1 // SPDX-License-Identifier: GPL-2.0 1 // SPDX-License-Identifier: GPL-2.0
2 /* 2 /*
3 * Copyright (C) 2007 Oracle. All rights rese 3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */ 4 */
5 5
6 #include <linux/blkdev.h> 6 #include <linux/blkdev.h>
7 #include <linux/module.h> 7 #include <linux/module.h>
8 #include <linux/fs.h> 8 #include <linux/fs.h>
9 #include <linux/pagemap.h> 9 #include <linux/pagemap.h>
10 #include <linux/highmem.h> 10 #include <linux/highmem.h>
11 #include <linux/time.h> 11 #include <linux/time.h>
12 #include <linux/init.h> 12 #include <linux/init.h>
13 #include <linux/seq_file.h> 13 #include <linux/seq_file.h>
14 #include <linux/string.h> 14 #include <linux/string.h>
15 #include <linux/backing-dev.h> 15 #include <linux/backing-dev.h>
16 #include <linux/mount.h> 16 #include <linux/mount.h>
17 #include <linux/writeback.h> 17 #include <linux/writeback.h>
18 #include <linux/statfs.h> 18 #include <linux/statfs.h>
19 #include <linux/compat.h> 19 #include <linux/compat.h>
20 #include <linux/parser.h> 20 #include <linux/parser.h>
21 #include <linux/ctype.h> 21 #include <linux/ctype.h>
22 #include <linux/namei.h> 22 #include <linux/namei.h>
23 #include <linux/miscdevice.h> 23 #include <linux/miscdevice.h>
24 #include <linux/magic.h> 24 #include <linux/magic.h>
25 #include <linux/slab.h> 25 #include <linux/slab.h>
26 #include <linux/ratelimit.h> 26 #include <linux/ratelimit.h>
27 #include <linux/crc32c.h> 27 #include <linux/crc32c.h>
28 #include <linux/btrfs.h> 28 #include <linux/btrfs.h>
29 #include <linux/security.h> <<
30 #include <linux/fs_parser.h> <<
31 #include <linux/swap.h> <<
32 #include "messages.h" <<
33 #include "delayed-inode.h" 29 #include "delayed-inode.h"
34 #include "ctree.h" 30 #include "ctree.h"
35 #include "disk-io.h" 31 #include "disk-io.h"
36 #include "transaction.h" 32 #include "transaction.h"
37 #include "btrfs_inode.h" 33 #include "btrfs_inode.h"
38 #include "direct-io.h" !! 34 #include "print-tree.h"
39 #include "props.h" 35 #include "props.h"
40 #include "xattr.h" 36 #include "xattr.h"
41 #include "bio.h" !! 37 #include "volumes.h"
42 #include "export.h" 38 #include "export.h"
43 #include "compression.h" 39 #include "compression.h"
>> 40 #include "rcu-string.h"
44 #include "dev-replace.h" 41 #include "dev-replace.h"
45 #include "free-space-cache.h" 42 #include "free-space-cache.h"
46 #include "backref.h" 43 #include "backref.h"
47 #include "space-info.h" 44 #include "space-info.h"
48 #include "sysfs.h" 45 #include "sysfs.h"
49 #include "zoned.h" 46 #include "zoned.h"
50 #include "tests/btrfs-tests.h" 47 #include "tests/btrfs-tests.h"
51 #include "block-group.h" 48 #include "block-group.h"
52 #include "discard.h" 49 #include "discard.h"
53 #include "qgroup.h" 50 #include "qgroup.h"
54 #include "raid56.h" 51 #include "raid56.h"
55 #include "fs.h" <<
56 #include "accessors.h" <<
57 #include "defrag.h" <<
58 #include "dir-item.h" <<
59 #include "ioctl.h" <<
60 #include "scrub.h" <<
61 #include "verity.h" <<
62 #include "super.h" <<
63 #include "extent-tree.h" <<
64 #define CREATE_TRACE_POINTS 52 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h> 53 #include <trace/events/btrfs.h>
66 54
67 static const struct super_operations btrfs_sup 55 static const struct super_operations btrfs_super_ops;
>> 56
>> 57 /*
>> 58 * Types for mounting the default subvolume and a subvolume explicitly
>> 59 * requested by subvol=/path. That way the callchain is straightforward and we
>> 60 * don't have to play tricks with the mount options and recursive calls to
>> 61 * btrfs_mount.
>> 62 *
>> 63 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
>> 64 */
68 static struct file_system_type btrfs_fs_type; 65 static struct file_system_type btrfs_fs_type;
>> 66 static struct file_system_type btrfs_root_fs_type;
69 67
70 static void btrfs_put_super(struct super_block !! 68 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
>> 69
>> 70 #ifdef CONFIG_PRINTK
>> 71
>> 72 #define STATE_STRING_PREFACE ": state "
>> 73 #define STATE_STRING_BUF_LEN (sizeof(STATE_STRING_PREFACE) + BTRFS_FS_STATE_COUNT)
>> 74
>> 75 /*
>> 76 * Characters to print to indicate error conditions or uncommon filesystem state.
