spec: prerelase 0.6 for testing cpuinfo and mmap overcommit

Change-Id: Iab5acc2c08ebe19251c37782cff87a4b5c914448

CMakeLists.txt

@@ -10,7 +10,7 @@ project(mckernel C ASM)
 set(MCKERNEL_VERSION "1.7.0")
 
 # See "Fedora Packaging Guidlines -- Versioning"
-set(MCKERNEL_RELEASE "0.4")
+set(MCKERNEL_RELEASE "0.6")
 
 set(CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/cmake/modules)
 # for rpmbuild

NEWS.md

@@ -535,3 +535,6 @@ Restrictions on McKernel
 
 25. The behaviors of madvise and mbind are changed to do nothing and
     report success as a workaround for Fugaku.
+
+26. mmap() allows unlimited overcommit. Note that it corresponds to
+    setting sysctl ``vm.overcommit_memory`` to 1.

arch/arm64/kernel/cpu.c

@@ -1279,7 +1279,7 @@ long ihk_mc_show_cpuinfo(char *buf, size_t buf_size, unsigned long read_off, int
 
 		/* generate strings */
 		loff += scnprintf(lbuf + loff, lbuf_size - loff,
-				  "processor\t: %d\n", cpuinfo->hwid);
+				  "processor\t: %d\n", i);
 		loff += scnprintf(lbuf + loff, lbuf_size - loff, "Features\t:");
 
 		for (j = 0; hwcap_str[j]; j++) {

arch/arm64/kernel/include/syscall_list.h

@@ -83,6 +83,7 @@ SYSCALL_HANDLED(175, geteuid)
 SYSCALL_HANDLED(176, getgid)
 SYSCALL_HANDLED(177, getegid)
 SYSCALL_HANDLED(178, gettid)
+SYSCALL_HANDLED(179, sysinfo)
 SYSCALL_DELEGATED(188, msgrcv)
 SYSCALL_DELEGATED(189, msgsnd)
 SYSCALL_DELEGATED(192, semtimedop)
@@ -143,3 +144,8 @@ SYSCALL_DELEGATED(1049, stat)
 SYSCALL_DELEGATED(1060, getpgrp)
 SYSCALL_HANDLED(1062, time)
 SYSCALL_DELEGATED(1069, epoll_wait)
+
+/* Do not edit the lines including this comment and
+ * EOF just after it because those are used as a
+ * robust marker for the autotest patch.
+ */

arch/arm64/kernel/syscall.c

@@ -142,17 +142,6 @@ int obtain_clone_cpuid(cpu_set_t *cpu_set, int use_last)
 	return min_cpu;
 }
 
-int
-arch_clear_host_user_space()
-{
-	struct thread *th = cpu_local_var(current);
-
-	/* XXX: might be unnecessary */
-	clear_host_pte(th->vm->region.user_start,
-			(th->vm->region.user_end - th->vm->region.user_start), 0);
-	return 0;
-}
-
 /* archtecture-depended syscall handlers */
 extern unsigned long do_fork(int clone_flags, unsigned long newsp,
 			     unsigned long parent_tidptr, unsigned long child_tidptr,
@@ -185,33 +174,6 @@ SYSCALL_DECLARE(clone)
 	return ret;
 }
 
-SYSCALL_DECLARE(rt_sigaction)
-{
-	int sig = ihk_mc_syscall_arg0(ctx);
-	const struct sigaction *act = (const struct sigaction *)ihk_mc_syscall_arg1(ctx);
-	struct sigaction *oact = (struct sigaction *)ihk_mc_syscall_arg2(ctx);
-	size_t sigsetsize = ihk_mc_syscall_arg3(ctx);
-	struct k_sigaction new_sa, old_sa;
-	int rc;
-
-	if (sigsetsize != sizeof(sigset_t))
-		return -EINVAL;
-
-	if(act)
-		if(copy_from_user(&new_sa.sa, act, sizeof new_sa.sa)){
-			goto fault;
-		}
-	rc = do_sigaction(sig, act? &new_sa: NULL, oact? &old_sa: NULL);
-	if(rc == 0 && oact)
-		if(copy_to_user(oact, &old_sa.sa, sizeof old_sa.sa)){
-			goto fault;
-		}
-
-	return rc;
-fault:
-	return -EFAULT;
-}
-
 SYSCALL_DECLARE(prctl)
 {
 	struct process *proc = cpu_local_var(current)->proc;
@@ -1368,70 +1330,6 @@ out:
 	return restart;
 }
 
-static struct sig_pending *
-getsigpending(struct thread *thread, int delflag){
-	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;
-
-	w = thread->sigmask.__val[0];
-
-	lock = &thread->sigcommon->lock;
-	head = &thread->sigcommon->sigpending;
-	for(;;) {
-		if (delflag) {
-			mcs_rwlock_writer_lock(lock, &mcs_rw_node);
-		}
-		else {
-			mcs_rwlock_reader_lock(lock, &mcs_rw_node);
-		}
-
-		list_for_each_entry_safe(pending, next, head, list){
-			if(!(pending->sigmask.__val[0] & w)){
-				if(delflag) 
-					list_del(&pending->list);
-
-				if (delflag) {
-					mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
-				}
-				else {
-					mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
-				}
-				return pending;
-			}
-		}
-
-		if (delflag) {
-			mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
-		}
-		else {
-			mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
-		}
-
-		if(lock == &thread->sigpendinglock)
-			return NULL;
-
-		lock = &thread->sigpendinglock;
-		head = &thread->sigpending;
-	}
-
-	return NULL;
-}
-
-struct sig_pending *
-hassigpending(struct thread *thread)
-{
-	if (list_empty(&thread->sigpending) &&
-	    list_empty(&thread->sigcommon->sigpending)) {
-		return NULL;
-	}
-
-	return getsigpending(thread, 0);
-}
-
 int
 interrupt_from_user(void *regs0)
 {
@@ -1455,191 +1353,6 @@ void save_syscall_return_value(int num, unsigned long rc)
 	}
 }
 
-void
-check_signal(unsigned long rc, void *regs0, int num)
-{
-	__check_signal(rc, regs0, num, 0);
-}
-
-void
-check_signal_irq_disabled(unsigned long rc, void *regs0, int num)
-{
-	__check_signal(rc, regs0, num, 1);
-}
-
-static void
-__check_signal(unsigned long rc, void *regs0, int num, int irq_disabled)
-{
-	ihk_mc_user_context_t *regs = regs0;
-	struct thread *thread;
-	struct sig_pending *pending;
-	int	irqstate;
-
-	if(clv == NULL)
-		return;
-	thread = cpu_local_var(current);
-
-	if(thread == NULL || thread->proc->pid == 0){
-		struct thread *t;
-
-		irqstate = cpu_disable_interrupt_save();
-		ihk_mc_spinlock_lock_noirq(&(cpu_local_var(runq_lock)));
-		list_for_each_entry(t, &(cpu_local_var(runq)), sched_list){
-			if(t->proc->pid <= 0)
-				continue;
-			if(t->status == PS_INTERRUPTIBLE &&
-			   hassigpending(t)){
-				t->status = PS_RUNNING;
-				break;
-			}
-		}
-		ihk_mc_spinlock_unlock_noirq(&(cpu_local_var(runq_lock)));
-		cpu_restore_interrupt(irqstate);
-		goto out;
-	}
-
-	if(regs != NULL && !interrupt_from_user(regs)) {
-		goto out;
-	}
-
-	if (list_empty(&thread->sigpending) &&
-	    list_empty(&thread->sigcommon->sigpending)) {
-		goto out;
-	}
-
-	for(;;){
-		/* When this function called from check_signal_irq_disabled,
-		 * return with interrupt invalid.
-		 * This is to eliminate signal loss.
-		 */
-		if (irq_disabled == 1) {
-			irqstate = cpu_disable_interrupt_save();
-		}
-		pending = getsigpending(thread, 1);
-		if(!pending) {
-			dkprintf("check_signal,queue is empty\n");
-			goto out;
-		}
-		if (irq_disabled == 1) {
-			cpu_restore_interrupt(irqstate);
-		}
-		if (do_signal(rc, regs, thread, pending, num)) {
-			num = -1;
-		}
-	}
-
-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 &&
-			     sig != SIGCONT) ||
-			    (k->sa.sa_handler != SIG_IGN &&
-			     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 &&
-						    sig != SIGCONT &&
-						    !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(void)
-{
-	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->group_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)
 {

arch/x86_64/kernel/include/syscall_list.h

@@ -74,6 +74,7 @@ SYSCALL_DELEGATED(89, readlink)
 SYSCALL_HANDLED(96, gettimeofday)
 SYSCALL_HANDLED(97, getrlimit)
 SYSCALL_HANDLED(98, getrusage)
+SYSCALL_HANDLED(99, sysinfo)
 SYSCALL_HANDLED(100, times)
 SYSCALL_HANDLED(101, ptrace)
 SYSCALL_HANDLED(102, getuid)
@@ -179,4 +180,8 @@ SYSCALL_HANDLED(802, linux_mlock)
 SYSCALL_HANDLED(803, suspend_threads)
 SYSCALL_HANDLED(804, resume_threads)
 SYSCALL_HANDLED(811, linux_spawn)
-/**** End of File ****/
+
+/* Do not edit the lines including this comment and
+ * EOF just after it because those are used as a
+ * robust marker for the autotest patch.
+ */

arch/x86_64/kernel/syscall.c

@@ -147,44 +147,6 @@ int obtain_clone_cpuid(cpu_set_t *cpu_set, int use_last) {
     return min_cpu;
 }
 
-int
-arch_clear_host_user_space()
-{
-	struct thread *th = cpu_local_var(current);
-
-	/* XXX: might be unnecessary */
-	clear_host_pte(th->vm->region.user_start,
-			(th->vm->region.user_end - th->vm->region.user_start), 0);
-	return 0;
-}
-
-SYSCALL_DECLARE(rt_sigaction)
-{
-	int sig = ihk_mc_syscall_arg0(ctx);
-	const struct sigaction *act = (const struct sigaction *)ihk_mc_syscall_arg1(ctx);
-	struct sigaction *oact = (struct sigaction *)ihk_mc_syscall_arg2(ctx);
-	size_t sigsetsize = ihk_mc_syscall_arg3(ctx);
-	struct k_sigaction new_sa, old_sa;
-	int rc;
-
-	if (sigsetsize != sizeof(sigset_t))
-		return -EINVAL;
-
-	if(act)
-		if(copy_from_user(&new_sa.sa, act, sizeof new_sa.sa)){
-			goto fault;
-		}
-	rc = do_sigaction(sig, act? &new_sa: NULL, oact? &old_sa: NULL);
-	if(rc == 0 && oact)
-		if(copy_to_user(oact, &old_sa.sa, sizeof old_sa.sa)){
-			goto fault;
-		}
-
-	return rc;
-fault:
-	return -EFAULT;
-}
-
 SYSCALL_DECLARE(prctl)
 {
 	struct process *proc = cpu_local_var(current)->proc;
@@ -1039,82 +1001,6 @@ out:
 	return restart;
 }
 
-static struct sig_pending *
-getsigpending(struct thread *thread, int delflag){
-	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;
-	struct k_sigaction *k;
-
-	w = thread->sigmask.__val[0];
-
-	lock = &thread->sigcommon->lock;
-	head = &thread->sigcommon->sigpending;
-	for(;;) {
-		if (delflag) {
-			mcs_rwlock_writer_lock(lock, &mcs_rw_node);
-		}
-		else {
-			mcs_rwlock_reader_lock(lock, &mcs_rw_node);
-		}
-
-		list_for_each_entry_safe(pending, next, head, list){
-			for(x = pending->sigmask.__val[0], sig = 0; x; sig++, x >>= 1);
-			k = thread->sigcommon->action + sig - 1;
-			if(delflag ||
-			   (sig != SIGCHLD &&
-			    sig != SIGURG &&
-			    sig != SIGCONT) ||
-			   (k->sa.sa_handler != (void *)1 &&
-			    k->sa.sa_handler != NULL)){
-				if(!(pending->sigmask.__val[0] & w)){
-					if(delflag) 
-						list_del(&pending->list);
-
-					if (delflag) {
-						mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
-					}
-					else {
-						mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
-					}
-					return pending;
-				}
-			}
-		}
-
-		if (delflag) {
-			mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
-		}
-		else {
-			mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
-		}
-
-		if(lock == &thread->sigpendinglock)
-			return NULL;
-
-		lock = &thread->sigpendinglock;
-		head = &thread->sigpending;
-	}
-
-	return NULL;
-}
-
-struct sig_pending *
-hassigpending(struct thread *thread)
-{
-	if (list_empty(&thread->sigpending) &&
-	    list_empty(&thread->sigcommon->sigpending)) {
-		return NULL;
-	}
-
-	return getsigpending(thread, 0);
-}
-
 int
 interrupt_from_user(void *regs0)
 {
@@ -1129,175 +1015,6 @@ void save_syscall_return_value(int num, unsigned long rc)
 	return;
 }
 
-/** \brief check arrived signals and processing
- *
- * @param rc    return value of syscall
- * @param regs0 context
- * @param num   syscall number (-1: Not called on exiting system call)
- */
-void
-check_signal(unsigned long rc, void *regs0, int num)
-{
-	struct x86_user_context *regs = regs0;
-	struct thread *thread;
-	struct sig_pending *pending;
-	int	irqstate;
-
-	if(clv == NULL)
-		return;
-	thread = cpu_local_var(current);
-
-	if(thread == NULL || thread == &cpu_local_var(idle)){
-		struct thread *t;
-
-		irqstate = cpu_disable_interrupt_save();
-		ihk_mc_spinlock_lock_noirq(&(cpu_local_var(runq_lock)));
-		list_for_each_entry(t, &(cpu_local_var(runq)), sched_list){
-			if(t == &cpu_local_var(idle))
-				continue;
-			if(t->status == PS_INTERRUPTIBLE &&
-			   hassigpending(t)){
-				t->status = PS_RUNNING;
-				break;
-			}
-		}
-		ihk_mc_spinlock_unlock_noirq(&(cpu_local_var(runq_lock)));
-		cpu_restore_interrupt(irqstate);
-		goto out;
-	}
-
-	if(regs != NULL && !interrupt_from_user(regs)) {
-		goto out;
-	}
-
-	if (list_empty(&thread->sigpending) &&
-	    list_empty(&thread->sigcommon->sigpending)) {
-		goto out;
-	}
-
-	for(;;){
-		pending = getsigpending(thread, 1);
-		if(!pending) {
-			dkprintf("check_signal,queue is empty\n");
-			goto out;
-		}
-
-		if (do_signal(rc, regs, thread, pending, num)) {
-			num = -1;
-		}
-	}
-
-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 &&
-			     sig != SIGCONT) ||
-			    (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 &&
-						    sig != SIGCONT &&
-						    !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(void)
-{
-	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->group_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)

executer/include/uprotocol.h

@@ -91,7 +91,10 @@ struct program_image_section {
 
 struct get_cpu_set_arg {
 	int nr_processes;
+	char *req_cpu_list;   // Requested by user-space
+	int req_cpu_list_len; // Lenght of request string
 	int *process_rank;
+	pid_t ppid;
 	void *cpu_set;
 	size_t cpu_set_size;	// Size in bytes
 	int *target_core;

executer/kernel/mcctrl/control.c

@@ -587,13 +587,14 @@ extern int mckernel_cpu_2_linux_cpu(struct mcctrl_usrdata *udp, int cpu_id);
 static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
 {
 	struct mcctrl_usrdata *udp = ihk_host_os_get_usrdata(os);
-	struct mcctrl_part_exec *pe;
+	struct mcctrl_part_exec *pe = NULL, *pe_itr;
 	struct get_cpu_set_arg req;
 	struct mcctrl_cpu_topology *cpu_top, *cpu_top_i;
 	struct cache_topology *cache_top;
 	int cpu, cpus_assigned, cpus_to_assign, cpu_prev;
 	int ret = 0;
 	int mcexec_linux_numa;
+	int pe_list_len = 0;
 	cpumask_t *mcexec_cpu_set = NULL;
 	cpumask_t *cpus_used = NULL;
 	cpumask_t *cpus_to_use = NULL;
@@ -613,24 +614,126 @@ static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
 		return -EINVAL;
 	}
 
-	pe = &udp->part_exec;
-
-	mutex_lock(&pe->lock);
-
 	if (copy_from_user(&req, (void *)arg, sizeof(req))) {
-		printk("%s: error copying user request\n", __FUNCTION__);
+		pr_err("%s: error copying user request\n", __func__);
 		ret = -EINVAL;
-		goto put_and_unlock_out;
+		goto put_out;
 	}
 
-	/* First process to enter CPU partitioning */
+	/* User requested CPU mask? */
-	if (pe->nr_processes == -1) {
+	if (req.req_cpu_list && req.req_cpu_list_len) {
+		char *cpu_list = NULL;
+
+		cpu_list = kmalloc(req.req_cpu_list_len, GFP_KERNEL);
+		if (!cpu_list) {
+			printk("%s: error: allocating CPU list\n", __FUNCTION__);
+			ret = -ENOMEM;
+			goto put_out;
+		}
+
+		if (copy_from_user(cpu_list,
+					req.req_cpu_list, req.req_cpu_list_len)) {
+			printk("%s: error copying CPU list request\n", __FUNCTION__);
+			kfree(cpu_list);
+			ret = -EINVAL;
+			goto put_out;
+		}
+
+		cpus_used = kmalloc(sizeof(cpumask_t), GFP_KERNEL);
+		cpus_to_use = kmalloc(sizeof(cpumask_t), GFP_KERNEL);
+		if (!cpus_to_use || !cpus_used) {
+			printk("%s: error: allocating CPU mask\n", __FUNCTION__);
+			ret = -ENOMEM;
+			kfree(cpu_list);
+			goto put_out;
+		}
+		memset(cpus_used, 0, sizeof(cpumask_t));
+		memset(cpus_to_use, 0, sizeof(cpumask_t));
+
+		/* Parse CPU list */
+		if (cpulist_parse(cpu_list, cpus_to_use) < 0) {
+			printk("%s: invalid CPUs requested: %s\n",
+				__FUNCTION__, cpu_list);
+			ret = -EINVAL;
+			kfree(cpu_list);
+			goto put_out;
+		}
+
+		memcpy(cpus_used, cpus_to_use, sizeof(cpumask_t));
+
+		/* Copy mask to user-space */
+		if (copy_to_user(req.cpu_set, cpus_used,
+					(req.cpu_set_size < sizeof(cpumask_t) ?
+					 req.cpu_set_size : sizeof(cpumask_t)))) {
+			printk("%s: error copying mask to user\n", __FUNCTION__);
+			ret = -EINVAL;
+			kfree(cpu_list);
+			goto put_out;
+		}
+
+		/* Copy IKC target core */
+		cpu = cpumask_next(-1, cpus_used);
+		if (copy_to_user(req.target_core, &cpu, sizeof(cpu))) {
+			printk("%s: error copying target core to user\n",
+					__FUNCTION__);
+			ret = -EINVAL;
+			kfree(cpu_list);
+			goto put_out;
+		}
+
+		/* Save in per-process structure */
+		memcpy(&ppd->cpu_set, cpus_used, sizeof(cpumask_t));
+		ppd->ikc_target_cpu = cpu;
+		printk("%s: %s -> target McKernel CPU: %d\n",
+			__func__, cpu_list, cpu);
+
+		ret = 0;
+		kfree(cpu_list);
+		goto put_out;
+	}
+
+	mutex_lock(&udp->part_exec_lock);
+	/* Find part_exec having same node_proxy */
+	list_for_each_entry_reverse(pe_itr, &udp->part_exec_list, chain) {
+		pe_list_len++;
+		if (pe_itr->node_proxy_pid == req.ppid) {
+			pe = pe_itr;
+			break;
+		}
+	}
+
+	if (!pe) {
+		/* First process to enter CPU partitioning */
+		pr_debug("%s: pe_list_len:%d\n", __func__, pe_list_len);
+		if (pe_list_len >= PE_LIST_MAXLEN) {
+			/* delete head entry of pe_list */
+			pe_itr = list_first_entry(&udp->part_exec_list,
+					struct mcctrl_part_exec, chain);
+			list_del(&pe_itr->chain);
+			kfree(pe_itr);
+		}
+
+		pe = kzalloc(sizeof(struct mcctrl_part_exec), GFP_KERNEL);
+		if (!pe) {
+			mutex_unlock(&udp->part_exec_lock);
+			ret = -ENOMEM;
+			goto put_out;
+		}
+		/* Init part_exec */
+		mutex_init(&pe->lock);
+		INIT_LIST_HEAD(&pe->pli_list);
 		pe->nr_processes = req.nr_processes;
 		pe->nr_processes_left = req.nr_processes;
+		pe->nr_processes_joined = 0;
+		pe->node_proxy_pid = req.ppid;
+
+		list_add_tail(&pe->chain, &udp->part_exec_list);
 		dprintk("%s: nr_processes: %d (partitioned exec starts)\n",
-				__FUNCTION__,
+				__func__, pe->nr_processes);
-				pe->nr_processes);
 	}
+	mutex_unlock(&udp->part_exec_lock);
+
+	mutex_lock(&pe->lock);
 
