Re: [PATCH 1/1] mpt3sas: Remove usage of dma_get_required_mask api

["Linux regression tracking (Thorsten Leemhuis)" <regressions@leemhuis.info>]

On 27.02.23 15:07, Thorsten Leemhuis wrote:
> On 22.02.23 17:44, Salvatore Bonaccorso wrote:
>> On Tue, Jan 24, 2023 at 10:37:23AM +0100, Salvatore Bonaccorso wrote:
>>> On Mon, Jan 02, 2023 at 08:06:41AM -0500, Martin K. Petersen wrote:
>>>>> is anything blocking mainline inclusion of this patch?
>>>> I app

[removing a footer from the quoting here, that accidentally was added
here in my previous mail; sorry!]

>>>> lied these to 6.2/scsi-fixes last week. The patches have been
>>>> sitting in a topic branch for a bit due to the three-way conflict
>>>> between fixes, queue, and upstream.
>>>
>>> It landed in 6.2-rc4 recently in fact. Thank you!
>>>
>>> Would it be posssible to backport the fix as well back to the stable
>>> series affected? 
>>>
>>> In Debian we have the reports as per https://bugs.debian.org/1022126
>>> where the issue was introduced back in 5.10.y. Context in
>>> https://lore.kernel.org/linux-scsi/CAK=zhgr=MYn=-mrz3gKUFoXG_+EQ796bHEWSdK88o1Aqamby7g@mail.gmail.com/
>>
>> Friendly ping on this, can this change be backported as well to the
>> relevant stable series? It would apply already cleanly to 6.1.y, but
>> due to 9df650963bf6 ("scsi: mpt3sas: Don't change DMA mask while
>> reallocating pools") it might need some additional review for the
>> older stable series (in particular of interest due to the above for
>> 5.10.y).
> 
> This afaics is a reasonable request for 6.1, as this seems to be
> (Salvatore, please correct me if I'm wrong) a regression caused by
> 0e0747de0ea3 ("scsi: mpt3sas: Fix return value check of
> dma_get_required_mask()"), which was merged for 6.0-rc7. Hence allow me
> to ask:
> 
> Sreekanth and Martin, is there a reason why this request (and the
> earlier one a month ago) was apparently met with silence? Or was
> progress made in between and I just missed it?

Salvatore, seems my inquiry didn't help. I'd suggest you ask the stable
maintainers yourself to pick this up for 6.1.y. See "Option 2" in


https://docs.kernel.org/process/stable-kernel-rules.html

> Salvatore, for 5.10 things are a bit more complicated, as someone would
> need to do the work. Sometimes that work is done by the driver
> developers and maintainers as well, but strictly speaking it's the duty
> of those that backported the change to 5.10.y. Didn't check who did that
> this case (the stable team?); but well, maybe let's sort this out for
> 6.1.y first.

Option 3 mentioned on above page might work for you here.

>> Thanks already! If the change for older series needs some additional
>> testing we might ask the affected users from the Debian bug 1022126 to
>> test on 5.10.y as well.

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat)
--
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.