mcoverlayfs: supported Linux kernel 4.0 or rhel kernel 3.10.0-327

add mcoverlayfs(linux-3.10.0-327.36.1.el7 base)

Makefile.in

@@ -1,5 +1,6 @@
 TARGET = @TARGET@
 SBINDIR = @SBINDIR@
+ETCDIR = @ETCDIR@
 MANDIR = @MANDIR@
 
 all::
@@ -48,6 +49,9 @@ 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; \
+		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

@@ -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"
@@ -883,13 +916,36 @@ 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) {
@@ -998,9 +1054,8 @@ 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"),
@@ -1012,14 +1067,14 @@ 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);
@@ -1494,7 +1549,8 @@ 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 +1564,8 @@ 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 +1574,15 @@ 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 +1604,9 @@ 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));

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

@@ -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/syscall_list.h

@@ -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)

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__)
@@ -84,20 +85,22 @@ void ihk_mc_free_pages(void *p, int npages)
 		pa_ops->free_page(p, npages);
 }
 
-void *ihk_mc_allocate(int size, enum ihk_mc_ap_flag flag)
+void *ihk_mc_allocate(int size, int flag)
 {
-	if (pa_ops && pa_ops->alloc)
-		return pa_ops->alloc(size, flag);
-	else
-		return ihk_mc_alloc_pages(1, flag);
+	if (!cpu_local_var(kmalloc_initialized)) {
+		kprintf("%s: error, kmalloc not yet initialized\n", __FUNCTION__);
+		return NULL;
+	}
+	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)
@@ -910,11 +913,17 @@ 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;
@@ -1105,6 +1114,7 @@ static int clear_range_l1(void *args0, pte_t *ptep, uint64_t base,
 	if (!(old & PFL1_FILEOFF) && args->free_physical) {
 		if (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;
 	}
@@ -1153,6 +1163,7 @@ 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)) {
 				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;
 		}
@@ -2261,13 +2272,18 @@ 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 +2299,22 @@ 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;
@@ -2413,8 +2440,18 @@ 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;

arch/x86/kernel/mikc.c

@@ -38,7 +38,7 @@ int ihk_mc_ikc_init_first_local(struct ihk_ikc_channel_desc *channel,
 	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),
@@ -219,6 +222,7 @@ 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 +269,31 @@ 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 | \
@@ -707,6 +730,8 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		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 +744,8 @@ 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);
 
@@ -751,6 +777,33 @@ 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)){
@@ -761,9 +814,6 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 			return;
 		}
 
-
-
-
 		usp = (unsigned long *)sigsp;
 		usp--;
 		*usp = (unsigned long)k->sa.sa_restorer;
@@ -1240,7 +1290,7 @@ done:
 	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;
 
@@ -1251,12 +1301,12 @@ 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 */
@@ -1387,9 +1437,8 @@ 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 +1562,7 @@ 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;
 

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=@ETCDIR@/irqbalance_mck
+ExecStart=/usr/sbin/irqbalance --foreground $IRQBALANCE_ARGS
+
+[Install]
+WantedBy=multi-user.target

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

@@ -13,27 +13,44 @@
 # 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=""
+
 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
+
+while getopts :i: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 [ ${INTERVAL} -le 0 ]
 		then
-			echo "invalid -i value"
+			echo "invalid -i value" >&2
 			exit 1
 		fi
 		;;
@@ -41,22 +58,24 @@ do
 		expr "${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}
+		;;
+	*)  echo "invalid option -${OPT}" >&2
 		exit 1
 	esac
 done
 
-mem="512M@0"
-cpus=""
 ihk_ikc_irq_core=0
 
 release=`uname -r`
@@ -66,10 +85,17 @@ 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
+enable_mcoverlay="no"
 if [ "${ENABLE_MCOVERLAYFS}" == "yes" ]; then
-	enable_mcoverlay=`if ( [ ${linux_version_code} -ge 262144 ] && [ ${linux_version_code} -lt 262400 ] ); then echo "yes"; else echo "no"; fi`
-else
-	enable_mcoverlay=no
+	if [ "${rhel_release}" == "" ]; then
+		if [ ${linux_version_code} -ge 262144 -a ${linux_version_code} -lt 262400 ]; then
+			enable_mcoverlay="yes"
+		fi
+	else
+		if [ ${linux_version_code} -eq 199168 -a ${rhel_release} -ge 327 ]; then
+			enable_mcoverlay="yes"
+		fi
+	fi
 fi
 
 if [ "$cpus" == "" ]; then
@@ -79,12 +105,7 @@ if [ "$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
-fi
-
-# Remove delegator if loaded
-if [ "`lsmod | grep mcctrl`" != "" ]; then
-        if ! rmmod mcctrl; then echo "error: removing mcctrl"; exit; fi
+	if [ "$cpus" == "" ]; then echo "error: no available CPUs on NUMA node 0?" >&2; exit 1; fi
 fi
 
 # Remove mcoverlay if loaded
@@ -95,13 +116,19 @@ if [ "$enable_mcoverlay" == "yes" ]; then
 		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; then echo "error: removing mcoverlay" >&2; exit 1; fi
 	fi
 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; exit 1; fi;
+fi
+
 # Load IHK if not loaded
 if [ "`lsmod | grep ihk`" == "" ]; then
-        if ! insmod ${KMODDIR}/ihk.ko; then echo "error: loading ihk"; exit; fi;
+        if ! insmod ${KMODDIR}/ihk.ko; then echo "error: loading ihk" >&2; exit 1; fi;
 fi
 
 # Load IHK-SMP if not loaded and reserve CPUs and memory
@@ -113,57 +140,61 @@ if [ "`lsmod | grep ihk_smp_x86`" == "" ]; then
                         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 ${mem}; then echo "error: reserving memory"; exit; fi
+        if [ "$ihk_irq" == "" ]; then echo "error: no IRQ available" >&2; exit 1; 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" >&2; exit 1; fi;
+        if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then echo "error: reserving CPUs" >&2; exit 1; fi
+        if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then echo "error: reserving memory" >&2; exit 1; 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
+        if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus" >&2; exit 1; fi
         cpus_allocated=`${SBINDIR}/ihkosctl 0 query cpu`
 		if [ "$cpus_allocated" == "" ]; then
-			if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then echo "error: reserving CPUs"; exit; fi
+			if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then echo "error: reserving CPUs" >&2; exit 1; fi
 		fi
 
-        if ! ${SBINDIR}/ihkosctl 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
+        if ! ${SBINDIR}/ihkosctl 0 query mem > /dev/null; then echo "error: querying memory" >&2; exit 1; fi
         mem_allocated=`${SBINDIR}/ihkosctl 0 query mem`
 		if [ "$mem_allocated" == "" ]; then
-			if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then echo "error: reserving memory"; exit; fi
+			if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then echo "error: reserving memory" >&2; exit 1; fi
 		fi
 fi
 
+# Load mcctrl if not loaded
+if [ "`lsmod | grep mcctrl`" == "" ]; then
+	if ! insmod ${KMODDIR}/mcctrl.ko; then echo "error: inserting mcctrl.ko" >&2; exit 1; 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
+        if ! ${SBINDIR}/ihkosctl 0 query cpu > /dev/null; then echo "error: querying cpus" >&2; exit 1; fi
         cpus=`${SBINDIR}/ihkosctl 0 query cpu`
-        if ! ${SBINDIR}/ihkosctl 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
+        if ! ${SBINDIR}/ihkosctl 0 query mem > /dev/null; then echo "error: querying memory" >&2; exit 1; fi
         mem=`${SBINDIR}/ihkosctl 0 query mem`
 
-        if ! ${SBINDIR}/ihkconfig 0 destroy 0; then echo "warning: destroy failed"; fi	
+        if ! ${SBINDIR}/ihkconfig 0 destroy 0; then echo "warning: destroy failed" >&2; fi	
 else
         # Otherwise query IHK-SMP for resources
-        if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus"; exit; fi
+        if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus" >&2; exit 1; fi
         cpus=`${SBINDIR}/ihkconfig 0 query cpu`
-        if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
+        if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then echo "error: querying memory" >&2; exit 1; fi
         mem=`${SBINDIR}/ihkconfig 0 query mem`	
 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
+if ! ${SBINDIR}/ihkconfig 0 create; then echo "error: create" >&2; exit; fi
+if ! ${SBINDIR}/ihkosctl 0 assign cpu ${cpus}; then echo "error: assign CPUs" >&2; exit 1; fi
+if ! ${SBINDIR}/ihkosctl 0 assign mem ${mem}; then echo "error: assign memory" >&2; exit 1; fi
+if ! ${SBINDIR}/ihkosctl 0 load ${KERNDIR}/mckernel.img; then echo "error: loading kernel image" >&2; exit 1; fi
+if ! ${SBINDIR}/ihkosctl 0 kargs "hidos ksyslogd=${LOGMODE}"; then echo "error: setting kernel arguments" >&2; exit 1; fi
+if ! ${SBINDIR}/ihkosctl 0 boot; then echo "error: booting" >&2; exit 1; fi
+if ! chown ${chown_option} /dev/mcd* /dev/mcos*; then echo "error: chowning device files" >&2; exit 1; fi
 
 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 ! mount -t tmpfs tmpfs /tmp/mcos; then echo "error: mount /tmp/mcos" >&2; exit 1; 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 ! mount --bind /proc /tmp/mcos/linux_proc; then echo "error: mount /tmp/mcos/linux_proc" >&2; exit 1; fi
+	if ! insmod ${KMODDIR}/mcoverlay.ko; then echo "error: inserting mcoverlay.ko" >&2; exit 1; fi
 	while [ ! -e /proc/mcos0 ]
 	do
 		sleep 1
@@ -171,7 +202,7 @@ if [ "$enable_mcoverlay" == "yes" ]; then
 	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 ! 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" >&2; exit 1; fi
 	mount --make-rprivate /proc
 	while [ ! -e /sys/devices/virtual/mcos/mcos0/sys ]
 	do
@@ -180,7 +211,7 @@ if [ "$enable_mcoverlay" == "yes" ]; then
 	if [ ! -e /tmp/mcos/mcos0_sys ]; then mkdir -p /tmp/mcos/mcos0_sys; fi
 	if [ ! -e /tmp/mcos/mcos0_sys_upper ]; then mkdir -p /tmp/mcos/mcos0_sys_upper; fi
 	if [ ! -e /tmp/mcos/mcos0_sys_work ]; then mkdir -p /tmp/mcos/mcos0_sys_work; fi
-	if ! mount -t mcoverlay mcoverlay -o lowerdir=/sys/devices/virtual/mcos/mcos0/sys:/sys,upperdir=/tmp/mcos/mcos0_sys_upper,workdir=/tmp/mcos/mcos0_sys_work,nocopyupw,nofscheck /tmp/mcos/mcos0_sys; then echo "error: mount /tmp/mcos/mcos0_sys"; exit; fi
+	if ! mount -t mcoverlay mcoverlay -o lowerdir=/sys/devices/virtual/mcos/mcos0/sys:/sys,upperdir=/tmp/mcos/mcos0_sys_upper,workdir=/tmp/mcos/mcos0_sys_work,nocopyupw,nofscheck /tmp/mcos/mcos0_sys; then echo "error: mount /tmp/mcos/mcos0_sys" >&2; exit 1; fi
 	mount --make-rprivate /sys
 	for cpuid in `find /sys/devices/system/cpu/* -maxdepth 0 -name "cpu[0123456789]*" -printf "%f "`; do
 		if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/cpu/$cpuid" ]; then
@@ -195,5 +226,25 @@ if [ "$enable_mcoverlay" == "yes" ]; then
 fi
 if [ ${LOGMODE} -ne 0 ]
 then
-	SBINDIR=${SBINDIR} ${SBINDIR}/mcklogd -i ${INTERVAL}
+# mcklogd survives when McKernel isn't shut down by mcstop+release.sh
+	pkill mcklogd    
+	SBINDIR=${SBINDIR} ${SBINDIR}/mcklogd -i ${INTERVAL} -f ${facility}
 fi
+
+# Start irqbalance with CPUs and IRQ for McKernel banned
+if [ "${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; exit 1; 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 $cmd;}}'; then echo "error: modifying /proc/irq/*/smp_affinity" >&2; exit 1; fi;
+
+    banirq=`cat /proc/interrupts| perl -e 'while(<>) { if(/^\s*(\d+).*IHK\-SMP\s*$/) {print $1;}}'`
+
+    sed "s/%mask%/$smp_affinity_mask/g" $ETCDIR/irqbalance_mck.in | sed "s/%banirq%/$banirq/g" > $ETCDIR/irqbalance_mck
+    if ! systemctl link $ETCDIR/irqbalance_mck.service >/dev/null 2>/dev/null; then echo "error: linking irqbalance_mck" >&2; exit 1; fi;
+    if ! systemctl start irqbalance_mck.service 2>/dev/null ; then echo "error: starting irqbalance_mck" >&2; exit 1; fi;
+#    echo cpus=$cpus mask=$smp_affinity_mask banirq=$banirq
+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 +18,47 @@ mem=""
 cpus=""
 
 # No SMP module? Exit.
-if [ "`lsmod | grep ihk_smp_x86`" == "" ]; then exit; fi
+if [ "`lsmod | grep ihk_smp_x86`" == "" ]; then exit 0; 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 ! ${SBINDIR}/ihkconfig 0 release cpu $cpus > /dev/null; then echo "error: releasing CPUs" >&2; exit 1; 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 ! ${SBINDIR}/ihkconfig 0 release mem $mem > /dev/null; then echo "error: releasing memory" >&2; exit 1; fi
 
 # Remove delegator if loaded
 if [ "`lsmod | grep mcctrl`" != "" ]; then
-        if ! rmmod mcctrl; then echo "error: removing mcctrl"; exit; fi
+        if ! rmmod mcctrl; then echo "error: removing mcctrl" >&2; exit 1; 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	
-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
+        if ! rmmod ihk_smp_x86; then echo "error: removing ihk_smp_x86" >&2; exit 1; fi
 fi
 
+# Remove core module
+if [ "`lsmod | grep -E 'ihk\s' | awk '{print $1}'`" != "" ]; then
+        if ! rmmod ihk; then echo "error: removing ihk" >&2; exit 1; fi
+fi
 
+# Stop mcklogd
+pkill mcklogd
+
+# 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 "error: stopping irqbalance_mck" >&2; exit 1; fi;
+    if ! systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null; then echo "error: disabling irqbalance_mck" >&2; exit 1; 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 "error: restoring /proc/irq/*/smp_affinity" >&2; exit 1; fi;
+    if ! systemctl start irqbalance.service; then echo "error: starting irqbalance" >&2; exit 1; fi;
+fi

configure

@@ -632,6 +632,7 @@ ENABLE_MCOVERLAYFS
 MANDIR
 KERNDIR
 KMODDIR
+ETCDIR
 SBINDIR
 BINDIR
 TARGET
@@ -3031,6 +3032,9 @@ 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
@@ -3882,11 +3886,12 @@ fi
 
 
 
+
 
 
 ac_config_headers="$ac_config_headers executer/config.h"
 
-ac_config_files="$ac_config_files Makefile executer/user/Makefile executer/kernel/mcctrl/Makefile executer/kernel/mcctrl/arch/x86_64/Makefile executer/kernel/mcoverlayfs/Makefile kernel/Makefile kernel/Makefile.build arch/x86/tools/mcreboot-attached-mic.sh arch/x86/tools/mcshutdown-attached-mic.sh arch/x86/tools/mcreboot-builtin-x86.sh arch/x86/tools/mcreboot-smp-x86.sh arch/x86/tools/mcstop+release-smp-x86.sh arch/x86/tools/mcshutdown-builtin-x86.sh arch/x86/tools/mcreboot.1:arch/x86/tools/mcreboot.1in"
+ac_config_files="$ac_config_files Makefile executer/user/Makefile 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 kernel/Makefile kernel/Makefile.build arch/x86/tools/mcreboot-attached-mic.sh arch/x86/tools/mcshutdown-attached-mic.sh 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/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"
 
 
 if test "x$enable_dcfa" = xyes; then :
@@ -4590,6 +4595,8 @@ do
     "executer/kernel/mcctrl/Makefile") CONFIG_FILES="$CONFIG_FILES executer/kernel/mcctrl/Makefile" ;;
     "executer/kernel/mcctrl/arch/x86_64/Makefile") CONFIG_FILES="$CONFIG_FILES executer/kernel/mcctrl/arch/x86_64/Makefile" ;;
     "executer/kernel/mcoverlayfs/Makefile") CONFIG_FILES="$CONFIG_FILES executer/kernel/mcoverlayfs/Makefile" ;;
+    "executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile") CONFIG_FILES="$CONFIG_FILES executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile" ;;
+    "executer/kernel/mcoverlayfs/linux-4.0.9/Makefile") CONFIG_FILES="$CONFIG_FILES executer/kernel/mcoverlayfs/linux-4.0.9/Makefile" ;;
     "kernel/Makefile") CONFIG_FILES="$CONFIG_FILES kernel/Makefile" ;;
     "kernel/Makefile.build") CONFIG_FILES="$CONFIG_FILES kernel/Makefile.build" ;;
     "arch/x86/tools/mcreboot-attached-mic.sh") CONFIG_FILES="$CONFIG_FILES arch/x86/tools/mcreboot-attached-mic.sh" ;;
@@ -4599,6 +4606,8 @@ do
     "arch/x86/tools/mcstop+release-smp-x86.sh") CONFIG_FILES="$CONFIG_FILES arch/x86/tools/mcstop+release-smp-x86.sh" ;;
     "arch/x86/tools/mcshutdown-builtin-x86.sh") CONFIG_FILES="$CONFIG_FILES arch/x86/tools/mcshutdown-builtin-x86.sh" ;;
     "arch/x86/tools/mcreboot.1") CONFIG_FILES="$CONFIG_FILES arch/x86/tools/mcreboot.1:arch/x86/tools/mcreboot.1in" ;;
+    "arch/x86/tools/irqbalance_mck.service") CONFIG_FILES="$CONFIG_FILES arch/x86/tools/irqbalance_mck.service" ;;
+    "arch/x86/tools/irqbalance_mck.in") CONFIG_FILES="$CONFIG_FILES arch/x86/tools/irqbalance_mck.in" ;;
     "kernel/Makefile.dcfa") CONFIG_FILES="$CONFIG_FILES kernel/Makefile.dcfa" ;;
 
   *) as_fn_error $? "invalid argument: \`$ac_config_target'" "$LINENO" 5;;

configure.ac

@@ -146,6 +146,9 @@ 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
@@ -278,6 +281,7 @@ AC_SUBST(KDIR)
 AC_SUBST(TARGET)
 AC_SUBST(BINDIR)
 AC_SUBST(SBINDIR)
+AC_SUBST(ETCDIR)
 AC_SUBST(KMODDIR)
 AC_SUBST(KERNDIR)
 AC_SUBST(MANDIR)
@@ -298,6 +302,8 @@ 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
 	kernel/Makefile
 	kernel/Makefile.build
 	arch/x86/tools/mcreboot-attached-mic.sh
@@ -307,6 +313,8 @@ AC_CONFIG_FILES([
 	arch/x86/tools/mcstop+release-smp-x86.sh
 	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/include/uprotocol.h

@@ -110,6 +110,13 @@ struct program_load_desc {
 };
 
 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 +135,17 @@ 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;

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
@@ -100,8 +100,6 @@ 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

@@ -255,7 +255,7 @@ 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/control.c

@@ -32,6 +32,8 @@
 #include <linux/fs.h>
 #include <linux/file.h>
 #include <linux/version.h>
+#include <linux/semaphore.h>
+#include <linux/interrupt.h>
 #include <asm/uaccess.h>
 #include <asm/delay.h>
 #include <asm/io.h>
@@ -80,7 +82,6 @@ static long mcexec_prepare_image(ihk_os_t os,
 	void *args, *envs;
 	long ret = 0;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
-	unsigned long flags;
 	struct mcctrl_per_proc_data *ppd = NULL;
 
 	if (copy_from_user(&desc, udesc,
@@ -123,52 +124,48 @@ static long mcexec_prepare_image(ihk_os_t os,
 	}
 
 	pdesc->args = (void*)virt_to_phys(args);
-	printk("args: 0x%lX\n", (unsigned long)pdesc->args);
-	printk("argc: %ld\n", *(long *)args);
+	dprintk("args: 0x%lX\n", (unsigned long)pdesc->args);
+	dprintk("argc: %ld\n", *(long *)args);
 	pdesc->envs = (void*)virt_to_phys(envs);
-	printk("envs: 0x%lX\n", (unsigned long)pdesc->envs);
-	printk("envc: %ld\n", *(long *)envs);
+	dprintk("envs: 0x%lX\n", (unsigned long)pdesc->envs);
+	dprintk("envc: %ld\n", *(long *)envs);
 
 	isp.msg = SCD_MSG_PREPARE_PROCESS;
 	isp.ref = pdesc->cpu;
 	isp.arg = virt_to_phys(pdesc);
 
-	printk("# of sections: %d\n", pdesc->num_sections);
-	printk("%p (%lx)\n", pdesc, isp.arg);
+	dprintk("# of sections: %d\n", pdesc->num_sections);
+	dprintk("%p (%lx)\n", pdesc, isp.arg);
 	
 	pdesc->status = 0;
 	mcctrl_ikc_send(os, pdesc->cpu, &isp);
 
-	wait_event_interruptible(usrdata->wq_prepare, pdesc->status);
+	while (wait_event_interruptible(usrdata->wq_prepare, pdesc->status) != 0);
 
 	if(pdesc->err < 0){
 		ret = pdesc->err;	
 		goto free_out;
 	}
 
-	ppd = kmalloc(sizeof(*ppd), GFP_ATOMIC);
+	ppd = mcctrl_get_per_proc_data(usrdata, task_tgid_vnr(current));
 	if (!ppd) {
-		printk("ERROR: allocating per process data\n");
-		ret = -ENOMEM;
+		printk("ERROR: no per process data for PID %d\n", task_tgid_vnr(current));
+		ret = -EINVAL;
 		goto free_out;
 	}
 
-	ppd->pid = pdesc->pid;
+	/* Update rpgtable */
 	ppd->rpgtable = pdesc->rpgtable;
-
-	flags = ihk_ikc_spinlock_lock(&usrdata->per_proc_list_lock);
-	list_add_tail(&ppd->list, &usrdata->per_proc_list);
-	ihk_ikc_spinlock_unlock(&usrdata->per_proc_list_lock, flags);
 	
-	dprintk("pid %d, rpgtable: 0x%lx added\n", 
-		ppd->pid, ppd->rpgtable);
-
 	if (copy_to_user(udesc, pdesc, sizeof(struct program_load_desc) + 
 	             sizeof(struct program_image_section) * desc.num_sections)) {
 		ret = -EFAULT;	
 		goto free_out;
 	}
 
+	dprintk("%s: pid %d, rpgtable: 0x%lx added\n", 
+		__FUNCTION__, ppd->pid, ppd->rpgtable);
+
 	ret = 0;
 
 free_out:
@@ -416,19 +413,200 @@ static long mcexec_get_cpu(ihk_os_t os)
 	return info->n_cpus;
 }
 
-int mcexec_syscall(struct mcctrl_channel *c, int pid, unsigned long arg)
+int mcctrl_add_per_proc_data(struct mcctrl_usrdata *ud, int pid, 
+	struct mcctrl_per_proc_data *ppd)
+{
+	struct mcctrl_per_proc_data *ppd_iter;
+	int hash = (pid & MCCTRL_PER_PROC_DATA_HASH_MASK);
+	int ret = 0;
+	unsigned long flags;
+
+	/* Check if data for this thread exists and add if not */
+	write_lock_irqsave(&ud->per_proc_data_hash_lock[hash], flags);
+	list_for_each_entry(ppd_iter, &ud->per_proc_data_hash[hash], hash) {
+		if (ppd_iter->pid == pid) {
+			ret = -EBUSY;
+			goto out;
+		}
+	}
+
+	list_add_tail(&ppd->hash, &ud->per_proc_data_hash[hash]);
+
+out:
+	write_unlock_irqrestore(&ud->per_proc_data_hash_lock[hash], flags);
+	return ret;
+}
+
+int mcctrl_delete_per_proc_data(struct mcctrl_usrdata *ud, int pid)
+{
+	struct mcctrl_per_proc_data *ppd_iter, *ppd = NULL;
+	int hash = (pid & MCCTRL_PER_PROC_DATA_HASH_MASK);
+	int ret = 0;
+	unsigned long flags;
+
+	write_lock_irqsave(&ud->per_proc_data_hash_lock[hash], flags);
+	list_for_each_entry(ppd_iter, &ud->per_proc_data_hash[hash], hash) {
+		if (ppd_iter->pid == pid) {
+			ppd = ppd_iter;
+			break;
+		}
+	}
+
+	if (!ppd) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	list_del(&ppd->hash);
+
+out:
+	write_unlock_irqrestore(&ud->per_proc_data_hash_lock[hash], flags);
+	return ret;
+}
+
+inline struct mcctrl_per_proc_data *mcctrl_get_per_proc_data(
+		struct mcctrl_usrdata *ud, int pid)
+{
+	struct mcctrl_per_proc_data *ppd_iter, *ppd = NULL;
+	int hash = (pid & MCCTRL_PER_PROC_DATA_HASH_MASK);
+	unsigned long flags;
+
+	/* Check if data for this process exists and return it */
+	read_lock_irqsave(&ud->per_proc_data_hash_lock[hash], flags);
+
+	list_for_each_entry(ppd_iter, &ud->per_proc_data_hash[hash], hash) {
+		if (ppd_iter->pid == pid) {
+			ppd = ppd_iter;
+			break;
+		}
+	}
+
+	read_unlock_irqrestore(&ud->per_proc_data_hash_lock[hash], flags);
+	return ppd;
+}
+
+/*
+ * Called indirectly from the IKC message handler.
+ */
+int mcexec_syscall(struct mcctrl_usrdata *ud, struct ikc_scd_packet *packet)
 {
 	struct wait_queue_head_list_node *wqhln = NULL;
 	struct wait_queue_head_list_node *wqhln_iter;
+	struct wait_queue_head_list_node *wqhln_alloc = NULL;
+	int pid = packet->pid;
 	unsigned long flags;
+	struct mcctrl_per_proc_data *ppd;
+
+	/* Look up per-process structure */
+	ppd = mcctrl_get_per_proc_data(ud, pid);
+
+	if (unlikely(!ppd)) {
+		kprintf("%s: ERROR: no per-process structure for PID %d??\n",
+			__FUNCTION__, task_tgid_vnr(current));
+			return 0;
+	}
+
+	dprintk("%s: (packet_handler) rtid: %d, ttid: %d, sys nr: %d\n",
+			__FUNCTION__,
+			packet->req.rtid,
+			packet->req.ttid,
+			packet->req.number);
+	/*
+	 * Three scenarios are possible:
+	 * - Find the designated thread if req->ttid is specified.
+	 * - Find any available thread if req->ttid is zero.
+	 * - Add a request element if no threads are available.
+	 */
+	flags = ihk_ikc_spinlock_lock(&ppd->wq_list_lock);
+
+	/* Is this a request for a specific thread? See if it's waiting */
+	if (unlikely(packet->req.ttid)) {
+		list_for_each_entry(wqhln_iter, &ppd->wq_list_exact, list) {
+			if (packet->req.ttid != task_pid_vnr(wqhln_iter->task))
+				continue;
 
-	/* Look up per-process wait queue head with pid */
-	flags = ihk_ikc_spinlock_lock(&c->wq_list_lock);
-	list_for_each_entry(wqhln_iter, &c->wq_list, list) {
-		if (wqhln_iter->pid == pid) {
 			wqhln = wqhln_iter;
 			break;
 		}
+		if (!wqhln) {
+			printk("%s: WARNING: no target thread found for exact request??\n",
+				__FUNCTION__);
+		}
+	}
+	/* Is there any thread available? */
+	else {
+		list_for_each_entry(wqhln_iter, &ppd->wq_list, list) {
+			if (wqhln_iter->task && !wqhln_iter->req) {
+				wqhln = wqhln_iter;
+				break;
+			}
+		}
+	}
+
+	/* If no match found, add request to pending request list */
+	if (unlikely(!wqhln)) {
+retry_alloc:
+		wqhln_alloc = kmalloc(sizeof(*wqhln), GFP_ATOMIC);
+		if (!wqhln_alloc) {
+			printk("WARNING: coudln't alloc wait queue head, retrying..\n");
+			goto retry_alloc;
+		}
+
+		wqhln = wqhln_alloc;
+		wqhln->req = 0;
+		wqhln->task = NULL;
+		init_waitqueue_head(&wqhln->wq_syscall);
+		list_add_tail(&wqhln->list, &ppd->wq_req_list);
+	}
+
+	wqhln->packet = packet;
+	wqhln->req = 1;
+	wake_up(&wqhln->wq_syscall);
+	ihk_ikc_spinlock_unlock(&ppd->wq_list_lock, flags);
+
+	return 0;
+}
+
+/*
+ * Called from an mcexec thread via ioctl().
+ */
+int mcexec_wait_syscall(ihk_os_t os, struct syscall_wait_desc *__user req)
+{
+	struct ikc_scd_packet *packet;
+	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
+	struct wait_queue_head_list_node *wqhln = NULL;
+	struct wait_queue_head_list_node *wqhln_iter;
+	int ret = 0;
+	unsigned long irqflags;
+	struct mcctrl_per_proc_data *ppd;
+
+	/* Look up per-process structure */
+	ppd = mcctrl_get_per_proc_data(usrdata, task_tgid_vnr(current));
+
+	if (unlikely(!ppd)) {
+		kprintf("%s: ERROR: no per-process structure for PID %d??\n",
+			__FUNCTION__, task_tgid_vnr(current));
+			return -EINVAL;
+	}
+
+	packet = (struct ikc_scd_packet *)mcctrl_get_per_thread_data(ppd, current);
+	if (packet) {
+		printk("%s: ERROR: packet %p is already registered for thread %d\n",
+				__FUNCTION__, packet, task_pid_vnr(current));
+		return -EBUSY;
+	}
+
+retry:
+	/* Prepare per-thread wait queue head or find a valid request */
+	irqflags = ihk_ikc_spinlock_lock(&ppd->wq_list_lock);
+	/* First see if there is a valid request already that is not yet taken */
+	list_for_each_entry(wqhln_iter, &ppd->wq_req_list, list) {
+		if (wqhln_iter->task == NULL && wqhln_iter->req) {
+			wqhln = wqhln_iter;
+			wqhln->task = current;
+			list_del(&wqhln->list);
+			break;
+		}
 	}
 
 	if (!wqhln) {
@@ -439,180 +617,86 @@ retry_alloc:
 			goto retry_alloc;
 		}
 
-		wqhln->pid = pid;
+		wqhln->task = current;
 		wqhln->req = 0;
 		init_waitqueue_head(&wqhln->wq_syscall);
-		list_add_tail(&wqhln->list, &c->wq_list);
+
+		/* Wait for a request.. */
+		list_add(&wqhln->list, &ppd->wq_list);
+		ihk_ikc_spinlock_unlock(&ppd->wq_list_lock, irqflags);
+
+		ret = wait_event_interruptible(wqhln->wq_syscall, wqhln->req);
+
+		/* Remove per-thread wait queue head */
+		irqflags = ihk_ikc_spinlock_lock(&ppd->wq_list_lock);
+		list_del(&wqhln->list);
 	}
+	ihk_ikc_spinlock_unlock(&ppd->wq_list_lock, irqflags);
 
-	wqhln->req = 1;
-	wake_up(&wqhln->wq_syscall);
-	ihk_ikc_spinlock_unlock(&c->wq_list_lock, flags);
-
-	return 0;
-}
-
-#ifndef DO_USER_MODE
-// static int remaining_job, base_cpu, job_pos;
-#endif
-
-// extern int num_channels;
-// extern int mcctrl_dma_abort;
-
-int mcexec_wait_syscall(ihk_os_t os, struct syscall_wait_desc *__user req)
-{
-	struct syscall_wait_desc swd;
-	struct mcctrl_channel *c;
-	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
-	struct wait_queue_head_list_node *wqhln;
-	struct wait_queue_head_list_node *wqhln_iter;
-	int ret = 0;
-	unsigned long irqflags;
-#ifndef DO_USER_MODE
-	unsigned long s, w, d;
-#endif
-	
-//printk("mcexec_wait_syscall swd=%p req=%p size=%d\n", &swd, req, sizeof(swd.cpu));
-	if (copy_from_user(&swd, req, sizeof(swd))) {
-		return -EFAULT;
-	}
-
-	if (swd.cpu >= usrdata->num_channels)
-		return -EINVAL;
-
-	c = get_peer_channel(usrdata, current);
-	if (c) {
-		printk("mcexec_wait_syscall:already registered. task %p ch %p\n",
-				current, c);
-		return -EBUSY;
-	}
-	c = usrdata->channels + swd.cpu;
-
-#ifdef DO_USER_MODE
-retry:
-	/* Prepare per-process wait queue head */
-retry_alloc:
-	wqhln = kmalloc(sizeof(*wqhln), GFP_KERNEL);
-	if (!wqhln) {
-		printk("WARNING: coudln't alloc wait queue head, retrying..\n");
-		goto retry_alloc;
-	}
-
-	wqhln->pid = swd.pid;	
-	wqhln->req = 0;
-	init_waitqueue_head(&wqhln->wq_syscall);
-	
-	irqflags = ihk_ikc_spinlock_lock(&c->wq_list_lock);
-	/* First see if there is one wait queue already */
-	list_for_each_entry(wqhln_iter, &c->wq_list, list) {
-		if (wqhln_iter->pid == task_tgid_vnr(current)) {
-			kfree(wqhln);
-			wqhln = wqhln_iter;
-			list_del(&wqhln->list);
-			break;
-		}
-	}
-	list_add_tail(&wqhln->list, &c->wq_list);
-	ihk_ikc_spinlock_unlock(&c->wq_list_lock, irqflags);
-
-	ret = wait_event_interruptible(wqhln->wq_syscall, wqhln->req);
-	
-
-	/* Remove per-process wait queue head */
-	irqflags = ihk_ikc_spinlock_lock(&c->wq_list_lock);
-	list_del(&wqhln->list);
-	ihk_ikc_spinlock_unlock(&c->wq_list_lock, irqflags);
 	if (ret && !wqhln->req) {
 		kfree(wqhln);
+		wqhln = NULL;
 		return -EINTR;
 	}
+
+	packet = wqhln->packet;
 	kfree(wqhln);
+	wqhln = NULL;
 
-	if (c->param.request_va->number == 61 &&
-			c->param.request_va->args[0] == swd.pid) {
+	dprintk("%s: tid: %d request from CPU %d\n",
+			__FUNCTION__, task_pid_vnr(current), packet->ref);
 
-		dprintk("pid: %d, tid: %d: SC %d, swd.cpu: %d, WARNING: wait4() for self?\n",
-				task_tgid_vnr(current),
-				task_pid_vnr(current);
-				c->param.request_va->number,
-				swd.cpu);
-
-		return -EINTR;
-	}
-
-#if 1
 	mb();
-	if (!c->param.request_va->valid) {
-printk("mcexec_wait_syscall:stray wakeup\n");
+	if (!packet->req.valid) {
+		printk("%s: ERROR: stray wakeup pid: %d, tid: %d: SC %lu\n",
+				__FUNCTION__,
+				task_tgid_vnr(current),
+				task_pid_vnr(current),
+				packet->req.number);
+		ihk_ikc_release_packet((struct ihk_ikc_free_packet *)packet,
+				(usrdata->channels + packet->ref)->c);
 		goto retry;
 	}
-#endif
-#else
-	while (1) {
-		c = usrdata->channels + swd.cpu;
-		ihk_get_tsc(s);
-		if (!usrdata->remaining_job) {
-			while (!(*c->param.doorbell_va)) {
-				mb();
-				cpu_relax();
-				ihk_get_tsc(w);
-				if (w > s + 1024UL * 1024 * 1024 * 10) {
-					return -EINTR;
-				}
-			}
-			d = (*c->param.doorbell_va) - 1;
-			*c->param.doorbell_va = 0;
 
