Print this page
    
10592 misc. metaslab and vdev related ZoL bug fixes
Portions contributed by: Jerry Jelinek <jerry.jelinek@joyent.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: Giuseppe Di Natale <guss80@gmail.com>
Reviewed by: George Melikov <mail@gmelikov.ru>
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed by: Tony Hutter <hutter2@llnl.gov>
Reviewed by: Kody Kantor <kody.kantor@joyent.com>
Approved by: Dan McDonald <danmcd@joyent.com>
    
      
        | Split | 
	Close | 
      
      | Expand all | 
      | Collapse all | 
    
    
          --- old/usr/src/uts/common/fs/zfs/sys/space_map.h
          +++ new/usr/src/uts/common/fs/zfs/sys/space_map.h
   1    1  /*
   2    2   * CDDL HEADER START
   3    3   *
   4    4   * The contents of this file are subject to the terms of the
   5    5   * Common Development and Distribution License (the "License").
   6    6   * You may not use this file except in compliance with the License.
   7    7   *
   8    8   * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9    9   * or http://www.opensolaris.org/os/licensing.
  10   10   * See the License for the specific language governing permissions
  11   11   * and limitations under the License.
  12   12   *
  13   13   * When distributing Covered Code, include this CDDL HEADER in each
  14   14   * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15   15   * If applicable, add the following below this CDDL HEADER, with the
  16   16   * fields enclosed by brackets "[]" replaced with your own identifying
  
    | 
      ↓ open down ↓ | 
    16 lines elided | 
    
      ↑ open up ↑ | 
  
  17   17   * information: Portions Copyright [yyyy] [name of copyright owner]
  18   18   *
  19   19   * CDDL HEADER END
  20   20   */
  21   21  /*
  22   22   * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  23   23   * Use is subject to license terms.
  24   24   */
  25   25  
  26   26  /*
  27      - * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
       27 + * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
  28   28   */
  29   29  
  30   30  #ifndef _SYS_SPACE_MAP_H
  31   31  #define _SYS_SPACE_MAP_H
  32   32  
  33   33  #include <sys/avl.h>
  34   34  #include <sys/range_tree.h>
  35   35  #include <sys/dmu.h>
  36   36  
  37   37  #ifdef  __cplusplus
  38   38  extern "C" {
  39   39  #endif
  40   40  
  41   41  /*
  42   42   * The size of the space map object has increased to include a histogram.
  43   43   * The SPACE_MAP_SIZE_V0 designates the original size and is used to
  44   44   * maintain backward compatibility.
  45   45   */
  46   46  #define SPACE_MAP_SIZE_V0       (3 * sizeof (uint64_t))
  47   47  #define SPACE_MAP_HISTOGRAM_SIZE        32
  
    | 
      ↓ open down ↓ | 
    10 lines elided | 
    
      ↑ open up ↑ | 
  
  48   48  
  49   49  /*
  50   50   * The space_map_phys is the on-disk representation of the space map.
  51   51   * Consumers of space maps should never reference any of the members of this
  52   52   * structure directly. These members may only be updated in syncing context.
  53   53   *
  54   54   * Note the smp_object is no longer used but remains in the structure
  55   55   * for backward compatibility.
  56   56   */
  57   57  typedef struct space_map_phys {
  58      -        uint64_t        smp_object;     /* on-disk space map object */
  59      -        uint64_t        smp_objsize;    /* size of the object */
  60      -        int64_t         smp_alloc;      /* space allocated from the map */
  61      -        uint64_t        smp_pad[5];     /* reserved */
       58 +        /* object number: not needed but kept for backwards compatibility */
       59 +        uint64_t        smp_object;
  62   60  
       61 +        /* length of the object in bytes */
       62 +        uint64_t        smp_length;
       63 +
       64 +        /* space allocated from the map */
       65 +        int64_t         smp_alloc;
       66 +
       67 +        /* reserved */
       68 +        uint64_t        smp_pad[5];
       69 +
  63   70          /*
  64   71           * The smp_histogram maintains a histogram of free regions. Each
  65   72           * bucket, smp_histogram[i], contains the number of free regions
  66   73           * whose size is:
  67   74           * 2^(i+sm_shift) <= size of free region in bytes < 2^(i+sm_shift+1)
  68   75           */
  69   76          uint64_t        smp_histogram[SPACE_MAP_HISTOGRAM_SIZE];
  70   77  } space_map_phys_t;
  71   78  
  72   79  /*
  73   80   * The space map object defines a region of space, its size, how much is
  
    | 
      ↓ open down ↓ | 
    1 lines elided | 
    
      ↑ open up ↑ | 
  
  74   81   * allocated, and the on-disk object that stores this information.
  75   82   * Consumers of space maps may only access the members of this structure.
  76   83   *
  77   84   * Note: the space_map may not be accessed concurrently; consumers
  78   85   * must provide external locking if required.
  79   86   */
  80   87  typedef struct space_map {
  81   88          uint64_t        sm_start;       /* start of map */
  82   89          uint64_t        sm_size;        /* size of map */
  83   90          uint8_t         sm_shift;       /* unit shift */
  84      -        uint64_t        sm_length;      /* synced length */
  85      -        int64_t         sm_alloc;       /* synced space allocated */
  86   91          objset_t        *sm_os;         /* objset for this map */
  87   92          uint64_t        sm_object;      /* object id for this map */
  88   93          uint32_t        sm_blksz;       /* block size for space map */
  89   94          dmu_buf_t       *sm_dbuf;       /* space_map_phys_t dbuf */
  90   95          space_map_phys_t *sm_phys;      /* on-disk space map */
  91   96  } space_map_t;
  92   97  
  93   98  /*
  94   99   * debug entry
  95  100   *
  96  101   *     2     2        10                     50
  97  102   *  +-----+-----+------------+----------------------------------+
  98  103   *  | 1 0 | act |  syncpass  |        txg (lower bits)          |
  99  104   *  +-----+-----+------------+----------------------------------+
 100  105   *   63 62 61 60 59        50 49                                0
 101  106   *
 102  107   *
 103  108   * one-word entry
 104  109   *
 105  110   *    1               47                   1           15
 106  111   *  +-----------------------------------------------------------+
 107  112   *  | 0 |   offset (sm_shift units)    | type |       run       |
 108  113   *  +-----------------------------------------------------------+
 109  114   *   63  62                          16   15   14               0
 110  115   *
 111  116   *
 112  117   * two-word entry
 113  118   *
 114  119   *     2     2               36                      24
 115  120   *  +-----+-----+---------------------------+-------------------+
 116  121   *  | 1 1 | pad |            run            |       vdev        |
 117  122   *  +-----+-----+---------------------------+-------------------+
 118  123   *   63 62 61 60 59                       24 23                 0
 119  124   *
 120  125   *     1                            63
 121  126   *  +------+----------------------------------------------------+
 122  127   *  | type |                      offset                        |
 123  128   *  +------+----------------------------------------------------+
 124  129   *     63   62                                                  0
 125  130   *
 126  131   * Note that a two-word entry will not strandle a block boundary.
 127  132   * If necessary, the last word of a block will be padded with a
 128  133   * debug entry (with act = syncpass = txg = 0).
 129  134   */
 130  135  
 131  136  typedef enum {
 132  137          SM_ALLOC,
 133  138          SM_FREE
 134  139  } maptype_t;
 135  140  
 136  141  typedef struct space_map_entry {
 137  142          maptype_t sme_type;
 138  143          uint32_t sme_vdev;      /* max is 2^24-1; SM_NO_VDEVID if not present */
 139  144          uint64_t sme_offset;    /* max is 2^63-1; units of sm_shift */
 140  145          uint64_t sme_run;       /* max is 2^36; units of sm_shift */
 141  146  } space_map_entry_t;
 142  147  
 143  148  #define SM_NO_VDEVID    (1 << SPA_VDEVBITS)
 144  149  
 145  150  /* one-word entry constants */
 146  151  #define SM_DEBUG_PREFIX 2
 147  152  #define SM_OFFSET_BITS  47
 148  153  #define SM_RUN_BITS     15
 149  154  
 150  155  /* two-word entry constants */
 151  156  #define SM2_PREFIX      3
 152  157  #define SM2_OFFSET_BITS 63
 153  158  #define SM2_RUN_BITS    36
 154  159  
 155  160  #define SM_PREFIX_DECODE(x)     BF64_DECODE(x, 62, 2)
 156  161  #define SM_PREFIX_ENCODE(x)     BF64_ENCODE(x, 62, 2)
 157  162  
 158  163  #define SM_DEBUG_ACTION_DECODE(x)       BF64_DECODE(x, 60, 2)
 159  164  #define SM_DEBUG_ACTION_ENCODE(x)       BF64_ENCODE(x, 60, 2)
 160  165  #define SM_DEBUG_SYNCPASS_DECODE(x)     BF64_DECODE(x, 50, 10)
 161  166  #define SM_DEBUG_SYNCPASS_ENCODE(x)     BF64_ENCODE(x, 50, 10)
 162  167  #define SM_DEBUG_TXG_DECODE(x)          BF64_DECODE(x, 0, 50)
 163  168  #define SM_DEBUG_TXG_ENCODE(x)          BF64_ENCODE(x, 0, 50)
 164  169  
 165  170  #define SM_OFFSET_DECODE(x)     BF64_DECODE(x, 16, SM_OFFSET_BITS)
 166  171  #define SM_OFFSET_ENCODE(x)     BF64_ENCODE(x, 16, SM_OFFSET_BITS)
 167  172  #define SM_TYPE_DECODE(x)       BF64_DECODE(x, 15, 1)
 168  173  #define SM_TYPE_ENCODE(x)       BF64_ENCODE(x, 15, 1)
 169  174  #define SM_RUN_DECODE(x)        (BF64_DECODE(x, 0, SM_RUN_BITS) + 1)
 170  175  #define SM_RUN_ENCODE(x)        BF64_ENCODE((x) - 1, 0, SM_RUN_BITS)
 171  176  #define SM_RUN_MAX              SM_RUN_DECODE(~0ULL)
 172  177  #define SM_OFFSET_MAX           SM_OFFSET_DECODE(~0ULL)
 173  178  
 174  179  #define SM2_RUN_DECODE(x)       (BF64_DECODE(x, SPA_VDEVBITS, SM2_RUN_BITS) + 1)
 175  180  #define SM2_RUN_ENCODE(x)       BF64_ENCODE((x) - 1, SPA_VDEVBITS, SM2_RUN_BITS)
 176  181  #define SM2_VDEV_DECODE(x)      BF64_DECODE(x, 0, SPA_VDEVBITS)
 177  182  #define SM2_VDEV_ENCODE(x)      BF64_ENCODE(x, 0, SPA_VDEVBITS)
 178  183  #define SM2_TYPE_DECODE(x)      BF64_DECODE(x, SM2_OFFSET_BITS, 1)
 179  184  #define SM2_TYPE_ENCODE(x)      BF64_ENCODE(x, SM2_OFFSET_BITS, 1)
 180  185  #define SM2_OFFSET_DECODE(x)    BF64_DECODE(x, 0, SM2_OFFSET_BITS)
 181  186  #define SM2_OFFSET_ENCODE(x)    BF64_ENCODE(x, 0, SM2_OFFSET_BITS)
  
    | 
      ↓ open down ↓ | 
    86 lines elided | 
    
      ↑ open up ↑ | 
  
 182  187  #define SM2_RUN_MAX             SM2_RUN_DECODE(~0ULL)
 183  188  #define SM2_OFFSET_MAX          SM2_OFFSET_DECODE(~0ULL)
 184  189  
 185  190  boolean_t sm_entry_is_debug(uint64_t e);
 186  191  boolean_t sm_entry_is_single_word(uint64_t e);
 187  192  boolean_t sm_entry_is_double_word(uint64_t e);
 188  193  
 189  194  typedef int (*sm_cb_t)(space_map_entry_t *sme, void *arg);
 190  195  
 191  196  int space_map_load(space_map_t *sm, range_tree_t *rt, maptype_t maptype);
 192      -int space_map_iterate(space_map_t *sm, sm_cb_t callback, void *arg);
      197 +int space_map_load_length(space_map_t *sm, range_tree_t *rt, maptype_t maptype,
      198 +    uint64_t length);
      199 +int space_map_iterate(space_map_t *sm, uint64_t length,
      200 +    sm_cb_t callback, void *arg);
 193  201  int space_map_incremental_destroy(space_map_t *sm, sm_cb_t callback, void *arg,
 194  202      dmu_tx_t *tx);
 195  203  
      204 +boolean_t space_map_histogram_verify(space_map_t *sm, range_tree_t *rt);
 196  205  void space_map_histogram_clear(space_map_t *sm);
 197  206  void space_map_histogram_add(space_map_t *sm, range_tree_t *rt,
 198  207      dmu_tx_t *tx);
 199  208  
 200      -void space_map_update(space_map_t *sm);
 201      -
 202  209  uint64_t space_map_object(space_map_t *sm);
 203      -uint64_t space_map_allocated(space_map_t *sm);
      210 +int64_t space_map_allocated(space_map_t *sm);
 204  211  uint64_t space_map_length(space_map_t *sm);
 205  212  
 206  213  void space_map_write(space_map_t *sm, range_tree_t *rt, maptype_t maptype,
 207  214      uint64_t vdev_id, dmu_tx_t *tx);
 208  215  uint64_t space_map_estimate_optimal_size(space_map_t *sm, range_tree_t *rt,
 209  216      uint64_t vdev_id);
 210  217  void space_map_truncate(space_map_t *sm, int blocksize, dmu_tx_t *tx);
 211  218  uint64_t space_map_alloc(objset_t *os, int blocksize, dmu_tx_t *tx);
 212  219  void space_map_free(space_map_t *sm, dmu_tx_t *tx);
 213  220  void space_map_free_obj(objset_t *os, uint64_t smobj, dmu_tx_t *tx);
 214  221  
 215  222  int space_map_open(space_map_t **smp, objset_t *os, uint64_t object,
 216  223      uint64_t start, uint64_t size, uint8_t shift);
 217  224  void space_map_close(space_map_t *sm);
 218  225  
 219      -int64_t space_map_alloc_delta(space_map_t *sm);
 220      -
 221  226  #ifdef  __cplusplus
 222  227  }
 223  228  #endif
 224  229  
 225  230  #endif  /* _SYS_SPACE_MAP_H */
    
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX