On Sat, 23 Dec 2023 14:06:42 +0800, Rex Zhang wrote:
> The drain_workqueue() is not in a locked context. In the multi-task case,
> it's possible to call queue_work() when drain_workqueue() is ongoing, then
> it can cause Call Trace due to pushing a work into a draining workqueue:
> Call Trace:
> <TASK>
> ? __warn+0x7d/0x140
> ? __queue_work+0x2b2/0x440
> ? report_bug+0x1f8/0x200
> ? handle_bug+0x3c/0x70
> ? exc_invalid_op+0x18/0x70
> ? asm_exc_invalid_op+0x1a/0x20
> ? __queue_work+0x2b2/0x440
> queue_work_on+0x28/0x30
> idxd_misc_thread+0x303/0x5a0 [idxd]
> ? __schedule+0x369/0xb40
> ? __pfx_irq_thread_fn+0x10/0x10
> ? irq_thread+0xbc/0x1b0
> irq_thread_fn+0x21/0x70
> irq_thread+0x102/0x1b0
> ? preempt_count_add+0x74/0xa0
> ? __pfx_irq_thread_dtor+0x10/0x10
> ? __pfx_irq_thread+0x10/0x10
> kthread+0x103/0x140
> ? __pfx_kthread+0x10/0x10
> ret_from_fork+0x31/0x50
> ? __pfx_kthread+0x10/0x10
> ret_from_fork_asm+0x1b/0x30
> </TASK>
> The original lock for event log was spinlock, drain_workqueue() can't
> be in a spinlocked context because it may cause task rescheduling. The
> spinlock was called in thread and irq thread context, the current usages
> does not require a spinlock for event log, so it's feasible to convert
> spinlock to mutex.
> For putting drain_workqueue() into a locked context, convert the spinlock
> to mutex, then lock drain_workqueue() by mutex.
>
> [...]
Applied, thanks!
[1/1] dmaengine: idxd: Convert spinlock to mutex to lock evl workqueue
commit: d5638de827cff0fce77007e426ec0ffdedf68a44
Best regards,
--
~Vinod
