linux-kernel - Re: [PATCH] scsi: core: use blk_mq_requeue_request in __scsi_queue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Wed, 28 Feb 2018 10:25:10 +0800
From:   "jianchao.wang" <jianchao.w.wang@...cle.com>
To:     Bart Van Assche <Bart.VanAssche@....com>,
        "jejb@...ux.vnet.ibm.com" <jejb@...ux.vnet.ibm.com>,
        "martin.petersen@...cle.com" <martin.petersen@...cle.com>
Cc:     "linux-scsi@...r.kernel.org" <linux-scsi@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "hch@....de" <hch@....de>
Subject: Re: [PATCH] scsi: core: use blk_mq_requeue_request in
 __scsi_queue_insert

Hi Bart

Thanks for your kindly response and precious time to review this.

On 02/28/2018 01:18 AM, Bart Van Assche wrote:
> On Tue, 2018-02-27 at 17:06 +0000, Bart Van Assche wrote:
>> On Tue, 2018-02-27 at 13:15 +0800, jianchao.wang wrote:
>>> Can you share more details about this ? 
>>
>> After having had another look, I think your patch is fine.
> 
> (replying to my own e-mail)
> 
> What I think is fine in your patch is that it skips the unprep and reprep
> when requeueing. However, there is a put_device(&sdev->sdev_gendev) call
> in scsi_mq_requeue_cmd() and your patch causes that put_device() call to
> be skipped when requeueing. An explanation is needed in the commit message
> why you think that removing that put_device() call is fine.

Your concern is right.
For the block legacy path in scsi core, the get_device(&sdev->sdev_gendev) is in prep.
So when it requeue the request w/ RQF_DONTPREP, the reference will not be got again.
However, for blk-mq patch in scsi core, the get_device(&sdev->sdev_gendev) in .get_budget,
so put_device is still needed here.

Thanks for your directive.
Jianchao