spec: prerelase 0.9 for testing libdwarf related package requirements

Change-Id: Iaaa116018505c4f89813883f5a99c8194cb4f99e

.gitmodules

@@ -1,3 +1,6 @@
 [submodule "ihk"]
 	path = ihk
 	url = https://github.com/RIKEN-SysSoft/ihk.git
+[submodule "executer/user/lib/libdwarf/libdwarf"]
+	path = executer/user/lib/libdwarf/libdwarf
+	url = https://github.com/bgerofi/libdwarf.git

CMakeLists.txt

@@ -10,7 +10,7 @@ project(mckernel C ASM)
 set(MCKERNEL_VERSION "1.7.0")
 
 # See "Fedora Packaging Guidlines -- Versioning"
-set(MCKERNEL_RELEASE "0.2")
+set(MCKERNEL_RELEASE "0.9")
 
 set(CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/cmake/modules)
 # for rpmbuild
@@ -34,6 +34,7 @@ include(GNUInstallDirs)
 include(CMakeParseArguments)
 include(Kbuild)
 include(CheckCCompilerFlag)
+include(AutoconfHelper)
 
 CHECK_C_COMPILER_FLAG(-Wno-implicit-fallthrough IMPLICIT_FALLTHROUGH)
 if(IMPLICIT_FALLTHROUGH)
@@ -119,9 +120,36 @@ find_package(PkgConfig REQUIRED)
 set(PKG_CONFIG_USE_CMAKE_PREFIX_PATH ON)
 
 find_library(LIBRT rt)
+if (NOT LIBRT)
+	message(FATAL_ERROR "error: couldn't find librt")
+endif()
 find_library(LIBNUMA numa)
+if (NOT LIBNUMA)
+	message(FATAL_ERROR "error: couldn't find libnuma")
+endif()
 find_library(LIBBFD bfd)
+if (NOT LIBBFD)
+	message(FATAL_ERROR "error: couldn't find libbfd")
+endif()
 find_library(LIBIBERTY iberty)
+if (NOT LIBIBERTY)
+	message(FATAL_ERROR "error: couldn't find libiberty")
+endif()
+
+find_library(LIBDWARF dwarf)
+
+if (NOT LIBDWARF)
+	if (CMAKE_CROSSCOMPILING)
+		message(FATAL_ERROR "Could not find libdwarf.so, install libdwarf-devel to ${CMAKE_FIND_ROOT_PATH}")
+	endif()
+	message("WARNING: libdwarf will be compiled locally")
+	enable_language(CXX)
+else()
+	# Note that libdwarf-devel provides /usr/include/libdwarf/dwarf.h
+	# but elfutils-devel provides /usr/include/dwarf.h
+	# while mcinspect.c performs "#include <dwarf.h>"
+	find_path(DWARF_H dwarf.h PATH_SUFFIXES libdwarf)
+endif()
 
 if (ENABLE_QLMPI)
 	find_package(MPI REQUIRED)
@@ -186,10 +214,14 @@ add_subdirectory(tools/crash)
 configure_file(scripts/mcreboot-smp.sh.in mcreboot.sh @ONLY)
 configure_file(scripts/mcstop+release-smp.sh.in mcstop+release.sh @ONLY)
 configure_file(scripts/mcreboot.1in mcreboot.1 @ONLY)
+configure_file(scripts/eclair-dump-backtrace.exp.in eclair-dump-backtrace.exp @ONLY)
 install(PROGRAMS
 		"${CMAKE_CURRENT_BINARY_DIR}/mcreboot.sh"
 		"${CMAKE_CURRENT_BINARY_DIR}/mcstop+release.sh"
 	DESTINATION "${CMAKE_INSTALL_SBINDIR}")
+install(PROGRAMS
+		"${CMAKE_CURRENT_BINARY_DIR}/eclair-dump-backtrace.exp"
+	DESTINATION "${CMAKE_INSTALL_BINDIR}")
 install(FILES "scripts/irqbalance_mck.in"
 	DESTINATION "${CMAKE_INSTALL_SYSCONFDIR}")
 install(FILES "${CMAKE_CURRENT_BINARY_DIR}/mcreboot.1"

NEWS.md

@@ -532,3 +532,9 @@ Restrictions on McKernel
 
 24. ihk_os_getperfevent() could time-out when invoked from Fujitsu TCS
     (job-scheduler).
+
+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.

README.md

@@ -64,7 +64,7 @@ First, the proxy process is compiled as a position independent binary, which ena
 
 For a smooth experience, we recommend the following combination of OS distributions and platforms:
 
-- CentOS 7.3+ running on Intel Xeon / Xeon Phi
+- CentOS 7.3+ running on Intel Xeon, Xeon Phi, Fujitsu A64FX
 
 
 ##### 1. Change SELinux settings
@@ -85,7 +85,12 @@ sudo reboot
 You will need the following packages installed:
 
 ~~~~
-sudo yum install cmake kernel-devel binutils-devel systemd-devel numactl-devel gcc make nasm git
+sudo yum install cmake kernel-devel binutils-devel systemd-devel numactl-devel gcc make nasm git libdwarf-devel
+~~~~
+
+Note that to install libdwarf-devel to RHEL-8.2, you need to enable the CodeReady Linux Builder (CLB) repository and the EPEL repository with the following commands:
+~~~~
+sudo subscription-manager repos --enable codeready-builder-for-rhel-8-$(/bin/arch)-rpms
 ~~~~
 
 Grant read permission to the System.map file of your kernel version:
@@ -128,7 +133,7 @@ The IHK kernel modules and McKernel kernel image should be installed under the *
 
 ###### 4.2 Install with rpm
 
-Configure, compile and build rpm:
+Build rpm:
 
 ~~~~
 mkdir -p build && cd build
@@ -179,7 +184,7 @@ IHK/McKernel booted.
 ~~~~
 
 
-##### 5. Run a simple program on McKernel
+##### 6. Run a simple program on McKernel
 
 The mcexec command line tool (which is also the Linux proxy process) can be used for executing applications on McKernel:
 
@@ -189,7 +194,7 @@ centos-vm
 ~~~~
 
 
-##### 6. Shutdown McKernel
+##### 7. Shutdown McKernel
 
 Finally, to shutdown McKernel and release CPU/memory resources back to Linux use the following command:
 
@@ -197,27 +202,27 @@ Finally, to shutdown McKernel and release CPU/memory resources back to Linux use
 sudo ./sbin/mcstop+release.sh
 ~~~~
 
-##### 7. Advanced: Enable Utility Thread offloading Interface (UTI)
+##### 8. Advanced: Enable Utility Thread offloading Interface (UTI)
 
 UTI enables a runtime such as MPI runtime to spawn utility threads such as MPI asynchronous progress threads to Linux cores.
 
-1. Install capstone
+###### 8.1 Install capstone
 
 Install EPEL capstone-devel:
 
 ~~~~
-sudo yum install epel-release
+sudo yum install https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm
 sudo yum install capstone-devel
 ~~~~
 
-2. Install syscall_intercept
+###### 8.2 Install syscall_intercept
 
 ~~~~
 git clone https://github.com/RIKEN-SysSoft/syscall_intercept.git
 cmake ../arch/aarch64 -DCMAKE_INSTALL_PREFIX=<syscall-intercept-install> -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=gcc -DTREAT_WARNINGS_AS_ERRORS=OFF
 ~~~~
 
-3. Install UTI for McKernel
+###### 8.3 Install UTI for McKernel
 
 Install:
 
@@ -228,19 +233,19 @@ mkdir build && cd build
 make && make install
 ~~~~
 
-4. Install McKernel
+###### 8.4 Install McKernel
 
 ~~~~
 CMAKE_PREFIX_PATH=<syscall-intercept-install> cmake -DCMAKE_INSTALL_PREFIX=${HOME}/ihk+mckernel -DENABLE_UTI=ON $HOME/src/ihk+mckernel/mckernel
 ~~~~
 
-5. Run executable
+###### 8.5 Run executable
 
 ~~~~
 mcexec --enable-uti <command>
 ~~~~
 
-6. Install UTI for Linux for performance comparison
+###### 8.6 Install UTI for Linux for performance comparison
 
 Install by make:
 

arch/arm64/kernel/cpu.c

@@ -1497,6 +1497,19 @@ struct thread *arch_switch_context(struct thread *prev, struct thread *next)
 		}
 	}
 #endif /*ENABLE_PERF*/
+
+#ifdef PROFILE_ENABLE
+	if (prev && prev->profile && prev->profile_start_ts != 0) {
+		prev->profile_elapsed_ts +=
+			(rdtsc() - prev->profile_start_ts);
+		prev->profile_start_ts = 0;
+	}
+
+	if (next->profile && next->profile_start_ts == 0) {
+		next->profile_start_ts = rdtsc();
+	}
+#endif
+
 	if (likely(prev)) {
 		tls_thread_switch(prev, next);
 
@@ -1867,6 +1880,11 @@ int arch_cpu_read_write_register(
 		ret = -1;
 	}
 
+	dkprintf("%s: MCCTRL_OS_CPU_%s_REGISTER: reg: 0x%lx, val: 0x%lx\n",
+			__FUNCTION__,
+			(op == MCCTRL_OS_CPU_READ_REGISTER ? "READ" : "WRITE"),
+			desc->addr, desc->val);
+
 	return ret;
 }
 

arch/arm64/kernel/include/rlimit.h

@@ -85,7 +85,11 @@ enum __rlimit_resource
   __RLIMIT_RTPRIO = 14,
 #define RLIMIT_RTPRIO __RLIMIT_RTPRIO
 
-  __RLIMIT_NLIMITS = 15,
+  /* timeout for RT tasks in us */
+  __RLIMIT_RTTIME = 15,
+#define RLIMIT_RTTIME __RLIMIT_RTTIME
+
+  __RLIMIT_NLIMITS = 16,
   __RLIM_NLIMITS = __RLIMIT_NLIMITS
 #define RLIMIT_NLIMITS __RLIMIT_NLIMITS
 #define RLIM_NLIMITS __RLIM_NLIMITS

arch/arm64/kernel/include/syscall_list.h

@@ -118,6 +118,7 @@ SYSCALL_HANDLED(241, perf_event_open)
 SYSCALL_DELEGATED(241, perf_event_open)
 #endif // PERF_ENABLE
 SYSCALL_HANDLED(260, wait4)
+SYSCALL_HANDLED(261, prlimit64)
 SYSCALL_HANDLED(270, process_vm_readv)
 SYSCALL_HANDLED(271, process_vm_writev)
 SYSCALL_HANDLED(281, execveat)
@@ -143,3 +144,9 @@ SYSCALL_HANDLED(1045, signalfd)
 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/irq-gic-v3.c

@@ -310,6 +310,9 @@ void handle_interrupt_gicv3(struct pt_regs *regs)
 {
 	uint64_t irqnr;
 	const int from_user = interrupt_from_user(regs);
+	struct cpu_local_var *v = get_this_cpu_local_var();
+	//unsigned long irqflags;
+	int do_check = 0;
 
 	irqnr = gic_read_iar();
 	cpu_enable_nmi();
@@ -323,10 +326,18 @@ void handle_interrupt_gicv3(struct pt_regs *regs)
 	}
 	set_cputime(from_user ? CPUTIME_MODE_K2U : CPUTIME_MODE_K2K_OUT);
 
-	/* for migration by IPI */
-	if (get_this_cpu_local_var()->flags & CPU_FLAG_NEED_MIGRATE) {
-		schedule();
+	//irqflags = ihk_mc_spinlock_lock(&v->runq_lock);
+	/* For migration by IPI or by timesharing */
+	if (v->flags &
+	    (CPU_FLAG_NEED_MIGRATE | CPU_FLAG_NEED_RESCHED)) {
+		v->flags &= ~CPU_FLAG_NEED_RESCHED;
+		do_check = 1;
+	}
+	//ihk_mc_spinlock_unlock(&v->runq_lock, irqflags);
+
+	if (do_check) {
 		check_signal(0, regs, 0);
+		schedule();
 	}
 }
 

arch/arm64/kernel/mikc.c

@@ -19,7 +19,7 @@ int ihk_mc_ikc_init_first_local(struct ihk_ikc_channel_desc *channel,
 
 	memset(channel, 0, sizeof(struct ihk_ikc_channel_desc));
 
-	mikc_queue_pages = ((4 * num_processors * MASTER_IKCQ_PKTSIZE)
+	mikc_queue_pages = ((8 * num_processors * MASTER_IKCQ_PKTSIZE)
 			+ (PAGE_SIZE - 1)) / PAGE_SIZE;
 
 	/* Place both sides in this side */

arch/arm64/kernel/perfctr_armv8pmu.c

@@ -590,7 +590,7 @@ static inline void armv8pmu_write_counter(int idx, uint32_t value)
 		 * count using the lower 32bits and we want an interrupt when
 		 * it overflows.
 		 */
-		uint64_t value64 = 0xffffffff00000000ULL | value;
+		uint64_t value64 = (int32_t)value;
 
 		write_sysreg(value64, pmccntr_el0);
 	}

arch/arm64/kernel/syscall.c

@@ -15,7 +15,6 @@
 #include <limits.h>
 #include <uio.h>
 #include <syscall.h>
-#include <bitops.h>
 #include <rusage_private.h>
 #include <ihk/debug.h>
 
@@ -58,13 +57,34 @@ extern int num_processors;
 int obtain_clone_cpuid(cpu_set_t *cpu_set, int use_last)
 {
 	int min_queue_len = -1;
-	int cpu, min_cpu = -1, uti_cpu = -1;
-	unsigned long irqstate;
+	int cpu, min_cpu = -1;
+#if 0
+	int uti_cpu = -1;
+#endif
+	unsigned long irqstate = 0;
 
-	irqstate = ihk_mc_spinlock_lock(&runq_reservation_lock);
+	int start, end, step;
+
+	if (use_last) {
+		start = num_processors - 1;
+		end = -1;
+		step = -1;
+	}
+	else {
+		start = 0;
+		end = num_processors;
+		step = 1;
+	}
+
+	if (!cpu_local_var(current)->proc->nr_processes) {
+		irqstate = ihk_mc_spinlock_lock(&runq_reservation_lock);
+	}
+	else {
+		irqstate = cpu_disable_interrupt_save();
+	}
 
 	/* Find the first allowed core with the shortest run queue */
-	for (cpu = 0; cpu < num_processors; ++cpu) {
+	for (cpu = start; cpu != end; cpu += step) {
 		struct cpu_local_var *v;
 
 		if (!CPU_ISSET(cpu, cpu_set))
@@ -75,11 +95,14 @@ int obtain_clone_cpuid(cpu_set_t *cpu_set, int use_last)
 		dkprintf("%s: cpu=%d,runq_len=%d,runq_reserved=%d\n",
 			 __func__, cpu, v->runq_len, v->runq_reserved);
 		if (min_queue_len == -1 ||
-		    v->runq_len + v->runq_reserved < min_queue_len) {
-			min_queue_len = v->runq_len + v->runq_reserved;
+		    //v->runq_len + v->runq_reserved < min_queue_len) {
+		    v->runq_len < min_queue_len) {
+			//min_queue_len = v->runq_len + v->runq_reserved;
+			min_queue_len = v->runq_len;
 			min_cpu = cpu;
 		}
 
+#if 0
 		/* Record the last tie CPU */
 		if (min_cpu != cpu &&
 		    v->runq_len + v->runq_reserved == min_queue_len) {
@@ -88,14 +111,15 @@ int obtain_clone_cpuid(cpu_set_t *cpu_set, int use_last)
 		dkprintf("%s: cpu=%d,runq_len=%d,runq_reserved=%d,min_cpu=%d,uti_cpu=%d\n",
 			 __func__, cpu, v->runq_len, v->runq_reserved,
 			 min_cpu, uti_cpu);
+#else
 
 		ihk_mc_spinlock_unlock_noirq(&v->runq_lock);
-#if 0
 		if (min_queue_len == 0)
 			break;
 #endif
 	}
 
+#if 0
 	min_cpu = use_last ? uti_cpu : min_cpu;
 	if (min_cpu != -1) {
 		if (get_cpu_local_var(min_cpu)->status != CPU_STATUS_RESERVED)
@@ -104,22 +128,20 @@ int obtain_clone_cpuid(cpu_set_t *cpu_set, int use_last)
 		__sync_fetch_and_add(&get_cpu_local_var(min_cpu)->runq_reserved,
 				     1);
 	}
-	ihk_mc_spinlock_unlock(&runq_reservation_lock, irqstate);
+#else
+	__sync_fetch_and_add(&get_cpu_local_var(min_cpu)->runq_reserved, 1);
+#endif
+
+	if (!cpu_local_var(current)->proc->nr_processes) {
+		ihk_mc_spinlock_unlock(&runq_reservation_lock, irqstate);
+	}
+	else {
+		cpu_restore_interrupt(irqstate);
+	}
 
 	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,
@@ -152,33 +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;
@@ -1335,70 +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)
 {
@@ -1422,191 +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)
 {
@@ -2043,7 +1789,7 @@ SYSCALL_DECLARE(mmap)
 		goto out;
 	}
 
-	addr = do_mmap(addr, len, prot, flags, fd, off0, 0, NULL);
+	addr = do_mmap(addr, len, prot, flags, fd, off0);
 
 	error = 0;
 out:
@@ -2137,7 +1883,9 @@ int do_process_vm_read_writev(int pid,
 	unsigned long lpage_left, rpage_left;
 	unsigned long lpsize, rpsize;
 	void *rva, *lva;
+#if 0
 	struct vm_range *range;
+#endif
 	struct mcs_rwlock_node_irqsave lock;
 	struct mcs_rwlock_node update_lock;
 
@@ -2150,8 +1898,9 @@ int do_process_vm_read_writev(int pid,
 		return -EINVAL;
 	}
 
+#if 0
 	/* Check if parameters are okay */
-	ihk_mc_spinlock_lock_noirq(&lthread->vm->memory_range_lock);
+	ihk_rwspinlock_read_lock_noirq(&lthread->vm->memory_range_lock);
 
 	range = lookup_process_memory_range(lthread->vm, 
 			(uintptr_t)local_iov, 
@@ -2173,11 +1922,12 @@ int do_process_vm_read_writev(int pid,
 
 	ret = 0;
 arg_out:
-	ihk_mc_spinlock_unlock_noirq(&lthread->vm->memory_range_lock);
+	ihk_rwspinlock_read_unlock_noirq(&lthread->vm->memory_range_lock);
 
 	if (ret != 0) {
 		goto out;
 	}
+#endif
 
 	for (li = 0; li < liovcnt; ++li) {
 		llen += local_iov[li].iov_len;
@@ -2242,7 +1992,7 @@ arg_out:
 		if (pli != li) {
 			struct vm_range *range;
 
-			ihk_mc_spinlock_lock_noirq(&lthread->vm->memory_range_lock);
+			ihk_rwspinlock_read_lock_noirq(&lthread->vm->memory_range_lock);
 
 			/* Is base valid? */
 			range = lookup_process_memory_range(lthread->vm,
@@ -2272,7 +2022,7 @@ arg_out:
 
 			ret = 0;
 pli_out:
-			ihk_mc_spinlock_unlock_noirq(&lthread->vm->memory_range_lock);
+			ihk_rwspinlock_read_unlock_noirq(&lthread->vm->memory_range_lock);
 
 			if (ret != 0) {
 				goto out;
@@ -2285,7 +2035,7 @@ pli_out:
 		if (pri != ri) {
 			struct vm_range *range;
 
-			ihk_mc_spinlock_lock_noirq(&rvm->memory_range_lock);
+			ihk_rwspinlock_read_lock_noirq(&rvm->memory_range_lock);
 
 			/* Is base valid? */
 			range = lookup_process_memory_range(rvm,
@@ -2315,7 +2065,7 @@ pli_out:
 
 			ret = 0;
 pri_out:
-			ihk_mc_spinlock_unlock_noirq(&rvm->memory_range_lock);
+			ihk_rwspinlock_read_unlock_noirq(&rvm->memory_range_lock);
 
 			if (ret != 0) {
 				goto out;

arch/arm64/kernel/vdso.c

@@ -155,7 +155,7 @@ int arch_map_vdso(struct process_vm *vm)
 	flag = VR_REMOTE | VR_PROT_READ;
 	flag |= VRFLAG_PROT_TO_MAXPROT(flag);
 	ret = add_process_memory_range(vm, start, end, vdso.vvar_phys, flag,
-				       NULL, 0, PAGE_SHIFT, NULL, &range);
+				       NULL, 0, PAGE_SHIFT, &range);
 	if (ret != 0){
 		dkprintf("ERROR: adding memory range for tod_data\n");
 		goto exit;
@@ -167,7 +167,7 @@ int arch_map_vdso(struct process_vm *vm)
 	flag = VR_REMOTE | VR_PROT_READ | VR_PROT_EXEC;
 	flag |= VRFLAG_PROT_TO_MAXPROT(flag);
 	ret = add_process_memory_range(vm, start, end, vdso.vdso_physlist[0], flag,
-				       NULL, 0, PAGE_SHIFT, NULL, &range);
+				       NULL, 0, PAGE_SHIFT, &range);
 	if (ret != 0) {
 		dkprintf("ERROR: adding memory range for vdso_text\n");
 

arch/x86_64/kernel/include/rlimit.h

@@ -84,7 +84,11 @@ enum __rlimit_resource
   __RLIMIT_RTPRIO = 14,
 #define RLIMIT_RTPRIO __RLIMIT_RTPRIO
 
-  __RLIMIT_NLIMITS = 15,
+  /* timeout for RT tasks in us */
+  __RLIMIT_RTTIME = 15,
+#define RLIMIT_RTTIME __RLIMIT_RTTIME
+
+  __RLIMIT_NLIMITS = 16,
   __RLIM_NLIMITS = __RLIMIT_NLIMITS
 #define RLIMIT_NLIMITS __RLIMIT_NLIMITS
 #define RLIM_NLIMITS __RLIM_NLIMITS

arch/x86_64/kernel/include/syscall_list.h

@@ -158,6 +158,7 @@ SYSCALL_HANDLED(289, signalfd4)
 #ifdef ENABLE_PERF
 SYSCALL_HANDLED(298, perf_event_open)
 #endif
+SYSCALL_HANDLED(302, prlimit64)
 #ifdef DCFA_KMOD
 SYSCALL_HANDLED(303, mod_call)
 #endif
@@ -180,4 +181,8 @@ 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

@@ -31,7 +31,6 @@
 #include <page.h>
 #include <limits.h>
 #include <syscall.h>
-#include <bitops.h>
 #include <rusage_private.h>
 #include <ihk/debug.h>
 
@@ -148,44 +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;
@@ -1040,82 +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)
 {
@@ -1130,175 +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)
@@ -1786,7 +1502,7 @@ recheck:
 		goto out;
 	}
 
-	addr = do_mmap(addr, len, prot, flags, fd, off0, 0, NULL);
+	addr = do_mmap(addr, len, prot, flags, fd, off0);
 
 	error = 0;
 out:
@@ -2115,7 +1831,7 @@ int arch_map_vdso(struct process_vm *vm)
 	vrflags |= VR_PROT_READ | VR_PROT_EXEC;
 	vrflags |= VRFLAG_PROT_TO_MAXPROT(vrflags);
 	error = add_process_memory_range(vm, (intptr_t)s, (intptr_t)e,
-			NOPHYS, vrflags, NULL, 0, PAGE_SHIFT, NULL, &range);
+			NOPHYS, vrflags, NULL, 0, PAGE_SHIFT, &range);
 	if (error) {
 		ekprintf("ERROR: adding memory range for vdso. %d\n", error);
 		goto out;
@@ -2147,8 +1863,7 @@ int arch_map_vdso(struct process_vm *vm)
 		vrflags |= VR_PROT_READ;
 		vrflags |= VRFLAG_PROT_TO_MAXPROT(vrflags);
 		error = add_process_memory_range(vm, (intptr_t)s, (intptr_t)e,
-				NOPHYS, vrflags, NULL, 0,
-				PAGE_SHIFT, NULL, &range);
+				NOPHYS, vrflags, NULL, 0, PAGE_SHIFT, &range);
 		if (error) {
 			ekprintf("ERROR: adding memory range for vvar. %d\n", error);
 			goto out;
@@ -2290,7 +2005,7 @@ int do_process_vm_read_writev(int pid,
 	}
 
 	/* Check if parameters are okay */
-	ihk_mc_spinlock_lock_noirq(&lthread->vm->memory_range_lock);
+	ihk_rwspinlock_read_lock_noirq(&lthread->vm->memory_range_lock);
 
 	range = lookup_process_memory_range(lthread->vm, 
 			(uintptr_t)local_iov, 
@@ -2312,7 +2027,7 @@ int do_process_vm_read_writev(int pid,
 
 	ret = 0;
 arg_out:
-	ihk_mc_spinlock_unlock_noirq(&lthread->vm->memory_range_lock);
+	ihk_rwspinlock_read_unlock_noirq(&lthread->vm->memory_range_lock);
 
 	if (ret != 0) {
 		goto out;
@@ -2381,7 +2096,7 @@ arg_out:
 		if (pli != li) {
 			struct vm_range *range;
 
-			ihk_mc_spinlock_lock_noirq(&lthread->vm->memory_range_lock);
+			ihk_rwspinlock_read_lock_noirq(&lthread->vm->memory_range_lock);
 
 			/* Is base valid? */
 			range = lookup_process_memory_range(lthread->vm,
@@ -2411,7 +2126,7 @@ arg_out:
 
 			ret = 0;
 pli_out:
-			ihk_mc_spinlock_unlock_noirq(&lthread->vm->memory_range_lock);
+			ihk_rwspinlock_read_unlock_noirq(&lthread->vm->memory_range_lock);
 
 			if (ret != 0) {
 				goto out;
@@ -2424,7 +2139,7 @@ pli_out:
 		if (pri != ri) {
 			struct vm_range *range;
 
-			ihk_mc_spinlock_lock_noirq(&rvm->memory_range_lock);
+			ihk_rwspinlock_read_lock_noirq(&rvm->memory_range_lock);
 
 			/* Is base valid? */
 			range = lookup_process_memory_range(rvm,
@@ -2454,7 +2169,7 @@ pli_out:
 
 			ret = 0;
 pri_out:
-			ihk_mc_spinlock_unlock_noirq(&rvm->memory_range_lock);
+			ihk_rwspinlock_read_unlock_noirq(&rvm->memory_range_lock);
 
 			if (ret != 0) {
 				goto out;

cmake/modules/AutoconfHelper.cmake

@@ -0,0 +1,383 @@
+# Helper functions for translating autoconf projects. Several functions
+# are lifted from the Mono sources
+
+include (CheckCSourceCompiles)
+include (CheckIncludeFile)
+include (TestBigEndian)
+include (CheckFunctionExists)
+include (CheckTypeSize)
+include (CheckCSourceRuns)
+
+
+# Function to get the version information from the configure.ac file in the
+# current directory. Its argument is the name of the library as passed to
+# AC_INIT. It will set the variables ${LIBNAME}_VERSION and ${LIBNAME}_SOVERSION
+function (ac_get_version libname)
+  string(TOUPPER "${libname}" libname_upper)
+
+  # Read the relevant content from configure.ac
+  file (STRINGS configure.ac tmp_configure_ac
+    REGEX "${libname_upper}_[_A-Z]+=[ \\t]*[0-9]+")
+
+  # Product version
+  string (REGEX REPLACE ".+MAJOR[_A-Z]+=([0-9]+).+MINOR[_A-Z]+=([0-9]+).+MICRO[_A-Z]+=([0-9]+).*"
+    "\\1.\\2.\\3" ${libname_upper}_VERSION "${tmp_configure_ac}")
+
+  # Library version for libtool
+  string (REGEX REPLACE ".+CURRENT=([0-9]+).+REVISION=([0-9]+).+AGE=([0-9]+).*"
+    "\\1.\\2.\\3" ${libname_upper}_SOVERSION "${tmp_configure_ac}")
+
+  # Checks if the string needs to be displayed
+  set (${libname_upper}_DISPLAYSTR_AUX
+    "Found ${libname} version ${${libname_upper}_VERSION}, soversion ${${libname_upper}_SOVERSION} from configure.ac"
+  )
+  if ((NOT ${libname_upper}_DISPLAYSTR) OR (NOT ${libname_upper}_DISPLAYSTR STREQUAL ${libname_upper}_DISPLAYSTR_AUX))
+    set (${libname_upper}_DISPLAYSTR ${${libname_upper}_DISPLAYSTR_AUX}
+      CACHE INTERNAL "Version string from ${libname}" FORCE)
+    message (STATUS ${${libname_upper}_DISPLAYSTR})
+  endif ()
+
+  # Export the result to the caller
+  set(${libname_upper}_VERSION "${${libname_upper}_VERSION}" PARENT_SCOPE)
+  set(${libname_upper}_SOVERSION "${${libname_upper}_SOVERSION}" PARENT_SCOPE)
+endfunction()
+
+
+# Also from mono's source code
+# Implementation of AC_CHECK_HEADERS
+# In addition, it also records the list of variables in the variable
+# 'autoheader_vars', and for each variable, a documentation string in the
+# variable ${var}_doc
+function(ac_check_headers)
+  foreach (header ${ARGV})
+	string(TOUPPER ${header} header_var)
+	string(REPLACE "." "_" header_var ${header_var})
+	string(REPLACE "/" "_" header_var ${header_var})
+	set(header_var "HAVE_${header_var}")
+	check_include_file (${header} ${header_var})
+	set("${header_var}_doc" "Define to 1 if you have the <${header}> header file." PARENT_SCOPE)
+	if (${header_var})
+	  set("${header_var}_defined" "1" PARENT_SCOPE)
+	endif()
+	set("${header_var}_val" "1" PARENT_SCOPE)
+	set (autoheader_vars ${autoheader_vars} ${header_var})
+  endforeach()
+  set (autoheader_vars ${autoheader_vars} PARENT_SCOPE)
+endfunction()
+
+# Function taken from mono's source code
+function (ac_check_funcs)
+  foreach (func ${ARGV})
+    string(TOUPPER ${func} var)
+    set(var "HAVE_${var}")
+    set(${var})
+    check_function_exists (${func} ${var})
+    set("${var}_doc" "Define to 1 if you have the '${func}' function." PARENT_SCOPE)
+    if (${var})
+      set("${var}_defined" "1" PARENT_SCOPE)
+      set(${var} yes PARENT_SCOPE)
+    endif()
+    set("${var}_val" "1" PARENT_SCOPE)
+    set (autoheader_vars ${autoheader_vars} ${var})
+  endforeach()
+  set (autoheader_vars ${autoheader_vars} PARENT_SCOPE)
+endfunction()
+
+
+# Specifically, this macro checks for stdlib.h', stdarg.h',
+# string.h', and float.h'; if the system has those, it probably
+# has the rest of the ANSI C header files.  This macro also checks
+# whether string.h' declares memchr' (and thus presumably the
+# other mem' functions), whether stdlib.h' declare free' (and
+# thus presumably malloc' and other related functions), and whether
+# the ctype.h' macros work on characters with the high bit set, as
+# ANSI C requires.
+function (ac_header_stdc)
+  if (STDC_HEADERS)
+    return()
+  endif()
+  message(STATUS "Looking for ANSI-C headers")
+  set(code "
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <float.h>
+
+int main(int argc, char **argv)
+{
+  void *ptr;
+  free((void*)1);
+  ptr = memchr((void*)1, 0, 0);
+
+  return (int)ptr;
+}
+")
+  # FIXME Check the ctype.h high bit
+  CHECK_C_SOURCE_COMPILES("${code}" STDC_HEADERS)
+  if (STDC_HEADERS)
+    set(STDC_HEADERS 1 PARENT_SCOPE)
+	message(STATUS "Looking for ANSI-C headers - found")
+  else()
+    message(STATUS "Looking for ANSI-C headers - not found")
+  endif()
+endfunction()
+
+
+# Also from the mono sources, kind of implements AC_SYS_LARGEFILE
+function (ac_sys_largefile)
+  CHECK_C_SOURCE_RUNS("
+#include <sys/types.h>
+#define BIG_OFF_T (((off_t)1<<62)-1+((off_t)1<<62))
+int main (int argc, char **argv) {
+    int big_off_t=((BIG_OFF_T%2147483629==721) &&
+                   (BIG_OFF_T%2147483647==1));
+    return big_off ? 0 : 1;
+}
+" HAVE_LARGE_FILE_SUPPORT)
+
+# Check if it makes sense to define _LARGE_FILES or _FILE_OFFSET_BITS
+  if (HAVE_LARGE_FILE_SUPPORT)
+    return()
+  endif()
+
+  set (_LARGE_FILE_EXTRA_SRC "
+#include <sys/types.h>
+int main (int argc, char **argv) {
+  return sizeof(off_t) == 8 ? 0 : 1;
+}
+")
+  CHECK_C_SOURCE_RUNS ("#define _LARGE_FILES\n${_LARGE_FILE_EXTRA_SRC}"
+    HAVE_USEFUL_D_LARGE_FILES)
+  if (NOT HAVE_USEFUL_D_LARGE_FILES)
+    if (NOT DEFINED HAVE_USEFUL_D_FILE_OFFSET_BITS)
+      set (SHOW_LARGE_FILE_WARNING TRUE)
+    endif ()
+    CHECK_C_SOURCE_RUNS ("#define _FILE_OFFSET_BITS 64\n${_LARGE_FILE_EXTRA_SRC}"
+      HAVE_USEFUL_D_FILE_OFFSET_BITS)
+	  if (HAVE_USEFUL_D_FILE_OFFSET_BITS)
+      set (_FILE_OFFSET_BITS 64 PARENT_SCOPE)
+	  elseif (SHOW_LARGE_FILE_WARNING)
+	    message (WARNING "No 64 bit file support through off_t available.")
+	  endif ()
+  else ()
+    set (_LARGE_FILES 1 PARENT_SCOPE)
+  endif ()
+endfunction ()
+
+
+# Quick way to set some basic variables
+# FIXME add support for variable number of arguments: only package and version are mandatory
+# arguments are package version bug_report tarname url
+function (ac_init)
+    set(package ${ARGV0})
+    set(version ${ARGV1})
+    set(bug_report ${ARGV2})
+    set(tarname ${ARGV3})
+    set(url ${ARGV4})
+    set(PACKAGE_NAME "\"${package}\"" PARENT_SCOPE)
+    set(PACKAGE_VERSION "\"${version}\"" PARENT_SCOPE)
+    set(VERSION "\"${version}\"" PARENT_SCOPE)
+    if(version)
+        set(PACKAGE_STRING "\"${package} ${version}\"" PARENT_SCOPE)
+    else()
+        set(PACKAGE_STRING "\"${package}\"" PARENT_SCOPE)
+    endif()
+
+    set(PACKAGE_BUGREPORT "\"${bug_report}\"" PARENT_SCOPE)
+
+    if(NOT tarname)
+        string(REGEX REPLACE "[^a-zA-Z0-9_]" "-" tarname "${package}")
+    endif()
+    set(PACKAGE_TARNAME "\"${tarname}\"" PARENT_SCOPE)
+
+    set(PACKAGE_URL "\"${url}\"" PARENT_SCOPE)
+endfunction()
+
+
+# Checks for the const keyword, defining "HAS_CONST_SUPPORT"
+# If it does not have support, defines "const" to 0 in the parent scope
+function (ac_c_const)
+  CHECK_C_SOURCE_COMPILES(
+    "int main(int argc, char **argv){const int r = 0;return r;}"
+    HAS_CONST_SUPPORT)
+  if (NOT HAS_CONST_SUPPORT)
+    set(const 0 PARENT_SCOPE)
+  endif()
+endfunction()
+
+
+# Inline keyword support. Defines "inline" in the parent scope to the
+# compiler internal keyword for inline in C
+# TODO write a better test!
+function (ac_c_inline)
+  if (MSVC)
+    set (inline __inline)
+  elseif(CMAKE_COMPILER_IS_GNUC)
+    set (inline __inline__)
+  endif()
+  set(inline "${inline}" PARENT_SCOPE)
+endfunction()
+
+
+# Test if you can safely include both <sys/time.h> and <time.h>
+function (ac_header_time)
+  CHECK_C_SOURCE_COMPILES(
+    "#include <sys/time.h>\n#include <time.h>\nint main(int argc, char **argv) { return 0; }"
+    TIME_WITH_SYS_TIME)
+  set(TIME_WITH_SYS_TIME ${TIME_WITH_SYS_TIME} PARENT_SCOPE)
+endfunction()
+
+
+# Native cpu byte order: 1 if big-endian (Motorola) or 0 if little-endian
+# (Intel), setting "WORDS_BIGENDIAN" to 1 if big endian
+function (ac_c_bigendian)
+  TEST_BIG_ENDIAN(HOST_BIGENDIAN)
+  if (HOST_BIGENDIAN)
+    set(WORDS_BIGENDIAN 1 PARENT_SCOPE)
+  endif()
+endfunction()
+
+
+# Check for off_t, setting "off_t" in the parent scope
+function(ac_type_off_t)
+  CHECK_TYPE_SIZE("off_t" SIZEOF_OFF_T)
+  if (NOT SIZEOF_OFF_T)
+    set(off_t "long int")
+  endif()
+  set(off_t ${off_t} PARENT_SCOPE)
+endfunction()
+
+
+# Check for size_t, setting "size_t" in the parent scope
+function(ac_type_size_t)
+  CHECK_TYPE_SIZE("size_t" SIZEOF_SIZE_T)
+  if (NOT SIZEOF_SIZE_T)
+    set(size_t "unsigned int")
+  endif()
+  set(size_t ${size_t} PARENT_SCOPE)
+endfunction()
+
+
+# Define "TM_IN_SYS_TIME" to 1 if <sys/time.h> declares "struct tm"
+function(ac_struct_tm)
+  CHECK_C_SOURCE_COMPILES(
+    "#include <sys/time.h>\nint main(int argc, char **argv) { struct tm x; return 0; }"
+    TM_IN_SYS_TIME
+  )
+  if (TM_IN_SYS_TIME)
+    set (TM_IN_SYS_TIME 1 PARENT_SCOPE)
+  endif()
+endfunction()
+
+
+# Obtain size of an 'type' and define as SIZEOF_TYPE
+function (ac_check_sizeof typename)
+  string(TOUPPER "SIZEOF_${typename}" varname)
+	string(REPLACE " " "_" varname "${varname}")
+  string(REPLACE "*" "p" varname "${varname}")
+  CHECK_TYPE_SIZE("${typename}" ${varname} BUILTIN_TYPES_ONLY)
+  if(NOT ${varname})
+    set(${varname} 0 PARENT_SCOPE)
+  endif()
+endfunction()
+
+
+# Check if the type exists, defines HAVE_<type>
+function (ac_check_type typename)
+  string(TOUPPER "${typename}" varname)
+	string(REPLACE " " "_" varname "${varname}")
+  string(REPLACE "*" "p" varname "${varname}")
+  CHECK_TYPE_SIZE("${typename}" ${varname})
+  if (NOT "${varname}" STREQUAL "")
+    set("HAVE_${varname}" 1 PARENT_SCOPE)
+    set("${varname}" "${typename}" PARENT_SCOPE)
+  else()
+    set("${varname}" "unknown" PARENT_SCOPE)
+  endif()
+endfunction()
+
+
+# Verifies if each type on the list exists, using the given prelude
+function (ac_check_types type_list prelude)
+  foreach(typename ${type_list})
+    string(TOUPPER "HAVE_${typename}" varname)
+    string(REPLACE " " "_" varname "${varname}")
+    string(REPLACE "*" "p" varname "${varname}")
+    CHECK_C_SOURCE_COMPILES("${prelude}\n ${typename} foo;" ${varname})
+  endforeach()
+endfunction()
+
+function(ac_path_prog variable prog_to_check_for value_if_not_found env_var)
+    find_program(${variable} NAMES ${prog_to_check_for} PATHS ENV ${env_var} NO_DEFAULT_PATH)
+    if(NOT ${variable})
+        message(STATUS "Looking for ${prog_to_check_for} - not found")
+        set(${variable} ${value_if_not_fount} PARENT_SCOPE)
+    else()
+        message(STATUS "Looking for ${prog_to_check_for} - ${variable}")
+        set(${variable} ${${variable}} PARENT_SCOPE)
+    endif()
+endfunction()
+
+# check if function func exists in library lib
+function(ac_check_lib lib func)
+    string(TOUPPER "HAVE_${func}" varname)
+    set(CMAKE_REQUIRED_LIBRARIES ${lib})
+    check_function_exists(${func} ${varname})
+    set(CMAKE_REQUIRED_LIBRARIES)
+endfunction()
+
+# check if source compiles without linking
+function(ac_try_compile SOURCE VAR)
+    set(CMAKE_TMP_DIR ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp)
+    if(NOT DEFINED "${VAR}")
+        file(WRITE
+            "${CMAKE_TMP_DIR}/src.c"
+            "${SOURCE}\n"
+        )
+
+        if(NOT CMAKE_REQUIRED_QUIET)
+            message(STATUS "Performing Test ${VAR}")
+        endif()
+        # Set up CMakeLists.txt for static library:
+        file(WRITE
+            ${CMAKE_TMP_DIR}/CMakeLists.txt
+            "add_library(compile STATIC src.c)"
+        )
+
+        # Configure:
+        execute_process(
+            COMMAND ${CMAKE_COMMAND} -G "${CMAKE_GENERATOR}" .
+            WORKING_DIRECTORY ${CMAKE_TMP_DIR}
+        )
+
+        # Build:
+        execute_process(
+            COMMAND ${CMAKE_COMMAND} --build ${CMAKE_TMP_DIR}
+            RESULT_VARIABLE RESVAR
+            OUTPUT_VARIABLE OUTPUT
+            ERROR_VARIABLE OUTPUT
+        )
+
+        # Set up result:
+        if(RESVAR EQUAL 0)
+            set(${VAR} 1 CACHE INTERNAL "Test ${VAR}")
+            if(NOT CMAKE_REQUIRED_QUIET)
+                message(STATUS "Performing Test ${VAR} - Success")
+            endif()
+
+            file(APPEND ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeOutput.log
+                "Performing C SOURCE FILE Test ${VAR} succeded with the following output:\n"
+                "${OUTPUT}\n"
+                "Source file was:\n${SOURCE}\n")
+        else()
+            if(NOT CMAKE_REQUIRED_QUIET)
+                message(STATUS "Performing Test ${VAR} - Failed")
+            endif()
+            set(${VAR} "" CACHE INTERNAL "Test ${VAR}")
+            file(APPEND ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeError.log
+            "Performing C SOURCE FILE Test ${VAR} failed with the following output:\n"
+            "${OUTPUT}\n"
+            "Source file was:\n${SOURCE}\n")
+        endif()
+    endif()
+endfunction()

cmake/modules/FindLibElf.cmake

@@ -0,0 +1,64 @@
+# - Try to find libelf
+# Once done this will define
+#
+#  LIBELF_FOUND - system has libelf
+#  LIBELF_INCLUDE_DIRS - the libelf include directory
+#  LIBELF_LIBRARIES - Link these to use libelf
+#  LIBELF_DEFINITIONS - Compiler switches required for using libelf
+#
+# This module reads hints about search locations from variables:
+#
+#  LIBELF_ROOT - Preferred installation prefix
+#
+#  Copyright (c) 2008 Bernhard Walle <bernhard.walle@gmx.de>
+#
+#  Redistribution and use is allowed according to the terms of the New
+#  BSD license.
+#  For details see the accompanying COPYING-CMAKE-SCRIPTS file.
+#
+
+
+if (LIBELF_LIBRARIES AND LIBELF_INCLUDE_DIRS)
+  set (LibElf_FIND_QUIETLY TRUE)
+endif (LIBELF_LIBRARIES AND LIBELF_INCLUDE_DIRS)
+
+find_path (LIBELF_INCLUDE_DIRS
+  NAMES
+    libelf/libelf.h libelf.h
+  HINTS
+    ${LIBELF_ROOT}
+  PATH_SUFFIXES
+    include
+	libelf/include
+)
+
+find_library (LIBELF_LIBRARIES
+  NAMES
+    elf libelf
+  HINTS
+    ${LIBELF_ROOT}
+  PATH_SUFFIXES
+    lib
+	libelf/lib
+)
+
+include (FindPackageHandleStandardArgs)
+
+
+# handle the QUIETLY and REQUIRED arguments and set LIBELF_FOUND to TRUE if all listed variables are TRUE
+FIND_PACKAGE_HANDLE_STANDARD_ARGS(LibElf DEFAULT_MSG
+  LIBELF_LIBRARIES
+  LIBELF_INCLUDE_DIRS)
+
+set(CMAKE_REQUIRED_LIBRARIES elf)
+include(CheckCXXSourceCompiles)
+check_cxx_source_compiles("#include <libelf.h>
+int main() {
+  Elf *e = (Elf*)0;
+  size_t sz;
+  elf_getshdrstrndx(e, &sz);
+  return 0;
+}" ELF_GETSHDRSTRNDX)
+unset(CMAKE_REQUIRED_LIBRARIES)
+
+mark_as_advanced(LIBELF_INCLUDE_DIRS LIBELF_LIBRARIES ELF_GETSHDRSTRNDX)

executer/include/uprotocol.h

@@ -91,8 +91,9 @@ 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

@@ -587,14 +587,13 @@ 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 = NULL, *pe_itr;
+	struct mcctrl_part_exec *pe;
 	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;
@@ -615,54 +614,96 @@ static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
 	}
 
 	if (copy_from_user(&req, (void *)arg, sizeof(req))) {
-		pr_err("%s: error copying user request\n", __func__);
+		printk("%s: error copying user request\n", __FUNCTION__);
 		ret = -EINVAL;
 		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;
-		}
-	}
+	/* User requested CPU mask? */
+	if (req.req_cpu_list && req.req_cpu_list_len) {
+		char *cpu_list = NULL;
 
-	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);
+		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;
 		}
-		/* 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",
-				__func__, pe->nr_processes);
+		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_unlock(&udp->part_exec_lock);
+
+	pe = &udp->part_exec;
 
 	mutex_lock(&pe->lock);
 
+	/* First process to enter CPU partitioning */
+	if (pe->nr_processes == -1) {
+		pe->nr_processes = req.nr_processes;
+		pe->nr_processes_left = req.nr_processes;
+		dprintk("%s: nr_processes: %d (partitioned exec starts)\n",
+				__FUNCTION__,
+				pe->nr_processes);
+	}
+
 	if (pe->nr_processes != req.nr_processes) {
 		printk("%s: error: requested number of processes"
 				" doesn't match current partitioned execution\n",
@@ -671,15 +712,7 @@ 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,
@@ -765,6 +798,8 @@ 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;
 		}
@@ -1012,8 +1047,16 @@ next_cpu:
 	/* Commit used cores to OS structure */
 	memcpy(&pe->cpus_used, cpus_used, sizeof(*cpus_used));
 
-	/* If not last process, wake up next process in list */
-	if (pe->nr_processes_left != 0) {
+	/* Reset if last process */
+	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);
@@ -1237,7 +1280,7 @@ int mcexec_syscall(struct mcctrl_usrdata *ud, struct ikc_scd_packet *packet)
 	ppd = mcctrl_get_per_proc_data(ud, pid);
 
 	if (unlikely(!ppd)) {
-		kprintf("%s: ERROR: no per-process structure for PID %d, "
+		dprintk("%s: ERROR: no per-process structure for PID %d, "
 				"syscall nr: %lu\n",
 				__FUNCTION__, pid, packet->req.number);
 
@@ -1452,7 +1495,7 @@ retry_alloc:
 			__FUNCTION__, task_pid_vnr(current), packet->ref);
 
 	mb();
-	if (!packet->req.valid) {
+	if (!smp_load_acquire(&packet->req.valid)) {
 		printk("%s: ERROR: stray wakeup pid: %d, tid: %d: SC %lu\n",
 				__FUNCTION__,
 				task_tgid_vnr(current),
@@ -1462,7 +1505,7 @@ retry_alloc:
 		goto retry;
 	}
 
-	packet->req.valid = 0; /* ack */
+	smp_store_release(&packet->req.valid,  0); /* ack */
 	dprintk("%s: system call: %d, args[0]: %lu, args[1]: %lu, args[2]: %lu, "
 			"args[3]: %lu, args[4]: %lu, args[5]: %lu\n",
 			__FUNCTION__,
@@ -3399,14 +3442,6 @@ long __mcctrl_control(ihk_os_t os, unsigned int req, unsigned long arg,
 	return -EINVAL;
 }
 
-/* Per-CPU register manipulation functions */
-struct mcctrl_os_cpu_response {
-	int done;
-	unsigned long val;
-	int err;
-	wait_queue_head_t wq;
-};
-
 int mcctrl_get_request_os_cpu(ihk_os_t os, int *ret_cpu)
 {
 	struct mcctrl_usrdata *usrdata;
@@ -3470,31 +3505,14 @@ out_put_ppd:
 	return ret;
 }
 
-void mcctrl_os_read_write_cpu_response(ihk_os_t os,
-		struct ikc_scd_packet *pisp)
-{
-	struct mcctrl_os_cpu_response *resp;
-
-	/* XXX: What if caller thread is unblocked by a signal
-	 * before this message arrives? */
-	resp = pisp->resp;
-	if (!resp) {
-		return;
-	}
-
-	resp->val = pisp->desc.val;
-	resp->done = 1;
-	resp->err = pisp->err;
-	wake_up_interruptible(&resp->wq);
-}
-
 int __mcctrl_os_read_write_cpu_register(ihk_os_t os, int cpu,
 		struct ihk_os_cpu_register *desc,
 		enum mcctrl_os_cpu_operation op)
 {
 	struct mcctrl_usrdata *udp = ihk_host_os_get_usrdata(os);
 	struct ikc_scd_packet isp;
-	struct mcctrl_os_cpu_response resp;
+	struct ihk_os_cpu_register *ldesc = NULL;
+	int do_free = 0;
 	int ret = -EINVAL;
 
 	if (!udp) {
@@ -3511,50 +3529,43 @@ int __mcctrl_os_read_write_cpu_register(ihk_os_t os, int cpu,
 
 	}
 
+	/* Keep a dynamic structure around that can
+	 * survive an early return due to a signal */
+	ldesc = kmalloc(sizeof(*ldesc), GFP_KERNEL);
+	if (!ldesc) {
+		printk("%s: ERROR: allocating cpu register desc\n", __FUNCTION__);
+		return -ENOMEM;
+	}
+	*ldesc = *desc;
+
 	memset(&isp, '\0', sizeof(struct ikc_scd_packet));
 	isp.msg = SCD_MSG_CPU_RW_REG;
 	isp.op = op;
-	isp.desc = *desc;
-	isp.resp = &resp;
+	isp.pdesc = virt_to_phys(ldesc);
 
-	resp.done = 0;
-	resp.err = 0;
-	init_waitqueue_head(&resp.wq);
-
-	mb();
-	ret = mcctrl_ikc_send(os, cpu, &isp);
-	if (ret < 0) {
+	ret = mcctrl_ikc_send_wait(os, cpu, &isp, 0, NULL, &do_free, 1, ldesc);
+	if (ret != 0) {
 		printk("%s: ERROR sending IKC msg: %d\n", __FUNCTION__, ret);
 		goto out;
 	}
 
-	/* Wait for response */
-	ret = wait_event_interruptible(resp.wq, resp.done);
-	if (ret < 0) {
-		printk("%s: ERROR after wait: %d\n", __FUNCTION__, ret);
-		goto out;
-	}
-
-	ret = resp.err;
-	if (ret != 0) {
-		printk("%s: ERROR receive: %d\n", __FUNCTION__, resp.err);
-		goto out;
-	}
-
 	/* Update if read */
-	if (ret == 0 && op == MCCTRL_OS_CPU_READ_REGISTER) {
-		desc->val = resp.val;
+	if (op == MCCTRL_OS_CPU_READ_REGISTER) {
+		desc->val = ldesc->val;
 	}
 
 	/* Notify caller (for future async implementation) */
 	atomic_set(&desc->sync, 1);
 
-	dprintk("%s: MCCTRL_OS_CPU_%s_REGISTER: reg: 0x%lx, val: 0x%lx\n",
+	dprintk("%s: MCCTRL_OS_CPU_%s_REGISTER: CPU: %d, addr_ext: 0x%lx, val: 0x%lx\n",
 		__FUNCTION__,
-		(op == MCCTRL_OS_CPU_READ_REGISTER ? "READ" : "WRITE"),
-		desc->addr, desc->val);
+		(op == MCCTRL_OS_CPU_READ_REGISTER ? "READ" : "WRITE"), cpu,
+		desc->addr_ext, desc->val);
 
 out:
+	if (do_free) {
+		kfree(ldesc);
+	}
 	return ret;
 }
 

executer/kernel/mcctrl/ikc.c

@@ -209,6 +209,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 	case SCD_MSG_SEND_SIGNAL_ACK:
 	case SCD_MSG_PROCFS_ANSWER:
 	case SCD_MSG_REMOTE_PAGE_FAULT_ANSWER:
+	case SCD_MSG_CPU_RW_REG_RESP:
 		mcctrl_wakeup_cb(__os, pisp);
 		break;
 
@@ -239,10 +240,6 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		get_vdso_info(__os, pisp->arg);
 		break;
 
-	case SCD_MSG_CPU_RW_REG_RESP:
-		mcctrl_os_read_write_cpu_response(__os, pisp);
-		break;
-
 	case SCD_MSG_EVENTFD:
 		dkprintf("%s: SCD_MSG_EVENTFD,pisp->eventfd_type=%d\n", __FUNCTION__, pisp->eventfd_type);
 		mcctrl_eventfd(__os, pisp);
@@ -465,7 +462,7 @@ int prepare_ikc_channels(ihk_os_t os)
 	int i;
 	int ret = 0;
 
-	usrdata = kzalloc(sizeof(struct mcctrl_usrdata), GFP_KERNEL);
+	usrdata = kzalloc(sizeof(struct mcctrl_usrdata), GFP_ATOMIC);
 	if (!usrdata) {
 		printk("%s: error: allocating mcctrl_usrdata\n", __FUNCTION__);
 		ret = -ENOMEM;
@@ -491,7 +488,7 @@ int prepare_ikc_channels(ihk_os_t os)
 	usrdata->num_channels = usrdata->cpu_info->n_cpus;
 	usrdata->channels = kzalloc(sizeof(struct mcctrl_channel) *
 			usrdata->num_channels,
-			GFP_KERNEL);
+			GFP_ATOMIC);
 
 	if (!usrdata->channels) {
 		printk("Error: cannot allocate channels.\n");
@@ -500,7 +497,7 @@ int prepare_ikc_channels(ihk_os_t os)
 	}
 
 	usrdata->ikc2linux = kzalloc(sizeof(struct ihk_ikc_channel_desc *) *
-			nr_cpu_ids, GFP_KERNEL);
+			nr_cpu_ids, GFP_ATOMIC);
 
 	if (!usrdata->ikc2linux) {
 		printk("Error: cannot allocate ikc2linux channels.\n");
@@ -516,7 +513,6 @@ 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]);
@@ -525,8 +521,10 @@ 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);
 
@@ -582,18 +580,6 @@ 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

@@ -154,7 +154,7 @@ struct ikc_scd_packet {
 
 		/* SCD_MSG_CPU_RW_REG */
 		struct {
-			struct ihk_os_cpu_register desc;
+			unsigned long pdesc; /* Physical addr of the descriptor */
 			enum mcctrl_os_cpu_operation op;
 			void *resp;
 		};
@@ -298,7 +298,6 @@ struct mcctrl_cpu_topology {
 	//struct mcctrl_usrdata *udp;
 	struct ihk_cpu_topology *saved;
 	int mckernel_cpu_id;
-	int mckernel_core_id;
 	cpumask_t core_siblings;
 	cpumask_t thread_siblings;
 
@@ -325,20 +324,13 @@ 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))
@@ -361,7 +353,6 @@ 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];
@@ -377,7 +368,7 @@ struct mcctrl_usrdata {
 	nodemask_t numa_online;
 	struct list_head cpu_topology_list;
 	struct list_head node_topology_list;
-	struct list_head part_exec_list;
+	struct mcctrl_part_exec part_exec;
 	int perf_event_num;
 };
 

executer/kernel/mcctrl/procfs.c

@@ -1130,7 +1130,6 @@ static const struct procfs_entry base_entry_stuff[] = {
 //	PROC_REG("cpuinfo",    S_IRUGO, NULL),
 #endif /* POSTK_DEBUG_ARCH_DEP_42 */
 //	PROC_REG("meminfo",    S_IRUGO, NULL),
-	PROC_REG("meminfo",    S_IRUGO, &mckernel_buff_io),
 //	PROC_REG("pagetypeinfo",S_IRUGO, NULL),
 //	PROC_REG("softirq",    S_IRUGO, NULL),
 	PROC_REG("stat",       0444, &mckernel_buff_io),

executer/kernel/mcctrl/syscall.c

@@ -228,18 +228,19 @@ static int __notify_syscall_requester(ihk_os_t os, struct ikc_scd_packet *packet
 	c = (usrdata->channels + packet->ref)->c;
 
 	/* If spinning, no need for IKC message */
-	if (__sync_bool_compare_and_swap(&res->req_thread_status,
+	if (cmpxchg(&res->req_thread_status,
 				IHK_SCD_REQ_THREAD_SPINNING,
-				IHK_SCD_REQ_THREAD_TO_BE_WOKEN)) {
+				IHK_SCD_REQ_THREAD_TO_BE_WOKEN) ==
+			IHK_SCD_REQ_THREAD_SPINNING) {
 		dprintk("%s: no need to send IKC message for PID %d\n",
-			   __FUNCTION__, packet->pid);
+				__FUNCTION__, packet->pid);
 		return ret;
 	}
 
 	/* Wait until the status goes back to IHK_SCD_REQ_THREAD_SPINNING or
 	   IHK_SCD_REQ_THREAD_DESCHEDULED because two wake-up attempts are competing.
 	   Note that mcexec_terminate_thread() and returning EINTR would compete. */
-	if (res->req_thread_status == IHK_SCD_REQ_THREAD_TO_BE_WOKEN) {
+	if (smp_load_acquire(&res->req_thread_status) == IHK_SCD_REQ_THREAD_TO_BE_WOKEN) {
 		printk("%s: INFO: someone else is waking up the McKernel thread, "
 				"pid: %d, req status: %lu, syscall nr: %lu\n",
 				__FUNCTION__, packet->pid,
@@ -247,9 +248,10 @@ static int __notify_syscall_requester(ihk_os_t os, struct ikc_scd_packet *packet
 	}
 
 	/* The thread is not spinning any more, make sure it's descheduled */
-	if (!__sync_bool_compare_and_swap(&res->req_thread_status,
+	if (cmpxchg(&res->req_thread_status,
 				IHK_SCD_REQ_THREAD_DESCHEDULED,
-				IHK_SCD_REQ_THREAD_TO_BE_WOKEN)) {
+				IHK_SCD_REQ_THREAD_TO_BE_WOKEN) !=
+			IHK_SCD_REQ_THREAD_DESCHEDULED) {
 		printk("%s: WARNING: inconsistent requester status, "
 				"pid: %d, req status: %lu, syscall nr: %lu\n",
 				__FUNCTION__, packet->pid,
@@ -537,7 +539,11 @@ retry_alloc:
 #define	USE_VM_INSERT_PFN	1
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
+#if defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(8, 2)
+static vm_fault_t rus_vm_fault(struct vm_fault *vmf)
+#else
 static int rus_vm_fault(struct vm_fault *vmf)
+#endif
 {
 	struct vm_area_struct *vma = vmf->vma;
 #else
@@ -970,6 +976,30 @@ static int pager_req_create(ihk_os_t os, int fd, uintptr_t result_pa)
 		goto out;
 	}
 
+	/* Shared memory hack */
+	{
+		char *pathbuf, *fullpath;
+		pathbuf = kmalloc(PATH_MAX, GFP_ATOMIC);
+		if (pathbuf) {
+			fullpath = d_path(&file->f_path, pathbuf, PATH_MAX);
+			if (!IS_ERR(fullpath)) {
+				if (!strncmp("/tmp/ompi.", fullpath, 10) ||
+						!strncmp("/dev/shm/", fullpath, 9) ||
+						(!strncmp("/var/opt/FJSVtcs/ple/daemonif/",
+							fullpath, 30) && !strstr(fullpath, "dstore_sm.lock"))) {
+					printk("%s: treating %s as a device file..\n",
+						__func__, fullpath);
+					kfree(pathbuf);
+
+					error = -ESRCH;
+					goto out;
+				}
+
+				kfree(pathbuf);
+			}
+		}
+	}
+
 	inode = file->f_path.dentry->d_inode;
 	if (!inode) {
 		error = -EBADF;
@@ -1029,7 +1059,7 @@ static int pager_req_create(ihk_os_t os, int fd, uintptr_t result_pa)
 			pager = newpager;
 			newpager = NULL;
 
-			/* Intel MPI library and shared memory "prefetch" */
+			/* Shared libraries prefetch */
 			{
 				char *pathbuf, *fullpath;
 
@@ -1037,15 +1067,7 @@ static int pager_req_create(ihk_os_t os, int fd, uintptr_t result_pa)
 				if (pathbuf) {
 					fullpath = d_path(&file->f_path, pathbuf, PATH_MAX);
 					if (!IS_ERR(fullpath)) {
-						if (!strncmp("/dev/shm/Intel_MPI", fullpath, 18)) {
-							mf_flags = (MF_PREMAP | MF_ZEROFILL);
-							dprintk("%s: filename: %s, premap & zerofill\n",
-									__FUNCTION__, fullpath);
-						}
-						else if (strstr(fullpath, "libmpi") ||
-								strstr(fullpath, "libiomp") ||
-								strstr(fullpath, "libpthread") ||
-								strstr(fullpath, "libc.so")) {
+						if (strstr(fullpath, ".so")) {
 							mf_flags = MF_PREFETCH;
 							dprintk("%s: filename: %s, prefetch\n",
 									__FUNCTION__, fullpath);
@@ -1417,6 +1439,26 @@ static int pager_req_map(ihk_os_t os, int fd, size_t len, off_t off,
 #define	ANY_WHERE 0
 	if (prot_and_flags & MAP_LOCKED) prot_and_flags |= MAP_POPULATE;
 
+	/* Shared memory hack */
+	{
+		char *pathbuf, *fullpath;
+		pathbuf = kmalloc(PATH_MAX, GFP_ATOMIC);
+		if (pathbuf) {
+			fullpath = d_path(&file->f_path, pathbuf, PATH_MAX);
+			if (!IS_ERR(fullpath)) {
+				if (!strncmp("/tmp/ompi.", fullpath, 10) ||
+						!strncmp("/dev/shm/", fullpath, 9) ||
+						!strncmp("/var/opt/FJSVtcs/ple/daemonif/",
+							fullpath, 30)) {
+					dprintk("%s: pre-populating %s..\n",
+						__func__, fullpath);
+					prot_and_flags |= MAP_POPULATE;
+				}
+				kfree(pathbuf);
+			}
+		}
+	}
+
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,5,0)
 	down_write(&current->mm->mmap_sem);
 
@@ -1430,7 +1472,12 @@ static int pager_req_map(ihk_os_t os, int fd, size_t len, off_t off,
 #endif
 
 	if (IS_ERR_VALUE(va)) {
-		printk("pager_req_map(%p,%d,%lx,%lx,%lx):do_mmap_pgoff failed. %d\n", os, fd, len, off, result_rpa, (int)va);
+		if ((int)va != -ENOTSUPP) {
+			pr_err("%s(%p,%d,%lx,%lx,%lx): "
+			       "do_mmap_pgoff failed. %d\n",
+			       __func__, os, fd, len, off, result_rpa,
+			       (int)va);
+		}
 		error = va;
 		goto out;
 	}
@@ -1579,10 +1626,27 @@ retry:
 #else
 		fault = handle_mm_fault(current->mm, vma, va, flags);
 #endif
+#ifdef SC_DEBUG
 		if (fault != 0) {
-			printk("%s: error: faulting %lx at off: %lu\n", 
-					__FUNCTION__, va, off);
+			char *pathbuf = NULL;
+			char *fullpath;
+
+			if (vma->vm_file) {
+				pathbuf = kmalloc(PATH_MAX, GFP_ATOMIC);
+				if (pathbuf) {
+					fullpath = d_path(&vma->vm_file->f_path,
+							pathbuf, PATH_MAX);
+					if (!IS_ERR(fullpath)) {
+						printk("%s: WARNING: couldn't fault 0x%lx"
+								" at off: %lu in %s\n",
+								__FUNCTION__, va, off, fullpath);
+					}
+
+					kfree(pathbuf);
+				}
+			}
 		}
+#endif
 
 		page_fault_attempted = 1;
 		goto retry;

executer/kernel/mcctrl/sysfs_files.c

@@ -12,7 +12,6 @@
  */
 
 #include <linux/kernel.h>
-#include <linux/hashtable.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
 #include <linux/version.h>
@@ -24,14 +23,6 @@
 #define wprintk(...) do { if (1) printk(KERN_WARNING __VA_ARGS__); } while (0)
 #define eprintk(...) do { if (1) printk(KERN_ERR __VA_ARGS__); } while (0)
 
-struct physical_core_id {
-	int linux_core_id;
-	int mckernel_core_id;
-	struct hlist_node next;
-};
-
-DEFINE_HASHTABLE(physical_core_id_map, 10);
-
 static ssize_t
 show_int(struct sysfsm_ops *ops, void *instance, void *buf, size_t size)
 {
@@ -197,9 +188,6 @@ static void free_cpu_topology(struct mcctrl_usrdata *udp)
 void free_topology_info(ihk_os_t os)
 {
 	struct mcctrl_usrdata *udp = ihk_host_os_get_usrdata(os);
-	int bkt;
-	struct hlist_node *tmp;
-	struct physical_core_id *cur;
 
 	if (!udp) {
 		pr_warn("%s: warning: mcctrl_usrdata not found\n", __func__);
@@ -209,11 +197,6 @@ void free_topology_info(ihk_os_t os)
 	free_node_topology(udp);
 	free_cpu_topology(udp);
 
-	hash_for_each_safe(physical_core_id_map, bkt, tmp, cur, next) {
-		hash_del(&cur->next);
-		kfree(cur);
-	}
-
 	return;
 } /* free_topology_info() */
 
@@ -365,11 +348,6 @@ static struct mcctrl_cpu_topology *get_one_cpu_topology(struct mcctrl_usrdata *u
 	struct mcctrl_cpu_topology *topology = NULL;
 	struct cache_topology *cache;
 	struct ihk_cache_topology *saved_cache;
-	int linux_core_id;
-	int mckernel_core_id;
-	struct physical_core_id *entry;
-	struct physical_core_id *cur;
-	static int nr_mckernel_core;
 
 	dprintk("get_one_cpu_topology(%p,%d)\n", udp, index);
 	topology = kmalloc(sizeof(*topology), GFP_KERNEL);
@@ -413,22 +391,6 @@ static struct mcctrl_cpu_topology *get_one_cpu_topology(struct mcctrl_usrdata *u
 		goto out;
 	}
 
-	linux_core_id = topology->saved->core_id;
-	mckernel_core_id = -1;
-	hash_for_each_possible(physical_core_id_map, cur, next, linux_core_id) {
-		mckernel_core_id = cur->mckernel_core_id;
-		break;
-	}
-	if (mckernel_core_id < 0) {
-		mckernel_core_id = nr_mckernel_core++;
-		entry = kmalloc(sizeof(struct physical_core_id), GFP_KERNEL);
-		entry->linux_core_id = linux_core_id;
-		entry->mckernel_core_id = mckernel_core_id;
-		hash_add(physical_core_id_map,
-			 &entry->next, entry->linux_core_id);
-	}
-	topology->mckernel_core_id = mckernel_core_id;
-
 	list_for_each_entry(saved_cache,
 			&topology->saved->cache_topology_list, chain) {
 		cache = get_cache_topology(udp, topology, saved_cache);
@@ -550,7 +512,7 @@ static void setup_cpu_sysfs_files(struct mcctrl_usrdata *udp,
 			"%s/cpu%d/topology/physical_package_id",
 			prefix, cpu_number);
 	sysfsm_createf(udp->os, SYSFS_SNOOPING_OPS_d32,
-			&cpu->mckernel_core_id, 0444,
+			&cpu->saved->core_id, 0444,
 			"%s/cpu%d/topology/core_id",
 			prefix, cpu_number);
 

executer/user/CMakeLists.txt

@@ -8,7 +8,7 @@ include_directories(
 	"${PROJECT_BINARY_DIR}/ihk/linux/include"
 )
 
-
+add_subdirectory(lib)
 add_library(libmcexec STATIC arch/${ARCH}/archdep.S)
 SET_TARGET_PROPERTIES(libmcexec PROPERTIES OUTPUT_NAME mcexec)
 set_property(TARGET libmcexec PROPERTY POSITION_INDEPENDENT_CODE ON)
@@ -20,6 +20,18 @@ target_include_directories(mcexec PUBLIC "${KERNEL_DIR}")
 set_property(TARGET mcexec PROPERTY POSITION_INDEPENDENT_CODE ON)
 set_property(TARGET mcexec PROPERTY LINK_FLAGS "-fPIE -pie")
 
+add_executable(mcinspect mcinspect.c)
+if (NOT LIBDWARF)
+	target_include_directories(mcinspect PRIVATE "${CMAKE_CURRENT_BINARY_DIR}/lib/")
+	target_include_directories(mcinspect PRIVATE
+		"${CMAKE_CURRENT_SOURCE_DIR}/lib/libdwarf/libdwarf/libdwarf/")
+	target_link_libraries(mcinspect dwarf z elf)
+else()
+	target_include_directories(mcinspect PRIVATE ${DWARF_H})
+	target_link_libraries(mcinspect ${LIBDWARF})
+endif()
+target_link_libraries(mcinspect ${LIBBFD})
+
 add_executable(eclair eclair.c arch/${ARCH}/arch-eclair.c)
 target_link_libraries(eclair ${LIBBFD})
 
@@ -74,10 +86,13 @@ add_library(ldump2mcdump SHARED ldump2mcdump.c)
 configure_file(vmcore2mckdump.in vmcore2mckdump @ONLY)
 
 configure_file(mcexec.1in mcexec.1 @ONLY)
+configure_file(mcps.in mcps @ONLY)
 
-install(TARGETS mcexec eclair
+install(TARGETS mcexec eclair mcinspect
 	DESTINATION "${CMAKE_INSTALL_BINDIR}")
-install(PROGRAMS "${CMAKE_CURRENT_BINARY_DIR}/vmcore2mckdump"
+install(PROGRAMS
+	"${CMAKE_CURRENT_BINARY_DIR}/vmcore2mckdump"
+	"${CMAKE_CURRENT_BINARY_DIR}/mcps"
 	DESTINATION "${CMAKE_INSTALL_BINDIR}")
 install(TARGETS sched_yield ldump2mcdump
 	DESTINATION "${CMAKE_INSTALL_LIBDIR}")

executer/user/lib/CMakeLists.txt

@@ -0,0 +1,3 @@
+if (NOT LIBDWARF)
+	add_subdirectory(libdwarf)
+endif()

executer/user/lib/libdwarf/CMakeLists.txt

@@ -0,0 +1,144 @@
+if (NOT LIBDWARF)
+	# view folders on supported IDEs
+	set_property(GLOBAL PROPERTY USE_FOLDERS ON)
+
+	# used when finding libelf
+	set_property(GLOBAL PROPERTY FIND_LIBRARY_USE_LIB64_PATHS TRUE)
+
+	find_package(LibElf REQUIRED)
+	list(APPEND CMAKE_REQUIRED_INCLUDES ${LIBELF_INCLUDE_DIRS})
+	include(AutoconfHelper)
+
+	ac_init()
+	ac_c_bigendian()
+	ac_check_headers(sys/types.h sys/stat.h stdlib.h string.h memory.h strings.h inttypes.h stdint.h unistd.h)
+	ac_check_headers(alloca.h elf.h elfaccess.h libelf.h libelf/libelf.h  sys/types.h sys/elf_386.h sys/elf_amd64.h sys/elf_sparc.h sys/ia64/elf.h)
+
+	#  The default libdwarf is the one with struct Elf
+	message(STATUS "Assuming struct Elf for the default libdwarf.h")
+
+	# Find out where the elf header is.
+	if(HAVE_ELF_H)
+		set(HAVE_LOCATION_OF_LIBELFHEADER "<elf.h>")
+	elseif(HAVE_LIBELF_H)
+		set(HAVE_LOCATION_OF_LIBELFHEADER "<libelf.h>")
+	elseif(HAVE_LIBELF_LIBELF_H)
+		set(HAVE_LOCATION_OF_LIBELFHEADER "<libelf/libelf.h>")
+	endif()
+
+	ac_check_lib(${LIBELF_LIBRARIES} elf64_getehdr)
+	ac_check_lib(${LIBELF_LIBRARIES} elf64_getshdr)
+
+	ac_try_compile("
+	int main()
+	{
+		__uint32_t p;
+		p = 3;
+		return 0;
+	}"
+	HAVE___UINT32_T)
+
+	ac_try_compile("
+	int main()
+	{
+		__uint64_t p;
+		p = 3;
+		return 0;
+	}"
+	HAVE___UINT64_T)
+
+	ac_try_compile("
+	#include <sys/types.h>
+	int main()
+	{
+		__uint32_t p;
+		p = 3;
+		return 0;
+	}"
+	HAVE___UINT32_T_IN_SYS_TYPES_H)
+
+	ac_try_compile("
+	#include <sys/types.h>
+	int main()
+	{
+		__uint64_t p;
+		p = 3;
+		return 0;
+	}"
+	HAVE___UINT64_T_IN_SYS_TYPES_H)
+
+	check_c_source_runs("
+	static unsigned foo( unsigned x, __attribute__ ((unused)) int y)
+	{
+		unsigned x2 = x + 1;
+		return x2;
+	}
+
+	int main(void) {
+		unsigned y = 0;
+		y = foo(12,y);
+		return 0;
+	}"
+	HAVE_UNUSED_ATTRIBUTE)
+
+	message(STATUS "Checking compiler supports __attribute__ unused... ${HAVE_UNUSED_ATTRIBUTE}")
+
+	ac_try_compile("
+	#include <zlib.h>
+	int main()
+	{
+		Bytef dest[100];
+		uLongf destlen = 100;
+		Bytef *src = 0;
+		uLong srclen = 3;
+		int res = uncompress(dest,&destlen,src,srclen);
+		if (res == Z_OK) {
+			 /* ALL IS WELL */
+		}
+		return 0;
+	}"
+	HAVE_ZLIB)
+	message(STATUS "Checking zlib.h usability... ${HAVE_ZLIB}")
+	set(dwfzlib $<$<BOOL:${HAVE_ZIB}>:"z")
+
+	configure_file(libdwarf/libdwarf/libdwarf.h.in libdwarf.h COPYONLY)
+	configure_file(libdwarf/libdwarf/config.h.in.cmake config.h)
+	set(DWARF_CONFIGURATION_FILES ${CMAKE_CURRENT_BINARY_DIR}/config.h)
+
+	set(DWARF_SOURCES dwarf_abbrev.c dwarf_alloc.c dwarf_arange.c dwarf_die_deliv.c dwarf_dnames.c dwarf_dsc.c dwarf_elf_access.c dwarf_error.c
+	dwarf_form.c dwarf_frame.c dwarf_frame2.c dwarf_funcs.c dwarf_gdbindex.c dwarf_global.c dwarf_groups.c dwarf_harmless.c dwarf_init_finish.c dwarf_leb.c
+	dwarf_line.c dwarf_loc.c dwarf_macro.c dwarf_macro5.c dwarf_original_elf_init.c dwarf_pubtypes.c dwarf_query.c dwarf_ranges.c dwarf_string.c dwarf_tied.c
+	dwarf_str_offsets.c
+	dwarf_tsearchhash.c dwarf_types.c dwarf_util.c dwarf_vars.c dwarf_weaks.c dwarf_xu_index.c dwarf_print_lines.c malloc_check.c pro_alloc.c pro_arange.c
+	pro_die.c pro_encode_nm.c pro_error.c pro_expr.c pro_finish.c pro_forms.c pro_funcs.c pro_frame.c pro_init.c pro_line.c pro_reloc.c pro_reloc_stream.c
+	pro_reloc_symbolic.c pro_pubnames.c pro_section.c pro_types.c pro_vars.c pro_macinfo.c pro_weaks.c)
+
+	set(DWARF_HEADERS dwarf.h dwarf_abbrev.h dwarf_alloc.h dwarf_arange.h dwarf_base_types.h dwarf_die_deliv.h dwarf_dnames.h dwarf_dsc.h
+	dwarf_elf_access.h dwarf_error.h dwarf_frame.h dwarf_funcs.h dwarf_gdbindex.h dwarf_global.h dwarf_harmless.h dwarf_incl.h dwarf_line.h dwarf_loc.h
+	dwarf_macro.h dwarf_macro5.h dwarf_opaque.h dwarf_reloc_arm.h dwarf_reloc_mips.h dwarf_reloc_ppc.h dwarf_reloc_ppc64.h dwarf_reloc_x86_64.h dwarf_tsearch.h
+	dwarf_str_offsets.h
+	dwarf_types.h dwarf_util.h dwarf_vars.h dwarf_weaks.h dwarf_xu_index.h dwgetopt.h libdwarfdefs.h malloc_check.h pro_alloc.h pro_arange.h pro_die.h
+	pro_encode_nm.h pro_error.h pro_expr.h pro_frame.h pro_incl.h pro_line.h pro_macinfo.h pro_opaque.h pro_reloc.h pro_reloc_stream.h pro_reloc_symbolic.h
+	pro_section.h pro_types.h pro_util.h)
+
+	SET(__SRCS "")
+		FOREACH(f ${DWARF_SOURCES})
+			LIST(APPEND __SRCS "libdwarf/libdwarf/${f}")
+		ENDFOREACH(f)
+	SET(DWARF_SOURCES ${__SRCS})
+
+	set(GENNAMES_SOURCES
+		libdwarf/libdwarf/gennames.c
+		libdwarf/libdwarf/dwgetopt.c
+		libdwarf/libdwarf/dwarf.h)
+	add_executable(gennames ${GENNAMES_SOURCES})
+	set(GENNAMES_OUTPUT dwarf_names.c dwarf_names.h dwarf_names_enum.h dwarf_names_new.h)
+
+	add_custom_command(OUTPUT ${GENNAMES_OUTPUT}
+		COMMAND gennames -s -i ${CMAKE_CURRENT_SOURCE_DIR}/libdwarf/libdwarf/ -o .
+		DEPENDS gennames libdwarf/libdwarf/libdwarf.h.in)
+
+	add_library(dwarf STATIC ${DWARF_SOURCES} ${GENNAMES_OUTPUT} ${DWARF_CONFIGURATION_FILES})
+	target_include_directories(dwarf PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/libdwarf/libdwarf/")
+	target_include_directories(dwarf BEFORE PRIVATE "${CMAKE_CURRENT_BINARY_DIR}")
+endif()

executer/user/libsched_yield.c

@@ -1,6 +1,7 @@
 #define _GNU_SOURCE
 #include <dlfcn.h>
 #include <sys/time.h>
+#include <sys/syscall.h>
 #include <sched.h>
 #include <unistd.h>
 #include <stdlib.h>
@@ -27,3 +28,110 @@ int sched_yield(void)
 
 	return 0;
 }
+
+#undef pthread_create
+
+typedef int (*__pthread_create_fn)(pthread_t *thread,
+		const pthread_attr_t *attr,
+		void *(*start_routine) (void *),
+		void *arg);
+
+static __pthread_create_fn orig_pthread_create = 0;
+
+
+int pthread_create(pthread_t *thread,
+		const pthread_attr_t *attr,
+		void *(*start_routine) (void *),
+		void *arg)
+{
+	if (!orig_pthread_create) {
+		orig_pthread_create =
+			(__pthread_create_fn)dlsym(RTLD_NEXT, "pthread_create");
+	}
+
+	/* CLONE_VM and newsp == parent_tidptr impiles pthread start routine addr */
+	syscall(__NR_clone, CLONE_VM, start_routine, start_routine, 0, 0, 0);
+
+	return orig_pthread_create(thread, attr, start_routine, arg);
+}
+
+#if 0
+#define PROCMAPS_MAX_LEN	131072
+
+char *addr_to_lib(void *addr, unsigned long *offset_in_lib)
+{
+	char maps_path[PATH_MAX];
+	char buf[PROCMAPS_MAX_LEN];
+	int fd;
+	void *start, *end;
+	char perms[4];
+	unsigned long offset;
+	unsigned long dev[2];
+	int inode;
+	char path[PATH_MAX];
+	char *line;
+
+	sprintf(maps_path,"/proc/self/maps");
+	fd = open(maps_path, O_RDONLY);
+	if (fd < 0) {
+		fprintf(stderr,"error: cannot open the memory maps, %s\n",
+				strerror(errno));
+		return NULL;
+	}
+
+	memset(buf, 0, PROCMAPS_MAX_LEN);
+	read(fd, buf, PROCMAPS_MAX_LEN);
+	line = strtok(buf, "\n");
+	while (line) {
+		memset(path, 0, sizeof(path));
+		sscanf(line, "%012lx-%012lx %4s %lx %lx:%lx %d\t\t\t%[^\n]",
+			&start, &end, perms, &offset, &dev[0], &dev[1], &inode, path);
+
+		if (start <= addr && end > addr) {
+			close(fd);
+			if (offset_in_lib)
+				*offset_in_lib = (unsigned long)(addr - start);
+			return strlen(path) > 0 ? strdup(path) : NULL;
+		}
+
+		line = strtok(NULL, "\n");
+	}
+
+	close(fd);
+	return NULL;
+}
+
+int pthread_create(pthread_t *thread,
+		const pthread_attr_t *attr,
+		void *(*start_routine) (void *),
+		void *arg)
+{
+	char *lib = NULL;
+	int util_thread = 1;
+	unsigned long offset;
+
+	if (!orig_pthread_create) {
+		orig_pthread_create =
+			(__pthread_create_fn)dlsym(RTLD_NEXT, "pthread_create");
+	}
+
+	lib = addr_to_lib(start_routine, &offset);
+	if (lib)
+		printf("%s: 0x%lx is in %s @ 0x%lx\n",
+				__func__, start_routine, lib, offset);
+
+	if (lib && (strstr(lib, "iomp") || strstr(lib, "psm"))) {
+		util_thread = 0;
+	}
+
+	if (util_thread) {
+		/* McKernel util_indicate_clone() */
+		syscall(731);
+	}
+
+	if (lib)
+		free(lib);
+
+	return orig_pthread_create(thread, attr, start_routine, arg);
+}
+#endif

executer/user/mcexec.c

@@ -1322,7 +1322,8 @@ void init_sigaction(void)
 
 	master_tid = gettid();
 	for (i = 1; i <= 64; i++) {
-		if (i != SIGKILL && i != SIGSTOP && i != SIGCHLD) {
+		if (i != SIGKILL && i != SIGSTOP && i != SIGCHLD &&
+				i != SIGTSTP && i != SIGTTIN && i != SIGTTOU) {
 			struct sigaction act;
 
 			sigaction(i, NULL, &act);
@@ -2211,6 +2212,64 @@ 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
@@ -2483,10 +2542,11 @@ 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;
@@ -2494,6 +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);
@@ -2502,6 +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) {

executer/user/mcps.in

@@ -0,0 +1,10 @@
+#!/bin/bash