>> 77 * RO is not an error.
>> 78 */
>> 79 static const char fs_state_chars[] = {
>> 80 [BTRFS_FS_STATE_ERROR] = 'E',
>> 81 [BTRFS_FS_STATE_REMOUNTING] = 'M',
>> 82 [BTRFS_FS_STATE_RO] = 0,
>> 83 [BTRFS_FS_STATE_TRANS_ABORTED] = 'A',
>> 84 [BTRFS_FS_STATE_DEV_REPLACING] = 'R',
>> 85 [BTRFS_FS_STATE_DUMMY_FS_INFO] = 0,
>> 86 [BTRFS_FS_STATE_NO_CSUMS] = 'C',
>> 87 [BTRFS_FS_STATE_LOG_CLEANUP_ERROR] = 'L',
>> 88 };
>> 89
>> 90 static void btrfs_state_to_string(const struct btrfs_fs_info *info, char *buf)
71 { 91 {
72 struct btrfs_fs_info *fs_info = btrfs_ !! 92 unsigned int bit;
>> 93 bool states_printed = false;
>> 94 unsigned long fs_state = READ_ONCE(info->fs_state);
>> 95 char *curr = buf;
73 96
74 btrfs_info(fs_info, "last unmount of f !! 97 memcpy(curr, STATE_STRING_PREFACE, sizeof(STATE_STRING_PREFACE));
75 close_ctree(fs_info); !! 98 curr += sizeof(STATE_STRING_PREFACE) - 1;
>> 99
>> 100 for_each_set_bit(bit, &fs_state, sizeof(fs_state)) {
>> 101 WARN_ON_ONCE(bit >= BTRFS_FS_STATE_COUNT);
>> 102 if ((bit < BTRFS_FS_STATE_COUNT) && fs_state_chars[bit]) {
>> 103 *curr++ = fs_state_chars[bit];
>> 104 states_printed = true;
>> 105 }
>> 106 }
>> 107
>> 108 /* If no states were printed, reset the buffer */
>> 109 if (!states_printed)
>> 110 curr = buf;
>> 111
>> 112 *curr++ = 0;
>> 113 }
>> 114 #endif
>> 115
>> 116 /*
>> 117 * Generally the error codes correspond to their respective errors, but there
>> 118 * are a few special cases.
>> 119 *
>> 120 * EUCLEAN: Any sort of corruption that we encounter. The tree-checker for
>> 121 * instance will return EUCLEAN if any of the blocks are corrupted in
>> 122 * a way that is problematic. We want to reserve EUCLEAN for these
>> 123 * sort of corruptions.
>> 124 *
>> 125 * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
>> 126 * need to use EROFS for this case. We will have no idea of the
>> 127 * original failure, that will have been reported at the time we tripped
>> 128 * over the error. Each subsequent error that doesn't have any context
>> 129 * of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
>> 130 */
>> 131 const char * __attribute_const__ btrfs_decode_error(int errno)
>> 132 {
>> 133 char *errstr = "unknown";
>> 134
>> 135 switch (errno) {
>> 136 case -ENOENT: /* -2 */
>> 137 errstr = "No such entry";
>> 138 break;
>> 139 case -EIO: /* -5 */
>> 140 errstr = "IO failure";
>> 141 break;
>> 142 case -ENOMEM: /* -12*/
>> 143 errstr = "Out of memory";
>> 144 break;
>> 145 case -EEXIST: /* -17 */
>> 146 errstr = "Object already exists";
>> 147 break;
>> 148 case -ENOSPC: /* -28 */
>> 149 errstr = "No space left";
>> 150 break;
>> 151 case -EROFS: /* -30 */
>> 152 errstr = "Readonly filesystem";
>> 153 break;
>> 154 case -EOPNOTSUPP: /* -95 */
>> 155 errstr = "Operation not supported";
>> 156 break;
>> 157 case -EUCLEAN: /* -117 */
>> 158 errstr = "Filesystem corrupted";
>> 159 break;
>> 160 case -EDQUOT: /* -122 */
>> 161 errstr = "Quota exceeded";
>> 162 break;
>> 163 }
>> 164
>> 165 return errstr;
>> 166 }
>> 167
>> 168 /*
>> 169 * __btrfs_handle_fs_error decodes expected errors from the caller and
>> 170 * invokes the appropriate error response.
>> 171 */
>> 172 __cold
>> 173 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
>> 174 unsigned int line, int errno, const char *fmt, ...)
>> 175 {
>> 176 struct super_block *sb = fs_info->sb;
>> 177 #ifdef CONFIG_PRINTK
>> 178 char statestr[STATE_STRING_BUF_LEN];
>> 179 const char *errstr;
>> 180 #endif
>> 181
>> 182 /*
>> 183 * Special case: if the error is EROFS, and we're already
>> 184 * under SB_RDONLY, then it is safe here.
>> 185 */
>> 186 if (errno == -EROFS && sb_rdonly(sb))
>> 187 return;
>> 188
>> 189 #ifdef CONFIG_PRINTK
>> 190 errstr = btrfs_decode_error(errno);
>> 191 btrfs_state_to_string(fs_info, statestr);
>> 192 if (fmt) {
>> 193 struct va_format vaf;
>> 194 va_list args;
>> 195
>> 196 va_start(args, fmt);
>> 197 vaf.fmt = fmt;
>> 198 vaf.va = &args;
>> 199
>> 200 pr_crit("BTRFS: error (device %s%s) in %s:%d: errno=%d %s (%pV)\n",
>> 201 sb->s_id, statestr, function, line, errno, errstr, &vaf);
>> 202 va_end(args);
>> 203 } else {
>> 204 pr_crit("BTRFS: error (device %s%s) in %s:%d: errno=%d %s\n",
>> 205 sb->s_id, statestr, function, line, errno, errstr);
>> 206 }
>> 207 #endif
>> 208
>> 209 /*
>> 210 * Today we only save the error info to memory. Long term we'll
>> 211 * also send it down to the disk
>> 212 */
>> 213 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
>> 214
>> 215 /* Don't go through full error handling during mount */
>> 216 if (!(sb->s_flags & SB_BORN))
>> 217 return;
>> 218
>> 219 if (sb_rdonly(sb))
>> 220 return;
>> 221
>> 222 btrfs_discard_stop(fs_info);
>> 223
>> 224 /* btrfs handle error by forcing the filesystem readonly */
>> 225 btrfs_set_sb_rdonly(sb);
>> 226 btrfs_info(fs_info, "forced readonly");
>> 227 /*
>> 228 * Note that a running device replace operation is not canceled here
>> 229 * although there is no way to update the progress. It would add the
>> 230 * risk of a deadlock, therefore the canceling is omitted. The only
>> 231 * penalty is that some I/O remains active until the procedure
>> 232 * completes. The next time when the filesystem is mounted writable
>> 233 * again, the device replace operation continues.
>> 234 */
76 } 235 }
77 236
78 /* Store the mount options related information !! 237 #ifdef CONFIG_PRINTK
79 struct btrfs_fs_context { !! 238 static const char * const logtypes[] = {
80 char *subvol_name; !! 239 "emergency",
81 u64 subvol_objectid; !! 240 "alert",
82 u64 max_inline; !! 241 "critical",
83 u32 commit_interval; !! 242 "error",
84 u32 metadata_ratio; !! 243 "warning",
85 u32 thread_pool_size; !! 244 "notice",
86 unsigned long long mount_opt; !! 245 "info",
87 unsigned long compress_type:4; !! 246 "debug",
88 unsigned int compress_level; <<
89 refcount_t refs; <<
90 }; 247 };
91 248
>> 249
>> 250 /*
>> 251 * Use one ratelimit state per log level so that a flood of less important
>> 252 * messages doesn't cause more important ones to be dropped.
