* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
-#define _GNU_SOURCE
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE 1
+#endif
#include <errno.h>
#include <fcntl.h>
#include <grp.h>
+#include <linux/unistd.h>
#include <pwd.h>
+#include <pthread.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
-#include <unistd.h>
-#include <linux/unistd.h>
#include <sys/mount.h>
#include <sys/param.h>
#include <sys/prctl.h>
#include <sys/socket.h>
#include <sys/syscall.h>
#include <sys/wait.h>
+#include <termios.h>
+#include <unistd.h>
#include <lxc/lxccontainer.h>
-#ifndef HAVE_DECL_PR_CAPBSET_DROP
-#define PR_CAPBSET_DROP 24
-#endif
-
-#ifndef HAVE_DECL_PR_SET_NO_NEW_PRIVS
-#define PR_SET_NO_NEW_PRIVS 38
-#endif
-
-#ifndef HAVE_DECL_PR_GET_NO_NEW_PRIVS
-#define PR_GET_NO_NEW_PRIVS 39
-#endif
-
#include "af_unix.h"
#include "attach.h"
#include "caps.h"
#include "lsm/lsm.h"
#include "lxclock.h"
#include "lxcseccomp.h"
+#include "macro.h"
#include "mainloop.h"
#include "namespace.h"
+#include "raw_syscalls.h"
+#include "syscall_wrappers.h"
#include "terminal.h"
#include "utils.h"
#include <sys/personality.h>
#endif
-#ifndef SOCK_CLOEXEC
-#define SOCK_CLOEXEC 02000000
-#endif
-
-#ifndef MS_REC
-#define MS_REC 16384
-#endif
-
-#ifndef MS_SLAVE
-#define MS_SLAVE (1 << 19)
-#endif
+lxc_log_define(attach, lxc);
-lxc_log_define(lxc_attach, lxc);
+/* Define default options if no options are supplied by the user. */
+static lxc_attach_options_t attach_static_default_options = LXC_ATTACH_OPTIONS_DEFAULT;
-/* /proc/pid-to-str/status\0 = (5 + 21 + 7 + 1) */
-#define __PROC_STATUS_LEN (5 + (LXC_NUMSTRLEN64) + 7 + 1)
static struct lxc_proc_context_info *lxc_proc_get_context_info(pid_t pid)
{
int ret;
bool found;
FILE *proc_file;
- char proc_fn[__PROC_STATUS_LEN];
+ char proc_fn[LXC_PROC_STATUS_LEN];
size_t line_bufsz = 0;
char *line = NULL;
struct lxc_proc_context_info *info = NULL;
/* Read capabilities. */
- ret = snprintf(proc_fn, __PROC_STATUS_LEN, "/proc/%d/status", pid);
- if (ret < 0 || ret >= __PROC_STATUS_LEN)
+ ret = snprintf(proc_fn, LXC_PROC_STATUS_LEN, "/proc/%d/status", pid);
+ if (ret < 0 || ret >= LXC_PROC_STATUS_LEN)
goto on_error;
proc_file = fopen(proc_fn, "r");
if (!proc_file) {
- SYSERROR("Could not open %s.", proc_fn);
+ SYSERROR("Could not open %s", proc_fn);
goto on_error;
}
info = calloc(1, sizeof(*info));
if (!info) {
- SYSERROR("Could not allocate memory.");
+ SYSERROR("Could not allocate memory");
fclose(proc_file);
return NULL;
}
found = false;
+
while (getline(&line, &line_bufsz, proc_file) != -1) {
ret = sscanf(line, "CapBnd: %llx", &info->capability_mask);
if (ret != EOF && ret == 1) {
fclose(proc_file);
if (!found) {
- SYSERROR("Could not read capability bounding set from %s.",
- proc_fn);
- errno = ENOENT;
+ ERROR("Could not read capability bounding set from %s",
+ proc_fn);
goto on_error;
}
for (i = 0; i < LXC_NS_MAX; i++) {
if (ctx->ns_fd[i] < 0)
continue;
+
close(ctx->ns_fd[i]);
ctx->ns_fd[i] = -EBADF;
}
lxc_proc_close_ns_fd(ctx);
free(ctx);
- ctx = NULL;
}
/**
static int in_same_namespace(pid_t pid1, pid_t pid2, const char *ns)
{
int ns_fd1 = -1, ns_fd2 = -1, ret = -1;
+ int saved_errno;
struct stat ns_st1, ns_st2;
ns_fd1 = lxc_preserve_ns(pid1, ns);
goto out;
/* processes are in the same namespace */
- ret = -EINVAL;
- if ((ns_st1.st_dev == ns_st2.st_dev ) && (ns_st1.st_ino == ns_st2.st_ino))
+ if ((ns_st1.st_dev == ns_st2.st_dev ) && (ns_st1.st_ino == ns_st2.st_ino)) {
+ ret = -EINVAL;
goto out;
+ }
/* processes are in different namespaces */
ret = ns_fd2;
ns_fd2 = -1;
out:
+ saved_errno = errno;
if (ns_fd1 >= 0)
close(ns_fd1);
+
if (ns_fd2 >= 0)
close(ns_fd2);
+ errno = saved_errno;
return ret;
}
ret = setns(ctx->ns_fd[i], ns_info[i].clone_flag);
if (ret < 0) {
SYSERROR("Failed to attach to %s namespace of %d",
- ns_info[i].proc_name, pid);
+ ns_info[i].proc_name, pid);
return -1;
}
ret = unshare(CLONE_NEWNS);
if (ret < 0) {
- SYSERROR("Failed to unshare mount namespace.");
+ SYSERROR("Failed to unshare mount namespace");
return -1;
}
if (detect_shared_rootfs()) {
if (mount(NULL, "/", NULL, MS_SLAVE | MS_REC, NULL)) {
- SYSERROR("Failed to make / rslave.");
+ SYSERROR("Failed to make / rslave");
ERROR("Continuing...");
}
}
/* Assume /proc is always mounted, so remount it. */
ret = umount2("/proc", MNT_DETACH);