+# IHK/McKernel mcps script.
+# author: Balazs Gerofi <bgerofi@riken.jp>
+#      Copyright (C) 2019 RIKEN
+#
+prefix="@prefix@"
+BINDIR="${prefix}/bin"
+KERNDIR="@MCKERNELDIR@"
+
+${BINDIR}/mcinspect --kernel=${KERNDIR}/mckernel.img --ps

kernel/devobj.c

@@ -170,7 +170,7 @@ static void devobj_free(struct memobj *memobj)
 
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
 	if (error) {
-		kprintf("%s(%p %lx): release failed. %d\n",
+		dkprintf("%s(%p %lx): release failed. %d\n",
 			__func__, obj, handle, error);
 		/* through */
 	}

kernel/fileobj.c

@@ -398,8 +398,9 @@ static void fileobj_free(struct memobj *memobj)
 		}
 	}
 
-	/* Pre-mapped? */
-	if (to_memobj(obj)->flags & MF_PREMAP) {
+	/* Pre-mapped zerofilled? */
+	if (to_memobj(obj)->flags & MF_PREMAP &&
+			to_memobj(obj)->flags & MF_ZEROFILL) {
 		int i;
 
 		for (i = 0; i < to_memobj(obj)->nr_pages; ++i) {
@@ -440,7 +441,7 @@ static void fileobj_free(struct memobj *memobj)
 
 	error = syscall_generic_forwarding(__NR_mmap, &ctx);
 	if (error) {
-		kprintf("%s(%p %lx): free failed. %d\n", __func__,
+		dkprintf("%s(%p %lx): free failed. %d\n", __func__,
 			obj, obj->handle, error);
 		/* through */
 	}
@@ -473,20 +474,24 @@ static void fileobj_do_pageio(void *args0)
 	ssize_t ss;
 	struct mcs_lock_node mcs_node;
 	int hash = (off >> PAGE_SHIFT) & FILEOBJ_PAGE_HASH_MASK;	
+	int attempts = 0;
 
-	mcs_lock_lock_noirq(&obj->page_hash_locks[hash],
-			&mcs_node);
+	mcs_lock_lock(&obj->page_hash_locks[hash], &mcs_node);
 	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page) {
 		goto out;
 	}
 
 	while (page->mode == PM_PAGEIO) {
-		mcs_lock_unlock_noirq(&obj->page_hash_locks[hash],
-				&mcs_node);
+		mcs_lock_unlock(&obj->page_hash_locks[hash], &mcs_node);
+		++attempts;
+		if (attempts > 49) {
+			dkprintf("%s: %s:%lu PM_PAGEIO loop %d -> schedule()\n",
+				__func__, to_memobj(obj)->path, off, attempts);
+			schedule();
+		}
 		cpu_pause();
-		mcs_lock_lock_noirq(&obj->page_hash_locks[hash],
-				&mcs_node);
+		mcs_lock_lock(&obj->page_hash_locks[hash], &mcs_node);
 	}
 
 	if (page->mode == PM_WILL_PAGEIO) {
@@ -499,8 +504,7 @@ static void fileobj_do_pageio(void *args0)
 		}
 		else {
 			page->mode = PM_PAGEIO;
-			mcs_lock_unlock_noirq(&obj->page_hash_locks[hash],
-					&mcs_node);
+			mcs_lock_unlock(&obj->page_hash_locks[hash], &mcs_node);
 
 			ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_READ;
 			ihk_mc_syscall_arg1(&ctx) = obj->handle;
@@ -512,8 +516,7 @@ static void fileobj_do_pageio(void *args0)
 					__FUNCTION__, obj->handle);
 			ss = syscall_generic_forwarding(__NR_mmap, &ctx);
 
-			mcs_lock_lock_noirq(&obj->page_hash_locks[hash],
-					&mcs_node);
+			mcs_lock_lock(&obj->page_hash_locks[hash], &mcs_node);
 			if (page->mode != PM_PAGEIO) {
 				kprintf("fileobj_do_pageio(%p,%lx,%lx):"
 						"invalid mode %x\n",
@@ -539,8 +542,7 @@ static void fileobj_do_pageio(void *args0)
 		page->mode = PM_DONE_PAGEIO;
 	}
 out:
-	mcs_lock_unlock_noirq(&obj->page_hash_locks[hash],
-			&mcs_node);
+	mcs_lock_unlock(&obj->page_hash_locks[hash], &mcs_node);
 	memobj_unref(&obj->memobj);		/* got fileobj_get_page() */
 	kfree(args0);
 	dkprintf("fileobj_do_pageio(%p,%lx,%lx):\n", obj, off, pgsize);
@@ -570,7 +572,8 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 	profile_event_add(PROFILE_page_fault_file, PAGE_SIZE);
 #endif // PROFILE_ENABLE
 
-	if (memobj->flags & MF_PREMAP) {
+	if (memobj->flags & MF_PREMAP &&
+			memobj->flags & MF_ZEROFILL) {
 		int page_ind = off >> PAGE_SHIFT;
 
 		if (!memobj->pages[page_ind]) {
@@ -587,8 +590,7 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 
 			/* Update the array but see if someone did it already and use
 			 * that if so */
-			if (!__sync_bool_compare_and_swap(&memobj->pages[page_ind],
-						NULL, virt)) {
+			if (cmpxchg(&memobj->pages[page_ind], NULL, virt) != NULL) {
 				ihk_mc_free_pages_user(virt, 1);
 			}
 			else {
@@ -609,8 +611,7 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 		goto out_nolock;
 	}
 
-	mcs_lock_lock_noirq(&obj->page_hash_locks[hash],
-			&mcs_node);
+	mcs_lock_lock(&obj->page_hash_locks[hash], &mcs_node);
 	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page || (page->mode == PM_WILL_PAGEIO)
 			|| (page->mode == PM_PAGEIO)) {
@@ -688,8 +689,7 @@ static int fileobj_get_page(struct memobj *memobj, off_t off,
 	*physp = page_to_phys(page);
 	virt = NULL;
 out:
-	mcs_lock_unlock_noirq(&obj->page_hash_locks[hash],
-			&mcs_node);
+	mcs_lock_unlock(&obj->page_hash_locks[hash], &mcs_node);
 out_nolock:
 	if (virt) {
 		ihk_mc_free_pages_user(virt, npages);
@@ -767,8 +767,7 @@ static int fileobj_lookup_page(struct memobj *memobj, off_t off,
 		return -ENOMEM;
 	}
 
-	mcs_lock_lock_noirq(&obj->page_hash_locks[hash],
-			&mcs_node);
+	mcs_lock_lock(&obj->page_hash_locks[hash], &mcs_node);
 
 	page = __fileobj_page_hash_lookup(obj, hash, off);
 	if (!page) {
@@ -779,8 +778,7 @@ static int fileobj_lookup_page(struct memobj *memobj, off_t off,
 	error = 0;
 
 out:
-	mcs_lock_unlock_noirq(&obj->page_hash_locks[hash],
-			&mcs_node);
+	mcs_lock_unlock(&obj->page_hash_locks[hash], &mcs_node);
 
 	dkprintf("fileobj_lookup_page(%p,%lx,%x,%p): %d \n",
 			obj, off, p2align, physp, error);

kernel/futex.c

@@ -311,6 +311,7 @@ static int futex_wake(uint32_t *uaddr, int fshared, int nr_wake, uint32_t bitset
 	struct plist_head *head;
 	union futex_key key = FUTEX_KEY_INIT;
 	int ret;
+	unsigned long irqstate;
 
 	if (!bitset)
 		return -EINVAL;
@@ -320,7 +321,7 @@ static int futex_wake(uint32_t *uaddr, int fshared, int nr_wake, uint32_t bitset
 		goto out;
 
 	hb = hash_futex(&key);
-	ihk_mc_spinlock_lock_noirq(&hb->lock);
+	irqstate = ihk_mc_spinlock_lock(&hb->lock);
 	head = &hb->chain;
 
 	plist_for_each_entry_safe(this, next, head, list) {
@@ -337,7 +338,7 @@ static int futex_wake(uint32_t *uaddr, int fshared, int nr_wake, uint32_t bitset
 		}
 	}
 
-	ihk_mc_spinlock_unlock_noirq(&hb->lock);
+	ihk_mc_spinlock_unlock(&hb->lock, irqstate);
 	put_futex_key(fshared, &key);
 out:
 	return ret;

kernel/host.c

@@ -142,7 +142,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 		if (add_process_memory_range(vm, s, e, NOPHYS, flags, NULL, 0,
 					pn->sections[i].len > LARGE_PAGE_SIZE ?
 					LARGE_PAGE_SHIFT : PAGE_SHIFT,
-					NULL, &range) != 0) {
+					&range) != 0) {
 			kprintf("ERROR: adding memory range for ELF section %i\n", i);
 			goto err;
 		}
@@ -284,7 +284,7 @@ int prepare_process_ranges_args_envs(struct thread *thread,
 	dkprintf("%s: args_envs: %d pages\n",
 			__func__, argenv_page_count);
 	if(add_process_memory_range(vm, addr, e, args_envs_p,
-				flags, NULL, 0, PAGE_SHIFT, NULL, NULL) != 0){
+				flags, NULL, 0, PAGE_SHIFT, NULL) != 0){
 		ihk_mc_free_pages_user(args_envs, argenv_page_count);
 		kprintf("ERROR: adding memory range for args/envs\n");
 		goto err;
@@ -500,6 +500,7 @@ static int process_msg_prepare_process(unsigned long rphys)
 		ihk_mc_unmap_memory(NULL, phys, sz);
 		return -ENOMEM;
 	}
+	sprintf(thread->pthread_routine, "%s", "[main]");
 	proc = thread->proc;
 	vm = thread->vm;
 
@@ -619,6 +620,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 {
 	struct ikc_scd_packet *packet = __packet;
 	struct ikc_scd_packet pckt;
+	struct ihk_os_cpu_register *cpu_desc;
 	struct ihk_ikc_channel_desc *resp_channel = cpu_local_var(ikc2linux);
 	int rc;
 	struct thread *thread;
@@ -689,11 +691,11 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 			ret = -EINVAL;
 			break;
 		}
-		thread_unlock(thread);
 
 		dkprintf("%s: SCD_MSG_WAKE_UP_SYSCALL_THREAD: waking up tid %d\n",
 			__FUNCTION__, packet->ttid);
 		waitq_wakeup(&thread->scd_wq);
+		thread_unlock(thread);
 		ret = 0;
 		break;
 
@@ -717,9 +719,11 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		dkprintf("remote page fault,pid=%d,va=%lx,reason=%x\n",
 			 thread->proc->pid, packet->fault_address,
 			 packet->fault_reason|PF_POPULATE);
+		preempt_disable();
 		pckt.err = page_fault_process_vm(thread->vm,
 					(void *)packet->fault_address,
 					packet->fault_reason|PF_POPULATE);
+		preempt_enable();
 
 #ifdef PROFILE_ENABLE
 		if (thread->profile) {
@@ -839,12 +843,17 @@ out_remote_pf:
 		break;
 
 	case SCD_MSG_CPU_RW_REG:
+		pp = ihk_mc_map_memory(NULL, packet->pdesc,
+				sizeof(struct ihk_os_cpu_register));
+		cpu_desc = (struct ihk_os_cpu_register *)ihk_mc_map_virtual(
+				pp, 1, PTATTR_WRITABLE | PTATTR_ACTIVE);
 
 		pckt.msg = SCD_MSG_CPU_RW_REG_RESP;
-		memcpy(&pckt.desc, &packet->desc,
-				sizeof(struct ihk_os_cpu_register));
-		pckt.resp = packet->resp;
-		pckt.err = arch_cpu_read_write_register(&pckt.desc, packet->op);
+		pckt.reply = packet->reply;
+		pckt.err = arch_cpu_read_write_register(cpu_desc, packet->op);
+
+		ihk_mc_unmap_virtual(cpu_desc, 1);
+		ihk_mc_unmap_memory(NULL, pp, sizeof(struct ihk_os_cpu_register));
 
 		ihk_ikc_send(resp_channel, &pckt, 0);
 		break;
@@ -895,7 +904,7 @@ void init_host_ikc2linux(int linux_cpu)
 		param.port = 503;
 		param.intr_cpu = linux_cpu;
 		param.pkt_size = sizeof(struct ikc_scd_packet);
-		param.queue_size = 2 * num_processors * sizeof(struct ikc_scd_packet);
+		param.queue_size = 4 * num_processors * sizeof(struct ikc_scd_packet);
 		if (param.queue_size < PAGE_SIZE * 4) {
 			param.queue_size = PAGE_SIZE * 4;
 		}

kernel/hugefileobj.c

@@ -139,7 +139,6 @@ static void hugefileobj_free(struct memobj *memobj)
 struct memobj_ops hugefileobj_ops = {
 	.free = hugefileobj_free,
 	.get_page = hugefileobj_get_page,
-
 };
 
 void hugefileobj_cleanup(void)

kernel/include/cls.h

@@ -100,6 +100,7 @@ struct cpu_local_var {
 	int in_interrupt;
 	int no_preempt;
 	int timer_enabled;
+	unsigned long nr_ctx_switches;
 	int kmalloc_initialized;
 	struct ihk_os_cpu_monitor *monitor;
 	struct rusage_percpu *rusage;

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/process.h

@@ -55,7 +55,6 @@
 #define	VR_MEMTYPE_MASK    0x0f000000
 #define VR_PAGEOUT	   0x10000000
 #define VR_DONTDUMP	   0x20000000
-#define VR_XPMEM	   0x40000000
 #define VR_WIPEONFORK	   0x80000000
 
 #define	PROT_TO_VR_FLAG(prot)	(((unsigned long)(prot) << 16) & VR_PROT_MASK)
@@ -506,10 +505,12 @@ struct process {
 			// PS_STOPPED -----+
 			// (PS_TRACED)
 
-	/* Store exit_status for a group of threads when stopped by SIGSTOP.
-	   exit_status can't be used because values of exit_status of threads
-	   might divert while the threads are exiting by group_exit(). */
-	int group_exit_status;
+	/* Store exit_status for a group of threads when stopped by SIGSTOP.  */
+	/* exit_status can't be used because values of exit_status of threads */
+	/* might divert while the threads are exiting by group_exit().	      */
+	/* The upper 4 bytes of group_exit_status is the confirmation flag of */
+	/* exit status. The lower 4 bytes is the exit status.		      */
+	unsigned long group_exit_status;
 
 	/* Manage ptraced processes in the separate list to make it easy to
 	   restore the orginal parent child relationship when 
@@ -609,6 +610,7 @@ struct thread {
 	// thread info
 	int cpu_id;
 	int tid;
+	char pthread_routine[PATH_MAX + 64];
 	int status;	// PS_RUNNING -> PS_EXITED (-> ZOMBIE / ptrace)
 			// |       ^       ^
 			// |       |       |
@@ -718,6 +720,7 @@ struct thread {
 	/* Syscall offload wait queue head */
 	struct waitq scd_wq;
 
+	unsigned long clone_pthread_start_routine;
 	int uti_state;
 	int mod_clone;
 	struct uti_attr *mod_clone_arg;
@@ -749,7 +752,7 @@ struct process_vm {
 	void *vvar_addr;
  	
 	ihk_spinlock_t page_table_lock;
-	ihk_spinlock_t memory_range_lock;
+	ihk_rwspinlock_t memory_range_lock;
     // to protect the followings:
     // 1. addition of process "memory range" (extend_process_region, add_process_memory_range)
     // 2. addition of process page table (allocate_pages, update_process_page_table)
@@ -805,7 +808,7 @@ int add_process_memory_range(struct process_vm *vm,
 		unsigned long start, unsigned long end,
 		unsigned long phys, unsigned long flag,
 		struct memobj *memobj, off_t offset,
-		int pgshift, void *private_data, struct vm_range **rp);
+		int pgshift, struct vm_range **rp);
 int remove_process_memory_range(struct process_vm *vm, unsigned long start,
 		unsigned long end, int *ro_freedp);
 int split_process_memory_range(struct process_vm *vm,

kernel/include/profile.h

@@ -5,10 +5,9 @@
 #define PROFILE_ENABLE
 
 #ifdef PROFILE_ENABLE
-#define PROFILE_SYSCALL_MAX                          300
+#define PROFILE_SYSCALL_MAX                          2000
 #define PROFILE_OFFLOAD_MAX   (PROFILE_SYSCALL_MAX << 1)
 #define PROFILE_EVENT_MIN            PROFILE_OFFLOAD_MAX
-#define __NR_profile                                 701
 
 #define PROF_JOB                       0x40000000
 #define PROF_PROC                      0x80000000
@@ -47,6 +46,8 @@ enum profile_event_type {
 	PROFILE_EVENT_MAX	/* Should be the last event type */
 };
 
+#define __NR_profile                   PROFILE_EVENT_MAX
+
 struct thread;
 struct process;
 

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
 
@@ -27,30 +28,17 @@ static inline int rusage_pgsize_to_pgtype(size_t pgsize)
 	case 12:
 		ret = IHK_OS_PGSIZE_4KB;
 		break;
-	case 16:
-		ret = IHK_OS_PGSIZE_64KB;
-		break;
 	case 21:
 		ret = IHK_OS_PGSIZE_2MB;
 		break;
-	case 25:
-		ret = IHK_OS_PGSIZE_32MB;
-		break;
 	case 30:
 		ret = IHK_OS_PGSIZE_1GB;
 		break;
-	case 34:
-		ret = IHK_OS_PGSIZE_16GB;
-		break;
-	case 29:
-		ret = IHK_OS_PGSIZE_512MB;
-		break;
-	case 42:
-		ret = IHK_OS_PGSIZE_4TB;
-		break;
 	default:
+#if 0 /* 64KB page goes here when using mckernel_rusage-compatible ihk_os_rusage */
 		kprintf("%s: Error: Unknown pgsize=%ld\n",
 			__func__, pgsize);
+#endif
 		break;
 	}
 
@@ -324,6 +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;

kernel/include/syscall.h

@@ -287,7 +287,7 @@ struct ikc_scd_packet {
 
 		/* SCD_MSG_CPU_RW_REG */
 		struct {
-			struct ihk_os_cpu_register desc;
+			unsigned long pdesc; /* Physical addr of the descriptor */
 			enum mcctrl_os_cpu_operation op;
 			void *resp;
 		};
@@ -508,7 +508,7 @@ enum set_cputime_mode {
 void set_cputime(enum set_cputime_mode mode);
 int do_munmap(void *addr, size_t len, int holding_memory_range_lock);
 intptr_t do_mmap(uintptr_t addr0, size_t len0, int prot, int flags, int fd,
-		off_t off0, const int vrf0, void *private_data);
+		off_t off0);
 void clear_host_pte(uintptr_t addr, size_t len, int holding_memory_range_lock);
 typedef int32_t key_t;
 int do_shmget(key_t key, size_t size, int shmflg);
@@ -651,4 +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/include/xpmem.h

@@ -27,20 +27,6 @@ int xpmem_remove_process_memory_range(struct process_vm *vm,
 	struct vm_range *vmr);
 int xpmem_fault_process_memory_range(struct process_vm *vm,
 	struct vm_range *vmr, unsigned long vaddr, uint64_t reason);
-int xpmem_update_process_page_table(struct process_vm *vm,
-	struct vm_range *vmr);
 
-struct xpmem_attachment {
-	mcs_rwlock_lock_t at_lock;	/* att lock */
-	unsigned long vaddr;	/* starting address of seg attached */
-	unsigned long at_vaddr;	/* address where seg is attached */
-	size_t at_size;		/* size of seg attachment */
-	struct vm_range *at_vmr;	/* vm_range where seg is attachment */
-	int flags;	/* att attributes and state */
-	ihk_atomic_t refcnt;	/* references to att */
-	struct xpmem_access_permit *ap;	/* associated access permit */
-	struct list_head att_list;	/* atts linked to access permit */
-	struct process_vm *vm;	/* process_vm attached to */
-};
 #endif /* _XPMEM_H */
 

kernel/include/xpmem_private.h

@@ -177,6 +177,19 @@ struct xpmem_access_permit {
 	struct list_head ap_hashlist;	/* access permit hash list */
 };
 
+struct xpmem_attachment {
+	mcs_rwlock_lock_t at_lock;	/* att lock */
+	unsigned long vaddr;	/* starting address of seg attached */
+	unsigned long at_vaddr;	/* address where seg is attached */
+	size_t at_size;		/* size of seg attachment */
+	struct vm_range *at_vmr;	/* vm_range where seg is attachment */
+	volatile int flags;	/* att attributes and state */
+	ihk_atomic_t refcnt;	/* references to att */
+	struct xpmem_access_permit *ap;	/* associated access permit */
+	struct list_head att_list;	/* atts linked to access permit */
+	struct process_vm *vm;	/* process_vm attached to */
+};
+
 struct xpmem_partition {
 	ihk_atomic_t n_opened;	/* # of /dev/xpmem opened */
 	struct xpmem_hashlist tg_hashtable[];	/* locks + tg hash lists */
@@ -318,7 +331,6 @@ static void xpmem_ap_deref(struct xpmem_access_permit *ap);
 static void xpmem_att_deref(struct xpmem_attachment *att);
 static int xpmem_validate_access(struct xpmem_access_permit *, off_t, size_t,
 	int, unsigned long *);
-static int is_remote_vm(struct process_vm *vm);
 
 /*
  * Inlines that mark an internal driver structure as being destroyable or not.

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;
 
@@ -1276,8 +1277,9 @@ static void unhandled_page_fault(struct thread *thread, void *fault_addr,
 
 	range = lookup_process_memory_range(vm, address, address+1);
 	if (range) {
-		__kprintf("address is in range, flag: 0x%lx\n",
-				range->flag);
+		__kprintf("address is in range, flag: 0x%lx (%s)\n",
+				range->flag,
+				range->memobj ? range->memobj->path : "");
 		ihk_mc_pt_print_pte(vm->address_space->page_table,
 				    (void *)address);
 	} else {
@@ -1600,9 +1602,11 @@ static void numa_distances_init()
 static ssize_t numa_sysfs_show_meminfo(struct sysfs_ops *ops,
 		void *instance, void *buf, size_t size)
 {
+#ifdef IHK_RBTREE_ALLOCATOR
 	struct ihk_mc_numa_node *node =
 		(struct ihk_mc_numa_node *)instance;
 	char *sbuf = (char *)buf;
+#endif
 	int len = 0;
 
 #ifdef IHK_RBTREE_ALLOCATOR
@@ -2559,7 +2563,7 @@ void ihk_mc_query_mem_user_page(void *dump_pase_info) {
 }
 
 void ihk_mc_query_mem_free_page(void *dump_pase_info) {
-
+#ifdef IHK_RBTREE_ALLOCATOR
 	struct free_chunk *chunk;
 	struct rb_node *node;
 	struct rb_root *free_chunks;
@@ -2629,7 +2633,7 @@ void ihk_mc_query_mem_free_page(void *dump_pase_info) {
 			}
 		}
 	}
-
+#endif
 	return;
 }
 

kernel/process.c

@@ -91,7 +91,7 @@ ihk_spinlock_t runq_reservation_lock;
 
 int idle_halt = 0;
 int allow_oversubscribe = 0;
-int time_sharing = 0;
+int time_sharing = 1;
 
 void
 init_process(struct process *proc, struct process *parent)
@@ -243,7 +243,7 @@ static int
 init_process_vm(struct process *owner, struct address_space *asp, struct process_vm *vm)
 {
 	int i;
-	ihk_mc_spinlock_init(&vm->memory_range_lock);
+	ihk_rwspinlock_init(&vm->memory_range_lock);
 	ihk_mc_spinlock_init(&vm->page_table_lock);
 
 	ihk_atomic_set(&vm->refcount, 1);
@@ -792,7 +792,7 @@ static int copy_user_ranges(struct process_vm *vm, struct process_vm *orgvm)
 	struct vm_range *last_insert;
 	struct copy_args args;
 
-	ihk_mc_spinlock_lock_noirq(&orgvm->memory_range_lock);
+	ihk_rwspinlock_read_lock_noirq(&orgvm->memory_range_lock);
 
 	/* Iterate original process' vm_range list and take a copy one-by-one */
 	last_insert = NULL;
@@ -854,7 +854,7 @@ static int copy_user_ranges(struct process_vm *vm, struct process_vm *orgvm)
 		// memory_stat_rss_add() is called in child-node, i.e. copy_user_pte()
 	}
 
-	ihk_mc_spinlock_unlock_noirq(&orgvm->memory_range_lock);
+	ihk_rwspinlock_read_unlock_noirq(&orgvm->memory_range_lock);
 
 	return 0;
 
@@ -881,7 +881,7 @@ err_rollback:
 		}
 	}
 
-	ihk_mc_spinlock_unlock_noirq(&orgvm->memory_range_lock);
+	ihk_rwspinlock_read_unlock_noirq(&orgvm->memory_range_lock);
 	return -1;
 }
 
@@ -1316,7 +1316,7 @@ int add_process_memory_range(struct process_vm *vm,
 		unsigned long start, unsigned long end,
 		unsigned long phys, unsigned long flag,
 		struct memobj *memobj, off_t offset,
-		int pgshift, void *private_data, struct vm_range **rp)
+		int pgshift, struct vm_range **rp)
 {
 	dkprintf("%s: start=%lx,end=%lx,phys=%lx,flag=%lx\n", __FUNCTION__, start, end, phys, flag);
 	struct vm_range *range;
@@ -1344,7 +1344,7 @@ int add_process_memory_range(struct process_vm *vm,
 	range->memobj = memobj;
 	range->objoff = offset;
 	range->pgshift = pgshift;
-	range->private_data = private_data;
+	range->private_data = NULL;
 
 	rc = 0;
 	if (phys == NOPHYS) {
@@ -1356,10 +1356,6 @@ int add_process_memory_range(struct process_vm *vm,
 	else if (flag & VR_IO_NOCACHE) {
 		rc = update_process_page_table(vm, range, phys, PTATTR_UNCACHABLE);
 	}
-	else if (flag & VR_XPMEM) {
-		range->memobj->flags |= MF_XPMEM;
-		rc = xpmem_update_process_page_table(vm, range);
-	}
 	else if (flag & VR_DEMAND_PAGING) {
 		dkprintf("%s: range: 0x%lx - 0x%lx is demand paging\n",
 				__FUNCTION__, range->start, range->end);
@@ -1387,8 +1383,7 @@ int add_process_memory_range(struct process_vm *vm,
 	}
 
 	/* Clear content! */
-	if (phys != NOPHYS
-			&& !(flag & (VR_REMOTE | VR_DEMAND_PAGING | VR_XPMEM))
+	if (phys != NOPHYS && !(flag & (VR_REMOTE | VR_DEMAND_PAGING))
 			&& ((flag & VR_PROT_MASK) != VR_PROT_NONE)) {
 #if 1
 		memset((void *)phys_to_virt(phys), 0, end - start);
@@ -2181,16 +2176,65 @@ static int do_page_fault_process_vm(struct process_vm *vm, void *fault_addr0, ui
 {
 	int error;
 	const uintptr_t fault_addr = (uintptr_t)fault_addr0;
-	struct vm_range *range;
+	struct vm_range *range = NULL;
 	struct thread *thread = cpu_local_var(current);
 	int locked = 0;
 
 	dkprintf("[%d]do_page_fault_process_vm(%p,%lx,%lx)\n",
 			ihk_mc_get_processor_id(), vm, fault_addr0, reason);
 	
+	/* grow stack */
+	if (fault_addr >= thread->vm->region.stack_start &&
+	    fault_addr < thread->vm->region.stack_end) {
+		range = lookup_process_memory_range(vm,
+						    thread->vm->region.stack_end - 1,
+						    thread->vm->region.stack_end);
+		if (range == NULL) {
+			error = -EFAULT;
+			ekprintf("%s: vm: %p, addr: %p, reason: %lx):"
+				 "stack not found: %d\n",
+				 __func__, vm, fault_addr0, reason, error);
+			goto out;
+		}
+
+		/* don't grow if replaced with hugetlbfs */
+		if (range->memobj) {
+			goto skip;
+		}
+
+		if (fault_addr >= range->start) {
+			goto skip;
+		}
+
+		if (thread->vm->is_memory_range_lock_taken == -1 ||
+		    thread->vm->is_memory_range_lock_taken !=
+		    ihk_mc_get_processor_id()) {
+			ihk_rwspinlock_write_lock_noirq(&vm->memory_range_lock);
+			locked = 1;
+		}
+
+		if (range->pgshift) {
+			range->start = fault_addr &
+				~((1UL << range->pgshift) - 1);
+		} else {
+			range->start = fault_addr & PAGE_MASK;
+		}
+
+		if (locked) {
+			ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
+			locked = 0;
+		}
+
+		dkprintf("%s: addr: %lx, reason: %lx, range: %lx-%lx:"
+			 "stack found\n",
+			 __func__, (unsigned long)fault_addr, reason,
+			 range->start, range->end);
+	}
+skip:
+
 	if (thread->vm->is_memory_range_lock_taken == -1 ||
 			thread->vm->is_memory_range_lock_taken != ihk_mc_get_processor_id()) {
-		ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_read_lock_noirq(&vm->memory_range_lock);
 		locked = 1;
 	} else {
 		dkprintf("%s: INFO: skip locking of memory_range_lock,pid=%d,tid=%d\n",
@@ -2202,13 +2246,16 @@ static int do_page_fault_process_vm(struct process_vm *vm, void *fault_addr0, ui
 		goto out;
 	}
 
-	range = lookup_process_memory_range(vm, fault_addr, fault_addr+1);
-	if (range == NULL) {
-		error = -EFAULT;
-		dkprintf("do_page_fault_process_vm(): vm: %p, addr: %p, reason: %lx):"
-				"out of range: %d\n",
-				vm, fault_addr0, reason, error);
-		goto out;
+	if (!range) {
+		range = lookup_process_memory_range(vm, fault_addr,
+						    fault_addr+1);
+		if (range == NULL) {
+			error = -EFAULT;
+			dkprintf("%s: vm: %p, addr: %p, reason: %lx):"
+				 "out of range: %d\n",
+				 __func__, vm, fault_addr0, reason, error);
+			goto out;
+		}
 	}
 
 	if (((range->flag & VR_PROT_MASK) == VR_PROT_NONE)
@@ -2305,7 +2352,7 @@ static int do_page_fault_process_vm(struct process_vm *vm, void *fault_addr0, ui
 	error = 0;
 out:
 	if (locked) {
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
 	}
 	dkprintf("[%d]do_page_fault_process_vm(%p,%lx,%lx): %d\n",
 			ihk_mc_get_processor_id(), vm, fault_addr0,
@@ -2324,10 +2371,12 @@ int page_fault_process_vm(struct process_vm *fault_vm, void *fault_addr, uint64_
 			break;
 		}
 
+		preempt_enable();
 		if (thread->pgio_fp) {
 			(*thread->pgio_fp)(thread->pgio_arg);
 			thread->pgio_fp = NULL;
 		}
+		preempt_disable();
 	}
 
 	return error;
@@ -2370,16 +2419,15 @@ int init_process_stack(struct thread *thread, struct program_load_desc *pn,
 		size = minsz;
 	}
 	size = (size + USER_STACK_PREPAGE_SIZE - 1) & USER_STACK_PAGE_MASK;
-	dkprintf("%s: stack_premap: %lu, rlim_cur: %lu, minsz: %lu, size: %lu\n",
-			__FUNCTION__,
-			pn->stack_premap,
-			proc->rlimit[MCK_RLIMIT_STACK].rlim_cur,
-			minsz, size);
-	start = (end - size) & USER_STACK_PAGE_MASK;
+	dkprintf("%s: stack_premap: %lu, rlim_cur: %lu, minsz: %lu, size: %lu, maxsz: %lx\n",
+		 __func__, pn->stack_premap,
+		 proc->rlimit[MCK_RLIMIT_STACK].rlim_cur,
+		 minsz, size, maxsz);
+	start = (end - minsz) & USER_STACK_PAGE_MASK;
 
 	/* Apply user allocation policy to stacks */
 	/* TODO: make threshold kernel or mcexec argument */
-	ap_flag = (size >= proc->mpol_threshold &&
+	ap_flag = (minsz >= proc->mpol_threshold &&
 		!(proc->mpol_flags & MPOL_NO_STACK)) ? IHK_MC_AP_USER : 0;
 	dkprintf("%s: max size: %lu, mapped size: %lu %s\n",
 			__FUNCTION__, size, minsz,
@@ -2404,8 +2452,7 @@ int init_process_stack(struct thread *thread, struct program_load_desc *pn,
 	vrflag |= VR_MAXPROT_READ | VR_MAXPROT_WRITE | VR_MAXPROT_EXEC;
 #define	NOPHYS	((uintptr_t)-1)
 	if ((rc = add_process_memory_range(thread->vm, start, end, NOPHYS,
-			vrflag, NULL, 0, USER_STACK_PAGE_SHIFT,
-			NULL, &range)) != 0) {
+			vrflag, NULL, 0, USER_STACK_PAGE_SHIFT, &range)) != 0) {
 		ihk_mc_free_pages_user(stack, minsz >> PAGE_SHIFT);
 		kprintf("%s: error addding process memory range: %d\n", rc);
 		return rc;
@@ -2512,7 +2559,7 @@ int init_process_stack(struct thread *thread, struct program_load_desc *pn,
 	ihk_mc_modify_user_context(thread->uctx, IHK_UCR_STACK_POINTER,
 	                           end + sizeof(unsigned long) * s_ind);
 	thread->vm->region.stack_end = end;
-	thread->vm->region.stack_start = start;
+	thread->vm->region.stack_start = (end - size) & USER_STACK_PAGE_MASK;
 
 	return 0;
 }
@@ -2569,7 +2616,7 @@ unsigned long extend_process_region(struct process_vm *vm,
 
 	if ((rc = add_process_memory_range(vm, end_allocated, new_end_allocated,
 					(p == 0 ? 0 : virt_to_phys(p)), flag, NULL, 0,
-					align_shift, NULL, NULL)) != 0) {
+					align_shift, NULL)) != 0) {
 		ihk_mc_free_pages_user(p, (new_end_allocated - end_allocated) >> PAGE_SHIFT);
 		return end_allocated;
 	}
@@ -2609,7 +2656,7 @@ void flush_process_memory(struct process_vm *vm)
 	int error;
 
 	dkprintf("flush_process_memory(%p)\n", vm);
-	ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_write_lock_noirq(&vm->memory_range_lock);
 	/* Let concurrent page faults know the VM will be gone */
 	vm->exiting = 1;
 	while ((node = next)) {
@@ -2627,7 +2674,7 @@ void flush_process_memory(struct process_vm *vm)
 			}
 		}
 	}
-	ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
 	dkprintf("flush_process_memory(%p):\n", vm);
 	return;
 }
@@ -2642,7 +2689,7 @@ void free_process_memory_ranges(struct process_vm *vm)
 		return;
 	}
 
-	ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_write_lock_noirq(&vm->memory_range_lock);
 	while ((node = next)) {
 		range = rb_entry(node, struct vm_range, vm_rb_node);
 		next = rb_next(node);
@@ -2655,7 +2702,7 @@ void free_process_memory_ranges(struct process_vm *vm)
 			/* through */
 		}
 	}
-	ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
 }
 
 static void free_thread_pages(struct thread *thread)
@@ -2745,7 +2792,7 @@ free_all_process_memory_range(struct process_vm *vm)
 	struct rb_node *node, *next = rb_first(&vm->vm_range_tree);
 	int error;
 
-	ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_write_lock_noirq(&vm->memory_range_lock);
 	while ((node = next)) {
 		range = rb_entry(node, struct vm_range, vm_rb_node);
