| 1 | // SPDX-License-Identifier: GPL-2.0 |
| 2 | /* |
| 3 | * Copyright (C) 2007 Oracle. All rights reserved. |
| 4 | */ |
| 5 | |
| 6 | #include <linux/blkdev.h> |
| 7 | #include <linux/module.h> |
| 8 | #include <linux/fs.h> |
| 9 | #include <linux/pagemap.h> |
| 10 | #include <linux/highmem.h> |
| 11 | #include <linux/time.h> |
| 12 | #include <linux/init.h> |
| 13 | #include <linux/seq_file.h> |
| 14 | #include <linux/string.h> |
| 15 | #include <linux/backing-dev.h> |
| 16 | #include <linux/mount.h> |
| 17 | #include <linux/writeback.h> |
| 18 | #include <linux/statfs.h> |
| 19 | #include <linux/compat.h> |
| 20 | #include <linux/parser.h> |
| 21 | #include <linux/ctype.h> |
| 22 | #include <linux/namei.h> |
| 23 | #include <linux/miscdevice.h> |
| 24 | #include <linux/magic.h> |
| 25 | #include <linux/slab.h> |
| 26 | #include <linux/ratelimit.h> |
| 27 | #include <linux/crc32c.h> |
| 28 | #include <linux/btrfs.h> |
| 29 | #include <linux/security.h> |
| 30 | #include <linux/fs_parser.h> |
| 31 | #include "messages.h" |
| 32 | #include "delayed-inode.h" |
| 33 | #include "ctree.h" |
| 34 | #include "disk-io.h" |
| 35 | #include "transaction.h" |
| 36 | #include "btrfs_inode.h" |
| 37 | #include "props.h" |
| 38 | #include "xattr.h" |
| 39 | #include "bio.h" |
| 40 | #include "export.h" |
| 41 | #include "compression.h" |
| 42 | #include "dev-replace.h" |
| 43 | #include "free-space-cache.h" |
| 44 | #include "backref.h" |
| 45 | #include "space-info.h" |
| 46 | #include "sysfs.h" |
| 47 | #include "zoned.h" |
| 48 | #include "tests/btrfs-tests.h" |
| 49 | #include "block-group.h" |
| 50 | #include "discard.h" |
| 51 | #include "qgroup.h" |
| 52 | #include "raid56.h" |
| 53 | #include "fs.h" |
| 54 | #include "accessors.h" |
| 55 | #include "defrag.h" |
| 56 | #include "dir-item.h" |
| 57 | #include "ioctl.h" |
| 58 | #include "scrub.h" |
| 59 | #include "verity.h" |
| 60 | #include "super.h" |
| 61 | #include "extent-tree.h" |
| 62 | #define CREATE_TRACE_POINTS |
| 63 | #include <trace/events/btrfs.h> |
| 64 | |
| 65 | static const struct super_operations btrfs_super_ops; |
| 66 | static struct file_system_type btrfs_fs_type; |
| 67 | |
| 68 | static void btrfs_put_super(struct super_block *sb) |
| 69 | { |
| 70 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 71 | |
| 72 | btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid); |
| 73 | close_ctree(fs_info); |
| 74 | } |
| 75 | |
| 76 | /* Store the mount options related information. */ |
| 77 | struct btrfs_fs_context { |
| 78 | char *subvol_name; |
| 79 | u64 subvol_objectid; |
| 80 | u64 max_inline; |
| 81 | u32 commit_interval; |
| 82 | u32 metadata_ratio; |
| 83 | u32 thread_pool_size; |
| 84 | unsigned long mount_opt; |
| 85 | unsigned long compress_type:4; |
| 86 | unsigned int compress_level; |
| 87 | refcount_t refs; |
| 88 | }; |
| 89 | |
| 90 | enum { |
| 91 | Opt_acl, |
| 92 | Opt_clear_cache, |
| 93 | Opt_commit_interval, |
| 94 | Opt_compress, |
| 95 | Opt_compress_force, |
| 96 | Opt_compress_force_type, |
| 97 | Opt_compress_type, |
| 98 | Opt_degraded, |
| 99 | Opt_device, |
| 100 | Opt_fatal_errors, |
| 101 | Opt_flushoncommit, |
| 102 | Opt_max_inline, |
| 103 | Opt_barrier, |
| 104 | Opt_datacow, |
| 105 | Opt_datasum, |
| 106 | Opt_defrag, |
| 107 | Opt_discard, |
| 108 | Opt_discard_mode, |
| 109 | Opt_ratio, |
| 110 | Opt_rescan_uuid_tree, |
| 111 | Opt_skip_balance, |
| 112 | Opt_space_cache, |
| 113 | Opt_space_cache_version, |
| 114 | Opt_ssd, |
| 115 | Opt_ssd_spread, |
| 116 | Opt_subvol, |
| 117 | Opt_subvol_empty, |
| 118 | Opt_subvolid, |
| 119 | Opt_thread_pool, |
| 120 | Opt_treelog, |
| 121 | Opt_user_subvol_rm_allowed, |
| 122 | Opt_norecovery, |
| 123 | |
| 124 | /* Rescue options */ |
| 125 | Opt_rescue, |
| 126 | Opt_usebackuproot, |
| 127 | Opt_nologreplay, |
| 128 | Opt_ignorebadroots, |
| 129 | Opt_ignoredatacsums, |
| 130 | Opt_rescue_all, |
| 131 | |
| 132 | /* Debugging options */ |
| 133 | Opt_enospc_debug, |
| 134 | #ifdef CONFIG_BTRFS_DEBUG |
| 135 | Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all, |
| 136 | #endif |
| 137 | #ifdef CONFIG_BTRFS_FS_REF_VERIFY |
| 138 | Opt_ref_verify, |
| 139 | #endif |
| 140 | Opt_err, |
| 141 | }; |
| 142 | |
| 143 | enum { |
| 144 | Opt_fatal_errors_panic, |
| 145 | Opt_fatal_errors_bug, |
| 146 | }; |
| 147 | |
| 148 | static const struct constant_table btrfs_parameter_fatal_errors[] = { |
| 149 | { "panic", Opt_fatal_errors_panic }, |
| 150 | { "bug", Opt_fatal_errors_bug }, |
| 151 | {} |
| 152 | }; |
| 153 | |
| 154 | enum { |
| 155 | Opt_discard_sync, |
| 156 | Opt_discard_async, |
| 157 | }; |
| 158 | |
| 159 | static const struct constant_table btrfs_parameter_discard[] = { |
| 160 | { "sync", Opt_discard_sync }, |
| 161 | { "async", Opt_discard_async }, |
| 162 | {} |
| 163 | }; |
| 164 | |
| 165 | enum { |
| 166 | Opt_space_cache_v1, |
| 167 | Opt_space_cache_v2, |
| 168 | }; |
| 169 | |
| 170 | static const struct constant_table btrfs_parameter_space_cache[] = { |
| 171 | { "v1", Opt_space_cache_v1 }, |
| 172 | { "v2", Opt_space_cache_v2 }, |
| 173 | {} |
| 174 | }; |
| 175 | |
| 176 | enum { |
| 177 | Opt_rescue_usebackuproot, |
| 178 | Opt_rescue_nologreplay, |
| 179 | Opt_rescue_ignorebadroots, |
| 180 | Opt_rescue_ignoredatacsums, |
| 181 | Opt_rescue_parameter_all, |
| 182 | }; |
| 183 | |
| 184 | static const struct constant_table btrfs_parameter_rescue[] = { |
| 185 | { "usebackuproot", Opt_rescue_usebackuproot }, |
| 186 | { "nologreplay", Opt_rescue_nologreplay }, |
| 187 | { "ignorebadroots", Opt_rescue_ignorebadroots }, |
| 188 | { "ibadroots", Opt_rescue_ignorebadroots }, |
| 189 | { "ignoredatacsums", Opt_rescue_ignoredatacsums }, |
| 190 | { "idatacsums", Opt_rescue_ignoredatacsums }, |
| 191 | { "all", Opt_rescue_parameter_all }, |
| 192 | {} |
| 193 | }; |
| 194 | |
| 195 | #ifdef CONFIG_BTRFS_DEBUG |
| 196 | enum { |
| 197 | Opt_fragment_parameter_data, |
| 198 | Opt_fragment_parameter_metadata, |
| 199 | Opt_fragment_parameter_all, |
| 200 | }; |
| 201 | |
| 202 | static const struct constant_table btrfs_parameter_fragment[] = { |
| 203 | { "data", Opt_fragment_parameter_data }, |
| 204 | { "metadata", Opt_fragment_parameter_metadata }, |
| 205 | { "all", Opt_fragment_parameter_all }, |
| 206 | {} |
| 207 | }; |
| 208 | #endif |
| 209 | |
| 210 | static const struct fs_parameter_spec btrfs_fs_parameters[] = { |
| 211 | fsparam_flag_no("acl", Opt_acl), |
| 212 | fsparam_flag_no("autodefrag", Opt_defrag), |
| 213 | fsparam_flag_no("barrier", Opt_barrier), |
| 214 | fsparam_flag("clear_cache", Opt_clear_cache), |
| 215 | fsparam_u32("commit", Opt_commit_interval), |
| 216 | fsparam_flag("compress", Opt_compress), |
| 217 | fsparam_string("compress", Opt_compress_type), |
| 218 | fsparam_flag("compress-force", Opt_compress_force), |
| 219 | fsparam_string("compress-force", Opt_compress_force_type), |
| 220 | fsparam_flag_no("datacow", Opt_datacow), |
| 221 | fsparam_flag_no("datasum", Opt_datasum), |
| 222 | fsparam_flag("degraded", Opt_degraded), |
| 223 | fsparam_string("device", Opt_device), |
| 224 | fsparam_flag_no("discard", Opt_discard), |
| 225 | fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard), |
| 226 | fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors), |
| 227 | fsparam_flag_no("flushoncommit", Opt_flushoncommit), |
| 228 | fsparam_string("max_inline", Opt_max_inline), |
| 229 | fsparam_u32("metadata_ratio", Opt_ratio), |
| 230 | fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree), |
| 231 | fsparam_flag("skip_balance", Opt_skip_balance), |
| 232 | fsparam_flag_no("space_cache", Opt_space_cache), |
| 233 | fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache), |
| 234 | fsparam_flag_no("ssd", Opt_ssd), |
| 235 | fsparam_flag_no("ssd_spread", Opt_ssd_spread), |
| 236 | fsparam_string("subvol", Opt_subvol), |
| 237 | fsparam_flag("subvol=", Opt_subvol_empty), |
| 238 | fsparam_u64("subvolid", Opt_subvolid), |
| 239 | fsparam_u32("thread_pool", Opt_thread_pool), |
| 240 | fsparam_flag_no("treelog", Opt_treelog), |
| 241 | fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed), |
| 242 | |
| 243 | /* Rescue options. */ |
| 244 | fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue), |
| 245 | /* Deprecated, with alias rescue=nologreplay */ |
| 246 | __fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL), |
| 247 | /* Deprecated, with alias rescue=usebackuproot */ |
| 248 | __fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL), |
| 249 | /* For compatibility only, alias for "rescue=nologreplay". */ |
| 250 | fsparam_flag("norecovery", Opt_norecovery), |
| 251 | |
| 252 | /* Debugging options. */ |
| 253 | fsparam_flag_no("enospc_debug", Opt_enospc_debug), |
| 254 | #ifdef CONFIG_BTRFS_DEBUG |
| 255 | fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment), |
| 256 | #endif |
| 257 | #ifdef CONFIG_BTRFS_FS_REF_VERIFY |
| 258 | fsparam_flag("ref_verify", Opt_ref_verify), |
| 259 | #endif |
| 260 | {} |
| 261 | }; |
| 262 | |
| 263 | /* No support for restricting writes to btrfs devices yet... */ |
| 264 | static inline blk_mode_t btrfs_open_mode(struct fs_context *fc) |
| 265 | { |
| 266 | return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES; |
| 267 | } |
| 268 | |
| 269 | static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param) |
| 270 | { |
| 271 | struct btrfs_fs_context *ctx = fc->fs_private; |
| 272 | struct fs_parse_result result; |
| 273 | int opt; |
| 274 | |
| 275 | opt = fs_parse(fc, btrfs_fs_parameters, param, &result); |
| 276 | if (opt < 0) |
| 277 | return opt; |
| 278 | |
| 279 | switch (opt) { |
| 280 | case Opt_degraded: |
| 281 | btrfs_set_opt(ctx->mount_opt, DEGRADED); |
| 282 | break; |
| 283 | case Opt_subvol_empty: |
| 284 | /* |
| 285 | * This exists because we used to allow it on accident, so we're |
| 286 | * keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow |
| 287 | * empty subvol= again"). |
| 288 | */ |
| 289 | break; |
| 290 | case Opt_subvol: |
| 291 | kfree(ctx->subvol_name); |
| 292 | ctx->subvol_name = kstrdup(param->string, GFP_KERNEL); |
| 293 | if (!ctx->subvol_name) |
| 294 | return -ENOMEM; |
| 295 | break; |
| 296 | case Opt_subvolid: |
| 297 | ctx->subvol_objectid = result.uint_64; |
| 298 | |
| 299 | /* subvolid=0 means give me the original fs_tree. */ |
| 300 | if (!ctx->subvol_objectid) |
| 301 | ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID; |
| 302 | break; |
| 303 | case Opt_device: { |
| 304 | struct btrfs_device *device; |
| 305 | blk_mode_t mode = btrfs_open_mode(fc); |
| 306 | |
| 307 | mutex_lock(&uuid_mutex); |
| 308 | device = btrfs_scan_one_device(param->string, mode, false); |
| 309 | mutex_unlock(&uuid_mutex); |
| 310 | if (IS_ERR(device)) |
| 311 | return PTR_ERR(device); |
| 312 | break; |
| 313 | } |
| 314 | case Opt_datasum: |
| 315 | if (result.negated) { |
| 316 | btrfs_set_opt(ctx->mount_opt, NODATASUM); |
| 317 | } else { |
| 318 | btrfs_clear_opt(ctx->mount_opt, NODATACOW); |
| 319 | btrfs_clear_opt(ctx->mount_opt, NODATASUM); |
| 320 | } |
| 321 | break; |
| 322 | case Opt_datacow: |
| 323 | if (result.negated) { |
| 324 | btrfs_clear_opt(ctx->mount_opt, COMPRESS); |
| 325 | btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS); |
| 326 | btrfs_set_opt(ctx->mount_opt, NODATACOW); |
| 327 | btrfs_set_opt(ctx->mount_opt, NODATASUM); |
| 328 | } else { |
| 329 | btrfs_clear_opt(ctx->mount_opt, NODATACOW); |
| 330 | } |
| 331 | break; |
| 332 | case Opt_compress_force: |
| 333 | case Opt_compress_force_type: |
| 334 | btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS); |
| 335 | fallthrough; |
| 336 | case Opt_compress: |
| 337 | case Opt_compress_type: |
| 338 | if (opt == Opt_compress || opt == Opt_compress_force) { |
| 339 | ctx->compress_type = BTRFS_COMPRESS_ZLIB; |
| 340 | ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL; |
| 341 | btrfs_set_opt(ctx->mount_opt, COMPRESS); |
| 342 | btrfs_clear_opt(ctx->mount_opt, NODATACOW); |
| 343 | btrfs_clear_opt(ctx->mount_opt, NODATASUM); |
| 344 | } else if (strncmp(param->string, "zlib", 4) == 0) { |
| 345 | ctx->compress_type = BTRFS_COMPRESS_ZLIB; |
| 346 | ctx->compress_level = |
| 347 | btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB, |
| 348 | param->string + 4); |
| 349 | btrfs_set_opt(ctx->mount_opt, COMPRESS); |
| 350 | btrfs_clear_opt(ctx->mount_opt, NODATACOW); |
