Fix evict() deadlock

Now that KM_SLEEP is not defined as GFP_NOFS there is the possibility
of synchronous reclaim deadlocks.  These deadlocks never existed in the
original OpenSolaris code because all memory reclaim on Solaris is done
asyncronously.  Linux does both synchronous (direct) and asynchronous
(indirect) reclaim.

This commit addresses a deadlock caused by inode eviction.  A KM_SLEEP
allocation may trigger direct memory reclaim and shrink the inode cache.
This can occur while a mutex in the array of ZFS_OBJ_HOLD mutexes is
held.  Through the ->shrink_icache_memory()->evict()->zfs_inactive()->
zfs_zinactive() call path the same mutex may be reacquired resulting
in a deadlock.  To avoid this deadlock the process must not reacquire
the mutex when it is already holding it.

This is a reasonable fix for now but longer term the ZFS_OBJ_HOLD
mutex locking should be reevaluated.  This infrastructure already
prevents us from ever using the Linux lock dependency analysis tools,
and it may limit scalability.

include/sys/zfs_znode.h

@@ -280,6 +280,8 @@ typedef struct znode {
 	mutex_tryenter(ZFS_OBJ_MUTEX((zsb), (obj_num)))
 #define	ZFS_OBJ_HOLD_EXIT(zsb, obj_num) \
 	mutex_exit(ZFS_OBJ_MUTEX((zsb), (obj_num)))
+#define	ZFS_OBJ_HOLD_OWNED(zsb, obj_num) \
+	mutex_owned(ZFS_OBJ_MUTEX((zsb), (obj_num)))
 
 /*
  * Macros to encode/decode ZFS stored time values from/to struct timespec

module/zfs/zfs_znode.c

@@ -980,13 +980,24 @@ zfs_zinactive(znode_t *zp)
 {
 	zfs_sb_t *zsb = ZTOZSB(zp);
 	uint64_t z_id = zp->z_id;
+	boolean_t drop_mutex = 0;
 
 	ASSERT(zp->z_sa_hdl);
 
 	/*
-	 * Don't allow a zfs_zget() while were trying to release this znode
+	 * Don't allow a zfs_zget() while were trying to release this znode.
+	 *
+	 * Linux allows direct memory reclaim which means that any KM_SLEEP
+	 * allocation may trigger inode eviction.  This can lead to a deadlock
+	 * through the ->shrink_icache_memory()->evict()->zfs_inactive()->
+	 * zfs_zinactive() call path.  To avoid this deadlock the process
+	 * must not reacquire the mutex when it is already holding it.
 	 */
-	ZFS_OBJ_HOLD_ENTER(zsb, z_id);
+	if (!ZFS_OBJ_HOLD_OWNED(zsb, z_id)) {
+		ZFS_OBJ_HOLD_ENTER(zsb, z_id);
+		drop_mutex = 1;
+	}
+
 	mutex_enter(&zp->z_lock);
 
 	/*
@@ -995,14 +1006,19 @@ zfs_zinactive(znode_t *zp)
 	 */
 	if (zp->z_unlinked) {
 		mutex_exit(&zp->z_lock);
-		ZFS_OBJ_HOLD_EXIT(zsb, z_id);
+
+		if (drop_mutex)
+			ZFS_OBJ_HOLD_EXIT(zsb, z_id);
+
 		zfs_rmnode(zp);
 		return;
 	}
 
 	mutex_exit(&zp->z_lock);
 	zfs_znode_dmu_fini(zp);
-	ZFS_OBJ_HOLD_EXIT(zsb, z_id);
+
+	if (drop_mutex)
+		ZFS_OBJ_HOLD_EXIT(zsb, z_id);
 }
 
 void