Align thread priority with Linux defaults

Under Linux filesystem threads responsible for handling I/O are
normally created with the maximum priority.  Non-I/O filesystem
processes run with the default priority.  ZFS should adopt the
same priority scheme under Linux to maintain good performance
and so that it will complete fairly when other Linux filesystems
are active.  The priorities have been updated to the following:

$ ps -eLo rtprio,cls,pid,pri,nice,cmd | egrep 'z_|spl_|zvol|arc|dbu|meta'
     -  TS 10743  19 -20 [spl_kmem_cache]
     -  TS 10744  19 -20 [spl_system_task]
     -  TS 10745  19 -20 [spl_dynamic_tas]
     -  TS 10764  19   0 [dbu_evict]
     -  TS 10765  19   0 [arc_prune]
     -  TS 10766  19   0 [arc_reclaim]
     -  TS 10767  19   0 [arc_user_evicts]
     -  TS 10768  19   0 [l2arc_feed]
     -  TS 10769  39   0 [z_unmount]
     -  TS 10770  39 -20 [zvol]
     -  TS 11011  39 -20 [z_null_iss]
     -  TS 11012  39 -20 [z_null_int]
     -  TS 11013  39 -20 [z_rd_iss]
     -  TS 11014  39 -20 [z_rd_int_0]
     -  TS 11022  38 -19 [z_wr_iss]
     -  TS 11023  39 -20 [z_wr_iss_h]
     -  TS 11024  39 -20 [z_wr_int_0]
     -  TS 11032  39 -20 [z_wr_int_h]
     -  TS 11033  39 -20 [z_fr_iss_0]
     -  TS 11041  39 -20 [z_fr_int]
     -  TS 11042  39 -20 [z_cl_iss]
     -  TS 11043  39 -20 [z_cl_int]
     -  TS 11044  39 -20 [z_ioctl_iss]
     -  TS 11045  39 -20 [z_ioctl_int]
     -  TS 11046  39 -20 [metaslab_group_]
     -  TS 11050  19   0 [z_iput]
     -  TS 11121  38 -19 [z_wr_iss]

Note that under Linux the meaning of a processes priority is inverted
with respect to illumos.  High values on Linux indicate a _low_ priority
while high value on illumos indicate a _high_ priority.

In order to preserve the logical meaning of the minclsyspri and
maxclsyspri macros when they are used by the illumos wrapper functions
their values have been inverted.  This way when changes are merged
from upstream illumos we won't need to remember to invert the macro.
It could also lead to confusion.

This patch depends on openzfs/spl#466.

Signed-off-by: Brian Behlendorf <[email protected]>
Signed-off-by: Ned Bass <[email protected]>
Closes #3607

Author: behlendorf

include/sys/zfs_context.h

@@ -233,6 +233,7 @@ typedef struct kthread {
 	kt_did_t	t_tid;
 	thread_func_t	t_func;
 	void *		t_arg;
+	pri_t		t_pri;
 } kthread_t;
 
 #define	curthread			zk_thread_current()
@@ -615,8 +616,12 @@ extern void delay(clock_t ticks);
 #define	max_ncpus	64
 #define	boot_ncpus	(sysconf(_SC_NPROCESSORS_ONLN))
 
-#define	minclsyspri	60
-#define	maxclsyspri	99
+/*
+ * Process priorities as defined by setpriority(2) and getpriority(2).
+ */
+#define	minclsyspri	19
+#define	maxclsyspri	-20
+#define	defclsyspri	0
 
 #define	CPU_SEQID	(pthread_self() & (max_ncpus - 1))
 

lib/libzpool/kernel.c

@@ -128,6 +128,7 @@ zk_thread_helper(void *arg)
 	VERIFY3S(pthread_mutex_lock(&kthread_lock), ==, 0);
 	kthread_nr++;
 	VERIFY3S(pthread_mutex_unlock(&kthread_lock), ==, 0);
+	(void) setpriority(PRIO_PROCESS, 0, kt->t_pri);
 
 	kt->t_tid = pthread_self();
 	((thread_func_arg_t) kt->t_func)(kt->t_arg);
@@ -151,6 +152,7 @@ zk_thread_create(caddr_t stk, size_t stksize, thread_func_t func, void *arg,
 	kt = umem_zalloc(sizeof (kthread_t), UMEM_NOFAIL);
 	kt->t_func = func;
 	kt->t_arg = arg;
+	kt->t_pri = pri;
 
