From: Andres Salomon <dilinger@queued.net>

We're never actually setting dma->substream to the current substream; that
means the dma->substream checks that we do in the suspend/resume path are
never satisfied, and the PRD registers are never correctly managed.  This
changes it so that we set the substream when constructing the specific bus
master DMA, and unsetting it when we tear down the BM's DMA.

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Jordan Crouse <jordan.crouse@amd.com>
Acked-by: Jaya Kumar <jayakumar.alsa@gmail.com>
Cc: Jaroslav Kysela <perex@suse.cz>
Cc: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 sound/pci/cs5535audio/cs5535audio_pcm.c |    2 ++
 1 files changed, 2 insertions(+)

diff -puN sound/pci/cs5535audio/cs5535audio_pcm.c~alsa-cs5535audio-correctly-set-dma-substream sound/pci/cs5535audio/cs5535audio_pcm.c
--- a/sound/pci/cs5535audio/cs5535audio_pcm.c~alsa-cs5535audio-correctly-set-dma-substream
+++ a/sound/pci/cs5535audio/cs5535audio_pcm.c
@@ -161,6 +161,7 @@ static int cs5535audio_build_dma_packets
 	jmpprd_addr = cpu_to_le32(lastdesc->addr +
 				  (sizeof(struct cs5535audio_dma_desc)*periods));
 
+	dma->substream = substream;
 	dma->period_bytes = period_bytes;
 	dma->periods = periods;
 	spin_lock_irq(&cs5535au->reg_lock);
@@ -238,6 +239,7 @@ static void cs5535audio_clear_dma_packet
 {
 	snd_dma_free_pages(&dma->desc_buf);
 	dma->desc_buf.area = NULL;
+	dma->substream = NULL;
 }
 
 static int snd_cs5535audio_hw_params(struct snd_pcm_substream *substream,
_