X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=fb672510b4ff6e718a71b581d4350c48ba46d577;hb=4d2b1e0a3f98d8cf93beb943b05868b55231006a;hp=efeba596658079606376811963586f61da49df3a;hpb=f6c51a8136de3f27e28caea2003e18f4bc4cb9a8;p=elogind.git diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index efeba5966..fb672510b 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -180,6 +180,7 @@ static bool arg_register = true; static bool arg_keep_unit = false; static char **arg_network_interfaces = NULL; static char **arg_network_macvlan = NULL; +static char **arg_network_ipvlan = NULL; static bool arg_network_veth = false; static const char *arg_network_bridge = NULL; static unsigned long arg_personality = 0xffffffffLU; @@ -211,6 +212,9 @@ static void help(void) { " --network-macvlan=INTERFACE\n" " Create a macvlan network interface based on an\n" " existing network interface to the container\n" + " --network-ipvlan=INTERFACE\n" + " Create a ipvlan network interface based on an\n" + " existing network interface to the container\n" " -n --network-veth Add a virtual ethernet connection between host\n" " and container\n" " --network-bridge=INTERFACE\n" @@ -285,6 +289,7 @@ static int parse_argv(int argc, char *argv[]) { ARG_KEEP_UNIT, ARG_NETWORK_INTERFACE, ARG_NETWORK_MACVLAN, + ARG_NETWORK_IPVLAN, ARG_NETWORK_BRIDGE, ARG_PERSONALITY, ARG_VOLATILE, @@ -319,6 +324,7 @@ static int parse_argv(int argc, char *argv[]) { { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT }, { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE }, { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN }, + { "network-ipvlan", required_argument, NULL, ARG_NETWORK_IPVLAN }, { "network-veth", no_argument, NULL, 'n' }, { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE }, { "personality", required_argument, NULL, ARG_PERSONALITY }, @@ -401,6 +407,13 @@ static int parse_argv(int argc, char *argv[]) { if (strv_extend(&arg_network_macvlan, optarg) < 0) return log_oom(); + arg_private_network = true; + break; + + case ARG_NETWORK_IPVLAN: + if (strv_extend(&arg_network_ipvlan, optarg) < 0) + return log_oom(); + /* fall through */ case ARG_PRIVATE_NETWORK: @@ -803,6 +816,7 @@ static int mount_all(const char *dest) { { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true }, { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, + { "tmpfs", "/tmp", "tmpfs", "mode=1777", MS_STRICTATIME, true }, #ifdef HAVE_SELINUX { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */ { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */ @@ -900,8 +914,12 @@ static int mount_binds(const char *dest, char **l, bool ro) { r = stat(where, &dest_st); if (r == 0) { - if ((source_st.st_mode & S_IFMT) != (dest_st.st_mode & S_IFMT)) { - log_error("The file types of %s and %s do not match. Refusing bind mount", *x, where); + if (S_ISDIR(source_st.st_mode) && !S_ISDIR(dest_st.st_mode)) { + log_error("Cannot bind mount directory %s on file %s.", *x, where); + return -EINVAL; + } + if (!S_ISDIR(source_st.st_mode) && S_ISDIR(dest_st.st_mode)) { + log_error("Cannot bind mount file %s on directory %s.", *x, where); return -EINVAL; } } else if (errno == ENOENT) { @@ -913,27 +931,18 @@ static int mount_binds(const char *dest, char **l, bool ro) { return -errno; } - /* Create the mount point, but be conservative -- refuse to create block - * and char devices. */ + /* Create the mount point. Any non-directory file can be + * mounted on any non-directory file (regular, fifo, socket, + * char, block). + */ if (S_ISDIR(source_st.st_mode)) { r = mkdir_label(where, 0755); if (r < 0 && errno != EEXIST) return log_error_errno(r, "Failed to create mount point %s: %m", where); - } else if (S_ISFIFO(source_st.st_mode)) { - r = mkfifo(where, 0644); - if (r < 0 && errno != EEXIST) - return log_error_errno(errno, "Failed to create mount point %s: %m", where); - } else if (S_ISSOCK(source_st.st_mode)) { - r = mknod(where, 0644 | S_IFSOCK, 0); - if (r < 0 && errno != EEXIST) - return log_error_errno(errno, "Failed to create mount point %s: %m", where); - } else if (S_ISREG(source_st.st_mode)) { + } else { r = touch(where); if (r < 0) return log_error_errno(r, "Failed to create mount point %s: %m", where); - } else { - log_error("Refusing to create mountpoint for file: %s", *x); - return -ENOTSUP; } if (mount(*x, where, "bind", MS_BIND, NULL) < 0) @@ -953,7 +962,7 @@ static int mount_cgroup_hierarchy(const char *dest, const char *controller, cons char *to; int r; - to = strappenda(dest, "/sys/fs/cgroup/", hierarchy); + to = strjoina(dest, "/sys/fs/cgroup/", hierarchy); r = path_is_mount_point(to, false); if (r < 0) @@ -963,9 +972,17 @@ static int mount_cgroup_hierarchy(const char *dest, const char *controller, cons mkdir_p(to, 0755); - if (mount("cgroup", to, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV|(read_only ? MS_RDONLY : 0), controller) < 0) + /* The superblock mount options of the mount point need to be + * identical to the hosts', and hence writable... */ + if (mount("cgroup", to, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, controller) < 0) return log_error_errno(errno, "Failed to mount to %s: %m", to); + /* ... hence let's only make the bind mount read-only, not the + * superblock. */ + if (read_only) { + if (mount(NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0) + return log_error_errno(errno, "Failed to remount %s read-only: %m", to); + } return 1; } @@ -987,7 +1004,7 @@ static int mount_cgroup(const char *dest) { if (r < 0) return log_error_errno(r, "Failed to determine our own cgroup path: %m"); - cgroup_root = strappenda(dest, "/sys/fs/cgroup"); + cgroup_root = strjoina(dest, "/sys/fs/cgroup"); if (mount("tmpfs", cgroup_root, "tmpfs", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, "mode=755") < 0) return log_error_errno(errno, "Failed to mount tmpfs to /sys/fs/cgroup: %m"); @@ -1035,17 +1052,17 @@ static int mount_cgroup(const char *dest) { } } - r = mount_cgroup_hierarchy(dest, "name=systemd", "systemd", false); + r = mount_cgroup_hierarchy(dest, "name=systemd,xattr", "systemd", false); if (r < 0) return r; /* Make our own cgroup a (writable) bind mount */ - systemd_own = strappenda(dest, "/sys/fs/cgroup/systemd", own_cgroup_path); + systemd_own = strjoina(dest, "/sys/fs/cgroup/systemd", own_cgroup_path); if (mount(systemd_own, systemd_own, NULL, MS_BIND, NULL) < 0) return log_error_errno(errno, "Failed to turn %s into a bind mount: %m", own_cgroup_path); /* And then remount the systemd cgroup root read-only */ - systemd_root = strappenda(dest, "/sys/fs/cgroup/systemd"); + systemd_root = strjoina(dest, "/sys/fs/cgroup/systemd"); if (mount(NULL, systemd_root, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0) return log_error_errno(errno, "Failed to mount cgroup root read-only: %m"); @@ -1198,7 +1215,7 @@ static int setup_volatile_state(const char *directory) { if (r < 0) return log_error_errno(r, "Failed to remount %s read-only: %m", directory); - p = strappenda(directory, "/var"); + p = strjoina(directory, "/var"); r = mkdir(p, 0755); if (r < 0 && errno != EEXIST) return log_error_errno(errno, "Failed to create %s: %m", directory); @@ -1234,8 +1251,8 @@ static int setup_volatile(const char *directory) { tmpfs_mounted = true; - f = strappenda(directory, "/usr"); - t = strappenda(template, "/usr"); + f = strjoina(directory, "/usr"); + t = strjoina(template, "/usr"); r = mkdir(t, 0755); if (r < 0 && errno != EEXIST) { @@ -1417,7 +1434,7 @@ static int setup_dev_console(const char *dest, const char *console) { * /dev/console. (Note that the major minor doesn't actually * matter here, since we mount it over anyway). */ - to = strappenda(dest, "/dev/console"); + to = strjoina(dest, "/dev/console"); if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) return log_error_errno(errno, "mknod() for /dev/console failed: %m"); @@ -2381,21 +2398,105 @@ static int setup_macvlan(pid_t pid) { return 0; } +static int setup_ipvlan(pid_t pid) { + _cleanup_udev_unref_ struct udev *udev = NULL; + _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; + char **i; + int r; + + if (!arg_private_network) + return 0; + + if (strv_isempty(arg_network_ipvlan)) + return 0; + + r = sd_rtnl_open(&rtnl, 0); + if (r < 0) + return log_error_errno(r, "Failed to connect to netlink: %m"); + + udev = udev_new(); + if (!udev) { + log_error("Failed to connect to udev."); + return -ENOMEM; + } + + STRV_FOREACH(i, arg_network_ipvlan) { + _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; + _cleanup_free_ char *n = NULL; + int ifi; + + ifi = parse_interface(udev, *i); + if (ifi < 0) + return ifi; + + r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0); + if (r < 0) + return log_error_errno(r, "Failed to allocate netlink message: %m"); + + r = sd_rtnl_message_append_u32(m, IFLA_LINK, ifi); + if (r < 0) + return log_error_errno(r, "Failed to add netlink interface index: %m"); + + n = strappend("iv-", *i); + if (!n) + return log_oom(); + + strshorten(n, IFNAMSIZ-1); + + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, n); + if (r < 0) + return log_error_errno(r, "Failed to add netlink interface name: %m"); + + r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); + if (r < 0) + return log_error_errno(r, "Failed to add netlink namespace field: %m"); + + r = sd_rtnl_message_open_container(m, IFLA_LINKINFO); + if (r < 0) + return log_error_errno(r, "Failed to open netlink container: %m"); + + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "ipvlan"); + if (r < 0) + return log_error_errno(r, "Failed to open netlink container: %m"); + + r = sd_rtnl_message_append_u16(m, IFLA_IPVLAN_MODE, IPVLAN_MODE_L2); + if (r < 0) + return log_error_errno(r, "Failed to add ipvlan mode: %m"); + + r = sd_rtnl_message_close_container(m); + if (r < 0) + return log_error_errno(r, "Failed to close netlink container: %m"); + + r = sd_rtnl_message_close_container(m); + if (r < 0) + return log_error_errno(r, "Failed to close netlink container: %m"); + + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) + return log_error_errno(r, "Failed to add new ipvlan interfaces: %m"); + } + + return 0; +} + static int setup_seccomp(void) { #ifdef HAVE_SECCOMP static const int blacklist[] = { SCMP_SYS(kexec_load), SCMP_SYS(open_by_handle_at), - SCMP_SYS(init_module), - SCMP_SYS(finit_module), - SCMP_SYS(delete_module), SCMP_SYS(iopl), SCMP_SYS(ioperm), SCMP_SYS(swapon), SCMP_SYS(swapoff), }; + static const int kmod_blacklist[] = { + SCMP_SYS(init_module), + SCMP_SYS(finit_module), + SCMP_SYS(delete_module), + }; + scmp_filter_ctx seccomp; unsigned i; int r; @@ -2420,6 +2521,20 @@ static int setup_seccomp(void) { } } + /* If the CAP_SYS_MODULE capability is not requested then + * we'll block the kmod syscalls too */ + if (!(arg_retain & (1ULL << CAP_SYS_MODULE))) { + for (i = 0; i < ELEMENTSOF(kmod_blacklist); i++) { + r = seccomp_rule_add(seccomp, SCMP_ACT_ERRNO(EPERM), kmod_blacklist[i], 0); + if (r == -EFAULT) + continue; /* unknown syscall */ + if (r < 0) { + log_error_errno(r, "Failed to block syscall: %m"); + goto finish; + } + } + } + /* Audit is broken in containers, much of the userspace audit hookup will fail if running inside a container. We don't @@ -2466,10 +2581,10 @@ static int setup_propagate(const char *root) { (void) mkdir_p("/run/systemd/nspawn/", 0755); (void) mkdir_p("/run/systemd/nspawn/propagate", 0600); - p = strappenda("/run/systemd/nspawn/propagate/", arg_machine); + p = strjoina("/run/systemd/nspawn/propagate/", arg_machine); (void) mkdir_p(p, 0600); - q = strappenda(root, "/run/systemd/nspawn/incoming"); + q = strjoina(root, "/run/systemd/nspawn/incoming"); mkdir_parents(q, 