workqueue: annotate racy sum_exec_runtime reads for CPU-intensive detection

The automatic CPU-intensive work item detection reads the worker task's se.sum_exec_runtime without a lock in wq_worker_running(), wq_worker_tick() and process_one_work(). The scheduler updates that field under the rq lock (from the tick via update_curr(), or cross-CPU via task_sched_runtime()), raising: BUG: KCSAN: data-race in wq_worker_running+0xa8/0xe8 race at unknown origin, with read to 0xffff0009a11d1df8 of 8 bytes by task 238535 on cpu 68: wq_worker_running schedule schedule_preempt_disabled __mutex_lock mutex_lock_nested cgroup_bpf_release process_one_work worker_thread kthread ret_from_fork value changed: 0x0000000088482ba0 -> 0x00000000884893c0 The value only feeds a heuristic, so the race is benign-ish. Unlike commit ecf5aad9a441 ("workqueue: annotate racy PWQ_STAT_CPU_TIME update in wq_worker_tick()") that only needs data_race(), these are plain reads whose result drives a subtraction and comparison, so use READ_ONCE() for a single, non-torn load, which also silences KCSAN.

Breno Leitao committed 4 months ago

Commit 5eaadeb1 parent ecf5aad1 file · +3 −3Browse files

Showing the changes to kernel/workqueue.c (1 of 1 file). Show all

Unified · Split

kernel/workqueue.c+3−3

@@ -1468,7 +1468,7 @@ void wq_worker_running(struct task_struct *task) · 1467 unmodified lines
1468	1468	* CPU intensive auto-detection cares about how long a work item hogged
1469	1469	* CPU without sleeping. Reset the starting timestamp on wakeup.
1470	1470	*/
1471		− worker->current_at = worker->task->se.sum_exec_runtime;
	1471	+ worker->current_at = READ_ONCE(worker->task->se.sum_exec_runtime);
1472	1472
1473	1473	WRITE_ONCE(worker->sleeping, 0);
1474	1474	}
@@ -1557,7 +1557,7 @@ void wq_worker_tick(struct task_struct *task) · 82 unmodified lines
1557	1557	* We probably want to make this prettier in the future.
1558	1558	*/
1559	1559	if ((worker->flags & WORKER_NOT_RUNNING) \|\| READ_ONCE(worker->sleeping) \|\|
1560		− worker->task->se.sum_exec_runtime - worker->current_at <
	1560	+ READ_ONCE(worker->task->se.sum_exec_runtime) - worker->current_at <
1561	1561	wq_cpu_intensive_thresh_us * NSEC_PER_USEC)
1562	1562	return;
1563	1563
@@ -3294,7 +3294,7 @@ __acquires(&pool->lock) · 1730 unmodified lines
3294	3294	worker->current_func = work->func;
3295	3295	worker->current_pwq = pwq;
3296	3296	if (worker->task)
3297		− worker->current_at = worker->task->se.sum_exec_runtime;
	3297	+ worker->current_at = READ_ONCE(worker->task->se.sum_exec_runtime);
3298	3298	worker->current_start = jiffies;
3299	3299	work_data = *work_data_bits(work);
3300	3300	worker->current_color = get_work_color(work_data);