2023-03-21 02:11:53

by Bard Liao

[permalink] [raw]
Subject: [PATCH] soundwire: intel: don't save hw_params for use in prepare

From: Pierre-Louis Bossart <[email protected]>

The existing code copies the hw_params pointer and reuses it later in
.prepare, specifically to re-initialize the ALH DMA channel
information that's lost in suspend-resume cycles.

This is not needed, we can directly access the information from the
substream/rtd - as done for the HDAudio DAIs in
sound/soc/sof/intel/hda-dai.c

In addition, using the saved pointer causes the suspend-resume test
cases to fail on specific platforms, depending on which version of GCC
is used. Péter Ujfalusi and I have spent long hours to root-cause this
problem that was reported by the Intel CI first with 6.2-rc1 and again
v6.3-rc1. In the latter case we were lucky that the problem was 100%
reproducible on local test devices, and found out that adding a
dev_dbg() or adding a call to usleep_range() just before accessing the
saved pointer "fixed" the issue. With errors appearing just by
changing the compiler version or minor changes in the code generated,
clearly we have a memory management Heisenbug.

The root-cause seems to be that the hw_params pointer is not
persistent. The soc-pcm code allocates the hw_params structure on the
stack, and passes it to the BE dailink hw_params and DAIs
hw_params. Saving such a pointer and reusing it later during the
.prepare stage cannot possibly work reliably, it's broken-by-design
since v5.10. It's astonishing that the problem was not seen earlier.

This simple fix will have to be back-ported to -stable, due to changes
to avoid the use of the get/set_dmadata routines this patch will only
apply on kernels older than v6.1.

Fixes: a5a0239c27fe ("soundwire: intel: reinitialize IP+DSP in .prepare(), but only when resuming")
Signed-off-by: Pierre-Louis Bossart <[email protected]>
Reviewed-by: Ranjani Sridharan <[email protected]>
Reviewed-by: Péter Ujfalusi <[email protected]>
Signed-off-by: Bard Liao <[email protected]>
---
drivers/soundwire/cadence_master.h | 2 --
drivers/soundwire/intel.c | 11 +++++++----
2 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/drivers/soundwire/cadence_master.h b/drivers/soundwire/cadence_master.h
index b653734085d9..27c56274217f 100644
--- a/drivers/soundwire/cadence_master.h
+++ b/drivers/soundwire/cadence_master.h
@@ -84,7 +84,6 @@ struct sdw_cdns_stream_config {
* @bus: Bus handle
* @stream_type: Stream type
* @link_id: Master link id
- * @hw_params: hw_params to be applied in .prepare step
* @suspended: status set when suspended, to be used in .prepare
* @paused: status set in .trigger, to be used in suspend
* @direction: stream direction
@@ -96,7 +95,6 @@ struct sdw_cdns_dai_runtime {
struct sdw_bus *bus;
enum sdw_stream_type stream_type;
int link_id;
- struct snd_pcm_hw_params *hw_params;
bool suspended;
bool paused;
int direction;
diff --git a/drivers/soundwire/intel.c b/drivers/soundwire/intel.c
index 046c67a2a39b..238acf5c97a9 100644
--- a/drivers/soundwire/intel.c
+++ b/drivers/soundwire/intel.c
@@ -725,7 +725,6 @@ static int intel_hw_params(struct snd_pcm_substream *substream,
dai_runtime->paused = false;
dai_runtime->suspended = false;
dai_runtime->pdi = pdi;
- dai_runtime->hw_params = params;

/* Inform DSP about PDI stream number */
ret = intel_params_stream(sdw, substream->stream, dai, params,
@@ -778,6 +777,11 @@ static int intel_prepare(struct snd_pcm_substream *substream,
}

if (dai_runtime->suspended) {
+ struct snd_soc_pcm_runtime *rtd = asoc_substream_to_rtd(substream);
+ struct snd_pcm_hw_params *hw_params;
+
+ hw_params = &rtd->dpcm[substream->stream].hw_params;
+
dai_runtime->suspended = false;

/*
@@ -789,7 +793,7 @@ static int intel_prepare(struct snd_pcm_substream *substream,
*/

/* configure stream */
- ch = params_channels(dai_runtime->hw_params);
+ ch = params_channels(hw_params);
if (substream->stream == SNDRV_PCM_STREAM_CAPTURE)
dir = SDW_DATA_DIR_RX;
else
@@ -801,7 +805,7 @@ static int intel_prepare(struct snd_pcm_substream *substream,

/* Inform DSP about PDI stream number */
ret = intel_params_stream(sdw, substream->stream, dai,
- dai_runtime->hw_params,
+ hw_params,
sdw->instance,
dai_runtime->pdi->intel_alh_id);
}
@@ -840,7 +844,6 @@ intel_hw_free(struct snd_pcm_substream *substream, struct snd_soc_dai *dai)
return ret;
}

- dai_runtime->hw_params = NULL;
dai_runtime->pdi = NULL;

return 0;
--
2.25.1



2023-04-12 10:07:18

by Vinod Koul

[permalink] [raw]
Subject: Re: [PATCH] soundwire: intel: don't save hw_params for use in prepare

On 21-03-23, 10:26, Bard Liao wrote:
> From: Pierre-Louis Bossart <[email protected]>
>
> The existing code copies the hw_params pointer and reuses it later in
> .prepare, specifically to re-initialize the ALH DMA channel
> information that's lost in suspend-resume cycles.
>
> This is not needed, we can directly access the information from the
> substream/rtd - as done for the HDAudio DAIs in
> sound/soc/sof/intel/hda-dai.c
>
> In addition, using the saved pointer causes the suspend-resume test
> cases to fail on specific platforms, depending on which version of GCC
> is used. P?ter Ujfalusi and I have spent long hours to root-cause this
> problem that was reported by the Intel CI first with 6.2-rc1 and again
> v6.3-rc1. In the latter case we were lucky that the problem was 100%
> reproducible on local test devices, and found out that adding a
> dev_dbg() or adding a call to usleep_range() just before accessing the
> saved pointer "fixed" the issue. With errors appearing just by
> changing the compiler version or minor changes in the code generated,
> clearly we have a memory management Heisenbug.
>
> The root-cause seems to be that the hw_params pointer is not
> persistent. The soc-pcm code allocates the hw_params structure on the
> stack, and passes it to the BE dailink hw_params and DAIs
> hw_params. Saving such a pointer and reusing it later during the
> .prepare stage cannot possibly work reliably, it's broken-by-design
> since v5.10. It's astonishing that the problem was not seen earlier.
>
> This simple fix will have to be back-ported to -stable, due to changes
> to avoid the use of the get/set_dmadata routines this patch will only
> apply on kernels older than v6.1.

Applied, thanks

--
~Vinod