lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Fri, 15 Oct 2021 11:22:58 -0500
From:   Pierre-Louis Bossart <pierre-louis.bossart@...ux.intel.com>
To:     Takashi Iwai <tiwai@...e.de>
Cc:     alsa-devel@...a-project.org,
        Kuninori Morimoto <kuninori.morimoto.gx@...esas.com>,
        open list <linux-kernel@...r.kernel.org>,
        Sameer Pujar <spujar@...dia.com>,
        Liam Girdwood <lgirdwood@...il.com>,
        Takashi Iwai <tiwai@...e.com>, vkoul@...nel.org,
        broonie@...nel.org, Gyeongtaek Lee <gt82.lee@...sung.com>,
        Peter Ujfalusi <peter.ujfalusi@...ux.intel.com>
Subject: Re: [RFC PATCH v3 05/13] ASoC: soc-pcm: align BE 'atomicity' with
 that of the FE


> The FE stream locks are necessary only two points: at adding and
> deleting the BE in the link.  We used dpcm_lock in other places, but
> those are superfluous or would make problem if converted to a FE
> stream lock.

I must be missing a fundamental concept here - possibly a set of concepts...

It is my understanding that the FE-BE connection can be updated
dynamically without any relationship to the usual ALSA steps, e.g. as a
result of a control being changed by a user.

So if you only protect the addition/removal, isn't there a case where
the for_each_dpcm_be() loop would either miss a BE or point to an
invalid one?

In other words, don't we need the *same* lock to be used
a) before changing and
b) walking through the list?

I also don't get what would happen if the dpcm_lock was converted to an
FE stream lock. It works fine in my tests, so if there's limitation I
didn't see it.

>>> In addition, a lock around dpcm_show_state() might be needed to be
>>> replaced with card->pcm_mutex, and we may need to revisit whether all
>>> other paths take card->pcm_mutex.
>>
>> What happens if we show the state while a trigger happens? That's my
>> main concern with using two separate locks (pcm_mutex and FE stream
>> lock) to protect the same list, there are still windows of time where
>> the list is not protected.
> 
> With the proper use of mutex, the list itself is protected.
> If we need to protect the concurrent access to each BE in the show
> method, an additional BE lock is needed in that part.  But that's a
> subtle issue, as the link traversal itself is protected by the mutex.

If I look at your patch2, dpcm_be_disconnect() protects the list removal
with the fe stream lock, but the show state is protected by both the
pcm_mutex and the fe stream lock.

I have not been able to figure out when you need
a) the pcm_mutex only
b) the fe stream lock only
c) both pcm_mutex and fe stream lock

Powered by blists - more mailing lists