From: Takashi Iwai <tiwai@xxxxxxx> commit cbea6e5a7772b7a5b80baa8f98fd77853487fd2a upstream. Currently we check only the substream->dma_buffer as the preset of the buffer configuration for verifying the availability of mmap. But a few drivers rather set up the buffer in the own way without the standard buffer preallocation using substream->dma_buffer, and they miss the proper checks. (Now it's working more or less fine as most of them are running only on x86). Actually, they may set up the runtime dma_buffer (referred via snd_pcm_get_dma_buf()) at the open callback, though. That is, this could have been used as the primary source. This patch changes the hw_support_mmap() function to check the runtime dma buffer at first. It's usually NULL with the standard buffer preallocation, and in that case, we continue checking substream->dma_buffer as fallback. Link: https://lore.kernel.org/r/20210809071829.22238-2-tiwai@xxxxxxx Signed-off-by: Takashi Iwai <tiwai@xxxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- sound/core/pcm_native.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) --- a/sound/core/pcm_native.c +++ b/sound/core/pcm_native.c @@ -243,13 +243,18 @@ int snd_pcm_info_user(struct snd_pcm_sub static bool hw_support_mmap(struct snd_pcm_substream *substream) { + struct snd_dma_buffer *dmabuf; + if (!(substream->runtime->hw.info & SNDRV_PCM_INFO_MMAP)) return false; if (substream->ops->mmap || substream->ops->page) return true; - switch (substream->dma_buffer.dev.type) { + dmabuf = snd_pcm_get_dma_buf(substream); + if (!dmabuf) + dmabuf = &substream->dma_buffer; + switch (dmabuf->dev.type) { case SNDRV_DMA_TYPE_UNKNOWN: /* we can't know the device, so just assume that the driver does * everything right @@ -259,7 +264,7 @@ static bool hw_support_mmap(struct snd_p case SNDRV_DMA_TYPE_VMALLOC: return true; default: - return dma_can_mmap(substream->dma_buffer.dev.dev); + return dma_can_mmap(dmabuf->dev.dev); } }
