configure.ac: Update version number

(1) Restart irqbalance when error occurs after it's stopped
(2) Restore /proc/irq/*/smp_affinity when error occurs after
    they're modified

Makefile.in

@@ -4,7 +4,7 @@ INCDIR =  @INCDIR@
 ETCDIR = @ETCDIR@
 MANDIR = @MANDIR@
 
-all: executer-mcctrl executer-mcoverlayfs executer-user mckernel
+all: executer-mcctrl executer-mcoverlayfs executer-user mckernel mck-tools
 
 executer-mcctrl:
 	+@(cd executer/kernel/mcctrl; $(MAKE) modules)
@@ -26,6 +26,9 @@ mckernel:
 		;; \
 	esac
 
+mck-tools:
+	+@(cd tools/mcstat; $(MAKE))
+
 install:
 	@(cd executer/kernel/mcctrl; $(MAKE) install)
 	@(cd executer/kernel/mcoverlayfs; $(MAKE) install)
@@ -60,6 +63,7 @@ install:
 		exit 1 \
 		;; \
 	esac
+	@(cd tools/mcstat/; $(MAKE) install)
 
 clean:
 	@(cd executer/kernel/mcctrl; $(MAKE) clean)
@@ -74,3 +78,4 @@ clean:
 		exit 1 \
 		;; \
 	esac
+	@(cd tools/mcstat; $(MAKE) clean)

arch/arm64/kernel/cpu.c

@@ -1172,8 +1172,6 @@ void arch_clone_thread(struct thread *othread, unsigned long pc,
 	asm("mrs %0, tpidr_el0" : "=r" (tls));
 	othread->tlsblock_base = nthread->tlsblock_base = tls;
 
-	/* copy fp_regs values from parent. */
-	save_fp_regs(othread);
 	if ((othread->fp_regs != NULL) && (check_and_allocate_fp_regs(nthread) == 0)) {
 		memcpy(nthread->fp_regs, othread->fp_regs, sizeof(fp_regs_struct));
 	}

arch/arm64/kernel/include/syscall_list.h

@@ -144,5 +144,3 @@ SYSCALL_HANDLED(1045, signalfd)
 SYSCALL_DELEGATED(1049, stat)
 SYSCALL_DELEGATED(1060, getpgrp)
 SYSCALL_DELEGATED(1062, time)
-SYSCALL_HANDLED(1071, vfork)
-SYSCALL_DELEGATED(1079, fork)

arch/arm64/kernel/memory.c

@@ -1807,7 +1807,6 @@ static int clear_range_l1(void *args0, pte_t *ptep, uint64_t base,
 			ihk_mc_free_pages_user(phys_to_virt(phys), npages);
 			dkprintf("%s: freeing regular page at 0x%lx\n", __FUNCTION__, base);
 		}
-		args->vm->currss -= PTL1_SIZE;
 	}
 	
 	return 0;
@@ -1887,7 +1886,6 @@ static int clear_range_middle(void *args0, pte_t *ptep, uint64_t base,
 				ihk_mc_free_pages_user(phys_to_virt(phys), npages);
 				dkprintf("%s(level=%d): freeing large page at 0x%lx\n", __FUNCTION__, level, base);
 			}
-			args->vm->currss -= tbl.pgsize;
 		}
 
 		return 0;

arch/arm64/kernel/syscall.c

@@ -13,6 +13,7 @@
 #include <hwcap.h>
 #include <prctl.h>
 #include <limits.h>
+#include <syscall.h>
 
 extern void ptrace_report_signal(struct thread *thread, int sig);
 extern void clear_single_step(struct thread *thread);
@@ -1322,6 +1323,17 @@ interrupt_from_user(void *regs0)
 	return((regs->pstate & PSR_MODE_MASK) == PSR_MODE_EL0t);
 }
 
+void save_syscall_return_value(int num, unsigned long rc)
+{
+	/*
+	 * Save syscall return value.
+	 */
+	if (cpu_local_var(current) && cpu_local_var(current)->uctx &&
+			num != __NR_rt_sigsuspend) {
+		ihk_mc_syscall_arg0(cpu_local_var(current)->uctx) = rc;
+	}
+}
+
 void
 check_signal(unsigned long rc, void *regs0, int num)
 {
@@ -1346,16 +1358,6 @@ __check_signal(unsigned long rc, void *regs0, int num, int irq_disabled)
 		return;
 	thread = cpu_local_var(current);
 
-	/**
-	 * If check_signal is called from syscall(), 
-	 * then save syscall return value.
-	 */
-	if((regs == NULL)&&(num != __NR_rt_sigsuspend)){ /* It's call from syscall! */
-	 	// Get user context through current thread
-	 	// and update syscall return.
-	 	ihk_mc_syscall_arg0(thread->uctx) = rc;
-	}
-
 	if(thread == NULL || thread->proc->pid == 0){
 		struct thread *t;
 		irqstate = ihk_mc_spinlock_lock(&(cpu_local_var(runq_lock)));
@@ -2497,4 +2499,15 @@ out:
 	return mpsr->phase_ret;
 }
 
+time_t time(void) {
+	struct timespec ats;
+
+	if (gettime_local_support) {
+		calculate_time_from_tsc(&ats);
+		return ats.tv_sec;
+	}
+
+	return (time_t)0;
+}
+
 /*** End of File ***/

arch/x86_64/kernel/cpu.c

@@ -849,6 +849,7 @@ 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) {
@@ -870,6 +871,19 @@ void show_context_stack(uintptr_t *rbp) {
 	return;
 }
 
+void interrupt_exit(struct x86_user_context *regs)
+{
+	if (interrupt_from_user(regs)) {
+		cpu_enable_interrupt();
+		check_sig_pending();
+		check_need_resched();
+		check_signal(0, regs, 0);
+	}
+	else {
+		check_sig_pending();
+	}
+}
+
 void handle_interrupt(int vector, struct x86_user_context *regs)
 {
 	struct ihk_mc_interrupt_handler *h;
@@ -992,12 +1006,8 @@ void handle_interrupt(int vector, struct x86_user_context *regs)
 		}
 	}
 
-	if(interrupt_from_user(regs)){
-		cpu_enable_interrupt();
-		check_signal(0, regs, 0);
-		check_need_resched();
-	}
-	set_cputime(0);
+	interrupt_exit(regs);
+	set_cputime(interrupt_from_user(regs)? 0: 1);
 
 	--v->in_interrupt;
 }
@@ -1012,13 +1022,9 @@ void gpe_handler(struct x86_user_context *regs)
 		panic("gpe_handler");
 	}
 	set_signal(SIGSEGV, regs, NULL);
-	if(interrupt_from_user(regs)){
-		cpu_enable_interrupt();
-		check_signal(0, regs, 0);
-		check_need_resched();
-	}
-	set_cputime(0);
-	// panic("GPF");
+	interrupt_exit(regs);
+	set_cputime(interrupt_from_user(regs)? 0: 1);
+	panic("GPF");
 }
 
 void debug_handler(struct x86_user_context *regs)
@@ -1045,12 +1051,8 @@ void debug_handler(struct x86_user_context *regs)
 	memset(&info, '\0', sizeof info);
 	info.si_code = si_code;
 	set_signal(SIGTRAP, regs, &info);
-	if(interrupt_from_user(regs)){
-		cpu_enable_interrupt();
-		check_signal(0, regs, 0);
-		check_need_resched();
-	}
-	set_cputime(0);
+	interrupt_exit(regs);
+	set_cputime(interrupt_from_user(regs)? 0: 1);
 }
 
 void int3_handler(struct x86_user_context *regs)
@@ -1067,12 +1069,8 @@ void int3_handler(struct x86_user_context *regs)
 	memset(&info, '\0', sizeof info);
 	info.si_code = TRAP_BRKPT;
 	set_signal(SIGTRAP, regs, &info);
-	if(interrupt_from_user(regs)){
-		cpu_enable_interrupt();
-		check_signal(0, regs, 0);
-		check_need_resched();
-	}
-	set_cputime(0);
+	interrupt_exit(regs);
+	set_cputime(interrupt_from_user(regs)? 0: 1);
 }
 
 void
@@ -1582,6 +1580,8 @@ void arch_show_interrupt_context(const void *reg)
 	__kprintf("%16lx %16lx %16lx %16lx\n",
 	        regs->cs, regs->ss, regs->rflags, regs->error);
 
+kprintf_unlock(irqflags);
+return;
 	arch_show_extended_context();
 
 	arch_print_pre_interrupt_stack(regs);

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

@@ -64,7 +64,6 @@ static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
 	return oldval;
 }
 
-#ifdef POSTK_DEBUG_ARCH_DEP_8 /* arch depend hide */
 static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
@@ -128,7 +127,6 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	}
 	return ret;
 }
-#endif	/* !POSTK_DEBUG_ARCH_DEP_8 */
 
 static inline int get_futex_value_locked(uint32_t *dest, uint32_t *from)
 {

arch/x86_64/kernel/include/signal.h

@@ -18,6 +18,11 @@
 #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/x86_64/kernel/include/syscall_list.h

@@ -56,7 +56,7 @@ SYSCALL_HANDLED(36, getitimer)
 SYSCALL_HANDLED(38, setitimer)
 SYSCALL_HANDLED(39, getpid)
 SYSCALL_HANDLED(56, clone)
-SYSCALL_DELEGATED(57, fork)
+SYSCALL_HANDLED(57, fork)
 SYSCALL_HANDLED(58, vfork)
 SYSCALL_HANDLED(59, execve)
 SYSCALL_HANDLED(60, exit)

arch/x86_64/kernel/interrupt.S

@@ -145,6 +145,8 @@ nmi:
 		movq	%rsp,%gs:PANIC_REGS+0x08
 
 		movl	nmi_mode(%rip),%eax
+		cmp $3,%rax
+		je  4f
 		cmp	$1,%rax
 		je	1f
 		cmp	$2,%rax
@@ -199,9 +201,9 @@ nmi:
 		movl	%eax,%gs:PANIC_REGS+0xA0
 		movq	$1,%gs:PANICED
 		call	ihk_mc_query_mem_areas
-1:
+4:
 		hlt
-		jmp	1b
+		jmp	4b
 
 .globl x86_syscall
 x86_syscall:

arch/x86_64/kernel/memory.c

@@ -493,7 +493,7 @@ uint64_t ihk_mc_pt_virt_to_pagemap(struct page_table *pt, unsigned long virt)
 
 	error = ihk_mc_pt_virt_to_phys(pt, (void *)virt, &phys);
 	if (error) {
-		return 0;
+		return PM_PSHIFT(PAGE_SHIFT);
 	}
 
 	pagemap = PM_PFRAME(phys >> PAGE_SHIFT);
@@ -1542,7 +1542,6 @@ static int clear_range_l1(void *args0, pte_t *ptep, uint64_t base,
 				dkprintf("%lx-,%s: calling memory_stat_rss_sub(),phys=%lx,size=%ld,pgsize=%ld\n", pte_get_phys(&old), __FUNCTION__, pte_get_phys(&old), PTL1_SIZE, PTL1_SIZE);
 				rusage_memory_stat_sub(args->memobj, PTL1_SIZE, PTL1_SIZE); 
 			}
-			args->vm->currss -= PTL1_SIZE;
 		}  else {
 			dkprintf("%s: !calling memory_stat_rss_sub(),virt=%lx,phys=%lx\n", __FUNCTION__, base, pte_get_phys(&old));
 		}
@@ -1611,7 +1610,6 @@ static int clear_range_l2(void *args0, pte_t *ptep, uint64_t base,
 					dkprintf("%lx-,%s: calling memory_stat_rss_sub(),phys=%lx,size=%ld,pgsize=%ld\n", pte_get_phys(&old), __FUNCTION__, pte_get_phys(&old), PTL2_SIZE, PTL2_SIZE);
 					rusage_memory_stat_sub(args->memobj, PTL2_SIZE, PTL2_SIZE); 
 				}
-				args->vm->currss -= PTL2_SIZE;
 			}
 		}
 
@@ -1693,7 +1691,6 @@ static int clear_range_l3(void *args0, pte_t *ptep, uint64_t base,
 					dkprintf("%lx-,%s: calling memory_stat_rss_sub(),phys=%lx,size=%ld,pgsize=%ld\n", pte_get_phys(&old), __FUNCTION__, pte_get_phys(&old), PTL3_SIZE, PTL3_SIZE);
 					rusage_memory_stat_sub(args->memobj, PTL3_SIZE, PTL3_SIZE); 
 				}
-				args->vm->currss -= PTL3_SIZE;
 			}
 		}
 

arch/x86_64/kernel/syscall.c

@@ -30,8 +30,9 @@
 #include <ihk/ikc.h>
 #include <page.h>
 #include <limits.h>
+#include <syscall.h>
 
-void terminate(int, int);
+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);
@@ -142,8 +143,6 @@ SYSCALL_DECLARE(rt_sigaction)
 	struct k_sigaction new_sa, old_sa;
 	int rc;
 
-	if(sig == SIGKILL || sig == SIGSTOP || sig <= 0 || sig > 64)
-		return -EINVAL;
 	if (sigsetsize != sizeof(sigset_t))
 		return -EINVAL;
 
@@ -251,8 +250,8 @@ SYSCALL_DECLARE(rt_sigreturn)
 		regs->gpr.rflags &= ~RFLAGS_TF;
 		info.si_code = TRAP_TRACE;
 		set_signal(SIGTRAP, regs, &info);
-		check_signal(0, regs, 0);
 		check_need_resched();
+		check_signal(0, regs, 0);
 	}
 
 	if(ksigsp.fpregs && xsavesize){
@@ -279,6 +278,7 @@ 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;
 
 #define RFLAGS_MASK (RFLAGS_CF | RFLAGS_PF | RFLAGS_AF | RFLAGS_ZF | \
@@ -807,6 +807,11 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 		regs->gpr.rip = (unsigned long)k->sa.sa_handler;
 		regs->gpr.rsp = (unsigned long)usp;
 
+		// check signal handler is ONESHOT
+		if (k->sa.sa_flags & SA_RESETHAND) {
+			k->sa.sa_handler = SIG_DFL; 
+		}
+
 		if(!(k->sa.sa_flags & SA_NODEFER))
 			thread->sigmask.__val[0] |= pending->sigmask.__val[0];
 		kfree(pending);
@@ -818,8 +823,8 @@ do_signal(unsigned long rc, void *regs0, struct thread *thread, struct sig_pendi
 			regs->gpr.rflags &= ~RFLAGS_TF;
 			info.si_code = TRAP_TRACE;
 			set_signal(SIGTRAP, regs, &info);
-			check_signal(0, regs, 0);
 			check_need_resched();
+			check_signal(0, regs, 0);
 		}
 	}
 	else {
@@ -1006,6 +1011,12 @@ interrupt_from_user(void *regs0)
 	return !(regs->gpr.rsp & 0x8000000000000000);
 }
 
+void save_syscall_return_value(int num, unsigned long rc)
+{
+	/* Empty on x86 */
+	return;
+}
+
 void
 check_signal(unsigned long rc, void *regs0, int num)
 {
@@ -1053,6 +1064,110 @@ out:
 	return;
 }
 
+static int
+check_sig_pending_thread(struct thread *thread)
+{
+	int found = 0;
+	struct list_head *head;
+	mcs_rwlock_lock_t *lock;
+	struct mcs_rwlock_node_irqsave mcs_rw_node;
+	struct sig_pending *next;
+	struct sig_pending *pending;
+	__sigset_t w;
+	__sigset_t x;
+	int sig = 0;
+	struct k_sigaction *k;
+	struct cpu_local_var *v;
+
+	v = get_this_cpu_local_var();
+	w = thread->sigmask.__val[0];
+
+	lock = &thread->sigcommon->lock;
+	head = &thread->sigcommon->sigpending;
+	for (;;) {
+		mcs_rwlock_reader_lock(lock, &mcs_rw_node);
+
+		list_for_each_entry_safe(pending, next, head, list){
+			for (x = pending->sigmask.__val[0], sig = 0; x;
+			     sig++, x >>= 1);
+			k = thread->sigcommon->action + sig - 1;
+			if ((sig != SIGCHLD && sig != SIGURG) ||
+			    (k->sa.sa_handler != (void *)1 &&
+			     k->sa.sa_handler != NULL)) {
+				if (!(pending->sigmask.__val[0] & w)) {
+					if (pending->interrupted == 0) {
+						pending->interrupted = 1;
+						found = 1;
+						if (sig != SIGCHLD &&
+						    sig != SIGURG &&
+						    !k->sa.sa_handler) {
+							found = 2;
+							break;
+						}
+					}
+				}
+			}
+		}
+
+		mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
+
+		if (found == 2) {
+			break;
+		}
+
+		if (lock == &thread->sigpendinglock) {
+			break;
+		}
+
+		lock = &thread->sigpendinglock;
+		head = &thread->sigpending;
+	}
+
+	if (found == 2) {
+		ihk_mc_spinlock_unlock(&v->runq_lock, v->runq_irqstate);
+		terminate_mcexec(0, sig);
+		return 1;
+	}
+	else if (found == 1) {
+		ihk_mc_spinlock_unlock(&v->runq_lock, v->runq_irqstate);
+		interrupt_syscall(thread, 0);
+		return 1;
+	}
+	return 0;
+}
+
+void
+check_sig_pending()
+{
+	struct thread *thread;
+	struct cpu_local_var *v;
+
+	if (clv == NULL)
+		return;
+
+	v = get_this_cpu_local_var();
+repeat:
+	v->runq_irqstate = ihk_mc_spinlock_lock(&v->runq_lock);
+	list_for_each_entry(thread, &(v->runq), sched_list) {
+
+		if (thread == NULL || thread == &cpu_local_var(idle)) {
+			continue;
+		}
+
+		if (thread->in_syscall_offload == 0) {
+			continue;
+		}
+
+		if (thread->proc->exit_status & 0x0000000100000000L) {
+			continue;
+		}
+
+		if (check_sig_pending_thread(thread))
+			goto repeat;
+	}
+	ihk_mc_spinlock_unlock(&v->runq_lock, v->runq_irqstate);
+}
+
 unsigned long
 do_kill(struct thread *thread, int pid, int tid, int sig, siginfo_t *info,
         int ptracecont)
@@ -1214,15 +1329,19 @@ done:
 		mcs_rwlock_reader_lock_noirq(&tproc->update_lock, &updatelock);
 		savelock = &tthread->sigpendinglock;
 		head = &tthread->sigpending;
-		if(sig == SIGKILL ||
-		   (tproc->status != PS_EXITED &&
-		    tproc->status != PS_ZOMBIE &&
-		    tthread->status != PS_EXITED)){
-			hold_thread(tthread);
+		mcs_rwlock_reader_lock_noirq(&tproc->threads_lock, &lock);
+		if (tthread->status != PS_EXITED &&
+			(sig == SIGKILL ||
+			 (tproc->status != PS_EXITED && tproc->status != PS_ZOMBIE))) {
+			if ((rc = hold_thread(tthread))) {
+				kprintf("%s: ERROR hold_thread returned %d,tid=%d\n", __FUNCTION__, rc, tthread->tid);
+				tthread = NULL;
+			}
 		}
 		else{
 			tthread = NULL;
 		}
+		mcs_rwlock_reader_unlock_noirq(&tproc->threads_lock, &lock);
 		mcs_rwlock_reader_unlock_noirq(&tproc->update_lock, &updatelock);
 		mcs_rwlock_reader_unlock_noirq(&thash->lock[hash], &lock);
 	}
@@ -1249,7 +1368,9 @@ done:
 	}
 
 	if (tthread->thread_offloaded) {
-		interrupt_syscall(tthread, sig);
+		if (!tthread->proc->nohost) {
+			interrupt_syscall(tthread, sig);
+		}
 		release_thread(tthread);
 		return 0;
 	}
@@ -1284,6 +1405,7 @@ done:
 				rc = -ENOMEM;
 			}
 			else{
+				memset(pending, 0, sizeof(struct sig_pending));
 				pending->sigmask.__val[0] = mask;
 				memcpy(&pending->info, info, sizeof(siginfo_t));
 				pending->ptracecont = ptracecont;
@@ -1307,9 +1429,6 @@ done:
 			ihk_mc_interrupt_cpu(get_x86_cpu_local_variable(tthread->cpu_id)->apic_id, 0xd0);
 		}
 
-		if(!tthread->proc->nohost)
-			interrupt_syscall(tthread, 0);
-
 		if (status != PS_RUNNING) {
 			if(sig == SIGKILL){
 				/* Wake up the target only when stopped by ptrace-reporting */
@@ -1486,6 +1605,16 @@ SYSCALL_DECLARE(clone)
                    ihk_mc_syscall_sp(ctx));
 }
 
+SYSCALL_DECLARE(fork)
+{
+	return do_fork(SIGCHLD, 0, 0, 0, 0, ihk_mc_syscall_pc(ctx), ihk_mc_syscall_sp(ctx));
+}
+
+SYSCALL_DECLARE(vfork)
+{
+	return do_fork(CLONE_VFORK|SIGCHLD, 0, 0, 0, 0, ihk_mc_syscall_pc(ctx), ihk_mc_syscall_sp(ctx));
+}
+
 SYSCALL_DECLARE(shmget)
 {
 	const key_t key = ihk_mc_syscall_arg0(ctx);
@@ -2540,4 +2669,14 @@ out:
 	return mpsr->phase_ret;
 }
 
+time_t time(void) {
+	struct syscall_request sreq IHK_DMA_ALIGN;
+	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;
+}
+
 /*** End of File ***/

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

@@ -35,7 +35,6 @@ error_exit() {
 		;&
 	tmp_mcos_created)
 		if [ "$enable_mcoverlay" == "yes" ]; then
-		    umask $umask_old
 			rm -rf /tmp/mcos
 		fi
 		;&
@@ -48,10 +47,6 @@ error_exit() {
 }
 fi
 
-# Change umask for /proc and /sys files 
-umask_dec=$(( 8#${umask_old} & 8#0002 ))
-umask 0`printf "%o" ${umask_dec}`
-
 if [ ! -e /tmp/mcos ]; then 
     mkdir -p /tmp/mcos; 
 fi
@@ -149,7 +144,3 @@ for cpuid in `find /sys/bus/cpu/devices/* -maxdepth 0 -name "cpu[0123456789]*" -
 	rm -rf /tmp/mcos/mcos0_sys/bus/cpu/devices/$cpuid
     fi
 done
-
-# Restore umask
-umask ${umask_old}
-

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

@@ -19,6 +19,7 @@ ETCDIR=@ETCDIR@
 KMODDIR="${prefix}/kmod"
 KERNDIR="${prefix}/@TARGET@/kernel"
 ENABLE_MCOVERLAYFS="@ENABLE_MCOVERLAYFS@"
+MCK_BUILDID=@BUILDID@
 
 mem="512M@0"
 cpus=""
@@ -44,8 +45,10 @@ fi
 turbo=""
 ihk_irq=""
 umask_old=`umask`
+idle_halt=""
+allow_oversubscribe=""
 
-while getopts :tk:c:m:o:f:r:q:i:d:e: OPT
+while getopts :tk:c:m:o:f:r:q:i:d:e:hO OPT
 do
 	case ${OPT} in
 	f)	facility=${OPTARG}
@@ -70,6 +73,10 @@ do
 		;;
 	i) mon_interval=${OPTARG}
 		;;
+	h) idle_halt="idle_halt"
+		;;
+	O) allow_oversubscribe="allow_oversubscribe"
+		;;
 	*)  echo "invalid option -${OPT}" >&2
 		exit 1
 	esac
@@ -90,6 +97,18 @@ error_exit() {
 	local status=$1
 
 	case $status in
+	irqbalance_mck_started)
+		if [ "${irqbalance_used}" == "yes" ]; then
+		    if [ "`systemctl status irqbalance_mck.service 2> /dev/null |grep -E 'Active: active'`" != "" ]; then
+			if ! systemctl stop irqbalance_mck.service 2>/dev/null; then
+			    echo "warning: failed to stop irqbalance_mck" >&2
+			fi
+			if ! systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null; then
+			    echo "warning: failed to disable irqbalance_mck" >&2
+			fi
+		    fi
+		fi
+		;&
 	mcos_sys_mounted)
 		if [ "$enable_mcoverlay" == "yes" ]; then
 			umount /tmp/mcos/mcos0_sys
@@ -117,7 +136,6 @@ error_exit() {
 		;&
 	tmp_mcos_created)
 		if [ "$enable_mcoverlay" == "yes" ]; then
-		    umask $umask_old
 			rm -rf /tmp/mcos
 		fi
 		;&
@@ -157,20 +175,20 @@ error_exit() {
 	ihk_loaded)
 		rmmod ihk 2>/dev/null || echo "warning: failed to remove ihk" >&2
 		;&
+	smp_affinity_modified)
+		umask $umask_old
+		if [ "${irqbalance_used}" == "yes" ]; then
+		    if ! perl -e '$tmpdir="/tmp/mcreboot"; @files = grep { -f } glob "$tmpdir/proc/irq/*/smp_affinity"; foreach $file (@files) { $dest = substr($file, length($tmpdir)); if (0) {print "cp $file $dest\n";} system("cp $file $dest 2>/dev/null"); }'; then
+			echo "warning: failed to restore /proc/irq/*/smp_affinity" >&2
+		    fi
+		    if [ -e /tmp/mcreboot ]; then rm -rf /tmp/mcreboot; fi
+		fi
+		;&
 	irqbalance_stopped)
-		if [ "`systemctl status irqbalance_mck.service 2> /dev/null |grep -E 'Active: active'`" != "" ]; then
-			if ! systemctl stop irqbalance_mck.service 2>/dev/null; then
-				echo "warning: failed to stop irqbalance_mck" >&2
-			fi
-			if ! systemctl disable irqbalance_mck.service >/dev/null 2>/dev/null; then
-				echo "warning: failed to disable irqbalance_mck" >&2
-			fi
-			if ! etcdir=@ETCDIR@ perl -e '$etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "$etcdir/proc/irq/*/smp_affinity"; foreach $file (@files) { $dest = substr($file, length($etcdir)); if(0) {print "cp $file $dest\n";} system("cp $file $dest 2>/dev/null"); }'; then
-				echo "warning: failed to restore /proc/irq/*/smp_affinity" >&2
-			fi
-			if ! systemctl start irqbalance.service; then
-				echo "warning: failed to start irqbalance" >&2;
-			fi
+		if [ "${irqbalance_used}" == "yes" ]; then
+		    if ! systemctl start irqbalance.service; then
+			echo "warning: failed to start irqbalance" >&2;
+		    fi
 		fi
 		;&
 	initial)
@@ -240,9 +258,9 @@ if [ "${irqbalance_used}" == "yes" ]; then
 		exit 1
     fi;
 
-    if ! etcdir=@ETCDIR@ perl -e 'use File::Copy qw(copy); $etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "/proc/irq/*/smp_affinity"; foreach $file (@files) { $rel = substr($file, 1); $dir=substr($rel, 0, length($rel)-length("/smp_affinity")); if(0) { print "cp $file $etcdir/$rel\n";} if(system("mkdir -p $etcdir/$dir")){ exit 1;} if(!copy($file,"$etcdir/$rel")){ exit 1;} }'; then
+    if ! perl -e 'use File::Copy qw(copy); $tmpdir="/tmp/mcreboot"; @files = grep { -f } glob "/proc/irq/*/smp_affinity"; foreach $file (@files) { $rel = substr($file, 1); $dir = substr($rel, 0, length($rel) - length("/smp_affinity")); if (system("mkdir -p $tmpdir/$dir")) { exit 1; } if (0) { print "cp $file $tmpdir/$rel\n"; } if (!copy($file,"$tmpdir/$rel")) { exit 1; } }'; then
 		echo "error: saving /proc/irq/*/smp_affinity" >&2
-		error_exit "mcos_sys_mounted"
+		error_exit "irqbalance_stopped"
 	fi;
 
 # Prevent /proc/irq/*/smp_affinity from getting zero after offlining
@@ -256,16 +274,20 @@ if [ "${irqbalance_used}" == "yes" ]; then
 
     if ! ncpus=$ncpus smp_affinity_mask=$smp_affinity_mask perl -e '@dirs = grep { -d } glob "/proc/irq/*"; foreach $dir (@dirs) { $hit = 0; $affinity_str = `cat $dir/smp_affinity`; chomp $affinity_str; @int32strs = split /,/, $affinity_str; @int32strs_mask=split /,/, $ENV{'smp_affinity_mask'}; for($i=0;$i <= $#int32strs_mask; $i++) { $int32strs_inv[$i] = sprintf("%08x",hex($int32strs_mask[$i])^0xffffffff); if($i == 0) { $len = int((($ENV{'ncpus'}%32)+3)/4); if($len != 0) { $int32strs_inv[$i] = substr($int32strs_inv[$i], -$len, $len); } } } $inv = join(",", @int32strs_inv); $nint32s = int(($ENV{'ncpus'}+31)/32); for($j = $nint32s - 1; $j >= 0; $j--) { if(hex($int32strs[$nint32s - 1 - $j]) & hex($int32strs_mask[$nint32s - 1 - $j])) { $hit = 1; }} if($hit == 1) { $cmd = "echo $inv > $dir/smp_affinity 2>/dev/null"; system $cmd;}}'; then
 		echo "error: modifying /proc/irq/*/smp_affinity" >&2
-		error_exit "mcos_sys_mounted"
+		error_exit "irqbalance_stopped"
 	fi
-
 fi
 
+# Set umask so that proc/sys files/directories created by
+# mcctrl.ko and mcreboot.sh have appropriate permission bits
+umask_dec=$(( 8#${umask_old} & 8#0002 ))
+umask 0`printf "%o" ${umask_dec}`
+
 # Load IHK if not loaded
 if ! grep -E 'ihk\s' /proc/modules &>/dev/null; then
 	if ! taskset -c 0 insmod ${KMODDIR}/ihk.ko 2>/dev/null; then
 		echo "error: loading ihk" >&2
-		error_exit "irqbalance_stopped"
+		error_exit "smp_affinity_modified"
 	fi
 fi
 
@@ -299,7 +321,7 @@ if ! grep ihk_smp_@ARCH@ /proc/modules &>/dev/null; then
 		error_exit "ihk_loaded"
 	fi
 
-	# Offline-reonline RAM (special case for OFP SNC-4 mode)
+	# 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$i/memory*/ -name "online" | while read f; do
@@ -313,6 +335,22 @@ if ! grep ihk_smp_@ARCH@ /proc/modules &>/dev/null; then
 			find /sys/devices/system/node/node$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
@@ -337,6 +375,13 @@ if ! grep mcctrl /proc/modules &>/dev/null; then
 	fi
 fi
 
+# Check that different versions of binaries/scripts are not mixed
+IHK_BUILDID=`${SBINDIR}/ihkconfig 0 get buildid`
+if [  "${IHK_BUILDID}" != "${MCK_BUILDID}" ]; then
+	echo "IHK build-id (${IHK_BUILDID}) didn't match McKernel build-id (${MCK_BUILDID})." >&2
+	exit 1
+fi
+
 # Destroy all LWK instances
 if ls /dev/mcos* 1>/dev/null 2>&1; then
 	for i in /dev/mcos*; do
@@ -387,7 +432,7 @@ if ! ${SBINDIR}/ihkosctl 0 load ${KERNDIR}/mckernel.img; then
 fi
 
 # Set kernel arguments
-if ! ${SBINDIR}/ihkosctl 0 kargs "hidos $turbo dump_level=${DUMP_LEVEL} $extra_kopts"; then
+if ! ${SBINDIR}/ihkosctl 0 kargs "hidos $turbo $idle_halt dump_level=${DUMP_LEVEL} $extra_kopts $allow_oversubscribe"; then
 	echo "error: setting kernel arguments" >&2
 	error_exit "os_created"
 fi
@@ -426,4 +471,7 @@ if [ "${irqbalance_used}" == "yes" ]; then
 #    echo cpus=$cpus ncpus=$ncpus banirq=$banirq
 fi
 
+# Restore umask
+umask ${umask_old}
+
 exit 0

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

@@ -48,6 +48,9 @@ if ls /dev/mcos* 1>/dev/null 2>&1; then
 	done
 fi
 
+# Allow ihkmond to flush kmsg buffer
+sleep 2.0
+
 # Query IHK-SMP resources and release them
 if ! ${SBINDIR}/ihkconfig 0 query cpu > /dev/null; then
 	echo "error: querying cpus" >&2
@@ -62,17 +65,23 @@ if [ "${cpus}" != "" ]; then
 	fi
 fi
 
-if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then
-	echo "error: querying memory" >&2
-	exit 1
-fi
+#if ! ${SBINDIR}/ihkconfig 0 query mem > /dev/null; then
+#	echo "error: querying memory" >&2
+#	exit 1
+#fi
+#
+#mem=`${SBINDIR}/ihkconfig 0 query mem`
+#if [ "${mem}" != "" ]; then
+#	if ! ${SBINDIR}/ihkconfig 0 release mem $mem > /dev/null; then
+#		echo "error: releasing memory" >&2
+#		exit 1
+#	fi
+#fi
 
-mem=`${SBINDIR}/ihkconfig 0 query mem`
-if [ "${mem}" != "" ]; then
-	if ! ${SBINDIR}/ihkconfig 0 release mem $mem > /dev/null; then
-		echo "error: releasing memory" >&2
-		exit 1
-	fi
+# Release all memory
+if ! ${SBINDIR}/ihkconfig 0 release mem "all" > /dev/null; then
+	echo "error: releasing memory" >&2
+	exit 1
 fi
 
 # Remove delegator if loaded
@@ -110,9 +119,10 @@ fi
 
 # Start irqbalance with the original settings
 if [ "${irqbalance_used}" != "" ]; then
-	if ! etcdir=@ETCDIR@ perl -e '$etcdir=$ENV{'etcdir'}; @files = grep { -f } glob "$etcdir/proc/irq/*/smp_affinity"; foreach $file (@files) { $dest = substr($file, length($etcdir)); if(0) {print "cp $file $dest\n";} system("cp $file $dest 2>/dev/null"); }'; then
+	if ! perl -e '$tmpdir="/tmp/mcreboot"; @files = grep { -f } glob "$tmpdir/proc/irq/*/smp_affinity"; foreach $file (@files) { $dest = substr($file, length($tmpdir)); if (0) {print "cp $file $dest\n";} system("cp $file $dest 2>/dev/null"); }'; then
 		echo "warning: failed to restore /proc/irq/*/smp_affinity" >&2
 	fi
+	if [ -e /tmp/mcreboot ]; then rm -rf /tmp/mcreboot; fi
 	if ! systemctl start irqbalance.service; then
 		echo "warning: failed to start irqbalance" >&2;
 	fi

config.h.in

@@ -3,21 +3,24 @@
 /* Path of install directory for binary */
 #undef BINDIR
 
+/* IHK build-id to confirm IHK and McKernel built at the same time are used */
+#undef BUILDID
+
 /* whether mcoverlayfs is enabled */
 #undef ENABLE_MCOVERLAYFS
 
 /* whether memdump feature is enabled */
 #undef ENABLE_MEMDUMP
 
+/* whether perf is enabled */
+#undef ENABLE_PERF
+
 /* whether qlmpi is enabled */
 #undef ENABLE_QLMPI
 
 /* whether rusage is enabled */
 #undef ENABLE_RUSAGE
 
-/* whether perf is enabled */
-#undef ENABLE_PERF
-
 /* Define to 1 if you have the <inttypes.h> header file. */
 #undef HAVE_INTTYPES_H
 

configure

@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.69 for mckernel 0.9.0.
+# Generated by GNU Autoconf 2.69 for mckernel 1.5.0.
 #
 #
 # 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='0.9.0'
-PACKAGE_STRING='mckernel 0.9.0'
+PACKAGE_VERSION='1.5.0'
+PACKAGE_STRING='mckernel 1.5.0'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -645,6 +645,7 @@ TARGET
 UNAME_R
 KDIR
 ARCH
+BUILDID
 XCC
 FGREP
 EGREP
@@ -1261,7 +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 0.9.0 to adapt to many kinds of systems.
+\`configure' configures mckernel 1.5.0 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1322,7 +1323,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of mckernel 0.9.0:";;
+     short | recursive ) echo "Configuration of mckernel 1.5.0:";;
    esac
   cat <<\_ACEOF
 
@@ -1430,7 +1431,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-mckernel configure 0.9.0
+mckernel configure 1.5.0
 generated by GNU Autoconf 2.69
 
 Copyright (C) 2012 Free Software Foundation, Inc.
@@ -1728,7 +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 0.9.0, which was
+It was created by mckernel $as_me 1.5.0, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   $ $0 $@
@@ -2081,12 +2082,12 @@ ac_compiler_gnu=$ac_cv_c_compiler_gnu
 
 
 
-IHK_VERSION=0.9.0
-MCKERNEL_VERSION=0.9.0
-DCFA_VERSION=0.9.0
-IHK_RELEASE_DATE=2013-11-18
-MCKERNEL_RELEASE_DATE=2013-11-18
-DCFA_RELEASE_DATE=2013-11-18
+IHK_VERSION=1.5.0
+MCKERNEL_VERSION=1.5.0
+DCFA_VERSION=DCFA_VERSION_m4
+IHK_RELEASE_DATE=2018-04-05
+MCKERNEL_RELEASE_DATE=2018-04-05
+DCFA_RELEASE_DATE=DCFA_RELEASE_DATE_m4
 
 
 
@@ -5012,6 +5013,20 @@ cat >>confdefs.h <<_ACEOF
 _ACEOF
 
 
+ABS_SRCDIR=$( cd $( dirname $0 ); pwd )
+IHK_ABS_SRCDIR=${ABS_SRCDIR}/../ihk
+BUILDID=$( cd $IHK_ABS_SRCDIR; if [ ! -d .git ]; then echo $IHK_VERSION; else bash -c 'git rev-list -1 HEAD | cut -c1-8'; fi )
+{ $as_echo "$as_me:${as_lineno-$LINENO}: BUILDID=$BUILDID" >&5
+$as_echo "$as_me: BUILDID=$BUILDID" >&6;}
+if test "x$BUILDID" != "x" ; then
+
+cat >>confdefs.h <<_ACEOF
+#define BUILDID "$BUILDID"
+_ACEOF
+
+fi
+
+
 
 
 
@@ -5045,7 +5060,7 @@ 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/$ARCH/Makefile executer/kernel/mcoverlayfs/Makefile executer/kernel/mcoverlayfs/linux-3.10.0-327.36.1.el7/Makefile executer/kernel/mcoverlayfs/linux-4.0.9/Makefile executer/kernel/mcoverlayfs/linux-4.6.7/Makefile executer/include/qlmpilib.h kernel/Makefile kernel/Makefile.build kernel/include/swapfmt.h arch/x86_64/tools/mcreboot-attached-mic.sh arch/x86_64/tools/mcshutdown-attached-mic.sh arch/x86_64/tools/mcreboot-builtin-x86.sh arch/x86_64/tools/mcreboot-smp-x86.sh arch/x86_64/tools/mcstop+release-smp-x86.sh arch/x86_64/tools/mcoverlay-destroy-smp-x86.sh arch/x86_64/tools/mcoverlay-create-smp-x86.sh arch/x86_64/tools/eclair-dump-backtrace.exp arch/x86_64/tools/mcshutdown-builtin-x86.sh arch/x86_64/tools/mcreboot.1:arch/x86_64/tools/mcreboot.1in arch/x86_64/tools/irqbalance_mck.service arch/x86_64/tools/irqbalance_mck.in"
+ac_config_files="$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/Makefile"
 
 
 if test "$TARGET" = "smp-x86"; then
@@ -5570,7 +5585,7 @@ cat >>$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 0.9.0, which was
+This file was extended by mckernel $as_me 1.5.0, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -5632,7 +5647,7 @@ _ACEOF
 cat >>$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 0.9.0
+mckernel config.status 1.5.0
 configured by $0, generated by GNU Autoconf 2.69,
   with options \\"\$ac_cs_config\\"
 
@@ -5782,6 +5797,7 @@ do
     "arch/x86_64/tools/mcreboot.1") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/mcreboot.1:arch/x86_64/tools/mcreboot.1in" ;;
     "arch/x86_64/tools/irqbalance_mck.service") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/irqbalance_mck.service" ;;
     "arch/x86_64/tools/irqbalance_mck.in") CONFIG_FILES="$CONFIG_FILES arch/x86_64/tools/irqbalance_mck.in" ;;
+    "tools/mcstat/Makefile") CONFIG_FILES="$CONFIG_FILES tools/mcstat/Makefile" ;;
     "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

@@ -1,11 +1,9 @@
 # configure.ac COPYRIGHT FUJITSU LIMITED 2015-2016
 AC_PREREQ(2.63)
-m4_define([IHK_VERSION_m4],[0.9.0])dnl
-m4_define([MCKERNEL_VERSION_m4],[0.9.0])dnl
-m4_define([DCFA_VERSION_m4],[0.9.0])dnl
-m4_define([IHK_RELEASE_DATE_m4],[2013-11-18])dnl
-m4_define([MCKERNEL_RELEASE_DATE_m4],[2013-11-18])dnl
-m4_define([DCFA_RELEASE_DATE_m4],[2013-11-18])dnl
+m4_define([IHK_VERSION_m4],[1.5.0])dnl
+m4_define([MCKERNEL_VERSION_m4],[1.5.0])dnl
+m4_define([IHK_RELEASE_DATE_m4],[2018-04-05])dnl
+m4_define([MCKERNEL_RELEASE_DATE_m4],[2018-04-05])dnl
 
 AC_INIT([mckernel], MCKERNEL_VERSION_m4)
 
@@ -502,6 +500,15 @@ fi
 AC_DEFINE_UNQUOTED(BINDIR,"$BINDIR",[Path of install directory for binary])
 AC_DEFINE_UNQUOTED(SBINDIR,"$SBINDIR",[Path of install directory for system binary])
 
+ABS_SRCDIR=$( cd $( dirname $0 ); pwd )
+IHK_ABS_SRCDIR=${ABS_SRCDIR}/../ihk
+BUILDID=$( cd $IHK_ABS_SRCDIR; if @<:@ ! -d .git @:>@; then echo $IHK_VERSION; else bash -c 'git rev-list -1 HEAD | cut -c1-8'; fi )
+AC_MSG_NOTICE([BUILDID=$BUILDID])
+if test "x$BUILDID" != "x" ; then
+   AC_DEFINE_UNQUOTED(BUILDID,"$BUILDID",[IHK build-id to confirm IHK and McKernel built at the same time are used])
+fi
+AC_SUBST(BUILDID)
+
 AC_SUBST(CC)
 AC_SUBST(XCC)
 AC_SUBST(ARCH)
@@ -563,6 +570,7 @@ AC_CONFIG_FILES([
 	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/Makefile
 ])
 
 if test "$TARGET" = "smp-x86"; then

executer/kernel/mcctrl/Makefile.in

@@ -9,7 +9,6 @@ IHK_BASE=$(src)/../../../../ihk
 
 obj-m += mcctrl.o
 
-# POSTK_DEBUG_ARCH_DEP_1, arch depend "-mcmodel"
 # POSTK_DEBUG_ARCH_DEP_83, arch depend translate_rva_to_rpa() move
 ccflags-y := -I$(IHK_BASE)/linux/include \
              -I$(IHK_BASE)/linux/include/ihk/arch/$(ARCH) \
@@ -24,9 +23,8 @@ ccflags-y := -I$(IHK_BASE)/linux/include \
              -I$(src)/../../../kernel/include \
              -DMCEXEC_PATH=\"$(BINDIR)/mcexec\"
 
-ifneq ($(ARCH), arm64)
-ccflags-y += -mno-red-zone -mcmodel=kernel
-endif
+# depending arch
+include @abs_builddir@/arch/$(ARCH)/Makefile
 
 mcctrl-y := driver.o control.o ikc.o syscall.o procfs.o binfmt_mcexec.o
 mcctrl-y += sysfs.o sysfs_files.o arch/$(ARCH)/archdeps.o

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

@@ -1 +1 @@
-# dummy file
+ccflags-y += -mno-red-zone -mcmodel=kernel

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

@@ -327,6 +327,14 @@ int translate_rva_to_rpa(ihk_os_t os, unsigned long rpt, unsigned long rva,
 			pgsize = 1UL << offsh;
 			rpa = pt[ix] & ((1UL << 52) - 1) & ~(pgsize - 1);
 			rpa |= rva & (pgsize - 1);
+
+			/* For GB pages, just report regular 2MB page */
+			if (offsh == 30) {
+				pgsize = 1UL << 21;
+				dprintk("%s: GB page translated 0x%lx -> 0x%lx, pgsize: %lu\n",
+						__FUNCTION__, rva, rpa, pgsize);
+			}
+
 			ihk_device_unmap_virtual(ihk_os_to_dev(os), pt, PAGE_SIZE);
 			ihk_device_unmap_memory(ihk_os_to_dev(os), phys, PAGE_SIZE);
 			error = 0;

executer/kernel/mcctrl/control.c

@@ -367,7 +367,7 @@ static long mcexec_debug_log(ihk_os_t os, unsigned long arg)
 }
 
 int mcexec_close_exec(ihk_os_t os);
-int mcexec_destroy_per_process_data(ihk_os_t os);
+int mcexec_destroy_per_process_data(ihk_os_t os, int pid);
 
 static void release_handler(ihk_os_t os, void *param)
 {
@@ -387,7 +387,7 @@ static void release_handler(ihk_os_t os, void *param)
 
 	mcexec_close_exec(os);
 
-	mcexec_destroy_per_process_data(os);
+	mcexec_destroy_per_process_data(os, info->pid);
 
 	memset(&isp, '\0', sizeof isp);
 	isp.msg = SCD_MSG_CLEANUP_PROCESS;
@@ -435,6 +435,7 @@ static long mcexec_start_image(ihk_os_t os,
 	struct mcctrl_channel *c;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct mcos_handler_info *info;
+	int ret = 0;
 
 	desc = kmalloc(sizeof(*desc), GFP_KERNEL);
 	if (!desc) {
@@ -445,17 +446,18 @@ static long mcexec_start_image(ihk_os_t os,
 
 	if (copy_from_user(desc, udesc,
 	                   sizeof(struct program_load_desc))) {
-		kfree(desc);
-		return -EFAULT;
+		ret = -EFAULT;
+		goto out;
 	}
 
 	info = new_mcos_handler_info(os, file);
 #ifdef POSTK_DEBUG_TEMP_FIX_64 /* host process is SIGKILLed fix. */
 	if (info == NULL) {
-		kfree(desc);
-		return -ENOMEM;
+		ret = -ENOMEM;
+		goto out;
 	}
 #endif /* POSTK_DEBUG_TEMP_FIX_64 */
+
 	info->pid = desc->pid;
 	info->cpu = desc->cpu;
 	ihk_os_register_release_handler(file, release_handler, info);
@@ -471,10 +473,14 @@ static long mcexec_start_image(ihk_os_t os,
 	isp.ref = desc->cpu;
 	isp.arg = desc->rprocess;
 
-	mcctrl_ikc_send(os, desc->cpu, &isp);
+	ret = mcctrl_ikc_send(os, desc->cpu, &isp);
+	if (ret < 0) {
+		printk("%s: error: sending IKC msg\n", __FUNCTION__);
+	}
 
+out:
 	kfree(desc);
-	return 0;
+	return ret;
 }
 
 static DECLARE_WAIT_QUEUE_HEAD(signalq);
@@ -632,6 +638,7 @@ static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
 
 	pli->task = current;
 	pli->ready = 0;
+	pli->timeout = 0;
 	init_waitqueue_head(&pli->pli_wq);
 
 	pli_next = NULL;
@@ -693,11 +700,50 @@ static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
 		dprintk("%s: pid: %d, waiting in list\n",
 				__FUNCTION__, task_tgid_vnr(current));
 		mutex_unlock(&pe->lock);
-		ret = wait_event_interruptible(pli->pli_wq, pli->ready);
+		/* Timeout period: 10 secs + (#procs * 0.1sec) */
+		ret = wait_event_interruptible_timeout(pli->pli_wq,
+				pli->ready,
+				msecs_to_jiffies(10000 + req.nr_processes * 100));
 		mutex_lock(&pe->lock);
-		if (ret != 0) {
+
+		/* First timeout task? Wake up everyone else,
+		 * but tell them we timed out */
+		if (ret == 0) {
+			printk("%s: error: pid: %d, timed out, waking everyone\n",
+					__FUNCTION__, task_tgid_vnr(current));
+			while (!list_empty(&pe->pli_list)) {
+				pli_next = list_first_entry(&pe->pli_list,
+						struct process_list_item, list);
+				list_del(&pli_next->list);
+				pli_next->ready = 1;
+				pli_next->timeout = 1;
+				wake_up_interruptible(&pli_next->pli_wq);
+			}
+
+			/* Reset process counter to start state */
+			pe->nr_processes = -1;
+			ret = -ETIMEDOUT;
 			goto put_and_unlock_out;
 		}
+
+		/* Interrupted or woken up by someone else due to time out? */
+		if (ret < 0 || pli->timeout) {
+			if (ret > 0) {
+				printk("%s: error: pid: %d, job startup timed out\n",
+						__FUNCTION__, task_tgid_vnr(current));
+				ret = -ETIMEDOUT;
+			}
+			goto put_and_unlock_out;
+		}
+
+		/* Incorrect wakeup state? */
+		if (!pli->ready) {
+			printk("%s: error: pid: %d, not ready but woken?\n",
+					__FUNCTION__, task_tgid_vnr(current));
+			ret = -EINVAL;
+			goto put_and_unlock_out;
+		}
+
 		dprintk("%s: pid: %d, woken up\n",
 				__FUNCTION__, task_tgid_vnr(current));
 	}
@@ -1201,8 +1247,8 @@ retry_alloc:
 
 	wqhln->packet = packet;
 	wqhln->req = 1;
-	ihk_ikc_spinlock_unlock(&ppd->wq_list_lock, flags);
 	wake_up(&wqhln->wq_syscall);
+	ihk_ikc_spinlock_unlock(&ppd->wq_list_lock, flags);
 
 	mcctrl_put_per_proc_data(ppd);
 
@@ -1344,7 +1390,6 @@ retry_alloc:
 			goto put_ppd_out;
 		}
 
-#ifdef POSTK_DEBUG_ARCH_DEP_46 /* user area direct access fix. */
 		if (copy_to_user(&req->cpu, &packet->ref, sizeof(req->cpu))) {
 			if (mcctrl_delete_per_thread_data(ppd, current) < 0) {
 				kprintf("%s: error deleting per-thread data\n", __FUNCTION__);
@@ -1352,9 +1397,6 @@ retry_alloc:
 			ret = -EINVAL;
 			goto put_ppd_out;
 		}
-#else /* POSTK_DEBUG_ARCH_DEP_46 */
-		req->cpu = packet->ref;
-#endif /* POSTK_DEBUG_ARCH_DEP_46 */
 
 		ret = 0;
 		goto put_ppd_out;
@@ -1680,12 +1722,12 @@ int mcexec_create_per_process_data(ihk_os_t os)
 	return 0;
 }
 
-int mcexec_destroy_per_process_data(ihk_os_t os)
+int mcexec_destroy_per_process_data(ihk_os_t os, int pid)
 {
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct mcctrl_per_proc_data *ppd = NULL;
 
-	ppd = mcctrl_get_per_proc_data(usrdata, task_tgid_vnr(current));
+	ppd = mcctrl_get_per_proc_data(usrdata, pid);
 
 	if (ppd) {
 		/* One for the reference and one for deallocation.
@@ -2431,7 +2473,9 @@ mcexec_terminate_thread(ihk_os_t os, unsigned long *param, struct file *file)
 	mcctrl_delete_per_thread_data(ppd, tsk);
 	__return_syscall(usrdata->os, packet, param[2], tid);
 	ihk_ikc_release_packet((struct ihk_ikc_free_packet *)packet,
-	                       (usrdata->channels + packet->ref)->c);
+						   (usrdata->ikc2linux[smp_processor_id()] ?
+							usrdata->ikc2linux[smp_processor_id()] :
+							usrdata->ikc2linux[0]));
 err:
 	if(ppd)
 		mcctrl_put_per_proc_data(ppd);

executer/kernel/mcctrl/driver.c

@@ -27,6 +27,7 @@
 #include <linux/miscdevice.h>
 #include <linux/slab.h>
 #include <linux/device.h>
+#include <linux/delay.h>
 #include "mcctrl.h"
 #include <ihk/ihk_host_user.h>
 
@@ -169,6 +170,14 @@ error_cleanup_channels:
 int mcctrl_os_shutdown_notifier(int os_index)
 {
 	if (os[os_index]) {
+		/* Wait for os running */
+		if (ihk_os_wait_for_status(os[os_index], IHK_OS_STATUS_RUNNING, 0, 200) != 0) {
+			printk("IHK: OS does not become RUNNING in shutdown. Force shutdown.\n");
+			/* send nmi to force shutdown */
+			ihk_os_send_nmi(os[os_index], 3);
+			mdelay(200);
+		}
+
 		sysfsm_cleanup(os[os_index]);
 		free_topology_info(os[os_index]);
 		ihk_os_unregister_user_call_handlers(os[os_index], mcctrl_uc + os_index);

executer/kernel/mcctrl/mcctrl.h

@@ -304,6 +304,7 @@ struct node_topology {
 
 struct process_list_item {
 	int ready;
+	int timeout;
 	struct task_struct *task;
 	struct list_head list;
 	wait_queue_head_t pli_wq;

executer/kernel/mcctrl/procfs.c

@@ -1019,7 +1019,8 @@ static const struct procfs_entry tid_entry_stuff[] = {
 
 static const struct procfs_entry pid_entry_stuff[] = {
 	PROC_REG("auxv",       S_IRUSR, NULL),
-	PROC_REG("cgroup",     S_IXUSR, 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",    S_IRUGO, NULL),
 //	PROC_REG("comm",       S_IRUGO|S_IWUSR, NULL),

executer/kernel/mcctrl/syscall.c

@@ -222,6 +222,14 @@ int translate_rva_to_rpa(ihk_os_t os, unsigned long rpt, unsigned long rva,
 			pgsize = 1UL << offsh;
 			rpa = pt[ix] & ((1UL << 52) - 1) & ~(pgsize - 1);
 			rpa |= rva & (pgsize - 1);
+
+			/* For GB pages, just report regular 2MB page */
+			if (offsh == 30) {
+				pgsize = 1UL << 21;
+				dprintk("%s: GB page translated 0x%lx -> 0x%lx, pgsize: %lu\n",
+						__FUNCTION__, rva, rpa, pgsize);
+			}
+
 			ihk_device_unmap_virtual(ihk_os_to_dev(os), pt, PAGE_SIZE);
 			ihk_device_unmap_memory(ihk_os_to_dev(os), phys, PAGE_SIZE);
 			error = 0;
@@ -799,7 +807,7 @@ static int rus_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 
 	packet = (struct ikc_scd_packet *)mcctrl_get_per_thread_data(ppd, current);
 	if (!packet) {
-		error = -ENOENT;
+		ret = VM_FAULT_SIGBUS;
 		printk("%s: no packet registered for TID %d\n",
 				__FUNCTION__, task_pid_vnr(current));
 		goto put_and_out;

executer/user/Makefile.in

@@ -11,7 +11,7 @@ MCKERNEL_INCDIR=@MCKERNEL_INCDIR@
 MCKERNEL_LIBDIR=@MCKERNEL_LIBDIR@
 KDIR ?= @KDIR@
 ARCH=@ARCH@
-CFLAGS=-Wall -O -I. -I$(VPATH)/arch/${ARCH}
+CFLAGS=-Wall -O -I. -I$(VPATH)/arch/${ARCH} -I${IHKDIR}
 LDFLAGS=@LDFLAGS@
 RPATH=$(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@
@@ -40,7 +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 -I${IHKDIR} $(CFLAGS) -o $@ $^ $(LIBS)
+	$(CC) -I.. -I. -I./arch/$(ARCH)/include -I$(VPATH)/.. -I$(VPATH) -I$(VPATH)/arch/$(ARCH)/include $(CFLAGS) -o $@ $^ $(LIBS)
 else
 eclair: eclair.c
 	$(CC) $(CFLAGS) -I${IHKDIR} -o $@ $^ $(LIBS)

executer/user/arch/arm64/Makefile.in

@@ -9,12 +9,15 @@ LIBS=@LIBS@
 
 all: $(TARGET)
 
