Skip to contentSkip to conversation

ALSA: hda: fix the missing ptr initialization

PATCH Superseded Vinod Koul 2 participants 6 messages Apr 28, 2016 Sound

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Vinod Koul

author
#0 lore raw

PATCH 1/2 ALSA: hda: fix the missing ptr initialization

ebus is a member of extended device and was never initialized, so do this at device creation.

---
 sound/hda/ext/hdac_ext_bus.c | 1 +
 1 file changed, 1 insertion(+)
sound/hda/ext/hdac_ext_bus.c +1 -0
index 2433f7c81472..64de0a3d6d93 100644
@@ -144,6 +144,7 @@ int snd_hdac_ext_bus_device_init(struct hdac_ext_bus *ebus, int addr)
 	if (!edev)
 		return -ENOMEM;
 	hdev = &edev->hdac;
+	edev->ebus = ebus;
 
 	snprintf(name, sizeof(name), "ehdaudio%dD%d", ebus->idx, addr);
 
Signature · 1 line
1.9.1
author
#1 lore raw

PATCH 2/2 ALSA: hda: fix to wait for RIRB & CORB DMA to set

From: Jeeja KP <[email protected]>

After setting the stop bit of RIRB/CORB DMA, we should wait for stop bit to be set.

---
 sound/hda/hdac_controller.c | 33 +++++++++++++++++++++++++++++++--
 1 file changed, 31 insertions(+), 2 deletions(-)
sound/hda/hdac_controller.c +31 -2
index 8c486235c905..cc020c1706e7 100644
@@ -80,6 +80,35 @@ void snd_hdac_bus_init_cmd_io(struct hdac_bus *bus)
 }
 EXPORT_SYMBOL_GPL(snd_hdac_bus_init_cmd_io);
 
+
+/* Stop CORB DMA and poll till they are stopped */
+static void hdac_stop_corb_dma(struct hdac_bus *bus)
+{
+	unsigned long timeout;
+
+	snd_hdac_chip_writew(bus, CORBCTL, 0);
+
+	udelay(3);
+	timeout = jiffies + msecs_to_jiffies(100);
+	while ((snd_hdac_chip_readb(bus, CORBCTL) & AZX_CORBCTL_RUN)
+		&& time_before(jiffies, timeout))
+		udelay(10);
+}
+
+/* Stop RIRB DMA and poll till they are stopped */
+static void hdac_stop_rirb_dma(struct hdac_bus *bus)
+{
+	unsigned long timeout;
+
+	snd_hdac_chip_writew(bus, RIRBCTL, 0);
+
+	udelay(3);
+	timeout = jiffies + msecs_to_jiffies(100);
+	while ((snd_hdac_chip_readb(bus, RIRBCTL) & AZX_RBCTL_DMA_EN)
+		&& time_before(jiffies, timeout))
+		udelay(10);
+}
+
 /**
  * snd_hdac_bus_stop_cmd_io - clean up CORB/RIRB buffers
  * @bus: HD-audio core bus
@@ -88,8 +117,8 @@ void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus)
 {
 	spin_lock_irq(&bus->reg_lock);
 	/* disable ringbuffer DMAs */
-	snd_hdac_chip_writeb(bus, RIRBCTL, 0);
-	snd_hdac_chip_writeb(bus, CORBCTL, 0);
+	hdac_stop_rirb_dma(bus);
+	hdac_stop_corb_dma(bus);
 	/* disable unsolicited responses */
 	snd_hdac_chip_updatel(bus, GCTL, AZX_GCTL_UNSOL, 0);
 	spin_unlock_irq(&bus->reg_lock);
Signature · 1 line
1.9.1
#2 lore raw

Re: [PATCH 2/2] ALSA: hda: fix to wait for RIRB & CORB DMA to set

Quoted 5 lines · Vinod Koul

From: Jeeja KP <[email protected]>

After setting the stop bit of RIRB/CORB DMA, we should wait for stop bit to be set.

What does this actually fix?

Quoted 47 lines
---
 sound/hda/hdac_controller.c | 33 +++++++++++++++++++++++++++++++--
 1 file changed, 31 insertions(+), 2 deletions(-)
sound/hda/hdac_controller.c +29 -0
index 8c486235c905..cc020c1706e7 100644
@@ -80,6 +80,35 @@ void snd_hdac_bus_init_cmd_io(struct hdac_bus *bus)
 }
 EXPORT_SYMBOL_GPL(snd_hdac_bus_init_cmd_io);
 
+
+/* Stop CORB DMA and poll till they are stopped */
+static void hdac_stop_corb_dma(struct hdac_bus *bus)
+{
+	unsigned long timeout;
+
+	snd_hdac_chip_writew(bus, CORBCTL, 0);
+
+	udelay(3);
+	timeout = jiffies + msecs_to_jiffies(100);
+	while ((snd_hdac_chip_readb(bus, CORBCTL) & AZX_CORBCTL_RUN)
+		&& time_before(jiffies, timeout))
+		udelay(10);
+}
+
+/* Stop RIRB DMA and poll till they are stopped */
+static void hdac_stop_rirb_dma(struct hdac_bus *bus)
+{
+	unsigned long timeout;
+
+	snd_hdac_chip_writew(bus, RIRBCTL, 0);
+
+	udelay(3);
+	timeout = jiffies + msecs_to_jiffies(100);
+	while ((snd_hdac_chip_readb(bus, RIRBCTL) & AZX_RBCTL_DMA_EN)
+		&& time_before(jiffies, timeout))
+		udelay(10);
+}
+
 /**
  * snd_hdac_bus_stop_cmd_io - clean up CORB/RIRB buffers
  * @bus: HD-audio core bus

sound/hda/hdac_controller.c void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus) · line 121

