X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=19fb086e7ab8ef2c9a67278f5329d17f10c80dd8;hp=73158a051831edf74a5db86ee8027b471a5e28c8;hb=d6797c920e9eb70f46a893c00fdd9ecb86d15f84;hpb=590b6b9188e75ba46c42995984a1c2fa06adb6d6 diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 73158a051..19fb086e7 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -84,6 +84,7 @@ #include "def.h" #include "rtnl-util.h" #include "udev-util.h" +#include "eventfd-util.h" #include "blkid-util.h" #include "gpt.h" #include "siphash24.h" @@ -92,6 +93,11 @@ #include "seccomp-util.h" #endif +typedef enum ContainerStatus { + CONTAINER_TERMINATED, + CONTAINER_REBOOTED +} ContainerStatus; + typedef enum LinkJournal { LINK_NO, LINK_AUTO, @@ -629,7 +635,7 @@ static int mount_all(const char *dest) { return r; } -static int mount_binds(const char *dest, char **l, unsigned long flags) { +static int mount_binds(const char *dest, char **l, bool ro) { char **x, **y; STRV_FOREACH_PAIR(x, y, l) { @@ -680,9 +686,12 @@ static int mount_binds(const char *dest, char **l, unsigned long flags) { return -errno; } - if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) { - log_error("mount(%s) failed: %m", where); - return -errno; + if (ro) { + r = bind_remount_recursive(where, true); + if (r < 0) { + log_error("Read-Only bind mount failed: %s", strerror(-r)); + return r; + } } } @@ -2569,6 +2578,76 @@ static int change_uid_gid(char **_home) { return 0; } +/* + * Return 0 in case the container is being rebooted, has been shut + * down or exited successfully. On failures a negative value is + * returned. + * + * The status of the container "CONTAINER_TERMINATED" or + * "CONTAINER_REBOOTED" will be saved in the container argument + */ +static int wait_for_container(pid_t pid, ContainerStatus *container) { + int r; + siginfo_t status; + + r = wait_for_terminate(pid, &status); + if (r < 0) + return r; + + switch (status.si_code) { + case CLD_EXITED: + r = status.si_status; + if (r == 0) { + if (!arg_quiet) + log_debug("Container %s exited successfully.", + arg_machine); + + *container = CONTAINER_TERMINATED; + } else { + log_error("Container %s failed with error code %i.", + arg_machine, status.si_status); + r = -1; + } + break; + + case CLD_KILLED: + if (status.si_status == SIGINT) { + if (!arg_quiet) + log_info("Container %s has been shut down.", + arg_machine); + + *container = CONTAINER_TERMINATED; + r = 0; + break; + } else if (status.si_status == SIGHUP) { + if (!arg_quiet) + log_info("Container %s is being rebooted.", + arg_machine); + + *container = CONTAINER_REBOOTED; + r = 0; + break; + } + /* CLD_KILLED fallthrough */ + + case CLD_DUMPED: + log_error("Container %s terminated by signal %s.", + arg_machine, signal_to_string(status.si_status)); + r = -1; + break; + + default: + log_error("Container %s failed due to unknown reason.", + arg_machine); + r = -1; + break; + } + + return r; +} + +static void nop_handler(int sig) {} + int main(int argc, char *argv[]) { _cleanup_free_ char *kdbus_domain = NULL, *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL; @@ -2580,8 +2659,8 @@ int main(int argc, char *argv[]) { const char *console = NULL; char veth_name[IFNAMSIZ]; bool secondary = false; + sigset_t mask, mask_chld; pid_t pid = 0; - sigset_t mask; log_parse_environment(); log_open(); @@ -2743,36 +2822,44 @@ int main(int argc, char *argv[]) { sd_notify(0, "READY=1"); assert_se(sigemptyset(&mask) == 0); + assert_se(sigemptyset(&mask_chld) == 0); + sigaddset(&mask_chld, SIGCHLD); sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); for (;;) { - int parent_ready_fd = -1, child_ready_fd = -1; - siginfo_t status; - eventfd_t x; - - parent_ready_fd = eventfd(0, EFD_CLOEXEC); - if (parent_ready_fd < 0) { - log_error("Failed to create event fd: %m"); + ContainerStatus container_status; + int eventfds[2] = { -1, -1 }; + struct sigaction sa = { + .sa_handler = nop_handler, + .sa_flags = SA_NOCLDSTOP, + }; + + /* Child can be killed before execv(), so handle SIGCHLD + * in order to interrupt parent's blocking calls and + * give it a chance to call wait() and terminate. */ + r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL); + if (r < 0) { + log_error("Failed to change the signal mask: %m"); goto finish; } - child_ready_fd = eventfd(0, EFD_CLOEXEC); - if (child_ready_fd < 0) { - log_error("Failed to create event fd: %m"); + r = sigaction(SIGCHLD, &sa, NULL); + if (r < 0) { + log_error("Failed to install SIGCHLD handler: %m"); goto finish; } - pid = syscall(__NR_clone, - SIGCHLD|CLONE_NEWNS| - (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)| - (arg_private_network ? CLONE_NEWNET : 0), NULL); + pid = clone_with_eventfd(SIGCHLD|CLONE_NEWNS| + (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)| + (arg_private_network ? CLONE_NEWNET : 0), eventfds); if (pid < 0) { if (errno == EINVAL) log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m"); else log_error("clone() failed: %m"); + r = pid; goto finish; } @@ -2857,15 +2944,17 @@ int main(int argc, char *argv[]) { /* Turn directory into bind mount */ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) { - log_error("Failed to make bind mount."); + log_error("Failed to make bind mount: %m"); goto child_fail; } - if (arg_read_only) - if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) { - log_error("Failed to make read-only."); + if (arg_read_only) { + k = bind_remount_recursive(arg_directory, true); + if (k < 0) { + log_error("Failed to make tree read-only: %s", strerror(-k)); goto child_fail; } + } if (mount_all(arg_directory) < 0) goto child_fail; @@ -2901,10 +2990,10 @@ int main(int argc, char *argv[]) { if (setup_journal(arg_directory) < 0) goto child_fail; - if (mount_binds(arg_directory, arg_bind, 0) < 0) + if (mount_binds(arg_directory, arg_bind, false) < 0) goto child_fail; - if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0) + if (mount_binds(arg_directory, arg_bind_ro, true) < 0) goto child_fail; if (setup_kdbus(arg_directory, kdbus_domain) < 0) @@ -2913,8 +3002,11 @@ int main(int argc, char *argv[]) { /* Tell the parent that we are ready, and that * it can cgroupify us to that we lack access * to certain devices and resources. */ - eventfd_write(child_ready_fd, 1); - child_ready_fd = safe_close(child_ready_fd); + r = eventfd_send_state(eventfds[1], + EVENTFD_CHILD_SUCCEEDED); + eventfds[1] = safe_close(eventfds[1]); + if (r < 0) + goto child_fail; if (chdir(arg_directory) < 0) { log_error("chdir(%s) failed: %m", arg_directory); @@ -3016,8 +3108,10 @@ int main(int argc, char *argv[]) { env_use = (char**) envp; /* Wait until the parent is ready with the setup, too... */ - eventfd_read(parent_ready_fd, &x); - parent_ready_fd = safe_close(parent_ready_fd); + r = eventfd_parent_succeeded(eventfds[0]); + eventfds[0] = safe_close(eventfds[0]); + if (r < 0) + goto child_fail; if (arg_boot) { char **a; @@ -3048,17 +3142,27 @@ int main(int argc, char *argv[]) { log_error("execv() failed: %m"); child_fail: + /* Tell the parent that the setup failed, so he + * can clean up resources and terminate. */ + if (eventfds[1] != -1) + eventfd_send_state(eventfds[1], + EVENTFD_CHILD_FAILED); _exit(EXIT_FAILURE); } fdset_free(fds); fds = NULL; - /* Wait until the child reported that it is ready with - * all it needs to do with privileges. After we got - * the notification we can make the process join its - * cgroup which might limit what it can do */ - eventfd_read(child_ready_fd, &x); + /* Wait for the child event: + * If EVENTFD_CHILD_FAILED, the child will terminate soon. + * If EVENTFD_CHILD_SUCCEEDED, the child is reporting that + * it is ready with all it needs to do with priviliges. + * After we got the notification we can make the process + * join its cgroup which might limit what it can do */ + r = eventfd_child_succeeded(eventfds[1]); + eventfds[1] = safe_close(eventfds[1]); + if (r < 0) + goto check_container_status; r = register_machine(pid); if (r < 0) @@ -3080,10 +3184,25 @@ int main(int argc, char *argv[]) { if (r < 0) goto finish; + /* Block SIGCHLD here, before notifying child. + * process_pty() will handle it with the other signals. */ + r = sigprocmask(SIG_BLOCK, &mask_chld, NULL); + if (r < 0) + goto finish; + + /* Reset signal to default */ + r = default_signals(SIGCHLD, -1); + if (r < 0) + goto finish; + /* Notify the child that the parent is ready with all - * its setup, and thtat the child can now hand over + * its setup, and that the child can now hand over * control to the code to run inside the container. */ - eventfd_write(parent_ready_fd, 1); + r = eventfd_send_state(eventfds[0], + EVENTFD_PARENT_SUCCEEDED); + eventfds[0] = safe_close(eventfds[0]); + if (r < 0) + goto finish; k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3); if (k < 0) { @@ -3097,51 +3216,20 @@ int main(int argc, char *argv[]) { /* Kill if it is not dead yet anyway */ terminate_machine(pid); +check_container_status: /* Redundant, but better safe than sorry */ kill(pid, SIGKILL); - k = wait_for_terminate(pid, &status); + r = wait_for_container(pid, &container_status); pid = 0; - if (k < 0) { + if (r < 0) { r = EXIT_FAILURE; break; - } - - if (status.si_code == CLD_EXITED) { - r = status.si_status; - if (status.si_status != 0) { - log_error("Container %s failed with error code %i.", arg_machine, status.si_status); - break; - } - - if (!arg_quiet) - log_debug("Container %s exited successfully.", arg_machine); - break; - } else if (status.si_code == CLD_KILLED && - status.si_status == SIGINT) { - - if (!arg_quiet) - log_info("Container %s has been shut down.", arg_machine); - r = 0; + } else if (container_status == CONTAINER_TERMINATED) break; - } else if (status.si_code == CLD_KILLED && - status.si_status == SIGHUP) { - if (!arg_quiet) - log_info("Container %s is being rebooted.", arg_machine); - continue; - } else if (status.si_code == CLD_KILLED || - status.si_code == CLD_DUMPED) { - - log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status)); - r = EXIT_FAILURE; - break; - } else { - log_error("Container %s failed due to unknown reason.", arg_machine); - r = EXIT_FAILURE; - break; - } + /* CONTAINER_REBOOTED, loop again */ } finish: