configure.ac: Update version number to 1.5.1-knl+hfi

Change-Id: Icbd08c9c5f65b22d007ec479a34acd20062e0e90

.gitignore

@@ -1,4 +1,3 @@
-*~
 *.o
 *.elf
 *.bin
@@ -15,3 +14,4 @@ elfboot/elfboot_test
 linux/executer/mcexec
 linux/mod_test*
 linux/target
+kernel/script/dwarf-extract-struct

.gitmodules

@@ -1,3 +0,0 @@
-[submodule "ihk"]
-	path = ihk
-	url = https://github.com/RIKEN-SysSoft/ihk.git

Makefile.in

@@ -1,5 +1,6 @@
 TARGET = @TARGET@
 SBINDIR = @SBINDIR@
+BINDIR = @BINDIR@
 INCDIR =  @INCDIR@
 ETCDIR = @ETCDIR@
 MANDIR = @MANDIR@
@@ -47,6 +48,7 @@ install:
 		mkdir -p -m 755 $(SBINDIR); \
 		install -m 755 arch/x86_64/tools/mcreboot-smp-x86.sh $(SBINDIR)/mcreboot.sh; \
 		install -m 755 arch/x86_64/tools/mcstop+release-smp-x86.sh $(SBINDIR)/mcstop+release.sh; \
+		install -m 755 arch/x86_64/tools/mpimcexec $(BINDIR)/mpimcexec; \
 		install -m 755 arch/x86_64/tools/mcoverlay-destroy-smp-x86.sh $(SBINDIR)/mcoverlay-destroy.sh; \
 		install -m 755 arch/x86_64/tools/mcoverlay-create-smp-x86.sh $(SBINDIR)/mcoverlay-create.sh; \
 		install -m 755 arch/x86_64/tools/eclair-dump-backtrace.exp $(SBINDIR)/eclair-dump-backtrace.exp;\
@@ -57,6 +59,7 @@ install:
 		install -m 644 kernel/include/swapfmt.h $(INCDIR); \
 		mkdir -p -m 755 $(MANDIR)/man1; \
 		install -m 644 arch/x86_64/tools/mcreboot.1 $(MANDIR)/man1/mcreboot.1; \
+		install -m 644 arch/x86_64/tools/mpimcexec.1 $(MANDIR)/man1/mpimcexec.1; \
 		;; \
 	    *) \
 		echo "unknown target $(TARGET)" >&2 \

arch/arm64/kernel/cpu.c

@@ -30,7 +30,6 @@
 #include <debug-monitors.h>
 #include <sysreg.h>
 #include <cpufeature.h>
-#include <debug.h>
 #ifdef POSTK_DEBUG_ARCH_DEP_65
 #include <hwcap.h>
 #endif /* POSTK_DEBUG_ARCH_DEP_65 */
@@ -40,10 +39,16 @@
 #include "postk_print_sysreg.c"
 
 #ifdef DEBUG_PRINT_CPU
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf kprintf
+#define ekprintf kprintf
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf kprintf
 #endif
 
+#define BUG_ON(condition) do { if (condition) { kprintf("PANIC: %s: %s(line:%d)\n",\
+				__FILE__, __FUNCTION__, __LINE__); panic(""); } } while(0)
+
 struct cpuinfo_arm64 cpuinfo_data[NR_CPUS];	/* index is logical cpuid */
 static unsigned int per_cpu_timer_val[NR_CPUS] = { 0 };
 
@@ -1278,6 +1283,7 @@ int ihk_mc_interrupt_cpu(int cpu, int vector)
 	return 0;
 }
 
+#ifdef POSTK_DEBUG_ARCH_DEP_22
 /*
  * @ref.impl linux-linaro/arch/arm64/kernel/process.c::tls_thread_switch()
  */
@@ -1303,13 +1309,14 @@ struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 	extern void perf_start(struct mc_perf_event *event);
 	extern void perf_reset(struct mc_perf_event *event);
 	struct thread *last;
+#ifdef POSTK_DEBUG_TEMP_FIX_41 /* early to wait4() wakeup for ptrace, fix. */
 	struct mcs_rwlock_node_irqsave lock;
+#endif /* POSTK_DEBUG_TEMP_FIX_41 */
 
 	/* Set up new TLS.. */
 	dkprintf("[%d] arch_switch_context: tlsblock_base: 0x%lX\n", 
 		 ihk_mc_get_processor_id(), next->tlsblock_base);
 
-#ifdef ENABLE_PERF
 	/* Performance monitoring inherit */
 	if(next->proc->monitoring_event) {
 		if(next->proc->perf_status == PP_RESET)
@@ -1319,10 +1326,10 @@ struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 			perf_start(next->proc->monitoring_event);
 		}
 	}
-#endif /*ENABLE_PERF*/
 	if (likely(prev)) {
 		tls_thread_switch(prev, next);
 
+#ifdef POSTK_DEBUG_TEMP_FIX_41 /* early to wait4() wakeup for ptrace, fix. */
 		mcs_rwlock_writer_lock(&prev->proc->update_lock, &lock);
 		if (prev->proc->status & (PS_DELAY_STOPPED | PS_DELAY_TRACED)) {
 			switch (prev->proc->status) {
@@ -1336,12 +1343,11 @@ struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 				break;
 			}
 			mcs_rwlock_writer_unlock(&prev->proc->update_lock, &lock);
-
-			/* Wake up the parent who tried wait4 and sleeping */
 			waitq_wakeup(&prev->proc->parent->waitpid_q);
 		} else {
 			mcs_rwlock_writer_unlock(&prev->proc->update_lock, &lock);
 		}
+#endif /* POSTK_DEBUG_TEMP_FIX_41 */
 
 		last = ihk_mc_switch_context(&prev->ctx, &next->ctx, prev);
 	}
@@ -1351,6 +1357,7 @@ struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 
 	return last;
 }
+#endif /* POSTK_DEBUG_ARCH_DEP_22 */
 
 /*@
   @ requires \valid(thread);
@@ -1432,7 +1439,8 @@ void copy_fp_regs(struct thread *from, struct thread *to)
 	}
 }
 
-void clear_fp_regs(void)
+void
+clear_fp_regs(struct thread *thread)
 {
 	if (likely(elf_hwcap & (HWCAP_FP | HWCAP_ASIMD))) {
 #ifdef CONFIG_ARM64_SVE
@@ -1469,7 +1477,7 @@ restore_fp_regs(struct thread *thread)
 	if (likely(elf_hwcap & (HWCAP_FP | HWCAP_ASIMD))) {
 		if (!thread->fp_regs) {
 			// only clear fpregs.
-			clear_fp_regs();
+			clear_fp_regs(thread);
 			return;
 		}
 		thread_fpsimd_load(thread);

arch/arm64/kernel/fpsimd.c

@@ -9,16 +9,20 @@
 #include <prctl.h>
 #include <cpufeature.h>
 #include <kmalloc.h>
-#include <debug.h>
-#include <process.h>
 
 //#define DEBUG_PRINT_FPSIMD
 
 #ifdef DEBUG_PRINT_FPSIMD
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf kprintf
+#define ekprintf kprintf
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf kprintf
 #endif
 
+#define BUG_ON(condition) do { if (condition) { kprintf("PANIC: %s: %s(line:%d)\n",\
+				__FILE__, __FUNCTION__, __LINE__); panic(""); } } while(0)
+
 #ifdef CONFIG_ARM64_SVE
 
 /* Maximum supported vector length across all CPUs (initially poisoned) */
@@ -69,6 +73,9 @@ static int get_nr_threads(struct process *proc)
 	return nr_threads;
 }
 
+extern void save_fp_regs(struct thread *thread);
+extern void clear_fp_regs(struct thread *thread);
+extern void restore_fp_regs(struct thread *thread);
 /* @ref.impl arch/arm64/kernel/fpsimd.c::sve_set_vector_length */
 int sve_set_vector_length(struct thread *thread,
 			unsigned long vl, unsigned long flags)
@@ -122,7 +129,7 @@ int sve_set_vector_length(struct thread *thread,
 			/* for self at prctl syscall */
 			if (thread == cpu_local_var(current)) {
 				save_fp_regs(thread);
-				clear_fp_regs();
+				clear_fp_regs(thread);
 				thread_sve_to_fpsimd(thread, &fp_regs);
 				sve_free(thread);
 

arch/arm64/kernel/gencore.c

@@ -7,7 +7,6 @@
 #include <process.h>
 #include <string.h>
 #include <elfcore.h>
-#include <debug.h>
 
 #define	align32(x) ((((x) + 3) / 4) * 4)
 #define	alignpage(x) ((((x) + (PAGE_SIZE) - 1) / (PAGE_SIZE)) * (PAGE_SIZE))
@@ -15,8 +14,11 @@
 //#define DEBUG_PRINT_GENCORE
 
 #ifdef DEBUG_PRINT_GENCORE
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...)	kprintf(__VA_ARGS__)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...)	do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
 #endif
 
 /*

arch/arm64/kernel/include/arch-lock.h

@@ -6,8 +6,6 @@
 
 #include <ihk/cpu.h>
 #include <ihk/atomic.h>
-#include "affinity.h"
-#include <lwk/compiler.h>
 
 //#define DEBUG_SPINLOCK
 //#define DEBUG_MCS_RWLOCK
@@ -154,8 +152,6 @@ typedef struct mcs_lock_node {
 	unsigned long irqsave;
 } __attribute__((aligned(64))) mcs_lock_node_t;
 
-typedef mcs_lock_node_t mcs_lock_t;
-
 static void mcs_lock_init(struct mcs_lock_node *node)
 {
 	node->locked = 0;
@@ -606,16 +602,4 @@ __mcs_rwlock_reader_unlock(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node_
 #endif
 }
 
-static inline int irqflags_can_interrupt(unsigned long flags)
-{
-#ifdef CONFIG_HAS_NMI
-#warning irqflags_can_interrupt needs testing/fixing on such a target
-	return flags > ICC_PMR_EL1_MASKED;
-#else
-	// PSTATE.DAIF I bit clear means interrupt is possible
-	return !(flags & (1 << 7));
-#endif
-}
-
-
 #endif /* !__HEADER_ARM64_COMMON_ARCH_LOCK_H */

arch/arm64/kernel/include/arch-perfctr.h

@@ -35,4 +35,38 @@ void arm64_disable_pmu(void);
 int armv8pmu_init(struct arm_pmu* cpu_pmu);
 
 /* TODO[PMU]: 共通部に定義があっても良い。今後の動向を見てここの定義を削除する */
+/*
+ * Generalized hardware cache events:
+ *
+ *       { L1-D, L1-I, LLC, ITLB, DTLB, BPU, NODE } x
+ *       { read, write, prefetch } x
+ *       { accesses, misses }
+ */
+enum perf_hw_cache_id {
+	PERF_COUNT_HW_CACHE_L1D			= 0,
+	PERF_COUNT_HW_CACHE_L1I			= 1,
+	PERF_COUNT_HW_CACHE_LL			= 2,
+	PERF_COUNT_HW_CACHE_DTLB		= 3,
+	PERF_COUNT_HW_CACHE_ITLB		= 4,
+	PERF_COUNT_HW_CACHE_BPU			= 5,
+	PERF_COUNT_HW_CACHE_NODE		= 6,
+
+	PERF_COUNT_HW_CACHE_MAX,		/* non-ABI */
+};
+
+enum perf_hw_cache_op_id {
+	PERF_COUNT_HW_CACHE_OP_READ		= 0,
+	PERF_COUNT_HW_CACHE_OP_WRITE		= 1,
+	PERF_COUNT_HW_CACHE_OP_PREFETCH		= 2,
+
+	PERF_COUNT_HW_CACHE_OP_MAX,		/* non-ABI */
+};
+
+enum perf_hw_cache_op_result_id {
+	PERF_COUNT_HW_CACHE_RESULT_ACCESS	= 0,
+	PERF_COUNT_HW_CACHE_RESULT_MISS		= 1,
+
+	PERF_COUNT_HW_CACHE_RESULT_MAX,		/* non-ABI */
+};
+
 #endif

arch/arm64/kernel/include/signal.h

@@ -9,11 +9,6 @@
 #define _NSIG_BPW	64
 #define _NSIG_WORDS (_NSIG / _NSIG_BPW)
 
-static inline int valid_signal(unsigned long sig)
-{
-	return sig <= _NSIG ? 1 : 0;
-}
-
 typedef unsigned long int __sigset_t;
 #define __sigmask(sig)  (((__sigset_t) 1) << ((sig) - 1))
 

arch/arm64/kernel/include/syscall_list.h

@@ -114,18 +114,14 @@ SYSCALL_HANDLED(236, get_mempolicy)
 SYSCALL_HANDLED(237, set_mempolicy)
 SYSCALL_HANDLED(238, migrate_pages)
 SYSCALL_HANDLED(239, move_pages)
-#ifdef PERF_ENABLE
 SYSCALL_HANDLED(241, perf_event_open)
-#endif // PERF_ENABLE
 SYSCALL_HANDLED(260, wait4)
 SYSCALL_HANDLED(270, process_vm_readv)
 SYSCALL_HANDLED(271, process_vm_writev)
-#ifdef PERF_ENABLE
 SYSCALL_HANDLED(601, pmc_init)
 SYSCALL_HANDLED(602, pmc_start)
 SYSCALL_HANDLED(603, pmc_stop)
 SYSCALL_HANDLED(604, pmc_reset)
-#endif // PERF_ENABLE
 SYSCALL_HANDLED(700, get_cpu_id)
 #ifdef PROFILE_ENABLE
 SYSCALL_HANDLED(__NR_profile, profile)

arch/arm64/kernel/irq-gic-v2.c

@@ -7,13 +7,15 @@
 #include <arch/cpu.h>
 #include <memory.h>
 #include <syscall.h>
-#include <debug.h>
 
 // #define DEBUG_GICV2
 
 #ifdef DEBUG_GICV2
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...)	kprintf(__VA_ARGS__)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
 #endif
 
 void *dist_base;

arch/arm64/kernel/irq-gic-v3.c

@@ -7,15 +7,17 @@
 #include <cputype.h>
 #include <process.h>
 #include <syscall.h>
-#include <debug.h>
 
 //#define DEBUG_GICV3
 
 #define USE_CAVIUM_THUNDER_X
 
 #ifdef DEBUG_GICV3
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...)	kprintf(__VA_ARGS__)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
 #endif
 
 #ifdef USE_CAVIUM_THUNDER_X

arch/arm64/kernel/memory.c

@@ -14,7 +14,9 @@
 #include <context.h>
 #include <kmalloc.h>
 #include <vdso.h>
-#include <debug.h>
+
+#define	dkprintf(...)	do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
 
 #define NOT_IMPLEMENTED()  do { kprintf("%s is not implemented\n", __func__); while(1);} while(0)
 
@@ -2922,12 +2924,17 @@ int read_process_vm(struct process_vm *vm, void *kdst, const void *usrc, size_t
 			return error;
 		}
 
-		if (!is_mckernel_memory(pa, pa + cpsize)) {
+#ifdef POSTK_DEBUG_TEMP_FIX_52 /* NUMA support(memory area determination) */
+		if (!is_mckernel_memory(pa)) {
+#else
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+#endif /* POSTK_DEBUG_TEMP_FIX_52 */
 			dkprintf("%s: pa is outside of LWK memory, to: %p, pa: %p,"
 				"cpsize: %d\n", __FUNCTION__, to, pa, cpsize);
 			va = ihk_mc_map_virtual(pa, 1, PTATTR_ACTIVE);
 			memcpy(to, va, cpsize);
-			ihk_mc_unmap_virtual(va, 1);
+			ihk_mc_unmap_virtual(va, 1, 1);
 		}
 		else {
 			va = phys_to_virt(pa);
@@ -3000,12 +3007,17 @@ int write_process_vm(struct process_vm *vm, void *udst, const void *ksrc, size_t
 			return error;
 		}
 
-		if (!is_mckernel_memory(pa, pa + cpsize)) {
+#ifdef POSTK_DEBUG_TEMP_FIX_52 /* NUMA support(memory area determination) */
+		if (!is_mckernel_memory(pa)) {
+#else
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+#endif /* POSTK_DEBUG_TEMP_FIX_52 */
 			dkprintf("%s: pa is outside of LWK memory, from: %p,"
 				"pa: %p, cpsize: %d\n", __FUNCTION__, from, pa, cpsize);
 			va = ihk_mc_map_virtual(pa, 1, PTATTR_WRITABLE|PTATTR_ACTIVE);
 			memcpy(va, from, cpsize);
-			ihk_mc_unmap_virtual(va, 1);
+			ihk_mc_unmap_virtual(va, 1, 1);
 		}
 		else {
 			va = phys_to_virt(pa);
@@ -3066,12 +3078,17 @@ int patch_process_vm(struct process_vm *vm, void *udst, const void *ksrc, size_t
 			return error;
 		}
 
-		if (!is_mckernel_memory(pa, pa + cpsize)) {
+#ifdef POSTK_DEBUG_TEMP_FIX_52 /* NUMA support(memory area determination) */
+		if (!is_mckernel_memory(pa)) {
+#else
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+#endif /* POSTK_DEBUG_TEMP_FIX_52 */
 			dkprintf("%s: pa is outside of LWK memory, from: %p,"
 				"pa: %p, cpsize: %d\n", __FUNCTION__, from, pa, cpsize);
 			va = ihk_mc_map_virtual(pa, 1, PTATTR_WRITABLE|PTATTR_ACTIVE);
 			memcpy(va, from, cpsize);
-			ihk_mc_unmap_virtual(va, 1);
+			ihk_mc_unmap_virtual(va, 1, 1);
 		}
 		else {
 			va = phys_to_virt(pa);

arch/arm64/kernel/perfctr.c

@@ -93,50 +93,21 @@ int ihk_mc_perfctr_init(int counter, uint64_t config, int mode)
 	return ret;
 }
 
-int ihk_mc_perfctr_start(unsigned long counter_mask)
+int ihk_mc_perfctr_start(int counter)
 {
-	int ret = 0;
-	int counter;
-	unsigned long counter_bit;
-
-	for (counter = 0, counter_bit = 1;
-	     counter_bit < counter_mask;
-	     counter++, counter_bit <<= 1) {
-		if (!(counter_mask & counter_bit))
-			continue;
-
-		ret = cpu_pmu.enable_counter(counter_mask);
-		if (ret < 0)
-			break;
-	}
-
-	return ret < 0 ? ret : 0;
+	int ret;
+	ret = cpu_pmu.enable_counter(counter);
+	return ret;
 }
 
-int ihk_mc_perfctr_stop(unsigned long counter_mask)
+int ihk_mc_perfctr_stop(int counter)
 {
-	int ret = 0;
-	int counter;
-	unsigned long counter_bit;
+	cpu_pmu.disable_counter(counter);
 
-	for (counter = 0, counter_bit = 1;
-	     counter_bit < counter_mask;
-	     counter++, counter_bit <<= 1) {
-		if (!(counter_mask & counter_bit))
-			continue;
-
-		ret = cpu_pmu.disable_counter(counter);
-		if (ret < 0)
-			break;
-
-		// ihk_mc_perfctr_startが呼ばれるときには、
-		// init系関数が呼ばれるのでdisableにする。
-		ret = cpu_pmu.disable_intens(counter);
-		if (ret < 0)
-			break;
-	}
-
-	return ret < 0 ? ret : 0;
+	// ihk_mc_perfctr_startが呼ばれるときには、
+	// init系関数が呼ばれるのでdisableにする。
+	cpu_pmu.disable_intens(counter);
+	return 0;
 }
 
 int ihk_mc_perfctr_reset(int counter)

arch/arm64/kernel/perfctr_armv8pmu.c

@@ -4,14 +4,16 @@
 #include <ihk/perfctr.h>
 #include <errno.h>
 #include <ihk/debug.h>
-#include <debug.h>
 
 #define BIT(nr) (1UL << (nr))
 
 //#define DEBUG_PRINT_PMU
 #ifdef DEBUG_PRINT_PMU
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...) kprintf(__VA_ARGS__)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
 #endif
 
 

arch/arm64/kernel/psci.c

@@ -21,13 +21,15 @@
 #include <ihk/debug.h>
 #include <compiler.h>
 #include <lwk/compiler.h>
-#include <debug.h>
 
 //#define DEBUG_PRINT_PSCI
 
 #ifdef DEBUG_PRINT_PSCI
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf(...) kprintf(__VA_ARGS__)
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) kprintf(__VA_ARGS__)
 #endif
 
 #define PSCI_POWER_STATE_TYPE_POWER_DOWN	1

arch/arm64/kernel/ptrace.c

@@ -11,17 +11,22 @@
 #include <hwcap.h>
 #include <string.h>
 #include <thread_info.h>
-#include <debug.h>
 
 //#define DEBUG_PRINT_SC
 
 #ifdef DEBUG_PRINT_SC
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf kprintf
+#define ekprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
 #endif
 
 #define NOT_IMPLEMENTED()  do { kprintf("%s is not implemented\n", __func__); while(1);} while(0)
 
+#define BUG_ON(condition) do { if (condition) { kprintf("PANIC: %s: %s(line:%d)\n",\
+				__FILE__, __FUNCTION__, __LINE__); panic(""); } } while(0)
+
 extern void save_debugreg(unsigned long *debugreg);
 extern unsigned long do_kill(struct thread *thread, int pid, int tid, int sig, struct siginfo *info, int ptracecont);
 extern int interrupt_from_user(void *);
@@ -954,7 +959,11 @@ void ptrace_report_signal(struct thread *thread, int sig)
 	}
 	thread->exit_status = sig;
 	/* Transition thread state */
+#ifdef POSTK_DEBUG_TEMP_FIX_41 /* early to wait4() wakeup for ptrace, fix. */
 	proc->status = PS_DELAY_TRACED;
+#else /* POSTK_DEBUG_TEMP_FIX_41 */
+	proc->status = PS_TRACED;
+#endif /* POSTK_DEBUG_TEMP_FIX_41 */
 	thread->status = PS_TRACED;
 	proc->ptrace &= ~PT_TRACE_SYSCALL;
 	if (sig == SIGSTOP || sig == SIGTSTP ||
@@ -973,6 +982,10 @@ void ptrace_report_signal(struct thread *thread, int sig)
 	info._sifields._sigchld.si_pid = thread->tid;
 	info._sifields._sigchld.si_status = thread->exit_status;
 	do_kill(cpu_local_var(current), parent_pid, -1, SIGCHLD, &info, 0);
+#ifndef POSTK_DEBUG_TEMP_FIX_41 /* early to wait4() wakeup for ptrace, fix. */
+	/* Wake parent (if sleeping in wait4()) */
+	waitq_wakeup(&proc->parent->waitpid_q);
+#endif /* !POSTK_DEBUG_TEMP_FIX_41 */
 
 	dkprintf("ptrace_report_signal,sleeping\n");
 	/* Sleep */

arch/arm64/kernel/syscall.c

@@ -14,8 +14,6 @@
 #include <prctl.h>
 #include <limits.h>
 #include <syscall.h>
-#include <uio.h>
-#include <debug.h>
 
 extern void ptrace_report_signal(struct thread *thread, int sig);
 extern void clear_single_step(struct thread *thread);
@@ -29,12 +27,18 @@ static void __check_signal(unsigned long rc, void *regs, int num, int irq_disabl
 //#define DEBUG_PRINT_SC
 
 #ifdef DEBUG_PRINT_SC
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf kprintf
+#define ekprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
 #endif
 
 #define NOT_IMPLEMENTED()  do { kprintf("%s is not implemented\n", __func__); while(1);} while(0)
 
+#define BUG_ON(condition) do { if (condition) { kprintf("PANIC: %s: %s(line:%d)\n",\
+				__FILE__, __FUNCTION__, __LINE__); panic(""); } } while(0)
+
 uintptr_t debug_constants[] = {
 	sizeof(struct cpu_local_var),
 	offsetof(struct cpu_local_var, current),
@@ -55,7 +59,7 @@ static int cpuid_head = 1;
 
 extern int num_processors;
 
-int obtain_clone_cpuid(cpu_set_t *cpu_set, int use_last) {
+int obtain_clone_cpuid(cpu_set_t *cpu_set) {
 	int min_queue_len = -1;
 	int i, min_cpu = -1;
 
@@ -1173,10 +1177,19 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 				/* Reap and set new signal_flags */
 				proc->signal_flags = SIGNAL_STOP_STOPPED;
 
+#ifdef POSTK_DEBUG_TEMP_FIX_41 /* early to wait4() wakeup for ptrace, fix. */
 				proc->status = PS_DELAY_STOPPED;
+#else /* POSTK_DEBUG_TEMP_FIX_41 */
+				proc->status = PS_STOPPED;
+#endif /* POSTK_DEBUG_TEMP_FIX_41 */
 				thread->status = PS_STOPPED;
 				mcs_rwlock_writer_unlock(&proc->update_lock, &lock);	
 
+#ifndef POSTK_DEBUG_TEMP_FIX_41 /* early to wait4() wakeup for ptrace, fix. */
+				/* Wake up the parent who tried wait4 and sleeping */
+				waitq_wakeup(&proc->parent->waitpid_q);
+#endif /* !POSTK_DEBUG_TEMP_FIX_41 */
+
 				dkprintf("do_signal(): pid: %d, tid: %d SIGSTOP, sleeping\n", 
 					proc->pid, thread->tid);
 				/* Sleep */
@@ -1193,10 +1206,19 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 			/* Update thread state in fork tree */
 			mcs_rwlock_writer_lock(&proc->update_lock, &lock);	
 			thread->exit_status = SIGTRAP;
+#ifdef POSTK_DEBUG_TEMP_FIX_41 /* early to wait4() wakeup for ptrace, fix. */
 			proc->status = PS_DELAY_TRACED;
+#else /* POSTK_DEBUG_TEMP_FIX_41 */
+			proc->status = PS_TRACED;
+#endif /* POSTK_DEBUG_TEMP_FIX_41 */
 			thread->status = PS_TRACED;
 			mcs_rwlock_writer_unlock(&proc->update_lock, &lock);	
 
+#ifndef POSTK_DEBUG_TEMP_FIX_41 /* early to wait4() wakeup for ptrace, fix. */
+			/* Wake up the parent who tried wait4 and sleeping */
+			waitq_wakeup(&thread->proc->parent->waitpid_q);
+#endif /* !POSTK_DEBUG_TEMP_FIX_41 */
+
 			/* Sleep */
 			dkprintf("do_signal,SIGTRAP,sleeping\n");
 
@@ -1572,7 +1594,7 @@ done:
 		return 0;
 	}
 
-	if (tthread->uti_state == UTI_STATE_RUNNING_IN_LINUX) {
+	if (tthread->thread_offloaded) {
 		interrupt_syscall(tthread, sig);
 		release_thread(tthread);
 		return 0;
@@ -1707,7 +1729,7 @@ SYSCALL_DECLARE(mmap)
 		| MAP_NONBLOCK		// 0x10000
 		;
 
-	const uintptr_t addr0 = ihk_mc_syscall_arg0(ctx);
+	const intptr_t addr0 = ihk_mc_syscall_arg0(ctx);
 	const size_t len0 = ihk_mc_syscall_arg1(ctx);
 	const int prot = ihk_mc_syscall_arg2(ctx);
 	const int flags0 = ihk_mc_syscall_arg3(ctx);
@@ -1716,7 +1738,7 @@ SYSCALL_DECLARE(mmap)
 	struct thread *thread = cpu_local_var(current);
 	struct vm_regions *region = &thread->vm->region;
 	int error;
-	uintptr_t addr = 0;
+	intptr_t addr = 0;
 	size_t len;
 	int flags = flags0;
 	size_t pgsize;
@@ -1779,9 +1801,8 @@ SYSCALL_DECLARE(mmap)
 		goto out;
 	}
 
-	if (addr < region->user_start
-			|| region->user_end <= addr
-			|| len > (region->user_end - region->user_start)) {
+	if ((flags & MAP_FIXED) && ((addr < region->user_start)
+			|| (region->user_end <= addr))) {
 		ekprintf("sys_mmap(%lx,%lx,%x,%x,%x,%lx):ENOMEM\n",
 				addr0, len0, prot, flags0, fd, off0);
 		error = -ENOMEM;

arch/arm64/kernel/vdso.c

@@ -14,13 +14,15 @@
 #include <ihk/debug.h>
 #include <ikc/queue.h>
 #include <vdso.h>
-#include <debug.h>
 
 //#define DEBUG_PRINT_VDSO
 
 #ifdef DEBUG_PRINT_VDSO
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...)	kprintf(__VA_ARGS__)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...)	do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
 #endif
 
 #ifdef POSTK_DEBUG_ARCH_DEP_52

arch/arm64/kernel/vdso/gettimeofday.c

@@ -1,7 +1,5 @@
 /* gettimeofday.c COPYRIGHT FUJITSU LIMITED 2016 */
 
-#include <affinity.h>
-#include <arch-memory.h>
 #include <time.h>
 #include <syscall.h>
 #include <registers.h>

arch/x86_64/elfboot/raw.lds

@@ -9,29 +9,29 @@ PHDRS
 SECTIONS
 {
 	. = SIZEOF_HEADERS;
-	. = ALIGN(4096);
+    . = ALIGN(4096);
 	.text :	{
-		*(.text)
+	      *(.text)
 	} :text
 	.data : {
-		*(.data)
-		*(.data.*)
+	      *(.data)
+	      *(.data.*)
 	} :data
 	.rodata : {
-		*(.rodata .rodata.*)
+	      *(.rodata .rodata.*)
 	} :data
 
 	. = ALIGN(8);
 	.bss : {
-		_bss_start = .;
-		*(.bss .bss.*)
-		_bss_end = .;
-		. = ALIGN(4096);
-		_stack_end = .;
-	} :data
+	     _bss_start = .;
+	     *(.bss .bss.*)
+	     _bss_end = .;
+         . = ALIGN(4096);
+         _stack_end = .;
+    } :data
 
 	/DISCARD/ : {
-		*(.eh_frame)
-		*(.note.gnu.build-id)
+	     *(.eh_frame)
+         *(.note.gnu.build-id)
 	}
-}
+}

arch/x86_64/kernel/cpu.c

@@ -31,7 +31,6 @@
 #include <prctl.h>
 #include <page.h>
 #include <kmalloc.h>
-#include <debug.h>
 
 #define LAPIC_ID            0x020
 #define LAPIC_TIMER         0x320
@@ -70,8 +69,11 @@
 //#define DEBUG_PRINT_CPU
 
 #ifdef DEBUG_PRINT_CPU
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf kprintf
+#define ekprintf kprintf
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf kprintf
 #endif
 
 static void *lapic_vp;
@@ -94,8 +96,6 @@ int gettime_local_support = 0;
 extern int ihk_mc_pt_print_pte(struct page_table *pt, void *virt);
 extern int kprintf(const char *format, ...);
 extern int interrupt_from_user(void *);
-extern void perf_start(struct mc_perf_event *event);
-extern void perf_reset(struct mc_perf_event *event);
 
 static struct idt_entry{
 	uint32_t desc[4];
@@ -847,6 +847,9 @@ void setup_x86_ap(void (*next_func)(void))
 }
 
 void arch_show_interrupt_context(const void *reg);
+void set_signal(int sig, void *regs, struct siginfo *info);
+void check_signal(unsigned long, void *, int);
+void check_sig_pending();
 extern void tlb_flush_handler(int vector);
 
 void __show_stack(uintptr_t *sp) {
@@ -874,7 +877,7 @@ void interrupt_exit(struct x86_user_context *regs)
 		cpu_enable_interrupt();
 		check_sig_pending();
 		check_need_resched();
-		check_signal(0, regs, -1);
+		check_signal(0, regs, 0);
 	}
 	else {
 		check_sig_pending();
@@ -1007,12 +1010,6 @@ void handle_interrupt(int vector, struct x86_user_context *regs)
 	set_cputime(interrupt_from_user(regs)? 0: 1);
 
 	--v->in_interrupt;
-
-	/* for migration by IPI */
-	if (v->flags & CPU_FLAG_NEED_MIGRATE) {
-		schedule();
-		check_signal(0, regs, 0);
-	}
 }
 
 void gpe_handler(struct x86_user_context *regs)
@@ -1228,6 +1225,13 @@ void cpu_pause(void)
 	asm volatile("pause" ::: "memory");
 }
 
+/* From: kernel-xppsl_1.5.2/arch/x86/include/asm/processor.h */
+/* REP NOP (PAUSE) is a good thing to insert into busy-wait loops. */
+void cpu_relax(void)
+{
+	asm volatile("rep; nop" ::: "memory");
+}
+
 /*@
   @ assigns \nothing;
   @ ensures \interrupt_disabled > 0;
@@ -1647,10 +1651,12 @@ int ihk_mc_interrupt_cpu(int cpu, int vector)
 	return 0;
 }
 
+#ifdef POSTK_DEBUG_ARCH_DEP_22
+extern void perf_start(struct mc_perf_event *event);
+extern void perf_reset(struct mc_perf_event *event);
 struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 {
 	struct thread *last;
-	struct mcs_rwlock_node_irqsave lock;
 
 	dkprintf("[%d] schedule: tlsblock_base: 0x%lX\n",
 	         ihk_mc_get_processor_id(), next->tlsblock_base);
@@ -1669,7 +1675,7 @@ struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 	}
 
 #ifdef PROFILE_ENABLE
-	if (prev && prev->profile && prev->profile_start_ts != 0) {
+	if (prev->profile && prev->profile_start_ts != 0) {
 		prev->profile_elapsed_ts +=
 			(rdtsc() - prev->profile_start_ts);
 		prev->profile_start_ts = 0;
@@ -1681,28 +1687,6 @@ struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 #endif
 
 	if (prev) {
-		mcs_rwlock_writer_lock(&prev->proc->update_lock, &lock);
-		if (prev->proc->status & (PS_DELAY_STOPPED | PS_DELAY_TRACED)) {
-			switch (prev->proc->status) {
-			case PS_DELAY_STOPPED:
-				prev->proc->status = PS_STOPPED;
-				break;
-			case PS_DELAY_TRACED:
-				prev->proc->status = PS_TRACED;
-				break;
-			default:
-				break;
-			}
-			mcs_rwlock_writer_unlock(&prev->proc->update_lock,
-						&lock);
-
-			/* Wake up the parent who tried wait4 and sleeping */
-			waitq_wakeup(&prev->proc->parent->waitpid_q);
-		} else {
-			mcs_rwlock_writer_unlock(&prev->proc->update_lock,
-						&lock);
-		}
-
 		last = ihk_mc_switch_context(&prev->ctx, &next->ctx, prev);
 	}
 	else {
@@ -1710,6 +1694,7 @@ struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 	}
 	return last;
 }
+#endif
 
 /*@
   @ requires \valid(thread);
@@ -1784,6 +1769,14 @@ void copy_fp_regs(struct thread *from, struct thread *to)
 	}
 }
 
+#ifdef POSTK_DEBUG_TEMP_FIX_19
+void
+clear_fp_regs(struct thread *thread)
+{
+	return;
+}
+#endif /* POSTK_DEBUG_TEMP_FIX_19 */
+
 /*@
   @ requires \valid(thread);
   @ assigns thread->fp_regs;
@@ -1791,11 +1784,8 @@ void copy_fp_regs(struct thread *from, struct thread *to)
 void
 restore_fp_regs(struct thread *thread)
 {
-	if (!thread->fp_regs) {
-		// only clear fpregs.
-		clear_fp_regs();
+	if (!thread->fp_regs)
 		return;
-	}
 
 	if (xsave_available) {
 		unsigned int low, high;
@@ -1814,13 +1804,6 @@ restore_fp_regs(struct thread *thread)
 	//release_fp_regs(thread);
 }
 
-void clear_fp_regs(void)
-{
-	struct cpu_local_var *v = get_this_cpu_local_var();
-
-	restore_fp_regs(&v->idle);
-}
-
 ihk_mc_user_context_t *lookup_user_context(struct thread *thread)
 {
 	ihk_mc_user_context_t *uctx = thread->uctx;

arch/x86_64/kernel/gencore.c

@@ -6,7 +6,6 @@
 #include <process.h>
 #include <string.h>
 #include <elfcore.h>
-#include <debug.h>
 
 #define	align32(x) ((((x) + 3) / 4) * 4)
 #define	alignpage(x) ((((x) + (PAGE_SIZE) - 1) / (PAGE_SIZE)) * (PAGE_SIZE))
@@ -14,16 +13,13 @@
 //#define DEBUG_PRINT_GENCORE
 
 #ifdef DEBUG_PRINT_GENCORE
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...)	kprintf(__VA_ARGS__)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...)	do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
 #endif
 
-/* Exclude reserved (mckernel's internal use), device file,
- * hole created by mprotect
- */
-#define GENCORE_RANGE_IS_INACCESSIBLE(range) \
-	    ((range->flag & (VR_RESERVED | VR_MEMTYPE_UC | VR_DONTDUMP)))
-
 /*
  * Generate a core file image, which consists of many chunks.
  * Returns an allocated table, an etnry of which is a pair of the address 
@@ -313,10 +309,12 @@ int gencore(struct thread *thread, void *regs,
 
 		dkprintf("start:%lx end:%lx flag:%lx objoff:%lx\n", 
 			 range->start, range->end, range->flag, range->objoff);
-
-		if (GENCORE_RANGE_IS_INACCESSIBLE(range)) {
+		/* We omit reserved areas because they are only for
+		   mckernel's internal use. */		   
+		if (range->flag & VR_RESERVED)
+			continue;
+		if (range->flag & VR_DONTDUMP)
 			continue;
-		}
 		/* We need a chunk for each page for a demand paging area.
 		   This can be optimized for spacial complexity but we would
 		   lose simplicity instead. */
@@ -405,9 +403,8 @@ int gencore(struct thread *thread, void *regs,
 		unsigned long flag = range->flag;
 		unsigned long size = range->end - range->start;
 
-		if (GENCORE_RANGE_IS_INACCESSIBLE(range)) {
+		if (range->flag & VR_RESERVED)
 			continue;
-		}
 
 		ph[i].p_type = PT_LOAD;
 		ph[i].p_flags = ((flag & VR_PROT_READ) ? PF_R : 0)
@@ -449,9 +446,8 @@ int gencore(struct thread *thread, void *regs,
 
 		unsigned long phys;
 
-		if (GENCORE_RANGE_IS_INACCESSIBLE(range)) {
+		if (range->flag & VR_RESERVED)
 			continue;
-		}
 		if (range->flag & VR_DEMAND_PAGING) {
 			/* Just an ad hoc kluge. */
 			unsigned long p, start, phys;

arch/x86_64/kernel/include/arch-futex.h

@@ -64,13 +64,12 @@ static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
 	return oldval;
 }
 
-static inline int futex_atomic_op_inuser(int encoded_op,
-					 int __user *uaddr)
+static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
-	int oparg = (encoded_op & 0x00fff000) >> 12;
-	int cmparg = encoded_op & 0xfff;
+	int oparg = (encoded_op << 8) >> 20;
+	int cmparg = (encoded_op << 20) >> 20;
 	int oldval = 0, ret, tem;
 
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))

arch/x86_64/kernel/include/arch-lock.h

@@ -6,7 +6,6 @@
 
 #include <ihk/cpu.h>
 #include <ihk/atomic.h>
-#include <lwk/compiler.h>
 
 //#define DEBUG_SPINLOCK
 //#define DEBUG_MCS_RWLOCK
@@ -38,58 +37,6 @@ static void ihk_mc_spinlock_init(ihk_spinlock_t *lock)
 }
 #define SPIN_LOCK_UNLOCKED { .head_tail = 0 }
 
-
-#ifdef DEBUG_SPINLOCK
-#define ihk_mc_spinlock_trylock_noirq(l) { int rc;						\
-__kprintf("[%d] call ihk_mc_spinlock_trylock_noirq %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
-rc = __ihk_mc_spinlock_trylock_noirq(l); \
- __kprintf("[%d] ret ihk_mc_spinlock_trylock_noirq\n", ihk_mc_get_processor_id()); rc; \
-}
-#else
-#define ihk_mc_spinlock_trylock_noirq __ihk_mc_spinlock_trylock_noirq
-#endif
-
-static int __ihk_mc_spinlock_trylock_noirq(ihk_spinlock_t *lock)
-{
-	ihk_spinlock_t cur = { .head_tail = lock->head_tail };
-	ihk_spinlock_t next = { .tickets.head = cur.tickets.head, .tickets.tail = cur.tickets.tail + 2 };
-	int success;
-
-	if (cur.tickets.head != cur.tickets.tail) {
-		return 0;
-	}
-
-	preempt_disable();
-
-	/* Use the same increment amount as other functions! */
-	success = __sync_bool_compare_and_swap((__ticketpair_t*)lock, cur.head_tail, next.head_tail);
-
-	if (!success) {
-		preempt_enable();
-	}
-	return success;
-}
-
-#ifdef DEBUG_SPINLOCK
-#define ihk_mc_spinlock_trylock(l, result) ({ unsigned long rc;		\
-__kprintf("[%d] call ihk_mc_spinlock_trylock %p %s:%d\n", ihk_mc_get_processor_id(), (l), __FILE__, __LINE__); \
- rc = __ihk_mc_spinlock_trylock(l, result);									\
-__kprintf("[%d] ret ihk_mc_spinlock_trylock\n", ihk_mc_get_processor_id()); rc;\
-})
-#else
-#define ihk_mc_spinlock_trylock __ihk_mc_spinlock_trylock
-#endif
-static unsigned long __ihk_mc_spinlock_trylock(ihk_spinlock_t *lock, int *result)
-{
-	unsigned long flags;
-	
-	flags = cpu_disable_interrupt_save();
-
-	*result = __ihk_mc_spinlock_trylock_noirq(lock);
-
-	return flags;
-}
-
 #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__); \
@@ -652,9 +599,4 @@ __mcs_rwlock_reader_unlock(struct mcs_rwlock_lock *lock, struct mcs_rwlock_node_
 #endif
 }
 
-static inline int irqflags_can_interrupt(unsigned long flags)
-{
-	return !!(flags & 0x200);
-}
-
 #endif

arch/x86_64/kernel/include/arch-memory.h

@@ -40,6 +40,12 @@
 #define LARGE_PAGE_MASK    (~((unsigned long)LARGE_PAGE_SIZE - 1))
 #define LARGE_PAGE_P2ALIGN (LARGE_PAGE_SHIFT - PAGE_SHIFT)
 
+#define GB_PAGE_SHIFT   30
+#define GB_PAGE_SIZE    (1UL << GB_PAGE_SHIFT)
+#define GB_PAGE_MASK    (~((unsigned long)GB_PAGE_SIZE - 1))
+#define GB_PAGE_P2ALIGN (GB_PAGE_SHIFT - PAGE_SHIFT)
+
+
 #define USER_END           0x0000800000000000UL
 #define TASK_UNMAPPED_BASE 0x00002AAAAAA00000UL
 

arch/x86_64/kernel/include/ihk/atomic.h

@@ -133,7 +133,7 @@ static inline void ihk_atomic64_inc(ihk_atomic64_t *v)
  * Note 2: xchg has side effect, so that attribute volatile is necessary,
  *	  but generally the primitive is invalid, *ptr is output argument. --ANK
  */
-#define __xg(x) ((volatile long *)(x))
+#define __xg(x) ((volatile typeof(x))(x))
 
 #define xchg4(ptr, x)						\
 ({									\

arch/x86_64/kernel/include/syscall_list.h

@@ -39,7 +39,7 @@ SYSCALL_HANDLED(15, rt_sigreturn)
 SYSCALL_HANDLED(16, ioctl)
 SYSCALL_DELEGATED(17, pread64)
 SYSCALL_DELEGATED(18, pwrite64)
-SYSCALL_DELEGATED(20, writev)
+SYSCALL_HANDLED(20, writev)
 SYSCALL_DELEGATED(21, access)
 SYSCALL_DELEGATED(23, select)
 SYSCALL_HANDLED(24, sched_yield)
@@ -114,7 +114,7 @@ SYSCALL_HANDLED(160, setrlimit)
 SYSCALL_HANDLED(164, settimeofday)
 SYSCALL_HANDLED(186, gettid)
 SYSCALL_HANDLED(200, tkill)
-SYSCALL_HANDLED(201, time)
+SYSCALL_DELEGATED(201, time)
 SYSCALL_HANDLED(202, futex)
 SYSCALL_HANDLED(203, sched_setaffinity)
 SYSCALL_HANDLED(204, sched_getaffinity)
@@ -161,7 +161,6 @@ SYSCALL_HANDLED(__NR_profile, profile)
 SYSCALL_HANDLED(730, util_migrate_inter_kernel)
 SYSCALL_HANDLED(731, util_indicate_clone)
 SYSCALL_HANDLED(732, get_system)
-SYSCALL_HANDLED(733, util_register_desc)
 
 /* McKernel Specific */
 SYSCALL_HANDLED(801, swapout)

arch/x86_64/kernel/memory.c

@@ -25,13 +25,15 @@
 #include <cls.h>
 #include <kmalloc.h>
 #include <rusage_private.h>
-#include <debug.h>
 
 //#define DEBUG
 
 #ifdef DEBUG
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...)	do { kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...)	do { kprintf(__VA_ARGS__); } while (0)
+#else
+#define dkprintf(...) do { } while (0)
+#define ekprintf(...) do { kprintf(__VA_ARGS__); } while (0)
 #endif
 
 static char *last_page;
@@ -168,6 +170,34 @@ static unsigned long setup_l3(struct page_table *pt,
 	return virt_to_phys(pt);
 }
 
+static void init_normal_area(struct page_table *pt)
+{
+	unsigned long map_start, map_end, phys, pt_phys;
+	int ident_index, virt_index;
+
+	/*
+	 * This has to start from 0x00, see load_file() in IHK-SMP.
+	 * For security reasons, we could skip holes in the LWK
+	 * assigned physical memory, but Linux mappings already map
+	 * those anyway.
+	 */
+	map_start = 0;
+	map_end = ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0);
+
+	ident_index = map_start >> PTL4_SHIFT;
+	virt_index = (MAP_ST_START >> PTL4_SHIFT) & (PT_ENTRIES - 1);
+
+	memset(pt, 0, sizeof(struct page_table));
+
+	for (phys = map_start; phys < map_end; phys += PTL4_SIZE) {
+		pt_phys = setup_l3(ihk_mc_alloc_pages(1, IHK_MC_AP_CRITICAL),
+				phys, map_start, map_end);
+
+		pt->entry[ident_index++] = pt_phys | PFL4_PDIR_ATTR;
+		pt->entry[virt_index++] = pt_phys | PFL4_PDIR_ATTR;
+	}
+}
+
 static struct page_table *__alloc_new_pt(ihk_mc_ap_flag ap_flag)
 {
 	struct page_table *newpt = ihk_mc_alloc_pages(1, ap_flag);
@@ -235,11 +265,6 @@ static unsigned long attr_to_l1attr(enum ihk_mc_pt_attribute attr)
 	}
 }
 
-#define PTLX_SHIFT(index) PTL ## index ## _SHIFT
-
-#define GET_VIRT_INDEX(virt, index, dest) \
-	dest = ((virt) >> PTLX_SHIFT(index)) & (PT_ENTRIES - 1)
-
 #define GET_VIRT_INDICES(virt, l4i, l3i, l2i, l1i) \
 	l4i = ((virt) >> PTL4_SHIFT) & (PT_ENTRIES - 1); \
 	l3i = ((virt) >> PTL3_SHIFT) & (PT_ENTRIES - 1); \
@@ -706,6 +731,26 @@ static void destroy_page_table(int level, struct page_table *pt)
 	return;
 }
 
+void ihk_mc_pt_destroy_pgd_subtree(struct page_table *pt, void *virt)
+{
+	int l4idx, l3idx, l2idx, l1idx;
+	unsigned long v = (unsigned long)virt;
+	struct page_table *lower;
+
+	GET_VIRT_INDICES(v, l4idx, l3idx, l2idx, l1idx);
+
+	if (!(pt->entry[l4idx] & PF_PRESENT))
+		return;
+
+	lower = (struct page_table *)
+		phys_to_virt(pt->entry[l4idx] & PT_PHYSMASK);
+	destroy_page_table(3, lower);
+
+	pt->entry[l4idx] = 0;
+	dkprintf("%s: virt: 0x%lx, l4idx: %d subtree destroyed\n",
+		__FUNCTION__, virt, l4idx);
+}
+
 void ihk_mc_pt_destroy(struct page_table *pt)
 {
 	const int level = 4;	/* PML4 */
@@ -1500,12 +1545,12 @@ static int clear_range_l1(void *args0, pte_t *ptep, uint64_t base,
 	if (page) {
 		dkprintf("%s: page=%p,is_in_memobj=%d,(old & PFL1_DIRTY)=%lx,memobj=%p,args->memobj->flags=%x\n", __FUNCTION__, page, page_is_in_memobj(page), (old & PFL1_DIRTY), args->memobj, args->memobj ? args->memobj->flags : -1);
 	}
-	if (page && page_is_in_memobj(page) && pte_is_dirty(&old, PTL1_SIZE) &&
-			args->memobj && !(args->memobj->flags & MF_ZEROFILL)) {
+	if (page && page_is_in_memobj(page) && (old & PFL1_DIRTY) && (args->memobj) &&
+			!(args->memobj->flags & MF_ZEROFILL)) {
 		memobj_flush_page(args->memobj, phys, PTL1_SIZE);
 	}
 
-	if (!pte_is_fileoff(&old, PTL1_SIZE)) {
+	if (!(old & PFL1_FILEOFF)) {
 		if(args->free_physical) {
 			if (!page) {
 				/* Anonymous || !XPMEM attach */
@@ -1567,11 +1612,11 @@ static int clear_range_l2(void *args0, pte_t *ptep, uint64_t base,
 			page = phys_to_page(phys);
 		}
 
-		if (page && page_is_in_memobj(page) && pte_is_dirty(&old, PTL2_SIZE)) {
+		if (page && page_is_in_memobj(page) && (old & PFL2_DIRTY)) {
 			memobj_flush_page(args->memobj, phys, PTL2_SIZE);
 		}
 
-		if (!pte_is_fileoff(&old, PTL2_SIZE)) {
+		if (!(old & PFL2_FILEOFF)) {
 			if(args->free_physical) {
 				if (!page) {
 					/* Anonymous || !XPMEM attach */
@@ -1648,13 +1693,13 @@ static int clear_range_l3(void *args0, pte_t *ptep, uint64_t base,
 			page = phys_to_page(phys);
 		}
 
-		if (page && page_is_in_memobj(page) && pte_is_dirty(&old, PTL3_SIZE)) {
+		if (page && page_is_in_memobj(page) && (old & PFL3_DIRTY)) {
 			memobj_flush_page(args->memobj, phys, PTL3_SIZE);
 		}
 
 		dkprintf("%s: phys=%ld, pte_get_phys(&old),PTL3_SIZE\n", __FUNCTION__, pte_get_phys(&old));
 
-		if (!pte_is_fileoff(&old, PTL3_SIZE)) {
+		if (!(old & PFL3_FILEOFF)) {
 			if(args->free_physical) {
 				if (!page) {
 					/* Anonymous || !XPMEM attach */
@@ -1942,6 +1987,28 @@ out:
 	return ptep;
 }
 
+pte_t *ihk_mc_pt_lookup_fault_pte(struct process_vm *vm, void *virt,
+		int pgshift, void **basep, size_t *sizep, int *p2alignp)
+{
+	int faulted = 0;
+	pte_t *ptep;
+
+retry:
+	ptep = ihk_mc_pt_lookup_pte(vm->address_space->page_table,
+			virt, pgshift, basep, sizep, p2alignp);
+	if (!faulted && (!ptep || !pte_is_present(ptep))) {
+		page_fault_process_vm(vm, virt, PF_POPULATE | PF_USER);
+		faulted = 1;
+		goto retry;
+	}
+
+	if (faulted && ptep && pte_is_present(ptep)) {
+		kprintf("%s: successfully faulted 0x%lx\n", __FUNCTION__, virt);
+	}
+
+	return ptep;
+}
+
 pte_t *ihk_mc_pt_lookup_pte(page_table_t pt, void *virt, int pgshift,
 		void **basep, size_t *sizep, int *p2alignp)
 {
@@ -2241,7 +2308,7 @@ out:
 
 int ihk_mc_pt_set_range(page_table_t pt, struct process_vm *vm, void *start, 
 		void *end, uintptr_t phys, enum ihk_mc_pt_attribute attr,
-						int pgshift, struct vm_range *range)
+		int pgshift, struct vm_range *range)
 {
 	int error;
 	struct set_range_args args;
@@ -2522,82 +2589,6 @@ static void init_fixed_area(struct page_table *pt)
 	return;
 }
 
-static void init_normal_area(struct page_table *pt)
-{
-	unsigned long map_start, map_end, phys;
-	void *virt;
-
-	map_start = ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0);
-	map_end = ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0);
-	virt = (void *)MAP_ST_START + map_start;
-
-	kprintf("map_start = %lx, map_end = %lx, virt %lx\n",
-		map_start, map_end, virt);
-
-	for (phys = map_start; phys < map_end; phys += LARGE_PAGE_SIZE) {
-		if (set_pt_large_page(pt, virt, phys, PTATTR_WRITABLE) != 0) {
-			kprintf("%s: error setting mapping for 0x%lx\n",
-					__func__, virt);
-		}
-		virt += LARGE_PAGE_SIZE;
-	}
-}
-
-static void init_linux_kernel_mapping(struct page_table *pt)
-{
-	unsigned long map_start, map_end, phys;
-	void *virt;
-	int nr_memory_chunks, chunk_id, numa_id;
-
-	/* In case of safe_kernel_map option (safe_kernel_map == 1),
-	 * processing to prevent destruction of the memory area on Linux side
-	 * is executed */
-	if (safe_kernel_map == 0) {
-		kprintf("Straight-map entire physical memory\n");
-
-		/* Map 2 TB for now */
-		map_start = 0;
-		map_end = 0x20000000000;
-
-		virt = (void *)LINUX_PAGE_OFFSET;
-
-		kprintf("Linux kernel virtual: 0x%lx - 0x%lx -> 0x%lx - 0x%lx\n",
-			LINUX_PAGE_OFFSET, LINUX_PAGE_OFFSET + map_end, 0, map_end);
-
-		for (phys = map_start; phys < map_end; phys += LARGE_PAGE_SIZE) {
-			if (set_pt_large_page(pt, virt, phys, PTATTR_WRITABLE) != 0) {
-				kprintf("%s: error setting mapping for 0x%lx\n", __FUNCTION__, virt);
-			}
-			virt += LARGE_PAGE_SIZE;
-		}
-	} else {
-		kprintf("Straight-map physical memory areas allocated to McKernel\n");
-
-		nr_memory_chunks = ihk_mc_get_nr_memory_chunks();
-		if (nr_memory_chunks == 0) {
-			kprintf("%s: ERROR: No memory chunk available.\n", __FUNCTION__);
-			return;
-		}
-
-		for (chunk_id = 0; chunk_id < nr_memory_chunks; chunk_id++) {
-			if (ihk_mc_get_memory_chunk(chunk_id, &map_start, &map_end, &numa_id)) {
-				kprintf("%s: ERROR: Memory chunk id (%d) out of range.\n", __FUNCTION__, chunk_id);
-				continue;
-			}
-
-			dkprintf("Linux kernel virtual: 0x%lx - 0x%lx -> 0x%lx - 0x%lx\n",
-					 LINUX_PAGE_OFFSET + map_start, LINUX_PAGE_OFFSET + map_end, map_start, map_end);
-
-			virt = (void *)(LINUX_PAGE_OFFSET + map_start);
-			for (phys = map_start; phys < map_end; phys += LARGE_PAGE_SIZE, virt += LARGE_PAGE_SIZE) {
-				if (set_pt_large_page(pt, virt, phys, PTATTR_WRITABLE) != 0) {
-					kprintf("%s: set_pt_large_page() failed for 0x%lx\n", __FUNCTION__, virt);
-				}
-			}
-		}
-	}
-}
-
 void init_text_area(struct page_table *pt)
 {
 	unsigned long __end, phys, virt;
@@ -2661,6 +2652,61 @@ void init_low_area(struct page_table *pt)
 	set_pt_large_page(pt, 0, 0, PTATTR_NO_EXECUTE|PTATTR_WRITABLE);
 }
 
+static void init_linux_kernel_mapping(struct page_table *pt)
+{
+	unsigned long map_start, map_end, phys;
+	void *virt;
+	int nr_memory_chunks, chunk_id, numa_id;
+	
+	/* In case of safe_kernel_map option (safe_kernel_map == 1), 
+	   processing to prevent destruction of the memory area on Linux side 
+	   is executed */
+	if (safe_kernel_map == 0) {
+		kprintf("Straight-map entire physical memory\n");
+
+		/* Map 2 TB for now */
+		map_start = 0;
+		map_end = 0x20000000000;
+
+		virt = (void *)LINUX_PAGE_OFFSET;
+
+		kprintf("Linux kernel virtual: 0x%lx - 0x%lx -> 0x%lx - 0x%lx\n",
+			LINUX_PAGE_OFFSET, LINUX_PAGE_OFFSET + map_end, 0, map_end);
+
+		for (phys = map_start; phys < map_end; phys += LARGE_PAGE_SIZE) {
+			if (set_pt_large_page(pt, virt, phys, PTATTR_WRITABLE) != 0) {
+				kprintf("%s: error setting mapping for 0x%lx\n", __FUNCTION__, virt);
+			}
+			virt += LARGE_PAGE_SIZE;
+		}
+	} else {
+		kprintf("Straight-map physical memory areas allocated to McKernel\n");
+
+		nr_memory_chunks = ihk_mc_get_nr_memory_chunks();
+		if (nr_memory_chunks == 0) {
+			kprintf("%s: ERROR: No memory chunk available.\n", __FUNCTION__);
+			return;
+		}
+
+		for (chunk_id = 0; chunk_id < nr_memory_chunks; chunk_id++) {
+			if (ihk_mc_get_memory_chunk(chunk_id, &map_start, &map_end, &numa_id)) {
+				kprintf("%s: ERROR: Memory chunk id (%d) out of range.\n", __FUNCTION__, chunk_id);
+				continue;
+			}
+			
+			dkprintf("Linux kernel virtual: 0x%lx - 0x%lx -> 0x%lx - 0x%lx\n",
+					 LINUX_PAGE_OFFSET + map_start, LINUX_PAGE_OFFSET + map_end, map_start, map_end);
+
+			virt = (void *)(LINUX_PAGE_OFFSET + map_start);
+			for (phys = map_start; phys < map_end; phys += LARGE_PAGE_SIZE, virt += LARGE_PAGE_SIZE) {
+				if (set_pt_large_page(pt, virt, phys, PTATTR_WRITABLE) != 0) {
+					kprintf("%s: set_pt_large_page() failed for 0x%lx\n", __FUNCTION__, virt);
+				}
+			}
+		}
+	}
+}
+
 static void init_vsyscall_area(struct page_table *pt)
 {
 	extern char vsyscall_page[];
@@ -2682,7 +2728,7 @@ void init_page_table(void)
 	init_pt = ihk_mc_alloc_pages(1, IHK_MC_AP_CRITICAL);
 	ihk_mc_spinlock_init(&init_pt_lock);
 	
-	memset(init_pt, 0, sizeof(*init_pt));
+	memset(init_pt, 0, sizeof(PAGE_SIZE));
 
 	/* Normal memory area */
 	init_normal_area(init_pt);
@@ -2916,12 +2962,17 @@ int read_process_vm(struct process_vm *vm, void *kdst, const void *usrc, size_t
 			return error;
 		}
 
-		if (!is_mckernel_memory(pa, pa + cpsize)) {
+#ifdef POSTK_DEBUG_TEMP_FIX_52 /* NUMA support(memory area determination) */
+		if (!is_mckernel_memory(pa)) {
+#else
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+#endif /* POSTK_DEBUG_TEMP_FIX_52 */
 			dkprintf("%s: pa is outside of LWK memory, to: %p, pa: %p,"
 				"cpsize: %d\n", __FUNCTION__, to, pa, cpsize);
 			va = ihk_mc_map_virtual(pa, 1, PTATTR_ACTIVE);
 			memcpy(to, va, cpsize);
-			ihk_mc_unmap_virtual(va, 1);
+			ihk_mc_unmap_virtual(va, 1, 1);
 		}
 		else {
 			va = phys_to_virt(pa);
@@ -2995,12 +3046,17 @@ int write_process_vm(struct process_vm *vm, void *udst, const void *ksrc, size_t
 			return error;
 		}
 
-		if (!is_mckernel_memory(pa, pa + cpsize)) {
+#ifdef POSTK_DEBUG_TEMP_FIX_52 /* NUMA support(memory area determination) */
+		if (!is_mckernel_memory(pa)) {
+#else
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+#endif /* POSTK_DEBUG_TEMP_FIX_52 */
 			dkprintf("%s: pa is outside of LWK memory, from: %p,"
 				"pa: %p, cpsize: %d\n", __FUNCTION__, from, pa, cpsize);
 			va = ihk_mc_map_virtual(pa, 1, PTATTR_ACTIVE);
 			memcpy(va, from, cpsize);
-			ihk_mc_unmap_virtual(va, 1);
+			ihk_mc_unmap_virtual(va, 1, 1);
 		}
 		else {
 			va = phys_to_virt(pa);
@@ -3061,12 +3117,17 @@ int patch_process_vm(struct process_vm *vm, void *udst, const void *ksrc, size_t
 			return error;
 		}
 
-		if (!is_mckernel_memory(pa, pa + cpsize)) {
+#ifdef POSTK_DEBUG_TEMP_FIX_52 /* NUMA support(memory area determination) */
+		if (!is_mckernel_memory(pa)) {
+#else
+		if (pa < ihk_mc_get_memory_address(IHK_MC_GMA_MAP_START, 0) ||
+			pa >= ihk_mc_get_memory_address(IHK_MC_GMA_MAP_END, 0)) {
+#endif /* POSTK_DEBUG_TEMP_FIX_52 */
 			dkprintf("%s: pa is outside of LWK memory, from: %p,"
 				"pa: %p, cpsize: %d\n", __FUNCTION__, from, pa, cpsize);
 			va = ihk_mc_map_virtual(pa, 1, PTATTR_ACTIVE);
 			memcpy(va, from, cpsize);
-			ihk_mc_unmap_virtual(va, 1);
+			ihk_mc_unmap_virtual(va, 1, 1);
 		}
 		else {
 			va = phys_to_virt(pa);

arch/x86_64/kernel/mikc.c

@@ -30,7 +30,7 @@ int ihk_mc_ikc_init_first_local(struct ihk_ikc_channel_desc *channel,
 
 	memset(channel, 0, sizeof(struct ihk_ikc_channel_desc));
 
-	mikc_queue_pages = ((4 * num_processors * MASTER_IKCQ_PKTSIZE)
+	mikc_queue_pages = ((2 * num_processors * MASTER_IKCQ_PKTSIZE)
 			+ (PAGE_SIZE - 1)) / PAGE_SIZE;
 
 	/* Place both sides in this side */

arch/x86_64/kernel/perfctr.c

@@ -16,16 +16,20 @@
 #include <registers.h>
 #include <mc_perf_event.h>
 #include <config.h>
-#include <debug.h>
 
 extern unsigned int *x86_march_perfmap;
 extern int running_on_kvm(void);
+#ifdef POSTK_DEBUG_TEMP_FIX_31
 int ihk_mc_perfctr_fixed_init(int counter, int mode);
+#endif/*POSTK_DEBUG_TEMP_FIX_31*/
 
 //#define PERFCTR_DEBUG
 #ifdef PERFCTR_DEBUG
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...)	do { kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...)	do { kprintf(__VA_ARGS__); } while (0)
+#else
+#define	dkprintf(...)	do { } while (0)
+#define	ekprintf(...)	do { kprintf(__VA_ARGS__); } while (0)
 #endif
 
 #define X86_CR4_PCE     0x00000100
@@ -39,11 +43,11 @@ int ihk_mc_perfctr_fixed_init(int counter, int mode);
 		}                                                               \
     } while(0)
 
-int perf_counters_discovered;
-int NUM_PERF_COUNTERS;
-unsigned long PERF_COUNTERS_MASK;
-int NUM_FIXED_PERF_COUNTERS;
-unsigned long FIXED_PERF_COUNTERS_MASK;
+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)
 {
@@ -74,17 +78,17 @@ void x86_init_perfctr(void)
 		op = 0x0a;
 		asm volatile("cpuid" : "=a"(eax),"=b"(ebx),"=c"(ecx),"=d"(edx):"a"(op));
 
-		NUM_PERF_COUNTERS = ((eax & 0xFF00) >> 8);
-		PERF_COUNTERS_MASK = (1 << NUM_PERF_COUNTERS) - 1;
+		X86_IA32_NUM_PERF_COUNTERS = ((eax & 0xFF00) >> 8);
+		X86_IA32_PERF_COUNTERS_MASK = (1 << X86_IA32_NUM_PERF_COUNTERS) - 1;
 		
-		NUM_FIXED_PERF_COUNTERS = (edx & 0x0F);
-		FIXED_PERF_COUNTERS_MASK =
-			((1UL << NUM_FIXED_PERF_COUNTERS) - 1) <<
-			BASE_FIXED_PERF_COUNTERS;
+		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("NUM_PERF_COUNTERS: %d, NUM_FIXED_PERF_COUNTERS: %d\n",
-				NUM_PERF_COUNTERS, NUM_FIXED_PERF_COUNTERS);
+		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 */
@@ -93,20 +97,20 @@ void x86_init_perfctr(void)
 	wrmsr(MSR_PERF_FIXED_CTRL, value);
 
 	/* Clear Generic Counter Control */
-	for (i = 0; i < NUM_PERF_COUNTERS; i++) {
+	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 |= PERF_COUNTERS_MASK;
-	value |= FIXED_PERF_COUNTERS_MASK;
+	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)
 {
-	if (counter < 0 || counter >= NUM_PERF_COUNTERS) {
+	if (counter < 0 || counter >= X86_IA32_NUM_PERF_COUNTERS) {
 		return -EINVAL;
 	}
 
@@ -145,14 +149,13 @@ static int set_pmc_x86_direct(int counter, long val)
 	val &= 0x000000ffffffffff; // 40bit Mask
 
 	cnt_bit = 1UL << counter;
-	if (cnt_bit & PERF_COUNTERS_MASK) {
+	if ( cnt_bit & X86_IA32_PERF_COUNTERS_MASK ) {
 		// set generic pmc
 		wrmsr(MSR_IA32_PMC0 + counter, val);
 	}
-	else if (cnt_bit & FIXED_PERF_COUNTERS_MASK) {
+	else if ( cnt_bit & X86_IA32_FIXED_PERF_COUNTERS_MASK ) {
 		// set fixed pmc
-		wrmsr(MSR_IA32_FIXED_CTR0 +
-			counter - BASE_FIXED_PERF_COUNTERS, val);
+		wrmsr(MSR_IA32_FIXED_CTR0 + counter - X86_IA32_BASE_FIXED_PERF_COUNTERS, val);
 	}
 	else {
 		return -EINVAL;
@@ -172,10 +175,10 @@ static int set_fixed_counter(int counter, int mode)
 {
 	unsigned long value = 0;
 	unsigned int  ctr_mask = 0xf;
-	int counter_idx = counter - BASE_FIXED_PERF_COUNTERS;
+	int counter_idx = counter - X86_IA32_BASE_FIXED_PERF_COUNTERS ;
 	unsigned int  set_val = 0;
 
-	if (counter_idx < 0 || counter_idx >= NUM_FIXED_PERF_COUNTERS) {
+	if (counter_idx < 0 || counter_idx >= X86_IA32_NUM_FIXED_PERF_COUNTERS) {
 		return -EINVAL;
 	}
 
@@ -205,13 +208,14 @@ int ihk_mc_perfctr_init_raw(int counter, uint64_t config, int mode)
 int ihk_mc_perfctr_init_raw(int counter, unsigned int code, int mode)
 #endif /*POSTK_DEBUG_TEMP_FIX_29*/
 {
+#ifdef POSTK_DEBUG_TEMP_FIX_31
 	// PAPI_REF_CYC counted by fixed counter
-	if (counter >= BASE_FIXED_PERF_COUNTERS &&
-		counter < BASE_FIXED_PERF_COUNTERS + NUM_FIXED_PERF_COUNTERS) {
+	if (counter >= X86_IA32_BASE_FIXED_PERF_COUNTERS) {
 		return ihk_mc_perfctr_fixed_init(counter, mode);
 	}
+#endif /*POSTK_DEBUG_TEMP_FIX_31*/
 
-	if (counter < 0 || counter >= NUM_PERF_COUNTERS) {
+	if (counter < 0 || counter >= X86_IA32_NUM_PERF_COUNTERS) {
 		return -EINVAL;
 	}
 
@@ -244,7 +248,7 @@ int ihk_mc_perfctr_init(int counter, enum ihk_perfctr_type type, int mode)
 	}
 #endif /*POSTK_DEBUG_TEMP_FIX_29*/
 
-	if (counter < 0 || counter >= NUM_PERF_COUNTERS) {
+	if (counter < 0 || counter >= X86_IA32_NUM_PERF_COUNTERS) {
 		return -EINVAL;
 	}
 	if (type < 0 || type >= PERFCTR_MAX_TYPE) {
@@ -296,11 +300,18 @@ int ihk_mc_perfctr_set_extra(struct mc_perf_event *event)
 extern void x86_march_perfctr_start(unsigned long counter_mask);
 #endif
 
+#ifdef POSTK_DEBUG_TEMP_FIX_30
+int ihk_mc_perfctr_start(int counter)
+#else
 int ihk_mc_perfctr_start(unsigned long counter_mask)
+#endif /*POSTK_DEBUG_TEMP_FIX_30*/
 {
 	int ret = 0;
 	unsigned long value = 0;
-	unsigned long mask = PERF_COUNTERS_MASK | FIXED_PERF_COUNTERS_MASK;
+	unsigned long mask = X86_IA32_PERF_COUNTERS_MASK | X86_IA32_FIXED_PERF_COUNTERS_MASK;
+#ifdef POSTK_DEBUG_TEMP_FIX_30
+	unsigned long counter_mask = 1UL << counter;
+#endif /*POSTK_DEBUG_TEMP_FIX_30*/
 
 	PERFCTR_CHKANDJUMP(counter_mask & ~mask, "counter_mask out of range", -EINVAL);
 
@@ -317,11 +328,18 @@ int ihk_mc_perfctr_start(unsigned long counter_mask)
 	goto fn_exit;
 }
 
+#ifdef POSTK_DEBUG_TEMP_FIX_30
+int ihk_mc_perfctr_stop(int counter)
+#else
 int ihk_mc_perfctr_stop(unsigned long counter_mask)
+#endif/*POSTK_DEBUG_TEMP_FIX_30*/
 {
 	int ret = 0;
 	unsigned long value;
-	unsigned long mask = PERF_COUNTERS_MASK | FIXED_PERF_COUNTERS_MASK;
+	unsigned long mask = X86_IA32_PERF_COUNTERS_MASK | X86_IA32_FIXED_PERF_COUNTERS_MASK;
+#ifdef POSTK_DEBUG_TEMP_FIX_30
+	unsigned long counter_mask = 1UL << counter;
+#endif/*POSTK_DEBUG_TEMP_FIX_30*/
 
 	PERFCTR_CHKANDJUMP(counter_mask & ~mask, "counter_mask out of range", -EINVAL);
 
@@ -358,10 +376,10 @@ int ihk_mc_perfctr_fixed_init(int counter, int mode)
 {
 	unsigned long value = 0;
 	unsigned int  ctr_mask = 0xf;
-	int counter_idx = counter - BASE_FIXED_PERF_COUNTERS;
+	int counter_idx = counter - X86_IA32_BASE_FIXED_PERF_COUNTERS ;
 	unsigned int  set_val = 0;
 
-	if (counter_idx < 0 || counter_idx >= NUM_FIXED_PERF_COUNTERS) {
+	if (counter_idx < 0 || counter_idx >= X86_IA32_NUM_FIXED_PERF_COUNTERS) {
 		return -EINVAL;
 	}
 
@@ -402,7 +420,7 @@ int ihk_mc_perfctr_read_mask(unsigned long counter_mask, unsigned long *value)
 {
 	int i, j;
 
-	for (i = 0, j = 0; i < NUM_PERF_COUNTERS && counter_mask;
+	for (i = 0, j = 0; i < X86_IA32_NUM_PERF_COUNTERS && counter_mask;
 	     i++, counter_mask >>= 1) {
 		if (counter_mask & 1) {
 			value[j++] = rdpmc(i);
@@ -422,14 +440,13 @@ unsigned long ihk_mc_perfctr_read(int counter)
 
 	cnt_bit = 1UL << counter;
 
-	if (cnt_bit & PERF_COUNTERS_MASK) {
+	if ( cnt_bit & X86_IA32_PERF_COUNTERS_MASK ) {
 		// read generic pmc
 		retval = rdpmc(counter);
 	}
-	else if (cnt_bit & FIXED_PERF_COUNTERS_MASK) {
+	else if ( cnt_bit & X86_IA32_FIXED_PERF_COUNTERS_MASK ) {
 		// read fixed pmc
-		retval = rdpmc((1 << 30) +
-			(counter - BASE_FIXED_PERF_COUNTERS));
+		retval = rdpmc((1 << 30) + (counter - X86_IA32_BASE_FIXED_PERF_COUNTERS));
 	}
 	else {
 		retval = -EINVAL;
@@ -451,12 +468,12 @@ unsigned long ihk_mc_perfctr_read_msr(int counter)
 
 	cnt_bit = 1UL << counter;
 
-	if (cnt_bit & PERF_COUNTERS_MASK) {
+	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 & FIXED_PERF_COUNTERS_MASK) {
+	else if ( cnt_bit & X86_IA32_FIXED_PERF_COUNTERS_MASK ) {
 		// read fixed pmc
 		idx = MSR_IA32_FIXED_CTR0 + counter;
 		retval = (unsigned long) rdmsr(idx);
@@ -489,8 +506,8 @@ int ihk_mc_perfctr_alloc_counter(unsigned int *type, unsigned long *config, unsi
 	}
 
 	// find avail generic counter
-	for (i = 0; i < NUM_PERF_COUNTERS; i++) {
-		if (!(pmc_status & (1 << i))) {
+       	for(i = 0; i < X86_IA32_NUM_PERF_COUNTERS; i++) {
+		if(!(pmc_status & (1 << i))) {
 			ret = i;
 			break;
 		}

arch/x86_64/kernel/syscall.c

@@ -31,11 +31,12 @@
 #include <page.h>
 #include <limits.h>
 #include <syscall.h>
-#include <debug.h>
 
 void terminate_mcexec(int, int);
 extern long do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact);
 long syscall(int num, ihk_mc_user_context_t *ctx);
+void set_signal(int sig, void *regs0, siginfo_t *info);
+void check_signal(unsigned long rc, void *regs0, int num);
 extern unsigned long do_fork(int, unsigned long, unsigned long, unsigned long,
 	unsigned long, unsigned long, unsigned long);
 extern int get_xsave_size();
@@ -44,8 +45,11 @@ extern uint64_t get_xsave_mask();
 //#define DEBUG_PRINT_SC
 
 #ifdef DEBUG_PRINT_SC
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf kprintf
+#define ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) kprintf(__VA_ARGS__)
 #endif
 
 uintptr_t debug_constants[] = {
@@ -88,45 +92,33 @@ static ptrdiff_t vdso_offset;
 
 extern int num_processors;
 
-int obtain_clone_cpuid(cpu_set_t *cpu_set, int use_last) {
+int obtain_clone_cpuid(cpu_set_t *cpu_set) {
 	int min_queue_len = -1;
-	int cpu, min_cpu = -1, uti_cpu = -1;
-	unsigned long irqstate;
-	
-	irqstate = ihk_mc_spinlock_lock(&runq_reservation_lock);
+	int cpu, min_cpu = -1;
+
 	/* Find the first allowed core with the shortest run queue */
 	for (cpu = 0; cpu < num_processors; ++cpu) {
 		struct cpu_local_var *v;
+		unsigned long irqstate;
 
 		if (!CPU_ISSET(cpu, cpu_set)) continue;
 
 		v = get_cpu_local_var(cpu);
-		ihk_mc_spinlock_lock_noirq(&v->runq_lock);
-		dkprintf("%s: cpu=%d,runq_len=%d,runq_reserved=%d\n", __FUNCTION__, cpu, v->runq_len, v->runq_reserved);
-		if (min_queue_len == -1 || v->runq_len + v->runq_reserved < min_queue_len) {
-			min_queue_len = v->runq_len + v->runq_reserved;
+		irqstate = ihk_mc_spinlock_lock(&v->runq_lock);
+		if (min_queue_len == -1 || v->runq_len < min_queue_len) {
+			min_queue_len = v->runq_len;
 			min_cpu = cpu;
 		}
+		ihk_mc_spinlock_unlock(&v->runq_lock, irqstate);
 
-		/* Record the last tie CPU */
-		if (min_cpu != cpu && v->runq_len + v->runq_reserved == min_queue_len) {
-			uti_cpu = cpu;
-		}
-		dkprintf("%s: cpu=%d,runq_len=%d,runq_reserved=%d,min_cpu=%d,uti_cpu=%d\n", __FUNCTION__, cpu, v->runq_len, v->runq_reserved, min_cpu, uti_cpu);
-		ihk_mc_spinlock_unlock_noirq(&v->runq_lock);
-#if 0
 		if (min_queue_len == 0)
 			break;
-#endif
 	}
 
-	min_cpu = use_last ? uti_cpu : min_cpu;
 	if (min_cpu != -1) {
 		if (get_cpu_local_var(min_cpu)->status != CPU_STATUS_RESERVED)
 			get_cpu_local_var(min_cpu)->status = CPU_STATUS_RESERVED;
-		__sync_fetch_and_add(&get_cpu_local_var(min_cpu)->runq_reserved, 1);
 	}
-	ihk_mc_spinlock_unlock(&runq_reservation_lock, irqstate);
 
     return min_cpu;
 }
@@ -259,7 +251,7 @@ SYSCALL_DECLARE(rt_sigreturn)
 		info.si_code = TRAP_TRACE;
 		set_signal(SIGTRAP, regs, &info);
 		check_need_resched();
-		check_signal(0, regs, -1);
+		check_signal(0, regs, 0);
 	}
 
 	if(ksigsp.fpregs && xsavesize){
@@ -284,6 +276,7 @@ SYSCALL_DECLARE(rt_sigreturn)
 }
 
 extern struct cpu_local_var *clv;
+extern unsigned long do_kill(struct thread *thread, int pid, int tid, int sig, struct siginfo *info, int ptracecont);
 extern void interrupt_syscall(struct thread *, int sig);
 extern void terminate(int, int);
 extern int num_processors;
@@ -537,32 +530,23 @@ void ptrace_report_signal(struct thread *thread, int sig)
 	dkprintf("ptrace_report_signal, tid=%d, pid=%d\n", thread->tid, thread->proc->pid);
 
 	mcs_rwlock_writer_lock(&proc->update_lock, &lock);	
-	if (!(thread->ptrace & PT_TRACED)) {
+	if(!(proc->ptrace & PT_TRACED)){
 		mcs_rwlock_writer_unlock(&proc->update_lock, &lock);
 		return;
 	}
-
-	/* Transition thread state */
 	thread->exit_status = sig;
+	/* Transition thread state */
+	proc->status = PS_TRACED;
 	thread->status = PS_TRACED;
-	thread->ptrace &= ~PT_TRACE_SYSCALL;
-	save_debugreg(thread->ptrace_debugreg);
+	proc->ptrace &= ~PT_TRACE_SYSCALL;
 	if (sig == SIGSTOP || sig == SIGTSTP ||
-	    sig == SIGTTIN || sig == SIGTTOU) {
-		thread->signal_flags |= SIGNAL_STOP_STOPPED;
-	}
-	else {
-		thread->signal_flags &= ~SIGNAL_STOP_STOPPED;
-	}
-
-	if (thread == proc->main_thread) {
-		proc->status = PS_DELAY_TRACED;
-		parent_pid = proc->parent->pid;
-	}
-	else {
-		parent_pid = thread->report_proc->pid;
-		waitq_wakeup(&thread->report_proc->waitpid_q);
+			sig == SIGTTIN || sig == SIGTTOU) {
+		proc->signal_flags |= SIGNAL_STOP_STOPPED;
+	} else {
+		proc->signal_flags &= ~SIGNAL_STOP_STOPPED;
 	}
+	parent_pid = proc->parent->pid;
+	save_debugreg(thread->ptrace_debugreg);
 	mcs_rwlock_writer_unlock(&proc->update_lock, &lock);
 
 	memset(&info, '\0', sizeof info);
@@ -571,6 +555,8 @@ void ptrace_report_signal(struct thread *thread, int sig)
 	info._sifields._sigchld.si_pid = thread->tid;
 	info._sifields._sigchld.si_status = thread->exit_status;
 	do_kill(cpu_local_var(current), parent_pid, -1, SIGCHLD, &info, 0);
+	/* Wake parent (if sleeping in wait4()) */
+	waitq_wakeup(&proc->parent->waitpid_q);
 
 	dkprintf("ptrace_report_signal,sleeping\n");
 	/* Sleep */
@@ -583,8 +569,9 @@ ptrace_arch_prctl(int pid, long code, long addr)
 {
 	long rc = -EIO;
 	struct thread *child;
+	struct mcs_rwlock_node_irqsave lock;
 
-	child = find_thread(pid, pid);
+	child = find_thread(pid, pid, &lock);
 	if (!child)
 		return -ESRCH;
 	if (child->proc->status & (PS_TRACED | PS_STOPPED)) {
@@ -626,7 +613,7 @@ ptrace_arch_prctl(int pid, long code, long addr)
 			break;
 		}
 	}
-	thread_unlock(child);
+	thread_unlock(child, &lock);
 
 	return rc;
 }
@@ -648,13 +635,11 @@ arch_ptrace(long request, int pid, long addr, long data)
 static int
 isrestart(int num, unsigned long rc, int sig, int restart)
 {
-	if (sig == SIGKILL || sig == SIGSTOP)
+	if(sig == SIGKILL || sig == SIGSTOP)
 		return 0;
-	if (num < 0 || rc != -EINTR)
+	if(num == 0 || rc != -EINTR)
 		return 0;
-	if (sig == SIGCHLD)
-		return 1;
-	switch (num) {
+	switch(num){
 	    case __NR_pause:
 	    case __NR_rt_sigsuspend:
 	    case __NR_rt_sigtimedwait:
@@ -675,12 +660,14 @@ isrestart(int num, unsigned long rc, int sig, int restart)
 	    case __NR_io_getevents:
 		return 0;
 	}
-	if (restart)
+	if(sig == SIGCHLD)
+		return 1;
+	if(restart)
 		return 1;
 	return 0;
 }
 
-int
+void
 do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pending *pending, int num)
 {
 	struct x86_user_context *regs = regs0;
@@ -692,15 +679,14 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 	int	ptraceflag = 0;
 	struct mcs_rwlock_node_irqsave lock;
 	struct mcs_rwlock_node_irqsave mcs_rw_node;
-	int restart = 0;
 
 	for(w = pending->sigmask.__val[0], sig = 0; w; sig++, w >>= 1);
 	dkprintf("do_signal(): tid=%d, pid=%d, sig=%d\n", thread->tid, proc->pid, sig);
 	orgsig = sig;
 
-	if ((thread->ptrace & PT_TRACED) &&
-	    pending->ptracecont == 0 &&
-	    sig != SIGKILL) {
+	if((proc->ptrace & PT_TRACED) &&
+	   pending->ptracecont == 0 &&
+	   sig != SIGKILL) {
 		ptraceflag = 1;
 		sig = SIGSTOP;
 	}
@@ -721,7 +707,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 	if(k->sa.sa_handler == SIG_IGN){
 		kfree(pending);
 		mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
-		goto out;
+		return;
 	}
 	else if(k->sa.sa_handler){
 		unsigned long *usp; /* user stack */
@@ -771,8 +757,9 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		memcpy(&ksigsp.sigstack, &thread->sigstack, sizeof(stack_t));
 		ksigsp.sigrc = rc;
 		ksigsp.num = num;
-		restart = isrestart(num, rc, sig, k->sa.sa_flags & SA_RESTART);
-		ksigsp.restart = restart;
+		ksigsp.restart = isrestart(num, rc, sig, k->sa.sa_flags & SA_RESTART);
+		if(num != 0 && rc == -EINTR && sig == SIGCHLD)
+			ksigsp.restart = 1;
 		if(xsavesize){
 			uint64_t xsave_mask = get_xsave_mask();
 			unsigned int low = (unsigned int)xsave_mask;
@@ -785,7 +772,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 				kfree(_kfpregs);
 				kprintf("do_signal,no space available\n");
 				terminate(0, sig);
-				goto out;
+				return;
 			}
 			kfpregs = (void *)((((unsigned long)_kfpregs) + 63) & ~63);
 			memset(kfpregs, '\0', xsavesize);
@@ -795,7 +782,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 				kfree(_kfpregs);
 				kprintf("do_signal,write_process_vm failed\n");
 				terminate(0, sig);
-				goto out;
+				return;
 			}
 			ksigsp.fpregs = (void *)fpregs;
 			kfree(_kfpregs);
@@ -807,7 +794,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 			mcs_rwlock_writer_unlock(&thread->sigcommon->lock, &mcs_rw_node);
 			kprintf("do_signal,write_process_vm failed\n");
 			terminate(0, sig);
-			goto out;
+			return;
 		}
 
 		usp = (unsigned long *)sigsp;
@@ -837,13 +824,12 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 			info.si_code = TRAP_TRACE;
 			set_signal(SIGTRAP, regs, &info);
 			check_need_resched();
-			check_signal(0, regs, -1);
+			check_signal(0, regs, 0);
 		}
 	}
 	else {
 		int	coredumped = 0;
 		siginfo_t info;
-		int ptc = pending->ptracecont;
 
 		if(ptraceflag){
 			if(thread->ptrace_recvsig)
@@ -870,37 +856,25 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 				info.si_code = CLD_STOPPED;
 				info._sifields._sigchld.si_pid = thread->proc->pid;
 				info._sifields._sigchld.si_status = (sig << 8) | 0x7f;
-				if (ptc == 2 &&
-				    thread != thread->proc->main_thread) {
-					thread->signal_flags =
-							    SIGNAL_STOP_STOPPED;
-					thread->status = PS_STOPPED;
-					thread->exit_status = SIGSTOP;
-					do_kill(thread,
-						thread->report_proc->pid, -1,
-						SIGCHLD, &info, 0);
-					waitq_wakeup(
-					       &thread->report_proc->waitpid_q);
-				}
-				else {
-					/* Update thread state in fork tree */
-					mcs_rwlock_writer_lock(
-						     &proc->update_lock, &lock);
-					proc->group_exit_status = SIGSTOP;
+				do_kill(cpu_local_var(current), thread->proc->parent->pid, -1, SIGCHLD, &info, 0);
+				dkprintf("do_signal,SIGSTOP,changing state\n");
 
-					/* Reap and set new signal_flags */
-					proc->main_thread->signal_flags =
-							    SIGNAL_STOP_STOPPED;
+				/* Update thread state in fork tree */
+				mcs_rwlock_writer_lock(&proc->update_lock, &lock);	
+				proc->group_exit_status = SIGSTOP;
 
-					proc->status = PS_DELAY_STOPPED;
-					thread->status = PS_STOPPED;
-					mcs_rwlock_writer_unlock(
-						     &proc->update_lock, &lock);
+				/* Reap and set new signal_flags */
+				proc->signal_flags = SIGNAL_STOP_STOPPED;
 
-					do_kill(thread,
-						thread->proc->parent->pid, -1,
-						SIGCHLD, &info, 0);
-				}
+				proc->status = PS_STOPPED;
+				thread->status = PS_STOPPED;
+				mcs_rwlock_writer_unlock(&proc->update_lock, &lock);	
+
+				/* Wake up the parent who tried wait4 and sleeping */
+				waitq_wakeup(&proc->parent->waitpid_q);
+
+				dkprintf("do_signal(): pid: %d, tid: %d SIGSTOP, sleeping\n", 
+					proc->pid, thread->tid);
 				/* Sleep */
 				schedule();
 				dkprintf("SIGSTOP(): woken up\n");
@@ -908,28 +882,19 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 			break;
 		case SIGTRAP:
 			dkprintf("do_signal,SIGTRAP\n");
-			if (!(thread->ptrace & PT_TRACED)) {
+			if(!(proc->ptrace & PT_TRACED)) {
 				goto core;
 			}
 
 			/* Update thread state in fork tree */
+			mcs_rwlock_writer_lock(&proc->update_lock, &lock);	
 			thread->exit_status = SIGTRAP;
+			proc->status = PS_TRACED;
 			thread->status = PS_TRACED;
-			if (thread == proc->main_thread) {
-				mcs_rwlock_writer_lock(&proc->update_lock,
-						       &lock);
-				proc->group_exit_status = SIGTRAP;
-				proc->status = PS_DELAY_TRACED;
-				mcs_rwlock_writer_unlock(&proc->update_lock,
-							 &lock);
-				do_kill(thread, thread->proc->parent->pid, -1,
-					SIGCHLD, &info, 0);
-			}
-			else {
-				do_kill(thread, thread->report_proc->pid, -1,
-					SIGCHLD, &info, 0);
-				waitq_wakeup(&thread->report_proc->waitpid_q);
-			}
+			mcs_rwlock_writer_unlock(&proc->update_lock, &lock);	
+
+			/* Wake up the parent who tried wait4 and sleeping */
+			waitq_wakeup(&thread->proc->parent->waitpid_q);
 
 			/* Sleep */
 			dkprintf("do_signal,SIGTRAP,sleeping\n");
@@ -944,7 +909,7 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 			info._sifields._sigchld.si_pid = proc->pid;
 			info._sifields._sigchld.si_status = 0x0000ffff;
 			do_kill(cpu_local_var(current), proc->parent->pid, -1, SIGCHLD, &info, 0);
-			proc->main_thread->signal_flags = SIGNAL_STOP_CONTINUED;
+			proc->signal_flags = SIGNAL_STOP_CONTINUED;
 			proc->status = PS_RUNNING;
 			dkprintf("do_signal,SIGCONT,do nothing\n");
 			break;
@@ -973,8 +938,6 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 			break;
 		}
 	}
-out:
-	return restart;
 }
 
 static struct sig_pending *
@@ -994,12 +957,10 @@ getsigpending(struct thread *thread, int delflag){
 	lock = &thread->sigcommon->lock;
 	head = &thread->sigcommon->sigpending;
 	for(;;) {
-		if (delflag) {
+		if (delflag)
 			mcs_rwlock_writer_lock(lock, &mcs_rw_node);
-		}
-		else {
+		else 
 			mcs_rwlock_reader_lock(lock, &mcs_rw_node);
-		}
 
 		list_for_each_entry_safe(pending, next, head, list){
 			for(x = pending->sigmask.__val[0], sig = 0; x; sig++, x >>= 1);
@@ -1012,23 +973,19 @@ getsigpending(struct thread *thread, int delflag){
 					if(delflag) 
 						list_del(&pending->list);
 
-					if (delflag) {
+					if (delflag)
 						mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
-					}
-					else {
+					else 
 						mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
-					}
 					return pending;
 				}
 			}
 		}
 
-		if (delflag) {
+		if (delflag)
 			mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
-		}
-		else {
+		else 
 			mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
-		}
 
 		if(lock == &thread->sigpendinglock)
 			return NULL;
@@ -1043,11 +1000,6 @@ getsigpending(struct thread *thread, int delflag){
 struct sig_pending *
 hassigpending(struct thread *thread)
 {
-	if (list_empty(&thread->sigpending) &&
-	    list_empty(&thread->sigcommon->sigpending)) {
-		return NULL;
-	}
-
 	return getsigpending(thread, 0);
 }
 
@@ -1065,12 +1017,6 @@ void save_syscall_return_value(int num, unsigned long rc)
 	return;
 }
 
-/** \brief check arrived signals and processing
- *
- * @param rc    return value of syscall
- * @param regs0 context
- * @param num   syscall number (-1: Not called on exiting system call)
- */
 void
 check_signal(unsigned long rc, void *regs0, int num)
 {
@@ -1104,11 +1050,6 @@ check_signal(unsigned long rc, void *regs0, int num)
 		goto out;
 	}
 
-	if (list_empty(&thread->sigpending) &&
-	    list_empty(&thread->sigcommon->sigpending)) {
-		goto out;
-	}
-
 	for(;;){
 		pending = getsigpending(thread, 1);
 		if(!pending) {
@@ -1116,9 +1057,7 @@ check_signal(unsigned long rc, void *regs0, int num)
 			goto out;
 		}
 
-		if (do_signal(rc, regs, thread, pending, num)) {
-			num = -1;
-		}
+		do_signal(rc, regs, thread, pending, num);
 	}
 
 out:
@@ -1198,7 +1137,7 @@ check_sig_pending_thread(struct thread *thread)
 }
 
 void
-check_sig_pending(void)
+check_sig_pending()
 {
 	struct thread *thread;
 	struct cpu_local_var *v;
@@ -1219,7 +1158,7 @@ repeat:
 			continue;
 		}
 
-		if (thread->proc->group_exit_status & 0x0000000100000000L) {
+		if (thread->proc->exit_status & 0x0000000100000000L) {
 			continue;
 		}
 
@@ -1428,8 +1367,7 @@ done:
 		return 0;
 	}
 
-	/* Forward signal to Linux by interrupt_syscall mechanism */
-	if (tthread->uti_state == UTI_STATE_RUNNING_IN_LINUX) {
+	if (tthread->thread_offloaded) {
 		if (!tthread->proc->nohost) {
 			interrupt_syscall(tthread, sig);
 		}
@@ -1446,10 +1384,10 @@ done:
 	   in check_signal */
 	rc = 0;
 	k = tthread->sigcommon->action + sig - 1;
-	if ((sig != SIGKILL && (tthread->ptrace & PT_TRACED)) ||
-	    (k->sa.sa_handler != (void *)1 &&
-	     (k->sa.sa_handler != NULL ||
-	      (sig != SIGCHLD && sig != SIGURG)))) {
+	if((sig != SIGKILL && (tproc->ptrace & PT_TRACED)) ||
+			(k->sa.sa_handler != (void *)1 &&
+			 (k->sa.sa_handler != NULL ||
+			  (sig != SIGCHLD && sig != SIGURG)))){
 		struct sig_pending *pending = NULL;
 		if (sig < 33) { // SIGRTMIN - SIGRTMAX
 			list_for_each_entry(pending, head, list){
@@ -1533,7 +1471,7 @@ set_signal(int sig, void *regs0, siginfo_t *info)
 
 SYSCALL_DECLARE(mmap)
 {
-	const unsigned int supported_flags = 0
+	const int supported_flags = 0
 		| MAP_SHARED		// 01
 		| MAP_PRIVATE		// 02
 		| MAP_FIXED		// 10
@@ -1541,7 +1479,7 @@ SYSCALL_DECLARE(mmap)
 		| MAP_LOCKED		// 2000
 		| MAP_POPULATE		// 8000
 		| MAP_HUGETLB		// 00040000
-		| (0x3FU << MAP_HUGE_SHIFT) // FC000000
+		| (0x3F << MAP_HUGE_SHIFT) // FC000000
 		;
 	const int ignored_flags = 0
 #ifdef	USE_NOCACHE_MMAP
@@ -1560,7 +1498,7 @@ SYSCALL_DECLARE(mmap)
 		| MAP_NONBLOCK		// 00010000
 		;
 
-	const uintptr_t addr0 = ihk_mc_syscall_arg0(ctx);
+	const intptr_t addr0 = ihk_mc_syscall_arg0(ctx);
 	const size_t len0 = ihk_mc_syscall_arg1(ctx);
 	const int prot = ihk_mc_syscall_arg2(ctx);
 	const int flags0 = ihk_mc_syscall_arg3(ctx);
@@ -1569,7 +1507,7 @@ SYSCALL_DECLARE(mmap)
 	struct thread *thread = cpu_local_var(current);
 	struct vm_regions *region = &thread->vm->region;
 	int error;
-	uintptr_t addr = 0;
+	intptr_t addr = 0;
 	size_t len;
 	int flags = flags0;
 	size_t pgsize;
@@ -1632,9 +1570,8 @@ SYSCALL_DECLARE(mmap)
 		goto out;
 	}
 
-	if (addr < region->user_start
-			|| region->user_end <= addr
-			|| len > (region->user_end - region->user_start)) {
+	if ((flags & MAP_FIXED) && ((addr < region->user_start)
+			|| (region->user_end <= addr))) {
 		ekprintf("sys_mmap(%lx,%lx,%x,%x,%x,%lx):ENOMEM\n",
 				addr0, len0, prot, flags0, fd, off0);
 		error = -ENOMEM;
@@ -1761,11 +1698,6 @@ SYSCALL_DECLARE(arch_prctl)
 	                     ihk_mc_syscall_arg1(ctx));
 }
 
-SYSCALL_DECLARE(time)
-{
-	return time();
-}
-
 static int vdso_get_vdso_info(void)
 {
 	int error;
@@ -2148,7 +2080,7 @@ int do_process_vm_read_writev(int pid,
 
 	range = lookup_process_memory_range(lthread->vm, 
 			(uintptr_t)local_iov, 
-			(uintptr_t)(local_iov + liovcnt));
+			(uintptr_t)(local_iov + liovcnt * sizeof(struct iovec)));
 
 	if (!range) {
 		ret = -EFAULT; 
@@ -2157,7 +2089,7 @@ int do_process_vm_read_writev(int pid,
 
 	range = lookup_process_memory_range(lthread->vm, 
 			(uintptr_t)remote_iov, 
-			(uintptr_t)(remote_iov + riovcnt));
+			(uintptr_t)(remote_iov + riovcnt * sizeof(struct iovec)));
 
 	if (!range) {
 		ret = -EFAULT; 
@@ -2433,6 +2365,8 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 			case 0:
 				memcpy(mpsr->virt_addr, mpsr->user_virt_addr,
 						sizeof(void *) * count);
+				memcpy(mpsr->status, mpsr->user_status,
+						sizeof(int) * count);
 				memcpy(mpsr->nodes, mpsr->user_nodes,
 						sizeof(int) * count);
 				memset(mpsr->ptep, 0, sizeof(pte_t) * count);
@@ -2452,38 +2386,41 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 			case 0:
 				memcpy(mpsr->virt_addr, mpsr->user_virt_addr,
 						sizeof(void *) * count);
+				memcpy(mpsr->status, mpsr->user_status,
+						sizeof(int) * count);
+			case 1:
 				memcpy(mpsr->nodes, mpsr->user_nodes,
 						sizeof(int) * count);
-				mpsr->nodes_ready = 1;
-				break;
-			case 1:
 				memset(mpsr->ptep, 0, sizeof(pte_t) * count);
 				memset(mpsr->status, 0, sizeof(int) * count);
 				memset(mpsr->nr_pages, 0, sizeof(int) * count);
 				memset(mpsr->dst_phys, 0,
 						sizeof(unsigned long) * count);
+				mpsr->nodes_ready = 1;
 				break;
 
 			default:
 				break;
 		}
 	}
-	else if (nr_cpus >= 4 && nr_cpus < 7) {
+	else if (nr_cpus >= 4 && nr_cpus < 8) {
 		switch (cpu_index) {
 			case 0:
 				memcpy(mpsr->virt_addr, mpsr->user_virt_addr,
 						sizeof(void *) * count);
 				break;
 			case 1:
+				memcpy(mpsr->status, mpsr->user_status,
+						sizeof(int) * count);
+				break;
+			case 2:
 				memcpy(mpsr->nodes, mpsr->user_nodes,
 						sizeof(int) * count);
 				mpsr->nodes_ready = 1;
 				break;
-			case 2:
+			case 3:
 				memset(mpsr->ptep, 0, sizeof(pte_t) * count);
 				memset(mpsr->status, 0, sizeof(int) * count);
-				break;
-			case 3:
 				memset(mpsr->nr_pages, 0, sizeof(int) * count);
 				memset(mpsr->dst_phys, 0,
 						sizeof(unsigned long) * count);
@@ -2493,7 +2430,7 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 				break;
 		}
 	}
-	else {
+	else if (nr_cpus >= 8) {
 		switch (cpu_index) {
 			case 0:
 				memcpy(mpsr->virt_addr, mpsr->user_virt_addr,
@@ -2505,23 +2442,28 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 						sizeof(void *) * (count / 2));
 				break;
 			case 2:
+				memcpy(mpsr->status, mpsr->user_status,
+						sizeof(int) * count);
+				break;
+			case 3:
 				memcpy(mpsr->nodes, mpsr->user_nodes,
 						sizeof(int) * count);
 				mpsr->nodes_ready = 1;
 				break;
-			case 3:
+			case 4:
 				memset(mpsr->ptep, 0, sizeof(pte_t) * count);
 				break;
-			case 4:
+			case 5:
 				memset(mpsr->status, 0, sizeof(int) * count);
 				break;
-			case 5:
+			case 6:
 				memset(mpsr->nr_pages, 0, sizeof(int) * count);
 				break;
-			case 6:
+			case 7:
 				memset(mpsr->dst_phys, 0,
 						sizeof(unsigned long) * count);
 				break;
+
 			default:
 				break;
 		}
@@ -2729,19 +2671,11 @@ out:
 
 time_t time(void) {
 	struct syscall_request sreq IHK_DMA_ALIGN;
-	struct timespec ats;
-	time_t ret = 0;
-
-	if (gettime_local_support) {
-		calculate_time_from_tsc(&ats);
-		ret = ats.tv_sec;
-	}
-	else {
-		sreq.number = __NR_time;
-		sreq.args[0] = (uintptr_t)NULL;
-		ret = (time_t)do_syscall(&sreq, ihk_mc_get_processor_id());
-	}
-
+	struct thread *thread = cpu_local_var(current);
+	time_t ret;
+	sreq.number = __NR_time;
+	sreq.args[0] = (uintptr_t)NULL;
+	ret = (time_t)do_syscall(&sreq, ihk_mc_get_processor_id(), thread->proc->pid);
 	return ret;
 }
 

arch/x86_64/kernel/vsyscall.c

@@ -31,6 +31,51 @@ struct tod_data_s tod_data
 	.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;

arch/x86_64/tools/mcoverlay-create-smp-x86.sh.in

@@ -43,8 +43,7 @@ error_exit() {
 		;;
 	esac
 
-	# Retun -EINVAL
-	exit -22
+	exit 1
 }
 fi
 
@@ -145,5 +144,3 @@ for cpuid in `find /sys/bus/cpu/devices/* -maxdepth 0 -name "cpu[0123456789]*" -
 	rm -rf /tmp/mcos/mcos0_sys/bus/cpu/devices/$cpuid
     fi
 done
-
-exit 0

arch/x86_64/tools/mcoverlay-destroy-smp-x86.sh.in

@@ -8,9 +8,6 @@ if grep mcoverlay /proc/modules &>/dev/null; then
     if [ -e /tmp/mcos ]; then rm -rf /tmp/mcos; fi
     if ! rmmod mcoverlay 2>/dev/null; then
 	echo "error: removing mcoverlay" >&2
-	# Return -EINVAL
-	exit -22
+	exit 1
     fi
 fi
-
-exit 0

arch/x86_64/tools/mcreboot-smp-x86.sh.in

@@ -12,7 +12,6 @@
 # the same set of resources as it used previously.
 # Note that the script does not output anything unless an error occurs.
 
-ret=1
 prefix="@prefix@"
 BINDIR="${prefix}/bin"
 SBINDIR="${prefix}/sbin"
@@ -50,7 +49,7 @@ umask_old=`umask`
 idle_halt=""
 allow_oversubscribe=""
 
-while getopts stk:c:m:o:f:r:q:i:d:e:hO OPT
+while getopts :stk:c:m:o:f:r:q:i:d:e:hO OPT
 do
 	case ${OPT} in
 	f)	facility=${OPTARG}
@@ -81,11 +80,14 @@ do
 		;;
 	O) allow_oversubscribe="allow_oversubscribe"
 		;;
-	\?) exit 1
-		;;
+	*)  echo "invalid option -${OPT}" >&2
+		exit 1
 	esac
 done
 
+redirect_kmsg=0
+turbo="turbo"
+
 # Start ihkmond
 pid=`pidof ihkmond`
 if [ "${pid}" != "" ]; then
@@ -208,8 +210,7 @@ error_exit() {
 		;;
 	esac
 
-	# Propagate exit status if any
-	exit $ret
+	exit 1
 }
 
 ihk_ikc_irq_core=0
@@ -235,7 +236,7 @@ if [ "${ENABLE_MCOVERLAYFS}" == "yes" ]; then
 			enable_mcoverlay="yes"
 		fi
 	else
-		if [ ${linux_version_code} -eq 199168 -a ${rhel_release} -ge 327 -a ${rhel_release} -le 862 ]; then
+		if [ ${linux_version_code} -eq 199168 -a ${rhel_release} -ge 327 -a ${rhel_release} -le 693 ]; then
 			enable_mcoverlay="yes"
 		fi
 		if [ ${linux_version_code} -ge 262144 -a ${linux_version_code} -lt 262400 ]; then
@@ -260,11 +261,7 @@ fi
 
 # Remove mcoverlay if loaded
 if [ "$enable_mcoverlay" == "yes" ]; then
-    ${SBINDIR}/mcoverlay-destroy.sh
-    ret=$?
-    if [ $ret -ne 0 ]; then
-	error_exit "initial"
-    fi
+    . ${SBINDIR}/mcoverlay-destroy.sh
 fi
 
 # Stop irqbalance
@@ -308,16 +305,25 @@ if ! grep -E 'ihk\s' /proc/modules &>/dev/null; then
 	fi
 fi
 
-# Increase swappiness so that we have better chance to allocate memory for IHK
-echo 100 > /proc/sys/vm/swappiness
+# Copy modules under /tmp to avoid loading from shared FS
+if mkdir -p /tmp/mcos-kmod; then
+	cp ${KMODDIR}/* /tmp/mcos-kmod/
+	KMODDIR="/tmp/mcos-kmod/"
+fi
 
-# Drop Linux caches to free memory
-sync && echo 3 > /proc/sys/vm/drop_caches
+# Fujitsu drops caches for us in between jobs so don't do it on OFP
+if [ "`hostname | grep "c[0-9][0-9][0-9][0-9].ofp"`" == "" ]; then
+	# Increase swappiness so that we have better chance to allocate memory for IHK
+	echo 100 > /proc/sys/vm/swappiness
 
-# Merge free memory areas into large, physically contigous ones
-echo 1 > /proc/sys/vm/compact_memory 2>/dev/null
+	# Drop Linux caches to free memory
+	sync && echo 3 > /proc/sys/vm/drop_caches
 
-sync
+	# Merge free memory areas into large, physically contigous ones
+	echo 1 > /proc/sys/vm/compact_memory 2>/dev/null
+
+	sync
+fi
 
 # Load IHK-SMP if not loaded and reserve CPUs and memory
 if ! grep ihk_smp_@ARCH@ /proc/modules &>/dev/null; then
@@ -338,41 +344,41 @@ if ! grep ihk_smp_@ARCH@ /proc/modules &>/dev/null; then
 		error_exit "ihk_loaded"
 	fi
 
-	# Offline-reonline RAM (special case for OFP SNC-4 flat mode)
-	if [ "`hostname | grep "c[0-9][0-9][0-9][0-9].ofp"`" != "" ] && [ "`cat /sys/devices/system/node/online`" == "0-7" ]; then
-		for i in  0 1 2 3; do
-			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
-				echo 0 > $f 2>&1 > /dev/null;
-			done
-			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
-				echo 1 > $f 2>&1 > /dev/null;
-			done
-		done
-		for i in 4 5 6 7; do
-			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
-				echo 0 > $f 2>&1 > /dev/null;
-			done
-		done
-		for i in 4 5 6 7; do
-			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
-				echo 1 > $f 2>&1 > /dev/null;
-			done
-		done
-	fi
-
-	# Offline-reonline RAM (special case for OFP Quadrant flat mode)
-	if [ "`hostname | grep "c[0-9][0-9][0-9][0-9].ofp"`" != "" ] && [ "`cat /sys/devices/system/node/online`" == "0-1" ]; then
-		for i in 1; do
-			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
-				echo 0 > $f 2>&1 > /dev/null;
-			done
-		done
-		for i in 1; do
-			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
-				echo 1 > $f 2>&1 > /dev/null;
-			done
-		done
-	fi
+# 	# Offline-reonline RAM (special case for OFP SNC-4 flat mode)
+# 	if [ "`hostname | grep "c[0-9][0-9][0-9][0-9].ofp"`" != "" ] && [ "`cat /sys/devices/system/node/online`" == "0-7" ]; then
+# 		for i in  0 1 2 3; do
+# 			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
+# 				echo 0 | tee $f 2>/dev/null 1>/dev/null
+# 			done
+# 			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
+# 				echo 1 | tee $f 2>/dev/null 1>/dev/null
+# 			done
+# 		done
+# 		for i in 4 5 6 7; do
+# 			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
+# 				echo 0 | tee $f 2>/dev/null 1>/dev/null
+# 			done
+# 		done
+# 		for i in 4 5 6 7; do
+# 			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
+# 				echo 1 | tee $f 2>/dev/null 1>/dev/null
+# 			done
+# 		done
+# 	fi
+#
+# 	# Offline-reonline RAM (special case for OFP Quadrant flat mode)
+# 	if [ "`hostname | grep "c[0-9][0-9][0-9][0-9].ofp"`" != "" ] && [ "`cat /sys/devices/system/node/online`" == "0-1" ]; then
+# 		for i in 1; do
+# 			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
+# 				echo 0 | tee $f 2>/dev/null 1>/dev/null
+# 			done
+# 		done
+# 		for i in 1; do
+# 			find /sys/devices/system/node/node$i/memory*/ -name "online" | while read f; do
+# 				echo 1 | tee $f 2>/dev/null 1>/dev/null
+# 			done
+# 		done
+# 	fi
 
 	if ! ${SBINDIR}/ihkconfig 0 reserve mem ${mem}; then
 		echo "error: reserving memory" >&2
@@ -467,11 +473,7 @@ fi
 
 # Overlay /proc, /sys with McKernel specific contents
 if [ "$enable_mcoverlay" == "yes" ]; then
-    ${SBINDIR}/mcoverlay-create.sh
-    ret=$?
-    if [ $ret -ne 0 ]; then
-	error_exit "os_created"
-    fi
+    . ${SBINDIR}/mcoverlay-create.sh
 fi
 
 # Start irqbalance with CPUs and IRQ for McKernel banned

arch/x86_64/tools/mcstop+release-smp-x86.sh.in

@@ -100,13 +100,7 @@ if grep mcctrl /proc/modules &>/dev/null; then
 fi
 
 # Remove mcoverlay if loaded
-${SBINDIR}/mcoverlay-destroy.sh
-ret=$?
-if [ $ret -ne 0 ]; then
-    echo "error: mcoverlay-destroy.sh" >&2
-    exit $ret
-fi
-
+. ${SBINDIR}/mcoverlay-destroy.sh
 
 # Remove SMP module
 if grep ihk_smp_@ARCH@ /proc/modules &>/dev/null; then

arch/x86_64/tools/mpimcexec.1in

@@ -0,0 +1,60 @@
+.\" Man page for mpimcexec
+.\"
+
+.TH MPIMCEXEC 1 "@MCKERNEL_RELEASE_DATE@" "Version @MCKERNEL_VERSION@" MCKERNEL @MCKERNEL_VERSION@"
+.SH NAME
+mpimcexec \- run an MPI application on McKernel
+.\"
+
+.\" ----------------------------  SYNOPSIS ----------------------------
+.SH SYNOPSIS
+.B mpimcexec \fR [\fIoptions\fR] \fI<command>\fR
+
+.\" ----------------------------  DESCRIPTION ----------------------------
+.SH DESCRIPTION
+mpimcexec is a wrapper script for running MPI applications on McKernel.
+It internally calls mpiexec to spawn mcexec on compute nodes, which in
+turn runs \fI<command>\fR on McKernel. mpimcexec specifies a number of
+mcexec arguments that enable high performance execution.
+
+.\" ----------------------------  OPTIONS ----------------------------
+.SH OPTIONS
+
+.TP
+.B	-ppn N, --ppn N, --ranks-per-node N
+Specify the number of MPI ranks per node. This argument is required.
+.TP
+.B	-n N, --n N, --ranks N
+Specify the number of total MPI ranks.
+e.g.,
+   $ mpimcexec -n 32 -ppn 4 ./a.out
+.br
+In the above example, 32 MPI processes are invoked
+on eight compute nodes each of which has four processes.
+.TP
+.B	--nodes N
+Specify the number of compute nodes.
+By default, all nodes, specified by "PJM --mpi proc" option, are used.
+.TP
+.B	--env, -env 
+Pass an additional environment variable
+.TP
+.B	-m N, --numa N
+Specify preferred NUMA node.
+.TP
+.B	-h <file name>, ---hostfile <file name>
+Specify a host file for MPI.
+.TP
+.B	--help
+Show help message.
+
+.PP
+.\" ----------------------------  SEE ALSO ----------------------------
+.SH SEE ALSO
+\fBmcexec\fR (1), \fBmpiexec\fR (1)
+
+.\" ----------------------------  AUTHORS ----------------------------
+.SH AUTHORS
+Copyright (C) 2018 McKernel Development Team, RIKEN, Japan
+
+

arch/x86_64/tools/mpimcexec.in

@@ -0,0 +1,147 @@
+#!/bin/bash
+#
+# OFP McKernel MPI wrapper script
+# author: Balazs Gerofi <bgerofi@riken.jp>
+#      Copyright (C) 2018  RIKEN R-CCS
+#
+
+prefix="@prefix@"
+BINDIR="${prefix}/bin"
+
+if [ "${BASH_VERSINFO[0]}" -lt 4 ]; then
+	echo "You need at least bash-4.0 to run this script." >&2
+	exit 1
+fi
+
+RANKS=""
+NODES=""
+PPN=""
+MPI_ENV=""
+COMMAND=""
+NUMA=""
+HOSTFILE=""
+
+if [ ! -z "${PJM_PROC_BY_NODE}" ]; then
+	PPN=${PJM_PROC_BY_NODE}
+elif [ ! -z "${MPI_LOCALNRANKS}" ]; then
+	PPN=${MPI_LOCALNRANKS}
+fi
+
+help_exit() {
+	echo ""
+	echo "Spawn an McKernel MPI job on Oakforest-PACS."
+	echo "usage: `basename $0` -ppn ranks_per_node [--nodes nodes] [-n ranks] [--env additional_environment]... command"
+	echo ""
+	echo "   -ppn | --ppn | --ranks-per-node         Number of MPI ranks per node (required)"
+	echo "   -n | --n | --ranks                      Total number of MPI ranks in the job"
+	echo "   --nodes                                 Number of nodes to be used"
+	echo "   --env | -env                            Pass an additional environment variable"
+	echo "   -m | --numa                             Preferred NUMA node(s)"
+	echo "   -h | --hostfile                         Host file for MPI"
+	echo "   --help                                  Show help message"
+	exit 1
+}
+
+# Parse options
+while true; do
+	case $1 in
+		-ppn | --ppn | --ranks-per-node )
+			if [ $# -lt 2 ]; then
+			    echo "error: needs an interger value for -ppn, --ppn, or --ranks-per-node option"
+			    help_exit
+			fi
+			PPN=$2
+			shift 2
+			;;
+		-n | --n | --ranks )
+			if [ $# -lt 2 ]; then
+			    echo "error: needs an interger value for -n, --n, or --ranks option"
+			    help_exit
+			fi
+			RANKS=$2
+			shift 2
+			;;
+		-m | --numa )
+			if [ $# -lt 2 ]; then
+			    echo "error: needs an interger value for -m or --numa option"
+			    help_exit
+			fi
+			NUMA="-m $2"
+			shift 2
+			;;
+		--nodes )
+			if [ $# -lt 2 ]; then
+			    echo "error: needs an interger value for --nodes option"
+			    help_exit
+			fi
+			NODES=$2
+			shift 2
+			;;
+		--env | -env )
+			if [ $# -lt 2 ]; then
+			    echo "error: needs an environment variable name for -env or --env option"
+			    help_exit
+			fi
+			if [ -z "`echo $2 | grep I_MPI_PIN`" ]; then
+				MPI_ENV=`echo "${MPI_ENV} -env $2" | xargs`
+			fi
+			shift 2
+			;;
+		-h | --hostfile )
+			if [ $# -lt 2 ]; then
+			    echo "error: needs a file name for -h or --hostfile option"
+			    help_exit
+			fi
+			HOSTFILE="-hostfile $2"
+			shift 2
+			;;
+		--help )
+			help_exit
+			;;
+		* )
+			COMMAND=$@
+			break
+			;;
+	esac
+done
+
+if [ -z ${PPN} ]; then
+	echo "error: please specify the number of ranks per node"
+	help_exit
+fi
+
+# Unless explicitly specified, use Fujitsu inherited value
+if [ -z ${NODES} ]; then
+	NODES=${PJM_VNODES}
+fi
+
+if [ -z ${RANKS} ] && [ -z ${NODES} ]; then
+	echo "error: please specify the total number of ranks or the number of nodes"
+	help_exit
+fi
+
+if [ "x${COMMAND}" = "x" ]; then
+	echo "error: please specify command"
+	help_exit
+fi
+
+# Calculate total job size if not specified
+if [ -z ${RANKS} ]; then
+	let RANKS=(${PPN}*${NODES})
+fi
+
+# Support direct SSH when not executed from Fujitsu job system
+if [ -z ${PJM_VNODES} ]; then
+	HOSTFILE="-launcher-exec ssh ${HOSTFILE}"
+fi
+
+export I_MPI_PIN=off
+export PSM2_RCVTHREAD=0
+export HFI_NO_CPUAFFINITY=1
+export I_MPI_COLL_INTRANODE_SHM_THRESHOLD=4194304
+export PSM2_MQ_RNDV_HFI_WINDOW=4194304
+export PSM2_MQ_EAGER_SDMA_SZ=65536
+export PSM2_MQ_RNDV_HFI_THRESH=200000
+
+mpirun ${HOSTFILE} -n ${RANKS} -ppn ${PPN} ${MPI_ENV} ${BINDIR}/mcexec -n ${PPN} ${NUMA} --enable-hfi1 --mpol-threshold=1M --stack-premap=4M,4G --extend-heap-by=8M --disable-sched-yield --mpol-shm-premap ${COMMAND}
+

config.h.in

@@ -54,6 +54,48 @@
 /* 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_umount, or 0 if exported */
+#undef MCCTRL_KSYM_sys_umount
+
+/* Define to address of kernel symbol sys_unshare, or 0 if exported */
+#undef MCCTRL_KSYM_sys_unshare
+
+/* 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_spec, or 0 if exported */
+#undef MCCTRL_KSYM_vdso_spec
+
+/* Define to address of kernel symbol vdso_start, or 0 if exported */
+#undef MCCTRL_KSYM_vdso_start
+
+/* Define to address of kernel symbol walk_page_range, or 0 if exported */
+#undef MCCTRL_KSYM_walk_page_range
+
+/* Define to address of kernel symbol zap_page_range, or 0 if exported */
+#undef MCCTRL_KSYM_zap_page_range
+
 /* McKernel specific headers */
 #undef MCKERNEL_INCDIR
 
@@ -86,6 +128,3 @@
 
 /* Define to 1 if you have the ANSI C header files. */
 #undef STDC_HEADERS
-
-/* whether or not syscall_intercept library is linked */
-#undef WITH_SYSCALL_INTERCEPT

configure

@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.69 for mckernel 1.6.0.
+# Generated by GNU Autoconf 2.69 for mckernel 1.5.1-knl+hfi.
 #
 #
 # Copyright (C) 1992-1996, 1998-2012 Free Software Foundation, Inc.
@@ -577,8 +577,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='mckernel'
 PACKAGE_TARNAME='mckernel'
-PACKAGE_VERSION='1.6.0'
-PACKAGE_STRING='mckernel 1.6.0'
+PACKAGE_VERSION='1.5.1-knl+hfi'
+PACKAGE_STRING='mckernel 1.5.1-knl+hfi'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -628,12 +628,9 @@ IHK_RELEASE_DATE
 DCFA_VERSION
 MCKERNEL_VERSION
 IHK_VERSION
-WITH_SYSCALL_INTERCEPT
 ENABLE_QLMPI
 ENABLE_RUSAGE
 ENABLE_MCOVERLAYFS
-LDFLAGS_SYSCALL_INTERCEPT
-CPPFLAGS_SYSCALL_INTERCEPT
 MANDIR
 KERNDIR
 KMODDIR
@@ -705,9 +702,6 @@ enable_option_checking
 with_mpi
 with_mpi_include
 with_mpi_lib
-with_syscall_intercept
-with_syscall_intercept_include
-with_syscall_intercept_lib
 with_kernelsrc
 with_target
 with_system_map
@@ -1268,7 +1262,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-\`configure' configures mckernel 1.6.0 to adapt to many kinds of systems.
+\`configure' configures mckernel 1.5.1-knl+hfi to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1329,7 +1323,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of mckernel 1.6.0:";;
+     short | recursive ) echo "Configuration of mckernel 1.5.1-knl+hfi:";;
    esac
   cat <<\_ACEOF
 
@@ -1352,15 +1346,6 @@ Optional Packages:
   --with-mpi-include=PATH specify path where mpi include directory can be
                           found
   --with-mpi-lib=PATH     specify path where mpi lib directory can be found
-  --with-syscall_intercept=PATH
-                          specify path where syscall_intercept include
-                          directory and lib directory can be found
-  --with-syscall_intercept-include=PATH
-                          specify path where syscall_intercept include
-                          directory can be found
-  --with-syscall_intercept-lib=PATH
-                          specify path where syscall_intercept lib directory
-                          can be found
   --with-kernelsrc=path   Path to 'kernel src', default is
                           /lib/modules/uname_r/build
   --with-target={attached-mic | builtin-mic | builtin-x86 | smp-x86}
@@ -1446,7 +1431,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-mckernel configure 1.6.0
+mckernel configure 1.5.1-knl+hfi
 generated by GNU Autoconf 2.69
 
 Copyright (C) 2012 Free Software Foundation, Inc.
@@ -1744,7 +1729,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by mckernel $as_me 1.6.0, which was
+It was created by mckernel $as_me 1.5.1-knl+hfi, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   $ $0 $@
@@ -2097,13 +2082,11 @@ ac_compiler_gnu=$ac_cv_c_compiler_gnu
 
 
 
-
-
-IHK_VERSION=1.6.0
-MCKERNEL_VERSION=1.6.0
+IHK_VERSION=1.5.1-knl+hfi
+MCKERNEL_VERSION=1.5.1-knl+hfi
 DCFA_VERSION=DCFA_VERSION_m4
-IHK_RELEASE_DATE=2018-11-11
-MCKERNEL_RELEASE_DATE=2018-11-11
+IHK_RELEASE_DATE=2019-05-14
+MCKERNEL_RELEASE_DATE=2019-05-14
 DCFA_RELEASE_DATE=DCFA_RELEASE_DATE_m4
 
 
@@ -3530,195 +3513,6 @@ fi
 
 
 
-
-# Check whether --with-syscall_intercept was given.
-if test "${with_syscall_intercept+set}" = set; then :
-  withval=$with_syscall_intercept; case "$withval" in #(
-  yes|no|'') :
-    { $as_echo "$as_me:${as_lineno-$LINENO}: WARNING: --without-syscall_intercept=PATH expects a valid PATH" >&5
-$as_echo "$as_me: WARNING: --without-syscall_intercept=PATH expects a valid PATH" >&2;}
-                          with_syscall_intercept="" ;; #(
-  *) :
-     ;;
-esac
-else
-  with_syscall_intercept=
-fi
-
-
-# Check whether --with-syscall_intercept-include was given.
-if test "${with_syscall_intercept_include+set}" = set; then :
-  withval=$with_syscall_intercept_include; case "$withval" in #(
-  yes|no|'') :
-    { $as_echo "$as_me:${as_lineno-$LINENO}: WARNING: --without-syscall_intercept-include=PATH expects a valid PATH" >&5
-$as_echo "$as_me: WARNING: --without-syscall_intercept-include=PATH expects a valid PATH" >&2;}
-                          with_syscall_intercept_include="" ;; #(
-  *) :
-     ;;
-esac
-fi
-
-
-# Check whether --with-syscall_intercept-lib was given.
-if test "${with_syscall_intercept_lib+set}" = set; then :
-  withval=$with_syscall_intercept_lib; case "$withval" in #(
-  yes|no|'') :
-    { $as_echo "$as_me:${as_lineno-$LINENO}: WARNING: --without-syscall_intercept-lib=PATH expects a valid PATH" >&5
-$as_echo "$as_me: WARNING: --without-syscall_intercept-lib=PATH expects a valid PATH" >&2;}
-                          with_syscall_intercept_lib="" ;; #(
-  *) :
-     ;;
-esac
-fi
-
-
-    # The args have been sanitized into empty/non-empty values above.
-    # Now append -I/-L args to CPPFLAGS/LDFLAGS, with more specific options
-    # taking priority
-
-    if test -n "${with_syscall_intercept_include}"; then :
-
-
-	if echo "$CPPFLAGS_SYSCALL_INTERCEPT" | $FGREP -e "\<-I${with_syscall_intercept_include}\>" >/dev/null 2>&1; then :
-  echo "CPPFLAGS_SYSCALL_INTERCEPT(='$CPPFLAGS_SYSCALL_INTERCEPT') contains '-I${with_syscall_intercept_include}', not appending" >&5
-else
-  echo "CPPFLAGS_SYSCALL_INTERCEPT(='$CPPFLAGS_SYSCALL_INTERCEPT') does not contain '-I${with_syscall_intercept_include}', appending" >&5
-		CPPFLAGS_SYSCALL_INTERCEPT="$CPPFLAGS_SYSCALL_INTERCEPT -I${with_syscall_intercept_include}"
-
-fi
-
-else
-  if test -n "${with_syscall_intercept}"; then :
-
-
-	if echo "$CPPFLAGS_SYSCALL_INTERCEPT" | $FGREP -e "\<-I${with_syscall_intercept}/include\>" >/dev/null 2>&1; then :
-  echo "CPPFLAGS_SYSCALL_INTERCEPT(='$CPPFLAGS_SYSCALL_INTERCEPT') contains '-I${with_syscall_intercept}/include', not appending" >&5
-else
-  echo "CPPFLAGS_SYSCALL_INTERCEPT(='$CPPFLAGS_SYSCALL_INTERCEPT') does not contain '-I${with_syscall_intercept}/include', appending" >&5
-		CPPFLAGS_SYSCALL_INTERCEPT="$CPPFLAGS_SYSCALL_INTERCEPT -I${with_syscall_intercept}/include"
-
-fi
-
-fi
-fi
-
-    if test -n "${with_syscall_intercept_lib}"; then :
-
-
-	if echo "$LDFLAGS_SYSCALL_INTERCEPT" | $FGREP -e "\<-L${with_syscall_intercept_lib} -Wl,-rpath,${with_syscall_intercept_lib}\>" >/dev/null 2>&1; then :
-  echo "LDFLAGS_SYSCALL_INTERCEPT(='$LDFLAGS_SYSCALL_INTERCEPT') contains '-L${with_syscall_intercept_lib} -Wl,-rpath,${with_syscall_intercept_lib}', not appending" >&5
-else
-  echo "LDFLAGS_SYSCALL_INTERCEPT(='$LDFLAGS_SYSCALL_INTERCEPT') does not contain '-L${with_syscall_intercept_lib} -Wl,-rpath,${with_syscall_intercept_lib}', appending" >&5
-		LDFLAGS_SYSCALL_INTERCEPT="$LDFLAGS_SYSCALL_INTERCEPT -L${with_syscall_intercept_lib} -Wl,-rpath,${with_syscall_intercept_lib}"
-
-fi
-
-else
-  if test -n "${with_syscall_intercept}"; then :
-
-
-	if echo "$LDFLAGS_SYSCALL_INTERCEPT" | $FGREP -e "\<-L${with_syscall_intercept}/lib -Wl,-rpath,${with_syscall_intercept}/lib\>" >/dev/null 2>&1; then :
-  echo "LDFLAGS_SYSCALL_INTERCEPT(='$LDFLAGS_SYSCALL_INTERCEPT') contains '-L${with_syscall_intercept}/lib -Wl,-rpath,${with_syscall_intercept}/lib', not appending" >&5
-else
-  echo "LDFLAGS_SYSCALL_INTERCEPT(='$LDFLAGS_SYSCALL_INTERCEPT') does not contain '-L${with_syscall_intercept}/lib -Wl,-rpath,${with_syscall_intercept}/lib', appending" >&5
-		LDFLAGS_SYSCALL_INTERCEPT="$LDFLAGS_SYSCALL_INTERCEPT -L${with_syscall_intercept}/lib -Wl,-rpath,${with_syscall_intercept}/lib"
-
-fi
-
-                  if test -d "${with_syscall_intercept}/lib64"; then :
-
-
-	if echo "$LDFLAGS_SYSCALL_INTERCEPT" | $FGREP -e "\<-L${with_syscall_intercept}/lib64 -Wl,-rpath,${with_syscall_intercept}/lib64\>" >/dev/null 2>&1; then :
-  echo "LDFLAGS_SYSCALL_INTERCEPT(='$LDFLAGS_SYSCALL_INTERCEPT') contains '-L${with_syscall_intercept}/lib64 -Wl,-rpath,${with_syscall_intercept}/lib64', not appending" >&5
-else
-  echo "LDFLAGS_SYSCALL_INTERCEPT(='$LDFLAGS_SYSCALL_INTERCEPT') does not contain '-L${with_syscall_intercept}/lib64 -Wl,-rpath,${with_syscall_intercept}/lib64', appending" >&5
-		LDFLAGS_SYSCALL_INTERCEPT="$LDFLAGS_SYSCALL_INTERCEPT -L${with_syscall_intercept}/lib64 -Wl,-rpath,${with_syscall_intercept}/lib64"
-
-fi
-
-fi
-
-fi
-
-fi
-
-    if test -n "${with_syscall_intercept}" || test -n "${with_syscall_intercept_include}" || test -n "${with_syscall_intercept_lib}"; then :
-  WITH_SYSCALL_INTERCEPT=yes
-else
-  WITH_SYSCALL_INTERCEPT=no
-fi
-
-
-if test "x$WITH_SYSCALL_INTERCEPT" == "xno" ; then
-	{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for syscall_no_intercept in -lsyscall_intercept" >&5
-$as_echo_n "checking for syscall_no_intercept in -lsyscall_intercept... " >&6; }
-if ${ac_cv_lib_syscall_intercept_syscall_no_intercept+:} false; then :
-  $as_echo_n "(cached) " >&6
-else
-  ac_check_lib_save_LIBS=$LIBS
-LIBS="-lsyscall_intercept -lcapstone -ldl $LIBS"
-cat confdefs.h - <<_ACEOF >conftest.$ac_ext
-/* end confdefs.h.  */
-
-/* Override any GCC internal prototype to avoid an error.
-   Use char because int might match the return type of a GCC
-   builtin and then its argument prototype would still apply.  */
-#ifdef __cplusplus
-extern "C"
-#endif
-char syscall_no_intercept ();
-int
-main ()
-{
-return syscall_no_intercept ();
-  ;
-  return 0;
-}
-_ACEOF
-if ac_fn_c_try_link "$LINENO"; then :
-  ac_cv_lib_syscall_intercept_syscall_no_intercept=yes
-else
-  ac_cv_lib_syscall_intercept_syscall_no_intercept=no
-fi
-rm -f core conftest.err conftest.$ac_objext \
-    conftest$ac_exeext conftest.$ac_ext
-LIBS=$ac_check_lib_save_LIBS
-fi
-{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_syscall_intercept_syscall_no_intercept" >&5
-$as_echo "$ac_cv_lib_syscall_intercept_syscall_no_intercept" >&6; }
-if test "x$ac_cv_lib_syscall_intercept_syscall_no_intercept" = xyes; then :
-  syscall_intercept_lib_found=yes
-else
-  syscall_intercept_lib_found=no
-fi
-
-	if test "x$syscall_intercept_lib_found" != "xyes"; then :
-  { $as_echo "$as_me:${as_lineno-$LINENO}: libsyscall_intercept.so not found" >&5
-$as_echo "$as_me: libsyscall_intercept.so not found" >&6;}
-fi
-
-	ac_fn_c_check_header_mongrel "$LINENO" "libsyscall_intercept_hook_point.h" "ac_cv_header_libsyscall_intercept_hook_point_h" "$ac_includes_default"
-if test "x$ac_cv_header_libsyscall_intercept_hook_point_h" = xyes; then :
-  syscall_intercept_header_found=yes
-else
-  syscall_intercept_header_found=no
-fi
-
-
-	if test "x$syscall_intercept_header_found" != "xyes"; then :
-  { $as_echo "$as_me:${as_lineno-$LINENO}: libsyscall_intercept_hook_point.h not found" >&5
-$as_echo "$as_me: libsyscall_intercept_hook_point.h not found" >&6;}
-fi
-
-	if test "x$syscall_intercept_lib_found" == "xyes" && test "x$syscall_intercept_header_found" == "xyes"; then :
-  WITH_SYSCALL_INTERCEPT=yes
-else
-  WITH_SYSCALL_INTERCEPT=no
-fi
-fi
-
-
-
 # Check whether --with-kernelsrc was given.
 if test "${with_kernelsrc+set}" = set; then :
   withval=$with_kernelsrc; WITH_KERNELSRC=$withval
@@ -4602,6 +4396,399 @@ KDIR="$WITH_KERNELSRC"
 UNAME_R="$WITH_UNAME_R"
 TARGET="$WITH_TARGET"
 
+MCCTRL_LINUX_SYMTAB=""
+case "X$WITH_SYSTEM_MAP" in
+    Xyes | Xno | X)
+	MCCTRL_LINUX_SYMTAB=""
+	;;
+    *)
+	MCCTRL_LINUX_SYMTAB="$WITH_SYSTEM_MAP"
+	;;
+esac
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for System.map" >&5
+$as_echo_n "checking for System.map... " >&6; }
+if test -r "$MCCTRL_LINUX_SYMTAB"; then
+	MCCTRL_LINUX_SYMTAB="$MCCTRL_LINUX_SYMTAB"
+elif test -r "/boot/System.map-`uname -r`"; then
+	MCCTRL_LINUX_SYMTAB="/boot/System.map-`uname -r`"
+elif test -r "$KDIR/System.map"; then
+	MCCTRL_LINUX_SYMTAB="$KDIR/System.map"
+fi
+
+if test "$MCCTRL_LINUX_SYMTAB" == ""; then
+	as_fn_error $? "could not find" "$LINENO" 5
+fi
+
+if test -z "`eval cat $MCCTRL_LINUX_SYMTAB`"; then
+	as_fn_error $? "could not read System.map file, no read permission?" "$LINENO" 5
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $MCCTRL_LINUX_SYMTAB" >&5
+$as_echo "$MCCTRL_LINUX_SYMTAB" >&6; }
+
+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>"
+
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol sys_mount" >&5
+$as_echo_n "checking System.map for symbol sys_mount... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " sys_mount\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_sys_mount\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_sys_mount $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol sys_umount" >&5
+$as_echo_n "checking System.map for symbol sys_umount... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " sys_umount\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_sys_umount\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_sys_umount $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol sys_unshare" >&5
+$as_echo_n "checking System.map for symbol sys_unshare... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " sys_unshare\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_sys_unshare\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_sys_unshare $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol zap_page_range" >&5
+$as_echo_n "checking System.map for symbol zap_page_range... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " zap_page_range\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_zap_page_range\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_zap_page_range $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol vdso_image_64" >&5
+$as_echo_n "checking System.map for symbol vdso_image_64... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " vdso_image_64\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_vdso_image_64\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_vdso_image_64 $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol vdso_start" >&5
+$as_echo_n "checking System.map for symbol vdso_start... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " vdso_start\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_vdso_start\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_vdso_start $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol vdso_end" >&5
+$as_echo_n "checking System.map for symbol vdso_end... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " vdso_end\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_vdso_end\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_vdso_end $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol vdso_pages" >&5
+$as_echo_n "checking System.map for symbol vdso_pages... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " vdso_pages\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_vdso_pages\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_vdso_pages $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol __vvar_page" >&5
+$as_echo_n "checking System.map for symbol __vvar_page... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __vvar_page\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab___vvar_page\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM___vvar_page $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol hpet_address" >&5
+$as_echo_n "checking System.map for symbol hpet_address... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " hpet_address\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_hpet_address\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_hpet_address $mcctrl_addr
+_ACEOF
+
+  fi
+
+# POSTK_DEBUG_ARCH_DEP_50, add:find kernel symbol.
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol vdso_spec" >&5
+$as_echo_n "checking System.map for symbol vdso_spec... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " vdso_spec\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_vdso_spec\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_vdso_spec $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol hv_clock" >&5
+$as_echo_n "checking System.map for symbol hv_clock... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " hv_clock\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_hv_clock\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_hv_clock $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol sys_readlink" >&5
+$as_echo_n "checking System.map for symbol sys_readlink... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " sys_readlink\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_sys_readlink\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_sys_readlink $mcctrl_addr
+_ACEOF
+
+  fi
+
+
+  { $as_echo "$as_me:${as_lineno-$LINENO}: checking System.map for symbol walk_page_range" >&5
+$as_echo_n "checking System.map for symbol walk_page_range... " >&6; }
+  mcctrl_addr=`eval $MCCTRL_LINUX_SYMTAB_CMD | grep " walk_page_range\$" | cut -d\  -f1`
+  if test -z $mcctrl_addr; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: not found" >&5
+$as_echo "not found" >&6; }
+  else
+    mcctrl_result=$mcctrl_addr
+    mcctrl_addr="0x$mcctrl_addr"
+
+      if `eval $MCCTRL_LINUX_SYMTAB_CMD | grep " __ksymtab_walk_page_range\$" >/dev/null`; then
+        mcctrl_result="exported"
+        mcctrl_addr="0"
+      fi
+
+    { $as_echo "$as_me:${as_lineno-$LINENO}: result: $mcctrl_result" >&5
+$as_echo "$mcctrl_result" >&6; }
+
+cat >>confdefs.h <<_ACEOF
+#define MCCTRL_KSYM_walk_page_range $mcctrl_addr
+_ACEOF
+
+  fi
+
+
 case $ENABLE_MEMDUMP in
 	yes|no|auto)
 		;;
@@ -4799,17 +4986,6 @@ else
 $as_echo "$as_me: perf is disabled" >&6;}
 fi
 
-if test "x$WITH_SYSCALL_INTERCEPT" = "xyes" ; then
-
-$as_echo "#define WITH_SYSCALL_INTERCEPT 1" >>confdefs.h
-
-	{ $as_echo "$as_me:${as_lineno-$LINENO}: syscall_intercept library is linked" >&5
-$as_echo "$as_me: syscall_intercept library is linked" >&6;}
-else
-	{ $as_echo "$as_me:${as_lineno-$LINENO}: syscall_intercept library isn't linked" >&5
-$as_echo "$as_me: syscall_intercept library isn't linked" >&6;}
-fi
-
 if test "x$MCKERNEL_INCDIR" != "x" ; then
 
 cat >>confdefs.h <<_ACEOF
@@ -4876,9 +5052,6 @@ fi
 
 
 
-
-
-
 
 
 
@@ -4887,14 +5060,9 @@ ac_config_headers="$ac_config_headers config.h"
 
