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;
Message-ID: <3836423e-8f0f-1278-38e5-d35fbd7298e4@kernel.org>
Date:   Fri, 28 Jul 2023 13:09:06 +0900
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.13.0
Subject: Re: Scsi_bus_resume+0x0/0x90 returns -5 when resuming from s3 sleep
Content-Language: en-US
To:     TW <dalzot@gmail.com>,
        Thorsten Leemhuis <regressions@leemhuis.info>
Cc:     regressions@lists.linux.dev,
        Mario Limonciello <mario.limonciello@amd.com>,
        Bart Van Assche <bvanassche@acm.org>,
        LKML <linux-kernel@vger.kernel.org>, stable@vger.kernel.org
References: <0e272abe-292d-d58f-cf80-55868e793abc@gmail.com>
 <6b66dd9a-8bd5-2882-9168-8e6e0848c454@leemhuis.info>
 <c70caa9e-164c-fee5-8f85-67f6d02373ab@kernel.org>
 <b0ed86e0-3e4a-d4d1-7b9d-c57f20538a80@gmail.com>
 <86435987-734e-c6c1-a857-1ba80da709fe@gmail.com>
 <48bc1736-5e4e-3a9b-3715-60509c333bb1@kernel.org>
 <f7e7b601-571f-bd2e-6410-a8a27e510c2f@gmail.com>
 <40365501-283a-408b-3514-48c29db36861@kernel.org>
 <d44e4057-0758-4cb8-ca29-02b5bb340b8b@gmail.com>
From:   Damien Le Moal <dlemoal@kernel.org>
Organization: Western Digital Research
In-Reply-To: <d44e4057-0758-4cb8-ca29-02b5bb340b8b@gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 7/28/23 11:49, TW wrote:
> Comparing to the 5.15 kernel which had almost no delay. They are HDDs 
> though, it was working flawlessly earlier but didn't want to top post 
> again. Tried a systemd suspend instead of from the Xfce4 logout menu and 
> everything works as intended. So I'd say that it's fixed now.

I think that asynchronous scsi suspend/resume using the PM infrastructure,
introduced in 5.16, is creating this additional delay. The reason is that now,
ata and scsi drivers do their resume operations from within the PM resume
process, resulting in some delays to reach "PM: suspend exit".

I am looking into improving this, but it may take some more time.

I will post officially the patch you tried and CC you. A "tested-by" tag from
you would be appreciated.

Thanks.

> 
> On 7/27/23 20:33, Damien Le Moal wrote:
>> "Slow coming back" -> Compared to which version of the kernel ? Do you 
>> have numbers ? If the devices are HDDs, resume will wait for these to 
>> spin up. That takes a while (about 10s normally).

-- 
Damien Le Moal
Western Digital Research