The code refactoring to move the WC page handling into the common memalloc layer caused a breakage for HD-audio HDMI. I overlooked that the driver is using the SG-buffer, which isn't covered by the patch.
This patch adds the mmap workaround for WC pages to SG-buffer handler. A caveat is that it falls back to the default handler by returning an error after setting the pgprot, so it won't work in all cases but merely for PCM (which is currently the only use case).
Fixes: 623c10108338 ("ALSA: memalloc: Fix pgprot for WC mmap on x86") Reported-and-tested-by: Andy Lavr andy.lavr@gmail.com Signed-off-by: Takashi Iwai tiwai@suse.de --- sound/core/sgbuf.c | 9 +++++++++ 1 file changed, 9 insertions(+)
diff --git a/sound/core/sgbuf.c b/sound/core/sgbuf.c index 9666a95f5a5c..8352a5cdb19f 100644 --- a/sound/core/sgbuf.c +++ b/sound/core/sgbuf.c @@ -183,10 +183,19 @@ static unsigned int snd_dma_sg_get_chunk_size(struct snd_dma_buffer *dmab, return size; }
+static int snd_dma_sg_mmap(struct snd_dma_buffer *dmab, + struct vm_area_struct *area) +{ + if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_WC_SG) + area->vm_page_prot = pgprot_writecombine(area->vm_page_prot); + return -ENOENT; /* continue with the default mmap handler */ +} + const struct snd_malloc_ops snd_dma_sg_ops = { .alloc = snd_dma_sg_alloc, .free = snd_dma_sg_free, .get_addr = snd_dma_sg_get_addr, .get_page = snd_dma_sg_get_page, .get_chunk_size = snd_dma_sg_get_chunk_size, + .mmap = snd_dma_sg_mmap, };