Problems with QCA9377 SDIO on NXP i.MX8MPlus SOM

[Gary Bisson]

Hi Lukas & Kalle,

On Tue, Jun 13, 2023 at 11:05:14AM +0300, Kalle Valo wrote:
> 
> EXTERNAL EMAIL: Be careful with attachments and links.
> 
> "Lukas F. Hartmann" <lukas at mntre.com> writes:
> 
> > I'm trying to get the QCA9377 SDIO module on Boundary Devices'
> > i.MX8MPlus SOM working with the ath10k mainline driver (from Linux git
> > tip). Currently it can scan and see some APs and authenticate, but not
> > associate (timeout). Several months ago I thought this was an SDIO
> > signalling issue with the SoC, but I'm not sure anymore, because I fixed
> > up the qcacld driver for the current Linux kernel and it works fine on
> > the same system (i.e. I can rmmod ath10k_sdio and insmod wlan.ko and it
> > works).
> >
> > I wonder if it is some kind of firmware/driver mismatch. I pasted the
> > relevant dmesg output after my signature below.
> >
> > This is after I tried to build my own board-2.bin including bdwlan30.bin
> > from qcacld-2.0-CNSS.LEA.NRT_3.0 (source:
> > https://github.com/8devices/qcacld-2.0/tree/CNSS.LEA.NRT_3.0/firmware_bin/sdio).
> > It gets picked up but doesn't seem to make a difference. I also tried
> > assembling my own firmware-5-sdio.bin from qwlan30.bin and otp30.bin but
> > this triggers a firmware crash (with register dump) when loading
> > ath10k_sdio.ko.

As far as I know, the ath10k driver only supports (deprecated) LEH
firmware [1]. That is information I got a long time ago and I can't seem
to find the email back. But basically what I was told is that the LEH &
LEA firmwares had different frames format to exchange data between the
host and the chip, therefore drivers & firmware have to match.
And at that time I was told that ath10k only supported the LEH firmware
for SDIO, which is why we haven't invested much time on it as LEH
firmware is 1) old/deprecated as qcom hasn't updated it for 6 years now
and 2) it contains bugs affecting the performance in our case (diversity
not working etc).

Since I don't have the insight info about the LEA vs. LEH firmware
behavior I'm afraid I can't help much but at least I would say that
trying LEA firmware on ath10k driver is expected to fail (timeout).

Maybe Kalle have more info on the matter and maybe the above doesn't
apply any more and ath10k has evolved to support LEA firmware.

> > The qcacld driver works with that firmware, though.
> >
> > Any pointers on how to debug this are most appreciated.
> 
> Trying significantly older ath10k from an older kernel is something to
> try in case there are recent regressions in ath10k.
> 
> I don't have an ath10k SDIO device myself so I don't know the state of
> SDIO support in ath10k. Does anyone else from the list know?

FYI we have a branch in our repo with the firmware that worked on ath10k
at some point [2].

Regards,
Gary

[1] https://github.com/boundarydevices/qca-firmware/tree/bd-sdmac-qcacld
[2] https://github.com/boundarydevices/qca-firmware/tree/bd-sdmac-ath10k