-../../libmcexec.a: archdep.o
-	$(AR) cr ../../libmcexec.a archdep.o 
+../../libmcexec.a: archdep.o arch_syscall.o
+	$(AR) cr ../../libmcexec.a archdep.o arch_syscall.o
 
 archdep.o: archdep.S
 	$(CC) -c -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -pthread $<
 
+arch_syscall.o: arch_syscall.c
+	$(CC) -c -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -pthread $<
+
 clean:
 	$(RM) $(TARGET) *.o
 

executer/user/arch/arm64/arch_syscall.c

@@ -0,0 +1,7 @@
+struct syscall_wait_desc;
+
+int
+archdep_syscall(struct syscall_wait_desc *w, long *ret)
+{
+	return -1;
+}

executer/user/arch/x86_64/Makefile.in

@@ -9,12 +9,15 @@ LIBS=@LIBS@
 
 all: $(TARGET)
 
-../../libmcexec.a: archdep.o
-	$(AR) cr ../../libmcexec.a archdep.o 
+../../libmcexec.a: archdep.o arch_syscall.o
+	$(AR) cr ../../libmcexec.a archdep.o arch_syscall.o
 
 archdep.o: archdep.S
 	$(CC) -c -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -pthread $<
 
+arch_syscall.o: arch_syscall.c
+	$(CC) -c -I${KDIR} $(CFLAGS) $(EXTRA_CFLAGS) -fPIE -pie -pthread $<
+
 clean:
 	$(RM) $(TARGET) *.o
 

executer/user/arch/x86_64/arch_syscall.c

@@ -0,0 +1,63 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <string.h>
+#include <sys/resource.h>
+#include <sys/syscall.h>
+#include <elf.h>
+#include <dirent.h>
+#include "../../../include/uprotocol.h"
+#include "../../archdep.h"
+
+//#define DEBUG
+#ifndef DEBUG
+#define __dprint(msg, ...)
+#define __dprintf(arg, ...)
+#define __eprint(msg, ...)
+#define __eprintf(format, ...)
+#else
+#define __dprint(msg, ...)  {printf("%s: " msg, __FUNCTION__);fflush(stdout);}
+#define __dprintf(format, ...)  {printf("%s: " format, __FUNCTION__, \
+					__VA_ARGS__);fflush(stdout);}
+#define __eprint(msg, ...)  {fprintf(stderr, "%s: " msg, __FUNCTION__);\
+					fflush(stderr);}
+#define __eprintf(format, ...)  {fprintf(stderr, "%s: " format, __FUNCTION__, \
+					__VA_ARGS__);fflush(stderr);}
+#endif
+
+extern char *chgpath(char *, char *);
+extern long do_strncpy_from_user(int, void *, void *, unsigned long);
+extern int fd;
+
+#define SET_ERR(ret) if (ret == -1) ret = -errno
+
+int
+archdep_syscall(struct syscall_wait_desc *w, long *ret)
+{
+	char *fn;
+	char pathbuf[PATH_MAX];
+	char tmpbuf[PATH_MAX];
+
+	switch (w->sr.number) {
+		case __NR_open:
+			*ret = do_strncpy_from_user(fd, pathbuf,
+			                       (void *)w->sr.args[0], PATH_MAX);
+			if (*ret >= PATH_MAX) {
+				*ret = -ENAMETOOLONG;
+			}
+			if (ret < 0) {
+				return 0;
+			}
+			__dprintf("open: %s\n", pathbuf);
+
+			fn = chgpath(pathbuf, tmpbuf);
+
+			*ret = open(fn, w->sr.args[1], w->sr.args[2]);
+			SET_ERR(*ret);
+			return 0;
+	}
+	return -1;
+}

executer/user/archdep.h

@@ -1,3 +1,4 @@
 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/mcexec.c

@@ -76,6 +76,7 @@
 #include <asm/prctl.h>
 #endif /* !POSTK_DEBUG_ARCH_DEP_77 */
 #include "../include/uprotocol.h"
+#include <ihk/ihk_host_user.h>
 #include <getopt.h>
 #include "archdep.h"
 #include "arch_args.h"
@@ -148,17 +149,25 @@ char **__glob_argv = 0;
 
 #ifdef ENABLE_MCOVERLAYFS
 #undef ENABLE_MCOVERLAYFS
-#ifndef RHEL_RELEASE_CODE
+
+// RedHat?
+#ifdef RHEL_RELEASE_CODE
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(3,10,0)
+#define ENABLE_MCOVERLAYFS 1
+#else
+#error "ERROR: your Linux kernel version on RHEL is not supported"
+#endif // LINUX_VERSION_CODE <= KERNEL_VERSION(3,10,0)
+
+// Other distro?
+#else
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,1,0)
 #define ENABLE_MCOVERLAYFS 1
 #endif // LINUX_VERSION_CODE == 4.0
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,6,0) && LINUX_VERSION_CODE < KERNEL_VERSION(4,7,0)
 #define ENABLE_MCOVERLAYFS 1
 #endif // LINUX_VERSION_CODE == 4.6
-#else
-#if RHEL_RELEASE_CODE <= RHEL_RELEASE_VERSION(7,3) 
-#define ENABLE_MCOVERLAYFS 1
-#endif // RHEL_RELEASE_CODE <= 7.3
+
 #endif // RHEL_RELEASE_CODE
 #endif // ENABLE_MCOVERLAYFS
 
@@ -198,7 +207,7 @@ struct thread_data_s;
 int main_loop(struct thread_data_s *);
 
 static int mcosid;
-static int fd;
+int fd;
 static char *exec_path = NULL;
 static char *altroot;
 static const char rlimit_stack_envname[] = "MCKERNEL_RLIMIT_STACK";
@@ -224,12 +233,13 @@ static int nr_processes = 0;
 static int nr_threads = -1;
 
 struct fork_sync {
-	pid_t pid;
 	int status;
+	volatile int success;
 	sem_t sem;
 };
 
 struct fork_sync_container {
+	pid_t pid;
 	struct fork_sync_container *next;
 	struct fork_sync *fs;
 };
@@ -1746,6 +1756,8 @@ static int
 opendev()
 {
 	int f;
+	char buildid[] = BUILDID;
+    char query_result[sizeof(BUILDID)];
 
 	sprintf(dev, "/dev/mcos%d", mcosid);
 
@@ -1757,6 +1769,18 @@ opendev()
 	}
 	fd = f;
 
+	if (ioctl(fd, IHK_OS_GET_BUILDID, query_result)) {
+		fprintf(stderr, "Error: IHK_OS_GET_BUILDID failed");
+		close(fd);
+		return -1;
+	}
+
+	if (strncmp(buildid, query_result, sizeof(buildid))) {
+		fprintf(stderr, "Error: build-id of mcexec (%s) didn't match that of IHK (%s)\n", buildid, query_result);
+		close(fd);
+		return -1;
+	}
+
 	return fd;
 }
 
@@ -1832,8 +1856,14 @@ int main(int argc, char **argv)
 		altroot = "/usr/linux-k1om-4.7/linux-k1om";
 	}
 
-	/* Disable address space layout randomization */
+	/* Disable READ_IMPLIES_EXEC */
 	persona = personality(0xffffffff);
+	if (persona & READ_IMPLIES_EXEC) {
+		persona &= ~READ_IMPLIES_EXEC;
+		persona = personality(persona);
+	}
+
+	/* Disable address space layout randomization */
 	__dprintf("persona=%08x\n", persona);
 	if ((persona & (PER_LINUX | ADDR_NO_RANDOMIZE)) == 0) {
 		CHKANDJUMP(getenv("MCEXEC_ADDR_NO_RANDOMIZE"), 1, "personality() and then execv() failed\n");
@@ -2510,20 +2540,11 @@ do_generic_syscall(
 
 	__dprintf("do_generic_syscall(%ld)\n", w->sr.number);
 
-#ifdef POSTK_DEBUG_TEMP_FIX_75 /* syscall return value check add. */
 	ret = syscall(w->sr.number, w->sr.args[0], w->sr.args[1], w->sr.args[2],
 		 w->sr.args[3], w->sr.args[4], w->sr.args[5]);
 	if (ret == -1) {
 		ret = -errno;
 	}
-#else /* POSTK_DEBUG_TEMP_FIX_75 */
-	errno = 0;
-	ret = syscall(w->sr.number, w->sr.args[0], w->sr.args[1], w->sr.args[2],
-		 w->sr.args[3], w->sr.args[4], w->sr.args[5]);
-	if (errno != 0) {
-		ret = -errno;
-	}
-#endif /* POSTK_DEBUG_TEMP_FIX_75 */
 
 	/* Overlayfs /sys/X directory lseek() problem work around */
 	if (w->sr.number == __NR_lseek && ret == -EINVAL) {
@@ -2985,7 +3006,7 @@ out:
 	return rc;
 }
 
-static long do_strncpy_from_user(int fd, void *dest, void *src, unsigned long n)
+long do_strncpy_from_user(int fd, void *dest, void *src, unsigned long n)
 {
 	struct strncpy_from_user_desc desc;
 	int ret;
@@ -3199,45 +3220,60 @@ int main_loop(struct thread_data_s *my_thread)
 		my_thread->remote_cpu = w.cpu;
 
 		switch (w.sr.number) {
-#ifdef POSTK_DEBUG_ARCH_DEP_13 /* arch depend hide */
-#ifdef __aarch64__
 		case __NR_openat:
 			/* initialize buffer */
 			memset(tmpbuf, '\0', sizeof(tmpbuf));
 			memset(pathbuf, '\0', sizeof(pathbuf));
 
 			/* check argument 1 dirfd */
-			if ((int)w.sr.args[0] != AT_FDCWD) {
+			ret = do_strncpy_from_user(fd, pathbuf,
+			                           (void *)w.sr.args[1],
+			                           PATH_MAX);
+			__dprintf("openat(dirfd == AT_FDCWD)\n");
+			if (ret >= PATH_MAX) {
+				ret = -ENAMETOOLONG;
+			}
+			if (ret < 0) {
+				do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+				break;
+			}
+
+			if ((int)w.sr.args[0] != AT_FDCWD &&
+			    pathbuf[0] != '/') {
 				/* dirfd != AT_FDCWD */
 				__dprintf("openat(dirfd != AT_FDCWD)\n");
-				snprintf(tmpbuf, sizeof(tmpbuf), "/proc/self/fd/%d", (int)w.sr.args[0]);
-				ret = readlink(tmpbuf, pathbuf, sizeof(pathbuf) - 1);
+				snprintf(tmpbuf, sizeof(tmpbuf),
+				         "/proc/self/fd/%d", (int)w.sr.args[0]);
+				ret = readlink(tmpbuf, pathbuf,
+				               sizeof(pathbuf) - 1);
+				if (ret == -1 &&
+				    (errno == ENOENT ||
+				     errno == EINVAL)) {
+					do_syscall_return(fd, cpu, -EBADF, 0, 0,
+					                  0, 0);
+					break;
+				}
 				if (ret < 0) {
-					do_syscall_return(fd, cpu, -errno, 0, 0, 0, 0);
+					do_syscall_return(fd, cpu, -errno, 0, 0,
+					                  0, 0);
 					break;
 				}
 				__dprintf("  %s -> %s\n", tmpbuf, pathbuf);
-				ret = do_strncpy_from_user(fd, tmpbuf, (void *)w.sr.args[1], PATH_MAX);
+				ret = do_strncpy_from_user(fd, tmpbuf,
+				                           (void *)w.sr.args[1],
+				                           PATH_MAX);
 				if (ret >= PATH_MAX) {
 					ret = -ENAMETOOLONG;
 				}
 				if (ret < 0) {
-					do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
+					do_syscall_return(fd, cpu, ret, 0, 0, 0,
+					                  0);
 					break;
 				}
 				strncat(pathbuf, "/", 1);
 				strncat(pathbuf, tmpbuf, strlen(tmpbuf) + 1);
-			} else {
-				/* dirfd == AT_FDCWD */
-				__dprintf("openat(dirfd == AT_FDCWD)\n");
-				ret = do_strncpy_from_user(fd, pathbuf, (void *)w.sr.args[1], PATH_MAX);
-				if (ret >= PATH_MAX) {
-					ret = -ENAMETOOLONG;
-				}
-				if (ret < 0) {
-					do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
-					break;
-				}
+			}
+			else {
 			}
 			__dprintf("openat: %s\n", pathbuf);
 
@@ -3247,44 +3283,6 @@ int main_loop(struct thread_data_s *my_thread)
 			SET_ERR(ret);
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
-#else /* __aarch64__ */
-		case __NR_open:
-			ret = do_strncpy_from_user(fd, pathbuf, (void *)w.sr.args[0], PATH_MAX);
-			if (ret >= PATH_MAX) {
-				ret = -ENAMETOOLONG;
-			}
-			if (ret < 0) {
-				do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
-				break;
-			}
-			__dprintf("open: %s\n", pathbuf);
-
-			fn = chgpath(pathbuf, tmpbuf);
-
-			ret = open(fn, w.sr.args[1], w.sr.args[2]);
-			SET_ERR(ret);
-			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
-			break;
-#endif /* __aarch64__ */
-#else /* POSTK_DEBUG_ARCH_DEP_13 */
-		case __NR_open:
-			ret = do_strncpy_from_user(fd, pathbuf, (void *)w.sr.args[0], PATH_MAX);
-			if (ret >= PATH_MAX) {
-				ret = -ENAMETOOLONG;
-			}
-			if (ret < 0) {
-				do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
-				break;
-			}
-			__dprintf("open: %s\n", pathbuf);
-
-			fn = chgpath(pathbuf, tmpbuf);
-
-			ret = open(fn, w.sr.args[1], w.sr.args[2]);
-			SET_ERR(ret);
-			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
-			break;
-#endif /* POSTK_DEBUG_ARCH_DEP_13 */
 
 		case __NR_futex:
 			ret = clock_gettime(w.sr.args[1], &tv);
@@ -3405,74 +3403,58 @@ gettid_out:
 			break;
 		}
 
-#ifdef POSTK_DEBUG_ARCH_DEP_13 /* arch depend hide */
-		case 1079: {
-#else /* POSTK_DEBUG_ARCH_DEP_13 */
-		case __NR_fork: {
-#endif /* POSTK_DEBUG_ARCH_DEP_13 */
+		case __NR_clone: {
 			struct fork_sync *fs;
-			struct fork_sync_container *fsc;
+			struct fork_sync_container *fsc = NULL;
 			struct fork_sync_container *fp;
 			struct fork_sync_container *fb;
 			int flag = w.sr.args[0];
 			int rc = -1;
 			pid_t pid;
 
+			if (flag == 1) {
+				pid = w.sr.args[1];
+				rc = 0;
+				pthread_mutex_lock(&fork_sync_mutex);
+				for (fp = fork_sync_top, fb = NULL; fp; fb = fp, fp = fp->next)
+					if (fp->pid == pid)
+						break;
+				if (fp) {
+					fs = fp->fs;
+					if (fb)
+						fb->next = fp->next;
+					else
+						fork_sync_top = fp->next;
+					fs->success = 1;
+					munmap(fs, sizeof(struct fork_sync));
+					free(fp);
+				}
+				pthread_mutex_unlock(&fork_sync_mutex);
+				do_syscall_return(fd, cpu, rc, 0, 0, 0, 0);
+				break;
+			}
+
+			fs = mmap(NULL, sizeof(struct fork_sync),
+			          PROT_READ | PROT_WRITE,
+			          MAP_SHARED | MAP_ANONYMOUS, -1, 0);
+			if (fs == (void *)-1) {
+				goto fork_err;
+			}
+			memset(fs, '\0', sizeof(struct fork_sync));
+			sem_init(&fs->sem, 1, 0);
+
 			fsc = malloc(sizeof(struct fork_sync_container));
+			if (!fsc) {
+				goto fork_err;
+			}
 			memset(fsc, '\0', sizeof(struct fork_sync_container));
 			pthread_mutex_lock(&fork_sync_mutex);
 			fsc->next = fork_sync_top;
 			fork_sync_top = fsc;
 			pthread_mutex_unlock(&fork_sync_mutex);
-			fsc->fs = fs = mmap(NULL, sizeof(struct fork_sync),
-			          PROT_READ | PROT_WRITE,
-			          MAP_SHARED | MAP_ANONYMOUS, -1, 0);
-			if(fs == (void *)-1){
-				goto fork_err;
-			}
+			fsc->fs = fs;
 
-			memset(fs, '\0', sizeof(struct fork_sync));
-			sem_init(&fs->sem, 1, 0);
-
-			if(flag){
-				int pipefds[2];
-
-				if(pipe(pipefds) == -1){
-					rc = -errno;
-					sem_destroy(&fs->sem);
-					goto fork_err;
-				}
-				pid = fork();
-				if(pid == 0){
-					close(pipefds[0]);
-					pid = fork();
-					if(pid != 0){
-						if (write(pipefds[1], &pid, sizeof pid) != sizeof(pid)) {
-							fprintf(stderr, "error: writing pipefds\n");
-						}
-						exit(0);
-					}
-				}
-				else if(pid != -1){
-					int npid;
-					int st;
-
-					close(pipefds[1]);
-					if (read(pipefds[0], &npid, sizeof npid) != sizeof(npid)) {
-						fprintf(stderr, "error: reading pipefds\n");
-					}
-					close(pipefds[0]);
-					waitpid(pid, &st, 0);
-					pid = npid;
-				}
-				else{
-					rc = -errno;
-					sem_destroy(&fs->sem);
-					goto fork_err;
-				}
-			}
-			else
-				pid = fork();
+			fsc->pid = pid = fork();
 
 			switch (pid) {
 			    /* Error */
@@ -3538,12 +3520,13 @@ gettid_out:
 
 fork_child_sync_pipe:
 				sem_post(&fs->sem);
+				sem_destroy(&fs->sem);
 				if (fs->status)
 					exit(1);
 
 				for (fp = fork_sync_top; fp;) {
 					fb = fp->next;
-					if (fp->fs)
+					if (fp->fs && fp->fs != fs)
 						munmap(fp->fs, sizeof(struct fork_sync));
 					free(fp);
 					fp = fb;
@@ -3555,6 +3538,16 @@ fork_child_sync_pipe:
 				ioctl(fd, MCEXEC_UP_NEW_PROCESS, &npdesc);
 
 				/* TODO: does the forked thread run in a pthread context? */
+				while (getppid() != 1 &&
+				       fs->success == 0) {
+					sched_yield();
+				}
+
+				if (fs->success == 0) {
+					exit(1);
+				}
+
+				munmap(fs, sizeof(struct fork_sync));
 				join_all_threads();
 
 				return ret;
@@ -3562,7 +3555,6 @@ fork_child_sync_pipe:
 				
 			    /* Parent */
 			    default:
-				fs->pid = pid;
 				while ((rc = sem_trywait(&fs->sem)) == -1 && (errno == EAGAIN || errno == EINTR)) {
 					int st;
 					int wrc;
@@ -3585,20 +3577,25 @@ fork_child_sync_pipe:
 				break;
 			}
 
-			sem_destroy(&fs->sem);
-			munmap(fs, sizeof(struct fork_sync));
 fork_err:
-			pthread_mutex_lock(&fork_sync_mutex);
-			for (fp = fork_sync_top, fb = NULL; fp; fb = fp, fp = fp->next)
-				if (fp == fsc)
-					break;
-			if (fp) {
-				if (fb)
-					fb->next = fsc->next;
-				else
-					fork_sync_top = fsc->next;
+			if (fs) {
+				sem_destroy(&fs->sem);
+				if (rc < 0) {
+					munmap(fs, sizeof(struct fork_sync));
+					pthread_mutex_lock(&fork_sync_mutex);
+					for (fp = fork_sync_top, fb = NULL; fp; fb = fp, fp = fp->next)
+						if (fp == fsc)
+							break;
+					if (fp) {
+						if (fb)
+							fb->next = fsc->next;
+						else
+							fork_sync_top = fsc->next;
+						free(fp);
+					}
+					pthread_mutex_unlock(&fork_sync_mutex);
+				}
 			}
-			pthread_mutex_unlock(&fork_sync_mutex);
 			do_syscall_return(fd, cpu, rc, 0, 0, 0, 0);
 			break;
 		}
@@ -4277,7 +4274,9 @@ return_linux_spawn:
 		}
 
 		default:
-			ret = do_generic_syscall(&w);
+			if (archdep_syscall(&w, &ret)) {
+				ret = do_generic_syscall(&w);
+			}
 			do_syscall_return(fd, cpu, ret, 0, 0, 0, 0);
 			break;
 

kernel/devobj.c

@@ -225,7 +225,7 @@ static void devobj_release(struct memobj *memobj)
 	return;
 }
 
-static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintptr_t *physp, unsigned long *flag)
+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_SIZE;
 	struct devobj *obj = to_devobj(memobj);

kernel/fileobj.c

@@ -191,7 +191,7 @@ static struct fileobj *obj_list_lookup(uintptr_t handle)
 /***********************************************************************
  * fileobj
  */
-int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
+int fileobj_create(int fd, struct memobj **objp, int *maxprotp, uintptr_t virt_addr)
 {
 	ihk_mc_user_context_t ctx;
 	struct pager_create_result result __attribute__((aligned(64)));	
@@ -265,7 +265,7 @@ int fileobj_create(int fd, struct memobj **objp, int *maxprotp)
 				/* Get the actual pages NUMA interleaved */
 				for (j = 0; j < nr_pages; ++j) {
 					mo->pages[j] = ihk_mc_alloc_aligned_pages_node_user(1,
-							PAGE_P2ALIGN, IHK_MC_AP_NOWAIT, node);
+							PAGE_P2ALIGN, IHK_MC_AP_NOWAIT, node, virt_addr);
 					if (!mo->pages[j]) {
 						kprintf("%s: ERROR: allocating pages[%d]\n",
 								__FUNCTION__, j);
@@ -558,7 +558,7 @@ out:
 }
 
 static int fileobj_get_page(struct memobj *memobj, off_t off,
-               int p2align, uintptr_t *physp, unsigned long *pflag)
+               int p2align, uintptr_t *physp, unsigned long *pflag, uintptr_t virt_addr)
 {
 	struct thread *proc = cpu_local_var(current);
 	struct fileobj *obj = to_fileobj(memobj);
@@ -571,7 +571,7 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 	struct mcs_rwlock_node mcs_node;
 	int hash = (off >> PAGE_SHIFT) & FILEOBJ_PAGE_HASH_MASK;	
 
-	dkprintf("fileobj_get_page(%p,%lx,%x,%p)\n", obj, off, p2align, physp);
+	dkprintf("fileobj_get_page(%p,%lx,%x,%x,%p)\n", obj, off, p2align, virt_addr, physp);
 	if (p2align != PAGE_P2ALIGN) {
 		return -ENOMEM;
 	}
@@ -584,13 +584,13 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 		int page_ind = off >> PAGE_SHIFT;
 
 		if (!memobj->pages[page_ind]) {
-			virt = ihk_mc_alloc_pages_user(1, IHK_MC_AP_NOWAIT | IHK_MC_AP_USER);
+			virt = ihk_mc_alloc_pages_user(1, IHK_MC_AP_NOWAIT | IHK_MC_AP_USER, virt_addr);
 
 			if (!virt) {
 				error = -ENOMEM;
-				kprintf("fileobj_get_page(%p,%lx,%x,%p):"
+				kprintf("fileobj_get_page(%p,%lx,%x,%x,%x,%p):"
 						"alloc failed. %d\n",
-						obj, off, p2align, physp,
+						obj, off, p2align, virt_addr, physp,
 						error);
 				goto out_nolock;
 			}
@@ -627,22 +627,22 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 		args = kmalloc(sizeof(*args), IHK_MC_AP_NOWAIT);
 		if (!args) {
 			error = -ENOMEM;
-			kprintf("fileobj_get_page(%p,%lx,%x,%p):"
+			kprintf("fileobj_get_page(%p,%lx,%x,%x,%p):"
 					"kmalloc failed. %d\n",
-					obj, off, p2align, physp, error);
+					obj, off, p2align, virt_addr, physp, error);
 			goto out;
 		}
 
 		if (!page) {
 			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 = ihk_mc_alloc_pages_user(npages, (IHK_MC_AP_NOWAIT |
+					(to_memobj(obj)->flags & MF_ZEROFILL) ? IHK_MC_AP_USER : 0), virt_addr);
 			if (!virt) {
 				error = -ENOMEM;
-				kprintf("fileobj_get_page(%p,%lx,%x,%p):"
+				kprintf("fileobj_get_page(%p,%lx,%x,%x,%p):"
 						"alloc failed. %d\n",
-						obj, off, p2align, physp,
+						obj, off, p2align, virt_addr, physp,
 						error);
 				goto out;
 			}
@@ -707,8 +707,8 @@ out_nolock:
 	if (args) {
 		kfree(args);
 	}
-	dkprintf("fileobj_get_page(%p,%lx,%x,%p): %d %lx\n",
-			obj, off, p2align, physp, error, phys);
+	dkprintf("fileobj_get_page(%p,%lx,%x,%x,%p): %d %lx\n",
+			obj, off, p2align, virt_addr, physp, error, phys);
 	return error;
 }
 

kernel/futex.c

@@ -673,9 +673,11 @@ static uint64_t futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q
 	xchg4(&(cpu_local_var(current)->status), PS_INTERRUPTIBLE);
 
 	/* Indicate spin sleep */
-	irqstate = ihk_mc_spinlock_lock(&thread->spin_sleep_lock);
-	thread->spin_sleep = 1;
-	ihk_mc_spinlock_unlock(&thread->spin_sleep_lock, irqstate);
+	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);
 	

kernel/host.c

@@ -79,7 +79,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 	unsigned long s, e, up;
 	char **argv;
 	char **a;
-	int i, n, argc, envc, args_envs_npages, l;
+	int i, n, argc, envc, args_envs_npages;
 	char **env;
 	int range_npages;
 	void *up_v;
@@ -148,7 +148,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 		}
 
 		if ((up_v = ihk_mc_alloc_pages_user(range_npages,
-						IHK_MC_AP_NOWAIT | ap_flags)) == NULL) {
+						IHK_MC_AP_NOWAIT | ap_flags, s)) == NULL) {
 			kprintf("ERROR: alloc pages for ELF section %i\n", i);
 			goto err;
 		}
@@ -259,7 +259,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 	e = addr + PAGE_SIZE * ARGENV_PAGE_COUNT;
 
 	if((args_envs = ihk_mc_alloc_pages_user(ARGENV_PAGE_COUNT,
-	                                        IHK_MC_AP_NOWAIT)) == NULL){
+	                                        IHK_MC_AP_NOWAIT, -1)) == NULL){
 		kprintf("ERROR: allocating pages for args/envs\n");
 		goto err;
 	}
@@ -349,21 +349,25 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 	// Update variables
 	argc = *((long *)(args_envs));
 	dkprintf("argc: %d\n", argc);
-
 	argv = (char **)(args_envs + (sizeof(long)));
-	if(proc->saved_cmdline){
+
+	if (proc->saved_cmdline) {
 		kfree(proc->saved_cmdline);
+		proc->saved_cmdline = NULL;
 		proc->saved_cmdline_len = 0;
 	}
-	for(a = argv, l = 0; *a; a++)
-		l += strlen(args_envs + (unsigned long)*a) + 1;
+
 	proc->saved_cmdline = kmalloc(p->args_len, IHK_MC_AP_NOWAIT);
-	if(!proc->saved_cmdline)
+	if (!proc->saved_cmdline) {
 		goto err;
-	proc->saved_cmdline_len = l;
-	for(a = argv, l = 0; *a; a++){
-		strcpy(proc->saved_cmdline + l, args_envs + (unsigned long)*a);
-		l += strlen(args_envs + (unsigned long)*a) + 1;
+	}
+
+	proc->saved_cmdline_len = p->args_len - ((argc + 1) * sizeof(char **));
+	memcpy(proc->saved_cmdline,
+			(char *)args_envs + ((argc + 1) * sizeof(char **)),
+			proc->saved_cmdline_len);
+
+	for (a = argv; *a; a++) {
 		*a = (char *)addr + (unsigned long)*a; // Process' address space!
 	}
 

kernel/include/cls.h

@@ -97,6 +97,8 @@ struct cpu_local_var {
 
 	ihk_spinlock_t smp_func_req_lock;
 	struct list_head smp_func_req_list;
+
+	struct process_vm *on_fork_vm;
 } __attribute__((aligned(64)));
 
 

kernel/include/futex.h

@@ -117,78 +117,9 @@
 #include <arch/system.h>
 #endif
 
-#ifdef POSTK_DEBUG_ARCH_DEP_8 /* arch depend hide */
-#else
-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 << 8) >> 20;
-	int cmparg = (encoded_op << 20) >> 20;
-	int oldval = 0, ret, tem;
-
-	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
-		oparg = 1 << oparg;
-
-#ifdef __UACCESS__
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
-		return -EFAULT;
-#endif
-
-	switch (op) {
-	case FUTEX_OP_SET:
-		__futex_atomic_op1("xchgl %0, %2", ret, oldval, uaddr, oparg);
-		break;
-	case FUTEX_OP_ADD:
-		__futex_atomic_op1("lock; xaddl %0, %2", ret, oldval,
-				   uaddr, oparg);
-		break;
-	case FUTEX_OP_OR:
-		__futex_atomic_op2("orl %4, %3", ret, oldval, uaddr, oparg);
-		break;
-	case FUTEX_OP_ANDN:
-		__futex_atomic_op2("andl %4, %3", ret, oldval, uaddr, ~oparg);
-		break;
-	case FUTEX_OP_XOR:
-		__futex_atomic_op2("xorl %4, %3", ret, oldval, uaddr, oparg);
-		break;
-	default:
-		ret = -ENOSYS;
-	}
-
-	if (!ret) {
-		switch (cmp) {
-		case FUTEX_OP_CMP_EQ:
-			ret = (oldval == cmparg);
-			break;
-		case FUTEX_OP_CMP_NE:
-			ret = (oldval != cmparg);
-			break;
-		case FUTEX_OP_CMP_LT:
-			ret = (oldval < cmparg);
-			break;
-		case FUTEX_OP_CMP_GE:
-			ret = (oldval >= cmparg);
-			break;
-		case FUTEX_OP_CMP_LE:
-			ret = (oldval <= cmparg);
-			break;
-		case FUTEX_OP_CMP_GT:
-			ret = (oldval > cmparg);
-			break;
-		default:
-			ret = -ENOSYS;
-		}
-	}
-	return ret;
-}
-
-#endif	/* arch depend hide */
 #endif // __KERNEL__
 #endif // _ASM_X86_FUTEX_H
 
-
-
 #define FUTEX_HASHBITS		8	/* 256 entries in each futex hash tbl */
 
 #define FUT_OFF_INODE    1 /* We set bit 0 if key has a reference on inode */

kernel/include/init.h

@@ -18,6 +18,7 @@ extern void mem_init(void);
 extern void ihk_ikc_master_init(void);
 extern void ap_init(void);
 extern void arch_ready(void);
+extern void done_init(void);
 extern void mc_ikc_test_init(void);
 extern void cpu_local_var_init(void);
 extern void kmalloc_init(void);

kernel/include/memobj.h

@@ -65,7 +65,7 @@ struct memobj {
 
 typedef void memobj_release_func_t(struct memobj *obj);
 typedef void memobj_ref_func_t(struct memobj *obj);
-typedef int memobj_get_page_func_t(struct memobj *obj, off_t off, int p2align, uintptr_t *physp, unsigned long *flag);
+typedef int memobj_get_page_func_t(struct memobj *obj, off_t off, int p2align, uintptr_t *physp, unsigned long *flag, uintptr_t virt_addr);
 typedef uintptr_t memobj_copy_page_func_t(struct memobj *obj, uintptr_t orgphys, int p2align);
 typedef int memobj_flush_page_func_t(struct memobj *obj, uintptr_t phys, size_t pgsize);
 typedef int memobj_invalidate_page_func_t(struct memobj *obj, uintptr_t phys, size_t pgsize);
@@ -96,10 +96,10 @@ static inline void memobj_ref(struct memobj *obj)
 }
 
 static inline int memobj_get_page(struct memobj *obj, off_t off,
-		int p2align, uintptr_t *physp, unsigned long *pflag)
+		int p2align, uintptr_t *physp, unsigned long *pflag, uintptr_t virt_addr)
 {
 	if (obj->ops->get_page) {
-		return (*obj->ops->get_page)(obj, off, p2align, physp, pflag);
+		return (*obj->ops->get_page)(obj, off, p2align, physp, pflag, virt_addr);
 	}
 	return -ENXIO;
 }
@@ -159,7 +159,7 @@ static inline int memobj_is_removable(struct memobj *obj)
 	return !!(obj->flags & MF_IS_REMOVABLE);
 }
 
-int fileobj_create(int fd, struct memobj **objp, int *maxprotp);
+int fileobj_create(int fd, struct memobj **objp, int *maxprotp, uintptr_t virt_addr);
 struct shmid_ds;
 int shmobj_create(struct shmid_ds *ds, struct memobj **objp);
 int zeroobj_create(struct memobj **objp);

kernel/include/process.h

@@ -277,6 +277,8 @@ struct resource_set {
 
 extern struct list_head	resource_set_list;
 extern mcs_rwlock_lock_t	resource_set_lock;
+extern int idle_halt;
+extern int allow_oversubscribe;
 
 struct process_hash {
 	struct list_head	list[HASH_SIZE];
@@ -390,7 +392,7 @@ struct vm_range {
 };
 
 struct vm_range_numa_policy {
-	struct list_head list;
+	struct rb_node policy_rb_node;
 	unsigned long start, end;
 	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
 	int numa_mem_policy;
@@ -419,6 +421,7 @@ struct mckfd {
 	long (*mmap_cb)(struct mckfd *, ihk_mc_user_context_t *);
 	int (*close_cb)(struct mckfd *, ihk_mc_user_context_t *);
 	int (*fcntl_cb)(struct mckfd *, ihk_mc_user_context_t *);
+	int (*dup_cb)(struct mckfd *, ihk_mc_user_context_t *);
 };
 
 #define SFD_CLOEXEC 02000000
@@ -436,6 +439,7 @@ struct sig_pending {
 	sigset_t sigmask;
 	siginfo_t info;
 	int ptracecont;
+	int interrupted;
 };
 
 typedef void pgio_func_t(void *arg);
@@ -486,7 +490,7 @@ struct process {
 			// V       +----   |
 			// PS_STOPPED -----+
 			// (PS_TRACED)
-	int exit_status; // only for zombie
+	unsigned long exit_status; // only for zombie
 
 	/* Store exit_status for a group of threads when stopped by SIGSTOP.
 	   exit_status can't be used because values of exit_status of threads
@@ -572,9 +576,6 @@ struct process {
 	int nr_processes; /* For partitioned execution */
 };
 
-void hold_thread(struct thread *ftn);
-void release_thread(struct thread *ftn);
-
 /*
  * Scheduling policies
  */
@@ -725,7 +726,7 @@ struct process_vm {
 	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
 	int numa_mem_policy;
 	/* Protected by memory_range_lock */
-	struct list_head vm_range_numa_policy_list;
+	struct rb_root vm_range_numa_policy_tree;
 	struct vm_range *range_cache[VM_RANGE_CACHE_SIZE];
 	int range_cache_ind;
 	struct swapinfo *swapinfo;
@@ -750,7 +751,7 @@ struct thread *create_thread(unsigned long user_pc,
 struct thread *clone_thread(struct thread *org, unsigned long pc,
                               unsigned long sp, int clone_flags);
 void destroy_thread(struct thread *thread);
-void hold_thread(struct thread *thread);
+int hold_thread(struct thread *thread);
 void release_thread(struct thread *thread);
 void flush_process_memory(struct process_vm *vm);
 void hold_process_vm(struct process_vm *vm);

kernel/include/rusage_private.h

@@ -35,6 +35,7 @@ rusage_rss_add(unsigned long size)
 	unsigned long newval;
 	unsigned long oldval;
 	unsigned long retval;
+	struct process_vm *vm;
 
 	newval = __sync_add_and_fetch(&rusage->rss_current, size);
 	oldval = rusage->memory_max_usage;
@@ -46,12 +47,28 @@ rusage_rss_add(unsigned long size)
 		}
 		oldval = retval;
 	}
+
+	/* process rss */
+	vm = cpu_local_var(on_fork_vm);
+	if (!vm) {
+		vm = cpu_local_var(current)->vm;
+	}
+
+	vm->currss += size;
+	if (vm->currss > vm->proc->maxrss) {
+		vm->proc->maxrss = vm->currss;
+	}
 }
 
 static inline void
 rusage_rss_sub(unsigned long size)
 {
+	struct process_vm *vm = cpu_local_var(current)->vm;
+
 	__sync_sub_and_fetch(&rusage->rss_current, size);
+
+	/* process rss */
+	vm->currss -= size;
 }
 
 static inline void memory_stat_rss_add(unsigned long size, int pgsize)

kernel/include/syscall.h

@@ -471,6 +471,8 @@ int arch_map_vdso(struct process_vm *vm);	/* arch dependent */
 int arch_setup_vdso(void);
 int arch_cpu_read_write_register(struct ihk_os_cpu_register *desc,
 		enum mcctrl_os_cpu_operation op);
+struct vm_range_numa_policy *vm_range_policy_search(struct process_vm *vm, uintptr_t addr);
+time_t time(void);
 
 #ifndef POSTK_DEBUG_ARCH_DEP_52
 #define VDSO_MAXPAGES 2

kernel/include/xpmem_private.h

@@ -252,6 +252,7 @@ extern struct xpmem_partition *xpmem_my_part;
 
 static int xpmem_ioctl(struct mckfd *mckfd, ihk_mc_user_context_t *ctx);
 static int xpmem_close(struct mckfd *mckfd, ihk_mc_user_context_t *ctx);
+static int xpmem_dup(struct mckfd *mckfd, ihk_mc_user_context_t *ctx);
 
 static int xpmem_init(void);
 static void xpmem_exit(void);
@@ -275,6 +276,7 @@ static int xpmem_release(xpmem_apid_t);
 static void xpmem_release_ap(struct xpmem_thread_group *,
 	struct xpmem_access_permit *);
 static void xpmem_release_aps_of_tg(struct xpmem_thread_group *ap_tg);
+static void xpmem_flush(struct mckfd *);
 
 static int xpmem_attach(struct mckfd *, xpmem_apid_t, off_t, size_t, 
 	unsigned long, int, int, unsigned long *);

kernel/init.c

@@ -144,6 +144,18 @@ static void parse_kargs(void)
 		}
 	}
 	ihk_mc_set_dump_level(dump_level);
+
+	/* idle_halt option */
+	ptr = find_command_line("idle_halt");
+	if (ptr) {
+		idle_halt = 1;
+	}
+
+	/* allow_oversubscribe option */
+	ptr = find_command_line("allow_oversubscribe");
+	if (ptr) {
+		allow_oversubscribe = 1;
+	}
 }
 
 extern void ihk_mc_get_boot_time(unsigned long *tv_sec, unsigned long *tv_nsec);
@@ -381,6 +393,7 @@ int main(void)
 
 	futex_init();
 
+	done_init();
 	kputs("IHK/McKernel booted.\n");
 
 #ifdef DCFA_KMOD

kernel/mem.c

@@ -76,7 +76,7 @@ static void *___kmalloc(int size, ihk_mc_ap_flag flag);
 static void ___kfree(void *ptr);
 
 static void *___ihk_mc_alloc_aligned_pages_node(int npages,
-		int p2align, ihk_mc_ap_flag flag, int node, int is_user);
+		int p2align, ihk_mc_ap_flag flag, int node, int is_user, uintptr_t virt_addr);
 static void *___ihk_mc_alloc_pages(int npages, ihk_mc_ap_flag flag, int is_user);
 static void ___ihk_mc_free_pages(void *p, int npages, int is_user);
 
@@ -193,7 +193,7 @@ struct pagealloc_track_entry *__pagealloc_track_find_entry(
 
 /* Top level routines called from macros */
 void *_ihk_mc_alloc_aligned_pages_node(int npages, int p2align,
-	ihk_mc_ap_flag flag, int node, int is_user,
+	ihk_mc_ap_flag flag, int node, int is_user, uintptr_t virt_addr,
 	char *file, int line)
 {
 	unsigned long irqflags;
@@ -201,7 +201,7 @@ void *_ihk_mc_alloc_aligned_pages_node(int npages, int p2align,
 	struct pagealloc_track_addr_entry *addr_entry;
 	int hash, addr_hash;
 	void *r = ___ihk_mc_alloc_aligned_pages_node(npages,
-					p2align, flag, node, is_user);
+					p2align, flag, node, is_user, virt_addr);
 
 	if (!memdebug || !pagealloc_track_initialized)
 		return r;
@@ -497,10 +497,10 @@ void pagealloc_memcheck(void)
 
 /* Actual allocation routines */
 static void *___ihk_mc_alloc_aligned_pages_node(int npages, int p2align,
-	ihk_mc_ap_flag flag, int node, int is_user)
+	ihk_mc_ap_flag flag, int node, int is_user, uintptr_t virt_addr)
 {
 	if (pa_ops)
-		return pa_ops->alloc_page(npages, p2align, flag, node, is_user);
+		return pa_ops->alloc_page(npages, p2align, flag, node, is_user, virt_addr);
 	else
 		return early_alloc_pages(npages);
 }
@@ -508,7 +508,7 @@ static void *___ihk_mc_alloc_aligned_pages_node(int npages, int p2align,
 static void *___ihk_mc_alloc_pages(int npages, ihk_mc_ap_flag flag,
 	int is_user)
 {
-	return ___ihk_mc_alloc_aligned_pages_node(npages, PAGE_P2ALIGN, flag, -1, is_user);
+	return ___ihk_mc_alloc_aligned_pages_node(npages, PAGE_P2ALIGN, flag, -1, is_user, -1);
 }
 
 static void ___ihk_mc_free_pages(void *p, int npages, int is_user)
@@ -544,7 +544,7 @@ static void reserve_pages(struct ihk_page_allocator_desc *pa_allocator,
 
 extern int cpu_local_var_initialized;
 static void *mckernel_allocate_aligned_pages_node(int npages, int p2align,
-		ihk_mc_ap_flag flag, int pref_node, int is_user)
+		ihk_mc_ap_flag flag, int pref_node, int is_user, uintptr_t virt_addr)
 {
 	unsigned long pa = 0;
 	int i, node;
@@ -553,6 +553,12 @@ static void *mckernel_allocate_aligned_pages_node(int npages, int p2align,
 #endif
 	int numa_id;
 
+	struct vm_range_numa_policy *range_policy_iter = NULL;
+	int numa_mem_policy = -1;
+	struct process_vm *vm;
+	struct vm_range *range = NULL;
+	int chk_shm = 0;
+
 	if(npages <= 0)
 		return NULL;
 
@@ -565,7 +571,23 @@ static void *mckernel_allocate_aligned_pages_node(int npages, int p2align,
 	/* No explicitly requested NUMA or user policy? */
 	if ((pref_node == -1) && (!(flag & IHK_MC_AP_USER) ||
 				cpu_local_var(current)->vm->numa_mem_policy == MPOL_DEFAULT)) {
-		goto distance_based;
+
+		if (virt_addr != -1) {
+			vm = cpu_local_var(current)->vm;
+			range_policy_iter = vm_range_policy_search(vm, virt_addr);
+			if (range_policy_iter) {
+				range = lookup_process_memory_range(vm, (uintptr_t)virt_addr, ((uintptr_t)virt_addr) + 1);
+				if (range) {
+					if( (range->memobj) && (range->memobj->flags == MF_SHM)) {
+						chk_shm = 1;
+					}
+				}
+			}
+		}
+
+
+		if ((!((range_policy_iter) && (range_policy_iter->numa_mem_policy != MPOL_DEFAULT))) && (chk_shm == 0))
+			goto distance_based;
 	}
 
 	node = ihk_mc_get_numa_id();
@@ -611,7 +633,28 @@ static void *mckernel_allocate_aligned_pages_node(int npages, int p2align,
 		}
 	}
 
-	switch (cpu_local_var(current)->vm->numa_mem_policy) {
+	if ((virt_addr != -1) && (chk_shm == 0)) {
+
+		vm = cpu_local_var(current)->vm;
+
+		if (!(range_policy_iter)) {
+			range_policy_iter = vm_range_policy_search(vm, virt_addr);
+		}
+
+		if (range_policy_iter) {
+			range = lookup_process_memory_range(vm, (uintptr_t)virt_addr, ((uintptr_t)virt_addr) + 1);
+			if ((range && (range->memobj->flags == MF_SHM))) {
+				chk_shm = 1;
+			} else {
+				numa_mem_policy = range_policy_iter->numa_mem_policy;
+			}
+		}
+	}
+
+	if (numa_mem_policy == -1)
+		numa_mem_policy = cpu_local_var(current)->vm->numa_mem_policy;
+
+	switch (numa_mem_policy) {
 		case MPOL_BIND:
 		case MPOL_PREFERRED:
 
@@ -1169,10 +1212,6 @@ static void page_fault_handler(void *fault_addr, uint64_t reason, void *regs)
 			info._sifields._sigfault.si_addr = fault_addr;
 			set_signal(SIGSEGV, regs, &info);
 		}
-		if(interrupt_from_user(regs)){
-			cpu_enable_interrupt();
-			check_signal(0, regs, 0);
-		}
 		goto out;
 	}
 
@@ -1181,8 +1220,12 @@ static void page_fault_handler(void *fault_addr, uint64_t reason, void *regs)
 out:
 	dkprintf("%s: addr: %p, reason: %lx, regs: %p -> error: %d\n",
 			__FUNCTION__, fault_addr, reason, regs, error);
-	check_need_resched();
-	set_cputime(0);
+	if(interrupt_from_user(regs)){
+		cpu_enable_interrupt();
+		check_need_resched();
+		check_signal(0, regs, 0);
+	}
+	set_cputime(interrupt_from_user(regs)? 0: 1);
 #ifdef PROFILE_ENABLE
 	if (thread->profile)
 		profile_event_add(PROFILE_page_fault, (rdtsc() - t_s));
@@ -1621,6 +1664,8 @@ void *ihk_mc_map_virtual(unsigned long phys, int npages,
 			ihk_pagealloc_free(vmap_allocator, virt_to_phys(p), npages);
 			return NULL;
 		}
+
+		flush_tlb_single((unsigned long)(p + (i << PAGE_SHIFT)));
 	}
 	barrier();
 	return (char *)p + offset;
@@ -1633,17 +1678,14 @@ void ihk_mc_unmap_virtual(void *va, int npages, int free_physical)
 	va = (void *)((unsigned long)va & PAGE_MASK);
 	for (i = 0; i < npages; i++) {
 		ihk_mc_pt_clear_page(NULL, (char *)va + (i << PAGE_SHIFT));
+		flush_tlb_single((unsigned long)(va + (i << PAGE_SHIFT)));
 	}
-#ifdef POSTK_DEBUG_TEMP_FIX_42 /* add unmap virtual tlb flush. */
-	flush_tlb();
-#endif /* POSTK_DEBUG_TEMP_FIX_42 */
-	
+
 #ifdef POSTK_DEBUG_TEMP_FIX_51 /* ihk_mc_unmap_virtual() free_physical disabled */
 	ihk_pagealloc_free(vmap_allocator, (unsigned long)va, npages);
 #else /* POSTK_DEBUG_TEMP_FIX_51 */
 	if (free_physical) {
 		ihk_pagealloc_free(vmap_allocator, (unsigned long)va, npages);
-		flush_tlb_single((unsigned long)va);
 	}
 #endif /* POSTK_DEBUG_TEMP_FIX_51 */
 }

kernel/pager.c

@@ -96,7 +96,7 @@ struct swapinfo {
 	struct arealist	swap_area;
 	struct arealist	mlock_area;
 	struct mlockcntnr mlock_container;
-#define UDATA_BUFSIZE	(8*1024)
+#define UDATA_BUFSIZE	PAGE_SIZE 
 	char	*swapfname;
 	char	*udata_buf;	/* To read-store data from Linux to user space */
 
@@ -295,7 +295,7 @@ static int
 pager_open(struct swapinfo *si, char *fname, int flag, int mode)
 {
 	int	fd;
-	strcpy(si->udata_buf, fname);
+	copy_to_user(si->udata_buf, fname, strlen(fname) + 1);
 	fd = linux_open(si->udata_buf, flag, mode);
 	return fd;
 }
@@ -303,22 +303,69 @@ pager_open(struct swapinfo *si, char *fname, int flag, int mode)
 static int
 pager_unlink(struct swapinfo *si, char *fname)
 {
-	strcpy(si->udata_buf, fname);
+	copy_to_user(si->udata_buf, fname, strlen(fname) + 1);
 	return linux_unlink(si->udata_buf);
 }
 
+static int 
+pager_copy_from_user(void * dst, void * from, size_t size, struct process_vm *vm)
+{
+	int ret;
+	void *virt;
+	unsigned long psize;
+	unsigned long rphys;
+	int faulted = 0;
+
+	if (size > PAGE_SIZE) {
+		ret = -EFAULT;
+		return ret;
+	}
+		
+retry_lookup:
+	/* remember page */
+	ret = ihk_mc_pt_virt_to_phys_size(vm->address_space->page_table,
+		dst, &rphys, &psize);
+
+	if (ret) {
+		uint64_t reason = PF_POPULATE | PF_WRITE | PF_USER;
+		void *addr= (void *)(((unsigned long)dst)& PAGE_MASK);
+
+		if (faulted) {
+			ret = -EFAULT;
+			return ret;
+		}
+
+		ret = page_fault_process_vm(vm, addr, reason);
+		if (ret) {
+			ret = -EFAULT;
+			return ret;
+		}
+
+		faulted = 1;
+		goto retry_lookup;
+	}
+
+	virt = phys_to_virt(rphys);
+
+	ret = copy_from_user(virt, from, size);
+
+	
+	return ret;
+}
+
 static ssize_t
-pager_read(struct swapinfo *si, int fd, void *start, size_t size)
+pager_read(struct swapinfo *si, int fd, void *start, size_t size,struct process_vm   *vm)
 {
 	ssize_t		off, sz, rs;
 
-	kprintf("pager_read: %lx (%lx)\n", start, size);
 	for (off = 0; off < size; off += sz) {
 		sz = size - off;
 		sz = (sz > UDATA_BUFSIZE) ? UDATA_BUFSIZE : sz;
 		rs = linux_read(fd, si->udata_buf, sz);
 		if (rs != sz) return rs;
-		copy_to_user(start + off, si->udata_buf, sz);
+		
+		rs = pager_copy_from_user(start + off, si->udata_buf, sz, vm); 
+		if (rs != 0) return rs;
 	}
 	return off;
 }
@@ -354,23 +401,26 @@ mlocklist_req(unsigned long start, unsigned long end, struct addrpair *addr, int
 static int
 mlocklist_morereq(struct swapinfo *si, unsigned long *start)
 {
-	struct areaent	*ent = si->mlock_area.tail;
+	struct areaent	went,*ent = si->mlock_area.tail;
+	copy_from_user(&went, ent, sizeof(struct areaent));
 
 	dkprintf("mlocklist_morereq: start = %ld and = %ld\n",
-		ent->pair[ent->count].start, ent->pair[ent->count].end);
-	if (ent->pair[ent->count].start != (unsigned long) -1) {
+		went.pair[went.count].start, went.pair[went.count].end);
+	if (went.pair[went.count].start != (unsigned long) -1) {
 		return 0;
 	}
-	*start = ent->pair[ent->count].end;
+	*start = went.pair[went.count].end;
 	return 1;
 }
 
 static int
 arealist_alloc(struct swapinfo *si, struct arealist *areap)
 {
+	struct areaent went;
 	areap->head = areap->tail = myalloc(si, sizeof(struct areaent));
 	if (areap->head == NULL) return -ENOMEM;
-	memset(areap->head, 0, sizeof(struct areaent));
+	memset(&went, 0, sizeof(struct areaent));
+	copy_to_user(areap->head, &went, sizeof(struct areaent));
 	return 0;
 }
 
@@ -402,7 +452,7 @@ arealist_free(struct arealist *area)
 static int
 arealist_get(struct swapinfo *si, struct addrpair **pair, struct arealist *area)
 {
-	struct areaent	*tmp;
+	struct areaent	*tmp,wtmp;
 	struct areaent	*tail = area->tail;
 	if (tail->count < MLOCKADDRS_SIZE - 1) { /* at least two entries are needed */
 		if (pair) *pair = &tail->pair[tail->count];
@@ -412,8 +462,10 @@ arealist_get(struct swapinfo *si, struct addrpair **pair, struct arealist *area)
 	if (tmp == NULL) {
 		return -1;
 	}
-	memset(tmp, 0, sizeof(struct areaent));
-	area->tail->next = tmp;
+	memset(&wtmp, 0, sizeof(struct areaent));
+	copy_to_user(tmp, &wtmp, sizeof(struct areaent));
+	copy_to_user(&(area->tail->next), &tmp, sizeof(struct areaent *));
+
 	area->tail = tmp;
 	if (pair) *pair = area->tail->pair;
 	return MLOCKADDRS_SIZE;
@@ -422,7 +474,10 @@ arealist_get(struct swapinfo *si, struct addrpair **pair, struct arealist *area)
 static void
 arealist_update(int cnt, struct arealist *area)
 {
-	area->tail->count += cnt;
+	int i;
+	copy_from_user(&i, &(area->tail->count), sizeof(int));
+	i += cnt;
+	copy_to_user(&(area->tail->count), &i, sizeof(int));
 	area->count += cnt;
 }
 
@@ -431,11 +486,13 @@ arealist_add(struct swapinfo *si, unsigned long start, unsigned long end,
              unsigned long flag, struct arealist *area)
 {
 	int	cc;
-	struct addrpair	*addr;
+	struct addrpair	*addr,waddr;
 
 	cc = arealist_get(si, &addr, area);
 	if (cc < 0) return -1;
-	addr->start = start; addr->end = end; addr->flag = flag;
+	waddr.start = start; waddr.end = end; waddr.flag = flag;
+	copy_to_user(addr, &waddr, sizeof(struct addrpair));
+	
 	arealist_update(1, area);
 	return 0;
 }
@@ -444,27 +501,31 @@ static int
 arealist_preparewrite(struct arealist *areap, struct swap_areainfo *info,
 		      ssize_t off, struct process_vm *vm, int flag)
 {
-	struct areaent		*ent;
+	struct areaent		*ent,went;
 	int			count = 0;
 	ssize_t			totsz = 0;
+	unsigned long pos;
 	struct page_table	*pt = vm->address_space->page_table;
 
 	for (ent = areap->head; ent != NULL; ent = ent->next) {
 		int i;
-		for (i = 0; i < ent->count; i++, count++) {
-			ssize_t sz = ent->pair[i].end - ent->pair[i].start; 
-			info[count].start = ent->pair[i].start;
-			info[count].end = ent->pair[i].end;
-			info[count].flag = ent->pair[i].flag;
+		copy_from_user(&went, ent, sizeof(struct areaent));
+		for (i = 0; i < went.count; i++, count++) {
+			ssize_t sz = went.pair[i].end - went.pair[i].start; 
+			copy_to_user(&(info[count].start), &(went.pair[i].start), sizeof(unsigned long));
+			copy_to_user(&(info[count].end), &(went.pair[i].end), sizeof(unsigned long));
+			copy_to_user(&(info[count].flag), &(went.pair[i].flag), sizeof(unsigned long));
 			if (flag) { /* position in file */
-				info[count].pos = off + totsz;
+				
+				pos = off + totsz;
 			} else { /* physical memory */
 				if (ihk_mc_pt_virt_to_phys(pt,
 						(void*) ent->pair[i].start,
-						 &info[count].pos)) {
+						 &pos)) {
 					kprintf("Cannot get phys\n");
 				}
 			}
+			copy_to_user(&(info[count].pos), &pos, sizeof(unsigned long));
 			totsz += sz;
 		}
 	}
@@ -489,6 +550,7 @@ arealist_print(char *msg, struct arealist *areap, int count)
 	for (ent = areap->head; ent != NULL; ent = ent->next) {
 		int i;
 		for (i = 0; i < ent->count; i++) {
+			
 			kprintf("\t%p -- %p\n",
 				(void*) ent->pair[i].start, (void*) ent->pair[i].end);
 		}
@@ -608,13 +670,13 @@ do_pagein(int flag)
 		extern int ihk_mc_pt_print_pte(struct page_table *pt, void *virt);
 		sz = si->swap_info[i].end - si->swap_info[i].start;
 		dkprintf("pagein: %016lx:%016lx sz(%lx)\n", si->swap_info[i].start, si->swap_info[i].end, sz);
-		rs = pager_read(si, fd, (void*) si->swap_info[i].start, sz);
+		rs = pager_read(si, fd, (void*) si->swap_info[i].start, sz, vm);
 		if (rs != sz) goto err;
 		// ihk_mc_pt_print_pte(vm->address_space->page_table, (void*) si->swap_info[i].start);
 	}
 	linux_close(fd);
 	print_region("after pagin", vm);
-	kprintf("do_pagein: done, currss(%lx)\n", vm->currss);
+	dkprintf("do_pagein: done, currss(%lx)\n", vm->currss);
 	vm->swapinfo = NULL;
 	kfree(si->swapfname);
 	kfree(si);
@@ -679,7 +741,8 @@ do_pageout(char *fname, void *buf, size_t size, int flag)
 		goto err;
 	}
 
-	fd = linux_open(fname, O_RDWR|O_CREAT|O_TRUNC, 0600);
+	copy_to_user(si->udata_buf, si->swapfname, strlen(si->swapfname) + 1);
+	fd = linux_open(si->udata_buf, O_RDWR|O_CREAT|O_TRUNC, 0600);
 	if (fd < 0) {
 		ekprintf("do_pageout: Cannot open/create file: %s\n", fname);
 		cc = fd;
@@ -752,10 +815,10 @@ do_pageout(char *fname, void *buf, size_t size, int flag)
 
 	/* preparing page store */
 	si->swphdr = myalloc(si, sizeof(struct swap_header));
-	strncpy(si->swphdr->magic, MCKERNEL_SWAP, SWAP_HLEN);
-	strncpy(si->swphdr->version, MCKERNEL_SWAP_VERSION, SWAP_HLEN);
-	si->swphdr->count_sarea = si->swap_area.count;
-	si->swphdr->count_marea = si->mlock_area.count;
+	copy_to_user(&(si->swphdr->magic), MCKERNEL_SWAP, SWAP_HLEN);
+	copy_to_user(&(si->swphdr->version), MCKERNEL_SWAP_VERSION, SWAP_HLEN);
+	copy_to_user(&(si->swphdr->count_sarea), &(si->swap_area.count), sizeof(unsigned int));
+	copy_to_user(&(si->swphdr->count_marea), &(si->mlock_area.count), sizeof(unsigned int));
 	if ((cc = pager_write(fd, si->swphdr, sizeof(struct swap_header)))
 	    != sizeof(struct swap_header)) {
 		if (cc >= 0)
@@ -779,8 +842,10 @@ do_pageout(char *fname, void *buf, size_t size, int flag)
 	if ((cc = arealist_write(fd, si->mlock_info, si->mlock_area.count)) < 0) goto err;
 	/* now pages are stored */
 	for (i = 0; i < si->swap_area.count; i++) {
-		sz = si->swap_info[i].end - si->swap_info[i].start;
-		if ((cc = pager_write(fd, (void*) si->swap_info[i].start, sz)) != sz) {
+		struct swap_areainfo sw_info;
+		copy_from_user(&sw_info, &(si->swap_info[i]), sizeof(struct swap_areainfo));
+		sz = sw_info.end - sw_info.start;
+		if ((cc = pager_write(fd, (void*) sw_info.start, sz)) != sz) {
 			if (cc >= 0)
 				cc = -EIO;
 			goto err;
@@ -792,10 +857,12 @@ do_pageout(char *fname, void *buf, size_t size, int flag)
 	}
 	kprintf("removing physical memory\n");
 	for (i = 0; i < si->swap_area.count; i++) {
+		struct swap_areainfo sw_info;
+		copy_from_user(&sw_info, &(si->swap_info[i]), sizeof(struct swap_areainfo));
 		cc = ihk_mc_pt_free_range(vm->address_space->page_table,
 					  vm,
-					  (void*) si->swap_info[i].start,
-					  (void*) si->swap_info[i].end, NULL);
+					  (void*) sw_info.start,
+					  (void*) sw_info.end, NULL);
 		if (cc < 0) {
 			kprintf("ihk_mc_pt_clear_range returns: %d\n", cc);
 		}
@@ -811,8 +878,10 @@ do_pageout(char *fname, void *buf, size_t size, int flag)
 	 * except TEXT, STACK, readonly pages, are not invalid.
 	 */
 	for (i = 0; i < si->swap_area.count; i++) {
-		sz = si->swap_info[i].end - si->swap_info[i].start;
-		cc = linux_munmap((void*) si->swap_info[i].start, sz, 0);
+		struct swap_areainfo sw_info;
+		copy_from_user(&sw_info, &(si->swap_info[i]), sizeof(struct swap_areainfo));
+		sz = sw_info.end - sw_info.start;
+		cc = linux_munmap((void*) sw_info.start, sz, 0);
 		if (cc < 0) {
 			kprintf("do_pageout: Cannot munmap: %lx len(%lx)\n",
 				si->swap_info[i].start, sz);

kernel/process.c

@@ -100,6 +100,9 @@ extern void perf_reset(struct mc_perf_event *event);
 struct list_head resource_set_list;
 mcs_rwlock_lock_t    resource_set_lock;
 
+int idle_halt = 0;
+int allow_oversubscribe = 0;
+
 void
 init_process(struct process *proc, struct process *parent)
 {
@@ -256,7 +259,7 @@ init_process_vm(struct process *owner, struct address_space *asp, struct process
 
 	ihk_atomic_set(&vm->refcount, 1);
 	vm->vm_range_tree = RB_ROOT;
-	INIT_LIST_HEAD(&vm->vm_range_numa_policy_list);
+	vm->vm_range_numa_policy_tree = RB_ROOT;
 	vm->address_space = asp;
 	vm->proc = owner;
 	vm->exiting = 0;
@@ -395,6 +398,7 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 	int termsig = clone_flags & 0xff;
 	struct process *proc = NULL;
 	struct address_space *asp = NULL;
+	struct cpu_local_var *v = get_this_cpu_local_var();
 
 	if ((thread = ihk_mc_alloc_pages(KERNEL_STACK_NR_PAGES,
 					IHK_MC_AP_NOWAIT)) == NULL) {
@@ -405,6 +409,9 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 	ihk_atomic_set(&thread->refcount, 2);
 	memcpy(&thread->cpu_set, &org->cpu_set, sizeof(thread->cpu_set));
 
+	/* New thread is in kernel until jumping to enter_user_mode */
+	thread->in_kernel = org->in_kernel;
+
 	/* NOTE: sp is the user mode stack! */
 	ihk_mc_init_user_process(&thread->ctx, &thread->uctx, ((char *)thread) +
 				 KERNEL_STACK_NR_PAGES * PAGE_SIZE, pc, sp);
@@ -478,12 +485,15 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 		dkprintf("fork(): copy_user_ranges()\n");
 		/* Copy user-space mappings.
 		 * TODO: do this with COW later? */
+		v->on_fork_vm = proc->vm;
 		if (copy_user_ranges(proc->vm, org->vm) != 0) {
 			release_address_space(asp);
+			v->on_fork_vm = NULL;
 			kfree(proc->vm);
 			kfree(proc);
 			goto err_free_proc;
 		}
+		v->on_fork_vm = NULL;
 
 		/* Copy mckfd list
 		   FIXME: Replace list manipulation with list_add() etc. */
@@ -507,13 +517,15 @@ clone_thread(struct thread *org, unsigned long pc, unsigned long sp,
 				mckfd->next = proc->mckfd;
 				proc->mckfd = mckfd;
 			}
+
+			if (mckfd->dup_cb) {
+				mckfd->dup_cb(mckfd, NULL);
+			}
 		}
 		ihk_mc_spinlock_unlock(&proc->mckfd_lock, irqstate);
 
 		thread->vm->vdso_addr = org->vm->vdso_addr;
 		thread->vm->vvar_addr = org->vm->vvar_addr;
-		thread->proc->maxrss = org->proc->maxrss;
-		thread->vm->currss = org->vm->currss;
 
 		thread->sigstack.ss_sp = org->sigstack.ss_sp;
 		thread->sigstack.ss_flags = org->sigstack.ss_flags;
@@ -646,7 +658,7 @@ static int copy_user_pte(void *arg0, page_table_t src_pt, pte_t *src_ptep, void
 
 		npages = pgsize / PAGE_SIZE;
 		virt = ihk_mc_alloc_aligned_pages_user(npages, pgalign,
-		                                       IHK_MC_AP_NOWAIT);
+		                                       IHK_MC_AP_NOWAIT, (uintptr_t)pgaddr);
 		if (!virt) {
 			kprintf("ERROR: copy_user_pte() allocating new page\n");
 			error = -ENOMEM;
@@ -1466,14 +1478,6 @@ static int remap_one_page(void *arg0, page_table_t pt, pte_t *ptep,
 	dkprintf("remap_one_page(%p,%p,%p %#lx,%p,%d)\n",
 			arg0, pt, ptep, *ptep, pgaddr, pgshift);
 
-	/* XXX: NYI: large pages */
-	if (pgsize != PAGE_SIZE) {
-		error = -E2BIG;
-		ekprintf("remap_one_page(%p,%p,%p %#lx,%p,%d):%d\n",
-				arg0, pt, ptep, *ptep, pgaddr, pgshift, error);
-		goto out;
-	}
-
 	off = args->off + ((uintptr_t)pgaddr - args->start);
 	pte_make_fileoff(off, 0, pgsize, &apte);
 
@@ -1509,6 +1513,7 @@ int remap_process_memory_range(struct process_vm *vm, struct vm_range *range,
 {
 	struct rfp_args args;
 	int error;
+	unsigned int retval;
 
 	dkprintf("remap_process_memory_range(%p,%p,%#lx,%#lx,%#lx)\n",
 			vm, range, start, end, off);
@@ -1519,6 +1524,13 @@ int remap_process_memory_range(struct process_vm *vm, struct vm_range *range,
 	args.off = off;
 	args.memobj = range->memobj;
 
+	retval = __sync_val_compare_and_swap(&range->pgshift, 0, PAGE_SHIFT);
+	if (retval != 0 && retval != PAGE_SHIFT) {
+		error = -E2BIG;
+		ekprintf("%s: pgshift is too big (%d)  failed:%d\n", __func__, retval, error);
+		goto out;
+	}
+		
 	error = visit_pte_range(vm->address_space->page_table, (void *)start,
 			(void *)end, range->pgshift, VPTEF_DEFAULT,
 			&remap_one_page, &args);
@@ -1765,7 +1777,7 @@ static int page_fault_process_memory_range(struct process_vm *vm, struct vm_rang
 				off = pte_get_off(ptep, pgsize);
 			}
 			error = memobj_get_page(range->memobj, off, p2align,
-                                       &phys, &memobj_flag);
+                                       &phys, &memobj_flag, fault_addr);
 			if (error) {
 				struct memobj *obj;
 
@@ -1787,7 +1799,7 @@ retry:
 			npages = pgsize / PAGE_SIZE;
 			virt = ihk_mc_alloc_aligned_pages_user(npages, p2align,
 					IHK_MC_AP_NOWAIT |
-					(range->flag & VR_AP_USER) ? IHK_MC_AP_USER : 0);
+					((range->flag & VR_AP_USER) ? IHK_MC_AP_USER : 0), fault_addr);
 			if (!virt && !range->pgshift && (pgsize != PAGE_SIZE)) {
 				error = arch_get_smaller_page_size(NULL, pgsize, &pgsize, &p2align);
 				if (error) {
@@ -1847,7 +1859,7 @@ retry:
 
 			npages = pgsize / PAGE_SIZE;
 			virt = ihk_mc_alloc_aligned_pages_user(npages, p2align,
-			                                      IHK_MC_AP_NOWAIT);
+			                                      IHK_MC_AP_NOWAIT, fault_addr);
 			if (!virt) {
 				error = -ENOMEM;
 				kprintf("page_fault_process_memory_range(%p,%lx-%lx %lx,%lx,%lx):cannot allocate copy page. %d\n", vm, range->start, range->end, range->flag, fault_addr, reason, error);
@@ -1936,9 +1948,6 @@ retry:
 		// memory_stat_rss_add() is called in downstream with !memobj check
 	}
 	flush_tlb_single(fault_addr);
-	vm->currss += pgsize;
-	if(vm->currss > vm->proc->maxrss)
-		vm->proc->maxrss = vm->currss;
 
 	error = 0;
 	page = NULL;
@@ -2123,6 +2132,8 @@ int init_process_stack(struct thread *thread, struct program_load_desc *pn,
 	struct process *proc = thread->proc;
 	unsigned long ap_flag;
 	struct vm_range *range;
+	int stack_populated_size = 0;
+	int stack_align_padding = 0;
 
 	/* Create stack range */
 	end = STACK_TOP(&thread->vm->region) & LARGE_PAGE_MASK;
@@ -2156,7 +2167,7 @@ int init_process_stack(struct thread *thread, struct program_load_desc *pn,
 			ap_flag ? "(IHK_MC_AP_USER)" : "");
 
 	stack = ihk_mc_alloc_aligned_pages_user(minsz >> PAGE_SHIFT,
-				LARGE_PAGE_P2ALIGN, IHK_MC_AP_NOWAIT | ap_flag);
+				LARGE_PAGE_P2ALIGN, IHK_MC_AP_NOWAIT | ap_flag, start);
 
 	if (!stack) {
 		kprintf("%s: error: couldn't allocate initial stack\n",
@@ -2194,22 +2205,29 @@ int init_process_stack(struct thread *thread, struct program_load_desc *pn,
 		return error;
 	}
 
-	// memory_stat_rss_add() is called in ihk_mc_pt_set_range();
+	/* Pre-compute populated size so that we can align stack
+	 * and verify the size at the end */
+	stack_align_padding = 0;
+	stack_populated_size = 16 /* Random */ +
+		AUXV_LEN * sizeof(unsigned long) /* AUXV */ +
+		(argc + 2) * sizeof(unsigned long) /* args + term NULL + argc */ +
+		(envc + 1) * sizeof(unsigned long); /* envs + term NULL */
 
 	/* set up initial stack frame */
 	p = (unsigned long *)(stack + minsz);
 	s_ind = -1;
 
-#ifdef POSTK_DEBUG_ARCH_DEP_15 /* userstack 16byte align */
-	if(!((envc + argc) % 2)){
-		p[s_ind--] = 0;
+	/* Align stack to 64 bytes */
+	while ((unsigned long)(stack + minsz -
+				stack_populated_size - stack_align_padding) & (0x40L - 1)) {
+		s_ind--;
+		stack_align_padding += sizeof(unsigned long);
 	}
-#endif /* POSTK_DEBUG_ARCH_DEP_15 */
 
 	/* "random" 16 bytes on the very top */
 	p[s_ind--] = 0x010101011;
 	p[s_ind--] = 0x010101011;
-	at_rand = end + sizeof(unsigned long) * s_ind;
+	at_rand = end + (s_ind + 1) * sizeof(unsigned long);
 
 	/* auxiliary vector */
 	/* If you add/delete entires, please increase/decrease
@@ -2261,6 +2279,20 @@ int init_process_stack(struct thread *thread, struct program_load_desc *pn,
 	/* argc */
 	p[s_ind] = argc;
 
+	if (((void *)&p[s_ind] != (void *)stack + minsz -
+				stack_populated_size - stack_align_padding)) {
+		kprintf("%s: WARNING: stack_populated_size mismatch (is AUXV_LEN up-to-date?): "
+				"&p[s_ind]: %lu, computed: %lu\n",
+				__FUNCTION__,
+				(unsigned long)&p[s_ind],
+				(unsigned long)stack + minsz -
+					stack_populated_size - stack_align_padding);
+	}
+
+	if ((unsigned long)&p[s_ind] & (0x40L - 1)) {
+		kprintf("%s: WARNING: stack alignment mismatch\n", __FUNCTION__);
+	}
+
 	ihk_mc_modify_user_context(thread->uctx, IHK_UCR_STACK_POINTER,
 	                           end + sizeof(unsigned long) * s_ind);
 	thread->vm->region.stack_end = end;
@@ -2300,7 +2332,7 @@ unsigned long extend_process_region(struct process_vm *vm,
 		p = ihk_mc_alloc_aligned_pages_user(
 				(new_end_allocated - end_allocated) >> PAGE_SHIFT,
 				align_p2align, IHK_MC_AP_NOWAIT |
-				(!(vm->proc->mpol_flags & MPOL_NO_HEAP) ? IHK_MC_AP_USER : 0));
+				(!(vm->proc->mpol_flags & MPOL_NO_HEAP) ? IHK_MC_AP_USER : 0), end_allocated);
 
 		if (!p) {
 			return end_allocated;
@@ -2542,14 +2574,15 @@ out:
 	return error;
 }
 
-void hold_thread(struct thread *thread)
+int hold_thread(struct thread *thread)
 {
 	if (thread->status == PS_EXITED) {
-		panic("hold_thread: already exited process");
+		kprintf("hold_thread: ERROR: already exited process,tid=%d\n", thread->tid);
+		return -ESRCH;
 	}
 
 	ihk_atomic_inc(&thread->refcount);
-	return;
+	return 0;
 }
 
 void
@@ -2596,21 +2629,30 @@ void destroy_thread(struct thread *thread)
 {
 	struct sig_pending *pending;
 	struct sig_pending *signext;
-	struct mcs_rwlock_node_irqsave lock;
+	struct mcs_rwlock_node_irqsave lock, updatelock;
 	struct process *proc = thread->proc;
 	struct resource_set *resource_set = cpu_local_var(resource_set);
 	int hash;
+	struct timespec ats;
 
 	hash = thread_hash(thread->tid);
 	mcs_rwlock_writer_lock(&resource_set->thread_hash->lock[hash], &lock);
 	list_del(&thread->hash_list);
 	mcs_rwlock_writer_unlock(&resource_set->thread_hash->lock[hash], &lock);
 
+	mcs_rwlock_writer_lock(&proc->update_lock, &updatelock);
+	tsc_to_ts(thread->system_tsc, &ats);
+	ts_add(&thread->proc->stime, &ats);
+	tsc_to_ts(thread->user_tsc, &ats);
+	ts_add(&thread->proc->utime, &ats);
+
 	mcs_rwlock_writer_lock(&proc->threads_lock, &lock);
 	list_del(&thread->siblings_list);
 	__release_tid(proc, thread);
 	mcs_rwlock_writer_unlock(&proc->threads_lock, &lock);
 
+	mcs_rwlock_writer_unlock(&proc->update_lock, &updatelock);
+
 	cpu_clear(thread->cpu_id, &thread->vm->address_space->cpu_set,
 	          &thread->vm->address_space->cpu_set_lock);
 	list_for_each_entry_safe(pending, signext, &thread->sigpending, list){
@@ -2639,20 +2681,11 @@ void destroy_thread(struct thread *thread)
 void release_thread(struct thread *thread)
 {
 	struct process_vm *vm;
-	struct mcs_rwlock_node_irqsave lock;
-	struct timespec ats;
 
 	if (!ihk_atomic_dec_and_test(&thread->refcount)) {
 		return;
 	}
 
-	mcs_rwlock_writer_lock(&thread->proc->update_lock, &lock);
-	tsc_to_ts(thread->system_tsc, &ats);
-	ts_add(&thread->proc->stime, &ats);
-	tsc_to_ts(thread->user_tsc, &ats);
-	ts_add(&thread->proc->utime, &ats);
-	mcs_rwlock_writer_unlock(&thread->proc->update_lock, &lock);
-
 	vm = thread->vm;
 
 #ifdef PROFILE_ENABLE
@@ -2891,7 +2924,7 @@ void sched_init(void)
 	ihk_mc_init_context(&idle_thread->ctx, NULL, idle);
 	ihk_mc_spinlock_init(&idle_thread->vm->memory_range_lock);
 	idle_thread->vm->vm_range_tree = RB_ROOT;
-	INIT_LIST_HEAD(&idle_thread->vm->vm_range_numa_policy_list);
+	idle_thread->vm->vm_range_numa_policy_tree = RB_ROOT;
 	idle_thread->proc->pid = 0;
 	idle_thread->tid = ihk_mc_get_processor_id();
 
@@ -3044,6 +3077,12 @@ void spin_sleep_or_schedule(void)
 	int woken = 0;
 	long irqstate;
 
+	/* Spinning disabled explicitly */
+	if (idle_halt) {
+		dkprintf("%s: idle_halt -> schedule()\n", __FUNCTION__);
+		goto out_schedule;
+	}
+
 	/* Try to spin sleep */
 	irqstate = ihk_mc_spinlock_lock(&thread->spin_sleep_lock);
 	if (thread->spin_sleep == 0) {
@@ -3092,6 +3131,7 @@ void spin_sleep_or_schedule(void)
 		cpu_pause();
 	}
 
+out_schedule:
 	schedule();
 }
 
@@ -3109,6 +3149,9 @@ void schedule(void)
 	}
 
 redo:
+	/* Reset for redo */
+	switch_ctx = 0;
+
 	cpu_local_var(runq_irqstate) = 
 		ihk_mc_spinlock_lock(&(get_this_cpu_local_var()->runq_lock));
 	v = get_this_cpu_local_var();
@@ -3130,7 +3173,10 @@ redo:
 		}
 	}
 
-	if (v->flags & CPU_FLAG_NEED_MIGRATE) {
+	/* Switch to idle() when prev is PS_EXITED since it always reaches release_thread() 
+	   because it always resumes from just after ihk_mc_switch_context() call. See #1029 */
+	if (v->flags & CPU_FLAG_NEED_MIGRATE ||
+	    prev->status == PS_EXITED) {
 		next = &cpu_local_var(idle);
 	} else {
 		/* Pick a new running process or one that has a pending signal */

kernel/procfs.c

@@ -358,7 +358,7 @@ void process_procfs_request(struct ikc_scd_packet *rpacket)
 			 *  08048000-08056000 r-xp 00000000 03:0c 64593   /usr/sbin/gpm
 			 */
 			written_now = snprintf(_buf, left, 
-					"%lx-%lx %s%s%s%s %lx %lx:%lx %d %s\n",
+					"%012lx-%012lx %s%s%s%s %lx %lx:%lx %d\t\t\t%s\n",
 					range->start, range->end,
 					range->flag & VR_PROT_READ ? "r" : "-",
 					range->flag & VR_PROT_WRITE ? "w" : "-",
@@ -369,6 +369,14 @@ void process_procfs_request(struct ikc_scd_packet *rpacket)
 					0UL,
 					0UL,
 					0,
+					range->start ==
+						(unsigned long)vm->vdso_addr ? "[vdso]" :
+					range->start ==
+						(unsigned long)vm->vvar_addr ? "[vsyscall]" :
+					range->flag & VR_STACK ? "[stack]" :
+					range->start >= vm->region.brk_start &&
+						range->end <= vm->region.brk_end_allocated ?
+						"[heap]" :
 					""
 					);
 			
@@ -422,12 +430,6 @@ void process_procfs_request(struct ikc_scd_packet *rpacket)
 	if (strcmp(p, "pagemap") == 0) {
 		uint64_t *_buf = (uint64_t *)buf;
 		uint64_t start, end;
-		
-		if (offset < PAGE_SIZE) {
-			kprintf("WARNING: /proc/pagemap queried for NULL page\n");
-			ans = 0;
-			goto end;
-		}
 
 		/* Check alignment */
 		if ((offset % sizeof(uint64_t) != 0) || 

kernel/shmobj.c

@@ -374,7 +374,7 @@ static void shmobj_ref(struct memobj *memobj)
 }
 
 static int shmobj_get_page(struct memobj *memobj, off_t off, int p2align,
-               uintptr_t *physp, unsigned long *pflag)
+               uintptr_t *physp, unsigned long *pflag, uintptr_t virt_addr)
 {
 	struct shmobj *obj = to_shmobj(memobj);
 	int error;
@@ -415,7 +415,7 @@ static int shmobj_get_page(struct memobj *memobj, off_t off, int p2align,
 	if (!page) {
 		npages = 1 << p2align;
 		virt = ihk_mc_alloc_aligned_pages_user(npages, p2align,
-				IHK_MC_AP_NOWAIT);
+				IHK_MC_AP_NOWAIT, virt_addr);
 		if (!virt) {
 			error = -ENOMEM;
 			ekprintf("shmobj_get_page(%p,%#lx,%d,%p):"

kernel/syscall.c

@@ -113,6 +113,7 @@ static ihk_spinlock_t tod_data_lock = SPIN_LOCK_UNLOCKED;
 static void calculate_time_from_tsc(struct timespec *ts);
 
 void check_signal(unsigned long, void *, int);
+void save_syscall_return_value(int num, unsigned long rc);
 void do_signal(long rc, void *regs, struct thread *thread, struct sig_pending *pending, int num);
 extern unsigned long do_kill(struct thread *thread, int pid, int tid, int sig, struct siginfo *info, int ptracecont);
 extern long alloc_debugreg(struct thread *thread);
@@ -319,10 +320,10 @@ long do_syscall(struct syscall_request *req, int cpu, int pid)
 			flags = cpu_disable_interrupt_save();
 
 			/* Try to sleep until notified */
-			if (__sync_bool_compare_and_swap(&res.req_thread_status,
-				IHK_SCD_REQ_THREAD_SPINNING,
-				IHK_SCD_REQ_THREAD_DESCHEDULED)) {
-
+			if (res.req_thread_status == IHK_SCD_REQ_THREAD_DESCHEDULED ||
+				__sync_bool_compare_and_swap(&res.req_thread_status,
+											 IHK_SCD_REQ_THREAD_SPINNING,
+											 IHK_SCD_REQ_THREAD_DESCHEDULED)) {
 				dkprintf("%s: tid %d waiting for syscall reply...\n",
 						__FUNCTION__, thread->tid);
 				waitq_init(&thread->scd_wq);
@@ -459,16 +460,10 @@ long do_syscall(struct syscall_request *req, int cpu, int pid)
 
 	/* -ERESTARTSYS indicates that the proxy process is gone
 	 * and the application should be terminated */
-#ifdef POSTK_DEBUG_TEMP_FIX_70 /* interrupt_syscall returned -ERESTARTSYS fix */
-	if (rc == -ERESTARTSYS && req->number != __NR_exit_group
-		&& req->number != __NR_kill) {
-#else /* POSTK_DEBUG_TEMP_FIX_70 */
-	if (rc == -ERESTARTSYS && req->number != __NR_exit_group) {
-#endif /* POSTK_DEBUG_TEMP_FIX_70 */
+	if (rc == -ERESTARTSYS) {
 		kprintf("%s: proxy PID %d is dead, terminate()\n",
 			__FUNCTION__, thread->proc->pid);
 		thread->proc->nohost = 1;
-		terminate(0, SIGKILL);
 	}
 
