X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=efeba596658079606376811963586f61da49df3a;hb=f6c51a8136de3f27e28caea2003e18f4bc4cb9a8;hp=ce9a9e84f69af363b33739a7bba64140f27f9f76;hpb=0dfaa0060711a8332c8eb9f1e10f48fe182d3650;p=elogind.git diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index ce9a9e84f..efeba5966 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -43,6 +43,8 @@ #include #include #include +#include +#include #ifdef HAVE_SELINUX #include @@ -216,7 +218,7 @@ static void help(void) { " and container and add it to an existing bridge on\n" " the host\n" " -p --port=[PROTOCOL:]HOSTPORT[:CONTAINERPORT]\n" - " Expose a container IP port ont the host\n" + " Expose a container IP port on the host\n" " -Z --selinux-context=SECLABEL\n" " Set the SELinux security context to be used by\n" " processes in the container\n" @@ -1029,7 +1031,7 @@ static int mount_cgroup(const char *dest) { return r; if (symlink(combined, target) < 0) - return log_error_errno(errno, "Failed to create symlink for combined hiearchy: %m"); + return log_error_errno(errno, "Failed to create symlink for combined hierarchy: %m"); } } @@ -1655,7 +1657,7 @@ static int watch_rtnl(sd_event *event, int recv_fd, union in_addr_union *exposed cmsg = CMSG_FIRSTHDR(&mh); assert(cmsg->cmsg_level == SOL_SOCKET); assert(cmsg->cmsg_type == SCM_RIGHTS); - assert(cmsg->cmsg_len = CMSG_LEN(sizeof(int))); + assert(cmsg->cmsg_len == CMSG_LEN(sizeof(int))); memcpy(&fd, CMSG_DATA(cmsg), sizeof(int)); r = sd_rtnl_open_fd(&rtnl, fd, 1, RTNLGRP_IPV4_IFADDR); @@ -2557,6 +2559,70 @@ static int setup_image(char **device_path, int *loop_nr) { return r; } +static int wait_for_block_device(struct udev *udev, dev_t devnum, struct udev_device **ret) { + _cleanup_udev_monitor_unref_ struct udev_monitor *monitor = NULL; + int r; + + assert(udev); + assert(ret); + + for (;;) { + _cleanup_udev_device_unref_ struct udev_device *d = NULL; + struct pollfd pfd = { + .events = POLLIN + }; + + d = udev_device_new_from_devnum(udev, 'b', devnum); + if (!d) + return log_oom(); + + r = udev_device_get_is_initialized(d); + if (r < 0) + return log_error_errno(r, "Failed to check if device is initialized: %m"); + if (r > 0) { + *ret = d; + d = NULL; + return 0; + } + d = udev_device_unref(d); + + if (!monitor) { + monitor = udev_monitor_new_from_netlink(udev, "udev"); + if (!monitor) + return log_oom(); + + r = udev_monitor_filter_add_match_subsystem_devtype(monitor, "block", NULL); + if (r < 0) + return log_error_errno(r, "Failed to add block match: %m"); + + r = udev_monitor_enable_receiving(monitor); + if (r < 0) + return log_error_errno(r, "Failed to turn on monitor: %m"); + + continue; + } + + pfd.fd = udev_monitor_get_fd(monitor); + if (pfd.fd < 0) + return log_error_errno(r, "Failed to get udev monitor fd: %m"); + + r = poll(&pfd, 1, -1); + if (r < 0) + return log_error_errno(errno, "Failed to wait for device initialization: %m"); + + d = udev_monitor_receive_device(monitor); + } + + return 0; +} + +#define PARTITION_TABLE_BLURB \ + "Note that the disk image needs to either contain only a single MBR partition of\n" \ + "type 0x83 that is marked bootable, or a sinlge GPT partition of type" \ + "0FC63DAF-8483-4772-8E79-3D69D8477DE4 or follow\n" \ + " http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/\n" \ + "to be bootable with systemd-nspawn." + static int dissect_image( int fd, char **root_device, bool *root_device_rw, @@ -2572,18 +2638,18 @@ static int dissect_image( #ifdef GPT_ROOT_SECONDARY int secondary_root_nr = -1; #endif - - _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL; + _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL, *generic = NULL; _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL; _cleanup_udev_device_unref_ struct udev_device *d = NULL; _cleanup_blkid_free_probe_ blkid_probe b = NULL; _cleanup_udev_unref_ struct udev *udev = NULL; struct udev_list_entry *first, *item; - bool home_rw = true, root_rw = true, secondary_root_rw = true, srv_rw = true; + bool home_rw = true, root_rw = true, secondary_root_rw = true, srv_rw = true, generic_rw = true; const char *pttype = NULL; blkid_partlist pl; struct stat st; int r; + bool is_gpt, is_mbr, multiple_generic = false; assert(fd >= 0); assert(root_device); @@ -2612,8 +2678,9 @@ static int dissect_image( errno = 0; r = blkid_do_safeprobe(b); if (r == -2 || r == 1) { - log_error("Failed to identify any partition table on %s.\n" - "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); + log_error("Failed to identify any partition table on\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); return -EINVAL; } else if (r != 0) { if (errno == 0) @@ -2623,9 +2690,14 @@ static int dissect_image( } blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL); - if (!streq_ptr(pttype, "gpt")) { - log_error("Image %s does not carry a GUID Partition Table.\n" - "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); + + is_gpt = streq_ptr(pttype, "gpt"); + is_mbr = streq_ptr(pttype, "dos"); + + if (!is_gpt && !is_mbr) { + log_error("No GPT or MBR partition table discovered on\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); return -EINVAL; } @@ -2646,9 +2718,9 @@ static int dissect_image( if (fstat(fd, &st) < 0) return log_error_errno(errno, "Failed to stat block device: %m"); - d = udev_device_new_from_devnum(udev, 'b', st.st_rdev); - if (!d) - return log_oom(); + r = wait_for_block_device(udev, st.st_rdev, &d); + if (r < 0) + return r; e = udev_enumerate_new(udev); if (!e) @@ -2665,9 +2737,8 @@ static int dissect_image( first = udev_enumerate_get_list_entry(e); udev_list_entry_foreach(item, first) { _cleanup_udev_device_unref_ struct udev_device *q; - const char *stype, *node; + const char *node; unsigned long long flags; - sd_id128_t type_id; blkid_partition pp; dev_t qn; int nr; @@ -2698,82 +2769,110 @@ static int dissect_image( continue; flags = blkid_partition_get_flags(pp); - if (flags & GPT_FLAG_NO_AUTO) - continue; nr = blkid_partition_get_partno(pp); if (nr < 0) continue; - stype = blkid_partition_get_type_string(pp); - if (!stype) - continue; + if (is_gpt) { + sd_id128_t type_id; + const char *stype; - if (sd_id128_from_string(stype, &type_id) < 0) - continue; + if (flags & GPT_FLAG_NO_AUTO) + continue; - if (sd_id128_equal(type_id, GPT_HOME)) { + stype = blkid_partition_get_type_string(pp); + if (!stype) + continue; - if (home && nr >= home_nr) + if (sd_id128_from_string(stype, &type_id) < 0) continue; - home_nr = nr; - home_rw = !(flags & GPT_FLAG_READ_ONLY); + if (sd_id128_equal(type_id, GPT_HOME)) { - free(home); - home = strdup(node); - if (!home) - return log_oom(); - } else if (sd_id128_equal(type_id, GPT_SRV)) { + if (home && nr >= home_nr) + continue; - if (srv && nr >= srv_nr) - continue; + home_nr = nr; + home_rw = !(flags & GPT_FLAG_READ_ONLY); - srv_nr = nr; - srv_rw = !(flags & GPT_FLAG_READ_ONLY); + r = free_and_strdup(&home, node); + if (r < 0) + return log_oom(); - free(srv); - srv = strdup(node); - if (!srv) - return log_oom(); - } + } else if (sd_id128_equal(type_id, GPT_SRV)) { + + if (srv && nr >= srv_nr) + continue; + + srv_nr = nr; + srv_rw = !