Re: Is request_firmware() really safe to call in resume callback when /usr/lib/firmware is on btrfs?

[Luis Chamberlain <mcgrof@kernel.org>]

On Fri, Aug 14, 2020 at 11:59:36PM +0200, Lukas Middendorf wrote:
> On 14/08/2020 18:37, Luis Chamberlain wrote:
> > On Fri, Aug 14, 2020 at 01:38:40PM +0200, Lukas Middendorf wrote:
> > > > If the firmware is *not* present for the si2168 driver and the device
> > > > has *not* been used yet you get a system freeze which you cannot recover
> > > > from, but only if you are *not* using a driver which also caches its
> > > > firmware already?
> > > 
> > > Yes, this is exactly what I wanted to say.
> > > 
> > 
> > OK great.. but..
> > 
> > > A new installation of Fedora 32 without firmware files
> > 
> > Fedora 32 comes with no firmware at all? Are you sure? How about your
> > wifi?
> 
> Fedora does come with firmware files for many devices (wifi, nouveau, …) but
> not for dvb devices. Firmware for the si2168 has to be installed as an extra
> package dvb-firmware from rpmfusion. When I talk about "no firmware files"
> or "install the firmware files" I mean the si2168 (and other dvb hardware)
> firmware files only. The nouveau firmware files are always present.

OK so how do you know that other firmware is not getting loaded or cached?

Other than checking kernel logs you can rm -rf /lib/firmware/ and then
only place the files you want to test.

> > > and with nouveau did
> > > not show my freeze problem. Installing either the firmware files or the
> > > nvidia driver started the freeze during resume.
> > 
> > Here now you say that if you install either the firmware files for
> > either si2168 or nouveau can cause a freeze,
> 
> no, I'm talking just about the firmware files for si2168

OK so if you install the firmware files of si2168 you do run into a
freeze, and this freeze happens on resume from suspend?

Without the si2168 firmware the freeze does not happen.

Is that right?

> In the case that I *don't* have the firmware files for si2168 (!) present it
> happens only if no other firmware is cached on suspend (in my case by the
> nouveau driver). No statement made about cases where the si2168 firmware
> file is present.

OK this statement is clear and is very different from the one I made
above.

But note, that this is true, how are you *sure* that no other firmware
other than nouveau is being used? What about wifi? or bluetooth?

> > You also clarify here your freeze happens on resume only. Is that right?
> > Never on suspend, but if the freeze happens, it happens only on resume?
> 
> Correct. I have not seen a freeze on suspend. It only happens on resume.

OK this is only if and only if you haven't used the si2168 device,
right? And since this is related to si2168 we know that even if you
don't use the si2168 device its function which calls to load firmware
*does* get called on resume, even though that same function was not
called on probe, as the device is not used.

If true, then the race to freeze here happens on resume against btrfs.

And is the firmware present or not in this case, in which the freeze
happens?

> > The actual case where you reach a freeze is still not clear yet. Let's
> > try to clarify this.
> 
> OK, let's try that again. To freeze during resume all of 1-4 has to be true:
> 1. /usr/lib/firmware is on btrfs
> 2. my Hauppauge WinTV-dualHD USB DVB tuner (contains si2168) is connected
> 3. have not actively used the tuner
> 4. any of the following cases:
> 4a) si2168 firmware not installed + nouveau driver not used + have not run
> "ls -R /usr/lib/firmware" before suspend
> 4b) firmware file installed + not run "cat /usr/lib/firmware/dvb*"
> 4c) firmware file installed + not run "ls -R /usr/lib/firmware" + not
> nouveau driver

OK perfect, now the next question to clarify is *are you sure* that no
other firmware is used, other than si2168 and nouveau?

> Not leading to a freeze is:
> 
> A: si2168 firmware not installed + nouveau driver used
> B: si2168 firmware not installed + run "ls -R /usr/lib/firmware" before
> suspend
> C: used the tuner before suspend (or tried to use, in case that the si2168
> firmware is not installed)
> D: using my patches with firmware_request_cache()
> E: si2168 firmware installed + "ls -R /usr/lib/firmware" + "cat
> /usr/lib/firmware/dvb*"
> F: si2168 firmware installed + nouveau driver used + "cat
> /usr/lib/firmware/dvb*"
> 
> I verified all cases again to make sure I was not remembering anything
> wrong.

This helps, thanks so much, now we'll have to write a reproducer, thanks
for the report!!

> The nouveau driver in use seems to be equivalent to running "ls -R
> /usr/lib/firmware" before suspend.
> 
> All the cases seem to boil down to:
> It freezes if the file system has to be accessed to list the content of
> /usr/lib/firmware or to read the si2168 firmware file

Let's confirm first whether or not your system is using other firmware
files too or not.

  Luis