ALSA: seq: More protection for concurrent write and ioctl races

This patch is an attempt for further hardening against races between
the concurrent write and ioctls.  The previous fix d15d662e89
("ALSA: seq: Fix racy pool initializations") covered the race of the
pool initialization at writer and the pool resize ioctl by the
client->ioctl_mutex (CVE-2018-1000004).  However, basically this mutex
should be applied more widely to the whole write operation for
avoiding the unexpected pool operations by another thread.

The only change outside snd_seq_write() is the additional mutex
argument to helper functions, so that we can unlock / relock the given
mutex temporarily during schedule() call for blocking write.

Fixes: d15d662e89 ("ALSA: seq: Fix racy pool initializations")
Reported-by: 范龙飞 <long7573@126.com>
Reported-by: Nicolai Stange <nstange@suse.de>
Reviewed-and-tested-by: Nicolai Stange <nstange@suse.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>

sound/core/seq/seq_clientmgr.c

@@ -910,7 +910,8 @@ int snd_seq_dispatch_event(struct snd_seq_event_cell *cell, int atomic, int hop)
 static int snd_seq_client_enqueue_event(struct snd_seq_client *client,
 					struct snd_seq_event *event,
 					struct file *file, int blocking,
-					int atomic, int hop)
+					int atomic, int hop,
+					struct mutex *mutexp)
 {
 	struct snd_seq_event_cell *cell;
 	int err;
@@ -948,7 +949,8 @@ static int snd_seq_client_enqueue_event(struct snd_seq_client *client,
 		return -ENXIO; /* queue is not allocated */
 
 	/* allocate an event cell */
-	err = snd_seq_event_dup(client->pool, event, &cell, !blocking || atomic, file);
+	err = snd_seq_event_dup(client->pool, event, &cell, !blocking || atomic,
+				file, mutexp);
 	if (err < 0)
 		return err;
 
@@ -1017,12 +1019,11 @@ static ssize_t snd_seq_write(struct file *file, const char __user *buf,
 		return -ENXIO;
 
 	/* allocate the pool now if the pool is not allocated yet */ 
-	if (client->pool->size > 0 && !snd_seq_write_pool_allocated(client)) {
 	mutex_lock(&client->ioctl_mutex);
+	if (client->pool->size > 0 && !snd_seq_write_pool_allocated(client)) {
 		err = snd_seq_pool_init(client->pool);
-		mutex_unlock(&client->ioctl_mutex);
 		if (err < 0)
-			return -ENOMEM;
+			goto out;
 	}
 
 	/* only process whole events */
@@ -1073,7 +1074,7 @@ static ssize_t snd_seq_write(struct file *file, const char __user *buf,
 		/* ok, enqueue it */
 		err = snd_seq_client_enqueue_event(client, &event, file,
 						   !(file->f_flags & O_NONBLOCK),
-						   0, 0);
+						   0, 0, &client->ioctl_mutex);
 		if (err < 0)
 			break;
 
@@ -1084,6 +1085,8 @@ static ssize_t snd_seq_write(struct file *file, const char __user *buf,
 		written += len;
 	}
 
+ out:
+	mutex_unlock(&client->ioctl_mutex);
 	return written ? written : err;
 }
 
@@ -2263,7 +2266,8 @@ static int kernel_client_enqueue(int client, struct snd_seq_event *ev,
 	if (! cptr->accept_output)
 		result = -EPERM;
 	else /* send it */
-		result = snd_seq_client_enqueue_event(cptr, ev, file, blocking, atomic, hop);
+		result = snd_seq_client_enqueue_event(cptr, ev, file, blocking,
+						      atomic, hop, NULL);
 
 	snd_seq_client_unlock(cptr);
 	return result;

sound/core/seq/seq_fifo.c

@@ -125,7 +125,7 @@ int snd_seq_fifo_event_in(struct snd_seq_fifo *f,
 		return -EINVAL;
 
 	snd_use_lock_use(&f->use_lock);
-	err = snd_seq_event_dup(f->pool, event, &cell, 1, NULL); /* always non-blocking */
+	err = snd_seq_event_dup(f->pool, event, &cell, 1, NULL, NULL); /* always non-blocking */
 	if (err < 0) {
 		if ((err == -ENOMEM) || (err == -EAGAIN))
 			atomic_inc(&f->overflow);

sound/core/seq/seq_memory.c

@@ -220,7 +220,8 @@ void snd_seq_cell_free(struct snd_seq_event_cell * cell)
  */
 static int snd_seq_cell_alloc(struct snd_seq_pool *pool,
 			      struct snd_seq_event_cell **cellp,
-			      int nonblock, struct file *file)
+			      int nonblock, struct file *file,
+			      struct mutex *mutexp)
 {
 	struct snd_seq_event_cell *cell;
 	unsigned long flags;
@@ -244,7 +245,11 @@ static int snd_seq_cell_alloc(struct snd_seq_pool *pool,
 		set_current_state(TASK_INTERRUPTIBLE);
 		add_wait_queue(&pool->output_sleep, &wait);
 		spin_unlock_irq(&pool->lock);
+		if (mutexp)
+			mutex_unlock(mutexp);
 		schedule();
+		if (mutexp)
+			mutex_lock(mutexp);
 		spin_lock_irq(&pool->lock);
 		remove_wait_queue(&pool->output_sleep, &wait);
 		/* interrupted? */
@@ -287,7 +292,7 @@ __error:
  */
 int snd_seq_event_dup(struct snd_seq_pool *pool, struct snd_seq_event *event,
 		      struct snd_seq_event_cell **cellp, int nonblock,
-		      struct file *file)
+		      struct file *file, struct mutex *mutexp)
 {
 	int ncells, err;
 	unsigned int extlen;
@@ -304,7 +309,7 @@ int snd_seq_event_dup(struct snd_seq_pool *pool, struct snd_seq_event *event,
 	if (ncells >= pool->total_elements)
 		return -ENOMEM;
 
-	err = snd_seq_cell_alloc(pool, &cell, nonblock, file);
+	err = snd_seq_cell_alloc(pool, &cell, nonblock, file, mutexp);
 	if (err < 0)
 		return err;
 
@@ -330,7 +335,8 @@ int snd_seq_event_dup(struct snd_seq_pool *pool, struct snd_seq_event *event,
 			int size = sizeof(struct snd_seq_event);
 			if (len < size)
 				size = len;
-			err = snd_seq_cell_alloc(pool, &tmp, nonblock, file);
+			err = snd_seq_cell_alloc(pool, &tmp, nonblock, file,
+						 mutexp);
 			if (err < 0)
 				goto __error;
 			if (cell->event.data.ext.ptr == NULL)

sound/core/seq/seq_memory.h

@@ -66,7 +66,8 @@ struct snd_seq_pool {
 void snd_seq_cell_free(struct snd_seq_event_cell *cell);
 
 int snd_seq_event_dup(struct snd_seq_pool *pool, struct snd_seq_event *event,
-		      struct snd_seq_event_cell **cellp, int nonblock, struct file *file);
+		      struct snd_seq_event_cell **cellp, int nonblock,
+		      struct file *file, struct mutex *mutexp);
 
 /* return number of unused (free) cells */
 static inline int snd_seq_unused_cells(struct snd_seq_pool *pool)