Google Git
Sign in
third-party-mirror / systemd / f229149d865bb087742add2415d0c7c321294591 / . / src / core / mount.c
blob: 5aa7da4433970b90414cfd931689111030acf32d [file] [log] [blame]
/* SPDX-License-Identifier: LGPL-2.1-or-later */
#include <errno.h>
#include <signal.h>
#include <stdio.h>
#include <sys/epoll.h>
#include "sd-messages.h"
#include "alloc-util.h"
#include "dbus-mount.h"
#include "dbus-unit.h"
#include "device.h"
#include "exit-status.h"
#include "format-util.h"
#include "fs-util.h"
#include "fstab-util.h"
#include "initrd-util.h"
#include "libmount-util.h"
#include "log.h"
#include "manager.h"
#include "mkdir-label.h"
#include "mount-setup.h"
#include "mount.h"
#include "mountpoint-util.h"
#include "parse-util.h"
#include "path-util.h"
#include "process-util.h"
#include "serialize.h"
#include "special.h"
#include "stat-util.h"
#include "string-table.h"
#include "string-util.h"
#include "strv.h"
#include "unit-name.h"
#include "unit.h"
#define RETRY_UMOUNT_MAX 32
static const UnitActiveState state_translation_table[_MOUNT_STATE_MAX] = {
[MOUNT_DEAD] = UNIT_INACTIVE,
[MOUNT_MOUNTING] = UNIT_ACTIVATING,
[MOUNT_MOUNTING_DONE] = UNIT_ACTIVATING,
[MOUNT_MOUNTED] = UNIT_ACTIVE,
[MOUNT_REMOUNTING] = UNIT_RELOADING,
[MOUNT_UNMOUNTING] = UNIT_DEACTIVATING,
[MOUNT_REMOUNTING_SIGTERM] = UNIT_RELOADING,
[MOUNT_REMOUNTING_SIGKILL] = UNIT_RELOADING,
[MOUNT_UNMOUNTING_SIGTERM] = UNIT_DEACTIVATING,
[MOUNT_UNMOUNTING_SIGKILL] = UNIT_DEACTIVATING,
[MOUNT_FAILED] = UNIT_FAILED,
[MOUNT_CLEANING] = UNIT_MAINTENANCE,
};
static int mount_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata);
static int mount_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata);
static void mount_enter_dead(Mount *m, MountResult f);
static void mount_enter_mounted(Mount *m, MountResult f);
static void mount_cycle_clear(Mount *m);
static int mount_process_proc_self_mountinfo(Manager *m);
static bool MOUNT_STATE_WITH_PROCESS(MountState state) {
return IN_SET(state,
MOUNT_MOUNTING,
MOUNT_MOUNTING_DONE,
MOUNT_REMOUNTING,
MOUNT_REMOUNTING_SIGTERM,
MOUNT_REMOUNTING_SIGKILL,
MOUNT_UNMOUNTING,
MOUNT_UNMOUNTING_SIGTERM,
MOUNT_UNMOUNTING_SIGKILL,
MOUNT_CLEANING);
}
static MountParameters* get_mount_parameters_fragment(Mount *m) {
assert(m);
if (m->from_fragment)
return &m->parameters_fragment;
return NULL;
}
static MountParameters* get_mount_parameters(Mount *m) {
assert(m);
if (m->from_proc_self_mountinfo)
return &m->parameters_proc_self_mountinfo;
return get_mount_parameters_fragment(m);
}
static bool mount_is_network(const MountParameters *p) {
assert(p);
if (fstab_test_option(p->options, "_netdev\0"))
return true;
if (p->fstype && fstype_is_network(p->fstype))
return true;
return false;
}
static bool mount_is_nofail(const Mount *m) {
assert(m);
if (!m->from_fragment)
return false;
return fstab_test_yes_no_option(m->parameters_fragment.options, "nofail\0" "fail\0");
}
static bool mount_is_loop(const MountParameters *p) {
assert(p);
if (fstab_test_option(p->options, "loop\0"))
return true;
return false;
}
static bool mount_is_bind(const MountParameters *p) {
assert(p);
if (fstab_test_option(p->options, "bind\0" "rbind\0"))
return true;
if (p->fstype && STR_IN_SET(p->fstype, "bind", "rbind"))
return true;
return false;
}
static bool mount_is_bound_to_device(Mount *m) {
const MountParameters *p;
assert(m);
/* Determines whether to place a Requires= or BindsTo= dependency on the backing device unit. We do
* this by checking for the x-systemd.device-bound mount option. Iff it is set we use BindsTo=,
* otherwise Requires=. But note that we might combine the latter with StopPropagatedFrom=, see
* below. */
p = get_mount_parameters(m);
if (!p)
return false;
return fstab_test_option(p->options, "x-systemd.device-bound\0");
}
static bool mount_propagate_stop(Mount *m) {
assert(m);
if (mount_is_bound_to_device(m)) /* If we are using BindsTo= the stop propagation is implicit, no need to bother */
return false;
return m->from_fragment; /* let's propagate stop whenever this is an explicitly configured unit,
* otherwise let's not bother. */
}
static bool mount_needs_quota(const MountParameters *p) {
assert(p);
if (p->fstype && !fstype_needs_quota(p->fstype))
return false;
if (mount_is_bind(p))
return false;
return fstab_test_option(p->options,
"usrquota\0" "grpquota\0" "quota\0" "usrjquota\0" "grpjquota\0");
}
static void mount_init(Unit *u) {
Mount *m = MOUNT(u);
assert(m);
assert(u);
assert(u->load_state == UNIT_STUB);
m->timeout_usec = u->manager->default_timeout_start_usec;
m->exec_context.std_output = u->manager->default_std_output;
m->exec_context.std_error = u->manager->default_std_error;
m->directory_mode = 0755;
/* We need to make sure that /usr/bin/mount is always called
* in the same process group as us, so that the autofs kernel
* side doesn't send us another mount request while we are
* already trying to comply its last one. */
m->exec_context.same_pgrp = true;
m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID;
u->ignore_on_isolate = true;
}
static int mount_arm_timer(Mount *m, usec_t usec) {
int r;
assert(m);
if (m->timer_event_source) {
r = sd_event_source_set_time(m->timer_event_source, usec);
if (r < 0)
return r;
return sd_event_source_set_enabled(m->timer_event_source, SD_EVENT_ONESHOT);
}
if (usec == USEC_INFINITY)
return 0;
r = sd_event_add_time(
UNIT(m)->manager->event,
&m->timer_event_source,
CLOCK_MONOTONIC,
usec, 0,
mount_dispatch_timer, m);
if (r < 0)
return r;
(void) sd_event_source_set_description(m->timer_event_source, "mount-timer");
return 0;
}
static void mount_unwatch_control_pid(Mount *m) {
assert(m);
if (m->control_pid <= 0)
return;
unit_unwatch_pid(UNIT(m), TAKE_PID(m->control_pid));
}
static void mount_parameters_done(MountParameters *p) {
assert(p);
p->what = mfree(p->what);
p->options = mfree(p->options);
p->fstype = mfree(p->fstype);
}
static void mount_done(Unit *u) {
Mount *m = MOUNT(u);
assert(m);
m->where = mfree(m->where);
mount_parameters_done(&m->parameters_proc_self_mountinfo);
mount_parameters_done(&m->parameters_fragment);
m->exec_runtime = exec_runtime_unref(m->exec_runtime, false);
exec_command_done_array(m->exec_command, _MOUNT_EXEC_COMMAND_MAX);
m->control_command = NULL;
dynamic_creds_unref(&m->dynamic_creds);
mount_unwatch_control_pid(m);
m->timer_event_source = sd_event_source_disable_unref(m->timer_event_source);
}
static int update_parameters_proc_self_mountinfo(
Mount *m,
const char *what,
const char *options,
const char *fstype) {
MountParameters *p;
int r, q, w;
p = &m->parameters_proc_self_mountinfo;
r = free_and_strdup(&p->what, what);
if (r < 0)
return r;
q = free_and_strdup(&p->options, options);
if (q < 0)
return q;
w = free_and_strdup(&p->fstype, fstype);
if (w < 0)
return w;
return r > 0 || q > 0 || w > 0;
}
static int mount_add_mount_dependencies(Mount *m) {
MountParameters *pm;
Unit *other;
Set *s;
int r;
assert(m);
if (!path_equal(m->where, "/")) {
_cleanup_free_ char *parent = NULL;
/* Adds in links to other mount points that might lie further up in the hierarchy */
r = path_extract_directory(m->where, &parent);
if (r < 0)
return r;
r = unit_require_mounts_for(UNIT(m), parent, UNIT_DEPENDENCY_IMPLICIT);
if (r < 0)
return r;
}
/* Adds in dependencies to other mount points that might be needed for the source path (if this is a bind mount
* or a loop mount) to be available. */
pm = get_mount_parameters_fragment(m);
if (pm && pm->what &&
path_is_absolute(pm->what) &&
(mount_is_bind(pm) || mount_is_loop(pm) || !mount_is_network(pm))) {
r = unit_require_mounts_for(UNIT(m), pm->what, UNIT_DEPENDENCY_FILE);
if (r < 0)
return r;
}
/* Adds in dependencies to other units that use this path or paths further down in the hierarchy */
s = manager_get_units_requiring_mounts_for(UNIT(m)->manager, m->where);
SET_FOREACH(other, s) {
if (other->load_state != UNIT_LOADED)
continue;
if (other == UNIT(m))
continue;
r = unit_add_dependency(other, UNIT_AFTER, UNIT(m), true, UNIT_DEPENDENCY_PATH);
if (r < 0)
return r;
if (UNIT(m)->fragment_path) {
/* If we have fragment configuration, then make this dependency required */
r = unit_add_dependency(other, UNIT_REQUIRES, UNIT(m), true, UNIT_DEPENDENCY_PATH);
if (r < 0)
return r;
}
}
return 0;
}
static int mount_add_device_dependencies(Mount *m) {
UnitDependencyMask mask;
MountParameters *p;
UnitDependency dep;
int r;
assert(m);
log_unit_trace(UNIT(m), "Processing implicit device dependencies");
p = get_mount_parameters(m);
if (!p) {
log_unit_trace(UNIT(m), "Missing mount parameters, skipping implicit device dependencies");
return 0;
}
if (!p->what) {
log_unit_trace(UNIT(m), "Missing mount source, skipping implicit device dependencies");
return 0;
}
if (mount_is_bind(p)) {
log_unit_trace(UNIT(m), "Mount unit is a bind mount, skipping implicit device dependencies");
return 0;
}
if (!is_device_path(p->what)) {
log_unit_trace(UNIT(m), "Mount source is not a device path, skipping implicit device dependencies");
return 0;
}
/* /dev/root is a really weird thing, it's not a real device, but just a path the kernel exports for
* the root file system specified on the kernel command line. Ignore it here. */
if (PATH_IN_SET(p->what, "/dev/root", "/dev/nfs")) {
log_unit_trace(UNIT(m), "Mount source is in /dev/root or /dev/nfs, skipping implicit device dependencies");
return 0;
}
if (path_equal(m->where, "/")) {
log_unit_trace(UNIT(m), "Mount destination is '/', skipping implicit device dependencies");
return 0;
}
/* Mount units from /proc/self/mountinfo are not bound to devices by default since they're subject to
* races when mounts are established by other tools with different backing devices than what we
* maintain. The user can still force this to be a BindsTo= dependency with an appropriate option (or
* udev property) so the mount units are automatically stopped when the device disappears
* suddenly. */
dep = mount_is_bound_to_device(m) ? UNIT_BINDS_TO : UNIT_REQUIRES;
/* We always use 'what' from /proc/self/mountinfo if mounted */
mask = m->from_proc_self_mountinfo ? UNIT_DEPENDENCY_MOUNTINFO : UNIT_DEPENDENCY_MOUNT_FILE;
r = unit_add_node_dependency(UNIT(m), p->what, dep, mask);
if (r < 0)
return r;
if (r > 0)
log_unit_trace(UNIT(m), "Added %s dependency on %s", unit_dependency_to_string(dep), p->what);
if (mount_propagate_stop(m)) {
r = unit_add_node_dependency(UNIT(m), p->what, UNIT_STOP_PROPAGATED_FROM, mask);
if (r < 0)
return r;
if (r > 0)
log_unit_trace(UNIT(m), "Added %s dependency on %s",
unit_dependency_to_string(UNIT_STOP_PROPAGATED_FROM), p->what);
}
r = unit_add_blockdev_dependency(UNIT(m), p->what, mask);
if (r > 0)
log_unit_trace(UNIT(m), "Added %s dependency on %s", unit_dependency_to_string(UNIT_AFTER), p->what);
return 0;
}
static int mount_add_quota_dependencies(Mount *m) {
MountParameters *p;
int r;
assert(m);
if (!MANAGER_IS_SYSTEM(UNIT(m)->manager))
return 0;
p = get_mount_parameters_fragment(m);
if (!p)
return 0;
if (!mount_needs_quota(p))
return 0;
r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_WANTS, SPECIAL_QUOTACHECK_SERVICE,
/* add_reference= */ true, UNIT_DEPENDENCY_FILE);
if (r < 0)
return r;
r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_WANTS, SPECIAL_QUOTAON_SERVICE,
/* add_reference= */true, UNIT_DEPENDENCY_FILE);
if (r < 0)
return r;
return 0;
}
static bool mount_is_extrinsic(Unit *u) {
MountParameters *p;
Mount *m = MOUNT(u);
assert(m);
/* Returns true for all units that are "magic" and should be excluded from the usual
* start-up and shutdown dependencies. We call them "extrinsic" here, as they are generally
* mounted outside of the systemd dependency logic. We shouldn't attempt to manage them
* ourselves but it's fine if the user operates on them with us. */
/* We only automatically manage mounts if we are in system mode */
if (MANAGER_IS_USER(u->manager))
return true;
p = get_mount_parameters(m);
if (p && fstab_is_extrinsic(m->where, p->options))
return true;
return false;
}
static int mount_add_default_ordering_dependencies(Mount *m, MountParameters *p, UnitDependencyMask mask) {
const char *after, *before, *e;
int r;
assert(m);
e = path_startswith(m->where, "/sysroot");
if (e && in_initrd()) {
/* All mounts under /sysroot need to happen later, at initrd-fs.target time. IOW,
* it's not technically part of the basic initrd filesystem itself, and so
* shouldn't inherit the default Before=local-fs.target dependency. However,
* these mounts still need to start after local-fs-pre.target, as a sync point
* for things like systemd-hibernate-resume@.service that should start before
* any mounts. */
after = SPECIAL_LOCAL_FS_PRE_TARGET;
before = isempty(e) ? SPECIAL_INITRD_ROOT_FS_TARGET : SPECIAL_INITRD_FS_TARGET;
} else if (in_initrd() && path_startswith(m->where, "/sysusr/usr")) {
after = SPECIAL_LOCAL_FS_PRE_TARGET;
before = SPECIAL_INITRD_USR_FS_TARGET;
} else if (mount_is_network(p)) {
after = SPECIAL_REMOTE_FS_PRE_TARGET;
before = SPECIAL_REMOTE_FS_TARGET;
} else {
after = SPECIAL_LOCAL_FS_PRE_TARGET;
before = SPECIAL_LOCAL_FS_TARGET;
}
if (!mount_is_nofail(m)) {
r = unit_add_dependency_by_name(UNIT(m), UNIT_BEFORE, before, /* add_reference= */ true, mask);
if (r < 0)
return r;
}
r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, after, /* add_reference= */ true, mask);
if (r < 0)
return r;
return unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET,
/* add_reference= */ true, mask);
}
static int mount_add_default_dependencies(Mount *m) {
UnitDependencyMask mask;
MountParameters *p;
int r;
assert(m);
if (!UNIT(m)->default_dependencies)
return 0;
/* We do not add any default dependencies to /, /usr or /run/initramfs/, since they are
* guaranteed to stay mounted the whole time, since our system is on it. Also, don't
* bother with anything mounted below virtual file systems, it's also going to be virtual,
* and hence not worth the effort. */
if (mount_is_extrinsic(UNIT(m)))
return 0;
p = get_mount_parameters(m);
if (!p)
return 0;
mask = m->from_proc_self_mountinfo ? UNIT_DEPENDENCY_MOUNTINFO : UNIT_DEPENDENCY_MOUNT_FILE;
r = mount_add_default_ordering_dependencies(m, p, mask);
if (r < 0)
return r;
if (mount_is_network(p)) {
/* We order ourselves after network.target. This is primarily useful at shutdown:
* services that take down the network should order themselves before
* network.target, so that they are shut down only after this mount unit is
* stopped. */
r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, SPECIAL_NETWORK_TARGET,
/* add_reference= */ true, mask);
if (r < 0)
return r;
/* We pull in network-online.target, and order ourselves after it. This is useful
* at start-up to actively pull in tools that want to be started before we start
* mounting network file systems, and whose purpose it is to delay this until the
* network is "up". */
r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_WANTS, UNIT_AFTER, SPECIAL_NETWORK_ONLINE_TARGET,
/* add_reference= */ true, mask);
if (r < 0)
return r;
}
/* If this is a tmpfs mount then we have to unmount it before we try to deactivate swaps */
if (streq_ptr(p->fstype, "tmpfs")) {
r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, SPECIAL_SWAP_TARGET,
/* add_reference= */ true, mask);
if (r < 0)
return r;
}
return 0;
}
static int mount_verify(Mount *m) {
_cleanup_free_ char *e = NULL;
MountParameters *p;
int r;
assert(m);
assert(UNIT(m)->load_state == UNIT_LOADED);
if (!m->from_fragment && !m->from_proc_self_mountinfo && !UNIT(m)->perpetual)
return -ENOENT;
r = unit_name_from_path(m->where, ".mount", &e);
if (r < 0)
return log_unit_error_errno(UNIT(m), r, "Failed to generate unit name from mount path: %m");
if (!unit_has_name(UNIT(m), e))
return log_unit_error_errno(UNIT(m), SYNTHETIC_ERRNO(ENOEXEC), "Where= setting doesn't match unit name. Refusing.");
if (mount_point_is_api(m->where) || mount_point_ignore(m->where))
return log_unit_error_errno(UNIT(m), SYNTHETIC_ERRNO(ENOEXEC), "Cannot create mount unit for API file system %s. Refusing.", m->where);
p = get_mount_parameters_fragment(m);
if (p && !p->what && !UNIT(m)->perpetual)
return log_unit_error_errno(UNIT(m), SYNTHETIC_ERRNO(ENOEXEC),
"What= setting is missing. Refusing.");
if (m->exec_context.pam_name && m->kill_context.kill_mode != KILL_CONTROL_GROUP)
return log_unit_error_errno(UNIT(m), SYNTHETIC_ERRNO(ENOEXEC), "Unit has PAM enabled. Kill mode must be set to control-group'. Refusing.");
return 0;
}
static int mount_add_non_exec_dependencies(Mount *m) {
int r;
assert(m);
/* We may be called due to this mount appearing in /proc/self/mountinfo, hence we clear all existing
* dependencies that were initialized from the unit file but whose final value really depends on the
* content of /proc/self/mountinfo. Some (such as m->where) might have become stale now. */
unit_remove_dependencies(UNIT(m), UNIT_DEPENDENCY_MOUNTINFO | UNIT_DEPENDENCY_MOUNT_FILE);
if (!m->where)
return 0;
/* Adds in all dependencies directly responsible for ordering the mount, as opposed to dependencies
* resulting from the ExecContext and such. */
r = mount_add_device_dependencies(m);
if (r < 0)
return r;
r = mount_add_mount_dependencies(m);
if (r < 0)
return r;
r = mount_add_quota_dependencies(m);
if (r < 0)
return r;
r = mount_add_default_dependencies(m);
if (r < 0)
return r;
return 0;
}
static int mount_add_extras(Mount *m) {
Unit *u = UNIT(m);
int r;
assert(m);
/* Note: this call might be called after we already have been loaded once (and even when it has already been
* activated), in case data from /proc/self/mountinfo has changed. This means all code here needs to be ready
* to run with an already set up unit. */
if (u->fragment_path)
m->from_fragment = true;
if (!m->where) {
r = unit_name_to_path(u->id, &m->where);
if (r == -ENAMETOOLONG)
log_unit_error_errno(u, r, "Failed to derive mount point path from unit name, because unit name is hashed. "
"Set \"Where=\" in the unit file explicitly.");
if (r < 0)
return r;
}
path_simplify(m->where);
if (!u->description) {
r = unit_set_description(u, m->where);
if (r < 0)
return r;
}
r = unit_patch_contexts(u);
if (r < 0)
return r;
r = unit_add_exec_dependencies(u, &m->exec_context);
if (r < 0)
return r;
r = unit_set_default_slice(u);
if (r < 0)
return r;
r = mount_add_non_exec_dependencies(m);
if (r < 0)
return r;
return 0;
}
static void mount_load_root_mount(Unit *u) {
assert(u);
if (!unit_has_name(u, SPECIAL_ROOT_MOUNT))
return;
u->perpetual = true;
u->default_dependencies = false;
/* The stdio/kmsg bridge socket is on /, in order to avoid a dep loop, don't use kmsg logging for -.mount */
MOUNT(u)->exec_context.std_output = EXEC_OUTPUT_NULL;
MOUNT(u)->exec_context.std_input = EXEC_INPUT_NULL;
if (!u->description)
u->description = strdup("Root Mount");
}
static int mount_load(Unit *u) {
Mount *m = MOUNT(u);
int r, q = 0;
assert(m);
assert(u);
assert(u->load_state == UNIT_STUB);
mount_load_root_mount(u);
bool fragment_optional = m->from_proc_self_mountinfo || u->perpetual;
r = unit_load_fragment_and_dropin(u, !fragment_optional);
/* Add in some extras. Note we do this in all cases (even if we failed to load the unit) when announced by the
* kernel, because we need some things to be set up no matter what when the kernel establishes a mount and thus
* we need to update the state in our unit to track it. After all, consider that we don't allow changing the
* 'slice' field for a unit once it is active. */
if (u->load_state == UNIT_LOADED || m->from_proc_self_mountinfo || u->perpetual)
q = mount_add_extras(m);
if (r < 0)
return r;
if (q < 0)
return q;
if (u->load_state != UNIT_LOADED)
return 0;
return mount_verify(m);
}
static void mount_set_state(Mount *m, MountState state) {
MountState old_state;
assert(m);
if (m->state != state)
bus_unit_send_pending_change_signal(UNIT(m), false);
old_state = m->state;
m->state = state;
if (!MOUNT_STATE_WITH_PROCESS(state)) {
m->timer_event_source = sd_event_source_disable_unref(m->timer_event_source);
mount_unwatch_control_pid(m);
m->control_command = NULL;
m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID;
}
if (state != old_state)
log_unit_debug(UNIT(m), "Changed %s -> %s", mount_state_to_string(old_state), mount_state_to_string(state));
unit_notify(UNIT(m), state_translation_table[old_state], state_translation_table[state],
m->reload_result == MOUNT_SUCCESS ? 0 : UNIT_NOTIFY_RELOAD_FAILURE);
}
static int mount_coldplug(Unit *u) {
Mount *m = MOUNT(u);
int r;
assert(m);
assert(m->state == MOUNT_DEAD);
if (m->deserialized_state == m->state)
return 0;
if (m->control_pid > 0 &&
pid_is_unwaited(m->control_pid) &&
MOUNT_STATE_WITH_PROCESS(m->deserialized_state)) {
r = unit_watch_pid(UNIT(m), m->control_pid, false);
if (r < 0)
return r;
r = mount_arm_timer(m, usec_add(u->state_change_timestamp.monotonic, m->timeout_usec));
if (r < 0)
return r;
}
if (!IN_SET(m->deserialized_state, MOUNT_DEAD, MOUNT_FAILED)) {
(void) unit_setup_dynamic_creds(u);
(void) unit_setup_exec_runtime(u);
}
mount_set_state(m, m->deserialized_state);
return 0;
}
static void mount_catchup(Unit *u) {
Mount *m = MOUNT(ASSERT_PTR(u));
assert(m);
/* Adjust the deserialized state. See comments in mount_process_proc_self_mountinfo(). */
if (m->from_proc_self_mountinfo)
switch (m->state) {
case MOUNT_DEAD:
case MOUNT_FAILED:
assert(m->control_pid == 0);
(void) unit_acquire_invocation_id(u);
mount_cycle_clear(m);
mount_enter_mounted(m, MOUNT_SUCCESS);
break;
case MOUNT_MOUNTING:
assert(m->control_pid > 0);
mount_set_state(m, MOUNT_MOUNTING_DONE);
break;
default:
break;
}
else
switch (m->state) {
case MOUNT_MOUNTING_DONE:
assert(m->control_pid > 0);
mount_set_state(m, MOUNT_MOUNTING);
break;
case MOUNT_MOUNTED:
assert(m->control_pid == 0);
mount_enter_dead(m, MOUNT_SUCCESS);
break;
default:
break;
}
}
static void mount_dump(Unit *u, FILE *f, const char *prefix) {
Mount *m = MOUNT(u);
MountParameters *p;
assert(m);
assert(f);
p = get_mount_parameters(m);
fprintf(f,
"%sMount State: %s\n"
"%sResult: %s\n"
"%sClean Result: %s\n"
"%sWhere: %s\n"
"%sWhat: %s\n"
"%sFile System Type: %s\n"
"%sOptions: %s\n"
"%sFrom /proc/self/mountinfo: %s\n"
"%sFrom fragment: %s\n"
"%sExtrinsic: %s\n"
"%sDirectoryMode: %04o\n"
"%sSloppyOptions: %s\n"
"%sLazyUnmount: %s\n"
"%sForceUnmount: %s\n"
"%sReadWriteOnly: %s\n"
"%sTimeoutSec: %s\n",
prefix, mount_state_to_string(m->state),
prefix, mount_result_to_string(m->result),
prefix, mount_result_to_string(m->clean_result),
prefix, m->where,
prefix, p ? strna(p->what) : "n/a",
prefix, p ? strna(p->fstype) : "n/a",
prefix, p ? strna(p->options) : "n/a",
prefix, yes_no(m->from_proc_self_mountinfo),
prefix, yes_no(m->from_fragment),
prefix, yes_no(mount_is_extrinsic(u)),
prefix, m->directory_mode,
prefix, yes_no(m->sloppy_options),
prefix, yes_no(m->lazy_unmount),
prefix, yes_no(m->force_unmount),
prefix, yes_no(m->read_write_only),
prefix, FORMAT_TIMESPAN(m->timeout_usec, USEC_PER_SEC));
if (m->control_pid > 0)
fprintf(f,
"%sControl PID: "PID_FMT"\n",
prefix, m->control_pid);
exec_context_dump(&m->exec_context, f, prefix);
kill_context_dump(&m->kill_context, f, prefix);
cgroup_context_dump(UNIT(m), f, prefix);
}
static int mount_spawn(Mount *m, ExecCommand *c, pid_t *_pid) {
_cleanup_(exec_params_clear) ExecParameters exec_params = {
.flags = EXEC_APPLY_SANDBOXING|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN,
.stdin_fd = -EBADF,
.stdout_fd = -EBADF,
.stderr_fd = -EBADF,
.exec_fd = -EBADF,
};
pid_t pid;
int r;
assert(m);
assert(c);
assert(_pid);
r = unit_prepare_exec(UNIT(m));
if (r < 0)
return r;
r = mount_arm_timer(m, usec_add(now(CLOCK_MONOTONIC), m->timeout_usec));
if (r < 0)
return r;
r = unit_set_exec_params(UNIT(m), &exec_params);
if (r < 0)
return r;
r = exec_spawn(UNIT(m),
c,
&m->exec_context,
&exec_params,
m->exec_runtime,
&m->dynamic_creds,
&pid);
if (r < 0)
return r;
r = unit_watch_pid(UNIT(m), pid, true);
if (r < 0)
return r;
*_pid = pid;
return 0;
}
static void mount_enter_dead(Mount *m, MountResult f) {
assert(m);
if (m->result == MOUNT_SUCCESS)
m->result = f;
unit_log_result(UNIT(m), m->result == MOUNT_SUCCESS, mount_result_to_string(m->result));
unit_warn_leftover_processes(UNIT(m), unit_log_leftover_process_stop);
mount_set_state(m, m->result != MOUNT_SUCCESS ? MOUNT_FAILED : MOUNT_DEAD);
m->exec_runtime = exec_runtime_unref(m->exec_runtime, true);
unit_destroy_runtime_data(UNIT(m), &m->exec_context);
unit_unref_uid_gid(UNIT(m), true);
dynamic_creds_destroy(&m->dynamic_creds);
/* Any dependencies based on /proc/self/mountinfo are now stale. Let's re-generate dependencies from
* .mount unit. */
(void) mount_add_non_exec_dependencies(m);
}
static void mount_enter_mounted(Mount *m, MountResult f) {
assert(m);
if (m->result == MOUNT_SUCCESS)
m->result = f;
mount_set_state(m, MOUNT_MOUNTED);
}
static void mount_enter_dead_or_mounted(Mount *m, MountResult f) {
assert(m);
/* Enter DEAD or MOUNTED state, depending on what the kernel currently says about the mount point. We use this
* whenever we executed an operation, so that our internal state reflects what the kernel says again, after all
* ultimately we just mirror the kernel's internal state on this. */
if (m->from_proc_self_mountinfo)
mount_enter_mounted(m, f);
else
mount_enter_dead(m, f);
}
static int state_to_kill_operation(MountState state) {
switch (state) {
case MOUNT_REMOUNTING_SIGTERM:
return KILL_RESTART;
case MOUNT_UNMOUNTING_SIGTERM:
return KILL_TERMINATE;
case MOUNT_REMOUNTING_SIGKILL:
case MOUNT_UNMOUNTING_SIGKILL:
return KILL_KILL;
default:
return _KILL_OPERATION_INVALID;
}
}
static void mount_enter_signal(Mount *m, MountState state, MountResult f) {
int r;
assert(m);
if (m->result == MOUNT_SUCCESS)
m->result = f;
r = unit_kill_context(
UNIT(m),
&m->kill_context,
state_to_kill_operation(state),
-1,
m->control_pid,
false);
if (r < 0)
goto fail;
if (r > 0) {
r = mount_arm_timer(m, usec_add(now(CLOCK_MONOTONIC), m->timeout_usec));
if (r < 0)
goto fail;
mount_set_state(m, state);
} else if (state == MOUNT_REMOUNTING_SIGTERM && m->kill_context.send_sigkill)
mount_enter_signal(m, MOUNT_REMOUNTING_SIGKILL, MOUNT_SUCCESS);
else if (IN_SET(state, MOUNT_REMOUNTING_SIGTERM, MOUNT_REMOUNTING_SIGKILL))
mount_enter_mounted(m, MOUNT_SUCCESS);
else if (state == MOUNT_UNMOUNTING_SIGTERM && m->kill_context.send_sigkill)
mount_enter_signal(m, MOUNT_UNMOUNTING_SIGKILL, MOUNT_SUCCESS);
else
mount_enter_dead_or_mounted(m, MOUNT_SUCCESS);
return;
fail:
log_unit_warning_errno(UNIT(m), r, "Failed to kill processes: %m");
mount_enter_dead_or_mounted(m, MOUNT_FAILURE_RESOURCES);
}
static void mount_enter_unmounting(Mount *m) {
int r;
assert(m);
/* Start counting our attempts */
if (!IN_SET(m->state,
MOUNT_UNMOUNTING,
MOUNT_UNMOUNTING_SIGTERM,
MOUNT_UNMOUNTING_SIGKILL))
m->n_retry_umount = 0;
m->control_command_id = MOUNT_EXEC_UNMOUNT;
m->control_command = m->exec_command + MOUNT_EXEC_UNMOUNT;
r = exec_command_set(m->control_command, UMOUNT_PATH, m->where, "-c", NULL);
if (r >= 0 && m->lazy_unmount)
r = exec_command_append(m->control_command, "-l", NULL);
if (r >= 0 && m->force_unmount)
r = exec_command_append(m->control_command, "-f", NULL);
if (r < 0)
goto fail;
mount_unwatch_control_pid(m);
r = mount_spawn(m, m->control_command, &m->control_pid);
if (r < 0)
goto fail;
mount_set_state(m, MOUNT_UNMOUNTING);
return;
fail:
log_unit_warning_errno(UNIT(m), r, "Failed to run 'umount' task: %m");
mount_enter_dead_or_mounted(m, MOUNT_FAILURE_RESOURCES);
}
static void mount_enter_mounting(Mount *m) {
int r;
MountParameters *p;
bool source_is_dir = true;
assert(m);
r = unit_fail_if_noncanonical(UNIT(m), m->where);
if (r < 0)
goto fail;
p = get_mount_parameters_fragment(m);
if (p && mount_is_bind(p)) {
r = is_dir(p->what, /* follow = */ true);
if (r < 0 && r != -ENOENT)
log_unit_info_errno(UNIT(m), r, "Failed to determine type of bind mount source '%s', ignoring: %m", p->what);
else if (r == 0)
source_is_dir = false;
}
if (source_is_dir)
(void) mkdir_p_label(m->where, m->directory_mode);
else
(void) touch_file(m->where, /* parents = */ true, USEC_INFINITY, UID_INVALID, GID_INVALID, MODE_INVALID);
if (source_is_dir)
unit_warn_if_dir_nonempty(UNIT(m), m->where);
unit_warn_leftover_processes(UNIT(m), unit_log_leftover_process_start);
m->control_command_id = MOUNT_EXEC_MOUNT;
m->control_command = m->exec_command + MOUNT_EXEC_MOUNT;
/* Create the source directory for bind-mounts if needed */
if (p && mount_is_bind(p)) {
r = mkdir_p_label(p->what, m->directory_mode);
/* mkdir_p_label() can return -EEXIST if the target path exists and is not a directory - which is
* totally OK, in case the user wants us to overmount a non-directory inode. Also -EROFS can be
* returned on read-only filesystem. Moreover, -EACCES (and also maybe -EPERM?) may be returned
* when the path is on NFS. See issue #24120. All such errors will be logged in the debug level. */
if (r < 0 && r != -EEXIST)
log_unit_full_errno(UNIT(m),
(r == -EROFS || ERRNO_IS_PRIVILEGE(r)) ? LOG_DEBUG : LOG_WARNING,
r, "Failed to make bind mount source '%s', ignoring: %m", p->what);
}
if (p) {
_cleanup_free_ char *opts = NULL;
r = fstab_filter_options(p->options, "nofail\0" "noauto\0" "auto\0", NULL, NULL, NULL, &opts);
if (r < 0)
goto fail;
r = exec_command_set(m->control_command, MOUNT_PATH, p->what, m->where, NULL);
if (r >= 0 && m->sloppy_options)
r = exec_command_append(m->control_command, "-s", NULL);
if (r >= 0 && m->read_write_only)
r = exec_command_append(m->control_command, "-w", NULL);
if (r >= 0 && p->fstype)
r = exec_command_append(m->control_command, "-t", p->fstype, NULL);
if (r >= 0 && !isempty(opts))
r = exec_command_append(m->control_command, "-o", opts, NULL);
} else
r = -ENOENT;
if (r < 0)
goto fail;
mount_unwatch_control_pid(m);
r = mount_spawn(m, m->control_command, &m->control_pid);
if (r < 0)
goto fail;
mount_set_state(m, MOUNT_MOUNTING);
return;
fail:
log_unit_warning_errno(UNIT(m), r, "Failed to run 'mount' task: %m");
mount_enter_dead_or_mounted(m, MOUNT_FAILURE_RESOURCES);
}
static void mount_set_reload_result(Mount *m, MountResult result) {
assert(m);
/* Only store the first error we encounter */
if (m->reload_result != MOUNT_SUCCESS)
return;
m->reload_result = result;
}
static void mount_enter_remounting(Mount *m) {
int r;
MountParameters *p;
assert(m);
/* Reset reload result when we are about to start a new remount operation */
m->reload_result = MOUNT_SUCCESS;
m->control_command_id = MOUNT_EXEC_REMOUNT;
m->control_command = m->exec_command + MOUNT_EXEC_REMOUNT;
p = get_mount_parameters_fragment(m);
if (p) {
const char *o;
if (p->options)
o = strjoina("remount,", p->options);
else
o = "remount";
r = exec_command_set(m->control_command, MOUNT_PATH,
p->what, m->where,
"-o", o, NULL);
if (r >= 0 && m->sloppy_options)
r = exec_command_append(m->control_command, "-s", NULL);
if (r >= 0 && m->read_write_only)
r = exec_command_append(m->control_command, "-w", NULL);
if (r >= 0 && p->fstype)
r = exec_command_append(m->control_command, "-t", p->fstype, NULL);
} else
r = -ENOENT;
if (r < 0)
goto fail;
mount_unwatch_control_pid(m);
r = mount_spawn(m, m->control_command, &m->control_pid);
if (r < 0)
goto fail;
mount_set_state(m, MOUNT_REMOUNTING);
return;
fail:
log_unit_warning_errno(UNIT(m), r, "Failed to run 'remount' task: %m");
mount_set_reload_result(m, MOUNT_FAILURE_RESOURCES);
mount_enter_dead_or_mounted(m, MOUNT_SUCCESS);
}
static void mount_cycle_clear(Mount *m) {
assert(m);
/* Clear all state we shall forget for this new cycle */
m->result = MOUNT_SUCCESS;
m->reload_result = MOUNT_SUCCESS;
exec_command_reset_status_array(m->exec_command, _MOUNT_EXEC_COMMAND_MAX);
UNIT(m)->reset_accounting = true;
}
static int mount_start(Unit *u) {
Mount *m = MOUNT(u);
int r;
assert(m);
/* We cannot fulfill this request right now, try again later
* please! */
if (IN_SET(m->state,
MOUNT_UNMOUNTING,
MOUNT_UNMOUNTING_SIGTERM,
MOUNT_UNMOUNTING_SIGKILL,
MOUNT_CLEANING))
return -EAGAIN;
/* Already on it! */
if (IN_SET(m->state, MOUNT_MOUNTING, MOUNT_MOUNTING_DONE))
return 0;
assert(IN_SET(m->state, MOUNT_DEAD, MOUNT_FAILED));
r = unit_acquire_invocation_id(u);
if (r < 0)
return r;
mount_cycle_clear(m);
mount_enter_mounting(m);
return 1;
}
static int mount_stop(Unit *u) {
Mount *m = MOUNT(u);
assert(m);
switch (m->state) {
case MOUNT_UNMOUNTING:
case MOUNT_UNMOUNTING_SIGKILL:
case MOUNT_UNMOUNTING_SIGTERM:
/* Already on it */
return 0;
case MOUNT_MOUNTING:
case MOUNT_MOUNTING_DONE:
case MOUNT_REMOUNTING:
/* If we are still waiting for /bin/mount, we go directly into kill mode. */
mount_enter_signal(m, MOUNT_UNMOUNTING_SIGTERM, MOUNT_SUCCESS);
return 0;
case MOUNT_REMOUNTING_SIGTERM:
/* If we are already waiting for a hung remount, convert this to the matching unmounting state */
mount_set_state(m, MOUNT_UNMOUNTING_SIGTERM);
return 0;
case MOUNT_REMOUNTING_SIGKILL:
/* as above */
mount_set_state(m, MOUNT_UNMOUNTING_SIGKILL);
return 0;
case MOUNT_MOUNTED:
mount_enter_unmounting(m);
return 1;
case MOUNT_CLEANING:
/* If we are currently cleaning, then abort it, brutally. */
mount_enter_signal(m, MOUNT_UNMOUNTING_SIGKILL, MOUNT_SUCCESS);
return 0;
default:
assert_not_reached();
}
}
static int mount_reload(Unit *u) {
Mount *m = MOUNT(u);
assert(m);
assert(m->state == MOUNT_MOUNTED);
mount_enter_remounting(m);
return 1;
}
static int mount_serialize(Unit *u, FILE *f, FDSet *fds) {
Mount *m = MOUNT(u);
assert(m);
assert(f);
assert(fds);
(void) serialize_item(f, "state", mount_state_to_string(m->state));
(void) serialize_item(f, "result", mount_result_to_string(m->result));
(void) serialize_item(f, "reload-result", mount_result_to_string(m->reload_result));
(void) serialize_item_format(f, "n-retry-umount", "%u", m->n_retry_umount);
if (m->control_pid > 0)
(void) serialize_item_format(f, "control-pid", PID_FMT, m->control_pid);
if (m->control_command_id >= 0)
(void) serialize_item(f, "control-command", mount_exec_command_to_string(m->control_command_id));
return 0;
}
static int mount_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) {
Mount *m = MOUNT(u);
int r;
assert(m);
assert(u);
assert(key);
assert(value);
assert(fds);
if (streq(key, "state")) {
MountState state;
state = mount_state_from_string(value);
if (state < 0)
log_unit_debug_errno(u, state, "Failed to parse state value: %s", value);
else
m->deserialized_state = state;
} else if (streq(key, "result")) {
MountResult f;
f = mount_result_from_string(value);
if (f < 0)
log_unit_debug_errno(u, f, "Failed to parse result value: %s", value);
else if (f != MOUNT_SUCCESS)
m->result = f;
} else if (streq(key, "reload-result")) {
MountResult f;
f = mount_result_from_string(value);
if (f < 0)
log_unit_debug_errno(u, f, "Failed to parse reload result value: %s", value);
else if (f != MOUNT_SUCCESS)
m->reload_result = f;
} else if (streq(key, "n-retry-umount")) {
r = safe_atou(value, &m->n_retry_umount);
if (r < 0)
log_unit_debug_errno(u, r, "Failed to parse n-retry-umount value: %s", value);
} else if (streq(key, "control-pid")) {
r = parse_pid(value, &m->control_pid);
if (r < 0)
log_unit_debug_errno(u, r, "Failed to parse control-pid value: %s", value);
} else if (streq(key, "control-command")) {
MountExecCommand id;
id = mount_exec_command_from_string(value);
if (id < 0)
log_unit_debug_errno(u, id, "Failed to parse exec-command value: %s", value);
else {
m->control_command_id = id;
m->control_command = m->exec_command + id;
}
} else
log_unit_debug(u, "Unknown serialization key: %s", key);
return 0;
}
_pure_ static UnitActiveState mount_active_state(Unit *u) {
assert(u);
return state_translation_table[MOUNT(u)->state];
}
_pure_ static const char *mount_sub_state_to_string(Unit *u) {
assert(u);
return mount_state_to_string(MOUNT(u)->state);
}
_pure_ static bool mount_may_gc(Unit *u) {
Mount *m = MOUNT(u);
assert(m);
if (m->from_proc_self_mountinfo)
return false;
return true;
}
static void mount_sigchld_event(Unit *u, pid_t pid, int code, int status) {
Mount *m = MOUNT(u);
MountResult f;
assert(m);
assert(pid >= 0);
if (pid != m->control_pid)
return;
/* So here's the thing, we really want to know before /usr/bin/mount or /usr/bin/umount exit whether
* they established/remove a mount. This is important when mounting, but even more so when unmounting
* since we need to deal with nested mounts and otherwise cannot safely determine whether to repeat
* the unmounts. In theory, the kernel fires /proc/self/mountinfo changes off before returning from
* the mount() or umount() syscalls, and thus we should see the changes to the proc file before we
* process the waitid() for the /usr/bin/(u)mount processes. However, this is unfortunately racy: we
* have to waitid() for processes using P_ALL (since we need to reap unexpected children that got
* reparented to PID 1), but when using P_ALL we might end up reaping processes that terminated just
* instants ago, i.e. already after our last event loop iteration (i.e. after the last point we might
* have noticed /proc/self/mountinfo events via epoll). This means event loop priorities for
* processing SIGCHLD vs. /proc/self/mountinfo IO events are not as relevant as we want. To fix that
* race, let's explicitly scan /proc/self/mountinfo before we start processing /usr/bin/(u)mount
* dying. It's ugly, but it makes our ordering systematic again, and makes sure we always see
* /proc/self/mountinfo changes before our mount/umount exits. */
(void) mount_process_proc_self_mountinfo(u->manager);
m->control_pid = 0;
if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL))
f = MOUNT_SUCCESS;
else if (code == CLD_EXITED)
f = MOUNT_FAILURE_EXIT_CODE;
else if (code == CLD_KILLED)
f = MOUNT_FAILURE_SIGNAL;
else if (code == CLD_DUMPED)
f = MOUNT_FAILURE_CORE_DUMP;
else
assert_not_reached();
if (IN_SET(m->state, MOUNT_REMOUNTING, MOUNT_REMOUNTING_SIGKILL, MOUNT_REMOUNTING_SIGTERM))
mount_set_reload_result(m, f);
else if (m->result == MOUNT_SUCCESS)
m->result = f;
if (m->control_command) {
exec_status_exit(&m->control_command->exec_status, &m->exec_context, pid, code, status);
m->control_command = NULL;
m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID;
}
unit_log_process_exit(
u,
"Mount process",
mount_exec_command_to_string(m->control_command_id),
f == MOUNT_SUCCESS,
code, status);
/* Note that due to the io event priority logic, we can be sure the new mountinfo is loaded
* before we process the SIGCHLD for the mount command. */
switch (m->state) {
case MOUNT_MOUNTING:
/* Our mount point has not appeared in mountinfo. Something went wrong. */
if (f == MOUNT_SUCCESS) {
/* Either /bin/mount has an unexpected definition of success,
* or someone raced us and we lost. */
log_unit_warning(UNIT(m), "Mount process finished, but there is no mount.");
f = MOUNT_FAILURE_PROTOCOL;
}
mount_enter_dead(m, f);
break;
case MOUNT_MOUNTING_DONE:
mount_enter_mounted(m, f);
break;
case MOUNT_REMOUNTING:
case MOUNT_REMOUNTING_SIGTERM:
case MOUNT_REMOUNTING_SIGKILL:
mount_enter_dead_or_mounted(m, MOUNT_SUCCESS);
break;
case MOUNT_UNMOUNTING:
if (f == MOUNT_SUCCESS && m->from_proc_self_mountinfo) {
/* Still a mount point? If so, let's try again. Most likely there were multiple mount points
* stacked on top of each other. We might exceed the timeout specified by the user overall,
* but we will stop as soon as any one umount times out. */
if (m->n_retry_umount < RETRY_UMOUNT_MAX) {
log_unit_debug(u, "Mount still present, trying again.");
m->n_retry_umount++;
mount_enter_unmounting(m);
} else {
log_unit_warning(u, "Mount still present after %u attempts to unmount, giving up.", m->n_retry_umount);
mount_enter_mounted(m, f);
}
} else
mount_enter_dead_or_mounted(m, f);
break;
case MOUNT_UNMOUNTING_SIGKILL:
case MOUNT_UNMOUNTING_SIGTERM:
mount_enter_dead_or_mounted(m, f);
break;
case MOUNT_CLEANING:
if (m->clean_result == MOUNT_SUCCESS)
m->clean_result = f;
mount_enter_dead(m, MOUNT_SUCCESS);
break;
default:
assert_not_reached();
}
/* Notify clients about changed exit status */
unit_add_to_dbus_queue(u);
}
static int mount_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) {
Mount *m = MOUNT(userdata);
assert(m);
assert(m->timer_event_source == source);
switch (m->state) {
case MOUNT_MOUNTING:
case MOUNT_MOUNTING_DONE:
log_unit_warning(UNIT(m), "Mounting timed out. Terminating.");
mount_enter_signal(m, MOUNT_UNMOUNTING_SIGTERM, MOUNT_FAILURE_TIMEOUT);
break;
case MOUNT_REMOUNTING:
log_unit_warning(UNIT(m), "Remounting timed out. Terminating remount process.");
mount_set_reload_result(m, MOUNT_FAILURE_TIMEOUT);
mount_enter_signal(m, MOUNT_REMOUNTING_SIGTERM, MOUNT_SUCCESS);
break;
case MOUNT_REMOUNTING_SIGTERM:
mount_set_reload_result(m, MOUNT_FAILURE_TIMEOUT);
if (m->kill_context.send_sigkill) {
log_unit_warning(UNIT(m), "Remounting timed out. Killing.");
mount_enter_signal(m, MOUNT_REMOUNTING_SIGKILL, MOUNT_SUCCESS);
} else {
log_unit_warning(UNIT(m), "Remounting timed out. Skipping SIGKILL. Ignoring.");
mount_enter_dead_or_mounted(m, MOUNT_SUCCESS);
}
break;
case MOUNT_REMOUNTING_SIGKILL:
mount_set_reload_result(m, MOUNT_FAILURE_TIMEOUT);
log_unit_warning(UNIT(m), "Mount process still around after SIGKILL. Ignoring.");
mount_enter_dead_or_mounted(m, MOUNT_SUCCESS);
break;
case MOUNT_UNMOUNTING:
log_unit_warning(UNIT(m), "Unmounting timed out. Terminating.");
mount_enter_signal(m, MOUNT_UNMOUNTING_SIGTERM, MOUNT_FAILURE_TIMEOUT);
break;
case MOUNT_UNMOUNTING_SIGTERM:
if (m->kill_context.send_sigkill) {
log_unit_warning(UNIT(m), "Mount process timed out. Killing.");
mount_enter_signal(m, MOUNT_UNMOUNTING_SIGKILL, MOUNT_FAILURE_TIMEOUT);
} else {
log_unit_warning(UNIT(m), "Mount process timed out. Skipping SIGKILL. Ignoring.");
mount_enter_dead_or_mounted(m, MOUNT_FAILURE_TIMEOUT);
}
break;
case MOUNT_UNMOUNTING_SIGKILL:
log_unit_warning(UNIT(m), "Mount process still around after SIGKILL. Ignoring.");
mount_enter_dead_or_mounted(m, MOUNT_FAILURE_TIMEOUT);
break;
case MOUNT_CLEANING:
log_unit_warning(UNIT(m), "Cleaning timed out. killing.");
if (m->clean_result == MOUNT_SUCCESS)
m->clean_result = MOUNT_FAILURE_TIMEOUT;
mount_enter_signal(m, MOUNT_UNMOUNTING_SIGKILL, 0);
break;
default:
assert_not_reached();
}
return 0;
}
static int mount_setup_new_unit(
Manager *m,
const char *name,
const char *what,
const char *where,
const char *options,
const char *fstype,
MountProcFlags *ret_flags,
Unit **ret) {
_cleanup_(unit_freep) Unit *u = NULL;
int r;
assert(m);
assert(name);
assert(ret_flags);
assert(ret);
r = unit_new_for_name(m, sizeof(Mount), name, &u);
if (r < 0)
return r;
r = free_and_strdup(&u->source_path, "/proc/self/mountinfo");
if (r < 0)
return r;
r = free_and_strdup(&MOUNT(u)->where, where);
if (r < 0)
return r;
r = update_parameters_proc_self_mountinfo(MOUNT(u), what, options, fstype);
if (r < 0)
return r;
/* This unit was generated because /proc/self/mountinfo reported it. Remember this, so that by the
* time we load the unit file for it (and thus add in extra deps right after) we know what source to
* attributes the deps to. */
MOUNT(u)->from_proc_self_mountinfo = true;
r = mount_add_non_exec_dependencies(MOUNT(u));
if (r < 0)
return r;
/* We have only allocated the stub now, let's enqueue this unit for loading now, so that everything
* else is loaded in now. */
unit_add_to_load_queue(u);
*ret_flags = MOUNT_PROC_IS_MOUNTED | MOUNT_PROC_JUST_MOUNTED | MOUNT_PROC_JUST_CHANGED;
*ret = TAKE_PTR(u);
return 0;
}
static int mount_setup_existing_unit(
Unit *u,
const char *what,
const char *where,
const char *options,
const char *fstype,
MountProcFlags *ret_flags) {
int r;
assert(u);
assert(ret_flags);
if (!MOUNT(u)->where) {
MOUNT(u)->where = strdup(where);
if (!MOUNT(u)->where)
return -ENOMEM;
}
/* In case we have multiple mounts established on the same mount point, let's merge flags set already
* for the current unit. Note that the flags field is reset on each iteration of reading
* /proc/self/mountinfo, hence we know for sure anything already set here is from the current
* iteration and thus worthy of taking into account. */
MountProcFlags flags =
MOUNT(u)->proc_flags | MOUNT_PROC_IS_MOUNTED;
r = update_parameters_proc_self_mountinfo(MOUNT(u), what, options, fstype);
if (r < 0)
return r;
if (r > 0)
flags |= MOUNT_PROC_JUST_CHANGED;
/* There are two conditions when we consider a mount point just mounted: when we haven't seen it in
* /proc/self/mountinfo before or when MOUNT_MOUNTING is our current state. Why bother with the
* latter? Shouldn't that be covered by the former? No, during reload it is not because we might then
* encounter a new /proc/self/mountinfo in combination with an old mount unit state (since it stems
* from the serialized state), and need to catch up. Since we know that the MOUNT_MOUNTING state is
* reached when we wait for the mount to appear we hence can assume that if we are in it, we are
* actually seeing it established for the first time. */
if (!MOUNT(u)->from_proc_self_mountinfo || MOUNT(u)->state == MOUNT_MOUNTING)
flags |= MOUNT_PROC_JUST_MOUNTED;
MOUNT(u)->from_proc_self_mountinfo = true;
if (IN_SET(u->load_state, UNIT_NOT_FOUND, UNIT_BAD_SETTING, UNIT_ERROR)) {
/* The unit was previously not found or otherwise not loaded. Now that the unit shows up in
* /proc/self/mountinfo we should reconsider it this, hence set it to UNIT_LOADED. */
u->load_state = UNIT_LOADED;
u->load_error = 0;
flags |= MOUNT_PROC_JUST_CHANGED;
}
if (FLAGS_SET(flags, MOUNT_PROC_JUST_CHANGED)) {
/* If things changed, then make sure that all deps are regenerated. Let's
* first remove all automatic deps, and then add in the new ones. */
r = mount_add_non_exec_dependencies(MOUNT(u));
if (r < 0)
return r;
}
*ret_flags = flags;
return 0;
}
static int mount_setup_unit(
Manager *m,
const char *what,
const char *where,
const char *options,
const char *fstype,
bool set_flags) {
_cleanup_free_ char *e = NULL;
MountProcFlags flags;
Unit *u;
int r;
assert(m);
assert(what);
assert(where);
assert(options);
assert(fstype);
/* Ignore API mount points. They should never be referenced in
* dependencies ever. */
if (mount_point_is_api(where) || mount_point_ignore(where))
return 0;
if (streq(fstype, "autofs"))
return 0;
/* probably some kind of swap, ignore */
if (!is_path(where))
return 0;
r = unit_name_from_path(where, ".mount", &e);
if (r < 0)
return log_struct_errno(
LOG_WARNING, r,
"MESSAGE_ID=" SD_MESSAGE_MOUNT_POINT_PATH_NOT_SUITABLE_STR,
"MOUNT_POINT=%s", where,
LOG_MESSAGE("Failed to generate valid unit name from mount point path '%s', ignoring mount point: %m",
where));
u = manager_get_unit(m, e);
if (u)
r = mount_setup_existing_unit(u, what, where, options, fstype, &flags);
else
/* First time we see this mount point meaning that it's not been initiated by a mount unit
* but rather by the sysadmin having called mount(8) directly. */
r = mount_setup_new_unit(m, e, what, where, options, fstype, &flags, &u);
if (r < 0)
return log_warning_errno(r, "Failed to set up mount unit for '%s': %m", where);
/* If the mount changed properties or state, let's notify our clients */
if (flags & (MOUNT_PROC_JUST_CHANGED|MOUNT_PROC_JUST_MOUNTED))
unit_add_to_dbus_queue(u);
if (set_flags)
MOUNT(u)->proc_flags = flags;
return 0;
}
static int mount_load_proc_self_mountinfo(Manager *m, bool set_flags) {
_cleanup_(mnt_free_tablep) struct libmnt_table *table = NULL;
_cleanup_(mnt_free_iterp) struct libmnt_iter *iter = NULL;
int r;
assert(m);
r = libmount_parse(NULL, NULL, &table, &iter);
if (r < 0)
return log_error_errno(r, "Failed to parse /proc/self/mountinfo: %m");
for (;;) {
struct libmnt_fs *fs;
const char *device, *path, *options, *fstype;
r = mnt_table_next_fs(table, iter, &fs);
if (r == 1)
break;
if (r < 0)
return log_error_errno(r, "Failed to get next entry from /proc/self/mountinfo: %m");
device = mnt_fs_get_source(fs);
path = mnt_fs_get_target(fs);
options = mnt_fs_get_options(fs);
fstype = mnt_fs_get_fstype(fs);
if (!device || !path)
continue;
device_found_node(m, device, DEVICE_FOUND_MOUNT, DEVICE_FOUND_MOUNT);
(void) mount_setup_unit(m, device, path, options, fstype, set_flags);
}
return 0;
}
static void mount_shutdown(Manager *m) {
assert(m);
m->mount_event_source = sd_event_source_disable_unref(m->mount_event_source);
mnt_unref_monitor(m->mount_monitor);
m->mount_monitor = NULL;
}
static int mount_get_timeout(Unit *u, usec_t *timeout) {
Mount *m = MOUNT(u);
usec_t t;
int r;
assert(m);
assert(u);
if (!m->timer_event_source)
return 0;
r = sd_event_source_get_time(m->timer_event_source, &t);
if (r < 0)
return r;
if (t == USEC_INFINITY)
return 0;
*timeout = t;
return 1;
}
static void mount_enumerate_perpetual(Manager *m) {
Unit *u;
int r;
assert(m);
/* Whatever happens, we know for sure that the root directory is around, and cannot go away. Let's
* unconditionally synthesize it here and mark it as perpetual. */
u = manager_get_unit(m, SPECIAL_ROOT_MOUNT);
if (!u) {
r = unit_new_for_name(m, sizeof(Mount), SPECIAL_ROOT_MOUNT, &u);
if (r < 0) {
log_error_errno(r, "Failed to allocate the special " SPECIAL_ROOT_MOUNT " unit: %m");
return;
}
}
u->perpetual = true;
MOUNT(u)->deserialized_state = MOUNT_MOUNTED;
unit_add_to_load_queue(u);
unit_add_to_dbus_queue(u);
}
static bool mount_is_mounted(Mount *m) {
assert(m);
return UNIT(m)->perpetual || FLAGS_SET(m->proc_flags, MOUNT_PROC_IS_MOUNTED);
}
static int mount_on_ratelimit_expire(sd_event_source *s, void *userdata) {
Manager *m = ASSERT_PTR(userdata);
Job *j;
/* Let's enqueue all start jobs that were previously skipped because of active ratelimit. */
HASHMAP_FOREACH(j, m->jobs) {
if (j->unit->type != UNIT_MOUNT)
continue;
job_add_to_run_queue(j);
}
/* By entering ratelimited state we made all mount start jobs not runnable, now rate limit is over so
* let's make sure we dispatch them in the next iteration. */
manager_trigger_run_queue(m);
return 0;
}
static void mount_enumerate(Manager *m) {
int r;
assert(m);
mnt_init_debug(0);
if (!m->mount_monitor) {
unsigned mount_rate_limit_burst = 5;
int fd;
m->mount_monitor = mnt_new_monitor();
if (!m->mount_monitor) {
log_oom();
goto fail;
}
r = mnt_monitor_enable_kernel(m->mount_monitor, 1);
if (r < 0) {
log_error_errno(r, "Failed to enable watching of kernel mount events: %m");
goto fail;
}
r = mnt_monitor_enable_userspace(m->mount_monitor, 1, NULL);
if (r < 0) {
log_error_errno(r, "Failed to enable watching of userspace mount events: %m");
goto fail;
}
/* mnt_unref_monitor() will close the fd */
fd = r = mnt_monitor_get_fd(m->mount_monitor);
if (r < 0) {
log_error_errno(r, "Failed to acquire watch file descriptor: %m");
goto fail;
}
r = sd_event_add_io(m->event, &m->mount_event_source, fd, EPOLLIN, mount_dispatch_io, m);
if (r < 0) {
log_error_errno(r, "Failed to watch mount file descriptor: %m");
goto fail;
}
r = sd_event_source_set_priority(m->mount_event_source, SD_EVENT_PRIORITY_NORMAL-10);
if (r < 0) {
log_error_errno(r, "Failed to adjust mount watch priority: %m");
goto fail;
}
/* Let users override the default (5 in 1s), as it stalls the boot sequence on busy systems. */
const char *e = secure_getenv("SYSTEMD_DEFAULT_MOUNT_RATE_LIMIT_BURST");
if (e) {
r = safe_atou(e, &mount_rate_limit_burst);
if (r < 0)
log_debug("Invalid value in $SYSTEMD_DEFAULT_MOUNT_RATE_LIMIT_BURST, ignoring: %s", e);
}
r = sd_event_source_set_ratelimit(m->mount_event_source, 1 * USEC_PER_SEC, mount_rate_limit_burst);
if (r < 0) {
log_error_errno(r, "Failed to enable rate limit for mount events: %m");
goto fail;
}
r = sd_event_source_set_ratelimit_expire_callback(m->mount_event_source, mount_on_ratelimit_expire);
if (r < 0) {
log_error_errno(r, "Failed to enable rate limit for mount events: %m");
goto fail;
}
(void) sd_event_source_set_description(m->mount_event_source, "mount-monitor-dispatch");
}
r = mount_load_proc_self_mountinfo(m, false);
if (r < 0)
goto fail;
return;
fail:
mount_shutdown(m);
}
static int drain_libmount(Manager *m) {
bool rescan = false;
int r;
assert(m);
/* Drain all events and verify that the event is valid.
*
* Note that libmount also monitors /run/mount mkdir if the directory does not exist yet. The mkdir
* may generate event which is irrelevant for us.
*
* error: r < 0; valid: r == 0, false positive: r == 1 */
do {
r = mnt_monitor_next_change(m->mount_monitor, NULL, NULL);
if (r < 0)
return log_error_errno(r, "Failed to drain libmount events: %m");
if (r == 0)
rescan = true;
} while (r == 0);
return rescan;
}
static int mount_process_proc_self_mountinfo(Manager *m) {
_cleanup_set_free_ Set *around = NULL, *gone = NULL;
const char *what;
int r;
assert(m);
r = drain_libmount(m);
if (r <= 0)
return r;
r = mount_load_proc_self_mountinfo(m, true);
if (r < 0) {
/* Reset flags, just in case, for later calls */
LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_MOUNT])
MOUNT(u)->proc_flags = 0;
return 0;
}
manager_dispatch_load_queue(m);
LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_MOUNT]) {
Mount *mount = MOUNT(u);
if (!mount_is_mounted(mount)) {
/* A mount point is not around right now. It
* might be gone, or might never have
* existed. */
if (mount->from_proc_self_mountinfo &&
mount->parameters_proc_self_mountinfo.what)
/* Remember that this device might just have disappeared */
if (set_put_strdup_full(&gone, &path_hash_ops_free, mount->parameters_proc_self_mountinfo.what) < 0)
log_oom(); /* we don't care too much about OOM here... */
mount->from_proc_self_mountinfo = false;
assert_se(update_parameters_proc_self_mountinfo(mount, NULL, NULL, NULL) >= 0);
switch (mount->state) {
case MOUNT_MOUNTED:
/* This has just been unmounted by somebody else, follow the state change. */
mount_enter_dead(mount, MOUNT_SUCCESS);
break;
case MOUNT_MOUNTING_DONE:
/* The mount command may add the corresponding proc mountinfo entry and
* then remove it because of an internal error. E.g., fuse.sshfs seems
* to do that when the connection fails. See #17617. To handle such the
* case, let's once set the state back to mounting. Then, the unit can
* correctly enter the failed state later in mount_sigchld(). */
mount_set_state(mount, MOUNT_MOUNTING);
break;
default:
break;
}
} else if (mount->proc_flags & (MOUNT_PROC_JUST_MOUNTED|MOUNT_PROC_JUST_CHANGED)) {
/* A mount point was added or changed */
switch (mount->state) {
case MOUNT_DEAD:
case MOUNT_FAILED:
/* This has just been mounted by somebody else, follow the state change, but let's
* generate a new invocation ID for this implicitly and automatically. */
(void) unit_acquire_invocation_id(u);
mount_cycle_clear(mount);
mount_enter_mounted(mount, MOUNT_SUCCESS);
break;
case MOUNT_MOUNTING:
mount_set_state(mount, MOUNT_MOUNTING_DONE);
break;
default:
/* Nothing really changed, but let's
* issue an notification call
* nonetheless, in case somebody is
* waiting for this. (e.g. file system
* ro/rw remounts.) */
mount_set_state(mount, mount->state);
break;
}
}
if (mount_is_mounted(mount) &&
mount->from_proc_self_mountinfo &&
mount->parameters_proc_self_mountinfo.what)
/* Track devices currently used */
if (set_put_strdup_full(&around, &path_hash_ops_free, mount->parameters_proc_self_mountinfo.what) < 0)
log_oom();
/* Reset the flags for later calls */
mount->proc_flags = 0;
}
SET_FOREACH(what, gone) {
if (set_contains(around, what))
continue;
/* Let the device units know that the device is no longer mounted */
device_found_node(m, what, DEVICE_NOT_FOUND, DEVICE_FOUND_MOUNT);
}
return 0;
}
static int mount_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) {
Manager *m = ASSERT_PTR(userdata);
assert(revents & EPOLLIN);
return mount_process_proc_self_mountinfo(m);
}
static void mount_reset_failed(Unit *u) {
Mount *m = MOUNT(u);
assert(m);
if (m->state == MOUNT_FAILED)
mount_set_state(m, MOUNT_DEAD);
m->result = MOUNT_SUCCESS;
m->reload_result = MOUNT_SUCCESS;
m->clean_result = MOUNT_SUCCESS;
}
static int mount_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) {
Mount *m = MOUNT(u);
assert(m);
return unit_kill_common(u, who, signo, -1, m->control_pid, error);
}
static int mount_control_pid(Unit *u) {
Mount *m = MOUNT(u);
assert(m);
return m->control_pid;
}
static int mount_clean(Unit *u, ExecCleanMask mask) {
_cleanup_strv_free_ char **l = NULL;
Mount *m = MOUNT(u);
int r;
assert(m);
assert(mask != 0);
if (m->state != MOUNT_DEAD)
return -EBUSY;
r = exec_context_get_clean_directories(&m->exec_context, u->manager->prefix, mask, &l);
if (r < 0)
return r;
if (strv_isempty(l))
return -EUNATCH;
mount_unwatch_control_pid(m);
m->clean_result = MOUNT_SUCCESS;
m->control_command = NULL;
m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID;
r = mount_arm_timer(m, usec_add(now(CLOCK_MONOTONIC), m->exec_context.timeout_clean_usec));
if (r < 0)
goto fail;
r = unit_fork_and_watch_rm_rf(u, l, &m->control_pid);
if (r < 0)
goto fail;
mount_set_state(m, MOUNT_CLEANING);
return 0;
fail:
log_unit_warning_errno(u, r, "Failed to initiate cleaning: %m");
m->clean_result = MOUNT_FAILURE_RESOURCES;
m->timer_event_source = sd_event_source_disable_unref(m->timer_event_source);
return r;
}
static int mount_can_clean(Unit *u, ExecCleanMask *ret) {
Mount *m = MOUNT(u);
assert(m);
return exec_context_get_clean_mask(&m->exec_context, ret);
}
static int mount_can_start(Unit *u) {
Mount *m = MOUNT(u);
int r;
assert(m);
r = unit_test_start_limit(u);
if (r < 0) {
mount_enter_dead(m, MOUNT_FAILURE_START_LIMIT_HIT);
return r;
}
return 1;
}
static const char* const mount_exec_command_table[_MOUNT_EXEC_COMMAND_MAX] = {
[MOUNT_EXEC_MOUNT] = "ExecMount",
[MOUNT_EXEC_UNMOUNT] = "ExecUnmount",
[MOUNT_EXEC_REMOUNT] = "ExecRemount",
};
DEFINE_STRING_TABLE_LOOKUP(mount_exec_command, MountExecCommand);
static const char* const mount_result_table[_MOUNT_RESULT_MAX] = {
[MOUNT_SUCCESS] = "success",
[MOUNT_FAILURE_RESOURCES] = "resources",
[MOUNT_FAILURE_TIMEOUT] = "timeout",
[MOUNT_FAILURE_EXIT_CODE] = "exit-code",
[MOUNT_FAILURE_SIGNAL] = "signal",
[MOUNT_FAILURE_CORE_DUMP] = "core-dump",
[MOUNT_FAILURE_START_LIMIT_HIT] = "start-limit-hit",
[MOUNT_FAILURE_PROTOCOL] = "protocol",
};
DEFINE_STRING_TABLE_LOOKUP(mount_result, MountResult);
const UnitVTable mount_vtable = {
.object_size = sizeof(Mount),
.exec_context_offset = offsetof(Mount, exec_context),
.cgroup_context_offset = offsetof(Mount, cgroup_context),
.kill_context_offset = offsetof(Mount, kill_context),
.exec_runtime_offset = offsetof(Mount, exec_runtime),
.dynamic_creds_offset = offsetof(Mount, dynamic_creds),
.sections =
"Unit\0"
"Mount\0"
"Install\0",
.private_section = "Mount",
.can_transient = true,
.can_fail = true,
.exclude_from_switch_root_serialization = true,
.init = mount_init,
.load = mount_load,
.done = mount_done,
.coldplug = mount_coldplug,
.catchup = mount_catchup,
.dump = mount_dump,
.start = mount_start,
.stop = mount_stop,
.reload = mount_reload,
.kill = mount_kill,
.clean = mount_clean,
.can_clean = mount_can_clean,
.serialize = mount_serialize,
.deserialize_item = mount_deserialize_item,
.active_state = mount_active_state,
.sub_state_to_string = mount_sub_state_to_string,
.will_restart = unit_will_restart_default,
.may_gc = mount_may_gc,
.is_extrinsic = mount_is_extrinsic,
.sigchld_event = mount_sigchld_event,
.reset_failed = mount_reset_failed,
.control_pid = mount_control_pid,
.bus_set_property = bus_mount_set_property,
.bus_commit_properties = bus_mount_commit_properties,
.get_timeout = mount_get_timeout,
.enumerate_perpetual = mount_enumerate_perpetual,
.enumerate = mount_enumerate,
.shutdown = mount_shutdown,
.status_message_formats = {
.starting_stopping = {
[0] = "Mounting %s...",
[1] = "Unmounting %s...",
},
.finished_start_job = {
[JOB_DONE] = "Mounted %s.",
[JOB_FAILED] = "Failed to mount %s.",
[JOB_TIMEOUT] = "Timed out mounting %s.",
},
.finished_stop_job = {
[JOB_DONE] = "Unmounted %s.",
[JOB_FAILED] = "Failed unmounting %s.",
[JOB_TIMEOUT] = "Timed out unmounting %s.",
},
},
.can_start = mount_can_start,
};