Pool allocation classes

Allocation Classes add the ability to have allocation classes in a
pool that are dedicated to serving specific block categories, such
as DDT data, metadata, and small file blocks. A pool can opt-in to
this feature by adding a 'special' or 'dedup' top-level VDEV.

Reviewed by: Pavel Zakharov <[email protected]>
Reviewed-by: Richard Laager <[email protected]>
Reviewed-by: Alek Pinchuk <[email protected]>
Reviewed-by: Håkan Johansson <[email protected]>
Reviewed-by: Andreas Dilger <[email protected]>
Reviewed-by: DHE <[email protected]>
Reviewed-by: Richard Elling <[email protected]>
Reviewed-by: Gregor Kopka <[email protected]>
Reviewed-by: Kash Pande <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Reviewed-by: Matthew Ahrens <[email protected]>
Signed-off-by: Don Brady <[email protected]>
Closes #5182 

1 files changed, 170 insertions, 17 deletions

diff --git a/cmd/ztest/ztest.c b/cmd/ztest/ztest.c
index 71d5ed646..83d057a74 100644
--- a/cmd/ztest/ztest.c
+++ b/cmd/ztest/ztest.c
@@ -20,11 +20,12 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  * Copyright (c) 2014 Integros [integros.com]
  * Copyright 2017 Joyent, Inc.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 /*
@@ -149,6 +150,12 @@ typedef struct ztest_shared_hdr {
 
 static ztest_shared_hdr_t *ztest_shared_hdr;
 
+enum ztest_class_state {
+	ZTEST_VDEV_CLASS_OFF,
+	ZTEST_VDEV_CLASS_ON,
+	ZTEST_VDEV_CLASS_RND
+};
+
 typedef struct ztest_shared_opts {
 	char zo_pool[ZFS_MAX_DATASET_NAME_LEN];
 	char zo_dir[ZFS_MAX_DATASET_NAME_LEN];
@@ -171,6 +178,7 @@ typedef struct ztest_shared_opts {
 	uint64_t zo_maxloops;
 	uint64_t zo_metaslab_force_ganging;
 	int zo_mmp_test;
+	int zo_special_vdevs;
 } ztest_shared_opts_t;
 
 static const ztest_shared_opts_t ztest_opts_defaults = {
@@ -194,6 +202,7 @@ static const ztest_shared_opts_t ztest_opts_defaults = {
 	.zo_time = 300,			/* 5 minutes */
 	.zo_maxloops = 50,		/* max loops during spa_freeze() */
 	.zo_metaslab_force_ganging = 32 << 10,
+	.zo_special_vdevs = ZTEST_VDEV_CLASS_RND,
 };
 
 extern uint64_t metaslab_force_ganging;
@@ -342,6 +351,7 @@ ztest_func_t ztest_dsl_dataset_promote_busy;
 ztest_func_t ztest_vdev_attach_detach;
 ztest_func_t ztest_vdev_LUN_growth;
 ztest_func_t ztest_vdev_add_remove;
+ztest_func_t ztest_vdev_class_add;
 ztest_func_t ztest_vdev_aux_add_remove;
 ztest_func_t ztest_split_pool;
 ztest_func_t ztest_reguid;
@@ -398,6 +408,7 @@ ztest_info_t ztest_info[] = {
 	ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes),
 	ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely),
 	ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime),
+	ZTI_INIT(ztest_vdev_class_add, 1, &ztest_opts.zo_vdevtime),
 	ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime),
 	ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes),
 	ZTI_INIT(ztest_remap_blocks, 1, &zopt_sometimes),
@@ -666,6 +677,7 @@ usage(boolean_t requested)
 	    "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n"
 	    "\t[-P passtime (default: %llu sec)] time per pass\n"
 	    "\t[-B alt_ztest (default: <none>)] alternate ztest path\n"
+	    "\t[-C vdev class state (default: random)] special=on|off|random\n"
 	    "\t[-o variable=value] ... set global variable to an unsigned\n"
 	    "\t    32-bit integer value\n"
 	    "\t[-G dump zfs_dbgmsg buffer before exiting due to an error\n"
@@ -691,6 +703,46 @@ usage(boolean_t requested)
 	exit(requested ? 0 : 1);
 }
 
