]> git.ipfire.org Git - people/ms/linux.git/commit - fs/exec.c
proc: Ensure we see the exit of each process tid exactly once
authorEric W. Biederman <ebiederm@xmission.com>
Sun, 19 Apr 2020 11:35:02 +0000 (06:35 -0500)
committerEric W. Biederman <ebiederm@xmission.com>
Tue, 28 Apr 2020 21:13:58 +0000 (16:13 -0500)
commit6b03d1304a32dc8450c7516000a0fe07bef7c446
tree440123d6108568366d2cd81c40e7b638ec57bc68
parent35fc0e3b0bd5be3b059e53ae90c4536ee4922330
proc: Ensure we see the exit of each process tid exactly once

When the thread group leader changes during exec and the old leaders
thread is reaped proc_flush_pid will flush the dentries for the entire
process because the leader still has it's original pid.

Fix this by exchanging the pids in an rcu safe manner,
and wrapping the code to do that up in a helper exchange_tids.

When I removed switch_exec_pids and introduced this behavior
in d73d65293e3e ("[PATCH] pidhash: kill switch_exec_pids") there
really was nothing that cared as flushing happened with
the cached dentry and de_thread flushed both of them on exec.

This lack of fully exchanging pids became a problem a few months later
when I introduced 48e6484d4902 ("[PATCH] proc: Rewrite the proc dentry
flush on exit optimization").  Which overlooked the de_thread case
was no longer swapping pids, and I was looking up proc dentries
by task->pid.

The current behavior isn't properly a bug as everything in proc will
continue to work correctly just a little bit less efficiently.  Fix
this just so there are no little surprise corner cases waiting to bite
people.

-- Oleg points out this could be an issue in next_tgid in proc where
   has_group_leader_pid is called, and reording some of the assignments
   should fix that.

-- Oleg points out this will break the 10 year old hack in __exit_signal.c
> /*
>  * This can only happen if the caller is de_thread().
>  * FIXME: this is the temporary hack, we should teach
>  * posix-cpu-timers to handle this case correctly.
>  */
> if (unlikely(has_group_leader_pid(tsk)))
> posix_cpu_timers_exit_group(tsk);

The code in next_tgid has been changed to use PIDTYPE_TGID,
and the posix cpu timers code has been fixed so it does not
need the 10 year old hack, so this should be safe to merge
now.

Link: https://lore.kernel.org/lkml/87h7x3ajll.fsf_-_@x220.int.ebiederm.org/
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Fixes: 48e6484d4902 ("[PATCH] proc: Rewrite the proc dentry flush on exit optimization").
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
fs/exec.c
include/linux/pid.h
kernel/pid.c