if (ret < 0) {
- SYSERROR("Failed to unmount /proc.");
+ SYSERROR("Failed to unmount /proc");
return -1;
}
ret = mount("none", "/proc", "proc", 0, NULL);
if (ret < 0) {
- SYSERROR("Failed to remount /proc.");
+ SYSERROR("Failed to remount /proc");
return -1;
}
*/
ret = umount2("/sys", MNT_DETACH);
if (ret < 0 && errno != EINVAL) {
- SYSERROR("Failed to unmount /sys.");
+ SYSERROR("Failed to unmount /sys");
return -1;
} else if (ret == 0) {
/* Remount it. */
ret = mount("none", "/sys", "sysfs", 0, NULL);
if (ret < 0) {
- SYSERROR("Failed to remount /sys.");
+ SYSERROR("Failed to remount /sys");
return -1;
}
}
if (ctx->capability_mask & (1LL << cap))
continue;
- if (prctl(PR_CAPBSET_DROP, cap, 0, 0, 0)) {
+ if (prctl(PR_CAPBSET_DROP, prctl_arg(cap), prctl_arg(0),
+ prctl_arg(0), prctl_arg(0))) {
SYSERROR("Failed to drop capability %d", cap);
return -1;
}
+
TRACE("Dropped capability %d", cap);
}
if (!extra_keep_store[i]) {
while (i > 0)
free(extra_keep_store[--i]);
+
free(extra_keep_store);
return -1;
}
free(extra_keep_store);
}
- SYSERROR("Failed to clear environment");
+ ERROR("Failed to clear environment");
return -1;
}
if (extra_keep_store[i]) {
ret = setenv(extra_keep[i], extra_keep_store[i], 1);
if (ret < 0)
- WARN("%s - Failed to set environment variable", strerror(errno));
+ SYSWARN("Failed to set environment variable");
}
+
free(extra_keep_store[i]);
}
+
free(extra_keep_store);
}
if (!path_kept) {
ret = setenv("PATH", "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", 1);
if (ret < 0)
- WARN("%s - Failed to set environment variable", strerror(errno));
+ SYSWARN("Failed to set environment variable");
}
}
ret = putenv("container=lxc");
if (ret < 0) {
- WARN("%s - Failed to set environment variable", strerror(errno));
+ SYSWARN("Failed to set environment variable");
return -1;
}
if (extra_env) {
for (; *extra_env; extra_env++) {
char *p;
+
/* We just assume the user knows what they are doing, so
* we don't do any checks.
*/
ret = putenv(p);
if (ret < 0)
- WARN("%s - Failed to set environment variable", strerror(errno));
+ SYSWARN("Failed to set environment variable");
}
}
int fd, ret;
pid_t pid;
int pipes[2];
- char *result = NULL;
+ FILE *pipe_f;
+ bool found = false;
+ size_t line_bufsz = 0;
+ char *line = NULL, *result = NULL;
/* We need to fork off a process that runs the getent program, and we
* need to capture its output, so we use a pipe for that purpose.
*/
- ret = pipe(pipes);
+ ret = pipe2(pipes, O_CLOEXEC);
if (ret < 0)
return NULL;
return NULL;
}
- if (pid) {
- int status;
- FILE *pipe_f;
- int found = 0;
- size_t line_bufsz = 0;
- char *line = NULL;
-
- close(pipes[1]);
-
- pipe_f = fdopen(pipes[0], "r");
- while (getline(&line, &line_bufsz, pipe_f) != -1) {
- int i;
- long value;
- char *token;
- char *endptr = NULL, *saveptr = NULL;
-
- /* If we already found something, just continue to read
- * until the pipe doesn't deliver any more data, but
- * don't modify the existing data structure.
- */
- if (found)
- continue;
-
- /* Trim line on the right hand side. */
- for (i = strlen(line); i > 0 && (line[i - 1] == '\n' || line[i - 1] == '\r'); --i)
- line[i - 1] = '\0';
-
- /* Split into tokens: first: user name. */
- token = strtok_r(line, ":", &saveptr);
- if (!token)
- continue;
- /* next: dummy password field */
- token = strtok_r(NULL, ":", &saveptr);
- if (!token)
- continue;
- /* next: user id */
- token = strtok_r(NULL, ":", &saveptr);
- value = token ? strtol(token, &endptr, 10) : 0;
- if (!token || !endptr || *endptr || value == LONG_MIN || value == LONG_MAX)
- continue;
- /* dummy sanity check: user id matches */
- if ((uid_t) value != uid)
- continue;
- /* skip fields: gid, gecos, dir, go to next field 'shell' */
- for (i = 0; i < 4; i++) {
- token = strtok_r(NULL, ":", &saveptr);
- if (!token)
- break;
- }
- if (!token)
- continue;
- free(result);
- result = strdup(token);
-
- /* Sanity check that there are no fields after that. */
- token = strtok_r(NULL, ":", &saveptr);
- if (token)
- continue;
-
- found = 1;
- }
-
- free(line);
- fclose(pipe_f);
- again:
- if (waitpid(pid, &status, 0) < 0) {
- if (errno == EINTR)
- goto again;
- free(result);
- return NULL;
- }
-
- /* Some sanity checks. If anything even hinted at going wrong,
- * we can't be sure we have a valid result, so we assume we
- * don't.
- */
-
- if (!WIFEXITED(status)) {
- free(result);
- return NULL;
- }
-
- if (WEXITSTATUS(status) != 0) {
- free(result);
- return NULL;
- }
-
- if (!found) {
- free(result);
- return NULL;
- }
-
- return result;
- } else {
+ if (!pid) {
char uid_buf[32];
char *arguments[] = {
"getent",
close(pipes[0]);
/* We want to capture stdout. */
- dup2(pipes[1], 1);
+ ret = dup2(pipes[1], STDOUT_FILENO);
close(pipes[1]);
+ if (ret < 0)
+ _exit(EXIT_FAILURE);
/* Get rid of stdin/stderr, so we try to associate it with
* /dev/null.
*/
- fd = open("/dev/null", O_RDWR);
+ fd = open_devnull();
if (fd < 0) {
- close(0);
- close(2);
+ close(STDIN_FILENO);
+ close(STDERR_FILENO);
} else {
- dup2(fd, 0);
- dup2(fd, 2);
+ (void)dup3(fd, STDIN_FILENO, O_CLOEXEC);
+ (void)dup3(fd, STDOUT_FILENO, O_CLOEXEC);
close(fd);
}
/* Finish argument list. */
- ret = snprintf(uid_buf, sizeof(uid_buf), "%ld", (long) uid);
- if (ret <= 0)
- exit(-1);
+ ret = snprintf(uid_buf, sizeof(uid_buf), "%ld", (long)uid);
+ if (ret <= 0 || ret >= sizeof(uid_buf))
+ _exit(EXIT_FAILURE);
/* Try to run getent program. */
- (void) execvp("getent", arguments);
- exit(-1);
+ (void)execvp("getent", arguments);
+ _exit(EXIT_FAILURE);
+ }
+
+ close(pipes[1]);
+
+ pipe_f = fdopen(pipes[0], "r");
+ while (getline(&line, &line_bufsz, pipe_f) != -1) {
+ int i;
+ long value;
+ char *token;
+ char *endptr = NULL, *saveptr = NULL;
+
+ /* If we already found something, just continue to read
+ * until the pipe doesn't deliver any more data, but
+ * don't modify the existing data structure.
+ */
+ if (found)
+ continue;
+
+ if (!line)
+ continue;
+
+ /* Trim line on the right hand side. */
+ for (i = strlen(line); i > 0 && (line[i - 1] == '\n' || line[i - 1] == '\r'); --i)
+ line[i - 1] = '\0';
+
+ /* Split into tokens: first: user name. */
+ token = strtok_r(line, ":", &saveptr);
+ if (!token)
+ continue;
+
+ /* next: dummy password field */
+ token = strtok_r(NULL, ":", &saveptr);
+ if (!token)
+ continue;
+
+ /* next: user id */
+ token = strtok_r(NULL, ":", &saveptr);
+ value = token ? strtol(token, &endptr, 10) : 0;
+ if (!token || !endptr || *endptr || value == LONG_MIN ||
+ value == LONG_MAX)
+ continue;
+
+ /* dummy sanity check: user id matches */
+ if ((uid_t)value != uid)
+ continue;
+
+ /* skip fields: gid, gecos, dir, go to next field 'shell' */
+ for (i = 0; i < 4; i++) {
+ token = strtok_r(NULL, ":", &saveptr);
+ if (!token)
+ continue;
+ }
+
+ if (!token)
+ continue;
+
+ free(result);
+ result = strdup(token);
+
+ /* Sanity check that there are no fields after that. */
+ token = strtok_r(NULL, ":", &saveptr);
+ if (token)
+ continue;
+
+ found = true;
+ }
+
+ free(line);
+ fclose(pipe_f);
+
+ ret = wait_for_pid(pid);
+ if (ret < 0) {
+ free(result);
+ return NULL;
+ }
+
+ if (!found) {
+ free(result);
+ return NULL;
}
+
+ return result;
}
static void lxc_attach_get_init_uidgid(uid_t *init_uid, gid_t *init_gid)
{
FILE *proc_file;
- char proc_fn[__PROC_STATUS_LEN];
+ char proc_fn[LXC_PROC_STATUS_LEN];
int ret;
char *line = NULL;
size_t line_bufsz = 0;
uid_t uid = (uid_t)-1;
gid_t gid = (gid_t)-1;
- /* Read capabilities. */
- snprintf(proc_fn, __PROC_STATUS_LEN, "/proc/%d/status", 1);
+ ret = snprintf(proc_fn, LXC_PROC_STATUS_LEN, "/proc/%d/status", 1);
+ if (ret < 0 || ret >= LXC_PROC_STATUS_LEN)
+ return;
proc_file = fopen(proc_fn, "r");
if (!proc_file)
if (ret != EOF && ret == 1)
gid = (gid_t)value;
}
+
if (uid != (uid_t)-1 && gid != (gid_t)-1)
break;
}
/* Only override arguments if we found something. */
if (uid != (uid_t)-1)
*init_uid = uid;
+
if (gid != (gid_t)-1)
*init_gid = gid;
*/
}
-/* Help the optimizer along if it doesn't know that exit always exits. */
-#define rexit(c) \
- do { \
- int __c = (c); \
- _exit(__c); \
- return __c; \
- } while (0)
-
-/* Define default options if no options are supplied by the user. */
-static lxc_attach_options_t attach_static_default_options = LXC_ATTACH_OPTIONS_DEFAULT;
-
-static bool fetch_seccomp(struct lxc_container *c,
- lxc_attach_options_t *options)
+static bool fetch_seccomp(struct lxc_container *c, lxc_attach_options_t *options)
{
+ int ret;
+ bool bret;
char *path;
if (!(options->namespaces & CLONE_NEWNS) ||
}