 	if (pe->nr_processes != req.nr_processes) {
 		printk("%s: error: requested number of processes"
@@ -640,7 +743,15 @@ static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
 		goto put_and_unlock_out;
 	}
 
+	if (pe->nr_processes_joined >= pe->nr_processes) {
+		printk("%s: too many processes have joined to the group of %d\n",
+				__func__, req.ppid);
+		ret = -EINVAL;
+		goto put_and_unlock_out;
+	}
+
 	--pe->nr_processes_left;
+	++pe->nr_processes_joined;
 	dprintk("%s: nr_processes: %d, nr_processes_left: %d\n",
 			__FUNCTION__,
 			pe->nr_processes,
@@ -726,8 +837,6 @@ static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
 				wake_up_interruptible(&pli_next->pli_wq);
 			}
 
-			/* Reset process counter to start state */
-			pe->nr_processes = -1;
 			ret = -ETIMEDOUT;
 			goto put_and_unlock_out;
 		}
@@ -975,16 +1084,8 @@ next_cpu:
 	/* Commit used cores to OS structure */
 	memcpy(&pe->cpus_used, cpus_used, sizeof(*cpus_used));
 
-	/* Reset if last process */
+	/* If not last process, wake up next process in list */
-	if (pe->nr_processes_left == 0) {
+	if (pe->nr_processes_left != 0) {
-		dprintk("%s: nr_processes: %d (partitioned exec ends)\n",
-				__FUNCTION__,
-				pe->nr_processes);
-		pe->nr_processes = -1;
-		memset(&pe->cpus_used, 0, sizeof(pe->cpus_used));
-	}
-	/* Otherwise wake up next process in list */
-	else {
 		++pe->process_rank;
 		pli_next = list_first_entry(&pe->pli_list,
 			struct process_list_item, list);
@@ -997,11 +1098,14 @@ next_cpu:
 	ret = 0;
 
 put_and_unlock_out:
+	mutex_unlock(&pe->lock);
+
+put_out:
+	mcctrl_put_per_proc_data(ppd);
+
 	kfree(cpus_to_use);
 	kfree(cpus_used);
 	kfree(mcexec_cpu_set);
-	mcctrl_put_per_proc_data(ppd);
-	mutex_unlock(&pe->lock);
 
 	return ret;
 }

executer/kernel/mcctrl/ikc.c

@@ -513,6 +513,7 @@ int prepare_ikc_channels(ihk_os_t os)
 
 	init_waitqueue_head(&usrdata->wq_procfs);
 	mutex_init(&usrdata->reserve_lock);
+	mutex_init(&usrdata->part_exec_lock);
 
 	for (i = 0; i < MCCTRL_PER_PROC_DATA_HASH_SIZE; ++i) {
 		INIT_LIST_HEAD(&usrdata->per_proc_data_hash[i]);
@@ -521,10 +522,8 @@ int prepare_ikc_channels(ihk_os_t os)
 
 	INIT_LIST_HEAD(&usrdata->cpu_topology_list);
 	INIT_LIST_HEAD(&usrdata->node_topology_list);
+	INIT_LIST_HEAD(&usrdata->part_exec_list);
 
-	mutex_init(&usrdata->part_exec.lock);
-	INIT_LIST_HEAD(&usrdata->part_exec.pli_list);
-	usrdata->part_exec.nr_processes = -1;
 	INIT_LIST_HEAD(&usrdata->wakeup_descs_list);
 	spin_lock_init(&usrdata->wakeup_descs_lock);
 
@@ -580,6 +579,18 @@ void destroy_ikc_channels(ihk_os_t os)
 
 	kfree(usrdata->channels);
 	kfree(usrdata->ikc2linux);
+
+	mutex_lock(&usrdata->part_exec_lock);
+	while (!list_empty(&usrdata->part_exec_list)) {
+		struct mcctrl_part_exec *pe;
+
+		pe = list_first_entry(&usrdata->part_exec_list,
+				struct mcctrl_part_exec, chain);
+		list_del(&pe->chain);
+		kfree(pe);
+	}
+	mutex_unlock(&usrdata->part_exec_lock);
+
 	kfree(usrdata);
 }
 

executer/kernel/mcctrl/mcctrl.h

@@ -324,13 +324,20 @@ struct process_list_item {
 	wait_queue_head_t pli_wq;
 };
 
+#define PE_LIST_MAXLEN 5
+
 struct mcctrl_part_exec {
 	struct mutex lock;	
 	int nr_processes;
+	/* number of processes to let in / out the synchronization point */
 	int nr_processes_left;
+	/* number of processes which have joined the partition */
+	int nr_processes_joined;
 	int process_rank;
+	pid_t node_proxy_pid;
 	cpumask_t cpus_used;
 	struct list_head pli_list;
+	struct list_head chain;
 };
 
 #define CPU_LONGS (((NR_CPUS) + (BITS_PER_LONG) - 1) / (BITS_PER_LONG))
@@ -353,6 +360,7 @@ struct mcctrl_usrdata {
 	int	job_pos;
 	int	mcctrl_dma_abort;
 	struct mutex reserve_lock;
+	struct mutex part_exec_lock;
 	unsigned long	last_thread_exec;
 	wait_queue_head_t wq_procfs;
 	struct list_head per_proc_data_hash[MCCTRL_PER_PROC_DATA_HASH_SIZE];
@@ -368,7 +376,7 @@ struct mcctrl_usrdata {
 	nodemask_t numa_online;
 	struct list_head cpu_topology_list;
 	struct list_head node_topology_list;
-	struct mcctrl_part_exec part_exec;
+	struct list_head part_exec_list;
 	int perf_event_num;
 };
 

executer/user/mcexec.c

@@ -2211,6 +2211,64 @@ int main(int argc, char **argv)
 
 	pthread_spin_init(&overlay_fd_lock, 0);
 
+	/* XXX: Fugaku: Fujitsu process placement fix */
+	if (getenv("FLIB_AFFINITY_ON_PROCESS")) {
+		char *cpu_s;
+		int flib_size;
+		char *flib_aff_orig, *flib_aff;
+		int cpu, off = 0;
+
+		flib_aff_orig = strdup(getenv("FLIB_AFFINITY_ON_PROCESS"));
+		if (!flib_aff_orig) {
+			fprintf(stderr, "error: dupping FLIB_AFFINITY_ON_PROCESS\n");
+			exit(EXIT_FAILURE);
+		}
+
+		flib_size = strlen(flib_aff_orig) * 2;
+
+		flib_aff = malloc(flib_size);
+		if (!flib_aff) {
+			fprintf(stderr, "error: allocating memory for "
+					"FLIB_AFFINITY_ON_PROCESS\n");
+			exit(EXIT_FAILURE);
+		}
+		memset(flib_aff, 0, flib_size);
+
+		cpu_s = strtok(flib_aff_orig, ",");
+		while (cpu_s) {
+			int ret;
+
+			/* "Shift" left by 12 CPUs */
+			cpu = atoi(cpu_s) - 12;
+
+			/* Prepend "," */
+			if (off > 0) {
+				ret = snprintf(flib_aff + off, flib_size - off, "%s", ",");
+				if (ret < 0) {
+					fprintf(stderr, "error: constructing "
+							"FLIB_AFFINITY_ON_PROCESS\n");
+					exit(EXIT_FAILURE);
+				}
+
+				off += ret;
+			}
+
+			ret = snprintf(flib_aff + off, flib_size - off, "%d", cpu);
+			if (ret < 0) {
+				fprintf(stderr, "error: constructing "
+						"FLIB_AFFINITY_ON_PROCESS\n");
+				exit(EXIT_FAILURE);
+			}
+
+			off += ret;
+			cpu_s = strtok(NULL, ",");
+		}
+
+		__dprintf("FLIB_AFFINITY_ON_PROCESS: %s -> %s\n",
+				getenv("FLIB_AFFINITY_ON_PROCESS"), flib_aff);
+		setenv("FLIB_AFFINITY_ON_PROCESS", flib_aff, 1);
+	}
+
 	ld_preload_init();
 
