Revert "brk(): larger allocation units internally"

This reverts commit c58ab0f648.

arch/x86/kernel/cpu.c

@@ -148,7 +148,7 @@ extern char page_fault[], general_protection_exception[];
 extern char debug_exception[], int3_exception[];
 
 uint64_t boot_pat_state = 0;
-int no_turbo = 0; /* May be updated by early parsing of kargs */
+int no_turbo = 1; /* May be updated by early parsing of kargs */
 
 extern int num_processors; /* kernel/ap.c */
 struct pvclock_vsyscall_time_info *pvti = NULL;
@@ -844,6 +844,25 @@ void set_signal(int sig, void *regs, struct siginfo *info);
 void check_signal(unsigned long, void *, int);
 extern void tlb_flush_handler(int vector);
 
+void __show_stack(uintptr_t *sp) {
+	while (((uintptr_t)sp >= 0xffff800000000000)
+			&& ((uintptr_t)sp <  0xffffffff80000000)) {
+		uintptr_t fp;
+		uintptr_t ip;
+
+		fp = sp[0];
+		ip = sp[1];
+		kprintf("IP: %016lx, SP: %016lx, FP: %016lx\n", ip, (uintptr_t)sp, fp);
+		sp = (void *)fp;
+	}
+	return;
+}
+
+void show_context_stack(uintptr_t *rbp) {
+	__show_stack(rbp);
+	return;
+}
+
 void handle_interrupt(int vector, struct x86_user_context *regs)
 {
 	struct ihk_mc_interrupt_handler *h;
@@ -952,6 +971,9 @@ void handle_interrupt(int vector, struct x86_user_context *regs)
 
 			tlb_flush_handler(vector);
 	} 
+	else if (vector == 133) {
+		show_context_stack((uintptr_t *)regs->gpr.rbp);
+	}
 	else {
 		list_for_each_entry(h, &handlers[vector - 32], list) {
 			if (h->func) {
@@ -1079,6 +1101,10 @@ unhandled_page_fault(struct thread *thread, void *fault_addr, void *regs)
 
 	kprintf_unlock(irqflags);
 
+	if (!(error & PF_USER)) {
+		panic("panic: kernel mode PF");
+	}
+
 	/* TODO */
 	ihk_mc_debug_show_interrupt_context(regs);
 
@@ -1736,7 +1762,7 @@ int arch_setup_pvclock(void)
 	npages = (size + PAGE_SIZE - 1) / PAGE_SIZE;
 	pvti_npages = npages;
 
-	pvti = allocate_pages(npages, IHK_MC_AP_NOWAIT);
+	pvti = ihk_mc_alloc_pages(npages, IHK_MC_AP_NOWAIT);
 	if (!pvti) {
 		ekprintf("arch_setup_pvclock: allocate_pages failed.\n");
 		return -ENOMEM;
@@ -1766,44 +1792,6 @@ void arch_start_pvclock(void)
 	return;
 } /* arch_start_pvclock() */
 
-static struct cpu_mapping *cpu_mapping = NULL;
-
-int arch_get_cpu_mapping(struct cpu_mapping **buf, int *nelemsp)
-{
-	int error;
-	size_t size;
-	int npages;
-	struct cpu_mapping *mapping;
-	int cpu;
-	struct x86_cpu_local_variables *v;
-
-	if (!cpu_mapping) {
-		size = sizeof(*mapping) * num_processors;
-		npages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
-		mapping = allocate_pages(npages, IHK_MC_AP_NOWAIT);
-		if (!mapping) {
-			error = -ENOMEM;
-			ekprintf("arch_get_cpu_mapping:allocate_pages failed. %d\n", error);
-			goto out;
-		}
-
-		for (cpu = 0; cpu < num_processors; ++cpu) {
-			v = get_x86_cpu_local_variable(cpu);
-			mapping[cpu].cpu_number = cpu;
-			mapping[cpu].hw_id = v->apic_id;
-		}
-
-		cpu_mapping = mapping;
-	}
-
-	error = 0;
-	*buf = cpu_mapping;
-	*nelemsp = num_processors;
-
-out:
-	return error;
-} /* arch_get_cpu_mapping() */
-
 #define KVM_CPUID_SIGNATURE	0x40000000
 
 int running_on_kvm(void) {

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

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

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

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

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

@@ -306,7 +306,7 @@ struct page_table;
 void set_pte(pte_t *ppte, unsigned long phys, enum ihk_mc_pt_attribute attr);
 pte_t *get_pte(struct page_table *pt, void *virt, enum ihk_mc_pt_attribute attr);
 
-void *early_alloc_page(void);
+void *early_alloc_pages(int nr_pages);
 void *get_last_early_heap(void);
 void flush_tlb(void);
 void flush_tlb_single(unsigned long addr);

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

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

arch/x86/kernel/include/syscall_list.h

@@ -22,7 +22,7 @@
 
 SYSCALL_HANDLED(0, read)
 SYSCALL_DELEGATED(1, write)
-SYSCALL_DELEGATED(2, open)
+SYSCALL_HANDLED(2, open)
 SYSCALL_HANDLED(3, close)
 SYSCALL_DELEGATED(4, stat)
 SYSCALL_DELEGATED(5, fstat)
@@ -150,5 +150,8 @@ SYSCALL_HANDLED(602, pmc_start)
 SYSCALL_HANDLED(603, pmc_stop)
 SYSCALL_HANDLED(604, pmc_reset)
 SYSCALL_HANDLED(700, get_cpu_id)
+#ifdef TRACK_SYSCALLS
+SYSCALL_HANDLED(701, syscall_offload_clr_cntrs)
+#endif // TRACK_SYSCALLS
 
 /**** End of File ****/

arch/x86/kernel/memory.c

@@ -31,11 +31,10 @@
 static char *last_page;
 extern char _head[], _end[];
 
-static struct ihk_mc_pa_ops *pa_ops;
-
 extern unsigned long x86_kernel_phys_base;
 
-void *early_alloc_page(void)
+/* Arch specific early allocation routine */
+void *early_alloc_pages(int nr_pages)
 {
 	void *p;
 
@@ -48,41 +47,14 @@ void *early_alloc_page(void)
 		panic("Early allocator is already finalized. Do not use it.\n");
 	}
 	p = last_page;
-	last_page += PAGE_SIZE;
+	last_page += (nr_pages * PAGE_SIZE);
 
 	return p;
 }
 
-void *arch_alloc_page(enum ihk_mc_ap_flag flag)
+void early_alloc_invalidate(void)
 {
-	if (pa_ops)
-		return pa_ops->alloc_page(1, PAGE_P2ALIGN, flag);
-	else
-		return early_alloc_page();
-}
-void arch_free_page(void *ptr)
-{
-	if (pa_ops)
-		pa_ops->free_page(ptr, 1);
-}
-
-void *ihk_mc_alloc_aligned_pages(int npages, int p2align, enum ihk_mc_ap_flag flag)
-{
-	if (pa_ops)
-		return pa_ops->alloc_page(npages, p2align, flag);
-	else
-		return NULL;
-}
-
-void *ihk_mc_alloc_pages(int npages, enum ihk_mc_ap_flag flag)
-{
-	return ihk_mc_alloc_aligned_pages(npages, PAGE_P2ALIGN, flag);
-}
-
-void ihk_mc_free_pages(void *p, int npages)
-{
-	if (pa_ops)
-		pa_ops->free_page(p, npages);
+	last_page = (void *)-1;
 }
 
 void *ihk_mc_allocate(int size, int flag)
@@ -175,7 +147,7 @@ static unsigned long setup_l3(struct page_table *pt,
 			pt->entry[i] = 0;
 			continue;
 		}
-		pt_phys = setup_l2(arch_alloc_page(IHK_MC_AP_CRITICAL), phys, start, end);
+		pt_phys = setup_l2(ihk_mc_alloc_pages(1, IHK_MC_AP_CRITICAL), phys, start, end);
 
 		pt->entry[i] = pt_phys | PFL3_PDIR_ATTR;
 	}
@@ -199,7 +171,7 @@ static void init_normal_area(struct page_table *pt)
 
 	for (phys = (map_start & ~(PTL4_SIZE - 1)); phys < map_end;
 	     phys += PTL4_SIZE) {
-		pt_phys = setup_l3(arch_alloc_page(IHK_MC_AP_CRITICAL), phys,
+		pt_phys = setup_l3(ihk_mc_alloc_pages(1, IHK_MC_AP_CRITICAL), phys,
 		                   map_start, map_end);
 
 		pt->entry[ident_index++] = pt_phys | PFL4_PDIR_ATTR;
@@ -209,7 +181,7 @@ static void init_normal_area(struct page_table *pt)
 
 static struct page_table *__alloc_new_pt(enum ihk_mc_ap_flag ap_flag)
 {
-	struct page_table *newpt = arch_alloc_page(ap_flag);
+	struct page_table *newpt = ihk_mc_alloc_pages(1, ap_flag);
 
 	if(newpt)
 		memset(newpt, 0, sizeof(struct page_table));
@@ -518,8 +490,10 @@ uint64_t ihk_mc_pt_virt_to_pagemap(struct page_table *pt, unsigned long virt)
 	return pagemap;
 }
 
-int ihk_mc_pt_virt_to_phys(struct page_table *pt,
-                           const void *virt, unsigned long *phys)
+int ihk_mc_pt_virt_to_phys_size(struct page_table *pt,
+                           const void *virt,
+						   unsigned long *phys,
+						   unsigned long *size)
 {
 	int l4idx, l3idx, l2idx, l1idx;
 	unsigned long v = (unsigned long)virt;
@@ -541,6 +515,7 @@ int ihk_mc_pt_virt_to_phys(struct page_table *pt,
 	if ((pt->entry[l3idx] & PFL3_SIZE)) {
 		*phys = pte_get_phys(&pt->entry[l3idx])
 			| (v & (PTL3_SIZE - 1));
+		if (size) *size = PTL3_SIZE;
 		return 0;
 	}
 	pt = phys_to_virt(pte_get_phys(&pt->entry[l3idx]));
@@ -551,6 +526,7 @@ int ihk_mc_pt_virt_to_phys(struct page_table *pt,
 	if ((pt->entry[l2idx] & PFL2_SIZE)) {
 		*phys = pte_get_phys(&pt->entry[l2idx])
 			| (v & (PTL2_SIZE - 1));
+		if (size) *size = PTL2_SIZE;
 		return 0;
 	}
 	pt = phys_to_virt(pte_get_phys(&pt->entry[l2idx]));
@@ -560,9 +536,17 @@ int ihk_mc_pt_virt_to_phys(struct page_table *pt,
 	}
 
 	*phys = pte_get_phys(&pt->entry[l1idx]) | (v & (PTL1_SIZE - 1));
+	if (size) *size = PTL1_SIZE;
 	return 0;
 }
 
+int ihk_mc_pt_virt_to_phys(struct page_table *pt,
+                           const void *virt, unsigned long *phys)
+{
+	return ihk_mc_pt_virt_to_phys_size(pt, virt, phys, NULL);
+}
+
+
 int ihk_mc_pt_print_pte(struct page_table *pt, void *virt)
 {
 	int l4idx, l3idx, l2idx, l1idx;
@@ -574,28 +558,34 @@ int ihk_mc_pt_print_pte(struct page_table *pt, void *virt)
 
 	GET_VIRT_INDICES(v, l4idx, l3idx, l2idx, l1idx);
 
+	__kprintf("l4 table: 0x%lX l4idx: %d \n", virt_to_phys(pt), l4idx);
 	if (!(pt->entry[l4idx] & PFL4_PRESENT)) {
 		__kprintf("0x%lX l4idx not present! \n", (unsigned long)virt);
-		__kprintf("l4 entry: 0x%lX\n", pt->entry[l4idx]);
 		return -EFAULT;
 	}
+	__kprintf("l4 entry: 0x%lX\n", pt->entry[l4idx]);
 	pt = phys_to_virt(pt->entry[l4idx] & PAGE_MASK);
 
 	__kprintf("l3 table: 0x%lX l3idx: %d \n", virt_to_phys(pt), l3idx);
 	if (!(pt->entry[l3idx] & PFL3_PRESENT)) {
 		__kprintf("0x%lX l3idx not present! \n", (unsigned long)virt);
-		__kprintf("l3 entry: 0x%lX\n", pt->entry[l3idx]);
 		return -EFAULT;
 	}
+	__kprintf("l3 entry: 0x%lX\n", pt->entry[l3idx]);
+	if ((pt->entry[l3idx] & PFL3_SIZE)) {
+		__kprintf("l3 entry is 1G page\n");
+		return 0;
+	}
 	pt = phys_to_virt(pt->entry[l3idx] & PAGE_MASK);
 	
 	__kprintf("l2 table: 0x%lX l2idx: %d \n", virt_to_phys(pt), l2idx);
 	if (!(pt->entry[l2idx] & PFL2_PRESENT)) {
 		__kprintf("0x%lX l2idx not present! \n", (unsigned long)virt);
-		__kprintf("l2 entry: 0x%lX\n", pt->entry[l2idx]);
 		return -EFAULT;
 	}
+	__kprintf("l2 entry: 0x%lX\n", pt->entry[l2idx]);
 	if ((pt->entry[l2idx] & PFL2_SIZE)) {
+		__kprintf("l2 entry is 2M page\n");
 		return 0;
 	}
 	pt = phys_to_virt(pt->entry[l2idx] & PAGE_MASK);
@@ -718,7 +708,7 @@ static void destroy_page_table(int level, struct page_table *pt)
 		}
 	}
 
-	arch_free_page(pt);
+	ihk_mc_free_pages(pt, 1);
 	return;
 }
 
@@ -1112,7 +1102,7 @@ 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)) {
+		if (!page || (page && page_unmap(page))) {
 			ihk_mc_free_pages(phys_to_virt(phys), 1);
 			dkprintf("%s: freeing regular page at 0x%lx\n", __FUNCTION__, base);
 		}
@@ -1161,7 +1151,7 @@ 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)) {
+			if (!page || (page && page_unmap(page))) {
 				ihk_mc_free_pages(phys_to_virt(phys), PTL2_SIZE/PTL1_SIZE);
 				dkprintf("%s: freeing large page at 0x%lx\n", __FUNCTION__, base);
 			}
@@ -1181,7 +1171,7 @@ static int clear_range_l2(void *args0, pte_t *ptep, uint64_t base,
 		*ptep = PTE_NULL;
 		remote_flush_tlb_cpumask(args->vm, base,
 				ihk_mc_get_processor_id());
-		arch_free_page(pt);
+		ihk_mc_free_pages(pt, 1);
 	}
 
 	return 0;
@@ -1226,7 +1216,7 @@ static int clear_range_l3(void *args0, pte_t *ptep, uint64_t base,
 		}
 
 		if (!(old & PFL3_FILEOFF) && args->free_physical) {
-			if (page && page_unmap(page)) {
+			if (!page || (page && page_unmap(page))) {
 				ihk_mc_free_pages(phys_to_virt(phys), PTL3_SIZE/PTL1_SIZE);
 			}
 			args->vm->currss -= PTL3_SIZE;
@@ -1245,7 +1235,7 @@ static int clear_range_l3(void *args0, pte_t *ptep, uint64_t base,
 		*ptep = PTE_NULL;
 		remote_flush_tlb_cpumask(args->vm, base,
 				ihk_mc_get_processor_id());
-		arch_free_page(pt);
+		ihk_mc_free_pages(pt, 1);
 	}
 
 	return 0;
@@ -1596,7 +1586,7 @@ retry:
 	error = 0;
 out:
 	if (newpt) {
-		arch_free_page(newpt);
+		ihk_mc_free_pages(newpt, 1);
 	}
 	dkprintf("set_range_l2(%lx,%lx,%lx): %d %lx\n",
 			base, start, end, error, *ptep);
@@ -1679,7 +1669,7 @@ retry:
 	error = 0;
 out:
 	if (newpt) {
-		arch_free_page(newpt);
+		ihk_mc_free_pages(newpt, 1);
 	}
 	dkprintf("set_range_l3(%lx,%lx,%lx): %d\n",
 			base, start, end, error, *ptep);
@@ -1737,7 +1727,7 @@ retry:
 	error = 0;
 out:
 	if (newpt) {
-		arch_free_page(newpt);
+		ihk_mc_free_pages(newpt, 1);
 	}
 	dkprintf("set_range_l4(%lx,%lx,%lx): %d %lx\n",
 			base, start, end, error, *ptep);
@@ -1789,9 +1779,19 @@ int ihk_mc_pt_set_pte(page_table_t pt, pte_t *ptep, size_t pgsize,
 		*ptep = phys | attr_to_l1attr(attr);
 	}
 	else if (pgsize == PTL2_SIZE) {
+		if (phys & (PTL2_SIZE - 1)) {
+			kprintf("%s: error: phys needs to be PTL2_SIZE aligned\n", __FUNCTION__);
+			error = -1;
+			goto out;
+		}
 		*ptep = phys | attr_to_l2attr(attr | PTATTR_LARGEPAGE);
 	}
 	else if ((pgsize == PTL3_SIZE) && (use_1gb_page)) {
+		if (phys & (PTL3_SIZE - 1)) {
+			kprintf("%s: error: phys needs to be PTL3_SIZE aligned\n", __FUNCTION__);
+			error = -1;
+			goto out;
+		}
 		*ptep = phys | attr_to_l3attr(attr | PTATTR_LARGEPAGE);
 	}
 	else {
@@ -2094,7 +2094,7 @@ static void init_vsyscall_area(struct page_table *pt)
 void init_page_table(void)
 {
 	check_available_page_size();
-	init_pt = arch_alloc_page(IHK_MC_AP_CRITICAL);
+	init_pt = ihk_mc_alloc_pages(1, IHK_MC_AP_CRITICAL);
 	ihk_mc_spinlock_init(&init_pt_lock);
 	
 	memset(init_pt, 0, sizeof(PAGE_SIZE));
@@ -2111,27 +2111,27 @@ void init_page_table(void)
 }
 
 extern void __reserve_arch_pages(unsigned long, unsigned long,
-                                 void (*)(unsigned long, unsigned long, int));
+		void (*)(struct ihk_page_allocator_desc *, 
+			unsigned long, unsigned long, int));
 
-void ihk_mc_reserve_arch_pages(unsigned long start, unsigned long end,
-                               void (*cb)(unsigned long, unsigned long, int))
+void ihk_mc_reserve_arch_pages(struct ihk_page_allocator_desc *pa_allocator,
+		unsigned long start, unsigned long end,
+		void (*cb)(struct ihk_page_allocator_desc *, 
+			unsigned long, unsigned long, int))
 {
 	/* Reserve Text + temporal heap */
-	cb(virt_to_phys(_head), virt_to_phys(get_last_early_heap()), 0);
+	cb(pa_allocator, virt_to_phys(_head), virt_to_phys(get_last_early_heap()), 0);
 	/* Reserve trampoline area to boot the second ap */
-	cb(ap_trampoline, ap_trampoline + AP_TRAMPOLINE_SIZE, 0);
+	cb(pa_allocator, ap_trampoline, ap_trampoline + AP_TRAMPOLINE_SIZE, 0);
 	/* Reserve the null page */
-	cb(0, PAGE_SIZE, 0);
-	/* Micro-arch specific */
+	cb(pa_allocator, 0, PAGE_SIZE, 0);
+	/* 
+	 * Micro-arch specific 
+	 * TODO: this does nothing in SMP mode, update it for KNC if necessary 
+	 */
 	__reserve_arch_pages(start, end, cb);
 }
 
-void ihk_mc_set_page_allocator(struct ihk_mc_pa_ops *ops)
-{
-	last_page = (void *)-1;
-	pa_ops = ops;
-}
-
 unsigned long virt_to_phys(void *v)
 {
 	unsigned long va = (unsigned long)v;
@@ -2383,8 +2383,18 @@ int write_process_vm(struct process_vm *vm, void *udst, const void *ksrc, size_t
 			return error;
 		}
 
-		va = phys_to_virt(pa);
-		memcpy(va, from, cpsize);
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+			dkprintf("%s: pa is outside of LWK memory, from: %p,"
+				"pa: %p, cpsize: %d\n", __FUNCTION__, from, pa, cpsize);
+			va = ihk_mc_map_virtual(pa, 1, PTATTR_ACTIVE);
+			memcpy(va, from, cpsize);
+			ihk_mc_unmap_virtual(va, 1, 1);
+		}
+		else {
+			va = phys_to_virt(pa);
+			memcpy(va, from, cpsize);
+		}
 
 		from += cpsize;
 		to += cpsize;

arch/x86/kernel/mikc.c

@@ -16,6 +16,7 @@
 #include <memory.h>
 #include <string.h>
 
+extern int num_processors;
 extern void arch_set_mikc_queue(void *r, void *w);
 ihk_ikc_ph_t arch_master_channel_packet_handler;
 
@@ -23,17 +24,23 @@ int ihk_mc_ikc_init_first_local(struct ihk_ikc_channel_desc *channel,
                                 ihk_ikc_ph_t packet_handler)
 {
 	struct ihk_ikc_queue_head *rq, *wq;
+	size_t mikc_queue_pages;
 
 	ihk_ikc_system_init(NULL);
 
 	memset(channel, 0, sizeof(struct ihk_ikc_channel_desc));
 
-	/* Place both sides in this side */
-	rq = arch_alloc_page(IHK_MC_AP_CRITICAL);
-	wq = arch_alloc_page(IHK_MC_AP_CRITICAL);
+	mikc_queue_pages = ((num_processors * MASTER_IKCQ_PKTSIZE)
+			+ (PAGE_SIZE - 1)) / PAGE_SIZE;
 
-	ihk_ikc_init_queue(rq, 0, 0, PAGE_SIZE, MASTER_IKCQ_PKTSIZE);
-	ihk_ikc_init_queue(wq, 0, 0, PAGE_SIZE, MASTER_IKCQ_PKTSIZE);
+	/* Place both sides in this side */
+	rq = ihk_mc_alloc_pages(mikc_queue_pages, IHK_MC_AP_CRITICAL);
+	wq = ihk_mc_alloc_pages(mikc_queue_pages, IHK_MC_AP_CRITICAL);
+
+	ihk_ikc_init_queue(rq, 0, 0,
+			mikc_queue_pages * PAGE_SIZE, MASTER_IKCQ_PKTSIZE);
+	ihk_ikc_init_queue(wq, 0, 0,
+			mikc_queue_pages * PAGE_SIZE, MASTER_IKCQ_PKTSIZE);
 
 	arch_master_channel_packet_handler = packet_handler;
 

arch/x86/kernel/syscall.c

@@ -70,71 +70,37 @@ static struct vdso vdso;
 static size_t container_size = 0;
 static ptrdiff_t vdso_offset;
 
-/*
-See dkprintf("BSP HW ID = %d, ", bsp_hw_id); (in ./mcos/kernel/ap.c)
+extern int num_processors;
 
-Core with BSP HW ID 224 is 1st logical core of last physical core.
-It                      boots first and is given SW-ID of 0
+int obtain_clone_cpuid(cpu_set_t *cpu_set) {
+	int min_queue_len = -1;
+	int cpu, min_cpu = -1;
 
-Core with BSP HW ID 0 is 1st logical core of 1st physical core. 
-It                      boots next and is given  SW-ID of 1.
-Core with BSP HW ID 1   boots next and is given  SW-ID of 2.
-Core with BSP HW ID 2   boots next and is given  SW-ID of 3.
-Core with BSP HW ID 3   boots next and is given  SW-ID of 4.
-...
-Core with BSP HW ID 220 is 1st logical core of 56-th physical core.
-It                      boots next and is given  SW-ID of 221.
-Core with BSP HW ID 221 boots next and is given  SW-ID of 222.
-Core with BSP HW ID 222 boots next and is given  SW-ID of 223.
-Core with BSP HW ID 223 boots next and is given  SW-ID of 224.
+	/* Find the first allowed core with the shortest run queue */
+	for (cpu = 0; cpu < num_processors; ++cpu) {
+		struct cpu_local_var *v;
+		unsigned long irqstate;
 
-Core with BSP HW ID 225 is 2nd logical core of last physical core.
-It                      boots next and is given  SW-ID of 225.
-Core with BSP HW ID 226 boots next and is given  SW-ID of 226.
-Core with BSP HW ID 227 boots next and is given  SW-ID of 227.
-*/
-ihk_spinlock_t cpuid_head_lock = 0;
-static int cpuid_head = 0;
+		if (!CPU_ISSET(cpu, cpu_set)) continue;
 
-/* archtecture-depended syscall handlers */
-int obtain_clone_cpuid() {
-    /* see above on BSP HW ID */
-	struct ihk_mc_cpu_info *cpu_info = ihk_mc_get_cpu_info();
-    int cpuid, nretry = 0;
-    ihk_mc_spinlock_lock_noirq(&cpuid_head_lock);
-	
-	/* Always start from 0 to fill in LWK cores linearily */
-	cpuid_head = 0;
- retry:
-    /* Try to obtain next physical core */
-    cpuid = cpuid_head;
+		v = get_cpu_local_var(cpu);
+		irqstate = ihk_mc_spinlock_lock(&v->runq_lock);
+		if (min_queue_len == -1 || v->runq_len < min_queue_len) {
+			min_queue_len = v->runq_len;
+			min_cpu = cpu;
+		}
+		ihk_mc_spinlock_unlock(&v->runq_lock, irqstate);
 
-    /* A hyper-threading core on the same physical core as
-       the parent process might be chosen. Use sched_setaffinity
-       if you want to skip that kind of busy physical core for
-       performance reason. */
-    cpuid_head += 1;
-    if(cpuid_head >= cpu_info->ncpus) {
-        cpuid_head = 0;
-    }
+		if (min_queue_len == 0)
+			break;
+	}
 
-    /* A hyper-threading core whose parent physical core has a
-       process on one of its hyper-threading core might
-       be chosen. Use sched_setaffinity if you want to skip that
-       kind of busy physical core for performance reason. */
-    if(get_cpu_local_var(cpuid)->status != CPU_STATUS_IDLE) {
-        nretry++;
-        if(nretry >= cpu_info->ncpus) {
-            cpuid = -1;
-            ihk_mc_spinlock_unlock_noirq(&cpuid_head_lock);
-            goto out;
-        }
-        goto retry; 
-    }
-	get_cpu_local_var(cpuid)->status = CPU_STATUS_RESERVED;
-    ihk_mc_spinlock_unlock_noirq(&cpuid_head_lock);
- out:
-    return cpuid;
+	if (min_cpu != -1) {
+		if (get_cpu_local_var(min_cpu)->status != CPU_STATUS_RESERVED)
+			get_cpu_local_var(min_cpu)->status = CPU_STATUS_RESERVED;
+	}
+
+    return min_cpu;
 }
 
 int
@@ -544,14 +510,14 @@ void ptrace_report_signal(struct thread *thread, int sig)
 	int parent_pid;
 	struct siginfo info;
 
-	dkprintf("ptrace_report_signal,pid=%d\n", thread->proc->pid);
+	dkprintf("ptrace_report_signal, tid=%d, pid=%d\n", thread->tid, thread->proc->pid);
 
 	mcs_rwlock_writer_lock(&proc->update_lock, &lock);	
 	if(!(proc->ptrace & PT_TRACED)){
 		mcs_rwlock_writer_unlock(&proc->update_lock, &lock);
 		return;
 	}
-	proc->exit_status = sig;
+	thread->exit_status = sig;
 	/* Transition thread state */
 	proc->status = PS_TRACED;
 	thread->status = PS_TRACED;
@@ -569,8 +535,8 @@ void ptrace_report_signal(struct thread *thread, int sig)
 	memset(&info, '\0', sizeof info);
 	info.si_signo = SIGCHLD;
 	info.si_code = CLD_TRAPPED;
-	info._sifields._sigchld.si_pid = thread->proc->pid;
-	info._sifields._sigchld.si_status = thread->proc->exit_status;
+	info._sifields._sigchld.si_pid = thread->tid;
+	info._sifields._sigchld.si_status = thread->exit_status;
 	do_kill(cpu_local_var(current), parent_pid, -1, SIGCHLD, &info, 0);
 	/* Wake parent (if sleeping in wait4()) */
 	waitq_wakeup(&proc->parent->waitpid_q);
@@ -695,10 +661,10 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 	int	orgsig;
 	int	ptraceflag = 0;
 	struct mcs_rwlock_node_irqsave lock;
-	unsigned long irqstate;
+	struct mcs_rwlock_node_irqsave mcs_rw_node;
 
 	for(w = pending->sigmask.__val[0], sig = 0; w; sig++, w >>= 1);
-	dkprintf("do_signal,pid=%d,sig=%d\n", proc->pid, sig);
+	dkprintf("do_signal(): tid=%d, pid=%d, sig=%d\n", thread->tid, proc->pid, sig);
 	orgsig = sig;
 
 	if((proc->ptrace & PT_TRACED) &&
@@ -718,12 +684,12 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		rc = regs->gpr.rax;
 	}
 
-	irqstate = ihk_mc_spinlock_lock(&thread->sigcommon->lock);
+	mcs_rwlock_writer_lock(&thread->sigcommon->lock, &mcs_rw_node);
 	k = thread->sigcommon->action + sig - 1;
 
 	if(k->sa.sa_handler == SIG_IGN){
 		kfree(pending);
-		ihk_mc_spinlock_unlock(&thread->sigcommon->lock, irqstate);
+		mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 		return;
 	}
 	else if(k->sa.sa_handler){
@@ -808,7 +774,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 
 		if(copy_to_user(sigsp, &ksigsp, sizeof ksigsp)){
 			kfree(pending);
-			ihk_mc_spinlock_unlock(&thread->sigcommon->lock, irqstate);
+			mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 			kprintf("do_signal,write_process_vm failed\n");
 			terminate(0, sig);
 			return;
@@ -827,7 +793,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		if(!(k->sa.sa_flags & SA_NODEFER))
 			thread->sigmask.__val[0] |= pending->sigmask.__val[0];
 		kfree(pending);
-		ihk_mc_spinlock_unlock(&thread->sigcommon->lock, irqstate);
+		mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 		if(regs->gpr.rflags & RFLAGS_TF){
 			struct siginfo info;
 
@@ -853,7 +819,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		}
 		else
 			kfree(pending);
-		ihk_mc_spinlock_unlock(&thread->sigcommon->lock, irqstate);
+		mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 		switch (sig) {
 		case SIGSTOP:
 		case SIGTSTP:
@@ -885,7 +851,8 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 				/* Wake up the parent who tried wait4 and sleeping */
 				waitq_wakeup(&proc->parent->waitpid_q);
 
-				dkprintf("do_signal,SIGSTOP,sleeping\n");
+				dkprintf("do_signal(): pid: %d, tid: %d SIGSTOP, sleeping\n", 
+					proc->pid, thread->tid);
 				/* Sleep */
 				schedule();
 				dkprintf("SIGSTOP(): woken up\n");
@@ -899,7 +866,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 
 			/* Update thread state in fork tree */
 			mcs_rwlock_writer_lock(&proc->update_lock, &lock);	
-			proc->exit_status = SIGTRAP;
+			thread->exit_status = SIGTRAP;
 			proc->status = PS_TRACED;
 			thread->status = PS_TRACED;
 			mcs_rwlock_writer_unlock(&proc->update_lock, &lock);	
@@ -953,11 +920,11 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 static struct sig_pending *
 getsigpending(struct thread *thread, int delflag){
 	struct list_head *head;
-	ihk_spinlock_t *lock;
+	mcs_rwlock_lock_t *lock;
+	struct mcs_rwlock_node_irqsave mcs_rw_node;
 	struct sig_pending *next;
 	struct sig_pending *pending;
 	__sigset_t w;
-	int	irqstate;
 	__sigset_t x;
 	int sig;
 	struct k_sigaction *k;
@@ -966,8 +933,12 @@ getsigpending(struct thread *thread, int delflag){
 
 	lock = &thread->sigcommon->lock;
 	head = &thread->sigcommon->sigpending;
-	for(;;){
-		irqstate = ihk_mc_spinlock_lock(lock);
+	for(;;) {
+		if (delflag)
+			mcs_rwlock_writer_lock(lock, &mcs_rw_node);
+		else 
+			mcs_rwlock_reader_lock(lock, &mcs_rw_node);
+
 		list_for_each_entry_safe(pending, next, head, list){
 			for(x = pending->sigmask.__val[0], sig = 0; x; sig++, x >>= 1);
 			k = thread->sigcommon->action + sig - 1;
@@ -976,17 +947,26 @@ getsigpending(struct thread *thread, int delflag){
 			   (k->sa.sa_handler != (void *)1 &&
 			    k->sa.sa_handler != NULL)){
 				if(!(pending->sigmask.__val[0] & w)){
-					if(delflag)
+					if(delflag) 
 						list_del(&pending->list);
-					ihk_mc_spinlock_unlock(lock, irqstate);
+
+					if (delflag)
+						mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
+					else 
+						mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
 					return pending;
 				}
 			}
 		}
-		ihk_mc_spinlock_unlock(lock, irqstate);
+
+		if (delflag)
+			mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
+		else 
+			mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
 
 		if(lock == &thread->sigpendinglock)
 			return NULL;
+
 		lock = &thread->sigpendinglock;
 		head = &thread->sigpending;
 	}
@@ -1034,22 +1014,25 @@ check_signal(unsigned long rc, void *regs0, int num)
 			}
 		}
 		ihk_mc_spinlock_unlock(&(cpu_local_var(runq_lock)), irqstate);
-		return;
+		goto out;
 	}
 
 	if(regs != NULL && !interrupt_from_user(regs)) {
-		return;
+		goto out;
 	}
 
 	for(;;){
 		pending = getsigpending(thread, 1);
 		if(!pending) {
 			dkprintf("check_signal,queue is empty\n");
-			return;
+			goto out;
 		}
 
 		do_signal(rc, regs, thread, pending, num);
 	}
+
+out:
+	return;
 }
 
 unsigned long
@@ -1063,7 +1046,8 @@ do_kill(struct thread *thread, int pid, int tid, int sig, siginfo_t *info,
 	struct thread *tthread = NULL;
 	int i;
 	__sigset_t mask;
-	ihk_spinlock_t *savelock = NULL;
+	mcs_rwlock_lock_t *savelock = NULL;
+	struct mcs_rwlock_node mcs_rw_node;
 	struct list_head *head = NULL;
 	int rc;
 	unsigned long irqstate = 0;
@@ -1247,7 +1231,7 @@ done:
 
 	doint = 0;
 
-	ihk_mc_spinlock_lock_noirq(savelock);
+	mcs_rwlock_writer_lock_noirq(savelock, &mcs_rw_node);
 
 	/* Put signal event even when handler is SIG_IGN or SIG_DFL
 	   because target ptraced thread must call ptrace_report_signal 
@@ -1286,7 +1270,7 @@ done:
 			}
 		}
 	}
-	ihk_mc_spinlock_unlock_noirq(savelock);
+	mcs_rwlock_writer_unlock_noirq(savelock, &mcs_rw_node);
 	cpu_restore_interrupt(irqstate);
 
 	if (doint && !(mask & tthread->sigmask.__val[0])) {

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

@@ -3,13 +3,13 @@
 # IHK SMP-x86 example boot script.
 # author: Balazs Gerofi <bgerofi@riken.jp>
 #      Copyright (C) 2014  RIKEN AICS
-# 
-# This is an example script for loading IHK, configuring a partition and 
-# booting McKernel on it. 
-# The script reserves half of the CPU cores and 512MB of RAM from NUMA node 0 
-# when IHK is loaded for the first time, otherwise it destroys the current 
-# McKernel instance and reboots it using the same set of resources as it used
-# previously. 
+#
+# This is an example script for loading IHK, configuring a partition and
+# booting McKernel on it.  Unless specific CPUs and memory are requested,
+# the script reserves half of the CPU cores and 512MB of RAM from
+# NUMA node 0 when IHK is loaded for the first time.
+# Otherwise, it destroys the current McKernel instance and reboots it using
+# the same set of resources as it used previously.
 # Note that the script does not output anything unless an error occurs.
 
 prefix="@prefix@"
@@ -23,18 +23,25 @@ ENABLE_MCOVERLAYFS="@ENABLE_MCOVERLAYFS@"
 mem="512M@0"
 cpus=""
 
+if [ "${BASH_VERSINFO[0]}" -lt 4 ]; then
+	echo "You need at least bash-4.0 to run this script." >&2
+	exit 1
+fi
+
 INTERVAL=1
 LOGMODE=0
 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"
+	irqbalance_used="yes"
 else
-irqbalance_used="no"
+	irqbalance_used="no"
 fi
 
-while getopts :i:k:c:m:o:f: OPT
+turbo=""
+
+while getopts :ti:k:c:m:o:f: OPT
 do
 	case ${OPT} in
 	f)	facility=${OPTARG}
@@ -71,11 +78,110 @@ do
 		;;
 	m) mem=${OPTARG}
 		;;
+	t) turbo="turbo"
+		;;
 	*)  echo "invalid option -${OPT}" >&2
 		exit 1
 	esac
 done
 
+#
+# Revert any state that has been initialized before the error occured.
+#
+error_exit() {
+	local status=$1
+
+	case $status in
+	mcos_sys_mounted)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			umount /tmp/mcos/mcos0_sys
+		fi
+		;&
+	mcos_proc_mounted)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			umount /tmp/mcos/mcos0_proc
+		fi
+		;&
+	mcoverlayfs_loaded)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			rmmod mcoverlay
+		fi
+		;&
+	linux_proc_bind_mounted)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			umount /tmp/mcos/linux_proc
+		fi
+		;&
+	tmp_mcos_mounted)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			umount /tmp/mcos
+		fi
+		;&
+	tmp_mcos_created)
+		if [ "$enable_mcoverlay" == "yes" ]; then
+			rm -rf /tmp/mcos
+		fi
+		;&
+	os_created)
+		# Destroy all LWK instances
+		if ls /dev/mcos* 1>/dev/null 2>&1; then
+			for i in /dev/mcos*; do
+				ind=`echo $i|cut -c10-`;
+				if ! ${SBINDIR}/ihkconfig 0 destroy $ind; then
+					echo "warning: failed to destroy LWK instance $ind" >&2
+				fi
+			done
+		fi
+		;&
+	mcctrl_loaded)
+		rmmod mcctrl || echo "warning: failed to remove mcctrl" >&2
+		;&
+	mem_reserved)
+		mem=`${SBINDIR}/ihkconfig 0 query mem`
+		if [ "${mem}" != "" ]; then
+			if ! ${SBINDIR}/ihkconfig 0 release mem $mem > /dev/null; then
+				echo "warning: failed to release memory" >&2
+			fi
+		fi
+		;&
+	cpus_reserved)
+		cpus=`${SBINDIR}/ihkconfig 0 query cpu`
+		if [ "${cpus}" != "" ]; then
+			if ! ${SBINDIR}/ihkconfig 0 release cpu $cpus > /dev/null; then
+				echo "warning: failed to release CPUs" >&2
+			fi
+		fi
+		;&
+	ihk_smp_loaded)
+		rmmod ihk_smp_x86 || echo "warning: failed to remove ihk_smp_x86" >&2
+		;&
+	ihk_loaded)
+		rmmod ihk || echo "warning: failed to remove ihk" >&2
+		;&
+	irqbalance_stopped)
+		if [ "`systemctl status irqbalance_mck.service 2> /dev/null |grep -E 'Active: active'`" != "" ]; then
+			if ! systemctl stop irqbalance_mck.service 2>/dev/null; then
+				echo "warning: failed to stop irqbalance_mck" >&2
+			fi
+			if ! systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null; then
+				echo "warning: failed to disable irqbalance_mck" >&2
+			fi
+			if ! etcdir=@ETCDIR@ perl -e '$etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "$etcdir/proc/irq/*/smp_affinity"; foreach $file (@files) { $dest = substr($file, length($etcdir)); if(0) {print "cp $file $dest\n";} system("cp $file $dest 2>/dev/null"); }'; then
+				echo "warning: failed to restore /proc/irq/*/smp_affinity" >&2
+			fi
+			if ! systemctl start irqbalance.service; then
+				echo "warning: failed to start irqbalance" >&2;
+			fi
+		fi
+		;&
+	initial)
+		# Nothing more to revert
+		;;
+	esac
+
+	exit 1
+}
+
 ihk_ikc_irq_core=0
 
 release=`uname -r`
@@ -84,13 +190,20 @@ minor=`echo ${release} | sed -e 's/^[0-9]*.\([0-9]*\).*/\1/'`
 patch=`echo ${release} | sed -e 's/^[0-9]*.[0-9]*.\([0-9]*\).*/\1/'`
 linux_version_code=`expr \( ${major} \* 65536 \) + \( ${minor} \* 256 \) + ${patch}`
 rhel_release=`echo ${release} | sed -e 's/^[0-9]*.[0-9]*.[0-9]*-\([0-9]*\).*/\1/'`
-if [ "${release}" == "${rhel_release}" ]; then rhel_release=""; fi
+if [ "${release}" == "${rhel_release}" ]; then
+	rhel_release="";
+fi
+
 enable_mcoverlay="no"
+
 if [ "${ENABLE_MCOVERLAYFS}" == "yes" ]; then
 	if [ "${rhel_release}" == "" ]; then
 		if [ ${linux_version_code} -ge 262144 -a ${linux_version_code} -lt 262400 ]; then
 			enable_mcoverlay="yes"
 		fi
+		if [ ${linux_version_code} -ge 263680 -a ${linux_version_code} -lt 263936 ]; then
+			enable_mcoverlay="yes"
+		fi
 	else
 		if [ ${linux_version_code} -eq 199168 -a ${rhel_release} -ge 327 ]; then
 			enable_mcoverlay="yes"
@@ -98,6 +211,7 @@ if [ "${ENABLE_MCOVERLAYFS}" == "yes" ]; then
 	fi
 fi
 
+# Figure out CPUs if not requested by user
 if [ "$cpus" == "" ]; then
 	# Get the number of CPUs on NUMA node 0
 	nr_cpus=`lscpu --parse | awk -F"," '{if ($4 == 0) print $4}' | wc -l`
@@ -105,7 +219,10 @@ if [ "$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?" >&2; exit 1; fi
+	if [ "$cpus" == "" ]; then
+		echo "error: no available CPUs on NUMA node 0?" >&2
+		exit 1
+	fi
 fi
 
 # Remove mcoverlay if loaded
@@ -116,135 +233,245 @@ 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" >&2; exit 1; fi
+		if ! rmmod mcoverlay; then
+			echo "error: removing mcoverlay" >&2
+			error_exit "initial"
+		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;
+    if ! systemctl stop irqbalance.service 2>/dev/null ; then
+		echo "error: stopping irqbalance" >&2
+		error_exit "initial"
+	fi;
+fi
+
+# Start mcklogd. Note that McKernel blocks when kmsg buffer is full
+# with '-k 1' until mcklogd unblocks it so starting mcklogd must preceed
+# booting McKernel
+if [ ${LOGMODE} -ne 0 ]; then
+	# Stop mcklogd which has survived McKernel shutdown because
+	# mcstop+release.sh is not used
+	pkill mcklogd
+	SBINDIR=${SBINDIR} ${SBINDIR}/mcklogd -i ${INTERVAL} -f ${facility}
 fi
 
 # Load IHK if not loaded
 if [ "`lsmod | grep ihk`" == "" ]; then
-        if ! insmod ${KMODDIR}/ihk.ko; then echo "error: loading ihk" >&2; exit 1; fi;
+	if ! insmod ${KMODDIR}/ihk.ko; then
+		echo "error: loading ihk" >&2
+		error_exit "irqbalance_stopped"
+	fi
 fi
 
+# Drop Linux caches to free memory
+sync && echo 3 > /proc/sys/vm/drop_caches
+
+# Merge free memory areas into large, physically contigous ones
+echo 1 > /proc/sys/vm/compact_memory 2>/dev/null
+
 # Load IHK-SMP if not loaded and reserve CPUs and memory
 if [ "`lsmod | grep ihk_smp_x86`" == "" ]; then
-        ihk_irq=""
-        for i in `seq 64 255`; do 
-                if [ ! -d /proc/irq/$i ] && [ "`cat /proc/interrupts | grep ":" | awk '{print $1}' | grep -o '[0-9]*' | grep -e '^$i$'`" == "" ]; then
-                        ihk_irq=$i 
-                        break
-                fi
-        done
-        if [ "$ihk_irq" == "" ]; then echo "error: no IRQ available" >&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 ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus" >&2; exit 1; fi
-        cpus_allocated=`${SBINDIR}/ihkosctl 0 query cpu`
-		if [ "$cpus_allocated" == "" ]; then
-			if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then echo "error: reserving CPUs" >&2; exit 1; fi
-		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 ${mem}; then echo "error: reserving memory" >&2; exit 1; fi
+	ihk_irq=""
+	for i in `seq 64 255`; do
+		if [ ! -d /proc/irq/$i ] && [ "`cat /proc/interrupts | grep ":" | awk '{print $1}' | grep -o '[0-9]*' | grep -e '^$i$'`" == "" ]; then
+			ihk_irq=$i
+			break
 		fi
+	done
+	if [ "$ihk_irq" == "" ]; then
+		echo "error: no IRQ available" >&2
+		error_exit "ihk_loaded"
+	fi
+	if ! insmod ${KMODDIR}/ihk-smp-x86.ko ihk_start_irq=$ihk_irq ihk_ikc_irq_core=$ihk_ikc_irq_core; then
+		echo "error: loading ihk-smp-x86" >&2
+		error_exit "ihk_loaded"
+	fi
+	if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then
+		echo "error: reserving CPUs" >&2;
+		error_exit "ihk_smp_loaded"
+	fi
+	if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then
+		echo "error: reserving memory" >&2
+		error_exit "cpus_reserved"
+	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
+	if ! insmod ${KMODDIR}/mcctrl.ko; then
+		echo "error: inserting mcctrl.ko" >&2
+		error_exit "mem_reserved"
+	fi
 fi
 
-# Check for existing OS instance and destroy
-if [ -c /dev/mcos0 ]; then
-        # Query CPU cores and memory of OS instance so that the same values are used as previously
-        if ! ${SBINDIR}/ihkosctl 0 query cpu > /dev/null; then echo "error: querying cpus" >&2; exit 1; fi
-        cpus=`${SBINDIR}/ihkosctl 0 query cpu`
-        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" >&2; fi	
-else
-        # Otherwise query IHK-SMP for resources
-        if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus" >&2; exit 1; fi
-        cpus=`${SBINDIR}/ihkconfig 0 query cpu`
-        if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then echo "error: querying memory" >&2; exit 1; fi
-        mem=`${SBINDIR}/ihkconfig 0 query mem`	
+# Destroy all LWK instances
+if ls /dev/mcos* 1>/dev/null 2>&1; then
+	for i in /dev/mcos*; do
+		ind=`echo $i|cut -c10-`;
+		if ! ${SBINDIR}/ihkconfig 0 destroy $ind; then
+			echo "error: destroying LWK instance $ind failed" >&2
+			error_exit "mcctrl_loaded"
+		fi
+	done
 fi
 
-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
+# Create OS instance
+if ! ${SBINDIR}/ihkconfig 0 create; then
+	echo "error: creating OS instance" >&2
+	error_exit "mcctrl_loaded"
+fi
 
+# Assign CPUs
+if ! ${SBINDIR}/ihkosctl 0 assign cpu ${cpus}; then
+	echo "error: assign CPUs" >&2
+	error_exit "os_created"
+fi
+
+# Assign memory
+if ! ${SBINDIR}/ihkosctl 0 assign mem ${mem}; then
+	echo "error: assign memory" >&2
+	error_exit "os_created"
+fi
+
+# Load kernel image
+if ! ${SBINDIR}/ihkosctl 0 load ${KERNDIR}/mckernel.img; then
+	echo "error: loading kernel image: ${KERNDIR}/mckernel.img" >&2
+	error_exit "os_created"
+fi
+
+# Set kernel arguments
+if ! ${SBINDIR}/ihkosctl 0 kargs "hidos ksyslogd=${LOGMODE} $turbo"; then
+	echo "error: setting kernel arguments" >&2
+	error_exit "os_created"
+fi
+
+# Boot OS instance
+if ! ${SBINDIR}/ihkosctl 0 boot; then
+	echo "error: booting" >&2
+	error_exit "os_created"
+fi
+
+# Set device file ownership
+if ! chown ${chown_option} /dev/mcd* /dev/mcos*; then
+	echo "warning: failed to chown device files" >&2
+fi
+
+# Overlay /proc, /sys with McKernel specific contents
 if [ "$enable_mcoverlay" == "yes" ]; then
 	if [ ! -e /tmp/mcos ]; then mkdir -p /tmp/mcos; fi
-	if ! mount -t tmpfs tmpfs /tmp/mcos; then echo "error: mount /tmp/mcos" >&2; exit 1; fi
+	if ! mount -t tmpfs tmpfs /tmp/mcos; then
+		echo "error: mount /tmp/mcos" >&2
+		error_exit "tmp_mcos_created"
+	fi
 	if [ ! -e /tmp/mcos/linux_proc ]; then mkdir -p /tmp/mcos/linux_proc; fi
-	if ! mount --bind /proc /tmp/mcos/linux_proc; then echo "error: mount /tmp/mcos/linux_proc" >&2; exit 1; fi
-	if ! insmod ${KMODDIR}/mcoverlay.ko; then echo "error: inserting mcoverlay.ko" >&2; exit 1; fi
+	if ! mount --bind /proc /tmp/mcos/linux_proc; then
+		echo "error: mount /tmp/mcos/linux_proc" >&2
+		error_exit "tmp_mcos_mounted"
+	fi
+	if ! insmod ${KMODDIR}/mcoverlay.ko; then
+		echo "error: inserting mcoverlay.ko" >&2
+		error_exit "linux_proc_bind_mounted"
+	fi
 	while [ ! -e /proc/mcos0 ]
 	do
-		sleep 1
+		sleep 0.1
 	done
 	if [ ! -e /tmp/mcos/mcos0_proc ]; then mkdir -p /tmp/mcos/mcos0_proc; fi
 	if [ ! -e /tmp/mcos/mcos0_proc_upper ]; then mkdir -p /tmp/mcos/mcos0_proc_upper; fi
 	if [ ! -e /tmp/mcos/mcos0_proc_work ]; then mkdir -p /tmp/mcos/mcos0_proc_work; fi
-	if ! mount -t mcoverlay mcoverlay -o lowerdir=/proc/mcos0:/proc,upperdir=/tmp/mcos/mcos0_proc_upper,workdir=/tmp/mcos/mcos0_proc_work,nocopyupw,nofscheck /tmp/mcos/mcos0_proc; then echo "error: mount /tmp/mcos/mcos0_proc" >&2; exit 1; fi
+	if ! mount -t mcoverlay mcoverlay -o lowerdir=/proc/mcos0:/proc,upperdir=/tmp/mcos/mcos0_proc_upper,workdir=/tmp/mcos/mcos0_proc_work,nocopyupw,nofscheck /tmp/mcos/mcos0_proc; then
+		echo "error: mounting /tmp/mcos/mcos0_proc" >&2
+		error_exit "mcoverlayfs_loaded"
+	fi
+	# TODO: How de we revert this in case of failure??
 	mount --make-rprivate /proc
-	while [ ! -e /sys/devices/virtual/mcos/mcos0/sys ]
+
+	while [ ! -e /sys/devices/virtual/mcos/mcos0/sys/setup_complete ]
 	do
-		sleep 1
+		sleep 0.1
 	done
 	if [ ! -e /tmp/mcos/mcos0_sys ]; then mkdir -p /tmp/mcos/mcos0_sys; fi
 	if [ ! -e /tmp/mcos/mcos0_sys_upper ]; then mkdir -p /tmp/mcos/mcos0_sys_upper; fi
 	if [ ! -e /tmp/mcos/mcos0_sys_work ]; then mkdir -p /tmp/mcos/mcos0_sys_work; fi
-	if ! mount -t mcoverlay mcoverlay -o lowerdir=/sys/devices/virtual/mcos/mcos0/sys:/sys,upperdir=/tmp/mcos/mcos0_sys_upper,workdir=/tmp/mcos/mcos0_sys_work,nocopyupw,nofscheck /tmp/mcos/mcos0_sys; then echo "error: mount /tmp/mcos/mcos0_sys" >&2; exit 1; fi
+	if ! mount -t mcoverlay mcoverlay -o lowerdir=/sys/devices/virtual/mcos/mcos0/sys:/sys,upperdir=/tmp/mcos/mcos0_sys_upper,workdir=/tmp/mcos/mcos0_sys_work,nocopyupw,nofscheck /tmp/mcos/mcos0_sys; then
+		echo "error: mount /tmp/mcos/mcos0_sys" >&2
+		error_exit "mcos_proc_mounted"
+	fi
+	# TODO: How de we revert this in case of failure??
 	mount --make-rprivate /sys
+
+	touch /tmp/mcos/mcos0_proc/mckernel
+
+	rm -rf /tmp/mcos/mcos0_sys/setup_complete
+
+	# Hide NUMA related files which are outside the LWK partition
 	for cpuid in `find /sys/devices/system/cpu/* -maxdepth 0 -name "cpu[0123456789]*" -printf "%f "`; do
 		if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/cpu/$cpuid" ]; then
 			rm -rf /tmp/mcos/mcos0_sys/devices/system/cpu/$cpuid
+			rm -rf /tmp/mcos/mcos0_sys/bus/cpu/devices/$cpuid
+			rm -rf /tmp/mcos/mcos0_sys/bus/cpu/drivers/processor/$cpuid
+		else
+			for nodeid in `find /sys/devices/system/cpu/$cpuid/* -maxdepth 0 -name "node[0123456789]*" -printf "%f "`; do
+				if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/cpu/$cpuid/$nodeid" ]; then
+					rm -f /tmp/mcos/mcos0_sys/devices/system/cpu/$cpuid/$nodeid
+				fi
+			done
 		fi
 	done
+	for nodeid in `find /sys/devices/system/node/* -maxdepth 0 -name "node[0123456789]*" -printf "%f "`; do
+		if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/node/$nodeid" ]; then
+			rm -rf /tmp/mcos/mcos0_sys/devices/system/node/$nodeid/*
+			rm -rf /tmp/mcos/mcos0_sys/bus/node/devices/$nodeid
+		else
+			# Delete non-existent symlinks
+			for cpuid in `find /sys/devices/system/node/$nodeid/* -maxdepth 0 -name "cpu[0123456789]*" -printf "%f "`; do
+				if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/node/$nodeid/$cpuid" ]; then
+					rm -f /tmp/mcos/mcos0_sys/devices/system/node/$nodeid/$cpuid
+				fi
+			done
+
+			rm -f /tmp/mcos/mcos0_sys/devices/system/node/$nodeid/memory*
+		fi
+	done
+	rm -f /tmp/mcos/mcos0_sys/devices/system/node/has_*
 	for cpuid in `find /sys/bus/cpu/devices/* -maxdepth 0 -name "cpu[0123456789]*" -printf "%f "`; do
 		if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/bus/cpu/devices/$cpuid" ]; then
 			rm -rf /tmp/mcos/mcos0_sys/bus/cpu/devices/$cpuid
 		fi
 	done
 fi
-if [ ${LOGMODE} -ne 0 ]
-then
-# mcklogd survives when McKernel isn't shut down by mcstop+release.sh
-	pkill mcklogd    
-	SBINDIR=${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;
+    if ! etcdir=@ETCDIR@ perl -e 'use File::Copy qw(copy); $etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "/proc/irq/*/smp_affinity"; foreach $file (@files) { $rel = substr($file, 1); $dir=substr($rel, 0, length($rel)-length("/smp_affinity")); if(0) { print "cp $file $etcdir/$rel\n";} if(system("mkdir -p $etcdir/$dir")){ exit 1;} if(!copy($file,"$etcdir/$rel")){ exit 1;} }'; then
+		echo "error: saving /proc/irq/*/smp_affinity" >&2
+		error_exit "mcos_sys_mounted"
+	fi;
 
     ncpus=`lscpu | grep -E '^CPU\(s\):' | awk '{print $2}'`
     smp_affinity_mask=`echo $cpus | ncpus=$ncpus perl -e 'while(<>){@tokens = split /,/;foreach $token (@tokens) {@nums = split /-/,$token; for($num = $nums[0]; $num <= $nums[$#nums]; $num++) {$ndx=int($num/32); $mask[$ndx] |= (1<<($num % 32))}}} $nint32s = int(($ENV{'ncpus'}+31)/32); for($j = $nint32s - 1; $j >= 0; $j--) { if($j != $nint32s - 1){print ",";} $nblks = $j == $nint32s - 1 ? int(($ENV{'ncpus'} % 32)/4) : 8; for($i = $nblks - 1;$i >= 0;$i--){ printf("%01x",($mask[$j] >> ($i*4)) & 0xf);}}'`
 
-    if ! ncpus=$ncpus smp_affinity_mask=$smp_affinity_mask perl -e '@dirs = grep { -d } glob "/proc/irq/*"; foreach $dir (@dirs) { $hit = 0; $affinity_str = `cat $dir/smp_affinity`; chomp $affinity_str; @int32strs = split /,/, $affinity_str; @int32strs_mask=split /,/, $ENV{'smp_affinity_mask'}; for($i=0;$i <= $#int32strs_mask; $i++) { $int32strs_inv[$i] = sprintf("%08x",hex($int32strs_mask[$i])^0xffffffff); if($i == 0) { $len = int((($ENV{'ncpus'}%32)+3)/4); $int32strs_inv[$i] = substr($int32strs_inv[$i], -$len, $len); } } $inv = join(",", @int32strs_inv); $nint32s = int(($ENV{'ncpus'}+31)/32); for($j = $nint32s - 1; $j >= 0; $j--) { if(hex($int32strs[$nint32s - 1 - $j]) & hex($int32strs_mask[$nint32s - 1 - $j])) { $hit = 1; }} if($hit == 1) { $cmd = "echo $inv > $dir/smp_affinity 2>/dev/null"; system $cmd;}}'; then echo "error: modifying /proc/irq/*/smp_affinity" >&2; exit 1; fi;
+    if ! ncpus=$ncpus smp_affinity_mask=$smp_affinity_mask perl -e '@dirs = grep { -d } glob "/proc/irq/*"; foreach $dir (@dirs) { $hit = 0; $affinity_str = `cat $dir/smp_affinity`; chomp $affinity_str; @int32strs = split /,/, $affinity_str; @int32strs_mask=split /,/, $ENV{'smp_affinity_mask'}; for($i=0;$i <= $#int32strs_mask; $i++) { $int32strs_inv[$i] = sprintf("%08x",hex($int32strs_mask[$i])^0xffffffff); if($i == 0) { $len = int((($ENV{'ncpus'}%32)+3)/4); $int32strs_inv[$i] = substr($int32strs_inv[$i], -$len, $len); } } $inv = join(",", @int32strs_inv); $nint32s = int(($ENV{'ncpus'}+31)/32); for($j = $nint32s - 1; $j >= 0; $j--) { if(hex($int32strs[$nint32s - 1 - $j]) & hex($int32strs_mask[$nint32s - 1 - $j])) { $hit = 1; }} if($hit == 1) { $cmd = "echo $inv > $dir/smp_affinity 2>/dev/null"; system $cmd;}}'; then
+		echo "error: modifying /proc/irq/*/smp_affinity" >&2
+		error_exit "mcos_sys_mounted"
+	fi
 
     banirq=`cat /proc/interrupts| perl -e 'while(<>) { if(/^\s*(\d+).*IHK\-SMP\s*$/) {print $1;}}'`
 
     sed "s/%mask%/$smp_affinity_mask/g" $ETCDIR/irqbalance_mck.in | sed "s/%banirq%/$banirq/g" > $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;
+    if ! systemctl link $ETCDIR/irqbalance_mck.service >/dev/null 2>/dev/null; then
+		echo "error: linking irqbalance_mck" >&2
+		error_exit "mcos_sys_mounted"
+	fi
+
+    if ! systemctl start irqbalance_mck.service 2>/dev/null ; then
+		echo "error: starting irqbalance_mck" >&2
+		error_exit "mcos_sys_mounted"
+	fi
 #    echo cpus=$cpus mask=$smp_affinity_mask banirq=$banirq
 fi
 

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

@@ -22,34 +22,76 @@ 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
+	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 cpu > /dev/null; then
+	echo "error: querying 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 $mem > /dev/null; then echo "error: releasing memory" >&2; exit 1; fi
+cpus=`${SBINDIR}/ihkconfig 0 query cpu`
+if [ "${cpus}" != "" ]; then
+	if ! ${SBINDIR}/ihkconfig 0 release cpu $cpus > /dev/null; then
+		echo "error: releasing CPUs" >&2
+		exit 1
+	fi
+fi
+
+if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then
+	echo "error: querying memory" >&2
+	exit 1
+fi
+
+mem=`${SBINDIR}/ihkconfig 0 query mem`
+if [ "${mem}" != "" ]; then
+	if ! ${SBINDIR}/ihkconfig 0 release mem $mem > /dev/null; then
+		echo "error: releasing memory" >&2
+		exit 1
+	fi
+fi
 
 # Remove delegator if loaded
 if [ "`lsmod | grep mcctrl`" != "" ]; then
-        if ! rmmod mcctrl; then echo "error: removing mcctrl" >&2; exit 1; fi
+	if ! rmmod mcctrl; then
+		echo "error: removing mcctrl" >&2
+		exit 1
+	fi
+fi
+
+# Remove mcoverlay if loaded
+if [ "`lsmod | grep mcoverlay`" != "" ]; then
+	if [ "`cat /proc/mounts | grep /tmp/mcos/mcos0_sys`" != "" ]; then umount -l /tmp/mcos/mcos0_sys; fi
+	if [ "`cat /proc/mounts | grep /tmp/mcos/mcos0_proc`" != "" ]; then umount -l /tmp/mcos/mcos0_proc; fi
+	if [ "`cat /proc/mounts | grep /tmp/mcos/linux_proc`" != "" ]; then umount -l /tmp/mcos/linux_proc; fi
+	if [ "`cat /proc/mounts | grep /tmp/mcos`" != "" ]; then umount -l /tmp/mcos; fi
+	if [ -e /tmp/mcos ]; then rm -rf /tmp/mcos; fi
+	if ! rmmod mcoverlay; then
+		echo "warning: failed to remove mcoverlay" >&2
+	fi
 fi
 
 # Remove SMP module
 if [ "`lsmod | grep ihk_smp_x86`" != "" ]; then
-        if ! rmmod ihk_smp_x86; then echo "error: removing ihk_smp_x86" >&2; exit 1; 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
+	if ! rmmod ihk; then
+		echo "error: removing ihk" >&2
+		exit 1
+	fi
 fi
 
 # Stop mcklogd
@@ -57,8 +99,17 @@ 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 $dest 2>/dev/null"); }' ; then echo "error: restoring /proc/irq/*/smp_affinity" >&2; exit 1; fi;
-    if ! systemctl start irqbalance.service; then echo "error: starting irqbalance" >&2; exit 1; fi;
+	if ! systemctl stop irqbalance_mck.service 2>/dev/null; then
+		echo "warning: failed to stop irqbalance_mck" >&2
+	fi
+	if ! systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null; then
+		echo "warning: failed to disable irqbalance_mck" >&2
+	fi
+	if ! etcdir=@ETCDIR@ perl -e '$etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "$etcdir/proc/irq/*/smp_affinity"; foreach $file (@files) { $dest = substr($file, length($etcdir)); if(0) {print "cp $file $dest\n";} system("cp $file $dest 2>/dev/null"); }'; then
+		echo "warning: failed to restore /proc/irq/*/smp_affinity" >&2
+	fi
+	if ! systemctl start irqbalance.service; then
+		echo "warning: failed to start irqbalance" >&2;
+	fi
 fi
+

configure

@@ -2922,6 +2922,7 @@ ac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $
 ac_compiler_gnu=$ac_cv_c_compiler_gnu
 
 	XCC=$CC
+	CFLAGS="$CFLAGS -ffreestanding -fno-tree-loop-distribute-patterns"
 	;;
     builtin-mic)
 	ARCH=k1om
@@ -3117,6 +3118,31 @@ _ACEOF
   fi
 
 
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol sys_umount" >&5
+$as_echo_n "checking System.map for symbol sys_umount... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " sys_umount\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_sys_umount\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_sys_umount $mcctrl_addr
+_ACEOF
+
+  fi
+
+
   { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol sys_unshare" >&5
 $as_echo_n "checking System.map for symbol sys_unshare... " >&6; }
   mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " sys_unshare\$" | cut -d\  -f1`
@@ -3887,11 +3913,12 @@ fi
 
 
 
+
 
 
 ac_config_headers="$ac_config_headers executer/config.h"
 
-ac_config_files="$ac_config_files Makefile executer/user/Makefile executer/kernel/mcctrl/Makefile executer/kernel/mcctrl/arch/x86_64/Makefile executer/kernel/mcoverlayfs/Makefile executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile executer/kernel/mcoverlayfs/linux-4.0.9/Makefile kernel/Makefile kernel/Makefile.build arch/x86/tools/mcreboot-attached-mic.sh arch/x86/tools/mcshutdown-attached-mic.sh arch/x86/tools/mcreboot-builtin-x86.sh arch/x86/tools/mcreboot-smp-x86.sh arch/x86/tools/mcstop+release-smp-x86.sh arch/x86/tools/mcshutdown-builtin-x86.sh arch/x86/tools/mcreboot.1:arch/x86/tools/mcreboot.1in arch/x86/tools/irqbalance_mck.service arch/x86/tools/irqbalance_mck.in"
+ac_config_files="$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 executer/kernel/mcoverlayfs/linux-4.6.7/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$enable_dcfa" = xyes; then :
@@ -4597,6 +4624,7 @@ do
     "executer/kernel/mcoverlayfs/Makefile") CONFIG_FILES="$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" ;;
+    "executer/kernel/mcoverlayfs/linux-4.6.7/Makefile") CONFIG_FILES="$CONFIG_FILES executer/kernel/mcoverlayfs/linux-4.6.7/Makefile" ;;
     "kernel/Makefile") CONFIG_FILES="$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" ;;

configure.ac

@@ -70,6 +70,7 @@ case $WITH_TARGET in
 	ARCH=`uname -m`
 	AC_PROG_CC
 	XCC=$CC
+	CFLAGS="$CFLAGS -ffreestanding -fno-tree-loop-distribute-patterns"
 	;;
     builtin-mic)
 	ARCH=k1om
@@ -221,6 +222,7 @@ AC_DEFUN([MCCTRL_FIND_KSYM],[
 ])
 
 MCCTRL_FIND_KSYM([sys_mount])
+MCCTRL_FIND_KSYM([sys_umount])
 MCCTRL_FIND_KSYM([sys_unshare])
 MCCTRL_FIND_KSYM([zap_page_range])
 MCCTRL_FIND_KSYM([vdso_image_64])
@@ -285,6 +287,7 @@ AC_SUBST(ETCDIR)
 AC_SUBST(KMODDIR)
 AC_SUBST(KERNDIR)
 AC_SUBST(MANDIR)
+AC_SUBST(CFLAGS)
 AC_SUBST(ENABLE_MCOVERLAYFS)
 
 AC_SUBST(IHK_VERSION)
@@ -304,6 +307,7 @@ AC_CONFIG_FILES([
 	executer/kernel/mcoverlayfs/Makefile
 	executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile
 	executer/kernel/mcoverlayfs/linux-4.0.9/Makefile
+	executer/kernel/mcoverlayfs/linux-4.6.7/Makefile
 	kernel/Makefile
 	kernel/Makefile.build
 	arch/x86/tools/mcreboot-attached-mic.sh

executer/config.h.in

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

executer/include/uprotocol.h

@@ -41,6 +41,8 @@
 #define MCEXEC_UP_NEW_PROCESS	 0x30a02909
 #define MCEXEC_UP_GET_CRED	 0x30a0290a
 #define MCEXEC_UP_GET_CREDV	 0x30a0290b
+#define MCEXEC_UP_GET_NODES  0x30a0290c
+#define MCEXEC_UP_GET_CPUSET  0x30a0290d
 
 #define MCEXEC_UP_PREPARE_DMA    0x30a02910
 #define MCEXEC_UP_FREE_DMA       0x30a02911
@@ -49,7 +51,8 @@
 #define MCEXEC_UP_CLOSE_EXEC     0x30a02913
 
 #define MCEXEC_UP_SYS_MOUNT      0x30a02914
-#define MCEXEC_UP_SYS_UNSHARE    0x30a02915
+#define MCEXEC_UP_SYS_UMOUNT     0x30a02915
+#define MCEXEC_UP_SYS_UNSHARE    0x30a02916
 
 #define MCEXEC_UP_DEBUG_LOG     0x40000000
 
@@ -77,6 +80,17 @@ struct program_image_section {
 #define SHELL_PATH_MAX_LEN	1024
 #define MCK_RLIM_MAX	20
 
+struct get_cpu_set_arg {
+	int nr_processes;
+	void *cpu_set;
+	size_t cpu_set_size;	// Size in bytes
+	int *target_core;
+};
+
+#define PLD_CPU_SET_MAX_CPUS 1024
+typedef unsigned long __cpu_set_unit;
+#define PLD_CPU_SET_SIZE (PLD_CPU_SET_MAX_CPUS / (8 * sizeof(__cpu_set_unit)))
+
 struct program_load_desc {
 	int num_sections;
 	int status;
@@ -106,6 +120,7 @@ struct program_load_desc {
 	struct rlimit rlimit[MCK_RLIM_MAX];
 	unsigned long interp_align;
 	char shell_path[SHELL_PATH_MAX_LEN];
+	__cpu_set_unit cpu_set[PLD_CPU_SET_SIZE];
 	struct program_image_section sections[0];
 };
 
@@ -196,6 +211,10 @@ struct sys_mount_desc {
 	void *data;
 };
 
+struct sys_umount_desc {
+	char *dir_name;
+};
+
 struct sys_unshare_desc {
 	unsigned long unshare_flags;
 };

executer/kernel/mcctrl/binfmt_mcexec.c

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

executer/kernel/mcctrl/control.c

@@ -34,6 +34,7 @@
 #include <linux/version.h>
 #include <linux/semaphore.h>
 #include <linux/interrupt.h>
+#include <linux/cpumask.h>
 #include <asm/uaccess.h>
 #include <asm/delay.h>
 #include <asm/io.h>
@@ -66,7 +67,18 @@ int (*mcctrl_sys_mount)(char *dev_name,char *dir_name, char *type, unsigned long
         (int_star_fn_char_char_char_ulong_void_t)
         MCCTRL_KSYM_sys_mount;
 #else // exported
-int (*mcctrl_sys_mount)(char *dev_name,char *dir_name, char *type, unsigned long flags, void *data) = NULL;
+int (*mcctrl_sys_mount)(char *dev_name,char *dir_name, char *type, unsigned long flags, void *data) = sys_mount;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM_sys_umount
+#if MCCTRL_KSYM_sys_umount
+typedef int (*int_fn_char_star_int_t)(char *, int);
+int (*mcctrl_sys_umount)(char *dir_name, int flags) =
+        (int_fn_char_star_int_t)
+        MCCTRL_KSYM_sys_umount;
+#else // exported
+int (*mcctrl_sys_umount)(char *dir_name, int flags) = sys_umount;
 #endif
 #endif
 
@@ -77,32 +89,49 @@ int mcctrl_ikc_set_recv_cpu(ihk_os_t os, int cpu);
 static long mcexec_prepare_image(ihk_os_t os,
                                  struct program_load_desc * __user udesc)
 {
-	struct program_load_desc desc, *pdesc;
+	struct program_load_desc *desc, *pdesc;
 	struct ikc_scd_packet isp;
 	void *args, *envs;
 	long ret = 0;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct mcctrl_per_proc_data *ppd = NULL;
+	int num_sections;
 
-	if (copy_from_user(&desc, udesc,
+	desc = kmalloc(sizeof(*desc), GFP_KERNEL);
+	if (!desc) {
+		printk("%s: error: allocating program_load_desc\n",
+			__FUNCTION__);
+		return -ENOMEM;
+	}
+
+	if (copy_from_user(desc, udesc,
 	                    sizeof(struct program_load_desc))) {
+		printk("%s: error: copying program_load_desc\n",
+			__FUNCTION__);
+		kfree(desc);
 		return -EFAULT;
 	}
-	if (desc.num_sections <= 0 || desc.num_sections > 16) {
-		printk("# of sections: %d\n", desc.num_sections);
+
+	num_sections = desc->num_sections;
+
+	if (num_sections <= 0 || num_sections > 16) {
+		printk("# of sections: %d\n", num_sections);
 		return -EINVAL;
 	}
 	pdesc = kmalloc(sizeof(struct program_load_desc) + 
 	                sizeof(struct program_image_section)
-	                * desc.num_sections, GFP_KERNEL);
-	memcpy(pdesc, &desc, sizeof(struct program_load_desc));
+	                * num_sections, GFP_KERNEL);
+	memcpy(pdesc, desc, sizeof(struct program_load_desc));
 	if (copy_from_user(pdesc->sections, udesc->sections,
 	                   sizeof(struct program_image_section)
-	                   * desc.num_sections)) {
+	                   * num_sections)) {
+		kfree(desc);
 		kfree(pdesc);
 		return -EFAULT;
 	}
 
+	kfree(desc);
+
 	pdesc->pid = task_tgid_vnr(current);
 
 	if (reserve_user_space(usrdata, &pdesc->user_start, &pdesc->user_end)) {
@@ -158,7 +187,7 @@ static long mcexec_prepare_image(ihk_os_t os,
 	ppd->rpgtable = pdesc->rpgtable;
 	
 	if (copy_to_user(udesc, pdesc, sizeof(struct program_load_desc) + 
-	             sizeof(struct program_image_section) * desc.num_sections)) {
+	             sizeof(struct program_image_section) * num_sections)) {
 		ret = -EFAULT;	
 		goto free_out;
 	}
@@ -264,8 +293,9 @@ int mcexec_transfer_image(ihk_os_t os, struct remote_transfer *__user upt)
 
 //extern unsigned long last_thread_exec;
 
-struct handlerinfo {
-	int	pid;
+struct release_handler_info {
+	int pid;
+	int cpu;
 };
 
 static long mcexec_debug_log(ihk_os_t os, unsigned long arg)
@@ -281,7 +311,7 @@ static long mcexec_debug_log(ihk_os_t os, unsigned long arg)
 
 static void release_handler(ihk_os_t os, void *param)
 {
-	struct handlerinfo *info = param;
+	struct release_handler_info *info = param;
 	struct ikc_scd_packet isp;
 	int os_ind = ihk_host_os_get_index(os);
 
@@ -289,10 +319,15 @@ static void release_handler(ihk_os_t os, void *param)
 	isp.msg = SCD_MSG_CLEANUP_PROCESS;
 	isp.pid = info->pid;
 
-	mcctrl_ikc_send(os, 0, &isp);
-	if(os_ind >= 0)
+	dprintk("%s: SCD_MSG_CLEANUP_PROCESS, info: %p, cpu: %d\n",
+			__FUNCTION__, info, info->cpu);
+	mcctrl_ikc_send(os, info->cpu, &isp);
+	if (os_ind >= 0) {
 		delete_pid_entry(os_ind, info->pid);
+	}
 	kfree(param);
+	dprintk("%s: SCD_MSG_CLEANUP_PROCESS, info: %p OK\n",
+			__FUNCTION__, info);
 }
 
 static long mcexec_newprocess(ihk_os_t os,
@@ -300,12 +335,12 @@ static long mcexec_newprocess(ihk_os_t os,
                               struct file *file)
 {
 	struct newprocess_desc desc;
-	struct handlerinfo *info;
+	struct release_handler_info *info;
 
 	if (copy_from_user(&desc, udesc, sizeof(struct newprocess_desc))) {
 		return -EFAULT;
 	}
-	info = kmalloc(sizeof(struct handlerinfo), GFP_KERNEL);
+	info = kmalloc(sizeof(struct release_handler_info), GFP_KERNEL);
 	info->pid = desc.pid;
 	ihk_os_register_release_handler(file, release_handler, info);
 	return 0;
@@ -315,33 +350,43 @@ static long mcexec_start_image(ihk_os_t os,
                                struct program_load_desc * __user udesc,
                                struct file *file)
 {
-	struct program_load_desc desc;
+	struct program_load_desc *desc;
 	struct ikc_scd_packet isp;
 	struct mcctrl_channel *c;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
-	struct handlerinfo *info;
+	struct release_handler_info *info;
 
-	if (copy_from_user(&desc, udesc,
+	desc = kmalloc(sizeof(*desc), GFP_KERNEL);
+	if (!desc) {
+		printk("%s: error: allocating program_load_desc\n",
+			__FUNCTION__);
+		return -ENOMEM;
+	}
+
+	if (copy_from_user(desc, udesc,
 	                   sizeof(struct program_load_desc))) {
+		kfree(desc);
 		return -EFAULT;
 	}
 
-	info = kmalloc(sizeof(struct handlerinfo), GFP_KERNEL);
-	info->pid = desc.pid;
+	info = kmalloc(sizeof(struct release_handler_info), GFP_KERNEL);
+	info->pid = desc->pid;
+	info->cpu = desc->cpu;
 	ihk_os_register_release_handler(file, release_handler, info);
 
-	c = usrdata->channels + desc.cpu;
+	c = usrdata->channels + desc->cpu;
 
-	mcctrl_ikc_set_recv_cpu(os, desc.cpu);
+	mcctrl_ikc_set_recv_cpu(os, desc->cpu);
 
-	usrdata->last_thread_exec = desc.cpu;
+	usrdata->last_thread_exec = desc->cpu;
 	
 	isp.msg = SCD_MSG_SCHEDULE_PROCESS;
-	isp.ref = desc.cpu;
-	isp.arg = desc.rprocess;
+	isp.ref = desc->cpu;
+	isp.arg = desc->rprocess;
 
-	mcctrl_ikc_send(os, desc.cpu, &isp);
+	mcctrl_ikc_send(os, desc->cpu, &isp);
 
+	kfree(desc);
 	return 0;
 }
 
@@ -413,6 +458,209 @@ static long mcexec_get_cpu(ihk_os_t os)
 	return info->n_cpus;
 }
 
+static long mcexec_get_nodes(ihk_os_t os)
+{
+	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
+
+	if (!usrdata || !usrdata->mem_info)
+		return -EINVAL;
+
+	return usrdata->mem_info->n_numa_nodes;
+}
+
+extern int linux_numa_2_mckernel_numa(struct mcctrl_usrdata *udp, int numa_id);
+extern int mckernel_cpu_2_linux_cpu(struct mcctrl_usrdata *udp, int cpu_id);
+
+static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
+{
+	struct mcctrl_usrdata *udp = ihk_host_os_get_usrdata(os);
+	struct mcctrl_part_exec *pe;
+	struct get_cpu_set_arg req;
+	struct cpu_topology *cpu_top, *cpu_top_i;
+	struct cache_topology *cache_top;
+	int cpu, cpus_assigned, cpus_to_assign, cpu_prev;
+	int ret = 0;
+	cpumask_t cpus_used;
+	cpumask_t cpus_to_use;
+	struct mcctrl_per_proc_data *ppd;
+
+	if (!udp) {
+		return -EINVAL;
+	}
+
+	/* Look up per-process structure */
+	ppd = mcctrl_get_per_proc_data(udp, task_tgid_vnr(current));
+	if (!ppd) {
+		return -EINVAL;
+	}
+
+	pe = &udp->part_exec;
+
+	if (copy_from_user(&req, (void *)arg, sizeof(req))) {
+		printk("%s: error copying user request\n", __FUNCTION__);
+		return -EINVAL;
+	}
+
+	mutex_lock(&pe->lock);
+
+	memcpy(&cpus_used, &pe->cpus_used, sizeof(cpumask_t));
+	memset(&cpus_to_use, 0, sizeof(cpus_to_use));
+
+	/* First process to enter CPU partitioning */
+	if (pe->nr_processes == -1) {
+		pe->nr_processes = req.nr_processes;
+		pe->nr_processes_left = req.nr_processes;
+		dprintk("%s: nr_processes: %d (partitioned exec starts)\n",
+				__FUNCTION__,
+				pe->nr_processes);
+	}
+
+	if (pe->nr_processes != req.nr_processes) {
+		printk("%s: error: requested number of processes"
+				" doesn't match current partitioned execution\n",
+				__FUNCTION__);
+		ret = -EINVAL;
+		goto unlock_out;
+	}
+
+	--pe->nr_processes_left;
+	dprintk("%s: nr_processes: %d, nr_processes_left: %d\n",
+			__FUNCTION__,
+			pe->nr_processes,
+			pe->nr_processes_left);
+
+	cpus_to_assign = udp->cpu_info->n_cpus / req.nr_processes;
+
+	/* Find the first unused CPU */
+	cpu = cpumask_next_zero(-1, &cpus_used);
+	if (cpu >= udp->cpu_info->n_cpus) {
+		printk("%s: error: no more CPUs available\n",
+				__FUNCTION__);
+		ret = -EINVAL;
+		goto unlock_out;
+	}
+
+	cpu_set(cpu, cpus_used);
+	cpu_set(cpu, cpus_to_use);
+	cpu_prev = cpu;
+	dprintk("%s: CPU %d assigned (first)\n", __FUNCTION__, cpu);
+
+	for (cpus_assigned = 1; cpus_assigned < cpus_to_assign;
+			++cpus_assigned) {
+		int node;
+
+		cpu_top = NULL;
+		/* Find the topology object of the last core assigned */
+		list_for_each_entry(cpu_top_i, &udp->cpu_topology_list, chain) {
+			if (cpu_top_i->mckernel_cpu_id == cpu_prev) {
+				cpu_top = cpu_top_i;
+				break;
+			}
+		}
+
+		if (!cpu_top) {
+			printk("%s: error: couldn't find CPU topology info\n",
+					__FUNCTION__);
+			ret = -EINVAL;
+			goto unlock_out;
+		}
+
+		/* Find a core sharing the same cache iterating caches from
+		 * the most inner one outwards */
+		list_for_each_entry(cache_top, &cpu_top->cache_list, chain) {
+			for_each_cpu(cpu, &cache_top->shared_cpu_map) {
+				if (!cpu_isset(cpu, cpus_used)) {
+					cpu_set(cpu, cpus_used);
+					cpu_set(cpu, cpus_to_use);
+					cpu_prev = cpu;
+					dprintk("%s: CPU %d assigned (same cache L%lu)\n",
+						__FUNCTION__, cpu, cache_top->saved->level);
+					goto next_cpu;
+				}
+			}
+		}
+
+		/* No CPU? Find a core from the same NUMA node */
+		node = linux_numa_2_mckernel_numa(udp,
+				cpu_to_node(mckernel_cpu_2_linux_cpu(udp, cpu_prev)));
+
+		for_each_cpu_not(cpu, &cpus_used) {
+			/* Invalid CPU? */
+			if (cpu >= udp->cpu_info->n_cpus)
+				break;
+
+			/* Found one */
+			if (node == linux_numa_2_mckernel_numa(udp,
+						cpu_to_node(mckernel_cpu_2_linux_cpu(udp, cpu)))) {
+				cpu_set(cpu, cpus_used);
+				cpu_set(cpu, cpus_to_use);
+				cpu_prev = cpu;
+				dprintk("%s: CPU %d assigned (same NUMA)\n",
+						__FUNCTION__, cpu);
+				goto next_cpu;
+			}
+		}
+
+		/* No CPU? Simply find the next unused one */
+		cpu = cpumask_next_zero(-1, &cpus_used);
+		if (cpu >= udp->cpu_info->n_cpus) {
+			printk("%s: error: no more CPUs available\n",
+					__FUNCTION__);
+			ret = -EINVAL;
+			goto unlock_out;
+		}
+
+		cpu_set(cpu, cpus_used);
+		cpu_set(cpu, cpus_to_use);
+		cpu_prev = cpu;
+		dprintk("%s: CPU %d assigned (unused)\n",
+				__FUNCTION__, cpu);
+
+next_cpu:
+		continue;
+	}
+
+	/* Found all cores, let user know */
+	if (copy_to_user(req.cpu_set, &cpus_to_use,
+				(req.cpu_set_size < sizeof(cpus_to_use) ?
+				 req.cpu_set_size : sizeof(cpus_to_use)))) {
+		printk("%s: error copying mask to user\n", __FUNCTION__);
+		ret = -EINVAL;
+		goto unlock_out;
+	}
+
+	cpu = cpumask_next(-1, &cpus_to_use);
+	if (copy_to_user(req.target_core, &cpu, sizeof(cpu))) {
+		printk("%s: error copying target core to user\n",
+				__FUNCTION__);
+		ret = -EINVAL;
+		goto unlock_out;
+	}
+
+	/* Save in per-process structure */
+	memcpy(&ppd->cpu_set, &cpus_to_use, sizeof(cpumask_t));
+	ppd->ikc_target_cpu = cpu;
+
+	/* Commit used cores to OS structure */
+	memcpy(&pe->cpus_used, &cpus_used, sizeof(cpus_used));
+
+	/* Reset if last process */
+	if (pe->nr_processes_left == 0) {
+		dprintk("%s: nr_processes: %d (partitioned exec ends)\n",
+				__FUNCTION__,
+				pe->nr_processes);
+		pe->nr_processes = -1;
+		memset(&pe->cpus_used, 0, sizeof(pe->cpus_used));
+	}
+
+	ret = 0;
+
+unlock_out:
+	mutex_unlock(&pe->lock);
+
+	return ret;
+}
+
 int mcctrl_add_per_proc_data(struct mcctrl_usrdata *ud, int pid, 
 	struct mcctrl_per_proc_data *ppd)
 {
@@ -501,9 +749,10 @@ int mcexec_syscall(struct mcctrl_usrdata *ud, struct ikc_scd_packet *packet)
 	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;
+		kprintf("%s: ERROR: no per-process structure for PID %d, "
+				"syscall nr: %lu\n",
+				__FUNCTION__, pid, packet->req.number);
+		return -1;
 	}
 
 	dprintk("%s: (packet_handler) rtid: %d, ttid: %d, sys nr: %d\n",
@@ -930,6 +1179,8 @@ int mcexec_open_exec(ihk_os_t os, char * __user filename)
 		INIT_LIST_HEAD(&ppd->wq_req_list);
 		INIT_LIST_HEAD(&ppd->wq_list_exact);
 		spin_lock_init(&ppd->wq_list_lock);
+		memset(&ppd->cpu_set, 0, sizeof(cpumask_t));
+		ppd->ikc_target_cpu = 0;
 
 		for (i = 0; i < MCCTRL_PER_THREAD_DATA_HASH_SIZE; ++i) {
 			INIT_LIST_HEAD(&ppd->per_thread_data_hash[i]);
@@ -1127,7 +1378,7 @@ long mcexec_sys_mount(struct sys_mount_desc *__user arg)
 	cap_raise(promoted->cap_effective, CAP_SYS_ADMIN);
 	original = override_creds(promoted);
 
-#if MCCTRL_KSYM_sys_mount
+#ifdef MCCTRL_KSYM_sys_mount
 	ret = mcctrl_sys_mount(desc.dev_name, desc.dir_name, desc.type,
 		desc.flags, desc.data);
 #else
@@ -1140,6 +1391,36 @@ long mcexec_sys_mount(struct sys_mount_desc *__user arg)
 	return ret;
 }
 
+long mcexec_sys_umount(struct sys_mount_desc *__user arg)
+{
+	struct sys_umount_desc desc;
+	struct cred *promoted;
+	const struct cred *original;
+	int ret;
+
+	if (copy_from_user(&desc, arg, sizeof(desc))) {
+		return -EFAULT;
+	}
+
+	promoted = prepare_creds();
+	if (!promoted) {
+		return -ENOMEM;
+	}
+	cap_raise(promoted->cap_effective, CAP_SYS_ADMIN);
+	original = override_creds(promoted);
+
+#ifdef MCCTRL_KSYM_sys_umount
+	ret = mcctrl_sys_umount(desc.dir_name, MNT_FORCE);
+#else
+	ret = -EFAULT;
+#endif
+
+	revert_creds(original);
+	put_cred(promoted);
+
+	return ret;
+}
+
 long mcexec_sys_unshare(struct sys_unshare_desc *__user arg)
 {
 	struct sys_unshare_desc desc;
@@ -1198,6 +1479,12 @@ long __mcctrl_control(ihk_os_t os, unsigned int req, unsigned long arg,
 	case MCEXEC_UP_GET_CPU:
 		return mcexec_get_cpu(os);
 
+	case MCEXEC_UP_GET_NODES:
+		return mcexec_get_nodes(os);
+
+	case MCEXEC_UP_GET_CPUSET:
+		return mcexec_get_cpuset(os, arg);
+
 	case MCEXEC_UP_STRNCPY_FROM_USER:
 		return mcexec_strncpy_from_user(os, 
 				(struct strncpy_from_user_desc *)arg);
@@ -1227,6 +1514,9 @@ long __mcctrl_control(ihk_os_t os, unsigned int req, unsigned long arg,
 	case MCEXEC_UP_SYS_MOUNT:
 		return mcexec_sys_mount((struct sys_mount_desc *)arg);
 
+	case MCEXEC_UP_SYS_UMOUNT:
+		return mcexec_sys_umount((struct sys_mount_desc *)arg);
+
 	case MCEXEC_UP_SYS_UNSHARE:
 		return mcexec_sys_unshare((struct sys_unshare_desc *)arg);
 

executer/kernel/mcctrl/driver.c

@@ -60,6 +60,8 @@ static struct ihk_os_user_call_handler mcctrl_uchs[] = {
 	{ .request = MCEXEC_UP_LOAD_SYSCALL, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SEND_SIGNAL, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_GET_CPU, .func = mcctrl_ioctl },
+	{ .request = MCEXEC_UP_GET_NODES, .func = mcctrl_ioctl },
+	{ .request = MCEXEC_UP_GET_CPUSET, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_STRNCPY_FROM_USER, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_NEW_PROCESS, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_PREPARE_DMA, .func = mcctrl_ioctl },
@@ -69,6 +71,7 @@ static struct ihk_os_user_call_handler mcctrl_uchs[] = {
 	{ .request = MCEXEC_UP_GET_CRED, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_GET_CREDV, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SYS_MOUNT, .func = mcctrl_ioctl },
+	{ .request = MCEXEC_UP_SYS_UMOUNT, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SYS_UNSHARE, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_DEBUG_LOG, .func = mcctrl_ioctl },
 };
@@ -129,11 +132,15 @@ error_cleanup_channels:
 
 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);
+	if (os[os_index]) {
+		sysfsm_cleanup(os[os_index]);
+		free_topology_info(os[os_index]);
+		ihk_os_unregister_user_call_handlers(os[os_index], mcctrl_uc + os_index);
+		destroy_ikc_channels(os[os_index]);
+		procfs_exit(os_index);
+	}
+
+	os[os_index] = NULL;
 
 	printk("mcctrl: OS ID %d shutdown event handled\n", os_index);
 	return 0;
@@ -151,11 +158,16 @@ static struct ihk_os_notifier mcctrl_os_notifier = {
 static int __init mcctrl_init(void)
 {
 	int ret = 0;
+	int i;
 
 #ifndef DO_USER_MODE
 	mcctrl_syscall_init();
 #endif
 
+	for (i = 0; i < OS_MAX_MINOR; ++i) {
+		os[i] = NULL;
+	}
+
 	rus_page_hash_init();
 
 	binfmt_mcexec_init();

executer/kernel/mcctrl/ikc.c

@@ -35,6 +35,16 @@
 
 #define REQUEST_SHIFT    16
 
+//#define DEBUG_IKC
+
+#ifdef DEBUG_IKC
+#define	dkprintf(...) kprintf(__VA_ARGS__)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) printk(__VA_ARGS__); } while (0)
+#define	ekprintf(...) printk(__VA_ARGS__)
+#endif
+
 //int num_channels;
 
 //struct mcctrl_channel *channels;
@@ -99,10 +109,6 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		get_vdso_info(__os, pisp->arg);
 		break;
 
-	case SCD_MSG_REPLY_GET_CPU_MAPPING:
-		reply_get_cpu_mapping(pisp->arg);
-		break;
-
 	default:
 		printk(KERN_ERR "mcctrl:syscall_packet_handler:"
 				"unknown message (%d.%d.%d.%d.%d.%#lx)\n",
@@ -168,93 +174,26 @@ int mcctrl_ikc_is_valid_thread(ihk_os_t os, int cpu)
 	}
 }
 
-//unsigned long *mcctrl_doorbell_va;
-//unsigned long mcctrl_doorbell_pa;
-
 static void mcctrl_ikc_init(ihk_os_t os, int cpu, unsigned long rphys, struct ihk_ikc_channel_desc *c)
 {
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct ikc_scd_packet packet;
 	struct mcctrl_channel *pmc = usrdata->channels + cpu;
-	unsigned long phys;
-	struct ikc_scd_init_param *rpm;
 
-	if(c->port == 502)
+	if (c->port == 502) {
 		pmc = usrdata->channels + usrdata->num_channels - 1;
-
-	if (!pmc) {
-		return;
 	}
 
-	printk("IKC init: cpu=%d port=%d\n", cpu, c->port);
-
-	phys = ihk_device_map_memory(ihk_os_to_dev(os), rphys,
-	                             sizeof(struct ikc_scd_init_param));
-#ifdef CONFIG_MIC
-	rpm = ioremap_wc(phys, sizeof(struct ikc_scd_init_param));
-#else
-	rpm = ihk_device_map_virtual(ihk_os_to_dev(os), phys, 
-	                             sizeof(struct ikc_scd_init_param),
-								 NULL, 0);
-#endif
-
-	pmc->param.request_va =
-		(void *)__get_free_pages(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(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);
-	memset(pmc->param.post_va, 0, PAGE_SIZE);
-	
-	pmc->param.response_rpa = rpm->response_page;
-	pmc->param.response_pa 
-		= ihk_device_map_memory(ihk_os_to_dev(os),
-		                        pmc->param.response_rpa,
-		                        PAGE_SIZE);
-#ifdef CONFIG_MIC							
-	pmc->param.response_va = ioremap_cache(pmc->param.response_pa,
-	                                       PAGE_SIZE);
-#else
-	pmc->param.response_va = ihk_device_map_virtual(ihk_os_to_dev(os), 
-	                                                pmc->param.response_pa,
-													PAGE_SIZE, NULL, 0);
-#endif
-
-	pmc->dma_buf = (void *)__get_free_pages(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;
-	rpm->post_page = pmc->param.post_pa;
+	if (!pmc) {
+		kprintf("%s: error: no channel found?\n", __FUNCTION__);
+		return;
+	}
 
 	packet.msg = SCD_MSG_INIT_CHANNEL_ACKED;
 	packet.ref = cpu;
 	packet.arg = rphys;
 
-	printk("Request: %lx, Response: %lx, Doorbell: %lx\n",
-	       pmc->param.request_pa, pmc->param.response_rpa,
-	       pmc->param.doorbell_pa);
-	printk("Request: %p, Response: %p, Doorbell: %p\n",
-	       pmc->param.request_va, pmc->param.response_va,
-	       pmc->param.doorbell_va);
-
 	ihk_ikc_send(pmc->c, &packet, 0);
-
-#ifdef CONFIG_MIC							
-	iounmap(rpm);
-#else
-	ihk_device_unmap_virtual(ihk_os_to_dev(os), rpm, 
-	                         sizeof(struct ikc_scd_init_param));
-#endif
-
-	ihk_device_unmap_memory(ihk_os_to_dev(os), phys,
-	                        sizeof(struct ikc_scd_init_param));
 }
 
 static int connect_handler(struct ihk_ikc_channel_info *param)
@@ -274,7 +213,7 @@ static int connect_handler(struct ihk_ikc_channel_info *param)
 	param->packet_handler = syscall_packet_handler;
 	
 	usrdata->channels[cpu].c = c;
-	kprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
+	dkprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
 
 	return 0;
 }
@@ -292,7 +231,7 @@ static int connect_handler2(struct ihk_ikc_channel_info *param)
 	param->packet_handler = syscall_packet_handler;
 	
 	usrdata->channels[cpu].c = c;
-	kprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
+	dkprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
 
 	return 0;
 }
@@ -301,7 +240,7 @@ static struct ihk_ikc_listen_param listen_param = {
 	.port = 501,
 	.handler = connect_handler,
 	.pkt_size = sizeof(struct ikc_scd_packet),
-	.queue_size = PAGE_SIZE,
+	.queue_size = PAGE_SIZE * 4,
 	.magic = 0x1129,
 };
 
@@ -309,33 +248,35 @@ static struct ihk_ikc_listen_param listen_param2 = {
 	.port = 502,
 	.handler = connect_handler2,
 	.pkt_size = sizeof(struct ikc_scd_packet),
-	.queue_size = PAGE_SIZE,
+	.queue_size = PAGE_SIZE * 4,
 	.magic = 0x1329,
 };
 
 int prepare_ikc_channels(ihk_os_t os)
 {
-	struct ihk_cpu_info *info;
-	struct mcctrl_usrdata   *usrdata;
+	struct mcctrl_usrdata *usrdata;
 	int i;
 
 	usrdata = kzalloc(sizeof(struct mcctrl_usrdata), GFP_KERNEL);
-	usrdata->mcctrl_doorbell_va = (void *)__get_free_page(GFP_KERNEL);
-	usrdata->mcctrl_doorbell_pa = virt_to_phys(usrdata->mcctrl_doorbell_va);
 
-	info = ihk_os_get_cpu_info(os);
-	if (!info) {
-		printk("Error: cannot retrieve CPU info.\n");
+	usrdata->cpu_info = ihk_os_get_cpu_info(os);
+	usrdata->mem_info = ihk_os_get_memory_info(os);
+
+	if (!usrdata->cpu_info || !usrdata->mem_info) {
+		printk("Error: cannot obtain OS CPU and memory information.\n");
 		return -EINVAL;
 	}
-	if (info->n_cpus < 1) {
+
+	if (usrdata->cpu_info->n_cpus < 1) {
 		printk("Error: # of cpu is invalid.\n");
 		return -EINVAL;
 	}
 
-	usrdata->num_channels = info->n_cpus + 1;
-	usrdata->channels = kzalloc(sizeof(struct mcctrl_channel) * usrdata->num_channels,
-	                   GFP_KERNEL);
+	usrdata->num_channels = usrdata->cpu_info->n_cpus + 1;
+	usrdata->channels = kzalloc(sizeof(struct mcctrl_channel) *
+			usrdata->num_channels,
+			GFP_KERNEL);
+
 	if (!usrdata->channels) {
 		printk("Error: cannot allocate channels.\n");
 		return -ENOMEM;
@@ -357,25 +298,15 @@ int prepare_ikc_channels(ihk_os_t os)
 	INIT_LIST_HEAD(&usrdata->cpu_topology_list);
 	INIT_LIST_HEAD(&usrdata->node_topology_list);
 
+	mutex_init(&usrdata->part_exec.lock);
+	usrdata->part_exec.nr_processes = -1;
+
 	return 0;
 }
 
 void __destroy_ikc_channel(ihk_os_t os, struct mcctrl_channel *pmc)
 {
-	free_pages((unsigned long)pmc->param.request_va,
-	           REQUEST_SHIFT - PAGE_SHIFT);
-	free_page((unsigned long)pmc->param.post_va);
-
-#ifdef CONFIG_MIC
-	iounmap(pmc->param.response_va);
-#else
-	ihk_device_unmap_virtual(ihk_os_to_dev(os), pmc->param.response_va, 
-	                         PAGE_SIZE);
-#endif
-	ihk_device_unmap_memory(ihk_os_to_dev(os),
-	                        pmc->param.response_pa, PAGE_SIZE);
-	free_pages((unsigned long)pmc->dma_buf,
-	           DMA_PIN_SHIFT - PAGE_SHIFT);
+	return;
 }
 
 void destroy_ikc_channels(ihk_os_t os)
@@ -383,6 +314,11 @@ void destroy_ikc_channels(ihk_os_t os)
 	int i;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 
+	if (!usrdata) {
+		printk("%s: WARNING: no mcctrl_usrdata found\n", __FUNCTION__);
+		return;
+	}
+
 	ihk_host_os_set_usrdata(os, NULL);
 
 	for (i = 0; i < usrdata->num_channels; i++) {
@@ -393,7 +329,6 @@ void destroy_ikc_channels(ihk_os_t os)
 			printk("Channel #%d freed.\n", i);
 		}
 	}
-	free_page((unsigned long)usrdata->mcctrl_doorbell_va);
 
 	kfree(usrdata->channels);
 	kfree(usrdata);

executer/kernel/mcctrl/mcctrl.h

@@ -59,8 +59,8 @@
 #define SCD_MSG_CLEANUP_PROCESS         0x9
 #define SCD_MSG_GET_VDSO_INFO           0xa
 
-#define SCD_MSG_GET_CPU_MAPPING         0xc
-#define SCD_MSG_REPLY_GET_CPU_MAPPING   0xd
+//#define SCD_MSG_GET_CPU_MAPPING         0xc
+//#define SCD_MSG_REPLY_GET_CPU_MAPPING   0xd
 
 #define	SCD_MSG_PROCFS_CREATE		0x10
 #define	SCD_MSG_PROCFS_DELETE		0x11
@@ -172,7 +172,6 @@ struct wait_queue_head_list_node {
 
 struct mcctrl_channel {
 	struct ihk_ikc_channel_desc *c;
-	struct syscall_params param;
 	struct ikc_scd_init_param init;
 	void *dma_buf;
 };
@@ -199,6 +198,8 @@ struct mcctrl_per_proc_data {
 
 	struct list_head per_thread_data_hash[MCCTRL_PER_THREAD_DATA_HASH_SIZE];
 	rwlock_t per_thread_data_hash_lock[MCCTRL_PER_THREAD_DATA_HASH_SIZE];
+	cpumask_t cpu_set;
+	int ikc_target_cpu;
 };
 
 struct sysfsm_req {
@@ -226,11 +227,6 @@ static inline int sysfs_inited(struct sysfsm_data *sdp)
 	return !!(sdp->sysfs_buf);
 } /* sysfs_inited() */
 
-struct cpu_mapping {
-	int cpu_number;
-	int hw_id;
-};
-
 struct cache_topology {
 	struct ihk_cache_topology *saved;
 	cpumask_t shared_cpu_map;
@@ -239,8 +235,9 @@ struct cache_topology {
 };
 
 struct cpu_topology {
-	struct cpu_mapping *cpu_mapping;
+	//struct mcctrl_usrdata *udp;
 	struct ihk_cpu_topology *saved;
+	int mckernel_cpu_id;
 	cpumask_t core_siblings;
 	cpumask_t thread_siblings;
 
@@ -248,13 +245,24 @@ struct cpu_topology {
 	struct list_head cache_list;
 };
 
+#define NODE_DISTANCE_S_SIZE	1024
+
 struct node_topology {
 	struct ihk_node_topology *saved;
+	int mckernel_numa_id;
+	char mckernel_numa_distance_s[NODE_DISTANCE_S_SIZE];
 	cpumask_t cpumap;
 
 	struct list_head chain;
 };
 
+struct mcctrl_part_exec {
+	struct mutex lock;	
+	int nr_processes;
+	int nr_processes_left;
+	cpumask_t cpus_used;
+};
+
 #define CPU_LONGS (((NR_CPUS) + (BITS_PER_LONG) - 1) / (BITS_PER_LONG))
 
 #define MCCTRL_PER_PROC_DATA_HASH_SHIFT 7
@@ -266,9 +274,7 @@ struct mcctrl_usrdata {
 	struct ihk_ikc_listen_param listen_param2;
 	ihk_os_t	os;
 	int	num_channels;
-	struct mcctrl_channel	*channels;
-	unsigned long	*mcctrl_doorbell_va;
-	unsigned long	mcctrl_doorbell_pa;
+	struct mcctrl_channel *channels;
 	int	remaining_job;
 	int	base_cpu;
 	int	job_pos;
@@ -282,12 +288,12 @@ struct mcctrl_usrdata {
 	void **keys;
 	struct sysfsm_data sysfsm_data;
 	unsigned long cpu_online[CPU_LONGS];
-	int cpu_mapping_elems;
-	int padding;
-	struct cpu_mapping *cpu_mapping;
-	long cpu_mapping_pa;
+	struct ihk_cpu_info *cpu_info;
+	struct ihk_mem_info *mem_info;
+	nodemask_t numa_online;
 	struct list_head cpu_topology_list;
 	struct list_head node_topology_list;
+	struct mcctrl_part_exec part_exec;
 };
 
 struct mcctrl_signal {
@@ -322,7 +328,7 @@ inline struct mcctrl_per_thread_data *mcctrl_get_per_thread_data(
 void __return_syscall(ihk_os_t os, struct ikc_scd_packet *packet, 
 		long ret, int stid);
 
-#define PROCFS_NAME_MAX 1000
+#define PROCFS_NAME_MAX 768
 
 struct procfs_read {
 	unsigned long pbuf;	/* physical address of the host buffer (request) */

executer/kernel/mcctrl/procfs.c

@@ -481,8 +481,9 @@ procfs_exit(int osnum)
 
 	down(&procfs_file_list_lock);
 	e = find_base_entry(osnum);
-	if(e)
+	if (e) {
 		delete_procfs_entries(e);
+	}
 	up(&procfs_file_list_lock);
 }
 
@@ -496,7 +497,7 @@ static ssize_t
 mckernel_procfs_read(struct file *file, char __user *buf, size_t nbytes,
 	       loff_t *ppos)
 {
-	struct inode * inode = file->f_path.dentry->d_inode;
+	struct inode * inode = file->f_inode;
 	char *kern_buffer = NULL;
 	int order = 0;
 	volatile struct procfs_read *r = NULL;
@@ -599,7 +600,7 @@ static ssize_t
 mckernel_procfs_write(struct file *file, const char __user *buf, size_t nbytes,
 	       loff_t *ppos)
 {
-	struct inode * inode = file->f_path.dentry->d_inode;
+	struct inode * inode = file->f_inode;
 	char *kern_buffer = NULL;
 	int order = 0;
 	volatile struct procfs_read *r = NULL;

executer/kernel/mcctrl/syscall.c

@@ -746,6 +746,18 @@ static struct list_head pager_list = LIST_HEAD_INIT(pager_list);
 struct pager_create_result {
 	uintptr_t	handle;
 	int		maxprot;
+	uint32_t flags;
+	size_t size;
+};
+
+enum {
+	/* for memobj.flags */
+	MF_HAS_PAGER	= 0x0001,
+	MF_SHMDT_OK	= 0x0002,
+	MF_IS_REMOVABLE	= 0x0004,
+	MF_PREFETCH = 0x0008,
+	MF_ZEROFILL = 0x0010,
+	MF_END
 };
 
 static int pager_req_create(ihk_os_t os, int fd, uintptr_t result_pa)
@@ -760,6 +772,7 @@ static int pager_req_create(ihk_os_t os, int fd, uintptr_t result_pa)
 	struct pager *newpager = NULL;
 	uintptr_t phys;
 	struct kstat st;
+	int mf_flags = 0;
 
 	dprintk("pager_req_create(%d,%lx)\n", fd, (long)result_pa);
 
@@ -827,6 +840,32 @@ static int pager_req_create(ihk_os_t os, int fd, uintptr_t result_pa)
 			list_add(&newpager->list, &pager_list);
 			pager = newpager;
 			newpager = NULL;
+
+			/* Intel MPI library and shared memory "prefetch" */
+			{
+				char *pathbuf, *fullpath;
+
+				pathbuf = kmalloc(PATH_MAX, GFP_TEMPORARY);
+				if (pathbuf) {
+					fullpath = d_path(&file->f_path, pathbuf, PATH_MAX);
+					if (!IS_ERR(fullpath)) {
+						if (!strncmp("/dev/shm/Intel_MPI", fullpath, 18)) {
+							//mf_flags = (MF_PREFETCH | MF_ZEROFILL);
+							mf_flags = (MF_ZEROFILL);
+							dprintk("%s: filename: %s, zerofill\n",
+									__FUNCTION__, fullpath);
+						}
+						else if (strstr(fullpath, "libmpi") != NULL) {
+							mf_flags = MF_PREFETCH;
+							dprintk("%s: filename: %s, prefetch\n",
+									__FUNCTION__, fullpath);
+						}
+					}
+
+					kfree(pathbuf);
+				}
+			}
+
 			break;
 		}
 
@@ -856,6 +895,8 @@ found:
 	resp = ihk_device_map_virtual(dev, phys, sizeof(*resp), NULL, 0);
 	resp->handle = (uintptr_t)pager;
 	resp->maxprot = maxprot;
+	resp->flags = mf_flags;
+	resp->size = st.size;
 	ihk_device_unmap_virtual(dev, resp, sizeof(*resp));
 	ihk_device_unmap_memory(dev, phys, sizeof(*resp));
 

executer/kernel/mcctrl/sysfs.c

@@ -18,7 +18,7 @@
 #include "mcctrl.h"
 #include "sysfs_msg.h"
 
-#define dprintk(...) do { if (0) printk(KERN_DEBUG __VA_ARGS__); } while (0)
+#define dprintk(...) do { if (0) printk(__VA_ARGS__); } while (0)
 #define wprintk(...) do { if (1) printk(KERN_WARNING __VA_ARGS__); } while (0)
 #define eprintk(...) do { if (1) printk(KERN_ERR __VA_ARGS__); } while (0)
 
@@ -1232,9 +1232,16 @@ sysfsm_cleanup(ihk_os_t os)
 	int error;
 	ihk_device_t dev = ihk_os_to_dev(os);
 	struct mcctrl_usrdata *udp = ihk_host_os_get_usrdata(os);
-	struct sysfsm_data *sdp = &udp->sysfsm_data;
+	struct sysfsm_data *sdp;
 	struct sysfsm_node *np;
 
+	if (!udp) {
+		printk("%s: WARNING: no mcctrl_usrdata found\n", __FUNCTION__);
+		return;
+	}
+
+	sdp = &udp->sysfsm_data;
+
 	dprintk("mcctrl:sysfsm_cleanup(%p)\n", os);
 
 	if (sdp->sysfs_buf) {
@@ -2095,9 +2102,16 @@ struct sysfsm_ops snooping_local_ops_s = {
 /**** local list ****/
 static ssize_t snooping_local_show_pbl(struct sysfsm_ops *ops, void *instance, void *buf, size_t bufsize)
 {
+	size_t ret;
 	const struct sysfsm_bitmap_param *p = instance;
 
-	return bitmap_scnlistprintf(buf, bufsize, p->ptr, p->nbits);
+	ret = bitmap_scnlistprintf(buf, bufsize, p->ptr, p->nbits);
+	if (ret < bufsize - 1) {
+		sprintf(buf + ret, "\n");
+		return ret + 1;
+	}
+
+	return 0;
 } /* snooping_local_show_pbl() */
 
 struct sysfsm_ops snooping_local_ops_pbl = {
@@ -2108,9 +2122,16 @@ struct sysfsm_ops snooping_local_ops_pbl = {
 /**** local map ****/
 static ssize_t snooping_local_show_pb(struct sysfsm_ops *ops, void *instance, void *buf, size_t bufsize)
 {
+	size_t ret;
 	const struct sysfsm_bitmap_param *p = instance;
 
-	return bitmap_scnprintf(buf, bufsize, p->ptr, p->nbits);
+	ret = bitmap_scnprintf(buf, bufsize, p->ptr, p->nbits);
+	if (ret < bufsize - 1) {
+		sprintf(buf + ret, "\n");
+		return ret + 1;
+	}
+
+	return 0;
 } /* snooping_local_show_pb() */
 
 struct sysfsm_ops snooping_local_ops_pb = {

executer/kernel/mcctrl/sysfs_files.c

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

executer/kernel/mcoverlayfs/Makefile.in

@@ -14,6 +14,9 @@ ifeq ($(BUILD_MODULE_TMP),org)
 ifeq ($(BUILD_MODULE),none)
 BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -ge 262144 -a ${LINUX_VERSION_CODE} -lt 262400 ]; then echo "linux-4.0.9"; else echo "none"; fi)
 endif
+ifeq ($(BUILD_MODULE),none)
+BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -ge 243680 -a ${LINUX_VERSION_CODE} -lt 263936 ]; then echo "linux-4.6.7"; else echo "none"; fi)
+endif
 endif
 ifeq ($(BUILD_MODULE_TMP),rhel)
 ifeq ($(BUILD_MODULE),none)
@@ -32,6 +35,7 @@ endif
 clean:
 	@(cd linux-3.10.0-327.36.1.el7; make clean)
 	@(cd linux-4.0.9; make clean)
+	@(cd linux-4.6.7; make clean)
 
 install:
 ifneq ($(BUILD_MODULE),none)

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

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

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

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

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

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

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

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

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

@@ -0,0 +1,222 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+
+//#define DEBUG
+#ifdef DEBUG
+#define OVL_DEBUG(format, ...)	pr_err("[DEBUG] %s(): " format, __FUNCTION__, ##__VA_ARGS__)
+#else
+#define OVL_DEBUG(format, ...)	{}
+#endif
+
+struct ovl_entry;
+
+enum ovl_path_type {
+	__OVL_PATH_PURE		= (1 << 0),
+	__OVL_PATH_UPPER	= (1 << 1),
+	__OVL_PATH_MERGE	= (1 << 2),
+};
+
+#define OVL_TYPE_UPPER(type)	((type) & __OVL_PATH_UPPER)
+#define OVL_TYPE_MERGE(type)	((type) & __OVL_PATH_MERGE)
+#define OVL_TYPE_PURE_UPPER(type) ((type) & __OVL_PATH_PURE)
+#define OVL_TYPE_MERGE_OR_LOWER(type) \
+	(OVL_TYPE_MERGE(type) || !OVL_TYPE_UPPER(type))
+
+#define OVL_XATTR_PRE_NAME "trusted.overlay."
+#define OVL_XATTR_PRE_LEN  16
+#define OVL_XATTR_OPAQUE   OVL_XATTR_PRE_NAME"opaque"
+
+enum ovl_opt_bit {
+	__OVL_OPT_DEFAULT	= 0,
+	__OVL_OPT_NOCOPYUPW	= (1 << 0),
+	__OVL_OPT_NOFSCHECK	= (1 << 1),
+};
+
+#define OVL_OPT_NOCOPYUPW(opt)	((opt) & __OVL_OPT_NOCOPYUPW)
+#define OVL_OPT_NOFSCHECK(opt)	((opt) & __OVL_OPT_NOFSCHECK)
+
+static inline int ovl_do_rmdir(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_rmdir(dir, dentry);
+	pr_debug("rmdir(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+static inline int ovl_do_unlink(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_unlink(dir, dentry, NULL);
+	pr_debug("unlink(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+static inline int ovl_do_link(struct dentry *old_dentry, struct inode *dir,
+			      struct dentry *new_dentry, bool debug)
+{
+	int err = vfs_link(old_dentry, dir, new_dentry, NULL);
+	if (debug) {
+		pr_debug("link(%pd2, %pd2) = %i\n",
+			 old_dentry, new_dentry, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_create(struct inode *dir, struct dentry *dentry,
+			     umode_t mode, bool debug)
+{
+	int err = vfs_create(dir, dentry, mode, true);
+	if (debug)
+		pr_debug("create(%pd2, 0%o) = %i\n", dentry, mode, err);
+	return err;
+}
+
+static inline int ovl_do_mkdir(struct inode *dir, struct dentry *dentry,
+			       umode_t mode, bool debug)
+{
+	int err = vfs_mkdir(dir, dentry, mode);
+	if (debug)
+		pr_debug("mkdir(%pd2, 0%o) = %i\n", dentry, mode, err);
+	return err;
+}
+
+static inline int ovl_do_mknod(struct inode *dir, struct dentry *dentry,
+			       umode_t mode, dev_t dev, bool debug)
+{
+	int err = vfs_mknod(dir, dentry, mode, dev);
+	if (debug) {
+		pr_debug("mknod(%pd2, 0%o, 0%o) = %i\n",
+			 dentry, mode, dev, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_symlink(struct inode *dir, struct dentry *dentry,
+				 const char *oldname, bool debug)
+{
+	int err = vfs_symlink(dir, dentry, oldname);
+	if (debug)
+		pr_debug("symlink(\"%s\", %pd2) = %i\n", oldname, dentry, err);
+	return err;
+}
+
+static inline int ovl_do_setxattr(struct dentry *dentry, const char *name,
+				  const void *value, size_t size, int flags)
+{
+	int err = vfs_setxattr(dentry, name, value, size, flags);
+	pr_debug("setxattr(%pd2, \"%s\", \"%*s\", 0x%x) = %i\n",
+		 dentry, name, (int) size, (char *) value, flags, err);
+	return err;
+}
+
+static inline int ovl_do_removexattr(struct dentry *dentry, const char *name)
+{
+	int err = vfs_removexattr(dentry, name);
+	pr_debug("removexattr(%pd2, \"%s\") = %i\n", dentry, name, err);
+	return err;
+}
+
+static inline int ovl_do_rename(struct inode *olddir, struct dentry *olddentry,
+				struct inode *newdir, struct dentry *newdentry,
+				unsigned int flags)
+{
+	int err;
+
+	pr_debug("rename2(%pd2, %pd2, 0x%x)\n",
+		 olddentry, newdentry, flags);
+
+	err = vfs_rename(olddir, olddentry, newdir, newdentry, NULL, flags);
+
+	if (err) {
+		pr_debug("...rename2(%pd2, %pd2, ...) = %i\n",
+			 olddentry, newdentry, err);
+	}
+	return err;
+}
+
+static inline int ovl_do_whiteout(struct inode *dir, struct dentry *dentry)
+{
+	int err = vfs_whiteout(dir, dentry);
+	pr_debug("whiteout(%pd2) = %i\n", dentry, err);
+	return err;
+}
+
+unsigned ovl_get_config_opt(struct dentry *dentry);
+void ovl_reset_ovl_entry(struct ovl_entry **oe, struct dentry *dentry);
+enum ovl_path_type ovl_path_type(struct dentry *dentry);
+u64 ovl_dentry_version_get(struct dentry *dentry);
+void ovl_dentry_version_inc(struct dentry *dentry);
+void ovl_path_upper(struct dentry *dentry, struct path *path);
+void ovl_path_lower(struct dentry *dentry, struct path *path);
+enum ovl_path_type ovl_path_real(struct dentry *dentry, struct path *path);
+int ovl_path_next(int idx, struct dentry *dentry, struct path *path);
+struct dentry *ovl_dentry_upper(struct dentry *dentry);
+struct dentry *ovl_dentry_lower(struct dentry *dentry);
+struct dentry *ovl_dentry_real(struct dentry *dentry);
+struct dentry *ovl_entry_real(struct ovl_entry *oe, bool *is_upper);
+struct vfsmount *ovl_entry_mnt_real(struct ovl_entry *oe, struct inode *inode,
+				    bool is_upper);
+struct ovl_dir_cache *ovl_dir_cache(struct dentry *dentry);
+bool ovl_is_default_permissions(struct inode *inode);
+void ovl_set_dir_cache(struct dentry *dentry, struct ovl_dir_cache *cache);
+struct dentry *ovl_workdir(struct dentry *dentry);
+int ovl_want_write(struct dentry *dentry);
+void ovl_drop_write(struct dentry *dentry);
+bool ovl_dentry_is_opaque(struct dentry *dentry);
+void ovl_dentry_set_opaque(struct dentry *dentry, bool opaque);
+bool ovl_is_whiteout(struct dentry *dentry);
+void ovl_dentry_update(struct dentry *dentry, struct dentry *upperdentry);
+struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
+			  unsigned int flags);
+struct file *ovl_path_open(struct path *path, int flags);
+
+struct dentry *ovl_upper_create(struct dentry *upperdir, struct dentry *dentry,
+				struct kstat *stat, const char *link);
+
+/* readdir.c */
+extern const struct file_operations ovl_dir_operations;
+int ovl_check_empty_dir(struct dentry *dentry, struct list_head *list);
+void ovl_cleanup_whiteouts(struct dentry *upper, struct list_head *list);
+void ovl_cache_free(struct list_head *list);
+int ovl_check_d_type_supported(struct path *realpath);
+
+/* inode.c */
+int ovl_setattr(struct dentry *dentry, struct iattr *attr);
+int ovl_permission(struct inode *inode, int mask);
+int ovl_setxattr(struct dentry *dentry, const char *name,
+		 const void *value, size_t size, int flags);
+ssize_t ovl_getxattr(struct dentry *dentry, const char *name,
+		     void *value, size_t size);
+ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size);
+int ovl_removexattr(struct dentry *dentry, const char *name);
+struct inode *ovl_d_select_inode(struct dentry *dentry, unsigned file_flags);
+
+struct inode *ovl_new_inode(struct super_block *sb, umode_t mode,
+			    struct ovl_entry *oe);
+static inline void ovl_copyattr(struct inode *from, struct inode *to)
+{
+	to->i_uid = from->i_uid;
+	to->i_gid = from->i_gid;
+	to->i_mode = from->i_mode;
+}
+
+/* dir.c */
+extern const struct inode_operations ovl_dir_inode_operations;
+struct dentry *ovl_lookup_temp(struct dentry *workdir, struct dentry *dentry);
+int ovl_create_real(struct inode *dir, struct dentry *newdentry,
+		    struct kstat *stat, const char *link,
+		    struct dentry *hardlink, bool debug);
+void ovl_cleanup(struct inode *dir, struct dentry *dentry);
+
+/* copy_up.c */
+int ovl_copy_up(struct dentry *dentry);
+int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry,
+		    struct path *lowerpath, struct kstat *stat);
+int ovl_copy_xattr(struct dentry *old, struct dentry *new, unsigned opt);
+int ovl_set_attr(struct dentry *upper, struct kstat *stat);

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

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

executer/user/eclair.c

@@ -17,6 +17,20 @@
 #include <sys/socket.h>
 #include <arpa/inet.h>
 
+/* From ihk/linux/include/ihk/ihk_host_user.h */
+#define PHYS_CHUNKS_DESC_SIZE 8192
+
+struct dump_mem_chunk {
+	unsigned long addr;
+	unsigned long size;
+};
+
+typedef struct dump_mem_chunks_s {
+	int nr_chunks;
+	struct dump_mem_chunk chunks[];
+} dump_mem_chunks_t;
+/* ---------- */
+
 #define CPU_TID_BASE 1000000
 
 struct options {
@@ -53,6 +67,7 @@ static volatile int f_done = 0;
 static bfd *symbfd = NULL;
 static bfd *dumpbfd = NULL;
 static asection *dumpscn = NULL;
+static dump_mem_chunks_t *mem_chunks;
 static int num_processors = -1;
 static asymbol **symtab = NULL;
 static ssize_t nsyms;
@@ -91,25 +106,35 @@ static uintptr_t virt_to_phys(uintptr_t va) {
 static int read_physmem(uintptr_t pa, void *buf, size_t size) {
 	off_t off;
 	bfd_boolean ok;
+	int i;
 
-	if (pa < dumpscn->vma) {
-		printf("read_physmem(%lx,%p,%lx):too small pa. vma %lx\n", pa, buf, size, dumpscn->vma);
-		return 1;
-	}
-	off = pa - dumpscn->vma;
-	if (off >= dumpscn->size) {
-		printf("read_physmem(%lx,%p,%lx):too large pa. vma %lx size %lx\n", pa, buf, size, dumpscn->vma, dumpscn->size);
-		return 1;
-	}
-	if ((dumpscn->size - off) < size) {
-		printf("read_physmem(%lx,%p,%lx):too large size. vma %lx size %lx\n", pa, buf, size, dumpscn->vma, dumpscn->size);
+	off = 0;
+	/* Check if pa is valid in any chunks and figure
+	 * out the global offset in dump section */
+	for (i = 0; i < mem_chunks->nr_chunks; ++i) {
+
+		if (mem_chunks->chunks[i].addr <= pa &&
+				((pa + size) <= (mem_chunks->chunks[i].addr +
+					mem_chunks->chunks[i].size))) {
+
+			off += (pa - mem_chunks->chunks[i].addr);
+			break;
+		}
+
+		off += mem_chunks->chunks[i].size;
+	}
+
+	if (i == mem_chunks->nr_chunks) {
+		printf("read_physmem: invalid addr 0x%lx\n", pa);
 		return 1;
 	}
+
 	ok = bfd_get_section_contents(dumpbfd, dumpscn, buf, off, size);
 	if (!ok) {
 		bfd_perror("read_physmem:bfd_get_section_contents");
 		return 1;
 	}
+
 	return 0;
 } /* read_physmem() */
 
@@ -508,6 +533,25 @@ static int setup_dump(char *fname) {
 		return 1;
 	}
 
+	mem_chunks = malloc(PHYS_CHUNKS_DESC_SIZE);
+	if (!mem_chunks) {
+		perror("allocating mem chunks descriptor: ");
+		return 1;
+	}
+
+	dumpscn = bfd_get_section_by_name(dumpbfd, "physchunks");
+	if (!dumpscn) {
+		bfd_perror("bfd_get_section_by_name");
+		return 1;
+	}
+
+	ok = bfd_get_section_contents(dumpbfd, dumpscn, mem_chunks,
+			0, PHYS_CHUNKS_DESC_SIZE);
+	if (!ok) {
+		bfd_perror("read_physmem:bfd_get_section_contents");
+		return 1;
+	}
+
 	dumpscn = bfd_get_section_by_name(dumpbfd, "physmem");
 	if (!dumpscn) {
 		bfd_perror("bfd_get_section_by_name");

executer/user/mcexec.c

@@ -41,6 +41,7 @@
 #include <sys/mman.h>
 #include <asm/unistd.h>
 #include <sched.h>
+#include <dirent.h>
 
 #include <termios.h>
 #include <sys/ioctl.h>
@@ -107,6 +108,9 @@ char **__glob_argv = 0;
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,1,0)
 #define ENABLE_MCOVERLAYFS 1
 #endif // LINUX_VERSION_CODE == 4.0
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,6,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,7,0)
+#define ENABLE_MCOVERLAYFS 1
+#endif // LINUX_VERSION_CODE == 4.6
 #else
 #if RHEL_RELEASE_CODE == RHEL_RELEASE_VERSION(7,2)
 #define ENABLE_MCOVERLAYFS 1
@@ -149,6 +153,10 @@ static const char rlimit_stack_envname[] = "MCKERNEL_RLIMIT_STACK";
 static int ischild;
 static int enable_vdso = 1;
 
+/* Partitioned execution (e.g., for MPI) */
+static int nr_processes = 0;
+static int nr_threads = -1;
+
 struct fork_sync {
 	pid_t pid;
 	int status;
@@ -498,7 +506,7 @@ retry:
 
 	/* Check whether the resolved path is a symlink */
 	if (lstat(path, &sb) == -1) {
-		fprintf(stderr, "lookup_exec_path(): error stat\n");
+		__dprintf(stderr, "lookup_exec_path(): error stat\n");
 		return errno;
 	}
 
@@ -1098,7 +1106,7 @@ static int reduce_stack(struct rlimit *orig_rlim, char *argv[])
 
 void print_usage(char **argv)
 {
-	fprintf(stderr, "Usage: %s [-c target_core] [<mcos-id>] (program) [args...]\n", argv[0]);
+	fprintf(stderr, "Usage: %s [-c target_core] [-n nr_partitions] [<mcos-id>] (program) [args...]\n", argv[0]);
 }
 
 void init_sigaction(void)
@@ -1148,75 +1156,41 @@ void init_worker_threads(int fd)
 
 #ifdef ENABLE_MCOVERLAYFS
 #define READ_BUFSIZE 1024
-static int isunshare(void)
+static int find_mount_prefix(char *prefix)
 {
-	int err = 0;
-	int ret;
-	int fd;
+	FILE *fp;
+	char *line = NULL;
+	size_t len = 0;
+	ssize_t read;
 	char proc_path[PATH_MAX];
-	ssize_t len_read;
-	char buf_read[READ_BUFSIZE + 1];
-	char *buf_read_off;
-	char *buf_find;
-	char buf_cmp[READ_BUFSIZE + 1];
-	char *buf_cmp_off;
-	ssize_t len_copy;
+	int ret = 0;
 
 	snprintf(proc_path, sizeof(proc_path), "/proc/%d/mounts", getpid());
-	fd = open(proc_path, O_RDONLY);
-	if (fd < 0) {
-		fprintf(stderr, "Error: Failed to open %s.\n", proc_path);
+
+	fp = fopen(proc_path, "r");
+	if (fp == NULL) {
 		return -1;
 	}
 
-	buf_cmp_off = buf_cmp;
-	while (1) {
-		len_read = read(fd, buf_read, READ_BUFSIZE);
-		if (len_read == -1) {
-			fprintf(stderr, "Error: Failed to read.\n");
-			err = -1;
-			break;
-		}
+	while ((read = getline(&line, &len, fp)) != -1) {
+		if (strlen(line) < strlen(prefix))
+			continue;
 
-		buf_read_off = buf_read;
-		while (1) {
-			if ((len_read - (buf_read_off - buf_read)) <= 0) {
-				break;
-			}
-			buf_find = memchr(buf_read_off, '\n', 
-				len_read - (buf_read_off - buf_read));
-			if (buf_find) {
-				len_copy = buf_find - buf_read_off;	
-			} else {
-				len_copy = len_read - (buf_read_off - buf_read);
-			}
-			memcpy(buf_cmp_off, buf_read_off, len_copy);
-			*(buf_cmp_off + len_copy) = '\0';
-
-			if (buf_find) {
-				buf_read_off = buf_read_off + len_copy + 1;
-				buf_cmp_off = buf_cmp;
-				ret = strncmp(buf_cmp, "mcoverlay /proc ", 16);
-				if (!ret) {
-					err = 1;
-					break;
-				}
-			} else {
-				buf_read_off = buf_read_off + len_copy;
-				buf_cmp_off = buf_cmp_off + len_copy;
-				break;
-			}
-		}
-
-		if (err == 1 || len_read == 0) {
+		if (!strncmp(line, prefix, strlen(prefix))) {
+			ret = 1;
 			break;
 		}
 	}
 
-	close(fd);
+	if (line)
+		free(line);
 
-	__dprintf("err=%d\n", err);
-	return err;
+	return ret;
+}
+
+static int isunshare(void)
+{
+	return find_mount_prefix("mcoverlay /proc ");
 }
 #endif // ENABLE_MCOVERLAYFS
 
@@ -1359,12 +1333,20 @@ int main(int argc, char **argv)
 	}
            
 	/* Parse options ("+" denotes stop at the first non-option) */
-	while ((opt = getopt_long(argc, argv, "+c:", mcexec_options, NULL)) != -1) {
+	while ((opt = getopt_long(argc, argv, "+c:n:t:", mcexec_options, NULL)) != -1) {
 		switch (opt) {
 			case 'c':
 				target_core = atoi(optarg);
 				break;
-			
+
+			case 'n':
+				nr_processes = atoi(optarg);
+				break;
+
+			case 't':
+				nr_threads = atoi(optarg);
+				break;
+
 			case 0:	/* long opt */
 				break;
 
@@ -1415,6 +1397,7 @@ int main(int argc, char **argv)
 	if (error == 0) {
 		struct sys_unshare_desc unshare_desc;
 		struct sys_mount_desc mount_desc;
+		struct sys_umount_desc umount_desc;
 
 		memset(&unshare_desc, '\0', sizeof unshare_desc);
 		memset(&mount_desc, '\0', sizeof mount_desc);
@@ -1426,6 +1409,53 @@ int main(int argc, char **argv)
 			return 1;
 		}
 
+		/*
+		 * Umount cgroup filesystems that may expose invalid NUMA
+		 * information
+		 */
+		if (find_mount_prefix("cgroup /sys/fs/cgroup/cpu,cpuacct")) {
+			umount_desc.dir_name = "/sys/fs/cgroup/cpu,cpuacct";
+
+			if (ioctl(fd, MCEXEC_UP_SYS_UMOUNT,
+						(unsigned long)&umount_desc) != 0) {
+				fprintf(stderr,
+						"WARNING: Failed to umount cgroup/cpu,cpuacct. (%s)\n",
+						strerror(errno));
+			}
+		}
+		else if (find_mount_prefix("cgroup /sys/fs/cgroup/cpu")) {
+			umount_desc.dir_name = "/sys/fs/cgroup/cpu";
+
+			if (ioctl(fd, MCEXEC_UP_SYS_UMOUNT,
+						(unsigned long)&umount_desc) != 0) {
+				fprintf(stderr,
+						"WARNING: Failed to umount cgroup/cpu. (%s)\n",
+						strerror(errno));
+			}
+		}
+
+		if (find_mount_prefix("cgroup /sys/fs/cgroup/cpuset")) {
+			umount_desc.dir_name = "/sys/fs/cgroup/cpuset";
+
+			if (ioctl(fd, MCEXEC_UP_SYS_UMOUNT,
+						(unsigned long)&umount_desc) != 0) {
+				fprintf(stderr,
+						"WARNING: Failed to umount cgroup/cpuset. (%s)\n",
+						strerror(errno));
+			}
+		}
+
+		if (find_mount_prefix("cgroup /sys/fs/cgroup/memory")) {
+			umount_desc.dir_name = "/sys/fs/cgroup/memory/";
+
+			if (ioctl(fd, MCEXEC_UP_SYS_UMOUNT,
+						(unsigned long)&umount_desc) != 0) {
+				fprintf(stderr,
+						"WARNING: Failed to umount cgroup/memory. (%s)\n",
+						strerror(errno));
+			}
+		}
+
 		sprintf(mcos_procdir, "/tmp/mcos/mcos%d_proc", mcosid);
 		mount_desc.dev_name = mcos_procdir;
 		mount_desc.dir_name = "/proc";
@@ -1532,9 +1562,15 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
-	n_threads = ncpu;
-	if (ncpu > 16) {
-		n_threads = 16;
+	if (nr_threads > 0) {
+		n_threads = nr_threads;
+	}
+	else if (getenv("OMP_NUM_THREADS")) {
+		/* Leave some headroom for helper threads.. */
+		n_threads = atoi(getenv("OMP_NUM_THREADS")) + 4;
+	}
+	else {
+		n_threads = ncpu;
 	}
 
 	/* 
@@ -1546,6 +1582,10 @@ int main(int argc, char **argv)
 	 * TODO: implement dynaic thread pool resizing.
 	 */
 	thread_data = (struct thread_data_s *)malloc(sizeof(struct thread_data_s) * (ncpu + 1));
+	if (!thread_data) {
+		fprintf(stderr, "error: allocating thread pool data\n");
+		return 1;
+	}
 	memset(thread_data, '\0', sizeof(struct thread_data_s) * (ncpu + 1));
 
 #if 0	
@@ -1580,6 +1620,24 @@ int main(int argc, char **argv)
 		exit(1);
 	}
 
+	/* Partitioned execution, obtain CPU set */
+	if (nr_processes > 0) {
+		struct get_cpu_set_arg cpu_set_arg;
+
+		cpu_set_arg.cpu_set = (void *)&desc->cpu_set;
+		cpu_set_arg.cpu_set_size = sizeof(desc->cpu_set);
+		cpu_set_arg.nr_processes = nr_processes;
+		cpu_set_arg.target_core = &target_core;
+		
+		if (ioctl(fd, MCEXEC_UP_GET_CPUSET, (void *)&cpu_set_arg) != 0) {
+			perror("getting CPU set for partitioned execution");
+			close(fd);
+			return 1;
+		}
+		
+		desc->cpu = target_core;
+	}
+	
 	if (ioctl(fd, MCEXEC_UP_PREPARE_IMAGE, (unsigned long)desc) != 0) {
 		perror("prepare");
 		close(fd);
@@ -1686,6 +1744,97 @@ do_generic_syscall(
 		ret = -errno;
 	}
 
+	/* Overlayfs /sys/X directory lseek() problem work around */
+	if (w->sr.number == __NR_lseek && ret == -EINVAL) {
+		char proc_path[512];
+		char path[512];
+		struct stat sb;
+
+		sprintf(proc_path, "/proc/self/fd/%d", (int)w->sr.args[0]);
+
+		/* Get filename */
+		if (readlink(proc_path, path, sizeof(path)) < 0) {
+			fprintf(stderr, "%s: error: readlink() failed for %s\n",
+				__FUNCTION__, proc_path);
+			goto out;
+		}
+
+		/* Not in /sys? */
+		if (strncmp(path, "/sys/", 5))
+			goto out;
+
+		/* Stat */
+		if (stat(path, &sb) < 0) {
+			fprintf(stderr, "%s: error stat() failed for %s\n",
+				__FUNCTION__, path);
+			goto out;
+		}
+
+		/* Not dir? */
+		if ((sb.st_mode & S_IFMT) != S_IFDIR)
+			goto out;
+
+		ret = 0;
+	}
+	/* Fake that nodeX in /sys/devices/system/node do not exist,
+	 * where X >= number of LWK NUMA nodes */
+	else if (w->sr.number == __NR_getdents && ret > 0) {
+		struct linux_dirent {
+			long           d_ino;
+			off_t          d_off;
+			unsigned short d_reclen;
+			char           d_name[];
+		};
+		struct linux_dirent *d;
+		char *buf = (char *)w->sr.args[1];
+		int bpos = 0;
+		int nodes,len;
+		char proc_path[PATH_MAX];
+		char path[PATH_MAX];
+
+		sprintf(proc_path, "/proc/self/fd/%d", (int)w->sr.args[0]);
+
+		/* Get filename */
+		if ((len = readlink(proc_path, path, sizeof(path))) < 0) {
+			fprintf(stderr, "%s: error: readlink() failed for %s\n",
+				__FUNCTION__, proc_path);
+			goto out;
+		}
+		path[len] = 0;
+
+		/* Not /sys/devices/system/node ? */
+		if (strcmp(path, "/sys/devices/system/node"))
+			goto out;
+
+		nodes = ioctl(fd, MCEXEC_UP_GET_NODES, 0);
+		if (nodes == -1) {
+			goto out;
+		}
+
+		d = (struct linux_dirent *) (buf + bpos);
+		for (bpos = 0; bpos < ret; ) {
+			int nodeid, tmp_reclen;
+			d = (struct linux_dirent *) (buf + bpos);
+
+			if (sscanf(d->d_name, "node%d", &nodeid) != 1) {
+				bpos += d->d_reclen;
+				continue;
+			}
+
+			if (nodeid >= nodes) {
+				tmp_reclen = d->d_reclen;
+				memmove(buf + bpos,
+						buf + bpos + tmp_reclen,
+						ret - bpos - tmp_reclen);
+				ret -= tmp_reclen;
+				continue;
+			}
+
+			bpos += d->d_reclen;
+		}
+	}
+
+out:
 	__dprintf("do_generic_syscall(%ld):%ld (%#lx)\n", w->sr.number, ret, ret);
 	return ret;
 }
@@ -1695,7 +1844,7 @@ kill_thread(unsigned long tid)
 {
 	int	i;
 
-	for (i = 0; i < n_threads; ++i) {
+	for (i = 0; i <= n_threads; ++i) {
 		if(thread_data[i].remote_tid == tid){
 			pthread_kill(thread_data[i].thread_id, LOCALSIG);
 			break;
@@ -1800,9 +1949,18 @@ int close_cloexec_fds(int mcos_fd)
 	return 0;
 }
 
+void chgdevpath(char *in, char *buf)
+{
+	if(!strcmp(in, "/dev/xpmem")){
+		sprintf(in, "/dev/null");
+	}
+}
+
 char *
 chgpath(char *in, char *buf)
 {
+	chgdevpath(in, buf);
+
 #ifdef ENABLE_MCOVERLAYFS
 	return in;
 #endif // ENABLE_MCOVERLAYFS
@@ -1954,21 +2112,6 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 #endif
 
 		case __NR_gettid:{
-			int mode = w.sr.args[0];
-			int remote_pid = w.sr.args[1];
-			int newcpuid = w.sr.args[2];
-			int oldcpuid = w.sr.args[3];
-			int wtid = thread_data[newcpuid].remote_tid;
-
-			if(mode == 0){
-				thread_data[ncpu].remote_tid = wtid;
-				thread_data[newcpuid].remote_tid = remote_pid;
-			}
-			else if(mode == 2){
-				thread_data[newcpuid].remote_tid = thread_data[oldcpuid].remote_tid;
-				thread_data[oldcpuid].remote_tid = wtid;
-			}
-
 			/*
 			 * Number of TIDs and the remote physical address where TIDs are
 			 * expected are passed in arg 4 and 5, respectively.
@@ -2002,7 +2145,7 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 				free(tids);
 			}
 gettid_out:
-			do_syscall_return(fd, cpu, thread_data[newcpuid].remote_tid, 0, 0, 0, 0);
+			do_syscall_return(fd, cpu, 0, 0, 0, 0, 0);
 			break;
 		}
 
@@ -2328,6 +2471,23 @@ return_execve1:
 
 					ret = 0;
 return_execve2:					
+#ifdef ENABLE_MCOVERLAYFS
+				{
+					struct sys_mount_desc mount_desc;
+
+					mount_desc.dev_name = NULL;
+					mount_desc.dir_name = "/proc";
+					mount_desc.type = NULL;
+					mount_desc.flags = MS_REMOUNT;
+					mount_desc.data = NULL;
+					if (ioctl(fd, MCEXEC_UP_SYS_MOUNT,
+								(unsigned long)&mount_desc) != 0) {
+						fprintf(stderr,
+								"WARNING: failed to remount /proc (%s)\n",
+								strerror(errno));
+					}
+				}
+#endif
 					do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 					break;
 

kernel/Makefile.build.in

@@ -3,15 +3,15 @@ SRC=$(VPATH)
 IHKDIR=$(IHKBASE)/$(TARGETDIR)
 OBJS = init.o mem.o debug.o mikc.o listeners.o ap.o syscall.o cls.o host.o
 OBJS += process.o copy.o waitq.o futex.o timer.o plist.o fileobj.o shmobj.o
-OBJS += zeroobj.o procfs.o devobj.o sysfs.o
+OBJS += zeroobj.o procfs.o devobj.o sysfs.o xpmem.o
 DEPSRCS=$(wildcard $(SRC)/*.c)
 
-CFLAGS += -I$(SRC)/include -D__KERNEL__ -g
+CFLAGS += -I$(SRC)/include -D__KERNEL__ -g -fno-omit-frame-pointer -fno-inline -fno-inline-small-functions
 LDFLAGS += -e arch_start
 IHKOBJ = ihk/ihk.o
 
 include $(SRC)/config/config.$(TARGET)
-include $(IHKBASE)/Makefile.common
+include @abs_builddir@/../../ihk/cokernel/Makefile.common
 
 # CFLAGS += -I$(SRC)/../arch/$(IHKARCH)/kernel/include -I$(SRC)/../lib/include
 

kernel/Makefile.in

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

kernel/ap.c

@@ -65,7 +65,7 @@ void ap_init(void)
 {
 	struct ihk_mc_cpu_info *cpu_info;
 	int i;
-	int bsp_hw_id;
+	int bsp_hw_id, bsp_cpu_id;
 
 	ihk_mc_init_ap();
 	init_delay();
@@ -78,13 +78,23 @@ void ap_init(void)
 		return;
 	}
 
-	kprintf("BSP HW ID = %d\n", bsp_hw_id);
+	bsp_cpu_id = 0;
+	for (i = 0; i < cpu_info->ncpus; ++i) {
+		if (cpu_info->hw_ids[i] == bsp_hw_id) {
+			bsp_cpu_id = i;
+			break;
+		}
+	}
+
+	kprintf("BSP: %d (HW ID: %d @ NUMA %d)\n", bsp_cpu_id,
+			bsp_hw_id, cpu_info->nodes[0]);
 
 	for (i = 0; i < cpu_info->ncpus; i++) {
 		if (cpu_info->hw_ids[i] == bsp_hw_id) {
 			continue;
 		}
-		kprintf("AP Booting: %d (HW ID: %d)\n", i, cpu_info->hw_ids[i]);
+		kprintf("AP Booting: %d (HW ID: %d @ NUMA %d)\n", i,
+			cpu_info->hw_ids[i], cpu_info->nodes[i]);
 		ihk_mc_boot_cpu(cpu_info->hw_ids[i], (unsigned long)ap_wait);
 
 		num_processors++;
@@ -199,7 +209,7 @@ cpu_sysfs_setup(void)
 	/* setup table */
 	info = kmalloc(sizeof(*info) * num_processors, IHK_MC_AP_CRITICAL);
 	for (cpu = 0; cpu < num_processors; ++cpu) {
-		info[cpu].online = 10+cpu;
+		info[cpu].online = 1;
 	}
 	fake_cpu_infos = info;
 

kernel/cls.c

@@ -23,7 +23,7 @@
 extern int num_processors;
 
 struct cpu_local_var *clv;
-static int cpu_local_var_initialized = 0;
+int cpu_local_var_initialized = 0;
 
 void cpu_local_var_init(void)
 {
@@ -32,7 +32,7 @@ void cpu_local_var_init(void)
 	z = sizeof(struct cpu_local_var) * num_processors;
 	z = (z + PAGE_SIZE - 1) >> PAGE_SHIFT;
 
-	clv = allocate_pages(z, IHK_MC_AP_CRITICAL);
+	clv = ihk_mc_alloc_pages(z, IHK_MC_AP_CRITICAL);
 	memset(clv, 0, z * PAGE_SIZE);
 	cpu_local_var_initialized = 1;
 }

kernel/debug.c

@@ -37,6 +37,8 @@ static void kprintf_wait(int len, unsigned long *flags_head, int *slide) {
 		if (head < tail) head += buf_len;
 		if (tail + len > buf_len) adj = buf_len - tail;
 		if (head > tail && head <= tail + len + adj) {
+			/* When proceeding tail (producer pointer) by len would
+			   cross head (consumer pointer) in ring-buffer */
 			if (mode != 1) {
 				*slide = 1;
 				break;
@@ -70,6 +72,9 @@ void kputs(char *buf)
 	
 	memcpy(kmsg_buf.str + kmsg_buf.tail, buf, len);
 	kmsg_buf.tail += len;
+	/* When proceeding tail (producer pointer) by len would
+	   cross head (consumer pointer) in ring-buffer, give up
+	   [head, tail] because the range is overwritten */
 	if (slide == 1) {
 		kmsg_buf.head = kmsg_buf.tail + 1;
 		if (kmsg_buf.head >= kmsg_buf.len) kmsg_buf.head = 0;
@@ -170,6 +175,17 @@ int kprintf(const char *format, ...)
 	return len;
 }
 
+/* mode:
+    0: mcklogd is not running.
+       When kmsg buffer is full, writer doesn't block
+       and overwrites the buffer.
+    1: mcklogd periodically retrieves kmsg.
+       When kmsg buffer is full, writer blocks until 
+       someone retrieves kmsg.
+    2: mcklogd periodically retrieves kmsg.
+       When kmsg buffer is full, writer doesn't block
+       and overwrites the buffer.
+*/
 void kmsg_init(int mode)
 {
 	ihk_mc_spinlock_init(&kmsg_lock);

kernel/devobj.c

@@ -127,6 +127,7 @@ int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxp
 
 	obj->memobj.ops = &devobj_ops;
 	obj->memobj.flags = MF_HAS_PAGER;
+	obj->memobj.size = len;
 	obj->handle = result.handle;
 	obj->ref = 1;
 	obj->pfn_pgoff = off / PAGE_SIZE;

kernel/fileobj.c

@@ -29,22 +29,27 @@
 #define	dkprintf(...)	do { if (0) kprintf(__VA_ARGS__); } while (0)
 #define	ekprintf(...)	kprintf(__VA_ARGS__)
 
-static ihk_spinlock_t fileobj_list_lock = SPIN_LOCK_UNLOCKED;
+mcs_rwlock_lock_t fileobj_list_lock =
+	{{{0}, MCS_RWLOCK_TYPE_COMMON_READER, 0, 0, 0, NULL}, NULL};
 static LIST_HEAD(fileobj_list);
 
+#define FILEOBJ_PAGE_HASH_SHIFT 9
+#define FILEOBJ_PAGE_HASH_SIZE (1 << FILEOBJ_PAGE_HASH_SHIFT)
+#define FILEOBJ_PAGE_HASH_MASK (FILEOBJ_PAGE_HASH_SIZE - 1)
+
 struct fileobj {
-	struct memobj		memobj;		/* must be first */
-	long			sref;
-	long			cref;
-	uintptr_t		handle;
-	struct list_head	page_list;
-	struct list_head	list;
+	struct memobj memobj;		/* must be first */
+	long sref;
+	long cref;
+	uintptr_t handle;
+	struct list_head list;
+	struct list_head page_hash[FILEOBJ_PAGE_HASH_SIZE];
+	mcs_rwlock_lock_t page_hash_locks[FILEOBJ_PAGE_HASH_SIZE];
 };
 
 static memobj_release_func_t fileobj_release;
 static memobj_ref_func_t fileobj_ref;
 static memobj_get_page_func_t fileobj_get_page;
-static memobj_copy_page_func_t fileobj_copy_page;
 static memobj_flush_page_func_t fileobj_flush_page;
 static memobj_invalidate_page_func_t fileobj_invalidate_page;
 static memobj_lookup_page_func_t fileobj_lookup_page;
@@ -53,7 +58,7 @@ static struct memobj_ops fileobj_ops = {
 	.release =	&fileobj_release,
 	.ref =		&fileobj_ref,
 	.get_page =	&fileobj_get_page,
-	.copy_page =	&fileobj_copy_page,
+	.copy_page =	NULL,
 	.flush_page =	&fileobj_flush_page,
 	.invalidate_page =	&fileobj_invalidate_page,
 	.lookup_page =	&fileobj_lookup_page,
@@ -72,28 +77,36 @@ static struct memobj *to_memobj(struct fileobj *fileobj)
 /***********************************************************************
  * page_list
  */
-static void page_list_init(struct fileobj *obj)
+static void fileobj_page_hash_init(struct fileobj *obj)
 {
-	INIT_LIST_HEAD(&obj->page_list);
+	int i;
+	for (i = 0; i < FILEOBJ_PAGE_HASH_SIZE; ++i) {
+		mcs_rwlock_init(&obj->page_hash_locks[i]);
+		INIT_LIST_HEAD(&obj->page_hash[i]);
+	}
 	return;
 }
 
-static void page_list_insert(struct fileobj *obj, struct page *page)
+/* NOTE: caller must hold page_hash_locks[hash] */
+static void __fileobj_page_hash_insert(struct fileobj *obj,
+		struct page *page, int hash)
 {
-	list_add(&page->list, &obj->page_list);
-	return;
+	list_add(&page->list, &obj->page_hash[hash]);
 }
 
-static void page_list_remove(struct fileobj *obj, struct page *page)
+/* NOTE: caller must hold page_hash_locks[hash] */
+static void __fileobj_page_hash_remove(struct page *page)
 {
 	list_del(&page->list);
 }
 
-static struct page *page_list_lookup(struct fileobj *obj, off_t off)
+/* NOTE: caller must hold page_hash_locks[hash] */
+static struct page *__fileobj_page_hash_lookup(struct fileobj *obj,
+		int hash, off_t off)
 {
 	struct page *page;
 
-	list_for_each_entry(page, &obj->page_list, list) {
+	list_for_each_entry(page, &obj->page_hash[hash], list) {
 		if ((page->mode != PM_WILL_PAGEIO)
 				&& (page->mode != PM_PAGEIO)
 				&& (page->mode != PM_DONE_PAGEIO)
@@ -104,6 +117,7 @@ static struct page *page_list_lookup(struct fileobj *obj, off_t off)
 					obj, off, page->mode);
 			panic("page_list_lookup:invalid obj page");
 		}
+
 		if (page->offset == off) {
 			goto out;
 		}
@@ -114,13 +128,22 @@ out:
 	return page;
 }
 
-static struct page *page_list_first(struct fileobj *obj)
+static struct page *fileobj_page_hash_first(struct fileobj *obj)
 {
-	if (list_empty(&obj->page_list)) {
-		return NULL;
+	int i;
+
+	for (i = 0; i < FILEOBJ_PAGE_HASH_SIZE; ++i) {
+		if (!list_empty(&obj->page_hash[i])) {
+			break;
+		}
 	}
 
-	return list_first_entry(&obj->page_list, struct page, list);
+	if (i != FILEOBJ_PAGE_HASH_SIZE) {
+		return list_first_entry(&obj->page_hash[i], struct page, list);
+	}
+	else {
+		return NULL;
+	}
 }
 
 /***********************************************************************
@@ -163,10 +186,11 @@ static struct fileobj *obj_list_lookup(uintptr_t handle)
 int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
 {
 	ihk_mc_user_context_t ctx;
-	struct pager_create_result result;	// XXX: assumes contiguous physical
+	struct pager_create_result result __attribute__((aligned(64)));	
 	int error;
 	struct fileobj *newobj  = NULL;
 	struct fileobj *obj;
+	struct mcs_rwlock_node node;
 
 	dkprintf("fileobj_create(%d)\n", fd);
 	newobj = kmalloc(sizeof(*newobj), IHK_MC_AP_NOWAIT);
@@ -179,6 +203,7 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
 	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_CREATE;
 	ihk_mc_syscall_arg1(&ctx) = fd;
 	ihk_mc_syscall_arg2(&ctx) = virt_to_phys(&result);
+	memset(&result, 0, sizeof(result));
 
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
 	if (error) {
@@ -192,23 +217,39 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
 	newobj->handle = result.handle;
 	newobj->sref = 1;
 	newobj->cref = 1;
-	page_list_init(newobj);
+	fileobj_page_hash_init(newobj);
 	ihk_mc_spinlock_init(&newobj->memobj.lock);
 
-	ihk_mc_spinlock_lock_noirq(&fileobj_list_lock);
+	mcs_rwlock_writer_lock_noirq(&fileobj_list_lock, &node);
 	obj = obj_list_lookup(result.handle);
 	if (!obj) {
 		obj_list_insert(newobj);
 		obj = newobj;
+		to_memobj(obj)->size = result.size;
+		to_memobj(obj)->flags |= result.flags;
+		to_memobj(obj)->status = MEMOBJ_READY;
+		if (to_memobj(obj)->flags & MF_PREFETCH) {
+			to_memobj(obj)->status = MEMOBJ_TO_BE_PREFETCHED;
+		}
 		newobj = NULL;
+		dkprintf("%s: new obj 0x%lx cref: %d, %s\n",
+			__FUNCTION__,
+			obj,
+			obj->cref,
+			to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
 	}
 	else {
 		++obj->sref;
 		++obj->cref;
 		memobj_unlock(&obj->memobj);	/* locked by obj_list_lookup() */
+		dkprintf("%s: existing obj 0x%lx cref: %d, %s\n",
+			__FUNCTION__,
+			obj,
+			obj->cref,
+			to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
 	}
 
-	ihk_mc_spinlock_unlock_noirq(&fileobj_list_lock);
+	mcs_rwlock_writer_unlock_noirq(&fileobj_list_lock, &node);
 
 	error = 0;
 	*objp = to_memobj(obj);
@@ -239,6 +280,7 @@ static void fileobj_release(struct memobj *memobj)
 	long free_sref = 0;
 	uintptr_t free_handle;
 	struct fileobj *free_obj = NULL;
+	struct mcs_rwlock_node node;
 
 	dkprintf("fileobj_release(%p %lx)\n", obj, obj->handle);
 
@@ -254,17 +296,34 @@ static void fileobj_release(struct memobj *memobj)
 	memobj_unlock(&obj->memobj);
 
 	if (free_obj) {
-		ihk_mc_spinlock_lock_noirq(&fileobj_list_lock);
+		dkprintf("%s: release obj 0x%lx cref: %d, free_obj: 0x%lx, %s\n",
+				__FUNCTION__,
+				obj,
+				obj->cref,
+				free_obj,
+				to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
+		mcs_rwlock_writer_lock_noirq(&fileobj_list_lock, &node);
 		/* zap page_list */
 		for (;;) {
 			struct page *page;
-			int count;
+			void *page_va;
 
-			page = page_list_first(obj);
+			page = fileobj_page_hash_first(obj);
 			if (!page) {
 				break;
 			}
-			page_list_remove(obj, page);
+			__fileobj_page_hash_remove(page);
+			page_va = phys_to_virt(page_to_phys(page));
+
+			if (ihk_atomic_read(&page->count) != 1) {
+				kprintf("%s: WARNING: page count for phys 0x%lx is invalid\n",
+					__FUNCTION__, page->phys);
+			}
+
+			if (page_unmap(page)) {
+				ihk_mc_free_pages(page_va, 1);
+			}
+#if 0
 			count = ihk_atomic_sub_return(1, &page->count);
 
 			if (!((page->mode == PM_WILL_PAGEIO)
@@ -281,10 +340,10 @@ static void fileobj_release(struct memobj *memobj)
 			}
 
 			page->mode = PM_NONE;
-			free_pages(phys_to_virt(page_to_phys(page)), 1);
+#endif
 		}
 		obj_list_remove(free_obj);
-		ihk_mc_spinlock_unlock_noirq(&fileobj_list_lock);
+		mcs_rwlock_writer_unlock_noirq(&fileobj_list_lock, &node);
 		kfree(free_obj);
 	}
 
@@ -330,83 +389,101 @@ static void fileobj_do_pageio(void *args0)
 	struct page *page;
 	ihk_mc_user_context_t ctx;
 	ssize_t ss;
+	struct mcs_rwlock_node mcs_node;
+	int hash = (off >> PAGE_SHIFT) & FILEOBJ_PAGE_HASH_MASK;	
 
-	memobj_lock(&obj->memobj);
-	page = page_list_lookup(obj, off);
+	mcs_rwlock_writer_lock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
+	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page) {
 		goto out;
 	}
 
 	while (page->mode == PM_PAGEIO) {
-		memobj_unlock(&obj->memobj);
+		mcs_rwlock_writer_unlock_noirq(&obj->page_hash_locks[hash],
+				&mcs_node);
 		cpu_pause();
-		memobj_lock(&obj->memobj);
+		mcs_rwlock_writer_lock_noirq(&obj->page_hash_locks[hash],
+				&mcs_node);
 	}
 
 	if (page->mode == PM_WILL_PAGEIO) {
-		page->mode = PM_PAGEIO;
-		memobj_unlock(&obj->memobj);
-
-		ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_READ;
-		ihk_mc_syscall_arg1(&ctx) = obj->handle;
-		ihk_mc_syscall_arg2(&ctx) = off;
-		ihk_mc_syscall_arg3(&ctx) = pgsize;
-		ihk_mc_syscall_arg4(&ctx) = page_to_phys(page);
-
-		ss = syscall_generic_forwarding(__NR_mmap, &ctx);
-
-		memobj_lock(&obj->memobj);
-		if (page->mode != PM_PAGEIO) {
-			kprintf("fileobj_do_pageio(%p,%lx,%lx):"
-					"invalid mode %x\n",
-					obj, off, pgsize, page->mode);
-			panic("fileobj_do_pageio:invalid page mode");
+		if (to_memobj(obj)->flags & MF_ZEROFILL) {
+			void *virt = phys_to_virt(page_to_phys(page));
+			memset(virt, 0, PAGE_SIZE);
 		}
+		else {
+			page->mode = PM_PAGEIO;
+			mcs_rwlock_writer_unlock_noirq(&obj->page_hash_locks[hash],
+					&mcs_node);
 
-		if (ss == 0) {
-			dkprintf("fileobj_do_pageio(%p,%lx,%lx):EOF? %ld\n",
-					obj, off, pgsize, ss);
-			page->mode = PM_PAGEIO_EOF;
-			goto out;
-		}
-		else if (ss != pgsize) {
-			kprintf("fileobj_do_pageio(%p,%lx,%lx):"
-					"read failed. %ld\n",
-					obj, off, pgsize, ss);
-			page->mode = PM_PAGEIO_ERROR;
-			goto out;
+			ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_READ;
+			ihk_mc_syscall_arg1(&ctx) = obj->handle;
+			ihk_mc_syscall_arg2(&ctx) = off;
+			ihk_mc_syscall_arg3(&ctx) = pgsize;
+			ihk_mc_syscall_arg4(&ctx) = page_to_phys(page);
+
+			dkprintf("%s: __NR_mmap for handle 0x%lx\n",
+					__FUNCTION__, obj->handle);
+			ss = syscall_generic_forwarding(__NR_mmap, &ctx);
+
+			mcs_rwlock_writer_lock_noirq(&obj->page_hash_locks[hash],
+					&mcs_node);
+			if (page->mode != PM_PAGEIO) {
+				kprintf("fileobj_do_pageio(%p,%lx,%lx):"
+						"invalid mode %x\n",
+						obj, off, pgsize, page->mode);
+				panic("fileobj_do_pageio:invalid page mode");
+			}
+
+			if (ss == 0) {
+				dkprintf("fileobj_do_pageio(%p,%lx,%lx):EOF? %ld\n",
+						obj, off, pgsize, ss);
+				page->mode = PM_PAGEIO_EOF;
+				goto out;
+			}
+			else if (ss != pgsize) {
+				kprintf("fileobj_do_pageio(%p,%lx,%lx):"
+						"read failed. %ld\n",
+						obj, off, pgsize, ss);
+				page->mode = PM_PAGEIO_ERROR;
+				goto out;
+			}
 		}
 
 		page->mode = PM_DONE_PAGEIO;
 	}
 out:
-	memobj_unlock(&obj->memobj);
+	mcs_rwlock_writer_unlock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
 	fileobj_release(&obj->memobj);		/* got fileobj_get_page() */
 	kfree(args0);
 	dkprintf("fileobj_do_pageio(%p,%lx,%lx):\n", obj, off, pgsize);
 	return;
 }
 
-static int fileobj_get_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *pflag)
+static int fileobj_get_page(struct memobj *memobj, off_t off,
+		int p2align, uintptr_t *physp, unsigned long *pflag)
 {
 	struct thread *proc = cpu_local_var(current);
 	struct fileobj *obj = to_fileobj(memobj);
-	int error;
+	int error = -1;
 	void *virt = NULL;
 	int npages;
 	uintptr_t phys = -1;
 	struct page *page;
 	struct pageio_args *args = NULL;
+	struct mcs_rwlock_node mcs_node;
+	int hash = (off >> PAGE_SHIFT) & FILEOBJ_PAGE_HASH_MASK;	
 
 	dkprintf("fileobj_get_page(%p,%lx,%x,%p)\n", obj, off, p2align, physp);
-
-	memobj_lock(&obj->memobj);
 	if (p2align != PAGE_P2ALIGN) {
-		error = -ENOMEM;
-		goto out;
+		return -ENOMEM;
 	}
 
-	page = page_list_lookup(obj, off);
+	mcs_rwlock_writer_lock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
+	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page || (page->mode == PM_WILL_PAGEIO)
 			|| (page->mode == PM_PAGEIO)) {
 		args = kmalloc(sizeof(*args), IHK_MC_AP_NOWAIT);
@@ -430,17 +507,19 @@ static int fileobj_get_page(struct memobj *memobj, off_t off, int p2align, uintp
 				goto out;
 			}
 			phys = virt_to_phys(virt);
-			page = phys_to_page(phys);
+			page = phys_to_page_insert_hash(phys);
 			if (page->mode != PM_NONE) {
 				panic("fileobj_get_page:invalid new page");
 			}
-			page->mode = PM_WILL_PAGEIO;
 			page->offset = off;
 			ihk_atomic_set(&page->count, 1);
-			page_list_insert(obj, page);
+			__fileobj_page_hash_insert(obj, page, hash);
+			page->mode = PM_WILL_PAGEIO;
 		}
 
+		memobj_lock(&obj->memobj);
 		++obj->cref;	/* for fileobj_do_pageio() */
+		memobj_unlock(&obj->memobj);
 
 		args->fileobj = obj;
 		args->objoff = off;
@@ -472,7 +551,8 @@ static int fileobj_get_page(struct memobj *memobj, off_t off, int p2align, uintp
 	*physp = page_to_phys(page);
 	virt = NULL;
 out:
-	memobj_unlock(&obj->memobj);
+	mcs_rwlock_writer_unlock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
 	if (virt) {
 		ihk_mc_free_pages(virt, npages);
 	}
@@ -484,76 +564,6 @@ out:
 	return error;
 }
 
-static uintptr_t fileobj_copy_page(
-		struct memobj *memobj, uintptr_t orgpa, int p2align)
-{
-	struct page *orgpage = phys_to_page(orgpa);
-	size_t pgsize = PAGE_SIZE << p2align;
-	int npages = 1 << p2align;
-	void *newkva = NULL;
-	uintptr_t newpa = -1;
-	void *orgkva;
-	int count;
-
-	dkprintf("fileobj_copy_page(%p,%lx,%d)\n", memobj, orgpa, p2align);
-	if (p2align != PAGE_P2ALIGN) {
-		panic("p2align");
-	}
-
-	memobj_lock(memobj);
-	for (;;) {
-		if (orgpage->mode != PM_MAPPED) {
-			kprintf("fileobj_copy_page(%p,%lx,%d):"
-					"invalid cow page. %x\n",
-					memobj, orgpa, p2align, orgpage->mode);
-			panic("fileobj_copy_page:invalid cow page");
-		}
-		count = ihk_atomic_read(&orgpage->count);
-		if (count == 2) {	// XXX: private only
-			list_del(&orgpage->list);
-			ihk_atomic_dec(&orgpage->count);
-			orgpage->mode = PM_NONE;
-			newpa = orgpa;
-			break;
-		}
-		if (count <= 0) {
-			kprintf("fileobj_copy_page(%p,%lx,%d):"
-					"orgpage count corrupted. %x\n",
-					memobj, orgpa, p2align, count);
-			panic("fileobj_copy_page:orgpage count corrupted");
-		}
-		if (newkva) {
-			orgkva = phys_to_virt(orgpa);
-			memcpy(newkva, orgkva, pgsize);
-			ihk_atomic_dec(&orgpage->count);
-			newpa = virt_to_phys(newkva);
-			page_map(phys_to_page(newpa));
-			newkva = NULL;	/* avoid ihk_mc_free_pages() */
-			break;
-		}
-
-		memobj_unlock(memobj);
-		newkva = ihk_mc_alloc_aligned_pages(npages, p2align,
-				IHK_MC_AP_NOWAIT);
-		if (!newkva) {
-			kprintf("fileobj_copy_page(%p,%lx,%d):"
-					"alloc page failed\n",
-					memobj, orgpa, p2align);
-			goto out;
-		}
-		memobj_lock(memobj);
-	}
-	memobj_unlock(memobj);
-
-out:
-	if (newkva) {
-		ihk_mc_free_pages(newkva, npages);
-	}
-	dkprintf("fileobj_copy_page(%p,%lx,%d): %lx\n",
-			memobj, orgpa, p2align, newpa);
-	return newpa;
-}
-
 static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 		size_t pgsize)
 {
@@ -562,7 +572,16 @@ static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 	ihk_mc_user_context_t ctx;
 	ssize_t ss;
 
+	if (to_memobj(obj)->flags & MF_ZEROFILL) {
+		return 0;
+	}
+
 	page = phys_to_page(phys);
+	if (!page) {
+		kprintf("%s: warning: tried to flush non-existing page for phys addr: 0x%lx\n", 
+			__FUNCTION__, phys);
+		return 0;
+	}
 	memobj_unlock(&obj->memobj);
 
 	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_WRITE;
@@ -585,63 +604,48 @@ static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 static int fileobj_invalidate_page(struct memobj *memobj, uintptr_t phys,
 		size_t pgsize)
 {
-	struct fileobj *obj = to_fileobj(memobj);
-	int error;
-	struct page *page;
-
 	dkprintf("fileobj_invalidate_page(%p,%#lx,%#lx)\n",
 			memobj, phys, pgsize);
 
-	if (!(page = phys_to_page(phys))
-			|| !(page = page_list_lookup(obj, page->offset))) {
-		error = 0;
-		goto out;
-	}
-
-	if (ihk_atomic_read(&page->count) == 1) {
-		if (page_unmap(page)) {
-			ihk_mc_free_pages(phys_to_virt(phys),
-					pgsize/PAGE_SIZE);
-		}
-	}
-
-	error = 0;
-out:
-	dkprintf("fileobj_invalidate_page(%p,%#lx,%#lx):%d\n",
-			memobj, phys, pgsize, error);
-	return error;
+	/* TODO: keep track of reverse mappings so that invalidation
+	 * can be performed */
+	kprintf("%s: WARNING: file mapping invalidation not supported\n",
+		__FUNCTION__);
+	return 0;
 }
 
-static int fileobj_lookup_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *pflag)
+static int fileobj_lookup_page(struct memobj *memobj, off_t off,
+		int p2align, uintptr_t *physp, unsigned long *pflag)
 {
 	struct fileobj *obj = to_fileobj(memobj);
-	int error;
-	uintptr_t phys = -1;
+	int error = -1;
 	struct page *page;
+	struct mcs_rwlock_node mcs_node;
+	int hash = (off >> PAGE_SHIFT) & FILEOBJ_PAGE_HASH_MASK;
 
 	dkprintf("fileobj_lookup_page(%p,%lx,%x,%p)\n", obj, off, p2align, physp);
 
-	memobj_lock(&obj->memobj);
 	if (p2align != PAGE_P2ALIGN) {
-		error = -ENOMEM;
-		goto out;
+		return -ENOMEM;
 	}
 
-	page = page_list_lookup(obj, off);
+	mcs_rwlock_reader_lock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
+
+	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page) {
-		error = -ENOENT;
-		dkprintf("fileobj_lookup_page(%p,%lx,%x,%p): page not found. %d\n", obj, off, p2align, physp, error);
 		goto out;
 	}
-	phys = page_to_phys(page);
 
+	*physp = page_to_phys(page);
 	error = 0;
-	if (physp) {
-		*physp = phys;
-	}
+
 out:
-	memobj_unlock(&obj->memobj);
-	dkprintf("fileobj_lookup_page(%p,%lx,%x,%p): %d %lx\n",
-			obj, off, p2align, physp, error, phys);
+	mcs_rwlock_reader_unlock_noirq(&obj->page_hash_locks[hash],
+			&mcs_node);
+
+	dkprintf("fileobj_lookup_page(%p,%lx,%x,%p): %d \n",
+			obj, off, p2align, physp, error);
 	return error;
 }
+

kernel/host.c

@@ -393,7 +393,9 @@ static int process_msg_prepare_process(unsigned long rphys)
 	memcpy_long(pn, p, sizeof(struct program_load_desc) 
 	            + sizeof(struct program_image_section) * n);
 
-	if((thread = create_thread(p->entry)) == NULL){
+	if ((thread = create_thread(p->entry,
+					(unsigned long *)&p->cpu_set,
+					sizeof(p->cpu_set))) == NULL) {
 		kfree(pn);
 		ihk_mc_unmap_virtual(p, npages, 1);
 		ihk_mc_unmap_memory(NULL, phys, sz);
@@ -534,31 +536,6 @@ extern void process_procfs_request(unsigned long rarg);
 extern void terminate_host(int pid);
 extern void debug_log(long);
 
-static void req_get_cpu_mapping(long req_rpa)
-{
-	size_t mapsize;
-	size_t size;
-	int npages;
-	long phys;
-	struct get_cpu_mapping_req *req;
-	struct cpu_mapping *buf;
-
-	size = sizeof(*req);
-	mapsize = size + (req_rpa & (PAGE_SIZE - 1));
-	npages = (mapsize + PAGE_SIZE - 1) >> PAGE_SHIFT;
-	phys = ihk_mc_map_memory(NULL, req_rpa, size);
-	req = ihk_mc_map_virtual(phys, npages, PTATTR_WRITABLE);
-
-	req->error = arch_get_cpu_mapping(&buf, &req->buf_elems);
-	if (!req->error) {
-		req->buf_rpa = virt_to_phys(buf);
-	}
-
-	ihk_mc_unmap_virtual(req, npages, 0);
-	ihk_mc_unmap_memory(NULL, phys, size);
-	return;
-} /* req_get_cpu_mapping() */
-
 static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
                                   void *__packet, void *ihk_os)
 {
@@ -604,24 +581,24 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		break;
 
 	case SCD_MSG_SCHEDULE_PROCESS:
-		cpuid = obtain_clone_cpuid();
-		if(cpuid == -1){
+		thread = (struct thread *)packet->arg;
+
+		cpuid = obtain_clone_cpuid(&thread->cpu_set);
+		if (cpuid == -1) {
 			kprintf("No CPU available\n");
 			ret = -1;
 			break;
 		}
-		dkprintf("SCD_MSG_SCHEDULE_PROCESS: %lx\n", packet->arg);
-		thread = (struct thread *)packet->arg;
-		proc = thread->proc;
 
-		settid(thread, 0, cpuid, -1, 0, NULL);
+		dkprintf("SCD_MSG_SCHEDULE_PROCESS: %lx\n", packet->arg);
+		proc = thread->proc;
+		thread->tid = proc->pid;
 		proc->status = PS_RUNNING;
 		thread->status = PS_RUNNING;
 		chain_thread(thread);
 		chain_process(proc);
 		runq_add_thread(thread, cpuid);
-					  
-		//cpu_local_var(next) = (struct thread *)packet->arg;
+
 		ret = 0;
 		break;
 
@@ -688,15 +665,6 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		ret = 0;
 		break;
 
-	case SCD_MSG_GET_CPU_MAPPING:
-		req_get_cpu_mapping(packet->arg);
-
-		pckt.msg = SCD_MSG_REPLY_GET_CPU_MAPPING;
-		pckt.arg = packet->arg;
-		syscall_channel_send(c, &pckt);
-		ret = 0;
-		break;
-
 	default:
 		kprintf("syscall_pakcet_handler:unknown message "
 				"(%d.%d.%d.%d.%d.%#lx)\n",
@@ -718,7 +686,7 @@ void init_host_syscall_channel(void)
 
 	param.port = 501;
 	param.pkt_size = sizeof(struct ikc_scd_packet);
-	param.queue_size = PAGE_SIZE;
+	param.queue_size = PAGE_SIZE * 4;
 	param.magic = 0x1129;
 	param.handler = syscall_packet_handler;
 
@@ -745,7 +713,7 @@ void init_host_syscall_channel2(void)
 
 	param.port = 502;
 	param.pkt_size = sizeof(struct ikc_scd_packet);
-	param.queue_size = PAGE_SIZE;
+	param.queue_size = PAGE_SIZE * 4;
 	param.magic = 0x1329;
 	param.handler = syscall_packet_handler;
 

kernel/include/init.h

@@ -16,7 +16,7 @@
 extern void arch_init(void);
 extern void kmsg_init(int);
 extern void mem_init(void);
-extern void ikc_master_init(void);
+extern void ihk_ikc_master_init(void);
 extern void ap_init(void);
 extern void arch_ready(void);
 extern void mc_ikc_test_init(void);
@@ -32,4 +32,6 @@ extern void cpu_sysfs_setup(void);
 
 extern char *find_command_line(char *name);
 
+extern int num_processors;
+
 #endif

kernel/include/memobj.h

@@ -32,13 +32,20 @@ enum {
 	MF_HAS_PAGER	= 0x0001,
 	MF_SHMDT_OK	= 0x0002,
 	MF_IS_REMOVABLE	= 0x0004,
+	MF_PREFETCH = 0x0008,
+	MF_ZEROFILL = 0x0010,
+	MF_END
 };
 
+#define MEMOBJ_READY              0
+#define MEMOBJ_TO_BE_PREFETCHED   1
+
 struct memobj {
-	struct memobj_ops *	ops;
-	uint32_t		flags;
-	int8_t			padding[4];
-	ihk_spinlock_t		lock;
+	struct memobj_ops *ops;
+	uint32_t flags;
+	uint32_t status;
+	size_t size;
+	ihk_spinlock_t lock;
 };
 
 typedef void memobj_release_func_t(struct memobj *obj);

kernel/include/page.h

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

kernel/include/pager.h

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

kernel/include/process.h

@@ -22,6 +22,7 @@
 #include <memobj.h>
 #include <affinity.h>
 #include <syscall.h>
+#include <bitops.h>
 
 #define VR_NONE            0x0
 #define VR_STACK           0x1
@@ -165,9 +166,74 @@
 
 #define NOPHYS ((uintptr_t)-1)
 
+#define PROCESS_NUMA_MASK_BITS 256
+
+/*
+ * Both the MPOL_* mempolicy mode and the MPOL_F_* optional mode flags are
+ * passed by the user to either set_mempolicy() or mbind() in an 'int' actual.
+ * The MPOL_MODE_FLAGS macro determines the legal set of optional mode flags.
+ */
+
+/* Policies */
+enum {
+	MPOL_DEFAULT,
+	MPOL_PREFERRED,
+	MPOL_BIND,
+	MPOL_INTERLEAVE,
+	MPOL_LOCAL,
+	MPOL_MAX,	/* always last member of enum */
+};
+
+enum mpol_rebind_step {
+	MPOL_REBIND_ONCE,	/* do rebind work at once(not by two step) */
+	MPOL_REBIND_STEP1,	/* first step(set all the newly nodes) */
+	MPOL_REBIND_STEP2,	/* second step(clean all the disallowed nodes)*/
+	MPOL_REBIND_NSTEP,
+};
+
+/* Flags for set_mempolicy */
+#define MPOL_F_STATIC_NODES	(1 << 15)
+#define MPOL_F_RELATIVE_NODES	(1 << 14)
+
+/*
+ * MPOL_MODE_FLAGS is the union of all possible optional mode flags passed to
+ * either set_mempolicy() or mbind().
+ */
+#define MPOL_MODE_FLAGS	(MPOL_F_STATIC_NODES | MPOL_F_RELATIVE_NODES)
+
+/* Flags for get_mempolicy */
+#define MPOL_F_NODE	(1<<0)	/* return next IL mode instead of node mask */
+#define MPOL_F_ADDR	(1<<1)	/* look up vma using address */
+#define MPOL_F_MEMS_ALLOWED (1<<2) /* return allowed memories */
+
+/* Flags for mbind */
+#define MPOL_MF_STRICT	(1<<0)	/* Verify existing pages in the mapping */
+#define MPOL_MF_MOVE	 (1<<1)	/* Move pages owned by this process to conform
+				   to policy */
+#define MPOL_MF_MOVE_ALL (1<<2)	/* Move every page to conform to policy */
+#define MPOL_MF_LAZY	 (1<<3)	/* Modifies '_MOVE:  lazy migrate on fault */
+#define MPOL_MF_INTERNAL (1<<4)	/* Internal flags start here */
+
+#define MPOL_MF_VALID	(MPOL_MF_STRICT   | 	\
+			 MPOL_MF_MOVE     | 	\
+			 MPOL_MF_MOVE_ALL)
+
+/*
+ * Internal flags that share the struct mempolicy flags word with
+ * "mode flags".  These flags are allocated from bit 0 up, as they
+ * are never OR'ed into the mode in mempolicy API arguments.
+ */
+#define MPOL_F_SHARED  (1 << 0)	/* identify shared policies */
+#define MPOL_F_LOCAL   (1 << 1)	/* preferred local allocation */
+#define MPOL_F_REBINDING (1 << 2)	/* identify policies in rebinding */
+#define MPOL_F_MOF	(1 << 3) /* this policy wants migrate on fault */
+#define MPOL_F_MORON	(1 << 4) /* Migrate On pte_numa Reference On Node */
+
 #include <waitq.h>
 #include <futex.h>
 
+//#define TRACK_SYSCALLS
+
 struct resource_set;
 struct process_hash;
 struct thread_hash;
@@ -305,6 +371,13 @@ struct vm_range {
 	int padding;
 };
 
+struct vm_range_numa_policy {
+	struct list_head list;
+	unsigned long start, end;
+	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
+	int numa_mem_policy;
+};
+
 struct vm_regions {
 	unsigned long vm_start, vm_end;
 	unsigned long text_start, text_end;
@@ -334,7 +407,7 @@ struct mckfd {
 #define SFD_NONBLOCK 04000
 
 struct sig_common {
-	ihk_spinlock_t	lock;
+	mcs_rwlock_lock_t lock;
 	ihk_atomic_t use;
 	struct k_sigaction action[_NSIG];
 	struct list_head sigpending;
@@ -395,7 +468,7 @@ struct process {
 			// V       +----   |
 			// PS_STOPPED -----+
 			// (PS_TRACED)
-	int exit_status;
+	int exit_status; // only for zombie
 
 	/* Store exit_status for a group of threads when stopped by SIGSTOP.
 	   exit_status can't be used because values of exit_status of threads
@@ -507,6 +580,7 @@ struct thread {
 			// PS_TRACED
 			// PS_INTERRPUTIBLE
 			// PS_UNINTERRUPTIBLE
+	int exit_status;
 
 	// process vm
 	struct process_vm *vm;
@@ -537,12 +611,20 @@ struct thread {
 	fp_regs_struct *fp_regs;
 	int in_syscall_offload;
 
+#ifdef TRACK_SYSCALLS
+	int socc_enabled;
+	uint64_t *syscall_times;
+	uint32_t *syscall_cnts;
+	uint64_t *offload_times;
+	uint32_t *offload_cnts;
+#endif // TRACK_SYSCALLS
+
 	// signal
 	struct sig_common *sigcommon;
 	sigset_t sigmask;
 	stack_t sigstack;
 	struct list_head sigpending;
-	ihk_spinlock_t	sigpendinglock;
+	mcs_rwlock_lock_t sigpendinglock;
 	volatile int sigevent;
 
 	// gpio
@@ -572,6 +654,8 @@ struct thread {
 	struct waitq scd_wq;
 };
 
+#define VM_RANGE_CACHE_SIZE	4
+
 struct process_vm {
 	struct address_space *address_space;
 	struct list_head vm_range_list;
@@ -594,6 +678,12 @@ struct process_vm {
 	int exiting;
 
 	long currss;
+	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
+	int numa_mem_policy;
+	/* Protected by memory_range_lock */
+	struct list_head vm_range_numa_policy_list;
+	struct vm_range *range_cache[VM_RANGE_CACHE_SIZE];
+	int range_cache_ind;
 };
 
 static inline int has_cap_ipc_lock(struct thread *th)
@@ -610,7 +700,8 @@ static inline int has_cap_sys_admin(struct thread *th)
 
 void hold_address_space(struct address_space *);
 void release_address_space(struct address_space *);
-struct thread *create_thread(unsigned long user_pc);
+struct thread *create_thread(unsigned long user_pc,
+		unsigned long *__cpu_set, size_t cpu_set_size);
 struct thread *clone_thread(struct thread *org, unsigned long pc,
                               unsigned long sp, int clone_flags);
 void destroy_thread(struct thread *thread);
@@ -690,7 +781,5 @@ 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

@@ -149,6 +149,10 @@ struct program_image_section {
 #define MCK_RLIMIT_SIGPENDING	14
 #define MCK_RLIMIT_STACK	15
 
+#define PLD_CPU_SET_MAX_CPUS 1024
+typedef unsigned long __cpu_set_unit;
+#define PLD_CPU_SET_SIZE (PLD_CPU_SET_MAX_CPUS / (8 * sizeof(__cpu_set_unit)))
+
 struct program_load_desc {
 	int num_sections;
 	int status;
@@ -178,6 +182,7 @@ struct program_load_desc {
 	struct rlimit rlimit[MCK_RLIM_MAX];
 	unsigned long interp_align;
 	char shell_path[SHELL_PATH_MAX_LEN];
+	__cpu_set_unit cpu_set[PLD_CPU_SET_SIZE];
 	struct program_image_section sections[0];
 };
 
@@ -331,7 +336,7 @@ void delete_proc_procfs_files(int pid);
 void create_os_procfs_files(void);
 void delete_os_procfs_files(void);
 
-#define PROCFS_NAME_MAX 1000
+#define PROCFS_NAME_MAX 768
 
 struct procfs_read {
 	unsigned long pbuf;	/* physical address of the host buffer (request) */
@@ -387,6 +392,7 @@ extern struct tod_data_s tod_data;	/* residing in arch-dependent file */
 
 void reset_cputime();
 void set_cputime(int mode);
+int do_munmap(void *addr, size_t len);
 intptr_t do_mmap(intptr_t addr0, size_t len0, int prot, int flags, int fd,
 		off_t off0);
 void clear_host_pte(uintptr_t addr, size_t len);

kernel/include/xpmem.h

@@ -0,0 +1,21 @@
+/**
+ * \file xpmem.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Structures and functions of xpmem
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef _XPMEM_H
+#define _XPMEM_H
+
+#include <ihk/context.h>
+
+#define XPMEM_DEV_PATH  "/dev/xpmem"
+
+extern int xpmem_open(ihk_mc_user_context_t *ctx);
+
+#endif /* _XPMEM_H */
+

kernel/include/xpmem_private.h

@@ -0,0 +1,388 @@
+/**
+ * \file xpmem_private.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Private Cross Partition Memory (XPMEM) structures and macros.
+ */
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (c) 2004-2007 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright 2009, 2010, 2014 Cray Inc. All Rights Reserved
+ * Copyright (c) 2014-2016 Los Alamos National Security, LCC. All rights
+ *                         reserved.
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef _XPMEM_PRIVATE_H
+#define _XPMEM_PRIVATE_H
+
+#include <mc_xpmem.h>
+#include <xpmem.h>
+
+#define XPMEM_CURRENT_VERSION           0x00026003
+
+//#define DEBUG_PRINT_XPMEM
+
+#ifdef DEBUG_PRINT_XPMEM
+#define dkprintf(...) kprintf(__VA_ARGS__)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#define XPMEM_DEBUG(format, a...) kprintf("[%d] %s: "format"\n", cpu_local_var(current)->proc->rgid, __func__, ##a)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#define XPMEM_DEBUG(format, a...) do { if (0) kprintf("\n"); } while (0)
+#endif
+
+//#define USE_DBUG_ON
+
+#ifdef USE_DBUG_ON
+#define DBUG_ON(condition) do { if (condition) kprintf("[%d] BUG: func=%s\n", cpu_local_var(current)->proc->rgid, __func__); } while (0)
+#else
+#define DBUG_ON(condition)
+#endif
+
+#define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
+
+#define min(x, y)	({                                              \
+	__typeof__(x) _min1 = (x);                                      \
+	__typeof__(y) _min2 = (y);                                      \
+	(void) (&_min1 == &_min2);                                      \
+	_min1 < _min2 ? _min1 : _min2;})
+
+#define max(x, y)	({                                              \
+	__typeof__(x) _max1 = (x);                                      \
+	__typeof__(y) _max2 = (y);                                      \
+	(void) (&_max1 == &_max2);                                      \
+	_max1 > _max2 ? _max1 : _max2;})
+
+#define MAX_ERRNO	4095
+
+#define IS_ERR_VALUE(x)	((x) >= (unsigned long)-MAX_ERRNO)
+
+static inline void * ERR_PTR(long error)
+{
+	return (void *)error;
+}
+
+static inline long PTR_ERR(const void *ptr)
+{
+	return (long)ptr;
+}
+
+static inline long IS_ERR(const void *ptr)
+{
+	return IS_ERR_VALUE((unsigned long)ptr);
+}
+
+static inline long IS_ERR_OR_NULL(const void *ptr)
+{
+	return !ptr || IS_ERR_VALUE((unsigned long)ptr);
+}
+
+/*
+ * Both the xpmem_segid_t and xpmem_apid_t are of type __s64 and designed
+ * to be opaque to the user. Both consist of the same underlying fields.
+ *
+ * The 'uniq' field is designed to give each segid or apid a unique value.
+ * Each type is only unique with respect to itself.
+ *
+ * An ID is never less than or equal to zero.
+ */
+struct xpmem_id {
+	pid_t tgid;		/* thread group that owns ID */
+	unsigned int uniq;	/* this value makes the ID unique */
+};
+
+typedef union {
+	struct xpmem_id xpmem_id;
+	xpmem_segid_t segid;
+	xpmem_apid_t apid;
+} xpmem_id_t;
+
+/* Shift INT_MAX by one so we can tell when we overflow. */
+#define XPMEM_MAX_UNIQ_ID	(INT_MAX >> 1)
+
+static inline pid_t xpmem_segid_to_tgid(xpmem_segid_t segid)
+{
+	DBUG_ON(segid <= 0);
+	return ((xpmem_id_t *)&segid)->xpmem_id.tgid;
+}
+
+static inline pid_t xpmem_apid_to_tgid(xpmem_apid_t apid)
+{
+	DBUG_ON(apid <= 0);
+	return ((xpmem_id_t *)&apid)->xpmem_id.tgid;
+}
+
+/*
+ * Hash Tables
+ *
+ * XPMEM utilizes hash tables to enable faster lookups of list entries.
+ * These hash tables are implemented as arrays. A simple modulus of the hash
+ * key yields the appropriate array index. A hash table's array element (i.e.,
+ * hash table bucket) consists of a hash list and the lock that protects it.
+ *
+ * XPMEM has the following two hash tables:
+ *
+ * table                bucket                                  key
+ * part->tg_hashtable   list of struct xpmem_thread_group       tgid
+ * tg->ap_hashtable     list of struct xpmem_access_permit      apid.uniq
+ */
+struct xpmem_hashlist {
+        mcs_rwlock_lock_t lock;	/* lock for hash list */
+        struct list_head list;	/* hash list */
+};
+
+#define XPMEM_TG_HASHTABLE_SIZE 8
+#define XPMEM_AP_HASHTABLE_SIZE 8
+
+static inline int xpmem_tg_hashtable_index(pid_t tgid)
+{
+	int index;
+
+	index = (unsigned int)tgid % XPMEM_TG_HASHTABLE_SIZE;
+
+	XPMEM_DEBUG("return: tgid=%lu, index=%d", tgid, index);
+
+	return index;
+}
+
+static inline int xpmem_ap_hashtable_index(xpmem_apid_t apid)
+{
+	int index;
+
+        DBUG_ON(apid <= 0);
+
+	index = ((xpmem_id_t *)&apid)->xpmem_id.uniq % XPMEM_AP_HASHTABLE_SIZE;
+
+	XPMEM_DEBUG("return: apid=%lu, index=%d", apid, index);
+
+	return index;
+}
+
+/*
+ * general internal driver structures
+ */
+struct xpmem_thread_group {
+	ihk_spinlock_t lock;	/* tg lock */
+	pid_t tgid;		/* tg's tgid */
+	uid_t uid;		/* tg's uid */
+	gid_t gid;		/* tg's gid */
+	volatile int flags;	/* tg attributes and state */
+	ihk_atomic_t uniq_segid;
+	ihk_atomic_t uniq_apid;
+	mcs_rwlock_lock_t seg_list_lock;
+	struct list_head seg_list;	/* tg's list of segs */
+	ihk_atomic_t refcnt;	/* references to tg */
+	ihk_atomic_t n_pinned;	/* #of pages pinned by this tg */
+	struct list_head tg_hashlist;	/* tg hash list */
+	struct thread *group_leader;	/* thread group leader */
+	struct process_vm *vm;		/* tg's mm */
+	ihk_atomic_t n_recall_PFNs;	/* #of recall of PFNs in progress */
+	struct xpmem_hashlist ap_hashtable[];	/* locks + ap hash lists */
+};
+
+struct xpmem_segment {
+	ihk_spinlock_t lock;	/* seg lock */
+	mcs_rwlock_lock_t seg_lock;	/* seg sema */
+	xpmem_segid_t segid;	/* unique segid */
+	unsigned long vaddr;	/* starting address */
+	size_t size;		/* size of seg */
+	int permit_type;	/* permission scheme */
+	void *permit_value;	/* permission data */
+	volatile int flags;	/* seg attributes and state */
+	ihk_atomic_t refcnt;	/* references to seg */
+	struct xpmem_thread_group *tg;	/* creator tg */
+	struct list_head ap_list;	/* local access permits of seg */
+	struct list_head seg_list;	/* tg's list of segs */
+};
+
+struct xpmem_access_permit {
+	ihk_spinlock_t lock;	/* access permit lock */
+	xpmem_apid_t apid;	/* unique apid */
+	int mode;		/* read/write mode */
+	volatile int flags;	/* access permit attributes and state */
+	ihk_atomic_t refcnt;	/* references to access permit */
+	struct xpmem_segment *seg;	/* seg permitted to be accessed */
+	struct xpmem_thread_group *tg;	/* access permit's tg */
+	struct list_head att_list;	/* atts of this access permit's seg */
+	struct list_head ap_list;	/* access permits linked to seg */
+	struct list_head ap_hashlist;	/* access permit hash list */
+};
+
+struct xpmem_attachment {
+	mcs_rwlock_lock_t at_lock;	/* att lock for serialization */
+	struct mcs_rwlock_node_irqsave at_irqsave;	/* att lock for serialization */
+	unsigned long vaddr;	/* starting address of seg attached */
+	unsigned long at_vaddr;	/* address where seg is attached */
+	size_t at_size;		/* size of seg attachment */
+	struct vm_range *at_vma;	/* vma where seg is attachment */
+	volatile int flags;	/* att attributes and state */
+	ihk_atomic_t refcnt;	/* references to att */
+	struct xpmem_access_permit *ap;	/* associated access permit */
+	struct list_head att_list;	/* atts linked to access permit */
+	struct process_vm *vm;	/* mm struct attached to */
+	mcs_rwlock_lock_t invalidate_lock;	/* to serialize page table invalidates */
+};
+
+struct xpmem_partition {
+	ihk_atomic_t n_opened;	/* # of /dev/xpmem opened */
+	struct xpmem_hashlist tg_hashtable[];	/* locks + tg hash lists */
+};
+
+#define XPMEM_FLAG_DESTROYING		0x00040 /* being destroyed */
+#define XPMEM_FLAG_DESTROYED		0x00080 /* 'being destroyed' finished */
+
+#define XPMEM_FLAG_VALIDPTEs		0x00200 /* valid PTEs exist */
+
+struct xpmem_perm {
+	uid_t uid;
+	gid_t gid;
+	unsigned long mode;
+};
+
+#define XPMEM_PERM_IRUSR 00400
+#define XPMEM_PERM_IWUSR 00200
+
+static int xpmem_ioctl(struct mckfd *mckfd, ihk_mc_user_context_t *ctx);
+static int xpmem_close( struct mckfd *mckfd, ihk_mc_user_context_t *ctx);
+
+static int xpmem_init(void);
+static void xpmem_exit(void);
+static int __xpmem_open(void);
+static void xpmem_destroy_tg(struct xpmem_thread_group *);
+
+static int xpmem_make(unsigned long, size_t, int, void *, xpmem_segid_t *);
+static xpmem_segid_t xpmem_make_segid(struct xpmem_thread_group *);
+
+static int xpmem_remove(xpmem_segid_t);
+static void xpmem_remove_seg(struct xpmem_thread_group *,
+        struct xpmem_segment *);
+
+static void xpmem_clear_PTEs(struct xpmem_segment *);
+
+extern struct xpmem_partition *xpmem_my_part;
+
+static struct xpmem_thread_group * __xpmem_tg_ref_by_tgid_nolock_internal(
+	pid_t, int, int);
+
+static inline struct xpmem_thread_group *__xpmem_tg_ref_by_tgid(
+	pid_t tgid,
+	int return_destroying)
+{
+	struct xpmem_thread_group *tg;
+	int index;
+	struct mcs_rwlock_node_irqsave lock;
+
+	XPMEM_DEBUG("call: tgid=%d, return_destroying=%d", 
+		tgid, return_destroying);
+
+	index = xpmem_tg_hashtable_index(tgid);
+	mcs_rwlock_reader_lock(&xpmem_my_part->tg_hashtable[index].lock, &lock);
+	tg = __xpmem_tg_ref_by_tgid_nolock_internal(tgid, index, 
+		return_destroying);
+        mcs_rwlock_reader_unlock(&xpmem_my_part->tg_hashtable[index].lock, 
+		&lock);
+
+	XPMEM_DEBUG("return: tg=0x%p", tg);
+
+        return tg;
+}
+
+static inline struct xpmem_thread_group *__xpmem_tg_ref_by_tgid_nolock(
+	pid_t tgid,
+	int return_destroying)
+{
+	struct xpmem_thread_group *tg;
+
+	XPMEM_DEBUG("call: tgid=%d, return_destroying=%d", 
+		tgid, return_destroying);
+
+        tg = __xpmem_tg_ref_by_tgid_nolock_internal(tgid, 
+		xpmem_tg_hashtable_index(tgid), return_destroying);
+
+	XPMEM_DEBUG("return: tg=0x%p", tg);
+
+        return tg;
+}
+
+#define xpmem_tg_ref_by_tgid(t)             __xpmem_tg_ref_by_tgid(t, 0)
+#define xpmem_tg_ref_by_tgid_all(t)         __xpmem_tg_ref_by_tgid(t, 1)
+#define xpmem_tg_ref_by_tgid_nolock(t)      __xpmem_tg_ref_by_tgid_nolock(t, 0)
+#define xpmem_tg_ref_by_tgid_all_nolock(t)  __xpmem_tg_ref_by_tgid_nolock(t, 1)
+
+static struct xpmem_thread_group * xpmem_tg_ref_by_segid(xpmem_segid_t);
+static void xpmem_tg_deref(struct xpmem_thread_group *);
+static struct xpmem_segment *xpmem_seg_ref_by_segid(struct xpmem_thread_group *,
+	xpmem_segid_t);
+static void xpmem_seg_deref(struct xpmem_segment *);
+
+/*
+ * Inlines that mark an internal driver structure as being destroyable or not.
+ * The idea is to set the refcnt to 1 at structure creation time and then
+ * drop that reference at the time the structure is to be destroyed.
+ */
+static inline void xpmem_tg_not_destroyable(
+	struct xpmem_thread_group *tg)
+{
+	ihk_atomic_set(&tg->refcnt, 1);
+
+	XPMEM_DEBUG("return: tg->refcnt=%d", tg->refcnt);
+}
+
+static inline void xpmem_tg_destroyable(
+	struct xpmem_thread_group *tg)
+{
+	XPMEM_DEBUG("call: ");
+
+	xpmem_tg_deref(tg);
+
+	XPMEM_DEBUG("return: ");
+}
+
+static inline void xpmem_seg_not_destroyable(
+	struct xpmem_segment *seg)
+{
+	ihk_atomic_set(&seg->refcnt, 1);
+
+	XPMEM_DEBUG("return: seg->refcnt=%d", seg->refcnt);
+}
+
+static inline void xpmem_seg_destroyable(
+	struct xpmem_segment *seg)
+{
+	XPMEM_DEBUG("call: ");
+
+	xpmem_seg_deref(seg);
+
+	XPMEM_DEBUG("return: ");
+}
+
+/*
+ * Inlines that increment the refcnt for the specified structure.
+ */
+static inline void xpmem_tg_ref(
+	struct xpmem_thread_group *tg)
+{
+	DBUG_ON(ihk_atomic_read(&tg->refcnt) <= 0);
+	ihk_atomic_inc(&tg->refcnt);
+
+	XPMEM_DEBUG("return: tg->refcnt=%d", tg->refcnt);
+}
+
+static inline void xpmem_seg_ref(
+	struct xpmem_segment *seg)
+{
+	DBUG_ON(ihk_atomic_read(&seg->refcnt) <= 0);
+	ihk_atomic_inc(&seg->refcnt);
+
+	XPMEM_DEBUG("return: seg->refcnt=%d", seg->refcnt);
+}
+
+#endif /* _XPMEM_PRIVATE_H */
+

kernel/init.c

@@ -108,11 +108,11 @@ static void dma_test(void)
 }
 #endif
 
-extern char *ihk_mc_get_kernel_args(void);
+extern char *ihk_get_kargs(void);
 
 char *find_command_line(char *name)
 {
-	char *cmdline = ihk_mc_get_kernel_args();
+	char *cmdline = ihk_get_kargs();
 
 	if (!cmdline) {
 		return NULL;
@@ -122,7 +122,7 @@ char *find_command_line(char *name)
 
 static void parse_kargs(void)
 {
-	kprintf("KCommand Line: %s\n", ihk_mc_get_kernel_args());
+	kprintf("KCommand Line: %s\n", ihk_get_kargs());
 
 	if (1) {
 		char *key = "osnum=";
@@ -254,7 +254,7 @@ static void rest_init(void)
 	time_init();
 	kmalloc_init();
 
-	ikc_master_init();
+	ihk_ikc_master_init();
 
 	proc_init();
 
@@ -320,7 +320,7 @@ static void setup_remote_snooping_samples(void)
 static void populate_sysfs(void)
 {
 	cpu_sysfs_setup();
-	setup_remote_snooping_samples();
+	//setup_remote_snooping_samples();
 } /* populate_sysfs() */
 
 int host_ikc_inited = 0;
@@ -373,6 +373,7 @@ int main(void)
 
 	kputs("IHK/McKernel started.\n");
 
+	ihk_set_kmsg(virt_to_phys(&kmsg_buf), IHK_KMSG_SIZE);
 	arch_init();
 
 	/*

kernel/mikc.c

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

kernel/process.c

@@ -74,7 +74,6 @@ init_process(struct process *proc, struct process *parent)
 {
 	/* These will be filled out when changing status */
 	proc->pid = -1;
-	proc->exit_status = -1;
 	proc->status = PS_RUNNING;
 
 	if(parent){
@@ -204,25 +203,45 @@ detach_address_space(struct address_space *asp, int pid)
 static int
 init_process_vm(struct process *owner, struct address_space *asp, struct process_vm *vm)
 {
+	int i;
 	ihk_mc_spinlock_init(&vm->memory_range_lock);
 	ihk_mc_spinlock_init(&vm->page_table_lock);
 
 	ihk_atomic_set(&vm->refcount, 1);
 	INIT_LIST_HEAD(&vm->vm_range_list);
+	INIT_LIST_HEAD(&vm->vm_range_numa_policy_list);
 	vm->address_space = asp;
 	vm->proc = owner;
 	vm->exiting = 0;
 
+	memset(&vm->numa_mask, 0, sizeof(vm->numa_mask));
+	for (i = 0; i < ihk_mc_get_nr_numa_nodes(); ++i) {
+		if (i >= PROCESS_NUMA_MASK_BITS) {
+			kprintf("%s: error: NUMA id is larger than mask size!\n",
+				__FUNCTION__);
+			break;
+		}
+		set_bit(i, &vm->numa_mask[0]);
+	}
+	vm->numa_mem_policy = MPOL_DEFAULT;
+
+	for (i = 0; i < VM_RANGE_CACHE_SIZE; ++i) {
+		vm->range_cache[i] = NULL;
+	}
+	vm->range_cache_ind = 0;
+
 	return 0;
 }
 
-struct thread *
-create_thread(unsigned long user_pc)
+struct thread *create_thread(unsigned long user_pc,
+		unsigned long *__cpu_set, size_t cpu_set_size)
 {
 	struct thread *thread;
 	struct process *proc;
 	struct process_vm *vm = NULL;
 	struct address_space *asp = NULL;
+	int cpu;
+	int cpu_set_empty = 1;
 
 	thread = ihk_mc_alloc_pages(KERNEL_STACK_NR_PAGES, IHK_MC_AP_NOWAIT);
 	if (!thread)
@@ -238,7 +257,22 @@ create_thread(unsigned long user_pc)
 	memset(vm, 0, sizeof(struct process_vm));
 	init_process(proc, cpu_local_var(resource_set)->pid1);
 
-	if (1) {
+	/* Use requested CPU cores */
+	for_each_set_bit(cpu, __cpu_set, cpu_set_size * BITS_PER_BYTE) {
+		if (cpu >= num_processors) {
+			kprintf("%s: invalid CPU requested in initial cpu_set\n",
+				__FUNCTION__);
+			goto err;
+		}
+
+		dkprintf("%s: pid: %d, CPU: %d\n",
+			__FUNCTION__, proc->pid, cpu); 
+		CPU_SET(cpu, &thread->cpu_set);
+		cpu_set_empty = 0;
+	}
+
+	/* Default allows all cores */
+	if (cpu_set_empty) {
 		struct ihk_mc_cpu_info *infop;
 		int i;
 
@@ -260,10 +294,10 @@ create_thread(unsigned long user_pc)
 	dkprintf("fork(): sigshared\n");
 
 	ihk_atomic_set(&thread->sigcommon->use, 1);
-	ihk_mc_spinlock_init(&thread->sigcommon->lock);
+	mcs_rwlock_init(&thread->sigcommon->lock);
 	INIT_LIST_HEAD(&thread->sigcommon->sigpending);
 
-	ihk_mc_spinlock_init(&thread->sigpendinglock);
+	mcs_rwlock_init(&thread->sigpendinglock);
 	INIT_LIST_HEAD(&thread->sigpending);
 
 	thread->sigstack.ss_sp = NULL;
@@ -280,6 +314,7 @@ create_thread(unsigned long user_pc)
 	if(init_process_vm(proc, asp, vm) != 0){
 		goto err;
 	}
+	thread->exit_status = -1;
 
 	cpu_set(ihk_mc_get_processor_id(), &thread->vm->address_space->cpu_set,
 			&thread->vm->address_space->cpu_set_lock);
@@ -371,6 +406,11 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 			kfree(proc);
 			goto err_free_proc;
 		}
+		memcpy(&proc->vm->numa_mask, &org->vm->numa_mask,
+				sizeof(proc->vm->numa_mask));
+		proc->vm->numa_mem_policy =
+			org->vm->numa_mem_policy;
+
 		thread->proc = proc;
 		thread->vm = proc->vm;
 
@@ -418,11 +458,11 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 		memcpy(thread->sigcommon->action, org->sigcommon->action,
 		       sizeof(struct k_sigaction) * _NSIG);
 		ihk_atomic_set(&thread->sigcommon->use, 1);
-		ihk_mc_spinlock_init(&thread->sigcommon->lock);
+		mcs_rwlock_init(&thread->sigcommon->lock);
 		INIT_LIST_HEAD(&thread->sigcommon->sigpending);
 		// TODO: copy signalfd
 	}
-	ihk_mc_spinlock_init(&thread->sigpendinglock);
+	mcs_rwlock_init(&thread->sigpendinglock);
 	INIT_LIST_HEAD(&thread->sigpending);
 	thread->sigmask = org->sigmask;
 
@@ -716,6 +756,7 @@ int join_process_memory_range(struct process_vm *vm,
 		struct vm_range *surviving, struct vm_range *merging)
 {
 	int error;
+	int i;
 
 	dkprintf("join_process_memory_range(%p,%lx-%lx,%lx-%lx)\n",
 			vm, surviving->start, surviving->end,
@@ -744,6 +785,10 @@ int join_process_memory_range(struct process_vm *vm,
 		memobj_release(merging->memobj);
 	}
 	list_del(&merging->list);
+	for (i = 0; i < VM_RANGE_CACHE_SIZE; ++i) {
+		if (vm->range_cache[i] == merging)
+			vm->range_cache[i] = surviving;
+	}
 	kfree(merging);
 
 	error = 0;
@@ -757,7 +802,7 @@ int free_process_memory_range(struct process_vm *vm, struct vm_range *range)
 {
 	const intptr_t start0 = range->start;
 	const intptr_t end0 = range->end;
-	int error;
+	int error, i;
 	intptr_t start;
 	intptr_t end;
 	struct vm_range *neighbor;
@@ -842,6 +887,10 @@ int free_process_memory_range(struct process_vm *vm, struct vm_range *range)
 	}
 
 	list_del(&range->list);
+	for (i = 0; i < VM_RANGE_CACHE_SIZE; ++i) {
+		if (vm->range_cache[i] == range)
+			vm->range_cache[i] = NULL;
+	}
 	kfree(range);
 
 	dkprintf("free_process_memory_range(%p,%lx-%lx): 0\n",
@@ -1058,6 +1107,7 @@ int add_process_memory_range(struct process_vm *vm,
 struct vm_range *lookup_process_memory_range(
 		struct process_vm *vm, uintptr_t start, uintptr_t end)
 {
+	int i;
 	struct vm_range *range = NULL;
 
 	dkprintf("lookup_process_memory_range(%p,%lx,%lx)\n", vm, start, end);
@@ -1066,6 +1116,16 @@ struct vm_range *lookup_process_memory_range(
 		goto out;
 	}
 
+	for (i = 0; i < VM_RANGE_CACHE_SIZE; ++i) {
+		int c_i = (i + vm->range_cache_ind) % VM_RANGE_CACHE_SIZE;
+		if (!vm->range_cache[c_i])
+			continue;
+
+		if (vm->range_cache[c_i]->start <= start &&
+			vm->range_cache[c_i]->end >= end)
+			return vm->range_cache[c_i];
+	}
+
 	list_for_each_entry(range, &vm->vm_range_list, list) {
 		if (end <= range->start) {
 			break;
@@ -1077,6 +1137,12 @@ struct vm_range *lookup_process_memory_range(
 
 	range = NULL;
 out:
+	if (range) {
+		vm->range_cache_ind = (vm->range_cache_ind - 1 + VM_RANGE_CACHE_SIZE)
+			% VM_RANGE_CACHE_SIZE;
+		vm->range_cache[vm->range_cache_ind] = range;
+	}
+
 	dkprintf("lookup_process_memory_range(%p,%lx,%lx): %p %lx-%lx\n",
 			vm, start, end, range,
 			range? range->start: 0, range? range->end: 0);
@@ -1318,6 +1384,11 @@ static int sync_one_page(void *arg0, page_table_t pt, pte_t *ptep,
 	flush_tlb_single((uintptr_t)pgaddr);	/* XXX: TLB flush */
 
 	phys = pte_get_phys(ptep);
+	if (args->memobj->flags & MF_ZEROFILL) {
+		error = 0;
+		goto out;
+	}
+
 	error = memobj_flush_page(args->memobj, phys, pgsize);
 	if (error) {
 		ekprintf("sync_one_page(%p,%p,%p %#lx,%p,%d):"
@@ -1345,11 +1416,19 @@ int sync_process_memory_range(struct process_vm *vm, struct vm_range *range,
 	args.memobj = range->memobj;
 
 	ihk_mc_spinlock_lock_noirq(&vm->page_table_lock);
-	memobj_lock(range->memobj);
+
+	if (!(range->memobj->flags & MF_ZEROFILL)) {
+		memobj_lock(range->memobj);
+	}
+
 	error = visit_pte_range(vm->address_space->page_table, (void *)start,
-	                        (void *)end, range->pgshift, VPTEF_SKIP_NULL,
-				&sync_one_page, &args);
-	memobj_unlock(range->memobj);
+			(void *)end, range->pgshift, VPTEF_SKIP_NULL,
+			&sync_one_page, &args);
+
+	if (!(range->memobj->flags & MF_ZEROFILL)) {
+		memobj_unlock(range->memobj);
+	}
+
 	ihk_mc_spinlock_unlock_noirq(&vm->page_table_lock);
 	if (error) {
 		ekprintf("sync_process_memory_range(%p,%p,%#lx,%#lx):"
@@ -1517,7 +1596,7 @@ static int page_fault_process_memory_range(struct process_vm *vm, struct vm_rang
 			}
 		}
 		if (phys == NOPHYS) {
-			void *virt;
+			void *virt = NULL;
 			size_t npages;
 
 retry:
@@ -1542,7 +1621,9 @@ retry:
 					__FUNCTION__, pgaddr, pgsize);
 			memset(virt, 0, pgsize);
 			phys = virt_to_phys(virt);
-			page_map(phys_to_page(phys));
+			if (phys_to_page(phys)) {
+				page_map(phys_to_page(phys));
+			}
 		}
 	}
 	else {
@@ -1554,10 +1635,12 @@ retry:
 	attr = arch_vrflag_to_ptattr(range->flag | memobj_flag, reason, ptep);
 
 	/*****/
-	if (((range->flag & VR_PRIVATE)
-				|| ((reason & PF_PATCH)
-					&& !(range->flag & VR_PROT_WRITE)))
-			&& (!page || page_is_in_memobj(page) || page_is_multi_mapped(page))) {
+	if (((range->flag & VR_PRIVATE) ||
+				((reason & PF_PATCH) && !(range->flag & VR_PROT_WRITE)))
+			&& ((!page && phys == NOPHYS) || (page &&
+					(page_is_in_memobj(page) ||
+					 page_is_multi_mapped(page))))) {
+
 		if (!(attr & PTATTR_DIRTY)) {
 			attr &= ~PTATTR_WRITABLE;
 		}
@@ -1637,10 +1720,9 @@ static int do_page_fault_process_vm(struct process_vm *vm, void *fault_addr0, ui
 	range = lookup_process_memory_range(vm, fault_addr, fault_addr+1);
 	if (range == NULL) {
 		error = -EFAULT;
-		dkprintf("[%d]do_page_fault_process_vm(%p,%lx,%lx):"
-				"out of range. %d\n",
-				ihk_mc_get_processor_id(), vm,
-				fault_addr0, reason, error);
+		dkprintf("do_page_fault_process_vm(): vm: %p, addr: %p, reason: %lx):"
+				"out of range: %d\n",
+				vm, fault_addr0, reason, error);
 		goto out;
 	}
 
@@ -1670,10 +1752,18 @@ static int do_page_fault_process_vm(struct process_vm *vm, void *fault_addr0, ui
 			kprintf("if (((range->flag & VR_PROT_MASK) == VR_PROT_NONE))\n");
 		if (((reason & PF_WRITE) && !(reason & PF_PATCH)))
 			kprintf("if (((reason & PF_WRITE) && !(reason & PF_PATCH)))\n");
-		if (!(range->flag & VR_PROT_WRITE))
+		if (!(range->flag & VR_PROT_WRITE)) {
 			kprintf("if (!(range->flag & VR_PROT_WRITE))\n");
-		if ((reason & PF_INSTR) && !(range->flag & VR_PROT_EXEC))
+			//kprintf("setting VR_PROT_WRITE\n");
+			//range->flag |= VR_PROT_WRITE;
+			//goto cont;
+		}
+		if ((reason & PF_INSTR) && !(range->flag & VR_PROT_EXEC)) {
 			kprintf("if ((reason & PF_INSTR) && !(range->flag & VR_PROT_EXEC))\n");
+			//kprintf("setting VR_PROT_EXEC\n");
+			//range->flag |= VR_PROT_EXEC;
+			//goto cont;
+		}
 		goto out;
 	}
 
@@ -2131,9 +2221,10 @@ int populate_process_memory(struct process_vm *vm, void *start, size_t len)
 	for (addr = (uintptr_t)start; addr < end; addr += PAGE_SIZE) {
 		error = page_fault_process_vm(vm, (void *)addr, reason);
 		if (error) {
-			ekprintf("populate_process_range:page_fault_process_vm"
-					"(%p,%lx,%lx) failed %d\n",
-					vm, addr, reason, error);
+			ekprintf("%s: WARNING: page_fault_process_vm(): vm: %p, "
+					"addr: %lx, reason: %lx, off: %lu, len: %lu returns %d\n",
+					__FUNCTION__, vm, addr, reason,
+					((void *)addr - start), len, error);
 			goto out;
 		}
 	}
@@ -2462,6 +2553,7 @@ void sched_init(void)
 	ihk_mc_init_context(&idle_thread->ctx, NULL, idle);
 	ihk_mc_spinlock_init(&idle_thread->vm->memory_range_lock);
 	INIT_LIST_HEAD(&idle_thread->vm->vm_range_list);
+	INIT_LIST_HEAD(&idle_thread->vm->vm_range_numa_policy_list);
 	idle_thread->proc->pid = 0;
 	idle_thread->tid = ihk_mc_get_processor_id();
 
@@ -2551,7 +2643,6 @@ 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);
@@ -2664,7 +2755,9 @@ redo:
 			restore_fp_regs(next);
 		}
 
-		ihk_mc_load_page_table(next->vm->address_space->page_table);
+		if (prev && prev->vm->address_space->page_table !=
+				next->vm->address_space->page_table)
+			ihk_mc_load_page_table(next->vm->address_space->page_table);
 
 		dkprintf("[%d] schedule: tlsblock_base: 0x%lX\n",
 		         ihk_mc_get_processor_id(), next->tlsblock_base);
@@ -2896,6 +2989,7 @@ find_thread(int pid, int tid, struct mcs_rwlock_node_irqsave *lock)
 	if(tid <= 0)
 		return NULL;
 	mcs_rwlock_reader_lock(&thash->lock[hash], lock);
+retry:
 	list_for_each_entry(thread, &thash->list[hash], hash_list){
 		if(thread->tid == tid){
 			if(pid <= 0)
@@ -2904,6 +2998,13 @@ find_thread(int pid, int tid, struct mcs_rwlock_node_irqsave *lock)
 				return thread;
 		}
 	}
+	/* If no thread with pid == tid was found, then we may be looking for a
+	 * specific thread (not the main thread of the process), try to find it
+	 * based on tid only */
+	if (pid > 0 && pid == tid) {
+		pid = 0;
+		goto retry;
+	}
 	mcs_rwlock_reader_unlock(&thash->lock[hash], lock);
 	return NULL;
 }

kernel/procfs.c

@@ -23,6 +23,8 @@
 #include <process.h>
 #include <page.h>
 #include <mman.h>
+#include <bitmap.h>
+#include <init.h>
 
 //#define DEBUG_PRINT_PROCFS
 
@@ -35,6 +37,7 @@
 extern int snprintf(char * buf, size_t size, const char *fmt, ...);
 extern int sprintf(char * buf, const char *fmt, ...);
 extern int sscanf(const char * buf, const char * fmt, ...);
+extern int scnprintf(char * buf, size_t size, const char *fmt, ...);
 
 extern int osnum;
 
@@ -404,12 +407,34 @@ process_procfs_request(unsigned long rarg)
 	/* 
 	 * mcos%d/PID/status
 	 */
+#define BITMASKS_BUF_SIZE	2048
 	if (strcmp(p, "status") == 0) {
+		extern int num_processors;	/* kernel/ap.c */
 		struct vm_range *range;
 		unsigned long lockedsize = 0;
-		char tmp[1024];
+		char *tmp;
+		char *bitmasks;
+		int bitmasks_offset = 0;
+		char *cpu_bitmask, *cpu_list, *numa_bitmask, *numa_list;
 		int len;
 
+		tmp = kmalloc(8192, IHK_MC_AP_CRITICAL);
+		if (!tmp) {
+			kprintf("%s: error allocating /proc/self/status buffer\n",
+				__FUNCTION__);
+			ans = 0;
+			goto end;
+		}
+
+		bitmasks = kmalloc(BITMASKS_BUF_SIZE, IHK_MC_AP_CRITICAL);
+		if (!tmp) {
+			kprintf("%s: error allocating /proc/self/status bitmaks buffer\n",
+				__FUNCTION__);
+			kfree(tmp);
+			ans = 0;
+			goto end;
+		}
+
 		ihk_mc_spinlock_lock_noirq(&proc->vm->memory_range_lock);
 		list_for_each_entry(range, &proc->vm->vm_range_list, list) {
 			if(range->flag & VR_LOCKED)
@@ -417,13 +442,42 @@ process_procfs_request(unsigned long rarg)
 		}
 		ihk_mc_spinlock_unlock_noirq(&proc->vm->memory_range_lock);
 
+		cpu_bitmask = &bitmasks[bitmasks_offset];
+		bitmasks_offset += bitmap_scnprintf(cpu_bitmask,
+				BITMASKS_BUF_SIZE - bitmasks_offset,
+				thread->cpu_set.__bits, num_processors);
+		bitmasks_offset++;
+
+		cpu_list = &bitmasks[bitmasks_offset];
+		bitmasks_offset += bitmap_scnlistprintf(cpu_list,
+				BITMASKS_BUF_SIZE - bitmasks_offset,
+				thread->cpu_set.__bits, __CPU_SETSIZE);
+		bitmasks_offset++;
+
+		numa_bitmask = &bitmasks[bitmasks_offset];
+		bitmasks_offset += bitmap_scnprintf(numa_bitmask,
+				BITMASKS_BUF_SIZE - bitmasks_offset,
+				proc->vm->numa_mask, PROCESS_NUMA_MASK_BITS);
+		bitmasks_offset++;
+
+		numa_list = &bitmasks[bitmasks_offset];
+		bitmasks_offset += bitmap_scnlistprintf(numa_list,
+				BITMASKS_BUF_SIZE - bitmasks_offset,
+				proc->vm->numa_mask, PROCESS_NUMA_MASK_BITS);
+		bitmasks_offset++;
+
 		sprintf(tmp,
 		        "Uid:\t%d\t%d\t%d\t%d\n"
 		        "Gid:\t%d\t%d\t%d\t%d\n"
-		        "VmLck:\t%9lu kB\n",
+		        "VmLck:\t%9lu kB\n"
+				"Cpus_allowed:\t%s\n"
+				"Cpus_allowed_list:\t%s\n"
+				"Mems_allowed:\t%s\n"
+				"Mems_allowed_list:\t%s\n",
 		        proc->ruid, proc->euid, proc->suid, proc->fsuid,
 		        proc->rgid, proc->egid, proc->sgid, proc->fsgid,
-		        (lockedsize + 1023) >> 10);
+		        (lockedsize + 1023) >> 10,
+				cpu_bitmask, cpu_list, numa_bitmask, numa_list);
 		len = strlen(tmp);
 		if (r->offset < len) {
 			if (r->offset + r->count < len) {
@@ -437,6 +491,8 @@ process_procfs_request(unsigned long rarg)
 			ans = 0;
 			eof = 1;
 		}
+		kfree(tmp);
+		kfree(bitmasks);
 		goto end;
 	}
 

kernel/shmobj.c

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

kernel/xpmem.c

@@ -0,0 +1,739 @@
+/**
+ * \file xpmem.c
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Cross Partition Memory (XPMEM) support.
+ */
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (c) 2004-2007 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright 2010, 2014 Cray Inc. All Rights Reserved
+ * Copyright 2015-2016 Los Alamos National Security, LLC. All rights reserved.
+ */
+/*
+ * HISTORY
+ */
+
+#include <errno.h>
+#include <kmalloc.h>
+#include <limits.h>
+#include <memobj.h>
+#include <mman.h>
+#include <string.h>
+#include <types.h>
+#include <vsprintf.h>
+#include <ihk/lock.h>
+#include <ihk/mm.h>
+#include <xpmem_private.h>
+
+
+struct xpmem_partition *xpmem_my_part = NULL;  /* pointer to this partition */
+
+
+int xpmem_open(
+	ihk_mc_user_context_t *ctx)
+{
+	const char *pathname = (const char *)ihk_mc_syscall_arg0(ctx);
+	int flags = (int)ihk_mc_syscall_arg1(ctx);
+	int ret;
+	struct thread *thread = cpu_local_var(current);
+	struct process *proc = thread->proc;
+	struct syscall_request request IHK_DMA_ALIGN;
+	int fd;
+	struct mckfd *mckfd;
+	long irqstate;
+
+	XPMEM_DEBUG("call: pathname=%s, flags=%d", pathname, flags);
+
+	if (!xpmem_my_part) {
+		ret = xpmem_init();
+		if (ret) {
+			return ret;
+		}
+	}
+
+	request.number = __NR_open;
+	request.args[0] = (unsigned long)pathname;
+	request.args[1] = flags;
+	fd = do_syscall(&request, ihk_mc_get_processor_id(), 0);
+	if(fd < 0){
+		XPMEM_DEBUG("__NR_open error: fd=%d", fd);
+		return fd;
+	}
+
+	ret = __xpmem_open();
+	if (ret) {
+		XPMEM_DEBUG("return: ret=%d", ret);
+		return ret;
+	}
+
+	mckfd = kmalloc(sizeof(struct mckfd), IHK_MC_AP_NOWAIT);
+	if(!mckfd) {
+		return -ENOMEM;
+	}
+	XPMEM_DEBUG("kmalloc(): mckfd=0x%p", mckfd);
+	memset(mckfd, 0, sizeof(struct mckfd));
+	mckfd->fd = fd;
+	mckfd->sig_no = -1;
+	mckfd->ioctl_cb = xpmem_ioctl;
+	mckfd->close_cb = xpmem_close;
+	irqstate = ihk_mc_spinlock_lock(&proc->mckfd_lock);
+
+	if(proc->mckfd == NULL) {
+		proc->mckfd = mckfd;
+		mckfd->next = NULL;
+	} else {
+		mckfd->next = proc->mckfd;
+		proc->mckfd = mckfd;
+	}
+
+	ihk_mc_spinlock_unlock(&proc->mckfd_lock, irqstate);
+
+	ihk_atomic_inc_return(&xpmem_my_part->n_opened);
+
+	XPMEM_DEBUG("return: ret=%d", mckfd->fd);
+
+	return mckfd->fd;
+}
+
+
+static int xpmem_ioctl(
+	struct mckfd *mckfd,
+	ihk_mc_user_context_t *ctx)
+{
+	int ret;
+	unsigned int cmd = ihk_mc_syscall_arg1(ctx);
+	unsigned long arg = ihk_mc_syscall_arg2(ctx);
+
+	XPMEM_DEBUG("call: cmd=0x%x, arg=0x%lx", cmd, arg);
+
+	switch (cmd) {
+	case XPMEM_CMD_VERSION: {
+		ret = XPMEM_CURRENT_VERSION;
+
+		XPMEM_DEBUG("return: cmd=0x%x, ret=0x%lx", cmd, ret);
+
+		return ret;
+	}
+	case XPMEM_CMD_MAKE: {
+		struct xpmem_cmd_make make_info;
+		xpmem_segid_t segid = 0;
+
+		if (copy_from_user(&make_info, (void __user *)arg, 
+			sizeof(struct xpmem_cmd_make)))
+			return -EFAULT;
+
+		ret = xpmem_make(make_info.vaddr, make_info.size, 
+			make_info.permit_type, 
+			(void *)make_info.permit_value, &segid);
+		if (ret != 0) {
+			XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+			return ret;
+		}
+
+		if (copy_to_user(&((struct xpmem_cmd_make __user *)arg)->segid, 
+			(void *)&segid, sizeof(xpmem_segid_t))) {
+			(void)xpmem_remove(segid);
+			return -EFAULT;
+		}
+
+		XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+
+		return ret;
+	}
+	case XPMEM_CMD_REMOVE: {
+		struct xpmem_cmd_remove remove_info;
+
+		if (copy_from_user(&remove_info, (void __user *)arg, 
+			sizeof(struct xpmem_cmd_remove)))
+			return -EFAULT;
+
+		ret = xpmem_remove(remove_info.segid);
+
+		XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+
+		return ret;
+	}
+	case XPMEM_CMD_GET: {
+		struct xpmem_cmd_get get_info;
+//		xpmem_apid_t apid = 0;
+
+		if (copy_from_user(&get_info, (void __user *)arg, 
+			sizeof(struct xpmem_cmd_get)))
+			return -EFAULT;
+
+//		ret = xpmem_get(get_info.segid, get_info.flags,
+//			get_info.permit_type,
+//			(void *)get_info.permit_value, &apid); // TODO
+		ret = -EINVAL;
+		if (ret != 0) {
+			XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+			return ret;
+		}
+
+//		if (copy_to_user(&((struct xpmem_cmd_get __user *)arg)->apid, 
+//			(void *)&apid, sizeof(xpmem_apid_t))) {
+//			(void)xpmem_release(apid);
+//			return -EFAULT;
+//		}
+
+		XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+
+		return ret;
+	}
+	case XPMEM_CMD_RELEASE: {
+		struct xpmem_cmd_release release_info;
+
+		if (copy_from_user(&release_info, (void __user *)arg,
+			sizeof(struct xpmem_cmd_release)))
+			return -EFAULT;
+
+//		ret = xpmem_release(release_info.apid); // TODO
+		ret = -EINVAL;
+
+		XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+
+		return ret;
+	}
+	case XPMEM_CMD_ATTACH: {
+		struct xpmem_cmd_attach attach_info;
+//		unsigned long at_vaddr = 0;
+
+		if (copy_from_user(&attach_info, (void __user *)arg, 
+			sizeof(struct xpmem_cmd_attach)))
+			return -EFAULT;
+
+//		ret = xpmem_attach(mckfd, attach_info.apid, attach_info.offset, 
+//			attach_info.size, attach_info.vaddr, 
+//			attach_info.fd, attach_info.flags, 
+//			&at_vaddr); // TODO
+		ret = -EINVAL;
+		if (ret != 0) {
+			XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+			return ret;
+		}
+
+//		if (copy_to_user(
+//			&((struct xpmem_cmd_attach __user *)arg)->vaddr, 
+//			(void *)&at_vaddr, sizeof(unsigned long))) {
+//			(void)xpmem_detach(at_vaddr);
+//			return -EFAULT;
+//		}
+
+		XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+
+		return ret;
+	}
+	case XPMEM_CMD_DETACH: {
+		struct xpmem_cmd_detach detach_info;
+
+		if (copy_from_user(&detach_info, (void __user *)arg, 
+			sizeof(struct xpmem_cmd_detach)))
+			return -EFAULT;
+
+//		ret = xpmem_detach(detach_info.vaddr); // TODO
+		ret = -EINVAL;
+
+		XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, ret);
+
+		return ret;
+	}
+	default:
+		break;
+	}
+
+	XPMEM_DEBUG("return: cmd=0x%x, ret=%d", cmd, -EINVAL);
+
+	return -EINVAL;
+}
+
+
+static int xpmem_close(
+	struct mckfd *mckfd,
+	ihk_mc_user_context_t *ctx)
+{
+	struct xpmem_thread_group *tg;
+	int index;
+	struct mcs_rwlock_node_irqsave lock;
+	int n_opened;
+
+	XPMEM_DEBUG("call: fd=%d", mckfd->fd);
+
+	n_opened = ihk_atomic_dec_return(&xpmem_my_part->n_opened);
+	if (n_opened) {
+		XPMEM_DEBUG("return: ret=%d, n_opened=%d", 0, n_opened);
+		return 0;
+	}
+	XPMEM_DEBUG("n_opened=%d", n_opened);
+
+	index = xpmem_tg_hashtable_index(cpu_local_var(current)->proc->pid);
+
+	mcs_rwlock_writer_lock(&xpmem_my_part->tg_hashtable[index].lock, &lock);
+
+	tg = xpmem_tg_ref_by_tgid_all_nolock(
+		cpu_local_var(current)->proc->pid);
+	if (!tg) {
+		mcs_rwlock_writer_unlock(
+			&xpmem_my_part->tg_hashtable[index].lock, &lock);
+		return 0;
+	}
+
+	list_del_init(&tg->tg_hashlist);
+
+	mcs_rwlock_writer_unlock(&xpmem_my_part->tg_hashtable[index].lock, 
+		&lock);
+
+	XPMEM_DEBUG("tg->vm=0x%p", tg->vm);
+
+	xpmem_destroy_tg(tg);
+
+	if (!n_opened) {
+		xpmem_exit();
+	}
+
+	XPMEM_DEBUG("return: ret=%d", 0);
+
+	return 0;
+}
+
+
+static int xpmem_init(void)
+{
+	int i;
+
+	XPMEM_DEBUG("call: ");
+
+	xpmem_my_part = kmalloc(sizeof(struct xpmem_partition) + 
+		sizeof(struct xpmem_hashlist) * XPMEM_TG_HASHTABLE_SIZE, 
+		IHK_MC_AP_NOWAIT);
+	if (xpmem_my_part == NULL) {
+		return -ENOMEM;
+	}
+	XPMEM_DEBUG("kmalloc(): xpmem_my_part=0x%p", xpmem_my_part);
+	memset(xpmem_my_part, 0, sizeof(struct xpmem_partition) + 
+		sizeof(struct xpmem_hashlist) * XPMEM_TG_HASHTABLE_SIZE);
+
+	for (i = 0; i < XPMEM_TG_HASHTABLE_SIZE; i++) {
+		mcs_rwlock_init(&xpmem_my_part->tg_hashtable[i].lock);
+		INIT_LIST_HEAD(&xpmem_my_part->tg_hashtable[i].list);
+	}
+
+	ihk_atomic_set(&xpmem_my_part->n_opened, 0);
+
+	XPMEM_DEBUG("return: ret=%d", 0);
+
+	return 0;
+}
+
+
+static void xpmem_exit(void)
+{
+	XPMEM_DEBUG("call: ");
+
+	if (xpmem_my_part) {
+		XPMEM_DEBUG("kfree(): 0x%p", xpmem_my_part);
+		kfree(xpmem_my_part);
+		xpmem_my_part = NULL;
+	}
+
+	XPMEM_DEBUG("return: ");
+}
+
+
+static int __xpmem_open(void)
+{
+	struct xpmem_thread_group *tg;
+	int index;
+	struct mcs_rwlock_node_irqsave lock;
+
+	XPMEM_DEBUG("call: ");
+
+	tg = xpmem_tg_ref_by_tgid(cpu_local_var(current)->proc->pid);
+	if (!IS_ERR(tg)) {
+		xpmem_tg_deref(tg);
+		XPMEM_DEBUG("return: ret=%d, tg=0x%p", 0, tg);
+		return 0;
+	}
+
+	tg = kmalloc(sizeof(struct xpmem_thread_group) + 
+		sizeof(struct xpmem_hashlist) * XPMEM_AP_HASHTABLE_SIZE, 
+		IHK_MC_AP_NOWAIT);
+	if (tg == NULL) {
+		return -ENOMEM;
+	}
+	XPMEM_DEBUG("kmalloc(): tg=0x%p", tg);
+	memset(tg, 0, sizeof(struct xpmem_thread_group) + 
+		sizeof(struct xpmem_hashlist) * XPMEM_AP_HASHTABLE_SIZE);
+
+	ihk_mc_spinlock_init(&tg->lock);
+	tg->tgid = cpu_local_var(current)->proc->pid;
+	tg->uid = cpu_local_var(current)->proc->ruid;
+	tg->gid = cpu_local_var(current)->proc->rgid;
+	ihk_atomic_set(&tg->uniq_segid, 0);
+	ihk_atomic_set(&tg->uniq_apid, 0);
+	mcs_rwlock_init(&tg->seg_list_lock);
+	INIT_LIST_HEAD(&tg->seg_list);
+	ihk_atomic_set(&tg->n_pinned, 0);
+	INIT_LIST_HEAD(&tg->tg_hashlist);
+	tg->vm = cpu_local_var(current)->vm;
+	ihk_atomic_set(&tg->n_recall_PFNs, 0);
+
+	for (index = 0; index < XPMEM_AP_HASHTABLE_SIZE; index++) {
+		mcs_rwlock_init(&tg->ap_hashtable[index].lock);
+		INIT_LIST_HEAD(&tg->ap_hashtable[index].list);
+	}
+
+	xpmem_tg_not_destroyable(tg);
+
+	index = xpmem_tg_hashtable_index(tg->tgid);
+	mcs_rwlock_writer_lock(&xpmem_my_part->tg_hashtable[index].lock, &lock);
+
+	list_add_tail(&tg->tg_hashlist, 
+		&xpmem_my_part->tg_hashtable[index].list);
+
+	mcs_rwlock_writer_unlock(&xpmem_my_part->tg_hashtable[index].lock, 
+		&lock);
+
+	tg->group_leader = cpu_local_var(current);
+
+	XPMEM_DEBUG("return: ret=%d", 0);
+
+	return 0;
+}
+
+
+static void xpmem_destroy_tg(
+	struct xpmem_thread_group *tg)
+{
+	XPMEM_DEBUG("call: tg=0x%p", tg);
+
+	XPMEM_DEBUG("tg->vm=0x%p", tg->vm);
+
+	xpmem_tg_destroyable(tg);
+	xpmem_tg_deref(tg);
+
+	XPMEM_DEBUG("return: ");
+}
+
+
+static int xpmem_make(
+	unsigned long vaddr,
+	size_t size,
+	int permit_type,
+	void *permit_value,
+	xpmem_segid_t *segid_p)
+{
+	xpmem_segid_t segid;
+	struct xpmem_thread_group *seg_tg;
+	struct xpmem_segment *seg;
+	struct mcs_rwlock_node_irqsave lock;
+
+	XPMEM_DEBUG("call: vaddr=0x%lx, size=%lu, permit_type=%d, " 
+		"permit_value=0%04lo", 
+		vaddr, size, permit_type, 
+		(unsigned long)(uintptr_t)permit_value);
+
+	if (permit_type != XPMEM_PERMIT_MODE ||
+		((unsigned long)(uintptr_t)permit_value & ~00777) || 
+		size == 0) {
+		XPMEM_DEBUG("return: ret=%d", -EINVAL);
+		return -EINVAL;
+	}
+
+	seg_tg = xpmem_tg_ref_by_tgid(cpu_local_var(current)->proc->pid);
+	if (IS_ERR(seg_tg)) {
+		DBUG_ON(PTR_ERR(seg_tg) != -ENOENT);
+		return -XPMEM_ERRNO_NOPROC;
+	}
+
+	/*
+	 * The start of the segment must be page aligned and it must be a
+	 * multiple of pages in size.
+	 */
+	if (offset_in_page(vaddr) != 0 || offset_in_page(size) != 0) {
+		xpmem_tg_deref(seg_tg);
+		XPMEM_DEBUG("return: ret=%d", -EINVAL);
+		return -EINVAL;
+	}
+
+	segid = xpmem_make_segid(seg_tg);
+	if (segid < 0) {
+		xpmem_tg_deref(seg_tg);
+		return segid;
+	}
+
+	/* create a new struct xpmem_segment structure with a unique segid */
+	seg = kmalloc(sizeof(struct xpmem_segment), IHK_MC_AP_NOWAIT);
+	if (seg == NULL) {
+		xpmem_tg_deref(seg_tg);
+		return -ENOMEM;
+	}
+	XPMEM_DEBUG("kmalloc(): seg=0x%p", seg);
+	memset(seg, 0, sizeof(struct xpmem_segment));
+
+	ihk_mc_spinlock_init(&seg->lock);
+	mcs_rwlock_init(&seg->seg_lock);
+	seg->segid = segid;
+	seg->vaddr = vaddr;
+	seg->size = size;
+	seg->permit_type = permit_type;
+	seg->permit_value = permit_value;
+	seg->tg = seg_tg;
+	INIT_LIST_HEAD(&seg->ap_list);
+	INIT_LIST_HEAD(&seg->seg_list);
+
+	xpmem_seg_not_destroyable(seg);
+
+	/* add seg to its tg's list of segs */
+	mcs_rwlock_writer_lock(&seg_tg->seg_list_lock, &lock);
+	list_add_tail(&seg->seg_list, &seg_tg->seg_list);
+	mcs_rwlock_writer_unlock(&seg_tg->seg_list_lock, &lock);
+
+	xpmem_tg_deref(seg_tg);
+
+	*segid_p = segid;
+
+	XPMEM_DEBUG("return: ret=%d, segid=0x%lx", 0, *segid_p);
+
+	return 0;
+}
+
+
+static xpmem_segid_t xpmem_make_segid(
+	struct xpmem_thread_group *seg_tg)
+{
+	struct xpmem_id segid;
+	xpmem_segid_t *segid_p = (xpmem_segid_t *)&segid;
+	int uniq;
+
+	XPMEM_DEBUG("call: seg_tg=0x%p, uniq_segid=%d", 
+		seg_tg, ihk_atomic_read(&seg_tg->uniq_segid));
+
+	DBUG_ON(sizeof(struct xpmem_id) != sizeof(xpmem_segid_t));
+
+	uniq = ihk_atomic_inc_return(&seg_tg->uniq_segid);
+	if (uniq > XPMEM_MAX_UNIQ_ID) {
+		ihk_atomic_dec(&seg_tg->uniq_segid);
+		return -EBUSY;
+	}
+
+	*segid_p = 0;
+	segid.tgid = seg_tg->tgid;
+	segid.uniq = (unsigned long)uniq;
+
+	DBUG_ON(*segid_p <= 0);
+
+	XPMEM_DEBUG("return: segid=0x%lx, segid.tgid=%d, segid.uniq=%d", 
+		segid, segid.tgid, segid.uniq);
+
+	return *segid_p;
+}
+
+
+static int xpmem_remove(
+	xpmem_segid_t segid)
+{
+	struct xpmem_thread_group *seg_tg;
+	struct xpmem_segment *seg;
+
+	XPMEM_DEBUG("call: segid=0x%lx", segid);
+
+	if (segid <= 0) {
+		XPMEM_DEBUG("return: ret=%d", -EINVAL);
+		return -EINVAL;
+	}
+
+	seg_tg = xpmem_tg_ref_by_segid(segid);
+	if (IS_ERR(seg_tg))
+		return PTR_ERR(seg_tg);
+
+	if (cpu_local_var(current)->proc->pid != seg_tg->tgid) {
+		xpmem_tg_deref(seg_tg);
+		XPMEM_DEBUG("return: ret=%d", -EACCES);
+		return -EACCES;
+	}
+
+	seg = xpmem_seg_ref_by_segid(seg_tg, segid);
+	if (IS_ERR(seg)) {
+		xpmem_tg_deref(seg_tg);
+		return PTR_ERR(seg);
+	}
+	DBUG_ON(seg->tg != seg_tg);
+
+	xpmem_remove_seg(seg_tg, seg);
+	xpmem_seg_deref(seg);
+	xpmem_tg_deref(seg_tg);
+
+	XPMEM_DEBUG("return: ret=%d", 0);
+
+	return 0;
+}
+
+
+static void xpmem_remove_seg(
+	struct xpmem_thread_group *seg_tg,
+	struct xpmem_segment *seg)
+{
+	DBUG_ON(ihk_atomic_read(&seg->refcnt) <= 0);
+	struct mcs_rwlock_node_irqsave seg_lock;
+	struct mcs_rwlock_node_irqsave lock;
+
+	XPMEM_DEBUG("call: tgid=%d, segid=0x%lx", seg_tg->tgid, seg->segid);
+
+	ihk_mc_spinlock_lock(&seg->lock);
+	if (seg->flags & XPMEM_FLAG_DESTROYING) {
+		ihk_mc_spinlock_unlock_noirq(&seg->lock);
+		schedule();
+		return;
+	}
+	seg->flags |= XPMEM_FLAG_DESTROYING;
+	ihk_mc_spinlock_unlock_noirq(&seg->lock);
+
+	mcs_rwlock_writer_lock(&seg->seg_lock, &seg_lock);
+
+	/* unpin pages and clear PTEs for each attachment to this segment */
+	xpmem_clear_PTEs(seg);
+
+	/* indicate that the segment has been destroyed */
+	ihk_mc_spinlock_lock(&seg->lock);
+	seg->flags |= XPMEM_FLAG_DESTROYED;
+	ihk_mc_spinlock_unlock_noirq(&seg->lock);
+
+	/* Remove segment structure from its tg's list of segs */
+	mcs_rwlock_writer_lock(&seg_tg->seg_list_lock, &lock);
+	list_del_init(&seg->seg_list);
+	mcs_rwlock_writer_unlock(&seg_tg->seg_list_lock, &lock);
+
+	mcs_rwlock_writer_unlock(&seg->seg_lock, &seg_lock);
+
+	xpmem_seg_destroyable(seg);
+
+	XPMEM_DEBUG("return: ");
+}
+
+
+static void xpmem_clear_PTEs(
+	struct xpmem_segment *seg)
+{
+	XPMEM_DEBUG("call: seg=0x%p", seg);
+
+//	xpmem_clear_PTEs_range(seg, seg->vaddr, seg->vaddr + seg->size, 0); // TODO
+
+	XPMEM_DEBUG("return: ");
+}
+
+
+static struct xpmem_thread_group * __xpmem_tg_ref_by_tgid_nolock_internal(
+	pid_t tgid,
+	int index,
+	int return_destroying)
+{
+	struct xpmem_thread_group *tg;
+
+	XPMEM_DEBUG("call: tgid=%d, index=%d, return_destroying=%d", 
+		tgid, index, return_destroying);
+
+	list_for_each_entry(tg, &xpmem_my_part->tg_hashtable[index].list,
+		tg_hashlist) {
+		if (tg->tgid == tgid) {
+			if ((tg->flags & XPMEM_FLAG_DESTROYING) &&
+				!return_destroying) {
+				continue;
+			}
+
+			xpmem_tg_ref(tg);
+
+			XPMEM_DEBUG("return: tg=0x%p", tg);
+			return tg;
+		}
+	}
+
+	XPMEM_DEBUG("return: tg=0x%p", ERR_PTR(-ENOENT));
+
+	return ERR_PTR(-ENOENT);
+}
+
+
+static struct xpmem_thread_group * xpmem_tg_ref_by_segid(
+	xpmem_segid_t segid)
+{
+	struct xpmem_thread_group *tg;
+
+	XPMEM_DEBUG("call: segid=0x%lx", segid);
+
+	tg = xpmem_tg_ref_by_tgid(xpmem_segid_to_tgid(segid));
+
+	XPMEM_DEBUG("return: tg=0x%p", tg);
+
+        return tg;
+}
+
+
+static void xpmem_tg_deref(
+	struct xpmem_thread_group *tg)
+{
+	XPMEM_DEBUG("call: tg=0x%p", tg);
+
+	DBUG_ON(ihk_atomic_read(&tg->refcnt) <= 0);
+	if (ihk_atomic_dec_return(&tg->refcnt) != 0) {
+		XPMEM_DEBUG("return: tg->refcnt=%d", tg->refcnt);
+		return;
+	}
+
+	XPMEM_DEBUG("kfree(): tg=0x%p", tg);
+	kfree(tg);
+
+	XPMEM_DEBUG("return: ");
+}
+
+
+static struct xpmem_segment * xpmem_seg_ref_by_segid(
+	struct xpmem_thread_group *seg_tg,
+	xpmem_segid_t segid)
+{
+	struct xpmem_segment *seg;
+	struct mcs_rwlock_node_irqsave lock;
+
+	XPMEM_DEBUG("call: seg_tg=0x%p, segid=0x%lx", seg_tg, segid);
+
+	mcs_rwlock_reader_lock(&seg_tg->seg_list_lock, &lock);
+
+	list_for_each_entry(seg, &seg_tg->seg_list, seg_list) {
+		if (seg->segid == segid) {
+			if (seg->flags & XPMEM_FLAG_DESTROYING)
+				continue;
+
+			xpmem_seg_ref(seg);
+			mcs_rwlock_reader_unlock(&seg_tg->seg_list_lock, &lock);
+			return seg;
+		}
+	}
+
+	mcs_rwlock_reader_unlock(&seg_tg->seg_list_lock, &lock);
+
+	return ERR_PTR(-ENOENT);
+}
+
+
+static void xpmem_seg_deref(
+	struct xpmem_segment *seg)
+{
+	XPMEM_DEBUG("call: seg=0x%p", seg);
+
+	DBUG_ON(ihk_atomic_read(&seg->refcnt) <= 0);
+	if (ihk_atomic_dec_return(&seg->refcnt) != 0) {
+		XPMEM_DEBUG("return: seg->refcnt=%d", seg->refcnt);
+		return;
+	}
+
+	DBUG_ON(!(seg->flags & XPMEM_FLAG_DESTROYING));
+
+	XPMEM_DEBUG("kfree(): seg=0x%p", seg);
+	kfree(seg);
+
+	XPMEM_DEBUG("return: ");
+}
+

