-/*-*- Mode: C; c-basic-offset: 8 -*-*/
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
/***
This file is part of systemd.
#include <string.h>
#include <libgen.h>
#include <assert.h>
+#include <unistd.h>
+#include <ftw.h>
#include "mount-setup.h"
#include "log.h"
#include "macro.h"
#include "util.h"
+#include "label.h"
+#include "set.h"
+#include "strv.h"
+
+#ifndef TTY_GID
+#define TTY_GID 5
+#endif
typedef struct MountPoint {
const char *what;
bool fatal;
} MountPoint;
+/* The first three entries we might need before SELinux is up. The
+ * other ones we can delay until SELinux is loaded. */
+#define N_EARLY_MOUNT 3
+
static const MountPoint mount_table[] = {
- { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
- { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
- { "devtmps", "/dev", "devtmpfs", "mode=755", MS_NOSUID, true },
- { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
- { "devpts", "/dev/pts", "devpts", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
- { "cgroup", "/cgroup/debug", "cgroup", "debug", MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
- { "debugfs", "/sys/kernel/debug", "debugfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
- { "binfmt_misc", "/proc/sys/fs/binfmt_misc", "binfmt_misc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
- { "mqueue", "/dev/mqueue", "mqueue", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
+ { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
+ { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
+ { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID, true },
+ { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV, true },
+ { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, false },
+ { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV, true },
+ { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
+ { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV, false },
+};
+
+/* These are API file systems that might be mounted by other software,
+ * we just list them here so that we know that we should ignore them */
+
+static const char * const ignore_paths[] = {
+ "/sys/fs/selinux",
+ "/selinux",
+ "/proc/bus/usb"
};
bool mount_point_is_api(const char *path) {
* should be ignored */
for (i = 0; i < ELEMENTSOF(mount_table); i ++)
- if (path_startswith(path, mount_table[i].where))
+ if (path_equal(path, mount_table[i].where))
+ return true;
+
+ return path_startswith(path, "/sys/fs/cgroup/");
+}
+
+bool mount_point_ignore(const char *path) {
+ unsigned i;
+
+ for (i = 0; i < ELEMENTSOF(ignore_paths); i++)
+ if (path_equal(path, ignore_paths[i]))
return true;
- return path_startswith(path, "/cgroup/");
+ return false;
}
-static int mount_one(const MountPoint *p) {
+static int mount_one(const MountPoint *p, bool relabel) {
int r;
assert(p);
- if ((r = path_is_mount_point(p->where)) < 0)
+ /* Relabel first, just in case */
+ if (relabel)
+ label_fix(p->where, true);
+
+ if ((r = path_is_mount_point(p->where, true)) < 0)
return r;
if (r > 0)
return p->fatal ? -errno : 0;
}
- return 0;
+ /* Relabel again, since we now mounted something fresh here */
+ if (relabel)
+ label_fix(p->where, false);
+
+ return 1;
}
-static int mount_cgroup_controllers(void) {
+int mount_setup_early(void) {
+ unsigned i;
+ int r = 0;
+
+ assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table));
+
+ /* Do a minimal mount of /proc and friends to enable the most
+ * basic stuff, such as SELinux */
+ for (i = 0; i < N_EARLY_MOUNT; i ++) {
+ int j;
+
+ j = mount_one(mount_table + i, false);
+ if (r == 0)
+ r = j;
+ }
+
+ return r;
+}
+
+int mount_cgroup_controllers(char ***join_controllers) {
int r;
FILE *f;
- char buf [256];
+ char buf[LINE_MAX];
+ Set *controllers;
- /* Mount all available cgroup controllers. */
+ /* Mount all available cgroup controllers that are built into the kernel. */
+
+ f = fopen("/proc/cgroups", "re");
+ if (!f) {
+ log_error("Failed to enumerate cgroup controllers: %m");
+ return 0;
+ }
- if (!(f = fopen("/proc/cgroups", "re")))
- return -ENOENT;
+ controllers = set_new(string_hash_func, string_compare_func);
+ if (!controllers) {
+ r = -ENOMEM;
+ log_error("Failed to allocate controller set.");
+ goto finish;
+ }
/* Ignore the header line */
(void) fgets(buf, sizeof(buf), f);
for (;;) {
- MountPoint p;
- char *controller, *where;
+ char *controller;
+ int enabled = 0;
- if (fscanf(f, "%ms %*i %*i %*i", &controller) != 1) {
+ if (fscanf(f, "%ms %*i %*i %i", &controller, &enabled) != 2) {
if (feof(f))
break;
goto finish;
}
- if (asprintf(&where, "/cgroup/%s", controller) < 0) {
+ if (!enabled) {
free(controller);
+ continue;
+ }
+
+ r = set_put(controllers, controller);
+ if (r < 0) {
+ log_error("Failed to add controller to set.");
+ free(controller);
+ goto finish;
+ }
+ }
+
+ for (;;) {
+ MountPoint p;
+ char *controller, *where, *options;
+ char ***k = NULL;
+
+ controller = set_steal_first(controllers);
+ if (!controller)
+ break;
+
+ if (join_controllers)
+ for (k = join_controllers; *k; k++)
+ if (strv_find(*k, controller))
+ break;
+
+ if (k && *k) {
+ char **i, **j;
+
+ for (i = *k, j = *k; *i; i++) {
+
+ if (!streq(*i, controller)) {
+ char *t;
+
+ t = set_remove(controllers, *i);
+ if (!t) {
+ free(*i);
+ continue;
+ }
+ free(t);
+ }
+
+ *(j++) = *i;
+ }
+
+ *j = NULL;
+
+ options = strv_join(*k, ",");
+ if (!options) {
+ log_error("Failed to join options");
+ free(controller);
+ r = -ENOMEM;
+ goto finish;
+ }
+
+ } else {
+ options = controller;
+ controller = NULL;
+ }
+
+ where = strappend("/sys/fs/cgroup/", options);
+ if (!where) {
+ log_error("Failed to build path");
+ free(options);
r = -ENOMEM;
goto finish;
}
p.what = "cgroup";
p.where = where;
p.type = "cgroup";
- p.options = controller;
+ p.options = options;
p.flags = MS_NOSUID|MS_NOEXEC|MS_NODEV;
p.fatal = false;
- r = mount_one(&p);
+ r = mount_one(&p, true);
free(controller);
free(where);
- if (r < 0)
+ if (r < 0) {
+ free(options);
goto finish;
+ }
+
+ if (r > 0 && k && *k) {
+ char **i;
+
+ for (i = *k; *i; i++) {
+ char *t;
+
+ t = strappend("/sys/fs/cgroup/", *i);
+ if (!t) {
+ log_error("Failed to build path");
+ r = -ENOMEM;
+ free(options);
+ goto finish;
+ }
+
+ r = symlink(options, t);
+ free(t);
+
+ if (r < 0 && errno != EEXIST) {
+ log_error("Failed to create symlink: %m");
+ r = -errno;
+ free(options);
+ goto finish;
+ }
+ }
+ }
+
+ free(options);
}
r = 0;
finish:
+ set_free_free(controllers);
+
fclose(f);
return r;
}
-int mount_setup(void) {
+static int symlink_and_label(const char *old_path, const char *new_path) {
+ int r;
+
+ assert(old_path);
+ assert(new_path);
+
+ if ((r = label_symlinkfile_set(new_path)) < 0)
+ return r;
+
+ if (symlink(old_path, new_path) < 0)
+ r = -errno;
+
+ label_file_clear();
+
+ return r;
+}
+
+static int nftw_cb(
+ const char *fpath,
+ const struct stat *sb,
+ int tflag,
+ struct FTW *ftwbuf) {
+
+ /* No need to label /dev twice in a row... */
+ if (ftwbuf->level == 0)
+ return 0;
+
+ label_fix(fpath, true);
+ return 0;
+};
+
+int mount_setup(bool loaded_policy) {
+
+ const char symlinks[] =
+ "/proc/kcore\0" "/dev/core\0"
+ "/proc/self/fd\0" "/dev/fd\0"
+ "/proc/self/fd/0\0" "/dev/stdin\0"
+ "/proc/self/fd/1\0" "/dev/stdout\0"
+ "/proc/self/fd/2\0" "/dev/stderr\0";
+
int r;
unsigned i;
+ const char *j, *k;
- for (i = 0; i < ELEMENTSOF(mount_table); i ++)
- if ((r = mount_one(mount_table+i)) < 0)
+ for (i = 0; i < ELEMENTSOF(mount_table); i ++) {
+ r = mount_one(mount_table + i, true);
+
+ if (r < 0)
return r;
+ }
+
+ /* Nodes in devtmpfs and /run need to be manually updated for
+ * the appropriate labels, after mounting. The other virtual
+ * API file systems like /sys and /proc do not need that, they
+ * use the same label for all their files. */
+ if (loaded_policy) {
+ usec_t before_relabel, after_relabel;
+ char timespan[FORMAT_TIMESPAN_MAX];
- return mount_cgroup_controllers();
+ before_relabel = now(CLOCK_MONOTONIC);
+
+ nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS);
+ nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS);
+
+ after_relabel = now(CLOCK_MONOTONIC);
+
+ log_info("Relabelled /dev and /run in %s.",
+ format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel));
+
+ }
+
+ /* Create a few default symlinks, which are normally created
+ * by udevd, but some scripts might need them before we start
+ * udevd. */
+ NULSTR_FOREACH_PAIR(j, k, symlinks)
+ symlink_and_label(j, k);
+
+ /* Create a few directories we always want around */
+ mkdir("/run/systemd", 0755);
+ mkdir("/run/systemd/system", 0755);
+
+ return 0;
}
~ianmdlvl / elogind.git / blobdiff