linux-kernel - Re: [PATCH v2 4/6] scsi: pm8001: Use sas_task_find

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

Message-ID: <5e570c35-ed98-a29b-736f-677aa896f654@huawei.com>
Date:   Fri, 30 Sep 2022 11:20:46 +0100
From:   John Garry <john.garry@...wei.com>
To:     Jinpu Wang <jinpu.wang@...os.com>, <damien.lemoal@....com>
CC:     <jejb@...ux.ibm.com>, <martin.petersen@...cle.com>, <hare@...e.de>,
        <linux-scsi@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <linuxarm@...wei.com>, <ipylypiv@...gle.com>,
        <changyuanl@...gle.com>, <hch@....de>, <yanaijie@...wei.com>
Subject: Re: [PATCH v2 4/6] scsi: pm8001: Use sas_task_find_rq() for tagging

On 30/09/2022 10:17, Jinpu Wang wrote:
> On Fri, Sep 30, 2022 at 11:03 AM John Garry<john.garry@...wei.com>  wrote:
>> The request associated with a scsi command coming from the block layer
>> has a unique tag, so use that when possible for getting a CCB.
>>
>> Unfortunately we don't support reserved commands in the SCSI midlayer yet,
>> so in the interim continue to manage those tags internally (along with
>> tags for private commands).
>>
>> Signed-off-by: John Garry<john.garry@...wei.com>
> Reviewed-by: Jack Wang<jinpu.wang@...os.com>
> nice, I would expect this can improve performance, do you have numbers?

Unfortunately my system hangs after I run for an appreciable period of 
time. I normally get around it by turning on much heavy debug options, 
but that would not be much use for performance testing.

But we did get considerable performance improvement for hisi_sas when we 
made the equivalent change.

Damien, if you are interested then sharing any results would be great.

BTW, I do notice that we still have this global lock in delivery path 
which should be removed at some stage:

int mvs_queue_command(struct sas_task *task, gfp_t gfp_flags)
{
	...

	spin_lock_irqsave(&mvi->lock, flags);
	rc = mvs_task_prep(task, mvi, is_tmf, tmf, &pass);
	...
	spin_unlock_irqrestore(&mvi->lock, flags);
}

That really will affect performance...

Thanks,
John