linux-kernel - Re: [PATCH] soundwire: fix initializing sysfs for same devices on different buses

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:   Wed, 4 Oct 2023 15:00:40 -0400
From:   Pierre-Louis Bossart <pierre-louis.bossart@...ux.intel.com>
To:     Mark Brown <broonie@...nel.org>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>,
        Vinod Koul <vkoul@...nel.org>,
        Bard Liao <yung-chuan.liao@...ux.intel.com>,
        Sanyog Kale <sanyog.r.kale@...el.com>,
        Philippe Ombredanne <pombredanne@...b.com>,
        Takashi Iwai <tiwai@...e.de>, alsa-devel@...a-project.org,
        linux-kernel@...r.kernel.org, stable@...r.kernel.org
Subject: Re: [PATCH] soundwire: fix initializing sysfs for same devices on
 different buses

On 10/4/23 11:40, Mark Brown wrote:
> On Wed, Oct 04, 2023 at 11:16:09AM -0400, Pierre-Louis Bossart wrote:
> 
>>> matching the name is fine (if you are matching it against an existing
>>> name) but expecting the name to be anything specific is not going to
>>> work as the name is dynamic and can/will change each boot.
> 
>> Not following, sorry.
> 
>> In the SoundWire context, the device name directly follows the ACPI or
>> Device Tree information, I don't really see how its name could change on
>> each boot (assuming no DSDT override or overlays of course). The
>> platform descriptors are pretty much fixed, aren't they?
> 
>> Intel and AMD make such assumptions on names for pretty much all machine
>> drivers, it's not really something new - probably 15+ years? Adding Mark
>> Brown in CC: to make sure he's aware of this thread.
> 
> FWIW DT is much less affected here since all the inter-device references
> are explicit in the DT (modulo needing to work around breakage) so we're
> not hard coding in the way ACPI so unfortunately requires.

Isn't there a contradiction between making "all inter-device references
explicit in the DT" and having a device name use an IDA, which cannot
possibly known ahead of time?

I think we keep circling on the differences between "Controller" and
"link" (aka bus). A Controller can have one or more links. A system can
have one or more controllers.

Intel platforms have one controller and 4 or more links.
QCOM platforms have one or more controllers with one link each.

I am not sure how this IDA-generated bus_id helps deal with these two
cases, since we can't really make any assumptions on how
controllers/links will be started and probed.

What we are missing is a hierarchical controller/link definition, IOW a
controller_id should be given to the master by a higher level instead of
using an IDA.