-			if (d < 0 || d >= usrdata->num_channels) {
-				d = 0;
-			}
-			usrdata->base_cpu = d;
-			usrdata->job_pos = 0;
-			usrdata->remaining_job = 1;
-		} else { 
-			usrdata->job_pos++;
-		}
-		
-		for (; usrdata->job_pos < usrdata->num_channels; usrdata->job_pos++) {
-			if (base_cpu + job_pos >= num_channels) {
-				c = usrdata->channels + 
-					(usrdata->base_cpu + usrdata->job_pos - usrdata->num_channels);
-			} else {
-				c = usrdata->channels + usrdata->base_cpu + usrdata->job_pos;
-			}
-			if (!c) {
-				continue;
-			}
-			if (c->param.request_va &&
-			    c->param.request_va->valid) {
-#endif
-				c->param.request_va->valid = 0; /* ack */
-				dprintk("SC #%lx, %lx\n",
-				        c->param.request_va->number,
-				        c->param.request_va->args[0]);
-				register_peer_channel(usrdata, current, c);
-				if (__do_in_kernel_syscall(os, c, c->param.request_va)) {
-					if (copy_to_user(&req->sr, c->param.request_va,
-							 sizeof(struct syscall_request))) {
-						deregister_peer_channel(usrdata, current, c);
-						return -EFAULT;
-					}
-					return 0;
-				}
-				deregister_peer_channel(usrdata, current, c);
-#ifdef	DO_USER_MODE
-				goto retry;
-#endif
-#ifndef DO_USER_MODE
-				if (usrdata->mcctrl_dma_abort) {
-					return -2;
-				}
-			}
-		}
-		usrdata->remaining_job = 0;
+	packet->req.valid = 0; /* ack */
+	dprintk("%s: system call: %d, args[0]: %lu, args[1]: %lu, args[2]: %lu, "
+			"args[3]: %lu, args[4]: %lu, args[5]: %lu\n",
+			__FUNCTION__,
+			packet->req.number,
+			packet->req.args[0],
+			packet->req.args[1],
+			packet->req.args[2],
+			packet->req.args[3],
+			packet->req.args[4],
+			packet->req.args[5]);
+	
+	if (mcctrl_add_per_thread_data(ppd, current, packet) < 0) {
+		kprintf("%s: error adding per-thread data\n", __FUNCTION__);
+		return -EINVAL;
 	}
-#endif
-	return 0;
+
+	if (__do_in_kernel_syscall(os, packet)) {
+		if (copy_to_user(&req->sr, &packet->req,
+					sizeof(struct syscall_request))) {
+
+			if (mcctrl_delete_per_thread_data(ppd, current) < 0) {
+				kprintf("%s: error deleting per-thread data\n", __FUNCTION__);
+				return -EINVAL;
+			}
+			return -EFAULT;
+		}
+		return 0;
+	}
+
+	ihk_ikc_release_packet((struct ihk_ikc_free_packet *)packet,
+			(usrdata->channels + packet->ref)->c);
+
+	if (mcctrl_delete_per_thread_data(ppd, current) < 0) {
+		kprintf("%s: error deleting per-thread data\n", __FUNCTION__);
+		return -EINVAL;
+	}
+
+	goto retry;
 }
 
 long mcexec_pin_region(ihk_os_t os, unsigned long *__user arg)
@@ -695,33 +779,6 @@ long mcexec_load_syscall(ihk_os_t os, struct syscall_load_desc *__user arg)
 #endif
 	
 	ihk_device_unmap_memory(ihk_os_to_dev(os), phys, desc.size);	
-	
-/*
-	ihk_dma_channel_t channel;
-	struct ihk_dma_request request;
-	unsigned long dma_status = 0;
-
-	channel = ihk_device_get_dma_channel(ihk_os_to_dev(os), 0);
-	if (!channel) {
-		return -EINVAL;
-	}
-
-	memset(&request, 0, sizeof(request));
-	request.src_os = os;
-	request.src_phys = desc.src;
-	request.dest_os = NULL;
-	request.dest_phys = desc.dest;
-	request.size = desc.size;
-	request.notify = (void *)virt_to_phys(&dma_status);
-	request.priv = (void *)1;
-
-	ihk_dma_request(channel, &request);
-
-	while (!dma_status) {
-		mb();
-		udelay(1);
-	}
-*/
 
 	return 0;
 }
@@ -729,80 +786,66 @@ long mcexec_load_syscall(ihk_os_t os, struct syscall_load_desc *__user arg)
 long mcexec_ret_syscall(ihk_os_t os, struct syscall_ret_desc *__user arg)
 {
 	struct syscall_ret_desc ret;
-	struct mcctrl_channel *mc;
+	struct ikc_scd_packet *packet;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
-#if 0	
-	ihk_dma_channel_t channel;
-	struct ihk_dma_request request;
-
-	channel = ihk_device_get_dma_channel(ihk_os_to_dev(os), 0);
-	if (!channel) {
-		return -EINVAL;
-	}
-#endif	
+	struct mcctrl_per_proc_data *ppd;
 
 	if (copy_from_user(&ret, arg, sizeof(struct syscall_ret_desc))) {
 		return -EFAULT;
 	}
-	mc = usrdata->channels + ret.cpu;
-	if (!mc) {
+
+	/* Look up per-process structure */
+	ppd = mcctrl_get_per_proc_data(usrdata, task_tgid_vnr(current));
+	if (!ppd) {
+		kprintf("%s: ERROR: no per-process structure for PID %d??\n", 
+				__FUNCTION__, task_tgid_vnr(current));
 		return -EINVAL;
 	}
-	deregister_peer_channel(usrdata, current, mc);
 
-	mc->param.response_va->ret = ret.ret;
+	packet = (struct ikc_scd_packet *)mcctrl_get_per_thread_data(ppd, current);
+	if (!packet) {
+		kprintf("%s: ERROR: no packet registered for TID %d\n", 
+			__FUNCTION__, task_pid_vnr(current));
+		return -EINVAL;
+	}
+
+	mcctrl_delete_per_thread_data(ppd, current);
 
 	if (ret.size > 0) {
 		/* Host => Accel. Write is fast. */
 		unsigned long phys;
 		void *rpm;
 
-		phys = ihk_device_map_memory(ihk_os_to_dev(os), ret.dest,
-		                             ret.size);
+		phys = ihk_device_map_memory(ihk_os_to_dev(os), ret.dest, ret.size);
 #ifdef CONFIG_MIC
 		rpm = ioremap_wc(phys, ret.size);
 #else
 		rpm = ihk_device_map_virtual(ihk_os_to_dev(os), phys, 
 		                             ret.size, NULL, 0);
 #endif
-		
 		if (copy_from_user(rpm, (void *__user)ret.src, ret.size)) {
 			return -EFAULT;
 		}
 
-		mb();
-		mc->param.response_va->status = 1;
-
 #ifdef CONFIG_MIC
 		iounmap(rpm);
 #else
 		ihk_device_unmap_virtual(ihk_os_to_dev(os), rpm, ret.size);
 #endif
 		ihk_device_unmap_memory(ihk_os_to_dev(os), phys, ret.size);
+	} 
 
-/*
-		memset(&request, 0, sizeof(request));
-		request.src_os = NULL;
-		request.src_phys = ret.src;
-		request.dest_os = os;
-		request.dest_phys = ret.dest;
-		request.size = ret.size;
-		request.notify_os = os;
-		request.notify = (void *)mc->param.response_rpa;
-		request.priv = (void *)1;
-		
-		ihk_dma_request(channel, &request);
-*/
-	} else {
-		mb();
-		mc->param.response_va->status = 1;
-	}
+	__return_syscall(os, packet, ret.ret, task_pid_vnr(current));
+
+	/* Free packet */
+	ihk_ikc_release_packet((struct ihk_ikc_free_packet *)packet,
+			(usrdata->channels + packet->ref)->c);
 
 	return 0;
 }
 
 LIST_HEAD(mckernel_exec_files);
-DEFINE_SPINLOCK(mckernel_exec_file_lock);
+DEFINE_SEMAPHORE(mckernel_exec_file_lock);
  
 
 struct mckernel_exec_file {
@@ -861,14 +904,53 @@ int mcexec_open_exec(ihk_os_t os, char * __user filename)
 	int retval;
 	int os_ind = ihk_host_os_get_index(os);
 	char *pathbuf, *fullpath;
+	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
+	struct mcctrl_per_proc_data *ppd = NULL;
+	int i;
 
 	if (os_ind < 0) {
 		return EINVAL;
 	}
 
+	ppd = mcctrl_get_per_proc_data(usrdata, task_tgid_vnr(current));
+
+	if (!ppd) {
+		ppd = kmalloc(sizeof(*ppd), GFP_KERNEL);
+		if (!ppd) {
+			printk("ERROR: allocating per process data\n");
+			return -ENOMEM;
+		}
+
+		ppd->pid = task_tgid_vnr(current);
+		/*
+		 * XXX: rpgtable will be updated in __do_in_kernel_syscall()
+		 * under case __NR_munmap
+		 */
+		INIT_LIST_HEAD(&ppd->wq_list);
+		INIT_LIST_HEAD(&ppd->wq_req_list);
+		INIT_LIST_HEAD(&ppd->wq_list_exact);
+		spin_lock_init(&ppd->wq_list_lock);
+
+		for (i = 0; i < MCCTRL_PER_THREAD_DATA_HASH_SIZE; ++i) {
+			INIT_LIST_HEAD(&ppd->per_thread_data_hash[i]);
+			rwlock_init(&ppd->per_thread_data_hash_lock[i]);
+		}
+
+		if (mcctrl_add_per_proc_data(usrdata, ppd->pid, ppd) < 0) {
+			printk("%s: error adding per process data\n", __FUNCTION__);
+			retval = EINVAL;
+			goto out_free_ppd;
+		}
+	}
+	else {
+		/* Only deallocate in case of an error if we added it above */
+		ppd = NULL;
+	}
+
 	pathbuf = kmalloc(PATH_MAX, GFP_TEMPORARY);
 	if (!pathbuf) {
-		return ENOMEM;
+		retval = ENOMEM;
+		goto out_error_drop_ppd;
 	}
 
 	file = open_exec(filename);
@@ -889,7 +971,7 @@ int mcexec_open_exec(ihk_os_t os, char * __user filename)
 		goto out_put_file;
 	}
 
-	spin_lock_irq(&mckernel_exec_file_lock);
+	down(&mckernel_exec_file_lock);
 	/* Find previous file (if exists) and drop it */
 	list_for_each_entry(mcef_iter, &mckernel_exec_files, list) {
 		if (mcef_iter->os == os && mcef_iter->pid == task_tgid_vnr(current)) {
@@ -900,7 +982,7 @@ int mcexec_open_exec(ihk_os_t os, char * __user filename)
 			break;
 		}
 	}
-	
+
 	/* Add new exec file to the list */
 	mcef->os = os;
 	mcef->pid = task_tgid_vnr(current);
@@ -910,19 +992,22 @@ int mcexec_open_exec(ihk_os_t os, char * __user filename)
 	/* Create /proc/self/exe entry */
 	add_pid_entry(os_ind, task_tgid_vnr(current));
 	proc_exe_link(os_ind, task_tgid_vnr(current), fullpath);
-	spin_unlock(&mckernel_exec_file_lock);
+	up(&mckernel_exec_file_lock);
 
 	dprintk("%d open_exec and holding file: %s\n", (int)task_tgid_vnr(current), filename);
 
 	kfree(pathbuf);
 
 	return 0;
-	
+
 out_put_file:
 	fput(file);
-
 out_error_free:
 	kfree(pathbuf);
+out_error_drop_ppd:
+	if (ppd) mcctrl_delete_per_proc_data(usrdata, ppd->pid);
+out_free_ppd:
+	if (ppd) kfree(ppd);
 	return -retval;
 }
 
@@ -932,12 +1017,29 @@ int mcexec_close_exec(ihk_os_t os)
 	struct mckernel_exec_file *mcef = NULL;
 	int found = 0;
 	int os_ind = ihk_host_os_get_index(os);	
+	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
+	struct mcctrl_per_proc_data *ppd = NULL;
+
+	ppd = mcctrl_get_per_proc_data(usrdata, task_tgid_vnr(current));
+
+	if (ppd) {
+		mcctrl_delete_per_proc_data(usrdata, ppd->pid);
+
+		dprintk("pid: %d, tid: %d: rpgtable for %d (0x%lx) removed\n",
+				task_tgid_vnr(current), current->pid, ppd->pid, ppd->rpgtable);
+
+		kfree(ppd);
+	}
+	else {
+		printk("WARNING: no per process data for pid %d ?\n", 
+				task_tgid_vnr(current));
+	}
 
 	if (os_ind < 0) {
 		return EINVAL;
 	}
 		
-	spin_lock_irq(&mckernel_exec_file_lock);
+	down(&mckernel_exec_file_lock);
 	list_for_each_entry(mcef, &mckernel_exec_files, list) {
 		if (mcef->os == os && mcef->pid == task_tgid_vnr(current)) {
 			allow_write_access(mcef->fp);
@@ -950,7 +1052,7 @@ int mcexec_close_exec(ihk_os_t os)
 		}
 	}
 
-	spin_unlock(&mckernel_exec_file_lock);
+	up(&mckernel_exec_file_lock);
 
 	return (found ? 0 : EINVAL);
 }

executer/kernel/mcctrl/driver.c

@@ -82,79 +82,109 @@ 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)
+{
+	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);
+
+	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;
+
 #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);
-		}
-	}
-
 	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>
@@ -40,16 +41,18 @@
 
 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);
 
+/* 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;
@@ -63,7 +66,7 @@ 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:
@@ -88,11 +91,8 @@ 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:
@@ -110,6 +110,14 @@ 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 +154,6 @@ 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;
 }
 
@@ -193,12 +199,13 @@ static void mcctrl_ikc_init(ihk_os_t os, int cpu, unsigned long rphys, struct ih
 #endif
 
 	pmc->param.request_va =
-		(void *)__get_free_pages(GFP_KERNEL,
+		(void *)__get_free_pages(in_interrupt() ? GFP_ATOMIC : 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_va = (void *)__get_free_page(in_interrupt() ? 
+			GFP_ATOMIC : 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);
@@ -218,8 +225,9 @@ static void mcctrl_ikc_init(ihk_os_t os, int cpu, unsigned long rphys, struct ih
 													PAGE_SIZE, NULL, 0);
 #endif
 
-	pmc->dma_buf = (void *)__get_free_pages(GFP_KERNEL,
-	                                        DMA_PIN_SHIFT - PAGE_SHIFT);
+	pmc->dma_buf = (void *)__get_free_pages(in_interrupt() ? 
+			GFP_ATOMIC : GFP_KERNEL,
+			DMA_PIN_SHIFT - PAGE_SHIFT);
 
 	rpm->request_page = pmc->param.request_pa;
 	rpm->doorbell_page = pmc->param.doorbell_pa;
@@ -265,9 +273,6 @@ 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);
 
@@ -286,9 +291,6 @@ 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);
 
@@ -315,7 +317,7 @@ int prepare_ikc_channels(ihk_os_t os)
 {
 	struct ihk_cpu_info *info;
 	struct mcctrl_usrdata   *usrdata;
-	int error;
+	int i;
 
 	usrdata = kzalloc(sizeof(struct mcctrl_usrdata), GFP_KERNEL);
 	usrdata->mcctrl_doorbell_va = (void *)__get_free_page(GFP_KERNEL);
@@ -347,17 +349,14 @@ int prepare_ikc_channels(ihk_os_t os)
 	memcpy(&usrdata->listen_param2, &listen_param2, sizeof listen_param2);
 	ihk_ikc_listen_port(os, &usrdata->listen_param2);
 
-	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;
-	}
-
 	return 0;
 }
 
@@ -396,7 +395,6 @@ void destroy_ikc_channels(ihk_os_t os)
 	}
 	free_page((unsigned long)usrdata->mcctrl_doorbell_va);
 
-	destroy_peer_channel_registry(usrdata);
 	kfree(usrdata->channels);
 	kfree(usrdata);
 }

executer/kernel/mcctrl/mcctrl.h

@@ -41,6 +41,7 @@
 #include <ikc/master.h>
 #include <ihk/msr.h>
 #include <linux/semaphore.h>
+#include <linux/rwlock.h>
 #include <linux/threads.h>
 #include "sysfs.h"
 
@@ -48,6 +49,7 @@
 #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
@@ -110,8 +112,9 @@ 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 +123,13 @@ struct ikc_scd_packet {
 			long sysfs_arg2;
 			long sysfs_arg3;
 		};
+
+		/* SCD_MSG_SCHEDULE_THREAD */
+		struct {
+			int ttid;
+		};
 	};
+	char padding[12];
 };
 
 struct mcctrl_priv { 
@@ -154,8 +163,11 @@ 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 {
@@ -163,15 +175,30 @@ struct mcctrl_channel {
 	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 list_head hash;
 	int pid;
 	unsigned long rpgtable;	/* per process, not per OS */
+
+	struct list_head wq_list;
+	struct list_head wq_req_list;
+	struct list_head wq_list_exact;
+	ihk_spinlock_t wq_list_lock;
+
+	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];
 };
 
 struct sysfsm_req {
@@ -230,6 +257,10 @@ struct node_topology {
 
 #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;
@@ -245,8 +276,9 @@ struct mcctrl_usrdata {
 	unsigned long	last_thread_exec;
 	wait_queue_head_t	wq_prepare;
 	
-	struct list_head per_proc_list;
-	ihk_spinlock_t per_proc_list_lock;
+	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];
@@ -273,12 +305,22 @@ 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);
+inline struct mcctrl_per_proc_data *mcctrl_get_per_proc_data(
+		struct mcctrl_usrdata *ud, int pid);
+
+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 1000
 
@@ -301,6 +343,7 @@ struct procfs_file {
 };
 
 void procfs_answer(unsigned int arg, int err);
+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);

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
 
@@ -81,7 +83,7 @@ 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)
@@ -375,67 +377,62 @@ _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 +448,7 @@ 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 +460,13 @@ 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 +477,13 @@ 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)
 		delete_procfs_entries(e);
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflag);
+	up(&procfs_file_list_lock);
 }
 
 /**
@@ -719,6 +714,57 @@ 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

@@ -14,6 +14,7 @@
 #include <linux/slab.h>
 #include <linux/device.h>
 #include <linux/version.h>
+#include <linux/interrupt.h>
 #include "mcctrl.h"
 #include "sysfs_msg.h"
 

executer/kernel/mcoverlayfs/Makefile.in

@@ -1,7 +1,3 @@
-KDIR ?= @KDIR@
-ARCH ?= @ARCH@
-KMODDIR=@KMODDIR@
-src = @abs_srcdir@
 ENABLE_MCOVERLAYFS=@ENABLE_MCOVERLAYFS@
 
 RELEASE=$(shell uname -r)
@@ -9,31 +5,36 @@ MAJOR=$(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 [ "${RHEL_RELEASE}" == "" ]; then echo "org"; else echo "rhel"; fi)
+BUILD_MODULE=none
 ifeq ($(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 ${LINUX_VERSION_CODE} -lt 262400 ]; then echo "linux-4.0.9"; else echo "none"; fi)
+endif
+endif
+ifeq ($(BUILD_MODULE_TMP),rhel)
+ifeq ($(BUILD_MODULE),none)
+BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -eq 199168 -a ${RHEL_RELEASE} -eq 327 ]; 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)
 
 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/user/eclair.c

@@ -167,6 +167,7 @@ enum {
 	CURRENT_OFFSET,
 	RUNQ_OFFSET,
 	CPU_STATUS_OFFSET,
+	IDLE_THREAD_OFFSET,
 
 	/* process */
 	CTX_OFFSET,
@@ -204,6 +205,7 @@ 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));
@@ -251,6 +253,64 @@ static int setup_threads(void) {
 	ihk_mc_switch_context = lookup_symbol("ihk_mc_switch_context");
 	if (0) printf("ihk_mc_switch_context: %lx\n", ihk_mc_switch_context);
 
+	/* Set up idle threads first */
+	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));
+
+		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 = pid;
+		ti->tid = tid;
+		ti->cpu = cpu;
+		ti->lcpu = cpu;
+		ti->process = thread;
+		ti->clv = v;
+		ti->x86_clv = locals + locals_span*cpu;
+
+		*titailp = ti;
+		titailp = &ti->next;
+	}
+
 	for (cpu = 0; cpu < num_processors; ++cpu) {
 		uintptr_t v;
 		uintptr_t head;

executer/user/mcexec.c

@@ -101,6 +101,19 @@ 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
+#else
+#if RHEL_RELEASE_CODE == RHEL_RELEASE_VERSION(7,2)
+#define ENABLE_MCOVERLAYFS 1
+#endif // RHEL_RELEASE_CODE == 7.2
+#endif // RHEL_RELEASE_CODE
+#endif // ENABLE_MCOVERLAYFS
+
 typedef unsigned char   cc_t;
 typedef unsigned int    speed_t;
 typedef unsigned int    tcflag_t;
@@ -375,7 +388,7 @@ 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 +406,27 @@ 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 +454,9 @@ retry:
 			}
 
 			free(tofree);
+			if(!found){
+				return ENOENT;
+			}
 			break;
 		}
 
@@ -654,7 +669,7 @@ 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 +683,10 @@ 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 +708,20 @@ 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 +730,20 @@ 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 +759,8 @@ void transfer_image(int fd, struct program_load_desc *desc)
 			}
 		}
 	}
+
+	return 0;
 }
 
 void print_desc(struct program_load_desc *desc)
@@ -837,7 +883,10 @@ 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 +897,7 @@ 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 +980,10 @@ 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 +1014,11 @@ 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;
 	}
@@ -1068,9 +1124,9 @@ 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;
@@ -1091,7 +1147,6 @@ 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)
 {
@@ -1163,7 +1218,6 @@ static int isunshare(void)
 	__dprintf("err=%d\n", err);
 	return err;
 }
-#endif
 #endif // ENABLE_MCOVERLAYFS
 
 #define MCK_RLIMIT_AS	0
@@ -1353,7 +1407,6 @@ 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];
@@ -1401,12 +1454,11 @@ 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 +1470,7 @@ 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;
 		}
@@ -1480,6 +1532,19 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
+	n_threads = ncpu;
+	if (ncpu > 16) {
+		n_threads = 16;
+	}
+
+	/* 
+	 * 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));
 	memset(thread_data, '\0', sizeof(struct thread_data_s) * (ncpu + 1));
 
@@ -1522,7 +1587,10 @@ 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 +1629,7 @@ 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);
 	}
 
@@ -1623,16 +1691,14 @@ do_generic_syscall(
 }
 
 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;
 		}
 	}
 }
@@ -1738,9 +1804,7 @@ char *
 chgpath(char *in, char *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 +1855,8 @@ 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 +1895,13 @@ 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){
@@ -1903,6 +1969,39 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 				thread_data[oldcpuid].remote_tid = wtid;
 			}
 
+			/*
+			 * 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;
+				}
+
+				for (i = 0; i < ncpu && i < w.sr.args[4]; ++i) {
+					tids[i] = thread_data[i].tid;
+				}
+
+				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, thread_data[newcpuid].remote_tid, 0, 0, 0, 0);
 			break;
 		}
@@ -1945,7 +2044,9 @@ 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 +2055,9 @@ 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 +2097,6 @@ 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 +2110,8 @@ 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 +2222,7 @@ 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 +2236,7 @@ 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 +2257,7 @@ 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 +2315,10 @@ 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) {
@@ -2262,6 +2370,53 @@ 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;
@@ -2295,7 +2450,9 @@ return_execve2:
 			break;
 
 		}
-		
+
+		thread_data[cpu].remote_tid = -1;
+
 		//pthread_mutex_unlock(lock);
 	}
 	__dprint("timed out.\n");

kernel/debug.c

@@ -110,6 +110,7 @@ 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);

kernel/devobj.c

@@ -78,51 +78,52 @@ 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;
@@ -140,11 +141,12 @@ 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 +166,8 @@ 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);
 
@@ -192,7 +196,7 @@ static void devobj_release(struct memobj *memobj)
 		}
 
 		if (obj->pfn_table) {
-			free_pages(obj->pfn_table, 1);
+			ihk_mc_free_pages(obj->pfn_table, pfn_npages);
 		}
 		kfree(free_obj);
 	}
@@ -204,7 +208,7 @@ 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 +220,7 @@ 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

@@ -182,7 +182,7 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
 
 	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;
 	}
 

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;
 
 /**

kernel/host.c

@@ -332,6 +332,9 @@ 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 +376,16 @@ 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;
@@ -385,7 +394,7 @@ static int process_msg_prepare_process(unsigned long rphys)
 	            + sizeof(struct program_image_section) * n);
 
 	if((thread = create_thread(p->entry)) == NULL){
-		ihk_mc_free(pn);
+		kfree(pn);
 		ihk_mc_unmap_virtual(p, npages, 1);
 		ihk_mc_unmap_memory(NULL, phys, sz);
 		return -ENOMEM;
@@ -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);
@@ -443,7 +452,7 @@ 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);
@@ -452,7 +461,7 @@ err:
 
 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_va = ihk_mc_alloc_pages(RESPONSE_PAGE_COUNT, 0);
 	lparam->response_pa = virt_to_phys(lparam->response_va);
 
 	pcp->request_page = 0;
@@ -521,12 +530,7 @@ 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 terminate_host(int pid);
 extern void debug_log(long);
 
@@ -561,6 +565,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	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 +577,17 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	} *sp, info;
 	unsigned long pp;
 	int cpuid;
+	int ret = 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;
@@ -600,19 +600,21 @@ 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;
+			ret = -1;
+			break;
 		}
 		dkprintf("SCD_MSG_SCHEDULE_PROCESS: %lx\n", packet->arg);
 		thread = (struct thread *)packet->arg;
 		proc = thread->proc;
 
-		settid(thread, 0, cpuid, -1);
+		settid(thread, 0, cpuid, -1, 0, NULL);
 		proc->status = PS_RUNNING;
 		thread->status = PS_RUNNING;
 		chain_thread(thread);
@@ -620,7 +622,29 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		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 +659,25 @@ 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;
+		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:
@@ -654,7 +685,8 @@ 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);
@@ -662,17 +694,21 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		pckt.msg = SCD_MSG_REPLY_GET_CPU_MAPPING;
 		pckt.arg = packet->arg;
 		syscall_channel_send(c, &pckt);
-		return 0;
+		ret = 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)

kernel/include/cls.h

@@ -19,11 +19,13 @@
  * 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 */
 };
 
 #include <ihk/lock.h>
@@ -38,8 +40,9 @@ 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;
@@ -73,6 +76,7 @@ struct cpu_local_var {
 	int in_interrupt;
 	int no_preempt;
 	int timer_enabled;
+	int kmalloc_initialized;
 } __attribute__((aligned(64)));
 
 

kernel/include/kmalloc.h

@@ -32,11 +32,10 @@ void *_kmalloc(int size, enum 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 __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

@@ -141,6 +141,7 @@ 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/process.h

@@ -29,6 +29,7 @@
 #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
@@ -160,7 +161,7 @@
 #endif
 
 #define USER_STACK_NR_PAGES 8192
-#define KERNEL_STACK_NR_PAGES 25
+#define KERNEL_STACK_NR_PAGES 32
 
 #define NOPHYS ((uintptr_t)-1)
 
@@ -319,12 +320,14 @@ 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
@@ -346,6 +349,11 @@ 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 +368,9 @@ 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. */
@@ -556,6 +567,9 @@ 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;
 };
 
 struct process_vm {
@@ -675,5 +689,8 @@ void chain_process(struct process *);
 void chain_thread(struct thread *);
 void proc_init();
 void set_timer();
+struct sig_pending *hassigpending(struct thread *thread);
+void settid(struct thread *thread, int mode, int newcpuid, int oldcpuid,
+	int nr_tids, int *tids);
 
 #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
@@ -117,28 +118,6 @@ 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;
@@ -210,13 +189,58 @@ 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 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;
+		};
+	};
+	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;

kernel/init.c

@@ -371,7 +371,7 @@ int main(void)
 	}
 	kmsg_init(mode);
 
-	kputs("MCK started.\n");
+	kputs("IHK/McKernel started.\n");
 
 	arch_init();
 
@@ -393,7 +393,7 @@ int main(void)
 
 	futex_init();
 
-	kputs("MCK/IHK booted.\n");
+	kputs("IHK/McKernel booted.\n");
 
 #ifdef DCFA_KMOD
 	mc_cmd_client_init();

kernel/mem.c

@@ -156,13 +156,17 @@ void sbox_write(int offset, unsigned int value);
 
 static void query_free_mem_interrupt_handler(void *priv)
 {
-#ifdef ATTACHED_MIC
-	dkprintf("query free mem handler!\n");
-
 	int pages = ihk_pagealloc_query_free(pa_allocator);
 	
-	dkprintf("free pages: %d\n", pages);
+	kprintf("McKernel free pages: %d\n", pages);
 
+	if (find_command_line("memdebug")) {
+		extern void kmalloc_memcheck(void);
+
+		kmalloc_memcheck();
+	}
+
+#ifdef ATTACHED_MIC
 	sbox_write(SBOX_SCRATCH0, pages);
 	sbox_write(SBOX_SCRATCH1, 1);
 #endif
@@ -265,6 +269,13 @@ void remote_flush_tlb_cpumask(struct process_vm *vm,
 		unsigned long tsc;
 		tsc = rdtsc() + 12884901888;  /* 1.2GHz =>10 sec */
 #endif
+		if (flush_entry->addr) {
+			flush_tlb_single(flush_entry->addr & PAGE_MASK);
+		}
+		/* Zero address denotes full TLB flush */
+		else {
+			flush_tlb();
+		}
 
 		/* Wait for all cores */
 		while (ihk_atomic_read(&flush_entry->pending) != 0) {
@@ -335,10 +346,9 @@ static void page_fault_handler(void *fault_addr, uint64_t reason, void *regs)
 			// no return
 		}
 
-		kprintf("[%d]page_fault_handler(%p,%lx,%p):"
-				"fault vm failed. %d, TID: %d\n",
-				ihk_mc_get_processor_id(), fault_addr,
-				reason, regs, error, thread->tid);
+		kprintf("%s fault VM failed for TID: %d, addr: 0x%lx, "
+				"reason: %d, error: %d\n", __FUNCTION__,
+				thread->tid, fault_addr, reason, error);
 		unhandled_page_fault(thread, fault_addr, regs);
 		preempt_enable();
 		memset(&info, '\0', sizeof info);
@@ -425,8 +435,9 @@ static void page_allocator_init(void)
 	
 	ihk_mc_reserve_arch_pages(pa_start, pa_end, reserve_pages);
 
-	kprintf("Available pages: %ld pages\n",
-	        ihk_pagealloc_count(pa_allocator));
+	kprintf("Available memory: %ld bytes in %ld pages\n",
+	        (ihk_pagealloc_count(pa_allocator) * PAGE_SIZE), 
+			ihk_pagealloc_count(pa_allocator));
 
 	/* Notify the ihk to use my page allocator */
 	ihk_mc_set_page_allocator(&allocator);
@@ -507,6 +518,9 @@ static void page_init(void)
 
 static char *memdebug = NULL;
 
+static void *___kmalloc(int size, enum ihk_mc_ap_flag flag);
+static void ___kfree(void *ptr);
+
 void register_kmalloc(void)
 {
 	if(memdebug){
@@ -636,60 +650,100 @@ void mem_init(void)
 	}
 }
 
-struct location {
-	struct location *next;
-	int line;
-	int cnt;
-	char file[0];
-};
+#define KMALLOC_TRACK_HASH_SHIFT	(8)
+#define KMALLOC_TRACK_HASH_SIZE     (1 << KMALLOC_TRACK_HASH_SHIFT)
+#define KMALLOC_TRACK_HASH_MASK     (KMALLOC_TRACK_HASH_SIZE - 1)
 
-struct alloc {
-	struct alloc *next;
-	struct malloc_header *p;
-	struct location *loc;
-	int size;
+struct list_head kmalloc_track_hash[KMALLOC_TRACK_HASH_SIZE];
+ihk_spinlock_t kmalloc_track_hash_locks[KMALLOC_TRACK_HASH_SIZE];
+
+struct list_head kmalloc_addr_hash[KMALLOC_TRACK_HASH_SIZE];
+ihk_spinlock_t kmalloc_addr_hash_locks[KMALLOC_TRACK_HASH_SIZE];
+
+int kmalloc_track_initialized = 0;
+int kmalloc_runcount = 0;
+
+struct kmalloc_track_addr_entry {
+	void *addr;
 	int runcount;
+	struct list_head list; /* track_entry's list */
+	struct kmalloc_track_entry *entry;
+	struct list_head hash; /* address hash */
 };
 
-#define HASHNUM 129
+struct kmalloc_track_entry {
+	char *file;
+	int line;
+	int size;
+	ihk_atomic_t alloc_count;
+	struct list_head hash;
+	struct list_head addr_list;
+	ihk_spinlock_t addr_list_lock;
+};
 
-static struct alloc *allochash[HASHNUM];
-static struct location *lochash[HASHNUM];
-static ihk_spinlock_t alloclock;
-int runcount;
-static unsigned char *page;
-static int space;
-
-static void *dalloc(unsigned long size)
+void kmalloc_init(void)
 {
-	void *r;
-	static int pos = 0;
-	unsigned long irqstate;
+	struct cpu_local_var *v = get_this_cpu_local_var();
 
-	irqstate = ihk_mc_spinlock_lock(&alloclock);
-	size = (size + 7) & 0xfffffffffffffff8L;
-	if (pos + size > space) {
-		page = allocate_pages(1, IHK_MC_AP_NOWAIT);
-		space = 4096;
-		pos = 0;
+	register_kmalloc();
+
+	INIT_LIST_HEAD(&v->free_list);
+	INIT_LIST_HEAD(&v->remote_free_list);
+	ihk_mc_spinlock_init(&v->remote_free_list_lock);
+
+	v->kmalloc_initialized = 1;
+
+	if (!kmalloc_track_initialized) {
+		int i;
+
+		memdebug = find_command_line("memdebug");
+
+		kmalloc_track_initialized = 1;
+		for (i = 0; i < KMALLOC_TRACK_HASH_SIZE; ++i) {
+			ihk_mc_spinlock_init(&kmalloc_track_hash_locks[i]);
+			INIT_LIST_HEAD(&kmalloc_track_hash[i]);
+			ihk_mc_spinlock_init(&kmalloc_addr_hash_locks[i]);
+			INIT_LIST_HEAD(&kmalloc_addr_hash[i]);
+		}
 	}
-	r = page + pos;
-	pos += size;
-	ihk_mc_spinlock_unlock(&alloclock, irqstate);
-
-	return r;
 }
 
+/* NOTE: Hash lock must be held */
+struct kmalloc_track_entry *__kmalloc_track_find_entry(
+		int size, char *file, int line)
+{
+	struct kmalloc_track_entry *entry_iter, *entry = NULL;
+	int hash = (strlen(file) + line + size) & KMALLOC_TRACK_HASH_MASK;
+
+	list_for_each_entry(entry_iter, &kmalloc_track_hash[hash], hash) {
+		if (!strcmp(entry_iter->file, file) &&
+				entry_iter->size == size &&
+				entry_iter->line == line) {
+			entry = entry_iter;
+			break;
+		}
+	}
+
+	if (entry) {
+		dkprintf("%s found entry %s:%d size: %d\n", __FUNCTION__,
+				file, line, size);
+	}
+	else {
+		dkprintf("%s couldn't find entry %s:%d size: %d\n", __FUNCTION__,
+				file, line, size);
+	}
+
+	return entry;
+}
+
+/* Top level routines called from macro */
 void *_kmalloc(int size, enum ihk_mc_ap_flag flag, char *file, int line)
 {
-	char *r = ___kmalloc(size, flag);
-	struct malloc_header *h;
-	unsigned long hash;
-	char *t;
-	struct location *lp;
-	struct alloc *ap;
-	unsigned long alcsize;
-	unsigned long chksize;
+	unsigned long irqflags;
+	struct kmalloc_track_entry *entry;
+	struct kmalloc_track_addr_entry *addr_entry;
+	int hash, addr_hash;
+	void *r = ___kmalloc(size, flag);
 
 	if (!memdebug)
 		return r;
@@ -697,177 +751,177 @@ void *_kmalloc(int size, enum ihk_mc_ap_flag flag, char *file, int line)
 	if (!r)
 		return r;
 
-	h = ((struct malloc_header *)r) - 1;
-	alcsize = h->size * sizeof(struct malloc_header);
-	chksize = alcsize - size;
-	memset(r + size, '\x5a', chksize);
+	hash = (strlen(file) + line + size) & KMALLOC_TRACK_HASH_MASK;
+	irqflags = ihk_mc_spinlock_lock(&kmalloc_track_hash_locks[hash]);
 
