[alsa-devel] Uninitialized id returned by saffirepro_both_clk_src_get.

[Takashi Sakamoto]

Hi Chris,

Thanks for your posting to alsa-devel. And sorry to include any bugs in
snd-bebob.

I reviewed your patch and realized that your patch cannot be applied to
current tree. The patch seems not to be based on the latest changes:

[alsa-devel] [PATCH 07/43] ALSA: bebob: More constify text arrays
http://mailman.alsa-project.org/pipermail/alsa-devel/2014-October/082580.html

[alsa-devel] [PATCH 08/43] ALSA: bebob: Use snd_ctl_enum_info()
http://mailman.alsa-project.org/pipermail/alsa-devel/2014-October/082574.html

I'll post revised patches later, just for our convenience.


Thanks

Takashi Sakamoto
o-takashi at sakamocchi.jp

On Oct 25 2014 20:40, Christian Vogel wrote:
> Hi,
> 
> there's a possibility to get a Oops caused by an uninitialized value
> in snd_bebob_stream_check_internal_clock for a SaffirePro running on the
> internal clock.
> 
> 	[   88.100531] BUG: unable to handle kernel paging request at 8a3c85fc          
> 	[   88.103808] IP: [<e8553aa0>] snd_bebob_stream_check_internal_clock+0x66/0x11e [snd_bebob]
> 
> ...which is dereferencing of clk_spec->labels[id] in...
> 
> 	sound/firewire/bebob/bebob_stream.c :
> 
> 	/* 1.The device has its own operation to switch source of clock */
> 	if (clk_spec) {
> 		err = clk_spec->get(bebob, &id);
> 		if (err < 0)
> 			dev_err(&bebob->unit->device,
> 				"fail to get clock source: %d\n", err);
> -->		else if (strncmp(clk_spec->labels[id], SND_BEBOB_CLOCK_INTERNAL,
> 				 strlen(SND_BEBOB_CLOCK_INTERNAL)) == 0)
> 			*internal = true;
> 		goto end;
> 	}
> 
> id is uninitialized, and will not be set by clk_spec->get (which is
> saffirepro_both_clk_src_get(), even if it returns ok(0).
> 
> Attached patch tries to clean up the logic in saffirepro_both_clk_src_get()
> and also adds a safety check to snd_bebob_stream_check_internal_clock().
> 
> Thanks for Takashi Sakamoto to whom I sent the patch initially and who
> suggested some cleanup to my code, reviewed the patch and suggested I send
> it to alsa-dev.