0755); mkdir_p(q, 0600); @@ -2559,63 +2674,6 @@ static int setup_image(char **device_path, int *loop_nr) { return r; } -static int wait_for_block_device(struct udev *udev, dev_t devnum, struct udev_device **ret) { - _cleanup_udev_monitor_unref_ struct udev_monitor *monitor = NULL; - int r; - - assert(udev); - assert(ret); - - for (;;) { - _cleanup_udev_device_unref_ struct udev_device *d = NULL; - struct pollfd pfd = { - .events = POLLIN - }; - - d = udev_device_new_from_devnum(udev, 'b', devnum); - if (!d) - return log_oom(); - - r = udev_device_get_is_initialized(d); - if (r < 0) - return log_error_errno(r, "Failed to check if device is initialized: %m"); - if (r > 0) { - *ret = d; - d = NULL; - return 0; - } - d = udev_device_unref(d); - - if (!monitor) { - monitor = udev_monitor_new_from_netlink(udev, "udev"); - if (!monitor) - return log_oom(); - - r = udev_monitor_filter_add_match_subsystem_devtype(monitor, "block", NULL); - if (r < 0) - return log_error_errno(r, "Failed to add block match: %m"); - - r = udev_monitor_enable_receiving(monitor); - if (r < 0) - return log_error_errno(r, "Failed to turn on monitor: %m"); - - continue; - } - - pfd.fd = udev_monitor_get_fd(monitor); - if (pfd.fd < 0) - return log_error_errno(r, "Failed to get udev monitor fd: %m"); - - r = poll(&pfd, 1, -1); - if (r < 0) - return log_error_errno(errno, "Failed to wait for device initialization: %m"); - - d = udev_monitor_receive_device(monitor); - } - - return 0; -} - #define PARTITION_TABLE_BLURB \ "Note that the disk image needs to either contain only a single MBR partition of\n" \ "type 0x83 that is marked bootable, or a sinlge GPT partition of type" \ @@ -2645,11 +2703,12 @@ static int dissect_image( _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; - bool is_gpt, is_mbr, multiple_generic = false; assert(fd >= 0); assert(root_device); @@ -2718,21 +2777,81 @@ static int dissect_image( if (fstat(fd, &st) < 0) return log_error_errno(errno, "Failed to stat block device: %m"); - r = wait_for_block_device(udev, st.st_rdev, &d); - if (r < 0) - return r; - - e = udev_enumerate_new(udev); - if (!e) + d = udev_device_new_from_devnum(udev, 'b', st.st_rdev); + if (!d) return log_oom(); - r = udev_enumerate_add_match_parent(e, d); - if (r < 0) - return log_oom(); + for (i = 0;; i++) { + int n, m; - 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); + 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) { @@ -2948,7 +3067,7 @@ static int mount_device(const char *what, const char *where, const char *directo rw = false; if (directory) - p = strappenda(where, directory); + p = strjoina(where, directory); else p = where; @@ -3508,7 +3627,6 @@ int main(int argc, char *argv[]) { } if (arg_ephemeral) { - _cleanup_release_lock_file_ LockFile original_lock = LOCK_FILE_INIT; char *np; /* If the specified path is a mount point we @@ -3584,7 +3702,7 @@ int main(int argc, char *argv[]) { } else { const char *p; - p = strappenda(arg_directory, + p = strjoina(arg_directory, argc > optind && path_is_absolute(argv[optind]) ? argv[optind] : "/usr/bin/"); if (access(p, F_OK) < 0) { log_error("Directory %s lacks the binary to execute or doesn't look like a binary tree. Refusing.", arg_directory); @@ -4044,6 +4162,10 @@ int main(int argc, char *argv[]) { if (r < 0) goto finish; + r = setup_ipvlan(pid); + if (r < 0) + goto finish; + r = register_machine(pid, ifi); if (r < 0) goto finish; @@ -4189,7 +4311,7 @@ finish: if (arg_machine) { const char *p; - p = strappenda("/run/systemd/nspawn/propagate/", arg_machine); + p = strjoina("/run/systemd/nspawn/propagate/", arg_machine); (void) rm_rf(p, false, true, false); } @@ -4201,6 +4323,7 @@ finish: strv_free(arg_setenv); strv_free(arg_network_interfaces); strv_free(arg_network_macvlan); + strv_free(arg_network_ipvlan); strv_free(arg_bind); strv_free(arg_bind_ro); strv_free(arg_tmpfs);