /proc/sys/vm/compact_memory 2>/dev/null
+
+sync
+
 # Load IHK-SMP if not loaded and reserve CPUs and memory
-if [ "`lsmod | grep ihk_smp_x86`" == "" ]; then
-        ihk_irq=""
-        for i in `seq 64 255`; do 
-                if [ ! -d /proc/irq/$i ] && [ "`cat /proc/interrupts | grep ":" | awk '{print $1}' | grep -o '[0-9]*' | grep -e '^$i$'`" == "" ]; then
-                        ihk_irq=$i 
-                        break
-                fi
-        done
-        if [ "$ihk_irq" == "" ]; then echo "error: no IRQ available"; exit; fi
-        if ! insmod ${KMODDIR}/ihk-smp-x86.ko ihk_start_irq=$ihk_irq ihk_ikc_irq_core=$ihk_ikc_irq_core; then echo "error: loading ihk-smp-x86"; exit; fi;
-        if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then echo "error: reserving CPUs"; exit; fi
-        if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then echo "error: reserving memory"; exit; fi
-# If loaded, but no resources allocated, get CPUs and memory
-else
-        if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus"; exit; fi
-        cpus_allocated=`${SBINDIR}/ihkosctl 0 query cpu`
-		if [ "$cpus_allocated" == "" ]; then
-			if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then echo "error: reserving CPUs"; exit; fi
+if ! grep ihk_smp_x86 /proc/modules &>/dev/null; then
+	ihk_irq=""
+	for i in `seq 64 255`; do
+		if [ ! -d /proc/irq/$i ] && [ "`cat /proc/interrupts | grep ":" | awk '{print $1}' | grep -o '[0-9]*' | grep -e '^$i$'`" == "" ]; then
+			ihk_irq=$i
+			break
 		fi
+	done
+	if [ "$ihk_irq" == "" ]; then
+		echo "error: no IRQ available" >&2
+		error_exit "ihk_loaded"
+	fi
+	if ! insmod ${KMODDIR}/ihk-smp-x86.ko ihk_start_irq=$ihk_irq ihk_ikc_irq_core=$ihk_ikc_irq_core 2>/dev/null; then
+		echo "error: loading ihk-smp-x86" >&2
+		error_exit "ihk_loaded"
+	fi
 
-        if ! ${SBINDIR}/ihkosctl 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
-        mem_allocated=`${SBINDIR}/ihkosctl 0 query mem`
-		if [ "$mem_allocated" == "" ]; then
-			if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then echo "error: reserving memory"; exit; fi
-		fi
+	# Free MCDRAM (special case for OFP SNC-4 mode)
+	if [ "`hostname | grep "c[0-9][0-9][0-9][0-9].ofp"`" != "" ] && [ "`cat /sys/devices/system/node/online`" == "0-7" ]; then
+		for i in 4 5 6 7; do
+			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
+				echo 0 > $f 2>&1 > /dev/null;
+			done
+			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
+				echo 1 > $f 2>&1 > /dev/null;
+			done
+		done
+	fi
+
+	if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then
+		echo "error: reserving memory" >&2
+		error_exit "ihk_smp_loaded"
+	fi
+	if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then
+		echo "error: reserving CPUs" >&2;
+		error_exit "mem_reserved"
+	fi
 fi
 
-# Check for existing OS instance and destroy
-if [ -c /dev/mcos0 ]; then
-        # Query CPU cores and memory of OS instance so that the same values are used as previously
-        if ! ${SBINDIR}/ihkosctl 0 query cpu > /dev/null; then echo "error: querying cpus"; exit; fi
-        cpus=`${SBINDIR}/ihkosctl 0 query cpu`
-        if ! ${SBINDIR}/ihkosctl 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
-        mem=`${SBINDIR}/ihkosctl 0 query mem`
-
-        if ! ${SBINDIR}/ihkconfig 0 destroy 0; then echo "warning: destroy failed"; fi	
-else
-        # Otherwise query IHK-SMP for resources
-        if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus"; exit; fi
-        cpus=`${SBINDIR}/ihkconfig 0 query cpu`
-        if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
-        mem=`${SBINDIR}/ihkconfig 0 query mem`	
+# Load mcctrl if not loaded
+if ! grep mcctrl /proc/modules &>/dev/null; then
+	if ! insmod ${KMODDIR}/mcctrl.ko 2>/dev/null; then
+		echo "error: inserting mcctrl.ko" >&2
+		error_exit "cpus_reserved"
+	fi
 fi
 
-if ! ${SBINDIR}/ihkconfig 0 create; then echo "error: create"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 assign cpu ${cpus}; then echo "error: assign CPUs"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 assign mem ${mem}; then echo "error: assign memory"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 load ${KERNDIR}/mckernel.img; then echo "error: loading kernel image"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 kargs "hidos ksyslogd=${LOGMODE}"; then echo "error: setting kernel arguments"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 boot; then echo "error: booting"; exit; fi
-if ! insmod ${KMODDIR}/mcctrl.ko; then echo "error: inserting mcctrl.ko"; exit; fi
-if ! chown `logname` /dev/mcd* /dev/mcos*; then echo "error: chowning device files"; exit; fi
+# Destroy all LWK instances
+if ls /dev/mcos* 1>/dev/null 2>&1; then
+	for i in /dev/mcos*; do
+		ind=`echo $i|cut -c10-`;
+		if ! ${SBINDIR}/ihkconfig 0 destroy $ind; then
+			echo "error: destroying LWK instance $ind failed" >&2
+			error_exit "mcctrl_loaded"
+		fi
+	done
+fi
 
+# Create OS instance
+if ! ${SBINDIR}/ihkconfig 0 create; then
+	echo "error: creating OS instance" >&2
+	error_exit "mcctrl_loaded"
+fi
+
+# Assign CPUs
+if ! ${SBINDIR}/ihkosctl 0 assign cpu ${cpus}; then
+	echo "error: assign CPUs" >&2
+	error_exit "os_created"
+fi
+
+# Assign memory
+if ! ${SBINDIR}/ihkosctl 0 assign mem ${mem}; then
+	echo "error: assign memory" >&2
+	error_exit "os_created"
+fi
+
+# Load kernel image
+if ! ${SBINDIR}/ihkosctl 0 load ${KERNDIR}/mckernel.img; then
+	echo "error: loading kernel image: ${KERNDIR}/mckernel.img" >&2
+	error_exit "os_created"
+fi
+
+# Set kernel arguments
+if ! ${SBINDIR}/ihkosctl 0 kargs "hidos ksyslogd=${LOGMODE} $turbo"; then
+	echo "error: setting kernel arguments" >&2
+	error_exit "os_created"
+fi
+
+# Boot OS instance
+if ! ${SBINDIR}/ihkosctl 0 boot; then
+	echo "error: booting" >&2
+	error_exit "os_created"
+fi
+
+# Set device file ownership
+if ! chown ${chown_option} /dev/mcd* /dev/mcos*; then
+	echo "warning: failed to chown device files" >&2
+fi
+
+# Overlay /proc, /sys with McKernel specific contents
 if [ "$enable_mcoverlay" == "yes" ]; then
-	if [ ! -e /tmp/mcos ]; then mkdir -p /tmp/mcos; fi
-	if ! mount -t tmpfs tmpfs /tmp/mcos; then echo "error: mount /tmp/mcos"; exit; fi
-	if [ ! -e /tmp/mcos/linux_proc ]; then mkdir -p /tmp/mcos/linux_proc; fi
-	if ! mount --bind /proc /tmp/mcos/linux_proc; then echo "error: mount /tmp/mcos/linux_proc"; exit; fi
-	if ! insmod ${KMODDIR}/mcoverlay.ko; then echo "error: inserting mcoverlay.ko"; exit; fi
+	if [ ! -e /tmp/mcos ]; then 
+		mkdir -p /tmp/mcos; 
+	fi
+	if ! mount -t tmpfs tmpfs /tmp/mcos; then
+		echo "error: mount /tmp/mcos" >&2
+		error_exit "tmp_mcos_created"
+	fi
+	if [ ! -e /tmp/mcos/linux_proc ]; then 
+		mkdir -p /tmp/mcos/linux_proc; 
+	fi
+	if ! mount --bind /proc /tmp/mcos/linux_proc; then
+		echo "error: mount /tmp/mcos/linux_proc" >&2
+		error_exit "tmp_mcos_mounted"
+	fi
+	if ! insmod ${KMODDIR}/mcoverlay.ko 2>/dev/null; then
+		echo "error: inserting mcoverlay.ko" >&2
+		error_exit "linux_proc_bind_mounted"
+	fi
 	while [ ! -e /proc/mcos0 ]
 	do
-		sleep 1
+		sleep 0.1
 	done
-	if [ ! -e /tmp/mcos/mcos0_proc ]; then mkdir -p /tmp/mcos/mcos0_proc; fi
-	if [ ! -e /tmp/mcos/mcos0_proc_upper ]; then mkdir -p /tmp/mcos/mcos0_proc_upper; fi
-	if [ ! -e /tmp/mcos/mcos0_proc_work ]; then mkdir -p /tmp/mcos/mcos0_proc_work; fi
-	if ! mount -t mcoverlay mcoverlay -o lowerdir=/proc/mcos0:/proc,upperdir=/tmp/mcos/mcos0_proc_upper,workdir=/tmp/mcos/mcos0_proc_work,nocopyupw,nofscheck /tmp/mcos/mcos0_proc; then echo "error: mount /tmp/mcos/mcos0_proc"; exit; fi
+	if [ ! -e /tmp/mcos/mcos0_proc ]; then 
+		mkdir -p /tmp/mcos/mcos0_proc; 
+	fi
+	if [ ! -e /tmp/mcos/mcos0_proc_upper ]; then 
+		mkdir -p /tmp/mcos/mcos0_proc_upper; 
+	fi
+	if [ ! -e /tmp/mcos/mcos0_proc_work ]; then 
+		mkdir -p /tmp/mcos/mcos0_proc_work; 
+	fi
+	if ! mount -t mcoverlay mcoverlay -o lowerdir=/proc/mcos0:/proc,upperdir=/tmp/mcos/mcos0_proc_upper,workdir=/tmp/mcos/mcos0_proc_work,nocopyupw,nofscheck /tmp/mcos/mcos0_proc; then
+		echo "error: mounting /tmp/mcos/mcos0_proc" >&2
+		error_exit "mcoverlayfs_loaded"
+	fi
+	# TODO: How de we revert this in case of failure??
 	mount --make-rprivate /proc
-	while [ ! -e /sys/devices/virtual/mcos/mcos0/sys ]
+
+	while [ ! -e /sys/devices/virtual/mcos/mcos0/sys/setup_complete ]
 	do
-		sleep 1
+		sleep 0.1
 	done
-	if [ ! -e /tmp/mcos/mcos0_sys ]; then mkdir -p /tmp/mcos/mcos0_sys; fi
-	if [ ! -e /tmp/mcos/mcos0_sys_upper ]; then mkdir -p /tmp/mcos/mcos0_sys_upper; fi
-	if [ ! -e /tmp/mcos/mcos0_sys_work ]; then mkdir -p /tmp/mcos/mcos0_sys_work; fi
-	if ! mount -t mcoverlay mcoverlay -o lowerdir=/sys/devices/virtual/mcos/mcos0/sys:/sys,upperdir=/tmp/mcos/mcos0_sys_upper,workdir=/tmp/mcos/mcos0_sys_work,nocopyupw,nofscheck /tmp/mcos/mcos0_sys; then echo "error: mount /tmp/mcos/mcos0_sys"; exit; fi
+	if [ ! -e /tmp/mcos/mcos0_sys ]; then 
+		mkdir -p /tmp/mcos/mcos0_sys; 
+	fi
+	if [ ! -e /tmp/mcos/mcos0_sys_upper ]; then 
+		mkdir -p /tmp/mcos/mcos0_sys_upper; 
+	fi
+	if [ ! -e /tmp/mcos/mcos0_sys_work ]; then 
+		mkdir -p /tmp/mcos/mcos0_sys_work; 
+	fi
+	if ! mount -t mcoverlay mcoverlay -o lowerdir=/sys/devices/virtual/mcos/mcos0/sys:/sys,upperdir=/tmp/mcos/mcos0_sys_upper,workdir=/tmp/mcos/mcos0_sys_work,nocopyupw,nofscheck /tmp/mcos/mcos0_sys; then
+		echo "error: mount /tmp/mcos/mcos0_sys" >&2
+		error_exit "mcos_proc_mounted"
+	fi
+	# TODO: How de we revert this in case of failure??
 	mount --make-rprivate /sys
+
+	touch /tmp/mcos/mcos0_proc/mckernel
+
+	rm -rf /tmp/mcos/mcos0_sys/setup_complete
+
+	# Hide NUMA related files which are outside the LWK partition
 	for cpuid in `find /sys/devices/system/cpu/* -maxdepth 0 -name "cpu[0123456789]*" -printf "%f "`; do
 		if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/cpu/$cpuid" ]; then
 			rm -rf /tmp/mcos/mcos0_sys/devices/system/cpu/$cpuid
+			rm -rf /tmp/mcos/mcos0_sys/bus/cpu/devices/$cpuid
+			rm -rf /tmp/mcos/mcos0_sys/bus/cpu/drivers/processor/$cpuid
+		else
+			for nodeid in `find /sys/devices/system/cpu/$cpuid/* -maxdepth 0 -name "node[0123456789]*" -printf "%f "`; do
+				if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/cpu/$cpuid/$nodeid" ]; then
+					rm -f /tmp/mcos/mcos0_sys/devices/system/cpu/$cpuid/$nodeid
+				fi
+			done
 		fi
 	done
+	for nodeid in `find /sys/devices/system/node/* -maxdepth 0 -name "node[0123456789]*" -printf "%f "`; do
+		if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/node/$nodeid" ]; then
+			rm -rf /tmp/mcos/mcos0_sys/devices/system/node/$nodeid/*
+			rm -rf /tmp/mcos/mcos0_sys/bus/node/devices/$nodeid
+		else
+			# Delete non-existent symlinks
+			for cpuid in `find /sys/devices/system/node/$nodeid/* -maxdepth 0 -name "cpu[0123456789]*" -printf "%f "`; do
+				if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/node/$nodeid/$cpuid" ]; then
+					rm -f /tmp/mcos/mcos0_sys/devices/system/node/$nodeid/$cpuid
+				fi
+			done
+
+			rm -f /tmp/mcos/mcos0_sys/devices/system/node/$nodeid/memory*
+		fi
+	done
+	rm -f /tmp/mcos/mcos0_sys/devices/system/node/has_*
 	for cpuid in `find /sys/bus/cpu/devices/* -maxdepth 0 -name "cpu[0123456789]*" -printf "%f "`; do
 		if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/bus/cpu/devices/$cpuid" ]; then
 			rm -rf /tmp/mcos/mcos0_sys/bus/cpu/devices/$cpuid
 		fi
 	done
 fi
-if [ ${LOGMODE} -ne 0 ]
-then
-	SBINDIR=${SBINDIR} ${SBINDIR}/mcklogd -i ${INTERVAL}
+
+# Start irqbalance with CPUs and IRQ for McKernel banned
+if [ "${irqbalance_used}" == "yes" ]; then
+    if ! etcdir=@ETCDIR@ perl -e 'use File::Copy qw(copy); $etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "/proc/irq/*/smp_affinity"; foreach $file (@files) { $rel = substr($file, 1); $dir=substr($rel, 0, length($rel)-length("/smp_affinity")); if(0) { print "cp $file $etcdir/$rel\n";} if(system("mkdir -p $etcdir/$dir")){ exit 1;} if(!copy($file,"$etcdir/$rel")){ exit 1;} }'; then
+		echo "error: saving /proc/irq/*/smp_affinity" >&2
+		error_exit "mcos_sys_mounted"
+	fi;
+
+    ncpus=`lscpu | grep -E '^CPU\(s\):' | awk '{print $2}'`
+    smp_affinity_mask=`echo $cpus | ncpus=$ncpus perl -e 'while(<>){@tokens = split /,/;foreach $token (@tokens) {@nums = split /-/,$token; for($num = $nums[0]; $num <= $nums[$#nums]; $num++) {$ndx=int($num/32); $mask[$ndx] |= (1<<($num % 32))}}} $nint32s = int(($ENV{'ncpus'}+31)/32); for($j = $nint32s - 1; $j >= 0; $j--) { if($j != $nint32s - 1){print ",";} $nblks = $j == $nint32s - 1 ? int(($ENV{'ncpus'} % 32)/4) : 8; for($i = $nblks - 1;$i >= 0;$i--){ printf("%01x",($mask[$j] >> ($i*4)) & 0xf);}}'`
+
+    if ! ncpus=$ncpus smp_affinity_mask=$smp_affinity_mask perl -e '@dirs = grep { -d } glob "/proc/irq/*"; foreach $dir (@dirs) { $hit = 0; $affinity_str = `cat $dir/smp_affinity`; chomp $affinity_str; @int32strs = split /,/, $affinity_str; @int32strs_mask=split /,/, $ENV{'smp_affinity_mask'}; for($i=0;$i <= $#int32strs_mask; $i++) { $int32strs_inv[$i] = sprintf("%08x",hex($int32strs_mask[$i])^0xffffffff); if($i == 0) { $len = int((($ENV{'ncpus'}%32)+3)/4); $int32strs_inv[$i] = substr($int32strs_inv[$i], -$len, $len); } } $inv = join(",", @int32strs_inv); $nint32s = int(($ENV{'ncpus'}+31)/32); for($j = $nint32s - 1; $j >= 0; $j--) { if(hex($int32strs[$nint32s - 1 - $j]) & hex($int32strs_mask[$nint32s - 1 - $j])) { $hit = 1; }} if($hit == 1) { $cmd = "echo $inv > $dir/smp_affinity 2>/dev/null"; system $cmd;}}'; then
+		echo "error: modifying /proc/irq/*/smp_affinity" >&2
+		error_exit "mcos_sys_mounted"
+	fi
+
+    banirq=`cat /proc/interrupts| perl -e 'while(<>) { if(/^\s*(\d+).*IHK\-SMP\s*$/) {print $1;}}'`
+
+    sed "s/%mask%/$smp_affinity_mask/g" $ETCDIR/irqbalance_mck.in | sed "s/%banirq%/$banirq/g" > /tmp/irqbalance_mck
+	systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null
+	if ! systemctl link $ETCDIR/irqbalance_mck.service >/dev/null 2>/dev/null; then
+		echo "error: linking irqbalance_mck" >&2
+		error_exit "mcos_sys_mounted"
+	fi
+
+    if ! systemctl start irqbalance_mck.service 2>/dev/null ; then
+		echo "error: starting irqbalance_mck" >&2
+		error_exit "mcos_sys_mounted"
+	fi
+#    echo cpus=$cpus mask=$smp_affinity_mask banirq=$banirq
 fi
+
+# Start mcklogd. Note that McKernel blocks when kmsg buffer is full
+# with '-k 1' until mcklogd unblocks it so starting mcklogd must preceed
+# booting McKernel
+if [ ${LOGMODE} -ne 0 ]; then
+	# Stop mcklogd which has survived McKernel shutdown because
+	# mcstop+release.sh is not used
+	pkill mcklogd
+	SBINDIR=${SBINDIR} ${SBINDIR}/mcklogd -i ${INTERVAL} -f ${facility}
+fi
+

arch/x86/tools/mcstop+release-smp-x86.sh.in

@@ -10,6 +10,7 @@
 prefix="@prefix@"
 BINDIR="@BINDIR@"
 SBINDIR="@SBINDIR@"
+ETCDIR=@ETCDIR@
 KMODDIR="@KMODDIR@"
 KERNDIR="@KERNDIR@"
 
@@ -17,31 +18,108 @@ mem=""
 cpus=""
 
 # No SMP module? Exit.
-if [ "`lsmod | grep ihk_smp_x86`" == "" ]; then exit; fi
+if ! grep ihk_smp_x86 /proc/modules &>/dev/null; then exit 0; fi
 
-# Remove delegator if loaded
-if [ "`lsmod | grep mcctrl`" != "" ]; then
-        if ! rmmod mcctrl; then echo "error: removing mcctrl"; exit; fi
-fi
-
-# Destroy all LWK instances
-for i in /dev/mcos*; do
-	ind=`echo $i|cut -c10-`;
-    if ! ${SBINDIR}/ihkconfig 0 destroy $ind; then echo "error: destroying LWK instance $ind failed"; exit; fi	
+# Stop mcklogd
+while pgrep "mcklogd" > /dev/null 2>&1;
+do
+	pkill -9 mcklogd
 done
 
-# Query IHK-SMP resources and release them
-if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus"; exit; fi
-cpus=`${SBINDIR}/ihkconfig 0 query cpu`
-if ! ${SBINDIR}/ihkconfig 0 release cpu $cpus > /dev/null; then echo "error: releasing CPUs"; exit; fi
-
-if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
-mem=`${SBINDIR}/ihkconfig 0 query mem`	
-if ! ${SBINDIR}/ihkconfig 0 release mem $mem > /dev/null; then echo "error: releasing memory"; exit; fi
-
-# Remove SMP module
-if [ "`lsmod | grep ihk_smp_x86`" != "" ]; then
-        if ! rmmod ihk_smp_x86; then echo "error: removing ihk_smp_x86"; exit; fi
+# Destroy all LWK instances
+if ls /dev/mcos* 1>/dev/null 2>&1; then
+	for i in /dev/mcos*; do
+		ind=`echo $i|cut -c10-`;
+		if ! ${SBINDIR}/ihkconfig 0 destroy $ind; then
+			echo "error: destroying LWK instance $ind failed" >&2
+			exit 1
+		fi
+	done
 fi
 
+# Query IHK-SMP resources and release them
+if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then
+	echo "error: querying cpus" >&2
+	exit 1
+fi
 
+cpus=`${SBINDIR}/ihkconfig 0 query cpu`
+if [ "${cpus}" != "" ]; then
+	if ! ${SBINDIR}/ihkconfig 0 release cpu $cpus > /dev/null; then
+		echo "error: releasing CPUs" >&2
+		exit 1
+	fi
+fi
+
+if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then
+	echo "error: querying memory" >&2
+	exit 1
+fi
+
+mem=`${SBINDIR}/ihkconfig 0 query mem`
+if [ "${mem}" != "" ]; then
+	if ! ${SBINDIR}/ihkconfig 0 release mem $mem > /dev/null; then
+		echo "error: releasing memory" >&2
+		exit 1
+	fi
+fi
+
+# Remove delegator if loaded
+if grep mcctrl /proc/modules &>/dev/null; then
+	if ! rmmod mcctrl 2>/dev/null; then
+		echo "error: removing mcctrl" >&2
+		exit 1
+	fi
+fi
+
+# Remove mcoverlay if loaded
+if grep mcoverlay /proc/modules &>/dev/null; then
+	if [ "`cat /proc/mounts | grep /tmp/mcos/mcos0_sys`" != "" ]; then umount -l /tmp/mcos/mcos0_sys; fi
+	if [ "`cat /proc/mounts | grep /tmp/mcos/mcos0_proc`" != "" ]; then umount -l /tmp/mcos/mcos0_proc; fi
+	if [ "`cat /proc/mounts | grep /tmp/mcos/linux_proc`" != "" ]; then umount -l /tmp/mcos/linux_proc; fi
+	if [ "`cat /proc/mounts | grep /tmp/mcos`" != "" ]; then umount -l /tmp/mcos; fi
+	if [ -e /tmp/mcos ]; then rm -rf /tmp/mcos; fi
+	if ! rmmod mcoverlay 2>/dev/null; then
+		echo "warning: failed to remove mcoverlay" >&2
+	fi
+fi
+
+# Remove SMP module
+if grep ihk_smp_x86 /proc/modules &>/dev/null; then
+	if ! rmmod ihk_smp_x86 2>/dev/null; then
+		echo "error: removing ihk_smp_x86" >&2
+		exit 1
+	fi
+fi
+
+# Remove core module
+if grep -E 'ihk\s' /proc/modules &>/dev/null; then
+	if ! rmmod ihk 2>/dev/null; then
+		echo "error: removing ihk" >&2
+		exit 1
+	fi
+fi
+
+# Start irqbalance with the original settings
+if [ "`systemctl status irqbalance_mck.service 2> /dev/null |grep -E 'Active: active'`" != "" ]; then
+	if ! systemctl stop irqbalance_mck.service 2>/dev/null; then
+		echo "warning: failed to stop irqbalance_mck" >&2
+	fi
+	if ! systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null; then
+		echo "warning: failed to disable irqbalance_mck" >&2
+	fi
+	if ! etcdir=@ETCDIR@ perl -e '$etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "$etcdir/proc/irq/*/smp_affinity"; foreach $file (@files) { $dest = substr($file, length($etcdir)); if(0) {print "cp $file $dest\n";} system("cp $file $dest 2>/dev/null"); }'; then
+		echo "warning: failed to restore /proc/irq/*/smp_affinity" >&2
+	fi
+	if ! systemctl start irqbalance.service; then
+		echo "warning: failed to start irqbalance" >&2;
+	fi
+fi
+
+# Re-enable ASLR
+if [ -f /tmp/mckernel_randomize_va_space ]; then
+	cat /tmp/mckernel_randomize_va_space > /proc/sys/kernel/randomize_va_space
+fi
+
+# Set back default swappiness
+echo 60 > /proc/sys/vm/swappiness

configure.ac

@@ -17,6 +17,13 @@ DCFA_RELEASE_DATE=DCFA_RELEASE_DATE_m4
 
 AC_PREFIX_DEFAULT([/opt/ppos])
 
+AC_CHECK_HEADER([numa.h],[numa_header_found=yes])
+AS_IF([test "x$numa_header_found" != "xyes"],
+	[AC_MSG_ERROR([Unable to find numa.h header file, missing numactl-devel?])])
+AC_CHECK_LIB([numa],[numa_run_on_node],[numa_lib_found=yes])
+AS_IF([test "x$numa_lib_found" != "xyes"],
+	[AC_MSG_ERROR([Unable to find NUMA library, missing numactl-devel?])])
+
 AC_ARG_WITH([kernelsrc],
   AC_HELP_STRING(
     [--with-kernelsrc=path],[Path to 'kernel src', default is /lib/modules/uname_r/build]),
@@ -48,6 +55,17 @@ AC_ARG_ENABLE([mcoverlayfs],
 	      [ENABLE_MCOVERLAYFS=$enableval],
 	      [ENABLE_MCOVERLAYFS=yes])
 
+AC_ARG_WITH([uname_r],
+  AC_HELP_STRING(
+    [--with-uname_r=uname_r],[Value of '`uname -r`' on the target platform, default is local value]),
+  [WITH_UNAME_R=$withval],[WITH_UNAME_R=yes])
+
+case "X$WITH_UNAME_R" in
+    Xyes | Xno | X)
+	WITH_UNAME_R='`uname -r`'
+	;;
+esac
+
 case "X$WITH_KERNELSRC" in
     Xyes | Xno | X)
 	WITH_KERNELSRC='/lib/modules/`uname -r`/build'
@@ -70,6 +88,7 @@ case $WITH_TARGET in
 	ARCH=`uname -m`
 	AC_PROG_CC
 	XCC=$CC
+	CFLAGS="$CFLAGS -ffreestanding -fno-tree-loop-distribute-patterns"
 	;;
     builtin-mic)
 	ARCH=k1om
@@ -146,6 +165,9 @@ case $WITH_TARGET in
 	if test "X$SBINDIR" = X; then
 		SBINDIR="$prefix/sbin"
 	fi
+	if test "X$ETCDIR" = X; then
+		ETCDIR="$prefix/etc"
+	fi
 	if test "X$KMODDIR" = X; then
 		KMODDIR="$prefix/kmod"
 	fi
@@ -159,6 +181,7 @@ case $WITH_TARGET in
 esac
 
 KDIR="$WITH_KERNELSRC"
+UNAME_R="$WITH_UNAME_R"
 TARGET="$WITH_TARGET"
 
 MCCTRL_LINUX_SYMTAB=""
@@ -218,6 +241,7 @@ AC_DEFUN([MCCTRL_FIND_KSYM],[
 ])
 
 MCCTRL_FIND_KSYM([sys_mount])
+MCCTRL_FIND_KSYM([sys_umount])
 MCCTRL_FIND_KSYM([sys_unshare])
 MCCTRL_FIND_KSYM([zap_page_range])
 MCCTRL_FIND_KSYM([vdso_image_64])
@@ -275,12 +299,15 @@ AC_SUBST(CC)
 AC_SUBST(XCC)
 AC_SUBST(ARCH)
 AC_SUBST(KDIR)
+AC_SUBST(UNAME_R)
 AC_SUBST(TARGET)
 AC_SUBST(BINDIR)
 AC_SUBST(SBINDIR)
+AC_SUBST(ETCDIR)
 AC_SUBST(KMODDIR)
 AC_SUBST(KERNDIR)
 AC_SUBST(MANDIR)
+AC_SUBST(CFLAGS)
 AC_SUBST(ENABLE_MCOVERLAYFS)
 
 AC_SUBST(IHK_VERSION)
@@ -298,6 +325,9 @@ AC_CONFIG_FILES([
 	executer/kernel/mcctrl/Makefile
 	executer/kernel/mcctrl/arch/x86_64/Makefile
 	executer/kernel/mcoverlayfs/Makefile
+	executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile
+	executer/kernel/mcoverlayfs/linux-4.0.9/Makefile
+	executer/kernel/mcoverlayfs/linux-4.6.7/Makefile
 	kernel/Makefile
 	kernel/Makefile.build
 	arch/x86/tools/mcreboot-attached-mic.sh
@@ -305,8 +335,11 @@ AC_CONFIG_FILES([
 	arch/x86/tools/mcreboot-builtin-x86.sh
 	arch/x86/tools/mcreboot-smp-x86.sh
 	arch/x86/tools/mcstop+release-smp-x86.sh
+	arch/x86/tools/eclair-dump-backtrace.exp
 	arch/x86/tools/mcshutdown-builtin-x86.sh
 	arch/x86/tools/mcreboot.1:arch/x86/tools/mcreboot.1in
+	arch/x86/tools/irqbalance_mck.service
+	arch/x86/tools/irqbalance_mck.in
 ])
 
 AS_IF([test "x$enable_dcfa" = xyes], [

executer/config.h.in

@@ -51,6 +51,9 @@
 /* Define to address of kernel symbol sys_readlink, or 0 if exported */
 #undef MCCTRL_KSYM_sys_readlink
 
+/* Define to address of kernel symbol sys_umount, or 0 if exported */
+#undef MCCTRL_KSYM_sys_umount
+
 /* Define to address of kernel symbol sys_unshare, or 0 if exported */
 #undef MCCTRL_KSYM_sys_unshare
 

executer/include/uprotocol.h

@@ -41,6 +41,8 @@
 #define MCEXEC_UP_NEW_PROCESS	 0x30a02909
 #define MCEXEC_UP_GET_CRED	 0x30a0290a
 #define MCEXEC_UP_GET_CREDV	 0x30a0290b
+#define MCEXEC_UP_GET_NODES  0x30a0290c
+#define MCEXEC_UP_GET_CPUSET  0x30a0290d
 
 #define MCEXEC_UP_PREPARE_DMA    0x30a02910
 #define MCEXEC_UP_FREE_DMA       0x30a02911
@@ -49,7 +51,8 @@
 #define MCEXEC_UP_CLOSE_EXEC     0x30a02913
 
 #define MCEXEC_UP_SYS_MOUNT      0x30a02914
-#define MCEXEC_UP_SYS_UNSHARE    0x30a02915
+#define MCEXEC_UP_SYS_UMOUNT     0x30a02915
+#define MCEXEC_UP_SYS_UNSHARE    0x30a02916
 
 #define MCEXEC_UP_DEBUG_LOG     0x40000000
 
@@ -77,6 +80,18 @@ struct program_image_section {
 #define SHELL_PATH_MAX_LEN	1024
 #define MCK_RLIM_MAX	20
 
+struct get_cpu_set_arg {
+	int nr_processes;
+	void *cpu_set;
+	size_t cpu_set_size;	// Size in bytes
+	int *target_core;
+	int *mcexec_linux_numa; // NUMA domain to bind mcexec to
+};
+
+#define PLD_CPU_SET_MAX_CPUS 1024
+typedef unsigned long __cpu_set_unit;
+#define PLD_CPU_SET_SIZE (PLD_CPU_SET_MAX_CPUS / (8 * sizeof(__cpu_set_unit)))
+
 struct program_load_desc {
 	int num_sections;
 	int status;
@@ -106,10 +121,18 @@ struct program_load_desc {
 	struct rlimit rlimit[MCK_RLIM_MAX];
 	unsigned long interp_align;
 	char shell_path[SHELL_PATH_MAX_LEN];
+	__cpu_set_unit cpu_set[PLD_CPU_SET_SIZE];
 	struct program_image_section sections[0];
 };
 
 struct syscall_request {
+	/* TID of requesting thread */
+	int rtid;
+	/*
+	 * TID of target thread. Remote page fault response needs to designate the
+	 * thread that must serve the request, 0 indicates any thread from the pool
+	 */
+	int ttid;
 	unsigned long valid;
 	unsigned long number;
 	unsigned long args[6];
@@ -128,8 +151,17 @@ struct syscall_load_desc {
 	unsigned long size;
 };
 
+#define IHK_SCD_REQ_THREAD_SPINNING         0
+#define IHK_SCD_REQ_THREAD_TO_BE_WOKEN      1
+#define IHK_SCD_REQ_THREAD_DESCHEDULED      2
+
 struct syscall_response {
+	/* TID of the thread that requested the service */
+	int ttid;
+	/* TID of the mcexec thread that is serving or has served the request */
+	int stid;
 	unsigned long status;
+	unsigned long req_thread_status;
 	long ret;
 	unsigned long fault_address;
 	unsigned long fault_reason;
@@ -180,8 +212,42 @@ struct sys_mount_desc {
 	void *data;
 };
 
+struct sys_umount_desc {
+	char *dir_name;
+};
+
 struct sys_unshare_desc {
 	unsigned long unshare_flags;
 };
 
+enum perf_ctrl_type {
+	PERF_CTRL_SET,
+	PERF_CTRL_GET,
+	PERF_CTRL_ENABLE,
+	PERF_CTRL_DISABLE,
+};
+
+struct perf_ctrl_desc {
+	enum perf_ctrl_type ctrl_type;
+	int status;
+	union {
+		/* for SET, GET */
+		struct {
+			unsigned int target_cntr;
+			unsigned long config;
+			unsigned long read_value;
+			unsigned disabled        :1,
+			         pinned          :1,
+			         exclude_user    :1,
+			         exclude_kernel  :1,
+			         exclude_hv      :1,
+			         exclude_idle    :1;
+		};
+
+		/* for START, STOP*/
+		struct {
+			unsigned long target_cntr_mask;
+		};
+	};
+};
 #endif

executer/kernel/mcctrl/arch/x86_64/archdeps.c

@@ -1,5 +1,5 @@
 #include <linux/version.h>
-#include "../../../../config.h"
+#include "../../config.h"
 #include "../../mcctrl.h"
 
 #ifdef MCCTRL_KSYM_vdso_image_64
@@ -64,6 +64,8 @@ reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp, unsign
 	unsigned long start = 0L;
 	unsigned long end;
 
+	mutex_lock(&usrdata->reserve_lock);
+
 #define	DESIRED_USER_END	0x800000000000
 #define	GAP_FOR_MCEXEC		0x008000000000UL
 	end = DESIRED_USER_END;
@@ -81,6 +83,8 @@ reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp, unsign
 	up_write(&current->mm->mmap_sem);
 #endif
 
+	mutex_unlock(&usrdata->reserve_lock);
+
 	if (IS_ERR_VALUE(start)) {
 		return start;
 	}
@@ -100,8 +104,6 @@ void get_vdso_info(ihk_os_t os, long vdso_rpa)
 	vdso_pa = ihk_device_map_memory(dev, vdso_rpa, sizeof(*vdso));
 	vdso = ihk_device_map_virtual(dev, vdso_pa, sizeof(*vdso), NULL, 0);
 
-	memset(vdso, 0, sizeof(*vdso));
-
 	/* VDSO pages */
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,16,0)
 	size = vdso_image->size;

executer/kernel/mcctrl/binfmt_mcexec.c

@@ -75,7 +75,7 @@ static int load_elf(struct linux_binprm *bprm
 	char buf[32];
 	int l;
 	int pass;
-	char pbuf[1024];
+	char *pbuf;
 	const char *path;
 
 	if(bprm->envc == 0)
@@ -88,6 +88,11 @@ static int load_elf(struct linux_binprm *bprm
 	if(elf_ex->e_ident[EI_CLASS] != ELFCLASS64)
 		return -ENOEXEC;
 
+	pbuf = kmalloc(1024, GFP_ATOMIC);
+	if (!pbuf) {
+		printk("%s: error: allocating pbuf\n", __FUNCTION__);
+		return -ENOMEM;
+	}
 	path = d_path(&bprm->file->f_path, pbuf, 1024);
 	if(!path || IS_ERR(path))
 		path = bprm->interp;
@@ -96,8 +101,10 @@ static int load_elf(struct linux_binprm *bprm
 	if(!cp ||
 	   !strcmp(cp, "/mcexec") ||
 	   !strcmp(cp, "/ihkosctl") ||
-	   !strcmp(cp, "/ihkconfig"))
+	   !strcmp(cp, "/ihkconfig")) {
+		kfree(pbuf);
 		return -ENOEXEC;
+	}
 
 	cnt[0] = bprm->argc;
 	cnt[1] = bprm->envc;
@@ -124,8 +131,10 @@ static int load_elf(struct linux_binprm *bprm
 				                        bprm->p, 1, 0, 1,
 				                        &page, NULL);
 #endif
-				if(rc <= 0)
+				if(rc <= 0) {
+					kfree(pbuf);
 					return -EFAULT;
+				}
 				addr = kmap_atomic(page
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,4,0)
 				                   , KM_USER0
@@ -199,21 +208,27 @@ static int load_elf(struct linux_binprm *bprm
 	for(ep = env; ep->name; ep++)
 		if(ep->val)
 			kfree(ep->val);
-	if(rc)
+	if(rc) {
+		kfree(pbuf);
 		return -ENOEXEC;
+	}
 
 	file = open_exec(MCEXEC_PATH);
-	if (IS_ERR(file))
+	if (IS_ERR(file)) {
+		kfree(pbuf);
 		return -ENOEXEC;
+	}
 
 	rc = remove_arg_zero(bprm);
 	if (rc){
 		fput(file);
+		kfree(pbuf);
 		return rc;
 	}
 	rc = copy_strings_kernel(1, &bprm->interp, bprm);
 	if (rc < 0){
 		fput(file);
+		kfree(pbuf);
 		return rc; 
 	}
 	bprm->argc++;
@@ -221,12 +236,14 @@ static int load_elf(struct linux_binprm *bprm
 	rc = copy_strings_kernel(1, &wp, bprm);
 	if (rc){
 		fput(file);
+		kfree(pbuf);
 		return rc; 
 	}
 	bprm->argc++;
 	rc = bprm_change_interp(MCEXEC_PATH, bprm);
 	if (rc < 0){
 		fput(file);
+		kfree(pbuf);
 		return rc;
 	}
 
@@ -236,8 +253,12 @@ static int load_elf(struct linux_binprm *bprm
 
 	rc = prepare_binprm(bprm);
 	if (rc < 0){
+		kfree(pbuf);
 		return rc;
 	}
+
+	kfree(pbuf);
+
 	return search_binary_handler(bprm
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,8,0)
 	                             , regs
@@ -255,7 +276,7 @@ void __init binfmt_mcexec_init(void)
 	insert_binfmt(&mcexec_format);
 }
 
-void __exit binfmt_mcexec_exit(void)
+void binfmt_mcexec_exit(void)
 {
 	unregister_binfmt(&mcexec_format);
 }

executer/kernel/mcctrl/driver.c

@@ -27,6 +27,7 @@
 #include <linux/slab.h>
 #include <linux/device.h>
 #include "mcctrl.h"
+#include <ihk/ihk_host_user.h>
 
 #define OS_MAX_MINOR 64
 
@@ -60,6 +61,8 @@ static struct ihk_os_user_call_handler mcctrl_uchs[] = {
 	{ .request = MCEXEC_UP_LOAD_SYSCALL, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SEND_SIGNAL, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_GET_CPU, .func = mcctrl_ioctl },
+	{ .request = MCEXEC_UP_GET_NODES, .func = mcctrl_ioctl },
+	{ .request = MCEXEC_UP_GET_CPUSET, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_STRNCPY_FROM_USER, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_NEW_PROCESS, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_PREPARE_DMA, .func = mcctrl_ioctl },
@@ -69,8 +72,15 @@ static struct ihk_os_user_call_handler mcctrl_uchs[] = {
 	{ .request = MCEXEC_UP_GET_CRED, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_GET_CREDV, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SYS_MOUNT, .func = mcctrl_ioctl },
+	{ .request = MCEXEC_UP_SYS_UMOUNT, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SYS_UNSHARE, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_DEBUG_LOG, .func = mcctrl_ioctl },
+	{ .request = IHK_OS_AUX_PERF_NUM, .func = mcctrl_ioctl },
+	{ .request = IHK_OS_AUX_PERF_SET, .func = mcctrl_ioctl },
+	{ .request = IHK_OS_AUX_PERF_GET, .func = mcctrl_ioctl },
+	{ .request = IHK_OS_AUX_PERF_ENABLE, .func = mcctrl_ioctl },
+	{ .request = IHK_OS_AUX_PERF_DISABLE, .func = mcctrl_ioctl },
+	{ .request = IHK_OS_AUX_PERF_DESTROY, .func = mcctrl_ioctl },
 };
 
 static struct ihk_os_user_call mcctrl_uc_proto = {
@@ -82,79 +92,118 @@ static struct ihk_os_user_call mcctrl_uc[OS_MAX_MINOR];
 
 static ihk_os_t os[OS_MAX_MINOR];
 
-ihk_os_t
-osnum_to_os(int n)
+ihk_os_t osnum_to_os(int n)
 {
 	return os[n];
 }
 
-static int __init mcctrl_init(void)
+/* OS event notifier implementation */
+int mcctrl_os_boot_notifier(int os_index)
 {
-	int	i;
 	int	rc;
 
-	rc = -ENOENT;
-	for(i = 0; i < OS_MAX_MINOR; i++){
-		os[i] = ihk_host_find_os(i, NULL);
-		if (os[i]) {
-			printk("OS #%d found.\n", i);
-			rc = 0;
-		}
-	}
-	if(rc){
-		printk("OS not found.\n");
-		return rc;
+	os[os_index] = ihk_host_find_os(os_index, NULL);
+	if (!os[os_index]) {
+		printk("mcctrl: error: OS ID %d couldn't be found\n", os_index);
+		return -EINVAL;
 	}
 
-	for(i = 0; i < OS_MAX_MINOR; i++){
-		if (os[i]) {
-			if (prepare_ikc_channels(os[i]) != 0) {
-				printk("Preparing syscall channels failed.\n");
-				os[i] = NULL;
-			}
-		}
+	if (prepare_ikc_channels(os[os_index]) != 0) {
+		printk("mcctrl: error: preparing IKC channels for OS %d\n", os_index);
+
+		os[os_index] = NULL;
+		return -EFAULT;
 	}
 
+	memcpy(mcctrl_uc + os_index, &mcctrl_uc_proto, sizeof mcctrl_uc_proto);
+
+	rc = ihk_os_register_user_call_handlers(os[os_index], mcctrl_uc + os_index);
+	if (rc < 0) {
+		destroy_ikc_channels(os[os_index]);
+		printk("mcctrl: error: registering callbacks for OS %d\n", os_index);
+
+		goto error_cleanup_channels;
+	}
+
+	procfs_init(os_index);
+	printk("mcctrl: OS ID %d boot event handled\n", os_index);
+
+	return 0;
+
+error_cleanup_channels:
+	destroy_ikc_channels(os[os_index]);
+
+	os[os_index] = NULL;
+	return rc;
+}
+
+int mcctrl_os_shutdown_notifier(int os_index)
+{
+	if (os[os_index]) {
+		sysfsm_cleanup(os[os_index]);
+		free_topology_info(os[os_index]);
+		ihk_os_unregister_user_call_handlers(os[os_index], mcctrl_uc + os_index);
+		destroy_ikc_channels(os[os_index]);
+		procfs_exit(os_index);
+	}
+
+	os[os_index] = NULL;
+
+	printk("mcctrl: OS ID %d shutdown event handled\n", os_index);
+	return 0;
+}
+
+static struct ihk_os_notifier_ops mcctrl_os_notifier_ops = {
+	.boot = mcctrl_os_boot_notifier,
+	.shutdown = mcctrl_os_shutdown_notifier,
+};
+
+static struct ihk_os_notifier mcctrl_os_notifier = {
+	.ops = &mcctrl_os_notifier_ops,
+};
+
+static int __init mcctrl_init(void)
+{
+	int ret = 0;
+	int i;
+
 #ifndef DO_USER_MODE
 	mcctrl_syscall_init();
 #endif
 
-	rus_page_hash_init();
-
-	for(i = 0; i < OS_MAX_MINOR; i++){
-		if (os[i]) {
-			memcpy(mcctrl_uc + i, &mcctrl_uc_proto, sizeof mcctrl_uc_proto);
-			rc = ihk_os_register_user_call_handlers(os[i], mcctrl_uc + i);
-			if(rc < 0){
-				destroy_ikc_channels(os[i]);
-				os[i] = NULL;
-			}
-			procfs_init(i);
-		}
+	for (i = 0; i < OS_MAX_MINOR; ++i) {
+		os[i] = NULL;
 	}
 
+	rus_page_hash_init();
+
 	binfmt_mcexec_init();
 
-	return 0;
+	if ((ret = ihk_host_register_os_notifier(&mcctrl_os_notifier)) != 0) {
+		printk("mcctrl: error: registering OS notifier\n");
+		goto error;
+	}
+
+	printk("mcctrl: initialized successfully.\n");
+	return ret;
+
+error:
+	binfmt_mcexec_exit();
+	rus_page_hash_put_pages();
+
+	return ret;
 }
 
 static void __exit mcctrl_exit(void)
 {
-	int	i;
-
-	binfmt_mcexec_exit();
-	printk("mcctrl: unregistered.\n");
-	for(i = 0; i < OS_MAX_MINOR; i++){
-		if(os[i]){
-			sysfsm_cleanup(os[i]);
-			free_topology_info(os[i]);
-			ihk_os_unregister_user_call_handlers(os[i], mcctrl_uc + i);
-			destroy_ikc_channels(os[i]);
-			procfs_exit(i);
-		}
+	if (ihk_host_deregister_os_notifier(&mcctrl_os_notifier) != 0) {
+		printk("mcctrl: warning: failed to deregister OS notifier??\n");
 	}
 
+	binfmt_mcexec_exit();
 	rus_page_hash_put_pages();
+
+	printk("mcctrl: unregistered.\n");
 }
 
 MODULE_LICENSE("GPL v2");

executer/kernel/mcctrl/ikc.c

@@ -27,6 +27,7 @@
 #include <linux/miscdevice.h>
 #include <linux/slab.h>
 #include <linux/string.h>
+#include <linux/interrupt.h>
 #include "mcctrl.h"
 #ifdef ATTACHED_MIC
 #include <sysdeps/mic/mic/micconst.h>
@@ -34,22 +35,37 @@
 
 #define REQUEST_SHIFT    16
 
+//#define DEBUG_IKC
+
+#ifdef DEBUG_IKC
+#define	dkprintf(...) kprintf(__VA_ARGS__)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) printk(__VA_ARGS__); } while (0)
+#define	ekprintf(...) printk(__VA_ARGS__)
+#endif
+
 //int num_channels;
 
 //struct mcctrl_channel *channels;
 
 void mcexec_prepare_ack(ihk_os_t os, unsigned long arg, int err);
 static void mcctrl_ikc_init(ihk_os_t os, int cpu, unsigned long rphys, struct ihk_ikc_channel_desc *c);
-int mcexec_syscall(struct mcctrl_channel *c, int pid, unsigned long arg);
+int mcexec_syscall(struct mcctrl_usrdata *ud, struct ikc_scd_packet *packet);
 void sig_done(unsigned long arg, int err);
+void mcctrl_perf_ack(ihk_os_t os, struct ikc_scd_packet *packet);
+void mcctrl_os_read_write_cpu_response(ihk_os_t os,
+		struct ikc_scd_packet *pisp);
 
+/* XXX: this runs in atomic context! */
 static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
                                   void *__packet, void *__os)
 {
 	struct ikc_scd_packet *pisp = __packet;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(__os);
+	int msg = pisp->msg;
 
-	switch (pisp->msg) {
+	switch (msg) {
 	case SCD_MSG_INIT_CHANNEL:
 		mcctrl_ikc_init(__os, pisp->ref, pisp->arg, c);
 		break;
@@ -63,11 +79,11 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		break;
 
 	case SCD_MSG_SYSCALL_ONESIDE:
-		mcexec_syscall(usrdata->channels + pisp->ref, pisp->pid, pisp->arg);
+		mcexec_syscall(usrdata, pisp);
 		break;
 
 	case SCD_MSG_PROCFS_ANSWER:
-		procfs_answer(pisp->arg, pisp->err);
+		procfs_answer(usrdata, pisp->pid);
 		break;
 
 	case SCD_MSG_SEND_SIGNAL:
@@ -88,19 +104,20 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		break;
 
 	case SCD_MSG_PROCFS_TID_CREATE:
-		add_tid_entry(ihk_host_os_get_index(__os), pisp->pid, pisp->arg);
-		break;
-
 	case SCD_MSG_PROCFS_TID_DELETE:
-		delete_tid_entry(ihk_host_os_get_index(__os), pisp->pid, pisp->arg);
+		procfsm_packet_handler(__os, pisp->msg, pisp->pid, pisp->arg);
 		break;
 
 	case SCD_MSG_GET_VDSO_INFO:
 		get_vdso_info(__os, pisp->arg);
 		break;
 
-	case SCD_MSG_REPLY_GET_CPU_MAPPING:
-		reply_get_cpu_mapping(pisp->arg);
+	case SCD_MSG_PERF_ACK:
+		mcctrl_perf_ack(__os, pisp);
+		break;
+
+	case SCD_MSG_CPU_RW_REG_RESP:
+		mcctrl_os_read_write_cpu_response(__os, pisp);
 		break;
 
 	default:
@@ -110,6 +127,14 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 				pisp->err, pisp->arg);
 		break;
 	}
+
+	/* 
+	 * SCD_MSG_SYSCALL_ONESIDE holds the packet and frees is it
+	 * mcexec_ret_syscall(), for the rest, free it here.
+	 */
+	if (msg != SCD_MSG_SYSCALL_ONESIDE) {
+		ihk_ikc_release_packet((struct ihk_ikc_free_packet *)__packet, c);
+	}
 	return 0;
 }
 
@@ -146,8 +171,6 @@ int mcctrl_ikc_set_recv_cpu(ihk_os_t os, int cpu)
 
 	ihk_ikc_channel_set_cpu(usrdata->channels[cpu].c,
 	                        ihk_ikc_get_processor_id());
-	kprintf("Setting the target to %d\n",
-	        ihk_ikc_get_processor_id());
 	return 0;
 }
 
@@ -162,91 +185,26 @@ int mcctrl_ikc_is_valid_thread(ihk_os_t os, int cpu)
 	}
 }
 
-//unsigned long *mcctrl_doorbell_va;
-//unsigned long mcctrl_doorbell_pa;
-
 static void mcctrl_ikc_init(ihk_os_t os, int cpu, unsigned long rphys, struct ihk_ikc_channel_desc *c)
 {
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct ikc_scd_packet packet;
 	struct mcctrl_channel *pmc = usrdata->channels + cpu;
-	unsigned long phys;
-	struct ikc_scd_init_param *rpm;
 
-	if(c->port == 502)
+	if (c->port == 502) {
 		pmc = usrdata->channels + usrdata->num_channels - 1;
-
-	if (!pmc) {
-		return;
 	}
 
-	printk("IKC init: cpu=%d port=%d\n", cpu, c->port);
-
-	phys = ihk_device_map_memory(ihk_os_to_dev(os), rphys,
-	                             sizeof(struct ikc_scd_init_param));
-#ifdef CONFIG_MIC
-	rpm = ioremap_wc(phys, sizeof(struct ikc_scd_init_param));
-#else
-	rpm = ihk_device_map_virtual(ihk_os_to_dev(os), phys, 
-	                             sizeof(struct ikc_scd_init_param),
-								 NULL, 0);
-#endif
-
-	pmc->param.request_va =
-		(void *)__get_free_pages(GFP_KERNEL,
-		                         REQUEST_SHIFT - PAGE_SHIFT);
-	pmc->param.request_pa = virt_to_phys(pmc->param.request_va);
-	pmc->param.doorbell_va = usrdata->mcctrl_doorbell_va;
-	pmc->param.doorbell_pa = usrdata->mcctrl_doorbell_pa;
-	pmc->param.post_va = (void *)__get_free_page(GFP_KERNEL);
-	pmc->param.post_pa = virt_to_phys(pmc->param.post_va);
-	memset(pmc->param.doorbell_va, 0, PAGE_SIZE);
-	memset(pmc->param.request_va, 0, PAGE_SIZE);
-	memset(pmc->param.post_va, 0, PAGE_SIZE);
-	
-	pmc->param.response_rpa = rpm->response_page;
-	pmc->param.response_pa 
-		= ihk_device_map_memory(ihk_os_to_dev(os),
-		                        pmc->param.response_rpa,
-		                        PAGE_SIZE);
-#ifdef CONFIG_MIC							
-	pmc->param.response_va = ioremap_cache(pmc->param.response_pa,
-	                                       PAGE_SIZE);
-#else
-	pmc->param.response_va = ihk_device_map_virtual(ihk_os_to_dev(os), 
-	                                                pmc->param.response_pa,
-													PAGE_SIZE, NULL, 0);
-#endif
-
-	pmc->dma_buf = (void *)__get_free_pages(GFP_KERNEL,
-	                                        DMA_PIN_SHIFT - PAGE_SHIFT);
-
-	rpm->request_page = pmc->param.request_pa;
-	rpm->doorbell_page = pmc->param.doorbell_pa;
-	rpm->post_page = pmc->param.post_pa;
+	if (!pmc) {
+		kprintf("%s: error: no channel found?\n", __FUNCTION__);
+		return;
+	}
 
 	packet.msg = SCD_MSG_INIT_CHANNEL_ACKED;
 	packet.ref = cpu;
 	packet.arg = rphys;
 
-	printk("Request: %lx, Response: %lx, Doorbell: %lx\n",
-	       pmc->param.request_pa, pmc->param.response_rpa,
-	       pmc->param.doorbell_pa);
-	printk("Request: %p, Response: %p, Doorbell: %p\n",
-	       pmc->param.request_va, pmc->param.response_va,
-	       pmc->param.doorbell_va);
-
 	ihk_ikc_send(pmc->c, &packet, 0);
-
-#ifdef CONFIG_MIC							
-	iounmap(rpm);
-#else
-	ihk_device_unmap_virtual(ihk_os_to_dev(os), rpm, 
-	                         sizeof(struct ikc_scd_init_param));
-#endif
-
-	ihk_device_unmap_memory(ihk_os_to_dev(os), phys,
-	                        sizeof(struct ikc_scd_init_param));
 }
 
 static int connect_handler(struct ihk_ikc_channel_info *param)
@@ -265,11 +223,8 @@ static int connect_handler(struct ihk_ikc_channel_info *param)
 	}
 	param->packet_handler = syscall_packet_handler;
 	
-	INIT_LIST_HEAD(&usrdata->channels[cpu].wq_list);
-	spin_lock_init(&usrdata->channels[cpu].wq_list_lock);
-
 	usrdata->channels[cpu].c = c;
-	kprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
+	dkprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
 
 	return 0;
 }
@@ -286,11 +241,8 @@ static int connect_handler2(struct ihk_ikc_channel_info *param)
 
 	param->packet_handler = syscall_packet_handler;
 	
-	INIT_LIST_HEAD(&usrdata->channels[cpu].wq_list);
-	spin_lock_init(&usrdata->channels[cpu].wq_list_lock);
-
 	usrdata->channels[cpu].c = c;
-	kprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
+	dkprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
 
 	return 0;
 }
@@ -299,7 +251,7 @@ static struct ihk_ikc_listen_param listen_param = {
 	.port = 501,
 	.handler = connect_handler,
 	.pkt_size = sizeof(struct ikc_scd_packet),
-	.queue_size = PAGE_SIZE,
+	.queue_size = PAGE_SIZE * 4,
 	.magic = 0x1129,
 };
 
@@ -307,76 +259,66 @@ static struct ihk_ikc_listen_param listen_param2 = {
 	.port = 502,
 	.handler = connect_handler2,
 	.pkt_size = sizeof(struct ikc_scd_packet),
-	.queue_size = PAGE_SIZE,
+	.queue_size = PAGE_SIZE * 4,
 	.magic = 0x1329,
 };
 
 int prepare_ikc_channels(ihk_os_t os)
 {
-	struct ihk_cpu_info *info;
-	struct mcctrl_usrdata   *usrdata;
-	int error;
+	struct mcctrl_usrdata *usrdata;
+	int i;
 
 	usrdata = kzalloc(sizeof(struct mcctrl_usrdata), GFP_KERNEL);
-	usrdata->mcctrl_doorbell_va = (void *)__get_free_page(GFP_KERNEL);
-	usrdata->mcctrl_doorbell_pa = virt_to_phys(usrdata->mcctrl_doorbell_va);
 
-	info = ihk_os_get_cpu_info(os);
-	if (!info) {
-		printk("Error: cannot retrieve CPU info.\n");
+	usrdata->cpu_info = ihk_os_get_cpu_info(os);
+	usrdata->mem_info = ihk_os_get_memory_info(os);
+
+	if (!usrdata->cpu_info || !usrdata->mem_info) {
+		printk("Error: cannot obtain OS CPU and memory information.\n");
 		return -EINVAL;
 	}
-	if (info->n_cpus < 1) {
+
+	if (usrdata->cpu_info->n_cpus < 1) {
 		printk("Error: # of cpu is invalid.\n");
 		return -EINVAL;
 	}
 
-	usrdata->num_channels = info->n_cpus + 1;
-	usrdata->channels = kzalloc(sizeof(struct mcctrl_channel) * usrdata->num_channels,
-	                   GFP_KERNEL);
+	usrdata->num_channels = usrdata->cpu_info->n_cpus + 1;
+	usrdata->channels = kzalloc(sizeof(struct mcctrl_channel) *
+			usrdata->num_channels,
+			GFP_KERNEL);
+
 	if (!usrdata->channels) {
 		printk("Error: cannot allocate channels.\n");
 		return -ENOMEM;
 	}
 
 	usrdata->os = os;
-	init_waitqueue_head(&usrdata->wq_prepare);
 	ihk_host_os_set_usrdata(os, usrdata);
 	memcpy(&usrdata->listen_param, &listen_param, sizeof listen_param);
 	ihk_ikc_listen_port(os, &usrdata->listen_param);
 	memcpy(&usrdata->listen_param2, &listen_param2, sizeof listen_param2);
 	ihk_ikc_listen_port(os, &usrdata->listen_param2);
+	init_waitqueue_head(&usrdata->wq_procfs);
+	mutex_init(&usrdata->reserve_lock);
 
-	INIT_LIST_HEAD(&usrdata->per_proc_list);
-	spin_lock_init(&usrdata->per_proc_list_lock);
+	for (i = 0; i < MCCTRL_PER_PROC_DATA_HASH_SIZE; ++i) {
+		INIT_LIST_HEAD(&usrdata->per_proc_data_hash[i]);
+		rwlock_init(&usrdata->per_proc_data_hash_lock[i]);
+	}
 
 	INIT_LIST_HEAD(&usrdata->cpu_topology_list);
 	INIT_LIST_HEAD(&usrdata->node_topology_list);
 
-	error = init_peer_channel_registry(usrdata);
-	if (error) {
-		return error;
-	}
+	mutex_init(&usrdata->part_exec.lock);
+	usrdata->part_exec.nr_processes = -1;
 
 	return 0;
 }
 
 void __destroy_ikc_channel(ihk_os_t os, struct mcctrl_channel *pmc)
 {
-	free_pages((unsigned long)pmc->param.request_va,
-	           REQUEST_SHIFT - PAGE_SHIFT);
-	free_page((unsigned long)pmc->param.post_va);
-
-#ifdef CONFIG_MIC
-	iounmap(pmc->param.response_va);
-#else
-	ihk_device_unmap_virtual(ihk_os_to_dev(os), pmc->param.response_va, 
-	                         PAGE_SIZE);
-#endif
-	ihk_device_unmap_memory(ihk_os_to_dev(os),
-	                        pmc->param.response_pa, PAGE_SIZE);
-	free_pages((unsigned long)pmc->dma_buf,
-	           DMA_PIN_SHIFT - PAGE_SHIFT);
+	return;
 }
 
 void destroy_ikc_channels(ihk_os_t os)
@@ -384,6 +326,11 @@ void destroy_ikc_channels(ihk_os_t os)
 	int i;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 
+	if (!usrdata) {
+		printk("%s: WARNING: no mcctrl_usrdata found\n", __FUNCTION__);
+		return;
+	}
+
 	ihk_host_os_set_usrdata(os, NULL);
 
 	for (i = 0; i < usrdata->num_channels; i++) {
@@ -391,12 +338,9 @@ void destroy_ikc_channels(ihk_os_t os)
 //			ihk_ikc_disconnect(usrdata->channels[i].c);
 			ihk_ikc_free_channel(usrdata->channels[i].c);
 			__destroy_ikc_channel(os, usrdata->channels + i);
-			printk("Channel #%d freed.\n", i);
 		}
 	}
-	free_page((unsigned long)usrdata->mcctrl_doorbell_va);
 
-	destroy_peer_channel_registry(usrdata);
 	kfree(usrdata->channels);
 	kfree(usrdata);
 }

executer/kernel/mcctrl/mcctrl.h

@@ -41,13 +41,16 @@
 #include <ikc/master.h>
 #include <ihk/msr.h>
 #include <linux/semaphore.h>
+#include <linux/rwlock.h>
 #include <linux/threads.h>
 #include "sysfs.h"
+#include "mcctrl_public.h"
 
 #define SCD_MSG_PREPARE_PROCESS         0x1
 #define SCD_MSG_PREPARE_PROCESS_ACKED   0x2
 #define SCD_MSG_PREPARE_PROCESS_NACKED  0x7
 #define SCD_MSG_SCHEDULE_PROCESS        0x3
+#define SCD_MSG_WAKE_UP_SYSCALL_THREAD  0x14
 
 #define SCD_MSG_INIT_CHANNEL            0x5
 #define SCD_MSG_INIT_CHANNEL_ACKED      0x6
@@ -57,8 +60,8 @@
 #define SCD_MSG_CLEANUP_PROCESS         0x9
 #define SCD_MSG_GET_VDSO_INFO           0xa
 
-#define SCD_MSG_GET_CPU_MAPPING         0xc
-#define SCD_MSG_REPLY_GET_CPU_MAPPING   0xd
+//#define SCD_MSG_GET_CPU_MAPPING         0xc
+//#define SCD_MSG_REPLY_GET_CPU_MAPPING   0xd
 
 #define	SCD_MSG_PROCFS_CREATE		0x10
 #define	SCD_MSG_PROCFS_DELETE		0x11
@@ -90,6 +93,12 @@
 #define SCD_MSG_PROCFS_TID_CREATE	0x44
 #define SCD_MSG_PROCFS_TID_DELETE	0x45
 
+#define SCD_MSG_PERF_CTRL               0x50
+#define SCD_MSG_PERF_ACK                0x51
+
+#define SCD_MSG_CPU_RW_REG              0x52
+#define SCD_MSG_CPU_RW_REG_RESP         0x53
+
 #define DMA_PIN_SHIFT                   21
 
 #define DO_USER_MODE
@@ -101,6 +110,12 @@ struct coretable {
 	unsigned long addr;
 };
 
+enum mcctrl_os_cpu_operation {
+	MCCTRL_OS_CPU_READ_REGISTER,
+	MCCTRL_OS_CPU_WRITE_REGISTER,
+	MCCTRL_OS_CPU_MAX_OP
+};
+
 struct ikc_scd_packet {
 	int msg;
 	int err;
@@ -110,8 +125,9 @@ struct ikc_scd_packet {
 			int ref;
 			int osnum;
 			int pid;
-			int padding;
 			unsigned long arg;
+			struct syscall_request req;
+			unsigned long resp_pa;
 		};
 
 		/* for SCD_MSG_SYSFS_* */
@@ -120,7 +136,20 @@ struct ikc_scd_packet {
 			long sysfs_arg2;
 			long sysfs_arg3;
 		};
+
+		/* SCD_MSG_SCHEDULE_THREAD */
+		struct {
+			int ttid;
+		};
+
+		/* SCD_MSG_CPU_RW_REG */
+		struct {
+			struct mcctrl_os_cpu_register desc;
+			enum mcctrl_os_cpu_operation op;
+			void *resp;
+		};
 	};
+	char padding[12];
 };
 
 struct mcctrl_priv { 
@@ -154,24 +183,48 @@ struct syscall_params {
 struct wait_queue_head_list_node {
 	struct list_head list;
 	wait_queue_head_t wq_syscall;
-	int pid;
+	struct task_struct *task;
+	/* Denotes an exclusive wait for requester TID rtid */
+	int rtid;
 	int req;
+	struct ikc_scd_packet *packet;
 };
 
 struct mcctrl_channel {
 	struct ihk_ikc_channel_desc *c;
-	struct syscall_params param;
 	struct ikc_scd_init_param init;
 	void *dma_buf;
-
-	struct list_head wq_list;
-	ihk_spinlock_t wq_list_lock;
 };
 
+struct mcctrl_per_thread_data {
+	struct list_head hash;
+	struct task_struct *task;
+	void *data;
+};
+
+#define MCCTRL_PER_THREAD_DATA_HASH_SHIFT 8
+#define MCCTRL_PER_THREAD_DATA_HASH_SIZE (1 << MCCTRL_PER_THREAD_DATA_HASH_SHIFT)
+#define MCCTRL_PER_THREAD_DATA_HASH_MASK (MCCTRL_PER_THREAD_DATA_HASH_SIZE - 1) 
+
 struct mcctrl_per_proc_data {
-	struct list_head list;
+	struct mcctrl_usrdata *ud;
+	struct list_head hash;
 	int pid;
 	unsigned long rpgtable;	/* per process, not per OS */
+
+	struct list_head wq_list; /* All these requests come from mcexec */
+	struct list_head wq_req_list; /* These requests come from IKC IRQ handler (can be processed by any threads) */
+	struct list_head wq_list_exact;  /* These requests come from IKC IRQ handler targeting a particular thread */
+
+	ihk_spinlock_t wq_list_lock;
+	wait_queue_head_t wq_prepare;
+	wait_queue_head_t wq_procfs;
+
+	struct list_head per_thread_data_hash[MCCTRL_PER_THREAD_DATA_HASH_SIZE];
+	rwlock_t per_thread_data_hash_lock[MCCTRL_PER_THREAD_DATA_HASH_SIZE];
+	cpumask_t cpu_set;
+	int ikc_target_cpu;
+	atomic_t refcount;
 };
 
 struct sysfsm_req {
@@ -199,11 +252,6 @@ static inline int sysfs_inited(struct sysfsm_data *sdp)
 	return !!(sdp->sysfs_buf);
 } /* sysfs_inited() */
 
-struct cpu_mapping {
-	int cpu_number;
-	int hw_id;
-};
-
 struct cache_topology {
 	struct ihk_cache_topology *saved;
 	cpumask_t shared_cpu_map;
@@ -212,8 +260,9 @@ struct cache_topology {
 };
 
 struct cpu_topology {
-	struct cpu_mapping *cpu_mapping;
+	//struct mcctrl_usrdata *udp;
 	struct ihk_cpu_topology *saved;
+	int mckernel_cpu_id;
 	cpumask_t core_siblings;
 	cpumask_t thread_siblings;
 
@@ -221,41 +270,56 @@ struct cpu_topology {
 	struct list_head cache_list;
 };
 
+#define NODE_DISTANCE_S_SIZE	1024
+
 struct node_topology {
 	struct ihk_node_topology *saved;
+	int mckernel_numa_id;
+	char mckernel_numa_distance_s[NODE_DISTANCE_S_SIZE];
 	cpumask_t cpumap;
 
 	struct list_head chain;
 };
 
+struct mcctrl_part_exec {
+	struct mutex lock;	
+	int nr_processes;
+	int nr_processes_left;
+	cpumask_t cpus_used;
+};
+
 #define CPU_LONGS (((NR_CPUS) + (BITS_PER_LONG) - 1) / (BITS_PER_LONG))
 
+#define MCCTRL_PER_PROC_DATA_HASH_SHIFT 7
+#define MCCTRL_PER_PROC_DATA_HASH_SIZE (1 << MCCTRL_PER_PROC_DATA_HASH_SHIFT)
+#define MCCTRL_PER_PROC_DATA_HASH_MASK (MCCTRL_PER_PROC_DATA_HASH_SIZE - 1)
+
 struct mcctrl_usrdata {
 	struct ihk_ikc_listen_param listen_param;
 	struct ihk_ikc_listen_param listen_param2;
 	ihk_os_t	os;
 	int	num_channels;
-	struct mcctrl_channel	*channels;
-	unsigned long	*mcctrl_doorbell_va;
-	unsigned long	mcctrl_doorbell_pa;
+	struct mcctrl_channel *channels;
 	int	remaining_job;
 	int	base_cpu;
 	int	job_pos;
 	int	mcctrl_dma_abort;
+	struct mutex reserve_lock;
 	unsigned long	last_thread_exec;
-	wait_queue_head_t	wq_prepare;
-	
-	struct list_head per_proc_list;
-	ihk_spinlock_t per_proc_list_lock;
+	wait_queue_head_t wq_procfs;
+	struct list_head per_proc_data_hash[MCCTRL_PER_PROC_DATA_HASH_SIZE];
+	rwlock_t per_proc_data_hash_lock[MCCTRL_PER_PROC_DATA_HASH_SIZE];
+
 	void **keys;
 	struct sysfsm_data sysfsm_data;
 	unsigned long cpu_online[CPU_LONGS];
-	int cpu_mapping_elems;
-	int padding;
-	struct cpu_mapping *cpu_mapping;
-	long cpu_mapping_pa;
+	struct ihk_cpu_info *cpu_info;
+	struct ihk_mem_info *mem_info;
+	nodemask_t numa_online;
 	struct list_head cpu_topology_list;
 	struct list_head node_topology_list;
+	struct mcctrl_part_exec part_exec;
+	int perf_event_num;
 };
 
 struct mcctrl_signal {
@@ -273,14 +337,25 @@ int mcctrl_ikc_is_valid_thread(ihk_os_t os, int cpu);
 ihk_os_t osnum_to_os(int n);
 
 /* syscall.c */
-int init_peer_channel_registry(struct mcctrl_usrdata *ud);
-void destroy_peer_channel_registry(struct mcctrl_usrdata *ud);
-int register_peer_channel(struct mcctrl_usrdata *ud, void *key, struct mcctrl_channel *ch);
-int deregister_peer_channel(struct mcctrl_usrdata *ud, void *key, struct mcctrl_channel *ch);
-struct mcctrl_channel *get_peer_channel(struct mcctrl_usrdata *ud, void *key);
-int __do_in_kernel_syscall(ihk_os_t os, struct mcctrl_channel *c, struct syscall_request *sc);
+int __do_in_kernel_syscall(ihk_os_t os, struct ikc_scd_packet *packet);
+int mcctrl_add_per_proc_data(struct mcctrl_usrdata *ud, int pid, 
+	struct mcctrl_per_proc_data *ppd);
+int mcctrl_delete_per_proc_data(struct mcctrl_usrdata *ud, int pid);
+struct mcctrl_per_proc_data *mcctrl_get_per_proc_data(
+		struct mcctrl_usrdata *ud, int pid);
+void mcctrl_put_per_proc_data(struct mcctrl_per_proc_data *ppd);
 
-#define PROCFS_NAME_MAX 1000
+int mcctrl_add_per_thread_data(struct mcctrl_per_proc_data* ppd,
+	struct task_struct *task, void *data);
+int mcctrl_delete_per_thread_data(struct mcctrl_per_proc_data* ppd,
+	struct task_struct *task);
+inline struct mcctrl_per_thread_data *mcctrl_get_per_thread_data(
+	struct mcctrl_per_proc_data *ppd, struct task_struct *task);
+
+void __return_syscall(ihk_os_t os, struct ikc_scd_packet *packet, 
+		long ret, int stid);
+
+#define PROCFS_NAME_MAX 768
 
 struct procfs_read {
 	unsigned long pbuf;	/* physical address of the host buffer (request) */
@@ -300,7 +375,8 @@ struct procfs_file {
 	char fname[PROCFS_NAME_MAX];	/* procfs filename (request) */
 };
 
-void procfs_answer(unsigned int arg, int err);
+void procfs_answer(struct mcctrl_usrdata *ud, int pid);
+int procfsm_packet_handler(void *os, int msg, int pid, unsigned long arg);
 void add_tid_entry(int osnum, int pid, int tid);
 void add_pid_entry(int osnum, int pid);
 void delete_tid_entry(int osnum, int pid, int tid);
@@ -347,4 +423,14 @@ struct get_cpu_mapping_req {
 	wait_queue_head_t wq;
 };
 
+struct ihk_perf_event_attr{
+	unsigned long config; 
+	unsigned disabled:1;
+	unsigned pinned:1;
+	unsigned exclude_user:1;
+	unsigned exclude_kernel:1;
+	unsigned exclude_hv:1;
+	unsigned exclude_idle:1;
+};
+
 #endif

executer/kernel/mcctrl/mcctrl_public.h

@@ -0,0 +1,20 @@
+#ifndef __MCCTRL_PUBLIC_H
+#define __MCCTRL_PUBLIC_H
+
+#include <ihk/ihk_host_user.h>
+#include <ikc/queue.h>
+
+struct mcctrl_os_cpu_register {
+	unsigned long addr;
+	unsigned long val;
+	unsigned long addr_ext;
+};
+
+int mcctrl_os_read_cpu_register(ihk_os_t os, int cpu,
+		struct mcctrl_os_cpu_register *desc);
+int mcctrl_os_write_cpu_register(ihk_os_t os, int cpu,
+		struct mcctrl_os_cpu_register *desc);
+int mcctrl_get_request_os_cpu(ihk_os_t *os, int *cpu);
+
+
+#endif // __MCCTRL_PUBLIC_H

executer/kernel/mcctrl/procfs.c

@@ -17,8 +17,10 @@
 #include <linux/uaccess.h>
 #include <linux/fs.h>
 #include <linux/resource.h>
+#include <linux/interrupt.h>
 #include "mcctrl.h"
 #include <linux/version.h>
+#include <linux/semaphore.h>
 
 //#define PROCFS_DEBUG
 
@@ -57,7 +59,6 @@ static const struct procfs_entry base_entry_stuff[];
 static const struct file_operations mckernel_forward_ro;
 static const struct file_operations mckernel_forward;
 
-static DECLARE_WAIT_QUEUE_HEAD(procfsq);
 static ssize_t mckernel_procfs_read(struct file *file, char __user *buf, 
 		size_t nbytes, loff_t *ppos);
 
@@ -81,7 +82,7 @@ struct procfs_list_entry {
  * file.
  */
 LIST_HEAD(procfs_file_list);
-static ihk_spinlock_t procfs_file_list_lock;
+DEFINE_SEMAPHORE(procfs_file_list_lock);
 
 static char *
 getpath(struct procfs_list_entry *e, char *buf, int bufsize)
@@ -104,14 +105,28 @@ getpath(struct procfs_list_entry *e, char *buf, int bufsize)
 /**
  * \brief Process SCD_MSG_PROCFS_ANSWER message.
  *
- * \param arg sent argument
- * \param err error info (redundant)
+ * \param ud mcctrl_usrdata pointer
+ * \param pid PID of the requesting process
  */
-void
-procfs_answer(unsigned int arg, int err)
+void procfs_answer(struct mcctrl_usrdata *ud, int pid)
 {
-	dprintk("procfs: received SCD_MSG_PROCFS_ANSWER message(err = %d).\n", err);
-	wake_up_interruptible(&procfsq);
+	struct mcctrl_per_proc_data *ppd = NULL;
+
+	if (pid > 0) {
+		ppd = mcctrl_get_per_proc_data(ud, pid);
+
+		if (unlikely(!ppd)) {
+			kprintf("%s: ERROR: no per-process structure for PID %d\n",
+					__FUNCTION__, pid);
+			return;
+		}
+	}
+
+	wake_up_all(pid > 0 ? &ppd->wq_procfs : &ud->wq_procfs);
+
+	if (pid > 0) {
+		mcctrl_put_per_proc_data(ppd);
+	}
 }
 
 static struct procfs_list_entry *
@@ -246,9 +261,11 @@ get_pid_cred(int pid)
 {
 	struct task_struct *task = NULL;
 
-	if(pid > 0){
+	if (pid > 0) {
+		rcu_read_lock();
 		task = pid_task(find_vpid(pid), PIDTYPE_PID);
-		if(task){
+		rcu_read_unlock();
+		if (task) {
 			return __task_cred(task);
 		}
 	}
@@ -375,67 +392,62 @@ _add_tid_entry(int osnum, int pid, int tid, const struct cred *cred)
 void
 add_tid_entry(int osnum, int pid, int tid)
 {
-	unsigned long irqflag;
 	const struct cred *cred = get_pid_cred(pid);
 
 	if(!cred)
 		return;
-	irqflag = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
+	down(&procfs_file_list_lock);
 	_add_tid_entry(osnum, pid, tid, cred);
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflag);
+	up(&procfs_file_list_lock);
 }
 
 void
 add_pid_entry(int osnum, int pid)
 {
 	struct procfs_list_entry *parent;
-	unsigned long irqflag;
 	const struct cred *cred = get_pid_cred(pid);
 
 	if(!cred)
 		return;
-	irqflag = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
+	down(&procfs_file_list_lock);
 	parent = get_pid_entry(osnum, pid);
 	add_procfs_entries(parent, pid_entry_stuff, cred->uid, cred->gid);
 	_add_tid_entry(osnum, pid, pid, cred);
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflag);
+	up(&procfs_file_list_lock);
 }
 
 void
 delete_tid_entry(int osnum, int pid, int tid)
 {
-	unsigned long irqflag;
 	struct procfs_list_entry *e;
 
-	irqflag = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
+	down(&procfs_file_list_lock);
 	e = find_tid_entry(osnum, pid, tid);
 	if(e)
 		delete_procfs_entries(e);
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflag);
+	up(&procfs_file_list_lock);
 }
 
 void
 delete_pid_entry(int osnum, int pid)
 {
-	unsigned long irqflag;
 	struct procfs_list_entry *e;
 
-	irqflag = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
+	down(&procfs_file_list_lock);
 	e = find_pid_entry(osnum, pid);
 	if(e)
 		delete_procfs_entries(e);
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflag);
+	up(&procfs_file_list_lock);
 }
 
 void
 proc_exe_link(int osnum, int pid, const char *path)
 {
 	struct procfs_list_entry *parent;
-	unsigned long irqflag;
 	kuid_t uid = KUIDT_INIT(0);
 	kgid_t gid = KGIDT_INIT(0);
 
-	irqflag = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
+	down(&procfs_file_list_lock);
 	parent = find_pid_entry(osnum, pid);
 	if(parent){
 		struct procfs_list_entry *task;
@@ -451,7 +463,7 @@ proc_exe_link(int osnum, int pid, const char *path)
 			                 uid, gid, path);
 		}
 	}
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflag);
+	up(&procfs_file_list_lock);
 }
 
 /**
@@ -463,14 +475,13 @@ void
 procfs_init(int osnum)
 {
 	struct procfs_list_entry *parent;
-	unsigned long irqflag;
 	kuid_t uid = KUIDT_INIT(0);
 	kgid_t gid = KGIDT_INIT(0);
 
-	irqflag = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
+	down(&procfs_file_list_lock);
 	parent = get_base_entry(osnum);
 	add_procfs_entries(parent, base_entry_stuff, uid, gid);
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflag);
+	up(&procfs_file_list_lock);
 }
 
 /**
@@ -481,14 +492,14 @@ procfs_init(int osnum)
 void
 procfs_exit(int osnum)
 {
-	unsigned long irqflag;
 	struct procfs_list_entry *e;
 
-	irqflag = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
+	down(&procfs_file_list_lock);
 	e = find_base_entry(osnum);
-	if(e)
+	if (e) {
 		delete_procfs_entries(e);
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflag);
+	}
+	up(&procfs_file_list_lock);
 }
 
 /**
@@ -497,36 +508,84 @@ procfs_exit(int osnum)
  * This function conforms to the 2) way of fs/proc/generic.c
  * from linux-2.6.39.4.
  */
-static ssize_t
-mckernel_procfs_read(struct file *file, char __user *buf, size_t nbytes,
-	       loff_t *ppos)
+static ssize_t __mckernel_procfs_read_write(
+		struct file *file,
+		char __user *buf, size_t nbytes,
+		loff_t *ppos, int read_write)
 {
-	struct inode * inode = file->f_path.dentry->d_inode;
+	struct inode * inode = file->f_inode;
 	char *kern_buffer = NULL;
 	int order = 0;
 	volatile struct procfs_read *r = NULL;
 	struct ikc_scd_packet isp;
-	int ret;
+	int ret, osnum, pid, retw;
 	unsigned long pbuf;
 	unsigned long count = nbytes;
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
 	struct proc_dir_entry *dp = PDE(inode);
 	struct procfs_list_entry *e = dp->data;
-#else	
+#else
 	struct procfs_list_entry *e = PDE_DATA(inode);
-#endif	
+#endif
 	loff_t offset = *ppos;
 	char pathbuf[PROCFS_NAME_MAX];
-	char *path;
+	char *path, *p;
+	ihk_os_t os = NULL;
+	struct mcctrl_usrdata *udp = NULL;
+	struct mcctrl_per_proc_data *ppd = NULL;
 
-	path = getpath(e, pathbuf, 256);
-	dprintk("mckernel_procfs_read: invoked for %s, offset: %lu, count: %d\n", 
-			path, offset, count); 
-	
 	if (count <= 0 || offset < 0) {
 		return 0;
 	}
-	
+
+	path = getpath(e, pathbuf, PROCFS_NAME_MAX);
+	dprintk("%s: invoked for %s, offset: %lu, count: %lu\n",
+			__FUNCTION__, path,
+			(unsigned long)offset, count);
+
+	/* Verify OS number */
+	ret = sscanf(path, "mcos%d/", &osnum);
+	if (ret != 1) {
+		printk("%s: error: couldn't determine OS number\n", __FUNCTION__);
+		return -EINVAL;
+	}
+
+	if (osnum != e->osnum) {
+		printk("%s: error: OS numbers don't match\n", __FUNCTION__);
+		return -EINVAL;
+	}
+
+	/* Is this request for a specific process? */
+	p = strchr(path, '/') + 1;
+	ret = sscanf(p, "%d/", &pid);
+	if (ret != 1) {
+		pid = -1;
+	}
+
+	os = osnum_to_os(osnum);
+	if (!os) {
+		printk("%s: error: no IHK OS data found for OS %d\n",
+				__FUNCTION__, osnum);
+		return -EINVAL;
+	}
+
+	udp = ihk_host_os_get_usrdata(os);
+	if (!udp) {
+		printk("%s: error: no MCCTRL data found for OS %d\n",
+				__FUNCTION__, osnum);
+		return -EINVAL;
+	}
+
+	if (pid > 0) {
+		ppd = mcctrl_get_per_proc_data(udp, pid);
+
+		if (unlikely(!ppd)) {
+			printk("%s: error: no per-process structure for PID %d",
+					__FUNCTION__, pid);
+			return -EINVAL;
+		}
+	}
+
 	while ((1 << order) < count) ++order;
 	if (order > 12) {
 		order -= 12;
@@ -538,10 +597,11 @@ mckernel_procfs_read(struct file *file, char __user *buf, size_t nbytes,
 	/* NOTE: we need physically contigous memory to pass through IKC */
 	kern_buffer = (char *)__get_free_pages(GFP_KERNEL, order);
 	if (!kern_buffer) {
-		printk("mckernel_procfs_read(): ERROR: allocating kernel buffer\n");
-		return -ENOMEM;
+		printk("%s: ERROR: allocating kernel buffer\n", __FUNCTION__);
+		ret = -ENOMEM;
+		goto out;
 	}
-	
+
 	pbuf = virt_to_phys(kern_buffer);
 
 	r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
@@ -555,152 +615,96 @@ mckernel_procfs_read(struct file *file, char __user *buf, size_t nbytes,
 	r->status = 0;
 	r->offset = offset;
 	r->count = count;
-	r->readwrite = 0;
+	r->readwrite = read_write;
 	strncpy((char *)r->fname, path, PROCFS_NAME_MAX);
 	isp.msg = SCD_MSG_PROCFS_REQUEST;
 	isp.ref = 0;
 	isp.arg = virt_to_phys(r);
-	
-	ret = mcctrl_ikc_send(osnum_to_os(e->osnum), 0, &isp);
-	
+	isp.pid = pid;
+
+	ret = mcctrl_ikc_send(osnum_to_os(e->osnum),
+			(pid > 0) ? ppd->ikc_target_cpu : 0, &isp);
+
 	if (ret < 0) {
 		goto out; /* error */
 	}
-	
+
 	/* Wait for a reply. */
 	ret = -EIO; /* default exit code */
-	dprintk("now wait for a relpy\n");
-	
-	/* Wait for the status field of the procfs_read structure set ready. */
-	if (wait_event_interruptible_timeout(procfsq, r->status != 0, HZ) == 0) {
-		kprintf("ERROR: mckernel_procfs_read: timeout (1 sec).\n");
+	dprintk("%s: waiting for reply\n", __FUNCTION__);
+
+retry_wait:
+	/* Wait for the status field of the procfs_read structure,
+	 * wait on per-process or OS specific data depending on
+	 * who the request is for.
+	 */
+	if (pid > 0) {
+		retw = wait_event_interruptible_timeout(ppd->wq_procfs,
+				r->status != 0, HZ);
+	}
+	else {
+		retw = wait_event_interruptible_timeout(udp->wq_procfs,
+				r->status != 0, HZ);
+	}
+
+	/* Timeout? */
+	if (retw == 0 && r->status == 0) {
+		printk("%s: error: timeout (1 sec)\n", __FUNCTION__);
 		goto out;
 	}
-	
-	/* Wake up and check the result. */
-	dprintk("mckernel_procfs_read: woke up. ret: %d, eof: %d\n", r->ret, r->eof);
-	
-	if (r->ret > 0) {
-		if (copy_to_user(buf, kern_buffer, r->ret)) {
-			kprintf("ERROR: mckernel_procfs_read: copy_to_user failed.\n");
-			ret = -EFAULT;
-			goto out;
-		}
+	/* Interrupted? */
+	else if (retw == -ERESTARTSYS) {
+		ret = -ERESTART;
+		goto out;
+	}
+	/* Were we woken up by a reply to another procfs request? */
+	else if (r->status == 0) {
+		/* TODO: r->status is not set atomically, we could be woken
+		 * up with status == 0 and it could change to 1 while in this
+		 * code, we could potentially miss the wake_up()... 
+		 */
+		printk("%s: stale wake-up, retrying\n", __FUNCTION__);
+		goto retry_wait;
+	}
 
+	/* Wake up and check the result. */
+	dprintk("%s: woke up. ret: %d, eof: %d\n",
+			__FUNCTION__, r->ret, r->eof);
+
+	if (r->ret > 0) {
+		if (read_write == 0) {
+			if (copy_to_user(buf, kern_buffer, r->ret)) {
+				printk("%s: ERROR: copy_to_user failed.\n", __FUNCTION__);
+				ret = -EFAULT;
+				goto out;
+			}
+		}
 		*ppos += r->ret;
 	}
 	ret = r->ret;
 
 out:
-	if(kern_buffer)
+	if (ppd)
+		mcctrl_put_per_proc_data(ppd);
+	if (kern_buffer)
 		free_pages((uintptr_t)kern_buffer, order);
-	if(r)
+	if (r)
 		kfree((void *)r);
-	
+
 	return ret;
 }
 
-static ssize_t
-mckernel_procfs_write(struct file *file, const char __user *buf, size_t nbytes,
-	       loff_t *ppos)
+static ssize_t mckernel_procfs_read(struct file *file,
+		char __user *buf, size_t nbytes, loff_t *ppos)
 {
-	struct inode * inode = file->f_path.dentry->d_inode;
-	char *kern_buffer = NULL;
-	int order = 0;
-	volatile struct procfs_read *r = NULL;
-	struct ikc_scd_packet isp;
-	int ret;
-	unsigned long pbuf;
-	unsigned long count = nbytes;
-#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
-	struct proc_dir_entry *dp = PDE(inode);
-	struct procfs_list_entry *e = dp->data;
-#else	
-	struct procfs_list_entry *e = PDE_DATA(inode);
-#endif	
-	loff_t offset = *ppos;
-	char pathbuf[PROCFS_NAME_MAX];
-	char *path;
+	return __mckernel_procfs_read_write(file, buf, nbytes, ppos, 0);
+}
 
-	path = getpath(e, pathbuf, 256);
-	dprintk("mckernel_procfs_read: invoked for %s, offset: %lu, count: %d\n", 
-			path, offset, count); 
-	
-	if (count <= 0 || offset < 0) {
-		return 0;
-	}
-	
-	while ((1 << order) < count) ++order;
-	if (order > 12) {
-		order -= 12;
-	}
-	else {
-		order = 1;
-	}
-
-	/* NOTE: we need physically contigous memory to pass through IKC */
-	kern_buffer = (char *)__get_free_pages(GFP_KERNEL, order);
-	if (!kern_buffer) {
-		printk("mckernel_procfs_read(): ERROR: allocating kernel buffer\n");
-		return -ENOMEM;
-	}
-	if (copy_from_user(kern_buffer, buf, nbytes)) {
-		ret = -EFAULT;
-		goto out;
-	}
-	
-	pbuf = virt_to_phys(kern_buffer);
-
-	r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
-	if (r == NULL) {
-		ret = -ENOMEM;
-		goto out;
-	}
-	dprintk("offset: %lx, count: %d, cpu: %d\n", offset, count, e->cpu);
-
-	r->pbuf = pbuf;
-	r->eof = 0;
-	r->ret = -EIO; /* default */
-	r->status = 0;
-	r->offset = offset;
-	r->count = count;
-	r->readwrite = 1;
-	strncpy((char *)r->fname, path, PROCFS_NAME_MAX);
-	isp.msg = SCD_MSG_PROCFS_REQUEST;
-	isp.ref = 0;
-	isp.arg = virt_to_phys(r);
-	
-	ret = mcctrl_ikc_send(osnum_to_os(e->osnum), 0, &isp);
-	
-	if (ret < 0) {
-		goto out; /* error */
-	}
-	
-	/* Wait for a reply. */
-	ret = -EIO; /* default exit code */
-	dprintk("now wait for a relpy\n");
-	
-	/* Wait for the status field of the procfs_read structure set ready. */
-	if (wait_event_interruptible_timeout(procfsq, r->status != 0, HZ) == 0) {
-		kprintf("ERROR: mckernel_procfs_read: timeout (1 sec).\n");
-		goto out;
-	}
-	
-	/* Wake up and check the result. */
-	dprintk("mckernel_procfs_read: woke up. ret: %d, eof: %d\n", r->ret, r->eof);
-	
-	if (r->ret > 0) {
-		*ppos += r->ret;
-	}
-	ret = r->ret;
-
-out:
-	if(kern_buffer)
-		free_pages((uintptr_t)kern_buffer, order);
-	if(r)
-		kfree((void *)r);
-	
-	return ret;
+static ssize_t mckernel_procfs_write(struct file *file,
+		const char __user *buf, size_t nbytes, loff_t *ppos)
+{
+	return __mckernel_procfs_read_write(file,
+			(char __user *)buf, nbytes, ppos, 1);
 }
 
 static loff_t
@@ -719,6 +723,57 @@ mckernel_procfs_lseek(struct file *file, loff_t offset, int orig)
 	return file->f_pos;
 }
 
+struct procfs_work {
+	void *os;
+	int msg;
+	int pid;
+	unsigned long arg;
+	struct work_struct work;
+};
+
+static void procfsm_work_main(struct work_struct *work0)
+{
+	struct procfs_work *work = container_of(work0, struct procfs_work, work);
+
+	switch (work->msg) {
+		case SCD_MSG_PROCFS_TID_CREATE:
+			add_tid_entry(ihk_host_os_get_index(work->os), work->pid, work->arg);
+			break;
+
+		case SCD_MSG_PROCFS_TID_DELETE:
+			delete_tid_entry(ihk_host_os_get_index(work->os), work->pid, work->arg);
+			break;
+
+		default:
+			printk("%s: unknown work: msg: %d, pid: %d, arg: %lu)\n",
+					__FUNCTION__, work->msg, work->pid, work->arg);
+			break;
+	}
+
+	kfree(work);
+	return;
+}
+
+int procfsm_packet_handler(void *os, int msg, int pid, unsigned long arg)
+{
+	struct procfs_work *work = NULL;
+
+	work = kzalloc(sizeof(*work), GFP_ATOMIC);
+	if (!work) {
+		printk("%s: kzalloc failed\n", __FUNCTION__);
+		return -1;
+	}
+
+	work->os = os;
+	work->msg = msg;
+	work->pid = pid;
+	work->arg = arg;
+	INIT_WORK(&work->work, &procfsm_work_main);
+
+	schedule_work(&work->work);
+	return 0;
+}
+
 static const struct file_operations mckernel_forward_ro = {
 	.llseek		= mckernel_procfs_lseek,
 	.read		= mckernel_procfs_read,

executer/kernel/mcctrl/sysfs.c

@@ -14,10 +14,11 @@
 #include <linux/slab.h>
 #include <linux/device.h>
 #include <linux/version.h>
+#include <linux/interrupt.h>
 #include "mcctrl.h"
 #include "sysfs_msg.h"
 
-#define dprintk(...) do { if (0) printk(KERN_DEBUG __VA_ARGS__); } while (0)
+#define dprintk(...) do { if (0) printk(__VA_ARGS__); } while (0)
 #define wprintk(...) do { if (1) printk(KERN_WARNING __VA_ARGS__); } while (0)
 #define eprintk(...) do { if (1) printk(KERN_ERR __VA_ARGS__); } while (0)
 
@@ -1231,9 +1232,16 @@ sysfsm_cleanup(ihk_os_t os)
 	int error;
 	ihk_device_t dev = ihk_os_to_dev(os);
 	struct mcctrl_usrdata *udp = ihk_host_os_get_usrdata(os);
-	struct sysfsm_data *sdp = &udp->sysfsm_data;
+	struct sysfsm_data *sdp;
 	struct sysfsm_node *np;
 
+	if (!udp) {
+		printk("%s: WARNING: no mcctrl_usrdata found\n", __FUNCTION__);
+		return;
+	}
+
+	sdp = &udp->sysfsm_data;
+
 	dprintk("mcctrl:sysfsm_cleanup(%p)\n", os);
 
 	if (sdp->sysfs_buf) {
@@ -2094,9 +2102,16 @@ struct sysfsm_ops snooping_local_ops_s = {
 /**** local list ****/
 static ssize_t snooping_local_show_pbl(struct sysfsm_ops *ops, void *instance, void *buf, size_t bufsize)
 {
+	size_t ret;
 	const struct sysfsm_bitmap_param *p = instance;
 
-	return bitmap_scnlistprintf(buf, bufsize, p->ptr, p->nbits);
+	ret = bitmap_scnlistprintf(buf, bufsize, p->ptr, p->nbits);
+	if (ret < bufsize - 1) {
+		sprintf(buf + ret, "\n");
+		return ret + 1;
+	}
+
+	return 0;
 } /* snooping_local_show_pbl() */
 
 struct sysfsm_ops snooping_local_ops_pbl = {
@@ -2107,9 +2122,16 @@ struct sysfsm_ops snooping_local_ops_pbl = {
 /**** local map ****/
 static ssize_t snooping_local_show_pb(struct sysfsm_ops *ops, void *instance, void *buf, size_t bufsize)
 {
+	size_t ret;
 	const struct sysfsm_bitmap_param *p = instance;
 
-	return bitmap_scnprintf(buf, bufsize, p->ptr, p->nbits);
+	ret = bitmap_scnprintf(buf, bufsize, p->ptr, p->nbits);
+	if (ret < bufsize - 1) {
+		sprintf(buf + ret, "\n");
+		return ret + 1;
+	}
+
+	return 0;
 } /* snooping_local_show_pb() */
 
 struct sysfsm_ops snooping_local_ops_pb = {

executer/kernel/mcctrl/sysfs_files.c

@@ -18,7 +18,7 @@
 #include "mcctrl.h"
 #include "sysfs_msg.h"
 
-#define dprintk(...) do { if (0) printk(KERN_DEBUG __VA_ARGS__); } while (0)
+#define dprintk(...) do { if (0) printk(__VA_ARGS__); } while (0)
 #define wprintk(...) do { if (1) printk(KERN_WARNING __VA_ARGS__); } while (0)
 #define eprintk(...) do { if (1) printk(KERN_ERR __VA_ARGS__); } while (0)
 
@@ -92,27 +92,19 @@ void setup_local_snooping_samples(ihk_os_t os)
 
 void setup_local_snooping_files(ihk_os_t os)
 {
-	struct ihk_cpu_info *info;
 	struct mcctrl_usrdata *udp = ihk_host_os_get_usrdata(os);
 	struct sysfsm_bitmap_param param;
 	static unsigned long cpu_offline = 0x0;
 	int i;
 	int error;
 
-	info = ihk_os_get_cpu_info(os);
-	if (!info) {
-		eprintk("mcctrl:ihk_os_get_cpu_info failed.\n");
-		return;
-	}
-
 	memset(udp->cpu_online, 0, sizeof(udp->cpu_online));
-	for (i = 0; i < info->n_cpus; i++) {
-		udp->cpu_online[i / BITS_PER_LONG] = 
-			udp->cpu_online[i / BITS_PER_LONG] | (1 << (i % BITS_PER_LONG));
+	for (i = 0; i < udp->cpu_info->n_cpus; i++) {
+		set_bit(i, udp->cpu_online);
 	}
 
 	param.nbits = CPU_LONGS * BITS_PER_LONG;
-	param.ptr = udp->cpu_online;
+	param.ptr = &udp->cpu_online;
 	dprintk("mcctrl:setup_local_snooping_files: CPU_LONGS=%d, BITS_PER_LONG=%d\n", 
 		CPU_LONGS, BITS_PER_LONG);
 
@@ -187,141 +179,122 @@ static void free_cpu_topology(struct mcctrl_usrdata *udp)
 	return;
 } /* free_cpu_topology() */
 
-static void free_cpu_mapping(struct mcctrl_usrdata *udp)
-{
-	ihk_device_t dev = ihk_os_to_dev(udp->os);
-	size_t size;
-
-	size = udp->cpu_mapping_elems * sizeof(struct cpu_mapping);
-	ihk_device_unmap_virtual(dev, udp->cpu_mapping, size);
-	ihk_device_unmap_memory(dev, udp->cpu_mapping_pa, size);
-
-	return;
-} /* free_cpu_mapping() */
-
 void free_topology_info(ihk_os_t os)
 {
 	struct mcctrl_usrdata *udp = ihk_host_os_get_usrdata(os);
 
+	if (!udp) {
+		printk("%s: WARNING: no mcctrl_usrdata found\n", __FUNCTION__);
+		return;
+	}
+
 	free_node_topology(udp);
 	free_cpu_topology(udp);
-	free_cpu_mapping(udp);
 
 	return;
 } /* free_topology_info() */
 
-void reply_get_cpu_mapping(long req_pa)
+/*
+ * CPU and NUMA node mapping conversion functions.
+ */
+int mckernel_cpu_2_linux_cpu(struct mcctrl_usrdata *udp, int cpu_id)
 {
-	struct get_cpu_mapping_req *req = phys_to_virt(req_pa);
+	return (cpu_id < udp->cpu_info->n_cpus) ?
+		udp->cpu_info->mapping[cpu_id] : -1;
+}
 
-	req->busy = 0;
-	wake_up(&req->wq);
-
-	return;
-} /* reply_get_cpu_mapping() */
-
-static int get_cpu_mapping(struct mcctrl_usrdata *udp)
+int mckernel_cpu_2_hw_id(struct mcctrl_usrdata *udp, int cpu_id)
 {
-	int error;
-	ihk_device_t dev = ihk_os_to_dev(udp->os);
-	struct get_cpu_mapping_req *req = NULL;
-	struct ikc_scd_packet packet;
-	size_t size;
+	return (cpu_id < udp->cpu_info->n_cpus) ?
+		udp->cpu_info->hw_ids[cpu_id] : -1;
+}
 
-	dprintk("get_cpu_mapping(%p)\n", udp);
-
-	req = kmalloc(sizeof(*req), GFP_KERNEL);
-	if (!req) {
-		error = -ENOMEM;
-		eprintk("mcctrl:get_cpu_mapping:kmalloc failed. %d\n", error);
-		goto out;
-	}
-
-	req->busy = 1;
-	req->error = -1;
-	init_waitqueue_head(&req->wq);
-
-	packet.msg = SCD_MSG_GET_CPU_MAPPING;
-	packet.arg = virt_to_phys(req);
-
-#define GET_CPU_MAPPING_CPU 0
-	error = mcctrl_ikc_send(udp->os, GET_CPU_MAPPING_CPU, &packet);
-	if (error) {
-		eprintk("mcctrl:get_cpu_mapping:"
-				"mcctrl_ikc_send failed. %d\n", error);
-		goto out;
-	}
-
-	error = wait_event_interruptible(req->wq, !req->busy);
-	if (error) {
-		eprintk("mcctrl:get_cpu_mapping:"
-				"wait_event_interruptible failed. %d\n", error);
-		req = NULL;	/* XXX */
-		goto out;
-	}
-
-	if (req->error) {
-		error = req->error;
-		eprintk("mcctrl:get_cpu_mapping:"
-				"SCD_MSG_GET_CPU_MAPPING failed. %d\n", error);
-		goto out;
-	}
-
-	size = req->buf_elems * sizeof(struct cpu_mapping);
-	udp->cpu_mapping_elems = req->buf_elems;
-	udp->cpu_mapping_pa = ihk_device_map_memory(dev, req->buf_rpa, size);
-	udp->cpu_mapping = ihk_device_map_virtual(
-			dev, udp->cpu_mapping_pa, size, NULL, 0);
-
-	error = 0;
-out:
-	dprintk("get_cpu_mapping(%p): %d\n", udp, error);
-	kfree(req);
-	return error;
-} /* get_cpu_mapping() */
-
-static int hwid_to_cpu(struct mcctrl_usrdata *udp, int hw_id)
+int linux_cpu_2_mckernel_cpu(struct mcctrl_usrdata *udp, int cpu_id)
 {
 	int i;
 
-	for (i = 0; i < udp->cpu_mapping_elems; ++i) {
-		if (udp->cpu_mapping[i].hw_id == hw_id) {
-			return udp->cpu_mapping[i].cpu_number;
+	for (i = 0; i < udp->cpu_info->n_cpus; ++i) {
+		if (udp->cpu_info->mapping[i] == cpu_id)
+			return i;
+	}
+
+	return -1;
+}
+
+#if 0
+int hw_id_2_mckernel_cpu(struct mcctrl_usrdata *udp, int hw_id)
+{
+	int i;
+
+	for (i = 0; i < udp->cpu_info->n_cpus; ++i) {
+		if (udp->cpu_info->hw_ids[i] == hw_id) {
+			return i;
 		}
 	}
 
 	return -1;
 }
 
+int hw_id_2_linux_cpu(struct mcctrl_usrdata *udp, int hw_id)
+{
+	int i;
+
+	for (i = 0; i < udp->cpu_info->n_cpus; ++i) {
+		if (udp->cpu_info->hw_ids[i] == hw_id) {
+			return mckernel_cpu_2_linux_cpu(udp, i);
+		}
+	}
+
+	return -1;
+}
+
+int linux_cpu_2_hw_id(struct mcctrl_usrdata *udp, int cpu)
+{
+	int mckernel_cpu = linux_cpu_2_mckernel_cpu(udp, cpu);
+
+	return (mckernel_cpu >= 0 && mckernel_cpu < udp->cpu_info->n_cpus) ?
+		udp->cpu_info->hw_ids[mckernel_cpu] : -1;
+}
+#endif
+
+int mckernel_numa_2_linux_numa(struct mcctrl_usrdata *udp, int numa_id)
+{
+	return (numa_id < udp->mem_info->n_numa_nodes) ?
+		udp->mem_info->numa_mapping[numa_id] : -1;
+}
+
+int linux_numa_2_mckernel_numa(struct mcctrl_usrdata *udp, int numa_id)
+{
+	int i;
+
+	for (i = 0; i < udp->mem_info->n_numa_nodes; ++i) {
+		if (udp->mem_info->numa_mapping[i] == numa_id)
+			return i;
+	}
+
+	return -1;
+}
+
+
+
 static int translate_cpumap(struct mcctrl_usrdata *udp,
 		cpumask_t *linmap, cpumask_t *mckmap)
 {
 	int error;
-	ihk_device_t dev = ihk_os_to_dev(udp->os);
 	int lincpu;
-	int hw_id;
 	int mckcpu;
 
 	dprintk("translate_cpumap(%p,%p,%p)\n", udp, linmap, mckmap);
 	cpumask_clear(mckmap);
 	for_each_cpu(lincpu, linmap) {
-		hw_id = ihk_device_linux_cpu_to_hw_id(dev, lincpu);
-		if (hw_id < 0) {
-			error = hw_id;
-			eprintk("mcctrl:translate_cpumap:"
-					"ihk_device_linux_cpu_to_hw_id failed."
-					" %d\n", error);
-			goto out;
-		}
+		mckcpu = linux_cpu_2_mckernel_cpu(udp, lincpu);
 
-		mckcpu = hwid_to_cpu(udp, hw_id);
 		if (mckcpu >= 0) {
 			cpumask_set_cpu(mckcpu, mckmap);
 		}
 	}
 
 	error = 0;
-out:
 	dprintk("translate_cpumap(%p,%p,%p): %d\n", udp, linmap, mckmap, error);
 	return error;
 } /* translate_cpumap() */
@@ -361,7 +334,7 @@ out:
 	return (error)? ERR_PTR(error): topo;
 } /* get_cache_topology() */
 
-static struct cpu_topology *get_cpu_topology_one(struct mcctrl_usrdata *udp,
+static struct cpu_topology *get_one_cpu_topology(struct mcctrl_usrdata *udp,
 		int index)
 {
 	int error;
@@ -370,41 +343,43 @@ static struct cpu_topology *get_cpu_topology_one(struct mcctrl_usrdata *udp,
 	struct cache_topology *cache;
 	struct ihk_cache_topology *saved_cache;
 
-	dprintk("get_cpu_topology_one(%p,%d)\n", udp, index);
+	dprintk("get_one_cpu_topology(%p,%d)\n", udp, index);
 	topology = kmalloc(sizeof(*topology), GFP_KERNEL);
 	if (!topology) {
 		error = -ENOMEM;
-		eprintk("mcctrl:get_cpu_topology_one:"
+		eprintk("mcctrl:get_one_cpu_topology:"
 				"kmalloc failed. %d\n", error);
 		goto out;
 	}
 
 	INIT_LIST_HEAD(&topology->cache_list);
-	topology->cpu_mapping = &udp->cpu_mapping[index];
+	topology->mckernel_cpu_id = index;
+	topology->saved = ihk_device_get_cpu_topology(dev, 
+			mckernel_cpu_2_hw_id(udp, index));
 
-	topology->saved = ihk_device_get_cpu_topology(
-			dev, topology->cpu_mapping->hw_id);
 	if (IS_ERR(topology->saved)) {
 		error = PTR_ERR(topology->saved);
-		eprintk("mcctrl:get_cpu_topology_one:"
+		eprintk("mcctrl:get_one_cpu_topology:"
 				"ihk_device_get_cpu_topology failed. %d\n",
 				error);
 		goto out;
 	}
 
-	error = translate_cpumap(udp, &topology->saved->core_siblings,
+	error = translate_cpumap(udp, 
+			&topology->saved->core_siblings,
 			&topology->core_siblings);
 	if (error) {
-		eprintk("mcctrl:get_cpu_topology_one:"
+		eprintk("mcctrl:get_one_cpu_topology:"
 				"translate_cpumap(core_siblings) failed."
 				" %d\n", error);
 		goto out;
 	}
 
-	error = translate_cpumap(udp, &topology->saved->thread_siblings,
+	error = translate_cpumap(udp, 
+			&topology->saved->thread_siblings,
 			&topology->thread_siblings);
 	if (error) {
-		eprintk("mcctrl:get_cpu_topology_one:"
+		eprintk("mcctrl:get_one_cpu_topology:"
 				"translate_cpumap(thread_siblings) failed."
 				" %d\n", error);
 		goto out;
@@ -415,7 +390,7 @@ static struct cpu_topology *get_cpu_topology_one(struct mcctrl_usrdata *udp,
 		cache = get_cache_topology(udp, topology, saved_cache);
 		if (IS_ERR(cache)) {
 			error = PTR_ERR(cache);
-			eprintk("mcctrl:get_cpu_topology_one:"
+			eprintk("mcctrl:get_one_cpu_topology:"
 					"get_cache_topology failed. %d\n",
 					error);
 			goto out;
@@ -429,10 +404,10 @@ out:
 	if (error && !IS_ERR_OR_NULL(topology)) {
 		free_cpu_topology_one(udp, topology);
 	}
-	dprintk("get_cpu_topology_one(%p,%d): %d %p\n",
+	dprintk("get_one_cpu_topology(%p,%d): %d %p\n",
 			udp, index, error, topology);
 	return (error)? ERR_PTR(error): topology;
-} /* get_cpu_topology_one() */
+} /* get_one_cpu_topology() */
 
 static int get_cpu_topology(struct mcctrl_usrdata *udp)
 {
@@ -441,12 +416,12 @@ static int get_cpu_topology(struct mcctrl_usrdata *udp)
 	struct cpu_topology *topology;
 
 	dprintk("get_cpu_topology(%p)\n", udp);
-	for (index = 0; index < udp->cpu_mapping_elems; ++index) {
-		topology = get_cpu_topology_one(udp, index);
+	for (index = 0; index < udp->cpu_info->n_cpus; ++index) {
+		topology = get_one_cpu_topology(udp, index);
 		if (IS_ERR(topology)) {
 			error = PTR_ERR(topology);
-			eprintk("mcctrl:get_cpu_topology:"
-					"get_cpu_topology_one failed. %d\n",
+			eprintk("mcctrl:get_cpu_topology: "
+					"get_one_cpu_topology failed. %d\n",
 					error);
 			goto out;
 		}
@@ -460,15 +435,15 @@ out:
 	return error;
 } /* get_cpu_topology() */
 
-static void setup_one_cache_files(struct mcctrl_usrdata *udp,
+static void setup_cpu_sysfs_cache_files(struct mcctrl_usrdata *udp,
 		struct cpu_topology *cpu, struct cache_topology *cache)
 {
 	char *prefix = "/sys/devices/system/cpu";
-	int cpu_number = cpu->cpu_mapping->cpu_number;
+	int cpu_number = cpu->mckernel_cpu_id;
 	int index = cache->saved->index;
 	struct sysfsm_bitmap_param param;
 
-	dprintk("setup_one_cache_files(%p,%p,%p)\n", udp, cpu, cache);
+	dprintk("setup_cpu_sysfs_cache_files(%p,%p,%p)\n", udp, cpu, cache);
 
 	sysfsm_createf(udp->os, SYSFS_SNOOPING_OPS_d64,
 			&cache->saved->level, 0444,
@@ -509,19 +484,19 @@ static void setup_one_cache_files(struct mcctrl_usrdata *udp,
 			"%s/cpu%d/cache/index%d/shared_cpu_list",
 			prefix, cpu_number, index);
 
-	dprintk("setup_one_cache_files(%p,%p,%p):\n", udp, cpu, cache);
+	dprintk("setup_cpu_sysfs_cache_files(%p,%p,%p):\n", udp, cpu, cache);
 	return;
-} /* setup_one_cache_files() */
+} /* setup_cpu_sysfs_cache_files() */
 
-static void setup_one_cpu_files(struct mcctrl_usrdata *udp,
+static void setup_cpu_sysfs_files(struct mcctrl_usrdata *udp,
 		struct cpu_topology *cpu)
 {
 	char *prefix = "/sys/devices/system/cpu";
-	int cpu_number = cpu->cpu_mapping->cpu_number;
+	int cpu_number = cpu->mckernel_cpu_id;
 	struct sysfsm_bitmap_param param;
 	struct cache_topology *cache;
 
-	dprintk("setup_one_cpu_files(%p,%p)\n", udp, cpu);
+	dprintk("setup_cpu_sysfs_files(%p,%p)\n", udp, cpu);
 
 	sysfsm_createf(udp->os, SYSFS_SNOOPING_OPS_d32,
 			&cpu->saved->physical_package_id, 0444,
@@ -553,41 +528,61 @@ static void setup_one_cpu_files(struct mcctrl_usrdata *udp,
 			prefix, cpu_number);
 
 	list_for_each_entry(cache, &cpu->cache_list, chain) {
-		setup_one_cache_files(udp, cpu, cache);
+		setup_cpu_sysfs_cache_files(udp, cpu, cache);
 	}
 
-	dprintk("setup_one_cpu_files(%p,%p):\n", udp, cpu);
+	dprintk("setup_cpu_sysfs_files(%p,%p):\n", udp, cpu);
 	return;
-} /* setup_one_cpu_files() */
+} /* setup_cpu_sysfs_files() */
 
-static void setup_cpu_files(struct mcctrl_usrdata *udp)
+static void setup_cpus_sysfs_files_node_link(struct mcctrl_usrdata *udp)
+{
+	int error;
+	int cpu;
+	struct sysfs_handle handle;
+
+	for (cpu = 0; cpu < udp->cpu_info->n_cpus; ++cpu) {
+		int node = linux_numa_2_mckernel_numa(udp,
+				cpu_to_node(mckernel_cpu_2_linux_cpu(udp, cpu)));
+
+		error = sysfsm_lookupf(udp->os, &handle,
+				"/sys/devices/system/node/node%d", node);
+		if (error) {
+			panic("sysfsm_lookupf: node for CPU");
+		}
+
+		error = sysfsm_symlinkf(udp->os, handle,
+				"/sys/devices/system/cpu/cpu%d/node%d",
+				cpu, node);
+		if (error) {
+			panic("sysfsm_symlinkf(CPU in node)");
+		}
+	}
+
+	error = 0;
+	return;
+}
+
+static void setup_cpus_sysfs_files(struct mcctrl_usrdata *udp)
 {
 	int error;
 	struct cpu_topology *cpu;
 
-	dprintk("setup_cpu_file(%p)\n", udp);
-	error = get_cpu_mapping(udp);
-	if (error) {
-		eprintk("mcctrl:setup_cpu_files:"
-				"get_cpu_mapping failed. %d\n", error);
-		goto out;
-	}
-
 	error = get_cpu_topology(udp);
 	if (error) {
-		eprintk("mcctrl:setup_cpu_files:"
+		eprintk("mcctrl:setup_cpus_sysfs_files:"
 				"get_cpu_topology failed. %d\n", error);
 		goto out;
 	}
 
 	list_for_each_entry(cpu, &udp->cpu_topology_list, chain) {
-		setup_one_cpu_files(udp, cpu);
+		setup_cpu_sysfs_files(udp, cpu);
 	}
 	error = 0;
 out:
 	dprintk("setup_cpu_file(%p):\n", udp);
 	return;
-} /* setup_cpu_files() */
+} /* setup_cpus_sysfs_files() */
 
 static struct node_topology *get_one_node_topology(struct mcctrl_usrdata *udp,
 		struct ihk_node_topology *saved)
@@ -629,8 +624,10 @@ static int get_node_topology(struct mcctrl_usrdata *udp)
 	struct node_topology *topology;
 
 	dprintk("get_node_topology(%p)\n", udp);
-	for (node = 0; ; ++node) {
-		saved = ihk_device_get_node_topology(dev, node);
+	for (node = 0; node < udp->mem_info->n_numa_nodes; ++node) {
+		saved = ihk_device_get_node_topology(dev,
+				mckernel_numa_2_linux_numa(udp, node));
+
 		if (IS_ERR(saved)) {
 			break;
 		}
@@ -647,6 +644,8 @@ static int get_node_topology(struct mcctrl_usrdata *udp)
 			goto out;
 		}
 
+		topology->mckernel_numa_id = node;
+
 		list_add(&topology->chain, &udp->node_topology_list);
 	}
 
@@ -659,6 +658,7 @@ out:
 static int setup_node_files(struct mcctrl_usrdata *udp)
 {
 	int error;
+	int node;
 	struct node_topology *p;
 	struct sysfsm_bitmap_param param;
 
@@ -670,16 +670,71 @@ static int setup_node_files(struct mcctrl_usrdata *udp)
 		goto out;
 	}
 
+	memset(&udp->numa_online, 0, sizeof(udp->numa_online));
+	for (node = 0; node < udp->mem_info->n_numa_nodes; ++node) {
+		node_set(node, udp->numa_online);
+	}
+
+	param.nbits = MAX_NUMNODES;
+	param.ptr = &udp->numa_online;
+	sysfsm_createf(udp->os, SYSFS_SNOOPING_OPS_pbl, &param, 0444,
+			"/sys/devices/system/node/online");
+	sysfsm_createf(udp->os, SYSFS_SNOOPING_OPS_pbl, &param, 0444,
+			"/sys/devices/system/node/possible");
+
 	list_for_each_entry(p, &udp->node_topology_list, chain) {
+		struct sysfs_handle handle;
+		int cpu;
+		size_t offset = 0;
 		param.nbits = nr_cpumask_bits;
 		param.ptr = &p->cpumap;
 
+		for (node = 0; node < udp->mem_info->n_numa_nodes; ++node) {
+			if (node > 0) {
+				offset += snprintf(&p->mckernel_numa_distance_s[offset],
+						NODE_DISTANCE_S_SIZE - offset, "%s", " ");
+			}
+			offset += snprintf(&p->mckernel_numa_distance_s[offset],
+					NODE_DISTANCE_S_SIZE - offset, "%d",
+					node_distance(
+						mckernel_numa_2_linux_numa(udp, p->mckernel_numa_id),
+						mckernel_numa_2_linux_numa(udp, node)
+						));
+		}
+
+		sysfsm_createf(udp->os, SYSFS_SNOOPING_OPS_s,
+				p->mckernel_numa_distance_s, 0444,
+				"/sys/devices/system/node/node%d/distance",
+				p->mckernel_numa_id);
+
 		sysfsm_createf(udp->os, SYSFS_SNOOPING_OPS_pb, &param, 0444,
 				"/sys/devices/system/node/node%d/cpumap",
-				p->saved->node_number);
+				p->mckernel_numa_id);
 		sysfsm_createf(udp->os, SYSFS_SNOOPING_OPS_pbl, &param, 0444,
 				"/sys/devices/system/node/node%d/cpulist",
-				p->saved->node_number);
+				p->mckernel_numa_id);
+
+		/* Add CPU symlinks for this node */
+		for (cpu = 0; cpu < udp->cpu_info->n_cpus; ++cpu) {
+			if (linux_numa_2_mckernel_numa(udp,
+						cpu_to_node(mckernel_cpu_2_linux_cpu(udp, cpu)))
+					!= p->mckernel_numa_id) {
+				continue;
+			}
+
+			error = sysfsm_lookupf(udp->os, &handle,
+					"/sys/devices/system/cpu/cpu%d", cpu);
+			if (error) {
+				panic("sysfsm_lookupf(CPU in node)");
+			}
+
+			error = sysfsm_symlinkf(udp->os, handle,
+					"/sys/devices/system/node/node%d/cpu%d",
+					p->mckernel_numa_id, cpu);
+			if (error) {
+				panic("sysfsm_symlinkf(CPU in node)");
+			}
+		}
 	}
 
 	error = 0;
@@ -1026,11 +1081,18 @@ void setup_sysfs_files(ihk_os_t os)
 		panic("sysfsm_unlinkf");
 	}
 
-	setup_local_snooping_samples(os);
+	//setup_local_snooping_samples(os);
 	setup_local_snooping_files(os);
-	setup_cpu_files(udp);
+	setup_cpus_sysfs_files(udp);
 	setup_node_files(udp);
-	setup_pci_files(udp);
+	setup_cpus_sysfs_files_node_link(udp);
+	//setup_pci_files(udp);
+
+	/* Indicate sysfs files setup completion for boot script */
+	error = sysfsm_mkdirf(os, NULL, "/sys/setup_complete");
+	if (error) {
+		panic("sysfsm_mkdir(complete)");
+	}
 
 	return;
 } /* setup_files() */

executer/kernel/mcoverlayfs/Makefile.in

@@ -1,39 +1,45 @@
-KDIR ?= @KDIR@
-ARCH ?= @ARCH@
-KMODDIR=@KMODDIR@
-src = @abs_srcdir@
 ENABLE_MCOVERLAYFS=@ENABLE_MCOVERLAYFS@
+RELEASE=@UNAME_R@
 
-RELEASE=$(shell uname -r)
 MAJOR=$(shell echo ${RELEASE} | sed -e 's/^\([0-9]*\).*/\1/')
 MINOR=$(shell echo ${RELEASE} | sed -e 's/^[0-9]*.\([0-9]*\).*/\1/')
 PATCH=$(shell echo ${RELEASE} | sed -e 's/^[0-9]*.[0-9]*.\([0-9]*\).*/\1/')
 LINUX_VERSION_CODE=$(shell expr \( ${MAJOR} \* 65536 \) + \( ${MINOR} \* 256 \) + ${PATCH})
-RHEL_RELEASE=$(shell echo ${RELEASE} | sed -e 's/^[0-9]*.[0-9]*.[0-9]*-\([0-9]*\).*/\1/')
-RHEL_RELEASE=$(shell if [ "${RELEASE}" == "${RHEL_RELEASE}" ]; then echo ""; else echo ${RHEL_RELEASE}; fi)
+RHEL_RELEASE_TMP=$(shell echo ${RELEASE} | sed -e 's/^[0-9]*.[0-9]*.[0-9]*-\([0-9]*\).*/\1/')
+RHEL_RELEASE=$(shell if [ "${RELEASE}" == "${RHEL_RELEASE_TMP}" ]; then echo ""; else echo ${RHEL_RELEASE_TMP}; fi)
+BUILD_MODULE_TMP=$(shell if [ "${RHEL_RELEASE}" == "" ]; then echo "org"; else echo "rhel"; fi)
+BUILD_MODULE=none
+#$(info "LINUX_VERSION_CODE: ${LINUX_VERSION_CODE}, RHEL_RELEASE: ${RHEL_RELEASE}")
 ifeq ($(ENABLE_MCOVERLAYFS),yes)
-ENABLE_BUILD=$(shell if ( [ ${LINUX_VERSION_CODE} -ge 262144 ] && [ ${LINUX_VERSION_CODE} -lt 262400 ] ); then echo "yes"; else echo "no"; fi)
-else
-ENABLE_BUILD=no
+ifeq ($(BUILD_MODULE_TMP),org)
+ifeq ($(BUILD_MODULE),none)
+BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -ge 262144 -a ${LINUX_VERSION_CODE} -lt 262400 ]; then echo "linux-4.0.9"; else echo "none"; fi)
+endif
+ifeq ($(BUILD_MODULE),none)
+BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -ge 243680 -a ${LINUX_VERSION_CODE} -lt 263936 ]; then echo "linux-4.6.7"; else echo "none"; fi)
+endif
+endif
+ifeq ($(BUILD_MODULE_TMP),rhel)
+ifeq ($(BUILD_MODULE),none)
+BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -eq 199168 -a ${RHEL_RELEASE} -ge 327 -a ${RHEL_RELEASE} -le 514 ]; then echo "linux-3.10.0-327.36.1.el7"; else echo "none"; fi)
+endif
+endif
 endif
-
-obj-m += mcoverlay.o
-
-mcoverlay-y := copy_up.o dir.o inode.o readdir.o super.o
 
 .PHONY: clean install modules
 
 modules:
-ifeq ($(ENABLE_BUILD),yes)
-	$(MAKE) -C $(KDIR) M=$(PWD) SUBDIRS=$(PWD) ARCH=$(ARCH) modules
+ifneq ($(BUILD_MODULE),none)
+	@(cd $(BUILD_MODULE); make modules)
 endif
 
 clean:
-	$(RM) .*.cmd *.mod.c *.o *.ko* Module.symvers modules.order -r .tmp*
+	@(cd linux-3.10.0-327.36.1.el7; make clean)
+	@(cd linux-4.0.9; make clean)
+	@(cd linux-4.6.7; make clean)
 
 install:
-ifeq ($(ENABLE_BUILD),yes)
-	mkdir -p -m 755 $(KMODDIR)
-	install -m 644 mcoverlay.ko $(KMODDIR)
+ifneq ($(BUILD_MODULE),none)
+	@(cd $(BUILD_MODULE); make install)
 endif
 

executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile.in

@@ -0,0 +1,21 @@
+KDIR ?= @KDIR@
+ARCH ?= @ARCH@
+KMODDIR = @KMODDIR@
+src = @abs_srcdir@
+
+obj-m += mcoverlay.o
+
+mcoverlay-y := copy_up.o dir.o inode.o readdir.o super.o
+
+.PHONY: clean install modules
+
+modules:
+	$(MAKE) -C $(KDIR) M=$(PWD) SUBDIRS=$(PWD) ARCH=$(ARCH) modules
+
+clean:
+	$(RM) .*.cmd *.mod.c *.o *.ko* Module.symvers modules.order -r .tmp*
+
+install:
+	mkdir -p -m 755 $(KMODDIR)
+	install -m 644 mcoverlay.ko $(KMODDIR)
+

executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/copy_up.c

@@ -0,0 +1,461 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/file.h>
+#include <linux/splice.h>
+#include <linux/xattr.h>
+#include <linux/security.h>
+#include <linux/uaccess.h>
+#include <linux/sched.h>
+#include <linux/namei.h>
+#include <linux/fdtable.h>
+#include <linux/ratelimit.h>
+#include "overlayfs.h"
+
+#define OVL_COPY_UP_CHUNK_SIZE (1 << 20)
+
+static unsigned ovl_check_copy_up = 1;
+module_param_named(check_copy_up, ovl_check_copy_up, uint,
+		   S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(ovl_check_copy_up,
+		 "Warn on copy-up when causing process also has a R/O fd open");
+
+static int ovl_check_fd(const void *data, struct file *f, unsigned fd)
+{
+	const struct dentry *dentry = data;
+
+	if (f->f_path.dentry == dentry)
+		pr_warn_ratelimited("overlayfs: Warning: Copying up %pD, but open R/O on fd %u which will cease to be coherent [pid=%d %s]\n",
+				    f, fd, current->pid, current->comm);
+	return 0;
+}
+
+/*
+ * Check the fds open by this process and warn if something like the following
+ * scenario is about to occur:
+ *
+ *	fd1 = open("foo", O_RDONLY);
+ *	fd2 = open("foo", O_RDWR);
+ */
+static void ovl_do_check_copy_up(struct dentry *dentry)
+{
+	if (ovl_check_copy_up)
+		iterate_fd(current->files, 0, ovl_check_fd, dentry);
+}
+
+int ovl_copy_xattr(struct dentry *old, struct dentry *new)
+{
+	ssize_t list_size, size, value_size = 0;
+	char *buf, *name, *value = NULL;
+	int uninitialized_var(error);
+
+	if (!old->d_inode->i_op->getxattr ||
+	    !new->d_inode->i_op->getxattr)
+		return 0;
+
+	list_size = vfs_listxattr(old, NULL, 0);
+	if (list_size <= 0) {
+		if (list_size == -EOPNOTSUPP)
+			return 0;
+		return list_size;
+	}
+
+	buf = kzalloc(list_size, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	list_size = vfs_listxattr(old, buf, list_size);
+	if (list_size <= 0) {
+		error = list_size;
+		goto out;
+	}
+
+	for (name = buf; name < (buf + list_size); name += strlen(name) + 1) {
+retry:
+		size = vfs_getxattr(old, name, value, value_size);
+		if (size == -ERANGE)
+			size = vfs_getxattr(old, name, NULL, 0);
+
+		if (size < 0) {
+			error = size;
+			break;
+		}
+
+		if (size > value_size) {
+			void *new;
+
+			new = krealloc(value, size, GFP_KERNEL);
+			if (!new) {
+				error = -ENOMEM;
+				break;
+			}
+			value = new;
+			value_size = size;
+			goto retry;
+		}
+
+		error = vfs_setxattr(new, name, value, size, 0);
+		if (error)
+			break;
+	}
+	kfree(value);
+out:
+	kfree(buf);
+	return error;
+}
+
+static int ovl_copy_up_data(struct path *old, struct path *new, loff_t len)
+{
+	struct file *old_file;
+	struct file *new_file;
+	loff_t old_pos = 0;
+	loff_t new_pos = 0;
+	int error = 0;
+
+	if (len == 0)
+		return 0;
+
+	old_file = ovl_path_open(old, O_RDONLY);
+	if (IS_ERR(old_file))
+		return PTR_ERR(old_file);
+
+	new_file = ovl_path_open(new, O_WRONLY);
+	if (IS_ERR(new_file)) {
+		error = PTR_ERR(new_file);
+		goto out_fput;
+	}
+
+	/* FIXME: copy up sparse files efficiently */
+	while (len) {
+		size_t this_len = OVL_COPY_UP_CHUNK_SIZE;
+		long bytes;
+
+		if (len < this_len)
+			this_len = len;
+
+		if (signal_pending_state(TASK_KILLABLE, current)) {
+			error = -EINTR;
+			break;
+		}
+
+		bytes = do_splice_direct(old_file, &old_pos,
+					 new_file, &new_pos,
+					 this_len, SPLICE_F_MOVE);
+		if (bytes <= 0) {
+			error = bytes;
+			break;
+		}
+		WARN_ON(old_pos != new_pos);
+
+		len -= bytes;
+	}
+
+	fput(new_file);
+out_fput:
+	fput(old_file);
+	return error;
+}
+
+static char *ovl_read_symlink(struct dentry *realdentry)
+{
+	int res;
+	char *buf;
+	struct inode *inode = realdentry->d_inode;
+	mm_segment_t old_fs;
+
+	res = -EINVAL;
+	if (!inode->i_op->readlink)
+		goto err;
+
+	res = -ENOMEM;
+	buf = (char *) __get_free_page(GFP_KERNEL);
+	if (!buf)
+		goto err;
+
+	old_fs = get_fs();
+	set_fs(get_ds());
+	/* The cast to a user pointer is valid due to the set_fs() */
+	res = inode->i_op->readlink(realdentry,
+				    (char __user *)buf, PAGE_SIZE - 1);
+	set_fs(old_fs);
+	if (res < 0) {
+		free_page((unsigned long) buf);
+		goto err;
+	}
+	buf[res] = '\0';
+
+	return buf;
+
+err:
+	return ERR_PTR(res);
+}
+
+static int ovl_set_timestamps(struct dentry *upperdentry, struct kstat *stat)
+{
+	struct iattr attr = {
+		.ia_valid =
+		     ATTR_ATIME | ATTR_MTIME | ATTR_ATIME_SET | ATTR_MTIME_SET,
+		.ia_atime = stat->atime,
+		.ia_mtime = stat->mtime,
+	};
+
+	return notify_change(upperdentry, &attr, NULL);
+}
+
+int ovl_set_attr(struct dentry *upperdentry, struct kstat *stat)
+{
+	int err = 0;
+
+	if (!S_ISLNK(stat->mode)) {
+		struct iattr attr = {
+			.ia_valid = ATTR_MODE,
+			.ia_mode = stat->mode,
+		};
+		err = notify_change(upperdentry, &attr, NULL);
+	}
+	if (!err) {
+		struct iattr attr = {
+			.ia_valid = ATTR_UID | ATTR_GID,
+			.ia_uid = stat->uid,
+			.ia_gid = stat->gid,
+		};
+		err = notify_change(upperdentry, &attr, NULL);
+	}
+	if (!err)
+		ovl_set_timestamps(upperdentry, stat);
+
+	return err;
+}
+
+static int ovl_copy_up_locked(struct dentry *workdir, struct dentry *upperdir,
+			      struct dentry *dentry, struct path *lowerpath,
+			      struct kstat *stat, struct iattr *attr,
+			      const char *link)
+{
+	struct inode *wdir = workdir->d_inode;
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *newdentry = NULL;
+	struct dentry *upper = NULL;
+	umode_t mode = stat->mode;
+	int err;
+
+	newdentry = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out;
+
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out1;
+
+	/* Can't properly set mode on creation because of the umask */
+	stat->mode &= S_IFMT;
+	err = ovl_create_real(wdir, newdentry, stat, link, NULL, true);
+	stat->mode = mode;
+	if (err)
+		goto out2;
+
+	if (S_ISREG(stat->mode)) {
+		struct path upperpath;
+		ovl_path_upper(dentry, &upperpath);
+		BUG_ON(upperpath.dentry != NULL);
+		upperpath.dentry = newdentry;
+
+		err = ovl_copy_up_data(lowerpath, &upperpath, stat->size);
+		if (err)
+			goto out_cleanup;
+	}
+
+	err = ovl_copy_xattr(lowerpath->dentry, newdentry);
+	if (err)
+		goto out_cleanup;
+
+	mutex_lock(&newdentry->d_inode->i_mutex);
+	err = ovl_set_attr(newdentry, stat);
+	if (!err && attr)
+		err = notify_change(newdentry, attr, NULL);
+	mutex_unlock(&newdentry->d_inode->i_mutex);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_do_rename(wdir, newdentry, udir, upper, 0);
+	if (err)
+		goto out_cleanup;
+
+	ovl_dentry_update(dentry, newdentry);
+	newdentry = NULL;
+
+	/*
+	 * Non-directores become opaque when copied up.
+	 */
+	if (!S_ISDIR(stat->mode))
+		ovl_dentry_set_opaque(dentry, true);
+out2:
+	dput(upper);
+out1:
+	dput(newdentry);
+out:
+	return err;
+
+out_cleanup:
+	ovl_cleanup(wdir, newdentry);
+	goto out;
+}
+
+/*
+ * Copy up a single dentry
+ *
+ * Directory renames only allowed on "pure upper" (already created on
+ * upper filesystem, never copied up).  Directories which are on lower or
+ * are merged may not be renamed.  For these -EXDEV is returned and
+ * userspace has to deal with it.  This means, when copying up a
+ * directory we can rely on it and ancestors being stable.
+ *
+ * Non-directory renames start with copy up of source if necessary.  The
+ * actual rename will only proceed once the copy up was successful.  Copy
+ * up uses upper parent i_mutex for exclusion.  Since rename can change
+ * d_parent it is possible that the copy up will lock the old parent.  At
+ * that point the file will have already been copied up anyway.
+ */
+int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry,
+		    struct path *lowerpath, struct kstat *stat,
+		    struct iattr *attr)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	int err;
+	struct kstat pstat;
+	struct path parentpath;
+	struct dentry *upperdir;
+	struct dentry *upperdentry;
+	const struct cred *old_cred;
+	struct cred *override_cred;
+	char *link = NULL;
+
+	if (WARN_ON(!workdir))
+		return -EROFS;
+
+	ovl_do_check_copy_up(lowerpath->dentry);
+
+	ovl_path_upper(parent, &parentpath);
+	upperdir = parentpath.dentry;
+
+	err = vfs_getattr(&parentpath, &pstat);
+	if (err)
+		return err;
+
+	if (S_ISLNK(stat->mode)) {
+		link = ovl_read_symlink(lowerpath->dentry);
+		if (IS_ERR(link))
+			return PTR_ERR(link);
+	}
+
+	err = -ENOMEM;
+	override_cred = prepare_creds();
+	if (!override_cred)
+		goto out_free_link;
+
+	override_cred->fsuid = stat->uid;
+	override_cred->fsgid = stat->gid;
+	/*
+	 * CAP_SYS_ADMIN for copying up extended attributes
+	 * CAP_DAC_OVERRIDE for create
+	 * CAP_FOWNER for chmod, timestamp update
+	 * CAP_FSETID for chmod
+	 * CAP_CHOWN for chown
+	 * CAP_MKNOD for mknod
+	 */
+	cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+	cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+	cap_raise(override_cred->cap_effective, CAP_FOWNER);
+	cap_raise(override_cred->cap_effective, CAP_FSETID);
+	cap_raise(override_cred->cap_effective, CAP_CHOWN);
+	cap_raise(override_cred->cap_effective, CAP_MKNOD);
+	old_cred = override_creds(override_cred);
+
+	err = -EIO;
+	if (lock_rename(workdir, upperdir) != NULL) {
+		pr_err("overlayfs: failed to lock workdir+upperdir\n");
+		goto out_unlock;
+	}
+	upperdentry = ovl_dentry_upper(dentry);
+	if (upperdentry) {
+		unlock_rename(workdir, upperdir);
+		err = 0;
+		/* Raced with another copy-up?  Do the setattr here */
+		if (attr) {
+			mutex_lock(&upperdentry->d_inode->i_mutex);
+			err = notify_change(upperdentry, attr, NULL);
+			mutex_unlock(&upperdentry->d_inode->i_mutex);
+		}
+		goto out_put_cred;
+	}
+
+	err = ovl_copy_up_locked(workdir, upperdir, dentry, lowerpath,
+				 stat, attr, link);
+	if (!err) {
+		/* Restore timestamps on parent (best effort) */
+		ovl_set_timestamps(upperdir, &pstat);
+	}
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out_put_cred:
+	revert_creds(old_cred);
+	put_cred(override_cred);
+
+out_free_link:
+	if (link)
+		free_page((unsigned long) link);
+
+	return err;
+}
+
+int ovl_copy_up(struct dentry *dentry)
+{
+	int err;
+
+	err = 0;
+	while (!err) {
+		struct dentry *next;
+		struct dentry *parent;
+		struct path lowerpath;
+		struct kstat stat;
+		enum ovl_path_type type = ovl_path_type(dentry);
+
+		if (OVL_TYPE_UPPER(type))
+			break;
+
+		next = dget(dentry);
+		/* find the topmost dentry not yet copied up */
+		for (;;) {
+			parent = dget_parent(next);
+
+			type = ovl_path_type(parent);
+			if (OVL_TYPE_UPPER(type))
+				break;
+
+			dput(next);
+			next = parent;
+		}
+
+		ovl_path_lower(next, &lowerpath);
+		err = vfs_getattr(&lowerpath, &stat);
+		if (!err)
+			err = ovl_copy_up_one(parent, next, &lowerpath, &stat, NULL);
+
+		dput(parent);
+		dput(next);
+	}
+
+	return err;
+}

executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/dir.c

@@ -0,0 +1,972 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/fs.h>
+#include <linux/namei.h>
+#include <linux/xattr.h>
+#include <linux/security.h>
+#include <linux/cred.h>
+#include "overlayfs.h"
+
+void ovl_cleanup(struct inode *wdir, struct dentry *wdentry)
+{
+	int err;
+
+	dget(wdentry);
+	if (S_ISDIR(wdentry->d_inode->i_mode))
+		err = ovl_do_rmdir(wdir, wdentry);
+	else
+		err = ovl_do_unlink(wdir, wdentry);
+	dput(wdentry);
+
+	if (err) {
+		pr_err("overlayfs: cleanup of '%pd2' failed (%i)\n",
+		       wdentry, err);
+	}
+}
+
+struct dentry *ovl_lookup_temp(struct dentry *workdir, struct dentry *dentry)
+{
+	struct dentry *temp;
+	char name[20];
+
+	snprintf(name, sizeof(name), "#%lx", (unsigned long) dentry);
+
+	temp = lookup_one_len(name, workdir, strlen(name));
+	if (!IS_ERR(temp) && temp->d_inode) {
+		pr_err("overlayfs: workdir/%s already exists\n", name);
+		dput(temp);
+		temp = ERR_PTR(-EIO);
+	}
+
+	return temp;
+}
+
+/* caller holds i_mutex on workdir */
+static struct dentry *ovl_whiteout(struct dentry *workdir,
+				   struct dentry *dentry)
+{
+	int err;
+	struct dentry *whiteout;
+	struct inode *wdir = workdir->d_inode;
+
+	whiteout = ovl_lookup_temp(workdir, dentry);
+	if (IS_ERR(whiteout))
+		return whiteout;
+
+	err = ovl_do_whiteout(wdir, whiteout);
+	if (err) {
+		dput(whiteout);
+		whiteout = ERR_PTR(err);
+	}
+
+	return whiteout;
+}
+
+int ovl_create_real(struct inode *dir, struct dentry *newdentry,
+		    struct kstat *stat, const char *link,
+		    struct dentry *hardlink, bool debug)
+{
+	int err;
+
+	if (newdentry->d_inode)
+		return -ESTALE;
+
+	if (hardlink) {
+		err = ovl_do_link(hardlink, dir, newdentry, debug);
+	} else {
+		switch (stat->mode & S_IFMT) {
+		case S_IFREG:
+			err = ovl_do_create(dir, newdentry, stat->mode, debug);
+			break;
+
+		case S_IFDIR:
+			err = ovl_do_mkdir(dir, newdentry, stat->mode, debug);
+			break;
+
+		case S_IFCHR:
+		case S_IFBLK:
+		case S_IFIFO:
+		case S_IFSOCK:
+			err = ovl_do_mknod(dir, newdentry,
+					   stat->mode, stat->rdev, debug);
+			break;
+
+		case S_IFLNK:
+			err = ovl_do_symlink(dir, newdentry, link, debug);
+			break;
+
+		default:
+			err = -EPERM;
+		}
+	}
+	if (!err && WARN_ON(!newdentry->d_inode)) {
+		/*
+		 * Not quite sure if non-instantiated dentry is legal or not.
+		 * VFS doesn't seem to care so check and warn here.
+		 */
+		err = -ENOENT;
+	}
+	return err;
+}
+
+static int ovl_set_opaque(struct dentry *upperdentry)
+{
+	return ovl_do_setxattr(upperdentry, OVL_XATTR_OPAQUE, "y", 1, 0);
+}
+
+static void ovl_remove_opaque(struct dentry *upperdentry)
+{
+	int err;
+
+	err = ovl_do_removexattr(upperdentry, OVL_XATTR_OPAQUE);
+	if (err) {
+		pr_warn("overlayfs: failed to remove opaque from '%s' (%i)\n",
+			upperdentry->d_name.name, err);
+	}
+}
+
+static int ovl_dir_getattr(struct vfsmount *mnt, struct dentry *dentry,
+			 struct kstat *stat)
+{
+	int err;
+	enum ovl_path_type type;
+	struct path realpath;
+
+	type = ovl_path_real(dentry, &realpath);
+	err = vfs_getattr(&realpath, stat);
+	if (err)
+		return err;
+
+	stat->dev = dentry->d_sb->s_dev;
+	stat->ino = dentry->d_inode->i_ino;
+
+	/*
+	 * It's probably not worth it to count subdirs to get the
+	 * correct link count.  nlink=1 seems to pacify 'find' and
+	 * other utilities.
+	 */
+	if (OVL_TYPE_MERGE(type))
+		stat->nlink = 1;
+
+	return 0;
+}
+
+static int ovl_create_upper(struct dentry *dentry, struct inode *inode,
+			    struct kstat *stat, const char *link,
+			    struct dentry *hardlink)
+{
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *newdentry;
+	int err;
+
+	mutex_lock_nested(&udir->i_mutex, I_MUTEX_PARENT);
+	newdentry = lookup_one_len(dentry->d_name.name, upperdir,
+				   dentry->d_name.len);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out_unlock;
+	err = ovl_create_real(udir, newdentry, stat, link, hardlink, false);
+	if (err)
+		goto out_dput;
+
+	ovl_dentry_version_inc(dentry->d_parent);
+	ovl_dentry_update(dentry, newdentry);
+	ovl_copyattr(newdentry->d_inode, inode);
+	d_instantiate(dentry, inode);
+	newdentry = NULL;
+out_dput:
+	dput(newdentry);
+out_unlock:
+	mutex_unlock(&udir->i_mutex);
+	return err;
+}
+
+static int ovl_lock_rename_workdir(struct dentry *workdir,
+				   struct dentry *upperdir)
+{
+	/* Workdir should not be the same as upperdir */
+	if (workdir == upperdir)
+		goto err;
+
+	/* Workdir should not be subdir of upperdir and vice versa */
+	if (lock_rename(workdir, upperdir) != NULL)
+		goto err_unlock;
+
+	return 0;
+
+err_unlock:
+	unlock_rename(workdir, upperdir);
+err:
+	pr_err("overlayfs: failed to lock workdir+upperdir\n");
+	return -EIO;
+}
+
+static struct dentry *ovl_clear_empty(struct dentry *dentry,
+				      struct list_head *list)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct path upperpath;
+	struct dentry *upper;
+	struct dentry *opaquedir;
+	struct kstat stat;
+	int err;
+
+	if (WARN_ON(!workdir))
+		return ERR_PTR(-EROFS);
+
+	err = ovl_lock_rename_workdir(workdir, upperdir);
+	if (err)
+		goto out;
+
+	ovl_path_upper(dentry, &upperpath);
+	err = vfs_getattr(&upperpath, &stat);
+	if (err)
+		goto out_unlock;
+
+	err = -ESTALE;
+	if (!S_ISDIR(stat.mode))
+		goto out_unlock;
+	upper = upperpath.dentry;
+	if (upper->d_parent->d_inode != udir)
+		goto out_unlock;
+
+	opaquedir = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(opaquedir);
+	if (IS_ERR(opaquedir))
+		goto out_unlock;
+
+	err = ovl_create_real(wdir, opaquedir, &stat, NULL, NULL, true);
+	if (err)
+		goto out_dput;
+
+	err = ovl_copy_xattr(upper, opaquedir);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_set_opaque(opaquedir);
+	if (err)
+		goto out_cleanup;
+
+	mutex_lock(&opaquedir->d_inode->i_mutex);
+	err = ovl_set_attr(opaquedir, &stat);
+	mutex_unlock(&opaquedir->d_inode->i_mutex);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_do_rename(wdir, opaquedir, udir, upper, RENAME_EXCHANGE);
+	if (err)
+		goto out_cleanup;
+
+	ovl_cleanup_whiteouts(upper, list);
+	ovl_cleanup(wdir, upper);
+	unlock_rename(workdir, upperdir);
+
+	/* dentry's upper doesn't match now, get rid of it */
+	d_drop(dentry);
+
+	return opaquedir;
+
+out_cleanup:
+	ovl_cleanup(wdir, opaquedir);
+out_dput:
+	dput(opaquedir);
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out:
+	return ERR_PTR(err);
+}
+
+static struct dentry *ovl_check_empty_and_clear(struct dentry *dentry)
+{
+	int err;
+	struct dentry *ret = NULL;
+	LIST_HEAD(list);
+
+	err = ovl_check_empty_dir(dentry, &list);
+	if (err)
+		ret = ERR_PTR(err);
+	else {
+		/*
+		 * If no upperdentry then skip clearing whiteouts.
+		 *
+		 * Can race with copy-up, since we don't hold the upperdir
+		 * mutex.  Doesn't matter, since copy-up can't create a
+		 * non-empty directory from an empty one.
+		 */
+		if (ovl_dentry_upper(dentry))
+			ret = ovl_clear_empty(dentry, &list);
+	}
+
+	ovl_cache_free(&list);
+
+	return ret;
+}
+
+static int ovl_create_over_whiteout(struct dentry *dentry, struct inode *inode,
+				    struct kstat *stat, const char *link,
+				    struct dentry *hardlink)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *upper;
+	struct dentry *newdentry;
+	int err;
+
+	if (WARN_ON(!workdir))
+		return -EROFS;
+
+	err = ovl_lock_rename_workdir(workdir, upperdir);
+	if (err)
+		goto out;
+
+	newdentry = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out_unlock;
+
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out_dput;
+
+	err = ovl_create_real(wdir, newdentry, stat, link, hardlink, true);
+	if (err)
+		goto out_dput2;
+
+	if (S_ISDIR(stat->mode)) {
+		err = ovl_set_opaque(newdentry);
+		if (err)
+			goto out_cleanup;
+
+		err = ovl_do_rename(wdir, newdentry, udir, upper,
+				    RENAME_EXCHANGE);
+		if (err)
+			goto out_cleanup;
+
+		ovl_cleanup(wdir, upper);
+	} else {
+		err = ovl_do_rename(wdir, newdentry, udir, upper, 0);
+		if (err)
+			goto out_cleanup;
+	}
+	ovl_dentry_version_inc(dentry->d_parent);
+	ovl_dentry_update(dentry, newdentry);
+	ovl_copyattr(newdentry->d_inode, inode);
+	d_instantiate(dentry, inode);
+	newdentry = NULL;
+out_dput2:
+	dput(upper);
+out_dput:
+	dput(newdentry);
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out:
+	return err;
+
+out_cleanup:
+	ovl_cleanup(wdir, newdentry);
+	goto out_dput2;
+}
+
+static int ovl_create_or_link(struct dentry *dentry, int mode, dev_t rdev,
+			      const char *link, struct dentry *hardlink)
+{
+	int err;
+	struct inode *inode;
+	struct kstat stat = {
+		.mode = mode,
+		.rdev = rdev,
+	};
+
+	err = -ENOMEM;
+	inode = ovl_new_inode(dentry->d_sb, mode, dentry->d_fsdata);
+	if (!inode)
+		goto out;
+
+	err = ovl_copy_up(dentry->d_parent);
+	if (err)
+		goto out_iput;
+
+	if (!ovl_dentry_is_opaque(dentry)) {
+		err = ovl_create_upper(dentry, inode, &stat, link, hardlink);
+	} else {
+		const struct cred *old_cred;
+		struct cred *override_cred;
+
+		err = -ENOMEM;
+		override_cred = prepare_creds();
+		if (!override_cred)
+			goto out_iput;
+
+		/*
+		 * CAP_SYS_ADMIN for setting opaque xattr
+		 * CAP_DAC_OVERRIDE for create in workdir, rename
+		 * CAP_FOWNER for removing whiteout from sticky dir
+		 */
+		cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		cap_raise(override_cred->cap_effective, CAP_FOWNER);
+		old_cred = override_creds(override_cred);
+
+		err = ovl_create_over_whiteout(dentry, inode, &stat, link,
+					       hardlink);
+
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+
+	if (!err)
+		inode = NULL;
+out_iput:
+	iput(inode);
+out:
+	return err;
+}
+
+static int ovl_create_object(struct dentry *dentry, int mode, dev_t rdev,
+			     const char *link)
+{
+	int err;
+
+	err = ovl_want_write(dentry);
+	if (!err) {
+		err = ovl_create_or_link(dentry, mode, rdev, link, NULL);
+		ovl_drop_write(dentry);
+	}
+
+	return err;
+}
+
+static int ovl_create(struct inode *dir, struct dentry *dentry, umode_t mode,
+		      bool excl)
+{
+	return ovl_create_object(dentry, (mode & 07777) | S_IFREG, 0, NULL);
+}
+
+static int ovl_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
+{
+	return ovl_create_object(dentry, (mode & 07777) | S_IFDIR, 0, NULL);
+}
+
+static int ovl_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
+		     dev_t rdev)
+{
+	/* Don't allow creation of "whiteout" on overlay */
+	if (S_ISCHR(mode) && rdev == WHITEOUT_DEV)
+		return -EPERM;
+
+	return ovl_create_object(dentry, mode, rdev, NULL);
+}
+
+static int ovl_symlink(struct inode *dir, struct dentry *dentry,
+		       const char *link)
+{
+	return ovl_create_object(dentry, S_IFLNK, 0, link);
+}
+
+static int ovl_link(struct dentry *old, struct inode *newdir,
+		    struct dentry *new)
+{
+	int err;
+	struct dentry *upper;
+
+	err = ovl_want_write(old);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(old);
+	if (err)
+		goto out_drop_write;
+
+	upper = ovl_dentry_upper(old);
+	err = ovl_create_or_link(new, upper->d_inode->i_mode, 0, NULL, upper);
+
+out_drop_write:
+	ovl_drop_write(old);
+out:
+	return err;
+}
+
+static int ovl_remove_and_whiteout(struct dentry *dentry, bool is_dir)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *whiteout;
+	struct dentry *upper;
+	struct dentry *opaquedir = NULL;
+	int err;
+	int flags = 0;
+
+	if (WARN_ON(!workdir))
+		return -EROFS;
+
+	if (is_dir) {
+		if (OVL_TYPE_MERGE_OR_LOWER(ovl_path_type(dentry))) {
+			opaquedir = ovl_check_empty_and_clear(dentry);
+			err = PTR_ERR(opaquedir);
+			if (IS_ERR(opaquedir))
+				goto out;
+		} else {
+			LIST_HEAD(list);
+
+			/*
+			 * When removing an empty opaque directory, then it
+			 * makes no sense to replace it with an exact replica of
+			 * itself.  But emptiness still needs to be checked.
+			 */
+			err = ovl_check_empty_dir(dentry, &list);
+			ovl_cache_free(&list);
+			if (err)
+				goto out;
+		}
+	}
+
+	err = ovl_lock_rename_workdir(workdir, upperdir);
+	if (err)
+		goto out_dput;
+
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out_unlock;
+
+	err = -ESTALE;
+	if ((opaquedir && upper != opaquedir) ||
+	    (!opaquedir && ovl_dentry_upper(dentry) &&
+	     upper != ovl_dentry_upper(dentry))) {
+		goto out_dput_upper;
+	}
+
+	whiteout = ovl_whiteout(workdir, dentry);
+	err = PTR_ERR(whiteout);
+	if (IS_ERR(whiteout))
+		goto out_dput_upper;
+
+	if (d_is_dir(upper))
+		flags = RENAME_EXCHANGE;
+
+	err = ovl_do_rename(wdir, whiteout, udir, upper, flags);
+	if (err)
+		goto kill_whiteout;
+	if (flags)
+		ovl_cleanup(wdir, upper);
+
+	ovl_dentry_version_inc(dentry->d_parent);
+out_d_drop:
+	d_drop(dentry);
+	dput(whiteout);
+out_dput_upper:
+	dput(upper);
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out_dput:
+	dput(opaquedir);
+out:
+	return err;
+
+kill_whiteout:
+	ovl_cleanup(wdir, whiteout);
+	goto out_d_drop;
+}
+
+static int ovl_remove_upper(struct dentry *dentry, bool is_dir)
+{
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *dir = upperdir->d_inode;
+	struct dentry *upper;
+	int err;
+
+	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out_unlock;
+
+	err = -ESTALE;
+	if (upper == ovl_dentry_upper(dentry)) {
+		if (is_dir)
+			err = vfs_rmdir(dir, upper);
+		else
+			err = vfs_unlink(dir, upper, NULL);
+		ovl_dentry_version_inc(dentry->d_parent);
+	}
+	dput(upper);
+
+	/*
+	 * Keeping this dentry hashed would mean having to release
+	 * upperpath/lowerpath, which could only be done if we are the
+	 * sole user of this dentry.  Too tricky...  Just unhash for
+	 * now.
+	 */
+	if (!err)
+		d_drop(dentry);
+out_unlock:
+	mutex_unlock(&dir->i_mutex);
+
+	return err;
+}
+
+static inline int ovl_check_sticky(struct dentry *dentry)
+{
+	struct inode *dir = ovl_dentry_real(dentry->d_parent)->d_inode;
+	struct inode *inode = ovl_dentry_real(dentry)->d_inode;
+
+	if (check_sticky(dir, inode))
+		return -EPERM;
+
+	return 0;
+}
+
+static int ovl_do_remove(struct dentry *dentry, bool is_dir)
+{
+	enum ovl_path_type type;
+	int err;
+
+	err = ovl_check_sticky(dentry);
+	if (err)
+		goto out;
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(dentry->d_parent);
+	if (err)
+		goto out_drop_write;
+
+	type = ovl_path_type(dentry);
+	if (OVL_TYPE_PURE_UPPER(type)) {
+		err = ovl_remove_upper(dentry, is_dir);
+	} else {
+		const struct cred *old_cred;
+		struct cred *override_cred;
+
+		err = -ENOMEM;
+		override_cred = prepare_creds();
+		if (!override_cred)
+			goto out_drop_write;
+
+		/*
+		 * CAP_SYS_ADMIN for setting xattr on whiteout, opaque dir
+		 * CAP_DAC_OVERRIDE for create in workdir, rename
+		 * CAP_FOWNER for removing whiteout from sticky dir
+		 * CAP_FSETID for chmod of opaque dir
+		 * CAP_CHOWN for chown of opaque dir
+		 */
+		cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		cap_raise(override_cred->cap_effective, CAP_FOWNER);
+		cap_raise(override_cred->cap_effective, CAP_FSETID);
+		cap_raise(override_cred->cap_effective, CAP_CHOWN);
+		old_cred = override_creds(override_cred);
+
+		err = ovl_remove_and_whiteout(dentry, is_dir);
+
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static int ovl_unlink(struct inode *dir, struct dentry *dentry)
+{
+	return ovl_do_remove(dentry, false);
+}
+
+static int ovl_rmdir(struct inode *dir, struct dentry *dentry)
+{
+	return ovl_do_remove(dentry, true);
+}
+
+static int ovl_rename2(struct inode *olddir, struct dentry *old,
+		       struct inode *newdir, struct dentry *new,
+		       unsigned int flags)
+{
+	int err;
+	enum ovl_path_type old_type;
+	enum ovl_path_type new_type;
+	struct dentry *old_upperdir;
+	struct dentry *new_upperdir;
+	struct dentry *olddentry;
+	struct dentry *newdentry;
+	struct dentry *trap;
+	bool old_opaque;
+	bool new_opaque;
+	bool new_create = false;
+	bool cleanup_whiteout = false;
+	bool overwrite = !(flags & RENAME_EXCHANGE);
+	bool is_dir = S_ISDIR(old->d_inode->i_mode);
+	bool new_is_dir = false;
+	struct dentry *opaquedir = NULL;
+	const struct cred *old_cred = NULL;
+	struct cred *override_cred = NULL;
+
+	err = -EINVAL;
+	if (flags & ~(RENAME_EXCHANGE | RENAME_NOREPLACE))
+		goto out;
+
+	flags &= ~RENAME_NOREPLACE;
+
+	err = ovl_check_sticky(old);
+	if (err)
+		goto out;
+
+	/* Don't copy up directory trees */
+	old_type = ovl_path_type(old);
+	err = -EXDEV;
+	if (OVL_TYPE_MERGE_OR_LOWER(old_type) && is_dir)
+		goto out;
+
+	if (new->d_inode) {
+		err = ovl_check_sticky(new);
+		if (err)
+			goto out;
+
+		if (S_ISDIR(new->d_inode->i_mode))
+			new_is_dir = true;
+
+		new_type = ovl_path_type(new);
+		err = -EXDEV;
+		if (!overwrite && OVL_TYPE_MERGE_OR_LOWER(new_type) && new_is_dir)
+			goto out;
+
+		err = 0;
+		if (!OVL_TYPE_UPPER(new_type) && !OVL_TYPE_UPPER(old_type)) {
+			if (ovl_dentry_lower(old)->d_inode ==
+			    ovl_dentry_lower(new)->d_inode)
+				goto out;
+		}
+		if (OVL_TYPE_UPPER(new_type) && OVL_TYPE_UPPER(old_type)) {
+			if (ovl_dentry_upper(old)->d_inode ==
+			    ovl_dentry_upper(new)->d_inode)
+				goto out;
+		}
+	} else {
+		if (ovl_dentry_is_opaque(new))
+			new_type = __OVL_PATH_UPPER;
+		else
+			new_type = __OVL_PATH_UPPER | __OVL_PATH_PURE;
+	}
+
+	err = ovl_want_write(old);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(old);
+	if (err)
+		goto out_drop_write;
+
+	err = ovl_copy_up(new->d_parent);
+	if (err)
+		goto out_drop_write;
+	if (!overwrite) {
+		err = ovl_copy_up(new);
+		if (err)
+			goto out_drop_write;
+	}
+
+	old_opaque = !OVL_TYPE_PURE_UPPER(old_type);
+	new_opaque = !OVL_TYPE_PURE_UPPER(new_type);
+
+	if (old_opaque || new_opaque) {
+		err = -ENOMEM;
+		override_cred = prepare_creds();
+		if (!override_cred)
+			goto out_drop_write;
+
+		/*
+		 * CAP_SYS_ADMIN for setting xattr on whiteout, opaque dir
+		 * CAP_DAC_OVERRIDE for create in workdir
+		 * CAP_FOWNER for removing whiteout from sticky dir
+		 * CAP_FSETID for chmod of opaque dir
+		 * CAP_CHOWN for chown of opaque dir
+		 */
+		cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		cap_raise(override_cred->cap_effective, CAP_FOWNER);
+		cap_raise(override_cred->cap_effective, CAP_FSETID);
+		cap_raise(override_cred->cap_effective, CAP_CHOWN);
+		old_cred = override_creds(override_cred);
+	}
+
+	if (overwrite && OVL_TYPE_MERGE_OR_LOWER(new_type) && new_is_dir) {
+		opaquedir = ovl_check_empty_and_clear(new);
+		err = PTR_ERR(opaquedir);
+		if (IS_ERR(opaquedir)) {
+			opaquedir = NULL;
+			goto out_revert_creds;
+		}
+	}
+
+	if (overwrite) {
+		if (old_opaque) {
+			if (new->d_inode || !new_opaque) {
+				/* Whiteout source */
+				flags |= RENAME_WHITEOUT;
+			} else {
+				/* Switch whiteouts */
+				flags |= RENAME_EXCHANGE;
+			}
+		} else if (is_dir && !new->d_inode && new_opaque) {
+			flags |= RENAME_EXCHANGE;
+			cleanup_whiteout = true;
+		}
+	}
+
+	old_upperdir = ovl_dentry_upper(old->d_parent);
+	new_upperdir = ovl_dentry_upper(new->d_parent);
+
+	trap = lock_rename(new_upperdir, old_upperdir);
+
+
+	olddentry = lookup_one_len(old->d_name.name, old_upperdir,
+				   old->d_name.len);
+	err = PTR_ERR(olddentry);
+	if (IS_ERR(olddentry))
+		goto out_unlock;
+
+	err = -ESTALE;
+	if (olddentry != ovl_dentry_upper(old))
+		goto out_dput_old;
+
+	newdentry = lookup_one_len(new->d_name.name, new_upperdir,
+				   new->d_name.len);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out_dput_old;
+
+	err = -ESTALE;
+	if (ovl_dentry_upper(new)) {
+		if (opaquedir) {
+			if (newdentry != opaquedir)
+				goto out_dput;
+		} else {
+			if (newdentry != ovl_dentry_upper(new))
+				goto out_dput;
+		}
+	} else {
+		new_create = true;
+		if (!d_is_negative(newdentry) &&
+		    (!new_opaque || !ovl_is_whiteout(newdentry)))
+			goto out_dput;
+	}
+
+	if (olddentry == trap)
+		goto out_dput;
+	if (newdentry == trap)
+		goto out_dput;
+
+	if (is_dir && !old_opaque && new_opaque) {
+		err = ovl_set_opaque(olddentry);
+		if (err)
+			goto out_dput;
+	}
+	if (!overwrite && new_is_dir && old_opaque && !new_opaque) {
+		err = ovl_set_opaque(newdentry);
+		if (err)
+			goto out_dput;
+	}
+
+	if (old_opaque || new_opaque) {
+		err = ovl_do_rename(old_upperdir->d_inode, olddentry,
+				    new_upperdir->d_inode, newdentry,
+				    flags);
+	} else {
+		/* No debug for the plain case */
+		BUG_ON(flags & ~RENAME_EXCHANGE);
+		err = vfs_rename(old_upperdir->d_inode, olddentry,
+				 new_upperdir->d_inode, newdentry,
+				 NULL, flags);
+	}
+
+	if (err) {
+		if (is_dir && !old_opaque && new_opaque)
+			ovl_remove_opaque(olddentry);
+		if (!overwrite && new_is_dir && old_opaque && !new_opaque)
+			ovl_remove_opaque(newdentry);
+		goto out_dput;
+	}
+
+	if (is_dir && old_opaque && !new_opaque)
+		ovl_remove_opaque(olddentry);
+	if (!overwrite && new_is_dir && !old_opaque && new_opaque)
+		ovl_remove_opaque(newdentry);
+
+	if (old_opaque != new_opaque) {
+		ovl_dentry_set_opaque(old, new_opaque);
+		if (!overwrite)
+			ovl_dentry_set_opaque(new, old_opaque);
+	}
+
+	if (cleanup_whiteout)
+		ovl_cleanup(old_upperdir->d_inode, newdentry);
+
+	ovl_dentry_version_inc(old->d_parent);
+	ovl_dentry_version_inc(new->d_parent);
+
+out_dput:
+	dput(newdentry);
+out_dput_old:
+	dput(olddentry);
+out_unlock:
+	unlock_rename(new_upperdir, old_upperdir);
+out_revert_creds:
+	if (old_opaque || new_opaque) {
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+out_drop_write:
+	ovl_drop_write(old);
+out:
+	dput(opaquedir);
+	return err;
+}
+
+static int ovl_rename(struct inode *olddir, struct dentry *old,
+		      struct inode *newdir, struct dentry *new)
+{
+	return ovl_rename2(olddir, old, newdir, new, 0);
+}
+
+const struct inode_operations_wrapper ovl_dir_inode_operations = {
+	.ops = {
+	.lookup		= ovl_lookup,
+	.mkdir		= ovl_mkdir,
+	.symlink	= ovl_symlink,
+	.unlink		= ovl_unlink,
+	.rmdir		= ovl_rmdir,
+	.rename		= ovl_rename,
+	.link		= ovl_link,
+	.setattr	= ovl_setattr,
+	.create		= ovl_create,
+	.mknod		= ovl_mknod,
+	.permission	= ovl_permission,
+	.getattr	= ovl_dir_getattr,
+	.setxattr	= ovl_setxattr,
+	.getxattr	= ovl_getxattr,
+	.listxattr	= ovl_listxattr,
+	.removexattr	= ovl_removexattr,
+	},
+	.rename2	= ovl_rename2,
+};

executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/inode.c

@@ -0,0 +1,442 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/xattr.h>
+#include "overlayfs.h"
+
+static int ovl_copy_up_last(struct dentry *dentry, struct iattr *attr,
+			    bool no_data)
+{
+	int err;
+	struct dentry *parent;
+	struct kstat stat;
+	struct path lowerpath;
+
+	parent = dget_parent(dentry);
+	err = ovl_copy_up(parent);
+	if (err)
+		goto out_dput_parent;
+
+	ovl_path_lower(dentry, &lowerpath);
+	err = vfs_getattr(&lowerpath, &stat);
+	if (err)
+		goto out_dput_parent;
+
+	if (no_data)
+		stat.size = 0;
+
+	err = ovl_copy_up_one(parent, dentry, &lowerpath, &stat, attr);
+
+out_dput_parent:
+	dput(parent);
+	return err;
+}
+
+int ovl_setattr(struct dentry *dentry, struct iattr *attr)
+{
+	int err;
+	struct dentry *upperdentry;
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(dentry);
+	if (!err) {
+		upperdentry = ovl_dentry_upper(dentry);
+
+		mutex_lock(&upperdentry->d_inode->i_mutex);
+		err = notify_change(upperdentry, attr, NULL);
+		mutex_unlock(&upperdentry->d_inode->i_mutex);
+	}
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static int ovl_getattr(struct vfsmount *mnt, struct dentry *dentry,
+			 struct kstat *stat)
+{
+	struct path realpath;
+
+	ovl_path_real(dentry, &realpath);
+	return vfs_getattr(&realpath, stat);
+}
+
+int ovl_permission(struct inode *inode, int mask)
+{
+	struct ovl_entry *oe;
+	struct dentry *alias = NULL;
+	struct inode *realinode;
+	struct dentry *realdentry;
+	bool is_upper;
+	int err;
+
+	if (S_ISDIR(inode->i_mode)) {
+		oe = inode->i_private;
+	} else if (mask & MAY_NOT_BLOCK) {
+		return -ECHILD;
+	} else {
+		/*
+		 * For non-directories find an alias and get the info
+		 * from there.
+		 */
+		alias = d_find_any_alias(inode);
+		if (WARN_ON(!alias))
+			return -ENOENT;
+
+		oe = alias->d_fsdata;
+	}
+
+	realdentry = ovl_entry_real(oe, &is_upper);
+
+	/* Careful in RCU walk mode */
+	realinode = ACCESS_ONCE(realdentry->d_inode);
+	if (!realinode) {
+		WARN_ON(!(mask & MAY_NOT_BLOCK));
+		err = -ENOENT;
+		goto out_dput;
+	}
+
+	if (mask & MAY_WRITE) {
+		umode_t mode = realinode->i_mode;
+
+		/*
+		 * Writes will always be redirected to upper layer, so
+		 * ignore lower layer being read-only.
+		 *
+		 * If the overlay itself is read-only then proceed
+		 * with the permission check, don't return EROFS.
+		 * This will only happen if this is the lower layer of
+		 * another overlayfs.
+		 *
+		 * If upper fs becomes read-only after the overlay was
+		 * constructed return EROFS to prevent modification of
+		 * upper layer.
+		 */
+		err = -EROFS;
+		if (is_upper && !IS_RDONLY(inode) && IS_RDONLY(realinode) &&
+		    (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
+			goto out_dput;
+	}
+
+	err = __inode_permission(realinode, mask);
+out_dput:
+	dput(alias);
+	return err;
+}
+
+
+struct ovl_link_data {
+	struct dentry *realdentry;
+	void *cookie;
+};
+
+static void *ovl_follow_link(struct dentry *dentry, struct nameidata *nd)
+{
+	void *ret;
+	struct dentry *realdentry;
+	struct inode *realinode;
+	struct ovl_link_data *data = NULL;
+
+	realdentry = ovl_dentry_real(dentry);
+	realinode = realdentry->d_inode;
+
+	if (WARN_ON(!realinode->i_op->follow_link))
+		return ERR_PTR(-EPERM);
+
+	if (realinode->i_op->put_link) {
+		data = kmalloc(sizeof(struct ovl_link_data), GFP_KERNEL);
+		if (!data)
+			return ERR_PTR(-ENOMEM);
+		data->realdentry = realdentry;
+	}
+
+	ret = realinode->i_op->follow_link(realdentry, nd);
+	if (IS_ERR(ret)) {
+		kfree(data);
+		return ret;
+	}
+
+	if (data)
+		data->cookie = ret;
+
+	return data;
+}
+
+static void ovl_put_link(struct dentry *dentry, struct nameidata *nd, void *c)
+{
+	struct inode *realinode;
+	struct ovl_link_data *data = c;
+
+	if (!data)
+		return;
+
+	realinode = data->realdentry->d_inode;
+	realinode->i_op->put_link(data->realdentry, nd, data->cookie);
+	kfree(data);
+}
+
+static int ovl_readlink(struct dentry *dentry, char __user *buf, int bufsiz)
+{
+	struct path realpath;
+	struct inode *realinode;
+
+	ovl_path_real(dentry, &realpath);
+	realinode = realpath.dentry->d_inode;
+
+	if (!realinode->i_op->readlink)
+		return -EINVAL;
+
+	touch_atime(&realpath);
+
+	return realinode->i_op->readlink(realpath.dentry, buf, bufsiz);
+}
+
+
+static bool ovl_is_private_xattr(const char *name)
+{
+	return strncmp(name, OVL_XATTR_PRE_NAME, OVL_XATTR_PRE_LEN) == 0;
+}
+
+int ovl_setxattr(struct dentry *dentry, const char *name,
+		 const void *value, size_t size, int flags)
+{
+	int err;
+	struct dentry *upperdentry;
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = -EPERM;
+	if (ovl_is_private_xattr(name))
+		goto out_drop_write;
+
+	err = ovl_copy_up(dentry);
+	if (err)
+		goto out_drop_write;
+
+	upperdentry = ovl_dentry_upper(dentry);
+	err = vfs_setxattr(upperdentry, name, value, size, flags);
+
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static bool ovl_need_xattr_filter(struct dentry *dentry,
+				  enum ovl_path_type type)
+{
+	if ((type & (__OVL_PATH_PURE | __OVL_PATH_UPPER)) == __OVL_PATH_UPPER)
+		return S_ISDIR(dentry->d_inode->i_mode);
+	else
+		return false;
+}
+
+ssize_t ovl_getxattr(struct dentry *dentry, const char *name,
+		     void *value, size_t size)
+{
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+
+	if (ovl_need_xattr_filter(dentry, type) && ovl_is_private_xattr(name))
+		return -ENODATA;
+
+	return vfs_getxattr(realpath.dentry, name, value, size);
+}
+
+ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size)
+{
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+	ssize_t res;
+	int off;
+
+	res = vfs_listxattr(realpath.dentry, list, size);
+	if (res <= 0 || size == 0)
+		return res;
+
+	if (!ovl_need_xattr_filter(dentry, type))
+		return res;
+
+	/* filter out private xattrs */
+	for (off = 0; off < res;) {
+		char *s = list + off;
+		size_t slen = strlen(s) + 1;
+
+		BUG_ON(off + slen > res);
+
+		if (ovl_is_private_xattr(s)) {
+			res -= slen;
+			memmove(s, s + slen, res - off);
+		} else {
+			off += slen;
+		}
+	}
+
+	return res;
+}
+
+int ovl_removexattr(struct dentry *dentry, const char *name)
+{
+	int err;
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = -ENODATA;
+	if (ovl_need_xattr_filter(dentry, type) && ovl_is_private_xattr(name))
+		goto out_drop_write;
+
+	if (!OVL_TYPE_UPPER(type)) {
+		err = vfs_getxattr(realpath.dentry, name, NULL, 0);
+		if (err < 0)
+			goto out_drop_write;
+
+		err = ovl_copy_up(dentry);
+		if (err)
+			goto out_drop_write;
+
+		ovl_path_upper(dentry, &realpath);
+	}
+
+	err = vfs_removexattr(realpath.dentry, name);
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static bool ovl_open_need_copy_up(int flags, enum ovl_path_type type,
+				  struct dentry *realdentry)
+{
+	if (OVL_TYPE_UPPER(type))
+		return false;
+
+	if (special_file(realdentry->d_inode->i_mode))
+		return false;
+
+	if (!(OPEN_FMODE(flags) & FMODE_WRITE) && !(flags & O_TRUNC))
+		return false;
+
+	return true;
+}
+
+static int ovl_dentry_open(struct dentry *dentry, struct file *file,
+		    const struct cred *cred)
+{
+	int err;
+	struct path realpath;
+	enum ovl_path_type type;
+	bool want_write = false;
+
+	type = ovl_path_real(dentry, &realpath);
+	if (!ovl_is_nocopyupw(dentry)) {
+		if (ovl_open_need_copy_up(file->f_flags, type,
+		    realpath.dentry)) {
+			want_write = true;
+			err = ovl_want_write(dentry);
+			if (err)
+				goto out;
+
+			if (file->f_flags & O_TRUNC)
+				err = ovl_copy_up_last(dentry, NULL, true);
+			else
+				err = ovl_copy_up(dentry);
+			if (err)
+				goto out_drop_write;
+
+			ovl_path_upper(dentry, &realpath);
+		}
+	}
+
+	err = vfs_open(&realpath, file, cred);
+out_drop_write:
+	if (want_write)
+		ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static const struct inode_operations_wrapper ovl_file_inode_operations = {
+	.ops = {
+	.setattr	= ovl_setattr,
+	.permission	= ovl_permission,
+	.getattr	= ovl_getattr,
+	.setxattr	= ovl_setxattr,
+	.getxattr	= ovl_getxattr,
+	.listxattr	= ovl_listxattr,
+	.removexattr	= ovl_removexattr,
+	},
+	.dentry_open	= ovl_dentry_open,
+};
+
+static const struct inode_operations ovl_symlink_inode_operations = {
+	.setattr	= ovl_setattr,
+	.follow_link	= ovl_follow_link,
+	.put_link	= ovl_put_link,
+	.readlink	= ovl_readlink,
+	.getattr	= ovl_getattr,
+	.setxattr	= ovl_setxattr,
+	.getxattr	= ovl_getxattr,
+	.listxattr	= ovl_listxattr,
+	.removexattr	= ovl_removexattr,
+};
+
+struct inode *ovl_new_inode(struct super_block *sb, umode_t mode,
+			    struct ovl_entry *oe)
+{
+	struct inode *inode;
+
+	inode = new_inode(sb);
+	if (!inode)
+		return NULL;
+
+	mode &= S_IFMT;
+
+	inode->i_ino = get_next_ino();
+	inode->i_mode = mode;
+	inode->i_flags |= S_NOATIME | S_NOCMTIME;
+
+	switch (mode) {
+	case S_IFDIR:
+		inode->i_private = oe;
+		inode->i_op = &ovl_dir_inode_operations.ops;
+		inode->i_fop = &ovl_dir_operations;
+		inode->i_flags |= S_IOPS_WRAPPER;
+		break;
+
+	case S_IFLNK:
+		inode->i_op = &ovl_symlink_inode_operations;
+		break;
+
+	case S_IFREG:
+	case S_IFSOCK:
+	case S_IFBLK:
+	case S_IFCHR:
+	case S_IFIFO:
+		inode->i_op = &ovl_file_inode_operations.ops;
+		inode->i_flags |= S_IOPS_WRAPPER;
+		break;
+
+	default:
+		WARN(1, "illegal file type: %i\n", mode);
+		iput(inode);
+		inode = NULL;
+	}
+
+	return inode;
+}

executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/overlayfs.h

@@ -0,0 +1,200 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+
+struct ovl_entry;
+
+enum ovl_path_type {
+	__OVL_PATH_PURE		= (1 << 0),
+	__OVL_PATH_UPPER	= (1 << 1),
+	__OVL_PATH_MERGE	= (1 << 2),
+};
+
+#define OVL_TYPE_UPPER(type)	((type) & __OVL_PATH_UPPER)
+#define OVL_TYPE_MERGE(type)	((type) & __OVL_PATH_MERGE)
+#define OVL_TYPE_PURE_UPPER(type) ((type) & __OVL_PATH_PURE)
+#define OVL_TYPE_MERGE_OR_LOWER(type) \
+	(OVL_TYPE_MERGE(type) || !OVL_TYPE_UPPER(type))
+
+#define OVL_XATTR_PRE_NAME "trusted.overlay."
+#define OVL_XATTR_PRE_LEN  16
+#define OVL_XATTR_OPAQUE   OVL_XATTR_PRE_NAME"opaque"
+
+static inline int ovl_do_rmdir(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_rmdir(dir, dentry);
+	pr_debug("rmdir(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+static inline int ovl_do_unlink(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_unlink(dir, dentry, NULL);
+	pr_debug("unlink(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+static inline int ovl_do_link(struct dentry *old_dentry, struct inode *dir,
+			      struct dentry *new_dentry, bool debug)
+{
+	int err = vfs_link(old_dentry, dir, new_dentry, NULL);
+	if (debug) {
+		pr_debug("link(%pd2, %pd2) = %i\n",
+			 old_dentry, new_dentry, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_create(struct inode *dir, struct dentry *dentry,
+			     umode_t mode, bool debug)
+{
+	int err = vfs_create(dir, dentry, mode, true);
+	if (debug)
+		pr_debug("create(%pd2, 0%o) = %i\n", dentry, mode, err);
+	return err;
+}
+
+static inline int ovl_do_mkdir(struct inode *dir, struct dentry *dentry,
+			       umode_t mode, bool debug)
+{
+	int err = vfs_mkdir(dir, dentry, mode);
+	if (debug)
+		pr_debug("mkdir(%pd2, 0%o) = %i\n", dentry, mode, err);
+	return err;
+}
+
+static inline int ovl_do_mknod(struct inode *dir, struct dentry *dentry,
+			       umode_t mode, dev_t dev, bool debug)
+{
+	int err = vfs_mknod(dir, dentry, mode, dev);
+	if (debug) {
+		pr_debug("mknod(%pd2, 0%o, 0%o) = %i\n",
+			 dentry, mode, dev, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_symlink(struct inode *dir, struct dentry *dentry,
+				 const char *oldname, bool debug)
+{
+	int err = vfs_symlink(dir, dentry, oldname);
+	if (debug)
+		pr_debug("symlink(\"%s\", %pd2) = %i\n", oldname, dentry, err);
+	return err;
+}
+
+static inline int ovl_do_setxattr(struct dentry *dentry, const char *name,
+				  const void *value, size_t size, int flags)
+{
+	int err = vfs_setxattr(dentry, name, value, size, flags);
+	pr_debug("setxattr(%pd2, \"%s\", \"%*s\", 0x%x) = %i\n",
+		 dentry, name, (int) size, (char *) value, flags, err);
+	return err;
+}
+
+static inline int ovl_do_removexattr(struct dentry *dentry, const char *name)
+{
+	int err = vfs_removexattr(dentry, name);
+	pr_debug("removexattr(%pd2, \"%s\") = %i\n", dentry, name, err);
+	return err;
+}
+
+static inline int ovl_do_rename(struct inode *olddir, struct dentry *olddentry,
+				struct inode *newdir, struct dentry *newdentry,
+				unsigned int flags)
+{
+	int err;
+
+	pr_debug("rename2(%pd2, %pd2, 0x%x)\n",
+		 olddentry, newdentry, flags);
+
+	err = vfs_rename(olddir, olddentry, newdir, newdentry, NULL, flags);
+
+	if (err) {
+		pr_debug("...rename2(%pd2, %pd2, ...) = %i\n",
+			 olddentry, newdentry, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_whiteout(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_whiteout(dir, dentry);
+	pr_debug("whiteout(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+bool ovl_is_nocopyupw(struct dentry *dentry);
+enum ovl_path_type ovl_path_type(struct dentry *dentry);
+u64 ovl_dentry_version_get(struct dentry *dentry);
+void ovl_dentry_version_inc(struct dentry *dentry);
+void ovl_path_upper(struct dentry *dentry, struct path *path);
+void ovl_path_lower(struct dentry *dentry, struct path *path);
+enum ovl_path_type ovl_path_real(struct dentry *dentry, struct path *path);
+int ovl_path_next(int idx, struct dentry *dentry, struct path *path);
+struct dentry *ovl_dentry_upper(struct dentry *dentry);
+struct dentry *ovl_dentry_lower(struct dentry *dentry);
+struct dentry *ovl_dentry_real(struct dentry *dentry);
+struct dentry *ovl_entry_real(struct ovl_entry *oe, bool *is_upper);
+struct ovl_dir_cache *ovl_dir_cache(struct dentry *dentry);
+void ovl_set_dir_cache(struct dentry *dentry, struct ovl_dir_cache *cache);
+struct dentry *ovl_workdir(struct dentry *dentry);
+int ovl_want_write(struct dentry *dentry);
+void ovl_drop_write(struct dentry *dentry);
+bool ovl_dentry_is_opaque(struct dentry *dentry);
+void ovl_dentry_set_opaque(struct dentry *dentry, bool opaque);
+bool ovl_is_whiteout(struct dentry *dentry);
+void ovl_dentry_update(struct dentry *dentry, struct dentry *upperdentry);
+struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
+			  unsigned int flags);
+struct file *ovl_path_open(struct path *path, int flags);
+
+struct dentry *ovl_upper_create(struct dentry *upperdir, struct dentry *dentry,
+				struct kstat *stat, const char *link);
+
+/* readdir.c */
+extern const struct file_operations ovl_dir_operations;
+int ovl_check_empty_dir(struct dentry *dentry, struct list_head *list);
+void ovl_cleanup_whiteouts(struct dentry *upper, struct list_head *list);
+void ovl_cache_free(struct list_head *list);
+
+/* inode.c */
+int ovl_setattr(struct dentry *dentry, struct iattr *attr);
+int ovl_permission(struct inode *inode, int mask);
+int ovl_setxattr(struct dentry *dentry, const char *name,
+		 const void *value, size_t size, int flags);
+ssize_t ovl_getxattr(struct dentry *dentry, const char *name,
+		     void *value, size_t size);
+ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size);
+int ovl_removexattr(struct dentry *dentry, const char *name);
+
+struct inode *ovl_new_inode(struct super_block *sb, umode_t mode,
+			    struct ovl_entry *oe);
+static inline void ovl_copyattr(struct inode *from, struct inode *to)
+{
+	to->i_uid = from->i_uid;
+	to->i_gid = from->i_gid;
+}
+
+/* dir.c */
+extern const struct inode_operations_wrapper ovl_dir_inode_operations;
+struct dentry *ovl_lookup_temp(struct dentry *workdir, struct dentry *dentry);
+int ovl_create_real(struct inode *dir, struct dentry *newdentry,
+		    struct kstat *stat, const char *link,
+		    struct dentry *hardlink, bool debug);
+void ovl_cleanup(struct inode *dir, struct dentry *dentry);
+
+/* copy_up.c */
+int ovl_copy_up(struct dentry *dentry);
+int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry,
+		    struct path *lowerpath, struct kstat *stat,
+		    struct iattr *attr);
+int ovl_copy_xattr(struct dentry *old, struct dentry *new);
+int ovl_set_attr(struct dentry *upper, struct kstat *stat);

executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/readdir.c

@@ -0,0 +1,588 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/namei.h>
+#include <linux/file.h>
+#include <linux/xattr.h>
+#include <linux/rbtree.h>
+#include <linux/security.h>
+#include <linux/cred.h>
+#include "overlayfs.h"
+
+struct ovl_cache_entry {
+	unsigned int len;
+	unsigned int type;
+	u64 ino;
+	struct list_head l_node;
+	struct rb_node node;
+	struct ovl_cache_entry *next_maybe_whiteout;
+	bool is_whiteout;
+	char name[];
+};
+
+struct ovl_dir_cache {
+	long refcount;
+	u64 version;
+	struct list_head entries;
+};
+
+struct dir_context {
+	const filldir_t actor;
+	//loff_t pos;
+};
+
+struct ovl_readdir_data {
+	struct dir_context ctx;
+	bool is_merge;
+	struct rb_root root;
+	struct list_head *list;
+	struct list_head middle;
+	struct ovl_cache_entry *first_maybe_whiteout;
+	int count;
+	int err;
+};
+
+struct ovl_dir_file {
+	bool is_real;
+	bool is_upper;
+	struct ovl_dir_cache *cache;
+	struct list_head *cursor;
+	struct file *realfile;
+	struct file *upperfile;
+};
+
+static struct ovl_cache_entry *ovl_cache_entry_from_node(struct rb_node *n)
+{
+	return container_of(n, struct ovl_cache_entry, node);
+}
+
+static struct ovl_cache_entry *ovl_cache_entry_find(struct rb_root *root,
+						    const char *name, int len)
+{
+	struct rb_node *node = root->rb_node;
+	int cmp;
+
+	while (node) {
+		struct ovl_cache_entry *p = ovl_cache_entry_from_node(node);
+
+		cmp = strncmp(name, p->name, len);
+		if (cmp > 0)
+			node = p->node.rb_right;
+		else if (cmp < 0 || len < p->len)
+			node = p->node.rb_left;
+		else
+			return p;
+	}
+
+	return NULL;
+}
+
+static struct ovl_cache_entry *ovl_cache_entry_new(struct ovl_readdir_data *rdd,
+						   const char *name, int len,
+						   u64 ino, unsigned int d_type)
+{
+	struct ovl_cache_entry *p;
+	size_t size = offsetof(struct ovl_cache_entry, name[len + 1]);
+
+	p = kmalloc(size, GFP_KERNEL);
+	if (!p)
+		return NULL;
+
+	memcpy(p->name, name, len);
+	p->name[len] = '\0';
+	p->len = len;
+	p->type = d_type;
+	p->ino = ino;
+	p->is_whiteout = false;
+
+	if (d_type == DT_CHR) {
+		p->next_maybe_whiteout = rdd->first_maybe_whiteout;
+		rdd->first_maybe_whiteout = p;
+	}
+	return p;
+}
+
+static int ovl_cache_entry_add_rb(struct ovl_readdir_data *rdd,
+				  const char *name, int len, u64 ino,
+				  unsigned int d_type)
+{
+	struct rb_node **newp = &rdd->root.rb_node;
+	struct rb_node *parent = NULL;
+	struct ovl_cache_entry *p;
+
+	while (*newp) {
+		int cmp;
+		struct ovl_cache_entry *tmp;
+
+		parent = *newp;
+		tmp = ovl_cache_entry_from_node(*newp);
+		cmp = strncmp(name, tmp->name, len);
+		if (cmp > 0)
+			newp = &tmp->node.rb_right;
+		else if (cmp < 0 || len < tmp->len)
+			newp = &tmp->node.rb_left;
+		else
+			return 0;
+	}
+
+	p = ovl_cache_entry_new(rdd, name, len, ino, d_type);
+	if (p == NULL)
+		return -ENOMEM;
+
+	list_add_tail(&p->l_node, rdd->list);
+	rb_link_node(&p->node, parent, newp);
+	rb_insert_color(&p->node, &rdd->root);
+
+	return 0;
+}
+
+static int ovl_fill_lower(struct ovl_readdir_data *rdd,
+			  const char *name, int namelen,
+			  loff_t offset, u64 ino, unsigned int d_type)
+{
+	struct ovl_cache_entry *p;
+
+	p = ovl_cache_entry_find(&rdd->root, name, namelen);
+	if (p) {
+		list_move_tail(&p->l_node, &rdd->middle);
+	} else {
+		p = ovl_cache_entry_new(rdd, name, namelen, ino, d_type);
+		if (p == NULL)
+			rdd->err = -ENOMEM;
+		else
+			list_add_tail(&p->l_node, &rdd->middle);
+	}
+
+	return rdd->err;
+}
+
+void ovl_cache_free(struct list_head *list)
+{
+	struct ovl_cache_entry *p;
+	struct ovl_cache_entry *n;
+
+	list_for_each_entry_safe(p, n, list, l_node)
+		kfree(p);
+
+	INIT_LIST_HEAD(list);
+}
+
+static void ovl_cache_put(struct ovl_dir_file *od, struct dentry *dentry)
+{
+	struct ovl_dir_cache *cache = od->cache;
+
+	WARN_ON(cache->refcount <= 0);
+	cache->refcount--;
+	if (!cache->refcount) {
+		if (ovl_dir_cache(dentry) == cache)
+			ovl_set_dir_cache(dentry, NULL);
+
+		ovl_cache_free(&cache->entries);
+		kfree(cache);
+	}
+}
+
+static int ovl_fill_merge(void *buf, const char *name, int namelen,
+			  loff_t offset, u64 ino, unsigned int d_type)
+{
+	struct dir_context *ctx = buf;
+	struct ovl_readdir_data *rdd =
+		container_of(ctx, struct ovl_readdir_data, ctx);
+
+	rdd->count++;
+	if (!rdd->is_merge)
+		return ovl_cache_entry_add_rb(rdd, name, namelen, ino, d_type);
+	else
+		return ovl_fill_lower(rdd, name, namelen, offset, ino, d_type);
+}
+
+static int ovl_check_whiteouts(struct dentry *dir, struct ovl_readdir_data *rdd)
+{
+	int err;
+	struct ovl_cache_entry *p;
+	struct dentry *dentry;
+	const struct cred *old_cred;
+	struct cred *override_cred;
+
+	override_cred = prepare_creds();
+	if (!override_cred)
+		return -ENOMEM;
+
+	/*
+	 * CAP_DAC_OVERRIDE for lookup
+	 */
+	cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+	old_cred = override_creds(override_cred);
+
+	err = mutex_lock_killable(&dir->d_inode->i_mutex);
+	if (!err) {
+		while (rdd->first_maybe_whiteout) {
+			p = rdd->first_maybe_whiteout;
+			rdd->first_maybe_whiteout = p->next_maybe_whiteout;
+			dentry = lookup_one_len(p->name, dir, p->len);
+			if (!IS_ERR(dentry)) {
+				p->is_whiteout = ovl_is_whiteout(dentry);
+				dput(dentry);
+			}
+		}
+		mutex_unlock(&dir->d_inode->i_mutex);
+	}
+	revert_creds(old_cred);
+	put_cred(override_cred);
+
+	return err;
+}
+
+static inline int ovl_dir_read(struct path *realpath,
+			       struct ovl_readdir_data *rdd)
+{
+	struct file *realfile;
+	int err;
+
+	realfile = ovl_path_open(realpath, O_RDONLY | O_DIRECTORY);
+	if (IS_ERR(realfile))
+		return PTR_ERR(realfile);
+
+	rdd->first_maybe_whiteout = NULL;
+	//rdd->ctx.pos = 0;
+	do {
+		rdd->count = 0;
+		rdd->err = 0;
+		err = vfs_readdir(realfile, rdd->ctx.actor, rdd);
+		if (err >= 0)
+			err = rdd->err;
+	} while (!err && rdd->count);
+
+	if (!err && rdd->first_maybe_whiteout)
+		err = ovl_check_whiteouts(realpath->dentry, rdd);
+
+	fput(realfile);
+
+	return err;
+}
+
+static void ovl_dir_reset(struct file *file)
+{
+	struct ovl_dir_file *od = file->private_data;
+	struct ovl_dir_cache *cache = od->cache;
+	struct dentry *dentry = file->f_path.dentry;
+	enum ovl_path_type type = ovl_path_type(dentry);
+
+	if (cache && ovl_dentry_version_get(dentry) != cache->version) {
+		ovl_cache_put(od, dentry);
+		od->cache = NULL;
+		od->cursor = NULL;
+	}
+	WARN_ON(!od->is_real && !OVL_TYPE_MERGE(type));
+	if (od->is_real && OVL_TYPE_MERGE(type))
+		od->is_real = false;
+}
+
+static int ovl_dir_read_merged(struct dentry *dentry, struct list_head *list)
+{
+	int err;
+	struct path realpath;
+	struct ovl_readdir_data rdd = {
+		.ctx.actor = ovl_fill_merge,
+		.list = list,
+		.root = RB_ROOT,
+		.is_merge = false,
+	};
+	int idx, next;
+
+	for (idx = 0; idx != -1; idx = next) {
+		next = ovl_path_next(idx, dentry, &realpath);
+
+		if (next != -1) {
+			err = ovl_dir_read(&realpath, &rdd);
+			if (err)
+				break;
+		} else {
+			/*
+			 * Insert lowest layer entries before upper ones, this
+			 * allows offsets to be reasonably constant
+			 */
+			list_add(&rdd.middle, rdd.list);
+			rdd.is_merge = true;
+			err = ovl_dir_read(&realpath, &rdd);
+			list_del(&rdd.middle);
+		}
+	}
+	return err;
+}
+
+static void ovl_seek_cursor(struct ovl_dir_file *od, loff_t pos)
+{
+	struct list_head *p;
+	loff_t off = 0;
+
+	list_for_each(p, &od->cache->entries) {
+		if (off >= pos)
+			break;
+		off++;
+	}
+	/* Cursor is safe since the cache is stable */
+	od->cursor = p;
+}
+
+static struct ovl_dir_cache *ovl_cache_get(struct dentry *dentry)
+{
+	int res;
+	struct ovl_dir_cache *cache;
+
+	cache = ovl_dir_cache(dentry);
+	if (cache && ovl_dentry_version_get(dentry) == cache->version) {
+		cache->refcount++;
+		return cache;
+	}
+	ovl_set_dir_cache(dentry, NULL);
+
+	cache = kzalloc(sizeof(struct ovl_dir_cache), GFP_KERNEL);
+	if (!cache)
+		return ERR_PTR(-ENOMEM);
+
+	cache->refcount = 1;
+	INIT_LIST_HEAD(&cache->entries);
+
+	res = ovl_dir_read_merged(dentry, &cache->entries);
+	if (res) {
+		ovl_cache_free(&cache->entries);
+		kfree(cache);
+		return ERR_PTR(res);
+	}
+
+	cache->version = ovl_dentry_version_get(dentry);
+	ovl_set_dir_cache(dentry, cache);
+
+	return cache;
+}
+
+static int ovl_readdir(struct file *file, void *buf, filldir_t filler)
+{
+	struct ovl_dir_file *od = file->private_data;
+	struct dentry *dentry = file->f_path.dentry;
+	struct ovl_cache_entry *p;
+	int res;
+
+	if (!file->f_pos)
+		ovl_dir_reset(file);
+
+	if (od->is_real) {
+		res = vfs_readdir(od->realfile, filler, buf);
+		file->f_pos = od->realfile->f_pos;
+
+		return res;
+	}
+
+	if (!od->cache) {
+		struct ovl_dir_cache *cache;
+
+		cache = ovl_cache_get(dentry);
+		if (IS_ERR(cache))
+			return PTR_ERR(cache);
+
+		od->cache = cache;
+		ovl_seek_cursor(od, file->f_pos);
+	}
+
+	while (od->cursor != &od->cache->entries) {
+		p = list_entry(od->cursor, struct ovl_cache_entry, l_node);
+		if (!p->is_whiteout)
+			if (filler(buf, p->name, p->len, file->f_pos, p->ino, p->type))
+				break;
+		od->cursor = p->l_node.next;
+		file->f_pos++;
+	}
+	return 0;
+}
+
+static loff_t ovl_dir_llseek(struct file *file, loff_t offset, int origin)
+{
+	loff_t res;
+	struct ovl_dir_file *od = file->private_data;
+
+	mutex_lock(&file_inode(file)->i_mutex);
+	if (!file->f_pos)
+		ovl_dir_reset(file);
+
+	if (od->is_real) {
+		res = vfs_llseek(od->realfile, offset, origin);
+		file->f_pos = od->realfile->f_pos;
+	} else {
+		res = -EINVAL;
+
+		switch (origin) {
+		case SEEK_CUR:
+			offset += file->f_pos;
+			break;
+		case SEEK_SET:
+			break;
+		default:
+			goto out_unlock;
+		}
+		if (offset < 0)
+			goto out_unlock;
+
+		if (offset != file->f_pos) {
+			file->f_pos = offset;
+			if (od->cache)
+				ovl_seek_cursor(od, offset);
+		}
+		res = offset;
+	}
+out_unlock:
+	mutex_unlock(&file_inode(file)->i_mutex);
+
+	return res;
+}
+
+static int ovl_dir_fsync(struct file *file, loff_t start, loff_t end,
+			 int datasync)
+{
+	struct ovl_dir_file *od = file->private_data;
+	struct dentry *dentry = file->f_path.dentry;
+	struct file *realfile = od->realfile;
+
+	/*
+	 * Need to check if we started out being a lower dir, but got copied up
+	 */
+	if (!od->is_upper && OVL_TYPE_UPPER(ovl_path_type(dentry))) {
+		struct inode *inode = file_inode(file);
+
+		realfile = lockless_dereference(od->upperfile);
+		if (!realfile) {
+			struct path upperpath;
+
+			ovl_path_upper(dentry, &upperpath);
+			realfile = ovl_path_open(&upperpath, O_RDONLY);
+			smp_mb__before_spinlock();
+			mutex_lock(&inode->i_mutex);
+			if (!od->upperfile) {
+				if (IS_ERR(realfile)) {
+					mutex_unlock(&inode->i_mutex);
+					return PTR_ERR(realfile);
+				}
+				od->upperfile = realfile;
+			} else {
+				/* somebody has beaten us to it */
+				if (!IS_ERR(realfile))
+					fput(realfile);
+				realfile = od->upperfile;
+			}
+			mutex_unlock(&inode->i_mutex);
+		}
+	}
+
+	return vfs_fsync_range(realfile, start, end, datasync);
+}
+
+static int ovl_dir_release(struct inode *inode, struct file *file)
+{
+	struct ovl_dir_file *od = file->private_data;
+
+	if (od->cache) {
+		mutex_lock(&inode->i_mutex);
+		ovl_cache_put(od, file->f_path.dentry);
+		mutex_unlock(&inode->i_mutex);
+	}
+	fput(od->realfile);
+	if (od->upperfile)
+		fput(od->upperfile);
+	kfree(od);
+
+	return 0;
+}
+
+static int ovl_dir_open(struct inode *inode, struct file *file)
+{
+	struct path realpath;
+	struct file *realfile;
+	struct ovl_dir_file *od;
+	enum ovl_path_type type;
+
+	od = kzalloc(sizeof(struct ovl_dir_file), GFP_KERNEL);
+	if (!od)
+		return -ENOMEM;
+
+	type = ovl_path_real(file->f_path.dentry, &realpath);
+	realfile = ovl_path_open(&realpath, file->f_flags);
+	if (IS_ERR(realfile)) {
+		kfree(od);
+		return PTR_ERR(realfile);
+	}
+	od->realfile = realfile;
+	od->is_real = !OVL_TYPE_MERGE(type);
+	od->is_upper = OVL_TYPE_UPPER(type);
+	file->private_data = od;
+
+	return 0;
+}
+
+const struct file_operations ovl_dir_operations = {
+	.read		= generic_read_dir,
+	.open		= ovl_dir_open,
+	.readdir	= ovl_readdir,
+	.llseek		= ovl_dir_llseek,
+	.fsync		= ovl_dir_fsync,
+	.release	= ovl_dir_release,
+};
+
+int ovl_check_empty_dir(struct dentry *dentry, struct list_head *list)
+{
+	int err;
+	struct ovl_cache_entry *p;
+
+	err = ovl_dir_read_merged(dentry, list);
+	if (err)
+		return err;
+
+	err = 0;
+
+	list_for_each_entry(p, list, l_node) {
+		if (p->is_whiteout)
+			continue;
+
+		if (p->name[0] == '.') {
+			if (p->len == 1)
+				continue;
+			if (p->len == 2 && p->name[1] == '.')
+				continue;
+		}
+		err = -ENOTEMPTY;
+		break;
+	}
+
+	return err;
+}
+
+void ovl_cleanup_whiteouts(struct dentry *upper, struct list_head *list)
+{
+	struct ovl_cache_entry *p;
+
+	mutex_lock_nested(&upper->d_inode->i_mutex, I_MUTEX_CHILD);
+	list_for_each_entry(p, list, l_node) {
+		struct dentry *dentry;
+
+		if (!p->is_whiteout)
+			continue;
+
+		dentry = lookup_one_len(p->name, upper, p->len);
+		if (IS_ERR(dentry)) {
+			pr_err("overlayfs: lookup '%s/%.*s' failed (%i)\n",
+			       upper->d_name.name, p->len, p->name,
+			       (int) PTR_ERR(dentry));
+			continue;
+		}
+		ovl_cleanup(upper->d_inode, dentry);
+		dput(dentry);
+	}
+	mutex_unlock(&upper->d_inode->i_mutex);
+}

executer/kernel/mcoverlayfs/linux-4.0.9/Makefile.in

@@ -0,0 +1,21 @@
+KDIR ?= @KDIR@
+ARCH ?= @ARCH@
+KMODDIR = @KMODDIR@
+src = @abs_srcdir@
+
+obj-m += mcoverlay.o
+
+mcoverlay-y := copy_up.o dir.o inode.o readdir.o super.o
+
+.PHONY: clean install modules
+
+modules:
+	$(MAKE) -C $(KDIR) M=$(PWD) SUBDIRS=$(PWD) ARCH=$(ARCH) modules
+
+clean:
+	$(RM) .*.cmd *.mod.c *.o *.ko* Module.symvers modules.order -r .tmp*
+
+install:
+	mkdir -p -m 755 $(KMODDIR)
+	install -m 644 mcoverlay.ko $(KMODDIR)
+

executer/kernel/mcoverlayfs/linux-4.6.7/Makefile.in

@@ -0,0 +1,21 @@
+KDIR ?= @KDIR@
+ARCH ?= @ARCH@
+KMODDIR = @KMODDIR@
+src = @abs_srcdir@
+
+obj-m += mcoverlay.o
+
+mcoverlay-y := copy_up.o dir.o inode.o readdir.o super.o
+
+.PHONY: clean install modules
+
+modules:
+	$(MAKE) -C $(KDIR) M=$(PWD) SUBDIRS=$(PWD) ARCH=$(ARCH) modules
+
+clean:
+	$(RM) .*.cmd *.mod.c *.o *.ko* Module.symvers modules.order -r .tmp*
+
+install:
+	mkdir -p -m 755 $(KMODDIR)
+	install -m 644 mcoverlay.ko $(KMODDIR)
+

executer/kernel/mcoverlayfs/linux-4.6.7/copy_up.c

@@ -0,0 +1,460 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/file.h>
+#include <linux/splice.h>
+#include <linux/xattr.h>
+#include <linux/security.h>
+#include <linux/uaccess.h>
+#include <linux/sched.h>
+#include <linux/namei.h>
+#include <linux/fdtable.h>
+#include <linux/ratelimit.h>
+#include "overlayfs.h"
+
+#define OVL_COPY_UP_CHUNK_SIZE (1 << 20)
+
+static bool __read_mostly ovl_check_copy_up;
+module_param_named(check_copy_up, ovl_check_copy_up, bool,
+		   S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(ovl_check_copy_up,
+		 "Warn on copy-up when causing process also has a R/O fd open");
+
+static int ovl_check_fd(const void *data, struct file *f, unsigned int fd)
+{
+	const struct dentry *dentry = data;
+
+	if (f->f_inode == d_inode(dentry))
+		pr_warn_ratelimited("overlayfs: Warning: Copying up %pD, but open R/O on fd %u which will cease to be coherent [pid=%d %s]\n",
+				    f, fd, current->pid, current->comm);
+	return 0;
+}
+
+/*
+ * Check the fds open by this process and warn if something like the following
+ * scenario is about to occur:
+ *
+ *	fd1 = open("foo", O_RDONLY);
+ *	fd2 = open("foo", O_RDWR);
+ */
+static void ovl_do_check_copy_up(struct dentry *dentry)
+{
+	if (ovl_check_copy_up)
+		iterate_fd(current->files, 0, ovl_check_fd, dentry);
+}
+
+int ovl_copy_xattr(struct dentry *old, struct dentry *new, unsigned opt)
+{
+	ssize_t list_size, size, value_size = 0;
+	char *buf, *name, *value = NULL;
+	int uninitialized_var(error);
+
+	if (!old->d_inode->i_op->getxattr ||
+	    !new->d_inode->i_op->getxattr)
+		return 0;
+
+	list_size = vfs_listxattr(old, NULL, 0);
+	if (list_size <= 0) {
+		if (list_size == -EOPNOTSUPP)
+			return 0;
+		return list_size;
+	}
+
+	buf = kzalloc(list_size, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	list_size = vfs_listxattr(old, buf, list_size);
+	if (list_size <= 0) {
+		error = list_size;
+		goto out;
+	}
+
+	for (name = buf; name < (buf + list_size); name += strlen(name) + 1) {
+retry:
+		size = vfs_getxattr(old, name, value, value_size);
+		if (size == -ERANGE)
+			size = vfs_getxattr(old, name, NULL, 0);
+
+		if (size < 0) {
+			if (OVL_OPT_NOFSCHECK(opt)) {
+				OVL_DEBUG("fail: old=%pd4, i_ino=%lu, name=%s\n", 
+					old, old->d_inode->i_ino, name);
+				continue;
+			} else {
+				error = size;
+				break;
+			}
+		}
+		OVL_DEBUG("success: old=%pd4, i_ino=%lu, name=%s\n", 
+			old, old->d_inode->i_ino, name);
+
+		if (size > value_size) {
+			void *new;
+
+			new = krealloc(value, size, GFP_KERNEL);
+			if (!new) {
+				error = -ENOMEM;
+				break;
+			}
+			value = new;
+			value_size = size;
+			goto retry;
+		}
+
+		error = vfs_setxattr(new, name, value, size, 0);
+		if (error)
+			break;
+	}
+	kfree(value);
+out:
+	kfree(buf);
+	return error;
+}
+
+static int ovl_copy_up_data(struct path *old, struct path *new, loff_t len)
+{
+	struct file *old_file;
+	struct file *new_file;
+	loff_t old_pos = 0;
+	loff_t new_pos = 0;
+	int error = 0;
+
+	if (len == 0)
+		return 0;
+
+	old_file = ovl_path_open(old, O_LARGEFILE | O_RDONLY);
+	if (IS_ERR(old_file))
+		return PTR_ERR(old_file);
+
+	new_file = ovl_path_open(new, O_LARGEFILE | O_WRONLY);
+	if (IS_ERR(new_file)) {
+		error = PTR_ERR(new_file);
+		goto out_fput;
+	}
+
+	/* FIXME: copy up sparse files efficiently */
+	while (len) {
+		size_t this_len = OVL_COPY_UP_CHUNK_SIZE;
+		long bytes;
+
+		if (len < this_len)
+			this_len = len;
+
+		if (signal_pending_state(TASK_KILLABLE, current)) {
+			error = -EINTR;
+			break;
+		}
+
+		bytes = do_splice_direct(old_file, &old_pos,
+					 new_file, &new_pos,
+					 this_len, SPLICE_F_MOVE);
+		if (bytes <= 0) {
+			error = bytes;
+			break;
+		}
+		WARN_ON(old_pos != new_pos);
+
+		len -= bytes;
+	}
+
+	fput(new_file);
+out_fput:
+	fput(old_file);
+	return error;
+}
+
+static char *ovl_read_symlink(struct dentry *realdentry)
+{
+	int res;
+	char *buf;
+	struct inode *inode = realdentry->d_inode;
+	mm_segment_t old_fs;
+
+	res = -EINVAL;
+	if (!inode->i_op->readlink)
+		goto err;
+
+	res = -ENOMEM;
+	buf = (char *) __get_free_page(GFP_KERNEL);
+	if (!buf)
+		goto err;
+
+	old_fs = get_fs();
+	set_fs(get_ds());
+	/* The cast to a user pointer is valid due to the set_fs() */
+	res = inode->i_op->readlink(realdentry,
+				    (char __user *)buf, PAGE_SIZE - 1);
+	set_fs(old_fs);
+	if (res < 0) {
+		free_page((unsigned long) buf);
+		goto err;
+	}
+	buf[res] = '\0';
+
+	return buf;
+
+err:
+	return ERR_PTR(res);
+}
+
+static int ovl_set_timestamps(struct dentry *upperdentry, struct kstat *stat)
+{
+	struct iattr attr = {
+		.ia_valid =
+		     ATTR_ATIME | ATTR_MTIME | ATTR_ATIME_SET | ATTR_MTIME_SET,
+		.ia_atime = stat->atime,
+		.ia_mtime = stat->mtime,
+	};
+
+	return notify_change(upperdentry, &attr, NULL);
+}
+
+int ovl_set_attr(struct dentry *upperdentry, struct kstat *stat)
+{
+	int err = 0;
+
+	if (!S_ISLNK(stat->mode)) {
+		struct iattr attr = {
+			.ia_valid = ATTR_MODE,
+			.ia_mode = stat->mode,
+		};
+		err = notify_change(upperdentry, &attr, NULL);
+	}
+	if (!err) {
+		struct iattr attr = {
+			.ia_valid = ATTR_UID | ATTR_GID,
+			.ia_uid = stat->uid,
+			.ia_gid = stat->gid,
+		};
+		err = notify_change(upperdentry, &attr, NULL);
+	}
+	if (!err)
+		ovl_set_timestamps(upperdentry, stat);
+
+	return err;
+}
+
+static int ovl_copy_up_locked(struct dentry *workdir, struct dentry *upperdir,
+			      struct dentry *dentry, struct path *lowerpath,
+			      struct kstat *stat, const char *link)
+{
+	struct inode *wdir = workdir->d_inode;
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *newdentry = NULL;
+	struct dentry *upper = NULL;
+	umode_t mode = stat->mode;
+	unsigned opt = ovl_get_config_opt(dentry);
+	int err;
+
+	newdentry = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out;
+
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out1;
+
+	/* Can't properly set mode on creation because of the umask */
+	stat->mode &= S_IFMT;
+	err = ovl_create_real(wdir, newdentry, stat, link, NULL, true);
+	stat->mode = mode;
+	if (err)
+		goto out2;
+
+	if (S_ISREG(stat->mode)) {
+		struct path upperpath;
+
+		ovl_path_upper(dentry, &upperpath);
+		BUG_ON(upperpath.dentry != NULL);
+		upperpath.dentry = newdentry;
+
+		err = ovl_copy_up_data(lowerpath, &upperpath, stat->size);
+		if (err)
+			goto out_cleanup;
+	}
+
+	err = ovl_copy_xattr(lowerpath->dentry, newdentry, opt);
+	if (err)
+		goto out_cleanup;
+
+	inode_lock(newdentry->d_inode);
+	err = ovl_set_attr(newdentry, stat);
+	inode_unlock(newdentry->d_inode);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_do_rename(wdir, newdentry, udir, upper, 0);
+	if (err)
+		goto out_cleanup;
+
+	ovl_dentry_update(dentry, newdentry);
+	newdentry = NULL;
+
+	/*
+	 * Non-directores become opaque when copied up.
+	 */
+	if (!S_ISDIR(stat->mode))
+		ovl_dentry_set_opaque(dentry, true);
+out2:
+	dput(upper);
+out1:
+	dput(newdentry);
+out:
+	return err;
+
+out_cleanup:
+	ovl_cleanup(wdir, newdentry);
+	goto out2;
+}
+
+/*
+ * Copy up a single dentry
+ *
+ * Directory renames only allowed on "pure upper" (already created on
+ * upper filesystem, never copied up).  Directories which are on lower or
+ * are merged may not be renamed.  For these -EXDEV is returned and
+ * userspace has to deal with it.  This means, when copying up a
+ * directory we can rely on it and ancestors being stable.
+ *
+ * Non-directory renames start with copy up of source if necessary.  The
+ * actual rename will only proceed once the copy up was successful.  Copy
+ * up uses upper parent i_mutex for exclusion.  Since rename can change
+ * d_parent it is possible that the copy up will lock the old parent.  At
+ * that point the file will have already been copied up anyway.
+ */
+int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry,
+		    struct path *lowerpath, struct kstat *stat)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	int err;
+	struct kstat pstat;
+	struct path parentpath;
+	struct dentry *upperdir;
+	struct dentry *upperdentry;
+	const struct cred *old_cred;
+	struct cred *override_cred;
+	char *link = NULL;
+
+	if (WARN_ON(!workdir))
+		return -EROFS;
+
+	ovl_do_check_copy_up(lowerpath->dentry);
+
+	ovl_path_upper(parent, &parentpath);
+	upperdir = parentpath.dentry;
+
+	err = vfs_getattr(&parentpath, &pstat);
+	if (err)
+		return err;
+
+	if (S_ISLNK(stat->mode)) {
+		link = ovl_read_symlink(lowerpath->dentry);
+		if (IS_ERR(link))
+			return PTR_ERR(link);
+	}
+
+	err = -ENOMEM;
+	override_cred = prepare_creds();
+	if (!override_cred)
+		goto out_free_link;
+
+	override_cred->fsuid = stat->uid;
+	override_cred->fsgid = stat->gid;
+	/*
+	 * CAP_SYS_ADMIN for copying up extended attributes
+	 * CAP_DAC_OVERRIDE for create
+	 * CAP_FOWNER for chmod, timestamp update
+	 * CAP_FSETID for chmod
+	 * CAP_CHOWN for chown
+	 * CAP_MKNOD for mknod
+	 */
+	cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+	cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+	cap_raise(override_cred->cap_effective, CAP_FOWNER);
+	cap_raise(override_cred->cap_effective, CAP_FSETID);
+	cap_raise(override_cred->cap_effective, CAP_CHOWN);
+	cap_raise(override_cred->cap_effective, CAP_MKNOD);
+	old_cred = override_creds(override_cred);
+
+	err = -EIO;
+	if (lock_rename(workdir, upperdir) != NULL) {
+		pr_err("overlayfs: failed to lock workdir+upperdir\n");
+		goto out_unlock;
+	}
+	upperdentry = ovl_dentry_upper(dentry);
+	if (upperdentry) {
+		/* Raced with another copy-up?  Nothing to do, then... */
+		err = 0;
+		goto out_unlock;
+	}
+
+	err = ovl_copy_up_locked(workdir, upperdir, dentry, lowerpath,
+				 stat, link);
+	if (!err) {
+		/* Restore timestamps on parent (best effort) */
+		ovl_set_timestamps(upperdir, &pstat);
+	}
+out_unlock:
+	unlock_rename(workdir, upperdir);
+	revert_creds(old_cred);
+	put_cred(override_cred);
+
+out_free_link:
+	if (link)
+		free_page((unsigned long) link);
+
+	return err;
+}
+
+int ovl_copy_up(struct dentry *dentry)
+{
+	int err;
+
+	err = 0;
+	while (!err) {
+		struct dentry *next;
+		struct dentry *parent;
+		struct path lowerpath;
+		struct kstat stat;
+		enum ovl_path_type type = ovl_path_type(dentry);
+
+		if (OVL_TYPE_UPPER(type))
+			break;
+
+		next = dget(dentry);
+		/* find the topmost dentry not yet copied up */
+		for (;;) {
+			parent = dget_parent(next);
+
+			type = ovl_path_type(parent);
+			if (OVL_TYPE_UPPER(type))
+				break;
+
+			dput(next);
+			next = parent;
+		}
+
+		ovl_path_lower(next, &lowerpath);
+		err = vfs_getattr(&lowerpath, &stat);
+		if (!err)
+			err = ovl_copy_up_one(parent, next, &lowerpath, &stat);
+
+		dput(parent);
+		dput(next);
+	}
+
+	return err;
+}

executer/kernel/mcoverlayfs/linux-4.6.7/dir.c

@@ -0,0 +1,969 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/fs.h>
+#include <linux/namei.h>
+#include <linux/xattr.h>
+#include <linux/security.h>
+#include <linux/cred.h>
+#include "overlayfs.h"
+
+void ovl_cleanup(struct inode *wdir, struct dentry *wdentry)
+{
+	int err;
+
+	dget(wdentry);
+	if (d_is_dir(wdentry))
+		err = ovl_do_rmdir(wdir, wdentry);
+	else
+		err = ovl_do_unlink(wdir, wdentry);
+	dput(wdentry);
+
+	if (err) {
+		pr_err("overlayfs: cleanup of '%pd2' failed (%i)\n",
+		       wdentry, err);
+	}
+}
+
+struct dentry *ovl_lookup_temp(struct dentry *workdir, struct dentry *dentry)
+{
+	struct dentry *temp;
+	char name[20];
+
+	snprintf(name, sizeof(name), "#%lx", (unsigned long) dentry);
+
+	temp = lookup_one_len(name, workdir, strlen(name));
+	if (!IS_ERR(temp) && temp->d_inode) {
+		pr_err("overlayfs: workdir/%s already exists\n", name);
+		dput(temp);
+		temp = ERR_PTR(-EIO);
+	}
+
+	return temp;
+}
+
+/* caller holds i_mutex on workdir */
+static struct dentry *ovl_whiteout(struct dentry *workdir,
+				   struct dentry *dentry)
+{
+	int err;
+	struct dentry *whiteout;
+	struct inode *wdir = workdir->d_inode;
+
+	whiteout = ovl_lookup_temp(workdir, dentry);
+	if (IS_ERR(whiteout))
+		return whiteout;
+
+	err = ovl_do_whiteout(wdir, whiteout);
+	if (err) {
+		dput(whiteout);
+		whiteout = ERR_PTR(err);
+	}
+
+	return whiteout;
+}
+
+int ovl_create_real(struct inode *dir, struct dentry *newdentry,
+		    struct kstat *stat, const char *link,
+		    struct dentry *hardlink, bool debug)
+{
+	int err;
+
+	if (newdentry->d_inode)
+		return -ESTALE;
+
+	if (hardlink) {
+		err = ovl_do_link(hardlink, dir, newdentry, debug);
+	} else {
+		switch (stat->mode & S_IFMT) {
+		case S_IFREG:
+			err = ovl_do_create(dir, newdentry, stat->mode, debug);
+			break;
+
+		case S_IFDIR:
+			err = ovl_do_mkdir(dir, newdentry, stat->mode, debug);
+			break;
+
+		case S_IFCHR:
+		case S_IFBLK:
+		case S_IFIFO:
+		case S_IFSOCK:
+			err = ovl_do_mknod(dir, newdentry,
+					   stat->mode, stat->rdev, debug);
+			break;
+
+		case S_IFLNK:
+			err = ovl_do_symlink(dir, newdentry, link, debug);
+			break;
+
+		default:
+			err = -EPERM;
+		}
+	}
+	if (!err && WARN_ON(!newdentry->d_inode)) {
+		/*
+		 * Not quite sure if non-instantiated dentry is legal or not.
+		 * VFS doesn't seem to care so check and warn here.
+		 */
+		err = -ENOENT;
+	}
+	return err;
+}
+
+static int ovl_set_opaque(struct dentry *upperdentry)
+{
+	return ovl_do_setxattr(upperdentry, OVL_XATTR_OPAQUE, "y", 1, 0);
+}
+
+static void ovl_remove_opaque(struct dentry *upperdentry)
+{
+	int err;
+
+	err = ovl_do_removexattr(upperdentry, OVL_XATTR_OPAQUE);
+	if (err) {
+		pr_warn("overlayfs: failed to remove opaque from '%s' (%i)\n",
+			upperdentry->d_name.name, err);
+	}
+}
+
+static int ovl_dir_getattr(struct vfsmount *mnt, struct dentry *dentry,
+			 struct kstat *stat)
+{
+	int err;
+	enum ovl_path_type type;
+	struct path realpath;
+
+	type = ovl_path_real(dentry, &realpath);
+	err = vfs_getattr(&realpath, stat);
+	if (err)
+		return err;
+
+	stat->dev = dentry->d_sb->s_dev;
+	stat->ino = dentry->d_inode->i_ino;
+
+	/*
+	 * It's probably not worth it to count subdirs to get the
+	 * correct link count.  nlink=1 seems to pacify 'find' and
+	 * other utilities.
+	 */
+	if (OVL_TYPE_MERGE(type))
+		stat->nlink = 1;
+
+	return 0;
+}
+
+static int ovl_create_upper(struct dentry *dentry, struct inode *inode,
+			    struct kstat *stat, const char *link,
+			    struct dentry *hardlink)
+{
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *newdentry;
+	int err;
+
+	inode_lock_nested(udir, I_MUTEX_PARENT);
+	newdentry = lookup_one_len(dentry->d_name.name, upperdir,
+				   dentry->d_name.len);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out_unlock;
+	err = ovl_create_real(udir, newdentry, stat, link, hardlink, false);
+	if (err)
+		goto out_dput;
+
+	ovl_dentry_version_inc(dentry->d_parent);
+	ovl_dentry_update(dentry, newdentry);
+	ovl_copyattr(newdentry->d_inode, inode);
+	d_instantiate(dentry, inode);
+	newdentry = NULL;
+out_dput:
+	dput(newdentry);
+out_unlock:
+	inode_unlock(udir);
+	return err;
+}
+
+static int ovl_lock_rename_workdir(struct dentry *workdir,
+				   struct dentry *upperdir)
+{
+	/* Workdir should not be the same as upperdir */
+	if (workdir == upperdir)
+		goto err;
+
+	/* Workdir should not be subdir of upperdir and vice versa */
+	if (lock_rename(workdir, upperdir) != NULL)
+		goto err_unlock;
+
+	return 0;
+
+err_unlock:
+	unlock_rename(workdir, upperdir);
+err:
+	pr_err("overlayfs: failed to lock workdir+upperdir\n");
+	return -EIO;
+}
+
+static struct dentry *ovl_clear_empty(struct dentry *dentry,
+				      struct list_head *list)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct path upperpath;
+	struct dentry *upper;
+	struct dentry *opaquedir;
+	struct kstat stat;
+	unsigned opt = ovl_get_config_opt(dentry);
+	int err;
+
+	if (WARN_ON(!workdir))
+		return ERR_PTR(-EROFS);
+
+	err = ovl_lock_rename_workdir(workdir, upperdir);
+	if (err)
+		goto out;
+
+	ovl_path_upper(dentry, &upperpath);
+	err = vfs_getattr(&upperpath, &stat);
+	if (err)
+		goto out_unlock;
+
+	err = -ESTALE;
+	if (!S_ISDIR(stat.mode))
+		goto out_unlock;
+	upper = upperpath.dentry;
+	if (upper->d_parent->d_inode != udir)
+		goto out_unlock;
+
+	opaquedir = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(opaquedir);
+	if (IS_ERR(opaquedir))
+		goto out_unlock;
+
+	err = ovl_create_real(wdir, opaquedir, &stat, NULL, NULL, true);
+	if (err)
+		goto out_dput;
+
+	err = ovl_copy_xattr(upper, opaquedir, opt);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_set_opaque(opaquedir);
+	if (err)
+		goto out_cleanup;
+
+	inode_lock(opaquedir->d_inode);
+	err = ovl_set_attr(opaquedir, &stat);
+	inode_unlock(opaquedir->d_inode);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_do_rename(wdir, opaquedir, udir, upper, RENAME_EXCHANGE);
+	if (err)
+		goto out_cleanup;
+
+	ovl_cleanup_whiteouts(upper, list);
+	ovl_cleanup(wdir, upper);
+	unlock_rename(workdir, upperdir);
+
+	/* dentry's upper doesn't match now, get rid of it */
+	d_drop(dentry);
+
+	return opaquedir;
+
+out_cleanup:
+	ovl_cleanup(wdir, opaquedir);
+out_dput:
+	dput(opaquedir);
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out:
+	return ERR_PTR(err);
+}
+
+static struct dentry *ovl_check_empty_and_clear(struct dentry *dentry)
+{
+	int err;
+	struct dentry *ret = NULL;
+	LIST_HEAD(list);
+
+	err = ovl_check_empty_dir(dentry, &list);
+	if (err)
+		ret = ERR_PTR(err);
+	else {
+		/*
+		 * If no upperdentry then skip clearing whiteouts.
+		 *
+		 * Can race with copy-up, since we don't hold the upperdir
+		 * mutex.  Doesn't matter, since copy-up can't create a
+		 * non-empty directory from an empty one.
+		 */
+		if (ovl_dentry_upper(dentry))
+			ret = ovl_clear_empty(dentry, &list);
+	}
+
+	ovl_cache_free(&list);
+
+	return ret;
+}
+
+static int ovl_create_over_whiteout(struct dentry *dentry, struct inode *inode,
+				    struct kstat *stat, const char *link,
+				    struct dentry *hardlink)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *upper;
+	struct dentry *newdentry;
+	int err;
+
+	if (WARN_ON(!workdir))
+		return -EROFS;
+
+	err = ovl_lock_rename_workdir(workdir, upperdir);
+	if (err)
+		goto out;
+
+	newdentry = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out_unlock;
+
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out_dput;
+
+	err = ovl_create_real(wdir, newdentry, stat, link, hardlink, true);
+	if (err)
+		goto out_dput2;
+
+	if (S_ISDIR(stat->mode)) {
+		err = ovl_set_opaque(newdentry);
+		if (err)
+			goto out_cleanup;
+
+		err = ovl_do_rename(wdir, newdentry, udir, upper,
+				    RENAME_EXCHANGE);
+		if (err)
+			goto out_cleanup;
+
+		ovl_cleanup(wdir, upper);
+	} else {
+		err = ovl_do_rename(wdir, newdentry, udir, upper, 0);
+		if (err)
+			goto out_cleanup;
+	}
+	ovl_dentry_version_inc(dentry->d_parent);
+	ovl_dentry_update(dentry, newdentry);
+	ovl_copyattr(newdentry->d_inode, inode);
+	d_instantiate(dentry, inode);
+	newdentry = NULL;
+out_dput2:
+	dput(upper);
+out_dput:
+	dput(newdentry);
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out:
+	return err;
+
+out_cleanup:
+	ovl_cleanup(wdir, newdentry);
+	goto out_dput2;
+}
+
+static int ovl_create_or_link(struct dentry *dentry, int mode, dev_t rdev,
+			      const char *link, struct dentry *hardlink)
+{
+	int err;
+	struct inode *inode;
+	struct kstat stat = {
+		.mode = mode,
+		.rdev = rdev,
+	};
+
+	err = -ENOMEM;
+	inode = ovl_new_inode(dentry->d_sb, mode, dentry->d_fsdata);
+	if (!inode)
+		goto out;
+
+	err = ovl_copy_up(dentry->d_parent);
+	if (err)
+		goto out_iput;
+
+	if (!ovl_dentry_is_opaque(dentry)) {
+		err = ovl_create_upper(dentry, inode, &stat, link, hardlink);
+	} else {
+		const struct cred *old_cred;
+		struct cred *override_cred;
+
+		err = -ENOMEM;
+		override_cred = prepare_creds();
+		if (!override_cred)
+			goto out_iput;
+
+		/*
+		 * CAP_SYS_ADMIN for setting opaque xattr
+		 * CAP_DAC_OVERRIDE for create in workdir, rename
+		 * CAP_FOWNER for removing whiteout from sticky dir
+		 */
+		cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		cap_raise(override_cred->cap_effective, CAP_FOWNER);
+		old_cred = override_creds(override_cred);
+
+		err = ovl_create_over_whiteout(dentry, inode, &stat, link,
+					       hardlink);
+
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+
+	if (!err)
+		inode = NULL;
+out_iput:
+	iput(inode);
+out:
+	return err;
+}
+
+static int ovl_create_object(struct dentry *dentry, int mode, dev_t rdev,
+			     const char *link)
+{
+	int err;
+
+	err = ovl_want_write(dentry);
+	if (!err) {
+		err = ovl_create_or_link(dentry, mode, rdev, link, NULL);
+		ovl_drop_write(dentry);
+	}
+
+	return err;
+}
+
+static int ovl_create(struct inode *dir, struct dentry *dentry, umode_t mode,
+		      bool excl)
+{
+	return ovl_create_object(dentry, (mode & 07777) | S_IFREG, 0, NULL);
+}
+
+static int ovl_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
+{
+	return ovl_create_object(dentry, (mode & 07777) | S_IFDIR, 0, NULL);
+}
+
+static int ovl_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
+		     dev_t rdev)
+{
+	/* Don't allow creation of "whiteout" on overlay */
+	if (S_ISCHR(mode) && rdev == WHITEOUT_DEV)
+		return -EPERM;
+
+	return ovl_create_object(dentry, mode, rdev, NULL);
+}
+
+static int ovl_symlink(struct inode *dir, struct dentry *dentry,
+		       const char *link)
+{
+	return ovl_create_object(dentry, S_IFLNK, 0, link);
+}
+
+static int ovl_link(struct dentry *old, struct inode *newdir,
+		    struct dentry *new)
+{
+	int err;
+	struct dentry *upper;
+
+	err = ovl_want_write(old);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(old);
+	if (err)
+		goto out_drop_write;
+
+	upper = ovl_dentry_upper(old);
+	err = ovl_create_or_link(new, upper->d_inode->i_mode, 0, NULL, upper);
+
+out_drop_write:
+	ovl_drop_write(old);
+out:
+	return err;
+}
+
+static int ovl_remove_and_whiteout(struct dentry *dentry, bool is_dir)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *whiteout;
+	struct dentry *upper;
+	struct dentry *opaquedir = NULL;
+	int err;
+	int flags = 0;
+
+	if (WARN_ON(!workdir))
+		return -EROFS;
+
+	if (is_dir) {
+		if (OVL_TYPE_MERGE_OR_LOWER(ovl_path_type(dentry))) {
+			opaquedir = ovl_check_empty_and_clear(dentry);
+			err = PTR_ERR(opaquedir);
+			if (IS_ERR(opaquedir))
+				goto out;
+		} else {
+			LIST_HEAD(list);
+
+			/*
+			 * When removing an empty opaque directory, then it
+			 * makes no sense to replace it with an exact replica of
+			 * itself.  But emptiness still needs to be checked.
+			 */
+			err = ovl_check_empty_dir(dentry, &list);
+			ovl_cache_free(&list);
+			if (err)
+				goto out;
+		}
+	}
+
+	err = ovl_lock_rename_workdir(workdir, upperdir);
+	if (err)
+		goto out_dput;
+
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out_unlock;
+
+	err = -ESTALE;
+	if ((opaquedir && upper != opaquedir) ||
+	    (!opaquedir && ovl_dentry_upper(dentry) &&
+	     upper != ovl_dentry_upper(dentry))) {
+		goto out_dput_upper;
+	}
+
+	whiteout = ovl_whiteout(workdir, dentry);
+	err = PTR_ERR(whiteout);
+	if (IS_ERR(whiteout))
+		goto out_dput_upper;
+
+	if (d_is_dir(upper))
+		flags = RENAME_EXCHANGE;
+
+	err = ovl_do_rename(wdir, whiteout, udir, upper, flags);
+	if (err)
+		goto kill_whiteout;
+	if (flags)
+		ovl_cleanup(wdir, upper);
+
+	ovl_dentry_version_inc(dentry->d_parent);
+out_d_drop:
+	d_drop(dentry);
+	dput(whiteout);
+out_dput_upper:
+	dput(upper);
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out_dput:
+	dput(opaquedir);
+out:
+	return err;
+
+kill_whiteout:
+	ovl_cleanup(wdir, whiteout);
+	goto out_d_drop;
+}
+
+static int ovl_remove_upper(struct dentry *dentry, bool is_dir)
+{
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *dir = upperdir->d_inode;
+	struct dentry *upper;
+	int err;
+
+	inode_lock_nested(dir, I_MUTEX_PARENT);
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out_unlock;
+
+	err = -ESTALE;
+	if (upper == ovl_dentry_upper(dentry)) {
+		if (is_dir)
+			err = vfs_rmdir(dir, upper);
+		else
+			err = vfs_unlink(dir, upper, NULL);
+		ovl_dentry_version_inc(dentry->d_parent);
+	}
+	dput(upper);
+
+	/*
+	 * Keeping this dentry hashed would mean having to release
+	 * upperpath/lowerpath, which could only be done if we are the
+	 * sole user of this dentry.  Too tricky...  Just unhash for
+	 * now.
+	 */
+	if (!err)
+		d_drop(dentry);
+out_unlock:
+	inode_unlock(dir);
+
+	return err;
+}
+
+static inline int ovl_check_sticky(struct dentry *dentry)
+{
+	struct inode *dir = ovl_dentry_real(dentry->d_parent)->d_inode;
+	struct inode *inode = ovl_dentry_real(dentry)->d_inode;
+
+	if (check_sticky(dir, inode))
+		return -EPERM;
+
+	return 0;
+}
+
+static int ovl_do_remove(struct dentry *dentry, bool is_dir)
+{
+	enum ovl_path_type type;
+	int err;
+
+	err = ovl_check_sticky(dentry);
+	if (err)
+		goto out;
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(dentry->d_parent);
+	if (err)
+		goto out_drop_write;
+
+	type = ovl_path_type(dentry);
+	if (OVL_TYPE_PURE_UPPER(type)) {
+		err = ovl_remove_upper(dentry, is_dir);
+	} else {
+		const struct cred *old_cred;
+		struct cred *override_cred;
+
+		err = -ENOMEM;
+		override_cred = prepare_creds();
+		if (!override_cred)
+			goto out_drop_write;
+
+		/*
+		 * CAP_SYS_ADMIN for setting xattr on whiteout, opaque dir
+		 * CAP_DAC_OVERRIDE for create in workdir, rename
+		 * CAP_FOWNER for removing whiteout from sticky dir
+		 * CAP_FSETID for chmod of opaque dir
+		 * CAP_CHOWN for chown of opaque dir
+		 */
+		cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		cap_raise(override_cred->cap_effective, CAP_FOWNER);
+		cap_raise(override_cred->cap_effective, CAP_FSETID);
+		cap_raise(override_cred->cap_effective, CAP_CHOWN);
+		old_cred = override_creds(override_cred);
+
+		err = ovl_remove_and_whiteout(dentry, is_dir);
+
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static int ovl_unlink(struct inode *dir, struct dentry *dentry)
+{
+	return ovl_do_remove(dentry, false);
+}
+
+static int ovl_rmdir(struct inode *dir, struct dentry *dentry)
+{
+	return ovl_do_remove(dentry, true);
+}
+
+static int ovl_rename2(struct inode *olddir, struct dentry *old,
+		       struct inode *newdir, struct dentry *new,
+		       unsigned int flags)
+{
+	int err;
+	enum ovl_path_type old_type;
+	enum ovl_path_type new_type;
+	struct dentry *old_upperdir;
+	struct dentry *new_upperdir;
+	struct dentry *olddentry;
+	struct dentry *newdentry;
+	struct dentry *trap;
+	bool old_opaque;
+	bool new_opaque;
+	bool cleanup_whiteout = false;
+	bool overwrite = !(flags & RENAME_EXCHANGE);
+	bool is_dir = d_is_dir(old);
+	bool new_is_dir = false;
+	struct dentry *opaquedir = NULL;
+	const struct cred *old_cred = NULL;
+	struct cred *override_cred = NULL;
+
+	err = -EINVAL;
+	if (flags & ~(RENAME_EXCHANGE | RENAME_NOREPLACE))
+		goto out;
+
+	flags &= ~RENAME_NOREPLACE;
+
+	err = ovl_check_sticky(old);
+	if (err)
+		goto out;
+
+	/* Don't copy up directory trees */
+	old_type = ovl_path_type(old);
+	err = -EXDEV;
+	if (OVL_TYPE_MERGE_OR_LOWER(old_type) && is_dir)
+		goto out;
+
+	if (new->d_inode) {
+		err = ovl_check_sticky(new);
+		if (err)
+			goto out;
+
+		if (d_is_dir(new))
+			new_is_dir = true;
+
+		new_type = ovl_path_type(new);
+		err = -EXDEV;
+		if (!overwrite && OVL_TYPE_MERGE_OR_LOWER(new_type) && new_is_dir)
+			goto out;
+
+		err = 0;
+		if (!OVL_TYPE_UPPER(new_type) && !OVL_TYPE_UPPER(old_type)) {
+			if (ovl_dentry_lower(old)->d_inode ==
+			    ovl_dentry_lower(new)->d_inode)
+				goto out;
+		}
+		if (OVL_TYPE_UPPER(new_type) && OVL_TYPE_UPPER(old_type)) {
+			if (ovl_dentry_upper(old)->d_inode ==
+			    ovl_dentry_upper(new)->d_inode)
+				goto out;
+		}
+	} else {
+		if (ovl_dentry_is_opaque(new))
+			new_type = __OVL_PATH_UPPER;
+		else
+			new_type = __OVL_PATH_UPPER | __OVL_PATH_PURE;
+	}
+
+	err = ovl_want_write(old);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(old);
+	if (err)
+		goto out_drop_write;
+
+	err = ovl_copy_up(new->d_parent);
+	if (err)
+		goto out_drop_write;
+	if (!overwrite) {
+		err = ovl_copy_up(new);
+		if (err)
+			goto out_drop_write;
+	}
+
+	old_opaque = !OVL_TYPE_PURE_UPPER(old_type);
+	new_opaque = !OVL_TYPE_PURE_UPPER(new_type);
+
+	if (old_opaque || new_opaque) {
+		err = -ENOMEM;
+		override_cred = prepare_creds();
+		if (!override_cred)
+			goto out_drop_write;
+
+		/*
+		 * CAP_SYS_ADMIN for setting xattr on whiteout, opaque dir
+		 * CAP_DAC_OVERRIDE for create in workdir
+		 * CAP_FOWNER for removing whiteout from sticky dir
+		 * CAP_FSETID for chmod of opaque dir
+		 * CAP_CHOWN for chown of opaque dir
+		 */
+		cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		cap_raise(override_cred->cap_effective, CAP_FOWNER);
+		cap_raise(override_cred->cap_effective, CAP_FSETID);
+		cap_raise(override_cred->cap_effective, CAP_CHOWN);
+		old_cred = override_creds(override_cred);
+	}
+
+	if (overwrite && OVL_TYPE_MERGE_OR_LOWER(new_type) && new_is_dir) {
+		opaquedir = ovl_check_empty_and_clear(new);
+		err = PTR_ERR(opaquedir);
+		if (IS_ERR(opaquedir)) {
+			opaquedir = NULL;
+			goto out_revert_creds;
+		}
+	}
+
+	if (overwrite) {
+		if (old_opaque) {
+			if (new->d_inode || !new_opaque) {
+				/* Whiteout source */
+				flags |= RENAME_WHITEOUT;
+			} else {
+				/* Switch whiteouts */
+				flags |= RENAME_EXCHANGE;
+			}
+		} else if (is_dir && !new->d_inode && new_opaque) {
+			flags |= RENAME_EXCHANGE;
+			cleanup_whiteout = true;
+		}
+	}
+
+	old_upperdir = ovl_dentry_upper(old->d_parent);
+	new_upperdir = ovl_dentry_upper(new->d_parent);
+
+	trap = lock_rename(new_upperdir, old_upperdir);
+
+
+	olddentry = lookup_one_len(old->d_name.name, old_upperdir,
+				   old->d_name.len);
+	err = PTR_ERR(olddentry);
+	if (IS_ERR(olddentry))
+		goto out_unlock;
+
+	err = -ESTALE;
+	if (olddentry != ovl_dentry_upper(old))
+		goto out_dput_old;
+
+	newdentry = lookup_one_len(new->d_name.name, new_upperdir,
+				   new->d_name.len);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out_dput_old;
+
+	err = -ESTALE;
+	if (ovl_dentry_upper(new)) {
+		if (opaquedir) {
+			if (newdentry != opaquedir)
+				goto out_dput;
+		} else {
+			if (newdentry != ovl_dentry_upper(new))
+				goto out_dput;
+		}
+	} else {
+		if (!d_is_negative(newdentry) &&
+		    (!new_opaque || !ovl_is_whiteout(newdentry)))
+			goto out_dput;
+	}
+
+	if (olddentry == trap)
+		goto out_dput;
+	if (newdentry == trap)
+		goto out_dput;
+
+	if (is_dir && !old_opaque && new_opaque) {
+		err = ovl_set_opaque(olddentry);
+		if (err)
+			goto out_dput;
+	}
+	if (!overwrite && new_is_dir && old_opaque && !new_opaque) {
+		err = ovl_set_opaque(newdentry);
+		if (err)
+			goto out_dput;
+	}
+
+	if (old_opaque || new_opaque) {
+		err = ovl_do_rename(old_upperdir->d_inode, olddentry,
+				    new_upperdir->d_inode, newdentry,
+				    flags);
+	} else {
+		/* No debug for the plain case */
+		BUG_ON(flags & ~RENAME_EXCHANGE);
+		err = vfs_rename(old_upperdir->d_inode, olddentry,
+				 new_upperdir->d_inode, newdentry,
+				 NULL, flags);
+	}
+
+	if (err) {
+		if (is_dir && !old_opaque && new_opaque)
+			ovl_remove_opaque(olddentry);
+		if (!overwrite && new_is_dir && old_opaque && !new_opaque)
+			ovl_remove_opaque(newdentry);
+		goto out_dput;
+	}
+
+	if (is_dir && old_opaque && !new_opaque)
+		ovl_remove_opaque(olddentry);
+	if (!overwrite && new_is_dir && !old_opaque && new_opaque)
+		ovl_remove_opaque(newdentry);
+
+	/*
+	 * Old dentry now lives in different location. Dentries in
+	 * lowerstack are stale. We cannot drop them here because
+	 * access to them is lockless. This could be only pure upper
+	 * or opaque directory - numlower is zero. Or upper non-dir
+	 * entry - its pureness is tracked by flag opaque.
+	 */
+	if (old_opaque != new_opaque) {
+		ovl_dentry_set_opaque(old, new_opaque);
+		if (!overwrite)
+			ovl_dentry_set_opaque(new, old_opaque);
+	}
+
+	if (cleanup_whiteout)
+		ovl_cleanup(old_upperdir->d_inode, newdentry);
+
+	ovl_dentry_version_inc(old->d_parent);
+	ovl_dentry_version_inc(new->d_parent);
+
+out_dput:
+	dput(newdentry);
+out_dput_old:
+	dput(olddentry);
+out_unlock:
+	unlock_rename(new_upperdir, old_upperdir);
+out_revert_creds:
+	if (old_opaque || new_opaque) {
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+out_drop_write:
+	ovl_drop_write(old);
+out:
+	dput(opaquedir);
+	return err;
+}
+
+const struct inode_operations ovl_dir_inode_operations = {
+	.lookup		= ovl_lookup,
+	.mkdir		= ovl_mkdir,
+	.symlink	= ovl_symlink,
+	.unlink		= ovl_unlink,
+	.rmdir		= ovl_rmdir,
+	.rename2	= ovl_rename2,
+	.link		= ovl_link,
+	.setattr	= ovl_setattr,
+	.create		= ovl_create,
+	.mknod		= ovl_mknod,
+	.permission	= ovl_permission,
+	.getattr	= ovl_dir_getattr,
+	.setxattr	= ovl_setxattr,
+	.getxattr	= ovl_getxattr,
+	.listxattr	= ovl_listxattr,
+	.removexattr	= ovl_removexattr,
+};

executer/kernel/mcoverlayfs/linux-4.6.7/inode.c

@@ -0,0 +1,494 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/xattr.h>
+#include "overlayfs.h"
+
+static int ovl_copy_up_truncate(struct dentry *dentry)
+{
+	int err;
+	struct dentry *parent;
+	struct kstat stat;
+	struct path lowerpath;
+
+	parent = dget_parent(dentry);
+	err = ovl_copy_up(parent);
+	if (err)
+		goto out_dput_parent;
+
+	ovl_path_lower(dentry, &lowerpath);
+	err = vfs_getattr(&lowerpath, &stat);
+	if (err)
+		goto out_dput_parent;
+
+	stat.size = 0;
+	err = ovl_copy_up_one(parent, dentry, &lowerpath, &stat);
+
+out_dput_parent:
+	dput(parent);
+	return err;
+}
+
+int ovl_setattr(struct dentry *dentry, struct iattr *attr)
+{
+	int err;
+	struct dentry *upperdentry;
+	unsigned opt = ovl_get_config_opt(dentry);
+
+	if (OVL_OPT_NOCOPYUPW(opt)) {
+		return 0;
+	}
+
+	/*
+	 * Check for permissions before trying to copy-up.  This is redundant
+	 * since it will be rechecked later by ->setattr() on upper dentry.  But
+	 * without this, copy-up can be triggered by just about anybody.
+	 *
+	 * We don't initialize inode->size, which just means that
+	 * inode_newsize_ok() will always check against MAX_LFS_FILESIZE and not
+	 * check for a swapfile (which this won't be anyway).
+	 */
+	err = inode_change_ok(dentry->d_inode, attr);
+	if (err)
+		return err;
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	if (attr->ia_valid & ATTR_SIZE) {
+		struct inode *realinode = d_inode(ovl_dentry_real(dentry));
+
+		err = -ETXTBSY;
+		if (atomic_read(&realinode->i_writecount) < 0)
+			goto out_drop_write;
+	}
+
+	err = ovl_copy_up(dentry);
+	if (!err) {
+		struct inode *winode = NULL;
+
+		upperdentry = ovl_dentry_upper(dentry);
+
+		if (attr->ia_valid & ATTR_SIZE) {
+			winode = d_inode(upperdentry);
+			err = get_write_access(winode);
+			if (err)
+				goto out_drop_write;
+		}
+
+		if (attr->ia_valid & (ATTR_KILL_SUID|ATTR_KILL_SGID))
+			attr->ia_valid &= ~ATTR_MODE;
+
+		inode_lock(upperdentry->d_inode);
+		err = notify_change(upperdentry, attr, NULL);
+		if (!err)
+			ovl_copyattr(upperdentry->d_inode, dentry->d_inode);
+		inode_unlock(upperdentry->d_inode);
+
+		if (winode)
+			put_write_access(winode);
+	}
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static int ovl_getattr(struct vfsmount *mnt, struct dentry *dentry,
+			 struct kstat *stat)
+{
+	struct path realpath;
+
+	ovl_path_real(dentry, &realpath);
+	return vfs_getattr(&realpath, stat);
+}
+
+int ovl_permission(struct inode *inode, int mask)
+{
+	struct ovl_entry *oe;
+	struct dentry *alias = NULL;
+	struct inode *realinode;
+	struct dentry *realdentry;
+	bool is_upper;
+	int err;
+
+	if (S_ISDIR(inode->i_mode)) {
+		oe = inode->i_private;
+	} else if (mask & MAY_NOT_BLOCK) {
+		return -ECHILD;
+	} else {
+		/*
+		 * For non-directories find an alias and get the info
+		 * from there.
+		 */
+		alias = d_find_any_alias(inode);
+		if (WARN_ON(!alias))
+			return -ENOENT;
+
+		oe = alias->d_fsdata;
+
+		ovl_reset_ovl_entry(&oe, alias);
+	}
+
+	realdentry = ovl_entry_real(oe, &is_upper);
+
+	if (ovl_is_default_permissions(inode)) {
+		struct kstat stat;
+		struct path realpath = { .dentry = realdentry };
+
+		if (mask & MAY_NOT_BLOCK)
+			return -ECHILD;
+
+		realpath.mnt = ovl_entry_mnt_real(oe, inode, is_upper);
+
+		err = vfs_getattr(&realpath, &stat);
+		if (err)
+			goto out_dput;
+
+		err = -ESTALE;
+		if ((stat.mode ^ inode->i_mode) & S_IFMT)
+			goto out_dput;
+
+		inode->i_mode = stat.mode;
+		inode->i_uid = stat.uid;
+		inode->i_gid = stat.gid;
+
+		err = generic_permission(inode, mask);
+		goto out_dput;
+	}
+
+	/* Careful in RCU walk mode */
+	realinode = ACCESS_ONCE(realdentry->d_inode);
+	if (!realinode) {
+		WARN_ON(!(mask & MAY_NOT_BLOCK));
+		err = -ENOENT;
+		goto out_dput;
+	}
+
+	if (mask & MAY_WRITE) {
+		umode_t mode = realinode->i_mode;
+
+		/*
+		 * Writes will always be redirected to upper layer, so
+		 * ignore lower layer being read-only.
+		 *
+		 * If the overlay itself is read-only then proceed
+		 * with the permission check, don't return EROFS.
+		 * This will only happen if this is the lower layer of
+		 * another overlayfs.
+		 *
+		 * If upper fs becomes read-only after the overlay was
+		 * constructed return EROFS to prevent modification of
+		 * upper layer.
+		 */
+		err = -EROFS;
+		if (is_upper && !IS_RDONLY(inode) && IS_RDONLY(realinode) &&
+		    (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
+			goto out_dput;
+	}
+
+	err = __inode_permission(realinode, mask);
+out_dput:
+	dput(alias);
+	return err;
+}
+
+static const char *ovl_get_link(struct dentry *dentry,
+				struct inode *inode,
+				struct delayed_call *done)
+{
+	struct dentry *realdentry;
+	struct inode *realinode;
+
+	if (!dentry)
+		return ERR_PTR(-ECHILD);
+
+	realdentry = ovl_dentry_real(dentry);
+	realinode = realdentry->d_inode;
+
+	if (WARN_ON(!realinode->i_op->get_link))
+		return ERR_PTR(-EPERM);
+
+	return realinode->i_op->get_link(realdentry, realinode, done);
+}
+
+static int ovl_readlink(struct dentry *dentry, char __user *buf, int bufsiz)
+{
+	struct path realpath;
+	struct inode *realinode;
+
+	ovl_path_real(dentry, &realpath);
+	realinode = realpath.dentry->d_inode;
+
+	if (!realinode->i_op->readlink)
+		return -EINVAL;
+
+	touch_atime(&realpath);
+
+	return realinode->i_op->readlink(realpath.dentry, buf, bufsiz);
+}
+
+
+static bool ovl_is_private_xattr(const char *name)
+{
+	return strncmp(name, OVL_XATTR_PRE_NAME, OVL_XATTR_PRE_LEN) == 0;
+}
+
+int ovl_setxattr(struct dentry *dentry, const char *name,
+		 const void *value, size_t size, int flags)
+{
+	int err;
+	struct dentry *upperdentry;
+	unsigned opt = ovl_get_config_opt(dentry);
+
+	if (OVL_OPT_NOCOPYUPW(opt)) {
+		return 0;
+	}
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = -EPERM;
+	if (ovl_is_private_xattr(name))
+		goto out_drop_write;
+
+	err = ovl_copy_up(dentry);
+	if (err)
+		goto out_drop_write;
+
+	upperdentry = ovl_dentry_upper(dentry);
+	err = vfs_setxattr(upperdentry, name, value, size, flags);
+
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static bool ovl_need_xattr_filter(struct dentry *dentry,
+				  enum ovl_path_type type)
+{
+	if ((type & (__OVL_PATH_PURE | __OVL_PATH_UPPER)) == __OVL_PATH_UPPER)
+		return S_ISDIR(dentry->d_inode->i_mode);
+	else
+		return false;
+}
+
+ssize_t ovl_getxattr(struct dentry *dentry, const char *name,
+		     void *value, size_t size)
+{
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+
+	if (ovl_need_xattr_filter(dentry, type) && ovl_is_private_xattr(name))
+		return -ENODATA;
+
+	return vfs_getxattr(realpath.dentry, name, value, size);
+}
+
+ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size)
+{
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+	ssize_t res;
+	int off;
+
+	res = vfs_listxattr(realpath.dentry, list, size);
+	if (res <= 0 || size == 0)
+		return res;
+
+	if (!ovl_need_xattr_filter(dentry, type))
+		return res;
+
+	/* filter out private xattrs */
+	for (off = 0; off < res;) {
+		char *s = list + off;
+		size_t slen = strlen(s) + 1;
+
+		BUG_ON(off + slen > res);
+
+		if (ovl_is_private_xattr(s)) {
+			res -= slen;
+			memmove(s, s + slen, res - off);
+		} else {
+			off += slen;
+		}
+	}
+
+	return res;
+}
+
+int ovl_removexattr(struct dentry *dentry, const char *name)
+{
+	int err;
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+	unsigned opt = ovl_get_config_opt(dentry);
+
+	if (OVL_OPT_NOCOPYUPW(opt)) {
+		return 0;
+	}
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = -ENODATA;
+	if (ovl_need_xattr_filter(dentry, type) && ovl_is_private_xattr(name))
+		goto out_drop_write;
+
+	if (!OVL_TYPE_UPPER(type)) {
+		err = vfs_getxattr(realpath.dentry, name, NULL, 0);
+		if (err < 0)
+			goto out_drop_write;
+
+		err = ovl_copy_up(dentry);
+		if (err)
+			goto out_drop_write;
+
+		ovl_path_upper(dentry, &realpath);
+	}
+
+	err = vfs_removexattr(realpath.dentry, name);
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static bool ovl_open_need_copy_up(int flags, enum ovl_path_type type,
+				  struct dentry *realdentry)
+{
+	if (OVL_TYPE_UPPER(type))
+		return false;
+
+	if (special_file(realdentry->d_inode->i_mode))
+		return false;
+
+	if (!(OPEN_FMODE(flags) & FMODE_WRITE) && !(flags & O_TRUNC))
+		return false;
+
+	return true;
+}
+
+struct inode *ovl_d_select_inode(struct dentry *dentry, unsigned file_flags)
+{
+	int err;
+	struct path realpath;
+	enum ovl_path_type type;
+	unsigned opt = ovl_get_config_opt(dentry);
+
+	if (d_is_dir(dentry))
+		return d_backing_inode(dentry);
+
+	type = ovl_path_real(dentry, &realpath);
+	if (!OVL_OPT_NOCOPYUPW(opt) && 
+		ovl_open_need_copy_up(file_flags, type, realpath.dentry)) {
+		OVL_DEBUG("copyup: realpath.dentry=%pd4, i_ino=%lu\n",
+			realpath.dentry, realpath.dentry->d_inode->i_ino);
+		err = ovl_want_write(dentry);
+		if (err)
+			return ERR_PTR(err);
+
+		if (file_flags & O_TRUNC)
+			err = ovl_copy_up_truncate(dentry);
+		else
+			err = ovl_copy_up(dentry);
+		ovl_drop_write(dentry);
+		if (err)
+			return ERR_PTR(err);
+
+		ovl_path_upper(dentry, &realpath);
+	}
+
+	if (realpath.dentry->d_flags & DCACHE_OP_SELECT_INODE)
+		return realpath.dentry->d_op->d_select_inode(realpath.dentry, file_flags);
+
+	if (OVL_OPT_NOFSCHECK(opt)) {
+		if (realpath.dentry->d_inode->i_sb->s_magic == SYSFS_MAGIC) {
+			OVL_DEBUG("sysfs: dentry=%pd4, i_ino=%lu\n", 
+				dentry, dentry->d_inode->i_ino);
+			OVL_DEBUG("sysfs: realpath.dentry=%pd4, i_ino=%lu\n", 
+				realpath.dentry, realpath.dentry->d_inode->i_ino);
+			if (!ovl_find_d_fsdata(dentry)) {
+				ovl_add_d_fsdata(dentry);
+				dentry->d_fsdata = realpath.dentry->d_fsdata;
+			}
+		}
+	}
+
+	return d_backing_inode(realpath.dentry);
+}
+
+static const struct inode_operations ovl_file_inode_operations = {
+	.setattr	= ovl_setattr,
+	.permission	= ovl_permission,
+	.getattr	= ovl_getattr,
+	.setxattr	= ovl_setxattr,
+	.getxattr	= ovl_getxattr,
+	.listxattr	= ovl_listxattr,
+	.removexattr	= ovl_removexattr,
+};
+
+static const struct inode_operations ovl_symlink_inode_operations = {
+	.setattr	= ovl_setattr,
+	.get_link	= ovl_get_link,
+	.readlink	= ovl_readlink,
+	.getattr	= ovl_getattr,
+	.setxattr	= ovl_setxattr,
+	.getxattr	= ovl_getxattr,
+	.listxattr	= ovl_listxattr,
+	.removexattr	= ovl_removexattr,
+};
+
+struct inode *ovl_new_inode(struct super_block *sb, umode_t mode,
+			    struct ovl_entry *oe)
+{
+	struct inode *inode;
+
+	inode = new_inode(sb);
+	if (!inode)
+		return NULL;
+
+	inode->i_ino = get_next_ino();
+	inode->i_mode = mode;
+	inode->i_flags |= S_NOATIME | S_NOCMTIME;
+
+	mode &= S_IFMT;
+	switch (mode) {
+	case S_IFDIR:
+		inode->i_private = oe;
+		inode->i_op = &ovl_dir_inode_operations;
+		inode->i_fop = &ovl_dir_operations;
+		break;
+
+	case S_IFLNK:
+		inode->i_op = &ovl_symlink_inode_operations;
+		break;
+
+	case S_IFREG:
+	case S_IFSOCK:
+	case S_IFBLK:
+	case S_IFCHR:
+	case S_IFIFO:
+		inode->i_op = &ovl_file_inode_operations;
+		break;
+
+	default:
+		WARN(1, "illegal file type: %i\n", mode);
+		iput(inode);
+		inode = NULL;
+	}
+
+	return inode;
+}

executer/kernel/mcoverlayfs/linux-4.6.7/overlayfs.h

@@ -0,0 +1,230 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+
+//#define DEBUG
+#ifdef DEBUG
+#define OVL_DEBUG(format, ...)	pr_err("[DEBUG] %s(): " format, __FUNCTION__, ##__VA_ARGS__)
+#else
+#define OVL_DEBUG(format, ...)	{}
+#endif
+
+struct ovl_entry;
+
+enum ovl_path_type {
+	__OVL_PATH_PURE		= (1 << 0),
+	__OVL_PATH_UPPER	= (1 << 1),
+	__OVL_PATH_MERGE	= (1 << 2),
+};
+
+#define OVL_TYPE_UPPER(type)	((type) & __OVL_PATH_UPPER)
+#define OVL_TYPE_MERGE(type)	((type) & __OVL_PATH_MERGE)
+#define OVL_TYPE_PURE_UPPER(type) ((type) & __OVL_PATH_PURE)
+#define OVL_TYPE_MERGE_OR_LOWER(type) \
+	(OVL_TYPE_MERGE(type) || !OVL_TYPE_UPPER(type))
+
+#define OVL_XATTR_PRE_NAME "trusted.overlay."
+#define OVL_XATTR_PRE_LEN  16
+#define OVL_XATTR_OPAQUE   OVL_XATTR_PRE_NAME"opaque"
+
+enum ovl_opt_bit {
+	__OVL_OPT_DEFAULT	= 0,
+	__OVL_OPT_NOCOPYUPW	= (1 << 0),
+	__OVL_OPT_NOFSCHECK	= (1 << 1),
+};
+
+#define OVL_OPT_NOCOPYUPW(opt)	((opt) & __OVL_OPT_NOCOPYUPW)
+#define OVL_OPT_NOFSCHECK(opt)	((opt) & __OVL_OPT_NOFSCHECK)
+
+struct ovl_d_fsdata {
+	struct list_head list;
+	struct dentry *d;
+	struct ovl_entry *oe;
+};
+
+static inline int ovl_do_rmdir(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_rmdir(dir, dentry);
+	pr_debug("rmdir(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+static inline int ovl_do_unlink(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_unlink(dir, dentry, NULL);
+	pr_debug("unlink(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+static inline int ovl_do_link(struct dentry *old_dentry, struct inode *dir,
+			      struct dentry *new_dentry, bool debug)
+{
+	int err = vfs_link(old_dentry, dir, new_dentry, NULL);
+	if (debug) {
+		pr_debug("link(%pd2, %pd2) = %i\n",
+			 old_dentry, new_dentry, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_create(struct inode *dir, struct dentry *dentry,
+			     umode_t mode, bool debug)
+{
+	int err = vfs_create(dir, dentry, mode, true);
+	if (debug)
+		pr_debug("create(%pd2, 0%o) = %i\n", dentry, mode, err);
+	return err;
+}
+
+static inline int ovl_do_mkdir(struct inode *dir, struct dentry *dentry,
+			       umode_t mode, bool debug)
+{
+	int err = vfs_mkdir(dir, dentry, mode);
+	if (debug)
+		pr_debug("mkdir(%pd2, 0%o) = %i\n", dentry, mode, err);
+	return err;
+}
+
+static inline int ovl_do_mknod(struct inode *dir, struct dentry *dentry,
+			       umode_t mode, dev_t dev, bool debug)
+{
+	int err = vfs_mknod(dir, dentry, mode, dev);
+	if (debug) {
+		pr_debug("mknod(%pd2, 0%o, 0%o) = %i\n",
+			 dentry, mode, dev, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_symlink(struct inode *dir, struct dentry *dentry,
+				 const char *oldname, bool debug)
+{
+	int err = vfs_symlink(dir, dentry, oldname);
+	if (debug)
+		pr_debug("symlink(\"%s\", %pd2) = %i\n", oldname, dentry, err);
+	return err;
+}
+
+static inline int ovl_do_setxattr(struct dentry *dentry, const char *name,
+				  const void *value, size_t size, int flags)
+{
+	int err = vfs_setxattr(dentry, name, value, size, flags);
+	pr_debug("setxattr(%pd2, \"%s\", \"%*s\", 0x%x) = %i\n",
+		 dentry, name, (int) size, (char *) value, flags, err);
+	return err;
+}
+
+static inline int ovl_do_removexattr(struct dentry *dentry, const char *name)
+{
+	int err = vfs_removexattr(dentry, name);
+	pr_debug("removexattr(%pd2, \"%s\") = %i\n", dentry, name, err);
+	return err;
+}
+
+static inline int ovl_do_rename(struct inode *olddir, struct dentry *olddentry,
+				struct inode *newdir, struct dentry *newdentry,
+				unsigned int flags)
+{
+	int err;
+
+	pr_debug("rename2(%pd2, %pd2, 0x%x)\n",
+		 olddentry, newdentry, flags);
+
+	err = vfs_rename(olddir, olddentry, newdir, newdentry, NULL, flags);
+
+	if (err) {
+		pr_debug("...rename2(%pd2, %pd2, ...) = %i\n",
+			 olddentry, newdentry, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_whiteout(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_whiteout(dir, dentry);
+	pr_debug("whiteout(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+unsigned ovl_get_config_opt(struct dentry *dentry);
+void ovl_reset_ovl_entry(struct ovl_entry **oe, struct dentry *dentry);
+struct ovl_entry *ovl_find_d_fsdata(struct dentry *dentry);
+int ovl_add_d_fsdata(struct dentry *dentry);
+enum ovl_path_type ovl_path_type(struct dentry *dentry);
+u64 ovl_dentry_version_get(struct dentry *dentry);
+void ovl_dentry_version_inc(struct dentry *dentry);
+void ovl_path_upper(struct dentry *dentry, struct path *path);
+void ovl_path_lower(struct dentry *dentry, struct path *path);
+enum ovl_path_type ovl_path_real(struct dentry *dentry, struct path *path);
+int ovl_path_next(int idx, struct dentry *dentry, struct path *path);
+struct dentry *ovl_dentry_upper(struct dentry *dentry);
+struct dentry *ovl_dentry_lower(struct dentry *dentry);
+struct dentry *ovl_dentry_real(struct dentry *dentry);
+struct dentry *ovl_entry_real(struct ovl_entry *oe, bool *is_upper);
+struct vfsmount *ovl_entry_mnt_real(struct ovl_entry *oe, struct inode *inode,
+				    bool is_upper);
+struct ovl_dir_cache *ovl_dir_cache(struct dentry *dentry);
+bool ovl_is_default_permissions(struct inode *inode);
+void ovl_set_dir_cache(struct dentry *dentry, struct ovl_dir_cache *cache);
+struct dentry *ovl_workdir(struct dentry *dentry);
+int ovl_want_write(struct dentry *dentry);
+void ovl_drop_write(struct dentry *dentry);
+bool ovl_dentry_is_opaque(struct dentry *dentry);
+void ovl_dentry_set_opaque(struct dentry *dentry, bool opaque);
+bool ovl_is_whiteout(struct dentry *dentry);
+void ovl_dentry_update(struct dentry *dentry, struct dentry *upperdentry);
+struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
+			  unsigned int flags);
+struct file *ovl_path_open(struct path *path, int flags);
+
+struct dentry *ovl_upper_create(struct dentry *upperdir, struct dentry *dentry,
+				struct kstat *stat, const char *link);
+
+/* readdir.c */
+extern const struct file_operations ovl_dir_operations;
+int ovl_check_empty_dir(struct dentry *dentry, struct list_head *list);
+void ovl_cleanup_whiteouts(struct dentry *upper, struct list_head *list);
+void ovl_cache_free(struct list_head *list);
+int ovl_check_d_type_supported(struct path *realpath);
+
+/* inode.c */
+int ovl_setattr(struct dentry *dentry, struct iattr *attr);
+int ovl_permission(struct inode *inode, int mask);
+int ovl_setxattr(struct dentry *dentry, const char *name,
+		 const void *value, size_t size, int flags);
+ssize_t ovl_getxattr(struct dentry *dentry, const char *name,
+		     void *value, size_t size);
+ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size);
+int ovl_removexattr(struct dentry *dentry, const char *name);
+struct inode *ovl_d_select_inode(struct dentry *dentry, unsigned file_flags);
+
+struct inode *ovl_new_inode(struct super_block *sb, umode_t mode,
+			    struct ovl_entry *oe);
+static inline void ovl_copyattr(struct inode *from, struct inode *to)
+{
+	to->i_uid = from->i_uid;
+	to->i_gid = from->i_gid;
+	to->i_mode = from->i_mode;
+}
+
+/* dir.c */
+extern const struct inode_operations ovl_dir_inode_operations;
+struct dentry *ovl_lookup_temp(struct dentry *workdir, struct dentry *dentry);
+int ovl_create_real(struct inode *dir, struct dentry *newdentry,
+		    struct kstat *stat, const char *link,
+		    struct dentry *hardlink, bool debug);
+void ovl_cleanup(struct inode *dir, struct dentry *dentry);
+
+/* copy_up.c */
+int ovl_copy_up(struct dentry *dentry);
+int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry,
+		    struct path *lowerpath, struct kstat *stat);
+int ovl_copy_xattr(struct dentry *old, struct dentry *new, unsigned opt);
+int ovl_set_attr(struct dentry *upper, struct kstat *stat);

executer/kernel/mcoverlayfs/linux-4.6.7/readdir.c

@@ -0,0 +1,616 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/namei.h>
+#include <linux/file.h>
+#include <linux/xattr.h>
+#include <linux/rbtree.h>
+#include <linux/security.h>
+#include <linux/cred.h>
+#include "overlayfs.h"
+
+struct ovl_cache_entry {
+	unsigned int len;
+	unsigned int type;
+	u64 ino;
+	struct list_head l_node;
+	struct rb_node node;
+	struct ovl_cache_entry *next_maybe_whiteout;
+	bool is_whiteout;
+	char name[];
+};
+
+struct ovl_dir_cache {
+	long refcount;
+	u64 version;
+	struct list_head entries;
+};
+
+struct ovl_readdir_data {
+	struct dir_context ctx;
+	bool is_lowest;
+	struct rb_root root;
+	struct list_head *list;
+	struct list_head middle;
+	struct ovl_cache_entry *first_maybe_whiteout;
+	int count;
+	int err;
+	bool d_type_supported;
+};
+
+struct ovl_dir_file {
+	bool is_real;
+	bool is_upper;
+	struct ovl_dir_cache *cache;
+	struct list_head *cursor;
+	struct file *realfile;
+	struct file *upperfile;
+};
+
+static struct ovl_cache_entry *ovl_cache_entry_from_node(struct rb_node *n)
+{
+	return container_of(n, struct ovl_cache_entry, node);
+}
+
+static struct ovl_cache_entry *ovl_cache_entry_find(struct rb_root *root,
+						    const char *name, int len)
+{
+	struct rb_node *node = root->rb_node;
+	int cmp;
+
+	while (node) {
+		struct ovl_cache_entry *p = ovl_cache_entry_from_node(node);
+
+		cmp = strncmp(name, p->name, len);
+		if (cmp > 0)
+			node = p->node.rb_right;
+		else if (cmp < 0 || len < p->len)
+			node = p->node.rb_left;
+		else
+			return p;
+	}
+
+	return NULL;
+}
+
+static struct ovl_cache_entry *ovl_cache_entry_new(struct ovl_readdir_data *rdd,
+						   const char *name, int len,
+						   u64 ino, unsigned int d_type)
+{
+	struct ovl_cache_entry *p;
+	size_t size = offsetof(struct ovl_cache_entry, name[len + 1]);
+
+	p = kmalloc(size, GFP_KERNEL);
+	if (!p)
+		return NULL;
+
+	memcpy(p->name, name, len);
+	p->name[len] = '\0';
+	p->len = len;
+	p->type = d_type;
+	p->ino = ino;
+	p->is_whiteout = false;
+
+	if (d_type == DT_CHR) {
+		p->next_maybe_whiteout = rdd->first_maybe_whiteout;
+		rdd->first_maybe_whiteout = p;
+	}
+	return p;
+}
+
+static int ovl_cache_entry_add_rb(struct ovl_readdir_data *rdd,
+				  const char *name, int len, u64 ino,
+				  unsigned int d_type)
+{
+	struct rb_node **newp = &rdd->root.rb_node;
+	struct rb_node *parent = NULL;
+	struct ovl_cache_entry *p;
+
+	while (*newp) {
+		int cmp;
+		struct ovl_cache_entry *tmp;
+
+		parent = *newp;
+		tmp = ovl_cache_entry_from_node(*newp);
+		cmp = strncmp(name, tmp->name, len);
+		if (cmp > 0)
+			newp = &tmp->node.rb_right;
+		else if (cmp < 0 || len < tmp->len)
+			newp = &tmp->node.rb_left;
+		else
+			return 0;
+	}
+
+	p = ovl_cache_entry_new(rdd, name, len, ino, d_type);
+	if (p == NULL)
+		return -ENOMEM;
+
+	list_add_tail(&p->l_node, rdd->list);
+	rb_link_node(&p->node, parent, newp);
+	rb_insert_color(&p->node, &rdd->root);
+
+	return 0;
+}
+
+static int ovl_fill_lowest(struct ovl_readdir_data *rdd,
+			   const char *name, int namelen,
+			   loff_t offset, u64 ino, unsigned int d_type)
+{
+	struct ovl_cache_entry *p;
+
+	p = ovl_cache_entry_find(&rdd->root, name, namelen);
+	if (p) {
+		list_move_tail(&p->l_node, &rdd->middle);
+	} else {
+		p = ovl_cache_entry_new(rdd, name, namelen, ino, d_type);
+		if (p == NULL)
+			rdd->err = -ENOMEM;
+		else
+			list_add_tail(&p->l_node, &rdd->middle);
+	}
+
+	return rdd->err;
+}
+
+void ovl_cache_free(struct list_head *list)
+{
+	struct ovl_cache_entry *p;
+	struct ovl_cache_entry *n;
+
+	list_for_each_entry_safe(p, n, list, l_node)
+		kfree(p);
+
+	INIT_LIST_HEAD(list);
+}
+
+static void ovl_cache_put(struct ovl_dir_file *od, struct dentry *dentry)
+{
+	struct ovl_dir_cache *cache = od->cache;
+
+	WARN_ON(cache->refcount <= 0);
+	cache->refcount--;
+	if (!cache->refcount) {
+		if (ovl_dir_cache(dentry) == cache)
+			ovl_set_dir_cache(dentry, NULL);
+
+		ovl_cache_free(&cache->entries);
+		kfree(cache);
+	}
+}
+
+static int ovl_fill_merge(struct dir_context *ctx, const char *name,
+			  int namelen, loff_t offset, u64 ino,
+			  unsigned int d_type)
+{
+	struct ovl_readdir_data *rdd =
+		container_of(ctx, struct ovl_readdir_data, ctx);
+
+	rdd->count++;
+	if (!rdd->is_lowest)
+		return ovl_cache_entry_add_rb(rdd, name, namelen, ino, d_type);
+	else
+		return ovl_fill_lowest(rdd, name, namelen, offset, ino, d_type);
+}
+
+static int ovl_check_whiteouts(struct dentry *dir, struct ovl_readdir_data *rdd)
+{
+	int err;
+	struct ovl_cache_entry *p;
+	struct dentry *dentry;
+	const struct cred *old_cred;
+	struct cred *override_cred;
+
+	override_cred = prepare_creds();
+	if (!override_cred)
+		return -ENOMEM;
+
+	/*
+	 * CAP_DAC_OVERRIDE for lookup
+	 */
+	cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+	old_cred = override_creds(override_cred);
+
+	err = mutex_lock_killable(&dir->d_inode->i_mutex);
+	if (!err) {
+		while (rdd->first_maybe_whiteout) {
+			p = rdd->first_maybe_whiteout;
+			rdd->first_maybe_whiteout = p->next_maybe_whiteout;
+			dentry = lookup_one_len(p->name, dir, p->len);
+			if (!IS_ERR(dentry)) {
+				p->is_whiteout = ovl_is_whiteout(dentry);
+				dput(dentry);
+			}
+		}
+		inode_unlock(dir->d_inode);
+	}
+	revert_creds(old_cred);
+	put_cred(override_cred);
+
+	return err;
+}
+
+static inline int ovl_dir_read(struct path *realpath,
+			       struct ovl_readdir_data *rdd)
+{
+	struct file *realfile;
+	int err;
+
+	realfile = ovl_path_open(realpath, O_RDONLY | O_DIRECTORY);
+	if (IS_ERR(realfile))
+		return PTR_ERR(realfile);
+
+	rdd->first_maybe_whiteout = NULL;
+	rdd->ctx.pos = 0;
+	do {
+		rdd->count = 0;
+		rdd->err = 0;
+		err = iterate_dir(realfile, &rdd->ctx);
+		if (err >= 0)
+			err = rdd->err;
+	} while (!err && rdd->count);
+
+	if (!err && rdd->first_maybe_whiteout)
+		err = ovl_check_whiteouts(realpath->dentry, rdd);
+
+	fput(realfile);
+
+	return err;
+}
+
+static void ovl_dir_reset(struct file *file)
+{
+	struct ovl_dir_file *od = file->private_data;
+	struct ovl_dir_cache *cache = od->cache;
+	struct dentry *dentry = file->f_path.dentry;
+	enum ovl_path_type type = ovl_path_type(dentry);
+
+	if (cache && ovl_dentry_version_get(dentry) != cache->version) {
+		ovl_cache_put(od, dentry);
+		od->cache = NULL;
+		od->cursor = NULL;
+	}
+	WARN_ON(!od->is_real && !OVL_TYPE_MERGE(type));
+	if (od->is_real && OVL_TYPE_MERGE(type))
+		od->is_real = false;
+}
+
+static int ovl_dir_read_merged(struct dentry *dentry, struct list_head *list)
+{
+	int err;
+	struct path realpath;
+	struct ovl_readdir_data rdd = {
+		.ctx.actor = ovl_fill_merge,
+		.list = list,
+		.root = RB_ROOT,
+		.is_lowest = false,
+	};
+	int idx, next;
+
+	for (idx = 0; idx != -1; idx = next) {
+		next = ovl_path_next(idx, dentry, &realpath);
+
+		if (next != -1) {
+			err = ovl_dir_read(&realpath, &rdd);
+			if (err)
+				break;
+		} else {
+			/*
+			 * Insert lowest layer entries before upper ones, this
+			 * allows offsets to be reasonably constant
+			 */
+			list_add(&rdd.middle, rdd.list);
+			rdd.is_lowest = true;
+			err = ovl_dir_read(&realpath, &rdd);
+			list_del(&rdd.middle);
+		}
+	}
+	return err;
+}
+
+static void ovl_seek_cursor(struct ovl_dir_file *od, loff_t pos)
+{
+	struct list_head *p;
+	loff_t off = 0;
+
+	list_for_each(p, &od->cache->entries) {
+		if (off >= pos)
+			break;
+		off++;
+	}
+	/* Cursor is safe since the cache is stable */
+	od->cursor = p;
+}
+
+static struct ovl_dir_cache *ovl_cache_get(struct dentry *dentry)
+{
+	int res;
+	struct ovl_dir_cache *cache;
+
+	cache = ovl_dir_cache(dentry);
+	if (cache && ovl_dentry_version_get(dentry) == cache->version) {
+		cache->refcount++;
+		return cache;
+	}
+	ovl_set_dir_cache(dentry, NULL);
+
+	cache = kzalloc(sizeof(struct ovl_dir_cache), GFP_KERNEL);
+	if (!cache)
+		return ERR_PTR(-ENOMEM);
+
+	cache->refcount = 1;
+	INIT_LIST_HEAD(&cache->entries);
+
+	res = ovl_dir_read_merged(dentry, &cache->entries);
+	if (res) {
+		ovl_cache_free(&cache->entries);
+		kfree(cache);
+		return ERR_PTR(res);
+	}
+
+	cache->version = ovl_dentry_version_get(dentry);
+	ovl_set_dir_cache(dentry, cache);
+
+	return cache;
+}
+
+static int ovl_iterate(struct file *file, struct dir_context *ctx)
+{
+	struct ovl_dir_file *od = file->private_data;
+	struct dentry *dentry = file->f_path.dentry;
+	struct ovl_cache_entry *p;
+
+	if (!ctx->pos)
+		ovl_dir_reset(file);
+
+	if (od->is_real)
+		return iterate_dir(od->realfile, ctx);
+
+	if (!od->cache) {
+		struct ovl_dir_cache *cache;
+
+		cache = ovl_cache_get(dentry);
+		if (IS_ERR(cache))
+			return PTR_ERR(cache);
+
+		od->cache = cache;
+		ovl_seek_cursor(od, ctx->pos);
+	}
+
+	while (od->cursor != &od->cache->entries) {
+		p = list_entry(od->cursor, struct ovl_cache_entry, l_node);
+		if (!p->is_whiteout)
+			if (!dir_emit(ctx, p->name, p->len, p->ino, p->type))
+				break;
+		od->cursor = p->l_node.next;
+		ctx->pos++;
+	}
+	return 0;
+}
+
+static loff_t ovl_dir_llseek(struct file *file, loff_t offset, int origin)
+{
+	loff_t res;
+	struct ovl_dir_file *od = file->private_data;
+
+	inode_lock(file_inode(file));
+	if (!file->f_pos)
+		ovl_dir_reset(file);
+
+	if (od->is_real) {
+		res = vfs_llseek(od->realfile, offset, origin);
+		file->f_pos = od->realfile->f_pos;
+	} else {
+		res = -EINVAL;
+
+		switch (origin) {
+		case SEEK_CUR:
+			offset += file->f_pos;
+			break;
+		case SEEK_SET:
+			break;
+		default:
+			goto out_unlock;
+		}
+		if (offset < 0)
+			goto out_unlock;
+
+		if (offset != file->f_pos) {
+			file->f_pos = offset;
+			if (od->cache)
+				ovl_seek_cursor(od, offset);
+		}
+		res = offset;
+	}
+out_unlock:
+	inode_unlock(file_inode(file));
+
+	return res;
+}
+
+static int ovl_dir_fsync(struct file *file, loff_t start, loff_t end,
+			 int datasync)
+{
+	struct ovl_dir_file *od = file->private_data;
+	struct dentry *dentry = file->f_path.dentry;
+	struct file *realfile = od->realfile;
+
+	/*
+	 * Need to check if we started out being a lower dir, but got copied up
+	 */
+	if (!od->is_upper && OVL_TYPE_UPPER(ovl_path_type(dentry))) {
+		struct inode *inode = file_inode(file);
+
+		realfile = lockless_dereference(od->upperfile);
+		if (!realfile) {
+			struct path upperpath;
+
+			ovl_path_upper(dentry, &upperpath);
+			realfile = ovl_path_open(&upperpath, O_RDONLY);
+			smp_mb__before_spinlock();
+			inode_lock(inode);
+			if (!od->upperfile) {
+				if (IS_ERR(realfile)) {
+					inode_unlock(inode);
+					return PTR_ERR(realfile);
+				}
+				od->upperfile = realfile;
+			} else {
+				/* somebody has beaten us to it */
+				if (!IS_ERR(realfile))
+					fput(realfile);
+				realfile = od->upperfile;
+			}
+			inode_unlock(inode);
+		}
+	}
+
+	return vfs_fsync_range(realfile, start, end, datasync);
+}
+
+static int ovl_dir_release(struct inode *inode, struct file *file)
+{
+	struct ovl_dir_file *od = file->private_data;
+
+	if (od->cache) {
+		inode_lock(inode);
+		ovl_cache_put(od, file->f_path.dentry);
+		inode_unlock(inode);
+	}
+	fput(od->realfile);
+	if (od->upperfile)
+		fput(od->upperfile);
+	kfree(od);
+
+	return 0;
+}
+
+static int ovl_dir_open(struct inode *inode, struct file *file)
+{
+	struct path realpath;
+	struct file *realfile;
+	struct ovl_dir_file *od;
+	enum ovl_path_type type;
+
+	od = kzalloc(sizeof(struct ovl_dir_file), GFP_KERNEL);
+	if (!od)
+		return -ENOMEM;
+
+	type = ovl_path_real(file->f_path.dentry, &realpath);
+	realfile = ovl_path_open(&realpath, file->f_flags);
+	if (IS_ERR(realfile)) {
+		kfree(od);
+		return PTR_ERR(realfile);
+	}
+	od->realfile = realfile;
+	od->is_real = !OVL_TYPE_MERGE(type);
+	od->is_upper = OVL_TYPE_UPPER(type);
+	file->private_data = od;
+
+	return 0;
+}
+
+const struct file_operations ovl_dir_operations = {
+	.read		= generic_read_dir,
+	.open		= ovl_dir_open,
+	.iterate	= ovl_iterate,
+	.llseek		= ovl_dir_llseek,
+	.fsync		= ovl_dir_fsync,
+	.release	= ovl_dir_release,
+};
+
+int ovl_check_empty_dir(struct dentry *dentry, struct list_head *list)
+{
+	int err;
+	struct ovl_cache_entry *p;
+
+	err = ovl_dir_read_merged(dentry, list);
+	if (err)
+		return err;
+
+	err = 0;
+
+	list_for_each_entry(p, list, l_node) {
+		if (p->is_whiteout)
+			continue;
+
+		if (p->name[0] == '.') {
+			if (p->len == 1)
+				continue;
+			if (p->len == 2 && p->name[1] == '.')
+				continue;
+		}
+		err = -ENOTEMPTY;
+		break;
+	}
+
+	return err;
+}
+
+void ovl_cleanup_whiteouts(struct dentry *upper, struct list_head *list)
+{
+	struct ovl_cache_entry *p;
+
+	inode_lock_nested(upper->d_inode, I_MUTEX_CHILD);
+	list_for_each_entry(p, list, l_node) {
+		struct dentry *dentry;
+
+		if (!p->is_whiteout)
+			continue;
+
+		dentry = lookup_one_len(p->name, upper, p->len);
+		if (IS_ERR(dentry)) {
+			pr_err("overlayfs: lookup '%s/%.*s' failed (%i)\n",
+			       upper->d_name.name, p->len, p->name,
+			       (int) PTR_ERR(dentry));
+			continue;
+		}
+		if (dentry->d_inode)
+			ovl_cleanup(upper->d_inode, dentry);
+		dput(dentry);
+	}
+	inode_unlock(upper->d_inode);
+}
+
+static int ovl_check_d_type(struct dir_context *ctx, const char *name,
+			  int namelen, loff_t offset, u64 ino,
+			  unsigned int d_type)
+{
+	struct ovl_readdir_data *rdd =
+		container_of(ctx, struct ovl_readdir_data, ctx);
+
+	/* Even if d_type is not supported, DT_DIR is returned for . and .. */
+	if (!strncmp(name, ".", namelen) || !strncmp(name, "..", namelen))
+		return 0;
+
+	if (d_type != DT_UNKNOWN)
+		rdd->d_type_supported = true;
+
+	return 0;
+}
+
+/*
+ * Returns 1 if d_type is supported, 0 not supported/unknown. Negative values
+ * if error is encountered.
+ */
+int ovl_check_d_type_supported(struct path *realpath)
+{
+	int err;
+	struct ovl_readdir_data rdd = {
+		.ctx.actor = ovl_check_d_type,
+		.d_type_supported = false,
+	};
+
+	err = ovl_dir_read(realpath, &rdd);
+	if (err)
+		return err;
+
+	return rdd.d_type_supported;
+}

executer/user/Makefile.in

@@ -6,14 +6,15 @@ VPATH=@abs_srcdir@
 TARGET=mcexec
 @uncomment_if_ENABLE_MEMDUMP@TARGET+=eclair
 LIBS=@LIBS@
+IHKDIR ?= $(VPATH)/../../../ihk/linux/include/
 
 all: $(TARGET)
 
 mcexec: mcexec.c
-	$(CC) -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -lrt -pthread -o $@ $^ $(EXTRA_OBJS)
+	$(CC) -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -lrt -lnuma -pthread -o $@ $^ $(EXTRA_OBJS)
 
 eclair: eclair.c
-	$(CC) $(CFLAGS) -o $@ $^ $(LIBS)
+	$(CC) $(CFLAGS) -I${IHKDIR} -o $@ $^ $(LIBS)
 
 clean:
 	$(RM) $(TARGET) *.o

executer/user/eclair.c

@@ -16,6 +16,8 @@
 #include <unistd.h>
 #include <sys/socket.h>
 #include <arpa/inet.h>
+#include <sys/ioctl.h>
+#include <ihk/ihk_host_user.h>
 
 #define CPU_TID_BASE 1000000
 
@@ -25,6 +27,10 @@ struct options {
 	char *kernel_path;
 	char *dump_path;
 	char *log_path;
+	int interactive;
+	int os_id;
+	int mcos_fd;
+	int print_idle;
 }; /* struct options */
 
 struct thread_info {
@@ -42,7 +48,7 @@ struct thread_info {
 	int tid;
 	int cpu;
 	int lcpu;
-	int padding;
+	int idle;
 	uintptr_t process;
 	uintptr_t clv;
 	uintptr_t x86_clv;
@@ -53,6 +59,7 @@ static volatile int f_done = 0;
 static bfd *symbfd = NULL;
 static bfd *dumpbfd = NULL;
 static asection *dumpscn = NULL;
+static dump_mem_chunks_t *mem_chunks;
 static int num_processors = -1;
 static asymbol **symtab = NULL;
 static ssize_t nsyms;
@@ -91,25 +98,35 @@ static uintptr_t virt_to_phys(uintptr_t va) {
 static int read_physmem(uintptr_t pa, void *buf, size_t size) {
 	off_t off;
 	bfd_boolean ok;
+	int i;
 
-	if (pa < dumpscn->vma) {
-		printf("read_physmem(%lx,%p,%lx):too small pa. vma %lx\n", pa, buf, size, dumpscn->vma);
-		return 1;
-	}
-	off = pa - dumpscn->vma;
-	if (off >= dumpscn->size) {
-		printf("read_physmem(%lx,%p,%lx):too large pa. vma %lx size %lx\n", pa, buf, size, dumpscn->vma, dumpscn->size);
-		return 1;
-	}
-	if ((dumpscn->size - off) < size) {
-		printf("read_physmem(%lx,%p,%lx):too large size. vma %lx size %lx\n", pa, buf, size, dumpscn->vma, dumpscn->size);
+	off = 0;
+	/* Check if pa is valid in any chunks and figure
+	 * out the global offset in dump section */
+	for (i = 0; i < mem_chunks->nr_chunks; ++i) {
+
+		if (mem_chunks->chunks[i].addr <= pa &&
+				((pa + size) <= (mem_chunks->chunks[i].addr +
+					mem_chunks->chunks[i].size))) {
+
+			off += (pa - mem_chunks->chunks[i].addr);
+			break;
+		}
+
+		off += mem_chunks->chunks[i].size;
+	}
+
+	if (i == mem_chunks->nr_chunks) {
+		printf("read_physmem: invalid addr 0x%lx\n", pa);
 		return 1;
 	}
+
 	ok = bfd_get_section_contents(dumpbfd, dumpscn, buf, off, size);
 	if (!ok) {
 		bfd_perror("read_physmem:bfd_get_section_contents");
 		return 1;
 	}
+
 	return 0;
 } /* read_physmem() */
 
@@ -125,7 +142,21 @@ static int read_mem(uintptr_t va, void *buf, size_t size) {
 		}
 		return 1;
 	}
-	error = read_physmem(pa, buf, size);
+
+	if (opt.interactive) {
+		dumpargs_t args;
+
+		args.cmd = DUMP_READ;
+		args.start = pa;
+		args.size = size;
+		args.buf = buf;
+
+		error = ioctl(opt.mcos_fd, IHK_OS_DUMP, &args);
+	}
+	else {
+		error = read_physmem(pa, buf, size);
+	}
+
 	if (error) {
 		perror("read_mem:read_physmem");
 		return 1;
@@ -167,6 +198,7 @@ enum {
 	CURRENT_OFFSET,
 	RUNQ_OFFSET,
 	CPU_STATUS_OFFSET,
+	IDLE_THREAD_OFFSET,
 
 	/* process */
 	CTX_OFFSET,
@@ -204,6 +236,7 @@ static int setup_constants(void) {
 		printf("CURRENT_OFFSET: %ld\n", K(CURRENT_OFFSET));
 		printf("RUNQ_OFFSET: %ld\n", K(RUNQ_OFFSET));
 		printf("CPU_STATUS_OFFSET: %ld\n", K(CPU_STATUS_OFFSET));
+		printf("IDLE_THREAD_OFFSET: %ld\n", K(IDLE_THREAD_OFFSET));
 		printf("CTX_OFFSET: %ld\n", K(CTX_OFFSET));
 		printf("SCHED_LIST_OFFSET: %ld\n", K(SCHED_LIST_OFFSET));
 		printf("PROC_OFFSET: %ld\n", K(PROC_OFFSET));
@@ -229,6 +262,7 @@ static int setup_threads(void) {
 		perror("num_processors");
 		return 1;
 	}
+	printf("%s: num_processors: %d\n", __FUNCTION__, num_processors);
 
 	error = read_symbol_64("locals", &locals);
 	if (error) {
@@ -315,15 +349,19 @@ static int setup_threads(void) {
 			ti->status = status;
 			ti->pid = pid;
 			ti->tid = tid;
-			ti->cpu = (thread == current)? cpu: -1;
+			ti->cpu = (thread == current) ? cpu : -1;
 			ti->lcpu = cpu;
 			ti->process = thread;
+			ti->idle = 0;
 			ti->clv = v;
 			ti->x86_clv = locals + locals_span*cpu;
 
 			*titailp = ti;
 			titailp = &ti->next;
 
+			if (!curr_thread)
+				curr_thread = ti;
+
 			error = read_64(entry, &entry);
 			if (error) {
 				perror("process2");
@@ -332,8 +370,78 @@ static int setup_threads(void) {
 		}
 	}
 
+	/* Set up idle threads */
+	if (opt.print_idle) {
+		for (cpu = 0; cpu < num_processors; ++cpu) {
+			uintptr_t v;
+			uintptr_t thread;
+			uintptr_t proc;
+			int pid;
+			int tid;
+			struct thread_info *ti;
+			int status;
+
+			v = clv + (cpu * K(CPU_LOCAL_VAR_SIZE));
+
+			error = read_64(v+K(CURRENT_OFFSET), &current);
+			if (error) {
+				perror("current");
+				return 1;
+			}
+
+			ti = malloc(sizeof(*ti));
+			if (!ti) {
+				perror("malloc");
+				return 1;
+			}
+
+			thread = v+K(IDLE_THREAD_OFFSET);
+
+			error = read_64(thread+K(PROC_OFFSET), &proc);
+			if (error) {
+				perror("proc");
+				return 1;
+			}
+
+			error = read_32(thread+K(STATUS_OFFSET), &status);
+			if (error) {
+				perror("status");
+				return 1;
+			}
+
+			error = read_32(proc+K(PID_OFFSET), &pid);
+			if (error) {
+				perror("pid");
+				return 1;
+			}
+
+			error = read_32(thread+K(TID_OFFSET), &tid);
+			if (error) {
+				perror("tid");
+				return 1;
+			}
+
+			ti->next = NULL;
+			ti->status = status;
+			ti->pid = 1;
+			ti->tid = 2000000000 + tid;
+			ti->cpu = (thread == current) ? cpu : -1;
+			ti->lcpu = cpu;
+			ti->process = thread;
+			ti->idle = 1;
+			ti->clv = v;
+			ti->x86_clv = locals + locals_span*cpu;
+
+			*titailp = ti;
+			titailp = &ti->next;
+
+			if (!curr_thread)
+				curr_thread = ti;
+		}
+	}
+
 	if (!tihead) {
-		printf("thread not found. cpu mode forcibly\n");
+		printf("No threads found, forcing CPU mode.\n");
 		opt.cpu = 1;
 	}
 
@@ -374,6 +482,7 @@ static int setup_threads(void) {
 			ti->tid = CPU_TID_BASE + cpu;
 			ti->cpu = cpu;
 			ti->process = current;
+			ti->idle = 1;
 			ti->clv = v;
 			ti->x86_clv = locals + locals_span*cpu;
 
@@ -386,7 +495,9 @@ static int setup_threads(void) {
 		printf("thread not found\n");
 		return 1;
 	}
-	curr_thread = tihead;
+
+	if (!curr_thread)
+		curr_thread = tihead;
 
 	return 0;
 } /* setup_threads() */
@@ -448,6 +559,25 @@ static int setup_dump(char *fname) {
 		return 1;
 	}
 
+	mem_chunks = malloc(PHYS_CHUNKS_DESC_SIZE);
+	if (!mem_chunks) {
+		perror("allocating mem chunks descriptor: ");
+		return 1;
+	}
+
+	dumpscn = bfd_get_section_by_name(dumpbfd, "physchunks");
+	if (!dumpscn) {
+		bfd_perror("bfd_get_section_by_name");
+		return 1;
+	}
+
+	ok = bfd_get_section_contents(dumpbfd, dumpscn, mem_chunks,
+			0, PHYS_CHUNKS_DESC_SIZE);
+	if (!ok) {
+		bfd_perror("read_physmem:bfd_get_section_contents");
+		return 1;
+	}
+
 	dumpscn = bfd_get_section_by_name(dumpbfd, "physmem");
 	if (!dumpscn) {
 		bfd_perror("bfd_get_section_by_name");
@@ -609,18 +739,21 @@ static void command(char *cmd, char *res) {
 					break;
 				}
 
+				//if (regs[17] > MAP_KERNEL) {}
 				pu8 = (void *)&regs;
 				for (i = 0; i < sizeof(regs)-4; ++i) {
 					rbp += sprintf(rbp, "%02x", pu8[i]);
 				}
 			}
 		}
+		/*
 		else if (!strcmp(p, "mffffffff80018a82,1")) {
 			rbp += sprintf(rbp, "b8");
 		}
 		else if (!strcmp(p, "mffffffff80018a82,9")) {
 			rbp += sprintf(rbp, "b8f2ffffff41564155");
 		}
+		*/
 		else if (!strncmp(p, "m", 1)) {
 			int n;
 			uintptr_t start;
@@ -716,33 +849,35 @@ static void command(char *cmd, char *res) {
 				break;
 			}
 			q = buf;
+			q += sprintf(q, "PID %d, ", ti->pid);
 			if (ti->status & PS_RUNNING) {
-				q += sprintf(q, "running on cpu%d", ti->cpu);
+				q += sprintf(q, "%srunning on cpu %d",
+					ti->idle ? "idle " : "", ti->lcpu);
 			}
 			else if (ti->status & (PS_INTERRUPTIBLE | PS_UNINTERRUPTIBLE)) {
-				q += sprintf(q, "waiting on cpu%d", ti->lcpu);
+				q += sprintf(q, "%swaiting on cpu %d",
+					ti->idle ? "idle " : "", ti->lcpu);
 			}
 			else if (ti->status & PS_STOPPED) {
-				q += sprintf(q, "stopped on cpu%d", ti->lcpu);
+				q += sprintf(q, "%sstopped on cpu %d",
+					ti->idle ? "idle " : "", ti->lcpu);
 			}
 			else if (ti->status & PS_TRACED) {
-				q += sprintf(q, "traced on cpu%d", ti->lcpu);
+				q += sprintf(q, "%straced on cpu %d",
+					ti->idle ? "idle " : "", ti->lcpu);
 			}
 			else if (ti->status == CS_IDLE) {
-				q += sprintf(q, "cpu%d idle", ti->cpu);
+				q += sprintf(q, "cpu %d idle", ti->cpu);
 			}
 			else if (ti->status == CS_RUNNING) {
-				q += sprintf(q, "cpu%d running", ti->cpu);
+				q += sprintf(q, "cpu %d running", ti->cpu);
 			}
 			else if (ti->status == CS_RESERVED) {
-				q += sprintf(q, "cpu%d reserved", ti->cpu);
+				q += sprintf(q, "cpu %d reserved", ti->cpu);
 			}
 			else {
 				q += sprintf(q, "status=%#x", ti->status);
 			}
-			if (ti->tid != ti->pid) {
-				q += sprintf(q, ",pid=%d", ti->pid);
-			}
 			rbp += print_hex(rbp, buf);
 		}
 	} while (0);
@@ -755,11 +890,12 @@ static void options(int argc, char *argv[]) {
 	memset(&opt, 0, sizeof(opt));
 	opt.kernel_path = "./mckernel.img";
 	opt.dump_path = "./mcdump";
+	opt.mcos_fd = -1;
 
 	for (;;) {
 		int c;
 
-		c = getopt(argc, argv, "cd:hk:");
+		c = getopt(argc, argv, "ilcd:hk:o:");
 		if (c < 0) {
 			break;
 		}
@@ -777,12 +913,32 @@ static void options(int argc, char *argv[]) {
 		case 'd':
 			opt.dump_path = optarg;
 			break;
+		case 'i':
+			opt.interactive = 1;
+			break;
+		case 'o':
+			opt.os_id = atoi(optarg);
+			break;
+		case 'l':
+			opt.print_idle = 1;
+			break;
 		}
 	}
 	if (optind < argc) {
 		opt.help = 1;
 	}
 
+	if (opt.interactive) {
+		char fn[128];
+		sprintf(fn, "/dev/mcos%d", opt.os_id);
+
+		opt.mcos_fd = open(fn, O_RDONLY);
+		if (opt.mcos_fd < 0) {
+			perror("open");
+			exit(1);
+		}
+	}
+
 	return;
 } /* options() */
 
@@ -865,7 +1021,7 @@ int main(int argc, char *argv[]) {
 	uint8_t sum;
 	uint8_t check;
 	static char lbuf[1024];
-	static char rbuf[1024];
+	static char rbuf[8192];
 	static char cbuf[3];
 	char *lbp;
 	char *p;

executer/user/mcexec.c

@@ -41,6 +41,7 @@
 #include <sys/mman.h>
 #include <asm/unistd.h>
 #include <sched.h>
+#include <dirent.h>
 
 #include <termios.h>
 #include <sys/ioctl.h>
@@ -65,6 +66,8 @@
 #include "../include/uprotocol.h"
 #include <getopt.h>
 #include "../config.h"
+#include <numa.h>
+#include <numaif.h>
 
 //#define DEBUG
 
@@ -101,6 +104,22 @@ int __glob_argc = -1;
 char **__glob_argv = 0;
 #endif
 
+#ifdef ENABLE_MCOVERLAYFS
+#undef ENABLE_MCOVERLAYFS
+#ifndef RHEL_RELEASE_CODE
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,1,0)
+#define ENABLE_MCOVERLAYFS 1
+#endif // LINUX_VERSION_CODE == 4.0
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,6,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,7,0)
+#define ENABLE_MCOVERLAYFS 1
+#endif // LINUX_VERSION_CODE == 4.6
+#else
+#if RHEL_RELEASE_CODE <= RHEL_RELEASE_VERSION(7,3) 
+#define ENABLE_MCOVERLAYFS 1
+#endif // RHEL_RELEASE_CODE <= 7.3
+#endif // RHEL_RELEASE_CODE
+#endif // ENABLE_MCOVERLAYFS
+
 typedef unsigned char   cc_t;
 typedef unsigned int    speed_t;
 typedef unsigned int    tcflag_t;
@@ -136,6 +155,10 @@ static const char rlimit_stack_envname[] = "MCKERNEL_RLIMIT_STACK";
 static int ischild;
 static int enable_vdso = 1;
 
+/* Partitioned execution (e.g., for MPI) */
+static int nr_processes = 0;
+static int nr_threads = -1;
+
 struct fork_sync {
 	pid_t pid;
 	int status;
@@ -375,7 +398,7 @@ struct program_load_desc *load_interp(struct program_load_desc *desc0, FILE *fp)
 
 unsigned char *dma_buf;
 
-int lookup_exec_path(char *filename, char *path, int max_len) 
+int lookup_exec_path(char *filename, char *path, int max_len, int execvp) 
 {
 	int found;
 	int error;
@@ -393,28 +416,27 @@ retry:
 
 			char *token, *string, *tofree;
 			char *PATH = getenv("COKERNEL_PATH");
-			if (!PATH) {
+
+			if (!execvp) {
+				if (strlen(filename) + 1 > max_len) {
+					return ENAMETOOLONG;
+				}
+				strcpy(path, filename);
+				error = access(path, X_OK);
+				if (error) {
+					return errno;
+				}
+				found = 1;
+				break;
+			}
+
+			if (!(PATH = getenv("COKERNEL_PATH"))) {
 				PATH = getenv("PATH");
 			}
 
 			if (strlen(filename) >= 255) {
 				return ENAMETOOLONG;
 			}
-			
-			/* See first whether file is available in current working dir */
-			error = access(filename, X_OK);
-			if (error == 0) {
-				__dprintf("lookup_exec_path(): found %s in cwd\n", filename);
-				error = snprintf(path, max_len, "%s", filename);
-
-				if (error < 0 || error >= max_len) {
-					fprintf(stderr, "lookup_exec_path(): array too small?\n");
-					return ENOMEM;
-				}
-
-				found = 1;
-				break;
-			}
 
 			__dprintf("PATH: %s\n", PATH);
 
@@ -442,6 +464,9 @@ retry:
 			}
 
 			free(tofree);
+			if(!found){
+				return ENOENT;
+			}
 			break;
 		}
 
@@ -483,7 +508,7 @@ retry:
 
 	/* Check whether the resolved path is a symlink */
 	if (lstat(path, &sb) == -1) {
-		fprintf(stderr, "lookup_exec_path(): error stat\n");
+		__dprintf(stderr, "lookup_exec_path(): error stat\n");
 		return errno;
 	}
 
@@ -654,7 +679,7 @@ int load_elf_desc(char *filename, struct program_load_desc **desc_p,
 	return 0;
 }
 
-void transfer_image(int fd, struct program_load_desc *desc)
+int transfer_image(int fd, struct program_load_desc *desc)
 {
 	struct remote_transfer pt;
 	unsigned long s, e, flen, rpa;
@@ -668,7 +693,10 @@ void transfer_image(int fd, struct program_load_desc *desc)
 		     + PAGE_SIZE - 1) & PAGE_MASK;
 		rpa = desc->sections[i].remote_pa;
 
-		fseek(fp, desc->sections[i].offset, SEEK_SET);
+		if (fseek(fp, desc->sections[i].offset, SEEK_SET) != 0) {
+			fprintf(stderr, "transfer_image(): error: seeking file position\n");
+			return -1;
+		}
 		flen = desc->sections[i].filesz;
 
 		__dprintf("seeked to %lx | size %ld\n",
@@ -690,7 +718,20 @@ void transfer_image(int fd, struct program_load_desc *desc)
 				if (lr > flen) {
 					lr = flen;
 				}
-				fread(dma_buf + l, 1, lr, fp); 
+				if (fread(dma_buf + l, 1, lr, fp) != lr) {
+					if (ferror(fp) > 0) {
+						fprintf(stderr, "transfer_image(): error: accessing file\n");
+						return -EINVAL;
+					}
+					else if (feof(fp) > 0) {
+						fprintf(stderr, "transfer_image(): file too short?\n");
+						return -EINVAL;
+					}
+					else {
+						/* TODO: handle smaller reads.. */
+						return -EINVAL;
+					}
+				}
 				flen -= lr;
 			} 
 			else if (flen > 0) {
@@ -699,7 +740,20 @@ void transfer_image(int fd, struct program_load_desc *desc)
 				} else {
 					lr = flen;
 				}
-				fread(dma_buf, 1, lr, fp);
+				if (fread(dma_buf, 1, lr, fp) != lr) {
+					if (ferror(fp) > 0) {
+						fprintf(stderr, "transfer_image(): error: accessing file\n");
+						return -EINVAL;
+					}
+					else if (feof(fp) > 0) {
+						fprintf(stderr, "transfer_image(): file too short?\n");
+						return -EINVAL;
+					}
+					else {
+						/* TODO: handle smaller reads.. */
+						return -EINVAL;
+					}
+				}
 				flen -= lr;
 			} 
 			s += PAGE_SIZE;
@@ -715,6 +769,8 @@ void transfer_image(int fd, struct program_load_desc *desc)
 			}
 		}
 	}
+
+	return 0;
 }
 
 void print_desc(struct program_load_desc *desc)
@@ -837,7 +893,10 @@ struct thread_data_s {
 	pthread_mutex_t *lock;
 	pthread_barrier_t *init_ready;
 } *thread_data;
+
 int ncpu;
+int n_threads;
+
 pid_t master_tid;
 
 pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
@@ -848,7 +907,7 @@ static void *main_loop_thread_func(void *arg)
 	struct thread_data_s *td = (struct thread_data_s *)arg;
 
 	td->tid = gettid();
-	td->remote_tid = (int)td->tid;
+	td->remote_tid = -1;
 	pthread_barrier_wait(&init_ready);
 	td->ret = main_loop(td->fd, td->cpu, td->lock);
 
@@ -931,7 +990,10 @@ act_signalfd4(struct syscall_wait_desc *w)
 			flags |= O_NONBLOCK;
 		if(tmp & SFD_CLOEXEC)
 			flags |= O_CLOEXEC;
-		pipe2(sfd->sigpipe, flags);
+		if (pipe2(sfd->sigpipe, flags) < 0) {
+			perror("pipe2 failed:");
+			return -1;
+		}
 		sfd->next = sigfdtop;
 		sigfdtop = sfd;
 		rc = sfd->sigpipe[0];
@@ -962,7 +1024,11 @@ act_signalfd4(struct syscall_wait_desc *w)
 			rc = -EBADF;
 		else{
 			info = (struct signalfd_siginfo *)w->sr.args[2];
-			write(sfd->sigpipe[1], info, sizeof(struct signalfd_siginfo));
+			if (write(sfd->sigpipe[1], info, sizeof(struct signalfd_siginfo))
+					!= sizeof(struct signalfd_siginfo)) {
+				fprintf(stderr, "error: writing sigpipe\n");
+				rc = -EBADF;
+			}
 		}
 		break;
 	}
@@ -1042,7 +1108,7 @@ static int reduce_stack(struct rlimit *orig_rlim, char *argv[])
 
 void print_usage(char **argv)
 {
-	fprintf(stderr, "Usage: %s [-c target_core] [<mcos-id>] (program) [args...]\n", argv[0]);
+	fprintf(stderr, "Usage: %s [-c target_core] [-n nr_partitions] [<mcos-id>] (program) [args...]\n", argv[0]);
 }
 
 void init_sigaction(void)
@@ -1068,9 +1134,9 @@ void init_worker_threads(int fd)
 	int i;
 
 	pthread_mutex_init(&lock, NULL);
-	pthread_barrier_init(&init_ready, NULL, ncpu + 2);
+	pthread_barrier_init(&init_ready, NULL, n_threads + 2);
 
-	for (i = 0; i <= ncpu; ++i) {
+	for (i = 0; i <= n_threads; ++i) {
 		int ret;
 
 		thread_data[i].fd = fd;
@@ -1091,79 +1157,43 @@ void init_worker_threads(int fd)
 }
 
 #ifdef ENABLE_MCOVERLAYFS
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,1,0)
 #define READ_BUFSIZE 1024
-static int isunshare(void)
+static int find_mount_prefix(char *prefix)
 {
-	int err = 0;
-	int ret;
-	int fd;
+	FILE *fp;
+	char *line = NULL;
+	size_t len = 0;
+	ssize_t read;
 	char proc_path[PATH_MAX];
-	ssize_t len_read;
-	char buf_read[READ_BUFSIZE + 1];
-	char *buf_read_off;
-	char *buf_find;
-	char buf_cmp[READ_BUFSIZE + 1];
-	char *buf_cmp_off;
-	ssize_t len_copy;
+	int ret = 0;
 
 	snprintf(proc_path, sizeof(proc_path), "/proc/%d/mounts", getpid());
-	fd = open(proc_path, O_RDONLY);
-	if (fd < 0) {
-		fprintf(stderr, "Error: Failed to open %s.\n", proc_path);
+
+	fp = fopen(proc_path, "r");
+	if (fp == NULL) {
 		return -1;
 	}
 
-	buf_cmp_off = buf_cmp;
-	while (1) {
-		len_read = read(fd, buf_read, READ_BUFSIZE);
-		if (len_read == -1) {
-			fprintf(stderr, "Error: Failed to read.\n");
-			err = -1;
-			break;
-		}
+	while ((read = getline(&line, &len, fp)) != -1) {
+		if (strlen(line) < strlen(prefix))
+			continue;
 
-		buf_read_off = buf_read;
-		while (1) {
-			if ((len_read - (buf_read_off - buf_read)) <= 0) {
-				break;
-			}
-			buf_find = memchr(buf_read_off, '\n', 
-				len_read - (buf_read_off - buf_read));
-			if (buf_find) {
-				len_copy = buf_find - buf_read_off;	
-			} else {
-				len_copy = len_read - (buf_read_off - buf_read);
-			}
-			memcpy(buf_cmp_off, buf_read_off, len_copy);
-			*(buf_cmp_off + len_copy) = '\0';
-
-			if (buf_find) {
-				buf_read_off = buf_read_off + len_copy + 1;
-				buf_cmp_off = buf_cmp;
-				ret = strncmp(buf_cmp, "mcoverlay /proc ", 16);
-				if (!ret) {
-					err = 1;
-					break;
-				}
-			} else {
-				buf_read_off = buf_read_off + len_copy;
-				buf_cmp_off = buf_cmp_off + len_copy;
-				break;
-			}
-		}
-
-		if (err == 1 || len_read == 0) {
+		if (!strncmp(line, prefix, strlen(prefix))) {
+			ret = 1;
 			break;
 		}
 	}
 
-	close(fd);
+	if (line)
+		free(line);
 
-	__dprintf("err=%d\n", err);
-	return err;
+	return ret;
+}
+
+static int isunshare(void)
+{
+	return find_mount_prefix("mcoverlay /proc ");
 }
-#endif
 #endif // ENABLE_MCOVERLAYFS
 
 #define MCK_RLIMIT_AS	0
@@ -1305,12 +1335,20 @@ int main(int argc, char **argv)
 	}
            
 	/* Parse options ("+" denotes stop at the first non-option) */
-	while ((opt = getopt_long(argc, argv, "+c:", mcexec_options, NULL)) != -1) {
+	while ((opt = getopt_long(argc, argv, "+c:n:t:", mcexec_options, NULL)) != -1) {
 		switch (opt) {
 			case 'c':
 				target_core = atoi(optarg);
 				break;
-			
+
+			case 'n':
+				nr_processes = atoi(optarg);
+				break;
+
+			case 't':
+				nr_threads = atoi(optarg);
+				break;
+
 			case 0:	/* long opt */
 				break;
 
@@ -1353,7 +1391,6 @@ int main(int argc, char **argv)
 	}
 
 #ifdef ENABLE_MCOVERLAYFS
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,1,0)
 	__dprintf("mcoverlay enable\n");
 	char mcos_procdir[PATH_MAX];
 	char mcos_sysdir[PATH_MAX];
@@ -1362,6 +1399,7 @@ int main(int argc, char **argv)
 	if (error == 0) {
 		struct sys_unshare_desc unshare_desc;
 		struct sys_mount_desc mount_desc;
+		struct sys_umount_desc umount_desc;
 
 		memset(&unshare_desc, '\0', sizeof unshare_desc);
 		memset(&mount_desc, '\0', sizeof mount_desc);
@@ -1373,6 +1411,53 @@ int main(int argc, char **argv)
 			return 1;
 		}
 
+		/*
+		 * Umount cgroup filesystems that may expose invalid NUMA
+		 * information
+		 */
+		if (find_mount_prefix("cgroup /sys/fs/cgroup/cpu,cpuacct")) {
+			umount_desc.dir_name = "/sys/fs/cgroup/cpu,cpuacct";
+
+			if (ioctl(fd, MCEXEC_UP_SYS_UMOUNT,
+						(unsigned long)&umount_desc) != 0) {
+				fprintf(stderr,
+						"WARNING: Failed to umount cgroup/cpu,cpuacct. (%s)\n",
+						strerror(errno));
+			}
+		}
+		else if (find_mount_prefix("cgroup /sys/fs/cgroup/cpu")) {
+			umount_desc.dir_name = "/sys/fs/cgroup/cpu";
+
+			if (ioctl(fd, MCEXEC_UP_SYS_UMOUNT,
+						(unsigned long)&umount_desc) != 0) {
+				fprintf(stderr,
+						"WARNING: Failed to umount cgroup/cpu. (%s)\n",
+						strerror(errno));
+			}
+		}
+
+		if (find_mount_prefix("cgroup /sys/fs/cgroup/cpuset")) {
+			umount_desc.dir_name = "/sys/fs/cgroup/cpuset";
+
+			if (ioctl(fd, MCEXEC_UP_SYS_UMOUNT,
+						(unsigned long)&umount_desc) != 0) {
+				fprintf(stderr,
+						"WARNING: Failed to umount cgroup/cpuset. (%s)\n",
+						strerror(errno));
+			}
+		}
+
+		if (find_mount_prefix("cgroup /sys/fs/cgroup/memory")) {
+			umount_desc.dir_name = "/sys/fs/cgroup/memory/";
+
+			if (ioctl(fd, MCEXEC_UP_SYS_UMOUNT,
+						(unsigned long)&umount_desc) != 0) {
+				fprintf(stderr,
+						"WARNING: Failed to umount cgroup/memory. (%s)\n",
+						strerror(errno));
+			}
+		}
+
 		sprintf(mcos_procdir, "/tmp/mcos/mcos%d_proc", mcosid);
 		mount_desc.dev_name = mcos_procdir;
 		mount_desc.dir_name = "/proc";
@@ -1401,12 +1486,11 @@ int main(int argc, char **argv)
 	} else if (error == -1) {
 		return 1;
 	}
-#endif
 #else
 	__dprintf("mcoverlay disable\n");
 #endif // ENABLE_MCOVERLAYFS
 
-	if (lookup_exec_path(argv[optind], path, sizeof(path)) != 0) {
+	if (lookup_exec_path(argv[optind], path, sizeof(path), 1) != 0) {
 		fprintf(stderr, "error: finding file: %s\n", argv[optind]);
 		return 1;
 	}
@@ -1418,7 +1502,7 @@ int main(int argc, char **argv)
 
 	/* Check whether shell script */
 	if (shell) {
-		if (lookup_exec_path(shell, shell_path, sizeof(shell_path)) != 0) {
+		if (lookup_exec_path(shell, shell_path, sizeof(shell_path), 0) != 0) {
 			fprintf(stderr, "error: finding file: %s\n", shell);
 			return 1;
 		}
@@ -1480,7 +1564,30 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
+	if (nr_threads > 0) {
+		n_threads = nr_threads;
+	}
+	else if (getenv("OMP_NUM_THREADS")) {
+		/* Leave some headroom for helper threads.. */
+		n_threads = atoi(getenv("OMP_NUM_THREADS")) + 4;
+	}
+	else {
+		n_threads = ncpu;
+	}
+
+	/* 
+	 * XXX: keep thread_data ncpu sized despite that there are only
+	 * n_threads worker threads in the pool so that signaling code
+	 * keeps working.
+	 *
+	 * TODO: fix signaling code to be independent of TIDs.
+	 * TODO: implement dynaic thread pool resizing.
+	 */
 	thread_data = (struct thread_data_s *)malloc(sizeof(struct thread_data_s) * (ncpu + 1));
+	if (!thread_data) {
+		fprintf(stderr, "error: allocating thread pool data\n");
+		return 1;
+	}
 	memset(thread_data, '\0', sizeof(struct thread_data_s) * (ncpu + 1));
 
 #if 0	
@@ -1515,6 +1622,53 @@ int main(int argc, char **argv)
 		exit(1);
 	}
 
+	/* Partitioned execution, obtain CPU set */
+	if (nr_processes > 0) {
+		struct get_cpu_set_arg cpu_set_arg;
+		int mcexec_linux_numa = 0;
+
+		cpu_set_arg.cpu_set = (void *)&desc->cpu_set;
+		cpu_set_arg.cpu_set_size = sizeof(desc->cpu_set);
+		cpu_set_arg.nr_processes = nr_processes;
+		cpu_set_arg.target_core = &target_core;
+		cpu_set_arg.mcexec_linux_numa = &mcexec_linux_numa;
+
+		if (ioctl(fd, MCEXEC_UP_GET_CPUSET, (void *)&cpu_set_arg) != 0) {
+			perror("getting CPU set for partitioned execution");
+			close(fd);
+			return 1;
+		}
+
+		desc->cpu = target_core;
+
+		/* This call may not succeed, but that is fine */
+		if (numa_run_on_node(mcexec_linux_numa) < 0) {
+			__dprint("%s: WARNING: couldn't bind to NUMA %d\n",
+				__FUNCTION__, mcexec_linux_numa);
+		}
+#ifdef DEBUG
+		else {
+			cpu_set_t cpuset;
+			char affinity[BUFSIZ];
+
+			CPU_ZERO(&cpuset);
+			if ((sched_getaffinity(0, sizeof(cpu_set_t), &cpuset)) != 0) {
+				perror("Error sched_getaffinity");
+				exit(1);
+			}
+
+			affinity[0] = '\0';
+			for (i = 0; i < 512; i++) {
+				if (CPU_ISSET(i, &cpuset) == 1) {
+					sprintf(affinity, "%s %d", affinity, i);
+				}
+			}
+			__dprint("%s: PID: %d affinity: %s\n",
+					__FUNCTION__, getpid(), affinity);
+		}
+#endif
+	}
+
 	if (ioctl(fd, MCEXEC_UP_PREPARE_IMAGE, (unsigned long)desc) != 0) {
 		perror("prepare");
 		close(fd);
@@ -1522,7 +1676,10 @@ int main(int argc, char **argv)
 	}
 
 	print_desc(desc);
-	transfer_image(fd, desc);
+	if (transfer_image(fd, desc) < 0) {
+		fprintf(stderr, "error: transferring image\n");
+		return -1;
+	}
 	fflush(stdout);
 	fflush(stderr);
 	
@@ -1561,7 +1718,7 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
-	for (i = 0; i <= ncpu; ++i) {
+	for (i = 0; i <= n_threads; ++i) {
 		pthread_join(thread_data[i].thread_id, NULL);
 	}
 
@@ -1618,21 +1775,111 @@ do_generic_syscall(
 		ret = -errno;
 	}
 
+	/* Overlayfs /sys/X directory lseek() problem work around */
+	if (w->sr.number == __NR_lseek && ret == -EINVAL) {
+		char proc_path[PATH_MAX];
+		char path[PATH_MAX];
+		struct stat sb;
+
+		sprintf(proc_path, "/proc/self/fd/%d", (int)w->sr.args[0]);
+
+		/* Get filename */
+		if (readlink(proc_path, path, sizeof(path)) < 0) {
+			fprintf(stderr, "%s: error: readlink() failed for %s\n",
+				__FUNCTION__, proc_path);
+			perror(": ");
+			goto out;
+		}
+
+		/* Not in /sys? */
+		if (strncmp(path, "/sys/", 5))
+			goto out;
+
+		/* Stat */
+		if (stat(path, &sb) < 0) {
+			fprintf(stderr, "%s: error stat() failed for %s\n",
+				__FUNCTION__, path);
+			goto out;
+		}
+
+		/* Not dir? */
+		if ((sb.st_mode & S_IFMT) != S_IFDIR)
+			goto out;
+
+		ret = 0;
+	}
+	/* Fake that nodeX in /sys/devices/system/node do not exist,
+	 * where X >= number of LWK NUMA nodes */
+	else if (w->sr.number == __NR_getdents && ret > 0) {
+		struct linux_dirent {
+			long           d_ino;
+			off_t          d_off;
+			unsigned short d_reclen;
+			char           d_name[];
+		};
+		struct linux_dirent *d;
+		char *buf = (char *)w->sr.args[1];
+		int bpos = 0;
+		int nodes,len;
+		char proc_path[PATH_MAX];
+		char path[PATH_MAX];
+
+		sprintf(proc_path, "/proc/self/fd/%d", (int)w->sr.args[0]);
+
+		/* Get filename */
+		if ((len = readlink(proc_path, path, sizeof(path))) < 0) {
+			fprintf(stderr, "%s: error: readlink() failed for %s\n",
+				__FUNCTION__, proc_path);
+			goto out;
+		}
+		path[len] = 0;
+
+		/* Not /sys/devices/system/node ? */
+		if (strcmp(path, "/sys/devices/system/node"))
+			goto out;
+
+		nodes = ioctl(fd, MCEXEC_UP_GET_NODES, 0);
+		if (nodes == -1) {
+			goto out;
+		}
+
+		d = (struct linux_dirent *) (buf + bpos);
+		for (bpos = 0; bpos < ret; ) {
+			int nodeid, tmp_reclen;
+			d = (struct linux_dirent *) (buf + bpos);
+
+			if (sscanf(d->d_name, "node%d", &nodeid) != 1) {
+				bpos += d->d_reclen;
+				continue;
+			}
+
+			if (nodeid >= nodes) {
+				tmp_reclen = d->d_reclen;
+				memmove(buf + bpos,
+						buf + bpos + tmp_reclen,
+						ret - bpos - tmp_reclen);
+				ret -= tmp_reclen;
+				continue;
+			}
+
+			bpos += d->d_reclen;
+		}
+	}
+
+out:
 	__dprintf("do_generic_syscall(%ld):%ld (%#lx)\n", w->sr.number, ret, ret);
 	return ret;
 }
 
 static void
-kill_thread(unsigned long cpu)
+kill_thread(unsigned long tid)
 {
-	if(cpu >= 0 && cpu < ncpu){
-		pthread_kill(thread_data[cpu].thread_id, LOCALSIG);
-	}
-	else{
-		int	i;
+	int	i;
 
-		for (i = 0; i < ncpu; ++i) {
+	for (i = 0; i <= n_threads; ++i) {
+		if(thread_data[i].remote_tid == tid){
 			pthread_kill(thread_data[i].thread_id, LOCALSIG);
+			break;
 		}
 	}
 }
@@ -1734,13 +1981,20 @@ int close_cloexec_fds(int mcos_fd)
 	return 0;
 }
 
+void chgdevpath(char *in, char *buf)
+{
+	if(!strcmp(in, "/dev/xpmem")){
+		sprintf(in, "/dev/null");
+	}
+}
+
 char *
 chgpath(char *in, char *buf)
 {
+	chgdevpath(in, buf);
+
 #ifdef ENABLE_MCOVERLAYFS
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,1,0)
 	return in;
-#endif
 #endif // ENABLE_MCOVERLAYFS
 	char	*fn = in;
 	struct stat	sb;
@@ -1791,6 +2045,8 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 		
 		//pthread_mutex_lock(lock);
 
+		thread_data[cpu].remote_tid = w.sr.rtid;
+
 		switch (w.sr.number) {
 		case __NR_open:
 			ret = do_strncpy_from_user(fd, pathbuf, (void *)w.sr.args[0], PATH_MAX);
@@ -1829,13 +2085,13 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 			sig = 0;
 			term = 0;
 			
+			do_syscall_return(fd, cpu, 0, 0, 0, 0, 0);
+
 			/* Drop executable file */
 			if ((ret = ioctl(fd, MCEXEC_UP_CLOSE_EXEC)) != 0) {
 				fprintf(stderr, "WARNING: close_exec() couldn't find exec file?\n");
 			}
 
-			do_syscall_return(fd, cpu, 0, 0, 0, 0, 0);
-
 			__dprintf("__NR_exit/__NR_exit_group: %ld (cpu_id: %d)\n",
 					w.sr.args[0], cpu);
 			if(w.sr.number == __NR_exit_group){
@@ -1888,22 +2144,40 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 #endif
 
 		case __NR_gettid:{
-			int mode = w.sr.args[0];
-			int remote_pid = w.sr.args[1];
-			int newcpuid = w.sr.args[2];
-			int oldcpuid = w.sr.args[3];
-			int wtid = thread_data[newcpuid].remote_tid;
+			/*
+			 * Number of TIDs and the remote physical address where TIDs are
+			 * expected are passed in arg 4 and 5, respectively.
+			 */
+			if (w.sr.args[4] > 0) {
+				struct remote_transfer trans;
+				int i = 0;
+				int *tids = malloc(sizeof(int) * w.sr.args[4]);
+				if (!tids) {
+					fprintf(stderr, "__NR_gettid(): error allocating TIDs\n");
+					goto gettid_out;
+				}
 
-			if(mode == 0){
-				thread_data[ncpu].remote_tid = wtid;
-				thread_data[newcpuid].remote_tid = remote_pid;
-			}
-			else if(mode == 2){
-				thread_data[newcpuid].remote_tid = thread_data[oldcpuid].remote_tid;
-				thread_data[oldcpuid].remote_tid = wtid;
-			}
+				for (i = 0; i < ncpu && i < w.sr.args[4]; ++i) {
+					tids[i] = thread_data[i].tid;
+				}
 
-			do_syscall_return(fd, cpu, thread_data[newcpuid].remote_tid, 0, 0, 0, 0);
+				for (; i < ncpu; ++i) {
+					tids[i] = 0;
+				}
+
+				trans.userp = (void*)tids;
+				trans.rphys = w.sr.args[5];
+				trans.size = sizeof(int) * w.sr.args[4];
+				trans.direction = MCEXEC_UP_TRANSFER_TO_REMOTE;
+
+				if (ioctl(fd, MCEXEC_UP_TRANSFER, &trans) != 0) {
+					fprintf(stderr, "__NR_gettid(): error transfering TIDs\n");
+				}
+
+				free(tids);
+			}
+gettid_out:
+			do_syscall_return(fd, cpu, 0, 0, 0, 0, 0);
 			break;
 		}
 
@@ -1945,7 +2219,9 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 					close(pipefds[0]);
 					pid = fork();
 					if(pid != 0){
-						write(pipefds[1], &pid, sizeof pid);
+						if (write(pipefds[1], &pid, sizeof pid) != sizeof(pid)) {
+							fprintf(stderr, "error: writing pipefds\n");
+						}
 						exit(0);
 					}
 				}
@@ -1954,7 +2230,9 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 					int st;
 
 					close(pipefds[1]);
-					read(pipefds[0], &npid, sizeof npid);
+					if (read(pipefds[0], &npid, sizeof npid) != sizeof(npid)) {
+						fprintf(stderr, "error: reading pipefds\n");
+					}
 					close(pipefds[0]);
 					waitpid(pid, &st, 0);
 					pid = npid;
@@ -1994,7 +2272,6 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 
 				/* Reinit signals and syscall threads */
 				init_sigaction();
-				init_worker_threads(fd);
 
 				__dprintf("pid(%d): signals and syscall threads OK\n", 
 						getpid());
@@ -2008,6 +2285,8 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 					goto fork_child_sync_pipe;
 				}
 
+				init_worker_threads(fd);
+
 fork_child_sync_pipe:
 				sem_post(&fs->sem);
 				if (fs->status)
@@ -2118,7 +2397,7 @@ fork_err:
 					shell = NULL;
 					filename = (char *)w.sr.args[1];
 					
-					if ((ret = lookup_exec_path(filename, path, sizeof(path))) 
+					if ((ret = lookup_exec_path(filename, path, sizeof(path), 0)) 
 						!= 0) {
 						goto return_execve1;
 					}
@@ -2132,7 +2411,7 @@ fork_err:
 					/* Check whether shell script */
 					if (shell) {
 						if ((ret = lookup_exec_path(shell, shell_path, 
-									sizeof(shell_path))) != 0) {
+									sizeof(shell_path), 0)) != 0) {
 							fprintf(stderr, "execve(): error: finding file: %s\n", shell);
 							goto return_execve1;
 						}
@@ -2153,6 +2432,7 @@ fork_err:
 						strcpy(desc->shell_path, shell_path);
 					}
 
+					desc->enable_vdso = enable_vdso;
 					__dprintf("execve(): load_elf_desc() for %s OK, num sections: %d\n",
 						path, desc->num_sections);
 
@@ -2210,7 +2490,10 @@ return_execve1:
 					
 					__dprintf("%s", "execve(): transfer ELF desc OK\n");
 
-					transfer_image(fd, desc);	
+					if (transfer_image(fd, desc) != 0) {
+						fprintf(stderr, "error: transferring image\n");
+						return -1;
+					}
 					__dprintf("%s", "execve(): image transferred\n");
 
 					if (close_cloexec_fds(fd) < 0) {
@@ -2220,6 +2503,23 @@ return_execve1:
 
 					ret = 0;
 return_execve2:					
+#ifdef ENABLE_MCOVERLAYFS
+				{
+					struct sys_mount_desc mount_desc;
+
+					mount_desc.dev_name = NULL;
+					mount_desc.dir_name = "/proc";
+					mount_desc.type = NULL;
+					mount_desc.flags = MS_REMOUNT;
+					mount_desc.data = NULL;
+					if (ioctl(fd, MCEXEC_UP_SYS_MOUNT,
+								(unsigned long)&mount_desc) != 0) {
+						fprintf(stderr,
+								"WARNING: failed to remount /proc (%s)\n",
+								strerror(errno));
+					}
+				}
+#endif
 					do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 					break;
 
@@ -2262,6 +2562,53 @@ return_execve2:
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
 
+		case __NR_setresuid:
+			ret = setresuid(w.sr.args[0], w.sr.args[1], w.sr.args[2]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setreuid:
+			ret = setreuid(w.sr.args[0], w.sr.args[1]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setuid:
+			ret = setuid(w.sr.args[0]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setresgid:
+			ret = setresgid(w.sr.args[0], w.sr.args[1], w.sr.args[2]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setregid:
+			ret = setregid(w.sr.args[0], w.sr.args[1]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setgid:
+			ret = setgid(w.sr.args[0]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setfsgid:
+			ret = setfsgid(w.sr.args[0]);
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
 		case __NR_close:
 			if(w.sr.args[0] == fd)
 				ret = -EBADF;
@@ -2289,13 +2636,33 @@ return_execve2:
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
 
+		case __NR_stat:
+			ret = do_strncpy_from_user(fd, pathbuf, (void *)w.sr.args[0], PATH_MAX);
+			if (ret >= PATH_MAX) {
+				ret = -ENAMETOOLONG;
+			}
+			if (ret < 0) {
+				do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+				break;
+			}
+
+			fn = chgpath(pathbuf, tmpbuf);
+
+			ret = stat(fn, (struct stat *)w.sr.args[1]);
+			__dprintf("stat: path=%s, ret=%ld\n", fn, ret);
+			SET_ERR(ret);
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
 		default:
 			ret = do_generic_syscall(&w);
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
 
 		}
-		
+
+		thread_data[cpu].remote_tid = -1;
+
 		//pthread_mutex_unlock(lock);
 	}
 	__dprint("timed out.\n");

kernel/Makefile.build.in

@@ -1,17 +1,22 @@
+ENABLE_RUSAGE=@ENABLE_RUSAGE@
 VPATH=@abs_srcdir@
 SRC=$(VPATH)
 IHKDIR=$(IHKBASE)/$(TARGETDIR)
 OBJS = init.o mem.o debug.o mikc.o listeners.o ap.o syscall.o cls.o host.o
 OBJS += process.o copy.o waitq.o futex.o timer.o plist.o fileobj.o shmobj.o
-OBJS += zeroobj.o procfs.o devobj.o sysfs.o
+OBJS += zeroobj.o procfs.o devobj.o sysfs.o xpmem.o rusage.o freeze.o
 DEPSRCS=$(wildcard $(SRC)/*.c)
+DOPT=
+ifeq ($(ENABLE_RUSAGE),yes)
+DOPT=-DENABLE_RUSAGE
+endif
 
-CFLAGS += -I$(SRC)/include -D__KERNEL__ -g
+CFLAGS += -I$(SRC)/include -D__KERNEL__ $(DOPT) -g -fno-omit-frame-pointer -fno-inline -fno-inline-small-functions
 LDFLAGS += -e arch_start
 IHKOBJ = ihk/ihk.o
 
 include $(SRC)/config/config.$(TARGET)
-include $(IHKBASE)/Makefile.common
+include @abs_builddir@/../../ihk/cokernel/Makefile.common
 
 # CFLAGS += -I$(SRC)/../arch/$(IHKARCH)/kernel/include -I$(SRC)/../lib/include
 

kernel/Makefile.in

@@ -9,7 +9,7 @@ V ?= $(VERBOSE)
 KERNEL = kernel.img
 KERNELS = $(addsuffix /$(KERNEL),$(addprefix $(O)/,$(BUILD_TARGET)))
 
-SUBCMD_OPTS = V='$(V)'
+SUBCMD_OPTS = V='$(V)' BUILD_IHK_COKERNEL=@abs_builddir@/../../ihk/cokernel
 
 $(if $(O),,$(error Specify the compilation target directory))
 #$(if $(shell ls $(IHKBASE)/Makefile),,\

kernel/ap.c

@@ -25,10 +25,27 @@
 #include <init.h>
 #include <march.h>
 #include <cls.h>
+#include <time.h>
+#include <syscall.h>
+#include <rusage.h>
+
+//#define DEBUG_PRINT_AP
+
+#ifdef DEBUG_PRINT_AP
+#define dkprintf(...) kprintf(__VA_ARGS__)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#endif
 
 int num_processors = 1;
 static volatile int ap_stop = 1;
 
+mcs_lock_node_t ap_syscall_semaphore;
+
+extern struct ihk_os_monitor *monitor;
+
 static void ap_wait(void)
 {
 	init_tick();
@@ -43,7 +60,11 @@ static void ap_wait(void)
 	arch_start_pvclock();
 
 	if (find_command_line("hidos")) {
+		mcs_lock_node_t mcs_node;
+
+		mcs_lock_lock_noirq(&ap_syscall_semaphore, &mcs_node);
 		init_host_syscall_channel();
+		mcs_lock_unlock_noirq(&ap_syscall_semaphore, &mcs_node);
 	}
 	
 	pc_ap_init();
@@ -57,6 +78,7 @@ static void ap_wait(void)
 void ap_start(void)
 {
 	init_tick();
+	mcs_lock_init(&ap_syscall_semaphore);
 	ap_stop = 0;
 	sync_tick();
 }
@@ -65,7 +87,7 @@ void ap_init(void)
 {
 	struct ihk_mc_cpu_info *cpu_info;
 	int i;
-	int bsp_hw_id;
+	int bsp_hw_id, bsp_cpu_id;
 
 	ihk_mc_init_ap();
 	init_delay();
@@ -78,18 +100,32 @@ void ap_init(void)
 		return;
 	}
 
-	kprintf("BSP HW ID = %d\n", bsp_hw_id);
+	bsp_cpu_id = 0;
+	for (i = 0; i < cpu_info->ncpus; ++i) {
+		if (cpu_info->hw_ids[i] == bsp_hw_id) {
+			bsp_cpu_id = i;
+			break;
+		}
+	}
+
+	kprintf("BSP: %d (HW ID: %d @ NUMA %d)\n", bsp_cpu_id,
+			bsp_hw_id, cpu_info->nodes[0]);
 
 	for (i = 0; i < cpu_info->ncpus; i++) {
 		if (cpu_info->hw_ids[i] == bsp_hw_id) {
 			continue;
 		}
-		kprintf("AP Booting: %d (HW ID: %d)\n", i, cpu_info->hw_ids[i]);
+		dkprintf("AP Booting: %d (HW ID: %d @ NUMA %d)\n", i,
+			cpu_info->hw_ids[i], cpu_info->nodes[i]);
 		ihk_mc_boot_cpu(cpu_info->hw_ids[i], (unsigned long)ap_wait);
 
 		num_processors++;
 	}
-	kprintf("AP Booting: Done\n");
+	kprintf("BSP: booted %d AP CPUs\n", cpu_info->ncpus - 1);
+#ifdef ENABLE_RUSAGE
+    rusage_num_threads = 0;
+    rusage_max_num_threads = 0;
+#endif
 }
 
 #include <sysfs.h>
@@ -199,7 +235,7 @@ cpu_sysfs_setup(void)
 	/* setup table */
 	info = kmalloc(sizeof(*info) * num_processors, IHK_MC_AP_CRITICAL);
 	for (cpu = 0; cpu < num_processors; ++cpu) {
-		info[cpu].online = 10+cpu;
+		info[cpu].online = 1;
 	}
 	fake_cpu_infos = info;
 

kernel/cls.c

@@ -23,17 +23,23 @@
 extern int num_processors;
 
 struct cpu_local_var *clv;
-static int cpu_local_var_initialized = 0;
+int cpu_local_var_initialized = 0;
+extern struct ihk_os_monitor *monitor;
 
 void cpu_local_var_init(void)
 {
 	int z;
+	int i;
 
 	z = sizeof(struct cpu_local_var) * num_processors;
 	z = (z + PAGE_SIZE - 1) >> PAGE_SHIFT;
 
-	clv = allocate_pages(z, IHK_MC_AP_CRITICAL);
+	clv = ihk_mc_alloc_pages(z, IHK_MC_AP_CRITICAL);
 	memset(clv, 0, z * PAGE_SIZE);
+
+	for(i = 0; i < num_processors; i++)
+		clv[i].monitor = monitor + i;
+
 	cpu_local_var_initialized = 1;
 }
 

kernel/debug.c

@@ -37,6 +37,8 @@ static void kprintf_wait(int len, unsigned long *flags_head, int *slide) {
 		if (head < tail) head += buf_len;
 		if (tail + len > buf_len) adj = buf_len - tail;
 		if (head > tail && head <= tail + len + adj) {
+			/* When proceeding tail (producer pointer) by len would
+			   cross head (consumer pointer) in ring-buffer */
 			if (mode != 1) {
 				*slide = 1;
 				break;
@@ -70,6 +72,9 @@ void kputs(char *buf)
 	
 	memcpy(kmsg_buf.str + kmsg_buf.tail, buf, len);
 	kmsg_buf.tail += len;
+	/* When proceeding tail (producer pointer) by len would
+	   cross head (consumer pointer) in ring-buffer, give up
+	   [head, tail] because the range is overwritten */
 	if (slide == 1) {
 		kmsg_buf.head = kmsg_buf.tail + 1;
 		if (kmsg_buf.head >= kmsg_buf.len) kmsg_buf.head = 0;
@@ -110,6 +115,7 @@ int __kprintf(const char *format, ...)
 	char buf[KPRINTF_LOCAL_BUF_LEN];
 
 	/* Copy into the local buf */
+	len = sprintf(buf, "[%3d]: ", ihk_mc_get_processor_id());
 	va_start(va, format);
 	len += vsnprintf(buf + len, KPRINTF_LOCAL_BUF_LEN - len - 2, format, va);
 	va_end(va);
@@ -169,6 +175,17 @@ int kprintf(const char *format, ...)
 	return len;
 }
 
+/* mode:
+    0: mcklogd is not running.
+       When kmsg buffer is full, writer doesn't block
+       and overwrites the buffer.
+    1: mcklogd periodically retrieves kmsg.
+       When kmsg buffer is full, writer blocks until 
+       someone retrieves kmsg.
+    2: mcklogd periodically retrieves kmsg.
+       When kmsg buffer is full, writer doesn't block
+       and overwrites the buffer.
+*/
 void kmsg_init(int mode)
 {
 	ihk_mc_spinlock_init(&kmsg_lock);

kernel/devobj.c

@@ -78,54 +78,56 @@ static struct memobj *to_memobj(struct devobj *devobj)
 /***********************************************************************
  * devobj
  */
-int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxprotp)
+int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxprotp,
+	int prot, int populate_flags)
 {
 	ihk_mc_user_context_t ctx;
 	struct pager_map_result result;	// XXX: assumes contiguous physical
 	int error;
 	struct devobj *obj  = NULL;
 	const size_t npages = (len + PAGE_SIZE - 1) / PAGE_SIZE;
+	const size_t pfn_npages = (npages / (PAGE_SIZE / sizeof(uintptr_t))) + 1;
 
-	dkprintf("devobj_create(%d,%lx,%lx)\n", fd, len, off);
-#define MAX_PAGES_IN_DEVOBJ (PAGE_SIZE / sizeof(uintptr_t))
-	if (npages > MAX_PAGES_IN_DEVOBJ) {
-		error = -EFBIG;
-		kprintf("devobj_create(%d,%lx,%lx):too large len. %d\n", fd, len, off, error);
-		goto out;
-	}
+	dkprintf("%s: fd: %d, len: %lu, off: %lu \n", __FUNCTION__, fd, len, off);
 
 	obj = kmalloc(sizeof(*obj), IHK_MC_AP_NOWAIT);
 	if (!obj) {
 		error = -ENOMEM;
-		kprintf("devobj_create(%d,%lx,%lx):kmalloc failed. %d\n", fd, len, off, error);
+		kprintf("%s: error: fd: %d, len: %lu, off: %lu kmalloc failed.\n", 
+			__FUNCTION__, fd, len, off);
 		goto out;
 	}
 	memset(obj, 0, sizeof(*obj));
 
-	obj->pfn_table = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	obj->pfn_table = ihk_mc_alloc_pages(pfn_npages, IHK_MC_AP_NOWAIT);
 	if (!obj->pfn_table) {
 		error = -ENOMEM;
-		kprintf("devobj_create(%d,%lx,%lx):allocate_pages failed. %d\n", fd, len, off, error);
+		kprintf("%s: error: fd: %d, len: %lu, off: %lu allocating PFN failed.\n", 
+			__FUNCTION__, fd, len, off);
 		goto out;
 	}
-	memset(obj->pfn_table, 0, 1*PAGE_SIZE);
+	memset(obj->pfn_table, 0, pfn_npages * PAGE_SIZE);
 
 	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_MAP;
 	ihk_mc_syscall_arg1(&ctx) = fd;
 	ihk_mc_syscall_arg2(&ctx) = len;
 	ihk_mc_syscall_arg3(&ctx) = off;
 	ihk_mc_syscall_arg4(&ctx) = virt_to_phys(&result);
+	ihk_mc_syscall_arg5(&ctx) = prot | populate_flags;
 
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
 	if (error) {
-		kprintf("devobj_create(%d,%lx,%lx):map failed. %d\n", fd, len, off, error);
+		kprintf("%s: error: fd: %d, len: %lu, off: %lu map failed.\n", 
+			__FUNCTION__, fd, len, off);
 		goto out;
 	}
-	dkprintf("devobj_create:handle: %lx\n", result.handle);
-	dkprintf("devobj_create:maxprot: %x\n", result.maxprot);
+
+	dkprintf("%s: fd: %d, len: %lu, off: %lu, handle: %p, maxprot: %x\n", 
+		__FUNCTION__, fd, len, off, result.handle, result.maxprot);
 
 	obj->memobj.ops = &devobj_ops;
-	obj->memobj.flags = MF_HAS_PAGER;
+	obj->memobj.flags = MF_HAS_PAGER | MF_DEV_FILE;
+	obj->memobj.size = len;
 	obj->handle = result.handle;
 	obj->ref = 1;
 	obj->pfn_pgoff = off / PAGE_SIZE;
@@ -140,11 +142,12 @@ int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxp
 out:
 	if (obj) {
 		if (obj->pfn_table) {
-			free_pages(obj->pfn_table, 1);
+			ihk_mc_free_pages(obj->pfn_table, pfn_npages);
 		}
 		kfree(obj);
 	}
-	dkprintf("devobj_create(%d,%lx,%lx): %d %p %x%d\n", fd, len, off, error, *objp, *maxprotp);
+	dkprintf("%s: ret: %d, fd: %d, len: %lu, off: %lu, handle: %p, maxprot: %x \n", 
+		__FUNCTION__, error, fd, len, off, result.handle, result.maxprot);
 	return error;
 }
 
@@ -164,6 +167,8 @@ static void devobj_release(struct memobj *memobj)
 	struct devobj *obj = to_devobj(memobj);
 	struct devobj *free_obj = NULL;
 	uintptr_t handle;
+	const size_t pfn_npages =
+		(obj->npages / (PAGE_SIZE / sizeof(uintptr_t))) + 1;
 
 	dkprintf("devobj_release(%p %lx)\n", obj, obj->handle);
 
@@ -176,23 +181,25 @@ static void devobj_release(struct memobj *memobj)
 	memobj_unlock(&obj->memobj);
 
 	if (free_obj) {
-		int error;
-		ihk_mc_user_context_t ctx;
+		if (!(free_obj->memobj.flags & MF_HOST_RELEASED)) {
+			int error;
+			ihk_mc_user_context_t ctx;
 
-		ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_UNMAP;
-		ihk_mc_syscall_arg1(&ctx) = handle;
-		ihk_mc_syscall_arg2(&ctx) = 1;
+			ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_UNMAP;
+			ihk_mc_syscall_arg1(&ctx) = handle;
+			ihk_mc_syscall_arg2(&ctx) = 1;
 
-		error = syscall_generic_forwarding(__NR_mmap, &ctx);
-		if (error) {
-			kprintf("devobj_release(%p %lx):"
-					"release failed. %d\n",
-					free_obj, handle, error);
-			/* through */
+			error = syscall_generic_forwarding(__NR_mmap, &ctx);
+			if (error) {
+				kprintf("devobj_release(%p %lx):"
+						"release failed. %d\n",
+						free_obj, handle, error);
+				/* through */
+			}
 		}
 
 		if (obj->pfn_table) {
-			free_pages(obj->pfn_table, 1);
+			ihk_mc_free_pages(obj->pfn_table, pfn_npages);
 		}
 		kfree(free_obj);
 	}
@@ -204,7 +211,7 @@ static void devobj_release(struct memobj *memobj)
 
 static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *flag)
 {
-	const off_t pgoff = off >> PAGE_SHIFT;
+	const off_t pgoff = off / PAGE_SIZE;
 	struct devobj *obj = to_devobj(memobj);
 	int error;
 	uintptr_t pfn;
@@ -216,7 +223,7 @@ static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintpt
 
 	if ((pgoff < obj->pfn_pgoff) || ((obj->pfn_pgoff + obj->npages) <= pgoff)) {
 		error = -EFBIG;
-		kprintf("devobj_get_page(%p %lx,%lx,%d): out of range. %d\n", memobj, obj->handle, off, p2align, error);
+		kprintf("%s: error: out of range: off: %lu, page off: %lu obj->npages: %d\n", __FUNCTION__, off, pgoff, obj->npages);
 		goto out;
 	}
 	ix = pgoff - obj->pfn_pgoff;

kernel/fileobj.c

@@ -29,22 +29,27 @@
 #define	dkprintf(...)	do { if (0) kprintf(__VA_ARGS__); } while (0)
 #define	ekprintf(...)	kprintf(__VA_ARGS__)
 
-static ihk_spinlock_t fileobj_list_lock = SPIN_LOCK_UNLOCKED;
+mcs_rwlock_lock_t fileobj_list_lock =
+	{{{0}, MCS_RWLOCK_TYPE_COMMON_READER, 0, 0, 0, NULL}, NULL};
 static LIST_HEAD(fileobj_list);
 
+#define FILEOBJ_PAGE_HASH_SHIFT 9
+#define FILEOBJ_PAGE_HASH_SIZE (1 << FILEOBJ_PAGE_HASH_SHIFT)
+#define FILEOBJ_PAGE_HASH_MASK (FILEOBJ_PAGE_HASH_SIZE - 1)
+
 struct fileobj {
-	struct memobj		memobj;		/* must be first */
-	long			sref;
-	long			cref;
-	uintptr_t		handle;
-	struct list_head	page_list;
-	struct list_head	list;
+	struct memobj memobj;		/* must be first */
+	long sref;
+	long cref;
+	uintptr_t handle;
+	struct list_head list;
+	struct list_head page_hash[FILEOBJ_PAGE_HASH_SIZE];
+	mcs_rwlock_lock_t page_hash_locks[FILEOBJ_PAGE_HASH_SIZE];
 };
 
 static memobj_release_func_t fileobj_release;
 static memobj_ref_func_t fileobj_ref;
 static memobj_get_page_func_t fileobj_get_page;
-static memobj_copy_page_func_t fileobj_copy_page;
 static memobj_flush_page_func_t fileobj_flush_page;
 static memobj_invalidate_page_func_t fileobj_invalidate_page;
 static memobj_lookup_page_func_t fileobj_lookup_page;
@@ -53,7 +58,7 @@ static struct memobj_ops fileobj_ops = {
 	.release =	&fileobj_release,
 	.ref =		&fileobj_ref,
 	.get_page =	&fileobj_get_page,
-	.copy_page =	&fileobj_copy_page,
+	.copy_page =	NULL,
 	.flush_page =	&fileobj_flush_page,
 	.invalidate_page =	&fileobj_invalidate_page,
 	.lookup_page =	&fileobj_lookup_page,
@@ -72,28 +77,36 @@ static struct memobj *to_memobj(struct fileobj *fileobj)
 /***********************************************************************
  * page_list
  */
-static void page_list_init(struct fileobj *obj)
+static void fileobj_page_hash_init(struct fileobj *obj)
 {
-	INIT_LIST_HEAD(&obj->page_list);
+	int i;
+	for (i = 0; i < FILEOBJ_PAGE_HASH_SIZE; ++i) {
+		mcs_rwlock_init(&obj->page_hash_locks[i]);
+		INIT_LIST_HEAD(&obj->page_hash[i]);
+	}
 	return;
 }
 
-static void page_list_insert(struct fileobj *obj, struct page *page)
+/* NOTE: caller must hold page_hash_locks[hash] */
+static void __fileobj_page_hash_insert(struct fileobj *obj,
+		struct page *page, int hash)
 {
-	list_add(&page->list, &obj->page_list);
-	return;
+	list_add(&page->list, &obj->page_hash[hash]);
 }
 
-static void page_list_remove(struct fileobj *obj, struct page *page)
+/* NOTE: caller must hold page_hash_locks[hash] */
+static void __fileobj_page_hash_remove(struct page *page)
 {
 	list_del(&page->list);
 }
 
-static struct page *page_list_lookup(struct fileobj *obj, off_t off)
+/* NOTE: caller must hold page_hash_locks[hash] */
+static struct page *__fileobj_page_hash_lookup(struct fileobj *obj,
+		int hash, off_t off)
 {
 	struct page *page;
 
-	list_for_each_entry(page, &obj->page_list, list) {
+	list_for_each_entry(page, &obj->page_hash[hash], list) {
 		if ((page->mode != PM_WILL_PAGEIO)
 				&& (page->mode != PM_PAGEIO)
 				&& (page->mode != PM_DONE_PAGEIO)
@@ -104,6 +117,7 @@ static struct page *page_list_lookup(struct fileobj *obj, off_t off)
 					obj, off, page->mode);
 			panic("page_list_lookup:invalid obj page");
 		}
+
 		if (page->offset == off) {
 			goto out;
 		}
@@ -114,13 +128,22 @@ out:
 	return page;
 }
 
-static struct page *page_list_first(struct fileobj *obj)
+static struct page *fileobj_page_hash_first(struct fileobj *obj)
 {
-	if (list_empty(&obj->page_list)) {
-		return NULL;
+	int i;
+
+	for (i = 0; i < FILEOBJ_PAGE_HASH_SIZE; ++i) {
+		if (!list_empty(&obj->page_hash[i])) {
+			break;
+		}
 	}
 
-	return list_first_entry(&obj->page_list, struct page, list);
+	if (i != FILEOBJ_PAGE_HASH_SIZE) {
+		return list_first_entry(&obj->page_hash[i], struct page, list);
+	}
+	else {
+		return NULL;
+	}
 }
 
 /***********************************************************************
@@ -163,10 +186,11 @@ static struct fileobj *obj_list_lookup(uintptr_t handle)
 int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
 {
 	ihk_mc_user_context_t ctx;
-	struct pager_create_result result;	// XXX: assumes contiguous physical
+	struct pager_create_result result __attribute__((aligned(64)));	
 	int error;
 	struct fileobj *newobj  = NULL;
 	struct fileobj *obj;
+	struct mcs_rwlock_node node;
 
 	dkprintf("fileobj_create(%d)\n", fd);
 	newobj = kmalloc(sizeof(*newobj), IHK_MC_AP_NOWAIT);
@@ -179,36 +203,53 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
 	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_CREATE;
 	ihk_mc_syscall_arg1(&ctx) = fd;
 	ihk_mc_syscall_arg2(&ctx) = virt_to_phys(&result);
+	memset(&result, 0, sizeof(result));
 
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
 	if (error) {
-		kprintf("fileobj_create(%d):create failed. %d\n", fd, error);
+		dkprintf("fileobj_create(%d):create failed. %d\n", fd, error);
 		goto out;
 	}
 
 	memset(newobj, 0, sizeof(*newobj));
 	newobj->memobj.ops = &fileobj_ops;
-	newobj->memobj.flags = MF_HAS_PAGER;
+	newobj->memobj.flags = MF_HAS_PAGER | MF_REG_FILE;
 	newobj->handle = result.handle;
 	newobj->sref = 1;
 	newobj->cref = 1;
-	page_list_init(newobj);
+	fileobj_page_hash_init(newobj);
 	ihk_mc_spinlock_init(&newobj->memobj.lock);
 
-	ihk_mc_spinlock_lock_noirq(&fileobj_list_lock);
+	mcs_rwlock_writer_lock_noirq(&fileobj_list_lock, &node);
 	obj = obj_list_lookup(result.handle);
 	if (!obj) {
 		obj_list_insert(newobj);
 		obj = newobj;
+		to_memobj(obj)->size = result.size;
+		to_memobj(obj)->flags |= result.flags;
+		to_memobj(obj)->status = MEMOBJ_READY;
+		if (to_memobj(obj)->flags & MF_PREFETCH) {
+			to_memobj(obj)->status = MEMOBJ_TO_BE_PREFETCHED;
+		}
 		newobj = NULL;
+		dkprintf("%s: new obj 0x%lx cref: %d, %s\n",
+			__FUNCTION__,
+			obj,
+			obj->cref,
+			to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
 	}
 	else {
 		++obj->sref;
 		++obj->cref;
 		memobj_unlock(&obj->memobj);	/* locked by obj_list_lookup() */
+		dkprintf("%s: existing obj 0x%lx cref: %d, %s\n",
+			__FUNCTION__,
+			obj,
+			obj->cref,
+			to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
 	}
 
-	ihk_mc_spinlock_unlock_noirq(&fileobj_list_lock);
+	mcs_rwlock_writer_unlock_noirq(&fileobj_list_lock, &node);
 
 	error = 0;
 	*objp = to_memobj(obj);
@@ -239,6 +280,7 @@ static void fileobj_release(struct memobj *memobj)
 	long free_sref = 0;
 	uintptr_t free_handle;
 	struct fileobj *free_obj = NULL;
+	struct mcs_rwlock_node node;
 
 	dkprintf("fileobj_release(%p %lx)\n", obj, obj->handle);
 
@@ -252,19 +294,39 @@ static void fileobj_release(struct memobj *memobj)
 	obj->sref -= free_sref;
 	free_handle = obj->handle;
 	memobj_unlock(&obj->memobj);
+	if (obj->memobj.flags & MF_HOST_RELEASED) {
+		free_sref = 0; // don't call syscall_generic_forwarding
+	}
 
 	if (free_obj) {
-		ihk_mc_spinlock_lock_noirq(&fileobj_list_lock);
+		dkprintf("%s: release obj 0x%lx cref: %d, free_obj: 0x%lx, %s\n",
+				__FUNCTION__,
+				obj,
+				obj->cref,
+				free_obj,
+				to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
+		mcs_rwlock_writer_lock_noirq(&fileobj_list_lock, &node);
 		/* zap page_list */
 		for (;;) {
 			struct page *page;
-			int count;
+			void *page_va;
 
-			page = page_list_first(obj);
+			page = fileobj_page_hash_first(obj);
 			if (!page) {
 				break;
 			}
-			page_list_remove(obj, page);
+			__fileobj_page_hash_remove(page);
+			page_va = phys_to_virt(page_to_phys(page));
+
+			if (ihk_atomic_read(&page->count) != 1) {
+				kprintf("%s: WARNING: page count for phys 0x%lx is invalid\n",
+					__FUNCTION__, page->phys);
+			}
+
+			if (page_unmap(page)) {
+				ihk_mc_free_pages(page_va, 1);
+			}
+#if 0
 			count = ihk_atomic_sub_return(1, &page->count);
 
 			if (!((page->mode == PM_WILL_PAGEIO)
@@ -281,10 +343,10 @@ static void fileobj_release(struct memobj *memobj)
 			}
 
 			page->mode = PM_NONE;
-			free_pages(phys_to_virt(page_to_phys(page)), 1);
+#endif
 		}
 		obj_list_remove(free_obj);
-		ihk_mc_spinlock_unlock_noirq(&fileobj_list_lock);
+		mcs_rwlock_writer_unlock_noirq(&fileobj_list_lock, &node);
 		kfree(free_obj);
 	}
 
@@ -330,83 +392,101 @@ static void fileobj_do_pageio(void *args0)
 	struct page *page;
 	ihk_mc_user_context_t ctx;
 	ssize_t ss;
+	struct mcs_rwlock_node mcs_node;
+	int hash = (off >> PAGE_SHIFT) & FILEOBJ_PAGE_HASH_MASK;	
 
-	memobj_lock(&obj->memobj);
-	page = page_list_lookup(obj, off);
+	mcs_rwlock_writer_lock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
+	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page) {
 		goto out;
 	}
 
 	while (page->mode == PM_PAGEIO) {
-		memobj_unlock(&obj->memobj);
+		mcs_rwlock_writer_unlock_noirq(&obj->page_hash_locks[hash],
+				&mcs_node);
 		cpu_pause();
-		memobj_lock(&obj->memobj);
+		mcs_rwlock_writer_lock_noirq(&obj->page_hash_locks[hash],
+				&mcs_node);
 	}
 
 	if (page->mode == PM_WILL_PAGEIO) {
-		page->mode = PM_PAGEIO;
-		memobj_unlock(&obj->memobj);
-
-		ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_READ;
-		ihk_mc_syscall_arg1(&ctx) = obj->handle;
-		ihk_mc_syscall_arg2(&ctx) = off;
-		ihk_mc_syscall_arg3(&ctx) = pgsize;
-		ihk_mc_syscall_arg4(&ctx) = page_to_phys(page);
-
-		ss = syscall_generic_forwarding(__NR_mmap, &ctx);
-
-		memobj_lock(&obj->memobj);
-		if (page->mode != PM_PAGEIO) {
-			kprintf("fileobj_do_pageio(%p,%lx,%lx):"
-					"invalid mode %x\n",
-					obj, off, pgsize, page->mode);
-			panic("fileobj_do_pageio:invalid page mode");
+		if (to_memobj(obj)->flags & MF_ZEROFILL) {
+			void *virt = phys_to_virt(page_to_phys(page));
+			memset(virt, 0, PAGE_SIZE);
 		}
+		else {
+			page->mode = PM_PAGEIO;
+			mcs_rwlock_writer_unlock_noirq(&obj->page_hash_locks[hash],
+					&mcs_node);
 
-		if (ss == 0) {
-			dkprintf("fileobj_do_pageio(%p,%lx,%lx):EOF? %ld\n",
-					obj, off, pgsize, ss);
-			page->mode = PM_PAGEIO_EOF;
-			goto out;
-		}
-		else if (ss != pgsize) {
-			kprintf("fileobj_do_pageio(%p,%lx,%lx):"
-					"read failed. %ld\n",
-					obj, off, pgsize, ss);
-			page->mode = PM_PAGEIO_ERROR;
-			goto out;
+			ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_READ;
+			ihk_mc_syscall_arg1(&ctx) = obj->handle;
+			ihk_mc_syscall_arg2(&ctx) = off;
+			ihk_mc_syscall_arg3(&ctx) = pgsize;
+			ihk_mc_syscall_arg4(&ctx) = page_to_phys(page);
+
+			dkprintf("%s: __NR_mmap for handle 0x%lx\n",
+					__FUNCTION__, obj->handle);
+			ss = syscall_generic_forwarding(__NR_mmap, &ctx);
+
+			mcs_rwlock_writer_lock_noirq(&obj->page_hash_locks[hash],
+					&mcs_node);
+			if (page->mode != PM_PAGEIO) {
+				kprintf("fileobj_do_pageio(%p,%lx,%lx):"
+						"invalid mode %x\n",
+						obj, off, pgsize, page->mode);
+				panic("fileobj_do_pageio:invalid page mode");
+			}
+
+			if (ss == 0) {
+				dkprintf("fileobj_do_pageio(%p,%lx,%lx):EOF? %ld\n",
+						obj, off, pgsize, ss);
+				page->mode = PM_PAGEIO_EOF;
+				goto out;
+			}
+			else if (ss != pgsize) {
+				kprintf("fileobj_do_pageio(%p,%lx,%lx):"
+						"read failed. %ld\n",
+						obj, off, pgsize, ss);
+				page->mode = PM_PAGEIO_ERROR;
+				goto out;
+			}
 		}
 
 		page->mode = PM_DONE_PAGEIO;
 	}
 out:
-	memobj_unlock(&obj->memobj);
+	mcs_rwlock_writer_unlock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
 	fileobj_release(&obj->memobj);		/* got fileobj_get_page() */
 	kfree(args0);
 	dkprintf("fileobj_do_pageio(%p,%lx,%lx):\n", obj, off, pgsize);
 	return;
 }
 
-static int fileobj_get_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *pflag)
+static int fileobj_get_page(struct memobj *memobj, off_t off,
+		int p2align, uintptr_t *physp, unsigned long *pflag)
 {
 	struct thread *proc = cpu_local_var(current);
 	struct fileobj *obj = to_fileobj(memobj);
-	int error;
+	int error = -1;
 	void *virt = NULL;
 	int npages;
 	uintptr_t phys = -1;
 	struct page *page;
 	struct pageio_args *args = NULL;
+	struct mcs_rwlock_node mcs_node;
+	int hash = (off >> PAGE_SHIFT) & FILEOBJ_PAGE_HASH_MASK;	
 
 	dkprintf("fileobj_get_page(%p,%lx,%x,%p)\n", obj, off, p2align, physp);
-
-	memobj_lock(&obj->memobj);
 	if (p2align != PAGE_P2ALIGN) {
-		error = -ENOMEM;
-		goto out;
+		return -ENOMEM;
 	}
 
-	page = page_list_lookup(obj, off);
+	mcs_rwlock_writer_lock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
+	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page || (page->mode == PM_WILL_PAGEIO)
 			|| (page->mode == PM_PAGEIO)) {
 		args = kmalloc(sizeof(*args), IHK_MC_AP_NOWAIT);
@@ -420,7 +500,10 @@ static int fileobj_get_page(struct memobj *memobj, off_t off, int p2align, uintp
 
 		if (!page) {
 			npages = 1 << p2align;
-			virt = ihk_mc_alloc_pages(npages, IHK_MC_AP_NOWAIT);
+
+			virt = ihk_mc_alloc_pages(npages, IHK_MC_AP_NOWAIT |
+					(to_memobj(obj)->flags & MF_ZEROFILL) ? IHK_MC_AP_USER : 0);
+
 			if (!virt) {
 				error = -ENOMEM;
 				kprintf("fileobj_get_page(%p,%lx,%x,%p):"
@@ -430,17 +513,19 @@ static int fileobj_get_page(struct memobj *memobj, off_t off, int p2align, uintp
 				goto out;
 			}
 			phys = virt_to_phys(virt);
-			page = phys_to_page(phys);
+			page = phys_to_page_insert_hash(phys);
 			if (page->mode != PM_NONE) {
 				panic("fileobj_get_page:invalid new page");
 			}
-			page->mode = PM_WILL_PAGEIO;
 			page->offset = off;
 			ihk_atomic_set(&page->count, 1);
-			page_list_insert(obj, page);
+			__fileobj_page_hash_insert(obj, page, hash);
+			page->mode = PM_WILL_PAGEIO;
 		}
 
+		memobj_lock(&obj->memobj);
 		++obj->cref;	/* for fileobj_do_pageio() */
+		memobj_unlock(&obj->memobj);
 
 		args->fileobj = obj;
 		args->objoff = off;
@@ -472,7 +557,8 @@ static int fileobj_get_page(struct memobj *memobj, off_t off, int p2align, uintp
 	*physp = page_to_phys(page);
 	virt = NULL;
 out:
-	memobj_unlock(&obj->memobj);
+	mcs_rwlock_writer_unlock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
 	if (virt) {
 		ihk_mc_free_pages(virt, npages);
 	}
@@ -484,76 +570,6 @@ out:
 	return error;
 }
 
-static uintptr_t fileobj_copy_page(
-		struct memobj *memobj, uintptr_t orgpa, int p2align)
-{
-	struct page *orgpage = phys_to_page(orgpa);
-	size_t pgsize = PAGE_SIZE << p2align;
-	int npages = 1 << p2align;
-	void *newkva = NULL;
-	uintptr_t newpa = -1;
-	void *orgkva;
-	int count;
-
-	dkprintf("fileobj_copy_page(%p,%lx,%d)\n", memobj, orgpa, p2align);
-	if (p2align != PAGE_P2ALIGN) {
-		panic("p2align");
-	}
-
-	memobj_lock(memobj);
-	for (;;) {
-		if (orgpage->mode != PM_MAPPED) {
-			kprintf("fileobj_copy_page(%p,%lx,%d):"
-					"invalid cow page. %x\n",
-					memobj, orgpa, p2align, orgpage->mode);
-			panic("fileobj_copy_page:invalid cow page");
-		}
-		count = ihk_atomic_read(&orgpage->count);
-		if (count == 2) {	// XXX: private only
-			list_del(&orgpage->list);
-			ihk_atomic_dec(&orgpage->count);
-			orgpage->mode = PM_NONE;
-			newpa = orgpa;
-			break;
-		}
-		if (count <= 0) {
-			kprintf("fileobj_copy_page(%p,%lx,%d):"
-					"orgpage count corrupted. %x\n",
-					memobj, orgpa, p2align, count);
-			panic("fileobj_copy_page:orgpage count corrupted");
-		}
-		if (newkva) {
-			orgkva = phys_to_virt(orgpa);
-			memcpy(newkva, orgkva, pgsize);
-			ihk_atomic_dec(&orgpage->count);
-			newpa = virt_to_phys(newkva);
-			page_map(phys_to_page(newpa));
-			newkva = NULL;	/* avoid ihk_mc_free_pages() */
-			break;
-		}
-
-		memobj_unlock(memobj);
-		newkva = ihk_mc_alloc_aligned_pages(npages, p2align,
-				IHK_MC_AP_NOWAIT);
-		if (!newkva) {
-			kprintf("fileobj_copy_page(%p,%lx,%d):"
-					"alloc page failed\n",
-					memobj, orgpa, p2align);
-			goto out;
-		}
-		memobj_lock(memobj);
-	}
-	memobj_unlock(memobj);
-
-out:
-	if (newkva) {
-		ihk_mc_free_pages(newkva, npages);
-	}
-	dkprintf("fileobj_copy_page(%p,%lx,%d): %lx\n",
-			memobj, orgpa, p2align, newpa);
-	return newpa;
-}
-
 static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 		size_t pgsize)
 {
@@ -562,7 +578,16 @@ static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 	ihk_mc_user_context_t ctx;
 	ssize_t ss;
 
+	if (to_memobj(obj)->flags & MF_ZEROFILL) {
+		return 0;
+	}
+
 	page = phys_to_page(phys);
+	if (!page) {
+		kprintf("%s: warning: tried to flush non-existing page for phys addr: 0x%lx\n", 
+			__FUNCTION__, phys);
+		return 0;
+	}
 	memobj_unlock(&obj->memobj);
 
 	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_WRITE;
@@ -585,63 +610,48 @@ static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 static int fileobj_invalidate_page(struct memobj *memobj, uintptr_t phys,
 		size_t pgsize)
 {
-	struct fileobj *obj = to_fileobj(memobj);
-	int error;
-	struct page *page;
-
 	dkprintf("fileobj_invalidate_page(%p,%#lx,%#lx)\n",
 			memobj, phys, pgsize);
 
-	if (!(page = phys_to_page(phys))
-			|| !(page = page_list_lookup(obj, page->offset))) {
-		error = 0;
-		goto out;
-	}
-
-	if (ihk_atomic_read(&page->count) == 1) {
-		if (page_unmap(page)) {
-			ihk_mc_free_pages(phys_to_virt(phys),
-					pgsize/PAGE_SIZE);
-		}
-	}
-
-	error = 0;
-out:
-	dkprintf("fileobj_invalidate_page(%p,%#lx,%#lx):%d\n",
-			memobj, phys, pgsize, error);
-	return error;
+	/* TODO: keep track of reverse mappings so that invalidation
+	 * can be performed */
+	kprintf("%s: WARNING: file mapping invalidation not supported\n",
+		__FUNCTION__);
+	return 0;
 }
 
-static int fileobj_lookup_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *pflag)
+static int fileobj_lookup_page(struct memobj *memobj, off_t off,
+		int p2align, uintptr_t *physp, unsigned long *pflag)
 {
 	struct fileobj *obj = to_fileobj(memobj);
-	int error;
-	uintptr_t phys = -1;
+	int error = -1;
 	struct page *page;
+	struct mcs_rwlock_node mcs_node;
+	int hash = (off >> PAGE_SHIFT) & FILEOBJ_PAGE_HASH_MASK;
 
 	dkprintf("fileobj_lookup_page(%p,%lx,%x,%p)\n", obj, off, p2align, physp);
 
-	memobj_lock(&obj->memobj);
 	if (p2align != PAGE_P2ALIGN) {
-		error = -ENOMEM;
-		goto out;
+		return -ENOMEM;
 	}
 
-	page = page_list_lookup(obj, off);
+	mcs_rwlock_reader_lock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
+
+	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page) {
-		error = -ENOENT;
-		dkprintf("fileobj_lookup_page(%p,%lx,%x,%p): page not found. %d\n", obj, off, p2align, physp, error);
 		goto out;
 	}
-	phys = page_to_phys(page);
 
+	*physp = page_to_phys(page);
 	error = 0;
-	if (physp) {
-		*physp = phys;
-	}
+
 out:
-	memobj_unlock(&obj->memobj);
-	dkprintf("fileobj_lookup_page(%p,%lx,%x,%p): %d %lx\n",
-			obj, off, p2align, physp, error, phys);
+	mcs_rwlock_reader_unlock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
+
+	dkprintf("fileobj_lookup_page(%p,%lx,%x,%p): %d \n",
+			obj, off, p2align, physp, error);
 	return error;
 }
+

kernel/freeze.c

@@ -0,0 +1,54 @@
+#include <kmsg.h>
+#include <string.h>
+#include <ihk/cpu.h>
+#include <ihk/debug.h>
+#include <cls.h>
+
+extern int nmi_mode;
+extern void mod_nmi_ctx(void *, void(*)());
+extern void lapic_ack();
+extern void __freeze();
+
+void
+freeze()
+{
+	struct ihk_os_monitor *monitor = cpu_local_var(monitor);
+
+	monitor->status_bak = monitor->status;
+	monitor->status = IHK_OS_MONITOR_KERNEL_FROZEN;
+	while (monitor->status == IHK_OS_MONITOR_KERNEL_FROZEN)
+		cpu_halt();
+	monitor->status = monitor->status_bak;
+}
+
+long
+freeze_thaw(void *nmi_ctx)
+{
+	struct ihk_os_monitor *monitor = cpu_local_var(monitor);
+
+	if (nmi_mode == 1) {
+		if (monitor->status != IHK_OS_MONITOR_KERNEL_FROZEN) {
+#if 1
+			mod_nmi_ctx(nmi_ctx, __freeze);
+			return 1;
+#else
+			unsigned long flags;
+
+			flags = cpu_disable_interrupt_save();
+			monitor->status_bak = monitor->status;
+			monitor->status = IHK_OS_MONITOR_KERNEL_FROZEN;
+			lapic_ack();
+			while (monitor->status == IHK_OS_MONITOR_KERNEL_FROZEN)
+				cpu_halt();
+			monitor->status = monitor->status_bak;
+			cpu_restore_interrupt(flags);
+#endif
+		}
+	}
+	else if(nmi_mode == 2) {
+		if (monitor->status == IHK_OS_MONITOR_KERNEL_FROZEN) {
+			monitor->status = IHK_OS_MONITOR_KERNEL_THAW;
+		}
+	}
+	return 0;
+}

kernel/futex.c

@@ -79,8 +79,6 @@
 #define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
 #endif
 
-extern struct sigpending *hassigpending(struct thread *thread);
-
 int futex_cmpxchg_enabled;
 
 /**
@@ -250,9 +248,13 @@ static int cmpxchg_futex_value_locked(uint32_t __user *uaddr, uint32_t uval, uin
 
 static int get_futex_value_locked(uint32_t *dest, uint32_t *from)
 {
-	/* RIKEN: futexes are always on not swappable pages */
-	*dest = getint_user((int *)from);
-
+	/*
+	 * Officially we should call:
+	 * return getint_user((int *)dest, (int *)from);
+	 *
+	 * but McKernel on x86 can just access user-space.
+	 */
+	*dest = *(volatile uint32_t *)from;
 	return 0;
 }
 
@@ -672,25 +674,32 @@ static uint64_t futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q
 				uint64_t timeout)
 {
 	uint64_t time_remain = 0;
+	unsigned long irqstate;
+	struct thread *thread = cpu_local_var(current);
 	/*
 	 * The task state is guaranteed to be set before another task can
-	 * wake it. set_current_state() is implemented using set_mb() and
-	 * queue_me() calls spin_unlock() upon completion, both serializing
-	 * access to the hash list and forcing another memory barrier.
+	 * wake it. 
+	 * queue_me() calls spin_unlock() upon completion, serializing
+	 * access to the hash list and forcing a memory barrier.
 	 */
 	xchg4(&(cpu_local_var(current)->status), PS_INTERRUPTIBLE);
+
+	/* Indicate spin sleep */
+	irqstate = ihk_mc_spinlock_lock(&thread->spin_sleep_lock);
+	thread->spin_sleep = 1;
+	ihk_mc_spinlock_unlock(&thread->spin_sleep_lock, irqstate);
+
 	queue_me(q, hb);
 	
 	if (!plist_node_empty(&q->list)) {
 		
-		/* RIKEN: use mcos timers */
 		if (timeout) {
 			dkprintf("futex_wait_queue_me(): tid: %d schedule_timeout()\n", cpu_local_var(current)->tid);
 			time_remain = schedule_timeout(timeout);
 		}
 		else {
 			dkprintf("futex_wait_queue_me(): tid: %d schedule()\n", cpu_local_var(current)->tid);
-			schedule();
+			spin_sleep_or_schedule();
 			time_remain = 0;
 		}
 		
@@ -699,6 +708,7 @@ static uint64_t futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q
 	
 	/* This does not need to be serialized */
 	cpu_local_var(current)->status = PS_RUNNING;
+	thread->spin_sleep = 0;
 	
 	return time_remain;
 }
@@ -745,14 +755,17 @@ static int futex_wait_setup(uint32_t __user *uaddr, uint32_t val, int fshared,
 	 */
 	q->key = FUTEX_KEY_INIT;
 	ret = get_futex_key(uaddr, fshared, &q->key);
-	if ((ret != 0))
+	if (ret != 0)
 		return ret;
 
 	*hb = queue_lock(q);
 
 	ret = get_futex_value_locked(&uval, uaddr);
-
-	/* RIKEN: get_futex_value_locked() always returns 0 on mckernel */
+	if (ret) {
+		queue_unlock(q, *hb);
+		put_futex_key(fshared, &q->key);
+		return ret;
+	}
 
 	if (uval != val) {
 		queue_unlock(q, *hb);
@@ -778,8 +791,6 @@ static int futex_wait(uint32_t __user *uaddr, int fshared,
 	q.bitset = bitset;
 	q.requeue_pi_key = NULL;
 
-	/* RIKEN: futex_wait_queue_me() calls schedule_timeout() if timer is set */
-
 retry:
 	/* Prepare to wait on uaddr. */
 	ret = futex_wait_setup(uaddr, val, fshared, &q, &hb);

kernel/host.c

@@ -23,14 +23,15 @@
 #include <ihk/debug.h>
 #include <ihk/ikc.h>
 #include <ikc/master.h>
-#include <syscall.h>
 #include <cls.h>
+#include <syscall.h>
 #include <process.h>
 #include <page.h>
 #include <mman.h>
 #include <init.h>
 #include <kmalloc.h>
 #include <sysfs.h>
+#include <ihk/perfctr.h>
 
