Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 1 | /* |
| 2 | * the_nilfs.c - the_nilfs shared structure. |
| 3 | * |
| 4 | * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation. |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or modify |
| 7 | * it under the terms of the GNU General Public License as published by |
| 8 | * the Free Software Foundation; either version 2 of the License, or |
| 9 | * (at your option) any later version. |
| 10 | * |
| 11 | * This program is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | * GNU General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU General Public License |
| 17 | * along with this program; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA |
| 19 | * |
| 20 | * Written by Ryusuke Konishi <ryusuke@osrg.net> |
| 21 | * |
| 22 | */ |
| 23 | |
| 24 | #include <linux/buffer_head.h> |
| 25 | #include <linux/slab.h> |
| 26 | #include <linux/blkdev.h> |
| 27 | #include <linux/backing-dev.h> |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 28 | #include <linux/crc32.h> |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 29 | #include "nilfs.h" |
| 30 | #include "segment.h" |
| 31 | #include "alloc.h" |
| 32 | #include "cpfile.h" |
| 33 | #include "sufile.h" |
| 34 | #include "dat.h" |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 35 | #include "segbuf.h" |
| 36 | |
Ryusuke Konishi | 33c8e57 | 2009-06-08 01:39:29 +0900 | [diff] [blame] | 37 | |
| 38 | static LIST_HEAD(nilfs_objects); |
| 39 | static DEFINE_SPINLOCK(nilfs_lock); |
| 40 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 41 | void nilfs_set_last_segment(struct the_nilfs *nilfs, |
| 42 | sector_t start_blocknr, u64 seq, __u64 cno) |
| 43 | { |
| 44 | spin_lock(&nilfs->ns_last_segment_lock); |
| 45 | nilfs->ns_last_pseg = start_blocknr; |
| 46 | nilfs->ns_last_seq = seq; |
| 47 | nilfs->ns_last_cno = cno; |
| 48 | spin_unlock(&nilfs->ns_last_segment_lock); |
| 49 | } |
| 50 | |
| 51 | /** |
| 52 | * alloc_nilfs - allocate the_nilfs structure |
| 53 | * @bdev: block device to which the_nilfs is related |
| 54 | * |
| 55 | * alloc_nilfs() allocates memory for the_nilfs and |
| 56 | * initializes its reference count and locks. |
| 57 | * |
| 58 | * Return Value: On success, pointer to the_nilfs is returned. |
| 59 | * On error, NULL is returned. |
| 60 | */ |
Ryusuke Konishi | 33c8e57 | 2009-06-08 01:39:29 +0900 | [diff] [blame] | 61 | static struct the_nilfs *alloc_nilfs(struct block_device *bdev) |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 62 | { |
| 63 | struct the_nilfs *nilfs; |
| 64 | |
| 65 | nilfs = kzalloc(sizeof(*nilfs), GFP_KERNEL); |
| 66 | if (!nilfs) |
| 67 | return NULL; |
| 68 | |
| 69 | nilfs->ns_bdev = bdev; |
| 70 | atomic_set(&nilfs->ns_count, 1); |
| 71 | atomic_set(&nilfs->ns_writer_refcount, -1); |
| 72 | atomic_set(&nilfs->ns_ndirtyblks, 0); |
| 73 | init_rwsem(&nilfs->ns_sem); |
Ryusuke Konishi | e59399d | 2009-06-08 01:39:32 +0900 | [diff] [blame] | 74 | init_rwsem(&nilfs->ns_super_sem); |
Ryusuke Konishi | aa7dfb8 | 2009-06-08 01:39:33 +0900 | [diff] [blame] | 75 | mutex_init(&nilfs->ns_mount_mutex); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 76 | mutex_init(&nilfs->ns_writer_mutex); |
Ryusuke Konishi | 33c8e57 | 2009-06-08 01:39:29 +0900 | [diff] [blame] | 77 | INIT_LIST_HEAD(&nilfs->ns_list); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 78 | INIT_LIST_HEAD(&nilfs->ns_supers); |
| 79 | spin_lock_init(&nilfs->ns_last_segment_lock); |
| 80 | nilfs->ns_gc_inodes_h = NULL; |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 81 | init_rwsem(&nilfs->ns_segctor_sem); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 82 | |
| 83 | return nilfs; |
| 84 | } |
| 85 | |
| 86 | /** |
Ryusuke Konishi | 33c8e57 | 2009-06-08 01:39:29 +0900 | [diff] [blame] | 87 | * find_or_create_nilfs - find or create nilfs object |
| 88 | * @bdev: block device to which the_nilfs is related |
| 89 | * |
| 90 | * find_nilfs() looks up an existent nilfs object created on the |
| 91 | * device and gets the reference count of the object. If no nilfs object |
| 92 | * is found on the device, a new nilfs object is allocated. |
| 93 | * |
| 94 | * Return Value: On success, pointer to the nilfs object is returned. |
| 95 | * On error, NULL is returned. |
| 96 | */ |
| 97 | struct the_nilfs *find_or_create_nilfs(struct block_device *bdev) |
| 98 | { |
| 99 | struct the_nilfs *nilfs, *new = NULL; |
| 100 | |
| 101 | retry: |
| 102 | spin_lock(&nilfs_lock); |
| 103 | list_for_each_entry(nilfs, &nilfs_objects, ns_list) { |
| 104 | if (nilfs->ns_bdev == bdev) { |
| 105 | get_nilfs(nilfs); |
| 106 | spin_unlock(&nilfs_lock); |
| 107 | if (new) |
| 108 | put_nilfs(new); |
| 109 | return nilfs; /* existing object */ |
| 110 | } |
| 111 | } |
| 112 | if (new) { |
| 113 | list_add_tail(&new->ns_list, &nilfs_objects); |
| 114 | spin_unlock(&nilfs_lock); |
| 115 | return new; /* new object */ |
| 116 | } |
| 117 | spin_unlock(&nilfs_lock); |
| 118 | |
| 119 | new = alloc_nilfs(bdev); |
| 120 | if (new) |
| 121 | goto retry; |
| 122 | return NULL; /* insufficient memory */ |
| 123 | } |
| 124 | |
| 125 | /** |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 126 | * put_nilfs - release a reference to the_nilfs |
| 127 | * @nilfs: the_nilfs structure to be released |
| 128 | * |
| 129 | * put_nilfs() decrements a reference counter of the_nilfs. |
| 130 | * If the reference count reaches zero, the_nilfs is freed. |
| 131 | */ |
| 132 | void put_nilfs(struct the_nilfs *nilfs) |
| 133 | { |
Ryusuke Konishi | 33c8e57 | 2009-06-08 01:39:29 +0900 | [diff] [blame] | 134 | spin_lock(&nilfs_lock); |
| 135 | if (!atomic_dec_and_test(&nilfs->ns_count)) { |
| 136 | spin_unlock(&nilfs_lock); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 137 | return; |
Ryusuke Konishi | 33c8e57 | 2009-06-08 01:39:29 +0900 | [diff] [blame] | 138 | } |
| 139 | list_del_init(&nilfs->ns_list); |
| 140 | spin_unlock(&nilfs_lock); |
| 141 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 142 | /* |
Ryusuke Konishi | 33c8e57 | 2009-06-08 01:39:29 +0900 | [diff] [blame] | 143 | * Increment of ns_count never occurs below because the caller |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 144 | * of get_nilfs() holds at least one reference to the_nilfs. |
| 145 | * Thus its exclusion control is not required here. |
| 146 | */ |
Ryusuke Konishi | 33c8e57 | 2009-06-08 01:39:29 +0900 | [diff] [blame] | 147 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 148 | might_sleep(); |
| 149 | if (nilfs_loaded(nilfs)) { |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 150 | nilfs_mdt_clear(nilfs->ns_sufile); |
| 151 | nilfs_mdt_destroy(nilfs->ns_sufile); |
| 152 | nilfs_mdt_clear(nilfs->ns_cpfile); |
| 153 | nilfs_mdt_destroy(nilfs->ns_cpfile); |
| 154 | nilfs_mdt_clear(nilfs->ns_dat); |
| 155 | nilfs_mdt_destroy(nilfs->ns_dat); |
| 156 | /* XXX: how and when to clear nilfs->ns_gc_dat? */ |
| 157 | nilfs_mdt_destroy(nilfs->ns_gc_dat); |
| 158 | } |
| 159 | if (nilfs_init(nilfs)) { |
| 160 | nilfs_destroy_gccache(nilfs); |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 161 | brelse(nilfs->ns_sbh[0]); |
| 162 | brelse(nilfs->ns_sbh[1]); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 163 | } |
| 164 | kfree(nilfs); |
| 165 | } |
| 166 | |
| 167 | static int nilfs_load_super_root(struct the_nilfs *nilfs, |
| 168 | struct nilfs_sb_info *sbi, sector_t sr_block) |
| 169 | { |
Ryusuke Konishi | c2698e5 | 2009-03-27 02:53:12 +0900 | [diff] [blame] | 170 | static struct lock_class_key dat_lock_key; |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 171 | struct buffer_head *bh_sr; |
| 172 | struct nilfs_super_root *raw_sr; |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 173 | struct nilfs_super_block **sbp = nilfs->ns_sbp; |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 174 | unsigned dat_entry_size, segment_usage_size, checkpoint_size; |
| 175 | unsigned inode_size; |
| 176 | int err; |
| 177 | |
| 178 | err = nilfs_read_super_root_block(sbi->s_super, sr_block, &bh_sr, 1); |
| 179 | if (unlikely(err)) |
| 180 | return err; |
| 181 | |
| 182 | down_read(&nilfs->ns_sem); |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 183 | dat_entry_size = le16_to_cpu(sbp[0]->s_dat_entry_size); |
| 184 | checkpoint_size = le16_to_cpu(sbp[0]->s_checkpoint_size); |
| 185 | segment_usage_size = le16_to_cpu(sbp[0]->s_segment_usage_size); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 186 | up_read(&nilfs->ns_sem); |
| 187 | |
| 188 | inode_size = nilfs->ns_inode_size; |
| 189 | |
| 190 | err = -ENOMEM; |
| 191 | nilfs->ns_dat = nilfs_mdt_new( |
| 192 | nilfs, NULL, NILFS_DAT_INO, NILFS_DAT_GFP); |
| 193 | if (unlikely(!nilfs->ns_dat)) |
| 194 | goto failed; |
| 195 | |
| 196 | nilfs->ns_gc_dat = nilfs_mdt_new( |
| 197 | nilfs, NULL, NILFS_DAT_INO, NILFS_DAT_GFP); |
| 198 | if (unlikely(!nilfs->ns_gc_dat)) |
| 199 | goto failed_dat; |
| 200 | |
| 201 | nilfs->ns_cpfile = nilfs_mdt_new( |
| 202 | nilfs, NULL, NILFS_CPFILE_INO, NILFS_CPFILE_GFP); |
| 203 | if (unlikely(!nilfs->ns_cpfile)) |
| 204 | goto failed_gc_dat; |
| 205 | |
| 206 | nilfs->ns_sufile = nilfs_mdt_new( |
| 207 | nilfs, NULL, NILFS_SUFILE_INO, NILFS_SUFILE_GFP); |
| 208 | if (unlikely(!nilfs->ns_sufile)) |
| 209 | goto failed_cpfile; |
| 210 | |
| 211 | err = nilfs_palloc_init_blockgroup(nilfs->ns_dat, dat_entry_size); |
| 212 | if (unlikely(err)) |
| 213 | goto failed_sufile; |
| 214 | |
| 215 | err = nilfs_palloc_init_blockgroup(nilfs->ns_gc_dat, dat_entry_size); |
| 216 | if (unlikely(err)) |
| 217 | goto failed_sufile; |
| 218 | |
Ryusuke Konishi | c2698e5 | 2009-03-27 02:53:12 +0900 | [diff] [blame] | 219 | lockdep_set_class(&NILFS_MDT(nilfs->ns_dat)->mi_sem, &dat_lock_key); |
| 220 | lockdep_set_class(&NILFS_MDT(nilfs->ns_gc_dat)->mi_sem, &dat_lock_key); |
| 221 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 222 | nilfs_mdt_set_shadow(nilfs->ns_dat, nilfs->ns_gc_dat); |
| 223 | nilfs_mdt_set_entry_size(nilfs->ns_cpfile, checkpoint_size, |
| 224 | sizeof(struct nilfs_cpfile_header)); |
| 225 | nilfs_mdt_set_entry_size(nilfs->ns_sufile, segment_usage_size, |
| 226 | sizeof(struct nilfs_sufile_header)); |
| 227 | |
| 228 | err = nilfs_mdt_read_inode_direct( |
| 229 | nilfs->ns_dat, bh_sr, NILFS_SR_DAT_OFFSET(inode_size)); |
| 230 | if (unlikely(err)) |
| 231 | goto failed_sufile; |
| 232 | |
| 233 | err = nilfs_mdt_read_inode_direct( |
| 234 | nilfs->ns_cpfile, bh_sr, NILFS_SR_CPFILE_OFFSET(inode_size)); |
| 235 | if (unlikely(err)) |
| 236 | goto failed_sufile; |
| 237 | |
| 238 | err = nilfs_mdt_read_inode_direct( |
| 239 | nilfs->ns_sufile, bh_sr, NILFS_SR_SUFILE_OFFSET(inode_size)); |
| 240 | if (unlikely(err)) |
| 241 | goto failed_sufile; |
| 242 | |
| 243 | raw_sr = (struct nilfs_super_root *)bh_sr->b_data; |
| 244 | nilfs->ns_nongc_ctime = le64_to_cpu(raw_sr->sr_nongc_ctime); |
| 245 | |
| 246 | failed: |
| 247 | brelse(bh_sr); |
| 248 | return err; |
| 249 | |
| 250 | failed_sufile: |
| 251 | nilfs_mdt_destroy(nilfs->ns_sufile); |
| 252 | |
| 253 | failed_cpfile: |
| 254 | nilfs_mdt_destroy(nilfs->ns_cpfile); |
| 255 | |
| 256 | failed_gc_dat: |
| 257 | nilfs_mdt_destroy(nilfs->ns_gc_dat); |
| 258 | |
| 259 | failed_dat: |
| 260 | nilfs_mdt_destroy(nilfs->ns_dat); |
| 261 | goto failed; |
| 262 | } |
| 263 | |
| 264 | static void nilfs_init_recovery_info(struct nilfs_recovery_info *ri) |
| 265 | { |
| 266 | memset(ri, 0, sizeof(*ri)); |
| 267 | INIT_LIST_HEAD(&ri->ri_used_segments); |
| 268 | } |
| 269 | |
| 270 | static void nilfs_clear_recovery_info(struct nilfs_recovery_info *ri) |
| 271 | { |
| 272 | nilfs_dispose_segment_list(&ri->ri_used_segments); |
| 273 | } |
| 274 | |
| 275 | /** |
| 276 | * load_nilfs - load and recover the nilfs |
| 277 | * @nilfs: the_nilfs structure to be released |
| 278 | * @sbi: nilfs_sb_info used to recover past segment |
| 279 | * |
| 280 | * load_nilfs() searches and load the latest super root, |
| 281 | * attaches the last segment, and does recovery if needed. |
| 282 | * The caller must call this exclusively for simultaneous mounts. |
| 283 | */ |
| 284 | int load_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi) |
| 285 | { |
| 286 | struct nilfs_recovery_info ri; |
| 287 | unsigned int s_flags = sbi->s_super->s_flags; |
| 288 | int really_read_only = bdev_read_only(nilfs->ns_bdev); |
| 289 | unsigned valid_fs; |
| 290 | int err = 0; |
| 291 | |
| 292 | nilfs_init_recovery_info(&ri); |
| 293 | |
| 294 | down_write(&nilfs->ns_sem); |
| 295 | valid_fs = (nilfs->ns_mount_state & NILFS_VALID_FS); |
| 296 | up_write(&nilfs->ns_sem); |
| 297 | |
| 298 | if (!valid_fs && (s_flags & MS_RDONLY)) { |
| 299 | printk(KERN_INFO "NILFS: INFO: recovery " |
| 300 | "required for readonly filesystem.\n"); |
| 301 | if (really_read_only) { |
| 302 | printk(KERN_ERR "NILFS: write access " |
| 303 | "unavailable, cannot proceed.\n"); |
| 304 | err = -EROFS; |
| 305 | goto failed; |
| 306 | } |
| 307 | printk(KERN_INFO "NILFS: write access will " |
| 308 | "be enabled during recovery.\n"); |
| 309 | sbi->s_super->s_flags &= ~MS_RDONLY; |
| 310 | } |
| 311 | |
| 312 | err = nilfs_search_super_root(nilfs, sbi, &ri); |
| 313 | if (unlikely(err)) { |
| 314 | printk(KERN_ERR "NILFS: error searching super root.\n"); |
| 315 | goto failed; |
| 316 | } |
| 317 | |
| 318 | err = nilfs_load_super_root(nilfs, sbi, ri.ri_super_root); |
| 319 | if (unlikely(err)) { |
| 320 | printk(KERN_ERR "NILFS: error loading super root.\n"); |
| 321 | goto failed; |
| 322 | } |
| 323 | |
| 324 | if (!valid_fs) { |
| 325 | err = nilfs_recover_logical_segments(nilfs, sbi, &ri); |
| 326 | if (unlikely(err)) { |
| 327 | nilfs_mdt_destroy(nilfs->ns_cpfile); |
| 328 | nilfs_mdt_destroy(nilfs->ns_sufile); |
| 329 | nilfs_mdt_destroy(nilfs->ns_dat); |
| 330 | goto failed; |
| 331 | } |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 332 | if (ri.ri_need_recovery == NILFS_RECOVERY_SR_UPDATED) |
| 333 | sbi->s_super->s_dirt = 1; |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 334 | } |
| 335 | |
| 336 | set_nilfs_loaded(nilfs); |
| 337 | |
| 338 | failed: |
| 339 | nilfs_clear_recovery_info(&ri); |
| 340 | sbi->s_super->s_flags = s_flags; |
| 341 | return err; |
| 342 | } |
| 343 | |
| 344 | static unsigned long long nilfs_max_size(unsigned int blkbits) |
| 345 | { |
| 346 | unsigned int max_bits; |
| 347 | unsigned long long res = MAX_LFS_FILESIZE; /* page cache limit */ |
| 348 | |
| 349 | max_bits = blkbits + NILFS_BMAP_KEY_BIT; /* bmap size limit */ |
| 350 | if (max_bits < 64) |
| 351 | res = min_t(unsigned long long, res, (1ULL << max_bits) - 1); |
| 352 | return res; |
| 353 | } |
| 354 | |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 355 | static int nilfs_store_disk_layout(struct the_nilfs *nilfs, |
| 356 | struct nilfs_super_block *sbp) |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 357 | { |
| 358 | if (le32_to_cpu(sbp->s_rev_level) != NILFS_CURRENT_REV) { |
| 359 | printk(KERN_ERR "NILFS: revision mismatch " |
| 360 | "(superblock rev.=%d.%d, current rev.=%d.%d). " |
| 361 | "Please check the version of mkfs.nilfs.\n", |
| 362 | le32_to_cpu(sbp->s_rev_level), |
| 363 | le16_to_cpu(sbp->s_minor_rev_level), |
| 364 | NILFS_CURRENT_REV, NILFS_MINOR_REV); |
| 365 | return -EINVAL; |
| 366 | } |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 367 | nilfs->ns_sbsize = le16_to_cpu(sbp->s_bytes); |
| 368 | if (nilfs->ns_sbsize > BLOCK_SIZE) |
| 369 | return -EINVAL; |
| 370 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 371 | nilfs->ns_inode_size = le16_to_cpu(sbp->s_inode_size); |
| 372 | nilfs->ns_first_ino = le32_to_cpu(sbp->s_first_ino); |
| 373 | |
| 374 | nilfs->ns_blocks_per_segment = le32_to_cpu(sbp->s_blocks_per_segment); |
| 375 | if (nilfs->ns_blocks_per_segment < NILFS_SEG_MIN_BLOCKS) { |
| 376 | printk(KERN_ERR "NILFS: too short segment. \n"); |
| 377 | return -EINVAL; |
| 378 | } |
| 379 | |
| 380 | nilfs->ns_first_data_block = le64_to_cpu(sbp->s_first_data_block); |
| 381 | nilfs->ns_nsegments = le64_to_cpu(sbp->s_nsegments); |
| 382 | nilfs->ns_r_segments_percentage = |
| 383 | le32_to_cpu(sbp->s_r_segments_percentage); |
| 384 | nilfs->ns_nrsvsegs = |
| 385 | max_t(unsigned long, NILFS_MIN_NRSVSEGS, |
| 386 | DIV_ROUND_UP(nilfs->ns_nsegments * |
| 387 | nilfs->ns_r_segments_percentage, 100)); |
| 388 | nilfs->ns_crc_seed = le32_to_cpu(sbp->s_crc_seed); |
| 389 | return 0; |
| 390 | } |
| 391 | |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 392 | static int nilfs_valid_sb(struct nilfs_super_block *sbp) |
| 393 | { |
| 394 | static unsigned char sum[4]; |
| 395 | const int sumoff = offsetof(struct nilfs_super_block, s_sum); |
| 396 | size_t bytes; |
| 397 | u32 crc; |
| 398 | |
| 399 | if (!sbp || le16_to_cpu(sbp->s_magic) != NILFS_SUPER_MAGIC) |
| 400 | return 0; |
| 401 | bytes = le16_to_cpu(sbp->s_bytes); |
| 402 | if (bytes > BLOCK_SIZE) |
| 403 | return 0; |
| 404 | crc = crc32_le(le32_to_cpu(sbp->s_crc_seed), (unsigned char *)sbp, |
| 405 | sumoff); |
| 406 | crc = crc32_le(crc, sum, 4); |
| 407 | crc = crc32_le(crc, (unsigned char *)sbp + sumoff + 4, |
| 408 | bytes - sumoff - 4); |
| 409 | return crc == le32_to_cpu(sbp->s_sum); |
| 410 | } |
| 411 | |
| 412 | static int nilfs_sb2_bad_offset(struct nilfs_super_block *sbp, u64 offset) |
| 413 | { |
| 414 | return offset < ((le64_to_cpu(sbp->s_nsegments) * |
| 415 | le32_to_cpu(sbp->s_blocks_per_segment)) << |
| 416 | (le32_to_cpu(sbp->s_log_block_size) + 10)); |
| 417 | } |
| 418 | |
| 419 | static void nilfs_release_super_block(struct the_nilfs *nilfs) |
| 420 | { |
| 421 | int i; |
| 422 | |
| 423 | for (i = 0; i < 2; i++) { |
| 424 | if (nilfs->ns_sbp[i]) { |
| 425 | brelse(nilfs->ns_sbh[i]); |
| 426 | nilfs->ns_sbh[i] = NULL; |
| 427 | nilfs->ns_sbp[i] = NULL; |
| 428 | } |
| 429 | } |
| 430 | } |
| 431 | |
| 432 | void nilfs_fall_back_super_block(struct the_nilfs *nilfs) |
| 433 | { |
| 434 | brelse(nilfs->ns_sbh[0]); |
| 435 | nilfs->ns_sbh[0] = nilfs->ns_sbh[1]; |
| 436 | nilfs->ns_sbp[0] = nilfs->ns_sbp[1]; |
| 437 | nilfs->ns_sbh[1] = NULL; |
| 438 | nilfs->ns_sbp[1] = NULL; |
| 439 | } |
| 440 | |
| 441 | void nilfs_swap_super_block(struct the_nilfs *nilfs) |
| 442 | { |
| 443 | struct buffer_head *tsbh = nilfs->ns_sbh[0]; |
| 444 | struct nilfs_super_block *tsbp = nilfs->ns_sbp[0]; |
| 445 | |
| 446 | nilfs->ns_sbh[0] = nilfs->ns_sbh[1]; |
| 447 | nilfs->ns_sbp[0] = nilfs->ns_sbp[1]; |
| 448 | nilfs->ns_sbh[1] = tsbh; |
| 449 | nilfs->ns_sbp[1] = tsbp; |
| 450 | } |
| 451 | |
| 452 | static int nilfs_load_super_block(struct the_nilfs *nilfs, |
| 453 | struct super_block *sb, int blocksize, |
| 454 | struct nilfs_super_block **sbpp) |
| 455 | { |
| 456 | struct nilfs_super_block **sbp = nilfs->ns_sbp; |
| 457 | struct buffer_head **sbh = nilfs->ns_sbh; |
| 458 | u64 sb2off = NILFS_SB2_OFFSET_BYTES(nilfs->ns_bdev->bd_inode->i_size); |
| 459 | int valid[2], swp = 0; |
| 460 | |
| 461 | sbp[0] = nilfs_read_super_block(sb, NILFS_SB_OFFSET_BYTES, blocksize, |
| 462 | &sbh[0]); |
| 463 | sbp[1] = nilfs_read_super_block(sb, sb2off, blocksize, &sbh[1]); |
| 464 | |
| 465 | if (!sbp[0]) { |
| 466 | if (!sbp[1]) { |
| 467 | printk(KERN_ERR "NILFS: unable to read superblock\n"); |
| 468 | return -EIO; |
| 469 | } |
| 470 | printk(KERN_WARNING |
| 471 | "NILFS warning: unable to read primary superblock\n"); |
| 472 | } else if (!sbp[1]) |
| 473 | printk(KERN_WARNING |
| 474 | "NILFS warning: unable to read secondary superblock\n"); |
| 475 | |
| 476 | valid[0] = nilfs_valid_sb(sbp[0]); |
| 477 | valid[1] = nilfs_valid_sb(sbp[1]); |
| 478 | swp = valid[1] && |
| 479 | (!valid[0] || |
| 480 | le64_to_cpu(sbp[1]->s_wtime) > le64_to_cpu(sbp[0]->s_wtime)); |
| 481 | |
| 482 | if (valid[swp] && nilfs_sb2_bad_offset(sbp[swp], sb2off)) { |
| 483 | brelse(sbh[1]); |
| 484 | sbh[1] = NULL; |
| 485 | sbp[1] = NULL; |
| 486 | swp = 0; |
| 487 | } |
| 488 | if (!valid[swp]) { |
| 489 | nilfs_release_super_block(nilfs); |
| 490 | printk(KERN_ERR "NILFS: Can't find nilfs on dev %s.\n", |
| 491 | sb->s_id); |
| 492 | return -EINVAL; |
| 493 | } |
| 494 | |
| 495 | if (swp) { |
| 496 | printk(KERN_WARNING "NILFS warning: broken superblock. " |
| 497 | "using spare superblock.\n"); |
| 498 | nilfs_swap_super_block(nilfs); |
| 499 | } |
| 500 | |
| 501 | nilfs->ns_sbwtime[0] = le64_to_cpu(sbp[0]->s_wtime); |
| 502 | nilfs->ns_sbwtime[1] = valid[!swp] ? le64_to_cpu(sbp[1]->s_wtime) : 0; |
| 503 | nilfs->ns_prot_seq = le64_to_cpu(sbp[valid[1] & !swp]->s_last_seq); |
| 504 | *sbpp = sbp[0]; |
| 505 | return 0; |
| 506 | } |
| 507 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 508 | /** |
| 509 | * init_nilfs - initialize a NILFS instance. |
| 510 | * @nilfs: the_nilfs structure |
| 511 | * @sbi: nilfs_sb_info |
| 512 | * @sb: super block |
| 513 | * @data: mount options |
| 514 | * |
| 515 | * init_nilfs() performs common initialization per block device (e.g. |
| 516 | * reading the super block, getting disk layout information, initializing |
| 517 | * shared fields in the_nilfs). It takes on some portion of the jobs |
| 518 | * typically done by a fill_super() routine. This division arises from |
| 519 | * the nature that multiple NILFS instances may be simultaneously |
| 520 | * mounted on a device. |
| 521 | * For multiple mounts on the same device, only the first mount |
| 522 | * invokes these tasks. |
| 523 | * |
| 524 | * Return Value: On success, 0 is returned. On error, a negative error |
| 525 | * code is returned. |
| 526 | */ |
| 527 | int init_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi, char *data) |
| 528 | { |
| 529 | struct super_block *sb = sbi->s_super; |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 530 | struct nilfs_super_block *sbp; |
| 531 | struct backing_dev_info *bdi; |
| 532 | int blocksize; |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 533 | int err; |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 534 | |
| 535 | down_write(&nilfs->ns_sem); |
| 536 | if (nilfs_init(nilfs)) { |
| 537 | /* Load values from existing the_nilfs */ |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 538 | sbp = nilfs->ns_sbp[0]; |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 539 | err = nilfs_store_magic_and_option(sb, sbp, data); |
| 540 | if (err) |
| 541 | goto out; |
| 542 | |
| 543 | blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size); |
| 544 | if (sb->s_blocksize != blocksize && |
| 545 | !sb_set_blocksize(sb, blocksize)) { |
| 546 | printk(KERN_ERR "NILFS: blocksize %d unfit to device\n", |
| 547 | blocksize); |
| 548 | err = -EINVAL; |
| 549 | } |
| 550 | sb->s_maxbytes = nilfs_max_size(sb->s_blocksize_bits); |
| 551 | goto out; |
| 552 | } |
| 553 | |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 554 | blocksize = sb_min_blocksize(sb, BLOCK_SIZE); |
| 555 | if (!blocksize) { |
| 556 | printk(KERN_ERR "NILFS: unable to set blocksize\n"); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 557 | err = -EINVAL; |
| 558 | goto out; |
| 559 | } |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 560 | err = nilfs_load_super_block(nilfs, sb, blocksize, &sbp); |
| 561 | if (err) |
| 562 | goto out; |
| 563 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 564 | err = nilfs_store_magic_and_option(sb, sbp, data); |
| 565 | if (err) |
| 566 | goto failed_sbh; |
| 567 | |
| 568 | blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size); |
| 569 | if (sb->s_blocksize != blocksize) { |
Martin K. Petersen | e1defc4 | 2009-05-22 17:17:49 -0400 | [diff] [blame] | 570 | int hw_blocksize = bdev_logical_block_size(sb->s_bdev); |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 571 | |
| 572 | if (blocksize < hw_blocksize) { |
| 573 | printk(KERN_ERR |
| 574 | "NILFS: blocksize %d too small for device " |
| 575 | "(sector-size = %d).\n", |
| 576 | blocksize, hw_blocksize); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 577 | err = -EINVAL; |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 578 | goto failed_sbh; |
| 579 | } |
| 580 | nilfs_release_super_block(nilfs); |
| 581 | sb_set_blocksize(sb, blocksize); |
| 582 | |
| 583 | err = nilfs_load_super_block(nilfs, sb, blocksize, &sbp); |
| 584 | if (err) |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 585 | goto out; |
| 586 | /* not failed_sbh; sbh is released automatically |
| 587 | when reloading fails. */ |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 588 | } |
| 589 | nilfs->ns_blocksize_bits = sb->s_blocksize_bits; |
| 590 | |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 591 | err = nilfs_store_disk_layout(nilfs, sbp); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 592 | if (err) |
| 593 | goto failed_sbh; |
| 594 | |
| 595 | sb->s_maxbytes = nilfs_max_size(sb->s_blocksize_bits); |
| 596 | |
| 597 | nilfs->ns_mount_state = le16_to_cpu(sbp->s_state); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 598 | |
| 599 | bdi = nilfs->ns_bdev->bd_inode_backing_dev_info; |
| 600 | if (!bdi) |
| 601 | bdi = nilfs->ns_bdev->bd_inode->i_mapping->backing_dev_info; |
| 602 | nilfs->ns_bdi = bdi ? : &default_backing_dev_info; |
| 603 | |
| 604 | /* Finding last segment */ |
| 605 | nilfs->ns_last_pseg = le64_to_cpu(sbp->s_last_pseg); |
| 606 | nilfs->ns_last_cno = le64_to_cpu(sbp->s_last_cno); |
| 607 | nilfs->ns_last_seq = le64_to_cpu(sbp->s_last_seq); |
| 608 | |
| 609 | nilfs->ns_seg_seq = nilfs->ns_last_seq; |
| 610 | nilfs->ns_segnum = |
| 611 | nilfs_get_segnum_of_block(nilfs, nilfs->ns_last_pseg); |
| 612 | nilfs->ns_cno = nilfs->ns_last_cno + 1; |
| 613 | if (nilfs->ns_segnum >= nilfs->ns_nsegments) { |
| 614 | printk(KERN_ERR "NILFS invalid last segment number.\n"); |
| 615 | err = -EINVAL; |
| 616 | goto failed_sbh; |
| 617 | } |
| 618 | /* Dummy values */ |
| 619 | nilfs->ns_free_segments_count = |
| 620 | nilfs->ns_nsegments - (nilfs->ns_segnum + 1); |
| 621 | |
| 622 | /* Initialize gcinode cache */ |
| 623 | err = nilfs_init_gccache(nilfs); |
| 624 | if (err) |
| 625 | goto failed_sbh; |
| 626 | |
| 627 | set_nilfs_init(nilfs); |
| 628 | err = 0; |
| 629 | out: |
| 630 | up_write(&nilfs->ns_sem); |
| 631 | return err; |
| 632 | |
| 633 | failed_sbh: |
Ryusuke Konishi | e339ad3 | 2009-04-06 19:01:59 -0700 | [diff] [blame] | 634 | nilfs_release_super_block(nilfs); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 635 | goto out; |
| 636 | } |
| 637 | |
| 638 | int nilfs_count_free_blocks(struct the_nilfs *nilfs, sector_t *nblocks) |
| 639 | { |
| 640 | struct inode *dat = nilfs_dat_inode(nilfs); |
| 641 | unsigned long ncleansegs; |
| 642 | int err; |
| 643 | |
| 644 | down_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ |
| 645 | err = nilfs_sufile_get_ncleansegs(nilfs->ns_sufile, &ncleansegs); |
| 646 | up_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ |
| 647 | if (likely(!err)) |
| 648 | *nblocks = (sector_t)ncleansegs * nilfs->ns_blocks_per_segment; |
| 649 | return err; |
| 650 | } |
| 651 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 652 | int nilfs_near_disk_full(struct the_nilfs *nilfs) |
| 653 | { |
| 654 | struct inode *sufile = nilfs->ns_sufile; |
| 655 | unsigned long ncleansegs, nincsegs; |
| 656 | int ret; |
| 657 | |
| 658 | ret = nilfs_sufile_get_ncleansegs(sufile, &ncleansegs); |
| 659 | if (likely(!ret)) { |
| 660 | nincsegs = atomic_read(&nilfs->ns_ndirtyblks) / |
| 661 | nilfs->ns_blocks_per_segment + 1; |
| 662 | if (ncleansegs <= nilfs->ns_nrsvsegs + nincsegs) |
| 663 | ret++; |
| 664 | } |
| 665 | return ret; |
| 666 | } |
| 667 | |
Ryusuke Konishi | 6dd4740 | 2009-06-08 01:39:31 +0900 | [diff] [blame] | 668 | /** |
| 669 | * nilfs_find_sbinfo - find existing nilfs_sb_info structure |
| 670 | * @nilfs: nilfs object |
| 671 | * @rw_mount: mount type (non-zero value for read/write mount) |
| 672 | * @cno: checkpoint number (zero for read-only mount) |
| 673 | * |
| 674 | * nilfs_find_sbinfo() returns the nilfs_sb_info structure which |
| 675 | * @rw_mount and @cno (in case of snapshots) matched. If no instance |
| 676 | * was found, NULL is returned. Although the super block instance can |
| 677 | * be unmounted after this function returns, the nilfs_sb_info struct |
| 678 | * is kept on memory until nilfs_put_sbinfo() is called. |
| 679 | */ |
| 680 | struct nilfs_sb_info *nilfs_find_sbinfo(struct the_nilfs *nilfs, |
| 681 | int rw_mount, __u64 cno) |
| 682 | { |
| 683 | struct nilfs_sb_info *sbi; |
| 684 | |
Ryusuke Konishi | e59399d | 2009-06-08 01:39:32 +0900 | [diff] [blame] | 685 | down_read(&nilfs->ns_super_sem); |
Ryusuke Konishi | 6dd4740 | 2009-06-08 01:39:31 +0900 | [diff] [blame] | 686 | /* |
| 687 | * The SNAPSHOT flag and sb->s_flags are supposed to be |
Ryusuke Konishi | e59399d | 2009-06-08 01:39:32 +0900 | [diff] [blame] | 688 | * protected with nilfs->ns_super_sem. |
Ryusuke Konishi | 6dd4740 | 2009-06-08 01:39:31 +0900 | [diff] [blame] | 689 | */ |
| 690 | sbi = nilfs->ns_current; |
| 691 | if (rw_mount) { |
| 692 | if (sbi && !(sbi->s_super->s_flags & MS_RDONLY)) |
| 693 | goto found; /* read/write mount */ |
| 694 | else |
| 695 | goto out; |
| 696 | } else if (cno == 0) { |
| 697 | if (sbi && (sbi->s_super->s_flags & MS_RDONLY)) |
| 698 | goto found; /* read-only mount */ |
| 699 | else |
| 700 | goto out; |
| 701 | } |
| 702 | |
| 703 | list_for_each_entry(sbi, &nilfs->ns_supers, s_list) { |
| 704 | if (nilfs_test_opt(sbi, SNAPSHOT) && |
| 705 | sbi->s_snapshot_cno == cno) |
| 706 | goto found; /* snapshot mount */ |
| 707 | } |
| 708 | out: |
Ryusuke Konishi | e59399d | 2009-06-08 01:39:32 +0900 | [diff] [blame] | 709 | up_read(&nilfs->ns_super_sem); |
Ryusuke Konishi | 6dd4740 | 2009-06-08 01:39:31 +0900 | [diff] [blame] | 710 | return NULL; |
| 711 | |
| 712 | found: |
| 713 | atomic_inc(&sbi->s_count); |
Ryusuke Konishi | e59399d | 2009-06-08 01:39:32 +0900 | [diff] [blame] | 714 | up_read(&nilfs->ns_super_sem); |
Ryusuke Konishi | 6dd4740 | 2009-06-08 01:39:31 +0900 | [diff] [blame] | 715 | return sbi; |
| 716 | } |
| 717 | |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 718 | int nilfs_checkpoint_is_mounted(struct the_nilfs *nilfs, __u64 cno, |
| 719 | int snapshot_mount) |
| 720 | { |
| 721 | struct nilfs_sb_info *sbi; |
| 722 | int ret = 0; |
| 723 | |
Ryusuke Konishi | e59399d | 2009-06-08 01:39:32 +0900 | [diff] [blame] | 724 | down_read(&nilfs->ns_super_sem); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 725 | if (cno == 0 || cno > nilfs->ns_cno) |
| 726 | goto out_unlock; |
| 727 | |
| 728 | list_for_each_entry(sbi, &nilfs->ns_supers, s_list) { |
| 729 | if (sbi->s_snapshot_cno == cno && |
| 730 | (!snapshot_mount || nilfs_test_opt(sbi, SNAPSHOT))) { |
| 731 | /* exclude read-only mounts */ |
| 732 | ret++; |
| 733 | break; |
| 734 | } |
| 735 | } |
| 736 | /* for protecting recent checkpoints */ |
| 737 | if (cno >= nilfs_last_cno(nilfs)) |
| 738 | ret++; |
| 739 | |
| 740 | out_unlock: |
Ryusuke Konishi | e59399d | 2009-06-08 01:39:32 +0900 | [diff] [blame] | 741 | up_read(&nilfs->ns_super_sem); |
Ryusuke Konishi | 8a9d219 | 2009-04-06 19:01:35 -0700 | [diff] [blame] | 742 | return ret; |
| 743 | } |