[PATCH 1/2] ASoC: dmaengine: Drop unused iov_iter for process callback
![](https://secure.gravatar.com/avatar/5b19e9d0e834ea10ef75803718ad564b.jpg?s=120&d=mm&r=g)
Passing the iov_iter to the process callback is rather buggy, as the iterator has been already processed for playback. Similarly, it makes the copy for capture buggy after the process callback reading the iterator out. Moreover, all existing process callbacks don't refer to the passed iterator at all. So, it's better to drop the argument from the process callback.
Fixes: 9bebd65443c1 ("ASoC: dmaengine: Use iov_iter for process callback, too") Suggested-by: Linus Torvalds torvalds@linux-foundation.org Link: https://lore.kernel.org/r/CAHk-=wje+VkXjjfVTmK-uJdG_M5=ar14QxAwK+XDiq07k_pzB... Signed-off-by: Takashi Iwai tiwai@suse.de --- include/sound/dmaengine_pcm.h | 2 +- sound/soc/atmel/mchp-pdmc.c | 2 +- sound/soc/soc-generic-dmaengine-pcm.c | 4 ++-- sound/soc/stm/stm32_sai_sub.c | 2 +- 4 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/include/sound/dmaengine_pcm.h b/include/sound/dmaengine_pcm.h index c9a8bce9a785..d70c55f17df7 100644 --- a/include/sound/dmaengine_pcm.h +++ b/include/sound/dmaengine_pcm.h @@ -142,7 +142,7 @@ struct snd_dmaengine_pcm_config { struct snd_pcm_substream *substream); int (*process)(struct snd_pcm_substream *substream, int channel, unsigned long hwoff, - struct iov_iter *buf, unsigned long bytes); + unsigned long bytes); dma_filter_fn compat_filter_fn; struct device *dma_dev; const char *chan_names[SNDRV_PCM_STREAM_LAST + 1]; diff --git a/sound/soc/atmel/mchp-pdmc.c b/sound/soc/atmel/mchp-pdmc.c index afe213a71212..dcc4e14b3dde 100644 --- a/sound/soc/atmel/mchp-pdmc.c +++ b/sound/soc/atmel/mchp-pdmc.c @@ -954,7 +954,7 @@ static int mchp_pdmc_dt_init(struct mchp_pdmc *dd) /* used to clean the channel index found on RHR's MSB */ static int mchp_pdmc_process(struct snd_pcm_substream *substream, int channel, unsigned long hwoff, - struct iov_iter *buf, unsigned long bytes) + unsigned long bytes) { struct snd_pcm_runtime *runtime = substream->runtime; u8 *dma_ptr = runtime->dma_area + hwoff + diff --git a/sound/soc/soc-generic-dmaengine-pcm.c b/sound/soc/soc-generic-dmaengine-pcm.c index ff2166525dbc..7a07fbf98e2e 100644 --- a/sound/soc/soc-generic-dmaengine-pcm.c +++ b/sound/soc/soc-generic-dmaengine-pcm.c @@ -296,7 +296,7 @@ static int dmaengine_copy(struct snd_soc_component *component, struct dmaengine_pcm *pcm = soc_component_to_pcm(component); int (*process)(struct snd_pcm_substream *substream, int channel, unsigned long hwoff, - struct iov_iter *buf, unsigned long bytes) = pcm->config->process; + unsigned long bytes) = pcm->config->process; bool is_playback = substream->stream == SNDRV_PCM_STREAM_PLAYBACK; void *dma_ptr = runtime->dma_area + hwoff + channel * (runtime->dma_bytes / runtime->channels); @@ -306,7 +306,7 @@ static int dmaengine_copy(struct snd_soc_component *component, return -EFAULT;
if (process) { - int ret = process(substream, channel, hwoff, buf, bytes); + int ret = process(substream, channel, hwoff, bytes); if (ret < 0) return ret; } diff --git a/sound/soc/stm/stm32_sai_sub.c b/sound/soc/stm/stm32_sai_sub.c index f9b5d5969155..0acc848c1f00 100644 --- a/sound/soc/stm/stm32_sai_sub.c +++ b/sound/soc/stm/stm32_sai_sub.c @@ -1246,7 +1246,7 @@ static const struct snd_soc_dai_ops stm32_sai_pcm_dai_ops2 = {
static int stm32_sai_pcm_process_spdif(struct snd_pcm_substream *substream, int channel, unsigned long hwoff, - struct iov_iter *buf, unsigned long bytes) + unsigned long bytes) { struct snd_pcm_runtime *runtime = substream->runtime; struct snd_soc_pcm_runtime *rtd = asoc_substream_to_rtd(substream);
![](https://secure.gravatar.com/avatar/5b19e9d0e834ea10ef75803718ad564b.jpg?s=120&d=mm&r=g)
While transitioning ASoC code for iov_iter usages, I kept the argument name as "buf" as the original code. But, iov_iter is an iterator, and using the name "buf" may be misleading: the crucial difference is that iov_iter can be proceeded after the operation, hence it can't be passed twice, while a simple "buffer" sounds as if reusable.
To make the usage clearer, rename the argument from "buf" to "iter". There is no functional changes, just names.
Fixes: 66201cacc33d ("ASoC: component: Add generic PCM copy ops") Suggested-by: Linus Torvalds torvalds@linux-foundation.org Link: https://lore.kernel.org/r/CAHk-=wje+VkXjjfVTmK-uJdG_M5=ar14QxAwK+XDiq07k_pzB... Signed-off-by: Takashi Iwai tiwai@suse.de --- include/sound/soc-component.h | 4 ++-- sound/soc/soc-component.c | 4 ++-- sound/soc/soc-generic-dmaengine-pcm.c | 6 +++--- 3 files changed, 7 insertions(+), 7 deletions(-)
diff --git a/include/sound/soc-component.h b/include/sound/soc-component.h index 17bea3144551..ceca69b46a82 100644 --- a/include/sound/soc-component.h +++ b/include/sound/soc-component.h @@ -139,7 +139,7 @@ struct snd_soc_component_driver { struct snd_pcm_audio_tstamp_report *audio_tstamp_report); int (*copy)(struct snd_soc_component *component, struct snd_pcm_substream *substream, int channel, - unsigned long pos, struct iov_iter *buf, + unsigned long pos, struct iov_iter *iter, unsigned long bytes); struct page *(*page)(struct snd_soc_component *component, struct snd_pcm_substream *substream, @@ -511,7 +511,7 @@ int snd_soc_pcm_component_ioctl(struct snd_pcm_substream *substream, int snd_soc_pcm_component_sync_stop(struct snd_pcm_substream *substream); int snd_soc_pcm_component_copy(struct snd_pcm_substream *substream, int channel, unsigned long pos, - struct iov_iter *buf, unsigned long bytes); + struct iov_iter *iter, unsigned long bytes); struct page *snd_soc_pcm_component_page(struct snd_pcm_substream *substream, unsigned long offset); int snd_soc_pcm_component_mmap(struct snd_pcm_substream *substream, diff --git a/sound/soc/soc-component.c b/sound/soc/soc-component.c index f18406dfa1e4..ba7c0ae82e00 100644 --- a/sound/soc/soc-component.c +++ b/sound/soc/soc-component.c @@ -1054,7 +1054,7 @@ int snd_soc_pcm_component_sync_stop(struct snd_pcm_substream *substream)
int snd_soc_pcm_component_copy(struct snd_pcm_substream *substream, int channel, unsigned long pos, - struct iov_iter *buf, unsigned long bytes) + struct iov_iter *iter, unsigned long bytes) { struct snd_soc_pcm_runtime *rtd = asoc_substream_to_rtd(substream); struct snd_soc_component *component; @@ -1065,7 +1065,7 @@ int snd_soc_pcm_component_copy(struct snd_pcm_substream *substream, if (component->driver->copy) return soc_component_ret(component, component->driver->copy(component, substream, - channel, pos, buf, bytes)); + channel, pos, iter, bytes));
return -EINVAL; } diff --git a/sound/soc/soc-generic-dmaengine-pcm.c b/sound/soc/soc-generic-dmaengine-pcm.c index 7a07fbf98e2e..d0653d775c87 100644 --- a/sound/soc/soc-generic-dmaengine-pcm.c +++ b/sound/soc/soc-generic-dmaengine-pcm.c @@ -290,7 +290,7 @@ static snd_pcm_uframes_t dmaengine_pcm_pointer( static int dmaengine_copy(struct snd_soc_component *component, struct snd_pcm_substream *substream, int channel, unsigned long hwoff, - struct iov_iter *buf, unsigned long bytes) + struct iov_iter *iter, unsigned long bytes) { struct snd_pcm_runtime *runtime = substream->runtime; struct dmaengine_pcm *pcm = soc_component_to_pcm(component); @@ -302,7 +302,7 @@ static int dmaengine_copy(struct snd_soc_component *component, channel * (runtime->dma_bytes / runtime->channels);
if (is_playback) - if (copy_from_iter(dma_ptr, bytes, buf) != bytes) + if (copy_from_iter(dma_ptr, bytes, iter) != bytes) return -EFAULT;
if (process) { @@ -312,7 +312,7 @@ static int dmaengine_copy(struct snd_soc_component *component, }
if (!is_playback) - if (copy_to_iter(dma_ptr, bytes, buf) != bytes) + if (copy_to_iter(dma_ptr, bytes, iter) != bytes) return -EFAULT;
return 0;
![](https://secure.gravatar.com/avatar/d930951cb00393ecf9c3db3a56d78fa9.jpg?s=120&d=mm&r=g)
On Thu, Aug 31, 2023 at 03:04:57PM +0200, Takashi Iwai wrote:
While transitioning ASoC code for iov_iter usages, I kept the argument name as "buf" as the original code. But, iov_iter is an iterator, and using the name "buf" may be misleading: the crucial difference is that iov_iter can be proceeded after the operation, hence it can't be passed twice, while a simple "buffer" sounds as if reusable.
Reviewed-by: Mark Brown broonie@kernel.org
![](https://secure.gravatar.com/avatar/d930951cb00393ecf9c3db3a56d78fa9.jpg?s=120&d=mm&r=g)
On Thu, Aug 31, 2023 at 03:04:56PM +0200, Takashi Iwai wrote:
Passing the iov_iter to the process callback is rather buggy, as the iterator has been already processed for playback. Similarly, it makes the copy for capture buggy after the process callback reading the iterator out. Moreover, all existing process callbacks don't refer to the passed iterator at all. So, it's better to drop the argument from the process callback.
Reviewed-by: Mark Brown broonie@kernel.org
participants (2)
-
Mark Brown
-
Takashi Iwai