linux-kernel - [PATCH 3/3] soundwire: intel: use pm_runtime

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>] [day] [month] [year] [list]

Message-Id: <20220621225641.221170-4-pierre-louis.bossart@linux.intel.com>
Date:   Tue, 21 Jun 2022 17:56:40 -0500
From:   Pierre-Louis Bossart <pierre-louis.bossart@...ux.intel.com>
To:     alsa-devel@...a-project.org
Cc:     tiwai@...e.de, broonie@...nel.org, vkoul@...nel.org,
        gregkh@...uxfoundation.org, srinivas.kandagatla@...aro.org,
        Bard liao <yung-chuan.liao@...ux.intel.com>,
        Pierre-Louis Bossart <pierre-louis.bossart@...ux.intel.com>,
        Rander Wang <rander.wang@...el.com>,
        Ranjani Sridharan <ranjani.sridharan@...ux.intel.com>,
        Péter Ujfalusi <peter.ujfalusi@...ux.intel.com>,
        Sanyog Kale <sanyog.r.kale@...el.com>,
        linux-kernel@...r.kernel.org (open list)
Subject: [PATCH 3/3] soundwire: intel: use pm_runtime_resume() on component probe

During the card registration, transactions on the SoundWire bus can be
initiated. If the ALSA card is registered after the bus suspends,
timeouts can be seen while reading/writing codec registers. This is
extremely easy to reproduce in driver bind/unbind tests.

In an initial experiment, the ASoC soc-component.c code was modified
to initiate a pm_runtime resume on a component probe. The results
showed this was too invasive. Instead this patch suggests resuming the
SoundWire component only.

Because of the parent-child hierarchy enforced by the pm_runtime
framework, it can be argued that the codec component probe should be
enough to resume all necessary devices, and indeed the same resume
will be applied to SoundWire codecs used on Intel platforms.

Calling pm_runtime_resume() on both the Intel and codec sides has the
benefit of resuming the bus without assuming any order during the card
registration. The first component on a dailink to be probed will
resume the bus. In addition, if a codec driver did not implement this
transition, the Intel component would still resume the bus and avoid
timeouts on card registration.

BugLink: https://github.com/thesofproject/linux/issues/3651
Reviewed-by: Rander Wang <rander.wang@...el.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@...ux.intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@...ux.intel.com>
Reviewed-by: Péter Ujfalusi <peter.ujfalusi@...ux.intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@...ux.intel.com>
---
 drivers/soundwire/intel.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/drivers/soundwire/intel.c b/drivers/soundwire/intel.c
index 2e7c27d303b42..c907bab12ee33 100644
--- a/drivers/soundwire/intel.c
+++ b/drivers/soundwire/intel.c
@@ -1051,6 +1051,23 @@ static int intel_trigger(struct snd_pcm_substream *substream, int cmd, struct sn
 	return ret;
 }

+static int intel_component_probe(struct snd_soc_component *component)
+{
+	int ret;
+
+	/*
+	 * make sure the device is pm_runtime_active before initiating
+	 * bus transactions during the card registration.
+	 * We use pm_runtime_resume() here, without taking a reference
+	 * and releasing it immediately.
+	 */
+	ret = pm_runtime_resume(component->dev);
+	if (ret < 0 && ret != -EACCES)
+		return ret;
+
+	return 0;
+}
+
 static int intel_component_dais_suspend(struct snd_soc_component *component)
 {
 	struct snd_soc_dai *dai;
@@ -1106,6 +1123,7 @@ static const struct snd_soc_dai_ops intel_pcm_dai_ops = {

 static const struct snd_soc_component_driver dai_component = {
 	.name           = "soundwire",
+	.probe		= intel_component_probe,
 	.suspend	= intel_component_dais_suspend
 };

-- 
2.34.1