[Buildroot] Firmware auto-loading does not work

[Guillaume GARDET - Oliséo]

Le 25/01/2017 à 14:52, Peter Korsgaard a écrit :
>>>>>> "Guillaume" == Guillaume GARDET <- Oliséo <guillaume.gardet at oliseo.fr>> writes:
>   > Le 25/01/2017 à 13:51, Peter Korsgaard a écrit :
>   >>>>>>> "Guillaume" == Guillaume GARDET <- Oliséo <guillaume.gardet at oliseo.fr>> writes:
>   >> > Hello,
>   >> > I am using buildroot 2016.05 (with a 2.6.37 kernel) and firmware
>   >> > auto-loading does not work. I need to do it manually (through
>   >> > sysfs). I tried mdev and eudev, both are not working.
>   >>
>   >> > Is there a known bug about that? In busybox, maybe?
>   >>
>   >> If you have tried both eudev and mdev, then it is unlikely to be a bug
>   >> in busybox (mdev). Are you using busybox modprobe applet or kmod?
>
>   > This is not a problem of module loading (it is working fine) but
>   > loading _firmware_ to devices for the WiFi dongle in my case.
>
> Heh, I wonder how I could misread your email so much - Sorry about that.
>
> First of all, our default busybox configuration has
> CONFIG_FEATURE_MDEV_LOAD_FIRMWARE disabled. Have you enabled it? Without
> that option, mdev doesn't do anything about firmware.

I though I checked that it was enabled but it seems it isn't.

>
> Less ancient (since 3.7, so 2012) Linux kernels directly loads firmware
> files from the kernel, so the support in mdev isn't needed, but as you
> are using an ancient kernel this doesn't work for you:
>
> commit abb139e75c2cdbb955e840d6331cb5863e409d0e
> Author: Linus Torvalds <torvalds at linux-foundation.org>
> Date:   Wed Oct 3 15:58:32 2012 -0700
>
>      firmware: teach the kernel to load firmware files directly from the filesystem
>
>      This is a first step in allowing people to by-pass udev for loading
>      device firmware.  Current versions of udev will deadlock (causing us to
>      block for the 30 second timeout) under some circumstances if the
>      firmware is loaded as part of the module initialization path, and this
>      is causing problems for media drivers in particular.
>
>      The current patch hardcodes the firmware path that udev uses by default,
>      and will fall back to the legacy udev mode if the firmware cannot be
>      found there.  We'd like to add support for both configuring the paths
>      and the fallback behaviour, but in the meantime this hopefully fixes the
>      immediate problem, while also giving us a way forward.
>
>      [ v2: Some VFS layer interface cleanups suggested by Al Viro ]
>      [ v3: use the default udev paths suggested by Kay Sievers ]
>
>      Suggested-by: Ivan Kalvachev <ikalvachev at gmail.com>
>      Acked-by: Greg KH <gregkh at linuxfoundation.org>
>      Acked-by: Al Viro <viro at zeniv.linux.org.uk>
>      Cc: Mauro Carvalho Chehab <mchehab at redhat.com>
>      Cc: Kay Sievers <kay at redhat.com>
>      Cc: Ming Lei <ming.lei at canonical.com>
>      Signed-off-by: Linus Torvalds <torvalds at linux-foundation.org>
>
> Looking at eudev git history, I see that the legacy user space loading
> support got removed in 2014, so that is why it isn't working with eudev
> either:
>
> commit 3b717594600fa717cdf9bcfd0c7c1b703b245482
> Author: Kay Sievers <kay at vrfy.org>
> Date:   Mon Sep 1 09:07:49 2014 -0400
>
>     udev: remove userspace firmware loading support
>
>     Signed-off-by: Anthony G. Basile <blueness at gentoo.org>

Thanks for this history.

>
> We should probably enable CONFIG_FEATURE_MDEV_LOAD_FIRMWARE in our
> default busybox.config to better support these old systems.
>
> Can you confirm that it works if you enable that option.
>

As soon as I get back the problematic hardware, I check that and come back to you.

Guillaume