Cleanup raidz expand pause variable usage

Signed-off-by: Don Brady <dev.fs.zfs@gmail.com>

cmd/ztest.c

@@ -275,7 +275,8 @@ extern uint_t dmu_object_alloc_chunk_shift;
 extern boolean_t zfs_force_some_double_word_sm_entries;
 extern unsigned long zio_decompress_fail_fraction;
 extern unsigned long zfs_reconstruct_indirect_damage_fraction;
-extern uint64_t raidz_expand_max_offset_pause;
+extern uint64_t raidz_expand_max_reflow_bytes;
+extern uint_t raidz_expand_pause_point;
 
 
 static ztest_shared_opts_t *ztest_shared_opts;
@@ -1226,21 +1227,21 @@ ztest_kill(ztest_shared_t *zs)
 	 * Before we kill ourselves, make sure that the config is updated.
 	 * See comment above spa_write_cachefile().
 	 */
-	if (raidz_expand_max_offset_pause) {
+	if (raidz_expand_pause_point != RAIDZ_EXPAND_PAUSE_NONE) {
 		if (mutex_tryenter(&spa_namespace_lock)) {
 			spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE,
 			    B_FALSE);
 			mutex_exit(&spa_namespace_lock);
 
 			ztest_scratch_state->zs_raidz_scratch_verify_pause =
-			    raidz_expand_max_offset_pause;
+			    raidz_expand_pause_point;
 		} else {
 			/*
 			 * Do not verify scratch object in case if
 			 * spa_namespace_lock cannot be acquired,
 			 * it can cause deadlock in spa_config_update().
 			 */
-			raidz_expand_max_offset_pause = 0;
+			raidz_expand_pause_point = RAIDZ_EXPAND_PAUSE_NONE;
 
 			return;
 		}
@@ -3949,7 +3950,7 @@ raidz_scratch_verify(void)
 	vdev_raidz_expand_t *vre;
 	vdev_t *raidvd;
 
-	ASSERT(raidz_expand_max_offset_pause == 0);
+	ASSERT(raidz_expand_pause_point == RAIDZ_EXPAND_PAUSE_NONE);
 
 	if (ztest_scratch_state->zs_raidz_scratch_verify_pause == 0)
 		return;
@@ -4031,7 +4032,7 @@ ztest_scratch_thread(void *arg)
 
 	/* wait up to 10 seconds */
 	for (int t = 100; t > 0; t -= 1) {
-		if (raidz_expand_max_offset_pause == 0)
+		if (raidz_expand_pause_point == RAIDZ_EXPAND_PAUSE_NONE)
 			thread_exit();
 
 		(void) poll(NULL, 0, 100);
@@ -4113,14 +4114,14 @@ ztest_vdev_raidz_attach(ztest_ds_t *zd, uint64_t id)
 	    0, 0, 1);
 
 	/*
-	 * 50% of the time, set raidz_expand_max_offset_pause to cause
+	 * 50% of the time, set raidz_expand_pause_point to cause
 	 * raidz_reflow_scratch_sync() and vdev_raidz_reflow_copy_scratch()
 	 * to pause at a certain point and then kill the test after 10
 	 * seconds so raidz_scratch_verify() can confirm consistency when
 	 * the pool is imported.
 	 */
 	if (ztest_random(2) == 0 && expected_error == 0) {
-		raidz_expand_max_offset_pause =
+		raidz_expand_pause_point =
 		    ztest_random(RAIDZ_EXPAND_PAUSE_SCRATCH_NOT_IN_USE) + 1;
 		scratch_thread = thread_create(NULL, 0, ztest_scratch_thread,
 		    ztest_shared, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri);
@@ -4140,13 +4141,13 @@ ztest_vdev_raidz_attach(ztest_ds_t *zd, uint64_t id)
 		    newpath, newsize, error, expected_error);
 	}
 
-	if (raidz_expand_max_offset_pause) {
+	if (raidz_expand_pause_point) {
 		if (error != 0) {
 			/*
 			 * Do not verify scratch object in case of error
 			 * returned by vdev attaching.
 			 */
-			raidz_expand_max_offset_pause = 0;
+			raidz_expand_pause_point = RAIDZ_EXPAND_PAUSE_NONE;
 		}
 
 		VERIFY0(thread_join(scratch_thread));
@@ -7635,8 +7636,10 @@ ztest_thread(void *arg)
 		/*
 		 * See if it's time to force a crash.
 		 */
-		if (now > zs->zs_thread_kill && !raidz_expand_max_offset_pause)
+		if (now > zs->zs_thread_kill &&
+		    raidz_expand_pause_point == RAIDZ_EXPAND_PAUSE_NONE) {
 			ztest_kill(zs);
+		}
 
 		/*
 		 * If we're getting ENOSPC with some regularity, stop.
@@ -8110,7 +8113,7 @@ ztest_raidz_expand_run(ztest_shared_t *zs, spa_t *spa)
 	/* Set our reflow target to 25%, 50% or 75% of allocated size */
 	uint_t multiple = ztest_random(3) + 1;
 	uint64_t reflow_max = (rzvd->vdev_stat.vs_alloc * multiple) / 4;
-	raidz_expand_max_offset_pause = reflow_max;
+	raidz_expand_max_reflow_bytes = reflow_max;
 
 	if (ztest_opts.zo_verbose >= 1) {
 		(void) printf("running raidz expansion test, killing when "
@@ -8194,7 +8197,7 @@ ztest_raidz_expand_run(ztest_shared_t *zs, spa_t *spa)
 	}
 
 	/* Reset the reflow pause before killing */
-	raidz_expand_max_offset_pause = 0;
+	raidz_expand_max_reflow_bytes = 0;
 
 	if (ztest_opts.zo_verbose >= 1) {
 		(void) printf("killing raidz expansion test after reflow "

include/sys/vdev_raidz.h

@@ -159,10 +159,18 @@ extern void spa_start_raidz_expansion_thread(spa_t *);
 extern int spa_raidz_expand_get_stats(spa_t *, pool_raidz_expand_stat_t *);
 extern int vdev_raidz_load(vdev_t *);
 
-/* Some of raidz scratch area states */
+/* RAIDZ scratch area pause points (for testing) */
 #define	RAIDZ_EXPAND_PAUSE_NONE	0
+#define	RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_1 1
+#define	RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_2 2
+#define	RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_3 3
 #define	RAIDZ_EXPAND_PAUSE_SCRATCH_VALID 4
 #define	RAIDZ_EXPAND_PAUSE_SCRATCH_REFLOWED 5
+#define	RAIDZ_EXPAND_PAUSE_SCRATCH_POST_REFLOW_1 6
+#define	RAIDZ_EXPAND_PAUSE_SCRATCH_POST_REFLOW_2 7
+#define	RAIDZ_EXPAND_PAUSE_SCRATCH_CRASH_COPY_1 8
+#define	RAIDZ_EXPAND_PAUSE_SCRATCH_CRASH_COPY_2 9
+#define	RAIDZ_EXPAND_PAUSE_SCRATCH_CRASH_COPY_3 10
 #define	RAIDZ_EXPAND_PAUSE_SCRATCH_NOT_IN_USE 11
 
 #ifdef	__cplusplus

man/man4/zfs.4

@@ -428,8 +428,8 @@ active.
 Max amount of memory to use for RAID-Z expansion I/O.
 This limits how much I/O can be outstanding at once.
 .
-.It Sy raidz_expand_max_offset_pause Ns = Ns Sy 0 Pq ulong
-For testing, pause RAID-Z expansion at this offset.
+.It Sy raidz_expand_max_reflow_bytes Ns = Ns Sy 0 Pq ulong
+For testing, pause RAID-Z expansion when reflow amount reaches this value.
 .
 .It Sy raidz_io_aggregate_rows Ns = Ns Sy 4 Pq ulong
 For expanded RAID-Z, aggregate reads that have more rows than this.

