* [PATCH] coredump: add core_pattern specifier for si_code
@ 2026-03-17 15:00 Emanuele Rocca
2026-03-18 9:45 ` Christian Brauner
0 siblings, 1 reply; 4+ messages in thread
From: Emanuele Rocca @ 2026-03-17 15:00 UTC (permalink / raw)
To: linux-kernel
Cc: Alexander Viro, Christian Brauner, Jan Kara, linux-fsdevel,
Mark Brown, Andrew Morton
The specifiers supported by core_pattern include the option to indicate the
signal number si_signo by using %s. Other than identifying which signal
generated a core dump (eg: 11 for SIGSEGV), it is useful to know the reason why
a certain signal was sent. The signal code si_code (eg: 2 for SEGV_ACCERR)
provides this information.
Adding the signal code to core_pattern can benefit in particular sysadmins who
pipe core dumps to user-space programs for later analysis. systemd-coredump(8)
is a notable example of such programs.
Signed-off-by: Emanuele Rocca <emanuele.rocca@arm.com>
---
Documentation/admin-guide/sysctl/kernel.rst | 1 +
fs/coredump.c | 5 +++++
2 files changed, 6 insertions(+)
diff --git a/Documentation/admin-guide/sysctl/kernel.rst b/Documentation/admin-guide/sysctl/kernel.rst
index 9aed74e65cf4..20177bd94514 100644
--- a/Documentation/admin-guide/sysctl/kernel.rst
+++ b/Documentation/admin-guide/sysctl/kernel.rst
@@ -170,6 +170,7 @@ core_pattern
%d dump mode, matches ``PR_SET_DUMPABLE`` and
``/proc/sys/fs/suid_dumpable``
%s signal number
+ %n signal code
%t UNIX time of dump
%h hostname
%e executable filename (may be shortened, could be changed by prctl etc)
diff --git a/fs/coredump.c b/fs/coredump.c
index 29df8aa19e2e..e78a889fe34c 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -400,6 +400,11 @@ static bool coredump_parse(struct core_name *cn, struct coredump_params *cprm,
err = cn_printf(cn, "%d",
cprm->siginfo->si_signo);
break;
+ /* code of the signal that caused the coredump */
+ case 'n':
+ err = cn_printf(cn, "%d",
+ cprm->siginfo->si_code);
+ break;
/* UNIX time of coredump */
case 't': {
time64_t time;
--
2.47.3
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH] coredump: add core_pattern specifier for si_code
2026-03-17 15:00 [PATCH] coredump: add core_pattern specifier for si_code Emanuele Rocca
@ 2026-03-18 9:45 ` Christian Brauner
2026-03-18 13:17 ` Emanuele Rocca
0 siblings, 1 reply; 4+ messages in thread
From: Christian Brauner @ 2026-03-18 9:45 UTC (permalink / raw)
To: Emanuele Rocca
Cc: linux-kernel, Alexander Viro, Jan Kara, linux-fsdevel, Mark Brown,
Andrew Morton
On Tue, Mar 17, 2026 at 04:00:06PM +0100, Emanuele Rocca wrote:
> The specifiers supported by core_pattern include the option to indicate the
> signal number si_signo by using %s. Other than identifying which signal
> generated a core dump (eg: 11 for SIGSEGV), it is useful to know the reason why
> a certain signal was sent. The signal code si_code (eg: 2 for SEGV_ACCERR)
> provides this information.
>
> Adding the signal code to core_pattern can benefit in particular sysadmins who
> pipe core dumps to user-space programs for later analysis. systemd-coredump(8)
> is a notable example of such programs.
>
> Signed-off-by: Emanuele Rocca <emanuele.rocca@arm.com>
> ---
> Documentation/admin-guide/sysctl/kernel.rst | 1 +
> fs/coredump.c | 5 +++++
> 2 files changed, 6 insertions(+)
>
> diff --git a/Documentation/admin-guide/sysctl/kernel.rst b/Documentation/admin-guide/sysctl/kernel.rst
> index 9aed74e65cf4..20177bd94514 100644
> --- a/Documentation/admin-guide/sysctl/kernel.rst
> +++ b/Documentation/admin-guide/sysctl/kernel.rst
> @@ -170,6 +170,7 @@ core_pattern
> %d dump mode, matches ``PR_SET_DUMPABLE`` and
> ``/proc/sys/fs/suid_dumpable``
> %s signal number
> + %n signal code
Request %F which installs pidfd for the coredumping process into the
coredump helper then use the pidfd info ioctl in you binary with
PIDFD_INFO_EXIT | PIDFD_INFO_COREDUMP and you can retrieve the exit
status including the signal that was sent to the process.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] coredump: add core_pattern specifier for si_code
2026-03-18 9:45 ` Christian Brauner
@ 2026-03-18 13:17 ` Emanuele Rocca
2026-03-19 10:31 ` Christian Brauner
0 siblings, 1 reply; 4+ messages in thread
From: Emanuele Rocca @ 2026-03-18 13:17 UTC (permalink / raw)
To: Christian Brauner
Cc: linux-kernel, Alexander Viro, Jan Kara, linux-fsdevel, Mark Brown,
Andrew Morton
Hi Christian,
On 2026-03-18 10:45, Christian Brauner wrote:
> Request %F which installs pidfd for the coredumping process into the
> coredump helper then use the pidfd info ioctl in you binary with
> PIDFD_INFO_EXIT | PIDFD_INFO_COREDUMP and you can retrieve the exit
> status including the signal that was sent to the process.
Correct me if I'm wrong, but PIDFD_INFO_COREDUMP seems to be setting
coredump_signal, which is siginfo->si_signo? That information is already
available by using the %s core_pattern specifier anyways.
What I am looking for is si_code instead. Perhaps a field could be added
to struct pidfs_attr and set in fs/pidfs.c similarly to what is currently
done for si_signo?
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] coredump: add core_pattern specifier for si_code
2026-03-18 13:17 ` Emanuele Rocca
@ 2026-03-19 10:31 ` Christian Brauner
0 siblings, 0 replies; 4+ messages in thread
From: Christian Brauner @ 2026-03-19 10:31 UTC (permalink / raw)
To: Emanuele Rocca, Jann Horn, Oleg Nesterov
Cc: linux-kernel, Alexander Viro, Jan Kara, linux-fsdevel, Mark Brown,
Andrew Morton
On Wed, Mar 18, 2026 at 02:17:31PM +0100, Emanuele Rocca wrote:
> Hi Christian,
>
> On 2026-03-18 10:45, Christian Brauner wrote:
> > Request %F which installs pidfd for the coredumping process into the
> > coredump helper then use the pidfd info ioctl in you binary with
> > PIDFD_INFO_EXIT | PIDFD_INFO_COREDUMP and you can retrieve the exit
> > status including the signal that was sent to the process.
>
> Correct me if I'm wrong, but PIDFD_INFO_COREDUMP seems to be setting
> coredump_signal, which is siginfo->si_signo? That information is already
> available by using the %s core_pattern specifier anyways.
>
> What I am looking for is si_code instead. Perhaps a field could be added
> to struct pidfs_attr and set in fs/pidfs.c similarly to what is currently
> done for si_signo?
I see. I don't have a problem with doing that. Let's rope in @Jann and
@Oleg to see whether they have any info-leak/security concerns around
this.
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-03-19 10:31 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-03-17 15:00 [PATCH] coredump: add core_pattern specifier for si_code Emanuele Rocca
2026-03-18 9:45 ` Christian Brauner
2026-03-18 13:17 ` Emanuele Rocca
2026-03-19 10:31 ` Christian Brauner
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox