[PATCH] ALSA: sh: aica: Fix spu_memset() stride

[PATCH] ALSA: sh: aica: Fix spu_memset() stride

[Florian Fuchs]

spu_memset() writes one 32-bit word per iteration, but advances the
destination offset by only one byte. spu_reset() calls it with
0x200000 / 4 words to clear the 2 MiB of sound RAM, so:

 - three out of four writel() calls target an unaligned address. The
   SH4 can't do that. Each one traps and is emulated by the unaligned
   access fixup. This call to spu_memset() causes 393216 unaligned
   accesses, visible in /proc/cpu/alignment.
 - the clear covered only about 512 KiB of sound RAM, not 2 MiB.

Advance the offset by the size of the written word, to clear the full
sound RAM using aligned writes.

Fixes: 198de43d758c ("[ALSA] Add ALSA support for the SEGA Dreamcast PCM device")
Signed-off-by: Florian Fuchs <fuchsfl@gmail.com>
---
The patch has been tested on the hardware and sound is also still
working well, the boot is faster by about 4s due to less unaligned
fixups.

Previously: 7.9s to boot:
[    2.900000] AICA AICA: ALSA Driver for Yamaha AICA Super Intelligent Sound Processor
..
[    7.872000] ALSA device list:
[    7.876000]   #1: Yamaha AICA Super Intelligent Sound Processor for SEGA Dreamcast

After: 3.8s to boot:
[    2.892000] AICA AICA: ALSA Driver for Yamaha AICA Super Intelligent Sound Processor
..
[    3.796000] ALSA device list:
[    3.800000]   #1: Yamaha AICA Super Intelligent Sound Processor for SEGA Dreamcast

 sound/sh/aica.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sound/sh/aica.c b/sound/sh/aica.c
index 078fc7fc08f9..04f80c8c89c0 100644
--- a/sound/sh/aica.c
+++ b/sound/sh/aica.c
@@ -94,7 +94,7 @@ static void spu_memset(u32 toi, u32 what, int length)
 		local_irq_save(flags);
 		writel(what, toi + SPU_MEMORY_BASE);
 		local_irq_restore(flags);
-		toi++;
+		toi += sizeof(what);
 	}
 }
 

base-commit: c1b1c25de0d6008c3fcabc942fba635ea3693c3f
-- 
2.43.0

Re: [PATCH] ALSA: sh: aica: Fix spu_memset() stride

[Adrian McMenamin]

On Thu, 22 Oct 2026 at 17:46, Florian Fuchs <fuchsfl@gmail.com> wrote:
>
> spu_memset() writes one 32-bit word per iteration, but advances the
> destination offset by only one byte. spu_reset() calls it with
> 0x200000 / 4 words to clear the 2 MiB of sound RAM, so:
>
>  - three out of four writel() calls target an unaligned address. The
>    SH4 can't do that. Each one traps and is emulated by the unaligned
>    access fixup. This call to spu_memset() causes 393216 unaligned
>    accesses, visible in /proc/cpu/alignment.
>  - the clear covered only about 512 KiB of sound RAM, not 2 MiB.
>
> Advance the offset by the size of the written word, to clear the full
> sound RAM using aligned writes.
>
> Fixes: 198de43d758c ("[ALSA] Add ALSA support for the SEGA Dreamcast PCM device")
> Signed-off-by: Florian Fuchs <fuchsfl@gmail.com>
> ---
> The patch has been tested on the hardware and sound is also still
> working well, the boot is faster by about 4s due to less unaligned
> fixups.
>
> Previously: 7.9s to boot:
> [    2.900000] AICA AICA: ALSA Driver for Yamaha AICA Super Intelligent Sound Processor
> ..
> [    7.872000] ALSA device list:
> [    7.876000]   #1: Yamaha AICA Super Intelligent Sound Processor for SEGA Dreamcast
>
> After: 3.8s to boot:
> [    2.892000] AICA AICA: ALSA Driver for Yamaha AICA Super Intelligent Sound Processor
> ..
> [    3.796000] ALSA device list:
> [    3.800000]   #1: Yamaha AICA Super Intelligent Sound Processor for SEGA Dreamcast
>
>  sound/sh/aica.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/sound/sh/aica.c b/sound/sh/aica.c
> index 078fc7fc08f9..04f80c8c89c0 100644
> --- a/sound/sh/aica.c
> +++ b/sound/sh/aica.c
> @@ -94,7 +94,7 @@ static void spu_memset(u32 toi, u32 what, int length)
>                 local_irq_save(flags);
>                 writel(what, toi + SPU_MEMORY_BASE);
>                 local_irq_restore(flags);
> -               toi++;
> +               toi += sizeof(what);
>         }
>  }
>
>
> base-commit: c1b1c25de0d6008c3fcabc942fba635ea3693c3f
> --
> 2.43.0
>

Yes, I don't know what I was thinking there - nobody who reviewed the
code picked it up either, though.

The trapping presumably stops the whole thing from falling over and
answers the "how on Earth did this ever work" question.

Acked-by: Adrian McMenamin <adrianmcmenamin@gmail.com>

Re: [PATCH] ALSA: sh: aica: Fix spu_memset() stride

[Takashi Iwai]

On Thu, 22 Oct 2026 18:32:41 +0200,
Florian Fuchs wrote:
> 
> spu_memset() writes one 32-bit word per iteration, but advances the
> destination offset by only one byte. spu_reset() calls it with
> 0x200000 / 4 words to clear the 2 MiB of sound RAM, so:
> 
>  - three out of four writel() calls target an unaligned address. The
>    SH4 can't do that. Each one traps and is emulated by the unaligned
>    access fixup. This call to spu_memset() causes 393216 unaligned
>    accesses, visible in /proc/cpu/alignment.
>  - the clear covered only about 512 KiB of sound RAM, not 2 MiB.
> 
> Advance the offset by the size of the written word, to clear the full
> sound RAM using aligned writes.
> 
> Fixes: 198de43d758c ("[ALSA] Add ALSA support for the SEGA Dreamcast PCM device")
> Signed-off-by: Florian Fuchs <fuchsfl@gmail.com>

Applied now.  Thanks.


Takashi