mcreboot/mcstop: proper error handling (revert previous state)

Merge free physical pages to create large, physically contiguous
blocks with the following command.

    echo 1 > /proc/sys/vm/compact_memory

Makefile.in

@@ -1,9 +1,11 @@
 TARGET = @TARGET@
 SBINDIR = @SBINDIR@
+ETCDIR = @ETCDIR@
 MANDIR = @MANDIR@
 
 all::
-	@(cd executer/kernel; make modules)
+	@(cd executer/kernel/mcctrl; make modules)
+	@(cd executer/kernel/mcoverlayfs; make modules)
 	@(cd executer/user; make)
 	@case "$(TARGET)" in \
 	    attached-mic | builtin-x86 | builtin-mic | smp-x86) \
@@ -16,7 +18,8 @@ all::
 	esac
 
 install::
-	@(cd executer/kernel; make install)
+	@(cd executer/kernel/mcctrl; make install)
+	@(cd executer/kernel/mcoverlayfs; make install)
 	@(cd executer/user; make install)
 	@case "$(TARGET)" in \
 	    attached-mic | builtin-x86 | builtin-mic | smp-x86) \
@@ -44,7 +47,11 @@ install::
 		;; \
 	    smp-x86) \
 		mkdir -p -m 755 $(SBINDIR); \
-		install -m 755 arch/x86/tools/mcreboot-smp-x86.sh $(SBINDIR)/mcreboot; \
+		install -m 755 arch/x86/tools/mcreboot-smp-x86.sh $(SBINDIR)/mcreboot.sh; \
+		install -m 755 arch/x86/tools/mcstop+release-smp-x86.sh $(SBINDIR)/mcstop+release.sh; \
+		mkdir -p -m 755 $(ETCDIR); \
+		install -m 644 arch/x86/tools/irqbalance_mck.service $(ETCDIR)/irqbalance_mck.service; \
+		install -m 644 arch/x86/tools/irqbalance_mck.in $(ETCDIR)/irqbalance_mck.in; \
 		mkdir -p -m 755 $(MANDIR)/man1; \
 		install -m 644 arch/x86/tools/mcreboot.1 $(MANDIR)/man1/mcreboot.1; \
 		;; \
@@ -55,7 +62,8 @@ install::
 	esac
 
 clean::
-	@(cd executer/kernel; make clean)
+	@(cd executer/kernel/mcctrl; make clean)
+	@(cd executer/kernel/mcoverlayfs; make clean)
 	@(cd executer/user; make clean)
 	@case "$(TARGET)" in \
 	    attached-mic | builtin-x86 | builtin-mic | smp-x86) \

arch/x86/kernel/context.S

@@ -10,7 +10,7 @@
  * HISTORY
  */
 
-#define X86_CPU_LOCAL_OFFSET_TSS    128
+#define X86_CPU_LOCAL_OFFSET_TSS    176
 #define X86_TSS_OFFSET_SP0          4
 #define X86_CPU_LOCAL_OFFSET_SP0 \
 		(X86_CPU_LOCAL_OFFSET_TSS + X86_TSS_OFFSET_SP0)

arch/x86/kernel/gencore.c

@@ -78,11 +78,11 @@ int get_prstatus_size(void)
  * \brief Fill a prstatus structure.
  *
  * \param head A pointer to a note structure.
- * \param proc A pointer to the current process structure.
+ * \param thread A pointer to the current thread structure.
  * \param regs0 A pointer to a x86_regs structure.
  */
 
-void fill_prstatus(struct note *head, struct process *proc, void *regs0)
+void fill_prstatus(struct note *head, struct thread *thread, void *regs0)
 {
 	void *name;
 	struct elf_prstatus64 *prstatus; 
@@ -160,11 +160,11 @@ int get_prpsinfo_size(void)
  * \brief Fill a prpsinfo structure.
  *
  * \param head A pointer to a note structure.
- * \param proc A pointer to the current process structure.
+ * \param thread A pointer to the current thread structure.
  * \param regs A pointer to a x86_regs structure.
  */
 
-void fill_prpsinfo(struct note *head, struct process *proc, void *regs)
+void fill_prpsinfo(struct note *head, struct thread *thread, void *regs)
 {
 	void *name;
 	struct elf_prpsinfo64 *prpsinfo;
@@ -176,8 +176,8 @@ void fill_prpsinfo(struct note *head, struct process *proc, void *regs)
 	memcpy(name, "CORE", sizeof("CORE"));
 	prpsinfo = (struct elf_prpsinfo64 *)(name + align32(sizeof("CORE")));
 
-	prpsinfo->pr_state = proc->ftn->status;
-	prpsinfo->pr_pid = proc->ftn->pid;
+	prpsinfo->pr_state = thread->status;
+	prpsinfo->pr_pid = thread->proc->pid;
 
 /*
   We leave most of the fields unfilled.
@@ -210,11 +210,11 @@ int get_auxv_size(void)
  * \brief Fill an AUXV structure.
  *
  * \param head A pointer to a note structure.
- * \param proc A pointer to the current process structure.
+ * \param thread A pointer to the current thread structure.
  * \param regs A pointer to a x86_regs structure.
  */
 
-void fill_auxv(struct note *head, struct process *proc, void *regs)
+void fill_auxv(struct note *head, struct thread *thread, void *regs)
 {
 	void *name;
 	void *auxv;
@@ -225,7 +225,7 @@ void fill_auxv(struct note *head, struct process *proc, void *regs)
 	name =  (void *) (head + 1);
 	memcpy(name, "CORE", sizeof("CORE"));
 	auxv = name + align32(sizeof("CORE"));
-	memcpy(auxv, proc->saved_auxv, sizeof(unsigned long) * AUXV_LEN);
+	memcpy(auxv, thread->proc->saved_auxv, sizeof(unsigned long) * AUXV_LEN);
 } 
 
 /**
@@ -243,23 +243,23 @@ int get_note_size(void)
  * \brief Fill the NOTE segment.
  *
  * \param head A pointer to a note structure.
- * \param proc A pointer to the current process structure.
+ * \param thread A pointer to the current thread structure.
  * \param regs A pointer to a x86_regs structure.
  */
 
-void fill_note(void *note, struct process *proc, void *regs)
+void fill_note(void *note, struct thread *thread, void *regs)
 {
-	fill_prstatus(note, proc, regs);
+	fill_prstatus(note, thread, regs);
 	note += get_prstatus_size();
-	fill_prpsinfo(note, proc, regs);
+	fill_prpsinfo(note, thread, regs);
 	note += get_prpsinfo_size();
-	fill_auxv(note, proc, regs);
+	fill_auxv(note, thread, regs);
 }
 
 /**
  * \brief Generate an image of the core file.
  *
- * \param proc A pointer to the current process structure.
+ * \param thread A pointer to the current thread structure.
  * \param regs A pointer to a x86_regs structure.
  * \param coretable(out) An array of core chunks.
  * \param chunks(out) Number of the entires of coretable.
@@ -271,7 +271,18 @@ void fill_note(void *note, struct process *proc, void *regs)
  * should be zero.
  */
 
-int gencore(struct process *proc, void *regs, 
+/*@
+  @ requires \valid(thread);
+  @ requires \valid(regs);
+  @ requires \valid(coretable);
+  @ requires \valid(chunks);
+  @ behavior success:
+  @   ensures \result == 0;
+  @   assigns coretable;
+  @ behavior failure:
+  @   ensures \result == -1;
+  @*/
+int gencore(struct thread *thread, void *regs, 
 	    struct coretable **coretable, int *chunks)
 {
 	struct coretable *ct = NULL;
@@ -279,7 +290,7 @@ int gencore(struct process *proc, void *regs,
 	Elf64_Phdr *ph = NULL;
 	void *note = NULL;
 	struct vm_range *range;
-	struct process_vm *vm = proc->vm;
+	struct process_vm *vm = thread->vm;
 	int segs = 1;	/* the first one is for NOTE */
 	int notesize, phsize, alignednotesize;
 	unsigned int offset = 0;
@@ -306,7 +317,7 @@ int gencore(struct process *proc, void *regs,
 			unsigned long p, phys;
 			int prevzero = 0;
 			for (p = range->start; p < range->end; p += PAGE_SIZE) {
-				if (ihk_mc_pt_virt_to_phys(proc->vm->page_table, 
+				if (ihk_mc_pt_virt_to_phys(thread->vm->address_space->page_table, 
 							    (void *)p, &phys) != 0) {
 					prevzero = 1;
 				} else {
@@ -326,7 +337,7 @@ int gencore(struct process *proc, void *regs,
 	dkprintf("we have %d segs and %d chunks.\n\n", segs, *chunks);
 
 	{
-		struct vm_regions region = proc->vm->region;
+		struct vm_regions region = thread->vm->region;
 
 		dkprintf("text:  %lx-%lx\n", region.text_start, region.text_end);
 		dkprintf("data:  %lx-%lx\n", region.data_start, region.data_end);
@@ -364,7 +375,7 @@ int gencore(struct process *proc, void *regs,
 		goto fail;
 	}
 	memset(note, 0, alignednotesize);
-	fill_note(note, proc, regs);
+	fill_note(note, thread, regs);
 
 	/* prgram header for NOTE segment is exceptional */
 	ph[0].p_type = PT_NOTE;
@@ -434,7 +445,7 @@ int gencore(struct process *proc, void *regs,
 
 			for (start = p = range->start; 
 			     p < range->end; p += PAGE_SIZE) {
-				if (ihk_mc_pt_virt_to_phys(proc->vm->page_table, 
+				if (ihk_mc_pt_virt_to_phys(thread->vm->address_space->page_table, 
 							    (void *)p, &phys) != 0) {
 					if (prevzero == 0) {
 						/* We begin a new chunk */
@@ -472,9 +483,9 @@ int gencore(struct process *proc, void *regs,
 				i++;
 			}		
 		} else {
-			if ((proc->vm->region.user_start <= range->start) &&
-			    (range->end <= proc->vm->region.user_end)) {
-				if (ihk_mc_pt_virt_to_phys(proc->vm->page_table, 
+			if ((thread->vm->region.user_start <= range->start) &&
+			    (range->end <= thread->vm->region.user_end)) {
+				if (ihk_mc_pt_virt_to_phys(thread->vm->address_space->page_table, 
 							   (void *)range->start, &phys) != 0) {
 					dkprintf("could not convert user virtual address %lx"
 						 "to physical address", range->start);
@@ -510,6 +521,10 @@ int gencore(struct process *proc, void *regs,
  * \param coretable An array of core chunks.
  */
 
+/*@
+  @ requires \valid(coretable);
+  @ assigns \nothing;
+  @*/
 void freecore(struct coretable **coretable)
 {
 	struct coretable *ct = *coretable;

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

@@ -0,0 +1,98 @@
+/**
+ * \file arch-bitops.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Find last set bit in word.
+ * \author Taku Shimosawa  <shimosawa@is.s.u-tokyo.ac.jp> \par
+ *      Copyright (C) 2011 - 2012  Taku Shimosawa
+ */
+/*
+ * HISTORY
+ */
+
+#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;
+	asm("bsrl %1,%0\n\t"
+	    "jnz 1f\n\t"
+	    "movl $-1,%0\n"
+	    "1:" : "=r" (r) : "rm" (x));
+
+	return r + 1;
+}
+
+/**
+ * ffs - find first set bit in word
+ * @x: the word to search
+ *
+ * This is defined the same way as the libc and compiler builtin ffs
+ * routines, therefore differs in spirit from the other bitops.
+ *
+ * ffs(value) returns 0 if value is 0 or the position of the first
+ * set bit if value is nonzero. The first (least significant) bit
+ * is at position 1.
+ */
+static inline int ffs(int x)
+{
+	int r;
+	asm("bsfl %1,%0\n\t"
+	    "jnz 1f\n\t"
+	    "movl $-1,%0\n"
+	    "1:" : "=r" (r) : "rm" (x));
+	return r + 1;
+}
+
+
+/**
+ * __ffs - find first set bit in word
+ * @word: The word to search
+ *
+ * Undefined if no bit exists, so code should check against 0 first.
+ */
+static inline unsigned long __ffs(unsigned long word)
+{
+	asm("bsf %1,%0"
+		: "=r" (word)
+		: "rm" (word));
+	return word;
+}
+
+/**
+ * ffz - find first zero bit in word
+ * @word: The word to search
+ *
+ * Undefined if no zero exists, so code should check against ~0UL first.
+ */
+static inline unsigned long ffz(unsigned long word)
+{
+	asm("bsf %1,%0"
+		: "=r" (word)
+		: "r" (~word));
+	return word;
+}
+
+
+#define ADDR (*(volatile long *)addr)
+
+static inline void set_bit(int nr, volatile unsigned long *addr)
+{
+	asm volatile("lock; btsl %1,%0"
+		     : "+m" (ADDR)
+		     : "Ir" (nr)
+		     : "memory");
+}
+
+static inline void clear_bit(int nr, volatile unsigned long *addr)
+{
+	asm volatile("lock; btrl %1,%0"
+		     : "+m" (ADDR)
+		     : "Ir" (nr)
+		     : "memory");
+}
+
+#endif

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

@@ -0,0 +1,67 @@
+/**
+ * \file futex.h
+ * Licence details are found in the file LICENSE.
+ *  
+ * \brief
+ * Futex adaptation to McKernel
+ *
+ * \author Balazs Gerofi  <bgerofi@riken.jp> \par
+ * Copyright (C) 2012  RIKEN AICS
+ *
+ *
+ * HISTORY:
+ *
+ */
+#ifndef _ARCH_FUTEX_H
+#define _ARCH_FUTEX_H
+#include <asm.h>
+
+#define __futex_atomic_op1(insn, ret, oldval, uaddr, oparg)	\
+	asm volatile("1:\t" insn "\n"				\
+		     "2:\t.section .fixup,\"ax\"\n"		\
+		     "3:\tmov\t%3, %1\n"			\
+		     "\tjmp\t2b\n"				\
+		     "\t.previous\n"				\
+		     _ASM_EXTABLE(1b, 3b)			\
+		     : "=r" (oldval), "=r" (ret), "+m" (*uaddr)	\
+		     : "i" (-EFAULT), "0" (oparg), "1" (0))
+
+#define __futex_atomic_op2(insn, ret, oldval, uaddr, oparg)	\
+	asm volatile("1:\tmovl	%2, %0\n"			\
+		     "\tmovl\t%0, %3\n"				\
+		     "\t" insn "\n"				\
+		     "2:\tlock; cmpxchgl %3, %2\n"		\
+		     "\tjnz\t1b\n"				\
+		     "3:\t.section .fixup,\"ax\"\n"		\
+		     "4:\tmov\t%5, %1\n"			\
+		     "\tjmp\t3b\n"				\
+		     "\t.previous\n"				\
+		     _ASM_EXTABLE(1b, 4b)			\
+		     _ASM_EXTABLE(2b, 4b)			\
+		     : "=&a" (oldval), "=&r" (ret),		\
+		       "+m" (*uaddr), "=&r" (tem)		\
+		     : "r" (oparg), "i" (-EFAULT), "1" (0))
+
+static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
+						int newval)
+{
+#ifdef __UACCESS__
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+		return -EFAULT;
+#endif
+
+	asm volatile("1:\tlock; cmpxchgl %3, %1\n"
+		     "2:\t.section .fixup, \"ax\"\n"
+		     "3:\tmov     %2, %0\n"
+		     "\tjmp     2b\n"
+		     "\t.previous\n"
+		     _ASM_EXTABLE(1b, 3b)
+		     : "=a" (oldval), "+m" (*uaddr)
+		     : "i" (-EFAULT), "r" (newval), "0" (oldval)
+		     : "memory"
+	);
+
+	return oldval;
+}
+
+#endif

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

@@ -5,15 +5,20 @@
 #define __HEADER_X86_COMMON_ARCH_LOCK
 
 #include <ihk/cpu.h>
+#include <ihk/atomic.h>
 
 //#define DEBUG_SPINLOCK
+//#define DEBUG_MCS_RWLOCK
 
-#ifdef DEBUG_SPINLOCK
+#if defined(DEBUG_SPINLOCK) || defined(DEBUG_MCS_RWLOCK)
 int __kprintf(const char *format, ...);
 #endif
 
 typedef int ihk_spinlock_t;
 
+extern void preempt_enable(void);
+extern void preempt_disable(void);
+
 #define IHK_STATIC_SPINLOCK_FUNCS
 
 static void ihk_mc_spinlock_init(ihk_spinlock_t *lock)
@@ -22,7 +27,17 @@ static void ihk_mc_spinlock_init(ihk_spinlock_t *lock)
 }
 #define SPIN_LOCK_UNLOCKED 0
 
-static void ihk_mc_spinlock_lock_noirq(ihk_spinlock_t *lock)
+#ifdef DEBUG_SPINLOCK
+#define ihk_mc_spinlock_lock_noirq(l) { \
+__kprintf("[%d] call ihk_mc_spinlock_lock_noirq %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__ihk_mc_spinlock_lock_noirq(l); \
+__kprintf("[%d] ret ihk_mc_spinlock_lock_noirq\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define ihk_mc_spinlock_lock_noirq __ihk_mc_spinlock_lock_noirq
+#endif
+
+static void __ihk_mc_spinlock_lock_noirq(ihk_spinlock_t *lock)
 {
 	int inc = 0x00010000;
 	int tmp;
@@ -41,10 +56,8 @@ static void ihk_mc_spinlock_lock_noirq(ihk_spinlock_t *lock)
 	             : "+Q" (inc), "+m" (*lock), "=r" (tmp) : : "memory", "cc");
 #endif
 
-#ifdef DEBUG_SPINLOCK
-	__kprintf("[%d] trying to grab lock: 0x%lX\n", 
-	          ihk_mc_get_processor_id(), lock);
-#endif
+	preempt_disable();
+
 	asm volatile("lock; xaddl %0, %1\n"
 			"movzwl %w0, %2\n\t"
 			"shrl $16, %0\n\t"
@@ -60,36 +73,455 @@ static void ihk_mc_spinlock_lock_noirq(ihk_spinlock_t *lock)
 			:
 			: "memory", "cc");
 
-#ifdef DEBUG_SPINLOCK
-	__kprintf("[%d] holding lock: 0x%lX\n", ihk_mc_get_processor_id(), lock);
-#endif
 }
 
-static unsigned long ihk_mc_spinlock_lock(ihk_spinlock_t *lock)
+#ifdef DEBUG_SPINLOCK
+#define ihk_mc_spinlock_lock(l) ({ unsigned long rc;\
+__kprintf("[%d] call ihk_mc_spinlock_lock %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+rc = __ihk_mc_spinlock_lock(l);\
+__kprintf("[%d] ret ihk_mc_spinlock_lock\n", ihk_mc_get_processor_id()); rc;\
+})
+#else
+#define ihk_mc_spinlock_lock __ihk_mc_spinlock_lock
+#endif
+static unsigned long __ihk_mc_spinlock_lock(ihk_spinlock_t *lock)
 {
 	unsigned long flags;
 	
 	flags = cpu_disable_interrupt_save();
 
-	ihk_mc_spinlock_lock_noirq(lock);
+	__ihk_mc_spinlock_lock_noirq(lock);
 
 	return flags;
 }
 
-static void ihk_mc_spinlock_unlock_noirq(ihk_spinlock_t *lock)
+#ifdef DEBUG_SPINLOCK
+#define ihk_mc_spinlock_unlock_noirq(l) { \
+__kprintf("[%d] call ihk_mc_spinlock_unlock_noirq %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__ihk_mc_spinlock_unlock_noirq(l); \
+__kprintf("[%d] ret ihk_mc_spinlock_unlock_noirq\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define ihk_mc_spinlock_unlock_noirq __ihk_mc_spinlock_unlock_noirq
+#endif
+static void __ihk_mc_spinlock_unlock_noirq(ihk_spinlock_t *lock)
 {
 	asm volatile ("lock incw %0" : "+m"(*lock) : : "memory", "cc");
+	
+	preempt_enable();
 }
 
-static void ihk_mc_spinlock_unlock(ihk_spinlock_t *lock, unsigned long flags)
+#ifdef DEBUG_SPINLOCK
+#define ihk_mc_spinlock_unlock(l, f) { \
+__kprintf("[%d] call ihk_mc_spinlock_unlock %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__ihk_mc_spinlock_unlock((l), (f)); \
+__kprintf("[%d] ret ihk_mc_spinlock_unlock\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define ihk_mc_spinlock_unlock __ihk_mc_spinlock_unlock
+#endif
+static void __ihk_mc_spinlock_unlock(ihk_spinlock_t *lock, unsigned long flags)
 {
-	ihk_mc_spinlock_unlock_noirq(lock);
+	__ihk_mc_spinlock_unlock_noirq(lock);
 
 	cpu_restore_interrupt(flags);
-#ifdef DEBUG_SPINLOCK
-	__kprintf("[%d] released lock: 0x%lX\n", ihk_mc_get_processor_id(), lock);
+}
+
+/* An implementation of the Mellor-Crummey Scott (MCS) lock */
+typedef struct mcs_lock_node {
+	unsigned long locked;
+	struct mcs_lock_node *next;
+} __attribute__((aligned(64))) mcs_lock_node_t;
+
+static void mcs_lock_init(struct mcs_lock_node *node)
+{
+	node->locked = 0;
+	node->next = NULL;
+}
+
+static void mcs_lock_lock(struct mcs_lock_node *lock,
+		struct mcs_lock_node *node)
+{
+	struct mcs_lock_node *pred;
+
+	node->next = NULL;
+	node->locked = 0;
+	pred = (struct mcs_lock_node *)xchg8((unsigned long *)&lock->next,
+			(unsigned long)node);
+
+	if (pred) {
+		node->locked = 1;
+		pred->next = node;
+		while (node->locked != 0) {
+			cpu_pause();
+		}
+	}
+}
+
+static void mcs_lock_unlock(struct mcs_lock_node *lock,
+		struct mcs_lock_node *node)
+{
+	if (node->next == NULL) {
+		struct mcs_lock_node *old = (struct mcs_lock_node *)
+			atomic_cmpxchg8((unsigned long *)&lock->next,
+					(unsigned long)node, (unsigned long)0);
+
+		if (old == node) {
+			return;
+		}
+
+		while (node->next == NULL) {
+			cpu_pause();
+		}
+	}
+
+	node->next->locked = 0;
+}
+
+// reader/writer lock
+typedef struct mcs_rwlock_node {
+	ihk_atomic_t count;	// num of readers (use only common reader)
+	char type;		// lock type
+#define MCS_RWLOCK_TYPE_COMMON_READER 0
+#define MCS_RWLOCK_TYPE_READER 1
+#define MCS_RWLOCK_TYPE_WRITER 2
+	char locked;		// lock
+#define MCS_RWLOCK_LOCKED	1
+#define MCS_RWLOCK_UNLOCKED	0
+	char dmy1;		// unused
+	char dmy2;		// unused
+	struct mcs_rwlock_node *next;
+} __attribute__((aligned(64))) mcs_rwlock_node_t;
+
+typedef struct mcs_rwlock_node_irqsave {
+	struct mcs_rwlock_node node;
+	unsigned long irqsave;
+} __attribute__((aligned(64))) mcs_rwlock_node_irqsave_t;
+
+typedef struct mcs_rwlock_lock {
+	struct mcs_rwlock_node reader;		/* common reader lock */
+	struct mcs_rwlock_node *node;		/* base */
+} __attribute__((aligned(64))) mcs_rwlock_lock_t;
+
+static void
+mcs_rwlock_init(struct mcs_rwlock_lock *lock)
+{
+	ihk_atomic_set(&lock->reader.count, 0);
+	lock->reader.type = MCS_RWLOCK_TYPE_COMMON_READER;
+	lock->node = NULL;
+}
+
+#ifdef DEBUG_MCS_RWLOCK
+#define mcs_rwlock_writer_lock_noirq(l, n) { \
+__kprintf("[%d] call mcs_rwlock_writer_lock_noirq %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__mcs_rwlock_writer_lock_noirq((l), (n)); \
+__kprintf("[%d] ret mcs_rwlock_writer_lock_noirq\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define mcs_rwlock_writer_lock_noirq __mcs_rwlock_writer_lock_noirq
 #endif
+static void
+__mcs_rwlock_writer_lock_noirq(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node *node)
+{
+	struct mcs_rwlock_node *pred;
+
+	preempt_disable();
+
+	node->type = MCS_RWLOCK_TYPE_WRITER;
+	node->next = NULL;
+
+	pred = (struct mcs_rwlock_node *)xchg8((unsigned long *)&lock->node,
+			(unsigned long)node);
+
+	if (pred) {
+		node->locked = MCS_RWLOCK_LOCKED;
+		pred->next = node;
+		while (node->locked != MCS_RWLOCK_UNLOCKED) {
+			cpu_pause();
+		}
+	}
+}
+
+static void
+mcs_rwlock_unlock_readers(struct mcs_rwlock_lock *lock)
+{
+	struct mcs_rwlock_node *p;
+	struct mcs_rwlock_node *f = NULL;
+	struct mcs_rwlock_node *n;
+	int breakf = 0;
+
+	ihk_atomic_inc(&lock->reader.count); // protect to unlock reader
+	for(p = &lock->reader; p->next; p = n){
+		n = p->next;
+		if(p->next->type == MCS_RWLOCK_TYPE_READER){
+			p->next = n->next;
+			if(lock->node == n){
+				struct mcs_rwlock_node *old;
+
+				old = (struct mcs_rwlock_node *)atomic_cmpxchg8(
+				       (unsigned long *)&lock->node,
+				       (unsigned long)n,
+				       (unsigned long)p);
+
+				if(old != n){ // couldn't change
+					while (n->next == NULL) {
+						cpu_pause();
+					}
+					p->next = n->next;
+				}
+				else{
+					breakf = 1;
+				}
+			}
+			else if(p->next == NULL){
+				while (n->next == NULL) {
+					cpu_pause();
+				}
+				p->next = n->next;
+			}
+			if(f){
+				ihk_atomic_inc(&lock->reader.count);
+				n->locked = MCS_RWLOCK_UNLOCKED;
+			}
+			else
+				f = n;
+			n = p;
+			if(breakf)
+				break;
+		}
+		if(n->next == NULL && lock->node != n){
+			while (n->next == NULL && lock->node != n) {
+				cpu_pause();
+			}
+		}
+	}
+
+	f->locked = MCS_RWLOCK_UNLOCKED;
+}
+
+#ifdef DEBUG_MCS_RWLOCK
+#define mcs_rwlock_writer_unlock_noirq(l, n) { \
+__kprintf("[%d] call mcs_rwlock_writer_unlock_noirq %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__mcs_rwlock_writer_unlock_noirq((l), (n)); \
+__kprintf("[%d] ret mcs_rwlock_writer_unlock_noirq\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define mcs_rwlock_writer_unlock_noirq __mcs_rwlock_writer_unlock_noirq
+#endif
+static void
+__mcs_rwlock_writer_unlock_noirq(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node *node)
+{
+	if (node->next == NULL) {
+		struct mcs_rwlock_node *old = (struct mcs_rwlock_node *)
+			atomic_cmpxchg8((unsigned long *)&lock->node,
+					(unsigned long)node, (unsigned long)0);
+
+		if (old == node) {
+			goto out;
+		}
+
+		while (node->next == NULL) {
+			cpu_pause();
+		}
+	}
+
+	if(node->next->type == MCS_RWLOCK_TYPE_READER){
+		lock->reader.next = node->next;
+		mcs_rwlock_unlock_readers(lock);
+	}
+	else{
+		node->next->locked = MCS_RWLOCK_UNLOCKED;
+	}
+
+out:
+	preempt_enable();
+}
+
+#ifdef DEBUG_MCS_RWLOCK
+#define mcs_rwlock_reader_lock_noirq(l, n) { \
+__kprintf("[%d] call mcs_rwlock_reader_lock_noirq %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__mcs_rwlock_reader_lock_noirq((l), (n)); \
+__kprintf("[%d] ret mcs_rwlock_reader_lock_noirq\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define mcs_rwlock_reader_lock_noirq __mcs_rwlock_reader_lock_noirq
+#endif
+
+static inline unsigned int
+atomic_inc_ifnot0(ihk_atomic_t *v)
+{
+	unsigned int *p = (unsigned int *)(&(v)->counter);
+	unsigned int old;
+	unsigned int new;
+	unsigned int val;
+
+	do{
+		if(!(old = *p))
+			break;
+		new = old + 1;
+		val = atomic_cmpxchg4(p, old, new);
+	}while(val != old);
+	return old;
+}
+
+static void
+__mcs_rwlock_reader_lock_noirq(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node *node)
+{
+	struct mcs_rwlock_node *pred;
+
+	preempt_disable();
+
+	node->type = MCS_RWLOCK_TYPE_READER;
+	node->next = NULL;
+	node->dmy1 = ihk_mc_get_processor_id();
+
+	pred = (struct mcs_rwlock_node *)xchg8((unsigned long *)&lock->node,
+			(unsigned long)node);
+
+	if (pred) {
+		if(pred == &lock->reader){
+			if(atomic_inc_ifnot0(&pred->count)){
+				struct mcs_rwlock_node *old;
+
+				old = (struct mcs_rwlock_node *)atomic_cmpxchg8(
+				       (unsigned long *)&lock->node,
+				       (unsigned long)node,
+				       (unsigned long)pred);
+
+				if (old == node) {
+					goto out;
+				}
+
+				while (node->next == NULL) {
+					cpu_pause();
+				}
+
+				node->locked = MCS_RWLOCK_LOCKED;
+				lock->reader.next = node;
+				mcs_rwlock_unlock_readers(lock);
+				ihk_atomic_dec(&pred->count);
+				goto out;
+			}
+		}
+		node->locked = MCS_RWLOCK_LOCKED;
+		pred->next = node;
+		while (node->locked != MCS_RWLOCK_UNLOCKED) {
+			cpu_pause();
+		}
+	}
+	else {
+		lock->reader.next = node;
+		mcs_rwlock_unlock_readers(lock);
+	}
+out:
+	return;
+}
+
+#ifdef DEBUG_MCS_RWLOCK
+#define mcs_rwlock_reader_unlock_noirq(l, n) { \
+__kprintf("[%d] call mcs_rwlock_reader_unlock_noirq %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__mcs_rwlock_reader_unlock_noirq((l), (n)); \
+__kprintf("[%d] ret mcs_rwlock_reader_unlock_noirq\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define mcs_rwlock_reader_unlock_noirq __mcs_rwlock_reader_unlock_noirq
+#endif
+static void
+__mcs_rwlock_reader_unlock_noirq(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node *node)
+{
+	if(ihk_atomic_dec_return(&lock->reader.count))
+		goto out;
+
+	if (lock->reader.next == NULL) {
+		struct mcs_rwlock_node *old;
+
+		old = (struct mcs_rwlock_node *)atomic_cmpxchg8(
+		       (unsigned long *)&lock->node,
+		       (unsigned long)&lock->reader,
+		       (unsigned long)0);
+
+		if (old == &lock->reader) {
+			goto out;
+		}
+
+		while (lock->reader.next == NULL) {
+			cpu_pause();
+		}
+	}
+
+	if(lock->reader.next->type == MCS_RWLOCK_TYPE_READER){
+		mcs_rwlock_unlock_readers(lock);
+	}
+	else{
+		lock->reader.next->locked = MCS_RWLOCK_UNLOCKED;
+	}
+
+out:
+	preempt_enable();
+}
+
+#ifdef DEBUG_MCS_RWLOCK
+#define mcs_rwlock_writer_lock(l, n) { \
+__kprintf("[%d] call mcs_rwlock_writer_lock %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__mcs_rwlock_writer_lock((l), (n)); \
+__kprintf("[%d] ret mcs_rwlock_writer_lock\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define mcs_rwlock_writer_lock __mcs_rwlock_writer_lock
+#endif
+static void
+__mcs_rwlock_writer_lock(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node_irqsave *node)
+{
+	node->irqsave = cpu_disable_interrupt_save();
+	__mcs_rwlock_writer_lock_noirq(lock, &node->node);
+}
+
+#ifdef DEBUG_MCS_RWLOCK
+#define mcs_rwlock_writer_unlock(l, n) { \
+__kprintf("[%d] call mcs_rwlock_writer_unlock %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__mcs_rwlock_writer_unlock((l), (n)); \
+__kprintf("[%d] ret mcs_rwlock_writer_unlock\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define mcs_rwlock_writer_unlock __mcs_rwlock_writer_unlock
+#endif
+static void
+__mcs_rwlock_writer_unlock(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node_irqsave *node)
+{
+	__mcs_rwlock_writer_unlock_noirq(lock, &node->node);
+	cpu_restore_interrupt(node->irqsave);
+}
+
+#ifdef DEBUG_MCS_RWLOCK
+#define mcs_rwlock_reader_lock(l, n) { \
+__kprintf("[%d] call mcs_rwlock_reader_lock %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__mcs_rwlock_reader_lock((l), (n)); \
+__kprintf("[%d] ret mcs_rwlock_reader_lock\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define mcs_rwlock_reader_lock __mcs_rwlock_reader_lock
+#endif
+static void
+__mcs_rwlock_reader_lock(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node_irqsave *node)
+{
+	node->irqsave = cpu_disable_interrupt_save();
+	__mcs_rwlock_reader_lock_noirq(lock, &node->node);
+}
+
+#ifdef DEBUG_MCS_RWLOCK
+#define mcs_rwlock_reader_unlock(l, n) { \
+__kprintf("[%d] call mcs_rwlock_reader_unlock %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
+__mcs_rwlock_reader_unlock((l), (n)); \
+__kprintf("[%d] ret mcs_rwlock_reader_unlock\n", ihk_mc_get_processor_id()); \
+}
+#else
+#define mcs_rwlock_reader_unlock __mcs_rwlock_reader_unlock
+#endif
+static void
+__mcs_rwlock_reader_unlock(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node_irqsave *node)
+{
+	__mcs_rwlock_reader_unlock_noirq(lock, &node->node);
+	cpu_restore_interrupt(node->irqsave);
 }
 
 #endif
-

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

@@ -22,6 +22,7 @@
 #define USER_CS_ENTRY      6
 #define USER_DS_ENTRY      7
 #define GLOBAL_TSS_ENTRY   8
+#define GETCPU_ENTRY       15
 
 #define KERNEL_CS          (KERNEL_CS_ENTRY * 8)
 #define KERNEL_DS          (KERNEL_DS_ENTRY * 8)
@@ -40,10 +41,12 @@
 #define LARGE_PAGE_P2ALIGN (LARGE_PAGE_SHIFT - PAGE_SHIFT)
 
 #define USER_END           0x0000800000000000UL
+#define TASK_UNMAPPED_BASE 0x00002AAAAAA00000UL
 #define MAP_ST_START       0xffff800000000000UL
 #define MAP_VMAP_START     0xfffff00000000000UL
 #define MAP_FIXED_START    0xffffffff70000000UL
 #define MAP_KERNEL_START   0xffffffff80000000UL
+#define STACK_TOP(region)  ((region)->user_end)
 
 #define MAP_VMAP_SIZE      0x0000000100000000UL
 
@@ -65,6 +68,8 @@
 
 #define	PF_PRESENT	((pte_t)0x01)	/* entry is valid */
 #define PF_WRITABLE	((pte_t)0x02)
+#define PFLX_PWT        ((pte_t)0x08)
+#define PFLX_PCD        ((pte_t)0x10)
 #define	PF_SIZE		((pte_t)0x80)	/* entry points large page */
 
 #define PFL4_PRESENT    ((pte_t)0x01)
@@ -74,8 +79,8 @@
 #define PFL3_PRESENT    ((pte_t)0x01)
 #define PFL3_WRITABLE   ((pte_t)0x02)
 #define PFL3_USER       ((pte_t)0x04)
-#define PFL3_PWT        ((pte_t)0x08)
-#define PFL3_PCD        ((pte_t)0x10)
+#define PFL3_PWT        PFLX_PWT
+#define PFL3_PCD        PFLX_PCD
 #define PFL3_ACCESSED   ((pte_t)0x20)
 #define PFL3_DIRTY      ((pte_t)0x40)
 #define PFL3_SIZE       ((pte_t)0x80)   /* Used in 1G page */
@@ -86,8 +91,8 @@
 #define PFL2_PRESENT    ((pte_t)0x01)
 #define PFL2_WRITABLE   ((pte_t)0x02)
 #define PFL2_USER       ((pte_t)0x04)
-#define PFL2_PWT        ((pte_t)0x08)
-#define PFL2_PCD        ((pte_t)0x10)
+#define PFL2_PWT        PFLX_PWT
+#define PFL2_PCD        PFLX_PCD
 #define PFL2_ACCESSED   ((pte_t)0x20)
 #define PFL2_DIRTY      ((pte_t)0x40)
 #define PFL2_SIZE       ((pte_t)0x80)   /* Used in 2M page */
@@ -98,8 +103,8 @@
 #define PFL1_PRESENT    ((pte_t)0x01)
 #define PFL1_WRITABLE   ((pte_t)0x02)
 #define PFL1_USER       ((pte_t)0x04)
-#define PFL1_PWT        ((pte_t)0x08)
-#define PFL1_PCD        ((pte_t)0x10)
+#define PFL1_PWT        PFLX_PWT
+#define PFL1_PCD        PFLX_PCD
 #define PFL1_ACCESSED   ((pte_t)0x20)
 #define PFL1_DIRTY      ((pte_t)0x40)
 #define PFL1_IGNORED_11 ((pte_t)1 << 11)
@@ -152,6 +157,8 @@ enum ihk_mc_pt_attribute {
 	PTATTR_WRITE_COMBINED = 0x40000,
 };
 
+enum ihk_mc_pt_attribute attr_mask;
+
 static inline int pte_is_null(pte_t *ptep)
 {
 	return (*ptep == PTE_NULL);
@@ -207,6 +214,27 @@ static inline off_t pte_get_off(pte_t *ptep, size_t pgsize)
 	return (off_t)(*ptep & PAGE_MASK);
 }
 
+static inline enum ihk_mc_pt_attribute pte_get_attr(pte_t *ptep, size_t pgsize)
+{
+	enum ihk_mc_pt_attribute attr;
+
+	attr = *ptep & attr_mask;
+	if (*ptep & PFLX_PWT) {
+		if (*ptep & PFLX_PCD) {
+			attr |= PTATTR_UNCACHABLE;
+		}
+		else {
+			attr |= PTATTR_WRITE_COMBINED;
+		}
+	}
+	if (((pgsize == PTL2_SIZE) && (*ptep & PFL2_SIZE))
+			|| ((pgsize == PTL3_SIZE) && (*ptep & PFL3_SIZE))) {
+		attr |= PTATTR_LARGEPAGE;
+	}
+
+	return attr;
+} /* pte_get_attr() */
+
 static inline void pte_make_null(pte_t *ptep, size_t pgsize)
 {
 	*ptep = PTE_NULL;
@@ -278,7 +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);
@@ -290,5 +318,5 @@ extern unsigned long ap_trampoline;
 #define AP_TRAMPOLINE_SIZE  0x2000
 
 /* Local is cachable */
-#define IHK_IKC_QUEUE_PT_ATTR (PTATTR_NO_EXECUTE | PTATTR_WRITABLE | PTATTR_UNCACHABLE)
+#define IHK_IKC_QUEUE_PT_ATTR (PTATTR_NO_EXECUTE | PTATTR_WRITABLE)
 #endif

arch/x86/kernel/include/arch/auxvec.h

@@ -0,0 +1,18 @@
+/**
+ * \file auxvec.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Declare architecture-dependent constants for auxiliary vector
+ * \author Gou Nakamura  <go.nakamura.yw@hitachi-solutions.com>
+ *      Copyright (C) 2016  RIKEN AICS
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef ARCH_AUXVEC_H
+#define ARCH_AUXVEC_H
+
+#define AT_SYSINFO_EHDR 33
+
+#endif

arch/x86/kernel/include/arch/cpu.h

@@ -0,0 +1,37 @@
+/**
+ * \file cpu.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Declare architecture-dependent types and functions to control CPU.
+ * \author Gou Nakamura  <go.nakamura.yw@hitachi-solutions.com>
+ *      Copyright (C) 2015  RIKEN AICS
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef ARCH_CPU_H
+#define ARCH_CPU_H
+
+#include <ihk/cpu.h>
+
+static inline void rmb(void)
+{
+	barrier();
+}
+
+static inline void wmb(void)
+{
+	barrier();
+}
+
+static unsigned long read_tsc(void)
+{
+	unsigned int low, high;
+
+	asm volatile("rdtsc" : "=a"(low), "=d"(high));
+
+	return (low | ((unsigned long)high << 32));
+}
+
+#endif /* ARCH_CPU_H */

arch/x86/kernel/include/arch/mm.h

@@ -0,0 +1,16 @@
+#ifndef __ARCH_MM_H
+#define __ARCH_MM_H
+
+struct process_vm;
+
+static inline void
+flush_nfo_tlb()
+{
+}
+
+static inline void
+flush_nfo_tlb_mm(struct process_vm *vm)
+{
+}
+
+#endif

arch/x86/kernel/include/arch/mman.h

@@ -27,6 +27,10 @@
 #define	MAP_STACK	0x00020000
 #define	MAP_HUGETLB	0x00040000
 
+#define MAP_HUGE_SHIFT  26
+#define MAP_HUGE_2MB    (21 << MAP_HUGE_SHIFT)
+#define MAP_HUGE_1GB    (30 << MAP_HUGE_SHIFT)
+
 /*
  * for mlockall()
  */

arch/x86/kernel/include/arch/shm.h

@@ -13,6 +13,11 @@
 #ifndef HEADER_ARCH_SHM_H
 #define HEADER_ARCH_SHM_H
 
+/* shmflg */
+#define SHM_HUGE_SHIFT  26
+#define SHM_HUGE_2MB    (21 << SHM_HUGE_SHIFT)
+#define SHM_HUGE_1GB    (30 << SHM_HUGE_SHIFT)
+
 struct ipc_perm {
 	key_t		key;
 	uid_t		uid;
@@ -34,7 +39,8 @@ struct shmid_ds {
 	pid_t		shm_cpid;
 	pid_t		shm_lpid;
 	uint64_t	shm_nattch;
-	uint8_t		padding[16];
+	uint8_t		padding[12];
+	int		init_pgshift;
 };
 
 #endif /* HEADER_ARCH_SHM_H */

arch/x86/kernel/include/cpulocal.h

@@ -22,7 +22,7 @@
  * - 4096 : kernel stack
  */
 
-#define X86_CPU_LOCAL_OFFSET_TSS    128
+#define X86_CPU_LOCAL_OFFSET_TSS    176
 #define X86_CPU_LOCAL_OFFSET_KSTACK 16
 #define X86_CPU_LOCAL_OFFSET_USTACK 24
 
@@ -39,13 +39,13 @@ struct x86_cpu_local_variables {
 	struct x86_desc_ptr gdt_ptr;
 	unsigned short pad[3];
 /* 48 */
-	uint64_t gdt[10];
-/* 128 */
+	uint64_t gdt[16];
+/* 176 */
 	struct tss64 tss;
-/* 232 */
+/* 280 */
 	unsigned long paniced;
 	uint64_t panic_regs[21];
-/* 408 */
+/* 456 */
 } __attribute__((packed));
 
 struct x86_cpu_local_variables *get_x86_cpu_local_variable(int id);

arch/x86/kernel/include/errno.h

@@ -1,40 +1,7 @@
-#ifndef _ASM_GENERIC_ERRNO_BASE_H
-#define _ASM_GENERIC_ERRNO_BASE_H
+#ifndef _ERRNO_BASE_H
+#define _ERRNO_BASE_H
 
-#define	EPERM		 1	/* Operation not permitted */
-#define	ENOENT		 2	/* No such file or directory */
-#define	ESRCH		 3	/* No such process */
-#define	EINTR		 4	/* Interrupted system call */
-#define	EIO		 5	/* I/O error */
-#define	ENXIO		 6	/* No such device or address */
-#define	E2BIG		 7	/* Argument list too long */
-#define	ENOEXEC		 8	/* Exec format error */
-#define	EBADF		 9	/* Bad file number */
-#define	ECHILD		10	/* No child processes */
-#define	EAGAIN		11	/* Try again */
-#define	ENOMEM		12	/* Out of memory */
-#define	EACCES		13	/* Permission denied */
-#define	EFAULT		14	/* Bad address */
-#define	ENOTBLK		15	/* Block device required */
-#define	EBUSY		16	/* Device or resource busy */
-#define	EEXIST		17	/* File exists */
-#define	EXDEV		18	/* Cross-device link */
-#define	ENODEV		19	/* No such device */
-#define	ENOTDIR		20	/* Not a directory */
-#define	EISDIR		21	/* Is a directory */
-#define	EINVAL		22	/* Invalid argument */
-#define	ENFILE		23	/* File table overflow */
-#define	EMFILE		24	/* Too many open files */
-#define	ENOTTY		25	/* Not a typewriter */
-#define	ETXTBSY		26	/* Text file busy */
-#define	EFBIG		27	/* File too large */
-#define	ENOSPC		28	/* No space left on device */
-#define	ESPIPE		29	/* Illegal seek */
-#define	EROFS		30	/* Read-only file system */
-#define	EMLINK		31	/* Too many links */
-#define	EPIPE		32	/* Broken pipe */
-#define	EDOM		33	/* Math argument out of domain of func */
-#define	ERANGE		34	/* Math result not representable */
+#include <generic-errno.h>
 
 #define	EDEADLK		35	/* Resource deadlock would occur */
 #define	ENAMETOOLONG	36	/* File name too long */
@@ -141,29 +108,4 @@
 
 #define ERFKILL		132	/* Operation not possible due to RF-kill */
 
-
-#ifdef __KERNEL__
-
-/* Should never be seen by user programs */
-#define ERESTARTSYS	512
-#define ERESTARTNOINTR	513
-#define ERESTARTNOHAND	514	/* restart if no handler.. */
-#define ENOIOCTLCMD	515	/* No ioctl command */
-#define ERESTART_RESTARTBLOCK 516 /* restart by calling sys_restart_syscall */
-
-/* Defined for the NFSv3 protocol */
-#define EBADHANDLE	521	/* Illegal NFS file handle */
-#define ENOTSYNC	522	/* Update synchronization mismatch */
-#define EBADCOOKIE	523	/* Cookie is stale */
-#define ENOTSUPP	524	/* Operation is not supported */
-#define ETOOSMALL	525	/* Buffer or request is too small */
-#define ESERVERFAULT	526	/* An untranslatable error occurred */
-#define EBADTYPE	527	/* Type not supported by server */
-#define EJUKEBOX	528	/* Request initiated, but will not complete before timeout */
-#define EIOCBQUEUED	529	/* iocb queued, will get completion event */
-#define EIOCBRETRY	530	/* iocb queued, will trigger a retry */
-
-#endif
-
-
 #endif

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

@@ -13,6 +13,10 @@
 #ifndef HEADER_X86_COMMON_IHK_ATOMIC_H
 #define HEADER_X86_COMMON_IHK_ATOMIC_H
  
+/***********************************************************************
+ * ihk_atomic_t
+ */
+
 typedef struct {
 	int counter;
 } ihk_atomic_t;
@@ -95,6 +99,30 @@ static inline int ihk_atomic_sub_return(int i, ihk_atomic_t *v)
 #define ihk_atomic_inc_return(v)  (ihk_atomic_add_return(1, v))
 #define ihk_atomic_dec_return(v)  (ihk_atomic_sub_return(1, v))
 
+/***********************************************************************
+ * ihk_atomic64_t
+ */
+
+typedef struct {
+	long counter64;
+} ihk_atomic64_t;
+
+#define IHK_ATOMIC64_INIT(i) { .counter64 = (i) }
+
+static inline long ihk_atomic64_read(const ihk_atomic64_t *v)
+{
+	return *(volatile long *)&(v)->counter64;
+}
+
+static inline void ihk_atomic64_inc(ihk_atomic64_t *v)
+{
+	asm volatile ("lock incq %0" : "+m"(v->counter64));
+}
+
+/***********************************************************************
+ * others
+ */
+
 /*
  * Note: no "lock" prefix even on SMP: xchg always implies lock anyway
  * Note 2: xchg has side effect, so that attribute volatile is necessary,
@@ -112,6 +140,17 @@ static inline int ihk_atomic_sub_return(int i, ihk_atomic_t *v)
 	__x;								\
 })
 
+static inline unsigned long xchg8(unsigned long *ptr, unsigned long x)
+{
+	unsigned long __x = (x);
+	asm volatile("xchgq %0,%1"
+			 : "=r" (__x)
+			 : "m" (*(volatile unsigned long*)(ptr)), "0" (__x)
+			 : "memory");
+
+	return __x;
+}
+
 #define __xchg(x, ptr, size)						\
 ({									\
 	__typeof(*(ptr)) __x = (x);					\
@@ -150,5 +189,30 @@ static inline int ihk_atomic_sub_return(int i, ihk_atomic_t *v)
 #define xchg(ptr, v)							\
 	__xchg((v), (ptr), sizeof(*ptr))
 
+static inline unsigned long atomic_cmpxchg8(unsigned long *addr,
+		unsigned long oldval,
+		unsigned long newval)
+{
+	asm volatile("lock; cmpxchgq %2, %1\n"
+		     : "=a" (oldval), "+m" (*addr)
+		     : "r" (newval), "0" (oldval)
+		     : "memory"
+	);
+
+	return oldval;
+}
+
+static inline unsigned long atomic_cmpxchg4(unsigned int *addr,
+		unsigned int oldval,
+		unsigned int newval)
+{
+	asm volatile("lock; cmpxchgl %2, %1\n"
+		     : "=a" (oldval), "+m" (*addr)
+		     : "r" (newval), "0" (oldval)
+		     : "memory"
+	);
+
+	return oldval;
+}
 
 #endif

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

@@ -31,9 +31,5 @@ typedef int64_t            off_t;
 
 #define NULL ((void *)0)
 
-#define BITS_PER_LONG_SHIFT 6 
-#define BITS_PER_LONG (1 << BITS_PER_LONG_SHIFT)
-
-
 #endif
 

arch/x86/kernel/include/prctl.h

@@ -0,0 +1,17 @@
+/**
+ * \file prctl.h
+ *  License details are found in the file LICENSE.
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef __ARCH_PRCTL_H
+#define __ARCH_PRCTL_H
+
+#define ARCH_SET_GS 0x1001
+#define ARCH_SET_FS 0x1002
+#define ARCH_GET_FS 0x1003
+#define ARCH_GET_GS 0x1004
+
+#endif

arch/x86/kernel/include/registers.h

@@ -58,6 +58,7 @@
 #define MSR_IA32_ENERGY_PERF_BIAS	0x000001b0
 #define MSR_NHM_TURBO_RATIO_LIMIT	0x000001ad
 #define MSR_IA32_CR_PAT			0x00000277
+#define MSR_IA32_XSS			0xda0
 
 
 #define CVAL(event, mask) \

arch/x86/kernel/include/rlimit.h

@@ -90,10 +90,6 @@ enum __rlimit_resource
 #define RLIM_NLIMITS __RLIM_NLIMITS
 };
 
-
-struct rlimit {
-	uint64_t rlim_cur;  /* Soft limit */
-	uint64_t rlim_max;  /* Hard limit (ceiling for rlim_cur) */
-};
+#include <generic-rlimit.h>
 
 #endif

arch/x86/kernel/include/syscall_list.h

@@ -20,12 +20,13 @@
  *     syscall_name[] only, no handler exists.
  */
 
-SYSCALL_DELEGATED(0, read)
+SYSCALL_HANDLED(0, read)
 SYSCALL_DELEGATED(1, write)
 SYSCALL_DELEGATED(2, open)
 SYSCALL_HANDLED(3, close)
 SYSCALL_DELEGATED(4, stat)
 SYSCALL_DELEGATED(5, fstat)
+SYSCALL_DELEGATED(7, poll)
 SYSCALL_DELEGATED(8, lseek)
 SYSCALL_HANDLED(9, mmap)
 SYSCALL_HANDLED(10, mprotect)
@@ -34,11 +35,12 @@ SYSCALL_HANDLED(12, brk)
 SYSCALL_HANDLED(13, rt_sigaction)
 SYSCALL_HANDLED(14, rt_sigprocmask)
 SYSCALL_HANDLED(15, rt_sigreturn)
-SYSCALL_DELEGATED(16, ioctl)
+SYSCALL_HANDLED(16, ioctl)
 SYSCALL_DELEGATED(17, pread64)
 SYSCALL_DELEGATED(18, pwrite64)
 SYSCALL_DELEGATED(20, writev)
 SYSCALL_DELEGATED(21, access)
+SYSCALL_DELEGATED(23, select)
 SYSCALL_HANDLED(24, sched_yield)
 SYSCALL_HANDLED(25, mremap)
 SYSCALL_HANDLED(26, msync)
@@ -48,6 +50,9 @@ SYSCALL_HANDLED(29, shmget)
 SYSCALL_HANDLED(30, shmat)
 SYSCALL_HANDLED(31, shmctl)
 SYSCALL_HANDLED(34, pause)
+SYSCALL_HANDLED(35, nanosleep)
+SYSCALL_HANDLED(36, getitimer)
+SYSCALL_HANDLED(38, setitimer)
 SYSCALL_HANDLED(39, getpid)
 SYSCALL_HANDLED(56, clone)
 SYSCALL_DELEGATED(57, fork)
@@ -57,12 +62,17 @@ SYSCALL_HANDLED(60, exit)
 SYSCALL_HANDLED(61, wait4)
 SYSCALL_HANDLED(62, kill)
 SYSCALL_DELEGATED(63, uname)
+SYSCALL_DELEGATED(65, semop)
 SYSCALL_HANDLED(67, shmdt)
-SYSCALL_DELEGATED(72, fcntl)
+SYSCALL_DELEGATED(69, msgsnd)
+SYSCALL_DELEGATED(70, msgrcv)
+SYSCALL_HANDLED(72, fcntl)
 SYSCALL_DELEGATED(79, getcwd)
 SYSCALL_DELEGATED(89, readlink)
-SYSCALL_DELEGATED(96, gettimeofday)
+SYSCALL_HANDLED(96, gettimeofday)
 SYSCALL_HANDLED(97, getrlimit)
+SYSCALL_HANDLED(98, getrusage)
+SYSCALL_HANDLED(100, times)
 SYSCALL_HANDLED(101, ptrace)
 SYSCALL_HANDLED(102, getuid)
 SYSCALL_HANDLED(104, getgid)
@@ -99,25 +109,36 @@ SYSCALL_HANDLED(151, mlockall)
 SYSCALL_HANDLED(152, munlockall)
 SYSCALL_HANDLED(158, arch_prctl)
 SYSCALL_HANDLED(160, setrlimit)
+SYSCALL_HANDLED(164, settimeofday)
 SYSCALL_HANDLED(186, gettid)
+SYSCALL_HANDLED(200, tkill)
 SYSCALL_DELEGATED(201, time)
 SYSCALL_HANDLED(202, futex)
 SYSCALL_HANDLED(203, sched_setaffinity)
 SYSCALL_HANDLED(204, sched_getaffinity)
+SYSCALL_DELEGATED(208, io_getevents)
 SYSCALL_HANDLED(216, remap_file_pages)
 SYSCALL_DELEGATED(217, getdents64)
 SYSCALL_HANDLED(218, set_tid_address)
+SYSCALL_DELEGATED(220, semtimedop)
+SYSCALL_HANDLED(228, clock_gettime)
+SYSCALL_DELEGATED(230, clock_nanosleep)
 SYSCALL_HANDLED(231, exit_group)
+SYSCALL_DELEGATED(232, epoll_wait)
 SYSCALL_HANDLED(234, tgkill)
 SYSCALL_HANDLED(237, mbind)
 SYSCALL_HANDLED(238, set_mempolicy)
 SYSCALL_HANDLED(239, get_mempolicy)
 SYSCALL_HANDLED(247, waitid)
 SYSCALL_HANDLED(256, migrate_pages)
+SYSCALL_DELEGATED(270, pselect6)
+SYSCALL_DELEGATED(271, ppoll)
 SYSCALL_HANDLED(273, set_robust_list)
 SYSCALL_HANDLED(279, move_pages)
+SYSCALL_DELEGATED(281, epoll_pwait)
 SYSCALL_HANDLED(282, signalfd)
 SYSCALL_HANDLED(289, signalfd4)
+SYSCALL_HANDLED(298, perf_event_open)
 #ifdef DCFA_KMOD
 SYSCALL_HANDLED(303, mod_call)
 #endif

arch/x86/kernel/interrupt.S

@@ -13,7 +13,7 @@
  * 2013/?? - bgerofi + shimosawa: handle rsp correctly for nested interrupts
  */
 
-#define X86_CPU_LOCAL_OFFSET_TSS    128
+#define X86_CPU_LOCAL_OFFSET_TSS    176
 #define X86_TSS_OFFSET_SP0          4
 #define X86_CPU_LOCAL_OFFSET_SP0 \
 		(X86_CPU_LOCAL_OFFSET_TSS + X86_TSS_OFFSET_SP0)
@@ -206,9 +206,12 @@ x86_syscall:
 		
 .globl enter_user_mode
 enter_user_mode:
+		callq	release_runq_lock
 		movq	$0,  %rdi
 		movq	%rsp, %rsi
-		call check_signal
+		call	check_signal
+		movq	$0,  %rdi
+		call	set_cputime
 		POP_ALL_REGS
 		addq	$8, %rsp
 		iretq

arch/x86/kernel/local.c

@@ -38,6 +38,11 @@ void init_processors_local(int max_id)
 	kprintf("locals = %p\n", locals);
 }
 
+/*@
+  @ requires \valid(id);
+  @ ensures \result == locals + (LOCALS_SPAN * id);
+  @ assigns \nothing;
+  @*/
 struct x86_cpu_local_variables *get_x86_cpu_local_variable(int id)
 {
 	return (struct x86_cpu_local_variables *)
@@ -98,6 +103,10 @@ void init_boot_processor_local(void)
 }
 
 /** IHK **/
+/*@
+  @ ensures \result == %gs;
+  @ assigns \nothing;
+  */
 int ihk_mc_get_processor_id(void)
 {
 	int id;
@@ -107,6 +116,10 @@ int ihk_mc_get_processor_id(void)
 	return id;
 }
 
+/*@
+  @ ensures \result == (locals + (LOCALS_SPAN * %gs))->apic_id;
+  @ assigns \nothing;
+  */
 int ihk_mc_get_hardware_processor_id(void)
 {
 	struct x86_cpu_local_variables *v =  get_x86_this_cpu_local();

arch/x86/kernel/mikc.c

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

arch/x86/kernel/perfctr.c

@@ -12,19 +12,72 @@
 #include <errno.h>
 #include <ihk/debug.h>
 #include <registers.h>
+#include <mc_perf_event.h>
 
 extern unsigned int *x86_march_perfmap;
+extern int running_on_kvm(void);
 
 #define X86_CR4_PCE     0x00000100
 
+int perf_counters_discovered = 0;
+int X86_IA32_NUM_PERF_COUNTERS = 0;
+unsigned long X86_IA32_PERF_COUNTERS_MASK = 0;
+int X86_IA32_NUM_FIXED_PERF_COUNTERS = 0;
+unsigned long X86_IA32_FIXED_PERF_COUNTERS_MASK = 0;
+
 void x86_init_perfctr(void)
 {
+	int i = 0;
 	unsigned long reg;
+	unsigned long value = 0;
+	uint64_t op;
+	uint64_t eax;
+	uint64_t ebx;
+	uint64_t ecx;
+	uint64_t edx;
+
+	/* Do not do it on KVM */
+	if (running_on_kvm()) return;
 
 	/* Allow PMC to be read from user space */
 	asm volatile("movq %%cr4, %0" : "=r"(reg));
 	reg |= X86_CR4_PCE;
 	asm volatile("movq %0, %%cr4" : : "r"(reg));
+	
+	/* Detect number of supported performance counters */
+	if (!perf_counters_discovered) {
+		/* See Table 35.2 - Architectural MSRs in Vol 3C */
+		op = 0x0a;
+		asm volatile("cpuid" : "=a"(eax),"=b"(ebx),"=c"(ecx),"=d"(edx):"a"(op));
+
+		X86_IA32_NUM_PERF_COUNTERS = ((eax & 0xFF00) >> 8);
+		X86_IA32_PERF_COUNTERS_MASK = (1 << X86_IA32_NUM_PERF_COUNTERS) - 1;
+		
+		X86_IA32_NUM_FIXED_PERF_COUNTERS = (edx & 0x0F);
+		X86_IA32_FIXED_PERF_COUNTERS_MASK = 
+			((1UL << X86_IA32_NUM_FIXED_PERF_COUNTERS) - 1) <<
+			X86_IA32_BASE_FIXED_PERF_COUNTERS;
+	
+		perf_counters_discovered = 1;
+		kprintf("X86_IA32_NUM_PERF_COUNTERS: %d, X86_IA32_NUM_FIXED_PERF_COUNTERS: %d\n",
+				X86_IA32_NUM_PERF_COUNTERS, X86_IA32_NUM_FIXED_PERF_COUNTERS);
+	}
+
+	/* Clear Fixed Counter Control */
+	value = rdmsr(MSR_PERF_FIXED_CTRL);
+	value &= 0xfffffffffffff000L;
+	wrmsr(MSR_PERF_FIXED_CTRL, value);
+
+	/* Clear Generic Counter Control */
+	for(i = 0; i < X86_IA32_NUM_PERF_COUNTERS; i++) {
+		wrmsr(MSR_IA32_PERFEVTSEL0 + i, 0);
+	}
+		
+	/* Enable PMC Control */
+	value = rdmsr(MSR_PERF_GLOBAL_CTRL);
+	value |= X86_IA32_PERF_COUNTERS_MASK;
+	value |= X86_IA32_FIXED_PERF_COUNTERS_MASK;
+	wrmsr(MSR_PERF_GLOBAL_CTRL, value);
 }
 
 static int set_perfctr_x86_direct(int counter, int mode, unsigned int value)
@@ -33,20 +86,53 @@ static int set_perfctr_x86_direct(int counter, int mode, unsigned int value)
 		return -EINVAL;
 	}
 
-	if (mode & PERFCTR_USER_MODE) {
+	// clear mode flags
+	value &= ~(3 << 16);
+
+	// set mode flags
+	if(mode & PERFCTR_USER_MODE) {
 		value |= 1 << 16;
-	}
-	if (mode & PERFCTR_KERNEL_MODE) {
+        }
+        if(mode & PERFCTR_KERNEL_MODE) {
 		value |= 1 << 17;
-	}
+        }
+
     //	wrmsr(MSR_PERF_GLOBAL_CTRL, 0);
 
 	value |= (1 << 22) | (1 << 18); /* EN */
+	value |= (1 << 20); /* Enable overflow interrupt */
 
 	wrmsr(MSR_IA32_PERFEVTSEL0 + counter, value);
 
-	kprintf("wrmsr: %d <= %x\n", MSR_PERF_GLOBAL_CTRL, 0);
-	kprintf("wrmsr: %d <= %x\n", MSR_IA32_PERFEVTSEL0 + counter, value);
+	//kprintf("wrmsr: %d <= %x\n", MSR_PERF_GLOBAL_CTRL, 0);
+	//kprintf("wrmsr: %d <= %x\n", MSR_IA32_PERFEVTSEL0 + counter, value);
+
+	return 0;
+}
+
+static int set_pmc_x86_direct(int counter, long val)
+{
+	unsigned long cnt_bit = 0;
+
+	if (counter < 0) {
+		return -EINVAL;
+	}
+
+	val &= 0x000000ffffffffff; // 40bit Mask
+
+	cnt_bit = 1UL << counter;
+	if ( cnt_bit & X86_IA32_PERF_COUNTERS_MASK ) {
+		// set generic pmc
+		wrmsr(MSR_IA32_PMC0 + counter, val);
+	}
+	else if ( cnt_bit & X86_IA32_FIXED_PERF_COUNTERS_MASK ) {
+		// set fixed pmc
+		wrmsr(MSR_IA32_FIXED_CTR0 + counter - X86_IA32_BASE_FIXED_PERF_COUNTERS, val);
+	}
+	else {
+		return -EINVAL;
+	}
+
 	return 0;
 }
 
@@ -57,6 +143,45 @@ static int set_perfctr_x86(int counter, int event, int mask, int inv, int count,
 	                              CVAL2(event, mask, inv, count));
 }
 
+static int set_fixed_counter(int counter, int mode)
+{
+	unsigned long value = 0;
+	unsigned int  ctr_mask = 0xf;
+	int counter_idx = counter - X86_IA32_BASE_FIXED_PERF_COUNTERS ;
+	unsigned int  set_val = 0;
+
+	if (counter_idx < 0 || counter_idx >= X86_IA32_NUM_FIXED_PERF_COUNTERS) {
+		return -EINVAL;
+	}
+
+	// clear specified fixed counter info
+	value = rdmsr(MSR_PERF_FIXED_CTRL);
+	ctr_mask <<= counter_idx * 4;
+	value &= ~ctr_mask;
+	
+	if (mode & PERFCTR_USER_MODE) {
+		set_val |= 1 << 1;
+	}
+	if (mode & PERFCTR_KERNEL_MODE) {
+		set_val |= 1;
+	}
+
+	set_val <<= counter_idx * 4;
+	value |= set_val;
+
+	wrmsr(MSR_PERF_FIXED_CTRL, value);
+
+	return 0;
+}
+
+int ihk_mc_perfctr_init_raw(int counter, unsigned int code, int mode)
+{
+	if (counter < 0 || counter >= X86_IA32_NUM_PERF_COUNTERS) {
+		return -EINVAL;
+	}
+
+	return set_perfctr_x86_direct(counter, mode, code);
+}
 int ihk_mc_perfctr_init(int counter, enum ihk_perfctr_type type, int mode)
 {
 	if (counter < 0 || counter >= X86_IA32_NUM_PERF_COUNTERS) {
@@ -78,14 +203,15 @@ extern void x86_march_perfctr_start(unsigned long counter_mask);
 
 int ihk_mc_perfctr_start(unsigned long counter_mask)
 {
-	unsigned int value = 0;
+	unsigned long value = 0;
+	unsigned long mask = X86_IA32_PERF_COUNTERS_MASK | X86_IA32_FIXED_PERF_COUNTERS_MASK;
 
 #ifdef HAVE_MARCH_PERFCTR_START
 	x86_march_perfctr_start(counter_mask);
 #endif
-	counter_mask &= ((1 << X86_IA32_NUM_PERF_COUNTERS) - 1);
+	counter_mask &= mask;
 	value = rdmsr(MSR_PERF_GLOBAL_CTRL);
-    value |= counter_mask;
+	value |= counter_mask;
 	wrmsr(MSR_PERF_GLOBAL_CTRL, value);
 
 	return 0;
@@ -93,25 +219,78 @@ int ihk_mc_perfctr_start(unsigned long counter_mask)
 
 int ihk_mc_perfctr_stop(unsigned long counter_mask)
 {
-	unsigned int value;
+	unsigned long value;
+	unsigned long mask = X86_IA32_PERF_COUNTERS_MASK | X86_IA32_FIXED_PERF_COUNTERS_MASK;
 
-	counter_mask &= ((1 << X86_IA32_NUM_PERF_COUNTERS) - 1);
+	counter_mask &= mask;
 	value = rdmsr(MSR_PERF_GLOBAL_CTRL);
 	value &= ~counter_mask;
 	wrmsr(MSR_PERF_GLOBAL_CTRL, value);
 
+	if(counter_mask >> 32 & 0x1) {
+		value = rdmsr(MSR_PERF_FIXED_CTRL);
+		value &= ~(0xf);
+		wrmsr(MSR_PERF_FIXED_CTRL, value);
+	}
+		
+	if(counter_mask >> 32 & 0x2) {
+		value = rdmsr(MSR_PERF_FIXED_CTRL);
+		value &= ~(0xf << 4);
+		wrmsr(MSR_PERF_FIXED_CTRL, value);
+	}
+
+	if(counter_mask >> 32 & 0x4) {
+		value = rdmsr(MSR_PERF_FIXED_CTRL);
+		value &= ~(0xf << 8);
+		wrmsr(MSR_PERF_FIXED_CTRL, value);
+	}
+
+	return 0;
+}
+
+// init for fixed counter
+int ihk_mc_perfctr_fixed_init(int counter, int mode)
+{
+	unsigned long value = 0;
+	unsigned int  ctr_mask = 0xf;
+	int counter_idx = counter - X86_IA32_BASE_FIXED_PERF_COUNTERS ;
+	unsigned int  set_val = 0;
+
+	if (counter_idx < 0 || counter_idx >= X86_IA32_NUM_FIXED_PERF_COUNTERS) {
+		return -EINVAL;
+	}
+
+	// clear specified fixed counter info
+	value = rdmsr(MSR_PERF_FIXED_CTRL);
+	ctr_mask <<= counter_idx * 4;
+	value &= ~ctr_mask;
+	
+	if (mode & PERFCTR_USER_MODE) {
+		set_val |= 1 << 1;
+	}
+	if (mode & PERFCTR_KERNEL_MODE) {
+		set_val |= 1;
+	}
+
+	// enable PMI on overflow
+	set_val |= 1 << 3;
+
+	set_val <<= counter_idx * 4;
+	value |= set_val;
+
+	wrmsr(MSR_PERF_FIXED_CTRL, value);
+
 	return 0;
 }
 
 int ihk_mc_perfctr_reset(int counter)
 {
-	if (counter < 0 || counter >= X86_IA32_NUM_PERF_COUNTERS) {
-		return -EINVAL;
-	}
+	return set_pmc_x86_direct(counter, 0);
+}
 
-	wrmsr(MSR_IA32_PMC0 + counter, 0);
-
-	return 0;
+int ihk_mc_perfctr_set(int counter, long val)
+{
+	return set_pmc_x86_direct(counter, val);
 }
 
 int ihk_mc_perfctr_read_mask(unsigned long counter_mask, unsigned long *value)
@@ -129,10 +308,87 @@ int ihk_mc_perfctr_read_mask(unsigned long counter_mask, unsigned long *value)
 
 unsigned long ihk_mc_perfctr_read(int counter)
 {
-	if (counter < 0 || counter >= X86_IA32_NUM_PERF_COUNTERS) {
+	unsigned long retval = 0;
+	unsigned long cnt_bit = 0;
+
+	if (counter < 0) {
 		return -EINVAL;
 	}
 
-	return rdpmc(counter);
+	cnt_bit = 1UL << counter;
+
+	if ( cnt_bit & X86_IA32_PERF_COUNTERS_MASK ) {
+		// read generic pmc
+		retval = rdpmc(counter);
+	}
+	else if ( cnt_bit & X86_IA32_FIXED_PERF_COUNTERS_MASK ) {
+		// read fixed pmc
+		retval = rdpmc((1 << 30) + (counter - X86_IA32_BASE_FIXED_PERF_COUNTERS));
+	}
+	else {
+		retval = -EINVAL;
+	}
+
+	return retval;
 }
 
+// read by rdmsr
+unsigned long ihk_mc_perfctr_read_msr(int counter)
+{
+	unsigned int idx = 0;
+	unsigned long retval = 0;
+	unsigned long cnt_bit = 0;
+
+	if (counter < 0) {
+		return -EINVAL;
+	}
+
+	cnt_bit = 1UL << counter;
+
+	if ( cnt_bit & X86_IA32_PERF_COUNTERS_MASK ) {
+		// read generic pmc
+		idx = MSR_IA32_PMC0 + counter;
+		retval = (unsigned long) rdmsr(idx);
+	}
+	else if ( cnt_bit & X86_IA32_FIXED_PERF_COUNTERS_MASK ) {
+		// read fixed pmc
+		idx = MSR_IA32_FIXED_CTR0 + counter;
+		retval = (unsigned long) rdmsr(idx);
+	}
+	else {
+		retval = -EINVAL;
+	}
+
+	return retval;
+}
+
+int ihk_mc_perfctr_alloc_counter(unsigned int *type, unsigned long *config, unsigned long pmc_status)
+{
+	int ret = -1;
+	int i = 0;
+
+	if(*type == PERF_TYPE_HARDWARE) {
+		switch(*config){
+		case PERF_COUNT_HW_INSTRUCTIONS :
+			*type = PERF_TYPE_RAW;
+			*config = 0x5300c0;
+			break;
+		default :
+			// Unexpected config
+			return -1;
+		}
+	}
+	else if(*type != PERF_TYPE_RAW) {
+		return -1;
+	}
+
+	// find avail generic counter
+       	for(i = 0; i < X86_IA32_NUM_PERF_COUNTERS; i++) {
+		if(!(pmc_status & (1 << i))) {
+			ret = i;
+			break;
+		}
+	}
+
+	return ret;
+}

arch/x86/kernel/vsyscall.c

@@ -18,20 +18,93 @@
  */
 
 #include <syscall.h>
+#include <ihk/atomic.h>
+#include <arch/cpu.h>
 
-extern int vsyscall_gettimeofday(void *tv, void *tz)
+extern int vsyscall_gettimeofday(struct timeval *tv, void *tz)
 	__attribute__ ((section (".vsyscall.gettimeofday")));
 
-int vsyscall_gettimeofday(void *tv, void *tz)
+struct tod_data_s tod_data
+		__attribute__ ((section(".vsyscall.gettimeofday.data"))) = {
+	.do_local =	0,
+	.version =	IHK_ATOMIC64_INIT(0),
+};
+
+static inline void cpu_pause_for_vsyscall(void)
+{
+	asm volatile ("pause" ::: "memory");
+	return;
+} /* cpu_pause_for_vsyscall() */
+
+static inline void calculate_time_from_tsc(struct timespec *ts)
+{
+	long ver;
+	unsigned long current_tsc;
+	__time_t sec_delta;
+	long ns_delta;
+
+	for (;;) {
+		while ((ver = ihk_atomic64_read(&tod_data.version)) & 1) {
+			/* settimeofday() is in progress */
+			cpu_pause_for_vsyscall();
+		}
+		rmb();
+		*ts = tod_data.origin;
+		rmb();
+		if (ver == ihk_atomic64_read(&tod_data.version)) {
+			break;
+		}
+
+		/* settimeofday() has intervened */
+		cpu_pause_for_vsyscall();
+	}
+
+	current_tsc = rdtsc();
+	sec_delta = current_tsc / tod_data.clocks_per_sec;
+	ns_delta = NS_PER_SEC * (current_tsc % tod_data.clocks_per_sec)
+		/ tod_data.clocks_per_sec;
+	/* calc. of ns_delta overflows if clocks_per_sec exceeds 18.44 GHz */
+
+	ts->tv_sec += sec_delta;
+	ts->tv_nsec += ns_delta;
+	if (ts->tv_nsec >= NS_PER_SEC) {
+		ts->tv_nsec -= NS_PER_SEC;
+		++ts->tv_sec;
+	}
+
+	return;
+} /* calculate_time_from_tsc() */
+
+int vsyscall_gettimeofday(struct timeval *tv, void *tz)
 {
 	int error;
+	struct timespec ats;
 
+	if (!tv && !tz) {
+		/* nothing to do */
+		return 0;
+	}
+
+	/* Do it locally if supported */
+	if (!tz && tod_data.do_local) {
+		calculate_time_from_tsc(&ats);
+
+		tv->tv_sec = ats.tv_sec;
+		tv->tv_usec = ats.tv_nsec / 1000;
+
+		return 0;
+	}
+
+	/* Otherwise syscall */
 	asm ("syscall" : "=a" (error)
 			: "a" (__NR_gettimeofday), "D" (tv), "S" (tz)
 			: "%rcx", "%r11", "memory");
 
+	if (error) {
+		*(int *)0 = 0;	/* i.e. raise(SIGSEGV) */
+	}
 	return error;
-}
+} /* vsyscall_gettimeofday() */
 
 extern long vsyscall_time(void *tp)
 	__attribute__ ((section (".vsyscall.time")));

arch/x86/tools/irqbalance_mck.in.in

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

arch/x86/tools/irqbalance_mck.service.in

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

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

@@ -3,77 +3,461 @@
 # IHK SMP-x86 example boot script.
 # author: Balazs Gerofi <bgerofi@riken.jp>
 #      Copyright (C) 2014  RIKEN AICS
-# 
-# This is an example script for loading IHK, configuring a partition and 
-# booting McKernel on it. 
-# The script reserves half of the CPU cores and 512MB of RAM from NUMA node 0 
-# when IHK is loaded for the first time, otherwise it destroys the current 
-# McKernel instance and reboots it using the same set of resources as it used
-# previously. 
+#
+# This is an example script for loading IHK, configuring a partition and
+# booting McKernel on it.  Unless specific CPUs and memory are requested,
+# the script reserves half of the CPU cores and 512MB of RAM from
+# NUMA node 0 when IHK is loaded for the first time.
+# Otherwise, it destroys the current McKernel instance and reboots it using
+# the same set of resources as it used previously.
 # Note that the script does not output anything unless an error occurs.
 
 prefix="@prefix@"
-BINDIR="@BINDIR@"
-SBINDIR="@SBINDIR@"
-KMODDIR="@KMODDIR@"
-KERNDIR="@KERNDIR@"
+BINDIR="${prefix}/bin"
+SBINDIR="${prefix}/sbin"
+ETCDIR=@ETCDIR@
+KMODDIR="${prefix}/kmod"
+KERNDIR="${prefix}/@TARGET@/kernel"
+ENABLE_MCOVERLAYFS="@ENABLE_MCOVERLAYFS@"
 
 mem="512M@0"
+cpus=""
 
-# Get the number of CPUs on NUMA node 0
-nr_cpus=`lscpu --parse | awk -F"," '{if ($4 == 0) print $4}' | wc -l`
+if [ "${BASH_VERSINFO[0]}" -lt 4 ]; then
+	echo "You need at least bash-4.0 to run this script." >&2
+	exit 1
+fi
 
-# Use the second half of the cores
-let nr_cpus="$nr_cpus / 2"
-cpus=`lscpu --parse | awk -F"," '{if ($4 == 0) print $1}' | tail -n $nr_cpus | xargs echo -n | sed 's/ /,/g'`
-if [ "$cpus" == "" ]; then echo "error: no available CPUs on NUMA node 0?"; exit; fi
+INTERVAL=1
+LOGMODE=0
+facility="LOG_LOCAL6"
+chown_option=`logname 2> /dev/null`
 
-# Remove delegator if loaded
-if [ "`lsmod | grep mcctrl`" != "" ]; then
-        if ! rmmod mcctrl; then echo "error: removing mcctrl"; exit; fi
+if [ "`systemctl status irqbalance_mck.service 2> /dev/null |grep -E 'Active: active'`" != "" -o "`systemctl status irqbalance.service 2> /dev/null |grep -E 'Active: active'`" != "" ]; then
+	irqbalance_used="yes"
+else
+	irqbalance_used="no"
+fi
+
+while getopts :i:k:c:m:o:f: OPT
+do
+	case ${OPT} in
+	f)	facility=${OPTARG}
+		;;
+	o)	chown_option=${OPTARG}
+		;;
+	i)	INTERVAL=${OPTARG}
+		expr "${INTERVAL}" + 1 > /dev/null 2>&1
+		if [ $? -ge 2 ]
+		then
+			echo "invalid -i value" >&2
+			exit 1
+		fi
+		if [ ${INTERVAL} -le 0 ]
+		then
+			echo "invalid -i value" >&2
+			exit 1
+		fi
+		;;
+	k)	LOGMODE=${OPTARG}
+		expr "${LOGMODE}" + 1 > /dev/null 2>&1
+		if [ $? -ge 2 ]
+		then
+			echo "invalid -k value" >&2
+			exit 1
+		fi
+		if [ ${LOGMODE} -lt 0 -o ${LOGMODE} -gt 2 ]
+		then
+			echo "invalid -k value" >&2
+			exit 1
+		fi
+		;;
+	c) cpus=${OPTARG}
+		;;
+	m) mem=${OPTARG}
+		;;
+	*)  echo "invalid option -${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`
+major=`echo ${release} | sed -e 's/^\([0-9]*\).*/\1/'`
+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
+
+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
+	else
+		if [ ${linux_version_code} -eq 199168 -a ${rhel_release} -ge 327 ]; then
+			enable_mcoverlay="yes"
+		fi
+	fi
+fi
+
+# Figure out CPUs if not requested by user
+if [ "$cpus" == "" ]; then
+	# Get the number of CPUs on NUMA node 0
+	nr_cpus=`lscpu --parse | awk -F"," '{if ($4 == 0) print $4}' | wc -l`
+
+	# Use the second half of the cores
+	let nr_cpus="$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
+fi
+
+# Remove mcoverlay if loaded
+if [ "$enable_mcoverlay" == "yes" ]; then
+	if [ "`lsmod | grep mcoverlay`" != "" ]; then
+		if [ "`cat /proc/mounts | grep /tmp/mcos/mcos0_sys`" != "" ]; then umount -l /tmp/mcos/mcos0_sys; fi
+		if [ "`cat /proc/mounts | grep /tmp/mcos/mcos0_proc`" != "" ]; then umount -l /tmp/mcos/mcos0_proc; fi
+		if [ "`cat /proc/mounts | grep /tmp/mcos/linux_proc`" != "" ]; then umount -l /tmp/mcos/linux_proc; fi
+		if [ "`cat /proc/mounts | grep /tmp/mcos`" != "" ]; then umount -l /tmp/mcos; fi
+		if [ -e /tmp/mcos ]; then rm -rf /tmp/mcos; fi
+		if ! rmmod mcoverlay; then
+			echo "error: removing mcoverlay" >&2
+			error_exit "initial"
+		fi
+	fi
+fi
+
+# Stop irqbalance
+if [ "${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
+		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"; exit; 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 $i`" == "" ]; then
-                        ihk_irq=$i 
-                        break
-                fi
-        done
-        if [ "$ihk_irq" == "" ]; then echo "error: no IRQ available"; exit; fi
-        if ! insmod ${KMODDIR}/ihk-smp-x86.ko ihk_start_irq=$ihk_irq; then echo "error: loading ihk-smp-x86"; exit; fi;
-        if ! ${SBINDIR}/ihkconfig 0 reserve cpu ${cpus}; then echo "error: reserving CPUs"; exit; fi
-        if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then echo "error: reserving memory"; exit; fi
+	ihk_irq=""
+	for i in `seq 64 255`; do
+		if [ ! -d /proc/irq/$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
 
-# Check for existing OS instance and destroy
-if [ -c /dev/mcos0 ]; then
-        # Query CPU cores and memory of OS instance so that the same values are used as previously
-        if ! ${SBINDIR}/ihkosctl 0 query cpu > /dev/null; then echo "error: querying cpus"; exit; fi
-        cpus=`${SBINDIR}/ihkosctl 0 query cpu`
-        if ! ${SBINDIR}/ihkosctl 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
-        mem=`${SBINDIR}/ihkosctl 0 query mem`
-
-        if ! ${SBINDIR}/ihkconfig 0 destroy 0; then echo "warning: destroy failed"; fi	
-else
-        # Otherwise query IHK-SMP for resources
-        if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then echo "error: querying cpus"; exit; fi
-        cpus=`${SBINDIR}/ihkconfig 0 query cpu`
-        if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then echo "error: querying memory"; exit; fi
-        mem=`${SBINDIR}/ihkconfig 0 query mem`	
+# Load mcctrl if not loaded
+if [ "`lsmod | grep mcctrl`" == "" ]; then
+	if ! insmod ${KMODDIR}/mcctrl.ko; then
+		echo "error: inserting mcctrl.ko" >&2
+		error_exit "mem_reserved"
+	fi
+fi
+
+# Destroy all LWK instances
+if ls /dev/mcos* 1>/dev/null 2>&1; then
+	for i in /dev/mcos*; do
+		ind=`echo $i|cut -c10-`;
+		if ! ${SBINDIR}/ihkconfig 0 destroy $ind; then
+			echo "error: destroying LWK instance $ind failed" >&2
+			error_exit "mcctrl_loaded"
+		fi
+	done
+fi
+
+# Create OS instance
+if ! ${SBINDIR}/ihkconfig 0 create; then
+	echo "error: creating OS instance" >&2
+	error_exit "mcctrl_loaded"
+fi
+
+# Assign CPUs
+if ! ${SBINDIR}/ihkosctl 0 assign cpu ${cpus}; then
+	echo "error: assign CPUs" >&2
+	error_exit "os_created"
+fi
+
+# Assign memory
+if ! ${SBINDIR}/ihkosctl 0 assign mem ${mem}; then
+	echo "error: assign memory" >&2
+	error_exit "os_created"
+fi
+
+# Load kernel image
+if ! ${SBINDIR}/ihkosctl 0 load ${KERNDIR}/mckernel.img; then
+	echo "error: loading kernel image: ${KERNDIR}/mckernel.img" >&2
+	error_exit "os_created"
+fi
+
+# Set kernel arguments
+if ! ${SBINDIR}/ihkosctl 0 kargs "hidos ksyslogd=${LOGMODE}"; 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
+		error_exit "tmp_mcos_created"
+	fi
+	if [ ! -e /tmp/mcos/linux_proc ]; then mkdir -p /tmp/mcos/linux_proc; fi
+	if ! mount --bind /proc /tmp/mcos/linux_proc; then
+		echo "error: mount /tmp/mcos/linux_proc" >&2
+		error_exit "tmp_mcos_mounted"
+	fi
+	if ! insmod ${KMODDIR}/mcoverlay.ko; then
+		echo "error: inserting mcoverlay.ko" >&2
+		error_exit "linux_proc_bind_mounted"
+	fi
+	while [ ! -e /proc/mcos0 ]
+	do
+		sleep 1
+	done
+	if [ ! -e /tmp/mcos/mcos0_proc ]; then mkdir -p /tmp/mcos/mcos0_proc; fi
+	if [ ! -e /tmp/mcos/mcos0_proc_upper ]; then mkdir -p /tmp/mcos/mcos0_proc_upper; fi
+	if [ ! -e /tmp/mcos/mcos0_proc_work ]; then mkdir -p /tmp/mcos/mcos0_proc_work; fi
+	if ! mount -t mcoverlay mcoverlay -o lowerdir=/proc/mcos0:/proc,upperdir=/tmp/mcos/mcos0_proc_upper,workdir=/tmp/mcos/mcos0_proc_work,nocopyupw,nofscheck /tmp/mcos/mcos0_proc; then
+		echo "error: mounting /tmp/mcos/mcos0_proc" >&2
+		error_exit "mcoverlayfs_loaded"
+	fi
+	# TODO: How de we revert this in case of failure??
+	mount --make-rprivate /proc
+
+	while [ ! -e /sys/devices/virtual/mcos/mcos0/sys/setup_complete ]
+	do
+		sleep 0.1
+	done
+	if [ ! -e /tmp/mcos/mcos0_sys ]; then mkdir -p /tmp/mcos/mcos0_sys; fi
+	if [ ! -e /tmp/mcos/mcos0_sys_upper ]; then mkdir -p /tmp/mcos/mcos0_sys_upper; fi
+	if [ ! -e /tmp/mcos/mcos0_sys_work ]; then mkdir -p /tmp/mcos/mcos0_sys_work; fi
+	if ! mount -t mcoverlay mcoverlay -o lowerdir=/sys/devices/virtual/mcos/mcos0/sys:/sys,upperdir=/tmp/mcos/mcos0_sys_upper,workdir=/tmp/mcos/mcos0_sys_work,nocopyupw,nofscheck /tmp/mcos/mcos0_sys; then
+		echo "error: mount /tmp/mcos/mcos0_sys" >&2
+		error_exit "mcos_proc_mounted"
+	fi
+	# TODO: How de we revert this in case of failure??
+	mount --make-rprivate /sys
+	rm -rf /tmp/mcos/mcos0_sys/setup_complete
+
+	# Hide NUMA related files which are outside the LWK partition
+	for cpuid in `find /sys/devices/system/cpu/* -maxdepth 0 -name "cpu[0123456789]*" -printf "%f "`; do
+		if [ ! -e "/sys/devices/virtual/mcos/mcos0/sys/devices/system/cpu/$cpuid" ]; then
+			rm -rf /tmp/mcos/mcos0_sys/devices/system/cpu/$cpuid
+		else
+			for nodeid in `find /sys/devices/system/cpu/$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
+		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
+	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
+
+# Start irqbalance with CPUs and IRQ for McKernel banned
+if [ "${irqbalance_used}" == "yes" ]; then
+    if ! etcdir=@ETCDIR@ perl -e 'use File::Copy qw(copy); $etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "/proc/irq/*/smp_affinity"; foreach $file (@files) { $rel = substr($file, 1); $dir=substr($rel, 0, length($rel)-length("/smp_affinity")); if(0) { print "cp $file $etcdir/$rel\n";} if(system("mkdir -p $etcdir/$dir")){ exit 1;} if(!copy($file,"$etcdir/$rel")){ exit 1;} }'; then
+		echo "error: saving /proc/irq/*/smp_affinity" >&2
+		error_exit "mcos_sys_mounted"
+	fi;
+
+    ncpus=`lscpu | grep -E '^CPU\(s\):' | awk '{print $2}'`
+    smp_affinity_mask=`echo $cpus | ncpus=$ncpus perl -e 'while(<>){@tokens = split /,/;foreach $token (@tokens) {@nums = split /-/,$token; for($num = $nums[0]; $num <= $nums[$#nums]; $num++) {$ndx=int($num/32); $mask[$ndx] |= (1<<($num % 32))}}} $nint32s = int(($ENV{'ncpus'}+31)/32); for($j = $nint32s - 1; $j >= 0; $j--) { if($j != $nint32s - 1){print ",";} $nblks = $j == $nint32s - 1 ? int(($ENV{'ncpus'} % 32)/4) : 8; for($i = $nblks - 1;$i >= 0;$i--){ printf("%01x",($mask[$j] >> ($i*4)) & 0xf);}}'`
+
+    if ! ncpus=$ncpus smp_affinity_mask=$smp_affinity_mask perl -e '@dirs = grep { -d } glob "/proc/irq/*"; foreach $dir (@dirs) { $hit = 0; $affinity_str = `cat $dir/smp_affinity`; chomp $affinity_str; @int32strs = split /,/, $affinity_str; @int32strs_mask=split /,/, $ENV{'smp_affinity_mask'}; for($i=0;$i <= $#int32strs_mask; $i++) { $int32strs_inv[$i] = sprintf("%08x",hex($int32strs_mask[$i])^0xffffffff); if($i == 0) { $len = int((($ENV{'ncpus'}%32)+3)/4); $int32strs_inv[$i] = substr($int32strs_inv[$i], -$len, $len); } } $inv = join(",", @int32strs_inv); $nint32s = int(($ENV{'ncpus'}+31)/32); for($j = $nint32s - 1; $j >= 0; $j--) { if(hex($int32strs[$nint32s - 1 - $j]) & hex($int32strs_mask[$nint32s - 1 - $j])) { $hit = 1; }} if($hit == 1) { $cmd = "echo $inv > $dir/smp_affinity 2>/dev/null"; system $cmd;}}'; then
+		echo "error: modifying /proc/irq/*/smp_affinity" >&2
+		error_exit "mcos_sys_mounted"
+	fi
+
+    banirq=`cat /proc/interrupts| perl -e 'while(<>) { if(/^\s*(\d+).*IHK\-SMP\s*$/) {print $1;}}'`
+
+    sed "s/%mask%/$smp_affinity_mask/g" $ETCDIR/irqbalance_mck.in | sed "s/%banirq%/$banirq/g" > $ETCDIR/irqbalance_mck
+    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
 
-if ! ${SBINDIR}/ihkconfig 0 create; then echo "error: create"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 assign cpu ${cpus}; then echo "error: assign CPUs"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 assign mem ${mem}; then echo "error: assign memory"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 load ${KERNDIR}/mckernel.img; then echo "error: loading kernel image"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 kargs hidos; then echo "error: setting kernel arguments"; exit; fi
-if ! ${SBINDIR}/ihkosctl 0 boot; then echo "error: booting"; exit; fi
-if ! insmod ${KMODDIR}/mcctrl.ko; then echo "error: inserting mcctrl.ko"; exit; fi

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

@@ -0,0 +1,115 @@
+#!/bin/bash
+
+# IHK SMP-x86 example McKernel unload script.
+# author: Balazs Gerofi <bgerofi@riken.jp>
+#      Copyright (C) 2015  RIKEN AICS
+# 
+# This is an example script for destroying McKernel and releasing IHK resources
+# Note that the script does no output anything unless an error occurs.
+
+prefix="@prefix@"
+BINDIR="@BINDIR@"
+SBINDIR="@SBINDIR@"
+ETCDIR=@ETCDIR@
+KMODDIR="@KMODDIR@"
+KERNDIR="@KERNDIR@"
+
+mem=""
+cpus=""
+
+# No SMP module? Exit.
+if [ "`lsmod | grep ihk_smp_x86`" == "" ]; then exit 0; fi
+
+# Destroy all LWK instances
+if ls /dev/mcos* 1>/dev/null 2>&1; then
+	for i in /dev/mcos*; do
+		ind=`echo $i|cut -c10-`;
+		if ! ${SBINDIR}/ihkconfig 0 destroy $ind; then
+			echo "error: destroying LWK instance $ind failed" >&2
+			exit 1
+		fi
+	done
+fi
+
+# Query IHK-SMP resources and release them
+if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then
+	echo "error: querying cpus" >&2
+	exit 1
+fi
+
+cpus=`${SBINDIR}/ihkconfig 0 query cpu`
+if [ "${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
+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
+fi
+
+# Remove core module
+if [ "`lsmod | grep -E 'ihk\s' | awk '{print $1}'`" != "" ]; then
+	if ! rmmod ihk; then
+		echo "error: removing ihk" >&2
+		exit 1
+	fi
+fi
+
+# Stop mcklogd
+pkill mcklogd
+
+# Start irqbalance with the original settings
+if [ "`systemctl status irqbalance_mck.service 2> /dev/null |grep -E 'Active: active'`" != "" ]; then
+	if ! systemctl stop irqbalance_mck.service 2>/dev/null; then
+		echo "warning: failed to stop irqbalance_mck" >&2
+	fi
+	if ! systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null; then
+		echo "warning: failed to disable irqbalance_mck" >&2
+	fi
+	if ! etcdir=@ETCDIR@ perl -e '$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.ac

@@ -27,10 +27,27 @@ AC_ARG_WITH([target],
     [--with-target={attached-mic | builtin-mic | builtin-x86 | smp-x86}],[target, default is attached-mic]),
   [WITH_TARGET=$withval],[WITH_TARGET=yes])
 
+AC_ARG_WITH([system_map],
+  AS_HELP_STRING(
+    [--with-system_map=path],[Path to 'System.map file', default is /boot/System.map-uname_r]),
+  [WITH_SYSTEM_MAP=$withval],[WITH_SYSTEM_MAP=yes])
+
 AC_ARG_ENABLE([dcfa],
   [AS_HELP_STRING(
   	[--enable-dcfa],[Enable DCFA modules])],[],[enable_dcfa=no])
 
+AC_ARG_ENABLE([memdump],
+	      AC_HELP_STRING([--enable-memdump],
+			     [enable dumping memory and analyzing a dump]),
+	      [ENABLE_MEMDUMP=$enableval],
+	      [ENABLE_MEMDUMP=default])
+
+AC_ARG_ENABLE([mcoverlayfs],
+	      AC_HELP_STRING([--enable-mcoverlayfs],
+			     [enable mcoverlayfs implementation]),
+	      [ENABLE_MCOVERLAYFS=$enableval],
+	      [ENABLE_MCOVERLAYFS=yes])
+
 case "X$WITH_KERNELSRC" in
     Xyes | Xno | X)
 	WITH_KERNELSRC='/lib/modules/`uname -r`/build'
@@ -49,9 +66,26 @@ fi
 test "x$prefix" = xNONE && prefix="$ac_default_prefix"
 
 case $WITH_TARGET in
-    attached-mic)
+    attached-mic|builtin-x86|smp-x86)
 	ARCH=`uname -m`
 	AC_PROG_CC
+	XCC=$CC
+	;;
+    builtin-mic)
+	ARCH=k1om
+	AC_CHECK_PROG(XCC,
+		[x86_64-$ARCH-linux-gcc],
+		[x86_64-$ARCH-linux-gcc],
+		[no])
+	CC=$XCC
+	;;
+    *)
+	AC_MSG_ERROR([target $WITH_TARGET is unknwon])
+	;;
+esac
+
+case $WITH_TARGET in
+    attached-mic)
 	if test "X$KERNDIR" = X; then
 		KERNDIR="$prefix/attached/kernel"
 	fi
@@ -69,12 +103,6 @@ case $WITH_TARGET in
 	fi
 	;;
     builtin-mic)
-	ARCH=k1om
-	AC_CHECK_PROG(XCC,
-		[x86_64-$ARCH-linux-gcc],
-		[x86_64-$ARCH-linux-gcc],
-		[no])
-	CC=$XCC
 	if test "X$KERNDIR" = X; then
 		KERNDIR="$prefix/attached/kernel"
 	fi
@@ -92,9 +120,6 @@ case $WITH_TARGET in
 	fi
 	;;
     builtin-x86)
-	ARCH=`uname -m`
-	AC_PROG_CC
-	XCC=$CC
 	if test "X$KERNDIR" = X; then
 		KERNDIR="$prefix/attached/kernel"
 	fi
@@ -112,9 +137,6 @@ case $WITH_TARGET in
 	fi
 	;;
     smp-x86)
-	ARCH=`uname -m`
-	AC_PROG_CC
-	XCC=$CC
 	if test "X$KERNDIR" = X; then
 		KERNDIR="$prefix/smp-x86/kernel"
 	fi
@@ -124,6 +146,9 @@ case $WITH_TARGET in
 	if test "X$SBINDIR" = X; then
 		SBINDIR="$prefix/sbin"
 	fi
+	if test "X$ETCDIR" = X; then
+		ETCDIR="$prefix/etc"
+	fi
 	if test "X$KMODDIR" = X; then
 		KMODDIR="$prefix/kmod"
 	fi
@@ -139,6 +164,116 @@ esac
 KDIR="$WITH_KERNELSRC"
 TARGET="$WITH_TARGET"
 
+MCCTRL_LINUX_SYMTAB=""
+case "X$WITH_SYSTEM_MAP" in
+    Xyes | Xno | X)
+	MCCTRL_LINUX_SYMTAB=""
+	;;
+    *)
+	MCCTRL_LINUX_SYMTAB="$WITH_SYSTEM_MAP"
+	;;
+esac
+
+AC_MSG_CHECKING([[for System.map]])
+if test -f "$MCCTRL_LINUX_SYMTAB"; then
+	MCCTRL_LINUX_SYMTAB="$MCCTRL_LINUX_SYMTAB"
+elif test -f "/boot/System.map-`uname -r`"; then
+	MCCTRL_LINUX_SYMTAB="/boot/System.map-`uname -r`"
+elif test -f "$KDIR/System.map"; then
+	MCCTRL_LINUX_SYMTAB="$KDIR/System.map"
+fi
+
+if test "$MCCTRL_LINUX_SYMTAB" == ""; then
+	AC_MSG_ERROR([could not find])
+fi
+
+if test -z "`eval cat $MCCTRL_LINUX_SYMTAB`"; then
+	AC_MSG_ERROR([could not read System.map file, no read permission?])
+fi
+AC_MSG_RESULT([$MCCTRL_LINUX_SYMTAB])
+
+MCCTRL_LINUX_SYMTAB_CMD="cat $MCCTRL_LINUX_SYMTAB"
+
+# MCCTRL_FIND_KSYM(SYMBOL)
+# ------------------------------------------------------
+# Search System.map for address of the given symbol and
+# do one of three things in config.h:
+# If not found, leave MCCTRL_KSYM_foo undefined
+# If found to be exported, "#define MCCTRL_KSYM_foo 0"
+# If found not to be exported, "#define MCCTRL_KSYM_foo 0x<value>"
+AC_DEFUN([MCCTRL_FIND_KSYM],[
+  AC_MSG_CHECKING([[System.map for symbol $1]])
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " $1\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    AC_MSG_RESULT([not found])
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+    m4_ifval([$2],[],[
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_$1\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+    ])
+    AC_MSG_RESULT([$mcctrl_result])
+    AC_DEFINE_UNQUOTED(MCCTRL_KSYM_[]$1,$mcctrl_addr,[Define to address of kernel symbol $1, or 0 if exported])
+  fi
+])
+
+MCCTRL_FIND_KSYM([sys_mount])
+MCCTRL_FIND_KSYM([sys_unshare])
+MCCTRL_FIND_KSYM([zap_page_range])
+MCCTRL_FIND_KSYM([vdso_image_64])
+MCCTRL_FIND_KSYM([vdso_start])
+MCCTRL_FIND_KSYM([vdso_end])
+MCCTRL_FIND_KSYM([vdso_pages])
+MCCTRL_FIND_KSYM([__vvar_page])
+MCCTRL_FIND_KSYM([hpet_address])
+MCCTRL_FIND_KSYM([hv_clock])
+MCCTRL_FIND_KSYM([sys_readlink])
+
+case $ENABLE_MEMDUMP in
+	yes|no|auto)
+		;;
+	default)
+		if test "x$WITH_TARGET" = "xsmp-x86" ; then
+			ENABLE_MEMDUMP=auto
+		else
+			ENABLE_MEMDUMP=no
+		fi
+		;;
+	*)
+		AC_MSG_ERROR([unknown memdump argument: $ENABLE_MEMDUMP])
+		;;
+esac
+
+if test "x$ENABLE_MEMDUMP" != "xno" ; then
+	enableval=yes
+	AC_CHECK_LIB([bfd],[bfd_init],[],[enableval=no])
+	AC_CHECK_HEADER([bfd.h],[],[enableval=no])
+
+	if test "x$ENABLE_MEMDUMP" = "xyes" -a "x$enableval" = "xno" ; then
+		AC_MSG_ERROR([memdump feature needs bfd.h and libbfd a.k.a bunutils-devel])
+	fi
+	ENABLE_MEMDUMP=$enableval
+fi
+
+if test "x$ENABLE_MEMDUMP" = "xyes" ; then
+	AC_MSG_NOTICE([memdump feature is enabled])
+	AC_DEFINE([ENABLE_MEMDUMP],[1],[whether memdump feature is enabled])
+	uncomment_if_ENABLE_MEMDUMP=''
+else
+	AC_MSG_NOTICE([memdump feature is disabled])
+	uncomment_if_ENABLE_MEMDUMP='#'
+fi
+
+if test "x$ENABLE_MCOVERLAYFS" = "xyes" ; then
+	AC_DEFINE([ENABLE_MCOVERLAYFS],[1],[whether mcoverlayfs is enabled])
+	AC_MSG_NOTICE([mcoverlayfs is enabled])
+else
+	AC_MSG_NOTICE([mcoverlayfs is disabled])
+fi
+
 AC_SUBST(CC)
 AC_SUBST(XCC)
 AC_SUBST(ARCH)
@@ -146,9 +281,11 @@ AC_SUBST(KDIR)
 AC_SUBST(TARGET)
 AC_SUBST(BINDIR)
 AC_SUBST(SBINDIR)
+AC_SUBST(ETCDIR)
 AC_SUBST(KMODDIR)
 AC_SUBST(KERNDIR)
 AC_SUBST(MANDIR)
+AC_SUBST(ENABLE_MCOVERLAYFS)
 
 AC_SUBST(IHK_VERSION)
 AC_SUBST(MCKERNEL_VERSION)
@@ -156,19 +293,28 @@ AC_SUBST(DCFA_VERSION)
 AC_SUBST(IHK_RELEASE_DATE)
 AC_SUBST(MCKERNEL_RELEASE_DATE)
 AC_SUBST(DCFA_RESEASE_DATE)
+AC_SUBST(uncomment_if_ENABLE_MEMDUMP)
 
+AC_CONFIG_HEADERS([executer/config.h])
 AC_CONFIG_FILES([
 	Makefile
 	executer/user/Makefile
-	executer/kernel/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
 ])
 
 AS_IF([test "x$enable_dcfa" = xyes], [

executer/config.h.in

@@ -0,0 +1,91 @@
+/* executer/config.h.in.  Generated from configure.ac by autoheader.  */
+
+/* whether mcoverlayfs is enabled */
+#undef ENABLE_MCOVERLAYFS
+
+/* whether memdump feature is enabled */
+#undef ENABLE_MEMDUMP
+
+/* Define to 1 if you have the <inttypes.h> header file. */
+#undef HAVE_INTTYPES_H
+
+/* Define to 1 if you have the `bfd' library (-lbfd). */
+#undef HAVE_LIBBFD
+
+/* Define to 1 if you have the <memory.h> header file. */
+#undef HAVE_MEMORY_H
+
+/* Define to 1 if you have the <stdint.h> header file. */
+#undef HAVE_STDINT_H
+
+/* Define to 1 if you have the <stdlib.h> header file. */
+#undef HAVE_STDLIB_H
+
+/* Define to 1 if you have the <strings.h> header file. */
+#undef HAVE_STRINGS_H
+
+/* Define to 1 if you have the <string.h> header file. */
+#undef HAVE_STRING_H
+
+/* Define to 1 if you have the <sys/stat.h> header file. */
+#undef HAVE_SYS_STAT_H
+
+/* Define to 1 if you have the <sys/types.h> header file. */
+#undef HAVE_SYS_TYPES_H
+
+/* Define to 1 if you have the <unistd.h> header file. */
+#undef HAVE_UNISTD_H
+
+/* Define to address of kernel symbol __vvar_page, or 0 if exported */
+#undef MCCTRL_KSYM___vvar_page
+
+/* Define to address of kernel symbol hpet_address, or 0 if exported */
+#undef MCCTRL_KSYM_hpet_address
+
+/* Define to address of kernel symbol hv_clock, or 0 if exported */
+#undef MCCTRL_KSYM_hv_clock
+
+/* Define to address of kernel symbol sys_mount, or 0 if exported */
+#undef MCCTRL_KSYM_sys_mount
+
+/* Define to address of kernel symbol sys_readlink, or 0 if exported */
+#undef MCCTRL_KSYM_sys_readlink
+
+/* Define to address of kernel symbol sys_unshare, or 0 if exported */
+#undef MCCTRL_KSYM_sys_unshare
+
+/* Define to address of kernel symbol vdso_end, or 0 if exported */
+#undef MCCTRL_KSYM_vdso_end
+
+/* Define to address of kernel symbol vdso_image_64, or 0 if exported */
+#undef MCCTRL_KSYM_vdso_image_64
+
+/* Define to address of kernel symbol vdso_pages, or 0 if exported */
+#undef MCCTRL_KSYM_vdso_pages
+
+/* Define to address of kernel symbol vdso_start, or 0 if exported */
+#undef MCCTRL_KSYM_vdso_start
+
+/* Define to address of kernel symbol zap_page_range, or 0 if exported */
+#undef MCCTRL_KSYM_zap_page_range
+
+/* Define to the address where bug reports for this package should be sent. */
+#undef PACKAGE_BUGREPORT
+
+/* Define to the full name of this package. */
+#undef PACKAGE_NAME
+
+/* Define to the full name and version of this package. */
+#undef PACKAGE_STRING
+
+/* Define to the one symbol short name of this package. */
+#undef PACKAGE_TARNAME
+
+/* Define to the home page for this package. */
+#undef PACKAGE_URL
+
+/* Define to the version of this package. */
+#undef PACKAGE_VERSION
+
+/* Define to 1 if you have the ANSI C header files. */
+#undef STDC_HEADERS

executer/include/uprotocol.h

@@ -48,6 +48,9 @@
 #define MCEXEC_UP_OPEN_EXEC      0x30a02912
 #define MCEXEC_UP_CLOSE_EXEC     0x30a02913
 
+#define MCEXEC_UP_SYS_MOUNT      0x30a02914
+#define MCEXEC_UP_SYS_UNSHARE    0x30a02915
+
 #define MCEXEC_UP_DEBUG_LOG     0x40000000
 
 #define MCEXEC_UP_TRANSFER_TO_REMOTE	0
@@ -83,6 +86,9 @@ struct program_load_desc {
 	int stack_prot;
 	int pgid;
 	int cred[8];
+	int reloc;
+	char enable_vdso;
+	char padding[7];
 	unsigned long entry;
 	unsigned long user_start;
 	unsigned long user_end;
@@ -104,6 +110,13 @@ struct program_load_desc {
 };
 
 struct syscall_request {
+	/* TID of requesting thread */
+	int rtid;
+	/*
+	 * TID of target thread. Remote page fault response needs to designate the
+	 * thread that must serve the request, 0 indicates any thread from the pool
+	 */
+	int ttid;
 	unsigned long valid;
 	unsigned long number;
 	unsigned long args[6];
@@ -122,8 +135,17 @@ struct syscall_load_desc {
 	unsigned long size;
 };
 
+#define IHK_SCD_REQ_THREAD_SPINNING         0
+#define IHK_SCD_REQ_THREAD_TO_BE_WOKEN      1
+#define IHK_SCD_REQ_THREAD_DESCHEDULED      2
+
 struct syscall_response {
+	/* TID of the thread that requested the service */
+	int ttid;
+	/* TID of the mcexec thread that is serving or has served the request */
+	int stid;
 	unsigned long status;
+	unsigned long req_thread_status;
 	long ret;
 	unsigned long fault_address;
 	unsigned long fault_reason;
@@ -166,4 +188,16 @@ struct newprocess_desc {
 	int pid;
 };
 
+struct sys_mount_desc {
+	char *dev_name;
+	char *dir_name;
+	char *type;
+	unsigned long flags;
+	void *data;
+};
+
+struct sys_unshare_desc {
+	unsigned long unshare_flags;
+};
+
 #endif

executer/kernel/Makefile.in

@@ -1,26 +0,0 @@
-KDIR ?= @KDIR@
-ARCH ?= @ARCH@
-src = @abs_srcdir@
-KMODDIR=@KMODDIR@
-BINDIR=@BINDIR@
-IHK_BASE=$(src)/../../../ihk
-
-obj-m += mcctrl.o
-
-ccflags-y := -I$(IHK_BASE)/linux/include -I$(IHK_BASE)/ikc/include -I$(IHK_BASE)/include -I$(src)/../include -mcmodel=kernel -mno-red-zone -DMCEXEC_PATH=\"$(BINDIR)/mcexec\"
-
-mcctrl-y := driver.o control.o ikc.o syscall.o procfs.o binfmt_mcexec.o
-
-KBUILD_EXTRA_SYMBOLS = @abs_builddir@/../../../ihk/linux/core/Module.symvers
-
-.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 mcctrl.ko $(KMODDIR)

executer/kernel/mcctrl.h

@@ -1,191 +0,0 @@
-/**
- * \file mcctrl.h
- *  License details are found in the file LICENSE.
- * \brief
- *  define data structure
- * \author Taku Shimosawa  <shimosawa@is.s.u-tokyo.ac.jp> \par
- *      Copyright (C) 2011 - 2012  Taku Shimosawa
- * \author Balazs Gerofi  <bgerofi@riken.jp> \par
- *      Copyright (C) 2012  RIKEN AICS
- * \author Gou Nakamura  <go.nakamura.yw@hitachi-solutions.com> \par
- *      Copyright (C) 2012 - 2013 Hitachi, Ltd.
- * \author Tomoki Shirasawa  <tomoki.shirasawa.kk@hitachi-solutions.com> \par
- *      Copyright (C) 2012 - 2013 Hitachi, Ltd.
- * \author Balazs Gerofi  <bgerofi@is.s.u-tokyo.ac.jp> \par
- *      Copyright (C) 2013  The University of Tokyo
- */
-/*
- * HISTORY:
- *  2013/11/07 hamada added <sys/resource.h> which is required by getrlimit(2)
- *  2013/10/21 nakamura exclude interpreter's segment from data region
- *  2013/10/11 nakamura mcexec: add a upper limit of the stack size
- *  2013/10/11 nakamura mcexec: add a path prefix for interpreter search
- *  2013/10/11 nakamura mcexec: add a interpreter invocation
- *  2013/10/08 nakamura add a AT_ENTRY entry to the auxiliary vector
- *  2013/09/02 shirasawa add terminate thread
- *  2013/08/19 shirasawa mcexec forward signal to MIC process
- *  2013/08/07 nakamura add page fault forwarding
- *  2013/07/26 shirasawa mcexec print signum or exit status
- *  2013/07/17 nakamura create more mcexec thread so that all cpu to be serviced
- *  2013/04/17 nakamura add generic system call forwarding
- */
-#ifndef HEADER_MCCTRL_H
-#define HEADER_MCCTRL_H
-
-#include <ihk/ihk_host_driver.h>
-#include <uprotocol.h>
-#include <linux/wait.h>
-#include <ihk/ikc.h>
-#include <ikc/master.h>
-
-#define SCD_MSG_PREPARE_PROCESS         0x1
-#define SCD_MSG_PREPARE_PROCESS_ACKED   0x2
-#define SCD_MSG_PREPARE_PROCESS_NACKED  0x7
-#define SCD_MSG_SCHEDULE_PROCESS        0x3
-
-#define SCD_MSG_INIT_CHANNEL            0x5
-#define SCD_MSG_INIT_CHANNEL_ACKED      0x6
-
-#define SCD_MSG_SYSCALL_ONESIDE         0x4
-#define SCD_MSG_SEND_SIGNAL     	0x8
-#define SCD_MSG_CLEANUP_PROCESS         0x9
-
-#define	SCD_MSG_PROCFS_CREATE		0x10
-#define	SCD_MSG_PROCFS_DELETE		0x11
-#define	SCD_MSG_PROCFS_REQUEST		0x12
-#define	SCD_MSG_PROCFS_ANSWER		0x13
-
-#define SCD_MSG_DEBUG_LOG		0x20
-
-#define DMA_PIN_SHIFT                   21
-
-#define DO_USER_MODE
-
-#define	__NR_coredump			999
-
-struct coretable {
-	int len;
-	unsigned long addr;
-};
-
-struct ikc_scd_packet {
-	int msg;
-	int ref;
-	int osnum;
-	int pid;
-	int err;
-	unsigned long arg;
-};
-
-struct mcctrl_priv { 
-	ihk_os_t os;
-	struct program_load_desc *desc;
-};
-
-struct ikc_scd_init_param {
-	unsigned long request_page;
-	unsigned long response_page;
-	unsigned long doorbell_page;
-	unsigned long post_page;
-};
-
-struct syscall_post {
-	unsigned long v[8];
-};
-
-struct syscall_params {
-	unsigned long request_pa;
-	struct syscall_request *request_va;
-	unsigned long response_rpa, response_pa;
-	struct syscall_response *response_va;
-	unsigned long post_pa;
-	struct syscall_post *post_va;
-	
-	unsigned long doorbell_pa;
-	unsigned long *doorbell_va;
-};
-
-struct wait_queue_head_list_node {
-	struct list_head list;
-	wait_queue_head_t wq_syscall;
-	int pid;
-	int req;
-};
-
-struct mcctrl_channel {
-	struct ihk_ikc_channel_desc *c;
-	struct syscall_params param;
-	struct ikc_scd_init_param init;
-	void *dma_buf;
-
-	struct list_head wq_list;
-	ihk_spinlock_t wq_list_lock;
-};
-
-struct mcctrl_per_proc_data {
-	struct list_head list;
-	int pid;
-	unsigned long rpgtable;	/* per process, not per OS */
-};
-
-struct mcctrl_usrdata {
-	struct ihk_ikc_listen_param listen_param;
-	struct ihk_ikc_listen_param listen_param2;
-	ihk_os_t	os;
-	int	num_channels;
-	struct mcctrl_channel	*channels;
-	unsigned long	*mcctrl_doorbell_va;
-	unsigned long	mcctrl_doorbell_pa;
-	int	remaining_job;
-	int	base_cpu;
-	int	job_pos;
-	int	mcctrl_dma_abort;
-	unsigned long	last_thread_exec;
-	wait_queue_head_t	wq_prepare;
-	
-	struct list_head per_proc_list;
-	ihk_spinlock_t per_proc_list_lock;
-	void **keys;
-};
-
-struct mcctrl_signal {
-	int	cond;
-	int	sig;
-	int	pid;
-	int	tid;
-	char	info[128];
-};
-
-int mcctrl_ikc_send(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp);
-int mcctrl_ikc_send_msg(ihk_os_t os, int cpu, int msg, int ref, unsigned long arg);
-int mcctrl_ikc_is_valid_thread(ihk_os_t os, int cpu);
-int reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp,
-		unsigned long *endp);
-
-/* syscall.c */
-int init_peer_channel_registry(struct mcctrl_usrdata *ud);
-int register_peer_channel(struct mcctrl_usrdata *ud, void *key, struct mcctrl_channel *ch);
-int deregister_peer_channel(struct mcctrl_usrdata *ud, void *key, struct mcctrl_channel *ch);
-struct mcctrl_channel *get_peer_channel(struct mcctrl_usrdata *ud, void *key);
-int __do_in_kernel_syscall(ihk_os_t os, struct mcctrl_channel *c, struct syscall_request *sc);
-
-#define PROCFS_NAME_MAX 1000
-
-struct procfs_read {
-	unsigned long pbuf;	/* physical address of the host buffer (request) */
-	unsigned long offset;	/* offset to read (request) */
-	int count;		/* bytes to read (request) */
-	int eof;		/* if eof is detected, 1 otherwise 0. (answer)*/
-	int ret;		/* read bytes (answer) */
-	int status;		/* non-zero if done (answer) */
-	int newcpu;		/* migrated new cpu (answer) */
-	char fname[PROCFS_NAME_MAX];	/* procfs filename (request) */
-};
-
-struct procfs_file {
-	int status;			/* status of processing (answer) */
-	int mode;			/* file mode (request) */
-	char fname[PROCFS_NAME_MAX];	/* procfs filename (request) */
-};
-
-#endif

executer/kernel/mcctrl/Makefile.in

@@ -0,0 +1,27 @@
+KDIR ?= @KDIR@
+ARCH ?= @ARCH@
+src = @abs_srcdir@
+KMODDIR=@KMODDIR@
+BINDIR=@BINDIR@
+IHK_BASE=$(src)/../../../../ihk
+
+obj-m += mcctrl.o
+
+ccflags-y := -I$(IHK_BASE)/linux/include -I$(IHK_BASE)/linux/include/ihk/arch/$(ARCH) -I$(IHK_BASE)/ikc/include -I$(IHK_BASE)/ikc/include/ikc/arch/$(ARCH) -I$(IHK_BASE)/include -I$(IHK_BASE)/include/arch/$(ARCH) -I$(src)/../../include -mcmodel=kernel -mno-red-zone -DMCEXEC_PATH=\"$(BINDIR)/mcexec\" -I@abs_builddir@
+
+mcctrl-y := driver.o control.o ikc.o syscall.o procfs.o binfmt_mcexec.o
+mcctrl-y += sysfs.o sysfs_files.o arch/$(ARCH)/archdeps.o
+
+KBUILD_EXTRA_SYMBOLS = @abs_builddir@/../../../../ihk/linux/core/Module.symvers
+
+.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 mcctrl.ko $(KMODDIR)

executer/kernel/mcctrl/arch/x86_64/Makefile.in

@@ -0,0 +1 @@
+# dummy file

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

@@ -0,0 +1,192 @@
+#include <linux/version.h>
+#include "../../config.h"
+#include "../../mcctrl.h"
+
+#ifdef MCCTRL_KSYM_vdso_image_64
+#if MCCTRL_KSYM_vdso_image_64
+struct vdso_image *vdso_image = (void *)MCCTRL_KSYM_vdso_image_64;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM_vdso_start
+#if MCCTRL_KSYM_vdso_start
+void *vdso_start = (void *)MCCTRL_KSYM_vdso_start;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM_vdso_end
+#if MCCTRL_KSYM_vdso_end
+void *vdso_end = (void *)MCCTRL_KSYM_vdso_end;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM_vdso_pages
+#if MCCTRL_KSYM_vdso_pages
+struct page **vdso_pages = (void *)MCCTRL_KSYM_vdso_pages;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM___vvar_page
+#if MCCTRL_KSYM___vvar_page
+void *__vvar_page = (void *)MCCTRL_KSYM___vvar_page;
+#endif
+#endif
+
+long *hpet_addressp
+#ifdef MCCTRL_KSYM_hpet_address
+#if MCCTRL_KSYM_hpet_address
+	= (void *)MCCTRL_KSYM_hpet_address;
+#else
+	= &hpet_address;
+#endif
+#else
+	= NULL;
+#endif
+
+void **hv_clockp
+#ifdef MCCTRL_KSYM_hv_clock
+#if MCCTRL_KSYM_hv_clock
+	= (void *)MCCTRL_KSYM_hv_clock;
+#else
+	= &hv_clock;
+#endif
+#else
+	= NULL;
+#endif
+
+unsigned long
+reserve_user_space_common(struct mcctrl_usrdata *usrdata, unsigned long start, unsigned long end);
+
+int
+reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp, unsigned long *endp)
+{
+	struct vm_area_struct *vma;
+	unsigned long start = 0L;
+	unsigned long end;
+
+#define	DESIRED_USER_END	0x800000000000
+#define	GAP_FOR_MCEXEC		0x008000000000UL
+	end = DESIRED_USER_END;
+	down_write(&current->mm->mmap_sem);
+	vma = find_vma(current->mm, 0);
+	if (vma) {
+		end = (vma->vm_start - GAP_FOR_MCEXEC) & ~(GAP_FOR_MCEXEC - 1);
+	}
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3,5,0)
+	up_write(&current->mm->mmap_sem);
+#endif
+	start = reserve_user_space_common(usrdata, start, end);
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,5,0)
+	up_write(&current->mm->mmap_sem);
+#endif
+
+	if (IS_ERR_VALUE(start)) {
+		return start;
+	}
+	*startp = start;
+	*endp = end;
+	return 0;
+}
+
+void get_vdso_info(ihk_os_t os, long vdso_rpa)
+{
+	ihk_device_t dev = ihk_os_to_dev(os);
+	long vdso_pa;
+	struct vdso *vdso;
+	size_t size;
+	int i;
+
+	vdso_pa = ihk_device_map_memory(dev, vdso_rpa, sizeof(*vdso));
+	vdso = ihk_device_map_virtual(dev, vdso_pa, sizeof(*vdso), NULL, 0);
+
+	/* VDSO pages */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3,16,0)
+	size = vdso_image->size;
+	vdso->vdso_npages = size >> PAGE_SHIFT;
+
+	if (vdso->vdso_npages > VDSO_MAXPAGES) {
+		vdso->vdso_npages = 0;
+		goto out;
+	}
+
+	for (i = 0; i < vdso->vdso_npages; ++i) {
+		vdso->vdso_physlist[i] = virt_to_phys(
+				vdso_image->data + (i * PAGE_SIZE));
+	}
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
+	size = vdso_end - vdso_start;
+	size = (size + PAGE_SIZE - 1) & PAGE_MASK;
+
+	vdso->vdso_npages = size >> PAGE_SHIFT;
+	if (vdso->vdso_npages > VDSO_MAXPAGES) {
+		vdso->vdso_npages = 0;
+		goto out;
+	}
+
+	for (i = 0; i < vdso->vdso_npages; ++i) {
+		vdso->vdso_physlist[i] = page_to_phys(vdso_pages[i]);
+	}
+#endif
+
+	/* VVAR page */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,5,0)
+	vdso->vvar_is_global = 0;
+	vdso->vvar_virt = (void *)(-3 * PAGE_SIZE);
+	vdso->vvar_phys = virt_to_phys(__vvar_page);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0)
+	vdso->vvar_is_global = 0;
+	vdso->vvar_virt = (void *)(-2 * PAGE_SIZE);
+	vdso->vvar_phys = virt_to_phys(__vvar_page);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,16,0)
+	vdso->vvar_is_global = 0;
+	vdso->vvar_virt = (void *)(vdso->vdso_npages * PAGE_SIZE);
+	vdso->vvar_phys = virt_to_phys(__vvar_page);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0)
+	vdso->vvar_is_global = 1;
+	vdso->vvar_virt = (void *)fix_to_virt(VVAR_PAGE);
+	vdso->vvar_phys = virt_to_phys(__vvar_page);
+#endif
+
+	/* HPET page */
+	if (hpet_addressp && *hpet_addressp) {
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,5,0)
+		vdso->hpet_is_global = 0;
+		vdso->hpet_virt = (void *)(-2 * PAGE_SIZE);
+		vdso->hpet_phys = *hpet_addressp;
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0)
+		vdso->hpet_is_global = 0;
+		vdso->hpet_virt = (void *)(-1 * PAGE_SIZE);
+		vdso->hpet_phys = *hpet_addressp;
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,16,0)
+		vdso->hpet_is_global = 0;
+		vdso->hpet_virt = (void *)((vdso->vdso_npages + 1) * PAGE_SIZE);
+		vdso->hpet_phys = *hpet_addressp;
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
+		vdso->hpet_is_global = 1;
+		vdso->hpet_virt = (void *)fix_to_virt(VSYSCALL_HPET);
+		vdso->hpet_phys = *hpet_addressp;
+#endif
+	}
+
+	/* struct pvlock_vcpu_time_info table */
+	if (hv_clockp && *hv_clockp) {
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,5,0)
+		vdso->pvti_is_global = 0;
+		vdso->pvti_virt = (void *)(-1 * PAGE_SIZE);
+		vdso->pvti_phys = virt_to_phys(*hv_clockp);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,8,0)
+		vdso->pvti_is_global = 1;
+		vdso->pvti_virt = (void *)fix_to_virt(PVCLOCK_FIXMAP_BEGIN);
+		vdso->pvti_phys = virt_to_phys(*hv_clockp);
+#endif
+	}
+
+out:
+	wmb();
+	vdso->busy = 0;
+
+	ihk_device_unmap_virtual(dev, vdso, sizeof(*vdso));
+	ihk_device_unmap_memory(dev, vdso_pa, sizeof(*vdso));
+	return;
+} /* get_vdso_info() */

executer/kernel/mcctrl/binfmt_mcexec.c

@@ -45,7 +45,6 @@ static int load_elf(struct linux_binprm *bprm
 #endif
 )
 {
-	char mcexec[BINPRM_BUF_SIZE];
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
 	const
 #endif
@@ -60,12 +59,8 @@ static int load_elf(struct linux_binprm *bprm
 		int	l;
 	} envdata;
 	envdata env[] = {
-		{.name = "MCEXEC"},
-#define env_mcexec (env[0].val)
 		{.name = "MCEXEC_WL"},
-#define env_mcexec_wl (env[1].val)
-		{.name = "MCEXEC_BL"},
-#define env_mcexec_bl (env[2].val)
+#define env_mcexec_wl (env[0].val)
 		{.name = NULL}
 	};
 	envdata *ep;
@@ -80,7 +75,11 @@ static int load_elf(struct linux_binprm *bprm
 	char buf[32];
 	int l;
 	int pass;
+	char *pbuf;
+	const char *path;
 
+	if(bprm->envc == 0)
+		return -ENOEXEC;
 	if(memcmp(elf_ex->e_ident, ELFMAG, SELFMAG) != 0)
 		return -ENOEXEC;
 	if(elf_ex->e_type != ET_EXEC && elf_ex->e_type != ET_DYN)
@@ -89,18 +88,29 @@ static int load_elf(struct linux_binprm *bprm
 	if(elf_ex->e_ident[EI_CLASS] != ELFCLASS64)
 		return -ENOEXEC;
 
-	cp = strrchr(bprm->interp, '/');
+	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;
+
+	cp = strrchr(path, '/');
 	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;
 	for(pass = 0; pass < 2; pass++){
 		p = bprm->p;
-		mode = cnt[0] == 0? (cnt[1] == 0? 2: 1): 0;
+		mode = cnt[0] == 0? 1: 0;
 		if(pass == 1){
 			for(ep = env; ep->name; ep++){
 				if(ep->l)
@@ -112,11 +122,19 @@ static int load_elf(struct linux_binprm *bprm
 		for(i = 0, st = 0; mode != 2;){
 			if(st == 0){
 				off = p & ~PAGE_MASK;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,6,0)
+				rc = get_user_pages_remote(current, bprm->mm,
+				                        bprm->p, 1, 0, 1,
+				                        &page, NULL);
+#else
 				rc = get_user_pages(current, bprm->mm,
 				                        bprm->p, 1, 0, 1,
 				                        &page, NULL);
-				if(rc <= 0)
+#endif
+				if(rc <= 0) {
+					kfree(pbuf);
 					return -EFAULT;
+				}
 				addr = kmap_atomic(page
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,4,0)
 				                   , KM_USER0
@@ -182,68 +200,52 @@ static int load_elf(struct linux_binprm *bprm
 		}
 	}
 
-	if(!env_mcexec || !strcmp(env_mcexec, "0") || !strcmp(env_mcexec, "off"))
-		rc = 1;
-	else{
-		rc = 0;
-		if(strchr(env_mcexec, '/') && strlen(env_mcexec) < BINPRM_BUF_SIZE)
-			strcpy(mcexec, env_mcexec);
-		else
-			strcpy(mcexec, MCEXEC_PATH);
-	}
-
-	if(rc);
-	else if(env_mcexec_wl)
-		rc = !pathcheck(bprm->interp, env_mcexec_wl);
-	else if(env_mcexec_bl)
-		rc = pathcheck(bprm->interp, env_mcexec_bl);
+	if(env_mcexec_wl)
+		rc = !pathcheck(path, env_mcexec_wl);
 	else
-		rc = pathcheck(bprm->interp, "/usr:/bin:/sbin:/opt");
+		rc = 1;
 
 	for(ep = env; ep->name; ep++)
 		if(ep->val)
 			kfree(ep->val);
-	if(rc)
+	if(rc) {
+		kfree(pbuf);
 		return -ENOEXEC;
+	}
 
-	file = open_exec(mcexec);
-	if (IS_ERR(file))
+	file = open_exec(MCEXEC_PATH);
+	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++;
-	wp = mcexec;
+	wp = MCEXEC_PATH;
 	rc = copy_strings_kernel(1, &wp, bprm);
 	if (rc){
 		fput(file);
+		kfree(pbuf);
 		return rc; 
 	}
 	bprm->argc++;
-#if 1
-	rc = bprm_change_interp(mcexec, bprm);
+	rc = bprm_change_interp(MCEXEC_PATH, bprm);
 	if (rc < 0){
 		fput(file);
+		kfree(pbuf);
 		return rc;
 	}
-#else
-	if(brpm->interp != bprm->filename)
-		kfree(brpm->interp);
-	kfree(brpm->filename);
-	bprm->filename = bprm->interp = kstrdup(mcexec, GFP_KERNEL);
-	if(!bprm->interp){
-		fput(file);
-		return -ENOMEM;
-	}
-#endif
 
 	allow_write_access(bprm->file);
 	fput(bprm->file);
@@ -251,8 +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
@@ -270,7 +276,7 @@ void __init binfmt_mcexec_init(void)
 	insert_binfmt(&mcexec_format);
 }
 
-void __exit binfmt_mcexec_exit(void)
+void binfmt_mcexec_exit(void)
 {
 	unregister_binfmt(&mcexec_format);
 }

executer/kernel/mcctrl/driver.c

@@ -25,6 +25,7 @@
 #include <linux/fs.h>
 #include <linux/miscdevice.h>
 #include <linux/slab.h>
+#include <linux/device.h>
 #include "mcctrl.h"
 
 #define OS_MAX_MINOR 64
@@ -67,6 +68,8 @@ static struct ihk_os_user_call_handler mcctrl_uchs[] = {
 	{ .request = MCEXEC_UP_CLOSE_EXEC, .func = mcctrl_ioctl },
 	{ .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_UNSHARE, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_DEBUG_LOG, .func = mcctrl_ioctl },
 };
 
@@ -79,71 +82,118 @@ static struct ihk_os_user_call mcctrl_uc[OS_MAX_MINOR];
 
 static ihk_os_t os[OS_MAX_MINOR];
 
-static int __init mcctrl_init(void)
+ihk_os_t osnum_to_os(int n)
+{
+	return os[n];
+}
+
+/* OS event notifier implementation */
+int mcctrl_os_boot_notifier(int os_index)
 {
-	int	i;
 	int	rc;
 
-	rc = -ENOENT;
-	for(i = 0; i < OS_MAX_MINOR; i++){
-		os[i] = ihk_host_find_os(i, NULL);
-		if (os[i]) {
-			printk("OS #%d found.\n", i);
-			rc = 0;
-		}
-	}
-	if(rc){
-		printk("OS not found.\n");
-		return rc;
+	os[os_index] = ihk_host_find_os(os_index, NULL);
+	if (!os[os_index]) {
+		printk("mcctrl: error: OS ID %d couldn't be found\n", os_index);
+		return -EINVAL;
 	}
 
-	for(i = 0; i < OS_MAX_MINOR; i++){
-		if (os[i]) {
-			if (prepare_ikc_channels(os[i]) != 0) {
-				printk("Preparing syscall channels failed.\n");
-				os[i] = NULL;
-			}
-		}
+	if (prepare_ikc_channels(os[os_index]) != 0) {
+		printk("mcctrl: error: preparing IKC channels for OS %d\n", os_index);
+
+		os[os_index] = NULL;
+		return -EFAULT;
 	}
 
+	memcpy(mcctrl_uc + os_index, &mcctrl_uc_proto, sizeof mcctrl_uc_proto);
+
+	rc = ihk_os_register_user_call_handlers(os[os_index], mcctrl_uc + os_index);
+	if (rc < 0) {
+		destroy_ikc_channels(os[os_index]);
+		printk("mcctrl: error: registering callbacks for OS %d\n", os_index);
+
+		goto error_cleanup_channels;
+	}
+
+	procfs_init(os_index);
+	printk("mcctrl: OS ID %d boot event handled\n", os_index);
+
+	return 0;
+
+error_cleanup_channels:
+	destroy_ikc_channels(os[os_index]);
+
+	os[os_index] = NULL;
+	return rc;
+}
+
+int mcctrl_os_shutdown_notifier(int os_index)
+{
+	if (os[os_index]) {
+		sysfsm_cleanup(os[os_index]);
+		free_topology_info(os[os_index]);
+		ihk_os_unregister_user_call_handlers(os[os_index], mcctrl_uc + os_index);
+		destroy_ikc_channels(os[os_index]);
+		procfs_exit(os_index);
+	}
+
+	os[os_index] = NULL;
+
+	printk("mcctrl: OS ID %d shutdown event handled\n", os_index);
+	return 0;
+}
+
+static struct ihk_os_notifier_ops mcctrl_os_notifier_ops = {
+	.boot = mcctrl_os_boot_notifier,
+	.shutdown = mcctrl_os_shutdown_notifier,
+};
+
+static struct ihk_os_notifier mcctrl_os_notifier = {
+	.ops = &mcctrl_os_notifier_ops,
+};
+
+static int __init mcctrl_init(void)
+{
+	int ret = 0;
+	int i;
+
 #ifndef DO_USER_MODE
 	mcctrl_syscall_init();
 #endif
 
-	rus_page_hash_init();
-
-	for(i = 0; i < OS_MAX_MINOR; i++){
-		if (os[i]) {
-			memcpy(mcctrl_uc + i, &mcctrl_uc_proto, sizeof mcctrl_uc_proto);
-			rc = ihk_os_register_user_call_handlers(os[i], mcctrl_uc + i);
-			if(rc < 0){
-				destroy_ikc_channels(os[i]);
-				os[i] = NULL;
-			}
-			procfs_init(i);
-		}
+	for (i = 0; i < OS_MAX_MINOR; ++i) {
+		os[i] = NULL;
 	}
 
+	rus_page_hash_init();
+
 	binfmt_mcexec_init();
 
-	return 0;
+	if ((ret = ihk_host_register_os_notifier(&mcctrl_os_notifier)) != 0) {
+		printk("mcctrl: error: registering OS notifier\n");
+		goto error;
+	}
+
+	printk("mcctrl: initialized successfully.\n");
+	return ret;
+
+error:
+	binfmt_mcexec_exit();
+	rus_page_hash_put_pages();
+
+	return ret;
 }
 
 static void __exit mcctrl_exit(void)
 {
-	int	i;
-
-	binfmt_mcexec_exit();
-	printk("mcctrl: unregistered.\n");
-	for(i = 0; i < OS_MAX_MINOR; i++){
-		if(os[i]){
-			ihk_os_unregister_user_call_handlers(os[i], mcctrl_uc + i);
-			destroy_ikc_channels(os[i]);
-			procfs_exit(i);
-		}
+	if (ihk_host_deregister_os_notifier(&mcctrl_os_notifier) != 0) {
+		printk("mcctrl: warning: failed to deregister OS notifier??\n");
 	}
 
+	binfmt_mcexec_exit();
 	rus_page_hash_put_pages();
+
+	printk("mcctrl: unregistered.\n");
 }
 
 MODULE_LICENSE("GPL v2");

executer/kernel/mcctrl/ikc.c

@@ -27,6 +27,7 @@
 #include <linux/miscdevice.h>
 #include <linux/slab.h>
 #include <linux/string.h>
+#include <linux/interrupt.h>
 #include "mcctrl.h"
 #ifdef ATTACHED_MIC
 #include <sysdeps/mic/mic/micconst.h>
@@ -34,25 +35,34 @@
 
 #define REQUEST_SHIFT    16
 
+//#define DEBUG_IKC
+
+#ifdef DEBUG_IKC
+#define	dkprintf(...) kprintf(__VA_ARGS__)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) printk(__VA_ARGS__); } while (0)
+#define	ekprintf(...) printk(__VA_ARGS__)
+#endif
+
 //int num_channels;
 
 //struct mcctrl_channel *channels;
 
 void mcexec_prepare_ack(ihk_os_t os, unsigned long arg, int err);
 static void mcctrl_ikc_init(ihk_os_t os, int cpu, unsigned long rphys, struct ihk_ikc_channel_desc *c);
-int mcexec_syscall(struct mcctrl_channel *c, int pid, unsigned long arg);
-void procfs_create(void *__os, int ref, int osnum, int pid, unsigned long arg);
-void procfs_delete(void *__os, int osnum, unsigned long arg);
-void procfs_answer(unsigned long arg, int err);
+int mcexec_syscall(struct mcctrl_usrdata *ud, struct ikc_scd_packet *packet);
 void sig_done(unsigned long arg, int err);
 
+/* XXX: this runs in atomic context! */
 static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
                                   void *__packet, void *__os)
 {
 	struct ikc_scd_packet *pisp = __packet;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(__os);
+	int msg = pisp->msg;
 
-	switch (pisp->msg) {
+	switch (msg) {
 	case SCD_MSG_INIT_CHANNEL:
 		mcctrl_ikc_init(__os, pisp->ref, pisp->arg, c);
 		break;
@@ -66,15 +76,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		break;
 
 	case SCD_MSG_SYSCALL_ONESIDE:
-		mcexec_syscall(usrdata->channels + pisp->ref, pisp->pid, pisp->arg);
-		break;
-
-	case SCD_MSG_PROCFS_CREATE:
-		procfs_create(__os, pisp->ref, pisp->osnum, pisp->pid, pisp->arg);
-		break;
-
-	case SCD_MSG_PROCFS_DELETE:
-		procfs_delete(__os, pisp->osnum, pisp->arg);
+		mcexec_syscall(usrdata, pisp);
 		break;
 
 	case SCD_MSG_PROCFS_ANSWER:
@@ -84,6 +86,43 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	case SCD_MSG_SEND_SIGNAL:
 		sig_done(pisp->arg, pisp->err);
 		break;
+
+	case SCD_MSG_SYSFS_REQ_CREATE:
+	case SCD_MSG_SYSFS_REQ_MKDIR:
+	case SCD_MSG_SYSFS_REQ_SYMLINK:
+	case SCD_MSG_SYSFS_REQ_LOOKUP:
+	case SCD_MSG_SYSFS_REQ_UNLINK:
+	case SCD_MSG_SYSFS_REQ_SETUP:
+	case SCD_MSG_SYSFS_RESP_SHOW:
+	case SCD_MSG_SYSFS_RESP_STORE:
+	case SCD_MSG_SYSFS_RESP_RELEASE:
+		sysfsm_packet_handler(__os, pisp->msg, pisp->err,
+				pisp->sysfs_arg1, pisp->sysfs_arg2);
+		break;
+
+	case SCD_MSG_PROCFS_TID_CREATE:
+	case SCD_MSG_PROCFS_TID_DELETE:
+		procfsm_packet_handler(__os, pisp->msg, pisp->pid, pisp->arg);
+		break;
+
+	case SCD_MSG_GET_VDSO_INFO:
+		get_vdso_info(__os, pisp->arg);
+		break;
+
+	default:
+		printk(KERN_ERR "mcctrl:syscall_packet_handler:"
+				"unknown message (%d.%d.%d.%d.%d.%#lx)\n",
+				pisp->msg, pisp->ref, pisp->osnum, pisp->pid,
+				pisp->err, pisp->arg);
+		break;
+	}
+
+	/* 
+	 * SCD_MSG_SYSCALL_ONESIDE holds the packet and frees is it
+	 * mcexec_ret_syscall(), for the rest, free it here.
+	 */
+	if (msg != SCD_MSG_SYSCALL_ONESIDE) {
+		ihk_ikc_release_packet((struct ihk_ikc_free_packet *)__packet, c);
 	}
 	return 0;
 }
@@ -121,8 +160,6 @@ int mcctrl_ikc_set_recv_cpu(ihk_os_t os, int cpu)
 
 	ihk_ikc_channel_set_cpu(usrdata->channels[cpu].c,
 	                        ihk_ikc_get_processor_id());
-	kprintf("Setting the target to %d\n",
-	        ihk_ikc_get_processor_id());
 	return 0;
 }
 
@@ -137,91 +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(GFP_KERNEL,
-		                         REQUEST_SHIFT - PAGE_SHIFT);
-	pmc->param.request_pa = virt_to_phys(pmc->param.request_va);
-	pmc->param.doorbell_va = usrdata->mcctrl_doorbell_va;
-	pmc->param.doorbell_pa = usrdata->mcctrl_doorbell_pa;
-	pmc->param.post_va = (void *)__get_free_page(GFP_KERNEL);
-	pmc->param.post_pa = virt_to_phys(pmc->param.post_va);
-	memset(pmc->param.doorbell_va, 0, PAGE_SIZE);
-	memset(pmc->param.request_va, 0, PAGE_SIZE);
-	memset(pmc->param.post_va, 0, PAGE_SIZE);
-	
-	pmc->param.response_rpa = rpm->response_page;
-	pmc->param.response_pa 
-		= ihk_device_map_memory(ihk_os_to_dev(os),
-		                        pmc->param.response_rpa,
-		                        PAGE_SIZE);
-#ifdef CONFIG_MIC							
-	pmc->param.response_va = ioremap_cache(pmc->param.response_pa,
-	                                       PAGE_SIZE);
-#else
-	pmc->param.response_va = ihk_device_map_virtual(ihk_os_to_dev(os), 
-	                                                pmc->param.response_pa,
-													PAGE_SIZE, NULL, 0);
-#endif
-
-	pmc->dma_buf = (void *)__get_free_pages(GFP_KERNEL,
-	                                        DMA_PIN_SHIFT - PAGE_SHIFT);
-
-	rpm->request_page = pmc->param.request_pa;
-	rpm->doorbell_page = pmc->param.doorbell_pa;
-	rpm->post_page = pmc->param.post_pa;
+	if (!pmc) {
+		kprintf("%s: error: no channel found?\n", __FUNCTION__);
+		return;
+	}
 
 	packet.msg = SCD_MSG_INIT_CHANNEL_ACKED;
 	packet.ref = cpu;
 	packet.arg = rphys;
 
-	printk("Request: %lx, Response: %lx, Doorbell: %lx\n",
-	       pmc->param.request_pa, pmc->param.response_rpa,
-	       pmc->param.doorbell_pa);
-	printk("Request: %p, Response: %p, Doorbell: %p\n",
-	       pmc->param.request_va, pmc->param.response_va,
-	       pmc->param.doorbell_va);
-
 	ihk_ikc_send(pmc->c, &packet, 0);
-
-#ifdef CONFIG_MIC							
-	iounmap(rpm);
-#else
-	ihk_device_unmap_virtual(ihk_os_to_dev(os), rpm, 
-	                         sizeof(struct ikc_scd_init_param));
-#endif
-
-	ihk_device_unmap_memory(ihk_os_to_dev(os), phys,
-	                        sizeof(struct ikc_scd_init_param));
 }
 
 static int connect_handler(struct ihk_ikc_channel_info *param)
@@ -240,11 +212,8 @@ static int connect_handler(struct ihk_ikc_channel_info *param)
 	}
 	param->packet_handler = syscall_packet_handler;
 	
-	INIT_LIST_HEAD(&usrdata->channels[cpu].wq_list);
-	spin_lock_init(&usrdata->channels[cpu].wq_list_lock);
-
 	usrdata->channels[cpu].c = c;
-	kprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
+	dkprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
 
 	return 0;
 }
@@ -261,11 +230,8 @@ static int connect_handler2(struct ihk_ikc_channel_info *param)
 
 	param->packet_handler = syscall_packet_handler;
 	
-	INIT_LIST_HEAD(&usrdata->channels[cpu].wq_list);
-	spin_lock_init(&usrdata->channels[cpu].wq_list_lock);
-
 	usrdata->channels[cpu].c = c;
-	kprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
+	dkprintf("syscall: MC CPU %d connected. c=%p\n", cpu, c);
 
 	return 0;
 }
@@ -288,27 +254,29 @@ static struct ihk_ikc_listen_param listen_param2 = {
 
 int prepare_ikc_channels(ihk_os_t os)
 {
-	struct ihk_cpu_info *info;
-	struct mcctrl_usrdata   *usrdata;
-	int error;
+	struct mcctrl_usrdata *usrdata;
+	int i;
 
 	usrdata = kzalloc(sizeof(struct mcctrl_usrdata), GFP_KERNEL);
-	usrdata->mcctrl_doorbell_va = (void *)__get_free_page(GFP_KERNEL);
-	usrdata->mcctrl_doorbell_pa = virt_to_phys(usrdata->mcctrl_doorbell_va);
 
-	info = ihk_os_get_cpu_info(os);
-	if (!info) {
-		printk("Error: cannot retrieve CPU info.\n");
+	usrdata->cpu_info = ihk_os_get_cpu_info(os);
+	usrdata->mem_info = ihk_os_get_memory_info(os);
+
+	if (!usrdata->cpu_info || !usrdata->mem_info) {
+		printk("Error: cannot obtain OS CPU and memory information.\n");
 		return -EINVAL;
 	}
-	if (info->n_cpus < 1) {
+
+	if (usrdata->cpu_info->n_cpus < 1) {
 		printk("Error: # of cpu is invalid.\n");
 		return -EINVAL;
 	}
 
-	usrdata->num_channels = info->n_cpus + 1;
-	usrdata->channels = kzalloc(sizeof(struct mcctrl_channel) * usrdata->num_channels,
-	                   GFP_KERNEL);
+	usrdata->num_channels = usrdata->cpu_info->n_cpus + 1;
+	usrdata->channels = kzalloc(sizeof(struct mcctrl_channel) *
+			usrdata->num_channels,
+			GFP_KERNEL);
+
 	if (!usrdata->channels) {
 		printk("Error: cannot allocate channels.\n");
 		return -ENOMEM;
@@ -322,33 +290,20 @@ int prepare_ikc_channels(ihk_os_t os)
 	memcpy(&usrdata->listen_param2, &listen_param2, sizeof listen_param2);
 	ihk_ikc_listen_port(os, &usrdata->listen_param2);
 
-	INIT_LIST_HEAD(&usrdata->per_proc_list);
-	spin_lock_init(&usrdata->per_proc_list_lock);
-
-	error = init_peer_channel_registry(usrdata);
-	if (error) {
-		return error;
+	for (i = 0; i < MCCTRL_PER_PROC_DATA_HASH_SIZE; ++i) {
+		INIT_LIST_HEAD(&usrdata->per_proc_data_hash[i]);
+		rwlock_init(&usrdata->per_proc_data_hash_lock[i]);
 	}
 
+	INIT_LIST_HEAD(&usrdata->cpu_topology_list);
+	INIT_LIST_HEAD(&usrdata->node_topology_list);
+
 	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)
@@ -356,6 +311,11 @@ 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++) {
@@ -366,7 +326,6 @@ 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

@@ -0,0 +1,388 @@
+/**
+ * \file mcctrl.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  define data structure
+ * \author Taku Shimosawa  <shimosawa@is.s.u-tokyo.ac.jp> \par
+ *      Copyright (C) 2011 - 2012  Taku Shimosawa
+ * \author Balazs Gerofi  <bgerofi@riken.jp> \par
+ *      Copyright (C) 2012  RIKEN AICS
+ * \author Gou Nakamura  <go.nakamura.yw@hitachi-solutions.com> \par
+ *      Copyright (C) 2012 - 2013 Hitachi, Ltd.
+ * \author Tomoki Shirasawa  <tomoki.shirasawa.kk@hitachi-solutions.com> \par
+ *      Copyright (C) 2012 - 2013 Hitachi, Ltd.
+ * \author Balazs Gerofi  <bgerofi@is.s.u-tokyo.ac.jp> \par
+ *      Copyright (C) 2013  The University of Tokyo
+ */
+/*
+ * HISTORY:
+ *  2013/11/07 hamada added <sys/resource.h> which is required by getrlimit(2)
+ *  2013/10/21 nakamura exclude interpreter's segment from data region
+ *  2013/10/11 nakamura mcexec: add a upper limit of the stack size
+ *  2013/10/11 nakamura mcexec: add a path prefix for interpreter search
+ *  2013/10/11 nakamura mcexec: add a interpreter invocation
+ *  2013/10/08 nakamura add a AT_ENTRY entry to the auxiliary vector
+ *  2013/09/02 shirasawa add terminate thread
+ *  2013/08/19 shirasawa mcexec forward signal to MIC process
+ *  2013/08/07 nakamura add page fault forwarding
+ *  2013/07/26 shirasawa mcexec print signum or exit status
+ *  2013/07/17 nakamura create more mcexec thread so that all cpu to be serviced
+ *  2013/04/17 nakamura add generic system call forwarding
+ */
+#ifndef HEADER_MCCTRL_H
+#define HEADER_MCCTRL_H
+
+#include <linux/fs.h>
+#include <ihk/ihk_host_driver.h>
+#include <linux/resource.h>
+#include <uprotocol.h>
+#include <linux/wait.h>
+#include <ihk/ikc.h>
+#include <ikc/master.h>
+#include <ihk/msr.h>
+#include <linux/semaphore.h>
+#include <linux/rwlock.h>
+#include <linux/threads.h>
+#include "sysfs.h"
+
+#define SCD_MSG_PREPARE_PROCESS         0x1
+#define SCD_MSG_PREPARE_PROCESS_ACKED   0x2
+#define SCD_MSG_PREPARE_PROCESS_NACKED  0x7
+#define SCD_MSG_SCHEDULE_PROCESS        0x3
+#define SCD_MSG_WAKE_UP_SYSCALL_THREAD  0x14
+
+#define SCD_MSG_INIT_CHANNEL            0x5
+#define SCD_MSG_INIT_CHANNEL_ACKED      0x6
+
+#define SCD_MSG_SYSCALL_ONESIDE         0x4
+#define SCD_MSG_SEND_SIGNAL     	0x8
+#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_PROCFS_CREATE		0x10
+#define	SCD_MSG_PROCFS_DELETE		0x11
+#define	SCD_MSG_PROCFS_REQUEST		0x12
+#define	SCD_MSG_PROCFS_ANSWER		0x13
+
+#define SCD_MSG_DEBUG_LOG		0x20
+
+#define SCD_MSG_SYSFS_REQ_CREATE        0x30
+/* #define SCD_MSG_SYSFS_RESP_CREATE    0x31 */
+#define SCD_MSG_SYSFS_REQ_MKDIR         0x32
+/* #define SCD_MSG_SYSFS_RESP_MKDIR     0x33 */
+#define SCD_MSG_SYSFS_REQ_SYMLINK       0x34
+/* #define SCD_MSG_SYSFS_RESP_SYMLINK   0x35 */
+#define SCD_MSG_SYSFS_REQ_LOOKUP        0x36
+/* #define SCD_MSG_SYSFS_RESP_LOOKUP    0x37 */
+#define SCD_MSG_SYSFS_REQ_UNLINK        0x38
+/* #define SCD_MSG_SYSFS_RESP_UNLINK    0x39 */
+#define SCD_MSG_SYSFS_REQ_SHOW          0x3a
+#define SCD_MSG_SYSFS_RESP_SHOW         0x3b
+#define SCD_MSG_SYSFS_REQ_STORE         0x3c
+#define SCD_MSG_SYSFS_RESP_STORE        0x3d
+#define SCD_MSG_SYSFS_REQ_RELEASE       0x3e
+#define SCD_MSG_SYSFS_RESP_RELEASE      0x3f
+#define SCD_MSG_SYSFS_REQ_SETUP         0x40
+#define SCD_MSG_SYSFS_RESP_SETUP        0x41
+/* #define SCD_MSG_SYSFS_REQ_CLEANUP    0x42 */
+/* #define SCD_MSG_SYSFS_RESP_CLEANUP   0x43 */
+#define SCD_MSG_PROCFS_TID_CREATE	0x44
+#define SCD_MSG_PROCFS_TID_DELETE	0x45
+
+#define DMA_PIN_SHIFT                   21
+
+#define DO_USER_MODE
+
+#define	__NR_coredump			999
+
+struct coretable {
+	int len;
+	unsigned long addr;
+};
+
+struct ikc_scd_packet {
+	int msg;
+	int err;
+	union {
+		/* for traditional SCD_MSG_* */
+		struct {
+			int ref;
+			int osnum;
+			int pid;
+			unsigned long arg;
+			struct syscall_request req;
+			unsigned long resp_pa;
+		};
+
+		/* for SCD_MSG_SYSFS_* */
+		struct {
+			long sysfs_arg1;
+			long sysfs_arg2;
+			long sysfs_arg3;
+		};
+
+		/* SCD_MSG_SCHEDULE_THREAD */
+		struct {
+			int ttid;
+		};
+	};
+	char padding[12];
+};
+
+struct mcctrl_priv { 
+	ihk_os_t os;
+	struct program_load_desc *desc;
+};
+
+struct ikc_scd_init_param {
+	unsigned long request_page;
+	unsigned long response_page;
+	unsigned long doorbell_page;
+	unsigned long post_page;
+};
+
+struct syscall_post {
+	unsigned long v[8];
+};
+
+struct syscall_params {
+	unsigned long request_pa;
+	struct syscall_request *request_va;
+	unsigned long response_rpa, response_pa;
+	struct syscall_response *response_va;
+	unsigned long post_pa;
+	struct syscall_post *post_va;
+	
+	unsigned long doorbell_pa;
+	unsigned long *doorbell_va;
+};
+
+struct wait_queue_head_list_node {
+	struct list_head list;
+	wait_queue_head_t wq_syscall;
+	struct task_struct *task;
+	/* Denotes an exclusive wait for requester TID rtid */
+	int rtid;
+	int req;
+	struct ikc_scd_packet *packet;
+};
+
+struct mcctrl_channel {
+	struct ihk_ikc_channel_desc *c;
+	struct ikc_scd_init_param init;
+	void *dma_buf;
+};
+
+struct mcctrl_per_thread_data {
+	struct list_head hash;
+	struct task_struct *task;
+	void *data;
+};
+
+#define MCCTRL_PER_THREAD_DATA_HASH_SHIFT 8
+#define MCCTRL_PER_THREAD_DATA_HASH_SIZE (1 << MCCTRL_PER_THREAD_DATA_HASH_SHIFT)
+#define MCCTRL_PER_THREAD_DATA_HASH_MASK (MCCTRL_PER_THREAD_DATA_HASH_SIZE - 1) 
+
+struct mcctrl_per_proc_data {
+	struct list_head hash;
+	int pid;
+	unsigned long rpgtable;	/* per process, not per OS */
+
+	struct list_head wq_list;
+	struct list_head wq_req_list;
+	struct list_head wq_list_exact;
+	ihk_spinlock_t wq_list_lock;
+
+	struct list_head per_thread_data_hash[MCCTRL_PER_THREAD_DATA_HASH_SIZE];
+	rwlock_t per_thread_data_hash_lock[MCCTRL_PER_THREAD_DATA_HASH_SIZE];
+};
+
+struct sysfsm_req {
+	int busy;
+	int padding;
+	long lresult;
+	wait_queue_head_t wq;
+};
+
+struct sysfsm_data {
+	size_t sysfs_bufsize;
+	void *sysfs_buf;
+	long sysfs_buf_rpa;
+	long sysfs_buf_pa;
+	struct kobject *sysfs_kobj;
+	struct sysfsm_node *sysfs_root;
+	struct semaphore sysfs_tree_sem;
+	struct semaphore sysfs_io_sem;
+	struct sysfsm_req sysfs_req;
+	ihk_os_t sysfs_os;
+};
+
+static inline int sysfs_inited(struct sysfsm_data *sdp)
+{
+	return !!(sdp->sysfs_buf);
+} /* sysfs_inited() */
+
+struct cache_topology {
+	struct ihk_cache_topology *saved;
+	cpumask_t shared_cpu_map;
+
+	struct list_head chain;
+};
+
+struct cpu_topology {
+	//struct mcctrl_usrdata *udp;
+	struct ihk_cpu_topology *saved;
+	int mckernel_cpu_id;
+	cpumask_t core_siblings;
+	cpumask_t thread_siblings;
+
+	struct list_head chain;
+	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;
+};
+
+#define CPU_LONGS (((NR_CPUS) + (BITS_PER_LONG) - 1) / (BITS_PER_LONG))
+
+#define MCCTRL_PER_PROC_DATA_HASH_SHIFT 7
+#define MCCTRL_PER_PROC_DATA_HASH_SIZE (1 << MCCTRL_PER_PROC_DATA_HASH_SHIFT)
+#define MCCTRL_PER_PROC_DATA_HASH_MASK (MCCTRL_PER_PROC_DATA_HASH_SIZE - 1)
+
+struct mcctrl_usrdata {
+	struct ihk_ikc_listen_param listen_param;
+	struct ihk_ikc_listen_param listen_param2;
+	ihk_os_t	os;
+	int	num_channels;
+	struct mcctrl_channel *channels;
+	int	remaining_job;
+	int	base_cpu;
+	int	job_pos;
+	int	mcctrl_dma_abort;
+	unsigned long	last_thread_exec;
+	wait_queue_head_t	wq_prepare;
+	
+	struct list_head per_proc_data_hash[MCCTRL_PER_PROC_DATA_HASH_SIZE];
+	rwlock_t per_proc_data_hash_lock[MCCTRL_PER_PROC_DATA_HASH_SIZE];
+
+	void **keys;
+	struct sysfsm_data sysfsm_data;
+	unsigned long cpu_online[CPU_LONGS];
+	struct ihk_cpu_info *cpu_info;
+	struct ihk_mem_info *mem_info;
+	struct list_head cpu_topology_list;
+	struct list_head node_topology_list;
+};
+
+struct mcctrl_signal {
+	int	cond;
+	int	sig;
+	int	pid;
+	int	tid;
+	char	info[128];
+};
+
+int mcctrl_ikc_send(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp);
+int mcctrl_ikc_send_msg(ihk_os_t os, int cpu, int msg, int ref, unsigned long arg);
+int mcctrl_ikc_is_valid_thread(ihk_os_t os, int cpu);
+
+ihk_os_t osnum_to_os(int n);
+
+/* syscall.c */
+int __do_in_kernel_syscall(ihk_os_t os, struct ikc_scd_packet *packet);
+int mcctrl_add_per_proc_data(struct mcctrl_usrdata *ud, int pid, 
+	struct mcctrl_per_proc_data *ppd);
+int mcctrl_delete_per_proc_data(struct mcctrl_usrdata *ud, int pid);
+inline struct mcctrl_per_proc_data *mcctrl_get_per_proc_data(
+		struct mcctrl_usrdata *ud, int pid);
+
+int mcctrl_add_per_thread_data(struct mcctrl_per_proc_data* ppd,
+	struct task_struct *task, void *data);
+int mcctrl_delete_per_thread_data(struct mcctrl_per_proc_data* ppd,
+	struct task_struct *task);
+inline struct mcctrl_per_thread_data *mcctrl_get_per_thread_data(
+	struct mcctrl_per_proc_data *ppd, struct task_struct *task);
+
+void __return_syscall(ihk_os_t os, struct ikc_scd_packet *packet, 
+		long ret, int stid);
+
+#define PROCFS_NAME_MAX 768
+
+struct procfs_read {
+	unsigned long pbuf;	/* physical address of the host buffer (request) */
+	unsigned long offset;	/* offset to read (request) */
+	int count;		/* bytes to read (request) */
+	int eof;		/* if eof is detected, 1 otherwise 0. (answer)*/
+	int ret;		/* read bytes (answer) */
+	int status;		/* non-zero if done (answer) */
+	int newcpu;		/* migrated new cpu (answer) */
+	int readwrite;		/* 0:read, 1:write */
+	char fname[PROCFS_NAME_MAX];	/* procfs filename (request) */
+};
+
+struct procfs_file {
+	int status;			/* status of processing (answer) */
+	int mode;			/* file mode (request) */
+	char fname[PROCFS_NAME_MAX];	/* procfs filename (request) */
+};
+
+void procfs_answer(unsigned int arg, int err);
+int procfsm_packet_handler(void *os, int msg, int pid, unsigned long arg);
+void add_tid_entry(int osnum, int pid, int tid);
+void add_pid_entry(int osnum, int pid);
+void delete_tid_entry(int osnum, int pid, int tid);
+void delete_pid_entry(int osnum, int pid);
+void proc_exe_link(int osnum, int pid, const char *path);
+void procfs_init(int osnum);
+void procfs_exit(int osnum);
+
+/* sysfs_files.c */
+void setup_sysfs_files(ihk_os_t os);
+void reply_get_cpu_mapping(long req_pa);
+void free_topology_info(ihk_os_t os);
+
+/* archdep.c */
+#define VDSO_MAXPAGES 2
+struct vdso {
+	long busy;
+	int vdso_npages;
+	char vvar_is_global;
+	char hpet_is_global;
+	char pvti_is_global;
+	char padding;
+	long vdso_physlist[VDSO_MAXPAGES];
+	void *vvar_virt;
+	long vvar_phys;
+	void *hpet_virt;
+	long hpet_phys;
+	void *pvti_virt;
+	long pvti_phys;
+};
+
+int reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp,
+		unsigned long *endp);
+void get_vdso_info(ihk_os_t os, long vdso_pa);
+
+struct get_cpu_mapping_req {
+	int busy;		/* INOUT: */
+	int error;		/* OUT: */
+	long buf_rpa;		/* OUT: physical address of struct cpu_mapping */
+	int buf_elems;		/* OUT: # of elements of buf */
+	int padding;
+
+	/* work for mcctrl */
+	wait_queue_head_t wq;
+};
+
+#endif

executer/kernel/mcctrl/procfs.c

@@ -0,0 +1,838 @@
+/**
+ * \file procfs.c
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  mcctrl procfs
+ * \author Naoki Hamada <nao@axe.bz> \par
+ * 	Copyright (C) 2014  AXE, Inc.
+ */
+/*
+ * HISTORY:
+ */
+
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/proc_fs.h>
+#include <linux/list.h>
+#include <linux/uaccess.h>
+#include <linux/fs.h>
+#include <linux/resource.h>
+#include <linux/interrupt.h>
+#include "mcctrl.h"
+#include <linux/version.h>
+#include <linux/semaphore.h>
+
+//#define PROCFS_DEBUG
+
+#ifdef PROCFS_DEBUG
+#define	dprintk(...)	printk(__VA_ARGS__)
+#else
+#define	dprintk(...)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,5,0)
+typedef uid_t kuid_t;
+typedef gid_t kgid_t;
+#endif
+
+struct procfs_entry {
+	char *name;
+	mode_t mode;
+	const struct file_operations *fops;
+};
+
+#define NOD(NAME, MODE, FOP) {				\
+	.name = (NAME),					\
+	.mode = MODE,					\
+	.fops  = FOP,					\
+}
+#define PROC_DIR(NAME, MODE)				\
+	NOD(NAME, (S_IFDIR|(MODE)), NULL)
+#define PROC_REG(NAME, MODE, fops)			\
+	NOD(NAME, (S_IFREG|(MODE)), fops)
+#define PROC_TERM					\
+	NOD(NULL, 0, NULL)
+
+static const struct procfs_entry tid_entry_stuff[];
+static const struct procfs_entry pid_entry_stuff[];
+static const struct procfs_entry base_entry_stuff[];
+static const struct file_operations mckernel_forward_ro;
+static const struct file_operations mckernel_forward;
+
+static DECLARE_WAIT_QUEUE_HEAD(procfsq);
+static ssize_t mckernel_procfs_read(struct file *file, char __user *buf, 
+		size_t nbytes, loff_t *ppos);
+
+/* A private data for the procfs driver. */
+struct procfs_list_entry;
+
+struct procfs_list_entry {
+	struct list_head list;
+	struct proc_dir_entry *entry;
+	struct procfs_list_entry *parent;
+	struct list_head children;
+	int osnum;
+	char *data;
+	char name[0];
+};
+
+/*
+ * In the procfs_file_list, mckenrel procfs files are
+ * listed in the manner that the leaf file is located 
+ * always nearer to the list top than its parent node 
+ * file.
+ */
+LIST_HEAD(procfs_file_list);
+DEFINE_SEMAPHORE(procfs_file_list_lock);
+
+static char *
+getpath(struct procfs_list_entry *e, char *buf, int bufsize)
+{
+	char	*w = buf + bufsize - 1;
+
+	*w = '\0';
+	for(;;){
+		int l = strlen(e->name);
+		w -= l;
+		memcpy(w, e->name, l);
+		e = e->parent;
+		if(!e)
+			return w;
+		w--;
+		*w = '/';
+	}
+}
+
+/**
+ * \brief Process SCD_MSG_PROCFS_ANSWER message.
+ *
+ * \param arg sent argument
+ * \param err error info (redundant)
+ */
+void
+procfs_answer(unsigned int arg, int err)
+{
+	dprintk("procfs: received SCD_MSG_PROCFS_ANSWER message(err = %d).\n", err);
+	wake_up_interruptible(&procfsq);
+}
+
+static struct procfs_list_entry *
+find_procfs_entry(struct procfs_list_entry *parent, const char *name)
+{
+	struct list_head *list;
+	struct procfs_list_entry *e;
+
+	if(parent == NULL)
+		list = &procfs_file_list;
+	else
+		list = &parent->children;
+
+	list_for_each_entry(e, list, list) {
+		if(!strcmp(e->name, name))
+			return e;
+	}
+
+	return NULL;
+}
+
+static void
+delete_procfs_entries(struct procfs_list_entry *top)
+{
+	struct procfs_list_entry *e;
+	struct procfs_list_entry *n;
+
+	list_del(&top->list);
+
+	list_for_each_entry_safe(e, n, &top->children, list) {
+		delete_procfs_entries(e);
+	}
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
+	e->entry->read_proc = NULL;
+	e->entry->data = NULL;
+#endif
+	remove_proc_entry(top->name, top->parent? top->parent->entry: NULL);
+	if(top->data)
+		kfree(top->data);
+	kfree(top);
+}
+
+static struct procfs_list_entry *
+add_procfs_entry(struct procfs_list_entry *parent, const char *name, int mode,
+                 kuid_t uid, kgid_t gid, const void *opaque)
+{
+	struct procfs_list_entry *e = find_procfs_entry(parent, name);
+	struct proc_dir_entry *pde;
+	struct proc_dir_entry *parent_pde = NULL;
+	int f_mode = mode & 0777;
+
+	if(e)
+		delete_procfs_entries(e);
+
+	e = kmalloc(sizeof(struct procfs_list_entry) + strlen(name) + 1,
+	            GFP_KERNEL);
+	if(!e){
+		kprintf("ERROR: not enough memory to create PROCFS entry.\n");
+		return NULL;
+	}
+	memset(e, '\0', sizeof(struct procfs_list_entry));
+	INIT_LIST_HEAD(&e->children);
+	strcpy(e->name, name);
+
+	if(parent)
+		parent_pde = parent->entry;
+
+	if (mode & S_IFDIR) {
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
+		pde = proc_mkdir(name, parent_pde);
+#else
+		pde = proc_mkdir_data(name, f_mode, parent_pde, e);
+#endif
+	}
+	else if ((mode & S_IFLNK) == S_IFLNK) {
+		pde = proc_symlink(name, parent_pde, (char *)opaque);
+	}
+	else {
+		const struct file_operations *fop;
+
+		if(opaque)
+			fop = (const struct file_operations *)opaque;
+		else if(mode & S_IWUSR)
+			fop = &mckernel_forward;
+		else
+			fop = &mckernel_forward_ro;
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
+		pde = create_proc_entry(name, f_mode, parent_pde);
+		if(pde)
+			pde->proc_fops = fop;
+#else
+		pde = proc_create_data(name, f_mode, parent_pde, fop, e);
+		if(pde)
+			proc_set_user(pde, uid, gid);
+#endif
+	}
+	if(!pde){
+		kprintf("ERROR: cannot create a PROCFS entry for %s.\n", name);
+		kfree(e);
+		return NULL;
+	}
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
+	pde->uid = uid;
+	pde->gid = gid;
+	pde->data = e;
+#endif
+
+	if(parent)
+		e->osnum = parent->osnum;
+	e->entry = pde;
+	e->parent = parent;
+	list_add(&(e->list), parent? &(parent->children): &procfs_file_list);
+
+	return e;
+}
+
+static void
+add_procfs_entries(struct procfs_list_entry *parent,
+                   const struct procfs_entry *entries, kuid_t uid, kgid_t gid)
+{
+	const struct procfs_entry *p;
+
+	for(p = entries; p->name; p++){
+		add_procfs_entry(parent, p->name, p->mode, uid, gid, p->fops);
+	}
+}
+
+static const struct cred *
+get_pid_cred(int pid)
+{
+	struct task_struct *task = NULL;
+
+	if(pid > 0){
+		task = pid_task(find_vpid(pid), PIDTYPE_PID);
+		if(task){
+			return __task_cred(task);
+		}
+	}
+	return NULL;
+}
+
+static struct procfs_list_entry *
+find_base_entry(int osnum)
+{
+	char name[12];
+
+	sprintf(name, "mcos%d", osnum);
+	return find_procfs_entry(NULL, name);
+}
+
+static struct procfs_list_entry *
+find_pid_entry(int osnum, int pid)
+{
+	struct procfs_list_entry *e;
+	char name[12];
+
+	if(!(e = find_base_entry(osnum)))
+		return NULL;
+	sprintf(name, "%d", pid);
+	return find_procfs_entry(e, name);
+}
+
+static struct procfs_list_entry *
+find_tid_entry(int osnum, int pid, int tid)
+{
+	struct procfs_list_entry *e;
+	char name[12];
+
+	if(!(e = find_pid_entry(osnum, pid)))
+		return NULL;
+	if(!(e = find_procfs_entry(e, "task")))
+		return NULL;
+	sprintf(name, "%d", tid);
+	return find_procfs_entry(e, name);
+}
+
+static struct procfs_list_entry *
+get_base_entry(int osnum)
+{
+	struct procfs_list_entry *e;
+	char name[12];
+	kuid_t uid = KUIDT_INIT(0);
+	kgid_t gid = KGIDT_INIT(0);
+
+	sprintf(name, "mcos%d", osnum);
+	e = find_procfs_entry(NULL, name);
+	if(!e){
+		e = add_procfs_entry(NULL, name, S_IFDIR | 0555,
+		                     uid, gid, NULL);
+		e->osnum = osnum;
+	}
+	return e;
+}
+
+static struct procfs_list_entry *
+get_pid_entry(int osnum, int pid)
+{
+	struct procfs_list_entry *parent;
+	struct procfs_list_entry *e;
+	char name[12];
+	kuid_t uid = KUIDT_INIT(0);
+	kgid_t gid = KGIDT_INIT(0);
+
+	sprintf(name, "mcos%d", osnum);
+	if(!(parent = find_procfs_entry(NULL, name)))
+		return NULL;
+	sprintf(name, "%d", pid);
+	e = find_procfs_entry(parent, name);
+	if(!e)
+		e = add_procfs_entry(parent, name, S_IFDIR | 0555,
+		                     uid, gid, NULL);
+	return e;
+}
+
+static struct procfs_list_entry *
+get_tid_entry(int osnum, int pid, int tid)
+{
+	struct procfs_list_entry *parent;
+	struct procfs_list_entry *e;
+	char name[12];
+	kuid_t uid = KUIDT_INIT(0);
+	kgid_t gid = KGIDT_INIT(0);
+
+	sprintf(name, "mcos%d", osnum);
+	if(!(parent = find_procfs_entry(NULL, name)))
+		return NULL;
+	sprintf(name, "%d", pid);
+	if(!(parent = find_procfs_entry(parent, name)))
+		return NULL;
+	if(!(parent = find_procfs_entry(parent, "task")))
+		return NULL;
+	sprintf(name, "%d", tid);
+	e = find_procfs_entry(parent, name);
+	if(!e)
+		e = add_procfs_entry(parent, name, S_IFDIR | 0555,
+		                     uid, gid, NULL);
+	return e;
+}
+
+static void
+_add_tid_entry(int osnum, int pid, int tid, const struct cred *cred)
+{
+	struct procfs_list_entry *parent;
+	struct procfs_list_entry *exe;
+
+	parent = get_tid_entry(osnum, pid, tid);
+	if(parent){
+		add_procfs_entries(parent, tid_entry_stuff,
+		                   cred->uid, cred->gid);
+		exe = find_procfs_entry(parent->parent->parent, "exe");
+		if(exe){
+			add_procfs_entry(parent, "exe", S_IFLNK | 0777,
+			                 cred->uid, cred->gid, exe->data);
+		}
+		
+	}
+}
+
+void
+add_tid_entry(int osnum, int pid, int tid)
+{
+	const struct cred *cred = get_pid_cred(pid);
+
+	if(!cred)
+		return;
+	down(&procfs_file_list_lock);
+	_add_tid_entry(osnum, pid, tid, cred);
+	up(&procfs_file_list_lock);
+}
+
+void
+add_pid_entry(int osnum, int pid)
+{
+	struct procfs_list_entry *parent;
+	const struct cred *cred = get_pid_cred(pid);
+
+	if(!cred)
+		return;
+	down(&procfs_file_list_lock);
+	parent = get_pid_entry(osnum, pid);
+	add_procfs_entries(parent, pid_entry_stuff, cred->uid, cred->gid);
+	_add_tid_entry(osnum, pid, pid, cred);
+	up(&procfs_file_list_lock);
+}
+
+void
+delete_tid_entry(int osnum, int pid, int tid)
+{
+	struct procfs_list_entry *e;
+
+	down(&procfs_file_list_lock);
+	e = find_tid_entry(osnum, pid, tid);
+	if(e)
+		delete_procfs_entries(e);
+	up(&procfs_file_list_lock);
+}
+
+void
+delete_pid_entry(int osnum, int pid)
+{
+	struct procfs_list_entry *e;
+
+	down(&procfs_file_list_lock);
+	e = find_pid_entry(osnum, pid);
+	if(e)
+		delete_procfs_entries(e);
+	up(&procfs_file_list_lock);
+}
+
+void
+proc_exe_link(int osnum, int pid, const char *path)
+{
+	struct procfs_list_entry *parent;
+	kuid_t uid = KUIDT_INIT(0);
+	kgid_t gid = KGIDT_INIT(0);
+
+	down(&procfs_file_list_lock);
+	parent = find_pid_entry(osnum, pid);
+	if(parent){
+		struct procfs_list_entry *task;
+		struct procfs_list_entry *e;
+
+		e = add_procfs_entry(parent, "exe", S_IFLNK | 0777, uid, gid,
+		                     path);
+		e->data = kmalloc(strlen(path) + 1, GFP_KERNEL);
+		strcpy(e->data, path);
+		task = find_procfs_entry(parent, "task");
+		list_for_each_entry(parent, &task->children, list) {
+			add_procfs_entry(parent, "exe", S_IFLNK | 0777,
+			                 uid, gid, path);
+		}
+	}
+	up(&procfs_file_list_lock);
+}
+
+/**
+ * \brief Initialization for procfs
+ *
+ * \param osnum os number
+ */
+void
+procfs_init(int osnum)
+{
+	struct procfs_list_entry *parent;
+	kuid_t uid = KUIDT_INIT(0);
+	kgid_t gid = KGIDT_INIT(0);
+
+	down(&procfs_file_list_lock);
+	parent = get_base_entry(osnum);
+	add_procfs_entries(parent, base_entry_stuff, uid, gid);
+	up(&procfs_file_list_lock);
+}
+
+/**
+ * \brief Finalization for procfs
+ *
+ * \param osnum os number
+ */
+void
+procfs_exit(int osnum)
+{
+	struct procfs_list_entry *e;
+
+	down(&procfs_file_list_lock);
+	e = find_base_entry(osnum);
+	if (e) {
+		delete_procfs_entries(e);
+	}
+	up(&procfs_file_list_lock);
+}
+
+/**
+ * \brief The callback funciton for McKernel procfs
+ *
+ * This function conforms to the 2) way of fs/proc/generic.c
+ * from linux-2.6.39.4.
+ */
+static ssize_t
+mckernel_procfs_read(struct file *file, char __user *buf, size_t nbytes,
+	       loff_t *ppos)
+{
+	struct inode * inode = file->f_path.dentry->d_inode;
+	char *kern_buffer = NULL;
+	int order = 0;
+	volatile struct procfs_read *r = NULL;
+	struct ikc_scd_packet isp;
+	int ret;
+	unsigned long pbuf;
+	unsigned long count = nbytes;
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
+	struct proc_dir_entry *dp = PDE(inode);
+	struct procfs_list_entry *e = dp->data;
+#else	
+	struct procfs_list_entry *e = PDE_DATA(inode);
+#endif	
+	loff_t offset = *ppos;
+	char pathbuf[PROCFS_NAME_MAX];
+	char *path;
+
+	path = getpath(e, pathbuf, 256);
+	dprintk("mckernel_procfs_read: invoked for %s, offset: %lu, count: %d\n", 
+			path, offset, count); 
+	
+	if (count <= 0 || offset < 0) {
+		return 0;
+	}
+	
+	while ((1 << order) < count) ++order;
+	if (order > 12) {
+		order -= 12;
+	}
+	else {
+		order = 1;
+	}
+
+	/* NOTE: we need physically contigous memory to pass through IKC */
+	kern_buffer = (char *)__get_free_pages(GFP_KERNEL, order);
+	if (!kern_buffer) {
+		printk("mckernel_procfs_read(): ERROR: allocating kernel buffer\n");
+		return -ENOMEM;
+	}
+	
+	pbuf = virt_to_phys(kern_buffer);
+
+	r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
+	if (r == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	r->pbuf = pbuf;
+	r->eof = 0;
+	r->ret = -EIO; /* default */
+	r->status = 0;
+	r->offset = offset;
+	r->count = count;
+	r->readwrite = 0;
+	strncpy((char *)r->fname, path, PROCFS_NAME_MAX);
+	isp.msg = SCD_MSG_PROCFS_REQUEST;
+	isp.ref = 0;
+	isp.arg = virt_to_phys(r);
+	
+	ret = mcctrl_ikc_send(osnum_to_os(e->osnum), 0, &isp);
+	
+	if (ret < 0) {
+		goto out; /* error */
+	}
+	
+	/* Wait for a reply. */
+	ret = -EIO; /* default exit code */
+	dprintk("now wait for a relpy\n");
+	
+	/* Wait for the status field of the procfs_read structure set ready. */
+	if (wait_event_interruptible_timeout(procfsq, r->status != 0, HZ) == 0) {
+		kprintf("ERROR: mckernel_procfs_read: timeout (1 sec).\n");
+		goto out;
+	}
+	
+	/* Wake up and check the result. */
+	dprintk("mckernel_procfs_read: woke up. ret: %d, eof: %d\n", r->ret, r->eof);
+	
+	if (r->ret > 0) {
+		if (copy_to_user(buf, kern_buffer, r->ret)) {
+			kprintf("ERROR: mckernel_procfs_read: copy_to_user failed.\n");
+			ret = -EFAULT;
+			goto out;
+		}
+
+		*ppos += r->ret;
+	}
+	ret = r->ret;
+
+out:
+	if(kern_buffer)
+		free_pages((uintptr_t)kern_buffer, order);
+	if(r)
+		kfree((void *)r);
+	
+	return ret;
+}
+
+static ssize_t
+mckernel_procfs_write(struct file *file, const char __user *buf, size_t nbytes,
+	       loff_t *ppos)
+{
+	struct inode * inode = file->f_path.dentry->d_inode;
+	char *kern_buffer = NULL;
+	int order = 0;
+	volatile struct procfs_read *r = NULL;
+	struct ikc_scd_packet isp;
+	int ret;
+	unsigned long pbuf;
+	unsigned long count = nbytes;
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
+	struct proc_dir_entry *dp = PDE(inode);
+	struct procfs_list_entry *e = dp->data;
+#else	
+	struct procfs_list_entry *e = PDE_DATA(inode);
+#endif	
+	loff_t offset = *ppos;
+	char pathbuf[PROCFS_NAME_MAX];
+	char *path;
+
+	path = getpath(e, pathbuf, 256);
+	dprintk("mckernel_procfs_read: invoked for %s, offset: %lu, count: %d\n", 
+			path, offset, count); 
+	
+	if (count <= 0 || offset < 0) {
+		return 0;
+	}
+	
+	while ((1 << order) < count) ++order;
+	if (order > 12) {
+		order -= 12;
+	}
+	else {
+		order = 1;
+	}
+
+	/* NOTE: we need physically contigous memory to pass through IKC */
+	kern_buffer = (char *)__get_free_pages(GFP_KERNEL, order);
+	if (!kern_buffer) {
+		printk("mckernel_procfs_read(): ERROR: allocating kernel buffer\n");
+		return -ENOMEM;
+	}
+	if (copy_from_user(kern_buffer, buf, nbytes)) {
+		ret = -EFAULT;
+		goto out;
+	}
+	
+	pbuf = virt_to_phys(kern_buffer);
+
+	r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
+	if (r == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	dprintk("offset: %lx, count: %d, cpu: %d\n", offset, count, e->cpu);
+
+	r->pbuf = pbuf;
+	r->eof = 0;
+	r->ret = -EIO; /* default */
+	r->status = 0;
+	r->offset = offset;
+	r->count = count;
+	r->readwrite = 1;
+	strncpy((char *)r->fname, path, PROCFS_NAME_MAX);
+	isp.msg = SCD_MSG_PROCFS_REQUEST;
+	isp.ref = 0;
+	isp.arg = virt_to_phys(r);
+	
+	ret = mcctrl_ikc_send(osnum_to_os(e->osnum), 0, &isp);
+	
+	if (ret < 0) {
+		goto out; /* error */
+	}
+	
+	/* Wait for a reply. */
+	ret = -EIO; /* default exit code */
+	dprintk("now wait for a relpy\n");
+	
+	/* Wait for the status field of the procfs_read structure set ready. */
+	if (wait_event_interruptible_timeout(procfsq, r->status != 0, HZ) == 0) {
+		kprintf("ERROR: mckernel_procfs_read: timeout (1 sec).\n");
+		goto out;
+	}
+	
+	/* Wake up and check the result. */
+	dprintk("mckernel_procfs_read: woke up. ret: %d, eof: %d\n", r->ret, r->eof);
+	
+	if (r->ret > 0) {
+		*ppos += r->ret;
+	}
+	ret = r->ret;
+
+out:
+	if(kern_buffer)
+		free_pages((uintptr_t)kern_buffer, order);
+	if(r)
+		kfree((void *)r);
+	
+	return ret;
+}
+
+static loff_t
+mckernel_procfs_lseek(struct file *file, loff_t offset, int orig)
+{
+	switch (orig) {
+	case 0:
+		file->f_pos = offset;
+		break;
+	case 1:
+		file->f_pos += offset;
+		break;
+	default:
+		return -EINVAL;
+	}
+	return file->f_pos;
+}
+
+struct procfs_work {
+	void *os;
+	int msg;
+	int pid;
+	unsigned long arg;
+	struct work_struct work;
+};
+
+static void procfsm_work_main(struct work_struct *work0)
+{
+	struct procfs_work *work = container_of(work0, struct procfs_work, work);
+
+	switch (work->msg) {
+		case SCD_MSG_PROCFS_TID_CREATE:
+			add_tid_entry(ihk_host_os_get_index(work->os), work->pid, work->arg);
+			break;
+
+		case SCD_MSG_PROCFS_TID_DELETE:
+			delete_tid_entry(ihk_host_os_get_index(work->os), work->pid, work->arg);
+			break;
+
+		default:
+			printk("%s: unknown work: msg: %d, pid: %d, arg: %lu)\n",
+					__FUNCTION__, work->msg, work->pid, work->arg);
+			break;
+	}
+
+	kfree(work);
+	return;
+}
+
+int procfsm_packet_handler(void *os, int msg, int pid, unsigned long arg)
+{
+	struct procfs_work *work = NULL;
+
+	work = kzalloc(sizeof(*work), GFP_ATOMIC);
+	if (!work) {
+		printk("%s: kzalloc failed\n", __FUNCTION__);
+		return -1;
+	}
+
+	work->os = os;
+	work->msg = msg;
+	work->pid = pid;
+	work->arg = arg;
+	INIT_WORK(&work->work, &procfsm_work_main);
+
+	schedule_work(&work->work);
+	return 0;
+}
+
+static const struct file_operations mckernel_forward_ro = {
+	.llseek		= mckernel_procfs_lseek,
+	.read		= mckernel_procfs_read,
+	.write		= NULL,
+};
+
+static const struct file_operations mckernel_forward = {
+	.llseek		= mckernel_procfs_lseek,
+	.read		= mckernel_procfs_read,
+	.write		= mckernel_procfs_write,
+};
+
+static const struct procfs_entry tid_entry_stuff[] = {
+//	PROC_REG("auxv",       S_IRUSR, NULL),
+//	PROC_REG("clear_refs", S_IWUSR, NULL),
+//	PROC_REG("cmdline",    S_IRUGO, NULL),
+//	PROC_REG("comm",       S_IRUGO|S_IWUSR, NULL),
+//	PROC_REG("environ",    S_IRUSR, NULL),
+//	PROC_LNK("exe",        mckernel_readlink),
+//	PROC_REG("limits",     S_IRUSR|S_IWUSR, NULL),
+//	PROC_REG("maps",       S_IRUGO, NULL),
+	PROC_REG("mem",        S_IRUSR|S_IWUSR, NULL),
+//	PROC_REG("pagemap",    S_IRUGO, NULL),
+//	PROC_REG("smaps",      S_IRUGO, NULL),
+	PROC_REG("stat",       S_IRUGO, NULL),
+//	PROC_REG("statm",      S_IRUGO, NULL),
+//	PROC_REG("status",     S_IRUGO, NULL),
+//	PROC_REG("syscall",    S_IRUGO, NULL),
+//	PROC_REG("wchan",      S_IRUGO, NULL),
+	PROC_TERM
+};
+
+static const struct procfs_entry pid_entry_stuff[] = {
+	PROC_REG("auxv",       S_IRUSR, NULL),
+	PROC_REG("cgroup",     S_IXUSR, NULL),
+//	PROC_REG("clear_refs", S_IWUSR, NULL),
+	PROC_REG("cmdline",    S_IRUGO, NULL),
+//	PROC_REG("comm",       S_IRUGO|S_IWUSR, NULL),
+//	PROC_REG("coredump_filter", S_IRUGO|S_IWUSR, NULL),
+	PROC_REG("cpuset",     S_IXUSR, NULL),
+//	PROC_REG("environ",    S_IRUSR, NULL),
+//	PROC_LNK("exe",        mckernel_readlink),
+//	PROC_REG("limits",     S_IRUSR|S_IWUSR, NULL),
+	PROC_REG("maps",       S_IRUGO, NULL),
+	PROC_REG("mem",        S_IRUSR|S_IWUSR, NULL),
+	PROC_REG("pagemap",    S_IRUGO, NULL),
+	PROC_REG("smaps",      S_IRUGO, NULL),
+//	PROC_REG("stat",       S_IRUGO, NULL),
+//	PROC_REG("statm",      S_IRUGO, NULL),
+	PROC_REG("status",     S_IRUGO, NULL),
+//	PROC_REG("syscall",    S_IRUGO, NULL),
+	PROC_DIR("task",       S_IRUGO|S_IXUGO),
+//	PROC_REG("wchan",      S_IRUGO, NULL),
+	PROC_TERM
+};
+
+static const struct procfs_entry base_entry_stuff[] = {
+//	PROC_REG("cmdline",    S_IRUGO, NULL),
+//	PROC_REG("cpuinfo",    S_IRUGO, NULL),
+//	PROC_REG("meminfo",    S_IRUGO, NULL),
+//	PROC_REG("pagetypeinfo",S_IRUGO, NULL),
+//	PROC_REG("softirq",    S_IRUGO, NULL),
+	PROC_REG("stat",       S_IRUGO, NULL),
+//	PROC_REG("uptime",     S_IRUGO, NULL),
+//	PROC_REG("version",    S_IRUGO, NULL),
+//	PROC_REG("vmallocinfo",S_IRUSR, NULL),
+//	PROC_REG("vmstat",     S_IRUGO, NULL),
+//	PROC_REG("zoneinfo",   S_IRUGO, NULL),
+	PROC_TERM
+};

executer/kernel/mcctrl/sysfs.h

@@ -0,0 +1,73 @@
+/**
+ * \file sysfs.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  sysfs framework API definitions
+ * \author Gou Nakamura  <go.nakamura.yw@hitachi-solutions.com> \par
+ * 	Copyright (C) 2016  RIKEN AICS
+ */
+/*
+ * HISTORY:
+ */
+
+#ifndef MCCTRL_SYSFS_H
+#define MCCTRL_SYSFS_H
+
+#define SYSFS_PATH_MAX 1024
+
+/* for sysfs_unlinkf() */
+#define SYSFS_UNLINK_KEEP_ANCESTOR      0x01
+
+
+struct sysfsm_ops {
+	ssize_t (*show)(struct sysfsm_ops *ops, void *instance, void *buf,
+			size_t bufsize);
+	ssize_t (*store)(struct sysfsm_ops *ops, void *instance,
+			const void *buf, size_t bufsize);
+	void (*release)(struct sysfsm_ops *ops, void *instance);
+};
+
+struct sysfs_handle {
+	long handle;
+};
+typedef struct sysfs_handle sysfs_handle_t;
+
+struct sysfsm_bitmap_param {
+	int nbits;
+	int padding;
+	void *ptr;
+};
+
+#define SYSFS_SPECIAL_OPS_MIN ((void *)1)
+#define SYSFS_SPECIAL_OPS_MAX ((void *)1000)
+
+#define SYSFS_SNOOPING_OPS_d32 ((void *)1)
+#define SYSFS_SNOOPING_OPS_d64 ((void *)2)
+#define SYSFS_SNOOPING_OPS_u32 ((void *)3)
+#define SYSFS_SNOOPING_OPS_u64 ((void *)4)
+#define SYSFS_SNOOPING_OPS_s ((void *)5)
+#define SYSFS_SNOOPING_OPS_pbl ((void *)6)
+#define SYSFS_SNOOPING_OPS_pb ((void *)7)
+#define SYSFS_SNOOPING_OPS_u32K ((void *)8)
+
+static inline int is_special_sysfs_ops(void *ops)
+{
+	return (((long)SYSFS_SPECIAL_OPS_MIN <= (long)ops)
+			&& ((long)ops <= (long)SYSFS_SPECIAL_OPS_MAX));
+}
+
+extern int sysfsm_createf(ihk_os_t os, struct sysfsm_ops *ops, void *instance,
+		int mode, const char *fmt, ...);
+extern int sysfsm_mkdirf(ihk_os_t os, sysfs_handle_t *dirhp,
+		const char *fmt, ...);
+extern int sysfsm_symlinkf(ihk_os_t os, sysfs_handle_t targeth,
+		const char *fmt, ...);
+extern int sysfsm_lookupf(ihk_os_t os, sysfs_handle_t *objhp,
+		const char *fmt, ...);
+extern int sysfsm_unlinkf(ihk_os_t os, int flags, const char *fmt, ...);
+
+extern void sysfsm_cleanup(ihk_os_t os);
+extern void sysfsm_packet_handler(void *os, int msg, int err, long arg1,
+		long arg2);
+
+#endif /* MCCTRL_SYSFS_H */

executer/kernel/mcctrl/sysfs_msg.h

@@ -0,0 +1,88 @@
+/**
+ * \file sysfs_msg.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *   message declarations for sysfs framework
+ * \author Gou Nakamura  <go.nakamura.yw@hitachi-solutions.com> \par
+ * 	Copyright (C) 2015  RIKEN AICS
+ */
+/*
+ * HISTORY:
+ */
+
+#ifndef MCKERNEL_SYSFS_MSG_H
+#define MCKERNEL_SYSFS_MSG_H
+
+#define SYSFS_PATH_MAX 1024
+
+struct sysfs_req_create_param {
+	int mode;
+	int error;
+	long client_ops;
+	long client_instance;
+	char path[SYSFS_PATH_MAX];
+	int padding;
+	int busy;
+}; /* struct sysfs_req_create_param */
+
+#define SYSFS_SPECIAL_OPS_MIN ((void *)1)
+#define SYSFS_SPECIAL_OPS_MAX ((void *)1000)
+
+#define SYSFS_SNOOPING_OPS_d32 ((void *)1)
+#define SYSFS_SNOOPING_OPS_d64 ((void *)2)
+#define SYSFS_SNOOPING_OPS_u32 ((void *)3)
+#define SYSFS_SNOOPING_OPS_u64 ((void *)4)
+#define SYSFS_SNOOPING_OPS_s ((void *)5)
+#define SYSFS_SNOOPING_OPS_pbl ((void *)6)
+#define SYSFS_SNOOPING_OPS_pb ((void *)7)
+#define SYSFS_SNOOPING_OPS_u32K ((void *)8)
+
+struct sysfs_req_mkdir_param {
+	int error;
+	int padding;
+	long handle;
+	char path[SYSFS_PATH_MAX];
+	int padding2;
+	int busy;
+}; /* struct sysfs_req_mkdir_param */
+
+struct sysfs_req_symlink_param {
+	int error;
+	int padding;
+	long target;
+	char path[SYSFS_PATH_MAX];
+	int padding2;
+	int busy;
+}; /* struct sysfs_req_symlink_param */
+
+struct sysfs_req_lookup_param {
+	int error;
+	int padding;
+	long handle;
+	char path[SYSFS_PATH_MAX];
+	int padding2;
+	int busy;
+}; /* struct sysfs_req_lookup_param */
+
+/* for sysfs_req_unlink_param.flags */
+#define SYSFS_UNLINK_KEEP_ANCESTOR      0x01
+
+struct sysfs_req_unlink_param {
+	int flags;
+	int error;
+	char path[SYSFS_PATH_MAX];
+	int padding;
+	int busy;
+}; /* struct sysfs_req_unlink_param */
+
+struct sysfs_req_setup_param {
+	int error;
+	int padding;
+	long buf_rpa;
+	long bufsize;
+	char padding3[SYSFS_PATH_MAX];
+	int padding2;
+	int busy;
+}; /* struct sysfs_req_setup_param */
+
+#endif /* MCKERNEL_SYSFS_MSG_H */

executer/kernel/mcoverlayfs/Makefile.in

@@ -0,0 +1,40 @@
+ENABLE_MCOVERLAYFS=@ENABLE_MCOVERLAYFS@
+
+RELEASE=$(shell uname -r)
+MAJOR=$(shell echo ${RELEASE} | sed -e 's/^\([0-9]*\).*/\1/')
+MINOR=$(shell echo ${RELEASE} | sed -e 's/^[0-9]*.\([0-9]*\).*/\1/')
+PATCH=$(shell echo ${RELEASE} | sed -e 's/^[0-9]*.[0-9]*.\([0-9]*\).*/\1/')
+LINUX_VERSION_CODE=$(shell expr \( ${MAJOR} \* 65536 \) + \( ${MINOR} \* 256 \) + ${PATCH})
+RHEL_RELEASE_TMP=$(shell echo ${RELEASE} | sed -e 's/^[0-9]*.[0-9]*.[0-9]*-\([0-9]*\).*/\1/')
+RHEL_RELEASE=$(shell if [ "${RELEASE}" == "${RHEL_RELEASE_TMP}" ]; then echo ""; else echo ${RHEL_RELEASE_TMP}; fi)
+BUILD_MODULE_TMP=$(shell if [ "${RHEL_RELEASE}" == "" ]; then echo "org"; else echo "rhel"; fi)
+BUILD_MODULE=none
+ifeq ($(ENABLE_MCOVERLAYFS),yes)
+ifeq ($(BUILD_MODULE_TMP),org)
+ifeq ($(BUILD_MODULE),none)
+BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -ge 262144 -a ${LINUX_VERSION_CODE} -lt 262400 ]; then echo "linux-4.0.9"; else echo "none"; fi)
+endif
+endif
+ifeq ($(BUILD_MODULE_TMP),rhel)
+ifeq ($(BUILD_MODULE),none)
+BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -eq 199168 -a ${RHEL_RELEASE} -eq 327 ]; then echo "linux-3.10.0-327.36.1.el7"; else echo "none"; fi)
+endif
+endif
+endif
+
+.PHONY: clean install modules
+
+modules:
+ifneq ($(BUILD_MODULE),none)
+	@(cd $(BUILD_MODULE); make modules)
+endif
+
+clean:
+	@(cd linux-3.10.0-327.36.1.el7; make clean)
+	@(cd linux-4.0.9; make clean)
+
+install:
+ifneq ($(BUILD_MODULE),none)
+	@(cd $(BUILD_MODULE); make install)
+endif
+

