expandable RAID-Z

          --- old/usr/src/uts/common/fs/zfs/vdev_root.c
          +++ new/usr/src/uts/common/fs/zfs/vdev_root.c

   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
  17   17   * information: Portions Copyright [yyyy] [name of copyright owner]
  18   18   *
  19   19   * CDDL HEADER END
  20   20   */
  21   21  /*
  22   22   * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
  23   23   * Use is subject to license terms.
  24   24   */
  25   25  
  26   26  #pragma ident   "%Z%%M% %I%     %E% SMI"
  27   27  
  28   28  #include <sys/zfs_context.h>
  29   29  #include <sys/spa.h>
  30   30  #include <sys/vdev_impl.h>
  31   31  #include <sys/zio.h>
  32   32  #include <sys/fs/zfs.h>
  33   33  
  34   34  /*
  35   35   * Virtual device vector for the pool's root vdev.
  36   36   */
  37   37  
  38   38  /*
  39   39   * We should be able to tolerate one failure with absolutely no damage
  40   40   * to our metadata.  Two failures will take out space maps, a bunch of
  41   41   * indirect block trees, meta dnodes, dnodes, etc.  Probably not a happy
  42   42   * place to live.  When we get smarter, we can liberalize this policy.
  43   43   * e.g. If we haven't lost two consecutive top-level vdevs, then we are
  44   44   * probably fine.  Adding bean counters during alloc/free can make this
  45   45   * future guesswork more accurate.
  46   46   */
  47   47  static int
  48   48  too_many_errors(vdev_t *vd, int numerrors)
  49   49  {
  50   50          ASSERT3U(numerrors, <=, vd->vdev_children);
  51   51          return (numerrors == vd->vdev_children);
  52   52  }
  53   53  
  54   54  static int
  55   55  vdev_root_open(vdev_t *vd, uint64_t *asize, uint64_t *ashift)
  56   56  {
  57   57          int c;
  58   58          int lasterror = 0;
  59   59          int numerrors = 0;
  60   60  
  61   61          if (vd->vdev_children == 0) {
  62   62                  vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
  63   63                  return (EINVAL);
  64   64          }
  65   65  
  66   66          for (c = 0; c < vd->vdev_children; c++) {
  67   67                  vdev_t *cvd = vd->vdev_child[c];
  68   68                  int error;
  69   69  
  70   70                  if ((error = vdev_open(cvd)) != 0) {
  71   71                          lasterror = error;
  72   72                          numerrors++;
  73   73                          continue;
  74   74                  }
  75   75          }
  76   76  
  77   77          if (numerrors > 0) {
  78   78                  if (!too_many_errors(vd, numerrors)) {
  79   79                          /* XXX - should not be explicitly setting this state */
  80   80                          vdev_set_state(vd, B_FALSE, VDEV_STATE_FAULTED,
  81   81                              VDEV_AUX_NO_REPLICAS);
  82   82                  } else {
  83   83                          vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
  84   84                          return (lasterror);
  85   85                  }
  86   86          }
  87   87  
  88   88          *asize = 0;
  89   89          *ashift = 0;
  90   90  
  91   91          return (0);
  92   92  }
  93   93  
  94   94  static void
  95   95  vdev_root_close(vdev_t *vd)
  96   96  {
  97   97          int c;
  98   98  
  99   99          for (c = 0; c < vd->vdev_children; c++)
 100  100                  vdev_close(vd->vdev_child[c]);
 101  101  }
 102  102  
 103  103  static void
 104  104  vdev_root_state_change(vdev_t *vd, int faulted, int degraded)
 105  105  {
 106  106          if (faulted) {
 107  107                  if (too_many_errors(vd, faulted))
 108  108                          vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
 109  109                              VDEV_AUX_NO_REPLICAS);
 110  110                  else
 111  111                          vdev_set_state(vd, B_FALSE, VDEV_STATE_FAULTED,
 112  112                              VDEV_AUX_NO_REPLICAS);
 113  113          } else if (degraded) {
 114  114                  vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
 115  115          } else {
 116  116                  vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE);
 117  117          }

 118  118  }
 119  119  
 120  120  vdev_ops_t vdev_root_ops = {
 121  121          vdev_root_open,
 122  122          vdev_root_close,
 123  123          NULL,
 124  124          vdev_default_asize,
 125  125          NULL,                   /* io_start - not applicable to the root */
 126  126          NULL,                   /* io_done - not applicable to the root */
 127  127          vdev_root_state_change,
      128 +        NULL,
 128  129          VDEV_TYPE_ROOT,         /* name of this vdev type */
 129  130          B_FALSE                 /* not a leaf vdev */
 130  131  };
    

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX