linux-kernel - Re: Query related to shutting down NVMe during system suspend

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: <24710253b2e34dfdd81ebe1f46b84652@codeaurora.org>
Date:   Mon, 07 Feb 2022 16:27:44 +0530
From:   nitirawa@...eaurora.org
To:     Vidya Sagar <vidyas@...dia.com>
Cc:     Keith Busch <kbusch@...nel.org>, rafael.j.wysocki@...el.com,
        hch@....de, bhelgaas@...gle.com, mmaddireddy@...dia.com,
        kthota@...dia.com, sagar.tv@...il.com, linux-pci@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: Query related to shutting down NVMe during system suspend

On 2022-02-01 22:28, Vidya Sagar wrote:
> Thanks for the super quick reply and I couldn't agree more.
> 
> On 2/1/2022 10:00 PM, Keith Busch wrote:
>> External email: Use caution opening links or attachments
>> 
>> 
>> On Tue, Feb 01, 2022 at 09:52:28PM +0530, Vidya Sagar wrote:
>>> Hi Rafael & Christoph,
>>> My query is regarding the comment and the code that follows after it 
>>> at
>>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/nvme/host/pci.c?h=v5.17-rc2#n3243
>>> What I understood from it is that, there is an underlying assumption
>>> that the power to the devices is not removed during the suspend call.
>>> In the case of device-tree based platforms like Tegra194, power is
>>> indeed removed to the devices during suspend-resume process. Hence, 
>>> the
>>> NVMe devices need to be taken through the shutdown path irrespective 
>>> of
>>> whether the ASPM states are enabled or not.
>>> I would like to hear from you the best method to follow to achieve 
>>> this.
>> 
>> Since platform makers can't converge on how to let a driver know what
>> it's supposed to do, I suggest we default to the simple shutdown 
>> suspend
>> all the time. We can add a module parameter to let a user request nvme
>> power management if they really want it. No matter what we do here,
>> someone is going to complain, but at least simple shutdown is safe...
>> 

Hi Vidya,

Are you planning to add module parameter based on above discussion. I 
see similar behaviour even with  qualcomm platform.

[  119.994092] nvme nvme0: I/O 9 QID 0 timeout, reset controller
[  120.006612] PM: dpm_run_callback(): pci_pm_resume+0x0/0xe4 returns 
-16
[  120.013502] nvme 0001:01:00.0: PM: pci_pm_resume+0x0/0xe4 returned 
-16 after 60059958 usecs
[  120.022239] nvme 0001:01:00.0: PM: failed to resume async: error -16

Regards,
Nitin