Skip to contentSkip to conversation

ALSA: compress: fix buffer handling and state validation

PATCH Applied Peter Ujfalusi 2 participants 6 messages Oct 19 – Oct 20

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v2 cover ALSA: compress: fix buffer handling and state validation

Hi,

Changes since v1:
- patch 1: do not free the old buffer in snd_compr_allocate_buffer(). SET_PARAMS can only be repeated in OPEN state after ops->set_params() has failed, so release the framework-owned buffer on that error path instead. The buffer is allocated once per successful SET_PARAMS and is never replaced (Vinod, Takashi)
- patch 2: reworded the commit message: the buffer is set up once and the next_track path does not re-allocate it. The geometry check is kept because ops->set_params() would otherwise receive buffer parameters that do not match the buffer in use. tinycompress always passes the open-time fragment_size/fragments on next_track, so existing users are not affected (Vinod)
- patches 3 and 4: added Acked-by from Vinod

The compress offload core has several edge cases in buffer management and state validation.

A SET_PARAMS that the driver rejects leaves the framework-owned buffer allocated while the stream stays in the OPEN state, so a following SET_PARAMS overwrites the pointer and leaks the buffer. During gapless playback, SET_PARAMS can also change the buffer geometry while the byte counters still refer to the current buffer. Separately, SET_METADATA can reach the driver before SET_PARAMS, and the origin_seqno task lookup can restart a task that is still ACTIVE.

Address these cases by:
- releasing the framework-owned buffer when ops->set_params() fails, so the buffer is allocated once per successful SET_PARAMS;
- not re-allocating the buffer on the gapless next_track path and rejecting a buffer geometry change there;
- rejecting SET_METADATA while the stream is OPEN; and
- allowing origin_seqno to reuse IDLE or FINISHED tasks, but not ACTIVE ones.

Regards,
Peter
---
Peter Ujfalusi (4):
ALSA: compress: fix buffer leak on failed SET_PARAMS
ALSA: compress: reject buffer geometry change on gapless next_track
ALSA: compress: add missing OPEN-state gate to SET_METADATA
ALSA: compress: reject restarting an ACTIVE task via origin_seqno

 sound/core/compress_offload.c | 32 +++++++++++++++++++++++++++-----
 1 file changed, 27 insertions(+), 5 deletions(-)
Signature · 1 line
2.56.0
#1 lore raw

PATCH v2 1/4 ALSA: compress: fix buffer leak on failed SET_PARAMS

When the driver has neither ops->copy nor a preallocated dma_buffer_p, snd_compr_allocate_buffer() kmallocs runtime->buffer. If ops->set_params() then rejects the parameters, SET_PARAMS fails, the stream stays in the OPEN state and keeps the buffer. The next SET_PARAMS allocates a new buffer and overwrites the pointer, so repeated SET_PARAMS calls with parameters that the driver rejects leak one buffer per call.

Release the framework-owned buffer when ops->set_params() fails on an OPEN stream, so that a failed SET_PARAMS leaves no buffer behind and the buffer is allocated only once per successful setup. Use the same ownership check as snd_compr_free().

Fixes b21c60a4edd2 ("ALSA: core: add support for compress_offload")
---
 sound/core/compress_offload.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
sound/core/compress_offload.c +8 -1
index 5c0db178d635..50301f72579e 100644
@@ -702,8 +702,15 @@ snd_compr_set_params(struct snd_compr_stream *stream, unsigned long arg)
 			return -ENOMEM;
 
 		retval = stream->ops->set_params(stream, params);
-		if (retval)
+		if (retval) {
+			/* the stream stays OPEN: drop the buffer of the rejected setup */
+			if (stream->runtime->state == SNDRV_PCM_STATE_OPEN &&
+			    !stream->runtime->dma_buffer_p) {
+				kfree(stream->runtime->buffer);
+				stream->runtime->buffer = NULL;
+			}
 			return retval;
+		}
 
 		if (stream->next_track)
 			return retval;
Signature · 1 line
2.56.0
#2 lore raw

PATCH v2 2/4 ALSA: compress: reject buffer geometry change on gapless next_track

In reply to Peter Ujfalusi

snd_compr_set_params() is invoked from the gapless next_track re-entry path, which calls snd_compr_allocate_buffer() again and lets buffer_size change while total_bytes_available and total_bytes_transferred still hold values computed for the previous buffer. snd_compr_calc_avail() then derives an avail/count value from those stale counters against the new buffer_size, and snd_compr_write_data() copy_from_user()s that many bytes into the new, potentially smaller buffer, overflowing it with user-controlled length and content.

The buffer is set up once, with the initial SET_PARAMS, and neither the core nor the DSP expects it to change for a gapless track transition. Do not call snd_compr_allocate_buffer() on the next_track path and reject a request that changes the buffer geometry, so that the codec parameters that the driver receives match the buffer in use.

Fixes 9727b490e543 ("ALSA: compress: add support for gapless playback")
---
 sound/core/compress_offload.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)
sound/core/compress_offload.c +13 -3
index 50301f72579e..d6863336f21c 100644
@@ -697,9 +697,19 @@ snd_compr_set_params(struct snd_compr_stream *stream, unsigned long arg)
 		if (retval)
 			return retval;
 
-		retval = snd_compr_allocate_buffer(stream, params);
-		if (retval)
-			return -ENOMEM;
+		if (stream->next_track) {
+			/*
+			 * total_bytes_available/transferred still refer to the
+			 * current buffer; its geometry must not change.
+			 */
+			if (params->buffer.fragment_size != stream->runtime->fragment_size ||
+			    params->buffer.fragments != stream->runtime->fragments)
+				return -EINVAL;
+		} else {
+			retval = snd_compr_allocate_buffer(stream, params);
+			if (retval)
+				return retval;
+		}
 
 		retval = stream->ops->set_params(stream, params);
 		if (retval) {
Signature · 1 line
2.56.0
#3 lore raw

PATCH v2 3/4 ALSA: compress: add missing OPEN-state gate to SET_METADATA

In reply to Peter Ujfalusi

snd_compr_set_metadata() calls stream->ops->set_metadata() straight after copy_from_user(), without ever checking runtime->state, even though the function's own comment says a parameter change should only be allowed once the stream has left the OPEN state. Every other metadata/param entry point in this file (SET_PARAMS, WRITE, ...) enforces that gate.

A codec driver's set_metadata callback can assume SET_PARAMS has already run and initialised its private state; calling it while the stream is still OPEN can operate on that uninitialised state.

Add the OPEN-state check the existing comment already documents.

Fixes 9727b490e543 ("ALSA: compress: add support for gapless playback")
---
 sound/core/compress_offload.c | 3 +++
 1 file changed, 3 insertions(+)
sound/core/compress_offload.c +3 -0
index d6863336f21c..a250c6b53996 100644
@@ -789,6 +789,9 @@ snd_compr_set_metadata(struct snd_compr_stream *stream, unsigned long arg)
 	* we should allow parameter change only when stream has been
 	* opened not in other cases
 	*/
+	if (stream->runtime->state == SNDRV_PCM_STATE_OPEN)
+		return -EBADFD;
+
 	if (copy_from_user(&metadata, (void __user *)arg, sizeof(metadata)))
 		return -EFAULT;
 
Signature · 1 line
2.56.0
#4 lore raw

PATCH v2 4/4 ALSA: compress: reject restarting an ACTIVE task via origin_seqno

In reply to Peter Ujfalusi

snd_compr_task_start_prepare() only rejects tasks with state >= SND_COMPRESS_TASK_STATE_FINISHED. Since the enum is IDLE(0) < ACTIVE(1) < FINISHED(2), this accepts ACTIVE tasks and rejects FINISHED ones, the opposite of what the origin_seqno mechanism needs: origin_seqno is meant to let a caller reuse a FINISHED task's buffers for a new job, not to restart a task that is still queued with the driver.

The seqno lookup path in snd_compr_task_start() has its own explicit "task->state != IDLE" guard, but the origin_seqno lookup path relies solely on this function, so it can restart an ACTIVE task, causing ops->task_start() to be called twice for the same task and runtime->active_tasks to be incremented twice without a matching finish/stop.

Accept IDLE and FINISHED, reject ACTIVE, on both entry paths.

Fixes 04177158cf98 ("ALSA: compress_offload: introduce accel operation mode")
---
 sound/core/compress_offload.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
sound/core/compress_offload.c +3 -1
index a250c6b53996..5e5e0f78f26d 100644
@@ -1206,7 +1206,9 @@ static int snd_compr_task_start_prepare(struct snd_compr_task_runtime *task,
 {
 	if (task == NULL)
 		return -EINVAL;
-	if (task->state >= SND_COMPRESS_TASK_STATE_FINISHED)
+	/* only an unqueued or already finished task may be (re)started */
+	if (task->state != SND_COMPRESS_TASK_STATE_IDLE &&
+	    task->state != SND_COMPRESS_TASK_STATE_FINISHED)
 		return -EBUSY;
 	if (utask->input_size > task->input->size)
 		return -EINVAL;
Signature · 1 line
2.56.0
Applied
#5 lore raw
In reply to Peter Ujfalusi
Quoted 37 lines · Peter Ujfalusi

Hi,

Changes since v1:
- patch 1: do not free the old buffer in snd_compr_allocate_buffer(). SET_PARAMS can only be repeated in OPEN state after ops->set_params() has failed, so release the framework-owned buffer on that error path instead. The buffer is allocated once per successful SET_PARAMS and is never replaced (Vinod, Takashi)
- patch 2: reworded the commit message: the buffer is set up once and the next_track path does not re-allocate it. The geometry check is kept because ops->set_params() would otherwise receive buffer parameters that do not match the buffer in use. tinycompress always passes the open-time fragment_size/fragments on next_track, so existing users are not affected (Vinod)
- patches 3 and 4: added Acked-by from Vinod

The compress offload core has several edge cases in buffer management and state validation.

A SET_PARAMS that the driver rejects leaves the framework-owned buffer allocated while the stream stays in the OPEN state, so a following SET_PARAMS overwrites the pointer and leaks the buffer. During gapless playback, SET_PARAMS can also change the buffer geometry while the byte counters still refer to the current buffer. Separately, SET_METADATA can reach the driver before SET_PARAMS, and the origin_seqno task lookup can restart a task that is still ACTIVE.

Address these cases by:
- releasing the framework-owned buffer when ops->set_params() fails, so the buffer is allocated once per successful SET_PARAMS;
- not re-allocating the buffer on the gapless next_track path and rejecting a buffer geometry change there;
- rejecting SET_METADATA while the stream is OPEN; and
- allowing origin_seqno to reuse IDLE or FINISHED tasks, but not ACTIVE ones.

Quoted 8 lines

Regards,
Peter
---
Peter Ujfalusi (4):
ALSA: compress: fix buffer leak on failed SET_PARAMS
ALSA: compress: reject buffer geometry change on gapless next_track
ALSA: compress: add missing OPEN-state gate to SET_METADATA
ALSA: compress: reject restarting an ACTIVE task via origin_seqno

Applied all four patches now. Thanks.

Takashi