sys/vfs/hammer/hammer_undo.c

   1 /*
   2  * Copyright (c) 2008 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   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  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  * $DragonFly: src/sys/vfs/hammer/hammer_undo.c,v 1.20 2008/07/18 00:19:53 dillon Exp $
  35  */
  36
  37 /*
  38  * HAMMER undo - undo buffer/FIFO management.
  39  */
  40
  41 #include "hammer.h"
  42
  43 static int hammer_und_rb_compare(hammer_undo_t node1, hammer_undo_t node2);
  44
  45 RB_GENERATE2(hammer_und_rb_tree, hammer_undo, rb_node,
  46              hammer_und_rb_compare, hammer_off_t, offset);
  47
  48 /*
  49  * Convert a zone-3 undo offset into a zone-2 buffer offset.
  50  */
  51 hammer_off_t
  52 hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t zone3_off, int *errorp)
  53 {
  54         hammer_volume_t root_volume;
  55         hammer_blockmap_t undomap;
  56         hammer_off_t result_offset;
  57         int i;
  58
  59         KKASSERT((zone3_off & HAMMER_OFF_ZONE_MASK) == HAMMER_ZONE_UNDO);
  60         root_volume = hammer_get_root_volume(hmp, errorp);
  61         if (*errorp)
  62                 return(0);
  63         undomap = &hmp->blockmap[HAMMER_ZONE_UNDO_INDEX];
  64         KKASSERT(HAMMER_ZONE_DECODE(undomap->alloc_offset) == HAMMER_ZONE_UNDO_INDEX);
  65         KKASSERT (zone3_off < undomap->alloc_offset);
  66
  67         i = (zone3_off & HAMMER_OFF_SHORT_MASK) / HAMMER_LARGEBLOCK_SIZE;
  68         result_offset = root_volume->ondisk->vol0_undo_array[i] +
  69                         (zone3_off & HAMMER_LARGEBLOCK_MASK64);
  70
  71         hammer_rel_volume(root_volume, 0);
  72         return(result_offset);
  73 }
  74
  75 /*
  76  * Generate an UNDO record for the block of data at the specified zone1
  77  * or zone2 offset.
  78  *
  79  * The recovery code will execute UNDOs in reverse order, allowing overlaps.
  80  * All the UNDOs are executed together so if we already laid one down we
  81  * do not have to lay another one down for the same range.
  82  */
  83 int
  84 hammer_generate_undo(hammer_transaction_t trans, hammer_io_t io,
  85                      hammer_off_t zone_off, void *base, int len)
  86 {
  87         hammer_mount_t hmp;
  88         hammer_volume_t root_volume;
  89         hammer_blockmap_t undomap;
  90         hammer_buffer_t buffer = NULL;
  91         hammer_fifo_undo_t undo;
  92         hammer_fifo_tail_t tail;
  93         hammer_off_t next_offset;
  94         int error;
  95         int bytes;
  96
  97         hmp = trans->hmp;
  98
  99         /*
 100          * Enter the offset into our undo history.  If there is an existing
 101          * undo we do not have to generate a new one.
 102          */
 103         if (hammer_enter_undo_history(hmp, zone_off, len) == EALREADY)
 104                 return(0);
 105
 106         root_volume = trans->rootvol;
 107         undomap = &hmp->blockmap[HAMMER_ZONE_UNDO_INDEX];
 108
 109         /* no undo recursion */
 110         hammer_modify_volume(NULL, root_volume, NULL, 0);
 111
 112         hammer_lock_ex(&hmp->undo_lock);
 113 again:
 114         /*
 115          * Allocate space in the FIFO
 116          */
 117         bytes = ((len + HAMMER_HEAD_ALIGN_MASK) & ~HAMMER_HEAD_ALIGN_MASK) +
 118                 sizeof(struct hammer_fifo_undo) +
 119                 sizeof(struct hammer_fifo_tail);
 120         if (hammer_undo_space(trans) < bytes + HAMMER_BUFSIZE*2)
 121                 panic("hammer: insufficient undo FIFO space!");
 122
 123         next_offset = undomap->next_offset;
 124
 125         /*
 126          * Wrap next_offset
 127          */
 128         if (undomap->next_offset == undomap->alloc_offset) {
 129                 next_offset = HAMMER_ZONE_ENCODE(HAMMER_ZONE_UNDO_INDEX, 0);
 130                 undomap->next_offset = next_offset;
 131         }
 132
 133         /*
 134          * This is a tail-chasing FIFO, when we hit the start of a new
 135          * buffer we don't have to read it in.
 136          */
 137         if ((next_offset & HAMMER_BUFMASK) == 0)
 138                 undo = hammer_bnew(hmp, next_offset, &error, &buffer);
 139         else
 140                 undo = hammer_bread(hmp, next_offset, &error, &buffer);
 141         if (error)
 142                 goto done;
 143
 144         hammer_modify_buffer(NULL, buffer, NULL, 0);
 145
 146         KKASSERT(undomap->next_offset == next_offset);
 147
 148         /*
 149          * The FIFO entry would cross a buffer boundary, PAD to the end
 150          * of the buffer and try again.  Due to our data alignment, the
 151          * worst case (smallest) PAD record is 8 bytes.  PAD records only
 152          * populate the first 8 bytes of hammer_fifo_head and the tail may
 153          * be at the same offset as the head.
 154          */
 155         if ((next_offset ^ (next_offset + bytes)) & ~HAMMER_BUFMASK64) {
 156                 bytes = HAMMER_BUFSIZE - ((int)next_offset & HAMMER_BUFMASK);
 157                 tail = (void *)((char *)undo + bytes - sizeof(*tail));
 158                 if ((void *)undo != (void *)tail) {
 159                         tail->tail_signature = HAMMER_TAIL_SIGNATURE;
 160                         tail->tail_type = HAMMER_HEAD_TYPE_PAD;
 161                         tail->tail_size = bytes;
 162                 }
 163                 undo->head.hdr_signature = HAMMER_HEAD_SIGNATURE;
 164                 undo->head.hdr_type = HAMMER_HEAD_TYPE_PAD;
 165                 undo->head.hdr_size = bytes;
 166                 /* NO CRC */
 167                 undomap->next_offset += bytes;
 168                 hammer_modify_buffer_done(buffer);
 169                 goto again;
 170         }
 171         if (hammer_debug_general & 0x0080)
 172                 kprintf("undo %016llx %d %d\n", next_offset, bytes, len);
 173
 174         /*
 175          * We're good, create the entry.
 176          */
 177         undo->head.hdr_signature = HAMMER_HEAD_SIGNATURE;
 178         undo->head.hdr_type = HAMMER_HEAD_TYPE_UNDO;
 179         undo->head.hdr_size = bytes;
 180         undo->head.reserved01 = 0;
 181         undo->head.hdr_crc = 0;
 182         undo->undo_offset = zone_off;
 183         undo->undo_data_bytes = len;
 184         bcopy(base, undo + 1, len);
 185
 186         tail = (void *)((char *)undo + bytes - sizeof(*tail));
 187         tail->tail_signature = HAMMER_TAIL_SIGNATURE;
 188         tail->tail_type = HAMMER_HEAD_TYPE_UNDO;
 189         tail->tail_size = bytes;
 190
 191         KKASSERT(bytes >= sizeof(undo->head));
 192         undo->head.hdr_crc = crc32(undo, HAMMER_FIFO_HEAD_CRCOFF) ^
 193                              crc32(&undo->head + 1, bytes - sizeof(undo->head));
 194         undomap->next_offset += bytes;
 195
 196         hammer_modify_buffer_done(buffer);
 197 done:
 198         hammer_modify_volume_done(root_volume);
 199         hammer_unlock(&hmp->undo_lock);
 200
 201         if (buffer)
 202                 hammer_rel_buffer(buffer, 0);
 203         return(error);
 204 }
 205
 206 /*
 207  * UNDO HISTORY API
 208  *
 209  * It is not necessary to layout an undo record for the same address space
 210  * multiple times.  Maintain a cache of recent undo's.
 211  */
 212
 213 /*
 214  * Enter an undo into the history.  Return EALREADY if the request completely
 215  * covers a previous request.
 216  */
 217 int
 218 hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset, int bytes)
 219 {
 220         hammer_undo_t node;
 221         hammer_undo_t onode;
 222
 223         node = RB_LOOKUP(hammer_und_rb_tree, &hmp->rb_undo_root, offset);
 224         if (node) {
 225                 TAILQ_REMOVE(&hmp->undo_lru_list, node, lru_entry);
 226                 TAILQ_INSERT_TAIL(&hmp->undo_lru_list, node, lru_entry);
 227                 if (bytes <= node->bytes)
 228                         return(EALREADY);
 229                 node->bytes = bytes;
 230                 return(0);
 231         }
 232         if (hmp->undo_alloc != HAMMER_MAX_UNDOS) {
 233                 node = &hmp->undos[hmp->undo_alloc++];
 234         } else {
 235                 node = TAILQ_FIRST(&hmp->undo_lru_list);
 236                 TAILQ_REMOVE(&hmp->undo_lru_list, node, lru_entry);
 237                 RB_REMOVE(hammer_und_rb_tree, &hmp->rb_undo_root, node);
 238         }
 239         node->offset = offset;
 240         node->bytes = bytes;
 241         TAILQ_INSERT_TAIL(&hmp->undo_lru_list, node, lru_entry);
 242         onode = RB_INSERT(hammer_und_rb_tree, &hmp->rb_undo_root, node);
 243         KKASSERT(onode == NULL);
 244         return(0);
 245 }
 246
 247 void
 248 hammer_clear_undo_history(hammer_mount_t hmp)
 249 {
 250         RB_INIT(&hmp->rb_undo_root);
 251         TAILQ_INIT(&hmp->undo_lru_list);
 252         hmp->undo_alloc = 0;
 253 }
 254
 255 /*
 256  * Return how much of the undo FIFO has been used
 257  *
 258  * The calculation includes undo FIFO space still reserved from a previous
 259  * flush (because it will still be run on recovery if a crash occurs and
 260  * we can't overwrite it yet).
 261  */
 262 int64_t
 263 hammer_undo_used(hammer_transaction_t trans)
 264 {
 265         hammer_blockmap_t cundomap;
 266         hammer_blockmap_t dundomap;
 267         int64_t max_bytes;
 268         int64_t bytes;
 269
 270         cundomap = &trans->hmp->blockmap[HAMMER_ZONE_UNDO_INDEX];
 271         dundomap = &trans->rootvol->ondisk->
 272                                 vol0_blockmap[HAMMER_ZONE_UNDO_INDEX];
 273
 274         if (dundomap->first_offset <= cundomap->next_offset) {
 275                 bytes = cundomap->next_offset - dundomap->first_offset;
 276         } else {
 277                 bytes = cundomap->alloc_offset - dundomap->first_offset +
 278                         (cundomap->next_offset & HAMMER_OFF_LONG_MASK);
 279         }
 280         max_bytes = cundomap->alloc_offset & HAMMER_OFF_SHORT_MASK;
 281         KKASSERT(bytes <= max_bytes);
 282         return(bytes);
 283 }
 284
 285 /*
 286  * Return how much of the undo FIFO is available for new records.
 287  */
 288 int64_t
 289 hammer_undo_space(hammer_transaction_t trans)
 290 {
 291         hammer_blockmap_t rootmap;
 292         int64_t max_bytes;
 293
 294         rootmap = &trans->hmp->blockmap[HAMMER_ZONE_UNDO_INDEX];
 295         max_bytes = rootmap->alloc_offset & HAMMER_OFF_SHORT_MASK;
 296         return(max_bytes - hammer_undo_used(trans));
 297 }
 298
 299 int64_t
 300 hammer_undo_max(hammer_mount_t hmp)
 301 {
 302         hammer_blockmap_t rootmap;
 303         int64_t max_bytes;
 304
 305         rootmap = &hmp->blockmap[HAMMER_ZONE_UNDO_INDEX];
 306         max_bytes = rootmap->alloc_offset & HAMMER_OFF_SHORT_MASK;
 307
 308         return(max_bytes);
 309 }
 310
 311 static int
 312 hammer_und_rb_compare(hammer_undo_t node1, hammer_undo_t node2)
 313 {
 314         if (node1->offset < node2->offset)
 315                 return(-1);
 316         if (node1->offset > node2->offset)
 317                 return(1);
 318         return(0);
 319 }
 320