sys/geom/raid/tr_raid0.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
   3  *
   4  * Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org>
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  *
  16  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
  17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
  20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  26  * SUCH DAMAGE.
  27  */
  28
  29 #include <sys/cdefs.h>
  30 __FBSDID("$FreeBSD$");
  31
  32 #include <sys/param.h>
  33 #include <sys/bio.h>
  34 #include <sys/endian.h>
  35 #include <sys/kernel.h>
  36 #include <sys/kobj.h>
  37 #include <sys/lock.h>
  38 #include <sys/malloc.h>
  39 #include <sys/mutex.h>
  40 #include <sys/systm.h>
  41 #include <geom/geom.h>
  42 #include "geom/raid/g_raid.h"
  43 #include "g_raid_tr_if.h"
  44
  45 static MALLOC_DEFINE(M_TR_RAID0, "tr_raid0_data", "GEOM_RAID RAID0 data");
  46
  47 struct g_raid_tr_raid0_object {
  48         struct g_raid_tr_object  trso_base;
  49         int                      trso_starting;
  50         int                      trso_stopped;
  51 };
  52
  53 static g_raid_tr_taste_t g_raid_tr_taste_raid0;
  54 static g_raid_tr_event_t g_raid_tr_event_raid0;
  55 static g_raid_tr_start_t g_raid_tr_start_raid0;
  56 static g_raid_tr_stop_t g_raid_tr_stop_raid0;
  57 static g_raid_tr_iostart_t g_raid_tr_iostart_raid0;
  58 static g_raid_tr_iodone_t g_raid_tr_iodone_raid0;
  59 static g_raid_tr_kerneldump_t g_raid_tr_kerneldump_raid0;
  60 static g_raid_tr_free_t g_raid_tr_free_raid0;
  61
  62 static kobj_method_t g_raid_tr_raid0_methods[] = {
  63         KOBJMETHOD(g_raid_tr_taste,     g_raid_tr_taste_raid0),
  64         KOBJMETHOD(g_raid_tr_event,     g_raid_tr_event_raid0),
  65         KOBJMETHOD(g_raid_tr_start,     g_raid_tr_start_raid0),
  66         KOBJMETHOD(g_raid_tr_stop,      g_raid_tr_stop_raid0),
  67         KOBJMETHOD(g_raid_tr_iostart,   g_raid_tr_iostart_raid0),
  68         KOBJMETHOD(g_raid_tr_iodone,    g_raid_tr_iodone_raid0),
  69         KOBJMETHOD(g_raid_tr_kerneldump,        g_raid_tr_kerneldump_raid0),
  70         KOBJMETHOD(g_raid_tr_free,      g_raid_tr_free_raid0),
  71         { 0, 0 }
  72 };
  73
  74 static struct g_raid_tr_class g_raid_tr_raid0_class = {
  75         "RAID0",
  76         g_raid_tr_raid0_methods,
  77         sizeof(struct g_raid_tr_raid0_object),
  78         .trc_enable = 1,
  79         .trc_priority = 100,
  80         .trc_accept_unmapped = 1
  81 };
  82
  83 static int
  84 g_raid_tr_taste_raid0(struct g_raid_tr_object *tr, struct g_raid_volume *volume)
  85 {
  86         struct g_raid_tr_raid0_object *trs;
  87
  88         trs = (struct g_raid_tr_raid0_object *)tr;
  89         if (tr->tro_volume->v_raid_level != G_RAID_VOLUME_RL_RAID0 ||
  90             tr->tro_volume->v_raid_level_qualifier != G_RAID_VOLUME_RLQ_NONE)
  91                 return (G_RAID_TR_TASTE_FAIL);
  92         trs->trso_starting = 1;
  93         return (G_RAID_TR_TASTE_SUCCEED);
  94 }
  95
  96 static int
  97 g_raid_tr_update_state_raid0(struct g_raid_volume *vol)
  98 {
  99         struct g_raid_tr_raid0_object *trs;
 100         struct g_raid_softc *sc;
 101         u_int s;
 102         int n, f;
 103
 104         sc = vol->v_softc;
 105         trs = (struct g_raid_tr_raid0_object *)vol->v_tr;
 106         if (trs->trso_stopped)
 107                 s = G_RAID_VOLUME_S_STOPPED;
 108         else if (trs->trso_starting)
 109                 s = G_RAID_VOLUME_S_STARTING;
 110         else {
 111                 n = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_ACTIVE);
 112                 f = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_FAILED);
 113                 if (n + f == vol->v_disks_count) {
 114                         if (f == 0)
 115                                 s = G_RAID_VOLUME_S_OPTIMAL;
 116                         else
 117                                 s = G_RAID_VOLUME_S_SUBOPTIMAL;
 118                 } else
 119                         s = G_RAID_VOLUME_S_BROKEN;
 120         }
 121         if (s != vol->v_state) {
 122                 g_raid_event_send(vol, G_RAID_VOLUME_S_ALIVE(s) ?
 123                     G_RAID_VOLUME_E_UP : G_RAID_VOLUME_E_DOWN,
 124                     G_RAID_EVENT_VOLUME);
 125                 g_raid_change_volume_state(vol, s);
 126                 if (!trs->trso_starting && !trs->trso_stopped)
 127                         g_raid_write_metadata(sc, vol, NULL, NULL);
 128         }
 129         return (0);
 130 }
 131
 132 static int
 133 g_raid_tr_event_raid0(struct g_raid_tr_object *tr,
 134     struct g_raid_subdisk *sd, u_int event)
 135 {
 136         struct g_raid_tr_raid0_object *trs;
 137         struct g_raid_softc *sc;
 138         struct g_raid_volume *vol;
 139         int state;
 140
 141         trs = (struct g_raid_tr_raid0_object *)tr;
 142         vol = tr->tro_volume;
 143         sc = vol->v_softc;
 144
 145         state = sd->sd_state;
 146         if (state != G_RAID_SUBDISK_S_NONE &&
 147             state != G_RAID_SUBDISK_S_FAILED &&
 148             state != G_RAID_SUBDISK_S_ACTIVE) {
 149                 G_RAID_DEBUG1(1, sc,
 150                     "Promote subdisk %s:%d from %s to ACTIVE.",
 151                     vol->v_name, sd->sd_pos,
 152                     g_raid_subdisk_state2str(sd->sd_state));
 153                 g_raid_change_subdisk_state(sd, G_RAID_SUBDISK_S_ACTIVE);
 154         }
 155         if (state != sd->sd_state &&
 156             !trs->trso_starting && !trs->trso_stopped)
 157                 g_raid_write_metadata(sc, vol, sd, NULL);
 158         g_raid_tr_update_state_raid0(vol);
 159         return (0);
 160 }
 161
 162 static int
 163 g_raid_tr_start_raid0(struct g_raid_tr_object *tr)
 164 {
 165         struct g_raid_tr_raid0_object *trs;
 166         struct g_raid_volume *vol;
 167
 168         trs = (struct g_raid_tr_raid0_object *)tr;
 169         vol = tr->tro_volume;
 170         trs->trso_starting = 0;
 171         g_raid_tr_update_state_raid0(vol);
 172         return (0);
 173 }
 174
 175 static int
 176 g_raid_tr_stop_raid0(struct g_raid_tr_object *tr)
 177 {
 178         struct g_raid_tr_raid0_object *trs;
 179         struct g_raid_volume *vol;
 180
 181         trs = (struct g_raid_tr_raid0_object *)tr;
 182         vol = tr->tro_volume;
 183         trs->trso_starting = 0;
 184         trs->trso_stopped = 1;
 185         g_raid_tr_update_state_raid0(vol);
 186         return (0);
 187 }
 188
 189 static void
 190 g_raid_tr_iostart_raid0(struct g_raid_tr_object *tr, struct bio *bp)
 191 {
 192         struct g_raid_volume *vol;
 193         struct g_raid_subdisk *sd;
 194         struct bio_queue_head queue;
 195         struct bio *cbp;
 196         char *addr;
 197         off_t offset, start, length, nstripe, remain;
 198         u_int no, strip_size;
 199
 200         vol = tr->tro_volume;
 201         if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL &&
 202             vol->v_state != G_RAID_VOLUME_S_SUBOPTIMAL) {
 203                 g_raid_iodone(bp, EIO);
 204                 return;
 205         }
 206         if (bp->bio_cmd == BIO_FLUSH) {
 207                 g_raid_tr_flush_common(tr, bp);
 208                 return;
 209         }
 210         if ((bp->bio_flags & BIO_UNMAPPED) != 0)
 211                 addr = NULL;
 212         else
 213                 addr = bp->bio_data;
 214         strip_size = vol->v_strip_size;
 215
 216         /* Stripe number. */
 217         nstripe = bp->bio_offset / strip_size;
 218         /* Start position in stripe. */
 219         start = bp->bio_offset % strip_size;
 220         /* Disk number. */
 221         no = nstripe % vol->v_disks_count;
 222         /* Stripe start position in disk. */
 223         offset = (nstripe / vol->v_disks_count) * strip_size;
 224         /* Length of data to operate. */
 225         remain = bp->bio_length;
 226
 227         bioq_init(&queue);
 228         do {
 229                 length = MIN(strip_size - start, remain);
 230                 cbp = g_clone_bio(bp);
 231                 if (cbp == NULL)
 232                         goto failure;
 233                 cbp->bio_offset = offset + start;
 234                 cbp->bio_length = length;
 235                 if ((bp->bio_flags & BIO_UNMAPPED) != 0 &&
 236                     bp->bio_cmd != BIO_DELETE) {
 237                         cbp->bio_ma_offset += (uintptr_t)addr;
 238                         cbp->bio_ma += cbp->bio_ma_offset / PAGE_SIZE;
 239                         cbp->bio_ma_offset %= PAGE_SIZE;
 240                         cbp->bio_ma_n = round_page(cbp->bio_ma_offset +
 241                             cbp->bio_length) / PAGE_SIZE;
 242                 } else
 243                         cbp->bio_data = addr;
 244                 cbp->bio_caller1 = &vol->v_subdisks[no];
 245                 bioq_insert_tail(&queue, cbp);
 246                 if (++no >= vol->v_disks_count) {
 247                         no = 0;
 248                         offset += strip_size;
 249                 }
 250                 remain -= length;
 251                 if (bp->bio_cmd != BIO_DELETE)
 252                         addr += length;
 253                 start = 0;
 254         } while (remain > 0);
 255         while ((cbp = bioq_takefirst(&queue)) != NULL) {
 256                 sd = cbp->bio_caller1;
 257                 cbp->bio_caller1 = NULL;
 258                 g_raid_subdisk_iostart(sd, cbp);
 259         }
 260         return;
 261 failure:
 262         while ((cbp = bioq_takefirst(&queue)) != NULL)
 263                 g_destroy_bio(cbp);
 264         if (bp->bio_error == 0)
 265                 bp->bio_error = ENOMEM;
 266         g_raid_iodone(bp, bp->bio_error);
 267 }
 268
 269 static int
 270 g_raid_tr_kerneldump_raid0(struct g_raid_tr_object *tr,
 271     void *virtual, vm_offset_t physical, off_t boffset, size_t blength)
 272 {
 273         struct g_raid_volume *vol;
 274         char *addr;
 275         off_t offset, start, length, nstripe, remain;
 276         u_int no, strip_size;
 277         int error;
 278
 279         vol = tr->tro_volume;
 280         if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL)
 281                 return (ENXIO);
 282         addr = virtual;
 283         strip_size = vol->v_strip_size;
 284
 285         /* Stripe number. */
 286         nstripe = boffset / strip_size;
 287         /* Start position in stripe. */
 288         start = boffset % strip_size;
 289         /* Disk number. */
 290         no = nstripe % vol->v_disks_count;
 291         /* Stripe tart position in disk. */
 292         offset = (nstripe / vol->v_disks_count) * strip_size;
 293         /* Length of data to operate. */
 294         remain = blength;
 295
 296         do {
 297                 length = MIN(strip_size - start, remain);
 298                 error = g_raid_subdisk_kerneldump(&vol->v_subdisks[no],
 299                     addr, 0, offset + start, length);
 300                 if (error != 0)
 301                         return (error);
 302                 if (++no >= vol->v_disks_count) {
 303                         no = 0;
 304                         offset += strip_size;
 305                 }
 306                 remain -= length;
 307                 addr += length;
 308                 start = 0;
 309         } while (remain > 0);
 310         return (0);
 311 }
 312
 313 static void
 314 g_raid_tr_iodone_raid0(struct g_raid_tr_object *tr,
 315     struct g_raid_subdisk *sd,struct bio *bp)
 316 {
 317         struct bio *pbp;
 318
 319         pbp = bp->bio_parent;
 320         if (pbp->bio_error == 0)
 321                 pbp->bio_error = bp->bio_error;
 322         g_destroy_bio(bp);
 323         pbp->bio_inbed++;
 324         if (pbp->bio_children == pbp->bio_inbed) {
 325                 pbp->bio_completed = pbp->bio_length;
 326                 g_raid_iodone(pbp, bp->bio_error);
 327         }
 328 }
 329
 330 static int
 331 g_raid_tr_free_raid0(struct g_raid_tr_object *tr)
 332 {
 333
 334         return (0);
 335 }
 336
 337 G_RAID_TR_DECLARE(raid0, "RAID0");