[alsa-devel] [PATCH] ALSA: seq: Enable 'use' locking in all configurations
The 'use' locking macros are no-ops if neither SMP or SND_DEBUG is enabled. This might once have been OK in non-preemptible configurations, but even in that case snd_seq_read() may sleep while relying on a 'use' lock. So always use the proper implementations.
Cc: stable@vger.kernel.org Signed-off-by: Ben Hutchings ben.hutchings@codethink.co.uk --- sound/core/seq/seq_lock.c | 4 ---- sound/core/seq/seq_lock.h | 12 ------------ 2 files changed, 16 deletions(-)
diff --git a/sound/core/seq/seq_lock.c b/sound/core/seq/seq_lock.c index 0ff7926a5a69..cda64b489e42 100644 --- a/sound/core/seq/seq_lock.c +++ b/sound/core/seq/seq_lock.c @@ -23,8 +23,6 @@ #include <sound/core.h> #include "seq_lock.h"
-#if defined(CONFIG_SMP) || defined(CONFIG_SND_DEBUG) - /* wait until all locks are released */ void snd_use_lock_sync_helper(snd_use_lock_t *lockp, const char *file, int line) { @@ -41,5 +39,3 @@ void snd_use_lock_sync_helper(snd_use_lock_t *lockp, const char *file, int line) } } EXPORT_SYMBOL(snd_use_lock_sync_helper); - -#endif diff --git a/sound/core/seq/seq_lock.h b/sound/core/seq/seq_lock.h index 54044bc2c9ef..ac38031c370e 100644 --- a/sound/core/seq/seq_lock.h +++ b/sound/core/seq/seq_lock.h @@ -3,8 +3,6 @@
#include <linux/sched.h>
-#if defined(CONFIG_SMP) || defined(CONFIG_SND_DEBUG) - typedef atomic_t snd_use_lock_t;
/* initialize lock */ @@ -20,14 +18,4 @@ typedef atomic_t snd_use_lock_t; void snd_use_lock_sync_helper(snd_use_lock_t *lock, const char *file, int line); #define snd_use_lock_sync(lockp) snd_use_lock_sync_helper(lockp, __BASE_FILE__, __LINE__)
-#else /* SMP || CONFIG_SND_DEBUG */ - -typedef spinlock_t snd_use_lock_t; /* dummy */ -#define snd_use_lock_init(lockp) /**/ -#define snd_use_lock_use(lockp) /**/ -#define snd_use_lock_free(lockp) /**/ -#define snd_use_lock_sync(lockp) /**/ - -#endif /* SMP || CONFIG_SND_DEBUG */ - #endif /* __SND_SEQ_LOCK_H */
On Wed, 18 Oct 2017 01:45:49 +0200, Ben Hutchings wrote:
The 'use' locking macros are no-ops if neither SMP or SND_DEBUG is enabled. This might once have been OK in non-preemptible configurations, but even in that case snd_seq_read() may sleep while relying on a 'use' lock. So always use the proper implementations.
Cc: stable@vger.kernel.org Signed-off-by: Ben Hutchings ben.hutchings@codethink.co.uk
Yeah, this conditional build makes little sense nowadays although it was meaningful in 20 years ago.
Applied now, thanks.
Takashi
participants (2)
-
Ben Hutchings
-
Takashi Iwai