[PATCH] fix de_thread() vs send_group_sigqueue() race
authorOleg Nesterov <oleg@tv-sign.ru>
Mon, 7 Nov 2005 18:12:43 +0000 (21:12 +0300)
committerLinus Torvalds <torvalds@g5.osdl.org>
Tue, 8 Nov 2005 20:58:38 +0000 (12:58 -0800)
commit329f7dba5f7dc3bc9a30ad00cf373d2e83115aa1
tree3d8c8b2fd07ad2c1df959421121451207373f78f
parenta52e8381c430896d3bd6065a34fda99cb5c74c82
[PATCH] fix de_thread() vs send_group_sigqueue() race

When non-leader thread does exec, de_thread calls release_task(leader) before
calling exit_itimers(). If local timer interrupt happens in between, it can
oops in send_group_sigqueue() while taking ->sighand->siglock == NULL.

However, we can't change send_group_sigqueue() to check p->signal != NULL,
because sys_timer_create() does get_task_struct() only in SIGEV_THREAD_ID
case. So it is possible that this task_struct was already freed and we can't
trust p->signal.

This patch changes de_thread() so that leader released after exit_itimers()
call.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
fs/exec.c