[PATCH] ASoC: ti: davinci-mcasp: Fix race condition during probe
From: Joao Paulo Goncalves joao.goncalves@toradex.com
When using davinci-mcasp as CPU DAI with simple-card, there are some conditions that cause simple-card to finish registering a sound card before davinci-mcasp finishes registering all sound components. This creates a non-working sound card from userspace with no problem indication apart from not being able to play/record audio on a PCM stream. The issue arises during simultaneous probe execution of both drivers. Specifically, the simple-card driver, awaiting a CPU DAI, proceeds as soon as davinci-mcasp registers its DAI. However, this process can lead to the client mutex lock (client_mutex in soc-core.c) being held or davinci-mcasp being preempted before PCM DMA registration on davinci-mcasp finishes. This situation occurs when the probes of both drivers run concurrently. Below is the code path for this condition. To solve the issue, defer davinci-mcasp CPU DAI registration to the last step in the audio part of it. This way, simple-card CPU DAI parsing will be deferred until all audio components are registered.
Fail Code Path:
simple-card.c: probe starts simple-card.c: simple_dai_link_of: simple_parse_node(..,cpu,..) returns EPROBE_DEFER, no CPU DAI yet davinci-mcasp.c: probe starts davinci-mcasp.c: devm_snd_soc_register_component() register CPU DAI simple-card.c: probes again, finish CPU DAI parsing and call devm_snd_soc_register_card() simple-card.c: finish probe davinci-mcasp.c: *dma_pcm_platform_register() register PCM DMA davinci-mcasp.c: probe finish
Cc: stable@vger.kernel.org Fixes: 9fbd58cf4ab0 ("ASoC: davinci-mcasp: Choose PCM driver based on configured DMA controller") Signed-off-by: Joao Paulo Goncalves joao.goncalves@toradex.com --- sound/soc/ti/davinci-mcasp.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/sound/soc/ti/davinci-mcasp.c b/sound/soc/ti/davinci-mcasp.c index b892d66f78470..1e760c3155213 100644 --- a/sound/soc/ti/davinci-mcasp.c +++ b/sound/soc/ti/davinci-mcasp.c @@ -2417,12 +2417,6 @@ static int davinci_mcasp_probe(struct platform_device *pdev)
mcasp_reparent_fck(pdev);
- ret = devm_snd_soc_register_component(&pdev->dev, &davinci_mcasp_component, - &davinci_mcasp_dai[mcasp->op_mode], 1); - - if (ret != 0) - goto err; - ret = davinci_mcasp_get_dma_type(mcasp); switch (ret) { case PCM_EDMA: @@ -2449,6 +2443,12 @@ static int davinci_mcasp_probe(struct platform_device *pdev) goto err; }
+ ret = devm_snd_soc_register_component(&pdev->dev, &davinci_mcasp_component, + &davinci_mcasp_dai[mcasp->op_mode], 1); + + if (ret != 0) + goto err; + no_audio: ret = davinci_mcasp_init_gpiochip(mcasp); if (ret) {
Hi Joao,
On Apr 17, 2024 at 15:41:38 -0300, Joao Paulo Goncalves wrote:
From: Joao Paulo Goncalves joao.goncalves@toradex.com
When using davinci-mcasp as CPU DAI with simple-card, there are some conditions that cause simple-card to finish registering a sound card before davinci-mcasp finishes registering all sound components. This creates a non-working sound card from userspace with no problem indication apart from not being able to play/record audio on a PCM stream. The issue arises during simultaneous probe execution of both drivers. Specifically, the simple-card driver, awaiting a CPU DAI, proceeds as soon as davinci-mcasp registers its DAI. However, this process can lead to the client mutex lock (client_mutex in soc-core.c) being held or davinci-mcasp being preempted before PCM DMA registration on davinci-mcasp finishes. This situation occurs when the probes of both drivers run concurrently. Below is the code path for this condition. To solve the issue, defer davinci-mcasp CPU DAI registration to the last step in the audio part of it. This way, simple-card CPU DAI parsing will be deferred until all audio components are registered.
Good catch, thanks for the fix.
