start: fix cgroup namespace preservation

Prior to this patch we raced with a very short-lived init process. Essentially,
the init process could exit before we had time to record the cgroup namespace
causing the container to abort and report ABORTING to the caller when it
actually started just fine. Let's not do this.

(This uses syscall(SYS_getpid) in the the child to retrieve the pid just in case
we're on an older glibc version and we end up in the namespace sharing branch
of the actual lxc_clone() call.)

Additionally this fixes the shortlived tests. They were faulty so far and
should have actually failed because of the cgroup namespace recording race but
the ret variable used to return from the function was not correctly
initialized. This fixes it.
Furthermore, the shortlived tests used the c->error_num variable to determine
success or failure but this is actually not correct when the container is
started daemonized.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

diff --git a/src/lxc/criu.c b/src/lxc/criu.c
index bb136d4cb176452d0c4bb03f2ec641eed246c196..3e9b4ebb7635dcfdfa3f40a33d87360fed0175d8 100644 (file)
--- a/src/lxc/criu.c
+++ b/src/lxc/criu.c
@@ -829,7 +829,7 @@ out_unlock:
  */
 static void do_restore(struct lxc_container *c, int status_pipe, struct migrate_opts *opts, char *criu_version)
 {
-       int fd;
+       int fd, ret;
        pid_t pid;
        struct lxc_handler *handler;
        int status = 0;
@@ -870,7 +870,11 @@ static void do_restore(struct lxc_container *c, int status_pipe, struct migrate_
                goto out_fini_handler;
        }
 
-       resolve_clone_flags(handler);
+       ret = resolve_clone_flags(handler);
+       if (ret < 0) {
+               ERROR("%s - Unsupported clone flag specified", strerror(errno));
+               goto out_fini_handler;
+       }
 
        if (pipe(pipes) < 0) {
                SYSERROR("pipe() failed");

diff --git a/src/lxc/start.c b/src/lxc/start.c
index 5eff606cbc98efa1ebdef8d62986d02a3aeb1d2b..f7b55844f5292a16fe867205660a7f2233bdb16f 100644 (file)
--- a/src/lxc/start.c
+++ b/src/lxc/start.c
@@ -818,7 +818,7 @@ void lxc_abort(const char *name, struct lxc_handler *handler)
 
 static int do_start(void *data)
 {
-       int ret;
+       int fd, ret;
        struct lxc_list *iterator;
        char path[PATH_MAX];
        struct lxc_handler *handler = data;
@@ -935,7 +935,7 @@ static int do_start(void *data)
         *
         *      8:cpuset:/
         */
-       if (cgns_supported()) {
+       if (handler->clone_flags & CLONE_NEWCGROUP) {
                if (unshare(CLONE_NEWCGROUP) < 0) {
                        INFO("Failed to unshare CLONE_NEWCGROUP.");
                        goto out_warn_father;
@@ -956,13 +956,31 @@ static int do_start(void *data)
 
        /* Setup the container, ip, names, utsname, ... */
        ret = lxc_setup_child(handler);
-       close(handler->data_sock[0]);
        close(handler->data_sock[1]);
        if (ret < 0) {
                ERROR("Failed to setup container \"%s\".", handler->name);
+               close(handler->data_sock[0]);
                goto out_warn_father;
        }
 
+       if (handler->clone_flags & CLONE_NEWCGROUP) {
+               fd = lxc_preserve_ns(syscall(SYS_getpid), "cgroup");
+               if (fd < 0) {
+                       ERROR("%s - Failed to preserve cgroup namespace", strerror(errno));
+                       close(handler->data_sock[0]);
+                       goto out_warn_father;
+               }
+
+               ret = lxc_abstract_unix_send_fds(handler->data_sock[0], &fd, 1, NULL, 0);
+               close(fd);
+               if (ret < 0) {
+                       ERROR("%s - Failed to preserve cgroup namespace", strerror(errno));
+                       close(handler->data_sock[0]);
+                       goto out_warn_father;
+               }
+       }
+       close(handler->data_sock[0]);
+
        /* Set the label to change to when we exec(2) the container's init. */
        if (lsm_process_label_set(NULL, handler->conf, 1, 1) < 0)
                goto out_warn_father;
@@ -1144,7 +1162,7 @@ static int lxc_recv_ttys_from_child(struct lxc_handler *handler)
        return ret;
 }
 
-void resolve_clone_flags(struct lxc_handler *handler)
+int resolve_clone_flags(struct lxc_handler *handler)
 {
        handler->clone_flags = CLONE_NEWPID | CLONE_NEWNS;
 
@@ -1155,18 +1173,34 @@ void resolve_clone_flags(struct lxc_handler *handler)
                if (!lxc_requests_empty_network(handler))
                        handler->clone_flags |= CLONE_NEWNET;
        } else {
-               INFO("Inheriting a NET namespace.");
+               INFO("Inheriting a net namespace.");
        }
 
        if (handler->conf->inherit_ns_fd[LXC_NS_IPC] == -1)
                handler->clone_flags |= CLONE_NEWIPC;
        else
-               INFO("Inheriting an IPC namespace.");
+               INFO("Inheriting an ipc namespace.");
 
        if (handler->conf->inherit_ns_fd[LXC_NS_UTS] == -1)
                handler->clone_flags |= CLONE_NEWUTS;
        else
-               INFO("Inheriting a UTS namespace.");
+               INFO("Inheriting uts namespace");
+
+       if (handler->conf->inherit_ns_fd[LXC_NS_PID] == -1)
+               handler->clone_flags |= CLONE_NEWPID;
+       else
+               INFO("Inheriting pid namespace");
+
+       if (cgns_supported()) {
+               if (handler->conf->inherit_ns_fd[LXC_NS_CGROUP] == -1)
+                       handler->clone_flags |= CLONE_NEWCGROUP;
+               else
+                       INFO("Inheriting cgroup namespace");
+       } else if (handler->conf->inherit_ns_fd[LXC_NS_CGROUP] >= 0) {
+                       return -EINVAL;
+       }
+
+       return 0;
 }
 
 /* lxc_spawn() performs crucial setup tasks and clone()s the new process which
@@ -1204,7 +1238,11 @@ static int lxc_spawn(struct lxc_handler *handler)
                return -1;
        }
 
-       resolve_clone_flags(handler);
+       ret = resolve_clone_flags(handler);
+       if (ret < 0) {
+               lxc_sync_fini(handler);
+               return -1;
+       }
 
        if (handler->clone_flags & CLONE_NEWNET) {
                if (!lxc_list_empty(&handler->conf->network)) {
@@ -1269,6 +1307,7 @@ static int lxc_spawn(struct lxc_handler *handler)
                 */
                flags &= ~CLONE_NEWNET;
        }
+
        handler->pid = lxc_clone(do_start, handler, flags);
        if (handler->pid < 0) {
                SYSERROR("Failed to clone a new set of namespaces.");
@@ -1399,6 +1438,17 @@ static int lxc_spawn(struct lxc_handler *handler)
                goto out_delete_net;
        }
 
+       if (handler->clone_flags & CLONE_NEWCGROUP) {
+               ret = lxc_abstract_unix_recv_fds(handler->data_sock[1],
+                                                &handler->nsfd[LXC_NS_CGROUP],
+                                                1, NULL, 0);
+               if (ret < 0) {
+                       ERROR("%s - Failed to preserve cgroup namespace", strerror(errno));
+                       goto out_delete_net;
+               }
+               DEBUG("Preserved cgroup namespace via fd %d", handler->nsfd[LXC_NS_CGROUP]);
+       }
+
        if (handler->ops->post_start(handler, handler->data))
                goto out_abort;
 

diff --git a/src/lxc/start.h b/src/lxc/start.h
index 28f9817bf2f15ab6d1e46f6c4d338009d3f65a51..e35e61f7e66f6e36457b79df811e3d743aeaf178 100644 (file)
--- a/src/lxc/start.h
+++ b/src/lxc/start.h
@@ -142,6 +142,6 @@ extern int lxc_check_inherited(struct lxc_conf *conf, bool closeall,
 extern int __lxc_start(const char *, struct lxc_handler *,
                       struct lxc_operations *, void *, const char *, bool);
 
-extern void resolve_clone_flags(struct lxc_handler *handler);
+extern int resolve_clone_flags(struct lxc_handler *handler);
 #endif
 

diff --git a/src/tests/shortlived.c b/src/tests/shortlived.c
index b5397e7eb7cce6187f3d93d0fa1e2876cc1d58a7..af5bb2eb7e0b0afabdd8883e698796111566bd1d 100644 (file)
--- a/src/tests/shortlived.c
+++ b/src/tests/shortlived.c
@@ -95,13 +95,13 @@ again:
 
 int main(int argc, char *argv[])
 {
-       int i;
+       int fd, i;
        const char *s;
        bool b;
        struct lxc_container *c;
        struct lxc_log log;
        char template[sizeof(P_tmpdir"/shortlived_XXXXXX")];
-       int ret = 0;
+       int ret = EXIT_FAILURE;
 
        strcpy(template, P_tmpdir"/shortlived_XXXXXX");
        i = lxc_make_tmpfile(template, false);
@@ -122,12 +122,10 @@ int main(int argc, char *argv[])
        if (lxc_log_init(&log))
                exit(EXIT_FAILURE);
 
-       ret = 1;
        /* test a real container */
        c = lxc_container_new(MYNAME, NULL);
        if (!c) {
                fprintf(stderr, "%d: error creating lxc_container %s\n", __LINE__, MYNAME);
-               ret = 1;
                goto out;
        }
 
@@ -136,8 +134,7 @@ int main(int argc, char *argv[])
                goto out;
        }
 
-       ret = create_container();
-       if (ret) {
+       if (create_container() < 0) {
                fprintf(stderr, "%d: failed to create a container\n", __LINE__);
                goto out;
        }
@@ -166,7 +163,7 @@ int main(int argc, char *argv[])
        }
 
        if (!c->set_config_item(c, "lxc.execute.cmd", "echo hello")) {
-               fprintf(stderr, "%d: failed setting lxc.init.cmd\n", __LINE__);
+               fprintf(stderr, "%d: failed setting lxc.execute.cmd\n", __LINE__);
                goto out;
        }
 
@@ -179,13 +176,6 @@ int main(int argc, char *argv[])
                        goto out;
                }
 
-               /* The container needs to exit with a successful error code. */
-               if (c->error_num != 0) {
-                       fprintf(stderr, "%d: %s exited successfully on %dth iteration\n", __LINE__, c->name, i);
-                       goto out;
-               }
-               fprintf(stderr, "%d: %s exited correctly with error code %d on %dth iteration\n", __LINE__, c->name, c->error_num, i);
-
                if (!c->wait(c, "STOPPED", 30)) {
                        fprintf(stderr, "%d: %s failed to wait on %dth iteration\n", __LINE__, c->name, i);
                        goto out;
@@ -199,13 +189,6 @@ int main(int argc, char *argv[])
                        goto out;
                }
 
-               /* The container needs to exit with a successful error code. */
-               if (c->error_num != 0) {
-                       fprintf(stderr, "%d: %s exited successfully on %dth iteration\n", __LINE__, c->name, i);
-                       goto out;
-               }
-               fprintf(stderr, "%d: %s exited correctly with error code %d on %dth iteration\n", __LINE__, c->name, c->error_num, i);
-
                if (!c->wait(c, "STOPPED", 30)) {
                        fprintf(stderr, "%d: %s failed to wait on %dth iteration\n", __LINE__, c->name, i);
                        goto out;
@@ -229,13 +212,6 @@ int main(int argc, char *argv[])
                        goto out;
                }
 
-               /* The container needs to exit with an error code.*/
-               if (c->error_num == 0) {
-                       fprintf(stderr, "%d: %s exited successfully on %dth iteration\n", __LINE__, c->name, i);
-                       goto out;
-               }
-               fprintf(stderr, "%d: %s exited correctly with error code %d on %dth iteration\n", __LINE__, c->name, c->error_num, i);
-
                if (!c->wait(c, "STOPPED", 30)) {
                        fprintf(stderr, "%d: %s failed to wait on %dth iteration\n", __LINE__, c->name, i);
                        goto out;
@@ -252,13 +228,6 @@ int main(int argc, char *argv[])
                        goto out;
                }
 
-               /* The container needs to exit with an error code.*/
-               if (c->error_num == 0) {
-                       fprintf(stderr, "%d: %s exited successfully on %dth iteration\n", __LINE__, c->name, i);
-                       goto out;
-               }
-               fprintf(stderr, "%d: %s exited correctly with error code %d on %dth iteration\n", __LINE__, c->name, c->error_num, i);
-
                if (!c->wait(c, "STOPPED", 30)) {
                        fprintf(stderr, "%d: %s failed to wait on %dth iteration\n", __LINE__, c->name, i);
                        goto out;
@@ -276,6 +245,21 @@ out:
                destroy_container();
        }
        lxc_container_put(c);
+
+       if (ret != 0) {
+               fd = open(template, O_RDONLY);
+               if (fd >= 0) {
+                       char buf[4096];
+                       ssize_t buflen;
+                       while ((buflen = read(fd, buf, 1024)) > 0) {
+                               buflen = write(STDERR_FILENO, buf, buflen);
+                               if (buflen <= 0)
+                                       break;
+                       }
+                       close(fd);
+               }
+       }
+
        unlink(template);
        exit(ret);
 }