]> git.karo-electronics.de Git - linux-beck.git/commit
sched: fix TASK_WAKEKILL vs SIGKILL race
authorOleg Nesterov <oleg@tv-sign.ru>
Sun, 8 Jun 2008 17:20:41 +0000 (21:20 +0400)
committerIngo Molnar <mingo@elte.hu>
Tue, 10 Jun 2008 09:37:25 +0000 (11:37 +0200)
commit16882c1e962b4be5122fc05aaf2afc10fd9e2d15
treec39cae4ae4874998d7e3486cd57d57613b05a89c
parent39b945a37bac2b692773a470890c8ba301485b15
sched: fix TASK_WAKEKILL vs SIGKILL race

schedule() has the special "TASK_INTERRUPTIBLE && signal_pending()" case,
this allows us to do

current->state = TASK_INTERRUPTIBLE;
schedule();

without fear to sleep with pending signal.

However, the code like

current->state = TASK_KILLABLE;
schedule();

is not right, schedule() doesn't take TASK_WAKEKILL into account. This means
that mutex_lock_killable(), wait_for_completion_killable(), down_killable(),
schedule_timeout_killable() can miss SIGKILL (and btw the second SIGKILL has
no effect).

Introduce the new helper, signal_pending_state(), and change schedule() to
use it. Hopefully it will have more users, that is why the task's state is
passed separately.

Note this "__TASK_STOPPED | __TASK_TRACED" check in signal_pending_state().
This is needed to preserve the current behaviour (ptrace_notify). I hope
this check will be removed soon, but this (afaics good) change needs the
separate discussion.

The fast path is "(state & (INTERRUPTIBLE | WAKEKILL)) + signal_pending(p)",
basically the same that schedule() does now. However, this patch of course
bloats schedule().

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
include/linux/sched.h
kernel/sched.c