Reviewed-by: Jai Luthra j-luthra@ti.com
Fail Code Path:
simple-card.c: probe starts simple-card.c: simple_dai_link_of: simple_parse_node(..,cpu,..) returns EPROBE_DEFER, no CPU DAI yet davinci-mcasp.c: probe starts davinci-mcasp.c: devm_snd_soc_register_component() register CPU DAI simple-card.c: probes again, finish CPU DAI parsing and call devm_snd_soc_register_card() simple-card.c: finish probe davinci-mcasp.c: *dma_pcm_platform_register() register PCM DMA davinci-mcasp.c: probe finish
Cc: stable@vger.kernel.org Fixes: 9fbd58cf4ab0 ("ASoC: davinci-mcasp: Choose PCM driver based on configured DMA controller") Signed-off-by: Joao Paulo Goncalves joao.goncalves@toradex.com
sound/soc/ti/davinci-mcasp.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/sound/soc/ti/davinci-mcasp.c b/sound/soc/ti/davinci-mcasp.c index b892d66f78470..1e760c3155213 100644 --- a/sound/soc/ti/davinci-mcasp.c +++ b/sound/soc/ti/davinci-mcasp.c @@ -2417,12 +2417,6 @@ static int davinci_mcasp_probe(struct platform_device *pdev)
mcasp_reparent_fck(pdev);
- ret = devm_snd_soc_register_component(&pdev->dev, &davinci_mcasp_component,
&davinci_mcasp_dai[mcasp->op_mode], 1);- if (ret != 0)
goto err;- ret = davinci_mcasp_get_dma_type(mcasp); switch (ret) { case PCM_EDMA:
@@ -2449,6 +2443,12 @@ static int davinci_mcasp_probe(struct platform_device *pdev) goto err; }
- ret = devm_snd_soc_register_component(&pdev->dev, &davinci_mcasp_component,
&davinci_mcasp_dai[mcasp->op_mode], 1);- if (ret != 0)
goto err;no_audio: ret = davinci_mcasp_init_gpiochip(mcasp); if (ret) { -- 2.34.1
On 17/04/2024 21:41, Joao Paulo Goncalves wrote:
From: Joao Paulo Goncalves joao.goncalves@toradex.com
When using davinci-mcasp as CPU DAI with simple-card, there are some conditions that cause simple-card to finish registering a sound card before davinci-mcasp finishes registering all sound components. This creates a non-working sound card from userspace with no problem indication apart from not being able to play/record audio on a PCM stream. The issue arises during simultaneous probe execution of both drivers. Specifically, the simple-card driver, awaiting a CPU DAI, proceeds as soon as davinci-mcasp registers its DAI. However, this process can lead to the client mutex lock (client_mutex in soc-core.c) being held or davinci-mcasp being preempted before PCM DMA registration on davinci-mcasp finishes. This situation occurs when the probes of both drivers run concurrently. Below is the code path for this condition. To solve the issue, defer davinci-mcasp CPU DAI registration to the last step in the audio part of it. This way, simple-card CPU DAI parsing will be deferred until all audio components are registered.
Fail Code Path:
simple-card.c: probe starts simple-card.c: simple_dai_link_of: simple_parse_node(..,cpu,..) returns EPROBE_DEFER, no CPU DAI yet davinci-mcasp.c: probe starts davinci-mcasp.c: devm_snd_soc_register_component() register CPU DAI simple-card.c: probes again, finish CPU DAI parsing and call devm_snd_soc_register_card() simple-card.c: finish probe davinci-mcasp.c: *dma_pcm_platform_register() register PCM DMA davinci-mcasp.c: probe finish
Interesting... Thanks for the details. Acked-by: Peter Ujfalusi peter.ujfalusi@gmail.com
Cc: stable@vger.kernel.org Fixes: 9fbd58cf4ab0 ("ASoC: davinci-mcasp: Choose PCM driver based on configured DMA controller")
Just to note that the DAI registration was always before the platform registration (ever since the DAI driver started to register the platform) and I think most TI (and probably other vendor's) driver does things this way. McASP does a bit of lifting by requesting a DMA channel to figure out the type of DMA...
Signed-off-by: Joao Paulo Goncalves joao.goncalves@toradex.com
sound/soc/ti/davinci-mcasp.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/sound/soc/ti/davinci-mcasp.c b/sound/soc/ti/davinci-mcasp.c index b892d66f78470..1e760c3155213 100644 --- a/sound/soc/ti/davinci-mcasp.c +++ b/sound/soc/ti/davinci-mcasp.c @@ -2417,12 +2417,6 @@ static int davinci_mcasp_probe(struct platform_device *pdev)
mcasp_reparent_fck(pdev);
- ret = devm_snd_soc_register_component(&pdev->dev, &davinci_mcasp_component,
&davinci_mcasp_dai[mcasp->op_mode], 1);- if (ret != 0)
goto err;- ret = davinci_mcasp_get_dma_type(mcasp); switch (ret) { case PCM_EDMA:
@@ -2449,6 +2443,12 @@ static int davinci_mcasp_probe(struct platform_device *pdev) goto err; }
- ret = devm_snd_soc_register_component(&pdev->dev, &davinci_mcasp_component,
&davinci_mcasp_dai[mcasp->op_mode], 1);- if (ret != 0)
goto err;no_audio: ret = davinci_mcasp_init_gpiochip(mcasp); if (ret) {
On Wed, 17 Apr 2024 15:41:38 -0300, Joao Paulo Goncalves wrote:
When using davinci-mcasp as CPU DAI with simple-card, there are some conditions that cause simple-card to finish registering a sound card before davinci-mcasp finishes registering all sound components. This creates a non-working sound card from userspace with no problem indication apart from not being able to play/record audio on a PCM stream. The issue arises during simultaneous probe execution of both drivers. Specifically, the simple-card driver, awaiting a CPU DAI, proceeds as soon as davinci-mcasp registers its DAI. However, this process can lead to the client mutex lock (client_mutex in soc-core.c) being held or davinci-mcasp being preempted before PCM DMA registration on davinci-mcasp finishes. This situation occurs when the probes of both drivers run concurrently. Below is the code path for this condition. To solve the issue, defer davinci-mcasp CPU DAI registration to the last step in the audio part of it. This way, simple-card CPU DAI parsing will be deferred until all audio components are registered.
[...]
Applied to
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git for-next
Thanks!
[1/1] ASoC: ti: davinci-mcasp: Fix race condition during probe commit: d18ca8635db2f88c17acbdf6412f26d4f6aff414
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks, Mark
participants (4)
-
Jai Luthra -
Joao Paulo Goncalves -
Mark Brown -
Péter Ujfalusi