 #ifdef ADD_ENVS_OPTION
@@ -2483,9 +2541,12 @@ int main(int argc, char **argv)
 
 		CPU_ZERO(&mcexec_cpu_set);
 
+		cpu_set_arg.req_cpu_list = NULL;
+		cpu_set_arg.req_cpu_list_len = 0;
 		cpu_set_arg.cpu_set = (void *)&desc->cpu_set;
 		cpu_set_arg.cpu_set_size = sizeof(desc->cpu_set);
 		cpu_set_arg.nr_processes = nr_processes;
+		cpu_set_arg.ppid = getppid();
 		cpu_set_arg.target_core = &target_core;
 		cpu_set_arg.process_rank = &process_rank;
 		cpu_set_arg.mcexec_linux_numa = &mcexec_linux_numa;
@@ -2493,6 +2554,16 @@ int main(int argc, char **argv)
 		cpu_set_arg.mcexec_cpu_set_size = sizeof(mcexec_cpu_set);
 		cpu_set_arg.ikc_mapped = &ikc_mapped;
 
+		/* Fugaku specific: Fujitsu CPU binding */
+		if (getenv("FLIB_AFFINITY_ON_PROCESS")) {
+			cpu_set_arg.req_cpu_list =
+				getenv("FLIB_AFFINITY_ON_PROCESS");
+			cpu_set_arg.req_cpu_list_len =
+				strlen(cpu_set_arg.req_cpu_list) + 1;
+			__dprintf("%s: requesting CPUs: %s\n",
+				__func__, cpu_set_arg.req_cpu_list);
+		}
+
 		if (ioctl(fd, MCEXEC_UP_GET_CPUSET, (void *)&cpu_set_arg) != 0) {
 			perror("getting CPU set for partitioned execution");
 			close(fd);
@@ -2501,6 +2572,12 @@ int main(int argc, char **argv)
 
 		desc->cpu = target_core;
 		desc->process_rank = process_rank;
+		/* Fugaku specific: Fujitsu node-local rank */
+		if (getenv("PLE_RANK_ON_NODE")) {
+			desc->process_rank = atoi(getenv("PLE_RANK_ON_NODE"));
+			__dprintf("%s: rank: %d, target CPU: %d\n",
+				__func__, desc->process_rank, desc->cpu);
+		}
 
 		/* Bind to CPU cores where the LWK process' IKC target maps to */
 		if (ikc_mapped && !no_bind_ikc_map) {

kernel/include/mman.h

@@ -72,4 +72,11 @@
 #define	MS_INVALIDATE	0x02
 #define	MS_SYNC		0x04
 
+/* include/uapi/linux/mman.h */
+#define OVERCOMMIT_GUESS		0
+#define OVERCOMMIT_ALWAYS		1
+#define OVERCOMMIT_NEVER		2
+
+extern int sysctl_overcommit_memory;
+
 #endif /* HEADER_MMAN_H */

kernel/include/rusage_private.h

@@ -11,6 +11,7 @@
 #include <ihk/ihk_monitor.h>
 #include <ihk/debug.h>
 #include <memory.h>
+#include <mman.h>
 
 #ifdef ENABLE_RUSAGE
 
@@ -56,6 +57,24 @@ rusage_total_memory_add(unsigned long size)
 #endif
 }
 
+static inline unsigned long
+rusage_get_total_memory()
+{
+	return rusage.total_memory;
+}
+
+static inline unsigned long
+rusage_get_free_memory()
+{
+	return rusage.total_memory - rusage.total_memory_usage;
+}
+
+static inline unsigned long
+rusage_get_usage_memory()
+{
+	return rusage.total_memory_usage;
+}
+
 static inline void
 rusage_rss_add(unsigned long size)
 {
@@ -293,6 +312,9 @@ rusage_check_overmap(size_t len, int pgshift)
 {
 	int npages = 0, remain_pages = 0;
 
+	if (sysctl_overcommit_memory == OVERCOMMIT_ALWAYS)
+		return 0;
+
 	npages = (len + (1UL << pgshift) - 1) >> pgshift;
 	remain_pages = (rusage.total_memory - rusage.total_memory_usage)
 			>> pgshift;
@@ -399,6 +421,24 @@ rusage_rss_add(unsigned long size)
 {
 }
 
+static inline unsigned long
+rusage_get_total_memory()
+{
+	return 0;
+}
+
+static inline unsigned long
+rusage_get_free_memory()
+{
+	return 0;
+}
+
+static inline unsigned long
+rusage_get_usage_memory()
+{
+	return 0;
+}
+
 static inline void
 rusage_rss_sub(unsigned long size)
 {

kernel/include/syscall.h

@@ -443,6 +443,22 @@ struct rusage {
 	long   ru_nivcsw;
 };
 
+struct sysinfo {
+	long uptime;             /* Seconds since boot */
+	unsigned long loads[3];  /* 1, 5, and 15 minute load averages */
+	unsigned long totalram;  /* Total usable main memory size */
+	unsigned long freeram;   /* Available memory size */
+	unsigned long sharedram; /* Amount of shared memory */
+	unsigned long bufferram; /* Memory used by buffers */
+	unsigned long totalswap; /* Total swap space size */
+	unsigned long freeswap;  /* swap space still available */
+	unsigned short procs;    /* Number of current processes */
+	unsigned long totalhigh; /* Total high memory size */
+	unsigned long freehigh;  /* Available high memory size */
+	unsigned int mem_unit;   /* Memory unit size in bytes */
+	char _f[20-2*sizeof(long)-sizeof(int)]; /* Padding for libc5 */
+};
+
 extern void terminate(int, int);
 
 struct tod_data_s {
@@ -635,4 +651,6 @@ extern int (*linux_clock_gettime)(clockid_t clk_id, struct timespec *tp);
 #define COREDUMP_TO_BE_WOKEN      2
 
 extern void terminate_host(int pid, struct thread *thread);
+struct sig_pending *getsigpending(struct thread *thread, int delflag);
+int interrupt_from_user(void *regs0);
 #endif

kernel/mem.c

@@ -62,6 +62,7 @@ extern int interrupt_from_user(void *);
 struct tlb_flush_entry tlb_flush_vector[IHK_TLB_FLUSH_IRQ_VECTOR_SIZE];
 
 int anon_on_demand = 0;
+int sysctl_overcommit_memory = OVERCOMMIT_ALWAYS;
 
 static struct ihk_mc_pa_ops *pa_ops;
 

kernel/syscall.c

@@ -120,7 +120,6 @@ extern int num_processors;
 extern unsigned long ihk_mc_get_ns_per_tsc(void);
 extern int ptrace_detach(int pid, int data);
 extern void debug_log(unsigned long);
-extern int arch_clear_host_user_space();
 extern long arch_ptrace(long request, int pid, long addr, long data);
 extern struct cpu_local_var *clv;
 
@@ -242,11 +241,6 @@ long do_syscall(struct syscall_request *req, int cpu)
 			unsigned long flags;
 			DECLARE_WAITQ_ENTRY(scd_wq_entry, cpu_local_var(current));
 
-			if (req->number == __NR_epoll_wait ||
-				req->number == __NR_epoll_pwait ||
-				req->number == __NR_ppoll)
-				goto schedule;
-
 			cpu_pause();
 
 			/* Spin if not preemptable */
@@ -275,7 +269,6 @@ long do_syscall(struct syscall_request *req, int cpu)
 				continue;
 			}
 
-schedule:
 			flags = cpu_disable_interrupt_save();
 
 			/* Try to sleep until notified */
@@ -2660,10 +2653,8 @@ static int do_execveat(ihk_mc_user_context_t *ctx, int dirfd,
 	}
 	
 	/* Clear host user space PTEs */
-	if (arch_clear_host_user_space()) {
+	clear_host_pte(vm->region.user_start,
-		kprintf("execve(): ERROR: clearing PTEs in host process\n");
+			(vm->region.user_end - vm->region.user_start), 0);
-		panic("");
-	}		
 
 	/* Request host to transfer ELF image */
 	request.number = __NR_execve;
@@ -5140,6 +5131,37 @@ SYSCALL_DECLARE(rt_sigsuspend)
 	return do_sigsuspend(thread, &wset);
 }
 
+SYSCALL_DECLARE(rt_sigaction)
+{
+	int sig = ihk_mc_syscall_arg0(ctx);
+	const struct sigaction *act =
+		(const struct sigaction *)ihk_mc_syscall_arg1(ctx);
+	struct sigaction *oact = (struct sigaction *)ihk_mc_syscall_arg2(ctx);
+	size_t sigsetsize = ihk_mc_syscall_arg3(ctx);
+	struct k_sigaction new_sa, old_sa;
+	int rc;
+
+	if (sigsetsize != sizeof(sigset_t))
+		return -EINVAL;
+
+	if (act) {
+		if (copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa))) {
+			goto fault;
+		}
+	}
+
+	rc = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
+	if (rc == 0 && oact) {
+		if (copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa))) {
+			goto fault;
+		}
+	}
+
+	return rc;
+fault:
+	return -EFAULT;
+}
+
 SYSCALL_DECLARE(sigaltstack)
 {
 	struct thread *thread = cpu_local_var(current);
@@ -6686,6 +6708,27 @@ SYSCALL_DECLARE(getrusage)
 	return 0;
 }
 
+SYSCALL_DECLARE(sysinfo)
+{
+	struct sysinfo *sysinfo = (struct sysinfo *)ihk_mc_syscall_arg0(ctx);
+	struct sysinfo __sysinfo;
+	int ret = 0;
+
+	memset(&__sysinfo, '\0', sizeof(struct sysinfo));
+
+	__sysinfo.totalram = rusage_get_total_memory();
+	__sysinfo.freeram = rusage_get_free_memory();
+	__sysinfo.mem_unit = 1; // always one unit for McKernel
+
+	if (copy_to_user(sysinfo, &__sysinfo, sizeof(struct sysinfo))) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+out:
+	return ret;
+}
+
 extern int ptrace_traceme(void);
 extern void set_single_step(struct thread *thread);
 
@@ -10320,11 +10363,7 @@ long syscall(int num, ihk_mc_user_context_t *ctx)
 	}
 #endif // PROFILE_ENABLE
 
-	/* Do not deschedule when returning from an event (e.g., MPI) */
+	if (smp_load_acquire(&v->flags) & CPU_FLAG_NEED_RESCHED) {
-	if (!(num == __NR_epoll_wait ||
-				num == __NR_epoll_pwait ||
-				num == __NR_ppoll) &&
-			smp_load_acquire(&v->flags) & CPU_FLAG_NEED_RESCHED) {
 		check_need_resched();
 	}
 
@@ -10345,3 +10384,275 @@ long syscall(int num, ihk_mc_user_context_t *ctx)
 
 	return l;
 }
+
+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 &&
+				 sig != SIGCONT) ||
+				(k->sa.sa_handler != SIG_IGN &&
+				 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 &&
+							sig != SIGCONT &&
+							!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;
+}
+
+struct sig_pending *
+getsigpending(struct thread *thread, int delflag)
+{
+	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;
+	struct k_sigaction *k;
+
+	w = thread->sigmask.__val[0];
+
+	lock = &thread->sigcommon->lock;
+	head = &thread->sigcommon->sigpending;
+	for (;;) {
+		if (delflag) {
+			mcs_rwlock_writer_lock(lock, &mcs_rw_node);
+		}
+		else {
+			mcs_rwlock_reader_lock(lock, &mcs_rw_node);
+		}
+
+		list_for_each_entry_safe(pending, next, head, list) {
+			for (x = pending->sigmask.__val[0], sig = 0; x;
+					sig++, x >>= 1) {
+			}
+			k = thread->sigcommon->action + sig - 1;
+			if (delflag ||
+			   (sig != SIGCHLD &&
+				sig != SIGURG &&
+				sig != SIGCONT) ||
+			   (k->sa.sa_handler != (void *)1 &&
+				k->sa.sa_handler != NULL)){
+				if (!(pending->sigmask.__val[0] & w)) {
+					if (delflag)
+						list_del(&pending->list);
+
+					if (delflag) {
+						mcs_rwlock_writer_unlock(
+							lock,
+							&mcs_rw_node);
+					}
+					else {
+						mcs_rwlock_reader_unlock(
+							lock,
+							&mcs_rw_node);
+					}
+					return pending;
+				}
+			}
+		}
+
+		if (delflag) {
+			mcs_rwlock_writer_unlock(lock, &mcs_rw_node);
+		}
+		else {
+			mcs_rwlock_reader_unlock(lock, &mcs_rw_node);
+		}
+
+		if (lock == &thread->sigpendinglock) {
+			return NULL;
+		}
+
+		lock = &thread->sigpendinglock;
+		head = &thread->sigpending;
+	}
+
+	return NULL;
+}
+
+struct sig_pending *
+hassigpending(struct thread *thread)
+{
+	if (list_empty(&thread->sigpending) &&
+		list_empty(&thread->sigcommon->sigpending)) {
+		return NULL;
+	}
+
+	return getsigpending(thread, 0);
+}
+
+void
+check_sig_pending(void)
+{
+	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->group_exit_status & 0x0000000100000000L) {
+			continue;
+		}
+
+		if (check_sig_pending_thread(thread))
+			goto repeat;
+	}
+	ihk_mc_spinlock_unlock(&v->runq_lock, v->runq_irqstate);
+}
+
+static void
+__check_signal(unsigned long rc, void *regs0, int num, int irq_disabled)
+{
+	ihk_mc_user_context_t *regs = regs0;
+	struct thread *thread;
+	struct sig_pending *pending;
+	int irqstate;
+
+	if (clv == NULL) {
+		return;
+	}
+	thread = cpu_local_var(current);
+
+	if (thread == NULL || thread->proc->pid == 0) {
+		struct thread *t;
+
+		irqstate = cpu_disable_interrupt_save();
+		ihk_mc_spinlock_lock_noirq(&(cpu_local_var(runq_lock)));
+		list_for_each_entry(t, &(cpu_local_var(runq)), sched_list) {
+			if (t->proc->pid <= 0) {
+				continue;
+			}
+			if (t->status == PS_INTERRUPTIBLE &&
+			   hassigpending(t)) {
+				t->status = PS_RUNNING;
+				break;
+			}
+		}
+		ihk_mc_spinlock_unlock_noirq(&(cpu_local_var(runq_lock)));
+		cpu_restore_interrupt(irqstate);
+		goto out;
+	}
+
+	if (regs != NULL && !interrupt_from_user(regs)) {
+		goto out;
+	}
+
+	if (list_empty(&thread->sigpending) &&
+		list_empty(&thread->sigcommon->sigpending)) {
+		goto out;
+	}
+
+	for (;;) {
+		/* When this function called from check_signal_irq_disabled,
+		 * return with interrupt invalid.
+		 * This is to eliminate signal loss.
+		 */
+		if (irq_disabled == 1) {
+			irqstate = cpu_disable_interrupt_save();
+		}
+		pending = getsigpending(thread, 1);
+		if (!pending) {
+			dkprintf("check_signal,queue is empty\n");
+			goto out;
+		}
+		if (irq_disabled == 1) {
+			cpu_restore_interrupt(irqstate);
+		}
+		if (do_signal(rc, regs, thread, pending, num)) {
+			num = -1;
+		}
+	}
+
+out:
+	return;
+}
+
+void
+check_signal(unsigned long rc, void *regs0, int num)
+{
+	__check_signal(rc, regs0, num, 0);
+}
+
+void
+check_signal_irq_disabled(unsigned long rc, void *regs0, int num)
+{
+	__check_signal(rc, regs0, num, 1);
+}
+

test/issues/1389/C1389.sh

@@ -0,0 +1,43 @@
+#/bin/sh
+
+USELTP=1
+USEOSTEST=0
+
+BOOTPARAM="-c 1-7 -m 10G@0,10G@1 -O"
+. ../../common.sh
+
+issue="1389"
+tid=01
+
+for tsuf in "01" "02.sh"
+do
+	tname=`printf "C${issue}T%02d" ${tid}`
+	echo "*** ${tname} start *******************************"
+	${MCEXEC} ./C1389T${tsuf}
+
+	if [ $? -eq 0 ]; then
+		echo "*** ${tname} PASSED ******************************"
+	else
+		echo "*** ${tname} FAILED ******************************"
+	fi
+	let tid++
+	echo ""
+done
+
+for tp in "mtest01 -p80" "mtest01 -p80 -w" "mem01"
+do
+	tname=`printf "C${issue}T%02d" ${tid}`
+	echo "*** ${tname} start *******************************"
+echo ${tp}
+	sudo $MCEXEC $LTPBIN/$tp 2>&1 | tee ${tname}.txt
+	ok=`grep PASS ${tname}.txt | wc -l`
+	ng=`grep FAIL ${tname}.txt | wc -l`
+	if [ $ng = 0 ]; then
+		echo "*** ${tname} PASSED ($ok)"
+	else
+		echo "*** ${tname} FAILED (ok=$ok ng=$ng)"
+	fi
+	let tid++
+	echo ""
+done
+

test/issues/1389/C1389T01.c

@@ -0,0 +1,65 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <linux/kernel.h>
+#include <sys/sysinfo.h>
+#include <errno.h>
+
+int main(void)
+{
+	struct sysinfo *info;
+	int ret = 0, rc = 0;
+	unsigned long assigned_mem = (20UL << 30);
+	unsigned long _totalram, _freeram;
+	unsigned int _mem_unit;
+
+	info = malloc(sizeof(struct sysinfo));
+	rc = sysinfo(info);
+	if (rc) {
+		perror("sysinfo fail: ");
+		ret = -1;
+		goto out;
+	}
+
+	_totalram = info->totalram;
+	_freeram = info->freeram;
+	_mem_unit = info->mem_unit;
+
+	// Check totalram
+	if (0.95 * assigned_mem < _totalram &&
+			_totalram < assigned_mem) {
+		printf("[OK] totalram: %ld\n", _totalram);
+	}
+	else {
+		printf("[NG] unexpected totalram: %ld\n", _totalram);
+		printf("  expected range: %ld - %ld\n",
+			(unsigned long)(0.95 * assigned_mem),
+			assigned_mem);
+		ret = -1;
+		goto out;
+	}
+
+	// Check freeram
+	if (0.95 * _totalram < _freeram &&
+			_freeram < _totalram) {
+		printf("[OK] freeram: %ld\n", _freeram);
+	}
+	else {
+		printf("[NG] unexpected freeram: %ld\n", _freeram);
+		ret = -1;
+		goto out;
+	}
+
+	// Check mem_unit
+	if (_mem_unit == 1) {
+		printf("[OK] mem_unit: %ld\n", _mem_unit);
+	}
+	else {
+		printf("[NG] unexpected mem_unit: %ld\n", _mem_unit);
+		ret = -1;
+		goto out;
+	}
+
+	free(info);
+out:
+	return ret;
+}

test/issues/1389/C1389T02.sh

@@ -0,0 +1,88 @@
+#!/bin/sh
+
+ASSIGNED_MEM=`echo '1024 * 1024 * 1024 * 20' | bc`
+cat /proc/meminfo > ./cur_meminfo.txt
+
+# Check MemTotal
+MemTotalTxt=`cat ./cur_meminfo.txt | grep MemTotal | awk '{print $(NF-1)}'`
+MemTotal=`echo "1024 * ${MemTotalTxt}" | bc`
+
+lower_limit=`echo "${ASSIGNED_MEM} * 0.95" | bc`
+lower_limit=${lower_limit%.*}
+upper_limit=${ASSIGNED_MEM}
+tgt=${MemTotal}
+
+if [ ${tgt} -ge ${lower_limit} -a ${tgt} -lt ${upper_limit} ]; then
+	echo "[OK] MemTotal: ${tgt}"
+else
+	echo "[NG] unexpected MemTotal: ${tgt}"
+	exit 1
+fi
+
+# Check MemFree
+MemFreeTxt=`cat ./cur_meminfo.txt | grep MemFree | awk '{print $(NF-1)}'`
+MemFree=`echo "1024 * ${MemFreeTxt}" | bc`
+
+lower_limit=`echo "${MemTotal} * 0.95" | bc`
+lower_limit=${lower_limit%.*}
+upper_limit=${MemTotal}
+tgt=${MemFree}
+
+if [ ${tgt} -ge ${lower_limit} -a ${tgt} -lt ${upper_limit} ]; then
+	echo "[OK] MemFree: ${tgt}"
+else
+	echo "[NG] unexpected MemFree: ${tgt}"
+	exit 1
+fi
+
+# Check SwapTotal
+SwapTotalTxt=`cat ./cur_meminfo.txt | grep SwapTotal | awk '{print $(NF-1)}'`
+SwapTotal=`echo "1024 * ${SwapTotalTxt}" | bc`
+
+tgt=${SwapTotal}
+
+if [ ${tgt} -eq 0 ]; then
+	echo "[OK] SwapTotal: ${tgt}"
+else
+	echo "[NG] unexpected SwapTotal: ${tgt}"
+	exit 1
+fi
+
+# Check SwapFree
+SwapFreeTxt=`cat ./cur_meminfo.txt | grep SwapFree | awk '{print $(NF-1)}'`
+SwapFree=`echo "1024 * ${SwapFreeTxt}" | bc`
+
+tgt=${SwapFree}
+
+if [ ${tgt} -eq 0 ]; then
+	echo "[OK] SwapFree: ${tgt}"
+else
+	echo "[NG] unexpected SwapFree: ${tgt}"
+	exit 1
+fi
+
+# Check CommitLimit
+CommitLimitTxt=`cat ./cur_meminfo.txt | grep CommitLimit | awk '{print $(NF-1)}'`
+CommitLimit=`echo "1024 * ${CommitLimitTxt}" | bc`
+
+tgt=${CommitLimit}
+
+if [ ${tgt} -eq ${MemFree} ]; then
+	echo "[OK] CommitLimit: ${tgt}"
+else
+	echo "[NG] unexpected CommitLimit: ${tgt}"
+	exit 1
+fi
+
+# Check Committed_AS
+Committed_ASTxt=`cat ./cur_meminfo.txt | grep Committed_AS | awk '{print $(NF-1)}'`
+Committed_AS=`echo "1024 * ${Committed_ASTxt}" | bc`
+
+tgt=${Committed_AS}
+
+if [ ${tgt} -eq $((${MemTotal} - ${MemFree})) ]; then
+	echo "[OK] Committed_AS: ${tgt}"
+else
+	echo "[NG] unexpected Committed_AS: ${tgt}"
+	exit 1
+fi

test/issues/1389/Makefile

@@ -0,0 +1,11 @@
+CFLAGS=-g
+LDFLAGS=
+
+TARGET=C1389T01
+
+all: $(TARGET)
+
+test: all
+	sh ./C1389.sh
+clean:
+	rm -f $(TARGET) *.o *.txt

test/issues/1389/README

@@ -0,0 +1,46 @@
+【Issue#1389 動作確認】
+□ テスト内容
+本Issueの対応で、sysinfo()と/proc/meminfo の下記項目においてMcKernelの情報を
+返すようにした。
+[sysinfo]
+  totalram : rusage_global の total_memory
+  freeram : rusage_global の total_memory - usage_memory
+  mem_unit : 常に1
+[/proc/meminfo]
+  MemTotal : rusage_global の total_memory
+  MemFree : rusage_global の total_memory - usage_memory
+  SwapTotal : 常に0
+  SwapFree : 常に0
+  CommitLimit : rusage_global の total_memory - usage_memory
+  Committed_AS : rusage_global の usage_memory
+
+1. 本Issueで対応したsysinfo() と /proc/meminfo の値が想定どおりとなっていることを確認
+McKernelへのメモリ割り当てを 10G@0,10G@1 とした状態で下記プログラムを実行
+C1389T01: sysinfo() で取得される値が下記のとおりであることを確認
+  totalram : 20GiB * 0.95 以上、20GiB 未満であること
+  freeram : totralram * 0.95 以上、totalram 未満であること
+  mem_unit : 1 であること
+C1389T02: /proc/meminfo から取得される値が下記のとおりであることを確認
+  MemTotal : 20GiB * 0.95 以上、20GiB 未満であること
+  MemFree : MemTotal * 0.95 以上、MemTotal 未満であること
+  SwapTotal : 0 であること
+  SwapFree : 0 であること
+  CommitLimit : MemFreeと同値であること
+  Committed_AS : MemTotal - MemFree と同値であること
+
+2. 本Issueで対応したsysinfo() の項目を利用する以下のLTPを実行し、PASSすることを確認
+  - mtest01 -p80
+  - mtest01 -p80 -w
+  - mem01
+
+□ 実行手順
+$ make test
+
+McKernelのインストール先や、OSTEST, LTPの配置場所は、
+$HOME/.mck_test_config を参照している
+.mck_test_config は、McKernelをビルドした際に生成されるmck_test_config.sample ファイルを
+$HOMEにコピーし、適宜編集する
+
+□ 実行結果
+x86_64_result.log aarch64_result.log 参照。
+すべての項目をPASSしていることを確認。

test/issues/1389/aarch64_result.log

@@ -0,0 +1,77 @@
+sh ./C1389.sh
+mcstop+release.sh ... done
+mcreboot.sh -c 1-7 -m 10G@0,10G@1 -O ... done
+*** C1389T01 start *******************************
+[OK] totalram: 21472215040
+[OK] freeram: 21461794816
+[OK] mem_unit: 1
+*** C1389T01 PASSED ******************************
+
+*** C1389T02 start *******************************
+[OK] MemTotal: 21472215040
+[OK] MemFree: 21452750848
+[OK] SwapTotal: 0
+[OK] SwapFree: 0
+[OK] CommitLimit: 21452750848
+[OK] Committed_AS: 19464192
+*** C1389T02 PASSED ******************************
+
+*** C1389T03 start *******************************
+mtest01 -p80
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+mtest01.c:134: INFO: Filling up 80% of free ram which is 16760729 kbytes
+mtest01.c:149: INFO: ... child 50935 starting
+mtest01.c:149: INFO: ... child 50944 starting
+mtest01.c:149: INFO: ... child 50953 starting
+mtest01.c:149: INFO: ... child 50962 starting
+mtest01.c:149: INFO: ... child 50971 starting
+mtest01.c:149: INFO: ... child 50980 starting
+mtest01.c:169: INFO: ... [t=300] 1434451968 bytes allocated only in child 50980
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 50935
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 50944
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 50953
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 50962
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 50971
+mtest01.c:221: PASS: 16760729 kbytes allocated
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1389T03 PASSED (1)
+
+*** C1389T04 start *******************************
+mtest01 -p80 -w
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+mtest01.c:134: INFO: Filling up 80% of free ram which is 16759654 kbytes
+mtest01.c:149: INFO: ... child 51017 starting
+mtest01.c:149: INFO: ... child 51026 starting
+mtest01.c:149: INFO: ... child 51035 starting
+mtest01.c:149: INFO: ... child 51044 starting
+mtest01.c:149: INFO: ... child 51053 starting
+mtest01.c:149: INFO: ... child 51062 starting
+mtest01.c:166: INFO: ... [t=300] 1433403392 bytes allocated and used in child 51062
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 51017
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 51026
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 51035
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 51044
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 51053
+mtest01.c:221: PASS: 16759654 kbytes allocated (and written to)
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1389T04 PASSED (1)
+
+*** C1389T05 start *******************************
+mem01
+mem01       0  TINFO  :  Free Mem:      20465 Mb
+mem01       0  TINFO  :  Free Swap:     0 Mb
+mem01       0  TINFO  :  Total Free:    20465 Mb
+mem01       0  TINFO  :  Total Tested:  3056 Mb
+mem01       0  TINFO  :  touching 3056MB of malloc'ed memory (linear)
+mem01       1  TPASS  :  malloc - alloc of 3056MB succeeded
+*** C1389T05 PASSED (1)

test/issues/1389/x86_64_result.log

@@ -0,0 +1,77 @@
+sh ./C1389.sh
+mcstop+release.sh ... done
+mcreboot.sh -c 1-7 -m 10G@0,10G@1 -O ... done
+*** C1389T01 start *******************************
+[OK] totalram: 21463240704
+[OK] freeram: 21459673088
+[OK] mem_unit: 1
+*** C1389T01 PASSED ******************************
+
+*** C1389T02 start *******************************
+[OK] MemTotal: 21463240704
+[OK] MemFree: 21453983744
+[OK] SwapTotal: 0
+[OK] SwapFree: 0
+[OK] CommitLimit: 21453983744
+[OK] Committed_AS: 9256960
+*** C1389T02 PASSED ******************************
+
+*** C1389T03 start *******************************
+mtest01 -p80
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+mtest01.c:134: INFO: Filling up 80% of free ram which is 16762396 kbytes
+mtest01.c:149: INFO: ... child 20718 starting
+mtest01.c:149: INFO: ... child 20727 starting
+mtest01.c:149: INFO: ... child 20736 starting
+mtest01.c:149: INFO: ... child 20745 starting
+mtest01.c:149: INFO: ... child 20754 starting
+mtest01.c:149: INFO: ... child 20763 starting
+mtest01.c:169: INFO: ... [t=300] 1436549120 bytes allocated only in child 20763
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 20718
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 20727
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 20736
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 20745
+mtest01.c:169: INFO: ... [t=299] 3145728000 bytes allocated only in child 20754
+mtest01.c:221: PASS: 16762396 kbytes allocated
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1389T03 PASSED (1)
+
+*** C1389T04 start *******************************
+mtest01 -p80 -w
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+mtest01.c:134: INFO: Filling up 80% of free ram which is 16761132 kbytes
+mtest01.c:149: INFO: ... child 20802 starting
+mtest01.c:149: INFO: ... child 20811 starting
+mtest01.c:149: INFO: ... child 20820 starting
+mtest01.c:149: INFO: ... child 20829 starting
+mtest01.c:149: INFO: ... child 20838 starting
+mtest01.c:149: INFO: ... child 20847 starting
+mtest01.c:166: INFO: ... [t=300] 1435500544 bytes allocated and used in child 20847
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 20802
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 20811
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 20820
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 20829
+mtest01.c:166: INFO: ... [t=299] 3145728000 bytes allocated and used in child 20838
+mtest01.c:221: PASS: 16761132 kbytes allocated (and written to)
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1389T04 PASSED (1)
+
+*** C1389T05 start *******************************
+mem01
+mem01       0  TINFO  :  Free Mem:      20463 Mb
+mem01       0  TINFO  :  Free Swap:     0 Mb
+mem01       0  TINFO  :  Total Free:    20463 Mb
+mem01       0  TINFO  :  Total Tested:  3056 Mb
+mem01       0  TINFO  :  touching 3056MB of malloc'ed memory (linear)
+mem01       1  TPASS  :  malloc - alloc of 3056MB succeeded
+*** C1389T05 PASSED (1)

test/issues/1457/C1457.py

@@ -0,0 +1,63 @@
+#
+# Test script for issue #1439
+#
+
+import os
+import sys
+import subprocess
+
+def get_command_result(cmd):
+    results = subprocess.Popen(
+        cmd, stdout=subprocess.PIPE,
+        shell=True).stdout.readlines()
+    return [str(x).rstrip("\n") for x in results]
+
+def enumerate_cpu(cpu_list):
+    allcpus = []
+    for ranged_cpu in cpu_list.split(','):
+        try:
+            cpu_begin, cpu_end = ranged_cpu.split('-')
+        except ValueError:
+            cpu_begin = cpu_end = ranged_cpu
+        for i in range(int(cpu_begin), int(cpu_end) + 1):
+            allcpus.append(i)
+    allcpus.sort()
+    return allcpus
+
+def get_online_cpu():
+    online_file = open('/sys/devices/system/cpu/online', 'r')
+    return online_file.readlines()[0].strip()
+
+def get_cpuinfo_processors():
+    processors = []
+    cpuinfo_file = open('/proc/cpuinfo', 'r')
+    for line in cpuinfo_file.readlines():
+        if line.startswith('processor'):
+            processor = int(line.strip().split()[-1])
+            processors.append(processor)
+    return processors
+
+def main():
+    onlines = enumerate_cpu(get_online_cpu())
+    processors = get_cpuinfo_processors()
+
+    print '# of online cpus:', len(onlines)
+    print '# of cpuinfo processors:', len(processors)
+    if len(onlines) != len(processors):
+        print 'ERROR: # of processors is not equal to # of cpus'
+        print 'FAIL'
+        exit()
+
+    i = 0
+    for cpu in processors:
+        print i, 'processor:', cpu
+        if i != cpu:
+            print 'ERROR: processor number is not ordered'
+            print 'FAIL'
+            exit()
+        i = i + 1
+
+    print 'SUCCESS'
+
+if __name__ == '__main__':
+    main()

test/issues/1457/C1457.sh

@@ -0,0 +1,7 @@
+#!/bin/bash
+
+. ${HOME}/.mck_test_config
+export MCK_DIR
+sudo ${MCK_DIR}/sbin/mcreboot.sh 1,3,5,7,9-12,50-58
+${MCK_DIR}/bin/mcexec python C1457.py
+sudo ${MCK_DIR}/sbin/mcstop+release.sh

test/issues/1457/Makefile

@@ -0,0 +1,2 @@
+test::
+	./C1457.shy

test/issues/1457/README

@@ -0,0 +1,22 @@
+□ テスト内容
+
+arm64環境における/proc/cpuinfoについて、以下を確かめる。
+
+(1) "processor" フィールドが/sys/devices/system/cpu/onlineファイルが示す
+CPUの数だけ存在すること。
+(2) "processor" フィールドの値が0からはじまる連番であること。
+
+
+□ 実行手順
+
+(1) $HOME/.mck_test_configを、MCK_DIRがMcKernelのインストール先を指すように編集する
+(2) 以下を実行する
+
+$ make test
+
+
+□ 確認方法
+
+標準出力にSUCCESSが出力されること
+
+

test/issues/1457/aarch64_result.log

@@ -0,0 +1,31 @@
+# of online cpus: 28
+# of cpuinfo processors: 28
+0 processor: 0
+1 processor: 1
+2 processor: 2
+3 processor: 3
+4 processor: 4
+5 processor: 5
+6 processor: 6
+7 processor: 7
+8 processor: 8
+9 processor: 9
+10 processor: 10
+11 processor: 11
+12 processor: 12
+13 processor: 13
+14 processor: 14
+15 processor: 15
+16 processor: 16
+17 processor: 17
+18 processor: 18
+19 processor: 19
+20 processor: 20
+21 processor: 21
+22 processor: 22
+23 processor: 23
+24 processor: 24
+25 processor: 25
+26 processor: 26
+27 processor: 27
+SUCCESS

test/issues/929/C929.sh

@@ -0,0 +1,145 @@
+#/bin/sh
+
+USELTP=0
+USEOSTEST=0
+
+. ../../common.sh
+
+issue="929"
+tid=01
+
+tname=`printf "C${issue}T%02d" ${tid}`
+echo "*** ${tname} start *******************************"
+TEST_CMD="mpirun -f ./hostfile -ppn 5 ${MCEXEC} -n 5 ./test_prog.sh"
+echo ${TEST_CMD}
+${TEST_CMD} &> ${tname}.txt
+mpi_ret=$?
+
+cat ./${tname}.txt
+started_num=`grep 'test_prog is started' ./${tname}.txt | wc -l`
+
+if [ ${mpi_ret} -eq 0 -a ${started_num} -eq 5 ]; then
+	echo "*** ${tname} PASSED ******************************"
+else
+	echo "*** ${tname} FAILED ******************************"
+fi
+let tid++
+echo ""
+
+tname=`printf "C${issue}T%02d" ${tid}`
+echo "*** ${tname} start *******************************"
+TEST_CMD="mpirun -f ./hostfile -ppn 5 ${MCEXEC} -n 3 ./test_prog.sh"
+echo ${TEST_CMD}
+${TEST_CMD} &> ${tname}.txt
+mpi_ret=$?
+
+cat ./${tname}.txt
+started_num=`grep 'test_prog is started' ./${tname}.txt | wc -l`
+
+if [ ${mpi_ret} -ne 0 -a ${started_num} -eq 3 ]; then
+	echo "*** ${tname} PASSED ******************************"
+else
+	echo "*** ${tname} FAILED ******************************"
+fi
+let tid++
+echo ""
+
+tname=`printf "C${issue}T%02d" ${tid}`
+echo "*** ${tname} start *******************************"
+TEST_CMD="mpirun -f ./hostfile -ppn 3 ${MCEXEC} -n 5 ./test_prog.sh"
+echo ${TEST_CMD}
+${TEST_CMD} &> ${tname}.txt
+mpi_ret=$?
+
+cat ./${tname}.txt
+started_num=`grep 'test_prog is started' ./${tname}.txt | wc -l`
+
+if [ ${mpi_ret} -ne 0 -a ${started_num} -eq 0 ]; then
+	echo "*** ${tname} PASSED ******************************"
+else
+	echo "*** ${tname} FAILED ******************************"
+fi
+let tid++
+echo ""
+
+tname=`printf "C${issue}T%02d" ${tid}`
+echo "*** ${tname} start *******************************"
+TEST_CMD="mpirun -f ./hostfile -ppn 6 ${MCEXEC} -n 3 ./test_prog.sh"
+echo ${TEST_CMD}
+${TEST_CMD} &> ${tname}.txt
+mpi_ret=$?
+
+cat ./${tname}.txt
+started_num=`grep 'test_prog is started' ./${tname}.txt | wc -l`
+
+if [ ${mpi_ret} -ne 0 -a ${started_num} -eq 3 ]; then
+	echo "*** ${tname} PASSED ******************************"
+else
+	echo "*** ${tname} FAILED ******************************"
+fi
+let tid++
+echo ""
+
+tname=`printf "C${issue}T%02d" ${tid}`
+echo "*** ${tname} start *******************************"
+TEST_CMD="mpirun -f ./hostfile -ppn 250 ${MCEXEC} -n 250 ./test_prog.sh"
+echo ${TEST_CMD}
+${TEST_CMD} &> ${tname}.txt
+mpi_ret=$?
+
+head -n 10 ./${tname}.txt
+echo "..."
+started_num=`grep 'test_prog is started' ./${tname}.txt | wc -l`
+
+if [ ${mpi_ret} -ne 0 -a ${started_num} -eq 0 ]; then
+	echo "*** ${tname} PASSED ******************************"
+else
+	echo "*** ${tname} FAILED ******************************"
+fi
+let tid++
+echo ""
+
+tname=`printf "C${issue}T%02d" ${tid}`
+echo "*** ${tname} start *******************************"
+ng=0
+TEST_CMD="mpirun -f ./hostfile -ppn 5 ${MCEXEC} -n 5 ./test_prog.sh"
+echo "** reboot mcrernel for check pe_list_len"
+mcreboot
+echo "** enable debug message in mcexec_get_cpuset"
+sudo sh -c "echo -n 'func mcexec_get_cpuset +p' > /sys/kernel/debug/dynamic_debug/control"
+echo ${TEST_CMD}
+for i in `seq 1 20`
+do
+	${TEST_CMD} &> ${tname}.txt
+	mpi_ret=$?
+	started_num=`grep 'test_prog is started' ./${tname}.txt | wc -l`
+	if [ ${mpi_ret} -eq 0 -a ${started_num} -eq 5 ]; then
+		echo "[OK] exec: $i"
+	else
+		echo "[NG] exec: $i"
+		let ng++
+	fi
+done
+echo "** check pe_list_len"
+dmesg --notime | grep "mcexec_get_cpuset: pe_list" | tail -n 20 | cut -f 2-3 -d ':' > ./pe_list_len.txt
+cat ./pe_list_len.txt | while read line
+do
+	len=`echo ${line} | cut -f 2 -d ':'`
+	if [ ${len} -ge 0 -a ${len} -le 5 ]; then
+		echo "[OK] ${line}"
+	else
+		echo "[NG] ${line}"
+		let ng++
+	fi
+done
+echo "** disable debug message in mcexec_get_cpuset"
+sudo sh -c "echo -n 'func mcexec_get_cpuset -p' > /sys/kernel/debug/dynamic_debug/control"
+
+if [ ${ng} -eq 0 ]; then
+	echo "*** ${tname} PASSED ******************************"
+else
+	echo "*** ${tname} FAILED ******************************"
+fi
+let tid++
+echo ""
+

test/issues/929/Makefile

@@ -0,0 +1,11 @@
+CFLAGS=-g
+LDFLAGS=
+
+TARGET=
+
+all: $(TARGET)
+
+test: all
+	./C929.sh
+clean:
+	rm -f $(TARGET) *.o *.txt

test/issues/929/README

@@ -0,0 +1,36 @@
+【Issue#929 動作確認】
+□ テスト内容
+1. mpirunで指定する-ppnと、mcexecで指定する-n の指定状況ごとに
+  想定どおりの動作となることを確認
+C929T01:
+  -ppn == -n の場合に、プログラムが実行され、mpirunが成功する
+
+C929T02:
+  -ppn > -n の場合に、プログラムの一部が実行され、mpirunが失敗する
+
+C929T03:
+  -ppn < -n の場合に、プログラムが実行されず、mpirunが失敗する
+
+C929T04:
+  -ppn が -n の整数倍である場合に、プログラムの一部が実行され、mpirunが失敗する
+
+C929T05:
+  -ppn と -n がMcKernelに割り当てたCPU数よりも大きい場合に、
+  プログラムが実行されず、mpirunが失敗する
+
+C929T06:
+  -ppn == -n での正常実行を20回連続で行った場合に、
+  プログラムが実行され、mpirunが成功する
+  また、mcctrlで管理しているpart_exec_list の要素数が5を超えない
+
+□ 実行手順
+$ make test
+
+McKernelのインストール先や、OSTEST, LTPの配置場所は、
+$HOME/.mck_test_config を参照している
+.mck_test_config は、McKernelをビルドした際に生成されるmck_test_config.sample ファイルを
+$HOMEにコピーし、適宜編集する
+
+□ 実行結果
+x86_64_result.log aarch64_result.log 参照。
+すべての項目をPASSしていることを確認。

test/issues/929/aarch64_result.log

@@ -0,0 +1,99 @@
+*** C929T01 start *******************************
+mpirun -f ./hostfile -ppn 5 /home/satoken/ihk+mckernel/bin/mcexec -n 5 ./test_prog.sh
+test_prog is started.
+test_prog is started.
+test_prog is started.
+test_prog is started.
+test_prog is started.
+*** C929T01 PASSED ******************************
+
+*** C929T02 start *******************************
+mpirun -f ./hostfile -ppn 5 /home/satoken/ihk+mckernel/bin/mcexec -n 3 ./test_prog.sh
+getting CPU set for partitioned execution: Invalid argument
+getting CPU set for partitioned execution: Invalid argument
+test_prog is started.
+test_prog is started.
+test_prog is started.
+*** C929T02 PASSED ******************************
+
+*** C929T03 start *******************************
+mpirun -f ./hostfile -ppn 3 /home/satoken/ihk+mckernel/bin/mcexec -n 5 ./test_prog.sh
+getting CPU set for partitioned execution: Connection timed out
+getting CPU set for partitioned execution: Connection timed out
+getting CPU set for partitioned execution: Connection timed out
+*** C929T03 PASSED ******************************
+
+*** C929T04 start *******************************
+mpirun -f ./hostfile -ppn 6 /home/satoken/ihk+mckernel/bin/mcexec -n 3 ./test_prog.sh
+getting CPU set for partitioned execution: Invalid argument
+getting CPU set for partitioned execution: Invalid argument
+getting CPU set for partitioned execution: Invalid argument
+test_prog is started.
+test_prog is started.
+test_prog is started.
+*** C929T04 PASSED ******************************
+
+*** C929T05 start *******************************
+mpirun -f ./hostfile -ppn 250 /home/satoken/ihk+mckernel/bin/mcexec -n 250 ./test_prog.sh
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+...
+*** C929T05 PASSED ******************************
+
+*** C929T06 start *******************************
+** reboot mcrernel for check pe_list_len
+mcreboot.sh -c 1-6,29-34 -m 50G@0,50G@1 -r 1-6:0+29-34:28 -O ... done
+** enable debug message in mcexec_get_cpuset
+mpirun -f ./hostfile -ppn 5 /home/satoken/ihk+mckernel/bin/mcexec -n 5 ./test_prog.sh
+[OK] exec: 1
+[OK] exec: 2
+[OK] exec: 3
+[OK] exec: 4
+[OK] exec: 5
+[OK] exec: 6
+[OK] exec: 7
+[OK] exec: 8
+[OK] exec: 9
+[OK] exec: 10
+[OK] exec: 11
+[OK] exec: 12
+[OK] exec: 13
+[OK] exec: 14
+[OK] exec: 15
+[OK] exec: 16
+[OK] exec: 17
+[OK] exec: 18
+[OK] exec: 19
+[OK] exec: 20
+** check pe_list_len
+[OK] pe_list_len:0
+[OK] pe_list_len:1
+[OK] pe_list_len:2
+[OK] pe_list_len:3
+[OK] pe_list_len:4
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+[OK] pe_list_len:5
+** disable debug message in mcexec_get_cpuset
+*** C929T06 PASSED ******************************
+

test/issues/929/hostfile

@@ -0,0 +1 @@
+localhost

test/issues/929/test_prog.sh

@@ -0,0 +1,3 @@
+#!/bin/sh
+
+echo "test_prog is started."

test/issues/929/x86_64_result.log

@@ -0,0 +1,74 @@
+*** C929T01 start *******************************
+mpirun -f ./hostfile -ppn 5 /home/satoken/ihk+mckernel/bin/mcexec -n 5 ./test_prog.sh
+test_prog is started.
+test_prog is started.
+test_prog is started.
+test_prog is started.
+test_prog is started.
+*** C929T01 PASSED ******************************
+
+*** C929T02 start *******************************
+mpirun -f ./hostfile -ppn 5 /home/satoken/ihk+mckernel/bin/mcexec -n 3 ./test_prog.sh
+getting CPU set for partitioned execution: Invalid argument
+getting CPU set for partitioned execution: Invalid argument
+test_prog is started.
+test_prog is started.
+test_prog is started.
+*** C929T02 PASSED ******************************
+
+*** C929T03 start *******************************
+mpirun -f ./hostfile -ppn 3 /home/satoken/ihk+mckernel/bin/mcexec -n 5 ./test_prog.sh
+getting CPU set for partitioned execution: Connection timed out
+getting CPU set for partitioned execution: Connection timed out
+getting CPU set for partitioned execution: Connection timed out
+*** C929T03 PASSED ******************************
+
+*** C929T04 start *******************************
+mpirun -f ./hostfile -ppn 6 /home/satoken/ihk+mckernel/bin/mcexec -n 3 ./test_prog.sh
+getting CPU set for partitioned execution: Invalid argument
+getting CPU set for partitioned execution: Invalid argument
+getting CPU set for partitioned execution: Invalid argument
+test_prog is started.
+test_prog is started.
+test_prog is started.
+*** C929T04 PASSED ******************************
+
+*** C929T05 start *******************************
+mpirun -f ./hostfile -ppn 250 /home/satoken/ihk+mckernel/bin/mcexec -n 250 ./test_prog.sh
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+error: nr_processes can't exceed nr. of CPUs
+...
+*** C929T05 PASSED ******************************
+
+*** C929T06 start *******************************
+mpirun -f ./hostfile -ppn 5 /home/satoken/ihk+mckernel/bin/mcexec -n 5 ./test_prog.sh
+[OK] exec: 1
+[OK] exec: 2
+[OK] exec: 3
+[OK] exec: 4
+[OK] exec: 5
+[OK] exec: 6
+[OK] exec: 7
+[OK] exec: 8
+[OK] exec: 9
+[OK] exec: 10
+[OK] exec: 11
+[OK] exec: 12
+[OK] exec: 13
+[OK] exec: 14
+[OK] exec: 15
+[OK] exec: 16
+[OK] exec: 17
+[OK] exec: 18
+[OK] exec: 19
+[OK] exec: 20
+*** C929T06 PASSED ******************************
+