Re: patch "firmware loader: allow disabling of udev as firmware loader" added to driver-core tree

[Shuah Khan <shuahkhan@gmail.com>]

On Wed, Aug 6, 2014 at 3:44 AM, Jean-Michel Hautbois
<jhautbois@gmail.com> wrote:
> 2014-08-06 11:24 GMT+02:00 Takashi Iwai <tiwai@suse.de>:
>> At Wed, 6 Aug 2014 11:10:27 +0200,
>> Jean-Michel Hautbois wrote:
>>>
>>> 2014-08-05 21:22 GMT+02:00 Shuah Khan <shuahkhan@gmail.com>:
>>> > On Tue, Aug 5, 2014 at 10:01 AM, Takashi Iwai <tiwai@suse.de> wrote:
>>> >>> Yes, it worked, if the firmware was in the /lib/firmware directory.
>>> >>
>>> >
>>> > ok it works when the firmware is in /lib/firmware. It sounds to me the
>>> > reason load fails when the firmware is under /sys/class/firmware is
>>> > fw_load_from_user_helper() returns -ENOENT when
>>> > CONFIG_FW_LOADER_USER_HELPER is disabled.
>>> >
>>> > It would be nice to see the entire dmesg with debug enabled though.
>>> >
>>> > -- Shuah
>>>
>>> It does not work with the 3.16 kernel even when firmware is in
>>> /lib/firmware, it worked before this patch is applied.
>>
>> Hm?  3.16 doesn't contain my patch yet.  It's merged for 3.17-rc1.
>
> Oh, you are right of course, I am on upstream kernel and I have your
> patch. I don't mean your match is causing the issue though ;-).
>

I think this is what is going on and this patch is the cause:
fw_load_from_user_helper() is a stub that returns -ENOENT with this
patch. As a result, in _request_firmware() after fw_get_filesystem_firmware()
fails to find the file, fw_load_from_user_helper() gets called and it
returns right
away with -ENOENT.

In some cases if rootfs mount is in progress, fw_load_from_user_helper()
steps into load the firmware.

A quick way to test is by adding a dev_info() in the fw_load_from_user_helper()
and see if that is path it is taking.

-- Shuah