[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <1645094128-17099-1-git-send-email-byungchul.park@lge.com>
Date: Thu, 17 Feb 2022 19:35:28 +0900
From: Byungchul Park <byungchul.park@....com>
To: torvalds@...ux-foundation.org
Cc: damien.lemoal@...nsource.wdc.com, linux-ide@...r.kernel.org,
adilger.kernel@...ger.ca, linux-ext4@...r.kernel.org,
mingo@...hat.com, linux-kernel@...r.kernel.org,
peterz@...radead.org, will@...nel.org, tglx@...utronix.de,
rostedt@...dmis.org, joel@...lfernandes.org, sashal@...nel.org,
daniel.vetter@...ll.ch, chris@...is-wilson.co.uk,
duyuyang@...il.com, johannes.berg@...el.com, tj@...nel.org,
tytso@....edu, willy@...radead.org, david@...morbit.com,
amir73il@...il.com, bfields@...ldses.org,
gregkh@...uxfoundation.org, kernel-team@....com,
linux-mm@...ck.org, akpm@...ux-foundation.org, mhocko@...nel.org,
minchan@...nel.org, hannes@...xchg.org, vdavydov.dev@...il.com,
sj@...nel.org, jglisse@...hat.com, dennis@...nel.org, cl@...ux.com,
penberg@...nel.org, rientjes@...gle.com, vbabka@...e.cz,
ngupta@...are.org, linux-block@...r.kernel.org, axboe@...nel.dk,
paolo.valente@...aro.org, josef@...icpanda.com,
linux-fsdevel@...r.kernel.org, viro@...iv.linux.org.uk,
jack@...e.cz, jack@...e.com, jlayton@...nel.org,
dan.j.williams@...el.com, hch@...radead.org, djwong@...nel.org,
dri-devel@...ts.freedesktop.org, airlied@...ux.ie,
rodrigosiqueiramelo@...il.com, melissa.srw@...il.com,
hamohammed.sa@...il.com
Subject: Re: Report in ata_scsi_port_error_handler()
<torvalds@...ux-foundation.org> wrote:
> On Tue, Feb 15, 2022 at 10:37 PM Damien Le Moal
> <damien.lemoal@...nsource.wdc.com> wrote:
> >
> > On 2/16/22 13:16, Byungchul Park wrote:
> > > [ 2.051040 ] ===================================================
> > > [ 2.051406 ] DEPT: Circular dependency has been detected.
> > > [ 2.051730 ] 5.17.0-rc1-00014-gcf3441bb2012 #2 Tainted: G W
> > > [ 2.051991 ] ---------------------------------------------------
> > > [ 2.051991 ] summary
> > > [ 2.051991 ] ---------------------------------------------------
> > > [ 2.051991 ] *** DEADLOCK ***
> > > [ 2.051991 ]
> > > [ 2.051991 ] context A
> > > [ 2.051991 ] [S] (unknown)(&(&ap->eh_wait_q)->dmap:0)
> > > [ 2.051991 ] [W] __raw_spin_lock_irq(&host->lock:0)
> > > [ 2.051991 ] [E] event(&(&ap->eh_wait_q)->dmap:0)
> > > [ 2.051991 ]
> > > [ 2.051991 ] context B
> > > [ 2.051991 ] [S] __raw_spin_lock_irqsave(&host->lock:0)
> > > [ 2.051991 ] [W] wait(&(&ap->eh_wait_q)->dmap:0)
> > > [ 2.051991 ] [E] spin_unlock(&host->lock:0)
> >
> > Sleeping with a spinlock held would be triggering warnings already, so
> > these reports seem bogus to me.
>
> Yeah, Matthew pointed out the same thing for another use-case, where
> it looks like DEPT is looking at the state at the wrong point (not at
> the scheduling point, but at prepare_to_sleep()).
>
> This ata_port_wait() is the exact same pattern, ie we have
>
> spin_lock_irqsave(ap->lock, flags);
>
> while (ap->pflags & (ATA_PFLAG_EH_PENDING | ATA_PFLAG_EH_IN_PROGRESS)) {
> prepare_to_wait(&ap->eh_wait_q, &wait, TASK_UNINTERRUPTIBLE);
> spin_unlock_irqrestore(ap->lock, flags);
> schedule();
>
> and DEPT has incorrectly taken it to mean that 'ap->lock' is held
> during the wait, when it is actually released before actually waiting.
>
> For the spin-locks, this is all very obvious (because they'd have been
> caught long ago by much simpler debug code), but the same
> prepare_to_wait -> wait pattern can most definitely happen with
> sleeping locks too, so they are all slightly suspect.
>
> And yes, the detailed reports are hard to read because the locations
> are given as "ata_port_wait_eh+0x52/0xc0". Running them through
> scripts/decode_stacktrace.sh to turn them into filename and line
> numbers - and also sort out inlining - would help a lot.
>
> Byungchul, could you fix those two issues? Some of your reports may
Of couse, that's what I should do. Thanks for your feedback.
> well be entirely valid, but the hard-to-read hex offsets and the
> knowledge that at least some of them are confused about how
> prepare_to_wait -> wait actually works makes the motivation to look at
> the details much less..
>
> Linus
Powered by blists - more mailing lists