+
+static void
+ztest_parse_name_value(const char *input, ztest_shared_opts_t *zo)
+{
+	char name[32];
+	char *value;
+	int state = ZTEST_VDEV_CLASS_RND;
+
+	(void) strlcpy(name, input, sizeof (name));
+
+	value = strchr(name, '=');
+	if (value == NULL) {
+		(void) fprintf(stderr, "missing value in property=value "
+		    "'-C' argument (%s)\n", input);
+		usage(B_FALSE);
+	}
+	*(value) = '\0';
+	value++;
+
+	if (strcmp(value, "on") == 0) {
+		state = ZTEST_VDEV_CLASS_ON;
+	} else if (strcmp(value, "off") == 0) {
+		state = ZTEST_VDEV_CLASS_OFF;
+	} else if (strcmp(value, "random") == 0) {
+		state = ZTEST_VDEV_CLASS_RND;
+	} else {
+		(void) fprintf(stderr, "invalid property value '%s'\n", value);
+		usage(B_FALSE);
+	}
+
+	if (strcmp(name, "special") == 0) {
+		zo->zo_special_vdevs = state;
+	} else {
+		(void) fprintf(stderr, "invalid property name '%s'\n", name);
+		usage(B_FALSE);
+	}
+	if (zo->zo_verbose >= 3)
+		(void) printf("%s vdev state is '%s'\n", name, value);
+}
+
 static void
 process_options(int argc, char **argv)
 {
@@ -704,7 +756,7 @@ process_options(int argc, char **argv)
 	bcopy(&ztest_opts_defaults, zo, sizeof (*zo));
 
 	while ((opt = getopt(argc, argv,
-	    "v:s:a:m:r:R:d:t:g:i:k:p:f:MVET:P:hF:B:o:G")) != EOF) {
+	    "v:s:a:m:r:R:d:t:g:i:k:p:f:MVET:P:hF:B:C:o:G")) != EOF) {
 		value = 0;
 		switch (opt) {
 		case 'v':
@@ -795,6 +847,9 @@ process_options(int argc, char **argv)
 		case 'B':
 			(void) strlcpy(altdir, optarg, sizeof (altdir));
 			break;
+		case 'C':
+			ztest_parse_name_value(optarg, zo);
+			break;
 		case 'o':
 			if (set_global_var(optarg) != 0)
 				usage(B_FALSE);
@@ -1022,13 +1077,16 @@ make_vdev_mirror(char *path, char *aux, char *pool, size_t size,
 
 static nvlist_t *
 make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift,
-    int log, int r, int m, int t)
+    const char *class, int r, int m, int t)
 {
 	nvlist_t *root, **child;
 	int c;
+	boolean_t log;
 
 	ASSERT(t > 0);
 
+	log = (class != NULL && strcmp(class, "log") == 0);
+
 	child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL);
 
 	for (c = 0; c < t; c++) {
@@ -1036,6 +1094,12 @@ make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift,
 		    r, m);
 		VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
 		    log) == 0);
+
+		if (class != NULL && class[0] != '\0') {
+			ASSERT(m > 1 || log);   /* expecting a mirror */
+			VERIFY(nvlist_add_string(child[c],
+			    ZPOOL_CONFIG_ALLOCATION_BIAS, class) == 0);
+		}
 	}
 
 	VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0);
@@ -1075,6 +1139,8 @@ ztest_random_spa_version(uint64_t initial_version)
 static int
 ztest_random_blocksize(void)
 {
+	ASSERT(ztest_spa->spa_max_ashift != 0);
+
 	/*
 	 * Choose a block size >= the ashift.
 	 * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks.
@@ -2722,7 +2788,7 @@ ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id)
 	/*
 	 * Attempt to create using a bad file.
 	 */
-	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1);
+	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1);
 	VERIFY3U(ENOENT, ==,
 	    spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL));
 	nvlist_free(nvroot);
@@ -2730,7 +2796,7 @@ ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id)
 	/*
 	 * Attempt to create using a bad mirror.
 	 */
-	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 2, 1);
+	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 2, 1);
 	VERIFY3U(ENOENT, ==,
 	    spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL));
 	nvlist_free(nvroot);
@@ -2740,7 +2806,7 @@ ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id)
 	 * what's in the nvroot; we should fail with EEXIST.
 	 */
 	(void) pthread_rwlock_rdlock(&ztest_name_lock);
-	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1);
+	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1);
 	VERIFY3U(EEXIST, ==,
 	    spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL));
 	nvlist_free(nvroot);
@@ -2816,7 +2882,7 @@ ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id)
 	(void) spa_destroy(name);
 
 	nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0,
-	    0, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1);
+	    NULL, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1);
 
 	/*
 	 * If we're configuring a RAIDZ device then make sure that the
@@ -2990,10 +3056,16 @@ ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id)
 	 * If we have slogs then remove them 1/4 of the time.
 	 */
 	if (spa_has_slogs(spa) && ztest_random(4) == 0) {
+		metaslab_group_t *mg;
+
 		/*
-		 * Grab the guid from the head of the log class rotor.
+		 * find the first real slog in log allocation class
 		 */
-		guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid;
+		mg =  spa_log_class(spa)->mc_rotor;
+		while (!mg->mg_vd->vdev_islog)
+			mg = mg->mg_next;
+
+		guid = mg->mg_vd->vdev_guid;
 
 		spa_config_exit(spa, SCL_VDEV, FTAG);
 
@@ -3024,12 +3096,11 @@ ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id)
 		spa_config_exit(spa, SCL_VDEV, FTAG);
 
 		/*
-		 * Make 1/4 of the devices be log devices.
+		 * Make 1/4 of the devices be log devices
 		 */
 		nvroot = make_vdev_root(NULL, NULL, NULL,
-		    ztest_opts.zo_vdev_size, 0,
-		    ztest_random(4) == 0, ztest_opts.zo_raidz,
-		    zs->zs_mirrors, 1);
+		    ztest_opts.zo_vdev_size, 0, (ztest_random(4) == 0) ?
+		    "log" : NULL, ztest_opts.zo_raidz, zs->zs_mirrors, 1);
 
 		error = spa_vdev_add(spa, nvroot);
 		nvlist_free(nvroot);
@@ -3048,6 +3119,83 @@ ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id)
 	mutex_exit(&ztest_vdev_lock);
 }
 
+/* ARGSUSED */
+void
+ztest_vdev_class_add(ztest_ds_t *zd, uint64_t id)
+{
+	ztest_shared_t *zs = ztest_shared;
+	spa_t *spa = ztest_spa;
+	uint64_t leaves;
+	nvlist_t *nvroot;
+	const char *class = (ztest_random(2) == 0) ?
+	    VDEV_ALLOC_BIAS_SPECIAL : VDEV_ALLOC_BIAS_DEDUP;
+	int error;
+
+	/*
+	 * By default add a special vdev 50% of the time
+	 */
+	if ((ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_OFF) ||
+	    (ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_RND &&
+	    ztest_random(2) == 0)) {
+		return;
+	}
+
+	mutex_enter(&ztest_vdev_lock);
+
+	/* Only test with mirrors */
+	if (zs->zs_mirrors < 2) {
+		mutex_exit(&ztest_vdev_lock);
+		return;
+	}
+
+	/* requires feature@allocation_classes */
+	if (!spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)) {
+		mutex_exit(&ztest_vdev_lock);
+		return;
+	}
+
+	leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz;
+
+	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
+	ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves;
+	spa_config_exit(spa, SCL_VDEV, FTAG);
+
+	nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
+	    class, ztest_opts.zo_raidz, zs->zs_mirrors, 1);
+
+	error = spa_vdev_add(spa, nvroot);
+	nvlist_free(nvroot);
+
+	if (error == ENOSPC)
+		ztest_record_enospc("spa_vdev_add");
+	else if (error != 0)
+		fatal(0, "spa_vdev_add() = %d", error);
+
+	/*
+	 * 50% of the time allow small blocks in the special class
+	 */
+	if (error == 0 &&
+	    spa_special_class(spa)->mc_groups == 1 && ztest_random(2) == 0) {
+		if (ztest_opts.zo_verbose >= 3)
+			(void) printf("Enabling special VDEV small blocks\n");
+		(void) ztest_dsl_prop_set_uint64(zd->zd_name,
+		    ZFS_PROP_SPECIAL_SMALL_BLOCKS, 32768, B_FALSE);
+	}
+
+	mutex_exit(&ztest_vdev_lock);
+
+	if (ztest_opts.zo_verbose >= 3) {
+		metaslab_class_t *mc;
+
+		if (strcmp(class, VDEV_ALLOC_BIAS_SPECIAL) == 0)
+			mc = spa_special_class(spa);
+		else
+			mc = spa_dedup_class(spa);
+		(void) printf("Added a %s mirrored vdev (of %d)\n",
+		    class, (int)mc->mc_groups);
+	}
+}
+
 /*
  * Verify that adding/removing aux devices (l2arc, hot spare) works as expected.
  */
@@ -3114,7 +3262,7 @@ ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id)
 		 * Add a new device.
 		 */
 		nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL,
-		    (ztest_opts.zo_vdev_size * 5) / 4, 0, 0, 0, 0, 1);
+		    (ztest_opts.zo_vdev_size * 5) / 4, 0, NULL, 0, 0, 1);
 		error = spa_vdev_add(spa, nvroot);
 
 		switch (error) {
@@ -3316,11 +3464,15 @@ ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id)
 	 * Locate this vdev.
 	 */
 	oldvd = rvd->vdev_child[top];
+
+	/* pick a child from the mirror */
 	if (zs->zs_mirrors >= 1) {
 		ASSERT(oldvd->vdev_ops == &vdev_mirror_ops);
 		ASSERT(oldvd->vdev_children >= zs->zs_mirrors);
 		oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz];
 	}
+
+	/* pick a child out of the raidz group */
 	if (ztest_opts.zo_raidz > 1) {
 		ASSERT(oldvd->vdev_ops == &vdev_raidz_ops);
 		ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz);
@@ -3422,7 +3574,7 @@ ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id)
 	 * Build the nvlist describing newpath.
 	 */
 	root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0,
-	    ashift, 0, 0, 0, 1);
+	    ashift, NULL, 0, 0, 1);
 
 	error = spa_vdev_attach(spa, oldguid, root, replacing);
 
@@ -3688,7 +3840,7 @@ ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id)
 		return;
 	}
 	ASSERT(psize > 0);
-	newsize = psize + psize / 8;
+	newsize = psize + MAX(psize / 8, SPA_MAXBLOCKSIZE);
 	ASSERT3U(newsize, >, psize);
 
 	if (ztest_opts.zo_verbose >= 6) {
@@ -7027,6 +7179,7 @@ make_random_props(void)
 	nvlist_t *props;
 
 	VERIFY0(nvlist_alloc(&props, NV_UNIQUE_NAME, 0));
+
 	if (ztest_random(2) == 0)
 		return (props);
 
@@ -7113,7 +7266,7 @@ ztest_init(ztest_shared_t *zs)
 	zs->zs_splits = 0;
 	zs->zs_mirrors = ztest_opts.zo_mirrors;
 	nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
-	    0, ztest_opts.zo_raidz, zs->zs_mirrors, 1);
+	    NULL, ztest_opts.zo_raidz, zs->zs_mirrors, 1);
 	props = make_random_props();
 
 	/*