Re: [PATCH v5 1/1] blktrace: switch trace spinlock to a raw spinlock

[Wander Costa]

On Mon, Dec 20, 2021 at 5:37 PM Jens Axboe <axboe@xxxxxxxxx> wrote:
>
> On 12/20/21 1:34 PM, Wander Costa wrote:
> > On Mon, Dec 20, 2021 at 5:24 PM Jens Axboe <axboe@xxxxxxxxx> wrote:
> >>
> >> On 12/20/21 12:49 PM, Wander Costa wrote:
> >>> On Mon, Dec 20, 2021 at 4:38 PM Jens Axboe <axboe@xxxxxxxxx> wrote:
> >>>>
> >>>> On 12/20/21 12:28 PM, Wander Lairson Costa wrote:
> >>>>> The running_trace_lock protects running_trace_list and is acquired
> >>>>> within the tracepoint which implies disabled preemption. The spinlock_t
> >>>>> typed lock can not be acquired with disabled preemption on PREEMPT_RT
> >>>>> because it becomes a sleeping lock.
> >>>>> The runtime of the tracepoint depends on the number of entries in
> >>>>> running_trace_list and has no limit. The blk-tracer is considered debug
> >>>>> code and higher latencies here are okay.
> >>>>
> >>>> You didn't put a changelog in here. Was this one actually compiled? Was
> >>>> it runtime tested?
> >>>
> >>> It feels like the changelog reached the inboxes after patch (at least
> >>> mine was so). Would you like that I send a v6 in the hope things
> >>> arrive in order?
> >>
> >> Not sure how you are sending them, but they don't appear to thread
> >> properly. But the changelog in the cover letter isn't really a
> >> changelog, it doesn't say what changed.
> >>
> >
> > Sorry, I think I was too brief in my explanation. I am backporting
> > this patch to the RHEL 9 kernel (which runs kernel 5.14). I mistakenly
> > generated the v4 patch from that tree, but it lacks this piece
> >
> > @@ -1608,9 +1608,9 @@ static int blk_trace_remove_queue(struct request_queue *q)
> >
> >         if (bt->trace_state == Blktrace_running) {
> >                 bt->trace_state = Blktrace_stopped;
> > -               spin_lock_irq(&running_trace_lock);
> > +               raw_spin_lock_irq(&running_trace_lock);
> >                 list_del_init(&bt->running_list);
> > -               spin_unlock_irq(&running_trace_lock);
> > +               raw_spin_unlock_irq(&running_trace_lock);
> >                 relay_flush(bt->rchan);
> >         }
> >
> > Causing the build error. v5 adds that. Sorry again for the confusion.
>
> Right, that's why I asked if a) you had even built this patch, and b) if
> you had tested it as well.
>

Yes, I had. But I had two versions of it. One for RHEL and one for
torvalds/master. I just picked the wrong branch when generating it.
I apologize for the mess once more.

> --
> Jens Axboe
>