-	for (hash = 0, t = file; *t; t++) {
-		hash <<= 1;
-		hash += *t;
+	entry = __kmalloc_track_find_entry(size, file, line);
+
+	if (!entry) {
+		entry = ___kmalloc(sizeof(*entry), IHK_MC_AP_NOWAIT);
+		if (!entry) {
+			kprintf("%s: ERROR: allocating tracking entry\n");
+			goto out;
+		}
+
+		entry->line = line;
+		entry->size = size;
+		ihk_atomic_set(&entry->alloc_count, 0);
+		ihk_mc_spinlock_init(&entry->addr_list_lock);
+		INIT_LIST_HEAD(&entry->addr_list);
+
+		entry->file = ___kmalloc(strlen(file) + 1, IHK_MC_AP_NOWAIT);
+		if (!entry->file) {
+			kprintf("%s: ERROR: allocating file string\n");
+			___kfree(entry);
+			ihk_mc_spinlock_unlock(&kmalloc_track_hash_locks[hash], irqflags);
+			goto out;
+		}
+
+		strcpy(entry->file, file);
+		entry->file[strlen(file)] = 0;
+		list_add(&entry->hash, &kmalloc_track_hash[hash]);
+		dkprintf("%s entry %s:%d size: %d added\n", __FUNCTION__,
+			file, line, size);
 	}
-	hash += line;
-	hash %= HASHNUM;
-	for (lp = lochash[hash]; lp; lp = lp->next)
-		if (lp->line == line &&
-		   !strcmp(lp->file, file))
-			break;
-	if (!lp) {
-		lp = dalloc(sizeof(struct location) + strlen(file) + 1);
-		memset(lp, '\0', sizeof(struct location));
-		lp->line = line;
-		strcpy(lp->file, file);
-		do {
-			lp->next = lochash[hash];
-		} while (!compare_and_swap(lochash + hash, (unsigned long)lp->next, (unsigned long)lp));
+	ihk_mc_spinlock_unlock(&kmalloc_track_hash_locks[hash], irqflags);
+
+	ihk_atomic_inc(&entry->alloc_count);
+
+	/* Add new addr entry for this allocation entry */
+	addr_entry = ___kmalloc(sizeof(*addr_entry), IHK_MC_AP_NOWAIT);
+	if (!addr_entry) {
+		kprintf("%s: ERROR: allocating addr entry\n");
+		goto out;
 	}
 
-	hash = (unsigned long)h % HASHNUM;
-	do {
-		for (ap = allochash[hash]; ap; ap = ap->next)
-			if (!ap->p)
-				break;
-	} while (ap && !compare_and_swap(&ap->p, 0UL, (unsigned long)h));
-	if (!ap) {
-		ap = dalloc(sizeof(struct alloc));
-		memset(ap, '\0', sizeof(struct alloc));
-		ap->p = h;
-		do {
-			ap->next = allochash[hash];
-		} while (!compare_and_swap(allochash + hash, (unsigned long)ap->next, (unsigned long)ap));
-	}
+	addr_entry->addr = r;
+	addr_entry->runcount = kmalloc_runcount;
+	addr_entry->entry = entry;
 
-	ap->loc = lp;
-	ap->size = size;
-	ap->runcount = runcount;
+	irqflags = ihk_mc_spinlock_lock(&entry->addr_list_lock);
+	list_add(&addr_entry->list, &entry->addr_list);
+	ihk_mc_spinlock_unlock(&entry->addr_list_lock, irqflags);
 
-	return r;
-}
+	/* Add addr entry to address hash */
+	addr_hash = ((unsigned long)r >> 5) & KMALLOC_TRACK_HASH_MASK;
+	irqflags = ihk_mc_spinlock_lock(&kmalloc_addr_hash_locks[addr_hash]);
+	list_add(&addr_entry->hash, &kmalloc_addr_hash[addr_hash]);
+	ihk_mc_spinlock_unlock(&kmalloc_addr_hash_locks[addr_hash], irqflags);
 
-int _memcheck(void *ptr, char *msg, char *file, int line, int flags)
-{
-	struct malloc_header *h = ((struct malloc_header *)ptr) - 1;
-	struct malloc_header *next;
-	unsigned long hash = (unsigned long)h % HASHNUM;
-	struct alloc *ap;
-	static unsigned long check = 0x5a5a5a5a5a5a5a5aUL;
-	unsigned long alcsize;
-	unsigned long chksize;
-
-
-	if (h->check != 0x5a5a5a5a) {
-		int i;
-		unsigned long max = 0;
-		unsigned long cur = (unsigned long)h;
-		struct alloc *maxap = NULL;
-
-		for (i = 0; i < HASHNUM; i++)
-			for (ap = allochash[i]; ap; ap = ap->next)
-				if ((unsigned long)ap->p < cur &&
-				   (unsigned long)ap->p > max) {
-					max = (unsigned long)ap->p;
-					maxap = ap;
-				}
-
-		kprintf("%s: detect buffer overrun, alc=%s:%d size=%ld h=%p, s=%ld\n", msg, maxap->loc->file, maxap->loc->line, maxap->size, maxap->p, maxap->p->size);
-		kprintf("broken header: h=%p next=%p size=%ld cpu_id=%d\n", h, h->next, h->size, h->cpu_id);
-	}
-
-	for (ap = allochash[hash]; ap; ap = ap->next)
-		if (ap->p == h)
-			break;
-	if (!ap) {
-		if(file)
-			kprintf("%s: address not found, %s:%d p=%p\n", msg, file, line, ptr);
-		else
-			kprintf("%s: address not found p=%p\n", msg, ptr);
-		return 1;
-	}
-
-	alcsize = h->size * sizeof(struct malloc_header);
-	chksize = alcsize - ap->size;
-	if (chksize > 8)
-		chksize = 8;
-	next = (struct malloc_header *)((char *)ptr + alcsize);
-
-	if (next->check != 0x5a5a5a5a ||
-	    memcmp((char *)ptr + ap->size, &check, chksize)) {
-		unsigned long buf = 0x5a5a5a5a5a5a5a5aUL;
-		unsigned char *p;
-		unsigned char *q;
-		memcpy(&buf, (char *)ptr + ap->size, chksize);
-		p = (unsigned char *)&(next->check);
-		q = (unsigned char *)&buf;
-
-		if (file)
-			kprintf("%s: broken, %s:%d alc=%s:%d %02x%02x%02x%02x%02x%02x%02x%02x %02x%02x%02x%02x size=%ld\n", msg, file, line, ap->loc->file, ap->loc->line, q[0], q[1], q[2], q[3], q[4], q[5], q[6], q[7], p[0], p[1], p[2], p[3], ap->size);
-		else
-			kprintf("%s: broken, alc=%s:%d %02x%02x%02x%02x%02x%02x%02x%02x %02x%02x%02x%02x size=%ld\n", msg, ap->loc->file, ap->loc->line, q[0], q[1], q[2], q[3], q[4], q[5], q[6], q[7], p[0], p[1], p[2], p[3], ap->size);
-
-
-		if (next->check != 0x5a5a5a5a)
-			kprintf("next->HEADER: next=%p size=%ld cpu_id=%d\n", next->next, next->size, next->cpu_id);
-
-		return 1;
-	}
-
-	if(flags & 1){
-		ap->p = NULL;
-		ap->loc = NULL;
-		ap->size = 0;
-	}
-	return 0;
-}
-
-int memcheckall()
-{
-	int i;
-	struct alloc *ap;
-	int r = 0;
-
-	for(i = 0; i < HASHNUM; i++)
-		for(ap = allochash[i]; ap; ap = ap->next)
-			if(ap->p)
-				r |= _memcheck(ap->p + 1, "memcheck", NULL, 0, 2);
-	return r;
-}
-
-int freecheck(int runcount)
-{
-	int i;
-	struct alloc *ap;
-	struct location *lp;
-	int r = 0;
-
-	for (i = 0; i < HASHNUM; i++)
-		for (lp = lochash[i]; lp; lp = lp->next)
-			lp->cnt = 0;
-
-	for (i = 0; i < HASHNUM; i++)
-		for (ap = allochash[i]; ap; ap = ap->next)
-			if (ap->p && ap->runcount == runcount) {
-				ap->loc->cnt++;
-				r++;
-			}
-
-	if (r) {
-		kprintf("memory leak?\n");
-		for (i = 0; i < HASHNUM; i++)
-			for (lp = lochash[i]; lp; lp = lp->next)
-				if (lp->cnt)
-					kprintf(" alc=%s:%d cnt=%d\n", lp->file, lp->line, lp->cnt);
-	}
+	dkprintf("%s addr_entry %p added\n", __FUNCTION__, r);
 
+out:
 	return r;
 }
 
 void _kfree(void *ptr, char *file, int line)
 {
-	if (memdebug)
-		_memcheck(ptr, "KFREE", file, line, 1);
+	unsigned long irqflags;
+	struct kmalloc_track_entry *entry;
+	struct kmalloc_track_addr_entry *addr_entry_iter, *addr_entry = NULL;
+	int hash;
+
+	if (!memdebug) {
+		goto out;
+	}
+
+	hash = ((unsigned long)ptr >> 5) & KMALLOC_TRACK_HASH_MASK;
+	irqflags = ihk_mc_spinlock_lock(&kmalloc_addr_hash_locks[hash]);
+	list_for_each_entry(addr_entry_iter,
+			&kmalloc_addr_hash[hash], hash) {
+		if (addr_entry_iter->addr == ptr) {
+			addr_entry = addr_entry_iter;
+			break;
+		}
+	}
+
+	if (addr_entry) {
+		list_del(&addr_entry->hash);
+	}
+	ihk_mc_spinlock_unlock(&kmalloc_addr_hash_locks[hash], irqflags);
+
+	if (!addr_entry) {
+		kprintf("%s: ERROR: kfree()ing invalid pointer\n", __FUNCTION__);
+		panic("panic");
+	}
+
+	entry = addr_entry->entry;
+
+	irqflags = ihk_mc_spinlock_lock(&entry->addr_list_lock);
+	list_del(&addr_entry->list);
+	ihk_mc_spinlock_unlock(&entry->addr_list_lock, irqflags);
+
+	dkprintf("%s addr_entry %p removed\n", __FUNCTION__, addr_entry->addr);
+	___kfree(addr_entry);
+
+	/* Do we need to remove tracking entry as well? */
+	if (!ihk_atomic_dec_and_test(&entry->alloc_count)) {
+		goto out;
+	}
+
+	hash = (strlen(entry->file) + entry->line + entry->size) &
+		KMALLOC_TRACK_HASH_MASK;
+	irqflags = ihk_mc_spinlock_lock(&kmalloc_track_hash_locks[hash]);
+	list_del(&entry->hash);
+	ihk_mc_spinlock_unlock(&kmalloc_track_hash_locks[hash], irqflags);
+
+	dkprintf("%s entry %s:%d size: %d removed\n", __FUNCTION__,
+			entry->file, entry->line, entry->size);
+	___kfree(entry->file);
+	___kfree(entry);
+
+out:
 	___kfree(ptr);
 }
 
+void kmalloc_memcheck(void)
+{
+	int i;
+	unsigned long irqflags;
+	struct kmalloc_track_entry *entry = NULL;
+
+	for (i = 0; i < KMALLOC_TRACK_HASH_SIZE; ++i) {
+		irqflags = ihk_mc_spinlock_lock(&kmalloc_track_hash_locks[i]);
+		list_for_each_entry(entry, &kmalloc_track_hash[i], hash) {
+			struct kmalloc_track_addr_entry *addr_entry = NULL;
+			int cnt = 0;
+
+			ihk_mc_spinlock_lock_noirq(&entry->addr_list_lock);
+			list_for_each_entry(addr_entry, &entry->addr_list, list) {
+
+			dkprintf("%s memory leak: %p @ %s:%d size: %d runcount: %d\n",
+				__FUNCTION__,
+				addr_entry->addr,
+				entry->file,
+				entry->line,
+				entry->size,
+				addr_entry->runcount);
+
+				if (kmalloc_runcount != addr_entry->runcount)
+					continue;
+
+				cnt++;
+			}
+			ihk_mc_spinlock_unlock_noirq(&entry->addr_list_lock);
+
+			if (!cnt)
+				continue;
+
+			kprintf("%s memory leak: %s:%d size: %d cnt: %d, runcount: %d\n",
+				__FUNCTION__,
+				entry->file,
+				entry->line,
+				entry->size,
+				cnt,
+				kmalloc_runcount);
+		}
+		ihk_mc_spinlock_unlock(&kmalloc_track_hash_locks[i], irqflags);
+	}
+
+	++kmalloc_runcount;
+}
+
+/* Redirection routines registered in alloc structure */
 void *__kmalloc(int size, enum ihk_mc_ap_flag flag)
 {
 	return kmalloc(size, flag);
@@ -878,160 +932,199 @@ void __kfree(void *ptr)
 	kfree(ptr);
 }
 
-void kmalloc_init(void)
+
+static void ___kmalloc_insert_chunk(struct list_head *free_list,
+	struct kmalloc_header *chunk)
 {
-	struct cpu_local_var *v = get_this_cpu_local_var();
-	struct malloc_header *h = &v->free_list;
-	int i;
+	struct kmalloc_header *chunk_iter, *next_chunk = NULL;
 
-	h->check = 0x5a5a5a5a;
-	h->next = &v->free_list;
-	h->size = 0;
-
-	register_kmalloc();
-
-	memdebug = find_command_line("memdebug");
-	for (i = 0; i < HASHNUM; i++) {
-		allochash[i] = NULL;
-		lochash[i] = NULL;
-	}
-	page = allocate_pages(16, IHK_MC_AP_NOWAIT);
-	space = 16 * 4096;
-	ihk_mc_spinlock_init(&alloclock);
-}
-
-void ____kfree(struct cpu_local_var *v, struct malloc_header *p)
-{
-	struct malloc_header *h = &v->free_list;
-	int combined = 0;
-
-	h = h->next;
-
-	while ((p < h || p > h->next) && h != &v->free_list) {
-		h = h->next;
-	}
-
-	if (h + h->size + 1 == p && h->size != 0) {
-		combined = 1;
-		h->size += p->size + 1;
-		h->check = 0x5a5a5a5a;
-	}
-	if (h->next == p + p->size + 1 && h->next->size != 0) {
-		if (combined) {
-			h->check = 0x5a5a5a5a;
-			h->size += h->next->size + 1;
-			h->next = h->next->next;
-		} else { 
-			p->check = 0x5a5a5a5a;
-			p->size += h->next->size + 1;
-			p->next = h->next->next;
-			h->next = p;
+	/* Find out where to insert */
+	list_for_each_entry(chunk_iter, free_list, list) {
+		if ((void *)chunk < (void *)chunk_iter) {
+			next_chunk = chunk_iter;
+			break;
 		}
-	} else if (!combined) {
-		p->next = h->next;
-		h->next = p;
 	}
+
+	/* Add in front of next */
+	if (next_chunk) {
+		list_add_tail(&chunk->list, &next_chunk->list);
+	}
+	/* Add after the head */
+	else {
+		list_add(&chunk->list, free_list);
+	}
+
+	return;
 }
 
-void *___kmalloc(int size, enum ihk_mc_ap_flag flag)
+static void ___kmalloc_init_chunk(struct kmalloc_header *h, int size)
 {
-	struct cpu_local_var *v = get_this_cpu_local_var();
-	struct malloc_header *h = &v->free_list, *prev, *p;
-	int u, req_page;
+	h->size = size;
+	h->front_magic = 0x5c5c5c5c;
+	h->end_magic = 0x6d6d6d6d;
+	h->cpu_id = ihk_mc_get_processor_id();
+}
 
-	p = (struct malloc_header *)xchg8((unsigned long *)&v->remote_free_list, 0L);
-	while(p){
-		struct malloc_header *n = p->next;
-		____kfree(v, p);
-		p = n; 
+static void ___kmalloc_consolidate_list(struct list_head *list)
+{
+	struct kmalloc_header *chunk_iter, *chunk, *next_chunk;
+
+reiterate:
+	chunk_iter = NULL;
+	chunk = NULL;
+
+	list_for_each_entry(next_chunk, list, list) {
+
+		if (chunk_iter && (((void *)chunk_iter + sizeof(struct kmalloc_header)
+						+ chunk_iter->size) == (void *)next_chunk)) {
+			chunk = chunk_iter;
+			break;
+		}
+
+		chunk_iter = next_chunk;
 	}
 
-	if (size >= PAGE_SIZE * 4) {
+	if (!chunk) {
+		return;
+	}
+
+	chunk->size += (next_chunk->size + sizeof(struct kmalloc_header));
+	list_del(&next_chunk->list);
+	goto reiterate;
+}
+
+
+void kmalloc_consolidate_free_list(void)
+{
+	struct kmalloc_header *chunk, *tmp;
+	unsigned long irqflags =
+		ihk_mc_spinlock_lock(&cpu_local_var(remote_free_list_lock));
+
+	/* Clean up remotely deallocated chunks */
+	list_for_each_entry_safe(chunk, tmp,
+			&cpu_local_var(remote_free_list), list) {
+
+		list_del(&chunk->list);
+		___kmalloc_insert_chunk(&cpu_local_var(free_list), chunk);
+	}
+
+	/* Free list lock ensures IRQs are disabled */
+	___kmalloc_consolidate_list(&cpu_local_var(free_list));
+
+	ihk_mc_spinlock_unlock(&cpu_local_var(remote_free_list_lock), irqflags);
+}
+
+#define KMALLOC_MIN_SHIFT   (5)
+#define KMALLOC_MIN_SIZE    (1 << KMALLOC_TRACK_HASH_SHIFT)
+#define KMALLOC_MIN_MASK    (KMALLOC_MIN_SIZE - 1)
+
+/* Actual low-level allocation routines */
+static void *___kmalloc(int size, enum ihk_mc_ap_flag flag)
+{
+	struct kmalloc_header *chunk_iter;
+	struct kmalloc_header *chunk = NULL;
+	int npages;
+	unsigned long kmalloc_irq_flags = cpu_disable_interrupt_save();
+
+	/* KMALLOC_MIN_SIZE bytes aligned size. */
+	if (size & KMALLOC_MIN_MASK) {
+		size = ((size + KMALLOC_MIN_SIZE - 1) & ~(KMALLOC_MIN_MASK));
+	}
+
+	chunk = NULL;
+	/* Find a chunk that is big enough */
+	list_for_each_entry(chunk_iter, &cpu_local_var(free_list), list) {
+		if (chunk_iter->size >= size) {
+			chunk = chunk_iter;
+			break;
+		}
+	}
+
+split_and_return:
+	/* Did we find one? */
+	if (chunk) {
+		/* Do we need to split it? Only if there is enough space for
+		 * another header and some actual content */
+		if (chunk->size > (size + sizeof(struct kmalloc_header))) {
+			struct kmalloc_header *leftover;
+
+			leftover = (struct kmalloc_header *)
+				((void *)chunk + sizeof(struct kmalloc_header) + size);
+			___kmalloc_init_chunk(leftover,
+				(chunk->size - size - sizeof(struct kmalloc_header)));
+			list_add(&leftover->list, &chunk->list);
+			chunk->size = size;
+		}
+
+		list_del(&chunk->list);
+		cpu_restore_interrupt(kmalloc_irq_flags);
+		return ((void *)chunk + sizeof(struct kmalloc_header));
+	}
+
+
+	/* Allocate new memory and add it to free list */
+	npages = (size + sizeof(struct kmalloc_header) + (PAGE_SIZE - 1))
+		>> PAGE_SHIFT;
+	chunk = ihk_mc_alloc_pages(npages, flag);
+
+	if (!chunk) {
+		cpu_restore_interrupt(kmalloc_irq_flags);
 		return NULL;
 	}
 
-	u = (size + sizeof(*h) - 1) / sizeof(*h);
+	___kmalloc_init_chunk(chunk,
+			(npages * PAGE_SIZE - sizeof(struct kmalloc_header)));
+	___kmalloc_insert_chunk(&cpu_local_var(free_list), chunk);
 
-	prev = h;
-	h = h->next;
-
-	while (1) {
-		if (h == &v->free_list) {
-			req_page = ((u + 2) * sizeof(*h) + PAGE_SIZE - 1)
-				>> PAGE_SHIFT;
-
-			h = allocate_pages(req_page, flag);
-			if(h == NULL) {
-				kprintf("kmalloc(%#x,%#x): out of memory\n", size, flag);
-				return NULL;
-			}
-			h->check = 0x5a5a5a5a;
-			prev->next = h;
-			h->size = (req_page * PAGE_SIZE) / sizeof(*h) - 2;
-			/* Guard entry */
-			p = h + h->size + 1;
-			p->check = 0x5a5a5a5a;
-			p->next = &v->free_list;
-			p->size = 0;
-			h->next = p;
-		}
-
-		if (h->size >= u) {
-			if (h->size == u || h->size == u + 1) {
-				prev->next = h->next;
-				h->cpu_id = ihk_mc_get_processor_id();
-
-				return h + 1;
-			} else { /* Divide */
-				h->size -= u + 1;
-				
-				p = h + h->size + 1;
-				p->check = 0x5a5a5a5a;
-				p->size = u;
-				p->cpu_id = ihk_mc_get_processor_id();
-
-				return p + 1;
-			}
-		}
-		prev = h;
-		h = h->next;
-	}
+	goto split_and_return;
 }
 
-void ___kfree(void *ptr)
+static void ___kfree(void *ptr)
 {
-	struct malloc_header *p = (struct malloc_header *)ptr;
-	struct cpu_local_var *v = get_cpu_local_var((--p)->cpu_id);
+	struct kmalloc_header *chunk =
+		(struct kmalloc_header*)(ptr - sizeof(struct kmalloc_header));
+	unsigned long kmalloc_irq_flags = cpu_disable_interrupt_save();
 
-	if(p->cpu_id == ihk_mc_get_processor_id()){
-		____kfree(v, p);
+	/* Sanity check */
+	if (chunk->front_magic != 0x5c5c5c5c || chunk->end_magic != 0x6d6d6d6d) {
+		kprintf("%s: memory corruption at address 0x%p\n", __FUNCTION__, ptr);
+		panic("panic");
 	}
-	else{
-		unsigned long oldval;
-		unsigned long newval;
-		unsigned long rval;
-		do{
-			p->next = v->remote_free_list;
-			oldval = (unsigned long)p->next;
-			newval = (unsigned long)p;
-			rval = atomic_cmpxchg8(
-			          (unsigned long *)&v->remote_free_list,
-			          oldval, newval);
-		}while(rval != oldval);
+
+	/* Does this chunk belong to this CPU? */
+	if (chunk->cpu_id == ihk_mc_get_processor_id()) {
+
+		___kmalloc_insert_chunk(&cpu_local_var(free_list), chunk);
+		___kmalloc_consolidate_list(&cpu_local_var(free_list));
 	}
+	else {
+		struct cpu_local_var *v = get_cpu_local_var(chunk->cpu_id);
+		unsigned long irqflags;
+
+		irqflags = ihk_mc_spinlock_lock(&v->remote_free_list_lock);
+		list_add(&chunk->list, &v->remote_free_list);
+		ihk_mc_spinlock_unlock(&v->remote_free_list_lock, irqflags);
+	}
+
+	cpu_restore_interrupt(kmalloc_irq_flags);
 }
 
-void print_free_list(void)
+
+void ___kmalloc_print_free_list(struct list_head *list)
 {
-	struct cpu_local_var *v = get_this_cpu_local_var();
-	struct malloc_header *h = &v->free_list;
+	struct kmalloc_header *chunk_iter;
+	unsigned long irqflags = kprintf_lock();
 
-	h = h->next;
-
-	kprintf("free_list : \n");
-	while (h != &v->free_list) {
-		kprintf("  %p : %p, %d ->\n", h, h->next, h->size);
-		h = h->next;
+	__kprintf("%s: [ \n", __FUNCTION__);
+	list_for_each_entry(chunk_iter, &cpu_local_var(free_list), list) {
+		__kprintf("%s: 0x%lx:%d (VA PFN: %lu, off: %lu)\n", __FUNCTION__,
+			(unsigned long)chunk_iter,
+			chunk_iter->size,
+			(unsigned long)chunk_iter >> PAGE_SHIFT,
+			(unsigned long)chunk_iter % PAGE_SIZE);
 	}
-	kprintf("\n");
+	__kprintf("%s: ] \n", __FUNCTION__);
+	kprintf_unlock(irqflags);
 }
+

kernel/process.c

@@ -53,7 +53,6 @@ static int copy_user_ranges(struct process_vm *vm, struct process_vm *orgvm);
 extern void release_fp_regs(struct thread *proc);
 extern void save_fp_regs(struct thread *proc);
 extern void restore_fp_regs(struct thread *proc);
-void settid(struct thread *proc, int mode, int newcpuid, int oldcpuid);
 extern void __runq_add_proc(struct thread *proc, int cpu_id);
 extern void terminate_host(int pid);
 extern void lapic_timer_enable(unsigned int clocks);
@@ -338,6 +337,10 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 		proc = org->proc;
 		thread->vm = org->vm;
 		thread->proc = proc;
+
+		thread->sigstack.ss_sp = NULL;
+		thread->sigstack.ss_flags = SS_DISABLE;
+		thread->sigstack.ss_size = 0;
 	}
 	/* fork() */
 	else {
@@ -383,9 +386,15 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 			goto err_free_proc;
 		}
 
+		thread->vm->vdso_addr = org->vm->vdso_addr;
+		thread->vm->vvar_addr = org->vm->vvar_addr;
 		thread->proc->maxrss = org->proc->maxrss;
 		thread->vm->currss = org->vm->currss;
 
+		thread->sigstack.ss_sp = org->sigstack.ss_sp;
+		thread->sigstack.ss_flags = org->sigstack.ss_flags;
+		thread->sigstack.ss_size = org->sigstack.ss_size;
+
 		dkprintf("fork(): copy_user_ranges() OK\n");
 	}
 
@@ -413,9 +422,6 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 		INIT_LIST_HEAD(&thread->sigcommon->sigpending);
 		// TODO: copy signalfd
 	}
-	thread->sigstack.ss_sp = NULL;
-	thread->sigstack.ss_flags = SS_DISABLE;
-	thread->sigstack.ss_size = 0;
 	ihk_mc_spinlock_init(&thread->sigpendinglock);
 	INIT_LIST_HEAD(&thread->sigpending);
 	thread->sigmask = org->sigmask;
@@ -566,6 +572,9 @@ static int copy_user_ranges(struct process_vm *vm, struct process_vm *orgvm)
 			break;
 		}
 
+		if(src_range->flag & VR_DONTFORK)
+			continue;
+
 		range = kmalloc(sizeof(struct vm_range), IHK_MC_AP_NOWAIT);
 		if (!range) {
 			goto err_rollback;
@@ -735,7 +744,7 @@ int join_process_memory_range(struct process_vm *vm,
 		memobj_release(merging->memobj);
 	}
 	list_del(&merging->list);
-	ihk_mc_free(merging);
+	kfree(merging);
 
 	error = 0;
 out:
@@ -831,8 +840,9 @@ int free_process_memory_range(struct process_vm *vm, struct vm_range *range)
 	if (range->memobj) {
 		memobj_release(range->memobj);
 	}
+
 	list_del(&range->list);
-	ihk_mc_free(range);
+	kfree(range);
 
 	dkprintf("free_process_memory_range(%p,%lx-%lx): 0\n",
 			vm, start0, end0);
@@ -958,7 +968,6 @@ enum ihk_mc_pt_attribute common_vrflag_to_ptattr(unsigned long flag, uint64_t fa
 	return attr;
 }
 
-/* XXX: インデントを揃える必要がある */
 int add_process_memory_range(struct process_vm *vm,
                              unsigned long start, unsigned long end,
                              unsigned long phys, unsigned long flag,
@@ -1529,6 +1538,8 @@ retry:
 				kprintf("page_fault_process_memory_range(%p,%lx-%lx %lx,%lx,%lx):cannot allocate new page. %d\n", vm, range->start, range->end, range->flag, fault_addr, reason, error);
 				goto out;
 			}
+			dkprintf("%s: clearing 0x%lx:%lu\n",
+					__FUNCTION__, pgaddr, pgsize);
 			memset(virt, 0, pgsize);
 			phys = virt_to_phys(virt);
 			page_map(phys_to_page(phys));
@@ -1561,6 +1572,8 @@ retry:
 				kprintf("page_fault_process_memory_range(%p,%lx-%lx %lx,%lx,%lx):cannot allocate copy page. %d\n", vm, range->start, range->end, range->flag, fault_addr, reason, error);
 				goto out;
 			}
+			dkprintf("%s: copying 0x%lx:%lu\n",
+				__FUNCTION__, pgaddr, pgsize);
 			memcpy(virt, phys_to_virt(phys), pgsize);
 
 			phys = virt_to_phys(virt);
@@ -1641,6 +1654,18 @@ static int do_page_fault_process_vm(struct process_vm *vm, void *fault_addr0, ui
 				"access denied. %d\n",
 				ihk_mc_get_processor_id(), vm,
 				fault_addr0, reason, error);
+		kprintf("%s: reason: %s%s%s%s%s%s%s\n", __FUNCTION__,
+			(reason & PF_PROT) ? "PF_PROT " : "",
+			(reason & PF_WRITE) ? "PF_WRITE " : "",
+			(reason & PF_USER) ? "PF_USER " : "",
+			(reason & PF_RSVD) ? "PF_RSVD " : "",
+			(reason & PF_INSTR) ? "PF_INSTR " : "",
+			(reason & PF_PATCH) ? "PF_PATCH " : "",
+			(reason & PF_POPULATE) ? "PF_POPULATE " : "");
+		kprintf("%s: range->flag & (%s%s%s)\n", __FUNCTION__,
+			(range->flag & VR_PROT_READ) ? "VR_PROT_READ " : "",
+			(range->flag & VR_PROT_WRITE) ? "VR_PROT_WRITE " : "",
+			(range->flag & VR_PROT_EXEC) ? "VR_PROT_EXEC " : "");
 		if (((range->flag & VR_PROT_MASK) == VR_PROT_NONE))
 			kprintf("if (((range->flag & VR_PROT_MASK) == VR_PROT_NONE))\n");
 		if (((reason & PF_WRITE) && !(reason & PF_PATCH)))
@@ -1868,14 +1893,14 @@ unsigned long extend_process_region(struct process_vm *vm,
 			aligned_end = (aligned_end + (LARGE_PAGE_SIZE - 1)) & LARGE_PAGE_MASK;
 			/* Fill in the gap between old_aligned_end and aligned_end
 			 * with regular pages */
-			if((p = allocate_pages((aligned_end - old_aligned_end) >> PAGE_SHIFT,
+			if((p = ihk_mc_alloc_pages((aligned_end - old_aligned_end) >> PAGE_SHIFT,
                                  IHK_MC_AP_NOWAIT)) == NULL){
 				return end;
 			}
 			if((rc = add_process_memory_range(vm, old_aligned_end,
                                         aligned_end, virt_to_phys(p), flag,
 					LARGE_PAGE_SHIFT)) != 0){
-				free_pages(p, (aligned_end - old_aligned_end) >> PAGE_SHIFT);
+				ihk_mc_free_pages(p, (aligned_end - old_aligned_end) >> PAGE_SHIFT);
 				return end;
 			}
 
@@ -1888,7 +1913,7 @@ unsigned long extend_process_region(struct process_vm *vm,
 				(LARGE_PAGE_SIZE - 1)) & LARGE_PAGE_MASK;
 		address = aligned_new_end;
 
-		if((p = allocate_pages((aligned_new_end - aligned_end + LARGE_PAGE_SIZE) >> PAGE_SHIFT,
+		if((p = ihk_mc_alloc_pages((aligned_new_end - aligned_end + LARGE_PAGE_SIZE) >> PAGE_SHIFT,
                             IHK_MC_AP_NOWAIT)) == NULL){
 			return end;
 		}
@@ -1896,16 +1921,16 @@ unsigned long extend_process_region(struct process_vm *vm,
 		p_aligned = ((unsigned long)p + (LARGE_PAGE_SIZE - 1)) & LARGE_PAGE_MASK;
 
 		if (p_aligned > (unsigned long)p) {
-			free_pages(p, (p_aligned - (unsigned long)p) >> PAGE_SHIFT);
+			ihk_mc_free_pages(p, (p_aligned - (unsigned long)p) >> PAGE_SHIFT);
 		}
-		free_pages(
+		ihk_mc_free_pages(
 			(void *)(p_aligned + aligned_new_end - aligned_end),
 			(LARGE_PAGE_SIZE - (p_aligned - (unsigned long)p)) >> PAGE_SHIFT);
 
 		if((rc = add_process_memory_range(vm, aligned_end,
                                aligned_new_end, virt_to_phys((void *)p_aligned),
                                flag, LARGE_PAGE_SHIFT)) != 0){
-			free_pages(p, (aligned_new_end - aligned_end + LARGE_PAGE_SIZE) >> PAGE_SHIFT);
+			ihk_mc_free_pages(p, (aligned_new_end - aligned_end + LARGE_PAGE_SIZE) >> PAGE_SHIFT);
 			return end;
 		}
 
@@ -1923,7 +1948,7 @@ unsigned long extend_process_region(struct process_vm *vm,
 	  p=0;
 	}else{
 
-	p = allocate_pages((aligned_new_end - aligned_end) >> PAGE_SHIFT, IHK_MC_AP_NOWAIT);
+	p = ihk_mc_alloc_pages((aligned_new_end - aligned_end) >> PAGE_SHIFT, IHK_MC_AP_NOWAIT);
 
 	if (!p) {
 		return end;
@@ -1932,7 +1957,7 @@ unsigned long extend_process_region(struct process_vm *vm,
 	if((rc = add_process_memory_range(vm, aligned_end, aligned_new_end,
                                       (p==0?0:virt_to_phys(p)), flag, NULL, 0,
 				      PAGE_SHIFT)) != 0){
-		free_pages(p, (aligned_new_end - aligned_end) >> PAGE_SHIFT);
+		ihk_mc_free_pages(p, (aligned_new_end - aligned_end) >> PAGE_SHIFT);
 		return end;
 	}
 
@@ -2045,6 +2070,7 @@ release_process(struct process *proc)
 		mcs_rwlock_writer_unlock(&parent->children_lock, &lock);
 	}
 
+	if (proc->tids) kfree(proc->tids);
 	kfree(proc);
 }
 
@@ -2150,6 +2176,23 @@ release_sigcommon(struct sig_common *sigcommon)
 	kfree(sigcommon);
 }
 
+/*
+ * Release the TID from the process' TID set corresponding to this thread.
+ * NOTE: threads_lock must be held.
+ */
+void __release_tid(struct process *proc, struct thread *thread) {
+	int i;
+
+	for (i = 0; i < proc->nr_tids; ++i) {
+		if (proc->tids[i].thread != thread) continue;
+
+		proc->tids[i].thread = NULL;
+		dkprintf("%s: tid %d has been released by %p\n",
+			__FUNCTION__, thread->tid, thread);
+		break;
+	}
+}
+
 void destroy_thread(struct thread *thread)
 {
 	struct sig_pending *pending;
@@ -2166,6 +2209,7 @@ void destroy_thread(struct thread *thread)
 
 	mcs_rwlock_writer_lock(&proc->threads_lock, &lock);
 	list_del(&thread->siblings_list);
+	__release_tid(proc, thread);
 	mcs_rwlock_writer_unlock(&proc->threads_lock, &lock);
 
 	cpu_clear(thread->cpu_id, &thread->vm->address_space->cpu_set,
@@ -2303,6 +2347,8 @@ static void idle(void)
 		}
 		if (v->status == CPU_STATUS_IDLE ||
 		    v->status == CPU_STATUS_RESERVED) {
+			/* No work to do? Consolidate the kmalloc free list */
+			kmalloc_consolidate_free_list();
 			cpu_safe_halt();
 		}
 		else {
@@ -2491,7 +2537,6 @@ static void do_migrate(void)
 		cur_v->runq_len -= 1;
 		old_cpu_id = req->thread->cpu_id;
 		req->thread->cpu_id = cpu_id;
-		settid(req->thread, 2, cpu_id, old_cpu_id);
 		list_add_tail(&req->thread->sched_list, &v->runq);
 		v->runq_len += 1;
 		
@@ -2506,6 +2551,7 @@ static void do_migrate(void)
 		v->flags |= CPU_FLAG_NEED_RESCHED;
 		ihk_mc_interrupt_cpu(get_x86_cpu_local_variable(cpu_id)->apic_id, 0xd1);
 		double_rq_unlock(cur_v, v, irqstate);
+		//settid(req->thread, 2, cpu_id, old_cpu_id, 0, NULL);
 
 ack:
 		waitq_wakeup(&req->wq);
@@ -2541,13 +2587,8 @@ void schedule(void)
 	struct thread *last;
 
 	if (cpu_local_var(no_preempt)) {
-		dkprintf("no schedule() while no preemption! \n");
-		return;
-	}
-
-	if (cpu_local_var(current)
-			&& cpu_local_var(current)->in_syscall_offload) {
-		dkprintf("no schedule() while syscall offload!\n");
+		kprintf("%s: WARNING can't schedule() while no preemption, cnt: %d\n",
+			__FUNCTION__, cpu_local_var(no_preempt));
 		return;
 	}
 
@@ -2576,9 +2617,10 @@ redo:
 	if (v->flags & CPU_FLAG_NEED_MIGRATE) {
 		next = &cpu_local_var(idle);
 	} else {
-		/* Pick a new running process */
+		/* Pick a new running process or one that has a pending signal */
 		list_for_each_entry_safe(thread, tmp, &(v->runq), sched_list) {
-			if (thread->status == PS_RUNNING) {
+			if (thread->status == PS_RUNNING ||
+				(thread->status == PS_INTERRUPTIBLE && hassigpending(thread))) {
 				next = thread;
 				break;
 			}
@@ -2704,9 +2746,11 @@ sched_wakeup_thread(struct thread *thread, int valid_states)
 	int spin_slept = 0;
 	unsigned long irqstate;
 	struct cpu_local_var *v = get_cpu_local_var(thread->cpu_id);
+	struct process *proc = thread->proc;
+	struct mcs_rwlock_node updatelock;
 
 	dkprintf("sched_wakeup_process,proc->pid=%d,valid_states=%08x,proc->status=%08x,proc->cpu_id=%d,my cpu_id=%d\n",
-			 thread->proc->pid, valid_states, thread->status, thread->cpu_id, ihk_mc_get_processor_id());
+			 proc->pid, valid_states, thread->status, thread->cpu_id, ihk_mc_get_processor_id());
 
 	irqstate = ihk_mc_spinlock_lock(&(thread->spin_sleep_lock));
 	if (thread->spin_sleep > 0) {
@@ -2726,7 +2770,10 @@ sched_wakeup_thread(struct thread *thread, int valid_states)
 	irqstate = ihk_mc_spinlock_lock(&(v->runq_lock));
 
 	if (thread->status & valid_states) {
-		xchg4((int *)(&thread->proc->status), PS_RUNNING);
+		mcs_rwlock_writer_lock_noirq(&proc->update_lock, &updatelock);
+		if(proc->status != PS_EXITED)
+			proc->status = PS_RUNNING;
+		mcs_rwlock_writer_unlock_noirq(&proc->update_lock, &updatelock);
 		xchg4((int *)(&thread->status), PS_RUNNING);
 		status = 0;
 	}

kernel/procfs.c

@@ -281,6 +281,13 @@ 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);

kernel/syscall.c

@@ -105,7 +105,6 @@ static void calculate_time_from_tsc(struct timespec *ts);
 void check_signal(unsigned long, void *, int);
 void do_signal(long rc, void *regs, struct thread *thread, struct sig_pending *pending, int num);
 extern unsigned long do_kill(struct thread *thread, int pid, int tid, int sig, struct siginfo *info, int ptracecont);
-extern struct sigpending *hassigpending(struct thread *thread);
 extern long alloc_debugreg(struct thread *thread);
 extern int num_processors;
 extern unsigned long ihk_mc_get_ns_per_tsc(void);
@@ -128,11 +127,9 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 static void do_mod_exit(int status);
 #endif
 
-static void send_syscall(struct syscall_request *req, int cpu, int pid)
+static void send_syscall(struct syscall_request *req, int cpu, int pid, struct syscall_response *res)
 {
-	struct ikc_scd_packet packet;
-	struct syscall_response *res;
-	struct syscall_params *scp;
+	struct ikc_scd_packet packet IHK_DMA_ALIGN;
 	struct ihk_ikc_channel_desc *syscall_channel;
 	int ret;
 
@@ -141,7 +138,6 @@ static void send_syscall(struct syscall_request *req, int cpu, int pid)
 	   req->number == __NR_kill){ // interrupt syscall
 		extern int num_processors;
 
-		scp = &get_cpu_local_var(0)->scp2;
 		syscall_channel = get_cpu_local_var(0)->syscall_channel2;
 		
 		/* XXX: is this really going to work if multiple processes 
@@ -153,34 +149,22 @@ static void send_syscall(struct syscall_request *req, int cpu, int pid)
 			pid = req->args[1];
 	}
 	else{
-		scp = &get_cpu_local_var(cpu)->scp;
 		syscall_channel = get_cpu_local_var(cpu)->syscall_channel;
 	}
-	res = scp->response_va;
 
 	res->status = 0;
 	req->valid = 0;
 
-#ifdef USE_DMA
-	memcpy_async(scp->request_pa,
-	             virt_to_phys(req), sizeof(*req), 0, &fin);
-
-	memcpy_async_wait(&scp->post_fin);
-	scp->post_va->v[0] = scp->post_idx;
-	memcpy_async_wait(&fin);
-#else
-	memcpy(scp->request_va, req, sizeof(*req));
-#endif
+	memcpy(&packet.req, req, sizeof(*req));
 
 	barrier();
-	scp->request_va->valid = 1;
-	*(unsigned int *)scp->doorbell_va = cpu + 1;
+	packet.req.valid = 1;
 
 #ifdef SYSCALL_BY_IKC
 	packet.msg = SCD_MSG_SYSCALL_ONESIDE;
 	packet.ref = cpu;
 	packet.pid = pid ? pid : cpu_local_var(current)->proc->pid;
-	packet.arg = scp->request_rpa;	
+	packet.resp_pa = virt_to_phys(res);
 	dkprintf("send syscall, nr: %d, pid: %d\n", req->number, packet.pid);
 
 	ret = ihk_ikc_send(syscall_channel, &packet, 0);
@@ -194,9 +178,8 @@ ihk_spinlock_t syscall_lock;
 
 long do_syscall(struct syscall_request *req, int cpu, int pid)
 {
-	struct syscall_response *res;
+	struct syscall_response res;
 	struct syscall_request req2 IHK_DMA_ALIGN;
-	struct syscall_params *scp;
 	int error;
 	long rc;
 	int islock = 0;
@@ -207,6 +190,9 @@ long do_syscall(struct syscall_request *req, int cpu, int pid)
 	dkprintf("SC(%d)[%3d] sending syscall\n",
 		ihk_mc_get_processor_id(),
 		req->number);
+	
+	irqstate = 0;	/* for avoidance of warning */
+	barrier();
 
 	if(req->number != __NR_exit_group){
 		if(proc->nohost && // host is down
@@ -216,55 +202,102 @@ long do_syscall(struct syscall_request *req, int cpu, int pid)
 		++thread->in_syscall_offload;
 	}
 
-	irqstate = 0;	/* for avoidance of warning */
 	if(req->number == __NR_exit_group ||
 	   req->number == __NR_gettid ||
 	   req->number == __NR_kill){ // interrupt syscall
-		scp = &get_cpu_local_var(0)->scp2;
 		islock = 1;
 		irqstate = ihk_mc_spinlock_lock(&syscall_lock);
 	}
-	else{
-		scp = &get_cpu_local_var(cpu)->scp;
-	}
-	res = scp->response_va;
+	/* The current thread is the requester and any thread from 
+	 * the pool may serve the request */
+	req->rtid = cpu_local_var(current)->tid;
+	req->ttid = 0;
+	res.req_thread_status = IHK_SCD_REQ_THREAD_SPINNING;
+	send_syscall(req, cpu, pid, &res);
 
-	send_syscall(req, cpu, pid);
-
-	dkprintf("SC(%d)[%3d] waiting for host.. \n", 
-	        ihk_mc_get_processor_id(),
-	        req->number);
+	dkprintf("%s: syscall num: %d waiting for Linux.. \n",
+		__FUNCTION__, req->number);
 	
 #define	STATUS_IN_PROGRESS	0
 #define	STATUS_COMPLETED	1
 #define	STATUS_PAGE_FAULT	3
-	while (res->status != STATUS_COMPLETED) {
-		while (res->status == STATUS_IN_PROGRESS) {
+	while (res.status != STATUS_COMPLETED) {
+		while (res.status == STATUS_IN_PROGRESS) {
+			struct cpu_local_var *v;
+			int do_schedule = 0;
+			long runq_irqstate;
+			unsigned long flags;
+			DECLARE_WAITQ_ENTRY(scd_wq_entry, cpu_local_var(current));
+
 			cpu_pause();
+
+			/* Spin if not preemptable */
+			if (cpu_local_var(no_preempt) || !thread->tid) {
+				continue;
+			}
+
+			/* Spin by default, but if re-schedule is requested let
+			 * the other thread run */
+			runq_irqstate =
+				ihk_mc_spinlock_lock(&(get_this_cpu_local_var()->runq_lock));
+			v = get_this_cpu_local_var();
+
+			if (v->flags & CPU_FLAG_NEED_RESCHED) {
+				do_schedule = 1;
+			}
+
+			ihk_mc_spinlock_unlock(&v->runq_lock, runq_irqstate);
+
+			if (!do_schedule) {
+				continue;
+			}
+
+			flags = cpu_disable_interrupt_save();
+
+			/* Try to sleep until notified */
+			if (__sync_bool_compare_and_swap(&res.req_thread_status,
+				IHK_SCD_REQ_THREAD_SPINNING,
+				IHK_SCD_REQ_THREAD_DESCHEDULED)) {
+
+				dkprintf("%s: tid %d waiting for syscall reply...\n",
+						__FUNCTION__, thread->tid);
+				waitq_init(&thread->scd_wq);
+				waitq_prepare_to_wait(&thread->scd_wq, &scd_wq_entry,
+					PS_INTERRUPTIBLE);
+				cpu_restore_interrupt(flags);
+				schedule();
+				waitq_finish_wait(&thread->scd_wq, &scd_wq_entry);
+			}
+
+			cpu_restore_interrupt(flags);
 		}
-	
-		if (res->status == STATUS_PAGE_FAULT) {
+
+		if (res.status == STATUS_PAGE_FAULT) {
 			dkprintf("STATUS_PAGE_FAULT in syscall, pid: %d\n", 
 					cpu_local_var(current)->proc->pid);
 			error = page_fault_process_vm(thread->vm,
-					(void *)res->fault_address,
-					res->fault_reason|PF_POPULATE);
+					(void *)res.fault_address,
+					res.fault_reason|PF_POPULATE);
 
 			/* send result */
 			req2.number = __NR_mmap;
 #define PAGER_RESUME_PAGE_FAULT	0x0101
 			req2.args[0] = PAGER_RESUME_PAGE_FAULT;
 			req2.args[1] = error;
+			/* The current thread is the requester and only the waiting thread
+			 * may serve the request */
+			req2.rtid = cpu_local_var(current)->tid;
+			req2.ttid = res.stid;
 
-			send_syscall(&req2, cpu, pid);
+			res.req_thread_status = IHK_SCD_REQ_THREAD_SPINNING;
+			send_syscall(&req2, cpu, pid, &res);
 		}
 	}
 
-	dkprintf("SC(%d)[%3d] got host reply: %d \n", 
-	        ihk_mc_get_processor_id(),
-	        req->number, res->ret);
+	dkprintf("%s: syscall num: %d got host reply: %d \n",
+		__FUNCTION__, req->number, res.ret);
 
-	rc = res->ret;
+	rc = res.ret;
 	if(islock){
 		ihk_mc_spinlock_unlock(&syscall_lock, irqstate);
 	}
@@ -795,7 +828,8 @@ terminate(int rc, int sig)
 	release_thread(mythread);
 	release_process_vm(vm);
 	schedule();
-	// no return
+	kprintf("%s: ERROR: returned from terminate() -> schedule()\n", __FUNCTION__);
+	panic("panic");
 }
 
 void
@@ -813,14 +847,15 @@ terminate_host(int pid)
 }
 
 void
-interrupt_syscall(int pid, int cpuid)
+interrupt_syscall(int pid, int tid)
 {
-	dkprintf("interrupt_syscall,target pid=%d,target cpuid=%d\n", pid, cpuid);
+	dkprintf("interrupt_syscall,target pid=%d,target tid=%d\n", pid, tid);
 	ihk_mc_user_context_t ctx;
 	long lerror;
 
+kprintf("interrupt_syscall pid=%d tid=%d\n", pid, tid);
 	ihk_mc_syscall_arg0(&ctx) = pid;
-	ihk_mc_syscall_arg1(&ctx) = cpuid;
+	ihk_mc_syscall_arg1(&ctx) = tid;
 
 	lerror = syscall_generic_forwarding(__NR_kill, &ctx);
 	if (lerror) {
@@ -883,8 +918,6 @@ static int do_munmap(void *addr, size_t len)
 	begin_free_pages_pending();
 	error = remove_process_memory_range(cpu_local_var(current)->vm,
 			(intptr_t)addr, (intptr_t)addr+len, &ro_freed);
-	// XXX: TLB flush
-	flush_tlb();
 	if (error || !ro_freed) {
 		clear_host_pte((uintptr_t)addr, len);
 	}
@@ -896,6 +929,8 @@ static int do_munmap(void *addr, size_t len)
 		}
 	}
 	finish_free_pages_pending();
+	dkprintf("%s: 0x%lx:%lu, error: %ld\n",
+		__FUNCTION__, addr, len, error);
 	return error;
 }
 
@@ -1043,31 +1078,29 @@ do_mmap(const intptr_t addr0, const size_t len0, const int prot,
 	vrflags |= PROT_TO_VR_FLAG(prot);
 	vrflags |= (flags & MAP_PRIVATE)? VR_PRIVATE: 0;
 	vrflags |= (flags & MAP_LOCKED)? VR_LOCKED: 0;
+	vrflags |= VR_DEMAND_PAGING;
 	if (flags & MAP_ANONYMOUS) {
-		if (0) {
-			/* dummy */
+		if (!anon_on_demand) {
+			populated_mapping = 1;
 		}
 #ifdef	USE_NOCACHE_MMAP
 #define	X_MAP_NOCACHE	MAP_32BIT
 		else if (flags & X_MAP_NOCACHE) {
+			vrflags &= ~VR_DEMAND_PAGING;
 			vrflags |= VR_IO_NOCACHE;
 		}
 #endif
-		else {
-			vrflags |= VR_DEMAND_PAGING;
-			if (!anon_on_demand) {
-				populated_mapping = 1;
-			}
-		}
-	}
-	else {
-		vrflags |= VR_DEMAND_PAGING;
 	}
 
 	if (flags & (MAP_POPULATE | MAP_LOCKED)) {
 		populated_mapping = 1;
 	}
 
+	/* XXX: Intel MPI 128MB mapping.. */
+	if (len == 134217728) {
+		populated_mapping = 0;
+	}
+
 	if (!(prot & PROT_WRITE)) {
 		error = set_host_vma(addr, len, PROT_READ);
 		if (error) {
@@ -1097,7 +1130,7 @@ do_mmap(const intptr_t addr0, const size_t len0, const int prot,
 		}
 #endif
 		if (error == -ESRCH) {
-			kprintf("do_mmap:hit non VREG\n");
+			dkprintf("do_mmap:hit non VREG\n");
 			/*
 			 * XXX: temporary:
 			 *
@@ -1108,10 +1141,17 @@ do_mmap(const intptr_t addr0, const size_t len0, const int prot,
 				vrflags &= ~VR_MEMTYPE_MASK;
 				vrflags |= VR_MEMTYPE_UC;
 			}
-			error = devobj_create(fd, len, off, &memobj, &maxprot);
+			error = devobj_create(fd, len, off, &memobj, &maxprot, 
+					prot, (flags & (MAP_POPULATE | MAP_LOCKED)));
+
+			if (!error) {
+				dkprintf("%s: device fd: %d off: %lu mapping at %p - %p\n", 
+						__FUNCTION__, fd, off, addr, addr + len); 
+			}
 		}
 		if (error) {
-			ekprintf("do_mmap:fileobj_create failed. %d\n", error);
+			kprintf("%s: error: file mapping failed, fd: %d, error: %d\n",
+					__FUNCTION__, error);
 			goto out;
 		}
 	}
@@ -1125,6 +1165,8 @@ do_mmap(const intptr_t addr0, const size_t len0, const int prot,
 			error = -ENOMEM;
 			goto out;
 		}
+		dkprintf("%s: 0x%x:%lu allocated %d pages, p2align: %lx\n",
+				__FUNCTION__, addr, len, npages, p2align);
 		phys = virt_to_phys(p);
 	}
 	else if (flags & MAP_SHARED) {
@@ -1160,10 +1202,10 @@ do_mmap(const intptr_t addr0, const size_t len0, const int prot,
 	error = add_process_memory_range(thread->vm, addr, addr+len, phys,
 			vrflags, memobj, off, pgshift);
 	if (error) {
-		ekprintf("do_mmap:add_process_memory_range"
-				"(%p,%lx,%lx,%lx,%lx,%d) failed %d\n",
-				thread->vm, addr, addr+len,
-				virt_to_phys(p), vrflags, pgshift, error);
+		kprintf("%s: add_process_memory_range failed for 0x%lx:%lu"
+				" flags: %lx, vrflags: %lx, pgshift: %d, error: %d\n",
+				__FUNCTION__, addr, addr+len,
+				flags, vrflags, pgshift, error);
 		goto out;
 	}
 
@@ -1181,9 +1223,12 @@ out:
 	if (!error && populated_mapping) {
 		error = populate_process_memory(thread->vm, (void *)addr, len);
 		if (error) {
-			ekprintf("do_mmap:populate_process_memory"
-					"(%p,%p,%lx) failed %d\n",
-					thread->vm, (void *)addr, len, error);
+			ekprintf("%s: error :populate_process_memory"
+					"vm: %p, addr: %p, len: %d (flags: %s%s) failed %d\n", __FUNCTION__,
+					thread->vm, (void *)addr, len,
+					(flags & MAP_POPULATE) ? "MAP_POPULATE " : "",
+					(flags & MAP_LOCKED) ? "MAP_LOCKED ": "",
+					error);
 			/*
 			 * In this case,
 			 * the mapping established by this call should be unmapped
@@ -1206,8 +1251,12 @@ out:
 	if (memobj) {
 		memobj_release(memobj);
 	}
-	dkprintf("do_mmap(%lx,%lx,%x,%x,%d,%lx): %ld %lx\n",
-			addr0, len0, prot, flags, fd, off0, error, addr);
+	dkprintf("%s: 0x%lx:%8lu, (req: 0x%lx:%lu), prot: %x, flags: %x, "
+			"fd: %d, off: %lu, error: %ld, addr: 0x%lx\n",
+			__FUNCTION__,
+			addr, len, addr0, len0, prot, flags,
+			fd, off0, error, addr);
+
 	return (!error)? addr: error;
 }
 
@@ -1438,8 +1487,8 @@ SYSCALL_DECLARE(getppid)
 	return thread->proc->ppid_parent->pid;
 }
 