 # POSTK_DEBUG_ARCH_DEP_37
 # AC_CONFIG_FILES arch dependfiles separate
-ac_config_files="$ac_config_files Makefile executer/user/Makefile executer/user/mcexec.1:executer/user/mcexec.1in executer/user/vmcore2mckdump executer/user/arch/$ARCH/Makefile executer/user/arch/x86_64/Makefile executer/kernel/mcctrl/Makefile executer/kernel/mcctrl/arch/$ARCH/Makefile executer/kernel/mcoverlayfs/Makefile executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile executer/kernel/mcoverlayfs/linux-4.0.9/Makefile executer/kernel/mcoverlayfs/linux-4.6.7/Makefile executer/include/qlmpilib.h kernel/Makefile kernel/Makefile.build kernel/include/swapfmt.h arch/x86_64/tools/mcreboot-attached-mic.sh arch/x86_64/tools/mcshutdown-attached-mic.sh arch/x86_64/tools/mcreboot-builtin-x86.sh arch/x86_64/tools/mcreboot-smp-x86.sh arch/x86_64/tools/mcstop+release-smp-x86.sh arch/x86_64/tools/mcoverlay-destroy-smp-x86.sh arch/x86_64/tools/mcoverlay-create-smp-x86.sh arch/x86_64/tools/eclair-dump-backtrace.exp arch/x86_64/tools/mcshutdown-builtin-x86.sh arch/x86_64/tools/mcreboot.1:arch/x86_64/tools/mcreboot.1in arch/x86_64/tools/irqbalance_mck.service arch/x86_64/tools/irqbalance_mck.in tools/mcstat/mcstat.1:tools/mcstat/mcstat.1in tools/mcstat/Makefile"
+ac_config_files="$ac_config_files Makefile executer/user/Makefile executer/user/mcexec.1:executer/user/mcexec.1in executer/user/vmcore2mckdump executer/user/arch/$ARCH/Makefile executer/user/arch/x86_64/Makefile executer/kernel/mcctrl/Makefile executer/kernel/mcctrl/arch/$ARCH/Makefile executer/kernel/mcoverlayfs/Makefile executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile executer/kernel/mcoverlayfs/linux-4.0.9/Makefile executer/kernel/mcoverlayfs/linux-4.6.7/Makefile executer/include/qlmpilib.h kernel/Makefile kernel/Makefile.build kernel/include/swapfmt.h arch/x86_64/tools/mcreboot-attached-mic.sh arch/x86_64/tools/mcshutdown-attached-mic.sh arch/x86_64/tools/mcreboot-builtin-x86.sh arch/x86_64/tools/mcreboot-smp-x86.sh arch/x86_64/tools/mcstop+release-smp-x86.sh arch/x86_64/tools/mcoverlay-destroy-smp-x86.sh arch/x86_64/tools/mcoverlay-create-smp-x86.sh arch/x86_64/tools/eclair-dump-backtrace.exp arch/x86_64/tools/mcshutdown-builtin-x86.sh arch/x86_64/tools/mcreboot.1:arch/x86_64/tools/mcreboot.1in arch/x86_64/tools/mpimcexec arch/x86_64/tools/mpimcexec.1:arch/x86_64/tools/mpimcexec.1in arch/x86_64/tools/irqbalance_mck.service arch/x86_64/tools/irqbalance_mck.in tools/mcstat/Makefile"
 
 
-if test -e "${ABS_SRCDIR}/test"; then
-ac_config_files="$ac_config_files mck_test_config.sample:test/mck_test_config.sample.in"
-
-fi
-
 if test "$TARGET" = "smp-x86"; then
 ac_config_files="$ac_config_files arch/x86_64/kernel/Makefile.arch"
 
@@ -5417,7 +5585,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by mckernel $as_me 1.6.0, which was
+This file was extended by mckernel $as_me 1.5.1-knl+hfi, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -5479,7 +5647,7 @@ _ACEOF
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`"
 ac_cs_version="\\
-mckernel config.status 1.6.0
+mckernel config.status 1.5.1-knl+hfi
 configured by $0, generated by GNU Autoconf 2.69,
   with options \\"\$ac_cs_config\\"
 
@@ -5627,11 +5795,11 @@ do
     "arch/x86_64/tools/eclair-dump-backtrace.exp") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/eclair-dump-backtrace.exp" ;;
     "arch/x86_64/tools/mcshutdown-builtin-x86.sh") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/mcshutdown-builtin-x86.sh" ;;
     "arch/x86_64/tools/mcreboot.1") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/mcreboot.1:arch/x86_64/tools/mcreboot.1in" ;;
+    "arch/x86_64/tools/mpimcexec") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/mpimcexec" ;;
+    "arch/x86_64/tools/mpimcexec.1") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/mpimcexec.1:arch/x86_64/tools/mpimcexec.1in" ;;
     "arch/x86_64/tools/irqbalance_mck.service") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/irqbalance_mck.service" ;;
     "arch/x86_64/tools/irqbalance_mck.in") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/irqbalance_mck.in" ;;
-    "tools/mcstat/mcstat.1") CONFIG_FILES="$CONFIG_FILES tools/mcstat/mcstat.1:tools/mcstat/mcstat.1in" ;;
     "tools/mcstat/Makefile") CONFIG_FILES="$CONFIG_FILES tools/mcstat/Makefile" ;;
-    "mck_test_config.sample") CONFIG_FILES="$CONFIG_FILES mck_test_config.sample:test/mck_test_config.sample.in" ;;
     "arch/x86_64/kernel/Makefile.arch") CONFIG_FILES="$CONFIG_FILES arch/x86_64/kernel/Makefile.arch" ;;
     "kernel/config/config.smp-arm64") CONFIG_FILES="$CONFIG_FILES kernel/config/config.smp-arm64" ;;
     "arch/arm64/kernel/vdso/Makefile") CONFIG_FILES="$CONFIG_FILES arch/arm64/kernel/vdso/Makefile" ;;

configure.ac

@@ -1,9 +1,9 @@
 # configure.ac COPYRIGHT FUJITSU LIMITED 2015-2016
 AC_PREREQ(2.63)
-m4_define([IHK_VERSION_m4],[1.6.0])dnl
-m4_define([MCKERNEL_VERSION_m4],[1.6.0])dnl
-m4_define([IHK_RELEASE_DATE_m4],[2018-11-11])dnl
-m4_define([MCKERNEL_RELEASE_DATE_m4],[2018-11-11])dnl
+m4_define([IHK_VERSION_m4],[1.5.1-knl+hfi])dnl
+m4_define([MCKERNEL_VERSION_m4],[1.5.1-knl+hfi])dnl
+m4_define([IHK_RELEASE_DATE_m4],[2019-05-14])dnl
+m4_define([MCKERNEL_RELEASE_DATE_m4],[2019-05-14])dnl
 
 AC_INIT([mckernel], MCKERNEL_VERSION_m4)
 
@@ -77,58 +77,6 @@ AC_DEFUN([PAC_SET_HEADER_LIB_PATH],[
           ])
 ])
 
-AC_DEFUN([PAC_SET_HEADER_LIB_PATH_SYSCALL_INTERCEPT],[
-    AC_ARG_WITH([$1],
-                [AC_HELP_STRING([--with-$1=PATH],
-                                [specify path where $1 include directory and lib directory can be found])],
-
-                [AS_CASE(["$withval"],
-                         [yes|no|''],
-                         [AC_MSG_WARN([--with[out]-$1=PATH expects a valid PATH])
-                          with_$1=""])],
-                [with_$1=$2])
-    AC_ARG_WITH([$1-include],
-                [AC_HELP_STRING([--with-$1-include=PATH],
-                                [specify path where $1 include directory can be found])],
-                [AS_CASE(["$withval"],
-                         [yes|no|''],
-                         [AC_MSG_WARN([--with[out]-$1-include=PATH expects a valid PATH])
-                          with_$1_include=""])],
-                [])
-    AC_ARG_WITH([$1-lib],
-                [AC_HELP_STRING([--with-$1-lib=PATH],
-                                [specify path where $1 lib directory can be found])],
-                [AS_CASE(["$withval"],
-                         [yes|no|''],
-                         [AC_MSG_WARN([--with[out]-$1-lib=PATH expects a valid PATH])
-                          with_$1_lib=""])],
-                [])
-
-    # The args have been sanitized into empty/non-empty values above.
-    # Now append -I/-L args to CPPFLAGS/LDFLAGS, with more specific options
-    # taking priority
-
-    AS_IF([test -n "${with_$1_include}"],
-          [PAC_APPEND_FLAG([-I${with_$1_include}],[CPPFLAGS_SYSCALL_INTERCEPT])],
-          [AS_IF([test -n "${with_$1}"],
-                 [PAC_APPEND_FLAG([-I${with_$1}/include],[CPPFLAGS_SYSCALL_INTERCEPT])])])
-
-    AS_IF([test -n "${with_$1_lib}"],
-          [PAC_APPEND_FLAG([-L${with_$1_lib} -Wl,-rpath,${with_$1_lib}],[LDFLAGS_SYSCALL_INTERCEPT])],
-          [AS_IF([test -n "${with_$1}"],
-                 dnl is adding lib64 by default really the right thing to do?  What if
-                 dnl we are on a 32-bit host that happens to have both lib dirs available?
-                 [PAC_APPEND_FLAG([-L${with_$1}/lib -Wl,-rpath,${with_$1}/lib],[LDFLAGS_SYSCALL_INTERCEPT])
-                  AS_IF([test -d "${with_$1}/lib64"],
-		        [PAC_APPEND_FLAG([-L${with_$1}/lib64 -Wl,-rpath,${with_$1}/lib64],[LDFLAGS_SYSCALL_INTERCEPT])])
-                 ])
-          ])
-
-    AS_IF([test -n "${with_$1}" || test -n "${with_$1_include}" || test -n "${with_$1_lib}"],
-	      [WITH_SYSCALL_INTERCEPT=yes],
-	      [WITH_SYSCALL_INTERCEPT=no])
-])
-
 IHK_VERSION=IHK_VERSION_m4
 MCKERNEL_VERSION=MCKERNEL_VERSION_m4
 DCFA_VERSION=DCFA_VERSION_m4
@@ -147,23 +95,6 @@ AS_IF([test "x$numa_lib_found" != "xyes"],
 
 PAC_SET_HEADER_LIB_PATH([mpi])
 
-PAC_SET_HEADER_LIB_PATH_SYSCALL_INTERCEPT([syscall_intercept])
-
-if test "x$WITH_SYSCALL_INTERCEPT" == "xno" ; then
-	AC_CHECK_LIB([syscall_intercept],[syscall_no_intercept],[syscall_intercept_lib_found=yes],[syscall_intercept_lib_found=no],[-lcapstone -ldl])
-	AS_IF([test "x$syscall_intercept_lib_found" != "xyes"],
-		[AC_MSG_NOTICE([libsyscall_intercept.so not found])])
-
-	AC_CHECK_HEADER([libsyscall_intercept_hook_point.h],[syscall_intercept_header_found=yes],[syscall_intercept_header_found=no])
-	AS_IF([test "x$syscall_intercept_header_found" != "xyes"],
-		[AC_MSG_NOTICE([libsyscall_intercept_hook_point.h not found])])
-
-	AS_IF([test "x$syscall_intercept_lib_found" == "xyes" && test "x$syscall_intercept_header_found" == "xyes"],
-		[WITH_SYSCALL_INTERCEPT=yes],
-		[WITH_SYSCALL_INTERCEPT=no])
-fi
-
-
 AC_ARG_WITH([kernelsrc],
   AC_HELP_STRING(
     [--with-kernelsrc=path],[Path to 'kernel src', default is /lib/modules/uname_r/build]),
@@ -408,6 +339,78 @@ KDIR="$WITH_KERNELSRC"
 UNAME_R="$WITH_UNAME_R"
 TARGET="$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 -r "$MCCTRL_LINUX_SYMTAB"; then
+	MCCTRL_LINUX_SYMTAB="$MCCTRL_LINUX_SYMTAB"
+elif test -r "/boot/System.map-`uname -r`"; then
+	MCCTRL_LINUX_SYMTAB="/boot/System.map-`uname -r`"
+elif test -r "$KDIR/System.map"; then
+	MCCTRL_LINUX_SYMTAB="$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_umount])
+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])
+# POSTK_DEBUG_ARCH_DEP_50, add:find kernel symbol.
+MCCTRL_FIND_KSYM([vdso_spec])
+MCCTRL_FIND_KSYM([hv_clock])
+MCCTRL_FIND_KSYM([sys_readlink])
+MCCTRL_FIND_KSYM([walk_page_range])
+
 case $ENABLE_MEMDUMP in
 	yes|no|auto)
 		;;
@@ -486,13 +489,6 @@ else
 	AC_MSG_NOTICE([perf is disabled])
 fi
 
-if test "x$WITH_SYSCALL_INTERCEPT" = "xyes" ; then
-	AC_DEFINE([WITH_SYSCALL_INTERCEPT],[1],[whether or not syscall_intercept library is linked])
-	AC_MSG_NOTICE([syscall_intercept library is linked])
-else
-	AC_MSG_NOTICE([syscall_intercept library isn't linked])
-fi
-
 if test "x$MCKERNEL_INCDIR" != "x" ; then
    AC_DEFINE_UNQUOTED(MCKERNEL_INCDIR,"$MCKERNEL_INCDIR",[McKernel specific headers])
 fi
@@ -530,12 +526,9 @@ AC_SUBST(KMODDIR)
 AC_SUBST(KERNDIR)
 AC_SUBST(MANDIR)
 AC_SUBST(CFLAGS)
-AC_SUBST(CPPFLAGS_SYSCALL_INTERCEPT)
-AC_SUBST(LDFLAGS_SYSCALL_INTERCEPT)
 AC_SUBST(ENABLE_MCOVERLAYFS)
 AC_SUBST(ENABLE_RUSAGE)
 AC_SUBST(ENABLE_QLMPI)
-AC_SUBST(WITH_SYSCALL_INTERCEPT)
 
 AC_SUBST(IHK_VERSION)
 AC_SUBST(MCKERNEL_VERSION)
@@ -575,18 +568,13 @@ AC_CONFIG_FILES([
 	arch/x86_64/tools/eclair-dump-backtrace.exp
 	arch/x86_64/tools/mcshutdown-builtin-x86.sh
 	arch/x86_64/tools/mcreboot.1:arch/x86_64/tools/mcreboot.1in
+	arch/x86_64/tools/mpimcexec
+	arch/x86_64/tools/mpimcexec.1:arch/x86_64/tools/mpimcexec.1in
 	arch/x86_64/tools/irqbalance_mck.service
 	arch/x86_64/tools/irqbalance_mck.in
-	tools/mcstat/mcstat.1:tools/mcstat/mcstat.1in
 	tools/mcstat/Makefile
 ])
 
-if test -e "${ABS_SRCDIR}/test"; then
-AC_CONFIG_FILES([
-mck_test_config.sample:test/mck_test_config.sample.in
-])
-fi
-
 if test "$TARGET" = "smp-x86"; then
 AC_CONFIG_FILES([
 	arch/x86_64/kernel/Makefile.arch

executer/include/uprotocol.h

@@ -55,14 +55,13 @@
 #define MCEXEC_UP_SYS_UMOUNT     0x30a02915
 #define MCEXEC_UP_SYS_UNSHARE    0x30a02916
 
-#define MCEXEC_UP_UTI_GET_CTX    0x30a02920
-#define MCEXEC_UP_UTI_SAVE_FS    0x30a02921
+#define MCEXEC_UP_UTIL_THREAD1   0x30a02920
+#define MCEXEC_UP_UTIL_THREAD2   0x30a02921
 #define MCEXEC_UP_SIG_THREAD     0x30a02922
 #define MCEXEC_UP_SYSCALL_THREAD 0x30a02924
 #define MCEXEC_UP_TERMINATE_THREAD 0x30a02925
 #define MCEXEC_UP_GET_NUM_POOL_THREADS  0x30a02926
 #define MCEXEC_UP_UTI_ATTR       0x30a02927
-#define MCEXEC_UP_RELEASE_USER_SPACE 0x30a02928
 
 #define MCEXEC_UP_DEBUG_LOG     0x40000000
 
@@ -111,10 +110,14 @@ typedef unsigned long __cpu_set_unit;
 #define MPOL_NO_BSS               0x04
 #define MPOL_SHM_PREMAP           0x08
 
+#define MCEXEC_HFI1               0x01
+
 struct program_load_desc {
 	int num_sections;
+	int status;
 	int cpu;
 	int pid;
+	int err;
 	int stack_prot;
 	int pgid;
 	int cred[8];
@@ -137,15 +140,15 @@ struct program_load_desc {
 	unsigned long envs_len;
 	struct rlimit rlimit[MCK_RLIM_MAX];
 	unsigned long interp_align;
+	unsigned long mcexec_flags;
 	unsigned long mpol_flags;
 	unsigned long mpol_threshold;
 	unsigned long heap_extension;
 	long stack_premap;
 	unsigned long mpol_bind_mask;
-	int uti_thread_rank; /* N-th clone() spawns a thread on Linux CPU */
-	int uti_use_last_cpu; /* Work-around not to share CPU with OpenMP thread */
 	int nr_processes;
 	int process_rank;
+	char shell_path[SHELL_PATH_MAX_LEN];
 	__cpu_set_unit cpu_set[PLD_CPU_SET_SIZE];
 	int profile;
 	struct program_image_section sections[0];
@@ -191,6 +194,7 @@ struct syscall_response {
 	long ret;
 	unsigned long fault_address;
 	unsigned long fault_reason;
+	void *private_data;
 };
 
 struct syscall_ret_desc {
@@ -246,28 +250,6 @@ struct sys_unshare_desc {
 	unsigned long unshare_flags;
 };
 
-struct release_user_space_desc {
-	unsigned long user_start;
-	unsigned long user_end;
-};
-
-struct terminate_thread_desc {
-	int pid;
-	int tid;
-
-	long code; 
-	/* 32------32  31--16  15--------8  7----0
-       exit_group          exit-status  signal */
-
-	unsigned long tsk; /* struct task_struct * */
-};
-
-struct rpgtable_desc {
-	uintptr_t rpgtable;
-	uintptr_t start;
-	uintptr_t len;
-};
-
 enum perf_ctrl_type {
 	PERF_CTRL_SET,
 	PERF_CTRL_GET,
@@ -277,7 +259,7 @@ enum perf_ctrl_type {
 
 struct perf_ctrl_desc {
 	enum perf_ctrl_type ctrl_type;
-	int err;
+	int status;
 	union {
 		/* for SET, GET */
 		struct {
@@ -317,10 +299,6 @@ struct perf_ctrl_desc {
 #define UTI_FLAG_HIGH_PRIORITY (1ULL<<12)
 #define UTI_FLAG_NON_COOPERATIVE (1ULL<<13)
 
-#define UTI_FLAG_PREFER_LWK (1ULL << 14)
-#define UTI_FLAG_PREFER_FWK (1ULL << 15)
-#define UTI_FLAG_FABRIC_INTR_AFFINITY (1ULL << 16)
-
 /* Linux default value is used */
 #define UTI_MAX_NUMA_DOMAINS (1024)
 
@@ -339,30 +317,6 @@ struct kuti_attr {
 
 struct uti_attr_desc {
 	unsigned long phys_attr;
-	char *uti_cpu_set_str; /* UTI_CPU_SET environmental variable */
-	size_t uti_cpu_set_len;
-};
-
-struct uti_ctx {
-	union {
-		char ctx[4096]; /* TODO: Get the size from config.h */
-		struct {
-			int uti_refill_tid;
-		};
-	};
-}; 
-
-struct uti_get_ctx_desc {
-	unsigned long rp_rctx; /* Remote physical address of remote context */
-	void *rctx; /* Remote context */
-	void *lctx; /* Local context */
-	int uti_refill_tid;
-	unsigned long key; /* OUT: struct task_struct* of mcexec thread, used to search struct host_thread */
-};
-
-struct uti_save_fs_desc {
-	void *rctx; /* Remote context */
-	void *lctx; /* Local context */
 };
 
 #endif

executer/include/uti.h

@@ -1,31 +0,0 @@
-#ifndef UTI_H_INCLUDED
-#define UTI_H_INCLUDED
-
-struct syscall_struct {
-	int number;
-	unsigned long args[6];
-	unsigned long ret;
-	unsigned long uti_clv; /* copy of a clv in McKernel */
-};
-
-#define UTI_SZ_SYSCALL_STACK 16
-
-/* Variables accessed by mcexec.c and syscall_intercept.c */
-struct uti_desc {
-	char lctx[4096]; /* TODO: Get the size from config.h */
-	char rctx[4096]; /* TODO: Get the size from config.h */
-	int mck_tid; /* TODO: Move this out for multiple migrated-to-Linux threads */
-	unsigned long key; /* struct task_struct* of mcexec thread, used to search struct host_thread */
-	int pid, tid; /* Used as the id of tracee when issuing MCEXEC_UP_TERMINATE_THREAD */
-	unsigned long uti_clv; /* copy of McKernel clv */
-
-	int fd; /* /dev/mcosX */
-	struct syscall_struct syscall_stack[UTI_SZ_SYSCALL_STACK]; /* stack of system call arguments and return values */
-	int syscall_stack_top; /* stack-pointer of syscall arguments list */
-	long syscalls[512], syscalls2[512]; /* Syscall profile counters */
-	int start_syscall_intercept; /* Used to sync between mcexec.c and syscall_intercept.c */
-};
-
-
-#endif
-

executer/kernel/mcctrl/arch/arm64/archdeps.c

@@ -1,7 +1,6 @@
 /* archdeps.c COPYRIGHT FUJITSU LIMITED 2016 */
 #include <linux/version.h>
 #include <linux/mm_types.h>
-#include <linux/kallsyms.h>
 #include <asm/vdso.h>
 #include "../../../config.h"
 #include "../../mcctrl.h"
@@ -18,31 +17,29 @@
 
 #define D(fmt, ...) printk("%s(%d) " fmt, __func__, __LINE__, ##__VA_ARGS__)
 
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 0, 0)
-void *vdso_start;
-void *vdso_end;
-static struct vm_special_mapping (*vdso_spec)[2];
+#ifdef MCCTRL_KSYM_vdso_start
+# if MCCTRL_KSYM_vdso_start
+void *vdso_start = (void *)MCCTRL_KSYM_vdso_start;
+# endif
+#else
+# error missing address of vdso_start.
 #endif
 
-int arch_symbols_init(void)
-{
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 0, 0)
-	vdso_start = (void *) kallsyms_lookup_name("vdso_start");
-	if (WARN_ON(!vdso_start))
-		return -EFAULT;
-
-	vdso_end = (void *) kallsyms_lookup_name("vdso_end");
-	if (WARN_ON(!vdso_end))
-		return -EFAULT;
-
-	vdso_spec = (void *) kallsyms_lookup_name("vdso_spec");
-	if (WARN_ON(!vdso_spec))
-		return -EFAULT;
+#ifdef MCCTRL_KSYM_vdso_end
+# if MCCTRL_KSYM_vdso_end
+void *vdso_end = (void *)MCCTRL_KSYM_vdso_end;
+# endif
+#else
+# error missing address of vdso_end.
 #endif
 
-	return 0;
-}
-
+#ifdef MCCTRL_KSYM_vdso_spec
+# if MCCTRL_KSYM_vdso_spec
+static struct vm_special_mapping (*vdso_spec)[2] = (void*)MCCTRL_KSYM_vdso_spec;
+# endif
+#else
+# error missing address of vdso_spec.
+#endif
 
 #ifdef POSTK_DEBUG_ARCH_DEP_52
 #define VDSO_MAXPAGES 1

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

@@ -1,6 +1,5 @@
 /* archdeps.c COPYRIGHT FUJITSU LIMITED 2016 */
 #include <linux/version.h>
-#include <linux/kallsyms.h>
 #include "../../../config.h"
 #include "../../mcctrl.h"
 
@@ -14,46 +13,57 @@
 #endif
 #endif /* POSTK_DEBUG_ARCH_DEP_83 */
 
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 16, 0)
-static struct vdso_image *vdso_image_64;
-#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 23)
-static void *vdso_start;
-static void *vdso_end;
-static struct page **vdso_pages;
+#ifdef MCCTRL_KSYM_vdso_image_64
+#if MCCTRL_KSYM_vdso_image_64
+struct vdso_image *vdso_image = (void *)MCCTRL_KSYM_vdso_image_64;
 #endif
-static void *__vvar_page;
-static long *hpet_address;
-static void **hv_clock;
-
-int arch_symbols_init(void)
-{
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 16, 0)
-	vdso_image_64 = (void *) kallsyms_lookup_name("vdso_image_64");
-	if (WARN_ON(!vdso_image_64))
-		return -EFAULT;
-#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 23)
-	vdso_start = (void *) kallsyms_lookup_name("vdso_start");
-	if (WARN_ON(!vdso_start))
-		return -EFAULT;
-
-	vdso_end = (void *) kallsyms_lookup_name("vdso_end");
-	if (WARN_ON(!vdso_end))
-		return -EFAULT;
-
-	vdso_pages = (void *) kallsyms_lookup_name("vdso_pages");
-	if (WARN_ON(!vdso_pages))
-		return -EFAULT;
 #endif
 
-	__vvar_page = (void *) kallsyms_lookup_name("__vvar_page");
-	if (WARN_ON(!__vvar_page))
-		return -EFAULT;
+#ifdef MCCTRL_KSYM_vdso_start
+#if MCCTRL_KSYM_vdso_start
+void *vdso_start = (void *)MCCTRL_KSYM_vdso_start;
+#endif
+#endif
 
-	hpet_address = (void *) kallsyms_lookup_name("hpet_address");
-	hv_clock = (void *) kallsyms_lookup_name("hv_clock");
-	return 0;
-}
+#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
 
 #ifdef POSTK_DEBUG_ARCH_DEP_52
 #define VDSO_MAXPAGES 2
@@ -128,7 +138,7 @@ void get_vdso_info(ihk_os_t os, long vdso_rpa)
 
 	/* VDSO pages */
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,16,0)
-	size = vdso_image_64->size;
+	size = vdso_image->size;
 	vdso->vdso_npages = size >> PAGE_SHIFT;
 
 	if (vdso->vdso_npages > VDSO_MAXPAGES) {
@@ -138,7 +148,7 @@ void get_vdso_info(ihk_os_t os, long vdso_rpa)
 
 	for (i = 0; i < vdso->vdso_npages; ++i) {
 		vdso->vdso_physlist[i] = virt_to_phys(
-				vdso_image_64->data + (i * PAGE_SIZE));
+				vdso_image->data + (i * PAGE_SIZE));
 	}
 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
 	size = vdso_end - vdso_start;
@@ -175,36 +185,36 @@ void get_vdso_info(ihk_os_t os, long vdso_rpa)
 #endif
 
 	/* HPET page */
-	if (hpet_address && *hpet_address) {
+	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_address;
+		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_address;
+		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_address;
+		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_address;
+		vdso->hpet_phys = *hpet_addressp;
 #endif
 	}
 
 	/* struct pvlock_vcpu_time_info table */
-	if (hv_clock && *hv_clock) {
+	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_clock);
+		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_clock);
+		vdso->pvti_phys = virt_to_phys(*hv_clockp);
 #endif
 	}
 
@@ -279,14 +289,6 @@ get_fs_ctx(void *ctx)
 	return tctx->fs;
 }
 
-unsigned long
-get_rsp_ctx(void *ctx)
-{
-	struct trans_uctx *tctx = ctx;
-
-	return tctx->rsp;
-}
-
 #ifdef POSTK_DEBUG_ARCH_DEP_83 /* arch depend translate_rva_to_rpa() move */
 int translate_rva_to_rpa(ihk_os_t os, unsigned long rpt, unsigned long rva,
 		unsigned long *rpap, unsigned long *pgsizep)

executer/kernel/mcctrl/binfmt_mcexec.c

@@ -125,6 +125,7 @@ static int load_elf(struct linux_binprm *bprm
 		for(i = 0, st = 0; mode != 2;){
 			if(st == 0){
 				off = p & ~PAGE_MASK;
+#ifdef POSTK_DEBUG_ARCH_DEP_41 /* HOST-Linux version switch add */
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,10,0)
 				rc = get_user_pages_remote(current, bprm->mm,
 				                        bprm->p, 1, FOLL_FORCE, &page, NULL, NULL);
@@ -140,6 +141,17 @@ static int load_elf(struct linux_binprm *bprm
 				                        bprm->p, 1, 0, 1,
 				                        &page, NULL);
 #endif
+#else /* POSTK_DEBUG_ARCH_DEP_41 */
+#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);
+#endif
+#endif /* POSTK_DEBUG_ARCH_DEP_41 */
 				if(rc <= 0) {
 					kfree(pbuf);
 					return -EFAULT;

executer/kernel/mcctrl/driver.c

@@ -28,7 +28,6 @@
 #include <linux/slab.h>
 #include <linux/device.h>
 #include <linux/delay.h>
-#include <linux/kallsyms.h>
 #include "mcctrl.h"
 #include <ihk/ihk_host_user.h>
 
@@ -44,6 +43,8 @@ extern void mcctrl_syscall_init(void);
 extern void procfs_init(int);
 extern void procfs_exit(int);
 
+extern void rus_page_hash_init(void);
+extern void rus_page_hash_put_pages(void);
 extern void uti_attr_finalize(void);
 extern void binfmt_mcexec_init(void);
 extern void binfmt_mcexec_exit(void);
@@ -83,14 +84,13 @@ static struct ihk_os_user_call_handler mcctrl_uchs[] = {
 	{ .request = MCEXEC_UP_SYS_MOUNT, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SYS_UMOUNT, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SYS_UNSHARE, .func = mcctrl_ioctl },
-	{ .request = MCEXEC_UP_UTI_GET_CTX, .func = mcctrl_ioctl },
-	{ .request = MCEXEC_UP_UTI_SAVE_FS, .func = mcctrl_ioctl },
+	{ .request = MCEXEC_UP_UTIL_THREAD1, .func = mcctrl_ioctl },
+	{ .request = MCEXEC_UP_UTIL_THREAD2, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SIG_THREAD, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_SYSCALL_THREAD, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_TERMINATE_THREAD, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_GET_NUM_POOL_THREADS, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_UTI_ATTR, .func = mcctrl_ioctl },
-	{ .request = MCEXEC_UP_RELEASE_USER_SPACE, .func = mcctrl_ioctl },
 	{ .request = MCEXEC_UP_DEBUG_LOG, .func = mcctrl_ioctl },
 	{ .request = IHK_OS_AUX_PERF_NUM, .func = mcctrl_ioctl },
 	{ .request = IHK_OS_AUX_PERF_SET, .func = mcctrl_ioctl },
@@ -178,7 +178,6 @@ int mcctrl_os_shutdown_notifier(int os_index)
 			mdelay(200);
 		}
 
-		pager_cleanup();
 		sysfsm_cleanup(os[os_index]);
 		free_topology_info(os[os_index]);
 		ihk_os_unregister_user_call_handlers(os[os_index], mcctrl_uc + os_index);
@@ -186,6 +185,9 @@ int mcctrl_os_shutdown_notifier(int os_index)
 		destroy_ikc_channels(os[os_index]);
 		procfs_exit(os_index);
 	}
+#ifdef POSTK_DEBUG_TEMP_FIX_35 /* in shutdown phase, rus_page_hash_put_pages() call added. */
+	rus_page_hash_put_pages();
+#endif /* POSTK_DEBUG_TEMP_FIX_35 */
 
 	os[os_index] = NULL;
 
@@ -212,68 +214,6 @@ static struct ihk_os_notifier mcctrl_os_notifier = {
 	.ops = &mcctrl_os_notifier_ops,
 };
 
-
-
-int (*mcctrl_sys_mount)(char *dev_name, char *dir_name, char *type,
-			unsigned long flags, void *data);
-int (*mcctrl_sys_umount)(char *dir_name, int flags);
-int (*mcctrl_sys_unshare)(unsigned long unshare_flags);
-long (*mcctrl_sched_setaffinity)(pid_t pid, const struct cpumask *in_mask);
-int (*mcctrl_sched_setscheduler_nocheck)(struct task_struct *p, int policy,
-					 const struct sched_param *param);
-
-ssize_t (*mcctrl_sys_readlink)(const char *path, char *buf,
-			       size_t bufsiz);
-void (*mcctrl_zap_page_range)(struct vm_area_struct *vma,
-			      unsigned long start,
-			      unsigned long size,
-			      struct zap_details *details);
-
-struct inode_operations *mcctrl_hugetlbfs_inode_operations;
-
-
-static int symbols_init(void)
-{
-	mcctrl_sys_mount = (void *) kallsyms_lookup_name("sys_mount");
-	if (WARN_ON(!mcctrl_sys_mount))
-		return -EFAULT;
-
-	mcctrl_sys_umount = (void *) kallsyms_lookup_name("sys_umount");
-	if (WARN_ON(!mcctrl_sys_umount))
-		return -EFAULT;
-
-	mcctrl_sys_unshare = (void *) kallsyms_lookup_name("sys_unshare");
-	if (WARN_ON(!mcctrl_sys_unshare))
-		return -EFAULT;
-
-	mcctrl_sched_setaffinity =
-		(void *) kallsyms_lookup_name("sched_setaffinity");
-	if (WARN_ON(!mcctrl_sched_setaffinity))
-		return -EFAULT;
-
-	mcctrl_sched_setscheduler_nocheck =
-		(void *) kallsyms_lookup_name("sched_setscheduler_nocheck");
-	if (WARN_ON(!mcctrl_sched_setscheduler_nocheck))
-		return -EFAULT;
-
-	mcctrl_sys_readlink =
-		(void *) kallsyms_lookup_name("sys_readlink");
-	if (WARN_ON(!mcctrl_sys_readlink))
-		return -EFAULT;
-
-	mcctrl_zap_page_range =
-		(void *) kallsyms_lookup_name("zap_page_range");
-	if (WARN_ON(!mcctrl_zap_page_range))
-		return -EFAULT;
-
-	mcctrl_hugetlbfs_inode_operations =
-		(void *) kallsyms_lookup_name("hugetlbfs_inode_operations");
-	if (WARN_ON(!mcctrl_hugetlbfs_inode_operations))
-		return -EFAULT;
-
-	return arch_symbols_init();
-}
-
 static int __init mcctrl_init(void)
 {
 	int ret = 0;
@@ -287,10 +227,9 @@ static int __init mcctrl_init(void)
 		os[i] = NULL;
 	}
 
-	binfmt_mcexec_init();
+	rus_page_hash_init();
 
-	if ((ret = symbols_init()))
-		goto error;
+	binfmt_mcexec_init();
 
 	if ((ret = ihk_host_register_os_notifier(&mcctrl_os_notifier)) != 0) {
 		printk("mcctrl: error: registering OS notifier\n");
@@ -302,6 +241,7 @@ static int __init mcctrl_init(void)
 
 error:
 	binfmt_mcexec_exit();
+	rus_page_hash_put_pages();
 
 	return ret;
 }
@@ -313,6 +253,7 @@ static void __exit mcctrl_exit(void)
 	}
 
 	binfmt_mcexec_exit();
+	rus_page_hash_put_pages();
 	uti_attr_finalize();
 
 	printk("mcctrl: unregistered.\n");

executer/kernel/mcctrl/ikc.c

@@ -49,125 +49,15 @@
 
 //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_usrdata *ud, struct ikc_scd_packet *packet);
 void sig_done(unsigned long arg, int err);
 void mcctrl_perf_ack(ihk_os_t os, struct ikc_scd_packet *packet);
-void mcctrl_futex_wake(struct ikc_scd_packet *pisp);
 void mcctrl_os_read_write_cpu_response(ihk_os_t os,
 		struct ikc_scd_packet *pisp);
 void mcctrl_eventfd(ihk_os_t os, struct ikc_scd_packet *pisp);
 
-/* Assumes usrdata->wakeup_descs_lock taken */
-static void mcctrl_wakeup_desc_cleanup(ihk_os_t os,
-		struct mcctrl_wakeup_desc *desc)
-{
-	int i;
-
-	list_del(&desc->chain);
-
-	for (i = 0; i < desc->free_addrs_count; i++) {
-		kfree(desc->free_addrs[i]);
-	}
-}
-
-static void mcctrl_wakeup_cb(ihk_os_t os, struct ikc_scd_packet *packet)
-{
-	struct mcctrl_wakeup_desc *desc = packet->reply;
-
-
-	WRITE_ONCE(desc->err, packet->err);
-
-	/*
-	 * Check if the other side is still waiting, and signal it we're done.
-	 *
-	 * Setting status needs to be done last because the other side could
-	 * wake up opportunistically between this set and the wake_up call.
-	 *
-	 * If the other side is no longer waiting, free the memory that was
-	 * left for us.
-	 */
-	if (cmpxchg(&desc->status, 0, 1)) {
-		struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
-		unsigned long flags;
-
-		spin_lock_irqsave(&usrdata->wakeup_descs_lock, flags);
-		mcctrl_wakeup_desc_cleanup(os, desc);
-		spin_unlock_irqrestore(&usrdata->wakeup_descs_lock, flags);
-		return;
-	}
-
-	wake_up_interruptible(&desc->wq);
-}
-
-int mcctrl_ikc_send_wait(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp,
-		long int timeout, struct mcctrl_wakeup_desc *desc,
-		int *do_frees, int free_addrs_count, ...)
-{
-	int ret, i;
-	int alloc_desc = (desc == NULL);
-	va_list ap;
-
-	if (free_addrs_count)
-		*do_frees = 1;
-	if (alloc_desc)
-		desc = kmalloc(sizeof(struct mcctrl_wakeup_desc) +
-			       (free_addrs_count + 1) * sizeof(void *),
-			       GFP_KERNEL);
-	if (!desc) {
-		pr_warn("%s: Could not allocate wakeup descriptor", __func__);
-		return -ENOMEM;
-	}
-	pisp->reply = desc;
-	va_start(ap, free_addrs_count);
-	for (i = 0; i < free_addrs_count; i++) {
-		desc->free_addrs[i] = va_arg(ap, void*);
-	}
-	va_end(ap);
-	if (alloc_desc)
-		desc->free_addrs[free_addrs_count++] = desc;
-	desc->free_addrs_count = free_addrs_count;
-	init_waitqueue_head(&desc->wq);
-	WRITE_ONCE(desc->err, 0);
-	WRITE_ONCE(desc->status, 0);
-
-	ret = mcctrl_ikc_send(os, cpu, pisp);
-	if (ret < 0) {
-		pr_warn("%s: mcctrl_ikc_send failed: %d\n", __func__, ret);
-		kfree(desc);
-		return ret;
-	}
-
-	if (timeout) {
-		ret = wait_event_interruptible_timeout(desc->wq,
-			desc->status, timeout);
-	} else {
-		ret = wait_event_interruptible(desc->wq, desc->status);
-	}
-
-	/*
-	 * Check if wait aborted (signal..) or timed out, and notify
-	 * the callback it will need to free things for us
-	 */
-	if (!cmpxchg(&desc->status, 0, 1)) {
-		struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
-		unsigned long flags;
-
-		spin_lock_irqsave(&usrdata->wakeup_descs_lock, flags);
-		list_add(&desc->chain, &usrdata->wakeup_descs_list);
-		spin_unlock_irqrestore(&usrdata->wakeup_descs_lock, flags);
-		if (do_frees)
-			*do_frees = 0;
-		return ret < 0 ? ret : -ETIME;
-	}
-
-	ret = READ_ONCE(desc->err);
-	if (alloc_desc)
-		kfree(desc);
-	return ret;
-}
-
-
 /* XXX: this runs in atomic context! */
 static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
                                   void *__packet, void *__os)
@@ -182,16 +72,25 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		break;
 
 	case SCD_MSG_PREPARE_PROCESS_ACKED:
-	case SCD_MSG_PERF_ACK:
-	case SCD_MSG_SEND_SIGNAL_ACK:
-	case SCD_MSG_PROCFS_ANSWER:
-		mcctrl_wakeup_cb(__os, pisp);
+		mcexec_prepare_ack(__os, pisp->arg, 0);
+		break;
+
+	case SCD_MSG_PREPARE_PROCESS_NACKED:
+		mcexec_prepare_ack(__os, pisp->arg, pisp->err);
 		break;
 
 	case SCD_MSG_SYSCALL_ONESIDE:
 		mcexec_syscall(usrdata, pisp);
 		break;
 
+	case SCD_MSG_PROCFS_ANSWER:
+		procfs_answer(usrdata, pisp->pid);
+		break;
+
+	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:
@@ -207,14 +106,17 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 
 	case SCD_MSG_PROCFS_TID_CREATE:
 	case SCD_MSG_PROCFS_TID_DELETE:
-		procfsm_packet_handler(__os, pisp->msg, pisp->pid, pisp->arg,
-				       pisp->resp_pa);
+		procfsm_packet_handler(__os, pisp->msg, pisp->pid, pisp->arg);
 		break;
 
 	case SCD_MSG_GET_VDSO_INFO:
 		get_vdso_info(__os, pisp->arg);
 		break;
 
+	case SCD_MSG_PERF_ACK:
+		mcctrl_perf_ack(__os, pisp);
+		break;
+
 	case SCD_MSG_CPU_RW_REG_RESP:
 		mcctrl_os_read_write_cpu_response(__os, pisp);
 		break;
@@ -224,10 +126,6 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		mcctrl_eventfd(__os, pisp);
 		break;
 
-	case SCD_MSG_FUTEX_WAKE:
-		mcctrl_futex_wake(pisp);
-		break;
-
 	default:
 		printk(KERN_ERR "mcctrl:syscall_packet_handler:"
 				"unknown message (%d.%d.%d.%d.%d.%#lx)\n",
@@ -259,15 +157,10 @@ static int dummy_packet_handler(struct ihk_ikc_channel_desc *c,
 
 int mcctrl_ikc_send(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp)
 {
-	struct mcctrl_usrdata *usrdata;
+	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 
-	if (cpu < 0 || os == NULL) {
-		return -EINVAL;
-	}
-
-	usrdata = ihk_host_os_get_usrdata(os);
-	if (usrdata == NULL || cpu >= usrdata->num_channels ||
-	    !usrdata->channels[cpu].c) {
+	if (cpu < 0 || os == NULL || usrdata == NULL ||
+	    cpu >= usrdata->num_channels || !usrdata->channels[cpu].c) {
 		return -EINVAL;
 	}
 	return ihk_ikc_send(usrdata->channels[cpu].c, pisp, 0);
@@ -461,8 +354,6 @@ int prepare_ikc_channels(ihk_os_t os)
 	mutex_init(&usrdata->part_exec.lock);
 	INIT_LIST_HEAD(&usrdata->part_exec.pli_list);
 	usrdata->part_exec.nr_processes = -1;
-	INIT_LIST_HEAD(&usrdata->wakeup_descs_list);
-	spin_lock_init(&usrdata->wakeup_descs_lock);
 
 	return 0;
 
@@ -484,9 +375,7 @@ void __destroy_ikc_channel(ihk_os_t os, struct mcctrl_channel *pmc)
 void destroy_ikc_channels(ihk_os_t os)
 {
 	int i;
-	unsigned long flags;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
-	struct mcctrl_wakeup_desc *mwd_entry, *mwd_next;
 
 	if (!usrdata) {
 		printk("%s: WARNING: no mcctrl_usrdata found\n", __FUNCTION__);
@@ -506,12 +395,6 @@ void destroy_ikc_channels(ihk_os_t os)
 			ihk_ikc_destroy_channel(usrdata->ikc2linux[i]);
 		}
 	}
-	spin_lock_irqsave(&usrdata->wakeup_descs_lock, flags);
-	list_for_each_entry_safe(mwd_entry, mwd_next,
-			&usrdata->wakeup_descs_list, chain) {
-		mcctrl_wakeup_desc_cleanup(os, mwd_entry);
-	}
-	spin_unlock_irqrestore(&usrdata->wakeup_descs_lock, flags);
 
 	kfree(usrdata->channels);
 	kfree(usrdata->ikc2linux);

executer/kernel/mcctrl/mcctrl.h

@@ -48,6 +48,7 @@
 
 #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
 
@@ -55,8 +56,7 @@
 #define SCD_MSG_INIT_CHANNEL_ACKED      0x6
 
 #define SCD_MSG_SYSCALL_ONESIDE         0x4
-#define SCD_MSG_SEND_SIGNAL             0x7
-#define SCD_MSG_SEND_SIGNAL_ACK         0x8
+#define SCD_MSG_SEND_SIGNAL     	0x8
 #define SCD_MSG_CLEANUP_PROCESS         0x9
 #define SCD_MSG_GET_VDSO_INFO           0xa
 
@@ -67,7 +67,6 @@
 #define	SCD_MSG_PROCFS_DELETE		0x11
 #define	SCD_MSG_PROCFS_REQUEST		0x12
 #define	SCD_MSG_PROCFS_ANSWER		0x13
-#define	SCD_MSG_PROCFS_RELEASE		0x15
 
 #define SCD_MSG_DEBUG_LOG		0x20
 
@@ -102,18 +101,23 @@
 #define SCD_MSG_CPU_RW_REG              0x52
 #define SCD_MSG_CPU_RW_REG_RESP         0x53
 
-#define SCD_MSG_FUTEX_WAKE              0x60
-
 #define DMA_PIN_SHIFT                   21
 
 #define DO_USER_MODE
 
 #define	__NR_coredump			999
 
+#ifdef POSTK_DEBUG_TEMP_FIX_61 /* Core table size and lseek return value to loff_t */
 struct coretable {
 	loff_t len;
 	unsigned long addr;
 };
+#else /* POSTK_DEBUG_TEMP_FIX_61 */
+struct coretable {
+	int len;
+	unsigned long addr;
+};
+#endif /* POSTK_DEBUG_TEMP_FIX_61 */
 
 enum mcctrl_os_cpu_operation {
 	MCCTRL_OS_CPU_READ_REGISTER,
@@ -121,16 +125,9 @@ enum mcctrl_os_cpu_operation {
 	MCCTRL_OS_CPU_MAX_OP
 };
 
-/* Used to wake-up a Linux thread futex_wait()-ing */
-struct uti_futex_resp {
-	int done;
-	wait_queue_head_t wq;
-};
-
 struct ikc_scd_packet {
 	int msg;
 	int err;
-	void *reply;
 	union {
 		/* for traditional SCD_MSG_* */
 		struct {
@@ -149,7 +146,7 @@ struct ikc_scd_packet {
 			long sysfs_arg3;
 		};
 
-		/* SCD_MSG_WAKE_UP_SYSCALL_THREAD */
+		/* SCD_MSG_SCHEDULE_THREAD */
 		struct {
 			int ttid;
 		};
@@ -165,17 +162,10 @@ struct ikc_scd_packet {
 		struct {
 			int eventfd_type;
 		};
-
-		/* SCD_MSG_FUTEX_WAKE */
-		struct {
-			void *resp;
-			int *spin_sleep; /* 1: waiting in linux_wait_event() 0: woken up by someone else */
-		} futex;
 	};
-	char padding[8];
+	char padding[12];
 };
 
-
 struct mcctrl_priv { 
 	ihk_os_t os;
 	struct program_load_desc *desc;
@@ -221,12 +211,9 @@ struct mcctrl_channel {
 };
 
 struct mcctrl_per_thread_data {
-	struct mcctrl_per_proc_data *ppd;
 	struct list_head hash;
 	struct task_struct *task;
 	void *data;
-	int tid; /* debug */
-	atomic_t refcount;
 };
 
 #define MCCTRL_PER_THREAD_DATA_HASH_SHIFT 8
@@ -244,6 +231,7 @@ struct mcctrl_per_proc_data {
 	struct list_head wq_list_exact;  /* These requests come from IKC IRQ handler targeting a particular thread */
 
 	ihk_spinlock_t wq_list_lock;
+	wait_queue_head_t wq_prepare;
 	wait_queue_head_t wq_procfs;
 
 	struct list_head per_thread_data_hash[MCCTRL_PER_THREAD_DATA_HASH_SIZE];
@@ -355,8 +343,6 @@ struct mcctrl_usrdata {
 	wait_queue_head_t wq_procfs;
 	struct list_head per_proc_data_hash[MCCTRL_PER_PROC_DATA_HASH_SIZE];
 	rwlock_t per_proc_data_hash_lock[MCCTRL_PER_PROC_DATA_HASH_SIZE];
-	struct list_head wakeup_descs_list;
-	spinlock_t wakeup_descs_lock;
 
 	void **keys;
 	struct sysfsm_data sysfsm_data;
@@ -382,60 +368,12 @@ 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);
 
-struct mcctrl_wakeup_desc {
-	int status;
-	int err;
-	wait_queue_head_t wq;
-	struct list_head chain;
-	int free_addrs_count;
-	void *free_addrs[];
-};
-
-/* ikc query-and-wait helper
- *
- * Arguments:
- *  - os, cpu and pisp as per mcctl_ikc_send()
- *  - timeout: time to wait for reply in ms
- *  - desc: if set, memory area to be used for desc.
- *    Care must be taken to leave room for variable-length array.
- *  - do_free: returns bool that specify if the caller should free
- *    its memory on error (e.g. if ikc_send failed in the first place,
- *    the reply has no chance of coming and memory should be free)
- *    Always true on success.
- *  - free_addrs_count & ...: addresses to kmalloc'd pointers that
- *    are referenced in the message and must be left intact if we
- *    abort to timeout/signal.
- */
-int mcctrl_ikc_send_wait(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp,
-	long int timeout, struct mcctrl_wakeup_desc *desc,
-	int *do_frees, int free_addrs_count, ...);
-
 ihk_os_t osnum_to_os(int n);
 
-/* look up symbols, plus arch-specific ones */
-extern int (*mcctrl_sys_mount)(char *dev_name, char *dir_name, char *type,
-			       unsigned long flags, void *data);
-extern int (*mcctrl_sys_umount)(char *dir_name, int flags);
-extern int (*mcctrl_sys_unshare)(unsigned long unshare_flags);
-extern long (*mcctrl_sched_setaffinity)(pid_t pid,
-					const struct cpumask *in_mask);
-extern int (*mcctrl_sched_setscheduler_nocheck)(struct task_struct *p,
-						int policy,
-						const struct sched_param *param);
-extern ssize_t (*mcctrl_sys_readlink)(const char *path, char *buf,
-				      size_t bufsiz);
-extern void (*mcctrl_zap_page_range)(struct vm_area_struct *vma,
-				     unsigned long start,
-				     unsigned long size,
-				     struct zap_details *details);
-extern struct inode_operations *mcctrl_hugetlbfs_inode_operations;
-
 /* syscall.c */
 void pager_add_process(void);
 void pager_remove_process(struct mcctrl_per_proc_data *ppd);
-void pager_cleanup(void);
 
-int __do_in_kernel_irq_syscall(ihk_os_t os, struct ikc_scd_packet *packet);
 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);
@@ -444,18 +382,20 @@ struct mcctrl_per_proc_data *mcctrl_get_per_proc_data(
 		struct mcctrl_usrdata *ud, int pid);
 void mcctrl_put_per_proc_data(struct mcctrl_per_proc_data *ppd);
 
-int mcctrl_add_per_thread_data(struct mcctrl_per_proc_data *ppd, void *data);
-void mcctrl_put_per_thread_data_unsafe(struct mcctrl_per_thread_data *ptd);
-void mcctrl_put_per_thread_data(struct mcctrl_per_thread_data* ptd);
+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);
 #ifdef POSTK_DEBUG_ARCH_DEP_56 /* Strange how to use inline declaration fix. */
-inline struct mcctrl_per_thread_data *mcctrl_get_per_thread_data(struct mcctrl_per_proc_data *ppd, struct task_struct *task)
+static inline struct mcctrl_per_thread_data *mcctrl_get_per_thread_data(
+	struct mcctrl_per_proc_data *ppd, struct task_struct *task)
 {
 	struct mcctrl_per_thread_data *ptd_iter, *ptd = NULL;
 	int hash = (((uint64_t)task >> 4) & MCCTRL_PER_THREAD_DATA_HASH_MASK);
 	unsigned long flags;
 
-	/* Check if data for this thread exists */
-	write_lock_irqsave(&ppd->per_thread_data_hash_lock[hash], flags);
+	/* Check if data for this thread exists and return it */
+	read_lock_irqsave(&ppd->per_thread_data_hash_lock[hash], flags);
 
 	list_for_each_entry(ptd_iter, &ppd->per_thread_data_hash[hash], hash) {
 		if (ptd_iter->task == task) {
@@ -464,27 +404,16 @@ inline struct mcctrl_per_thread_data *mcctrl_get_per_thread_data(struct mcctrl_p
 		}
 	}
 
-	if (ptd) {
-		if (atomic_read(&ptd->refcount) <= 0) {
-			printk("%s: ERROR: use-after-free detected (%d)", __FUNCTION__, atomic_read(&ptd->refcount));
-			ptd = NULL;
-			goto out;
-		}
-		atomic_inc(&ptd->refcount);
-	}
-
- out:
-	write_unlock_irqrestore(&ppd->per_thread_data_hash_lock[hash], flags);
-	return ptd;
+	read_unlock_irqrestore(&ppd->per_thread_data_hash_lock[hash], flags);
+	return ptd ? ptd->data : NULL;
 }
 #else /* POSTK_DEBUG_ARCH_DEP_56 */
-inline struct mcctrl_per_thread_data *mcctrl_get_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);
 #endif /* POSTK_DEBUG_ARCH_DEP_56 */
-int mcctrl_clear_pte_range(uintptr_t start, uintptr_t len);
 
 void __return_syscall(ihk_os_t os, struct ikc_scd_packet *packet, 
 		long ret, int stid);
-int clear_pte_range(uintptr_t start, uintptr_t len);
 
 int mcctrl_os_alive(void);
 
@@ -496,6 +425,7 @@ struct procfs_read {
 	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) */
@@ -508,8 +438,7 @@ struct procfs_file {
 };
 
 void procfs_answer(struct mcctrl_usrdata *ud, int pid);
-int procfsm_packet_handler(void *os, int msg, int pid, unsigned long arg,
-			   unsigned long resp_pa);
+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);
@@ -545,9 +474,7 @@ struct vdso {
 
 int reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp,
 		unsigned long *endp);
-int release_user_space(uintptr_t start, uintptr_t len);
 void get_vdso_info(ihk_os_t os, long vdso_pa);
-int arch_symbols_init(void);
 
 struct get_cpu_mapping_req {
 	int busy;		/* INOUT: */

executer/kernel/mcctrl/procfs.c

@@ -103,6 +103,33 @@ getpath(struct procfs_list_entry *e, char *buf, int bufsize)
 	}
 }
 
+/**
+ * \brief Process SCD_MSG_PROCFS_ANSWER message.
+ *
+ * \param ud mcctrl_usrdata pointer
+ * \param pid PID of the requesting process
+ */
+void procfs_answer(struct mcctrl_usrdata *ud, int pid)
+{
+	struct mcctrl_per_proc_data *ppd = NULL;
+
+	if (pid > 0) {
+		ppd = mcctrl_get_per_proc_data(ud, pid);
+
+		if (unlikely(!ppd)) {
+			kprintf("%s: ERROR: no per-process structure for PID %d\n",
+					__FUNCTION__, pid);
+			return;
+		}
+	}
+
+	wake_up_all(pid > 0 ? &ppd->wq_procfs : &ud->wq_procfs);
+
+	if (pid > 0) {
+		mcctrl_put_per_proc_data(ppd);
+	}
+}
+
 static struct procfs_list_entry *
 find_procfs_entry(struct procfs_list_entry *parent, const char *name)
 {
@@ -294,8 +321,6 @@ get_base_entry(int osnum)
 	if(!e){
 		e = add_procfs_entry(NULL, name, S_IFDIR | 0555,
 		                     uid, gid, NULL);
-		if (!e)
-			return NULL;
 		e->osnum = osnum;
 	}
 	return e;
@@ -431,8 +456,6 @@ proc_exe_link(int osnum, int pid, const char *path)
 
 		e = add_procfs_entry(parent, "exe", S_IFLNK | 0777, uid, gid,
 		                     path);
-		if (!e)
-			goto out;
 		e->data = kmalloc(strlen(path) + 1, GFP_KERNEL);
 		strcpy(e->data, path);
 		task = find_procfs_entry(parent, "task");
@@ -441,7 +464,6 @@ proc_exe_link(int osnum, int pid, const char *path)
 			                 uid, gid, path);
 		}
 	}
-out:
 	up(&procfs_file_list_lock);
 }
 
@@ -487,6 +509,7 @@ procfs_exit(int osnum)
  * This function conforms to the 2) way of fs/proc/generic.c
  * from linux-2.6.39.4.
  */
+#ifdef POSTK_DEBUG_TEMP_FIX_43 /* Fixed an issue that failed pread / pwrite of size larger than 4MB */
 static ssize_t __mckernel_procfs_read_write(
 		struct file *file,
 		char __user *buf, size_t nbytes,
@@ -497,7 +520,7 @@ static ssize_t __mckernel_procfs_read_write(
 	int order = 0;
 	volatile struct procfs_read *r = NULL;
 	struct ikc_scd_packet isp;
-	int ret, osnum, pid;
+	int ret, osnum, pid, retw;
 	unsigned long pbuf;
 	size_t count = nbytes;
 	size_t copy_size = 0;
@@ -592,11 +615,11 @@ static ssize_t __mckernel_procfs_read_write(
 
 	while (count > 0) {
 		int this_len = min_t(ssize_t, count, copy_size);
-		int do_free;
 
 		r->pbuf = pbuf;
 		r->eof = 0;
 		r->ret = -EIO; /* default */
+		r->status = 0;
 		r->offset = offset;
 		r->count = this_len;
 		r->readwrite = read_write;
@@ -606,26 +629,50 @@ static ssize_t __mckernel_procfs_read_write(
 		isp.arg = virt_to_phys(r);
 		isp.pid = pid;
 
-		ret = mcctrl_ikc_send_wait(osnum_to_os(e->osnum),
-					   (pid > 0) ? ppd->ikc_target_cpu : 0,
-					   &isp, HZ, NULL, &do_free, 1, r);
-
-		if (!do_free && ret >= 0) {
-			ret = -EIO;
-		}
+		ret = mcctrl_ikc_send(osnum_to_os(e->osnum),
+				(pid > 0) ? ppd->ikc_target_cpu : 0, &isp);
 
 		if (ret < 0) {
-			if (ret == -ETIME) {
-				pr_info("%s: error: timeout (1 sec)\n",
-				       __func__);
-			}
-			else if (ret == -ERESTARTSYS) {
-				ret = -ERESTART;
-			}
-			if (!do_free)
-				r = NULL;
+			goto out; /* error */
+		}
+
+		/* Wait for a reply. */
+		ret = -EIO; /* default exit code */
+		dprintk("%s: waiting for reply\n", __FUNCTION__);
+
+retry_wait:
+		/* Wait for the status field of the procfs_read structure,
+		 * wait on per-process or OS specific data depending on
+		 * who the request is for.
+		 */
+		if (pid > 0) {
+			retw = wait_event_interruptible_timeout(ppd->wq_procfs,
+					r->status != 0, HZ);
+		}
+		else {
+			retw = wait_event_interruptible_timeout(udp->wq_procfs,
+					r->status != 0, HZ);
+		}
+
+		/* Timeout? */
+		if (retw == 0 && r->status == 0) {
+			printk("%s: error: timeout (1 sec)\n", __FUNCTION__);
 			goto out;
 		}
+		/* Interrupted? */
+		else if (retw == -ERESTARTSYS) {
+			ret = -ERESTART;
+			goto out;
+		}
+		/* Were we woken up by a reply to another procfs request? */
+		else if (r->status == 0) {
+			/* TODO: r->status is not set atomically, we could be woken
+			 * up with status == 0 and it could change to 1 while in this
+			 * code, we could potentially miss the wake_up()... 
+			 */
+			printk("%s: stale wake-up, retrying\n", __FUNCTION__);
+			goto retry_wait;
+		}
 
 		/* Wake up and check the result. */
 		dprintk("%s: woke up. ret: %d, eof: %d\n",
@@ -670,6 +717,193 @@ out:
 
 	return ret;
 }
+#else /* POSTK_DEBUG_TEMP_FIX_43 */
+static ssize_t __mckernel_procfs_read_write(
+		struct file *file,
+		char __user *buf, size_t nbytes,
+		loff_t *ppos, int read_write)
+{
+	struct inode * inode = file->f_inode;
+	char *kern_buffer = NULL;
+	int order = 0;
+	volatile struct procfs_read *r = NULL;
+	struct ikc_scd_packet isp;
+	int ret, osnum, pid, retw;
+	unsigned long pbuf;
+	unsigned long count = nbytes;
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)
+	struct proc_dir_entry *dp = PDE(inode);
+	struct procfs_list_entry *e = dp->data;
+#else
+	struct procfs_list_entry *e = PDE_DATA(inode);
+#endif
+	loff_t offset = *ppos;
+	char pathbuf[PROCFS_NAME_MAX];
+	char *path, *p;
+	ihk_os_t os = NULL;
+	struct mcctrl_usrdata *udp = NULL;
+	struct mcctrl_per_proc_data *ppd = NULL;
+
+	if (count <= 0 || offset < 0) {
+		return 0;
+	}
+
+	path = getpath(e, pathbuf, PROCFS_NAME_MAX);
+	dprintk("%s: invoked for %s, offset: %lu, count: %lu\n",
+			__FUNCTION__, path,
+			(unsigned long)offset, count);
+
+	/* Verify OS number */
+	ret = sscanf(path, "mcos%d/", &osnum);
+	if (ret != 1) {
+		printk("%s: error: couldn't determine OS number\n", __FUNCTION__);
+		return -EINVAL;
+	}
+
+	if (osnum != e->osnum) {
+		printk("%s: error: OS numbers don't match\n", __FUNCTION__);
+		return -EINVAL;
+	}
+
+	/* Is this request for a specific process? */
+	p = strchr(path, '/') + 1;
+	ret = sscanf(p, "%d/", &pid);
+	if (ret != 1) {
+		pid = -1;
+	}
+
+	os = osnum_to_os(osnum);
+	if (!os) {
+		printk("%s: error: no IHK OS data found for OS %d\n",
+				__FUNCTION__, osnum);
+		return -EINVAL;
+	}
+
+	udp = ihk_host_os_get_usrdata(os);
+	if (!udp) {
+		printk("%s: error: no MCCTRL data found for OS %d\n",
+				__FUNCTION__, osnum);
+		return -EINVAL;
+	}
+
+	if (pid > 0) {
+		ppd = mcctrl_get_per_proc_data(udp, pid);
+
+		if (unlikely(!ppd)) {
+			printk("%s: error: no per-process structure for PID %d",
+					__FUNCTION__, pid);
+			return -EINVAL;
+		}
+	}
+
+	while ((1 << order) < count) ++order;
+	if (order > 12) {
+		order -= 12;
+	}
+	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("%s: ERROR: allocating kernel buffer\n", __FUNCTION__);
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	pbuf = virt_to_phys(kern_buffer);
+
+	r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
+	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 = read_write;
+	strncpy((char *)r->fname, path, PROCFS_NAME_MAX);
+	isp.msg = SCD_MSG_PROCFS_REQUEST;
+	isp.ref = 0;
+	isp.arg = virt_to_phys(r);
+	isp.pid = pid;
+
+	ret = mcctrl_ikc_send(osnum_to_os(e->osnum),
+			(pid > 0) ? ppd->ikc_target_cpu : 0, &isp);
+
+	if (ret < 0) {
+		goto out; /* error */
+	}
+
+	/* Wait for a reply. */
+	ret = -EIO; /* default exit code */
+	dprintk("%s: waiting for reply\n", __FUNCTION__);
+
+retry_wait:
+	/* Wait for the status field of the procfs_read structure,
+	 * wait on per-process or OS specific data depending on
+	 * who the request is for.
+	 */
+	if (pid > 0) {
+		retw = wait_event_interruptible_timeout(ppd->wq_procfs,
+				r->status != 0, 5 * HZ);
+	}
+	else {
+		retw = wait_event_interruptible_timeout(udp->wq_procfs,
+				r->status != 0, 5 * HZ);
+	}
+
+	/* Timeout? */
+	if (retw == 0 && r->status == 0) {
+		printk("%s: error: timeout (1 sec)\n", __FUNCTION__);
+		goto out;
+	}
+	/* Interrupted? */
+	else if (retw == -ERESTARTSYS) {
+		ret = -ERESTART;
+		goto out;
+	}
+	/* Were we woken up by a reply to another procfs request? */
+	else if (r->status == 0) {
+		/* TODO: r->status is not set atomically, we could be woken
+		 * up with status == 0 and it could change to 1 while in this
+		 * code, we could potentially miss the wake_up()... 
+		 */
+		printk("%s: stale wake-up, retrying\n", __FUNCTION__);
+		goto retry_wait;
+	}
+
+	/* Wake up and check the result. */
+	dprintk("%s: woke up. ret: %d, eof: %d\n",
+			__FUNCTION__, r->ret, r->eof);
+
+	if (r->ret > 0) {
+		if (read_write == 0) {
+			if (copy_to_user(buf, kern_buffer, r->ret)) {
+				printk("%s: ERROR: copy_to_user failed.\n", __FUNCTION__);
+				ret = -EFAULT;
+				goto out;
+			}
+		}
+		*ppos += r->ret;
+	}
+	ret = r->ret;
+
+out:
+	if (ppd)
+		mcctrl_put_per_proc_data(ppd);
+	if (kern_buffer)
+		free_pages((uintptr_t)kern_buffer, order);
+	if (r)
+		kfree((void *)r);
+
+	return ret;
+}
+#endif /* POSTK_DEBUG_TEMP_FIX_43 */
 
 static ssize_t mckernel_procfs_read(struct file *file,
 		char __user *buf, size_t nbytes, loff_t *ppos)
@@ -705,48 +939,33 @@ struct procfs_work {
 	int msg;
 	int pid;
 	unsigned long arg;
-	unsigned long resp_pa;
 	struct work_struct work;
 };
 
 static void procfsm_work_main(struct work_struct *work0)
 {
 	struct procfs_work *work = container_of(work0, struct procfs_work, work);
-	unsigned long phys;
-	int *done;
 
 	switch (work->msg) {
-	case SCD_MSG_PROCFS_TID_CREATE:
-		add_tid_entry(ihk_host_os_get_index(work->os),
-				work->pid, work->arg);
-		phys = ihk_device_map_memory(ihk_os_to_dev(work->os),
-					     work->resp_pa, sizeof(int));
-		done = ihk_device_map_virtual(ihk_os_to_dev(work->os),
-					      phys, sizeof(int), NULL, 0);
-		*done = 1;
-		ihk_device_unmap_virtual(ihk_os_to_dev(work->os),
-						 done, sizeof(int));
-		ihk_device_unmap_memory(ihk_os_to_dev(work->os),
-					phys, sizeof(int));
-		break;
+		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;
+		case SCD_MSG_PROCFS_TID_DELETE:
+			delete_tid_entry(ihk_host_os_get_index(work->os), work->pid, work->arg);
+			break;
 
-	default:
-		pr_warn("%s: unknown work: msg: %d, pid: %d, arg: %lu)\n",
-			__func__, work->msg, 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,
-			   unsigned long resp_pa)
+int procfsm_packet_handler(void *os, int msg, int pid, unsigned long arg)
 {
 	struct procfs_work *work = NULL;
 
@@ -760,7 +979,6 @@ int procfsm_packet_handler(void *os, int msg, int pid, unsigned long arg,
 	work->msg = msg;
 	work->pid = pid;
 	work->arg = arg;
-	work->resp_pa = resp_pa;
 	INIT_WORK(&work->work, &procfsm_work_main);
 
 	schedule_work(&work->work);
@@ -779,303 +997,6 @@ static const struct file_operations mckernel_forward = {
 	.write		= mckernel_procfs_write,
 };
 
-#define PA_NULL (-1L)
-
-struct mckernel_procfs_buffer_info {
-	unsigned long top_pa;
-	unsigned long cur_pa;
-	ihk_os_t os;
-	int pid;
-	char path[0];
-};
-
-struct mckernel_procfs_buffer {
-	unsigned long next_pa;
-	unsigned long pos;
-	unsigned long size;
-	char buf[0];
-};
-
-static int mckernel_procfs_buff_open(struct inode *inode, struct file *file)
-{
-	struct mckernel_procfs_buffer_info *info;
-	int pid;
-	int ret;
-	char *path;
-	char *path_buf;
-	char *p;
-	ihk_os_t os;
-#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
-
-	os = osnum_to_os(e->osnum);
-	if (!os) {
-		return -EINVAL;
-	}
-	path_buf = kmalloc(PROCFS_NAME_MAX, GFP_KERNEL);
-	if (!path_buf) {
-		return -ENOMEM;
-	}
-	path = getpath(e, path_buf, PROCFS_NAME_MAX);
-	p = strchr(path, '/') + 1;
-	ret = sscanf(p, "%d/", &pid);
-	if (ret != 1) {
-		pid = -1;
-	}
-
-	info = kmalloc(sizeof(struct mckernel_procfs_buffer_info) +
-		       strlen(path) + 1, GFP_KERNEL);
-	if (!info) {
-		kfree(path_buf);
-		return -ENOMEM;
-	}
-	info->top_pa = PA_NULL;
-	info->cur_pa = PA_NULL;
-	info->os = os;
-	info->pid = pid;
-	strcpy(info->path, path);
-	file->private_data = info;
-
-	kfree(path_buf);
-	return 0;
-}
-
-static int mckernel_procfs_buff_release(struct inode *inode, struct file *file)
-{
-	struct mckernel_procfs_buffer_info *info = file->private_data;
-	int rc = 0;
-
-	if (!info) {
-		return -EIO;
-	}
-
-	file->private_data = NULL;
-	if (info->top_pa != PA_NULL) {
-		int ret;
-		struct procfs_read *r = NULL;
-		struct ikc_scd_packet isp;
-		int do_free;
-
-		r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
-		if (r == NULL) {
-			rc = -ENOMEM;
-			goto out;
-		}
-		memset(r, '\0', sizeof(struct procfs_read));
-		r->pbuf = info->top_pa;
-		r->ret = -EIO; /* default */
-		r->fname[0] = '\0';
-		isp.msg = SCD_MSG_PROCFS_RELEASE;
-		isp.ref = 0;
-		isp.arg = virt_to_phys(r);
-		isp.pid = 0;
-
-		rc = -EIO;
-		ret = mcctrl_ikc_send_wait(info->os, 0,
-					   &isp, 5 * HZ, NULL, &do_free, 1, r);
-
-		if (!do_free && ret >= 0) {
-			ret = -EIO;
-		}
-
-		if (ret < 0) {
-			rc = ret;
-			if (ret == -ETIME) {
-				pr_info("%s: error: timeout (1 sec)\n",
-				       __func__);
-			}
-			else if (ret == -ERESTARTSYS) {
-				rc = -ERESTART;
-			}
-			if (!do_free)
-				r = NULL;
-			goto out;
-		}
-
-		if (r->ret < 0) {
-			rc = r->ret;
-			goto out;
-		}
-		rc = 0;
-out:
-		if (r)
-			kfree((void *)r);
-	}
-	kfree(info);
-	return rc;
-}
-
-static ssize_t mckernel_procfs_buff_read(struct file *file, char __user *ubuf,
-					 size_t nbytes, loff_t *ppos)
-{
-	struct mckernel_procfs_buffer_info *info = file->private_data;
-	unsigned long phys;
-	struct mckernel_procfs_buffer *buf;
-	int pos = *ppos;
-	ssize_t l = 0;
-	int done = 0;
-	ihk_os_t os;
-
-	if (nbytes <= 0 || *ppos < 0) {
-		return 0;
-	}
-
-	if (!info) {
-		return -EIO;
-	}
-
-	os = info->os;
-	if (info->top_pa == PA_NULL) {
-		int ret;
-		int pid = info->pid;
-		struct procfs_read *r = NULL;
-		struct ikc_scd_packet isp;
-		struct mcctrl_usrdata *udp = NULL;
-		struct mcctrl_per_proc_data *ppd = NULL;
-		int do_free;
-
-		udp = ihk_host_os_get_usrdata(os);
-		if (!udp) {
-			pr_err("%s: no MCCTRL data found for OS\n",
-					__func__);
-			return -EINVAL;
-		}
-
-		if (pid > 0) {
-			ppd = mcctrl_get_per_proc_data(udp, pid);
-
-			if (unlikely(!ppd)) {
-				pr_err("%s: no per-process structure for PID %d",
-						__func__, pid);
-				return -EINVAL;
-			}
-		}
-
-		r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
-		if (r == NULL) {
-			l = -ENOMEM;
-			done = 1;
-			goto out;
-		}
-		memset(r, '\0', sizeof(struct procfs_read));
-		r->pbuf = PA_NULL;
-		r->ret = -EIO; /* default */
-		strncpy((char *)r->fname, info->path, PROCFS_NAME_MAX);
-		isp.msg = SCD_MSG_PROCFS_REQUEST;
-		isp.ref = 0;
-		isp.arg = virt_to_phys(r);
-		isp.pid = pid;
-
-		l = -EIO;
-		done = 1;
-		ret = mcctrl_ikc_send_wait(os,
-					   (pid > 0) ? ppd->ikc_target_cpu : 0,
-					   &isp, 5 * HZ, NULL, &do_free, 1, r);
-
-		if (!do_free && ret >= 0) {
-			ret = -EIO;
-		}
-
-		if (ret < 0) {
-			l = ret;
-			if (ret == -ETIME) {
-				pr_info("%s: error: timeout (1 sec)\n",
-				       __func__);
-			}
-			else if (ret == -ERESTARTSYS) {
-				l = -ERESTART;
-			}
-			if (!do_free)
-				r = NULL;
-			goto out;
-		}
-
-		if (r->ret < 0) {
-			l = r->ret;
-			goto out;
-		}
-
-		done = 0;
-		l = 0;
-		info->top_pa = info->cur_pa = r->pbuf;
-
-out:
-		if (ppd)
-			mcctrl_put_per_proc_data(ppd);
-		if (r)
-			kfree((void *)r);
-	}
-
-	if (info->cur_pa == PA_NULL) {
-		info->cur_pa = info->top_pa;
-	}
-
-	while (!done && info->cur_pa != PA_NULL) {
-		long bpos;
-		long bsize;
-
-		phys = ihk_device_map_memory(ihk_os_to_dev(os), info->cur_pa,
-					     PAGE_SIZE);
-#ifdef CONFIG_MIC
-		buf = ioremap_wc(phys, PAGE_SIZE);
-#else
-		buf = ihk_device_map_virtual(ihk_os_to_dev(os), phys,
-					     PAGE_SIZE, NULL, 0);
-#endif
-
-		if (pos < buf->pos) {
-			info->cur_pa = info->top_pa;
-			goto rep;
-		}
-
-		if (pos >= buf->pos + buf->size) {
-			info->cur_pa = buf->next_pa;
-			goto rep;
-		}
-
-		bpos = pos - buf->pos;
-		bsize = (buf->pos + buf->size) - pos;
-		if (bsize > (nbytes - l)) {
-			bsize = nbytes - l;
-		}
-		if (copy_to_user(ubuf, buf->buf + bpos, bsize)) {
-			done = 1;
-			pos = *ppos;
-			l = -EFAULT;
-		}
-		else {
-			ubuf += bsize;
-			pos += bsize;
-			l += bsize;
-			if (l == nbytes) {
-				done = 1;
-			}
-		}
-rep:
-#ifdef CONFIG_MIC
-		iounmap(buf);
-#else
-		ihk_device_unmap_virtual(ihk_os_to_dev(os), buf, PAGE_SIZE);
-#endif
-		ihk_device_unmap_memory(ihk_os_to_dev(os), phys, PAGE_SIZE);
-	};
-
-	*ppos = pos;
-	return l;
-}
-
-static const struct file_operations mckernel_buff_io = {
-	.llseek		= mckernel_procfs_lseek,
-	.read		= mckernel_procfs_buff_read,
-	.write		= NULL,
-	.open		= mckernel_procfs_buff_open,
-	.release	= mckernel_procfs_buff_release,
-};
-
 static const struct procfs_entry tid_entry_stuff[] = {
 //	PROC_REG("auxv",       S_IRUSR, NULL),
 //	PROC_REG("clear_refs", S_IWUSR, NULL),
@@ -1085,10 +1006,10 @@ static const struct procfs_entry tid_entry_stuff[] = {
 //	PROC_LNK("exe",        mckernel_readlink),
 //	PROC_REG("limits",     S_IRUSR|S_IWUSR, NULL),
 //	PROC_REG("maps",       S_IRUGO, NULL),
-	PROC_REG("mem",        0600, NULL),
+	PROC_REG("mem",        S_IRUSR|S_IWUSR, NULL),
 //	PROC_REG("pagemap",    S_IRUGO, NULL),
 //	PROC_REG("smaps",      S_IRUGO, NULL),
-	PROC_REG("stat",       0444, &mckernel_buff_io),
+	PROC_REG("stat",       S_IRUGO, NULL),
 //	PROC_REG("statm",      S_IRUGO, NULL),
 //	PROC_REG("status",     S_IRUGO, NULL),
 //	PROC_REG("syscall",    S_IRUGO, NULL),
@@ -1097,26 +1018,26 @@ static const struct procfs_entry tid_entry_stuff[] = {
 };
 
 static const struct procfs_entry pid_entry_stuff[] = {
-	PROC_REG("auxv",       0400, &mckernel_buff_io),
+	PROC_REG("auxv",       S_IRUSR, NULL),
 	/* Support the case where McKernel process retrieves its job-id under the Fujitsu TCS suite. */
 //	PROC_REG("cgroup",     S_IXUSR, NULL),
 //	PROC_REG("clear_refs", S_IWUSR, NULL),
-	PROC_REG("cmdline",    0444, &mckernel_buff_io),
-	PROC_REG("comm",       0644, &mckernel_buff_io),
+	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_IRUGO, 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",       0444, &mckernel_buff_io),
-	PROC_REG("mem",        0400, NULL),
-	PROC_REG("pagemap",    0444, NULL),
-//	PROC_REG("smaps",      S_IRUGO, NULL),
-//	PROC_REG("stat",       0444, &mckernel_buff_io),
+	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",     0444, &mckernel_buff_io),
+	PROC_REG("status",     S_IRUGO, NULL),
 //	PROC_REG("syscall",    S_IRUGO, NULL),
-	PROC_DIR("task",       0555),
+	PROC_DIR("task",       S_IRUGO|S_IXUGO),
 //	PROC_REG("wchan",      S_IRUGO, NULL),
 	PROC_TERM
 };
@@ -1124,14 +1045,14 @@ static const struct procfs_entry pid_entry_stuff[] = {
 static const struct procfs_entry base_entry_stuff[] = {
 //	PROC_REG("cmdline",    S_IRUGO, NULL),
 #ifdef POSTK_DEBUG_ARCH_DEP_42 /* /proc/cpuinfo support added. */
-	PROC_REG("cpuinfo",    0444, &mckernel_buff_io),
+	PROC_REG("cpuinfo",    S_IRUGO, NULL),
 #else /* POSTK_DEBUG_ARCH_DEP_42 */
 //	PROC_REG("cpuinfo",    S_IRUGO, NULL),
 #endif /* POSTK_DEBUG_ARCH_DEP_42 */
 //	PROC_REG("meminfo",    S_IRUGO, NULL),
 //	PROC_REG("pagetypeinfo",S_IRUGO, NULL),
 //	PROC_REG("softirq",    S_IRUGO, NULL),
-	PROC_REG("stat",       0444, &mckernel_buff_io),
+	PROC_REG("stat",       S_IRUGO, NULL),
 //	PROC_REG("uptime",     S_IRUGO, NULL),
 //	PROC_REG("version",    S_IRUGO, NULL),
 //	PROC_REG("vmallocinfo",S_IRUSR, NULL),

executer/kernel/mcctrl/sysfs_files.c

@@ -790,7 +790,6 @@ out:
 	return error;
 } /* setup_node_files() */
 
-#ifdef SETUP_PCI_FILES
 static int read_file(void *buf, size_t size, char *fmt, va_list ap)
 {
 	int error;
@@ -799,6 +798,7 @@ static int read_file(void *buf, size_t size, char *fmt, va_list ap)
 	int n;
 	struct file *fp = NULL;
 	loff_t off;
+	mm_segment_t ofs;
 	ssize_t ss;
 
 	dprintk("read_file(%p,%ld,%s,%p)\n", buf, size, fmt, ap);
@@ -824,14 +824,13 @@ static int read_file(void *buf, size_t size, char *fmt, va_list ap)
 	}
 
 	off = 0;
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 14, 0)
-	ss = kernel_read(fp, buf, size, &off);
-#else
-	ss = kernel_read(fp, off, buf, size);
-#endif
+	ofs = get_fs();
+	set_fs(KERNEL_DS);
+	ss = vfs_read(fp, buf, size, &off);
+	set_fs(ofs);
 	if (ss < 0) {
 		error = ss;
-		eprintk("mcctrl:read_file:kernel_read failed. %d\n", error);
+		eprintk("mcctrl:read_file:vfs_read failed. %d\n", error);
 		goto out;
 	}
 	if (ss >= size) {
@@ -893,6 +892,16 @@ out:
 	return error;
 } /* read_long() */
 
+#ifdef MCCTRL_KSYM_sys_readlink
+static ssize_t (*mcctrl_sys_readlink)(const char *path, char *buf,
+		size_t bufsiz)
+#if MCCTRL_KSYM_sys_readlink
+	= (void *)MCCTRL_KSYM_sys_readlink;
+#else
+	= &sys_readlink;
+#endif
+#endif
+
 static int read_link(char *buf, size_t bufsize, char *fmt, ...)
 {
 	int error;
@@ -942,14 +951,30 @@ out:
 	return error;
 } /* read_link() */
 
+#ifdef POSTK_DEBUG_TEMP_FIX_22 /* iterate_dir() deadlock */
 static int setup_one_pci(struct mcctrl_usrdata *udp, const char *name)
 {
+#else /* POSTK_DEBUG_TEMP_FIX_22 */
+static int setup_one_pci(void *arg0, const char *name, int namlen,
+		loff_t offset, u64 ino, unsigned d_type)
+{
+	struct mcctrl_usrdata *udp = arg0;
+#endif /* POSTK_DEBUG_TEMP_FIX_22 */
 	int error;
 	char *buf = NULL;
 	long node;
 	struct sysfsm_bitmap_param param;
 
+#ifdef POSTK_DEBUG_TEMP_FIX_22 /* iterate_dir() deadlock */
 	dprintk("setup_one_pci(%p,%s)\n", udp, name);
+#else /* POSTK_DEBUG_TEMP_FIX_22 */
+	dprintk("setup_one_pci(%p,%s,%d,%#lx,%#lx,%d)\n",
+			arg0, name, namlen, (long)offset, (long)ino, d_type);
+	if (namlen != 12) {
+		error = 0;
+		goto out;
+	}
+#endif /* POSTK_DEBUG_TEMP_FIX_22 */
 
 	buf = (void *)__get_free_pages(GFP_KERNEL, 0);
 	if (!buf) {
@@ -1001,39 +1026,26 @@ static int setup_one_pci(struct mcctrl_usrdata *udp, const char *name)
 	error = 0;
 out:
 	free_pages((long)buf, 0);
+#ifdef POSTK_DEBUG_TEMP_FIX_22 /* iterate_dir() deadlock */
 	dprintk("setup_one_pci(%p,%s): %d\n", udp, name, error);
+#else /* POSTK_DEBUG_TEMP_FIX_22 */
+	dprintk("setup_one_pci(%p,%s,%d,%#lx,%#lx,%d): %d\n",
+			arg0, name, namlen, (long)offset, (long)ino, d_type,
+			error);
+#endif /* POSTK_DEBUG_TEMP_FIX_22 */
 	return error;
 } /* setup_one_pci() */
 
+#ifdef POSTK_DEBUG_TEMP_FIX_22 /* iterate_dir() deadlock */
 LIST_HEAD(pci_file_name_list);
 struct pci_file_name {
 	char *name;
 	struct list_head chain;
 };
 
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0) || \
-	(defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(7, 5))
-struct mcctrl_filler_args {
-	struct dir_context ctx;
-	void *buf;
-};
-
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 19, 0)
-static int pci_file_name_gen(struct dir_context *ctx, const char *name,
-		int namlen, loff_t offset, u64 ino, unsigned int d_type)
-#else
-static int pci_file_name_gen(void *ctx, const char *name,
-		int namlen, loff_t offset, u64 ino, unsigned int d_type)
-#endif
-{
-	struct mcctrl_filler_args *args
-		= container_of(ctx, struct mcctrl_filler_args, ctx);
-	void *buf = args->buf;
-#else
 static int pci_file_name_gen(void *buf, const char *name, int namlen,
 		loff_t offset, u64 ino, unsigned d_type)
 {
-#endif
 	struct pci_file_name *p;
 	int error = -1;
 
@@ -1071,31 +1083,56 @@ out:
 			buf, name, namlen, (long)offset, (long)ino, d_type, error);
 	return error;
 }
+#endif /* POSTK_DEBUG_TEMP_FIX_22 */
 
-static inline int mcctrl_vfs_readdir(struct file *file, filldir_t filler,
-				     void *buf)
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3,11,0)
+typedef int (*mcctrl_filldir_t)(void *buf, const char *name, int namlen,
+		loff_t offset, u64 ino, unsigned d_type);
+
+struct mcctrl_filler_args {
+	struct dir_context ctx;
+	mcctrl_filldir_t filler;
+	void *buf;
+};
+
+static int mcctrl_filler(struct dir_context *ctx, const char *name,
+		int namlen, loff_t offset, u64 ino, unsigned d_type)
+{
+	struct mcctrl_filler_args *args
+		= container_of(ctx, struct mcctrl_filler_args, ctx);
+
+	return (*args->filler)(args->buf, name, namlen, offset, ino, d_type);
+} /* mcctrl_filler() */
+
+static inline int mcctrl_vfs_readdir(struct file *file,
+		mcctrl_filldir_t filler, void *buf)
 {
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0) || \
-	(defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(7, 5))
 	struct mcctrl_filler_args args = {
-		.ctx.actor = filler,
+		.ctx.actor = &mcctrl_filler,
+		.filler = (void *)filler,
 		.buf = buf,
 	};
 
 	return iterate_dir(file, &args.ctx);
-#else
-	return vfs_readdir(file, filler, buf);
-#endif
 } /* mcctrl_vfs_readdir() */
+#else
+static inline int mcctrl_vfs_readdir(struct file *file, filldir_t filler,
+		void *buf)
+{
+	return vfs_readdir(file, filler, buf);
+} /* mcctrl_vfs_readdir() */
+#endif
 
 static int setup_pci_files(struct mcctrl_usrdata *udp)
 {
 	int error;
 	int er;
 	struct file *fp = NULL;
+#ifdef POSTK_DEBUG_TEMP_FIX_22 /* iterate_dir() deadlock */
 	int ret = 0;
 	struct pci_file_name *cur;
 	struct pci_file_name *next;
+#endif /* POSTK_DEBUG_TEMP_FIX_22 */
 
 	dprintk("setup_pci_files(%p)\n", udp);
 	fp = filp_open("/sys/bus/pci/devices", O_DIRECTORY, 0);
@@ -1105,13 +1142,18 @@ static int setup_pci_files(struct mcctrl_usrdata *udp)
 		goto out;
 	}
 
+#ifdef POSTK_DEBUG_TEMP_FIX_22 /* iterate_dir() deadlock */
 	error = mcctrl_vfs_readdir(fp, &pci_file_name_gen, udp);
+#else /* POSTK_DEBUG_TEMP_FIX_22 */
+	error = mcctrl_vfs_readdir(fp, &setup_one_pci, udp);
+#endif /* POSTK_DEBUG_TEMP_FIX_22 */
 	if (error) {
 		eprintk("mcctrl:setup_pci_files:"
 				"mcctrl_vfs_readdir failed. %d\n", error);
 		goto out;
 	}
 
+#ifdef POSTK_DEBUG_TEMP_FIX_22 /* iterate_dir() deadlock */
 	list_for_each_entry_safe(cur, next, &pci_file_name_list, chain) {
 		if (!ret) {
 			ret = setup_one_pci(udp, cur->name);
@@ -1120,6 +1162,7 @@ static int setup_pci_files(struct mcctrl_usrdata *udp)
 		kfree(cur->name);
 		kfree(cur);
 	}
+#endif /* POSTK_DEBUG_TEMP_FIX_22 */
 
 	error = 0;
 out:
@@ -1133,7 +1176,6 @@ out:
 	dprintk("setup_pci_files(%p): %d\n", udp, error);
 	return error;
 } /* setup_pci_files() */
-#endif // SETUP_PCI_FILES
 
 void setup_sysfs_files(ihk_os_t os)
 {
@@ -1173,9 +1215,7 @@ void setup_sysfs_files(ihk_os_t os)
 	setup_cpus_sysfs_files(udp);
 	setup_node_files(udp);
 	setup_cpus_sysfs_files_node_link(udp);
-#ifdef SETUP_PCI_FILES
-	setup_pci_files(udp);
-#endif
+	//setup_pci_files(udp);
 
 	/* Indicate sysfs files setup completion for boot script */
 	error = sysfsm_mkdirf(os, NULL, "/sys/setup_complete");

executer/kernel/mcoverlayfs/Makefile.in

@@ -21,7 +21,7 @@ endif
 endif
 ifeq ($(BUILD_MODULE_TMP),rhel)
 ifeq ($(BUILD_MODULE),none)
-BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -eq 199168 -a ${RHEL_RELEASE} -ge 327 -a ${RHEL_RELEASE} -le 862 ]; then echo "linux-3.10.0-327.36.1.el7"; else echo "none"; fi)
+BUILD_MODULE=$(shell if [ ${LINUX_VERSION_CODE} -eq 199168 -a ${RHEL_RELEASE} -ge 327 -a ${RHEL_RELEASE} -le 693 ]; then echo "linux-3.10.0-327.36.1.el7"; else echo "none"; fi)
 endif
 ifeq ($(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)

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

@@ -15,7 +15,6 @@
 #include <linux/rbtree.h>
 #include <linux/security.h>
 #include <linux/cred.h>
-#include <linux/version.h>
 #include "overlayfs.h"
 
 struct ovl_cache_entry {
@@ -35,18 +34,10 @@ struct ovl_dir_cache {
 	struct list_head entries;
 };
 
-/* vfs_readdir vs. iterate_dir compat */
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0) || \
-	(defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(7, 5))
-#define USE_ITERATE_DIR 1
-#endif
-
-#ifndef USE_ITERATE_DIR
 struct dir_context {
 	const filldir_t actor;
 	//loff_t pos;
 };
-#endif
 
 struct ovl_readdir_data {
 	struct dir_context ctx;
@@ -265,11 +256,7 @@ static inline int ovl_dir_read(struct path *realpath,
 	do {
 		rdd->count = 0;
 		rdd->err = 0;
-#ifdef USE_ITERATE_DIR
-		err = iterate_dir(realfile, &rdd->ctx);
-#else
 		err = vfs_readdir(realfile, rdd->ctx.actor, rdd);
-#endif
 		if (err >= 0)
 			err = rdd->err;
 	} while (!err && rdd->count);
@@ -378,22 +365,6 @@ static struct ovl_dir_cache *ovl_cache_get(struct dentry *dentry)
 	return cache;
 }
 
-#ifdef USE_ITERATE_DIR
-struct iterate_wrapper {
-	struct dir_context ctx;
-	filldir_t actor;
-	void *buf;
-};
-
-static int ovl_wrap_readdir(void *ctx, const char *name, int namelen,
-			    loff_t offset, u64 ino, unsigned int d_type)
-{
-	struct iterate_wrapper *w = ctx;
-
-	return w->actor(w->buf, name, namelen, offset, ino, d_type);
-}
-#endif
-
 static int ovl_readdir(struct file *file, void *buf, filldir_t filler)
 {
 	struct ovl_dir_file *od = file->private_data;
@@ -405,16 +376,7 @@ static int ovl_readdir(struct file *file, void *buf, filldir_t filler)
 		ovl_dir_reset(file);
 
 	if (od->is_real) {
-#ifdef USE_ITERATE_DIR
-		struct iterate_wrapper w = {
-			.ctx.actor = ovl_wrap_readdir,
-			.actor = filler,
-			.buf = buf,
-		};
-		res = iterate_dir(od->realfile, &w.ctx);
-#else
 		res = vfs_readdir(od->realfile, filler, buf);
-#endif
 		file->f_pos = od->realfile->f_pos;
 
 		return res;

executer/user/Makefile.in

@@ -13,8 +13,6 @@ KDIR ?= @KDIR@
 ARCH=@ARCH@
 CFLAGS=-Wall -O -I. -I$(VPATH)/arch/${ARCH} -I${IHKDIR} -I@abs_builddir@/../../../ihk/linux/include
 LDFLAGS=@LDFLAGS@
-CPPFLAGS_SYSCALL_INTERCEPT=@CPPFLAGS_SYSCALL_INTERCEPT@
-LDFLAGS_SYSCALL_INTERCEPT=@LDFLAGS_SYSCALL_INTERCEPT@
 RPATH=$(shell echo $(LDFLAGS)|awk '{for(i=1;i<=NF;i++){if($$i~/^-L/){w=$$i;sub(/^-L/,"-Wl,-rpath,",w);print w}}}')
 VPATH=@abs_srcdir@
 TARGET=mcexec libsched_yield ldump2mcdump.so
@@ -23,17 +21,12 @@ LIBS=@LIBS@
 IHKDIR ?= $(VPATH)/../../../ihk/linux/include/
 MCEXEC_LIBS=-lmcexec -lrt -lnuma -pthread -L@abs_builddir@/../../../ihk/linux/user -lihk -Wl,-rpath,$(MCKERNEL_LIBDIR)
 ENABLE_QLMPI=@ENABLE_QLMPI@
-WITH_SYSCALL_INTERCEPT=@WITH_SYSCALL_INTERCEPT@
 
 ifeq ($(ENABLE_QLMPI),yes)
 	MCEXEC_LIBS += -lmpi
 	TARGET+= libqlmpi.so ql_server ql_mpiexec_start ql_mpiexec_finalize ql_talker libqlfort.so
 endif
 
-ifeq ($(WITH_SYSCALL_INTERCEPT),yes)
-	TARGET += syscall_intercept.so
-endif
-
 ifeq ($(ARCH), arm64)
 CFLAGS += $(foreach i, $(shell seq 1 100), $(addprefix -DPOSTK_DEBUG_ARCH_DEP_, $(i)))
 CFLAGS += $(foreach i, $(shell seq 1 100), $(addprefix -DPOSTK_DEBUG_TEMP_FIX_, $(i)))
@@ -47,7 +40,7 @@ mcexec: mcexec.c libmcexec.a
 # POSTK_DEBUG_ARCH_DEP_34, eclair arch depend separate.
 ifeq ($(ARCH), arm64)
 eclair: eclair.c arch/$(ARCH)/arch-eclair.c
-	$(CC) -I.. -I. -I./arch/$(ARCH)/include -I$(VPATH)/.. -I$(VPATH) -I$(VPATH)/arch/$(ARCH)/include $(CFLAGS) -o $@ $^ $(LIBS) -ldl -lz
+	$(CC) -I.. -I. -I./arch/$(ARCH)/include -I$(VPATH)/.. -I$(VPATH) -I$(VPATH)/arch/$(ARCH)/include $(CFLAGS) -o $@ $^ $(LIBS)
 else
 eclair: eclair.c arch/$(ARCH)/arch-eclair.c
 	$(CC) -I.. -I$(VPATH) -I$(VPATH)/arch/$(ARCH)/include $(CFLAGS) -o $@ $^ $(LIBS)
@@ -59,12 +52,6 @@ ldump2mcdump.so: ldump2mcdump.c
 libsched_yield: libsched_yield.c
 	$(CC) -shared -fPIC -Wl,-soname,sched_yield.so.1 -o libsched_yield.so.1.0.0 $^ -lc -ldl
 
-syscall_intercept.so: syscall_intercept.c libsyscall_intercept_arch.a
-	$(CC) $(CPPFLAGS_SYSCALL_INTERCEPT) -g -O2 $(LDFLAGS_SYSCALL_INTERCEPT) -lsyscall_intercept -fpic -shared -L. -lsyscall_intercept_arch $^ -o $@ 
-
-libsyscall_intercept_arch.a::
-	+(cd arch/${ARCH}; $(MAKE))
-
 libmcexec.a::
 	+(cd arch/${ARCH}; $(MAKE))
 
@@ -112,9 +99,6 @@ ifeq ($(ENABLE_QLMPI),yes)
 	install -m 755 ql_mpiexec_start $(BINDIR)
 	install -m 755 ql_mpiexec_finalize $(BINDIR)
 	install -m 755 ql_talker $(SBINDIR)
-endif
-ifeq ($(WITH_SYSCALL_INTERCEPT),yes)
-	install -m 755 syscall_intercept.so $(MCKERNEL_LIBDIR)
 endif
 	@uncomment_if_ENABLE_MEMDUMP@install -m 755 eclair $(BINDIR)
 	@uncomment_if_ENABLE_MEMDUMP@install -m 755 vmcore2mckdump $(BINDIR)

executer/user/arch/x86_64/Makefile.in

@@ -4,7 +4,7 @@ BINDIR=@BINDIR@
 KDIR ?= @KDIR@
 CFLAGS=-Wall -O -I.
 VPATH=@abs_srcdir@
-TARGET=../../libmcexec.a ../../libsyscall_intercept_arch.a
+TARGET=../../libmcexec.a
 LIBS=@LIBS@
 
 all: $(TARGET)
@@ -18,12 +18,6 @@ archdep.o: archdep.S
 arch_syscall.o: arch_syscall.c
 	$(CC) -c -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -pthread $<
 
-../../libsyscall_intercept_arch.a: archdep_c.o
-	$(AR) cr ../../libsyscall_intercept_arch.a archdep_c.o
-
-archdep_c.o: archdep_c.c
-	$(CC) -c -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -pthread $<
-
 clean:
 	$(RM) $(TARGET) *.o
 

executer/user/arch/x86_64/arch-eclair.c

@@ -42,7 +42,7 @@ int print_kregs(char *rbp, size_t rbp_size, const struct arch_kregs *kregs)
 	}
 
 	for (i = 0; i < sizeof(regs_1)/sizeof(regs_1[0]); i++) {	/* rsi, rdi, rbp, rsp */
-		ret = print_bin(rbp, rbp_size, regs_1 + i, sizeof(regs_1[0]));
+		ret = print_bin(rbp, rbp_size, (void *)regs_1[i], sizeof(regs_1[0]));
 		if (ret < 0) {
 			return ret;
 		}
@@ -62,7 +62,7 @@ int print_kregs(char *rbp, size_t rbp_size, const struct arch_kregs *kregs)
 	}
 
 	for (i = 0; i < sizeof(regs_2)/sizeof(regs_2[0]); i++) {	/* r12-r15 */
-		ret = print_bin(rbp, rbp_size, regs_2 + i, sizeof(regs_2[0]));
+		ret = print_bin(rbp, rbp_size, (void *)regs_2[i], sizeof(regs_2[0]));
 		if (ret < 0) {
 			return ret;
 		}

executer/user/arch/x86_64/arch_args.h

@@ -67,12 +67,6 @@ get_syscall_arg6(syscall_args *args)
 	return args->r9;
 }
 
-static inline unsigned long
-get_syscall_rip(syscall_args *args)
-{
-	return args->rip;
-}
-
 static inline void
 set_syscall_number(syscall_args *args, unsigned long value)
 {

executer/user/arch/x86_64/arch_syscall.c

@@ -48,7 +48,7 @@ archdep_syscall(struct syscall_wait_desc *w, long *ret)
 			if (*ret >= PATH_MAX) {
 				*ret = -ENAMETOOLONG;
 			}
-			if (*ret < 0) {
+			if (ret < 0) {
 				return 0;
 			}
 			__dprintf("open: %s\n", pathbuf);

executer/user/arch/x86_64/archdep.S

@@ -1,22 +1,15 @@
 /*
-Calling convention:
-  arg: rdi, rsi, rdx, rcx, r8, r9
-  ret: rax
+arg: rdi, rsi, rdx, rcx, r8, r9
+ret: rax
 
-  rdi: fd
-  rsi: cmd
-  rdx: param
-  rcx: save area
-  r8: new thread context
-
-Syscam call convention:
-  syscall number: rax
-  arg: rdi, rsi, rdx, r10, r8, r9
-  return addr: rcx
-
-  rdi:	 fd
-  rsi:	cmd
-  rdx:	param
+rax syscall number
+syscall: (rax:num) rdi rsi rdx r10 r8 r9 (rcx:ret addr)
+fd, cmd, param
+rdi: fd
+rsi: cmd
+rdx: param
+rcx: save area
+r8: new thread context
 */
 
 .global switch_ctx
@@ -98,7 +91,6 @@ switch_ctx:
 
 1:
 	mov	$0xffffffffffffffff,%eax
-	retq
 2:
 	pushq	%rax
 	movq	$158,%rax /* arch_prctl */
@@ -154,3 +146,4 @@ compare_and_swap_int:
 	lock
 	cmpxchgl	%edx,0(%rdi)
 	retq
+

executer/user/arch/x86_64/archdep_c.c

@@ -1,52 +0,0 @@
-/*
-function call convention
-rdi, rsi, rdx, rcx, r8, r9: 	IN	arguments
-rax:				OUT	return value
-
-syscall convention:
-rax:				IN	syscall number
-rdi, rsi, rdx, r10, r8, r9:	IN	arguments
-rax:				OUT	return value
-rcx, r11:			CLOBBER
-*/
-long uti_syscall6(long syscall_number, long arg0, long arg1, long arg2, long arg3, long arg4, long arg5)
-{
-	long ret;
-	asm volatile ("movq %[arg3],%%r10; movq %[arg4],%%r8; movq %[arg5],%%r9; syscall"
-				  : "=a" (ret)
-				  : "a" (syscall_number),
-					"D" (arg0), "S" (arg1), "d" (arg2),
-					[arg3] "g" (arg3), [arg4] "g" (arg4), [arg5] "g" (arg5)
-				  : "rcx", "r11", "r10", "r8", "r9", "memory");
-	return ret;
-}
-
-long uti_syscall3(long syscall_number, long arg0, long arg1, long arg2)
-{
-	long ret;
-	asm volatile ("syscall"
-				  : "=a" (ret)
-				  : "a" (syscall_number), "D" (arg0), "S" (arg1), "d" (arg2)
-				  : "rcx", "r11", "memory");
-	return ret;
-}
-
-long uti_syscall1(long syscall_number, long arg0)
-{
-	long ret;
-	asm volatile ("syscall"
-				  : "=a" (ret)
-				  : "a" (syscall_number), "D" (arg0)
-				  : "rcx", "r11", "memory");
-	return ret;
-}
-
-long uti_syscall0(long syscall_number)
-{
-	long ret;
-	asm volatile ("syscall"
-				  : "=a" (ret)
-				  : "a" (syscall_number)
-				  : "rcx", "r11", "memory");
-	return ret;
-}

executer/user/archdep.h

@@ -1,6 +1,4 @@
-#include "../include/uprotocol.h"
-
-extern int switch_ctx(int fd, unsigned long cmd, struct uti_save_fs_desc *desc, void *lctx, void *rctx);
+extern int switch_ctx(int fd, unsigned long cmd, void **param, void *lctx, void *rctx);
 extern unsigned long compare_and_swap(unsigned long *addr, unsigned long old, unsigned long new);
 extern unsigned int compare_and_swap_int(unsigned int *addr, unsigned int old, unsigned int new);
 extern int archdep_syscall(struct syscall_wait_desc *w, long *ret);

executer/user/archdep_uti.h

@@ -1,5 +0,0 @@
-extern long uti_syscall6(long syscall_number, long arg0, long arg1, long arg2, long arg3, long arg4, long arg5);
-extern long uti_syscall3(long syscall_number, long arg0, long arg1, long arg2);
-extern long uti_syscall1(long syscall_number, long arg0);
-extern long uti_syscall0(long syscall_number);
- 

executer/user/libsched_yield.c

@@ -11,9 +11,7 @@
 
 typedef int (*int_void_fn)(void);
 
-#if 0
 static int_void_fn orig_sched_yield = 0;
-#endif
 
 int sched_yield(void)
 {

executer/user/mcexec.1in

@@ -73,6 +73,13 @@ e.g.: 10k means 10Kibyte, 100M 100Mibyte, 1G 1Gibyte
 Enable system call profiling. After the execution, profiling
 information may be obtained by the ihkosctl tool.
 
+.TP
+.B	-m N
+Specify the NUMA memory policy. In the case of Quadrant&Flat mode, NUMA node
+0 is CPU cores and NUMA node 1 is MCDRAM. Thus, option "-m 1"
+means that user's memory areas are assigned in MCDRAM.
+
+
 .TP
 .B	--mpol-no-heap, --mpol-no-stack, --mpol-no-bss, 
 Disregard NUMA memory policy in the heap/stack/BSS areas.
@@ -93,7 +100,7 @@ This option eliminates potential kernel resource contention by
 avoiding page faults in the shared memory region.
 
 .TP
-.B	-m N, --mpol-threshold=N
+.B	-M N, --mpol-threshold=N
 Specify the threshold of memory size for respecting the memory
 allocation policy in NUMA machines. If the size of memory allocation
 is smaller than the one specified in this option, the memory area is

executer/user/syscall_intercept.c

@@ -1,139 +0,0 @@
-#include <libsyscall_intercept_hook_point.h>
-#include <errno.h>
-#include <stdio.h>
-#include <stdint.h>
-#include <syscall.h>
-#include <sys/time.h>
-#include <sys/resource.h>
-#include "../include/uprotocol.h"
-#include "../include/uti.h"
-#include "./archdep_uti.h"
-
-static struct uti_desc uti_desc;
-
-#define DEBUG_UTI
-
-static int
-hook(long syscall_number,
-	 long arg0, long arg1,
-	 long arg2, long arg3,
-	 long arg4, long arg5,
-	 long *result)
-{
-	//return 1; /* debug */
-	int tid = uti_syscall0(__NR_gettid);
-	struct terminate_thread_desc term_desc;
-	unsigned long code;
-	int stack_top;
-		
-	if (!uti_desc.start_syscall_intercept) {
-		return 1; /* System call isn't taken over */
-	}
-	if (tid != uti_desc.mck_tid) {
-		if (uti_desc.syscalls2 && syscall_number >= 0 && syscall_number < 512) {
-			uti_desc.syscalls2[syscall_number]++;
-		}
-		return 1;
-	}
-#ifdef DEBUG_UTI
-	if (uti_desc.syscalls && syscall_number >= 0 && syscall_number < 512) {
-		uti_desc.syscalls[syscall_number]++;
-	}
-#endif
-
-	switch (syscall_number) {
-	case __NR_gettid:
-		*result = uti_desc.mck_tid;
-		return 0;
-	case __NR_futex:
-	case __NR_brk:
-	case __NR_mmap:
-	case __NR_munmap:
-	case __NR_mprotect:
-	case __NR_mremap:
-		/* Overflow check */
-		if (uti_desc.syscall_stack_top == -1) {
-			*result = -ENOMEM;
-			return 0;
-		}
-	
-		/* Sanity check */
-		if (uti_desc.syscall_stack_top < 0 || uti_desc.syscall_stack_top >= UTI_SZ_SYSCALL_STACK) {
-			*result = -EINVAL;
-			return 0;
-		}
-
-		/* Store the return value in the stack to prevent it from getting corrupted 
-		   when an interrupt happens just after ioctl() and before copying the return
-		   value to *result */
-		stack_top = __sync_fetch_and_sub(&uti_desc.syscall_stack_top, 1);
-		
-		uti_desc.syscall_stack[stack_top].number = syscall_number;
-		uti_desc.syscall_stack[stack_top].args[0] = arg0;
-		uti_desc.syscall_stack[stack_top].args[1] = arg1;
-		uti_desc.syscall_stack[stack_top].args[2] = arg2;
-		uti_desc.syscall_stack[stack_top].args[3] = arg3;
-		uti_desc.syscall_stack[stack_top].args[4] = arg4;
-		uti_desc.syscall_stack[stack_top].args[5] = arg5;
-		uti_desc.syscall_stack[stack_top].uti_clv = uti_desc.uti_clv;
-		uti_desc.syscall_stack[stack_top].ret = -EINVAL;
-
-		uti_syscall3(__NR_ioctl, uti_desc.fd, MCEXEC_UP_SYSCALL_THREAD, (long)(uti_desc.syscall_stack + stack_top));
-		*result = uti_desc.syscall_stack[stack_top].ret;
-		
-		/* push syscall_struct list */
-		__sync_fetch_and_add(&uti_desc.syscall_stack_top, 1);
-
-		return 0; /* System call is taken over */
-	case __NR_exit_group:
-		code = 0x100000000;
-		goto make_remote_thread_exit;
-	case __NR_exit:
-		code = 0;
-	make_remote_thread_exit:
-		/* Make migrated-to-Linux thread on the McKernel side call do_exit() or terminate() */
-		term_desc.pid = uti_desc.pid;
-		term_desc.tid = uti_desc.tid; /* tid of mcexec */
-		term_desc.code = code | ((arg0 & 255) << 8);
-		term_desc.tsk = uti_desc.key;
-
-		uti_syscall3(__NR_ioctl, uti_desc.fd, MCEXEC_UP_TERMINATE_THREAD, (long)&term_desc);
-		return 1;
-	case __NR_clone:
-	case __NR_fork:
-	case __NR_vfork:
-	case __NR_execve:
-		*result = -ENOSYS;
-		return 0;
-#if 0 /* debug */
-	case __NR_set_robust_list:
-		*result = -ENOSYS;
-		return 0;
-#endif
-	case 888:
-        *result = (long)&uti_desc;
-		return 0;
-	default:
-		return 1;
-	}
-
-	return 0;
-}
-
-static __attribute__((constructor)) void
-init(void)
-{
-	/* Set up the callback function */
-	intercept_hook_point = hook;
-
-	/* Initialize uti_desc */
-	uti_desc.syscall_stack_top = UTI_SZ_SYSCALL_STACK - 1;
-
-	/* Pass address of uti_desc to McKernel */
-	uti_syscall1(733, (unsigned long)&uti_desc);
-}
-
-static __attribute__((destructor)) void
-dtor(void)
-{
-}

kernel/Makefile.build.in

@@ -6,8 +6,9 @@ IHKDIR=$(IHKBASE)/$(TARGETDIR)
 OBJS = init.o mem.o debug.o mikc.o listeners.o ap.o syscall.o cls.o host.o
 OBJS += process.o copy.o waitq.o futex.o timer.o plist.o fileobj.o shmobj.o
 OBJS += zeroobj.o procfs.o devobj.o sysfs.o xpmem.o profile.o freeze.o
-OBJS += rbtree.o hugefileobj.o
+OBJS += rbtree.o
 OBJS += pager.o
+OBJS += file_ops.o user_sdma.o sdma.o user_exp_rcv.o chip.o
 # POSTK_DEBUG_ARCH_DEP_18 coredump arch separation.
 DEPSRCS=$(wildcard $(SRC)/*.c)
 
@@ -19,7 +20,7 @@ endif
 
 CFLAGS += -I$(SRC)/include -I@abs_builddir@/../ -I@abs_builddir@/include -D__KERNEL__  -g -fno-omit-frame-pointer -fno-inline -fno-inline-small-functions
 ifneq ($(ARCH), arm64)
-CFLAGS += -mcmodel=large -mno-red-zone -mno-sse
+CFLAGS += -mcmodel=large -mno-red-zone
 endif
 LDFLAGS += -e arch_start
 IHKOBJ = ihk/ihk.o

kernel/ap.c

@@ -29,13 +29,15 @@
 #include <time.h>
 #include <syscall.h>
 #include <rusage_private.h>
-#include <debug.h>
 
 //#define DEBUG_PRINT_AP
 
 #ifdef DEBUG_PRINT_AP
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf(...) do { kprintf(__VA_ARGS__); } while (0)
+#define ekprintf(...) do { kprintf(__VA_ARGS__); } while (0)
+#else
+#define dkprintf(...) do { } while (0)
+#define ekprintf(...) do { kprintf(__VA_ARGS__); } while (0)
 #endif
 
 int num_processors = 1;
@@ -66,6 +68,11 @@ static void ap_wait(void)
 		init_host_ikc2mckernel();
 		init_host_ikc2linux(ikc_cpu);
 		mcs_lock_unlock_noirq(&ap_syscall_semaphore, &mcs_node);
+
+		{
+			extern void hfi1_kmalloc_cache_prealloc(void);
+			hfi1_kmalloc_cache_prealloc();
+		}
 	}
 	
 	/* one of them listens */
@@ -207,10 +214,8 @@ store_fake_cpu_info(struct sysfs_ops *ops0, void *instance, void *buf,
 
 static struct fake_cpu_info_ops show_fci_online = {
 	.member = ONLINE,
-	.ops = {
-		.show = &show_fake_cpu_info,
-		.store = &store_fake_cpu_info,
-	},
+	.ops.show = &show_fake_cpu_info,
+	.ops.store = &store_fake_cpu_info,
 };
 
 void

kernel/chip.c

@@ -0,0 +1,126 @@
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+/*
+ * This file contains all of the code that is specific to the HFI chip,
+ * or what we use of them.
+ */
+
+#include <hfi1/hfi.h>
+#include <hfi1/chip_registers.h>
+#include <hfi1/chip.h>
+
+//#define DEBUG_PRINT_CHIP
+
+#ifdef DEBUG_PRINT_CHIP
+#define dkprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if(0) kprintf(__VA_ARGS__); } while (0)
+#endif
+
+
+/*
+ * index is the index into the receive array
+ */
+void hfi1_put_tid(struct hfi1_devdata *dd, u32 index,
+		  u32 type, unsigned long pa, u16 order)
+{
+	u64 reg;
+	void __iomem *base = (dd->rcvarray_wc ? dd->rcvarray_wc :
+			      (dd->kregbase1 + RCV_ARRAY));
+
+	if (!(dd->flags & HFI1_PRESENT))
+		goto done;
+
+	if (type == PT_INVALID) {
+		pa = 0;
+	} else if (type > PT_INVALID) {
+		kprintf("unexpected receive array type %u for index %u, not handled\n",
+			type, index);
+		goto done;
+	}
+
+#ifdef TIDRDMA_DEBUG
+	hfi1_cdbg(TID, "type %s, index 0x%x, pa 0x%lx, bsize 0x%lx",
+		  pt_name(type), index, pa, (unsigned long)order);
+#endif
+
+#define RT_ADDR_SHIFT 12	/* 4KB kernel address boundary */
+	reg = RCV_ARRAY_RT_WRITE_ENABLE_SMASK
+		| (u64)order << RCV_ARRAY_RT_BUF_SIZE_SHIFT
+		| ((pa >> RT_ADDR_SHIFT) & RCV_ARRAY_RT_ADDR_MASK)
+					<< RCV_ARRAY_RT_ADDR_SHIFT;
+	dkprintf("type %d, index 0x%x, pa 0x%lx, bsize 0x%lx, reg 0x%llx\n",
+		type, index, pa, (unsigned long)order, reg);
+	writeq(reg, base + (index * 8));
+
+	if (type == PT_EAGER)
+		/*
+		 * Eager entries are written one-by-one so we have to push them
+		 * after we write the entry.
+		 */
+		flush_wc();
+done:
+	return;
+}
+
+void hfi1_clear_tids(struct hfi1_ctxtdata *rcd)
+{
+	struct hfi1_devdata *dd = rcd->dd;
+	u32 i;
+
+#if 0
+	/* this could be optimized */
+	for (i = rcd->eager_base; i < rcd->eager_base +
+		     rcd->egrbufs.alloced; i++)
+		hfi1_put_tid(dd, i, PT_INVALID, 0, 0);
+#endif
+	for (i = rcd->expected_base;
+			i < rcd->expected_base + rcd->expected_count; i++)
+		hfi1_put_tid(dd, i, PT_INVALID, 0, 0);
+}
+

kernel/config/attached-mic.lds

@@ -1,28 +1,24 @@
 PHDRS 
 {
 	text PT_LOAD FLAGS(5);
-	data PT_LOAD FLAGS(7);
+    data PT_LOAD FLAGS(7);
 }
 SECTIONS
 {
 	. = 0xffffffff80001000;
-	_head = .;
+    _head = .;
 
-	.text : {
-		*(.text);
-	} : text
+    .text : {
+        *(.text);
+    } : text
 
-	. = ALIGN(4096);
+    . = ALIGN(4096);
 	.data : {
-		*(.data)
-		*(.data.*)
-		. = ALIGN(8);
-		__start___verbose = .;
-		*(__verbose);
-		__stop___verbose = .;
+	      *(.data)
+	      *(.data.*)
 	} :data
 	.rodata : {
-		*(.rodata .rodata.*)
+	      *(.rodata .rodata.*)
 	} :data
 
 	.vsyscall : ALIGN(0x1000) {
@@ -41,14 +37,14 @@ SECTIONS
 		. = ALIGN(4096);
 	} : data = 0xf4
 
-	.bss : {
-		*(.bss .bss.*)
-	}
-	. = ALIGN(4096);
-	_end = .;
+    .bss : {
+          *(.bss .bss.*)
+    }
+    . = ALIGN(4096);
+    _end = .;
 
 	/DISCARD/ : {
-		*(.eh_frame)
-		*(.note.gnu.build-id)
+	     *(.eh_frame)
+         *(.note.gnu.build-id)
 	}
 }

kernel/config/builtin-mic.lds

@@ -1,28 +1,24 @@
 PHDRS 
 {
 	text PT_LOAD FLAGS(5);
-	data PT_LOAD FLAGS(7);
+    data PT_LOAD FLAGS(7);
 }
 SECTIONS
 {
 	. = 0xffffffff80001000;
-	_head = .;
+    _head = .;
 
-	.text : {
-		*(.text);
-	} : text
+    .text : {
+        *(.text);
+    } : text
 
-	. = ALIGN(4096);
+    . = ALIGN(4096);
 	.data : {
-		*(.data)
-		*(.data.*)
-		. = ALIGN(8);
-		__start___verbose = .;
-		*(__verbose);
-		__stop___verbose = .;
+	      *(.data)
+	      *(.data.*)
 	} :data
 	.rodata : {
-		*(.rodata .rodata.*)
+	      *(.rodata .rodata.*)
 	} :data
 
 	.vsyscall : ALIGN(0x1000) {
@@ -41,14 +37,14 @@ SECTIONS
 		. = ALIGN(4096);
 	} : data = 0xf4
 
-	.bss : {
-		*(.bss .bss.*)
-	}
-	. = ALIGN(4096);
-	_end = .;
+    .bss : {
+          *(.bss .bss.*)
+    }
+    . = ALIGN(4096);
+    _end = .;
 
 	/DISCARD/ : {
-		*(.eh_frame)
-		*(.note.gnu.build-id)
+	     *(.eh_frame)
+         *(.note.gnu.build-id)
 	}
 }

kernel/config/builtin-x86.lds

@@ -1,28 +1,24 @@
 PHDRS 
 {
 	text PT_LOAD FLAGS(5);
-	data PT_LOAD FLAGS(7);
+    data PT_LOAD FLAGS(7);
 }
 SECTIONS
 {
 	. = 0xffffffff80001000;
-	_head = .;
+    _head = .;
 
-	.text : {
-		*(.text);
-	} : text
+    .text : {
+        *(.text);
+    } : text
 
-	. = ALIGN(4096);
+    . = ALIGN(4096);
 	.data : {
-		*(.data)
-		*(.data.*)
-		. = ALIGN(8);
-		__start___verbose = .;
-		*(__verbose);
-		__stop___verbose = .;
+	      *(.data)
+	      *(.data.*)
 	} :data
 	.rodata : {
-		*(.rodata .rodata.*)
+	      *(.rodata .rodata.*)
 	} :data
 
 	.vsyscall : ALIGN(0x1000) {
@@ -41,10 +37,10 @@ SECTIONS
 		. = ALIGN(4096);
 	} : data = 0xf4
 
-	.bss : {
-		*(.bss .bss.*)
-	}
-	. = ALIGN(4096);
-	_end = .;
+    .bss : {
+          *(.bss .bss.*)
+    }
+    . = ALIGN(4096);
+    _end = .;
 
 }

kernel/config/smp-arm64_type1.lds

@@ -16,10 +16,6 @@ SECTIONS
 	.data : {
 		*(.data)
 		*(.data.*)
-		. = ALIGN(8);
-		__start___verbose = .;
-		*(__verbose);
-		__stop___verbose = .;
 	} :data
 	.rodata : {
 		*(.rodata .rodata.*)

kernel/config/smp-arm64_type2.lds

@@ -16,10 +16,6 @@ SECTIONS
 	.data : {
 		*(.data)
 		*(.data.*)
-		. = ALIGN(8);
-		__start___verbose = .;
-		*(__verbose);
-		__stop___verbose = .;
 	} :data
 	.rodata : {
 		*(.rodata .rodata.*)

kernel/config/smp-arm64_type3.lds

@@ -16,10 +16,6 @@ SECTIONS
 	.data : {
 		*(.data)
 		*(.data.*)
-		. = ALIGN(8);
-		__start___verbose = .;
-		*(__verbose);
-		__stop___verbose = .;
 	} :data
 	.rodata : {
 		*(.rodata .rodata.*)

kernel/config/smp-arm64_type4.lds

@@ -16,10 +16,6 @@ SECTIONS
 	.data : {
 		*(.data)
 		*(.data.*)
-		. = ALIGN(8);
-		__start___verbose = .;
-		*(__verbose);
-		__stop___verbose = .;
 	} :data
 	.rodata : {
 		*(.rodata .rodata.*)

kernel/config/smp-x86.lds

@@ -1,28 +1,24 @@
 PHDRS 
 {
 	text PT_LOAD FLAGS(5);
-	data PT_LOAD FLAGS(7);
+    data PT_LOAD FLAGS(7);
 }
 SECTIONS
 {
 	. = 0xFFFFFFFFFE801000;
-	_head = .;
+    _head = .;
 
-	.text : {
-		*(.text);
-	} : text
+    .text : {
+        *(.text);
+    } : text
 
-	. = ALIGN(4096);
+    . = ALIGN(4096);
 	.data : {
-		*(.data)
-		*(.data.*)
-		. = ALIGN(8);
-		__start___verbose = .;
-		*(__verbose);
-		__stop___verbose = .;
+	      *(.data)
+	      *(.data.*)
 	} :data
 	.rodata : {
-		*(.rodata .rodata.*)
+	      *(.rodata .rodata.*)
 	} :data
 
 	.vsyscall : ALIGN(0x1000) {
@@ -41,9 +37,9 @@ SECTIONS
 		. = ALIGN(4096);
 	} : data = 0xf4
 
-	.bss : {
-		*(.bss .bss.*)
-	}
-	. = ALIGN(4096);
-	_end = .;
+    .bss : {
+          *(.bss .bss.*)
+    }
+    . = ALIGN(4096);
+    _end = .;
 }

kernel/debug.c

@@ -18,9 +18,6 @@
 #include <ihk/lock.h>
 #include <ihk/monitor.h>
 #include <errno.h>
-#include <sysfs.h>
-#include <debug.h>
-#include <limits.h>
 
 struct ihk_kmsg_buf *kmsg_buf;
 
@@ -87,8 +84,7 @@ void kputs(char *buf)
 	debug_spin_unlock_irqrestore(&kmsg_buf->lock, flags_inner);
 	kprintf_unlock(flags_outer);
 
-	if (irqflags_can_interrupt(flags_outer) &&
-			DEBUG_KMSG_USED > IHK_KMSG_HIGH_WATER_MARK) {
+	if (DEBUG_KMSG_USED > IHK_KMSG_HIGH_WATER_MARK) {
 		eventfd(IHK_OS_EVENTFD_TYPE_KMSG);
 		ihk_mc_delay_us(IHK_KMSG_NOTIFY_DELAY);
 	}
@@ -127,8 +123,8 @@ int __kprintf(const char *format, ...)
 	}
 
 	debug_spin_unlock_irqrestore(&kmsg_buf->lock, flags_inner);
-	if (irqflags_can_interrupt(flags_inner) &&
-			DEBUG_KMSG_USED > IHK_KMSG_HIGH_WATER_MARK) {
+
+	if (DEBUG_KMSG_USED > IHK_KMSG_HIGH_WATER_MARK) {
 		eventfd(IHK_OS_EVENTFD_TYPE_KMSG);
 		ihk_mc_delay_us(IHK_KMSG_NOTIFY_DELAY);
 	}
@@ -169,8 +165,7 @@ int kprintf(const char *format, ...)
 	debug_spin_unlock_irqrestore(&kmsg_buf->lock, flags_inner);
 	kprintf_unlock(flags_outer);
 
-	if (irqflags_can_interrupt(flags_outer) &&
-			DEBUG_KMSG_USED > IHK_KMSG_HIGH_WATER_MARK) {
+	if (DEBUG_KMSG_USED > IHK_KMSG_HIGH_WATER_MARK) {
 		eventfd(IHK_OS_EVENTFD_TYPE_KMSG);
 		ihk_mc_delay_us(IHK_KMSG_NOTIFY_DELAY);
 	}
@@ -183,147 +178,3 @@ void kmsg_init()
 {
 	ihk_mc_spinlock_init(&kmsg_lock);
 }
-
-extern struct ddebug __start___verbose[];
-extern struct ddebug __stop___verbose[];
-
-static ssize_t dynamic_debug_sysfs_show(struct sysfs_ops *ops,
-		void *instance, void *buf, size_t size)
-{
-	struct ddebug *dbg;
-	ssize_t n = 0;
-
-	n = snprintf(buf, size, "# filename:lineno function flags format\n");
-
-	for (dbg = __start___verbose; dbg < __stop___verbose; dbg++) {
-		n += snprintf(buf + n, size - n, "%s:%d %s =%s\n",
-				dbg->file, dbg->line, dbg->func,
-				dbg->flags ? "p" : "_");
-
-		if (n >= size)
-			break;
-	}
-
-	return n;
-}
-
-static ssize_t dynamic_debug_sysfs_store(struct sysfs_ops *ops,
-		void *instance, void *buf, size_t size)
-{
-	char *cur = buf;
-	char *file = NULL, *func = NULL;
-	long int line_start = 0, line_end = INT_MAX;
-	int set_flag = -1;
-	struct ddebug *dbg;
-
-
-	// assume line was new-line terminated and squash last newline
-	cur[size-1] = '\0';
-
-	/* basic line parsing, combinaisons of:
-	 *   file <file>
-	 *   func <func>
-	 *   line <line|line-line|line-|-line>
-	 *   and must end with [+-=][p_] (set/clear print flag)
-	 */
-again:
-	while (cur && cur < ((char *)buf) + size && *cur) {
-		dkprintf("looking at %.*s, size left %d\n",
-			size - (cur - (char *)buf), cur,
-			(char *)buf - cur + size);
-
-		if (strncmp(cur, "func ", 5) == 0) {
-			cur += 5;
-			func = cur;
-		} else if (strncmp(cur, "file ", 5) == 0) {
-			cur += 5;
-			file = cur;
-		} else if (strncmp(cur, "line ", 5) == 0) {
-			cur += 5;
-			if (*cur != '-') {
-				line_start = strtol(cur, &cur, 0);
-			}
-			if (*cur != '-') {
-				line_end = line_start;
-			} else {
-				cur++;
-				if (*cur == ' ' || *cur == '\0') {
-					line_end = INT_MAX;
-				} else {
-					line_end = strtol(cur, &cur, 0);
-				}
-			}
-		} else if (strchr("+-=", *cur)) {
-			switch ((*cur) + 256 * (*(cur+1))) {
-			case '+' + 256*'p':
-			case '=' + 256*'p':
-				set_flag = DDEBUG_PRINT;
-				break;
-			case '-' + 256*'p':
-			case '=' + 256*'_':
-				set_flag = DDEBUG_NONE;
-				break;
-			default:
-				kprintf("invalid flag: %.*s\n",
-					size - (cur - (char *)buf), cur);
-				return -EINVAL;
-			}
-			/* XXX check 3rd char is end of input or \n or ; */
-			cur += 3;
-			break;
-
-		} else {
-			kprintf("dynamic debug control: unrecognized keyword: %.*s\n",
-				size - (cur - (char *)buf), cur);
-			return -EINVAL;
-		}
-		cur = strpbrk(cur, " \n");
-		if (cur) {
-			*cur = '\0';
-			cur++;
-		}
-	}
-	dkprintf("func %s, file %s, lines %d-%d, flag %x\n",
-		func, file, line_start, line_end, set_flag);
-
-	if (set_flag < 0) {
-		kprintf("dynamic debug control: no flag set?\n");
-		return -EINVAL;
-	}
-	if (!func && !file) {
-		kprintf("at least file or func should be set\n");
-		return -EINVAL;
-	}
-
-	for (dbg = __start___verbose; dbg < __stop___verbose; dbg++) {
-		/* TODO: handle wildcards */
-		if ((!func || strcmp(func, dbg->func) == 0) &&
-		    (!file || strcmp(file, dbg->file) == 0) &&
-		    dbg->line >= line_start &&
-		    dbg->line <= line_end) {
-			dbg->flags = set_flag;
-		}
-	}
-
-	if (cur && cur < ((char *)buf) + size && *cur)
-		goto again;
-
-	return size;
-}
-
-static struct sysfs_ops dynamic_debug_sysfs_ops = {
-	.show = &dynamic_debug_sysfs_show,
-	.store = &dynamic_debug_sysfs_store,
-};
-
-void dynamic_debug_sysfs_setup(void)
-{
-	int error;
-
-	error = sysfs_createf(&dynamic_debug_sysfs_ops, NULL, 0644,
-			      "/sys/kernel/debug/dynamic_debug/control");
-	if (error) {
-		kprintf("%s: ERROR: creating dynamic_debug/control sysfs file",
-			__func__);
-	}
-}

kernel/devobj.c

@@ -36,13 +36,15 @@
 #include <syscall.h>
 #include <process.h>
 #include <rusage_private.h>
-#include <debug.h>
 
 //#define DEBUG_PRINT_DEVOBJ
 
 #ifdef DEBUG_PRINT_DEVOBJ
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...) kprintf(__VA_ARGS__)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
 #endif
 
 
@@ -52,15 +54,16 @@ struct devobj {
 	uintptr_t	handle;
 	off_t		pfn_pgoff;
 	uintptr_t *	pfn_table;
-	ihk_spinlock_t  pfn_table_lock;
 	size_t		npages;
 };
 
-static memobj_free_func_t devobj_free;
+static memobj_release_func_t devobj_release;
+static memobj_ref_func_t devobj_ref;
 static memobj_get_page_func_t devobj_get_page;
 
 static struct memobj_ops devobj_ops = {
-	.free =		&devobj_free,
+	.release =	&devobj_release,
+	.ref =		&devobj_ref,
 	.get_page =	&devobj_get_page,
 };
 
@@ -85,9 +88,12 @@ int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxp
 	int error;
 	struct devobj *obj  = NULL;
 	const size_t npages = (len + PAGE_SIZE - 1) / PAGE_SIZE;
+#ifdef POSTK_DEBUG_TEMP_FIX_36
 	const size_t uintptr_per_page = (PAGE_SIZE / sizeof(uintptr_t));
-	const size_t pfn_npages =
-		(npages + uintptr_per_page - 1) / uintptr_per_page;
+	const size_t pfn_npages = (npages + uintptr_per_page - 1) / uintptr_per_page;
+#else
+	const size_t pfn_npages = (npages / (PAGE_SIZE / sizeof(uintptr_t))) + 1;
+#endif /*POSTK_DEBUG_TEMP_FIX_36*/
 
 	dkprintf("%s: fd: %d, len: %lu, off: %lu \n", __FUNCTION__, fd, len, off);
 
@@ -116,8 +122,6 @@ int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxp
 	ihk_mc_syscall_arg4(&ctx) = virt_to_phys(&result);
 	ihk_mc_syscall_arg5(&ctx) = prot | populate_flags;
 
-	memset(&result, 0, sizeof(result));
-
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
 	if (error) {
 		kprintf("%s: error: fd: %d, len: %lu, off: %lu map failed.\n", 
@@ -131,7 +135,6 @@ int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxp
 	obj->memobj.ops = &devobj_ops;
 	obj->memobj.flags = MF_HAS_PAGER | MF_DEV_FILE;
 	obj->memobj.size = len;
-	ihk_atomic_set(&obj->memobj.refcnt, 1);
 	obj->handle = result.handle;
 
 	dkprintf("%s: path=%s\n", __FUNCTION__, result.path);
@@ -145,9 +148,10 @@ int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxp
 		strncpy(obj->memobj.path, result.path, PATH_MAX);
 	}
 
-	obj->pfn_pgoff = off >> PAGE_SHIFT;
+	obj->ref = 1;
+	obj->pfn_pgoff = off / PAGE_SIZE;
 	obj->npages = npages;
-	ihk_mc_spinlock_init(&obj->pfn_table_lock);
+	ihk_mc_spinlock_init(&obj->memobj.lock);
 
 	error = 0;
 	*objp = to_memobj(obj);
@@ -166,50 +170,81 @@ out:
 	return error;
 }
 
-static void devobj_free(struct memobj *memobj)
+static void devobj_ref(struct memobj *memobj)
 {
 	struct devobj *obj = to_devobj(memobj);
+
+	dkprintf("devobj_ref(%p %lx):\n", obj, obj->handle);
+	memobj_lock(&obj->memobj);
+	++obj->ref;
+	memobj_unlock(&obj->memobj);
+	return;
+}
+
+static void devobj_release(struct memobj *memobj)
+{
+	struct devobj *obj = to_devobj(memobj);
+	struct devobj *free_obj = NULL;
 	uintptr_t handle;
-	const size_t uintptr_per_page = (PAGE_SIZE / sizeof(uintptr_t));
+#ifndef POSTK_DEBUG_TEMP_FIX_36
 	const size_t pfn_npages =
-		(obj->npages + uintptr_per_page - 1) / uintptr_per_page;
-	int error;
-	ihk_mc_user_context_t ctx;
+		(obj->npages / (PAGE_SIZE / sizeof(uintptr_t))) + 1;
+#endif /*!POSTK_DEBUG_TEMP_FIX_36*/
 
-	dkprintf("%s(%p %lx)\n", __func__, obj, obj->handle);
+	dkprintf("devobj_release(%p %lx)\n", obj, obj->handle);
 
+	memobj_lock(&obj->memobj);
+	--obj->ref;
+	if (obj->ref <= 0) {
+		free_obj = obj;
+	}
 	handle = obj->handle;
+	memobj_unlock(&obj->memobj);
 
-	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_UNMAP;
-	ihk_mc_syscall_arg1(&ctx) = handle;
-	ihk_mc_syscall_arg2(&ctx) = 1;
+	if (free_obj) {
+		if (!(free_obj->memobj.flags & MF_HOST_RELEASED)) {
+			int error;
+			ihk_mc_user_context_t ctx;
 
-	error = syscall_generic_forwarding(__NR_mmap, &ctx);
-	if (error) {
-		kprintf("%s(%p %lx): release failed. %d\n",
-			__func__, obj, handle, error);
-		/* through */
+			ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_UNMAP;
+			ihk_mc_syscall_arg1(&ctx) = handle;
+			ihk_mc_syscall_arg2(&ctx) = 1;
+
+			error = syscall_generic_forwarding(__NR_mmap, &ctx);
+			if (error) {
+				kprintf("devobj_release(%p %lx):"
+						"release failed. %d\n",
+						free_obj, handle, error);
+				/* through */
+			}
+		}
+
+		if (obj->pfn_table) {
+			// Don't call memory_stat_rss_sub() because devobj related pages don't reside in main memory
+#ifdef POSTK_DEBUG_TEMP_FIX_36
+			const size_t uintptr_per_page = (PAGE_SIZE / sizeof(uintptr_t));
+			const size_t pfn_npages = (obj->npages + uintptr_per_page - 1) / uintptr_per_page;
+			ihk_mc_free_pages(obj->pfn_table, pfn_npages);
+#else
+			ihk_mc_free_pages(obj->pfn_table, pfn_npages);
+#endif /*POSTK_DEBUG_TEMP_FIX_36*/
+		}
+
+		if (to_memobj(free_obj)->path) {
+			kfree(to_memobj(free_obj)->path);
+		}
+
+		kfree(free_obj);
 	}
 
-	if (obj->pfn_table) {
-		// Don't call memory_stat_rss_sub() because devobj related
-		// pages don't reside in main memory
-		ihk_mc_free_pages(obj->pfn_table, pfn_npages);
-	}
-
-	if (to_memobj(obj)->path) {
-		kfree(to_memobj(obj)->path);
-	}
-
-	kfree(obj);
-
-	dkprintf("%s(%p %lx):free\n", __func__, obj, handle);
+	dkprintf("devobj_release(%p %lx):free %p\n",
+			obj, handle, free_obj);
 	return;
 }
 
 static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *flag, uintptr_t virt_addr)
 {
-	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;
@@ -227,14 +262,17 @@ static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintpt
 	ix = pgoff - obj->pfn_pgoff;
 	dkprintf("ix: %ld\n", ix);
 
+	memobj_lock(&obj->memobj);
+	pfn = obj->pfn_table[ix];
 #ifdef PROFILE_ENABLE
 	profile_event_add(PROFILE_page_fault_dev_file, PAGE_SIZE);
 #endif // PROFILE_ENABLE
-	pfn = obj->pfn_table[ix];
 	if (!(pfn & PFN_VALID)) {
+		memobj_unlock(&obj->memobj);
+
 		ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_PFN;
 		ihk_mc_syscall_arg1(&ctx) = obj->handle;
-		ihk_mc_syscall_arg2(&ctx) = off & ~(PAGE_SIZE - 1);
+		ihk_mc_syscall_arg2(&ctx) = pgoff << PAGE_SHIFT;
 		ihk_mc_syscall_arg3(&ctx) = virt_to_phys(&pfn);
 
 		error = syscall_generic_forwarding(__NR_mmap, &ctx);
@@ -265,9 +303,11 @@ static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintpt
 			dkprintf("devobj_get_page(%p %lx,%lx,%d):PFN_PRESENT after %#lx\n", memobj, obj->handle, off, p2align, pfn);
 		}
 
