shinys000114/linux
1
0
Fork 0
mirror of https://github.com/hardkernel/linux.git synced 2026-06-07 03:15:31 +09:00
Code Issues Packages Projects Releases Wiki Activity

ANDROID: sched: Don't allow frozen asymmetric tasks to remain on the rq

Browse Source 
If a task with a restricted possible CPU mask and PF_FROZEN or
PF_FREEZER_SKIP set blocks, then we must not put it back on the runqueue
to handle a signal because this could lead to migration failures later
on if the suspending CPU is not capable of running it.

Return such a task to the runqueue only if a fatal signal is pending,
and otherwise allow the task to block.

Bug: 202918514
Signed-off-by: Will Deacon <willdeacon@google.com>
Change-Id: I04cc9e65751f2bffc556c4da9ef02fe386764324
This commit is contained in:
Will Deacon
2022-02-09 18:11:25 +00:00
parent 6ebb3c5057
commit ccabb0e365
1 changed files with 18 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
kernel/sched/core.c
View File

@@ -6223,6 +6223,23 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
#endif /* CONFIG_SCHED_CORE */
static bool __task_can_run(struct task_struct *prev)
{
if (__fatal_signal_pending(prev))
return true;
if (!frozen_or_skipped(prev))
return true;
/*
* We can't safely go back on the runqueue if we're an asymmetric
* task skipping the freezer. Doing so can lead to migration failures
* later on if there aren't any suitable CPUs left around for us to
* move to.
*/
return task_cpu_possible_mask(prev) == cpu_possible_mask;
}
/*
* Constants for the sched_mode argument of __schedule().
*
@@ -6334,7 +6351,7 @@ static void __sched notrace __schedule(unsigned int sched_mode)
*/
prev_state = READ_ONCE(prev->__state);
if (!(sched_mode & SM_MASK_PREEMPT) && prev_state) {
if (signal_pending_state(prev_state, prev)) {
if (signal_pending_state(prev_state, prev) && __task_can_run(prev)) {
WRITE_ONCE(prev->__state, TASK_RUNNING);
} else {
prev->sched_contributes_to_load =