Stefan Roese | 9eefe2a | 2009-03-19 15:35:05 +0100 | [diff] [blame] | 1 | /* |
| 2 | * This file is part of UBIFS. |
| 3 | * |
| 4 | * Copyright (C) 2006-2008 Nokia Corporation. |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or modify it |
| 7 | * under the terms of the GNU General Public License version 2 as published by |
| 8 | * the Free Software Foundation. |
| 9 | * |
| 10 | * This program is distributed in the hope that it will be useful, but WITHOUT |
| 11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
| 12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for |
| 13 | * more details. |
| 14 | * |
| 15 | * You should have received a copy of the GNU General Public License along with |
| 16 | * this program; if not, write to the Free Software Foundation, Inc., 51 |
| 17 | * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA |
| 18 | * |
| 19 | * Author: Adrian Hunter |
| 20 | */ |
| 21 | |
| 22 | #include "ubifs.h" |
| 23 | |
| 24 | /* |
| 25 | * An orphan is an inode number whose inode node has been committed to the index |
| 26 | * with a link count of zero. That happens when an open file is deleted |
| 27 | * (unlinked) and then a commit is run. In the normal course of events the inode |
| 28 | * would be deleted when the file is closed. However in the case of an unclean |
| 29 | * unmount, orphans need to be accounted for. After an unclean unmount, the |
| 30 | * orphans' inodes must be deleted which means either scanning the entire index |
| 31 | * looking for them, or keeping a list on flash somewhere. This unit implements |
| 32 | * the latter approach. |
| 33 | * |
| 34 | * The orphan area is a fixed number of LEBs situated between the LPT area and |
| 35 | * the main area. The number of orphan area LEBs is specified when the file |
| 36 | * system is created. The minimum number is 1. The size of the orphan area |
| 37 | * should be so that it can hold the maximum number of orphans that are expected |
| 38 | * to ever exist at one time. |
| 39 | * |
| 40 | * The number of orphans that can fit in a LEB is: |
| 41 | * |
| 42 | * (c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64) |
| 43 | * |
| 44 | * For example: a 15872 byte LEB can fit 1980 orphans so 1 LEB may be enough. |
| 45 | * |
| 46 | * Orphans are accumulated in a rb-tree. When an inode's link count drops to |
| 47 | * zero, the inode number is added to the rb-tree. It is removed from the tree |
| 48 | * when the inode is deleted. Any new orphans that are in the orphan tree when |
| 49 | * the commit is run, are written to the orphan area in 1 or more orphan nodes. |
| 50 | * If the orphan area is full, it is consolidated to make space. There is |
| 51 | * always enough space because validation prevents the user from creating more |
| 52 | * than the maximum number of orphans allowed. |
| 53 | */ |
| 54 | |
| 55 | /** |
| 56 | * tot_avail_orphs - calculate total space. |
| 57 | * @c: UBIFS file-system description object |
| 58 | * |
| 59 | * This function returns the number of orphans that can be written in half |
| 60 | * the total space. That leaves half the space for adding new orphans. |
| 61 | */ |
| 62 | static int tot_avail_orphs(struct ubifs_info *c) |
| 63 | { |
| 64 | int avail_lebs, avail; |
| 65 | |
| 66 | avail_lebs = c->orph_lebs; |
| 67 | avail = avail_lebs * |
| 68 | ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64)); |
| 69 | return avail / 2; |
| 70 | } |
| 71 | |
| 72 | /** |
| 73 | * ubifs_clear_orphans - erase all LEBs used for orphans. |
| 74 | * @c: UBIFS file-system description object |
| 75 | * |
| 76 | * If recovery is not required, then the orphans from the previous session |
| 77 | * are not needed. This function locates the LEBs used to record |
| 78 | * orphans, and un-maps them. |
| 79 | */ |
| 80 | int ubifs_clear_orphans(struct ubifs_info *c) |
| 81 | { |
| 82 | int lnum, err; |
| 83 | |
| 84 | for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) { |
| 85 | err = ubifs_leb_unmap(c, lnum); |
| 86 | if (err) |
| 87 | return err; |
| 88 | } |
| 89 | c->ohead_lnum = c->orph_first; |
| 90 | c->ohead_offs = 0; |
| 91 | return 0; |
| 92 | } |
| 93 | |
| 94 | /** |
| 95 | * insert_dead_orphan - insert an orphan. |
| 96 | * @c: UBIFS file-system description object |
| 97 | * @inum: orphan inode number |
| 98 | * |
| 99 | * This function is a helper to the 'do_kill_orphans()' function. The orphan |
| 100 | * must be kept until the next commit, so it is added to the rb-tree and the |
| 101 | * deletion list. |
| 102 | */ |
| 103 | static int insert_dead_orphan(struct ubifs_info *c, ino_t inum) |
| 104 | { |
| 105 | struct ubifs_orphan *orphan, *o; |
| 106 | struct rb_node **p, *parent = NULL; |
| 107 | |
| 108 | orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_KERNEL); |
| 109 | if (!orphan) |
| 110 | return -ENOMEM; |
| 111 | orphan->inum = inum; |
| 112 | |
| 113 | p = &c->orph_tree.rb_node; |
| 114 | while (*p) { |
| 115 | parent = *p; |
| 116 | o = rb_entry(parent, struct ubifs_orphan, rb); |
| 117 | if (inum < o->inum) |
| 118 | p = &(*p)->rb_left; |
| 119 | else if (inum > o->inum) |
| 120 | p = &(*p)->rb_right; |
| 121 | else { |
| 122 | /* Already added - no problem */ |
| 123 | kfree(orphan); |
| 124 | return 0; |
| 125 | } |
| 126 | } |
| 127 | c->tot_orphans += 1; |
| 128 | rb_link_node(&orphan->rb, parent, p); |
| 129 | rb_insert_color(&orphan->rb, &c->orph_tree); |
| 130 | list_add_tail(&orphan->list, &c->orph_list); |
| 131 | orphan->dnext = c->orph_dnext; |
| 132 | c->orph_dnext = orphan; |
| 133 | dbg_mnt("ino %lu, new %d, tot %d", (unsigned long)inum, |
| 134 | c->new_orphans, c->tot_orphans); |
| 135 | return 0; |
| 136 | } |
| 137 | |
| 138 | /** |
| 139 | * do_kill_orphans - remove orphan inodes from the index. |
| 140 | * @c: UBIFS file-system description object |
| 141 | * @sleb: scanned LEB |
| 142 | * @last_cmt_no: cmt_no of last orphan node read is passed and returned here |
| 143 | * @outofdate: whether the LEB is out of date is returned here |
| 144 | * @last_flagged: whether the end orphan node is encountered |
| 145 | * |
| 146 | * This function is a helper to the 'kill_orphans()' function. It goes through |
| 147 | * every orphan node in a LEB and for every inode number recorded, removes |
| 148 | * all keys for that inode from the TNC. |
| 149 | */ |
| 150 | static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb, |
| 151 | unsigned long long *last_cmt_no, int *outofdate, |
| 152 | int *last_flagged) |
| 153 | { |
| 154 | struct ubifs_scan_node *snod; |
| 155 | struct ubifs_orph_node *orph; |
| 156 | unsigned long long cmt_no; |
| 157 | ino_t inum; |
| 158 | int i, n, err, first = 1; |
| 159 | |
| 160 | list_for_each_entry(snod, &sleb->nodes, list) { |
| 161 | if (snod->type != UBIFS_ORPH_NODE) { |
| 162 | ubifs_err("invalid node type %d in orphan area at " |
| 163 | "%d:%d", snod->type, sleb->lnum, snod->offs); |
| 164 | dbg_dump_node(c, snod->node); |
| 165 | return -EINVAL; |
| 166 | } |
| 167 | |
| 168 | orph = snod->node; |
| 169 | |
| 170 | /* Check commit number */ |
| 171 | cmt_no = le64_to_cpu(orph->cmt_no) & LLONG_MAX; |
| 172 | /* |
| 173 | * The commit number on the master node may be less, because |
| 174 | * of a failed commit. If there are several failed commits in a |
| 175 | * row, the commit number written on orphan nodes will continue |
| 176 | * to increase (because the commit number is adjusted here) even |
| 177 | * though the commit number on the master node stays the same |
| 178 | * because the master node has not been re-written. |
| 179 | */ |
| 180 | if (cmt_no > c->cmt_no) |
| 181 | c->cmt_no = cmt_no; |
| 182 | if (cmt_no < *last_cmt_no && *last_flagged) { |
| 183 | /* |
| 184 | * The last orphan node had a higher commit number and |
| 185 | * was flagged as the last written for that commit |
| 186 | * number. That makes this orphan node, out of date. |
| 187 | */ |
| 188 | if (!first) { |
| 189 | ubifs_err("out of order commit number %llu in " |
| 190 | "orphan node at %d:%d", |
| 191 | cmt_no, sleb->lnum, snod->offs); |
| 192 | dbg_dump_node(c, snod->node); |
| 193 | return -EINVAL; |
| 194 | } |
| 195 | dbg_rcvry("out of date LEB %d", sleb->lnum); |
| 196 | *outofdate = 1; |
| 197 | return 0; |
| 198 | } |
| 199 | |
| 200 | if (first) |
| 201 | first = 0; |
| 202 | |
| 203 | n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3; |
| 204 | for (i = 0; i < n; i++) { |
| 205 | inum = le64_to_cpu(orph->inos[i]); |
| 206 | dbg_rcvry("deleting orphaned inode %lu", |
| 207 | (unsigned long)inum); |
| 208 | err = ubifs_tnc_remove_ino(c, inum); |
| 209 | if (err) |
| 210 | return err; |
| 211 | err = insert_dead_orphan(c, inum); |
| 212 | if (err) |
| 213 | return err; |
| 214 | } |
| 215 | |
| 216 | *last_cmt_no = cmt_no; |
| 217 | if (le64_to_cpu(orph->cmt_no) & (1ULL << 63)) { |
| 218 | dbg_rcvry("last orph node for commit %llu at %d:%d", |
| 219 | cmt_no, sleb->lnum, snod->offs); |
| 220 | *last_flagged = 1; |
| 221 | } else |
| 222 | *last_flagged = 0; |
| 223 | } |
| 224 | |
| 225 | return 0; |
| 226 | } |
| 227 | |
| 228 | /** |
| 229 | * kill_orphans - remove all orphan inodes from the index. |
| 230 | * @c: UBIFS file-system description object |
| 231 | * |
| 232 | * If recovery is required, then orphan inodes recorded during the previous |
| 233 | * session (which ended with an unclean unmount) must be deleted from the index. |
| 234 | * This is done by updating the TNC, but since the index is not updated until |
| 235 | * the next commit, the LEBs where the orphan information is recorded are not |
| 236 | * erased until the next commit. |
| 237 | */ |
| 238 | static int kill_orphans(struct ubifs_info *c) |
| 239 | { |
| 240 | unsigned long long last_cmt_no = 0; |
| 241 | int lnum, err = 0, outofdate = 0, last_flagged = 0; |
| 242 | |
| 243 | c->ohead_lnum = c->orph_first; |
| 244 | c->ohead_offs = 0; |
| 245 | /* Check no-orphans flag and skip this if no orphans */ |
| 246 | if (c->no_orphs) { |
| 247 | dbg_rcvry("no orphans"); |
| 248 | return 0; |
| 249 | } |
| 250 | /* |
| 251 | * Orph nodes always start at c->orph_first and are written to each |
| 252 | * successive LEB in turn. Generally unused LEBs will have been unmapped |
| 253 | * but may contain out of date orphan nodes if the unmap didn't go |
| 254 | * through. In addition, the last orphan node written for each commit is |
| 255 | * marked (top bit of orph->cmt_no is set to 1). It is possible that |
| 256 | * there are orphan nodes from the next commit (i.e. the commit did not |
| 257 | * complete successfully). In that case, no orphans will have been lost |
| 258 | * due to the way that orphans are written, and any orphans added will |
| 259 | * be valid orphans anyway and so can be deleted. |
| 260 | */ |
| 261 | for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) { |
| 262 | struct ubifs_scan_leb *sleb; |
| 263 | |
| 264 | dbg_rcvry("LEB %d", lnum); |
| 265 | sleb = ubifs_scan(c, lnum, 0, c->sbuf); |
| 266 | if (IS_ERR(sleb)) { |
| 267 | sleb = ubifs_recover_leb(c, lnum, 0, c->sbuf, 0); |
| 268 | if (IS_ERR(sleb)) { |
| 269 | err = PTR_ERR(sleb); |
| 270 | break; |
| 271 | } |
| 272 | } |
| 273 | err = do_kill_orphans(c, sleb, &last_cmt_no, &outofdate, |
| 274 | &last_flagged); |
| 275 | if (err || outofdate) { |
| 276 | ubifs_scan_destroy(sleb); |
| 277 | break; |
| 278 | } |
| 279 | if (sleb->endpt) { |
| 280 | c->ohead_lnum = lnum; |
| 281 | c->ohead_offs = sleb->endpt; |
| 282 | } |
| 283 | ubifs_scan_destroy(sleb); |
| 284 | } |
| 285 | return err; |
| 286 | } |
| 287 | |
| 288 | /** |
| 289 | * ubifs_mount_orphans - delete orphan inodes and erase LEBs that recorded them. |
| 290 | * @c: UBIFS file-system description object |
| 291 | * @unclean: indicates recovery from unclean unmount |
| 292 | * @read_only: indicates read only mount |
| 293 | * |
| 294 | * This function is called when mounting to erase orphans from the previous |
| 295 | * session. If UBIFS was not unmounted cleanly, then the inodes recorded as |
| 296 | * orphans are deleted. |
| 297 | */ |
| 298 | int ubifs_mount_orphans(struct ubifs_info *c, int unclean, int read_only) |
| 299 | { |
| 300 | int err = 0; |
| 301 | |
| 302 | c->max_orphans = tot_avail_orphs(c); |
| 303 | |
| 304 | if (!read_only) { |
| 305 | c->orph_buf = vmalloc(c->leb_size); |
| 306 | if (!c->orph_buf) |
| 307 | return -ENOMEM; |
| 308 | } |
| 309 | |
| 310 | if (unclean) |
| 311 | err = kill_orphans(c); |
| 312 | else if (!read_only) |
| 313 | err = ubifs_clear_orphans(c); |
| 314 | |
| 315 | return err; |
| 316 | } |