-void
-settid(struct thread *thread, int mode, int newcpuid, int oldcpuid)
+void settid(struct thread *thread, int mode, int newcpuid, int oldcpuid,
+	int nr_tids, int *tids)
 {
 	struct syscall_request request IHK_DMA_ALIGN;
 	unsigned long rc;
@@ -1449,6 +1498,12 @@ settid(struct thread *thread, int mode, int newcpuid, int oldcpuid)
 	request.args[1] = thread->proc->pid;
 	request.args[2] = newcpuid;
 	request.args[3] = oldcpuid;
+	/*
+	 * If nr_tids is non-zero, tids should point to an array of ints
+	 * where the thread ids of the mcexec process are expected.
+	 */
+	request.args[4] = nr_tids;
+	request.args[5] = virt_to_phys(tids);
 	rc = do_syscall(&request, ihk_mc_get_processor_id(), thread->proc->pid);
 	if (mode != 2) {
 		thread->tid = rc;
@@ -1853,7 +1908,61 @@ unsigned long do_fork(int clone_flags, unsigned long newsp,
 	        &new->vm->address_space->cpu_set_lock);
 
 	if (clone_flags & CLONE_VM) {
-		settid(new, 1, cpuid, -1);
+		int *tids = NULL;
+		int i;
+		struct mcs_rwlock_node_irqsave lock;
+
+		mcs_rwlock_writer_lock(&newproc->threads_lock, &lock);
+		/* Obtain mcexec TIDs if not known yet */
+		if (!newproc->nr_tids) {
+			tids = kmalloc(sizeof(int) * num_processors, IHK_MC_AP_NOWAIT);
+			if (!tids) {
+				mcs_rwlock_writer_unlock(&newproc->threads_lock, &lock);
+				release_cpuid(cpuid);
+				return -ENOMEM;
+			}
+
+			newproc->tids = kmalloc(sizeof(struct mcexec_tid) * num_processors, IHK_MC_AP_NOWAIT);
+			if (!newproc->tids) {
+				mcs_rwlock_writer_unlock(&newproc->threads_lock, &lock);
+				kfree(tids);
+				release_cpuid(cpuid);
+				return -ENOMEM;
+			}
+
+			settid(new, 1, cpuid, -1, num_processors, tids);
+
+			for (i = 0; (i < num_processors) && tids[i]; ++i) {
+				dkprintf("%s: tid[%d]: %d\n", __FUNCTION__, i, tids[i]);
+				newproc->tids[i].tid = tids[i];
+				newproc->tids[i].thread = NULL;
+				++newproc->nr_tids;
+			}
+
+			kfree(tids);
+		}
+
+		/* Find an unused TID */
+retry_tid:
+		for (i = 0; i < newproc->nr_tids; ++i) {
+			if (!newproc->tids[i].thread) {
+				if (!__sync_bool_compare_and_swap(
+					&newproc->tids[i].thread, NULL, new)) {
+					goto retry_tid;
+				}
+				new->tid = newproc->tids[i].tid;
+				dkprintf("%s: tid %d assigned to %p\n", __FUNCTION__, new->tid, new);
+				break;
+			}
+		}
+
+		/* TODO: spawn more mcexec threads */
+		if (!new->tid) {
+			kprintf("%s: no more TIDs available\n");
+			panic("");
+		}
+
+		mcs_rwlock_writer_unlock(&newproc->threads_lock, &lock);
 	}
 	/* fork() a new process on the host */
 	else {
@@ -1873,7 +1982,7 @@ unsigned long do_fork(int clone_flags, unsigned long newsp,
 		}
 
 		/* In a single threaded process TID equals to PID */
-		settid(new, 0, cpuid, -1);
+		new->tid = newproc->pid;
 		new->vm->address_space->pids[0] = new->proc->pid;
 
 		dkprintf("fork(): new pid: %d\n", new->proc->pid);
@@ -2242,9 +2351,8 @@ SYSCALL_DECLARE(setfsgid)
 	unsigned long newfsgid;
 	struct syscall_request request IHK_DMA_ALIGN;
 
-	request.number = __NR_setfsuid;
+	request.number = __NR_setfsgid;
 	request.args[0] = fsgid;
-	request.args[1] = 0;
 	newfsgid = do_syscall(&request, ihk_mc_get_processor_id(), 0);
 	do_setresgid();
 	return newfsgid;
@@ -2470,6 +2578,31 @@ SYSCALL_DECLARE(close)
 	return rc;
 }
 
+SYSCALL_DECLARE(fcntl)
+{
+	int fd = ihk_mc_syscall_arg0(ctx);
+	// int cmd = ihk_mc_syscall_arg1(ctx);
+	long rc;
+	struct thread *thread = cpu_local_var(current);
+	struct process *proc = thread->proc;
+	struct mckfd *fdp;
+	long irqstate;
+
+	irqstate = ihk_mc_spinlock_lock(&proc->mckfd_lock);
+	for(fdp = proc->mckfd; fdp; fdp = fdp->next)
+		if(fdp->fd == fd)
+			break;
+	ihk_mc_spinlock_unlock(&proc->mckfd_lock, irqstate);
+
+	if(fdp && fdp->fcntl_cb){
+		rc = fdp->fcntl_cb(fdp, ctx);
+	}
+	else{
+		rc = syscall_generic_forwarding(__NR_fcntl, ctx);
+	}
+	return rc;
+}
+
 SYSCALL_DECLARE(rt_sigprocmask)
 {
 	int how = ihk_mc_syscall_arg0(ctx);
@@ -2621,26 +2754,12 @@ perf_counter_alloc(struct mc_perf_event *event)
 	struct perf_event_attr *attr = &event->attr;
 	struct mc_perf_event *leader = event->group_leader;
 
-	if(attr->type == PERF_TYPE_HARDWARE) {
-
-		event->counter_id = ihk_mc_perfctr_alloc_counter(leader->pmc_status); 
-	
-	} else if(attr->type == PERF_TYPE_RAW) {
-		// PAPI_REF_CYC counted by fixed counter
-		if((attr->config & 0x0000ffff) == 0x00000300) {
-			event->counter_id = 2 + X86_IA32_BASE_FIXED_PERF_COUNTERS;
-			return ret;
-		}
-
-		event->counter_id = ihk_mc_perfctr_alloc_counter(leader->pmc_status); 
-	} else {
-		// Not supported type.
-		ret = -1;
-	}
+	ret = ihk_mc_perfctr_alloc_counter(&attr->type, &attr->config, leader->pmc_status); 
 
 	if(ret >= 0) {
-		leader->pmc_status |= 1UL << event->counter_id;
+		leader->pmc_status |= 1UL << ret;
 	}
+	event->counter_id = ret;
 
 	return ret;
 }
@@ -2649,7 +2768,6 @@ int
 perf_counter_start(struct mc_perf_event *event)
 {
 	int ret = 0;
-	enum ihk_perfctr_type type;
 	struct perf_event_attr *attr = &event->attr;
 	int mode = 0x00;
 
@@ -2660,52 +2778,34 @@ perf_counter_start(struct mc_perf_event *event)
 		mode |= PERFCTR_USER_MODE;
 	}
 
-	if(attr->type == PERF_TYPE_HARDWARE) {
-		switch(attr->config){
-		case PERF_COUNT_HW_CPU_CYCLES :
-			type = APT_TYPE_CYCLE;
-			break;
-		case PERF_COUNT_HW_INSTRUCTIONS :
-			type = APT_TYPE_INSTRUCTIONS;
-			break;
-		default :
-			// Not supported config.
-			type = PERFCTR_MAX_TYPE;
-		}
-
-		ret = ihk_mc_perfctr_init(event->counter_id, type, mode);
-		ihk_mc_perfctr_set(event->counter_id, event->sample_freq * -1);
-		ihk_mc_perfctr_start(1UL << event->counter_id);
-	
-	} else if(attr->type == PERF_TYPE_RAW) {
-		// PAPI_REF_CYC counted by fixed counter
-		if(event->counter_id >= X86_IA32_BASE_FIXED_PERF_COUNTERS) {
-			ret = ihk_mc_perfctr_fixed_init(event->counter_id, mode);
-			ihk_mc_perfctr_set(event->counter_id, event->sample_freq * -1);
-			ihk_mc_perfctr_start(1UL << event->counter_id);
-			return ret;
-		}
-
+	if(event->counter_id >= 0 && event->counter_id < X86_IA32_NUM_PERF_COUNTERS) {
 		ret = ihk_mc_perfctr_init_raw(event->counter_id, attr->config, mode);
-		ihk_mc_perfctr_set(event->counter_id, event->sample_freq * -1);
 		ihk_mc_perfctr_start(1UL << event->counter_id);
-	} else {
-		// Not supported type.
+	}
+	else if(event->counter_id >= X86_IA32_BASE_FIXED_PERF_COUNTERS &&
+		event->counter_id < X86_IA32_BASE_FIXED_PERF_COUNTERS + X86_IA32_NUM_FIXED_PERF_COUNTERS) {
+		ret = ihk_mc_perfctr_fixed_init(event->counter_id, mode);
+		ihk_mc_perfctr_start(1UL << event->counter_id);
+	}
+	else {
 		ret = -1;
 	}
-
+		
 	return ret;
 }
 
 unsigned long perf_event_read_value(struct mc_perf_event *event)
 {
 	unsigned long rtn_count = 0;
+	unsigned long pmc_count = 0;
 	int counter_id = event->counter_id;
 
-	if(event->pid == 0)
-		event->count = ihk_mc_perfctr_read(counter_id);
+	if(event->pid == 0) {
+		pmc_count = ihk_mc_perfctr_read(counter_id) + event->attr.sample_freq;
+		pmc_count &= 0x000000ffffffffffL; // 40bit MASK
+	}
 
-	rtn_count += event->count;
+	rtn_count += event->count + pmc_count;
 
 	if(event->attr.inherit)
 		rtn_count += event->child_count_total;
@@ -2922,11 +3022,21 @@ perf_ioctl(struct mckfd *sfd, ihk_mc_user_context_t *ctx)
                 break;
         case PERF_EVENT_IOC_RESET:
 		// TODO: reset other process
-		ihk_mc_perfctr_reset(counter_id);
+		ihk_mc_perfctr_set(counter_id, event->attr.sample_freq * -1);
+		event->count = 0L;
                 break;
         case PERF_EVENT_IOC_REFRESH:
 		// TODO: refresh other process
-		ihk_mc_perfctr_set(counter_id, event->sample_freq * -1);
+		
+		// not supported on inherited events
+		if(event->attr.inherit)
+			return -EINVAL;
+
+		event->count += event->attr.sample_freq;
+		ihk_mc_perfctr_set(counter_id, event->attr.sample_freq * -1);
+
+		perf_start(event);
+
 		break;
 	default :
 		return -1;
@@ -2945,6 +3055,28 @@ perf_close(struct mckfd *sfd, ihk_mc_user_context_t *ctx)
 	return 0;
 }
 
+static int
+perf_fcntl(struct mckfd *sfd, ihk_mc_user_context_t *ctx)
+{
+	int cmd = ihk_mc_syscall_arg1(ctx);
+	long arg = ihk_mc_syscall_arg2(ctx);
+	int rc = 0;
+
+	switch(cmd) {
+	case 10: // F_SETSIG
+		sfd->sig_no = arg;
+		break;
+	case 0xf: // F_SETOWN_EX
+		break;
+	default : 
+		break;
+	}
+
+	rc = syscall_generic_forwarding(__NR_fcntl, ctx);
+
+	return rc;
+}
+
 static long
 perf_mmap(struct mckfd *sfd, ihk_mc_user_context_t *ctx)
 {
@@ -2963,6 +3095,7 @@ perf_mmap(struct mckfd *sfd, ihk_mc_user_context_t *ctx)
 
 	// setup perf_event_mmap_page
 	page = (struct perf_event_mmap_page *)rc;
+	page->data_head = 16;
 	page->cap_user_rdpmc = 1;
 
 	return rc;
@@ -3014,7 +3147,7 @@ SYSCALL_DECLARE(perf_event_open)
 
 	event->sample_freq = attr->sample_freq;
 	event->nr_siblings = 0;
-	event->count = 0;
+	event->count = 0L;
 	event->child_count_total = 0;
 	event->parent = NULL;
 	event->pid = pid;
@@ -3050,10 +3183,12 @@ SYSCALL_DECLARE(perf_event_open)
 	if(!sfd)
 		return -ENOMEM;
 	sfd->fd = fd;
+	sfd->sig_no = -1;
 	sfd->read_cb = perf_read;
 	sfd->ioctl_cb = perf_ioctl;
 	sfd->close_cb = perf_close;
 	sfd->mmap_cb = perf_mmap;
+	sfd->fcntl_cb = perf_fcntl;
 	sfd->data = (long)event;
 	irqstate = ihk_mc_spinlock_lock(&proc->mckfd_lock);
 
@@ -3413,6 +3548,90 @@ SYSCALL_DECLARE(mincore)
 	return 0;
 } /* sys_mincore() */
 
+static int
+set_memory_range_flag(struct vm_range *range, unsigned long arg)
+{
+	range->flag |= arg;
+	return 0;
+}
+
+static int
+clear_memory_range_flag(struct vm_range *range, unsigned long arg)
+{
+	range->flag &= ~arg;
+	return 0;
+}
+
+static int
+change_attr_process_memory_range(struct process_vm *vm,
+                                 uintptr_t start, uintptr_t end,
+                                 int (*change_proc)(struct vm_range *,
+                                                    unsigned long),
+                                 unsigned long arg)
+{
+	uintptr_t addr;
+	int error;
+	struct vm_range *range;
+	struct vm_range *prev;
+	struct vm_range *next;
+	int join_flag = 0;
+
+	error = 0;
+	range = lookup_process_memory_range(vm, start, start + PAGE_SIZE);
+	if(!range){
+		error = -ENOMEM;
+		goto out;
+	}
+
+	prev = previous_process_memory_range(vm, range);
+	if(!prev)
+		prev = range;
+	for (addr = start; addr < end; addr = range->end) {
+		if (range->start < addr) {
+			if((error = split_process_memory_range(vm, range, addr, &range))) {
+				break;
+			}
+		}
+		if (end < range->end) {
+			if((error = split_process_memory_range(vm, range, end, NULL))) {
+				break;
+			}
+		}
+
+		if(!(error = change_proc(range, arg))){
+			break;
+		}
+		range = next_process_memory_range(vm, range);
+	}
+	if(error){
+		next = next_process_memory_range(vm, range);
+		if(!next)
+			next = range;
+	}
+	else{
+		next = range;
+	}
+
+	while(prev != next){
+		int wkerr;
+
+		range = next_process_memory_range(vm, prev);
+		if(!range)
+			break;
+		wkerr = join_process_memory_range(vm, prev, range);
+		if(range == next)
+			join_flag = 1;
+		if (wkerr) {
+			if(join_flag)
+				break;
+			prev = range;
+		}
+	}
+
+out:
+	return error;
+}
+
 SYSCALL_DECLARE(madvise)
 {
 	const uintptr_t start = (uintptr_t)ihk_mc_syscall_arg0(ctx);
@@ -3521,6 +3740,7 @@ SYSCALL_DECLARE(madvise)
 				goto out;
 			}
 		}
+		else if(advice == MADV_DONTFORK || advice == MADV_DOFORK);
 		else if (!range->memobj || !memobj_has_pager(range->memobj)) {
 			dkprintf("[%d]sys_madvise(%lx,%lx,%x):has not pager"
 					"[%lx-%lx) %lx\n",
@@ -3565,6 +3785,27 @@ SYSCALL_DECLARE(madvise)
 		}
 	}
 
+	if(advice == MADV_DONTFORK){
+		error = change_attr_process_memory_range(thread->vm, start, end,
+		                                         set_memory_range_flag,
+		                                         VR_DONTFORK);
+		if(error){
+			goto out;
+		}
+	}
+	if(advice == MADV_DOFORK){
+		error = change_attr_process_memory_range(thread->vm, start, end,
+		                                         clear_memory_range_flag,
+		                                         VR_DONTFORK);
+		if(error){
+			goto out;
+		}
+	}
+	if(advice == MADV_DONTFORK ||
+	   advice == MADV_DOFORK){
+		error = syscall_generic_forwarding(__NR_madvise, ctx);
+	}
+
 	error = 0;
 out:
 	ihk_mc_spinlock_unlock_noirq(&thread->vm->memory_range_lock);
@@ -5540,6 +5781,10 @@ SYSCALL_DECLARE(sched_setaffinity)
 	int empty_set = 1; 
 	extern int num_processors;
 
+	if (!u_cpu_set) {
+		return -EINVAL;
+	}
+
 	if (sizeof(k_cpu_set) > len) {
 		memset(&k_cpu_set, 0, sizeof(k_cpu_set));
 	}
@@ -5547,7 +5792,7 @@ SYSCALL_DECLARE(sched_setaffinity)
 	len = MIN2(len, sizeof(k_cpu_set));
 
 	if (copy_from_user(&k_cpu_set, u_cpu_set, len)) {
-		kprintf("%s:%d copy_from_user failed.\n", __FILE__, __LINE__);
+		dkprintf("%s: error: copy_from_user failed for %p:%d\n", __FUNCTION__, u_cpu_set, len);
 		return -EFAULT;
 	}
 

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/zeroobj.c

@@ -172,6 +172,10 @@ 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/include/ihk/mm.h

@@ -103,7 +103,7 @@ void ihk_mc_clean_micpa(void);
 void *ihk_mc_alloc_aligned_pages(int npages, int p2align, enum ihk_mc_ap_flag flag);
 void *ihk_mc_alloc_pages(int npages, enum ihk_mc_ap_flag flag);
 void ihk_mc_free_pages(void *p, int npages);
-void *ihk_mc_allocate(int size, enum ihk_mc_ap_flag flag);
+void *ihk_mc_allocate(int size, int flag);
 void ihk_mc_free(void *p);
 
 void *arch_alloc_page(enum ihk_mc_ap_flag flag);

lib/include/ihk/perfctr.h

@@ -54,11 +54,11 @@ int ihk_mc_perfctr_start(unsigned long counter_mask);
 int ihk_mc_perfctr_stop(unsigned long counter_mask);
 int ihk_mc_perfctr_fixed_init(int counter, int mode);
 int ihk_mc_perfctr_reset(int counter);
-int ihk_mc_perfctr_set(int counter, unsigned long value);
+int ihk_mc_perfctr_set(int counter, long value);
 int ihk_mc_perfctr_read_mask(unsigned long counter_mask, unsigned long *value);
 unsigned long ihk_mc_perfctr_read(int counter);
 unsigned long ihk_mc_perfctr_read_msr(int counter);
-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);
 
 #endif