diff options
author | Zbigniew Jędrzejewski-Szmek <zbyszek@in.waw.pl> | 2016-12-10 01:08:13 -0500 |
---|---|---|
committer | GitHub <noreply@github.com> | 2016-12-10 01:08:13 -0500 |
commit | 4a5567d5d6ab01974dd089eb8907fecd6aff4fcf (patch) | |
tree | 4f40cdd195f08022e2846d5c9757147ca436fede | |
parent | 2e1f244efd2dfc1a60d032bef3d88b9ba6e0444b (diff) | |
parent | 58abb66f4b9b0b3a16fe29211454d9936d35c35d (diff) |
Merge pull request #4795 from poettering/dissect
Generalize image dissection logic of nspawn, and make it useful for other tools.
-rw-r--r-- | .gitignore | 2 | ||||
-rw-r--r-- | Makefile.am | 46 | ||||
-rw-r--r-- | man/systemd-nspawn.xml | 17 | ||||
-rw-r--r-- | src/basic/fileio.c | 19 | ||||
-rw-r--r-- | src/basic/fileio.h | 2 | ||||
-rw-r--r-- | src/basic/fs-util.h | 7 | ||||
-rw-r--r-- | src/basic/rm-rf.h | 2 | ||||
-rw-r--r-- | src/core/umount.c | 19 | ||||
-rw-r--r-- | src/cryptsetup/cryptsetup.c | 2 | ||||
l--------- | src/dissect/Makefile | 1 | ||||
-rw-r--r-- | src/dissect/dissect.c | 271 | ||||
-rw-r--r-- | src/libudev/libudev.c | 4 | ||||
-rw-r--r-- | src/machine/image-dbus.c | 165 | ||||
-rw-r--r-- | src/machine/image-dbus.h | 1 | ||||
-rw-r--r-- | src/machine/machine-dbus.c | 40 | ||||
-rw-r--r-- | src/machine/machine-dbus.h | 2 | ||||
-rw-r--r-- | src/machine/machinectl.c | 33 | ||||
-rw-r--r-- | src/machine/machined-dbus.c | 52 | ||||
-rw-r--r-- | src/machine/org.freedesktop.machine1.conf | 8 | ||||
-rw-r--r-- | src/nspawn/nspawn.c | 797 | ||||
-rw-r--r-- | src/shared/dissect-image.c | 1078 | ||||
-rw-r--r-- | src/shared/dissect-image.h | 90 | ||||
-rw-r--r-- | src/shared/gpt.h | 17 | ||||
-rw-r--r-- | src/shared/loop-util.c | 166 | ||||
-rw-r--r-- | src/shared/loop-util.h | 41 | ||||
-rw-r--r-- | src/test/test-dissect-image.c | 66 |
26 files changed, 2234 insertions, 714 deletions
diff --git a/.gitignore b/.gitignore index 2e39f65e43..016ba625e3 100644 --- a/.gitignore +++ b/.gitignore @@ -67,6 +67,7 @@ /systemd-debug-generator /systemd-delta /systemd-detect-virt +/systemd-dissect /systemd-escape /systemd-export /systemd-firstboot @@ -180,6 +181,7 @@ /test-dhcp-option /test-dhcp-server /test-dhcp6-client +/test-dissect-image /test-dns-domain /test-dns-packet /test-dnssec diff --git a/Makefile.am b/Makefile.am index 99adfe4480..9bc18d39d6 100644 --- a/Makefile.am +++ b/Makefile.am @@ -404,6 +404,11 @@ rootlibexec_PROGRAMS = \ systemd-socket-proxyd \ systemd-update-done +if HAVE_BLKID +rootlibexec_PROGRAMS += \ + systemd-dissect +endif + if HAVE_UTMP rootlibexec_PROGRAMS += \ systemd-update-utmp @@ -1044,6 +1049,8 @@ libshared_la_SOURCES = \ src/shared/machine-image.h \ src/shared/machine-pool.c \ src/shared/machine-pool.h \ + src/shared/loop-util.c \ + src/shared/loop-util.h \ src/shared/resolve-util.c \ src/shared/resolve-util.h \ src/shared/bus-unit-util.c \ @@ -1055,7 +1062,9 @@ libshared_la_SOURCES = \ src/shared/fdset.c \ src/shared/fdset.h \ src/shared/nsflags.h \ - src/shared/nsflags.c + src/shared/nsflags.c \ + src/shared/dissect-image.c \ + src/shared/dissect-image.h if HAVE_UTMP libshared_la_SOURCES += \ @@ -1078,7 +1087,9 @@ libshared_la_CFLAGS = \ $(AM_CFLAGS) \ $(ACL_CFLAGS) \ $(LIBIDN_CFLAGS) \ - $(SECCOMP_CFLAGS) + $(SECCOMP_CFLAGS) \ + $(BLKID_CFLAGS) \ + $(LIBCRYPTSETUP_CFLAGS) libshared_la_LIBADD = \ libsystemd-internal.la \ @@ -1087,7 +1098,9 @@ libshared_la_LIBADD = \ libudev-internal.la \ $(ACL_LIBS) \ $(LIBIDN_LIBS) \ - $(SECCOMP_LIBS) + $(SECCOMP_LIBS) \ + $(BLKID_LIBS) \ + $(LIBCRYPTSETUP_LIBS) rootlibexec_LTLIBRARIES += \ libsystemd-shared.la @@ -1109,6 +1122,8 @@ libsystemd_shared_la_CFLAGS = \ $(ACL_CFLAGS) \ $(LIBIDN_CFLAGS) \ $(SECCOMP_CFLAGS) \ + $(BLKID_CFLAGS) \ + $(LIBCRYPTSETUP_CFLAGS) \ -fvisibility=default # We can't use libshared_la_LIBADD here because it would @@ -1120,7 +1135,9 @@ libsystemd_shared_la_LIBADD = \ $(libudev_internal_la_LIBADD) \ $(ACL_LIBS) \ $(LIBIDN_LIBS) \ - $(SECCOMP_LIBS) + $(SECCOMP_LIBS) \ + $(BLKID_LIBS) \ + $(LIBCRYPTSETUP_LIBS) libsystemd_shared_la_LDFLAGS = \ $(AM_LDFLAGS) \ @@ -1458,7 +1475,8 @@ manual_tests += \ test-btrfs \ test-acd \ test-ipv4ll-manual \ - test-ask-password-api + test-ask-password-api \ + test-dissect-image unsafe_tests = \ test-hostname \ @@ -2069,6 +2087,17 @@ test_ask_password_api_SOURCES = \ test_ask_password_api_LDADD = \ libsystemd-shared.la +test_dissect_image_SOURCES = \ + src/test/test-dissect-image.c + +test_dissect_image_CFLAGS = \ + $(AM_CFLAGS) \ + $(BLKID_CFLAGS) + +test_dissect_image_LDADD = \ + libsystemd-shared.la \ + $(BLKID_LIBS) + test_signal_util_SOURCES = \ src/test/test-signal-util.c @@ -3056,6 +3085,13 @@ systemd_notify_LDADD = \ libsystemd-shared.la # ------------------------------------------------------------------------------ +systemd_dissect_SOURCES = \ + src/dissect/dissect.c + +systemd_dissect_LDADD = \ + libsystemd-shared.la + +# ------------------------------------------------------------------------------ systemd_path_SOURCES = \ src/path/path.c diff --git a/man/systemd-nspawn.xml b/man/systemd-nspawn.xml index cd0a90d82f..2bc81ea1aa 100644 --- a/man/systemd-nspawn.xml +++ b/man/systemd-nspawn.xml @@ -235,17 +235,34 @@ identified by the partition types defined by the <ulink url="http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/">Discoverable Partitions Specification</ulink>.</para></listitem> + + <listitem><para>No partition table, and a single file system spanning the whole image.</para></listitem> </itemizedlist> <para>On GPT images, if an EFI System Partition (ESP) is discovered, it is automatically mounted to <filename>/efi</filename> (or <filename>/boot</filename> as fallback) in case a directory by this name exists and is empty.</para> + <para>Partitions encrypted with LUKS are automatically decrypted. Also, on GPT images dm-verity data integrity + hash partitions are set up if the root hash for them is specified using the <option>--root-hash=</option> + option.</para> + <para>Any other partitions, such as foreign partitions or swap partitions are not mounted. May not be specified together with <option>--directory=</option>, <option>--template=</option>.</para></listitem> </varlistentry> <varlistentry> + <term><option>--root-hash=</option></term> + + <listitem><para>Takes a data integrity (dm-verity) root hash specified in hexadecimal. This option enables data + integrity checks using dm-verity, if the used image contains the appropriate integrity data (see above). The + specified hash must match the root hash of integrity data, and is usually at least 256bits (and hence 64 + hexadecimal characters) long (in case of SHA256 for example). If this option is not specified, but a file with + the <filename>.roothash</filename> suffix is found next to the image file, bearing otherwise the same name the + root hash is read from it and automatically used.</para></listitem> + </varlistentry> + + <varlistentry> <term><option>-a</option></term> <term><option>--as-pid2</option></term> diff --git a/src/basic/fileio.c b/src/basic/fileio.c index 1615456659..c43b0583a4 100644 --- a/src/basic/fileio.c +++ b/src/basic/fileio.c @@ -1409,3 +1409,22 @@ int read_nul_string(FILE *f, char **ret) { return 0; } + +int mkdtemp_malloc(const char *template, char **ret) { + char *p; + + assert(template); + assert(ret); + + p = strdup(template); + if (!p) + return -ENOMEM; + + if (!mkdtemp(p)) { + free(p); + return -errno; + } + + *ret = p; + return 0; +} diff --git a/src/basic/fileio.h b/src/basic/fileio.h index b58c83e64a..17b38a5d60 100644 --- a/src/basic/fileio.h +++ b/src/basic/fileio.h @@ -88,3 +88,5 @@ int open_tmpfile_linkable(const char *target, int flags, char **ret_path); int link_tmpfile(int fd, const char *path, const char *target); int read_nul_string(FILE *f, char **ret); + +int mkdtemp_malloc(const char *template, char **ret); diff --git a/src/basic/fs-util.h b/src/basic/fs-util.h index 0d925c6b84..5fe5c71ff0 100644 --- a/src/basic/fs-util.h +++ b/src/basic/fs-util.h @@ -84,3 +84,10 @@ enum { }; int chase_symlinks(const char *path_with_prefix, const char *root, unsigned flags, char **ret); + +/* Useful for usage with _cleanup_(), removes a directory and frees the pointer */ +static inline void rmdir_and_free(char *p) { + (void) rmdir(p); + free(p); +} +DEFINE_TRIVIAL_CLEANUP_FUNC(char*, rmdir_and_free); diff --git a/src/basic/rm-rf.h b/src/basic/rm-rf.h index f693a5bb7c..e13f7003e3 100644 --- a/src/basic/rm-rf.h +++ b/src/basic/rm-rf.h @@ -33,8 +33,6 @@ int rm_rf(const char *path, RemoveFlags flags); /* Useful for usage with _cleanup_(), destroys a directory and frees the pointer */ static inline void rm_rf_physical_and_free(char *p) { - if (!p) - return; (void) rm_rf(p, REMOVE_ROOT|REMOVE_PHYSICAL); free(p); } diff --git a/src/core/umount.c b/src/core/umount.c index 1e5459ed80..2f4b12bdb9 100644 --- a/src/core/umount.c +++ b/src/core/umount.c @@ -344,24 +344,29 @@ static int delete_loopback(const char *device) { } static int delete_dm(dev_t devnum) { - _cleanup_close_ int fd = -1; - int r; + struct dm_ioctl dm = { - .version = {DM_VERSION_MAJOR, - DM_VERSION_MINOR, - DM_VERSION_PATCHLEVEL}, + .version = { + DM_VERSION_MAJOR, + DM_VERSION_MINOR, + DM_VERSION_PATCHLEVEL + }, .data_size = sizeof(dm), .dev = devnum, }; + _cleanup_close_ int fd = -1; + assert(major(devnum) != 0); fd = open("/dev/mapper/control", O_RDWR|O_CLOEXEC); if (fd < 0) return -errno; - r = ioctl(fd, DM_DEV_REMOVE, &dm); - return r >= 0 ? 0 : -errno; + if (ioctl(fd, DM_DEV_REMOVE, &dm) < 0) + return -errno; + + return 0; } static int mount_points_list_umount(MountPoint **head, bool *changed, bool log_error) { diff --git a/src/cryptsetup/cryptsetup.c b/src/cryptsetup/cryptsetup.c index 01e7ee9973..c7fec609df 100644 --- a/src/cryptsetup/cryptsetup.c +++ b/src/cryptsetup/cryptsetup.c @@ -651,7 +651,7 @@ int main(int argc, char *argv[]) { k = crypt_init(&cd, arg_header); } else k = crypt_init(&cd, argv[3]); - if (k) { + if (k != 0) { log_error_errno(k, "crypt_init() failed: %m"); goto finish; } diff --git a/src/dissect/Makefile b/src/dissect/Makefile new file mode 120000 index 0000000000..d0b0e8e008 --- /dev/null +++ b/src/dissect/Makefile @@ -0,0 +1 @@ +../Makefile
\ No newline at end of file diff --git a/src/dissect/dissect.c b/src/dissect/dissect.c new file mode 100644 index 0000000000..e3c96b7407 --- /dev/null +++ b/src/dissect/dissect.c @@ -0,0 +1,271 @@ +/*** + This file is part of systemd. + + Copyright 2016 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <fcntl.h> +#include <stdio.h> +#include <getopt.h> + +#include "architecture.h" +#include "dissect-image.h" +#include "hexdecoct.h" +#include "log.h" +#include "loop-util.h" +#include "string-util.h" +#include "util.h" + +static enum { + ACTION_DISSECT, + ACTION_MOUNT, +} arg_action = ACTION_DISSECT; +static const char *arg_image = NULL; +static const char *arg_path = NULL; +static DissectImageFlags arg_flags = DISSECT_IMAGE_DISCARD_ON_LOOP; +static void *arg_root_hash = NULL; +static size_t arg_root_hash_size = 0; + +static void help(void) { + printf("%s [OPTIONS...] IMAGE\n" + "%s [OPTIONS...] --mount IMAGE PATH\n" + "Dissect a file system OS image.\n\n" + " -h --help Show this help\n" + " --version Show package version\n" + " -m --mount Mount the image to the specified directory\n" + " -r --read-only Mount read-only\n" + " --discard=MODE Choose 'discard' mode (disabled, loop, all, crypto)\n" + " --root-hash=HASH Specify root hash for verity\n", + program_invocation_short_name, + program_invocation_short_name); +} + +static int parse_argv(int argc, char *argv[]) { + + enum { + ARG_VERSION = 0x100, + ARG_DISCARD, + ARG_ROOT_HASH, + }; + + static const struct option options[] = { + { "help", no_argument, NULL, 'h' }, + { "version", no_argument, NULL, ARG_VERSION }, + { "mount", no_argument, NULL, 'm' }, + { "read-only", no_argument, NULL, 'r' }, + { "discard", required_argument, NULL, ARG_DISCARD }, + { "root-hash", required_argument, NULL, ARG_ROOT_HASH }, + {} + }; + + int c, r; + + assert(argc >= 0); + assert(argv); + + while ((c = getopt_long(argc, argv, "hmr", options, NULL)) >= 0) { + + switch (c) { + + case 'h': + help(); + return 0; + + case ARG_VERSION: + return version(); + + case 'm': + arg_action = ACTION_MOUNT; + break; + + case 'r': + arg_flags |= DISSECT_IMAGE_READ_ONLY; + break; + + case ARG_DISCARD: + if (streq(optarg, "disabled")) + arg_flags &= ~(DISSECT_IMAGE_DISCARD_ON_LOOP|DISSECT_IMAGE_DISCARD|DISSECT_IMAGE_DISCARD_ON_CRYPTO); + else if (streq(optarg, "loop")) + arg_flags = (arg_flags & ~(DISSECT_IMAGE_DISCARD|DISSECT_IMAGE_DISCARD_ON_CRYPTO)) | DISSECT_IMAGE_DISCARD_ON_LOOP; + else if (streq(optarg, "all")) + arg_flags = (arg_flags & ~(DISSECT_IMAGE_DISCARD_ON_CRYPTO)) | DISSECT_IMAGE_DISCARD_ON_LOOP | DISSECT_IMAGE_DISCARD; + else if (streq(optarg, "crypt")) + arg_flags |= DISSECT_IMAGE_DISCARD_ON_LOOP | DISSECT_IMAGE_DISCARD | DISSECT_IMAGE_DISCARD_ON_CRYPTO; + else { + log_error("Unknown --discard= parameter: %s", optarg); + return -EINVAL; + } + + break; + + case ARG_ROOT_HASH: { + void *p; + size_t l; + + r = unhexmem(optarg, strlen(optarg), &p, &l); + if (r < 0) + return log_error_errno(r, "Failed to parse root hash: %s", optarg); + if (l < sizeof(sd_id128_t)) { + log_error("Root hash must be at least 128bit long: %s", optarg); + free(p); + return -EINVAL; + } + + free(arg_root_hash); + arg_root_hash = p; + arg_root_hash_size = l; + break; + } + + case '?': + return -EINVAL; + + default: + assert_not_reached("Unhandled option"); + } + + } + + switch (arg_action) { + + case ACTION_DISSECT: + if (optind + 1 != argc) { + log_error("Expected a file path as only argument."); + return -EINVAL; + } + + arg_image = argv[optind]; + arg_flags |= DISSECT_IMAGE_READ_ONLY; + break; + + case ACTION_MOUNT: + if (optind + 2 != argc) { + log_error("Expected a file path and mount point path as only arguments."); + return -EINVAL; + } + + arg_image = argv[optind]; + arg_path = argv[optind + 1]; + break; + + default: + assert_not_reached("Unknown action."); + } + + return 1; +} + +int main(int argc, char *argv[]) { + _cleanup_(loop_device_unrefp) LoopDevice *d = NULL; + _cleanup_(decrypted_image_unrefp) DecryptedImage *di = NULL; + _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL; + int r; + + log_parse_environment(); + log_open(); + + r = parse_argv(argc, argv); + if (r <= 0) + goto finish; + + r = loop_device_make_by_path(arg_image, (arg_flags & DISSECT_IMAGE_READ_ONLY) ? O_RDONLY : O_RDWR, &d); + if (r < 0) { + log_error_errno(r, "Failed to set up loopback device: %m"); + goto finish; + } + + r = dissect_image(d->fd, arg_root_hash, arg_root_hash_size, &m); + if (r == -ENOPKG) { + log_error_errno(r, "Couldn't identify a suitable partition table or file system in %s.", arg_image); + goto finish; + } + if (r == -EADDRNOTAVAIL) { + log_error_errno(r, "No root partition for specified root hash found in %s.", arg_image); + goto finish; + } + if (r < 0) { + log_error_errno(r, "Failed to dissect image: %m"); + goto finish; + } + + switch (arg_action) { + + case ACTION_DISSECT: { + unsigned i; + + for (i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) { + DissectedPartition *p = m->partitions + i; + int k; + + if (!p->found) + continue; + + printf("Found %s '%s' partition", + p->rw ? "writable" : "read-only", + partition_designator_to_string(i)); + + if (p->fstype) + printf(" of type %s", p->fstype); + + if (p->architecture != _ARCHITECTURE_INVALID) + printf(" for %s", architecture_to_string(p->architecture)); + + k = PARTITION_VERITY_OF(i); + if (k >= 0) + printf(" %s verity", m->partitions[k].found ? "with" : "without"); + + if (p->partno >= 0) + printf(" on partition #%i", p->partno); + + if (p->node) + printf(" (%s)", p->node); + + putchar('\n'); + } + + break; + } + + case ACTION_MOUNT: + r = dissected_image_decrypt_interactively(m, NULL, arg_root_hash, arg_root_hash_size, arg_flags, &di); + if (r < 0) + goto finish; + + r = dissected_image_mount(m, arg_path, arg_flags); + if (r < 0) { + log_error_errno(r, "Failed to mount image: %m"); + goto finish; + } + + if (di) { + r = decrypted_image_relinquish(di); + if (r < 0) { + log_error_errno(r, "Failed to relinquish DM devices: %m"); + goto finish; + } + } + + loop_device_relinquish(d); + break; + + default: + assert_not_reached("Unknown action."); + } + +finish: + free(arg_root_hash); + return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS; +} diff --git a/src/libudev/libudev.c b/src/libudev/libudev.c index 57ce749e07..d8e13288b0 100644 --- a/src/libudev/libudev.c +++ b/src/libudev/libudev.c @@ -97,8 +97,10 @@ _public_ struct udev *udev_new(void) { _cleanup_fclose_ FILE *f = NULL; udev = new0(struct udev, 1); - if (udev == NULL) + if (!udev) { + errno = -ENOMEM; return NULL; + } udev->refcount = 1; f = fopen("/etc/udev/udev.conf", "re"); diff --git a/src/machine/image-dbus.c b/src/machine/image-dbus.c index 867bbc467b..e2fb882393 100644 --- a/src/machine/image-dbus.c +++ b/src/machine/image-dbus.c @@ -17,14 +17,23 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ +#include <sys/mount.h> + #include "alloc-util.h" #include "bus-label.h" #include "bus-util.h" +#include "copy.h" +#include "dissect-image.h" #include "fd-util.h" +#include "fileio.h" +#include "fs-util.h" #include "image-dbus.h" #include "io-util.h" +#include "loop-util.h" #include "machine-image.h" +#include "mount-util.h" #include "process-util.h" +#include "raw-clone.h" #include "strv.h" #include "user-util.h" @@ -279,6 +288,161 @@ int bus_image_method_set_limit( return sd_bus_reply_method_return(message, NULL); } +#define EXIT_NOT_FOUND 2 + +static int directory_image_get_os_release(Image *image, char ***ret, sd_bus_error *error) { + + _cleanup_free_ char *path = NULL; + _cleanup_close_ int fd = -1; + int r; + + assert(image); + assert(ret); + + r = chase_symlinks("/etc/os-release", image->path, CHASE_PREFIX_ROOT, &path); + if (r == -ENOENT) + r = chase_symlinks("/usr/lib/os-release", image->path, CHASE_PREFIX_ROOT, &path); + if (r == -ENOENT) + return sd_bus_error_setf(error, SD_BUS_ERROR_FAILED, "Image does not contain OS release information"); + if (r < 0) + return sd_bus_error_set_errnof(error, r, "Failed to resolve %s: %m", image->path); + + r = load_env_file_pairs(NULL, path, NULL, ret); + if (r < 0) + return sd_bus_error_set_errnof(error, r, "Failed to open %s: %m", path); + + return 0; +} + +static int raw_image_get_os_release(Image *image, char ***ret, sd_bus_error *error) { + _cleanup_(rmdir_and_freep) char *t = NULL; + _cleanup_(loop_device_unrefp) LoopDevice *d = NULL; + _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL; + _cleanup_(sigkill_waitp) pid_t child = 0; + _cleanup_close_pair_ int pair[2] = { -1, -1 }; + _cleanup_fclose_ FILE *f = NULL; + _cleanup_strv_free_ char **v = NULL; + siginfo_t si; + int r; + + assert(image); + assert(ret); + + r = mkdtemp_malloc("/tmp/machined-root-XXXXXX", &t); + if (r < 0) + return sd_bus_error_set_errnof(error, r, "Failed to create temporary directory: %m"); + + r = loop_device_make_by_path(image->path, O_RDONLY, &d); + if (r < 0) + return sd_bus_error_set_errnof(error, r, "Failed to set up loop block device for %s: %m", image->path); + + r = dissect_image(d->fd, NULL, 0, &m); + if (r == -ENOPKG) + return sd_bus_error_set_errnof(error, r, "Disk image %s not understood: %m", image->path); + if (r < 0) + return sd_bus_error_set_errnof(error, r, "Failed to dissect image %s: %m", image->path); + + if (pipe2(pair, O_CLOEXEC) < 0) + return sd_bus_error_set_errnof(error, errno, "Failed to create communication pipe: %m"); + + child = raw_clone(SIGCHLD|CLONE_NEWNS); + if (child < 0) + return sd_bus_error_set_errnof(error, errno, "Failed to fork(): %m"); + + if (child == 0) { + int fd; + + pair[0] = safe_close(pair[0]); + + /* Make sure we never propagate to the host */ + if (mount(NULL, "/", NULL, MS_SLAVE | MS_REC, NULL) < 0) + _exit(EXIT_FAILURE); + + r = dissected_image_mount(m, t, DISSECT_IMAGE_READ_ONLY); + if (r < 0) + _exit(EXIT_FAILURE); + + r = mount_move_root(t); + if (r < 0) + _exit(EXIT_FAILURE); + + fd = open("/etc/os-release", O_RDONLY|O_CLOEXEC|O_NOCTTY); + if (fd < 0 && errno == ENOENT) { + fd = open("/usr/lib/os-release", O_RDONLY|O_CLOEXEC|O_NOCTTY); + if (fd < 0 && errno == ENOENT) + _exit(EXIT_NOT_FOUND); + } + if (fd < 0) + _exit(EXIT_FAILURE); + + r = copy_bytes(fd, pair[1], (uint64_t) -1, false); + if (r < 0) + _exit(EXIT_FAILURE); + + _exit(EXIT_SUCCESS); + } + + pair[1] = safe_close(pair[1]); + + f = fdopen(pair[0], "re"); + if (!f) + return -errno; + + pair[0] = -1; + + r = load_env_file_pairs(f, "os-release", NULL, &v); + if (r < 0) + return r; + + r = wait_for_terminate(child, &si); + if (r < 0) + return sd_bus_error_set_errnof(error, r, "Failed to wait for child: %m"); + child = 0; + if (si.si_code == CLD_EXITED && si.si_status == EXIT_NOT_FOUND) + return sd_bus_error_setf(error, SD_BUS_ERROR_FAILED, "Image does not contain OS release information"); + if (si.si_code != CLD_EXITED || si.si_status != EXIT_SUCCESS) + return sd_bus_error_setf(error, SD_BUS_ERROR_FAILED, "Child died abnormally."); + + *ret = v; + v = NULL; + + return 0; +} + +int bus_image_method_get_os_release( + sd_bus_message *message, + void *userdata, + sd_bus_error *error) { + + _cleanup_release_lock_file_ LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT; + _cleanup_strv_free_ char **v = NULL; + Image *image = userdata; + int r; + + r = image_path_lock(image->path, LOCK_SH|LOCK_NB, &tree_global_lock, &tree_local_lock); + if (r < 0) + return sd_bus_error_set_errnof(error, r, "Failed to lock image: %m"); + + switch (image->type) { + + case IMAGE_DIRECTORY: + case IMAGE_SUBVOLUME: + r = directory_image_get_os_release(image, &v, error); + break; + + case IMAGE_RAW: + r = raw_image_get_os_release(image, &v, error); + break; + + default: + assert_not_reached("Unknown image type"); + } + if (r < 0) + return r; + + return bus_reply_pair_array(message, v); +} + const sd_bus_vtable image_vtable[] = { SD_BUS_VTABLE_START(0), SD_BUS_PROPERTY("Name", "s", NULL, offsetof(Image, name), 0), @@ -296,6 +460,7 @@ const sd_bus_vtable