Illumos #3537

3537 want pool io kstats

Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Eric Schrock <eric.schrock@delphix.com>
Reviewed by: Sa?o Kiselkov <skiselkov.ml@gmail.com>
Reviewed by: Garrett D'Amore <garrett@damore.org>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Gordon Ross <gwr@nexenta.com>

References:
  http://www.illumos.org/issues/3537
  illumos/illumos-gate@c3a6601

Ported by: Cyril Plisko <cyril.plisko@mountall.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

Porting Notes:

1. The patch was restructured to take advantage of the existing
   spa statistics infrastructure.  To accomplish this the kstat
   was moved in to spa->io_stats and the init/destroy code moved
   to spa_stats.c.

2. The I/O kstat was simply named <pool> which conflicted with the
   pool directory we had already created.  Therefore it was renamed
   to <pool>/io

3. An update handler was added to allow the kstat to be zeroed.

include/sys/spa.h

@@ -550,6 +550,7 @@ typedef struct spa_stats {
 	spa_stats_history_t	read_history;
 	spa_stats_history_t	txg_history;
 	spa_stats_history_t	tx_assign_histogram;
+	spa_stats_history_t	io_history;
 } spa_stats_t;
 
 typedef enum txg_state {

include/sys/zfs_context.h

@@ -344,10 +344,16 @@ extern void cv_broadcast(kcondvar_t *cv);
 /*
  * kstat creation, installation and deletion
  */
-extern kstat_t *kstat_create(char *, int,
+extern kstat_t *kstat_create(const char *, int,
-    char *, char *, uchar_t, ulong_t, uchar_t);
+    const char *, const char *, uchar_t, ulong_t, uchar_t);
 extern void kstat_install(kstat_t *);
 extern void kstat_delete(kstat_t *);
+extern void kstat_waitq_enter(kstat_io_t *);
+extern void kstat_waitq_exit(kstat_io_t *);
+extern void kstat_runq_enter(kstat_io_t *);
+extern void kstat_runq_exit(kstat_io_t *);
+extern void kstat_waitq_to_runq(kstat_io_t *);
+extern void kstat_runq_back_to_waitq(kstat_io_t *);
 extern void kstat_set_raw_ops(kstat_t *ksp,
     int (*headers)(char *buf, size_t size),
     int (*data)(char *buf, size_t size, void *data),

lib/libzpool/kernel.c

@@ -224,8 +224,8 @@ zk_thread_join(kt_did_t tid)
  */
 /*ARGSUSED*/
 kstat_t *
-kstat_create(char *module, int instance, char *name, char *class,
+kstat_create(const char *module, int instance, const char *name,
-    uchar_t type, ulong_t ndata, uchar_t ks_flag)
+    const char *class, uchar_t type, ulong_t ndata, uchar_t ks_flag)
 {
 	return (NULL);
 }
@@ -241,6 +241,35 @@ kstat_delete(kstat_t *ksp)
 {}
 
 /*ARGSUSED*/
+void
+kstat_waitq_enter(kstat_io_t *kiop)
+{}
+
+/*ARGSUSED*/
+void
+kstat_waitq_exit(kstat_io_t *kiop)
+{}
+
+/*ARGSUSED*/
+void
+kstat_runq_enter(kstat_io_t *kiop)
+{}
+
+/*ARGSUSED*/
+void
+kstat_runq_exit(kstat_io_t *kiop)
+{}
+
+/*ARGSUSED*/
+void
+kstat_waitq_to_runq(kstat_io_t *kiop)
+{}
+
+/*ARGSUSED*/
+void
+kstat_runq_back_to_waitq(kstat_io_t *kiop)
+{}
+
 void
 kstat_set_raw_ops(kstat_t *ksp,
     int (*headers)(char *buf, size_t size),

module/zfs/spa_stats.c

@@ -608,12 +608,61 @@ spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs)
 	atomic_inc_64(&((kstat_named_t *)ssh->private)[idx].value.ui64);
 }
 
+/*
+ * ==========================================================================
+ * SPA IO History Routines
+ * ==========================================================================
+ */
+static int
+spa_io_history_update(kstat_t *ksp, int rw)
+{
+	if (rw == KSTAT_WRITE)
+		memset(ksp->ks_data, 0, ksp->ks_data_size);
+
+	return (0);
+}
+
+static void
+spa_io_history_init(spa_t *spa)
+{
+	spa_stats_history_t *ssh = &spa->spa_stats.io_history;
+	char name[KSTAT_STRLEN];
+	kstat_t *ksp;
+
+	mutex_init(&ssh->lock, NULL, MUTEX_DEFAULT, NULL);
+
+	(void) snprintf(name, KSTAT_STRLEN, "zfs/%s", spa_name(spa));
+	name[KSTAT_STRLEN-1] = '\0';
+
+	ksp = kstat_create(name, 0, "io", "disk", KSTAT_TYPE_IO, 1, 0);
+	ssh->kstat = ksp;
+
+	if (ksp) {
+		ksp->ks_lock = &ssh->lock;
+		ksp->ks_private = spa;
+		ksp->ks_update = spa_io_history_update;
+		kstat_install(ksp);
+	}
+}
+
+static void
+spa_io_history_destroy(spa_t *spa)
+{
+	spa_stats_history_t *ssh = &spa->spa_stats.io_history;
+
+	if (ssh->kstat)
+		kstat_delete(ssh->kstat);
+
+	mutex_destroy(&ssh->lock);
+}
+
 void
 spa_stats_init(spa_t *spa)
 {
 	spa_read_history_init(spa);
 	spa_txg_history_init(spa);
 	spa_tx_assign_init(spa);
+	spa_io_history_init(spa);
 }
 
 void
@@ -622,6 +671,7 @@ spa_stats_destroy(spa_t *spa)
 	spa_tx_assign_destroy(spa);
 	spa_txg_history_destroy(spa);
 	spa_read_history_destroy(spa);
+	spa_io_history_destroy(spa);
 }
 
 #if defined(_KERNEL) && defined(HAVE_SPL)

module/zfs/vdev_queue.c

@@ -29,8 +29,10 @@
 
 #include <sys/zfs_context.h>
 #include <sys/vdev_impl.h>
+#include <sys/spa_impl.h>
 #include <sys/zio.h>
 #include <sys/avl.h>
+#include <sys/kstat.h>
 
 /*
  * These tunables are for performance analysis.
@@ -164,15 +166,72 @@ vdev_queue_fini(vdev_t *vd)
 static void
 vdev_queue_io_add(vdev_queue_t *vq, zio_t *zio)
 {
+	spa_t *spa = zio->io_spa;
+	spa_stats_history_t *ssh = &spa->spa_stats.io_history;
+
 	avl_add(&vq->vq_deadline_tree, zio);
 	avl_add(zio->io_vdev_tree, zio);
+
+	if (ssh->kstat != NULL) {
+		mutex_enter(&ssh->lock);
+		kstat_waitq_enter(ssh->kstat->ks_data);
+		mutex_exit(&ssh->lock);
+	}
 }
 
 static void
 vdev_queue_io_remove(vdev_queue_t *vq, zio_t *zio)
 {
+	spa_t *spa = zio->io_spa;
+	spa_stats_history_t *ssh = &spa->spa_stats.io_history;
+
 	avl_remove(&vq->vq_deadline_tree, zio);
 	avl_remove(zio->io_vdev_tree, zio);
+
+	if (ssh->kstat != NULL) {
+		mutex_enter(&ssh->lock);
+		kstat_waitq_exit(ssh->kstat->ks_data);
+		mutex_exit(&ssh->lock);
+	}
+}
+
+static void
+vdev_queue_pending_add(vdev_queue_t *vq, zio_t *zio)
+{
+	spa_t *spa = zio->io_spa;
+	spa_stats_history_t *ssh = &spa->spa_stats.io_history;
+
+	avl_add(&vq->vq_pending_tree, zio);
+
+	if (ssh->kstat != NULL) {
+		mutex_enter(&ssh->lock);
+		kstat_runq_enter(ssh->kstat->ks_data);
+		mutex_exit(&ssh->lock);
+	}
+}
+
+static void
+vdev_queue_pending_remove(vdev_queue_t *vq, zio_t *zio)
+{
+	spa_t *spa = zio->io_spa;
+	spa_stats_history_t *ssh = &spa->spa_stats.io_history;
+
+	avl_remove(&vq->vq_pending_tree, zio);
+
+	if (ssh->kstat != NULL) {
+		kstat_io_t *ksio = ssh->kstat->ks_data;
+
+		mutex_enter(&ssh->lock);
+		kstat_runq_exit(ksio);
+		if (zio->io_type == ZIO_TYPE_READ) {
+			ksio->reads++;
+			ksio->nread += zio->io_size;
+		} else if (zio->io_type == ZIO_TYPE_WRITE) {
+			ksio->writes++;
+			ksio->nwritten += zio->io_size;
+		}
+		mutex_exit(&ssh->lock);
+	}
 }
 
 static void
@@ -351,7 +410,7 @@ again:
 			zio_execute(dio);
 		} while (dio != lio);
 
-		avl_add(&vq->vq_pending_tree, aio);
+		vdev_queue_pending_add(vq, aio);
 		list_remove(&vq->vq_io_list, vi);
 
 		return (aio);
@@ -374,7 +433,7 @@ again:
 		goto again;
 	}
 
-	avl_add(&vq->vq_pending_tree, fio);
+	vdev_queue_pending_add(vq, fio);
 
 	return (fio);
 }
@@ -431,7 +490,7 @@ vdev_queue_io_done(zio_t *zio)
 
 	mutex_enter(&vq->vq_lock);
 
-	avl_remove(&vq->vq_pending_tree, zio);
+	vdev_queue_pending_remove(vq, zio);
 
 	zio->io_delta = gethrtime() - zio->io_timestamp;
 	vq->vq_io_complete_ts = gethrtime();