4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2016 by Delphix. All rights reserved.
24 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
27 #include <sys/zfs_context.h>
29 #include <sys/dmu_tx.h>
30 #include <sys/space_map.h>
31 #include <sys/metaslab_impl.h>
32 #include <sys/vdev_impl.h>
34 #include <sys/spa_impl.h>
35 #include <sys/zfeature.h>
37 #define WITH_DF_BLOCK_ALLOCATOR
39 #define GANG_ALLOCATION(flags) \
40 ((flags) & (METASLAB_GANG_CHILD | METASLAB_GANG_HEADER))
43 * Metaslab granularity, in bytes. This is roughly similar to what would be
44 * referred to as the "stripe size" in traditional RAID arrays. In normal
45 * operation, we will try to write this amount of data to a top-level vdev
46 * before moving on to the next one.
48 unsigned long metaslab_aliquot
= 512 << 10;
50 uint64_t metaslab_gang_bang
= SPA_MAXBLOCKSIZE
+ 1; /* force gang blocks */
53 * The in-core space map representation is more compact than its on-disk form.
54 * The zfs_condense_pct determines how much more compact the in-core
55 * space map representation must be before we compact it on-disk.
56 * Values should be greater than or equal to 100.
58 int zfs_condense_pct
= 200;
61 * Condensing a metaslab is not guaranteed to actually reduce the amount of
62 * space used on disk. In particular, a space map uses data in increments of
63 * MAX(1 << ashift, space_map_blksz), so a metaslab might use the
64 * same number of blocks after condensing. Since the goal of condensing is to
65 * reduce the number of IOPs required to read the space map, we only want to
66 * condense when we can be sure we will reduce the number of blocks used by the
67 * space map. Unfortunately, we cannot precisely compute whether or not this is
68 * the case in metaslab_should_condense since we are holding ms_lock. Instead,
69 * we apply the following heuristic: do not condense a spacemap unless the
70 * uncondensed size consumes greater than zfs_metaslab_condense_block_threshold
73 int zfs_metaslab_condense_block_threshold
= 4;
76 * The zfs_mg_noalloc_threshold defines which metaslab groups should
77 * be eligible for allocation. The value is defined as a percentage of
78 * free space. Metaslab groups that have more free space than
79 * zfs_mg_noalloc_threshold are always eligible for allocations. Once
80 * a metaslab group's free space is less than or equal to the
81 * zfs_mg_noalloc_threshold the allocator will avoid allocating to that
82 * group unless all groups in the pool have reached zfs_mg_noalloc_threshold.
83 * Once all groups in the pool reach zfs_mg_noalloc_threshold then all
84 * groups are allowed to accept allocations. Gang blocks are always
85 * eligible to allocate on any metaslab group. The default value of 0 means
86 * no metaslab group will be excluded based on this criterion.
88 int zfs_mg_noalloc_threshold
= 0;
91 * Metaslab groups are considered eligible for allocations if their
92 * fragmenation metric (measured as a percentage) is less than or equal to
93 * zfs_mg_fragmentation_threshold. If a metaslab group exceeds this threshold
94 * then it will be skipped unless all metaslab groups within the metaslab
95 * class have also crossed this threshold.
97 int zfs_mg_fragmentation_threshold
= 85;
100 * Allow metaslabs to keep their active state as long as their fragmentation
101 * percentage is less than or equal to zfs_metaslab_fragmentation_threshold. An
102 * active metaslab that exceeds this threshold will no longer keep its active
103 * status allowing better metaslabs to be selected.
105 int zfs_metaslab_fragmentation_threshold
= 70;
108 * When set will load all metaslabs when pool is first opened.
110 int metaslab_debug_load
= 0;
113 * When set will prevent metaslabs from being unloaded.
115 int metaslab_debug_unload
= 0;
118 * Minimum size which forces the dynamic allocator to change
119 * it's allocation strategy. Once the space map cannot satisfy
120 * an allocation of this size then it switches to using more
121 * aggressive strategy (i.e search by size rather than offset).
123 uint64_t metaslab_df_alloc_threshold
= SPA_OLD_MAXBLOCKSIZE
;
126 * The minimum free space, in percent, which must be available
127 * in a space map to continue allocations in a first-fit fashion.
128 * Once the space map's free space drops below this level we dynamically
129 * switch to using best-fit allocations.
131 int metaslab_df_free_pct
= 4;
134 * Percentage of all cpus that can be used by the metaslab taskq.
136 int metaslab_load_pct
= 50;
139 * Determines how many txgs a metaslab may remain loaded without having any
140 * allocations from it. As long as a metaslab continues to be used we will
143 int metaslab_unload_delay
= TXG_SIZE
* 2;
146 * Max number of metaslabs per group to preload.
148 int metaslab_preload_limit
= SPA_DVAS_PER_BP
;
151 * Enable/disable preloading of metaslab.
153 int metaslab_preload_enabled
= B_TRUE
;
156 * Enable/disable fragmentation weighting on metaslabs.
158 int metaslab_fragmentation_factor_enabled
= B_TRUE
;
161 * Enable/disable lba weighting (i.e. outer tracks are given preference).
163 int metaslab_lba_weighting_enabled
= B_TRUE
;
166 * Enable/disable metaslab group biasing.
168 int metaslab_bias_enabled
= B_TRUE
;
172 * Enable/disable segment-based metaslab selection.
174 int zfs_metaslab_segment_weight_enabled
= B_TRUE
;
177 * When using segment-based metaslab selection, we will continue
178 * allocating from the active metaslab until we have exhausted
179 * zfs_metaslab_switch_threshold of its buckets.
181 int zfs_metaslab_switch_threshold
= 2;
184 * Internal switch to enable/disable the metaslab allocation tracing
187 #ifdef _METASLAB_TRACING
188 boolean_t metaslab_trace_enabled
= B_TRUE
;
192 * Maximum entries that the metaslab allocation tracing facility will keep
193 * in a given list when running in non-debug mode. We limit the number
194 * of entries in non-debug mode to prevent us from using up too much memory.
195 * The limit should be sufficiently large that we don't expect any allocation
196 * to every exceed this value. In debug mode, the system will panic if this
197 * limit is ever reached allowing for further investigation.
199 #ifdef _METASLAB_TRACING
200 uint64_t metaslab_trace_max_entries
= 5000;
203 static uint64_t metaslab_weight(metaslab_t
*);
204 static void metaslab_set_fragmentation(metaslab_t
*);
206 #ifdef _METASLAB_TRACING
207 kmem_cache_t
*metaslab_alloc_trace_cache
;
211 * ==========================================================================
213 * ==========================================================================
216 metaslab_class_create(spa_t
*spa
, metaslab_ops_t
*ops
)
218 metaslab_class_t
*mc
;
220 mc
= kmem_zalloc(sizeof (metaslab_class_t
), KM_SLEEP
);
225 mutex_init(&mc
->mc_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
226 refcount_create_tracked(&mc
->mc_alloc_slots
);
232 metaslab_class_destroy(metaslab_class_t
*mc
)
234 ASSERT(mc
->mc_rotor
== NULL
);
235 ASSERT(mc
->mc_alloc
== 0);
236 ASSERT(mc
->mc_deferred
== 0);
237 ASSERT(mc
->mc_space
== 0);
238 ASSERT(mc
->mc_dspace
== 0);
240 refcount_destroy(&mc
->mc_alloc_slots
);
241 mutex_destroy(&mc
->mc_lock
);
242 kmem_free(mc
, sizeof (metaslab_class_t
));
246 metaslab_class_validate(metaslab_class_t
*mc
)
248 metaslab_group_t
*mg
;
252 * Must hold one of the spa_config locks.
254 ASSERT(spa_config_held(mc
->mc_spa
, SCL_ALL
, RW_READER
) ||
255 spa_config_held(mc
->mc_spa
, SCL_ALL
, RW_WRITER
));
257 if ((mg
= mc
->mc_rotor
) == NULL
)
262 ASSERT(vd
->vdev_mg
!= NULL
);
263 ASSERT3P(vd
->vdev_top
, ==, vd
);
264 ASSERT3P(mg
->mg_class
, ==, mc
);
265 ASSERT3P(vd
->vdev_ops
, !=, &vdev_hole_ops
);
266 } while ((mg
= mg
->mg_next
) != mc
->mc_rotor
);
272 metaslab_class_space_update(metaslab_class_t
*mc
, int64_t alloc_delta
,
273 int64_t defer_delta
, int64_t space_delta
, int64_t dspace_delta
)
275 atomic_add_64(&mc
->mc_alloc
, alloc_delta
);
276 atomic_add_64(&mc
->mc_deferred
, defer_delta
);
277 atomic_add_64(&mc
->mc_space
, space_delta
);
278 atomic_add_64(&mc
->mc_dspace
, dspace_delta
);
282 metaslab_class_get_alloc(metaslab_class_t
*mc
)
284 return (mc
->mc_alloc
);
288 metaslab_class_get_deferred(metaslab_class_t
*mc
)
290 return (mc
->mc_deferred
);
294 metaslab_class_get_space(metaslab_class_t
*mc
)
296 return (mc
->mc_space
);
300 metaslab_class_get_dspace(metaslab_class_t
*mc
)
302 return (spa_deflate(mc
->mc_spa
) ? mc
->mc_dspace
: mc
->mc_space
);
306 metaslab_class_histogram_verify(metaslab_class_t
*mc
)
308 vdev_t
*rvd
= mc
->mc_spa
->spa_root_vdev
;
312 if ((zfs_flags
& ZFS_DEBUG_HISTOGRAM_VERIFY
) == 0)
315 mc_hist
= kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE
,
318 for (int c
= 0; c
< rvd
->vdev_children
; c
++) {
319 vdev_t
*tvd
= rvd
->vdev_child
[c
];
320 metaslab_group_t
*mg
= tvd
->vdev_mg
;
323 * Skip any holes, uninitialized top-levels, or
324 * vdevs that are not in this metalab class.
326 if (tvd
->vdev_ishole
|| tvd
->vdev_ms_shift
== 0 ||
327 mg
->mg_class
!= mc
) {
331 for (i
= 0; i
< RANGE_TREE_HISTOGRAM_SIZE
; i
++)
332 mc_hist
[i
] += mg
->mg_histogram
[i
];
335 for (i
= 0; i
< RANGE_TREE_HISTOGRAM_SIZE
; i
++)
336 VERIFY3U(mc_hist
[i
], ==, mc
->mc_histogram
[i
]);
338 kmem_free(mc_hist
, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE
);
342 * Calculate the metaslab class's fragmentation metric. The metric
343 * is weighted based on the space contribution of each metaslab group.
344 * The return value will be a number between 0 and 100 (inclusive), or
345 * ZFS_FRAG_INVALID if the metric has not been set. See comment above the
346 * zfs_frag_table for more information about the metric.
349 metaslab_class_fragmentation(metaslab_class_t
*mc
)
351 vdev_t
*rvd
= mc
->mc_spa
->spa_root_vdev
;
352 uint64_t fragmentation
= 0;
354 spa_config_enter(mc
->mc_spa
, SCL_VDEV
, FTAG
, RW_READER
);
356 for (int c
= 0; c
< rvd
->vdev_children
; c
++) {
357 vdev_t
*tvd
= rvd
->vdev_child
[c
];
358 metaslab_group_t
*mg
= tvd
->vdev_mg
;
361 * Skip any holes, uninitialized top-levels, or
362 * vdevs that are not in this metalab class.
364 if (tvd
->vdev_ishole
|| tvd
->vdev_ms_shift
== 0 ||
365 mg
->mg_class
!= mc
) {
370 * If a metaslab group does not contain a fragmentation
371 * metric then just bail out.
373 if (mg
->mg_fragmentation
== ZFS_FRAG_INVALID
) {
374 spa_config_exit(mc
->mc_spa
, SCL_VDEV
, FTAG
);
375 return (ZFS_FRAG_INVALID
);
379 * Determine how much this metaslab_group is contributing
380 * to the overall pool fragmentation metric.
382 fragmentation
+= mg
->mg_fragmentation
*
383 metaslab_group_get_space(mg
);
385 fragmentation
/= metaslab_class_get_space(mc
);
387 ASSERT3U(fragmentation
, <=, 100);
388 spa_config_exit(mc
->mc_spa
, SCL_VDEV
, FTAG
);
389 return (fragmentation
);
393 * Calculate the amount of expandable space that is available in
394 * this metaslab class. If a device is expanded then its expandable
395 * space will be the amount of allocatable space that is currently not
396 * part of this metaslab class.
399 metaslab_class_expandable_space(metaslab_class_t
*mc
)
401 vdev_t
*rvd
= mc
->mc_spa
->spa_root_vdev
;
404 spa_config_enter(mc
->mc_spa
, SCL_VDEV
, FTAG
, RW_READER
);
405 for (int c
= 0; c
< rvd
->vdev_children
; c
++) {
406 vdev_t
*tvd
= rvd
->vdev_child
[c
];
407 metaslab_group_t
*mg
= tvd
->vdev_mg
;
409 if (tvd
->vdev_ishole
|| tvd
->vdev_ms_shift
== 0 ||
410 mg
->mg_class
!= mc
) {
415 * Calculate if we have enough space to add additional
416 * metaslabs. We report the expandable space in terms
417 * of the metaslab size since that's the unit of expansion.
419 space
+= P2ALIGN(tvd
->vdev_max_asize
- tvd
->vdev_asize
,
420 1ULL << tvd
->vdev_ms_shift
);
422 spa_config_exit(mc
->mc_spa
, SCL_VDEV
, FTAG
);
427 metaslab_compare(const void *x1
, const void *x2
)
429 const metaslab_t
*m1
= (const metaslab_t
*)x1
;
430 const metaslab_t
*m2
= (const metaslab_t
*)x2
;
432 int cmp
= AVL_CMP(m2
->ms_weight
, m1
->ms_weight
);
436 IMPLY(AVL_CMP(m1
->ms_start
, m2
->ms_start
) == 0, m1
== m2
);
438 return (AVL_CMP(m1
->ms_start
, m2
->ms_start
));
442 * Verify that the space accounting on disk matches the in-core range_trees.
445 metaslab_verify_space(metaslab_t
*msp
, uint64_t txg
)
447 spa_t
*spa
= msp
->ms_group
->mg_vd
->vdev_spa
;
448 uint64_t allocated
= 0;
449 uint64_t sm_free_space
, msp_free_space
;
451 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
453 if ((zfs_flags
& ZFS_DEBUG_METASLAB_VERIFY
) == 0)
457 * We can only verify the metaslab space when we're called
458 * from syncing context with a loaded metaslab that has an allocated
459 * space map. Calling this in non-syncing context does not
460 * provide a consistent view of the metaslab since we're performing
461 * allocations in the future.
463 if (txg
!= spa_syncing_txg(spa
) || msp
->ms_sm
== NULL
||
467 sm_free_space
= msp
->ms_size
- space_map_allocated(msp
->ms_sm
) -
468 space_map_alloc_delta(msp
->ms_sm
);
471 * Account for future allocations since we would have already
472 * deducted that space from the ms_freetree.
474 for (int t
= 0; t
< TXG_CONCURRENT_STATES
; t
++) {
476 range_tree_space(msp
->ms_alloctree
[(txg
+ t
) & TXG_MASK
]);
479 msp_free_space
= range_tree_space(msp
->ms_tree
) + allocated
+
480 msp
->ms_deferspace
+ range_tree_space(msp
->ms_freedtree
);
482 VERIFY3U(sm_free_space
, ==, msp_free_space
);
486 * ==========================================================================
488 * ==========================================================================
491 * Update the allocatable flag and the metaslab group's capacity.
492 * The allocatable flag is set to true if the capacity is below
493 * the zfs_mg_noalloc_threshold or has a fragmentation value that is
494 * greater than zfs_mg_fragmentation_threshold. If a metaslab group
495 * transitions from allocatable to non-allocatable or vice versa then the
496 * metaslab group's class is updated to reflect the transition.
499 metaslab_group_alloc_update(metaslab_group_t
*mg
)
501 vdev_t
*vd
= mg
->mg_vd
;
502 metaslab_class_t
*mc
= mg
->mg_class
;
503 vdev_stat_t
*vs
= &vd
->vdev_stat
;
504 boolean_t was_allocatable
;
505 boolean_t was_initialized
;
507 ASSERT(vd
== vd
->vdev_top
);
509 mutex_enter(&mg
->mg_lock
);
510 was_allocatable
= mg
->mg_allocatable
;
511 was_initialized
= mg
->mg_initialized
;
513 mg
->mg_free_capacity
= ((vs
->vs_space
- vs
->vs_alloc
) * 100) /
516 mutex_enter(&mc
->mc_lock
);
519 * If the metaslab group was just added then it won't
520 * have any space until we finish syncing out this txg.
521 * At that point we will consider it initialized and available
522 * for allocations. We also don't consider non-activated
523 * metaslab groups (e.g. vdevs that are in the middle of being removed)
524 * to be initialized, because they can't be used for allocation.
526 mg
->mg_initialized
= metaslab_group_initialized(mg
);
527 if (!was_initialized
&& mg
->mg_initialized
) {
529 } else if (was_initialized
&& !mg
->mg_initialized
) {
530 ASSERT3U(mc
->mc_groups
, >, 0);
533 if (mg
->mg_initialized
)
534 mg
->mg_no_free_space
= B_FALSE
;
537 * A metaslab group is considered allocatable if it has plenty
538 * of free space or is not heavily fragmented. We only take
539 * fragmentation into account if the metaslab group has a valid
540 * fragmentation metric (i.e. a value between 0 and 100).
542 mg
->mg_allocatable
= (mg
->mg_activation_count
> 0 &&
543 mg
->mg_free_capacity
> zfs_mg_noalloc_threshold
&&
544 (mg
->mg_fragmentation
== ZFS_FRAG_INVALID
||
545 mg
->mg_fragmentation
<= zfs_mg_fragmentation_threshold
));
548 * The mc_alloc_groups maintains a count of the number of
549 * groups in this metaslab class that are still above the
550 * zfs_mg_noalloc_threshold. This is used by the allocating
551 * threads to determine if they should avoid allocations to
552 * a given group. The allocator will avoid allocations to a group
553 * if that group has reached or is below the zfs_mg_noalloc_threshold
554 * and there are still other groups that are above the threshold.
555 * When a group transitions from allocatable to non-allocatable or
556 * vice versa we update the metaslab class to reflect that change.
557 * When the mc_alloc_groups value drops to 0 that means that all
558 * groups have reached the zfs_mg_noalloc_threshold making all groups
559 * eligible for allocations. This effectively means that all devices
560 * are balanced again.
562 if (was_allocatable
&& !mg
->mg_allocatable
)
563 mc
->mc_alloc_groups
--;
564 else if (!was_allocatable
&& mg
->mg_allocatable
)
565 mc
->mc_alloc_groups
++;
566 mutex_exit(&mc
->mc_lock
);
568 mutex_exit(&mg
->mg_lock
);
572 metaslab_group_create(metaslab_class_t
*mc
, vdev_t
*vd
)
574 metaslab_group_t
*mg
;
576 mg
= kmem_zalloc(sizeof (metaslab_group_t
), KM_SLEEP
);
577 mutex_init(&mg
->mg_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
578 avl_create(&mg
->mg_metaslab_tree
, metaslab_compare
,
579 sizeof (metaslab_t
), offsetof(struct metaslab
, ms_group_node
));
582 mg
->mg_activation_count
= 0;
583 mg
->mg_initialized
= B_FALSE
;
584 mg
->mg_no_free_space
= B_TRUE
;
585 refcount_create_tracked(&mg
->mg_alloc_queue_depth
);
587 mg
->mg_taskq
= taskq_create("metaslab_group_taskq", metaslab_load_pct
,
588 maxclsyspri
, 10, INT_MAX
, TASKQ_THREADS_CPU_PCT
| TASKQ_DYNAMIC
);
594 metaslab_group_destroy(metaslab_group_t
*mg
)
596 ASSERT(mg
->mg_prev
== NULL
);
597 ASSERT(mg
->mg_next
== NULL
);
599 * We may have gone below zero with the activation count
600 * either because we never activated in the first place or
601 * because we're done, and possibly removing the vdev.
603 ASSERT(mg
->mg_activation_count
<= 0);
605 taskq_destroy(mg
->mg_taskq
);
606 avl_destroy(&mg
->mg_metaslab_tree
);
607 mutex_destroy(&mg
->mg_lock
);
608 refcount_destroy(&mg
->mg_alloc_queue_depth
);
609 kmem_free(mg
, sizeof (metaslab_group_t
));
613 metaslab_group_activate(metaslab_group_t
*mg
)
615 metaslab_class_t
*mc
= mg
->mg_class
;
616 metaslab_group_t
*mgprev
, *mgnext
;
618 ASSERT(spa_config_held(mc
->mc_spa
, SCL_ALLOC
, RW_WRITER
));
620 ASSERT(mc
->mc_rotor
!= mg
);
621 ASSERT(mg
->mg_prev
== NULL
);
622 ASSERT(mg
->mg_next
== NULL
);
623 ASSERT(mg
->mg_activation_count
<= 0);
625 if (++mg
->mg_activation_count
<= 0)
628 mg
->mg_aliquot
= metaslab_aliquot
* MAX(1, mg
->mg_vd
->vdev_children
);
629 metaslab_group_alloc_update(mg
);
631 if ((mgprev
= mc
->mc_rotor
) == NULL
) {
635 mgnext
= mgprev
->mg_next
;
636 mg
->mg_prev
= mgprev
;
637 mg
->mg_next
= mgnext
;
638 mgprev
->mg_next
= mg
;
639 mgnext
->mg_prev
= mg
;
645 metaslab_group_passivate(metaslab_group_t
*mg
)
647 metaslab_class_t
*mc
= mg
->mg_class
;
648 metaslab_group_t
*mgprev
, *mgnext
;
650 ASSERT(spa_config_held(mc
->mc_spa
, SCL_ALLOC
, RW_WRITER
));
652 if (--mg
->mg_activation_count
!= 0) {
653 ASSERT(mc
->mc_rotor
!= mg
);
654 ASSERT(mg
->mg_prev
== NULL
);
655 ASSERT(mg
->mg_next
== NULL
);
656 ASSERT(mg
->mg_activation_count
< 0);
660 taskq_wait_outstanding(mg
->mg_taskq
, 0);
661 metaslab_group_alloc_update(mg
);
663 mgprev
= mg
->mg_prev
;
664 mgnext
= mg
->mg_next
;
669 mc
->mc_rotor
= mgnext
;
670 mgprev
->mg_next
= mgnext
;
671 mgnext
->mg_prev
= mgprev
;
679 metaslab_group_initialized(metaslab_group_t
*mg
)
681 vdev_t
*vd
= mg
->mg_vd
;
682 vdev_stat_t
*vs
= &vd
->vdev_stat
;
684 return (vs
->vs_space
!= 0 && mg
->mg_activation_count
> 0);
688 metaslab_group_get_space(metaslab_group_t
*mg
)
690 return ((1ULL << mg
->mg_vd
->vdev_ms_shift
) * mg
->mg_vd
->vdev_ms_count
);
694 metaslab_group_histogram_verify(metaslab_group_t
*mg
)
697 vdev_t
*vd
= mg
->mg_vd
;
698 uint64_t ashift
= vd
->vdev_ashift
;
701 if ((zfs_flags
& ZFS_DEBUG_HISTOGRAM_VERIFY
) == 0)
704 mg_hist
= kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE
,
707 ASSERT3U(RANGE_TREE_HISTOGRAM_SIZE
, >=,
708 SPACE_MAP_HISTOGRAM_SIZE
+ ashift
);
710 for (int m
= 0; m
< vd
->vdev_ms_count
; m
++) {
711 metaslab_t
*msp
= vd
->vdev_ms
[m
];
713 if (msp
->ms_sm
== NULL
)
716 for (i
= 0; i
< SPACE_MAP_HISTOGRAM_SIZE
; i
++)
717 mg_hist
[i
+ ashift
] +=
718 msp
->ms_sm
->sm_phys
->smp_histogram
[i
];
721 for (i
= 0; i
< RANGE_TREE_HISTOGRAM_SIZE
; i
++)
722 VERIFY3U(mg_hist
[i
], ==, mg
->mg_histogram
[i
]);
724 kmem_free(mg_hist
, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE
);
728 metaslab_group_histogram_add(metaslab_group_t
*mg
, metaslab_t
*msp
)
730 metaslab_class_t
*mc
= mg
->mg_class
;
731 uint64_t ashift
= mg
->mg_vd
->vdev_ashift
;
733 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
734 if (msp
->ms_sm
== NULL
)
737 mutex_enter(&mg
->mg_lock
);
738 for (int i
= 0; i
< SPACE_MAP_HISTOGRAM_SIZE
; i
++) {
739 mg
->mg_histogram
[i
+ ashift
] +=
740 msp
->ms_sm
->sm_phys
->smp_histogram
[i
];
741 mc
->mc_histogram
[i
+ ashift
] +=
742 msp
->ms_sm
->sm_phys
->smp_histogram
[i
];
744 mutex_exit(&mg
->mg_lock
);
748 metaslab_group_histogram_remove(metaslab_group_t
*mg
, metaslab_t
*msp
)
750 metaslab_class_t
*mc
= mg
->mg_class
;
751 uint64_t ashift
= mg
->mg_vd
->vdev_ashift
;
753 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
754 if (msp
->ms_sm
== NULL
)
757 mutex_enter(&mg
->mg_lock
);
758 for (int i
= 0; i
< SPACE_MAP_HISTOGRAM_SIZE
; i
++) {
759 ASSERT3U(mg
->mg_histogram
[i
+ ashift
], >=,
760 msp
->ms_sm
->sm_phys
->smp_histogram
[i
]);
761 ASSERT3U(mc
->mc_histogram
[i
+ ashift
], >=,
762 msp
->ms_sm
->sm_phys
->smp_histogram
[i
]);
764 mg
->mg_histogram
[i
+ ashift
] -=
765 msp
->ms_sm
->sm_phys
->smp_histogram
[i
];
766 mc
->mc_histogram
[i
+ ashift
] -=
767 msp
->ms_sm
->sm_phys
->smp_histogram
[i
];
769 mutex_exit(&mg
->mg_lock
);
773 metaslab_group_add(metaslab_group_t
*mg
, metaslab_t
*msp
)
775 ASSERT(msp
->ms_group
== NULL
);
776 mutex_enter(&mg
->mg_lock
);
779 avl_add(&mg
->mg_metaslab_tree
, msp
);
780 mutex_exit(&mg
->mg_lock
);
782 mutex_enter(&msp
->ms_lock
);
783 metaslab_group_histogram_add(mg
, msp
);
784 mutex_exit(&msp
->ms_lock
);
788 metaslab_group_remove(metaslab_group_t
*mg
, metaslab_t
*msp
)
790 mutex_enter(&msp
->ms_lock
);
791 metaslab_group_histogram_remove(mg
, msp
);
792 mutex_exit(&msp
->ms_lock
);
794 mutex_enter(&mg
->mg_lock
);
795 ASSERT(msp
->ms_group
== mg
);
796 avl_remove(&mg
->mg_metaslab_tree
, msp
);
797 msp
->ms_group
= NULL
;
798 mutex_exit(&mg
->mg_lock
);
802 metaslab_group_sort(metaslab_group_t
*mg
, metaslab_t
*msp
, uint64_t weight
)
805 * Although in principle the weight can be any value, in
806 * practice we do not use values in the range [1, 511].
808 ASSERT(weight
>= SPA_MINBLOCKSIZE
|| weight
== 0);
809 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
811 mutex_enter(&mg
->mg_lock
);
812 ASSERT(msp
->ms_group
== mg
);
813 avl_remove(&mg
->mg_metaslab_tree
, msp
);
814 msp
->ms_weight
= weight
;
815 avl_add(&mg
->mg_metaslab_tree
, msp
);
816 mutex_exit(&mg
->mg_lock
);
820 * Calculate the fragmentation for a given metaslab group. We can use
821 * a simple average here since all metaslabs within the group must have
822 * the same size. The return value will be a value between 0 and 100
823 * (inclusive), or ZFS_FRAG_INVALID if less than half of the metaslab in this
824 * group have a fragmentation metric.
827 metaslab_group_fragmentation(metaslab_group_t
*mg
)
829 vdev_t
*vd
= mg
->mg_vd
;
830 uint64_t fragmentation
= 0;
831 uint64_t valid_ms
= 0;
833 for (int m
= 0; m
< vd
->vdev_ms_count
; m
++) {
834 metaslab_t
*msp
= vd
->vdev_ms
[m
];
836 if (msp
->ms_fragmentation
== ZFS_FRAG_INVALID
)
840 fragmentation
+= msp
->ms_fragmentation
;
843 if (valid_ms
<= vd
->vdev_ms_count
/ 2)
844 return (ZFS_FRAG_INVALID
);
846 fragmentation
/= valid_ms
;
847 ASSERT3U(fragmentation
, <=, 100);
848 return (fragmentation
);
852 * Determine if a given metaslab group should skip allocations. A metaslab
853 * group should avoid allocations if its free capacity is less than the
854 * zfs_mg_noalloc_threshold or its fragmentation metric is greater than
855 * zfs_mg_fragmentation_threshold and there is at least one metaslab group
856 * that can still handle allocations. If the allocation throttle is enabled
857 * then we skip allocations to devices that have reached their maximum
858 * allocation queue depth unless the selected metaslab group is the only
859 * eligible group remaining.
862 metaslab_group_allocatable(metaslab_group_t
*mg
, metaslab_group_t
*rotor
,
865 spa_t
*spa
= mg
->mg_vd
->vdev_spa
;
866 metaslab_class_t
*mc
= mg
->mg_class
;
869 * We can only consider skipping this metaslab group if it's
870 * in the normal metaslab class and there are other metaslab
871 * groups to select from. Otherwise, we always consider it eligible
874 if (mc
!= spa_normal_class(spa
) || mc
->mc_groups
<= 1)
878 * If the metaslab group's mg_allocatable flag is set (see comments
879 * in metaslab_group_alloc_update() for more information) and
880 * the allocation throttle is disabled then allow allocations to this
881 * device. However, if the allocation throttle is enabled then
882 * check if we have reached our allocation limit (mg_alloc_queue_depth)
883 * to determine if we should allow allocations to this metaslab group.
884 * If all metaslab groups are no longer considered allocatable
885 * (mc_alloc_groups == 0) or we're trying to allocate the smallest
886 * gang block size then we allow allocations on this metaslab group
887 * regardless of the mg_allocatable or throttle settings.
889 if (mg
->mg_allocatable
) {
890 metaslab_group_t
*mgp
;
892 uint64_t qmax
= mg
->mg_max_alloc_queue_depth
;
894 if (!mc
->mc_alloc_throttle_enabled
)
898 * If this metaslab group does not have any free space, then
899 * there is no point in looking further.
901 if (mg
->mg_no_free_space
)
904 qdepth
= refcount_count(&mg
->mg_alloc_queue_depth
);
907 * If this metaslab group is below its qmax or it's
908 * the only allocatable metasable group, then attempt
909 * to allocate from it.
911 if (qdepth
< qmax
|| mc
->mc_alloc_groups
== 1)
913 ASSERT3U(mc
->mc_alloc_groups
, >, 1);
916 * Since this metaslab group is at or over its qmax, we
917 * need to determine if there are metaslab groups after this
918 * one that might be able to handle this allocation. This is
919 * racy since we can't hold the locks for all metaslab
920 * groups at the same time when we make this check.
922 for (mgp
= mg
->mg_next
; mgp
!= rotor
; mgp
= mgp
->mg_next
) {
923 qmax
= mgp
->mg_max_alloc_queue_depth
;
925 qdepth
= refcount_count(&mgp
->mg_alloc_queue_depth
);
928 * If there is another metaslab group that
929 * might be able to handle the allocation, then
930 * we return false so that we skip this group.
932 if (qdepth
< qmax
&& !mgp
->mg_no_free_space
)
937 * We didn't find another group to handle the allocation
938 * so we can't skip this metaslab group even though
939 * we are at or over our qmax.
943 } else if (mc
->mc_alloc_groups
== 0 || psize
== SPA_MINBLOCKSIZE
) {
950 * ==========================================================================
951 * Range tree callbacks
952 * ==========================================================================
956 * Comparison function for the private size-ordered tree. Tree is sorted
957 * by size, larger sizes at the end of the tree.
960 metaslab_rangesize_compare(const void *x1
, const void *x2
)
962 const range_seg_t
*r1
= x1
;
963 const range_seg_t
*r2
= x2
;
964 uint64_t rs_size1
= r1
->rs_end
- r1
->rs_start
;
965 uint64_t rs_size2
= r2
->rs_end
- r2
->rs_start
;
967 int cmp
= AVL_CMP(rs_size1
, rs_size2
);
971 return (AVL_CMP(r1
->rs_start
, r2
->rs_start
));
975 * Create any block allocator specific components. The current allocators
976 * rely on using both a size-ordered range_tree_t and an array of uint64_t's.
979 metaslab_rt_create(range_tree_t
*rt
, void *arg
)
981 metaslab_t
*msp
= arg
;
983 ASSERT3P(rt
->rt_arg
, ==, msp
);
984 ASSERT(msp
->ms_tree
== NULL
);
986 avl_create(&msp
->ms_size_tree
, metaslab_rangesize_compare
,
987 sizeof (range_seg_t
), offsetof(range_seg_t
, rs_pp_node
));
991 * Destroy the block allocator specific components.
994 metaslab_rt_destroy(range_tree_t
*rt
, void *arg
)
996 metaslab_t
*msp
= arg
;
998 ASSERT3P(rt
->rt_arg
, ==, msp
);
999 ASSERT3P(msp
->ms_tree
, ==, rt
);
1000 ASSERT0(avl_numnodes(&msp
->ms_size_tree
));
1002 avl_destroy(&msp
->ms_size_tree
);
1006 metaslab_rt_add(range_tree_t
*rt
, range_seg_t
*rs
, void *arg
)
1008 metaslab_t
*msp
= arg
;
1010 ASSERT3P(rt
->rt_arg
, ==, msp
);
1011 ASSERT3P(msp
->ms_tree
, ==, rt
);
1012 VERIFY(!msp
->ms_condensing
);
1013 avl_add(&msp
->ms_size_tree
, rs
);
1017 metaslab_rt_remove(range_tree_t
*rt
, range_seg_t
*rs
, void *arg
)
1019 metaslab_t
*msp
= arg
;
1021 ASSERT3P(rt
->rt_arg
, ==, msp
);
1022 ASSERT3P(msp
->ms_tree
, ==, rt
);
1023 VERIFY(!msp
->ms_condensing
);
1024 avl_remove(&msp
->ms_size_tree
, rs
);
1028 metaslab_rt_vacate(range_tree_t
*rt
, void *arg
)
1030 metaslab_t
*msp
= arg
;
1032 ASSERT3P(rt
->rt_arg
, ==, msp
);
1033 ASSERT3P(msp
->ms_tree
, ==, rt
);
1036 * Normally one would walk the tree freeing nodes along the way.
1037 * Since the nodes are shared with the range trees we can avoid
1038 * walking all nodes and just reinitialize the avl tree. The nodes
1039 * will be freed by the range tree, so we don't want to free them here.
1041 avl_create(&msp
->ms_size_tree
, metaslab_rangesize_compare
,
1042 sizeof (range_seg_t
), offsetof(range_seg_t
, rs_pp_node
));
1045 static range_tree_ops_t metaslab_rt_ops
= {
1047 metaslab_rt_destroy
,
1054 * ==========================================================================
1055 * Common allocator routines
1056 * ==========================================================================
1060 * Return the maximum contiguous segment within the metaslab.
1063 metaslab_block_maxsize(metaslab_t
*msp
)
1065 avl_tree_t
*t
= &msp
->ms_size_tree
;
1068 if (t
== NULL
|| (rs
= avl_last(t
)) == NULL
)
1071 return (rs
->rs_end
- rs
->rs_start
);
1074 static range_seg_t
*
1075 metaslab_block_find(avl_tree_t
*t
, uint64_t start
, uint64_t size
)
1077 range_seg_t
*rs
, rsearch
;
1080 rsearch
.rs_start
= start
;
1081 rsearch
.rs_end
= start
+ size
;
1083 rs
= avl_find(t
, &rsearch
, &where
);
1085 rs
= avl_nearest(t
, where
, AVL_AFTER
);
1091 #if defined(WITH_FF_BLOCK_ALLOCATOR) || \
1092 defined(WITH_DF_BLOCK_ALLOCATOR) || \
1093 defined(WITH_CF_BLOCK_ALLOCATOR)
1095 * This is a helper function that can be used by the allocator to find
1096 * a suitable block to allocate. This will search the specified AVL
1097 * tree looking for a block that matches the specified criteria.
1100 metaslab_block_picker(avl_tree_t
*t
, uint64_t *cursor
, uint64_t size
,
1103 range_seg_t
*rs
= metaslab_block_find(t
, *cursor
, size
);
1105 while (rs
!= NULL
) {
1106 uint64_t offset
= P2ROUNDUP(rs
->rs_start
, align
);
1108 if (offset
+ size
<= rs
->rs_end
) {
1109 *cursor
= offset
+ size
;
1112 rs
= AVL_NEXT(t
, rs
);
1116 * If we know we've searched the whole map (*cursor == 0), give up.
1117 * Otherwise, reset the cursor to the beginning and try again.
1123 return (metaslab_block_picker(t
, cursor
, size
, align
));
1125 #endif /* WITH_FF/DF/CF_BLOCK_ALLOCATOR */
1127 #if defined(WITH_FF_BLOCK_ALLOCATOR)
1129 * ==========================================================================
1130 * The first-fit block allocator
1131 * ==========================================================================
1134 metaslab_ff_alloc(metaslab_t
*msp
, uint64_t size
)
1137 * Find the largest power of 2 block size that evenly divides the
1138 * requested size. This is used to try to allocate blocks with similar
1139 * alignment from the same area of the metaslab (i.e. same cursor
1140 * bucket) but it does not guarantee that other allocations sizes
1141 * may exist in the same region.
1143 uint64_t align
= size
& -size
;
1144 uint64_t *cursor
= &msp
->ms_lbas
[highbit64(align
) - 1];
1145 avl_tree_t
*t
= &msp
->ms_tree
->rt_root
;
1147 return (metaslab_block_picker(t
, cursor
, size
, align
));
1150 static metaslab_ops_t metaslab_ff_ops
= {
1154 metaslab_ops_t
*zfs_metaslab_ops
= &metaslab_ff_ops
;
1155 #endif /* WITH_FF_BLOCK_ALLOCATOR */
1157 #if defined(WITH_DF_BLOCK_ALLOCATOR)
1159 * ==========================================================================
1160 * Dynamic block allocator -
1161 * Uses the first fit allocation scheme until space get low and then
1162 * adjusts to a best fit allocation method. Uses metaslab_df_alloc_threshold
1163 * and metaslab_df_free_pct to determine when to switch the allocation scheme.
1164 * ==========================================================================
1167 metaslab_df_alloc(metaslab_t
*msp
, uint64_t size
)
1170 * Find the largest power of 2 block size that evenly divides the
1171 * requested size. This is used to try to allocate blocks with similar
1172 * alignment from the same area of the metaslab (i.e. same cursor
1173 * bucket) but it does not guarantee that other allocations sizes
1174 * may exist in the same region.
1176 uint64_t align
= size
& -size
;
1177 uint64_t *cursor
= &msp
->ms_lbas
[highbit64(align
) - 1];
1178 range_tree_t
*rt
= msp
->ms_tree
;
1179 avl_tree_t
*t
= &rt
->rt_root
;
1180 uint64_t max_size
= metaslab_block_maxsize(msp
);
1181 int free_pct
= range_tree_space(rt
) * 100 / msp
->ms_size
;
1183 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1184 ASSERT3U(avl_numnodes(t
), ==, avl_numnodes(&msp
->ms_size_tree
));
1186 if (max_size
< size
)
1190 * If we're running low on space switch to using the size
1191 * sorted AVL tree (best-fit).
1193 if (max_size
< metaslab_df_alloc_threshold
||
1194 free_pct
< metaslab_df_free_pct
) {
1195 t
= &msp
->ms_size_tree
;
1199 return (metaslab_block_picker(t
, cursor
, size
, 1ULL));
1202 static metaslab_ops_t metaslab_df_ops
= {
1206 metaslab_ops_t
*zfs_metaslab_ops
= &metaslab_df_ops
;
1207 #endif /* WITH_DF_BLOCK_ALLOCATOR */
1209 #if defined(WITH_CF_BLOCK_ALLOCATOR)
1211 * ==========================================================================
1212 * Cursor fit block allocator -
1213 * Select the largest region in the metaslab, set the cursor to the beginning
1214 * of the range and the cursor_end to the end of the range. As allocations
1215 * are made advance the cursor. Continue allocating from the cursor until
1216 * the range is exhausted and then find a new range.
1217 * ==========================================================================
1220 metaslab_cf_alloc(metaslab_t
*msp
, uint64_t size
)
1222 range_tree_t
*rt
= msp
->ms_tree
;
1223 avl_tree_t
*t
= &msp
->ms_size_tree
;
1224 uint64_t *cursor
= &msp
->ms_lbas
[0];
1225 uint64_t *cursor_end
= &msp
->ms_lbas
[1];
1226 uint64_t offset
= 0;
1228 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1229 ASSERT3U(avl_numnodes(t
), ==, avl_numnodes(&rt
->rt_root
));
1231 ASSERT3U(*cursor_end
, >=, *cursor
);
1233 if ((*cursor
+ size
) > *cursor_end
) {
1236 rs
= avl_last(&msp
->ms_size_tree
);
1237 if (rs
== NULL
|| (rs
->rs_end
- rs
->rs_start
) < size
)
1240 *cursor
= rs
->rs_start
;
1241 *cursor_end
= rs
->rs_end
;
1250 static metaslab_ops_t metaslab_cf_ops
= {
1254 metaslab_ops_t
*zfs_metaslab_ops
= &metaslab_cf_ops
;
1255 #endif /* WITH_CF_BLOCK_ALLOCATOR */
1257 #if defined(WITH_NDF_BLOCK_ALLOCATOR)
1259 * ==========================================================================
1260 * New dynamic fit allocator -
1261 * Select a region that is large enough to allocate 2^metaslab_ndf_clump_shift
1262 * contiguous blocks. If no region is found then just use the largest segment
1264 * ==========================================================================
1268 * Determines desired number of contiguous blocks (2^metaslab_ndf_clump_shift)
1269 * to request from the allocator.
1271 uint64_t metaslab_ndf_clump_shift
= 4;
1274 metaslab_ndf_alloc(metaslab_t
*msp
, uint64_t size
)
1276 avl_tree_t
*t
= &msp
->ms_tree
->rt_root
;
1278 range_seg_t
*rs
, rsearch
;
1279 uint64_t hbit
= highbit64(size
);
1280 uint64_t *cursor
= &msp
->ms_lbas
[hbit
- 1];
1281 uint64_t max_size
= metaslab_block_maxsize(msp
);
1283 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1284 ASSERT3U(avl_numnodes(t
), ==, avl_numnodes(&msp
->ms_size_tree
));
1286 if (max_size
< size
)
1289 rsearch
.rs_start
= *cursor
;
1290 rsearch
.rs_end
= *cursor
+ size
;
1292 rs
= avl_find(t
, &rsearch
, &where
);
1293 if (rs
== NULL
|| (rs
->rs_end
- rs
->rs_start
) < size
) {
1294 t
= &msp
->ms_size_tree
;
1296 rsearch
.rs_start
= 0;
1297 rsearch
.rs_end
= MIN(max_size
,
1298 1ULL << (hbit
+ metaslab_ndf_clump_shift
));
1299 rs
= avl_find(t
, &rsearch
, &where
);
1301 rs
= avl_nearest(t
, where
, AVL_AFTER
);
1305 if ((rs
->rs_end
- rs
->rs_start
) >= size
) {
1306 *cursor
= rs
->rs_start
+ size
;
1307 return (rs
->rs_start
);
1312 static metaslab_ops_t metaslab_ndf_ops
= {
1316 metaslab_ops_t
*zfs_metaslab_ops
= &metaslab_ndf_ops
;
1317 #endif /* WITH_NDF_BLOCK_ALLOCATOR */
1321 * ==========================================================================
1323 * ==========================================================================
1327 * Wait for any in-progress metaslab loads to complete.
1330 metaslab_load_wait(metaslab_t
*msp
)
1332 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1334 while (msp
->ms_loading
) {
1335 ASSERT(!msp
->ms_loaded
);
1336 cv_wait(&msp
->ms_load_cv
, &msp
->ms_lock
);
1341 metaslab_load(metaslab_t
*msp
)
1344 boolean_t success
= B_FALSE
;
1346 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1347 ASSERT(!msp
->ms_loaded
);
1348 ASSERT(!msp
->ms_loading
);
1350 msp
->ms_loading
= B_TRUE
;
1353 * If the space map has not been allocated yet, then treat
1354 * all the space in the metaslab as free and add it to the
1357 if (msp
->ms_sm
!= NULL
)
1358 error
= space_map_load(msp
->ms_sm
, msp
->ms_tree
, SM_FREE
);
1360 range_tree_add(msp
->ms_tree
, msp
->ms_start
, msp
->ms_size
);
1362 success
= (error
== 0);
1363 msp
->ms_loading
= B_FALSE
;
1366 ASSERT3P(msp
->ms_group
, !=, NULL
);
1367 msp
->ms_loaded
= B_TRUE
;
1369 for (int t
= 0; t
< TXG_DEFER_SIZE
; t
++) {
1370 range_tree_walk(msp
->ms_defertree
[t
],
1371 range_tree_remove
, msp
->ms_tree
);
1373 msp
->ms_max_size
= metaslab_block_maxsize(msp
);
1375 cv_broadcast(&msp
->ms_load_cv
);
1380 metaslab_unload(metaslab_t
*msp
)
1382 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1383 range_tree_vacate(msp
->ms_tree
, NULL
, NULL
);
1384 msp
->ms_loaded
= B_FALSE
;
1385 msp
->ms_weight
&= ~METASLAB_ACTIVE_MASK
;
1386 msp
->ms_max_size
= 0;
1390 metaslab_init(metaslab_group_t
*mg
, uint64_t id
, uint64_t object
, uint64_t txg
,
1393 vdev_t
*vd
= mg
->mg_vd
;
1394 objset_t
*mos
= vd
->vdev_spa
->spa_meta_objset
;
1398 ms
= kmem_zalloc(sizeof (metaslab_t
), KM_SLEEP
);
1399 mutex_init(&ms
->ms_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1400 cv_init(&ms
->ms_load_cv
, NULL
, CV_DEFAULT
, NULL
);
1402 ms
->ms_start
= id
<< vd
->vdev_ms_shift
;
1403 ms
->ms_size
= 1ULL << vd
->vdev_ms_shift
;
1406 * We only open space map objects that already exist. All others
1407 * will be opened when we finally allocate an object for it.
1410 error
= space_map_open(&ms
->ms_sm
, mos
, object
, ms
->ms_start
,
1411 ms
->ms_size
, vd
->vdev_ashift
, &ms
->ms_lock
);
1414 kmem_free(ms
, sizeof (metaslab_t
));
1418 ASSERT(ms
->ms_sm
!= NULL
);
1422 * We create the main range tree here, but we don't create the
1423 * other range trees until metaslab_sync_done(). This serves
1424 * two purposes: it allows metaslab_sync_done() to detect the
1425 * addition of new space; and for debugging, it ensures that we'd
1426 * data fault on any attempt to use this metaslab before it's ready.
1428 ms
->ms_tree
= range_tree_create(&metaslab_rt_ops
, ms
, &ms
->ms_lock
);
1429 metaslab_group_add(mg
, ms
);
1431 metaslab_set_fragmentation(ms
);
1434 * If we're opening an existing pool (txg == 0) or creating
1435 * a new one (txg == TXG_INITIAL), all space is available now.
1436 * If we're adding space to an existing pool, the new space
1437 * does not become available until after this txg has synced.
1438 * The metaslab's weight will also be initialized when we sync
1439 * out this txg. This ensures that we don't attempt to allocate
1440 * from it before we have initialized it completely.
1442 if (txg
<= TXG_INITIAL
)
1443 metaslab_sync_done(ms
, 0);
1446 * If metaslab_debug_load is set and we're initializing a metaslab
1447 * that has an allocated space map object then load the its space
1448 * map so that can verify frees.
1450 if (metaslab_debug_load
&& ms
->ms_sm
!= NULL
) {
1451 mutex_enter(&ms
->ms_lock
);
1452 VERIFY0(metaslab_load(ms
));
1453 mutex_exit(&ms
->ms_lock
);
1457 vdev_dirty(vd
, 0, NULL
, txg
);
1458 vdev_dirty(vd
, VDD_METASLAB
, ms
, txg
);
1467 metaslab_fini(metaslab_t
*msp
)
1469 metaslab_group_t
*mg
= msp
->ms_group
;
1471 metaslab_group_remove(mg
, msp
);
1473 mutex_enter(&msp
->ms_lock
);
1474 VERIFY(msp
->ms_group
== NULL
);
1475 vdev_space_update(mg
->mg_vd
, -space_map_allocated(msp
->ms_sm
),
1477 space_map_close(msp
->ms_sm
);
1479 metaslab_unload(msp
);
1480 range_tree_destroy(msp
->ms_tree
);
1481 range_tree_destroy(msp
->ms_freeingtree
);
1482 range_tree_destroy(msp
->ms_freedtree
);
1484 for (int t
= 0; t
< TXG_SIZE
; t
++) {
1485 range_tree_destroy(msp
->ms_alloctree
[t
]);
1488 for (int t
= 0; t
< TXG_DEFER_SIZE
; t
++) {
1489 range_tree_destroy(msp
->ms_defertree
[t
]);
1492 ASSERT0(msp
->ms_deferspace
);
1494 mutex_exit(&msp
->ms_lock
);
1495 cv_destroy(&msp
->ms_load_cv
);
1496 mutex_destroy(&msp
->ms_lock
);
1498 kmem_free(msp
, sizeof (metaslab_t
));
1501 #define FRAGMENTATION_TABLE_SIZE 17
1504 * This table defines a segment size based fragmentation metric that will
1505 * allow each metaslab to derive its own fragmentation value. This is done
1506 * by calculating the space in each bucket of the spacemap histogram and
1507 * multiplying that by the fragmetation metric in this table. Doing
1508 * this for all buckets and dividing it by the total amount of free
1509 * space in this metaslab (i.e. the total free space in all buckets) gives
1510 * us the fragmentation metric. This means that a high fragmentation metric
1511 * equates to most of the free space being comprised of small segments.
1512 * Conversely, if the metric is low, then most of the free space is in
1513 * large segments. A 10% change in fragmentation equates to approximately
1514 * double the number of segments.
1516 * This table defines 0% fragmented space using 16MB segments. Testing has
1517 * shown that segments that are greater than or equal to 16MB do not suffer
1518 * from drastic performance problems. Using this value, we derive the rest
1519 * of the table. Since the fragmentation value is never stored on disk, it
1520 * is possible to change these calculations in the future.
1522 int zfs_frag_table
[FRAGMENTATION_TABLE_SIZE
] = {
1542 * Calclate the metaslab's fragmentation metric. A return value
1543 * of ZFS_FRAG_INVALID means that the metaslab has not been upgraded and does
1544 * not support this metric. Otherwise, the return value should be in the
1548 metaslab_set_fragmentation(metaslab_t
*msp
)
1550 spa_t
*spa
= msp
->ms_group
->mg_vd
->vdev_spa
;
1551 uint64_t fragmentation
= 0;
1553 boolean_t feature_enabled
= spa_feature_is_enabled(spa
,
1554 SPA_FEATURE_SPACEMAP_HISTOGRAM
);
1556 if (!feature_enabled
) {
1557 msp
->ms_fragmentation
= ZFS_FRAG_INVALID
;
1562 * A null space map means that the entire metaslab is free
1563 * and thus is not fragmented.
1565 if (msp
->ms_sm
== NULL
) {
1566 msp
->ms_fragmentation
= 0;
1571 * If this metaslab's space map has not been upgraded, flag it
1572 * so that we upgrade next time we encounter it.
1574 if (msp
->ms_sm
->sm_dbuf
->db_size
!= sizeof (space_map_phys_t
)) {
1575 uint64_t txg
= spa_syncing_txg(spa
);
1576 vdev_t
*vd
= msp
->ms_group
->mg_vd
;
1579 * If we've reached the final dirty txg, then we must
1580 * be shutting down the pool. We don't want to dirty
1581 * any data past this point so skip setting the condense
1582 * flag. We can retry this action the next time the pool
1585 if (spa_writeable(spa
) && txg
< spa_final_dirty_txg(spa
)) {
1586 msp
->ms_condense_wanted
= B_TRUE
;
1587 vdev_dirty(vd
, VDD_METASLAB
, msp
, txg
+ 1);
1588 spa_dbgmsg(spa
, "txg %llu, requesting force condense: "
1589 "ms_id %llu, vdev_id %llu", txg
, msp
->ms_id
,
1592 msp
->ms_fragmentation
= ZFS_FRAG_INVALID
;
1596 for (int i
= 0; i
< SPACE_MAP_HISTOGRAM_SIZE
; i
++) {
1598 uint8_t shift
= msp
->ms_sm
->sm_shift
;
1600 int idx
= MIN(shift
- SPA_MINBLOCKSHIFT
+ i
,
1601 FRAGMENTATION_TABLE_SIZE
- 1);
1603 if (msp
->ms_sm
->sm_phys
->smp_histogram
[i
] == 0)
1606 space
= msp
->ms_sm
->sm_phys
->smp_histogram
[i
] << (i
+ shift
);
1609 ASSERT3U(idx
, <, FRAGMENTATION_TABLE_SIZE
);
1610 fragmentation
+= space
* zfs_frag_table
[idx
];
1614 fragmentation
/= total
;
1615 ASSERT3U(fragmentation
, <=, 100);
1617 msp
->ms_fragmentation
= fragmentation
;
1621 * Compute a weight -- a selection preference value -- for the given metaslab.
1622 * This is based on the amount of free space, the level of fragmentation,
1623 * the LBA range, and whether the metaslab is loaded.
1626 metaslab_space_weight(metaslab_t
*msp
)
1628 metaslab_group_t
*mg
= msp
->ms_group
;
1629 vdev_t
*vd
= mg
->mg_vd
;
1630 uint64_t weight
, space
;
1632 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1633 ASSERT(!vd
->vdev_removing
);
1636 * The baseline weight is the metaslab's free space.
1638 space
= msp
->ms_size
- space_map_allocated(msp
->ms_sm
);
1640 if (metaslab_fragmentation_factor_enabled
&&
1641 msp
->ms_fragmentation
!= ZFS_FRAG_INVALID
) {
1643 * Use the fragmentation information to inversely scale
1644 * down the baseline weight. We need to ensure that we
1645 * don't exclude this metaslab completely when it's 100%
1646 * fragmented. To avoid this we reduce the fragmented value
1649 space
= (space
* (100 - (msp
->ms_fragmentation
- 1))) / 100;
1652 * If space < SPA_MINBLOCKSIZE, then we will not allocate from
1653 * this metaslab again. The fragmentation metric may have
1654 * decreased the space to something smaller than
1655 * SPA_MINBLOCKSIZE, so reset the space to SPA_MINBLOCKSIZE
1656 * so that we can consume any remaining space.
1658 if (space
> 0 && space
< SPA_MINBLOCKSIZE
)
1659 space
= SPA_MINBLOCKSIZE
;
1664 * Modern disks have uniform bit density and constant angular velocity.
1665 * Therefore, the outer recording zones are faster (higher bandwidth)
1666 * than the inner zones by the ratio of outer to inner track diameter,
1667 * which is typically around 2:1. We account for this by assigning
1668 * higher weight to lower metaslabs (multiplier ranging from 2x to 1x).
1669 * In effect, this means that we'll select the metaslab with the most
1670 * free bandwidth rather than simply the one with the most free space.
1672 if (!vd
->vdev_nonrot
&& metaslab_lba_weighting_enabled
) {
1673 weight
= 2 * weight
- (msp
->ms_id
* weight
) / vd
->vdev_ms_count
;
1674 ASSERT(weight
>= space
&& weight
<= 2 * space
);
1678 * If this metaslab is one we're actively using, adjust its
1679 * weight to make it preferable to any inactive metaslab so
1680 * we'll polish it off. If the fragmentation on this metaslab
1681 * has exceed our threshold, then don't mark it active.
1683 if (msp
->ms_loaded
&& msp
->ms_fragmentation
!= ZFS_FRAG_INVALID
&&
1684 msp
->ms_fragmentation
<= zfs_metaslab_fragmentation_threshold
) {
1685 weight
|= (msp
->ms_weight
& METASLAB_ACTIVE_MASK
);
1688 WEIGHT_SET_SPACEBASED(weight
);
1693 * Return the weight of the specified metaslab, according to the segment-based
1694 * weighting algorithm. The metaslab must be loaded. This function can
1695 * be called within a sync pass since it relies only on the metaslab's
1696 * range tree which is always accurate when the metaslab is loaded.
1699 metaslab_weight_from_range_tree(metaslab_t
*msp
)
1701 uint64_t weight
= 0;
1702 uint32_t segments
= 0;
1704 ASSERT(msp
->ms_loaded
);
1706 for (int i
= RANGE_TREE_HISTOGRAM_SIZE
- 1; i
>= SPA_MINBLOCKSHIFT
;
1708 uint8_t shift
= msp
->ms_group
->mg_vd
->vdev_ashift
;
1709 int max_idx
= SPACE_MAP_HISTOGRAM_SIZE
+ shift
- 1;
1712 segments
+= msp
->ms_tree
->rt_histogram
[i
];
1715 * The range tree provides more precision than the space map
1716 * and must be downgraded so that all values fit within the
1717 * space map's histogram. This allows us to compare loaded
1718 * vs. unloaded metaslabs to determine which metaslab is
1719 * considered "best".
1724 if (segments
!= 0) {
1725 WEIGHT_SET_COUNT(weight
, segments
);
1726 WEIGHT_SET_INDEX(weight
, i
);
1727 WEIGHT_SET_ACTIVE(weight
, 0);
1735 * Calculate the weight based on the on-disk histogram. This should only
1736 * be called after a sync pass has completely finished since the on-disk
1737 * information is updated in metaslab_sync().
1740 metaslab_weight_from_spacemap(metaslab_t
*msp
)
1742 uint64_t weight
= 0;
1744 for (int i
= SPACE_MAP_HISTOGRAM_SIZE
- 1; i
>= 0; i
--) {
1745 if (msp
->ms_sm
->sm_phys
->smp_histogram
[i
] != 0) {
1746 WEIGHT_SET_COUNT(weight
,
1747 msp
->ms_sm
->sm_phys
->smp_histogram
[i
]);
1748 WEIGHT_SET_INDEX(weight
, i
+
1749 msp
->ms_sm
->sm_shift
);
1750 WEIGHT_SET_ACTIVE(weight
, 0);
1758 * Compute a segment-based weight for the specified metaslab. The weight
1759 * is determined by highest bucket in the histogram. The information
1760 * for the highest bucket is encoded into the weight value.
1763 metaslab_segment_weight(metaslab_t
*msp
)
1765 metaslab_group_t
*mg
= msp
->ms_group
;
1766 uint64_t weight
= 0;
1767 uint8_t shift
= mg
->mg_vd
->vdev_ashift
;
1769 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1772 * The metaslab is completely free.
1774 if (space_map_allocated(msp
->ms_sm
) == 0) {
1775 int idx
= highbit64(msp
->ms_size
) - 1;
1776 int max_idx
= SPACE_MAP_HISTOGRAM_SIZE
+ shift
- 1;
1778 if (idx
< max_idx
) {
1779 WEIGHT_SET_COUNT(weight
, 1ULL);
1780 WEIGHT_SET_INDEX(weight
, idx
);
1782 WEIGHT_SET_COUNT(weight
, 1ULL << (idx
- max_idx
));
1783 WEIGHT_SET_INDEX(weight
, max_idx
);
1785 WEIGHT_SET_ACTIVE(weight
, 0);
1786 ASSERT(!WEIGHT_IS_SPACEBASED(weight
));
1791 ASSERT3U(msp
->ms_sm
->sm_dbuf
->db_size
, ==, sizeof (space_map_phys_t
));
1794 * If the metaslab is fully allocated then just make the weight 0.
1796 if (space_map_allocated(msp
->ms_sm
) == msp
->ms_size
)
1799 * If the metaslab is already loaded, then use the range tree to
1800 * determine the weight. Otherwise, we rely on the space map information
1801 * to generate the weight.
1803 if (msp
->ms_loaded
) {
1804 weight
= metaslab_weight_from_range_tree(msp
);
1806 weight
= metaslab_weight_from_spacemap(msp
);
1810 * If the metaslab was active the last time we calculated its weight
1811 * then keep it active. We want to consume the entire region that
1812 * is associated with this weight.
1814 if (msp
->ms_activation_weight
!= 0 && weight
!= 0)
1815 WEIGHT_SET_ACTIVE(weight
, WEIGHT_GET_ACTIVE(msp
->ms_weight
));
1820 * Determine if we should attempt to allocate from this metaslab. If the
1821 * metaslab has a maximum size then we can quickly determine if the desired
1822 * allocation size can be satisfied. Otherwise, if we're using segment-based
1823 * weighting then we can determine the maximum allocation that this metaslab
1824 * can accommodate based on the index encoded in the weight. If we're using
1825 * space-based weights then rely on the entire weight (excluding the weight
1829 metaslab_should_allocate(metaslab_t
*msp
, uint64_t asize
)
1831 boolean_t should_allocate
;
1833 if (msp
->ms_max_size
!= 0)
1834 return (msp
->ms_max_size
>= asize
);
1836 if (!WEIGHT_IS_SPACEBASED(msp
->ms_weight
)) {
1838 * The metaslab segment weight indicates segments in the
1839 * range [2^i, 2^(i+1)), where i is the index in the weight.
1840 * Since the asize might be in the middle of the range, we
1841 * should attempt the allocation if asize < 2^(i+1).
1843 should_allocate
= (asize
<
1844 1ULL << (WEIGHT_GET_INDEX(msp
->ms_weight
) + 1));
1846 should_allocate
= (asize
<=
1847 (msp
->ms_weight
& ~METASLAB_WEIGHT_TYPE
));
1849 return (should_allocate
);
1852 metaslab_weight(metaslab_t
*msp
)
1854 vdev_t
*vd
= msp
->ms_group
->mg_vd
;
1855 spa_t
*spa
= vd
->vdev_spa
;
1858 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1861 * This vdev is in the process of being removed so there is nothing
1862 * for us to do here.
1864 if (vd
->vdev_removing
) {
1865 ASSERT0(space_map_allocated(msp
->ms_sm
));
1866 ASSERT0(vd
->vdev_ms_shift
);
1870 metaslab_set_fragmentation(msp
);
1873 * Update the maximum size if the metaslab is loaded. This will
1874 * ensure that we get an accurate maximum size if newly freed space
1875 * has been added back into the free tree.
1878 msp
->ms_max_size
= metaslab_block_maxsize(msp
);
1881 * Segment-based weighting requires space map histogram support.
1883 if (zfs_metaslab_segment_weight_enabled
&&
1884 spa_feature_is_enabled(spa
, SPA_FEATURE_SPACEMAP_HISTOGRAM
) &&
1885 (msp
->ms_sm
== NULL
|| msp
->ms_sm
->sm_dbuf
->db_size
==
1886 sizeof (space_map_phys_t
))) {
1887 weight
= metaslab_segment_weight(msp
);
1889 weight
= metaslab_space_weight(msp
);
1895 metaslab_activate(metaslab_t
*msp
, uint64_t activation_weight
)
1897 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
1899 if ((msp
->ms_weight
& METASLAB_ACTIVE_MASK
) == 0) {
1900 metaslab_load_wait(msp
);
1901 if (!msp
->ms_loaded
) {
1902 int error
= metaslab_load(msp
);
1904 metaslab_group_sort(msp
->ms_group
, msp
, 0);
1909 msp
->ms_activation_weight
= msp
->ms_weight
;
1910 metaslab_group_sort(msp
->ms_group
, msp
,
1911 msp
->ms_weight
| activation_weight
);
1913 ASSERT(msp
->ms_loaded
);
1914 ASSERT(msp
->ms_weight
& METASLAB_ACTIVE_MASK
);
1920 metaslab_passivate(metaslab_t
*msp
, uint64_t weight
)
1922 ASSERTV(uint64_t size
= weight
& ~METASLAB_WEIGHT_TYPE
);
1925 * If size < SPA_MINBLOCKSIZE, then we will not allocate from
1926 * this metaslab again. In that case, it had better be empty,
1927 * or we would be leaving space on the table.
1929 ASSERT(!WEIGHT_IS_SPACEBASED(msp
->ms_weight
) ||
1930 size
>= SPA_MINBLOCKSIZE
||
1931 range_tree_space(msp
->ms_tree
) == 0);
1932 ASSERT0(weight
& METASLAB_ACTIVE_MASK
);
1934 msp
->ms_activation_weight
= 0;
1935 metaslab_group_sort(msp
->ms_group
, msp
, weight
);
1936 ASSERT((msp
->ms_weight
& METASLAB_ACTIVE_MASK
) == 0);
1940 * Segment-based metaslabs are activated once and remain active until
1941 * we either fail an allocation attempt (similar to space-based metaslabs)
1942 * or have exhausted the free space in zfs_metaslab_switch_threshold
1943 * buckets since the metaslab was activated. This function checks to see
1944 * if we've exhaused the zfs_metaslab_switch_threshold buckets in the
1945 * metaslab and passivates it proactively. This will allow us to select a
1946 * metaslab with a larger contiguous region, if any, remaining within this
1947 * metaslab group. If we're in sync pass > 1, then we continue using this
1948 * metaslab so that we don't dirty more block and cause more sync passes.
1951 metaslab_segment_may_passivate(metaslab_t
*msp
)
1953 spa_t
*spa
= msp
->ms_group
->mg_vd
->vdev_spa
;
1955 if (WEIGHT_IS_SPACEBASED(msp
->ms_weight
) || spa_sync_pass(spa
) > 1)
1959 * Since we are in the middle of a sync pass, the most accurate
1960 * information that is accessible to us is the in-core range tree
1961 * histogram; calculate the new weight based on that information.
1963 uint64_t weight
= metaslab_weight_from_range_tree(msp
);
1964 int activation_idx
= WEIGHT_GET_INDEX(msp
->ms_activation_weight
);
1965 int current_idx
= WEIGHT_GET_INDEX(weight
);
1967 if (current_idx
<= activation_idx
- zfs_metaslab_switch_threshold
)
1968 metaslab_passivate(msp
, weight
);
1972 metaslab_preload(void *arg
)
1974 metaslab_t
*msp
= arg
;
1975 spa_t
*spa
= msp
->ms_group
->mg_vd
->vdev_spa
;
1976 fstrans_cookie_t cookie
= spl_fstrans_mark();
1978 ASSERT(!MUTEX_HELD(&msp
->ms_group
->mg_lock
));
1980 mutex_enter(&msp
->ms_lock
);
1981 metaslab_load_wait(msp
);
1982 if (!msp
->ms_loaded
)
1983 (void) metaslab_load(msp
);
1984 msp
->ms_selected_txg
= spa_syncing_txg(spa
);
1985 mutex_exit(&msp
->ms_lock
);
1986 spl_fstrans_unmark(cookie
);
1990 metaslab_group_preload(metaslab_group_t
*mg
)
1992 spa_t
*spa
= mg
->mg_vd
->vdev_spa
;
1994 avl_tree_t
*t
= &mg
->mg_metaslab_tree
;
1997 if (spa_shutting_down(spa
) || !metaslab_preload_enabled
) {
1998 taskq_wait_outstanding(mg
->mg_taskq
, 0);
2002 mutex_enter(&mg
->mg_lock
);
2004 * Load the next potential metaslabs
2006 for (msp
= avl_first(t
); msp
!= NULL
; msp
= AVL_NEXT(t
, msp
)) {
2008 * We preload only the maximum number of metaslabs specified
2009 * by metaslab_preload_limit. If a metaslab is being forced
2010 * to condense then we preload it too. This will ensure
2011 * that force condensing happens in the next txg.
2013 if (++m
> metaslab_preload_limit
&& !msp
->ms_condense_wanted
) {
2017 VERIFY(taskq_dispatch(mg
->mg_taskq
, metaslab_preload
,
2018 msp
, TQ_SLEEP
) != TASKQID_INVALID
);
2020 mutex_exit(&mg
->mg_lock
);
2024 * Determine if the space map's on-disk footprint is past our tolerance
2025 * for inefficiency. We would like to use the following criteria to make
2028 * 1. The size of the space map object should not dramatically increase as a
2029 * result of writing out the free space range tree.
2031 * 2. The minimal on-disk space map representation is zfs_condense_pct/100
2032 * times the size than the free space range tree representation
2033 * (i.e. zfs_condense_pct = 110 and in-core = 1MB, minimal = 1.1.MB).
2035 * 3. The on-disk size of the space map should actually decrease.
2037 * Checking the first condition is tricky since we don't want to walk
2038 * the entire AVL tree calculating the estimated on-disk size. Instead we
2039 * use the size-ordered range tree in the metaslab and calculate the
2040 * size required to write out the largest segment in our free tree. If the
2041 * size required to represent that segment on disk is larger than the space
2042 * map object then we avoid condensing this map.
2044 * To determine the second criterion we use a best-case estimate and assume
2045 * each segment can be represented on-disk as a single 64-bit entry. We refer
2046 * to this best-case estimate as the space map's minimal form.
2048 * Unfortunately, we cannot compute the on-disk size of the space map in this
2049 * context because we cannot accurately compute the effects of compression, etc.
2050 * Instead, we apply the heuristic described in the block comment for
2051 * zfs_metaslab_condense_block_threshold - we only condense if the space used
2052 * is greater than a threshold number of blocks.
2055 metaslab_should_condense(metaslab_t
*msp
)
2057 space_map_t
*sm
= msp
->ms_sm
;
2059 uint64_t size
, entries
, segsz
, object_size
, optimal_size
, record_size
;
2060 dmu_object_info_t doi
;
2061 uint64_t vdev_blocksize
= 1ULL << msp
->ms_group
->mg_vd
->vdev_ashift
;
2063 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
2064 ASSERT(msp
->ms_loaded
);
2067 * Use the ms_size_tree range tree, which is ordered by size, to
2068 * obtain the largest segment in the free tree. We always condense
2069 * metaslabs that are empty and metaslabs for which a condense
2070 * request has been made.
2072 rs
= avl_last(&msp
->ms_size_tree
);
2073 if (rs
== NULL
|| msp
->ms_condense_wanted
)
2077 * Calculate the number of 64-bit entries this segment would
2078 * require when written to disk. If this single segment would be
2079 * larger on-disk than the entire current on-disk structure, then
2080 * clearly condensing will increase the on-disk structure size.
2082 size
= (rs
->rs_end
- rs
->rs_start
) >> sm
->sm_shift
;
2083 entries
= size
/ (MIN(size
, SM_RUN_MAX
));
2084 segsz
= entries
* sizeof (uint64_t);
2086 optimal_size
= sizeof (uint64_t) * avl_numnodes(&msp
->ms_tree
->rt_root
);
2087 object_size
= space_map_length(msp
->ms_sm
);
2089 dmu_object_info_from_db(sm
->sm_dbuf
, &doi
);
2090 record_size
= MAX(doi
.doi_data_block_size
, vdev_blocksize
);
2092 return (segsz
<= object_size
&&
2093 object_size
>= (optimal_size
* zfs_condense_pct
/ 100) &&
2094 object_size
> zfs_metaslab_condense_block_threshold
* record_size
);
2098 * Condense the on-disk space map representation to its minimized form.
2099 * The minimized form consists of a small number of allocations followed by
2100 * the entries of the free range tree.
2103 metaslab_condense(metaslab_t
*msp
, uint64_t txg
, dmu_tx_t
*tx
)
2105 spa_t
*spa
= msp
->ms_group
->mg_vd
->vdev_spa
;
2106 range_tree_t
*condense_tree
;
2107 space_map_t
*sm
= msp
->ms_sm
;
2109 ASSERT(MUTEX_HELD(&msp
->ms_lock
));
2110 ASSERT3U(spa_sync_pass(spa
), ==, 1);
2111 ASSERT(msp
->ms_loaded
);
2114 spa_dbgmsg(spa
, "condensing: txg %llu, msp[%llu] %p, vdev id %llu, "
2115 "spa %s, smp size %llu, segments %lu, forcing condense=%s", txg
,
2116 msp
->ms_id
, msp
, msp
->ms_group
->mg_vd
->vdev_id
,
2117 msp
->ms_group
->mg_vd
->vdev_spa
->spa_name
,
2118 space_map_length(msp
->ms_sm
), avl_numnodes(&msp
->ms_tree
->rt_root
),
2119 msp
->ms_condense_wanted
? "TRUE" : "FALSE");
2121 msp
->ms_condense_wanted
= B_FALSE
;
2124 * Create an range tree that is 100% allocated. We remove segments
2125 * that have been freed in this txg, any deferred frees that exist,
2126 * and any allocation in the future. Removing segments should be
2127 * a relatively inexpensive operation since we expect these trees to
2128 * have a small number of nodes.
2130 condense_tree
= range_tree_create(NULL
, NULL
, &msp
->ms_lock
);
2131 range_tree_add(condense_tree
, msp
->ms_start
, msp
->ms_size
);
2134 * Remove what's been freed in this txg from the condense_tree.
2135 * Since we're in sync_pass 1, we know that all the frees from
2136 * this txg are in the freeingtree.
2138 range_tree_walk(msp
->ms_freeingtree
, range_tree_remove
, condense_tree
);
2140 for (int t
= 0; t
< TXG_DEFER_SIZE
; t
++) {
2141 range_tree_walk(msp
->ms_defertree
[t
],
2142 range_tree_remove
, condense_tree
);
2145 for (int t
= 1; t
< TXG_CONCURRENT_STATES
; t
++) {
2146 range_tree_walk(msp
->ms_alloctree
[(txg
+ t
) & TXG_MASK
],
2147 range_tree_remove
, condense_tree
);
2151 * We're about to drop the metaslab's lock thus allowing
2152 * other consumers to change it's content. Set the
2153 * metaslab's ms_condensing flag to ensure that
2154 * allocations on this metaslab do not occur while we're
2155 * in the middle of committing it to disk. This is only critical
2156 * for the ms_tree as all other range trees use per txg
2157 * views of their content.
2159 msp
->ms_condensing
= B_TRUE
;
2161 mutex_exit(&msp
->ms_lock
);
2162 space_map_truncate(sm
, tx
);
2163 mutex_enter(&msp
->ms_lock
);
2166 * While we would ideally like to create a space map representation
2167 * that consists only of allocation records, doing so can be
2168 * prohibitively expensive because the in-core free tree can be
2169 * large, and therefore computationally expensive to subtract
2170 * from the condense_tree. Instead we sync out two trees, a cheap
2171 * allocation only tree followed by the in-core free tree. While not
2172 * optimal, this is typically close to optimal, and much cheaper to
2175 space_map_write(sm
, condense_tree
, SM_ALLOC
, tx
);
2176 range_tree_vacate(condense_tree
, NULL
, NULL
);
2177 range_tree_destroy(condense_tree
);
2179 space_map_write(sm
, msp
->ms_tree
, SM_FREE
, tx
);
2180 msp
->ms_condensing
= B_FALSE
;
2184 * Write a metaslab to disk in the context of the specified transaction group.
2187 metaslab_sync(metaslab_t
*msp
, uint64_t txg
)
2189 metaslab_group_t
*mg
= msp
->ms_group
;
2190 vdev_t
*vd
= mg
->mg_vd
;
2191 spa_t
*spa
= vd
->vdev_spa
;
2192 objset_t
*mos
= spa_meta_objset(spa
);
2193 range_tree_t
*alloctree
= msp
->ms_alloctree
[txg
& TXG_MASK
];
2195 uint64_t object
= space_map_object(msp
->ms_sm
);
2197 ASSERT(!vd
->vdev_ishole
);
2200 * This metaslab has just been added so there's no work to do now.
2202 if (msp
->ms_freeingtree
== NULL
) {
2203 ASSERT3P(alloctree
, ==, NULL
);
2207 ASSERT3P(alloctree
, !=, NULL
);
2208 ASSERT3P(msp
->ms_freeingtree
, !=, NULL
);
2209 ASSERT3P(msp
->ms_freedtree
, !=, NULL
);
2212 * Normally, we don't want to process a metaslab if there
2213 * are no allocations or frees to perform. However, if the metaslab
2214 * is being forced to condense and it's loaded, we need to let it
2217 if (range_tree_space(alloctree
) == 0 &&
2218 range_tree_space(msp
->ms_freeingtree
) == 0 &&
2219 !(msp
->ms_loaded
&& msp
->ms_condense_wanted
))
2223 VERIFY(txg
<= spa_final_dirty_txg(spa
));
2226 * The only state that can actually be changing concurrently with
2227 * metaslab_sync() is the metaslab's ms_tree. No other thread can
2228 * be modifying this txg's alloctree, freeingtree, freedtree, or
2229 * space_map_phys_t. Therefore, we only hold ms_lock to satify
2230 * space map ASSERTs. We drop it whenever we call into the DMU,
2231 * because the DMU can call down to us (e.g. via zio_free()) at
2235 tx
= dmu_tx_create_assigned(spa_get_dsl(spa
), txg
);
2237 if (msp
->ms_sm
== NULL
) {
2238 uint64_t new_object
;
2240 new_object
= space_map_alloc(mos
, tx
);
2241 VERIFY3U(new_object
, !=, 0);
2243 VERIFY0(space_map_open(&msp
->ms_sm
, mos
, new_object
,
2244 msp
->ms_start
, msp
->ms_size
, vd
->vdev_ashift
,
2246 ASSERT(msp
->ms_sm
!= NULL
);
2249 mutex_enter(&msp
->ms_lock
);
2252 * Note: metaslab_condense() clears the space map's histogram.
2253 * Therefore we must verify and remove this histogram before
2256 metaslab_group_histogram_verify(mg
);
2257 metaslab_class_histogram_verify(mg
->mg_class
);
2258 metaslab_group_histogram_remove(mg
, msp
);
2260 if (msp
->ms_loaded
&& spa_sync_pass(spa
) == 1 &&
2261 metaslab_should_condense(msp
)) {
2262 metaslab_condense(msp
, txg
, tx
);
2264 space_map_write(msp
->ms_sm
, alloctree
, SM_ALLOC
, tx
);
2265 space_map_write(msp
->ms_sm
, msp
->ms_freeingtree
, SM_FREE
, tx
);
2268 if (msp
->ms_loaded
) {
2270 * When the space map is loaded, we have an accruate
2271 * histogram in the range tree. This gives us an opportunity
2272 * to bring the space map's histogram up-to-date so we clear
2273 * it first before updating it.