Quoted 7 lines
sound/hda/hdac_controller.c +2 -2
@@ -88,8 +117,8 @@ void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus)
 {
 	spin_lock_irq(&bus->reg_lock);
 	/* disable ringbuffer DMAs */
-	snd_hdac_chip_writeb(bus, RIRBCTL, 0);
-	snd_hdac_chip_writeb(bus, CORBCTL, 0);
+	hdac_stop_rirb_dma(bus);
+	hdac_stop_corb_dma(bus);

Doing these one after another sequentially is a waste of time. Clear two once, then sync them.

thanks,

Takashi

author
#3 lore raw

Re: [PATCH 2/2] ALSA: hda: fix to wait for RIRB & CORB DMA to set

Quoted 9 lines · Takashi Iwai
Quoted 5 lines · Vinod Koul

From: Jeeja KP <[email protected]>

After setting the stop bit of RIRB/CORB DMA, we should wait for stop bit to be set.

What does this actually fix?

We some some stablity issues on SKL that were attributed to DMAs not being quisced properly so recommendation was to to wait till DMAs are stopped

Quoted 11 lines

sound/hda/hdac_controller.c void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus) · line 121

Quoted 8 lines
sound/hda/hdac_controller.c +2 -2
@@ -88,8 +117,8 @@ void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus)
 {
 	spin_lock_irq(&bus->reg_lock);
 	/* disable ringbuffer DMAs */
-	snd_hdac_chip_writeb(bus, RIRBCTL, 0);
-	snd_hdac_chip_writeb(bus, CORBCTL, 0);
+	hdac_stop_rirb_dma(bus);
+	hdac_stop_corb_dma(bus);

Doing these one after another sequentially is a waste of time. Clear two once, then sync them.

Yes agreed, that sounds better, something like adding a new wait function:

	hdac_stop_rirb_dma(bus);
	hdac_stop_corb_dma(bus);
	hdac_wait_for_cmd_dmas(bus);

Thanks

Signature · 1 line
~Vinod
#4 lore raw

Re: [PATCH 2/2] ALSA: hda: fix to wait for RIRB & CORB DMA to set

Quoted 7 lines · Vinod Koul
Quoted 5 lines · Takashi Iwai
Quoted 3 lines · Vinod Koul

From: Jeeja KP <[email protected]>

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

After setting the stop bit of RIRB/CORB DMA, we should wait for stop bit to be set.

What does this actually fix?

We some some stablity issues on SKL that were attributed to DMAs not being quisced properly so recommendation was to to wait till DMAs are stopped

Then please write it up in the changelog.

And, did it actually improve the stability? That's the biggest question :)

Quoted 12 lines
Quoted 11 lines

sound/hda/hdac_controller.c void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus) · line 121

Quoted 8 lines
sound/hda/hdac_controller.c +2 -2
@@ -88,8 +117,8 @@ void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus)
 {
 	spin_lock_irq(&bus->reg_lock);
 	/* disable ringbuffer DMAs */
-	snd_hdac_chip_writeb(bus, RIRBCTL, 0);
-	snd_hdac_chip_writeb(bus, CORBCTL, 0);
+	hdac_stop_rirb_dma(bus);
+	hdac_stop_corb_dma(bus);

Doing these one after another sequentially is a waste of time. Clear two once, then sync them.

Quoted 5 lines

Yes agreed, that sounds better, something like adding a new wait function:

	hdac_stop_rirb_dma(bus);
	hdac_stop_corb_dma(bus);
	hdac_wait_for_cmd_dmas(bus);

Yep, that looks better.

thanks,

Takashi

author
#5 lore raw

Re: [PATCH 2/2] ALSA: hda: fix to wait for RIRB & CORB DMA to set

Quoted 12 lines · Takashi Iwai
Quoted 10 lines · Vinod Koul
Quoted 8 lines · Takashi Iwai
Quoted 5 lines · Vinod Koul

From: Jeeja KP <[email protected]>

After setting the stop bit of RIRB/CORB DMA, we should wait for stop bit to be set.

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

What does this actually fix?

We some some stablity issues on SKL that were attributed to DMAs not being quisced properly so recommendation was to to wait till DMAs are stopped

Then please write it up in the changelog.

Sure thing

Quoted 3 lines

And, did it actually improve the stability? That's the biggest question :)

Yes it did :)

Quoted 12 lines
Quoted 12 lines
Quoted 11 lines

sound/hda/hdac_controller.c void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus) · line 121

Quoted 8 lines
sound/hda/hdac_controller.c +2 -2
@@ -88,8 +117,8 @@ void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus)
 {
 	spin_lock_irq(&bus->reg_lock);
 	/* disable ringbuffer DMAs */
-	snd_hdac_chip_writeb(bus, RIRBCTL, 0);
-	snd_hdac_chip_writeb(bus, CORBCTL, 0);
+	hdac_stop_rirb_dma(bus);
+	hdac_stop_corb_dma(bus);

Doing these one after another sequentially is a waste of time. Clear two once, then sync them.

Quoted 7 lines
Quoted 5 lines

Yes agreed, that sounds better, something like adding a new wait function:

	hdac_stop_rirb_dma(bus);
	hdac_stop_corb_dma(bus);
	hdac_wait_for_cmd_dmas(bus);

Yep, that looks better.

Okay will update and post v2

Signature · 1 line
~Vinod