>> 253 */
>> 254 static struct ratelimit_state printk_limits[] = {
>> 255 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 256 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 257 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 258 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 259 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 260 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 261 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 262 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 263 };
>> 264
>> 265 void __cold _btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
>> 266 {
>> 267 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
>> 268 struct va_format vaf;
>> 269 va_list args;
>> 270 int kern_level;
>> 271 const char *type = logtypes[4];
>> 272 struct ratelimit_state *ratelimit = &printk_limits[4];
>> 273
>> 274 va_start(args, fmt);
>> 275
>> 276 while ((kern_level = printk_get_level(fmt)) != 0) {
>> 277 size_t size = printk_skip_level(fmt) - fmt;
>> 278
>> 279 if (kern_level >= '' && kern_level <= '7') {
>> 280 memcpy(lvl, fmt, size);
>> 281 lvl[size] = '\0';
>> 282 type = logtypes[kern_level - ''];
>> 283 ratelimit = &printk_limits[kern_level - ''];
>> 284 }
>> 285 fmt += size;
>> 286 }
>> 287
>> 288 vaf.fmt = fmt;
>> 289 vaf.va = &args;
>> 290
>> 291 if (__ratelimit(ratelimit)) {
>> 292 if (fs_info) {
>> 293 char statestr[STATE_STRING_BUF_LEN];
>> 294
>> 295 btrfs_state_to_string(fs_info, statestr);
>> 296 _printk("%sBTRFS %s (device %s%s): %pV\n", lvl, type,
>> 297 fs_info->sb->s_id, statestr, &vaf);
>> 298 } else {
>> 299 _printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
>> 300 }
>> 301 }
>> 302
>> 303 va_end(args);
>> 304 }
>> 305 #endif
>> 306
>> 307 #if BITS_PER_LONG == 32
>> 308 void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
>> 309 {
>> 310 if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
>> 311 btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
>> 312 btrfs_warn(fs_info,
>> 313 "due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
>> 314 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
>> 315 btrfs_warn(fs_info,
>> 316 "please consider upgrading to 64bit kernel/hardware");
>> 317 }
>> 318 }
>> 319
>> 320 void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
>> 321 {
>> 322 if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
>> 323 btrfs_err(fs_info, "reached 32bit limit for logical addresses");
>> 324 btrfs_err(fs_info,
>> 325 "due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
>> 326 BTRFS_32BIT_MAX_FILE_SIZE >> 40);
>> 327 btrfs_err(fs_info,
>> 328 "please consider upgrading to 64bit kernel/hardware");
>> 329 }
>> 330 }
>> 331 #endif
>> 332
>> 333 /*
>> 334 * We only mark the transaction aborted and then set the file system read-only.
>> 335 * This will prevent new transactions from starting or trying to join this
>> 336 * one.
>> 337 *
>> 338 * This means that error recovery at the call site is limited to freeing
>> 339 * any local memory allocations and passing the error code up without
>> 340 * further cleanup. The transaction should complete as it normally would
>> 341 * in the call path but will return -EIO.
>> 342 *
>> 343 * We'll complete the cleanup in btrfs_end_transaction and
>> 344 * btrfs_commit_transaction.
>> 345 */
>> 346 __cold
>> 347 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
>> 348 const char *function,
>> 349 unsigned int line, int errno)
>> 350 {
>> 351 struct btrfs_fs_info *fs_info = trans->fs_info;
>> 352
>> 353 WRITE_ONCE(trans->aborted, errno);
>> 354 WRITE_ONCE(trans->transaction->aborted, errno);
>> 355 /* Wake up anybody who may be waiting on this transaction */
>> 356 wake_up(&fs_info->transaction_wait);
>> 357 wake_up(&fs_info->transaction_blocked_wait);
>> 358 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
>> 359 }
>> 360 /*
>> 361 * __btrfs_panic decodes unexpected, fatal errors from the caller,
>> 362 * issues an alert, and either panics or BUGs, depending on mount options.
>> 363 */
>> 364 __cold
>> 365 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
>> 366 unsigned int line, int errno, const char *fmt, ...)
>> 367 {
>> 368 char *s_id = "<unknown>";
>> 369 const char *errstr;
>> 370 struct va_format vaf = { .fmt = fmt };
>> 371 va_list args;
>> 372
>> 373 if (fs_info)
>> 374 s_id = fs_info->sb->s_id;
>> 375
>> 376 va_start(args, fmt);
>> 377 vaf.va = &args;
>> 378
>> 379 errstr = btrfs_decode_error(errno);
>> 380 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
>> 381 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
>> 382 s_id, function, line, &vaf, errno, errstr);
>> 383
>> 384 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
>> 385 function, line, &vaf, errno, errstr);
>> 386 va_end(args);
>> 387 /* Caller calls BUG() */
>> 388 }
>> 389
>> 390 static void btrfs_put_super(struct super_block *sb)
>> 391 {
>> 392 close_ctree(btrfs_sb(sb));
>> 393 }
>> 394
92 enum { 395 enum {
93 Opt_acl, !! 396 Opt_acl, Opt_noacl,
94 Opt_clear_cache, 397 Opt_clear_cache,
95 Opt_commit_interval, 398 Opt_commit_interval,
96 Opt_compress, 399 Opt_compress,
97 Opt_compress_force, 400 Opt_compress_force,
98 Opt_compress_force_type, 401 Opt_compress_force_type,
99 Opt_compress_type, 402 Opt_compress_type,
100 Opt_degraded, 403 Opt_degraded,
101 Opt_device, 404 Opt_device,
102 Opt_fatal_errors, 405 Opt_fatal_errors,
103 Opt_flushoncommit, !! 406 Opt_flushoncommit, Opt_noflushoncommit,
104 Opt_max_inline, 407 Opt_max_inline,
105 Opt_barrier, !! 408 Opt_barrier, Opt_nobarrier,
106 Opt_datacow, !! 409 Opt_datacow, Opt_nodatacow,
107 Opt_datasum, !! 410 Opt_datasum, Opt_nodatasum,
108 Opt_defrag, !! 411 Opt_defrag, Opt_nodefrag,
109 Opt_discard, !! 412 Opt_discard, Opt_nodiscard,
110 Opt_discard_mode, 413 Opt_discard_mode,
>> 414 Opt_norecovery,
111 Opt_ratio, 415 Opt_ratio,
112 Opt_rescan_uuid_tree, 416 Opt_rescan_uuid_tree,
113 Opt_skip_balance, 417 Opt_skip_balance,
114 Opt_space_cache, !! 418 Opt_space_cache, Opt_no_space_cache,
115 Opt_space_cache_version, 419 Opt_space_cache_version,
116 Opt_ssd, !! 420 Opt_ssd, Opt_nossd,
117 Opt_ssd_spread, !! 421 Opt_ssd_spread, Opt_nossd_spread,
118 Opt_subvol, 422 Opt_subvol,
119 Opt_subvol_empty, 423 Opt_subvol_empty,
120 Opt_subvolid, 424 Opt_subvolid,
121 Opt_thread_pool, 425 Opt_thread_pool,
122 Opt_treelog, !! 426 Opt_treelog, Opt_notreelog,
123 Opt_user_subvol_rm_allowed, 427 Opt_user_subvol_rm_allowed,
124 Opt_norecovery, <<
125 428
126 /* Rescue options */ 429 /* Rescue options */
127 Opt_rescue, 430 Opt_rescue,
128 Opt_usebackuproot, 431 Opt_usebackuproot,
129 Opt_nologreplay, 432 Opt_nologreplay,
>> 433 Opt_ignorebadroots,
>> 434 Opt_ignoredatacsums,
>> 435 Opt_rescue_all,
>> 436
>> 437 /* Deprecated options */
>> 438 Opt_recovery,
>> 439 Opt_inode_cache, Opt_noinode_cache,
130 440
131 /* Debugging options */ 441 /* Debugging options */
132 Opt_enospc_debug, !! 442 Opt_check_integrity,
>> 443 Opt_check_integrity_including_extent_data,
>> 444 Opt_check_integrity_print_mask,
>> 445 Opt_enospc_debug, Opt_noenospc_debug,
133 #ifdef CONFIG_BTRFS_DEBUG 446 #ifdef CONFIG_BTRFS_DEBUG
134 Opt_fragment, Opt_fragment_data, Opt_f !! 447 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
135 #endif 448 #endif
136 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 449 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
137 Opt_ref_verify, 450 Opt_ref_verify,
138 #endif 451 #endif
139 Opt_err, 452 Opt_err,
140 }; 453 };
141 454
142 enum { !! 455 static const match_table_t tokens = {
143 Opt_fatal_errors_panic, !! 456 {Opt_acl, "acl"},
144 Opt_fatal_errors_bug, !! 457 {Opt_noacl, "noacl"},
145 }; !! 458 {Opt_clear_cache, "clear_cache"},
146 !! 459 {Opt_commit_interval, "commit=%u"},
147 static const struct constant_table btrfs_param !! 460 {Opt_compress, "compress"},
