Re: [PATCH ath-next 1/3] wifi: ath10k: set the PMF peer flag on WCN3990

[Loic Poulain]

Hi Conn, Shunsuke,

On Tue, Nov 3, 2026 at 12:57 AM Conn O'Griofa <connogriofa@xxxxxxxxx> wrote:
>
> On 02/11/2026 22:51, Nagashima, Shunsuke wrote:
>
> > Aleksei, Conn: a firmware file update only helps devices that load that
> > file. If you have time, could you make the same change to the
> > firmware-N.bin your device loads, test it, and send it to
> > linux-firmware?
>
>
> I can confirm that the unmodified driver with a patched firmware fixes
> the issue on QCA9377 as well. I appended the mfp-support flag to the
> existing flags for both firmware variants that are relevant to my device:
>
> ./ath10k-fwencoder --modify --features=ignore-otp-result,mfp-support
> firmware-5.bin
> ./ath10k-fwencoder --modify
> --features=wowlan,ignore-otp-result,mfp-support firmware-6.bin
>
> Testing each firmware individually confirms the flag is now present, and
> both firmware versions can now reach 300Mbps.
>
> ath10k_pci 0000:02:00.0: firmware ver WLAN.TF.1.0-00002-QCATFSWPZ-5 api
> 5 features ignore-otp,mfpcrc32 18db2397
> ath10k_pci 0000:02:00.0: firmware ver WLAN.TF.2.1-00021-QCARMSWP-1 api 6
> features wowlan,ignore-otp,mfpcrc32 02bceae2
>
> I would be happy to submit the modified firmware, but I am skeptical
> that it would be accepted? I'm not sure if I've ever observed
> unofficial/modified wifi firmware to ever have been accepted into
> linux-firmware (which seems completely understandable, given the risks).

I think having this defined directly in firmware-5.bin makes more
sense, as the feature is primarily tied to a specific firmware
implementation rather than the driver itself. If PMF support was
introduced at some point in the WCN3990 firmware distributed through
linux-firmware, then advertising that capability through the
corresponding firmware-5.bin entry seems like the right place to do
it.

Compared to enabling it unconditionally in the driver, this also
prevents older firmware packages that may not implement the feature
from being incorrectly treated as PMF-capable. In other words, the PMF
capability should follow the firmware version, not the driver. As a
result, hardware using the same firmware image would naturally expose
the same PMF capability, while older firmware revisions would continue
to behave correctly without requiring any special handling in the
driver.

For change request I think the best for this is probably to follow
board like change:
https://wireless.docs.kernel.org/en/latest/en/users/drivers/ath10k/boardfiles.html

Please, keep me in CC.

Regards,
Loic