(flags & GPT_FLAG_READ_ONLY); + + r = free_and_strdup(&srv, node); + if (r < 0) + return log_oom(); + } #ifdef GPT_ROOT_NATIVE - else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { + else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { - if (root && nr >= root_nr) - continue; + if (root && nr >= root_nr) + continue; - root_nr = nr; - root_rw = !(flags & GPT_FLAG_READ_ONLY); + root_nr = nr; + root_rw = !(flags & GPT_FLAG_READ_ONLY); - free(root); - root = strdup(node); - if (!root) - return log_oom(); - } + r = free_and_strdup(&root, node); + if (r < 0) + return log_oom(); + } #endif #ifdef GPT_ROOT_SECONDARY - else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY)) { + else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY)) { + + if (secondary_root && nr >= secondary_root_nr) + continue; + + secondary_root_nr = nr; + secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY); + + r = free_and_strdup(&secondary_root, node); + if (r < 0) + return log_oom(); + } +#endif + else if (sd_id128_equal(type_id, GPT_LINUX_GENERIC)) { + + if (generic) + multiple_generic = true; + else { + generic_rw = !(flags & GPT_FLAG_READ_ONLY); + + r = free_and_strdup(&generic, node); + if (r < 0) + return log_oom(); + } + } + + } else if (is_mbr) { + int type; - if (secondary_root && nr >= secondary_root_nr) + if (flags != 0x80) /* Bootable flag */ continue; - secondary_root_nr = nr; - secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY); + type = blkid_partition_get_type(pp); + if (type != 0x83) /* Linux partition */ + continue; + if (generic) + multiple_generic = true; + else { + generic_rw = true; - free(secondary_root); - secondary_root = strdup(node); - if (!secondary_root) - return log_oom(); + r = free_and_strdup(&root, node); + if (r < 0) + return log_oom(); + } } -#endif - } - - if (!root && !secondary_root) { - log_error("Failed to identify root partition in disk image %s.\n" - "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); - return -EINVAL; } if (root) { @@ -2788,6 +2887,31 @@ static int dissect_image( *root_device_rw = secondary_root_rw; *secondary = true; + } else if (generic) { + + /* There were no partitions with precise meanings + * around, but we found generic partitions. In this + * case, if there's only one, we can go ahead and boot + * it, otherwise we bail out, because we really cannot + * make any sense of it. */ + + if (multiple_generic) { + log_error("Identified multiple bootable Linux partitions on\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); + return -EINVAL; + } + + *root_device = generic; + generic = NULL; + + *root_device_rw = generic_rw; + *secondary = false; + } else { + log_error("Failed to identify root partition in disk image\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); + return -EINVAL; } if (home) { @@ -2915,7 +3039,7 @@ static void loop_remove(int nr, int *image_fd) { if (image_fd && *image_fd >= 0) { r = ioctl(*image_fd, LOOP_CLR_FD); if (r < 0) - log_warning_errno(errno, "Failed to close loop image: %m"); + log_debug_errno(errno, "Failed to close loop image: %m"); *image_fd = safe_close(*image_fd); } @@ -2927,7 +3051,7 @@ static void loop_remove(int nr, int *image_fd) { r = ioctl(control, LOOP_CTL_REMOVE, nr); if (r < 0) - log_warning_errno(errno, "Failed to remove loop %d: %m", nr); + log_debug_errno(errno, "Failed to remove loop %d: %m", nr); } static int spawn_getent(const char *database, const char *key, pid_t *rpid) { @@ -3273,7 +3397,7 @@ static int determine_names(void) { return -ENOENT; } - if (i->type == IMAGE_GPT) + if (i->type == IMAGE_RAW) r = set_sanitized_path(&arg_image, i->path); else r = set_sanitized_path(&arg_directory, i->path); @@ -3337,6 +3461,7 @@ int main(int argc, char *argv[]) { pid_t pid = 0; int ret = EXIT_SUCCESS; union in_addr_union exposed = {}; + _cleanup_release_lock_file_ LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT; log_parse_environment(); log_open(); @@ -3382,20 +3507,8 @@ int main(int argc, char *argv[]) { goto finish; } - if (arg_template) { - r = btrfs_subvol_snapshot(arg_template, arg_directory, arg_read_only, true); - if (r == -EEXIST) { - if (!arg_quiet) - log_info("Directory %s already exists, not populating from template %s.", arg_directory, arg_template); - } else if (r < 0) { - log_error_errno(r, "Couldn't create snapshort %s from %s: %m", arg_directory, arg_template); - goto finish; - } else { - if (!arg_quiet) - log_info("Populated %s from template %s.", arg_directory, arg_template); - } - - } else if (arg_ephemeral) { + if (arg_ephemeral) { + _cleanup_release_lock_file_ LockFile original_lock = LOCK_FILE_INIT; char *np; /* If the specified path is a mount point we @@ -3418,6 +3531,12 @@ int main(int argc, char *argv[]) { goto finish; } + r = image_path_lock(np, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock); + if (r < 0) { + log_error_errno(r, "Failed to lock %s: %m", np); + goto finish; + } + r = btrfs_subvol_snapshot(arg_directory, np, arg_read_only, true); if (r < 0) { free(np); @@ -3429,6 +3548,31 @@ int main(int argc, char *argv[]) { arg_directory = np; remove_subvol = true; + + } else { + r = image_path_lock(arg_directory, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock); + if (r == -EBUSY) { + log_error_errno(r, "Directory tree %s is currently busy.", arg_directory); + goto finish; + } + if (r < 0) { + log_error_errno(r, "Failed to lock %s: %m", arg_directory); + return r; + } + + if (arg_template) { + r = btrfs_subvol_snapshot(arg_template, arg_directory, arg_read_only, true); + if (r == -EEXIST) { + if (!arg_quiet) + log_info("Directory %s already exists, not populating from template %s.", arg_directory, arg_template); + } else if (r < 0) { + log_error_errno(r, "Couldn't create snapshot %s from %s: %m", arg_directory, arg_template); + goto finish; + } else { + if (!arg_quiet) + log_info("Populated %s from template %s.", arg_directory, arg_template); + } + } } if (arg_boot) { @@ -3455,6 +3599,16 @@ int main(int argc, char *argv[]) { assert(arg_image); assert(!arg_template); + r = image_path_lock(arg_image, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock); + if (r == -EBUSY) { + r = log_error_errno(r, "Disk image %s is currently busy.", arg_image); + goto finish; + } + if (r < 0) { + r = log_error_errno(r, "Failed to create image lock: %m"); + goto finish; + } + if (!mkdtemp(template)) { log_error_errno(errno, "Failed to create temporary directory: %m"); r = -errno; @@ -3917,9 +4071,10 @@ int main(int argc, char *argv[]) { _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; char last_char = 0; - sd_notify(false, - "READY=1\n" - "STATUS=Container running."); + sd_notifyf(false, + "READY=1\n" + "STATUS=Container running.\n" + "X_NSPAWN_LEADER_PID=" PID_FMT, pid); r = sd_event_new(&event); if (r < 0) { @@ -4034,7 +4189,7 @@ finish: if (arg_machine) { const char *p; - p = strappenda("/run/systemd/nspawn/propagate", arg_machine); + p = strappenda("/run/systemd/nspawn/propagate/", arg_machine); (void) rm_rf(p, false, true, false); }