 #ifdef PROFILE_ENABLE
@@ -963,6 +958,28 @@ SYSCALL_DECLARE(waitid)
 	return 0;
 }
 
+void terminate_mcexec(int rc, int sig)
+{
+	unsigned long old_exit_status;
+	unsigned long exit_status;
+	struct thread *mythread = cpu_local_var(current);
+	struct process *proc = mythread->proc;
+	struct syscall_request request IHK_DMA_ALIGN;
+
+	if ((old_exit_status = proc->exit_status) & 0x0000000100000000L)
+		return;
+	exit_status = 0x0000000100000000L | ((rc & 0x00ff) << 8) | (sig & 0xff);
+	if (!__sync_bool_compare_and_swap(&proc->exit_status,
+	                                  old_exit_status, exit_status))
+		return;
+	if (!proc->nohost) {
+		request.number = __NR_exit_group;
+		request.args[0] = proc->exit_status;
+		proc->nohost = 1;
+		do_syscall(&request, ihk_mc_get_processor_id(), proc->pid);
+	}
+}
+
 void terminate(int rc, int sig)
 {
 	struct resource_set *resource_set = cpu_local_var(resource_set);
@@ -980,7 +997,6 @@ void terminate(int rc, int sig)
 	int i;
 	int n;
 	int *ids = NULL;
-	struct syscall_request request IHK_DMA_ALIGN;
 	int exit_status;
 
 	// sync perf info
@@ -988,13 +1004,15 @@ void terminate(int rc, int sig)
 		sync_child_event(proc->monitoring_event);
 
 	// clean up threads
-	mcs_rwlock_reader_lock(&proc->threads_lock, &lock); // conflict clone
 	mcs_rwlock_writer_lock_noirq(&proc->update_lock, &updatelock);
+	mcs_rwlock_writer_lock(&proc->threads_lock, &lock); // conflict clone
 	if (proc->status == PS_EXITED) {
-		mcs_rwlock_writer_unlock_noirq(&proc->update_lock, &updatelock);
-		mcs_rwlock_reader_unlock(&proc->threads_lock, &lock);
+		dkprintf("%s: PID: %d, TID: %d PS_EXITED already\n",
+				__FUNCTION__, proc->pid, mythread->tid);
 		preempt_disable();
 		mythread->status = PS_EXITED;
+		mcs_rwlock_writer_unlock(&proc->threads_lock, &lock);
+		mcs_rwlock_writer_unlock_noirq(&proc->update_lock, &updatelock);
 		release_thread(mythread);
 		preempt_enable();
 		schedule();
@@ -1002,10 +1020,15 @@ void terminate(int rc, int sig)
 		return;
 	}
 
-	exit_status = mythread->exit_status = ((rc & 0x00ff) << 8) | (sig & 0xff);
+	dkprintf("%s: PID: %d, TID: %d setting PS_EXITED\n",
+			__FUNCTION__, proc->pid, mythread->tid);
+	exit_status = ((rc & 0x00ff) << 8) | (sig & 0xff);
+	mythread->exit_status = exit_status;
 	proc->status = PS_EXITED;
+	mcs_rwlock_writer_unlock(&proc->threads_lock, &lock);
 	mcs_rwlock_writer_unlock_noirq(&proc->update_lock, &updatelock);
-	mcs_rwlock_reader_unlock(&proc->threads_lock, &lock);
+
+	terminate_mcexec(rc, sig);
 
 	mcs_rwlock_writer_lock(&proc->threads_lock, &lock);
 	list_del(&mythread->siblings_list);
@@ -1151,23 +1174,13 @@ void terminate(int rc, int sig)
 #endif
 
 	// clean up memory
-	if (!proc->nohost) {
-		request.number = __NR_exit_group;
-		request.args[0] = exit_status;
-#ifdef POSTK_DEBUG_TEMP_FIX_48 /* nohost flag missed fix */
-		proc->nohost = 1;
-		do_syscall(&request, ihk_mc_get_processor_id(), proc->pid);
-#else /* POSTK_DEBUG_TEMP_FIX_48 */
-		do_syscall(&request, ihk_mc_get_processor_id(), proc->pid);
-		proc->nohost = 1;
-#endif /* POSTK_DEBUG_TEMP_FIX_48 */
-	}
 
-	proc->exit_status = exit_status;
 	finalize_process(proc);
 
 	preempt_disable();
+	mcs_rwlock_writer_lock(&proc->threads_lock, &lock);
 	mythread->status = PS_EXITED;
+	mcs_rwlock_writer_unlock(&proc->threads_lock, &lock);
 	release_thread(mythread);
 	release_process_vm(vm);
 	preempt_enable();
@@ -1479,7 +1492,7 @@ do_mmap(const intptr_t addr0, const size_t len0, const int prot,
 	maxprot = PROT_READ | PROT_WRITE | PROT_EXEC;
 	if (!(flags & MAP_ANONYMOUS)) {
 		off = off0;
-		error = fileobj_create(fd, &memobj, &maxprot);
+		error = fileobj_create(fd, &memobj, &maxprot, addr0);
 #ifdef ATTACHED_MIC
 		/*
 		 * XXX: refuse device mapping in attached-mic now:
@@ -1542,7 +1555,7 @@ do_mmap(const intptr_t addr0, const size_t len0, const int prot,
 		}
 
 		p = ihk_mc_alloc_aligned_pages_user(npages, p2align,
-				IHK_MC_AP_NOWAIT | ap_flag);
+				IHK_MC_AP_NOWAIT | ap_flag, addr0);
 		if (p == NULL) {
 			dkprintf("%s: warning: failed to allocate %d contiguous pages "
 					" (bytes: %lu, pgshift: %d), enabling demand paging\n",
@@ -1966,9 +1979,7 @@ static void settid(struct thread *thread, int nr_tids, int *tids)
 	int ret;
 	struct syscall_request request IHK_DMA_ALIGN;
 
-#ifdef POSTK_DEBUG_ARCH_DEP_58 /* settid() arguments area 0 clear add. */
 	memset(&request, 0, sizeof(request));
-#endif /* POSTK_DEBUG_ARCH_DEP_58 */
 
 	request.number = __NR_gettid;
 	/*
@@ -2152,11 +2163,9 @@ SYSCALL_DECLARE(execve)
 {
 	int error;
 	long ret;
-	char *empty_envp[1] = {NULL};
 	const char *filename = (const char *)ihk_mc_syscall_arg0(ctx);
 	char **argv = (char **)ihk_mc_syscall_arg1(ctx);
-	char **envp = (char **)ihk_mc_syscall_arg2(ctx) ? 
-		(char **)ihk_mc_syscall_arg2(ctx) : empty_envp;
+	char **envp = (char **)ihk_mc_syscall_arg2(ctx);
 
 	char *argv_flat = NULL;
 	int argv_flat_len = 0;
@@ -2202,13 +2211,8 @@ SYSCALL_DECLARE(execve)
 	if (ret != 0) {
 		dkprintf("execve(): ERROR: host failed to load elf header, errno: %d\n", 
 				ret);
-#ifdef POSTK_DEBUG_TEMP_FIX_10 /* sys_execve() memleak fix. */
 		ret = -ret;
-		goto desc_free;
-#else /* POSTK_DEBUG_TEMP_FIX_10 */
-		ihk_mc_free_pages(desc, 4);
-		return -ret;
-#endif /* POSTK_DEBUG_TEMP_FIX_10 */
+		goto end;
 	}
 
 	dkprintf("execve(): ELF desc received, num sections: %d\n",
@@ -2221,7 +2225,7 @@ SYSCALL_DECLARE(execve)
 	/* Flatten argv and envp into kernel-space buffers */
 	argv_flat_len = flatten_strings_from_user(-1, (desc->shell_path[0] ? 
 				desc->shell_path : NULL), argv, &argv_flat);
-	if (argv_flat_len == 0) {
+	if (argv_flat_len < 0) {
 		char *kfilename;
 		int len = strlen_user(filename);
 
@@ -2231,17 +2235,12 @@ SYSCALL_DECLARE(execve)
 		kprintf("ERROR: no argv for executable: %s?\n", kfilename? kfilename: "");
 		if(kfilename)
 			kfree(kfilename);
-#ifdef POSTK_DEBUG_TEMP_FIX_10 /* sys_execve() memleak fix. */
-		ret = -EINVAL;
-		goto desc_free;
-#else /* POSTK_DEBUG_TEMP_FIX_10 */
-		ihk_mc_free_pages(desc, 4);
-		return -EINVAL;
-#endif /* POSTK_DEBUG_TEMP_FIX_10 */
+		ret = argv_flat_len;
+		goto end;
 	}
 
 	envp_flat_len = flatten_strings_from_user(-1, NULL, envp, &envp_flat);
-	if (envp_flat_len == 0) {
+	if (envp_flat_len < 0) {
 		char *kfilename;
 		int len = strlen_user(filename);
 
@@ -2251,12 +2250,8 @@ SYSCALL_DECLARE(execve)
 		kprintf("ERROR: no envp for executable: %s?\n", kfilename? kfilename: "");
 		if(kfilename)
 			kfree(kfilename);
-#ifdef POSTK_DEBUG_TEMP_FIX_10 /* sys_execve() memleak fix. */
-		ret = -EINVAL;
-		goto argv_free;
-#else /* POSTK_DEBUG_TEMP_FIX_10 */
-		return -EINVAL;
-#endif /* POSTK_DEBUG_TEMP_FIX_10 */
+		ret = envp_flat_len;
+		goto end;
 	}
 
 	if (cpu_local_var(current)->proc->ptrace) {
@@ -2293,11 +2288,8 @@ SYSCALL_DECLARE(execve)
 	request.args[2] = sizeof(struct program_load_desc) + 
 		sizeof(struct program_image_section) * desc->num_sections;
 
-	ret = do_syscall(&request, ihk_mc_get_processor_id(), 0);
-
-	if (ret != 0) {
-		kprintf("execve(): PANIC: host failed to load elf image\n");
-		panic("");
+	if ((ret = do_syscall(&request, ihk_mc_get_processor_id(), 0)) != 0) {
+		goto end;
 	}
 
 	for(i = 0; i < _NSIG; i++){
@@ -2320,18 +2312,14 @@ SYSCALL_DECLARE(execve)
 	dkprintf("execve(): switching to new process\n");
 	proc->execed = 1;
 	
-#ifdef POSTK_DEBUG_TEMP_FIX_10 /* sys_execve() memleak fix. */
 	ret = 0;
+end:
 	if (envp_flat) {
 		kfree(envp_flat);
 	}
-
-argv_free:
 	if (argv_flat) {
 		kfree(argv_flat);
 	}
-
-desc_free:
 	ihk_mc_free_pages(desc, 4);
 
 	if (!ret) {
@@ -2343,21 +2331,6 @@ desc_free:
 			cpu_local_var(current));
 	}
 	return ret;
-#else /* POSTK_DEBUG_TEMP_FIX_10 */
-	ihk_mc_free_pages(desc, 4);
-	kfree(argv_flat);
-	kfree(envp_flat);
-	
-	/* Lock run queue because enter_user_mode expects to release it */
-	cpu_local_var(runq_irqstate) = 
-		ihk_mc_spinlock_lock(&(get_this_cpu_local_var()->runq_lock));
-
-	ihk_mc_switch_context(NULL, &cpu_local_var(current)->ctx, 
-		cpu_local_var(current));
-
-	/* Never reach here */
-	return 0;
-#endif /* POSTK_DEBUG_TEMP_FIX_10 */
 }
 
 unsigned long do_fork(int clone_flags, unsigned long newsp,
@@ -2374,6 +2347,7 @@ unsigned long do_fork(int clone_flags, unsigned long newsp,
 	struct syscall_request request1 IHK_DMA_ALIGN;
 	int ptrace_event = 0;
 	int termsig = clone_flags & 0x000000ff;
+	const struct ihk_mc_cpu_info *cpu_info = ihk_mc_get_cpu_info();
 
     dkprintf("do_fork,flags=%08x,newsp=%lx,ptidptr=%lx,ctidptr=%lx,tls=%lx,curpc=%lx,cursp=%lx",
             clone_flags, newsp, parent_tidptr, child_tidptr, tlsblock_base, curpc, cursp);
@@ -2413,6 +2387,11 @@ unsigned long do_fork(int clone_flags, unsigned long newsp,
 		return -EINVAL;
 	}
 
+	if (!allow_oversubscribe && rusage->num_threads >= cpu_info->ncpus) {
+		kprintf("%s: ERROR: CPU oversubscription is not allowed. Specify -O option in mcreboot.sh to allow it.\n", __FUNCTION__);
+        return -EINVAL;
+	}
+
 	cpuid = obtain_clone_cpuid(&old->cpu_set);
     if (cpuid == -1) {
 		kprintf("do_fork,core not available\n");
@@ -2482,28 +2461,25 @@ retry_tid:
 			}
 		}
 
+		mcs_rwlock_writer_unlock(&newproc->threads_lock, &lock);
+
 		/* TODO: spawn more mcexec threads */
 		if (!new->tid) {
+			release_cpuid(cpuid);
 			kprintf("%s: no more TIDs available\n");
-			panic("");
+			return -ENOMEM;
 		}
-
-		mcs_rwlock_writer_unlock(&newproc->threads_lock, &lock);
 	}
 	/* fork() a new process on the host */
 	else {
-		request1.number = __NR_fork;
+		request1.number = __NR_clone;
 		request1.args[0] = 0;
 		if(clone_flags & CLONE_PARENT){
 			if(oldproc->ppid_parent->pid != 1)
 				request1.args[0] = clone_flags;
 		}
 		newproc->pid = do_syscall(&request1, ihk_mc_get_processor_id(), 0);
-#ifdef POSTK_DEBUG_TEMP_FIX_12 /* __NR_fork retval check fix. */
 		if (newproc->pid < 0) {
-#else /* POSTK_DEBUG_TEMP_FIX_12 */
-		if (newproc->pid == -1) {
-#endif /* POSTK_DEBUG_TEMP_FIX_12 */
 			kprintf("ERROR: forking host process\n");
 			
 			/* TODO: clean-up new */
@@ -2651,6 +2627,13 @@ retry_tid:
 		old->tid,
 		new->tid);
 
+	if (!(clone_flags & CLONE_VM)) {
+		request1.number = __NR_clone;
+		request1.args[0] = 1;
+		request1.args[1] = new->tid;
+		do_syscall(&request1, ihk_mc_get_processor_id(), 0);
+	}
+
 	runq_add_thread(new, cpuid);
 
 	if (ptrace_event) {
@@ -2660,11 +2643,6 @@ retry_tid:
 	return new->tid;
 }
 
-SYSCALL_DECLARE(vfork)
-{
-    return do_fork(CLONE_VFORK|SIGCHLD, 0, 0, 0, 0, ihk_mc_syscall_pc(ctx), ihk_mc_syscall_sp(ctx));
-}
-
 SYSCALL_DECLARE(set_tid_address)
 {
 	cpu_local_var(current)->clear_child_tid = 
@@ -3093,6 +3071,13 @@ do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
 	struct mcs_rwlock_node_irqsave mcs_rw_node;
 	ihk_mc_user_context_t ctx0;
 
+	if (!valid_signal(sig) || sig < 1) {
+		return -EINVAL;
+	}
+	if (act && (sig == SIGKILL || sig == SIGSTOP)) {
+		return -EINVAL;
+	}
+
 	mcs_rwlock_writer_lock(&thread->sigcommon->lock, &mcs_rw_node);
 	k = thread->sigcommon->action + sig - 1;
 	if(oact)
@@ -3817,6 +3802,57 @@ perf_mmap(struct mckfd *sfd, ihk_mc_user_context_t *ctx)
 	return rc;
 }
 
+struct vm_range_numa_policy *vm_range_policy_search(struct process_vm *vm, uintptr_t addr)
+{
+	struct rb_root *root = &vm->vm_range_numa_policy_tree;
+	struct rb_node *node = root->rb_node;
+	struct vm_range_numa_policy *numa_policy = NULL;
+
+	while (node) {
+		numa_policy = rb_entry(node, struct vm_range_numa_policy, policy_rb_node);
+		if (addr < numa_policy->start) {
+			node = node->rb_left;
+		} else if (addr >= numa_policy->end) {
+			node = node->rb_right;
+		} else {
+			return numa_policy;
+		}
+	}
+
+	return NULL;
+}
+
+static int vm_policy_insert(struct process_vm *vm, struct vm_range_numa_policy *newrange)
+{
+	struct rb_root *root = &vm->vm_range_numa_policy_tree;
+	struct rb_node **new = &(root->rb_node), *parent = NULL;
+	struct vm_range_numa_policy *range;
+
+	while (*new) {
+		range = rb_entry(*new, struct vm_range_numa_policy, policy_rb_node);
+		parent = *new;
+		if (newrange->end <= range->start) {
+			new = &((*new)->rb_left);
+		} else if (newrange->start >= range->end) {
+			new = &((*new)->rb_right);
+		} else {
+			ekprintf("vm_range_insert(%p,%lx-%lx (nodemask)%lx (policy)%d): overlap %lx-%lx (nodemask)%lx (policy)%d\n",
+					vm, newrange->start, newrange->end, newrange->numa_mask, newrange->numa_mem_policy,
+					range->start, range->end, range->numa_mask, range->numa_mem_policy);
+			return -EFAULT;
+		}
+	}
+
+	dkprintf("vm_range_insert: %p,%p: %lx-%lx (nodemask)%lx (policy)%d\n", 
+				vm, newrange, newrange->start, newrange->end, newrange->numa_mask, newrange->numa_mem_policy);
+
+	rb_link_node(&newrange->policy_rb_node, parent, new);
+	rb_insert_color(&newrange->policy_rb_node, root);
+
+	return 0;
+}
+
+
 struct mc_perf_event*
 mc_perf_event_alloc(struct perf_event_attr *attr)
 {
@@ -3926,7 +3962,7 @@ SYSCALL_DECLARE(perf_event_open)
 	}
 
 	if (not_supported_flag) {
-		return -1;
+		return -ENOENT;
 	}
 
 	event = mc_perf_event_alloc((struct perf_event_attr*)attr);
@@ -4684,30 +4720,6 @@ struct shminfo the_shminfo = {
 };
 struct shm_info the_shm_info = { 0, };
 
-time_t time(void) {
-#ifndef POSTK_DEBUG_ARCH_DEP_13 /* arch depend tmp hide */
-	struct syscall_request sreq IHK_DMA_ALIGN;
-	struct thread *thread = cpu_local_var(current);
-#endif /* POSTK_DEBUG_ARCH_DEP_13 */
-
-#ifdef POSTK_DEBUG_ARCH_DEP_49 /* time() local calculate added. */
-	struct timespec ats;
-
-	if (gettime_local_support) {
-		calculate_time_from_tsc(&ats);
-		return ats.tv_sec;
-	}
-#endif /* POSTK_DEBUG_ARCH_DEP_49 */
-
-#ifdef POSTK_DEBUG_ARCH_DEP_13 /* arch depend tmp hide */
-	return (time_t)0;
-#else /* POSTK_DEBUG_ARCH_DEP_13 */
-	sreq.number = __NR_time;
-	sreq.args[0] = (uintptr_t)NULL;
-	return (time_t)do_syscall(&sreq, ihk_mc_get_processor_id(), thread->proc->pid);
-#endif /* POSTK_DEBUG_ARCH_DEP_13 */
-}
-
 static int make_shmid(struct shmobj *obj)
 {
 	return ((int)obj->index << 16) | obj->ds.shm_perm.seq;
@@ -5475,16 +5487,16 @@ do_exit(int code)
 		      FUTEX_WAKE, 1, 0, NULL, 0, 0, 1);
 	}
 
-	mcs_rwlock_reader_lock(&proc->threads_lock, &lock);
+	mcs_rwlock_writer_lock(&proc->threads_lock, &lock);
 	if(proc->status == PS_EXITED){
-		mcs_rwlock_reader_unlock(&proc->threads_lock, &lock);
+		mcs_rwlock_writer_unlock(&proc->threads_lock, &lock);
 		terminate(exit_status, 0);
 		return;
 	}
 	preempt_disable();
 	thread->status = PS_EXITED;
 	sync_child_event(thread->proc->monitoring_event);
-	mcs_rwlock_reader_unlock(&proc->threads_lock, &lock);
+	mcs_rwlock_writer_unlock(&proc->threads_lock, &lock);
 	release_thread(thread);
 	preempt_enable();
 
@@ -5681,15 +5693,8 @@ SYSCALL_DECLARE(getrusage)
 		kusage.ru_maxrss = proc->maxrss / 1024;
 	}
 	else if(who == RUSAGE_CHILDREN){
-#ifdef POSTK_DEBUG_TEMP_FIX_72 /* fix RUSAGE_CHILDREN time */
 		ts_to_tv(&kusage.ru_utime, &proc->utime_children);
 		ts_to_tv(&kusage.ru_stime, &proc->stime_children);
-#else /* POSTK_DEBUG_TEMP_FIX_72 */
-		tsc_to_ts(thread->user_tsc, &ats);
-		ts_to_tv(&kusage.ru_utime, &ats);
-		tsc_to_ts(thread->system_tsc, &ats);
-		ts_to_tv(&kusage.ru_stime, &ats);
-#endif /* POSTK_DEBUG_TEMP_FIX_72 */
 
 		kusage.ru_maxrss = proc->maxrss_children / 1024;
 	}
@@ -6095,6 +6100,13 @@ static int ptrace_attach(int pid)
 		error = -ESRCH;
 		goto out;
 	}
+
+	if (proc->pid == pid) {
+		thread_unlock(thread, &lock);
+		error = -EPERM;
+		goto out;
+	}
+
 	child = thread->proc;
 	dkprintf("ptrace_attach(): pid requested:%d, thread->tid:%d, thread->proc->pid=%d, thread->proc->parent=%p\n", pid, thread->tid, thread->proc->pid, thread->proc->parent);
 
@@ -6112,7 +6124,6 @@ static int ptrace_attach(int pid)
 	}
 
 	parent = child->parent;
-#ifdef POSTK_DEBUG_TEMP_FIX_53 /* attach for child-process fix. */
 	dkprintf("ptrace_attach() parent->pid=%d\n", parent->pid);
 
 	mcs_rwlock_writer_lock_noirq(&parent->children_lock, &childlock);
@@ -6124,23 +6135,6 @@ static int ptrace_attach(int pid)
 	list_add_tail(&child->siblings_list, &proc->children_list);
 	child->parent = proc;
 	mcs_rwlock_writer_unlock_noirq(&proc->children_lock, &childlock);
-#else /* POSTK_DEBUG_TEMP_FIX_53 */
-	/* XXX: tmp */
-	if (parent != proc) {
-
-		dkprintf("ptrace_attach() parent->pid=%d\n", parent->pid);
-
-		mcs_rwlock_writer_lock_noirq(&parent->children_lock, &childlock);
-		list_del(&child->siblings_list);
-		list_add_tail(&child->ptraced_siblings_list, &parent->ptraced_children_list);
-		mcs_rwlock_writer_unlock_noirq(&parent->children_lock, &childlock);
-
-		mcs_rwlock_writer_lock_noirq(&proc->children_lock, &childlock);
-		list_add_tail(&child->siblings_list, &proc->children_list);
-		child->parent = proc;
-		mcs_rwlock_writer_unlock_noirq(&proc->children_lock, &childlock);
-	}
-#endif /* POSTK_DEBUG_TEMP_FIX_53 */
 
 	child->ptrace = PT_TRACED | PT_TRACE_EXEC;
 
@@ -8109,8 +8103,7 @@ SYSCALL_DECLARE(mbind)
 	int error = 0;
 	int bit;
 	struct vm_range *range;
-	struct vm_range_numa_policy *range_policy, *range_policy_iter;
-	struct vm_range_numa_policy *range_policy_next = NULL;
+	struct vm_range_numa_policy *range_policy, *range_policy_iter = NULL;
 	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
 
 	dkprintf("%s: addr: 0x%lx, len: %lu, mode: 0x%x, "
@@ -8321,17 +8314,12 @@ SYSCALL_DECLARE(mbind)
 		case MPOL_INTERLEAVE:
 		case MPOL_PREFERRED:
 			/* Adjust any overlapping range settings and add new one */
-			range_policy_next = NULL;
-			list_for_each_entry(range_policy_iter,
-					&vm->vm_range_numa_policy_list, list) {
+			range_policy_iter = vm_range_policy_search(vm, addr);
+			if (range_policy_iter) {
 				int adjusted = 0;
 				unsigned long orig_end = range_policy_iter->end;
 
-				if (range_policy_iter->end < addr)
-					continue;
-
-				/* Special case of entirely overlapping */
-				if (range_policy_iter->start == addr &&
+				if (range_policy_iter->start == addr &&	
 						range_policy_iter->end == addr + len) {
 					range_policy = range_policy_iter;
 					goto mbind_update_only;
@@ -8348,7 +8336,7 @@ SYSCALL_DECLARE(mbind)
 				if (orig_end > addr + len) {
 					if (adjusted) {
 						/* Add a new entry after */
-						range_policy = kmalloc(sizeof(*range_policy),
+						range_policy = kmalloc(sizeof(struct vm_range_numa_policy),
 								IHK_MC_AP_NOWAIT);
 						if (!range_policy) {
 							dkprintf("%s: error allocating range_policy\n",
@@ -8357,31 +8345,24 @@ SYSCALL_DECLARE(mbind)
 							goto unlock_out;
 						}
 
-						memcpy(range_policy, range_policy_iter,
-								sizeof(*range_policy));
+						RB_CLEAR_NODE(&range_policy->policy_rb_node);
 						range_policy->start = addr + len;
 						range_policy->end = orig_end;
-						list_add(&range_policy->list,
-								&range_policy_iter->list);
-						range_policy_next = range_policy;
-						break;
+
+						error = vm_policy_insert(vm, range_policy);
+						if (error) {
+							kprintf("%s: ERROR: could not insert range: %d\n",__FUNCTION__, error);
+							return error;
+						}
 					}
 					else {
 						range_policy_iter->start = addr + len;
-						range_policy_next = range_policy_iter;
-						break;
 					}
 				}
-
-				/* Next one in ascending address order? */
-				if (range_policy_iter->start >= addr + len) {
-					range_policy_next = range_policy_iter;
-					break;
-				}
 			}
 
 			/* Add a new entry */
-			range_policy = kmalloc(sizeof(*range_policy),
+			range_policy = kmalloc(sizeof(struct vm_range_numa_policy),
 					IHK_MC_AP_NOWAIT);
 			if (!range_policy) {
 				dkprintf("%s: error allocating range_policy\n",
@@ -8390,17 +8371,14 @@ SYSCALL_DECLARE(mbind)
 				goto unlock_out;
 			}
 
-			memset(range_policy, 0, sizeof(*range_policy));
+			RB_CLEAR_NODE(&range_policy->policy_rb_node);
 			range_policy->start = addr;
 			range_policy->end = addr + len;
 
-			if (range_policy_next) {
-				list_add_tail(&range_policy->list,
-						&range_policy_next->list);
-			}
-			else {
-				list_add_tail(&range_policy->list,
-						&vm->vm_range_numa_policy_list);
+			error = vm_policy_insert(vm, range_policy);
+			if (error) {
+				kprintf("%s: ERROR: could not insert range: %d\n",__FUNCTION__, error);
+				return error;
 			}
 
 mbind_update_only:
@@ -8441,8 +8419,6 @@ SYSCALL_DECLARE(set_mempolicy)
 	struct process_vm *vm = cpu_local_var(current)->vm;
 	int error = 0;
 	int bit, valid_mask;
-	struct vm_range_numa_policy *range_policy_iter;
-	struct vm_range_numa_policy *range_policy_next = NULL;
 	DECLARE_BITMAP(numa_mask, PROCESS_NUMA_MASK_BITS);
 
 	memset(numa_mask, 0, sizeof(numa_mask));
@@ -8463,6 +8439,8 @@ SYSCALL_DECLARE(set_mempolicy)
 		}
 	}
 
+	mode &= ~MPOL_MODE_FLAGS;
+
 	switch (mode) {
 		case MPOL_DEFAULT:
 			if (nodemask && nodemask_bits) {
@@ -8486,6 +8464,13 @@ SYSCALL_DECLARE(set_mempolicy)
 				set_bit(bit, vm->numa_mask);
 			}
 
+#if 0
+			/* In man, "MPOL_DEFAULT mode deletes a process memory policy 
+			   other than the default and interprets that the memory policy" 
+			   falls back to the system default policy ", but not to delete 
+			   the NUMA memory policy.
+			   There was no processing of Linux's same name command. */
+
 			/* Delete all range settings */
 			ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
 			list_for_each_entry_safe(range_policy_iter, range_policy_next,
@@ -8494,6 +8479,7 @@ SYSCALL_DECLARE(set_mempolicy)
 				kfree(range_policy_iter);
 			}
 			ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+#endif
 
 			vm->numa_mem_policy = mode;
 			error = 0;
@@ -8641,7 +8627,6 @@ SYSCALL_DECLARE(get_mempolicy)
 
 	/* Address range specific? */
 	if (flags & MPOL_F_ADDR) {
-		struct vm_range_numa_policy *range_policy_iter;
 		struct vm_range *range;
 
 		ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
@@ -8653,16 +8638,8 @@ SYSCALL_DECLARE(get_mempolicy)
 			goto out;
 		}
 
-		list_for_each_entry(range_policy_iter,
-					&vm->vm_range_numa_policy_list, list) {
-			if (range_policy_iter->start > addr ||
-				range_policy_iter->end <= addr) {
-				continue;
-			}
+		range_policy = vm_range_policy_search(vm, addr);
 
-			range_policy = range_policy_iter;
-			break;
-		}
 		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
 	}
 
@@ -9288,6 +9265,7 @@ set_cputime(int mode)
 		return;
 	}
 
+	cpu_disable_interrupt();
 	tsc = rdtsc();
 	if(thread->base_tsc != 0){
 		unsigned long dtsc = tsc - thread->base_tsc;
@@ -9368,6 +9346,7 @@ set_cputime(int mode)
 			}
 		}
 	}
+	cpu_enable_interrupt();
 }
 
 long syscall(int num, ihk_mc_user_context_t *ctx)
@@ -9386,6 +9365,7 @@ long syscall(int num, ihk_mc_user_context_t *ctx)
 #endif // DISABLE_SCHED_YIELD
 		set_cputime(1);
 
+//kprintf("syscall=%d\n", num);
 #ifdef PROFILE_ENABLE
 	if (thread->profile && thread->profile_start_ts) {
 		unsigned long ts = rdtsc();
@@ -9396,6 +9376,7 @@ long syscall(int num, ihk_mc_user_context_t *ctx)
 
 	if(cpu_local_var(current)->proc->status == PS_EXITED &&
 	   (num != __NR_exit && num != __NR_exit_group)){
+		save_syscall_return_value(num, -EINVAL);
 		check_signal(-EINVAL, NULL, 0);
 		set_cputime(0);
 		return -EINVAL;
@@ -9456,22 +9437,7 @@ long syscall(int num, ihk_mc_user_context_t *ctx)
 		l = ihk_mc_syscall_ret(ctx);
 	}
 
-#if defined(POSTK_DEBUG_TEMP_FIX_60) && defined(POSTK_DEBUG_TEMP_FIX_56)
-	check_signal(l, NULL, num);
-#elif defined(POSTK_DEBUG_TEMP_FIX_60) /* sched_yield called check_signal fix. */
-	if (num != __NR_futex) {
-		check_signal(l, NULL, num);
-	}
-#elif defined(POSTK_DEBUG_TEMP_FIX_56) /* in futex_wait() signal handring fix. */
-	if (num != __NR_sched_yield) {
-		check_signal(l, NULL, num);
-	}
-#else /* POSTK_DEBUG_TEMP_FIX_60 && POSTK_DEBUG_TEMP_FIX_56 */
-	if (!list_empty(&thread->sigpending) ||
-	    !list_empty(&thread->sigcommon->sigpending)) {
-		check_signal(l, NULL, num);
-	}
-#endif /* POSTK_DEBUG_TEMP_FIX_60 && POSTK_DEBUG_TEMP_FIX_56 */
+	save_syscall_return_value(num, l);
 
 #ifdef PROFILE_ENABLE
 	{
@@ -9511,10 +9477,20 @@ long syscall(int num, ihk_mc_user_context_t *ctx)
 	}
 #endif /* POSTK_DEBUG_TEMP_FIX_60 && POSTK_DEBUG_TEMP_FIX_56 */
 
+	if (!list_empty(&thread->sigpending) ||
+	    !list_empty(&thread->sigcommon->sigpending)) {
+		check_signal(l, NULL, num);
+	}
+
 #ifdef DISABLE_SCHED_YIELD
 	if (num != __NR_sched_yield)
 #endif // DISABLE_SCHED_YIELD
 		set_cputime(0);
 
+	if (thread->proc->nohost) { // mcexec termination was detected
+		terminate(0, SIGKILL);
+	}
+//kprintf("syscall=%d returns %lx(%ld)\n", num, l, l);
+
 	return l;
 }

kernel/xpmem.c

@@ -34,7 +34,6 @@
 #include <ihk/mm.h>
 #include <xpmem_private.h>
 
-
 struct xpmem_partition *xpmem_my_part = NULL;  /* pointer to this partition */
 
 
@@ -107,6 +106,7 @@ int xpmem_open(
 	mckfd->sig_no = -1;
 	mckfd->ioctl_cb = xpmem_ioctl;
 	mckfd->close_cb = xpmem_close;
+	mckfd->dup_cb = xpmem_dup;
 	mckfd->data = (long)proc;
 	irqstate = ihk_mc_spinlock_lock(&proc->mckfd_lock);
 
@@ -276,16 +276,11 @@ static int xpmem_ioctl(
 	return -EINVAL;
 }
 
-
 static int xpmem_close(
 	struct mckfd *mckfd,
 	ihk_mc_user_context_t *ctx)
 {
 	int n_opened;
-	struct process *proc = (struct process *)mckfd->data;
-	struct xpmem_thread_group *tg;
-	int index;
-	struct mcs_rwlock_node_irqsave lock;
 
 	XPMEM_DEBUG("call: fd=%d, pid=%d, rgid=%d", 
 		mckfd->fd, proc->pid, proc->rgid);
@@ -293,37 +288,11 @@ static int xpmem_close(
 	n_opened = ihk_atomic_dec_return(&xpmem_my_part->n_opened);
 	XPMEM_DEBUG("n_opened=%d", n_opened);
 
-	index = xpmem_tg_hashtable_index(proc->pid);
-
-	mcs_rwlock_writer_lock(&xpmem_my_part->tg_hashtable[index].lock, &lock);
-
-	tg = xpmem_tg_ref_by_tgid_all_nolock(proc->pid);
-	if (IS_ERR(tg)) {
-		mcs_rwlock_writer_unlock(
-			&xpmem_my_part->tg_hashtable[index].lock, &lock);
-		return 0;
+	if (mckfd->data) {
+		/* release my xpmem-objects */
+		xpmem_flush(mckfd);
 	}
 
-	list_del_init(&tg->tg_hashlist);
-
-	mcs_rwlock_writer_unlock(&xpmem_my_part->tg_hashtable[index].lock, 
-		&lock);
-
-	XPMEM_DEBUG("tg->vm=0x%p", tg->vm);
-
-	ihk_mc_spinlock_lock_noirq(&tg->lock);
-	tg->flags |= XPMEM_FLAG_DESTROYING;
-	ihk_mc_spinlock_unlock_noirq(&tg->lock);
-
-	xpmem_release_aps_of_tg(tg);
-	xpmem_remove_segs_of_tg(tg);
-
-	ihk_mc_spinlock_lock_noirq(&tg->lock);
-	tg->flags |= XPMEM_FLAG_DESTROYED;
-	ihk_mc_spinlock_unlock_noirq(&tg->lock);
-
-	xpmem_destroy_tg(tg);
-
 	if (!n_opened) {
 		xpmem_exit();
 	}
@@ -333,6 +302,15 @@ static int xpmem_close(
 	return 0;
 }
 
+static int xpmem_dup(
+	struct mckfd *mckfd,
+	ihk_mc_user_context_t *ctx)
+{
+	mckfd->data = 0;
+	ihk_atomic_inc_return(&xpmem_my_part->n_opened);
+
+	return 0;
+}
 
 static int xpmem_init(void)
 {
@@ -987,6 +965,44 @@ static void xpmem_release_aps_of_tg(
 	XPMEM_DEBUG("return: ");
 }
 
+static void xpmem_flush(struct mckfd *mckfd)
+{
+	struct process *proc = (struct process *)mckfd->data;
+	struct xpmem_thread_group *tg;
+	int index;
+	struct mcs_rwlock_node_irqsave lock;
+
+	index = xpmem_tg_hashtable_index(proc->pid);
+
+	mcs_rwlock_writer_lock(&xpmem_my_part->tg_hashtable[index].lock, &lock);
+
+	tg = xpmem_tg_ref_by_tgid_all_nolock(proc->pid);
+	if (IS_ERR(tg)) {
+		mcs_rwlock_writer_unlock(
+			&xpmem_my_part->tg_hashtable[index].lock, &lock);
+		return;
+	}
+
+	list_del_init(&tg->tg_hashlist);
+
+	mcs_rwlock_writer_unlock(&xpmem_my_part->tg_hashtable[index].lock, 
+		&lock);
+
+	XPMEM_DEBUG("tg->vm=0x%p", tg->vm);
+
+	ihk_mc_spinlock_lock_noirq(&tg->lock);
+	tg->flags |= XPMEM_FLAG_DESTROYING;
+	ihk_mc_spinlock_unlock_noirq(&tg->lock);
+
+	xpmem_release_aps_of_tg(tg);
+	xpmem_remove_segs_of_tg(tg);
+
+	ihk_mc_spinlock_lock_noirq(&tg->lock);
+	tg->flags |= XPMEM_FLAG_DESTROYED;
+	ihk_mc_spinlock_unlock_noirq(&tg->lock);
+
+	xpmem_destroy_tg(tg);
+}
 
 static int xpmem_attach(
 	struct mckfd *mckfd,
@@ -1431,11 +1447,12 @@ static void xpmem_detach_att(
 		att->at_vaddr, att->at_vaddr + 1);
 
 	if (!range || range->start > att->at_vaddr) {
-		DBUG_ON(1);
+		ihk_mc_spinlock_lock_noirq(&ap->lock);
+		list_del_init(&att->att_list);
+		ihk_mc_spinlock_unlock_noirq(&ap->lock);
 		mcs_rwlock_writer_unlock(&att->at_lock, &at_lock);
 		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
-		ekprintf("%s: ERROR: lookup_process_memory_range() failed\n", 
-			__FUNCTION__);
+		xpmem_att_destroyable(att);
 		XPMEM_DEBUG("return: range=%p");
 		return;
 	}

kernel/zeroobj.c

@@ -169,7 +169,7 @@ out:
 }
 
 static int zeroobj_get_page(struct memobj *memobj, off_t off, int p2align,
-		uintptr_t *physp, unsigned long *pflag)
+		uintptr_t *physp, unsigned long *pflag, uintptr_t virt_addr)
 {
 	int error;
 	struct zeroobj *obj = to_zeroobj(memobj);

lib/include/ihk/mm.h

@@ -90,7 +90,7 @@ void ihk_mc_reserve_arch_pages(struct ihk_page_allocator_desc *pa_allocator,
 			unsigned long, unsigned long, int));
 
 struct ihk_mc_pa_ops {
-	void *(*alloc_page)(int, int, ihk_mc_ap_flag, int node, int is_user);
+	void *(*alloc_page)(int, int, ihk_mc_ap_flag, int node, int is_user, uintptr_t virt_addr);
 	void (*free_page)(void *, int, int is_user);
 
 	void *(*alloc)(int, ihk_mc_ap_flag);
@@ -115,33 +115,33 @@ int ihk_mc_free_micpa(unsigned long mic_pa);
 void ihk_mc_clean_micpa(void);
 
 void *_ihk_mc_alloc_aligned_pages_node(int npages, int p2align,
-	ihk_mc_ap_flag flag, int node, int is_user, char *file, int line);
+	ihk_mc_ap_flag flag, int node, int is_user, uintptr_t virt_addr, char *file, int line);
 #define ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, node) ({\
-void *r = _ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, node, IHK_MC_PG_KERNEL, __FILE__, __LINE__);\
+void *r = _ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, node, IHK_MC_PG_KERNEL, -1, __FILE__, __LINE__);\
 r;\
 })
-#define ihk_mc_alloc_aligned_pages_node_user(npages, p2align, flag, node) ({\
-void *r = _ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, node, IHK_MC_PG_USER, __FILE__, __LINE__);\
+#define ihk_mc_alloc_aligned_pages_node_user(npages, p2align, flag, node, virt_addr) ({\
+void *r = _ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, node, IHK_MC_PG_USER, virt_addr, __FILE__, __LINE__);\
 r;\
 })
 
 #define ihk_mc_alloc_aligned_pages(npages, p2align, flag) ({\
-void *r = _ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, -1, IHK_MC_PG_KERNEL, __FILE__, __LINE__);\
+void *r = _ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, -1, IHK_MC_PG_KERNEL, -1, __FILE__, __LINE__);\
 r;\
 })
 
-#define ihk_mc_alloc_aligned_pages_user(npages, p2align, flag) ({\
-void *r = _ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, -1, IHK_MC_PG_USER, __FILE__, __LINE__);\
+#define ihk_mc_alloc_aligned_pages_user(npages, p2align, flag, virt_addr) ({\
+void *r = _ihk_mc_alloc_aligned_pages_node(npages, p2align, flag, -1, IHK_MC_PG_USER, virt_addr, __FILE__, __LINE__);\
 r;\
 })
 
 #define ihk_mc_alloc_pages(npages, flag) ({\
-void *r = _ihk_mc_alloc_aligned_pages_node(npages, PAGE_P2ALIGN, flag, -1, IHK_MC_PG_KERNEL, __FILE__, __LINE__);\
+void *r = _ihk_mc_alloc_aligned_pages_node(npages, PAGE_P2ALIGN, flag, -1, IHK_MC_PG_KERNEL, -1, __FILE__, __LINE__);\
 r;\
 })
 
-#define ihk_mc_alloc_pages_user(npages, flag) ({\
-void *r = _ihk_mc_alloc_aligned_pages_node(npages, PAGE_P2ALIGN, flag, -1, IHK_MC_PG_USER, __FILE__, __LINE__);\
+#define ihk_mc_alloc_pages_user(npages, flag, virt_addr) ({\
+void *r = _ihk_mc_alloc_aligned_pages_node(npages, PAGE_P2ALIGN, flag, -1, IHK_MC_PG_USER, virt_addr, __FILE__, __LINE__);\
 r;\
 })
 

lib/string.c

@@ -280,6 +280,18 @@ int flatten_strings_from_user(int nr_strings, char *first, char **strings, char
 	long r;
 	int n, ret;
 
+	/* When strings is NULL, make array one NULL */
+	if (!strings) {
+		full_len = sizeof(long) + sizeof(char *);
+		_flat = kmalloc(full_len, IHK_MC_AP_NOWAIT);
+		if (!_flat) {
+			return -ENOMEM;
+		}
+		memset(_flat, 0, full_len);
+		*flat = (char *)_flat;
+		return full_len;
+	}
+
 	/* How many strings do we have? */
 	if (nr_strings == -1) {
 		nr_strings = 0;

test/dump/README

@@ -0,0 +1,23 @@
+===================
+Advance preparation
+===================
+1)Implement patch of destroy_mem.patch
+cd mckernel
+patch -p0 < destroy_mem.patch
+make
+make install
+
+
+2)Compile command execution processing
+cd mckernel/test/dump/mcexec_test_proc/
+make
+
+==========
+How to run
+==========
+#Run McKernel led dump test
+./go_mck_dump_test.sh
+
+#Run Linux led dump test
+./go_linux_dump_test.sh
+

test/error_handling/Issue1011/1011.patch

@@ -0,0 +1,13 @@
+diff --git a/executer/user/mcexec.c b/executer/user/mcexec.c
+index 2d0d50e..9f109aa 100644
+--- a/executer/user/mcexec.c
++++ b/executer/user/mcexec.c
+@@ -470,7 +470,7 @@ struct program_load_desc *load_interp(struct program_load_desc *desc0, FILE *fp)
+ 			j++;
+ 		}
+ 	}
+-	desc->num_sections = j;
++	desc->num_sections = -1; // Test for num_secionts check Issue#1011
+ 
+ 	desc->entry = hdr.e_entry;
+ 	desc->interp_align = align;

test/error_handling/Issue1011/README

@@ -0,0 +1,17 @@
+■ Issue#1011 動作確認
+1. ファイルの説明
+  1011.patch   mcexec からmcctrl.ko に渡すprogram_load_des構造体のnum_sections に
+               常に-1 を設定するパッチ
+
+2. 確認方法
+  1. 上記パッチを適用していない状態でMcKernelを起動する
+  2. mcexec hostname を実行し、ホスト名が表示されることを確認する
+  3. 上記パッチをMcKernelに適用後、ビルドと起動を行う
+  4. mcexec hostname を実行し、ホスト名が表示されず、"prepare: Invalid argument"が
+     コンソールに出力されることを確認する
+  5. /var/log/messages に"kernel: mcexec_prepare_image: ERROR: # of sections: -1"が
+     出力されていることを確認する
+
+3. 確認結果
+  上記確認方法で期待される動作が確認できたため、問題ない。
+

test/error_handling/Issue727/727.patch

@@ -0,0 +1,15 @@
+diff --git a/executer/user/mcexec.c b/executer/user/mcexec.c
+index 2d0d50e..70856e7 100644
+--- a/executer/user/mcexec.c
++++ b/executer/user/mcexec.c
+@@ -3732,7 +3732,9 @@ return_execve1:
+ 
+ 				/* Copy program image phase */
+ 				case 2:
+-					
++					fprintf(stderr, "execve killed\n");
++					fflush(stderr);
++					kill(getpid(), SIGKILL);
+ 					ret = -1;
+ 					/* Alloc descriptor */
+ 					desc = malloc(w.sr.args[2]);

test/error_handling/Issue727/README

@@ -0,0 +1,17 @@
+■ Issue#727 動作確認
+1. ファイルの説明
+727.patch     再現確認用パッチ mcexec において、execve phase 2 を SIGKILL
+              終了させる
+exec.c        確認用テストプログラム exec ls する
+patch-off.log パッチ非適用時の動作確認結果 ls の結果が表示されれば OK
+patch-on.log  パッチ適用時の動作確認結果 強制終了し、McKernel が PANIC して
+              いなければ OK
+
+2. 確認用テストプログラムのコンパイル方法
+gcc -o exec exec.c を行い、実行ファイル exec を得る
+
+3. 確認結果
+patch-on.log において、mcexec が強制終了し、McKernelのログに PANIC の表示が
+無いため、確認結果は OK。
+また、パッチ非適用時 (patch-off.log)、正常に ls の結果が表示されているので OK。
+以上より、確認結果は問題無い。

test/error_handling/Issue727/exec.c

@@ -0,0 +1,10 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <errno.h>
+int
+main(int argc, char **argv)
+{
+	execlp("ls", "ls", NULL);
+}

test/error_handling/Issue727/patch-off.log

@@ -0,0 +1,33 @@
+スクリプトは Wed Nov 29 11:22:32 2017
+ に開始しました[?1034hbash-4.2$ ../../../../mic/mcexec ./exec
+727.patch  exec  exec.c  patch-off.log
+bash-4.2$ ../../../../mic/ihkosctl 0kmsg kmsg
+IHK/McKernel started.
+[ -1]: no_execute_available: 1
+[ -1]: X86_IA32_NUM_PERF_COUNTERS: 4, X86_IA32_NUM_FIXED_PERF_COUNTERS: 3
+[ -1]: Invariant TSC supported.
+[ -1]: setup_x86 done.
+[ -1]: ns_per_tsc: 384
+[ -1]: KCommand Line: hidos  dump_level=24 
+[ -1]: Physical memory: 0x1002c7000 - 0x140000000, 1070829568 bytes, 261433 pages available @ NUMA: 0
+[ -1]: Physical memory: 0x880000000 - 0x8c0000000, 1073741824 bytes, 262144 pages available @ NUMA: 1
+[ -1]: NUMA: 0, Linux NUMA: 0, type: 1, available bytes: 1070829568, pages: 261433
+[ -1]: NUMA: 1, Linux NUMA: 1, type: 1, available bytes: 1073741824, pages: 262144
+[ -1]: NUMA 0 distances: 0 (10), 1 (21), 
+[ -1]: NUMA 1 distances: 1 (10), 0 (21), 
+[ -1]: map_fixed: phys: 0x90000 => 0xffffffff70015000 (2 pages)
+[ -1]: Trampoline area: 0x90000 
+[ -1]: map_fixed: phys: 0x0 => 0xffffffff70017000 (1 pages)
+[ -1]: # of cpus : 7
+[ -1]: locals = ffff8001002eb000
+[  0]: BSP: 0 (HW ID: 2 @ NUMA 0)
+[  0]: BSP: booted 6 AP CPUs
+[  0]: Master channel init acked.
+[  0]: vdso is enabled
+IHK/McKernel booted.
+
+bash-4.2$ シェルから脱出するには "exit" を使用してください。
+bash-4.2$ exit
+
+スクリプトは Wed Nov 29 11:22:57 2017
+ に終了しました

test/error_handling/Issue727/patch-on.log

@@ -0,0 +1,35 @@
+スクリプトは Wed Nov 29 11:25:01 2017
+ に開始しました[?1034hbash-4.2$ ../../../../mic/mcexec ./exec
+execve killed
+強制終了
+bash-4.2$ ../../../../mic/ihkosctl 0 kmsg
+IHK/McKernel started.
+[ -1]: no_execute_available: 1
+[ -1]: X86_IA32_NUM_PERF_COUNTERS: 4, X86_IA32_NUM_FIXED_PERF_COUNTERS: 3
+[ -1]: Invariant TSC supported.
+[ -1]: setup_x86 done.
+[ -1]: ns_per_tsc: 384
+[ -1]: KCommand Line: hidos  dump_level=24 
+[ -1]: Physical memory: 0x1002c7000 - 0x140000000, 1070829568 bytes, 261433 pages available @ NUMA: 0
+[ -1]: Physical memory: 0x880000000 - 0x8c0000000, 1073741824 bytes, 262144 pages available @ NUMA: 1
+[ -1]: NUMA: 0, Linux NUMA: 0, type: 1, available bytes: 1070829568, pages: 261433
+[ -1]: NUMA: 1, Linux NUMA: 1, type: 1, available bytes: 1073741824, pages: 262144
+[ -1]: NUMA 0 distances: 0 (10), 1 (21), 
+[ -1]: NUMA 1 distances: 1 (10), 0 (21), 
+[ -1]: map_fixed: phys: 0x90000 => 0xffffffff70015000 (2 pages)
+[ -1]: Trampoline area: 0x90000 
+[ -1]: map_fixed: phys: 0x0 => 0xffffffff70017000 (1 pages)
+[ -1]: # of cpus : 7
+[ -1]: locals = ffff8001002eb000
+[  0]: BSP: 0 (HW ID: 2 @ NUMA 0)
+[  0]: BSP: booted 6 AP CPUs
+[  0]: Master channel init acked.
+[  0]: vdso is enabled
+IHK/McKernel booted.
+[  0]: do_syscall: proxy PID 14276 is dead, terminate()
+
+bash-4.2$ シェルから脱出するには "exit" を使用してください。
+bash-4.2$ exit
+
+スクリプトは Wed Nov 29 11:25:33 2017
+ に終了しました

test/error_handling/Issue873/README

@@ -0,0 +1,12 @@
+■ Issue#873 動作確認
+1. ファイルの説明
+  mck_boot_test.sh   McKernelの停止→起動 を100回繰り返すスクリプト
+                     すべてで成功した場合には[OK]メッセージを、途中で失敗した場合には[NG] メッセージを出力する
+
+2. 確認方法
+  1. mck_boot_test.sh のMCK_DIR にMcKernelのインストールディレクトリを指定する
+  2. sh ./mck_boot_test.sh  を実行する
+  3. [OK]メッセージがコンソールに出力されることを確認する
+
+3. 確認結果
+  上記の確認方法で期待される動作が確認できたため、問題ない。

test/error_handling/Issue873/mck_boot_test.sh

@@ -0,0 +1,18 @@
+#!/bin/sh
+
+MCK_DIR=/home/satoken/ppos
+REP_NUM=100
+
+for i in `seq 1 ${REP_NUM}`
+do
+	sudo ${MCK_DIR}/sbin/mcstop+release.sh
+	sleep 1
+	sudo ${MCK_DIR}/sbin/mcreboot.sh
+
+	if [ $? -ne 0 ]; then
+		echo "[NG] failed to boot Mckernel  :${i}"
+		exit 1
+	fi
+done
+
+echo "[OK] succeed to boot McKernel ${i} times"

test/mbind/README

@@ -0,0 +1,19 @@
+===================
+Advance preparation
+===================
+1)Implement patch of test_trace_mem.patch
+cd mckernel
+patch -p0 < test_trace_mem.patch
+make
+make install
+
+
+2)Compile command execution processing
+cd mckernel/test/mbind/mcexec_test_proc/
+make
+
+==========
+How to run
+==========
+./go_mbind_test.sh
+

test/mbind/chk_mbind_result.sh

@@ -0,0 +1,58 @@
+#!/bin/sh
+
+DEFAULT_POLICY_KIND="<default policy>"
+#SHARED_POLICY_KIND="<default policy:Mapping of MAP_SHARED>"
+NUMA_NODE_POLICY_KIND="<NUMA node policy>"
+
+FILE_NAME=$1
+CHK_LOG_FILE="./result/${FILE_NAME}.log"
+
+source "./testcases/${FILE_NAME}.txt"
+CHK_POLICY_KIND=${POLICY_KIND}
+
+SET_MEM_POLICY=`grep "OK:set_mempolicy" $CHK_LOG_FILE | grep -o '(MPOL.*)'`
+SET_POLICY_NUM=`grep -c1 "OK:mbind" $CHK_LOG_FILE`
+
+for exec_num in `seq 0 $((SET_POLICY_NUM - 1))`
+do
+
+	if [ $exec_num -lt 10 ]; then
+		NUMA_NODE_ADDR=`grep "OK:mbind" $CHK_LOG_FILE | grep -e "0$exec_num]" | grep -o '(0x.*000)'`
+		NUMA_NODE_POLICY=`grep "OK:mbind" $CHK_LOG_FILE | grep -e "0$exec_num]" | grep -o '(MPOL.*)'`
+	else
+		NUMA_NODE_ADDR=`grep "OK:mbind" $CHK_LOG_FILE | grep -e "$exec_num]" | grep -o '(0x.*000)'`
+		NUMA_NODE_POLICY=`grep "OK:mbind" $CHK_LOG_FILE | grep -e "$exec_num]" | grep -o '(MPOL.*)'`
+	fi
+
+	if [ "$CHK_POLICY_KIND" = "$DEFAULT_POLICY_KIND" ]; then
+		SET_MEM_POLICY_NUM=`grep -v $NUMA_NODE_ADDR $CHK_LOG_FILE | grep -e "$CHK_POLICY_KIND" | grep -ce "$SET_MEM_POLICY"`
+		if [ $SET_MEM_POLICY_NUM -gt 0 ]; then
+			echo "OK:" $exec_num $CHK_POLICY_KIND" - not address" $NUMA_NODE_ADDR "test policy" $SET_MEM_POLICY "allocate num:" $SET_MEM_POLICY_NUM
+			exit 0
+		else
+			echo "NG:" $exec_num $CHK_POLICY_KIND" - not address" $NUMA_NODE_ADDR "test policy" $SET_MEM_POLICY "allocate num:" $SET_MEM_POLICY_NUM
+			exit 1
+		fi
+	fi
+
+	ALLOCATE_POLICY=`grep "mckernel_allocate_aligned_pages_node" $CHK_LOG_FILE | grep -e $NUMA_NODE_ADDR | grep -e "$CHK_POLICY_KIND" | grep -o '(MPOL.*)'`
+
+	if [ "$CHK_POLICY_KIND" = "$NUMA_NODE_POLICY_KIND" ]; then
+		if [ $NUMA_NODE_POLICY != $ALLOCATE_POLICY ]; then
+			echo "NG:" $exec_num $CHK_POLICY_KIND" - address" $NUMA_NODE_ADDR "test policy" $NUMA_NODE_POLICY "allocate policy" $ALLOCATE_POLICY
+			exit 1
+		else
+			echo "OK:" $exec_num $CHK_POLICY_KIND" - address" $NUMA_NODE_ADDR "test policy" $NUMA_NODE_POLICY "allocate policy" $ALLOCATE_POLICY
+		fi
+	else
+		if [ $SET_MEM_POLICY != $ALLOCATE_POLICY ]; then
+			echo "NG:" $exec_num $CHK_POLICY_KIND" - address" $NUMA_NODE_ADDR "test policy" $SET_MEM_POLICY "allocate policy" $ALLOCATE_POLICY
+			exit 1
+		else
+			echo "OK:" $exec_num $CHK_POLICY_KIND" - address" $NUMA_NODE_ADDR "test policy" $SET_MEM_POLICY "allocate policy" $ALLOCATE_POLICY
+		fi
+	fi
+done
+
+exit 0
+

test/mbind/config

@@ -0,0 +1,3 @@
+MCMOD_DIR=$HOME/ppos
+
+export MCMOD_DIR

test/mbind/go_mbind_test.sh

@@ -0,0 +1,26 @@
+#!/bin/sh
+START_NG_TEST_NO=0085
+
+for test_case in `ls -1 ./testcases/*.txt`
+do
+	case_name=`basename ${test_case} .txt`
+
+	logfile="./result/${case_name}.log"
+	./mbind_test.sh ${test_case} &> ${logfile}
+
+	if [ $? -eq 0 ]; then
+	        ./chk_mbind_result.sh ${case_name}
+        	if [ $? -eq 0 ]; then
+               		echo "[OK] ${case_name} is done."
+        	else
+                	echo "[NG] failed to test ${case_name}. Please check ${logfile}"
+        	fi
+	else
+		test_number=`basename ${test_case} _mbind.txt`
+		if [ $test_number -ge $START_NG_TEST_NO ]; then
+			echo "[OK] ${case_name} is done(NG test case)."
+		else
+			echo "[NG] failed to test ${case_name}. Please check ${logfile}"
+		fi
+	fi
+done

test/mbind/mbind_test.sh

@@ -0,0 +1,47 @@
+#!/bin/sh
+
+if [ $# -lt 1 ]; then
+        echo "Error: too few arguments."
+        echo "usage: `basename $0` <param_file>"
+fi
+
+# read config
+source ./config
+
+# read testcase param
+source $1
+
+# mcexec processのkill
+./utils/kill_mcexec.sh &> /dev/null
+
+# stop mckernel
+sudo ${MCMOD_DIR}/sbin/mcstop+release.sh
+
+sleep 1
+
+# boot mckernel
+echo "${MCMOD_DIR}/sbin/mcreboot.sh ${MCRBT_OPT%,}"
+sudo ${MCMOD_DIR}/sbin/mcreboot.sh ${MCRBT_OPT%,}
+
+sleep 1
+
+if [ ! -e "/dev/mcos0" ]; then
+	echo "Error: failed to mcreboot"
+	exit 1
+fi
+
+# exec mckernel test program
+echo "${MCMOD_DIR}/bin/mcexec ${USR_PROC}"
+${MCMOD_DIR}/bin/mcexec ${USR_PROC} 
+
+#if [ $? -eq 0 ]; then
+if [ $? == 0 ]; then
+	sleep 1
+
+	echo "${MCMOD_DIR}/sbin/ihkosctl ${OS_IDX} kmsg"
+	sudo ${MCMOD_DIR}/sbin/ihkosctl ${OS_IDX} kmsg
+	exit 0
+else
+        echo "Error: faild to mcexec"
+        exit 1
+fi

test/mbind/mcexec_test_proc/Makefile

@@ -0,0 +1,7 @@
+OBJS = exec_setmempolicy_mbind exec_setmempolicy_mbind_shared
+LDFLAGS = -lnuma
+
+all: $(OBJS)
+
+clean:
+	rm $(OBJS)

test/mbind/mcexec_test_proc/exec_setmempolicy_mbind.c

@@ -0,0 +1,208 @@
+#include <stdio.h>
+#include <assert.h>
+#include <sys/mman.h>
+#include <numa.h>
+#include <numaif.h>
+
+#define PAGE_SIZE	(4096)
+
+typedef struct func_setmem_para {
+	int set_mode;
+	int dummy;
+	unsigned long set_nodemask;
+	unsigned long set_maxnode;
+} set_mem_para;
+
+typedef struct func_mbind_para {
+	int set_mode;
+	int loop_cnt;
+	unsigned long set_nodemask;
+	unsigned long set_maxnode;
+	unsigned flags;
+} mbind_para;
+
+
+typedef struct func_para {
+	set_mem_para para1;
+	mbind_para   para2;
+} main_para;
+
+
+
+char *mempolicy [] = {
+	"MPOL_DEFAULT",
+	"MPOL_PREFERRED",
+	"MPOL_BIND",
+	"MPOL_INTERLEAVE"
+};
+
+int func_set_mempolicy(set_mem_para* inpara)
+{
+	int rst = -1;
+	int set_mode = inpara->set_mode;
+	unsigned long set_nodemask = inpara->set_nodemask;
+	unsigned long set_maxnode = inpara->set_maxnode;
+	int mode = set_mode & 0x00000003;
+
+	rst = set_mempolicy(set_mode, &set_nodemask, set_maxnode);
+
+	printf("-----\n");
+	if (rst < 0) {
+		printf("NG:set_mempolicy - mode:(%s) nodemask:0x%x maxnode:%d rst:%d\n"
+			,mempolicy[mode] ,set_nodemask ,set_maxnode, rst);
+		//assert(0 && "set_mempolicy() failed");
+	} else {
+		printf("OK:set_mempolicy - mode:(%s) nodemask:0x%x maxnode:%d\n"
+			,mempolicy[mode] ,set_nodemask ,set_maxnode);
+	}
+	printf("-----\n");
+
+	return rst;
+}
+
+int func_mbind(mbind_para* inpara)
+{
+	int rst = -1;
+	unsigned char *addr = NULL;
+	int get_mode = 0;
+	int i = 0;
+	unsigned long mem_len = PAGE_SIZE;
+
+	int set_mode = inpara->set_mode;
+	unsigned long set_nodemask = inpara->set_nodemask;
+	unsigned long set_maxnode = inpara->set_maxnode;
+	unsigned flags = inpara->flags;
+	int loop_cnt = inpara->loop_cnt;
+	int mode = set_mode & 0x00000003;
+
+	for (i = 0; i < loop_cnt; i++) {
+
+		addr = mmap(0, mem_len, (PROT_READ | PROT_WRITE),
+				(MAP_ANONYMOUS | MAP_PRIVATE), 0, 0);
+		if (addr == (void *) -1) {
+			printf("[%02d] NG:mmap - len:%d prot:0x%x flags:0x%x\n"
+				,i ,mem_len ,(PROT_READ | PROT_WRITE) ,(MAP_ANONYMOUS | MAP_PRIVATE));
+			//assert(0 && "mmap() failed");
+			return -1;
+		} else {
+//			printf("[%02d] OK:mmap - addr:(0x%016lx) len:%d prot:0x%x flags:0x%x\n"
+//				,i ,addr ,mem_len ,(PROT_READ | PROT_WRITE) ,(MAP_ANONYMOUS | MAP_PRIVATE));
+		}
+
+		if ((inpara->set_mode & 0x000000ff) == 0xff) {
+			switch ((i & 0x3)) {
+				case MPOL_PREFERRED:
+					set_mode = ((set_mode & 0xffffff00) | MPOL_PREFERRED);
+					set_nodemask = inpara->set_nodemask;
+					flags = 0;
+					mode = MPOL_PREFERRED;
+					break;
+
+				case MPOL_BIND:
+					set_mode = ((set_mode & 0xffffff00) | MPOL_BIND);
+					set_nodemask = inpara->set_nodemask;
+					flags = 0;
+					mode = MPOL_BIND;
+					break;
+
+				case MPOL_INTERLEAVE:
+					set_mode = ((set_mode & 0xffffff00) | MPOL_INTERLEAVE);
+					set_nodemask = inpara->set_nodemask;
+					flags = 0;
+					mode = MPOL_INTERLEAVE;
+					break;
+
+				case MPOL_DEFAULT:
+				default:
+					set_mode = ((set_mode & 0xffffff00) | MPOL_DEFAULT);
+					set_nodemask = 0;
+					flags = MPOL_MF_STRICT;
+					mode = MPOL_DEFAULT;
+					break;
+			}
+		}
+
+		rst = mbind(addr, mem_len, set_mode, &set_nodemask, set_maxnode, flags);
+		if (rst < 0) {
+			printf("[%02d] NG:mbind - addr:(0x%016lx) len:%d mode:(%s) nodemask:0x%x maxnode:%d flags:%d rst:%d\n"
+				,i ,addr ,mem_len ,mempolicy[mode] ,set_nodemask ,set_maxnode ,flags ,rst);
+			//assert(0 && "mbind() failed");
+			return -1;
+		} else {
+			printf("[%02d] OK:mbind - addr:(0x%016lx) len:%d mode:(%s) nodemask:0x%x maxnode:%d flags:%d\n"
+				,i ,addr ,mem_len ,mempolicy[mode] ,set_nodemask ,set_maxnode ,flags);
+		}
+
+		rst = get_mempolicy(&get_mode, NULL, 0, addr, MPOL_F_ADDR);
+		if(rst < 0) {
+			printf("[%02d] NG:get_mempolicy - addr:(0x%016lx) rst:%d\n"
+				,i ,addr , rst);
+			//assert(0 && "get_mempolicy failed");
+			return -1;
+		} else {
+			printf("[%02d] OK:get_mempolicy - addr:(0x%016lx) mode:(%s)\n"
+				,i ,addr ,mempolicy[get_mode]);
+		}
+
+		rst = munmap(addr, mem_len);
+		if (rst < 0) {
+			printf("[%02d] NG:munmap - addr:(0x%016lx) len:%d\n"
+				,i ,addr ,mem_len);
+		} else {
+//			printf("[%02d] OK:munmap - addr:(0x%016lx) len:%d\n"
+//				,i ,addr ,mem_len);
+		}
+
+		addr = mmap(addr, mem_len, (PROT_READ | PROT_WRITE),
+				(MAP_FIXED | MAP_ANONYMOUS | MAP_PRIVATE), 0, 0);
+		if (addr == (void *) -1) {
+			printf("[%02d] NG:mmap - len:%d prot:0x%x flags:0x%x\n"
+				,i ,mem_len ,(PROT_READ | PROT_WRITE) ,(MAP_FIXED | MAP_ANONYMOUS | MAP_PRIVATE));
+			//assert(0 && "mmap() failed");
+			return -1;
+		} else {
+//			printf("[%02d] OK:mmap - addr:(0x%016lx) len:%d prot:0x%x flags:0x%x\n"
+//				,i ,addr ,mem_len ,(PROT_READ | PROT_WRITE) ,(MAP_FIXED | MAP_ANONYMOUS | MAP_PRIVATE));
+		}
+		printf("-----\n");
+
+	}
+
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	main_para inpara;
+	int rst = -1;
+
+	if (argc == 9 ) {
+
+		inpara.para1.set_mode = strtol(argv[1], NULL, 16);
+		inpara.para1.set_nodemask = strtoul(argv[2], NULL, 16);
+		inpara.para1.set_maxnode = strtol(argv[3], NULL, 10);
+		rst = func_set_mempolicy(&inpara.para1);
+		if (rst == 0) {
+			inpara.para2.set_mode = strtol(argv[4], NULL, 16);
+			inpara.para2.set_nodemask =  strtoul(argv[5], NULL, 16);
+			inpara.para2.set_maxnode = strtoul(argv[6], NULL, 10);
+			inpara.para2.flags = strtoul(argv[7], NULL, 16);
+			inpara.para2.loop_cnt = strtol(argv[8], NULL, 10);
+			rst = func_mbind(&inpara.para2);
+		}
+	} else {
+		printf("NG: Invalid number of parameters(%d)\n",(argc-1));
+		printf("   parameter 1 : set_mempolicy(mode)\n");
+		printf("   parameter 2 : set_mempolicy(nodemask)\n");
+		printf("   parameter 3 : set_mempolicy(maxnode)\n");
+		printf("   parameter 4 : mbind(mode) 0xff - all mode\n");
+		printf("   parameter 5 : mbind(nodemask)\n");
+		printf("   parameter 6 : mbind(maxnode)\n");
+		printf("   parameter 7 : mbind(flags)\n");
+		printf("   parameter 8 : Number of mbind executed\n");
+		printf("   example) ./exec_setmempolicy_mbind 0x1 0x1 2 0x2 0x1 2 0x0 1\n");
+	}
+
+	return rst;
+}
+

test/mbind/mcexec_test_proc/exec_setmempolicy_mbind_shared.c

@@ -0,0 +1,208 @@
+#include <stdio.h>
+#include <assert.h>
+#include <sys/mman.h>
+#include <numa.h>
+#include <numaif.h>
+
+#define PAGE_SIZE	(4096)
+
+typedef struct func_setmem_para {
+	int set_mode;
+	int dummy;
+	unsigned long set_nodemask;
+	unsigned long set_maxnode;
+} set_mem_para;
+
+typedef struct func_mbind_para {
+	int set_mode;
+	int loop_cnt;
+	unsigned long set_nodemask;
+	unsigned long set_maxnode;
+	unsigned flags;
+} mbind_para;
+
+
+typedef struct func_para {
+	set_mem_para para1;
+	mbind_para   para2;
+} main_para;
+
+
+
+char *mempolicy [] = {
+	"MPOL_DEFAULT",
+	"MPOL_PREFERRED",
+	"MPOL_BIND",
+	"MPOL_INTERLEAVE"
+};
+
+int func_set_mempolicy(set_mem_para* inpara)
+{
+	int rst = -1;
+	int set_mode = inpara->set_mode;
+	unsigned long set_nodemask = inpara->set_nodemask;
+	unsigned long set_maxnode = inpara->set_maxnode;
+	int mode = set_mode & 0x00000003;
+
+	rst = set_mempolicy(set_mode, &set_nodemask, set_maxnode);
+
+	printf("-----\n");
+	if (rst < 0) {
+		printf("NG:set_mempolicy - mode:(%s) nodemask:0x%x maxnode:%d rst:%d\n"
+			,mempolicy[mode] ,set_nodemask ,set_maxnode, rst);
+		//assert(0 && "set_mempolicy() failed");
+	} else {
+		printf("OK:set_mempolicy - mode:(%s) nodemask:0x%x maxnode:%d\n"
+			,mempolicy[mode] ,set_nodemask ,set_maxnode);
+	}
+	printf("-----\n");
+
+	return rst;
+}
+
+int func_mbind(mbind_para* inpara)
+{
+	int rst = -1;
+	unsigned char *addr = NULL;
+	int get_mode = 0;
+	int i = 0;
+	unsigned long mem_len = PAGE_SIZE;
+
+	int set_mode = inpara->set_mode;
+	unsigned long set_nodemask = inpara->set_nodemask;
+	unsigned long set_maxnode = inpara->set_maxnode;
+	unsigned flags = inpara->flags;
+	int loop_cnt = inpara->loop_cnt;
+	int mode = set_mode & 0x00000003;
+
+	for (i = 0; i < loop_cnt; i++) {
+
+		addr = mmap(0, mem_len, (PROT_READ | PROT_WRITE),
+				(MAP_ANONYMOUS | MAP_SHARED | MAP_POPULATE), 0, 0);
+		if (addr == (void *) -1) {
+			printf("[%02d] NG:mmap - len:%d prot:0x%x flags:0x%x\n"
+				,i ,mem_len ,(PROT_READ | PROT_WRITE) ,(MAP_ANONYMOUS | MAP_SHARED | MAP_POPULATE));
+			//assert(0 && "mmap() failed");
+			return -1;
+		} else {
+//			printf("[%02d] OK:mmap - addr:(0x%016lx) len:%d prot:0x%x flags:0x%x\n"
+//				,i ,addr ,mem_len ,(PROT_READ | PROT_WRITE) ,(MAP_ANONYMOUS | MAP_SHARED | MAP_POPULATE));
+		}
+
+		if ((inpara->set_mode & 0x000000ff) == 0xff) {
+			switch ((i & 0x3)) {
+				case MPOL_PREFERRED:
+					set_mode = ((set_mode & 0xffffff00) | MPOL_PREFERRED);
+					set_nodemask = inpara->set_nodemask;
+					flags = 0;
+					mode = MPOL_PREFERRED;
+					break;
+
+				case MPOL_BIND:
+					set_mode = ((set_mode & 0xffffff00) | MPOL_BIND);
+					set_nodemask = inpara->set_nodemask;
+					flags = 0;
+					mode = MPOL_BIND;
+					break;
+
+				case MPOL_INTERLEAVE:
+					set_mode = ((set_mode & 0xffffff00) | MPOL_INTERLEAVE);
+					set_nodemask = inpara->set_nodemask;
+					flags = 0;
+					mode = MPOL_INTERLEAVE;
+					break;
+
+				case MPOL_DEFAULT:
+				default:
+					set_mode = ((set_mode & 0xffffff00) | MPOL_DEFAULT);
+					set_nodemask = 0;
+					flags = MPOL_MF_STRICT;
+					mode = MPOL_DEFAULT;
+					break;
+			}
+		}
+
+		rst = mbind(addr, mem_len, set_mode, &set_nodemask, set_maxnode, flags);
+		if (rst < 0) {
+			printf("[%02d] NG:mbind - addr:(0x%016lx) len:%d mode:(%s) nodemask:0x%x maxnode:%d flags:%d rst:%d\n"
+				,i ,addr ,mem_len ,mempolicy[mode] ,set_nodemask ,set_maxnode ,flags ,rst);
+			//assert(0 && "mbind() failed");
+			return -1;
+		} else {
+			printf("[%02d] OK:mbind - addr:(0x%016lx) len:%d mode:(%s) nodemask:0x%x maxnode:%d flags:%d\n"
+				,i ,addr ,mem_len ,mempolicy[mode] ,set_nodemask ,set_maxnode ,flags);
+		}
+
+		rst = get_mempolicy(&get_mode, NULL, 0, addr, MPOL_F_ADDR);
+		if(rst < 0) {
+			printf("[%02d] NG:get_mempolicy - addr:(0x%016lx) rst:%d\n"
+				,i ,addr , rst);
+			//assert(0 && "get_mempolicy failed");
+			return -1;
+		} else {
+			printf("[%02d] OK:get_mempolicy - addr:(0x%016lx) mode:(%s)\n"
+				,i ,addr ,mempolicy[get_mode]);
+		}
+
+		rst = munmap(addr, mem_len);
+		if (rst < 0) {
+			printf("[%02d] NG:munmap - addr:(0x%016lx) len:%d\n"
+				,i ,addr ,mem_len);
+		} else {
+//			printf("[%02d] OK:munmap - addr:(0x%016lx) len:%d\n"
+//				,i ,addr ,mem_len);
+		}
+
+		addr = mmap(addr, mem_len, (PROT_READ | PROT_WRITE),
+				(MAP_FIXED | MAP_ANONYMOUS | MAP_SHARED | MAP_POPULATE), 0, 0);
+		if (addr == (void *) -1) {
+			printf("[%02d] NG:mmap - len:%d prot:0x%x flags:0x%x\n"
+				,i ,mem_len ,(PROT_READ | PROT_WRITE) ,(MAP_FIXED | MAP_ANONYMOUS | MAP_SHARED | MAP_POPULATE));
+			//assert(0 && "mmap() failed");
+			return -1;
+		} else {
+//			printf("[%02d] OK:mmap - addr:(0x%016lx) len:%d prot:0x%x flags:0x%x\n"
+//				,i ,addr ,mem_len ,(PROT_READ | PROT_WRITE) ,(MAP_FIXED | MAP_ANONYMOUS | MAP_SHARED | MAP_POPULATE));
+		}
+		printf("-----\n");
+
+	}
+
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	main_para inpara;
+	int rst = -1;
+
+	if (argc == 9 ) {
+
+		inpara.para1.set_mode = strtol(argv[1], NULL, 16);
+		inpara.para1.set_nodemask = strtoul(argv[2], NULL, 16);
+		inpara.para1.set_maxnode = strtol(argv[3], NULL, 10);
+		rst = func_set_mempolicy(&inpara.para1);
+		if (rst == 0) {
+			inpara.para2.set_mode = strtol(argv[4], NULL, 16);
+			inpara.para2.set_nodemask =  strtoul(argv[5], NULL, 16);
+			inpara.para2.set_maxnode = strtoul(argv[6], NULL, 10);
+			inpara.para2.flags = strtoul(argv[7], NULL, 16);
+			inpara.para2.loop_cnt = strtol(argv[8], NULL, 10);
+			rst = func_mbind(&inpara.para2);
+		}
+	} else {
+		printf("NG: Invalid number of parameters(%d)\n",(argc-1));
+		printf("   parameter 1 : set_mempolicy(mode)\n");
+		printf("   parameter 2 : set_mempolicy(nodemask)\n");
+		printf("   parameter 3 : set_mempolicy(maxnode)\n");
+		printf("   parameter 4 : mbind(mode) 0xff - all mode\n");
+		printf("   parameter 5 : mbind(nodemask)\n");
+		printf("   parameter 6 : mbind(maxnode)\n");
+		printf("   parameter 7 : mbind(flags)\n");
+		printf("   parameter 8 : Number of mbind executed\n");
+		printf("   example) ./exec_setmempolicy_mbind 0x1 0x1 2 0x2 0x1 2 0x0 1\n");
+	}
+
+	return rst;
+}
+

test/mbind/test_trace_mem.patch

@@ -0,0 +1,75 @@
+diff --git kernel/mem.c kernel/mem.c
+index 62cb206..5bfb6d6 100644
+--- kernel/mem.c
++++ kernel/mem.c
+@@ -542,6 +542,15 @@ static void reserve_pages(struct ihk_page_allocator_desc *pa_allocator,
+ 	ihk_pagealloc_reserve(pa_allocator, start, end);
+ }
+ 
++#if 1   /* Trace for DEBUG */
++char *mempolicy [] = {
++	"MPOL_DEFAULT",
++	"MPOL_PREFERRED",
++	"MPOL_BIND",
++	"MPOL_INTERLEAVE"
++};
++#endif
++
+ extern int cpu_local_var_initialized;
+ static void *mckernel_allocate_aligned_pages_node(int npages, int p2align,
+ 		ihk_mc_ap_flag flag, int pref_node, int is_user, uintptr_t virt_addr)
+@@ -585,6 +594,23 @@ static void *mckernel_allocate_aligned_pages_node(int npages, int p2align,
+ 			}
+ 		}
+ 
++#if 1	/* Trace for DEBUG */
++		if (!((range_policy_iter) && (range_policy_iter->numa_mem_policy != MPOL_DEFAULT))) {
++			if ((range_policy_iter) && (range_policy_iter->numa_mem_policy == MPOL_DEFAULT)) {
++				if (chk_shm == 0) {
++					kprintf("%s[%d]: addr(0x%016lx) policy(%s) <NUMA node policy>\n"
++						,__FUNCTION__ ,__LINE__ ,virt_addr ,mempolicy[(range_policy_iter->numa_mem_policy & 0x3)]);
++				}
++			} else {
++				if ((cpu_local_var(current)->vm->numa_mem_policy == MPOL_DEFAULT) && (virt_addr != -1)) {
++					if (virt_addr) {
++						kprintf("%s[%d]: addr(0x%016lx) policy(%s) <default policy>\n"
++							,__FUNCTION__ ,__LINE__ ,virt_addr ,mempolicy[(cpu_local_var(current)->vm->numa_mem_policy & 0x3)]);
++					}
++				}
++			}
++		}
++#endif
+ 
+ 		if ((!((range_policy_iter) && (range_policy_iter->numa_mem_policy != MPOL_DEFAULT))) && (chk_shm == 0))
+ 			goto distance_based;
+@@ -647,10 +673,30 @@ static void *mckernel_allocate_aligned_pages_node(int npages, int p2align,
+ 				chk_shm = 1;
+ 			} else {
+ 				numa_mem_policy = range_policy_iter->numa_mem_policy;
++
++#if 1	/* Trace for DEBUG */
++				kprintf("%s[%d]: addr(0x%016lx) policy(%s) <NUMA node policy>\n"
++					,__FUNCTION__ ,__LINE__ ,virt_addr ,mempolicy[(numa_mem_policy & 0x3)]);
++#endif
++
+ 			}
+ 		}
+ 	}
+ 
++#if 1	/* Trace for DEBUG */
++	if (numa_mem_policy == -1) {
++		if (chk_shm == 1) {
++			kprintf("%s[%d]: addr(0x%016lx) policy(%s) <default policy:Mapping of MAP_SHARED>\n"
++				,__FUNCTION__ ,__LINE__ ,virt_addr ,mempolicy[(cpu_local_var(current)->vm->numa_mem_policy & 0x3)]);
++		} else {
++			if (virt_addr) {
++				kprintf("%s[%d]: addr(0x%016lx) policy(%s) <default policy>\n"
++					,__FUNCTION__ ,__LINE__ ,virt_addr ,mempolicy[(cpu_local_var(current)->vm->numa_mem_policy & 0x3)]);
++			}
++		}
++	}
++#endif
++
+ 	if (numa_mem_policy == -1)
+ 		numa_mem_policy = cpu_local_var(current)->vm->numa_mem_policy;
+ 

test/mbind/testcases/0001_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 1 0x8000 0x0 1 1 1"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0002_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 1 0x8001 0x1 1 0 1"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0003_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 1 0x8002 0x1 1 0 1"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0004_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 1 0x8003 0x1 1 0 1"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0005_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 1 0x80ff 0x1 1 0 4"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0006_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 1 0x80ff 0x1 1 0 40"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0007_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 1 0x8000 0x0 1 1 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0008_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8001 0x1 1 0x8000 0x0 1 1 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0009_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8002 0x1 1 0x8000 0x0 1 1 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0010_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8003 0x1 1 0x8000 0x0 1 1 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0011_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind_shared 0x8000 0x0 1 0x8003 0x1 1 0 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0012_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind_shared 0x8001 0x1 1 0x8002 0x1 1 0 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0013_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind_shared 0x8002 0x1 1 0x8001 0x1 1 0 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0014_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind_shared 0x8003 0x1 1 0x8000 0x0 1 0 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0015_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 2 0x8000 0x0 2 1 1"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0016_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 2 0x8001 0x3 2 0 1"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0017_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 2 0x8002 0x3 2 0 1"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0018_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 2 0x8003 0x3 2 0 1"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0019_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 2 0x80ff 0x3 2 0 4"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0020_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 2 0x80ff 0x3 2 0 40"
+OS_IDX=0
+POLICY_KIND="<NUMA node policy>"
+#POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0021_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8000 0x0 2 0x8000 0x0 2 1 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0022_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8001 0x3 2 0x8000 0x0 2 1 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0023_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8002 0x3 2 0x8000 0x0 2 1 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"

test/mbind/testcases/0024_mbind.txt

@@ -0,0 +1,6 @@
+MCRBT_OPT="-m `./utils/gen_mem_chunks.sh "0 1" 32M 1`"
+USR_PROC="mcexec_test_proc/exec_setmempolicy_mbind 0x8003 0x3 2 0x8000 0x0 2 1 1"
+OS_IDX=0
+#POLICY_KIND="<NUMA node policy>"
+POLICY_KIND="<default policy>"
+#POLICY_KIND="<default policy:Mapping of MAP_SHARED>"