[regression] Re: 5.14 rc6 broken for QCA6390 on Dell XPS 13 9310

[Kalle Valo]

(Adding Linus and Greg)

Kalle Valo <kvalo at codeaurora.org> writes:

> (Adding Dave and Jakub to CC)
>
> Nicolas Schichan <nschichan at freebox.fr> writes:
>
>> On 23/08/2021 14:53, Kalle Valo wrote:
>>> (adding linux-wireless, regressions and some mhi folks)
>>> 
>>> Wren Turkal <wt at penguintechs.org> writes:
>>> 
>>>> I have no device for the wifi on my computer. 5.13 worked pretty well.
>>>> I also found others reporting similar issues:
>>>> https://forum.manjaro.org/t/kernel5-14-rc6-ath11k-on-qca6390-regression-mhi-probe-failure/79362
>>>>
>>>> Here's info about the hardware on my laptop:
>>>> ➜  ~ sudo dmidecode -s system-product-name
>>>> XPS 13 9310
>>>>
>>>> ➜  ~ sudo dmidecode -s bios-version
>>>> 3.0.4
>>> 
>>> Do you get the same mhi error as in the forum post?
>>> 
>>> qcom_mhi_qrtr: probe of mhi0_IPCR failed with error -22
>>> 
>>> MHI folks, any ideas? I have XPS 13 9310 myself but I'm not able to test
>>> v5.14-rc6 kernel right now.
>>
>> Hello,
>>
>> I have hit the same issue on Linux 5.14-rc7:
>>
>> [   14.780566] qcom_mhi_qrtr: probe of mhi0_IPCR failed with error -22
>>
>> This looks to be caused by commit:
>>
>> ce78ffa3ef16 net: really fix the build...
>>
>> Reverting this commit results in ath11k being able to load without errors.
>
> Thanks, this helps a lot. Commit ce78ffa3ef16 was introduced in
> v5.14-rc5 so I assume that's the first release which broke ath11k.
>
> Dave, what build issue did you fix with commit ce78ffa3ef16? Could we
> revert the commit? The problem is that, unless we fix this issue, with
> v5.14 WLAN will be broken on a quite few newer Dell laptops.

I did some investiation and I suspect that commit ce78ffa3ef16 ("net:
really fix the build...")[1] is for handling a conflict between net-next
and char-misc-next trees related to Loic's commit 0092a1e3f763 ("bus:
mhi: Add inbound buffers allocation flag"). Greg mentions this in a
commit[2]:

commit 813272ed5238b37c81e448b302048e8008570121
Merge: de0534df9347 36a21d51725a
Author:     Greg Kroah-Hartman <gregkh at linuxfoundation.org>

    Merge 5.14-rc5 into char-misc-next
    
    We need the fixes in here as well, and resolves some merge issues with
    the mhi codebase.

If my assumption is correct when I propose that we revert commit
ce78ffa3ef16 for v5.14, AFAICS it's not really needed for v5.14 (commit
0092a1e3f763 is in -next, not in Linus' tree yet) and commit
ce78ffa3ef16 breaks ath11k. And the conflict between net-next and
char-misc-next can be then later fixed during the merge window.

Will this work for everyone? If no objections, I'll submit the revert to
Linus later today. We are getting really close to final v5.14 release so
not much time left to fix this.

Just to reiterate why the urgency: commit ce78ffa3ef16 broke ath11k in
v5.14-rc5, users have reported (and I have confirmed) that at least
QCA6390 support is broken but I suspect all Qualcomm Wi-Fi 6 devices
supported by ath11k are currently broken.

[1] https://git.kernel.org/linus/ce78ffa3ef16

[2] https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc.git/commit/?h=char-misc-next&id=813272ed5238b37c81e448b302048e8008570121

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches