Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230601155652.1157611-1-kai.heng.feng@canonical.com> <164576ab-4e68-ca5d-0c9e-d756588cdbb5@acm.org>
In-Reply-To: <164576ab-4e68-ca5d-0c9e-d756588cdbb5@acm.org>
From:   Kai-Heng Feng <kai.heng.feng@canonical.com>
Date:   Fri, 2 Jun 2023 09:00:51 +0800
Message-ID: <CAAd53p6S=nzxgwBky6daJZ8wa-HaBODpjwLPYZ52g8FPXbbACw@mail.gmail.com>
Subject: Re: [PATCH v5] scsi: core: Wait until device is fully resumed before
 doing rescan
To:     Bart Van Assche <bvanassche@acm.org>
Cc:     jejb@linux.ibm.com, martin.petersen@oracle.com, dlemoal@kernel.org,
        bblock@linux.ibm.com, acelan.kao@canonical.com,
        linux-pm@vger.kernel.org, linux-scsi@vger.kernel.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Fri, Jun 2, 2023 at 3:48=E2=80=AFAM Bart Van Assche <bvanassche@acm.org>=
 wrote:
>
> On 6/1/23 08:56, Kai-Heng Feng wrote:
> > During system resuming process, the resuming order is from top to down.
> > Namely, the ATA host is resumed before disks connected to it.
> >
> > When an EH is scheduled while ATA host is resumed and disk device is
> > still suspended, the device_lock hold by scsi_rescan_device() is never
> > released so the dpm_resume() of the disk is blocked forerver, therefore
> > the system can never be resumed back.
> >
> > That's because scsi_attach_vpd() is expecting the disk device is in
> > operational state, as it doesn't work on suspended device.
> >
> > To avoid such deadlock, wait until the scsi device is fully resumed,
> > before continuing the rescan process.
>
> Why doesn't scsi_attach_vpd() support runtime power management? Calling
> scsi_attach_vpd() should result in a call of sdev_runtime_resume(),

It's system-wide resume in this context, so it's dpm_resume() waiting
for the lock to be released by scsi_rescan_device().

Kai-Heng

> isn't it?
>
> Thanks,
>
> Bart.
>