2275 space_map_histogram_clear(msp
->ms_sm
);
2276 space_map_histogram_add(msp
->ms_sm
, msp
->ms_tree
, tx
);
2279 * Since we've cleared the histogram we need to add back
2280 * any free space that has already been processed, plus
2281 * any deferred space. This allows the on-disk histogram
2282 * to accurately reflect all free space even if some space
2283 * is not yet available for allocation (i.e. deferred).
2285 space_map_histogram_add(msp
->ms_sm
, msp
->ms_freedtree
, tx
);
2288 * Add back any deferred free space that has not been
2289 * added back into the in-core free tree yet. This will
2290 * ensure that we don't end up with a space map histogram
2291 * that is completely empty unless the metaslab is fully
2294 for (int t
= 0; t
< TXG_DEFER_SIZE
; t
++) {
2295 space_map_histogram_add(msp
->ms_sm
,
2296 msp
->ms_defertree
[t
], tx
);
2301 * Always add the free space from this sync pass to the space
2302 * map histogram. We want to make sure that the on-disk histogram
2303 * accounts for all free space. If the space map is not loaded,
2304 * then we will lose some accuracy but will correct it the next
2305 * time we load the space map.
2307 space_map_histogram_add(msp
->ms_sm
, msp
->ms_freeingtree
, tx
);
2309 metaslab_group_histogram_add(mg
, msp
);
2310 metaslab_group_histogram_verify(mg
);
2311 metaslab_class_histogram_verify(mg
->mg_class
);
2314 * For sync pass 1, we avoid traversing this txg's free range tree
2315 * and instead will just swap the pointers for freeingtree and
2316 * freedtree. We can safely do this since the freed_tree is
2317 * guaranteed to be empty on the initial pass.
2319 if (spa_sync_pass(spa
) == 1) {
2320 range_tree_swap(&msp
->ms_freeingtree
, &msp
->ms_freedtree
);
2322 range_tree_vacate(msp
->ms_freeingtree
,
2323 range_tree_add
, msp
->ms_freedtree
);
2325 range_tree_vacate(alloctree
, NULL
, NULL
);
2327 ASSERT0(range_tree_space(msp
->ms_alloctree
[txg
& TXG_MASK
]));
2328 ASSERT0(range_tree_space(msp
->ms_alloctree
[TXG_CLEAN(txg
) & TXG_MASK
]));
2329 ASSERT0(range_tree_space(msp
->ms_freeingtree
));
2331 mutex_exit(&msp
->ms_lock
);
2333 if (object
!= space_map_object(msp
->ms_sm
)) {
2334 object
= space_map_object(msp
->ms_sm
);
2335 dmu_write(mos
, vd
->vdev_ms_array
, sizeof (uint64_t) *
2336 msp
->ms_id
, sizeof (uint64_t), &object
, tx
);
2342 * Called after a transaction group has completely synced to mark
2343 * all of the metaslab's free space as usable.
2346 metaslab_sync_done(metaslab_t
*msp
, uint64_t txg
)
2348 metaslab_group_t
*mg
= msp
->ms_group
;
2349 vdev_t
*vd
= mg
->mg_vd
;
2350 spa_t
*spa
= vd
->vdev_spa
;
2351 range_tree_t
**defer_tree
;
2352 int64_t alloc_delta
, defer_delta
;
2353 boolean_t defer_allowed
= B_TRUE
;
2355 ASSERT(!vd
->vdev_ishole
);
2357 mutex_enter(&msp
->ms_lock
);
2360 * If this metaslab is just becoming available, initialize its
2361 * range trees and add its capacity to the vdev.
2363 if (msp
->ms_freedtree
== NULL
) {
2364 for (int t
= 0; t
< TXG_SIZE
; t
++) {
2365 ASSERT(msp
->ms_alloctree
[t
] == NULL
);
2367 msp
->ms_alloctree
[t
] = range_tree_create(NULL
, msp
,
2371 ASSERT3P(msp
->ms_freeingtree
, ==, NULL
);
2372 msp
->ms_freeingtree
= range_tree_create(NULL
, msp
,
2375 ASSERT3P(msp
->ms_freedtree
, ==, NULL
);
2376 msp
->ms_freedtree
= range_tree_create(NULL
, msp
,
2379 for (int t
= 0; t
< TXG_DEFER_SIZE
; t
++) {
2380 ASSERT(msp
->ms_defertree
[t
] == NULL
);
2382 msp
->ms_defertree
[t
] = range_tree_create(NULL
, msp
,
2386 vdev_space_update(vd
, 0, 0, msp
->ms_size
);
2389 defer_tree
= &msp
->ms_defertree
[txg
% TXG_DEFER_SIZE
];
2391 uint64_t free_space
= metaslab_class_get_space(spa_normal_class(spa
)) -
2392 metaslab_class_get_alloc(spa_normal_class(spa
));
2393 if (free_space
<= spa_get_slop_space(spa
)) {
2394 defer_allowed
= B_FALSE
;
2398 alloc_delta
= space_map_alloc_delta(msp
->ms_sm
);
2399 if (defer_allowed
) {
2400 defer_delta
= range_tree_space(msp
->ms_freedtree
) -
2401 range_tree_space(*defer_tree
);
2403 defer_delta
-= range_tree_space(*defer_tree
);
2406 vdev_space_update(vd
, alloc_delta
+ defer_delta
, defer_delta
, 0);
2409 * If there's a metaslab_load() in progress, wait for it to complete
2410 * so that we have a consistent view of the in-core space map.
2412 metaslab_load_wait(msp
);
2415 * Move the frees from the defer_tree back to the free
2416 * range tree (if it's loaded). Swap the freed_tree and the
2417 * defer_tree -- this is safe to do because we've just emptied out
2420 range_tree_vacate(*defer_tree
,
2421 msp
->ms_loaded
? range_tree_add
: NULL
, msp
->ms_tree
);
2422 if (defer_allowed
) {
2423 range_tree_swap(&msp
->ms_freedtree
, defer_tree
);
2425 range_tree_vacate(msp
->ms_freedtree
,
2426 msp
->ms_loaded
? range_tree_add
: NULL
, msp
->ms_tree
);
2429 space_map_update(msp
->ms_sm
);
2431 msp
->ms_deferspace
+= defer_delta
;
2432 ASSERT3S(msp
->ms_deferspace
, >=, 0);
2433 ASSERT3S(msp
->ms_deferspace
, <=, msp
->ms_size
);
2434 if (msp
->ms_deferspace
!= 0) {
2436 * Keep syncing this metaslab until all deferred frees
2437 * are back in circulation.
2439 vdev_dirty(vd
, VDD_METASLAB
, msp
, txg
+ 1);
2443 * Calculate the new weights before unloading any metaslabs.
2444 * This will give us the most accurate weighting.
2446 metaslab_group_sort(mg
, msp
, metaslab_weight(msp
));
2449 * If the metaslab is loaded and we've not tried to load or allocate
2450 * from it in 'metaslab_unload_delay' txgs, then unload it.
2452 if (msp
->ms_loaded
&&
2453 msp
->ms_selected_txg
+ metaslab_unload_delay
< txg
) {
2455 for (int t
= 1; t
< TXG_CONCURRENT_STATES
; t
++) {
2456 VERIFY0(range_tree_space(
2457 msp
->ms_alloctree
[(txg
+ t
) & TXG_MASK
]));
2460 if (!metaslab_debug_unload
)
2461 metaslab_unload(msp
);
2464 mutex_exit(&msp
->ms_lock
);
2468 metaslab_sync_reassess(metaslab_group_t
*mg
)
2470 metaslab_group_alloc_update(mg
);
2471 mg
->mg_fragmentation
= metaslab_group_fragmentation(mg
);
2474 * Preload the next potential metaslabs
2476 metaslab_group_preload(mg
);
2480 metaslab_distance(metaslab_t
*msp
, dva_t
*dva
)
2482 uint64_t ms_shift
= msp
->ms_group
->mg_vd
->vdev_ms_shift
;
2483 uint64_t offset
= DVA_GET_OFFSET(dva
) >> ms_shift
;
2484 uint64_t start
= msp
->ms_id
;
2486 if (msp
->ms_group
->mg_vd
->vdev_id
!= DVA_GET_VDEV(dva
))
2487 return (1ULL << 63);
2490 return ((start
- offset
) << ms_shift
);
2492 return ((offset
- start
) << ms_shift
);
2497 * ==========================================================================
2498 * Metaslab allocation tracing facility
2499 * ==========================================================================
2501 #ifdef _METASLAB_TRACING
2502 kstat_t
*metaslab_trace_ksp
;
2503 kstat_named_t metaslab_trace_over_limit
;
2506 metaslab_alloc_trace_init(void)
2508 ASSERT(metaslab_alloc_trace_cache
== NULL
);
2509 metaslab_alloc_trace_cache
= kmem_cache_create(
2510 "metaslab_alloc_trace_cache", sizeof (metaslab_alloc_trace_t
),
2511 0, NULL
, NULL
, NULL
, NULL
, NULL
, 0);
2512 metaslab_trace_ksp
= kstat_create("zfs", 0, "metaslab_trace_stats",
2513 "misc", KSTAT_TYPE_NAMED
, 1, KSTAT_FLAG_VIRTUAL
);
2514 if (metaslab_trace_ksp
!= NULL
) {
2515 metaslab_trace_ksp
->ks_data
= &metaslab_trace_over_limit
;
2516 kstat_named_init(&metaslab_trace_over_limit
,
2517 "metaslab_trace_over_limit", KSTAT_DATA_UINT64
);
2518 kstat_install(metaslab_trace_ksp
);
2523 metaslab_alloc_trace_fini(void)
2525 if (metaslab_trace_ksp
!= NULL
) {
2526 kstat_delete(metaslab_trace_ksp
);
2527 metaslab_trace_ksp
= NULL
;
2529 kmem_cache_destroy(metaslab_alloc_trace_cache
);
2530 metaslab_alloc_trace_cache
= NULL
;
2534 * Add an allocation trace element to the allocation tracing list.
2537 metaslab_trace_add(zio_alloc_list_t
*zal
, metaslab_group_t
*mg
,
2538 metaslab_t
*msp
, uint64_t psize
, uint32_t dva_id
, uint64_t offset
)
2540 metaslab_alloc_trace_t
*mat
;
2542 if (!metaslab_trace_enabled
)
2546 * When the tracing list reaches its maximum we remove
2547 * the second element in the list before adding a new one.
2548 * By removing the second element we preserve the original
2549 * entry as a clue to what allocations steps have already been
2552 if (zal
->zal_size
== metaslab_trace_max_entries
) {
2553 metaslab_alloc_trace_t
*mat_next
;
2555 panic("too many entries in allocation list");
2557 atomic_inc_64(&metaslab_trace_over_limit
.value
.ui64
);
2559 mat_next
= list_next(&zal
->zal_list
, list_head(&zal
->zal_list
));
2560 list_remove(&zal
->zal_list
, mat_next
);
2561 kmem_cache_free(metaslab_alloc_trace_cache
, mat_next
);
2564 mat
= kmem_cache_alloc(metaslab_alloc_trace_cache
, KM_SLEEP
);
2565 list_link_init(&mat
->mat_list_node
);
2568 mat
->mat_size
= psize
;
2569 mat
->mat_dva_id
= dva_id
;
2570 mat
->mat_offset
= offset
;
2571 mat
->mat_weight
= 0;
2574 mat
->mat_weight
= msp
->ms_weight
;
2577 * The list is part of the zio so locking is not required. Only
2578 * a single thread will perform allocations for a given zio.
2580 list_insert_tail(&zal
->zal_list
, mat
);
2583 ASSERT3U(zal
->zal_size
, <=, metaslab_trace_max_entries
);
2587 metaslab_trace_init(zio_alloc_list_t
*zal
)
2589 list_create(&zal
->zal_list
, sizeof (metaslab_alloc_trace_t
),
2590 offsetof(metaslab_alloc_trace_t
, mat_list_node
));
2595 metaslab_trace_fini(zio_alloc_list_t
*zal
)
2597 metaslab_alloc_trace_t
*mat
;
2599 while ((mat
= list_remove_head(&zal
->zal_list
)) != NULL
)
2600 kmem_cache_free(metaslab_alloc_trace_cache
, mat
);
2601 list_destroy(&zal
->zal_list
);
2606 #define metaslab_trace_add(zal, mg, msp, psize, id, off)
2609 metaslab_alloc_trace_init(void)
2614 metaslab_alloc_trace_fini(void)
2619 metaslab_trace_init(zio_alloc_list_t
*zal
)
2624 metaslab_trace_fini(zio_alloc_list_t
*zal
)
2628 #endif /* _METASLAB_TRACING */
2631 * ==========================================================================
2632 * Metaslab block operations
2633 * ==========================================================================
2637 metaslab_group_alloc_increment(spa_t
*spa
, uint64_t vdev
, void *tag
, int flags
)
2639 if (!(flags
& METASLAB_ASYNC_ALLOC
) ||
2640 flags
& METASLAB_DONT_THROTTLE
)
2643 metaslab_group_t
*mg
= vdev_lookup_top(spa
, vdev
)->vdev_mg
;
2644 if (!mg
->mg_class
->mc_alloc_throttle_enabled
)
2647 (void) refcount_add(&mg
->mg_alloc_queue_depth
, tag
);
2651 metaslab_group_alloc_decrement(spa_t
*spa
, uint64_t vdev
, void *tag
, int flags
)
2653 if (!(flags
& METASLAB_ASYNC_ALLOC
) ||
2654 flags
& METASLAB_DONT_THROTTLE
)
2657 metaslab_group_t
*mg
= vdev_lookup_top(spa
, vdev
)->vdev_mg
;
2658 if (!mg
->mg_class
->mc_alloc_throttle_enabled
)
2661 (void) refcount_remove(&mg
->mg_alloc_queue_depth
, tag
);
2665 metaslab_group_alloc_verify(spa_t
*spa
, const blkptr_t
*bp
, void *tag
)
2668 const dva_t
*dva
= bp
->blk_dva
;
2669 int ndvas
= BP_GET_NDVAS(bp
);
2671 for (int d
= 0; d
< ndvas
; d
++) {
2672 uint64_t vdev
= DVA_GET_VDEV(&dva
[d
]);
2673 metaslab_group_t
*mg
= vdev_lookup_top(spa
, vdev
)->vdev_mg
;
2674 VERIFY(refcount_not_held(&mg
->mg_alloc_queue_depth
, tag
));
2680 metaslab_block_alloc(metaslab_t
*msp
, uint64_t size
, uint64_t txg
)
2683 range_tree_t
*rt
= msp
->ms_tree
;
2684 metaslab_class_t
*mc
= msp
->ms_group
->mg_class
;
2686 VERIFY(!msp
->ms_condensing
);
2688 start
= mc
->mc_ops
->msop_alloc(msp
, size
);
2689 if (start
!= -1ULL) {
2690 metaslab_group_t
*mg
= msp
->ms_group
;
2691 vdev_t
*vd
= mg
->mg_vd
;
2693 VERIFY0(P2PHASE(start
, 1ULL << vd
->vdev_ashift
));
2694 VERIFY0(P2PHASE(size
, 1ULL << vd
->vdev_ashift
));
2695 VERIFY3U(range_tree_space(rt
) - size
, <=, msp
->ms_size
);
2696 range_tree_remove(rt
, start
, size
);
2698 if (range_tree_space(msp
->ms_alloctree
[txg
& TXG_MASK
]) == 0)
2699 vdev_dirty(mg
->mg_vd
, VDD_METASLAB
, msp
, txg
);
2701 range_tree_add(msp
->ms_alloctree
[txg
& TXG_MASK
], start
, size
);
2703 /* Track the last successful allocation */
2704 msp
->ms_alloc_txg
= txg
;
2705 metaslab_verify_space(msp
, txg
);
2709 * Now that we've attempted the allocation we need to update the
2710 * metaslab's maximum block size since it may have changed.
2712 msp
->ms_max_size
= metaslab_block_maxsize(msp
);
2717 metaslab_group_alloc_normal(metaslab_group_t
*mg
, zio_alloc_list_t
*zal
,
2718 uint64_t asize
, uint64_t txg
, uint64_t min_distance
, dva_t
*dva
, int d
)
2720 metaslab_t
*msp
= NULL
;
2721 uint64_t offset
= -1ULL;
2722 uint64_t activation_weight
;
2723 uint64_t target_distance
;
2726 activation_weight
= METASLAB_WEIGHT_PRIMARY
;
2727 for (i
= 0; i
< d
; i
++) {
2728 if (DVA_GET_VDEV(&dva
[i
]) == mg
->mg_vd
->vdev_id
) {
2729 activation_weight
= METASLAB_WEIGHT_SECONDARY
;
2734 metaslab_t
*search
= kmem_alloc(sizeof (*search
), KM_SLEEP
);
2735 search
->ms_weight
= UINT64_MAX
;
2736 search
->ms_start
= 0;
2738 boolean_t was_active
;
2739 avl_tree_t
*t
= &mg
->mg_metaslab_tree
;
2742 mutex_enter(&mg
->mg_lock
);
2745 * Find the metaslab with the highest weight that is less
2746 * than what we've already tried. In the common case, this
2747 * means that we will examine each metaslab at most once.
2748 * Note that concurrent callers could reorder metaslabs
2749 * by activation/passivation once we have dropped the mg_lock.
2750 * If a metaslab is activated by another thread, and we fail
2751 * to allocate from the metaslab we have selected, we may
2752 * not try the newly-activated metaslab, and instead activate
2753 * another metaslab. This is not optimal, but generally
2754 * does not cause any problems (a possible exception being
2755 * if every metaslab is completely full except for the
2756 * the newly-activated metaslab which we fail to examine).
2758 msp
= avl_find(t
, search
, &idx
);
2760 msp
= avl_nearest(t
, idx
, AVL_AFTER
);
2761 for (; msp
!= NULL
; msp
= AVL_NEXT(t
, msp
)) {
2763 if (!metaslab_should_allocate(msp
, asize
)) {
2764 metaslab_trace_add(zal
, mg
, msp
, asize
, d
,
2770 * If the selected metaslab is condensing, skip it.
2772 if (msp
->ms_condensing
)
2775 was_active
= msp
->ms_weight
& METASLAB_ACTIVE_MASK
;
2776 if (activation_weight
== METASLAB_WEIGHT_PRIMARY
)
2779 target_distance
= min_distance
+
2780 (space_map_allocated(msp
->ms_sm
) != 0 ? 0 :
2783 for (i
= 0; i
< d
; i
++) {
2784 if (metaslab_distance(msp
, &dva
[i
]) <
2791 mutex_exit(&mg
->mg_lock
);
2793 kmem_free(search
, sizeof (*search
));
2796 search
->ms_weight
= msp
->ms_weight
;
2797 search
->ms_start
= msp
->ms_start
+ 1;
2799 mutex_enter(&msp
->ms_lock
);
2802 * Ensure that the metaslab we have selected is still
2803 * capable of handling our request. It's possible that
2804 * another thread may have changed the weight while we
2805 * were blocked on the metaslab lock. We check the
2806 * active status first to see if we need to reselect
2809 if (was_active
&& !(msp
->ms_weight
& METASLAB_ACTIVE_MASK
)) {
2810 mutex_exit(&msp
->ms_lock
);
2814 if ((msp
->ms_weight
& METASLAB_WEIGHT_SECONDARY
) &&
2815 activation_weight
== METASLAB_WEIGHT_PRIMARY
) {
2816 metaslab_passivate(msp
,
2817 msp
->ms_weight
& ~METASLAB_ACTIVE_MASK
);
2818 mutex_exit(&msp
->ms_lock
);
2822 if (metaslab_activate(msp
, activation_weight
) != 0) {
2823 mutex_exit(&msp
->ms_lock
);
2826 msp
->ms_selected_txg
= txg
;
2829 * Now that we have the lock, recheck to see if we should
2830 * continue to use this metaslab for this allocation. The
2831 * the metaslab is now loaded so metaslab_should_allocate() can
2832 * accurately determine if the allocation attempt should
2835 if (!metaslab_should_allocate(msp
, asize
)) {
2836 /* Passivate this metaslab and select a new one. */
2837 metaslab_trace_add(zal
, mg
, msp
, asize
, d
,
2844 * If this metaslab is currently condensing then pick again as
2845 * we can't manipulate this metaslab until it's committed
2848 if (msp
->ms_condensing
) {
2849 metaslab_trace_add(zal
, mg
, msp
, asize
, d
,
2851 mutex_exit(&msp
->ms_lock
);
2855 offset
= metaslab_block_alloc(msp
, asize
, txg
);
2856 metaslab_trace_add(zal
, mg
, msp
, asize
, d
, offset
);
2858 if (offset
!= -1ULL) {
2859 /* Proactively passivate the metaslab, if needed */
2860 metaslab_segment_may_passivate(msp
);
2864 ASSERT(msp
->ms_loaded
);
2867 * We were unable to allocate from this metaslab so determine
2868 * a new weight for this metaslab. Now that we have loaded
2869 * the metaslab we can provide a better hint to the metaslab
2872 * For space-based metaslabs, we use the maximum block size.
2873 * This information is only available when the metaslab
2874 * is loaded and is more accurate than the generic free
2875 * space weight that was calculated by metaslab_weight().
2876 * This information allows us to quickly compare the maximum
2877 * available allocation in the metaslab to the allocation
2878 * size being requested.
2880 * For segment-based metaslabs, determine the new weight
2881 * based on the highest bucket in the range tree. We
2882 * explicitly use the loaded segment weight (i.e. the range
2883 * tree histogram) since it contains the space that is
2884 * currently available for allocation and is accurate
2885 * even within a sync pass.
2887 if (WEIGHT_IS_SPACEBASED(msp
->ms_weight
)) {
2888 uint64_t weight
= metaslab_block_maxsize(msp
);
2889 WEIGHT_SET_SPACEBASED(weight
);
2890 metaslab_passivate(msp
, weight
);
2892 metaslab_passivate(msp
,
2893 metaslab_weight_from_range_tree(msp
));
2897 * We have just failed an allocation attempt, check
2898 * that metaslab_should_allocate() agrees. Otherwise,
2899 * we may end up in an infinite loop retrying the same
2902 ASSERT(!metaslab_should_allocate(msp
, asize
));
2903 mutex_exit(&msp
->ms_lock
);
2905 mutex_exit(&msp
->ms_lock
);
2906 kmem_free(search
, sizeof (*search
));
2911 metaslab_group_alloc(metaslab_group_t
*mg
, zio_alloc_list_t
*zal
,
2912 uint64_t asize
, uint64_t txg
, uint64_t min_distance
, dva_t
*dva
, int d
)
2915 ASSERT(mg
->mg_initialized
);
2917 offset
= metaslab_group_alloc_normal(mg
, zal
, asize
, txg
,
2918 min_distance
, dva
, d
);
2920 mutex_enter(&mg
->mg_lock
);
2921 if (offset
== -1ULL) {
2922 mg
->mg_failed_allocations
++;
2923 metaslab_trace_add(zal
, mg
, NULL
, asize
, d
,
2924 TRACE_GROUP_FAILURE
);
2925 if (asize
== SPA_GANGBLOCKSIZE
) {
2927 * This metaslab group was unable to allocate
2928 * the minimum gang block size so it must be out of
2929 * space. We must notify the allocation throttle
2930 * to start skipping allocation attempts to this
2931 * metaslab group until more space becomes available.
2932 * Note: this failure cannot be caused by the
2933 * allocation throttle since the allocation throttle
2934 * is only responsible for skipping devices and
2935 * not failing block allocations.
2937 mg
->mg_no_free_space
= B_TRUE
;
2940 mg
->mg_allocations
++;
2941 mutex_exit(&mg
->mg_lock
);
2946 * If we have to write a ditto block (i.e. more than one DVA for a given BP)
2947 * on the same vdev as an existing DVA of this BP, then try to allocate it
2948 * at least (vdev_asize / (2 ^ ditto_same_vdev_distance_shift)) away from the
2951 int ditto_same_vdev_distance_shift
= 3;
2954 * Allocate a block for the specified i/o.
2957 metaslab_alloc_dva(spa_t
*spa
, metaslab_class_t
*mc
, uint64_t psize
,
2958 dva_t
*dva
, int d
, dva_t
*hintdva
, uint64_t txg
, int flags
,
2959 zio_alloc_list_t
*zal
)
2961 metaslab_group_t
*mg
, *fast_mg
, *rotor
;
2963 boolean_t try_hard
= B_FALSE
;
2965 ASSERT(!DVA_IS_VALID(&dva
[d
]));
2968 * For testing, make some blocks above a certain size be gang blocks.
2970 if (psize
>= metaslab_gang_bang
&& (ddi_get_lbolt() & 3) == 0) {
2971 metaslab_trace_add(zal
, NULL
, NULL
, psize
, d
, TRACE_FORCE_GANG
);
2972 return (SET_ERROR(ENOSPC
));
2976 * Start at the rotor and loop through all mgs until we find something.
2977 * Note that there's no locking on mc_rotor or mc_aliquot because
2978 * nothing actually breaks if we miss a few updates -- we just won't
2979 * allocate quite as evenly. It all balances out over time.
2981 * If we are doing ditto or log blocks, try to spread them across
2982 * consecutive vdevs. If we're forced to reuse a vdev before we've
2983 * allocated all of our ditto blocks, then try and spread them out on
2984 * that vdev as much as possible. If it turns out to not be possible,
2985 * gradually lower our standards until anything becomes acceptable.
2986 * Also, allocating on consecutive vdevs (as opposed to random vdevs)
2987 * gives us hope of containing our fault domains to something we're
2988 * able to reason about. Otherwise, any two top-level vdev failures
2989 * will guarantee the loss of data. With consecutive allocation,
2990 * only two adjacent top-level vdev failures will result in data loss.
2992 * If we are doing gang blocks (hintdva is non-NULL), try to keep
2993 * ourselves on the same vdev as our gang block header. That
2994 * way, we can hope for locality in vdev_cache, plus it makes our
2995 * fault domains something tractable.
2998 vd
= vdev_lookup_top(spa
, DVA_GET_VDEV(&hintdva
[d
]));
3001 * It's possible the vdev we're using as the hint no
3002 * longer exists (i.e. removed). Consult the rotor when
3008 if (flags
& METASLAB_HINTBP_AVOID
&&
3009 mg
->mg_next
!= NULL
)
3014 } else if (d
!= 0) {
3015 vd
= vdev_lookup_top(spa
, DVA_GET_VDEV(&dva
[d
- 1]));
3016 mg
= vd
->vdev_mg
->mg_next
;
3017 } else if (flags
& METASLAB_FASTWRITE
) {
3018 mg
= fast_mg
= mc
->mc_rotor
;
3021 if (fast_mg
->mg_vd
->vdev_pending_fastwrite
<
3022 mg
->mg_vd
->vdev_pending_fastwrite
)
3024 } while ((fast_mg
= fast_mg
->mg_next
) != mc
->mc_rotor
);
3031 * If the hint put us into the wrong metaslab class, or into a
3032 * metaslab group that has been passivated, just follow the rotor.
3034 if (mg
->mg_class
!= mc
|| mg
->mg_activation_count
<= 0)
3040 boolean_t allocatable
;
3042 ASSERT(mg
->mg_activation_count
== 1);
3046 * Don't allocate from faulted devices.
3049 spa_config_enter(spa
, SCL_ZIO
, FTAG
, RW_READER
);
3050 allocatable
= vdev_allocatable(vd
);
3051 spa_config_exit(spa
, SCL_ZIO
, FTAG
);
3053 allocatable
= vdev_allocatable(vd
);
3057 * Determine if the selected metaslab group is eligible
3058 * for allocations. If we're ganging then don't allow
3059 * this metaslab group to skip allocations since that would
3060 * inadvertently return ENOSPC and suspend the pool
3061 * even though space is still available.
3063 if (allocatable
&& !GANG_ALLOCATION(flags
) && !try_hard
) {
3064 allocatable
= metaslab_group_allocatable(mg
, rotor
,
3069 metaslab_trace_add(zal
, mg
, NULL
, psize
, d
,
3070 TRACE_NOT_ALLOCATABLE
);
3074 ASSERT(mg
->mg_initialized
);
3077 * Avoid writing single-copy data to a failing,
3078 * non-redundant vdev, unless we've already tried all
3081 if ((vd
->vdev_stat
.vs_write_errors
> 0 ||
3082 vd
->vdev_state
< VDEV_STATE_HEALTHY
) &&
3083 d
== 0 && !try_hard
&& vd
->vdev_children
== 0) {
3084 metaslab_trace_add(zal
, mg
, NULL
, psize
, d
,
3089 ASSERT(mg
->mg_class
== mc
);
3092 * If we don't need to try hard, then require that the
3093 * block be 1/8th of the device away from any other DVAs
3094 * in this BP. If we are trying hard, allow any offset
3095 * to be used (distance=0).
3097 uint64_t distance
= 0;
3099 distance
= vd
->vdev_asize
>>
3100 ditto_same_vdev_distance_shift
;
3101 if (distance
<= (1ULL << vd
->vdev_ms_shift
))
3105 uint64_t asize
= vdev_psize_to_asize(vd
, psize
);
3106 ASSERT(P2PHASE(asize
, 1ULL << vd
->vdev_ashift
) == 0);
3108 uint64_t offset
= metaslab_group_alloc(mg
, zal
, asize
, txg
,
3111 if (offset
!= -1ULL) {
3113 * If we've just selected this metaslab group,
3114 * figure out whether the corresponding vdev is
3115 * over- or under-used relative to the pool,
3116 * and set an allocation bias to even it out.
3118 * Bias is also used to compensate for unequally
3119 * sized vdevs so that space is allocated fairly.
3121 if (mc
->mc_aliquot
== 0 && metaslab_bias_enabled
) {
3122 vdev_stat_t
*vs
= &vd
->vdev_stat
;
3123 int64_t vs_free
= vs
->vs_space
- vs
->vs_alloc
;
3124 int64_t mc_free
= mc
->mc_space
- mc
->mc_alloc
;
3128 * Calculate how much more or less we should
3129 * try to allocate from this device during
3130 * this iteration around the rotor.
3132 * This basically introduces a zero-centered
3133 * bias towards the devices with the most
3134 * free space, while compensating for vdev
3138 * vdev V1 = 16M/128M
3139 * vdev V2 = 16M/128M
3140 * ratio(V1) = 100% ratio(V2) = 100%
3142 * vdev V1 = 16M/128M
3143 * vdev V2 = 64M/128M
3144 * ratio(V1) = 127% ratio(V2) = 72%
3146 * vdev V1 = 16M/128M
3147 * vdev V2 = 64M/512M
3148 * ratio(V1) = 40% ratio(V2) = 160%
3150 ratio
= (vs_free
* mc
->mc_alloc_groups
* 100) /
3152 mg
->mg_bias
= ((ratio
- 100) *
3153 (int64_t)mg
->mg_aliquot
) / 100;
3154 } else if (!metaslab_bias_enabled
) {
3158 if ((flags
& METASLAB_FASTWRITE
) ||
3159 atomic_add_64_nv(&mc
->mc_aliquot
, asize
) >=
3160 mg
->mg_aliquot
+ mg
->mg_bias
) {
3161 mc
->mc_rotor
= mg
->mg_next
;
3165 DVA_SET_VDEV(&dva
[d
], vd
->vdev_id
);
3166 DVA_SET_OFFSET(&dva
[d
], offset
);
3167 DVA_SET_GANG(&dva
[d
],
3168 ((flags
& METASLAB_GANG_HEADER
) ? 1 : 0));
3169 DVA_SET_ASIZE(&dva
[d
], asize
);
3171 if (flags
& METASLAB_FASTWRITE
) {
3172 atomic_add_64(&vd
->vdev_pending_fastwrite
,
3179 mc
->mc_rotor
= mg
->mg_next
;
3181 } while ((mg
= mg
->mg_next
) != rotor
);
3184 * If we haven't tried hard, do so now.
3191 bzero(&dva
[d
], sizeof (dva_t
));
3193 metaslab_trace_add(zal
, rotor
, NULL
, psize
, d
, TRACE_ENOSPC
);
3194 return (SET_ERROR(ENOSPC
));
3198 * Free the block represented by DVA in the context of the specified
3199 * transaction group.
3202 metaslab_free_dva(spa_t
*spa
, const dva_t
*dva
, uint64_t txg
, boolean_t now
)
3204 uint64_t vdev
= DVA_GET_VDEV(dva
);
3205 uint64_t offset
= DVA_GET_OFFSET(dva
);
3206 uint64_t size
= DVA_GET_ASIZE(dva
);
3210 if (txg
> spa_freeze_txg(spa
))
3213 if ((vd
= vdev_lookup_top(spa
, vdev
)) == NULL
|| !DVA_IS_VALID(dva
) ||
3214 (offset
>> vd
->vdev_ms_shift
) >= vd
->vdev_ms_count
) {
3215 zfs_panic_recover("metaslab_free_dva(): bad DVA %llu:%llu:%llu",
3216 (u_longlong_t
)vdev
, (u_longlong_t
)offset
,
3217 (u_longlong_t
)size
);
3221 msp
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
3223 if (DVA_GET_GANG(dva
))
3224 size
= vdev_psize_to_asize(vd
, SPA_GANGBLOCKSIZE
);
3226 mutex_enter(&msp
->ms_lock
);
3229 range_tree_remove(msp
->ms_alloctree
[txg
& TXG_MASK
],
3232 VERIFY(!msp
->ms_condensing
);
3233 VERIFY3U(offset
, >=, msp
->ms_start
);
3234 VERIFY3U(offset
+ size
, <=, msp
->ms_start
+ msp
->ms_size
);
3235 VERIFY3U(range_tree_space(msp
->ms_tree
) + size
, <=,
3237 VERIFY0(P2PHASE(offset
, 1ULL << vd
->vdev_ashift
));
3238 VERIFY0(P2PHASE(size
, 1ULL << vd
->vdev_ashift
));
3239 range_tree_add(msp
->ms_tree
, offset
, size
);
3240 msp
->ms_max_size
= metaslab_block_maxsize(msp
);
3242 VERIFY3U(txg
, ==, spa
->spa_syncing_txg
);
3243 if (range_tree_space(msp
->ms_freeingtree
) == 0)
3244 vdev_dirty(vd
, VDD_METASLAB
, msp
, txg
);
3245 range_tree_add(msp
->ms_freeingtree
, offset
, size
);
3248 mutex_exit(&msp
->ms_lock
);
3252 * Intent log support: upon opening the pool after a crash, notify the SPA
3253 * of blocks that the intent log has allocated for immediate write, but
3254 * which are still considered free by the SPA because the last transaction
3255 * group didn't commit yet.
3258 metaslab_claim_dva(spa_t
*spa
, const dva_t
*dva
, uint64_t txg
)
3260 uint64_t vdev
= DVA_GET_VDEV(dva
);
3261 uint64_t offset
= DVA_GET_OFFSET(dva
);
3262 uint64_t size
= DVA_GET_ASIZE(dva
);
3267 ASSERT(DVA_IS_VALID(dva
));
3269 if ((vd
= vdev_lookup_top(spa
, vdev
)) == NULL
||
3270 (offset
>> vd
->vdev_ms_shift
) >= vd
->vdev_ms_count
)
3271 return (SET_ERROR(ENXIO
));
3273 msp
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
3275 if (DVA_GET_GANG(dva
))
3276 size
= vdev_psize_to_asize(vd
, SPA_GANGBLOCKSIZE
);
3278 mutex_enter(&msp
->ms_lock
);
3280 if ((txg
!= 0 && spa_writeable(spa
)) || !msp
->ms_loaded
)
3281 error
= metaslab_activate(msp
, METASLAB_WEIGHT_SECONDARY
);
3283 if (error
== 0 && !range_tree_contains(msp
->ms_tree
, offset
, size
))
3284 error
= SET_ERROR(ENOENT
);
3286 if (error
|| txg
== 0) { /* txg == 0 indicates dry run */
3287 mutex_exit(&msp
->ms_lock
);
3291 VERIFY(!msp
->ms_condensing
);
3292 VERIFY0(P2PHASE(offset
, 1ULL << vd
->vdev_ashift
));
3293 VERIFY0(P2PHASE(size
, 1ULL << vd
->vdev_ashift
));
3294 VERIFY3U(range_tree_space(msp
->ms_tree
) - size
, <=, msp
->ms_size
);
3295 range_tree_remove(msp
->ms_tree
, offset
, size
);
3297 if (spa_writeable(spa
)) { /* don't dirty if we're zdb(1M) */
3298 if (range_tree_space(msp
->ms_alloctree
[txg
& TXG_MASK
]) == 0)
3299 vdev_dirty(vd
, VDD_METASLAB
, msp
, txg
);
3300 range_tree_add(msp
->ms_alloctree
[txg
& TXG_MASK
], offset
, size
);
3303 mutex_exit(&msp
->ms_lock
);
3309 * Reserve some allocation slots. The reservation system must be called
3310 * before we call into the allocator. If there aren't any available slots
3311 * then the I/O will be throttled until an I/O completes and its slots are
3312 * freed up. The function returns true if it was successful in placing
3316 metaslab_class_throttle_reserve(metaslab_class_t
*mc
, int slots
, zio_t
*zio
,
3319 uint64_t available_slots
= 0;
3320 boolean_t slot_reserved
= B_FALSE
;
3322 ASSERT(mc
->mc_alloc_throttle_enabled
);
3323 mutex_enter(&mc
->mc_lock
);
3325 uint64_t reserved_slots
= refcount_count(&mc
->mc_alloc_slots
);
3326 if (reserved_slots
< mc
->mc_alloc_max_slots
)
3327 available_slots
= mc
->mc_alloc_max_slots
- reserved_slots
;
3329 if (slots
<= available_slots
|| GANG_ALLOCATION(flags
)) {
3331 * We reserve the slots individually so that we can unreserve
3332 * them individually when an I/O completes.
3334 for (int d
= 0; d
< slots
; d
++) {
3335 reserved_slots
= refcount_add(&mc
->mc_alloc_slots
, zio
);
3337 zio
->io_flags
|= ZIO_FLAG_IO_ALLOCATING
;
3338 slot_reserved
= B_TRUE
;
3341 mutex_exit(&mc
->mc_lock
);
3342 return (slot_reserved
);
3346 metaslab_class_throttle_unreserve(metaslab_class_t
*mc
, int slots
, zio_t
*zio
)
3348 ASSERT(mc
->mc_alloc_throttle_enabled
);
3349 mutex_enter(&mc
->mc_lock
);
3350 for (int d
= 0; d
< slots
; d
++) {
3351 (void) refcount_remove(&mc
->mc_alloc_slots
, zio
);
3353 mutex_exit(&mc
->mc_lock
);
3357 metaslab_alloc(spa_t
*spa
, metaslab_class_t
*mc
, uint64_t psize
, blkptr_t
*bp
,
3358 int ndvas
, uint64_t txg
, blkptr_t
*hintbp
, int flags
,
3359 zio_alloc_list_t
*zal
, zio_t
*zio
)
3361 dva_t
*dva
= bp
->blk_dva
;
3362 dva_t
*hintdva
= hintbp
->blk_dva
;
3365 ASSERT(bp
->blk_birth
== 0);
3366 ASSERT(BP_PHYSICAL_BIRTH(bp
) == 0);
3368 spa_config_enter(spa
, SCL_ALLOC
, FTAG
, RW_READER
);
3370 if (mc
->mc_rotor
== NULL
) { /* no vdevs in this class */
3371 spa_config_exit(spa
, SCL_ALLOC
, FTAG
);
3372 return (SET_ERROR(ENOSPC
));
3375 ASSERT(ndvas
> 0 && ndvas
<= spa_max_replication(spa
));
3376 ASSERT(BP_GET_NDVAS(bp
) == 0);
3377 ASSERT(hintbp
== NULL
|| ndvas
<= BP_GET_NDVAS(hintbp
));
3378 ASSERT3P(zal
, !=, NULL
);
3380 for (int d
= 0; d
< ndvas
; d
++) {
3381 error
= metaslab_alloc_dva(spa
, mc
, psize
, dva
, d
, hintdva
,
3384 for (d
--; d
>= 0; d
--) {
3385 metaslab_free_dva(spa
, &dva
[d
], txg
, B_TRUE
);
3386 metaslab_group_alloc_decrement(spa
,
3387 DVA_GET_VDEV(&dva
[d
]), zio
, flags
);
3388 bzero(&dva
[d
], sizeof (dva_t
));
3390 spa_config_exit(spa
, SCL_ALLOC
, FTAG
);
3394 * Update the metaslab group's queue depth
3395 * based on the newly allocated dva.
3397 metaslab_group_alloc_increment(spa
,
3398 DVA_GET_VDEV(&dva
[d
]), zio
, flags
);
3403 ASSERT(BP_GET_NDVAS(bp
) == ndvas
);
3405 spa_config_exit(spa
, SCL_ALLOC
, FTAG
);
3407 BP_SET_BIRTH(bp
, txg
, 0);
3413 metaslab_free(spa_t
*spa
, const blkptr_t
*bp
, uint64_t txg
, boolean_t now
)
3415 const dva_t
*dva
= bp
->blk_dva
;
3416 int ndvas
= BP_GET_NDVAS(bp
);
3418 ASSERT(!BP_IS_HOLE(bp
));
3419 ASSERT(!now
|| bp
->blk_birth
>= spa_syncing_txg(spa
));
3421 spa_config_enter(spa
, SCL_FREE
, FTAG
, RW_READER
);
3423 for (int d
= 0; d
< ndvas
; d
++)
3424 metaslab_free_dva(spa
, &dva
[d
], txg
, now
);
3426 spa_config_exit(spa
, SCL_FREE
, FTAG
);
3430 metaslab_claim(spa_t
*spa
, const blkptr_t
*bp
, uint64_t txg
)
3432 const dva_t
*dva
= bp
->blk_dva
;
3433 int ndvas
= BP_GET_NDVAS(bp
);
3436 ASSERT(!BP_IS_HOLE(bp
));
3440 * First do a dry run to make sure all DVAs are claimable,
3441 * so we don't have to unwind from partial failures below.
3443 if ((error
= metaslab_claim(spa
, bp
, 0)) != 0)
3447 spa_config_enter(spa
, SCL_ALLOC
, FTAG
, RW_READER
);
3449 for (int d
= 0; d
< ndvas
; d
++)
3450 if ((error
= metaslab_claim_dva(spa
, &dva
[d
], txg
)) != 0)
3453 spa_config_exit(spa
, SCL_ALLOC
, FTAG
);
3455 ASSERT(error
== 0 || txg
== 0);
3461 metaslab_fastwrite_mark(spa_t
*spa
, const blkptr_t
*bp
)
3463 const dva_t
*dva
= bp
->blk_dva
;
3464 int ndvas
= BP_GET_NDVAS(bp
);
3465 uint64_t psize
= BP_GET_PSIZE(bp
);
3469 ASSERT(!BP_IS_HOLE(bp
));
3470 ASSERT(!BP_IS_EMBEDDED(bp
));
3473 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
3475 for (d
= 0; d
< ndvas
; d
++) {
3476 if ((vd
= vdev_lookup_top(spa
, DVA_GET_VDEV(&dva
[d
]))) == NULL
)
3478 atomic_add_64(&vd
->vdev_pending_fastwrite
, psize
);
3481 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
3485 metaslab_fastwrite_unmark(spa_t
*spa
, const blkptr_t
*bp
)
3487 const dva_t
*dva
= bp
->blk_dva
;
3488 int ndvas
= BP_GET_NDVAS(bp
);
3489 uint64_t psize
= BP_GET_PSIZE(bp
);
3493 ASSERT(!BP_IS_HOLE(bp
));
3494 ASSERT(!BP_IS_EMBEDDED(bp
));
3497 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
3499 for (d
= 0; d
< ndvas
; d
++) {
3500 if ((vd
= vdev_lookup_top(spa
, DVA_GET_VDEV(&dva
[d
]))) == NULL
)
3502 ASSERT3U(vd
->vdev_pending_fastwrite
, >=, psize
);
3503 atomic_sub_64(&vd
->vdev_pending_fastwrite
, psize
);
3506 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
3510 metaslab_check_free(spa_t
*spa
, const blkptr_t
*bp
)
3512 if ((zfs_flags
& ZFS_DEBUG_ZIO_FREE
) == 0)
3515 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
3516 for (int i
= 0; i
< BP_GET_NDVAS(bp
); i
++) {
3517 uint64_t vdev
= DVA_GET_VDEV(&bp
->blk_dva
[i
]);
3518 vdev_t
*vd
= vdev_lookup_top(spa
, vdev
);
3519 uint64_t offset
= DVA_GET_OFFSET(&bp
->blk_dva
[i
]);
3520 uint64_t size
= DVA_GET_ASIZE(&bp
->blk_dva
[i
]);
3521 metaslab_t
*msp
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
3524 range_tree_verify(msp
->ms_tree
, offset
, size
);
3526 range_tree_verify(msp
->ms_freeingtree
, offset
, size
);
3527 range_tree_verify(msp
->ms_freedtree
, offset
, size
);
3528 for (int j
= 0; j
< TXG_DEFER_SIZE
; j
++)
3529 range_tree_verify(msp
->ms_defertree
[j
], offset
, size
);
3531 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
3534 #if defined(_KERNEL) && defined(HAVE_SPL)
3536 module_param(metaslab_aliquot
, ulong
, 0644);
3537 MODULE_PARM_DESC(metaslab_aliquot
,
3538 "allocation granularity (a.k.a. stripe size)");
3540 module_param(metaslab_debug_load
, int, 0644);
3541 MODULE_PARM_DESC(metaslab_debug_load
,
3542 "load all metaslabs when pool is first opened");
3544 module_param(metaslab_debug_unload
, int, 0644);
3545 MODULE_PARM_DESC(metaslab_debug_unload
,
3546 "prevent metaslabs from being unloaded");
3548 module_param(metaslab_preload_enabled
, int, 0644);
3549 MODULE_PARM_DESC(metaslab_preload_enabled
,
3550 "preload potential metaslabs during reassessment");
3552 module_param(zfs_mg_noalloc_threshold
, int, 0644);
3553 MODULE_PARM_DESC(zfs_mg_noalloc_threshold
,
3554 "percentage of free space for metaslab group to allow allocation");
3556 module_param(zfs_mg_fragmentation_threshold
, int, 0644);
3557 MODULE_PARM_DESC(zfs_mg_fragmentation_threshold
,
3558 "fragmentation for metaslab group to allow allocation");
3560 module_param(zfs_metaslab_fragmentation_threshold
, int, 0644);
3561 MODULE_PARM_DESC(zfs_metaslab_fragmentation_threshold
,
3562 "fragmentation for metaslab to allow allocation");
3564 module_param(metaslab_fragmentation_factor_enabled
, int, 0644);
3565 MODULE_PARM_DESC(metaslab_fragmentation_factor_enabled
,
3566 "use the fragmentation metric to prefer less fragmented metaslabs");
3568 module_param(metaslab_lba_weighting_enabled
, int, 0644);
3569 MODULE_PARM_DESC(metaslab_lba_weighting_enabled
,
3570 "prefer metaslabs with lower LBAs");
3572 module_param(metaslab_bias_enabled
, int, 0644);
3573 MODULE_PARM_DESC(metaslab_bias_enabled
,
3574 "enable metaslab group biasing");
3576 module_param(zfs_metaslab_segment_weight_enabled
, int, 0644);
3577 MODULE_PARM_DESC(zfs_metaslab_segment_weight_enabled
,
3578 "enable segment-based metaslab selection");
3580 module_param(zfs_metaslab_switch_threshold
, int, 0644);
3581 MODULE_PARM_DESC(zfs_metaslab_switch_threshold
,
3582 "segment-based metaslab selection maximum buckets before switching");
3583 #endif /* _KERNEL && HAVE_SPL */