Skip to contentSkip to conversation

Yet another round of SDCA fixes

PATCH In review Charles Keepax 2 participants 9 messages Apr 9 – Apr 10

Subsystems (3)
lore
Share & tools
lore

Charles Keepax

author maintainer
#0 lore raw

cover Yet another round of SDCA fixes

Another round of SDCA fixes a couple of fix to the IRQ cleanup from Richard, and a minor tweak to the IRQ handling from me and an update to the comment on a fixme.

Thanks,
Charles

Changes since v1:
- Move cancel_work_sync() on boot_work to bus remove.
- Add fix for kernel doc error
- Add patch to update text of FIXME

Charles Keepax (3):
ASoC: SDCA: Correct kernel doc for sdca_irq_cleanup()
ASoC: SDCA: Tidy up irq_enable_flags()/sdca_irq_disable()
ASoC: SDCA: Update text of FIXME

Richard Fitzgerald (1):
ASoC: SDCA: Fix cleanup inversion in class driver

 sound/soc/sdca/sdca_class.c      | 34 ++++++++++++++++++++------------
 sound/soc/sdca/sdca_interrupts.c | 24 ++++++++++++++--------
 2 files changed, 37 insertions(+), 21 deletions(-)
Signature · 1 line
2.47.3
author maintainer
#1 lore raw

PATCH v2 1/4 ASoC: SDCA: Correct kernel doc for sdca_irq_cleanup()

Fix typo of function argument name in the kernel doc.

Fixes 0b8757b220f9 ("ASoC: SDCA: Unregister IRQ handlers on module remove")

---

New since v1.

 sound/soc/sdca/sdca_interrupts.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
sound/soc/sdca/sdca_interrupts.c +1 -1
index 5cdabf8ae9da3..475aebcafe3f1 100644
@@ -555,7 +555,7 @@ EXPORT_SYMBOL_NS_GPL(sdca_irq_populate, "SND_SOC_SDCA");
 
 /**
  * sdca_irq_cleanup - Free all the individual IRQs for an SDCA Function
- * @sdev: Device pointer against which the sdca_interrupt_info was allocated.
+ * @dev: Device pointer against which the sdca_interrupt_info was allocated.
  * @function: Pointer to the SDCA Function.
  * @info: Pointer to the SDCA interrupt info for this device.
  *
Signature · 1 line
2.47.3
author maintainer
#2 lore raw

PATCH v2 2/4 ASoC: SDCA: Fix cleanup inversion in class driver

In reply to Charles Keepax

From: Richard Fitzgerald <[email protected]>

Fix inverted cleanup of the SoundWire IRQ and the function drivers that use it.

The devm cleanup function to call sdca_dev_unregister_functions() was being registered at the end of class_sdw_probe(). The bus core creates the parent SoundWire IRQ handler after class_sdw_probe() has returned, and it registers a devm cleanup handler at the same time.

This led to a cleanup inversion where the devm cleanup for the parent Soundwire IRQ runs before the handler that removes the function drivers. So the parent IRQ is destroyed before the function drivers had a chance to do any cleanup and remove their IRQ handlers.

Move the registrations of the function driver cleanup into class_boot_work() after the function drivers are registered, so that it runs before the cleanup of the parent SoundWire IRQ handler.

Fixes 2d877d0659cb ("ASoC: SDCA: Add basic SDCA class driver")

---

Changes since v1:
- Move cancel_work_sync() on boot_work to bus remove.

 sound/soc/sdca/sdca_class.c | 34 +++++++++++++++++++++-------------
 1 file changed, 21 insertions(+), 13 deletions(-)
sound/soc/sdca/sdca_class.c +21 -13
index 7af4e5d1b347f..6e9b66f718019 100644
@@ -137,6 +137,13 @@ static const struct regmap_config class_dev_regmap_config = {
 	.unlock			= class_regmap_unlock,
 };
 
+static void class_remove_functions(void *data)
+{
+	struct sdca_class_drv *drv = data;
+
+	sdca_dev_unregister_functions(drv->sdw);
+}
+
 static void class_boot_work(struct work_struct *work)
 {
 	struct sdca_class_drv *drv = container_of(work,
@@ -157,6 +164,11 @@ static void class_boot_work(struct work_struct *work)
 	if (ret)
 		goto err;
 
+	/* Ensure function drivers are removed before the IRQ is destroyed */
+	ret = devm_add_action_or_reset(drv->dev, class_remove_functions, drv);
+	if (ret)
+		goto err;
+
 	dev_dbg(drv->dev, "boot work complete\n");
 
 	pm_runtime_mark_last_busy(drv->dev);
@@ -168,15 +180,6 @@ static void class_boot_work(struct work_struct *work)
 	pm_runtime_put_sync(drv->dev);
 }
 
-static void class_dev_remove(void *data)
-{
-	struct sdca_class_drv *drv = data;
-
-	cancel_work_sync(&drv->boot_work);
-
-	sdca_dev_unregister_functions(drv->sdw);
-}
-
 static int class_sdw_probe(struct sdw_slave *sdw, const struct sdw_device_id *id)
 {
 	struct device *dev = &sdw->dev;
@@ -230,15 +233,19 @@ static int class_sdw_probe(struct sdw_slave *sdw, const struct sdw_device_id *id
 	if (ret)
 		return ret;
 
-	ret = devm_add_action_or_reset(dev, class_dev_remove, drv);
-	if (ret)
-		return ret;
-
 	queue_work(system_long_wq, &drv->boot_work);
 
 	return 0;
 }
 
+static void class_sdw_remove(struct sdw_slave *sdw)
+{
+	struct device *dev = &sdw->dev;
+	struct sdca_class_drv *drv = dev_get_drvdata(dev);
+
+	cancel_work_sync(&drv->boot_work);
+}
+
 static int class_suspend(struct device *dev)
 {
 	struct sdca_class_drv *drv = dev_get_drvdata(dev);
@@ -330,6 +337,7 @@ static struct sdw_driver class_sdw_driver = {
 	},
 
 	.probe		= class_sdw_probe,
+	.remove		= class_sdw_remove,
 	.id_table	= class_sdw_id,
 	.ops		= &class_sdw_ops,
 };
Signature · 1 line
2.47.3
author maintainer
#3 lore raw

PATCH v2 3/4 ASoC: SDCA: Tidy up irq_enable_flags()/sdca_irq_disable()

In reply to Charles Keepax

In irq_enable_flags() and sdca_irq_disable() there is a NULL check on the interrupt data pointer, however this is just pulled from an array so can never be NULL. This was likely left over from an earlier version that looked up the data in a different way. Replace the check with checking for the IRQ itself being non-zero.

Whilst here also drop the sdca_interrupt structure down into the loop within the function to better match the style of the rest of the code in this file.

---

No changes since v1.

 sound/soc/sdca/sdca_interrupts.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
sound/soc/sdca/sdca_interrupts.c +4 -6
index 475aebcafe3f1..3d7eb19967243 100644
@@ -630,13 +630,12 @@ EXPORT_SYMBOL_NS_GPL(sdca_irq_allocate, "SND_SOC_SDCA");
 static void irq_enable_flags(struct sdca_function_data *function,
 			     struct sdca_interrupt_info *info, bool early)
 {
-	struct sdca_interrupt *interrupt;
 	int i;
 
 	for (i = 0; i < SDCA_MAX_INTERRUPTS; i++) {
-		interrupt = &info->irqs[i];
+		struct sdca_interrupt *interrupt = &info->irqs[i];
 
-		if (!interrupt || interrupt->function != function)
+		if (!interrupt->irq || interrupt->function != function)
 			continue;
 
 		switch (SDCA_CTL_TYPE(interrupt->entity->type,
@@ -689,13 +688,12 @@ EXPORT_SYMBOL_NS_GPL(sdca_irq_enable, "SND_SOC_SDCA");
 void sdca_irq_disable(struct sdca_function_data *function,
 		      struct sdca_interrupt_info *info)
 {
-	struct sdca_interrupt *interrupt;
 	int i;
 
 	for (i = 0; i < SDCA_MAX_INTERRUPTS; i++) {
-		interrupt = &info->irqs[i];
+		struct sdca_interrupt *interrupt = &info->irqs[i];
 
-		if (!interrupt || interrupt->function != function)
+		if (!interrupt->irq || interrupt->function != function)
 			continue;
 
 		disable_irq(interrupt->irq);
Signature · 1 line
2.47.3
author maintainer
#4 lore raw

PATCH 4/4 ASoC: SDCA: Update text of FIXME

In reply to Charles Keepax

A couple of attempts to correct this FIXME have been sent upstream but the situation is not quite a simple as the FIXME implies. Update the FIXME to include a better description of the situation.

---

New since v1.

 sound/soc/sdca/sdca_interrupts.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
sound/soc/sdca/sdca_interrupts.c +11 -1
index 3d7eb19967243..6e10b4e660d96 100644
@@ -119,7 +119,17 @@ static irqreturn_t function_status_handler(int irq, void *data)
 	for_each_set_bit(mask, &status, BITS_PER_BYTE) {
 		switch (BIT(mask)) {
 		case SDCA_CTL_ENTITY_0_FUNCTION_NEEDS_INITIALIZATION:
-			//FIXME: Add init writes
+/*
+ * FIXME: Should this do init writes?
+ *
+ * Currently init writes/cache sync are done from the suspend/resume
+ * infrastructure. It is unclear in what situations one would receive this
+ * IRQ outside of that flow. Presumably it would be something like the chip
+ * crashing. In that case however doing the init writes and a cache sync might
+ * not be sufficient, for example if the failure was during audio playback
+ * there could be ordering constraints on the register writes to restore the
+ * state that are not handled by a simple cache sync.
+ */
 			break;
 		case SDCA_CTL_ENTITY_0_FUNCTION_FAULT:
 			dev_err(dev, "function fault\n");
Signature · 1 line
2.47.3
maintainer
#5 lore raw

Re: [PATCH 4/4] ASoC: SDCA: Update text of FIXME

Quoted 3 lines · Charles Keepax

A couple of attempts to correct this FIXME have been sent upstream but the situation is not quite a simple as the FIXME implies. Update the FIXME to include a better description of the situation.

Please don't try to mix versions in the middle of a series, that is just going to confuse both people and tooling. The versioning applies to the series as a whole.

maintainer
#6 lore raw

Re: [PATCH 4/4] ASoC: SDCA: Update text of FIXME

Quoted 8 lines · Mark Brown
Quoted 3 lines · Charles Keepax

A couple of attempts to correct this FIXME have been sent upstream but the situation is not quite a simple as the FIXME implies. Update the FIXME to include a better description of the situation.

Please don't try to mix versions in the middle of a series, that is just going to confuse both people and tooling. The versioning applies to the series as a whole.

Yeah, I actually can't convince things to interpret this as part of the series but the threading means the actual series gets in the way of handling the patch by itself.

author maintainer
#7 lore raw

Re: [PATCH 4/4] ASoC: SDCA: Update text of FIXME

Quoted 6 lines · Mark Brown
Quoted 5 lines · Mark Brown
Quoted 3 lines · Charles Keepax

A couple of attempts to correct this FIXME have been sent upstream but the situation is not quite a simple as the FIXME implies. Update the FIXME to include a better description of the situation.

Quoted 7 lines
Quoted 3 lines

Please don't try to mix versions in the middle of a series, that is just going to confuse both people and tooling. The versioning applies to the series as a whole.

Yeah, I actually can't convince things to interpret this as part of the series but the threading means the actual series gets in the way of handling the patch by itself.

Sorry I didn't actually mean to do that. I will resend the extra patch.

Thanks,
Charles

maintainer
#8 lore raw
In reply to Charles Keepax
Quoted 10 lines · Charles Keepax

Yet another round of SDCA fixes

Another round of SDCA fixes a couple of fix to the IRQ cleanup from Richard, and a minor tweak to the IRQ handling from me and an update to the comment on a fixme.

Thanks,
Charles

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git for-7.1

Thanks!

[1/4] ASoC: SDCA: Correct kernel doc for sdca_irq_cleanup()

      https://git.kernel.org/broonie/sound/c/c271b0815f45
[2/4] ASoC: SDCA: Fix cleanup inversion in class driver
      https://git.kernel.org/broonie/sound/c/7936490e0473
[3/4] ASoC: SDCA: Tidy up irq_enable_flags()/sdca_irq_disable()
      https://git.kernel.org/broonie/sound/c/87ceac0a98e9

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