diff mbox

[trusty,1/1] UBUNTU: SAUCE: kthread: Do not leave kthread_create() immediately upon SIGKILL.

Message ID 201403291539.GBF15045.CVPONGSVVtU@I-love.SAKURA.ne.jp
State New
Headers show

Commit Message

Tetsuo Handa March 29, 2014, 6:39 a.m. UTC 
From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>

Commit 786235ee "kthread: make kthread_create() killable" changed to
leave kthread_create() as soon as receiving SIGKILL. But this change
caused boot failures if systemd-udevd worker process received SIGKILL
due to systemd's hardcoded 30 seconds timeout while loading fusion
driver using finit_module() [1].

Linux kernel people think that the systemd's hardcoded timeout is a
systemd bug. But systemd people think that loading of kernel module
needs more than 30 seconds is a kernel module's bug.

Although Linux kernel people are expecting fusion driver module not
to take more than 30 seconds, it will definitely not in time for
trusty kernel. Also, nobody can prove that fusion driver module is
the only case which is affected by commit 786235ee.

Therefore, this patch changes kthread_create() to wait for up to 10
seconds after receiving SIGKILL, unless chosen by the OOM killer,
in order to give the kthreadd a chance to complete the request.
The side effect of this patch is that current thread's response to
SIGKILL is delayed for a bit (likely less than a second, unlikely
10 seconds).

  [1] https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1276705

Reported-by: Pierre Fersing <pierref@pierref.org>
Reported-by: Faidon Liambotis <paravoid@debian.org>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
---
 kernel/kthread.c |   12 ++++++++++++
 1 files changed, 12 insertions(+), 0 deletions(-)

Comments

Tim Gardner March 29, 2014, 3:52 p.m. UTC | #1 
Tetsuo - thanks for this.
Joseph Salisbury March 31, 2014, 10:09 p.m. UTC | #2 
On 03/29/2014 11:52 AM, Tim Gardner wrote:
> Tetsuo - thanks for this.
>
I built a Trusty master-next kernel and requested testing in the bug[0].

[0] http://pad.lv/1276705
diff mbox

Patch

diff --git a/kernel/kthread.c b/kernel/kthread.c
index b5ae3ee..52ae7dc 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -292,6 +292,17 @@  struct task_struct *kthread_create_on_node(int (*threadfn)(void *data),
 	 * new kernel thread.
 	 */
 	if (unlikely(wait_for_completion_killable(&done))) {
+		int i = 0;
+
+		/*
+		 * I got SIGKILL, but wait for 10 more seconds for completion
+		 * unless chosen by the OOM killer. This delay is there as a
+		 * workaround for boot failure caused by SIGKILL upon device
+		 * driver initialization timeout.
+		 */
+		while (i++ < 10 && !test_tsk_thread_flag(current, TIF_MEMDIE))
+			if (wait_for_completion_timeout(&done, HZ))
+				goto ready;
 		/*
 		 * If I was SIGKILLed before kthreadd (or new kernel thread)
 		 * calls complete(), leave the cleanup of this structure to
@@ -305,6 +316,7 @@  struct task_struct *kthread_create_on_node(int (*threadfn)(void *data),
 		 */
 		wait_for_completion(&done);
 	}
+ready:
 	task = create->result;
 	if (!IS_ERR(task)) {
 		static const struct sched_param param = { .sched_priority = 0 };