summaryrefslogtreecommitdiff
path: root/src
diff options
context:
space:
mode:
Diffstat (limited to 'src')
-rw-r--r--src/basic/cgroup-util.c14
-rw-r--r--src/basic/cgroup-util.h4
-rw-r--r--src/basic/gunicode.c2
-rw-r--r--src/basic/rm-rf.c14
-rw-r--r--src/coredump/coredump.c2
-rw-r--r--src/journal/journal-file.c166
-rw-r--r--src/journal/journal-vacuum.c2
-rw-r--r--src/journal/journalctl.c15
-rw-r--r--src/journal/journald-server.c52
-rw-r--r--src/journal/journald-server.h2
-rw-r--r--src/libsystemd/sd-bus/bus-error.c8
-rw-r--r--src/login/systemd-user.m42
-rw-r--r--src/nspawn/nspawn-cgroup.c53
-rw-r--r--src/nspawn/nspawn-cgroup.h2
-rw-r--r--src/nspawn/nspawn.c2
-rw-r--r--src/test/test-cgroup-util.c25
16 files changed, 276 insertions, 89 deletions
diff --git a/src/basic/cgroup-util.c b/src/basic/cgroup-util.c
index 37e6928a46..cede835920 100644
--- a/src/basic/cgroup-util.c
+++ b/src/basic/cgroup-util.c
@@ -2514,6 +2514,20 @@ int cg_blkio_weight_parse(const char *s, uint64_t *ret) {
return 0;
}
+bool is_cgroup_fs(const struct statfs *s) {
+ return is_fs_type(s, CGROUP_SUPER_MAGIC) ||
+ is_fs_type(s, CGROUP2_SUPER_MAGIC);
+}
+
+bool fd_is_cgroup_fs(int fd) {
+ struct statfs s;
+
+ if (fstatfs(fd, &s) < 0)
+ return -errno;
+
+ return is_cgroup_fs(&s);
+}
+
static const char *cgroup_controller_table[_CGROUP_CONTROLLER_MAX] = {
[CGROUP_CONTROLLER_CPU] = "cpu",
[CGROUP_CONTROLLER_CPUACCT] = "cpuacct",
diff --git a/src/basic/cgroup-util.h b/src/basic/cgroup-util.h
index 7529c9719e..0aa27c4cd7 100644
--- a/src/basic/cgroup-util.h
+++ b/src/basic/cgroup-util.h
@@ -23,6 +23,7 @@
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
+#include <sys/statfs.h>
#include <sys/types.h>
#include "def.h"
@@ -254,3 +255,6 @@ CGroupController cgroup_controller_from_string(const char *s) _pure_;
int cg_weight_parse(const char *s, uint64_t *ret);
int cg_cpu_shares_parse(const char *s, uint64_t *ret);
int cg_blkio_weight_parse(const char *s, uint64_t *ret);
+
+bool is_cgroup_fs(const struct statfs *s);
+bool fd_is_cgroup_fs(int fd);
diff --git a/src/basic/gunicode.c b/src/basic/gunicode.c
index 542110503f..e6ac0545a4 100644
--- a/src/basic/gunicode.c
+++ b/src/basic/gunicode.c
@@ -26,7 +26,7 @@
char *
utf8_prev_char (const char *p)
{
- while (1)
+ for (;;)
{
p--;
if ((*p & 0xc0) != 0x80)
diff --git a/src/basic/rm-rf.c b/src/basic/rm-rf.c
index 43816fd1bb..baa70c2c8d 100644
--- a/src/basic/rm-rf.c
+++ b/src/basic/rm-rf.c
@@ -27,6 +27,7 @@
#include <unistd.h>
#include "btrfs-util.h"
+#include "cgroup-util.h"
#include "fd-util.h"
#include "log.h"
#include "macro.h"
@@ -36,9 +37,14 @@
#include "stat-util.h"
#include "string-util.h"
+static bool is_physical_fs(const struct statfs *sfs) {
+ return !is_temporary_fs(sfs) && !is_cgroup_fs(sfs);
+}
+
int rm_rf_children(int fd, RemoveFlags flags, struct stat *root_dev) {
_cleanup_closedir_ DIR *d = NULL;
int ret = 0, r;
+ struct statfs sfs;
assert(fd >= 0);
@@ -47,13 +53,13 @@ int rm_rf_children(int fd, RemoveFlags flags, struct stat *root_dev) {
if (!(flags & REMOVE_PHYSICAL)) {
- r = fd_is_temporary_fs(fd);
+ r = fstatfs(fd, &sfs);
if (r < 0) {
safe_close(fd);
- return r;
+ return -errno;
}
- if (!r) {
+ if (is_physical_fs(&sfs)) {
/* We refuse to clean physical file systems
* with this call, unless explicitly
* requested. This is extra paranoia just to
@@ -210,7 +216,7 @@ int rm_rf(const char *path, RemoveFlags flags) {
if (statfs(path, &s) < 0)
return -errno;
- if (!is_temporary_fs(&s)) {
+ if (is_physical_fs(&s)) {
log_error("Attempted to remove disk file system, and we can't allow that.");
return -EPERM;
}
diff --git a/src/coredump/coredump.c b/src/coredump/coredump.c
index db60d0af7a..a982c204be 100644
--- a/src/coredump/coredump.c
+++ b/src/coredump/coredump.c
@@ -589,7 +589,7 @@ static int get_mount_namespace_leader(pid_t pid, pid_t *container_pid) {
if (r < 0)
return r;
- while (1) {
+ for (;;) {
ino_t parent_mntns;
r = get_process_ppid(cpid, &ppid);
diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c
index 349ef74e81..49199b269f 100644
--- a/src/journal/journal-file.c
+++ b/src/journal/journal-file.c
@@ -568,8 +568,8 @@ static int journal_file_verify_header(JournalFile *f) {
return -ENODATA;
if (f->writable) {
- uint8_t state;
sd_id128_t machine_id;
+ uint8_t state;
int r;
r = sd_id128_get_machine(&machine_id);
@@ -590,6 +590,14 @@ static int journal_file_verify_header(JournalFile *f) {
log_debug("Journal file %s has unknown state %i.", f->path, state);
return -EBUSY;
}
+
+ /* Don't permit appending to files from the future. Because otherwise the realtime timestamps wouldn't
+ * be strictly ordered in the entries in the file anymore, and we can't have that since it breaks
+ * bisection. */
+ if (le64toh(f->header->tail_entry_realtime) > now(CLOCK_REALTIME)) {
+ log_debug("Journal file %s is from the future, refusing to append new data to it that'd be older.", f->path);
+ return -ETXTBSY;
+ }
}
f->compress_xz = JOURNAL_HEADER_COMPRESSED_XZ(f->header);
@@ -747,12 +755,16 @@ int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset
assert(ret);
/* Objects may only be located at multiple of 64 bit */
- if (!VALID64(offset))
+ if (!VALID64(offset)) {
+ log_debug("Attempt to move to object at non-64bit boundary: %" PRIu64, offset);
return -EBADMSG;
+ }
/* Object may not be located in the file header */
- if (offset < le64toh(f->header->header_size))
+ if (offset < le64toh(f->header->header_size)) {
+ log_debug("Attempt to move to object located in file header: %" PRIu64, offset);
return -EBADMSG;
+ }
r = journal_file_move_to(f, type, false, offset, sizeof(ObjectHeader), &t);
if (r < 0)
@@ -761,17 +773,29 @@ int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset
o = (Object*) t;
s = le64toh(o->object.size);
- if (s < sizeof(ObjectHeader))
+ if (s == 0) {
+ log_debug("Attempt to move to uninitialized object: %" PRIu64, offset);
+ return -EBADMSG;
+ }
+ if (s < sizeof(ObjectHeader)) {
+ log_debug("Attempt to move to overly short object: %" PRIu64, offset);
return -EBADMSG;
+ }
- if (o->object.type <= OBJECT_UNUSED)
+ if (o->object.type <= OBJECT_UNUSED) {
+ log_debug("Attempt to move to object with invalid type: %" PRIu64, offset);
return -EBADMSG;
+ }
- if (s < minimum_header_size(o))
+ if (s < minimum_header_size(o)) {
+ log_debug("Attempt to move to truncated object: %" PRIu64, offset);
return -EBADMSG;
+ }
- if (type > OBJECT_UNUSED && o->object.type != type)
+ if (type > OBJECT_UNUSED && o->object.type != type) {
+ log_debug("Attempt to move to object of unexpected type: %" PRIu64, offset);
return -EBADMSG;
+ }
if (s > sizeof(ObjectHeader)) {
r = journal_file_move_to(f, type, false, offset, s, &t);
@@ -2472,6 +2496,37 @@ int journal_file_compare_locations(JournalFile *af, JournalFile *bf) {
return 0;
}
+static int bump_array_index(uint64_t *i, direction_t direction, uint64_t n) {
+
+ /* Increase or decrease the specified index, in the right direction. */
+
+ if (direction == DIRECTION_DOWN) {
+ if (*i >= n - 1)
+ return 0;
+
+ (*i) ++;
+ } else {
+ if (*i <= 0)
+ return 0;
+
+ (*i) --;
+ }
+
+ return 1;
+}
+
+static bool check_properly_ordered(uint64_t new_offset, uint64_t old_offset, direction_t direction) {
+
+ /* Consider it an error if any of the two offsets is uninitialized */
+ if (old_offset == 0 || new_offset == 0)
+ return false;
+
+ /* If we go down, the new offset must be larger than the old one. */
+ return direction == DIRECTION_DOWN ?
+ new_offset > old_offset :
+ new_offset < old_offset;
+}
+
int journal_file_next_entry(
JournalFile *f,
uint64_t p,
@@ -2502,36 +2557,34 @@ int journal_file_next_entry(
if (r <= 0)
return r;
- if (direction == DIRECTION_DOWN) {
- if (i >= n - 1)
- return 0;
-
- i++;
- } else {
- if (i <= 0)
- return 0;
-
- i--;
- }
+ r = bump_array_index(&i, direction, n);
+ if (r <= 0)
+ return r;
}
/* And jump to it */
- r = generic_array_get(f,
- le64toh(f->header->entry_array_offset),
- i,
- ret, &ofs);
- if (r == -EBADMSG && direction == DIRECTION_DOWN) {
- /* Special case: when we iterate throught the journal file linearly, and hit an entry we can't read,
- * consider this the end of the journal file. */
- log_debug_errno(r, "Encountered entry we can't read while iterating through journal file. Considering this the end of the file.");
- return 0;
+ for (;;) {
+ r = generic_array_get(f,
+ le64toh(f->header->entry_array_offset),
+ i,
+ ret, &ofs);
+ if (r > 0)
+ break;
+ if (r != -EBADMSG)
+ return r;
+
+ /* OK, so this entry is borked. Most likely some entry didn't get synced to disk properly, let's see if
+ * the next one might work for us instead. */
+ log_debug_errno(r, "Entry item %" PRIu64 " is bad, skipping over it.", i);
+
+ r = bump_array_index(&i, direction, n);
+ if (r <= 0)
+ return r;
}
- if (r <= 0)
- return r;
- if (p > 0 &&
- (direction == DIRECTION_DOWN ? ofs <= p : ofs >= p)) {
- log_debug("%s: entry array corrupted at entry %" PRIu64, f->path, i);
+ /* Ensure our array is properly ordered. */
+ if (p > 0 && !check_properly_ordered(ofs, p, direction)) {
+ log_debug("%s: entry array not properly ordered at entry %" PRIu64, f->path, i);
return -EBADMSG;
}
@@ -2548,9 +2601,9 @@ int journal_file_next_entry_for_data(
direction_t direction,
Object **ret, uint64_t *offset) {
- uint64_t n, i;
- int r;
+ uint64_t i, n, ofs;
Object *d;
+ int r;
assert(f);
assert(p > 0 || !o);
@@ -2582,25 +2635,39 @@ int journal_file_next_entry_for_data(
if (r <= 0)
return r;
- if (direction == DIRECTION_DOWN) {
- if (i >= n - 1)
- return 0;
+ r = bump_array_index(&i, direction, n);
+ if (r <= 0)
+ return r;
+ }
- i++;
- } else {
- if (i <= 0)
- return 0;
+ for (;;) {
+ r = generic_array_get_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ i,
+ ret, &ofs);
+ if (r > 0)
+ break;
+ if (r != -EBADMSG)
+ return r;
- i--;
- }
+ log_debug_errno(r, "Data entry item %" PRIu64 " is bad, skipping over it.", i);
+
+ r = bump_array_index(&i, direction, n);
+ if (r <= 0)
+ return r;
+ }
+ /* Ensure our array is properly ordered. */
+ if (p > 0 && check_properly_ordered(ofs, p, direction)) {
+ log_debug("%s data entry array not properly ordered at entry %" PRIu64, f->path, i);
+ return -EBADMSG;
}
- return generic_array_get_plus_one(f,
- le64toh(d->data.entry_offset),
- le64toh(d->data.entry_array_offset),
- i,
- ret, offset);
+ if (offset)
+ *offset = ofs;
+
+ return 1;
}
int journal_file_move_to_entry_by_offset_for_data(
@@ -3271,7 +3338,8 @@ int journal_file_open_reliably(
-EBUSY, /* unclean shutdown */
-ESHUTDOWN, /* already archived */
-EIO, /* IO error, including SIGBUS on mmap */
- -EIDRM /* File has been deleted */))
+ -EIDRM, /* File has been deleted */
+ -ETXTBSY)) /* File is from the future */
return r;
if ((flags & O_ACCMODE) == O_RDONLY)
diff --git a/src/journal/journal-vacuum.c b/src/journal/journal-vacuum.c
index f09dc66e03..12ce2fd56c 100644
--- a/src/journal/journal-vacuum.c
+++ b/src/journal/journal-vacuum.c
@@ -343,7 +343,7 @@ finish:
free(list[i].filename);
free(list);
- log_full(verbose ? LOG_INFO : LOG_DEBUG, "Vacuuming done, freed %s of archived journals on disk.", format_bytes(sbytes, sizeof(sbytes), freed));
+ log_full(verbose ? LOG_INFO : LOG_DEBUG, "Vacuuming done, freed %s of archived journals from %s.", format_bytes(sbytes, sizeof(sbytes), freed), directory);
return r;
}
diff --git a/src/journal/journalctl.c b/src/journal/journalctl.c
index 4350925fb0..7f997487b4 100644
--- a/src/journal/journalctl.c
+++ b/src/journal/journalctl.c
@@ -1091,8 +1091,10 @@ static int discover_next_boot(sd_journal *j,
r = sd_journal_previous(j);
if (r < 0)
return r;
- else if (r == 0)
+ else if (r == 0) {
+ log_debug("Whoopsie! We found a boot ID but can't read its last entry.");
return -ENODATA; /* This shouldn't happen. We just came from this very boot ID. */
+ }
r = sd_journal_get_realtime_usec(j, &next_boot->last);
if (r < 0)
@@ -1112,7 +1114,7 @@ static int get_boots(
bool skip_once;
int r, count = 0;
- BootId *head = NULL, *tail = NULL;
+ BootId *head = NULL, *tail = NULL, *id;
const bool advance_older = boot_id && offset <= 0;
sd_id128_t previous_boot_id;
@@ -1203,6 +1205,13 @@ static int get_boots(
break;
}
} else {
+ LIST_FOREACH(boot_list, id, head) {
+ if (sd_id128_equal(id->id, current->id)) {
+ /* boot id already stored, something wrong with the journal files */
+ /* exiting as otherwise this problem would cause forever loop */
+ goto finish;
+ }
+ }
LIST_INSERT_AFTER(boot_list, head, tail, current);
tail = current;
current = NULL;
@@ -2257,7 +2266,7 @@ int main(int argc, char *argv[]) {
if (r < 0)
goto finish;
- printf("Archived and active journals take up %s on disk.\n",
+ printf("Archived and active journals take up %s in the file system.\n",
format_bytes(sbytes, sizeof(sbytes), bytes));
goto finish;
}
diff --git a/src/journal/journald-server.c b/src/journal/journald-server.c
index f01cf1d937..381182fa2c 100644
--- a/src/journal/journald-server.c
+++ b/src/journal/journald-server.c
@@ -595,52 +595,86 @@ static void server_cache_hostname(Server *s) {
static bool shall_try_append_again(JournalFile *f, int r) {
switch(r) {
+
case -E2BIG: /* Hit configured limit */
case -EFBIG: /* Hit fs limit */
case -EDQUOT: /* Quota limit hit */
case -ENOSPC: /* Disk full */
log_debug("%s: Allocation limit reached, rotating.", f->path);
return true;
+
case -EIO: /* I/O error of some kind (mmap) */
log_warning("%s: IO error, rotating.", f->path);
return true;
+
case -EHOSTDOWN: /* Other machine */
log_info("%s: Journal file from other machine, rotating.", f->path);
return true;
+
case -EBUSY: /* Unclean shutdown */
log_info("%s: Unclean shutdown, rotating.", f->path);
return true;
+
case -EPROTONOSUPPORT: /* Unsupported feature */
log_info("%s: Unsupported feature, rotating.", f->path);
return true;
+
case -EBADMSG: /* Corrupted */
case -ENODATA: /* Truncated */
case -ESHUTDOWN: /* Already archived */
log_warning("%s: Journal file corrupted, rotating.", f->path);
return true;
+
case -EIDRM: /* Journal file has been deleted */
log_warning("%s: Journal file has been deleted, rotating.", f->path);
return true;
+
+ case -ETXTBSY: /* Journal file is from the future */
+ log_warning("%s: Journal file is from the future, rotating.", f->path);
+ return true;
+
default:
return false;
}
}
static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n, int priority) {
+ bool vacuumed = false, rotate = false;
+ struct dual_timestamp ts;
JournalFile *f;
- bool vacuumed = false;
int r;
assert(s);
assert(iovec);
assert(n > 0);
- f = find_journal(s, uid);
- if (!f)
- return;
+ /* Get the closest, linearized time we have for this log event from the event loop. (Note that we do not use
+ * the source time, and not even the time the event was originally seen, but instead simply the time we started
+ * processing it, as we want strictly linear ordering in what we write out.) */
+ assert_se(sd_event_now(s->event, CLOCK_REALTIME, &ts.realtime) >= 0);
+ assert_se(sd_event_now(s->event, CLOCK_MONOTONIC, &ts.monotonic) >= 0);
+
+ if (ts.realtime < s->last_realtime_clock) {
+ /* When the time jumps backwards, let's immediately rotate. Of course, this should not happen during
+ * regular operation. However, when it does happen, then we should make sure that we start fresh files
+ * to ensure that the entries in the journal files are strictly ordered by time, in order to ensure
+ * bisection works correctly. */
+
+ log_debug("Time jumped backwards, rotating.");
+ rotate = true;
+ } else {
+
+ f = find_journal(s, uid);
+ if (!f)
+ return;
+
+ if (journal_file_rotate_suggested(f, s->max_file_usec)) {
+ log_debug("%s: Journal header limits reached or header out-of-date, rotating.", f->path);
+ rotate = true;
+ }
+ }
- if (journal_file_rotate_suggested(f, s->max_file_usec)) {
- log_debug("%s: Journal header limits reached or header out-of-date, rotating.", f->path);
+ if (rotate) {
server_rotate(s);
server_vacuum(s, false, false);
vacuumed = true;
@@ -650,7 +684,9 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned
return;
}
- r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
+ s->last_realtime_clock = ts.realtime;
+
+ r = journal_file_append_entry(f, &ts, iovec, n, &s->seqnum, NULL, NULL);
if (r >= 0) {
server_schedule_sync(s, priority);
return;
@@ -669,7 +705,7 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned
return;
log_debug("Retrying write.");
- r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
+ r = journal_file_append_entry(f, &ts, iovec, n, &s->seqnum, NULL, NULL);
if (r < 0)
log_error_errno(r, "Failed to write entry (%d items, %zu bytes) despite vacuuming, ignoring: %m", n, IOVEC_TOTAL_SIZE(iovec, n));
else
diff --git a/src/journal/journald-server.h b/src/journal/journald-server.h
index dfb5724794..cc68a0a690 100644
--- a/src/journal/journald-server.h
+++ b/src/journal/journald-server.h
@@ -149,6 +149,8 @@ struct Server {
char *cgroup_root;
usec_t watchdog_usec;
+
+ usec_t last_realtime_clock;
};
#define SERVER_MACHINE_ID(s) ((s)->machine_id_field + strlen("_MACHINE_ID="))
diff --git a/src/libsystemd/sd-bus/bus-error.c b/src/libsystemd/sd-bus/bus-error.c
index 26219bdeed..378f7a377a 100644
--- a/src/libsystemd/sd-bus/bus-error.c
+++ b/src/libsystemd/sd-bus/bus-error.c
@@ -70,11 +70,9 @@ BUS_ERROR_MAP_ELF_REGISTER const sd_bus_error_map bus_standard_errors[] = {
SD_BUS_ERROR_MAP_END
};
-/* GCC maps this magically to the beginning and end of the BUS_ERROR_MAP section.
- * Hide them; for currently unknown reasons they get exported to the shared libries
- * even without being listed in the sym file. */
-extern const sd_bus_error_map __start_BUS_ERROR_MAP[] _hidden_;
-extern const sd_bus_error_map __stop_BUS_ERROR_MAP[] _hidden_;
+/* GCC maps this magically to the beginning and end of the BUS_ERROR_MAP section */
+extern const sd_bus_error_map __start_BUS_ERROR_MAP[];
+extern const sd_bus_error_map __stop_BUS_ERROR_MAP[];
/* Additional maps registered with sd_bus_error_add_map() are in this
* NULL terminated array */
diff --git a/src/login/systemd-user.m4 b/src/login/systemd-user.m4
index fe38b24fef..e33963b125 100644
--- a/src/login/systemd-user.m4
+++ b/src/login/systemd-user.m4
@@ -2,6 +2,8 @@
#
# Used by systemd --user instances.
+account required pam_unix.so
+
m4_ifdef(`HAVE_SELINUX',
session required pam_selinux.so close
session required pam_selinux.so nottys open
diff --git a/src/nspawn/nspawn-cgroup.c b/src/nspawn/nspawn-cgroup.c
index 6793df1286..fd0578b85c 100644
--- a/src/nspawn/nspawn-cgroup.c
+++ b/src/nspawn/nspawn-cgroup.c
@@ -25,27 +25,18 @@
#include "mkdir.h"
#include "mount-util.h"
#include "nspawn-cgroup.h"
+#include "rm-rf.h"
#include "string-util.h"
#include "strv.h"
#include "util.h"
-int chown_cgroup(pid_t pid, uid_t uid_shift) {
- _cleanup_free_ char *path = NULL, *fs = NULL;
+static int chown_cgroup_path(const char *path, uid_t uid_shift) {
_cleanup_close_ int fd = -1;
const char *fn;
- int r;
- r = cg_pid_get_path(NULL, pid, &path);
- if (r < 0)
- return log_error_errno(r, "Failed to get container cgroup path: %m");
-
- r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, path, NULL, &fs);
- if (r < 0)
- return log_error_errno(r, "Failed to get file system path for container cgroup: %m");
-
- fd = open(fs, O_RDONLY|O_CLOEXEC|O_DIRECTORY);
+ fd = open(path, O_RDONLY|O_CLOEXEC|O_DIRECTORY);
if (fd < 0)
- return log_error_errno(errno, "Failed to open %s: %m", fs);
+ return -errno;
FOREACH_STRING(fn,
".",
@@ -63,7 +54,27 @@ int chown_cgroup(pid_t pid, uid_t uid_shift) {
return 0;
}
-int sync_cgroup(pid_t pid, CGroupUnified unified_requested) {
+int chown_cgroup(pid_t pid, uid_t uid_shift) {
+ _cleanup_free_ char *path = NULL, *fs = NULL;
+ _cleanup_close_ int fd = -1;
+ int r;
+
+ r = cg_pid_get_path(NULL, pid, &path);
+ if (r < 0)
+ return log_error_errno(r, "Failed to get container cgroup path: %m");
+
+ r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, path, NULL, &fs);
+ if (r < 0)
+ return log_error_errno(r, "Failed to get file system path for container cgroup: %m");
+
+ r = chown_cgroup_path(fs, uid_shift);
+ if (r < 0)
+ return log_error_errno(r, "Failed to chown() cgroup %s: %m", fs);
+
+ return 0;
+}
+
+int sync_cgroup(pid_t pid, CGroupUnified unified_requested, uid_t arg_uid_shift) {
_cleanup_free_ char *cgroup = NULL;
char tree[] = "/tmp/unifiedXXXXXX", pid_string[DECIMAL_STR_MAX(pid) + 1];
bool undo_mount = false;
@@ -101,14 +112,26 @@ int sync_cgroup(pid_t pid, CGroupUnified unified_requested) {
undo_mount = true;
+ /* If nspawn dies abruptly the cgroup hierarchy created below
+ * its unit isn't cleaned up. So, let's remove it
+ * https://github.com/systemd/systemd/pull/4223#issuecomment-252519810 */
+ fn = strjoina(tree, cgroup);
+ (void) rm_rf(fn, REMOVE_ROOT|REMOVE_ONLY_DIRECTORIES);
+
fn = strjoina(tree, cgroup, "/cgroup.procs");
(void) mkdir_parents(fn, 0755);
sprintf(pid_string, PID_FMT, pid);
r = write_string_file(fn, pid_string, 0);
- if (r < 0)
+ if (r < 0) {
log_error_errno(r, "Failed to move process: %m");
+ goto finish;
+ }
+ fn = strjoina(tree, cgroup);
+ r = chown_cgroup_path(fn, arg_uid_shift);
+ if (r < 0)
+ log_error_errno(r, "Failed to chown() cgroup %s: %m", fn);
finish:
if (undo_mount)
(void) umount_verbose(tree);
diff --git a/src/nspawn/nspawn-cgroup.h b/src/nspawn/nspawn-cgroup.h
index dc33da8abe..fa4321ab43 100644
--- a/src/nspawn/nspawn-cgroup.h
+++ b/src/nspawn/nspawn-cgroup.h
@@ -25,5 +25,5 @@
#include "cgroup-util.h"
int chown_cgroup(pid_t pid, uid_t uid_shift);
-int sync_cgroup(pid_t pid, CGroupUnified unified_requested);
+int sync_cgroup(pid_t pid, CGroupUnified unified_requested, uid_t uid_shift);
int create_subcgroup(pid_t pid, CGroupUnified unified_requested);
diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c
index d95204f71e..14af51fc0e 100644
--- a/src/nspawn/nspawn.c
+++ b/src/nspawn/nspawn.c
@@ -3879,7 +3879,7 @@ static int run(int master,
return r;
}
- r = sync_cgroup(*pid, arg_unified_cgroup_hierarchy);
+ r = sync_cgroup(*pid, arg_unified_cgroup_hierarchy, arg_uid_shift);
if (r < 0)
return r;
diff --git a/src/test/test-cgroup-util.c b/src/test/test-cgroup-util.c
index 43f8906172..c24c784e9b 100644
--- a/src/test/test-cgroup-util.c
+++ b/src/test/test-cgroup-util.c
@@ -24,6 +24,7 @@
#include "formats-util.h"
#include "parse-util.h"
#include "process-util.h"
+#include "stat-util.h"
#include "string-util.h"
#include "test-helper.h"
#include "user-util.h"
@@ -309,6 +310,28 @@ static void test_mask_supported(void) {
printf("'%s' is supported: %s\n", cgroup_controller_to_string(c), yes_no(m & CGROUP_CONTROLLER_TO_MASK(c)));
}
+static void test_is_cgroup_fs(void) {
+ struct statfs sfs;
+ assert_se(statfs("/sys/fs/cgroup", &sfs) == 0);
+ if (is_temporary_fs(&sfs))
+ assert_se(statfs("/sys/fs/cgroup/systemd", &sfs) == 0);
+ assert_se(is_cgroup_fs(&sfs));
+}
+
+static void test_fd_is_cgroup_fs(void) {
+ int fd;
+
+ fd = open("/sys/fs/cgroup", O_RDONLY|O_DIRECTORY|O_CLOEXEC|O_NOFOLLOW);
+ assert_se(fd >= 0);
+ if (fd_is_temporary_fs(fd)) {
+ fd = safe_close(fd);
+ fd = open("/sys/fs/cgroup/systemd", O_RDONLY|O_DIRECTORY|O_CLOEXEC|O_NOFOLLOW);
+ assert_se(fd >= 0);
+ }
+ assert_se(fd_is_cgroup_fs(fd));
+ fd = safe_close(fd);
+}
+
int main(void) {
test_path_decode_unit();
test_path_get_unit();
@@ -324,6 +347,8 @@ int main(void) {
test_slice_to_path();
test_shift_path();
TEST_REQ_RUNNING_SYSTEMD(test_mask_supported());
+ TEST_REQ_RUNNING_SYSTEMD(test_is_cgroup_fs());
+ TEST_REQ_RUNNING_SYSTEMD(test_fd_is_cgroup_fs());
return 0;
}
generated by cgit v1.2.3-54-g00ecf (git 2.45.2) at 2024-07-28 11:16:20 +0000