Illumos #3581 spa_zio_taskq[ZIO_TYPE_FREE][ZIO_TASKQ_ISSUE]->tq_lock contention

3581 spa_zio_taskq[ZIO_TYPE_FREE][ZIO_TASKQ_ISSUE]->tq_lock is piping hot

Reviewed by: Matthew Ahrens <[email protected]>
Reviewed by: George Wilson <[email protected]>
Reviewed by: Christopher Siden <[email protected]>
Reviewed by: Gordon Ross <[email protected]>
Approved by: Richard Lowe <[email protected]>

References:
  illumos/illumos-gate@ec94d32
  https://illumos.org/issues/3581

Notes for Linux port:

Earlier commit 08d08eb reduced contention on this taskq lock by simply
reducing the number of z_fr_iss threads from 100 to one-per-CPU.  We
also optimized the taskq implementation in zfsonlinux/spl@3c6ed54.
These changes significantly improved unlink performance to acceptable
levels.

This patch further reduces time spent spinning on this lock by
randomly dispatching the work items over multiple independent task
queues.  The Illumos ZFS developers stated that this lock contention
only arose after "3329 spa_sync() spends 10-20% of its time in
spa_free_sync_cb()" was landed.  It's not clear if 3329 affects the
Linux port or not.  I didn't see spa_free_sync_cb() show up in
oprofile sessions while unlinking large files, but I may just not
have used the right test case.

I tested unlinking a 1 TB of data with and without the patch and
didn't observe a meaningful difference in elapsed time.  However,
oprofile showed that the percent time spent in taskq_thread() was
reduced from about 16% to about 5%.  Aside from a possible slight
performance benefit this may be worth landing if only for the sake of
maintaining consistency with upstream.

Ported-by: Ned Bass <[email protected]>
Closes #1327

1 files changed, 12 insertions, 4 deletions

diff --git a/include/sys/spa_impl.h b/include/sys/spa_impl.h
index 5bd0e0a9c..1b12b4e3a 100644
--- a/include/sys/spa_impl.h
+++ b/include/sys/spa_impl.h
@@ -80,16 +80,16 @@ typedef struct spa_config_dirent {
 	char		*scd_path;
 } spa_config_dirent_t;
 
-enum zio_taskq_type {
+typedef enum zio_taskq_type {
 	ZIO_TASKQ_ISSUE = 0,
 	ZIO_TASKQ_ISSUE_HIGH,
 	ZIO_TASKQ_INTERRUPT,
 	ZIO_TASKQ_INTERRUPT_HIGH,
 	ZIO_TASKQ_TYPES
-};
+} zio_taskq_type_t;
 
 /*
- * State machine for the zpool-pooname process.  The states transitions
+ * State machine for the zpool-poolname process.  The states transitions
  * are done as follows:
  *
  *	From		   To			Routine
@@ -107,6 +107,11 @@ typedef enum spa_proc_state {
 	SPA_PROC_GONE		/* spa_thread() is exiting, spa_proc = &p0 */
 } spa_proc_state_t;
 
+typedef struct spa_taskqs {
+	uint_t stqs_count;
+	taskq_t **stqs_taskq;
+} spa_taskqs_t;
+
 struct spa {
 	/*
 	 * Fields protected by spa_namespace_lock.
@@ -125,7 +130,7 @@ struct spa {
 	uint8_t		spa_sync_on;		/* sync threads are running */
 	spa_load_state_t spa_load_state;	/* current load operation */
 	uint64_t	spa_import_flags;	/* import specific flags */
-	taskq_t		*spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
+	spa_taskqs_t	spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
 	dsl_pool_t	*spa_dsl_pool;
 	boolean_t	spa_is_initializing;	/* true while opening pool */
 	metaslab_class_t *spa_normal_class;	/* normal data class */
@@ -243,6 +248,9 @@ struct spa {
 
 extern char *spa_config_path;
 
+extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
+    task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent);
+
 #ifdef	__cplusplus
 }
 #endif