X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn.c;h=2c1144a7f5b6f3b9d22f133d1d3cccd7afb7e2a4;hp=cd528deb79314e487c0d042a96000be02d952e6d;hb=ff01d048b4c1455241c894cf7982662c9d28fd34;hpb=fd14078a3ab2110cd10e5eb55cdaeecfa51a189c diff --git a/src/nspawn.c b/src/nspawn.c index cd528deb7..2c1144a7f 100644 --- a/src/nspawn.c +++ b/src/nspawn.c @@ -36,6 +36,7 @@ #include #include #include +#include #include "log.h" #include "util.h" @@ -43,15 +44,20 @@ #include "cgroup-util.h" #include "sd-daemon.h" #include "strv.h" +#include "loopback-setup.h" static char *arg_directory = NULL; +static char *arg_user = NULL; +static bool arg_private_network = false; 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" - " -D --directory=NAME Root directory for the container\n", + " -D --directory=NAME Root directory for the container\n" + " -u --user=USER Run the command under specified user or uid\n" + " --private-network Disable network in container\n", program_invocation_short_name); return 0; @@ -59,10 +65,16 @@ static int help(void) { static int parse_argv(int argc, char *argv[]) { + enum { + ARG_PRIVATE_NETWORK = 0x100 + }; + static const struct option options[] = { - { "help", no_argument, NULL, 'h' }, - { "directory", required_argument, NULL, 'D' }, - { NULL, 0, NULL, 0 } + { "help", no_argument, NULL, 'h' }, + { "directory", required_argument, NULL, 'D' }, + { "user", required_argument, NULL, 'u' }, + { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK }, + { NULL, 0, NULL, 0 } }; int c; @@ -70,7 +82,7 @@ static int parse_argv(int argc, char *argv[]) { assert(argc >= 0); assert(argv); - while ((c = getopt_long(argc, argv, "+hD:", options, NULL)) >= 0) { + while ((c = getopt_long(argc, argv, "+hD:u:", options, NULL)) >= 0) { switch (c) { @@ -87,6 +99,19 @@ static int parse_argv(int argc, char *argv[]) { break; + case 'u': + free(arg_user); + if (!(arg_user = strdup(optarg))) { + log_error("Failed to duplicate user name."); + return -ENOMEM; + } + + break; + + case ARG_PRIVATE_NETWORK: + arg_private_network = true; + break; + case '?': return -EINVAL; @@ -111,23 +136,25 @@ static int mount_all(const char *dest) { } MountPoint; static const MountPoint mount_table[] = { - { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, - { "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */ - { "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ - { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, true }, - { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID, true }, - { "/dev/pts", "/dev/pts", "bind", NULL, MS_BIND, true }, - { "tmpfs", "/dev/.run", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, + { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, + { "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */ + { "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ + { "/sys", "/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */ + { "/sys", "/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ + { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID, true }, + { "/dev/pts", "/dev/pts", "bind", NULL, MS_BIND, true }, + { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV, true }, #ifdef HAVE_SELINUX - { "selinux", "/selinux", "selinuxfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, false }, + { "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND, false }, /* Bind mount first */ + { "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */ #endif }; unsigned k; int r = 0; + char *where; for (k = 0; k < ELEMENTSOF(mount_table); k++) { - char *where; int t; if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) { @@ -167,6 +194,13 @@ static int mount_all(const char *dest) { free(where); } + /* Fix the timezone, if possible */ + if (asprintf(&where, "%s/%s", dest, "/etc/localtime") >= 0) { + mount("/etc/localtime", where, "bind", MS_BIND, NULL); + mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); + free(where); + } + return r; } @@ -292,7 +326,6 @@ static int copy_devnodes(const char *dest, const char *console) { } finish: - umask(u); return r; @@ -325,7 +358,7 @@ static int drop_capabilities(void) { unsigned long l; - for (l = 0; l <= MAX(63LU, (unsigned long) CAP_LAST_CAP); l ++) { + for (l = 0; l <= MAX(63LU, (unsigned long) CAP_LAST_CAP); l++) { unsigned i; for (i = 0; i < ELEMENTSOF(retain); i++) @@ -340,7 +373,7 @@ static int drop_capabilities(void) { /* If this capability is not known, EINVAL * will be returned, let's ignore this. */ if (errno == EINVAL) - continue; + break; log_error("PR_CAPBSET_DROP failed: %m"); return -errno; @@ -367,6 +400,7 @@ static int is_os_tree(const char *path) { #define BUFFER_SIZE 1024 static int process_pty(int master, sigset_t *mask) { + char in_buffer[BUFFER_SIZE], out_buffer[BUFFER_SIZE]; size_t in_buffer_full = 0, out_buffer_full = 0; struct epoll_event stdin_ev, stdout_ev, master_ev, signal_ev; @@ -457,11 +491,13 @@ static int process_pty(int master, sigset_t *mask) { if ((n = read(signal_fd, &sfsi, sizeof(sfsi))) != sizeof(sfsi)) { if (n >= 0) { + log_error("Failed to read from signalfd: invalid block size"); r = -EIO; goto finish; } if (errno != EINTR && errno != EAGAIN) { + log_error("Failed to read from signalfd: %m"); r = -errno; goto finish; } @@ -474,7 +510,7 @@ static int process_pty(int master, sigset_t *mask) { if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0) ioctl(master, TIOCSWINSZ, &ws); } else { - r = -EINTR; + r = 0; goto finish; } } @@ -494,6 +530,7 @@ static int process_pty(int master, sigset_t *mask) { stdin_readable = false; else { log_error("read(): %m"); + r = -errno; goto finish; } } else @@ -508,6 +545,7 @@ static int process_pty(int master, sigset_t *mask) { master_writable = false; else { log_error("write(): %m"); + r = -errno; goto finish; } @@ -526,6 +564,7 @@ static int process_pty(int master, sigset_t *mask) { master_readable = false; else { log_error("read(): %m"); + r = -errno; goto finish; } } else @@ -540,6 +579,7 @@ static int process_pty(int master, sigset_t *mask) { stdout_writable = false; else { log_error("write(): %m"); + r = -errno; goto finish; } @@ -670,7 +710,7 @@ int main(int argc, char *argv[]) { sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); - if ((pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS, NULL)) < 0) { + if ((pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL)) < 0) { log_error("clone() failed: %m"); goto finish; } @@ -679,14 +719,19 @@ int main(int argc, char *argv[]) { /* child */ const char *hn; + const char *home = NULL; + uid_t uid = (uid_t) -1; + gid_t gid = (gid_t) -1; const char *envp[] = { - "HOME=/root", "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", - NULL, + NULL, /* TERM */ + NULL, /* HOME */ + NULL, /* USER */ + NULL, /* LOGNAME */ NULL }; - envp[2] = strv_find_prefix(environ, "TERM="); + envp[1] = strv_find_prefix(environ, "TERM="); close_nointr_nofail(master); @@ -707,6 +752,10 @@ int main(int argc, char *argv[]) { if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) goto child_fail; + /* Mark / as private, in case somebody marked it shared */ + if (mount(NULL, "/", NULL, MS_PRIVATE|MS_REC, NULL) < 0) + goto child_fail; + if (mount_all(arg_directory) < 0) goto child_fail; @@ -738,18 +787,60 @@ int main(int argc, char *argv[]) { goto child_fail; } - umask(0002); + umask(0022); + + loopback_setup(); if (drop_capabilities() < 0) goto child_fail; + if (arg_user) { + + if (get_user_creds((const char**)&arg_user, &uid, &gid, &home) < 0) { + log_error("get_user_creds() failed: %m"); + goto child_fail; + } + + if (mkdir_parents(home, 0775) < 0) { + log_error("mkdir_parents() failed: %m"); + goto child_fail; + } + + if (safe_mkdir(home, 0775, uid, gid) < 0) { + log_error("safe_mkdir() 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; + } + } + + if ((asprintf((char**)(envp + 2), "HOME=%s", home? home: "/root") < 0) || + (asprintf((char**)(envp + 3), "USER=%s", arg_user? arg_user : "root") < 0) || + (asprintf((char**)(envp + 4), "LOGNAME=%s", arg_user? arg_user : "root") < 0)) { + log_error("Out of memory"); + goto child_fail; + } + if ((hn = file_name_from_path(arg_directory))) sethostname(hn, strlen(hn)); if (argc > optind) execvpe(argv[optind], argv + optind, (char**) envp); else { - chdir("/root"); + chdir(home ? home : "/root"); execle("/bin/bash", "-bash", NULL, (char**) envp); }