 		next = rb_next(node);
@@ -2758,7 +2805,7 @@ free_all_process_memory_range(struct process_vm *vm)
 			/* through */
 		}
 	}
-	ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
 }
 
 void
@@ -2814,6 +2861,7 @@ int populate_process_memory(struct process_vm *vm, void *start, size_t len)
 	uintptr_t addr;
 
 	end = (uintptr_t)start + len;
+	preempt_disable();
 	for (addr = (uintptr_t)start; addr < end; addr += PAGE_SIZE) {
 		error = page_fault_process_vm(vm, (void *)addr, reason);
 		if (error) {
@@ -2827,6 +2875,7 @@ int populate_process_memory(struct process_vm *vm, void *start, size_t len)
 
 	error = 0;
 out:
+	preempt_enable();
 	return error;
 }
 
@@ -3189,7 +3238,7 @@ void sched_init(void)
 	               &idle_thread->proc->children_list);
 
 	ihk_mc_init_context(&idle_thread->ctx, NULL, idle);
-	ihk_mc_spinlock_init(&idle_thread->vm->memory_range_lock);
+	ihk_rwspinlock_init(&idle_thread->vm->memory_range_lock);
 	idle_thread->vm->vm_range_tree = RB_ROOT;
 	idle_thread->vm->vm_range_numa_policy_tree = RB_ROOT;
 	idle_thread->proc->pid = 0;
@@ -3299,8 +3348,11 @@ static void do_migrate(void)
 
 		dkprintf("%s: migrated TID %d from CPU %d to CPU %d\n",
 			__FUNCTION__, req->thread->tid, old_cpu_id, cpu_id);
-		
+
 		v->flags |= CPU_FLAG_NEED_RESCHED;
+		/* Kick scheduler on target CPU */
+		ihk_mc_interrupt_cpu(cpu_id, ihk_mc_get_vector(IHK_GV_IKC));
+
 		waitq_wakeup(&req->wq);
 		double_rq_unlock(cur_v, v, irqstate);
 		continue;
@@ -3326,7 +3378,7 @@ void set_timer(int runq_locked)
 	}
 
 	list_for_each_entry(thread, &v->runq, sched_list) {
-		if (thread->status != PS_RUNNING) {
+		if (thread->status != PS_RUNNING && !thread->spin_sleep) {
 			continue;
 		}
 		num_running++;
@@ -3336,7 +3388,7 @@ void set_timer(int runq_locked)
 	if (num_running > 1 || v->current->itimer_enabled ||
 	    !list_empty(&v->backlog_list)) {
 		if (!cpu_local_var(timer_enabled)) {
-			lapic_timer_enable(/*10000000*/1000000);
+			lapic_timer_enable(1000000);
 			cpu_local_var(timer_enabled) = 1;
 		}
 	}
@@ -3516,8 +3568,11 @@ void schedule(void)
 	set_timer(1);
 
 	if (switch_ctx) {
-		dkprintf("schedule: %d => %d \n",
-		        prev ? prev->tid : 0, next ? next->tid : 0);
+		++cpu_local_var(nr_ctx_switches);
+		dkprintf("%s: %d => %d [ctx sws: %lu]\n",
+				__func__,
+				prev ? prev->tid : 0, next ? next->tid : 0,
+				cpu_local_var(nr_ctx_switches));
 
 		if (prev && prev->ptrace_debugreg) {
 			save_debugreg(prev->ptrace_debugreg);
@@ -3677,6 +3732,11 @@ int __sched_wakeup_thread(struct thread *thread,
 
 		/* Make interrupt_exit() call schedule() */
 		v->flags |= CPU_FLAG_NEED_RESCHED;
+
+		/* Make sure to check if timer needs to be re-enabled */
+		if (thread->cpu_id == ihk_mc_get_processor_id()) {
+			set_timer(1);
+		}
 	}
 	else {
 		status = -EINVAL;

kernel/procfs.c

@@ -26,7 +26,6 @@
 #include <mman.h>
 #include <bitmap.h>
 #include <init.h>
-#include <rusage_private.h>
 
 //#define DEBUG_PRINT_PROCFS
 
@@ -360,28 +359,6 @@ static int _process_procfs_request(struct ikc_scd_packet *rpacket, int *result)
 		goto end;
 	}
 #endif /* POSTK_DEBUG_ARCH_DEP_42 */
-	else if (!strcmp(p, "meminfo")) {
-		ans = snprintf(buf, count,
-				"MemTotal:      %10d kB\n"
-				"MemFree:       %10d kB\n"
-				"SwapTotal:     %10d kB\n"
-				"SwapFree:      %10d kB\n"
-				"CommitLimit:   %10d kB\n"
-				"Committed_AS:  %10d kB\n",
-				rusage_get_total_memory() >> 10,
-				rusage_get_free_memory() >> 10,
-				0, 0,
-				rusage_get_free_memory() >> 10,
-				rusage_get_usage_memory() >> 10);
-
-		if (ans < 0 || ans > count ||
-		    buf_add(&buf_top, &buf_cur, buf, ans) < 0) {
-			goto err;
-		}
-
-		ans = 0;
-		goto end;
-	}
 	else {
 		kprintf("unsupported procfs entry: %s\n", p);
 		goto end;
@@ -459,7 +436,7 @@ static int _process_procfs_request(struct ikc_scd_packet *rpacket, int *result)
 	if (strcmp(p, "maps") == 0) {
 		struct vm_range *range;
 
-		if (!ihk_mc_spinlock_trylock_noirq(&vm->memory_range_lock)) {
+		if (!ihk_rwspinlock_read_trylock_noirq(&vm->memory_range_lock)) {
 			if (!result) {
 				if ((err = procfs_backlog(vm, rpacket))) {
 					goto err;
@@ -505,14 +482,14 @@ static int _process_procfs_request(struct ikc_scd_packet *rpacket, int *result)
 
 			if (ans < 0 || ans > count ||
 			    buf_add(&buf_top, &buf_cur, buf, ans) < 0) {
-				ihk_mc_spinlock_unlock_noirq(
+				ihk_rwspinlock_read_unlock_noirq(
 							&vm->memory_range_lock);
 				goto err;
 			}
 			range = next_process_memory_range(vm, range);
 		}
 
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
 
 		ans = 0;
 		goto end;
@@ -535,7 +512,7 @@ static int _process_procfs_request(struct ikc_scd_packet *rpacket, int *result)
 		start = (offset / sizeof(uint64_t)) << PAGE_SHIFT;
 		end = start + ((count / sizeof(uint64_t)) << PAGE_SHIFT);
 
-		if (!ihk_mc_spinlock_trylock_noirq(&vm->memory_range_lock)) {
+		if (!ihk_rwspinlock_read_trylock_noirq(&vm->memory_range_lock)) {
 			if (!result) {
 				if ((err = procfs_backlog(vm, rpacket))) {
 					goto err;
@@ -555,7 +532,7 @@ static int _process_procfs_request(struct ikc_scd_packet *rpacket, int *result)
 			++_buf;
 		}
 
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
 
 		dprintf("/proc/pagemap: 0x%lx - 0x%lx, count: %d\n", 
 			start, end, count);
@@ -587,7 +564,7 @@ static int _process_procfs_request(struct ikc_scd_packet *rpacket, int *result)
 			goto err;
 		}
 
-		if (!ihk_mc_spinlock_trylock_noirq(&vm->memory_range_lock)) {
+		if (!ihk_rwspinlock_read_trylock_noirq(&vm->memory_range_lock)) {
 			if (!result) {
 				if ((err = procfs_backlog(vm, rpacket))) {
 					goto err;
@@ -604,7 +581,7 @@ static int _process_procfs_request(struct ikc_scd_packet *rpacket, int *result)
 				lockedsize += range->end - range->start;
 			range = next_process_memory_range(vm, range);
 		}
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
 
 		cpu_bitmask = &bitmasks[bitmasks_offset];
 		bitmasks_offset += bitmap_scnprintf(cpu_bitmask,

kernel/profile.c

@@ -106,153 +106,108 @@ void profile_event_add(enum profile_event_type type, uint64_t tsc)
 	event->tsc += tsc;
 }
 
-void profile_print_thread_stats(struct thread *thread)
+static void print_profile_events(struct profile_event *profile_events,
+	char *full_hdr_fmt,
+	char *hdr_prefix,
+	int id,
+	unsigned long elapsed_ts)
 {
 	int i;
 	unsigned long flags;
 
+	flags = kprintf_lock();
+	__kprintf(full_hdr_fmt, id, elapsed_ts / 1000);
+
+	__kprintf("%3s: %5s (%3s,%20s): %6s %7s offl: %6s %7s (%6s)\n",
+		"ID", "<num>", "num", "(syscall/event) name", "cnt",
+		"cycles", "cnt", "cycles", "perc");
+
+	for (i = 0; i < PROFILE_SYSCALL_MAX; ++i) {
+		if (!profile_events[i].cnt &&
+				!profile_events[i + PROFILE_SYSCALL_MAX].cnt)
+			continue;
+
+		__kprintf("%s: %4d (%3d,%20s): %6u %6luk "
+				"offl: %6u %6luk (%2d.%2d%%)\n",
+				hdr_prefix,
+				id,
+				i,
+				syscall_name[i],
+				profile_events[i].cnt,
+				profile_events[i].tsc / 1000,
+				profile_events[i + PROFILE_SYSCALL_MAX].cnt,
+				profile_events[i + PROFILE_SYSCALL_MAX].tsc /
+					1000,
+				(profile_events[i].tsc ?
+				 profile_events[i].tsc * 100
+				 / elapsed_ts : 0),
+				(profile_events[i].tsc ?
+				 (profile_events[i].tsc * 10000
+				  / elapsed_ts) % 100 : 0)
+				);
+	}
+
+	for (i = PROFILE_EVENT_MIN; i < PROFILE_EVENT_MAX; ++i) {
+		if (!profile_events[i].cnt)
+			continue;
+
+		__kprintf("%s: %4d (%24s): %6u %6lu\n",
+				hdr_prefix,
+				id,
+				profile_event_names[i - PROFILE_EVENT_MIN],
+				profile_events[i].cnt,
+				(profile_events[i].tsc /
+				 (profile_events[i].cnt ?
+				  profile_events[i].cnt : 1))
+				,
+				(profile_events[i].tsc &&
+				 elapsed_ts ?
+				 profile_events[i].tsc * 100
+				 / elapsed_ts : 0),
+				(profile_events[i].tsc &&
+				 elapsed_ts ?
+				 (profile_events[i].tsc * 10000
+				  / elapsed_ts) % 100 : 0)
+				);
+	}
+
+	kprintf_unlock(flags);
+}
+
+
+void profile_print_thread_stats(struct thread *thread)
+{
 	if (!thread->profile_events)
 		return;
 
 	/* Not yet accumulated period? */
 	if (thread->profile_start_ts) {
-		thread->profile_elapsed_ts += (rdtsc() - thread->profile_start_ts);
+		thread->profile_elapsed_ts += (rdtsc() -
+			thread->profile_start_ts);
 	}
 
-	flags = kprintf_lock();
-
-	__kprintf("TID: %4d elapsed cycles (excluding idle): %luk\n",
-			thread->tid,
-			thread->profile_elapsed_ts / 1000);
-
-	for (i = 0; i < PROFILE_SYSCALL_MAX; ++i) {
-		if (!thread->profile_events[i].cnt &&
-				!thread->profile_events[i + PROFILE_SYSCALL_MAX].cnt)
-			continue;
-
-		__kprintf("TID: %4d (%3d,%20s): %6u %6luk offl: %6u %6luk (%2d.%2d%%)\n",
-				thread->tid,
-				i,
-				syscall_name[i],
-				thread->profile_events[i].cnt,
-				(thread->profile_events[i].tsc /
-				 (thread->profile_events[i].cnt ?
-				  thread->profile_events[i].cnt : 1))
-				/ 1000,
-				thread->profile_events[i + PROFILE_SYSCALL_MAX].cnt,
-				(thread->profile_events[i + PROFILE_SYSCALL_MAX].tsc /
-				 (thread->profile_events[i + PROFILE_SYSCALL_MAX].cnt ?
-				  thread->profile_events[i + PROFILE_SYSCALL_MAX].cnt : 1))
-				/ 1000,
-				 (thread->profile_events[i].tsc ?
-				  thread->profile_events[i].tsc * 100
-				  / thread->profile_elapsed_ts : 0),
-				 (thread->profile_events[i].tsc ?
-				  (thread->profile_events[i].tsc * 10000
-				  / thread->profile_elapsed_ts) % 100 : 0)
-				);
-	}
-
-	for (i = PROFILE_EVENT_MIN; i < PROFILE_EVENT_MAX; ++i) {
-
-		if (!thread->profile_events[i].cnt)
-			continue;
-
-		__kprintf("TID: %4d (%24s): %6u %6luk \n",
-				thread->tid,
-				profile_event_names[i - PROFILE_EVENT_MIN],
-				thread->profile_events[i].cnt,
-				(thread->profile_events[i].tsc /
-				 (thread->profile_events[i].cnt ?
-				  thread->profile_events[i].cnt : 1))
-				/ 1000,
-				(thread->profile_events[i].tsc ?
-				 thread->profile_events[i].tsc * 100
-				 / thread->profile_elapsed_ts : 0),
-				(thread->profile_events[i].tsc ?
-				 (thread->profile_events[i].tsc * 10000
-				  / thread->profile_elapsed_ts) % 100 : 0)
-				);
-	}
-
-
-	kprintf_unlock(flags);
+	print_profile_events(thread->profile_events,
+		"TID: %4d elapsed cycles (excluding idle): %luk\n",
+		"TID",
+		thread->tid,
+		thread->profile_elapsed_ts);
 }
 
 void profile_print_proc_stats(struct process *proc)
 {
-	int i;
-	unsigned long flags;
-
 	if (!proc->profile_events || !proc->profile_elapsed_ts)
 		return;
 
-	flags = kprintf_lock();
-	__kprintf("PID: %4d elapsed cycles for all threads (excluding idle): %luk\n",
-			proc->pid,
-			proc->profile_elapsed_ts / 1000);
-
-	for (i = 0; i < PROFILE_SYSCALL_MAX; ++i) {
-		if (!proc->profile_events[i].cnt &&
-				!proc->profile_events[i + PROFILE_SYSCALL_MAX].cnt)
-			continue;
-
-		__kprintf("PID: %4d (%3d,%20s): %6u %6luk offl: %6u %6luk (%2d.%2d%%)\n",
-				proc->pid,
-				i,
-				syscall_name[i],
-				proc->profile_events[i].cnt,
-				(proc->profile_events[i].tsc /
-				 (proc->profile_events[i].cnt ?
-				  proc->profile_events[i].cnt : 1))
-				/ 1000,
-				proc->profile_events[i + PROFILE_SYSCALL_MAX].cnt,
-				(proc->profile_events[i + PROFILE_SYSCALL_MAX].tsc /
-				 (proc->profile_events[i + PROFILE_SYSCALL_MAX].cnt ?
-				  proc->profile_events[i + PROFILE_SYSCALL_MAX].cnt : 1))
-				/ 1000,
-				(proc->profile_events[i].tsc ?
-				 proc->profile_events[i].tsc * 100
-				 / proc->profile_elapsed_ts : 0),
-				(proc->profile_events[i].tsc ?
-				 (proc->profile_events[i].tsc * 10000
-				  / proc->profile_elapsed_ts) % 100 : 0)
-				);
-	}
-
-	for (i = PROFILE_EVENT_MIN; i < PROFILE_EVENT_MAX; ++i) {
-
-		if (!proc->profile_events[i].cnt)
-			continue;
-
-//		__kprintf("PID: %4d (%24s): %6u %6luk \n",
-		__kprintf("PID: %4d (%24s): %6u %6lu \n",
-				proc->pid,
-				profile_event_names[i - PROFILE_EVENT_MIN],
-				proc->profile_events[i].cnt,
-				(proc->profile_events[i].tsc /
-				 (proc->profile_events[i].cnt ?
-				  proc->profile_events[i].cnt : 1))
-//				/ 1000
-				,
-				(proc->profile_events[i].tsc &&
-				 proc->profile_elapsed_ts ?
-				 proc->profile_events[i].tsc * 100
-				 / proc->profile_elapsed_ts : 0),
-				(proc->profile_events[i].tsc &&
-				 proc->profile_elapsed_ts ?
-				 (proc->profile_events[i].tsc * 10000
-				  / proc->profile_elapsed_ts) % 100 : 0)
-				);
-	}
-
-	kprintf_unlock(flags);
+	print_profile_events(proc->profile_events,
+		"PID: %4d elapsed cycles for all threads (excluding idle): %luk\n",
+		"PID",
+		proc->pid,
+		proc->profile_elapsed_ts);
 }
 
 int profile_accumulate_and_print_job_events(struct process *proc)
 {
 	int i;
-	unsigned long flags;
 	struct mcs_lock_node mcs_node;
 
 	mcs_lock_lock(&job_profile_lock, &mcs_node);
@@ -297,68 +252,20 @@ int profile_accumulate_and_print_job_events(struct process *proc)
 
 	/* Last process? */
 	if (job_nr_processes_left == 0) {
-		flags = kprintf_lock();
-		__kprintf("JOB: (%2d) elapsed cycles for all threads (excluding idle): %luk\n",
-				job_nr_processes,
-				job_elapsed_ts / 1000);
 
-		for (i = 0; i < PROFILE_SYSCALL_MAX; ++i) {
-			if (!job_profile_events[i].cnt &&
-					!job_profile_events[i + PROFILE_SYSCALL_MAX].cnt)
-				continue;
-
-			__kprintf("JOB: (%2d) (%3d,%20s): %6u %6luk offl: %6u %6luk (%2d.%2d%%)\n",
-					job_nr_processes,
-					i,
-					syscall_name[i],
-					job_profile_events[i].cnt,
-					(job_profile_events[i].tsc /
-					 (job_profile_events[i].cnt ?
-					  job_profile_events[i].cnt : 1))
-					/ 1000,
-					job_profile_events[i + PROFILE_SYSCALL_MAX].cnt,
-					(job_profile_events[i + PROFILE_SYSCALL_MAX].tsc /
-					 (job_profile_events[i + PROFILE_SYSCALL_MAX].cnt ?
-					  job_profile_events[i + PROFILE_SYSCALL_MAX].cnt : 1))
-					/ 1000,
-					(job_profile_events[i].tsc ?
-					 job_profile_events[i].tsc * 100
-					 / job_elapsed_ts : 0),
-					(job_profile_events[i].tsc ?
-					 (job_profile_events[i].tsc * 10000
-					  / job_elapsed_ts) % 100 : 0)
-					);
-
-			job_profile_events[i].tsc = 0;
-			job_profile_events[i].cnt = 0;
-			job_profile_events[i + PROFILE_SYSCALL_MAX].tsc = 0;
-			job_profile_events[i + PROFILE_SYSCALL_MAX].cnt = 0;
-		}
-
-		for (i = PROFILE_EVENT_MIN; i < PROFILE_EVENT_MAX; ++i) {
-
-			if (!job_profile_events[i].cnt)
-				continue;
-
-			__kprintf("JOB: (%2d) (%24s): %6u %6luk \n",
-					job_nr_processes,
-					profile_event_names[i - PROFILE_EVENT_MIN],
-					job_profile_events[i].cnt,
-					(job_profile_events[i].tsc /
-					 (job_profile_events[i].cnt ?
-					  job_profile_events[i].cnt : 1))
-					/ 1000);
-
-			job_profile_events[i].tsc = 0;
-			job_profile_events[i].cnt = 0;
-		}
-
-		kprintf_unlock(flags);
+		print_profile_events(job_profile_events,
+			"JOB: (%2d) elapsed cycles for all threads (excluding idle): %luk\n",
+			"JOB",
+			job_nr_processes,
+			job_elapsed_ts);
 
 		/* Reset job process indicators */
 		job_nr_processes = -1;
 		job_nr_processes_left = -1;
 		job_elapsed_ts = 0;
+
+		memset(job_profile_events, 0, sizeof(*job_profile_events) *
+			PROFILE_EVENT_MAX);
 	}
 
 	mcs_lock_unlock(&job_profile_lock, &mcs_node);

kernel/xpmem.c

@@ -423,11 +423,6 @@ static int xpmem_make(
 	struct xpmem_thread_group *seg_tg;
 	struct xpmem_segment *seg;
 	struct mcs_rwlock_node_irqsave lock;
-	struct process_vm *vm = cpu_local_var(current)->vm;
-	int ret;
-	pte_t *seg_pte = NULL;
-	size_t pgsize = 0, seg_size = 0;
-	unsigned long pf_addr;
 
 	XPMEM_DEBUG("call: vaddr=0x%lx, size=0x%lx, permit_type=%d, " 
 		"permit_value=0%04lo", 
@@ -459,27 +454,6 @@ static int xpmem_make(
 		return -EINVAL;
 	}
 
-	/* Page-in segment area */
-	pf_addr = vaddr;
-	while (pf_addr < vaddr + size) {
-		ret = page_fault_process_vm(vm, (void *)pf_addr,
-			PF_POPULATE | PF_WRITE | PF_USER);
-		if (ret) {
-			xpmem_tg_deref(seg_tg);
-			return -ENOENT;
-		}
-		seg_pte = xpmem_vaddr_to_pte(vm, pf_addr, &pgsize);
-		if (!seg_pte || pte_is_null(seg_pte)) {
-			xpmem_tg_deref(seg_tg);
-			return -ENOENT;
-		}
-		pf_addr += pgsize;
-		seg_size += pgsize;
-	}
-	if (seg_size > size) {
-		size = seg_size;
-	}
-
 	segid = xpmem_make_segid(seg_tg);
 	if (segid < 0) {
 		xpmem_tg_deref(seg_tg);
@@ -1037,6 +1011,7 @@ static int xpmem_attach(
 	struct xpmem_segment *seg;
 	struct xpmem_attachment *att;
 	struct mcs_rwlock_node_irqsave at_lock;
+	struct vm_range *vmr;
 	struct process_vm *vm = cpu_local_var(current)->vm;
 
 	XPMEM_DEBUG("call: apid=0x%lx, offset=0x%lx, size=0x%lx, vaddr=0x%lx, " 
@@ -1130,38 +1105,58 @@ static int xpmem_attach(
 	if (flags & MAP_FIXED) {
 		struct vm_range *existing_vmr;
 
-		ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_read_lock_noirq(&vm->memory_range_lock);
 
 		existing_vmr = lookup_process_memory_range(vm, vaddr, 
 			vaddr + size);
 
+		ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
+
 		for (; existing_vmr && existing_vmr->start < vaddr + size;
 			existing_vmr = next_process_memory_range(vm, 
 			existing_vmr)) {
 			if (xpmem_is_private_data(existing_vmr)) {
 				ret = -EINVAL;
-				ihk_mc_spinlock_unlock_noirq(
-					&vm->memory_range_lock);
 				goto out_2;
 			}
 		}
-
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
 	}
 
 	flags |= MAP_ANONYMOUS;
 	XPMEM_DEBUG("do_mmap(): vaddr=0x%lx, size=0x%lx, prot_flags=0x%lx, " 
 		"flags=0x%lx, fd=%d, offset=0x%lx", 
 		vaddr, size, prot_flags, flags, mckfd->fd, offset);
-	/* The new range is associated with shmobj because of
+	/* The new range uses on-demand paging and is associated with shmobj because of 
 	   MAP_ANONYMOUS && !MAP_PRIVATE && MAP_SHARED */
-	at_vaddr = do_mmap(vaddr, size, prot_flags, flags, mckfd->fd,
-			offset, VR_XPMEM, att);
+	at_vaddr = do_mmap(vaddr, size, prot_flags, flags, mckfd->fd, offset);
 	if (IS_ERR((void *)(uintptr_t)at_vaddr)) {
 		ret = at_vaddr;
 		goto out_2;
 	}
 	XPMEM_DEBUG("at_vaddr=0x%lx", at_vaddr);
+	att->at_vaddr = at_vaddr;
+
+	ihk_rwspinlock_read_lock_noirq(&vm->memory_range_lock);
+
+	vmr = lookup_process_memory_range(vm, at_vaddr, at_vaddr + 1);
+
+	/* To identify pages of XPMEM attachment for rusage accounting */
+	if(vmr->memobj) {
+		vmr->memobj->flags |= MF_XPMEM;
+	} else {
+		ekprintf("%s: vmr->memobj equals to NULL\n", __FUNCTION__);
+	}
+
+	ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
+
+	if (!vmr) {
+		ret = -ENOENT;
+		goto out_2;
+	}
+	vmr->private_data = att;
+
+
+	att->at_vmr = vmr;
 
 	*at_vaddr_p = at_vaddr + offset_in_page(att->vaddr);
 
@@ -1187,6 +1182,7 @@ out_1:
 	return ret;
 }
 
+
 static int xpmem_detach(
 	unsigned long at_vaddr)
 {
@@ -1199,18 +1195,18 @@ static int xpmem_detach(
 
 	XPMEM_DEBUG("call: at_vaddr=0x%lx", at_vaddr);
 
-	ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_write_lock_noirq(&vm->memory_range_lock);
 
 	range = lookup_process_memory_range(vm, at_vaddr, at_vaddr + 1);
 
 	if (!range || range->start > at_vaddr) {
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
 		return 0;
 	}
 
 	att = (struct xpmem_attachment *)range->private_data;
 	if (att == NULL) {
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
 		return -EINVAL;
 	}
 
@@ -1220,7 +1216,7 @@ static int xpmem_detach(
 
 	if (att->flags & XPMEM_FLAG_DESTROYING) {
 		mcs_rwlock_writer_unlock(&att->at_lock, &at_lock);
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
 		xpmem_att_deref(att);
 		return 0;
 	}
@@ -1233,7 +1229,7 @@ static int xpmem_detach(
 		att->flags &= ~XPMEM_FLAG_DESTROYING;
 		xpmem_ap_deref(ap);
 		mcs_rwlock_writer_unlock(&att->at_lock, &at_lock);
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
 		xpmem_att_deref(att);
 		return -EACCES;
 	}
@@ -1253,7 +1249,7 @@ static int xpmem_detach(
 		ekprintf("%s: ERROR: xpmem_vm_munmap() failed %d\n", 
 			__FUNCTION__, ret);
 	}
-	ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_write_unlock_noirq(&vm->memory_range_lock);
 	DBUG_ON(ret != 0);
 
 	att->flags &= ~XPMEM_FLAG_VALIDPTEs;
@@ -1416,21 +1412,24 @@ static void xpmem_detach_att(
 
 	XPMEM_DEBUG("call: apid=0x%lx, att=0x%p", ap->apid, att);
 
-	XPMEM_DEBUG("detaching att->vm=0x%p", (void *)att->vm);
+	XPMEM_DEBUG("detaching current->vm=0x%p, att->vm=0x%p", 
+		(void *)cpu_local_var(current)->vm, (void *)att->vm);
+
+	vm = cpu_local_var(current)->vm ? cpu_local_var(current)->vm : att->vm;
+
+	ihk_rwspinlock_read_lock_noirq(&vm->memory_range_lock);
 
 	mcs_rwlock_writer_lock(&att->at_lock, &at_lock);
 
 	if (att->flags & XPMEM_FLAG_DESTROYING) {
 		mcs_rwlock_writer_unlock(&att->at_lock, &at_lock);
+		ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
 		XPMEM_DEBUG("return: XPMEM_FLAG_DESTROYING");
 		return;
 	}
 	att->flags |= XPMEM_FLAG_DESTROYING;
 
-	vm = att->vm;
-	ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
-
-	range = lookup_process_memory_range(vm,
+	range = lookup_process_memory_range(cpu_local_var(current)->vm,
 		att->at_vaddr, att->at_vaddr + 1);
 
 	if (!range || range->start > att->at_vaddr) {
@@ -1438,7 +1437,7 @@ static void xpmem_detach_att(
 		list_del_init(&att->att_list);
 		ihk_mc_spinlock_unlock_noirq(&ap->lock);
 		mcs_rwlock_writer_unlock(&att->at_lock, &at_lock);
-		ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+		ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
 		xpmem_att_destroyable(att);
 		XPMEM_DEBUG("return: range=%p");
 		return;
@@ -1467,13 +1466,14 @@ static void xpmem_detach_att(
 
 	XPMEM_DEBUG("xpmem_vm_munmap(): start=0x%lx, len=0x%lx", 
 		range->start, att->at_size);
-	ret = xpmem_vm_munmap(vm, (void *)range->start, att->at_size);
+	ret = xpmem_vm_munmap(cpu_local_var(current)->vm, (void *)range->start, 
+		att->at_size);
 	if (ret) {
 		ekprintf("%s: ERROR: xpmem_vm_munmap() failed %d\n", 
 			__FUNCTION__, ret);
 	}
 
-	ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
+	ihk_rwspinlock_read_unlock_noirq(&vm->memory_range_lock);
 
 	xpmem_att_destroyable(att);
 
@@ -1576,7 +1576,7 @@ static void xpmem_clear_PTEs_of_att(
 	XPMEM_DEBUG("call: att=0x%p, start=0x%lx, end=0x%lx", 
 		att, start, end);
 
-	ihk_mc_spinlock_lock_noirq(&att->vm->memory_range_lock);
+	ihk_rwspinlock_read_lock_noirq(&att->vm->memory_range_lock);
 	mcs_rwlock_writer_lock(&att->at_lock, &at_lock);
 
 	if (att->flags & XPMEM_FLAG_VALIDPTEs) {
@@ -1636,7 +1636,7 @@ static void xpmem_clear_PTEs_of_att(
 	}
 out:
 	mcs_rwlock_writer_unlock(&att->at_lock, &at_lock);
-	ihk_mc_spinlock_unlock_noirq(&att->vm->memory_range_lock);
+	ihk_rwspinlock_read_unlock_noirq(&att->vm->memory_range_lock);
 
 	XPMEM_DEBUG("return: ");
 }
@@ -1664,11 +1664,16 @@ int xpmem_remove_process_memory_range(
 
 	xpmem_att_ref(att);
 
+	ihk_rwspinlock_read_lock_noirq(
+		&cpu_local_var(current)->vm->memory_range_lock);
+
 	mcs_rwlock_writer_lock(&att->at_lock, &at_lock);
 
 	if (att->flags & XPMEM_FLAG_DESTROYING) {
+		mcs_rwlock_writer_unlock(&att->at_lock, &at_lock);
+		xpmem_att_deref(att);
 		XPMEM_DEBUG("already cleaned up");
-		goto out;
+		return 0;
 	}
 
 	if (vmr->start == att->at_vaddr &&
@@ -1697,7 +1702,7 @@ int xpmem_remove_process_memory_range(
 	else {
 		remaining_vaddr = vmr->end;
 		remaining_vmr = lookup_process_memory_range(
-			vm, remaining_vaddr - 1,
+			cpu_local_var(current)->vm, remaining_vaddr - 1, 
 			remaining_vaddr);
 		if (!remaining_vmr) {
 			ekprintf("%s: ERROR: vm_range is NULL\n", __FUNCTION__);
@@ -1718,7 +1723,7 @@ int xpmem_remove_process_memory_range(
 	}
 
 	remaining_vmr = lookup_process_memory_range(
-		vm, remaining_vaddr,
+		cpu_local_var(current)->vm, remaining_vaddr, 
 		remaining_vaddr + 1);
 	if (!remaining_vmr) {
 		ekprintf("%s: ERROR: vm_range is NULL\n", __FUNCTION__);
@@ -1742,6 +1747,9 @@ int xpmem_remove_process_memory_range(
 out:
 	mcs_rwlock_writer_unlock(&att->at_lock, &at_lock);
 
+	ihk_rwspinlock_read_unlock_noirq(
+		&cpu_local_var(current)->vm->memory_range_lock);
+
 	xpmem_att_deref(att);
 
 	XPMEM_DEBUG("return: ret=%d", 0);
@@ -1870,117 +1878,6 @@ out_1:
 	return ret;
 }
 
-int xpmem_update_process_page_table(
-	struct process_vm *vm, struct vm_range *vmr)
-{
-	int ret = 0;
-	unsigned long seg_vaddr = 0;
-	unsigned long vaddr = vmr->start;
-	pte_t *pte = NULL;
-	pte_t *seg_pte = NULL;
-	struct xpmem_thread_group *ap_tg;
-	struct xpmem_thread_group *seg_tg;
-	struct xpmem_access_permit *ap;
-	struct xpmem_attachment *att;
-	struct xpmem_segment *seg;
-	size_t seg_pgsize;
-	size_t pgsize;
-
-	XPMEM_DEBUG("call: vmr=0x%p", vmr);
-
-	att = (struct xpmem_attachment *)vmr->private_data;
-	if (att == NULL) {
-		return -EFAULT;
-	}
-
-	xpmem_att_ref(att);
-	ap = att->ap;
-	xpmem_ap_ref(ap);
-	ap_tg = ap->tg;
-	xpmem_tg_ref(ap_tg);
-
-	if ((ap->flags & XPMEM_FLAG_DESTROYING) ||
-		(ap_tg->flags & XPMEM_FLAG_DESTROYING)) {
-		ret = -EFAULT;
-		goto out_1;
-	}
-
-	DBUG_ON(cpu_local_var(current)->proc->pid != ap_tg->tgid);
-	DBUG_ON(ap->mode != XPMEM_RDWR);
-
-	seg = ap->seg;
-	xpmem_seg_ref(seg);
-	seg_tg = seg->tg;
-	xpmem_tg_ref(seg_tg);
-
-	if ((seg->flags & XPMEM_FLAG_DESTROYING) ||
-		(seg_tg->flags & XPMEM_FLAG_DESTROYING)) {
-		ret = -ENOENT;
-		goto out_2;
-	}
-
-	att->at_vaddr = vmr->start;
-	att->at_vmr = vmr;
-
-	if ((att->flags & XPMEM_FLAG_DESTROYING) ||
-		(ap_tg->flags & XPMEM_FLAG_DESTROYING) ||
-		(seg_tg->flags & XPMEM_FLAG_DESTROYING)) {
-		goto out_2;
-	}
-
-	seg_vaddr = (att->vaddr & PAGE_MASK) + (vaddr - att->at_vaddr);
-	XPMEM_DEBUG("vaddr=%lx, seg_vaddr=%lx", vaddr, seg_vaddr);
-	while (vaddr < vmr->end) {
-		ret = xpmem_ensure_valid_page(seg, seg_vaddr);
-		if (ret != 0) {
-			goto out_2;
-		}
-
-		seg_pte = xpmem_vaddr_to_pte(seg_tg->vm, seg_vaddr,
-				&seg_pgsize);
-
-		if (seg_pte && !pte_is_null(seg_pte)) {
-			pte = xpmem_vaddr_to_pte(cpu_local_var(current)->vm,
-					vaddr, &pgsize);
-			if (pte && !pte_is_null(pte)) {
-				if (*seg_pte != *pte) {
-					ret = -EFAULT;
-					ekprintf("%s: ERROR: pte mismatch: "
-						"0x%lx != 0x%lx\n",
-						__func__, *seg_pte, *pte);
-				}
-
-				ihk_atomic_dec(&seg->tg->n_pinned);
-				goto out_2;
-			}
-
-			ret = xpmem_remap_pte(vm, vmr, vaddr,
-					0, seg, seg_vaddr);
-			if (ret) {
-				ekprintf("%s: ERROR: xpmem_remap_pte() failed %d\n",
-					__func__, ret);
-			}
-		}
-		flush_tlb_single(vaddr);
-		att->flags |= XPMEM_FLAG_VALIDPTEs;
-
-		seg_vaddr += seg_pgsize;
-		vaddr += seg_pgsize;
-	}
-
-out_2:
-	xpmem_tg_deref(seg_tg);
-	xpmem_seg_deref(seg);
-
-out_1:
-	xpmem_att_deref(att);
-	xpmem_ap_deref(ap);
-	xpmem_tg_deref(ap_tg);
-
-	XPMEM_DEBUG("return: ret=%d", ret);
-
-	return ret;
-}
 
 static int xpmem_remap_pte(
 	struct process_vm *vm,
@@ -2008,13 +1905,13 @@ static int xpmem_remap_pte(
 		"seg_vaddr=0x%lx", 
 		vmr, vaddr, reason, seg->segid, seg_vaddr);
 
-	if (is_remote_vm(seg_tg->vm)) {
-		ihk_mc_spinlock_lock_noirq(&seg_tg->vm->memory_range_lock);
-	}
+	ihk_rwspinlock_read_lock_noirq(&seg_tg->vm->memory_range_lock);
 
 	seg_vmr = lookup_process_memory_range(seg_tg->vm, seg_vaddr, 
 		seg_vaddr + 1);
 
+	ihk_rwspinlock_read_unlock_noirq(&seg_tg->vm->memory_range_lock);
+
 	if (!seg_vmr) {
 		ret = -EFAULT;
 		ekprintf("%s: ERROR: lookup_process_memory_range() failed\n", 
@@ -2025,7 +1922,6 @@ static int xpmem_remap_pte(
 	seg_pte = ihk_mc_pt_lookup_pte(seg_tg->vm->address_space->page_table, 
 		(void *)seg_vaddr, seg_vmr->pgshift, &seg_pgaddr, &seg_pgsize, 
 		&seg_p2align);
-
 	if (!seg_pte) {
 		ret = -EFAULT;
 		ekprintf("%s: ERROR: ihk_mc_pt_lookup_pte() failed\n", 
@@ -2049,34 +1945,31 @@ static int xpmem_remap_pte(
 	att_attr = arch_vrflag_to_ptattr(vmr->flag, reason, att_pte);
 	XPMEM_DEBUG("att_attr=0x%lx", att_attr);
 
-	if (att_pte && !pgsize_is_contiguous(seg_pgsize)) {
-		ret = ihk_mc_pt_set_pte(vm->address_space->page_table, att_pte,
-			seg_pgsize, seg_phys, att_attr);
+	if (att_pte) {
+		ret = ihk_mc_pt_set_pte(vm->address_space->page_table, att_pte, 
+			att_pgsize, seg_phys, att_attr);
 		if (ret) {
 			ret = -EFAULT;
-			ekprintf("%s: ERROR: ihk_mc_pt_set_pte() failed %d\n",
-				__func__, ret);
+			ekprintf("%s: ERROR: ihk_mc_pt_set_pte() failed %d\n", 
+				__FUNCTION__, ret);
 			goto out;
 		}
+		// memory_stat_rss_add() is called by the process hosting the memory area
 	}
 	else {
-		ret = ihk_mc_pt_set_range(vm->address_space->page_table, vm,
-				att_pgaddr, att_pgaddr + seg_pgsize,
-				seg_phys, att_attr,
-				pgsize_to_pgshift(seg_pgsize), vmr, 1);
+		ret = ihk_mc_pt_set_range(vm->address_space->page_table, vm, 
+			att_pgaddr, att_pgaddr + att_pgsize, seg_phys, att_attr,
+								  vmr->pgshift, vmr, 0);
 		if (ret) {
 			ret = -EFAULT;
 			ekprintf("%s: ERROR: ihk_mc_pt_set_range() failed %d\n",
-				 __func__, ret);
+				 __FUNCTION__, ret);
 			goto out;
 		}
+		// memory_stat_rss_add() is called by the process hosting the memory area
 	}
 
 out:
-	if (is_remote_vm(seg_tg->vm)) {
-		ihk_mc_spinlock_unlock_noirq(&seg_tg->vm->memory_range_lock);
-	}
-
 	XPMEM_DEBUG("return: ret=%d", ret);
 
 	return ret;
@@ -2133,7 +2026,8 @@ static pte_t * xpmem_vaddr_to_pte(
 	}
 
 out:
-	return pte;
+
+        return pte;
 }
 
 
@@ -2143,35 +2037,37 @@ static int xpmem_pin_page(
 	struct process_vm *src_vm,
 	unsigned long vaddr)
 {
-	int ret = 0;
+	int ret;
 	struct vm_range *range;
 
 	XPMEM_DEBUG("call: tgid=%d, vaddr=0x%lx", tg->tgid, vaddr);
 
-	if (is_remote_vm(src_vm)) {
-		ihk_mc_spinlock_lock_noirq(&src_vm->memory_range_lock);
-	}
+	ihk_rwspinlock_read_lock_noirq(&src_vm->memory_range_lock);
 
 	range = lookup_process_memory_range(src_vm, vaddr, vaddr + 1);
 
+	ihk_rwspinlock_read_unlock_noirq(&src_vm->memory_range_lock);
+
 	if (!range || range->start > vaddr) {
-		ret = -ENOENT;
-		goto out;
+		return -ENOENT;
 	}
 
 	if (xpmem_is_private_data(range)) {
-		ret = -ENOENT;
-		goto out;
+		return -ENOENT;
 	}
 
-	ihk_atomic_inc(&tg->n_pinned);
-out:
-	if (is_remote_vm(src_vm)) {
-		ihk_mc_spinlock_unlock_noirq(&src_vm->memory_range_lock);
+	ret = page_fault_process_vm(src_vm, (void *)vaddr, 
+		PF_POPULATE | PF_WRITE | PF_USER);
+	if (!ret) {
+		ihk_atomic_inc(&tg->n_pinned);
+	}
+	else {
+		return -ENOENT;
 	}
 
 	XPMEM_DEBUG("return: ret=%d", ret);
-	return ret;
+
+        return ret;
 }
 
 
@@ -2181,24 +2077,30 @@ static void xpmem_unpin_pages(
 	unsigned long vaddr,
 	size_t size)
 {
+	int n_pgs = (((offset_in_page(vaddr) + (size)) + (PAGE_SIZE - 1)) >> 
+		PAGE_SHIFT);
 	int n_pgs_unpinned = 0;
 	size_t vsize = 0;
-	unsigned long end = vaddr + size;
 	pte_t *pte = NULL;
 
 	XPMEM_DEBUG("call: segid=0x%lx, vaddr=0x%lx, size=0x%lx", 
 		seg->segid, vaddr, size);
 
+	XPMEM_DEBUG("n_pgs=%d", n_pgs);
+
 	vaddr &= PAGE_MASK;
 
-	while (vaddr < end) {
+	while (n_pgs > 0) {
 		pte = xpmem_vaddr_to_pte(vm, vaddr, &vsize);
 		if (pte && !pte_is_null(pte)) {
 			n_pgs_unpinned++;
-			vaddr += vsize;
+			vaddr += PAGE_SIZE;
+			n_pgs--;
 		}
 		else {
-			vaddr = ((vaddr + vsize) & (~(vsize - 1)));
+			vsize = ((vaddr + vsize) & (~(vsize - 1)));
+			n_pgs -= (vsize - vaddr) / PAGE_SIZE;
+			vaddr = vsize;
 		}
 	}
 
@@ -2400,15 +2302,3 @@ static int xpmem_validate_access(
 	return 0;
 }
 
-static int is_remote_vm(struct process_vm *vm)
-{
-	int ret = 0;
-
-	if (cpu_local_var(current)->proc->vm != vm) {
-		/* vm is not mine */
-		ret = 1;
-	}
-
-	return ret;
-}
-

lib/include/ihk/lock.h

@@ -15,6 +15,162 @@
 
 #include <arch-lock.h>
 
+
+/* Simple read/write spinlock implementation */
+#define IHK_RWSPINLOCK_WRITELOCKED	(0xffU << 24)
+typedef struct {
+	ihk_atomic_t v;
+} __attribute__((aligned(4))) ihk_rwspinlock_t;
+
+static void ihk_rwspinlock_init(ihk_rwspinlock_t *lock)
+{
+	ihk_atomic_set(&lock->v, 0);
+}
+
+static inline void __ihk_rwspinlock_read_lock(ihk_rwspinlock_t *lock)
+{
+	int desired_old_val;
+	int new_val;
+
+	/*
+	 * Atomically increase number of readers,
+	 * but make sure no writer is holding the lock.
+	 */
+	for (;;) {
+		desired_old_val = ihk_atomic_read(&lock->v);
+		desired_old_val &= ~(IHK_RWSPINLOCK_WRITELOCKED);
+		new_val = desired_old_val + 1;
+
+		/* Only if we have not reached the max number of readers */
+		if (likely((uint32_t)new_val < IHK_RWSPINLOCK_WRITELOCKED)) {
+			if (likely(cmpxchg(&lock->v.counter, desired_old_val, new_val) ==
+					desired_old_val))
+				return;
+		}
+	}
+}
+
+static inline int __ihk_rwspinlock_read_trylock(ihk_rwspinlock_t *lock)
+{
+	int desired_old_val;
+	int new_val;
+
+	/*
+	 * Atomically try to increase number of readers,
+	 * but make sure no writer is holding the lock.
+	 */
+	desired_old_val = ihk_atomic_read(&lock->v);
+	desired_old_val &= ~(IHK_RWSPINLOCK_WRITELOCKED);
+	new_val = desired_old_val + 1;
+
+	/* Only if we have not reached the max number of readers */
+	if (likely((uint32_t)new_val < IHK_RWSPINLOCK_WRITELOCKED)) {
+		if (likely(cmpxchg(&lock->v.counter, desired_old_val, new_val) ==
+					desired_old_val))
+			return 1;
+	}
+
+	return 0;
+}
+
+
+static inline void __ihk_rwspinlock_read_unlock(ihk_rwspinlock_t *lock)
+{
+	ihk_atomic_dec((ihk_atomic_t *)&lock->v);
+}
+
+static inline void __ihk_rwspinlock_write_lock(ihk_rwspinlock_t *lock)
+{
+	/*
+	 * Atomically switch to write-locked state,
+	 * but make sure no one else is holding the lock.
+	 */
+	for (;;) {
+		if (likely(cmpxchg(&lock->v.counter,
+					0, IHK_RWSPINLOCK_WRITELOCKED) == 0))
+			return;
+		cpu_pause();
+	}
+}
+
+static inline void __ihk_rwspinlock_write_unlock(ihk_rwspinlock_t *lock)
+{
+	smp_store_release(&(lock->v.counter), 0);
+}
+
+/* User facing functions */
+static inline void ihk_rwspinlock_read_lock_noirq(ihk_rwspinlock_t *lock)
+{
+	preempt_disable();
+	__ihk_rwspinlock_read_lock(lock);
+}
+
+static inline int ihk_rwspinlock_read_trylock_noirq(ihk_rwspinlock_t *lock)
+{
+	int rc;
+
+	preempt_disable();
+	rc = __ihk_rwspinlock_read_trylock(lock);
+	if (!rc) {
+		preempt_enable();
+	}
+
+	return rc;
+}
+
+static inline void ihk_rwspinlock_write_lock_noirq(ihk_rwspinlock_t *lock)
+{
+	preempt_disable();
+	__ihk_rwspinlock_write_lock(lock);
+}
+
+static inline void ihk_rwspinlock_read_unlock_noirq(ihk_rwspinlock_t *lock)
+{
+	__ihk_rwspinlock_read_unlock(lock);
+	preempt_enable();
+}
+
+static inline void ihk_rwspinlock_write_unlock_noirq(ihk_rwspinlock_t *lock)
+{
+	__ihk_rwspinlock_write_unlock(lock);
+	preempt_enable();
+}
+
+
+static inline
+unsigned long ihk_rwspinlock_read_lock(ihk_rwspinlock_t *lock)
+{
+	unsigned long irqstate = cpu_disable_interrupt_save();
+
+	ihk_rwspinlock_read_lock_noirq(lock);
+	return irqstate;
+}
+
+static inline
+unsigned long ihk_rwspinlock_write_lock(ihk_rwspinlock_t *lock)
+{
+	unsigned long irqstate = cpu_disable_interrupt_save();
+
+	ihk_rwspinlock_write_lock_noirq(lock);
+	return irqstate;
+}
+
+static inline void ihk_rwspinlock_read_unlock(ihk_rwspinlock_t *lock,
+	unsigned long irqstate)
+{
+	ihk_rwspinlock_read_unlock_noirq(lock);
+	cpu_restore_interrupt(irqstate);
+}
+
+static inline void ihk_rwspinlock_write_unlock(ihk_rwspinlock_t *lock,
+	unsigned long irqstate)
+{
+	ihk_rwspinlock_write_unlock_noirq(lock);
+	cpu_restore_interrupt(irqstate);
+}
+
+
+
 #ifndef ARCH_MCS_LOCK
 /* An architecture independent implementation of the
  * Mellor-Crummey Scott (MCS) lock */

scripts/eclair-dump-backtrace.exp.in

@@ -0,0 +1,96 @@
+#!/usr/bin/expect
+
+set INST_DIR "@prefix@"
+
+spawn $INST_DIR/bin/eclair -k $INST_DIR/smp-@ARCH@/kernel/mckernel.img -i -l
+
+set state "init"
+set thread_id 0
+set timeout -1
+
+expect {
+	"Quit anyway? (y or n) " {
+		send "y\r"
+		wait
+		exit 0
+	}
+	"Unknown thread*" {
+		send "quit\r"
+		exp_continue
+	}
+	"in ?? ()" {
+		switch -- $state {
+			"thread_chosen" {
+				set state "thread_skip"
+			}
+			"thread_bt" {
+				set state "thread_skip"
+			}
+		}
+
+		exp_continue
+	}
+	"(eclair) " {
+		switch -- $state {
+			"init" {
+				set state "threads_list"
+				send "info threads\r"
+			}
+			"threads_list" {
+				incr thread_id
+				set state "thread_chosen"
+				send "thread $thread_id\r"
+			}
+			"thread_bt" {
+				incr thread_id
+				set state "thread_chosen"
+				send "thread $thread_id\r"
+			}
+			"thread_skip" {
+				incr thread_id
+				set state "thread_chosen"
+				send "thread $thread_id\r"
+			}
+			"thread_chosen" {
+				set state "thread_bt"
+				send "bt\r"
+			}
+		}
+
+		exp_continue
+	}
+	"Type <RET> for more, q to quit, c to continue without paging--" {
+		switch -- $state {
+			"threads_list" {
+				send "\r"
+			}
+			"thread_bt" {
+				send "q\r"
+			}
+			"thread_skip" {
+				send "q\r"
+			}
+		}
+		exp_continue
+	}
+	"Type <return> to continue, or q <return> to quit" {
+		switch -- $state {
+			"threads_list" {
+				send "\r"
+			}
+			"thread_bt" {
+				send "\r"
+			}
+			"thread_skip" {
+				send "q\r"
+			}
+		}
+		exp_continue
+	}
+	" not known." {
+		expect "(eclair) " { send "quit\r" }
+		expect "Quit anyway? (y or n) " { send "y\r" }
+		exit 0
+	}
+}
+

scripts/mckernel.spec.in

@@ -14,7 +14,7 @@
 %{!?kernel_version: %global kernel_version @UNAME_R@}
 %{!?kernel_dir: %global kernel_dir /usr/src/kernels/%{kernel_version}}
 %define krequires %(echo %{kernel_version} | sed "s/.%{_target_cpu}$//")
-%define ktag %(echo %{krequires} | tr '-' '_' | sed -e 's/\.el[0-9]*$//' | sed -e 's/\.\([a-zA-Z]\)/_\1/')
+%define ktag %(echo %{krequires} | tr '-' '_' | sed -e 's/\.el[0-9_]*$//' | sed -e 's/\.\([a-zA-Z]\)/_\1/')
 
 Name: mckernel
 Version: @MCKERNEL_VERSION@
@@ -23,6 +23,8 @@ Summary: IHK/McKernel
 License: GPLv2
 Source0: mckernel-%{version}.tar.gz
 
+Requires: systemd-libs numactl-libs libdwarf
+
 # kernel_module_package macro does not handle cross build...
 %if "%{?_host_cpu}" == "x86_64" && "%{?_target_cpu}" == "aarch64"
 %define cross_compile 1
@@ -34,7 +36,7 @@ Requires: kernel = %{krequires}
 Requires(post):   /usr/sbin/depmod
 Requires(postun): /usr/sbin/depmod
 %else
-BuildRequires: systemd-devel numactl-devel binutils-devel kernel-devel
+BuildRequires: systemd-devel numactl-devel binutils-devel kernel-devel libdwarf-devel
 # Friendly reminder of the fact that kernel-rpm-macros is no longer included in kernel-devel
 %if 0%{?rhel} >= 8
 BuildRequires: redhat-rpm-config kernel-rpm-macros elfutils-libelf-devel kmod
@@ -96,6 +98,9 @@ This package contains headers and libraries required for build apps using IHK/Mc
 %{_sbindir}/ihkmond
 %{_bindir}/mcexec
 %{_bindir}/eclair
+%{_bindir}/eclair-dump-backtrace.exp
+%{_bindir}/mcinspect
+%{_bindir}/mcps
 %{_bindir}/vmcore2mckdump
 %{_bindir}/mcstat
 %{_libdir}/libihk.so

test/issues/1259/C1259.sh

@@ -1,187 +0,0 @@
-#!/usr/bin/bash
-
-USELTP=0
-USEOSTEST=0
-
-XPMEM_DIR=$HOME/usr
-XPMEM_BUILD_DIR=/home/satoken/xpmem
-
-arch=`uname -p`
-if [ -f "./${arch}_config" ]; then
-	. ./${arch}_config
-else
-	echo "$1 is unexpected arch"
-	exit 1
-fi
-
-. ../../common.sh
-
-sudo insmod ${XPMEM_DIR}/lib/modules/`uname -r`/xpmem.ko
-sudo chmod og+rw /dev/xpmem
-
-issue=1259
-tid=01
-ng=0
-echo "*** C${issue}T${tid} start *******************************"
-echo "** xpmem_attach to Huge mapped memory range"
-echo "** end of range is aligned with Large page size"
-for pgshift in ${PGSHIFT_LIST[@]}
-do
-	${IHKOSCTL} 0 clear_kmsg
-	log_file="./C${issue}T${tid}_${pgshift}.log"
-	echo pageshift: ${pgshift}
-	${MCEXEC} ./huge_page_xpmem ${pgshift} 2 0 > ${log_file}
-	${IHKOSCTL} 0 kmsg >> ${log_file}
-
-	EXPECT_PGSIZE=`grep EXPECT_PAGE_SIZE ${log_file} | awk '{ print $2; }'`
-
-	SEG_ADDR=`grep parent: ${log_file} | awk '{ print $3; }'`
-	SEG_PGSIZE=`cat ${log_file} | awk '/OK/,/DONE/' | \
-grep -o "large_page_allocation.*${SEG_ADDR}.*" | awk '{ print $5; }'`
-
-	XPMEM_ADDR=`grep xpmem_attachment_addr ${log_file} | awk '{ print $3; }'`
-	XPMEM_PGSIZE=`grep -o "xpmem_page_attach.*${XPMEM_ADDR}.*" ${log_file} | awk '{ print $5; }'`
-
-	if [ "${SEG_PGSIZE}" = "${EXPECT_PGSIZE}" ]; then
-		echo "** [ OK ] seg_addr ($SEG_ADDR) is allocated until xpmem_attach"
-	else
-		echo "** [ NG ] seg_addr ($SEG_ADDR) is NOT allocated until xpmem_attach"
-		let ng++
-	fi
-	if [ "${XPMEM_PGSIZE}" = "${EXPECT_PGSIZE}" ]; then
-		echo "** [ OK ] xpmem_addr ($XPMEM_ADDR) is allocated using large pages"
-	else
-		echo "** [ NG ] xpmem_addr ($XPMEM_ADDR) is NOT allocated using large pages"
-		let ng++
-	fi
-done
-if [ ${ng} -eq 0 ]; then
-	echo "*** C${issue}T${tid}: PASSED"
-else
-	echo "*** C${issue}T${tid}: FAILED"
-fi
-echo ""
-
-tid=02
-ng=0
-echo "*** C${issue}T${tid} start *******************************"
-echo "** xpmem_attach to Huge mapped memory range"
-echo "** end of range is NOT aligned with Large page size"
-for pgshift in ${PGSHIFT_LIST[@]}
-do
-	${IHKOSCTL} 0 clear_kmsg
-	log_file="./C${issue}T${tid}_${pgshift}.log"
-	echo pageshift: ${pgshift}
-	${MCEXEC} ./huge_page_xpmem ${pgshift} 2 ${SMALL_PGSIZE} > ${log_file}
-	${IHKOSCTL} 0 kmsg >> ${log_file}
-
-	EXPECT_PGSIZE=`grep EXPECT_PAGE_SIZE ${log_file} | awk '{ print $2; }'`
-
-	SEG_ADDR=`grep parent: ${log_file} | awk '{ print $3; }'`
-	SEG_PGSIZE=`cat ${log_file} | awk '/OK/,/DONE/' | \
-grep -o "large_page_allocation.*${SEG_ADDR}.*" | awk '{ print $5; }'`
-
-	XPMEM_ADDR=`grep xpmem_attachment_addr ${log_file} | awk '{ print $3; }'`
-	XPMEM_PGSIZE=`grep -o "xpmem_page_attach.*${XPMEM_ADDR}.*" ${log_file} | awk '{ print $5; }'`
-
-	if [ "${SEG_PGSIZE}" = "${EXPECT_PGSIZE}" ]; then
-		echo "** [ OK ] seg_addr ($SEG_ADDR) is allocated until xpmem_attach"
-	else
-		echo "** [ NG ] seg_addr ($SEG_ADDR) is NOT allocated until xpmem_attach"
-		let ng++
-	fi
-	if [ "${XPMEM_PGSIZE}" = "${EXPECT_PGSIZE}" ]; then
-		echo "** [ OK ] xpmem_addr ($XPMEM_ADDR) is allocated using large pages"
-	else
-		echo "** [ NG ] xpmem_addr ($XPMEM_ADDR) is NOT allocated using large pages"
-		let ng++
-	fi
-done
-if [ ${ng} -eq 0 ]; then
-	echo "*** C${issue}T${tid}: PASSED"
-else
-	echo "*** C${issue}T${tid}: FAILED"
-fi
-echo ""
-
-tid=03
-ng=0
-echo "*** C${issue}T${tid} start *******************************"
-echo "** xpmem_attach to small mapped memory range"
-${IHKOSCTL} 0 clear_kmsg
-log_file="./C${issue}T${tid}.log"
-echo pageshift: small page
-${MCEXEC} ./huge_page_xpmem -1 2 0 > ${log_file}
-${IHKOSCTL} 0 kmsg >> ${log_file}
-
-EXPECT_PGSIZE=`grep EXPECT_PAGE_SIZE ${log_file} | awk '{ print $2; }'`
-
-XPMEM_ADDR=`grep xpmem_attachment_addr ${log_file} | awk '{ print $3; }'`
-XPMEM_PGSIZE=`grep -o "xpmem_page_attach.*${XPMEM_ADDR}.*" ${log_file} | awk '{ print $5; }'`
-
-if [ "${XPMEM_PGSIZE}" = "${EXPECT_PGSIZE}" ]; then
-	echo "** [ OK ] xpmem_addr ($XPMEM_ADDR) is allocated using small pages"
-else
-	echo "** [ NG ] xpmem_addr ($XPMEM_ADDR) is NOT allocated using small pages"
-	ng=1
-fi
-if [ ${ng} -eq 0 ]; then
-	echo "*** C${issue}T${tid}: PASSED"
-else
-	echo "*** C${issue}T${tid}: FAILED"
-fi
-echo ""
-
-tid=04
-ng=0
-echo "*** C${issue}T${tid} start *******************************"
-echo "** xpmem_attach to multi pagesize range"
-pgshift=${PGSHIFT_LIST[0]}
-${IHKOSCTL} 0 clear_kmsg
-log_file="./C${issue}T${tid}_${pgshift}.log"
-echo pageshift: ${pgshift}
-${MCEXEC} ./multi_vmr_xpmem ${pgshift} 1 > ${log_file}
-${IHKOSCTL} 0 kmsg >> ${log_file}
-
-EXPECT_PGSIZE=`grep EXPECT_PAGE_SIZE ${log_file} | awk '{ print $2; }'`
-
-XPMEM_ADDR=`grep xpmem_large ${log_file} | awk '{ print $3; }'`
-XPMEM_PGSIZE=`grep -o "xpmem_page_attach.*${XPMEM_ADDR}.*" ${log_file} | awk '{ print $5; }'`
-
-if [ "${XPMEM_PGSIZE}" = "${EXPECT_PGSIZE}" ]; then
-	echo "** [ OK ] xpmem_addr ($XPMEM_ADDR) is allocated using large pages"
-else
-	echo "** [ NG ] xpmem_addr ($XPMEM_ADDR) is NOT allocated using large pages"
-	let ng++
-fi
-if [ ${ng} -eq 0 ]; then
-	echo "*** C${issue}T${tid}: PASSED"
-else
-	echo "*** C${issue}T${tid}: FAILED"
-fi
-echo ""
-
-tid=05
-ng=0
-echo "*** C${issue}T${tid} start *******************************"
-echo "** xpmem testsuite"
-cwd=`pwd`
-cd ${XPMEM_BUILD_DIR}/test
-${cwd}/mc_run.sh
-cd ${cwd}
-
-# xpmem basic test
-${MCEXEC} ./XTP_001
-${MCEXEC} ./XTP_002
-${MCEXEC} ./XTP_003
-${MCEXEC} ./XTP_004
-${MCEXEC} ./XTP_005
-${MCEXEC} ./XTP_006
-sleep 3
-${MCEXEC} ./XTP_007
-${MCEXEC} ./XTP_008
-${MCEXEC} ./XTP_009
-${MCEXEC} ./XTP_010
-${MCEXEC} ./XTP_011
-
-sudo rmmod xpmem.ko

test/issues/1259/Makefile

@@ -1,12 +0,0 @@
-XPMEM_DIR=$(HOME)/usr
-CPPFLAGS=-I$(XPMEM_DIR)/include
-LDFLAGS=-L$(XPMEM_DIR)/lib -Wl,-rpath -Wl,$(XPMEM_DIR)/lib -lxpmem
-
-TARGET=huge_page_xpmem multi_vmr_xpmem XTP_001 XTP_002 XTP_003 XTP_004 XTP_005 XTP_006 XTP_007 XTP_008 XTP_009 XTP_010 XTP_011
-
-all: $(TARGET)
-test: all
-	bash ./C1259.sh
-clean:
-	rm -f $(TARGET) C*.log
-

test/issues/1259/README

@@ -1,104 +0,0 @@
-【Issue#1259 動作確認】
-□ テスト内容
-1. Large pageでマップされたメモリ領域でのxpmemの動作確認
-C1259T01:
-	Large pageでマップされたメモリ領域に対してxpmem_attachを行った場合、
-	attach先の領域がxpmem_makeでページインされることを確認する
-	また、xpmemでもLarge pageが利用されることを確認する
-
-C1259T02:
-	Large pageでかつ、最後のページがLarge pageサイズでアラインされていない
-	メモリ領域に対してxpmem_attachを行った場合、
-	xpmemでもLarge pageが利用されることを確認する
-
-C1259T03:
-	Small pageでマップされたメモリ領域に対してxpmem_attachを行った場合、
-	xpmemでもSmall pageが利用されることを確認する
-
-C1259T04:
-	small - large - small のように、異なるページサイズの複数のvm_rangeから
-	構成されるメモリ領域に対してxpmem_attach を行った場合、
-	xpmemでも同じ構成でLarge pageが利用されることを確認する
-
-2. xpmemのテストスイートによる動作確認
-xpmemに付属するテストスイートをMcKernelで実行し、PASSすることを確認する
-
-3. xpmemの基本操作の確認
-xpmemで操作するメモリ領域は、Large pageでマップする
-
-XTP_001: 単一プロセスでのXPMEM操作
-	1. 実行したプロセスがxpmem_make -> xpmem_get -> xpmem_attach -> xpmem_detach -> xpmem_remove
-
-XTP_002: 子プロセスでのXPMEM操作
-	1. 親プロセスがfork()
-	2. 子プロセスがxpmem_make -> xpmem_get -> xpmem_attach -> xpmem_detach ->xpmem_remove
-	3. 子プロセス終了後、親プロセスが終了
-
-XTP_003: 親プロセスがmakeした共有領域への子プロセスによるXPMEM操作
-	1. 親プロセスがxpmem_make
-	2. fork()で子プロセスを作成
-	3. 子プロセスで、xpmem_get -> xpmem_attach -> 値(TEST_VAL)の設定 -> xpmem_detach
-	4. 子プロセスが終了
-	5. 親プロセスが、子プロセスによって設定された値(TEST_VAL)を確認
-	6. 親プロセスがxpmem_remove
-
-XTP_004: fork()後に親プロセスがmakeした共有領域への子プロセスによるXPMEM操作
-	1. fork()で子プロセスを作成
-	2. 親プロセスがxpmem_make
-	3. 子プロセスで、xpmem_get -> xpmem_attach -> 値(TEST_VAL)の設定 -> xpmem_detach
-	4. 子プロセスが終了
-	5. 親プロセスが、子プロセスによって設定された値(TEST_VAL)を確認
-	6. 親プロセスがxpmem_remove
-
-XTP_005: 子プロセスがxpmem_attach後、xpmem_detachをせずに終了
-	1. 親プロセスがxpmem_make
-	2. fork()で子プロセスを作成
-	3. 子プロセスで、xpmem_get -> xpmem_attach
-	4. 子プロセスが終了
-	5. 親プロセスがxpmem_remove
-
-XTP_006: 子プロセスがXPMEM操作を行う時点で、xpmem_makeをした親プロセスが終了している
-	1. 親プロセスがxpmem_make
-	2. fork()で子プロセスを作成
-	3. 親プロセスが終了
-	4. 子プロセスで、xpmem_get (失敗)
-	5. 子プロセスが終了
-
-XTP_007: xpmem_make 呼び出しの異常系
-	1. xpmem_make の第1引数に不正なアドレスを指定する (失敗)
-	2. 1度xpmem_make を実施したメモリ領域に対して、再度xpmem_make を行う (成功)
-
-XTP_008: xpmem_get 呼び出しの異常系
-	1. xpmem_get の第1引数に不正なsegidを指定する (失敗)
-	2. 1度xpmem_get を実施したsegidで、再度xpmem_get を行う (成功)
-
-XTP_009: xpmem_attach 呼び出しの異常系
-	1. xpmem_attach の第1引数に不正なapidを指定する (失敗)
-	2. 1度xpmem_attach を実施したapidで、再度xpmem_attach を行う (成功)
-
-XTP_010: xpmem_detach 呼び出しの異常系
-	1. xpmem_detach の第1引数に不正なアドレスを指定する (成功)
-	2. 1度xpmem_detach を実施したメモリ領域に対して、再度xpmem_detach を行う (成功)
-
-XTP_011: xpmem_remove 呼び出しの異常系
-	1. xpmem_remove の第1引数に不正なsegidを指定する (失敗)
-	2. 1度xpmem_remove を実施したsegidで、再度xpmem_remove を行う (失敗)
-
-□ 実行手順
-1. xpmemのインストールディレクトリをMakefileとC1259.sh中のXPMEM_DIRに記載する
-2. xpmemのビルドディレクトリをC1259.sh中のXPMEM_BUILD_DIRに記載する
-3. 下記の手順でテストを実行する
-$ cd <mckernel>
-$ patch -p0 < test/issues/1259/large_page.patch
-(build mckernel)
-$ cd test/issues/1259
-$ 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/1259/XTP_001.c

@@ -1,60 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	OKNG(segid == -1, "xpmem_make");
-
-	apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-	OKNG(apid == -1, "xpmem_get");
-
-	addr.apid = apid;
-	addr.offset = 0;
-	attach = xpmem_attach(addr, SZ_MEM, NULL);
-	OKNG(attach == (void *)-1, "xpmem_attach");
-
-	rc = xpmem_detach(attach);
-	OKNG(rc == -1, "xpmem_detach");
-
-	rc = xpmem_remove(segid);
-	OKNG(rc == -1, "xpmem_remove");
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_002.c

@@ -1,76 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		rc = xpmem_init();
-		CHKANDJUMP(rc != 0, "xpmem_init in child");
-
-		segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE,
-			(void *)0666);
-		OKNG(segid == -1, "xpmem_make in child");
-
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		OKNG(apid == -1, "xpmem_get in child");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		OKNG(attach == (void *)-1, "xpmem_attach in child");
-
-		rc = xpmem_detach(attach);
-		OKNG(rc == -1, "xpmem_detach in child");
-
-		rc = xpmem_remove(segid);
-		OKNG(rc == -1, "xpmem_remove in child");
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_003.c

@@ -1,79 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	OKNG(segid == -1, "xpmem_make");
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		OKNG(apid == -1, "xpmem_get in child");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		OKNG(attach == (void *)-1, "xpmem_attach in child");
-
-		*((unsigned long *)attach) = TEST_VAL;
-
-		rc = xpmem_detach(attach);
-		OKNG(rc == -1, "xpmem_detach in child");
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-
-		OKNG(*((unsigned long *)mem) != TEST_VAL, "validate TEST_VAL");
-
-		rc = xpmem_remove(segid);
-		OKNG(rc == -1, "xpmem_remove");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_004.c

@@ -1,112 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-#define BUFF_SIZE 1024
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-	key_t key = ftok(argv[0], 0);
-	int shmid;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	shmid = shmget(key, SZ_MEM, IPC_CREAT | 0660);
-	CHKANDJUMP(shmid == -1, "shmget");
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		void *shm = shmat(shmid, NULL, 0);
-
-		CHKANDJUMP(shm == (void *)-1, "shmat in child");
-
-		while ((segid = *(xpmem_segid_t *)shm) == 0) {
-		};
-
-		rc = shmdt(shm);
-		CHKANDJUMP(rc == -1, "shmdt");
-
-		rc = xpmem_init();
-		CHKANDJUMP(rc != 0, "xpmem_init in child");
-
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		OKNG(apid == -1, "xpmem_get in child");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		OKNG(attach == (void *)-1, "xpmem_attach in child");
-
-		*((unsigned long *)attach) = TEST_VAL;
-
-		rc = xpmem_detach(attach);
-		OKNG(rc == -1, "xpmem_detach in child");
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		void *shm = shmat(shmid, NULL, 0);
-		struct shmid_ds buf;
-
-		CHKANDJUMP(shm == (void *)-1, "shmat in parent");
-		rc = xpmem_init();
-		CHKANDJUMP(rc != 0, "xpmem_init");
-
-		segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE,
-			(void *)0666);
-		OKNG(segid == -1, "xpmem_make");
-
-		*(xpmem_segid_t *)shm = segid;
-
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-
-		OKNG(*((unsigned long *)mem) != TEST_VAL, "validate TEST_VAL");
-
-		rc = shmctl(shmid, IPC_RMID, &buf);
-		CHKANDJUMP(rc == -1, "shmctl");
-
-		rc = shmdt(shm);
-		CHKANDJUMP(rc == -1, "shmdt");
-
-		rc = xpmem_remove(segid);
-		OKNG(rc == -1, "xpmem_remove");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_005.c

@@ -1,76 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	OKNG(segid == -1, "xpmem_make");
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		OKNG(apid == -1, "xpmem_get in child");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		OKNG(attach == (void *)-1, "xpmem_attach in child");
-
-		*((unsigned long *)attach) = TEST_VAL;
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-
-		OKNG(*((unsigned long *)mem) != TEST_VAL, "validate TEST_VAL");
-
-		rc = xpmem_remove(segid);
-		OKNG(rc == -1, "xpmem_remove");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_006.c

@@ -1,64 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	OKNG(segid == -1, "xpmem_make");
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		sleep(1); /* wait for parent process exit */
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		OKNG(apid != -1,
-			"xpmem_get in child failed (parent process exited already");
-		fflush(0);
-
-	} else {
-		/* Parent process */
-		_exit(0);
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_007.c

@@ -1,89 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-#define BAD_ADDRESS ((void *)-1)
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(BAD_ADDRESS, SZ_MEM, XPMEM_PERMIT_MODE,
-		(void *)0666);
-	OKNG(segid != -1, "xpmem_make failed (invalid address)");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	CHKANDJUMP(segid == -1, "xpmem_make");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	OKNG(segid == -1, "xpmem_make succeed(do twice to same address)");
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		CHKANDJUMP(apid == -1, "xpmem_get in child");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		CHKANDJUMP(attach == (void *)-1, "xpmem_attach in child");
-
-		*((unsigned long *)attach) = TEST_VAL;
-
-		rc = xpmem_detach(attach);
-		CHKANDJUMP(rc == -1, "xpmem_detach in child");
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-
-		CHKANDJUMP(*((unsigned long *)mem) != TEST_VAL,
-			"validate TEST_VAL");
-
-		rc = xpmem_remove(segid);
-		CHKANDJUMP(rc == -1, "xpmem_remove");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_008.c

@@ -1,89 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-#define BAD_SEGID -1
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	CHKANDJUMP(segid == -1, "xpmem_make");
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		apid = xpmem_get(BAD_SEGID, XPMEM_RDWR, XPMEM_PERMIT_MODE,
-			NULL);
-		OKNG(apid != -1, "xpmem_get in child failed (invalid segid)");
-
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		CHKANDJUMP(apid == -1, "xpmem_get in child");
-
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		OKNG(apid == -1, "xpmem_get in child (do twice to same segid");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		CHKANDJUMP(attach == (void *)-1, "xpmem_attach in child");
-
-		*((unsigned long *)attach) = TEST_VAL;
-
-		rc = xpmem_detach(attach);
-		CHKANDJUMP(rc == -1, "xpmem_detach in child");
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-
-		CHKANDJUMP(*((unsigned long *)mem) != TEST_VAL,
-			"validate TEST_VAL");
-
-		rc = xpmem_remove(segid);
-		CHKANDJUMP(rc == -1, "xpmem_remove");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_009.c

@@ -1,92 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	CHKANDJUMP(segid == -1, "xpmem_make");
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		CHKANDJUMP(apid == -1, "xpmem_get in child");
-
-		addr.apid = -1;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		OKNG(attach != (void *)-1,
-			"xpmem_attach in childi failed (invalid apid)");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		CHKANDJUMP(attach == (void *)-1, "xpmem_attach in child");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		OKNG(attach == (void *)-1,
-			"xpmem_attach in child succeed (do twice to same apid)");
-
-		*((unsigned long *)attach) = TEST_VAL;
-
-		rc = xpmem_detach(attach);
-		CHKANDJUMP(rc == -1, "xpmem_detach in child");
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-
-		CHKANDJUMP(*((unsigned long *)mem) != TEST_VAL,
-			"validate TEST_VAL");
-
-		rc = xpmem_remove(segid);
-		CHKANDJUMP(rc == -1, "xpmem_remove");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_010.c

@@ -1,90 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-#define BAD_ADDRESS ((void *) -1)
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	CHKANDJUMP(segid == -1, "xpmem_make");
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		CHKANDJUMP(apid == -1, "xpmem_get in child");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		CHKANDJUMP(attach == (void *)-1, "xpmem_attach in child");
-
-		*((unsigned long *)attach) = TEST_VAL;
-
-		rc = xpmem_detach(BAD_ADDRESS);
-		OKNG(rc == -1,
-			"xpmem_detach in child succeed (invalid address)");
-
-		rc = xpmem_detach(attach);
-		CHKANDJUMP(rc == -1, "xpmem_detach in child");
-
-		rc = xpmem_detach(attach);
-		OKNG(rc == -1,
-			"xpmem_detach in child succeed (do twice to same address)");
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-
-		CHKANDJUMP(*((unsigned long *)mem) != TEST_VAL,
-			"validate TEST_VAL");
-
-		rc = xpmem_remove(segid);
-		CHKANDJUMP(rc == -1, "xpmem_remove");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/XTP_011.c

@@ -1,88 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include <libgen.h>
-#include "util2.h"
-
-#define BAD_SEGID -1
-
-int main(int argc, char **argv)
-{
-	void *mem, *attach;
-	int rc = 0;
-	int status;
-	pid_t pid;
-	xpmem_segid_t segid;
-	xpmem_apid_t apid;
-	struct xpmem_addr addr;
-
-	printf("*** %s start ***\n", basename(argv[0]));
-
-	mem = mmap(0, SZ_MEM, PROT_READ | PROT_WRITE,
-		MAP_ANONYMOUS | MAP_PRIVATE |
-		MAP_HUGETLB | (LARGE_PAGE_SHIFT << MAP_HUGE_SHIFT), -1, 0);
-	CHKANDJUMP(mem == NULL, "mmap");
-	memset(mem, 0, SZ_MEM);
-
-	rc = xpmem_init();
-	CHKANDJUMP(rc != 0, "xpmem_init");
-
-	segid = xpmem_make(mem, SZ_MEM, XPMEM_PERMIT_MODE, (void *)0666);
-	CHKANDJUMP(segid == -1, "xpmem_make");
-
-	fflush(0);
-	pid = fork();
-	CHKANDJUMP(pid == -1, "fork failed\n");
-
-	if (pid == 0) {
-		/* Child process */
-		apid = xpmem_get(segid, XPMEM_RDWR, XPMEM_PERMIT_MODE, NULL);
-		CHKANDJUMP(apid == -1, "xpmem_get in child");
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, SZ_MEM, NULL);
-		CHKANDJUMP(attach == (void *)-1, "xpmem_attach in child");
-
-		*((unsigned long *)attach) = TEST_VAL;
-
-		rc = xpmem_detach(attach);
-		CHKANDJUMP(rc == -1, "xpmem_detach in child");
-
-		fflush(0);
-		_exit(0);
-	} else {
-		/* Parent process */
-		rc = waitpid(pid, &status, 0);
-		CHKANDJUMP(rc == -1, "waitpid failed\n");
-
-		CHKANDJUMP(*((unsigned long *)mem) != TEST_VAL,
-			"validate TEST_VAL");
-
-		rc = xpmem_remove(BAD_SEGID);
-		OKNG(rc != -1, "xpmem_remove failed (invalid segid)");
-
-		rc = xpmem_remove(segid);
-		CHKANDJUMP(rc == -1, "xpmem_remove");
-
-		rc = xpmem_remove(segid);
-		OKNG(rc != -1, "xpmem_remove failed (do twice to same segid)");
-	}
-
-	printf("*** %s PASSED\n\n", basename(argv[0]));
-	return 0;
-
-fn_fail:
-	printf("*** %s FAILED\n\n", basename(argv[0]));
-
-	return -1;
-}

test/issues/1259/aarch64_config

@@ -1,2 +0,0 @@
-PGSHIFT_LIST=(21 29 34)
-SMALL_PGSIZE=65536

test/issues/1259/aarch64_result.log

@@ -1,184 +0,0 @@
-*** C1259T01 start *******************************
-** xpmem_attach to Huge mapped memory range
-** end of range is aligned with Large page size
-pageshift: 21
-** [ OK ] seg_addr (100000400000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (100000400000) is allocated using large pages
-pageshift: 29
-** [ OK ] seg_addr (100020000000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (100020000000) is allocated using large pages
-pageshift: 34
-** [ OK ] seg_addr (100400000000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (100400000000) is allocated using large pages
-*** C1259T01: PASSED
-
-*** C1259T02 start *******************************
-** xpmem_attach to Huge mapped memory range
-** end of range is NOT aligned with Large page size
-pageshift: 21
-** [ OK ] seg_addr (100000400000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (100000400000) is allocated using large pages
-pageshift: 29
-** [ OK ] seg_addr (100020000000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (100020000000) is allocated using large pages
-pageshift: 34
-** [ OK ] seg_addr (100400000000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (100400000000) is allocated using large pages
-*** C1259T02: PASSED
-
-*** C1259T03 start *******************************
-** xpmem_attach to small mapped memory range
-pageshift: small page
-** [ OK ] xpmem_addr (100000210000) is allocated using small pages
-*** C1259T03: PASSED
-
-*** C1259T04 start *******************************
-** xpmem_attach to multi pagesize range
-pageshift: 21
-** [ OK ] xpmem_addr (100000600000) is allocated using large pages
-*** C1259T04: PASSED
-
-*** C1259T05 start *******************************
-** xpmem testsuite
-XPMEM version = 26003
-
-==== test_base STARTS ====
-xpmem_proc1: mypid = 38514
-xpmem_proc1: sharing 262144 bytes
-xpmem_proc1: segid = 200009672 at 0x100000210000
-
-xpmem_proc2: mypid = 38711
-xpmem_proc2: segid = 200009672
-xpmem_proc2: attached at 0x100000210000
-xpmem_proc2: adding 1 to all elems
-
-xpmem_proc1: verifying data...done
-==== test_base PASSED ====
-
-==== test_two_attach STARTS ====
-xpmem_proc1: mypid = 39028
-xpmem_proc1: sharing 262144 bytes
-xpmem_proc1: segid = 200009874 at 0x100000210000
-
-xpmem_proc2: mypid = 39233
-xpmem_proc2: segid = 200009874
-xpmem_proc2: attached at 0x100000210000
-xpmem_proc2: attached at 0x100000250000
-xpmem_proc2: adding 1 to all elems using 0x100000210000
-xpmem_proc2: adding 1 to all elems using 0x100000250000
-
-xpmem_proc1: verifying data...done
-==== test_two_attach PASSED ====
-
-==== test_two_shares STARTS ====
-xpmem_proc1: mypid = 39429
-xpmem_proc1: sharing 2 segments, 262144 bytes each
-xpmem_proc1: segid[0] = 200009a05 at 0x100000210000
-xpmem_proc1: segid[1] = 400009a05 at 0x100000250000
-
-xpmem_proc2: mypid = 39625
-xpmem_proc2: segid[0] = 200009a05
-xpmem_proc2: segid[1] = 400009a05
-xpmem_proc2: data[0] attached at 0x100000210000
-xpmem_proc2: data[1] attached at 0x100000250000
-xpmem_proc2: adding 1 to all elems using 0x100000210000
-xpmem_proc2: adding 1 to all elems using 0x100000250000
-
-xpmem_proc1: verifying data...done
-==== test_two_shares PASSED ====
-
-==== test_fork STARTS ====
-xpmem_proc1: mypid = 39831
-xpmem_proc1: sharing 262144 bytes
-xpmem_proc1: segid = 200009b97 at 0x100000210000
-
-xpmem_proc2: mypid = 40027
-xpmem_proc2: segid = 200009b97
-xpmem_proc2: attached at 0x100000220000
-xpmem_proc2: reading to pin pages
-xpmem_proc2: waiting for COW...
-
-xpmem_proc1: forking a child
-xpmem_proc1: adding 1 to all elems to induce COW
-
-xpmem_child: hello from pid 40224
-
-xpmem_proc1: give control back to xpmem_proc2
-
-xpmem_proc2: adding 1 to all elems
-
-xpmem_proc1: verifying data...done
-==== test_fork PASSED ====
-
-*** XTP_001 start ***
- [OK] xpmem_make
- [OK] xpmem_get
- [OK] xpmem_attach
- [OK] xpmem_detach
- [OK] xpmem_remove
-*** XTP_001 PASSED
-
-*** XTP_002 start ***
- [OK] xpmem_make in child
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] xpmem_remove in child
-*** XTP_002 PASSED
-
-*** XTP_003 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_003 PASSED
-
-*** XTP_004 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_004 PASSED
-
-*** XTP_005 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_005 PASSED
-
-*** XTP_006 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child failed (parent process exited already
-*** XTP_006 PASSED
-
-*** XTP_007 start ***
- [OK] xpmem_make failed (invalid address)
- [OK] xpmem_make succeed(do twice to same address)
-*** XTP_007 PASSED
-
-*** XTP_008 start ***
- [OK] xpmem_get in child failed (invalid segid)
- [OK] xpmem_get in child (do twice to same segid
-*** XTP_008 PASSED
-
-*** XTP_009 start ***
- [OK] xpmem_attach in childi failed (invalid apid)
- [OK] xpmem_attach in child succeed (do twice to same apid)
-*** XTP_009 PASSED
-
-*** XTP_010 start ***
- [OK] xpmem_detach in child succeed (invalid address)
- [OK] xpmem_detach in child succeed (do twice to same address)
-*** XTP_010 PASSED
-
-*** XTP_011 start ***
- [OK] xpmem_remove failed (invalid segid)
- [OK] xpmem_remove failed (do twice to same segid)
-*** XTP_011 PASSED
-

test/issues/1259/huge_page_xpmem.c

@@ -1,182 +0,0 @@
-#include <unistd.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include "util.h"
-
-#define DEBUG
-
-#define SHM_SIZE (1UL << 12)
-
-#define MAP_HUGE_SHIFT 26
-#define KEYWORD 0x12345678UL
-
-void usage(void)
-{
-	printf("Usage: huge_page_map: <pgshift> <pgnum> <pgoffset>\n");
-	printf("\tpgshift : pageshift of map area (Using MAP_HUGETLB)\n");
-	printf("\t            -1 means using small pagesize\n");
-	printf("\tpgnum   : number of page of map area\n");
-	printf("\tpgoffset: offset of last page\n");
-}
-
-
-void *mmap_flag(size_t mapsize, int page_shift)
-{
-	char *addr_mmap;
-	int flags = MAP_ANONYMOUS | MAP_PRIVATE;
-
-	if (page_shift >= 0) {
-		/* mean use MAP_HUGETLB */
-		flags |= MAP_HUGETLB | (page_shift << MAP_HUGE_SHIFT);
-	}
-
-	addr_mmap = mmap(0, mapsize,
-			PROT_READ | PROT_WRITE,
-			flags, -1, 0);
-
-	return addr_mmap;
-}
-
-int main(int argc, char **argv)
-{
-	void *mem;
-	int ret = 0;
-	pid_t pid;
-	int status;
-	key_t key = ftok(argv[0], 0);
-	void *shm;
-	int shmid;
-	xpmem_segid_t segid;
-	struct shmid_ds shmctl_buf;
-	int pgshift, pgnum;
-	size_t pgsize, map_size, pgoffset;
-
-	if (argc < 4) {
-		printf("Err: Too few arguments\n");
-		usage();
-		return -1;
-	}
-
-	pgshift = atoi(argv[1]);
-	pgnum = atoi(argv[2]);
-	pgoffset = atol(argv[3]);
-	if (pgshift > 0) {
-		pgsize = (1UL << pgshift);
-	} else {
-		pgsize = getpagesize();
-	}
-
-	if (pgoffset > 0) {
-		map_size = (pgsize * (pgnum - 1)) + pgoffset;
-	} else {
-		map_size = pgsize * pgnum;
-	}
-
-	shmid = shmget(key, SHM_SIZE, IPC_CREAT | 0660);
-	CHKANDJUMP(shmid == -1, EXIT_FAILURE, "shmget failed: %s\n",
-		   strerror(errno));
-
-	printf("EXPECT_PAGE_SIZE: 0x%lx\n", pgsize);
-	fflush(stdout);
-
-	pid = fork();
-	CHKANDJUMP(pid == -1, EXIT_FAILURE, "fork failed\n");
-	if (pid == 0) {
-		xpmem_apid_t apid;
-		struct xpmem_addr addr;
-		void *attach;
-
-		shm = shmat(shmid, NULL, 0);
-		CHKANDJUMP(shm == (void *)-1, EXIT_FAILURE,
-			   "shmat failed: %s\n", strerror(errno));
-
-		while ((segid = *(xpmem_segid_t *)shm) == 0) {
-		};
-
-		ret = shmdt(shm);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "shmdt failed\n");
-
-		apid = xpmem_get(segid, XPMEM_RDWR,
-				 XPMEM_PERMIT_MODE, NULL);
-		CHKANDJUMP(apid == -1, EXIT_FAILURE, "xpmem_get failed: %s\n",
-			   strerror(errno));
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, pgsize * pgnum, NULL);
-
-		CHKANDJUMP(attach == (void *)-1, EXIT_FAILURE,
-			   "xpmem_attach failed: %s\n", strerror(errno));
-
-		printf("child: xpmem_attachment_addr: %lx\n",
-		       attach);
-
-		*((unsigned long *)attach) = KEYWORD;
-		if (pgnum > 1 && pgshift > 0) {
-			*((unsigned long *)(attach +
-				(1UL << pgshift))) = KEYWORD;
-		}
-		*((unsigned long *)(attach + map_size
-			- sizeof(unsigned long *))) = KEYWORD;
-
-		ret = xpmem_detach(attach);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "xpmem_detach failed\n");
-
-		exit(0);
-	} else {
-		mem = mmap_flag(map_size, pgshift);
-		CHKANDJUMP(mem == MAP_FAILED, EXIT_FAILURE, "mmap failed\n");
-		printf("parent: anonymous_map_addr: %lx - %lx\n",
-		       mem, mem + map_size);
-
-		shm = shmat(shmid, NULL, 0);
-
-		CHKANDJUMP(shm == (void *)-1, EXIT_FAILURE,
-			   "shmat failed: %s\n", strerror(errno));
-
-		segid = xpmem_make(mem, map_size, XPMEM_PERMIT_MODE,
-				   (void *)0666);
-		CHKANDJUMP(segid == -1, EXIT_FAILURE,
-			   "xpmem_ioctl failed: %s\n", strerror(errno));
-
-		*(xpmem_segid_t *)shm = segid;
-
-		ret = waitpid(pid, &status, 0);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "waitpid failed\n");
-
-		NG(*(unsigned long *)mem == KEYWORD,
-			"HEAD of xpmem area is INVALID. isn't shared?\n");
-		if (pgnum > 1 && pgshift > 0) {
-			NG(*((unsigned long *)(mem +
-				(1UL << pgshift))) == KEYWORD,
-			"MIDDLE of xpmem area is INVALID. isn't shared?\n");
-		}
-		NG(*((unsigned long *)(mem + map_size
-				- sizeof(unsigned long *))) == KEYWORD,
-			"TAIL of xpmem area is INVALID. isn't shared?\n");
-		printf("xpmem area is shared: OK\n");
-
-		ret = shmctl(shmid, IPC_RMID, &shmctl_buf);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "shmctl failed\n");
-
-		ret = shmdt(shm);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "shmdt failed\n");
-
-		ret = xpmem_remove(segid);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "xpmem_remove failed\n");
-	}
-
-	ret = 0;
- out:
-	return ret;
-}

test/issues/1259/large_page.patch

@@ -1,99 +0,0 @@
-diff --git arch/arm64/kernel/memory.c arch/arm64/kernel/memory.c
-index a84bc21..7368ada 100644
---- arch/arm64/kernel/memory.c
-+++ arch/arm64/kernel/memory.c
-@@ -2701,6 +2701,13 @@ int set_range_l1(void *args0, pte_t *ptep, uintptr_t base, uintptr_t start,
- 	ptl1_set(ptep, pte);
- 
- 	error = 0;
-+
-+	if (args->attr[0] & PTE_CONT &&
-+	    __page_offset(base, PTL1_CONT_SIZE) == 0) {
-+		kprintf("%s: large_page_allocation, addr: %016lx, size: 0x%lx , phys: %lx\n",
-+			__func__, base, PTL1_CONT_SIZE, phys);
-+	}
-+
- 	// call memory_stat_rss_add() here because pgshift is resolved here
- 	if (!(args->attr[0] & PTE_CONT)) {
- 		if (rusage_memory_stat_add(args->range, phys,
-@@ -2810,6 +2817,17 @@ retry:
- 					level);
- 
- 				error = 0;
-+
-+				if (args->attr[level-1] & PTE_CONT) {
-+					if (__page_offset(base, tbl.cont_pgsize) == 0) {
-+						kprintf("%s: large_page_allocation, addr: %016lx, size: 0x%lx , phys: %lx\n",
-+							__func__, base, tbl.cont_pgsize, phys);
-+					}
-+				} else {
-+					kprintf("%s: large_page_allocation, addr: %016lx, size: 0x%lx , phys: %lx\n",
-+						__func__, base, tbl.pgsize, phys);
-+				}
-+
- 				dkprintf("set_range_middle(%lx,%lx,%lx,%d):"
- 					 "large page. %d %lx\n",
- 					 base, start, end, level, error, *ptep);
-diff --git arch/x86_64/kernel/memory.c arch/x86_64/kernel/memory.c
-index df545e1..633e390 100644
---- arch/x86_64/kernel/memory.c
-+++ arch/x86_64/kernel/memory.c
-@@ -1931,6 +1931,10 @@ retry:
- 			dkprintf("set_range_l2(%lx,%lx,%lx):"
- 					"2MiB page. %d %lx\n",
- 					base, start, end, error, *ptep);
-+
-+			kprintf("%s: large_page_allocation, addr: %016lx, size: 0x%lx\n",
-+				__func__, base, PTL2_SIZE);
-+
- 			// Call memory_stat_rss_add() here because pgshift is resolved here
- 			if (rusage_memory_stat_add(args->range, phys, PTL2_SIZE, PTL2_SIZE)) {
- 				dkprintf("%lx+,%s: calling memory_stat_rss_add(),base=%lx,phys=%lx,size=%ld,pgsize=%ld\n", phys, __FUNCTION__, base, phys, PTL2_SIZE, PTL2_SIZE);
-@@ -2020,6 +2024,9 @@ retry:
- 					"1GiB page. %d %lx\n",
- 					base, start, end, error, *ptep);
- 
-+			kprintf("%s: large_page_allocation, addr: %016lx, size: 0x%lx\n",
-+				__func__, base, PTL3_SIZE);
-+
- 			// Call memory_stat_rss_add() here because pgshift is resolved here
- 			if (rusage_memory_stat_add(args->range, phys, PTL3_SIZE, PTL3_SIZE)) {
- 				dkprintf("%lx+,%s: calling memory_stat_rss_add(),base=%lx,phys=%lx,size=%ld,pgsize=%ld\n", phys, __FUNCTION__, base, phys, PTL3_SIZE, PTL3_SIZE);
-diff --git kernel/process.c kernel/process.c
-index 809f5e0..cba9e5a 100644
---- kernel/process.c
-+++ kernel/process.c
-@@ -2059,6 +2059,12 @@ retry:
- 		}
- 
- 		dkprintf("%s: attr=%x\n", __FUNCTION__, attr);
-+
-+		if (pgsize > PAGE_SIZE) {
-+			kprintf("large_page_allocation, addr: %016lx, size: %d, phys: %lx\n",
-+				pgaddr, pgsize, phys);
-+		}
-+
- 		error = ihk_mc_pt_set_pte(vm->address_space->page_table, ptep,
- 		                          pgsize, phys, attr);
- 		if (error) {
-diff --git kernel/xpmem.c kernel/xpmem.c
-index e1d0231..c9da711 100644
---- kernel/xpmem.c
-+++ kernel/xpmem.c
-@@ -514,6 +514,7 @@ static int xpmem_make(
- 	*segid_p = segid;
- 
- 	XPMEM_DEBUG("return: ret=%d, segid=0x%lx", 0, *segid_p);
-+	kprintf("%s: DONE\n", __func__);
- 
- 	return 0;
- }
-@@ -1994,6 +1995,8 @@ int xpmem_update_process_page_table(
- 		flush_tlb_single(vaddr);
- 		att->flags |= XPMEM_FLAG_VALIDPTEs;
- 
-+		kprintf("%s: xpmem_page_attach, addr: %016lx, size: 0x%lx\n",
-+					__func__, vaddr, seg_pgsize);
- 		seg_vaddr += seg_pgsize;
- 		vaddr += seg_pgsize;
- 	}

test/issues/1259/mc_run.sh

@@ -1,15 +0,0 @@
-#!/usr/bin/env bash
-
-test -e /tmp/xpmem.share && rm -f /tmp/xpmem.share
-test -e /tmp/xpmem.lock && rm -f /tmp/xpmem.lock
-
-# create TMP_SHARE_SIZE bytes defined in xpmem_test.h
-for i in `seq 0 31` ; do
-	echo -n 0 >> /tmp/xpmem.share
-done
-echo 0 > /tmp/xpmem.lock
-
-# Run the main test app
-mcexec $PWD/xpmem_master
-exit 0
-

test/issues/1259/multi_vmr_xpmem.c

@@ -1,187 +0,0 @@
-#include <unistd.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/shm.h>
-#include <xpmem.h>
-#include "util.h"
-
-#define DEBUG
-
-#define SHM_SIZE (1UL << 12)
-
-#define MAP_HUGE_SHIFT 26
-#define KEYWORD 0x12345678UL
-
-void usage(void)
-{
-	printf("Usage: multi_vmr_xpmem: <pgshift> <pgnum>\n");
-	printf("\tpgshift : pageshift of map area (Using MAP_HUGETLB)\n");
-	printf("\t            -1 means using small pagesize\n");
-	printf("\tpgnum   : number of page of map area\n");
-}
-
-
-void *mmap_flag(size_t mapsize, int page_shift)
-{
-	char *addr_mmap;
-	int flags = MAP_ANONYMOUS | MAP_PRIVATE;
-
-	if (page_shift >= 0) {
-		/* mean use MAP_HUGETLB */
-		flags |= MAP_HUGETLB | (page_shift << MAP_HUGE_SHIFT);
-	}
-
-	addr_mmap = mmap(0, mapsize * 2,
-			PROT_READ | PROT_WRITE,
-			flags, -1, 0);
-
-	/* Make sure that area before addr_map is available to
-	 * MAP_FIXED map
-	 */
-	return addr_mmap + mapsize;
-}
-
-int main(int argc, char **argv)
-{
-	void *mem, *mem_1, *mem_2;
-	int ret = 0;
-	pid_t pid;
-	int status;
-	key_t key = ftok(argv[0], 10);
-	void *shm;
-	int shmid;
-	xpmem_segid_t segid;
-	struct shmid_ds shmctl_buf;
-	int pgshift, pgnum;
-	size_t extr_size, pgsize, map_size;
-
-	if (argc < 3) {
-		printf("Err: Too few arguments\n");
-		usage();
-		return -1;
-	}
-
-	pgshift = atoi(argv[1]);
-	pgnum = atoi(argv[2]);
-	extr_size = getpagesize() * 3;
-	if (pgshift > 0) {
-		pgsize = (1UL << pgshift);
-	} else {
-		pgsize = getpagesize();
-	}
-
-	map_size = pgsize * pgnum;
-
-	shmid = shmget(key, SHM_SIZE, IPC_CREAT | 0660);
-	CHKANDJUMP(shmid == -1, EXIT_FAILURE, "shmget failed: %s\n",
-		   strerror(errno));
-
-	printf("EXPECT_PAGE_SIZE: 0x%lx\n", pgsize);
-	fflush(stdout);
-
-	pid = fork();
-	CHKANDJUMP(pid == -1, EXIT_FAILURE, "fork failed\n");
-	if (pid == 0) {
-		xpmem_apid_t apid;
-		struct xpmem_addr addr;
-		void *attach;
-
-		shm = shmat(shmid, NULL, 0);
-		CHKANDJUMP(shm == (void *)-1, EXIT_FAILURE,
-			   "shmat failed: %s\n", strerror(errno));
-
-		while ((segid = *(xpmem_segid_t *)shm) == 0) {
-		};
-		ret = shmdt(shm);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "shmdt failed\n");
-
-		apid = xpmem_get(segid, XPMEM_RDWR,
-				 XPMEM_PERMIT_MODE, NULL);
-		CHKANDJUMP(apid == -1, EXIT_FAILURE, "xpmem_get failed: %s\n",
-			   strerror(errno));
-
-		addr.apid = apid;
-		addr.offset = 0;
-		attach = xpmem_attach(addr, map_size + (extr_size * 2), NULL);
-
-		CHKANDJUMP(attach == (void *)-1, EXIT_FAILURE,
-			   "xpmem_attach failed: %s\n", strerror(errno));
-
-		printf("child: xpmem_attachment_addr: %lx - %lx\n",
-		       attach, attach + map_size + (extr_size * 2));
-		printf("child: xpmem_large: %lx\n", attach + extr_size);
-
-		*((unsigned long *)attach) = KEYWORD;
-		*((unsigned long *)(attach + extr_size)) = KEYWORD;
-		*((unsigned long *)(attach + extr_size * 2 + map_size
-			- sizeof(unsigned long *))) = KEYWORD;
-
-		ret = xpmem_detach(attach);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "xpmem_detach failed\n");
-
-		exit(0);
-	} else {
-		mem = mmap_flag(map_size, pgshift);
-		CHKANDJUMP(mem == MAP_FAILED, EXIT_FAILURE, "mmap failed\n");
-		mem_1 = mmap(mem - extr_size, extr_size,
-				PROT_READ | PROT_WRITE,
-				MAP_ANONYMOUS | MAP_PRIVATE | MAP_FIXED,
-				-1, 0);
-		mem_2 = mmap(mem + map_size, extr_size,
-				PROT_READ | PROT_WRITE,
-				MAP_ANONYMOUS | MAP_PRIVATE | MAP_FIXED,
-				 -1, 0);
-		if ((mem_1 + extr_size != mem) || (mem_2 != mem + map_size)) {
-			printf("vm_range is NOT contignuous!!\n");
-			exit(1);
-		}
-		printf("parent: anonymous_map_addr: %lx - %lx\n",
-		       mem_1, mem_2 + extr_size);
-
-		shm = shmat(shmid, NULL, 0);
-
-		CHKANDJUMP(shm == (void *)-1, EXIT_FAILURE,
-			   "shmat failed: %s\n", strerror(errno));
-
-		segid = xpmem_make(mem_1, map_size + (extr_size * 2),
-				XPMEM_PERMIT_MODE, (void *)0666);
-		CHKANDJUMP(segid == -1, EXIT_FAILURE,
-			   "xpmem_ioctl failed: %s\n", strerror(errno));
-
-		*(xpmem_segid_t *)shm = segid;
-
-		ret = waitpid(pid, &status, 0);
-printf("child exited\n");
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "waitpid failed\n");
-
-		NG(*(unsigned long *)mem_1 == KEYWORD,
-			"HEAD of xpmem area is INVALID. isn't shared?\n");
-		NG(*(unsigned long *)mem == KEYWORD,
-			"MIDDLE of xpmem area is INVALID. isn't shared?\n");
-		NG(*((unsigned long *)(mem_2 + extr_size
-				- sizeof(unsigned long *))) == KEYWORD,
-			"TAIL of xpmem area is INVALID. isn't shared?\n");
-		printf("xpmem area is shared: OK\n");
-
-		ret = shmctl(shmid, IPC_RMID, &shmctl_buf);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "shmctl failed\n");
-
-		ret = shmdt(shm);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "shmdt failed\n");
-
-		ret = xpmem_remove(segid);
-		CHKANDJUMP(ret == -1, EXIT_FAILURE, "xpmem_remove failed\n");
-	}
-
-	ret = 0;
- out:
-	return ret;
-}

test/issues/1259/util.h

@@ -1,28 +0,0 @@
-#ifndef __UTIL_H_INCLUDED__
-#define __UTIL_H_INCLUDED__
-
-#define CHKANDJUMP(cond, err, ...) do { \
-	if (cond) {			\
-		printf(__VA_ARGS__);   \
-		ret = err;		\
-		goto out;		\
-	}				\
-} while (0)
-
-#define _OKNG(verb, jump, cond, fmt, args...) do {	\
-	if (cond) {					\
-		if (verb)				\
-			printf("[ OK ] " fmt, ##args);	\
-	} else {					\
-		printf("[ NG ] " fmt, ##args);		\
-		if (jump)				\
-			goto out;			\
-	}						\
-} while (0)
-
-#define OKNG(args...) _OKNG(1, 1, ##args)
-#define NG(args...) _OKNG(0, 1, ##args)
-#define OKNGNOJUMP(args...) _OKNG(1, 0, ##args)
-
-#endif
-

test/issues/1259/util2.h

@@ -1,31 +0,0 @@
-#define CHKANDJUMP(cond, ...)	do {		\
-	if (cond) {				\
-		fprintf(stderr, " [NG] ");	\
-		fprintf(stderr, __VA_ARGS__);	\
-		fprintf(stderr, " failed\n");	\
-		goto fn_fail;			\
-	}					\
-} while (0)
-
-#define OKNG(cond, ...)		do {		\
-	if (cond) {				\
-		CHKANDJUMP(cond, __VA_ARGS__);	\
-	} else {				\
-		fprintf(stdout, " [OK] ");	\
-		fprintf(stdout, __VA_ARGS__);	\
-		fprintf(stdout, "\n");		\
-	}					\
-} while (0)
-
-
-#ifdef __aarch64__
-#define LARGE_PAGE_SHIFT 21
-#elif defined(__x86_64__)
-#define LARGE_PAGE_SHIFT 21
-#else
-#error "Non-compliant architecture."
-#endif
-
-#define MAP_HUGE_SHIFT 26
-#define SZ_MEM (2 * (1ULL << LARGE_PAGE_SHIFT))
-#define TEST_VAL 0x1129

test/issues/1259/x86_64_config

@@ -1,2 +0,0 @@
-PGSHIFT_LIST=(21 30)
-SMALL_PGSIZE=4096

test/issues/1259/x86_64_result.log

@@ -1,178 +0,0 @@
-*** C1259T01 start *******************************
-** xpmem_attach to Huge mapped memory range
-** end of range is aligned with Large page size
-pageshift: 21
-** [ OK ] seg_addr (2aaaab000000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (2aaaab000000) is allocated using large pages
-pageshift: 30
-** [ OK ] seg_addr (2aaac0000000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (2aaac0000000) is allocated using large pages
-*** C1259T01: PASSED
-
-*** C1259T02 start *******************************
-** xpmem_attach to Huge mapped memory range
-** end of range is NOT aligned with Large page size
-pageshift: 21
-** [ OK ] seg_addr (2aaaab000000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (2aaaab000000) is allocated using large pages
-pageshift: 30
-** [ OK ] seg_addr (2aaac0000000) is allocated until xpmem_attach
-** [ OK ] xpmem_addr (2aaac0000000) is allocated using large pages
-*** C1259T02: PASSED
-
-*** C1259T03 start *******************************
-** xpmem_attach to small mapped memory range
-pageshift: small page
-** [ OK ] xpmem_addr (2aaaaafee000) is allocated using small pages
-*** C1259T03: PASSED
-
-*** C1259T04 start *******************************
-** xpmem_attach to multi pagesize range
-pageshift: 21
-** [ OK ] xpmem_addr (2aaaab200000) is allocated using large pages
-*** C1259T04: PASSED
-
-*** C1259T05 start *******************************
-** xpmem testsuite
-XPMEM version = 26003
-
-==== test_base STARTS ====
-xpmem_proc1: mypid = 20070
-xpmem_proc1: sharing 16384 bytes
-xpmem_proc1: segid = 200004e66 at 0x2aaaaafee000
-
-xpmem_proc2: mypid = 20490
-xpmem_proc2: segid = 200004e66
-xpmem_proc2: attached at 0x2aaaaafee000
-xpmem_proc2: adding 1 to all elems
-
-xpmem_proc1: verifying data...done
-==== test_base PASSED ====
-
-==== test_two_attach STARTS ====
-xpmem_proc1: mypid = 20913
-xpmem_proc1: sharing 16384 bytes
-xpmem_proc1: segid = 2000051b1 at 0x2aaaaafee000
-
-xpmem_proc2: mypid = 21336
-xpmem_proc2: segid = 2000051b1
-xpmem_proc2: attached at 0x2aaaaafee000
-xpmem_proc2: attached at 0x2aaaaaff2000
-xpmem_proc2: adding 1 to all elems using 0x2aaaaafee000
-xpmem_proc2: adding 1 to all elems using 0x2aaaaaff2000
-
-xpmem_proc1: verifying data...done
-==== test_two_attach PASSED ====
-
-==== test_two_shares STARTS ====
-xpmem_proc1: mypid = 21758
-xpmem_proc1: sharing 2 segments, 16384 bytes each
-xpmem_proc1: segid[0] = 2000054fe at 0x2aaaaafee000
-xpmem_proc1: segid[1] = 4000054fe at 0x2aaaaaff2000
-
-xpmem_proc2: mypid = 22179
-xpmem_proc2: segid[0] = 2000054fe
-xpmem_proc2: segid[1] = 4000054fe
-xpmem_proc2: data[0] attached at 0x2aaaaafee000
-xpmem_proc2: data[1] attached at 0x2aaaaaff2000
-xpmem_proc2: adding 1 to all elems using 0x2aaaaafee000
-xpmem_proc2: adding 1 to all elems using 0x2aaaaaff2000
-
-xpmem_proc1: verifying data...done
-==== test_two_shares PASSED ====
-
-==== test_fork STARTS ====
-xpmem_proc1: mypid = 22599
-xpmem_proc1: sharing 16384 bytes
-xpmem_proc1: segid = 200005847 at 0x2aaaaafee000
-
-xpmem_proc2: mypid = 23022
-xpmem_proc2: segid = 200005847
-xpmem_proc2: attached at 0x2aaaaafef000
-xpmem_proc2: reading to pin pages
-xpmem_proc2: waiting for COW...
-
-xpmem_proc1: forking a child
-xpmem_proc1: adding 1 to all elems to induce COW
-
-xpmem_proc1: give control back to xpmem_proc2
-xpmem_child: hello from pid 23443
-
-
-xpmem_proc2: adding 1 to all elems
-
-xpmem_proc1: verifying data...done
-==== test_fork PASSED ====
-
-*** XTP_001 start ***
- [OK] xpmem_make
- [OK] xpmem_get
- [OK] xpmem_attach
- [OK] xpmem_detach
- [OK] xpmem_remove
-*** XTP_001 PASSED
-
-*** XTP_002 start ***
- [OK] xpmem_make in child
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] xpmem_remove in child
-*** XTP_002 PASSED
-
-*** XTP_003 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_003 PASSED
-
-*** XTP_004 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_004 PASSED
-
-*** XTP_005 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_005 PASSED
-
-*** XTP_006 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child failed (parent process exited already
-*** XTP_006 PASSED
-
-*** XTP_007 start ***
- [OK] xpmem_make failed (invalid address)
- [OK] xpmem_make succeed(do twice to same address)
-*** XTP_007 PASSED
-
-*** XTP_008 start ***
- [OK] xpmem_get in child failed (invalid segid)
- [OK] xpmem_get in child (do twice to same segid
-*** XTP_008 PASSED
-
-*** XTP_009 start ***
- [OK] xpmem_attach in childi failed (invalid apid)
- [OK] xpmem_attach in child succeed (do twice to same apid)
-*** XTP_009 PASSED
-
-*** XTP_010 start ***
- [OK] xpmem_detach in child succeed (invalid address)
- [OK] xpmem_detach in child succeed (do twice to same address)
-*** XTP_010 PASSED
-
-*** XTP_011 start ***
- [OK] xpmem_remove failed (invalid segid)
- [OK] xpmem_remove failed (do twice to same segid)
-*** XTP_011 PASSED
-

test/issues/1330/README

@@ -1,22 +0,0 @@
-【Issue#1330 動作確認】
-□ テスト内容
-1. xpmemの基本操作の確認
-mckernel/test/xpmemのテストを実行し、xpmemの基本操作が
-正常に動作することを確認する
-テスト内容の詳細はmckernel/test/xpmem/README を参照
-
-□ 実行手順
-1. mckernel/test/xpmem に移動する
-1. xpmemのインストールディレクトリをMakefileとgo_test.sh中のXPMEM_DIRに記載する
-2. xpmemのビルドディレクトリをgo_test.sh中のXPMEM_BUILD_DIRに記載する
-3. 下記の手順でテストを実行する
-$ 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/1330/aarch64_result.log

@@ -1,151 +0,0 @@
-*** XPMEM_TESTSUITE start *******************************
-XPMEM version = 26003
-
-==== test_base STARTS ====
-xpmem_proc1: mypid = 3962
-xpmem_proc1: sharing 262144 bytes
-xpmem_proc1: segid = 200000f7a at 0x100000210000
-
-xpmem_proc2: mypid = 4161
-xpmem_proc2: segid = 200000f7a
-xpmem_proc2: attached at 0x100000210000
-xpmem_proc2: adding 1 to all elems
-
-xpmem_proc1: verifying data...done
-==== test_base PASSED ====
-
-==== test_two_attach STARTS ====
-xpmem_proc1: mypid = 4472
-xpmem_proc1: sharing 262144 bytes
-xpmem_proc1: segid = 200001178 at 0x100000210000
-
-xpmem_proc2: mypid = 4670
-xpmem_proc2: segid = 200001178
-xpmem_proc2: attached at 0x100000210000
-xpmem_proc2: attached at 0x100000250000
-xpmem_proc2: adding 1 to all elems using 0x100000210000
-xpmem_proc2: adding 1 to all elems using 0x100000250000
-
-xpmem_proc1: verifying data...done
-==== test_two_attach PASSED ====
-
-==== test_two_shares STARTS ====
-xpmem_proc1: mypid = 4871
-xpmem_proc1: sharing 2 segments, 262144 bytes each
-xpmem_proc1: segid[0] = 200001307 at 0x100000210000
-xpmem_proc1: segid[1] = 400001307 at 0x100000250000
-
-xpmem_proc2: mypid = 5068
-xpmem_proc2: segid[0] = 200001307
-xpmem_proc2: segid[1] = 400001307
-xpmem_proc2: data[0] attached at 0x100000210000
-xpmem_proc2: data[1] attached at 0x100000250000
-xpmem_proc2: adding 1 to all elems using 0x100000210000
-xpmem_proc2: adding 1 to all elems using 0x100000250000
-
-xpmem_proc1: verifying data...done
-==== test_two_shares PASSED ====
-
-==== test_fork STARTS ====
-xpmem_proc1: mypid = 5271
-xpmem_proc1: sharing 262144 bytes
-xpmem_proc1: segid = 200001497 at 0x100000210000
-
-xpmem_proc2: mypid = 5468
-xpmem_proc2: segid = 200001497
-xpmem_proc2: attached at 0x100000220000
-xpmem_proc2: reading to pin pages
-xpmem_proc2: waiting for COW...
-
-xpmem_proc1: forking a child
-xpmem_proc1: adding 1 to all elems to induce COW
-
-xpmem_child: hello from pid 5789
-
-xpmem_proc1: give control back to xpmem_proc2
-
-xpmem_proc2: adding 1 to all elems
-
-xpmem_proc1: verifying data...done
-==== test_fork PASSED ====
-
-*** XTP_001 start ***
- [OK] xpmem_make
- [OK] xpmem_get
- [OK] xpmem_attach
- [OK] xpmem_detach
- [OK] xpmem_remove
-*** XTP_001 PASSED
-
-*** XTP_002 start ***
- [OK] xpmem_make in child
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] xpmem_remove in child
-*** XTP_002 PASSED
-
-*** XTP_003 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_003 PASSED
-
-*** XTP_004 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_004 PASSED
-
-*** XTP_005 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_005 PASSED
-
-*** XTP_006 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child failed (parent process exited already
-*** XTP_006 PASSED
-
-*** XTP_007 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_007 PASSED
-
-*** XTP_901 start ***
- [OK] xpmem_make failed (invalid address)
- [OK] xpmem_make succeed(do twice to same address)
-*** XTP_901 PASSED
-
-*** XTP_902 start ***
- [OK] xpmem_get in child failed (invalid segid)
- [OK] xpmem_get in child (do twice to same segid
-*** XTP_902 PASSED
-
-*** XTP_903 start ***
- [OK] xpmem_attach in childi failed (invalid apid)
- [OK] xpmem_attach in child succeed (do twice to same apid)
-*** XTP_903 PASSED
-
-*** XTP_904 start ***
- [OK] xpmem_detach in child succeed (invalid address)
- [OK] xpmem_detach in child succeed (do twice to same address)
-*** XTP_904 PASSED
-
-*** XTP_905 start ***
- [OK] xpmem_remove failed (invalid segid)
- [OK] xpmem_remove failed (do twice to same segid)
-*** XTP_905 PASSED
-

test/issues/1330/x86_64_result.log

@@ -1,151 +0,0 @@
-*** XPMEM_TESTSUITE start *******************************
-XPMEM version = 26003
-
-==== test_base STARTS ====
-xpmem_proc1: mypid = 4598
-xpmem_proc1: sharing 16384 bytes
-xpmem_proc1: segid = 2000011f6 at 0x2aaaaafee000
-
-xpmem_proc2: mypid = 5018
-xpmem_proc2: segid = 2000011f6
-xpmem_proc2: attached at 0x2aaaaafee000
-xpmem_proc2: adding 1 to all elems
-
-xpmem_proc1: verifying data...done
-==== test_base PASSED ====
-
-==== test_two_attach STARTS ====
-xpmem_proc1: mypid = 5438
-xpmem_proc1: sharing 16384 bytes
-xpmem_proc1: segid = 20000153e at 0x2aaaaafee000
-
-xpmem_proc2: mypid = 5858
-xpmem_proc2: segid = 20000153e
-xpmem_proc2: attached at 0x2aaaaafee000
-xpmem_proc2: attached at 0x2aaaaaff2000
-xpmem_proc2: adding 1 to all elems using 0x2aaaaafee000
-xpmem_proc2: adding 1 to all elems using 0x2aaaaaff2000
-
-xpmem_proc1: verifying data...done
-==== test_two_attach PASSED ====
-
-==== test_two_shares STARTS ====
-xpmem_proc1: mypid = 6278
-xpmem_proc1: sharing 2 segments, 16384 bytes each
-xpmem_proc1: segid[0] = 200001886 at 0x2aaaaafee000
-xpmem_proc1: segid[1] = 400001886 at 0x2aaaaaff2000
-
-xpmem_proc2: mypid = 6698
-xpmem_proc2: segid[0] = 200001886
-xpmem_proc2: segid[1] = 400001886
-xpmem_proc2: data[0] attached at 0x2aaaaafee000
-xpmem_proc2: data[1] attached at 0x2aaaaaff2000
-xpmem_proc2: adding 1 to all elems using 0x2aaaaafee000
-xpmem_proc2: adding 1 to all elems using 0x2aaaaaff2000
-
-xpmem_proc1: verifying data...done
-==== test_two_shares PASSED ====
-
-==== test_fork STARTS ====
-xpmem_proc1: mypid = 7118
-xpmem_proc1: sharing 16384 bytes
-xpmem_proc1: segid = 200001bce at 0x2aaaaafee000
-
-xpmem_proc2: mypid = 7538
-xpmem_proc2: segid = 200001bce
-xpmem_proc2: attached at 0x2aaaaafef000
-xpmem_proc2: reading to pin pages
-xpmem_proc2: waiting for COW...
-
-xpmem_proc1: forking a child
-xpmem_proc1: adding 1 to all elems to induce COW
-
-xpmem_proc1: give control back to xpmem_proc2
-
-xpmem_child: hello from pid 7958
-
-xpmem_proc2: adding 1 to all elems
-
-xpmem_proc1: verifying data...done
-==== test_fork PASSED ====
-
-*** XTP_001 start ***
- [OK] xpmem_make
- [OK] xpmem_get
- [OK] xpmem_attach
- [OK] xpmem_detach
- [OK] xpmem_remove
-*** XTP_001 PASSED
-
-*** XTP_002 start ***
- [OK] xpmem_make in child
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] xpmem_remove in child
-*** XTP_002 PASSED
-
-*** XTP_003 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_003 PASSED
-
-*** XTP_004 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] xpmem_detach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_004 PASSED
-
-*** XTP_005 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_005 PASSED
-
-*** XTP_006 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child failed (parent process exited already
-*** XTP_006 PASSED
-
-*** XTP_007 start ***
- [OK] xpmem_make
- [OK] xpmem_get in child
- [OK] xpmem_attach in child
- [OK] validate TEST_VAL
- [OK] xpmem_remove
-*** XTP_007 PASSED
-
-*** XTP_901 start ***
- [OK] xpmem_make failed (invalid address)
- [OK] xpmem_make succeed(do twice to same address)
-*** XTP_901 PASSED
-
-*** XTP_902 start ***
- [OK] xpmem_get in child failed (invalid segid)
- [OK] xpmem_get in child (do twice to same segid
-*** XTP_902 PASSED
-
-*** XTP_903 start ***
- [OK] xpmem_attach in childi failed (invalid apid)
- [OK] xpmem_attach in child succeed (do twice to same apid)
-*** XTP_903 PASSED
-
-*** XTP_904 start ***
- [OK] xpmem_detach in child succeed (invalid address)
- [OK] xpmem_detach in child succeed (do twice to same address)
-*** XTP_904 PASSED
-
-*** XTP_905 start ***
- [OK] xpmem_remove failed (invalid segid)
- [OK] xpmem_remove failed (do twice to same segid)
-*** XTP_905 PASSED
-

test/issues/1377/C1377.sh

@@ -0,0 +1,56 @@
+#!/bin/sh
+USELTP=1
+USEOSTEST=0
+
+. ../../common.sh
+
+################################################################################
+if [ ! -f "$LTPBIN/dirtyc0w" ]; then
+	echo BAD environment: LTP is too old >&2
+	exit 1
+fi
+
+echo "*** C1377T01 start"
+ng=0
+ok=0
+tp=dirtyc0w
+for ((i=0; i<20; i++)); do
+	for ((j=0; j<50; j++)); do
+		sudo PATH=$PATH:$LTPBIN $MCEXEC $LTPBIN/$tp > $tp.txt 2>&1
+		wok=`grep PASS $tp.txt | wc -l`
+		wng=`grep FAIL $tp.txt | wc -l`
+		if [ $wng != 0 ]; then
+			echo -n '*'
+			ng=`expr $ng + 1`
+		elif [ $wok == 0 ]; then
+			echo -n '?'
+		else
+			echo -n '.'
+			ok=`expr $ok + 1`
+		fi
+	done
+	echo
+done
+
+if [ $ng != 0 ]; then
+	echo "*** C1377T01 FAILED ok: $ok, ng: $ng"
+else
+	echo "*** C1377T01 PASS ok: $ok"
+fi
+
+
+
+for i in wait02:02 wait401:03 waitpid01:04 waitpid02:05 waitpid04:06 \
+	 waitpid05:07 waitpid06:08 waitpid07:09 waitpid08:10 waitpid09:11 \
+	 waitpid10:12 waitpid11:13 waitpid12:14 waitpid13:15; do
+	tp=`echo $i|sed 's/:.*//'`
+	id=`echo $i|sed 's/.*://'`
+	sudo PATH=$PATH:$LTPBIN $MCEXEC $LTPBIN/$tp 2>&1 | tee $tp.txt
+	ok=`grep PASS $tp.txt | wc -l`
+	ng=`grep FAIL $tp.txt | wc -l`
+	if [ $ng = 0 ]; then
+		echo "*** C1377T$id: $tp PASS ($ok)"
+	else
+		echo "*** C1377T$id: $tp FAIL (ok=$ok ng=$ng)"
+	fi
+done

test/issues/1377/C1377_arm64.txt

@@ -0,0 +1,179 @@
+Script started on Tue 16 Jun 2020 02:42:48 PM JST
+[shirasawa@apollo04 1377]$ make test
+sh ./C1377.sh
+mcstop+release.sh ... done
+mcreboot.sh -c 2-31 -m 2G@0,2G@1 -O ... done
+*** C1377T01 start
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+*** C1377T01 PASS ok: 1000
+wait02      1  TPASS  :  wait(&status) returned 78343
+*** C1377T02: wait02 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+wait401.c:55: PASS: waitpid() returned correct pid 78457
+wait401.c:64: PASS: WIFEXITED() is set in status
+wait401.c:69: PASS: WEXITSTATUS() == 0
+
+Summary:
+passed   3
+failed   0
+skipped  0
+warnings 0
+*** C1377T03: wait401 PASS (3)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid01.c:51: PASS: waitpid() returned correct pid 78571
+waitpid01.c:60: PASS: WIFSIGNALED() set in status
+waitpid01.c:68: PASS: WTERMSIG() == SIGALRM
+
+Summary:
+passed   3
+failed   0
+skipped  0
+warnings 0
+*** C1377T04: waitpid01 PASS (3)
+waitpid02    1  TPASS  :  received expected pid
+waitpid02    2  TPASS  :  received expected signal
+waitpid02    3  TPASS  :  received expected exit value
+*** C1377T05: waitpid02 PASS (3)
+waitpid04    1  TPASS  :  condition 1 test passed
+waitpid04    2  TPASS  :  condition 2 test passed
+waitpid04    3  TPASS  :  condition 3 test passed
+*** C1377T06: waitpid04 PASS (3)
+waitpid05    1  TPASS  :  received expected pid.
+waitpid05    2  TPASS  :  received expected exit number.
+waitpid05    3  TPASS  :  received expected pid.
+waitpid05    4  TPASS  :  received expected exit number.
+waitpid05    5  TPASS  :  received expected pid.
+waitpid05    6  TPASS  :  received expected exit number.
+waitpid05    7  TPASS  :  received expected pid.
+waitpid05    8  TPASS  :  received expected exit number.
+waitpid05    9  TPASS  :  received expected pid.
+waitpid05   10  TPASS  :  received expected exit number.
+waitpid05   11  TPASS  :  received expected pid.
+waitpid05   12  TPASS  :  received expected exit number.
+waitpid05   13  TPASS  :  received expected pid.
+waitpid05   14  TPASS  :  received expected exit number.
+waitpid05   15  TPASS  :  received expected pid.
+waitpid05   16  TPASS  :  received expected exit number.
+waitpid05   17  TPASS  :  received expected pid.
+waitpid05   18  TPASS  :  received expected exit number.
+waitpid05   19  TPASS  :  received expected pid.
+waitpid05   20  TPASS  :  received expected exit number.
+waitpid05   21  TPASS  :  received expected pid.
+waitpid05   22  TPASS  :  received expected exit number.
+waitpid05   23  TPASS  :  received expected pid.
+waitpid05   24  TPASS  :  received expected exit number.
+*** C1377T07: waitpid05 PASS (24)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid06.c:54: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T08: waitpid06 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid07.c:59: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T09: waitpid07 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid_common.h:142: INFO: Sending SIGCONT to 80073
+waitpid_common.h:142: INFO: Sending SIGCONT to 80105
+waitpid_common.h:142: INFO: Sending SIGCONT to 80137
+waitpid_common.h:142: INFO: Sending SIGCONT to 80169
+waitpid_common.h:142: INFO: Sending SIGCONT to 80201
+waitpid_common.h:142: INFO: Sending SIGCONT to 80233
+waitpid_common.h:142: INFO: Sending SIGCONT to 80266
+waitpid_common.h:142: INFO: Sending SIGCONT to 80298
+waitpid08.c:62: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T10: waitpid08 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid09.c:83: PASS: waitpid(pid, WNOHANG) = 0 for a running child
+waitpid09.c:123: PASS: waitpid(pid, WNOHANG) = pid for an exited child
+waitpid09.c:143: PASS: waitpid(-1, 0) = -1 with ECHILD if no children
+waitpid09.c:162: PASS: waitpid(-1, WNOHANG) = -1 with ECHILD if no children
+
+Summary:
+passed   4
+failed   0
+skipped  0
+warnings 0
+*** C1377T11: waitpid09 PASS (4)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid10.c:62: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T12: waitpid10 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid11.c:60: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T13: waitpid11 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid12.c:70: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T14: waitpid12 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid_common.h:142: INFO: Sending SIGCONT to 85038
+waitpid_common.h:142: INFO: Sending SIGCONT to 85070
+waitpid_common.h:142: INFO: Sending SIGCONT to 85102
+waitpid_common.h:142: INFO: Sending SIGCONT to 85134
+waitpid_common.h:142: INFO: Sending SIGCONT to 84910
+waitpid_common.h:142: INFO: Sending SIGCONT to 84942
+waitpid_common.h:142: INFO: Sending SIGCONT to 84974
+waitpid_common.h:142: INFO: Sending SIGCONT to 85006
+waitpid13.c:70: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T15: waitpid13 PASS (1)
+[shirasawa@apollo04 1377]$ exit
+exit
+
+Script done on Tue 16 Jun 2020 03:02:48 PM JST

test/issues/1377/C1377_x86_64.txt

@@ -0,0 +1,179 @@
+Script started on Tue Jun 16 15:36:06 2020
+bash-4.2$ make test
+sh ./C1377.sh
+mcstop+release.sh ... done
+mcreboot.sh -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 ... done
+*** C1377T01 start
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+..................................................
+*** C1377T01 PASS ok: 1000
+wait02      1  TPASS  :  wait(&status) returned 32240
+*** C1377T02: wait02 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+wait401.c:55: PASS: waitpid() returned correct pid 32347
+wait401.c:64: PASS: WIFEXITED() is set in status
+wait401.c:69: PASS: WEXITSTATUS() == 0
+
+Summary:
+passed   3
+failed   0
+skipped  0
+warnings 0
+*** C1377T03: wait401 PASS (3)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid01.c:51: PASS: waitpid() returned correct pid 32454
+waitpid01.c:60: PASS: WIFSIGNALED() set in status
+waitpid01.c:68: PASS: WTERMSIG() == SIGALRM
+
+Summary:
+passed   3
+failed   0
+skipped  0
+warnings 0
+*** C1377T04: waitpid01 PASS (3)
+waitpid02    1  TPASS  :  received expected pid
+waitpid02    2  TPASS  :  received expected signal
+waitpid02    3  TPASS  :  received expected exit value
+*** C1377T05: waitpid02 PASS (3)
+waitpid04    1  TPASS  :  condition 1 test passed
+waitpid04    2  TPASS  :  condition 2 test passed
+waitpid04    3  TPASS  :  condition 3 test passed
+*** C1377T06: waitpid04 PASS (3)
+waitpid05    1  TPASS  :  received expected pid.
+waitpid05    2  TPASS  :  received expected exit number.
+waitpid05    3  TPASS  :  received expected pid.
+waitpid05    4  TPASS  :  received expected exit number.
+waitpid05    5  TPASS  :  received expected pid.
+waitpid05    6  TPASS  :  received expected exit number.
+waitpid05    7  TPASS  :  received expected pid.
+waitpid05    8  TPASS  :  received expected exit number.
+waitpid05    9  TPASS  :  received expected pid.
+waitpid05   10  TPASS  :  received expected exit number.
+waitpid05   11  TPASS  :  received expected pid.
+waitpid05   12  TPASS  :  received expected exit number.
+waitpid05   13  TPASS  :  received expected pid.
+waitpid05   14  TPASS  :  received expected exit number.
+waitpid05   15  TPASS  :  received expected pid.
+waitpid05   16  TPASS  :  received expected exit number.
+waitpid05   17  TPASS  :  received expected pid.
+waitpid05   18  TPASS  :  received expected exit number.
+waitpid05   19  TPASS  :  received expected pid.
+waitpid05   20  TPASS  :  received expected exit number.
+waitpid05   21  TPASS  :  received expected pid.
+waitpid05   22  TPASS  :  received expected exit number.
+waitpid05   23  TPASS  :  received expected pid.
+waitpid05   24  TPASS  :  received expected exit number.
+*** C1377T07: waitpid05 PASS (24)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid06.c:54: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T08: waitpid06 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid07.c:59: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T09: waitpid07 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid_common.h:142: INFO: Sending SIGCONT to 1524
+waitpid_common.h:142: INFO: Sending SIGCONT to 1554
+waitpid_common.h:142: INFO: Sending SIGCONT to 1584
+waitpid_common.h:142: INFO: Sending SIGCONT to 1614
+waitpid_common.h:142: INFO: Sending SIGCONT to 1644
+waitpid_common.h:142: INFO: Sending SIGCONT to 1674
+waitpid_common.h:142: INFO: Sending SIGCONT to 1704
+waitpid_common.h:142: INFO: Sending SIGCONT to 1734
+waitpid08.c:62: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T10: waitpid08 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid09.c:83: PASS: waitpid(pid, WNOHANG) = 0 for a running child
+waitpid09.c:123: PASS: waitpid(pid, WNOHANG) = pid for an exited child
+waitpid09.c:143: PASS: waitpid(-1, 0) = -1 with ECHILD if no children
+waitpid09.c:162: PASS: waitpid(-1, WNOHANG) = -1 with ECHILD if no children
+
+Summary:
+passed   4
+failed   0
+skipped  0
+warnings 0
+*** C1377T11: waitpid09 PASS (4)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid10.c:62: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T12: waitpid10 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid11.c:60: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T13: waitpid11 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid12.c:70: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T14: waitpid12 PASS (1)
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+waitpid_common.h:142: INFO: Sending SIGCONT to 6219
+waitpid_common.h:142: INFO: Sending SIGCONT to 6249
+waitpid_common.h:142: INFO: Sending SIGCONT to 6279
+waitpid_common.h:142: INFO: Sending SIGCONT to 6309
+waitpid_common.h:142: INFO: Sending SIGCONT to 6099
+waitpid_common.h:142: INFO: Sending SIGCONT to 6129
+waitpid_common.h:142: INFO: Sending SIGCONT to 6159
+waitpid_common.h:142: INFO: Sending SIGCONT to 6189
+waitpid13.c:70: PASS: Test PASSED
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1377T15: waitpid13 PASS (1)
+bash-4.2$ exit
+exit
+
+Script done on Tue Jun 16 15:55:52 2020

test/issues/1377/Makefile

@@ -0,0 +1,9 @@
+CC = gcc
+
+all::
+
+test:: all
+	sh ./C1377.sh
+
+clean::
+	rm -f $(TARGET) *.o

test/issues/1377/README

@@ -0,0 +1,38 @@
+【Issue#1377 動作確認】
+□ テスト内容
+1. Issue 指摘事項の再現確認
+   現象はタイミングに依存し、ARM64環境でdirtyc0wを繰り返し実行した際に、
+   数回〜数十回に１回程度発生する。
+   問題の発生はタイミングに依存するため、dirtyc0wを十分な回数繰り返し
+   実行し、問題が再発しないことを確認する。
+
+C1377T01 dirtyc0wを1000回連続実行し、全てPASSすることを確認
+
+2. LTP を用いて既存処理に影響しないことを確認
+   プロセスの終了ステータスを変更したため、修正が既存処理に影響しないことを
+   wait系システムコールのLTPを用いて確認する。
+C1377T02 wait02: waitの基本機能の確認
+C1377T03 wait401: wait4の基本機能の確認
+C1377T04 waitpid01: waitpidの基本機能の確認
+C1377T05 waitpid02: waitpidの基本機能の確認
+C1377T06 waitpid04: waitpidの基本機能の確認
+C1377T07 waitpid05: waitpidの基本機能の確認
+C1377T08 waitpid06: waitpidの基本機能の確認
+C1377T09 waitpid07: waitpidの基本機能の確認
+C1377T10 waitpid08: waitpidの基本機能の確認
+C1377T11 waitpid09: waitpidの基本機能の確認
+C1377T12 waitpid10: waitpidの基本機能の確認
+C1377T13 waitpid11: waitpidの基本機能の確認
+C1377T14 waitpid12: waitpidの基本機能の確認
+C1377T15 waitpid13: waitpidの基本機能の確認
+
+□ 実行手順
+$ make test
+
+McKernelのインストール先や LTP の配置場所は、$HOME/.mck_test_config を
+参照する。.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、適宜編集すること。
+
+□ 実行結果
+C1377_x86_64.txt(x86_64実行結果)、C1377_arm64.txt(arm64実行結果)参照。
+全ての項目が PASS していることを確認。

test/issues/1380/C1380.sh

@@ -0,0 +1,26 @@
+#/bin/sh
+
+USELTP=1
+USEOSTEST=0
+
+. ../../common.sh
+
+issue="1380"
+tid=01
+
+for tp in tgkill03 tgkill01 tkill01 tkill02
+do
+	tname=`printf "C${issue}T%02d" ${tid}`
+	echo "*** ${tname} start *******************************"
+	sudo $MCEXEC $LTPBIN/$tp 2>&1 | tee $tp.txt
+	ok=`grep PASS $tp.txt | wc -l`
+	ng=`grep FAIL $tp.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/1380/Makefile

@@ -6,6 +6,6 @@ TARGET=
 all: $(TARGET)
 
 test: all
-	./C929.sh
+	./C1380.sh
 clean:
 	rm -f $(TARGET) *.o *.txt

test/issues/1380/README

@@ -0,0 +1,21 @@
+【Issue#1324 動作確認】
+□ テスト内容
+1. 以下のLTPを用いてIssueで報告された症状が発生しないことを確認
+  - tgkill03
+
+3. 以下のLTPを用いて既存機能に影響が無いことを確認
+  - tgkill01
+  - tkill01
+  - tkill02
+
+□ 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/1380/aarch64_result.log

@@ -0,0 +1,37 @@
+*** C1380T01 start *******************************
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+tgkill03.c:92: PASS: Invalid tgid failed as expected: EINVAL
+tgkill03.c:92: PASS: Invalid tid failed as expected: EINVAL
+tgkill03.c:92: PASS: Invalid signal failed as expected: EINVAL
+tgkill03.c:92: PASS: Defunct tid failed as expected: ESRCH
+tgkill03.c:92: PASS: Defunct tgid failed as expected: ESRCH
+tgkill03.c:99: PASS: Valid tgkill call succeeded
+
+Summary:
+passed   6
+failed   0
+skipped  0
+warnings 0
+*** C1380T01 PASSED (6)
+
+*** C1380T02 start *******************************
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+tgkill01.c:99: PASS: SIGUSR1 delivered to correct threads
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1380T02 PASSED (1)
+
+*** C1380T03 start *******************************
+tkill01     1  TPASS  :  tkill call succeeded
+tkill01     2  TPASS  :  tkill call succeeded
+*** C1380T03 PASSED (2)
+
+*** C1380T04 start *******************************
+tkill02     1  TPASS  :  tkill(-1, SIGUSR1) failed as expected: TEST_ERRNO=EINVAL(22): Invalid argument
+tkill02     2  TPASS  :  tkill(29822, SIGUSR1) failed as expected: TEST_ERRNO=ESRCH(3): No such process
+*** C1380T04 PASSED (2)
+

test/issues/1380/x86_64_result.log

@@ -0,0 +1,37 @@
+*** C1380T01 start *******************************
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+tgkill03.c:92: PASS: Invalid tgid failed as expected: EINVAL
+tgkill03.c:92: PASS: Invalid tid failed as expected: EINVAL
+tgkill03.c:92: PASS: Invalid signal failed as expected: EINVAL
+tgkill03.c:92: PASS: Defunct tid failed as expected: ESRCH
+tgkill03.c:92: PASS: Defunct tgid failed as expected: ESRCH
+tgkill03.c:99: PASS: Valid tgkill call succeeded
+
+Summary:
+passed   6
+failed   0
+skipped  0
+warnings 0
+*** C1380T01 PASSED (6)
+
+*** C1380T02 start *******************************
+tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s
+tgkill01.c:99: PASS: SIGUSR1 delivered to correct threads
+
+Summary:
+passed   1
+failed   0
+skipped  0
+warnings 0
+*** C1380T02 PASSED (1)
+
+*** C1380T03 start *******************************
+tkill01     1  TPASS  :  tkill call succeeded
+tkill01     2  TPASS  :  tkill call succeeded
+*** C1380T03 PASSED (2)
+
+*** C1380T04 start *******************************
+tkill02     1  TPASS  :  tkill(-1, SIGUSR1) failed as expected: TEST_ERRNO=EINVAL(22): Invalid argument
+tkill02     2  TPASS  :  tkill(32768, SIGUSR1) failed as expected: TEST_ERRNO=ESRCH(3): No such process
+*** C1380T04 PASSED (2)
+

test/issues/1399/C1399.sh

@@ -0,0 +1,41 @@
+#/bin/sh
+
+USELTP=1
+USEOSTEST=0
+
+. ../../common.sh
+
+issue="1399"
+tid=01
+
+for tno in 01
+do
+	tname=`printf "C${issue}T%02d" ${tid}`
+	echo "*** ${tname} start *******************************"
+	sudo ${MCEXEC} ./C1399T${tno}
+
+	if [ $? -eq 0 ]; then
+		echo "*** ${tname} PASSED ******************************"
+	else
+		echo "*** ${tname} FAILED ******************************"
+	fi
+	let tid++
+	echo ""
+done
+
+for tp in set_robust_list01 get_robust_list01
+do
+	tname=`printf "C${issue}T%02d" ${tid}`
+	echo "*** ${tname} start *******************************"
+	sudo $MCEXEC $LTPBIN/$tp 2>&1 | tee $tp.txt
+	ok=`grep PASS $tp.txt | wc -l`
+	ng=`grep FAIL $tp.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/1399/C1399T01.c

@@ -0,0 +1,36 @@
+#include <stdio.h>
+#include <unistd.h>
+#include <linux/futex.h>
+#include <syscall.h>
+#include <errno.h>
+
+int main(int argc, void *argv[])
+{
+	struct robust_list_head rlh;
+	size_t len = sizeof(struct robust_list_head);
+	int rc = 0, ret = 0;
+
+	errno = 0;
+	rc = syscall(__NR_set_robust_list, &rlh, len + 1);
+	if (rc == -1 && errno == EINVAL) {
+		printf("[OK] invalid len (1 greater than correct): EINVAL\n");
+	} else {
+		printf("[NG] invalid len (1 greater than correct): Succeed\n");
+		ret = -1;
+		goto out;
+	}
+
+	errno = 0;
+	rc = syscall(__NR_set_robust_list, &rlh, len - 1);
+	if (rc == -1 && errno == EINVAL) {
+		printf("[OK] invalid len (1 less than correct): EINVAL\n");
+	} else {
+		printf("[NG] invalid len (1 less than correct): Succeed\n");
+		ret = -1;
+		goto out;
+	}
+
+out:
+	return ret;
+}
+

test/issues/1399/Makefile

@@ -0,0 +1,11 @@
+CFLAGS=-g
+LDFLAGS=
+
+TARGET=C1399T01
+
+all: $(TARGET)
+
+test: all
+	./C1399.sh
+clean:
+	rm -f $(TARGET) *.o *.txt

test/issues/1399/README

@@ -0,0 +1,24 @@
+【Issue#1399 動作確認】
+□ テスト内容
+1. 下記のテストプログラムを実行し、症状が発生しないことを確認する
+C1399T01:
+  1. struct robust_list_head のサイズよりも1大きい値をサイズとして指定して
+    set_robsut_list を呼び出し、EINVALで失敗することを確認
+  2. struct robust_list_head のサイズよりも1小さい値をサイズとして指定して
+    set_robsut_list を呼び出し、EINVALで失敗することを確認
+
+2. 以下のLTPを用いて既存のrobust_list機能に影響が無いことを確認
+  - set_robst_list01
+  - get_robst_list01
+
+□ 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/1399/aarch64_result.log

@@ -0,0 +1,18 @@
+*** C1399T01 start *******************************
+[OK] invalid len (1 greater than correct): EINVAL
+[OK] invalid len (1 less than correct): EINVAL
+*** C1399T01 PASSED ******************************
+
+*** C1399T02 start *******************************
+set_robust_list01    1  TPASS  :  set_robust_list: retval = -1 (expected -1), errno = 22 (expected 22)
+set_robust_list01    2  TPASS  :  set_robust_list: retval = 0 (expected 0), errno = 0 (expected 0)
+*** C1399T02 PASSED (2)
+
+*** C1399T03 start *******************************
+get_robust_list01    1  TPASS  :  get_robust_list failed as expected with EFAULT
+get_robust_list01    2  TPASS  :  get_robust_list failed as expected with EFAULT
+get_robust_list01    3  TPASS  :  get_robust_list failed as expected with ESRCH
+get_robust_list01    4  TPASS  :  get_robust_list succeeded
+get_robust_list01    5  TPASS  :  get_robust_list failed as expected with EPERM
+*** C1399T03 PASSED (5)
+

test/issues/1399/x86_64_result.log

@@ -0,0 +1,18 @@
+*** C1399T01 start *******************************
+[OK] invalid len (1 greater than correct): EINVAL
+[OK] invalid len (1 less than correct): EINVAL
+*** C1399T01 PASSED ******************************
+
+*** C1399T02 start *******************************
+set_robust_list01    1  TPASS  :  set_robust_list: retval = -1 (expected -1), errno = 22 (expected 22)
+set_robust_list01    2  TPASS  :  set_robust_list: retval = 0 (expected 0), errno = 0 (expected 0)
+*** C1399T02 PASSED (2)
+
+*** C1399T03 start *******************************
+get_robust_list01    1  TPASS  :  get_robust_list failed as expected with EFAULT
+get_robust_list01    2  TPASS  :  get_robust_list failed as expected with EFAULT
+get_robust_list01    3  TPASS  :  get_robust_list failed as expected with ESRCH
+get_robust_list01    4  TPASS  :  get_robust_list succeeded
+get_robust_list01    5  TPASS  :  get_robust_list failed as expected with EPERM
+*** C1399T03 PASSED (5)
+

test/issues/1425/C1425.sh

@@ -0,0 +1,43 @@
+#/bin/sh
+
+USELTP=1
+USEOSTEST=0
+
+. ../../common.sh
+
+issue="1425"
+tid=01
+
+STOPSIG_LIST="TSTP TTIN TTOU"
+
+for signame in ${STOPSIG_LIST}
+do
+	tname=`printf "C${issue}T%02d" ${tid}`
+	echo "*** ${tname} start *******************************"
+	sh ./check_stopsig.sh ${MCEXEC} ${signame}
+
+	if [ $? -eq 0 ]; then
+		echo "*** ${tname} PASSED ******************************"
+	else
+		echo "*** ${tname} FAILED ******************************"
+	fi
+	let tid++
+	echo ""
+done
+
+for tp in kill01 kill02 kill06 kill07 kill08 kill09 signal01 signal03 signal04 signal05
+do
+	tname=`printf "C${issue}T%02d" ${tid}`
+	echo "*** ${tname} start *******************************"
+	sudo $MCEXEC $LTPBIN/$tp 2>&1 | tee $tp.txt
+	ok=`grep PASS $tp.txt | wc -l`
+	ng=`grep FAIL $tp.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/1425/Makefile

@@ -0,0 +1,11 @@
+CFLAGS=-g
+LDFLAGS=
+
+TARGET=
+
+all: $(TARGET)
+
+test: all
+	sh ./C1425.sh
+clean:
+	rm -f $(TARGET) *.o *.txt