module/zfs/vdev_raidz.c

@@ -142,13 +142,18 @@
 }
 
 /*
- * For testing only: logical offset at which to pause the raidz expansion.
+ * For testing only: pause the raidz expansion after reflowing this amount.
  * (accessed by ZTS and ztest)
  */
 #ifdef	_KERNEL
 static
 #endif	/* _KERNEL */
-unsigned long raidz_expand_max_offset_pause = 0;
+unsigned long raidz_expand_max_reflow_bytes = 0;
+
+/*
+ * For testing only: pause the raidz expansion at a certain point.
+ */
+uint_t raidz_expand_pause_point = 0;
 
 /*
  * Maximum amount of copy io's outstanding at once.
@@ -3771,13 +3776,13 @@ raidz_reflow_impl(vdev_t *vd, vdev_raidz_expand_t *vre, range_tree_t *rt,
 }
 
 /*
- * For testing.
+ * For testing (ztest specific)
  */
 static void
-raidz_expand_pause(uint64_t progress)
+raidz_expand_pause(uint_t pause_point)
 {
-	while (raidz_expand_max_offset_pause != 0 &&
-	    raidz_expand_max_offset_pause <= progress)
+	while (raidz_expand_pause_point != 0 &&
+	    raidz_expand_pause_point <= pause_point)
 		delay(hz);
 }
 
@@ -3834,7 +3839,7 @@ raidz_reflow_scratch_sync(void *arg, dmu_tx_t *tx)
 		abds[i] = abd_alloc_linear(read_size, B_FALSE);
 	}
 
-	raidz_expand_pause(1);
+	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_1);
 
 	/*
 	 * If we have already written the scratch area then we must read from
@@ -3893,10 +3898,11 @@ raidz_reflow_scratch_sync(void *arg, dmu_tx_t *tx)
 		return;
 	}
 
+	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_2);
+
 	/*
 	 * Reflow in memory.
 	 */
-	raidz_expand_pause(2);
 	uint64_t logical_sectors = logical_size >> ashift;
 	for (int i = raidvd->vdev_children - 1; i < logical_sectors; i++) {
 		int oldchild = i % (raidvd->vdev_children - 1);
@@ -3946,7 +3952,8 @@ raidz_reflow_scratch_sync(void *arg, dmu_tx_t *tx)
 
 	zfs_dbgmsg("reflow: wrote %llu bytes (logical) to scratch area",
 	    (long long)logical_size);
-	raidz_expand_pause(3);
+
+	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_3);
 
 	/*
 	 * Update uberblock to indicate that scratch space is valid.  This is
@@ -4032,7 +4039,7 @@ raidz_reflow_scratch_sync(void *arg, dmu_tx_t *tx)
 	    (long long)logical_size,
 	    (long long)spa->spa_ubsync.ub_timestamp);
 
-	raidz_expand_pause(6);
+	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_POST_REFLOW_1);
 
 	/*
 	 * Update progress.
@@ -4050,7 +4057,7 @@ raidz_reflow_scratch_sync(void *arg, dmu_tx_t *tx)
 	 */
 	raidz_reflow_sync(spa, tx);
 
-	raidz_expand_pause(7);
+	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_POST_REFLOW_2);
 }
 
 /*
@@ -4080,7 +4087,7 @@ vdev_raidz_reflow_copy_scratch(spa_t *spa)
 		abds[i] = abd_alloc_linear(write_size, B_FALSE);
 	}
 
-	raidz_expand_pause(8);
+	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_CRASH_COPY_1);
 	pio = zio_root(spa, NULL, NULL, 0);
 	for (int i = 0; i < raidvd->vdev_children; i++) {
 		/*
@@ -4095,7 +4102,7 @@ vdev_raidz_reflow_copy_scratch(spa_t *spa)
 		    raidz_scratch_child_done, pio));
 	}
 	zio_wait(pio);
-	raidz_expand_pause(9);
+	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_CRASH_COPY_2);
 
 	/*
 	 * Overwrite real location with reflow'ed data.
@@ -4118,7 +4125,7 @@ vdev_raidz_reflow_copy_scratch(spa_t *spa)
 	for (int i = 0; i < raidvd->vdev_children; i++)
 		abd_free(abds[i]);
 	kmem_free(abds, raidvd->vdev_children * sizeof (abd_t *));
-	raidz_expand_pause(10);
+	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_CRASH_COPY_3);
 
 	/*
 	 * Update uberblock.
@@ -4272,14 +4279,12 @@ spa_raidz_expand_cb(void *arg, zthr_t *zthr)
 
 			/*
 			 * If requested, pause the reflow when the amount
-			 * specified by raidz_expand_max_offset_pause is reached
+			 * specified by raidz_expand_max_reflow_bytes is reached
 			 *
 			 * This pause is only used during testing or debugging.
-			 *
-			 * XXX Rename once we confirm that we want bytes
 			 */
-			while (raidz_expand_max_offset_pause != 0 &&
-			    raidz_expand_max_offset_pause <=
+			while (raidz_expand_max_reflow_bytes != 0 &&
+			    raidz_expand_max_reflow_bytes <=
 			    vre->vre_bytes_copied && !zthr_iscancelled(zthr)) {
 				delay(hz);
 			}
@@ -4775,8 +4780,8 @@ vdev_ops_t vdev_raidz_ops = {
 	.vdev_op_leaf = B_FALSE			/* not a leaf vdev */
 };
 
-ZFS_MODULE_PARAM(zfs_vdev, raidz_, expand_max_offset_pause, ULONG, ZMOD_RW,
-	"For testing, pause RAIDZ expansion at this offset");
+ZFS_MODULE_PARAM(zfs_vdev, raidz_, expand_max_reflow_bytes, ULONG, ZMOD_RW,
+	"For testing, pause RAIDZ expansion after reflowing this many bytes");
 ZFS_MODULE_PARAM(zfs_vdev, raidz_, expand_max_copy_bytes, ULONG, ZMOD_RW,
 	"Max amount of concurrent i/o for RAIDZ expansion");
 ZFS_MODULE_PARAM(zfs_vdev, raidz_, io_aggregate_rows, ULONG, ZMOD_RW,

tests/zfs-tests/include/tunables.cfg

@@ -63,7 +63,7 @@ MULTIHOST_IMPORT_INTERVALS	multihost.import_intervals	zfs_multihost_import_inter
 MULTIHOST_INTERVAL		multihost.interval		zfs_multihost_interval
 OVERRIDE_ESTIMATE_RECORDSIZE	send.override_estimate_recordsize	zfs_override_estimate_recordsize
 PREFETCH_DISABLE		prefetch.disable		zfs_prefetch_disable
-RAIDZ_EXPAND_MAX_OFFSET_PAUSE	vdev.expand_max_offset_pause	raidz_expand_max_offset_pause
+RAIDZ_EXPAND_MAX_REFLOW_BYTES	vdev.expand_max_reflow_bytes	raidz_expand_max_reflow_bytes
 REBUILD_SCRUB_ENABLED		rebuild_scrub_enabled		zfs_rebuild_scrub_enabled
 REMOVAL_SUSPEND_PROGRESS	removal_suspend_progress	zfs_removal_suspend_progress
 REMOVE_MAX_SEGMENT		remove_max_segment		zfs_remove_max_segment

