[alsa-devel] [PATCH] ALSA: HDA: Lessen CPU usage when waiting for chip to respond
When an IRQ for some reason gets lost, we wait up to a second using udelay, which is CPU intensive. This patch improves the situation by waiting about 30 ms in the CPU intensive mode, then stepping down to using msleep(2) instead. In essence, we trade some granularity in exchange for less CPU consumption when the waiting time is a bit longer.
As a result, PulseAudio should no longer be killed by the kernel for taking up to much RT-prio CPU time. At least not for *this* reason.
Signed-off-by: David Henningsson david.henningsson@canonical.com --- sound/pci/hda/hda_intel.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
Hi Arun,
Can you check if this patch resolves your problem with PulseAudio getting killed by the kernel? If so, we should apply it to the kernel, perhaps even to stable.
// David
diff --git a/sound/pci/hda/hda_intel.c b/sound/pci/hda/hda_intel.c index 7b6a823..0e7c3f1 100644 --- a/sound/pci/hda/hda_intel.c +++ b/sound/pci/hda/hda_intel.c @@ -783,11 +783,13 @@ static unsigned int azx_rirb_get_response(struct hda_bus *bus, { struct azx *chip = bus->private_data; unsigned long timeout; + unsigned long loopcounter; int do_poll = 0;
again: timeout = jiffies + msecs_to_jiffies(1000); - for (;;) { + + for (loopcounter = 0;; loopcounter++) { if (chip->polling_mode || do_poll) { spin_lock_irq(&chip->reg_lock); azx_update_rirb(chip); @@ -803,7 +805,7 @@ static unsigned int azx_rirb_get_response(struct hda_bus *bus, } if (time_after(jiffies, timeout)) break; - if (bus->needs_damn_long_delay) + if (bus->needs_damn_long_delay || loopcounter > 3000) msleep(2); /* temporary workaround */ else { udelay(10);
At Fri, 4 May 2012 11:05:55 +0200, David Henningsson wrote:
When an IRQ for some reason gets lost, we wait up to a second using udelay, which is CPU intensive. This patch improves the situation by waiting about 30 ms in the CPU intensive mode, then stepping down to using msleep(2) instead. In essence, we trade some granularity in exchange for less CPU consumption when the waiting time is a bit longer.
As a result, PulseAudio should no longer be killed by the kernel for taking up to much RT-prio CPU time. At least not for *this* reason.
Signed-off-by: David Henningsson david.henningsson@canonical.com
sound/pci/hda/hda_intel.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
Hi Arun,
Can you check if this patch resolves your problem with PulseAudio getting killed by the kernel? If so, we should apply it to the kernel, perhaps even to stable.
The patch looks good to me. I'm going to apply it for 3.4 kernel tree (with Cc to stable, too) if it's confirmed to improve actually. Even if not, it's still good to apply, but this kind of change is a 3.5 kernel material.
So, Arun, please let me know whether it works for you.
thanks,
Takashi
// David
diff --git a/sound/pci/hda/hda_intel.c b/sound/pci/hda/hda_intel.c index 7b6a823..0e7c3f1 100644 --- a/sound/pci/hda/hda_intel.c +++ b/sound/pci/hda/hda_intel.c @@ -783,11 +783,13 @@ static unsigned int azx_rirb_get_response(struct hda_bus *bus, { struct azx *chip = bus->private_data; unsigned long timeout;
unsigned long loopcounter; int do_poll = 0;
again: timeout = jiffies + msecs_to_jiffies(1000);
- for (;;) {
- for (loopcounter = 0;; loopcounter++) { if (chip->polling_mode || do_poll) { spin_lock_irq(&chip->reg_lock); azx_update_rirb(chip);
@@ -803,7 +805,7 @@ static unsigned int azx_rirb_get_response(struct hda_bus *bus, } if (time_after(jiffies, timeout)) break;
if (bus->needs_damn_long_delay)
if (bus->needs_damn_long_delay || loopcounter > 3000) msleep(2); /* temporary workaround */-- 1.7.9.5
Alsa-devel mailing list Alsa-devel@alsa-project.org http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
On Fri, 2012-05-04 at 11:05 +0200, David Henningsson wrote:
When an IRQ for some reason gets lost, we wait up to a second using udelay, which is CPU intensive. This patch improves the situation by waiting about 30 ms in the CPU intensive mode, then stepping down to using msleep(2) instead. In essence, we trade some granularity in exchange for less CPU consumption when the waiting time is a bit longer.
As a result, PulseAudio should no longer be killed by the kernel for taking up to much RT-prio CPU time. At least not for *this* reason.
Signed-off-by: David Henningsson david.henningsson@canonical.com
sound/pci/hda/hda_intel.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
Hi Arun,
Can you check if this patch resolves your problem with PulseAudio getting killed by the kernel? If so, we should apply it to the kernel, perhaps even to stable.
// David
Thanks, this fixes the problem for me. Don't know what the official procedure for these things is, but fwiw:
Signed-off-by: Arun Raghavan arun.raghavan@collabora.co.uk
diff --git a/sound/pci/hda/hda_intel.c b/sound/pci/hda/hda_intel.c index 7b6a823..0e7c3f1 100644 --- a/sound/pci/hda/hda_intel.c +++ b/sound/pci/hda/hda_intel.c @@ -783,11 +783,13 @@ static unsigned int azx_rirb_get_response(struct hda_bus *bus, { struct azx *chip = bus->private_data; unsigned long timeout;
unsigned long loopcounter; int do_poll = 0;
again: timeout = jiffies + msecs_to_jiffies(1000);
- for (;;) {
- for (loopcounter = 0;; loopcounter++) { if (chip->polling_mode || do_poll) { spin_lock_irq(&chip->reg_lock); azx_update_rirb(chip);
@@ -803,7 +805,7 @@ static unsigned int azx_rirb_get_response(struct hda_bus *bus, } if (time_after(jiffies, timeout)) break;
if (bus->needs_damn_long_delay)
if (bus->needs_damn_long_delay || loopcounter > 3000) msleep(2); /* temporary workaround */
At Wed, 09 May 2012 01:35:52 +0530, Arun Raghavan wrote:
On Fri, 2012-05-04 at 11:05 +0200, David Henningsson wrote:
When an IRQ for some reason gets lost, we wait up to a second using udelay, which is CPU intensive. This patch improves the situation by waiting about 30 ms in the CPU intensive mode, then stepping down to using msleep(2) instead. In essence, we trade some granularity in exchange for less CPU consumption when the waiting time is a bit longer.
As a result, PulseAudio should no longer be killed by the kernel for taking up to much RT-prio CPU time. At least not for *this* reason.
Signed-off-by: David Henningsson david.henningsson@canonical.com
sound/pci/hda/hda_intel.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
Hi Arun,
Can you check if this patch resolves your problem with PulseAudio getting killed by the kernel? If so, we should apply it to the kernel, perhaps even to stable.
// David
Thanks, this fixes the problem for me. Don't know what the official procedure for these things is, but fwiw:
Signed-off-by: Arun Raghavan arun.raghavan@collabora.co.uk
I guess you meant tested-by tag? In anyway I took the patch now with Cc to stable. It'll be included in the next pull request to Linus.
thanks,
Takashi
diff --git a/sound/pci/hda/hda_intel.c b/sound/pci/hda/hda_intel.c index 7b6a823..0e7c3f1 100644 --- a/sound/pci/hda/hda_intel.c +++ b/sound/pci/hda/hda_intel.c @@ -783,11 +783,13 @@ static unsigned int azx_rirb_get_response(struct hda_bus *bus, { struct azx *chip = bus->private_data; unsigned long timeout;
unsigned long loopcounter; int do_poll = 0;
again: timeout = jiffies + msecs_to_jiffies(1000);
- for (;;) {
- for (loopcounter = 0;; loopcounter++) { if (chip->polling_mode || do_poll) { spin_lock_irq(&chip->reg_lock); azx_update_rirb(chip);
@@ -803,7 +805,7 @@ static unsigned int azx_rirb_get_response(struct hda_bus *bus, } if (time_after(jiffies, timeout)) break;
if (bus->needs_damn_long_delay)
if (bus->needs_damn_long_delay || loopcounter > 3000) msleep(2); /* temporary workaround */
Alsa-devel mailing list Alsa-devel@alsa-project.org http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
On Wed, 2012-05-09 at 12:51 +0200, Takashi Iwai wrote:
At Wed, 09 May 2012 01:35:52 +0530, Arun Raghavan wrote:
On Fri, 2012-05-04 at 11:05 +0200, David Henningsson wrote:
When an IRQ for some reason gets lost, we wait up to a second using udelay, which is CPU intensive. This patch improves the situation by waiting about 30 ms in the CPU intensive mode, then stepping down to using msleep(2) instead. In essence, we trade some granularity in exchange for less CPU consumption when the waiting time is a bit longer.
As a result, PulseAudio should no longer be killed by the kernel for taking up to much RT-prio CPU time. At least not for *this* reason.
Signed-off-by: David Henningsson david.henningsson@canonical.com
sound/pci/hda/hda_intel.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
Hi Arun,
Can you check if this patch resolves your problem with PulseAudio getting killed by the kernel? If so, we should apply it to the kernel, perhaps even to stable.
// David
Thanks, this fixes the problem for me. Don't know what the official procedure for these things is, but fwiw:
Signed-off-by: Arun Raghavan arun.raghavan@collabora.co.uk
I guess you meant tested-by tag? In anyway I took the patch now with Cc to stable. It'll be included in the next pull request to Linus.
I did, and now I know what tag to use the next time. :)
Thanks, Arun
participants (3)
-
Arun Raghavan -
David Henningsson -
Takashi Iwai