148 { "panic", Opt_fatal_errors_panic }, !! 461 {Opt_compress_type, "compress=%s"},
149 { "bug", Opt_fatal_errors_bug }, !! 462 {Opt_compress_force, "compress-force"},
150 {} !! 463 {Opt_compress_force_type, "compress-force=%s"},
151 }; !! 464 {Opt_degraded, "degraded"},
>> 465 {Opt_device, "device=%s"},
>> 466 {Opt_fatal_errors, "fatal_errors=%s"},
>> 467 {Opt_flushoncommit, "flushoncommit"},
>> 468 {Opt_noflushoncommit, "noflushoncommit"},
>> 469 {Opt_inode_cache, "inode_cache"},
>> 470 {Opt_noinode_cache, "noinode_cache"},
>> 471 {Opt_max_inline, "max_inline=%s"},
>> 472 {Opt_barrier, "barrier"},
>> 473 {Opt_nobarrier, "nobarrier"},
>> 474 {Opt_datacow, "datacow"},
>> 475 {Opt_nodatacow, "nodatacow"},
>> 476 {Opt_datasum, "datasum"},
>> 477 {Opt_nodatasum, "nodatasum"},
>> 478 {Opt_defrag, "autodefrag"},
>> 479 {Opt_nodefrag, "noautodefrag"},
>> 480 {Opt_discard, "discard"},
>> 481 {Opt_discard_mode, "discard=%s"},
>> 482 {Opt_nodiscard, "nodiscard"},
>> 483 {Opt_norecovery, "norecovery"},
>> 484 {Opt_ratio, "metadata_ratio=%u"},
>> 485 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
>> 486 {Opt_skip_balance, "skip_balance"},
>> 487 {Opt_space_cache, "space_cache"},
>> 488 {Opt_no_space_cache, "nospace_cache"},
>> 489 {Opt_space_cache_version, "space_cache=%s"},
>> 490 {Opt_ssd, "ssd"},
>> 491 {Opt_nossd, "nossd"},
>> 492 {Opt_ssd_spread, "ssd_spread"},
>> 493 {Opt_nossd_spread, "nossd_spread"},
>> 494 {Opt_subvol, "subvol=%s"},
>> 495 {Opt_subvol_empty, "subvol="},
>> 496 {Opt_subvolid, "subvolid=%s"},
>> 497 {Opt_thread_pool, "thread_pool=%u"},
>> 498 {Opt_treelog, "treelog"},
>> 499 {Opt_notreelog, "notreelog"},
>> 500 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
152 501
153 enum { !! 502 /* Rescue options */
154 Opt_discard_sync, !! 503 {Opt_rescue, "rescue=%s"},
155 Opt_discard_async, <<
156 }; <<
157 <<
158 static const struct constant_table btrfs_param <<
159 { "sync", Opt_discard_sync }, <<
160 { "async", Opt_discard_async }, <<
161 {} <<
162 }; <<
163 <<
164 enum { <<
165 Opt_space_cache_v1, <<
166 Opt_space_cache_v2, <<
167 }; <<
168 <<
169 static const struct constant_table btrfs_param <<
170 { "v1", Opt_space_cache_v1 }, <<
171 { "v2", Opt_space_cache_v2 }, <<
172 {} <<
173 }; <<
174 <<
175 enum { <<
176 Opt_rescue_usebackuproot, <<
177 Opt_rescue_nologreplay, <<
178 Opt_rescue_ignorebadroots, <<
179 Opt_rescue_ignoredatacsums, <<
180 Opt_rescue_ignoremetacsums, <<
181 Opt_rescue_ignoresuperflags, <<
182 Opt_rescue_parameter_all, <<
183 }; <<
184 <<
185 static const struct constant_table btrfs_param <<
186 { "usebackuproot", Opt_rescue_usebacku <<
187 { "nologreplay", Opt_rescue_nologrepla <<
188 { "ignorebadroots", Opt_rescue_ignoreb <<
189 { "ibadroots", Opt_rescue_ignorebadroo <<
190 { "ignoredatacsums", Opt_rescue_ignore <<
191 { "ignoremetacsums", Opt_rescue_ignore <<
192 { "ignoresuperflags", Opt_rescue_ignor <<
193 { "idatacsums", Opt_rescue_ignoredatac <<
194 { "imetacsums", Opt_rescue_ignoremetac <<
195 { "isuperflags", Opt_rescue_ignoresupe <<
196 { "all", Opt_rescue_parameter_all }, <<
197 {} <<
198 }; <<
199 <<
200 #ifdef CONFIG_BTRFS_DEBUG <<
201 enum { <<
202 Opt_fragment_parameter_data, <<
203 Opt_fragment_parameter_metadata, <<
204 Opt_fragment_parameter_all, <<
205 }; <<
206 <<
207 static const struct constant_table btrfs_param <<
208 { "data", Opt_fragment_parameter_data <<
209 { "metadata", Opt_fragment_parameter_m <<
210 { "all", Opt_fragment_parameter_all }, <<
211 {} <<
212 }; <<
213 #endif <<
214 <<
215 static const struct fs_parameter_spec btrfs_fs <<
216 fsparam_flag_no("acl", Opt_acl), <<
217 fsparam_flag_no("autodefrag", Opt_defr <<
218 fsparam_flag_no("barrier", Opt_barrier <<
219 fsparam_flag("clear_cache", Opt_clear_ <<
220 fsparam_u32("commit", Opt_commit_inter <<
221 fsparam_flag("compress", Opt_compress) <<
222 fsparam_string("compress", Opt_compres <<
223 fsparam_flag("compress-force", Opt_com <<
224 fsparam_string("compress-force", Opt_c <<
225 fsparam_flag_no("datacow", Opt_datacow <<
226 fsparam_flag_no("datasum", Opt_datasum <<
227 fsparam_flag("degraded", Opt_degraded) <<
228 fsparam_string("device", Opt_device), <<
229 fsparam_flag_no("discard", Opt_discard <<
230 fsparam_enum("discard", Opt_discard_mo <<
231 fsparam_enum("fatal_errors", Opt_fatal <<
232 fsparam_flag_no("flushoncommit", Opt_f <<
233 fsparam_string("max_inline", Opt_max_i <<
234 fsparam_u32("metadata_ratio", Opt_rati <<
235 fsparam_flag("rescan_uuid_tree", Opt_r <<
236 fsparam_flag("skip_balance", Opt_skip_ <<
237 fsparam_flag_no("space_cache", Opt_spa <<
238 fsparam_enum("space_cache", Opt_space_ <<
239 fsparam_flag_no("ssd", Opt_ssd), <<
240 fsparam_flag_no("ssd_spread", Opt_ssd_ <<
241 fsparam_string("subvol", Opt_subvol), <<
242 fsparam_flag("subvol=", Opt_subvol_emp <<
243 fsparam_u64("subvolid", Opt_subvolid), <<
244 fsparam_u32("thread_pool", Opt_thread_ <<
245 fsparam_flag_no("treelog", Opt_treelog <<
246 fsparam_flag("user_subvol_rm_allowed", <<
247 <<
248 /* Rescue options. */ <<
249 fsparam_enum("rescue", Opt_rescue, btr <<
250 /* Deprecated, with alias rescue=nolog 504 /* Deprecated, with alias rescue=nologreplay */
251 __fsparam(NULL, "nologreplay", Opt_nol !! 505 {Opt_nologreplay, "nologreplay"},
252 /* Deprecated, with alias rescue=useba 506 /* Deprecated, with alias rescue=usebackuproot */
253 __fsparam(NULL, "usebackuproot", Opt_u !! 507 {Opt_usebackuproot, "usebackuproot"},
254 /* For compatibility only, alias for " <<
255 fsparam_flag("norecovery", Opt_norecov <<
256 508
257 /* Debugging options. */ !! 509 /* Deprecated options */
258 fsparam_flag_no("enospc_debug", Opt_en !! 510 {Opt_recovery, "recovery"},
>> 511
>> 512 /* Debugging options */
>> 513 {Opt_check_integrity, "check_int"},
>> 514 {Opt_check_integrity_including_extent_data, "check_int_data"},
>> 515 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
>> 516 {Opt_enospc_debug, "enospc_debug"},
>> 517 {Opt_noenospc_debug, "noenospc_debug"},
259 #ifdef CONFIG_BTRFS_DEBUG 518 #ifdef CONFIG_BTRFS_DEBUG
260 fsparam_enum("fragment", Opt_fragment, !! 519 {Opt_fragment_data, "fragment=data"},
>> 520 {Opt_fragment_metadata, "fragment=metadata"},
>> 521 {Opt_fragment_all, "fragment=all"},
261 #endif 522 #endif
262 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 523 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
263 fsparam_flag("ref_verify", Opt_ref_ver !! 524 {Opt_ref_verify, "ref_verify"},
264 #endif 525 #endif
265 {} !! 526 {Opt_err, NULL},
>> 527 };
>> 528
>> 529 static const match_table_t rescue_tokens = {
>> 530 {Opt_usebackuproot, "usebackuproot"},
>> 531 {Opt_nologreplay, "nologreplay"},
>> 532 {Opt_ignorebadroots, "ignorebadroots"},
>> 533 {Opt_ignorebadroots, "ibadroots"},
>> 534 {Opt_ignoredatacsums, "ignoredatacsums"},
>> 535 {Opt_ignoredatacsums, "idatacsums"},
>> 536 {Opt_rescue_all, "all"},
>> 537 {Opt_err, NULL},
266 }; 538 };
267 539
268 /* No support for restricting writes to btrfs !! 540 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
269 static inline blk_mode_t btrfs_open_mode(struc !! 541 const char *opt_name)
270 { 542 {
271 return sb_open_mode(fc->sb_flags) & ~B !! 543 if (fs_info->mount_opt & opt) {
>> 544 btrfs_err(fs_info, "%s must be used with ro mount option",
>> 545 opt_name);
>> 546 return true;
>> 547 }
>> 548 return false;
272 } 549 }
273 550
274 static int btrfs_parse_param(struct fs_context !! 551 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
275 { 552 {
276 struct btrfs_fs_context *ctx = fc->fs_ !! 553 char *opts;
277 struct fs_parse_result result; !! 554 char *orig;
278 int opt; !! 555 char *p;
>> 556 substring_t args[MAX_OPT_ARGS];
>> 557 int ret = 0;
279 558
280 opt = fs_parse(fc, btrfs_fs_parameters !! 559 opts = kstrdup(options, GFP_KERNEL);
281 if (opt < 0) !! 560 if (!opts)
282 return opt; !! 561 return -ENOMEM;