tests/zfs-tests/tests/functional/raidz/raidz_expand_001_pos.ksh

@@ -62,7 +62,7 @@ function cleanup
 	done
 
 	log_must set_tunable32 PREFETCH_DISABLE $prefetch_disable
-	log_must set_tunable64 RAIDZ_EXPAND_MAX_OFFSET_PAUSE 0
+	log_must set_tunable64 RAIDZ_EXPAND_MAX_REFLOW_BYTES 0
 }
 
 function wait_expand_paused
@@ -144,7 +144,7 @@ function test_scrub # <pool> <parity> <dir>
 
 	randbyte=$(( ((RANDOM<<15) + RANDOM) % \
 		(dev_size_mb * (devs-1) * 1024 * 1024) ))
-	log_must set_tunable64 RAIDZ_EXPAND_MAX_OFFSET_PAUSE $randbyte
+	log_must set_tunable64 RAIDZ_EXPAND_MAX_REFLOW_BYTES $randbyte
 	log_must zpool attach $TESTPOOL ${raid}-0 $dir/dev-$devs
 	wait_expand_paused
 
@@ -175,7 +175,7 @@ function test_scrub # <pool> <parity> <dir>
 	log_must check_pool_status $pool "errors" "No known data errors"
 
 	log_must zpool clear $pool
-	log_must set_tunable64 RAIDZ_EXPAND_MAX_OFFSET_PAUSE 0
+	log_must set_tunable64 RAIDZ_EXPAND_MAX_REFLOW_BYTES 0
 	log_must zpool wait -t raidz_expand $TESTPOOL
 }
 

tests/zfs-tests/tests/functional/raidz/raidz_expand_003_pos.ksh

@@ -36,13 +36,13 @@
 #	2. For each parity value [1..3]
 #	    - create raidz pool with minimum block device files required
 #	    - create couple of datasets with different recordsize and fill it
-#	    - set raidz expand offset pause
+#	    - set raidz expand maximum reflow bytes
 #	    - start randwritecomp on one of the datasets files
 #	    - attach new device to the pool
-#	    - wait until reflow offset is equal to raidz expand pause offset
+#	    - wait for reflow bytes to reach the maximum
 #	    - kill randwritecomp
 #	    - verify pool
-#	    - set raidz expand offset to max value to complete raidz expansion
+#	    - set reflow bytes to max value to complete the expansion
 
 typeset -r devs=10
 typeset -r dev_size_mb=128
@@ -60,7 +60,7 @@ function cleanup
 	done
 
 	log_must set_tunable32 EMBEDDED_SLOG_MIN_MS $embedded_slog_min_ms
-	log_must set_tunable64 RAIDZ_EXPAND_MAX_OFFSET_PAUSE 0
+	log_must set_tunable64 RAIDZ_EXPAND_MAX_REFLOW_BYTES 0
 }
 
 function wait_expand_paused
@@ -105,7 +105,7 @@ for nparity in 1 2 3; do
 		pool_size=$(get_pool_prop size $pool)
 		# Pause at random location near the end of vdev
 		pause=$((((RANDOM << 15) + RANDOM) % pool_size))
-		log_must set_tunable64 RAIDZ_EXPAND_MAX_OFFSET_PAUSE $pause
+		log_must set_tunable64 RAIDZ_EXPAND_MAX_REFLOW_BYTES $pause
 
 		log_bkgrnd randwritecomp /$pool/fs/file
 		pid0=$!
@@ -125,9 +125,9 @@ for nparity in 1 2 3; do
 		log_must check_pool_status $pool "scan" "with 0 errors"
 		log_must check_pool_status $pool "scan" "repaired 0B"
 
-		# Set pause past largest possible offset for this pool
+		# Set pause past largest possible value for this pool
 		pause=$((devs*dev_size_mb*1024*1024))
-		log_must set_tunable64 RAIDZ_EXPAND_MAX_OFFSET_PAUSE $pause
+		log_must set_tunable64 RAIDZ_EXPAND_MAX_REFLOW_BYTES $pause
 
 		log_must zpool wait -t raidz_expand $pool
 	done