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

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

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

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

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

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

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

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

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

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

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

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

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

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

executer/kernel/mcoverlayfs/linux-4.0.9/copy_up.c

@@ -0,0 +1,416 @@
+/*
+ *
+ * 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/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 "overlayfs.h"
+
+#define OVL_COPY_UP_CHUNK_SIZE (1 << 20)
+
+int ovl_copy_xattr(struct dentry *old, struct dentry *new)
+{
+	ssize_t list_size, size;
+	char *buf, *name, *value;
+	int 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;
+
+	error = -ENOMEM;
+	value = kmalloc(XATTR_SIZE_MAX, GFP_KERNEL);
+	if (!value)
+		goto out;
+
+	list_size = vfs_listxattr(old, buf, list_size);
+	if (list_size <= 0) {
+		error = list_size;
+		goto out_free_value;
+	}
+
+	for (name = buf; name < (buf + list_size); name += strlen(name) + 1) {
+		size = vfs_getxattr(old, name, value, XATTR_SIZE_MAX);
+		if (size <= 0) {
+			error = size;
+			goto out_free_value;
+		}
+		error = vfs_setxattr(new, name, value, size, 0);
+		if (error)
+			goto out_free_value;
+	}
+
+out_free_value:
+	kfree(value);
+out:
+	kfree(buf);
+	return error;
+}
+
+static int ovl_copy_up_data(struct path *old, struct path *new, loff_t len)
+{
+	struct file *old_file;
+	struct file *new_file;
+	loff_t old_pos = 0;
+	loff_t new_pos = 0;
+	int error = 0;
+
+	if (len == 0)
+		return 0;
+
+	old_file = ovl_path_open(old, O_RDONLY);
+	if (IS_ERR(old_file))
+		return PTR_ERR(old_file);
+
+	new_file = ovl_path_open(new, O_WRONLY);
+	if (IS_ERR(new_file)) {
+		error = PTR_ERR(new_file);
+		goto out_fput;
+	}
+
+	/* FIXME: copy up sparse files efficiently */
+	while (len) {
+		size_t this_len = OVL_COPY_UP_CHUNK_SIZE;
+		long bytes;
+
+		if (len < this_len)
+			this_len = len;
+
+		if (signal_pending_state(TASK_KILLABLE, current)) {
+			error = -EINTR;
+			break;
+		}
+
+		bytes = do_splice_direct(old_file, &old_pos,
+					 new_file, &new_pos,
+					 this_len, SPLICE_F_MOVE);
+		if (bytes <= 0) {
+			error = bytes;
+			break;
+		}
+		WARN_ON(old_pos != new_pos);
+
+		len -= bytes;
+	}
+
+	fput(new_file);
+out_fput:
+	fput(old_file);
+	return error;
+}
+
+static char *ovl_read_symlink(struct dentry *realdentry)
+{
+	int res;
+	char *buf;
+	struct inode *inode = realdentry->d_inode;
+	mm_segment_t old_fs;
+
+	res = -EINVAL;
+	if (!inode->i_op->readlink)
+		goto err;
+
+	res = -ENOMEM;
+	buf = (char *) __get_free_page(GFP_KERNEL);
+	if (!buf)
+		goto err;
+
+	old_fs = get_fs();
+	set_fs(get_ds());
+	/* The cast to a user pointer is valid due to the set_fs() */
+	res = inode->i_op->readlink(realdentry,
+				    (char __user *)buf, PAGE_SIZE - 1);
+	set_fs(old_fs);
+	if (res < 0) {
+		free_page((unsigned long) buf);
+		goto err;
+	}
+	buf[res] = '\0';
+
+	return buf;
+
+err:
+	return ERR_PTR(res);
+}
+
+static int ovl_set_timestamps(struct dentry *upperdentry, struct kstat *stat)
+{
+	struct iattr attr = {
+		.ia_valid =
+		     ATTR_ATIME | ATTR_MTIME | ATTR_ATIME_SET | ATTR_MTIME_SET,
+		.ia_atime = stat->atime,
+		.ia_mtime = stat->mtime,
+	};
+
+	return notify_change(upperdentry, &attr, NULL);
+}
+
+int ovl_set_attr(struct dentry *upperdentry, struct kstat *stat)
+{
+	int err = 0;
+
+	if (!S_ISLNK(stat->mode)) {
+		struct iattr attr = {
+			.ia_valid = ATTR_MODE,
+			.ia_mode = stat->mode,
+		};
+		err = notify_change(upperdentry, &attr, NULL);
+	}
+	if (!err) {
+		struct iattr attr = {
+			.ia_valid = ATTR_UID | ATTR_GID,
+			.ia_uid = stat->uid,
+			.ia_gid = stat->gid,
+		};
+		err = notify_change(upperdentry, &attr, NULL);
+	}
+	if (!err)
+		ovl_set_timestamps(upperdentry, stat);
+
+	return err;
+}
+
+static int ovl_copy_up_locked(struct dentry *workdir, struct dentry *upperdir,
+			      struct dentry *dentry, struct path *lowerpath,
+			      struct kstat *stat, struct iattr *attr,
+			      const char *link)
+{
+	struct inode *wdir = workdir->d_inode;
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *newdentry = NULL;
+	struct dentry *upper = NULL;
+	umode_t mode = stat->mode;
+	int err;
+
+	newdentry = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out;
+
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out1;
+
+	/* Can't properly set mode on creation because of the umask */
+	stat->mode &= S_IFMT;
+	err = ovl_create_real(wdir, newdentry, stat, link, NULL, true);
+	stat->mode = mode;
+	if (err)
+		goto out2;
+
+	if (S_ISREG(stat->mode)) {
+		struct path upperpath;
+		ovl_path_upper(dentry, &upperpath);
+		BUG_ON(upperpath.dentry != NULL);
+		upperpath.dentry = newdentry;
+
+		err = ovl_copy_up_data(lowerpath, &upperpath, stat->size);
+		if (err)
+			goto out_cleanup;
+	}
+
+	err = ovl_copy_xattr(lowerpath->dentry, newdentry);
+	if (err)
+		goto out_cleanup;
+
+	mutex_lock(&newdentry->d_inode->i_mutex);
+	err = ovl_set_attr(newdentry, stat);
+	if (!err && attr)
+		err = notify_change(newdentry, attr, NULL);
+	mutex_unlock(&newdentry->d_inode->i_mutex);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_do_rename(wdir, newdentry, udir, upper, 0);
+	if (err)
+		goto out_cleanup;
+
+	ovl_dentry_update(dentry, newdentry);
+	newdentry = NULL;
+
+	/*
+	 * Non-directores become opaque when copied up.
+	 */
+	if (!S_ISDIR(stat->mode))
+		ovl_dentry_set_opaque(dentry, true);
+out2:
+	dput(upper);
+out1:
+	dput(newdentry);
+out:
+	return err;
+
+out_cleanup:
+	ovl_cleanup(wdir, newdentry);
+	goto out;
+}
+
+/*
+ * Copy up a single dentry
+ *
+ * Directory renames only allowed on "pure upper" (already created on
+ * upper filesystem, never copied up).  Directories which are on lower or
+ * are merged may not be renamed.  For these -EXDEV is returned and
+ * userspace has to deal with it.  This means, when copying up a
+ * directory we can rely on it and ancestors being stable.
+ *
+ * Non-directory renames start with copy up of source if necessary.  The
+ * actual rename will only proceed once the copy up was successful.  Copy
+ * up uses upper parent i_mutex for exclusion.  Since rename can change
+ * d_parent it is possible that the copy up will lock the old parent.  At
+ * that point the file will have already been copied up anyway.
+ */
+int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry,
+		    struct path *lowerpath, struct kstat *stat,
+		    struct iattr *attr)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	int err;
+	struct kstat pstat;
+	struct path parentpath;
+	struct dentry *upperdir;
+	struct dentry *upperdentry;
+	const struct cred *old_cred;
+	struct cred *override_cred;
+	char *link = NULL;
+
+	if (WARN_ON(!workdir))
+		return -EROFS;
+
+	ovl_path_upper(parent, &parentpath);
+	upperdir = parentpath.dentry;
+
+	err = vfs_getattr(&parentpath, &pstat);
+	if (err)
+		return err;
+
+	if (S_ISLNK(stat->mode)) {
+		link = ovl_read_symlink(lowerpath->dentry);
+		if (IS_ERR(link))
+			return PTR_ERR(link);
+	}
+
+	err = -ENOMEM;
+	override_cred = prepare_creds();
+	if (!override_cred)
+		goto out_free_link;
+
+	override_cred->fsuid = stat->uid;
+	override_cred->fsgid = stat->gid;
+	/*
+	 * CAP_SYS_ADMIN for copying up extended attributes
+	 * CAP_DAC_OVERRIDE for create
+	 * CAP_FOWNER for chmod, timestamp update
+	 * CAP_FSETID for chmod
+	 * CAP_CHOWN for chown
+	 * CAP_MKNOD for mknod
+	 */
+	cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+	cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+	cap_raise(override_cred->cap_effective, CAP_FOWNER);
+	cap_raise(override_cred->cap_effective, CAP_FSETID);
+	cap_raise(override_cred->cap_effective, CAP_CHOWN);
+	cap_raise(override_cred->cap_effective, CAP_MKNOD);
+	old_cred = override_creds(override_cred);
+
+	err = -EIO;
+	if (lock_rename(workdir, upperdir) != NULL) {
+		pr_err("overlayfs: failed to lock workdir+upperdir\n");
+		goto out_unlock;
+	}
+	upperdentry = ovl_dentry_upper(dentry);
+	if (upperdentry) {
+		unlock_rename(workdir, upperdir);
+		err = 0;
+		/* Raced with another copy-up?  Do the setattr here */
+		if (attr) {
+			mutex_lock(&upperdentry->d_inode->i_mutex);
+			err = notify_change(upperdentry, attr, NULL);
+			mutex_unlock(&upperdentry->d_inode->i_mutex);
+		}
+		goto out_put_cred;
+	}
+
+	err = ovl_copy_up_locked(workdir, upperdir, dentry, lowerpath,
+				 stat, attr, link);
+	if (!err) {
+		/* Restore timestamps on parent (best effort) */
+		ovl_set_timestamps(upperdir, &pstat);
+	}
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out_put_cred:
+	revert_creds(old_cred);
+	put_cred(override_cred);
+
+out_free_link:
+	if (link)
+		free_page((unsigned long) link);
+
+	return err;
+}
+
+int ovl_copy_up(struct dentry *dentry)
+{
+	int err;
+
+	err = 0;
+	while (!err) {
+		struct dentry *next;
+		struct dentry *parent;
+		struct path lowerpath;
+		struct kstat stat;
+		enum ovl_path_type type = ovl_path_type(dentry);
+
+		if (OVL_TYPE_UPPER(type))
+			break;
+
+		next = dget(dentry);
+		/* find the topmost dentry not yet copied up */
+		for (;;) {
+			parent = dget_parent(next);
+
+			type = ovl_path_type(parent);
+			if (OVL_TYPE_UPPER(type))
+				break;
+
+			dput(next);
+			next = parent;
+		}
+
+		ovl_path_lower(next, &lowerpath);
+		err = vfs_getattr(&lowerpath, &stat);
+		if (!err)
+			err = ovl_copy_up_one(parent, next, &lowerpath, &stat, NULL);
+
+		dput(parent);
+		dput(next);
+	}
+
+	return err;
+}

executer/kernel/mcoverlayfs/linux-4.0.9/dir.c

@@ -0,0 +1,951 @@
+/*
+ *
+ * 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;
+
+	mutex_lock_nested(&udir->i_mutex, I_MUTEX_PARENT);
+	newdentry = lookup_one_len(dentry->d_name.name, upperdir,
+				   dentry->d_name.len);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out_unlock;
+	err = ovl_create_real(udir, newdentry, stat, link, hardlink, false);
+	if (err)
+		goto out_dput;
+
+	ovl_dentry_version_inc(dentry->d_parent);
+	ovl_dentry_update(dentry, newdentry);
+	ovl_copyattr(newdentry->d_inode, inode);
+	d_instantiate(dentry, inode);
+	newdentry = NULL;
+out_dput:
+	dput(newdentry);
+out_unlock:
+	mutex_unlock(&udir->i_mutex);
+	return err;
+}
+
+static int ovl_lock_rename_workdir(struct dentry *workdir,
+				   struct dentry *upperdir)
+{
+	/* Workdir should not be the same as upperdir */
+	if (workdir == upperdir)
+		goto err;
+
+	/* Workdir should not be subdir of upperdir and vice versa */
+	if (lock_rename(workdir, upperdir) != NULL)
+		goto err_unlock;
+
+	return 0;
+
+err_unlock:
+	unlock_rename(workdir, upperdir);
+err:
+	pr_err("overlayfs: failed to lock workdir+upperdir\n");
+	return -EIO;
+}
+
+static struct dentry *ovl_clear_empty(struct dentry *dentry,
+				      struct list_head *list)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct path upperpath;
+	struct dentry *upper;
+	struct dentry *opaquedir;
+	struct kstat stat;
+	int err;
+
+	if (WARN_ON(!workdir))
+		return ERR_PTR(-EROFS);
+
+	err = ovl_lock_rename_workdir(workdir, upperdir);
+	if (err)
+		goto out;
+
+	ovl_path_upper(dentry, &upperpath);
+	err = vfs_getattr(&upperpath, &stat);
+	if (err)
+		goto out_unlock;
+
+	err = -ESTALE;
+	if (!S_ISDIR(stat.mode))
+		goto out_unlock;
+	upper = upperpath.dentry;
+	if (upper->d_parent->d_inode != udir)
+		goto out_unlock;
+
+	opaquedir = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(opaquedir);
+	if (IS_ERR(opaquedir))
+		goto out_unlock;
+
+	err = ovl_create_real(wdir, opaquedir, &stat, NULL, NULL, true);
+	if (err)
+		goto out_dput;
+
+	err = ovl_copy_xattr(upper, opaquedir);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_set_opaque(opaquedir);
+	if (err)
+		goto out_cleanup;
+
+	mutex_lock(&opaquedir->d_inode->i_mutex);
+	err = ovl_set_attr(opaquedir, &stat);
+	mutex_unlock(&opaquedir->d_inode->i_mutex);
+	if (err)
+		goto out_cleanup;
+
+	err = ovl_do_rename(wdir, opaquedir, udir, upper, RENAME_EXCHANGE);
+	if (err)
+		goto out_cleanup;
+
+	ovl_cleanup_whiteouts(upper, list);
+	ovl_cleanup(wdir, upper);
+	unlock_rename(workdir, upperdir);
+
+	/* dentry's upper doesn't match now, get rid of it */
+	d_drop(dentry);
+
+	return opaquedir;
+
+out_cleanup:
+	ovl_cleanup(wdir, opaquedir);
+out_dput:
+	dput(opaquedir);
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out:
+	return ERR_PTR(err);
+}
+
+static struct dentry *ovl_check_empty_and_clear(struct dentry *dentry)
+{
+	int err;
+	struct dentry *ret = NULL;
+	LIST_HEAD(list);
+
+	err = ovl_check_empty_dir(dentry, &list);
+	if (err)
+		ret = ERR_PTR(err);
+	else {
+		/*
+		 * If no upperdentry then skip clearing whiteouts.
+		 *
+		 * Can race with copy-up, since we don't hold the upperdir
+		 * mutex.  Doesn't matter, since copy-up can't create a
+		 * non-empty directory from an empty one.
+		 */
+		if (ovl_dentry_upper(dentry))
+			ret = ovl_clear_empty(dentry, &list);
+	}
+
+	ovl_cache_free(&list);
+
+	return ret;
+}
+
+static int ovl_create_over_whiteout(struct dentry *dentry, struct inode *inode,
+				    struct kstat *stat, const char *link,
+				    struct dentry *hardlink)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *upper;
+	struct dentry *newdentry;
+	int err;
+
+	if (WARN_ON(!workdir))
+		return -EROFS;
+
+	err = ovl_lock_rename_workdir(workdir, upperdir);
+	if (err)
+		goto out;
+
+	newdentry = ovl_lookup_temp(workdir, dentry);
+	err = PTR_ERR(newdentry);
+	if (IS_ERR(newdentry))
+		goto out_unlock;
+
+	upper = lookup_one_len(dentry->d_name.name, upperdir,
+			       dentry->d_name.len);
+	err = PTR_ERR(upper);
+	if (IS_ERR(upper))
+		goto out_dput;
+
+	err = ovl_create_real(wdir, newdentry, stat, link, hardlink, true);
+	if (err)
+		goto out_dput2;
+
+	if (S_ISDIR(stat->mode)) {
+		err = ovl_set_opaque(newdentry);
+		if (err)
+			goto out_cleanup;
+
+		err = ovl_do_rename(wdir, newdentry, udir, upper,
+				    RENAME_EXCHANGE);
+		if (err)
+			goto out_cleanup;
+
+		ovl_cleanup(wdir, upper);
+	} else {
+		err = ovl_do_rename(wdir, newdentry, udir, upper, 0);
+		if (err)
+			goto out_cleanup;
+	}
+	ovl_dentry_version_inc(dentry->d_parent);
+	ovl_dentry_update(dentry, newdentry);
+	ovl_copyattr(newdentry->d_inode, inode);
+	d_instantiate(dentry, inode);
+	newdentry = NULL;
+out_dput2:
+	dput(upper);
+out_dput:
+	dput(newdentry);
+out_unlock:
+	unlock_rename(workdir, upperdir);
+out:
+	return err;
+
+out_cleanup:
+	ovl_cleanup(wdir, newdentry);
+	goto out_dput2;
+}
+
+static int ovl_create_or_link(struct dentry *dentry, int mode, dev_t rdev,
+			      const char *link, struct dentry *hardlink)
+{
+	int err;
+	struct inode *inode;
+	struct kstat stat = {
+		.mode = mode,
+		.rdev = rdev,
+	};
+
+	err = -ENOMEM;
+	inode = ovl_new_inode(dentry->d_sb, mode, dentry->d_fsdata);
+	if (!inode)
+		goto out;
+
+	err = ovl_copy_up(dentry->d_parent);
+	if (err)
+		goto out_iput;
+
+	if (!ovl_dentry_is_opaque(dentry)) {
+		err = ovl_create_upper(dentry, inode, &stat, link, hardlink);
+	} else {
+		const struct cred *old_cred;
+		struct cred *override_cred;
+
+		err = -ENOMEM;
+		override_cred = prepare_creds();
+		if (!override_cred)
+			goto out_iput;
+
+		/*
+		 * CAP_SYS_ADMIN for setting opaque xattr
+		 * CAP_DAC_OVERRIDE for create in workdir, rename
+		 * CAP_FOWNER for removing whiteout from sticky dir
+		 */
+		cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		cap_raise(override_cred->cap_effective, CAP_FOWNER);
+		old_cred = override_creds(override_cred);
+
+		err = ovl_create_over_whiteout(dentry, inode, &stat, link,
+					       hardlink);
+
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+
+	if (!err)
+		inode = NULL;
+out_iput:
+	iput(inode);
+out:
+	return err;
+}
+
+static int ovl_create_object(struct dentry *dentry, int mode, dev_t rdev,
+			     const char *link)
+{
+	int err;
+
+	err = ovl_want_write(dentry);
+	if (!err) {
+		err = ovl_create_or_link(dentry, mode, rdev, link, NULL);
+		ovl_drop_write(dentry);
+	}
+
+	return err;
+}
+
+static int ovl_create(struct inode *dir, struct dentry *dentry, umode_t mode,
+		      bool excl)
+{
+	return ovl_create_object(dentry, (mode & 07777) | S_IFREG, 0, NULL);
+}
+
+static int ovl_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
+{
+	return ovl_create_object(dentry, (mode & 07777) | S_IFDIR, 0, NULL);
+}
+
+static int ovl_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
+		     dev_t rdev)
+{
+	/* Don't allow creation of "whiteout" on overlay */
+	if (S_ISCHR(mode) && rdev == WHITEOUT_DEV)
+		return -EPERM;
+
+	return ovl_create_object(dentry, mode, rdev, NULL);
+}
+
+static int ovl_symlink(struct inode *dir, struct dentry *dentry,
+		       const char *link)
+{
+	return ovl_create_object(dentry, S_IFLNK, 0, link);
+}
+
+static int ovl_link(struct dentry *old, struct inode *newdir,
+		    struct dentry *new)
+{
+	int err;
+	struct dentry *upper;
+
+	err = ovl_want_write(old);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(old);
+	if (err)
+		goto out_drop_write;
+
+	upper = ovl_dentry_upper(old);
+	err = ovl_create_or_link(new, upper->d_inode->i_mode, 0, NULL, upper);
+
+out_drop_write:
+	ovl_drop_write(old);
+out:
+	return err;
+}
+
+static int ovl_remove_and_whiteout(struct dentry *dentry, bool is_dir)
+{
+	struct dentry *workdir = ovl_workdir(dentry);
+	struct inode *wdir = workdir->d_inode;
+	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
+	struct inode *udir = upperdir->d_inode;
+	struct dentry *whiteout;
+	struct dentry *upper;
+	struct dentry *opaquedir = NULL;
+	int err;
+
+	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;
+
+	whiteout = ovl_whiteout(workdir, dentry);
+	err = PTR_ERR(whiteout);
+	if (IS_ERR(whiteout))
+		goto out_unlock;
+
+	upper = ovl_dentry_upper(dentry);
+	if (!upper) {
+		upper = lookup_one_len(dentry->d_name.name, upperdir,
+				       dentry->d_name.len);
+		err = PTR_ERR(upper);
+		if (IS_ERR(upper))
+			goto kill_whiteout;
+
+		err = ovl_do_rename(wdir, whiteout, udir, upper, 0);
+		dput(upper);
+		if (err)
+			goto kill_whiteout;
+	} else {
+		int flags = 0;
+
+		if (opaquedir)
+			upper = opaquedir;
+		err = -ESTALE;
+		if (upper->d_parent != upperdir)
+			goto kill_whiteout;
+
+		if (is_dir)
+			flags |= RENAME_EXCHANGE;
+
+		err = ovl_do_rename(wdir, whiteout, udir, upper, flags);
+		if (err)
+			goto kill_whiteout;
+
+		if (is_dir)
+			ovl_cleanup(wdir, upper);
+	}
+	ovl_dentry_version_inc(dentry->d_parent);
+out_d_drop:
+	d_drop(dentry);
+	dput(whiteout);
+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 = ovl_dentry_upper(dentry);
+	int err;
+
+	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
+	err = -ESTALE;
+	if (upper->d_parent == upperdir) {
+		/* Don't let d_delete() think it can reset d_inode */
+		dget(upper);
+		if (is_dir)
+			err = vfs_rmdir(dir, upper);
+		else
+			err = vfs_unlink(dir, upper, NULL);
+		dput(upper);
+		ovl_dentry_version_inc(dentry->d_parent);
+	}
+
+	/*
+	 * 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.
+	 */
+	d_drop(dentry);
+	mutex_unlock(&dir->i_mutex);
+
+	return err;
+}
+
+static inline int ovl_check_sticky(struct dentry *dentry)
+{
+	struct inode *dir = ovl_dentry_real(dentry->d_parent)->d_inode;
+	struct inode *inode = ovl_dentry_real(dentry)->d_inode;
+
+	if (check_sticky(dir, inode))
+		return -EPERM;
+
+	return 0;
+}
+
+static int ovl_do_remove(struct dentry *dentry, bool is_dir)
+{
+	enum ovl_path_type type;
+	int err;
+
+	err = ovl_check_sticky(dentry);
+	if (err)
+		goto out;
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = ovl_copy_up(dentry->d_parent);
+	if (err)
+		goto out_drop_write;
+
+	type = ovl_path_type(dentry);
+	if (OVL_TYPE_PURE_UPPER(type)) {
+		err = ovl_remove_upper(dentry, is_dir);
+	} else {
+		const struct cred *old_cred;
+		struct cred *override_cred;
+
+		err = -ENOMEM;
+		override_cred = prepare_creds();
+		if (!override_cred)
+			goto out_drop_write;
+
+		/*
+		 * CAP_SYS_ADMIN for setting xattr on whiteout, opaque dir
+		 * CAP_DAC_OVERRIDE for create in workdir, rename
+		 * CAP_FOWNER for removing whiteout from sticky dir
+		 * CAP_FSETID for chmod of opaque dir
+		 * CAP_CHOWN for chown of opaque dir
+		 */
+		cap_raise(override_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		cap_raise(override_cred->cap_effective, CAP_FOWNER);
+		cap_raise(override_cred->cap_effective, CAP_FSETID);
+		cap_raise(override_cred->cap_effective, CAP_CHOWN);
+		old_cred = override_creds(override_cred);
+
+		err = ovl_remove_and_whiteout(dentry, is_dir);
+
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static int ovl_unlink(struct inode *dir, struct dentry *dentry)
+{
+	return ovl_do_remove(dentry, false);
+}
+
+static int ovl_rmdir(struct inode *dir, struct dentry *dentry)
+{
+	return ovl_do_remove(dentry, true);
+}
+
+static int ovl_rename2(struct inode *olddir, struct dentry *old,
+		       struct inode *newdir, struct dentry *new,
+		       unsigned int flags)
+{
+	int err;
+	enum ovl_path_type old_type;
+	enum ovl_path_type new_type;
+	struct dentry *old_upperdir;
+	struct dentry *new_upperdir;
+	struct dentry *olddentry;
+	struct dentry *newdentry;
+	struct dentry *trap;
+	bool old_opaque;
+	bool new_opaque;
+	bool new_create = false;
+	bool cleanup_whiteout = false;
+	bool overwrite = !(flags & RENAME_EXCHANGE);
+	bool is_dir = 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 = ovl_dentry_upper(old);
+	newdentry = ovl_dentry_upper(new);
+	if (newdentry) {
+		if (opaquedir) {
+			newdentry = opaquedir;
+			opaquedir = NULL;
+		} else {
+			dget(newdentry);
+		}
+	} else {
+		new_create = true;
+		newdentry = lookup_one_len(new->d_name.name, new_upperdir,
+					   new->d_name.len);
+		err = PTR_ERR(newdentry);
+		if (IS_ERR(newdentry))
+			goto out_unlock;
+	}
+
+	err = -ESTALE;
+	if (olddentry->d_parent != old_upperdir)
+		goto out_dput;
+	if (newdentry->d_parent != new_upperdir)
+		goto out_dput;
+	if (olddentry == trap)
+		goto out_dput;
+	if (newdentry == trap)
+		goto out_dput;
+
+	if (is_dir && !old_opaque && new_opaque) {
+		err = ovl_set_opaque(olddentry);
+		if (err)
+			goto out_dput;
+	}
+	if (!overwrite && new_is_dir && old_opaque && !new_opaque) {
+		err = ovl_set_opaque(newdentry);
+		if (err)
+			goto out_dput;
+	}
+
+	if (old_opaque || new_opaque) {
+		err = ovl_do_rename(old_upperdir->d_inode, olddentry,
+				    new_upperdir->d_inode, newdentry,
+				    flags);
+	} else {
+		/* No debug for the plain case */
+		BUG_ON(flags & ~RENAME_EXCHANGE);
+		err = vfs_rename(old_upperdir->d_inode, olddentry,
+				 new_upperdir->d_inode, newdentry,
+				 NULL, flags);
+	}
+
+	if (err) {
+		if (is_dir && !old_opaque && new_opaque)
+			ovl_remove_opaque(olddentry);
+		if (!overwrite && new_is_dir && old_opaque && !new_opaque)
+			ovl_remove_opaque(newdentry);
+		goto out_dput;
+	}
+
+	if (is_dir && old_opaque && !new_opaque)
+		ovl_remove_opaque(olddentry);
+	if (!overwrite && new_is_dir && !old_opaque && new_opaque)
+		ovl_remove_opaque(newdentry);
+
+	if (old_opaque != new_opaque) {
+		ovl_dentry_set_opaque(old, new_opaque);
+		if (!overwrite)
+			ovl_dentry_set_opaque(new, old_opaque);
+	}
+
+	if (cleanup_whiteout)
+		ovl_cleanup(old_upperdir->d_inode, newdentry);
+
+	ovl_dentry_version_inc(old->d_parent);
+	ovl_dentry_version_inc(new->d_parent);
+
+out_dput:
+	dput(newdentry);
+out_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-4.0.9/inode.c

@@ -0,0 +1,438 @@
+/*
+ *
+ * Copyright (C) 2011 Novell Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/xattr.h>
+#include "overlayfs.h"
+
+static int ovl_copy_up_last(struct dentry *dentry, struct iattr *attr,
+			    bool no_data)
+{
+	int err;
+	struct dentry *parent;
+	struct kstat stat;
+	struct path lowerpath;
+
+	parent = dget_parent(dentry);
+	err = ovl_copy_up(parent);
+	if (err)
+		goto out_dput_parent;
+
+	ovl_path_lower(dentry, &lowerpath);
+	err = vfs_getattr(&lowerpath, &stat);
+	if (err)
+		goto out_dput_parent;
+
+	if (no_data)
+		stat.size = 0;
+
+	err = ovl_copy_up_one(parent, dentry, &lowerpath, &stat, attr);
+
+out_dput_parent:
+	dput(parent);
+	return err;
+}
+
+int ovl_setattr(struct dentry *dentry, struct iattr *attr)
+{
+	int err;
+	struct dentry *upperdentry;
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	upperdentry = ovl_dentry_upper(dentry);
+	if (upperdentry) {
+		mutex_lock(&upperdentry->d_inode->i_mutex);
+		err = notify_change(upperdentry, attr, NULL);
+		mutex_unlock(&upperdentry->d_inode->i_mutex);
+	} else {
+		err = ovl_copy_up_last(dentry, attr, false);
+	}
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static int ovl_getattr(struct vfsmount *mnt, struct dentry *dentry,
+			 struct kstat *stat)
+{
+	struct path realpath;
+
+	ovl_path_real(dentry, &realpath);
+	return vfs_getattr(&realpath, stat);
+}
+
+int ovl_permission(struct inode *inode, int mask)
+{
+	struct ovl_entry *oe;
+	struct dentry *alias = NULL;
+	struct inode *realinode;
+	struct dentry *realdentry;
+	bool is_upper;
+	int err;
+
+	if (S_ISDIR(inode->i_mode)) {
+		oe = inode->i_private;
+	} else if (mask & MAY_NOT_BLOCK) {
+		return -ECHILD;
+	} else {
+		/*
+		 * For non-directories find an alias and get the info
+		 * from there.
+		 */
+		alias = d_find_any_alias(inode);
+		if (WARN_ON(!alias))
+			return -ENOENT;
+
+		oe = alias->d_fsdata;
+	}
+
+	realdentry = ovl_entry_real(oe, &is_upper);
+
+	/* Careful in RCU walk mode */
+	realinode = ACCESS_ONCE(realdentry->d_inode);
+	if (!realinode) {
+		WARN_ON(!(mask & MAY_NOT_BLOCK));
+		err = -ENOENT;
+		goto out_dput;
+	}
+
+	if (mask & MAY_WRITE) {
+		umode_t mode = realinode->i_mode;
+
+		/*
+		 * Writes will always be redirected to upper layer, so
+		 * ignore lower layer being read-only.
+		 *
+		 * If the overlay itself is read-only then proceed
+		 * with the permission check, don't return EROFS.
+		 * This will only happen if this is the lower layer of
+		 * another overlayfs.
+		 *
+		 * If upper fs becomes read-only after the overlay was
+		 * constructed return EROFS to prevent modification of
+		 * upper layer.
+		 */
+		err = -EROFS;
+		if (is_upper && !IS_RDONLY(inode) && IS_RDONLY(realinode) &&
+		    (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
+			goto out_dput;
+	}
+
+	err = __inode_permission(realinode, mask);
+out_dput:
+	dput(alias);
+	return err;
+}
+
+
+struct ovl_link_data {
+	struct dentry *realdentry;
+	void *cookie;
+};
+
+static void *ovl_follow_link(struct dentry *dentry, struct nameidata *nd)
+{
+	void *ret;
+	struct dentry *realdentry;
+	struct inode *realinode;
+
+	realdentry = ovl_dentry_real(dentry);
+	realinode = realdentry->d_inode;
+
+	if (WARN_ON(!realinode->i_op->follow_link))
+		return ERR_PTR(-EPERM);
+
+	ret = realinode->i_op->follow_link(realdentry, nd);
+	if (IS_ERR(ret))
+		return ret;
+
+	if (realinode->i_op->put_link) {
+		struct ovl_link_data *data;
+
+		data = kmalloc(sizeof(struct ovl_link_data), GFP_KERNEL);
+		if (!data) {
+			realinode->i_op->put_link(realdentry, nd, ret);
+			return ERR_PTR(-ENOMEM);
+		}
+		data->realdentry = realdentry;
+		data->cookie = ret;
+
+		return data;
+	} else {
+		return NULL;
+	}
+}
+
+static void ovl_put_link(struct dentry *dentry, struct nameidata *nd, void *c)
+{
+	struct inode *realinode;
+	struct ovl_link_data *data = c;
+
+	if (!data)
+		return;
+
+	realinode = data->realdentry->d_inode;
+	realinode->i_op->put_link(data->realdentry, nd, data->cookie);
+	kfree(data);
+}
+
+static int ovl_readlink(struct dentry *dentry, char __user *buf, int bufsiz)
+{
+	struct path realpath;
+	struct inode *realinode;
+
+	ovl_path_real(dentry, &realpath);
+	realinode = realpath.dentry->d_inode;
+
+	if (!realinode->i_op->readlink)
+		return -EINVAL;
+
+	touch_atime(&realpath);
+
+	return realinode->i_op->readlink(realpath.dentry, buf, bufsiz);
+}
+
+
+static bool ovl_is_private_xattr(const char *name)
+{
+	return strncmp(name, OVL_XATTR_PRE_NAME, OVL_XATTR_PRE_LEN) == 0;
+}
+
+int ovl_setxattr(struct dentry *dentry, const char *name,
+		 const void *value, size_t size, int flags)
+{
+	int err;
+	struct dentry *upperdentry;
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = -EPERM;
+	if (ovl_is_private_xattr(name))
+		goto out_drop_write;
+
+	err = ovl_copy_up(dentry);
+	if (err)
+		goto out_drop_write;
+
+	upperdentry = ovl_dentry_upper(dentry);
+	err = vfs_setxattr(upperdentry, name, value, size, flags);
+
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static bool ovl_need_xattr_filter(struct dentry *dentry,
+				  enum ovl_path_type type)
+{
+	if ((type & (__OVL_PATH_PURE | __OVL_PATH_UPPER)) == __OVL_PATH_UPPER)
+		return S_ISDIR(dentry->d_inode->i_mode);
+	else
+		return false;
+}
+
+ssize_t ovl_getxattr(struct dentry *dentry, const char *name,
+		     void *value, size_t size)
+{
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+
+	if (ovl_need_xattr_filter(dentry, type) && ovl_is_private_xattr(name))
+		return -ENODATA;
+
+	return vfs_getxattr(realpath.dentry, name, value, size);
+}
+
+ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size)
+{
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+	ssize_t res;
+	int off;
+
+	res = vfs_listxattr(realpath.dentry, list, size);
+	if (res <= 0 || size == 0)
+		return res;
+
+	if (!ovl_need_xattr_filter(dentry, type))
+		return res;
+
+	/* filter out private xattrs */
+	for (off = 0; off < res;) {
+		char *s = list + off;
+		size_t slen = strlen(s) + 1;
+
+		BUG_ON(off + slen > res);
+
+		if (ovl_is_private_xattr(s)) {
+			res -= slen;
+			memmove(s, s + slen, res - off);
+		} else {
+			off += slen;
+		}
+	}
+
+	return res;
+}
+
+int ovl_removexattr(struct dentry *dentry, const char *name)
+{
+	int err;
+	struct path realpath;
+	enum ovl_path_type type = ovl_path_real(dentry, &realpath);
+
+	err = ovl_want_write(dentry);
+	if (err)
+		goto out;
+
+	err = -ENODATA;
+	if (ovl_need_xattr_filter(dentry, type) && ovl_is_private_xattr(name))
+		goto out_drop_write;
+
+	if (!OVL_TYPE_UPPER(type)) {
+		err = vfs_getxattr(realpath.dentry, name, NULL, 0);
+		if (err < 0)
+			goto out_drop_write;
+
+		err = ovl_copy_up(dentry);
+		if (err)
+			goto out_drop_write;
+
+		ovl_path_upper(dentry, &realpath);
+	}
+
+	err = vfs_removexattr(realpath.dentry, name);
+out_drop_write:
+	ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static bool ovl_open_need_copy_up(int flags, enum ovl_path_type type,
+				  struct dentry *realdentry)
+{
+	if (OVL_TYPE_UPPER(type))
+		return false;
+
+	if (special_file(realdentry->d_inode->i_mode))
+		return false;
+
+	if (!(OPEN_FMODE(flags) & FMODE_WRITE) && !(flags & O_TRUNC))
+		return false;
+
+	return true;
+}
+
+static int ovl_dentry_open(struct dentry *dentry, struct file *file,
+		    const struct cred *cred)
+{
+	int err;
+	struct path realpath;
+	enum ovl_path_type type;
+	bool want_write = false;
+
+	type = ovl_path_real(dentry, &realpath);
+	if (!ovl_is_nocopyupw(dentry)) {
+		if (ovl_open_need_copy_up(file->f_flags, type, realpath.dentry)) {
+			want_write = true;
+			err = ovl_want_write(dentry);
+			if (err)
+				goto out;
+
+			if (file->f_flags & O_TRUNC)
+				err = ovl_copy_up_last(dentry, NULL, true);
+			else
+				err = ovl_copy_up(dentry);
+			if (err)
+				goto out_drop_write;
+
+			ovl_path_upper(dentry, &realpath);
+		}
+	}
+
+	err = vfs_open(&realpath, file, cred);
+out_drop_write:
+	if (want_write)
+		ovl_drop_write(dentry);
+out:
+	return err;
+}
+
+static const struct inode_operations 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,
+	.dentry_open	= ovl_dentry_open,
+};
+
+static const struct inode_operations ovl_symlink_inode_operations = {
+	.setattr	= ovl_setattr,
+	.follow_link	= ovl_follow_link,
+	.put_link	= ovl_put_link,
+	.readlink	= ovl_readlink,
+	.getattr	= ovl_getattr,
+	.setxattr	= ovl_setxattr,
+	.getxattr	= ovl_getxattr,
+	.listxattr	= ovl_listxattr,
+	.removexattr	= ovl_removexattr,
+};
+
+struct inode *ovl_new_inode(struct super_block *sb, umode_t mode,
+			    struct ovl_entry *oe)
+{
+	struct inode *inode;
+
+	inode = new_inode(sb);
+	if (!inode)
+		return NULL;
+
+	mode &= S_IFMT;
+
+	inode->i_ino = get_next_ino();
+	inode->i_mode = mode;
+	inode->i_flags |= S_NOATIME | S_NOCMTIME;
+
+	switch (mode) {
+	case S_IFDIR:
+		inode->i_private = oe;
+		inode->i_op = &ovl_dir_inode_operations;
+		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-4.0.9/overlayfs.h

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

executer/kernel/mcoverlayfs/linux-4.0.9/readdir.c

@@ -0,0 +1,557 @@
+/*
+ *
+ * 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;
+	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_merge;
+	struct rb_root root;
+	struct list_head *list;
+	struct list_head middle;
+	struct dentry *dir;
+	int count;
+	int err;
+};
+
+struct ovl_dir_file {
+	bool is_real;
+	bool is_upper;
+	struct ovl_dir_cache *cache;
+	struct list_head *cursor;
+	struct file *realfile;
+	struct file *upperfile;
+};
+
+static struct ovl_cache_entry *ovl_cache_entry_from_node(struct rb_node *n)
+{
+	return container_of(n, struct ovl_cache_entry, node);
+}
+
+static struct ovl_cache_entry *ovl_cache_entry_find(struct rb_root *root,
+						    const char *name, int len)
+{
+	struct rb_node *node = root->rb_node;
+	int cmp;
+
+	while (node) {
+		struct ovl_cache_entry *p = ovl_cache_entry_from_node(node);
+
+		cmp = strncmp(name, p->name, len);
+		if (cmp > 0)
+			node = p->node.rb_right;
+		else if (cmp < 0 || len < p->len)
+			node = p->node.rb_left;
+		else
+			return p;
+	}
+
+	return NULL;
+}
+
+static struct ovl_cache_entry *ovl_cache_entry_new(struct dentry *dir,
+						   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) {
+		struct dentry *dentry;
+		const struct cred *old_cred;
+		struct cred *override_cred;
+
+		override_cred = prepare_creds();
+		if (!override_cred) {
+			kfree(p);
+			return NULL;
+		}
+
+		/*
+		 * CAP_DAC_OVERRIDE for lookup
+		 */
+		cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
+		old_cred = override_creds(override_cred);
+
+		dentry = lookup_one_len(name, dir, len);
+		if (!IS_ERR(dentry)) {
+			p->is_whiteout = ovl_is_whiteout(dentry);
+			dput(dentry);
+		}
+		revert_creds(old_cred);
+		put_cred(override_cred);
+	}
+	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->dir, name, len, ino, d_type);
+	if (p == NULL)
+		return -ENOMEM;
+
+	list_add_tail(&p->l_node, rdd->list);
+	rb_link_node(&p->node, parent, newp);
+	rb_insert_color(&p->node, &rdd->root);
+
+	return 0;
+}
+
+static int ovl_fill_lower(struct ovl_readdir_data *rdd,
+			  const char *name, int namelen,
+			  loff_t offset, u64 ino, unsigned int d_type)
+{
+	struct ovl_cache_entry *p;
+
+	p = ovl_cache_entry_find(&rdd->root, name, namelen);
+	if (p) {
+		list_move_tail(&p->l_node, &rdd->middle);
+	} else {
+		p = ovl_cache_entry_new(rdd->dir, 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_merge)
+		return ovl_cache_entry_add_rb(rdd, name, namelen, ino, d_type);
+	else
+		return ovl_fill_lower(rdd, name, namelen, offset, ino, d_type);
+}
+
+static 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->dir = realpath->dentry;
+	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);
+	fput(realfile);
+
+	return err;
+}
+
+static void ovl_dir_reset(struct file *file)
+{
+	struct ovl_dir_file *od = file->private_data;
+	struct ovl_dir_cache *cache = od->cache;
+	struct dentry *dentry = file->f_path.dentry;
+	enum ovl_path_type type = ovl_path_type(dentry);
+
+	if (cache && ovl_dentry_version_get(dentry) != cache->version) {
+		ovl_cache_put(od, dentry);
+		od->cache = NULL;
+		od->cursor = NULL;
+	}
+	WARN_ON(!od->is_real && !OVL_TYPE_MERGE(type));
+	if (od->is_real && OVL_TYPE_MERGE(type))
+		od->is_real = false;
+}
+
+static int ovl_dir_read_merged(struct dentry *dentry, struct list_head *list)
+{
+	int err;
+	struct path realpath;
+	struct ovl_readdir_data rdd = {
+		.ctx.actor = ovl_fill_merge,
+		.list = list,
+		.root = RB_ROOT,
+		.is_merge = false,
+	};
+	int idx, next;
+
+	for (idx = 0; idx != -1; idx = next) {
+		next = ovl_path_next(idx, dentry, &realpath);
+
+		if (next != -1) {
+			err = ovl_dir_read(&realpath, &rdd);
+			if (err)
+				break;
+		} else {
+			/*
+			 * Insert lowest layer entries before upper ones, this
+			 * allows offsets to be reasonably constant
+			 */
+			list_add(&rdd.middle, rdd.list);
+			rdd.is_merge = true;
+			err = ovl_dir_read(&realpath, &rdd);
+			list_del(&rdd.middle);
+		}
+	}
+	return err;
+}
+
+static void ovl_seek_cursor(struct ovl_dir_file *od, loff_t pos)
+{
+	struct list_head *p;
+	loff_t off = 0;
+
+	list_for_each(p, &od->cache->entries) {
+		if (off >= pos)
+			break;
+		off++;
+	}
+	/* Cursor is safe since the cache is stable */
+	od->cursor = p;
+}
+
+static struct ovl_dir_cache *ovl_cache_get(struct dentry *dentry)
+{
+	int res;
+	struct ovl_dir_cache *cache;
+
+	cache = ovl_dir_cache(dentry);
+	if (cache && ovl_dentry_version_get(dentry) == cache->version) {
+		cache->refcount++;
+		return cache;
+	}
+	ovl_set_dir_cache(dentry, NULL);
+
+	cache = kzalloc(sizeof(struct ovl_dir_cache), GFP_KERNEL);
+	if (!cache)
+		return ERR_PTR(-ENOMEM);
+
+	cache->refcount = 1;
+	INIT_LIST_HEAD(&cache->entries);
+
+	res = ovl_dir_read_merged(dentry, &cache->entries);
+	if (res) {
+		ovl_cache_free(&cache->entries);
+		kfree(cache);
+		return ERR_PTR(res);
+	}
+
+	cache->version = ovl_dentry_version_get(dentry);
+	ovl_set_dir_cache(dentry, cache);
+
+	return cache;
+}
+
+static int ovl_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;
+
+	mutex_lock(&file_inode(file)->i_mutex);
+	if (!file->f_pos)
+		ovl_dir_reset(file);
+
+	if (od->is_real) {
+		res = vfs_llseek(od->realfile, offset, origin);
+		file->f_pos = od->realfile->f_pos;
+	} else {
+		res = -EINVAL;
+
+		switch (origin) {
+		case SEEK_CUR:
+			offset += file->f_pos;
+			break;
+		case SEEK_SET:
+			break;
+		default:
+			goto out_unlock;
+		}
+		if (offset < 0)
+			goto out_unlock;
+
+		if (offset != file->f_pos) {
+			file->f_pos = offset;
+			if (od->cache)
+				ovl_seek_cursor(od, offset);
+		}
+		res = offset;
+	}
+out_unlock:
+	mutex_unlock(&file_inode(file)->i_mutex);
+
+	return res;
+}
+
+static int ovl_dir_fsync(struct file *file, loff_t start, loff_t end,
+			 int datasync)
+{
+	struct ovl_dir_file *od = file->private_data;
+	struct dentry *dentry = file->f_path.dentry;
+	struct file *realfile = od->realfile;
+
+	/*
+	 * Need to check if we started out being a lower dir, but got copied up
+	 */
+	if (!od->is_upper && OVL_TYPE_UPPER(ovl_path_type(dentry))) {
+		struct inode *inode = file_inode(file);
+
+		realfile = lockless_dereference(od->upperfile);
+		if (!realfile) {
+			struct path upperpath;
+
+			ovl_path_upper(dentry, &upperpath);
+			realfile = ovl_path_open(&upperpath, O_RDONLY);
+			smp_mb__before_spinlock();
+			mutex_lock(&inode->i_mutex);
+			if (!od->upperfile) {
+				if (IS_ERR(realfile)) {
+					mutex_unlock(&inode->i_mutex);
+					return PTR_ERR(realfile);
+				}
+				od->upperfile = realfile;
+			} else {
+				/* somebody has beaten us to it */
+				if (!IS_ERR(realfile))
+					fput(realfile);
+				realfile = od->upperfile;
+			}
+			mutex_unlock(&inode->i_mutex);
+		}
+	}
+
+	return vfs_fsync_range(realfile, start, end, datasync);
+}
+
+static int ovl_dir_release(struct inode *inode, struct file *file)
+{
+	struct ovl_dir_file *od = file->private_data;
+
+	if (od->cache) {
+		mutex_lock(&inode->i_mutex);
+		ovl_cache_put(od, file->f_path.dentry);
+		mutex_unlock(&inode->i_mutex);
+	}
+	fput(od->realfile);
+	if (od->upperfile)
+		fput(od->upperfile);
+	kfree(od);
+
+	return 0;
+}
+
+static int ovl_dir_open(struct inode *inode, struct file *file)
+{
+	struct path realpath;
+	struct file *realfile;
+	struct ovl_dir_file *od;
+	enum ovl_path_type type;
+
+	od = kzalloc(sizeof(struct ovl_dir_file), GFP_KERNEL);
+	if (!od)
+		return -ENOMEM;
+
+	type = ovl_path_real(file->f_path.dentry, &realpath);
+	realfile = ovl_path_open(&realpath, file->f_flags);
+	if (IS_ERR(realfile)) {
+		kfree(od);
+		return PTR_ERR(realfile);
+	}
+	od->realfile = realfile;
+	od->is_real = !OVL_TYPE_MERGE(type);
+	od->is_upper = OVL_TYPE_UPPER(type);
+	file->private_data = od;
+
+	return 0;
+}
+
+const struct file_operations ovl_dir_operations = {
+	.read		= generic_read_dir,
+	.open		= ovl_dir_open,
+	.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;
+
+	mutex_lock_nested(&upper->d_inode->i_mutex, I_MUTEX_CHILD);
+	list_for_each_entry(p, list, l_node) {
+		struct dentry *dentry;
+
+		if (!p->is_whiteout)
+			continue;
+
+		dentry = lookup_one_len(p->name, upper, p->len);
+		if (IS_ERR(dentry)) {
+			pr_err("overlayfs: lookup '%s/%.*s' failed (%i)\n",
+			       upper->d_name.name, p->len, p->name,
+			       (int) PTR_ERR(dentry));
+			continue;
+		}
+		ovl_cleanup(upper->d_inode, dentry);
+		dput(dentry);
+	}
+	mutex_unlock(&upper->d_inode->i_mutex);
+}

executer/kernel/procfs.c

@@ -1,488 +0,0 @@
-/**
- * \file procfs.c
- *  License details are found in the file LICENSE.
- * \brief
- *  mcctrl procfs
- * \author Naoki Hamada <nao@axe.bz> \par
- * 	Copyright (C) 2014  AXE, Inc.
- */
-/*
- * HISTORY:
- */
-
-#include <linux/slab.h>
-#include <linux/string.h>
-#include <linux/proc_fs.h>
-#include <linux/list.h>
-#include <linux/uaccess.h>
-#include <linux/fs.h>
-#include <linux/resource.h>
-#include "mcctrl.h"
-#include <linux/version.h>
-
-//#define PROCFS_DEBUG
-
-#ifdef PROCFS_DEBUG
-#define	dprintk(...)	printk(__VA_ARGS__)
-#else
-#define	dprintk(...)
-#endif
-
-static DECLARE_WAIT_QUEUE_HEAD(procfsq);
-static ssize_t mckernel_procfs_read(struct file *file, char __user *buf, 
-		size_t nbytes, loff_t *ppos);
-
-/* A private data for the procfs driver. */
-struct procfs_list_entry;
-
-struct procfs_list_entry {
-	struct list_head list;
-	struct proc_dir_entry *entry;
-	struct procfs_list_entry *parent;
-	ihk_os_t os;
-	int osnum;
-	int pid;
-	int cpu;
-	char fname[PROCFS_NAME_MAX];
-};
-
-/*
- * In the procfs_file_list, mckenrel procfs files are
- * listed in the manner that the leaf file is located 
- * always nearer to the list top than its parent node 
- * file.
- */
-
-LIST_HEAD(procfs_file_list);
-static ihk_spinlock_t procfs_file_list_lock;
-
-loff_t mckernel_procfs_lseek(struct file *file, loff_t offset, int orig)
-{
-	switch (orig) {
-	case 0:
-		file->f_pos = offset;
-		break;
-	case 1:
-		file->f_pos += offset;
-		break;
-	default:
-		return -EINVAL;
-	}
-	return file->f_pos;
-}
-
-static const struct file_operations mckernel_procfs_file_operations = {
-	.llseek		= mckernel_procfs_lseek,
-	.read		= mckernel_procfs_read,
-	.write		= NULL,
-};
-
-
-/**
- * \brief Return specified procfs entry. 
- *
- * \param p a name of the procfs file
- * \param osnum os number
- * \param mode if zero create a directory otherwise a file
- *
- * return value: NULL: Something wrong has occurred.
- *               otherwise: address of the proc_dir_entry structure of the procfs file
- *
- * p should not be NULL nor terminated by "/".
- *
- * We create a procfs entry if there is not already one.
- * This process is recursive to the root of the procfs tree.
- */
-/*
- * XXX: Two or more entries which have same name can be created.
- *
- * get_procfs_list_entry() avoids creating an entry which has already been created.
- * But, it allows creating an entry which is being created by another thread.
- *
- * This problem occurred when two requests which created files with a common
- * ancestor directory which was not explicitly created were racing.
- */
-
-static struct procfs_list_entry *get_procfs_list_entry(char *p, int osnum, int mode)
-{
-	char *r;
-	struct proc_dir_entry *pde = NULL;
-	struct procfs_list_entry *e, *ret = NULL, *parent = NULL;
-	char name[PROCFS_NAME_MAX];
-	unsigned long irqflags;
-
-	dprintk("get_procfs_list_entry: %s for osnum %d mode %o\n", p, osnum, mode);
-	irqflags = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
-	list_for_each_entry(e, &procfs_file_list, list) {
-		if (e == NULL) {
-			kprintf("ERROR: The procfs_file_list has a null entry.\n");
-			return NULL;
-		}
-		if (strncmp(e->fname, p, PROCFS_NAME_MAX) == 0) {
-			/* We found the entry */
-			ret = e;
-			break;
-		}
-	}
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflags);
-	if (ret != NULL) {
-		return ret;
-	}
-	r = strrchr(p, '/');
-	if (r != NULL) {
-		/* We have non-null parent dir. */
-		strncpy(name, p, r - p);
-		name[r - p] = '\0'; 
-		parent = get_procfs_list_entry(name, osnum, 0);
-		if (parent == NULL) {
-			/* We counld not get a parent procfs entry. Give up.*/
-			return NULL;
-		}
-	}
-	ret = kmalloc(sizeof(struct procfs_list_entry), GFP_KERNEL);
-	if (ret == NULL) {
-		kprintf("ERROR: not enough memory to create PROCFS entry.\n");
-		return NULL;
-	}
-	/* Fill the fname field of the entry */
-	strncpy(ret->fname, p, PROCFS_NAME_MAX);
-
-	if (r != NULL) {
-		strncpy(name, r + 1, p + PROCFS_NAME_MAX - r - 1);
-	} else {
-		strncpy(name, p, PROCFS_NAME_MAX);
-	}
-	if (mode == 0) {
-#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
-		pde = proc_mkdir(name, parent ? parent->entry : NULL);
-#else		
-		pde = proc_mkdir_data(name, 0555, parent ? parent->entry : NULL, ret);
-#endif		
-	} else {	
-#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
-		pde = create_proc_entry(name, mode, parent->entry);
-		if (pde) 
-			pde->proc_fops = &mckernel_procfs_file_operations;
-#else		
-		pde = proc_create_data(name, mode, parent->entry, 
-				&mckernel_procfs_file_operations, ret);
-#endif		
-	}
-	if (pde == NULL) {
-		kprintf("ERROR: cannot create a PROCFS entry for %s.\n", p);
-		kfree(ret);
-		return NULL;
-	}
-#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
-	pde->data = ret;
-#endif
-	ret->osnum = osnum;
-	ret->entry = pde;
-	ret->parent = parent;
-
-	irqflags = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
-	list_add(&(ret->list), &procfs_file_list);
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflags);
-
-	dprintk("get_procfs_list_entry: %s done\n", p);
-	return ret;
-}
-
-/**
- * \brief Create a procfs entry.
- *
- * \param __os (opeque) os variable
- * \param ref cpuid of the requesting mckernel process
- * \param osnum osnum of the requesting mckernel process
- * \param pid pid of the requesting mckernel process
- * \param arg sent argument
- */
-
-void procfs_create(void *__os, int ref, int osnum, int pid, unsigned long arg)
-{
-	struct procfs_list_entry *e;
-	ihk_device_t dev = ihk_os_to_dev(__os);
-	unsigned long parg;
-	struct procfs_file *f;
-	int mode;
-	char name[PROCFS_NAME_MAX];
-
-	dprintk("procfs_create: osnum: %d, cpu: %d, pid: %d\n", osnum, ref, pid);
-
-	parg = ihk_device_map_memory(dev, arg, sizeof(struct procfs_file));
-	f = ihk_device_map_virtual(dev, parg, sizeof(struct procfs_file), NULL, 0);
-
-	dprintk("name: %s mode: %o\n", f->fname, f->mode);
-
-	strncpy(name, f->fname, PROCFS_NAME_MAX);
-	mode = f->mode;
-
-	if (name[PROCFS_NAME_MAX - 1] != '\0') {
-			printk("ERROR: procfs_creat: file name not properly terminated.\n");
-			goto quit;
-	}
-	e = get_procfs_list_entry(name, osnum, mode);
-	if (e == NULL) {
-		printk("ERROR: could not create a procfs entry for %s.\n", name);
-		goto quit;
-	}
-
-	e->os = __os;
-	e->cpu = ref;
-	e->pid = pid;
-
-quit:
-	f->status = 1; /* Now the peer can free the data. */
-	ihk_device_unmap_virtual(dev, f, sizeof(struct procfs_file));
-	ihk_device_unmap_memory(dev, parg, sizeof(struct procfs_file));
-	dprintk("procfs_create: done\n");
-}
-
-/**
- * \brief Delete a procfs entry.
- *
- * \param __os (opaque) os variable
- * \param osnum os number
- * \param arg sent argument
- */
-
-void procfs_delete(void *__os, int osnum, unsigned long arg)
-{
-	ihk_device_t dev = ihk_os_to_dev(__os);
-	unsigned long parg;
-	struct procfs_file *f;
-	struct procfs_list_entry *e;
-	struct procfs_list_entry *parent = NULL;
-	char name[PROCFS_NAME_MAX];
-	char *r;
-	unsigned long irqflags;
-
-	dprintk("procfs_delete: \n");
-	parg = ihk_device_map_memory(dev, arg, sizeof(struct procfs_file));
-	f = ihk_device_map_virtual(dev, parg, sizeof(struct procfs_file), NULL, 0);
-	dprintk("fname: %s.\n", f->fname);
-	irqflags = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
-	list_for_each_entry(e, &procfs_file_list, list) {
-		if ((strncmp(e->fname, f->fname, PROCFS_NAME_MAX) == 0) &&
-		    (e->osnum == osnum)) {
-			list_del(&e->list);
-#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
-			e->entry->read_proc = NULL;
-			e->entry->data = NULL;
-#endif
-			parent = e->parent;
-			kfree(e);
-			r = strrchr(f->fname, '/');
-			if (r == NULL) {
-				strncpy(name, f->fname, PROCFS_NAME_MAX);
-			} else {
-				strncpy(name, r + 1, PROCFS_NAME_MAX);
-			}
-			dprintk("found and remove %s from the list.\n", name);
-			remove_proc_entry(name, parent->entry);
-			break;
-		}
-	}
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflags);
-	f->status = 1; /* Now the peer can free the data. */
-	ihk_device_unmap_virtual(dev, f, sizeof(struct procfs_file));
-	ihk_device_unmap_memory(dev, parg, sizeof(struct procfs_file));
-	dprintk("procfs_delete: done\n");
-}
-
-/**
- * \brief Process SCD_MSG_PROCFS_ANSWER message.
- *
- * \param arg sent argument
- * \param err error info (redundant)
- */
-
-void procfs_answer(unsigned int arg, int err)
-{
-	dprintk("procfs: received SCD_MSG_PROCFS_ANSWER message(err = %d).\n", err);
-	wake_up_interruptible(&procfsq);
-}
-
-/**
- * \brief The callback funciton for McKernel procfs
- *
- * This function conforms to the 2) way of fs/proc/generic.c
- * from linux-2.6.39.4.
- */
-static ssize_t
-mckernel_procfs_read(struct file *file, char __user *buf, size_t nbytes,
-	       loff_t *ppos)
-{
-	struct inode * inode = file->f_path.dentry->d_inode;
-	char *kern_buffer;
-	int order = 0;
-	volatile struct procfs_read *r;
-	struct ikc_scd_packet isp;
-	int ret, retrycount = 0;
-	unsigned long pbuf;
-	unsigned long count = nbytes;
-#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
-	struct proc_dir_entry *dp = PDE(inode);
-	struct procfs_list_entry *e = dp->data;
-#else	
-	struct procfs_list_entry *e = PDE_DATA(inode);
-#endif	
-	loff_t offset = *ppos;
-
-	dprintk("mckernel_procfs_read: invoked for %s, offset: %lu, count: %d\n", 
-			e->fname, offset, count); 
-	
-	if (count <= 0 || offset < 0) {
-		return 0;
-	}
-	
-	while ((1 << order) < count) ++order;
-	if (order > 12) {
-		order -= 12;
-	}
-	else {
-		order = 1;
-	}
-
-	/* NOTE: we need physically contigous memory to pass through IKC */
-	kern_buffer = (char *)__get_free_pages(GFP_KERNEL, order);
-	if (!kern_buffer) {
-		printk("mckernel_procfs_read(): ERROR: allocating kernel buffer\n");
-		return -ENOMEM;
-	}
-	
-	pbuf = virt_to_phys(kern_buffer);
-
-	r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
-	if (r == NULL) {
-		return -ENOMEM;
-	}
-retry:
-	dprintk("offset: %lx, count: %d, cpu: %d\n", offset, count, e->cpu);
-
-	r->pbuf = pbuf;
-	r->eof = 0;
-	r->ret = -EIO; /* default */
-	r->status = 0;
-	r->offset = offset;
-	r->count = count;
-	strncpy((char *)r->fname, e->fname, PROCFS_NAME_MAX);
-	isp.msg = SCD_MSG_PROCFS_REQUEST;
-	isp.ref = e->cpu;
-	isp.arg = virt_to_phys(r);
-	
-	ret = mcctrl_ikc_send(e->os, e->cpu, &isp);
-	
-	if (ret < 0) {
-		goto out; /* error */
-	}
-	
-	/* Wait for a reply. */
-	ret = -EIO; /* default exit code */
-	dprintk("now wait for a relpy\n");
-	
-	/* Wait for the status field of the procfs_read structure set ready. */
-	if (wait_event_interruptible_timeout(procfsq, r->status != 0, HZ) == 0) {
-		kprintf("ERROR: mckernel_procfs_read: timeout (1 sec).\n");
-		goto out;
-	}
-	
-	/* Wake up and check the result. */
-	dprintk("mckernel_procfs_read: woke up. ret: %d, eof: %d\n", r->ret, r->eof);
-	if ((r->ret == 0) && (r->eof != 1)) {
-		/* A miss-hit caused by migration has occurred.
-		 * We simply retry the query with a new CPU.
-		 */
-		if (retrycount++ > 10) {
-			kprintf("ERROR: mckernel_procfs_read: excessive retry.\n");
-			goto out;
-		}
-		e->cpu = r->newcpu;
-		dprintk("retry\n");
-		goto retry;
-	}
-	
-	if (r->ret > 0) {
-		if (copy_to_user(buf, kern_buffer, r->ret)) {
-			kprintf("ERROR: mckernel_procfs_read: copy_to_user failed.\n");
-			ret = -EFAULT;
-			goto out;
-		}
-
-		*ppos += r->ret;
-	}
-	ret = r->ret;
-
-out:
-	free_pages((uintptr_t)kern_buffer, order);
-	kfree((void *)r);
-	
-	return ret;
-}
-
-/**
- * \brief Initialization for procfs
- *
- * \param osnum os number
- */
-
-void procfs_init(int osnum) {
-}
-
-/**
- * \brief Finalization for procfs
- *
- * \param osnum os number
- */
-
-void procfs_exit(int osnum) {
-	char buf[20], *r;
-	int error;
-	mm_segment_t old_fs = get_fs();
-	struct kstat stat;
-	struct procfs_list_entry *parent;
-	struct procfs_list_entry *e, *temp = NULL;
-	unsigned long irqflags;
-
-	dprintk("remove remaining mckernel procfs files.\n");
-
-	irqflags = ihk_ikc_spinlock_lock(&procfs_file_list_lock);
-	list_for_each_entry_safe(e, temp, &procfs_file_list, list) {
-		if (e->osnum == osnum) {
-			dprintk("found entry for %s.\n", e->fname);
-			list_del(&e->list);
-#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
-			e->entry->read_proc = NULL;
-			e->entry->data = NULL;
-#endif
-			parent = e->parent;
-			r = strrchr(e->fname, '/');
-			if (r == NULL) {
-				r = e->fname;
-			} else {
-				r += 1;
-			}
-			if (parent) {
-				remove_proc_entry(r, parent->entry);
-			}
-			dprintk("free the entry\n");
-			kfree(e);
-		}
-		dprintk("iterate it.\n");
-	}
-	ihk_ikc_spinlock_unlock(&procfs_file_list_lock, irqflags);
-
-	sprintf(buf, "/proc/mcos%d", osnum);
-
-	set_fs(KERNEL_DS);
-	error = vfs_stat (buf, &stat);
-	set_fs(old_fs);
-	if (error != 0) {
-		return;
-	}
-
-	printk("procfs_exit: We have to remove unexpectedly remaining %s.\n", buf);
-
-	/* remove remnant of previous mcos%d */
-	remove_proc_entry(buf + 6, NULL);
-}

executer/user/Makefile.in

@@ -1,13 +1,19 @@
 CC=@CC@
 BINDIR=@BINDIR@
-CFLAGS=-Wall -O -fPIE -pie
+KDIR ?= @KDIR@
+CFLAGS=-Wall -O -I.
 VPATH=@abs_srcdir@
 TARGET=mcexec
+@uncomment_if_ENABLE_MEMDUMP@TARGET+=eclair
+LIBS=@LIBS@
 
 all: $(TARGET)
 
 mcexec: mcexec.c
-	$(CC) $(CFLAGS) $(EXTRA_CFLAGS) -pthread -o $@ $^ $(EXTRA_OBJS)
+	$(CC) -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -lrt -pthread -o $@ $^ $(EXTRA_OBJS)
+
+eclair: eclair.c
+	$(CC) $(CFLAGS) -o $@ $^ $(LIBS)
 
 clean:
 	$(RM) $(TARGET) *.o
@@ -17,4 +23,5 @@ clean:
 install:
 	mkdir -p -m 755 $(BINDIR)
 	install -m 755 mcexec $(BINDIR)
+	@uncomment_if_ENABLE_MEMDUMP@install -m 755 eclair $(BINDIR)
 

executer/user/mcexec.c

@@ -59,7 +59,12 @@
 #include <semaphore.h>
 #include <signal.h>
 #include <sys/signalfd.h>
+#include <sys/mount.h>
+#include <include/generated/uapi/linux/version.h>
+#include <sys/user.h>
 #include "../include/uprotocol.h"
+#include <getopt.h>
+#include "../config.h"
 
 //#define DEBUG
 
@@ -96,6 +101,19 @@ int __glob_argc = -1;
 char **__glob_argv = 0;
 #endif
 
+#ifdef ENABLE_MCOVERLAYFS
+#undef ENABLE_MCOVERLAYFS
+#ifndef RHEL_RELEASE_CODE
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,1,0)
+#define ENABLE_MCOVERLAYFS 1
+#endif // LINUX_VERSION_CODE == 4.0
+#else
+#if RHEL_RELEASE_CODE == RHEL_RELEASE_VERSION(7,2)
+#define ENABLE_MCOVERLAYFS 1
+#endif // RHEL_RELEASE_CODE == 7.2
+#endif // RHEL_RELEASE_CODE
+#endif // ENABLE_MCOVERLAYFS
+
 typedef unsigned char   cc_t;
 typedef unsigned int    speed_t;
 typedef unsigned int    tcflag_t;
@@ -129,6 +147,7 @@ static char *exec_path = NULL;
 static char *altroot;
 static const char rlimit_stack_envname[] = "MCKERNEL_RLIMIT_STACK";
 static int ischild;
+static int enable_vdso = 1;
 
 struct fork_sync {
 	pid_t pid;
@@ -183,6 +202,9 @@ struct program_load_desc *load_elf(FILE *fp, char **interp_pathp)
 	
 	desc = malloc(sizeof(struct program_load_desc)
 	              + sizeof(struct program_image_section) * nhdrs);
+	memset(desc, '\0', sizeof(struct program_load_desc)
+	                   + sizeof(struct program_image_section) * nhdrs);
+	desc->shell_path[0] = '\0';
 	fseek(fp, hdr.e_phoff, SEEK_SET);
 	j = 0;
 	desc->num_sections = nhdrs;
@@ -242,6 +264,8 @@ struct program_load_desc *load_elf(FILE *fp, char **interp_pathp)
 	}
 	desc->pid = getpid();
 	desc->pgid = getpgid(0);
+	if(*interp_pathp)
+		desc->reloc = hdr.e_type == ET_DYN;
 	desc->entry = hdr.e_entry;
 	ioctl(fd, MCEXEC_UP_GET_CREDV, desc->cred);
 	desc->at_phdr = load_addr + hdr.e_phoff;
@@ -364,7 +388,7 @@ struct program_load_desc *load_interp(struct program_load_desc *desc0, FILE *fp)
 
 unsigned char *dma_buf;
 
-int lookup_exec_path(char *filename, char *path, int max_len) 
+int lookup_exec_path(char *filename, char *path, int max_len, int execvp) 
 {
 	int found;
 	int error;
@@ -382,28 +406,27 @@ retry:
 
 			char *token, *string, *tofree;
 			char *PATH = getenv("COKERNEL_PATH");
-			if (!PATH) {
+
+			if (!execvp) {
+				if (strlen(filename) + 1 > max_len) {
+					return ENAMETOOLONG;
+				}
+				strcpy(path, filename);
+				error = access(path, X_OK);
+				if (error) {
+					return errno;
+				}
+				found = 1;
+				break;
+			}
+
+			if (!(PATH = getenv("COKERNEL_PATH"))) {
 				PATH = getenv("PATH");
 			}
 
 			if (strlen(filename) >= 255) {
 				return ENAMETOOLONG;
 			}
-			
-			/* See first whether file is available in current working dir */
-			error = access(filename, X_OK);
-			if (error == 0) {
-				__dprintf("lookup_exec_path(): found %s in cwd\n", filename);
-				error = snprintf(path, max_len, "%s", filename);
-
-				if (error < 0 || error >= max_len) {
-					fprintf(stderr, "lookup_exec_path(): array too small?\n");
-					return ENOMEM;
-				}
-
-				found = 1;
-				break;
-			}
 
 			__dprintf("PATH: %s\n", PATH);
 
@@ -431,6 +454,9 @@ retry:
 			}
 
 			free(tofree);
+			if(!found){
+				return ENOENT;
+			}
 			break;
 		}
 
@@ -477,7 +503,7 @@ retry:
 	}
 
 	if ((sb.st_mode & S_IFMT) == S_IFLNK) {
-		char *link_path = malloc(max_len);
+		link_path = malloc(max_len);
 		if (!link_path) {
 			fprintf(stderr, "lookup_exec_path(): error allocating\n");
 			return ENOMEM;
@@ -488,9 +514,18 @@ retry:
 			fprintf(stderr, "lookup_exec_path(): error readlink\n");
 			return EINVAL;
 		}
+		link_path[error] = '\0';
 
 		__dprintf("lookup_exec_path(): %s is link -> %s\n", path, link_path);
 
+		if(link_path[0] != '/'){
+			char *t = strrchr(path, '/');
+			if(t){
+				t++;
+				strcpy(t, link_path);
+				strcpy(link_path, path);
+			}
+		}
 		filename = link_path;
 		goto retry; 
 	}
@@ -634,10 +669,7 @@ int load_elf_desc(char *filename, struct program_load_desc **desc_p,
 	return 0;
 }
 
-#define PAGE_SIZE 4096
-#define PAGE_MASK ~((unsigned long)PAGE_SIZE - 1)
-
-void transfer_image(int fd, struct program_load_desc *desc)
+int transfer_image(int fd, struct program_load_desc *desc)
 {
 	struct remote_transfer pt;
 	unsigned long s, e, flen, rpa;
@@ -651,13 +683,17 @@ void transfer_image(int fd, struct program_load_desc *desc)
 		     + PAGE_SIZE - 1) & PAGE_MASK;
 		rpa = desc->sections[i].remote_pa;
 
-		fseek(fp, desc->sections[i].offset, SEEK_SET);
+		if (fseek(fp, desc->sections[i].offset, SEEK_SET) != 0) {
+			fprintf(stderr, "transfer_image(): error: seeking file position\n");
+			return -1;
+		}
 		flen = desc->sections[i].filesz;
 
 		__dprintf("seeked to %lx | size %ld\n",
 		          desc->sections[i].offset, flen);
 
 		while (s < e) {
+			memset(&pt, '\0', sizeof pt);
 			pt.rphys = rpa;
 			pt.userp = dma_buf;
 			pt.size = PAGE_SIZE;
@@ -672,7 +708,20 @@ void transfer_image(int fd, struct program_load_desc *desc)
 				if (lr > flen) {
 					lr = flen;
 				}
-				fread(dma_buf + l, 1, lr, fp); 
+				if (fread(dma_buf + l, 1, lr, fp) != lr) {
+					if (ferror(fp) > 0) {
+						fprintf(stderr, "transfer_image(): error: accessing file\n");
+						return -EINVAL;
+					}
+					else if (feof(fp) > 0) {
+						fprintf(stderr, "transfer_image(): file too short?\n");
+						return -EINVAL;
+					}
+					else {
+						/* TODO: handle smaller reads.. */
+						return -EINVAL;
+					}
+				}
 				flen -= lr;
 			} 
 			else if (flen > 0) {
@@ -681,7 +730,20 @@ void transfer_image(int fd, struct program_load_desc *desc)
 				} else {
 					lr = flen;
 				}
-				fread(dma_buf, 1, lr, fp);
+				if (fread(dma_buf, 1, lr, fp) != lr) {
+					if (ferror(fp) > 0) {
+						fprintf(stderr, "transfer_image(): error: accessing file\n");
+						return -EINVAL;
+					}
+					else if (feof(fp) > 0) {
+						fprintf(stderr, "transfer_image(): file too short?\n");
+						return -EINVAL;
+					}
+					else {
+						/* TODO: handle smaller reads.. */
+						return -EINVAL;
+					}
+				}
 				flen -= lr;
 			} 
 			s += PAGE_SIZE;
@@ -697,6 +759,8 @@ void transfer_image(int fd, struct program_load_desc *desc)
 			}
 		}
 	}
+
+	return 0;
 }
 
 void print_desc(struct program_load_desc *desc)
@@ -761,7 +825,7 @@ int flatten_strings(int nr_strings, char *first, char **strings, char **flat)
 	}
 
 	/* Count full length */
-	full_len = sizeof(int) + sizeof(char *); // Counter and terminating NULL
+	full_len = sizeof(long) + sizeof(char *); // Counter and terminating NULL
 	if (first) {
 		full_len += sizeof(char *) + strlen(first) + 1; 
 	}
@@ -771,6 +835,8 @@ int flatten_strings(int nr_strings, char *first, char **strings, char **flat)
 		full_len += sizeof(char *) + strlen(strings[string_i]) + 1; 
 	}
 
+	full_len = (full_len + sizeof(long) - 1) & ~(sizeof(long) - 1);
+
 	_flat = (char *)malloc(full_len);
 	if (!_flat) {
 		return 0;
@@ -779,14 +845,14 @@ int flatten_strings(int nr_strings, char *first, char **strings, char **flat)
 	memset(_flat, 0, full_len);
 
 	/* Number of strings */
-	*((int*)_flat) = nr_strings + (first ? 1 : 0);
+	*((long *)_flat) = nr_strings + (first ? 1 : 0);
 	
 	// Actual offset
-	flat_offset = sizeof(int) + sizeof(char *) * (nr_strings + 1 + 
+	flat_offset = sizeof(long) + sizeof(char *) * (nr_strings + 1 + 
 			(first ? 1 : 0)); 
 
 	if (first) {
-		*((char **)(_flat + sizeof(int))) = (void *)flat_offset;
+		*((char **)(_flat + sizeof(long))) = (void *)flat_offset;
 		memcpy(_flat + flat_offset, first, strlen(first) + 1);
 		flat_offset += strlen(first) + 1;
 	}
@@ -794,7 +860,7 @@ int flatten_strings(int nr_strings, char *first, char **strings, char **flat)
 	for (string_i = 0; string_i < nr_strings; ++string_i) {
 		
 		/* Fabricate the string */
-		*((char **)(_flat + sizeof(int) + (string_i + (first ? 1 : 0)) 
+		*((char **)(_flat + sizeof(long) + (string_i + (first ? 1 : 0)) 
 					* sizeof(char *))) = (void *)flat_offset;
 		memcpy(_flat + flat_offset, strings[string_i], strlen(strings[string_i]) + 1);
 		flat_offset += strlen(strings[string_i]) + 1;
@@ -817,7 +883,10 @@ struct thread_data_s {
 	pthread_mutex_t *lock;
 	pthread_barrier_t *init_ready;
 } *thread_data;
+
 int ncpu;
+int n_threads;
+
 pid_t master_tid;
 
 pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
@@ -828,7 +897,7 @@ static void *main_loop_thread_func(void *arg)
 	struct thread_data_s *td = (struct thread_data_s *)arg;
 
 	td->tid = gettid();
-	td->remote_tid = (int)td->tid;
+	td->remote_tid = -1;
 	pthread_barrier_wait(&init_ready);
 	td->ret = main_loop(td->fd, td->cpu, td->lock);
 
@@ -877,6 +946,7 @@ sendsig(int sig, siginfo_t *siginfo, void *context)
 		remote_tid = -1;
 	}
 
+	memset(&sigdesc, '\0', sizeof sigdesc);
 	sigdesc.cpu = cpu;
 	sigdesc.pid = (int)pid;
 	sigdesc.tid = remote_tid;
@@ -903,13 +973,17 @@ act_signalfd4(struct syscall_wait_desc *w)
 	switch(mode){
 	    case 0: /* new signalfd */
 		sfd = malloc(sizeof(struct sigfd));
+		memset(sfd, '\0', sizeof(struct sigfd));
 		tmp = w->sr.args[1];
 		flags = 0;
 		if(tmp & SFD_NONBLOCK)
 			flags |= O_NONBLOCK;
 		if(tmp & SFD_CLOEXEC)
 			flags |= O_CLOEXEC;
-		pipe2(sfd->sigpipe, flags);
+		if (pipe2(sfd->sigpipe, flags) < 0) {
+			perror("pipe2 failed:");
+			return -1;
+		}
 		sfd->next = sigfdtop;
 		sigfdtop = sfd;
 		rc = sfd->sigpipe[0];
@@ -940,7 +1014,11 @@ act_signalfd4(struct syscall_wait_desc *w)
 			rc = -EBADF;
 		else{
 			info = (struct signalfd_siginfo *)w->sr.args[2];
-			write(sfd->sigpipe[1], info, sizeof(struct signalfd_siginfo));
+			if (write(sfd->sigpipe[1], info, sizeof(struct signalfd_siginfo))
+					!= sizeof(struct signalfd_siginfo)) {
+				fprintf(stderr, "error: writing sigpipe\n");
+				rc = -EBADF;
+			}
 		}
 		break;
 	}
@@ -1046,9 +1124,9 @@ void init_worker_threads(int fd)
 	int i;
 
 	pthread_mutex_init(&lock, NULL);
-	pthread_barrier_init(&init_ready, NULL, ncpu + 2);
+	pthread_barrier_init(&init_ready, NULL, n_threads + 2);
 
-	for (i = 0; i <= ncpu; ++i) {
+	for (i = 0; i <= n_threads; ++i) {
 		int ret;
 
 		thread_data[i].fd = fd;
@@ -1068,6 +1146,80 @@ void init_worker_threads(int fd)
 	pthread_barrier_wait(&init_ready);
 }
 
+#ifdef ENABLE_MCOVERLAYFS
+#define READ_BUFSIZE 1024
+static int isunshare(void)
+{
+	int err = 0;
+	int ret;
+	int fd;
+	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;
+
+	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);
+		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;
+		}
+
+		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) {
+			break;
+		}
+	}
+
+	close(fd);
+
+	__dprintf("err=%d\n", err);
+	return err;
+}
+#endif // ENABLE_MCOVERLAYFS
+
 #define MCK_RLIMIT_AS	0
 #define MCK_RLIMIT_CORE	1
 #define MCK_RLIMIT_CPU	2
@@ -1138,6 +1290,24 @@ static int rlimits[] = {
 
 char dev[64];
 
+static struct option mcexec_options[] = {
+	{
+		.name =		"disable-vdso",
+		.has_arg =	no_argument,
+		.flag =		&enable_vdso,
+		.val =		0,
+	},
+	{
+		.name =		"enable-vdso",
+		.has_arg =	no_argument,
+		.flag =		&enable_vdso,
+		.val =		1,
+	},
+
+	/* end */
+	{ NULL, 0, NULL, 0, },
+};
+
 int main(int argc, char **argv)
 {
 //	int fd;
@@ -1189,12 +1359,15 @@ int main(int argc, char **argv)
 	}
            
 	/* Parse options ("+" denotes stop at the first non-option) */
-	while ((opt = getopt(argc, argv, "+c:")) != -1) {
+	while ((opt = getopt_long(argc, argv, "+c:", mcexec_options, NULL)) != -1) {
 		switch (opt) {
 			case 'c':
 				target_core = atoi(optarg);
 				break;
 			
+			case 0:	/* long opt */
+				break;
+
 			default: /* '?' */
 				print_usage(argv);
 				exit(EXIT_FAILURE);
@@ -1233,7 +1406,59 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
-	if (lookup_exec_path(argv[optind], path, sizeof(path)) != 0) {
+#ifdef ENABLE_MCOVERLAYFS
+	__dprintf("mcoverlay enable\n");
+	char mcos_procdir[PATH_MAX];
+	char mcos_sysdir[PATH_MAX];
+
+	error = isunshare();
+	if (error == 0) {
+		struct sys_unshare_desc unshare_desc;
+		struct sys_mount_desc mount_desc;
+
+		memset(&unshare_desc, '\0', sizeof unshare_desc);
+		memset(&mount_desc, '\0', sizeof mount_desc);
+		unshare_desc.unshare_flags = CLONE_NEWNS;
+		if (ioctl(fd, MCEXEC_UP_SYS_UNSHARE, 
+			(unsigned long)&unshare_desc) != 0) {
+			fprintf(stderr, "Error: Failed to unshare. (%s)\n", 
+				strerror(errno));
+			return 1;
+		}
+
+		sprintf(mcos_procdir, "/tmp/mcos/mcos%d_proc", mcosid);
+		mount_desc.dev_name = mcos_procdir;
+		mount_desc.dir_name = "/proc";
+		mount_desc.type = NULL;
+		mount_desc.flags = MS_BIND;
+		mount_desc.data = NULL;
+		if (ioctl(fd, MCEXEC_UP_SYS_MOUNT, 
+			(unsigned long)&mount_desc) != 0) {
+			fprintf(stderr, "Error: Failed to mount /proc. (%s)\n", 
+				strerror(errno));
+			return 1;
+		}
+
+		sprintf(mcos_sysdir, "/tmp/mcos/mcos%d_sys", mcosid);
+		mount_desc.dev_name = mcos_sysdir;
+		mount_desc.dir_name = "/sys";
+		mount_desc.type = NULL;
+		mount_desc.flags = MS_BIND;
+		mount_desc.data = NULL;
+		if (ioctl(fd, MCEXEC_UP_SYS_MOUNT, 
+			(unsigned long)&mount_desc) != 0) {
+			fprintf(stderr, "Error: Failed to mount /sys. (%s)\n", 
+				strerror(errno));
+			return 1;
+		}
+	} else if (error == -1) {
+		return 1;
+	}
+#else
+	__dprintf("mcoverlay disable\n");
+#endif // ENABLE_MCOVERLAYFS
+
+	if (lookup_exec_path(argv[optind], path, sizeof(path), 1) != 0) {
 		fprintf(stderr, "error: finding file: %s\n", argv[optind]);
 		return 1;
 	}
@@ -1245,7 +1470,7 @@ int main(int argc, char **argv)
 
 	/* Check whether shell script */
 	if (shell) {
-		if (lookup_exec_path(shell, shell_path, sizeof(shell_path)) != 0) {
+		if (lookup_exec_path(shell, shell_path, sizeof(shell_path), 0) != 0) {
 			fprintf(stderr, "error: finding file: %s\n", shell);
 			return 1;
 		}
@@ -1271,6 +1496,8 @@ int main(int argc, char **argv)
 	//print_flat(args);
 
 	desc->cpu = target_core;
+	desc->enable_vdso = enable_vdso;
+
 	p = getenv(rlimit_stack_envname);
 	if (p) {
 		errno = 0;
@@ -1305,6 +1532,19 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
+	n_threads = ncpu;
+	if (ncpu > 16) {
+		n_threads = 16;
+	}
+
+	/* 
+	 * XXX: keep thread_data ncpu sized despite that there are only
+	 * n_threads worker threads in the pool so that signaling code
+	 * keeps working.
+	 *
+	 * TODO: fix signaling code to be independent of TIDs.
+	 * TODO: implement dynaic thread pool resizing.
+	 */
 	thread_data = (struct thread_data_s *)malloc(sizeof(struct thread_data_s) * (ncpu + 1));
 	memset(thread_data, '\0', sizeof(struct thread_data_s) * (ncpu + 1));
 
@@ -1347,7 +1587,10 @@ int main(int argc, char **argv)
 	}
 
 	print_desc(desc);
-	transfer_image(fd, desc);
+	if (transfer_image(fd, desc) < 0) {
+		fprintf(stderr, "error: transferring image\n");
+		return -1;
+	}
 	fflush(stdout);
 	fflush(stderr);
 	
@@ -1386,7 +1629,7 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
-	for (i = 0; i <= ncpu; ++i) {
+	for (i = 0; i <= n_threads; ++i) {
 		pthread_join(thread_data[i].thread_id, NULL);
 	}
 
@@ -1400,6 +1643,7 @@ void do_syscall_return(int fd, int cpu,
 {
 	struct syscall_ret_desc desc;
 
+	memset(&desc, '\0', sizeof desc);
 	desc.cpu = cpu;
 	desc.ret = ret;
 	desc.src = src;
@@ -1416,6 +1660,7 @@ void do_syscall_load(int fd, int cpu, unsigned long dest, unsigned long src,
 {
 	struct syscall_load_desc desc;
 
+	memset(&desc, '\0', sizeof desc);
 	desc.cpu = cpu;
 	desc.src = src;
 	desc.dest = dest;
@@ -1446,16 +1691,14 @@ do_generic_syscall(
 }
 
 static void
-kill_thread(unsigned long cpu)
+kill_thread(unsigned long tid)
 {
-	if(cpu >= 0 && cpu < ncpu){
-		pthread_kill(thread_data[cpu].thread_id, LOCALSIG);
-	}
-	else{
-		int	i;
+	int	i;
 
-		for (i = 0; i < ncpu; ++i) {
+	for (i = 0; i <= n_threads; ++i) {
+		if(thread_data[i].remote_tid == tid){
 			pthread_kill(thread_data[i].thread_id, LOCALSIG);
+			break;
 		}
 	}
 }
@@ -1465,6 +1708,7 @@ static long do_strncpy_from_user(int fd, void *dest, void *src, unsigned long n)
 	struct strncpy_from_user_desc desc;
 	int ret;
 
+	memset(&desc, '\0', sizeof desc);
 	desc.dest = dest;
 	desc.src = src;
 	desc.n = n;
@@ -1559,6 +1803,9 @@ int close_cloexec_fds(int mcos_fd)
 char *
 chgpath(char *in, char *buf)
 {
+#ifdef ENABLE_MCOVERLAYFS
+	return in;
+#endif // ENABLE_MCOVERLAYFS
 	char	*fn = in;
 	struct stat	sb;
 
@@ -1588,10 +1835,11 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 	char *fn;
 	int sig;
 	int term;
-	struct timeval tv;
+	struct timespec tv;
 	char pathbuf[PATH_MAX];
 	char tmpbuf[PATH_MAX];
 
+	memset(&w, '\0', sizeof w);
 	w.cpu = cpu;
 	w.pid = getpid();
 
@@ -1607,6 +1855,8 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 		
 		//pthread_mutex_lock(lock);
 
+		thread_data[cpu].remote_tid = w.sr.rtid;
+
 		switch (w.sr.number) {
 		case __NR_open:
 			ret = do_strncpy_from_user(fd, pathbuf, (void *)w.sr.args[0], PATH_MAX);
@@ -1627,13 +1877,13 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 			break;
 
 		case __NR_futex:
-			ret = gettimeofday(&tv, NULL);
+			ret = clock_gettime(w.sr.args[1], &tv);
 			SET_ERR(ret);
-			__dprintf("gettimeofday=%016ld,%09ld\n",
+			__dprintf("clock_gettime=%016ld,%09ld\n",
 					tv.tv_sec,
-					tv.tv_usec);
+					tv.tv_nsec);
 			do_syscall_return(fd, cpu, ret, 1, (unsigned long)&tv,
-			                  w.sr.args[0], sizeof(struct timeval));
+			                  w.sr.args[0], sizeof(struct timespec));
 			break;
 
 		case __NR_kill: // interrupt syscall
@@ -1645,13 +1895,13 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 			sig = 0;
 			term = 0;
 			
+			do_syscall_return(fd, cpu, 0, 0, 0, 0, 0);
+
 			/* Drop executable file */
 			if ((ret = ioctl(fd, MCEXEC_UP_CLOSE_EXEC)) != 0) {
 				fprintf(stderr, "WARNING: close_exec() couldn't find exec file?\n");
 			}
 
-			do_syscall_return(fd, cpu, 0, 0, 0, 0, 0);
-
 			__dprintf("__NR_exit/__NR_exit_group: %ld (cpu_id: %d)\n",
 					w.sr.args[0], cpu);
 			if(w.sr.number == __NR_exit_group){
@@ -1689,17 +1939,12 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 			return w.sr.args[0];
 
 		case __NR_mmap:
+		case __NR_munmap:
 		case __NR_mprotect:
 			/* reserved for internal use */
 			do_syscall_return(fd, cpu, -ENOSYS, 0, 0, 0, 0);
 			break;
 
-		case __NR_munmap:
-			ret = madvise((void *)w.sr.args[0], w.sr.args[1], MADV_DONTNEED);
-			SET_ERR(ret);
-			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
-			break;
-
 #ifdef USE_SYSCALL_MOD_CALL
 		case 303:{
 			__dprintf("mcexec.c,mod_cal,mod=%ld,cmd=%ld\n", w.sr.args[0], w.sr.args[1]);
@@ -1709,22 +1954,40 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 #endif
 
 		case __NR_gettid:{
-			int mode = w.sr.args[0];
-			int remote_pid = w.sr.args[1];
-			int newcpuid = w.sr.args[2];
-			int oldcpuid = w.sr.args[3];
-			int wtid = thread_data[newcpuid].remote_tid;
+			/*
+			 * Number of TIDs and the remote physical address where TIDs are
+			 * expected are passed in arg 4 and 5, respectively.
+			 */
+			if (w.sr.args[4] > 0) {
+				struct remote_transfer trans;
+				int i = 0;
+				int *tids = malloc(sizeof(int) * w.sr.args[4]);
+				if (!tids) {
+					fprintf(stderr, "__NR_gettid(): error allocating TIDs\n");
+					goto gettid_out;
+				}
 
-			if(mode == 0){
-				thread_data[ncpu].remote_tid = wtid;
-				thread_data[newcpuid].remote_tid = remote_pid;
-			}
-			else if(mode == 2){
-				thread_data[newcpuid].remote_tid = thread_data[oldcpuid].remote_tid;
-				thread_data[oldcpuid].remote_tid = wtid;
-			}
+				for (i = 0; i < ncpu && i < w.sr.args[4]; ++i) {
+					tids[i] = thread_data[i].tid;
+				}
 
-			do_syscall_return(fd, cpu, thread_data[newcpuid].remote_tid, 0, 0, 0, 0);
+				for (; i < ncpu; ++i) {
+					tids[i] = 0;
+				}
+
+				trans.userp = (void*)tids;
+				trans.rphys = w.sr.args[5];
+				trans.size = sizeof(int) * w.sr.args[4];
+				trans.direction = MCEXEC_UP_TRANSFER_TO_REMOTE;
+
+				if (ioctl(fd, MCEXEC_UP_TRANSFER, &trans) != 0) {
+					fprintf(stderr, "__NR_gettid(): error transfering TIDs\n");
+				}
+
+				free(tids);
+			}
+gettid_out:
+			do_syscall_return(fd, cpu, 0, 0, 0, 0, 0);
 			break;
 		}
 
@@ -1733,6 +1996,7 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 			struct fork_sync_container *fsc;
 			struct fork_sync_container *fp;
 			struct fork_sync_container *fb;
+			int flag = w.sr.args[0];
 			int rc = -1;
 			pid_t pid;
 
@@ -1752,7 +2016,45 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 			memset(fs, '\0', sizeof(struct fork_sync));
 			sem_init(&fs->sem, 1, 0);
 
-			pid = fork();
+			if(flag){
+				int pipefds[2];
+
+				if(pipe(pipefds) == -1){
+					rc = -errno;
+					sem_destroy(&fs->sem);
+					goto fork_err;
+				}
+				pid = fork();
+				if(pid == 0){
+					close(pipefds[0]);
+					pid = fork();
+					if(pid != 0){
+						if (write(pipefds[1], &pid, sizeof pid) != sizeof(pid)) {
+							fprintf(stderr, "error: writing pipefds\n");
+						}
+						exit(0);
+					}
+				}
+				else if(pid != -1){
+					int npid;
+					int st;
+
+					close(pipefds[1]);
+					if (read(pipefds[0], &npid, sizeof npid) != sizeof(npid)) {
+						fprintf(stderr, "error: reading pipefds\n");
+					}
+					close(pipefds[0]);
+					waitpid(pid, &st, 0);
+					pid = npid;
+				}
+				else{
+					rc = -errno;
+					sem_destroy(&fs->sem);
+					goto fork_err;
+				}
+			}
+			else
+				pid = fork();
 
 			switch (pid) {
 			    /* Error */
@@ -1780,7 +2082,6 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 
 				/* Reinit signals and syscall threads */
 				init_sigaction();
-				init_worker_threads(fd);
 
 				__dprintf("pid(%d): signals and syscall threads OK\n", 
 						getpid());
@@ -1794,6 +2095,8 @@ int main_loop(int fd, int cpu, pthread_mutex_t *lock)
 					goto fork_child_sync_pipe;
 				}
 
+				init_worker_threads(fd);
+
 fork_child_sync_pipe:
 				sem_post(&fs->sem);
 				if (fs->status)
@@ -1879,7 +2182,7 @@ fork_err:
 			}
 
 			if(ret != pid) {
-				fprintf(stderr, "ERROR: waiting for %lu\n", w.sr.args[0]);
+				fprintf(stderr, "ERROR: waiting for %lu rc=%d errno=%d\n", w.sr.args[0], ret, errno);
 			}
 
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
@@ -1895,15 +2198,16 @@ fork_err:
 				char path[1024];
 				char *filename;
 				int ret;
-				char *shell = NULL;
+				char *shell;
 				char shell_path[1024];
 
 				/* Load descriptor phase */
 				case 1:
 					
+					shell = NULL;
 					filename = (char *)w.sr.args[1];
 					
-					if ((ret = lookup_exec_path(filename, path, sizeof(path))) 
+					if ((ret = lookup_exec_path(filename, path, sizeof(path), 0)) 
 						!= 0) {
 						goto return_execve1;
 					}
@@ -1917,7 +2221,7 @@ fork_err:
 					/* Check whether shell script */
 					if (shell) {
 						if ((ret = lookup_exec_path(shell, shell_path, 
-									sizeof(shell_path))) != 0) {
+									sizeof(shell_path), 0)) != 0) {
 							fprintf(stderr, "execve(): error: finding file: %s\n", shell);
 							goto return_execve1;
 						}
@@ -1938,6 +2242,7 @@ fork_err:
 						strcpy(desc->shell_path, shell_path);
 					}
 
+					desc->enable_vdso = enable_vdso;
 					__dprintf("execve(): load_elf_desc() for %s OK, num sections: %d\n",
 						path, desc->num_sections);
 
@@ -1978,6 +2283,7 @@ return_execve1:
 						fprintf(stderr, "execve(): error allocating desc\n");
 						goto return_execve2;
 					}
+					memset(desc, '\0', w.sr.args[2]);
 
 					/* Copy descriptor from co-kernel side */
 					trans.userp = (void*)desc;
@@ -1994,7 +2300,10 @@ return_execve1:
 					
 					__dprintf("%s", "execve(): transfer ELF desc OK\n");
 
-					transfer_image(fd, desc);	
+					if (transfer_image(fd, desc) != 0) {
+						fprintf(stderr, "error: transferring image\n");
+						return -1;
+					}
 					__dprintf("%s", "execve(): image transferred\n");
 
 					if (close_cloexec_fds(fd) < 0) {
@@ -2020,6 +2329,11 @@ return_execve2:
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
 
+		case __NR_perf_event_open:
+			ret = open("/dev/null", O_RDONLY);
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
 		case __NR_rt_sigaction:
 			act_sigaction(&w);
 			do_syscall_return(fd, cpu, 0, 0, 0, 0, 0);
@@ -2041,6 +2355,53 @@ return_execve2:
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
 
+		case __NR_setresuid:
+			ret = setresuid(w.sr.args[0], w.sr.args[1], w.sr.args[2]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setreuid:
+			ret = setreuid(w.sr.args[0], w.sr.args[1]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setuid:
+			ret = setuid(w.sr.args[0]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setresgid:
+			ret = setresgid(w.sr.args[0], w.sr.args[1], w.sr.args[2]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setregid:
+			ret = setregid(w.sr.args[0], w.sr.args[1]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setgid:
+			ret = setgid(w.sr.args[0]);
+			if(ret == -1)
+				ret = -errno;
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
+		case __NR_setfsgid:
+			ret = setfsgid(w.sr.args[0]);
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
 		case __NR_close:
 			if(w.sr.args[0] == fd)
 				ret = -EBADF;
@@ -2049,13 +2410,34 @@ return_execve2:
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
 
+		case __NR_readlink:
+			ret = do_strncpy_from_user(fd, pathbuf, (void *)w.sr.args[0], PATH_MAX);
+			if (ret >= PATH_MAX) {
+				ret = -ENAMETOOLONG;
+			}
+			if (ret < 0) {
+				do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+				break;
+			}
+
+			fn = chgpath(pathbuf, tmpbuf);
+
+			ret = readlink(fn, (char *)w.sr.args[1], w.sr.args[2]);
+			__dprintf("readlink: path=%s, buf=%s, ret=%ld\n", 
+				fn, (char *)w.sr.args[1], ret);
+			SET_ERR(ret);
+			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+			break;
+
 		default:
 			ret = do_generic_syscall(&w);
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
 
 		}
-		
+
+		thread_data[cpu].remote_tid = -1;
+
 		//pthread_mutex_unlock(lock);
 	}
 	__dprint("timed out.\n");

kernel/Makefile.build.dcfa

@@ -3,7 +3,7 @@ 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
 DEPSRCS=$(wildcard $(SRC)/*.c)
 
-CFLAGS += -I$(SRC)/include -mcmodel=kernel -D__KERNEL__
+CFLAGS += -I$(SRC)/include -D__KERNEL__
 CFLAGS += -DKNC_MAP_MICPA $(EXTRA_CFLAGS)
 
 ifeq ("$(DCFA_MODE)", "kmod")

kernel/Makefile.build.in

@@ -3,10 +3,10 @@ SRC=$(VPATH)
 IHKDIR=$(IHKBASE)/$(TARGETDIR)
 OBJS = init.o mem.o debug.o mikc.o listeners.o ap.o syscall.o cls.o host.o
 OBJS += process.o copy.o waitq.o futex.o timer.o plist.o fileobj.o shmobj.o
-OBJS += zeroobj.o procfs.o devobj.o
+OBJS += zeroobj.o procfs.o devobj.o sysfs.o
 DEPSRCS=$(wildcard $(SRC)/*.c)
 
-CFLAGS += -I$(SRC)/include -mcmodel=kernel -D__KERNEL__ -g
+CFLAGS += -I$(SRC)/include -D__KERNEL__ -g
 LDFLAGS += -e arch_start
 IHKOBJ = ihk/ihk.o
 

kernel/ap.c

@@ -24,20 +24,23 @@
 #include <process.h>
 #include <init.h>
 #include <march.h>
+#include <cls.h>
 
 int num_processors = 1;
 static volatile int ap_stop = 1;
 
 static void ap_wait(void)
 {
-	wrmsr(MSR_IA32_TIME_STAMP_COUNTER, 0);
-	
+	init_tick();
 	while (ap_stop) {
 		barrier();
 		cpu_pause();
 	}
+	sync_tick();
+
 	kmalloc_init();
 	sched_init();
+	arch_start_pvclock();
 
 	if (find_command_line("hidos")) {
 		init_host_syscall_channel();
@@ -53,19 +56,20 @@ static void ap_wait(void)
 
 void ap_start(void)
 {
+	init_tick();
 	ap_stop = 0;
+	sync_tick();
 }
 
 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();
 	
-	wrmsr(MSR_IA32_TIME_STAMP_COUNTER, 0);
-
 	cpu_info = ihk_mc_get_cpu_info();
 	bsp_hw_id = ihk_mc_get_hardware_processor_id();
 
@@ -74,18 +78,164 @@ void ap_init(void)
 		return;
 	}
 
-	kprintf("BSP HW ID = %d, ", bsp_hw_id);
-	kprintf("AP Booting :");
+	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 @ 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);
-		kprintf(" %d", cpu_info->hw_ids[i]);
 
 		num_processors++;
 	}
-	kprintf(" .. Done\n");
+	kprintf("AP Booting: Done\n");
 }
 
+#include <sysfs.h>
+#include <kmalloc.h>
+#include <string.h>
+#include <vsprintf.h>
+
+static ssize_t
+show_int(struct sysfs_ops *ops, void *instance, void *buf, size_t size)
+{
+	int *p = instance;
+
+	return snprintf(buf, size, "%d\n", *p);
+}/* show_int() */
+
+struct sysfs_ops show_int_ops = {
+	.show = &show_int,
+};
+
+struct fake_cpu_info {
+	int online;
+};
+
+static struct fake_cpu_info *fake_cpu_infos = NULL;
+
+enum fake_cpu_info_member {
+	ONLINE,
+};
+
+struct fake_cpu_info_ops {
+	enum fake_cpu_info_member member;
+	struct sysfs_ops ops;
+};
+
+static ssize_t
+show_fake_cpu_info(struct sysfs_ops *ops0, void *instance, void *buf,
+		size_t size)
+{
+	struct fake_cpu_info_ops *ops
+		= container_of(ops0, struct fake_cpu_info_ops, ops);
+	struct fake_cpu_info *info = instance;
+	ssize_t n;
+
+	switch (ops->member) {
+	case ONLINE:
+		n = snprintf(buf, size, "%d\n", info->online);
+		break;
+	default:
+		n = -EINVAL;
+		break;
+	}
+
+	if (n >= size) {
+		n = -ENOSPC;
+	}
+
+	return n;
+} /* show_fake_cpu_info() */
+
+static ssize_t
+store_fake_cpu_info(struct sysfs_ops *ops0, void *instance, void *buf,
+		size_t size)
+{
+	struct fake_cpu_info_ops *ops
+		= container_of(ops0, struct fake_cpu_info_ops, ops);
+	struct fake_cpu_info *info = instance;
+	ssize_t n;
+
+	switch (ops->member) {
+	case ONLINE:
+		kprintf("NYI:store_fake_cpu_info(%p,%p,%p,%ld): "
+				"online %d --> \"%.*s\"\n",
+				ops0, instance, buf, size, info->online,
+				(int)size, buf);
+		n = size;
+		break;
+	default:
+		n = -EIO;
+		break;
+	}
+
+	return n;
+} /* store_fake_cpu_info() */
+
+static struct fake_cpu_info_ops show_fci_online = {
+	.member = ONLINE,
+	.ops.show = &show_fake_cpu_info,
+	.ops.store = &store_fake_cpu_info,
+};
+
+void
+cpu_sysfs_setup(void)
+{
+	int error;
+	int cpu;
+	sysfs_handle_t targeth;
+	struct fake_cpu_info *info;
+
+	/* sample of simple variable **********************************/
+	error = sysfs_createf(&show_int_ops, &num_processors, 0444,
+			"/sys/devices/system/cpu/num_processors");
+	if (error) {
+		panic("cpu_sysfs_setup:sysfs_createf(num_processors) failed\n");
+	}
+
+	/* sample of more complex variable ****************************/
+	/* setup table */
+	info = kmalloc(sizeof(*info) * num_processors, IHK_MC_AP_CRITICAL);
+	for (cpu = 0; cpu < num_processors; ++cpu) {
+		info[cpu].online = 10+cpu;
+	}
+	fake_cpu_infos = info;
+
+	/* setup sysfs tree */
+	for (cpu = 0; cpu < num_processors; ++cpu) {
+		/* online */
+		error = sysfs_createf(&show_fci_online.ops,
+				&fake_cpu_infos[cpu], 0644,
+				"/sys/devices/system/cpu/cpu%d/online", cpu);
+		if (error) {
+			panic("cpu_sysfs_setup:sysfs_createf failed\n");
+		}
+
+		/* link to cpu%d */
+		error = sysfs_lookupf(&targeth,
+				"/sys/devices/system/cpu/cpu%d", cpu);
+		if (error) {
+			panic("cpu_sysfs_setup:sysfs_lookupf failed\n");
+		}
+
+		error = sysfs_symlinkf(targeth, "/sys/bus/cpu/devices/cpu%d",
+				cpu);
+		if (error) {
+			panic("cpu_sysfs_setup:sysfs_symlinkf failed\n");
+		}
+	}
+
+	return;
+} /* cpu_sysfs_setup() */

kernel/cls.c

@@ -23,6 +23,7 @@
 extern int num_processors;
 
 struct cpu_local_var *clv;
+static int cpu_local_var_initialized = 0;
 
 void cpu_local_var_init(void)
 {
@@ -31,11 +32,24 @@ 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;
 }
 
 struct cpu_local_var *get_cpu_local_var(int id)
 {
 	return clv + id;
 }
+
+void preempt_enable(void)
+{
+	if (cpu_local_var_initialized)
+		--cpu_local_var(no_preempt);
+}
+
+void preempt_disable(void)
+{
+	if (cpu_local_var_initialized)
+		++cpu_local_var(no_preempt);
+}

kernel/config/attached-mic.lds

@@ -26,6 +26,7 @@ SECTIONS
 
 		. = vsyscall_page + 0x000;
 		*(.vsyscall.gettimeofday)
+		*(.vsyscall.gettimeofday.*)
 
 		. = vsyscall_page + 0x400;
 		*(.vsyscall.time)

kernel/config/builtin-mic.lds

@@ -26,6 +26,7 @@ SECTIONS
 
 		. = vsyscall_page + 0x000;
 		*(.vsyscall.gettimeofday)
+		*(.vsyscall.gettimeofday.*)
 
 		. = vsyscall_page + 0x400;
 		*(.vsyscall.time)

kernel/config/builtin-x86.lds

@@ -26,6 +26,7 @@ SECTIONS
 
 		. = vsyscall_page + 0x000;
 		*(.vsyscall.gettimeofday)
+		*(.vsyscall.gettimeofday.*)
 
 		. = vsyscall_page + 0x400;
 		*(.vsyscall.time)

kernel/config/config.attached-mic

@@ -1,6 +1,5 @@
 CC = /usr/linux-k1om-4.7/bin/x86_64-k1om-linux-gcc
 LD = /usr/linux-k1om-4.7/bin/x86_64-k1om-linux-ld
 
-CFLAGS += -mno-sse -mno-mmx -mno-sse2 -mno-3dnow
 LDFLAGS += -m elf_k1om -T $(SRC)/config/attached-mic.lds
 LDFLAGS_MKIMAGE = -m elf_k1om

kernel/config/config.builtin-mic

@@ -3,6 +3,5 @@ LD = /usr/linux-k1om-4.7/bin/x86_64-k1om-linux-ld
 OBJDUMP = /usr/linux-k1om-4.7/bin/x86_64-k1om-linux-objdump
 OBJCOPY = /usr/linux-k1om-4.7/bin/x86_64-k1om-linux-objcopy
 
-CFLAGS += -mno-sse -mno-mmx -mno-sse2 -mno-3dnow
 LDFLAGS += -m elf_k1om -T $(SRC)/config/builtin-mic.lds
 LDFLAGS_MKIMAGE = -m elf_k1om

kernel/config/config.builtin-x86

@@ -1,2 +1 @@
-CFLAGS += -mno-sse -mno-mmx -mno-sse2 -mno-3dnow
 LDFLAGS += -T $(SRC)/config/builtin-x86.lds

kernel/config/config.smp-x86

@@ -1,2 +1 @@
-CFLAGS += -mno-sse -mno-mmx -mno-sse2 -mno-3dnow
 LDFLAGS += -T $(SRC)/config/smp-x86.lds

kernel/config/smp-x86.lds

@@ -26,6 +26,7 @@ SECTIONS
 
 		. = vsyscall_page + 0x000;
 		*(.vsyscall.gettimeofday)
+		*(.vsyscall.gettimeofday.*)
 
 		. = vsyscall_page + 0x400;
 		*(.vsyscall.time)
@@ -41,9 +42,4 @@ SECTIONS
     }
     . = ALIGN(4096);
     _end = .;
-
-	/DISCARD/ : {
-	     *(.eh_frame)
-         *(.note.gnu.build-id)
-	}
 }

kernel/debug.c

@@ -22,13 +22,46 @@ extern int vsnprintf(char *buf, size_t size, const char *fmt, va_list args);
 extern int sprintf(char * buf, const char *fmt, ...);
 static ihk_spinlock_t kmsg_lock;
 
+static unsigned long kprintf_lock_head(void);
+static void kprintf_unlock_head(unsigned long irqflags);
+
+static void kprintf_wait(int len, unsigned long *flags_head, int *slide) {
+	int head, tail, buf_len, mode, adj;
+
+	mode = kmsg_buf.mode;
+	while (1) {
+		adj = 0;
+		tail = kmsg_buf.tail;
+		buf_len = kmsg_buf.len;
+		head = kmsg_buf.head;
+		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;
+			} else {
+				kprintf_unlock_head(*flags_head);
+				*flags_head = kprintf_lock_head();
+			}
+		} else {
+			break;
+		}
+	}
+}
+
 /* TODO: lock */
 void kputs(char *buf)
 {
 	int len = strlen(buf);
-	unsigned long flags;
+	int slide = 0;
+	unsigned long flags_tail, flags_head;
 
-	flags = ihk_mc_spinlock_lock(&kmsg_lock);
+	flags_tail = kprintf_lock();
+	flags_head = kprintf_lock_head();
+	kprintf_wait(len, &flags_head, &slide);
 
 	if (len + kmsg_buf.tail > kmsg_buf.len) {
 		kmsg_buf.tail = 0;
@@ -39,60 +72,57 @@ void kputs(char *buf)
 	
 	memcpy(kmsg_buf.str + kmsg_buf.tail, buf, len);
 	kmsg_buf.tail += len;
-
-	ihk_mc_spinlock_unlock(&kmsg_lock, flags);
+	/* 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;
+	}
+	kprintf_unlock_head(flags_head);
+	kprintf_unlock(flags_tail);
 }
 
 #define KPRINTF_LOCAL_BUF_LEN 1024
 
 unsigned long kprintf_lock(void)
 {
-	return ihk_mc_spinlock_lock(&kmsg_lock);
+	return __ihk_mc_spinlock_lock(&kmsg_lock);
 }
 
 void kprintf_unlock(unsigned long irqflags)
 {
-	ihk_mc_spinlock_unlock(&kmsg_lock, irqflags);
+	__ihk_mc_spinlock_unlock(&kmsg_lock, irqflags);
+}
+
+static unsigned long kprintf_lock_head(void)
+{
+	return __ihk_mc_spinlock_lock(&kmsg_buf.lock);
+}
+
+static void kprintf_unlock_head(unsigned long irqflags)
+{
+	__ihk_mc_spinlock_unlock(&kmsg_buf.lock, irqflags);
 }
 
 /* Caller must hold kmsg_lock! */
 int __kprintf(const char *format, ...)
 {
 	int len = 0;
+	int slide = 0;
 	va_list va;
+	unsigned long flags_head;
 	char buf[KPRINTF_LOCAL_BUF_LEN];
 
-	/* Copy into the local buf */
-	va_start(va, format);
-	len += vsnprintf(buf + len, KPRINTF_LOCAL_BUF_LEN - len - 2, format, va);
-	va_end(va);
-
-	/* Append to kmsg buffer */
-	if (kmsg_buf.tail + len > kmsg_buf.len) {
-		kmsg_buf.tail = 0;
-	}
-
-	memcpy(kmsg_buf.str + kmsg_buf.tail, buf, len);
-	kmsg_buf.tail += len;
-
-	return len;
-}
-
-int kprintf(const char *format, ...)
-{
-	int len = 0;
-	va_list va;
-	unsigned long flags;
-	char buf[KPRINTF_LOCAL_BUF_LEN];
-
-	flags = ihk_mc_spinlock_lock(&kmsg_lock);
-
 	/* Copy into the local buf */
 	len = sprintf(buf, "[%3d]: ", ihk_mc_get_processor_id());
 	va_start(va, format);
 	len += vsnprintf(buf + len, KPRINTF_LOCAL_BUF_LEN - len - 2, format, va);
 	va_end(va);
 
+	flags_head = kprintf_lock_head();
+	kprintf_wait(len, &flags_head, &slide);
+
 	/* Append to kmsg buffer */
 	if (kmsg_buf.tail + len > kmsg_buf.len) {
 		kmsg_buf.tail = 0;
@@ -100,16 +130,69 @@ int kprintf(const char *format, ...)
 
 	memcpy(kmsg_buf.str + kmsg_buf.tail, buf, len);
 	kmsg_buf.tail += len;
+	if (slide == 1) {
+		kmsg_buf.head = kmsg_buf.tail + 1;
+		if (kmsg_buf.head >= kmsg_buf.len) kmsg_buf.head = 0;
+	}
 
-	ihk_mc_spinlock_unlock(&kmsg_lock, flags);
+	kprintf_unlock_head(flags_head);
+	return len;
+}
+
+int kprintf(const char *format, ...)
+{
+	int len = 0;
+	int slide = 0;
+	va_list va;
+	unsigned long flags_tail, flags_head;
+	char buf[KPRINTF_LOCAL_BUF_LEN];
+
+	/* Copy into the local buf */
+	len = sprintf(buf, "[%3d]: ", ihk_mc_get_processor_id());
+	va_start(va, format);
+	len += vsnprintf(buf + len, KPRINTF_LOCAL_BUF_LEN - len - 2, format, va);
+	va_end(va);
+
+	flags_tail = kprintf_lock();
+	flags_head = kprintf_lock_head();
+	kprintf_wait(len, &flags_head, &slide);
+
+	/* Append to kmsg buffer */
+	if (kmsg_buf.tail + len > kmsg_buf.len) {
+		kmsg_buf.tail = 0;
+	}
+
+	memcpy(kmsg_buf.str + kmsg_buf.tail, buf, len);
+	kmsg_buf.tail += len;
+	if (slide == 1) {
+		kmsg_buf.head = kmsg_buf.tail + 1;
+		if (kmsg_buf.head >= kmsg_buf.len) kmsg_buf.head = 0;
+	}
+
+	kprintf_unlock_head(flags_head);
+	kprintf_unlock(flags_tail);
 
 	return len;
 }
 
-void kmsg_init(void)
+/* 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);
 	kmsg_buf.tail = 0;
 	kmsg_buf.len = sizeof(kmsg_buf.str);
+	kmsg_buf.head = 0;
+	kmsg_buf.mode = mode;
+	ihk_mc_spinlock_init(&kmsg_buf.lock);
 	memset(kmsg_buf.str, 0, kmsg_buf.len);
 }

kernel/devobj.c

@@ -78,51 +78,52 @@ static struct memobj *to_memobj(struct devobj *devobj)
 /***********************************************************************
  * devobj
  */
-int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxprotp)
+int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxprotp,
+	int prot, int populate_flags)
 {
 	ihk_mc_user_context_t ctx;
 	struct pager_map_result result;	// XXX: assumes contiguous physical
 	int error;
 	struct devobj *obj  = NULL;
 	const size_t npages = (len + PAGE_SIZE - 1) / PAGE_SIZE;
+	const size_t pfn_npages = (npages / (PAGE_SIZE / sizeof(uintptr_t))) + 1;
 
-	dkprintf("devobj_create(%d,%lx,%lx)\n", fd, len, off);
-#define MAX_PAGES_IN_DEVOBJ (PAGE_SIZE / sizeof(uintptr_t))
-	if (npages > MAX_PAGES_IN_DEVOBJ) {
-		error = -EFBIG;
-		kprintf("devobj_create(%d,%lx,%lx):too large len. %d\n", fd, len, off, error);
-		goto out;
-	}
+	dkprintf("%s: fd: %d, len: %lu, off: %lu \n", __FUNCTION__, fd, len, off);
 
 	obj = kmalloc(sizeof(*obj), IHK_MC_AP_NOWAIT);
 	if (!obj) {
 		error = -ENOMEM;
-		kprintf("devobj_create(%d,%lx,%lx):kmalloc failed. %d\n", fd, len, off, error);
+		kprintf("%s: error: fd: %d, len: %lu, off: %lu kmalloc failed.\n", 
+			__FUNCTION__, fd, len, off);
 		goto out;
 	}
 	memset(obj, 0, sizeof(*obj));
 
-	obj->pfn_table = allocate_pages(1, IHK_MC_AP_NOWAIT);
+	obj->pfn_table = ihk_mc_alloc_pages(pfn_npages, IHK_MC_AP_NOWAIT);
 	if (!obj->pfn_table) {
 		error = -ENOMEM;
-		kprintf("devobj_create(%d,%lx,%lx):allocate_pages failed. %d\n", fd, len, off, error);
+		kprintf("%s: error: fd: %d, len: %lu, off: %lu allocating PFN failed.\n", 
+			__FUNCTION__, fd, len, off);
 		goto out;
 	}
-	memset(obj->pfn_table, 0, 1*PAGE_SIZE);
+	memset(obj->pfn_table, 0, pfn_npages * PAGE_SIZE);
 
 	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_MAP;
 	ihk_mc_syscall_arg1(&ctx) = fd;
 	ihk_mc_syscall_arg2(&ctx) = len;
 	ihk_mc_syscall_arg3(&ctx) = off;
 	ihk_mc_syscall_arg4(&ctx) = virt_to_phys(&result);
+	ihk_mc_syscall_arg5(&ctx) = prot | populate_flags;
 
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
 	if (error) {
-		kprintf("devobj_create(%d,%lx,%lx):map failed. %d\n", fd, len, off, error);
+		kprintf("%s: error: fd: %d, len: %lu, off: %lu map failed.\n", 
+			__FUNCTION__, fd, len, off);
 		goto out;
 	}
-	dkprintf("devobj_create:handle: %lx\n", result.handle);
-	dkprintf("devobj_create:maxprot: %x\n", result.maxprot);
+
+	dkprintf("%s: fd: %d, len: %lu, off: %lu, handle: %p, maxprot: %x\n", 
+		__FUNCTION__, fd, len, off, result.handle, result.maxprot);
 
 	obj->memobj.ops = &devobj_ops;
 	obj->memobj.flags = MF_HAS_PAGER;
@@ -140,11 +141,12 @@ int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxp
 out:
 	if (obj) {
 		if (obj->pfn_table) {
-			free_pages(obj->pfn_table, 1);
+			ihk_mc_free_pages(obj->pfn_table, pfn_npages);
 		}
 		kfree(obj);
 	}
-	dkprintf("devobj_create(%d,%lx,%lx): %d %p %x%d\n", fd, len, off, error, *objp, *maxprotp);
+	dkprintf("%s: ret: %d, fd: %d, len: %lu, off: %lu, handle: %p, maxprot: %x \n", 
+		__FUNCTION__, error, fd, len, off, result.handle, result.maxprot);
 	return error;
 }
 
@@ -164,6 +166,8 @@ static void devobj_release(struct memobj *memobj)
 	struct devobj *obj = to_devobj(memobj);
 	struct devobj *free_obj = NULL;
 	uintptr_t handle;
+	const size_t pfn_npages =
+		(obj->npages / (PAGE_SIZE / sizeof(uintptr_t))) + 1;
 
 	dkprintf("devobj_release(%p %lx)\n", obj, obj->handle);
 
@@ -192,7 +196,7 @@ static void devobj_release(struct memobj *memobj)
 		}
 
 		if (obj->pfn_table) {
-			free_pages(obj->pfn_table, 1);
+			ihk_mc_free_pages(obj->pfn_table, pfn_npages);
 		}
 		kfree(free_obj);
 	}
@@ -204,7 +208,7 @@ static void devobj_release(struct memobj *memobj)
 
 static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *flag)
 {
-	const off_t pgoff = off >> PAGE_SHIFT;
+	const off_t pgoff = off / PAGE_SIZE;
 	struct devobj *obj = to_devobj(memobj);
 	int error;
 	uintptr_t pfn;
@@ -216,7 +220,7 @@ static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintpt
 
 	if ((pgoff < obj->pfn_pgoff) || ((obj->pfn_pgoff + obj->npages) <= pgoff)) {
 		error = -EFBIG;
-		kprintf("devobj_get_page(%p %lx,%lx,%d): out of range. %d\n", memobj, obj->handle, off, p2align, error);
+		kprintf("%s: error: out of range: off: %lu, page off: %lu obj->npages: %d\n", __FUNCTION__, off, pgoff, obj->npages);
 		goto out;
 	}
 	ix = pgoff - obj->pfn_pgoff;

kernel/fileobj.c

@@ -47,6 +47,7 @@ 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;
 
 static struct memobj_ops fileobj_ops = {
 	.release =	&fileobj_release,
@@ -55,6 +56,7 @@ static struct memobj_ops fileobj_ops = {
 	.copy_page =	&fileobj_copy_page,
 	.flush_page =	&fileobj_flush_page,
 	.invalidate_page =	&fileobj_invalidate_page,
+	.lookup_page =	&fileobj_lookup_page,
 };
 
 static struct fileobj *to_fileobj(struct memobj *memobj)
@@ -180,7 +182,7 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
 
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
 	if (error) {
-		kprintf("fileobj_create(%d):create failed. %d\n", fd, error);
+		dkprintf("fileobj_create(%d):create failed. %d\n", fd, error);
 		goto out;
 	}
 
@@ -256,13 +258,24 @@ static void fileobj_release(struct memobj *memobj)
 		/* zap page_list */
 		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, 1);
+			}
+#if 0
 			count = ihk_atomic_sub_return(1, &page->count);
 
 			if (!((page->mode == PM_WILL_PAGEIO)
@@ -279,7 +292,7 @@ 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);
@@ -387,7 +400,7 @@ out:
 
 static int fileobj_get_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *pflag)
 {
-	struct process *proc = cpu_local_var(current);
+	struct thread *proc = cpu_local_var(current);
 	struct fileobj *obj = to_fileobj(memobj);
 	int error;
 	void *virt = NULL;
@@ -428,7 +441,7 @@ 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");
 			}
@@ -500,10 +513,10 @@ static uintptr_t fileobj_copy_page(
 
 	memobj_lock(memobj);
 	for (;;) {
-		if (orgpage->mode != PM_MAPPED) {
+		if (!orgpage || orgpage->mode != PM_MAPPED) {
 			kprintf("fileobj_copy_page(%p,%lx,%d):"
 					"invalid cow page. %x\n",
-					memobj, orgpa, p2align, orgpage->mode);
+					memobj, orgpa, p2align, orgpage ? orgpage->mode : 0);
 			panic("fileobj_copy_page:invalid cow page");
 		}
 		count = ihk_atomic_read(&orgpage->count);
@@ -525,7 +538,9 @@ static uintptr_t fileobj_copy_page(
 			memcpy(newkva, orgkva, pgsize);
 			ihk_atomic_dec(&orgpage->count);
 			newpa = virt_to_phys(newkva);
-			page_map(phys_to_page(newpa));
+			if (phys_to_page(newpa)) {
+				page_map(phys_to_page(newpa));
+			}
 			newkva = NULL;	/* avoid ihk_mc_free_pages() */
 			break;
 		}
@@ -561,6 +576,11 @@ static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 	ssize_t ss;
 
 	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;
@@ -609,3 +629,37 @@ out:
 			memobj, phys, pgsize, error);
 	return error;
 }
+
+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;
+	struct page *page;
+
+	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;
+	}
+
+	page = page_list_lookup(obj, 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);
+
+	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);
+	return error;
+}

kernel/futex.c

@@ -103,7 +103,7 @@ int futex_cmpxchg_enabled;
 struct futex_q {
 	struct plist_node list;
 
-	struct process *task;
+	struct thread *task;
 	ihk_spinlock_t *lock_ptr;
 	union futex_key key;
 	union futex_key *requeue_pi_key;
@@ -151,7 +151,7 @@ static inline int match_futex(union futex_key *key1, union futex_key *key2)
  */
 static void get_futex_key_refs(union futex_key *key)
 {
-	/* RIKEN: only !fshared futexes... */
+	/* RIKEN: no swapping in McKernel */
 	return;
 }
 
@@ -161,7 +161,7 @@ static void get_futex_key_refs(union futex_key *key)
  */
 static void drop_futex_key_refs(union futex_key *key)
 {
-	/* RIKEN: only !fshared futexes... */
+	/* RIKEN: no swapping in McKernel */
 	return;
 }
 /**
@@ -183,6 +183,7 @@ static int
 get_futex_key(uint32_t *uaddr, int fshared, union futex_key *key)
 {
 	unsigned long address = (unsigned long)uaddr;
+	unsigned long phys;
 	struct process_vm *mm = cpu_local_var(current)->vm;
 
 	/*
@@ -201,15 +202,31 @@ get_futex_key(uint32_t *uaddr, int fshared, union futex_key *key)
 	 *        but access_ok() should be faster than find_vma()
 	 */
 	if (!fshared) {
-
 		key->private.mm = mm;
 		key->private.address = address;
 		get_futex_key_refs(key);
 		return 0;
 	}
 
-	/* RIKEN: No shared futex support... */
-	return -EFAULT;
+	key->both.offset |= FUT_OFF_MMSHARED;
+
+retry_v2p:
+	/* Just use physical address of page, McKernel does not do swapping */
+	if (ihk_mc_pt_virt_to_phys(mm->address_space->page_table, 
+				(void *)uaddr, &phys)) { 
+
+		/* Check if we can fault in page */
+		if (page_fault_process_vm(mm, uaddr, PF_POPULATE | PF_WRITE | PF_USER)) {
+			kprintf("error: get_futex_key() virt to phys translation failed\n");
+			return -EFAULT;
+		}
+
+		goto retry_v2p;
+	}
+	key->shared.phys = (void *)phys;
+	key->shared.pgoff = 0;
+
+	return 0;
 }
 
 
@@ -232,7 +249,7 @@ static int cmpxchg_futex_value_locked(uint32_t __user *uaddr, uint32_t uval, uin
 static int get_futex_value_locked(uint32_t *dest, uint32_t *from)
 {
 	/* RIKEN: futexes are always on not swappable pages */
-	*dest = *from;
+	*dest = getint_user((int *)from);
 
 	return 0;
 }
@@ -243,7 +260,7 @@ static int get_futex_value_locked(uint32_t *dest, uint32_t *from)
  */
 static void wake_futex(struct futex_q *q)
 {
-	struct process *p = q->task;
+	struct thread *p = q->task;
 
 	/*
 	 * We set q->lock_ptr = NULL _before_ we wake up the task. If
@@ -263,7 +280,8 @@ static void wake_futex(struct futex_q *q)
 	barrier();
 	q->lock_ptr = NULL;
 
-	sched_wakeup_process(p, PS_NORMAL);
+	dkprintf("wake_futex(): waking up tid %d\n", p->tid);
+	sched_wakeup_thread(p, PS_NORMAL);
 }
 
 /*
@@ -658,23 +676,27 @@ static uint64_t futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q
 	 * queue_me() calls spin_unlock() upon completion, both serializing
 	 * access to the hash list and forcing another memory barrier.
 	 */
-	xchg4(&(cpu_local_var(current)->ftn->status), PS_INTERRUPTIBLE);
+	xchg4(&(cpu_local_var(current)->status), PS_INTERRUPTIBLE);
 	queue_me(q, hb);
 	
 	if (!plist_node_empty(&q->list)) {
 		
 		/* RIKEN: use mcos timers */
 		if (timeout) {
+			dkprintf("futex_wait_queue_me(): tid: %d schedule_timeout()\n", cpu_local_var(current)->tid);
 			time_remain = schedule_timeout(timeout);
 		}
 		else {
+			dkprintf("futex_wait_queue_me(): tid: %d schedule()\n", cpu_local_var(current)->tid);
 			schedule();
 			time_remain = 0;
 		}
+		
+		dkprintf("futex_wait_queue_me(): tid: %d woken up\n", cpu_local_var(current)->tid);
 	}
 	
 	/* This does not need to be serialized */
-	cpu_local_var(current)->ftn->status = PS_RUNNING;
+	cpu_local_var(current)->status = PS_RUNNING;
 	
 	return time_remain;
 }
@@ -775,6 +797,11 @@ retry:
 	if (timeout && !time_remain)
 		goto out_put_key;
 
+	if (hassigpending(cpu_local_var(current))) {
+		ret = -EINTR;
+		goto out_put_key;
+	}
+
 	/* RIKEN: no signals */
 	put_futex_key(fshared, &q.key);
 	goto retry;
@@ -786,17 +813,10 @@ out:
 }
 
 int futex(uint32_t *uaddr, int op, uint32_t val, uint64_t timeout,
-		uint32_t *uaddr2, uint32_t val2, uint32_t val3)
+		uint32_t *uaddr2, uint32_t val2, uint32_t val3, int fshared)
 {
 	int clockrt, ret = -ENOSYS;
 	int cmd = op & FUTEX_CMD_MASK;
-	int fshared = 0;
-
-	/* RIKEN: Assume address space private futexes. 
-	if (!(op & FUTEX_PRIVATE_FLAG)) {
-		fshared = 1;
-	}
-	*/
 
 	clockrt = op & FUTEX_CLOCK_REALTIME;
 	if (clockrt && cmd != FUTEX_WAIT_BITSET && cmd != FUTEX_WAIT_REQUEUE_PI)
@@ -817,8 +837,7 @@ int futex(uint32_t *uaddr, int op, uint32_t val, uint64_t timeout,
 		ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, NULL, 0);
 		break;
 	case FUTEX_CMP_REQUEUE:
-		ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, &val3,
-				    0);
+		ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, &val3, 0);
 		break;
 	case FUTEX_WAKE_OP:
 		ret = futex_wake_op(uaddr, fshared, uaddr2, val, val2, val3);

kernel/host.c

@@ -30,6 +30,7 @@
 #include <mman.h>
 #include <init.h>
 #include <kmalloc.h>
+#include <sysfs.h>
 
 //#define DEBUG_PRINT_HOST
 
@@ -39,11 +40,11 @@
 #define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
 #endif
 
-void check_mapping_for_proc(struct process *proc, unsigned long addr)
+void check_mapping_for_proc(struct thread *thread, unsigned long addr)
 {
 	unsigned long __phys;
 
-	if (ihk_mc_pt_virt_to_phys(proc->vm->page_table, (void*)addr, &__phys)) {
+	if (ihk_mc_pt_virt_to_phys(thread->vm->address_space->page_table, (void*)addr, &__phys)) {
 		kprintf("check_map: no mapping for 0x%lX\n", addr);
 	}
 	else {
@@ -60,7 +61,7 @@ void check_mapping_for_proc(struct process *proc, unsigned long addr)
  * NOTE: if args, args_len, envs, envs_len are zero, 
  * the function constructs them based on the descriptor 
  */
-int prepare_process_ranges_args_envs(struct process *proc, 
+int prepare_process_ranges_args_envs(struct thread *thread, 
 		struct program_load_desc *pn,
 		struct program_load_desc *p,
 		enum ihk_mc_pt_attribute attr,
@@ -80,15 +81,21 @@ int prepare_process_ranges_args_envs(struct process *proc,
 	unsigned long flags;
 	uintptr_t interp_obase = -1;
 	uintptr_t interp_nbase = -1;
+	size_t map_size;
+	struct process *proc = thread->proc;
+	struct process_vm *vm = proc->vm;
+	struct address_space *as = vm->address_space;
+	long aout_base;
+	int error;
 	
 	n = p->num_sections;
 
+	aout_base = (pn->reloc)? vm->region.map_end: 0;
 	for (i = 0; i < n; i++) {
-		
 		if (pn->sections[i].interp && (interp_nbase == (uintptr_t)-1)) {
 			interp_obase = pn->sections[i].vaddr;
 			interp_obase -= (interp_obase % pn->interp_align);
-			interp_nbase = proc->vm->region.map_start;
+			interp_nbase = vm->region.map_end;
 			interp_nbase = (interp_nbase + pn->interp_align - 1)
 				& ~(pn->interp_align - 1);
 		}
@@ -98,6 +105,10 @@ int prepare_process_ranges_args_envs(struct process *proc,
 			pn->sections[i].vaddr += interp_nbase;
 			p->sections[i].vaddr = pn->sections[i].vaddr;
 		}
+		else{
+			pn->sections[i].vaddr += aout_base;
+			p->sections[i].vaddr = pn->sections[i].vaddr;
+		}
 		s = (pn->sections[i].vaddr) & PAGE_MASK;
 		e = (pn->sections[i].vaddr + pn->sections[i].len
 				+ PAGE_SIZE - 1) & PAGE_MASK;
@@ -113,7 +124,8 @@ int prepare_process_ranges_args_envs(struct process *proc,
 		} 
 		
 		up = virt_to_phys(up_v);
-		if (add_process_memory_range(proc, s, e, up, flags, NULL, 0) != 0) {
+		if (add_process_memory_range(vm, s, e, up, flags, NULL, 0,
+					PAGE_SHIFT) != 0) {
 			ihk_mc_free_pages(up_v, range_npages);
 			kprintf("ERROR: adding memory range for ELF section %i\n", i);
 			goto err;
@@ -122,14 +134,14 @@ int prepare_process_ranges_args_envs(struct process *proc,
 		{
 			void *_virt = (void *)s;
 			unsigned long _phys;
-			if (ihk_mc_pt_virt_to_phys(proc->vm->page_table, 
+			if (ihk_mc_pt_virt_to_phys(as->page_table, 
 						_virt, &_phys)) {
 				kprintf("ERROR: no mapping for 0x%lX\n", _virt);
 			}
 			for (_virt = (void *)s + PAGE_SIZE; 
 					(unsigned long)_virt < e; _virt += PAGE_SIZE) {
 				unsigned long __phys;
-				if (ihk_mc_pt_virt_to_phys(proc->vm->page_table, 
+				if (ihk_mc_pt_virt_to_phys(as->page_table, 
 							_virt, &__phys)) {
 					kprintf("ERROR: no mapping for 0x%lX\n", _virt);
 					panic("mapping");
@@ -148,23 +160,27 @@ int prepare_process_ranges_args_envs(struct process *proc,
 
 		/* TODO: Maybe we need flag */
 		if (pn->sections[i].interp) {
-			proc->vm->region.map_end = e;
+			vm->region.map_end = e;
 		}
 		else if (i == 0) {
-			proc->vm->region.text_start = s;
-			proc->vm->region.text_end = e;
+			vm->region.text_start = s;
+			vm->region.text_end = e;
 		} 
 		else if (i == 1) {
-			proc->vm->region.data_start = s;
-			proc->vm->region.data_end = e;
+			vm->region.data_start = s;
+			vm->region.data_end = e;
 		} 
 		else {
-			proc->vm->region.data_start =
-				(s < proc->vm->region.data_start ? 
-				 s : proc->vm->region.data_start);
-			proc->vm->region.data_end = 
-				(e > proc->vm->region.data_end ? 
-				 e : proc->vm->region.data_end);
+			vm->region.data_start =
+				(s < vm->region.data_start ? 
+				 s : vm->region.data_start);
+			vm->region.data_end = 
+				(e > vm->region.data_end ? 
+				 e : vm->region.data_end);
+		}
+
+		if (aout_base) {
+			vm->region.map_end = e;
 		}
 	}
 
@@ -172,17 +188,22 @@ int prepare_process_ranges_args_envs(struct process *proc,
 		pn->entry -= interp_obase;
 		pn->entry += interp_nbase;
 		p->entry = pn->entry;
-		ihk_mc_modify_user_context(proc->uctx, IHK_UCR_PROGRAM_COUNTER, 
-				pn->entry);
+		ihk_mc_modify_user_context(thread->uctx,
+		                           IHK_UCR_PROGRAM_COUNTER, 
+		                           pn->entry);
 	}
 
-	proc->vm->region.brk_start = proc->vm->region.brk_end =
-		proc->vm->region.data_end;
+	if (aout_base) {
+		pn->at_phdr += aout_base;
+		pn->at_entry += aout_base;
+	}
+
+	vm->region.brk_start = vm->region.brk_end = vm->region.data_end;
 
 	/* Map, copy and update args and envs */
 	flags = VR_PROT_READ | VR_PROT_WRITE;
 	flags |= VRFLAG_PROT_TO_MAXPROT(flags);
-	addr = proc->vm->region.map_start - PAGE_SIZE * SCD_RESERVED_COUNT;
+	addr = vm->region.map_start - PAGE_SIZE * SCD_RESERVED_COUNT;
 	e = addr + PAGE_SIZE * ARGENV_PAGE_COUNT;
 
 	if((args_envs = ihk_mc_alloc_pages(ARGENV_PAGE_COUNT, IHK_MC_AP_NOWAIT)) == NULL){
@@ -191,8 +212,8 @@ int prepare_process_ranges_args_envs(struct process *proc,
 	}
 	args_envs_p = virt_to_phys(args_envs);
 
-	if(add_process_memory_range(proc, addr, e, args_envs_p,
-				flags, NULL, 0) != 0){
+	if(add_process_memory_range(vm, addr, e, args_envs_p,
+				flags, NULL, 0, PAGE_SHIFT) != 0){
 		ihk_mc_free_pages(args_envs, ARGENV_PAGE_COUNT);
 		kprintf("ERROR: adding memory range for args/envs\n");
 		goto err;
@@ -205,7 +226,8 @@ int prepare_process_ranges_args_envs(struct process *proc,
 	/* Only map remote address if it wasn't specified as an argument */
 	if (!args) {
 		// Map in remote physical addr of args and copy it
-		args_envs_npages = (p->args_len + PAGE_SIZE - 1) >> PAGE_SHIFT;
+		map_size = ((uintptr_t)p->args & (PAGE_SIZE - 1)) + p->args_len;
+		args_envs_npages = (map_size + PAGE_SIZE - 1) >> PAGE_SHIFT;
 		dkprintf("args_envs_npages: %d\n", args_envs_npages);
 		args_envs_rp = ihk_mc_map_memory(NULL, 
 				(unsigned long)p->args, p->args_len);
@@ -222,9 +244,9 @@ int prepare_process_ranges_args_envs(struct process *proc,
 		p->args_len = args_len;
 	}
 
-	dkprintf("args copy, nr: %d\n", *((int*)args_envs_r));
+	dkprintf("args copy, nr: %d\n", *((long *)args_envs_r));
 
-	memcpy_long(args_envs, args_envs_r, p->args_len + 8);
+	memcpy_long(args_envs, args_envs_r, p->args_len + sizeof(long) - 1);
 
 	/* Only unmap remote address if it wasn't specified as an argument */
 	if (!args) {
@@ -238,7 +260,8 @@ int prepare_process_ranges_args_envs(struct process *proc,
 	/* Only map remote address if it wasn't specified as an argument */
 	if (!envs) {
 		// Map in remote physical addr of envs and copy it after args
-		args_envs_npages = (p->envs_len + PAGE_SIZE - 1) >> PAGE_SHIFT;
+		map_size = ((uintptr_t)p->envs & (PAGE_SIZE - 1)) + p->envs_len;
+		args_envs_npages = (map_size + PAGE_SIZE - 1) >> PAGE_SHIFT;
 		dkprintf("args_envs_npages: %d\n", args_envs_npages);
 		args_envs_rp = ihk_mc_map_memory(NULL, (unsigned long)p->envs, 
 				p->envs_len);
@@ -256,9 +279,9 @@ int prepare_process_ranges_args_envs(struct process *proc,
 		p->envs_len = envs_len;
 	}
 
-	dkprintf("envs copy, nr: %d\n", *((int*)args_envs_r));
+	dkprintf("envs copy, nr: %d\n", *((long *)args_envs_r));
 
-	memcpy_long(args_envs + p->args_len, args_envs_r, p->envs_len + 8);
+	memcpy_long(args_envs + p->args_len, args_envs_r, p->envs_len + sizeof(long) - 1);
 
 	/* Only map remote address if it wasn't specified as an argument */
 	if (!envs) {
@@ -268,10 +291,10 @@ int prepare_process_ranges_args_envs(struct process *proc,
 	flush_tlb();
 
 	// Update variables
-	argc = *((int*)(args_envs));
+	argc = *((long *)(args_envs));
 	dkprintf("argc: %d\n", argc);
 
-	argv = (char **)(args_envs + (sizeof(int)));
+	argv = (char **)(args_envs + (sizeof(long)));
 	if(proc->saved_cmdline){
 		kfree(proc->saved_cmdline);
 		proc->saved_cmdline_len = 0;
@@ -288,24 +311,35 @@ int prepare_process_ranges_args_envs(struct process *proc,
 		*a = (char *)addr + (unsigned long)*a; // Process' address space!
 	}
 
-	envc = *((int*)(args_envs + p->args_len));
+	envc = *((long *)(args_envs + p->args_len));
 	dkprintf("envc: %d\n", envc);
 
-	env = (char **)(args_envs + p->args_len + sizeof(int));
+	env = (char **)(args_envs + p->args_len + sizeof(long));
 	while (*env) {
 		char **_env = env;
 		//dkprintf("%s\n", args_envs + p->args_len + (unsigned long)*env);
 		*env = (char *)addr + p->args_len + (unsigned long)*env;
 		env = ++_env;
 	}
-	env = (char **)(args_envs + p->args_len + sizeof(int));
+	env = (char **)(args_envs + p->args_len + sizeof(long));
 
 	dkprintf("env OK\n");
 