 //#define DEBUG_PRINT_HOST
 
@@ -125,7 +126,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 		
 		up = virt_to_phys(up_v);
 		if (add_process_memory_range(vm, s, e, up, flags, NULL, 0,
-					PAGE_SHIFT) != 0) {
+					PAGE_SHIFT, NULL) != 0) {
 			ihk_mc_free_pages(up_v, range_npages);
 			kprintf("ERROR: adding memory range for ELF section %i\n", i);
 			goto err;
@@ -213,7 +214,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 	args_envs_p = virt_to_phys(args_envs);
 
 	if(add_process_memory_range(vm, addr, e, args_envs_p,
-				flags, NULL, 0, PAGE_SHIFT) != 0){
+				flags, NULL, 0, PAGE_SHIFT, NULL) != 0){
 		ihk_mc_free_pages(args_envs, ARGENV_PAGE_COUNT);
 		kprintf("ERROR: adding memory range for args/envs\n");
 		goto err;
@@ -332,6 +333,9 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 			goto err;
 		}
 	}
+	else {
+		vm->vdso_addr = NULL;
+	}
 
 	p->rprocess = (unsigned long)thread;
 	p->rpgtable = virt_to_phys(as->page_table);
@@ -373,10 +377,16 @@ static int process_msg_prepare_process(unsigned long rphys)
 	}
 
 	n = p->num_sections;
+	if (n > 16) {
+		kprintf("%s: ERROR: more ELF sections than 16??\n",
+			__FUNCTION__);
+		return -ENOMEM;
+	}
 	dkprintf("# of sections: %d\n", n);
 
-	if((pn = ihk_mc_allocate(sizeof(struct program_load_desc) 
-	       + sizeof(struct program_image_section) * n, IHK_MC_AP_NOWAIT)) == NULL){
+	if((pn = kmalloc(sizeof(struct program_load_desc) 
+					+ sizeof(struct program_image_section) * n,
+					IHK_MC_AP_NOWAIT)) == NULL){
 		ihk_mc_unmap_virtual(p, npages, 0);
 		ihk_mc_unmap_memory(NULL, phys, sz);
 		return -ENOMEM;
@@ -384,8 +394,10 @@ static int process_msg_prepare_process(unsigned long rphys)
 	memcpy_long(pn, p, sizeof(struct program_load_desc) 
 	            + sizeof(struct program_image_section) * n);
 
-	if((thread = create_thread(p->entry)) == NULL){
-		ihk_mc_free(pn);
+	if ((thread = create_thread(p->entry,
+					(unsigned long *)&p->cpu_set,
+					sizeof(p->cpu_set))) == NULL) {
+		kfree(pn);
 		ihk_mc_unmap_virtual(p, npages, 1);
 		ihk_mc_unmap_memory(NULL, phys, sz);
 		return -ENOMEM;
@@ -423,9 +435,6 @@ static int process_msg_prepare_process(unsigned long rphys)
 	vm->region.map_end = vm->region.map_start;
 	memcpy(proc->rlimit, pn->rlimit, sizeof(struct rlimit) * MCK_RLIM_MAX);
 
-	/* TODO: Clear it at the proper timing */
-	cpu_local_var(scp).post_idx = 0;
-
 	if (prepare_process_ranges_args_envs(thread, pn, p, attr, 
 				NULL, 0, NULL, 0) != 0) {
 		kprintf("error: preparing process ranges, args, envs, stack\n");
@@ -435,7 +444,7 @@ static int process_msg_prepare_process(unsigned long rphys)
 	dkprintf("new process : %p [%d] / table : %p\n", proc, proc->pid,
 	        vm->address_space->page_table);
 
-	ihk_mc_free(pn);
+	kfree(pn);
 
 	ihk_mc_unmap_virtual(p, npages, 1);
 	ihk_mc_unmap_memory(NULL, phys, sz);
@@ -443,77 +452,13 @@ static int process_msg_prepare_process(unsigned long rphys)
 
 	return 0;
 err:
-	ihk_mc_free(pn);
+	kfree(pn);
 	ihk_mc_unmap_virtual(p, npages, 1);
 	ihk_mc_unmap_memory(NULL, phys, sz);
 	destroy_thread(thread);
 	return -ENOMEM;
 }
 
-static void process_msg_init(struct ikc_scd_init_param *pcp, struct syscall_params *lparam)
-{
-	lparam->response_va = allocate_pages(RESPONSE_PAGE_COUNT, 0);
-	lparam->response_pa = virt_to_phys(lparam->response_va);
-
-	pcp->request_page = 0;
-	pcp->doorbell_page = 0;
-	pcp->response_page = lparam->response_pa;
-}
-
-static void process_msg_init_acked(struct ihk_ikc_channel_desc *c, unsigned long pphys)
-{
-	struct ikc_scd_init_param *param = phys_to_virt(pphys);
-	struct syscall_params *lparam;
-	enum ihk_mc_pt_attribute attr;
-
-	attr = PTATTR_NO_EXECUTE | PTATTR_WRITABLE | PTATTR_FOR_USER;
-
-	lparam = &cpu_local_var(scp);
-	if(cpu_local_var(syscall_channel2) == c)
-		lparam = &cpu_local_var(scp2);
-	lparam->request_rpa = param->request_page;
-	lparam->request_pa = ihk_mc_map_memory(NULL, param->request_page,
-	                                       REQUEST_PAGE_COUNT * PAGE_SIZE);
-	if((lparam->request_va = ihk_mc_map_virtual(lparam->request_pa,
-	                                        REQUEST_PAGE_COUNT,
-	                                        attr)) == NULL){
-		// TODO: 
-		panic("ENOMEM");
-	}
-
-	lparam->doorbell_rpa = param->doorbell_page;
-	lparam->doorbell_pa = ihk_mc_map_memory(NULL, param->doorbell_page,
-	                                        DOORBELL_PAGE_COUNT * 
-	                                        PAGE_SIZE);
-	if((lparam->doorbell_va = ihk_mc_map_virtual(lparam->doorbell_pa,
-	                                         DOORBELL_PAGE_COUNT,
-	                                         attr)) == NULL){
-		// TODO: 
-		panic("ENOMEM");
-	}
-
-	lparam->post_rpa = param->post_page;
-	lparam->post_pa = ihk_mc_map_memory(NULL, param->post_page,
-	                                    PAGE_SIZE);
-	if((lparam->post_va = ihk_mc_map_virtual(lparam->post_pa, 1,
-	                                     attr)) == NULL){
-		// TODO: 
-		panic("ENOMEM");
-	}
-
-	lparam->post_fin = 1;
-
-	dkprintf("Syscall parameters: (%d)\n", ihk_mc_get_processor_id());
-	dkprintf(" Response: %lx, %p\n",
-	        lparam->response_pa, lparam->response_va);
-	dkprintf(" Request : %lx, %lx, %p\n",
-	        lparam->request_pa, lparam->request_rpa, lparam->request_va);
-	dkprintf(" Doorbell: %lx, %lx, %p\n",
-	        lparam->doorbell_pa, lparam->doorbell_rpa, lparam->doorbell_va);
-	dkprintf(" Post: %lx, %lx, %p\n",
-	        lparam->post_pa, lparam->post_rpa, lparam->post_va);
-}
-
 static void syscall_channel_send(struct ihk_ikc_channel_desc *c,
                                  struct ikc_scd_packet *packet)
 {
@@ -521,46 +466,17 @@ static void syscall_channel_send(struct ihk_ikc_channel_desc *c,
 }
 
 extern unsigned long do_kill(struct thread *, int, int, int, struct siginfo *, int ptracecont);
-extern void settid(struct thread *proc, int mode, int newcpuid, int oldcpuid);
-
-extern void process_procfs_request(unsigned long rarg);
-extern int memcheckall();
-extern int freecheck(int runcount);
-extern int runcount;
+extern void process_procfs_request(struct ikc_scd_packet *rpacket);
 extern void terminate_host(int pid);
 extern void debug_log(long);
 
-static void req_get_cpu_mapping(long req_rpa)
-{
-	size_t mapsize;
-	size_t size;
-	int npages;
-	long phys;
-	struct get_cpu_mapping_req *req;
-	struct cpu_mapping *buf;
-
-	size = sizeof(*req);
-	mapsize = size + (req_rpa & (PAGE_SIZE - 1));
-	npages = (mapsize + PAGE_SIZE - 1) >> PAGE_SHIFT;
-	phys = ihk_mc_map_memory(NULL, req_rpa, size);
-	req = ihk_mc_map_virtual(phys, npages, PTATTR_WRITABLE);
-
-	req->error = arch_get_cpu_mapping(&buf, &req->buf_elems);
-	if (!req->error) {
-		req->buf_rpa = virt_to_phys(buf);
-	}
-
-	ihk_mc_unmap_virtual(req, npages, 0);
-	ihk_mc_unmap_memory(NULL, phys, size);
-	return;
-} /* req_get_cpu_mapping() */
-
 static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
                                   void *__packet, void *ihk_os)
 {
 	struct ikc_scd_packet *packet = __packet;
 	struct ikc_scd_packet pckt;
 	int rc;
+	struct mcs_rwlock_node_irqsave lock;
 	struct thread *thread;
 	struct process *proc;
 	struct mcctrl_signal {
@@ -572,22 +488,18 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	} *sp, info;
 	unsigned long pp;
 	int cpuid;
+	int ret = 0;
+	struct perf_ctrl_desc *pcd;
+	unsigned int mode = 0;
 
 	switch (packet->msg) {
 	case SCD_MSG_INIT_CHANNEL_ACKED:
 		dkprintf("SCD_MSG_INIT_CHANNEL_ACKED\n");
-		process_msg_init_acked(c, packet->arg);
-		return 0;
+		ret = 0;
+		break;
 
 	case SCD_MSG_PREPARE_PROCESS:
 
-		if (find_command_line("memdebug")) {
-			memcheckall();
-			if (runcount)
-				freecheck(runcount);
-			runcount++;
-		}
-
 		if((rc = process_msg_prepare_process(packet->arg)) == 0){
 			pckt.msg = SCD_MSG_PREPARE_PROCESS_ACKED;
 			pckt.err = 0;
@@ -600,27 +512,51 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		pckt.arg = packet->arg;
 		syscall_channel_send(c, &pckt);
 
-		return 0;
+		ret = 0;
+		break;
 
 	case SCD_MSG_SCHEDULE_PROCESS:
-		cpuid = obtain_clone_cpuid();
-		if(cpuid == -1){
-			kprintf("No CPU available\n");
-			return -1;
-		}
-		dkprintf("SCD_MSG_SCHEDULE_PROCESS: %lx\n", packet->arg);
 		thread = (struct thread *)packet->arg;
-		proc = thread->proc;
 
-		settid(thread, 0, cpuid, -1);
+		cpuid = obtain_clone_cpuid(&thread->cpu_set);
+		if (cpuid == -1) {
+			kprintf("No CPU available\n");
+			ret = -1;
+			break;
+		}
+
+		dkprintf("SCD_MSG_SCHEDULE_PROCESS: %lx\n", packet->arg);
+		proc = thread->proc;
+		thread->tid = proc->pid;
 		proc->status = PS_RUNNING;
 		thread->status = PS_RUNNING;
 		chain_thread(thread);
 		chain_process(proc);
 		runq_add_thread(thread, cpuid);
-					  
-		//cpu_local_var(next) = (struct thread *)packet->arg;
-		return 0;
+
+		ret = 0;
+		break;
+
+	/*
+	 * Used for syscall offload reply message to explicitly schedule in
+	 * the waiting thread
+	 */
+	case SCD_MSG_WAKE_UP_SYSCALL_THREAD:
+		thread = find_thread(0, packet->ttid, &lock);
+		if (!thread) {
+			kprintf("%s: WARNING: no thread for SCD reply? TID: %d\n",
+				__FUNCTION__, packet->ttid);
+			ret = -EINVAL;
+			break;
+		}
+		thread_unlock(thread, &lock);
+
+		dkprintf("%s: SCD_MSG_WAKE_UP_SYSCALL_THREAD: waking up tid %d\n",
+			__FUNCTION__, packet->ttid);
+		waitq_wakeup(&thread->scd_wq);
+		ret = 0;
+		break;
+
 	case SCD_MSG_SEND_SIGNAL:
 		pp = ihk_mc_map_memory(NULL, packet->arg, sizeof(struct mcctrl_signal));
 		sp = (struct mcctrl_signal *)ihk_mc_map_virtual(pp, 1, PTATTR_WRITABLE | PTATTR_ACTIVE);
@@ -635,18 +571,25 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 
 		rc = do_kill(NULL, info.pid, info.tid, info.sig, &info.info, 0);
 		kprintf("SCD_MSG_SEND_SIGNAL: do_kill(pid=%d, tid=%d, sig=%d)=%d\n", info.pid, info.tid, info.sig, rc);
-		return 0;
+		ret = 0;
+		break;
+
 	case SCD_MSG_PROCFS_REQUEST:
-		process_procfs_request(packet->arg);
-		return 0;
+		process_procfs_request(packet);
+		ret = 0;
+		break;
+
 	case SCD_MSG_CLEANUP_PROCESS:
 		dkprintf("SCD_MSG_CLEANUP_PROCESS pid=%d\n", packet->pid);
 		terminate_host(packet->pid);
-		return 0;
+		ret = 0;
+		break;
+
 	case SCD_MSG_DEBUG_LOG:
 		dkprintf("SCD_MSG_DEBUG_LOG code=%lx\n", packet->arg);
 		debug_log(packet->arg);
-		return 0;
+		ret = 0;
+		break;
 
 	case SCD_MSG_SYSFS_REQ_SHOW:
 	case SCD_MSG_SYSFS_REQ_STORE:
@@ -654,25 +597,76 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		sysfss_packet_handler(c, packet->msg, packet->err,
 				packet->sysfs_arg1, packet->sysfs_arg2,
 				packet->sysfs_arg3);
-		return 0;
+		ret = 0;
+		break;
 
-	case SCD_MSG_GET_CPU_MAPPING:
-		req_get_cpu_mapping(packet->arg);
+	case SCD_MSG_PERF_CTRL:
+		pp = ihk_mc_map_memory(NULL, packet->arg, sizeof(struct perf_ctrl_desc));
+		pcd = (struct perf_ctrl_desc *)ihk_mc_map_virtual(pp, 1, PTATTR_WRITABLE | PTATTR_ACTIVE);
 
-		pckt.msg = SCD_MSG_REPLY_GET_CPU_MAPPING;
+		switch (pcd->ctrl_type) {
+		case PERF_CTRL_SET:
+			if (!pcd->exclude_kernel) {
+				mode |= PERFCTR_KERNEL_MODE;
+			}
+			if (!pcd->exclude_user) {
+				mode |= PERFCTR_USER_MODE;
+			}
+			ihk_mc_perfctr_init_raw(pcd->target_cntr, pcd->config, mode);
+			ihk_mc_perfctr_stop(1 << pcd->target_cntr);
+			ihk_mc_perfctr_reset(pcd->target_cntr);
+			break;
+
+		case PERF_CTRL_ENABLE:
+			ihk_mc_perfctr_start(pcd->target_cntr_mask);
+			break;
+			
+		case PERF_CTRL_DISABLE:
+			ihk_mc_perfctr_stop(pcd->target_cntr_mask);
+			break;
+
+		case PERF_CTRL_GET:
+			pcd->read_value = ihk_mc_perfctr_read(pcd->target_cntr);
+			break;
+			
+		default:
+			kprintf("%s: SCD_MSG_PERF_CTRL unexpected ctrl_type\n", __FUNCTION__);
+		}
+
+		ihk_mc_unmap_virtual(pcd, 1, 0);
+		ihk_mc_unmap_memory(NULL, pp, sizeof(struct perf_ctrl_desc));
+
+		pckt.msg = SCD_MSG_PERF_ACK;
+		pckt.err = 0;
 		pckt.arg = packet->arg;
-		syscall_channel_send(c, &pckt);
-		return 0;
+		ihk_ikc_send(c, &pckt, 0);
+
+		ret = 0;
+		break;
+
+	case SCD_MSG_CPU_RW_REG:
+
+		pckt.msg = SCD_MSG_CPU_RW_REG_RESP;
+		memcpy(&pckt.desc, &packet->desc,
+				sizeof(struct mcctrl_os_cpu_register));
+		pckt.resp = packet->resp;
+		pckt.err = arch_cpu_read_write_register(&pckt.desc, pckt.op);
+
+		ihk_ikc_send(c, &pckt, 0);
+		break;
 
 	default:
 		kprintf("syscall_pakcet_handler:unknown message "
 				"(%d.%d.%d.%d.%d.%#lx)\n",
 				packet->msg, packet->ref, packet->osnum,
 				packet->pid, packet->err, packet->arg);
-		return 0;
+		ret = 0;
+		break;
 
 	}
-	return 0;
+
+	ihk_ikc_release_packet((struct ihk_ikc_free_packet *)packet, c);
+	return ret;
 }
 
 void init_host_syscall_channel(void)
@@ -682,7 +676,7 @@ void init_host_syscall_channel(void)
 
 	param.port = 501;
 	param.pkt_size = sizeof(struct ikc_scd_packet);
-	param.queue_size = PAGE_SIZE;
+	param.queue_size = PAGE_SIZE * 4;
 	param.magic = 0x1129;
 	param.handler = syscall_packet_handler;
 
@@ -695,7 +689,6 @@ void init_host_syscall_channel(void)
 
 	get_this_cpu_local_var()->syscall_channel = param.channel;
 
-	process_msg_init(&cpu_local_var(iip), &cpu_local_var(scp));
 	pckt.msg = SCD_MSG_INIT_CHANNEL;
 	pckt.ref = ihk_mc_get_processor_id();
 	pckt.arg = virt_to_phys(&cpu_local_var(iip));
@@ -709,7 +702,7 @@ void init_host_syscall_channel2(void)
 
 	param.port = 502;
 	param.pkt_size = sizeof(struct ikc_scd_packet);
-	param.queue_size = PAGE_SIZE;
+	param.queue_size = PAGE_SIZE * 4;
 	param.magic = 0x1329;
 	param.handler = syscall_packet_handler;
 
@@ -722,7 +715,6 @@ void init_host_syscall_channel2(void)
 
 	get_this_cpu_local_var()->syscall_channel2 = param.channel;
 
-	process_msg_init(&cpu_local_var(iip2), &cpu_local_var(scp2));
 	pckt.msg = SCD_MSG_INIT_CHANNEL;
 	pckt.ref = ihk_mc_get_processor_id();
 	pckt.arg = virt_to_phys(&cpu_local_var(iip2));

kernel/include/cls.h

@@ -19,11 +19,33 @@
  * CPU Local Storage (cls)
  */
 
-struct malloc_header {
-	unsigned int check;
+struct kmalloc_header {
+	unsigned int front_magic;
 	unsigned int cpu_id;
-	struct malloc_header *next;
-	unsigned long size;
+	struct list_head list;
+	int size; /* The size of this chunk without the header */
+	unsigned int end_magic;
+	/* 32 bytes */
+};
+
+#define IHK_OS_MONITOR_NOT_BOOT 0
+#define IHK_OS_MONITOR_IDLE 1
+#define IHK_OS_MONITOR_USER 2
+#define IHK_OS_MONITOR_KERNEL 3
+#define IHK_OS_MONITOR_KERNEL_HEAVY 4
+#define IHK_OS_MONITOR_KERNEL_OFFLOAD 5
+#define IHK_OS_MONITOR_KERNEL_FREEZING 8
+#define IHK_OS_MONITOR_KERNEL_FROZEN 9
+#define IHK_OS_MONITOR_KERNEL_THAW 10
+#define IHK_OS_MONITOR_PANIC 99
+
+struct ihk_os_monitor {
+	int status;
+	int status_bak;
+	unsigned long counter;
+	unsigned long ocounter;
+	unsigned long user_tsc;
+	unsigned long system_tsc;
 };
 
 #include <ihk/lock.h>
@@ -38,8 +60,9 @@ extern ihk_spinlock_t	cpu_status_lock;
 
 struct cpu_local_var {
 	/* malloc */
-	struct malloc_header free_list;
-	struct malloc_header *remote_free_list;
+	struct list_head free_list;
+	struct list_head remote_free_list;
+	ihk_spinlock_t remote_free_list_lock;
 
 	struct thread idle;
 	struct process idle_proc;
@@ -53,11 +76,9 @@ struct cpu_local_var {
 	size_t runq_len;
 
 	struct ihk_ikc_channel_desc *syscall_channel;
-	struct syscall_params scp;
 	struct ikc_scd_init_param iip;
 
 	struct ihk_ikc_channel_desc *syscall_channel2;
-	struct syscall_params scp2;
 	struct ikc_scd_init_param iip2;
 	struct resource_set *resource_set;
 	
@@ -73,6 +94,8 @@ struct cpu_local_var {
 	int in_interrupt;
 	int no_preempt;
 	int timer_enabled;
+	int kmalloc_initialized;
+	struct ihk_os_monitor *monitor;
 } __attribute__((aligned(64)));
 
 

kernel/include/init.h

@@ -16,7 +16,7 @@
 extern void arch_init(void);
 extern void kmsg_init(int);
 extern void mem_init(void);
-extern void ikc_master_init(void);
+extern void ihk_ikc_master_init(void);
 extern void ap_init(void);
 extern void arch_ready(void);
 extern void mc_ikc_test_init(void);
@@ -29,7 +29,11 @@ extern void init_host_syscall_channel2(void);
 extern void sched_init(void);
 extern void pc_ap_init(void);
 extern void cpu_sysfs_setup(void);
+extern void rusage_sysfs_setup(void);
+extern void status_sysfs_setup(void);
 
 extern char *find_command_line(char *name);
 
+extern int num_processors;
+
 #endif

kernel/include/kmalloc.h

@@ -28,15 +28,14 @@ r;\
 })
 #define kfree(ptr) _kfree(ptr, __FILE__, __LINE__)
 #define memcheck(ptr, msg) _memcheck(ptr, msg, __FILE__, __LINE__, 0)
-void *_kmalloc(int size, enum ihk_mc_ap_flag flag, char *file, int line);
+void *_kmalloc(int size, ihk_mc_ap_flag flag, char *file, int line);
 void _kfree(void *ptr, char *file, int line);
-void *__kmalloc(int size, enum ihk_mc_ap_flag flag);
+void *__kmalloc(int size, ihk_mc_ap_flag flag);
 void __kfree(void *ptr);
-void *___kmalloc(int size, enum ihk_mc_ap_flag flag);
-void ___kfree(void *ptr);
 
 int _memcheck(void *ptr, char *msg, char *file, int line, int free);
 int memcheckall();
 int freecheck(int runcount);
+void kmalloc_consolidate_free_list(void);
 
 #endif

kernel/include/memobj.h

@@ -32,13 +32,23 @@ enum {
 	MF_HAS_PAGER	= 0x0001,
 	MF_SHMDT_OK	= 0x0002,
 	MF_IS_REMOVABLE	= 0x0004,
+	MF_PREFETCH = 0x0008,
+	MF_ZEROFILL = 0x0010,
+	MF_REG_FILE = 0x1000,
+	MF_DEV_FILE = 0x2000,
+	MF_HOST_RELEASED = 0x80000000,
+	MF_END
 };
 
+#define MEMOBJ_READY              0
+#define MEMOBJ_TO_BE_PREFETCHED   1
+
 struct memobj {
-	struct memobj_ops *	ops;
-	uint32_t		flags;
-	int8_t			padding[4];
-	ihk_spinlock_t		lock;
+	struct memobj_ops *ops;
+	uint32_t flags;
+	uint32_t status;
+	size_t size;
+	ihk_spinlock_t lock;
 };
 
 typedef void memobj_release_func_t(struct memobj *obj);
@@ -141,6 +151,7 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp);
 struct shmid_ds;
 int shmobj_create(struct shmid_ds *ds, struct memobj **objp);
 int zeroobj_create(struct memobj **objp);
-int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxprotp);
+int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxprotp,
+	int prot, int populate_flags);
 
 #endif /* HEADER_MEMOBJ_H */

kernel/include/page.h

@@ -17,8 +17,9 @@
 
 struct page {
 	struct list_head	list;
+	struct list_head	hash;
 	uint8_t			mode;
-	uint8_t			padding[3];
+	uint64_t		phys;
 	ihk_atomic_t		count;
 	off_t			offset;
 };
@@ -38,9 +39,8 @@ enum page_mode {
 struct page *phys_to_page(uintptr_t phys);
 uintptr_t page_to_phys(struct page *page);
 int page_unmap(struct page *page);
+struct page *phys_to_page_insert_hash(uint64_t phys);
 
-void *allocate_pages(int npages, enum ihk_mc_ap_flag flag);
-void free_pages(void *va, int npages);
 void begin_free_pages_pending(void);
 void finish_free_pages_pending(void);
 

kernel/include/pager.h

@@ -30,7 +30,8 @@ enum pager_op {
 struct pager_create_result {
 	uintptr_t	handle;
 	int		maxprot;
-	int8_t		padding[4];
+	uint32_t flags;
+	size_t size;
 };
 
 /*

kernel/include/process.h

@@ -22,6 +22,7 @@
 #include <memobj.h>
 #include <affinity.h>
 #include <syscall.h>
+#include <bitops.h>
 
 #define VR_NONE            0x0
 #define VR_STACK           0x1
@@ -29,6 +30,7 @@
 #define VR_IO_NOCACHE      0x100
 #define VR_REMOTE          0x200
 #define VR_WRITE_COMBINED  0x400
+#define VR_DONTFORK        0x800
 #define VR_DEMAND_PAGING   0x1000
 #define	VR_PRIVATE         0x2000
 #define	VR_LOCKED          0x4000
@@ -160,10 +162,73 @@
 #endif
 
 #define USER_STACK_NR_PAGES 8192
-#define KERNEL_STACK_NR_PAGES 25
+#define KERNEL_STACK_NR_PAGES 32
 
 #define NOPHYS ((uintptr_t)-1)
 
+#define PROCESS_NUMA_MASK_BITS 256
+
+/*
+ * Both the MPOL_* mempolicy mode and the MPOL_F_* optional mode flags are
+ * passed by the user to either set_mempolicy() or mbind() in an 'int' actual.
+ * The MPOL_MODE_FLAGS macro determines the legal set of optional mode flags.
+ */
+
+/* Policies */
+enum {
+	MPOL_DEFAULT,
+	MPOL_PREFERRED,
+	MPOL_BIND,
+	MPOL_INTERLEAVE,
+	MPOL_LOCAL,
+	MPOL_MAX,	/* always last member of enum */
+};
+
+enum mpol_rebind_step {
+	MPOL_REBIND_ONCE,	/* do rebind work at once(not by two step) */
+	MPOL_REBIND_STEP1,	/* first step(set all the newly nodes) */
+	MPOL_REBIND_STEP2,	/* second step(clean all the disallowed nodes)*/
+	MPOL_REBIND_NSTEP,
+};
+
+/* Flags for set_mempolicy */
+#define MPOL_F_STATIC_NODES	(1 << 15)
+#define MPOL_F_RELATIVE_NODES	(1 << 14)
+
+/*
+ * MPOL_MODE_FLAGS is the union of all possible optional mode flags passed to
+ * either set_mempolicy() or mbind().
+ */
+#define MPOL_MODE_FLAGS	(MPOL_F_STATIC_NODES | MPOL_F_RELATIVE_NODES)
+
+/* Flags for get_mempolicy */
+#define MPOL_F_NODE	(1<<0)	/* return next IL mode instead of node mask */
+#define MPOL_F_ADDR	(1<<1)	/* look up vma using address */
+#define MPOL_F_MEMS_ALLOWED (1<<2) /* return allowed memories */
+
+/* Flags for mbind */
+#define MPOL_MF_STRICT	(1<<0)	/* Verify existing pages in the mapping */
+#define MPOL_MF_MOVE	 (1<<1)	/* Move pages owned by this process to conform
+				   to policy */
+#define MPOL_MF_MOVE_ALL (1<<2)	/* Move every page to conform to policy */
+#define MPOL_MF_LAZY	 (1<<3)	/* Modifies '_MOVE:  lazy migrate on fault */
+#define MPOL_MF_INTERNAL (1<<4)	/* Internal flags start here */
+
+#define MPOL_MF_VALID	(MPOL_MF_STRICT   | 	\
+			 MPOL_MF_MOVE     | 	\
+			 MPOL_MF_MOVE_ALL)
+
+/*
+ * Internal flags that share the struct mempolicy flags word with
+ * "mode flags".  These flags are allocated from bit 0 up, as they
+ * are never OR'ed into the mode in mempolicy API arguments.
+ */
+#define MPOL_F_SHARED  (1 << 0)	/* identify shared policies */
+#define MPOL_F_LOCAL   (1 << 1)	/* preferred local allocation */
+#define MPOL_F_REBINDING (1 << 2)	/* identify policies in rebinding */
+#define MPOL_F_MOF	(1 << 3) /* this policy wants migrate on fault */
+#define MPOL_F_MORON	(1 << 4) /* Migrate On pte_numa Reference On Node */
+
 #include <waitq.h>
 #include <futex.h>
 
@@ -177,6 +242,28 @@ struct process_vm;
 struct vm_regions;
 struct vm_range;
 
+//#define TRACK_SYSCALLS
+
+#ifdef TRACK_SYSCALLS
+#define TRACK_SYSCALLS_MAX               300
+#define __NR_track_syscalls              701
+
+#define TRACK_SYSCALLS_CLEAR             0x01
+#define TRACK_SYSCALLS_ON                0x02
+#define TRACK_SYSCALLS_OFF               0x04
+#define TRACK_SYSCALLS_PRINT             0x08
+#define TRACK_SYSCALLS_PRINT_PROC        0x10
+
+void track_syscalls_print_thread_stats(struct thread *thread);
+void track_syscalls_print_proc_stats(struct process *proc);
+void track_syscalls_accumulate_counters(struct thread *thread,
+		struct process *proc);
+void track_syscalls_alloc_counters(struct thread *thread);
+void track_syscalls_dealloc_thread_counters(struct thread *thread);
+void track_syscalls_dealloc_proc_counters(struct process *proc);
+#endif // TRACK_SYSCALLS
+
+
 #define HASH_SIZE	73
 
 struct resource_set {
@@ -302,6 +389,14 @@ struct vm_range {
 	off_t objoff;
 	int pgshift;	/* page size. 0 means THP */
 	int padding;
+	void *private_data;
+};
+
+struct vm_range_numa_policy {
+	struct list_head list;
+	unsigned long start, end;
+	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
+	int numa_mem_policy;
 };
 
 struct vm_regions {
@@ -319,19 +414,21 @@ struct process_vm;
 struct mckfd {
 	struct mckfd *next;
 	int fd;
+	int sig_no;
 	long data;
 	void *opt;
 	long (*read_cb)(struct mckfd *, ihk_mc_user_context_t *);
 	int (*ioctl_cb)(struct mckfd *, ihk_mc_user_context_t *);
 	long (*mmap_cb)(struct mckfd *, ihk_mc_user_context_t *);
 	int (*close_cb)(struct mckfd *, ihk_mc_user_context_t *);
+	int (*fcntl_cb)(struct mckfd *, ihk_mc_user_context_t *);
 };
 
 #define SFD_CLOEXEC 02000000
 #define SFD_NONBLOCK 04000
 
 struct sig_common {
-	ihk_spinlock_t	lock;
+	mcs_rwlock_lock_t lock;
 	ihk_atomic_t use;
 	struct k_sigaction action[_NSIG];
 	struct list_head sigpending;
@@ -346,6 +443,11 @@ struct sig_pending {
 
 typedef void pgio_func_t(void *arg);
 
+struct mcexec_tid {
+	int tid;
+	struct thread *thread;
+};
+
 /* Represents a node in the process fork tree, it may exist even after the 
  * corresponding process exited due to references from the parent and/or 
  * children and is used for implementing wait/waitpid without having a 
@@ -360,6 +462,9 @@ struct process {
 	// threads and children
 	struct list_head threads_list;
 	mcs_rwlock_lock_t threads_lock; // lock for threads_list
+	/* TID set of proxy process */
+	struct mcexec_tid *tids;
+	int nr_tids;
 
 	/* The ptracing process behave as the parent of the ptraced process
 	   after using PTRACE_ATTACH except getppid. So we save it here. */
@@ -384,7 +489,7 @@ struct process {
 			// V       +----   |
 			// PS_STOPPED -----+
 			// (PS_TRACED)
-	int exit_status;
+	int exit_status; // only for zombie
 
 	/* Store exit_status for a group of threads when stopped by SIGSTOP.
 	   exit_status can't be used because values of exit_status of threads
@@ -455,6 +560,13 @@ struct process {
 #define PP_COUNT 2
 #define PP_STOP 3
 	struct mc_perf_event *monitoring_event;
+#ifdef TRACK_SYSCALLS
+	mcs_lock_node_t st_lock;
+	uint64_t *syscall_times;
+	uint32_t *syscall_cnts;
+	uint64_t *offload_times;
+	uint32_t *offload_cnts;
+#endif // TRACK_SYSCALLS
 };
 
 void hold_thread(struct thread *ftn);
@@ -496,6 +608,7 @@ struct thread {
 			// PS_TRACED
 			// PS_INTERRPUTIBLE
 			// PS_UNINTERRUPTIBLE
+	int exit_status;
 
 	// process vm
 	struct process_vm *vm;
@@ -526,12 +639,20 @@ struct thread {
 	fp_regs_struct *fp_regs;
 	int in_syscall_offload;
 
+#ifdef TRACK_SYSCALLS
+	int track_syscalls;
+	uint64_t *syscall_times;
+	uint32_t *syscall_cnts;
+	uint64_t *offload_times;
+	uint32_t *offload_cnts;
+#endif // TRACK_SYSCALLS
+
 	// signal
 	struct sig_common *sigcommon;
 	sigset_t sigmask;
 	stack_t sigstack;
 	struct list_head sigpending;
-	ihk_spinlock_t	sigpendinglock;
+	mcs_rwlock_lock_t sigpendinglock;
 	volatile int sigevent;
 
 	// gpio
@@ -544,9 +665,14 @@ struct thread {
 	struct sig_pending *ptrace_sendsig;
 
 	// cpu time
+	/*
 	struct timespec stime;
 	struct timespec utime;
 	struct timespec btime;
+	*/
+	unsigned long system_tsc;
+	unsigned long user_tsc;
+	unsigned long base_tsc;
 	int times_update;
 	int in_kernel;
 
@@ -556,8 +682,13 @@ struct thread {
 	struct itimerval itimer_prof;
 	struct timespec itimer_virtual_value;
 	struct timespec itimer_prof_value;
+
+	/* Syscall offload wait queue head */
+	struct waitq scd_wq;
 };
 
+#define VM_RANGE_CACHE_SIZE	4
+
 struct process_vm {
 	struct address_space *address_space;
 	struct list_head vm_range_list;
@@ -580,6 +711,12 @@ struct process_vm {
 	int exiting;
 
 	long currss;
+	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
+	int numa_mem_policy;
+	/* Protected by memory_range_lock */
+	struct list_head vm_range_numa_policy_list;
+	struct vm_range *range_cache[VM_RANGE_CACHE_SIZE];
+	int range_cache_ind;
 };
 
 static inline int has_cap_ipc_lock(struct thread *th)
@@ -596,7 +733,8 @@ static inline int has_cap_sys_admin(struct thread *th)
 
 void hold_address_space(struct address_space *);
 void release_address_space(struct address_space *);
-struct thread *create_thread(unsigned long user_pc);
+struct thread *create_thread(unsigned long user_pc,
+		unsigned long *__cpu_set, size_t cpu_set_size);
 struct thread *clone_thread(struct thread *org, unsigned long pc,
                               unsigned long sp, int clone_flags);
 void destroy_thread(struct thread *thread);
@@ -611,9 +749,10 @@ void free_process_memory_ranges(struct process_vm *vm);
 int populate_process_memory(struct process_vm *vm, void *start, size_t len);
 
 int add_process_memory_range(struct process_vm *vm,
-                             unsigned long start, unsigned long end,
-                             unsigned long phys, unsigned long flag,
-			     struct memobj *memobj, off_t objoff, int pgshift);
+		unsigned long start, unsigned long end,
+		unsigned long phys, unsigned long flag,
+		struct memobj *memobj, off_t offset,
+		int pgshift, struct vm_range **rp);
 int remove_process_memory_range(struct process_vm *vm, unsigned long start,
 		unsigned long end, int *ro_freedp);
 int split_process_memory_range(struct process_vm *vm,
@@ -653,9 +792,11 @@ extern enum ihk_mc_pt_attribute arch_vrflag_to_ptattr(unsigned long flag, uint64
 enum ihk_mc_pt_attribute common_vrflag_to_ptattr(unsigned long flag, uint64_t fault, pte_t *ptep);
 
 void schedule(void);
+void spin_sleep_or_schedule(void);
 void runq_add_thread(struct thread *thread, int cpu_id);
 void runq_del_thread(struct thread *thread, int cpu_id);
 int sched_wakeup_thread(struct thread *thread, int valid_states);
+int sched_wakeup_thread_locked(struct thread *thread, int valid_states);
 
 void sched_request_migrate(int cpu_id, struct thread *thread);
 void check_need_resched(void);
@@ -675,5 +816,6 @@ void chain_process(struct process *);
 void chain_thread(struct thread *);
 void proc_init();
 void set_timer();
+struct sig_pending *hassigpending(struct thread *thread);
 
 #endif

kernel/include/rusage.h

@@ -0,0 +1,79 @@
+#ifndef __RUSAGE_H
+#define __RUSAGE_H
+
+#define RUSAGE_DEFAULT_SIZE 10
+
+enum RUSAGE_MEMBER {
+    RUSAGE_RSS,
+    RUSAGE_CACHE,
+    RUSAGE_RSS_HUGE,
+    RUSAGE_MAPPED_FILE,
+    RUSAGE_MAX_USAGE,
+    RUSAGE_KMEM_USAGE,
+    RUSAGE_KMAX_USAGE,
+    RUSAGE_NUM_NUMA_NODES,
+    RUSAGE_NUMA_STAT,
+    RUSAGE_HUGETLB ,
+    RUSAGE_HUGETLB_MAX ,
+    RUSAGE_STAT_SYSTEM ,
+    RUSAGE_STAT_USER ,
+    RUSAGE_USAGE ,
+    RUSAGE_USAGE_PER_CPU ,
+    RUSAGE_NUM_THREADS ,
+    RUSAGE_MAX_NUM_THREADS
+};
+
+struct  r_data{
+    unsigned long pid;
+    unsigned long rss;
+    unsigned long cache;
+    unsigned long rss_huge;
+    unsigned long mapped_file;
+    unsigned long max_usage;
+    unsigned long kmem_usage;
+    unsigned long kmax_usage;
+    unsigned long hugetlb;
+    unsigned long hugetlb_max;
+    unsigned long stat_system;
+    unsigned long stat_user;
+    unsigned long usage;
+    struct r_data *next;
+} ;
+typedef struct r_data rusage_data;
+
+rusage_data *rdata[RUSAGE_DEFAULT_SIZE];
+unsigned long rusage_max_num_threads;
+unsigned long rusage_num_threads;
+
+enum ihk_os_status {
+    IHK_STATUS_INACTIVE,
+    IHK_STATUS_BOOTING,
+    IHK_STATUS_RUNNING,
+    IHK_STATUS_SHUTDOWN,
+    IHK_STATUS_PANIC,
+    IHK_STATUS_HUNGUP,
+    IHK_STATUS_FREEZING,
+    IHK_STATUS_FROZEN,
+};
+
+enum ihk_os_status os_status;
+unsigned long sys_delegate_count;
+enum sys_delegate_state_enum {
+	ENTER_KERNEL,
+	EXIT_KERNEL,
+};
+enum sys_delegate_state_enum sys_delegate_state;
+
+unsigned long rusage_rss[sizeof(cpu_set_t)/8];
+unsigned long rusage_rss_max;
+long rusage_rss_current;
+unsigned long rusage_kmem_usage;
+unsigned long rusage_kmem_max_usage;
+unsigned long rusage_hugetlb_usage;
+unsigned long rusage_hugetlb_max_usage;
+unsigned long rusage_numa_stat[1024];
+unsigned long rusage_max_memory;
+
+#define RUSAGE_MEM_LIMIT 2000000
+
+#endif

kernel/include/syscall.h

@@ -31,6 +31,7 @@
 #define SCD_MSG_PREPARE_PROCESS_ACKED   0x2
 #define SCD_MSG_PREPARE_PROCESS_NACKED  0x7
 #define SCD_MSG_SCHEDULE_PROCESS        0x3
+#define SCD_MSG_WAKE_UP_SYSCALL_THREAD  0x14
 
 #define SCD_MSG_INIT_CHANNEL            0x5
 #define SCD_MSG_INIT_CHANNEL_ACKED      0x6
@@ -72,6 +73,13 @@
 /* #define SCD_MSG_SYSFS_RESP_CLEANUP   0x43 */
 #define SCD_MSG_PROCFS_TID_CREATE	0x44
 #define SCD_MSG_PROCFS_TID_DELETE	0x45
+#define SCD_MSG_EVENT_SIGNAL		0x46
+
+#define SCD_MSG_PERF_CTRL               0x50
+#define SCD_MSG_PERF_ACK                0x51
+
+#define SCD_MSG_CPU_RW_REG              0x52
+#define SCD_MSG_CPU_RW_REG_RESP         0x53
 
 /* Cloning flags.  */
 # define CSIGNAL       0x000000ff /* Signal mask to be sent at exit.  */
@@ -117,28 +125,6 @@ struct user_desc {
 	unsigned int  lm:1;
 };
 
-struct ikc_scd_packet {
-	int msg;
-	int err;
-	union {
-		/* for traditional SCD_MSG_* */
-		struct {
-			int ref;
-			int osnum;
-			int pid;
-			int padding;
-			unsigned long arg;
-		};
-
-		/* for SCD_MSG_SYSFS_* */
-		struct {
-			long sysfs_arg1;
-			long sysfs_arg2;
-			long sysfs_arg3;
-		};
-	};
-};
-
 struct program_image_section {
 	unsigned long vaddr;
 	unsigned long len;
@@ -170,6 +156,10 @@ struct program_image_section {
 #define MCK_RLIMIT_SIGPENDING	14
 #define MCK_RLIMIT_STACK	15
 
+#define PLD_CPU_SET_MAX_CPUS 1024
+typedef unsigned long __cpu_set_unit;
+#define PLD_CPU_SET_SIZE (PLD_CPU_SET_MAX_CPUS / (8 * sizeof(__cpu_set_unit)))
+
 struct program_load_desc {
 	int num_sections;
 	int status;
@@ -199,6 +189,7 @@ struct program_load_desc {
 	struct rlimit rlimit[MCK_RLIM_MAX];
 	unsigned long interp_align;
 	char shell_path[SHELL_PATH_MAX_LEN];
+	__cpu_set_unit cpu_set[PLD_CPU_SET_SIZE];
 	struct program_image_section sections[0];
 };
 
@@ -210,13 +201,77 @@ struct ikc_scd_init_param {
 };
 
 struct syscall_request {
+	/* TID of requesting thread */
+	int rtid;
+	/*
+	 * TID of target thread. Remote page fault response needs to designate the
+	 * thread that must serve the request, 0 indicates any thread from the pool
+	 */
+	int ttid;
 	unsigned long valid;
 	unsigned long number;
 	unsigned long args[6];
 };
 
+struct mcctrl_os_cpu_register {
+	unsigned long addr;
+	unsigned long val;
+	unsigned long addr_ext;
+};
+
+enum mcctrl_os_cpu_operation {
+	MCCTRL_OS_CPU_READ_REGISTER,
+	MCCTRL_OS_CPU_WRITE_REGISTER,
+	MCCTRL_OS_CPU_MAX_OP
+};
+
+struct ikc_scd_packet {
+	int msg;
+	int err;
+	union {
+		/* for traditional SCD_MSG_* */
+		struct {
+			int ref;
+			int osnum;
+			int pid;
+			unsigned long arg;
+			struct syscall_request req;
+			unsigned long resp_pa;
+		};
+
+		/* for SCD_MSG_SYSFS_* */
+		struct {
+			long sysfs_arg1;
+			long sysfs_arg2;
+			long sysfs_arg3;
+		};
+
+		/* SCD_MSG_SCHEDULE_THREAD */
+		struct {
+			int ttid;
+		};
+
+		/* SCD_MSG_CPU_RW_REG */
+		struct {
+			struct mcctrl_os_cpu_register desc;
+			enum mcctrl_os_cpu_operation op;
+			void *resp;
+		};
+	};
+	char padding[12];
+};
+
+#define IHK_SCD_REQ_THREAD_SPINNING         0
+#define IHK_SCD_REQ_THREAD_TO_BE_WOKEN      1
+#define IHK_SCD_REQ_THREAD_DESCHEDULED      2
+
 struct syscall_response {
+	/* TID of the thread that requested the service */
+	int ttid;
+	/* TID of the mcexec thread that is serving the request */
+	int stid;
 	unsigned long status;
+	unsigned long req_thread_status;
 	long ret;
 	unsigned long fault_address;
 	unsigned long fault_reason;
@@ -226,22 +281,6 @@ struct syscall_post {
 	unsigned long v[8];
 };
 
-struct syscall_params {
-	unsigned long request_rpa, request_pa;
-	struct syscall_request *request_va;
-	unsigned long response_pa;
-	struct syscall_response *response_va;
-
-	unsigned long doorbell_rpa, doorbell_pa;
-	unsigned long *doorbell_va;
-
-	unsigned int post_idx;
-	unsigned long post_rpa, post_pa;
-	struct syscall_post *post_va;
-	unsigned long post_fin;
-	struct syscall_post post_buf IHK_DMA_ALIGN;
-};
-
 #define SYSCALL_DECLARE(name) long sys_##name(int n, ihk_mc_user_context_t *ctx)
 #define SYSCALL_HEADER struct syscall_request request IHK_DMA_ALIGN; \
 	request.number = n
@@ -307,7 +346,7 @@ void delete_proc_procfs_files(int pid);
 void create_os_procfs_files(void);
 void delete_os_procfs_files(void);
 
-#define PROCFS_NAME_MAX 1000
+#define PROCFS_NAME_MAX 768
 
 struct procfs_read {
 	unsigned long pbuf;	/* physical address of the host buffer (request) */
@@ -361,8 +400,32 @@ struct tod_data_s {
 };
 extern struct tod_data_s tod_data;	/* residing in arch-dependent file */
 
+static inline void tsc_to_ts(unsigned long tsc, struct timespec *ts)
+{
+	time_t sec_delta;
+	long ns_delta;
+
+	sec_delta = tsc / tod_data.clocks_per_sec;
+	ns_delta = NS_PER_SEC * (tsc % tod_data.clocks_per_sec)
+	           / tod_data.clocks_per_sec;
+	/* calc. of ns_delta overflows if clocks_per_sec exceeds 18.44 GHz */
+
+	ts->tv_sec = sec_delta;
+	ts->tv_nsec = ns_delta;
+	if (ts->tv_nsec >= NS_PER_SEC) {
+		ts->tv_nsec -= NS_PER_SEC;
+		++ts->tv_sec;
+	}
+}
+
+static inline unsigned long timespec_to_jiffy(const struct timespec *ats)
+{
+	return ats->tv_sec * 100 + ats->tv_nsec / 10000000;
+}
+
 void reset_cputime();
 void set_cputime(int mode);
+int do_munmap(void *addr, size_t len);
 intptr_t do_mmap(intptr_t addr0, size_t len0, int prot, int flags, int fd,
 		off_t off0);
 void clear_host_pte(uintptr_t addr, size_t len);
@@ -371,6 +434,8 @@ int do_shmget(key_t key, size_t size, int shmflg);
 struct process_vm;
 int arch_map_vdso(struct process_vm *vm);	/* arch dependent */
 int arch_setup_vdso(void);
+int arch_cpu_read_write_register(struct mcctrl_os_cpu_register *desc,
+		enum mcctrl_os_cpu_operation op);
 
 #define VDSO_MAXPAGES 2
 struct vdso {
@@ -407,4 +472,34 @@ struct get_cpu_mapping_req {
 #endif
 };
 
+enum perf_ctrl_type {
+	PERF_CTRL_SET,
+	PERF_CTRL_GET,
+	PERF_CTRL_ENABLE,
+	PERF_CTRL_DISABLE,
+};
+
+struct perf_ctrl_desc {
+	enum perf_ctrl_type ctrl_type;
+	int status;
+	union {
+		/* for SET, GET */
+		struct {
+			unsigned int target_cntr;
+			unsigned long config;
+			unsigned long read_value;
+			unsigned disabled        :1,
+			         pinned          :1,
+			         exclude_user    :1,
+			         exclude_kernel  :1,
+			         exclude_hv      :1,
+			         exclude_idle    :1;
+		};
+
+		/* for START, STOP*/
+		struct {
+			unsigned long target_cntr_mask;
+		};
+	};
+};
 #endif

kernel/include/waitq.h

@@ -27,6 +27,8 @@ typedef int (*waitq_func_t)(struct waitq_entry *wait, unsigned mode,
 
 int default_wake_function(struct waitq_entry *wait, unsigned mode, int flags,
 			              void *key);
+int locked_wake_function(struct waitq_entry *wait, unsigned mode, int flags,
+			              void *key);
 
 typedef struct waitq {
 	ihk_spinlock_t lock;
@@ -57,6 +59,13 @@ typedef struct waitq_entry {
 	    .link  = { &(name).link, &(name).link }			\
 	}
 
+#define DECLARE_WAITQ_ENTRY_LOCKED(name, tsk)           \
+	waitq_entry_t name = {								\
+	    .private  = tsk,								\
+		.func = locked_wake_function,					\
+	    .link  = { &(name).link, &(name).link }			\
+	}
+
 extern void waitq_init(waitq_t *waitq);
 extern void waitq_init_entry(waitq_entry_t *entry, struct thread *proc);
 extern int waitq_active(waitq_t *waitq);

kernel/include/xpmem.h

@@ -0,0 +1,26 @@
+/**
+ * \file xpmem.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Structures and functions of xpmem
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef _XPMEM_H
+#define _XPMEM_H
+
+#include <process.h>
+#include <ihk/context.h>
+
+#define XPMEM_DEV_PATH  "/dev/xpmem"
+
+extern int xpmem_open(ihk_mc_user_context_t *ctx);
+extern int xpmem_remove_process_memory_range(struct process_vm *vm, 
+	struct vm_range *vmr);
+extern int xpmem_fault_process_memory_range(struct process_vm *vm, 
+	struct vm_range *vmr, unsigned long vaddr, uint64_t reason);
+
+#endif /* _XPMEM_H */
+

kernel/include/xpmem_private.h

@@ -0,0 +1,490 @@
+/**
+ * \file xpmem_private.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Private Cross Partition Memory (XPMEM) structures and macros.
+ */
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (c) 2004-2007 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright 2009, 2010, 2014 Cray Inc. All Rights Reserved
+ * Copyright (c) 2014-2016 Los Alamos National Security, LCC. All rights
+ *                         reserved.
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef _XPMEM_PRIVATE_H
+#define _XPMEM_PRIVATE_H
+
+#include <mc_xpmem.h>
+#include <xpmem.h>
+
+#define XPMEM_CURRENT_VERSION           0x00026003
+
+//#define DEBUG_PRINT_XPMEM
+
+#ifdef DEBUG_PRINT_XPMEM
+#define dkprintf(...) kprintf(__VA_ARGS__)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#define XPMEM_DEBUG(format, a...) kprintf("[%d] %s: "format"\n", cpu_local_var(current)->proc->rgid, __func__, ##a)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#define XPMEM_DEBUG(format, a...) do { if (0) kprintf("\n"); } while (0)
+#endif
+
+//#define USE_DBUG_ON
+
+#ifdef USE_DBUG_ON
+#define DBUG_ON(condition) do { if (condition) kprintf("[%d] BUG: func=%s\n", cpu_local_var(current)->proc->rgid, __func__); } while (0)
+#else
+#define DBUG_ON(condition)
+#endif
+
+#define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
+
+#define min(x, y)	({                                              \
+	__typeof__(x) _min1 = (x);                                      \
+	__typeof__(y) _min2 = (y);                                      \
+	(void) (&_min1 == &_min2);                                      \
+	_min1 < _min2 ? _min1 : _min2;})
+
+#define max(x, y)	({                                              \
+	__typeof__(x) _max1 = (x);                                      \
+	__typeof__(y) _max2 = (y);                                      \
+	(void) (&_max1 == &_max2);                                      \
+	_max1 > _max2 ? _max1 : _max2;})
+
+#define MAX_ERRNO	4095
+
+#define IS_ERR_VALUE(x)	((x) >= (unsigned long)-MAX_ERRNO)
+
+static inline void * ERR_PTR(long error)
+{
+	return (void *)error;
+}
+
+static inline long PTR_ERR(const void *ptr)
+{
+	return (long)ptr;
+}
+
+static inline long IS_ERR(const void *ptr)
+{
+	return IS_ERR_VALUE((unsigned long)ptr);
+}
+
+static inline long IS_ERR_OR_NULL(const void *ptr)
+{
+	return !ptr || IS_ERR_VALUE((unsigned long)ptr);
+}
+
+/*
+ * Both the xpmem_segid_t and xpmem_apid_t are of type __s64 and designed
+ * to be opaque to the user. Both consist of the same underlying fields.
+ *
+ * The 'uniq' field is designed to give each segid or apid a unique value.
+ * Each type is only unique with respect to itself.
+ *
+ * An ID is never less than or equal to zero.
+ */
+struct xpmem_id {
+	pid_t tgid;		/* thread group that owns ID */
+	unsigned int uniq;	/* this value makes the ID unique */
+};
+
+typedef union {
+	struct xpmem_id xpmem_id;
+	xpmem_segid_t segid;
+	xpmem_apid_t apid;
+} xpmem_id_t;
+
+/* Shift INT_MAX by one so we can tell when we overflow. */
+#define XPMEM_MAX_UNIQ_ID	(INT_MAX >> 1)
+
+static inline pid_t xpmem_segid_to_tgid(xpmem_segid_t segid)
+{
+	DBUG_ON(segid <= 0);
+	return ((xpmem_id_t *)&segid)->xpmem_id.tgid;
+}
+
+static inline pid_t xpmem_apid_to_tgid(xpmem_apid_t apid)
+{
+	DBUG_ON(apid <= 0);
+	return ((xpmem_id_t *)&apid)->xpmem_id.tgid;
+}
+
+/*
+ * Hash Tables
+ *
+ * XPMEM utilizes hash tables to enable faster lookups of list entries.
+ * These hash tables are implemented as arrays. A simple modulus of the hash
+ * key yields the appropriate array index. A hash table's array element (i.e.,
+ * hash table bucket) consists of a hash list and the lock that protects it.
+ *
+ * XPMEM has the following two hash tables:
+ *
+ * table                bucket                                  key
+ * part->tg_hashtable   list of struct xpmem_thread_group       tgid
+ * tg->ap_hashtable     list of struct xpmem_access_permit      apid.uniq
+ */
+struct xpmem_hashlist {
+        mcs_rwlock_lock_t lock;	/* lock for hash list */
+        struct list_head list;	/* hash list */
+};
+
+#define XPMEM_TG_HASHTABLE_SIZE 8
+#define XPMEM_AP_HASHTABLE_SIZE 8
+
+static inline int xpmem_tg_hashtable_index(pid_t tgid)
+{
+	int index;
+
+	index = (unsigned int)tgid % XPMEM_TG_HASHTABLE_SIZE;
+
+	XPMEM_DEBUG("return: tgid=%lu, index=%d", tgid, index);
+
+	return index;
+}
+
+static inline int xpmem_ap_hashtable_index(xpmem_apid_t apid)
+{
+	int index;
+
+        DBUG_ON(apid <= 0);
+
+	index = ((xpmem_id_t *)&apid)->xpmem_id.uniq % XPMEM_AP_HASHTABLE_SIZE;
+
+	XPMEM_DEBUG("return: apid=0x%lx, index=%d", apid, index);
+
+	return index;
+}
+
+/*
+ * general internal driver structures
+ */
+struct xpmem_thread_group {
+	ihk_spinlock_t lock;	/* tg lock */
+	pid_t tgid;		/* tg's tgid */
+	uid_t uid;		/* tg's uid */
+	gid_t gid;		/* tg's gid */
+	volatile int flags;	/* tg attributes and state */
+	ihk_atomic_t uniq_segid;	/* segid uniq */
+	ihk_atomic_t uniq_apid;	/* apid uniq */
+	mcs_rwlock_lock_t seg_list_lock;	/* tg's list of segs lock */
+	struct list_head seg_list;	/* tg's list of segs */
+	ihk_atomic_t refcnt;	/* references to tg */
+	ihk_atomic_t n_pinned;	/* #of pages pinned by this tg */
+	struct list_head tg_hashlist;	/* tg hash list */
+	struct thread *group_leader;	/* thread group leader */
+	struct process_vm *vm;		/* tg's process_vm */
+	struct xpmem_hashlist ap_hashtable[];	/* locks + ap hash lists */
+};
+
+struct xpmem_segment {
+	ihk_spinlock_t lock;	/* seg lock */
+	xpmem_segid_t segid;	/* unique segid */
+	unsigned long vaddr;	/* starting address */
+	size_t size;		/* size of seg */
+	int permit_type;	/* permission scheme */
+	void *permit_value;	/* permission data */
+	volatile int flags;	/* seg attributes and state */
+	ihk_atomic_t refcnt;	/* references to seg */
+	struct xpmem_thread_group *tg;	/* creator tg */
+	struct list_head ap_list;	/* local access permits of seg */
+	struct list_head seg_list;	/* tg's list of segs */
+};
+
+struct xpmem_access_permit {
+	ihk_spinlock_t lock;	/* access permit lock */
+	xpmem_apid_t apid;	/* unique apid */
+	int mode;		/* read/write mode */
+	volatile int flags;	/* access permit attributes and state */
+	ihk_atomic_t refcnt;	/* references to access permit */
+	struct xpmem_segment *seg;	/* seg permitted to be accessed */
+	struct xpmem_thread_group *tg;	/* access permit's tg */
+	struct list_head att_list;	/* atts of this access permit's seg */
+	struct list_head ap_list;	/* access permits linked to seg */
+	struct list_head ap_hashlist;	/* access permit hash list */
+};
+
+struct xpmem_attachment {
+	mcs_rwlock_lock_t at_lock;	/* att lock */
+	unsigned long vaddr;	/* starting address of seg attached */
+	unsigned long at_vaddr;	/* address where seg is attached */
+	size_t at_size;		/* size of seg attachment */
+	struct vm_range *at_vmr;	/* vm_range where seg is attachment */
+	volatile int flags;	/* att attributes and state */
+	ihk_atomic_t refcnt;	/* references to att */
+	struct xpmem_access_permit *ap;	/* associated access permit */
+	struct list_head att_list;	/* atts linked to access permit */
+	struct process_vm *vm;	/* process_vm attached to */
+};
+
+struct xpmem_partition {
+	ihk_atomic_t n_opened;	/* # of /dev/xpmem opened */
+	struct xpmem_hashlist tg_hashtable[];	/* locks + tg hash lists */
+};
+
+#define XPMEM_FLAG_DESTROYING		0x00040 /* being destroyed */
+#define XPMEM_FLAG_DESTROYED		0x00080 /* 'being destroyed' finished */
+
+#define XPMEM_FLAG_VALIDPTEs		0x00200 /* valid PTEs exist */
+
+struct xpmem_perm {
+	uid_t uid;
+	gid_t gid;
+	unsigned long mode;
+};
+
+#define XPMEM_PERM_IRUSR 00400
+#define XPMEM_PERM_IWUSR 00200
+
+extern struct xpmem_partition *xpmem_my_part;
+
+static int xpmem_ioctl(struct mckfd *mckfd, ihk_mc_user_context_t *ctx);
+static int xpmem_close(struct mckfd *mckfd, ihk_mc_user_context_t *ctx);
+
+static int xpmem_init(void);
+static void xpmem_exit(void);
+static int __xpmem_open(void);
+static void xpmem_destroy_tg(struct xpmem_thread_group *);
+
+static int xpmem_make(unsigned long, size_t, int, void *, xpmem_segid_t *);
+static xpmem_segid_t xpmem_make_segid(struct xpmem_thread_group *);
+
+static int xpmem_remove(xpmem_segid_t);
+static void xpmem_remove_seg(struct xpmem_thread_group *,
+        struct xpmem_segment *);
+static void xpmem_remove_segs_of_tg(struct xpmem_thread_group *seg_tg);
+
+static int xpmem_get(xpmem_segid_t, int, int, void *, xpmem_apid_t *);
+static int xpmem_check_permit_mode(int, struct xpmem_segment *);
+static int xpmem_perms(struct xpmem_perm *, short);
+static xpmem_apid_t xpmem_make_apid(struct xpmem_thread_group *);
+
+static int xpmem_release(xpmem_apid_t);
+static void xpmem_release_ap(struct xpmem_thread_group *,
+	struct xpmem_access_permit *);
+static void xpmem_release_aps_of_tg(struct xpmem_thread_group *ap_tg);
+
+static int xpmem_attach(struct mckfd *, xpmem_apid_t, off_t, size_t, 
+	unsigned long, int, int, unsigned long *);
+
+static int xpmem_detach(unsigned long);
+static int xpmem_vm_munmap(struct process_vm *vm, void *addr, size_t len);
+static int xpmem_remove_process_range(struct process_vm *vm, 
+	unsigned long start, unsigned long end, int *ro_freedp);
+static int xpmem_free_process_memory_range(struct process_vm *vm,
+	struct vm_range *range);
+static void xpmem_detach_att(struct xpmem_access_permit *, 
+	struct xpmem_attachment *);
+static void xpmem_clear_PTEs(struct xpmem_segment *);
+static void xpmem_clear_PTEs_range(struct xpmem_segment *, unsigned long,
+	unsigned long);
+static void xpmem_clear_PTEs_of_ap(struct xpmem_access_permit *, unsigned long, 
+	unsigned long);
+static void xpmem_clear_PTEs_of_att(struct xpmem_attachment *, unsigned long, 
+	unsigned long);
+
+static int xpmem_remap_pte(struct process_vm *, struct vm_range *,
+	unsigned long, uint64_t, struct xpmem_segment *, unsigned long);
+
+static int xpmem_ensure_valid_page(struct xpmem_segment *, unsigned long);
+static pte_t * xpmem_vaddr_to_pte(struct process_vm *, unsigned long, 
+	size_t *pgsize);
+static int xpmem_pin_page(struct xpmem_thread_group *, struct thread *,
+	struct process_vm *, unsigned long);
+static void xpmem_unpin_pages(struct xpmem_segment *, struct process_vm *, 
+	unsigned long, size_t);
+
+static struct xpmem_thread_group * __xpmem_tg_ref_by_tgid_nolock_internal(
+	pid_t, int, int);
+
+static inline struct xpmem_thread_group *__xpmem_tg_ref_by_tgid(
+	pid_t tgid,
+	int return_destroying)
+{
+	struct xpmem_thread_group *tg;
+	int index;
+	struct mcs_rwlock_node_irqsave lock;
+
+	XPMEM_DEBUG("call: tgid=%d, return_destroying=%d", 
+		tgid, return_destroying);
+
+	index = xpmem_tg_hashtable_index(tgid);
+	mcs_rwlock_reader_lock(&xpmem_my_part->tg_hashtable[index].lock, &lock);
+	tg = __xpmem_tg_ref_by_tgid_nolock_internal(tgid, index, 
+		return_destroying);
+        mcs_rwlock_reader_unlock(&xpmem_my_part->tg_hashtable[index].lock, 
+		&lock);
+
+	XPMEM_DEBUG("return: tg=0x%p", tg);
+
+        return tg;
+}
+
+static inline struct xpmem_thread_group *__xpmem_tg_ref_by_tgid_nolock(
+	pid_t tgid,
+	int return_destroying)
+{
+	struct xpmem_thread_group *tg;
+
+	XPMEM_DEBUG("call: tgid=%d, return_destroying=%d", 
+		tgid, return_destroying);
+
+        tg = __xpmem_tg_ref_by_tgid_nolock_internal(tgid, 
+		xpmem_tg_hashtable_index(tgid), return_destroying);
+
+	XPMEM_DEBUG("return: tg=0x%p", tg);
+
+        return tg;
+}
+
+#define xpmem_tg_ref_by_tgid(t)             __xpmem_tg_ref_by_tgid(t, 0)
+#define xpmem_tg_ref_by_tgid_all(t)         __xpmem_tg_ref_by_tgid(t, 1)
+#define xpmem_tg_ref_by_tgid_nolock(t)      __xpmem_tg_ref_by_tgid_nolock(t, 0)
+#define xpmem_tg_ref_by_tgid_all_nolock(t)  __xpmem_tg_ref_by_tgid_nolock(t, 1)
+
+static struct xpmem_thread_group * xpmem_tg_ref_by_segid(xpmem_segid_t);
+static struct xpmem_thread_group * xpmem_tg_ref_by_apid(xpmem_apid_t);
+static void xpmem_tg_deref(struct xpmem_thread_group *);
+static struct xpmem_segment *xpmem_seg_ref_by_segid(struct xpmem_thread_group *,
+	xpmem_segid_t);
+static void xpmem_seg_deref(struct xpmem_segment *);
+static struct xpmem_access_permit * xpmem_ap_ref_by_apid(
+	struct xpmem_thread_group *, xpmem_apid_t);
+static void xpmem_ap_deref(struct xpmem_access_permit *);
+static void xpmem_att_deref(struct xpmem_attachment *);
+static int xpmem_validate_access(struct xpmem_access_permit *, off_t, size_t,
+	int, unsigned long *);
+
+/*
+ * Inlines that mark an internal driver structure as being destroyable or not.
+ * The idea is to set the refcnt to 1 at structure creation time and then
+ * drop that reference at the time the structure is to be destroyed.
+ */
+static inline void xpmem_tg_not_destroyable(
+	struct xpmem_thread_group *tg)
+{
+	ihk_atomic_set(&tg->refcnt, 1);
+
+	XPMEM_DEBUG("return: tg->refcnt=%d", tg->refcnt);
+}
+
+static inline void xpmem_tg_destroyable(
+	struct xpmem_thread_group *tg)
+{
+	XPMEM_DEBUG("call: ");
+
+	xpmem_tg_deref(tg);
+
+	XPMEM_DEBUG("return: ");
+}
+
+static inline void xpmem_seg_not_destroyable(
+	struct xpmem_segment *seg)
+{
+	ihk_atomic_set(&seg->refcnt, 1);
+
+	XPMEM_DEBUG("return: seg->refcnt=%d", seg->refcnt);
+}
+
+static inline void xpmem_seg_destroyable(
+	struct xpmem_segment *seg)
+{
+	XPMEM_DEBUG("call: ");
+
+	xpmem_seg_deref(seg);
+
+	XPMEM_DEBUG("return: ");
+}
+
+static inline void xpmem_ap_not_destroyable(
+	struct xpmem_access_permit *ap)
+{
+	ihk_atomic_set(&ap->refcnt, 1);
+
+	XPMEM_DEBUG("return: ap->refcnt=%d", ap->refcnt);
+}
+
+static inline void xpmem_ap_destroyable(
+	struct xpmem_access_permit *ap)
+{
+	XPMEM_DEBUG("call: ");
+
+	xpmem_ap_deref(ap);
+
+	XPMEM_DEBUG("return: ");
+}
+
+static inline void xpmem_att_not_destroyable(
+	struct xpmem_attachment *att)
+{
+	ihk_atomic_set(&att->refcnt, 1);
+
+	XPMEM_DEBUG("return: att->refcnt=%d", att->refcnt);
+}
+
+static inline void xpmem_att_destroyable(
+	struct xpmem_attachment *att)
+{
+	XPMEM_DEBUG("call: ");
+
+	xpmem_att_deref(att);
+
+	XPMEM_DEBUG("return: ");
+}
+
+/*
+ * Inlines that increment the refcnt for the specified structure.
+ */
+static inline void xpmem_tg_ref(
+	struct xpmem_thread_group *tg)
+{
+	DBUG_ON(ihk_atomic_read(&tg->refcnt) <= 0);
+	ihk_atomic_inc(&tg->refcnt);
+
+	XPMEM_DEBUG("return: tg->refcnt=%d", tg->refcnt);
+}
+
+static inline void xpmem_seg_ref(
+	struct xpmem_segment *seg)
+{
+	DBUG_ON(ihk_atomic_read(&seg->refcnt) <= 0);
+	ihk_atomic_inc(&seg->refcnt);
+
+	XPMEM_DEBUG("return: seg->refcnt=%d", seg->refcnt);
+}
+
+static inline void xpmem_ap_ref(
+	struct xpmem_access_permit *ap)
+{
+	DBUG_ON(ihk_atomic_read(&ap->refcnt) <= 0);
+	ihk_atomic_inc(&ap->refcnt);
+
+	XPMEM_DEBUG("return: ap->refcnt=%d", ap->refcnt);
+}
+
+static inline void xpmem_att_ref(
+	struct xpmem_attachment *att)
+{
+	DBUG_ON(ihk_atomic_read(&att->refcnt) <= 0);
+	ihk_atomic_inc(&att->refcnt);
+
+	XPMEM_DEBUG("return: att->refcnt=%d", att->refcnt);
+}
+
+static inline int xpmem_is_private_data(
+	struct vm_range *vmr)
+{
+        return (vmr->private_data != NULL);
+}
+
+#endif /* _XPMEM_PRIVATE_H */
+

kernel/init.c

@@ -31,6 +31,7 @@
 #include <cls.h>
 #include <syscall.h>
 #include <sysfs.h>
+#include <rusage.h>
 
 //#define IOCTL_FUNC_EXTENSION
 #ifdef IOCTL_FUNC_EXTENSION
@@ -108,11 +109,11 @@ static void dma_test(void)
 }
 #endif
 
-extern char *ihk_mc_get_kernel_args(void);
+extern char *ihk_get_kargs(void);
 
 char *find_command_line(char *name)
 {
-	char *cmdline = ihk_mc_get_kernel_args();
+	char *cmdline = ihk_get_kargs();
 
 	if (!cmdline) {
 		return NULL;
@@ -122,7 +123,7 @@ char *find_command_line(char *name)
 
 static void parse_kargs(void)
 {
-	kprintf("KCommand Line: %s\n", ihk_mc_get_kernel_args());
+	kprintf("KCommand Line: %s\n", ihk_get_kargs());
 
 	if (1) {
 		char *key = "osnum=";
@@ -239,6 +240,31 @@ static void time_init(void)
 	return;
 }
 
+struct ihk_os_monitor *monitor;
+
+static void monitor_init()
+{
+	int z;
+	unsigned long phys;
+
+	z = sizeof(struct ihk_os_monitor) * num_processors;
+	z = (z + PAGE_SIZE -1) >> PAGE_SHIFT;
+	monitor = ihk_mc_alloc_pages(z, IHK_MC_AP_CRITICAL);
+	memset(monitor, 0, z * PAGE_SIZE);
+	phys = virt_to_phys(monitor);
+	ihk_set_monitor(phys, sizeof(struct ihk_os_monitor) * num_processors);
+}
+
+int nmi_mode;
+
+static void nmi_init()
+{
+	unsigned long phys;
+
+	phys = virt_to_phys(&nmi_mode);
+	ihk_set_nmi_mode_addr(phys);
+}
+
 static void rest_init(void)
 {
 	handler_init();
@@ -250,11 +276,13 @@ static void rest_init(void)
 	//pc_test();
 
 	ap_init();
+	monitor_init();
 	cpu_local_var_init();
+	nmi_init();
 	time_init();
 	kmalloc_init();
 
-	ikc_master_init();
+	ihk_ikc_master_init();
 
 	proc_init();
 
@@ -320,7 +348,11 @@ static void setup_remote_snooping_samples(void)
 static void populate_sysfs(void)
 {
 	cpu_sysfs_setup();
-	setup_remote_snooping_samples();
+#ifdef ENABLE_RUSAGE
+	rusage_sysfs_setup();
+	status_sysfs_setup();
+#endif
+	//setup_remote_snooping_samples();
 } /* populate_sysfs() */
 
 int host_ikc_inited = 0;
@@ -336,11 +368,8 @@ static void post_init(void)
 	}
 
 	if (find_command_line("hidos")) {
-		extern ihk_spinlock_t syscall_lock;
-
 		init_host_syscall_channel();
 		init_host_syscall_channel2();
-		ihk_mc_spinlock_init(&syscall_lock);
 	}
 
 	arch_setup_vdso();
@@ -364,6 +393,21 @@ int main(void)
 	char *ptr;
 	int mode = 0;
 
+#ifdef ENABLE_RUSAGE
+	int i;
+	os_status = IHK_STATUS_INACTIVE;
+	rusage_hugetlb_usage = 0;
+	rusage_hugetlb_max_usage = 0;
+	for (i = 0; i < sizeof(cpu_set_t)/8; i++) {
+		rusage_rss[i] = 0;
+	}
+	for (i = 0; i < 1024; i++) {
+		rusage_numa_stat[i] = 0;
+	}
+	rusage_rss_current = 0;
+	rusage_rss_max = 0;
+#endif
+
 	ptr = find_command_line("ksyslogd=");
 	if (ptr) {
 	    mode = ptr[9] - 0x30;
@@ -371,8 +415,11 @@ int main(void)
 	}
 	kmsg_init(mode);
 
-	kputs("MCK started.\n");
-
+	kputs("IHK/McKernel started.\n");
+#ifdef ENABLE_RUSAGE
+	os_status = IHK_STATUS_BOOTING;
+#endif
+	ihk_set_kmsg(virt_to_phys(&kmsg_buf), IHK_KMSG_SIZE);
 	arch_init();
 
 	/*
@@ -393,7 +440,10 @@ int main(void)
 
 	futex_init();
 
-	kputs("MCK/IHK booted.\n");
+	kputs("IHK/McKernel booted.\n");
+#ifdef ENABLE_RUSAGE
+	os_status = IHK_STATUS_RUNNING;
+#endif
 
 #ifdef DCFA_KMOD
 	mc_cmd_client_init();

kernel/mikc.c

@@ -21,7 +21,7 @@ static struct ihk_ikc_channel_desc *mchannel;
 static int arch_master_channel_packet_handler(struct ihk_ikc_channel_desc *,
                                          void *__packet, void *arg);
 
-void ikc_master_init(void)
+void ihk_ikc_master_init(void)
 {
 	mchannel = kmalloc(sizeof(struct ihk_ikc_channel_desc) +
 	                   sizeof(struct ihk_ikc_master_packet),

kernel/procfs.c

@@ -17,12 +17,14 @@
 #include <ihk/debug.h>
 #include <ihk/ikc.h>
 #include <ikc/master.h>
-#include <syscall.h>
 #include <cls.h>
+#include <syscall.h>
 #include <kmalloc.h>
 #include <process.h>
 #include <page.h>
 #include <mman.h>
+#include <bitmap.h>
+#include <init.h>
 
 //#define DEBUG_PRINT_PROCFS
 
@@ -35,6 +37,7 @@
 extern int snprintf(char * buf, size_t size, const char *fmt, ...);
 extern int sprintf(char * buf, const char *fmt, ...);
 extern int sscanf(const char * buf, const char * fmt, ...);
+extern int scnprintf(char * buf, size_t size, const char *fmt, ...);
 
 extern int osnum;
 
@@ -73,11 +76,11 @@ procfs_delete_thread(struct thread *thread)
  *
  * \param rarg returned argument
  */
-void
-process_procfs_request(unsigned long rarg)
+void process_procfs_request(struct ikc_scd_packet *rpacket)
 {
+	unsigned long rarg = rpacket->arg;
 	unsigned long parg, pbuf;
-        struct thread *thread = NULL;
+	struct thread *thread = NULL;
 	struct process *proc = NULL;
 	struct process_vm *vm = NULL;
 	struct procfs_read *r;
@@ -158,7 +161,7 @@ process_procfs_request(unsigned long rarg)
 	 */
 	ret = sscanf(p, "%d/", &pid);
 	if (ret == 1) {
-		struct mcs_rwlock_node tlock;
+		struct mcs_rwlock_node_irqsave tlock;
 		int tids;
 		struct thread *thread1 = NULL;
 
@@ -175,7 +178,7 @@ process_procfs_request(unsigned long rarg)
 		else
 			tid = pid;
 
-		mcs_rwlock_reader_lock_noirq(&proc->threads_lock, &tlock);
+		mcs_rwlock_reader_lock(&proc->threads_lock, &tlock);
 		list_for_each_entry(thread, &proc->threads_list, siblings_list){
 			if(thread->tid == tid)
 				break;
@@ -185,15 +188,15 @@ process_procfs_request(unsigned long rarg)
 		if(thread == NULL){
 			kprintf("process_procfs_request: no such tid %d-%d\n", pid, tid);
 			if(tids){
+				mcs_rwlock_reader_unlock(&proc->threads_lock, &tlock);
 				process_unlock(proc, &lock);
-				mcs_rwlock_reader_unlock_noirq(&proc->threads_lock, &tlock);
 				goto end;
 			}
 			thread = thread1;
 		}
 		if(thread)
 			hold_thread(thread);
-		mcs_rwlock_reader_unlock_noirq(&proc->threads_lock, &tlock);
+		mcs_rwlock_reader_unlock(&proc->threads_lock, &tlock);
 		hold_process(proc);
 		vm = proc->vm;
 		if(vm)
@@ -281,6 +284,13 @@ process_procfs_request(unsigned long rarg)
 					ans = -EIO;
 				goto end;
 			}
+
+			if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+					pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+				ans = -EIO;
+				goto end;
+			}
+
 			va = phys_to_virt(pa);
 			if(readwrite)
 				memcpy(va, buf + ans, size);
@@ -397,12 +407,34 @@ process_procfs_request(unsigned long rarg)
 	/* 
 	 * mcos%d/PID/status
 	 */
+#define BITMASKS_BUF_SIZE	2048
 	if (strcmp(p, "status") == 0) {
+		extern int num_processors;	/* kernel/ap.c */
 		struct vm_range *range;
 		unsigned long lockedsize = 0;
-		char tmp[1024];
+		char *tmp;
+		char *bitmasks;
+		int bitmasks_offset = 0;
+		char *cpu_bitmask, *cpu_list, *numa_bitmask, *numa_list;
 		int len;
 
+		tmp = kmalloc(8192, IHK_MC_AP_CRITICAL);
+		if (!tmp) {
+			kprintf("%s: error allocating /proc/self/status buffer\n",
+				__FUNCTION__);
+			ans = 0;
+			goto end;
+		}
+
+		bitmasks = kmalloc(BITMASKS_BUF_SIZE, IHK_MC_AP_CRITICAL);
+		if (!tmp) {
+			kprintf("%s: error allocating /proc/self/status bitmaks buffer\n",
+				__FUNCTION__);
+			kfree(tmp);
+			ans = 0;
+			goto end;
+		}
+
 		ihk_mc_spinlock_lock_noirq(&proc->vm->memory_range_lock);
 		list_for_each_entry(range, &proc->vm->vm_range_list, list) {
 			if(range->flag & VR_LOCKED)
@@ -410,13 +442,42 @@ process_procfs_request(unsigned long rarg)
 		}
 		ihk_mc_spinlock_unlock_noirq(&proc->vm->memory_range_lock);
 
+		cpu_bitmask = &bitmasks[bitmasks_offset];
+		bitmasks_offset += bitmap_scnprintf(cpu_bitmask,
+				BITMASKS_BUF_SIZE - bitmasks_offset,
+				thread->cpu_set.__bits, num_processors);
+		bitmasks_offset++;
+
+		cpu_list = &bitmasks[bitmasks_offset];
+		bitmasks_offset += bitmap_scnlistprintf(cpu_list,
+				BITMASKS_BUF_SIZE - bitmasks_offset,
+				thread->cpu_set.__bits, __CPU_SETSIZE);
+		bitmasks_offset++;
+
+		numa_bitmask = &bitmasks[bitmasks_offset];
+		bitmasks_offset += bitmap_scnprintf(numa_bitmask,
+				BITMASKS_BUF_SIZE - bitmasks_offset,
+				proc->vm->numa_mask, PROCESS_NUMA_MASK_BITS);
+		bitmasks_offset++;
+
+		numa_list = &bitmasks[bitmasks_offset];
+		bitmasks_offset += bitmap_scnlistprintf(numa_list,
+				BITMASKS_BUF_SIZE - bitmasks_offset,
+				proc->vm->numa_mask, PROCESS_NUMA_MASK_BITS);
+		bitmasks_offset++;
+
 		sprintf(tmp,
 		        "Uid:\t%d\t%d\t%d\t%d\n"
 		        "Gid:\t%d\t%d\t%d\t%d\n"
-		        "VmLck:\t%9lu kB\n",
+		        "VmLck:\t%9lu kB\n"
+				"Cpus_allowed:\t%s\n"
+				"Cpus_allowed_list:\t%s\n"
+				"Mems_allowed:\t%s\n"
+				"Mems_allowed_list:\t%s\n",
 		        proc->ruid, proc->euid, proc->suid, proc->fsuid,
 		        proc->rgid, proc->egid, proc->sgid, proc->fsgid,
-		        (lockedsize + 1023) >> 10);
+		        (lockedsize + 1023) >> 10,
+				cpu_bitmask, cpu_list, numa_bitmask, numa_list);
 		len = strlen(tmp);
 		if (r->offset < len) {
 			if (r->offset + r->count < len) {
@@ -430,6 +491,8 @@ process_procfs_request(unsigned long rarg)
 			ans = 0;
 			eof = 1;
 		}
+		kfree(tmp);
+		kfree(bitmasks);
 		goto end;
 	}
 
@@ -570,6 +633,7 @@ dataunavail:
 
 	packet.msg = SCD_MSG_PROCFS_ANSWER;
 	packet.arg = rarg;
+	packet.pid = rpacket->pid;
 
 	ret = ihk_ikc_send(syscall_channel, &packet, 0);
 	if (ret < 0) {

kernel/rusage.c

@@ -0,0 +1,389 @@
+/**
+ * \file rusage.c
+ */
+#include <types.h>
+#include <kmsg.h>
+#include <ihk/cpu.h>
+#include <ihk/mm.h>
+#include <ihk/debug.h>
+#include <process.h>
+#include <init.h>
+#include <march.h>
+#include <cls.h>
+#include <time.h>
+#include <syscall.h>
+#include <string.h>
+#include <rusage.h>
+
+//#define DEBUG_PRINT_AP
+
+#ifdef DEBUG_PRINT_AP
+#define dkprintf(...) kprintf(__VA_ARGS__)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#endif
+
+extern int num_processors ;
+static volatile int ap_stop = 1;
+
+mcs_lock_node_t ap_syscall_semaphore;
+
+extern struct ihk_os_monitor *monitor;
+
+
+#ifdef ENABLE_RUSAGE
+/* count total rss */
+unsigned long count_rss () {
+	int i;
+	unsigned long val = 0;
+	for(i = 0; i < sizeof(cpu_set_t)/8; i++){
+		val += rusage_rss[i];
+	}
+	return val;
+}
+
+/* count total cache */
+unsigned long count_cache () {
+   return 0;
+}
+
+/* count total rss_huge */
+unsigned long count_rss_huge () {
+	return 0;
+}
+
+/* count total mapped_file */
+unsigned long count_mapped_file () {
+	return 0;
+}
+
+/* count total max_usage */
+unsigned long count_max_usage() {
+	return rusage_rss_max;
+}
+
+/* count total kmem_usage */
+unsigned long count_kmem_usage() {
+	return 0;
+}
+
+/* count total kmax_usage */
+unsigned long count_kmem_max_usage() {
+	return 0;
+}
+
+#endif 
+
+#include <sysfs.h>
+#include <vsprintf.h>
+
+#ifdef ENABLE_RUSAGE
+char* strcat_rusage(char *s1, char *s2) {
+	int i;
+	int j;
+	for (i = 0; s1[i] != '\0'; i++); //skip chars.
+	for (j = 0; s2[j] != '\0'; j++) {
+		s1[i+j] = s2[j];
+	}
+	s1[i+j] = '\0';
+	return s1;	
+}
+
+static ssize_t
+show_rusage_memory_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf, size, "rss %lu\ncache %lu\nrss_huge %lu\nmapped_file %lu\n",
+			count_rss(),
+			count_cache(),
+			count_rss_huge(),
+			count_mapped_file()
+	);
+}
+
+static ssize_t 
+show_rusage_memory_max_usage_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf,size,"%lu\n",count_max_usage());
+}
+
+static ssize_t 
+show_rusage_memory_kmem_usage_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf,size,"%lu\n",count_kmem_usage());
+}
+
+static ssize_t 
+show_rusage_memory_kmem_max_usage_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf,size,"%lu\n",count_kmem_max_usage());
+}
+
+static ssize_t 
+show_rusage_num_numa_nodes_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf,size,"%d\n",ihk_mc_get_nr_numa_nodes());
+}
+
+static ssize_t 
+show_rusage_memory_numa_stat_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	int i;
+	int num_numa;
+	char tmp_buf1[1024];
+	char tmp_buf2[1024];
+	unsigned long total = 0;
+	memset(tmp_buf1, 0, 1024);
+	num_numa = ihk_mc_get_nr_numa_nodes();
+
+	for (i = 0; i < num_numa; i++) {
+		total += rusage_numa_stat[i];
+	}
+	sprintf(tmp_buf1, "total=%lu ", total);
+
+	for (i = 0; i < num_numa; i++) {
+		sprintf(tmp_buf2, "N%d=%lu ", i, rusage_numa_stat[i]);
+		strcat_rusage(tmp_buf1, tmp_buf2);
+		memset(tmp_buf2, 0, 1024);
+	}
+	return snprintf(buf, size, "%s\n", tmp_buf1);
+}
+
+static ssize_t 
+show_rusage_hugetlb_usage_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf, size, "%lu\n", rusage_hugetlb_usage);
+}
+
+static ssize_t 
+show_rusage_hugetlb_max_usage_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf, size, "%lu\n", rusage_hugetlb_max_usage);
+}
+static ssize_t 
+show_rusage_cpuacct_stat_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	struct timespec uts;
+	struct timespec sts;
+	int i;
+	int r = 0;
+
+	uts.tv_sec = 0;
+	uts.tv_nsec = 0;
+	sts.tv_sec = 0;
+	sts.tv_nsec = 0;
+	if (monitor)
+		for (i = 0; i < num_processors; i++) {
+			struct timespec ats;
+
+			tsc_to_ts(monitor[i].user_tsc, &ats);
+			ts_add(&uts, &ats);
+			tsc_to_ts(monitor[i].system_tsc, &ats);
+			ts_add(&sts, &ats);
+		}
+	r = snprintf(buf, size, "user %lu\n", timespec_to_jiffy(&uts));
+	r += snprintf(strchr(buf, '\0'), size - r, "system %lu\n",
+				  timespec_to_jiffy(&sts));
+	return r;
+}
+static ssize_t 
+show_rusage_cpuacct_usage_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	struct timespec uts;
+	int i;
+	int r = 0;
+
+	uts.tv_sec = 0;
+	uts.tv_nsec = 0;
+	if (monitor)
+		for (i = 0; i < num_processors; i++) {
+			struct timespec ats;
+
+			tsc_to_ts(monitor[i].user_tsc + monitor[i].system_tsc,
+					  &ats);
+			ts_add(&uts, &ats);
+		}
+	if (uts.tv_sec)
+		r = snprintf(buf, size, "%lu%09lu\n", uts.tv_sec, uts.tv_nsec);
+	else
+		r = snprintf(buf, size, "%lu\n", uts.tv_nsec);
+	return r;
+}
+
+
+static ssize_t 
+show_rusage_cpuacct_usage_percpu_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	struct timespec uts;
+	int i;
+	int r = 0;
+
+	((char *)buf)[0] = '\0';
+	for (i = 0; i < num_processors; i++) {
+		if (monitor) {
+			tsc_to_ts(monitor[i].user_tsc + monitor[i].system_tsc,
+					  &uts);
+		}
+		else {
+			uts.tv_sec = 0;
+			uts.tv_nsec = 0;
+		}
+		if (uts.tv_sec)
+			r += snprintf(strchr(buf, '\0'), size - r,
+						  "%lu%09lu ", uts.tv_sec, uts.tv_nsec);
+		else
+			r += snprintf(strchr(buf, '\0'), size - r,
+						  "%lu ", uts.tv_nsec);
+	}
+	((char *)buf)[r - 1] = '\n';
+	return r;
+}
+
+/* callback funciton of rusage(threads) sysfs */
+static ssize_t
+show_rusage_num_threads_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf, size, "%lu\n", rusage_num_threads);
+}
+
+/* callback funciton of rusage(max threads) sysfs */
+static ssize_t
+show_rusage_max_num_threads_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf, size, "%lu\n", rusage_max_num_threads);
+}
+
+/* definition of sysfs ops */
+struct sysfs_ops show_rusage_memory = {
+	.show = &show_rusage_memory_data,
+};
+struct sysfs_ops show_rusage_memory_max_usage = {
+	.show = &show_rusage_memory_max_usage_data,
+};
+struct sysfs_ops show_rusage_memory_kmem_usage = {
+	.show = &show_rusage_memory_kmem_usage_data,
+};
+struct sysfs_ops show_rusage_memory_kmem_max_usage = {
+	.show = &show_rusage_memory_kmem_max_usage_data,
+};
+struct sysfs_ops show_rusage_num_numa_nodes = {
+	.show = &show_rusage_num_numa_nodes_data,
+};
+struct sysfs_ops show_rusage_memory_numa_stat = {
+	.show = &show_rusage_memory_numa_stat_data,
+};
+struct sysfs_ops show_rusage_hugetlb_usage = {
+	.show = &show_rusage_hugetlb_usage_data,
+};
+struct sysfs_ops show_rusage_hugetlb_max_usage = {
+	.show = &show_rusage_hugetlb_max_usage_data,
+};
+struct sysfs_ops show_rusage_cpuacct_stat = {
+	.show = &show_rusage_cpuacct_stat_data,
+};
+struct sysfs_ops show_rusage_cpuacct_usage = {
+	.show = &show_rusage_cpuacct_usage_data,
+};
+struct sysfs_ops show_rusage_num_threads = {
+	.show = &show_rusage_num_threads_data,
+};
+struct sysfs_ops show_rusage_cpuacct_usage_percpu = {
+	.show = &show_rusage_cpuacct_usage_percpu_data,
+};
+struct sysfs_ops show_rusage_max_num_threads = {
+	.show = &show_rusage_max_num_threads_data,
+};
+
+/* create sysfs files for rusage. */
+void rusage_sysfs_setup(void) {
+	int error;
+	error = sysfs_createf(&show_rusage_memory, &rdata, 0444,
+		"/sys/fs/cgroup/memory/memory.stat");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_memory_max_usage, &rdata, 0444,
+		"/sys/fs/cgroup/memory/memory.max_usage_in_bytes");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_memory_kmem_usage, &rdata, 0444,
+		"/sys/fs/cgroup/memory/memory.kmem.usage_in_bytes");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_memory_kmem_max_usage, &rdata, 0444,
+		"/sys/fs/cgroup/memory/memory.kmem.max_usage_in_bytes");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_num_numa_nodes, &rdata, 0444,
+		"/sys/fs/cgroup/cpu/num_numa_nodes.txt");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_memory_numa_stat, &rdata, 0444,
+		"/sys/fs/cgroup/memory/memory.numa_stat");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_hugetlb_usage, &rdata, 0444,
+		"/sys/fs/cgroup/hugetlb/hugetlb.1GB.usage_in_bytes");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_hugetlb_max_usage, &rdata, 0444,
+		"/sys/fs/cgroup/hugetlb/hugetlb.1GB.max_usage_in_bytes");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_cpuacct_stat, &rdata, 0444,
+		"/sys/fs/cgroup/cpuacct/cpuacct.stat");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_cpuacct_usage, &rdata, 0444,
+		"/sys/fs/cgroup/cpuacct/cpuacct.usage");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_cpuacct_usage_percpu, &rdata, 0444,
+		"/sys/fs/cgroup/cpuacct/cpuacct.usage_percpu");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_num_threads, &rdata, 0444,
+		"/sys/fs/cgroup/num_threads");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+	error = sysfs_createf(&show_rusage_max_num_threads, &rdata, 0444,
+		"/sys/fs/cgroup/max_num_threads");
+	if (error) {
+		panic("rusage_sysfs_setup:sysfs_createf() failed\n");
+	}
+}
+
+/* callback funciton of os_status sysfs */
+static ssize_t
+show_ihk_status_data(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	return snprintf(buf, size, "%d\n", os_status);
+}
+
+struct sysfs_ops show_ihk_status = {
+	.show = &show_ihk_status_data,
+};
+
+/* create sysfs files for monitoring status.*/
+void status_sysfs_setup(void) {
+	int error;
+	error = sysfs_createf(&show_ihk_status, &rdata, 0444,
+		"/sys/fs/cgroup/mck_status");
+	if (error) {
+		panic("status_sysfs_setup:sysfs_createf() failed\n");
+	}
+}
+#endif
+

kernel/shmobj.c

@@ -179,6 +179,7 @@ int shmobj_create(struct shmid_ds *ds, struct memobj **objp)
 
 	memset(obj, 0, sizeof(*obj));
 	obj->memobj.ops = &shmobj_ops;
+	obj->memobj.size = ds->shm_segsz;
 	obj->ds = *ds;
 	obj->ds.shm_perm.seq = the_seq++;
 	obj->ds.shm_nattch = 1;
@@ -240,14 +241,24 @@ void shmobj_destroy(struct shmobj *obj)
 	npages = (size_t)1 << (obj->pgshift - PAGE_SHIFT);
 	for (;;) {
 		struct page *page;
-		int count;
+		void *page_va;
 
 		page = page_list_first(obj);
 		if (!page) {
 			break;
 		}
 		page_list_remove(obj, page);
+		page_va = phys_to_virt(page_to_phys(page));
 
+		if (ihk_atomic_read(&page->count) != 1) {
+			kprintf("%s: WARNING: page count for phys 0x%lx is invalid\n",
+					__FUNCTION__, page->phys);
+		}
+
+		if (page_unmap(page)) {
+			ihk_mc_free_pages(page_va, npages);
+		}
+#if 0
 		dkprintf("shmobj_destroy(%p):"
 				"release page. %p %#lx %d %d",
 				obj, page, page_to_phys(page),
@@ -265,7 +276,8 @@ void shmobj_destroy(struct shmobj *obj)
 		}
 
 		page->mode = PM_NONE;
-		free_pages(phys_to_virt(page_to_phys(page)), npages);
+		ihk_mc_free_pages(phys_to_virt(page_to_phys(page)), npages);
+#endif
 	}
 	if (obj->index < 0) {
 		kfree(obj);
@@ -404,7 +416,7 @@ static int shmobj_get_page(struct memobj *memobj, off_t off, int p2align,
 			goto out;
 		}
 		phys = virt_to_phys(virt);
-		page = phys_to_page(phys);
+		page = phys_to_page_insert_hash(phys);
 		if (page->mode != PM_NONE) {
 			fkprintf("shmobj_get_page(%p,%#lx,%d,%p):"
 					"page %p %#lx %d %d %#lx\n",

kernel/sysfs.c

@@ -75,7 +75,7 @@ sysfs_createf(struct sysfs_ops *ops, void *instance, int mode,
 	dkprintf("sysfs_createf(%p,%p,%#o,%s,...)\n",
 			ops, instance, mode, fmt);
 
-	param = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	param = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
 	if (!param) {
 		error = -ENOMEM;
 		ekprintf("sysfs_createf:allocate_pages failed. %d\n", error);
@@ -134,7 +134,7 @@ sysfs_createf(struct sysfs_ops *ops, void *instance, int mode,
 	error = 0;
 out:
 	if (param) {
-		free_pages(param, 1);
+		ihk_mc_free_pages(param, 1);
 	}
 	if (error) {
 		ekprintf("sysfs_createf(%p,%p,%#o,%s,...): %d\n",
@@ -156,7 +156,7 @@ sysfs_mkdirf(sysfs_handle_t *dirhp, const char *fmt, ...)
 
 	dkprintf("sysfs_mkdirf(%p,%s,...)\n", dirhp, fmt);
 
-	param = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	param = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
 	if (!param) {
 		error = -ENOMEM;
 		ekprintf("sysfs_mkdirf:allocate_pages failed. %d\n", error);
@@ -208,7 +208,7 @@ sysfs_mkdirf(sysfs_handle_t *dirhp, const char *fmt, ...)
 
 out:
 	if (param) {
-		free_pages(param, 1);
+		ihk_mc_free_pages(param, 1);
 	}
 	if (error) {
 		ekprintf("sysfs_mkdirf(%p,%s,...): %d\n", dirhp, fmt, error);
@@ -229,7 +229,7 @@ sysfs_symlinkf(sysfs_handle_t targeth, const char *fmt, ...)
 
 	dkprintf("sysfs_symlinkf(%#lx,%s,...)\n", targeth.handle, fmt);
 
-	param = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	param = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
 	if (!param) {
 		error = -ENOMEM;
 		ekprintf("sysfs_symlinkf:allocate_pages failed. %d\n", error);
@@ -279,7 +279,7 @@ sysfs_symlinkf(sysfs_handle_t targeth, const char *fmt, ...)
 	error = 0;
 out:
 	if (param) {
-		free_pages(param, 1);
+		ihk_mc_free_pages(param, 1);
 	}
 	if (error) {
 		ekprintf("sysfs_symlinkf(%#lx,%s,...): %d\n",
@@ -301,7 +301,7 @@ sysfs_lookupf(sysfs_handle_t *objhp, const char *fmt, ...)
 
 	dkprintf("sysfs_lookupf(%p,%s,...)\n", objhp, fmt);
 
-	param = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	param = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
 	if (!param) {
 		error = -ENOMEM;
 		ekprintf("sysfs_lookupf:allocate_pages failed. %d\n", error);
@@ -353,7 +353,7 @@ sysfs_lookupf(sysfs_handle_t *objhp, const char *fmt, ...)
 
 out:
 	if (param) {
-		free_pages(param, 1);
+		ihk_mc_free_pages(param, 1);
 	}
 	if (error) {
 		ekprintf("sysfs_lookupf(%p,%s,...): %d\n", objhp, fmt, error);
@@ -374,7 +374,7 @@ sysfs_unlinkf(int flags, const char *fmt, ...)
 
 	dkprintf("sysfs_unlinkf(%#x,%s,...)\n", flags, fmt);
 
-	param = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	param = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
 	if (!param) {
 		error = -ENOMEM;
 		ekprintf("sysfs_unlinkf:allocate_pages failed. %d\n", error);
@@ -423,7 +423,7 @@ sysfs_unlinkf(int flags, const char *fmt, ...)
 	error = 0;
 out:
 	if (param) {
-		free_pages(param, 1);
+		ihk_mc_free_pages(param, 1);
 	}
 	if (error) {
 		ekprintf("sysfs_unlinkf(%#x,%s,...): %d\n", flags, fmt, error);
@@ -601,14 +601,14 @@ sysfs_init(void)
 	}
 
 	sysfs_data_bufsize = PAGE_SIZE;
-	sysfs_data_buf = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	sysfs_data_buf = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
 	if (!sysfs_data_buf) {
 		error = -ENOMEM;
 		ekprintf("sysfs_init:allocate_pages(buf) failed. %d\n", error);
 		goto out;
 	}
 
-	param = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	param = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
 	if (!param) {
 		error = -ENOMEM;
 		ekprintf("sysfs_init:allocate_pages(param) failed. %d\n",
@@ -644,7 +644,7 @@ sysfs_init(void)
 	error = 0;
 out:
 	if (param) {
-		free_pages(param, 1);
+		ihk_mc_free_pages(param, 1);
 	}
 	if (error) {
 		ekprintf("sysfs_init(): %d\n", error);

kernel/timer.c

@@ -54,136 +54,75 @@ void init_timers(void)
 }
 
 uint64_t schedule_timeout(uint64_t timeout)
-{	
-	struct waitq_entry my_wait;
-	struct timer my_timer;
+{
 	struct thread *thread = cpu_local_var(current);
-	int irqstate;
-	int spin_sleep;
-
-	irqstate = ihk_mc_spinlock_lock(&thread->spin_sleep_lock);
-	dkprintf("schedule_timeout() spin sleep timeout: %lu\n", timeout);
-	spin_sleep = ++thread->spin_sleep;
-	ihk_mc_spinlock_unlock(&thread->spin_sleep_lock, irqstate);
+	long irqstate;
 
 	/* Spin sleep.. */
 	for (;;) {
+		int need_schedule;
+		struct cpu_local_var *v = get_this_cpu_local_var();
 		uint64_t t_s = rdtsc();
 		uint64_t t_e;
-		int spin_over = 0;
-		
+
 		irqstate = ihk_mc_spinlock_lock(&thread->spin_sleep_lock);
-		
+
 		/* Woken up by someone? */
-		if (thread->spin_sleep < 1) {
+		if (thread->spin_sleep == 0) {
 			t_e = rdtsc();
 
-			spin_over = 1;
 			if ((t_e - t_s) < timeout) {
 				timeout -= (t_e - t_s);
 			}
 			else {
 				timeout = 1;
 			}
+
+			ihk_mc_spinlock_unlock(&thread->spin_sleep_lock, irqstate);
+			break;
 		}
-		
+
 		ihk_mc_spinlock_unlock(&thread->spin_sleep_lock, irqstate);
 
-		if (!spin_over) {
-			t_s = rdtsc();
-			int need_schedule;
-			struct cpu_local_var *v = get_this_cpu_local_var();
-			int irqstate = ihk_mc_spinlock_lock(&(v->runq_lock));
-			need_schedule = v->runq_len > 1 ? 1 : 0;
+		/* Give a chance to another thread (if any) in case the core is
+		 * oversubscribed, but make sure we will be re-scheduled */
+		irqstate = ihk_mc_spinlock_lock(&(v->runq_lock));
+		need_schedule = v->runq_len > 1 ? 1 : 0;
+
+		if (need_schedule) {
+			xchg4(&(cpu_local_var(current)->status), PS_RUNNING);
 			ihk_mc_spinlock_unlock(&(v->runq_lock), irqstate);
+			schedule();
 
-			/* Give a chance to another thread (if any) in case the core is
-			 * oversubscribed, but make sure we will be re-scheduled */
-			if (need_schedule) {
-				xchg4(&(cpu_local_var(current)->status), PS_RUNNING);
-				schedule();
-				xchg4(&(cpu_local_var(current)->status), 
-						PS_INTERRUPTIBLE);
-			}
-			else {
-				/* Spin wait */
-				while ((rdtsc() - t_s) < LOOP_TIMEOUT) {
-					cpu_pause();
-				}
-
-				if (timeout < LOOP_TIMEOUT) {
-					timeout = 0;
-					spin_over = 1;
-				}
-				else {
-					timeout -= LOOP_TIMEOUT;
-				}
-			}
+			/* Recheck if woken */
+			continue;
+		}
+		else {
+			ihk_mc_spinlock_unlock(&(v->runq_lock), irqstate);
 		}
-		
-		if (spin_over) {
-			dkprintf("schedule_timeout() spin woken up, timeout: %lu\n", 
-					timeout);
-			
-			/* Give a chance to another thread (if any) in case we timed out, 
-			 * but make sure we will be re-scheduled */
-			if (timeout == 0) {
-				int need_schedule;
-				struct cpu_local_var *v = get_this_cpu_local_var();
 
-				int irqstate = 
-					ihk_mc_spinlock_lock(&(v->runq_lock));
-				need_schedule = v->runq_len > 1 ? 1 : 0;
-				ihk_mc_spinlock_unlock(&(v->runq_lock), irqstate);
+		/* Spin wait */
+		while ((rdtsc() - t_s) < LOOP_TIMEOUT) {
+			cpu_pause();
+		}
 
-				if (need_schedule) {
-					xchg4(&(cpu_local_var(current)->status), PS_RUNNING);
-					schedule();
-					xchg4(&(cpu_local_var(current)->status), 
-							PS_INTERRUPTIBLE);
-				}
-			}
-			
+		/* Time out? */
+		if (timeout < LOOP_TIMEOUT) {
+			timeout = 0;
+
+			/* We are not sleeping any more */
 			irqstate = ihk_mc_spinlock_lock(&thread->spin_sleep_lock);
-			if (spin_sleep == thread->spin_sleep) {
-				--thread->spin_sleep;
-			}
+			thread->spin_sleep = 0;
 			ihk_mc_spinlock_unlock(&thread->spin_sleep_lock, irqstate);
 
-			return timeout;
+			break;
+		}
+		else {
+			timeout -= LOOP_TIMEOUT;
 		}
 	}
 
-	/* Init waitq and wait entry for this timer */
-	my_timer.timeout = (timeout < LOOP_TIMEOUT) ? LOOP_TIMEOUT : timeout;
-	my_timer.thread = cpu_local_var(current);
-	waitq_init(&my_timer.processes);
-	waitq_init_entry(&my_wait, cpu_local_var(current));
-
-	/* Add ourself to the timer queue */
-	ihk_mc_spinlock_lock_noirq(&timers_lock);
-	list_add_tail(&my_timer.list, &timers);
-
-	dkprintf("schedule_timeout() sleep timeout: %lu\n", my_timer.timeout);
-
-	/* Add ourself to the waitqueue and sleep */ 
-	waitq_prepare_to_wait(&my_timer.processes, &my_wait, PS_INTERRUPTIBLE);
-	ihk_mc_spinlock_unlock_noirq(&timers_lock);
-	schedule();
-	waitq_finish_wait(&my_timer.processes, &my_wait);
-
-	ihk_mc_spinlock_lock_noirq(&timers_lock);
-	
-	/* Waken up by someone else then timeout? */
-	if (my_timer.timeout) {
-		list_del(&my_timer.list);
-	}
-	ihk_mc_spinlock_unlock_noirq(&timers_lock);
-
-	dkprintf("schedule_timeout() woken up, timeout: %lu\n", 
-			my_timer.timeout);
-
-	return my_timer.timeout;
+	return timeout;
 }
 
 

kernel/waitq.c

@@ -22,6 +22,13 @@ default_wake_function(waitq_entry_t *entry, unsigned mode,
 	return sched_wakeup_thread(entry->private, PS_NORMAL);
 }
 
+int
+locked_wake_function(waitq_entry_t *entry, unsigned mode,
+					  int flags, void *key)
+{
+	return sched_wakeup_thread_locked(entry->private, PS_NORMAL);
+}
+
 void
 waitq_init(waitq_t *waitq)
 {

kernel/zeroobj.c

@@ -102,6 +102,7 @@ static int alloc_zeroobj(void)
 
 	memset(obj, 0, sizeof(*obj));
 	obj->memobj.ops = &zeroobj_ops;
+	obj->memobj.size = 0;
 	page_list_init(obj);
 	ihk_mc_spinlock_init(&obj->memobj.lock);
 
@@ -112,7 +113,7 @@ static int alloc_zeroobj(void)
 		goto out;
 	}
 	phys = virt_to_phys(virt);
-	page = phys_to_page(phys);
+	page = phys_to_page_insert_hash(phys);
 
 	if (page->mode != PM_NONE) {
 		fkprintf("alloc_zeroobj():"
@@ -172,6 +173,10 @@ static int zeroobj_get_page(struct memobj *memobj, off_t off, int p2align,
 	struct zeroobj *obj = to_zeroobj(memobj);
 	struct page *page;
 
+	/* Don't bother about zero page, page fault handler will
+	 * allocate and clear pages */
+	return 0;
+
 	dkprintf("zeroobj_get_page(%p,%#lx,%d,%p)\n",
 			memobj, off, p2align, physp);
 	if (off & ~PAGE_MASK) {

lib/abort.c

@@ -1,8 +1,17 @@
 #include <ihk/debug.h>
 #include <ihk/cpu.h>
+#include <cls.h>
+#include <ihk/rusage.h>
+
+extern struct cpu_local_var *clv;
 
 void panic(const char *msg)
 {
+	if (clv) {
+		struct ihk_os_monitor *monitor = cpu_local_var(monitor);
+
+		monitor->status = IHK_OS_MONITOR_PANIC;
+	}
 	cpu_disable_interrupt();
 
 	kprintf(msg);

lib/bitops.c

@@ -140,3 +140,58 @@ found:
 	return result + ffz(tmp);
 }
 
+/**
+ * hweightN - returns the hamming weight of a N-bit word
+ * @x: the word to weigh
+ *
+ * The Hamming Weight of a number is the total number of bits set in it.
+ */
+
+unsigned int __sw_hweight32(unsigned int w)
+{
+#ifdef ARCH_HAS_FAST_MULTIPLIER
+	w -= (w >> 1) & 0x55555555;
+	w =  (w & 0x33333333) + ((w >> 2) & 0x33333333);
+	w =  (w + (w >> 4)) & 0x0f0f0f0f;
+	return (w * 0x01010101) >> 24;
+#else
+	unsigned int res = w - ((w >> 1) & 0x55555555);
+	res = (res & 0x33333333) + ((res >> 2) & 0x33333333);
+	res = (res + (res >> 4)) & 0x0F0F0F0F;
+	res = res + (res >> 8);
+	return (res + (res >> 16)) & 0x000000FF;
+#endif
+}
+
+unsigned int __sw_hweight16(unsigned int w)
+{
+	unsigned int res = w - ((w >> 1) & 0x5555);
+	res = (res & 0x3333) + ((res >> 2) & 0x3333);
+	res = (res + (res >> 4)) & 0x0F0F;
+	return (res + (res >> 8)) & 0x00FF;
+}
+
+unsigned int __sw_hweight8(unsigned int w)
+{
+	unsigned int res = w - ((w >> 1) & 0x55);
+	res = (res & 0x33) + ((res >> 2) & 0x33);
+	return (res + (res >> 4)) & 0x0F;
+}
+
+unsigned long __sw_hweight64(uint64_t w)
+{
+#ifdef ARCH_HAS_FAST_MULTIPLIER
+	w -= (w >> 1) & 0x5555555555555555ul;
+	w =  (w & 0x3333333333333333ul) + ((w >> 2) & 0x3333333333333333ul);
+	w =  (w + (w >> 4)) & 0x0f0f0f0f0f0f0f0ful;
+	return (w * 0x0101010101010101ul) >> 56;
+#else
+	uint64_t res = w - ((w >> 1) & 0x5555555555555555ul);
+	res = (res & 0x3333333333333333ul) + ((res >> 2) & 0x3333333333333333ul);
+	res = (res + (res >> 4)) & 0x0F0F0F0F0F0F0F0Ful;
+	res = res + (res >> 8);
+	res = res + (res >> 16);
+	return (res + (res >> 32)) & 0x00000000000000FFul;
+#endif
+}
+

lib/include/bitmap.h

@@ -0,0 +1,307 @@
+#ifndef __LINUX_BITMAP_H
+#define __LINUX_BITMAP_H
+
+#include <types.h>
+#include <bitops.h>
+#include <string.h>
+
+/*
+ * bitmaps provide bit arrays that consume one or more unsigned
+ * longs.  The bitmap interface and available operations are listed
+ * here, in bitmap.h
+ *
+ * Function implementations generic to all architectures are in
+ * lib/bitmap.c.  Functions implementations that are architecture
+ * specific are in various include/asm-<arch>/bitops.h headers
+ * and other arch/<arch> specific files.
+ *
+ * See lib/bitmap.c for more details.
+ */
+
+/*
+ * The available bitmap operations and their rough meaning in the
+ * case that the bitmap is a single unsigned long are thus:
+ *
+ * Note that nbits should be always a compile time evaluable constant.
+ * Otherwise many inlines will generate horrible code.
+ *
+ * bitmap_zero(dst, nbits)			*dst = 0UL
+ * bitmap_fill(dst, nbits)			*dst = ~0UL
+ * bitmap_copy(dst, src, nbits)			*dst = *src
+ * bitmap_and(dst, src1, src2, nbits)		*dst = *src1 & *src2
+ * bitmap_or(dst, src1, src2, nbits)		*dst = *src1 | *src2
+ * bitmap_xor(dst, src1, src2, nbits)		*dst = *src1 ^ *src2
+ * bitmap_andnot(dst, src1, src2, nbits)	*dst = *src1 & ~(*src2)
+ * bitmap_complement(dst, src, nbits)		*dst = ~(*src)
+ * bitmap_equal(src1, src2, nbits)		Are *src1 and *src2 equal?
+ * bitmap_intersects(src1, src2, nbits) 	Do *src1 and *src2 overlap?
+ * bitmap_subset(src1, src2, nbits)		Is *src1 a subset of *src2?
+ * bitmap_empty(src, nbits)			Are all bits zero in *src?
+ * bitmap_full(src, nbits)			Are all bits set in *src?
+ * bitmap_weight(src, nbits)			Hamming Weight: number set bits
+ * bitmap_set(dst, pos, nbits)			Set specified bit area
+ * bitmap_clear(dst, pos, nbits)		Clear specified bit area
+ * bitmap_find_next_zero_area(buf, len, pos, n, mask)	Find bit free area
+ * bitmap_shift_right(dst, src, n, nbits)	*dst = *src >> n
+ * bitmap_shift_left(dst, src, n, nbits)	*dst = *src << n
+ * bitmap_remap(dst, src, old, new, nbits)	*dst = map(old, new)(src)
+ * bitmap_bitremap(oldbit, old, new, nbits)	newbit = map(old, new)(oldbit)
+ * bitmap_onto(dst, orig, relmap, nbits)	*dst = orig relative to relmap
+ * bitmap_fold(dst, orig, sz, nbits)		dst bits = orig bits mod sz
+ * bitmap_scnprintf(buf, len, src, nbits)	Print bitmap src to buf
+ * bitmap_parse(buf, buflen, dst, nbits)	Parse bitmap dst from kernel buf
+ * bitmap_parse_user(ubuf, ulen, dst, nbits)	Parse bitmap dst from user buf
+ * bitmap_scnlistprintf(buf, len, src, nbits)	Print bitmap src as list to buf
+ * bitmap_parselist(buf, dst, nbits)		Parse bitmap dst from kernel buf
+ * bitmap_parselist_user(buf, dst, nbits)	Parse bitmap dst from user buf
+ * bitmap_find_free_region(bitmap, bits, order)	Find and allocate bit region
+ * bitmap_release_region(bitmap, pos, order)	Free specified bit region
+ * bitmap_allocate_region(bitmap, pos, order)	Allocate specified bit region
+ */
+
+/*
+ * Also the following operations in asm/bitops.h apply to bitmaps.
+ *
+ * set_bit(bit, addr)			*addr |= bit
+ * clear_bit(bit, addr)			*addr &= ~bit
+ * change_bit(bit, addr)		*addr ^= bit
+ * test_bit(bit, addr)			Is bit set in *addr?
+ * test_and_set_bit(bit, addr)		Set bit and return old value
+ * test_and_clear_bit(bit, addr)	Clear bit and return old value
+ * test_and_change_bit(bit, addr)	Change bit and return old value
+ * find_first_zero_bit(addr, nbits)	Position first zero bit in *addr
+ * find_first_bit(addr, nbits)		Position first set bit in *addr
+ * find_next_zero_bit(addr, nbits, bit)	Position next zero bit in *addr >= bit
+ * find_next_bit(addr, nbits, bit)	Position next set bit in *addr >= bit
+ */
+
+/*
+ * The DECLARE_BITMAP(name,bits) macro, in linux/types.h, can be used
+ * to declare an array named 'name' of just enough unsigned longs to
+ * contain all bit positions from 0 to 'bits' - 1.
+ */
+
+/*
+ * lib/bitmap.c provides these functions:
+ */
+
+#define __user 
+#define __force
+#define u32 uint32_t
+
+extern int __bitmap_empty(const unsigned long *bitmap, int bits);
+extern int __bitmap_full(const unsigned long *bitmap, int bits);
+extern int __bitmap_equal(const unsigned long *bitmap1,
+                	const unsigned long *bitmap2, int bits);
+extern void __bitmap_complement(unsigned long *dst, const unsigned long *src,
+			int bits);
+extern void __bitmap_shift_right(unsigned long *dst,
+                        const unsigned long *src, int shift, int bits);
+extern void __bitmap_shift_left(unsigned long *dst,
+                        const unsigned long *src, int shift, int bits);
+extern int __bitmap_and(unsigned long *dst, const unsigned long *bitmap1,
+			const unsigned long *bitmap2, int bits);
+extern void __bitmap_or(unsigned long *dst, const unsigned long *bitmap1,
+			const unsigned long *bitmap2, int bits);
+extern void __bitmap_xor(unsigned long *dst, const unsigned long *bitmap1,
+			const unsigned long *bitmap2, int bits);
+extern int __bitmap_andnot(unsigned long *dst, const unsigned long *bitmap1,
+			const unsigned long *bitmap2, int bits);
+extern int __bitmap_intersects(const unsigned long *bitmap1,
+			const unsigned long *bitmap2, int bits);
+extern int __bitmap_subset(const unsigned long *bitmap1,
+			const unsigned long *bitmap2, int bits);
+extern int __bitmap_weight(const unsigned long *bitmap, int bits);
+
+extern void bitmap_set(unsigned long *map, int i, int len);
+extern void bitmap_clear(unsigned long *map, int start, int nr);
+extern unsigned long bitmap_find_next_zero_area(unsigned long *map,
+					 unsigned long size,
+					 unsigned long start,
+					 unsigned int nr,
+					 unsigned long align_mask);
+
+extern int bitmap_scnprintf(char *buf, unsigned int len,
+			const unsigned long *src, int nbits);
+extern int __bitmap_parse(const char *buf, unsigned int buflen, int is_user,
+			unsigned long *dst, int nbits);
+extern int bitmap_parse_user(const char __user *ubuf, unsigned int ulen,
+			unsigned long *dst, int nbits);
+extern int bitmap_scnlistprintf(char *buf, unsigned int len,
+			const unsigned long *src, int nbits);
+extern int bitmap_parselist(const char *buf, unsigned long *maskp,
+			int nmaskbits);
+extern int bitmap_parselist_user(const char __user *ubuf, unsigned int ulen,
+			unsigned long *dst, int nbits);
+extern void bitmap_remap(unsigned long *dst, const unsigned long *src,
+		const unsigned long *old, const unsigned long *new, int bits);
+extern int bitmap_bitremap(int oldbit,
+		const unsigned long *old, const unsigned long *new, int bits);
+extern void bitmap_onto(unsigned long *dst, const unsigned long *orig,
+		const unsigned long *relmap, int bits);
+extern void bitmap_fold(unsigned long *dst, const unsigned long *orig,
+		int sz, int bits);
+extern int bitmap_find_free_region(unsigned long *bitmap, int bits, int order);
+extern void bitmap_release_region(unsigned long *bitmap, int pos, int order);
+extern int bitmap_allocate_region(unsigned long *bitmap, int pos, int order);
+extern int bitmap_ord_to_pos(const unsigned long *bitmap, int n, int bits);
+
+#define BITMAP_FIRST_WORD_MASK(start) (~0UL << ((start) % BITS_PER_LONG))
+#define BITMAP_LAST_WORD_MASK(nbits)					\
+(									\
+	((nbits) % BITS_PER_LONG) ?					\
+		(1UL<<((nbits) % BITS_PER_LONG))-1 : ~0UL		\
+)
+
+#define small_const_nbits(nbits) \
+	(__builtin_constant_p(nbits) && (nbits) <= BITS_PER_LONG)
+
+static inline void bitmap_zero(unsigned long *dst, int nbits)
+{
+	if (small_const_nbits(nbits))
+		*dst = 0UL;
+	else {
+		int len = BITS_TO_LONGS(nbits) * sizeof(unsigned long);
+		memset(dst, 0, len);
+	}
+}
+
+static inline void bitmap_fill(unsigned long *dst, int nbits)
+{
+	size_t nlongs = BITS_TO_LONGS(nbits);
+	if (!small_const_nbits(nbits)) {
+		int len = (nlongs - 1) * sizeof(unsigned long);
+		memset(dst, 0xff,  len);
+	}
+	dst[nlongs - 1] = BITMAP_LAST_WORD_MASK(nbits);
+}
+
+static inline void bitmap_copy(unsigned long *dst, const unsigned long *src,
+			int nbits)
+{
+	if (small_const_nbits(nbits))
+		*dst = *src;
+	else {
+		int len = BITS_TO_LONGS(nbits) * sizeof(unsigned long);
+		memcpy(dst, src, len);
+	}
+}
+
+static inline int bitmap_and(unsigned long *dst, const unsigned long *src1,
+			const unsigned long *src2, int nbits)
+{
+	if (small_const_nbits(nbits))
+		return (*dst = *src1 & *src2) != 0;
+	return __bitmap_and(dst, src1, src2, nbits);
+}
+
+static inline void bitmap_or(unsigned long *dst, const unsigned long *src1,
+			const unsigned long *src2, int nbits)
+{
+	if (small_const_nbits(nbits))
+		*dst = *src1 | *src2;
+	else
+		__bitmap_or(dst, src1, src2, nbits);
+}
+
+static inline void bitmap_xor(unsigned long *dst, const unsigned long *src1,
+			const unsigned long *src2, int nbits)
+{
+	if (small_const_nbits(nbits))
+		*dst = *src1 ^ *src2;
+	else
+		__bitmap_xor(dst, src1, src2, nbits);
+}
+
+static inline int bitmap_andnot(unsigned long *dst, const unsigned long *src1,
+			const unsigned long *src2, int nbits)
+{
+	if (small_const_nbits(nbits))
+		return (*dst = *src1 & ~(*src2)) != 0;
+	return __bitmap_andnot(dst, src1, src2, nbits);
+}
+
+static inline void bitmap_complement(unsigned long *dst, const unsigned long *src,
+			int nbits)
+{
+	if (small_const_nbits(nbits))
+		*dst = ~(*src) & BITMAP_LAST_WORD_MASK(nbits);
+	else
+		__bitmap_complement(dst, src, nbits);
+}
+
+static inline int bitmap_equal(const unsigned long *src1,
+			const unsigned long *src2, int nbits)
+{
+	if (small_const_nbits(nbits))
+		return ! ((*src1 ^ *src2) & BITMAP_LAST_WORD_MASK(nbits));
+	else
+		return __bitmap_equal(src1, src2, nbits);
+}
+
+static inline int bitmap_intersects(const unsigned long *src1,
+			const unsigned long *src2, int nbits)
+{
+	if (small_const_nbits(nbits))
+		return ((*src1 & *src2) & BITMAP_LAST_WORD_MASK(nbits)) != 0;
+	else
+		return __bitmap_intersects(src1, src2, nbits);
+}
+
+static inline int bitmap_subset(const unsigned long *src1,
+			const unsigned long *src2, int nbits)
+{
+	if (small_const_nbits(nbits))
+		return ! ((*src1 & ~(*src2)) & BITMAP_LAST_WORD_MASK(nbits));
+	else
+		return __bitmap_subset(src1, src2, nbits);
+}
+
+static inline int bitmap_empty(const unsigned long *src, int nbits)
+{
+	if (small_const_nbits(nbits))
+		return ! (*src & BITMAP_LAST_WORD_MASK(nbits));
+	else
+		return __bitmap_empty(src, nbits);
+}
+
+static inline int bitmap_full(const unsigned long *src, int nbits)
+{
+	if (small_const_nbits(nbits))
+		return ! (~(*src) & BITMAP_LAST_WORD_MASK(nbits));
+	else
+		return __bitmap_full(src, nbits);
+}
+
+static inline int bitmap_weight(const unsigned long *src, int nbits)
+{
+	if (small_const_nbits(nbits))
+		return hweight_long(*src & BITMAP_LAST_WORD_MASK(nbits));
+	return __bitmap_weight(src, nbits);
+}
+
+static inline void bitmap_shift_right(unsigned long *dst,
+			const unsigned long *src, int n, int nbits)
+{
+	if (small_const_nbits(nbits))
+		*dst = *src >> n;
+	else
+		__bitmap_shift_right(dst, src, n, nbits);
+}
+
+static inline void bitmap_shift_left(unsigned long *dst,
+			const unsigned long *src, int n, int nbits)
+{
+	if (small_const_nbits(nbits))
+		*dst = (*src << n) & BITMAP_LAST_WORD_MASK(nbits);
+	else
+		__bitmap_shift_left(dst, src, n, nbits);
+}
+
+static inline int bitmap_parse(const char *buf, unsigned int buflen,
+			unsigned long *maskp, int nmaskbits)
+{
+	return __bitmap_parse(buf, buflen, 0, maskp, nmaskbits);
+}
+
+#endif /* __LINUX_BITMAP_H */

lib/include/bitops.h

@@ -27,6 +27,31 @@ unsigned long find_first_bit(const unsigned long *addr,
 unsigned long find_first_zero_bit(const unsigned long *addr, 
 				  unsigned long size);
 
+static inline int test_bit(int nr, const void *addr)
+{
+	const uint32_t *p = (const uint32_t *)addr;
+	return ((1UL << (nr & 31)) & (p[nr >> 5])) != 0;
+}
+
+extern unsigned int __sw_hweight32(unsigned int w);
+extern unsigned int __sw_hweight16(unsigned int w);
+extern unsigned int __sw_hweight8(unsigned int w);
+extern unsigned long __sw_hweight64(uint64_t w);
+
+static inline unsigned long hweight_long(unsigned long w)
+{
+	return sizeof(w) == 4 ? __sw_hweight32(w) : __sw_hweight64(w);
+}
+
+#define BIT(nr)			(1UL << (nr))
+#define BIT_WORD(nr)		((nr) / BITS_PER_LONG)
+#define BITS_PER_BYTE		8
+
+#define __ALIGN_KERNEL_MASK(x, mask)	(((x) + (mask)) & ~(mask))
+#define __ALIGN_MASK(x, mask)	__ALIGN_KERNEL_MASK((x), (mask))
+#define __ALIGN_KERNEL(x, a)		__ALIGN_KERNEL_MASK(x, (typeof(x))(a) - 1)
+#define ALIGN(x, a)		__ALIGN_KERNEL((x), (a))
+
 #endif /*__ASSEMBLY__*/
 
 #include <arch-bitops.h>

lib/include/ihk/cpu.h

@@ -49,12 +49,17 @@ struct ihk_mc_cpu_info {
 	int ncpus;
 	int *hw_ids;
 	int *nodes;
+	int *linux_cpu_ids;
 };
 
 struct ihk_mc_cpu_info *ihk_mc_get_cpu_info(void);
 void ihk_mc_boot_cpu(int cpuid, unsigned long pc);
 int ihk_mc_get_processor_id(void);
 int ihk_mc_get_hardware_processor_id(void);
+int ihk_mc_get_numa_id(void);
+int ihk_mc_get_nr_cores();
+int ihk_mc_get_core(int id, unsigned long *linux_core_id, unsigned long *apic_id,
+                    int *numa_id);
 
 void ihk_mc_delay_us(int us);
 void ihk_mc_set_syscall_handler(long (*handler)(int, ihk_mc_user_context_t *));