configure.ac: Update version number

Change-Id: Ia497306551aa103d80eb5a307ca7196940ea7e14

arch/arm64/kernel/syscall.c

@@ -1801,8 +1801,9 @@ SYSCALL_DECLARE(mmap)
 		goto out;
 	}
 
-	if ((flags & MAP_FIXED) && ((addr < region->user_start)
-			|| (region->user_end <= addr))) {
+	if (addr < region->user_start
+			|| region->user_end <= addr
+			|| len > (region->user_end - region->user_start)) {
 		ekprintf("sys_mmap(%lx,%lx,%x,%x,%x,%lx):ENOMEM\n",
 				addr0, len0, prot, flags0, fd, off0);
 		error = -ENOMEM;

arch/x86_64/kernel/syscall.c

@@ -1570,8 +1570,9 @@ SYSCALL_DECLARE(mmap)
 		goto out;
 	}
 
-	if ((flags & MAP_FIXED) && ((addr < region->user_start)
-			|| (region->user_end <= addr))) {
+	if (addr < region->user_start
+			|| region->user_end <= addr
+			|| len > (region->user_end - region->user_start)) {
 		ekprintf("sys_mmap(%lx,%lx,%x,%x,%x,%lx):ENOMEM\n",
 				addr0, len0, prot, flags0, fd, off0);
 		error = -ENOMEM;

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

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

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

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

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

@@ -12,6 +12,7 @@
 # the same set of resources as it used previously.
 # Note that the script does not output anything unless an error occurs.
 
+ret=1
 prefix="@prefix@"
 BINDIR="${prefix}/bin"
 SBINDIR="${prefix}/sbin"
@@ -204,7 +205,8 @@ error_exit() {
 		;;
 	esac
 
-	exit 1
+	# Propagate exit status if any
+	exit $ret
 }
 
 ihk_ikc_irq_core=0
@@ -255,7 +257,11 @@ fi
 
 # Remove mcoverlay if loaded
 if [ "$enable_mcoverlay" == "yes" ]; then
-    . ${SBINDIR}/mcoverlay-destroy.sh
+    ${SBINDIR}/mcoverlay-destroy.sh
+    ret=$?
+    if [ $ret -ne 0 ]; then
+	error_exit "initial"
+    fi
 fi
 
 # Stop irqbalance
@@ -458,7 +464,11 @@ fi
 
 # Overlay /proc, /sys with McKernel specific contents
 if [ "$enable_mcoverlay" == "yes" ]; then
-    . ${SBINDIR}/mcoverlay-create.sh
+    ${SBINDIR}/mcoverlay-create.sh
+    ret=$?
+    if [ $ret -ne 0 ]; then
+	error_exit "os_created"
+    fi
 fi
 
 # Start irqbalance with CPUs and IRQ for McKernel banned

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

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

configure

@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.69 for mckernel 1.5.0.
+# Generated by GNU Autoconf 2.69 for mckernel 1.5.1.
 #
 #
 # Copyright (C) 1992-1996, 1998-2012 Free Software Foundation, Inc.
@@ -577,8 +577,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='mckernel'
 PACKAGE_TARNAME='mckernel'
-PACKAGE_VERSION='1.5.0'
-PACKAGE_STRING='mckernel 1.5.0'
+PACKAGE_VERSION='1.5.1'
+PACKAGE_STRING='mckernel 1.5.1'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -1262,7 +1262,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-\`configure' configures mckernel 1.5.0 to adapt to many kinds of systems.
+\`configure' configures mckernel 1.5.1 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1323,7 +1323,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of mckernel 1.5.0:";;
+     short | recursive ) echo "Configuration of mckernel 1.5.1:";;
    esac
   cat <<\_ACEOF
 
@@ -1431,7 +1431,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-mckernel configure 1.5.0
+mckernel configure 1.5.1
 generated by GNU Autoconf 2.69
 
 Copyright (C) 2012 Free Software Foundation, Inc.
@@ -1729,7 +1729,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by mckernel $as_me 1.5.0, which was
+It was created by mckernel $as_me 1.5.1, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   $ $0 $@
@@ -2082,11 +2082,11 @@ ac_compiler_gnu=$ac_cv_c_compiler_gnu
 
 
 
-IHK_VERSION=1.5.0
-MCKERNEL_VERSION=1.5.0
+IHK_VERSION=1.5.1
+MCKERNEL_VERSION=1.5.1
 DCFA_VERSION=DCFA_VERSION_m4
-IHK_RELEASE_DATE=2018-04-05
-MCKERNEL_RELEASE_DATE=2018-04-05
+IHK_RELEASE_DATE=2018-07-09
+MCKERNEL_RELEASE_DATE=2018-07-09
 DCFA_RELEASE_DATE=DCFA_RELEASE_DATE_m4
 
 
@@ -5585,7 +5585,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by mckernel $as_me 1.5.0, which was
+This file was extended by mckernel $as_me 1.5.1, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -5647,7 +5647,7 @@ _ACEOF
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`"
 ac_cs_version="\\
-mckernel config.status 1.5.0
+mckernel config.status 1.5.1
 configured by $0, generated by GNU Autoconf 2.69,
   with options \\"\$ac_cs_config\\"
 

configure.ac

@@ -1,9 +1,9 @@
 # configure.ac COPYRIGHT FUJITSU LIMITED 2015-2016
 AC_PREREQ(2.63)
-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
+m4_define([IHK_VERSION_m4],[1.5.1])dnl
+m4_define([MCKERNEL_VERSION_m4],[1.5.1])dnl
+m4_define([IHK_RELEASE_DATE_m4],[2018-07-09])dnl
+m4_define([MCKERNEL_RELEASE_DATE_m4],[2018-07-09])dnl
 
 AC_INIT([mckernel], MCKERNEL_VERSION_m4)
 

executer/include/uprotocol.h

@@ -111,10 +111,8 @@ typedef unsigned long __cpu_set_unit;
 
 struct program_load_desc {
 	int num_sections;
-	int status;
 	int cpu;
 	int pid;
-	int err;
 	int stack_prot;
 	int pgid;
 	int cred[8];
@@ -253,7 +251,6 @@ enum perf_ctrl_type {
 
 struct perf_ctrl_desc {
 	enum perf_ctrl_type ctrl_type;
-	int status;
 	union {
 		/* for SET, GET */
 		struct {

executer/kernel/mcctrl/control.c

@@ -107,6 +107,7 @@ static long mcexec_prepare_image(ihk_os_t os,
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct mcctrl_per_proc_data *ppd = NULL;
 	int num_sections;
+	int free_ikc_pointers = 1;
 
 	desc = kmalloc(sizeof(*desc), GFP_KERNEL);
 	if (!desc) {
@@ -140,9 +141,9 @@ static long mcexec_prepare_image(ihk_os_t os,
 		goto put_and_free_out;
 	}
 
-	pdesc = kmalloc(sizeof(struct program_load_desc) + 
-	                sizeof(struct program_image_section)
-	                * num_sections, GFP_KERNEL);
+	pdesc = kmalloc(sizeof(struct program_load_desc) +
+			sizeof(struct program_image_section) * num_sections,
+			GFP_KERNEL);
 	memcpy(pdesc, desc, sizeof(struct program_load_desc));
 
 	if (copy_from_user(pdesc->sections, udesc->sections,
@@ -167,10 +168,10 @@ static long mcexec_prepare_image(ihk_os_t os,
 		ret = -EFAULT;
 		goto put_and_free_out;
 	}
-	
+
 	envs = kmalloc(pdesc->envs_len, GFP_KERNEL);
 	if (copy_from_user(envs, pdesc->envs, pdesc->envs_len)) {
-		ret = -EFAULT;	
+		ret = -EFAULT;
 		goto put_and_free_out;
 	}
 
@@ -187,36 +188,24 @@ static long mcexec_prepare_image(ihk_os_t os,
 
 	dprintk("# of sections: %d\n", pdesc->num_sections);
 	dprintk("%p (%lx)\n", pdesc, isp.arg);
-	
-	pdesc->status = 0;
-	mb();
-	ret = mcctrl_ikc_send(os, pdesc->cpu, &isp);
-	if(ret < 0) {
-		printk("%s: ERROR mcctrl_ikc_send: %d\n", __FUNCTION__, ret);
-		goto put_and_free_out;
-	}
 
-	ret = wait_event_interruptible(ppd->wq_prepare, pdesc->status);
+	ret = mcctrl_ikc_send_wait(os, pdesc->cpu, &isp, 0, NULL,
+				   &free_ikc_pointers, 3, pdesc, args, envs);
 	if (ret < 0) {
-		printk("%s: ERROR after wait: %d\n", __FUNCTION__, ret);
-		goto put_and_free_out;
-	}
-
-	if (pdesc->err < 0) {
-		ret = pdesc->err;	
+		/* either send or remote prepare_process failed */
 		goto put_and_free_out;
 	}
 
 	/* Update rpgtable */
 	ppd->rpgtable = pdesc->rpgtable;
-	
-	if (copy_to_user(udesc, pdesc, sizeof(struct program_load_desc) + 
+
+	if (copy_to_user(udesc, pdesc, sizeof(struct program_load_desc) +
 	             sizeof(struct program_image_section) * num_sections)) {
-		ret = -EFAULT;	
+		ret = -EFAULT;
 		goto put_and_free_out;
 	}
 
-	dprintk("%s: pid %d, rpgtable: 0x%lx added\n", 
+	dprintk("%s: pid %d, rpgtable: 0x%lx added\n",
 		__FUNCTION__, ppd->pid, ppd->rpgtable);
 
 	ret = 0;
@@ -224,10 +213,12 @@ static long mcexec_prepare_image(ihk_os_t os,
 put_and_free_out:
 	mcctrl_put_per_proc_data(ppd);
 free_out:
-	kfree(args);
-	kfree(pdesc);
-	kfree(envs);
-	kfree(desc);
+	if (free_ikc_pointers) {
+		kfree(args);
+		kfree(pdesc);
+		kfree(envs);
+		kfree(desc);
+	}
 
 	return ret;
 }
@@ -485,25 +476,33 @@ out:
 
 static DECLARE_WAIT_QUEUE_HEAD(signalq);
 
+struct mcctrl_signal_desc {
+	struct mcctrl_signal msig;
+	struct mcctrl_wakeup_desc wakeup;
+	void *addrs[1];
+};
+
 static long mcexec_send_signal(ihk_os_t os, struct signal_desc *sigparam)
 {
 	struct ikc_scd_packet isp;
 	struct mcctrl_channel *c;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct signal_desc sig;
-	struct mcctrl_signal msig[2];
+	struct mcctrl_signal_desc *desc;
 	struct mcctrl_signal *msigp;
-	int rc;
+	int rc, do_free;
 
 	if (copy_from_user(&sig, sigparam, sizeof(struct signal_desc))) {
 		return -EFAULT;
 	}
 
-	msigp = msig;
-	if(((unsigned long)msig & 0xfffffffffffff000L) !=
-	   ((unsigned long)(msig + 1) & 0xfffffffffffff000L))
-		msigp++;
-	memset(msigp, '\0', sizeof msig);
+	desc = kmalloc(sizeof(*desc), GFP_KERNEL);
+	if (!desc) {
+		return -ENOMEM;
+	}
+
+	msigp = &desc->msig;
+	memset(msigp, '\0', sizeof(*msigp));
 	msigp->sig = sig.sig;
 	msigp->pid = sig.pid;
 	msigp->tid = sig.tid;
@@ -515,25 +514,19 @@ static long mcexec_send_signal(ihk_os_t os, struct signal_desc *sigparam)
 	isp.pid = sig.pid;
 	isp.arg = virt_to_phys(msigp);
 
-	if ((rc = mcctrl_ikc_send(os, sig.cpu, &isp)) < 0) {
+	rc = mcctrl_ikc_send_wait(os, sig.cpu, &isp, 0, &desc->wakeup,
+				  &do_free, 1, desc);
+	if (rc < 0) {
 		printk("mcexec_send_signal: mcctrl_ikc_send ret=%d\n", rc);
+		if (do_free)
+			kfree(desc);
 		return rc;
 	}
-	wait_event_interruptible(signalq, msigp->cond != 0);
 
+	kfree(desc);
 	return 0;
 }
 
-void
-sig_done(unsigned long arg, int err)
-{
-	struct mcctrl_signal *msigp;
-
-	msigp = phys_to_virt(arg);
-	msigp->cond = 1;
-	wake_up_interruptible(&signalq);
-}
-
 static long mcexec_get_cpu(ihk_os_t os)
 {
 	struct ihk_cpu_info *info;
@@ -1691,7 +1684,6 @@ int mcexec_create_per_process_data(ihk_os_t os)
 	INIT_LIST_HEAD(&ppd->wq_list);
 	INIT_LIST_HEAD(&ppd->wq_req_list);
 	INIT_LIST_HEAD(&ppd->wq_list_exact);
-	init_waitqueue_head(&ppd->wq_prepare);
 	init_waitqueue_head(&ppd->wq_procfs);
 	spin_lock_init(&ppd->wq_list_lock);
 	memset(&ppd->cpu_set, 0, sizeof(cpumask_t));
@@ -2019,58 +2011,65 @@ long mcctrl_perf_num(ihk_os_t os, unsigned long arg)
 	return 0;
 }
 
+struct mcctrl_perf_ctrl_desc {
+	struct perf_ctrl_desc desc;
+	struct mcctrl_wakeup_desc wakeup;
+	void *addrs[1];
+};
+#define wakeup_desc_of_perf_desc(_desc) \
+	(&container_of((_desc), struct mcctrl_perf_ctrl_desc, desc)->wakeup)
+
 long mcctrl_perf_set(ihk_os_t os, struct ihk_perf_event_attr *__user arg)
 {
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct ikc_scd_packet isp;
-	struct perf_ctrl_desc *perf_desc = NULL;
+	struct perf_ctrl_desc *perf_desc;
 	struct ihk_perf_event_attr attr;
 	struct ihk_cpu_info *info = ihk_os_get_cpu_info(os);
 	int ret = 0;
 	int i = 0, j = 0;
+	int need_free;
 
 	for (i = 0; i < usrdata->perf_event_num; i++) {
-		if (copy_from_user(&attr, &arg[i], sizeof(struct ihk_perf_event_attr))) {
-			printk("%s: error: copying ihk_perf_event_attr from user\n",
-			       __FUNCTION__);
+		ret = copy_from_user(&attr, &arg[i],
+				     sizeof(struct ihk_perf_event_attr));
+		if (ret) {
+			pr_err("%s: error: copying ihk_perf_event_attr from user\n",
+			       __func__);
 			return -EINVAL;
 		}
 
-		for (j = 0; j < info->n_cpus; j++) {
-			perf_desc = kmalloc(sizeof(struct perf_ctrl_desc), GFP_KERNEL);
-			if (!perf_desc) {
-				printk("%s: error: allocating perf_ctrl_desc\n",
-				       __FUNCTION__);
-				return -ENOMEM;
-			}
-			memset(perf_desc, '\0', sizeof(struct perf_ctrl_desc));
-
-			perf_desc->ctrl_type = PERF_CTRL_SET;
-			perf_desc->status = 0;
-			perf_desc->target_cntr = i;
-			perf_desc->config = attr.config;
-			perf_desc->exclude_kernel = attr.exclude_kernel;
-			perf_desc->exclude_user = attr.exclude_user;
-
-			memset(&isp, '\0', sizeof(struct ikc_scd_packet));
-			isp.msg = SCD_MSG_PERF_CTRL;
-			isp.arg = virt_to_phys(perf_desc);
-
-			if ((ret = mcctrl_ikc_send(os, j, &isp)) < 0) {
-				printk("%s: mcctrl_ikc_send ret=%d\n", __FUNCTION__, ret);
-				kfree(perf_desc);
-				return -EINVAL;
-			}
-
-			ret = wait_event_interruptible(perfctrlq, perf_desc->status);
-			if (ret < 0) {
-				printk("%s: ERROR after wait: %d\n", __FUNCTION__, ret);
-				kfree(perf_desc);
-				return -EINVAL;
-			}
-				
-			kfree(perf_desc);
+		perf_desc = kmalloc(sizeof(struct mcctrl_perf_ctrl_desc),
+				    GFP_KERNEL);
+		if (!perf_desc) {
+			return -ENOMEM;
 		}
+		memset(perf_desc, '\0', sizeof(struct perf_ctrl_desc));
+
+		perf_desc->ctrl_type = PERF_CTRL_SET;
+		perf_desc->target_cntr = i;
+		perf_desc->config = attr.config;
+		perf_desc->exclude_kernel = attr.exclude_kernel;
+		perf_desc->exclude_user = attr.exclude_user;
+
+		memset(&isp, '\0', sizeof(struct ikc_scd_packet));
+		isp.msg = SCD_MSG_PERF_CTRL;
+		isp.arg = virt_to_phys(perf_desc);
+
+		for (j = 0; j < info->n_cpus; j++) {
+			ret = mcctrl_ikc_send_wait(os, j, &isp, 0,
+					wakeup_desc_of_perf_desc(perf_desc),
+					&need_free, 1, perf_desc);
+			if (ret < 0) {
+				pr_warn("%s: mcctrl_ikc_send_wait ret=%d\n",
+					__func__, ret);
+				if (need_free)
+					kfree(perf_desc);
+				return -EINVAL;
+			}
+		}
+
+		kfree(perf_desc);
 	}
 
 	return usrdata->perf_event_num;
@@ -2080,48 +2079,46 @@ long mcctrl_perf_get(ihk_os_t os, unsigned long *__user arg)
 {
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct ikc_scd_packet isp;
-	struct perf_ctrl_desc *perf_desc = NULL;
+	struct perf_ctrl_desc *perf_desc;
 	struct ihk_cpu_info *info = ihk_os_get_cpu_info(os);
 	unsigned long value_sum = 0;
 	int ret = 0;
 	int i = 0, j = 0;
+	int need_free;
 
 	for (i = 0; i < usrdata->perf_event_num; i++) {
-		for (j = 0; j < info->n_cpus; j++) {
-			perf_desc = kmalloc(sizeof(struct perf_ctrl_desc), GFP_KERNEL);
-			if (!perf_desc) {
-				printk("%s: error: allocating perf_ctrl_desc\n",
-				       __FUNCTION__);
-				return -ENOMEM;
-			}
-			memset(perf_desc, '\0', sizeof(struct perf_ctrl_desc));
-
-			perf_desc->ctrl_type = PERF_CTRL_GET;
-			perf_desc->status = 0;
-			perf_desc->target_cntr = i;
-
-			memset(&isp, '\0', sizeof(struct ikc_scd_packet));
-			isp.msg = SCD_MSG_PERF_CTRL;
-			isp.arg = virt_to_phys(perf_desc);
-
-			if ((ret = mcctrl_ikc_send(os, j, &isp)) < 0) {
-				printk("%s: mcctrl_ikc_send ret=%d\n", __FUNCTION__, ret);
-				kfree(perf_desc);
-				return -EINVAL;
-			}
-
-			ret = wait_event_interruptible(perfctrlq, perf_desc->status);
-			if (ret < 0) {
-				printk("%s: ERROR after wait: %d\n", __FUNCTION__, ret);
-				kfree(perf_desc);
-				return -EINVAL;
-			}
-			value_sum += perf_desc->read_value;
-			kfree(perf_desc);
+		perf_desc = kmalloc(sizeof(struct mcctrl_perf_ctrl_desc),
+				    GFP_KERNEL);
+		if (!perf_desc) {
+			return -ENOMEM;
 		}
+		memset(perf_desc, '\0', sizeof(struct perf_ctrl_desc));
+
+		perf_desc->ctrl_type = PERF_CTRL_GET;
+		perf_desc->target_cntr = i;
+
+		memset(&isp, '\0', sizeof(struct ikc_scd_packet));
+		isp.msg = SCD_MSG_PERF_CTRL;
+		isp.arg = virt_to_phys(perf_desc);
+
+		for (j = 0; j < info->n_cpus; j++) {
+			ret = mcctrl_ikc_send_wait(os, j, &isp, 0,
+					wakeup_desc_of_perf_desc(perf_desc),
+					&need_free, 1, perf_desc);
+			if (ret < 0) {
+				pr_warn("%s: mcctrl_ikc_send_wait ret=%d\n",
+					__func__, ret);
+				if (need_free)
+					kfree(perf_desc);
+				return -EINVAL;
+			}
+
+			value_sum += perf_desc->read_value;
+		}
+		kfree(perf_desc);
 		if (copy_to_user(&arg[i], &value_sum, sizeof(unsigned long))) {
 			printk("%s: error: copying read_value to user\n",
-			       __FUNCTION__);
+			       __func__);
 			return -EINVAL;
 		}
 		value_sum = 0;
@@ -2134,46 +2131,44 @@ long mcctrl_perf_enable(ihk_os_t os)
 {
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct ikc_scd_packet isp;
-	struct perf_ctrl_desc *perf_desc = NULL;
+	struct perf_ctrl_desc *perf_desc;
 	struct ihk_cpu_info *info = ihk_os_get_cpu_info(os);
 	unsigned int cntr_mask = 0;
 	int ret = 0;
 	int i = 0, j = 0;
+	int need_free;
 
 	for (i = 0; i < usrdata->perf_event_num; i++) {
 		cntr_mask |= 1 << i;
 	}
-	for (j = 0; j < info->n_cpus; j++) {
-		perf_desc = kmalloc(sizeof(struct perf_ctrl_desc), GFP_KERNEL);
-		if (!perf_desc) {
-			printk("%s: error: allocating perf_ctrl_desc\n",
-			       __FUNCTION__);
-			return -ENOMEM;
-		}
-		memset(perf_desc, '\0', sizeof(struct perf_ctrl_desc));
-
-		perf_desc->ctrl_type = PERF_CTRL_ENABLE;
-		perf_desc->status = 0;
-		perf_desc->target_cntr_mask = cntr_mask;
-
-		memset(&isp, '\0', sizeof(struct ikc_scd_packet));
-		isp.msg = SCD_MSG_PERF_CTRL;
-		isp.arg = virt_to_phys(perf_desc);
-
-		if ((ret = mcctrl_ikc_send(os, j, &isp)) < 0) {
-			printk("%s: mcctrl_ikc_send ret=%d\n", __FUNCTION__, ret);
-			kfree(perf_desc);
-			return -EINVAL;
-		}
-
-		ret = wait_event_interruptible(perfctrlq, perf_desc->status);
-		if (ret < 0) {
-			printk("%s: ERROR after wait: %d\n", __FUNCTION__, ret);
-			kfree(perf_desc);
-			return -EINVAL;
-		}
-		kfree(perf_desc);
+	perf_desc = kmalloc(sizeof(struct mcctrl_perf_ctrl_desc), GFP_KERNEL);
+	if (!perf_desc) {
+		return -ENOMEM;
 	}
+	memset(perf_desc, '\0', sizeof(struct perf_ctrl_desc));
+
+	perf_desc->ctrl_type = PERF_CTRL_ENABLE;
+	perf_desc->target_cntr_mask = cntr_mask;
+
+	memset(&isp, '\0', sizeof(struct ikc_scd_packet));
+	isp.msg = SCD_MSG_PERF_CTRL;
+	isp.arg = virt_to_phys(perf_desc);
+
+	for (j = 0; j < info->n_cpus; j++) {
+		ret = mcctrl_ikc_send_wait(os, j, &isp, 0,
+					   wakeup_desc_of_perf_desc(perf_desc),
+					   &need_free, 1, perf_desc);
+
+		if (ret < 0) {
+			pr_warn("%s: mcctrl_ikc_send_wait ret=%d\n",
+				__func__, ret);
+			if (need_free)
+				kfree(perf_desc);
+			return -EINVAL;
+		}
+
+	}
+	kfree(perf_desc);
 
 	return 0;
 }
@@ -2182,46 +2177,43 @@ long mcctrl_perf_disable(ihk_os_t os)
 {
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
 	struct ikc_scd_packet isp;
-	struct perf_ctrl_desc *perf_desc = NULL;
+	struct perf_ctrl_desc *perf_desc;
 	struct ihk_cpu_info *info = ihk_os_get_cpu_info(os);
 	unsigned int cntr_mask = 0;
 	int ret = 0;
 	int i = 0, j = 0;
+	int need_free;
 
 	for (i = 0; i < usrdata->perf_event_num; i++) {
 		cntr_mask |= 1 << i;
 	}
-	for (j = 0; j < info->n_cpus; j++) {
-		perf_desc = kmalloc(sizeof(struct perf_ctrl_desc), GFP_KERNEL);
-		if (!perf_desc) {
-			printk("%s: error: allocating perf_ctrl_desc\n",
-			       __FUNCTION__);
-			return -ENOMEM;
-		}
-		memset(perf_desc, '\0', sizeof(struct perf_ctrl_desc));
-
-		perf_desc->ctrl_type = PERF_CTRL_DISABLE;
-		perf_desc->status = 0;
-		perf_desc->target_cntr_mask = cntr_mask;
-
-		memset(&isp, '\0', sizeof(struct ikc_scd_packet));
-		isp.msg = SCD_MSG_PERF_CTRL;
-		isp.arg = virt_to_phys(perf_desc);
-
-		if ((ret = mcctrl_ikc_send(os, j, &isp)) < 0) {
-			printk("%s: mcctrl_ikc_send ret=%d\n", __FUNCTION__, ret);
-			kfree(perf_desc);
-			return -EINVAL;
-		}
-
-		ret = wait_event_interruptible(perfctrlq, perf_desc->status);
-		if (ret < 0) {
-			printk("%s: ERROR after wait: %d\n", __FUNCTION__, ret);
-			kfree(perf_desc);
-			return -EINVAL;
-		}
-		kfree(perf_desc);
+	perf_desc = kmalloc(sizeof(struct mcctrl_perf_ctrl_desc), GFP_KERNEL);
+	if (!perf_desc) {
+		return -ENOMEM;
 	}
+	memset(perf_desc, '\0', sizeof(struct perf_ctrl_desc));
+
+	perf_desc->ctrl_type = PERF_CTRL_DISABLE;
+	perf_desc->target_cntr_mask = cntr_mask;
+
+	memset(&isp, '\0', sizeof(struct ikc_scd_packet));
+	isp.msg = SCD_MSG_PERF_CTRL;
+	isp.arg = virt_to_phys(perf_desc);
+
+	for (j = 0; j < info->n_cpus; j++) {
+		ret = mcctrl_ikc_send_wait(os, j, &isp, 0,
+				wakeup_desc_of_perf_desc(perf_desc),
+				&need_free, 1, perf_desc);
+		if (ret < 0) {
+			pr_warn("%s: mcctrl_ikc_send_wait ret=%d\n",
+				__func__, ret);
+			if (need_free)
+				kfree(perf_desc);
+			return -EINVAL;
+		}
+
+	}
+	kfree(perf_desc);
 
 	return 0;
 }
@@ -2233,15 +2225,6 @@ long mcctrl_perf_destroy(ihk_os_t os)
 	return 0;
 }
 
-void mcctrl_perf_ack(ihk_os_t os, struct ikc_scd_packet *packet)
-{
-	struct perf_ctrl_desc *perf_desc = phys_to_virt(packet->arg);
-
-	perf_desc->status = 1;
-	wake_up_interruptible(&perfctrlq);
-
-}
-
 /* Compose LWK-specific rusage structure */
 long mcctrl_getrusage(ihk_os_t ihk_os, struct mcctrl_ioctl_getrusage_desc *__user _desc)
 {
@@ -2925,28 +2908,6 @@ long __mcctrl_control(ihk_os_t os, unsigned int req, unsigned long arg,
 	return -EINVAL;
 }
 
-void mcexec_prepare_ack(ihk_os_t os, unsigned long arg, int err)
-{
-	struct program_load_desc *desc = phys_to_virt(arg);
-	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
-	struct mcctrl_per_proc_data *ppd = NULL;
-
-	ppd = mcctrl_get_per_proc_data(usrdata, desc->pid);
-	if (!ppd) {
-		printk("%s: ERROR: no per process data for PID %d\n",
-			__FUNCTION__, desc->pid);
-		return;
-	}
-
-	desc->err = err;
-	desc->status = 1;
-	mb();
-	
-	wake_up_all(&ppd->wq_prepare);
-	mcctrl_put_per_proc_data(ppd);
-}
-
-
 /* Per-CPU register manipulation functions */
 struct mcctrl_os_cpu_response {
 	int done;

executer/kernel/mcctrl/ikc.c

@@ -49,15 +49,123 @@
 
 //struct mcctrl_channel *channels;
 
-void mcexec_prepare_ack(ihk_os_t os, unsigned long arg, int err);
 static void mcctrl_ikc_init(ihk_os_t os, int cpu, unsigned long rphys, struct ihk_ikc_channel_desc *c);
 int mcexec_syscall(struct mcctrl_usrdata *ud, struct ikc_scd_packet *packet);
 void sig_done(unsigned long arg, int err);
-void mcctrl_perf_ack(ihk_os_t os, struct ikc_scd_packet *packet);
 void mcctrl_os_read_write_cpu_response(ihk_os_t os,
 		struct ikc_scd_packet *pisp);
 void mcctrl_eventfd(ihk_os_t os, struct ikc_scd_packet *pisp);
 
+/* Assumes usrdata->wakeup_descs_lock taken */
+static void mcctrl_wakeup_desc_cleanup(ihk_os_t os,
+		struct mcctrl_wakeup_desc *desc)
+{
+	int i;
+
+	list_del(&desc->chain);
+
+	for (i = 0; i < desc->free_addrs_count; i++) {
+		kfree(desc->free_addrs[i]);
+	}
+}
+
+static void mcctrl_wakeup_cb(ihk_os_t os, struct ikc_scd_packet *packet)
+{
+	struct mcctrl_wakeup_desc *desc = packet->reply;
+
+
+	WRITE_ONCE(desc->err, packet->err);
+
+	/*
+	 * Check if the other side is still waiting, and signal it we're done.
+	 *
+	 * Setting status needs to be done last because the other side could
+	 * wake up opportunistically between this set and the wake_up call.
+	 *
+	 * If the other side is no longer waiting, free the memory that was
+	 * left for us.
+	 */
+	if (cmpxchg(&desc->status, 0, 1)) {
+		struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
+		unsigned long flags;
+
+		spin_lock_irqsave(&usrdata->wakeup_descs_lock, flags);
+		mcctrl_wakeup_desc_cleanup(os, desc);
+		spin_unlock_irqrestore(&usrdata->wakeup_descs_lock, flags);
+		return;
+	}
+
+	wake_up_interruptible(&desc->wq);
+}
+
+int mcctrl_ikc_send_wait(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp,
+		long int timeout, struct mcctrl_wakeup_desc *desc,
+		int *do_frees, int free_addrs_count, ...)
+{
+	int ret, i;
+	int alloc_desc = (desc == NULL);
+	va_list ap;
+
+	if (free_addrs_count)
+		*do_frees = 1;
+	if (alloc_desc)
+		desc = kmalloc(sizeof(struct mcctrl_wakeup_desc) +
+			       (free_addrs_count + 1) * sizeof(void *),
+			       GFP_KERNEL);
+	if (!desc) {
+		pr_warn("%s: Could not allocate wakeup descriptor", __func__);
+		return -ENOMEM;
+	}
+	pisp->reply = desc;
+	va_start(ap, free_addrs_count);
+	for (i = 0; i < free_addrs_count; i++) {
+		desc->free_addrs[i] = va_arg(ap, void*);
+	}
+	va_end(ap);
+	if (alloc_desc)
+		desc->free_addrs[free_addrs_count++] = desc;
+	desc->free_addrs_count = free_addrs_count;
+	init_waitqueue_head(&desc->wq);
+	WRITE_ONCE(desc->err, 0);
+	WRITE_ONCE(desc->status, 0);
+
+	ret = mcctrl_ikc_send(os, cpu, pisp);
+	if (ret < 0) {
+		pr_warn("%s: mcctrl_ikc_send failed: %d\n", __func__, ret);
+		kfree(desc);
+		return ret;
+	}
+
+	if (timeout) {
+		ret = wait_event_interruptible_timeout(desc->wq,
+			desc->status, timeout);
+	} else {
+		ret = wait_event_interruptible(desc->wq, desc->status);
+	}
+
+	/*
+	 * Check if wait aborted (signal..) or timed out, and notify
+	 * the callback it will need to free things for us
+	 */
+	if (!cmpxchg(&desc->status, 0, 1)) {
+		struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
+		unsigned long flags;
+
+		spin_lock_irqsave(&usrdata->wakeup_descs_lock, flags);
+		list_add(&desc->chain, &usrdata->wakeup_descs_list);
+		spin_unlock_irqrestore(&usrdata->wakeup_descs_lock, flags);
+		if (free_addrs_count)
+			*do_frees = 0;
+		return ret < 0 ? ret : -ETIME;
+	}
+
+	ret = READ_ONCE(desc->err);
+	if (alloc_desc)
+		kfree(desc);
+	return ret;
+}
+
+
 /* XXX: this runs in atomic context! */
 static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
                                   void *__packet, void *__os)
@@ -72,11 +180,9 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		break;
 
 	case SCD_MSG_PREPARE_PROCESS_ACKED:
-		mcexec_prepare_ack(__os, pisp->arg, 0);
-		break;
-
-	case SCD_MSG_PREPARE_PROCESS_NACKED:
-		mcexec_prepare_ack(__os, pisp->arg, pisp->err);
+	case SCD_MSG_PERF_ACK:
+	case SCD_MSG_SEND_SIGNAL_ACK:
+		mcctrl_wakeup_cb(__os, pisp);
 		break;
 
 	case SCD_MSG_SYSCALL_ONESIDE:
@@ -87,9 +193,6 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		procfs_answer(usrdata, pisp->pid);
 		break;
 
-	case SCD_MSG_SEND_SIGNAL:
-		sig_done(pisp->arg, pisp->err);
-		break;
 
 	case SCD_MSG_SYSFS_REQ_CREATE:
 	case SCD_MSG_SYSFS_REQ_MKDIR:
@@ -113,10 +216,6 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		get_vdso_info(__os, pisp->arg);
 		break;
 
-	case SCD_MSG_PERF_ACK:
-		mcctrl_perf_ack(__os, pisp);
-		break;
-
 	case SCD_MSG_CPU_RW_REG_RESP:
 		mcctrl_os_read_write_cpu_response(__os, pisp);
 		break;
@@ -157,10 +256,15 @@ static int dummy_packet_handler(struct ihk_ikc_channel_desc *c,
 
 int mcctrl_ikc_send(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp)
 {
-	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
+	struct mcctrl_usrdata *usrdata;
 
-	if (cpu < 0 || os == NULL || usrdata == NULL ||
-	    cpu >= usrdata->num_channels || !usrdata->channels[cpu].c) {
+	if (cpu < 0 || os == NULL) {
+		return -EINVAL;
+	}
+
+	usrdata = ihk_host_os_get_usrdata(os);
+	if (usrdata == NULL || cpu >= usrdata->num_channels ||
+	    !usrdata->channels[cpu].c) {
 		return -EINVAL;
 	}
 	return ihk_ikc_send(usrdata->channels[cpu].c, pisp, 0);
@@ -354,6 +458,8 @@ int prepare_ikc_channels(ihk_os_t os)
 	mutex_init(&usrdata->part_exec.lock);
 	INIT_LIST_HEAD(&usrdata->part_exec.pli_list);
 	usrdata->part_exec.nr_processes = -1;
+	INIT_LIST_HEAD(&usrdata->wakeup_descs_list);
+	spin_lock_init(&usrdata->wakeup_descs_lock);
 
 	return 0;
 
@@ -375,7 +481,9 @@ void __destroy_ikc_channel(ihk_os_t os, struct mcctrl_channel *pmc)
 void destroy_ikc_channels(ihk_os_t os)
 {
 	int i;
+	unsigned long flags;
 	struct mcctrl_usrdata *usrdata = ihk_host_os_get_usrdata(os);
+	struct mcctrl_wakeup_desc *mwd_entry, *mwd_next;
 
 	if (!usrdata) {
 		printk("%s: WARNING: no mcctrl_usrdata found\n", __FUNCTION__);
@@ -395,6 +503,12 @@ void destroy_ikc_channels(ihk_os_t os)
 			ihk_ikc_destroy_channel(usrdata->ikc2linux[i]);
 		}
 	}
+	spin_lock_irqsave(&usrdata->wakeup_descs_lock, flags);
+	list_for_each_entry_safe(mwd_entry, mwd_next,
+			&usrdata->wakeup_descs_list, chain) {
+		mcctrl_wakeup_desc_cleanup(os, mwd_entry);
+	}
+	spin_unlock_irqrestore(&usrdata->wakeup_descs_lock, flags);
 
 	kfree(usrdata->channels);
 	kfree(usrdata->ikc2linux);

executer/kernel/mcctrl/mcctrl.h

@@ -48,7 +48,6 @@
 
 #define SCD_MSG_PREPARE_PROCESS         0x1
 #define SCD_MSG_PREPARE_PROCESS_ACKED   0x2
-#define SCD_MSG_PREPARE_PROCESS_NACKED  0x7
 #define SCD_MSG_SCHEDULE_PROCESS        0x3
 #define SCD_MSG_WAKE_UP_SYSCALL_THREAD  0x14
 
@@ -56,7 +55,8 @@
 #define SCD_MSG_INIT_CHANNEL_ACKED      0x6
 
 #define SCD_MSG_SYSCALL_ONESIDE         0x4
-#define SCD_MSG_SEND_SIGNAL     	0x8
+#define SCD_MSG_SEND_SIGNAL             0x7
+#define SCD_MSG_SEND_SIGNAL_ACK         0x8
 #define SCD_MSG_CLEANUP_PROCESS         0x9
 #define SCD_MSG_GET_VDSO_INFO           0xa
 
@@ -128,6 +128,7 @@ enum mcctrl_os_cpu_operation {
 struct ikc_scd_packet {
 	int msg;
 	int err;
+	void *reply;
 	union {
 		/* for traditional SCD_MSG_* */
 		struct {
@@ -163,9 +164,10 @@ struct ikc_scd_packet {
 			int eventfd_type;
 		};
 	};
-	char padding[12];
+	char padding[8];
 };
 
+
 struct mcctrl_priv { 
 	ihk_os_t os;
 	struct program_load_desc *desc;
@@ -231,7 +233,6 @@ struct mcctrl_per_proc_data {
 	struct list_head wq_list_exact;  /* These requests come from IKC IRQ handler targeting a particular thread */
 
 	ihk_spinlock_t wq_list_lock;
-	wait_queue_head_t wq_prepare;
 	wait_queue_head_t wq_procfs;
 
 	struct list_head per_thread_data_hash[MCCTRL_PER_THREAD_DATA_HASH_SIZE];
@@ -342,6 +343,8 @@ struct mcctrl_usrdata {
 	wait_queue_head_t wq_procfs;
 	struct list_head per_proc_data_hash[MCCTRL_PER_PROC_DATA_HASH_SIZE];
 	rwlock_t per_proc_data_hash_lock[MCCTRL_PER_PROC_DATA_HASH_SIZE];
+	struct list_head wakeup_descs_list;
+	spinlock_t wakeup_descs_lock;
 
 	void **keys;
 	struct sysfsm_data sysfsm_data;
@@ -367,6 +370,34 @@ int mcctrl_ikc_send(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp);
 int mcctrl_ikc_send_msg(ihk_os_t os, int cpu, int msg, int ref, unsigned long arg);
 int mcctrl_ikc_is_valid_thread(ihk_os_t os, int cpu);
 
+struct mcctrl_wakeup_desc {
+	int status;
+	int err;
+	wait_queue_head_t wq;
+	struct list_head chain;
+	int free_addrs_count;
+	void *free_addrs[];
+};
+
+/* ikc query-and-wait helper
+ *
+ * Arguments:
+ *  - os, cpu and pisp as per mcctl_ikc_send()
+ *  - timeout: time to wait for reply in ms
+ *  - desc: if set, memory area to be used for desc.
+ *    Care must be taken to leave room for variable-length array.
+ *  - do_free: returns bool that specify if the caller should free
+ *    its memory on error (e.g. if ikc_send failed in the first place,
+ *    the reply has no chance of coming and memory should be free)
+ *    Always true on success.
+ *  - free_addrs_count & ...: addresses to kmalloc'd pointers that
+ *    are referenced in the message and must be left intact if we
+ *    abort to timeout/signal.
+ */
+int mcctrl_ikc_send_wait(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp,
+	long int timeout, struct mcctrl_wakeup_desc *desc,
+	int *do_frees, int free_addrs_count, ...);
+
 ihk_os_t osnum_to_os(int n);
 
 /* syscall.c */

kernel/host.c

@@ -597,14 +597,9 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 
 	case SCD_MSG_PREPARE_PROCESS:
 
-		if((rc = process_msg_prepare_process(packet->arg)) == 0){
-			pckt.msg = SCD_MSG_PREPARE_PROCESS_ACKED;
-			pckt.err = 0;
-		}
-		else{
-			pckt.msg = SCD_MSG_PREPARE_PROCESS_NACKED;
-			pckt.err = rc;
-		}
+		pckt.err = process_msg_prepare_process(packet->arg);
+		pckt.msg = SCD_MSG_PREPARE_PROCESS_ACKED;
+		pckt.reply = packet->reply;
 		pckt.ref = packet->ref;
 		pckt.arg = packet->arg;
 		syscall_channel_send(resp_channel, &pckt);
@@ -660,10 +655,11 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		memcpy(&info, sp, sizeof(struct mcctrl_signal));
 		ihk_mc_unmap_virtual(sp, 1, 0);
 		ihk_mc_unmap_memory(NULL, pp, sizeof(struct mcctrl_signal));
-		pckt.msg = SCD_MSG_SEND_SIGNAL;
+		pckt.msg = SCD_MSG_SEND_SIGNAL_ACK;
 		pckt.err = 0;
 		pckt.ref = packet->ref;
 		pckt.arg = packet->arg;
+		pckt.reply = packet->reply;
 		syscall_channel_send(resp_channel, &pckt);
 
 		rc = do_kill(NULL, info.pid, info.tid, info.sig, &info.info, 0);
@@ -736,6 +732,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		pckt.msg = SCD_MSG_PERF_ACK;
 		pckt.err = 0;
 		pckt.arg = packet->arg;
+		pckt.reply = packet->reply;
 		ihk_ikc_send(resp_channel, &pckt, 0);
 
 		ret = 0;

kernel/include/kmalloc.h

@@ -34,7 +34,7 @@ void *__kmalloc(int size, ihk_mc_ap_flag flag);
 void __kfree(void *ptr);
 
 int _memcheck(void *ptr, char *msg, char *file, int line, int free);
-int memcheckall();
+int memcheckall(void);
 int freecheck(int runcount);
 void kmalloc_consolidate_free_list(void);
 

kernel/include/process.h

@@ -825,8 +825,8 @@ struct process *find_process(int pid, struct mcs_rwlock_node_irqsave *lock);
 void process_unlock(struct process *proc, struct mcs_rwlock_node_irqsave *lock);
 void chain_process(struct process *);
 void chain_thread(struct thread *);
-void proc_init();
-void set_timer();
+void proc_init(void);
+void set_timer(void);
 struct sig_pending *hassigpending(struct thread *thread);
 
 #endif

kernel/include/syscall.h

@@ -30,7 +30,6 @@
 
 #define SCD_MSG_PREPARE_PROCESS         0x1
 #define SCD_MSG_PREPARE_PROCESS_ACKED   0x2
-#define SCD_MSG_PREPARE_PROCESS_NACKED  0x7
 #define SCD_MSG_SCHEDULE_PROCESS        0x3
 #define SCD_MSG_WAKE_UP_SYSCALL_THREAD  0x14
 
@@ -38,8 +37,9 @@
 #define SCD_MSG_INIT_CHANNEL_ACKED      0x6
 
 #define SCD_MSG_SYSCALL_ONESIDE         0x4
-#define SCD_MSG_SEND_SIGNAL		0x8
-#define	SCD_MSG_CLEANUP_PROCESS		0x9
+#define SCD_MSG_SEND_SIGNAL             0x7
+#define SCD_MSG_SEND_SIGNAL_ACK         0x8
+#define SCD_MSG_CLEANUP_PROCESS         0x9
 #define SCD_MSG_GET_VDSO_INFO           0xa
 
 #define SCD_MSG_GET_CPU_MAPPING         0xc
@@ -168,10 +168,8 @@ typedef unsigned long __cpu_set_unit;
 
 struct program_load_desc {
 	int num_sections;
-	int status;
 	int cpu;
 	int pid;
-	int err;
 	int stack_prot;
 	int pgid;
 	int cred[8];
@@ -241,6 +239,7 @@ enum mcctrl_os_cpu_operation {
 struct ikc_scd_packet {
 	int msg;
 	int err;
+	void *reply;
 	union {
 		/* for traditional SCD_MSG_* */
 		struct {
@@ -276,7 +275,7 @@ struct ikc_scd_packet {
 			int eventfd_type;
 		};
 	};
-	char padding[12];
+	char padding[8];
 };
 
 #define IHK_SCD_REQ_THREAD_SPINNING         0
@@ -340,7 +339,7 @@ struct syscall_post {
 #define SYSCALL_FOOTER return do_syscall(&request, ihk_mc_get_processor_id(), 0)
 
 extern long do_syscall(struct syscall_request *req, int cpu, int pid);
-extern int obtain_clone_cpuid();
+int obtain_clone_cpuid(cpu_set_t *cpu_set);
 extern long syscall_generic_forwarding(int n, ihk_mc_user_context_t *ctx);
 
 #define DECLARATOR(number,name)		__NR_##name = number,
@@ -458,7 +457,7 @@ static inline unsigned long timespec_to_jiffy(const struct timespec *ats)
 	return ats->tv_sec * 100 + ats->tv_nsec / 10000000;
 }
 
-void reset_cputime();
+void reset_cputime(void);
 void set_cputime(int mode);
 int do_munmap(void *addr, size_t len);
 intptr_t do_mmap(intptr_t addr0, size_t len0, int prot, int flags, int fd,
@@ -520,7 +519,6 @@ enum perf_ctrl_type {
 
 struct perf_ctrl_desc {
 	enum perf_ctrl_type ctrl_type;
-	int status;
 	union {
 		/* for SET, GET */
 		struct {

kernel/mem.c

@@ -1746,11 +1746,16 @@ static void rusage_init()
 {
 	int npages;
 	unsigned long phys;
+	const struct ihk_mc_cpu_info *cpu_info = ihk_mc_get_cpu_info();
+
+	if (!cpu_info) {
+		panic("rusage_init: PANIC: ihk_mc_get_cpu_info returned NULL");
+	}
 
 	npages = (sizeof(struct rusage_global) + PAGE_SIZE -1) >> PAGE_SHIFT;
 	rusage = ihk_mc_alloc_pages(npages, IHK_MC_AP_CRITICAL);
 	memset(rusage, 0, npages * PAGE_SIZE);
-	rusage->num_processors = num_processors;
+	rusage->num_processors = cpu_info->ncpus;
 	rusage->num_numa_nodes = ihk_mc_get_nr_numa_nodes();
 	rusage->ns_per_tsc = ihk_mc_get_ns_per_tsc();
 	phys = virt_to_phys(rusage);

kernel/procfs.c

@@ -369,6 +369,7 @@ void process_procfs_request(struct ikc_scd_packet *rpacket)
 					0UL,
 					0UL,
 					0,
+					range->memobj && range->memobj->path ? range->memobj->path :
 					range->start ==
 						(unsigned long)vm->vdso_addr ? "[vdso]" :
 					range->start ==

kernel/syscall.c

@@ -2135,6 +2135,9 @@ static void munmap_all(void)
 	/* free vm_ranges which do_munmap() failed to remove. */
 	free_process_memory_ranges(thread->vm);
 
+	/* reset vm_region's map area */
+	thread->vm->region.map_end = thread->vm->region.map_start;
+
 	return;
 } /* munmap_all() */
 
@@ -2237,21 +2240,22 @@ SYSCALL_DECLARE(execve)
 		goto end;
 	}
 
-	if (cpu_local_var(current)->proc->ptrace) {
+	if (thread->proc->ptrace) {
 		ihk_mc_syscall_ret(ctx) = 0;
-		ptrace_syscall_event(cpu_local_var(current));
+		ptrace_syscall_event(thread);
 	}
 
 	/* Unmap all memory areas of the process, userspace will be gone */
 	munmap_all();
 
-	ihk_mc_init_user_process(&cpu_local_var(current)->ctx, 
-			&cpu_local_var(current)->uctx,
-			((char *)cpu_local_var(current)) + 
+	/* Code assumes no process switch from here on */
+	preempt_disable();
+	ihk_mc_init_user_process(&thread->ctx, &thread->uctx,
+			((char *)thread) +
 			KERNEL_STACK_NR_PAGES * PAGE_SIZE, desc->entry, 0);
 
 	/* Create virtual memory ranges and update args/envs */
-	if (prepare_process_ranges_args_envs(cpu_local_var(current), desc, desc, 
+	if (prepare_process_ranges_args_envs(thread, desc, desc,
 				PTATTR_NO_EXECUTE | PTATTR_WRITABLE | PTATTR_FOR_USER,
 				argv_flat, argv_flat_len, envp_flat, envp_flat_len) != 0) {
 		kprintf("execve(): PANIC: preparing ranges, args, envs, stack\n");
@@ -2286,7 +2290,7 @@ SYSCALL_DECLARE(execve)
 	clear_fp_regs(thread);
 #endif /* POSTK_DEBUG_TEMP_FIX_19 */
 
-	error = ptrace_report_exec(cpu_local_var(current));
+	error = ptrace_report_exec(thread);
 	if(error) {
 		kprintf("execve(): ERROR: ptrace_report_exec()\n");
 	}
@@ -2309,10 +2313,16 @@ end:
 		/* 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));
+		preempt_enable();
 
-		ihk_mc_switch_context(NULL, &cpu_local_var(current)->ctx, 
-			cpu_local_var(current));
+		ihk_mc_switch_context(NULL, &thread->ctx, thread);
+
+		/* not reached */
+		return -EFAULT;
 	}
+
+	/* no preempt_enable, errors can only happen before we disabled it */
+
 	return ret;
 }
 
@@ -2467,7 +2477,7 @@ retry_tid:
 			
 			/* TODO: clean-up new */
 			release_cpuid(cpuid);
-			return -EFAULT;
+			return newproc->pid;
 		}
 
 		/* In a single threaded process TID equals to PID */
@@ -6786,22 +6796,14 @@ SYSCALL_DECLARE(sched_getaffinity)
 	int ret;
 
 	dkprintf("%s() len: %d, mask: %p\n", __FUNCTION__, len, u_cpu_set);
-#ifdef POSTK_DEBUG_TEMP_FIX_5 /* sched_getaffinity arguments check add (S64FX_10) */
 	if (len * 8 < num_processors) {
-		kprintf("%s:%d Too small buffer.\n", __FILE__, __LINE__);
+		dkprintf("%s: Too small buffer.\n", __func__);
 		return -EINVAL;
 	}
 	if (len & (sizeof(unsigned long)-1)) {
-		kprintf("%s:%d Size not align to unsigned long.\n", __FILE__, __LINE__);
+		dkprintf("%s: Size not align to unsigned long.\n", __func__);
 		return -EINVAL;
 	}
-#else /* POSTK_DEBUG_TEMP_FIX_5 */
-	if (!len || u_cpu_set == (cpu_set_t *)-1)
-		return -EINVAL;
-
-	if ((len * BITS_PER_BYTE) < __CPU_SETSIZE)
-		return -EINVAL;
-#endif /* POSTK_DEBUG_TEMP_FIX_5 */
 
 	len = MIN2(len, sizeof(k_cpu_set));
 
@@ -6836,11 +6838,8 @@ SYSCALL_DECLARE(sched_getaffinity)
 	}
 
 	dkprintf("%s() len: %d, ret: %d\n", __FUNCTION__, len, ret);
-#ifdef POSTK_DEBUG_TEMP_FIX_58 /* sched_getafifnity return value fix */
+
 	return ret;
-#else /* POSTK_DEBUG_TEMP_FIX_58 */
-	return len;
-#endif /* POSTK_DEBUG_TEMP_FIX_58 */
 }
 
 SYSCALL_DECLARE(get_cpu_id)
@@ -7705,8 +7704,8 @@ SYSCALL_DECLARE(mremap)
 	const size_t newsize0 = ihk_mc_syscall_arg2(ctx);
 	const int flags = ihk_mc_syscall_arg3(ctx);
 	const uintptr_t newaddr = ihk_mc_syscall_arg4(ctx);
-	const ssize_t oldsize = (oldsize0 + PAGE_SIZE - 1) & PAGE_MASK;
-	const ssize_t newsize = (newsize0 + PAGE_SIZE - 1) & PAGE_MASK;
+	const size_t oldsize = (oldsize0 + PAGE_SIZE - 1) & PAGE_MASK;
+	const size_t newsize = (newsize0 + PAGE_SIZE - 1) & PAGE_MASK;
 	const uintptr_t oldstart = oldaddr;
 	const uintptr_t oldend = oldstart + oldsize;
 	struct thread *thread = cpu_local_var(current);
@@ -7725,9 +7724,9 @@ SYSCALL_DECLARE(mremap)
 			oldaddr, oldsize0, newsize0, flags, newaddr);
 	ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
 
+	/* check arguments */
 	if ((oldaddr & ~PAGE_MASK)
-			|| (oldsize < 0)
-			|| (newsize <= 0)
+			|| (newsize == 0)
 			|| (flags & ~(MREMAP_MAYMOVE | MREMAP_FIXED))
 			|| ((flags & MREMAP_FIXED)
 				&& !(flags & MREMAP_MAYMOVE))
@@ -7740,6 +7739,24 @@ SYSCALL_DECLARE(mremap)
 		goto out;
 	}
 
+	if (oldend < oldstart) {
+		error = -EINVAL;
+		ekprintf("sys_mremap(%#lx,%#lx,%#lx,%#x,%#lx):"
+				"old range overflow. %d\n",
+				oldaddr, oldsize0, newsize0,
+				flags, newaddr, error);
+		goto out;
+	}
+
+	if (newsize > (vm->region.user_end - vm->region.user_start)) {
+		error = -ENOMEM;
+		ekprintf("sys_mremap(%#lx,%#lx,%#lx,%#x,%#lx):"
+				"cannot allocate. %d\n",
+				oldaddr, oldsize0, newsize0,
+				flags, newaddr, error);
+		goto out;
+	}
+
 	/* check original mapping */
 	range = lookup_process_memory_range(vm, oldstart, oldstart+PAGE_SIZE);
 	if (!range || (oldstart < range->start) || (range->end < oldend)
@@ -7754,15 +7771,6 @@ SYSCALL_DECLARE(mremap)
 		goto out;
 	}
 
-	if (oldend < oldstart) {
-		error = -EINVAL;
-		ekprintf("sys_mremap(%#lx,%#lx,%#lx,%#x,%#lx):"
-				"old range overflow. %d\n",
-				oldaddr, oldsize0, newsize0, flags, newaddr,
-				error);
-		goto out;
-	}
-
 	/* determine new mapping range */
 	need_relocate = 0;
 	if (flags & MREMAP_FIXED) {

lib/include/ihk/cpu.h

@@ -60,9 +60,9 @@ void ihk_mc_boot_cpu(int cpuid, unsigned long pc);
 int ihk_mc_get_processor_id(void);
 int ihk_mc_get_hardware_processor_id(void);
 int ihk_mc_get_numa_id(void);
-int ihk_mc_get_nr_cores();
-int ihk_mc_get_nr_linux_cores();
-int ihk_mc_get_osnum();
+int ihk_mc_get_nr_cores(void);
+int ihk_mc_get_nr_linux_cores(void);
+int ihk_mc_get_osnum(void);
 int ihk_mc_get_core(int id, unsigned long *linux_core_id, unsigned long *apic_id,
                     int *numa_id);
 int ihk_mc_get_ikc_cpu(int id);
@@ -78,7 +78,7 @@ void ihk_mc_init_context(ihk_mc_kernel_context_t *new_ctx,
                          void (*next_function)(void));
 
 int ihk_mc_get_extra_reg_id(unsigned long hw_config, unsigned long hw_config_ext);
-unsigned int ihk_mc_get_nr_extra_regs();
+unsigned int ihk_mc_get_nr_extra_regs(void);
 int ihk_mc_get_extra_reg_idx(int id);
 unsigned int ihk_mc_get_extra_reg_msr(int id);
 unsigned long ihk_mc_get_extra_reg_event(int id);

lib/include/ihk/mm.h

@@ -224,7 +224,7 @@ void remote_flush_tlb_array_cpumask(struct process_vm *vm,
 		int cpu_id);
 
 int ihk_get_kmsg_buf(unsigned long *addr, unsigned long *size);
-char *ihk_get_kargs();
+char *ihk_get_kargs(void);
 
 int ihk_set_monitor(unsigned long addr, unsigned long size);
 int ihk_set_rusage(unsigned long addr, unsigned long size);

test/840/C840.sh

@@ -1,152 +0,0 @@
-#!/bin/sh
-if ! sudo ls /sys/kernel/debug | grep kmemleak > /dev/null 2>&1; then
-	echo kmemleak: not found >&2
-	exit 1
-fi
-
-echo 'C840T01... '
-ng=0
-sync
-sudo /sbin/sysctl vm.drop_caches=3 > /dev/null 2>&1
-./ihkosctl 0 clear_kmsg
-sudo dmesg -c > /dev/null
-sudo sh -c 'echo clear > /sys/kernel/debug/kmemleak'
-./mcexec ./C840T01
-if [ `sudo cat /sys/kernel/debug/kmemleak | wc -l` != 0 ]; then
-	echo 'C840T01: NG (kmemleak)'
-	ng=1
-fi
-if ! dmesg | grep 'remote_page_fault:interrupted. -512' > /dev/null 2>&1; then
-	echo 'C840T01: WARN (remote_page_fault)'
-	ng=1
-fi
-if ! ./ihkosctl 0 kmsg | grep 'is dead, terminate()' > /dev/null 2>&1; then
-	echo 'C840T01: WARN (syscall offloading)'
-	ng=1
-fi
-if [ $ng = 0 ]; then
-	echo C840T01: OK
-fi
-
-echo 'C840T02... '
-ng=0
-sync
-sudo /sbin/sysctl vm.drop_caches=3 > /dev/null 2>&1
-./ihkosctl 0 clear_kmsg
-sudo dmesg -c > /dev/null
-sudo sh -c 'echo clear > /sys/kernel/debug/kmemleak'
-./mcexec ./C840T02
-if [ `sudo cat /sys/kernel/debug/kmemleak | wc -l` != 0 ]; then
-	echo 'C840T02: NG (kmemleak)'
-	ng=1
-fi
-if dmesg | grep 'remote_page_fault:interrupted. -512' > /dev/null 2>&1; then
-	echo 'C840T02: WARN (remote_page_fault)'
-	ng=1
-fi
-if ! ./ihkosctl 0 kmsg | grep 'is dead, terminate()' > /dev/null 2>&1; then
-	echo 'C840T02: WARN (syscall offloading)'
-	ng=1
-fi
-if [ $ng = 0 ]; then
-	echo C840T02: OK
-fi
-
-echo 'C840T03... '
-ng=0
-sync
-sudo /sbin/sysctl vm.drop_caches=3 > /dev/null 2>&1
-./ihkosctl 0 clear_kmsg
-sudo dmesg -c > /dev/null
-sudo sh -c 'echo clear > /sys/kernel/debug/kmemleak'
-./mcexec ./C840T03
-if [ `sudo cat /sys/kernel/debug/kmemleak | wc -l` != 0 ]; then
-	echo 'C840T03: NG (kmemleak)'
-	ng=1
-fi
-if dmesg | grep 'remote_page_fault:interrupted. -512' > /dev/null 2>&1; then
-	echo 'C840T03: WARN (remote_page_fault)'
-	ng=1
-fi
-if ./ihkosctl 0 kmsg | grep 'is dead, terminate()' > /dev/null 2>&1; then
-	echo 'C840T03: WARN (syscall offloading)'
-	ng=1
-fi
-if [ $ng = 0 ]; then
-	echo C840T03: OK
-fi
-
-echo 'C840T04... '
-ng=0
-sync
-sudo /sbin/sysctl vm.drop_caches=3 > /dev/null 2>&1
-./ihkosctl 0 clear_kmsg
-sudo dmesg -c > /dev/null
-sudo sh -c 'echo clear > /sys/kernel/debug/kmemleak'
-timeout -s 9 2 ./mcexec ./C840T04
-sleep 2
-if [ `sudo cat /sys/kernel/debug/kmemleak | wc -l` != 0 ]; then
-	echo 'C840T04: NG (kmemleak)'
-	ng=1
-fi
-if ! dmesg | grep 'remote_page_fault:interrupted. -512' > /dev/null 2>&1; then
-	echo 'C840T04: WARN (remote_page_fault)'
-	ng=1
-fi
-if ! ./ihkosctl 0 kmsg | grep 'is dead, terminate()' > /dev/null 2>&1; then
-	echo 'C840T04: WARN (syscall offloading)'
-	ng=1
-fi
-if [ $ng = 0 ]; then
-	echo C840T04: OK
-fi
-
-echo 'C840T05... '
-ng=0
-sync
-sudo /sbin/sysctl vm.drop_caches=3 > /dev/null 2>&1
-./ihkosctl 0 clear_kmsg
-sudo dmesg -c > /dev/null
-sudo sh -c 'echo clear > /sys/kernel/debug/kmemleak'
-timeout -s 9 2 ./mcexec ./C840T05
-sleep 2
-if [ `sudo cat /sys/kernel/debug/kmemleak | wc -l` != 0 ]; then
-	echo 'C840T05: NG (kmemleak)'
-	ng=1
-fi
-if dmesg | grep 'remote_page_fault:interrupted. -512' > /dev/null 2>&1; then
-	echo 'C840T05: WARN (remote_page_fault)'
-	ng=1
-fi
-if ! ./ihkosctl 0 kmsg | grep 'is dead, terminate()' > /dev/null 2>&1; then
-	echo 'C840T05: WARN (syscall offloading)'
-	ng=1
-fi
-if [ $ng = 0 ]; then
-	echo C840T05: OK
-fi
-
-echo 'C840T06... '
-ng=0
-sync
-sudo /sbin/sysctl vm.drop_caches=3 > /dev/null 2>&1
-./ihkosctl 0 clear_kmsg
-sudo dmesg -c > /dev/null
-sudo sh -c 'echo clear > /sys/kernel/debug/kmemleak'
-timeout -s 9 2 ./mcexec ./C840T06
-sleep 2
-if [ `sudo cat /sys/kernel/debug/kmemleak | wc -l` != 0 ]; then
-	echo 'C840T06: NG (kmemleak)'
-	ng=1
-fi
-if dmesg | grep 'remote_page_fault:interrupted. -512' > /dev/null 2>&1; then
-	echo 'C840T06: WARN (remote_page_fault)'
-	ng=1
-fi
-if ./ihkosctl 0 kmsg | grep 'is dead, terminate()' > /dev/null 2>&1; then
-	echo 'C840T06: WARN (syscall offloading)'
-	ng=1
-fi
-if [ $ng = 0 ]; then
-	echo C840T06: OK
-fi

test/840/C840.txt

@@ -1,28 +0,0 @@
-スクリプトは Wed Jun  6 14:38:21 2018
- に開始しました[?1034hbash-4.2$ sh C840.sh
-C840T01... 
-Terminate by signal 9
-C840.sh: 14 行: 22464 強制終了            ./mcexec ./C840T01
-C840T01: OK
-C840T02... 
-Terminate by signal 9
-C840.sh: 38 行: 22500 強制終了            ./mcexec ./C840T02
-C840T02: OK
-C840T03... 
-Terminate by signal 9
-C840.sh: 62 行: 22535 強制終了            ./mcexec ./C840T03
-C840T03: OK
-C840T04... 
-C840.sh: 86 行: 22570 強制終了            timeout -s 9 2 ./mcexec ./C840T04
-C840T04: OK
-C840T05... 
-C840.sh: 111 行: 22598 強制終了            timeout -s 9 2 ./mcexec ./C840T05
-C840T05: OK
-C840T06... 
-C840.sh: 136 行: 22626 強制終了            timeout -s 9 2 ./mcexec ./C840T06
-C840T06: OK
-bash-4.2$ exit
-exit
-
-スクリプトは Wed Jun  6 14:38:51 2018
- に終了しました

test/840/C840T01.c

@@ -1,46 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-#include <sys/mman.h>
-#include <fcntl.h>
-#include <signal.h>
-#include <errno.h>
-
-int
-main(int argc, char **argv)
-{
-	int fd;
-	void *p;
-	long l;
-	pid_t pid;
-
-	pid = fork();
-	if (pid == 0) {
-		sleep(1);
-		kill(getppid(), SIGKILL);
-		exit(0);
-	}
-
-	fd = open("rpf.data", O_RDONLY);
-	if (fd == -1) {
-		perror("open(rpf.data)");
-		exit(1);
-	}
-	p = mmap(NULL, 512*1024*1024, PROT_READ, MAP_PRIVATE, fd, 0);
-	if (p == (void *)-1) {
-		perror("mmap");
-		exit(1);
-	}
-	close(fd);
-	fd = open("rpf.out", O_WRONLY|O_CREAT|O_TRUNC, 0600);
-	if (fd == -1) {
-		perror("open(fpt.out)");
-		exit(1);
-	}
-	l = write(fd, p, 512*1024*1024);
-	printf("write=%ld\n", l);
-	close(fd);
-	munmap(p, 512*1024*1024);
-	exit(0);
-}

test/840/C840T02.c

@@ -1,27 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-#include <sys/mman.h>
-#include <fcntl.h>
-#include <signal.h>
-#include <errno.h>
-
-int
-main(int argc, char **argv)
-{
-	pid_t pid;
-	int pfd[2];
-	char c;
-
-	pid = fork();
-	if (pid == 0) {
-		sleep(1);
-		kill(getppid(), SIGKILL);
-		exit(0);
-	}
-
-	pipe(pfd);
-	read(pfd[0], &c, 1);
-	exit(0);
-}

test/840/C840T03.c

@@ -1,24 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-#include <sys/mman.h>
-#include <fcntl.h>
-#include <signal.h>
-#include <errno.h>
-
-int
-main(int argc, char **argv)
-{
-	pid_t pid;
-
-	pid = fork();
-	if (pid == 0) {
-		sleep(1);
-		kill(getppid(), SIGKILL);
-		exit(0);
-	}
-
-	for(;;);
-	exit(0);
-}

test/840/C840T04.c

@@ -1,38 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-#include <sys/mman.h>
-#include <fcntl.h>
-#include <signal.h>
-#include <errno.h>
-
-int
-main(int argc, char **argv)
-{
-	int fd;
-	void *p;
-	long l;
-
-	fd = open("rpf.data", O_RDONLY);
-	if (fd == -1) {
-		perror("open(rpf.data)");
-		exit(1);
-	}
-	p = mmap(NULL, 512*1024*1024, PROT_READ, MAP_PRIVATE, fd, 0);
-	if (p == (void *)-1) {
-		perror("mmap");
-		exit(1);
-	}
-	close(fd);
-	fd = open("rpf.out", O_WRONLY|O_CREAT|O_TRUNC, 0600);
-	if (fd == -1) {
-		perror("open(fpt.out)");
-		exit(1);
-	}
-	l = write(fd, p, 512*1024*1024);
-	printf("write=%ld\n", l);
-	close(fd);
-	munmap(p, 512*1024*1024);
-	exit(0);
-}

test/840/C840T05.c

@@ -1,19 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-#include <sys/mman.h>
-#include <fcntl.h>
-#include <signal.h>
-#include <errno.h>
-
-int
-main(int argc, char **argv)
-{
-	int pfd[2];
-	char c;
-
-	pipe(pfd);
-	read(pfd[0], &c, 1);
-	exit(0);
-}

test/840/C840T06.c

@@ -1,15 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-#include <sys/mman.h>
-#include <fcntl.h>
-#include <signal.h>
-#include <errno.h>
-
-int
-main(int argc, char **argv)
-{
-	for(;;);
-	exit(0);
-}

test/926/C926.c

@@ -1,587 +0,0 @@
-#include <stdlib.h>
-#include <stdio.h>
-#include <unistd.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/socket.h>
-#include <sys/mman.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-
-int
-main(int argc, char **argv)
-{
-	pid_t pid1;
-	pid_t pid2;
-	pid_t pid3;
-	int st;
-	int p[2];
-	int shmid;
-	int *sp;
-	key_t key;
-	int valid1;
-	int valid2;
-	int valid3;
-	char c;
-	int result;
-	struct shmid_ds buf;
-
-	key = ftok(argv[0], 0);
-
-	printf("C926T01... ");
-	fflush(stdout);
-	valid1 = 1;
-	valid2 = 2;
-	valid3 = 2;
-	if (socketpair(AF_UNIX, SOCK_STREAM, 0, p) == -1) {
-		perror("socketpair");
-		exit(1);
-	}
-
-	if ((pid1 = fork()) == 0) {
-		close(p[0]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 0
-		sp = shmat(shmid, NULL, 0);
-		*sp = 1;
-		// step 1
-		st = *sp == valid1? 1: 0;
-		shmdt(sp);
-		write(p[1], &c, 1);
-		exit(st);
-	}
-
-	if ((pid2 = fork()) == 0) {
-		close(p[1]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 2
-		read(p[0], &c, 1);
-		sp = shmat(shmid, NULL, 0);
-		(*sp)++;
-		// step 3
-		st = *sp == valid2? 1: 0;
-		shmdt(sp);
-		// step 4
-		exit(st);
-	}
-
-	close(p[0]);
-	close(p[1]);
-	result = 0;
-	waitpid(pid1, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-	waitpid(pid2, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if ((pid3 = fork()) == 0) {
-		// step 5
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		sp = shmat(shmid, NULL, 0);
-		st = *sp == valid3? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	waitpid(pid3, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if (result == 3) {
-		printf("OK\n");
-	}
-	else {
-		printf("NG\n");
-	}
-
-	printf("C926T02... ");
-	fflush(stdout);
-	valid1 = 1;
-	valid2 = 2;
-	valid3 = 2;
-	if (socketpair(AF_UNIX, SOCK_STREAM, 0, p) == -1) {
-		perror("socketpair");
-		exit(1);
-	}
-
-	if ((pid1 = fork()) == 0) {
-		close(p[0]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 0
-		sp = shmat(shmid, NULL, 0);
-		*sp = 1;
-		// step 1
-		st = *sp == valid1? 1: 0;
-		exit(st);
-	}
-
-	if ((pid2 = fork()) == 0) {
-		close(p[1]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 2
-		read(p[0], &c, 1);
-		sp = shmat(shmid, NULL, 0);
-		(*sp)++;
-		// step 3
-		st = *sp == valid2? 1: 0;
-		shmdt(sp);
-		// step 4
-		exit(st);
-	}
-
-	close(p[0]);
-	close(p[1]);
-	result = 0;
-	waitpid(pid1, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-	waitpid(pid2, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if ((pid3 = fork()) == 0) {
-		// step 5
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		sp = shmat(shmid, NULL, 0);
-		st = *sp == valid3? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	waitpid(pid3, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if (result == 3) {
-		printf("OK\n");
-	}
-	else {
-		printf("NG\n");
-	}
-
-	printf("C926T03... ");
-	fflush(stdout);
-	valid1 = 1;
-	valid2 = 1;
-	valid3 = 1;
-	if (socketpair(AF_UNIX, SOCK_STREAM, 0, p) == -1) {
-		perror("socketpair");
-		exit(1);
-	}
-
-	if ((pid1 = fork()) == 0) {
-		close(p[0]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 0
-		sp = shmat(shmid, NULL, 0);
-		*sp = 1;
-		// step 1
-		shmctl(shmid, IPC_RMID, &buf);
-		write(p[1], &c, 1);
-
-		// step3
-		read(p[1], &c, 1);
-		st = *sp == valid1? 1: 0;
-		shmdt(sp);
-		write(p[1], &c, 1);
-		exit(st);
-	}
-
-	if ((pid2 = fork()) == 0) {
-		close(p[1]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 2
-		read(p[0], &c, 1);
-		sp = shmat(shmid, NULL, 0);
-		(*sp)++;
-		write(p[0], &c, 1);
-		// step 4
-		read(p[0], &c, 1);
-		st = *sp == valid2? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	close(p[0]);
-	close(p[1]);
-	result = 0;
-	waitpid(pid1, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-	waitpid(pid2, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if ((pid3 = fork()) == 0) {
-		// step 5
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		sp = shmat(shmid, NULL, 0);
-		st = *sp == valid3? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	waitpid(pid3, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if (result == 3) {
-		printf("OK\n");
-	}
-	else {
-		printf("NG\n");
-	}
-
-	printf("C926T04... ");
-	fflush(stdout);
-	valid1 = 1;
-	valid2 = 1;
-	valid3 = 1;
-	if (socketpair(AF_UNIX, SOCK_STREAM, 0, p) == -1) {
-		perror("socketpair");
-		exit(1);
-	}
-
-	if ((pid1 = fork()) == 0) {
-		close(p[0]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 0
-		sp = shmat(shmid, NULL, 0);
-		*sp = 1;
-		// step 1
-		shmctl(shmid, IPC_RMID, &buf);
-		write(p[1], &c, 1);
-
-		// step4
-		read(p[1], &c, 1);
-		st = *sp == valid1? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	if ((pid2 = fork()) == 0) {
-		close(p[1]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 2
-		read(p[0], &c, 1);
-		sp = shmat(shmid, NULL, 0);
-		(*sp)++;
-		// step 3
-		st = *sp == valid2? 1: 0;
-		shmdt(sp);
-		write(p[0], &c, 1);
-		exit(st);
-	}
-
-	close(p[0]);
-	close(p[1]);
-	result = 0;
-	waitpid(pid1, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-	waitpid(pid2, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if ((pid3 = fork()) == 0) {
-		// step 5
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		sp = shmat(shmid, NULL, 0);
-		st = *sp == valid3? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	waitpid(pid3, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if (result == 3) {
-		printf("OK\n");
-	}
-	else {
-		printf("NG\n");
-	}
-
-	printf("C926T05... ");
-	fflush(stdout);
-	valid1 = 1;
-	valid2 = 1;
-	valid3 = 1;
-	if (socketpair(AF_UNIX, SOCK_STREAM, 0, p) == -1) {
-		perror("socketpair");
-		exit(1);
-	}
-
-	if ((pid1 = fork()) == 0) {
-		close(p[0]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 0
-		sp = shmat(shmid, NULL, 0);
-		*sp = 1;
-		// step 1
-		shmctl(shmid, IPC_RMID, &buf);
-		// step2
-		st = *sp == valid1? 1: 0;
-		shmdt(sp);
-		write(p[1], &c, 1);
-		exit(st);
-	}
-
-	if ((pid2 = fork()) == 0) {
-		close(p[1]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 3
-		read(p[0], &c, 1);
-		sp = shmat(shmid, NULL, 0);
-		(*sp)++;
-		// step 4
-		st = *sp == valid2? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	close(p[0]);
-	close(p[1]);
-	result = 0;
-	waitpid(pid1, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-	waitpid(pid2, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if ((pid3 = fork()) == 0) {
-		// step 5
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		sp = shmat(shmid, NULL, 0);
-		st = *sp == valid3? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	waitpid(pid3, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if (result == 3) {
-		printf("OK\n");
-	}
-	else {
-		printf("NG\n");
-	}
-
-	printf("C926T06... ");
-	fflush(stdout);
-	valid1 = 1;
-	valid2 = 1;
-	valid3 = 1;
-	if (socketpair(AF_UNIX, SOCK_STREAM, 0, p) == -1) {
-		perror("socketpair");
-		exit(1);
-	}
-
-	if ((pid1 = fork()) == 0) {
-		close(p[0]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 0
-		sp = shmat(shmid, NULL, 0);
-		*sp = 1;
-		// step 1
-		shmctl(shmid, IPC_RMID, &buf);
-		// step2
-		st = *sp == valid1? 1: 0;
-		exit(st);
-	}
-
-	if ((pid2 = fork()) == 0) {
-		close(p[1]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 3
-		read(p[0], &c, 1);
-		sp = shmat(shmid, NULL, 0);
-		(*sp)++;
-		// step 4
-		st = *sp == valid2? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	close(p[0]);
-	close(p[1]);
-	result = 0;
-	waitpid(pid1, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-	waitpid(pid2, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if ((pid3 = fork()) == 0) {
-		// step 5
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		sp = shmat(shmid, NULL, 0);
-		st = *sp == valid3? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	waitpid(pid3, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if (result == 3) {
-		printf("OK\n");
-	}
-	else {
-		printf("NG\n");
-	}
-
-	printf("C926T07... ");
-	fflush(stdout);
-	valid1 = 2;
-	valid2 = 2;
-	valid3 = 0;
-	if (socketpair(AF_UNIX, SOCK_STREAM, 0, p) == -1) {
-		perror("socketpair");
-		exit(1);
-	}
-
-	if ((pid1 = fork()) == 0) {
-		close(p[0]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 0
-		sp = shmat(shmid, NULL, 0);
-		*sp = 1;
-		write(p[1], &c, 1);
-		// step 2
-		read(p[1], &c, 1);
-		shmctl(shmid, IPC_RMID, &buf);
-		// step2
-		st = *sp == valid1? 1: 0;
-		exit(st);
-	}
-
-	if ((pid2 = fork()) == 0) {
-		close(p[1]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 1
-		read(p[0], &c, 1);
-		sp = shmat(shmid, NULL, 0);
-		(*sp)++;
-		write(p[0], &c, 1);
-		// step 3
-		read(p[0], &c, 1);
-		st = *sp == valid2? 1: 0;
-		shmdt(sp);
-		// step 4
-		exit(st);
-	}
-
-	close(p[0]);
-	close(p[1]);
-	result = 0;
-	waitpid(pid1, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-	waitpid(pid2, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if ((pid3 = fork()) == 0) {
-		// step 5
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		sp = shmat(shmid, NULL, 0);
-		st = *sp == valid3? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	waitpid(pid3, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if (result == 3) {
-		printf("OK\n");
-	}
-	else {
-		printf("NG\n");
-	}
-
-	printf("C926T08... ");
-	fflush(stdout);
-	valid1 = 2;
-	valid2 = 2;
-	valid3 = 2;
-	if (socketpair(AF_UNIX, SOCK_STREAM, 0, p) == -1) {
-		perror("socketpair");
-		exit(1);
-	}
-
-	if ((pid1 = fork()) == 0) {
-		close(p[0]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 0
-		sp = shmat(shmid, NULL, 0);
-		*sp = 1;
-		write(p[1], &c, 1);
-		// step 2
-		read(p[1], &c, 1);
-		// step2
-		st = *sp == valid1? 1: 0;
-		exit(st);
-	}
-
-	if ((pid2 = fork()) == 0) {
-		close(p[1]);
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		// step 1
-		read(p[0], &c, 1);
-		sp = shmat(shmid, NULL, 0);
-		(*sp)++;
-		write(p[0], &c, 1);
-		// step 3
-		read(p[0], &c, 1);
-		st = *sp == valid2? 1: 0;
-		shmdt(sp);
-		// step 4
-		exit(st);
-	}
-
-	close(p[0]);
-	close(p[1]);
-	result = 0;
-	waitpid(pid1, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-	waitpid(pid2, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if ((pid3 = fork()) == 0) {
-		// step 5
-		shmid = shmget(key, 4096, IPC_CREAT | 0660);
-		sp = shmat(shmid, NULL, 0);
-		st = *sp == valid3? 1: 0;
-		shmdt(sp);
-		exit(st);
-	}
-
-	waitpid(pid3, &st, 0);
-	if (WIFEXITED(st))
-		result += WEXITSTATUS(st);
-
-	if (result == 3) {
-		printf("OK\n");
-	}
-	else {
-		printf("NG\n");
-	}
-
-	exit(0);
-}

test/926/C926.txt

@@ -1,15 +0,0 @@
-スクリプトは Wed Jun  6 14:39:24 2018
- に開始しました[?1034hbash-4.2$ ./mcexec C./C926
-C926T01... OK
-C926T02... OK
-C926T03... OK
-C926T04... OK
-C926T05... OK
-C926T06... OK
-C926T07... OK
-C926T08... OK
-bash-4.2$ exit
-exit
-
-スクリプトは Wed Jun  6 14:39:38 2018
- に終了しました

test/issues/1121/C1121.sh

@@ -0,0 +1,108 @@
+#!/bin/sh
+BIN=
+SBIN=
+OSTEST=
+LTP=
+BOOTPARAM="-c 1-7,9-15,17-23,25-31 -m 10G@0,10G@1 -r 1-7:0+9-15:8+17-23:16+25-31:24"
+
+if [ -f ../../../config.h ]; then
+	str=`grep "^#define BINDIR " ../../../config.h | head -1 | sed 's/^#define BINDIR /BINDIR=/'`
+	eval $str
+fi
+if [ "x$BINDIR" = x ];then
+	BINDIR="$BIN"
+fi
+
+if [ -f ../../../Makefile ]; then
+	str=`grep ^SBINDIR ../../../Makefile | head -1 | sed 's/ //g'`
+	eval $str
+fi
+if [ "x$SBINDIR" = x ];then
+	SBINDIR="$SBIN"
+fi
+
+if [ -f $HOME/ltp/testcases/bin/sched_setaffinity01 ]; then
+	LTPDIR=$HOME/ltp/testcases
+fi
+if [ "x$LTPDIR" = x ]; then
+	LTPDIR="$LTP"
+fi
+
+if [ -f $HOME/ostest/bin/test_mck ]; then
+	OSTESTDIR=$HOME/ostest/
+fi
+if [ "x$OSTESTDIR" = x ]; then
+	OSTESTDIR="$OSTEST"
+fi
+
+if [ ! -x $SBINDIR/mcstop+release.sh ]; then
+	echo mcstop+releas: not found >&2
+	exit 1
+fi
+echo -n "mcstop+release.sh ... "
+sudo $SBINDIR/mcstop+release.sh
+echo "done"
+
+if [ ! -x $SBINDIR/mcreboot.sh ]; then
+	echo mcreboot: not found >&2
+	exit 1
+fi
+echo -n "mcreboot.sh $BOOTPARAM ... "
+sudo $SBINDIR/mcreboot.sh $BOOTPARAM
+echo "done"
+
+if [ ! -x $BINDIR/mcexec ]; then
+	echo mcexec: not found >&2
+	exit 1
+fi
+
+tid=001
+echo "*** RT_$tid start *******************************"
+sudo $BINDIR/mcexec $OSTESTDIR/bin/test_mck -s sched_setaffinity -n 0 -- -p 20 2>&1 | tee ./RT_${tid}.txt
+if grep "RESULT: ok" ./RT_${tid}.txt > /dev/null 2>&1 ; then
+	echo "*** RT_$tid: PASSED"
+else
+	echo "*** RT_$tid: FAILED"
+fi
+echo ""
+
+tid=002
+echo "*** RT_$tid start *******************************"
+sudo $BINDIR/mcexec $OSTESTDIR/bin/test_mck -s sched_setaffinity -n 1 -- -p 20 2>&1 | tee ./RT_${tid}.txt
+if grep "RESULT: ok" ./RT_${tid}.txt > /dev/null 2>&1 ; then
+	echo "*** RT_$tid: PASSED"
+else
+	echo "*** RT_$tid: FAILED"
+fi
+echo ""
+
+tid=003
+echo "*** RT_$tid start *******************************"
+sudo $BINDIR/mcexec $OSTESTDIR/bin/test_mck -s sched_getaffinity -n 3 -- -p 20 2>&1 | tee ./RT_${tid}.txt
+if grep "RESULT: ok" ./RT_${tid}.txt > /dev/null 2>&1 ; then
+	echo "*** RT_$tid: PASSED"
+else
+	echo "*** RT_$tid: FAILED"
+fi
+echo ""
+
+tid=004
+echo "*** RT_$tid start *******************************"
+sudo $BINDIR/mcexec $OSTESTDIR/bin/test_mck -s sched_getaffinity -n 5 -- -p 20 2>&1 | tee ./RT_${tid}.txt
+if grep "RESULT: ok" ./RT_${tid}.txt > /dev/null 2>&1 ; then
+	echo "*** RT_$tid: PASSED"
+else
+	echo "*** RT_$tid: FAILED"
+fi
+echo ""
+
+tid=001
+echo "*** LT_$tid start *******************************"
+sudo $BINDIR/mcexec $LTPDIR/bin/sched_setaffinity01 2>&1 | tee ./LT_${tid}.txt
+ok=`grep TPASS LT_${tid}.txt | wc -l`
+ng=`grep TFAIL LT_${tid}.txt | wc -l`
+if [ $ng = 0 ]; then
+	echo "*** LT_$tid: PASSED (ok:$ok)"
+else
+	echo "*** LT_$tid: FAILED (ok:$ok, ng:$ng)"
+fi

test/issues/1121/Makefile

@@ -0,0 +1,14 @@
+CC = gcc
+TARGET=
+
+CPPFLAGS =
+LDFLAGS =
+
+all: $(TARGET)
+
+test: all
+	@sh ./C1121.sh
+
+clean:
+	rm -f $(TARGET) *.o
+

test/issues/1121/README

@@ -0,0 +1,36 @@
+【Issue#1121 動作確認】
+□ テスト内容
+1. Issueで報告された再現プログラムでの確認
+RT_001: ostest-sched_setaffinity.000
+	自プロセスのaffinityの設定と取得が行えることを確認
+
+RT_002: ostest-sched_setaffinity.001
+	自プロセスのaffinityを設定し、CPUのmigrateができることを確認
+
+RT_003: ostest-sched_getaffinity.003
+	存在しないプロセスIDを指定したsched_getaffinity()を実行し、
+	-1 が返り、errnoにESRCH が設定されることを確認
+
+RT_004: ostest-sched_getaffinity.005
+	子プロセスを生成し、以下を確認
+	 - 子プロセスが親プロセスのaffinityを取得
+	 - 子プロセスが自プロセスのaffinityを取得
+	 - 親プロセス、自プロセスから取得したaffinityが等しいことを確認
+
+2. 既存のsched_setaffinity機能に影響がないことをLTPを用いて確認
+LT_001: ltp-sched_setaffinity01
+	異常な引数を指定した際に返される値と
+	設定されるerrnoが正しいことを確認 (OK 4件)
+
+□ 実行手順
+$ make test
+
+実行できない場合は、C1121.shの以下の行を適切に書き換えた後に実行。
+BIN=          mcexec が存在するパス
+SBIN=         mcreboot.sh が存在するパス
+OSTEST=       OSTESTが存在するパス
+LTP=          LTPが存在するパス
+
+□ 実行結果
+result.log 参照。
+すべての項目をPASSしていることを確認。

test/issues/1121/result.log

@@ -0,0 +1,35 @@
+*** RT_001 start *******************************
+TEST_SUITE: sched_setaffinity
+TEST_NUMBER: 0
+ARGS: -p 20
+RESULT: ok
+*** RT_001: PASSED
+
+*** RT_002 start *******************************
+TEST_SUITE: sched_setaffinity
+TEST_NUMBER: 1
+ARGS: -p 20
+RESULT: ok
+*** RT_002: PASSED
+
+*** RT_003 start *******************************
+TEST_SUITE: sched_getaffinity
+TEST_NUMBER: 3
+ARGS: -p 20
+sched_getaffinity result:-1, errno:3 (expect error is "ESRCH"=3)
+RESULT: ok
+*** RT_003: PASSED
+
+*** RT_004 start *******************************
+TEST_SUITE: sched_getaffinity
+TEST_NUMBER: 5
+ARGS: -p 20
+RESULT: ok
+*** RT_004: PASSED
+
+*** LT_001 start *******************************
+sched_setaffinity01    1  TPASS  :  expected failure with 'Bad address'
+sched_setaffinity01    2  TPASS  :  expected failure with 'Invalid argument'
+sched_setaffinity01    3  TPASS  :  expected failure with 'No such process'
+sched_setaffinity01    4  TPASS  :  expected failure with 'Operation not permitted'
+*** LT_001: PASSED (ok:4)

test/issues/731/C731.sh

@@ -0,0 +1,66 @@
+#!/bin/sh
+BIN=
+SBIN=
+LTP=
+BOOTPARAM="-c 1-7,17-23,9-15,25-31 -m 10G@0,10G@1"
+
+if [ -f ../../../config.h ]; then
+	str=`grep "^#define BINDIR " ../../../config.h | head -1 | sed 's/^#define BINDIR /BINDIR=/'`
+	eval $str
+fi
+if [ "x$BINDIR" = x ];then
+	BINDIR="$BIN"
+fi
+
+if [ -f ../../../Makefile ]; then
+	str=`grep ^SBINDIR ../../../Makefile | head -1 | sed 's/ //g'`
+	eval $str
+fi
+if [ "x$SBINDIR" = x ];then
+	SBINDIR="$SBIN"
+fi
+
+if [ -f $HOME/ltp/testcases/bin/fork01 ]; then
+	LTPDIR=$HOME/ltp/testcases
+fi
+if [ "x$LTPDIR" = x ]; then
+	LTPDIR="$LTP"
+fi
+
+if ! lsmod | grep mcctrl > /dev/null 2>&1; then
+	if [ ! -x $SBINDIR/mcreboot.sh ]; then
+		echo no mcreboot found >&2
+		exit 1
+	fi
+	sudo $SBINDIR/mcreboot.sh $BOOTPARAM
+fi
+
+if [ ! -x $BINDIR/mcexec ]; then
+	echo no mcexec found >&2
+	exit 1
+fi
+
+sudo $BINDIR/mcexec ./g310a 2>&1 | tee g310a.txt
+if grep "fork: Permission denied" g310a.txt > /dev/null 2>&1 ; then
+	echo "*** C731T001: g310a OK"
+else
+	echo "*** C731T001: g310a NG"
+fi
+
+if [ x$LTPDIR = x ]; then
+	echo no LTP found >&2
+	exit 1
+fi
+
+for i in 01:002 02:003 03:004 04:005 07:006 08:007; do
+	tp=`echo $i|sed 's/:.*//'`
+	id=`echo $i|sed 's/.*://'`
+	$BINDIR/mcexec $LTPDIR/bin/fork$tp 2>&1 | tee fork$tp.txt
+	ok=`grep TPASS fork$tp.txt | wc -l`
+	ng=`grep TFAIL fork$tp.txt | wc -l`
+	if [ $ng = 0 ]; then
+		echo "*** C731T$id: fork$tp OK ($ok)"
+	else
+		echo "*** C731T$id: fork$tp NG (ok=$ok ng=%ng)"
+	fi
+done

test/issues/731/C731.txt

@@ -0,0 +1,46 @@
+Script started on Thu Jun 28 09:16:36 2018
+bash-4.2$ make test
+gcc -o g310a g310a.c -Wall -g
+sh ./C731.sh
+Error: Failed to open /dev/mcos0.
+ERROR: opening /dev/mcos0
+fork(): error with child process after fork
+fork: Permission denied
+waitpid: No child processes
+*** C731T001: g310a OK
+fork01      1  TPASS  :  fork() returned 14513
+fork01      2  TPASS  :  child pid and fork() return agree: 14513
+*** C731T002: fork01 OK (2)
+fork02      0  TINFO  :  Inside parent
+fork02      0  TINFO  :  Inside child
+fork02      0  TINFO  :  exit status of wait 0
+fork02      1  TPASS  :  test 1 PASSED
+*** C731T003: fork02 OK (1)
+fork03      0  TINFO  :  process id in parent of child from fork : 14669
+fork03      1  TPASS  :  test 1 PASSED
+*** C731T004: fork03 OK (1)
+fork04      1  TPASS  :  Env var TERM unchanged after fork(): xterm
+fork04      2  TPASS  :  Env var NoTSetzWq unchanged after fork(): getenv() does not find variable set
+fork04      3  TPASS  :  Env var TESTPROG unchanged after fork(): FRKTCS04
+*** C731T005: fork04 OK (3)
+fork07      0  TINFO  :  Forking 100 children
+fork07      0  TINFO  :  Forked all 100 children, now collecting
+fork07      0  TINFO  :  Collected all 100 children
+fork07      1  TPASS  :  100/100 children read correctly from an inheritted fd
+*** C731T006: fork07 OK (1)
+fork08      0  TINFO  :  parent forksval: 1
+fork08      0  TINFO  :  parent forksval: 2
+fork08      0  TINFO  :  second child got char: b
+fork08      1  TPASS  :  Test passed in childnumber 2
+fork08      0  TINFO  :  exit status of wait  expected 0 got 0
+fork08      1  TPASS  :  parent test PASSED
+fork08      0  TINFO  :  exit status of wait  expected 0 got 0
+fork08      2  TPASS  :  parent test PASSED
+fork08      0  TINFO  :  exit status of wait  expected 0 got 0
+fork08      3  TPASS  :  parent test PASSED
+fork08      0  TINFO  :  Number of processes forked is 2
+*** C731T007: fork08 OK (4)
+bash-4.2$ exit
+exit
+
+Script done on Thu Jun 28 09:17:21 2018

test/issues/731/Makefile

@@ -0,0 +1,13 @@
+CC=gcc
+TARGET=g310a
+
+all:: $(TARGET)
+
+g310a: g310a.c
+	$(CC) -o g310a g310a.c -Wall -g
+
+test:: $(TARGET)
+	sh ./C731.sh
+
+clean::
+	rm -f *.o $(TARGET)

test/issues/731/README

@@ -0,0 +1,35 @@
+【Issue#731 動作確認】
+□ テスト内容
+1. g310aの実行確認
+C731T001 g310aを実行し、forkが Permission denied でエラーとなることを確認
+
+2. 既存fork機能に影響しないことをLTPを用いて確認
+C731T002 fork01 の実行確認
+  fork後に子プロセスのプロセスIDが正しいことを確認 (OK 2件)
+C731T003 fork02 の実行確認
+  fork後にwaitを行い、waitが子プロセスのPIDを返却することを確認(OK 1件)
+C731T004 fork03 の実行確認
+  子プロセスがfork後に計算を行えること、また子プロセスでのfork返却値が0で
+  あることを確認 (OK 1件)
+C731T005 fork04 の実行確認
+  forkで生成した子プロセスが環境変数を変更しても、親プロセス側の
+  環境変数に変化が無いことを確認 (OK 3件)
+C731T006 fork07 の実行確認
+  forkした子プロセスに、親プロセスからファイルディスクリプタを
+  引き継いでいることを確認 (OK 1件)
+C731T007 fork08 の実行確認
+  forkした複数の子プロセスが、それぞれ親プロセスから引き継いだファイル
+  ディスクリプタを別個に操作できることを確認
+  (ある子プロセスがcloseしても別な子プロセスがI/O可能) (OK 4件)
+
+□ 実行手順
+$ make test
+
+実行できない場合は、C731.shの以下の行を適切に書き換えた後に実行。
+BIN=          mcexec が存在するパス
+SBIN=         mcreboot.sh が存在するパス
+LTP=          LTPが存在するパス
+
+□ 実行結果
+C731.txt 参照。
+全ての項目が OK となっていることを確認。

test/issues/731/fork01.txt

@@ -0,0 +1,2 @@
+fork01      1  TPASS  :  fork() returned 14513
+fork01      2  TPASS  :  child pid and fork() return agree: 14513

test/issues/731/fork02.txt

@@ -0,0 +1,4 @@
+fork02      0  TINFO  :  Inside parent
+fork02      0  TINFO  :  Inside child
+fork02      0  TINFO  :  exit status of wait 0
+fork02      1  TPASS  :  test 1 PASSED

test/issues/731/fork03.txt

@@ -0,0 +1,2 @@
+fork03      0  TINFO  :  process id in parent of child from fork : 14669
+fork03      1  TPASS  :  test 1 PASSED

test/issues/731/fork04.txt

@@ -0,0 +1,3 @@
+fork04      1  TPASS  :  Env var TERM unchanged after fork(): xterm
+fork04      2  TPASS  :  Env var NoTSetzWq unchanged after fork(): getenv() does not find variable set
+fork04      3  TPASS  :  Env var TESTPROG unchanged after fork(): FRKTCS04

test/issues/731/fork07.txt

@@ -0,0 +1,4 @@
+fork07      0  TINFO  :  Forking 100 children
+fork07      0  TINFO  :  Forked all 100 children, now collecting
+fork07      0  TINFO  :  Collected all 100 children
+fork07      1  TPASS  :  100/100 children read correctly from an inheritted fd

test/issues/731/fork08.txt

@@ -0,0 +1,11 @@
+fork08      0  TINFO  :  parent forksval: 1
+fork08      0  TINFO  :  parent forksval: 2
+fork08      0  TINFO  :  second child got char: b
+fork08      1  TPASS  :  Test passed in childnumber 2
+fork08      0  TINFO  :  exit status of wait  expected 0 got 0
+fork08      1  TPASS  :  parent test PASSED
+fork08      0  TINFO  :  exit status of wait  expected 0 got 0
+fork08      2  TPASS  :  parent test PASSED
+fork08      0  TINFO  :  exit status of wait  expected 0 got 0
+fork08      3  TPASS  :  parent test PASSED
+fork08      0  TINFO  :  Number of processes forked is 2

test/issues/731/g310a.c

@@ -0,0 +1,55 @@
+/*
+ * g310a: If superuser try to fork() after seteuid(bin), ...
+ */
+
+#include <pwd.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <sys/wait.h>
+
+int main(void)
+{
+	int error;
+	struct passwd *pwd;
+	pid_t pid;
+	int ws;
+
+	if (geteuid()) {
+		printf("not a superuser\n");
+		return 2;
+	}
+
+	pwd = getpwnam("bin");
+	if (!pwd) {
+		perror("getpwnam");
+		return 1;
+	}
+
+	error = seteuid(pwd->pw_uid);
+	if (error) {
+		perror("seteuid");
+		return 1;
+	}
+
+	pid = fork();
+	if (pid == -1) {
+		perror("fork");
+	}
+
+	if (!pid) {
+		return 0;
+	}
+
+	pid = waitpid(pid, &ws, 0);
+	if (pid == -1) {
+		perror("waitpid");
+		return 1;
+	}
+	if (ws) {
+		printf("ws: %#x\n", ws);
+		return 1;
+	}
+
+	printf("done.\n");
+	return 0;
+}

test/issues/731/g310a.txt

@@ -0,0 +1,5 @@
+Error: Failed to open /dev/mcos0.
+ERROR: opening /dev/mcos0
+fork(): error with child process after fork
+fork: Permission denied
+waitpid: No child processes

test/rusage/Makefile

@@ -1,18 +1,18 @@
 .SUFFIXES:	# Clear suffixes
-.SUFFIXES:	.c 
+.SUFFIXES:	.c
 
 CC = gcc
 
 CPPFLAGS = -I$(HOME)/project/os/install/include
 CCFLAGS = -g
-LDFLAGS = -L$(HOME)/project/os/install/lib -lihk -Wl,-rpath -Wl,$(HOME)/project/os/install/lib
-EXES = 
-SRCS = 
+LDFLAGS = -L$(HOME)/project/os/install/lib -lihk -Wl,-rpath -Wl,$(HOME)/project/os/install/lib -lpthread
+EXES =
+SRCS =
 OBJS = $(SRCS:.c=.o)
 
 CPPFLAGSMCK = -I$(HOME)/usr/include
 CCFLAGSMCK = -g -O0
-LDFLAGSMCK = -static 
+LDFLAGSMCK = -static -lpthread
 SRCSMCK = $(shell ls rusage*.c)
 EXESMCK = $(SRCSMCK:.c=)
 OBJSMCK = $(SRCSMCK:.c=.o)
@@ -163,5 +163,17 @@ rusage103: rusage103.o
 rusage103.o: rusage103.c
 	$(CC) $(CCFLAGS) $(CPPFLAGS) -c $<
 
+%_mck: %_mck.o
+	$(CC) -o $@ $^ $(LDFLAGSMCK)
+
+%_mck.o:: %_mck.c
+	$(CC) $(CCFLAGSMCK) $(CPPFLAGSMCK) -c $<
+
+%_lin.o:: %_lin.c
+	$(CC) $(CCFLAGS) $(CPPFLAGS) -c $<
+
+%_lin: %_lin.o
+	$(CC) -o $@ $^ $(LDFLAGS)
+
 clean:
 	rm -f core $(EXES) $(OBJS) $(EXESMCK) $(OBJSMCK)

test/rusage/README

@@ -102,15 +102,16 @@ rusage021: Run npb bt-mz.S.4
 2-ppn x 2-tpn x 2-node (wallaby{14,15}) [OK]
 2-ppn x 1-tpn x 2-node (polaris,kochab) [OK]
 
-rusage100: Test ihk_os_getrusage() 
+rusage100: Test ihk_os_getrusage()
 anon mmap,num_threads=1 [OK]
 
 rusage101: Test ihk_os_getrusage()
 anon mmap,num_threads=2 [OK]
 
-rusage102: Test ihk_os_getrusage()  
+rusage102: Test ihk_os_getrusage()
 file map,num_threads=1 [OK]
 
 rusage103: Test ihk_os_getrusage()
 anon mmap@numa#1 [OK]
 
+rusage104: Test ihk_os_getrusage(), user time per CPU

test/rusage/run.sh

@@ -17,13 +17,18 @@ case ${testname} in
 	printf "*** Enable debug messages in rusage.h, memory.c, fileobj.c, shmobj.c, process.c by defining DEBUG macro, e.g. #define RUSAGE_DEBUG and then recompile IHK/McKernel.\n"
 	printf "*** Install xpmem by git-clone https://github.com/hjelmn/xpmem.\n"
 	;;
-    rusage10?)
+    rusage100 | rusage101 | rusage102 | rusage103)
 	printf "*** Refer to rusage100.patch to add syscall #900 by editing syscall_list.h and syscall.c and recompile IHK/McKernel.\n"
 	;;
+    rusage104)
+	printf "*** Apply rusage104.patch to enable syscall #900"
+	printf "which reports rusage values.\n"
+	;;
     *)
 	printf "*** Enable debug messages in rusage.h, memory.c, fileobj.c, shmobj.c, process.c by defining DEBUG macro, e.g. #define RUSAGE_DEBUG and then recompile IHK/McKernel.\n"
 	;;
 esac
+
 read -p "*** Hit return when ready!" key
 
 case ${testname} in
@@ -45,6 +50,11 @@ case ${testname} in
 	bn=npb/NPB3.3.1-MZ/NPB3.3-MZ-MPI/bin/bt-mz.S.4
 	perl -e 'print "polaris:2\nkochab:2\n"' > ./hostfile
 	;;
+    rusage104)
+	bn=${testname}
+	make clean > /dev/null 2> /dev/null
+	make ${bn}_mck ${bn}_lin
+	;;
     *)
 	bn=${testname}
 	make clean > /dev/null 2> /dev/null
@@ -141,8 +151,11 @@ case ${testname} in
     rusage103)
 	bootopt="-m 256M@1"
 	;;
+    rusage104)
+	bootopt="-c 1,2,3 -m 256M"
+	;;
     *)
-	echo Unknown test case 
+	echo Unknown test case
 	exit 255
 esac
 
@@ -199,6 +212,12 @@ else
 	    echo "================================================" >> ./${testname}.log
 	    sudo ${install}/sbin/ihkosctl 0 kmsg >> ./${testname}.log
 	    ;;
+	rusage104)
+	    ${install}/bin/mcexec ${mcexecopt} ./${bn}_mck
+	    ${install}/bin/mcexec ${mcexecopt} ./${bn}_lin
+	    sudo ${install}/sbin/ihkosctl 0 kmsg > ./${testname}.log
+	    grep user ./${testname}.log
+	    ;;
 	*)
 	    ${install}/bin/mcexec ${mcexecopt} ./${bn} ${testopt}
 	    sudo ${install}/sbin/ihkosctl 0 kmsg > ./${testname}.log
@@ -206,10 +225,13 @@ else
 fi
 
 case ${testname} in
-    rusage10?)
+    rusage100 | rusage101 | rusage102 | rusage103)
 	printf "*** Check the ihk_os_getrusage() result (the first part of ${testname}.log) matches with the syscall #900 result (the second part) \n"
 	;;
-
+    rusage104)
+	printf "*** It behaves as expected when there's no [NG] and "
+	printf "\"All tests finished\" is shown\n"
+	;;
     *)
 	printf "*** cat ${testname}.log (kmsg) > ./match.pl to confirm there's no stray add/sub.\n"
 	printf "*** Look ${testname}.log (kmsg) to confirm memory_stat_*[*] returned to zero when the last thread exits.\n"

test/rusage/rusage104.patch

@@ -0,0 +1,47 @@
+diff --git arch/x86_64/kernel/include/syscall_list.h arch/x86_64/kernel/include/syscall_list.h
+index 48b1ea0..3717bb2 100644
+--- arch/x86_64/kernel/include/syscall_list.h
++++ arch/x86_64/kernel/include/syscall_list.h
+@@ -168,4 +168,6 @@ SYSCALL_HANDLED(802, linux_mlock)
+ SYSCALL_HANDLED(803, suspend_threads)
+ SYSCALL_HANDLED(804, resume_threads)
+ SYSCALL_HANDLED(811, linux_spawn)
++
++SYSCALL_HANDLED(900, dump_rusage)
+ /**** End of File ****/
+diff --git kernel/syscall.c kernel/syscall.c
+index 8dc0a0e..9969a4b 100644
+--- kernel/syscall.c
++++ kernel/syscall.c
+@@ -9477,3 +9477,31 @@ long syscall(int num, ihk_mc_user_context_t *ctx)
+ 
+ 	return l;
+ }
++
++SYSCALL_DECLARE(dump_rusage)
++{
++	/* rusage debug */
++	int i;
++	for(i = 0; i < IHK_MAX_NUM_PGSIZES; i++) {
++		kprintf("memory_stat_rss[%d]=%ld\n", i, rusage->memory_stat_rss[i]);
++	}
++	for(i = 0; i < IHK_MAX_NUM_PGSIZES; i++) {
++		kprintf("memory_stat_mapped_file[%d]=%ld\n", i, rusage->memory_stat_mapped_file[i]);
++	}
++	kprintf("memory_max_usage=%ld\n", rusage->memory_max_usage);
++	kprintf("memory_kmem_usage=%ld\n", rusage->memory_kmem_usage);
++	kprintf("memory_kmem_max_usage=%ld\n", rusage->memory_kmem_max_usage);
++	for (i = 0; i < rusage->num_numa_nodes; i++) {
++		kprintf("memory_numa_stat[%d]=%ld\n", i, rusage->memory_numa_stat[i]);
++	}
++	kprintf("ns_per_tsc=%ld\n", rusage->ns_per_tsc);
++	for (i = 0; i < rusage->num_processors; i++) {
++		kprintf("cpu[%d].user_tsc=%ld\n", i, rusage->cpu[i].user_tsc);
++		kprintf("cpu[%d].system_tsc=%ld\n", i, rusage->cpu[i].system_tsc);
++	}
++
++	kprintf("num_threads=%d\n", rusage->num_threads);
++	kprintf("max_num_threads=%d\n", rusage->max_num_threads);
++
++	return 0;
++}

test/rusage/rusage104_lin.c

@@ -0,0 +1,38 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <errno.h>
+#include "ihklib.h"
+#include "mckernel/ihklib_rusage.h"
+#include "util.h"
+
+#define DELAY0 (100UL * 1000 * 1000)
+#define DELAY1 (200UL * 1000 * 1000)
+#define SCALE 1.5
+#define WITHIN_RANGE(x, y, s) (x >= y && x <= y * s)
+
+int main(int argc, char **argv)
+{
+	int ret = 0;
+	struct mckernel_rusage rusage;
+
+	if ((ret = ihk_os_getrusage(0, &rusage, sizeof(rusage)))) {
+		fprintf(stderr, "%s: ihk_os_getrusage failed\n", __func__);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	OKNG(WITHIN_RANGE(rusage.cpuacct_usage_percpu[1], DELAY0, SCALE),
+	     "cpu 0: user time: expected: %ld nsec, reported: %ld nsec\n",
+	     DELAY0, rusage.cpuacct_usage_percpu[1]);
+	OKNG(WITHIN_RANGE(rusage.cpuacct_usage_percpu[2], DELAY1, SCALE),
+	     "cpu 1: user time: expected: %ld nsec, reported: %ld nsec\n",
+	     DELAY1, rusage.cpuacct_usage_percpu[2]);
+
+	printf("All tests finished\n");
+
+ out:
+	return ret;
+}
+

test/rusage/rusage104_mck.c

@@ -0,0 +1,221 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <sys/mman.h>
+#include <unistd.h>
+#include <string.h>
+#include <pthread.h>
+#include <errno.h>
+#include <sys/mman.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+
+
+#define DELAY0 (100UL * 1000 * 1000)
+#define DELAY1 (200UL * 1000 * 1000)
+
+struct thr_arg {
+	unsigned long delay;
+};
+
+struct thr_arg thr_arg[2] = { { .delay = DELAY0 }, { .delay = DELAY1 } };
+pthread_t thr[2];
+
+#define DIFFNSEC(end, start) ((end.tv_sec - start.tv_sec) * 1000000000UL + \
+		(end.tv_nsec - start.tv_nsec))
+#define TIMER_KIND CLOCK_MONOTONIC_RAW /* CLOCK_THREAD_CPUTIME_ID */
+
+static int print_cpu_last_executed_on(void)
+{
+	int ret = 0;
+	char fn[256];
+	char *result;
+	pid_t tid = syscall(SYS_gettid);
+	int fd;
+	int offset;
+	int amount = 0;
+	char *list;
+	char *token;
+	int i;
+	int cpu;
+
+	sprintf(fn, "/proc/%d/task/%d/stat", getpid(), (int)tid);
+	fd = open(fn, O_RDONLY);
+	if (fd == -1) {
+		printf("open() failed\n");
+		goto fn_fail;
+	}
+
+	result = malloc(65536);
+	if (result == NULL) {
+		printf("malloc() failed");
+		goto fn_fail;
+	}
+
+	offset = 0;
+	while (1) {
+		amount = read(fd, result + offset, 65536);
+		//		printf("amount=%d\n", amount);
+		if (amount == -1) {
+			printf("read() failed");
+			goto fn_fail;
+		}
+		if (amount == 0) {
+			goto eof;
+		}
+		offset += amount;
+	}
+ eof:;
+	//printf("result:%s\n", result);
+
+	list = result;
+	for (i = 0; i < 39; i++) {
+		token = strsep(&list, " ");
+	}
+
+	cpu = sched_getcpu();
+	if (cpu == -1) {
+		printf("getcpu() failed\n");
+		goto fn_fail;
+	}
+
+	printf("[INFO] stat-cpu=%02d,sched_getcpu=%02d,tid=%d\n",
+		token ? atoi(token) : -1, cpu, tid);
+ fn_exit:
+	free(result);
+	return ret;
+ fn_fail:
+	ret = -1;
+	goto fn_exit;
+}
+
+static inline void asm_loop(unsigned long n)
+{
+	int j;
+
+	for (j = 0; j < (n); j++) {
+		asm volatile(
+			     "movq $0, %%rcx\n\t"
+			     "1:\t"
+			     "addq $1, %%rcx\n\t"
+			     "cmpq $99, %%rcx\n\t"
+			     "jle 1b\n\t"
+			     :
+			     :
+			     : "rcx", "cc");
+	}
+}
+
+double nspw; /* nsec per work */
+unsigned long nsec;
+
+void fwq_init(void)
+{
+	struct timespec start, end;
+	int i;
+
+	clock_gettime(TIMER_KIND, &start);
+#define N_INIT 10000000
+	asm_loop(N_INIT);
+	clock_gettime(TIMER_KIND, &end);
+	nsec = DIFFNSEC(end, start);
+	nspw = nsec / (double)N_INIT;
+}
+
+#if 1
+void fwq(long delay_nsec)
+{
+	if (delay_nsec < 0) {
+		return;
+	}
+	asm_loop(delay_nsec / nspw);
+}
+#else
+/* For machines with large core-to-core performance variation (e.g. OFP) */
+void fwq(long delay_nsec)
+{
+	struct timespec start, end;
+
+	if (delay_nsec < 0) {
+		return;
+	}
+	clock_gettime(TIMER_KIND, &start);
+
+	while (1) {
+		clock_gettime(TIMER_KIND, &end);
+		if (DIFFNSEC(end, start) >= delay_nsec) {
+			break;
+		}
+		asm_loop(2); /* ~150 ns per iteration on FOP */
+	}
+}
+#endif
+
+void *util_thread(void *_arg)
+{
+	struct thr_arg *arg = (struct thr_arg *)_arg;
+
+	print_cpu_last_executed_on();
+	fwq(arg->delay);
+	pthread_exit(NULL);
+}
+
+int main(int argc, char **argv)
+{
+	int i;
+	int ret = 0;
+	cpu_set_t cpuset;
+	pthread_attr_t attr[2];
+
+	fwq_init();
+
+	/* Migrate to cpu#0 */
+	CPU_ZERO(&cpuset);
+	CPU_SET(0, &cpuset);
+	sched_setaffinity(0, sizeof(cpu_set_t), &cpuset);
+	print_cpu_last_executed_on();
+
+	for (i = 0; i < 2; i++) {
+		CPU_ZERO(&cpuset);
+		CPU_SET(i + 1, &cpuset);
+
+		if ((ret = pthread_attr_init(&attr[i]))) {
+			printf("%s: ERROR: pthread_attr_init failed (%d)\n",
+				__func__, ret);
+			ret = -EINVAL;
+			goto out;
+		}
+
+		if ((ret = pthread_attr_setaffinity_np(&attr[i],
+			sizeof(cpu_set_t), &cpuset))) {
+			printf("%s: ERROR: pthread_attr_setaffinity_np "
+				"failed (%d)\n", __func__, ret);
+			ret = -EINVAL;
+			goto out;
+		}
+
+		if ((ret = pthread_create(&thr[i], &attr[i],
+			util_thread, &thr_arg[i]))) {
+			fprintf(stderr, "ERROR: pthread_create failed (%d)\n",
+				ret);
+			ret = -EINVAL;
+			goto out;
+		}
+	}
+
+	for (i = 0; i < 2; i++) {
+		pthread_join(thr[i], NULL);
+	}
+
+	if ((ret = syscall(900))) {
+		fprintf(stderr, "%s: syscall failed\n", __func__);
+		ret = -EINVAL;
+		goto out;
+	}
+ out:
+	return ret;
+}
+

test/rusage/util.h

@@ -0,0 +1,44 @@
+#ifndef __UTIL_H_INCLUDED__
+#define __UTIL_H_INCLUDED__
+
+#define DEBUG
+
+#ifdef DEBUG
+#define dprintf(...) do {			 \
+	char msg[1024];			 \
+	sprintf(msg, __VA_ARGS__);		 \
+	fprintf(stderr, "%s,%s", __func__, msg); \
+} while (0)
+#else
+#define dprintf(...) do {  } while (0)
+#endif
+
+#define eprintf(...) do {			 \
+	char msg[1024];			 \
+	sprintf(msg, __VA_ARGS__);		 \
+	fprintf(stderr, "%s,%s", __func__, msg); \
+} while (0)
+
+#define CHKANDJUMP(cond, err, ...) do { \
+	if (cond) {			\
+		eprintf(__VA_ARGS__);   \
+		ret = err;		\
+		goto fn_fail;		\
+	}				\
+} while (0)
+
+
+#define OKNG(cond, ...) do {				\
+	if (cond) {					\
+		printf("[ OK ] ");			\
+		printf(__VA_ARGS__);			\
+	} else {					\
+		printf("[ NG ] ");			\
+		printf(__VA_ARGS__);			\
+		ret = -EINVAL;				\
+		goto out;				\
+	}						\
+} while (0)
+
+#endif
+