include/sys/metaslab_impl.h

   1 /*
   2  * CDDL HEADER START
   3  *
   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.
   7  *
   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.
  12  *
  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]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  23  * Use is subject to license terms.
  24  */
  25
  26 /*
  27  * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
  28  */
  29
  30 #ifndef _SYS_METASLAB_IMPL_H
  31 #define _SYS_METASLAB_IMPL_H
  32
  33 #include <sys/metaslab.h>
  34 #include <sys/space_map.h>
  35 #include <sys/range_tree.h>
  36 #include <sys/vdev.h>
  37 #include <sys/txg.h>
  38 #include <sys/avl.h>
  39
  40 #ifdef  __cplusplus
  41 extern "C" {
  42 #endif
  43
  44 /*
  45  * A metaslab class encompasses a category of allocatable top-level vdevs.
  46  * Each top-level vdev is associated with a metaslab group which defines
  47  * the allocatable region for that vdev. Examples of these categories include
  48  * "normal" for data block allocations (i.e. main pool allocations) or "log"
  49  * for allocations designated for intent log devices (i.e. slog devices).
  50  * When a block allocation is requested from the SPA it is associated with a
  51  * metaslab_class_t, and only top-level vdevs (i.e. metaslab groups) belonging
  52  * to the class can be used to satisfy that request. Allocations are done
  53  * by traversing the metaslab groups that are linked off of the mc_rotor field.
  54  * This rotor points to the next metaslab group where allocations will be
  55  * attempted. Allocating a block is a 3 step process -- select the metaslab
  56  * group, select the metaslab, and then allocate the block. The metaslab
  57  * class defines the low-level block allocator that will be used as the
  58  * final step in allocation. These allocators are pluggable allowing each class
  59  * to use a block allocator that best suits that class.
  60  */
  61 struct metaslab_class {
  62         kmutex_t                mc_lock;
  63         spa_t                   *mc_spa;
  64         metaslab_group_t        *mc_rotor;
  65         metaslab_ops_t          *mc_ops;
  66         uint64_t                mc_aliquot;
  67
  68         /*
  69          * Track the number of metaslab groups that have been initialized
  70          * and can accept allocations. An initialized metaslab group is
  71          * one has been completely added to the config (i.e. we have
  72          * updated the MOS config and the space has been added to the pool).
  73          */
  74         uint64_t                mc_groups;
  75
  76         /*
  77          * Toggle to enable/disable the allocation throttle.
  78          */
  79         boolean_t               mc_alloc_throttle_enabled;
  80
  81         /*
  82          * The allocation throttle works on a reservation system. Whenever
  83          * an asynchronous zio wants to perform an allocation it must
  84          * first reserve the number of blocks that it wants to allocate.
  85          * If there aren't sufficient slots available for the pending zio
  86          * then that I/O is throttled until more slots free up. The current
  87          * number of reserved allocations is maintained by the mc_alloc_slots
  88          * refcount. The mc_alloc_max_slots value determines the maximum
  89          * number of allocations that the system allows. Gang blocks are
  90          * allowed to reserve slots even if we've reached the maximum
  91          * number of allocations allowed.
  92          */
  93         uint64_t                mc_alloc_max_slots;
  94         refcount_t              mc_alloc_slots;
  95
  96         uint64_t                mc_alloc_groups; /* # of allocatable groups */
  97
  98         uint64_t                mc_alloc;       /* total allocated space */
  99         uint64_t                mc_deferred;    /* total deferred frees */
 100         uint64_t                mc_space;       /* total space (alloc + free) */
 101         uint64_t                mc_dspace;      /* total deflated space */
 102         uint64_t                mc_histogram[RANGE_TREE_HISTOGRAM_SIZE];
 103 };
 104
 105 /*
 106  * Metaslab groups encapsulate all the allocatable regions (i.e. metaslabs)
 107  * of a top-level vdev. They are linked together to form a circular linked
 108  * list and can belong to only one metaslab class. Metaslab groups may become
 109  * ineligible for allocations for a number of reasons such as limited free
 110  * space, fragmentation, or going offline. When this happens the allocator will
 111  * simply find the next metaslab group in the linked list and attempt
 112  * to allocate from that group instead.
 113  */
 114 struct metaslab_group {
 115         kmutex_t                mg_lock;
 116         avl_tree_t              mg_metaslab_tree;
 117         uint64_t                mg_aliquot;
 118         boolean_t               mg_allocatable;         /* can we allocate? */
 119
 120         /*
 121          * A metaslab group is considered to be initialized only after
 122          * we have updated the MOS config and added the space to the pool.
 123          * We only allow allocation attempts to a metaslab group if it
 124          * has been initialized.
 125          */
 126         boolean_t               mg_initialized;
 127
 128         uint64_t                mg_free_capacity;       /* percentage free */
 129         int64_t                 mg_bias;
 130         int64_t                 mg_activation_count;
 131         metaslab_class_t        *mg_class;
 132         vdev_t                  *mg_vd;
 133         taskq_t                 *mg_taskq;
 134         metaslab_group_t        *mg_prev;
 135         metaslab_group_t        *mg_next;
 136
 137         /*
 138          * Each metaslab group can handle mg_max_alloc_queue_depth allocations
 139          * which are tracked by mg_alloc_queue_depth. It's possible for a
 140          * metaslab group to handle more allocations than its max. This
 141          * can occur when gang blocks are required or when other groups
 142          * are unable to handle their share of allocations.
 143          */
 144         uint64_t                mg_max_alloc_queue_depth;
 145         refcount_t              mg_alloc_queue_depth;
 146
 147         /*
 148          * A metalab group that can no longer allocate the minimum block
 149          * size will set mg_no_free_space. Once a metaslab group is out
 150          * of space then its share of work must be distributed to other
 151          * groups.
 152          */
 153         boolean_t               mg_no_free_space;
 154
 155         uint64_t                mg_allocations;
 156         uint64_t                mg_failed_allocations;
 157         uint64_t                mg_fragmentation;
 158         uint64_t                mg_histogram[RANGE_TREE_HISTOGRAM_SIZE];
 159 };
 160
 161 /*
 162  * This value defines the number of elements in the ms_lbas array. The value
 163  * of 64 was chosen as it covers all power of 2 buckets up to UINT64_MAX.
 164  * This is the equivalent of highbit(UINT64_MAX).
 165  */
 166 #define MAX_LBAS        64
 167
 168 /*
 169  * Each metaslab maintains a set of in-core trees to track metaslab operations.
 170  * The in-core free tree (ms_tree) contains the current list of free segments.
 171  * As blocks are allocated, the allocated segment are removed from the ms_tree
 172  * and added to a per txg allocation tree (ms_alloctree). As blocks are freed,
 173  * they are added to the per txg free tree (ms_freetree). These per txg
 174  * trees allow us to process all allocations and frees in syncing context
 175  * where it is safe to update the on-disk space maps. One additional in-core
 176  * tree is maintained to track deferred frees (ms_defertree). Once a block
 177  * is freed it will move from the ms_freetree to the ms_defertree. A deferred
 178  * free means that a block has been freed but cannot be used by the pool
 179  * until TXG_DEFER_SIZE transactions groups later. For example, a block
 180  * that is freed in txg 50 will not be available for reallocation until
 181  * txg 52 (50 + TXG_DEFER_SIZE).  This provides a safety net for uberblock
 182  * rollback. A pool could be safely rolled back TXG_DEFERS_SIZE
 183  * transactions groups and ensure that no block has been reallocated.
 184  *
 185  * The simplified transition diagram looks like this:
 186  *
 187  *
 188  *      ALLOCATE
 189  *         |
 190  *         V
 191  *    free segment (ms_tree) --------> ms_alloctree ----> (write to space map)
 192  *         ^
 193  *         |
 194  *         |                           ms_freetree <--- FREE
 195  *         |                                 |
 196  *         |                                 |
 197  *         |                                 |
 198  *         +----------- ms_defertree <-------+---------> (write to space map)
 199  *
 200  *
 201  * Each metaslab's space is tracked in a single space map in the MOS,
 202  * which is only updated in syncing context. Each time we sync a txg,
 203  * we append the allocs and frees from that txg to the space map.
 204  * The pool space is only updated once all metaslabs have finished syncing.
 205  *
 206  * To load the in-core free tree we read the space map from disk.
 207  * This object contains a series of alloc and free records that are
 208  * combined to make up the list of all free segments in this metaslab. These
 209  * segments are represented in-core by the ms_tree and are stored in an
 210  * AVL tree.
 211  *
 212  * As the space map grows (as a result of the appends) it will
 213  * eventually become space-inefficient. When the metaslab's in-core free tree
 214  * is zfs_condense_pct/100 times the size of the minimal on-disk
 215  * representation, we rewrite it in its minimized form. If a metaslab
 216  * needs to condense then we must set the ms_condensing flag to ensure
 217  * that allocations are not performed on the metaslab that is being written.
 218  */
 219 struct metaslab {
 220         kmutex_t        ms_lock;
 221         kcondvar_t      ms_load_cv;
 222         space_map_t     *ms_sm;
 223         metaslab_ops_t  *ms_ops;
 224         uint64_t        ms_id;
 225         uint64_t        ms_start;
 226         uint64_t        ms_size;
 227         uint64_t        ms_fragmentation;
 228
 229         range_tree_t    *ms_alloctree[TXG_SIZE];
 230         range_tree_t    *ms_freetree[TXG_SIZE];
 231         range_tree_t    *ms_defertree[TXG_DEFER_SIZE];
 232         range_tree_t    *ms_tree;
 233
 234         boolean_t       ms_condensing;  /* condensing? */
 235         boolean_t       ms_condense_wanted;
 236         boolean_t       ms_loaded;
 237         boolean_t       ms_loading;
 238
 239         int64_t         ms_deferspace;  /* sum of ms_defermap[] space   */
 240         uint64_t        ms_weight;      /* weight vs. others in group   */
 241         uint64_t        ms_access_txg;
 242
 243         /*
 244          * The metaslab block allocators can optionally use a size-ordered
 245          * range tree and/or an array of LBAs. Not all allocators use
 246          * this functionality. The ms_size_tree should always contain the
 247          * same number of segments as the ms_tree. The only difference
 248          * is that the ms_size_tree is ordered by segment sizes.
 249          */
 250         avl_tree_t      ms_size_tree;
 251         uint64_t        ms_lbas[MAX_LBAS];
 252
 253         metaslab_group_t *ms_group;     /* metaslab group               */
 254         avl_node_t      ms_group_node;  /* node in metaslab group tree  */
 255         txg_node_t      ms_txg_node;    /* per-txg dirty metaslab links */
 256 };
 257
 258 #ifdef  __cplusplus
 259 }
 260 #endif
 261
 262 #endif  /* _SYS_METASLAB_IMPL_H */