linux-kernel - Re: [alsa-devel] [PATCH 1/3 v1] ALSA: usb-audio: more tolerant packetsize

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:   Thu, 1 Dec 2016 12:50:39 +0100
From:   Clemens Ladisch <clemens@...isch.de>
To:     Takashi Iwai <tiwai@...e.de>
Cc:     Jiada Wang <jiada_wang@...tor.com>, alsa-devel@...a-project.org,
        apape@...adit-jv.com, linux-kernel@...r.kernel.org,
        Mark_Craske@...tor.com
Subject: Re: [alsa-devel] [PATCH 1/3 v1] ALSA: usb-audio: more tolerant
 packetsize

Takashi Iwai wrote:
> Clemens Ladisch wrote:
>> Takashi Iwai wrote:
>>> [...]
>>> In the commit mentioned above, we changed the logic to take +25%
>>> frequency as the basis, and it my *reduce* if ep->maxpacksize is lower
>>> than that.
>>>
>>> OTOH, if ep->maxpacksize is sane, we can rely on it rather than the
>>> implicit +25% frequency.  That said, maybe we can check
>>> ep->maxpacksize whether it fits within the expected range, then adapt
>>> it, or take +25% freq as fallback?
>>
>> You are describing how the current code behaves.  The +25% limit _is_
>> what the code takes as the expected range.
>
> Well, the question is what is the "sane" range.  +25% doesn't fit for
> some devices.

The USB audio specification _requires_ that there is as little jitter
as possible.

It's no surprise that some device violates the specification.  But
we don't know what the actual error is; whether we could adjust the
packet size for this particular device only, or increase the limit
for all devices, or use a completely different workaround.

> If maxpacksize fits without +100% as this patch suggests, can we rely
> on it instead?

The packet size affect the following computations, like the number of
packets per URB.  I don't know how bad the effects would be.


Regards,
Clemens