[PATCH 0/2] ALSA: pcm: stream linking locking fixes
Two patches fixing locking issues for SNDRV_PCM_IOCTL_LINK handling: first adds a check preventing linking a stream to itself, second quiets lockdep warning about nested locks.
Michał Mirosław (2): ALSA: pcm: disallow linking stream to itself ALSA: pcm: fix snd_pcm_link() lockdep splat
sound/core/pcm_native.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-)
Add and use snd_pcm_stream_lock_nested() in snd_pcm_link/unlink implementation. The code is fine, but generates a lockdep complaint:
============================================ WARNING: possible recursive locking detected 5.7.1mq+ #381 Tainted: G O -------------------------------------------- pulseaudio/4180 is trying to acquire lock: ffff888402d6f508 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xda8/0xee0 [snd_pcm]
but task is already holding lock: ffff8883f7a8cf18 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xe4e/0xee0 [snd_pcm]
other info that might help us debug this: Possible unsafe locking scenario:
CPU0 ---- lock(&group->lock); lock(&group->lock);
*** DEADLOCK ***
May be due to missing lock nesting notation
2 locks held by pulseaudio/4180: #0: ffffffffa1a05190 (snd_pcm_link_rwsem){++++}-{3:3}, at: snd_pcm_common_ioctl+0xca0/0xee0 [snd_pcm] #1: ffff8883f7a8cf18 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xe4e/0xee0 [snd_pcm] [...]
Cc: stable@vger.kernel.org Fixes: f57f3df03a8e ("ALSA: pcm: More fine-grained PCM link locking") Signed-off-by: Michał Mirosław mirq-linux@rere.qmqm.pl --- sound/core/pcm_native.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)
diff --git a/sound/core/pcm_native.c b/sound/core/pcm_native.c index 3ad399cb6f30..5b36881f7095 100644 --- a/sound/core/pcm_native.c +++ b/sound/core/pcm_native.c @@ -138,6 +138,16 @@ void snd_pcm_stream_lock_irq(struct snd_pcm_substream *substream) } EXPORT_SYMBOL_GPL(snd_pcm_stream_lock_irq);
+static void snd_pcm_stream_lock_nested(struct snd_pcm_substream *substream) +{ + struct snd_pcm_group *group = &substream->self_group; + + if (substream->pcm->nonatomic) + mutex_lock_nested(&group->mutex, SINGLE_DEPTH_NESTING); + else + spin_lock_nested(&group->lock, SINGLE_DEPTH_NESTING); +} + /** * snd_pcm_stream_unlock_irq - Unlock the PCM stream * @substream: PCM substream @@ -2200,7 +2210,7 @@ static int snd_pcm_link(struct snd_pcm_substream *substream, int fd) snd_pcm_stream_unlock_irq(substream);
snd_pcm_group_lock_irq(target_group, nonatomic); - snd_pcm_stream_lock(substream1); + snd_pcm_stream_lock_nested(substream1); snd_pcm_group_assign(substream1, target_group); refcount_inc(&target_group->refs); snd_pcm_stream_unlock(substream1); @@ -2216,7 +2226,7 @@ static int snd_pcm_link(struct snd_pcm_substream *substream, int fd)
static void relink_to_local(struct snd_pcm_substream *substream) { - snd_pcm_stream_lock(substream); + snd_pcm_stream_lock_nested(substream); snd_pcm_group_assign(substream, &substream->self_group); snd_pcm_stream_unlock(substream); }
On Mon, 08 Jun 2020 12:06:32 +0200, Michał Mirosław wrote:
Add and use snd_pcm_stream_lock_nested() in snd_pcm_link/unlink implementation. The code is fine, but generates a lockdep complaint:
============================================ WARNING: possible recursive locking detected 5.7.1mq+ #381 Tainted: G O
pulseaudio/4180 is trying to acquire lock: ffff888402d6f508 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xda8/0xee0 [snd_pcm]
but task is already holding lock: ffff8883f7a8cf18 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xe4e/0xee0 [snd_pcm]
other info that might help us debug this: Possible unsafe locking scenario:
CPU0 ----
lock(&group->lock); lock(&group->lock);
*** DEADLOCK ***
May be due to missing lock nesting notation
2 locks held by pulseaudio/4180: #0: ffffffffa1a05190 (snd_pcm_link_rwsem){++++}-{3:3}, at: snd_pcm_common_ioctl+0xca0/0xee0 [snd_pcm] #1: ffff8883f7a8cf18 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xe4e/0xee0 [snd_pcm] [...]
Cc: stable@vger.kernel.org Fixes: f57f3df03a8e ("ALSA: pcm: More fine-grained PCM link locking") Signed-off-by: Michał Mirosław mirq-linux@rere.qmqm.pl
Applied now. Thanks.
Takashi
Prevent SNDRV_PCM_IOCTL_LINK linking stream to itself - the code can't handle it. Fixed commit is not where bug was introduced, but changes the context significantly.
Cc: stable@vger.kernel.org Fixes: 0888c321de70 ("pcm_native: switch to fdget()/fdput()") Signed-off-by: Michał Mirosław mirq-linux@rere.qmqm.pl --- sound/core/pcm_native.c | 6 ++++++ 1 file changed, 6 insertions(+)
diff --git a/sound/core/pcm_native.c b/sound/core/pcm_native.c index aef860256278..3ad399cb6f30 100644 --- a/sound/core/pcm_native.c +++ b/sound/core/pcm_native.c @@ -2166,6 +2166,12 @@ static int snd_pcm_link(struct snd_pcm_substream *substream, int fd) } pcm_file = f.file->private_data; substream1 = pcm_file->substream; + + if (substream == substream1) { + res = -EDEADLK; + goto _badf; + } + group = kzalloc(sizeof(*group), GFP_KERNEL); if (!group) { res = -ENOMEM;
On Mon, 08 Jun 2020 12:06:32 +0200, Michał Mirosław wrote:
Prevent SNDRV_PCM_IOCTL_LINK linking stream to itself - the code can't handle it. Fixed commit is not where bug was introduced, but changes the context significantly.
Cc: stable@vger.kernel.org Fixes: 0888c321de70 ("pcm_native: switch to fdget()/fdput()") Signed-off-by: Michał Mirosław mirq-linux@rere.qmqm.pl
Thanks for the fix. Just a minor point:
@@ -2166,6 +2166,12 @@ static int snd_pcm_link(struct snd_pcm_substream *substream, int fd) } pcm_file = f.file->private_data; substream1 = pcm_file->substream;
- if (substream == substream1) {
res = -EDEADLK;
We've never used this error code, hence it may confuse the user-space side. I'd use a more standard -EINVAL instead; the error is basically an invalid argument, after all.
thanks,
Takashi
Prevent SNDRV_PCM_IOCTL_LINK linking stream to itself - the code can't handle it. Fixed commit is not where bug was introduced, but changes the context significantly.
Cc: stable@vger.kernel.org Fixes: 0888c321de70 ("pcm_native: switch to fdget()/fdput()") Signed-off-by: Michał Mirosław mirq-linux@rere.qmqm.pl --- v2: EDEADLK -> EINVAL --- sound/core/pcm_native.c | 6 ++++++ 1 file changed, 6 insertions(+)
diff --git a/sound/core/pcm_native.c b/sound/core/pcm_native.c index aef860256278..434e7b604bad 100644 --- a/sound/core/pcm_native.c +++ b/sound/core/pcm_native.c @@ -2166,6 +2166,12 @@ static int snd_pcm_link(struct snd_pcm_substream *substream, int fd) } pcm_file = f.file->private_data; substream1 = pcm_file->substream; + + if (substream == substream1) { + res = -EINVAL; + goto _badf; + } + group = kzalloc(sizeof(*group), GFP_KERNEL); if (!group) { res = -ENOMEM;
On Mon, 08 Jun 2020 18:50:39 +0200, Michał Mirosław wrote:
Prevent SNDRV_PCM_IOCTL_LINK linking stream to itself - the code can't handle it. Fixed commit is not where bug was introduced, but changes the context significantly.
Cc: stable@vger.kernel.org Fixes: 0888c321de70 ("pcm_native: switch to fdget()/fdput()") Signed-off-by: Michał Mirosław mirq-linux@rere.qmqm.pl
v2: EDEADLK -> EINVAL
Applied now. Thanks!
Takashi
participants (2)
-
Michał Mirosław
-
Takashi Iwai