sys/geom/raid/tr_raid0.c

   1 /*-
   2  * Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org>
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  *
  14  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
  15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
  18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  24  * SUCH DAMAGE.
  25  */
  26
  27 #include <sys/cdefs.h>
  28 __FBSDID("$FreeBSD$");
  29
  30 #include <sys/param.h>
  31 #include <sys/bio.h>
  32 #include <sys/endian.h>
  33 #include <sys/kernel.h>
  34 #include <sys/kobj.h>
  35 #include <sys/lock.h>
  36 #include <sys/malloc.h>
  37 #include <sys/mutex.h>
  38 #include <sys/systm.h>
  39 #include <geom/geom.h>
  40 #include "geom/raid/g_raid.h"
  41 #include "g_raid_tr_if.h"
  42
  43 static MALLOC_DEFINE(M_TR_RAID0, "tr_raid0_data", "GEOM_RAID RAID0 data");
  44
  45 struct g_raid_tr_raid0_object {
  46         struct g_raid_tr_object  trso_base;
  47         int                      trso_starting;
  48         int                      trso_stopped;
  49 };
  50
  51 static g_raid_tr_taste_t g_raid_tr_taste_raid0;
  52 static g_raid_tr_event_t g_raid_tr_event_raid0;
  53 static g_raid_tr_start_t g_raid_tr_start_raid0;
  54 static g_raid_tr_stop_t g_raid_tr_stop_raid0;
  55 static g_raid_tr_iostart_t g_raid_tr_iostart_raid0;
  56 static g_raid_tr_iodone_t g_raid_tr_iodone_raid0;
  57 static g_raid_tr_kerneldump_t g_raid_tr_kerneldump_raid0;
  58 static g_raid_tr_free_t g_raid_tr_free_raid0;
  59
  60 static kobj_method_t g_raid_tr_raid0_methods[] = {
  61         KOBJMETHOD(g_raid_tr_taste,     g_raid_tr_taste_raid0),
  62         KOBJMETHOD(g_raid_tr_event,     g_raid_tr_event_raid0),
  63         KOBJMETHOD(g_raid_tr_start,     g_raid_tr_start_raid0),
  64         KOBJMETHOD(g_raid_tr_stop,      g_raid_tr_stop_raid0),
  65         KOBJMETHOD(g_raid_tr_iostart,   g_raid_tr_iostart_raid0),
  66         KOBJMETHOD(g_raid_tr_iodone,    g_raid_tr_iodone_raid0),
  67         KOBJMETHOD(g_raid_tr_kerneldump,        g_raid_tr_kerneldump_raid0),
  68         KOBJMETHOD(g_raid_tr_free,      g_raid_tr_free_raid0),
  69         { 0, 0 }
  70 };
  71
  72 static struct g_raid_tr_class g_raid_tr_raid0_class = {
  73         "RAID0",
  74         g_raid_tr_raid0_methods,
  75         sizeof(struct g_raid_tr_raid0_object),
  76         .trc_enable = 1,
  77         .trc_priority = 100,
  78         .trc_accept_unmapped = 1
  79 };
  80
  81 static int
  82 g_raid_tr_taste_raid0(struct g_raid_tr_object *tr, struct g_raid_volume *volume)
  83 {
  84         struct g_raid_tr_raid0_object *trs;
  85
  86         trs = (struct g_raid_tr_raid0_object *)tr;
  87         if (tr->tro_volume->v_raid_level != G_RAID_VOLUME_RL_RAID0 ||
  88             tr->tro_volume->v_raid_level_qualifier != G_RAID_VOLUME_RLQ_NONE)
  89                 return (G_RAID_TR_TASTE_FAIL);
  90         trs->trso_starting = 1;
  91         return (G_RAID_TR_TASTE_SUCCEED);
  92 }
  93
  94 static int
  95 g_raid_tr_update_state_raid0(struct g_raid_volume *vol)
  96 {
  97         struct g_raid_tr_raid0_object *trs;
  98         struct g_raid_softc *sc;
  99         u_int s;
 100         int n, f;
 101
 102         sc = vol->v_softc;
 103         trs = (struct g_raid_tr_raid0_object *)vol->v_tr;
 104         if (trs->trso_stopped)
 105                 s = G_RAID_VOLUME_S_STOPPED;
 106         else if (trs->trso_starting)
 107                 s = G_RAID_VOLUME_S_STARTING;
 108         else {
 109                 n = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_ACTIVE);
 110                 f = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_FAILED);
 111                 if (n + f == vol->v_disks_count) {
 112                         if (f == 0)
 113                                 s = G_RAID_VOLUME_S_OPTIMAL;
 114                         else
 115                                 s = G_RAID_VOLUME_S_SUBOPTIMAL;
 116                 } else
 117                         s = G_RAID_VOLUME_S_BROKEN;
 118         }
 119         if (s != vol->v_state) {
 120                 g_raid_event_send(vol, G_RAID_VOLUME_S_ALIVE(s) ?
 121                     G_RAID_VOLUME_E_UP : G_RAID_VOLUME_E_DOWN,
 122                     G_RAID_EVENT_VOLUME);
 123                 g_raid_change_volume_state(vol, s);
 124                 if (!trs->trso_starting && !trs->trso_stopped)
 125                         g_raid_write_metadata(sc, vol, NULL, NULL);
 126         }
 127         return (0);
 128 }
 129
 130 static int
 131 g_raid_tr_event_raid0(struct g_raid_tr_object *tr,
 132     struct g_raid_subdisk *sd, u_int event)
 133 {
 134         struct g_raid_tr_raid0_object *trs;
 135         struct g_raid_softc *sc;
 136         struct g_raid_volume *vol;
 137         int state;
 138
 139         trs = (struct g_raid_tr_raid0_object *)tr;
 140         vol = tr->tro_volume;
 141         sc = vol->v_softc;
 142
 143         state = sd->sd_state;
 144         if (state != G_RAID_SUBDISK_S_NONE &&
 145             state != G_RAID_SUBDISK_S_FAILED &&
 146             state != G_RAID_SUBDISK_S_ACTIVE) {
 147                 G_RAID_DEBUG1(1, sc,
 148                     "Promote subdisk %s:%d from %s to ACTIVE.",
 149                     vol->v_name, sd->sd_pos,
 150                     g_raid_subdisk_state2str(sd->sd_state));
 151                 g_raid_change_subdisk_state(sd, G_RAID_SUBDISK_S_ACTIVE);
 152         }
 153         if (state != sd->sd_state &&
 154             !trs->trso_starting && !trs->trso_stopped)
 155                 g_raid_write_metadata(sc, vol, sd, NULL);
 156         g_raid_tr_update_state_raid0(vol);
 157         return (0);
 158 }
 159
 160 static int
 161 g_raid_tr_start_raid0(struct g_raid_tr_object *tr)
 162 {
 163         struct g_raid_tr_raid0_object *trs;
 164         struct g_raid_volume *vol;
 165
 166         trs = (struct g_raid_tr_raid0_object *)tr;
 167         vol = tr->tro_volume;
 168         trs->trso_starting = 0;
 169         g_raid_tr_update_state_raid0(vol);
 170         return (0);
 171 }
 172
 173 static int
 174 g_raid_tr_stop_raid0(struct g_raid_tr_object *tr)
 175 {
 176         struct g_raid_tr_raid0_object *trs;
 177         struct g_raid_volume *vol;
 178
 179         trs = (struct g_raid_tr_raid0_object *)tr;
 180         vol = tr->tro_volume;
 181         trs->trso_starting = 0;
 182         trs->trso_stopped = 1;
 183         g_raid_tr_update_state_raid0(vol);
 184         return (0);
 185 }
 186
 187 static void
 188 g_raid_tr_iostart_raid0(struct g_raid_tr_object *tr, struct bio *bp)
 189 {
 190         struct g_raid_volume *vol;
 191         struct g_raid_subdisk *sd;
 192         struct bio_queue_head queue;
 193         struct bio *cbp;
 194         char *addr;
 195         off_t offset, start, length, nstripe, remain;
 196         u_int no, strip_size;
 197
 198         vol = tr->tro_volume;
 199         if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL &&
 200             vol->v_state != G_RAID_VOLUME_S_SUBOPTIMAL) {
 201                 g_raid_iodone(bp, EIO);
 202                 return;
 203         }
 204         if (bp->bio_cmd == BIO_FLUSH) {
 205                 g_raid_tr_flush_common(tr, bp);
 206                 return;
 207         }
 208         if ((bp->bio_flags & BIO_UNMAPPED) != 0)
 209                 addr = NULL;
 210         else
 211                 addr = bp->bio_data;
 212         strip_size = vol->v_strip_size;
 213
 214         /* Stripe number. */
 215         nstripe = bp->bio_offset / strip_size;
 216         /* Start position in stripe. */
 217         start = bp->bio_offset % strip_size;
 218         /* Disk number. */
 219         no = nstripe % vol->v_disks_count;
 220         /* Stripe start position in disk. */
 221         offset = (nstripe / vol->v_disks_count) * strip_size;
 222         /* Length of data to operate. */
 223         remain = bp->bio_length;
 224
 225         bioq_init(&queue);
 226         do {
 227                 length = MIN(strip_size - start, remain);
 228                 cbp = g_clone_bio(bp);
 229                 if (cbp == NULL)
 230                         goto failure;
 231                 cbp->bio_offset = offset + start;
 232                 cbp->bio_length = length;
 233                 if ((bp->bio_flags & BIO_UNMAPPED) != 0 &&
 234                     bp->bio_cmd != BIO_DELETE) {
 235                         cbp->bio_ma_offset += (uintptr_t)addr;
 236                         cbp->bio_ma += cbp->bio_ma_offset / PAGE_SIZE;
 237                         cbp->bio_ma_offset %= PAGE_SIZE;
 238                         cbp->bio_ma_n = round_page(cbp->bio_ma_offset +
 239                             cbp->bio_length) / PAGE_SIZE;
 240                 } else
 241                         cbp->bio_data = addr;
 242                 cbp->bio_caller1 = &vol->v_subdisks[no];
 243                 bioq_insert_tail(&queue, cbp);
 244                 if (++no >= vol->v_disks_count) {
 245                         no = 0;
 246                         offset += strip_size;
 247                 }
 248                 remain -= length;
 249                 if (bp->bio_cmd != BIO_DELETE)
 250                         addr += length;
 251                 start = 0;
 252         } while (remain > 0);
 253         while ((cbp = bioq_takefirst(&queue)) != NULL) {
 254                 sd = cbp->bio_caller1;
 255                 cbp->bio_caller1 = NULL;
 256                 g_raid_subdisk_iostart(sd, cbp);
 257         }
 258         return;
 259 failure:
 260         while ((cbp = bioq_takefirst(&queue)) != NULL)
 261                 g_destroy_bio(cbp);
 262         if (bp->bio_error == 0)
 263                 bp->bio_error = ENOMEM;
 264         g_raid_iodone(bp, bp->bio_error);
 265 }
 266
 267 static int
 268 g_raid_tr_kerneldump_raid0(struct g_raid_tr_object *tr,
 269     void *virtual, vm_offset_t physical, off_t boffset, size_t blength)
 270 {
 271         struct g_raid_volume *vol;
 272         char *addr;
 273         off_t offset, start, length, nstripe, remain;
 274         u_int no, strip_size;
 275         int error;
 276
 277         vol = tr->tro_volume;
 278         if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL)
 279                 return (ENXIO);
 280         addr = virtual;
 281         strip_size = vol->v_strip_size;
 282
 283         /* Stripe number. */
 284         nstripe = boffset / strip_size;
 285         /* Start position in stripe. */
 286         start = boffset % strip_size;
 287         /* Disk number. */
 288         no = nstripe % vol->v_disks_count;
 289         /* Stripe tart position in disk. */
 290         offset = (nstripe / vol->v_disks_count) * strip_size;
 291         /* Length of data to operate. */
 292         remain = blength;
 293
 294         do {
 295                 length = MIN(strip_size - start, remain);
 296                 error = g_raid_subdisk_kerneldump(&vol->v_subdisks[no],
 297                     addr, 0, offset + start, length);
 298                 if (error != 0)
 299                         return (error);
 300                 if (++no >= vol->v_disks_count) {
 301                         no = 0;
 302                         offset += strip_size;
 303                 }
 304                 remain -= length;
 305                 addr += length;
 306                 start = 0;
 307         } while (remain > 0);
 308         return (0);
 309 }
 310
 311 static void
 312 g_raid_tr_iodone_raid0(struct g_raid_tr_object *tr,
 313     struct g_raid_subdisk *sd,struct bio *bp)
 314 {
 315         struct bio *pbp;
 316
 317         pbp = bp->bio_parent;
 318         if (pbp->bio_error == 0)
 319                 pbp->bio_error = bp->bio_error;
 320         g_destroy_bio(bp);
 321         pbp->bio_inbed++;
 322         if (pbp->bio_children == pbp->bio_inbed) {
 323                 pbp->bio_completed = pbp->bio_length;
 324                 g_raid_iodone(pbp, bp->bio_error);
 325         }
 326 }
 327
 328 static int
 329 g_raid_tr_free_raid0(struct g_raid_tr_object *tr)
 330 {
 331
 332         return (0);
 333 }
 334
 335 G_RAID_TR_DECLARE(raid0, "RAID0");