coredump: use %d in kernel core pattern

The kernel provides %d which is documented as
"dump mode—same as value returned by prctl(2) PR_GET_DUMPABLE".

We already query /proc/pid/auxv for this information, but unfortunately this
check is subject to a race, because the crashed process may be replaced by an
attacker before we read this data, for example replacing a SUID process that
was killed by a signal with another process that is not SUID, tricking us into
making the coredump of the original process readable by the attacker.

With this patch, we effectively add one more check to the list of conditions
that need be satisfied if we are to make the coredump accessible to the user.

Reportedy-by: Qualys Security Advisory <qsa@qualys.com>

In principle, %d might return a value other than 0, 1, or 2 in the future.
Thus, we accept those, but emit a notice.
This commit is contained in:
Zbigniew Jędrzejewski-Szmek
2025-04-29 14:47:59 +02:00
parent 49f1f2d4a7
commit 0c49e0049b
4 changed files with 36 additions and 4 deletions

View File

@@ -329,6 +329,18 @@ COREDUMP_FILENAME=/var/lib/systemd/coredump/core.Web….552351.….zst
</listitem>
</varlistentry>
<varlistentry>
<term><varname>COREDUMP_DUMPABLE=</varname></term>
<listitem><para>The <constant>PR_GET_DUMPABLE</constant> field as reported by the kernel, see
<citerefentry
project='man-pages'><refentrytitle>prctl</refentrytitle><manvolnum>2</manvolnum></citerefentry>.
</para>
<xi:include href="version-info.xml" xpointer="v258"/>
</listitem>
</varlistentry>
<varlistentry>
<term><varname>COREDUMP_OPEN_FDS=</varname></term>

View File

@@ -103,6 +103,7 @@ typedef enum {
_META_ARGV_REQUIRED,
/* The fields below were added to kernel/core_pattern at later points, so they might be missing. */
META_ARGV_HOSTNAME = _META_ARGV_REQUIRED, /* %h: hostname */
META_ARGV_DUMPABLE, /* %d: as set by the kernel */
/* If new fields are added, they should be added here, to maintain compatibility
* with callers which don't know about the new fields. */
_META_ARGV_MAX,
@@ -131,6 +132,7 @@ static const char * const meta_field_names[_META_MAX] = {
[META_ARGV_TIMESTAMP] = "COREDUMP_TIMESTAMP=",
[META_ARGV_RLIMIT] = "COREDUMP_RLIMIT=",
[META_ARGV_HOSTNAME] = "COREDUMP_HOSTNAME=",
[META_ARGV_DUMPABLE] = "COREDUMP_DUMPABLE=",
[META_COMM] = "COREDUMP_COMM=",
[META_EXE] = "COREDUMP_EXE=",
[META_UNIT] = "COREDUMP_UNIT=",
@@ -141,6 +143,7 @@ typedef struct Context {
PidRef pidref;
uid_t uid;
gid_t gid;
unsigned dumpable;
int signo;
uint64_t rlimit;
bool is_pid1;
@@ -437,14 +440,16 @@ static int grant_user_access(int core_fd, const Context *context) {
if (r < 0)
return r;
/* We allow access if we got all the data and at_secure is not set and
* the uid/gid matches euid/egid. */
/* We allow access if %d/dumpable on the command line was exactly 1, we got all the data,
* at_secure is not set, and the uid/gid match euid/egid. */
bool ret =
context->dumpable == 1 &&
at_secure == 0 &&
uid != UID_INVALID && euid != UID_INVALID && uid == euid &&
gid != GID_INVALID && egid != GID_INVALID && gid == egid;
log_debug("Will %s access (uid="UID_FMT " euid="UID_FMT " gid="GID_FMT " egid="GID_FMT " at_secure=%s)",
log_debug("Will %s access (dumpable=%u uid="UID_FMT " euid="UID_FMT " gid="GID_FMT " egid="GID_FMT " at_secure=%s)",
ret ? "permit" : "restrict",
context->dumpable,
uid, euid, gid, egid, yes_no(at_secure));
return ret;
}
@@ -1087,6 +1092,16 @@ static int context_parse_iovw(Context *context, struct iovec_wrapper *iovw) {
if (r < 0)
log_warning_errno(r, "Failed to parse resource limit \"%s\", ignoring: %m", context->meta[META_ARGV_RLIMIT]);
/* The value is set to contents of /proc/sys/fs/suid_dumpable, which we set to 2,
* if the process is marked as not dumpable, see PR_SET_DUMPABLE(2const). */
if (context->meta[META_ARGV_DUMPABLE]) {
r = safe_atou(context->meta[META_ARGV_DUMPABLE], &context->dumpable);
if (r < 0)
return log_error_errno(r, "Failed to parse dumpable field \"%s\": %m", context->meta[META_ARGV_DUMPABLE]);
if (context->dumpable > 2)
log_notice("Got unexpected %%d/dumpable value %u.", context->dumpable);
}
unit = context->meta[META_UNIT];
context->is_pid1 = streq(context->meta[META_ARGV_PID], "1") || streq_ptr(unit, SPECIAL_INIT_SCOPE);
context->is_journald = streq_ptr(unit, SPECIAL_JOURNALD_SERVICE);

View File

@@ -13,7 +13,7 @@
# the core dump.
#
# See systemd-coredump(8) and core(5).
kernel.core_pattern=|{{LIBEXECDIR}}/systemd-coredump %P %u %g %s %t %c %h
kernel.core_pattern=|{{LIBEXECDIR}}/systemd-coredump %P %u %g %s %t %c %h %d
# Allow 16 coredumps to be dispatched in parallel by the kernel.
# We collect metadata from /proc/%P/, and thus need to make sure the crashed

View File

@@ -198,12 +198,17 @@ journalctl -b -n 1 --output=export --output-fields=MESSAGE,COREDUMP COREDUMP_EXE
/usr/lib/systemd/systemd-coredump --backtrace $$ 0 0 6 1679509900 12345
journalctl -b -n 1 --output=export --output-fields=MESSAGE,COREDUMP COREDUMP_EXE="/usr/bin/test-dump" |
/usr/lib/systemd/systemd-coredump --backtrace $$ 0 0 6 1679509901 12345 mymachine
journalctl -b -n 1 --output=export --output-fields=MESSAGE,COREDUMP COREDUMP_EXE="/usr/bin/test-dump" |
/usr/lib/systemd/systemd-coredump --backtrace $$ 0 0 6 1679509902 12345 youmachine 1
# Wait a bit for the coredumps to get processed
timeout 30 bash -c "while [[ \$(coredumpctl list -q --no-legend $$ | wc -l) -lt 2 ]]; do sleep 1; done"
coredumpctl info $$
coredumpctl info COREDUMP_TIMESTAMP=1679509900000000
coredumpctl info COREDUMP_TIMESTAMP=1679509901000000
coredumpctl info COREDUMP_HOSTNAME="mymachine"
coredumpctl info COREDUMP_TIMESTAMP=1679509902000000
coredumpctl info COREDUMP_HOSTNAME="youmachine"
coredumpctl info COREDUMP_DUMPABLE="1"
# This used to cause a stack overflow
systemd-run -t --property CoredumpFilter=all ls /tmp