Message-Id: <20230705123018.30903-1-johan+linaro@kernel.org>
Date:   Wed,  5 Jul 2023 14:30:10 +0200
From:   Johan Hovold <johan+linaro@...nel.org>
To:     Mark Brown <broonie@...nel.org>, Vinod Koul <vkoul@...nel.org>
Cc:     Bard Liao <yung-chuan.liao@...ux.intel.com>,
        Pierre-Louis Bossart <pierre-louis.bossart@...ux.intel.com>,
        Sanyog Kale <sanyog.r.kale@...el.com>,
        Srinivas Kandagatla <srinivas.kandagatla@...aro.org>,
        Banajit Goswami <bgoswami@...cinc.com>,
        Liam Girdwood <lgirdwood@...il.com>,
        Jaroslav Kysela <perex@...ex.cz>,
        Takashi Iwai <tiwai@...e.com>, alsa-devel@...a-project.org,
        linux-kernel@...r.kernel.org,
        Johan Hovold <johan+linaro@...nel.org>
Subject: [PATCH 0/8] ASoC/soundwire/qdsp6/wcd: fix leaks and probe deferral

I've been hitting a race during boot which breaks probe of the sound
card on the Lenovo ThinkPad X13s as I've previously reported here:

	https://lore.kernel.org/all/ZIHMMFtuDtvdpFAZ@hovoldconsulting.com/

The immediate issue appeared to be a probe deferral that was turned into
a hard failure, but addressing that in itself only made things worse as
it exposed further bugs.

I was hoping someone more familiar with the code in question would look
into this, but as this affects users of the X13s and breaks audio on my
machine every fifth boot or so, I decided to investigate it myself.

As expected, the Qualcomm codec drivers are broken and specifically leak
resources on component remove, which in turn breaks sound card probe
deferrals.

The source of the deferral itself appears to be legitimate and was
simply due to some audio component not yet having been registered due to
random changes in timing during boot.

These issues can most easily be reproduced by simply blacklisting the
q6apm_dai module and loading it manually after boot.

The sound card probe deferral also exposes a bug in the soundwire
subsystem, which uses completion structures for signalling that a device
has been enumerated on the bus and initialised. The way this is
implemented prevents reprobed codec drivers from learning that the
soundwire devices are still attached, which causes probe to fail.

Included are also two patches that suppresses error messages on
component probe deferral to avoid spamming the logs during boot.

These patches should preferably all go through the ASoC tree even if
merging the soundwire fix separately also works.

Note the ASoC tree already has the following related fixes:

	https://lore.kernel.org/lkml/20230630120318.6571-1-johan+linaro@kernel.org/
	https://lore.kernel.org/lkml/20230630142717.5314-1-johan+linaro@kernel.org/
	https://lore.kernel.org/lkml/20230701094723.29379-1-johan+linaro@kernel.org/
	https://lore.kernel.org/lkml/20230703124701.11734-1-johan+linaro@kernel.org/

Johan


Johan Hovold (8):
  soundwire: fix enumeration completion
  ASoC: qdsp6: audioreach: fix topology probe deferral
  ASoC: codecs: wcd938x: fix missing clsh ctrl error handling
  ASoC: codecs: wcd938x: fix resource leaks on component remove
  ASoC: codecs: wcd934x: fix resource leaks on component remove
  ASoC: codecs: wcd-mbhc-v2: fix resource leaks on component remove
  ASoC: topology: suppress probe deferral errors
  ASoC: core: suppress probe deferral errors

 drivers/soundwire/bus.c         |  8 ++---
 sound/soc/codecs/wcd-mbhc-v2.c  | 57 ++++++++++++++++++++++---------
 sound/soc/codecs/wcd934x.c      | 12 +++++++
 sound/soc/codecs/wcd938x.c      | 59 +++++++++++++++++++++++++++++----
 sound/soc/qcom/qdsp6/topology.c |  4 +--
 sound/soc/soc-core.c            |  6 ++--
 sound/soc/soc-topology.c        | 10 ++++--
 7 files changed, 122 insertions(+), 34 deletions(-)

-- 
2.39.3

Powered by blists - more mailing lists