On 6/10/22 00:25, Greg KH wrote:

On Fri, Jun 10, 2022 at 10:35:36AM +0800, Bard Liao wrote:

...

From: Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com

The same debug message is replicated multiple time, add __func__ to figure out what link is ignored.

Signed-off-by: Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com Reviewed-by: Ranjani Sridharan ranjani.sridharan@linux.intel.com Reviewed-by: Péter Ujfalusi peter.ujfalusi@linux.intel.com Signed-off-by: Bard Liao yung-chuan.liao@linux.intel.com

---

drivers/soundwire/intel.c | 28 ++++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-)

diff --git a/drivers/soundwire/intel.c b/drivers/soundwire/intel.c index 505c5ef061e3..808e2f320052 100644 --- a/drivers/soundwire/intel.c +++ b/drivers/soundwire/intel.c @@ -1328,8 +1328,8 @@ int intel_link_startup(struct auxiliary_device *auxdev)

if (bus->prop.hw_disabled) { dev_info(dev,

• 	 "SoundWire master %d is disabled, ignoring\n",
  

• 	 sdw->instance);
  

• 	 "%s: SoundWire master %d is disabled, ignoring\n",
  

• 	 __func__, sdw->instance);
  

This is not a debug message, please make it such if you want to have __func__ And even then, it's not needed as you can get that from the kernel automatically.

Sorry, I don't understand the feedback at all.

This message was added precisely to figure out why the expected programming sequence was not followed, only to discover that we have devices with spurious PCI wakes handled below. Without this added difference with __func__, we wouldn't know if the issue happened during the expected/regular programming sequence or not.

...

return 0;

}

@@ -1489,8 +1489,8 @@ int intel_link_process_wakeen_event(struct auxiliary_device *auxdev) bus = &sdw->cdns.bus;

if (bus->prop.hw_disabled || !sdw->startup_done) {

• dev_dbg(dev, "SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	bus->link_id);
  

• dev_dbg(dev, "%s: SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	__func__, bus->link_id);
  

  return 0; }

@@ -1549,8 +1549,8 @@ static int __maybe_unused intel_pm_prepare(struct device *dev) int ret;

if (bus->prop.hw_disabled || !sdw->startup_done) {

• dev_dbg(dev, "SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	bus->link_id);
  

• dev_dbg(dev, "%s: SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	__func__, bus->link_id);
  

Not needed, it is provided automatically if you ask the kernel for this. Same for all other instances in this patch.

provided how? Your comment is a bit cryptic here.

...

...

...

diff --git a/drivers/soundwire/intel.c b/drivers/soundwire/intel.c index 505c5ef061e3..808e2f320052 100644 --- a/drivers/soundwire/intel.c +++ b/drivers/soundwire/intel.c @@ -1328,8 +1328,8 @@ int intel_link_startup(struct auxiliary_device *auxdev)

if (bus->prop.hw_disabled) { dev_info(dev,

• 	 "SoundWire master %d is disabled, ignoring\n",
  

• 	 sdw->instance);
  

• 	 "%s: SoundWire master %d is disabled, ignoring\n",
  

• 	 __func__, sdw->instance);
  

This is not a debug message, please make it such if you want to have __func__ And even then, it's not needed as you can get that from the kernel automatically.

Sorry, I don't understand the feedback at all.

dev_info() is not a way to send debug messages.

If you want this to be only for debugging, use dev_dbg(). And when you use that, you get the __func__ location for free in the output already if you want that.

...

This message was added precisely to figure out why the expected programming sequence was not followed, only to discover that we have devices with spurious PCI wakes handled below. Without this added difference with __func__, we wouldn't know if the issue happened during the expected/regular programming sequence or not.

Perhaps make the text unique then? Why would an informational message need a function name. Drivers should be quiet when all is going well. If something is not going well, dev_info() is not the kernel log level to be sending something out at.

This is not a debug message - unlike all the others. The BIOS and/or the user may disable a specific link and it's useful to know.

The intent of __func__ was precisely to make the text unique across all cases where we test for hw_disabled, but you have a point that this specific info message should be clearer. The text was actually already different from the 'disable or not-started' cases, I might have done one too many edits in this file. Will fix.

...

...

...

return 0;

}

@@ -1489,8 +1489,8 @@ int intel_link_process_wakeen_event(struct auxiliary_device *auxdev) bus = &sdw->cdns.bus;

if (bus->prop.hw_disabled || !sdw->startup_done) {

• dev_dbg(dev, "SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	bus->link_id);
  

• dev_dbg(dev, "%s: SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	__func__, bus->link_id);
  

  return 0; }

@@ -1549,8 +1549,8 @@ static int __maybe_unused intel_pm_prepare(struct device *dev) int ret;

if (bus->prop.hw_disabled || !sdw->startup_done) {

• dev_dbg(dev, "SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	bus->link_id);
  

• dev_dbg(dev, "%s: SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	__func__, bus->link_id);
  

Not needed, it is provided automatically if you ask the kernel for this. Same for all other instances in this patch.

provided how? Your comment is a bit cryptic here.

the dynamic debug code in the kernel already adds the function name where the message was sent from, if you want to know this in userspace. Please read the documentation for details (I think the key is the 'f' flag to be enabled in userspace).

So adding __func__ to dev_dbg() calls are redundant and never needed.

Oh wow, I had no idea - and I am not the only one :-)

Here's what I get with a grep on the kernel code

git grep __func__ | grep dev_dbg | wc -l

3122

Doing this for sound/ gives me 356, and that's for single-line logs only.

There's clearly something that was missed by tons of people.

We use an 'sof-dyndbg.conf' file that we provide to users and our CI to log what's needed for SOF debug, it's trivial to add a +f option.

options foo dyndbg=+pf

Thanks for the feedback, much appreciated. -Pierre

...

...

...

...

...

@@ -1549,8 +1549,8 @@ static int __maybe_unused intel_pm_prepare(struct device *dev) int ret;

if (bus->prop.hw_disabled || !sdw->startup_done) {

• dev_dbg(dev, "SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	bus->link_id);
  

• dev_dbg(dev, "%s: SoundWire master %d is disabled or not-started, ignoring\n",
  

• 	__func__, bus->link_id);
  

Not needed, it is provided automatically if you ask the kernel for this. Same for all other instances in this patch.

provided how? Your comment is a bit cryptic here.

the dynamic debug code in the kernel already adds the function name where the message was sent from, if you want to know this in userspace. Please read the documentation for details (I think the key is the 'f' flag to be enabled in userspace).

So adding __func__ to dev_dbg() calls are redundant and never needed.

Oh wow, I had no idea - and I am not the only one :-)

Here's what I get with a grep on the kernel code

git grep __func__ | grep dev_dbg | wc -l

3122

Doing this for sound/ gives me 356, and that's for single-line logs only.

There's clearly something that was missed by tons of people.

Yes, it's not commonly known so cleanups are always welcome. Perhaps some of those usages predate the feature being added as well.

Yep, looks like no one in our team knew about this. We've changed our scripts and I started cleaning this up, changes will be posted upstream soon.

https://github.com/thesofproject/linux/pull/3695

This patch will be reverted in our tree, completely unnecessary indeed.

On Fri, Jun 10, 2022 at 10:35:37AM +0800, Bard Liao wrote:

When a pipeline is split into FE and BE parts, the BE pipeline may need to be triggered separately in the BE trigger op. So add the trigger callback in the link_res ops that will be invoked during BE DAI trigger.

Signed-off-by: Bard Liao yung-chuan.liao@linux.intel.com Reviewed-by: Rander Wang rander.wang@intel.com Reviewed-by: Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com Reviewed-by: Ranjani Sridharan ranjani.sridharan@linux.intel.com

---

drivers/soundwire/intel.c | 8 ++++++++ include/linux/soundwire/sdw_intel.h | 1 + 2 files changed, 9 insertions(+)

diff --git a/drivers/soundwire/intel.c b/drivers/soundwire/intel.c index 808e2f320052..3f3c2c99cb8e 100644 --- a/drivers/soundwire/intel.c +++ b/drivers/soundwire/intel.c @@ -1004,9 +1004,17 @@ static int intel_trigger(struct snd_pcm_substream *substream, int cmd, struct sn { struct sdw_cdns *cdns = snd_soc_dai_get_drvdata(dai); struct sdw_intel *sdw = cdns_to_intel(cdns);

• struct sdw_intel_link_res *res = sdw->link_res; struct sdw_cdns_dma_data *dma; int ret = 0;

• /* The .trigger callback is used to send required IPC to audio

• * firmware. The .free_stream callback will still be called
  

• * by intel_free_stream() in the TRIGGER_SUSPEND case.
  

• */
  

• if (res->ops && res->ops->trigger)

• res->ops->trigger(dai, cmd, substream->stream);
  

• dma = snd_soc_dai_get_dma_data(dai, substream); if (!dma) { dev_err(dai->dev, "failed to get dma data in %s\n",

diff --git a/include/linux/soundwire/sdw_intel.h b/include/linux/soundwire/sdw_intel.h index 67e0d3e750b5..f638707fd712 100644 --- a/include/linux/soundwire/sdw_intel.h +++ b/include/linux/soundwire/sdw_intel.h @@ -119,6 +119,7 @@ struct sdw_intel_ops { struct sdw_intel_stream_params_data *params_data); int (*free_stream)(struct device *dev, struct sdw_intel_stream_free_data *free_data);

• int (*trigger)(struct snd_soc_dai *dai, int cmd, int stream);

};

/**

2.17.1

Where is the in-kernel user of this new callback? Without that, there is no need for this, NOR is there a way to properly review this commit.

sorry,

greg k-h