* Can anyone test with AMD onboard or HDMI/DP?
@ 2022-06-17 12:29 Takashi Iwai
2022-06-17 12:40 ` Péter Ujfalusi
2022-06-17 13:56 ` Péter Ujfalusi
0 siblings, 2 replies; 4+ messages in thread
From: Takashi Iwai @ 2022-06-17 12:29 UTC (permalink / raw)
To: alsa-devel
Hi,
can anyone have an AMD onboard audio device and/or AMD HDMI/DP output
for testing a patch below with 5.18.x or 5.19-rc kernels? It's a
pending fix (for 5.18+), but currently it can't be verified whether it
causes a regression on the actual audio I/O (while it fixes the kernel
crash).
If the generic allocator still doesn't work as expected here, it
should show some audio stuttering or such effect.
Thanks!
Takashi
-- 8< --
From: Takashi Iwai <tiwai@suse.de>
Subject: [PATCH] ALSA: memalloc: Drop x86-specific hack for WC allocations
The recent report for a crash on Haswell machines implied that the
x86-specific (rather hackish) implementation for write-cache memory
buffer allocation in ALSA core is buggy with the recent kernel in some
corner cases. This patch drops the x86-specific implementation and
uses the standard dma_alloc_wc() & co generically for avoiding the bug
and also for simplification.
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=216112
Signed-off-by: Takashi Iwai <tiwai@suse.de>
---
sound/core/memalloc.c | 23 +----------------------
1 file changed, 1 insertion(+), 22 deletions(-)
diff --git a/sound/core/memalloc.c b/sound/core/memalloc.c
index 15dc7160ba34..8cfdaee77905 100644
--- a/sound/core/memalloc.c
+++ b/sound/core/memalloc.c
@@ -431,33 +431,17 @@ static const struct snd_malloc_ops snd_dma_iram_ops = {
*/
static void *snd_dma_dev_alloc(struct snd_dma_buffer *dmab, size_t size)
{
- void *p;
-
- p = dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
-#ifdef CONFIG_X86
- if (p && dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
- set_memory_wc((unsigned long)p, PAGE_ALIGN(size) >> PAGE_SHIFT);
-#endif
- return p;
+ return dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
}
static void snd_dma_dev_free(struct snd_dma_buffer *dmab)
{
-#ifdef CONFIG_X86
- if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
- set_memory_wb((unsigned long)dmab->area,
- PAGE_ALIGN(dmab->bytes) >> PAGE_SHIFT);
-#endif
dma_free_coherent(dmab->dev.dev, dmab->bytes, dmab->area, dmab->addr);
}
static int snd_dma_dev_mmap(struct snd_dma_buffer *dmab,
struct vm_area_struct *area)
{
-#ifdef CONFIG_X86
- if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
- area->vm_page_prot = pgprot_writecombine(area->vm_page_prot);
-#endif
return dma_mmap_coherent(dmab->dev.dev, area,
dmab->area, dmab->addr, dmab->bytes);
}
@@ -471,10 +455,6 @@ static const struct snd_malloc_ops snd_dma_dev_ops = {
/*
* Write-combined pages
*/
-#ifdef CONFIG_X86
-/* On x86, share the same ops as the standard dev ops */
-#define snd_dma_wc_ops snd_dma_dev_ops
-#else /* CONFIG_X86 */
static void *snd_dma_wc_alloc(struct snd_dma_buffer *dmab, size_t size)
{
return dma_alloc_wc(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
@@ -497,7 +477,6 @@ static const struct snd_malloc_ops snd_dma_wc_ops = {
.free = snd_dma_wc_free,
.mmap = snd_dma_wc_mmap,
};
-#endif /* CONFIG_X86 */
#ifdef CONFIG_SND_DMA_SGBUF
static void *snd_dma_sg_fallback_alloc(struct snd_dma_buffer *dmab, size_t size);
--
2.35.3
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: Can anyone test with AMD onboard or HDMI/DP?
2022-06-17 12:29 Can anyone test with AMD onboard or HDMI/DP? Takashi Iwai
@ 2022-06-17 12:40 ` Péter Ujfalusi
2022-06-17 13:56 ` Péter Ujfalusi
1 sibling, 0 replies; 4+ messages in thread
From: Péter Ujfalusi @ 2022-06-17 12:40 UTC (permalink / raw)
To: alsa-devel
On 17/06/2022 15:29, Takashi Iwai wrote:
> Hi,
>
> can anyone have an AMD onboard audio device and/or AMD HDMI/DP output
> for testing a patch below with 5.18.x or 5.19-rc kernels? It's a
> pending fix (for 5.18+), but currently it can't be verified whether it
> causes a regression on the actual audio I/O (while it fixes the kernel
> crash).
>
> If the generic allocator still doesn't work as expected here, it
> should show some audio stuttering or such effect.
But the fallback was needed for some machines using SOF to be able to
load the firmware...
like this:
https://github.com/thesofproject/linux/issues/3609
>
> Thanks!
>
> Takashi
>
> -- 8< --
> From: Takashi Iwai <tiwai@suse.de>
> Subject: [PATCH] ALSA: memalloc: Drop x86-specific hack for WC allocations
>
> The recent report for a crash on Haswell machines implied that the
> x86-specific (rather hackish) implementation for write-cache memory
> buffer allocation in ALSA core is buggy with the recent kernel in some
> corner cases. This patch drops the x86-specific implementation and
> uses the standard dma_alloc_wc() & co generically for avoiding the bug
> and also for simplification.
>
> BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=216112
> Signed-off-by: Takashi Iwai <tiwai@suse.de>
> ---
> sound/core/memalloc.c | 23 +----------------------
> 1 file changed, 1 insertion(+), 22 deletions(-)
>
> diff --git a/sound/core/memalloc.c b/sound/core/memalloc.c
> index 15dc7160ba34..8cfdaee77905 100644
> --- a/sound/core/memalloc.c
> +++ b/sound/core/memalloc.c
> @@ -431,33 +431,17 @@ static const struct snd_malloc_ops snd_dma_iram_ops = {
> */
> static void *snd_dma_dev_alloc(struct snd_dma_buffer *dmab, size_t size)
> {
> - void *p;
> -
> - p = dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> -#ifdef CONFIG_X86
> - if (p && dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> - set_memory_wc((unsigned long)p, PAGE_ALIGN(size) >> PAGE_SHIFT);
> -#endif
> - return p;
> + return dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> }
>
> static void snd_dma_dev_free(struct snd_dma_buffer *dmab)
> {
> -#ifdef CONFIG_X86
> - if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> - set_memory_wb((unsigned long)dmab->area,
> - PAGE_ALIGN(dmab->bytes) >> PAGE_SHIFT);
> -#endif
> dma_free_coherent(dmab->dev.dev, dmab->bytes, dmab->area, dmab->addr);
> }
>
> static int snd_dma_dev_mmap(struct snd_dma_buffer *dmab,
> struct vm_area_struct *area)
> {
> -#ifdef CONFIG_X86
> - if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> - area->vm_page_prot = pgprot_writecombine(area->vm_page_prot);
> -#endif
> return dma_mmap_coherent(dmab->dev.dev, area,
> dmab->area, dmab->addr, dmab->bytes);
> }
> @@ -471,10 +455,6 @@ static const struct snd_malloc_ops snd_dma_dev_ops = {
> /*
> * Write-combined pages
> */
> -#ifdef CONFIG_X86
> -/* On x86, share the same ops as the standard dev ops */
> -#define snd_dma_wc_ops snd_dma_dev_ops
> -#else /* CONFIG_X86 */
> static void *snd_dma_wc_alloc(struct snd_dma_buffer *dmab, size_t size)
> {
> return dma_alloc_wc(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> @@ -497,7 +477,6 @@ static const struct snd_malloc_ops snd_dma_wc_ops = {
> .free = snd_dma_wc_free,
> .mmap = snd_dma_wc_mmap,
> };
> -#endif /* CONFIG_X86 */
>
> #ifdef CONFIG_SND_DMA_SGBUF
> static void *snd_dma_sg_fallback_alloc(struct snd_dma_buffer *dmab, size_t size);
--
Péter
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Can anyone test with AMD onboard or HDMI/DP?
2022-06-17 12:29 Can anyone test with AMD onboard or HDMI/DP? Takashi Iwai
2022-06-17 12:40 ` Péter Ujfalusi
@ 2022-06-17 13:56 ` Péter Ujfalusi
2022-06-17 14:37 ` Takashi Iwai
1 sibling, 1 reply; 4+ messages in thread
From: Péter Ujfalusi @ 2022-06-17 13:56 UTC (permalink / raw)
To: Takashi Iwai, alsa-devel
Hi Takashi,
On 17/06/2022 15:29, Takashi Iwai wrote:
> Hi,
Try #2, first mail did not made it through, I think.
> can anyone have an AMD onboard audio device and/or AMD HDMI/DP output
> for testing a patch below with 5.18.x or 5.19-rc kernels? It's a
> pending fix (for 5.18+), but currently it can't be verified whether it
> causes a regression on the actual audio I/O (while it fixes the kernel
> crash).
>
> If the generic allocator still doesn't work as expected here, it
> should show some audio stuttering or such effect.
But the fallback was needed for some machines using SOF to be able to
load the firmware...
like this:
https://github.com/thesofproject/linux/issues/3609
>
>
> Thanks!
>
> Takashi
>
> -- 8< --
> From: Takashi Iwai <tiwai@suse.de>
> Subject: [PATCH] ALSA: memalloc: Drop x86-specific hack for WC allocations
>
> The recent report for a crash on Haswell machines implied that the
> x86-specific (rather hackish) implementation for write-cache memory
> buffer allocation in ALSA core is buggy with the recent kernel in some
> corner cases. This patch drops the x86-specific implementation and
> uses the standard dma_alloc_wc() & co generically for avoiding the bug
> and also for simplification.
>
> BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=216112
> Signed-off-by: Takashi Iwai <tiwai@suse.de>
> ---
> sound/core/memalloc.c | 23 +----------------------
> 1 file changed, 1 insertion(+), 22 deletions(-)
>
> diff --git a/sound/core/memalloc.c b/sound/core/memalloc.c
> index 15dc7160ba34..8cfdaee77905 100644
> --- a/sound/core/memalloc.c
> +++ b/sound/core/memalloc.c
> @@ -431,33 +431,17 @@ static const struct snd_malloc_ops snd_dma_iram_ops = {
> */
> static void *snd_dma_dev_alloc(struct snd_dma_buffer *dmab, size_t size)
> {
> - void *p;
> -
> - p = dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> -#ifdef CONFIG_X86
> - if (p && dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> - set_memory_wc((unsigned long)p, PAGE_ALIGN(size) >> PAGE_SHIFT);
> -#endif
> - return p;
> + return dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> }
>
> static void snd_dma_dev_free(struct snd_dma_buffer *dmab)
> {
> -#ifdef CONFIG_X86
> - if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> - set_memory_wb((unsigned long)dmab->area,
> - PAGE_ALIGN(dmab->bytes) >> PAGE_SHIFT);
> -#endif
> dma_free_coherent(dmab->dev.dev, dmab->bytes, dmab->area, dmab->addr);
> }
>
> static int snd_dma_dev_mmap(struct snd_dma_buffer *dmab,
> struct vm_area_struct *area)
> {
> -#ifdef CONFIG_X86
> - if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> - area->vm_page_prot = pgprot_writecombine(area->vm_page_prot);
> -#endif
> return dma_mmap_coherent(dmab->dev.dev, area,
> dmab->area, dmab->addr, dmab->bytes);
> }
> @@ -471,10 +455,6 @@ static const struct snd_malloc_ops snd_dma_dev_ops = {
> /*
> * Write-combined pages
> */
> -#ifdef CONFIG_X86
> -/* On x86, share the same ops as the standard dev ops */
> -#define snd_dma_wc_ops snd_dma_dev_ops
> -#else /* CONFIG_X86 */
> static void *snd_dma_wc_alloc(struct snd_dma_buffer *dmab, size_t size)
> {
> return dma_alloc_wc(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> @@ -497,7 +477,6 @@ static const struct snd_malloc_ops snd_dma_wc_ops = {
> .free = snd_dma_wc_free,
> .mmap = snd_dma_wc_mmap,
> };
> -#endif /* CONFIG_X86 */
>
> #ifdef CONFIG_SND_DMA_SGBUF
> static void *snd_dma_sg_fallback_alloc(struct snd_dma_buffer *dmab, size_t size);
--
Péter
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Can anyone test with AMD onboard or HDMI/DP?
2022-06-17 13:56 ` Péter Ujfalusi
@ 2022-06-17 14:37 ` Takashi Iwai
0 siblings, 0 replies; 4+ messages in thread
From: Takashi Iwai @ 2022-06-17 14:37 UTC (permalink / raw)
To: Péter Ujfalusi; +Cc: alsa-devel
On Fri, 17 Jun 2022 15:56:31 +0200,
Péter Ujfalusi wrote:
>
> Hi Takashi,
>
> On 17/06/2022 15:29, Takashi Iwai wrote:
> > Hi,
>
> Try #2, first mail did not made it through, I think.
>
> > can anyone have an AMD onboard audio device and/or AMD HDMI/DP output
> > for testing a patch below with 5.18.x or 5.19-rc kernels? It's a
> > pending fix (for 5.18+), but currently it can't be verified whether it
> > causes a regression on the actual audio I/O (while it fixes the kernel
> > crash).
> >
> > If the generic allocator still doesn't work as expected here, it
> > should show some audio stuttering or such effect.
>
> But the fallback was needed for some machines using SOF to be able to
> load the firmware...
> like this:
> https://github.com/thesofproject/linux/issues/3609
The bug isn't about the fallback of SG buffer we've already resolved,
but rather about the continuous WC page allocations that happen for
the CORB/RIRB or the position buffer on some devices like AMD.
The Intel HD-audio doesn't hit the problem.
Takashi
>
> >
> >
> > Thanks!
> >
> > Takashi
> >
> > -- 8< --
> > From: Takashi Iwai <tiwai@suse.de>
> > Subject: [PATCH] ALSA: memalloc: Drop x86-specific hack for WC allocations
> >
> > The recent report for a crash on Haswell machines implied that the
> > x86-specific (rather hackish) implementation for write-cache memory
> > buffer allocation in ALSA core is buggy with the recent kernel in some
> > corner cases. This patch drops the x86-specific implementation and
> > uses the standard dma_alloc_wc() & co generically for avoiding the bug
> > and also for simplification.
> >
> > BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=216112
> > Signed-off-by: Takashi Iwai <tiwai@suse.de>
> > ---
> > sound/core/memalloc.c | 23 +----------------------
> > 1 file changed, 1 insertion(+), 22 deletions(-)
> >
> > diff --git a/sound/core/memalloc.c b/sound/core/memalloc.c
> > index 15dc7160ba34..8cfdaee77905 100644
> > --- a/sound/core/memalloc.c
> > +++ b/sound/core/memalloc.c
> > @@ -431,33 +431,17 @@ static const struct snd_malloc_ops snd_dma_iram_ops = {
> > */
> > static void *snd_dma_dev_alloc(struct snd_dma_buffer *dmab, size_t size)
> > {
> > - void *p;
> > -
> > - p = dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> > -#ifdef CONFIG_X86
> > - if (p && dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> > - set_memory_wc((unsigned long)p, PAGE_ALIGN(size) >> PAGE_SHIFT);
> > -#endif
> > - return p;
> > + return dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> > }
> >
> > static void snd_dma_dev_free(struct snd_dma_buffer *dmab)
> > {
> > -#ifdef CONFIG_X86
> > - if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> > - set_memory_wb((unsigned long)dmab->area,
> > - PAGE_ALIGN(dmab->bytes) >> PAGE_SHIFT);
> > -#endif
> > dma_free_coherent(dmab->dev.dev, dmab->bytes, dmab->area, dmab->addr);
> > }
> >
> > static int snd_dma_dev_mmap(struct snd_dma_buffer *dmab,
> > struct vm_area_struct *area)
> > {
> > -#ifdef CONFIG_X86
> > - if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC)
> > - area->vm_page_prot = pgprot_writecombine(area->vm_page_prot);
> > -#endif
> > return dma_mmap_coherent(dmab->dev.dev, area,
> > dmab->area, dmab->addr, dmab->bytes);
> > }
> > @@ -471,10 +455,6 @@ static const struct snd_malloc_ops snd_dma_dev_ops = {
> > /*
> > * Write-combined pages
> > */
> > -#ifdef CONFIG_X86
> > -/* On x86, share the same ops as the standard dev ops */
> > -#define snd_dma_wc_ops snd_dma_dev_ops
> > -#else /* CONFIG_X86 */
> > static void *snd_dma_wc_alloc(struct snd_dma_buffer *dmab, size_t size)
> > {
> > return dma_alloc_wc(dmab->dev.dev, size, &dmab->addr, DEFAULT_GFP);
> > @@ -497,7 +477,6 @@ static const struct snd_malloc_ops snd_dma_wc_ops = {
> > .free = snd_dma_wc_free,
> > .mmap = snd_dma_wc_mmap,
> > };
> > -#endif /* CONFIG_X86 */
> >
> > #ifdef CONFIG_SND_DMA_SGBUF
> > static void *snd_dma_sg_fallback_alloc(struct snd_dma_buffer *dmab, size_t size);
>
> --
> Péter
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2022-06-17 17:19 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2022-06-17 12:29 Can anyone test with AMD onboard or HDMI/DP? Takashi Iwai
2022-06-17 12:40 ` Péter Ujfalusi
2022-06-17 13:56 ` Péter Ujfalusi
2022-06-17 14:37 ` Takashi Iwai
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox