Force custom firmware for a mt7915e card (AW7915-NPD)

For Developers
1

I have been running into some problems getting a stable performance from my AsiaRF AW7915-NPD card, ever since kernel version 5.15 (which was the last working version for me).

That being said, this isn't a problem, specific to OpenWrt, but to any distro using kernel past 5.15.

I lack the skills and knowledge (and perhaps the tools) to diagnose what exactly is causing this but you can see my other thread (which at this point is a few years old).

I have tried downloading different eeprom but I have my doubts that anything in /lib/firmware/mediatek/ is actually used by the driver, or maybe the driver is using some other EEPROM, or maybe there is a race condition somewhere during the initialization because sometimes, like 1 out of 15 reboots, the card works, but when it works, it is actually very stable.

The rest of the reboots, I get random crashes, freezes, kernel panics within a minute or two after booting up to OpenWrt, sometimes it does not crash but the card still fails to work - all this happens more than 90% of the time.

I am still unsure what is causing this but as a part of my wider investigation, I have been somewhat fixated that it must be the driver not loading the correct eeprom because I have compared:

/sys/kernel/debug/ieee80211/phy0/mt76/eeprom

with

/lib/firmware/mediatek/mt7915_eeprom.bin

and they appear to be different. They differ even when the card boots properly.

I suspect there might be an eeprom on the card itself which the driver reads, meaning that any custom firmware in /lib/firmware/mediatek/ gets skipped.

My question is: should there be any difference in the eeproms, i.e. how does the mt76 driver work with mt7915e and most importantly, how do I guarantee that the correct eeprom gets loaded?

2

Rename file so it can not load to check?

3

Which file?

/sys/kernel/debug/ieee80211/phy0/mt76/eeprom
is showing the currently loaded and running firmware. If it is being taken from the card itself, I am not entirely certain where to look.

/lib/firmware/mediatek/mt7915_eeprom.bin
is the custom firmware, which comes with the kmod-mt7915-firmware package

If the running firmware is taken from /lib/firmware/mediatek/ , would it not be the same? The fact that it isn't makes me think it is being loaded from elsewhere.

4

Answer in ques%ion?

5

That comparison is not valid, /sys/kernel/debug/ieee80211/phy0/mt76/eeprom, is pointing to info in the running kernel. Its not a copy of a loaded file.

What platform are you using? I am running a similar card, AW7915-NP1, without problems in an 'Extreme Networks WS-AP3825i'.

2 Likes
6

I am using a x86-based platform.
If the comparison is not valid, what is the purpose of having downloaded the kmod-7915-firmware package? I am still struggling to understand how the firmware is being utilized and whether the card has onboard eeprom?

7

Why you keep throwing dirt and hope something sticks? Describe problem you are facing, with technical details.

8

Because I am not sure what exactly is causing it. I keep getting upon boot the following messages (they repeat until the whole system crashes)

[   35.798718] mt7915e 0000:04:00.0: Retry message 00001eed (seq 7)
[   56.224088] mt7915e 0000:04:00.0: Message 00001eed (seq 7) timeout

However every once in like 10-15 restarts, the wifi adapter actually gets properly initialized and works as intended without any hitch for days on end, until reboot. If I reboot it, I have to reboot it again like dozens of times before it starts functioning. As you can imagine, this issue cannot be consistently reproduced. I have tried different firmware, does not work.

I just don't know where to look. I found the following thread on github which describes the issue:

and it is said there the problem should have been more or less fixed in 24.10.1, however I am still getting the same old error even with 24.10.1.

The latest message there suggests there might be a need to apply 3 patches:
"The only thing missing is this patch or rather the series of 3 patches
https://patchwork.kernel.org/project/linux-wireless/patch/20250402004528.1036715-3-mail@david-bauer.net/"
However, I am unsure what I need to do because it is way outside the scope of what I have previously done.

As described there (and in the beginning of this thread), the problem began after kernel 5.15, and it is not OpenWrt-specific. It is a regression in the kernel.

1 Like
9

I am now trying to add some patches to mt76/mt7915/mcu.c as outlined here:

I have ran:

git clone https://git.openwrt.org/openwrt/openwrt.git
cd openwrt
git checkout v24.10.1

then checked the version for mt76 in makefile, then ran

git clone https://github.com/openwrt/mt76.git ~/custom-mt76
cd ~/custom-mt76
git checkout e5fef138524e63314cb96ff8314048d175294e95

after that, I made the necessary changes in mt76/mt7915/mcu.c, then edited the mt76 makefile in ~/openwrt so it points to the local source instead of using the repo.

However, before I go ahead and compile the image, I am worried that it will make my custom installation unable to install official packages from the repo due to hash mismatch.
Am I correct and if so, how could I avoid this?