linux-kernel - Re: [PATCH] ASoC/HDA cs35l56: Remove redundant IRQ handling

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]

Message-ID: <6eb95724-9ba3-410f-a42e-e1013c449458@sirena.org.uk>
Date: Wed, 4 Sep 2024 18:27:10 +0100
From: Mark Brown <broonie@...nel.org>
To: Simon Trimmer <simont@...nsource.cirrus.com>
Cc: tiwai@...e.com, linux-sound@...r.kernel.org,
	alsa-devel@...a-project.org, linux-kernel@...r.kernel.org,
	patches@...nsource.cirrus.com
Subject: Re: [PATCH] ASoC/HDA cs35l56: Remove redundant IRQ handling

On Wed, Sep 04, 2024 at 02:46:30PM +0100, Simon Trimmer wrote:
> On Wed, Sep 04, 2024 at 12:25:00PM +0000, Mark Brown wrote:
> > On Wed, Sep 04, 2024 at 12:07:00PM +0000, Simon Trimmer wrote:

> > > The IRQ handling in the cs35l56 driver was purely informational. It was
> > > not necessary to support the HDA or ASoC driver functionality and added
> > > unnecessary complexity to the drivers.

> > Given that the code is there now and has been since the driver was
> > introduced about 18 months ago what's the ongoing cost of having it?
> > The information it's providing is notification of hardware faults,
> > reporting those does seem useful.

> Originally we were expecting to use the IRQ mechanism for an event logging
> stream that would function in a similar manner to compressed streams to be
> able to get an information feed for debug and tuning tools, but those were
> never created and the logging infrastructure not implemented.

Right.  Though ideally we might actually do something about some of
the errors that are reported.

> It's quite a spread of code and a lot of complexity in the regular execution
> paths managing them / synchronizing the contexts, there is more going on in
> the SoundWire bus variant compared to the conventional i2c/spi that it is
> hard to justify maintaining it all for a couple of log messages - in the
> event that someone did encounter the two situations being reported the
> regmap dump would point us to the cause pretty quickly.

I'm not sure how many end users are going to get as far as talking to
you in the event that they have issues - people often won't get as far
as trying to contact their distros or upstreams.  Even errors in dmesg
are pretty obscure but they're more discoverable than interpreting the
regmap, people would need to identify that they need to look at the chip
first and actually be experiencing the problem when they figure that
out.  Ideally we'd hae better handling for this (I did note that the
latest Iron Devices driver will back off speaker volume during a thermal
warning which isn't a terrible idea, though there's potential issues).

It sounds like the only real concern is the Soundwire stuff (I2C and SPI
interrupt stuff should generally be trivial?) - if that's the case I'd
be more inclined to only pull out the Soundwire bits and leave the
support there for the simpler buses.

Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)