The second (final) part of v4.0

.github/workflows/actuated-aarch64-test.yaml

@@ -0,0 +1,52 @@
+name: Actuated aarch64 test
+
+on: [push, pull_request]
+
+# Cancel any preceding run on the pull request.
+concurrency:
+  group: actuated-test-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/criu-dev' }}
+
+jobs:
+  build:
+    # Actuated runners are not available in all repositories.
+    if: ${{ github.repository == 'checkpoint-restore/criu' }}
+    # The memory size and the number of CPUs can be freely selected.
+    # 3GB and 4 CPUs seems to be enough according to the result from 'vmmeter'.
+    runs-on: actuated-arm64-4cpu-3gb
+    strategy:
+      matrix:
+        target: [GCC=1, CLANG=1]
+
+    steps:
+    # https://gist.github.com/alexellis/1f33e581c75e11e161fe613c46180771#file-metering-gha-md
+    # vmmeter start
+    - name: Prepare arkade
+      uses: alexellis/arkade-get@master
+      with:
+        crane: latest
+        print-summary: false
+
+    - name: Install vmmeter
+      run: |
+        crane export --platform linux/arm64 ghcr.io/openfaasltd/vmmeter:latest | sudo tar -xvf - -C /usr/local/bin
+
+    - name: Run vmmeter
+      uses: self-actuated/vmmeter-action@master
+    # vmmeter end
+
+    - uses: actions/checkout@v4
+    - name: Run Tests ${{ matrix.target }}
+      # Following tests are failing on the actuated VMs:
+      #  ./change_mnt_context --pidfile=change_mnt_context.pid --outfile=change_mnt_context.out
+      #   45: ERR: change_mnt_context.c:23: mount (errno = 22 (Invalid argument))
+      #
+      # In combination with '--remote-lazy-pages' following error occurs:
+      #  138: FAIL: maps05.c:84: Data corrupted at page 1639 (errno = 11 (Resource temporarily unavailable))
+      run: |
+        # The 'sched_policy00' needs the following:
+        sudo sysctl -w kernel.sched_rt_runtime_us=-1
+        # etc/hosts entry is needed for netns_lock_iptables
+        echo "127.0.0.1   localhost" | sudo tee -a /etc/hosts
+        sudo -E make -C scripts/ci local ${{ matrix.target }} RUN_TESTS=1 \
+          ZDTM_OPTS="-x zdtm/static/change_mnt_context -x zdtm/static/maps05"

coredump/coredump

@@ -1,4 +1,5 @@
 #!/usr/bin/env python3
+import platform
 import argparse
 import os
 import sys
@@ -36,6 +37,10 @@ def main():
 
     opts = vars(parser.parse_args())
 
+    if platform.machine() != 'x86_64':
+        print('ERROR: %s only supported on x86_64' % sys.argv[0])
+        sys.exit(1)
+
     try:
         coredump(opts)
     except SystemExit as error:

crit/crit/__main__.py

@@ -323,12 +323,12 @@ def explore_rss(opts):
         pvmi = -1
         for pm in pms[1:]:
             pstr = '\t%lx / %-8d' % (pm['vaddr'], pm['nr_pages'])
-            while vmas[vmi]['end'] <= pm['vaddr']:
+            while vmi < len(vmas) and vmas[vmi]['end'] <= pm['vaddr']:
                 vmi += 1
 
             pme = pm['vaddr'] + (pm['nr_pages'] << 12)
             vstr = ''
-            while vmas[vmi]['start'] < pme:
+            while vmi < len(vmas) and vmas[vmi]['start'] < pme:
                 vma = vmas[vmi]
                 if vmi == pvmi:
                     vstr += ' ~'

criu/cgroup.c

@@ -28,7 +28,6 @@
 #include "images/cgroup.pb-c.h"
 #include "kerndat.h"
 #include "linux/mount.h"
-#include "syscall.h"
 
 /*
  * This structure describes set of controller groups
@@ -581,37 +580,42 @@ static int __new_open_cgroupfs(struct cg_ctl *cc)
 	int fsfd, fd;
 	char *name;
 
-	fsfd = sys_fsopen(fstype, 0);
+	fsfd = cr_fsopen(fstype, 0);
 	if (fsfd < 0) {
 		pr_perror("Unable to open the cgroup file system");
 		return -1;
 	}
 
 	if (strstartswith(cc->name, namestr)) {
-		if (sys_fsconfig(fsfd, FSCONFIG_SET_STRING, "name", cc->name + strlen(namestr), 0)) {
+		if (cr_fsconfig(fsfd, FSCONFIG_SET_STRING, "name", cc->name + strlen(namestr), 0)) {
+			fsfd_dump_messages(fsfd);
 			pr_perror("Unable to configure the cgroup (%s) file system", cc->name);
 			goto err;
 		}
 	} else if (cc->name[0] != 0) { /* cgroup v1 */
 		char *saveptr = NULL, *buf = strdupa(cc->name);
 		name = strtok_r(buf, ",", &saveptr);
 		while (name) {
-			if (sys_fsconfig(fsfd, FSCONFIG_SET_FLAG, name, NULL, 0)) {
+			if (cr_fsconfig(fsfd, FSCONFIG_SET_FLAG, name, NULL, 0)) {
+				fsfd_dump_messages(fsfd);
 				pr_perror("Unable to configure the cgroup (%s) file system", name);
 				goto err;
 			}
 			name = strtok_r(NULL, ",", &saveptr);
 		}
 	}
 
-	if (sys_fsconfig(fsfd, FSCONFIG_CMD_CREATE, NULL, NULL, 0)) {
+	if (cr_fsconfig(fsfd, FSCONFIG_CMD_CREATE, NULL, NULL, 0)) {
+		fsfd_dump_messages(fsfd);
 		pr_perror("Unable to create the cgroup (%s) file system", cc->name);
 		goto err;
 	}
 
-	fd = sys_fsmount(fsfd, 0, 0);
-	if (fd < 0)
+	fd = cr_fsmount(fsfd, 0, 0);
+	if (fd < 0) {
+		fsfd_dump_messages(fsfd);
 		pr_perror("Unable to mount the cgroup (%s) file system", cc->name);
+	}
 	close(fsfd);
 
 	return fd;

criu/cr-check.c

@@ -54,7 +54,6 @@
 #include "restorer.h"
 #include "uffd.h"
 #include "linux/aio_abi.h"
-#include "syscall.h"
 #include "mount-v2.h"
 
 #include "images/inventory.pb-c.h"
@@ -1437,18 +1436,18 @@ static int ovl_mount(void)
 {
 	int tmpfs, fsfd, ovl;
 
-	fsfd = sys_fsopen("tmpfs", 0);
+	fsfd = cr_fsopen("tmpfs", 0);
 	if (fsfd == -1) {
 		pr_perror("Unable to fsopen tmpfs");
 		return -1;
 	}
 
-	if (sys_fsconfig(fsfd, FSCONFIG_CMD_CREATE, NULL, NULL, 0) == -1) {
+	if (cr_fsconfig(fsfd, FSCONFIG_CMD_CREATE, NULL, NULL, 0) == -1) {
 		pr_perror("Unable to create tmpfs mount");
 		return -1;
 	}
 
-	tmpfs = sys_fsmount(fsfd, 0, 0);
+	tmpfs = cr_fsmount(fsfd, 0, 0);
 	if (tmpfs == -1) {
 		pr_perror("Unable to mount tmpfs");
 		return -1;
@@ -1475,23 +1474,23 @@ static int ovl_mount(void)
 		return -1;
 	}
 
-	fsfd = sys_fsopen("overlay", 0);
+	fsfd = cr_fsopen("overlay", 0);
 	if (fsfd == -1) {
 		pr_perror("Unable to fsopen overlayfs");
 		return -1;
 	}
-	if (sys_fsconfig(fsfd, FSCONFIG_SET_STRING, "source", "test", 0) == -1 ||
-	    sys_fsconfig(fsfd, FSCONFIG_SET_STRING, "lowerdir", "/tmp/l", 0) == -1 ||
-	    sys_fsconfig(fsfd, FSCONFIG_SET_STRING, "upperdir", "/tmp/u", 0) == -1 ||
-	    sys_fsconfig(fsfd, FSCONFIG_SET_STRING, "workdir", "/tmp/w", 0) == -1) {
+	if (cr_fsconfig(fsfd, FSCONFIG_SET_STRING, "source", "test", 0) == -1 ||
+	    cr_fsconfig(fsfd, FSCONFIG_SET_STRING, "lowerdir", "/tmp/l", 0) == -1 ||
+	    cr_fsconfig(fsfd, FSCONFIG_SET_STRING, "upperdir", "/tmp/u", 0) == -1 ||
+	    cr_fsconfig(fsfd, FSCONFIG_SET_STRING, "workdir", "/tmp/w", 0) == -1) {
 		pr_perror("Unable to configure overlayfs");
 		return -1;
 	}
-	if (sys_fsconfig(fsfd, FSCONFIG_CMD_CREATE, NULL, NULL, 0) == -1) {
+	if (cr_fsconfig(fsfd, FSCONFIG_CMD_CREATE, NULL, NULL, 0) == -1) {
 		pr_perror("Unable to create overlayfs");
 		return -1;
 	}
-	ovl = sys_fsmount(fsfd, 0, 0);
+	ovl = cr_fsmount(fsfd, 0, 0);
 	if (ovl == -1) {
 		pr_perror("Unable to mount overlayfs");
 		return -1;

criu/fault-injection.c

@@ -1,6 +1,7 @@
 #include <stdlib.h>
 #include "criu-log.h"
 #include "fault-injection.h"
+#include "seize.h"
 
 enum faults fi_strategy;
 
@@ -21,5 +22,13 @@ int fault_injection_init(void)
 	}
 
 	fi_strategy = start;
+
+	switch (fi_strategy) {
+	case FI_DISABLE_FREEZE_CGROUP:
+		dont_use_freeze_cgroup();
+		break;
+	default:
+		break;
+	};
 	return 0;
 }

criu/include/fault-injection.h

@@ -21,6 +21,8 @@ enum faults {
 	FI_CORRUPT_EXTREGS = 134,
 	FI_DONT_USE_PAGEMAP_SCAN = 135,
 	FI_DUMP_CRASH = 136,
+	FI_DISABLE_FREEZE_CGROUP = 137,
+	FI_PLUGIN_CUDA_FORCE_ENABLE = 138,
 	FI_MAX,
 };
 

criu/include/seize.h

@@ -2,11 +2,13 @@
 #define __CR_SEIZE_H__
 
 extern int collect_pstree(void);
+struct pstree_item;
 extern void pstree_switch_state(struct pstree_item *root_item, int st);
 extern const char *get_real_freezer_state(void);
 extern bool alarm_timeouted(void);
 
 extern char *task_comm_info(pid_t pid, char *comm, size_t size);
 extern char *__task_comm_info(pid_t pid);
+extern void dont_use_freeze_cgroup(void);
 
 #endif

criu/include/util.h

@@ -387,6 +387,11 @@ static inline void print_stack_trace(pid_t pid)
 
 extern int mount_detached_fs(const char *fsname);
 
+extern int cr_fsopen(const char *fsname, unsigned int flags);
+extern int cr_fsconfig(int fd, unsigned int cmd, const char *key, const char *value, int aux);
+extern int cr_fsmount(int fd, unsigned int flags, unsigned int attr_flags);
+extern void fsfd_dump_messages(int fd);
+
 extern char *get_legacy_iptables_bin(bool ipv6, bool restore);
 
 extern int set_opts_cap_eff(void);

criu/kerndat.c

@@ -834,7 +834,7 @@ static int kerndat_detect_stack_guard_gap(void)
 		 * (see kernel commit 1be7107fbe18ee).
 		 *
 		 * Same time there was semi-complete
-		 * patch released which hitted a number
+		 * patch released which hit a number
 		 * of repos (Ubuntu, Fedora) where instead
 		 * of PAGE_SIZE the 1M gap is cut off.
 		 */