/* Remove current setting. */
- if (!c->set_config_item(c, "lxc.seccomp", "") &&
- !c->set_config_item(c, "lxc.seccomp.profile", "")) {
+ if (!c->set_config_item(c, "lxc.seccomp.profile", "") &&
+ !c->set_config_item(c, "lxc.seccomp", ""))
return false;
- }
/* Fetch the current profile path over the cmd interface. */
path = c->get_running_config_item(c, "lxc.seccomp.profile");
if (!path) {
- INFO("Failed to get running config item for lxc.seccomp.profile");
+ INFO("Failed to retrieve lxc.seccomp.profile");
+
path = c->get_running_config_item(c, "lxc.seccomp");
- }
- if (!path) {
- INFO("Failed to get running config item for lxc.seccomp");
- return true;
+ if (!path) {
+ INFO("Failed to retrieve lxc.seccomp");
+ return true;
+ }
}
/* Copy the value into the new lxc_conf. */
- if (!c->set_config_item(c, "lxc.seccomp.profile", path)) {
- free(path);
- return false;
- }
+ bret = c->set_config_item(c, "lxc.seccomp.profile", path);
free(path);
+ if (!bret)
+ return false;
/* Attempt to parse the resulting config. */
- if (lxc_read_seccomp_config(c->lxc_conf) < 0) {
- ERROR("Error reading seccomp policy.");
+ ret = lxc_read_seccomp_config(c->lxc_conf);
+ if (ret < 0) {
+ ERROR("Failed to retrieve seccomp policy");
return false;
}
- INFO("Retrieved seccomp policy.");
+ INFO("Retrieved seccomp policy");
return true;
}
static bool no_new_privs(struct lxc_container *c, lxc_attach_options_t *options)
{
+ bool bret;
char *val;
/* Remove current setting. */
- if (!c->set_config_item(c, "lxc.no_new_privs", ""))
+ if (!c->set_config_item(c, "lxc.no_new_privs", "")) {
+ INFO("Failed to unset lxc.no_new_privs");
return false;
+ }
/* Retrieve currently active setting. */
val = c->get_running_config_item(c, "lxc.no_new_privs");
if (!val) {
- INFO("Failed to get running config item for lxc.no_new_privs.");
+ INFO("Failed to retrieve lxc.no_new_privs");
return false;
}
/* Set currently active setting. */
- if (!c->set_config_item(c, "lxc.no_new_privs", val)) {
- free(val);
- return false;
- }
+ bret = c->set_config_item(c, "lxc.no_new_privs", val);
free(val);
-
- return true;
+ return bret;
}
static signed long get_personality(const char *name, const char *lxcpath)
struct attach_clone_payload {
int ipc_socket;
- int pty_fd;
+ int terminal_slave_fd;
lxc_attach_options_t *options;
struct lxc_proc_context_info *init_ctx;
lxc_attach_exec_t exec_function;
static void lxc_put_attach_clone_payload(struct attach_clone_payload *p)
{
if (p->ipc_socket >= 0) {
- shutdown(p->ipc_socket, SHUT_RDWR);
close(p->ipc_socket);
p->ipc_socket = -EBADF;
}
- if (p->pty_fd >= 0) {
- close(p->pty_fd);
- p->pty_fd = -EBADF;
+ if (p->terminal_slave_fd >= 0) {
+ close(p->terminal_slave_fd);
+ p->terminal_slave_fd = -EBADF;
}
if (p->init_ctx) {
int fd, lsm_fd, ret;
uid_t new_uid;
gid_t new_gid;
+ uid_t ns_root_uid = 0;
+ gid_t ns_root_gid = 0;
lxc_attach_options_t* options = payload->options;
struct lxc_proc_context_info* init_ctx = payload->init_ctx;
bool needs_lsm = (options->namespaces & CLONE_NEWNS) &&
ret = lxc_attach_remount_sys_proc();
if (ret < 0)
goto on_error;
+
TRACE("Remounted \"/proc\" and \"/sys\"");
}
new_personality = init_ctx->personality;
else
new_personality = options->personality;
+
ret = personality(new_personality);
if (ret < 0)
goto on_error;
+
TRACE("Set new personality");
}
#endif
ret = lxc_attach_drop_privs(init_ctx);
if (ret < 0)
goto on_error;
+
TRACE("Dropped capabilities");
}
options->extra_keep_env);
if (ret < 0)
goto on_error;
+
TRACE("Set up environment");
/* This remark only affects fully unprivileged containers:
*/
if (needs_lsm) {
ret = lxc_abstract_unix_recv_fds(payload->ipc_socket, &lsm_fd, 1, NULL, 0);
- if (ret <= 0)
+ if (ret <= 0) {
+ if (ret < 0)
+ SYSERROR("Failed to receive lsm label fd");
+
goto on_error;
+ }
+
TRACE("Received LSM label file descriptor %d from parent", lsm_fd);
}
if (options->stdin_fd > 0 && isatty(options->stdin_fd)) {
- ret = lxc_make_controlling_pty(options->stdin_fd);
+ ret = lxc_make_controlling_terminal(options->stdin_fd);
if (ret < 0)
goto on_error;
}
- /* Set {u,g}id. */
- new_uid = 0;
- new_gid = 0;
- /* Ignore errors, we will fall back to root in that case (/proc was not
- * mounted etc.).
- */
- if (options->namespaces & CLONE_NEWUSER)
- lxc_attach_get_init_uidgid(&new_uid, &new_gid);
+ if (options->namespaces & CLONE_NEWUSER) {
+ /* Check whether nsuid 0 has a mapping. */
+ ns_root_uid = get_ns_uid(0);
- if (options->uid != (uid_t)-1)
- new_uid = options->uid;
- if (options->gid != (gid_t)-1)
- new_gid = options->gid;
+ /* Check whether nsgid 0 has a mapping. */
+ ns_root_gid = get_ns_gid(0);
- /* Try to set the {u,g}id combination. */
- if (new_uid != 0 || new_gid != 0 || options->namespaces & CLONE_NEWUSER) {
- ret = lxc_switch_uid_gid(new_uid, new_gid);
- if (ret < 0)
+ /* If there's no mapping for nsuid 0 try to retrieve the nsuid
+ * init was started with.
+ */
+ if (ns_root_uid == LXC_INVALID_UID)
+ lxc_attach_get_init_uidgid(&ns_root_uid, &ns_root_gid);
+
+ if (ns_root_uid == LXC_INVALID_UID)
goto on_error;
- ret = lxc_setgroups(0, NULL);
- if (ret < 0)
+ if (!lxc_switch_uid_gid(ns_root_uid, ns_root_gid))
goto on_error;
}
+ if (!lxc_setgroups(0, NULL) && errno != EPERM)
+ goto on_error;
+
+ /* Set {u,g}id. */
+ if (options->uid != LXC_INVALID_UID)
+ new_uid = options->uid;
+ else
+ new_uid = ns_root_uid;
+
+ if (options->gid != LXC_INVALID_GID)
+ new_gid = options->gid;
+ else
+ new_gid = ns_root_gid;
+
if ((init_ctx->container && init_ctx->container->lxc_conf &&
init_ctx->container->lxc_conf->no_new_privs) ||
(options->attach_flags & LXC_ATTACH_NO_NEW_PRIVS)) {
- ret = prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0);
+ ret = prctl(PR_SET_NO_NEW_PRIVS, prctl_arg(1), prctl_arg(0),
+ prctl_arg(0), prctl_arg(0));
if (ret < 0)
goto on_error;
+
TRACE("Set PR_SET_NO_NEW_PRIVS");
}
/* Change into our new LSM profile. */
on_exec = options->attach_flags & LXC_ATTACH_LSM_EXEC ? true : false;
+
ret = lsm_process_label_set_at(lsm_fd, init_ctx->lsm_label, on_exec);
close(lsm_fd);
if (ret < 0)
goto on_error;
+
TRACE("Set %s LSM label to \"%s\"", lsm_name(), init_ctx->lsm_label);
}
ret = lxc_seccomp_load(init_ctx->container->lxc_conf);
if (ret < 0)
goto on_error;
+
TRACE("Loaded seccomp profile");
}
- shutdown(payload->ipc_socket, SHUT_RDWR);
+
close(payload->ipc_socket);
payload->ipc_socket = -EBADF;
lxc_proc_put_context_info(init_ctx);
+ payload->init_ctx = NULL;
/* The following is done after the communication socket is shut down.
* That way, all errors that might (though unlikely) occur up until this
* may want to make sure the fds are closed, for example.
*/
if (options->stdin_fd >= 0 && options->stdin_fd != STDIN_FILENO)
- dup2(options->stdin_fd, STDIN_FILENO);
+ (void)dup2(options->stdin_fd, STDIN_FILENO);
if (options->stdout_fd >= 0 && options->stdout_fd != STDOUT_FILENO)
- dup2(options->stdout_fd, STDOUT_FILENO);
+ (void)dup2(options->stdout_fd, STDOUT_FILENO);
if (options->stderr_fd >= 0 && options->stderr_fd != STDERR_FILENO)
- dup2(options->stderr_fd, STDERR_FILENO);
+ (void)dup2(options->stderr_fd, STDERR_FILENO);
/* close the old fds */
if (options->stdin_fd > STDERR_FILENO)
* here, ignore errors.
*/
for (fd = STDIN_FILENO; fd <= STDERR_FILENO; fd++) {
- int flags;
-
- flags = fcntl(fd, F_GETFL);
- if (flags < 0)
- continue;
-
- if ((flags & FD_CLOEXEC) == 0)
- continue;
-
- ret = fcntl(fd, F_SETFL, flags & ~FD_CLOEXEC);
+ ret = fd_cloexec(fd, false);
if (ret < 0) {
SYSERROR("Failed to clear FD_CLOEXEC from file descriptor %d", fd);
goto on_error;
}
}
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY) {
- ret = lxc_login_pty(payload->pty_fd);
+ if (options->attach_flags & LXC_ATTACH_TERMINAL) {
+ ret = lxc_terminal_prepare_login(payload->terminal_slave_fd);
if (ret < 0) {
- SYSERROR("Failed to prepare pty file descriptor %d", payload->pty_fd);
+ SYSERROR("Failed to prepare terminal file descriptor %d", payload->terminal_slave_fd);
goto on_error;
}
- TRACE("Prepared pty file descriptor %d", payload->pty_fd);
+
+ TRACE("Prepared terminal file descriptor %d", payload->terminal_slave_fd);
}
+ /* Avoid unnecessary syscalls. */
+ if (new_uid == ns_root_uid)
+ new_uid = LXC_INVALID_UID;
+
+ if (new_gid == ns_root_gid)
+ new_gid = LXC_INVALID_GID;
+
+ if (!lxc_switch_uid_gid(new_uid, new_gid))
+ goto on_error;
+
/* We're done, so we can now do whatever the user intended us to do. */
- rexit(payload->exec_function(payload->exec_payload));
+ _exit(payload->exec_function(payload->exec_payload));
on_error:
lxc_put_attach_clone_payload(payload);
- rexit(EXIT_FAILURE);
+ _exit(EXIT_FAILURE);
}
-static int lxc_attach_pty(struct lxc_conf *conf, struct lxc_terminal *pty)
+static int lxc_attach_terminal(struct lxc_conf *conf,
+ struct lxc_terminal *terminal)
{
int ret;
- lxc_terminal_init(pty);
+ lxc_terminal_init(terminal);
- ret = lxc_terminal_create(pty);
+ ret = lxc_terminal_create(terminal);
if (ret < 0) {
- SYSERROR("Failed to create pty");
+ ERROR("Failed to create terminal");
return -1;
}
/* Shift ttys to container. */
- ret = lxc_terminal_map_ids(conf, pty);
+ ret = lxc_terminal_map_ids(conf, terminal);
if (ret < 0) {
- ERROR("Failed to shift pty");
+ ERROR("Failed to chown terminal");
goto on_error;
}
return 0;
on_error:
- lxc_terminal_delete(pty);
- lxc_terminal_conf_free(pty);
+ lxc_terminal_delete(terminal);
+ lxc_terminal_conf_free(terminal);
return -1;
}
-static int lxc_attach_pty_mainloop_init(struct lxc_terminal *pty,
- struct lxc_epoll_descr *descr)
+static int lxc_attach_terminal_mainloop_init(struct lxc_terminal *terminal,
+ struct lxc_epoll_descr *descr)
{
int ret;
return -1;
}
- ret = lxc_terminal_mainloop_add(descr, pty);
+ ret = lxc_terminal_mainloop_add(descr, terminal);
if (ret < 0) {
ERROR("Failed to add handlers to mainloop");
lxc_mainloop_close(descr);
return 0;
}
-static inline void lxc_attach_pty_close_master(struct lxc_terminal *pty)
+static inline void lxc_attach_terminal_close_master(struct lxc_terminal *terminal)
{
- if (pty->master < 0)
+ if (terminal->master < 0)
return;
- close(pty->master);
- pty->master = -EBADF;
+ close(terminal->master);
+ terminal->master = -EBADF;
}
-static inline void lxc_attach_pty_close_slave(struct lxc_terminal *pty)
+static inline void lxc_attach_terminal_close_slave(struct lxc_terminal *terminal)
{
- if (pty->slave < 0)
+ if (terminal->slave < 0)
return;
- close(pty->slave);
- pty->slave = -EBADF;
+ close(terminal->slave);
+ terminal->slave = -EBADF;
}
-static inline void lxc_attach_pty_close_peer(struct lxc_terminal *pty)
+static inline void lxc_attach_terminal_close_peer(struct lxc_terminal *terminal)
{
- if (pty->peer < 0)
+ if (terminal->peer < 0)
return;
- close(pty->peer);
- pty->peer = -EBADF;
+ close(terminal->peer);
+ terminal->peer = -EBADF;
}
-static inline void lxc_attach_pty_close_log(struct lxc_terminal *pty)
+static inline void lxc_attach_terminal_close_log(struct lxc_terminal *terminal)
{
- if (pty->log_fd < 0)
+ if (terminal->log_fd < 0)
return;
- close(pty->log_fd);
- pty->log_fd = -EBADF;
+ close(terminal->log_fd);
+ terminal->log_fd = -EBADF;
}
int lxc_attach(const char *name, const char *lxcpath,
signed long personality;
pid_t attached_pid, init_pid, pid;
struct lxc_proc_context_info *init_ctx;
- struct lxc_terminal pty;
+ struct lxc_terminal terminal;
struct lxc_conf *conf;
struct attach_clone_payload payload = {0};
ret = access("/proc/self/ns", X_OK);
if (ret) {
- ERROR("Does this kernel version support namespaces?");
+ SYSERROR("Does this kernel version support namespaces?");
return -1;
}
init_pid = lxc_cmd_get_init_pid(name, lxcpath);
if (init_pid < 0) {
- ERROR("Failed to get init pid.");
+ ERROR("Failed to get init pid");
return -1;
}
init_ctx->container->lxc_conf = lxc_conf_init();
if (!init_ctx->container->lxc_conf) {
lxc_proc_put_context_info(init_ctx);
- return -ENOMEM;
+ return -1;
}
}
conf = init_ctx->container->lxc_conf;
if (!fetch_seccomp(init_ctx->container, options))
- WARN("Failed to get seccomp policy.");
+ WARN("Failed to get seccomp policy");
if (!no_new_privs(init_ctx->container, options))
- WARN("Could not determine whether PR_SET_NO_NEW_PRIVS is set.");
+ WARN("Could not determine whether PR_SET_NO_NEW_PRIVS is set");
cwd = getcwd(NULL, 0);
}
pid = lxc_raw_getpid();
+
for (i = 0; i < LXC_NS_MAX; i++) {
- int j, saved_errno;
+ int j;
if (options->namespaces & ns_info[i].clone_flag)
init_ctx->ns_fd[i] = lxc_preserve_ns(init_pid, ns_info[i].proc_name);
init_ctx->ns_fd[i] = in_same_namespace(pid, init_pid, ns_info[i].proc_name);
else
continue;
+
if (init_ctx->ns_fd[i] >= 0)
continue;
}
/* We failed to preserve the namespace. */
- saved_errno = errno;
+ SYSERROR("Failed to attach to %s namespace of %d",
+ ns_info[i].proc_name, pid);
+
/* Close all already opened file descriptors before we return an
* error, so we don't leak them.
*/
for (j = 0; j < i; j++)
close(init_ctx->ns_fd[j]);
- errno = saved_errno;
- SYSERROR("Failed to attach to %s namespace of %d",
- ns_info[i].proc_name, pid);
free(cwd);
lxc_proc_put_context_info(init_ctx);
return -1;
}
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY) {
- ret = lxc_attach_pty(conf, &pty);
+ if (options->attach_flags & LXC_ATTACH_TERMINAL) {
+ ret = lxc_attach_terminal(conf, &terminal);
if (ret < 0) {
- ERROR("Failed to allocate pty");
+ ERROR("Failed to setup new terminal");
free(cwd);
lxc_proc_put_context_info(init_ctx);
return -1;
}
- pty.log_fd = options->log_fd;
+ terminal.log_fd = options->log_fd;
} else {
- lxc_terminal_init(&pty);
+ lxc_terminal_init(&terminal);
}
/* Create a socket pair for IPC communication; set SOCK_CLOEXEC in order
* just fork()s away without exec'ing directly after, the socket fd will
* exist in the forked process from the other thread and any close() in
* our own child process will not really cause the socket to close
- * properly, potentiall causing the parent to hang.
+ * properly, potentially causing the parent to hang.
*
* For this reason, while IPC is still active, we have to use shutdown()
* if the child exits prematurely in order to signal that the socket is
*/
ret = socketpair(PF_LOCAL, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets);
if (ret < 0) {
- SYSERROR("Could not set up required IPC mechanism for attaching.");
+ SYSERROR("Could not set up required IPC mechanism for attaching");
free(cwd);
lxc_proc_put_context_info(init_ctx);
return -1;
*/
pid = fork();
if (pid < 0) {
- SYSERROR("Failed to create first subprocess.");
+ SYSERROR("Failed to create first subprocess");
free(cwd);
lxc_proc_put_context_info(init_ctx);
return -1;
close(ipc_sockets[1]);
free(cwd);
lxc_proc_close_ns_fd(init_ctx);
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY)
- lxc_attach_pty_close_slave(&pty);
+ if (options->attach_flags & LXC_ATTACH_TERMINAL)
+ lxc_attach_terminal_close_slave(&terminal);
/* Attach to cgroup, if requested. */
if (options->attach_flags & LXC_ATTACH_MOVE_TO_CGROUP) {
- if (!cgroup_attach(name, lxcpath, pid))
+ struct cgroup_ops *cgroup_ops;
+
+ cgroup_ops = cgroup_init(conf);
+ if (!cgroup_ops)
goto on_error;
- TRACE("Moved intermediate process %d into container's "
- "cgroups", pid);
+
+ if (!cgroup_ops->attach(cgroup_ops, name, lxcpath, pid))
+ goto on_error;
+
+ cgroup_exit(cgroup_ops);
+ TRACE("Moved intermediate process %d into container's cgroups", pid);
}
/* Setup /proc limits */
goto on_error;
}
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY) {
- ret = lxc_attach_pty_mainloop_init(&pty, &descr);
+ if (options->attach_flags & LXC_ATTACH_TERMINAL) {
+ ret = lxc_attach_terminal_mainloop_init(&terminal, &descr);
if (ret < 0)
goto on_error;
- TRACE("Initalized pty mainloop");
+
+ TRACE("Initialized terminal mainloop");
}
/* Let the child process know to go ahead. */
ret = lxc_write_nointr(ipc_sockets[0], &status, sizeof(status));
if (ret != sizeof(status))
goto close_mainloop;
+
TRACE("Told intermediate process to start initializing");
/* Get pid of attached process from intermediate process. */
ret = lxc_read_nointr(ipc_sockets[0], &attached_pid, sizeof(attached_pid));
if (ret != sizeof(attached_pid))
goto close_mainloop;
+
TRACE("Received pid %d of attached process in parent pid namespace", attached_pid);
/* Ignore SIGKILL (CTRL-C) and SIGQUIT (CTRL-\) - issue #313. */
ret = wait_for_pid(pid);
if (ret < 0)
goto close_mainloop;
+
TRACE("Intermediate process %d exited", pid);
/* We will always have to reap the attached process now. */
labelfd = lsm_process_label_fd_get(attached_pid, on_exec);
if (labelfd < 0)
goto close_mainloop;
+
TRACE("Opened LSM label file descriptor %d", labelfd);
/* Send child fd of the LSM security module to write to. */
ret = lxc_abstract_unix_send_fds(ipc_sockets[0], &labelfd, 1, NULL, 0);
- close(labelfd);
if (ret <= 0) {
- SYSERROR("%d", (int)ret);
+ if (ret < 0)
+ SYSERROR("Failed to send lsm label fd");
+
+ close(labelfd);
goto close_mainloop;
}
+
+ close(labelfd);
TRACE("Sent LSM label file descriptor %d to child", labelfd);
}
ret_parent = 0;
to_cleanup_pid = -1;
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY) {
+
+ if (options->attach_flags & LXC_ATTACH_TERMINAL) {
ret = lxc_mainloop(&descr, -1);
if (ret < 0) {
ret_parent = -1;
}
close_mainloop:
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY)
+ if (options->attach_flags & LXC_ATTACH_TERMINAL)
lxc_mainloop_close(&descr);
on_error:
if (to_cleanup_pid > 0)
(void)wait_for_pid(to_cleanup_pid);
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY) {
- lxc_terminal_delete(&pty);
- lxc_terminal_conf_free(&pty);
+ if (options->attach_flags & LXC_ATTACH_TERMINAL) {
+ lxc_terminal_delete(&terminal);
+ lxc_terminal_conf_free(&terminal);
}
+
lxc_proc_put_context_info(init_ctx);
return ret_parent;
}
/* close unneeded file descriptors */
close(ipc_sockets[0]);
ipc_sockets[0] = -EBADF;
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY) {
- lxc_attach_pty_close_master(&pty);
- lxc_attach_pty_close_peer(&pty);
- lxc_attach_pty_close_log(&pty);
+
+ if (options->attach_flags & LXC_ATTACH_TERMINAL) {
+ lxc_attach_terminal_close_master(&terminal);
+ lxc_attach_terminal_close_peer(&terminal);
+ lxc_attach_terminal_close_log(&terminal);
}
/* Wait for the parent to have setup cgroups. */
if (ret != sizeof(status)) {
shutdown(ipc_sockets[1], SHUT_RDWR);
lxc_proc_put_context_info(init_ctx);
- rexit(-1);
+ _exit(EXIT_FAILURE);
}
+
TRACE("Intermediate process starting to initialize");
/* Attach now, create another subprocess later, since pid namespaces
ERROR("Failed to enter namespaces");
shutdown(ipc_sockets[1], SHUT_RDWR);
lxc_proc_put_context_info(init_ctx);
- rexit(-1);
+ _exit(EXIT_FAILURE);
}
+
/* close namespace file descriptors */
lxc_proc_close_ns_fd(init_ctx);
payload.ipc_socket = ipc_sockets[1];
payload.options = options;
payload.init_ctx = init_ctx;
- payload.pty_fd = pty.slave;
+ payload.terminal_slave_fd = terminal.slave;
payload.exec_function = exec_function;
payload.exec_payload = exec_payload;
SYSERROR("Failed to clone attached process");
shutdown(ipc_sockets[1], SHUT_RDWR);
lxc_proc_put_context_info(init_ctx);
- rexit(-1);
+ _exit(EXIT_FAILURE);
}
if (pid == 0) {
+ if (options->attach_flags & LXC_ATTACH_TERMINAL) {
+ ret = pthread_sigmask(SIG_SETMASK,
+ &terminal.tty_state->oldmask, NULL);
+ if (ret < 0) {
+ SYSERROR("Failed to reset signal mask");
+ _exit(EXIT_FAILURE);
+ }
+ }
+
ret = attach_child_main(&payload);
if (ret < 0)
ERROR("Failed to exec");
+
_exit(EXIT_FAILURE);
}
- if (options->attach_flags & LXC_ATTACH_ALLOCATE_PTY)
- lxc_attach_pty_close_slave(&pty);
+
+ if (options->attach_flags & LXC_ATTACH_TERMINAL)
+ lxc_attach_terminal_close_slave(&terminal);
/* Tell grandparent the pid of the pid of the newly created child. */
ret = lxc_write_nointr(ipc_sockets[1], &pid, sizeof(pid));
*/
shutdown(ipc_sockets[1], SHUT_RDWR);
lxc_proc_put_context_info(init_ctx);
- rexit(-1);
+ _exit(EXIT_FAILURE);
}
+
TRACE("Sending pid %d of attached process", pid);
/* The rest is in the hands of the initial and the attached process. */
lxc_proc_put_context_info(init_ctx);
- rexit(0);
+ _exit(0);
}
-int lxc_attach_run_command(void* payload)
+int lxc_attach_run_command(void *payload)
{
- lxc_attach_command_t* cmd = (lxc_attach_command_t*)payload;
+ int ret = -1;
+ lxc_attach_command_t *cmd = payload;
- execvp(cmd->program, cmd->argv);
- SYSERROR("Failed to exec \"%s\".", cmd->program);
- return -1;
+ ret = execvp(cmd->program, cmd->argv);
+ if (ret < 0) {
+ switch (errno) {
+ case ENOEXEC:
+ ret = 126;
+ break;
+ case ENOENT:
+ ret = 127;
+ break;
+ }
+ }
+
+ SYSERROR("Failed to exec \"%s\"", cmd->program);
+ return ret;
}
int lxc_attach_run_shell(void* payload)
{
uid_t uid;
- struct passwd *passwd;
+ struct passwd pwent;
+ struct passwd *pwentp = NULL;
char *user_shell;
+ char *buf;
+ size_t bufsize;
+ int ret;
/* Ignore payload parameter. */
(void)payload;
uid = getuid();
- passwd = getpwuid(uid);
+
+ bufsize = sysconf(_SC_GETPW_R_SIZE_MAX);
+ if (bufsize == -1)
+ bufsize = 1024;
+
+ buf = malloc(bufsize);
+ if (buf) {
+ ret = getpwuid_r(uid, &pwent, buf, bufsize, &pwentp);
+ if (!pwentp) {
+ if (ret == 0)
+ WARN("Could not find matched password record");
+
+ WARN("Failed to get password record - %u", uid);
+ }
+ }
/* This probably happens because of incompatible nss implementations in
* host and container (remember, this code is still using the host's
* the information by spawning a [getent passwd uid] process and parsing
* the result.
*/
- if (!passwd)
+ if (!pwentp)
user_shell = lxc_attach_getpwshell(uid);
else
- user_shell = passwd->pw_shell;
+ user_shell = pwent.pw_shell;
+
if (user_shell)
execlp(user_shell, user_shell, (char *)NULL);
* on /bin/sh as a default shell.
*/
execlp("/bin/sh", "/bin/sh", (char *)NULL);
+
SYSERROR("Failed to execute shell");
- if (!passwd)
+ if (!pwentp)
free(user_shell);
+
+ free(buf);
return -1;
}