X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=73158a051831edf74a5db86ee8027b471a5e28c8;hp=097d7f779f0d781ba73e6662e48f35265fec4c8b;hb=590b6b9188e75ba46c42995984a1c2fa06adb6d6;hpb=0cb9fbcd44517ec90b2a678876194607beab5dec diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 097d7f779..73158a051 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -86,6 +86,7 @@ #include "udev-util.h" #include "blkid-util.h" #include "gpt.h" +#include "siphash24.h" #ifdef HAVE_SECCOMP #include "seccomp-util.h" @@ -656,7 +657,7 @@ static int mount_binds(const char *dest, char **l, unsigned long flags) { return r; } } else { - log_error("Failed to bind mount %s: %s", *x, strerror(errno)); + log_error("Failed to bind mount %s: %m", *x); return -errno; } /* Create the mount point, but be conservative -- refuse to create block @@ -768,6 +769,15 @@ static int setup_resolv_conf(const char *dest) { return 0; } +static char* id128_format_as_uuid(sd_id128_t id, char s[37]) { + + snprintf(s, 37, + "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x", + SD_ID128_FORMAT_VAL(id)); + + return s; +} + static int setup_boot_id(const char *dest) { _cleanup_free_ char *from = NULL, *to = NULL; sd_id128_t rnd = {}; @@ -793,10 +803,7 @@ static int setup_boot_id(const char *dest) { return r; } - snprintf(as_uuid, sizeof(as_uuid), - "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x", - SD_ID128_FORMAT_VAL(rnd)); - char_array_0(as_uuid); + id128_format_as_uuid(rnd, as_uuid); r = write_string_file(from, as_uuid); if (r < 0) { @@ -985,7 +992,7 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { /* Store away the fd in the socket, so that it stays open as * long as we run the child */ k = sendmsg(kmsg_socket, &mh, MSG_DONTWAIT|MSG_NOSIGNAL); - close_nointr_nofail(fd); + safe_close(fd); if (k < 0) { log_error("Failed to send FIFO fd: %m"); @@ -1136,10 +1143,8 @@ static int setup_journal(const char *directory) { } else if (access(p, F_OK) < 0) return 0; - if (dir_is_empty(q) == 0) { - log_error("%s not empty.", q); - return -ENOTEMPTY; - } + if (dir_is_empty(q) == 0) + log_warning("%s is not empty, proceeding anyway.", q); r = mkdir_p(q, 0755); if (r < 0) { @@ -1258,7 +1263,7 @@ static int register_machine(pid_t pid) { return r; } - r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 8, + r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 10, /* Allow the container to * access and create the API * device nodes, so that @@ -1277,7 +1282,18 @@ static int register_machine(pid_t pid) { * container to ever create * these device nodes. */ "/dev/pts/ptmx", "rw", - "char-pts", "rw"); + "char-pts", "rw", + /* Allow the container + * access to all kdbus + * devices. Again, the + * container cannot create + * these nodes, only use + * them. We use a pretty + * open match here, so that + * the kernel API can still + * change. */ + "char-kdbus", "rw", + "char-kdbus/*", "rw"); if (r < 0) { log_error("Failed to add device whitelist: %s", strerror(-r)); return r; @@ -1388,9 +1404,46 @@ static int reset_audit_loginuid(void) { return 0; } +#define HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2) + +static int get_mac(struct ether_addr *mac) { + int r; + + uint8_t result[8]; + size_t l, sz; + uint8_t *v; + + l = strlen(arg_machine); + sz = sizeof(sd_id128_t) + l; + v = alloca(sz); + + /* fetch some persistent data unique to the host */ + r = sd_id128_get_machine((sd_id128_t*) v); + if (r < 0) + return r; + + /* combine with some data unique (on this host) to this + * container instance */ + memcpy(v + sizeof(sd_id128_t), arg_machine, l); + + /* Let's hash the host machine ID plus the container name. We + * use a fixed, but originally randomly created hash key here. */ + siphash24(result, v, sz, HASH_KEY.bytes); + + assert_cc(ETH_ALEN <= sizeof(result)); + memcpy(mac->ether_addr_octet, result, ETH_ALEN); + + /* see eth_random_addr in the kernel */ + mac->ether_addr_octet[0] &= 0xfe; /* clear multicast bit */ + mac->ether_addr_octet[0] |= 0x02; /* set local assignment bit (IEEE802) */ + + return 0; +} + static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; + struct ether_addr mac; int r; if (!arg_private_network) @@ -1405,9 +1458,14 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { memcpy(iface_name, "vb-", 3); else memcpy(iface_name, "ve-", 3); - strncpy(iface_name+3, arg_machine, IFNAMSIZ - 3); + r = get_mac(&mac); + if (r < 0) { + log_error("Failed to generate predictable MAC address for host0"); + return r; + } + r = sd_rtnl_open(&rtnl, 0); if (r < 0) { log_error("Failed to connect to netlink: %s", strerror(-r)); @@ -1432,13 +1490,7 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { return r; } - r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "veth"); - if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); - return r; - } - - r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA); + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "veth"); if (r < 0) { log_error("Failed to open netlink container: %s", strerror(-r)); return r; @@ -1456,6 +1508,12 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { return r; } + r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac); + if (r < 0) { + log_error("Failed to add netlink MAC address: %s", strerror(-r)); + return r; + } + r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); if (r < 0) { log_error("Failed to add netlink namespace field: %s", strerror(-r)); @@ -1521,6 +1579,12 @@ static int setup_bridge(const char veth_name[]) { return r; } + r = sd_rtnl_message_link_set_flags(m, IFF_UP, IFF_UP); + if (r < 0) { + log_error("Failed to set IFF_UP flag: %s", strerror(-r)); + return r; + } + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, veth_name); if (r < 0) { log_error("Failed to add netlink interface name field: %s", strerror(-r)); @@ -1691,13 +1755,7 @@ static int setup_macvlan(pid_t pid) { return r; } - r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "macvlan"); - if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); - return r; - } - - r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA); + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "macvlan"); if (r < 0) { log_error("Failed to open netlink container: %s", strerror(-r)); return r; @@ -1880,9 +1938,9 @@ static int setup_image(char **device_path, int *loop_nr) { static int dissect_image( int fd, - char **root_device, - char **home_device, - char **srv_device, + char **root_device, bool *root_device_rw, + char **home_device, bool *home_device_rw, + char **srv_device, bool *srv_device_rw, bool *secondary) { #ifdef HAVE_BLKID @@ -1893,6 +1951,7 @@ static int dissect_image( _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; const char *pttype = NULL; blkid_partlist pl; struct stat st; @@ -1982,6 +2041,7 @@ static int dissect_image( udev_list_entry_foreach(item, first) { _cleanup_udev_device_unref_ struct udev_device *q; const char *stype, *node; + unsigned long long flags; sd_id128_t type_id; blkid_partition pp; dev_t qn; @@ -2012,6 +2072,10 @@ static int dissect_image( if (!pp) continue; + flags = blkid_partition_get_flags(pp); + if (flags & GPT_FLAG_NO_AUTO) + continue; + nr = blkid_partition_get_partno(pp); if (nr < 0) continue; @@ -2029,6 +2093,8 @@ static int dissect_image( continue; home_nr = nr; + home_rw = !(flags & GPT_FLAG_READ_ONLY); + free(home); home = strdup(node); if (!home) @@ -2039,6 +2105,8 @@ static int dissect_image( continue; srv_nr = nr; + srv_rw = !(flags & GPT_FLAG_READ_ONLY); + free(srv); srv = strdup(node); if (!srv) @@ -2051,6 +2119,8 @@ static int dissect_image( continue; root_nr = nr; + root_rw = !(flags & GPT_FLAG_READ_ONLY); + free(root); root = strdup(node); if (!root) @@ -2064,6 +2134,9 @@ static int dissect_image( continue; secondary_root_nr = nr; + secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY); + + free(secondary_root); secondary_root = strdup(node); if (!secondary_root) @@ -2081,21 +2154,29 @@ static int dissect_image( 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; } if (home) { *home_device = home; home = NULL; + + *home_device_rw = home_rw; } if (srv) { *srv_device = srv; srv = NULL; + + *srv_device_rw = srv_rw; } return 0; @@ -2105,7 +2186,7 @@ static int dissect_image( #endif } -static int mount_device(const char *what, const char *where, const char *directory) { +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; @@ -2114,6 +2195,9 @@ static int mount_device(const char *what, const char *where, const char *directo assert(what); assert(where); + if (arg_read_only) + rw = false; + if (directory) p = strappenda(where, directory); else @@ -2156,7 +2240,7 @@ static int mount_device(const char *what, const char *where, const char *directo return -ENOTSUP; } - if (mount(what, p, fstype, arg_read_only ? MS_NODEV|MS_RDONLY : 0, NULL) < 0) { + if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0) { log_error("Failed to mount %s: %m", what); return -errno; } @@ -2168,13 +2252,17 @@ static int mount_device(const char *what, const char *where, const char *directo #endif } -static int mount_devices(const char *where, const char *root_device, const char *home_device, const char *srv_device) { +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) { int r; assert(where); if (root_device) { - r = mount_device(root_device, arg_directory, NULL); + r = mount_device(root_device, arg_directory, NULL, root_device_rw); if (r < 0) { log_error("Failed to mount root directory: %s", strerror(-r)); return r; @@ -2182,7 +2270,7 @@ static int mount_devices(const char *where, const char *root_device, const char } if (home_device) { - r = mount_device(home_device, arg_directory, "/home"); + r = mount_device(home_device, arg_directory, "/home", home_device_rw); if (r < 0) { log_error("Failed to mount home directory: %s", strerror(-r)); return r; @@ -2190,7 +2278,7 @@ static int mount_devices(const char *where, const char *root_device, const char } if (srv_device) { - r = mount_device(srv_device, arg_directory, "/srv"); + r = mount_device(srv_device, arg_directory, "/srv", srv_device_rw); if (r < 0) { log_error("Failed to mount server data directory: %s", strerror(-r)); return r; @@ -2208,8 +2296,7 @@ static void loop_remove(int nr, int *image_fd) { if (image_fd && *image_fd >= 0) { ioctl(*image_fd, LOOP_CLR_FD); - close_nointr_nofail(*image_fd); - *image_fd = -1; + *image_fd = safe_close(*image_fd); } control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); @@ -2244,9 +2331,9 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) { _exit(EXIT_FAILURE); if (pipe_fds[0] > 2) - close_nointr_nofail(pipe_fds[0]); + safe_close(pipe_fds[0]); if (pipe_fds[1] > 2) - close_nointr_nofail(pipe_fds[1]); + safe_close(pipe_fds[1]); nullfd = open("/dev/null", O_RDWR); if (nullfd < 0) @@ -2259,18 +2346,17 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) { _exit(EXIT_FAILURE); if (nullfd > 2) - close_nointr_nofail(nullfd); + safe_close(nullfd); reset_all_signal_handlers(); close_all_fds(NULL, 0); - execle("/usr/bin/getent", "getenv", database, key, NULL, &empty_env); - execle("/usr/getent", "getenv", database, key, NULL, &empty_env); + execle("/usr/bin/getent", "getent", database, key, NULL, &empty_env); + execle("/bin/getent", "getent", database, key, NULL, &empty_env); _exit(EXIT_FAILURE); } - close_nointr_nofail(pipe_fds[1]); - pipe_fds[1] = -1; + pipe_fds[1] = safe_close(pipe_fds[1]); *rpid = pid; @@ -2278,15 +2364,13 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) { } static int change_uid_gid(char **_home) { - - _cleanup_strv_free_ char **passwd = NULL; char line[LINE_MAX], *w, *x, *state, *u, *g, *h; _cleanup_free_ uid_t *uids = NULL; _cleanup_free_ char *home = NULL; _cleanup_fclose_ FILE *f = NULL; _cleanup_close_ int fd = -1; unsigned n_uids = 0; - size_t sz, l; + size_t sz = 0, l; uid_t uid; gid_t gid; pid_t pid; @@ -2453,7 +2537,7 @@ static int change_uid_gid(char **_home) { } r = mkdir_safe(home, 0755, uid, gid); - if (r < 0) { + if (r < 0 && r != -EEXIST) { log_error("Failed to make home directory: %s", strerror(-r)); return r; } @@ -2488,8 +2572,9 @@ static int change_uid_gid(char **_home) { int main(int argc, char *argv[]) { _cleanup_free_ char *kdbus_domain = NULL, *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL; + bool root_device_rw = true, home_device_rw = true, srv_device_rw = true; _cleanup_close_ int master = -1, kdbus_fd = -1, image_fd = -1; - _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 }; + _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }; _cleanup_fdset_free_ FDSet *fds = NULL; int r = EXIT_FAILURE, k, n_fd_passed, loop_nr = -1; const char *console = NULL; @@ -2605,7 +2690,7 @@ int main(int argc, char *argv[]) { goto finish; } - r = dissect_image(image_fd, &root_device, &home_device, &srv_device, &secondary); + r = dissect_image(image_fd, &root_device, &root_device_rw, &home_device, &home_device_rw, &srv_device, &srv_device_rw, &secondary); if (r < 0) goto finish; } @@ -2713,15 +2798,13 @@ int main(int argc, char *argv[]) { if (envp[n_env]) n_env ++; - close_nointr_nofail(master); - master = -1; + master = safe_close(master); close_nointr(STDIN_FILENO); close_nointr(STDOUT_FILENO); close_nointr(STDERR_FILENO); - close_nointr_nofail(kmsg_socket_pair[0]); - kmsg_socket_pair[0] = -1; + kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]); reset_all_signal_handlers(); @@ -2731,7 +2814,7 @@ int main(int argc, char *argv[]) { k = open_terminal(console, O_RDWR); if (k != STDIN_FILENO) { if (k >= 0) { - close_nointr_nofail(k); + safe_close(k); k = -EINVAL; } @@ -2766,7 +2849,10 @@ int main(int argc, char *argv[]) { goto child_fail; } - if (mount_devices(arg_directory, root_device, home_device, srv_device) < 0) + if (mount_devices(arg_directory, + root_device, root_device_rw, + home_device, home_device_rw, + srv_device, srv_device_rw) < 0) goto child_fail; /* Turn directory into bind mount */ @@ -2801,8 +2887,7 @@ int main(int argc, char *argv[]) { if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0) goto child_fail; - close_nointr_nofail(kmsg_socket_pair[1]); - kmsg_socket_pair[1] = -1; + kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]); if (setup_boot_id(arg_directory) < 0) goto child_fail; @@ -2829,8 +2914,7 @@ int main(int argc, char *argv[]) { * it can cgroupify us to that we lack access * to certain devices and resources. */ eventfd_write(child_ready_fd, 1); - close_nointr_nofail(child_ready_fd); - child_ready_fd = -1; + child_ready_fd = safe_close(child_ready_fd); if (chdir(arg_directory) < 0) { log_error("chdir(%s) failed: %m", arg_directory); @@ -2874,7 +2958,9 @@ int main(int argc, char *argv[]) { } if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) { - if (asprintf((char**)(envp + n_env++), "container_uuid=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(arg_uuid)) < 0) { + char as_uuid[37]; + + if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0) { log_oom(); goto child_fail; } @@ -2931,8 +3017,7 @@ int main(int argc, char *argv[]) { /* Wait until the parent is ready with the setup, too... */ eventfd_read(parent_ready_fd, &x); - close_nointr_nofail(parent_ready_fd); - parent_ready_fd = -1; + parent_ready_fd = safe_close(parent_ready_fd); if (arg_boot) { char **a; @@ -2970,7 +3055,7 @@ int main(int argc, char *argv[]) { fds = NULL; /* Wait until the child reported that it is ready with - * all it needs to do with priviliges. After we got + * all it needs to do with privileges. After we got * the notification we can make the process join its * cgroup which might limit what it can do */ eventfd_read(child_ready_fd, &x);