-	p->rprocess = (unsigned long)proc;
-	p->rpgtable = virt_to_phys(proc->vm->page_table);
+	if (pn->enable_vdso) {
+		error = arch_map_vdso(vm);
+		if (error) {
+			kprintf("ERROR: mapping vdso pages. %d\n", error);
+			goto err;
+		}
+	}
+	else {
+		vm->vdso_addr = NULL;
+	}
 
-	if (init_process_stack(proc, pn, argc, argv, envc, env) != 0) {
+	p->rprocess = (unsigned long)thread;
+	p->rpgtable = virt_to_phys(as->page_table);
+
+	if (init_process_stack(thread, pn, argc, argv, envc, env) != 0) {
 		goto err;
 	}
 
@@ -324,7 +358,9 @@ static int process_msg_prepare_process(unsigned long rphys)
 	unsigned long phys, sz;
 	struct program_load_desc *p, *pn;
 	int npages, n;
+	struct thread *thread;
 	struct process *proc;
+	struct process_vm *vm;
 	enum ihk_mc_pt_attribute attr;
 
 	attr = PTATTR_NO_EXECUTE | PTATTR_WRITABLE | PTATTR_FOR_USER;
@@ -340,10 +376,16 @@ static int process_msg_prepare_process(unsigned long rphys)
 	}
 
 	n = p->num_sections;
+	if (n > 16) {
+		kprintf("%s: ERROR: more ELF sections than 16??\n",
+			__FUNCTION__);
+		return -ENOMEM;
+	}
 	dkprintf("# of sections: %d\n", n);
 
-	if((pn = ihk_mc_allocate(sizeof(struct program_load_desc) 
-	       + sizeof(struct program_image_section) * n, IHK_MC_AP_NOWAIT)) == NULL){
+	if((pn = kmalloc(sizeof(struct program_load_desc) 
+					+ sizeof(struct program_image_section) * n,
+					IHK_MC_AP_NOWAIT)) == NULL){
 		ihk_mc_unmap_virtual(p, npages, 0);
 		ihk_mc_unmap_memory(NULL, phys, sz);
 		return -ENOMEM;
@@ -351,43 +393,58 @@ 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((proc = create_process(p->entry)) == NULL){
-		ihk_mc_free(pn);
+	if((thread = create_thread(p->entry)) == NULL){
+		kfree(pn);
 		ihk_mc_unmap_virtual(p, npages, 1);
 		ihk_mc_unmap_memory(NULL, phys, sz);
 		return -ENOMEM;
 	}
-	proc->ftn->pid = pn->pid;
-	proc->ftn->pgid = pn->pgid;
+	proc = thread->proc;
+	vm = thread->vm;
 
-	proc->ftn->ruid = pn->cred[0];
-	proc->ftn->euid = pn->cred[1];
-	proc->ftn->suid = pn->cred[2];
-	proc->ftn->fsuid = pn->cred[3];
-	proc->ftn->rgid = pn->cred[4];
-	proc->ftn->egid = pn->cred[5];
-	proc->ftn->sgid = pn->cred[6];
-	proc->ftn->fsgid = pn->cred[7];
+	proc->pid = pn->pid;
+	proc->vm->address_space->pids[0] = pn->pid;
+	proc->pgid = pn->pgid;
+	proc->ruid = pn->cred[0];
+	proc->euid = pn->cred[1];
+	proc->suid = pn->cred[2];
+	proc->fsuid = pn->cred[3];
+	proc->rgid = pn->cred[4];
+	proc->egid = pn->cred[5];
+	proc->sgid = pn->cred[6];
+	proc->fsgid = pn->cred[7];
+	proc->termsig = SIGCHLD;
 
-	proc->vm->region.user_start = pn->user_start;
-	proc->vm->region.user_end = pn->user_end;
-	proc->vm->region.map_start = (USER_END / 3) & LARGE_PAGE_MASK;
-	proc->vm->region.map_end = proc->vm->region.map_start;
+	vm->region.user_start = pn->user_start;
+	vm->region.user_end = pn->user_end;
+	if(vm->region.user_end > USER_END)
+		vm->region.user_end = USER_END;
+	if(vm->region.user_start != 0UL ||
+	   vm->region.user_end < TASK_UNMAPPED_BASE){
+		vm->region.map_start = 
+			(vm->region.user_start +
+			 (vm->region.user_end - vm->region.user_start) / 3) &
+			LARGE_PAGE_MASK;
+	}
+	else{
+		vm->region.map_start = TASK_UNMAPPED_BASE;
+	}
+	vm->region.map_end = vm->region.map_start;
 	memcpy(proc->rlimit, pn->rlimit, sizeof(struct rlimit) * MCK_RLIM_MAX);
 
 	/* TODO: Clear it at the proper timing */
 	cpu_local_var(scp).post_idx = 0;
 
-	if (prepare_process_ranges_args_envs(proc, pn, p, attr, 
+	if (prepare_process_ranges_args_envs(thread, pn, p, attr, 
 				NULL, 0, NULL, 0) != 0) {
 		kprintf("error: preparing process ranges, args, envs, stack\n");
 		goto err;
 	}
 
-	dkprintf("new process : %p [%d] / table : %p\n", proc, proc->ftn->pid,
-	        proc->vm->page_table);
+	dkprintf("new process : %p [%d] / table : %p\n", proc, proc->pid,
+	        vm->address_space->page_table);
 
-	ihk_mc_free(pn);
+	kfree(pn);
 
 	ihk_mc_unmap_virtual(p, npages, 1);
 	ihk_mc_unmap_memory(NULL, phys, sz);
@@ -395,17 +452,16 @@ static int process_msg_prepare_process(unsigned long rphys)
 
 	return 0;
 err:
-	ihk_mc_free(pn);
+	kfree(pn);
 	ihk_mc_unmap_virtual(p, npages, 1);
 	ihk_mc_unmap_memory(NULL, phys, sz);
-	free_process_memory(proc);
-	destroy_process(proc);
+	destroy_thread(thread);
 	return -ENOMEM;
 }
 
 static void process_msg_init(struct ikc_scd_init_param *pcp, struct syscall_params *lparam)
 {
-	lparam->response_va = allocate_pages(RESPONSE_PAGE_COUNT, 0);
+	lparam->response_va = ihk_mc_alloc_pages(RESPONSE_PAGE_COUNT, 0);
 	lparam->response_pa = virt_to_phys(lparam->response_va);
 
 	pcp->request_page = 0;
@@ -415,7 +471,7 @@ static void process_msg_init(struct ikc_scd_init_param *pcp, struct syscall_para
 
 static void process_msg_init_acked(struct ihk_ikc_channel_desc *c, unsigned long pphys)
 {
-	struct ikc_scd_init_param *param = (void *)pphys;
+	struct ikc_scd_init_param *param = phys_to_virt(pphys);
 	struct syscall_params *lparam;
 	enum ihk_mc_pt_attribute attr;
 
@@ -473,13 +529,8 @@ static void syscall_channel_send(struct ihk_ikc_channel_desc *c,
 	ihk_ikc_send(c, packet, 0);
 }
 
-extern unsigned long do_kill(int, int, int, struct siginfo *, int ptracecont);
-extern void settid(struct process *proc, int mode, int newcpuid, int oldcpuid);
-
+extern unsigned long do_kill(struct thread *, int, int, int, struct siginfo *, int ptracecont);
 extern void process_procfs_request(unsigned long rarg);
-extern int memcheckall();
-extern int freecheck(int runcount);
-extern int runcount;
 extern void terminate_host(int pid);
 extern void debug_log(long);
 
@@ -489,6 +540,8 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	struct ikc_scd_packet *packet = __packet;
 	struct ikc_scd_packet pckt;
 	int rc;
+	struct mcs_rwlock_node_irqsave lock;
+	struct thread *thread;
 	struct process *proc;
 	struct mcctrl_signal {
 		int	cond;
@@ -499,22 +552,17 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	} *sp, info;
 	unsigned long pp;
 	int cpuid;
+	int ret = 0;
 
 	switch (packet->msg) {
 	case SCD_MSG_INIT_CHANNEL_ACKED:
 		dkprintf("SCD_MSG_INIT_CHANNEL_ACKED\n");
 		process_msg_init_acked(c, packet->arg);
-		return 0;
+		ret = 0;
+		break;
 
 	case SCD_MSG_PREPARE_PROCESS:
 
-		if (find_command_line("memdebug")) {
-			memcheckall();
-			if (runcount)
-				freecheck(runcount);
-			runcount++;
-		}
-
 		if((rc = process_msg_prepare_process(packet->arg)) == 0){
 			pckt.msg = SCD_MSG_PREPARE_PROCESS_ACKED;
 			pckt.err = 0;
@@ -527,23 +575,50 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		pckt.arg = packet->arg;
 		syscall_channel_send(c, &pckt);
 
-		return 0;
+		ret = 0;
+		break;
 
 	case SCD_MSG_SCHEDULE_PROCESS:
 		cpuid = obtain_clone_cpuid();
 		if(cpuid == -1){
 			kprintf("No CPU available\n");
-			return -1;
+			ret = -1;
+			break;
 		}
 		dkprintf("SCD_MSG_SCHEDULE_PROCESS: %lx\n", packet->arg);
-		proc = (struct process *)packet->arg;
-
-		settid(proc, 0, cpuid, -1);
-		proc->ftn->status = PS_RUNNING;
-		runq_add_proc(proc, cpuid);
+		thread = (struct thread *)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 process *)packet->arg;
-		return 0;
+		//cpu_local_var(next) = (struct thread *)packet->arg;
+		ret = 0;
+		break;
+
+	/*
+	 * Used for syscall offload reply message to explicitly schedule in
+	 * the waiting thread
+	 */
+	case SCD_MSG_WAKE_UP_SYSCALL_THREAD:
+		thread = find_thread(0, packet->ttid, &lock);
+		if (!thread) {
+			kprintf("%s: WARNING: no thread for SCD reply? TID: %d\n",
+				__FUNCTION__, packet->ttid);
+			ret = -EINVAL;
+			break;
+		}
+		thread_unlock(thread, &lock);
+
+		dkprintf("%s: SCD_MSG_WAKE_UP_SYSCALL_THREAD: waking up tid %d\n",
+			__FUNCTION__, packet->ttid);
+		waitq_wakeup(&thread->scd_wq);
+		ret = 0;
+		break;
+
 	case SCD_MSG_SEND_SIGNAL:
 		pp = ihk_mc_map_memory(NULL, packet->arg, sizeof(struct mcctrl_signal));
 		sp = (struct mcctrl_signal *)ihk_mc_map_virtual(pp, 1, PTATTR_WRITABLE | PTATTR_ACTIVE);
@@ -556,22 +631,49 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		pckt.arg = packet->arg;
 		syscall_channel_send(c, &pckt);
 
-		rc = do_kill(info.pid, info.tid, info.sig, &info.info, 0);
+		rc = do_kill(NULL, info.pid, info.tid, info.sig, &info.info, 0);
 		kprintf("SCD_MSG_SEND_SIGNAL: do_kill(pid=%d, tid=%d, sig=%d)=%d\n", info.pid, info.tid, info.sig, rc);
-		return 0;
+		ret = 0;
+		break;
+
 	case SCD_MSG_PROCFS_REQUEST:
 		process_procfs_request(packet->arg);
-		return 0;
+		ret = 0;
+		break;
+
 	case SCD_MSG_CLEANUP_PROCESS:
 		dkprintf("SCD_MSG_CLEANUP_PROCESS pid=%d\n", packet->pid);
 		terminate_host(packet->pid);
-		return 0;
+		ret = 0;
+		break;
+
 	case SCD_MSG_DEBUG_LOG:
 		dkprintf("SCD_MSG_DEBUG_LOG code=%lx\n", packet->arg);
 		debug_log(packet->arg);
-		return 0;
+		ret = 0;
+		break;
+
+	case SCD_MSG_SYSFS_REQ_SHOW:
+	case SCD_MSG_SYSFS_REQ_STORE:
+	case SCD_MSG_SYSFS_REQ_RELEASE:
+		sysfss_packet_handler(c, packet->msg, packet->err,
+				packet->sysfs_arg1, packet->sysfs_arg2,
+				packet->sysfs_arg3);
+		ret = 0;
+		break;
+
+	default:
+		kprintf("syscall_pakcet_handler:unknown message "
+				"(%d.%d.%d.%d.%d.%#lx)\n",
+				packet->msg, packet->ref, packet->osnum,
+				packet->pid, packet->err, packet->arg);
+		ret = 0;
+		break;
+
 	}
-	return 0;
+
+	ihk_ikc_release_packet((struct ihk_ikc_free_packet *)packet, c);
+	return ret;
 }
 
 void init_host_syscall_channel(void)

kernel/include/auxvec.h

@@ -1,6 +1,8 @@
 #ifndef _LINUX_AUXVEC_H
 #define _LINUX_AUXVEC_H
 
+#include <arch/auxvec.h>
+
 /* Symbolic values for the entries in the auxiliary table
    put on the initial stack */
 #define AT_NULL   0	/* end of vector */

kernel/include/cls.h

@@ -19,11 +19,13 @@
  * CPU Local Storage (cls)
  */
 
-struct malloc_header {
-	unsigned int check;
+struct kmalloc_header {
+	unsigned int front_magic;
 	unsigned int cpu_id;
-	struct malloc_header *next;
-	unsigned long size;
+	struct list_head list;
+	int size; /* The size of this chunk without the header */
+	unsigned int end_magic;
+	/* 32 bytes */
 };
 
 #include <ihk/lock.h>
@@ -38,15 +40,18 @@ extern ihk_spinlock_t	cpu_status_lock;
 
 struct cpu_local_var {
 	/* malloc */
-	struct malloc_header free_list;
-	ihk_spinlock_t free_list_lock;
+	struct list_head free_list;
+	struct list_head remote_free_list;
+	ihk_spinlock_t remote_free_list_lock;
 
-	struct process idle;
-	struct fork_tree_node idle_ftn;
+	struct thread idle;
+	struct process idle_proc;
 	struct process_vm idle_vm;
+	struct address_space idle_asp;
 
 	ihk_spinlock_t runq_lock;
-	struct process *current;
+	unsigned long runq_irqstate;
+	struct thread *current;
 	struct list_head runq;
 	size_t runq_len;
 
@@ -57,6 +62,7 @@ struct cpu_local_var {
 	struct ihk_ikc_channel_desc *syscall_channel2;
 	struct syscall_params scp2;
 	struct ikc_scd_init_param iip2;
+	struct resource_set *resource_set;
 	
 	int status;
 	int fs;
@@ -67,6 +73,10 @@ struct cpu_local_var {
 
 	ihk_spinlock_t migq_lock;
 	struct list_head migq;
+	int in_interrupt;
+	int no_preempt;
+	int timer_enabled;
+	int kmalloc_initialized;
 } __attribute__((aligned(64)));
 
 

kernel/include/futex.h

@@ -99,6 +99,8 @@
 
 #ifdef __KERNEL__
 
+#define __user 
+
 /* We don't deal with uaccess at the moment, because x86 can access
  * userspace directly, we rely on glibc and the app developers.
  */
@@ -106,42 +108,14 @@
 #include <arch/uaccess.h>
 #endif
 
-#include <asm.h>
 #include <errno.h>
-
-#define __user 
+#include <arch-futex.h>
 
 #if 0
 #include <arch/processor.h>
 #include <arch/system.h>
 #endif
 
-#define __futex_atomic_op1(insn, ret, oldval, uaddr, oparg)	\
-	asm volatile("1:\t" insn "\n"				\
-		     "2:\t.section .fixup,\"ax\"\n"		\
-		     "3:\tmov\t%3, %1\n"			\
-		     "\tjmp\t2b\n"				\
-		     "\t.previous\n"				\
-		     _ASM_EXTABLE(1b, 3b)			\
-		     : "=r" (oldval), "=r" (ret), "+m" (*uaddr)	\
-		     : "i" (-EFAULT), "0" (oparg), "1" (0))
-
-#define __futex_atomic_op2(insn, ret, oldval, uaddr, oparg)	\
-	asm volatile("1:\tmovl	%2, %0\n"			\
-		     "\tmovl\t%0, %3\n"				\
-		     "\t" insn "\n"				\
-		     "2:\tlock; cmpxchgl %3, %2\n"		\
-		     "\tjnz\t1b\n"				\
-		     "3:\t.section .fixup,\"ax\"\n"		\
-		     "4:\tmov\t%5, %1\n"			\
-		     "\tjmp\t3b\n"				\
-		     "\t.previous\n"				\
-		     _ASM_EXTABLE(1b, 4b)			\
-		     _ASM_EXTABLE(2b, 4b)			\
-		     : "=&a" (oldval), "=&r" (ret),		\
-		       "+m" (*uaddr), "=&r" (tem)		\
-		     : "r" (oparg), "i" (-EFAULT), "1" (0))
-
 static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
@@ -206,28 +180,6 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	return ret;
 }
 
-static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
-						int newval)
-{
-#ifdef __UACCESS__
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
-		return -EFAULT;
-#endif
-
-	asm volatile("1:\tlock; cmpxchgl %3, %1\n"
-		     "2:\t.section .fixup, \"ax\"\n"
-		     "3:\tmov     %2, %0\n"
-		     "\tjmp     2b\n"
-		     "\t.previous\n"
-		     _ASM_EXTABLE(1b, 3b)
-		     : "=a" (oldval), "+m" (*uaddr)
-		     : "i" (-EFAULT), "r" (newval), "0" (oldval)
-		     : "memory"
-	);
-
-	return oldval;
-}
-
 #endif // __KERNEL__
 #endif // _ASM_X86_FUTEX_H
 
@@ -241,13 +193,11 @@ static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
 struct process_vm;
 
 union futex_key {
-#if 0
 	struct {
 		unsigned long pgoff;
-		struct inode *inode;
+		void *phys;
 		int offset;
 	} shared;
-#endif	
 	struct {
 		unsigned long address;
 		struct process_vm *mm;
@@ -261,6 +211,7 @@ union futex_key {
 };
 
 #define FUTEX_KEY_INIT (union futex_key) { .both = { .ptr = NULL } }
+#define FUT_OFF_MMSHARED 2
 
 extern int futex_init(void);
 
@@ -272,7 +223,8 @@ futex(
 	uint64_t				timeout,
 	uint32_t __user *		uaddr2,
 	uint32_t				val2,
-	uint32_t				val3
+	uint32_t				val3,
+	int                     fshared
 );
 
 

kernel/include/generic-rlimit.h

@@ -0,0 +1,23 @@
+/**
+ * \file rlimit.h
+ *  License details are found in the file LICENSE.
+ * \brief
+ *  Kinds of resource limit
+ * \author Taku Shimosawa  <shimosawa@is.s.u-tokyo.ac.jp> \par
+ * Copyright (C) 2011 - 2012  Taku Shimosawa
+ */
+/*
+ * HISTORY
+ */
+
+#ifndef __GENERIC_RLIMIT_H
+#define __GENERIC_RLIMIT_H
+
+typedef uint64_t rlim_t;
+
+struct rlimit {
+	rlim_t rlim_cur;  /* Soft limit */
+	rlim_t rlim_max;  /* Hard limit (ceiling for rlim_cur) */
+};
+
+#endif

kernel/include/init.h

@@ -14,7 +14,7 @@
 #define INIT_H
 
 extern void arch_init(void);
-extern void kmsg_init(void);
+extern void kmsg_init(int);
 extern void mem_init(void);
 extern void ikc_master_init(void);
 extern void ap_init(void);
@@ -28,6 +28,7 @@ extern void init_host_syscall_channel(void);
 extern void init_host_syscall_channel2(void);
 extern void sched_init(void);
 extern void pc_ap_init(void);
+extern void cpu_sysfs_setup(void);
 
 extern char *find_command_line(char *name);
 

kernel/include/kmalloc.h

@@ -14,19 +14,28 @@
 #define __HEADER_KMALLOC_H
 
 #include <ihk/mm.h>
+#include <cls.h>
 
-#define kmalloc(size, flag) _kmalloc(size, flag, __FILE__, __LINE__)
+void panic(const char *);
+int kprintf(const char *format, ...);
+
+#define kmalloc(size, flag) ({\
+void *r = _kmalloc(size, flag, __FILE__, __LINE__);\
+if(r == NULL){\
+kprintf("kmalloc: out of memory %s:%d no_preempt=%d\n", __FILE__, __LINE__, cpu_local_var(no_preempt)); \
+}\
+r;\
+})
 #define kfree(ptr) _kfree(ptr, __FILE__, __LINE__)
 #define memcheck(ptr, msg) _memcheck(ptr, msg, __FILE__, __LINE__, 0)
 void *_kmalloc(int size, enum ihk_mc_ap_flag flag, char *file, int line);
 void _kfree(void *ptr, char *file, int line);
 void *__kmalloc(int size, enum ihk_mc_ap_flag flag);
 void __kfree(void *ptr);
-void *___kmalloc(int size, enum ihk_mc_ap_flag flag);
-void ___kfree(void *ptr);
 
 int _memcheck(void *ptr, char *msg, char *file, int line, int free);
 int memcheckall();
 int freecheck(int runcount);
+void kmalloc_consolidate_free_list(void);
 
 #endif

kernel/include/kmsg.h

@@ -16,6 +16,6 @@
 void kputs(char *buf);
 int kprintf(const char *format, ...);
 
-void kmsg_init(void);
+void kmsg_init(int);
 
 #endif

kernel/include/lwk/futex.h

@@ -92,7 +92,8 @@ futex(
 	uint64_t			timeout,
 	uint32_t __user *		uaddr2,
 	uint32_t			val2,
-	uint32_t			val3
+	uint32_t			val3,
+	int                 fshared
 );
 
 extern long

kernel/include/memobj.h

@@ -31,6 +31,7 @@ enum {
 	/* for memobj.flags */
 	MF_HAS_PAGER	= 0x0001,
 	MF_SHMDT_OK	= 0x0002,
+	MF_IS_REMOVABLE	= 0x0004,
 };
 
 struct memobj {
@@ -46,6 +47,7 @@ typedef int memobj_get_page_func_t(struct memobj *obj, off_t off, int p2align, u
 typedef uintptr_t memobj_copy_page_func_t(struct memobj *obj, uintptr_t orgphys, int p2align);
 typedef int memobj_flush_page_func_t(struct memobj *obj, uintptr_t phys, size_t pgsize);
 typedef int memobj_invalidate_page_func_t(struct memobj *obj, uintptr_t phys, size_t pgsize);
+typedef int memobj_lookup_page_func_t(struct memobj *obj, off_t off, int p2align, uintptr_t *physp, unsigned long *flag);
 
 struct memobj_ops {
 	memobj_release_func_t *		release;
@@ -54,6 +56,7 @@ struct memobj_ops {
 	memobj_copy_page_func_t *	copy_page;
 	memobj_flush_page_func_t *	flush_page;
 	memobj_invalidate_page_func_t *	invalidate_page;
+	memobj_lookup_page_func_t *	lookup_page;
 };
 
 static inline void memobj_release(struct memobj *obj)
@@ -105,6 +108,15 @@ static inline int memobj_invalidate_page(struct memobj *obj, uintptr_t phys,
 	return 0;
 }
 
+static inline int memobj_lookup_page(struct memobj *obj, off_t off,
+		int p2align, uintptr_t *physp, unsigned long *pflag)
+{
+	if (obj->ops->lookup_page) {
+		return (*obj->ops->lookup_page)(obj, off, p2align, physp, pflag);
+	}
+	return -ENXIO;
+}
+
 static inline void memobj_lock(struct memobj *obj)
 {
 	ihk_mc_spinlock_lock_noirq(&obj->lock);
@@ -120,10 +132,16 @@ static inline int memobj_has_pager(struct memobj *obj)
 	return !!(obj->flags & MF_HAS_PAGER);
 }
 
+static inline int memobj_is_removable(struct memobj *obj)
+{
+	return !!(obj->flags & MF_IS_REMOVABLE);
+}
+
 int fileobj_create(int fd, struct memobj **objp, int *maxprotp);
 struct shmid_ds;
 int shmobj_create(struct shmid_ds *ds, struct memobj **objp);
 int zeroobj_create(struct memobj **objp);
-int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxprotp);
+int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxprotp,
+	int prot, int populate_flags);
 
 #endif /* HEADER_MEMOBJ_H */

kernel/include/page.h

@@ -17,8 +17,9 @@
 
 struct page {
 	struct list_head	list;
+	struct list_head	hash;
 	uint8_t			mode;
-	uint8_t			padding[3];
+	uint64_t		phys;
 	ihk_atomic_t		count;
 	off_t			offset;
 };
@@ -38,9 +39,8 @@ enum page_mode {
 struct page *phys_to_page(uintptr_t phys);
 uintptr_t page_to_phys(struct page *page);
 int page_unmap(struct page *page);
+struct page *phys_to_page_insert_hash(uint64_t phys);
 
-void *allocate_pages(int npages, enum ihk_mc_ap_flag flag);
-void free_pages(void *va, int npages);
 void begin_free_pages_pending(void);
 void finish_free_pages_pending(void);
 
@@ -66,4 +66,6 @@ static inline int page_is_multi_mapped(struct page *page)
 	return (ihk_atomic_read(&page->count) > 1);
 }
 
+/* Should we take page faults on ANONYMOUS mappings? */
+extern int anon_on_demand;
 #endif

kernel/include/process.h

@@ -22,6 +22,7 @@
 #include <memobj.h>
 #include <affinity.h>
 #include <syscall.h>
+#include <bitops.h>
 
 #define VR_NONE            0x0
 #define VR_STACK           0x1
@@ -29,6 +30,7 @@
 #define VR_IO_NOCACHE      0x100
 #define VR_REMOTE          0x200
 #define VR_WRITE_COMBINED  0x400
+#define VR_DONTFORK        0x800
 #define VR_DEMAND_PAGING   0x1000
 #define	VR_PRIVATE         0x2000
 #define	VR_LOCKED          0x4000
@@ -51,6 +53,7 @@
 #define	VRFLAG_PROT_TO_MAXPROT(vrflag)	(((vrflag) & VR_PROT_MASK) << 4)
 #define	VRFLAG_MAXPROT_TO_PROT(vrflag)	(((vrflag) & VR_MAXPROT_MASK) >> 4)
 
+// struct process.status, struct thread.status
 #define PS_RUNNING           0x1
 #define PS_INTERRUPTIBLE     0x2
 #define PS_UNINTERRUPTIBLE   0x4
@@ -58,15 +61,19 @@
 #define PS_EXITED            0x10
 #define PS_STOPPED           0x20
 #define PS_TRACED            0x40 /* Set to "not running" by a ptrace related event */
+#define PS_STOPPING          0x80
+#define PS_TRACING           0x100
 
 #define PS_NORMAL	(PS_INTERRUPTIBLE | PS_UNINTERRUPTIBLE)
 
+// struct process.ptrace
 #define PT_TRACED 0x80     /* The process is ptraced */
 #define PT_TRACE_EXEC 0x100 /* Trace execve(2) */
 #define PT_TRACE_SYSCALL_ENTER 0x200 /* Trace syscall enter */
 #define PT_TRACE_SYSCALL_EXIT  0x400 /* Trace syscall exit */
 #define PT_TRACE_SYSCALL_MASK  (PT_TRACE_SYSCALL_ENTER | PT_TRACE_SYSCALL_EXIT)
 
+// ptrace(2) request
 #define	PTRACE_TRACEME 0
 #define	PTRACE_PEEKTEXT 1
 #define	PTRACE_PEEKDATA 2
@@ -95,6 +102,7 @@
 #define	PTRACE_GETREGSET 0x4204
 #define	PTRACE_SETREGSET 0x4205
 
+// ptrace(2) options
 #define PTRACE_O_TRACESYSGOOD 1
 #define PTRACE_O_TRACEFORK 2
 #define PTRACE_O_TRACEVFORK 4
@@ -104,6 +112,7 @@
 #define PTRACE_O_TRACEEXIT 0x40
 #define PTRACE_O_MASK 0x7f
 
+// ptrace(2) events
 #define	PTRACE_EVENT_FORK 1
 #define	PTRACE_EVENT_VFORK 2
 #define	PTRACE_EVENT_CLONE 3
@@ -153,11 +162,136 @@
 #endif
 
 #define USER_STACK_NR_PAGES 8192
-#define KERNEL_STACK_NR_PAGES 25
+#define KERNEL_STACK_NR_PAGES 32
+
+#define NOPHYS ((uintptr_t)-1)
+
+#define PROCESS_NUMA_MASK_BITS 64
+
+/*
+ * 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>
 
+struct resource_set;
+struct process_hash;
+struct thread_hash;
+struct address_space;
+struct process;
+struct thread;
+struct process_vm;
+struct vm_regions;
+struct vm_range;
+
+#define HASH_SIZE	73
+
+struct resource_set {
+	struct list_head	list;
+	char			*path;
+	struct process_hash	*process_hash;
+	struct thread_hash	*thread_hash;
+	struct list_head	phys_mem_list;
+	mcs_rwlock_lock_t		phys_mem_lock;
+	cpu_set_t		cpu_set;
+	mcs_rwlock_lock_t		cpu_set_lock;
+	struct process		*pid1;
+};
+
+extern struct list_head	resource_set_list;
+extern mcs_rwlock_lock_t	resource_set_lock;
+
+struct process_hash {
+	struct list_head	list[HASH_SIZE];
+	mcs_rwlock_lock_t		lock[HASH_SIZE];
+};
+
+static inline int
+process_hash(int pid)
+{
+	return pid % HASH_SIZE;
+}
+
+static inline int
+thread_hash(int tid)
+{
+	return tid % HASH_SIZE;
+}
+
+struct thread_hash {
+	struct list_head	list[HASH_SIZE];
+	mcs_rwlock_lock_t		lock[HASH_SIZE];
+};
+
+struct address_space {
+	struct page_table	*page_table;
+	void			*opt;
+	void			(*free_cb)(struct address_space *, void *);
+	ihk_atomic_t		refcount;
+	cpu_set_t cpu_set;
+	ihk_spinlock_t cpu_set_lock;
+	int			nslots;
+	int			pids[];
+};
+
 struct user_fpregs_struct
 {
 	unsigned short cwd;
@@ -223,7 +357,7 @@ struct user
 	unsigned long int  u_debugreg [8];
 };
 
-#define	AUXV_LEN	16
+#define	AUXV_LEN	18
 
 struct vm_range {
 	struct list_head list;
@@ -231,9 +365,12 @@ struct vm_range {
 	unsigned long flag;
 	struct memobj *memobj;
 	off_t objoff;
+	int pgshift;	/* page size. 0 means THP */
+	int padding;
 };
 
 struct vm_regions {
+	unsigned long vm_start, vm_end;
 	unsigned long text_start, text_end;
 	unsigned long data_start, data_end;
 	unsigned long brk_start, brk_end;
@@ -244,19 +381,27 @@ struct vm_regions {
 
 struct process_vm;
 
-struct sigfd {
-	struct sigfd *next;
+struct mckfd {
+	struct mckfd *next;
 	int fd;
-	__sigset_t mask;
+	int sig_no;
+	long data;
+	void *opt;
+	long (*read_cb)(struct mckfd *, ihk_mc_user_context_t *);
+	int (*ioctl_cb)(struct mckfd *, ihk_mc_user_context_t *);
+	long (*mmap_cb)(struct mckfd *, ihk_mc_user_context_t *);
+	int (*close_cb)(struct mckfd *, ihk_mc_user_context_t *);
+	int (*fcntl_cb)(struct mckfd *, ihk_mc_user_context_t *);
 };
+
 #define SFD_CLOEXEC 02000000
 #define SFD_NONBLOCK 04000
 
-struct sig_handler {
+struct sig_common {
 	ihk_spinlock_t	lock;
-	ihk_atomic_t	use;
-	struct sigfd *sigfd;
+	ihk_atomic_t use;
 	struct k_sigaction action[_NSIG];
+	struct list_head sigpending;
 };
 
 struct sig_pending {
@@ -266,27 +411,68 @@ struct sig_pending {
 	int ptracecont;
 };
 
-struct sig_shared {
-	ihk_spinlock_t  lock;
-	ihk_atomic_t    use;
-	struct list_head sigpending;
-};
-
 typedef void pgio_func_t(void *arg);
 
+struct mcexec_tid {
+	int tid;
+	struct thread *thread;
+};
+
 /* Represents a node in the process fork tree, it may exist even after the 
  * corresponding process exited due to references from the parent and/or 
  * children and is used for implementing wait/waitpid without having a 
  * special "init" process */
-struct fork_tree_node {
-	ihk_spinlock_t lock;
-	ihk_atomic_t refcount;
-	int exit_status;
-	int status;
+struct process {
+	struct list_head hash_list;
+	mcs_rwlock_lock_t update_lock; // lock for parent, status, cpu time...
 
-	struct process *owner;
+	// process vm
+	struct process_vm *vm;
+
+	// threads and children
+	struct list_head threads_list;
+	mcs_rwlock_lock_t threads_lock; // lock for threads_list
+	/* TID set of proxy process */
+	struct mcexec_tid *tids;
+	int nr_tids;
+
+	/* The ptracing process behave as the parent of the ptraced process
+	   after using PTRACE_ATTACH except getppid. So we save it here. */
+	struct process *parent;
+	struct process *ppid_parent;
+	struct list_head children_list;
+	struct list_head ptraced_children_list;
+	mcs_rwlock_lock_t children_lock; // lock for children_list and ptraced_children_list
+	struct list_head siblings_list; // lock parent
+	struct list_head ptraced_siblings_list; // lock ppid_parent
+
+	ihk_atomic_t refcount;
+
+	// process status and exit status
+	int status;	// PS_RUNNING -> PS_EXITED -> PS_ZOMBIE
+			// |       ^       ^
+			// |       |---+   |
+			// V           |   |
+			// PS_STOPPING-)---+
+			// (PS_TRACING)|   |
+			// |           |   |
+			// V       +----   |
+			// PS_STOPPED -----+
+			// (PS_TRACED)
+	int exit_status;
+
+	/* 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
+	   might divert while the threads are exiting by group_exit(). */
+	int group_exit_status;
+
+	/* Manage ptraced processes in the separate list to make it easy to
+	   restore the orginal parent child relationship when 
+	   performing PTRACE_DETACH */
+	struct waitq waitpid_q;
+
+	// process info and credentials etc.
 	int pid;
-	int tid;
 	int pgid;
 	int ruid;
 	int euid;
@@ -296,50 +482,58 @@ struct fork_tree_node {
 	int egid;
 	int sgid;
 	int fsgid;
-	
-	struct fork_tree_node *parent;
-	struct list_head children;
-	struct list_head siblings_list;
-	
-    /* The ptracing process behave as the parent of the ptraced process
-       after using PTRACE_ATTACH except getppid. So we save it here. */
-	struct fork_tree_node *ppid_parent;
+	int execed;
+	int nohost;
+	int nowait;
+	struct rlimit rlimit[MCK_RLIM_MAX];
+	unsigned long saved_auxv[AUXV_LEN];
+	char *saved_cmdline;
+	long saved_cmdline_len;
 
-    /* Manage ptraced processes in the separate list to make it easy to
-       restore the orginal parent child relationship when 
-       performing PTRACE_DETACH */
-	struct list_head ptrace_children;
-	struct list_head ptrace_siblings_list;
+	/* Store ptrace flags.
+	 * The lower 8 bits are PTRACE_O_xxx of the PTRACE_SETOPTIONS request.
+	 * Other bits are for inner use of the McKernel.
+	 */
+	int ptrace;
 
-	struct waitq waitpid_q;
+	/* Store ptrace event message.
+	 * PTRACE_O_xxx will store event message here.
+	 * PTRACE_GETEVENTMSG will get from here.
+	 */
+	unsigned long ptrace_eventmsg;
 
-    /* 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
-       might divert while the threads are exiting by group_exit(). */
-    int group_exit_status;
+	/* Store event related to signal. For example, 
+	   it represents that the proceess has been resumed by SIGCONT. */
+	int signal_flags;
 
-    /* Store ptrace flags.
-     * The lower 8 bits are PTRACE_O_xxx of the PTRACE_SETOPTIONS request.
-     * Other bits are for inner use of the McKernel.
-     */
-    int ptrace;
+	/* Store signal sent to parent when the process terminates. */
+	int termsig;
 
-    /* Store ptrace event message.
-       PTRACE_O_xxx will store event message here.
-       PTRACE_GETEVENTMSG will get from here.
-     */
-    unsigned long ptrace_eventmsg;
+	ihk_spinlock_t mckfd_lock;
+	struct mckfd *mckfd;
 
-    /* Store event related to signal. For example, 
-       it represents that the proceess has been resumed by SIGCONT. */
-    int signal_flags;
+	// cpu time (summary)
+	struct timespec stime;
+	struct timespec utime;
 
-    /* Store signal sent to parent when the process terminates. */
-    int termsig;
+	// cpu time (children)
+	struct timespec stime_children;
+	struct timespec utime_children;
+
+	long maxrss;
+	long maxrss_children;
+
+	// perf_event
+	int perf_status;
+#define PP_NONE 0
+#define PP_RESET 1
+#define PP_COUNT 2
+#define PP_STOP 3
+	struct mc_perf_event *monitoring_event;
 };
 
-void hold_fork_tree_node(struct fork_tree_node *ftn);
-void release_fork_tree_node(struct fork_tree_node *ftn);
+void hold_thread(struct thread *ftn);
+void release_thread(struct thread *ftn);
 
 /*
  * Scheduling policies
@@ -364,100 +558,150 @@ struct sched_param {
 	int sched_priority;
 };
 
-struct process {
+struct thread {
+	struct list_head hash_list;
+	// thread info
 	int cpu_id;
+	int tid;
+	int status;	// PS_RUNNING -> PS_EXITED
+			// |       ^       ^
+			// |       |       |
+			// V       |       |
+			// PS_STOPPED------+
+			// PS_TRACED
+			// PS_INTERRPUTIBLE
+			// PS_UNINTERRUPTIBLE
 
-	ihk_atomic_t refcount;
+	// process vm
 	struct process_vm *vm;
 
+	// context
 	ihk_mc_kernel_context_t ctx;
 	ihk_mc_user_context_t  *uctx;
 	
+	// sibling
+	struct process *proc;
+	struct list_head siblings_list; // lock process
+
 	// Runqueue list entry
-	struct list_head sched_list;  
+	struct list_head sched_list;	// lock cls
 	int sched_policy;
 	struct sched_param sched_param;
 	
 	ihk_spinlock_t spin_sleep_lock;
 	int spin_sleep;
 
-	struct thread {
-		int	*clear_child_tid;
-		unsigned long tlsblock_base, tlsblock_limit;
-	} thread;
+	ihk_atomic_t refcount;
 
-	volatile int sigevent;
-	int nohost;
-	int execed;
+	int	*clear_child_tid;
+	unsigned long tlsblock_base, tlsblock_limit;
+
+	// thread info
+	cpu_set_t cpu_set;
+	fp_regs_struct *fp_regs;
+	int in_syscall_offload;
+
+	// signal
+	struct sig_common *sigcommon;
 	sigset_t sigmask;
 	stack_t sigstack;
-	ihk_spinlock_t	sigpendinglock;
 	struct list_head sigpending;
-	struct sig_shared *sigshared;
-	struct sig_handler *sighandler;
+	ihk_spinlock_t	sigpendinglock;
+	volatile int sigevent;
 
-	struct rlimit rlimit[MCK_RLIM_MAX];
+	// gpio
 	pgio_func_t *pgio_fp;
 	void *pgio_arg;
 
-	struct fork_tree_node *ftn;
-
-	cpu_set_t cpu_set;
-	unsigned long saved_auxv[AUXV_LEN];
-
+	// for ptrace
 	unsigned long *ptrace_debugreg;	/* debug registers for ptrace */
 	struct sig_pending *ptrace_recvsig;
 	struct sig_pending *ptrace_sendsig;
-	fp_regs_struct *fp_regs;
-	char *saved_cmdline;
-	long saved_cmdline_len;
+
+	// cpu time
+	struct timespec stime;
+	struct timespec utime;
+	struct timespec btime;
+	int times_update;
+	int in_kernel;
+
+	// interval timers
+	int itimer_enabled;
+	struct itimerval itimer_virtual;
+	struct itimerval itimer_prof;
+	struct timespec itimer_virtual_value;
+	struct timespec itimer_prof_value;
+
+	/* Syscall offload wait queue head */
+	struct waitq scd_wq;
 };
 
 struct process_vm {
-	ihk_atomic_t refcount;
-
-	struct page_table *page_table;
+	struct address_space *address_space;
 	struct list_head vm_range_list;
 	struct vm_regions region;
-	struct process *owner_process;		/* process that reside on the same page */
+	struct process *proc;		/* process that reside on the same page */
+	void *opt;
+	void (*free_cb)(struct process_vm *, void *);
+	void *vdso_addr;
+	void *vvar_addr;
  	
-    ihk_spinlock_t page_table_lock;
-    ihk_spinlock_t memory_range_lock;
+	ihk_spinlock_t page_table_lock;
+	ihk_spinlock_t memory_range_lock;
     // to protect the followings:
     // 1. addition of process "memory range" (extend_process_region, add_process_memory_range)
     // 2. addition of process page table (allocate_pages, update_process_page_table)
     // note that physical memory allocator (ihk_mc_alloc_pages, ihk_pagealloc_alloc)
     // is protected by its own lock (see ihk/manycore/generic/page_alloc.c)
 
-	cpu_set_t cpu_set;
-	ihk_spinlock_t cpu_set_lock;
+	ihk_atomic_t refcount;
 	int exiting;
+
+	long currss;
+	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
+	int numa_mem_policy;
 };
 
+static inline int has_cap_ipc_lock(struct thread *th)
+{
+	/* CAP_IPC_LOCK (= 14) */
+	return !(th->proc->euid);
+}
 
-struct process *create_process(unsigned long user_pc);
-struct process *clone_process(struct process *org, unsigned long pc,
+static inline int has_cap_sys_admin(struct thread *th)
+{
+	/* CAP_SYS_ADMIN (= 21) */
+	return !(th->proc->euid);
+}
+
+void hold_address_space(struct address_space *);
+void release_address_space(struct address_space *);
+struct thread *create_thread(unsigned long user_pc);
+struct thread *clone_thread(struct thread *org, unsigned long pc,
                               unsigned long sp, int clone_flags);
-void destroy_process(struct process *proc);
-void hold_process(struct process *proc);
-void release_process(struct process *proc);
-void flush_process_memory(struct process *proc);
-void free_process_memory(struct process *proc);
-void free_process_memory_ranges(struct process *proc);
-int populate_process_memory(struct process *proc, void *start, size_t len);
+void destroy_thread(struct thread *thread);
+void hold_thread(struct thread *thread);
+void release_thread(struct thread *thread);
+void flush_process_memory(struct process_vm *vm);
+void hold_process_vm(struct process_vm *vm);
+void release_process_vm(struct process_vm *vm);
+void hold_process(struct process *);
+void release_process(struct process *);
+void free_process_memory_ranges(struct process_vm *vm);
+int populate_process_memory(struct process_vm *vm, void *start, size_t len);
 
-int add_process_memory_range(struct process *process,
+int add_process_memory_range(struct process_vm *vm,
                              unsigned long start, unsigned long end,
                              unsigned long phys, unsigned long flag,
-			     struct memobj *memobj, off_t objoff);
-int remove_process_memory_range(struct process *process, unsigned long start,
+			     struct memobj *memobj, off_t objoff, int pgshift);
+int remove_process_memory_range(struct process_vm *vm, unsigned long start,
 		unsigned long end, int *ro_freedp);
-int split_process_memory_range(struct process *process,
+int split_process_memory_range(struct process_vm *vm,
 		struct vm_range *range, uintptr_t addr, struct vm_range **splitp);
-int join_process_memory_range(struct process *process, struct vm_range *surviving,
+int join_process_memory_range(struct process_vm *vm, struct vm_range *surviving,
 		struct vm_range *merging);
 int change_prot_process_memory_range(
-		struct process *process, struct vm_range *range,
+		struct process_vm *vm, struct vm_range *range,
 		unsigned long newflag);
 int remap_process_memory_range(struct process_vm *vm, struct vm_range *range,
 		uintptr_t start, uintptr_t end, off_t off);
@@ -476,31 +720,41 @@ int extend_up_process_memory_range(struct process_vm *vm,
 
 int page_fault_process_vm(struct process_vm *fault_vm, void *fault_addr,
 		uint64_t reason);
-int remove_process_region(struct process *proc,
+int remove_process_region(struct process_vm *vm,
                           unsigned long start, unsigned long end);
 struct program_load_desc;
-int init_process_stack(struct process *process, struct program_load_desc *pn,
+int init_process_stack(struct thread *thread, struct program_load_desc *pn,
                         int argc, char **argv, 
                         int envc, char **env);
-unsigned long extend_process_region(struct process *proc,
+unsigned long extend_process_region(struct process_vm *vm,
                                     unsigned long start, unsigned long end,
                                     unsigned long address, unsigned long flag);
 extern enum ihk_mc_pt_attribute arch_vrflag_to_ptattr(unsigned long flag, uint64_t fault, pte_t *ptep);
 enum ihk_mc_pt_attribute common_vrflag_to_ptattr(unsigned long flag, uint64_t fault, pte_t *ptep);
 
 void schedule(void);
-void runq_add_proc(struct process *proc, int cpu_id);
-void runq_del_proc(struct process *proc, int cpu_id);
-int sched_wakeup_process(struct process *proc, int valid_states);
+void runq_add_thread(struct thread *thread, int cpu_id);
+void runq_del_thread(struct thread *thread, int cpu_id);
+int sched_wakeup_thread(struct thread *thread, int valid_states);
 
-void sched_request_migrate(int cpu_id, struct process *proc);
+void sched_request_migrate(int cpu_id, struct thread *thread);
 void check_need_resched(void);
 
 void cpu_set(int cpu, cpu_set_t *cpu_set, ihk_spinlock_t *lock);
 void cpu_clear(int cpu, cpu_set_t *cpu_set, ihk_spinlock_t *lock);
+void cpu_clear_and_set(int c_cpu, int s_cpu, 
+		cpu_set_t *cpu_set, ihk_spinlock_t *lock);
 
-struct process *findthread_and_lock(int pid, int tid, ihk_spinlock_t **savelock, unsigned long *irqstate);
-void process_unlock(void *savelock, unsigned long irqstate);
 void release_cpuid(int cpuid);
 
+struct thread *find_thread(int pid, int tid, struct mcs_rwlock_node_irqsave *lock);
+void thread_unlock(struct thread *thread, struct mcs_rwlock_node_irqsave *lock);
+struct process *find_process(int pid, struct mcs_rwlock_node_irqsave *lock);
+void process_unlock(struct process *proc, struct mcs_rwlock_node_irqsave *lock);
+void chain_process(struct process *);
+void chain_thread(struct thread *);
+void proc_init();
+void set_timer();
+struct sig_pending *hassigpending(struct thread *thread);
+
 #endif

kernel/include/shm.h

@@ -25,6 +25,7 @@ enum {
 	IPC_CREAT	= 01000,
 	IPC_EXCL	= 02000,
 
+	SHM_HUGETLB	= 04000,
 	SHM_RDONLY	= 010000,
 	SHM_RND		= 020000,
 	SHM_REMAP	= 040000,
@@ -46,11 +47,14 @@ enum {
 	SHM_INFO	= 14,
 };
 
+struct shmlock_user;
+
 struct shmobj {
 	struct memobj		memobj;		/* must be first */
 	int			index;
-	uint8_t			padding[4];
+	int			pgshift;
 	size_t			real_segsz;
+	struct shmlock_user *	user;
 	struct shmid_ds		ds;
 	struct list_head	page_list;
 	struct list_head	chain;		/* shmobj_list */
@@ -75,9 +79,33 @@ struct shm_info {
 	uint64_t	swap_successes;
 };
 
+struct shmlock_user {
+	uid_t ruid;
+	int padding;
+	size_t locked;
+
+	struct list_head chain;
+};
+
+extern ihk_spinlock_t shmlock_users_lock_body;
+
+static inline void shmlock_users_lock(void)
+{
+	ihk_mc_spinlock_lock_noirq(&shmlock_users_lock_body);
+	return;
+}
+
+static inline void shmlock_users_unlock(void)
+{
+	ihk_mc_spinlock_unlock_noirq(&shmlock_users_lock_body);
+	return;
+}
+
 void shmobj_list_lock(void);
 void shmobj_list_unlock(void);
 int shmobj_create_indexed(struct shmid_ds *ds, struct shmobj **objp);
 void shmobj_destroy(struct shmobj *obj);
+void shmlock_user_free(struct shmlock_user *user);
+int shmlock_user_get(uid_t ruid, struct shmlock_user **userp);
 
 #endif /* HEADER_SHM_H */