kernel/zeroobj.c

@@ -102,6 +102,7 @@ static int alloc_zeroobj(void)
 
 	memset(obj, 0, sizeof(*obj));
 	obj->memobj.ops = &zeroobj_ops;
+	obj->memobj.size = 0;
 	page_list_init(obj);
 	ihk_mc_spinlock_init(&obj->memobj.lock);
 
@@ -112,7 +113,7 @@ static int alloc_zeroobj(void)
 		goto out;
 	}
 	phys = virt_to_phys(virt);
-	page = phys_to_page(phys);
+	page = phys_to_page_insert_hash(phys);
 
 	if (page->mode != PM_NONE) {
 		fkprintf("alloc_zeroobj():"

lib/bitops.c

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

lib/include/bitmap.h

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

lib/include/bitops.h

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

lib/include/ihk/cpu.h

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

lib/include/ihk/mm.h

@@ -72,8 +72,11 @@ struct ihk_mc_memory_node {
 
 unsigned long ihk_mc_get_memory_address(enum ihk_mc_gma_type, int);
 
-void ihk_mc_reserve_arch_pages(unsigned long start, unsigned long end,
-                               void (*cb)(unsigned long, unsigned long, int));
+struct ihk_page_allocator_desc;
+void ihk_mc_reserve_arch_pages(struct ihk_page_allocator_desc *pa_allocator,
+		unsigned long start, unsigned long end,
+		void (*cb)(struct ihk_page_allocator_desc *, 
+			unsigned long, unsigned long, int));
 
 struct ihk_mc_pa_ops {
 	void *(*alloc_page)(int, int, enum ihk_mc_ap_flag);
@@ -100,14 +103,28 @@ void ihk_mc_map_micpa(unsigned long host_pa, unsigned long* mic_pa);
 int ihk_mc_free_micpa(unsigned long mic_pa);
 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_alloc_aligned_pages(int npages, int p2align,
+	enum ihk_mc_ap_flag flag, char *file, int line);
+#define ihk_mc_alloc_aligned_pages(npages, p2align, flag) ({\
+void *r = _ihk_mc_alloc_aligned_pages(npages, p2align, flag, __FILE__, __LINE__);\
+r;\
+})
+
+void *_ihk_mc_alloc_pages(int npages, enum ihk_mc_ap_flag flag,
+		char *file, int line);
+#define ihk_mc_alloc_pages(npages, flag) ({\
+void *r = _ihk_mc_alloc_pages(npages, flag, __FILE__, __LINE__);\
+r;\
+})
+
+void _ihk_mc_free_pages(void *ptr, int npages, char *file, int line);
+#define ihk_mc_free_pages(p, npages) ({\
+_ihk_mc_free_pages(p, npages, __FILE__, __LINE__);\
+})
+
 void *ihk_mc_allocate(int size, int flag);
 void ihk_mc_free(void *p);
 
-void *arch_alloc_page(enum ihk_mc_ap_flag flag);
-void arch_free_page(void *ptr);
 int arch_get_smaller_page_size(void *args, size_t origsize, size_t *sizep, int *p2alignp);
 
 typedef void *page_table_t;
@@ -147,13 +164,30 @@ struct page_table *ihk_mc_pt_create(enum ihk_mc_ap_flag ap_flag);
 /* XXX: proper use of struct page_table and page_table_t is unknown */
 void ihk_mc_pt_destroy(struct page_table *pt);
 void ihk_mc_load_page_table(struct page_table *pt);
+int ihk_mc_pt_virt_to_phys_size(struct page_table *pt,
+                           const void *virt,
+						   unsigned long *phys,
+						   unsigned long *size);
 int ihk_mc_pt_virt_to_phys(struct page_table *pt,
                            const void *virt, unsigned long *phys);
 uint64_t ihk_mc_pt_virt_to_pagemap(struct page_table *pt, unsigned long virt);
 
+int ihk_mc_get_nr_numa_nodes(void);
+struct smp_coreset;
+int ihk_mc_get_numa_node(int id, int *linux_numa_id, int *type);
+int ihk_mc_get_numa_distance(int i, int j);
+int ihk_mc_get_nr_memory_chunks(void);
+int ihk_mc_get_memory_chunk(int id,
+	unsigned long *start,
+	unsigned long *end,
+	int *numa_id);
+
 void remote_flush_tlb_cpumask(struct process_vm *vm, 
 		unsigned long addr, int cpu_id);
 
+int ihk_set_kmsg(unsigned long addr, unsigned long size);
+char *ihk_get_kargs();
+
 extern void (*__tlb_flush_handler)(int vector);
 
 struct tlb_flush_entry {

lib/include/ihk/page_alloc.h

@@ -14,14 +14,30 @@
 #ifndef __HEADER_GENERIC_IHK_PAGE_ALLOC
 #define __HEADER_GENERIC_IHK_PAGE_ALLOC
 
+#include <list.h>
+
+/* XXX: Physical memory management shouldn't be part of IHK */
+struct node_distance {
+	int id;
+	int distance;
+};
+
+struct ihk_mc_numa_node {
+	int id;
+	int linux_numa_id;
+	int type;
+	struct list_head allocators;
+	struct node_distance *nodes_by_distance;
+};
+
 struct ihk_page_allocator_desc {
-	unsigned long start;
+	unsigned long start, end;
 	unsigned int last;
 	unsigned int count;
 	unsigned int flag;
 	unsigned int shift;
-	ihk_spinlock_t lock;
-	unsigned int pad;
+	mcs_lock_node_t lock;
+	struct list_head list;
 	
 	unsigned long map[0];
 };

lib/include/mc_xpmem.h

@@ -0,0 +1,153 @@
+/**
+ * \file mc_xpmem.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Cross Partition Memory (XPMEM) structures and macros.
+ */
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (c) 2004-2007 Silicon Graphics, Inc.  All Rights Reserved.
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef _MC_XPMEM_H
+#define _MC_XPMEM_H
+
+#ifndef __KERNEL__
+#include <sys/types.h>
+#endif
+
+/*
+ * _IOC definitions for McKernel
+ */
+#define _IOC_NRBITS	8
+#define _IOC_TYPEBITS	8
+
+#define _IOC_SIZEBITS	14
+
+#define _IOC_DIRBITS	2
+
+#define _IOC_NRSHIFT	0
+#define _IOC_TYPESHIFT	(_IOC_NRSHIFT+_IOC_NRBITS)
+#define _IOC_SIZESHIFT	(_IOC_TYPESHIFT+_IOC_TYPEBITS)
+#define _IOC_DIRSHIFT	(_IOC_SIZESHIFT+_IOC_SIZEBITS)
+
+#define _IOC_NONE	0U
+
+#define _IOC(dir,type,nr,size) \
+	(((dir)  << _IOC_DIRSHIFT) | \
+	 ((type) << _IOC_TYPESHIFT) | \
+	 ((nr)   << _IOC_NRSHIFT) | \
+	 ((size) << _IOC_SIZESHIFT))
+
+#define _IO(type,nr) _IOC(_IOC_NONE,(type),(nr),0)
+
+/*
+ * basic argument type definitions for McKernel
+ */
+typedef uint64_t u64;
+typedef uint64_t __u64;
+typedef int64_t __s64;
+
+/*
+ * basic argument type definitions
+ */
+typedef __s64 xpmem_segid_t;	/* segid returned from xpmem_make() */
+typedef __s64 xpmem_apid_t;	/* apid returned from xpmem_get() */
+
+struct xpmem_addr {
+	xpmem_apid_t apid;	/* apid that represents memory */
+	off_t offset;		/* offset into apid's memory */
+};
+
+#define XPMEM_MAXADDR_SIZE	(size_t)(-1L)
+
+/*
+ * path to XPMEM device
+ */
+#define XPMEM_DEV_PATH  "/dev/xpmem"
+
+/*
+ * The following are the possible XPMEM related errors.
+ */
+#define XPMEM_ERRNO_NOPROC	2004	/* unknown thread due to fork() */
+
+/*
+ * flags for segment permissions
+ */
+#define XPMEM_RDONLY	0x1
+#define XPMEM_RDWR	0x2
+
+/*
+ * Valid permit_type values for xpmem_make().
+ */
+#define XPMEM_PERMIT_MODE	0x1
+
+/*
+ * ioctl() commands used to interface to the kernel module.
+ */
+#define XPMEM_IOC_MAGIC		'x'
+#define XPMEM_CMD_VERSION	_IO(XPMEM_IOC_MAGIC, 0)
+#define XPMEM_CMD_MAKE		_IO(XPMEM_IOC_MAGIC, 1)
+#define XPMEM_CMD_REMOVE	_IO(XPMEM_IOC_MAGIC, 2)
+#define XPMEM_CMD_GET		_IO(XPMEM_IOC_MAGIC, 3)
+#define XPMEM_CMD_RELEASE	_IO(XPMEM_IOC_MAGIC, 4)
+#define XPMEM_CMD_ATTACH	_IO(XPMEM_IOC_MAGIC, 5)
+#define XPMEM_CMD_DETACH	_IO(XPMEM_IOC_MAGIC, 6)
+
+/*
+ * Structures used with the preceding ioctl() commands to pass data.
+ */
+struct xpmem_cmd_make {
+	__u64 vaddr;
+	size_t size;
+	int permit_type;
+	__u64 permit_value;
+	xpmem_segid_t segid;	/* returned on success */
+};
+
+struct xpmem_cmd_remove {
+	xpmem_segid_t segid;
+};
+
+struct xpmem_cmd_get {
+	xpmem_segid_t segid;
+	int flags;
+	int permit_type;
+	__u64 permit_value;
+	xpmem_apid_t apid;	/* returned on success */
+};
+
+struct xpmem_cmd_release {
+	xpmem_apid_t apid;
+};
+
+struct xpmem_cmd_attach {
+	xpmem_apid_t apid;
+	off_t offset;
+	size_t size;
+	__u64 vaddr;
+	int fd;
+	int flags;
+};
+
+struct xpmem_cmd_detach {
+	__u64 vaddr;
+};
+
+#ifndef __KERNEL__
+extern int xpmem_version(void);
+extern xpmem_segid_t xpmem_make(void *, size_t, int, void *);
+extern int xpmem_remove(xpmem_segid_t);
+extern xpmem_apid_t xpmem_get(xpmem_segid_t, int, int, void *);
+extern int xpmem_release(xpmem_apid_t);
+extern void *xpmem_attach(struct xpmem_addr, size_t, void *);
+extern int xpmem_detach(void *);
+#endif
+
+#endif /* _MC_XPMEM_H */

lib/include/string.h

@@ -14,6 +14,7 @@
 #define __STRING_H
 
 #include <types.h>
+#include <arch-string.h>
 
 size_t strlen(const char *p);
 size_t strnlen(const char *p, size_t maxlen);
@@ -29,6 +30,17 @@ void *memcpy_long(void *dest, const void *src, size_t n);
 int memcmp(const void *s1, const void *s2, size_t n);
 void *memset(void *s, int n, size_t l);
 
+#ifdef ARCH_FAST_MEMCPY
+#define fast_memcpy	__inline_memcpy
+#else
+#define fast_memcpy memcpy
+#endif
+
+extern int snprintf(char * buf, size_t size, const char *fmt, ...);
+extern int sprintf(char * buf, const char *fmt, ...);
+extern int sscanf(const char * buf, const char * fmt, ...);
+extern int scnprintf(char * buf, size_t size, const char *fmt, ...);
+
 unsigned long strtol(const char *cp, char **endp, unsigned int base);
 int flatten_strings(int nr_strings, char *first, char **strings, char **flat);
 int flatten_strings_from_user(int nr_strings, char *first, char **strings, char **flat);

lib/page_alloc.c

@@ -52,7 +52,7 @@ void *__ihk_pagealloc_init(unsigned long start, unsigned long size,
 		desc = initial;
 		*pdescsize = descsize;
 	} else {
-		desc = (void *)allocate_pages(descsize, IHK_MC_AP_CRITICAL);
+		desc = (void *)ihk_mc_alloc_pages(descsize, IHK_MC_AP_CRITICAL);
 	}
 	if (!desc) {
 		kprintf("IHK: failed to allocate page-allocator-desc "\
@@ -64,15 +64,16 @@ void *__ihk_pagealloc_init(unsigned long start, unsigned long size,
 	memset(desc, 0, descsize * PAGE_SIZE);
 
 	desc->start = start;
+	desc->end = start + size;
 	desc->last = 0;
 	desc->count = mapaligned >> 3;
 	desc->shift = page_shift;
 	desc->flag = flag;
 
-	kprintf("Page allocator: %lx - %lx (%d)\n", start, start + size,
-	        page_shift);
+	//kprintf("page allocator @ %lx - %lx (%d)\n", start, start + size,
+	//        page_shift);
 
-	ihk_mc_spinlock_init(&desc->lock);
+	mcs_lock_init(&desc->lock);
 
 	/* Reserve align padding area */
 	for (i = mapsize; i < mapaligned * 8; i++) {
@@ -92,18 +93,18 @@ void ihk_pagealloc_destroy(void *__desc)
 {
 	struct ihk_page_allocator_desc *desc = __desc;
 
-	free_pages(desc, desc->flag);
+	ihk_mc_free_pages(desc, desc->flag);
 }
 
 static unsigned long __ihk_pagealloc_large(struct ihk_page_allocator_desc *desc,
                                            int npages, int p2align)
 {
-	unsigned long flags;
 	unsigned int i, j, mi;
 	int nblocks;
 	int nfrags;
 	unsigned long mask;
-	int mialign;
+	unsigned long align_mask = ((PAGE_SIZE << p2align) - 1);
+	mcs_lock_node_t node;
 
 	nblocks = (npages / 64);
 	mask = -1;
@@ -112,14 +113,13 @@ static unsigned long __ihk_pagealloc_large(struct ihk_page_allocator_desc *desc,
 		++nblocks;
 		mask = (1UL << nfrags) - 1;
 	}
-	mialign = (p2align <= 6)? 1: (1 << (p2align - 6));
 
-	flags = ihk_mc_spinlock_lock(&desc->lock);
+	mcs_lock_lock(&desc->lock, &node);
 	for (i = 0, mi = desc->last; i < desc->count; i++, mi++) {
 		if (mi >= desc->count) {
 			mi = 0;
 		}
-		if ((mi + nblocks >= desc->count) || (mi % mialign)) {
+		if ((mi + nblocks >= desc->count) || (ADDRESS(desc, mi, 0) & align_mask)) {
 			continue;
 		}
 		for (j = mi; j < mi + nblocks - 1; j++) {
@@ -132,11 +132,11 @@ static unsigned long __ihk_pagealloc_large(struct ihk_page_allocator_desc *desc,
 				desc->map[j] = (unsigned long)-1;
 			}
 			desc->map[j] |= mask;
-			ihk_mc_spinlock_unlock(&desc->lock, flags);
+			mcs_lock_unlock(&desc->lock, &node);
 			return ADDRESS(desc, mi, 0);
 		}
 	}
-	ihk_mc_spinlock_unlock(&desc->lock, flags);
+	mcs_lock_unlock(&desc->lock, &node);
 
 	return 0;
 }
@@ -146,8 +146,9 @@ unsigned long ihk_pagealloc_alloc(void *__desc, int npages, int p2align)
 	struct ihk_page_allocator_desc *desc = __desc;
 	unsigned int i, mi;
 	int j;
-	unsigned long v, mask, flags;
+	unsigned long v, mask;
 	int jalign;
+	mcs_lock_node_t node;
 
 	if ((npages >= 32) || (p2align >= 5)) {
 		return __ihk_pagealloc_large(desc, npages, p2align);
@@ -156,7 +157,7 @@ unsigned long ihk_pagealloc_alloc(void *__desc, int npages, int p2align)
 	mask = (1UL << npages) - 1;
 	jalign = (p2align <= 0)? 1: (1 << p2align);
 
-	flags = ihk_mc_spinlock_lock(&desc->lock);
+	mcs_lock_lock(&desc->lock, &node);
 	for (i = 0, mi = desc->last; i < desc->count; i++, mi++) {
 		if (mi >= desc->count) {
 			mi = 0;
@@ -173,12 +174,12 @@ unsigned long ihk_pagealloc_alloc(void *__desc, int npages, int p2align)
 			if (!(v & (mask << j))) { /* free */
 				desc->map[mi] |= (mask << j);
 
-				ihk_mc_spinlock_unlock(&desc->lock, flags);
+				mcs_lock_unlock(&desc->lock, &node);
 				return ADDRESS(desc, mi, j);
 			}
 		}
 	}
-	ihk_mc_spinlock_unlock(&desc->lock, flags);
+	mcs_lock_unlock(&desc->lock, &node);
 
 	/* We use null pointer for failure */
 	return 0;
@@ -188,7 +189,7 @@ void ihk_pagealloc_reserve(void *__desc, unsigned long start, unsigned long end)
 {
 	int i, n;
 	struct ihk_page_allocator_desc *desc = __desc;
-	unsigned long flags;
+	mcs_lock_node_t node;
 
 	n = (end + (1 << desc->shift) - 1 - desc->start) >> desc->shift;
 	i = ((start - desc->start) >> desc->shift);
@@ -196,7 +197,7 @@ void ihk_pagealloc_reserve(void *__desc, unsigned long start, unsigned long end)
 		return;
 	}
 
-	flags = ihk_mc_spinlock_lock(&desc->lock);
+	mcs_lock_lock(&desc->lock, &node);
 	for (; i < n; i++) {
 		if (!(i & 63) && i + 63 < n) {
 			desc->map[MAP_INDEX(i)] = (unsigned long)-1L;
@@ -205,7 +206,7 @@ void ihk_pagealloc_reserve(void *__desc, unsigned long start, unsigned long end)
 			desc->map[MAP_INDEX(i)] |= (1UL << MAP_BIT(i));
 		}
 	}
-	ihk_mc_spinlock_unlock(&desc->lock, flags);
+	mcs_lock_unlock(&desc->lock, &node);
 }
 
 void ihk_pagealloc_free(void *__desc, unsigned long address, int npages)
@@ -213,24 +214,24 @@ void ihk_pagealloc_free(void *__desc, unsigned long address, int npages)
 	struct ihk_page_allocator_desc *desc = __desc;
 	int i;
 	unsigned mi;
-	unsigned long flags;
+	mcs_lock_node_t node;
 
 	/* XXX: Parameter check */
-	flags = ihk_mc_spinlock_lock(&desc->lock);
+	mcs_lock_lock(&desc->lock, &node);
 	mi = (address - desc->start) >> desc->shift;
 	for (i = 0; i < npages; i++, mi++) {
 		desc->map[MAP_INDEX(mi)] &= ~(1UL << MAP_BIT(mi));
 	}
-	ihk_mc_spinlock_unlock(&desc->lock, flags);
+	mcs_lock_unlock(&desc->lock, &node);
 }
 
 unsigned long ihk_pagealloc_count(void *__desc)
 {
 	struct ihk_page_allocator_desc *desc = __desc;
 	unsigned long i, j, n = 0;
-	unsigned long flags;
+	mcs_lock_node_t node;
 
-	flags = ihk_mc_spinlock_lock(&desc->lock);
+	mcs_lock_lock(&desc->lock, &node);
 	/* XXX: Very silly counting */
 	for (i = 0; i < desc->count; i++) {
 		for (j = 0; j < 64; j++) {
@@ -239,7 +240,7 @@ unsigned long ihk_pagealloc_count(void *__desc)
 			}
 		}
 	}
-	ihk_mc_spinlock_unlock(&desc->lock, flags);
+	mcs_lock_unlock(&desc->lock, &node);
 	
 	return n;
 }
@@ -249,10 +250,11 @@ int ihk_pagealloc_query_free(void *__desc)
 	struct ihk_page_allocator_desc *desc = __desc;
 	unsigned int mi;
 	int j;
-	unsigned long v, flags;
+	unsigned long v;
 	int npages = 0;
+	mcs_lock_node_t node;
 
-	flags = ihk_mc_spinlock_lock(&desc->lock);
+	mcs_lock_lock(&desc->lock, &node);
 	for (mi = 0; mi < desc->count; mi++) {
 		
 		v = desc->map[mi];
@@ -265,7 +267,7 @@ int ihk_pagealloc_query_free(void *__desc)
 			}
 		}
 	}
-	ihk_mc_spinlock_unlock(&desc->lock, flags);
+	mcs_lock_unlock(&desc->lock, &node);
 
 	return npages;
 }
@@ -275,11 +277,12 @@ void __ihk_pagealloc_zero_free_pages(void *__desc)
 	struct ihk_page_allocator_desc *desc = __desc;
 	unsigned int mi;
 	int j;
-	unsigned long v, flags;
+	unsigned long v;
+	mcs_lock_node_t node;
 
 kprintf("zeroing free memory... ");
 
-	flags = ihk_mc_spinlock_lock(&desc->lock);
+	mcs_lock_lock(&desc->lock, &node);
 	for (mi = 0; mi < desc->count; mi++) {
 		
 		v = desc->map[mi];
@@ -293,7 +296,7 @@ kprintf("zeroing free memory... ");
 			}
 		}
 	}
-	ihk_mc_spinlock_unlock(&desc->lock, flags);
+	mcs_lock_unlock(&desc->lock, &node);
 
 kprintf("\nzeroing done\n");
 }