Skip to contentSkip to conversation

ALSA: hdac: stream fixes

PATCH Applied Vinod Koul 2 participants 8 messages Aug 4, 2015 Sound

lore
Share & tools
lore

Vinod Koul

author
#0 lore raw

cover ALSA: hdac: stream fixes

Hi Takashi,

I had few off fixes on hdac piled up so time to send them out.

First one is a fix for stream assignment for decoupled mode used in SKL driver. Second one is static checker fix reported by Dan Carpenter Last is a potential issue which our team reported so a WA for that. The command DMAs do not finish instantaneously, so we should per spec wait till DMA finishes

Jeeja KP (2):
ALSA: HDA: Fix stream assignment for host in decoupled mode
ALSA: HDA: Dont check return for snd_hdac_chip_readl

Vinod Koul (1):
ALSA: HDA: wait for RIRB, CORB DMA to finish

 sound/hda/ext/hdac_ext_controller.c |  6 ------
 sound/hda/ext/hdac_ext_stream.c     |  2 +-
 sound/hda/hdac_controller.c         | 14 ++++++++++++++
 3 files changed, 15 insertions(+), 7 deletions(-)
Signature · 1 line
1.9.1
author
#1 lore raw

PATCH 1/3 ALSA: HDA: Fix stream assignment for host in decoupled mode

From: Jeeja KP <[email protected]>

This fixes issue in assigning host stream in case of decoupled mode. The check to verify if the stream is already in use was wrong so fix that

---
 sound/hda/ext/hdac_ext_stream.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
sound/hda/ext/hdac_ext_stream.c +1 -1
index f8ffbdbb450d..3de47dd1a76d 100644
@@ -299,7 +299,7 @@ hdac_ext_host_stream_assign(struct hdac_ext_bus *ebus,
 		if (stream->direction != substream->stream)
 			continue;
 
-		if (stream->opened) {
+		if (!stream->opened) {
 			if (!hstream->decoupled)
 				snd_hdac_ext_stream_decouple(ebus, hstream, true);
 			res = hstream;
Signature · 1 line
1.9.1
Applied
#2 lore raw

Re: [PATCH 1/3] ALSA: HDA: Fix stream assignment for host in decoupled mode

Quoted 9 lines · Vinod Koul

From: Jeeja KP <[email protected]>

This fixes issue in assigning host stream in case of decoupled mode. The check to verify if the stream is already in use was wrong so fix that

Applied, thanks.

Takashi

Quoted 20 lines
---
 sound/hda/ext/hdac_ext_stream.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
sound/hda/ext/hdac_ext_stream.c +1 -1
index f8ffbdbb450d..3de47dd1a76d 100644
@@ -299,7 +299,7 @@ hdac_ext_host_stream_assign(struct hdac_ext_bus *ebus,
 		if (stream->direction != substream->stream)
 			continue;
 
-		if (stream->opened) {
+		if (!stream->opened) {
 			if (!hstream->decoupled)
 				snd_hdac_ext_stream_decouple(ebus, hstream, true);
 			res = hstream;
Signature · 1 line
1.9.1
author
#3 lore raw

PATCH 2/3 ALSA: HDA: Dont check return for snd_hdac_chip_readl

In reply to Vinod Koul

From: Jeeja KP <[email protected]>

The snd_hdac_chip_readl return can never be less than zeros, so no point in checking for the return value

This fixes following static checker warnings in
snd_hdac_ext_bus_parse_capabilities

sound/hda/ext/hdac_ext_controller.c:47
snd_hdac_ext_bus_parse_capabilities()
warn: unsigned 'offset' is never less than zero.

sound/hda/ext/hdac_ext_controller.c:54
snd_hdac_ext_bus_parse_capabilities()
warn: unsigned 'cur_cap' is never less than zero.

---
 sound/hda/ext/hdac_ext_controller.c | 6 ------
 1 file changed, 6 deletions(-)
sound/hda/ext/hdac_ext_controller.c +0 -6
index b2da19b60f4e..358f16195483 100644
@@ -44,16 +44,10 @@ int snd_hdac_ext_bus_parse_capabilities(struct hdac_ext_bus *ebus)
 
 	offset = snd_hdac_chip_readl(bus, LLCH);
 
-	if (offset < 0)
-		return -EIO;
-
 	/* Lets walk the linked capabilities list */
 	do {
 		cur_cap = _snd_hdac_chip_read(l, bus, offset);
 
-		if (cur_cap < 0)
-			return -EIO;
-
 		dev_dbg(bus->dev, "Capability version: 0x%x\n",
 				((cur_cap & AZX_CAP_HDR_VER_MASK) >> AZX_CAP_HDR_VER_OFF));
 
Signature · 1 line
1.9.1
Applied
#4 lore raw

Re: [PATCH 2/3] ALSA: HDA: Dont check return for snd_hdac_chip_readl

Quoted 13 lines · Vinod Koul

From: Jeeja KP <[email protected]>

The snd_hdac_chip_readl return can never be less than zeros, so no point in checking for the return value

This fixes following static checker warnings in
snd_hdac_ext_bus_parse_capabilities

sound/hda/ext/hdac_ext_controller.c:47
snd_hdac_ext_bus_parse_capabilities()
warn: unsigned 'offset' is never less than zero.

Quoted 7 lines

sound/hda/ext/hdac_ext_controller.c:54
snd_hdac_ext_bus_parse_capabilities()
warn: unsigned 'cur_cap' is never less than zero.

Applied, thanks.

Takashi

Quoted 28 lines
---
 sound/hda/ext/hdac_ext_controller.c | 6 ------
 1 file changed, 6 deletions(-)
sound/hda/ext/hdac_ext_controller.c +0 -6
index b2da19b60f4e..358f16195483 100644
@@ -44,16 +44,10 @@ int snd_hdac_ext_bus_parse_capabilities(struct hdac_ext_bus *ebus)
 
 	offset = snd_hdac_chip_readl(bus, LLCH);
 
-	if (offset < 0)
-		return -EIO;
-
 	/* Lets walk the linked capabilities list */
 	do {
 		cur_cap = _snd_hdac_chip_read(l, bus, offset);
 
-		if (cur_cap < 0)
-			return -EIO;
-
 		dev_dbg(bus->dev, "Capability version: 0x%x\n",
 				((cur_cap & AZX_CAP_HDR_VER_MASK) >> AZX_CAP_HDR_VER_OFF));
 
Signature · 1 line
1.9.1
author
#5 lore raw

PATCH 3/3 ALSA: HDA: wait for RIRB, CORB DMA to finish

In reply to Vinod Koul

HDA spec says that RORB and CORB DMA stop will take some time to complete. So we should wait till the DMAs are stopped.

Although the current controllers don't have multilinks so doesn't impact much, but SKL onwards we have multiple links so waiting for DMAs to stop makes better sense.

---
 sound/hda/hdac_controller.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
sound/hda/hdac_controller.c +14 -0
index b5a17cb510a0..3b5d07174d79 100644
@@ -86,10 +86,24 @@ EXPORT_SYMBOL_GPL(snd_hdac_bus_init_cmd_io);
  */
 void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus)
 {
+	unsigned long timeout;
+
 	spin_lock_irq(&bus->reg_lock);
 	/* disable ringbuffer DMAs */
 	snd_hdac_chip_writeb(bus, RIRBCTL, 0);
 	snd_hdac_chip_writeb(bus, CORBCTL, 0);
+
+	/* poll DMAs to check if they stopped or not */
+
+	timeout = jiffies + msecs_to_jiffies(100);
+	while ((snd_hdac_chip_readb(bus, RIRBCTL) & AZX_RBCTL_DMA_EN) &&
+	       time_before(jiffies, timeout))
+		usleep_range(500, 1000);
+	timeout = jiffies + msecs_to_jiffies(100);
+	while ((snd_hdac_chip_readb(bus, CORBCTL) & AZX_CORBCTL_RUN) &&
+	       time_before(jiffies, timeout))
+		usleep_range(500, 1000);
+
 	/* 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
#6 lore raw

Re: [PATCH 3/3] ALSA: HDA: wait for RIRB, CORB DMA to finish

Quoted 26 lines · Vinod Koul

HDA spec says that RORB and CORB DMA stop will take some time to complete. So we should wait till the DMAs are stopped.

Although the current controllers don't have multilinks so doesn't impact much, but SKL onwards we have multiple links so waiting for DMAs to stop makes better sense.

---
 sound/hda/hdac_controller.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
sound/hda/hdac_controller.c +2 -0
index b5a17cb510a0..3b5d07174d79 100644
@@ -86,10 +86,24 @@ EXPORT_SYMBOL_GPL(snd_hdac_bus_init_cmd_io);
  */
 void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus)
 {
+	unsigned long timeout;
+
 	spin_lock_irq(&bus->reg_lock);

sound/hda/hdac_controller.c EXPORT_SYMBOL_GPL(snd_hdac_bus_init_cmd_io); · line 101

Quoted 10 lines
sound/hda/hdac_controller.c +7 -0
 	/* disable ringbuffer DMAs */
 	snd_hdac_chip_writeb(bus, RIRBCTL, 0);
 	snd_hdac_chip_writeb(bus, CORBCTL, 0);
+
+	/* poll DMAs to check if they stopped or not */
+
+	timeout = jiffies + msecs_to_jiffies(100);
+	while ((snd_hdac_chip_readb(bus, RIRBCTL) & AZX_RBCTL_DMA_EN) &&
+	       time_before(jiffies, timeout))
+		usleep_range(500, 1000);

You must not use *sleep() inside atomic context.

Takashi

author
#7 lore raw

Re: [PATCH 3/3] ALSA: HDA: wait for RIRB, CORB DMA to finish

Quoted 11 lines · Takashi Iwai
Quoted 11 lines
sound/hda/hdac_controller.c +3 -0
@@ -86,10 +86,24 @@ EXPORT_SYMBOL_GPL(snd_hdac_bus_init_cmd_io);
  */
 void snd_hdac_bus_stop_cmd_io(struct hdac_bus *bus)
 {
+	unsigned long timeout;
+
 	spin_lock_irq(&bus->reg_lock);
 	/* disable ringbuffer DMAs */
 	snd_hdac_chip_writeb(bus, RIRBCTL, 0);
 	snd_hdac_chip_writeb(bus, CORBCTL, 0);
+
Quoted 8 lines

sound/hda/hdac_controller.c EXPORT_SYMBOL_GPL(snd_hdac_bus_init_cmd_io); · line 101

Quoted 6 lines
sound/hda/hdac_controller.c +6 -0
+	/* poll DMAs to check if they stopped or not */
+
+	timeout = jiffies + msecs_to_jiffies(100);
+	while ((snd_hdac_chip_readb(bus, RIRBCTL) & AZX_RBCTL_DMA_EN) &&
+	       time_before(jiffies, timeout))
+		usleep_range(500, 1000);

You must not use *sleep() inside atomic context.

Right, not sure why it didnt crib when we tested, will send updated one soon

Thanks for the super quick review :)

Signature · 1 line
~Vinod