 	VERIFY0(pthread_attr_init(&attr));
 	VERIFY0(pthread_attr_setdetachstate(&attr, detachstate));

lib/libzpool/taskq.c

@@ -308,7 +308,7 @@ taskq_create(const char *name, int nthreads, pri_t pri,
 
 	for (t = 0; t < nthreads; t++)
 		VERIFY((tq->tq_threadlist[t] = thread_create(NULL, 0,
-		    taskq_thread, tq, TS_RUN, NULL, 0, 0)) != NULL);
+		    taskq_thread, tq, TS_RUN, NULL, 0, pri)) != NULL);
 
 	return (tq);
 }
@@ -371,7 +371,7 @@ taskq_cancel_id(taskq_t *tq, taskqid_t id)
 void
 system_taskq_init(void)
 {
-	system_taskq = taskq_create("system_taskq", 64, minclsyspri, 4, 512,
+	system_taskq = taskq_create("system_taskq", 64, maxclsyspri, 4, 512,
 	    TASKQ_DYNAMIC | TASKQ_PREPOPULATE);
 }
 

module/zfs/arc.c

@@ -5431,7 +5431,7 @@ arc_init(void)
 	mutex_init(&arc_prune_mtx, NULL, MUTEX_DEFAULT, NULL);
 	bzero(&arc_eviction_hdr, sizeof (arc_buf_hdr_t));
 
-	arc_prune_taskq = taskq_create("arc_prune", max_ncpus, minclsyspri,
+	arc_prune_taskq = taskq_create("arc_prune", max_ncpus, defclsyspri,
 	    max_ncpus, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
 
 	arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED,
@@ -5444,10 +5444,10 @@ arc_init(void)
 	}
 
 	(void) thread_create(NULL, 0, arc_reclaim_thread, NULL, 0, &p0,
-	    TS_RUN, minclsyspri);
+	    TS_RUN, defclsyspri);
 
 	(void) thread_create(NULL, 0, arc_user_evicts_thread, NULL, 0, &p0,
-	    TS_RUN, minclsyspri);
+	    TS_RUN, defclsyspri);
 
 	arc_dead = FALSE;
 	arc_warm = B_FALSE;
@@ -6954,7 +6954,7 @@ l2arc_start(void)
 		return;
 
 	(void) thread_create(NULL, 0, l2arc_feed_thread, NULL, 0, &p0,
-	    TS_RUN, minclsyspri);
+	    TS_RUN, defclsyspri);
 }
 
 void

module/zfs/dbuf.c

@@ -413,7 +413,7 @@ dbuf_init(void)
 	 * All entries are queued via taskq_dispatch_ent(), so min/maxalloc
 	 * configuration is not required.
 	 */
-	dbu_evict_taskq = taskq_create("dbu_evict", 1, minclsyspri, 0, 0, 0);
+	dbu_evict_taskq = taskq_create("dbu_evict", 1, defclsyspri, 0, 0, 0);
 }
 
 void

module/zfs/dmu_objset.c

@@ -1839,7 +1839,7 @@ dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj,
 	ntasks = dmu_find_threads;
 	if (ntasks == 0)
 		ntasks = vdev_count_leaves(dp->dp_spa) * 4;
-	tq = taskq_create("dmu_objset_find", ntasks, minclsyspri, ntasks,
+	tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks,
 	    INT_MAX, 0);
 	if (tq == NULL) {
 		kmem_free(dcp, sizeof (*dcp));

module/zfs/dsl_pool.c

@@ -170,7 +170,7 @@ dsl_pool_open_impl(spa_t *spa, uint64_t txg)
 	mutex_init(&dp->dp_lock, NULL, MUTEX_DEFAULT, NULL);
 	cv_init(&dp->dp_spaceavail_cv, NULL, CV_DEFAULT, NULL);
 
-	dp->dp_iput_taskq = taskq_create("z_iput", max_ncpus, minclsyspri,
+	dp->dp_iput_taskq = taskq_create("z_iput", max_ncpus, defclsyspri,
 	    max_ncpus * 8, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
 
 	return (dp);

module/zfs/metaslab.c

@@ -492,7 +492,7 @@ metaslab_group_create(metaslab_class_t *mc, vdev_t *vd)
 	mg->mg_activation_count = 0;
 
 	mg->mg_taskq = taskq_create("metaslab_group_taskq", metaslab_load_pct,
-	    minclsyspri, 10, INT_MAX, TASKQ_THREADS_CPU_PCT | TASKQ_DYNAMIC);
+	    maxclsyspri, 10, INT_MAX, TASKQ_THREADS_CPU_PCT | TASKQ_DYNAMIC);
 
 	return (mg);
 }

module/zfs/spa.c

@@ -898,11 +898,13 @@ spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
 			pri_t pri = maxclsyspri;
 			/*
 			 * The write issue taskq can be extremely CPU
-			 * intensive.  Run it at slightly lower priority
-			 * than the other taskqs.
+			 * intensive.  Run it at slightly less important
+			 * priority than the other taskqs.  Under Linux this
+			 * means incrementing the priority value on platforms
+			 * like illumos it should be decremented.
 			 */
 			if (t == ZIO_TYPE_WRITE && q == ZIO_TASKQ_ISSUE)
-				pri--;
+				pri++;
 
 			tq = taskq_create_proc(name, value, pri, 50,
 			    INT_MAX, spa->spa_proc, flags);

module/zfs/txg.c

@@ -205,15 +205,15 @@ txg_sync_start(dsl_pool_t *dp)
 	tx->tx_threads = 2;
 
 	tx->tx_quiesce_thread = thread_create(NULL, 0, txg_quiesce_thread,
-	    dp, 0, &p0, TS_RUN, minclsyspri);
+	    dp, 0, &p0, TS_RUN, defclsyspri);
 
 	/*
 	 * The sync thread can need a larger-than-default stack size on
 	 * 32-bit x86.  This is due in part to nested pools and
 	 * scrub_visitbp() recursion.
 	 */
 	tx->tx_sync_thread = thread_create(NULL, 32<<10, txg_sync_thread,
-	    dp, 0, &p0, TS_RUN, minclsyspri);
+	    dp, 0, &p0, TS_RUN, defclsyspri);
 
 	mutex_exit(&tx->tx_sync_lock);
 }
@@ -445,7 +445,7 @@ txg_dispatch_callbacks(dsl_pool_t *dp, uint64_t txg)
 			 * Commit callback taskq hasn't been created yet.
 			 */
 			tx->tx_commit_cb_taskq = taskq_create("tx_commit_cb",
-			    max_ncpus, minclsyspri, max_ncpus, max_ncpus * 2,
+			    max_ncpus, defclsyspri, max_ncpus, max_ncpus * 2,
 			    TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
 		}
 

module/zfs/zfs_ctldir.c

@@ -1009,7 +1009,7 @@ zfsctl_shares_lookup(struct inode *dip, char *name, struct inode **ipp,
 void
 zfsctl_init(void)
 {
-	zfs_expire_taskq = taskq_create("z_unmount", 1, maxclsyspri,
+	zfs_expire_taskq = taskq_create("z_unmount", 1, defclsyspri,
 	    1, 8, TASKQ_PREPOPULATE);
 }
 

module/zfs/zil.c

@@ -1888,7 +1888,7 @@ zil_open(objset_t *os, zil_get_data_t *get_data)
 	ASSERT(list_is_empty(&zilog->zl_lwb_list));
 
 	zilog->zl_get_data = get_data;
-	zilog->zl_clean_taskq = taskq_create("zil_clean", 1, minclsyspri,
+	zilog->zl_clean_taskq = taskq_create("zil_clean", 1, defclsyspri,
 	    2, 2, TASKQ_PREPOPULATE);
 
 	return (zilog);