image_vtable[] = { SD_BUS_METHOD("Clone", "sb", NULL, bus_image_method_clone, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("MarkReadOnly", "b", NULL, bus_image_method_mark_read_only, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("SetLimit", "t", NULL, bus_image_method_set_limit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetOSRelease", NULL, "a{ss}", bus_image_method_get_os_release, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_VTABLE_END }; diff --git a/src/machine/image-dbus.h b/src/machine/image-dbus.h index b62da996c6..bc8a6c3400 100644 --- a/src/machine/image-dbus.h +++ b/src/machine/image-dbus.h @@ -33,3 +33,4 @@ int bus_image_method_rename(sd_bus_message *message, void *userdata, sd_bus_erro int bus_image_method_clone(sd_bus_message *message, void *userdata, sd_bus_error *error); int bus_image_method_mark_read_only(sd_bus_message *message, void *userdata, sd_bus_error *error); int bus_image_method_set_limit(sd_bus_message *message, void *userdata, sd_bus_error *error); +int bus_image_method_get_os_release(sd_bus_message *message, void *userdata, sd_bus_error *error); diff --git a/src/machine/machine-dbus.c b/src/machine/machine-dbus.c index 28e4867cb3..af745b6567 100644 --- a/src/machine/machine-dbus.c +++ b/src/machine/machine-dbus.c @@ -356,11 +356,11 @@ int bus_machine_method_get_addresses(sd_bus_message *message, void *userdata, sd return sd_bus_send(NULL, reply, NULL); } +#define EXIT_NOT_FOUND 2 + int bus_machine_method_get_os_release(sd_bus_message *message, void *userdata, sd_bus_error *error) { - _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; _cleanup_strv_free_ char **l = NULL; Machine *m = userdata; - char **k, **v; int r; assert(message); @@ -394,7 +394,7 @@ int bus_machine_method_get_os_release(sd_bus_message *message, void *userdata, s return sd_bus_error_set_errnof(error, errno, "Failed to fork(): %m"); if (child == 0) { - _cleanup_close_ int fd = -1; + int fd = -1; pair[0] = safe_close(pair[0]); @@ -402,12 +402,14 @@ int bus_machine_method_get_os_release(sd_bus_message *message, void *userdata, s if (r < 0) _exit(EXIT_FAILURE); - fd = open("/etc/os-release", O_RDONLY|O_CLOEXEC); - if (fd < 0) { - fd = open("/usr/lib/os-release", O_RDONLY|O_CLOEXEC); - if (fd < 0) - _exit(EXIT_FAILURE); + fd = open("/etc/os-release", O_RDONLY|O_CLOEXEC|O_NOCTTY); + if (fd < 0 && errno == ENOENT) { + fd = open("/usr/lib/os-release", O_RDONLY|O_CLOEXEC|O_NOCTTY); + if (fd < 0 && errno == ENOENT) + _exit(EXIT_NOT_FOUND); } + if (fd < 0) + _exit(EXIT_FAILURE); r = copy_bytes(fd, pair[1], (uint64_t) -1, false); if (r < 0) @@ -431,6 +433,8 @@ int bus_machine_method_get_os_release(sd_bus_message *message, void *userdata, s r = wait_for_terminate(child, &si); if (r < 0) return sd_bus_error_set_errnof(error, r, "Failed to wait for child: %m"); + if (si.si_code == CLD_EXITED && si.si_status == EXIT_NOT_FOUND) + return sd_bus_error_setf(error, SD_BUS_ERROR_FAILED, "Machine does not contain OS release information"); if (si.si_code != CLD_EXITED || si.si_status != EXIT_SUCCESS) return sd_bus_error_setf(error, SD_BUS_ERROR_FAILED, "Child died abnormally."); @@ -441,25 +445,7 @@ int bus_machine_method_get_os_release(sd_bus_message *message, void *userdata, s return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Requesting OS release data is only supported on container machines."); } - r = sd_bus_message_new_method_return(message, &reply); - if (r < 0) - return r; - - r = sd_bus_message_open_container(reply, 'a', "{ss}"); - if (r < 0) - return r; - - STRV_FOREACH_PAIR(k, v, l) { - r = sd_bus_message_append(reply, "{ss}", *k, *v); - if (r < 0) - return r; - } - - r = sd_bus_message_close_container(reply); - if (r < 0) - return r; - - return sd_bus_send(NULL, reply, NULL); + return bus_reply_pair_array(message, l); } int bus_machine_method_open_pty(sd_bus_message *message, void *userdata, sd_bus_error *error) { diff --git a/src/machine/machine-dbus.h b/src/machine/machine-dbus.h index 241b23c7ec..c513783480 100644 --- a/src/machine/machine-dbus.h +++ b/src/machine/machine-dbus.h @@ -42,3 +42,5 @@ int bus_machine_method_open_root_directory(sd_bus_message *message, void *userda int machine_send_signal(Machine *m, bool new_machine); int machine_send_create_reply(Machine *m, sd_bus_error *error); + +int bus_reply_pair_array(sd_bus_message *m, char **l); diff --git a/src/machine/machinectl.c b/src/machine/machinectl.c index 9c754b4327..3294ea7821 100644 --- a/src/machine/machinectl.c +++ b/src/machine/machinectl.c @@ -138,7 +138,7 @@ static void clean_machine_info(MachineInfo *machines, size_t n_machines) { free(machines); } -static int get_os_release_property(sd_bus *bus, const char *name, const char *query, ...) { +static int call_get_os_release(sd_bus *bus, const char *method, const char *name, const char *query, ...) { _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; const char *k, *v, *iter, **query_res = NULL; size_t count = 0, awaited_args = 0; @@ -153,12 +153,13 @@ static int get_os_release_property(sd_bus *bus, const char *name, const char *qu awaited_args++; query_res = newa0(const char *, awaited_args); - r = sd_bus_call_method(bus, - "org.freedesktop.machine1", - "/org/freedesktop/machine1", - "org.freedesktop.machine1.Manager", - "GetMachineOSRelease", - NULL, &reply, "s", name); + r = sd_bus_call_method( + bus, + "org.freedesktop.machine1", + "/org/freedesktop/machine1", + "org.freedesktop.machine1.Manager", + method, + NULL, &reply, "s", name); if (r < 0) return r; @@ -193,7 +194,7 @@ static int get_os_release_property(sd_bus *bus, const char *name, const char *qu val = strdup(query_res[count]); if (!val) { va_end(ap); - return log_oom(); + return -ENOMEM; } *out = val; } @@ -249,8 +250,12 @@ static int list_machines(int argc, char *argv[], void *userdata) { machines[n_machines].os = NULL; machines[n_machines].version_id = NULL; - r = get_os_release_property(bus, name, - "ID\0" "VERSION_ID\0", + r = call_get_os_release( + bus, + "GetMachineOSRelease", + name, + "ID\0" + "VERSION_ID\0", &machines[n_machines].os, &machines[n_machines].version_id); if (r < 0) @@ -610,7 +615,7 @@ static int print_addresses(sd_bus *bus, const char *name, int ifi, const char *p return 0; } -static int print_os_release(sd_bus *bus, const char *name, const char *prefix) { +static int print_os_release(sd_bus *bus, const char *method, const char *name, const char *prefix) { _cleanup_free_ char *pretty = NULL; int r; @@ -618,7 +623,7 @@ static int print_os_release(sd_bus *bus, const char *name, const char *prefix) { assert(name); assert(prefix); - r = get_os_release_property(bus, name, "PRETTY_NAME\0", &pretty, NULL); + r = call_get_os_release(bus, method, name, "PRETTY_NAME\0", &pretty, NULL); if (r < 0) return r; @@ -729,7 +734,7 @@ static void print_machine_status_info(sd_bus *bus, MachineStatusInfo *i) { "\n\t ", ALL_IP_ADDRESSES); - print_os_release(bus, i->name, "\t OS: "); + print_os_release(bus, "GetMachineOSRelease", i->name, "\t OS: "); if (i->unit) { printf("\t Unit: %s\n", i->unit); @@ -927,6 +932,8 @@ static void print_image_status_info(sd_bus *bus, ImageStatusInfo *i) { if (i->path) printf("\t Path: %s\n", i->path); + print_os_release(bus, "GetImageOSRelease", i->name, "\t OS: "); + printf("\t RO: %s%s%s\n", i->read_only ? ansi_highlight_red() : "", i->read_only ? "read-only" : "writable", diff --git a/src/machine/machined-dbus.c b/src/machine/machined-dbus.c index 3ee3938ebb..fd9e5b56fc 100644 --- a/src/machine/machined-dbus.c +++ b/src/machine/machined-dbus.c @@ -825,6 +825,30 @@ static int method_mark_image_read_only(sd_bus_message *message, void *userdata, return bus_image_method_mark_read_only(message, i, error); } +static int method_get_image_os_release(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_(image_unrefp) Image *i = NULL; + const char *name; + int r; + + assert(message); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + if (!image_name_is_valid(name)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Image name '%s' is invalid.", name); + + r = image_find(name, &i); + if (r < 0) + return r; + if (r == 0) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_IMAGE, "No image '%s' known", name); + + i->userdata = userdata; + return bus_image_method_get_os_release(message, i, error); +} + static int clean_pool_done(Operation *operation, int ret, sd_bus_error *error) { _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; _cleanup_fclose_ FILE *f = NULL; @@ -1396,6 +1420,7 @@ const sd_bus_vtable manager_vtable[] = { SD_BUS_METHOD("RenameImage", "ss", NULL, method_rename_image, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("CloneImage", "ssb", NULL, method_clone_image, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("MarkImageReadOnly", "sb", NULL, method_mark_image_read_only, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetImageOSRelease", "s", "a{ss}", method_get_image_os_release, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("SetPoolLimit", "t", NULL, method_set_pool_limit, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("SetImageLimit", "st", NULL, method_set_image_limit, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("CleanPool", "s", "a(st)", method_clean_pool, SD_BUS_VTABLE_UNPRIVILEGED), @@ -1804,3 +1829,30 @@ int manager_add_machine(Manager *m, const char *name, Machine **_machine) { return 0; } + +int bus_reply_pair_array(sd_bus_message *m, char **l) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + char **k, **v; + int r; + + r = sd_bus_message_new_method_return(m, &reply); + if (r < 0) + return r; + + r = sd_bus_message_open_container(reply, 'a', "{ss}"); + if (r < 0) + return r; + + STRV_FOREACH_PAIR(k, v, l) { + r = sd_bus_message_append(reply, "{ss}", *k, *v); + if (r < 0) + return r; + } + + r = sd_bus_message_close_container(reply); + if (r < 0) + return r; + + return sd_bus_send(NULL, reply, NULL); + +} diff --git a/src/machine/org.freedesktop.machine1.conf b/src/machine/org.freedesktop.machine1.conf index 562b9d3cc0..82ebfba50c 100644 --- a/src/machine/org.freedesktop.machine1.conf +++ b/src/machine/org.freedesktop.machine1.conf @@ -118,6 +118,10 @@ <allow send_destination="org.freedesktop.machine1" send_interface="org.freedesktop.machine1.Manager" + send_member="GetImageOSRelease"/> + + <allow send_destination="org.freedesktop.machine1" + send_interface="org.freedesktop.machine1.Manager" send_member="CleanPool"/> <allow send_destination="org.freedesktop.machine1" @@ -192,6 +196,10 @@ send_interface="org.freedesktop.machine1.Image" send_member="MarkReadOnly"/> + <allow send_destination="org.freedesktop.machine1" + send_interface="org.freedesktop.machine1.Image" + send_member="GetOSRelease"/> + <allow receive_sender="org.freedesktop.machine1"/> </policy> diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index f3303756a3..d701f2158d 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -53,6 +53,7 @@ #include "cgroup-util.h" #include "copy.h" #include "dev-setup.h" +#include "dissect-image.h" #include "env-util.h" #include "fd-util.h" #include "fdset.h" @@ -60,9 +61,11 @@ #include "format-util.h" #include "fs-util.h" #include "gpt.h" +#include "hexdecoct.h" #include "hostname-util.h" #include "id128-util.h" #include "log.h" +#include "loop-util.h" #include "loopback-setup.h" #include "machine-image.h" #include "macro.h" @@ -198,6 +201,8 @@ static bool arg_notify_ready = false; static bool arg_use_cgns = true; static unsigned long arg_clone_ns_flags = CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS; static MountSettingsMask arg_mount_settings = MOUNT_APPLY_APIVFS_RO; +static void *arg_root_hash = NULL; +static size_t arg_root_hash_size = 0; static void help(void) { printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" @@ -211,6 +216,7 @@ static void help(void) { " -x --ephemeral Run container with snapshot of root directory, and\n" " remove it after exit\n" " -i --image=PATH File system device or disk image for the container\n" + " --root-hash=HASH Specify verity root hash\n" " -a --as-pid2 Maintain a stub init as PID1, invoke binary as PID2\n" " -b --boot Boot up full system (i.e. invoke init)\n" " --chdir=PATH Set working directory in the container\n" @@ -422,6 +428,7 @@ static int parse_argv(int argc, char *argv[]) { ARG_CHDIR, ARG_PRIVATE_USERS_CHOWN, ARG_NOTIFY_READY, + ARG_ROOT_HASH, }; static const struct option options[] = { @@ -471,6 +478,7 @@ static int parse_argv(int argc, char *argv[]) { { "settings", required_argument, NULL, ARG_SETTINGS }, { "chdir", required_argument, NULL, ARG_CHDIR }, { "notify-ready", required_argument, NULL, ARG_NOTIFY_READY }, + { "root-hash", required_argument, NULL, ARG_ROOT_HASH }, {} }; @@ -1014,6 +1022,25 @@ static int parse_argv(int argc, char *argv[]) { arg_settings_mask |= SETTING_NOTIFY_READY; break; + case ARG_ROOT_HASH: { + void *k; + size_t l; + + r = unhexmem(optarg, strlen(optarg), &k, &l); + if (r < 0) + return log_error_errno(r, "Failed to parse root hash: %s", optarg); + if (l < sizeof(sd_id128_t)) { + log_error("Root hash must be at least 128bit long: %s", optarg); + free(k); + return -EINVAL; + } + + free(arg_root_hash); + arg_root_hash = k; + arg_root_hash_size = l; + break; + } + case '?': return -EINVAL; @@ -1771,546 +1798,6 @@ static int setup_propagate(const char *root) { return mount_verbose(LOG_ERR, NULL, q, NULL, MS_SLAVE, NULL); } -static int setup_image(char **device_path, int *loop_nr) { - struct loop_info64 info = { - .lo_flags = LO_FLAGS_AUTOCLEAR|LO_FLAGS_PARTSCAN - }; - _cleanup_close_ int fd = -1, control = -1, loop = -1; - _cleanup_free_ char* loopdev = NULL; - struct stat st; - int r, nr; - - assert(device_path); - assert(loop_nr); - assert(arg_image); - - fd = open(arg_image, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY); - if (fd < 0) - return log_error_errno(errno, "Failed to open %s: %m", arg_image); - - if (fstat(fd, &st) < 0) - return log_error_errno(errno, "Failed to stat %s: %m", arg_image); - - if (S_ISBLK(st.st_mode)) { - char *p; - - p = strdup(arg_image); - if (!p) - return log_oom(); - - *device_path = p; - - *loop_nr = -1; - - r = fd; - fd = -1; - - return r; - } - - if (!S_ISREG(st.st_mode)) { - log_error("%s is not a regular file or block device.", arg_image); - return -EINVAL; - } - - control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); - if (control < 0) - return log_error_errno(errno, "Failed to open /dev/loop-control: %m"); - - nr = ioctl(control, LOOP_CTL_GET_FREE); - if (nr < 0) - return log_error_errno(errno, "Failed to allocate loop device: %m"); - - if (asprintf(&loopdev, "/dev/loop%i", nr) < 0) - return log_oom(); - - loop = open(loopdev, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY); - if (loop < 0) - return log_error_errno(errno, "Failed to open loop device %s: %m", loopdev); - - if (ioctl(loop, LOOP_SET_FD, fd) < 0) - return log_error_errno(errno, "Failed to set loopback file descriptor on %s: %m", loopdev); - - if (arg_read_only) - info.lo_flags |= LO_FLAGS_READ_ONLY; - - if (ioctl(loop, LOOP_SET_STATUS64, &info) < 0) - return log_error_errno(errno, "Failed to set loopback settings on %s: %m", loopdev); - - *device_path = loopdev; - loopdev = NULL; - - *loop_nr = nr; - - r = loop; - loop = -1; - - return r; -} - -#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 single 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, - char **home_device, bool *home_device_rw, - char **srv_device, bool *srv_device_rw, - char **esp_device, - bool *secondary) { - -#ifdef HAVE_BLKID - int home_nr = -1, srv_nr = -1, esp_nr = -1; -#ifdef GPT_ROOT_NATIVE - int root_nr = -1; -#endif -#ifdef GPT_ROOT_SECONDARY - int secondary_root_nr = -1; -#endif - _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL, *esp = 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, generic_rw = true; - bool is_gpt, is_mbr, multiple_generic = false; - const char *pttype = NULL; - blkid_partlist pl; - struct stat st; - unsigned i; - int r; - - assert(fd >= 0); - assert(root_device); - assert(home_device); - assert(srv_device); - assert(esp_device); - assert(secondary); - assert(arg_image); - - b = blkid_new_probe(); - if (!b) - return log_oom(); - - errno = 0; - r = blkid_probe_set_device(b, fd, 0, 0); - if (r != 0) { - if (errno == 0) - return log_oom(); - - return log_error_errno(errno, "Failed to set device on blkid probe: %m"); - } - - blkid_probe_enable_partitions(b, 1); - blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS); - - errno = 0; - r = blkid_do_safeprobe(b); - if (r == -2 || r == 1) { - 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) - errno = EIO; - return log_error_errno(errno, "Failed to probe: %m"); - } - - (void) blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL); - - 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; - } - - errno = 0; - pl = blkid_probe_get_partitions(b); - if (!pl) { - if (errno == 0) - return log_oom(); - - log_error("Failed to list partitions of %s", arg_image); - return -errno; - } - - udev = udev_new(); - if (!udev) - return log_oom(); - - 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(); - - for (i = 0;; i++) { - int n, m; - - if (i >= 10) { - log_error("Kernel partitions never appeared."); - return -ENXIO; - } - - e = udev_enumerate_new(udev); - if (!e) - return log_oom(); - - r = udev_enumerate_add_match_parent(e, d); - if (r < 0) - return log_oom(); - - r = udev_enumerate_scan_devices(e); - if (r < 0) - return log_error_errno(r, "Failed to scan for partition devices of %s: %m", arg_image); - - /* Count the partitions enumerated by the kernel */ - n = 0; - first = udev_enumerate_get_list_entry(e); - udev_list_entry_foreach(item, first) - n++; - - /* Count the partitions enumerated by blkid */ - m = blkid_partlist_numof_partitions(pl); - if (n == m + 1) - break; - if (n > m + 1) { - log_error("blkid and kernel partition list do not match."); - return -EIO; - } - if (n < m + 1) { - unsigned j; - - /* The kernel has probed fewer partitions than - * blkid? Maybe the kernel prober is still - * running or it got EBUSY because udev - * already opened the device. Let's reprobe - * the device, which is a synchronous call - * that waits until probing is complete. */ - - for (j = 0; j < 20; j++) { - - r = ioctl(fd, BLKRRPART, 0); - if (r < 0) - r = -errno; - if (r >= 0 || r != -EBUSY) - break; - - /* If something else has the device - * open, such as an udev rule, the - * ioctl will return EBUSY. Since - * there's no way to wait until it - * isn't busy anymore, let's just wait - * a bit, and try again. - * - * This is really something they - * should fix in the kernel! */ - - usleep(50 * USEC_PER_MSEC); - } - - if (r < 0) - return log_error_errno(r, "Failed to reread partition table: %m"); - } - - e = udev_enumerate_unref(e); - } - - first = udev_enumerate_get_list_entry(e); - udev_list_entry_foreach(item, first) { - _cleanup_udev_device_unref_ struct udev_device *q; - const char *node; - unsigned long long flags; - blkid_partition pp; - dev_t qn; - int nr; - - errno = 0; - q = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item)); - if (!q) { - if (!errno) - errno = ENOMEM; - - return log_error_errno(errno, "Failed to get partition device of %s: %m", arg_image); - } - - qn = udev_device_get_devnum(q); - if (major(qn) == 0) - continue; - - if (st.st_rdev == qn) - continue; - - node = udev_device_get_devnode(q); - if (!node) - continue; - - pp = blkid_partlist_devno_to_partition(pl, qn); - if (!pp) - continue; - - flags = blkid_partition_get_flags(pp); - - nr = blkid_partition_get_partno(pp); - if (nr < 0) - continue; - - if (is_gpt) { - sd_id128_t type_id; - const char *stype; - - if (flags & GPT_FLAG_NO_AUTO) - continue; - - stype = blkid_partition_get_type_string(pp); - if (!stype) - continue; - - if (sd_id128_from_string(stype, &type_id) < 0) - continue; - - if (sd_id128_equal(type_id, GPT_HOME)) { - - if (home && nr >= home_nr) - continue; - - home_nr = nr; - home_rw = !(flags & GPT_FLAG_READ_ONLY); - - r = free_and_strdup(&home, node); - if (r < 0) - 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(); - } else if (sd_id128_equal(type_id, GPT_ESP)) { - - if (esp && nr >= esp_nr) - continue; - - esp_nr = nr; - - r = free_and_strdup(&esp, node); - if (r < 0) - return log_oom(); - } -#ifdef GPT_ROOT_NATIVE - else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { - - if (root && nr >= root_nr) - continue; - - root_nr = nr; - root_rw = !(flags & GPT_FLAG_READ_ONLY); - - 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)) { - - 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 (flags != 0x80) /* Bootable flag */ - continue; - - type = blkid_partition_get_type(pp); - if (type != 0x83) /* Linux partition */ - continue; - - if (generic) - multiple_generic = true; - else { - generic_rw = true; - - r = free_and_strdup(&root, node); - if (r < 0) - return log_oom(); - } - } - } - - if (root) { - *root_device = root; - root = NULL; - - *root_device_rw = root_rw; - *secondary = false; - } else if (secondary_root) { - *root_device = secondary_root; - secondary_root = NULL; - - *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) { - *home_device = home; - home = NULL; - - *home_device_rw = home_rw; - } - - if (srv) { - *srv_device = srv; - srv = NULL; - - *srv_device_rw = srv_rw; - } - - if (esp) { - *esp_device = esp; - esp = NULL; - } - - return 0; -#else - log_error("--image= is not supported, compiled without blkid support."); - return -EOPNOTSUPP; -#endif -} - -static int mount_device(const char *what, const char *where, const char *directory, bool rw) { -#ifdef HAVE_BLKID - _cleanup_blkid_free_probe_ blkid_probe b = NULL; - const char *fstype, *p, *options; - int r; - - assert(what); - assert(where); - - if (arg_read_only) - rw = false; - - if (directory) - p = strjoina(where, directory); - else - p = where; - - errno = 0; - b = blkid_new_probe_from_filename(what); - if (!b) { - if (errno == 0) - return log_oom(); - return log_error_errno(errno, "Failed to allocate prober for %s: %m", what); - } - - blkid_probe_enable_superblocks(b, 1); - blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE); - - errno = 0; - r = blkid_do_safeprobe(b); - if (r == -1 || r == 1) { - log_error("Cannot determine file system type of %s", what); - return -EINVAL; - } else if (r != 0) { - if (errno == 0) - errno = EIO; - return log_error_errno(errno, "Failed to probe %s: %m", what); - } - - errno = 0; - if (blkid_probe_lookup_value(b, "TYPE", &fstype, NULL) < 0) { - if (errno == 0) - errno = EINVAL; - log_error("Failed to determine file system type of %s", what); - return -errno; - } - - if (streq(fstype, "crypto_LUKS")) { - log_error("nspawn currently does not support LUKS disk images."); - return -EOPNOTSUPP; - } - - /* If this is a loopback device then let's mount the image with discard, so that the underlying file remains - * sparse when possible. */ - if (STR_IN_SET(fstype, "btrfs", "ext4", "vfat", "xfs")) { - const char *l; - - l = path_startswith(what, "/dev"); - if (l && startswith(l, "loop")) - options = "discard"; - } - - return mount_verbose(LOG_ERR, what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), options); -#else - log_error("--image= is not supported, compiled without blkid support."); - return -EOPNOTSUPP; -#endif -} - static int setup_machine_id(const char *directory) { const char *etc_machine_id; sd_id128_t id; @@ -2370,83 +1857,6 @@ static int recursive_chown(const char *directory, uid_t shift, uid_t range) { return r; } -static int mount_devices( - const char *where, - const char *root_device, bool root_device_rw, - const char *home_device, bool home_device_rw, - const char *srv_device, bool srv_device_rw, - const char *esp_device) { - int r; - - assert(where); - - if (root_device) { - r = mount_device(root_device, arg_directory, NULL, root_device_rw); - if (r < 0) - return log_error_errno(r, "Failed to mount root directory: %m"); - } - - if (home_device) { - r = mount_device(home_device, arg_directory, "/home", home_device_rw); - if (r < 0) - return log_error_errno(r, "Failed to mount home directory: %m"); - } - - if (srv_device) { - r = mount_device(srv_device, arg_directory, "/srv", srv_device_rw); - if (r < 0) - return log_error_errno(r, "Failed to mount server data directory: %m"); - } - - if (esp_device) { - const char *mp, *x; - - /* Mount the ESP to /efi if it exists and is empty. If it doesn't exist, use /boot instead. */ - - mp = "/efi"; - x = strjoina(arg_directory, mp); - r = dir_is_empty(x); - if (r == -ENOENT) { - mp = "/boot"; - x = strjoina(arg_directory, mp); - r = dir_is_empty(x); - } - - if (r > 0) { - r = mount_device(esp_device, arg_directory, mp, true); - if (r < 0) - return log_error_errno(r, "Failed to mount ESP: %m"); - } - } - - return 0; -} - -static void loop_remove(int nr, int *image_fd) { - _cleanup_close_ int control = -1; - int r; - - if (nr < 0) - return; - - if (image_fd && *image_fd >= 0) { - r = ioctl(*image_fd, LOOP_CLR_FD); - if (r < 0) - log_debug_errno(errno, "Failed to close loop image: %m"); - *image_fd = safe_close(*image_fd); - } - - control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); - if (control < 0) { - log_warning_errno(errno, "Failed to open /dev/loop-control: %m"); - return; - } - - r = ioctl(control, LOOP_CTL_REMOVE, nr); - if (r < 0) - log_debug_errno(errno, "Failed to remove loop %d: %m", nr); -} - /* * Return values: * < 0 : wait_for_terminate() failed to get the state of the @@ -2568,10 +1978,22 @@ static int determine_names(void) { } if (!arg_machine) { + if (arg_directory && path_equal(arg_directory, "/")) arg_machine = gethostname_malloc(); - else - arg_machine = strdup(basename(arg_image ?: arg_directory)); + else { + if (arg_image) { + char *e; + + arg_machine = strdup(basename(arg_image)); + + /* Truncate suffix if there is one */ + e = endswith(arg_machine, ".raw"); + if (e) + *e = 0; + } else + arg_machine = strdup(basename(arg_directory)); + } if (!arg_machine) return log_oom(); @@ -2921,10 +2343,7 @@ static int outer_child( Barrier *barrier, const char *directory, const char *console, - const char *root_device, bool root_device_rw, - const char *home_device, bool home_device_rw, - const char *srv_device, bool srv_device_rw, - const char *esp_device, + DissectedImage *dissected_image, bool interactive, bool secondary, int pid_socket, @@ -2984,13 +2403,11 @@ static int outer_child( if (r < 0) return r; - r = mount_devices(directory, - root_device, root_device_rw, - home_device, home_device_rw, - srv_device, srv_device_rw, - esp_device); - if (r < 0) - return r; + if (dissected_image) { + r = dissected_image_mount(dissected_image, directory, DISSECT_IMAGE_DISCARD_ON_LOOP|(arg_read_only ? DISSECT_IMAGE_READ_ONLY : 0)); + if (r < 0) + return r; + } r = determine_uid_shift(directory); if (r < 0) @@ -3607,10 +3024,7 @@ static int load_settings(void) { static int run(int master, const char* console, - const char *root_device, bool root_device_rw, - const char *home_device, bool home_device_rw, - const char *srv_device, bool srv_device_rw, - const char *esp_device, + DissectedImage *dissected_image, bool interactive, bool secondary, FDSet *fds, @@ -3717,10 +3131,7 @@ static int run(int master, r = outer_child(&barrier, arg_directory, console, - root_device, root_device_rw, - home_device, home_device_rw, - srv_device, srv_device_rw, - esp_device, + dissected_image, interactive, secondary, pid_socket_pair[1], @@ -4025,13 +3436,59 @@ static int run(int master, return 1; /* loop again */ } +static int load_root_hash(const char *image) { + _cleanup_free_ char *text = NULL; + char *fn, *n, *e; + void *k; + size_t l; + int r; + + assert_se(image); + + /* Try to load the root hash from a file next to the image file if it exists. */ + + if (arg_root_hash) + return 0; + + fn = new(char, strlen(image) + strlen(".roothash") + 1); + if (!fn) + return log_oom(); + + n = stpcpy(fn, image); + e = endswith(fn, ".raw"); + if (e) + n = e; + + strcpy(n, ".roothash"); + + r = read_one_line_file(fn, &text); + if (r == -ENOENT) + return 0; + if (r < 0) { + log_warning_errno(r, "Failed to read %s, ignoring: %m", fn); + return 0; + } + + r = unhexmem(text, strlen(text), &k, &l); + if (r < 0) + return log_error_errno(r, "Invalid root hash: %s", text); + if (l < sizeof(sd_id128_t)) { + free(k); + return log_error_errno(r, "Root hash too short: %s", text); + } + + arg_root_hash = k; + arg_root_hash_size = l; + + return 0; +} + int main(int argc, char *argv[]) { - _cleanup_free_ char *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL, *esp_device = NULL, *console = NULL; - bool root_device_rw = true, home_device_rw = true, srv_device_rw = true; - _cleanup_close_ int master = -1, image_fd = -1; + _cleanup_free_ char *console = NULL; + _cleanup_close_ int master = -1; _cleanup_fdset_free_ FDSet *fds = NULL; - int r, n_fd_passed, loop_nr = -1, ret = EXIT_SUCCESS; + int r, n_fd_passed, ret = EXIT_SUCCESS; char veth_name[IFNAMSIZ] = ""; bool secondary = false, remove_directory = false, remove_image = false; pid_t pid = 0; @@ -4039,6 +3496,9 @@ int main(int argc, char *argv[]) { _cleanup_release_lock_file_ LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT; bool interactive, veth_created = false, remove_tmprootdir = false; char tmprootdir[] = "/tmp/nspawn-root-XXXXXX"; + _cleanup_(loop_device_unrefp) LoopDevice *loop = NULL; + _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; + _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; log_parse_environment(); log_open(); @@ -4237,6 +3697,10 @@ int main(int argc, char *argv[]) { r = log_error_errno(r, "Failed to create image lock: %m"); goto finish; } + + r = load_root_hash(arg_image); + if (r < 0) + goto finish; } if (!mkdtemp(tmprootdir)) { @@ -4252,18 +3716,41 @@ int main(int argc, char *argv[]) { goto finish; } - image_fd = setup_image(&device_path, &loop_nr); - if (image_fd < 0) { - r = image_fd; + r = loop_device_make_by_path(arg_image, arg_read_only ? O_RDONLY : O_RDWR, &loop); + if (r < 0) { + log_error_errno(r, "Failed to set up loopback block device: %m"); + goto finish; + } + + r = dissect_image(loop->fd, arg_root_hash, arg_root_hash_size, &dissected_image); + if (r == -ENOPKG) { + log_error_errno(r, "Could not find a suitable file system or partition table in image: %s", arg_image); + + log_notice("Note that the disk image needs to\n" + " a) either contain only a single MBR partition of type 0x83 that is marked bootable\n" + " b) or contain a single GPT partition of type 0FC63DAF-8483-4772-8E79-3D69D8477DE4\n" + " c) or follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/\n" + " d) or contain a file system without a partition table\n" + "in order to be bootable with systemd-nspawn."); + goto finish; + } + if (r == -EADDRNOTAVAIL) { + log_error_errno(r, "No root partition for specified root hash found."); + goto finish; + } + if (r == -EOPNOTSUPP) { + log_error_errno(r, "--image= is not supported, compiled without blkid support."); + goto finish; + } + if (r < 0) { + log_error_errno(r, "Failed to dissect image: %m"); goto finish; } - r = dissect_image(image_fd, - &root_device, &root_device_rw, - &home_device, &home_device_rw, - &srv_device, &srv_device_rw, - &esp_device, - &secondary); + if (!arg_root_hash && dissected_image->can_verity) + log_notice("Note: image %s contains verity information, but no root hash specified! Proceeding without integrity checking.", arg_image); + + r = dissected_image_decrypt_interactively(dissected_image, NULL, arg_root_hash, arg_root_hash_size, 0, &decrypted_image); if (r < 0) goto finish; @@ -4317,10 +3804,7 @@ int main(int argc, char *argv[]) { for (;;) { r = run(master, console, - root_device, root_device_rw, - home_device, home_device_rw, - srv_device, srv_device_rw, - esp_device, + dissected_image, interactive, secondary, fds, veth_name, &veth_created, @@ -4347,8 +3831,6 @@ finish: if (pid > 0) (void) wait_for_terminate(pid, NULL); - loop_remove(loop_nr, &image_fd); - if (remove_directory && arg_directory) { int k; @@ -4395,6 +3877,7 @@ finish: strv_free(arg_parameters); custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts); expose_port_free_all(arg_expose_ports); + free(arg_root_hash); return r < 0 ? EXIT_FAILURE : ret; } diff --git a/src/shared/dissect-image.c b/src/shared/dissect-image.c new file mode 100644 index 0000000000..257af78781 --- /dev/null +++ b/src/shared/dissect-image.c @@ -0,0 +1,1078 @@ +/*** + This file is part of systemd. + + Copyright 2016 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#ifdef HAVE_LIBCRYPTSETUP +#include <libcryptsetup.h> +#endif +#include <linux/dm-ioctl.h> +#include <sys/mount.h> + +#include "architecture.h" +#include "ask-password-api.h" +#include "blkid-util.h" +#include "dissect-image.h" +#include "fd-util.h" +#include "gpt.h" +#include "mount-util.h" +#include "path-util.h" +#include "stat-util.h" +#include "stdio-util.h" +#include "string-table.h" +#include "string-util.h" +#include "udev-util.h" + +static int probe_filesystem(const char *node, char **ret_fstype) { +#ifdef HAVE_BLKID + _cleanup_blkid_free_probe_ blkid_probe b = NULL; + const char *fstype; + int r; + + b = blkid_new_probe_from_filename(node); + if (!b) + return -ENOMEM; + + blkid_probe_enable_superblocks(b, 1); + blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE); + + errno = 0; + r = blkid_do_safeprobe(b); + if (r == -2 || r == 1) { + log_debug("Failed to identify any partition type on partition %s", node); + goto not_found; + } + if (r != 0) { + if (errno == 0) + return -EIO; + + return -errno; + } + + (void) blkid_probe_lookup_value(b, "TYPE", &fstype, NULL); + + if (fstype) { + char *t; + + t = strdup(fstype); + if (!t) + return -ENOMEM; + + *ret_fstype = t; + return 1; + } + +not_found: + *ret_fstype = NULL; + return 0; +#else + return -EOPNOTSUPP; +#endif +} + +int dissect_image(int fd, const void *root_hash, size_t root_hash_size, DissectedImage **ret) { + +#ifdef HAVE_BLKID + sd_id128_t root_uuid = SD_ID128_NULL, verity_uuid = SD_ID128_NULL; + _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL; + bool is_gpt, is_mbr, generic_rw, multiple_generic = false; + _cleanup_udev_device_unref_ struct udev_device *d = NULL; + _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL; + _cleanup_blkid_free_probe_ blkid_probe b = NULL; + _cleanup_udev_unref_ struct udev *udev = NULL; + _cleanup_free_ char *generic_node = NULL; + const char *pttype = NULL, *usage = NULL; + struct udev_list_entry *first, *item; + blkid_partlist pl; + int r, generic_nr; + struct stat st; + unsigned i; + + assert(fd >= 0); + assert(ret); + assert(root_hash || root_hash_size == 0); + + /* Probes a disk image, and returns information about what it found in *ret. + * + * Returns -ENOPKG if no suitable partition table or file system could be found. + * Returns -EADDRNOTAVAIL if a root hash was specified but no matching root/verity partitions found. */ + + if (root_hash) { + /* If a root hash is supplied, then we use the root partition that has a UUID that match the first + * 128bit of the root hash. And we use the verity partition that has a UUID that match the final + * 128bit. */ + + if (root_hash_size < sizeof(sd_id128_t)) + return -EINVAL; + + memcpy(&root_uuid, root_hash, sizeof(sd_id128_t)); + memcpy(&verity_uuid, (const uint8_t*) root_hash + root_hash_size - sizeof(sd_id128_t), sizeof(sd_id128_t)); + + if (sd_id128_is_null(root_uuid)) + return -EINVAL; + if (sd_id128_is_null(verity_uuid)) + return -EINVAL; + } + + if (fstat(fd, &st) < 0) + return -errno; + + if (!S_ISBLK(st.st_mode)) + return -ENOTBLK; + + b = blkid_new_probe(); + if (!b) + return -ENOMEM; + + errno = 0; + r = blkid_probe_set_device(b, fd, 0, 0); + if (r != 0) { + if (errno == 0) + return -ENOMEM; + + return -errno; + } + + blkid_probe_enable_superblocks(b, 1); + blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE|BLKID_SUBLKS_USAGE); + blkid_probe_enable_partitions(b, 1); + blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS); + + errno = 0; + r = blkid_do_safeprobe(b); + if (r == -2 || r == 1) { + log_debug("Failed to identify any partition table."); + return -ENOPKG; + } + if (r != 0) { + if (errno == 0) + return -EIO; + + return -errno; + } + + m = new0(DissectedImage, 1); + if (!m) + return -ENOMEM; + + (void) blkid_probe_lookup_value(b, "USAGE", &usage, NULL); + if (STRPTR_IN_SET(usage, "filesystem", "crypto")) { + _cleanup_free_ char *t = NULL, *n = NULL; + const char *fstype = NULL; + + /* OK, we have found a file system, that's our root partition then. */ + (void) blkid_probe_lookup_value(b, "TYPE", &fstype, NULL); + + if (fstype) { + t = strdup(fstype); + if (!t) + return -ENOMEM; + } + + if (asprintf(&n, "/dev/block/%u:%u", major(st.st_rdev), minor(st.st_rdev)) < 0) + return -ENOMEM; + + m->partitions[PARTITION_ROOT] = (DissectedPartition) { + .found = true, + .rw = true, + .partno = -1, + .architecture = _ARCHITECTURE_INVALID, + .fstype = t, + .node = n, + }; + + t = n = NULL; + + m->encrypted = streq(fstype, "crypto_LUKS"); + + *ret = m; + m = NULL; + + return 0; + } + + (void) blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL); + if (!pttype) + return -ENOPKG; + + is_gpt = streq_ptr(pttype, "gpt"); + is_mbr = streq_ptr(pttype, "dos"); + + if (!is_gpt && !is_mbr) + return -ENOPKG; + + errno = 0; + pl = blkid_probe_get_partitions(b); + if (!pl) { + if (errno == 0) + return -ENOMEM; + + return -errno; + } + + udev = udev_new(); + if (!udev) + return -errno; + + d = udev_device_new_from_devnum(udev, 'b', st.st_rdev); + if (!d) + return -ENOMEM; + + for (i = 0;; i++) { + int n, z; + + if (i >= 10) { + log_debug("Kernel partitions never appeared."); + return -ENXIO; + } + + e = udev_enumerate_new(udev); + if (!e) + return -errno; + + r = udev_enumerate_add_match_parent(e, d); + if (r < 0) + return r; + + r = udev_enumerate_scan_devices(e); + if (r < 0) + return r; + + /* Count the partitions enumerated by the kernel */ + n = 0; + first = udev_enumerate_get_list_entry(e); + udev_list_entry_foreach(item, first) + n++; + + /* Count the partitions enumerated by blkid */ + z = blkid_partlist_numof_partitions(pl); + if (n == z + 1) + break; + if (n > z + 1) { + log_debug("blkid and kernel partition list do not match."); + return -EIO; + } + if (n < z + 1) { + unsigned j; + + /* The kernel has probed fewer partitions than blkid? Maybe the kernel prober is still running + * or it got EBUSY because udev already opened the device. Let's reprobe the device, which is a + * synchronous call that waits until probing is complete. */ + + for (j = 0; j < 20; j++) { + + r = ioctl(fd, BLKRRPART, 0); + if (r < 0) + r = -errno; + if (r >= 0 || r != -EBUSY) + break; + + /* If something else has the device open, such as an udev rule, the ioctl will return + * EBUSY. Since there's no way to wait until it isn't busy anymore, let's just wait a + * bit, and try again. + * + * This is really something they should fix in the kernel! */ + + usleep(50 * USEC_PER_MSEC); + } + + if (r < 0) + return r; + } + + e = udev_enumerate_unref(e); + } + + first = udev_enumerate_get_list_entry(e); + udev_list_entry_foreach(item, first) { + _cleanup_udev_device_unref_ struct udev_device *q; + unsigned long long flags; + blkid_partition pp; + const char *node; + dev_t qn; + int nr; + + q = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item)); + if (!q) + return -errno; + + qn = udev_device_get_devnum(q); + if (major(qn) == 0) + continue; + + if (st.st_rdev == qn) + continue; + + node = udev_device_get_devnode(q); + if (!node) + continue; + + pp = blkid_partlist_devno_to_partition(pl, qn); + if (!pp) + continue; + + flags = blkid_partition_get_flags(pp); + + nr = blkid_partition_get_partno(pp); + if (nr < 0) + continue; + + if (is_gpt) { + int designator = _PARTITION_DESIGNATOR_INVALID, architecture = _ARCHITECTURE_INVALID; + const char *stype, *sid, *fstype = NULL; + sd_id128_t type_id, id; + bool rw = true; + + if (flags & GPT_FLAG_NO_AUTO) + continue; + + sid = blkid_partition_get_uuid(pp); + if (!sid) + continue; + if (sd_id128_from_string(sid, &id) < 0) + continue; + + stype = blkid_partition_get_type_string(pp); + if (!stype) + continue; + if (sd_id128_from_string(stype, &type_id) < 0) + continue; + + if (sd_id128_equal(type_id, GPT_HOME)) { + designator = PARTITION_HOME; + rw = !(flags & GPT_FLAG_READ_ONLY); + } else if (sd_id128_equal(type_id, GPT_SRV)) { + designator = PARTITION_SRV; + rw = !(flags & GPT_FLAG_READ_ONLY); + } else if (sd_id128_equal(type_id, GPT_ESP)) { + designator = PARTITION_ESP; + fstype = "vfat"; + } +#ifdef GPT_ROOT_NATIVE + else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { + + /* If a root ID is specified, ignore everything but the root id */ + if (!sd_id128_is_null(root_uuid) && !sd_id128_equal(root_uuid, id)) + continue; + + designator = PARTITION_ROOT; + architecture = native_architecture(); + rw = !(flags & GPT_FLAG_READ_ONLY); + } +#ifdef GPT_ROOT_NATIVE_VERITY + else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE_VERITY)) { + + m->can_verity = true; + + /* Ignore verity unless a root hash is specified */ + if (sd_id128_is_null(verity_uuid) || !sd_id128_equal(verity_uuid, id)) + continue; + + designator = PARTITION_ROOT_VERITY; + fstype = "DM_verity_hash"; + architecture = native_architecture(); + rw = false; + } +#endif +#endif +#ifdef GPT_ROOT_SECONDARY + else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY)) { + + /* If a root ID is specified, ignore everything but the root id */ + if (!sd_id128_is_null(root_uuid) && !sd_id128_equal(root_uuid, id)) + continue; + + designator = PARTITION_ROOT_SECONDARY; + architecture = SECONDARY_ARCHITECTURE; + rw = !(flags & GPT_FLAG_READ_ONLY); + } +#ifdef GPT_ROOT_SECONDARY_VERITY + else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY_VERITY)) { + + m->can_verity = true; + + /* Ignore verity unless root has is specified */ + if (sd_id128_is_null(verity_uuid) || !sd_id128_equal(verity_uuid, id)) + continue; + + designator = PARTITION_ROOT_SECONDARY_VERITY; + fstype = "DM_verity_hash"; + architecture = SECONDARY_ARCHITECTURE; + rw = false; + } +#endif +#endif + else if (sd_id128_equal(type_id, GPT_SWAP)) { + designator = PARTITION_SWAP; + fstype = "swap"; + } else if (sd_id128_equal(type_id, GPT_LINUX_GENERIC)) { + + if (generic_node) + multiple_generic = true; + else { + generic_nr = nr; + generic_rw = !(flags & GPT_FLAG_READ_ONLY); + generic_node = strdup(node); + if (!generic_node) + return -ENOMEM; + } + } + + if (designator != _PARTITION_DESIGNATOR_INVALID) { + _cleanup_free_ char *t = NULL, *n = NULL; + + /* First one wins */ + if (m->partitions[designator].found) + continue; + + if (fstype) { + t = strdup(fstype); + if (!t) + return -ENOMEM; + } + + n = strdup(node); + if (!n) + return -ENOMEM; + + m->partitions[designator] = (DissectedPartition) { + .found = true, + .partno = nr, + .rw = rw, + .architecture = architecture, + .node = n, + .fstype = t, + }; + + n = t = NULL; + } + + } else if (is_mbr) { + + if (flags != 0x80) /* Bootable flag */ + continue; + + if (blkid_partition_get_type(pp) != 0x83) /* Linux partition */ + continue; + + if (generic_node) + multiple_generic = true; + else { + generic_nr = nr; + generic_rw = true; + generic_node = strdup(node); + if (!generic_node) + return -ENOMEM; + } + } + } + + if (!m->partitions[PARTITION_ROOT].found) { + /* No root partition found? Then let's see if ther's one for the secondary architecture. And if not + * either, then check if there's a single generic one, and use that. */ + + if (m->partitions[PARTITION_ROOT_VERITY].found) + return -ENXIO; + + if (m->partitions[PARTITION_ROOT_SECONDARY].found) { + m->partitions[PARTITION_ROOT] = m->partitions[PARTITION_ROOT_SECONDARY]; + zero(m->partitions[PARTITION_ROOT_SECONDARY]); + + m->partitions[PARTITION_ROOT_VERITY] = m->partitions[PARTITION_ROOT_SECONDARY_VERITY]; + zero(m->partitions[PARTITION_ROOT_SECONDARY_VERITY]); + + } else if (generic_node && !root_hash) { + + if (multiple_generic) + return -ENOTUNIQ; + + m->partitions[PARTITION_ROOT] = (DissectedPartition) { + .found = true, + .rw = generic_rw, + .partno = generic_nr, + .architecture = _ARCHITECTURE_INVALID, + .node = generic_node, + }; + + generic_node = NULL; + } else + return -ENXIO; + } + + assert(m->partitions[PARTITION_ROOT].found); + + if (root_hash) { + if (!m->partitions[PARTITION_ROOT_VERITY].found) + return -EADDRNOTAVAIL; + + /* If we found the primary root with the hash, then we definitely want to suppress any secondary root + * (which would be weird, after all the root hash should only be assigned to one pair of + * partitions... */ + m->partitions[PARTITION_ROOT_SECONDARY].found = false; + m->partitions[PARTITION_ROOT_SECONDARY_VERITY].found = false; + + /* If we found a verity setup, then the root partition is necessarily read-only. */ + m->partitions[PARTITION_ROOT].rw = false; + + m->verity = true; + } + + blkid_free_probe(b); + b = NULL; + + /* Fill in file system types if we don't know them yet. */ + for (i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) { + DissectedPartition *p = m->partitions + i; + + if (!p->found) + continue; + + if (!p->fstype && p->node) { + r = probe_filesystem(p->node, &p->fstype); + if (r < 0) + return r; + } + + if (streq_ptr(p->fstype, "crypto_LUKS")) + m->encrypted = true; + } + + *ret = m; + m = NULL; + + return 0; +#else + return -EOPNOTSUPP; +#endif +} + +DissectedImage* dissected_image_unref(DissectedImage *m) { + unsigned i; + + if (!m) + return NULL; + + for (i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) { + free(m->partitions[i].fstype); + free(m->partitions[i].node); + free(m->partitions[i].decrypted_fstype); + free(m->partitions[i].decrypted_node); + } + + free(m); + return NULL; +} + +static int is_loop_device(const char *path) { + char s[strlen("/sys/dev/block/") + DECIMAL_STR_MAX(dev_t) + 1 + DECIMAL_STR_MAX(dev_t) + strlen("/../loop/")]; + struct stat st; + + assert(path); + + if (stat(path, &st) < 0) + return -errno; + + if (!S_ISBLK(st.st_mode)) + return -ENOTBLK; + + xsprintf(s, "/sys/dev/block/%u:%u/loop/", major(st.st_rdev), minor(st.st_rdev)); + if (access(s, F_OK) < 0) { + if (errno != ENOENT) + return -errno; + + /* The device itself isn't a loop device, but maybe it's a partition and its parent is? */ + xsprintf(s, "/sys/dev/block/%u:%u/../loop/", major(st.st_rdev), minor(st.st_rdev)); + if (access(s, F_OK) < 0) + return errno == ENOENT ? false : -errno; + } + + return true; +} + +static int mount_partition( + DissectedPartition *m, + const char *where, + const char *directory, + DissectImageFlags flags) { + + const char *p, *options = NULL, *node, *fstype; + bool rw; + + assert(m); + assert(where); + + node = m->decrypted_node ?: m->node; + fstype = m->decrypted_fstype ?: m->fstype; + + if (!m->found || !node || !fstype) + return 0; + + /* Stacked encryption? Yuck */ + if (streq_ptr(fstype, "crypto_LUKS")) + return -ELOOP; + + rw = m->rw && !(flags & DISSECT_IMAGE_READ_ONLY); + + if (directory) + p = strjoina(where, directory); + else + p = where; + + /* If requested, turn on discard support. */ + if (STR_IN_SET(fstype, "btrfs", "ext4", "vfat", "xfs") && + ((flags & DISSECT_IMAGE_DISCARD) || + ((flags & DISSECT_IMAGE_DISCARD_ON_LOOP) && is_loop_device(m->node)))) + options = "discard"; + + return mount_verbose(LOG_DEBUG, node, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), options); +} + +int dissected_image_mount(DissectedImage *m, const char *where, DissectImageFlags flags) { + int r; + + assert(m); + assert(where); + + if (!m->partitions[PARTITION_ROOT].found) + return -ENXIO; + + r = mount_partition(m->partitions + PARTITION_ROOT, where, NULL, flags); + if (r < 0) + return r; + + r = mount_partition(m->partitions + PARTITION_HOME, where, "/home", flags); + if (r < 0) + return r; + + r = mount_partition(m->partitions + PARTITION_SRV, where, "/srv", flags); + if (r < 0) + return r; + + if (m->partitions[PARTITION_ESP].found) { + const char *mp, *x; + + /* Mount the ESP to /efi if it exists and is empty. If it doesn't exist, use /boot instead. */ + + mp = "/efi"; + x = strjoina(where, mp); + r = dir_is_empty(x); + if (r == -ENOENT) { + mp = "/boot"; + x = strjoina(where, mp); + r = dir_is_empty(x); + } + if (r > 0) { + r = mount_partition(m->partitions + PARTITION_ESP, where, mp, flags); + if (r < 0) + return r; + } + } + + return 0; +} + +#ifdef HAVE_LIBCRYPTSETUP +typedef struct DecryptedPartition { + struct crypt_device *device; + char *name; + bool relinquished; +} DecryptedPartition; + +struct DecryptedImage { + DecryptedPartition *decrypted; + size_t n_decrypted; + size_t n_allocated; +}; +#endif + +DecryptedImage* decrypted_image_unref(DecryptedImage* d) { +#ifdef HAVE_LIBCRYPTSETUP + size_t i; + int r; + + if (!d) + return NULL; + + for (i = 0; i < d->n_decrypted; i++) { + DecryptedPartition *p = d->decrypted + i; + + if (p->device && p->name && !p->relinquished) { + r = crypt_deactivate(p->device, p->name); + if (r < 0) + log_debug_errno(r, "Failed to deactivate encrypted partition %s", p->name); + } + + if (p->device) + crypt_free(p->device); + free(p->name); + } + + free(d); +#endif + return NULL; +} + +#ifdef HAVE_LIBCRYPTSETUP + +static int make_dm_name_and_node(const void *original_node, const char *suffix, char **ret_name, char **ret_node) { + _cleanup_free_ char *name = NULL, *node = NULL; + const char *base; + + assert(original_node); + assert(suffix); + assert(ret_name); + assert(ret_node); + + base = strrchr(original_node, '/'); + if (!base) + return -EINVAL; + base++; + if (isempty(base)) + return -EINVAL; + + name = strjoin(base, suffix); + if (!name) + return -ENOMEM; + if (!filename_is_valid(name)) + return -EINVAL; + + node = strjoin(crypt_get_dir(), "/", name); + if (!node) + return -ENOMEM; + + *ret_name = name; + *ret_node = node; + + name = node = NULL; + return 0; +} + +static int decrypt_partition( + DissectedPartition *m, + const char *passphrase, + DissectImageFlags flags, + DecryptedImage *d) { + + _cleanup_free_ char *node = NULL, *name = NULL; + struct crypt_device *cd; + int r; + + assert(m); + assert(d); + + if (!m->found || !m->node || !m->fstype) + return 0; + + if (!streq(m->fstype, "crypto_LUKS")) + return 0; + + r = make_dm_name_and_node(m->node, "-decrypted", &name, &node); + if (r < 0) + return r; + + if (!GREEDY_REALLOC0(d->decrypted, d->n_allocated, d->n_decrypted + 1)) + return -ENOMEM; + + r = crypt_init(&cd, m->node); + if (r < 0) + return r; + + r = crypt_load(cd, CRYPT_LUKS1, NULL); + if (r < 0) + goto fail; + + r = crypt_activate_by_passphrase(cd, name, CRYPT_ANY_SLOT, passphrase, strlen(passphrase), + ((flags & DISSECT_IMAGE_READ_ONLY) ? CRYPT_ACTIVATE_READONLY : 0) | + ((flags & DISSECT_IMAGE_DISCARD_ON_CRYPTO) ? CRYPT_ACTIVATE_ALLOW_DISCARDS : 0)); + if (r == -EPERM) { + r = -EKEYREJECTED; + goto fail; + } + if (r < 0) + goto fail; + + d->decrypted[d->n_decrypted].name = name; + name = NULL; + + d->decrypted[d->n_decrypted].device = cd; + d->n_decrypted++; + + m->decrypted_node = node; + node = NULL; + + return 0; + +fail: + crypt_free(cd); + return r; +} + +static int verity_partition( + DissectedPartition *m, + DissectedPartition *v, + const void *root_hash, + size_t root_hash_size, + DissectImageFlags flags, + DecryptedImage *d) { + + _cleanup_free_ char *node = NULL, *name = NULL; + struct crypt_device *cd; + int r; + + assert(m); + assert(v); + + if (!root_hash) + return 0; + + if (!m->found || !m->node || !m->fstype) + return 0; + if (!v->found || !v->node || !v->fstype) + return 0; + + if (!streq(v->fstype, "DM_verity_hash")) + return 0; + + r = make_dm_name_and_node(m->node, "-verity", &name, &node); + if (r < 0) + return r; + + if (!GREEDY_REALLOC0(d->decrypted, d->n_allocated, d->n_decrypted + 1)) + return -ENOMEM; + + r = crypt_init(&cd, v->node); + if (r < 0) + return r; + + r = crypt_load(cd, CRYPT_VERITY, NULL); + if (r < 0) + goto fail; + + r = crypt_set_data_device(cd, m->node); + if (r < 0) + goto fail; + + r = crypt_activate_by_volume_key(cd, name, root_hash, root_hash_size, CRYPT_ACTIVATE_READONLY); + if (r < 0) + goto fail; + + d->decrypted[d->n_decrypted].name = name; + name = NULL; + + d->decrypted[d->n_decrypted].device = cd; + d->n_decrypted++; + + m->decrypted_node = node; + node = NULL; + + return 0; + +fail: + crypt_free(cd); + return r; +} +#endif + +int dissected_image_decrypt( + DissectedImage *m, + const char *passphrase, + const void *root_hash, + size_t root_hash_size, + DissectImageFlags flags, + DecryptedImage **ret) { + + _cleanup_(decrypted_image_unrefp) DecryptedImage *d = NULL; +#ifdef HAVE_LIBCRYPTSETUP + unsigned i; + int r; +#endif + + assert(m); + assert(root_hash || root_hash_size == 0); + + /* Returns: + * + * = 0 → There was nothing to decrypt + * > 0 → Decrypted successfully + * -ENOKEY → There's some to decrypt but no key was supplied + * -EKEYREJECTED → Passed key was not correct + */ + + if (root_hash && root_hash_size < sizeof(sd_id128_t)) + return -EINVAL; + + if (!m->encrypted && !m->verity) { + *ret = NULL; + return 0; + } + +#ifdef HAVE_LIBCRYPTSETUP + if (m->encrypted && !passphrase) + return -ENOKEY; + + d = new0(DecryptedImage, 1); + if (!d) + return -ENOMEM; + + for (i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) { + DissectedPartition *p = m->partitions + i; + int k; + + if (!p->found) + continue; + + r = decrypt_partition(p, passphrase, flags, d); + if (r < 0) + return r; + + k = PARTITION_VERITY_OF(i); + if (k >= 0) { + r = verity_partition(p, m->partitions + k, root_hash, root_hash_size, flags, d); + if (r < 0) + return r; + } + + if (!p->decrypted_fstype && p->decrypted_node) { + r = probe_filesystem(p->decrypted_node, &p->decrypted_fstype); + if (r < 0) + return r; + } + } + + *ret = d; + d = NULL; + + return 1; +#else + return -EOPNOTSUPP; +#endif +} + +int dissected_image_decrypt_interactively( + DissectedImage *m, + const char *passphrase, + const void *root_hash, + size_t root_hash_size, + DissectImageFlags flags, + DecryptedImage **ret) { + + _cleanup_strv_free_erase_ char **z = NULL; + int n = 3, r; + + if (passphrase) + n--; + + for (;;) { + r = dissected_image_decrypt(m, passphrase, root_hash, root_hash_size, flags, ret); + if (r >= 0) + return r; + if (r == -EKEYREJECTED) + log_error_errno(r, "Incorrect passphrase, try again!"); + else if (r != -ENOKEY) { + log_error_errno(r, "Failed to decrypt image: %m"); + return r; + } + + if (--n < 0) { + log_error("Too many retries."); + return -EKEYREJECTED; + } + + z = strv_free(z); + + r = ask_password_auto("Please enter image passphrase!", NULL, "dissect", "dissect", USEC_INFINITY, 0, &z); + if (r < 0) + return log_error_errno(r, "Failed to query for passphrase: %m"); + + passphrase = z[0]; + } +} + +#ifdef HAVE_LIBCRYPTSETUP +static int deferred_remove(DecryptedPartition *p) { + + struct dm_ioctl dm = { + .version = { + DM_VERSION_MAJOR, + DM_VERSION_MINOR, + DM_VERSION_PATCHLEVEL + }, + .data_size = sizeof(dm), + .flags = DM_DEFERRED_REMOVE, + }; + + _cleanup_close_ int fd = -1; + + assert(p); + + /* Unfortunately, libcryptsetup doesn't provide a proper API for this, hence call the ioctl() directly. */ + + fd = open("/dev/mapper/control", O_RDWR|O_CLOEXEC); + if (fd < 0) + return -errno; + + strncpy(dm.name, p->name, sizeof(dm.name)); + + if (ioctl(fd, DM_DEV_REMOVE, &dm)) + return -errno; + + return 0; +} +#endif + +int decrypted_image_relinquish(DecryptedImage *d) { + +#ifdef HAVE_LIBCRYPTSETUP + size_t i; + int r; +#endif + + assert(d); + + /* Turns on automatic removal after the last use ended for all DM devices of this image, and sets a boolean so + * that we don't clean it up ourselves either anymore */ + +#ifdef HAVE_LIBCRYPTSETUP + for (i = 0; i < d->n_decrypted; i++) { + DecryptedPartition *p = d->decrypted + i; + + if (p->relinquished) + continue; + + r = deferred_remove(p); + if (r < 0) + return log_debug_errno(r, "Failed to mark %s for auto-removal: %m", p->name); + + p->relinquished = true; + } +#endif + + return 0; +} + +static const char *const partition_designator_table[] = { + [PARTITION_ROOT] = "root", + [PARTITION_ROOT_SECONDARY] = "root-secondary", + [PARTITION_HOME] = "home", + [PARTITION_SRV] = "srv", + [PARTITION_ESP] = "esp", + [PARTITION_SWAP] = "swap", + [PARTITION_ROOT_VERITY] = "root-verity", + [PARTITION_ROOT_SECONDARY_VERITY] = "root-secondary-verity", +}; + +DEFINE_STRING_TABLE_LOOKUP(partition_designator, int); diff --git a/src/shared/dissect-image.h b/src/shared/dissect-image.h new file mode 100644 index 0000000000..902c8d4a37 --- /dev/null +++ b/src/shared/dissect-image.h @@ -0,0 +1,90 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2016 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +#include "macro.h" + +typedef struct DissectedImage DissectedImage; +typedef struct DissectedPartition DissectedPartition; +typedef struct DecryptedImage DecryptedImage; + +struct DissectedPartition { + bool found:1; + bool rw:1; + int partno; /* -1 if there was no partition and the images contains a file system directly */ + int architecture; /* Intended architecture: either native, secondary or unset (-1). */ + char *fstype; + char *node; + char *decrypted_node; + char *decrypted_fstype; +}; + +enum { + PARTITION_ROOT, + PARTITION_ROOT_SECONDARY, /* Secondary architecture */ + PARTITION_HOME, + PARTITION_SRV, + PARTITION_ESP, + PARTITION_SWAP, + PARTITION_ROOT_VERITY, /* verity data for the PARTITION_ROOT partition */ + PARTITION_ROOT_SECONDARY_VERITY, /* verity data for the PARTITION_ROOT_SECONDARY partition */ + _PARTITION_DESIGNATOR_MAX, + _PARTITION_DESIGNATOR_INVALID = -1 +}; + +static inline int PARTITION_VERITY_OF(int p) { + if (p == PARTITION_ROOT) + return PARTITION_ROOT_VERITY; + if (p == PARTITION_ROOT_SECONDARY) + return PARTITION_ROOT_SECONDARY_VERITY; + return _PARTITION_DESIGNATOR_INVALID; +} + +typedef enum DissectImageFlags { + DISSECT_IMAGE_READ_ONLY = 1, + DISSECT_IMAGE_DISCARD_ON_LOOP = 2, /* Turn on "discard" if on loop device and file system supports it */ + DISSECT_IMAGE_DISCARD = 4, /* Turn on "discard" if file system supports it, on all block devices */ + DISSECT_IMAGE_DISCARD_ON_CRYPTO = 8, /* Turn on "discard" also on crypto devices */ +} DissectImageFlags; + +struct DissectedImage { + bool encrypted:1; + bool verity:1; /* verity available and usable */ + bool can_verity:1; /* verity available, but not necessarily used */ + DissectedPartition partitions[_PARTITION_DESIGNATOR_MAX]; +}; + +int dissect_image(int fd, const void *root_hash, size_t root_hash_size, DissectedImage **ret); + +DissectedImage* dissected_image_unref(DissectedImage *m); +DEFINE_TRIVIAL_CLEANUP_FUNC(DissectedImage*, dissected_image_unref); + +int dissected_image_decrypt(DissectedImage *m, const char *passphrase, const void *root_hash, size_t root_hash_size, DissectImageFlags flags, DecryptedImage **ret); +int dissected_image_decrypt_interactively(DissectedImage *m, const char *passphrase, const void *root_hash, size_t root_hash_size, DissectImageFlags flags, DecryptedImage **ret); +int dissected_image_mount(DissectedImage *m, const char *dest, DissectImageFlags flags); + +DecryptedImage* decrypted_image_unref(DecryptedImage *p); +DEFINE_TRIVIAL_CLEANUP_FUNC(DecryptedImage*, decrypted_image_unref); +int decrypted_image_relinquish(DecryptedImage *d); + +const char* partition_designator_to_string(int i) _const_; +int partition_designator_from_string(const char *name) _pure_; diff --git a/src/shared/gpt.h b/src/shared/gpt.h index 55b41bbcd8..13d80d611c 100644 --- a/src/shared/gpt.h +++ b/src/shared/gpt.h @@ -32,28 +32,43 @@ #define GPT_ROOT_ARM SD_ID128_MAKE(69,da,d7,10,2c,e4,4e,3c,b1,6c,21,a1,d4,9a,be,d3) #define GPT_ROOT_ARM_64 SD_ID128_MAKE(b9,21,b0,45,1d,f0,41,c3,af,44,4c,6f,28,0d,3f,ae) #define GPT_ROOT_IA64 SD_ID128_MAKE(99,3d,8d,3d,f8,0e,42,25,85,5a,9d,af,8e,d7,ea,97) - #define GPT_ESP SD_ID128_MAKE(c1,2a,73,28,f8,1f,11,d2,ba,4b,00,a0,c9,3e,c9,3b) #define GPT_SWAP SD_ID128_MAKE(06,57,fd,6d,a4,ab,43,c4,84,e5,09,33,c8,4b,4f,4f) #define GPT_HOME SD_ID128_MAKE(93,3a,c7,e1,2e,b4,4f,13,b8,44,0e,14,e2,ae,f9,15) #define GPT_SRV SD_ID128_MAKE(3b,8f,84,25,20,e0,4f,3b,90,7f,1a,25,a7,6f,98,e8) +/* Verity partitions for the root partitions above (we only define them for the root partitions, because only they are + * are commonly read-only and hence suitable for verity). */ +#define GPT_ROOT_X86_VERITY SD_ID128_MAKE(d1,3c,5d,3b,b5,d1,42,2a,b2,9f,94,54,fd,c8,9d,76) +#define GPT_ROOT_X86_64_VERITY SD_ID128_MAKE(2c,73,57,ed,eb,d2,46,d9,ae,c1,23,d4,37,ec,2b,f5) +#define GPT_ROOT_ARM_VERITY SD_ID128_MAKE(73,86,cd,f2,20,3c,47,a9,a4,98,f2,ec,ce,45,a2,d6) +#define GPT_ROOT_ARM_64_VERITY SD_ID128_MAKE(df,33,00,ce,d6,9f,4c,92,97,8c,9b,fb,0f,38,d8,20) +#define GPT_ROOT_IA64_VERITY SD_ID128_MAKE(86,ed,10,d5,b6,07,45,bb,89,57,d3,50,f2,3d,05,71) + + #if defined(__x86_64__) # define GPT_ROOT_NATIVE GPT_ROOT_X86_64 # define GPT_ROOT_SECONDARY GPT_ROOT_X86 +# define GPT_ROOT_NATIVE_VERITY GPT_ROOT_X86_64_VERITY +# define GPT_ROOT_SECONDARY_VERITY GPT_ROOT_X86_VERITY #elif defined(__i386__) # define GPT_ROOT_NATIVE GPT_ROOT_X86 +# define GPT_ROOT_NATIVE_VERITY GPT_ROOT_X86_VERITY #endif #if defined(__ia64__) # define GPT_ROOT_NATIVE GPT_ROOT_IA64 +# define GPT_ROOT_NATIVE_VERITY GPT_ROOT_IA64_VERITY #endif #if defined(__aarch64__) && (__BYTE_ORDER != __BIG_ENDIAN) # define GPT_ROOT_NATIVE GPT_ROOT_ARM_64 # define GPT_ROOT_SECONDARY GPT_ROOT_ARM +# define GPT_ROOT_NATIVE_VERITY GPT_ROOT_ARM_64_VERITY +# define GPT_ROOT_SECONDARY_VERITY GPT_ROOT_ARM_VERITY #elif defined(__arm__) && (__BYTE_ORDER != __BIG_ENDIAN) # define GPT_ROOT_NATIVE GPT_ROOT_ARM +# define GPT_ROOT_NATIVE_VERITY GPT_ROOT_ARM_VERITY #endif /* Flags we recognize on the root, swap, home and srv partitions when diff --git a/src/shared/loop-util.c b/src/shared/loop-util.c new file mode 100644 index 0000000000..047e213634 --- /dev/null +++ b/src/shared/loop-util.c @@ -0,0 +1,166 @@ +/*** + This file is part of systemd. + + Copyright 2016 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <fcntl.h> +#include <linux/loop.h> +#include <sys/ioctl.h> +#include <sys/stat.h> + +#include "alloc-util.h" +#include "fd-util.h" +#include "loop-util.h" + +int loop_device_make(int fd, int open_flags, LoopDevice **ret) { + const struct loop_info64 info = { + .lo_flags = LO_FLAGS_AUTOCLEAR|LO_FLAGS_PARTSCAN|(open_flags == O_RDONLY ? LO_FLAGS_READ_ONLY : 0), + }; + + _cleanup_close_ int control = -1, loop = -1; + _cleanup_free_ char *loopdev = NULL; + struct stat st; + LoopDevice *d; + int nr; + + assert(fd >= 0); + assert(ret); + assert(IN_SET(open_flags, O_RDWR, O_RDONLY)); + + if (fstat(fd, &st) < 0) + return -errno; + + if (S_ISBLK(st.st_mode)) { + int copy; + + /* If this is already a block device, store a copy of the fd as it is */ + + copy = fcntl(fd, F_DUPFD_CLOEXEC, 3); + if (copy < 0) + return -errno; + + d = new0(LoopDevice, 1); + if (!d) + return -ENOMEM; + + *d = (LoopDevice) { + .fd = copy, + .nr = -1, + }; + + *ret = d; + + return 0; + } + + if (!S_ISREG(st.st_mode)) + return -EINVAL; + + control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); + if (control < 0) + return -errno; + + nr = ioctl(control, LOOP_CTL_GET_FREE); + if (nr < 0) + return -errno; + + if (asprintf(&loopdev, "/dev/loop%i", nr) < 0) + return -ENOMEM; + + loop = open(loopdev, O_CLOEXEC|O_NONBLOCK|O_NOCTTY|open_flags); + if (loop < 0) + return -errno; + + if (ioctl(loop, LOOP_SET_FD, fd) < 0) + return -errno; + + if (ioctl(loop, LOOP_SET_STATUS64, &info) < 0) + return -errno; + + d = new(LoopDevice, 1); + if (!d) + return -ENOMEM; + + *d = (LoopDevice) { + .fd = loop, + .node = loopdev, + .nr = nr, + }; + + loop = -1; + loopdev = NULL; + + *ret = d; + + return (*ret)->fd; +} + +int loop_device_make_by_path(const char *path, int open_flags, LoopDevice **ret) { + _cleanup_close_ int fd = -1; + + assert(path); + assert(ret); + assert(IN_SET(open_flags, O_RDWR, O_RDONLY)); + + fd = open(path, O_CLOEXEC|O_NONBLOCK|O_NOCTTY|open_flags); + if (fd < 0) + return -errno; + + return loop_device_make(fd, open_flags, ret); +} + +LoopDevice* loop_device_unref(LoopDevice *d) { + if (!d) + return NULL; + + if (d->fd >= 0) { + + if (d->nr >= 0 && !d->relinquished) { + if (ioctl(d->fd, LOOP_CLR_FD) < 0) + log_debug_errno(errno, "Failed to clear loop device: %m"); + + } + + safe_close(d->fd); + } + + if (d->nr >= 0 && !d->relinquished) { + _cleanup_close_ int control = -1; + + control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); + if (control < 0) + log_debug_errno(errno, "Failed to open loop control device: %m"); + else { + if (ioctl(control, LOOP_CTL_REMOVE, d->nr) < 0) + log_debug_errno(errno, "Failed to remove loop device: %m"); + } + } + + free(d->node); + free(d); + + return NULL; +} + +void loop_device_relinquish(LoopDevice *d) { + assert(d); + + /* Don't attempt to clean up the loop device anymore from this point on. Leave the clean-ing up to the kernel + * itself, using the loop device "auto-clear" logic we already turned on when creating the device. */ + + d->relinquished = true; +} diff --git a/src/shared/loop-util.h b/src/shared/loop-util.h new file mode 100644 index 0000000000..45fead5f18 --- /dev/null +++ b/src/shared/loop-util.h @@ -0,0 +1,41 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2016 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "macro.h" + +typedef struct LoopDevice LoopDevice; + +/* Some helpers for setting up loopback block devices */ + +struct LoopDevice { + int fd; + int nr; + char *node; + bool relinquished; +}; + +int loop_device_make(int fd, int open_flags, LoopDevice **ret); +int loop_device_make_by_path(const char *path, int open_flags, LoopDevice **ret); + +LoopDevice* loop_device_unref(LoopDevice *d); +DEFINE_TRIVIAL_CLEANUP_FUNC(LoopDevice*, loop_device_unref); + +void loop_device_relinquish(LoopDevice *d); diff --git a/src/test/test-dissect-image.c b/src/test/test-dissect-image.c new file mode 100644 index 0000000000..0512a15e88 --- /dev/null +++ b/src/test/test-dissect-image.c @@ -0,0 +1,66 @@ +/*** + This file is part of systemd. + + Copyright 2016 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <fcntl.h> +#include <stdio.h> + +#include "dissect-image.h" +#include "log.h" +#include "loop-util.h" +#include "string-util.h" + +int main(int argc, char *argv[]) { + _cleanup_(loop_device_unrefp) LoopDevice *d = NULL; + _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL; + int r, i; + + log_set_max_level(LOG_DEBUG); + + if (argc < 2) { + log_error("Requires one command line argument."); + return EXIT_FAILURE; + } + + r = loop_device_make_by_path(argv[1], O_RDONLY, &d); + if (r < 0) { + log_error_errno(r, "Failed to set up loopback device: %m"); + return EXIT_FAILURE; + } + + r = dissect_image(d->fd, NULL, 0, &m); + if (r < 0) { + log_error_errno(r, "Failed to dissect image: %m"); + return EXIT_FAILURE; + } + + for (i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) { + + if (!m->partitions[i].found) + continue; + + printf("Found %s partition, %s of type %s at #%i (%s)\n", + partition_designator_to_string(i), + m->partitions[i].rw ? "writable" : "read-only", + strna(m->partitions[i].fstype), + m->partitions[i].partno, + strna(m->partitions[i].node)); + } + + return EXIT_SUCCESS; +} |