| 351 | btrfs_clear_opt(ctx->mount_opt, NODATASUM); |
| 352 | } else if (strncmp(param->string, "lzo", 3) == 0) { |
| 353 | ctx->compress_type = BTRFS_COMPRESS_LZO; |
| 354 | ctx->compress_level = 0; |
| 355 | btrfs_set_opt(ctx->mount_opt, COMPRESS); |
| 356 | btrfs_clear_opt(ctx->mount_opt, NODATACOW); |
| 357 | btrfs_clear_opt(ctx->mount_opt, NODATASUM); |
| 358 | } else if (strncmp(param->string, "zstd", 4) == 0) { |
| 359 | ctx->compress_type = BTRFS_COMPRESS_ZSTD; |
| 360 | ctx->compress_level = |
| 361 | btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD, |
| 362 | param->string + 4); |
| 363 | btrfs_set_opt(ctx->mount_opt, COMPRESS); |
| 364 | btrfs_clear_opt(ctx->mount_opt, NODATACOW); |
| 365 | btrfs_clear_opt(ctx->mount_opt, NODATASUM); |
| 366 | } else if (strncmp(param->string, "no", 2) == 0) { |
| 367 | ctx->compress_level = 0; |
| 368 | ctx->compress_type = 0; |
| 369 | btrfs_clear_opt(ctx->mount_opt, COMPRESS); |
| 370 | btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS); |
| 371 | } else { |
| 372 | btrfs_err(NULL, "unrecognized compression value %s", |
| 373 | param->string); |
| 374 | return -EINVAL; |
| 375 | } |
| 376 | break; |
| 377 | case Opt_ssd: |
| 378 | if (result.negated) { |
| 379 | btrfs_set_opt(ctx->mount_opt, NOSSD); |
| 380 | btrfs_clear_opt(ctx->mount_opt, SSD); |
| 381 | btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD); |
| 382 | } else { |
| 383 | btrfs_set_opt(ctx->mount_opt, SSD); |
| 384 | btrfs_clear_opt(ctx->mount_opt, NOSSD); |
| 385 | } |
| 386 | break; |
| 387 | case Opt_ssd_spread: |
| 388 | if (result.negated) { |
| 389 | btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD); |
| 390 | } else { |
| 391 | btrfs_set_opt(ctx->mount_opt, SSD); |
| 392 | btrfs_set_opt(ctx->mount_opt, SSD_SPREAD); |
| 393 | btrfs_clear_opt(ctx->mount_opt, NOSSD); |
| 394 | } |
| 395 | break; |
| 396 | case Opt_barrier: |
| 397 | if (result.negated) |
| 398 | btrfs_set_opt(ctx->mount_opt, NOBARRIER); |
| 399 | else |
| 400 | btrfs_clear_opt(ctx->mount_opt, NOBARRIER); |
| 401 | break; |
| 402 | case Opt_thread_pool: |
| 403 | if (result.uint_32 == 0) { |
| 404 | btrfs_err(NULL, "invalid value 0 for thread_pool"); |
| 405 | return -EINVAL; |
| 406 | } |
| 407 | ctx->thread_pool_size = result.uint_32; |
| 408 | break; |
| 409 | case Opt_max_inline: |
| 410 | ctx->max_inline = memparse(param->string, NULL); |
| 411 | break; |
| 412 | case Opt_acl: |
| 413 | if (result.negated) { |
| 414 | fc->sb_flags &= ~SB_POSIXACL; |
| 415 | } else { |
| 416 | #ifdef CONFIG_BTRFS_FS_POSIX_ACL |
| 417 | fc->sb_flags |= SB_POSIXACL; |
| 418 | #else |
| 419 | btrfs_err(NULL, "support for ACL not compiled in"); |
| 420 | return -EINVAL; |
| 421 | #endif |
| 422 | } |
| 423 | /* |
| 424 | * VFS limits the ability to toggle ACL on and off via remount, |
| 425 | * despite every file system allowing this. This seems to be |
| 426 | * an oversight since we all do, but it'll fail if we're |
| 427 | * remounting. So don't set the mask here, we'll check it in |
| 428 | * btrfs_reconfigure and do the toggling ourselves. |
| 429 | */ |
| 430 | if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE) |
| 431 | fc->sb_flags_mask |= SB_POSIXACL; |
| 432 | break; |
| 433 | case Opt_treelog: |
| 434 | if (result.negated) |
| 435 | btrfs_set_opt(ctx->mount_opt, NOTREELOG); |
| 436 | else |
| 437 | btrfs_clear_opt(ctx->mount_opt, NOTREELOG); |
| 438 | break; |
| 439 | case Opt_nologreplay: |
| 440 | btrfs_warn(NULL, |
| 441 | "'nologreplay' is deprecated, use 'rescue=nologreplay' instead"); |
| 442 | btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); |
| 443 | break; |
| 444 | case Opt_norecovery: |
| 445 | btrfs_info(NULL, |
| 446 | "'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'"); |
| 447 | btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); |
| 448 | break; |
| 449 | case Opt_flushoncommit: |
| 450 | if (result.negated) |
| 451 | btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT); |
| 452 | else |
| 453 | btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT); |
| 454 | break; |
| 455 | case Opt_ratio: |
| 456 | ctx->metadata_ratio = result.uint_32; |
| 457 | break; |
| 458 | case Opt_discard: |
| 459 | if (result.negated) { |
| 460 | btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC); |
| 461 | btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC); |
| 462 | btrfs_set_opt(ctx->mount_opt, NODISCARD); |
| 463 | } else { |
| 464 | btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC); |
| 465 | btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC); |
| 466 | } |
| 467 | break; |
| 468 | case Opt_discard_mode: |
| 469 | switch (result.uint_32) { |
| 470 | case Opt_discard_sync: |
| 471 | btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC); |
| 472 | btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC); |
| 473 | break; |
| 474 | case Opt_discard_async: |
| 475 | btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC); |
| 476 | btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC); |
| 477 | break; |
| 478 | default: |
| 479 | btrfs_err(NULL, "unrecognized discard mode value %s", |
| 480 | param->key); |
| 481 | return -EINVAL; |
| 482 | } |
| 483 | btrfs_clear_opt(ctx->mount_opt, NODISCARD); |
| 484 | break; |
| 485 | case Opt_space_cache: |
| 486 | if (result.negated) { |
| 487 | btrfs_set_opt(ctx->mount_opt, NOSPACECACHE); |
| 488 | btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE); |
| 489 | btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE); |
| 490 | } else { |
| 491 | btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE); |
| 492 | btrfs_set_opt(ctx->mount_opt, SPACE_CACHE); |
| 493 | } |
| 494 | break; |
| 495 | case Opt_space_cache_version: |
| 496 | switch (result.uint_32) { |
| 497 | case Opt_space_cache_v1: |
| 498 | btrfs_set_opt(ctx->mount_opt, SPACE_CACHE); |
| 499 | btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE); |
| 500 | break; |
| 501 | case Opt_space_cache_v2: |
| 502 | btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE); |
| 503 | btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE); |
| 504 | break; |
| 505 | default: |
| 506 | btrfs_err(NULL, "unrecognized space_cache value %s", |
| 507 | param->key); |
| 508 | return -EINVAL; |
| 509 | } |
| 510 | break; |
| 511 | case Opt_rescan_uuid_tree: |
| 512 | btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE); |
| 513 | break; |
| 514 | case Opt_clear_cache: |
| 515 | btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE); |
| 516 | break; |
| 517 | case Opt_user_subvol_rm_allowed: |
| 518 | btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED); |
| 519 | break; |
| 520 | case Opt_enospc_debug: |
| 521 | if (result.negated) |
| 522 | btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG); |
| 523 | else |
| 524 | btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG); |
| 525 | break; |
| 526 | case Opt_defrag: |
| 527 | if (result.negated) |
| 528 | btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG); |
| 529 | else |
| 530 | btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG); |
| 531 | break; |
| 532 | case Opt_usebackuproot: |
| 533 | btrfs_warn(NULL, |
| 534 | "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead"); |
| 535 | btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT); |
| 536 | |
| 537 | /* If we're loading the backup roots we can't trust the space cache. */ |
| 538 | btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE); |
| 539 | break; |
| 540 | case Opt_skip_balance: |
| 541 | btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE); |
| 542 | break; |
| 543 | case Opt_fatal_errors: |
| 544 | switch (result.uint_32) { |
| 545 | case Opt_fatal_errors_panic: |
| 546 | btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR); |
| 547 | break; |
| 548 | case Opt_fatal_errors_bug: |
| 549 | btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR); |
| 550 | break; |
| 551 | default: |
| 552 | btrfs_err(NULL, "unrecognized fatal_errors value %s", |
| 553 | param->key); |
| 554 | return -EINVAL; |
| 555 | } |
| 556 | break; |
| 557 | case Opt_commit_interval: |
| 558 | ctx->commit_interval = result.uint_32; |
| 559 | if (ctx->commit_interval == 0) |
| 560 | ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
| 561 | break; |
| 562 | case Opt_rescue: |
| 563 | switch (result.uint_32) { |
| 564 | case Opt_rescue_usebackuproot: |
| 565 | btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT); |
| 566 | break; |
| 567 | case Opt_rescue_nologreplay: |
| 568 | btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); |
| 569 | break; |
| 570 | case Opt_rescue_ignorebadroots: |
| 571 | btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS); |
| 572 | break; |
| 573 | case Opt_rescue_ignoredatacsums: |
| 574 | btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS); |
| 575 | break; |
| 576 | case Opt_rescue_parameter_all: |
| 577 | btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS); |
| 578 | btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS); |
| 579 | btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); |
| 580 | break; |
| 581 | default: |
| 582 | btrfs_info(NULL, "unrecognized rescue option '%s'", |
| 583 | param->key); |
| 584 | return -EINVAL; |
| 585 | } |
| 586 | break; |
| 587 | #ifdef CONFIG_BTRFS_DEBUG |
| 588 | case Opt_fragment: |
| 589 | switch (result.uint_32) { |
| 590 | case Opt_fragment_parameter_all: |
| 591 | btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA); |
| 592 | btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA); |
| 593 | break; |
| 594 | case Opt_fragment_parameter_metadata: |
| 595 | btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA); |
| 596 | break; |
| 597 | case Opt_fragment_parameter_data: |
| 598 | btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA); |
| 599 | break; |
| 600 | default: |
| 601 | btrfs_info(NULL, "unrecognized fragment option '%s'", |
| 602 | param->key); |
| 603 | return -EINVAL; |
| 604 | } |
| 605 | break; |
| 606 | #endif |
| 607 | #ifdef CONFIG_BTRFS_FS_REF_VERIFY |
| 608 | case Opt_ref_verify: |
| 609 | btrfs_set_opt(ctx->mount_opt, REF_VERIFY); |
| 610 | break; |
| 611 | #endif |
| 612 | default: |
| 613 | btrfs_err(NULL, "unrecognized mount option '%s'", param->key); |
| 614 | return -EINVAL; |
| 615 | } |
| 616 | |
| 617 | return 0; |
| 618 | } |
| 619 | |
| 620 | /* |
| 621 | * Some options only have meaning at mount time and shouldn't persist across |
| 622 | * remounts, or be displayed. Clear these at the end of mount and remount code |
| 623 | * paths. |
| 624 | */ |
| 625 | static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) |
| 626 | { |
| 627 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); |
| 628 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
| 629 | btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE); |
| 630 | } |
| 631 | |
| 632 | static bool check_ro_option(struct btrfs_fs_info *fs_info, |
| 633 | unsigned long mount_opt, unsigned long opt, |
| 634 | const char *opt_name) |
| 635 | { |
| 636 | if (mount_opt & opt) { |
| 637 | btrfs_err(fs_info, "%s must be used with ro mount option", |
| 638 | opt_name); |
| 639 | return true; |
| 640 | } |
| 641 | return false; |
| 642 | } |
| 643 | |
| 644 | bool btrfs_check_options(struct btrfs_fs_info *info, unsigned long *mount_opt, |
| 645 | unsigned long flags) |
| 646 | { |
| 647 | bool ret = true; |
| 648 | |
| 649 | if (!(flags & SB_RDONLY) && |
| 650 | (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") || |
| 651 | check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") || |
| 652 | check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))) |
| 653 | ret = false; |
| 654 | |
| 655 | if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) && |
| 656 | !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) && |
| 657 | !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) { |
| 658 | btrfs_err(info, "cannot disable free-space-tree"); |
| 659 | ret = false; |
| 660 | } |
| 661 | if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) && |
| 662 | !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) { |
| 663 | btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature"); |
| 664 | ret = false; |
| 665 | } |
| 666 | |
| 667 | if (btrfs_check_mountopts_zoned(info, mount_opt)) |
| 668 | ret = false; |
| 669 | |
| 670 | if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) { |
| 671 | if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) |
| 672 | btrfs_info(info, "disk space caching is enabled"); |
| 673 | if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) |
| 674 | btrfs_info(info, "using free-space-tree"); |
| 675 | } |
| 676 | |
| 677 | return ret; |
| 678 | } |
| 679 | |
| 680 | /* |
| 681 | * This is subtle, we only call this during open_ctree(). We need to pre-load |
| 682 | * the mount options with the on-disk settings. Before the new mount API took |
| 683 | * effect we would do this on mount and remount. With the new mount API we'll |
| 684 | * only do this on the initial mount. |
| 685 | * |
| 686 | * This isn't a change in behavior, because we're using the current state of the |
| 687 | * file system to set the current mount options. If you mounted with special |
| 688 | * options to disable these features and then remounted we wouldn't revert the |
| 689 | * settings, because mounting without these features cleared the on-disk |
| 690 | * settings, so this being called on re-mount is not needed. |
| 691 | */ |
| 692 | void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info) |
| 693 | { |
| 694 | if (fs_info->sectorsize < PAGE_SIZE) { |
| 695 | btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); |
| 696 | if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { |
| 697 | btrfs_info(fs_info, |
| 698 | "forcing free space tree for sector size %u with page size %lu", |
| 699 | fs_info->sectorsize, PAGE_SIZE); |
| 700 | btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE); |
| 701 | } |
| 702 | } |
| 703 | |
| 704 | /* |
| 705 | * At this point our mount options are populated, so we only mess with |
| 706 | * these settings if we don't have any settings already. |
| 707 | */ |
| 708 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE)) |
| 709 | return; |
| 710 | |
| 711 | if (btrfs_is_zoned(fs_info) && |
| 712 | btrfs_free_space_cache_v1_active(fs_info)) { |
| 713 | btrfs_info(fs_info, "zoned: clearing existing space cache"); |
| 714 | btrfs_set_super_cache_generation(fs_info->super_copy, 0); |
| 715 | return; |
| 716 | } |
| 717 | |
| 718 | if (btrfs_test_opt(fs_info, SPACE_CACHE)) |
| 719 | return; |
| 720 | |
| 721 | if (btrfs_test_opt(fs_info, NOSPACECACHE)) |
| 722 | return; |
| 723 | |
| 724 | /* |
| 725 | * At this point we don't have explicit options set by the user, set |
| 726 | * them ourselves based on the state of the file system. |
| 727 | */ |
| 728 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) |
| 729 | btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE); |
| 730 | else if (btrfs_free_space_cache_v1_active(fs_info)) |
| 731 | btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE); |
| 732 | } |
| 733 | |
| 734 | static void set_device_specific_options(struct btrfs_fs_info *fs_info) |
| 735 | { |
| 736 | if (!btrfs_test_opt(fs_info, NOSSD) && |
| 737 | !fs_info->fs_devices->rotating) |
| 738 | btrfs_set_opt(fs_info->mount_opt, SSD); |
| 739 | |
| 740 | /* |
| 741 | * For devices supporting discard turn on discard=async automatically, |
| 742 | * unless it's already set or disabled. This could be turned off by |
| 743 | * nodiscard for the same mount. |
| 744 | * |
| 745 | * The zoned mode piggy backs on the discard functionality for |
| 746 | * resetting a zone. There is no reason to delay the zone reset as it is |
| 747 | * fast enough. So, do not enable async discard for zoned mode. |
| 748 | */ |
| 749 | if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || |
| 750 | btrfs_test_opt(fs_info, DISCARD_ASYNC) || |
| 751 | btrfs_test_opt(fs_info, NODISCARD)) && |
| 752 | fs_info->fs_devices->discardable && |
| 753 | !btrfs_is_zoned(fs_info)) |
| 754 | btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC); |
| 755 | } |
| 756 | |
| 757 | char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info, |
| 758 | u64 subvol_objectid) |
| 759 | { |
| 760 | struct btrfs_root *root = fs_info->tree_root; |
| 761 | struct btrfs_root *fs_root = NULL; |
| 762 | struct btrfs_root_ref *root_ref; |
| 763 | struct btrfs_inode_ref *inode_ref; |
| 764 | struct btrfs_key key; |
| 765 | struct btrfs_path *path = NULL; |
| 766 | char *name = NULL, *ptr; |
| 767 | u64 dirid; |
| 768 | int len; |
| 769 | int ret; |
| 770 | |
| 771 | path = btrfs_alloc_path(); |
| 772 | if (!path) { |
| 773 | ret = -ENOMEM; |
| 774 | goto err; |
| 775 | } |
| 776 | |
| 777 | name = kmalloc(PATH_MAX, GFP_KERNEL); |
| 778 | if (!name) { |
| 779 | ret = -ENOMEM; |
| 780 | goto err; |
| 781 | } |
| 782 | ptr = name + PATH_MAX - 1; |
| 783 | ptr[0] = '\0'; |
| 784 | |
| 785 | /* |
| 786 | * Walk up the subvolume trees in the tree of tree roots by root |
| 787 | * backrefs until we hit the top-level subvolume. |
| 788 | */ |
| 789 | while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) { |
| 790 | key.objectid = subvol_objectid; |
| 791 | key.type = BTRFS_ROOT_BACKREF_KEY; |
| 792 | key.offset = (u64)-1; |
| 793 | |
| 794 | ret = btrfs_search_backwards(root, &key, path); |
| 795 | if (ret < 0) { |
| 796 | goto err; |
| 797 | } else if (ret > 0) { |
| 798 | ret = -ENOENT; |
| 799 | goto err; |
| 800 | } |
| 801 | |
| 802 | subvol_objectid = key.offset; |
| 803 | |
| 804 | root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
| 805 | struct btrfs_root_ref); |
| 806 | len = btrfs_root_ref_name_len(path->nodes[0], root_ref); |
| 807 | ptr -= len + 1; |
| 808 | if (ptr < name) { |
| 809 | ret = -ENAMETOOLONG; |
| 810 | goto err; |
| 811 | } |
| 812 | read_extent_buffer(path->nodes[0], ptr + 1, |
| 813 | (unsigned long)(root_ref + 1), len); |
| 814 | ptr[0] = '/'; |
| 815 | dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref); |
| 816 | btrfs_release_path(path); |
| 817 | |
| 818 | fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true); |
| 819 | if (IS_ERR(fs_root)) { |
| 820 | ret = PTR_ERR(fs_root); |
| 821 | fs_root = NULL; |
| 822 | goto err; |
| 823 | } |
| 824 | |
| 825 | /* |
| 826 | * Walk up the filesystem tree by inode refs until we hit the |
| 827 | * root directory. |
| 828 | */ |
| 829 | while (dirid != BTRFS_FIRST_FREE_OBJECTID) { |
| 830 | key.objectid = dirid; |
| 831 | key.type = BTRFS_INODE_REF_KEY; |
| 832 | key.offset = (u64)-1; |
| 833 | |
| 834 | ret = btrfs_search_backwards(fs_root, &key, path); |
| 835 | if (ret < 0) { |
| 836 | goto err; |
| 837 | } else if (ret > 0) { |
| 838 | ret = -ENOENT; |
| 839 | goto err; |
| 840 | } |
| 841 | |
| 842 | dirid = key.offset; |
| 843 | |
| 844 | inode_ref = btrfs_item_ptr(path->nodes[0], |
| 845 | path->slots[0], |
| 846 | struct btrfs_inode_ref); |
| 847 | len = btrfs_inode_ref_name_len(path->nodes[0], |
| 848 | inode_ref); |
| 849 | ptr -= len + 1; |
| 850 | if (ptr < name) { |
| 851 | ret = -ENAMETOOLONG; |
| 852 | goto err; |
| 853 | } |
| 854 | read_extent_buffer(path->nodes[0], ptr + 1, |
| 855 | (unsigned long)(inode_ref + 1), len); |
| 856 | ptr[0] = '/'; |
| 857 | btrfs_release_path(path); |
| 858 | } |
| 859 | btrfs_put_root(fs_root); |
| 860 | fs_root = NULL; |
| 861 | } |
| 862 | |
| 863 | btrfs_free_path(path); |
| 864 | if (ptr == name + PATH_MAX - 1) { |
| 865 | name[0] = '/'; |
| 866 | name[1] = '\0'; |
| 867 | } else { |
| 868 | memmove(name, ptr, name + PATH_MAX - ptr); |
| 869 | } |
| 870 | return name; |
| 871 | |
| 872 | err: |
| 873 | btrfs_put_root(fs_root); |
| 874 | btrfs_free_path(path); |
| 875 | kfree(name); |
| 876 | return ERR_PTR(ret); |
| 877 | } |
| 878 | |
| 879 | static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid) |
| 880 | { |
| 881 | struct btrfs_root *root = fs_info->tree_root; |
| 882 | struct btrfs_dir_item *di; |
| 883 | struct btrfs_path *path; |
| 884 | struct btrfs_key location; |
| 885 | struct fscrypt_str name = FSTR_INIT("default", 7); |
| 886 | u64 dir_id; |
| 887 | |
| 888 | path = btrfs_alloc_path(); |
| 889 | if (!path) |
| 890 | return -ENOMEM; |
| 891 | |
| 892 | /* |
| 893 | * Find the "default" dir item which points to the root item that we |
| 894 | * will mount by default if we haven't been given a specific subvolume |
| 895 | * to mount. |
| 896 | */ |
| 897 | dir_id = btrfs_super_root_dir(fs_info->super_copy); |
| 898 | di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0); |
| 899 | if (IS_ERR(di)) { |
| 900 | btrfs_free_path(path); |
| 901 | return PTR_ERR(di); |
| 902 | } |
| 903 | if (!di) { |
| 904 | /* |
| 905 | * Ok the default dir item isn't there. This is weird since |
| 906 | * it's always been there, but don't freak out, just try and |
| 907 | * mount the top-level subvolume. |
| 908 | */ |
| 909 | btrfs_free_path(path); |
| 910 | *objectid = BTRFS_FS_TREE_OBJECTID; |
| 911 | return 0; |
| 912 | } |
| 913 | |
| 914 | btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); |
| 915 | btrfs_free_path(path); |
| 916 | *objectid = location.objectid; |
| 917 | return 0; |
| 918 | } |
| 919 | |
| 920 | static int btrfs_fill_super(struct super_block *sb, |
| 921 | struct btrfs_fs_devices *fs_devices, |
| 922 | void *data) |
| 923 | { |
| 924 | struct inode *inode; |
| 925 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 926 | int err; |
| 927 | |
| 928 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
| 929 | sb->s_magic = BTRFS_SUPER_MAGIC; |
| 930 | sb->s_op = &btrfs_super_ops; |
| 931 | sb->s_d_op = &btrfs_dentry_operations; |
| 932 | sb->s_export_op = &btrfs_export_ops; |
| 933 | #ifdef CONFIG_FS_VERITY |
| 934 | sb->s_vop = &btrfs_verityops; |
| 935 | #endif |
| 936 | sb->s_xattr = btrfs_xattr_handlers; |
| 937 | sb->s_time_gran = 1; |
| 938 | sb->s_iflags |= SB_I_CGROUPWB; |
| 939 | |
| 940 | err = super_setup_bdi(sb); |
| 941 | if (err) { |
| 942 | btrfs_err(fs_info, "super_setup_bdi failed"); |
| 943 | return err; |
| 944 | } |
| 945 | |
| 946 | err = open_ctree(sb, fs_devices, (char *)data); |
| 947 | if (err) { |
| 948 | btrfs_err(fs_info, "open_ctree failed"); |
| 949 | return err; |
| 950 | } |
| 951 | |
| 952 | inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root); |
| 953 | if (IS_ERR(inode)) { |
| 954 | err = PTR_ERR(inode); |
| 955 | btrfs_handle_fs_error(fs_info, err, NULL); |
| 956 | goto fail_close; |
| 957 | } |
| 958 | |
| 959 | sb->s_root = d_make_root(inode); |
| 960 | if (!sb->s_root) { |
| 961 | err = -ENOMEM; |
| 962 | goto fail_close; |
| 963 | } |
| 964 | |
| 965 | sb->s_flags |= SB_ACTIVE; |
| 966 | return 0; |
| 967 | |
| 968 | fail_close: |
| 969 | close_ctree(fs_info); |
| 970 | return err; |
| 971 | } |
| 972 | |
| 973 | int btrfs_sync_fs(struct super_block *sb, int wait) |
| 974 | { |
| 975 | struct btrfs_trans_handle *trans; |
| 976 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 977 | struct btrfs_root *root = fs_info->tree_root; |
| 978 | |
| 979 | trace_btrfs_sync_fs(fs_info, wait); |
| 980 | |
| 981 | if (!wait) { |
| 982 | filemap_flush(fs_info->btree_inode->i_mapping); |
| 983 | return 0; |
| 984 | } |
| 985 | |
| 986 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); |
| 987 | |
| 988 | trans = btrfs_attach_transaction_barrier(root); |
| 989 | if (IS_ERR(trans)) { |
| 990 | /* no transaction, don't bother */ |
| 991 | if (PTR_ERR(trans) == -ENOENT) { |
| 992 | /* |
| 993 | * Exit unless we have some pending changes |
| 994 | * that need to go through commit |
| 995 | */ |
| 996 | if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT, |
| 997 | &fs_info->flags)) |
| 998 | return 0; |
| 999 | /* |
| 1000 | * A non-blocking test if the fs is frozen. We must not |
| 1001 | * start a new transaction here otherwise a deadlock |
| 1002 | * happens. The pending operations are delayed to the |
| 1003 | * next commit after thawing. |
| 1004 | */ |
| 1005 | if (sb_start_write_trylock(sb)) |
| 1006 | sb_end_write(sb); |
| 1007 | else |
| 1008 | return 0; |
| 1009 | trans = btrfs_start_transaction(root, 0); |
| 1010 | } |
| 1011 | if (IS_ERR(trans)) |
| 1012 | return PTR_ERR(trans); |
| 1013 | } |
| 1014 | return btrfs_commit_transaction(trans); |
| 1015 | } |
| 1016 | |
| 1017 | static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed) |
| 1018 | { |
| 1019 | seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s); |
| 1020 | *printed = true; |
| 1021 | } |
| 1022 | |
| 1023 | static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) |
| 1024 | { |
| 1025 | struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb); |
| 1026 | const char *compress_type; |
| 1027 | const char *subvol_name; |
| 1028 | bool printed = false; |
| 1029 | |
| 1030 | if (btrfs_test_opt(info, DEGRADED)) |
| 1031 | seq_puts(seq, ",degraded"); |
| 1032 | if (btrfs_test_opt(info, NODATASUM)) |
| 1033 | seq_puts(seq, ",nodatasum"); |
| 1034 | if (btrfs_test_opt(info, NODATACOW)) |
| 1035 | seq_puts(seq, ",nodatacow"); |
| 1036 | if (btrfs_test_opt(info, NOBARRIER)) |
| 1037 | seq_puts(seq, ",nobarrier"); |
| 1038 | if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE) |
| 1039 | seq_printf(seq, ",max_inline=%llu", info->max_inline); |
| 1040 | if (info->thread_pool_size != min_t(unsigned long, |
| 1041 | num_online_cpus() + 2, 8)) |
| 1042 | seq_printf(seq, ",thread_pool=%u", info->thread_pool_size); |
| 1043 | if (btrfs_test_opt(info, COMPRESS)) { |
| 1044 | compress_type = btrfs_compress_type2str(info->compress_type); |
| 1045 | if (btrfs_test_opt(info, FORCE_COMPRESS)) |
| 1046 | seq_printf(seq, ",compress-force=%s", compress_type); |
| 1047 | else |
| 1048 | seq_printf(seq, ",compress=%s", compress_type); |
| 1049 | if (info->compress_level) |
| 1050 | seq_printf(seq, ":%d", info->compress_level); |
| 1051 | } |
| 1052 | if (btrfs_test_opt(info, NOSSD)) |
| 1053 | seq_puts(seq, ",nossd"); |
| 1054 | if (btrfs_test_opt(info, SSD_SPREAD)) |
| 1055 | seq_puts(seq, ",ssd_spread"); |
| 1056 | else if (btrfs_test_opt(info, SSD)) |
| 1057 | seq_puts(seq, ",ssd"); |
| 1058 | if (btrfs_test_opt(info, NOTREELOG)) |
| 1059 | seq_puts(seq, ",notreelog"); |
| 1060 | if (btrfs_test_opt(info, NOLOGREPLAY)) |
| 1061 | print_rescue_option(seq, "nologreplay", &printed); |
| 1062 | if (btrfs_test_opt(info, USEBACKUPROOT)) |
| 1063 | print_rescue_option(seq, "usebackuproot", &printed); |
| 1064 | if (btrfs_test_opt(info, IGNOREBADROOTS)) |
| 1065 | print_rescue_option(seq, "ignorebadroots", &printed); |
| 1066 | if (btrfs_test_opt(info, IGNOREDATACSUMS)) |
| 1067 | print_rescue_option(seq, "ignoredatacsums", &printed); |
| 1068 | if (btrfs_test_opt(info, FLUSHONCOMMIT)) |
| 1069 | seq_puts(seq, ",flushoncommit"); |
| 1070 | if (btrfs_test_opt(info, DISCARD_SYNC)) |
| 1071 | seq_puts(seq, ",discard"); |
| 1072 | if (btrfs_test_opt(info, DISCARD_ASYNC)) |
| 1073 | seq_puts(seq, ",discard=async"); |
| 1074 | if (!(info->sb->s_flags & SB_POSIXACL)) |
| 1075 | seq_puts(seq, ",noacl"); |
| 1076 | if (btrfs_free_space_cache_v1_active(info)) |
| 1077 | seq_puts(seq, ",space_cache"); |
| 1078 | else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE)) |
| 1079 | seq_puts(seq, ",space_cache=v2"); |
| 1080 | else |
| 1081 | seq_puts(seq, ",nospace_cache"); |
| 1082 | if (btrfs_test_opt(info, RESCAN_UUID_TREE)) |
| 1083 | seq_puts(seq, ",rescan_uuid_tree"); |
| 1084 | if (btrfs_test_opt(info, CLEAR_CACHE)) |
| 1085 | seq_puts(seq, ",clear_cache"); |
| 1086 | if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED)) |
| 1087 | seq_puts(seq, ",user_subvol_rm_allowed"); |
| 1088 | if (btrfs_test_opt(info, ENOSPC_DEBUG)) |
| 1089 | seq_puts(seq, ",enospc_debug"); |
| 1090 | if (btrfs_test_opt(info, AUTO_DEFRAG)) |
| 1091 | seq_puts(seq, ",autodefrag"); |
| 1092 | if (btrfs_test_opt(info, SKIP_BALANCE)) |
| 1093 | seq_puts(seq, ",skip_balance"); |
| 1094 | if (info->metadata_ratio) |
| 1095 | seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio); |
| 1096 | if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR)) |
| 1097 | seq_puts(seq, ",fatal_errors=panic"); |
| 1098 | if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL) |
| 1099 | seq_printf(seq, ",commit=%u", info->commit_interval); |
| 1100 | #ifdef CONFIG_BTRFS_DEBUG |
| 1101 | if (btrfs_test_opt(info, FRAGMENT_DATA)) |
| 1102 | seq_puts(seq, ",fragment=data"); |
| 1103 | if (btrfs_test_opt(info, FRAGMENT_METADATA)) |
| 1104 | seq_puts(seq, ",fragment=metadata"); |
| 1105 | #endif |
| 1106 | if (btrfs_test_opt(info, REF_VERIFY)) |
| 1107 | seq_puts(seq, ",ref_verify"); |
| 1108 | seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root)); |
| 1109 | subvol_name = btrfs_get_subvol_name_from_objectid(info, |
| 1110 | btrfs_root_id(BTRFS_I(d_inode(dentry))->root)); |
| 1111 | if (!IS_ERR(subvol_name)) { |
| 1112 | seq_puts(seq, ",subvol="); |
| 1113 | seq_escape(seq, subvol_name, " \t\n\\"); |
| 1114 | kfree(subvol_name); |
| 1115 | } |
| 1116 | return 0; |
| 1117 | } |
| 1118 | |
| 1119 | /* |
| 1120 | * subvolumes are identified by ino 256 |
| 1121 | */ |
| 1122 | static inline int is_subvolume_inode(struct inode *inode) |
| 1123 | { |
| 1124 | if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) |
| 1125 | return 1; |
| 1126 | return 0; |
| 1127 | } |
| 1128 | |
| 1129 | static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid, |
| 1130 | struct vfsmount *mnt) |
| 1131 | { |
| 1132 | struct dentry *root; |
| 1133 | int ret; |
| 1134 | |
| 1135 | if (!subvol_name) { |
| 1136 | if (!subvol_objectid) { |
| 1137 | ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb), |
| 1138 | &subvol_objectid); |
| 1139 | if (ret) { |
| 1140 | root = ERR_PTR(ret); |
| 1141 | goto out; |
| 1142 | } |
| 1143 | } |
| 1144 | subvol_name = btrfs_get_subvol_name_from_objectid( |
| 1145 | btrfs_sb(mnt->mnt_sb), subvol_objectid); |
| 1146 | if (IS_ERR(subvol_name)) { |
| 1147 | root = ERR_CAST(subvol_name); |
| 1148 | subvol_name = NULL; |
| 1149 | goto out; |
| 1150 | } |
| 1151 | |
| 1152 | } |
| 1153 | |
| 1154 | root = mount_subtree(mnt, subvol_name); |
| 1155 | /* mount_subtree() drops our reference on the vfsmount. */ |
| 1156 | mnt = NULL; |
| 1157 | |
| 1158 | if (!IS_ERR(root)) { |
| 1159 | struct super_block *s = root->d_sb; |
| 1160 | struct btrfs_fs_info *fs_info = btrfs_sb(s); |
| 1161 | struct inode *root_inode = d_inode(root); |
| 1162 | u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root); |
| 1163 | |
| 1164 | ret = 0; |
| 1165 | if (!is_subvolume_inode(root_inode)) { |
| 1166 | btrfs_err(fs_info, "'%s' is not a valid subvolume", |
| 1167 | subvol_name); |
| 1168 | ret = -EINVAL; |
| 1169 | } |
| 1170 | if (subvol_objectid && root_objectid != subvol_objectid) { |
| 1171 | /* |
| 1172 | * This will also catch a race condition where a |
| 1173 | * subvolume which was passed by ID is renamed and |
| 1174 | * another subvolume is renamed over the old location. |
| 1175 | */ |
| 1176 | btrfs_err(fs_info, |
| 1177 | "subvol '%s' does not match subvolid %llu", |
| 1178 | subvol_name, subvol_objectid); |
| 1179 | ret = -EINVAL; |
| 1180 | } |
| 1181 | if (ret) { |
| 1182 | dput(root); |
| 1183 | root = ERR_PTR(ret); |
| 1184 | deactivate_locked_super(s); |
| 1185 | } |
| 1186 | } |
| 1187 | |
| 1188 | out: |
| 1189 | mntput(mnt); |
| 1190 | kfree(subvol_name); |
| 1191 | return root; |
| 1192 | } |
| 1193 | |
| 1194 | static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info, |
| 1195 | u32 new_pool_size, u32 old_pool_size) |
| 1196 | { |
| 1197 | if (new_pool_size == old_pool_size) |
| 1198 | return; |
| 1199 | |
| 1200 | fs_info->thread_pool_size = new_pool_size; |
| 1201 | |
| 1202 | btrfs_info(fs_info, "resize thread pool %d -> %d", |
| 1203 | old_pool_size, new_pool_size); |
| 1204 | |
| 1205 | btrfs_workqueue_set_max(fs_info->workers, new_pool_size); |
| 1206 | btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size); |
| 1207 | btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size); |
| 1208 | workqueue_set_max_active(fs_info->endio_workers, new_pool_size); |
| 1209 | workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size); |
| 1210 | btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size); |
| 1211 | btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size); |
| 1212 | btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size); |
| 1213 | } |
| 1214 | |
| 1215 | static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info, |
| 1216 | unsigned long old_opts, int flags) |
| 1217 | { |
| 1218 | if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && |
| 1219 | (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || |
| 1220 | (flags & SB_RDONLY))) { |
| 1221 | /* wait for any defraggers to finish */ |
| 1222 | wait_event(fs_info->transaction_wait, |
| 1223 | (atomic_read(&fs_info->defrag_running) == 0)); |
| 1224 | if (flags & SB_RDONLY) |
| 1225 | sync_filesystem(fs_info->sb); |
| 1226 | } |
| 1227 | } |
| 1228 | |
| 1229 | static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info, |
| 1230 | unsigned long old_opts) |
| 1231 | { |
| 1232 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
| 1233 | |
| 1234 | /* |
| 1235 | * We need to cleanup all defragable inodes if the autodefragment is |
| 1236 | * close or the filesystem is read only. |
| 1237 | */ |
| 1238 | if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && |
| 1239 | (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) { |
| 1240 | btrfs_cleanup_defrag_inodes(fs_info); |
| 1241 | } |
| 1242 | |
| 1243 | /* If we toggled discard async */ |
| 1244 | if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) && |
| 1245 | btrfs_test_opt(fs_info, DISCARD_ASYNC)) |
| 1246 | btrfs_discard_resume(fs_info); |
| 1247 | else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) && |
| 1248 | !btrfs_test_opt(fs_info, DISCARD_ASYNC)) |
| 1249 | btrfs_discard_cleanup(fs_info); |
| 1250 | |
| 1251 | /* If we toggled space cache */ |
| 1252 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) |
| 1253 | btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); |
| 1254 | } |
| 1255 | |
| 1256 | static int btrfs_remount_rw(struct btrfs_fs_info *fs_info) |
| 1257 | { |
| 1258 | int ret; |
| 1259 | |
| 1260 | if (BTRFS_FS_ERROR(fs_info)) { |
| 1261 | btrfs_err(fs_info, |
| 1262 | "remounting read-write after error is not allowed"); |
| 1263 | return -EINVAL; |
| 1264 | } |
| 1265 | |
| 1266 | if (fs_info->fs_devices->rw_devices == 0) |
| 1267 | return -EACCES; |
| 1268 | |
| 1269 | if (!btrfs_check_rw_degradable(fs_info, NULL)) { |
| 1270 | btrfs_warn(fs_info, |
| 1271 | "too many missing devices, writable remount is not allowed"); |
| 1272 | return -EACCES; |
| 1273 | } |
| 1274 | |
| 1275 | if (btrfs_super_log_root(fs_info->super_copy) != 0) { |
| 1276 | btrfs_warn(fs_info, |
| 1277 | "mount required to replay tree-log, cannot remount read-write"); |
| 1278 | return -EINVAL; |
| 1279 | } |
| 1280 | |
| 1281 | /* |
| 1282 | * NOTE: when remounting with a change that does writes, don't put it |
| 1283 | * anywhere above this point, as we are not sure to be safe to write |
| 1284 | * until we pass the above checks. |
| 1285 | */ |
| 1286 | ret = btrfs_start_pre_rw_mount(fs_info); |
| 1287 | if (ret) |
| 1288 | return ret; |
| 1289 | |
| 1290 | btrfs_clear_sb_rdonly(fs_info->sb); |
| 1291 | |
| 1292 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
| 1293 | |
| 1294 | /* |
| 1295 | * If we've gone from readonly -> read-write, we need to get our |
| 1296 | * sync/async discard lists in the right state. |
| 1297 | */ |
| 1298 | btrfs_discard_resume(fs_info); |
| 1299 | |
| 1300 | return 0; |
| 1301 | } |
| 1302 | |
| 1303 | static int btrfs_remount_ro(struct btrfs_fs_info *fs_info) |
| 1304 | { |
| 1305 | /* |
| 1306 | * This also happens on 'umount -rf' or on shutdown, when the |
| 1307 | * filesystem is busy. |
| 1308 | */ |
| 1309 | cancel_work_sync(&fs_info->async_reclaim_work); |
| 1310 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
| 1311 | |
| 1312 | btrfs_discard_cleanup(fs_info); |
| 1313 | |
| 1314 | /* Wait for the uuid_scan task to finish */ |
| 1315 | down(&fs_info->uuid_tree_rescan_sem); |
| 1316 | /* Avoid complains from lockdep et al. */ |
| 1317 | up(&fs_info->uuid_tree_rescan_sem); |
| 1318 | |
| 1319 | btrfs_set_sb_rdonly(fs_info->sb); |
| 1320 | |
| 1321 | /* |
| 1322 | * Setting SB_RDONLY will put the cleaner thread to sleep at the next |
| 1323 | * loop if it's already active. If it's already asleep, we'll leave |
| 1324 | * unused block groups on disk until we're mounted read-write again |
| 1325 | * unless we clean them up here. |
| 1326 | */ |
| 1327 | btrfs_delete_unused_bgs(fs_info); |
| 1328 | |
| 1329 | /* |
| 1330 | * The cleaner task could be already running before we set the flag |
| 1331 | * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make |
| 1332 | * sure that after we finish the remount, i.e. after we call |
| 1333 | * btrfs_commit_super(), the cleaner can no longer start a transaction |
| 1334 | * - either because it was dropping a dead root, running delayed iputs |
| 1335 | * or deleting an unused block group (the cleaner picked a block |
| 1336 | * group from the list of unused block groups before we were able to |
| 1337 | * in the previous call to btrfs_delete_unused_bgs()). |
| 1338 | */ |
| 1339 | wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE); |
| 1340 | |
| 1341 | /* |
| 1342 | * We've set the superblock to RO mode, so we might have made the |
| 1343 | * cleaner task sleep without running all pending delayed iputs. Go |
| 1344 | * through all the delayed iputs here, so that if an unmount happens |
| 1345 | * without remounting RW we don't end up at finishing close_ctree() |
| 1346 | * with a non-empty list of delayed iputs. |
| 1347 | */ |
| 1348 | btrfs_run_delayed_iputs(fs_info); |
| 1349 | |
| 1350 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
| 1351 | btrfs_scrub_cancel(fs_info); |
| 1352 | btrfs_pause_balance(fs_info); |
| 1353 | |
| 1354 | /* |
| 1355 | * Pause the qgroup rescan worker if it is running. We don't want it to |
| 1356 | * be still running after we are in RO mode, as after that, by the time |
| 1357 | * we unmount, it might have left a transaction open, so we would leak |
| 1358 | * the transaction and/or crash. |
| 1359 | */ |
| 1360 | btrfs_qgroup_wait_for_completion(fs_info, false); |
| 1361 | |
| 1362 | return btrfs_commit_super(fs_info); |
| 1363 | } |
| 1364 | |
| 1365 | static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx) |
| 1366 | { |
| 1367 | fs_info->max_inline = ctx->max_inline; |
| 1368 | fs_info->commit_interval = ctx->commit_interval; |
| 1369 | fs_info->metadata_ratio = ctx->metadata_ratio; |
| 1370 | fs_info->thread_pool_size = ctx->thread_pool_size; |
| 1371 | fs_info->mount_opt = ctx->mount_opt; |
| 1372 | fs_info->compress_type = ctx->compress_type; |
| 1373 | fs_info->compress_level = ctx->compress_level; |
| 1374 | } |
| 1375 | |
| 1376 | static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx) |
| 1377 | { |
| 1378 | ctx->max_inline = fs_info->max_inline; |
| 1379 | ctx->commit_interval = fs_info->commit_interval; |
| 1380 | ctx->metadata_ratio = fs_info->metadata_ratio; |
| 1381 | ctx->thread_pool_size = fs_info->thread_pool_size; |
| 1382 | ctx->mount_opt = fs_info->mount_opt; |
| 1383 | ctx->compress_type = fs_info->compress_type; |
| 1384 | ctx->compress_level = fs_info->compress_level; |
| 1385 | } |
| 1386 | |
| 1387 | #define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \ |
| 1388 | do { \ |
| 1389 | if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \ |
| 1390 | btrfs_raw_test_opt(fs_info->mount_opt, opt)) \ |
| 1391 | btrfs_info(fs_info, fmt, ##args); \ |
| 1392 | } while (0) |
| 1393 | |
| 1394 | #define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \ |
| 1395 | do { \ |
| 1396 | if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \ |
| 1397 | !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \ |
| 1398 | btrfs_info(fs_info, fmt, ##args); \ |
| 1399 | } while (0) |
| 1400 | |
| 1401 | static void btrfs_emit_options(struct btrfs_fs_info *info, |
| 1402 | struct btrfs_fs_context *old) |
| 1403 | { |
| 1404 | btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum"); |
| 1405 | btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts"); |
| 1406 | btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum"); |
| 1407 | btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations"); |
| 1408 | btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme"); |
| 1409 | btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers"); |
| 1410 | btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log"); |
| 1411 | btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time"); |
| 1412 | btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit"); |
| 1413 | btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard"); |
| 1414 | btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard"); |
| 1415 | btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree"); |
| 1416 | btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching"); |
| 1417 | btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache"); |
| 1418 | btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag"); |
| 1419 | btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data"); |
| 1420 | btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata"); |
| 1421 | btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification"); |
| 1422 | btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time"); |
| 1423 | btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots"); |
| 1424 | btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums"); |
| 1425 | |
| 1426 | btrfs_info_if_unset(info, old, NODATACOW, "setting datacow"); |
| 1427 | btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations"); |
| 1428 | btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme"); |
| 1429 | btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers"); |
| 1430 | btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log"); |
| 1431 | btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching"); |
| 1432 | btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree"); |
| 1433 | btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag"); |
| 1434 | btrfs_info_if_unset(info, old, COMPRESS, "use no compression"); |
| 1435 | |
| 1436 | /* Did the compression settings change? */ |
| 1437 | if (btrfs_test_opt(info, COMPRESS) && |
| 1438 | (!old || |
| 1439 | old->compress_type != info->compress_type || |
| 1440 | old->compress_level != info->compress_level || |
| 1441 | (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) && |
| 1442 | btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) { |
| 1443 | const char *compress_type = btrfs_compress_type2str(info->compress_type); |
| 1444 | |
| 1445 | btrfs_info(info, "%s %s compression, level %d", |
| 1446 | btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use", |
| 1447 | compress_type, info->compress_level); |
| 1448 | } |
| 1449 | |
| 1450 | if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE) |
| 1451 | btrfs_info(info, "max_inline set to %llu", info->max_inline); |
| 1452 | } |
| 1453 | |
| 1454 | static int btrfs_reconfigure(struct fs_context *fc) |
| 1455 | { |
| 1456 | struct super_block *sb = fc->root->d_sb; |
| 1457 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 1458 | struct btrfs_fs_context *ctx = fc->fs_private; |
| 1459 | struct btrfs_fs_context old_ctx; |
| 1460 | int ret = 0; |
| 1461 | bool mount_reconfigure = (fc->s_fs_info != NULL); |
| 1462 | |
| 1463 | btrfs_info_to_ctx(fs_info, &old_ctx); |
| 1464 | |
| 1465 | /* |
| 1466 | * This is our "bind mount" trick, we don't want to allow the user to do |
| 1467 | * anything other than mount a different ro/rw and a different subvol, |
| 1468 | * all of the mount options should be maintained. |
| 1469 | */ |
| 1470 | if (mount_reconfigure) |
| 1471 | ctx->mount_opt = old_ctx.mount_opt; |
| 1472 | |
| 1473 | sync_filesystem(sb); |
| 1474 | set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); |
| 1475 | |
| 1476 | if (!mount_reconfigure && |
| 1477 | !btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags)) |
| 1478 | return -EINVAL; |
| 1479 | |
| 1480 | ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY)); |
| 1481 | if (ret < 0) |
| 1482 | return ret; |
| 1483 | |
| 1484 | btrfs_ctx_to_info(fs_info, ctx); |
| 1485 | btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags); |
| 1486 | btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size, |
| 1487 | old_ctx.thread_pool_size); |
| 1488 | |
| 1489 | if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) != |
| 1490 | (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && |
| 1491 | (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) { |
| 1492 | btrfs_warn(fs_info, |
| 1493 | "remount supports changing free space tree only from RO to RW"); |
| 1494 | /* Make sure free space cache options match the state on disk. */ |
| 1495 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
| 1496 | btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE); |
| 1497 | btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); |
| 1498 | } |
| 1499 | if (btrfs_free_space_cache_v1_active(fs_info)) { |
| 1500 | btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE); |
| 1501 | btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE); |
| 1502 | } |
| 1503 | } |
| 1504 | |
| 1505 | ret = 0; |
| 1506 | if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY)) |
| 1507 | ret = btrfs_remount_ro(fs_info); |
| 1508 | else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY)) |
| 1509 | ret = btrfs_remount_rw(fs_info); |
| 1510 | if (ret) |
| 1511 | goto restore; |
| 1512 | |
| 1513 | /* |
| 1514 | * If we set the mask during the parameter parsing VFS would reject the |
| 1515 | * remount. Here we can set the mask and the value will be updated |
| 1516 | * appropriately. |
| 1517 | */ |
| 1518 | if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL)) |
| 1519 | fc->sb_flags_mask |= SB_POSIXACL; |
| 1520 | |
| 1521 | btrfs_emit_options(fs_info, &old_ctx); |
| 1522 | wake_up_process(fs_info->transaction_kthread); |
| 1523 | btrfs_remount_cleanup(fs_info, old_ctx.mount_opt); |
| 1524 | btrfs_clear_oneshot_options(fs_info); |
| 1525 | clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); |
| 1526 | |
| 1527 | return 0; |
| 1528 | restore: |
| 1529 | btrfs_ctx_to_info(fs_info, &old_ctx); |
| 1530 | btrfs_remount_cleanup(fs_info, old_ctx.mount_opt); |
| 1531 | clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); |
| 1532 | return ret; |
| 1533 | } |
| 1534 | |
| 1535 | /* Used to sort the devices by max_avail(descending sort) */ |
| 1536 | static int btrfs_cmp_device_free_bytes(const void *a, const void *b) |
| 1537 | { |
| 1538 | const struct btrfs_device_info *dev_info1 = a; |
| 1539 | const struct btrfs_device_info *dev_info2 = b; |
| 1540 | |
| 1541 | if (dev_info1->max_avail > dev_info2->max_avail) |
| 1542 | return -1; |
| 1543 | else if (dev_info1->max_avail < dev_info2->max_avail) |
| 1544 | return 1; |
| 1545 | return 0; |
| 1546 | } |
| 1547 | |
| 1548 | /* |
| 1549 | * sort the devices by max_avail, in which max free extent size of each device |
| 1550 | * is stored.(Descending Sort) |
| 1551 | */ |
| 1552 | static inline void btrfs_descending_sort_devices( |
| 1553 | struct btrfs_device_info *devices, |
| 1554 | size_t nr_devices) |
| 1555 | { |
| 1556 | sort(devices, nr_devices, sizeof(struct btrfs_device_info), |
| 1557 | btrfs_cmp_device_free_bytes, NULL); |
| 1558 | } |
| 1559 | |
| 1560 | /* |
| 1561 | * The helper to calc the free space on the devices that can be used to store |
| 1562 | * file data. |
| 1563 | */ |
| 1564 | static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info, |
| 1565 | u64 *free_bytes) |
| 1566 | { |
| 1567 | struct btrfs_device_info *devices_info; |
| 1568 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
| 1569 | struct btrfs_device *device; |
| 1570 | u64 type; |
| 1571 | u64 avail_space; |
| 1572 | u64 min_stripe_size; |
| 1573 | int num_stripes = 1; |
| 1574 | int i = 0, nr_devices; |
| 1575 | const struct btrfs_raid_attr *rattr; |
| 1576 | |
| 1577 | /* |
| 1578 | * We aren't under the device list lock, so this is racy-ish, but good |
| 1579 | * enough for our purposes. |
| 1580 | */ |
| 1581 | nr_devices = fs_info->fs_devices->open_devices; |
| 1582 | if (!nr_devices) { |
| 1583 | smp_mb(); |
| 1584 | nr_devices = fs_info->fs_devices->open_devices; |
| 1585 | ASSERT(nr_devices); |
| 1586 | if (!nr_devices) { |
| 1587 | *free_bytes = 0; |
| 1588 | return 0; |
| 1589 | } |
| 1590 | } |
| 1591 | |
| 1592 | devices_info = kmalloc_array(nr_devices, sizeof(*devices_info), |
| 1593 | GFP_KERNEL); |
| 1594 | if (!devices_info) |
| 1595 | return -ENOMEM; |
| 1596 | |
| 1597 | /* calc min stripe number for data space allocation */ |
| 1598 | type = btrfs_data_alloc_profile(fs_info); |
| 1599 | rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)]; |
| 1600 | |
| 1601 | if (type & BTRFS_BLOCK_GROUP_RAID0) |
| 1602 | num_stripes = nr_devices; |
| 1603 | else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK) |
| 1604 | num_stripes = rattr->ncopies; |
| 1605 | else if (type & BTRFS_BLOCK_GROUP_RAID10) |
| 1606 | num_stripes = 4; |
| 1607 | |
| 1608 | /* Adjust for more than 1 stripe per device */ |
| 1609 | min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN; |
| 1610 | |
| 1611 | rcu_read_lock(); |
| 1612 | list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { |
| 1613 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, |
| 1614 | &device->dev_state) || |
| 1615 | !device->bdev || |
| 1616 | test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) |
| 1617 | continue; |
| 1618 | |
| 1619 | if (i >= nr_devices) |
| 1620 | break; |
| 1621 | |
| 1622 | avail_space = device->total_bytes - device->bytes_used; |
| 1623 | |
| 1624 | /* align with stripe_len */ |
| 1625 | avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN); |
| 1626 | |
| 1627 | /* |
| 1628 | * Ensure we have at least min_stripe_size on top of the |
| 1629 | * reserved space on the device. |
| 1630 | */ |
| 1631 | if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size) |
| 1632 | continue; |
| 1633 | |
| 1634 | avail_space -= BTRFS_DEVICE_RANGE_RESERVED; |
| 1635 | |
| 1636 | devices_info[i].dev = device; |
| 1637 | devices_info[i].max_avail = avail_space; |
| 1638 | |
| 1639 | i++; |
| 1640 | } |
| 1641 | rcu_read_unlock(); |
| 1642 | |
| 1643 | nr_devices = i; |
| 1644 | |
| 1645 | btrfs_descending_sort_devices(devices_info, nr_devices); |
| 1646 | |
| 1647 | i = nr_devices - 1; |
| 1648 | avail_space = 0; |
| 1649 | while (nr_devices >= rattr->devs_min) { |
| 1650 | num_stripes = min(num_stripes, nr_devices); |
| 1651 | |
| 1652 | if (devices_info[i].max_avail >= min_stripe_size) { |
| 1653 | int j; |
| 1654 | u64 alloc_size; |
| 1655 | |
| 1656 | avail_space += devices_info[i].max_avail * num_stripes; |
| 1657 | alloc_size = devices_info[i].max_avail; |
| 1658 | for (j = i + 1 - num_stripes; j <= i; j++) |
| 1659 | devices_info[j].max_avail -= alloc_size; |
| 1660 | } |
| 1661 | i--; |
| 1662 | nr_devices--; |
| 1663 | } |
| 1664 | |
| 1665 | kfree(devices_info); |
| 1666 | *free_bytes = avail_space; |
| 1667 | return 0; |
| 1668 | } |
| 1669 | |
| 1670 | /* |
| 1671 | * Calculate numbers for 'df', pessimistic in case of mixed raid profiles. |
| 1672 | * |
| 1673 | * If there's a redundant raid level at DATA block groups, use the respective |
| 1674 | * multiplier to scale the sizes. |
| 1675 | * |
| 1676 | * Unused device space usage is based on simulating the chunk allocator |
| 1677 | * algorithm that respects the device sizes and order of allocations. This is |
| 1678 | * a close approximation of the actual use but there are other factors that may |
| 1679 | * change the result (like a new metadata chunk). |
| 1680 | * |
| 1681 | * If metadata is exhausted, f_bavail will be 0. |
| 1682 | */ |
| 1683 | static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
| 1684 | { |
| 1685 | struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); |
| 1686 | struct btrfs_super_block *disk_super = fs_info->super_copy; |
| 1687 | struct btrfs_space_info *found; |
| 1688 | u64 total_used = 0; |
| 1689 | u64 total_free_data = 0; |
| 1690 | u64 total_free_meta = 0; |
| 1691 | u32 bits = fs_info->sectorsize_bits; |
| 1692 | __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid; |
| 1693 | unsigned factor = 1; |
| 1694 | struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; |
| 1695 | int ret; |
| 1696 | u64 thresh = 0; |
| 1697 | int mixed = 0; |
| 1698 | |
| 1699 | list_for_each_entry(found, &fs_info->space_info, list) { |
| 1700 | if (found->flags & BTRFS_BLOCK_GROUP_DATA) { |
| 1701 | int i; |
| 1702 | |
| 1703 | total_free_data += found->disk_total - found->disk_used; |
| 1704 | total_free_data -= |
| 1705 | btrfs_account_ro_block_groups_free_space(found); |
| 1706 | |
| 1707 | for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) { |
| 1708 | if (!list_empty(&found->block_groups[i])) |
| 1709 | factor = btrfs_bg_type_to_factor( |
| 1710 | btrfs_raid_array[i].bg_flag); |
| 1711 | } |
| 1712 | } |
| 1713 | |
| 1714 | /* |
| 1715 | * Metadata in mixed block group profiles are accounted in data |
| 1716 | */ |
| 1717 | if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) { |
| 1718 | if (found->flags & BTRFS_BLOCK_GROUP_DATA) |
| 1719 | mixed = 1; |
| 1720 | else |
| 1721 | total_free_meta += found->disk_total - |
| 1722 | found->disk_used; |
| 1723 | } |
| 1724 | |
| 1725 | total_used += found->disk_used; |
| 1726 | } |
| 1727 | |
| 1728 | buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor); |
| 1729 | buf->f_blocks >>= bits; |
| 1730 | buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits); |
| 1731 | |
| 1732 | /* Account global block reserve as used, it's in logical size already */ |
| 1733 | spin_lock(&block_rsv->lock); |
| 1734 | /* Mixed block groups accounting is not byte-accurate, avoid overflow */ |
| 1735 | if (buf->f_bfree >= block_rsv->size >> bits) |
| 1736 | buf->f_bfree -= block_rsv->size >> bits; |
| 1737 | else |
| 1738 | buf->f_bfree = 0; |
| 1739 | spin_unlock(&block_rsv->lock); |
| 1740 | |
| 1741 | buf->f_bavail = div_u64(total_free_data, factor); |
| 1742 | ret = btrfs_calc_avail_data_space(fs_info, &total_free_data); |
| 1743 | if (ret) |
| 1744 | return ret; |
| 1745 | buf->f_bavail += div_u64(total_free_data, factor); |
| 1746 | buf->f_bavail = buf->f_bavail >> bits; |
| 1747 | |
| 1748 | /* |
| 1749 | * We calculate the remaining metadata space minus global reserve. If |
| 1750 | * this is (supposedly) smaller than zero, there's no space. But this |
| 1751 | * does not hold in practice, the exhausted state happens where's still |
| 1752 | * some positive delta. So we apply some guesswork and compare the |
| 1753 | * delta to a 4M threshold. (Practically observed delta was ~2M.) |
| 1754 | * |
| 1755 | * We probably cannot calculate the exact threshold value because this |
| 1756 | * depends on the internal reservations requested by various |
| 1757 | * operations, so some operations that consume a few metadata will |
| 1758 | * succeed even if the Avail is zero. But this is better than the other |
| 1759 | * way around. |
| 1760 | */ |
| 1761 | thresh = SZ_4M; |
| 1762 | |
| 1763 | /* |
| 1764 | * We only want to claim there's no available space if we can no longer |
| 1765 | * allocate chunks for our metadata profile and our global reserve will |
| 1766 | * not fit in the free metadata space. If we aren't ->full then we |
| 1767 | * still can allocate chunks and thus are fine using the currently |
| 1768 | * calculated f_bavail. |
| 1769 | */ |
| 1770 | if (!mixed && block_rsv->space_info->full && |
| 1771 | (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size)) |
| 1772 | buf->f_bavail = 0; |
| 1773 | |
| 1774 | buf->f_type = BTRFS_SUPER_MAGIC; |
| 1775 | buf->f_bsize = fs_info->sectorsize; |
| 1776 | buf->f_namelen = BTRFS_NAME_LEN; |
| 1777 | |
| 1778 | /* We treat it as constant endianness (it doesn't matter _which_) |
| 1779 | because we want the fsid to come out the same whether mounted |
| 1780 | on a big-endian or little-endian host */ |
| 1781 | buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); |
| 1782 | buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); |
| 1783 | /* Mask in the root object ID too, to disambiguate subvols */ |
| 1784 | buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32; |
| 1785 | buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root); |
| 1786 | |
| 1787 | return 0; |
| 1788 | } |
| 1789 | |
| 1790 | static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc) |
| 1791 | { |
| 1792 | struct btrfs_fs_info *p = fc->s_fs_info; |
| 1793 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 1794 | |
| 1795 | return fs_info->fs_devices == p->fs_devices; |
| 1796 | } |
| 1797 | |
| 1798 | static int btrfs_get_tree_super(struct fs_context *fc) |
| 1799 | { |
| 1800 | struct btrfs_fs_info *fs_info = fc->s_fs_info; |
| 1801 | struct btrfs_fs_context *ctx = fc->fs_private; |
| 1802 | struct btrfs_fs_devices *fs_devices = NULL; |
| 1803 | struct block_device *bdev; |
| 1804 | struct btrfs_device *device; |
| 1805 | struct super_block *sb; |
| 1806 | blk_mode_t mode = btrfs_open_mode(fc); |
| 1807 | int ret; |
| 1808 | |
| 1809 | btrfs_ctx_to_info(fs_info, ctx); |
| 1810 | mutex_lock(&uuid_mutex); |
| 1811 | |
| 1812 | /* |
| 1813 | * With 'true' passed to btrfs_scan_one_device() (mount time) we expect |
| 1814 | * either a valid device or an error. |
| 1815 | */ |
| 1816 | device = btrfs_scan_one_device(fc->source, mode, true); |
| 1817 | ASSERT(device != NULL); |
| 1818 | if (IS_ERR(device)) { |
| 1819 | mutex_unlock(&uuid_mutex); |
| 1820 | return PTR_ERR(device); |
| 1821 | } |
| 1822 | |
| 1823 | fs_devices = device->fs_devices; |
| 1824 | fs_info->fs_devices = fs_devices; |
| 1825 | |
| 1826 | ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type); |
| 1827 | mutex_unlock(&uuid_mutex); |
| 1828 | if (ret) |
| 1829 | return ret; |
| 1830 | |
| 1831 | if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) { |
| 1832 | ret = -EACCES; |
| 1833 | goto error; |
| 1834 | } |
| 1835 | |
| 1836 | bdev = fs_devices->latest_dev->bdev; |
| 1837 | |
| 1838 | /* |
| 1839 | * From now on the error handling is not straightforward. |
| 1840 | * |
| 1841 | * If successful, this will transfer the fs_info into the super block, |
| 1842 | * and fc->s_fs_info will be NULL. However if there's an existing |
| 1843 | * super, we'll still have fc->s_fs_info populated. If we error |
| 1844 | * completely out it'll be cleaned up when we drop the fs_context, |
| 1845 | * otherwise it's tied to the lifetime of the super_block. |
| 1846 | */ |
| 1847 | sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc); |
| 1848 | if (IS_ERR(sb)) { |
| 1849 | ret = PTR_ERR(sb); |
| 1850 | goto error; |
| 1851 | } |
| 1852 | |
| 1853 | set_device_specific_options(fs_info); |
| 1854 | |
| 1855 | if (sb->s_root) { |
| 1856 | btrfs_close_devices(fs_devices); |
| 1857 | if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY) |
| 1858 | ret = -EBUSY; |
| 1859 | } else { |
| 1860 | snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev); |
| 1861 | shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id); |
| 1862 | btrfs_sb(sb)->bdev_holder = &btrfs_fs_type; |
| 1863 | ret = btrfs_fill_super(sb, fs_devices, NULL); |
| 1864 | } |
| 1865 | |
| 1866 | if (ret) { |
| 1867 | deactivate_locked_super(sb); |
| 1868 | return ret; |
| 1869 | } |
| 1870 | |
| 1871 | btrfs_clear_oneshot_options(fs_info); |
| 1872 | |
| 1873 | fc->root = dget(sb->s_root); |
| 1874 | return 0; |
| 1875 | |
| 1876 | error: |
| 1877 | btrfs_close_devices(fs_devices); |
| 1878 | return ret; |
| 1879 | } |
| 1880 | |
| 1881 | /* |
| 1882 | * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes |
| 1883 | * with different ro/rw options") the following works: |
| 1884 | * |
| 1885 | * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo |
| 1886 | * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar |
| 1887 | * |
| 1888 | * which looks nice and innocent but is actually pretty intricate and deserves |
| 1889 | * a long comment. |
| 1890 | * |
| 1891 | * On another filesystem a subvolume mount is close to something like: |
| 1892 | * |
| 1893 | * (iii) # create rw superblock + initial mount |
| 1894 | * mount -t xfs /dev/sdb /opt/ |
| 1895 | * |
| 1896 | * # create ro bind mount |
| 1897 | * mount --bind -o ro /opt/foo /mnt/foo |
| 1898 | * |
| 1899 | * # unmount initial mount |
| 1900 | * umount /opt |
| 1901 | * |
| 1902 | * Of course, there's some special subvolume sauce and there's the fact that the |
| 1903 | * sb->s_root dentry is really swapped after mount_subtree(). But conceptually |
| 1904 | * it's very close and will help us understand the issue. |
| 1905 | * |
| 1906 | * The old mount API didn't cleanly distinguish between a mount being made ro |
| 1907 | * and a superblock being made ro. The only way to change the ro state of |
| 1908 | * either object was by passing ms_rdonly. If a new mount was created via |
| 1909 | * mount(2) such as: |
| 1910 | * |
| 1911 | * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null); |
| 1912 | * |
| 1913 | * the MS_RDONLY flag being specified had two effects: |
| 1914 | * |
| 1915 | * (1) MNT_READONLY was raised -> the resulting mount got |
| 1916 | * @mnt->mnt_flags |= MNT_READONLY raised. |
| 1917 | * |
| 1918 | * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems |
| 1919 | * made the superblock ro. Note, how SB_RDONLY has the same value as |
| 1920 | * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2). |
| 1921 | * |
| 1922 | * Creating a subtree mount via (iii) ends up leaving a rw superblock with a |
| 1923 | * subtree mounted ro. |
| 1924 | * |
| 1925 | * But consider the effect on the old mount API on btrfs subvolume mounting |
| 1926 | * which combines the distinct step in (iii) into a single step. |
| 1927 | * |
| 1928 | * By issuing (i) both the mount and the superblock are turned ro. Now when (ii) |
| 1929 | * is issued the superblock is ro and thus even if the mount created for (ii) is |
| 1930 | * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro |
| 1931 | * to rw for (ii) which it did using an internal remount call. |
| 1932 | * |
| 1933 | * IOW, subvolume mounting was inherently complicated due to the ambiguity of |
| 1934 | * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate |
| 1935 | * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when |
| 1936 | * passed by mount(8) to mount(2). |
| 1937 | * |
| 1938 | * Enter the new mount API. The new mount API disambiguates making a mount ro |
| 1939 | * and making a superblock ro. |
| 1940 | * |
| 1941 | * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either |
| 1942 | * fsmount() or mount_setattr() this is a pure VFS level change for a |
| 1943 | * specific mount or mount tree that is never seen by the filesystem itself. |
| 1944 | * |
| 1945 | * (4) To turn a superblock ro the "ro" flag must be used with |
| 1946 | * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem |
| 1947 | * in fc->sb_flags. |
| 1948 | * |
| 1949 | * This disambiguation has rather positive consequences. Mounting a subvolume |
| 1950 | * ro will not also turn the superblock ro. Only the mount for the subvolume |
| 1951 | * will become ro. |
| 1952 | * |
| 1953 | * So, if the superblock creation request comes from the new mount API the |
| 1954 | * caller must have explicitly done: |
| 1955 | * |
| 1956 | * fsconfig(FSCONFIG_SET_FLAG, "ro") |
| 1957 | * fsmount/mount_setattr(MOUNT_ATTR_RDONLY) |
| 1958 | * |
| 1959 | * IOW, at some point the caller must have explicitly turned the whole |
| 1960 | * superblock ro and we shouldn't just undo it like we did for the old mount |
| 1961 | * API. In any case, it lets us avoid the hack in the new mount API. |
| 1962 | * |
| 1963 | * Consequently, the remounting hack must only be used for requests originating |
| 1964 | * from the old mount API and should be marked for full deprecation so it can be |
| 1965 | * turned off in a couple of years. |
| 1966 | * |
| 1967 | * The new mount API has no reason to support this hack. |
| 1968 | */ |
| 1969 | static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc) |
| 1970 | { |
| 1971 | struct vfsmount *mnt; |
| 1972 | int ret; |
| 1973 | const bool ro2rw = !(fc->sb_flags & SB_RDONLY); |
| 1974 | |
| 1975 | /* |
| 1976 | * We got an EBUSY because our SB_RDONLY flag didn't match the existing |
| 1977 | * super block, so invert our setting here and retry the mount so we |
| 1978 | * can get our vfsmount. |
| 1979 | */ |
| 1980 | if (ro2rw) |
| 1981 | fc->sb_flags |= SB_RDONLY; |
| 1982 | else |
| 1983 | fc->sb_flags &= ~SB_RDONLY; |
| 1984 | |
| 1985 | mnt = fc_mount(fc); |
| 1986 | if (IS_ERR(mnt)) |
| 1987 | return mnt; |
| 1988 | |
| 1989 | if (!fc->oldapi || !ro2rw) |
| 1990 | return mnt; |
| 1991 | |
| 1992 | /* We need to convert to rw, call reconfigure. */ |
| 1993 | fc->sb_flags &= ~SB_RDONLY; |
| 1994 | down_write(&mnt->mnt_sb->s_umount); |
| 1995 | ret = btrfs_reconfigure(fc); |
| 1996 | up_write(&mnt->mnt_sb->s_umount); |
| 1997 | if (ret) { |
| 1998 | mntput(mnt); |
| 1999 | return ERR_PTR(ret); |
| 2000 | } |
| 2001 | return mnt; |
| 2002 | } |
| 2003 | |
| 2004 | static int btrfs_get_tree_subvol(struct fs_context *fc) |
| 2005 | { |
| 2006 | struct btrfs_fs_info *fs_info = NULL; |
| 2007 | struct btrfs_fs_context *ctx = fc->fs_private; |
| 2008 | struct fs_context *dup_fc; |
| 2009 | struct dentry *dentry; |
| 2010 | struct vfsmount *mnt; |
| 2011 | |
| 2012 | /* |
| 2013 | * Setup a dummy root and fs_info for test/set super. This is because |
| 2014 | * we don't actually fill this stuff out until open_ctree, but we need |
| 2015 | * then open_ctree will properly initialize the file system specific |
| 2016 | * settings later. btrfs_init_fs_info initializes the static elements |
| 2017 | * of the fs_info (locks and such) to make cleanup easier if we find a |
| 2018 | * superblock with our given fs_devices later on at sget() time. |
| 2019 | */ |
| 2020 | fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL); |
| 2021 | if (!fs_info) |
| 2022 | return -ENOMEM; |
| 2023 | |
| 2024 | fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL); |
| 2025 | fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL); |
| 2026 | if (!fs_info->super_copy || !fs_info->super_for_commit) { |
| 2027 | btrfs_free_fs_info(fs_info); |
| 2028 | return -ENOMEM; |
| 2029 | } |
| 2030 | btrfs_init_fs_info(fs_info); |
| 2031 | |
| 2032 | dup_fc = vfs_dup_fs_context(fc); |
| 2033 | if (IS_ERR(dup_fc)) { |
| 2034 | btrfs_free_fs_info(fs_info); |
| 2035 | return PTR_ERR(dup_fc); |
| 2036 | } |
| 2037 | |
| 2038 | /* |
| 2039 | * When we do the sget_fc this gets transferred to the sb, so we only |
| 2040 | * need to set it on the dup_fc as this is what creates the super block. |
| 2041 | */ |
| 2042 | dup_fc->s_fs_info = fs_info; |
| 2043 | |
| 2044 | /* |
| 2045 | * We'll do the security settings in our btrfs_get_tree_super() mount |
| 2046 | * loop, they were duplicated into dup_fc, we can drop the originals |
| 2047 | * here. |
| 2048 | */ |
| 2049 | security_free_mnt_opts(&fc->security); |
| 2050 | fc->security = NULL; |
| 2051 | |
| 2052 | mnt = fc_mount(dup_fc); |
| 2053 | if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) |
| 2054 | mnt = btrfs_reconfigure_for_mount(dup_fc); |
| 2055 | put_fs_context(dup_fc); |
| 2056 | if (IS_ERR(mnt)) |
| 2057 | return PTR_ERR(mnt); |
| 2058 | |
| 2059 | /* |
| 2060 | * This free's ->subvol_name, because if it isn't set we have to |
| 2061 | * allocate a buffer to hold the subvol_name, so we just drop our |
| 2062 | * reference to it here. |
| 2063 | */ |
| 2064 | dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt); |
| 2065 | ctx->subvol_name = NULL; |
| 2066 | if (IS_ERR(dentry)) |
| 2067 | return PTR_ERR(dentry); |
| 2068 | |
| 2069 | fc->root = dentry; |
| 2070 | return 0; |
| 2071 | } |
| 2072 | |
| 2073 | static int btrfs_get_tree(struct fs_context *fc) |
| 2074 | { |
| 2075 | /* |
| 2076 | * Since we use mount_subtree to mount the default/specified subvol, we |
| 2077 | * have to do mounts in two steps. |
| 2078 | * |
| 2079 | * First pass through we call btrfs_get_tree_subvol(), this is just a |
| 2080 | * wrapper around fc_mount() to call back into here again, and this time |
| 2081 | * we'll call btrfs_get_tree_super(). This will do the open_ctree() and |
| 2082 | * everything to open the devices and file system. Then we return back |
| 2083 | * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and |
| 2084 | * from there we can do our mount_subvol() call, which will lookup |
| 2085 | * whichever subvol we're mounting and setup this fc with the |
| 2086 | * appropriate dentry for the subvol. |
| 2087 | */ |
| 2088 | if (fc->s_fs_info) |
| 2089 | return btrfs_get_tree_super(fc); |
| 2090 | return btrfs_get_tree_subvol(fc); |
| 2091 | } |
| 2092 | |
| 2093 | static void btrfs_kill_super(struct super_block *sb) |
| 2094 | { |
| 2095 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 2096 | kill_anon_super(sb); |
| 2097 | btrfs_free_fs_info(fs_info); |
| 2098 | } |
| 2099 | |
| 2100 | static void btrfs_free_fs_context(struct fs_context *fc) |
| 2101 | { |
| 2102 | struct btrfs_fs_context *ctx = fc->fs_private; |
| 2103 | struct btrfs_fs_info *fs_info = fc->s_fs_info; |
| 2104 | |
| 2105 | if (fs_info) |
| 2106 | btrfs_free_fs_info(fs_info); |
| 2107 | |
| 2108 | if (ctx && refcount_dec_and_test(&ctx->refs)) { |
| 2109 | kfree(ctx->subvol_name); |
| 2110 | kfree(ctx); |
| 2111 | } |
| 2112 | } |
| 2113 | |
| 2114 | static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc) |
| 2115 | { |
| 2116 | struct btrfs_fs_context *ctx = src_fc->fs_private; |
| 2117 | |
| 2118 | /* |
| 2119 | * Give a ref to our ctx to this dup, as we want to keep it around for |
| 2120 | * our original fc so we can have the subvolume name or objectid. |
| 2121 | * |
| 2122 | * We unset ->source in the original fc because the dup needs it for |
| 2123 | * mounting, and then once we free the dup it'll free ->source, so we |
| 2124 | * need to make sure we're only pointing to it in one fc. |
| 2125 | */ |
| 2126 | refcount_inc(&ctx->refs); |
| 2127 | fc->fs_private = ctx; |
| 2128 | fc->source = src_fc->source; |
| 2129 | src_fc->source = NULL; |
| 2130 | return 0; |
| 2131 | } |
| 2132 | |
| 2133 | static const struct fs_context_operations btrfs_fs_context_ops = { |
| 2134 | .parse_param = btrfs_parse_param, |
| 2135 | .reconfigure = btrfs_reconfigure, |
| 2136 | .get_tree = btrfs_get_tree, |
| 2137 | .dup = btrfs_dup_fs_context, |
| 2138 | .free = btrfs_free_fs_context, |
| 2139 | }; |
| 2140 | |
| 2141 | static int btrfs_init_fs_context(struct fs_context *fc) |
| 2142 | { |
| 2143 | struct btrfs_fs_context *ctx; |
| 2144 | |
| 2145 | ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL); |
| 2146 | if (!ctx) |
| 2147 | return -ENOMEM; |
| 2148 | |
| 2149 | refcount_set(&ctx->refs, 1); |
| 2150 | fc->fs_private = ctx; |
| 2151 | fc->ops = &btrfs_fs_context_ops; |
| 2152 | |
| 2153 | if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { |
| 2154 | btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx); |
| 2155 | } else { |
| 2156 | ctx->thread_pool_size = |
| 2157 | min_t(unsigned long, num_online_cpus() + 2, 8); |
| 2158 | ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
| 2159 | ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
| 2160 | } |
| 2161 | |
| 2162 | #ifdef CONFIG_BTRFS_FS_POSIX_ACL |
| 2163 | fc->sb_flags |= SB_POSIXACL; |
| 2164 | #endif |
| 2165 | fc->sb_flags |= SB_I_VERSION; |
| 2166 | |
| 2167 | return 0; |
| 2168 | } |
| 2169 | |
| 2170 | static struct file_system_type btrfs_fs_type = { |
| 2171 | .owner = THIS_MODULE, |
| 2172 | .name = "btrfs", |
| 2173 | .init_fs_context = btrfs_init_fs_context, |
| 2174 | .parameters = btrfs_fs_parameters, |
| 2175 | .kill_sb = btrfs_kill_super, |
| 2176 | .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP, |
| 2177 | }; |
| 2178 | |
| 2179 | MODULE_ALIAS_FS("btrfs"); |
| 2180 | |
| 2181 | static int btrfs_control_open(struct inode *inode, struct file *file) |
| 2182 | { |
| 2183 | /* |
| 2184 | * The control file's private_data is used to hold the |
| 2185 | * transaction when it is started and is used to keep |
| 2186 | * track of whether a transaction is already in progress. |
| 2187 | */ |
| 2188 | file->private_data = NULL; |
| 2189 | return 0; |
| 2190 | } |
| 2191 | |
| 2192 | /* |
| 2193 | * Used by /dev/btrfs-control for devices ioctls. |
| 2194 | */ |
| 2195 | static long btrfs_control_ioctl(struct file *file, unsigned int cmd, |
| 2196 | unsigned long arg) |
| 2197 | { |
| 2198 | struct btrfs_ioctl_vol_args *vol; |
| 2199 | struct btrfs_device *device = NULL; |
| 2200 | dev_t devt = 0; |
| 2201 | int ret = -ENOTTY; |
| 2202 | |
| 2203 | if (!capable(CAP_SYS_ADMIN)) |
| 2204 | return -EPERM; |
| 2205 | |
| 2206 | vol = memdup_user((void __user *)arg, sizeof(*vol)); |
| 2207 | if (IS_ERR(vol)) |
| 2208 | return PTR_ERR(vol); |
| 2209 | ret = btrfs_check_ioctl_vol_args_path(vol); |
| 2210 | if (ret < 0) |
| 2211 | goto out; |
| 2212 | |
| 2213 | switch (cmd) { |
| 2214 | case BTRFS_IOC_SCAN_DEV: |
| 2215 | mutex_lock(&uuid_mutex); |
| 2216 | /* |
| 2217 | * Scanning outside of mount can return NULL which would turn |
| 2218 | * into 0 error code. |
| 2219 | */ |
| 2220 | device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false); |
| 2221 | ret = PTR_ERR_OR_ZERO(device); |
| 2222 | mutex_unlock(&uuid_mutex); |
| 2223 | break; |
| 2224 | case BTRFS_IOC_FORGET_DEV: |
| 2225 | if (vol->name[0] != 0) { |
| 2226 | ret = lookup_bdev(vol->name, &devt); |
| 2227 | if (ret) |
| 2228 | break; |
| 2229 | } |
| 2230 | ret = btrfs_forget_devices(devt); |
| 2231 | break; |
| 2232 | case BTRFS_IOC_DEVICES_READY: |
| 2233 | mutex_lock(&uuid_mutex); |
| 2234 | /* |
| 2235 | * Scanning outside of mount can return NULL which would turn |
| 2236 | * into 0 error code. |
| 2237 | */ |
| 2238 | device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false); |
| 2239 | if (IS_ERR_OR_NULL(device)) { |
| 2240 | mutex_unlock(&uuid_mutex); |
| 2241 | ret = PTR_ERR(device); |
| 2242 | break; |
| 2243 | } |
| 2244 | ret = !(device->fs_devices->num_devices == |
| 2245 | device->fs_devices->total_devices); |
| 2246 | mutex_unlock(&uuid_mutex); |
| 2247 | break; |
| 2248 | case BTRFS_IOC_GET_SUPPORTED_FEATURES: |
| 2249 | ret = btrfs_ioctl_get_supported_features((void __user*)arg); |
| 2250 | break; |
| 2251 | } |
| 2252 | |
| 2253 | out: |
| 2254 | kfree(vol); |
| 2255 | return ret; |
| 2256 | } |
| 2257 | |
| 2258 | static int btrfs_freeze(struct super_block *sb) |
| 2259 | { |
| 2260 | struct btrfs_trans_handle *trans; |
| 2261 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 2262 | struct btrfs_root *root = fs_info->tree_root; |
| 2263 | |
| 2264 | set_bit(BTRFS_FS_FROZEN, &fs_info->flags); |
| 2265 | /* |
| 2266 | * We don't need a barrier here, we'll wait for any transaction that |
| 2267 | * could be in progress on other threads (and do delayed iputs that |
| 2268 | * we want to avoid on a frozen filesystem), or do the commit |
| 2269 | * ourselves. |
| 2270 | */ |
| 2271 | trans = btrfs_attach_transaction_barrier(root); |
| 2272 | if (IS_ERR(trans)) { |
| 2273 | /* no transaction, don't bother */ |
| 2274 | if (PTR_ERR(trans) == -ENOENT) |
| 2275 | return 0; |
| 2276 | return PTR_ERR(trans); |
| 2277 | } |
| 2278 | return btrfs_commit_transaction(trans); |
| 2279 | } |
| 2280 | |
| 2281 | static int check_dev_super(struct btrfs_device *dev) |
| 2282 | { |
| 2283 | struct btrfs_fs_info *fs_info = dev->fs_info; |
| 2284 | struct btrfs_super_block *sb; |
| 2285 | u64 last_trans; |
| 2286 | u16 csum_type; |
| 2287 | int ret = 0; |
| 2288 | |
| 2289 | /* This should be called with fs still frozen. */ |
| 2290 | ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags)); |
| 2291 | |
| 2292 | /* Missing dev, no need to check. */ |
| 2293 | if (!dev->bdev) |
| 2294 | return 0; |
| 2295 | |
| 2296 | /* Only need to check the primary super block. */ |
| 2297 | sb = btrfs_read_dev_one_super(dev->bdev, 0, true); |
| 2298 | if (IS_ERR(sb)) |
| 2299 | return PTR_ERR(sb); |
| 2300 | |
| 2301 | /* Verify the checksum. */ |
| 2302 | csum_type = btrfs_super_csum_type(sb); |
| 2303 | if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) { |
| 2304 | btrfs_err(fs_info, "csum type changed, has %u expect %u", |
| 2305 | csum_type, btrfs_super_csum_type(fs_info->super_copy)); |
| 2306 | ret = -EUCLEAN; |
| 2307 | goto out; |
| 2308 | } |
| 2309 | |
| 2310 | if (btrfs_check_super_csum(fs_info, sb)) { |
| 2311 | btrfs_err(fs_info, "csum for on-disk super block no longer matches"); |
| 2312 | ret = -EUCLEAN; |
| 2313 | goto out; |
| 2314 | } |
| 2315 | |
| 2316 | /* Btrfs_validate_super() includes fsid check against super->fsid. */ |
| 2317 | ret = btrfs_validate_super(fs_info, sb, 0); |
| 2318 | if (ret < 0) |
| 2319 | goto out; |
| 2320 | |
| 2321 | last_trans = btrfs_get_last_trans_committed(fs_info); |
| 2322 | if (btrfs_super_generation(sb) != last_trans) { |
| 2323 | btrfs_err(fs_info, "transid mismatch, has %llu expect %llu", |
| 2324 | btrfs_super_generation(sb), last_trans); |
| 2325 | ret = -EUCLEAN; |
| 2326 | goto out; |
| 2327 | } |
| 2328 | out: |
| 2329 | btrfs_release_disk_super(sb); |
| 2330 | return ret; |
| 2331 | } |
| 2332 | |
| 2333 | static int btrfs_unfreeze(struct super_block *sb) |
| 2334 | { |
| 2335 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 2336 | struct btrfs_device *device; |
| 2337 | int ret = 0; |
| 2338 | |
| 2339 | /* |
| 2340 | * Make sure the fs is not changed by accident (like hibernation then |
| 2341 | * modified by other OS). |
| 2342 | * If we found anything wrong, we mark the fs error immediately. |
| 2343 | * |
| 2344 | * And since the fs is frozen, no one can modify the fs yet, thus |
| 2345 | * we don't need to hold device_list_mutex. |
| 2346 | */ |
| 2347 | list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) { |
| 2348 | ret = check_dev_super(device); |
| 2349 | if (ret < 0) { |
| 2350 | btrfs_handle_fs_error(fs_info, ret, |
| 2351 | "super block on devid %llu got modified unexpectedly", |
| 2352 | device->devid); |
| 2353 | break; |
| 2354 | } |
| 2355 | } |
| 2356 | clear_bit(BTRFS_FS_FROZEN, &fs_info->flags); |
| 2357 | |
| 2358 | /* |
| 2359 | * We still return 0, to allow VFS layer to unfreeze the fs even the |
| 2360 | * above checks failed. Since the fs is either fine or read-only, we're |
| 2361 | * safe to continue, without causing further damage. |
| 2362 | */ |
| 2363 | return 0; |
| 2364 | } |
| 2365 | |
| 2366 | static int btrfs_show_devname(struct seq_file *m, struct dentry *root) |
| 2367 | { |
| 2368 | struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); |
| 2369 | |
| 2370 | /* |
| 2371 | * There should be always a valid pointer in latest_dev, it may be stale |
| 2372 | * for a short moment in case it's being deleted but still valid until |
| 2373 | * the end of RCU grace period. |
| 2374 | */ |
| 2375 | rcu_read_lock(); |
| 2376 | seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\"); |
| 2377 | rcu_read_unlock(); |
| 2378 | |
| 2379 | return 0; |
| 2380 | } |
| 2381 | |
| 2382 | static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc) |
| 2383 | { |
| 2384 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 2385 | const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps); |
| 2386 | |
| 2387 | trace_btrfs_extent_map_shrinker_count(fs_info, nr); |
| 2388 | |
| 2389 | return nr; |
| 2390 | } |
| 2391 | |
| 2392 | static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc) |
| 2393 | { |
| 2394 | const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan); |
| 2395 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
| 2396 | |
| 2397 | return btrfs_free_extent_maps(fs_info, nr_to_scan); |
| 2398 | } |
| 2399 | |
| 2400 | static const struct super_operations btrfs_super_ops = { |
| 2401 | .drop_inode = btrfs_drop_inode, |
| 2402 | .evict_inode = btrfs_evict_inode, |
| 2403 | .put_super = btrfs_put_super, |
| 2404 | .sync_fs = btrfs_sync_fs, |
| 2405 | .show_options = btrfs_show_options, |
| 2406 | .show_devname = btrfs_show_devname, |
| 2407 | .alloc_inode = btrfs_alloc_inode, |
| 2408 | .destroy_inode = btrfs_destroy_inode, |
| 2409 | .free_inode = btrfs_free_inode, |
| 2410 | .statfs = btrfs_statfs, |
| 2411 | .freeze_fs = btrfs_freeze, |
| 2412 | .unfreeze_fs = btrfs_unfreeze, |
| 2413 | .nr_cached_objects = btrfs_nr_cached_objects, |
| 2414 | .free_cached_objects = btrfs_free_cached_objects, |
| 2415 | }; |
| 2416 | |
| 2417 | static const struct file_operations btrfs_ctl_fops = { |
| 2418 | .open = btrfs_control_open, |
| 2419 | .unlocked_ioctl = btrfs_control_ioctl, |
| 2420 | .compat_ioctl = compat_ptr_ioctl, |
| 2421 | .owner = THIS_MODULE, |
| 2422 | .llseek = noop_llseek, |
| 2423 | }; |
| 2424 | |
| 2425 | static struct miscdevice btrfs_misc = { |
| 2426 | .minor = BTRFS_MINOR, |
| 2427 | .name = "btrfs-control", |
| 2428 | .fops = &btrfs_ctl_fops |
| 2429 | }; |
| 2430 | |
| 2431 | MODULE_ALIAS_MISCDEV(BTRFS_MINOR); |
| 2432 | MODULE_ALIAS("devname:btrfs-control"); |
| 2433 | |
| 2434 | static int __init btrfs_interface_init(void) |
| 2435 | { |
| 2436 | return misc_register(&btrfs_misc); |
| 2437 | } |
| 2438 | |
| 2439 | static __cold void btrfs_interface_exit(void) |
| 2440 | { |
| 2441 | misc_deregister(&btrfs_misc); |
| 2442 | } |
| 2443 | |
| 2444 | static int __init btrfs_print_mod_info(void) |
| 2445 | { |
| 2446 | static const char options[] = "" |
| 2447 | #ifdef CONFIG_BTRFS_DEBUG |
| 2448 | ", debug=on" |
| 2449 | #endif |
| 2450 | #ifdef CONFIG_BTRFS_ASSERT |
| 2451 | ", assert=on" |
| 2452 | #endif |
| 2453 | #ifdef CONFIG_BTRFS_FS_REF_VERIFY |
| 2454 | ", ref-verify=on" |
| 2455 | #endif |
| 2456 | #ifdef CONFIG_BLK_DEV_ZONED |
| 2457 | ", zoned=yes" |
| 2458 | #else |
| 2459 | ", zoned=no" |
| 2460 | #endif |
| 2461 | #ifdef CONFIG_FS_VERITY |
| 2462 | ", fsverity=yes" |
| 2463 | #else |
| 2464 | ", fsverity=no" |
| 2465 | #endif |
| 2466 | ; |
| 2467 | pr_info("Btrfs loaded%s\n", options); |
| 2468 | return 0; |
| 2469 | } |
| 2470 | |
| 2471 | static int register_btrfs(void) |
| 2472 | { |
| 2473 | return register_filesystem(&btrfs_fs_type); |
| 2474 | } |
| 2475 | |
| 2476 | static void unregister_btrfs(void) |
| 2477 | { |
| 2478 | unregister_filesystem(&btrfs_fs_type); |
| 2479 | } |
| 2480 | |
| 2481 | /* Helper structure for long init/exit functions. */ |
| 2482 | struct init_sequence { |
| 2483 | int (*init_func)(void); |
| 2484 | /* Can be NULL if the init_func doesn't need cleanup. */ |
| 2485 | void (*exit_func)(void); |
| 2486 | }; |
| 2487 | |
| 2488 | static const struct init_sequence mod_init_seq[] = { |
| 2489 | { |
| 2490 | .init_func = btrfs_props_init, |
| 2491 | .exit_func = NULL, |
| 2492 | }, { |
| 2493 | .init_func = btrfs_init_sysfs, |
| 2494 | .exit_func = btrfs_exit_sysfs, |
| 2495 | }, { |
| 2496 | .init_func = btrfs_init_compress, |
| 2497 | .exit_func = btrfs_exit_compress, |
| 2498 | }, { |
| 2499 | .init_func = btrfs_init_cachep, |
| 2500 | .exit_func = btrfs_destroy_cachep, |
| 2501 | }, { |
| 2502 | .init_func = btrfs_transaction_init, |
| 2503 | .exit_func = btrfs_transaction_exit, |
| 2504 | }, { |
| 2505 | .init_func = btrfs_ctree_init, |
| 2506 | .exit_func = btrfs_ctree_exit, |
| 2507 | }, { |
| 2508 | .init_func = btrfs_free_space_init, |
| 2509 | .exit_func = btrfs_free_space_exit, |
| 2510 | }, { |
| 2511 | .init_func = extent_state_init_cachep, |
| 2512 | .exit_func = extent_state_free_cachep, |
| 2513 | }, { |
| 2514 | .init_func = extent_buffer_init_cachep, |
| 2515 | .exit_func = extent_buffer_free_cachep, |
| 2516 | }, { |
| 2517 | .init_func = btrfs_bioset_init, |
| 2518 | .exit_func = btrfs_bioset_exit, |
| 2519 | }, { |
| 2520 | .init_func = extent_map_init, |
| 2521 | .exit_func = extent_map_exit, |
| 2522 | }, { |
| 2523 | .init_func = ordered_data_init, |
| 2524 | .exit_func = ordered_data_exit, |
| 2525 | }, { |
| 2526 | .init_func = btrfs_delayed_inode_init, |
| 2527 | .exit_func = btrfs_delayed_inode_exit, |
| 2528 | }, { |
| 2529 | .init_func = btrfs_auto_defrag_init, |
| 2530 | .exit_func = btrfs_auto_defrag_exit, |
| 2531 | }, { |
| 2532 | .init_func = btrfs_delayed_ref_init, |
| 2533 | .exit_func = btrfs_delayed_ref_exit, |
| 2534 | }, { |
| 2535 | .init_func = btrfs_prelim_ref_init, |
| 2536 | .exit_func = btrfs_prelim_ref_exit, |
| 2537 | }, { |
| 2538 | .init_func = btrfs_interface_init, |
| 2539 | .exit_func = btrfs_interface_exit, |
| 2540 | }, { |
| 2541 | .init_func = btrfs_print_mod_info, |
| 2542 | .exit_func = NULL, |
| 2543 | }, { |
| 2544 | .init_func = btrfs_run_sanity_tests, |
| 2545 | .exit_func = NULL, |
| 2546 | }, { |
| 2547 | .init_func = register_btrfs, |
| 2548 | .exit_func = unregister_btrfs, |
| 2549 | } |
| 2550 | }; |
| 2551 | |
| 2552 | static bool mod_init_result[ARRAY_SIZE(mod_init_seq)]; |
| 2553 | |
| 2554 | static __always_inline void btrfs_exit_btrfs_fs(void) |
| 2555 | { |
| 2556 | int i; |
| 2557 | |
| 2558 | for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) { |
| 2559 | if (!mod_init_result[i]) |
| 2560 | continue; |
| 2561 | if (mod_init_seq[i].exit_func) |
| 2562 | mod_init_seq[i].exit_func(); |
| 2563 | mod_init_result[i] = false; |
| 2564 | } |
| 2565 | } |
| 2566 | |
| 2567 | static void __exit exit_btrfs_fs(void) |
| 2568 | { |
| 2569 | btrfs_exit_btrfs_fs(); |
| 2570 | btrfs_cleanup_fs_uuids(); |
| 2571 | } |
| 2572 | |
| 2573 | static int __init init_btrfs_fs(void) |
| 2574 | { |
| 2575 | int ret; |
| 2576 | int i; |
| 2577 | |
| 2578 | for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) { |
| 2579 | ASSERT(!mod_init_result[i]); |
| 2580 | ret = mod_init_seq[i].init_func(); |
| 2581 | if (ret < 0) { |
| 2582 | btrfs_exit_btrfs_fs(); |
| 2583 | return ret; |
| 2584 | } |
| 2585 | mod_init_result[i] = true; |
| 2586 | } |
| 2587 | return 0; |
| 2588 | } |
| 2589 | |
| 2590 | late_initcall(init_btrfs_fs); |
| 2591 | module_exit(exit_btrfs_fs) |
| 2592 | |
| 2593 | MODULE_LICENSE("GPL"); |
| 2594 | MODULE_SOFTDEP("pre: crc32c"); |
| 2595 | MODULE_SOFTDEP("pre: xxhash64"); |
| 2596 | MODULE_SOFTDEP("pre: sha256"); |
| 2597 | MODULE_SOFTDEP("pre: blake2b-256"); |