[PATCH] ASoC: amd:Replacing MSI with Legacy IRQ model
When we try to play and capture simultaneously we see that interrupts are genrated but our handler is not being acknowledged, After investigating further more in detail on this issue we found that IRQ delivery via MSI from the ACP IP is unreliable and so sometimes interrupt generated will not be acknowledged so MSI model shouldn't be used and using legacy IRQs will resolve interrupt handling issue.
This patch replaces MSI interrupt handling with legacy IRQ model.
Issue can be reproduced easily by running below python script:
import subprocess import time import threading
def do2(): cmd = 'aplay -f dat -D hw:2,1 /dev/zero -d 1' subprocess.call(cmd, stdin=subprocess.PIPE, stderr=subprocess.PIPE, shell=True) print('Play Done')
def run(): for i in range(1000): do2()
def do(i): cmd = 'arecord -f dat -D hw:2,2 /dev/null -d 1' subprocess.call(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True) print(datetime.datetime.now(), i)
t = threading.Thread(target=run) t.start() for i in range(1000): do(i)
t.join()
After applying this patch issue is resolved.
Signed-off-by: Ravulapati Vishnu vardhan rao Vishnuvardhanrao.Ravulapati@amd.com --- sound/soc/amd/raven/pci-acp3x.c | 16 +++------------- 1 file changed, 3 insertions(+), 13 deletions(-)
diff --git a/sound/soc/amd/raven/pci-acp3x.c b/sound/soc/amd/raven/pci-acp3x.c index 8c138e490f0c..d3536fd6a124 100644 --- a/sound/soc/amd/raven/pci-acp3x.c +++ b/sound/soc/amd/raven/pci-acp3x.c @@ -140,21 +140,14 @@ static int snd_acp3x_probe(struct pci_dev *pci, goto release_regions; }
- /* check for msi interrupt support */ - ret = pci_enable_msi(pci); - if (ret) - /* msi is not enabled */ - irqflags = IRQF_SHARED; - else - /* msi is enabled */ - irqflags = 0; + irqflags = IRQF_SHARED;
addr = pci_resource_start(pci, 0); adata->acp3x_base = devm_ioremap(&pci->dev, addr, pci_resource_len(pci, 0)); if (!adata->acp3x_base) { ret = -ENOMEM; - goto disable_msi; + goto release_regions; } pci_set_master(pci); pci_set_drvdata(pci, adata); @@ -162,7 +155,7 @@ static int snd_acp3x_probe(struct pci_dev *pci, adata->pme_en = rv_readl(adata->acp3x_base + mmACP_PME_EN); ret = acp3x_init(adata); if (ret) - goto disable_msi; + goto release_regions;
val = rv_readl(adata->acp3x_base + mmACP_I2S_PIN_CONFIG); switch (val) { @@ -251,8 +244,6 @@ static int snd_acp3x_probe(struct pci_dev *pci, de_init: if (acp3x_deinit(adata->acp3x_base)) dev_err(&pci->dev, "ACP de-init failed\n"); -disable_msi: - pci_disable_msi(pci); release_regions: pci_release_regions(pci); disable_pci: @@ -311,7 +302,6 @@ static void snd_acp3x_remove(struct pci_dev *pci) dev_err(&pci->dev, "ACP de-init failed\n"); pm_runtime_forbid(&pci->dev); pm_runtime_get_noresume(&pci->dev); - pci_disable_msi(pci); pci_release_regions(pci); pci_disable_device(pci); }
On Tue, 22 Dec 2020 12:59:18 +0100, Ravulapati Vishnu vardhan rao wrote:
When we try to play and capture simultaneously we see that interrupts are genrated but our handler is not being acknowledged, After investigating further more in detail on this issue we found that IRQ delivery via MSI from the ACP IP is unreliable and so sometimes interrupt generated will not be acknowledged so MSI model shouldn't be used and using legacy IRQs will resolve interrupt handling issue.
This patch replaces MSI interrupt handling with legacy IRQ model.
Issue can be reproduced easily by running below python script:
import subprocess import time import threading
def do2(): cmd = 'aplay -f dat -D hw:2,1 /dev/zero -d 1' subprocess.call(cmd, stdin=subprocess.PIPE, stderr=subprocess.PIPE, shell=True) print('Play Done')
def run(): for i in range(1000): do2()
def do(i): cmd = 'arecord -f dat -D hw:2,2 /dev/null -d 1' subprocess.call(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True) print(datetime.datetime.now(), i)
t = threading.Thread(target=run) t.start() for i in range(1000): do(i)
t.join()
After applying this patch issue is resolved.
Signed-off-by: Ravulapati Vishnu vardhan rao Vishnuvardhanrao.Ravulapati@amd.com
Is this specific to Raven, i.e. Renoir doesn't need the same fix? If so, it should be mentioned in the patch description.
thanks,
Takashi
[AMD Official Use Only - Internal Distribution Only]
On Tue, 22 Dec 2020 12:59:18 +0100, Ravulapati Vishnu vardhan rao wrote:
When we try to play and capture simultaneously we see that interrupts are genrated but our handler is not being acknowledged, After investigating further more in detail on this issue we found that IRQ delivery via MSI from the ACP IP is unreliable and so sometimes interrupt generated will not be acknowledged so MSI model shouldn't be used and using legacy IRQs will resolve interrupt handling issue.
This patch replaces MSI interrupt handling with legacy IRQ model.
Issue can be reproduced easily by running below python script:
import subprocess import time import threading
def do2(): cmd = 'aplay -f dat -D hw:2,1 /dev/zero -d 1' subprocess.call(cmd, stdin=subprocess.PIPE, stderr=subprocess.PIPE, shell=True) print('Play Done')
def run(): for i in range(1000): do2()
def do(i): cmd = 'arecord -f dat -D hw:2,2 /dev/null -d 1' subprocess.call(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True) print(datetime.datetime.now(), i)
t = threading.Thread(target=run) t.start() for i in range(1000): do(i)
t.join()
After applying this patch issue is resolved.
Signed-off-by: Ravulapati Vishnu vardhan rao Vishnuvardhanrao.Ravulapati@amd.com
Is this specific to Raven, i.e. Renoir doesn't need >the same fix? If so, it should be mentioned in the patch >description.
We will send this fix as separate patch for Renoir.
Thanks, Vishnu
Dne 22. 12. 20 v 12:59 Ravulapati Vishnu vardhan rao napsal(a):
When we try to play and capture simultaneously we see that interrupts are genrated but our handler is not being acknowledged, After investigating further more in detail on this issue we found that IRQ delivery via MSI from the ACP IP is unreliable and so sometimes interrupt generated will not be acknowledged so MSI model shouldn't be used and using legacy IRQs will resolve interrupt handling issue.
What is the real culprit? It's hw bug? If not, it would be better to fix the PCI code or the irq handler.
Jaroslav
On Sun, 27 Dec 2020 19:03:28 +0100, Jaroslav Kysela wrote:
Dne 22. 12. 20 v 12:59 Ravulapati Vishnu vardhan rao napsal(a):
When we try to play and capture simultaneously we see that interrupts are genrated but our handler is not being acknowledged, After investigating further more in detail on this issue we found that IRQ delivery via MSI from the ACP IP is unreliable and so sometimes interrupt generated will not be acknowledged so MSI model shouldn't be used and using legacy IRQs will resolve interrupt handling issue.
What is the real culprit? It's hw bug? If not, it would be better to fix the PCI code or the irq handler.
I'm not surprised if it's a hardware "feature". There are lots of deny list and quirks about disabling MSI in HD-audio controller code, after all.
Takashi
On Tue, 22 Dec 2020 17:29:18 +0530, Ravulapati Vishnu vardhan rao wrote:
When we try to play and capture simultaneously we see that interrupts are genrated but our handler is not being acknowledged, After investigating further more in detail on this issue we found that IRQ delivery via MSI from the ACP IP is unreliable and so sometimes interrupt generated will not be acknowledged so MSI model shouldn't be used and using legacy IRQs will resolve interrupt handling issue.
[...]
Applied to
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git for-next
Thanks!
[1/1] ASoC: amd:Replacing MSI with Legacy IRQ model commit: a523e1538fdd5f00ea3289cc0b3c6c1785b89814
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
participants (5)
-
Jaroslav Kysela -
Mark Brown -
Ravulapati Vishnu vardhan rao -
RAVULAPATI, VISHNU VARDHAN RAO
-
Takashi Iwai