X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=9a9ed9dc6e0f7e3e6d2759b6ce50711697c194d9;hp=c0f99482c075b443bcb981f6db5f2d2bd422a1ce;hb=3d94f76c99da13e5603831d0b278f8c8c21bcb02;hpb=88212f7bd19e30d73874aa160e789e300a643860 diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index c0f99482c..9a9ed9dc6 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -33,8 +33,6 @@ #include #include #include -#include -#include #include #include #include @@ -42,10 +40,28 @@ #include #include #include +#include +#include +#include +#include +#include -#include -#include +#ifdef HAVE_SELINUX +#include +#endif + +#ifdef HAVE_SECCOMP +#include +#endif +#ifdef HAVE_BLKID +#include +#endif + +#include "sd-daemon.h" +#include "sd-bus.h" +#include "sd-id128.h" +#include "sd-rtnl.h" #include "log.h" #include "util.h" #include "mkdir.h" @@ -56,15 +72,24 @@ #include "strv.h" #include "path-util.h" #include "loopback-setup.h" -#include "sd-id128.h" #include "dev-setup.h" #include "fdset.h" #include "build.h" #include "fileio.h" #include "bus-util.h" - -#ifndef TTY_GID -#define TTY_GID 5 +#include "bus-error.h" +#include "ptyfwd.h" +#include "bus-kernel.h" +#include "env-util.h" +#include "def.h" +#include "rtnl-util.h" +#include "udev-util.h" +#include "blkid-util.h" +#include "gpt.h" +#include "siphash24.h" + +#ifdef HAVE_SECCOMP +#include "seccomp-util.h" #endif typedef enum LinkJournal { @@ -78,6 +103,8 @@ static char *arg_directory = NULL; static char *arg_user = NULL; static sd_id128_t arg_uuid = {}; static char *arg_machine = NULL; +static const char *arg_selinux_context = NULL; +static const char *arg_selinux_apifs_context = NULL; static const char *arg_slice = NULL; static bool arg_private_network = false; static bool arg_read_only = false; @@ -108,31 +135,69 @@ static uint64_t arg_retain = (1ULL << CAP_SYS_RESOURCE) | (1ULL << CAP_SYS_BOOT) | (1ULL << CAP_AUDIT_WRITE) | - (1ULL << CAP_AUDIT_CONTROL); + (1ULL << CAP_AUDIT_CONTROL) | + (1ULL << CAP_MKNOD); static char **arg_bind = NULL; static char **arg_bind_ro = NULL; +static char **arg_setenv = NULL; +static bool arg_quiet = false; +static bool arg_share_system = false; +static bool arg_register = true; +static bool arg_keep_unit = false; +static char **arg_network_interfaces = NULL; +static char **arg_network_macvlan = NULL; +static bool arg_network_veth = false; +static const char *arg_network_bridge = NULL; +static unsigned long arg_personality = 0xffffffffLU; +static const char *arg_image = NULL; static int help(void) { printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" "Spawn a minimal namespace container for debugging, testing and building.\n\n" - " -h --help Show this help\n" - " --version Print version string\n" - " -D --directory=NAME Root directory for the container\n" - " -b --boot Boot up full system (i.e. invoke init)\n" - " -u --user=USER Run the command under specified user or uid\n" - " --uuid=UUID Set a specific machine UUID for the container\n" - " -M --machine=NAME Set the machine name for the container\n" - " -S --slice=SLICE Place the container in the specified slice\n" - " --private-network Disable network in container\n" - " --read-only Mount the root directory read-only\n" - " --capability=CAP In addition to the default, retain specified\n" - " capability\n" - " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n" - " -j Equivalent to --link-journal=host\n" - " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n" - " the container\n" - " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n", + " -h --help Show this help\n" + " --version Print version string\n" + " -q --quiet Do not show status information\n" + " -D --directory=PATH Root directory for the container\n" + " -i --image=PATH File system device or image for the container\n" + " -b --boot Boot up full system (i.e. invoke init)\n" + " -u --user=USER Run the command under specified user or uid\n" + " -M --machine=NAME Set the machine name for the container\n" + " --uuid=UUID Set a specific machine UUID for the container\n" + " -S --slice=SLICE Place the container in the specified slice\n" + " --private-network Disable network in container\n" + " --network-interface=INTERFACE\n" + " Assign an existing network interface to the\n" + " container\n" + " --network-macvlan=INTERFACE\n" + " Create a macvlan network interface based on an\n" + " existing network interface to the container\n" + " --network-veth Add a virtual ethernet connection between host\n" + " and container\n" + " --network-bridge=INTERFACE\n" + " Add a virtual ethernet connection between host\n" + " and container and add it to an existing bridge on\n" + " the host\n" + " -Z --selinux-context=SECLABEL\n" + " Set the SELinux security context to be used by\n" + " processes in the container\n" + " -L --selinux-apifs-context=SECLABEL\n" + " Set the SELinux security context to be used by\n" + " API/tmpfs file systems in the container\n" + " --capability=CAP In addition to the default, retain specified\n" + " capability\n" + " --drop-capability=CAP Drop the specified capability from the default set\n" + " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n" + " -j Equivalent to --link-journal=host\n" + " --read-only Mount the root directory read-only\n" + " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n" + " the container\n" + " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n" + " --setenv=NAME=VALUE Pass an environment variable to PID 1\n" + " --share-system Share system namespaces with host\n" + " --register=BOOLEAN Register container as machine\n" + " --keep-unit Do not register a scope for the machine, reuse\n" + " the service unit nspawn is running in\n", program_invocation_short_name); return 0; @@ -146,41 +211,65 @@ static int parse_argv(int argc, char *argv[]) { ARG_UUID, ARG_READ_ONLY, ARG_CAPABILITY, + ARG_DROP_CAPABILITY, ARG_LINK_JOURNAL, ARG_BIND, - ARG_BIND_RO + ARG_BIND_RO, + ARG_SETENV, + ARG_SHARE_SYSTEM, + ARG_REGISTER, + ARG_KEEP_UNIT, + ARG_NETWORK_INTERFACE, + ARG_NETWORK_MACVLAN, + ARG_NETWORK_VETH, + ARG_NETWORK_BRIDGE, + ARG_PERSONALITY, }; static const struct option options[] = { - { "help", no_argument, NULL, 'h' }, - { "version", no_argument, NULL, ARG_VERSION }, - { "directory", required_argument, NULL, 'D' }, - { "user", required_argument, NULL, 'u' }, - { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK }, - { "boot", no_argument, NULL, 'b' }, - { "uuid", required_argument, NULL, ARG_UUID }, - { "read-only", no_argument, NULL, ARG_READ_ONLY }, - { "capability", required_argument, NULL, ARG_CAPABILITY }, - { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, - { "bind", required_argument, NULL, ARG_BIND }, - { "bind-ro", required_argument, NULL, ARG_BIND_RO }, - { "machine", required_argument, NULL, 'M' }, - { "slice", required_argument, NULL, 'S' }, - { NULL, 0, NULL, 0 } + { "help", no_argument, NULL, 'h' }, + { "version", no_argument, NULL, ARG_VERSION }, + { "directory", required_argument, NULL, 'D' }, + { "user", required_argument, NULL, 'u' }, + { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK }, + { "boot", no_argument, NULL, 'b' }, + { "uuid", required_argument, NULL, ARG_UUID }, + { "read-only", no_argument, NULL, ARG_READ_ONLY }, + { "capability", required_argument, NULL, ARG_CAPABILITY }, + { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY }, + { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, + { "bind", required_argument, NULL, ARG_BIND }, + { "bind-ro", required_argument, NULL, ARG_BIND_RO }, + { "machine", required_argument, NULL, 'M' }, + { "slice", required_argument, NULL, 'S' }, + { "setenv", required_argument, NULL, ARG_SETENV }, + { "selinux-context", required_argument, NULL, 'Z' }, + { "selinux-apifs-context", required_argument, NULL, 'L' }, + { "quiet", no_argument, NULL, 'q' }, + { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM }, + { "register", required_argument, NULL, ARG_REGISTER }, + { "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-veth", no_argument, NULL, ARG_NETWORK_VETH }, + { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE }, + { "personality", required_argument, NULL, ARG_PERSONALITY }, + { "image", required_argument, NULL, 'i' }, + {} }; int c, r; + uint64_t plus = 0, minus = 0; assert(argc >= 0); assert(argv); - while ((c = getopt_long(argc, argv, "+hD:u:bM:jS:", options, NULL)) >= 0) { + while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:qi:", options, NULL)) >= 0) { switch (c) { case 'h': - help(); - return 0; + return help(); case ARG_VERSION: puts(PACKAGE_STRING); @@ -191,12 +280,16 @@ static int parse_argv(int argc, char *argv[]) { free(arg_directory); arg_directory = canonicalize_file_name(optarg); if (!arg_directory) { - log_error("Failed to canonicalize root directory."); + log_error("Invalid root directory: %m"); return -ENOMEM; } break; + case 'i': + arg_image = optarg; + break; + case 'u': free(arg_user); arg_user = strdup(optarg); @@ -205,6 +298,29 @@ static int parse_argv(int argc, char *argv[]) { break; + case ARG_NETWORK_BRIDGE: + arg_network_bridge = optarg; + + /* fall through */ + + case ARG_NETWORK_VETH: + arg_network_veth = true; + arg_private_network = true; + break; + + case ARG_NETWORK_INTERFACE: + if (strv_extend(&arg_network_interfaces, optarg) < 0) + return log_oom(); + + arg_private_network = true; + break; + + case ARG_NETWORK_MACVLAN: + if (strv_extend(&arg_network_macvlan, optarg) < 0) + return log_oom(); + + /* fall through */ + case ARG_PRIVATE_NETWORK: arg_private_network = true; break; @@ -222,46 +338,69 @@ static int parse_argv(int argc, char *argv[]) { break; case 'S': - arg_slice = strdup(optarg); + arg_slice = optarg; break; case 'M': - if (!hostname_is_valid(optarg)) { - log_error("Invalid machine name: %s", optarg); - return -EINVAL; + if (isempty(optarg)) { + free(arg_machine); + arg_machine = NULL; + } else { + + if (!hostname_is_valid(optarg)) { + log_error("Invalid machine name: %s", optarg); + return -EINVAL; + } + + free(arg_machine); + arg_machine = strdup(optarg); + if (!arg_machine) + return log_oom(); + + break; } - free(arg_machine); - arg_machine = strdup(optarg); - if (!arg_machine) - return log_oom(); + case 'Z': + arg_selinux_context = optarg; + break; + case 'L': + arg_selinux_apifs_context = optarg; break; case ARG_READ_ONLY: arg_read_only = true; break; - case ARG_CAPABILITY: { + case ARG_CAPABILITY: + case ARG_DROP_CAPABILITY: { char *state, *word; size_t length; FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) { + _cleanup_free_ char *t; cap_value_t cap; - char *t; t = strndup(word, length); if (!t) return log_oom(); - if (cap_from_name(t, &cap) < 0) { - log_error("Failed to parse capability %s.", t); - free(t); - return -EINVAL; - } + if (streq(t, "all")) { + if (c == ARG_CAPABILITY) + plus = (uint64_t) -1; + else + minus = (uint64_t) -1; + } else { + if (cap_from_name(t, &cap) < 0) { + log_error("Failed to parse capability %s.", t); + return -EINVAL; + } - free(t); - arg_retain |= 1ULL << (uint64_t) cap; + if (c == ARG_CAPABILITY) + plus |= 1ULL << (uint64_t) cap; + else + minus |= 1ULL << (uint64_t) cap; + } } break; @@ -314,24 +453,92 @@ static int parse_argv(int argc, char *argv[]) { r = strv_extend(x, a); if (r < 0) - return r; + return log_oom(); r = strv_extend(x, b); if (r < 0) - return r; + return log_oom(); + + break; + } + case ARG_SETENV: { + char **n; + + if (!env_assignment_is_valid(optarg)) { + log_error("Environment variable assignment '%s' is not valid.", optarg); + return -EINVAL; + } + + n = strv_env_set(arg_setenv, optarg); + if (!n) + return log_oom(); + + strv_free(arg_setenv); + arg_setenv = n; break; } + case 'q': + arg_quiet = true; + break; + + case ARG_SHARE_SYSTEM: + arg_share_system = true; + break; + + case ARG_REGISTER: + r = parse_boolean(optarg); + if (r < 0) { + log_error("Failed to parse --register= argument: %s", optarg); + return r; + } + + arg_register = r; + break; + + case ARG_KEEP_UNIT: + arg_keep_unit = true; + break; + + case ARG_PERSONALITY: + + arg_personality = personality_from_string(optarg); + if (arg_personality == 0xffffffffLU) { + log_error("Unknown or unsupported personality '%s'.", optarg); + return -EINVAL; + } + + break; + case '?': return -EINVAL; default: - log_error("Unknown option code %c", c); - return -EINVAL; + assert_not_reached("Unhandled option"); } } + if (arg_share_system) + arg_register = false; + + if (arg_boot && arg_share_system) { + log_error("--boot and --share-system may not be combined."); + return -EINVAL; + } + + if (arg_keep_unit && cg_pid_get_owner_uid(0, NULL) >= 0) { + log_error("--keep-unit may not be used when invoked from a user session."); + return -EINVAL; + } + + if (arg_directory && arg_image) { + log_error("--directory= and --image= may not be combined."); + return -EINVAL; + } + + arg_retain = (arg_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus; + return 1; } @@ -366,6 +573,10 @@ static int mount_all(const char *dest) { for (k = 0; k < ELEMENTSOF(mount_table); k++) { _cleanup_free_ char *where = NULL; +#ifdef HAVE_SELINUX + _cleanup_free_ char *options = NULL; +#endif + const char *o; int t; where = strjoin(dest, "/", mount_table[k].where, NULL); @@ -388,11 +599,24 @@ static int mount_all(const char *dest) { mkdir_p(where, 0755); +#ifdef HAVE_SELINUX + if (arg_selinux_apifs_context && + (streq_ptr(mount_table[k].what, "tmpfs") || streq_ptr(mount_table[k].what, "devpts"))) { + options = strjoin(mount_table[k].options, ",context=\"", arg_selinux_apifs_context, "\"", NULL); + if (!options) + return log_oom(); + + o = options; + } else +#endif + o = mount_table[k].options; + + if (mount(mount_table[k].what, where, mount_table[k].type, mount_table[k].flags, - mount_table[k].options) < 0 && + o) < 0 && mount_table[k].fatal) { log_error("mount(%s) failed: %m", where); @@ -409,39 +633,46 @@ static int mount_binds(const char *dest, char **l, unsigned long flags) { char **x, **y; STRV_FOREACH_PAIR(x, y, l) { - _cleanup_free_ char *where = NULL; + char *where; struct stat source_st, dest_st; + int r; if (stat(*x, &source_st) < 0) { - log_error("failed to stat %s: %m", *x); + log_error("Failed to stat %s: %m", *x); return -errno; } - where = strjoin(dest, "/", *y, NULL); - if (!where) - return log_oom(); - - if (stat(where, &dest_st) == 0) { + where = strappenda(dest, *y); + 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); return -EINVAL; } - } else { - /* Create the mount point, but be conservative -- refuse to create block - * and char devices. */ - if (S_ISDIR(source_st.st_mode)) - mkdir_p_label(where, 0755); - else if (S_ISFIFO(source_st.st_mode)) - mkfifo(where, 0644); - else if (S_ISSOCK(source_st.st_mode)) - mknod(where, 0644 | S_IFSOCK, 0); - else if (S_ISREG(source_st.st_mode)) - touch(where); - else { - log_error("Refusing to create mountpoint for file: %s", *x); - return -ENOTSUP; + } else if (errno == ENOENT) { + r = mkdir_parents_label(where, 0755); + if (r < 0) { + log_error("Failed to bind mount %s: %s", *x, strerror(-r)); + return r; } + } else { + log_error("Failed to bind mount %s: %s", *x, strerror(errno)); + return -errno; + } + /* Create the mount point, but be conservative -- refuse to create block + * and char devices. */ + if (S_ISDIR(source_st.st_mode)) + mkdir_label(where, 0755); + else if (S_ISFIFO(source_st.st_mode)) + mkfifo(where, 0644); + else if (S_ISSOCK(source_st.st_mode)) + mknod(where, 0644 | S_IFSOCK, 0); + else if (S_ISREG(source_st.st_mode)) + touch(where); + else { + log_error("Refusing to create mountpoint for file: %s", *x); + return -ENOTSUP; } if (mount(*x, where, "bind", MS_BIND, NULL) < 0) { @@ -540,12 +771,15 @@ static int setup_resolv_conf(const char *dest) { static int setup_boot_id(const char *dest) { _cleanup_free_ char *from = NULL, *to = NULL; - sd_id128_t rnd; + sd_id128_t rnd = {}; char as_uuid[37]; int r; assert(dest); + if (arg_share_system) + return 0; + /* Generate a new randomized boot ID, so that each boot-up of * the container gets a new one */ @@ -600,40 +834,30 @@ static int copy_devnodes(const char *dest) { u = umask(0000); NULSTR_FOREACH(d, devnodes) { - struct stat st; _cleanup_free_ char *from = NULL, *to = NULL; + struct stat st; - asprintf(&from, "/dev/%s", d); - asprintf(&to, "%s/dev/%s", dest, d); - - if (!from || !to) { - log_oom(); - - if (r == 0) - r = -ENOMEM; - - break; - } + from = strappend("/dev/", d); + to = strjoin(dest, "/dev/", d, NULL); + if (!from || !to) + return log_oom(); if (stat(from, &st) < 0) { if (errno != ENOENT) { log_error("Failed to stat %s: %m", from); - if (r == 0) - r = -errno; + return -errno; } } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) { log_error("%s is not a char or block device, cannot copy", from); - if (r == 0) - r = -EIO; + return -EIO; } else if (mknod(to, st.st_mode, st.st_rdev) < 0) { log_error("mknod(%s) failed: %m", dest); - if (r == 0) - r = -errno; + return -errno; } } @@ -656,23 +880,19 @@ static int setup_ptmx(const char *dest) { } static int setup_dev_console(const char *dest, const char *console) { + _cleanup_umask_ mode_t u; + const char *to; struct stat st; - _cleanup_free_ char *to = NULL; int r; - _cleanup_umask_ mode_t u; assert(dest); assert(console); u = umask(0000); - if (stat(console, &st) < 0) { - log_error("Failed to stat %s: %m", console); + if (stat("/dev/null", &st) < 0) { + log_error("Failed to stat /dev/null: %m"); return -errno; - - } else if (!S_ISCHR(st.st_mode)) { - log_error("/dev/console is not a char device"); - return -EIO; } r = chmod_and_chown(console, 0600, 0, 0); @@ -681,16 +901,15 @@ static int setup_dev_console(const char *dest, const char *console) { return r; } - if (asprintf(&to, "%s/dev/console", dest) < 0) - return log_oom(); - /* We need to bind mount the right tty to /dev/console since * ptys can only exist on pts file systems. To have something - * to bind mount things on we create a device node first, that - * has the right major/minor (note that the major minor - * doesn't actually matter here, since we mount it over - * anyway). */ + * to bind mount things on we create a device node first, and + * use /dev/null for that since we the cgroups device policy + * allows us to create that freely, while we cannot create + * /dev/console. (Note that the major minor doesn't actually + * matter here, since we mount it over anyway). */ + to = strappenda(dest, "/dev/console"); if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) { log_error("mknod() for /dev/console failed: %m"); return -errno; @@ -767,7 +986,7 @@ 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"); @@ -781,6 +1000,9 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { static int setup_hostname(void) { + if (arg_share_system) + return 0; + if (sethostname(arg_machine, strlen(arg_machine)) < 0) return -errno; @@ -788,14 +1010,11 @@ static int setup_hostname(void) { } static int setup_journal(const char *directory) { - sd_id128_t machine_id; + sd_id128_t machine_id, this_id; _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL; char *id; int r; - if (arg_link_journal == LINK_NO) - return 0; - p = strappend(directory, "/etc/machine-id"); if (!p) return log_oom(); @@ -819,6 +1038,24 @@ static int setup_journal(const char *directory) { return r; } + r = sd_id128_get_machine(&this_id); + if (r < 0) { + log_error("Failed to retrieve machine ID: %s", strerror(-r)); + return r; + } + + if (sd_id128_equal(machine_id, this_id)) { + log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR, + "Host and machine ids are equal (%s): refusing to link journals", id); + if (arg_link_journal == LINK_AUTO) + return 0; + return + -EEXIST; + } + + if (arg_link_journal == LINK_NO) + return 0; + free(p); p = strappend("/var/log/journal/", id); q = strjoin(directory, "/var/log/journal/", id, NULL); @@ -919,310 +1156,1440 @@ static int setup_journal(const char *directory) { return 0; } +static int setup_kdbus(const char *dest, const char *path) { + const char *p; + + if (!path) + return 0; + + p = strappenda(dest, "/dev/kdbus"); + if (mkdir(p, 0755) < 0) { + log_error("Failed to create kdbus path: %m"); + return -errno; + } + + if (mount(path, p, "bind", MS_BIND, NULL) < 0) { + log_error("Failed to mount kdbus domain path: %m"); + return -errno; + } + + return 0; +} + static int drop_capabilities(void) { return capability_bounding_set_drop(~arg_retain, false); } -static int process_pty(int master, pid_t pid, sigset_t *mask) { - - char in_buffer[LINE_MAX], out_buffer[LINE_MAX]; - size_t in_buffer_full = 0, out_buffer_full = 0; - struct epoll_event stdin_ev, stdout_ev, master_ev, signal_ev; - bool stdin_readable = false, stdout_writable = false, master_readable = false, master_writable = false; - int ep = -1, signal_fd = -1, r; - bool tried_orderly_shutdown = false; - - assert(master >= 0); - assert(pid > 0); - assert(mask); +static int register_machine(pid_t pid) { + _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_bus_unref_ sd_bus *bus = NULL; + int r; - fd_nonblock(STDIN_FILENO, 1); - fd_nonblock(STDOUT_FILENO, 1); - fd_nonblock(master, 1); + if (!arg_register) + return 0; - signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC); - if (signal_fd < 0) { - log_error("signalfd(): %m"); - r = -errno; - goto finish; + r = sd_bus_default_system(&bus); + if (r < 0) { + log_error("Failed to open system bus: %s", strerror(-r)); + return r; } - ep = epoll_create1(EPOLL_CLOEXEC); - if (ep < 0) { - log_error("Failed to create epoll: %m"); - r = -errno; - goto finish; - } + if (arg_keep_unit) { + r = sd_bus_call_method( + bus, + "org.freedesktop.machine1", + "/org/freedesktop/machine1", + "org.freedesktop.machine1.Manager", + "RegisterMachine", + &error, + NULL, + "sayssus", + arg_machine, + SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), + "nspawn", + "container", + (uint32_t) pid, + strempty(arg_directory)); + } else { + _cleanup_bus_message_unref_ sd_bus_message *m = NULL; + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.machine1", + "/org/freedesktop/machine1", + "org.freedesktop.machine1.Manager", + "CreateMachine"); + if (r < 0) { + log_error("Failed to create message: %s", strerror(-r)); + return r; + } - /* We read from STDIN only if this is actually a TTY, - * otherwise we assume non-interactivity. */ - if (isatty(STDIN_FILENO)) { - zero(stdin_ev); - stdin_ev.events = EPOLLIN|EPOLLET; - stdin_ev.data.fd = STDIN_FILENO; + r = sd_bus_message_append( + m, + "sayssus", + arg_machine, + SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), + "nspawn", + "container", + (uint32_t) pid, + strempty(arg_directory)); + if (r < 0) { + log_error("Failed to append message arguments: %s", strerror(-r)); + return r; + } - if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0) { - log_error("Failed to register STDIN in epoll: %m"); - r = -errno; - goto finish; + r = sd_bus_message_open_container(m, 'a', "(sv)"); + if (r < 0) { + log_error("Failed to open container: %s", strerror(-r)); + return r; } - } - zero(stdout_ev); - stdout_ev.events = EPOLLOUT|EPOLLET; - stdout_ev.data.fd = STDOUT_FILENO; + if (!isempty(arg_slice)) { + r = sd_bus_message_append(m, "(sv)", "Slice", "s", arg_slice); + if (r < 0) { + log_error("Failed to append slice: %s", strerror(-r)); + return r; + } + } - zero(master_ev); - master_ev.events = EPOLLIN|EPOLLOUT|EPOLLET; - master_ev.data.fd = master; + r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict"); + if (r < 0) { + log_error("Failed to add device policy: %s", strerror(-r)); + return r; + } - zero(signal_ev); - signal_ev.events = EPOLLIN; - signal_ev.data.fd = signal_fd; + r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 10, + /* Allow the container to + * access and create the API + * device nodes, so that + * PrivateDevices= in the + * container can work + * fine */ + "/dev/null", "rwm", + "/dev/zero", "rwm", + "/dev/full", "rwm", + "/dev/random", "rwm", + "/dev/urandom", "rwm", + "/dev/tty", "rwm", + /* Allow the container + * access to ptys. However, + * do not permit the + * container to ever create + * these device nodes. */ + "/dev/pts/ptmx", "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; + } - if (epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0) { - if (errno != EPERM) { - log_error("Failed to register stdout in epoll: %m"); - r = -errno; - goto finish; + r = sd_bus_message_close_container(m); + if (r < 0) { + log_error("Failed to close container: %s", strerror(-r)); + return r; } - /* stdout without epoll support. Likely redirected to regular file. */ - stdout_writable = true; + + r = sd_bus_call(bus, m, 0, &error, NULL); } - if (epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 || - epoll_ctl(ep, EPOLL_CTL_ADD, signal_fd, &signal_ev) < 0) { - log_error("Failed to register fds in epoll: %m"); - r = -errno; - goto finish; + if (r < 0) { + log_error("Failed to register machine: %s", bus_error_message(&error, r)); + return r; } - for (;;) { - struct epoll_event ev[16]; - ssize_t k; - int i, nfds; + return 0; +} - nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1); - if (nfds < 0) { +static int terminate_machine(pid_t pid) { + _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_bus_message_unref_ sd_bus_message *reply = NULL; + _cleanup_bus_unref_ sd_bus *bus = NULL; + const char *path; + int r; - if (errno == EINTR || errno == EAGAIN) - continue; + if (!arg_register) + return 0; - log_error("epoll_wait(): %m"); - r = -errno; - goto finish; - } + r = sd_bus_default_system(&bus); + if (r < 0) { + log_error("Failed to open system bus: %s", strerror(-r)); + return r; + } - assert(nfds >= 1); + r = sd_bus_call_method( + bus, + "org.freedesktop.machine1", + "/org/freedesktop/machine1", + "org.freedesktop.machine1.Manager", + "GetMachineByPID", + &error, + &reply, + "u", + (uint32_t) pid); + if (r < 0) { + /* Note that the machine might already have been + * cleaned up automatically, hence don't consider it a + * failure if we cannot get the machine object. */ + log_debug("Failed to get machine: %s", bus_error_message(&error, r)); + return 0; + } - for (i = 0; i < nfds; i++) { - if (ev[i].data.fd == STDIN_FILENO) { + r = sd_bus_message_read(reply, "o", &path); + if (r < 0) + return bus_log_parse_error(r); - if (ev[i].events & (EPOLLIN|EPOLLHUP)) - stdin_readable = true; + r = sd_bus_call_method( + bus, + "org.freedesktop.machine1", + path, + "org.freedesktop.machine1.Machine", + "Terminate", + &error, + NULL, + NULL); + if (r < 0) { + log_debug("Failed to terminate machine: %s", bus_error_message(&error, r)); + return 0; + } - } else if (ev[i].data.fd == STDOUT_FILENO) { + return 0; +} - if (ev[i].events & (EPOLLOUT|EPOLLHUP)) - stdout_writable = true; +static int reset_audit_loginuid(void) { + _cleanup_free_ char *p = NULL; + int r; - } else if (ev[i].data.fd == master) { + if (arg_share_system) + return 0; - if (ev[i].events & (EPOLLIN|EPOLLHUP)) - master_readable = true; + r = read_one_line_file("/proc/self/loginuid", &p); + if (r == -ENOENT) + return 0; + if (r < 0) { + log_error("Failed to read /proc/self/loginuid: %s", strerror(-r)); + return r; + } - if (ev[i].events & (EPOLLOUT|EPOLLHUP)) - master_writable = true; + /* Already reset? */ + if (streq(p, "4294967295")) + return 0; - } else if (ev[i].data.fd == signal_fd) { - struct signalfd_siginfo sfsi; - ssize_t n; + r = write_string_file("/proc/self/loginuid", "4294967295"); + if (r < 0) { + log_error("Failed to reset audit login UID. This probably means that your kernel is too\n" + "old and you have audit enabled. Note that the auditing subsystem is known to\n" + "be incompatible with containers on old kernels. Please make sure to upgrade\n" + "your kernel or to off auditing with 'audit=0' on the kernel command line before\n" + "using systemd-nspawn. Sleeping for 5s... (%s)\n", strerror(-r)); - n = read(signal_fd, &sfsi, sizeof(sfsi)); - if (n != sizeof(sfsi)) { + sleep(5); + } - if (n >= 0) { - log_error("Failed to read from signalfd: invalid block size"); - r = -EIO; - goto finish; - } + return 0; +} - if (errno != EINTR && errno != EAGAIN) { - log_error("Failed to read from signalfd: %m"); - r = -errno; - goto finish; - } - } else { +#define HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2) - if (sfsi.ssi_signo == SIGWINCH) { - struct winsize ws; +static int get_mac(struct ether_addr *mac) { + int r; - /* The window size changed, let's forward that. */ - if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0) - ioctl(master, TIOCSWINSZ, &ws); - } else if (sfsi.ssi_signo == SIGTERM && arg_boot && !tried_orderly_shutdown) { + uint8_t result[8]; + size_t l, sz; + uint8_t *v; - log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination."); + l = strlen(arg_machine); + sz = sizeof(sd_id128_t) + l; + v = alloca(sz); - /* This only works for systemd... */ - tried_orderly_shutdown = true; - kill(pid, SIGRTMIN+3); + /* fetch some persistent data unique to the host */ + r = sd_id128_get_machine((sd_id128_t*) v); + if (r < 0) + return r; - } else { - r = 0; - goto finish; - } - } - } - } + /* combine with some data unique (on this host) to this + * container instance */ + memcpy(v + sizeof(sd_id128_t), arg_machine, l); - while ((stdin_readable && in_buffer_full <= 0) || - (master_writable && in_buffer_full > 0) || - (master_readable && out_buffer_full <= 0) || - (stdout_writable && out_buffer_full > 0)) { + /* 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); - if (stdin_readable && in_buffer_full < LINE_MAX) { + assert_cc(ETH_ALEN <= sizeof(result)); + memcpy(mac->ether_addr_octet, result, ETH_ALEN); - k = read(STDIN_FILENO, in_buffer + in_buffer_full, LINE_MAX - in_buffer_full); - if (k < 0) { + /* 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) */ - if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) - stdin_readable = false; - else { - log_error("read(): %m"); - r = -errno; - goto finish; - } - } else - in_buffer_full += (size_t) k; - } + return 0; +} - if (master_writable && in_buffer_full > 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; - k = write(master, in_buffer, in_buffer_full); - if (k < 0) { + if (!arg_private_network) + return 0; - if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) - master_writable = false; - else { - log_error("write(): %m"); - r = -errno; - goto finish; - } + if (!arg_network_veth) + return 0; + + /* Use two different interface name prefixes depending whether + * we are in bridge mode or not. */ + if (arg_network_bridge) + 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)); + return r; + } + + r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0); + if (r < 0) { + log_error("Failed to allocate netlink message: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, iface_name); + if (r < 0) { + log_error("Failed to add netlink interface name: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_open_container(m, IFLA_LINKINFO); + if (r < 0) { + log_error("Failed to open netlink container: %s", strerror(-r)); + 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); + if (r < 0) { + log_error("Failed to open netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_open_container(m, VETH_INFO_PEER); + if (r < 0) { + log_error("Failed to open netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, "host0"); + if (r < 0) { + log_error("Failed to add netlink interface name: %s", strerror(-r)); + 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)); + return r; + } + + r = sd_rtnl_message_close_container(m); + if (r < 0) { + log_error("Failed to close netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_close_container(m); + if (r < 0) { + log_error("Failed to close netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_close_container(m); + if (r < 0) { + log_error("Failed to close netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) { + log_error("Failed to add new veth interfaces: %s", strerror(-r)); + return r; + } + + return 0; +} + +static int setup_bridge(const char veth_name[]) { + _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; + _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; + int r, bridge; + + if (!arg_private_network) + return 0; + + if (!arg_network_veth) + return 0; + + if (!arg_network_bridge) + return 0; + + bridge = (int) if_nametoindex(arg_network_bridge); + if (bridge <= 0) { + log_error("Failed to resolve interface %s: %m", arg_network_bridge); + return -errno; + } + + r = sd_rtnl_open(&rtnl, 0); + if (r < 0) { + log_error("Failed to connect to netlink: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, 0); + if (r < 0) { + log_error("Failed to allocate netlink message: %s", strerror(-r)); + 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)); + return r; + } + + r = sd_rtnl_message_append_u32(m, IFLA_MASTER, bridge); + if (r < 0) { + log_error("Failed to add netlink master field: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) { + log_error("Failed to add veth interface to bridge: %s", strerror(-r)); + return r; + } + + return 0; +} + +static int parse_interface(struct udev *udev, const char *name) { + _cleanup_udev_device_unref_ struct udev_device *d = NULL; + char ifi_str[2 + DECIMAL_STR_MAX(int)]; + int ifi; + + ifi = (int) if_nametoindex(name); + if (ifi <= 0) { + log_error("Failed to resolve interface %s: %m", name); + return -errno; + } + + sprintf(ifi_str, "n%i", ifi); + d = udev_device_new_from_device_id(udev, ifi_str); + if (!d) { + log_error("Failed to get udev device for interface %s: %m", name); + return -errno; + } + + if (udev_device_get_is_initialized(d) <= 0) { + log_error("Network interface %s is not initialized yet.", name); + return -EBUSY; + } + + return ifi; +} + +static int move_network_interfaces(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_interfaces)) + return 0; + + r = sd_rtnl_open(&rtnl, 0); + if (r < 0) { + log_error("Failed to connect to netlink: %s", strerror(-r)); + return r; + } + + udev = udev_new(); + if (!udev) { + log_error("Failed to connect to udev."); + return -ENOMEM; + } + + STRV_FOREACH(i, arg_network_interfaces) { + _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; + int ifi; + + ifi = parse_interface(udev, *i); + if (ifi < 0) + return ifi; + + r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, ifi); + if (r < 0) { + log_error("Failed to allocate netlink message: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); + if (r < 0) { + log_error("Failed to append namespace PID to netlink message: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) { + log_error("Failed to move interface %s to namespace: %s", *i, strerror(-r)); + return r; + } + } + + return 0; +} + +static int setup_macvlan(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_macvlan)) + return 0; + + r = sd_rtnl_open(&rtnl, 0); + if (r < 0) { + log_error("Failed to connect to netlink: %s", strerror(-r)); + return r; + } + + udev = udev_new(); + if (!udev) { + log_error("Failed to connect to udev."); + return -ENOMEM; + } + + STRV_FOREACH(i, arg_network_macvlan) { + _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) { + log_error("Failed to allocate netlink message: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_u32(m, IFLA_LINK, ifi); + if (r < 0) { + log_error("Failed to add netlink interface index: %s", strerror(-r)); + return r; + } + + n = strappend("mv-", *i); + if (!n) + return log_oom(); + + strshorten(n, IFNAMSIZ-1); + + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, n); + if (r < 0) { + log_error("Failed to add netlink interface name: %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)); + return r; + } + + r = sd_rtnl_message_open_container(m, IFLA_LINKINFO); + if (r < 0) { + log_error("Failed to open netlink container: %s", strerror(-r)); + 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); + if (r < 0) { + log_error("Failed to open netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_u32(m, IFLA_MACVLAN_MODE, MACVLAN_MODE_BRIDGE); + if (r < 0) { + log_error("Failed to append macvlan mode: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_close_container(m); + if (r < 0) { + log_error("Failed to close netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_close_container(m); + if (r < 0) { + log_error("Failed to close netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) { + log_error("Failed to add new macvlan interfaces: %s", strerror(-r)); + return r; + } + } + + return 0; +} + +static int audit_still_doesnt_work_in_containers(void) { + +#ifdef HAVE_SECCOMP + scmp_filter_ctx seccomp; + int r; + + /* + Audit is broken in containers, much of the userspace audit + hookup will fail if running inside a container. We don't + care and just turn off creation of audit sockets. + + This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail + with EAFNOSUPPORT which audit userspace uses as indication + that audit is disabled in the kernel. + */ + + seccomp = seccomp_init(SCMP_ACT_ALLOW); + if (!seccomp) + return log_oom(); + + r = seccomp_add_secondary_archs(seccomp); + if (r < 0) { + log_error("Failed to add secondary archs to seccomp filter: %s", strerror(-r)); + goto finish; + } + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EAFNOSUPPORT), + SCMP_SYS(socket), + 2, + SCMP_A0(SCMP_CMP_EQ, AF_NETLINK), + SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT)); + if (r < 0) { + log_error("Failed to add audit seccomp rule: %s", strerror(-r)); + goto finish; + } + + r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0); + if (r < 0) { + log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r)); + goto finish; + } + + r = seccomp_load(seccomp); + if (r < 0) + log_error("Failed to install seccomp audit filter: %s", strerror(-r)); + +finish: + seccomp_release(seccomp); + return r; +#else + return 0; +#endif + +} + +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); + + fd = open(arg_image, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY); + if (fd < 0) { + log_error("Failed to open %s: %m", arg_image); + return -errno; + } + + if (fstat(fd, &st) < 0) { + log_error("Failed to stat %s: %m", arg_image); + return -errno; + } + + 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: %m", arg_image); + return -EINVAL; + } + + control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); + if (control < 0) { + log_error("Failed to open /dev/loop-control: %m"); + return -errno; + } + + nr = ioctl(control, LOOP_CTL_GET_FREE); + if (nr < 0) { + log_error("Failed to allocate loop device: %m"); + return -errno; + } + + 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) { + log_error("Failed to open loop device %s: %m", loopdev); + return -errno; + } + + if (ioctl(loop, LOOP_SET_FD, fd) < 0) { + log_error("Failed to set loopback file descriptor on %s: %m", loopdev); + return -errno; + } + + if (arg_read_only) + info.lo_flags |= LO_FLAGS_READ_ONLY; + + if (ioctl(loop, LOOP_SET_STATUS64, &info) < 0) { + log_error("Failed to set loopback settings on %s: %m", loopdev); + return -errno; + } + + *device_path = loopdev; + loopdev = NULL; + + *loop_nr = nr; + + r = loop; + loop = -1; + + return r; +} + +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, + bool *secondary) { + +#ifdef HAVE_BLKID + int home_nr = -1, root_nr = -1, secondary_root_nr = -1, srv_nr = -1; + _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = 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; + const char *pttype = NULL; + blkid_partlist pl; + struct stat st; + int r; + + assert(fd >= 0); + assert(root_device); + assert(home_device); + assert(srv_device); + assert(secondary); + + 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(); + + log_error("Failed to set device on blkid probe: %m"); + return -errno; + } + + 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 %s.\n" + "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); + return -EINVAL; + } else if (r != 0) { + if (errno == 0) + errno = EIO; + log_error("Failed to probe: %m"); + return -errno; + } + + blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL); + if (!streq_ptr(pttype, "gpt")) { + log_error("Image %s does not carry a GUID Partition Table.\n" + "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); + 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) { + log_error("Failed to stat block device: %m"); + return -errno; + } + + d = udev_device_new_from_devnum(udev, 'b', st.st_rdev); + if (!d) + return log_oom(); + + 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) { + log_error("Failed to scan for partition devices of %s: %s", arg_image, strerror(-r)); + return r; + } + + first = udev_enumerate_get_list_entry(e); + udev_list_entry_foreach(item, first) { + _cleanup_udev_device_unref_ struct udev_device *q; + const char *stype, *node; + unsigned long long flags; + sd_id128_t type_id; + 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; + + log_error("Failed to get partition device of %s: %m", arg_image); + 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); + if (flags & GPT_FLAG_NO_AUTO) + continue; + + nr = blkid_partition_get_partno(pp); + if (nr < 0) + continue; + + stype = blkid_partition_get_type_string(pp); + if (!stype) + continue; + + if (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); + + free(home); + home = strdup(node); + if (!home) + 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); + + free(srv); + srv = strdup(node); + if (!srv) + 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); + + free(root); + root = strdup(node); + if (!root) + 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); + + + free(secondary_root); + secondary_root = strdup(node); + if (!secondary_root) + return log_oom(); + } +#endif + } + + if (!root && !secondary_root) { + log_error("Failed to identify root partition in disk image %s.\n" + "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); + return -EINVAL; + } + + if (root) { + *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; +#else + log_error("--image= is not supported, compiled without blkid support."); + return -ENOTSUP; +#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; + int r; + + assert(what); + assert(where); + + if (arg_read_only) + rw = false; + + if (directory) + p = strappenda(where, directory); + else + p = where; + + errno = 0; + b = blkid_new_probe_from_filename(what); + if (!b) { + if (errno == 0) + return log_oom(); + log_error("Failed to allocate prober for %s: %m", what); + return -errno; + } + + 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; + log_error("Failed to probe %s: %m", what); + return -errno; + } + + 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 -ENOTSUP; + } + + if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0) { + log_error("Failed to mount %s: %m", what); + return -errno; + } + + return 0; +#else + log_error("--image= is not supported, compiled without blkid support."); + return -ENOTSUP; +#endif +} + +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, root_device_rw); + if (r < 0) { + log_error("Failed to mount root directory: %s", strerror(-r)); + return r; + } + } + + if (home_device) { + 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; + } + } + + if (srv_device) { + 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; + } + } + + return 0; +} + +static void loop_remove(int nr, int *image_fd) { + _cleanup_close_ int control = -1; + + if (nr < 0) + return; + + if (image_fd && *image_fd >= 0) { + ioctl(*image_fd, LOOP_CLR_FD); + *image_fd = safe_close(*image_fd); + } + + control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); + if (control < 0) + return; + + ioctl(control, LOOP_CTL_REMOVE, nr); +} + +static int spawn_getent(const char *database, const char *key, pid_t *rpid) { + int pipe_fds[2]; + pid_t pid; + + assert(database); + assert(key); + assert(rpid); + + if (pipe2(pipe_fds, O_CLOEXEC) < 0) { + log_error("Failed to allocate pipe: %m"); + return -errno; + } + + pid = fork(); + if (pid < 0) { + log_error("Failed to fork getent child: %m"); + return -errno; + } else if (pid == 0) { + int nullfd; + char *empty_env = NULL; + + if (dup3(pipe_fds[1], STDOUT_FILENO, 0) < 0) + _exit(EXIT_FAILURE); + + if (pipe_fds[0] > 2) + safe_close(pipe_fds[0]); + if (pipe_fds[1] > 2) + safe_close(pipe_fds[1]); + + nullfd = open("/dev/null", O_RDWR); + if (nullfd < 0) + _exit(EXIT_FAILURE); + + if (dup3(nullfd, STDIN_FILENO, 0) < 0) + _exit(EXIT_FAILURE); + + if (dup3(nullfd, STDERR_FILENO, 0) < 0) + _exit(EXIT_FAILURE); + + if (nullfd > 2) + safe_close(nullfd); + + reset_all_signal_handlers(); + close_all_fds(NULL, 0); + + execle("/usr/bin/getent", "getent", database, key, NULL, &empty_env); + execle("/bin/getent", "getent", database, key, NULL, &empty_env); + _exit(EXIT_FAILURE); + } + + pipe_fds[1] = safe_close(pipe_fds[1]); - } else { - assert(in_buffer_full >= (size_t) k); - memmove(in_buffer, in_buffer + k, in_buffer_full - k); - in_buffer_full -= k; - } - } + *rpid = pid; - if (master_readable && out_buffer_full < LINE_MAX) { + return pipe_fds[0]; +} - k = read(master, out_buffer + out_buffer_full, LINE_MAX - out_buffer_full); - if (k < 0) { +static int change_uid_gid(char **_home) { + 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; + uid_t uid; + gid_t gid; + pid_t pid; + int r; - if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) - master_readable = false; - else { - log_error("read(): %m"); - r = -errno; - goto finish; - } - } else - out_buffer_full += (size_t) k; - } + assert(_home); - if (stdout_writable && out_buffer_full > 0) { + if (!arg_user || streq(arg_user, "root") || streq(arg_user, "0")) { + /* Reset everything fully to 0, just in case */ - k = write(STDOUT_FILENO, out_buffer, out_buffer_full); - if (k < 0) { + if (setgroups(0, NULL) < 0) { + log_error("setgroups() failed: %m"); + return -errno; + } - if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) - stdout_writable = false; - else { - log_error("write(): %m"); - r = -errno; - goto finish; - } + if (setresgid(0, 0, 0) < 0) { + log_error("setregid() failed: %m"); + return -errno; + } - } else { - assert(out_buffer_full >= (size_t) k); - memmove(out_buffer, out_buffer + k, out_buffer_full - k); - out_buffer_full -= k; - } - } + if (setresuid(0, 0, 0) < 0) { + log_error("setreuid() failed: %m"); + return -errno; } + + *_home = NULL; + return 0; } -finish: - if (ep >= 0) - close_nointr_nofail(ep); + /* First, get user credentials */ + fd = spawn_getent("passwd", arg_user, &pid); + if (fd < 0) + return fd; - if (signal_fd >= 0) - close_nointr_nofail(signal_fd); + f = fdopen(fd, "r"); + if (!f) + return log_oom(); + fd = -1; - return r; -} + if (!fgets(line, sizeof(line), f)) { -static int register_machine(void) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; - _cleanup_bus_unref_ sd_bus *bus = NULL; - int r; + if (!ferror(f)) { + log_error("Failed to resolve user %s.", arg_user); + return -ESRCH; + } + + log_error("Failed to read from getent: %m"); + return -errno; + } + + truncate_nl(line); - r = sd_bus_open_system(&bus); + wait_for_terminate_and_warn("getent passwd", pid); + + x = strchr(line, ':'); + if (!x) { + log_error("/etc/passwd entry has invalid user field."); + return -EIO; + } + + u = strchr(x+1, ':'); + if (!u) { + log_error("/etc/passwd entry has invalid password field."); + return -EIO; + } + + u++; + g = strchr(u, ':'); + if (!g) { + log_error("/etc/passwd entry has invalid UID field."); + return -EIO; + } + + *g = 0; + g++; + x = strchr(g, ':'); + if (!x) { + log_error("/etc/passwd entry has invalid GID field."); + return -EIO; + } + + *x = 0; + h = strchr(x+1, ':'); + if (!h) { + log_error("/etc/passwd entry has invalid GECOS field."); + return -EIO; + } + + h++; + x = strchr(h, ':'); + if (!x) { + log_error("/etc/passwd entry has invalid home directory field."); + return -EIO; + } + + *x = 0; + + r = parse_uid(u, &uid); if (r < 0) { - log_error("Failed to open system bus: %s", strerror(-r)); - return r; + log_error("Failed to parse UID of user."); + return -EIO; } - r = sd_bus_call_method( - bus, - "org.freedesktop.machine1", - "/org/freedesktop/machine1", - "org.freedesktop.machine1.Manager", - "CreateMachine", - &error, - NULL, - "sayssusa(sv)", - arg_machine, - SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), - "nspawn", - "container", - (uint32_t) 0, - strempty(arg_directory), - !isempty(arg_slice), "Slice", "s", arg_slice); + r = parse_gid(g, &gid); + if (r < 0) { + log_error("Failed to parse GID of user."); + return -EIO; + } + + home = strdup(h); + if (!home) + return log_oom(); + + /* Second, get group memberships */ + fd = spawn_getent("initgroups", arg_user, &pid); + if (fd < 0) + return fd; + + fclose(f); + f = fdopen(fd, "r"); + if (!f) + return log_oom(); + fd = -1; + + if (!fgets(line, sizeof(line), f)) { + if (!ferror(f)) { + log_error("Failed to resolve user %s.", arg_user); + return -ESRCH; + } + + log_error("Failed to read from getent: %m"); + return -errno; + } + + truncate_nl(line); + + wait_for_terminate_and_warn("getent initgroups", pid); + + /* Skip over the username and subsequent separator whitespace */ + x = line; + x += strcspn(x, WHITESPACE); + x += strspn(x, WHITESPACE); + + FOREACH_WORD(w, l, x, state) { + char c[l+1]; + + memcpy(c, w, l); + c[l] = 0; + + if (!GREEDY_REALLOC(uids, sz, n_uids+1)) + return log_oom(); + + r = parse_uid(c, &uids[n_uids++]); + if (r < 0) { + log_error("Failed to parse group data from getent."); + return -EIO; + } + } + + r = mkdir_parents(home, 0775); if (r < 0) { - log_error("Failed to register machine: %s", error.message ? error.message : strerror(-r)); + log_error("Failed to make home root directory: %s", strerror(-r)); return r; } - return 0; -} + r = mkdir_safe(home, 0755, uid, gid); + if (r < 0 && r != -EEXIST) { + log_error("Failed to make home directory: %s", strerror(-r)); + return r; + } + + fchown(STDIN_FILENO, uid, gid); + fchown(STDOUT_FILENO, uid, gid); + fchown(STDERR_FILENO, uid, gid); + + if (setgroups(n_uids, uids) < 0) { + log_error("Failed to set auxiliary groups: %m"); + return -errno; + } + + if (setresgid(gid, gid, gid) < 0) { + log_error("setregid() failed: %m"); + return -errno; + } -static bool audit_enabled(void) { - int fd; + if (setresuid(uid, uid, uid) < 0) { + log_error("setreuid() failed: %m"); + return -errno; + } - fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_AUDIT); - if (fd >= 0) { - close_nointr_nofail(fd); - return true; + if (_home) { + *_home = home; + home = NULL; } - return false; + + return 0; } int main(int argc, char *argv[]) { - pid_t pid = 0; - int r = EXIT_FAILURE, k; - _cleanup_close_ int master = -1; - int n_fd_passed; + + _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_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; - struct termios saved_attr, raw_attr; + char veth_name[IFNAMSIZ]; + bool secondary = false; + pid_t pid = 0; sigset_t mask; - bool saved_attr_valid = false; - struct winsize ws; - int kmsg_socket_pair[2] = { -1, -1 }; - _cleanup_fdset_free_ FDSet *fds = NULL; log_parse_environment(); log_open(); @@ -1235,24 +2602,25 @@ int main(int argc, char *argv[]) { goto finish; } - if (arg_directory) { - char *p; + if (!arg_image) { + if (arg_directory) { + char *p; - p = path_make_absolute_cwd(arg_directory); - free(arg_directory); - arg_directory = p; - } else - arg_directory = get_current_dir_name(); + p = path_make_absolute_cwd(arg_directory); + free(arg_directory); + arg_directory = p; + } else + arg_directory = get_current_dir_name(); - if (!arg_directory) { - log_error("Failed to determine path, please use -D."); - goto finish; + if (!arg_directory) { + log_error("Failed to determine path, please use -D."); + goto finish; + } + path_kill_slashes(arg_directory); } - path_kill_slashes(arg_directory); - if (!arg_machine) { - arg_machine = strdup(path_get_file_name(arg_directory)); + arg_machine = strdup(basename(arg_image ? arg_image : arg_directory)); if (!arg_machine) { log_oom(); goto finish; @@ -1275,23 +2643,6 @@ int main(int argc, char *argv[]) { goto finish; } - if (arg_boot && audit_enabled()) { - log_warning("The kernel auditing subsystem is known to be incompatible with containers.\n" - "Please make sure to turn off auditing with 'audit=0' on the kernel command\n" - "line before using systemd-nspawn. Sleeping for 5s...\n"); - sleep(5); - } - - if (path_equal(arg_directory, "/")) { - log_error("Spawning container on root directory not supported."); - goto finish; - } - - if (path_is_os_tree(arg_directory) <= 0) { - log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory); - goto finish; - } - log_close(); n_fd_passed = sd_listen_fds(false); if (n_fd_passed > 0) { @@ -1304,6 +2655,54 @@ int main(int argc, char *argv[]) { fdset_close_others(fds); log_open(); + if (arg_directory) { + if (path_equal(arg_directory, "/")) { + log_error("Spawning container on root directory not supported."); + goto finish; + } + + if (arg_boot) { + if (path_is_os_tree(arg_directory) <= 0) { + log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory); + goto finish; + } + } else { + const char *p; + + p = strappenda(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); + goto finish; + + } + } + } else { + char template[] = "/tmp/nspawn-root-XXXXXX"; + + if (!mkdtemp(template)) { + log_error("Failed to create temporary directory: %m"); + r = -errno; + goto finish; + } + + arg_directory = strdup(template); + if (!arg_directory) { + r = log_oom(); + goto finish; + } + + image_fd = setup_image(&device_path, &loop_nr); + if (image_fd < 0) { + r = image_fd; + goto finish; + } + + 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; + } + master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); if (master < 0) { log_error("Failed to acquire pseudo tty: %m"); @@ -1316,26 +2715,36 @@ int main(int argc, char *argv[]) { goto finish; } - log_info("Spawning namespace container on %s (console is %s).", arg_directory, console); - - if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0) - ioctl(master, TIOCSWINSZ, &ws); + if (!arg_quiet) + log_info("Spawning container %s on %s. Press ^] three times within 1s to abort execution.", arg_machine, arg_image ? arg_image : arg_directory); if (unlockpt(master) < 0) { log_error("Failed to unlock tty: %m"); goto finish; } - if (tcgetattr(STDIN_FILENO, &saved_attr) >= 0) { - saved_attr_valid = true; + if (access("/dev/kdbus/control", F_OK) >= 0) { + + if (arg_share_system) { + kdbus_domain = strdup("/dev/kdbus"); + if (!kdbus_domain) { + log_oom(); + goto finish; + } + } else { + const char *ns; - raw_attr = saved_attr; - cfmakeraw(&raw_attr); - raw_attr.c_lflag &= ~ECHO; + ns = strappenda("machine-", arg_machine); + kdbus_fd = bus_kernel_create_domain(ns, &kdbus_domain); + if (r < 0) + log_debug("Failed to create kdbus domain: %s", strerror(-r)); + else + log_debug("Successfully created kdbus domain as %s", kdbus_domain); + } } if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { - log_error("Failed to create kmsg socket pair."); + log_error("Failed to create kmsg socket pair: %m"); goto finish; } @@ -1346,21 +2755,26 @@ int main(int argc, char *argv[]) { assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); for (;;) { + int parent_ready_fd = -1, child_ready_fd = -1; siginfo_t status; - int pipefd[2], pipefd2[2]; + eventfd_t x; - if (pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) { - log_error("pipe2(): %m"); + parent_ready_fd = eventfd(0, EFD_CLOEXEC); + if (parent_ready_fd < 0) { + log_error("Failed to create event fd: %m"); goto finish; } - if (pipe2(pipefd2, O_NONBLOCK|O_CLOEXEC) < 0) { - log_error("pipe2(): %m"); - close_pipe(pipefd); + child_ready_fd = eventfd(0, EFD_CLOEXEC); + if (child_ready_fd < 0) { + log_error("Failed to create event fd: %m"); goto finish; } - pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL); + pid = syscall(__NR_clone, + SIGCHLD|CLONE_NEWNS| + (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)| + (arg_private_network ? CLONE_NEWNET : 0), NULL); if (pid < 0) { if (errno == EINVAL) log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m"); @@ -1372,12 +2786,10 @@ int main(int argc, char *argv[]) { if (pid == 0) { /* child */ - const char *home = NULL; - uid_t uid = (uid_t) -1; - gid_t gid = (gid_t) -1; + _cleanup_free_ char *home = NULL; unsigned n_env = 2; const char *envp[] = { - "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", + "PATH=" DEFAULT_PATH_SPLIT_USR, "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */ NULL, /* TERM */ NULL, /* HOME */ @@ -1388,32 +2800,19 @@ int main(int argc, char *argv[]) { NULL, /* LISTEN_PID */ NULL }; + char **env_use; envp[n_env] = strv_find_prefix(environ, "TERM="); if (envp[n_env]) n_env ++; - /* Wait for the parent process to log our PID */ - close_nointr_nofail(pipefd[1]); - fd_wait_for_event(pipefd[0], POLLHUP, -1); - close_nointr_nofail(pipefd[0]); - - close_nointr_nofail(master); - master = -1; - - if (saved_attr_valid) { - if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) { - log_error("Failed to set terminal attributes: %m"); - goto child_fail; - } - } + 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(); @@ -1423,7 +2822,7 @@ 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; } @@ -1442,17 +2841,14 @@ int main(int argc, char *argv[]) { goto child_fail; } + if (reset_audit_loginuid() < 0) + goto child_fail; + if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) { log_error("PR_SET_PDEATHSIG failed: %m"); goto child_fail; } - close_pipe(pipefd2); - - r = register_machine(); - if (r < 0) - goto finish; - /* Mark everything as slave, so that we still * receive mounts from the real root, but don't * propagate mounts to the real root. */ @@ -1461,6 +2857,12 @@ int main(int argc, char *argv[]) { goto child_fail; } + 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 */ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) { log_error("Failed to make bind mount."); @@ -1484,14 +2886,16 @@ int main(int argc, char *argv[]) { dev_setup(arg_directory); + if (audit_still_doesnt_work_in_containers() < 0) + goto child_fail; + if (setup_dev_console(arg_directory, console) < 0) goto child_fail; 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; @@ -1511,6 +2915,15 @@ int main(int argc, char *argv[]) { if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0) goto child_fail; + if (setup_kdbus(arg_directory, kdbus_domain) < 0) + goto child_fail; + + /* Tell the parent that we are ready, and that + * it can cgroupify us to that we lack access + * to certain devices and resources. */ + eventfd_write(child_ready_fd, 1); + child_ready_fd = safe_close(child_ready_fd); + if (chdir(arg_directory) < 0) { log_error("chdir(%s) failed: %m", arg_directory); goto child_fail; @@ -1533,68 +2946,17 @@ int main(int argc, char *argv[]) { umask(0022); - loopback_setup(); + if (arg_private_network) + loopback_setup(); if (drop_capabilities() < 0) { log_error("drop_capabilities() failed: %m"); goto child_fail; } - if (arg_user) { - - /* Note that this resolves user names - * inside the container, and hence - * accesses the NSS modules from the - * container and not the host. This is - * a bit weird... */ - - if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) { - log_error("get_user_creds() failed: %m"); - goto child_fail; - } - - if (mkdir_parents_label(home, 0775) < 0) { - log_error("mkdir_parents_label() failed: %m"); - goto child_fail; - } - - if (mkdir_safe_label(home, 0775, uid, gid) < 0) { - log_error("mkdir_safe_label() failed: %m"); - goto child_fail; - } - - if (initgroups((const char*)arg_user, gid) < 0) { - log_error("initgroups() failed: %m"); - goto child_fail; - } - - if (setresgid(gid, gid, gid) < 0) { - log_error("setregid() failed: %m"); - goto child_fail; - } - - if (setresuid(uid, uid, uid) < 0) { - log_error("setreuid() failed: %m"); - goto child_fail; - } - } else { - /* Reset everything fully to 0, just in case */ - - if (setgroups(0, NULL) < 0) { - log_error("setgroups() failed: %m"); - goto child_fail; - } - - if (setresgid(0, 0, 0) < 0) { - log_error("setregid() failed: %m"); - goto child_fail; - } - - if (setresuid(0, 0, 0) < 0) { - log_error("setreuid() failed: %m"); - goto child_fail; - } - } + r = change_uid_gid(&home); + if (r < 0) + goto child_fail; if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) || (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) || @@ -1626,6 +2988,43 @@ int main(int argc, char *argv[]) { setup_hostname(); + if (arg_personality != 0xffffffffLU) { + if (personality(arg_personality) < 0) { + log_error("personality() failed: %m"); + goto child_fail; + } + } else if (secondary) { + if (personality(PER_LINUX32) < 0) { + log_error("personality() failed: %m"); + goto child_fail; + } + } + +#ifdef HAVE_SELINUX + if (arg_selinux_context) + if (setexeccon((security_context_t) arg_selinux_context) < 0) { + log_error("setexeccon(\"%s\") failed: %m", arg_selinux_context); + goto child_fail; + } +#endif + + if (!strv_isempty(arg_setenv)) { + char **n; + + n = strv_env_merge(2, envp, arg_setenv); + if (!n) { + log_oom(); + goto child_fail; + } + + env_use = n; + } else + env_use = (char**) envp; + + /* Wait until the parent is ready with the setup, too... */ + eventfd_read(parent_ready_fd, &x); + parent_ready_fd = safe_close(parent_ready_fd); + if (arg_boot) { char **a; size_t l; @@ -1637,18 +3036,19 @@ int main(int argc, char *argv[]) { memcpy(a + 1, argv + optind, l * sizeof(char*)); a[0] = (char*) "/usr/lib/systemd/systemd"; - execve(a[0], a, (char**) envp); + execve(a[0], a, env_use); a[0] = (char*) "/lib/systemd/systemd"; - execve(a[0], a, (char**) envp); + execve(a[0], a, env_use); a[0] = (char*) "/sbin/init"; - execve(a[0], a, (char**) envp); + execve(a[0], a, env_use); } else if (argc > optind) - execvpe(argv[optind], argv + optind, (char**) envp); + execvpe(argv[optind], argv + optind, env_use); else { chdir(home ? home : "/root"); - execle("/bin/bash", "-bash", NULL, (char**) envp); + execle("/bin/bash", "-bash", NULL, env_use); + execle("/bin/sh", "-sh", NULL, env_use); } log_error("execv() failed: %m"); @@ -1657,25 +3057,58 @@ int main(int argc, char *argv[]) { _exit(EXIT_FAILURE); } - log_info("Init process in the container running as PID %lu.", (unsigned long) pid); - close_nointr_nofail(pipefd[0]); - close_nointr_nofail(pipefd[1]); - - /* Wait for the child process to establish cgroup hierarchy */ - close_nointr_nofail(pipefd2[1]); - fd_wait_for_event(pipefd2[0], POLLHUP, -1); - close_nointr_nofail(pipefd2[0]); - fdset_free(fds); fds = NULL; - if (process_pty(master, pid, &mask) < 0) + /* Wait until the child reported that it is ready with + * all it needs to do with priviliges. After we got + * the notification we can make the process join its + * cgroup which might limit what it can do */ + eventfd_read(child_ready_fd, &x); + + r = register_machine(pid); + if (r < 0) + goto finish; + + r = move_network_interfaces(pid); + if (r < 0) + goto finish; + + r = setup_veth(pid, veth_name); + if (r < 0) + goto finish; + + r = setup_bridge(veth_name); + if (r < 0) + goto finish; + + r = setup_macvlan(pid); + if (r < 0) goto finish; - if (saved_attr_valid) - tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); + /* Notify the child that the parent is ready with all + * its setup, and thtat the child can now hand over + * control to the code to run inside the container. */ + eventfd_write(parent_ready_fd, 1); + + k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3); + if (k < 0) { + r = EXIT_FAILURE; + break; + } + + if (!arg_quiet) + putc('\n', stdout); + + /* Kill if it is not dead yet anyway */ + terminate_machine(pid); + + /* Redundant, but better safe than sorry */ + kill(pid, SIGKILL); k = wait_for_terminate(pid, &status); + pid = 0; + if (k < 0) { r = EXIT_FAILURE; break; @@ -1684,45 +3117,53 @@ int main(int argc, char *argv[]) { if (status.si_code == CLD_EXITED) { r = status.si_status; if (status.si_status != 0) { - log_error("Container failed with error code %i.", status.si_status); + log_error("Container %s failed with error code %i.", arg_machine, status.si_status); break; } - log_debug("Container exited successfully."); + if (!arg_quiet) + log_debug("Container %s exited successfully.", arg_machine); break; } else if (status.si_code == CLD_KILLED && status.si_status == SIGINT) { - log_info("Container has been shut down."); + + if (!arg_quiet) + log_info("Container %s has been shut down.", arg_machine); r = 0; break; } else if (status.si_code == CLD_KILLED && status.si_status == SIGHUP) { - log_info("Container is being rebooted."); + + if (!arg_quiet) + log_info("Container %s is being rebooted.", arg_machine); continue; } else if (status.si_code == CLD_KILLED || status.si_code == CLD_DUMPED) { - log_error("Container terminated by signal %s.", signal_to_string(status.si_status)); + log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status)); r = EXIT_FAILURE; break; } else { - log_error("Container failed due to unknown reason."); + log_error("Container %s failed due to unknown reason.", arg_machine); r = EXIT_FAILURE; break; } } finish: - if (saved_attr_valid) - tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); - - close_pipe(kmsg_socket_pair); + loop_remove(loop_nr, &image_fd); if (pid > 0) kill(pid, SIGKILL); free(arg_directory); free(arg_machine); + free(arg_user); + strv_free(arg_setenv); + strv_free(arg_network_interfaces); + strv_free(arg_network_macvlan); + strv_free(arg_bind); + strv_free(arg_bind_ro); return r; }