Re: [PATCH] scsi: ufs: core: move some irq handling back to hardirq (with time limit)

[André Draszik]

On Thu, 2025-07-24 at 09:02 -0700, Bart Van Assche wrote:
> On 7/24/25 2:54 AM, André Draszik wrote:
> > @@ -5656,19 +5689,39 @@ static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
> >   	WARN_ONCE(completed_reqs & ~hba->outstanding_reqs,
> >   		  "completed: %#lx; outstanding: %#lx\n", completed_reqs,
> >   		  hba->outstanding_reqs);
> > -	if (queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT) {
> > -		/* Do not complete polled requests from interrupt context. */
> > +	if (time_limit) {
> > +		/* Do not complete polled requests from hardirq context. */
> >   		ufshcd_clear_polled(hba, &completed_reqs);
> >   	}
> 
> This if-statement and the code inside the if-statement probably can be
> left out. This if-statement was introduced at a time when the block
> layer did not support completing polled requests from interrupt context.
> I think that commit b99182c501c3 ("bio: add pcpu caching for non-polling
> bio_put") enabled support for completing polled requests from interrupt
> context. Since this patch touches that if-statement, how about removing
> it with a separate patch that comes before this patch? Polling can be
> enabled by adding --hipri=1 to the fio command line and by using an I/O
> engine that supports polling, e.g. pvsync2 or io_uring.

Bart, thank you for taking the time to explain and the background info on
this, very helpful!

Cheers,
Andre'