+		memobj_lock(&obj->memobj);
 		obj->pfn_table[ix] = pfn;
 		// Don't call memory_stat_rss_add() because devobj related pages don't reside in main memory
 	}
+	memobj_unlock(&obj->memobj);
 
 	if (!(pfn & PFN_PRESENT)) {
 		kprintf("devobj_get_page(%p %lx,%lx,%d):not present. %lx\n", memobj, obj->handle, off, p2align, pfn);

kernel/file_ops.c

@@ -0,0 +1,291 @@
+#include <hfi1/file_ops.h>
+#include <hfi1/hfi.h>
+#include <hfi1/user_sdma.h>
+#include <hfi1/sdma.h>
+#include <hfi1/ihk_hfi1_common.h>
+#include <hfi1/user_exp_rcv.h>
+#include <errno.h>
+
+//#define DEBUG_PRINT_FOPS
+
+#ifdef DEBUG_PRINT_FOPS
+#define	dkprintf(...) kprintf(__VA_ARGS__)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...) kprintf(__VA_ARGS__)
+#endif
+
+long hfi1_file_ioctl(void *private_data, unsigned int cmd,
+			    unsigned long arg, unsigned long t_s)
+{
+	struct hfi1_filedata *fd = private_data;
+	struct hfi1_ctxtdata *uctxt = fd->uctxt;
+	struct hfi1_tid_info tinfo;
+	unsigned long addr;
+	int ret = -ENOTSUPP;
+
+	hfi1_cdbg(IOCTL, "IOCTL recv: 0x%x", cmd);
+	if (cmd != HFI1_IOCTL_ASSIGN_CTXT &&
+	    cmd != HFI1_IOCTL_GET_VERS &&
+	    !uctxt)
+		return -EINVAL;
+
+	switch (cmd) {
+	case HFI1_IOCTL_ASSIGN_CTXT:
+#if 0	
+		if (uctxt)
+			return -EINVAL;
+
+		if (copy_from_user(&uinfo,
+				   (struct hfi1_user_info __user *)arg,
+				   sizeof(uinfo)))
+			return -EFAULT;
+
+		ret = assign_ctxt(fp, &uinfo);
+		if (ret < 0)
+			return ret;
+		ret = setup_ctxt(fp);
+		if (ret)
+			return ret;
+		ret = user_init(fp);
+#endif
+		dkprintf("%s: HFI1_IOCTL_ASSIGN_CTXT \n", __FUNCTION__);
+		break;
+	case HFI1_IOCTL_CTXT_INFO:
+#if 0
+		ret = get_ctxt_info(fp, (void __user *)(unsigned long)arg,
+				    sizeof(struct hfi1_ctxt_info));
+#endif
+		dkprintf("%s: HFI1_IOCTL_CTXT_INFO \n", __FUNCTION__);
+		break;
+	case HFI1_IOCTL_USER_INFO:
+#if 0
+		ret = get_base_info(fp, (void __user *)(unsigned long)arg,
+				    sizeof(struct hfi1_base_info));
+#endif
+		dkprintf("%s: HFI1_IOCTL_USER_INFO \n", __FUNCTION__);
+		break;
+	case HFI1_IOCTL_CREDIT_UPD:
+#if 0	
+		if (uctxt)
+			sc_return_credits(uctxt->sc);
+#endif
+		dkprintf("%s: HFI1_IOCTL_CREDIT_UPD \n", __FUNCTION__);
+		break;
+
+	case HFI1_IOCTL_TID_UPDATE:
+		dkprintf("%s: HFI1_IOCTL_TID_UPDATE \n", __FUNCTION__);
+		if (copy_from_user(&tinfo,
+				   (struct hfi11_tid_info __user *)arg,
+				   sizeof(tinfo)))
+			return -EFAULT;
+
+		ret = hfi1_user_exp_rcv_setup(fd, &tinfo);
+		if (!ret) {
+			/*
+			 * Copy the number of tidlist entries we used
+			 * and the length of the buffer we registered.
+			 * These fields are adjacent in the structure so
+			 * we can copy them at the same time.
+			 */
+			addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
+			if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
+					 sizeof(tinfo.tidcnt) +
+					 sizeof(tinfo.length)))
+				ret = -EFAULT;
+		}
+		break;
+
+	case HFI1_IOCTL_TID_FREE:
+		dkprintf("%s: HFI1_IOCTL_TID_FREE \n", __FUNCTION__);
+		if (copy_from_user(&tinfo,
+				   (struct hfi11_tid_info __user *)arg,
+				   sizeof(tinfo)))
+			return -EFAULT;
+
+		ret = hfi1_user_exp_rcv_clear(fd, &tinfo);
+		if (ret)
+			break;
+		addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
+		if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
+				 sizeof(tinfo.tidcnt)))
+			ret = -EFAULT;
+		break;
+
+	case HFI1_IOCTL_TID_INVAL_READ:
+		dkprintf("%s: HFI1_IOCTL_TID_INVAL_READ \n", __FUNCTION__);
+		if (copy_from_user(&tinfo,
+				   (struct hfi11_tid_info __user *)arg,
+				   sizeof(tinfo)))
+			return -EFAULT;
+
+		ret = hfi1_user_exp_rcv_invalid(fd, &tinfo);
+		if (ret)
+			break;
+		addr = arg + offsetof(struct hfi1_tid_info, tidcnt);
+		if (copy_to_user((void __user *)addr, &tinfo.tidcnt,
+				 sizeof(tinfo.tidcnt)))
+			ret = -EFAULT;
+		break;
+
+	case HFI1_IOCTL_RECV_CTRL:
+#if 0
+		ret = get_user(uval, (int __user *)arg);
+		if (ret != 0)
+			return -EFAULT;
+		ret = manage_rcvq(uctxt, fd->subctxt, uval);
+#endif
+		dkprintf("%s: HFI1_IOCTL_RECV_CTRL \n", __FUNCTION__);
+		break;
+
+	case HFI1_IOCTL_POLL_TYPE:
+#if 0
+		ret = get_user(uval, (int __user *)arg);
+		if (ret != 0)
+			return -EFAULT;
+		uctxt->poll_type = (typeof(uctxt->poll_type))uval;
+#endif
+		dkprintf("%s: HFI1_IOCTL_POLL_TYPE \n", __FUNCTION__);
+		break;
+
+	case HFI1_IOCTL_ACK_EVENT:
+#if 0
+		ret = get_user(ul_uval, (unsigned long __user *)arg);
+		if (ret != 0)
+			return -EFAULT;
+		ret = user_event_ack(uctxt, fd->subctxt, ul_uval);
+#endif
+		dkprintf("%s: HFI1_IOCTL_ACK_EVENT \n", __FUNCTION__);
+		break;
+
+	case HFI1_IOCTL_SET_PKEY:
+#if 0
+		ret = get_user(uval16, (u16 __user *)arg);
+		if (ret != 0)
+			return -EFAULT;
+		if (HFI1_CAP_IS_USET(PKEY_CHECK))
+			ret = set_ctxt_pkey(uctxt, fd->subctxt, uval16);
+		else
+			return -EPERM;
+#endif
+		ret = -ENODEV;
+		dkprintf("%s: HFI1_IOCTL_SET_PKEY \n", __FUNCTION__);
+		break;
+
+	case HFI1_IOCTL_CTXT_RESET: {
+#if 0
+		struct send_context *sc;
+		struct hfi1_devdata *dd;
+
+		if (!uctxt || !uctxt->dd || !uctxt->sc)
+			return -EINVAL;
+
+		/*
+		 * There is no protection here. User level has to
+		 * guarantee that no one will be writing to the send
+		 * context while it is being re-initialized.
+		 * If user level breaks that guarantee, it will break
+		 * it's own context and no one else's.
+		 */
+		dd = uctxt->dd;
+		sc = uctxt->sc;
+		/*
+		 * Wait until the interrupt handler has marked the
+		 * context as halted or frozen. Report error if we time
+		 * out.
+		 */
+		wait_event_interruptible_timeout(
+			sc->halt_wait, (sc->flags & SCF_HALTED),
+			msecs_to_jiffies(SEND_CTXT_HALT_TIMEOUT));
+		if (!(sc->flags & SCF_HALTED))
+			return -ENOLCK;
+
+		/*
+		 * If the send context was halted due to a Freeze,
+		 * wait until the device has been "unfrozen" before
+		 * resetting the context.
+		 */
+		if (sc->flags & SCF_FROZEN) {
+			wait_event_interruptible_timeout(
+				dd->event_queue,
+				!(ACCESS_ONCE(dd->flags) & HFI1_FROZEN),
+				msecs_to_jiffies(SEND_CTXT_HALT_TIMEOUT));
+			if (dd->flags & HFI1_FROZEN)
+				return -ENOLCK;
+
+			if (dd->flags & HFI1_FORCED_FREEZE)
+				/*
+				 * Don't allow context reset if we are into
+				 * forced freeze
+				 */
+				return -ENODEV;
+
+			sc_disable(sc);
+			ret = sc_enable(sc);
+			hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_ENB,
+				     uctxt->ctxt);
+		} else {
+			ret = sc_restart(sc);
+		}
+		if (!ret)
+			sc_return_credits(sc);
+		break;
+#endif
+		dkprintf("%s: HFI1_IOCTL_CTXT_RESET \n", __FUNCTION__);
+		break;
+	}
+
+	case HFI1_IOCTL_GET_VERS:
+#if 0
+		uval = HFI1_USER_SWVERSION;
+		if (put_user(uval, (int __user *)arg))
+			return -EFAULT;
+#endif
+		dkprintf("%s: HFI1_IOCTL_GET_VERS \n", __FUNCTION__);
+		break;
+
+	default:
+		return -ENOTSUPP;
+	}
+	return ret;
+}
+
+ssize_t hfi1_aio_write(void *private_data, const struct iovec *iovec, unsigned long dim)
+{
+	struct hfi1_filedata *fd = private_data;
+	struct hfi1_user_sdma_pkt_q *pq = fd->pq;
+	struct hfi1_user_sdma_comp_q *cq = fd->cq;
+	int done = 0, reqs = 0;
+
+	if (!cq || !pq)
+		return -EIO;
+
+	if (!dim)
+		return -EINVAL;
+
+	hfi1_cdbg(SDMA, "SDMA request from %u:%u (%lu)",
+		fd->uctxt->ctxt, fd->subctxt, dim);
+
+	if (atomic_read(&pq->n_reqs) == pq->n_max_reqs)
+		return -ENOSPC;
+
+	while (dim) {
+		int ret;
+		unsigned long count = 0;
+
+		ret = hfi1_user_sdma_process_request(
+			private_data, (struct iovec *)(iovec + done),
+			dim, &count);
+		if (ret) {
+			reqs = ret;
+			break;
+		}
+		dim -= count;
+		done += count;
+		reqs++;
+	}
+
+	return reqs;
+}
+

kernel/fileobj.c

@@ -27,13 +27,15 @@
 #include <string.h>
 #include <syscall.h>
 #include <rusage_private.h>
-#include <debug.h>
 
 //#define DEBUG_PRINT_FILEOBJ
 
 #ifdef DEBUG_PRINT_FILEOBJ
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define	dkprintf(...)	do { if (1) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
+#else
+#define	dkprintf(...)	do { if (0) kprintf(__VA_ARGS__); } while (0)
+#define	ekprintf(...)	kprintf(__VA_ARGS__)
 #endif
 
 mcs_lock_t fileobj_list_lock;
@@ -45,21 +47,24 @@ static LIST_HEAD(fileobj_list);
 
 struct fileobj {
 	struct memobj memobj;		/* must be first */
-	uint64_t sref;
+	long sref;
+	long cref;
 	uintptr_t handle;
 	struct list_head list;
 	struct list_head page_hash[FILEOBJ_PAGE_HASH_SIZE];
 	mcs_lock_t page_hash_locks[FILEOBJ_PAGE_HASH_SIZE];
 };
 
-static memobj_free_func_t fileobj_free;
+static memobj_release_func_t fileobj_release;
+static memobj_ref_func_t fileobj_ref;
 static memobj_get_page_func_t fileobj_get_page;
 static memobj_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 = {
-	.free =	&fileobj_free,
+	.release =	&fileobj_release,
+	.ref =		&fileobj_ref,
 	.get_page =	&fileobj_get_page,
 	.copy_page =	NULL,
 	.flush_page =	&fileobj_flush_page,
@@ -165,22 +170,22 @@ static void obj_list_remove(struct fileobj *obj)
 /* return NULL or locked fileobj */
 static struct fileobj *obj_list_lookup(uintptr_t handle)
 {
+	struct fileobj *obj;
 	struct fileobj *p;
 
+	obj = NULL;
 	list_for_each_entry(p, &fileobj_list, list) {
 		if (p->handle == handle) {
-			/* for the interval between last put and fileobj_free
-			 * taking list_lock
-			 */
-			if (memobj_ref(&p->memobj) <= 1) {
-				ihk_atomic_dec(&p->memobj.refcnt);
-				continue;
+			memobj_lock(&p->memobj);
+			if (p->cref > 0) {
+				obj = p;
+				break;
 			}
-			return p;
+			memobj_unlock(&p->memobj);
 		}
 	}
 
-	return NULL;
+	return obj;
 }
 
 /***********************************************************************
@@ -195,7 +200,13 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp, uintptr_t virt_a
 	struct fileobj *obj;
 	struct mcs_lock_node node;
 
-	dkprintf("%s(%d)\n", __func__, fd);
+	dkprintf("fileobj_create(%d)\n", fd);
+	newobj = kmalloc(sizeof(*newobj), IHK_MC_AP_NOWAIT);
+	if (!newobj) {
+		error = -ENOMEM;
+		kprintf("fileobj_create(%d):kmalloc failed. %d\n", fd, error);
+		goto out;
+	}
 
 	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_CREATE;
 	ihk_mc_syscall_arg1(&ctx) = fd;
@@ -203,41 +214,20 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp, uintptr_t virt_a
 	memset(&result, 0, sizeof(result));
 
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
-
 	if (error) {
-		/* -ESRCH doesn't mean an error but requesting a fall
-		 * back to treat the file as a device file
-		 */
-		if (error != -ESRCH) {
-			kprintf("%s(%d):create failed. %d\n",
-				__func__, fd, error);
-		}
+		dkprintf("fileobj_create(%d):create failed. %d\n", fd, error);
 		goto out;
 	}
 
-	if (result.flags & MF_HUGETLBFS) {
-		return hugefileobj_pre_create(&result, objp, maxprotp);
-	}
-
-	mcs_lock_lock(&fileobj_list_lock, &node);
-	obj = obj_list_lookup(result.handle);
-	if (obj)
-		goto found;
-	mcs_lock_unlock(&fileobj_list_lock, &node);
-
-	// not found: alloc new object and lookup again
-	newobj = kmalloc(sizeof(*newobj), IHK_MC_AP_NOWAIT);
-	if (!newobj) {
-		error = -ENOMEM;
-		kprintf("%s(%d):kmalloc failed. %d\n", __func__, fd, error);
-		goto out;
-	}
 	memset(newobj, 0, sizeof(*newobj));
 	newobj->memobj.ops = &fileobj_ops;
 	newobj->memobj.flags = MF_HAS_PAGER | MF_REG_FILE;
 	newobj->handle = result.handle;
 
+	newobj->sref = 1;
+	newobj->cref = 1;
 	fileobj_page_hash_init(newobj);
+	ihk_mc_spinlock_init(&newobj->memobj.lock);
 
 	mcs_lock_lock_noirq(&fileobj_list_lock, &node);
 	obj = obj_list_lookup(result.handle);
@@ -247,8 +237,6 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp, uintptr_t virt_a
 		to_memobj(obj)->size = result.size;
 		to_memobj(obj)->flags |= result.flags;
 		to_memobj(obj)->status = MEMOBJ_READY;
-		ihk_atomic_set(&to_memobj(obj)->refcnt, 1);
-		obj->sref = 1;
 		if (to_memobj(obj)->flags & MF_PREFETCH) {
 			to_memobj(obj)->status = MEMOBJ_TO_BE_PREFETCHED;
 		}
@@ -317,17 +305,20 @@ error_cleanup:
 		}
 
 		newobj = NULL;
-		dkprintf("%s: new obj 0x%lx %s\n",
+		dkprintf("%s: new obj 0x%lx cref: %d, %s\n",
 			__FUNCTION__,
 			obj,
+			obj->cref,
 			to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
 	}
 	else {
-found:
-		obj->sref++;
-		dkprintf("%s: existing obj 0x%lx, %s\n",
+		++obj->sref;
+		++obj->cref;
+		memobj_unlock(&obj->memobj);	/* locked by obj_list_lookup() */
+		dkprintf("%s: existing obj 0x%lx cref: %d, %s\n",
 			__FUNCTION__,
 			obj,
+			obj->cref,
 			to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
 	}
 
@@ -341,111 +332,152 @@ out:
 	if (newobj) {
 		kfree(newobj);
 	}
-	dkprintf("%s(%d):%d %p %x\n", __func__, fd, error, *objp, *maxprotp);
+	dkprintf("fileobj_create(%d):%d %p %x\n", fd, error, *objp, *maxprotp);
 	return error;
 }
 
-static void fileobj_free(struct memobj *memobj)
+static void fileobj_ref(struct memobj *memobj)
 {
 	struct fileobj *obj = to_fileobj(memobj);
-	struct mcs_lock_node node;
-	int error;
-	ihk_mc_user_context_t ctx;
 
-
-	dkprintf("%s: free obj 0x%lx, %s\n", __func__,
-		 obj, to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
-
-	mcs_lock_lock_noirq(&fileobj_list_lock, &node);
-	obj_list_remove(obj);
-	mcs_lock_unlock_noirq(&fileobj_list_lock, &node);
-
-	/* zap page_list */
-	for (;;) {
-		struct page *page;
-		void *page_va;
-		uintptr_t phys;
-
-		page = fileobj_page_hash_first(obj);
-		if (!page) {
-			break;
-		}
-		__fileobj_page_hash_remove(page);
-		phys = page_to_phys(page);
-		page_va = phys_to_virt(phys);
-		/* Count must be one because set to one on the first
-		 * get_page() invoking fileobj_do_pageio and incremented by
-		 * the second get_page() reaping the pageio and decremented
-		 * by clear_range().
-		 */
-		if (ihk_atomic_read(&page->count) != 1) {
-			kprintf("%s: WARNING: page count is %d for phys 0x%lx is invalid, flags: 0x%lx\n",
-				__func__, ihk_atomic_read(&page->count),
-				page->phys, to_memobj(obj)->flags);
-		}
-		else if (page_unmap(page)) {
-			ihk_mc_free_pages_user(page_va, 1);
-			/* Track change in page->count for !MF_PREMAP pages.
-			 * It is decremented here or in clear_range()
-			 */
-			dkprintf("%lx-,%s: calling memory_stat_rss_sub(),phys=%lx,size=%ld,pgsize=%ld\n",
-				 phys, __func__, phys, PAGE_SIZE, PAGE_SIZE);
-			rusage_memory_stat_mapped_file_sub(PAGE_SIZE,
-							   PAGE_SIZE);
-		}
-	}
-
-	/* Pre-mapped? */
-	if (to_memobj(obj)->flags & MF_PREMAP) {
-		int i;
-
-		for (i = 0; i < to_memobj(obj)->nr_pages; ++i) {
-			if (to_memobj(obj)->pages[i]) {
-				dkprintf("%s: pages[i]=%p\n", __func__, i,
-					 to_memobj(obj)->pages[i]);
-				// Track change in fileobj->pages[] for MF_PREMAP pages
-				// Note that page_unmap() isn't called for MF_PREMAP in
-				// free_process_memory_range() --> ihk_mc_pt_free_range()
-				dkprintf("%lx-,%s: memory_stat_rss_sub,phys=%lx,size=%ld,pgsize=%ld\n",
-					 virt_to_phys(to_memobj(obj)->pages[i]),
-					 __func__,
-					 virt_to_phys(to_memobj(obj)->pages[i]),
-					 PAGE_SIZE, PAGE_SIZE);
-				rusage_memory_stat_mapped_file_sub(PAGE_SIZE,
-								   PAGE_SIZE);
-				ihk_mc_free_pages_user(to_memobj(obj)->pages[i],
-						       1);
-			}
-		}
-
-		kfree(to_memobj(obj)->pages);
-	}
-
-	if (to_memobj(obj)->path) {
-		dkprintf("%s: %s\n", __func__, to_memobj(obj)->path);
-		kfree(to_memobj(obj)->path);
-	}
-
-	/* linux side
-	 * sref is necessary because handle is used as key, so there could
-	 * be a new mckernel pager with the same handle being created as
-	 * this one is being destroyed
-	 */
-	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_RELEASE;
-	ihk_mc_syscall_arg1(&ctx) = obj->handle;
-	ihk_mc_syscall_arg2(&ctx) = obj->sref;
-
-	error = syscall_generic_forwarding(__NR_mmap, &ctx);
-	if (error) {
-		kprintf("%s(%p %lx): free failed. %d\n", __func__,
-			obj, obj->handle, error);
-		/* through */
-	}
-
-	dkprintf("%s(%p %lx):free\n", __func__, obj, obj->handle);
-	kfree(obj);
+	dkprintf("fileobj_ref(%p %lx):\n", obj, obj->handle);
+	memobj_lock(&obj->memobj);
+	++obj->cref;
+	memobj_unlock(&obj->memobj);
 	return;
+}
 
+static void fileobj_release(struct memobj *memobj)
+{
+	struct fileobj *obj = to_fileobj(memobj);
+	long free_sref = 0;
+	uintptr_t free_handle;
+	struct fileobj *free_obj = NULL;
+	struct mcs_lock_node node;
+
+	dkprintf("fileobj_release(%p %lx)\n", obj, obj->handle);
+
+	memobj_lock(&obj->memobj);
+	--obj->cref;
+	free_sref = obj->sref - 1;	/* surplus sref */
+	if (obj->cref <= 0) {
+		free_sref = obj->sref;
+		free_obj = obj;
+	}
+	obj->sref -= free_sref;
+	free_handle = obj->handle;
+	memobj_unlock(&obj->memobj);
+	if (obj->memobj.flags & MF_HOST_RELEASED) {
+		free_sref = 0; // don't call syscall_generic_forwarding
+	}
+
+	if (free_obj) {
+		dkprintf("%s: release obj 0x%lx cref: %d, free_obj: 0x%lx, %s\n",
+				__FUNCTION__,
+				obj,
+				obj->cref,
+				free_obj,
+				to_memobj(obj)->flags & MF_ZEROFILL ? "zerofill" : "");
+		mcs_lock_lock_noirq(&fileobj_list_lock, &node);
+		/* zap page_list */
+		for (;;) {
+			struct page *page;
+			void *page_va;
+			uintptr_t phys;
+
+			page = fileobj_page_hash_first(obj);
+			if (!page) {
+				break;
+			}
+			__fileobj_page_hash_remove(page);
+			phys = page_to_phys(page);
+			page_va = phys_to_virt(phys);
+
+			/* Count must be one because set to one on the first get_page() invoking fileobj_do_pageio and
+			   incremented by the second get_page() reaping the pageio and decremented by clear_range().
+			 */
+			if (ihk_atomic_read(&page->count) != 1) {
+				kprintf("%s: WARNING: page count is %d for phys 0x%lx is invalid, flags: 0x%lx\n",
+						__FUNCTION__,
+					ihk_atomic_read(&page->count),
+					page->phys,
+					to_memobj(free_obj)->flags);
+			}
+			else if (page_unmap(page)) {
+				ihk_mc_free_pages_user(page_va, 1);
+				/* Track change in page->count for !MF_PREMAP pages. It is decremented here or in clear_range() */
+				dkprintf("%lx-,%s: calling memory_stat_rss_sub(),phys=%lx,size=%ld,pgsize=%ld\n", phys, __FUNCTION__, phys, PAGE_SIZE, PAGE_SIZE);
+				rusage_memory_stat_mapped_file_sub(PAGE_SIZE, PAGE_SIZE); 
+			}
+#if 0
+			count = ihk_atomic_sub_return(1, &page->count);
+
+			if (!((page->mode == PM_WILL_PAGEIO)
+					|| (page->mode == PM_DONE_PAGEIO)
+					|| (page->mode == PM_PAGEIO_EOF)
+					|| (page->mode == PM_PAGEIO_ERROR)
+					|| ((page->mode == PM_MAPPED)
+						&& (count <= 0)))) {
+				kprintf("fileobj_release(%p %lx): "
+					       "mode %x, count %d, off %lx\n",
+					       obj, obj->handle, page->mode,
+					       count, page->offset);
+				panic("fileobj_release");
+			}
+
+			page->mode = PM_NONE;
+#endif
+		}
+
+		/* Pre-mapped? */
+		if (to_memobj(free_obj)->flags & MF_PREMAP) {
+			int i;
+			for (i = 0; i < to_memobj(free_obj)->nr_pages; ++i) {
+				if (to_memobj(free_obj)->pages[i]) {
+					dkprintf("%s: pages[i]=%p\n", __FUNCTION__, i, to_memobj(free_obj)->pages[i]);
+					// Track change in fileobj->pages[] for MF_PREMAP pages
+					// Note that page_unmap() isn't called for MF_PREMAP in 
+					// free_process_memory_range() --> ihk_mc_pt_free_range()
+					dkprintf("%lx-,%s: memory_stat_rss_sub,phys=%lx,size=%ld,pgsize=%ld\n",
+							virt_to_phys(to_memobj(free_obj)->pages[i]), __FUNCTION__, virt_to_phys(to_memobj(free_obj)->pages[i]), PAGE_SIZE, PAGE_SIZE);
+					rusage_memory_stat_mapped_file_sub(PAGE_SIZE, PAGE_SIZE);
+					ihk_mc_free_pages_user(to_memobj(free_obj)->pages[i], 1);
+				}
+			}
+
+			kfree(to_memobj(free_obj)->pages);
+		}
+
+		if (to_memobj(free_obj)->path) {
+			dkprintf("%s: %s\n", __FUNCTION__, to_memobj(free_obj)->path);
+			kfree(to_memobj(free_obj)->path);
+		}
+
+		obj_list_remove(free_obj);
+		mcs_lock_unlock_noirq(&fileobj_list_lock, &node);
+		kfree(free_obj);
+	}
+
+	if (free_sref) {
+		int error;
+		ihk_mc_user_context_t ctx;
+
+		ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_RELEASE;
+		ihk_mc_syscall_arg1(&ctx) = free_handle;
+		ihk_mc_syscall_arg2(&ctx) = free_sref;
+
+		error = syscall_generic_forwarding(__NR_mmap, &ctx);
+		if (error) {
+			kprintf("fileobj_release(%p %lx):"
+					"release %ld failed. %d\n",
+					obj, free_handle, free_sref, error);
+			/* through */
+		}
+	}
+
+	dkprintf("fileobj_release(%p %lx):free %ld %p\n",
+			obj, free_handle, free_sref, free_obj);
+	return;
 }
 
 struct pageio_args {
@@ -538,7 +570,7 @@ static void fileobj_do_pageio(void *args0)
 out:
 	mcs_lock_unlock_noirq(&obj->page_hash_locks[hash],
 			&mcs_node);
-	memobj_unref(&obj->memobj);		/* got fileobj_get_page() */
+	fileobj_release(&obj->memobj);		/* got fileobj_get_page() */
 	kfree(args0);
 	dkprintf("fileobj_do_pageio(%p,%lx,%lx):\n", obj, off, pgsize);
 	return;
@@ -624,9 +656,7 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 			npages = 1 << p2align;
 
 			virt = ihk_mc_alloc_pages_user(npages, (IHK_MC_AP_NOWAIT |
-					((to_memobj(obj)->flags & MF_ZEROFILL) ?
-						IHK_MC_AP_USER : 0)),
-					virt_addr);
+					(to_memobj(obj)->flags & MF_ZEROFILL) ? IHK_MC_AP_USER : 0), virt_addr);
 			if (!virt) {
 				error = -ENOMEM;
 				kprintf("fileobj_get_page(%p,%lx,%x,%x,%p):"
@@ -651,7 +681,9 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 			page->mode = PM_WILL_PAGEIO;
 		}
 
-		memobj_ref(&obj->memobj);
+		memobj_lock(&obj->memobj);
+		++obj->cref;	/* for fileobj_do_pageio() */
+		memobj_unlock(&obj->memobj);
 
 		args->fileobj = obj;
 		args->objoff = off;
@@ -712,6 +744,10 @@ static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 		return 0;
 	}
 
+	if (memobj->flags & MF_HOST_RELEASED) {
+		return 0;
+	}
+
 	page = phys_to_page(phys);
 	if (!page) {
 		kprintf("%s: warning: tried to flush non-existing page for phys addr: 0x%lx\n", 
@@ -719,6 +755,8 @@ static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 		return 0;
 	}
 
+	memobj_unlock(&obj->memobj);
+
 	ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_WRITE;
 	ihk_mc_syscall_arg1(&ctx) = obj->handle;
 	ihk_mc_syscall_arg2(&ctx) = page->offset;
@@ -733,6 +771,7 @@ static int fileobj_flush_page(struct memobj *memobj, uintptr_t phys,
 		/* through */
 	}
 
+	memobj_lock(&obj->memobj);
 	return 0;
 }
 

kernel/futex.c

@@ -70,22 +70,15 @@
 #include <cls.h>
 #include <kmsg.h>
 #include <timer.h>
-#include <debug.h>
-#include <syscall.h>
 
 //#define DEBUG_PRINT_FUTEX
 
 #ifdef DEBUG_PRINT_FUTEX
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
-#define uti_dkprintf(...) do { ((clv_override && linux_printk) ? (*linux_printk) : kprintf)(__VA_ARGS__); } while (0)
+#define dkprintf kprintf
 #else
-#define uti_dkprintf(...) do { } while (0)
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
 #endif
-#define uti_kprintf(...) do { ((clv_override && linux_printk) ? (*linux_printk) : kprintf)(__VA_ARGS__); } while (0)
 
-
-unsigned long ihk_mc_get_ns_per_tsc(void);
 int futex_cmpxchg_enabled;
 
 /**
@@ -115,9 +108,6 @@ struct futex_q {
 	union futex_key key;
 	union futex_key *requeue_pi_key;
 	uint32_t bitset;
-
-	/* Used to wake-up a thread running on a Linux CPU */
-	void *uti_futex_resp; 
 };
 
 /*
@@ -190,12 +180,11 @@ static void drop_futex_key_refs(union futex_key *key)
  * lock_page() might sleep, the caller should not hold a spinlock.
  */
 static int
-get_futex_key(uint32_t *uaddr, int fshared, union futex_key *key, struct cpu_local_var *clv_override)
+get_futex_key(uint32_t *uaddr, int fshared, union futex_key *key)
 {
 	unsigned long address = (unsigned long)uaddr;
 	unsigned long phys;
-	struct thread *thread = cpu_local_var_with_override(current, clv_override);
-	struct process_vm *mm = thread->vm;
+	struct process_vm *mm = cpu_local_var(current)->vm;
 
 	/*
 	 * The futex address must be "naturally" aligned.
@@ -261,7 +250,7 @@ static int cmpxchg_futex_value_locked(uint32_t __user *uaddr, uint32_t uval, uin
  * The hash bucket lock must be held when this is called.
  * Afterwards, the futex_q must not be accessed.
  */
-static void wake_futex(struct futex_q *q, struct cpu_local_var *clv_override)
+static void wake_futex(struct futex_q *q)
 {
 	struct thread *p = q->task;
 
@@ -283,31 +272,8 @@ static void wake_futex(struct futex_q *q, struct cpu_local_var *clv_override)
 	barrier();
 	q->lock_ptr = NULL;
 
-
-	if (q->uti_futex_resp) { 
-		int rc;
-		uti_dkprintf("wake_futex(): waking up migrated-to-Linux thread (tid %d),uti_futex_resp=%p\n", p->tid, q->uti_futex_resp);
-		/* TODO: Add the case when a Linux thread waking up another Linux thread */
-		if (clv_override) {
-			uti_dkprintf("%s: ERROR: A Linux thread is waking up migrated-to-Linux thread\n", __FUNCTION__);
-		}
-		if (p->spin_sleep == 0) {
-			uti_dkprintf("%s: INFO: woken up by someone else\n", __FUNCTION__);
-		}
-
-		struct ikc_scd_packet pckt;
-		struct ihk_ikc_channel_desc *resp_channel = cpu_local_var_with_override(ikc2linux, clv_override);
-		pckt.msg = SCD_MSG_FUTEX_WAKE;
-		pckt.futex.resp = q->uti_futex_resp;
-		pckt.futex.spin_sleep = &p->spin_sleep;
-		rc = ihk_ikc_send(resp_channel, &pckt, 0);
-		if (rc) {
-			uti_dkprintf("%s: ERROR: ihk_ikc_send returned %d, resp_channel=%p\n", __FUNCTION__, rc, resp_channel);
-		}
-	} else {
-		uti_dkprintf("wake_futex(): waking up McKernel thread (tid %d)\n", p->tid);
-		sched_wakeup_thread(p, PS_NORMAL);
-	}
+	dkprintf("wake_futex(): waking up tid %d\n", p->tid);
+	sched_wakeup_thread(p, PS_NORMAL);
 }
 
 /*
@@ -337,7 +303,7 @@ double_unlock_hb(struct futex_hash_bucket *hb1, struct futex_hash_bucket *hb2)
 /*
  * Wake up waiters matching bitset queued on this futex (uaddr).
  */
-static int futex_wake(uint32_t *uaddr, int fshared, int nr_wake, uint32_t bitset, struct cpu_local_var *clv_override)
+static int futex_wake(uint32_t *uaddr, int fshared, int nr_wake, uint32_t bitset)
 {
 	struct futex_hash_bucket *hb;
 	struct futex_q *this, *next;
@@ -348,7 +314,7 @@ static int futex_wake(uint32_t *uaddr, int fshared, int nr_wake, uint32_t bitset
 	if (!bitset)
 		return -EINVAL;
 
-	ret = get_futex_key(uaddr, fshared, &key, clv_override);
+	ret = get_futex_key(uaddr, fshared, &key);
 	if ((ret != 0))
 		goto out;
 
@@ -364,7 +330,7 @@ static int futex_wake(uint32_t *uaddr, int fshared, int nr_wake, uint32_t bitset
 			if (!(this->bitset & bitset))
 				continue;
 
-			wake_futex(this, clv_override);
+			wake_futex(this);
 			if (++ret >= nr_wake)
 				break;
 		}
@@ -382,8 +348,7 @@ out:
  */
 static int
 futex_wake_op(uint32_t *uaddr1, int fshared, uint32_t *uaddr2,
-			  int nr_wake, int nr_wake2, int op,
-			  struct cpu_local_var *clv_override)
+	      int nr_wake, int nr_wake2, int op)
 {
 	union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT;
 	struct futex_hash_bucket *hb1, *hb2;
@@ -392,10 +357,10 @@ futex_wake_op(uint32_t *uaddr1, int fshared, uint32_t *uaddr2,
 	int ret, op_ret;
 
 retry:
-	ret = get_futex_key(uaddr1, fshared, &key1, clv_override);
+	ret = get_futex_key(uaddr1, fshared, &key1);
 	if ((ret != 0))
 		goto out;
-	ret = get_futex_key(uaddr2, fshared, &key2, clv_override);
+	ret = get_futex_key(uaddr2, fshared, &key2);
 	if ((ret != 0))
 		goto out_put_key1;
 
@@ -429,7 +394,7 @@ retry_private:
 
 	plist_for_each_entry_safe(this, next, head, list) {
 		if (match_futex (&this->key, &key1)) {
-			wake_futex(this, clv_override);
+			wake_futex(this);
 			if (++ret >= nr_wake)
 				break;
 		}
@@ -441,7 +406,7 @@ retry_private:
 		op_ret = 0;
 		plist_for_each_entry_safe(this, next, head, list) {
 			if (match_futex (&this->key, &key2)) {
-				wake_futex(this, clv_override);
+				wake_futex(this);
 				if (++op_ret >= nr_wake2)
 					break;
 			}
@@ -504,7 +469,7 @@ void requeue_futex(struct futex_q *q, struct futex_hash_bucket *hb1,
  */
 static int futex_requeue(uint32_t *uaddr1, int fshared, uint32_t *uaddr2,
 			 int nr_wake, int nr_requeue, uint32_t *cmpval,
-						 int requeue_pi, struct cpu_local_var *clv_override)
+			 int requeue_pi)
 {
 	union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT;
 	int drop_count = 0, task_count = 0, ret;
@@ -512,10 +477,10 @@ static int futex_requeue(uint32_t *uaddr1, int fshared, uint32_t *uaddr2,
 	struct plist_head *head1;
 	struct futex_q *this, *next;
 
-	ret = get_futex_key(uaddr1, fshared, &key1, clv_override);
+	ret = get_futex_key(uaddr1, fshared, &key1);
 	if ((ret != 0))
 		goto out;
-	ret = get_futex_key(uaddr2, fshared, &key2, clv_override);
+	ret = get_futex_key(uaddr2, fshared, &key2);
 	if ((ret != 0))
 		goto out_put_key1;
 
@@ -550,7 +515,7 @@ static int futex_requeue(uint32_t *uaddr1, int fshared, uint32_t *uaddr2,
 		 */
 		/* RIKEN: no requeue_pi at this moment */
 		if (++task_count <= nr_wake) {
-			wake_futex(this, clv_override);
+			wake_futex(this);
 			continue;
 		}
 
@@ -609,7 +574,7 @@ queue_unlock(struct futex_q *q, struct futex_hash_bucket *hb)
  * state is implicit in the state of woken task (see futex_wait_requeue_pi() for
  * an example).
  */
-static inline void queue_me(struct futex_q *q, struct futex_hash_bucket *hb, struct cpu_local_var *clv_override)
+static inline void queue_me(struct futex_q *q, struct futex_hash_bucket *hb)
 {
 	int prio;
 
@@ -630,7 +595,7 @@ static inline void queue_me(struct futex_q *q, struct futex_hash_bucket *hb, str
 	q->list.plist.spinlock = &hb->lock;
 #endif
 	plist_add(&q->list, &hb->chain);
-	q->task = cpu_local_var_with_override(current, clv_override);
+	q->task = cpu_local_var(current);
 	ihk_mc_spinlock_unlock_noirq(&hb->lock);
 }
 
@@ -693,64 +658,46 @@ retry:
 /* RIKEN: this function has been rewritten so that it returns the remaining
  * time in case we are waken.
  */
-static int64_t futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q,
-				   uint64_t timeout, struct cpu_local_var *clv_override)
+static uint64_t futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q,
+				uint64_t timeout)
 {
-	int64_t time_remain = 0;
+	uint64_t time_remain = 0;
 	unsigned long irqstate;
-	struct thread *thread = cpu_local_var_with_override(current, clv_override);
+	struct thread *thread = cpu_local_var(current);
 	/*
 	 * The task state is guaranteed to be set before another task can
 	 * wake it. 
 	 * queue_me() calls spin_unlock() upon completion, serializing
 	 * access to the hash list and forcing a memory barrier.
 	 */
-	xchg4(&(thread->status), PS_INTERRUPTIBLE);
+	xchg4(&(cpu_local_var(current)->status), PS_INTERRUPTIBLE);
 
-	/* Indicate spin sleep. Note that schedule_timeout() with
-	 * idle_halt should use spin sleep because sleep with timeout
-	 * is not implemented.
-	 */
-	if (!idle_halt || timeout) {
+	/* Indicate spin sleep */
+	if (!idle_halt) {
 		irqstate = ihk_mc_spinlock_lock(&thread->spin_sleep_lock);
 		thread->spin_sleep = 1;
 		ihk_mc_spinlock_unlock(&thread->spin_sleep_lock, irqstate);
 	}
 
-	queue_me(q, hb, clv_override);
+	queue_me(q, hb);
 	
 	if (!plist_node_empty(&q->list)) {
-		if (clv_override) {
-			uti_dkprintf("%s: tid: %d is trying to sleep\n", __FUNCTION__, thread->tid);
-			/* Note that the unit of timeout is nsec */
-			time_remain = (*linux_wait_event)(q->uti_futex_resp, timeout);
-			
-			/* Note that time_remain == 0 indicates contidion evaluated to false after the timeout elapsed */
-			if (time_remain < 0) {
-				if (time_remain == -ERESTARTSYS) { /* Interrupted by signal */
-					uti_dkprintf("%s: DEBUG: wait_event returned -ERESTARTSYS\n", __FUNCTION__);
-				} else {
-					uti_kprintf("%s: ERROR: wait_event returned %d\n", __FUNCTION__, time_remain);
-				}
-			}
-			uti_dkprintf("%s: tid: %d woken up\n", __FUNCTION__, thread->tid);
-		} else {
 		
 		if (timeout) {
-			dkprintf("futex_wait_queue_me(): tid: %d schedule_timeout()\n", thread->tid);
+			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", thread->tid);
+			dkprintf("futex_wait_queue_me(): tid: %d schedule()\n", cpu_local_var(current)->tid);
 			spin_sleep_or_schedule();
 			time_remain = 0;
 		}
-		dkprintf("futex_wait_queue_me(): tid: %d woken up\n", thread->tid);
-		}
+		
+		dkprintf("futex_wait_queue_me(): tid: %d woken up\n", cpu_local_var(current)->tid);
 	}
 	
 	/* This does not need to be serialized */
-	thread->status = PS_RUNNING;
+	cpu_local_var(current)->status = PS_RUNNING;
 	thread->spin_sleep = 0;
 	
 	return time_remain;
@@ -774,8 +721,7 @@ static int64_t futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q
  * <1 - -EFAULT or -EWOULDBLOCK (uaddr does not contain val) and hb is unlcoked
  */
 static int futex_wait_setup(uint32_t __user *uaddr, uint32_t val, int fshared,
-			    struct futex_q *q, struct futex_hash_bucket **hb,
-			    struct cpu_local_var *clv_override)
+			   struct futex_q *q, struct futex_hash_bucket **hb)
 {
 	uint32_t uval;
 	int ret;
@@ -798,7 +744,7 @@ static int futex_wait_setup(uint32_t __user *uaddr, uint32_t val, int fshared,
 	 * rare, but normal.
 	 */
 	q->key = FUTEX_KEY_INIT;
-	ret = get_futex_key(uaddr, fshared, &q->key, clv_override);
+	ret = get_futex_key(uaddr, fshared, &q->key);
 	if (ret != 0)
 		return ret;
 
@@ -822,59 +768,49 @@ static int futex_wait_setup(uint32_t __user *uaddr, uint32_t val, int fshared,
 }
 
 static int futex_wait(uint32_t __user *uaddr, int fshared,
-		      uint32_t val, uint64_t timeout, uint32_t bitset, int clockrt,
-		      struct cpu_local_var *clv_override)
+		      uint32_t val, uint64_t timeout, uint32_t bitset, int clockrt)
 {
 	struct futex_hash_bucket *hb;
 	struct futex_q q;
-	int64_t time_remain;
+	uint64_t time_remain;
 	int ret;
 
 	if (!bitset)
 		return -EINVAL;
 
 #ifdef PROFILE_ENABLE
-	if (cpu_local_var_with_override(current, clv_override)->profile &&
-		cpu_local_var_with_override(current, clv_override)->profile_start_ts) {
-		cpu_local_var_with_override(current, clv_override)->profile_elapsed_ts +=
-			(rdtsc() - cpu_local_var_with_override(current, clv_override)->profile_start_ts);
-		cpu_local_var_with_override(current, clv_override)->profile_start_ts = 0;
+	if (cpu_local_var(current)->profile &&
+			cpu_local_var(current)->profile_start_ts) {
+		cpu_local_var(current)->profile_elapsed_ts +=
+			(rdtsc() - cpu_local_var(current)->profile_start_ts);
+		cpu_local_var(current)->profile_start_ts = 0;
 	}
 #endif
 
 	q.bitset = bitset;
 	q.requeue_pi_key = NULL;
-	q.uti_futex_resp = cpu_local_var_with_override(uti_futex_resp, clv_override);
 
 retry:
 	/* Prepare to wait on uaddr. */
-	ret = futex_wait_setup(uaddr, val, fshared, &q, &hb, clv_override);
-	if (ret) {
-		uti_dkprintf("%s: tid=%d futex_wait_setup returns zero, no need to sleep\n", __FUNCTION__, cpu_local_var_with_override(current, clv_override)->tid);
+	ret = futex_wait_setup(uaddr, val, fshared, &q, &hb);
+	if (ret)
 		goto out;
-	}
 
 	/* queue_me and wait for wakeup, timeout, or a signal. */
-	time_remain = futex_wait_queue_me(hb, &q, timeout, clv_override);
+	time_remain = futex_wait_queue_me(hb, &q, timeout);
 
 	/* If we were woken (and unqueued), we succeeded, whatever. */
 	ret = 0;
-	if (!unqueue_me(&q)) {
-		uti_dkprintf("%s: tid=%d unqueued\n", __FUNCTION__, cpu_local_var_with_override(current, clv_override)->tid);
+	if (!unqueue_me(&q))
 		goto out_put_key;
-	}
 	ret = -ETIMEDOUT;
 
 	/* RIKEN: timer expired case (indicated by !time_remain) */
-	if (timeout && !time_remain) {
-		uti_dkprintf("%s: tid=%d timer expired\n", __FUNCTION__, cpu_local_var_with_override(current, clv_override)->tid);
+	if (timeout && !time_remain)
 		goto out_put_key;
-	}
 
-	/* RIKEN: futex_wait_queue_me() returns -ERESTARTSYS when waiting on Linux CPU and woken up by signal */
-	if (hassigpending(cpu_local_var_with_override(current, clv_override)) || time_remain == -ERESTARTSYS) {
+	if (hassigpending(cpu_local_var(current))) {
 		ret = -EINTR;
-		uti_dkprintf("%s: tid=%d woken up by signal\n", __FUNCTION__, cpu_local_var_with_override(current, clv_override)->tid);
 		goto out_put_key;
 	}
 
@@ -886,22 +822,19 @@ out_put_key:
 	put_futex_key(fshared, &q.key);
 out:
 #ifdef PROFILE_ENABLE
-	if (cpu_local_var_with_override(current, clv_override)->profile) {
-		cpu_local_var_with_override(current, clv_override)->profile_start_ts = rdtsc();
+	if (cpu_local_var(current)->profile) {
+		cpu_local_var(current)->profile_start_ts = rdtsc();
 	}
 #endif
 	return ret;
 }
 
 int futex(uint32_t *uaddr, int op, uint32_t val, uint64_t timeout,
-	  uint32_t *uaddr2, uint32_t val2, uint32_t val3, int fshared,
-	  struct cpu_local_var *clv_override)
+		uint32_t *uaddr2, uint32_t val2, uint32_t val3, int fshared)
 {
 	int clockrt, ret = -ENOSYS;
 	int cmd = op & FUTEX_CMD_MASK;
 
-	uti_dkprintf("%s: uaddr=%p, op=%x, val=%x, timeout=%ld, uaddr2=%p, val2=%x, val3=%x, fshared=%d, clv=%p\n", __FUNCTION__, uaddr, op, val, timeout, uaddr2, val2, val3, fshared, clv_override);
-
 	clockrt = op & FUTEX_CLOCK_REALTIME;
 	if (clockrt && cmd != FUTEX_WAIT_BITSET && cmd != FUTEX_WAIT_REQUEUE_PI)
 		return -ENOSYS;
@@ -910,21 +843,21 @@ int futex(uint32_t *uaddr, int op, uint32_t val, uint64_t timeout,
 	case FUTEX_WAIT:
 		val3 = FUTEX_BITSET_MATCH_ANY;
 	case FUTEX_WAIT_BITSET:
-		ret = futex_wait(uaddr, fshared, val, timeout, val3, clockrt, clv_override);
+		ret = futex_wait(uaddr, fshared, val, timeout, val3, clockrt);
 		break;
 	case FUTEX_WAKE:
 		val3 = FUTEX_BITSET_MATCH_ANY;
 	case FUTEX_WAKE_BITSET:
-		ret = futex_wake(uaddr, fshared, val, val3, clv_override);
+		ret = futex_wake(uaddr, fshared, val, val3);
 		break;
 	case FUTEX_REQUEUE:
-		ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, NULL, 0, clv_override);
+		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, clv_override);
+		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, clv_override);
+		ret = futex_wake_op(uaddr, fshared, uaddr2, val, val2, val3);
 		break;
 	/* RIKEN: these calls are not supported for now.	
 	case FUTEX_LOCK_PI:

kernel/host.c

@@ -34,13 +34,13 @@
 #include <sysfs.h>
 #include <ihk/perfctr.h>
 #include <rusage_private.h>
-#include <debug.h>
 
 //#define DEBUG_PRINT_HOST
 
 #ifdef DEBUG_PRINT_HOST
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
+#define dkprintf kprintf
+#else
+#define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0)
 #endif
 
 /* Linux channel table, indexec by Linux CPU id */
@@ -78,6 +78,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 	unsigned long args_envs_p, args_envs_rp;
 	unsigned long s, e, up;
 	char **argv;
+	char **a;
 	int i, n, argc, envc, args_envs_npages;
 	char **env;
 	int range_npages;
@@ -305,7 +306,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 
 	/* Only unmap remote address if it wasn't specified as an argument */
 	if (!args) {
-		ihk_mc_unmap_virtual(args_envs_r, args_envs_npages);
+		ihk_mc_unmap_virtual(args_envs_r, args_envs_npages, 0);
 		ihk_mc_unmap_memory(NULL, args_envs_rp, p->args_len);
 	}
 	flush_tlb();
@@ -340,7 +341,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 
 	/* Only map remote address if it wasn't specified as an argument */
 	if (!envs) {
-		ihk_mc_unmap_virtual(args_envs_r, args_envs_npages);
+		ihk_mc_unmap_virtual(args_envs_r, args_envs_npages, 0);
 		ihk_mc_unmap_memory(NULL, args_envs_rp, p->envs_len);
 	}
 	flush_tlb();
@@ -356,13 +357,12 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 		proc->saved_cmdline_len = 0;
 	}
 
-	proc->saved_cmdline_len = p->args_len - ((argc + 2) * sizeof(char **));
-	proc->saved_cmdline = kmalloc(proc->saved_cmdline_len,
-				      IHK_MC_AP_NOWAIT);
+	proc->saved_cmdline = kmalloc(p->args_len, IHK_MC_AP_NOWAIT);
 	if (!proc->saved_cmdline) {
 		goto err;
 	}
 
+	proc->saved_cmdline_len = p->args_len - ((argc + 2) * sizeof(char **));
 	memcpy(proc->saved_cmdline,
 			(char *)args_envs + ((argc + 2) * sizeof(char **)),
 			proc->saved_cmdline_len);
@@ -370,18 +370,21 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 			__FUNCTION__,
 			proc->saved_cmdline);
 
-	for (i = 0; i < argc; i++) {
-		// Process' address space!
-		argv[i] = (char *)addr + (unsigned long)argv[i];
+	for (a = argv; *a; a++) {
+		*a = (char *)addr + (unsigned long)*a; // Process' address space!
 	}
 
 	envc = *((long *)(args_envs + p->args_len));
 	dkprintf("envc: %d\n", envc);
 
 	env = (char **)(args_envs + p->args_len + sizeof(long));
-	for (i = 0; i < envc; i++) {
-		env[i] = addr + p->args_len + env[i];
+	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(long));
 
 	dkprintf("env OK\n");
 
@@ -446,7 +449,7 @@ static int process_msg_prepare_process(unsigned long rphys)
 	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);
+		ihk_mc_unmap_virtual(p, npages, 0);
 		ihk_mc_unmap_memory(NULL, phys, sz);
 		return -ENOMEM;
 	}
@@ -457,7 +460,7 @@ static int process_msg_prepare_process(unsigned long rphys)
 					(unsigned long *)&p->cpu_set,
 					sizeof(p->cpu_set))) == NULL) {
 		kfree(pn);
-		ihk_mc_unmap_virtual(p, npages);
+		ihk_mc_unmap_virtual(p, npages, 1);
 		ihk_mc_unmap_memory(NULL, phys, sz);
 		return -ENOMEM;
 	}
@@ -476,6 +479,7 @@ static int process_msg_prepare_process(unsigned long rphys)
 	proc->sgid = pn->cred[6];
 	proc->fsgid = pn->cred[7];
 	proc->termsig = SIGCHLD;
+	proc->mcexec_flags = pn->mcexec_flags;
 	proc->mpol_flags = pn->mpol_flags;
 	proc->mpol_threshold = pn->mpol_threshold;
 	proc->nr_processes = pn->nr_processes;
@@ -502,9 +506,6 @@ static int process_msg_prepare_process(unsigned long rphys)
 		vm->numa_mem_policy = MPOL_BIND;
 	}
 
-	proc->uti_thread_rank = pn->uti_thread_rank;
-	proc->uti_use_last_cpu = pn->uti_use_last_cpu;
-
 #ifdef PROFILE_ENABLE
 	proc->profile = pn->profile;
 	thread->profile = pn->profile;
@@ -543,14 +544,14 @@ static int process_msg_prepare_process(unsigned long rphys)
 
 	kfree(pn);
 
-	ihk_mc_unmap_virtual(p, npages);
+	ihk_mc_unmap_virtual(p, npages, 1);
 	ihk_mc_unmap_memory(NULL, phys, sz);
 	flush_tlb();
 
 	return 0;
 err:
 	kfree(pn);
-	ihk_mc_unmap_virtual(p, npages);
+	ihk_mc_unmap_virtual(p, npages, 1);
 	ihk_mc_unmap_memory(NULL, phys, sz);
 	destroy_thread(thread);
 	return -ENOMEM;
@@ -563,6 +564,7 @@ static void syscall_channel_send(struct ihk_ikc_channel_desc *c,
 }
 
 extern unsigned long do_kill(struct thread *, int, int, int, struct siginfo *, int ptracecont);
+extern void process_procfs_request(struct ikc_scd_packet *rpacket);
 extern void terminate_host(int pid);
 extern void debug_log(long);
 
@@ -573,6 +575,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	struct ikc_scd_packet pckt;
 	struct ihk_ikc_channel_desc *resp_channel = cpu_local_var(ikc2linux);
 	int rc;
+	struct mcs_rwlock_node_irqsave lock;
 	struct thread *thread;
 	struct process *proc;
 	struct mcctrl_signal {
@@ -596,9 +599,14 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 
 	case SCD_MSG_PREPARE_PROCESS:
 
-		pckt.err = process_msg_prepare_process(packet->arg);
-		pckt.msg = SCD_MSG_PREPARE_PROCESS_ACKED;
-		pckt.reply = packet->reply;
+		if((rc = process_msg_prepare_process(packet->arg)) == 0){
+			pckt.msg = SCD_MSG_PREPARE_PROCESS_ACKED;
+			pckt.err = 0;
+		}
+		else{
+			pckt.msg = SCD_MSG_PREPARE_PROCESS_NACKED;
+			pckt.err = rc;
+		}
 		pckt.ref = packet->ref;
 		pckt.arg = packet->arg;
 		syscall_channel_send(resp_channel, &pckt);
@@ -609,7 +617,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	case SCD_MSG_SCHEDULE_PROCESS:
 		thread = (struct thread *)packet->arg;
 
-		cpuid = obtain_clone_cpuid(&thread->cpu_set, 0);
+		cpuid = obtain_clone_cpuid(&thread->cpu_set);
 		if (cpuid == -1) {
 			kprintf("No CPU available\n");
 			ret = -1;
@@ -633,14 +641,14 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	 * the waiting thread
 	 */
 	case SCD_MSG_WAKE_UP_SYSCALL_THREAD:
-		thread = find_thread(0, packet->ttid);
+		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);
+		thread_unlock(thread, &lock);
 
 		dkprintf("%s: SCD_MSG_WAKE_UP_SYSCALL_THREAD: waking up tid %d\n",
 			__FUNCTION__, packet->ttid);
@@ -652,13 +660,12 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		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);
 		memcpy(&info, sp, sizeof(struct mcctrl_signal));
-		ihk_mc_unmap_virtual(sp, 1);
+		ihk_mc_unmap_virtual(sp, 1, 0);
 		ihk_mc_unmap_memory(NULL, pp, sizeof(struct mcctrl_signal));
-		pckt.msg = SCD_MSG_SEND_SIGNAL_ACK;
+		pckt.msg = SCD_MSG_SEND_SIGNAL;
 		pckt.err = 0;
 		pckt.ref = packet->ref;
 		pckt.arg = packet->arg;
-		pckt.reply = packet->reply;
 		syscall_channel_send(resp_channel, &pckt);
 
 		rc = do_kill(NULL, info.pid, info.tid, info.sig, &info.info, 0);
@@ -667,14 +674,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		break;
 
 	case SCD_MSG_PROCFS_REQUEST:
-	case SCD_MSG_PROCFS_RELEASE:
-		pckt.msg = SCD_MSG_PROCFS_ANSWER;
-		pckt.ref = packet->ref;
-		pckt.arg = packet->arg;
-		pckt.err = process_procfs_request(packet);
-		pckt.reply = packet->reply;
-		pckt.pid = packet->pid;
-		syscall_channel_send(resp_channel, &pckt);
+		process_procfs_request(packet);
 		ret = 0;
 		break;
 
@@ -711,26 +711,17 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 			if (!pcd->exclude_user) {
 				mode |= PERFCTR_USER_MODE;
 			}
-
-			ret = ihk_mc_perfctr_init_raw(pcd->target_cntr, pcd->config, mode);
-			if (ret != 0) {
-				break;
-			}
-
-			ret = ihk_mc_perfctr_stop(1 << pcd->target_cntr);
-			if (ret != 0) {
-				break;
-			}
-
-			ret = ihk_mc_perfctr_reset(pcd->target_cntr);
+			ihk_mc_perfctr_init_raw(pcd->target_cntr, pcd->config, mode);
+			ihk_mc_perfctr_stop(1 << pcd->target_cntr);
+			ihk_mc_perfctr_reset(pcd->target_cntr);
 			break;
 
 		case PERF_CTRL_ENABLE:
-			ret = ihk_mc_perfctr_start(pcd->target_cntr_mask);
+			ihk_mc_perfctr_start(pcd->target_cntr_mask);
 			break;
 			
 		case PERF_CTRL_DISABLE:
-			ret = ihk_mc_perfctr_stop(pcd->target_cntr_mask);
+			ihk_mc_perfctr_stop(pcd->target_cntr_mask);
 			break;
 
 		case PERF_CTRL_GET:
@@ -741,15 +732,15 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 			kprintf("%s: SCD_MSG_PERF_CTRL unexpected ctrl_type\n", __FUNCTION__);
 		}
 
-		ihk_mc_unmap_virtual(pcd, 1);
+		ihk_mc_unmap_virtual(pcd, 1, 0);
 		ihk_mc_unmap_memory(NULL, pp, sizeof(struct perf_ctrl_desc));
 
 		pckt.msg = SCD_MSG_PERF_ACK;
-		pckt.err = ret;
+		pckt.err = 0;
 		pckt.arg = packet->arg;
-		pckt.reply = packet->reply;
 		ihk_ikc_send(resp_channel, &pckt, 0);
 
+		ret = 0;
 		break;
 
 	case SCD_MSG_CPU_RW_REG:

kernel/hugefileobj.c

@@ -1,303 +0,0 @@
-#include <memobj.h>
-#include <ihk/mm.h>
-#include <kmsg.h>
-#include <kmalloc.h>
-#include <string.h>
-#include <debug.h>
-
-#if DEBUG_HUGEFILEOBJ
-#undef DDEBUG_DEFAULT
-#define DDEBUG_DEFAULT DDEBUG_PRINT
-#endif
-
-struct hugefilechunk {
-	struct list_head list;
-	off_t pgoff;
-	int npages;
-	void *mem;
-};
-
-struct hugefileobj {
-	struct memobj memobj;
-	size_t pgsize;
-	uintptr_t handle;
-	unsigned int pgshift;
-	struct list_head chunk_list;
-	ihk_spinlock_t chunk_lock;
-	struct list_head obj_list;
-};
-
-static ihk_spinlock_t hugefileobj_list_lock;
-static LIST_HEAD(hugefileobj_list);
-
-static struct hugefileobj *to_hugefileobj(struct memobj *memobj)
-{
-	return (struct hugefileobj *)memobj;
-}
-
-static struct memobj *to_memobj(struct hugefileobj *obj)
-{
-	return &obj->memobj;
-}
-
-static struct hugefileobj *hugefileobj_lookup(uintptr_t handle)
-{
-	struct hugefileobj *p;
-
-	list_for_each_entry(p, &hugefileobj_list, obj_list) {
-		if (p->handle == handle) {
-			/* for the interval between last put and fileobj_free
-			 * taking list_lock
-			 */
-			if (memobj_ref(&p->memobj) <= 1) {
-				ihk_atomic_dec(&p->memobj.refcnt);
-				continue;
-			}
-			return p;
-		}
-	}
-
-	return NULL;
-}
-
-static int hugefileobj_get_page(struct memobj *memobj, off_t off,
-				int p2align, uintptr_t *physp,
-				unsigned long *pflag, uintptr_t virt_addr)
-{
-	struct hugefileobj *obj = to_hugefileobj(memobj);
-	struct hugefilechunk *chunk;
-	off_t pgoff;
-
-	if (p2align != obj->pgshift - PTL1_SHIFT) {
-		kprintf("%s: p2align %d but expected %d\n",
-			__func__, p2align, obj->pgshift - PTL1_SHIFT);
-		return -ENOMEM;
-	}
-
-	pgoff = off >> obj->pgshift;
-	ihk_mc_spinlock_lock_noirq(&obj->chunk_lock);
-	list_for_each_entry(chunk, &obj->chunk_list, list) {
-		if (pgoff >= chunk->pgoff + chunk->npages)
-			continue;
-		if (pgoff >= chunk->pgoff)
-			break;
-		kprintf("%s: no segment found for pgoff %lx (obj %p)\n",
-			__func__, pgoff, obj);
-		chunk = NULL;
-		break;
-	}
-	ihk_mc_spinlock_unlock_noirq(&obj->chunk_lock);
-	if (!chunk)
-		return -EIO;
-
-	*physp = virt_to_phys(chunk->mem + (off - chunk->pgoff * PAGE_SIZE));
-
-	return 0;
-}
-
-static void hugefileobj_free(struct memobj *memobj)
-{
-	struct hugefileobj *obj = to_hugefileobj(memobj);
-	struct hugefilechunk *chunk, *next;
-
-	dkprintf("Destroying hugefileobj %p\n", memobj);
-
-	ihk_mc_spinlock_lock_noirq(&hugefileobj_list_lock);
-	list_del(&obj->obj_list);
-	ihk_mc_spinlock_unlock_noirq(&hugefileobj_list_lock);
-
-	kfree(memobj->path);
-	/* don't bother with chunk_lock, memobj refcounting makes this safe */
-	list_for_each_entry_safe(chunk, next, &obj->chunk_list, list) {
-		ihk_mc_free_pages_user(chunk->mem, chunk->npages);
-		kfree(chunk);
-	}
-	kfree(memobj);
-}
-
-struct memobj_ops hugefileobj_ops = {
-	.free = hugefileobj_free,
-	.get_page = hugefileobj_get_page,
-
-};
-
-void hugefileobj_cleanup(void)
-{
-	struct hugefileobj *obj;
-	int refcnt;
-
-	while (true) {
-		ihk_mc_spinlock_lock_noirq(&hugefileobj_list_lock);
-		if (list_empty(&hugefileobj_list)) {
-			ihk_mc_spinlock_unlock_noirq(&hugefileobj_list_lock);
-			break;
-		}
-		obj = list_first_entry(&hugefileobj_list, struct hugefileobj,
-				       obj_list);
-		ihk_mc_spinlock_unlock_noirq(&hugefileobj_list_lock);
-
-		if ((refcnt = memobj_unref(to_memobj(obj))) != 0) {
-			kprintf("%s: obj %p had refcnt %ld > 1, destroying anyway\n",
-				__func__, obj, refcnt + 1);
-			hugefileobj_free(to_memobj(obj));
-		}
-	}
-}
-
-int hugefileobj_pre_create(struct pager_create_result *result,
-			   struct memobj **objp, int *maxprotp)
-{
-	struct hugefileobj *obj;
-
-	ihk_mc_spinlock_lock_noirq(&hugefileobj_list_lock);
-	obj = hugefileobj_lookup(result->handle);
-	if (obj)
-		goto out_unlock;
-
-	obj = kmalloc(sizeof(*obj), IHK_MC_AP_NOWAIT);
-	if (!obj)
-		return -ENOMEM;
-
-	obj->handle = result->handle;
-	obj->pgsize = result->size;
-	obj->pgshift = 0;
-	INIT_LIST_HEAD(&obj->chunk_list);
-	ihk_mc_spinlock_init(&obj->chunk_lock);
-	obj->memobj.flags = result->flags;
-	obj->memobj.status = MEMOBJ_TO_BE_PREFETCHED;
-	obj->memobj.ops = &hugefileobj_ops;
-	/* keep mapping around when process is gone */
-	ihk_atomic_set(&obj->memobj.refcnt, 2);
-	if (result->path[0]) {
-		obj->memobj.path = kmalloc(PATH_MAX, IHK_MC_AP_NOWAIT);
-		if (!obj->memobj.path) {
-			kfree(obj);
-			return -ENOMEM;
-		}
-		strncpy(obj->memobj.path, result->path, PATH_MAX);
-	}
-
-	list_add(&obj->obj_list, &hugefileobj_list);
-out_unlock:
-	ihk_mc_spinlock_unlock_noirq(&hugefileobj_list_lock);
-
-	*maxprotp = result->maxprot;
-	*objp = to_memobj(obj);
-
-	return 0;
-}
-
-int hugefileobj_create(struct memobj *memobj, size_t len, off_t off,
-		       int *pgshiftp, uintptr_t virt_addr)
-{
-	struct hugefileobj *obj = to_hugefileobj(memobj);
-	struct hugefilechunk *chunk = NULL, *old_chunk = NULL;
-	int p2align;
-	unsigned int pgshift;
-	int npages, npages_left;
-	void *v;
-	off_t pgoff, next_pgoff;
-	int error;
-
-	error = arch_get_smaller_page_size(NULL, obj->pgsize + 1, NULL,
-					   &p2align);
-	if (error)
-		return error;
-	pgshift = p2align + PTL1_SHIFT;
-	if (1 << pgshift != obj->pgsize) {
-		dkprintf("invalid hugefileobj pagesize: %d\n",
-			obj->pgsize);
-		return -EINVAL;
-	}
-
-	if (len & ((1 << pgshift) - 1)) {
-		dkprintf("invalid hugetlbfs mmap size %d (pagesize %d)\n",
-			len, 1 << pgshift);
-		obj->pgshift = 0;
-		return -EINVAL;
-	}
-	if (off & ((1 << pgshift) - 1)) {
-		dkprintf("invalid hugetlbfs mmap offset %d (pagesize %d)\n",
-			off, 1 << pgshift);
-		obj->pgshift = 0;
-		return -EINVAL;
-	}
-
-
-	ihk_mc_spinlock_lock_noirq(&obj->chunk_lock);
-	if (obj->pgshift && obj->pgshift != pgshift) {
-		kprintf("pgshift changed between two calls on same inode?! had %d now %d\n",
-			obj->pgshift, pgshift);
-		ihk_mc_spinlock_unlock_noirq(&obj->chunk_lock);
-		return -EINVAL;
-	}
-	obj->pgshift = pgshift;
-
-	/* Prealloc upfront, we need to fail here if not enough memory. */
-	if (!list_empty(&obj->chunk_list))
-		old_chunk = list_first_entry(&obj->chunk_list,
-					     struct hugefilechunk, list);
-	pgoff = off >> PAGE_SHIFT;
-	npages_left = len >> PAGE_SHIFT;
-	npages = npages_left;
-	while (npages_left) {
-		while (old_chunk &&
-				pgoff >= old_chunk->pgoff + old_chunk->npages) {
-			if (list_is_last(&old_chunk->list, &obj->chunk_list)) {
-				old_chunk = NULL;
-				break;
-			}
-			old_chunk = list_entry(old_chunk->list.next,
-					       struct hugefilechunk, list);
-		}
-		if (old_chunk) {
-			next_pgoff = old_chunk->pgoff + old_chunk->npages;
-			if (pgoff >= old_chunk->pgoff && pgoff < next_pgoff) {
-				npages_left -= next_pgoff - pgoff;
-				pgoff = next_pgoff;
-				continue;
-			}
-		}
-		if (!chunk) {
-			chunk = kmalloc(sizeof(*chunk), IHK_MC_AP_NOWAIT);
-		}
-		if (!chunk) {
-			kprintf("could not allocate hugefileobj chunk\n");
-			return -ENOMEM;
-		}
-		if (npages > npages_left)
-			npages = npages_left;
-		v = ihk_mc_alloc_aligned_pages_user(npages, p2align,
-				IHK_MC_AP_NOWAIT | IHK_MC_AP_USER, virt_addr);
-		if (!v) {
-			if (npages == 1) {
-				dkprintf("could not allocate more pages wth pgshift %d\n",
-					 pgshift);
-				kfree(chunk);
-				/* caller will cleanup the rest */
-				return -ENOMEM;
-			}
-			/* exponential backoff, try less aggressive? */
-			npages /= 2;
-			continue;
-		}
-		memset(v, 0, npages * PAGE_SIZE);
-		chunk->npages = npages;
-		chunk->mem = v;
-		chunk->pgoff = pgoff;
-		/* ordered list: insert before next (bigger) element */
-		if (old_chunk)
-			list_add(&chunk->list, old_chunk->list.prev);
-		else
-			list_add(&chunk->list, obj->chunk_list.prev);
-		pgoff += npages;
-		npages_left -= npages;
-	}
-	obj->memobj.size = len;
-
-	ihk_mc_spinlock_unlock_noirq(&obj->chunk_lock);
-
-	*pgshiftp = pgshift;
-
-	return 0;
-}

kernel/include/cls.h

@@ -19,10 +19,17 @@
  * CPU Local Storage (cls)
  */
 
+struct kmalloc_cache_header {
+	struct kmalloc_cache_header *next;
+};
+
 struct kmalloc_header {
 	unsigned int front_magic;
 	int cpu_id;
-	struct list_head list;
+	union {
+		struct list_head list;
+		struct kmalloc_cache_header *cache;
+	};
 	int size; /* The size of this chunk without the header */
 	unsigned int end_magic;
 	/* 32 bytes */
@@ -74,7 +81,6 @@ struct cpu_local_var {
 	struct thread *current;
 	struct list_head runq;
 	size_t runq_len;
-	size_t runq_reserved; /* Number of threads which are about to be added to runq */
 
 	struct ihk_ikc_channel_desc *ikc2linux;
 
@@ -101,8 +107,11 @@ struct cpu_local_var {
 
 	struct process_vm *on_fork_vm;
 
-	/* UTI */
-	void *uti_futex_resp;
+	/* HFI1 related per-core kmalloc caches */
+	struct kmalloc_cache_header txreq_cache;
+	struct kmalloc_cache_header tids_cache;
+	struct kmalloc_cache_header tidlist_cache;
+	struct kmalloc_cache_header tid_node_cache;
 } __attribute__((aligned(64)));
 
 
@@ -114,6 +123,4 @@ static struct cpu_local_var *get_this_cpu_local_var(void)
 
 #define cpu_local_var(name) get_this_cpu_local_var()->name
 
-#define cpu_local_var_with_override(name, clv_override) (clv_override ? clv_override->name : get_this_cpu_local_var()->name)
-
 #endif

kernel/include/debug.h

@@ -1,54 +0,0 @@
-#ifndef DEBUG_H
-#define DEBUG_H
-
-#include "lwk/compiler.h"
-
-void panic(const char *);
-
-/* when someone has a lot of time, add attribute __printf(1, 2) to kprintf */
-int kprintf(const char *format, ...);
-
-struct ddebug {
-	const char *file;
-	const char *func;
-	const char *fmt;
-	unsigned int line:24;
-	unsigned int flags:8;
-} __aligned(8);
-
-#define DDEBUG_NONE  0x0
-#define DDEBUG_PRINT 0x1
-
-#define DDEBUG_DEFAULT DDEBUG_NONE
-
-#define DDEBUG_SYMBOL()                                  \
-	static struct ddebug __aligned(8)                \
-	__attribute__((section("__verbose"))) ddebug = { \
-		.file = __FILE__,                        \
-		.func = __func__,                        \
-		.line = __LINE__,                        \
-		.flags = DDEBUG_DEFAULT,                 \
-	}
-
-#define DDEBUG_TEST ddebug.flags
-
-
-
-#define dkprintf(fmt, args...)        \
-do {                                  \
-	DDEBUG_SYMBOL();              \
-	if (DDEBUG_TEST)              \
-		kprintf(fmt, ##args); \
-} while (0)
-#define ekprintf(fmt, args...) kprintf(fmt, ##args)
-
-#define BUG_ON(condition) do {                         \
-	if (condition) {                               \
-		kprintf("PANIC: %s: %s(line:%d)\n",    \
-			__FILE__, __func__, __LINE__); \
-		panic("");                             \
-	}                                              \
-} while (0)
-#define BUILD_BUG_ON(condition) ((void)sizeof(char[1 - 2*!!(condition)]))
-
-#endif

kernel/include/futex.h

@@ -63,7 +63,7 @@
 #define FUTEX_OP_ANDN		3	/* *(int *)UADDR2 &= ~OPARG; */
 #define FUTEX_OP_XOR		4	/* *(int *)UADDR2 ^= OPARG; */
 
-#define FUTEX_OP_OPARG_SHIFT	8U	/* Use (1 << OPARG) instead of OPARG. */
+#define FUTEX_OP_OPARG_SHIFT	8	/* Use (1 << OPARG) instead of OPARG.  */
 
 #define FUTEX_OP_CMP_EQ		0	/* if (oldval == CMPARG) wake */
 #define FUTEX_OP_CMP_NE		1	/* if (oldval != CMPARG) wake */
@@ -150,7 +150,6 @@ union futex_key {
 
 extern int futex_init(void);
 
-struct cpu_local_var;
 extern int
 futex(
 	uint32_t __user *		uaddr,
@@ -160,8 +159,7 @@ futex(
 	uint32_t __user *		uaddr2,
 	uint32_t				val2,
 	uint32_t				val3,
-	int                     fshared,
-	struct cpu_local_var *clv_override
+	int                     fshared
 );
 
 

kernel/include/hfi1/chip.h

@@ -0,0 +1,60 @@
+#ifndef _CHIP_H
+#define _CHIP_H
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+/*
+ * This file contains all of the defines that is specific to the HFI chip
+ */
+
+#define MAX_EXPECTED_BUFFER    (2048 * 1024)
+
+void hfi1_put_tid(struct hfi1_devdata *dd, u32 index,
+                  u32 type, unsigned long pa, u16 order);
+void hfi1_clear_tids(struct hfi1_ctxtdata *rcd);
+
+#endif /* _CHIP_H */

kernel/include/hfi1/chip_registers.h

@@ -0,0 +1,64 @@
+#ifndef DEF_CHIP_REG
+#define DEF_CHIP_REG
+
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#define CORE            0x000000000000
+
+
+#define RXE                     (CORE + 0x000001000000)
+
+
+#define RCV_ARRAY (RXE + 0x000000200000)
+#define RCV_ARRAY_CNT (RXE + 0x000000000018)
+#define RCV_ARRAY_RT_ADDR_MASK 0xFFFFFFFFFull
+#define RCV_ARRAY_RT_ADDR_SHIFT 0
+#define RCV_ARRAY_RT_BUF_SIZE_SHIFT 36
+#define RCV_ARRAY_RT_WRITE_ENABLE_SMASK 0x8000000000000000ull
+
+#endif          /* DEF_CHIP_REG */

kernel/include/hfi1/common.h

@@ -0,0 +1,411 @@
+/*
+ * Copyright(c) 2015, 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *  - Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  - Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#ifndef _COMMON_H
+#define _COMMON_H
+
+#ifdef __HFI1_ORIG__
+#include "update/hfi1_user.h"
+#else
+#include <hfi1/hfi1_user.h>
+#endif /* __HFI1_ORIG__ */
+
+/*
+ * This file contains defines, structures, etc. that are used
+ * to communicate between kernel and user code.
+ */
+
+/* version of protocol header (known to chip also). In the long run,
+ * we should be able to generate and accept a range of version numbers;
+ * for now we only accept one, and it's compiled in.
+ */
+#define IPS_PROTO_VERSION 2
+
+/*
+ * These are compile time constants that you may want to enable or disable
+ * if you are trying to debug problems with code or performance.
+ * HFI1_VERBOSE_TRACING define as 1 if you want additional tracing in
+ * fast path code
+ * HFI1_TRACE_REGWRITES define as 1 if you want register writes to be
+ * traced in fast path code
+ * _HFI1_TRACING define as 0 if you want to remove all tracing in a
+ * compilation unit
+ */
+
+/*
+ * If a packet's QP[23:16] bits match this value, then it is
+ * a PSM packet and the hardware will expect a KDETH header
+ * following the BTH.
+ */
+#define DEFAULT_KDETH_QP 0x80
+
+/* driver/hw feature set bitmask */
+#define HFI1_CAP_USER_SHIFT      24
+#define HFI1_CAP_MASK            ((1UL << HFI1_CAP_USER_SHIFT) - 1)
+/* locked flag - if set, only HFI1_CAP_WRITABLE_MASK bits can be set */
+#define HFI1_CAP_LOCKED_SHIFT    63
+#define HFI1_CAP_LOCKED_MASK     0x1ULL
+#define HFI1_CAP_LOCKED_SMASK    (HFI1_CAP_LOCKED_MASK << HFI1_CAP_LOCKED_SHIFT)
+/* extra bits used between kernel and user processes */
+#define HFI1_CAP_MISC_SHIFT      (HFI1_CAP_USER_SHIFT * 2)
+#define HFI1_CAP_MISC_MASK       ((1ULL << (HFI1_CAP_LOCKED_SHIFT - \
+					   HFI1_CAP_MISC_SHIFT)) - 1)
+
+#define HFI1_CAP_KSET(cap) ({ hfi1_cap_mask |= HFI1_CAP_##cap; hfi1_cap_mask; })
+#define HFI1_CAP_KCLEAR(cap)						\
+	({								\
+		hfi1_cap_mask &= ~HFI1_CAP_##cap;			\
+		hfi1_cap_mask;						\
+	})
+#define HFI1_CAP_USET(cap)						\
+	({								\
+		hfi1_cap_mask |= (HFI1_CAP_##cap << HFI1_CAP_USER_SHIFT); \
+		hfi1_cap_mask;						\
+		})
+#define HFI1_CAP_UCLEAR(cap)						\
+	({								\
+		hfi1_cap_mask &= ~(HFI1_CAP_##cap << HFI1_CAP_USER_SHIFT); \
+		hfi1_cap_mask;						\
+	})
+#define HFI1_CAP_SET(cap)						\
+	({								\
+		hfi1_cap_mask |= (HFI1_CAP_##cap | (HFI1_CAP_##cap <<	\
+						  HFI1_CAP_USER_SHIFT)); \
+		hfi1_cap_mask;						\
+	})
+#define HFI1_CAP_CLEAR(cap)						\
+	({								\
+		hfi1_cap_mask &= ~(HFI1_CAP_##cap |			\
+				  (HFI1_CAP_##cap << HFI1_CAP_USER_SHIFT)); \
+		hfi1_cap_mask;						\
+	})
+#define HFI1_CAP_LOCK()							\
+	({ hfi1_cap_mask |= HFI1_CAP_LOCKED_SMASK; hfi1_cap_mask; })
+#define HFI1_CAP_LOCKED() (!!(hfi1_cap_mask & HFI1_CAP_LOCKED_SMASK))
+/*
+ * The set of capability bits that can be changed after initial load
+ * This set is the same for kernel and user contexts. However, for
+ * user contexts, the set can be further filtered by using the
+ * HFI1_CAP_RESERVED_MASK bits.
+ */
+#define HFI1_CAP_WRITABLE_MASK   (HFI1_CAP_SDMA_AHG |			\
+				  HFI1_CAP_HDRSUPP |			\
+				  HFI1_CAP_MULTI_PKT_EGR |		\
+				  HFI1_CAP_NODROP_RHQ_FULL |		\
+				  HFI1_CAP_NODROP_EGR_FULL |		\
+				  HFI1_CAP_ALLOW_PERM_JKEY |		\
+				  HFI1_CAP_STATIC_RATE_CTRL |		\
+				  HFI1_CAP_PRINT_UNIMPL |		\
+				  HFI1_CAP_TID_UNMAP |			\
+				  HFI1_CAP_OPFN |			\
+				  HFI1_CAP_TID_RDMA)
+/*
+ * A set of capability bits that are "global" and are not allowed to be
+ * set in the user bitmask.
+ */
+#define HFI1_CAP_RESERVED_MASK   ((HFI1_CAP_SDMA |			\
+				   HFI1_CAP_USE_SDMA_HEAD |		\
+				   HFI1_CAP_EXTENDED_PSN |		\
+				   HFI1_CAP_PRINT_UNIMPL |		\
+				   HFI1_CAP_NO_INTEGRITY |		\
+				   HFI1_CAP_PKEY_CHECK |		\
+				   HFI1_CAP_TID_RDMA |			\
+				   HFI1_CAP_OPFN) <<			\
+				  HFI1_CAP_USER_SHIFT)
+/*
+ * Set of capabilities that need to be enabled for kernel context in
+ * order to be allowed for user contexts, as well.
+ */
+#define HFI1_CAP_MUST_HAVE_KERN (HFI1_CAP_STATIC_RATE_CTRL)
+/* Default enabled capabilities (both kernel and user) */
+#define HFI1_CAP_MASK_DEFAULT    (HFI1_CAP_HDRSUPP |			\
+				 HFI1_CAP_NODROP_RHQ_FULL |		\
+				 HFI1_CAP_NODROP_EGR_FULL |		\
+				 HFI1_CAP_SDMA |			\
+				 HFI1_CAP_PRINT_UNIMPL |		\
+				 HFI1_CAP_STATIC_RATE_CTRL |		\
+				 HFI1_CAP_PKEY_CHECK |			\
+				 HFI1_CAP_MULTI_PKT_EGR |		\
+				 HFI1_CAP_EXTENDED_PSN |		\
+				 ((HFI1_CAP_HDRSUPP |			\
+				   HFI1_CAP_MULTI_PKT_EGR |		\
+				   HFI1_CAP_STATIC_RATE_CTRL |		\
+				   HFI1_CAP_PKEY_CHECK |		\
+				   HFI1_CAP_EARLY_CREDIT_RETURN) <<	\
+				  HFI1_CAP_USER_SHIFT))
+/*
+ * A bitmask of kernel/global capabilities that should be communicated
+ * to user level processes.
+ */
+#define HFI1_CAP_K2U (HFI1_CAP_SDMA |			\
+		     HFI1_CAP_EXTENDED_PSN |		\
+		     HFI1_CAP_PKEY_CHECK |		\
+		     HFI1_CAP_NO_INTEGRITY)
+
+#define HFI1_USER_SWVERSION ((HFI1_USER_SWMAJOR << HFI1_SWMAJOR_SHIFT) | \
+			     HFI1_USER_SWMINOR)
+
+#ifndef HFI1_KERN_TYPE
+#define HFI1_KERN_TYPE 0
+#endif
+
+/*
+ * Similarly, this is the kernel version going back to the user.  It's
+ * slightly different, in that we want to tell if the driver was built as
+ * part of a Intel release, or from the driver from openfabrics.org,
+ * kernel.org, or a standard distribution, for support reasons.
+ * The high bit is 0 for non-Intel and 1 for Intel-built/supplied.
+ *
+ * It's returned by the driver to the user code during initialization in the
+ * spi_sw_version field of hfi1_base_info, so the user code can in turn
+ * check for compatibility with the kernel.
+*/
+#define HFI1_KERN_SWVERSION ((HFI1_KERN_TYPE << 31) | HFI1_USER_SWVERSION)
+
+/*
+ * Define the driver version number.  This is something that refers only
+ * to the driver itself, not the software interfaces it supports.
+ */
+#ifndef HFI1_DRIVER_VERSION_BASE
+#define HFI1_DRIVER_VERSION_BASE "0.9-294"
+#endif
+
+/* create the final driver version string */
+#ifdef HFI1_IDSTR
+#define HFI1_DRIVER_VERSION HFI1_DRIVER_VERSION_BASE " " HFI1_IDSTR
+#else
+#define HFI1_DRIVER_VERSION HFI1_DRIVER_VERSION_BASE
+#endif
+
+/*
+ * Diagnostics can send a packet by writing the following
+ * struct to the diag packet special file.
+ *
+ * This allows a custom PBC qword, so that special modes and deliberate
+ * changes to CRCs can be used.
+ */
+#define _DIAG_PKT_VERS 1
+struct diag_pkt {
+	__u16 version;		/* structure version */
+	__u16 unit;		/* which device */
+	__u16 sw_index;		/* send sw index to use */
+	__u16 len;		/* data length, in bytes */
+	__u16 port;		/* port number */
+	__u16 unused;
+	__u32 flags;		/* call flags */
+	__u64 data;		/* user data pointer */
+	__u64 pbc;		/* PBC for the packet */
+};
+
+/* diag_pkt flags */
+#define F_DIAGPKT_WAIT 0x1	/* wait until packet is sent */
+
+/*
+ * The next set of defines are for packet headers, and chip register
+ * and memory bits that are visible to and/or used by user-mode software.
+ */
+
+/*
+ * Receive Header Flags
+ */
+#define RHF_PKT_LEN_SHIFT	0
+#define RHF_PKT_LEN_MASK	0xfffull
+#define RHF_PKT_LEN_SMASK (RHF_PKT_LEN_MASK << RHF_PKT_LEN_SHIFT)
+
+#define RHF_RCV_TYPE_SHIFT	12
+#define RHF_RCV_TYPE_MASK	0x7ull
+#define RHF_RCV_TYPE_SMASK (RHF_RCV_TYPE_MASK << RHF_RCV_TYPE_SHIFT)
+
+#define RHF_USE_EGR_BFR_SHIFT	15
+#define RHF_USE_EGR_BFR_MASK	0x1ull
+#define RHF_USE_EGR_BFR_SMASK (RHF_USE_EGR_BFR_MASK << RHF_USE_EGR_BFR_SHIFT)
+
+#define RHF_EGR_INDEX_SHIFT	16
+#define RHF_EGR_INDEX_MASK	0x7ffull
+#define RHF_EGR_INDEX_SMASK (RHF_EGR_INDEX_MASK << RHF_EGR_INDEX_SHIFT)
+
+#define RHF_DC_INFO_SHIFT	27
+#define RHF_DC_INFO_MASK	0x1ull
+#define RHF_DC_INFO_SMASK (RHF_DC_INFO_MASK << RHF_DC_INFO_SHIFT)
+
+#define RHF_RCV_SEQ_SHIFT	28
+#define RHF_RCV_SEQ_MASK	0xfull
+#define RHF_RCV_SEQ_SMASK (RHF_RCV_SEQ_MASK << RHF_RCV_SEQ_SHIFT)
+
+#define RHF_EGR_OFFSET_SHIFT	32
+#define RHF_EGR_OFFSET_MASK	0xfffull
+#define RHF_EGR_OFFSET_SMASK (RHF_EGR_OFFSET_MASK << RHF_EGR_OFFSET_SHIFT)
+#define RHF_HDRQ_OFFSET_SHIFT	44
+#define RHF_HDRQ_OFFSET_MASK	0x1ffull
+#define RHF_HDRQ_OFFSET_SMASK (RHF_HDRQ_OFFSET_MASK << RHF_HDRQ_OFFSET_SHIFT)
+#define RHF_K_HDR_LEN_ERR	(0x1ull << 53)
+#define RHF_DC_UNC_ERR		(0x1ull << 54)
+#define RHF_DC_ERR		(0x1ull << 55)
+#define RHF_RCV_TYPE_ERR_SHIFT	56
+#define RHF_RCV_TYPE_ERR_MASK	0x7ul
+#define RHF_RCV_TYPE_ERR_SMASK (RHF_RCV_TYPE_ERR_MASK << RHF_RCV_TYPE_ERR_SHIFT)
+#define RHF_TID_ERR		(0x1ull << 59)
+#define RHF_LEN_ERR		(0x1ull << 60)
+#define RHF_ECC_ERR		(0x1ull << 61)
+#define RHF_VCRC_ERR		(0x1ull << 62)
+#define RHF_ICRC_ERR		(0x1ull << 63)
+
+#define RHF_ERROR_SMASK 0xffe0000000000000ull		/* bits 63:53 */
+
+/* RHF receive types */
+#define RHF_RCV_TYPE_EXPECTED 0
+#define RHF_RCV_TYPE_EAGER    1
+#define RHF_RCV_TYPE_IB       2 /* normal IB, IB Raw, or IPv6 */
+#define RHF_RCV_TYPE_ERROR    3
+#define RHF_RCV_TYPE_BYPASS   4
+#define RHF_RCV_TYPE_INVALID5 5
+#define RHF_RCV_TYPE_INVALID6 6
+#define RHF_RCV_TYPE_INVALID7 7
+
+/* RHF receive type error - expected packet errors */
+#define RHF_RTE_EXPECTED_FLOW_SEQ_ERR	0x2
+#define RHF_RTE_EXPECTED_FLOW_GEN_ERR	0x4
+
+/* RHF receive type error - eager packet errors */
+#define RHF_RTE_EAGER_NO_ERR		0x0
+
+/* RHF receive type error - IB packet errors */
+#define RHF_RTE_IB_NO_ERR		0x0
+
+/* RHF receive type error - error packet errors */
+#define RHF_RTE_ERROR_NO_ERR		0x0
+#define RHF_RTE_ERROR_OP_CODE_ERR	0x1
+#define RHF_RTE_ERROR_KHDR_MIN_LEN_ERR	0x2
+#define RHF_RTE_ERROR_KHDR_HCRC_ERR	0x3
+#define RHF_RTE_ERROR_KHDR_KVER_ERR	0x4
+#define RHF_RTE_ERROR_CONTEXT_ERR	0x5
+#define RHF_RTE_ERROR_KHDR_TID_ERR	0x6
+
+/* RHF receive type error - bypass packet errors */
+#define RHF_RTE_BYPASS_NO_ERR		0x0
+
+/* IB - LRH header constants */
+#define HFI1_LRH_GRH 0x0003      /* 1. word of IB LRH - next header: GRH */
+#define HFI1_LRH_BTH 0x0002      /* 1. word of IB LRH - next header: BTH */
+
+/* misc. */
+#define SIZE_OF_CRC 1
+
+#define LIM_MGMT_P_KEY       0x7FFF
+#define FULL_MGMT_P_KEY      0xFFFF
+
+#define DEFAULT_P_KEY LIM_MGMT_P_KEY
+#define HFI1_FECN_SHIFT 31
+#define HFI1_FECN_MASK 1
+#define HFI1_FECN_SMASK BIT(HFI1_FECN_SHIFT)
+#define HFI1_BECN_SHIFT 30
+#define HFI1_BECN_MASK 1
+#define HFI1_BECN_SMASK BIT(HFI1_BECN_SHIFT)
+
+#define HFI1_PSM_IOC_BASE_SEQ 0x0
+
+/* Number of BTH.PSN bits used for sequence number in expected rcvs */
+#define HFI1_KDETH_BTH_SEQ_SHIFT 11
+#define HFI1_KDETH_BTH_SEQ_MASK (BIT(HFI1_KDETH_BTH_SEQ_SHIFT) - 1)
+
+static inline __u64 rhf_to_cpu(const __le32 *rbuf)
+{
+	return __le64_to_cpu(*((__le64 *)rbuf));
+}
+
+static inline u64 rhf_err_flags(u64 rhf)
+{
+	return rhf & RHF_ERROR_SMASK;
+}
+
+static inline u32 rhf_rcv_type(u64 rhf)
+{
+	return (rhf >> RHF_RCV_TYPE_SHIFT) & RHF_RCV_TYPE_MASK;
+}
+
+static inline u32 rhf_rcv_type_err(u64 rhf)
+{
+	return (rhf >> RHF_RCV_TYPE_ERR_SHIFT) & RHF_RCV_TYPE_ERR_MASK;
+}
+
+/* return size is in bytes, not DWORDs */
+static inline u32 rhf_pkt_len(u64 rhf)
+{
+	return ((rhf & RHF_PKT_LEN_SMASK) >> RHF_PKT_LEN_SHIFT) << 2;
+}
+
+static inline u32 rhf_egr_index(u64 rhf)
+{
+	return (rhf >> RHF_EGR_INDEX_SHIFT) & RHF_EGR_INDEX_MASK;
+}
+
+static inline u32 rhf_rcv_seq(u64 rhf)
+{
+	return (rhf >> RHF_RCV_SEQ_SHIFT) & RHF_RCV_SEQ_MASK;
+}
+
+/* returned offset is in DWORDS */
+static inline u32 rhf_hdrq_offset(u64 rhf)
+{
+	return (rhf >> RHF_HDRQ_OFFSET_SHIFT) & RHF_HDRQ_OFFSET_MASK;
+}
+
+static inline u64 rhf_use_egr_bfr(u64 rhf)
+{
+	return rhf & RHF_USE_EGR_BFR_SMASK;
+}
+
+static inline u64 rhf_dc_info(u64 rhf)
+{
+	return rhf & RHF_DC_INFO_SMASK;
+}
+
+static inline u32 rhf_egr_buf_offset(u64 rhf)
+{
+	return (rhf >> RHF_EGR_OFFSET_SHIFT) & RHF_EGR_OFFSET_MASK;
+}
+#endif /* _COMMON_H */

kernel/include/hfi1/file_ops.h

@@ -0,0 +1,9 @@
+#ifndef _HFI1_FILE_OPS_H_
+#define _HFI1_FILE_OPS_H_
+
+#include <ihk/types.h>
+#include <uio.h>
+
+ssize_t hfi1_aio_write(void *private_data, const struct iovec *iovec, unsigned long dim);
+
+#endif

kernel/include/hfi1/hfi1_generated_ctxtdata.h

@@ -0,0 +1,41 @@
+struct hfi1_ctxtdata {
+	union {
+		char whole_struct[1160];
+		struct {
+			char padding0[144];
+			u16 ctxt;
+		};
+		struct {
+			char padding1[168];
+			u32 rcv_array_groups;
+		};
+		struct {
+			char padding2[172];
+			u32 eager_base;
+		};
+		struct {
+			char padding3[176];
+			u32 expected_count;
+		};
+		struct {
+			char padding4[180];
+			u32 expected_base;
+		};
+		struct {
+			char padding5[184];
+			struct exp_tid_set tid_group_list;
+		};
+		struct {
+			char padding6[208];
+			struct exp_tid_set tid_used_list;
+		};
+		struct {
+			char padding7[232];
+			struct exp_tid_set tid_full_list;
+		};
+		struct {
+			char padding8[392];
+			struct hfi1_devdata *dd;
+		};
+	};
+};

kernel/include/hfi1/hfi1_generated_devdata.h

@@ -0,0 +1,65 @@
+struct hfi1_devdata {
+	union {
+		char whole_struct[7808];
+		struct {
+			char padding0[3368];
+			u8 *kregbase1;
+		};
+		struct {
+			char padding1[3376];
+			resource_size_t physaddr;
+		};
+		struct {
+			char padding2[3704];
+			u64 default_desc1;
+		};
+		struct {
+			char padding3[3736];
+			dma_addr_t sdma_pad_phys;
+		};
+		struct {
+			char padding4[3760];
+			struct sdma_engine *per_sdma;
+		};
+		struct {
+			char padding5[3768];
+			struct sdma_vl_map *sdma_map;
+		};
+		struct {
+			char padding6[3816];
+			void *piobase;
+		};
+		struct {
+			char padding7[3824];
+			void *rcvarray_wc;
+		};
+		struct {
+			char padding8[4040];
+			long unsigned int *events;
+		};
+		struct {
+			char padding9[4076];
+			u32 chip_rcv_contexts;
+		};
+		struct {
+			char padding10[4080];
+			u32 chip_rcv_array_count;
+		};
+		struct {
+			char padding11[7264];
+			struct hfi1_pportdata *pport;
+		};
+		struct {
+			char padding12[7296];
+			u16 flags;
+		};
+		struct {
+			char padding13[7299];
+			u8 first_dyn_alloc_ctxt;
+		};
+		struct {
+			char padding14[7368];
+			u64 sc2vl[4];
+		};
+	};
+};

kernel/include/hfi1/hfi1_generated_filedata.h

@@ -0,0 +1,49 @@
+struct hfi1_filedata {
+	union {
+		char whole_struct[104];
+		struct {
+			char padding0[0];
+			struct hfi1_devdata *dd;
+		};
+		struct {
+			char padding1[8];
+			struct hfi1_ctxtdata *uctxt;
+		};
+		struct {
+			char padding2[16];
+			struct hfi1_user_sdma_comp_q *cq;
+		};
+		struct {
+			char padding3[24];
+			struct hfi1_user_sdma_pkt_q *pq;
+		};
+		struct {
+			char padding4[32];
+			u16 subctxt;
+		};
+		struct {
+			char padding5[56];
+			struct tid_rb_node **entry_to_rb;
+		};
+		struct {
+			char padding6[64];
+			spinlock_t tid_lock;
+		};
+		struct {
+			char padding7[72];
+			u32 tid_used;
+		};
+		struct {
+			char padding8[80];
+			u32 *invalid_tids;
+		};
+		struct {
+			char padding9[88];
+			u32 invalid_tid_idx;
+		};
+		struct {
+			char padding10[92];
+			spinlock_t invalid_lock;
+		};
+	};
+};

kernel/include/hfi1/hfi1_generated_hfi1_user_sdma_pkt_q.h

@@ -0,0 +1,29 @@
+struct hfi1_user_sdma_pkt_q {
+	union {
+		char whole_struct[352];
+		struct {
+			char padding0[4];
+			u16 n_max_reqs;
+		};
+		struct {
+			char padding1[8];
+			atomic_t n_reqs;
+		};
+		struct {
+			char padding2[16];
+			struct hfi1_devdata *dd;
+		};
+		struct {
+			char padding3[32];
+			struct user_sdma_request *reqs;
+		};
+		struct {
+			char padding4[40];
+			long unsigned int *req_in_use;
+		};
+		struct {
+			char padding5[288];
+			enum pkt_q_sdma_state state;
+		};
+	};
+};