>> 562 orig = opts;
283 563
284 switch (opt) { !! 564 while ((p = strsep(&opts, ":")) != NULL) {
285 case Opt_degraded: !! 565 int token;
286 btrfs_set_opt(ctx->mount_opt, <<
287 break; <<
288 case Opt_subvol_empty: <<
289 /* <<
290 * This exists because we used <<
291 * keeping it to maintain ABI. <<
292 * empty subvol= again"). <<
293 */ <<
294 break; <<
295 case Opt_subvol: <<
296 kfree(ctx->subvol_name); <<
297 ctx->subvol_name = kstrdup(par <<
298 if (!ctx->subvol_name) <<
299 return -ENOMEM; <<
300 break; <<
301 case Opt_subvolid: <<
302 ctx->subvol_objectid = result. <<
303 566
304 /* subvolid=0 means give me th !! 567 if (!*p)
305 if (!ctx->subvol_objectid) !! 568 continue;
306 ctx->subvol_objectid = !! 569 token = match_token(p, rescue_tokens, args);
307 break; !! 570 switch (token){
308 case Opt_device: { !! 571 case Opt_usebackuproot:
309 struct btrfs_device *device; !! 572 btrfs_info(info,
310 blk_mode_t mode = btrfs_open_m !! 573 "trying to use backup root at mount time");
>> 574 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
>> 575 break;
>> 576 case Opt_nologreplay:
>> 577 btrfs_set_and_info(info, NOLOGREPLAY,
>> 578 "disabling log replay at mount time");
>> 579 break;
>> 580 case Opt_ignorebadroots:
>> 581 btrfs_set_and_info(info, IGNOREBADROOTS,
>> 582 "ignoring bad roots");
>> 583 break;
>> 584 case Opt_ignoredatacsums:
>> 585 btrfs_set_and_info(info, IGNOREDATACSUMS,
>> 586 "ignoring data csums");
>> 587 break;
>> 588 case Opt_rescue_all:
>> 589 btrfs_info(info, "enabling all of the rescue options");
>> 590 btrfs_set_and_info(info, IGNOREDATACSUMS,
>> 591 "ignoring data csums");
>> 592 btrfs_set_and_info(info, IGNOREBADROOTS,
>> 593 "ignoring bad roots");
>> 594 btrfs_set_and_info(info, NOLOGREPLAY,
>> 595 "disabling log replay at mount time");
>> 596 break;
>> 597 case Opt_err:
>> 598 btrfs_info(info, "unrecognized rescue option '%s'", p);
>> 599 ret = -EINVAL;
>> 600 goto out;
>> 601 default:
>> 602 break;
>> 603 }
311 604
312 mutex_lock(&uuid_mutex); <<
313 device = btrfs_scan_one_device <<
314 mutex_unlock(&uuid_mutex); <<
315 if (IS_ERR(device)) <<
316 return PTR_ERR(device) <<
317 break; <<
318 } 605 }
319 case Opt_datasum: !! 606 out:
320 if (result.negated) { !! 607 kfree(orig);
321 btrfs_set_opt(ctx->mou !! 608 return ret;
322 } else { !! 609 }
323 btrfs_clear_opt(ctx->m <<
324 btrfs_clear_opt(ctx->m <<
325 } <<
326 break; <<
327 case Opt_datacow: <<
328 if (result.negated) { <<
329 btrfs_clear_opt(ctx->m <<
330 btrfs_clear_opt(ctx->m <<
331 btrfs_set_opt(ctx->mou <<
332 btrfs_set_opt(ctx->mou <<
333 } else { <<
334 btrfs_clear_opt(ctx->m <<
335 } <<
336 break; <<
337 case Opt_compress_force: <<
338 case Opt_compress_force_type: <<
339 btrfs_set_opt(ctx->mount_opt, <<
340 fallthrough; <<
341 case Opt_compress: <<
342 case Opt_compress_type: <<
343 /* <<
344 * Provide the same semantics <<
345 * context, specifying the "co <<
346 * "force-compress" without th <<
347 * "compress-force=[no|none]" <<
348 */ <<
349 if (opt != Opt_compress_force <<
350 btrfs_clear_opt(ctx->m <<
351 610
352 if (opt == Opt_compress || opt !! 611 /*
353 ctx->compress_type = B !! 612 * Regular mount options parser. Everything that is needed only when
354 ctx->compress_level = !! 613 * reading in a new superblock is parsed here.
355 btrfs_set_opt(ctx->mou !! 614 * XXX JDM: This needs to be cleaned up for remount.
356 btrfs_clear_opt(ctx->m !! 615 */
357 btrfs_clear_opt(ctx->m !! 616 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
358 } else if (strncmp(param->stri !! 617 unsigned long new_flags)
359 ctx->compress_type = B !! 618 {
360 ctx->compress_level = !! 619 substring_t args[MAX_OPT_ARGS];
361 btrfs_compress !! 620 char *p, *num;
362 !! 621 int intarg;
363 btrfs_set_opt(ctx->mou !! 622 int ret = 0;
364 btrfs_clear_opt(ctx->m !! 623 char *compress_type;
365 btrfs_clear_opt(ctx->m !! 624 bool compress_force = false;
366 } else if (strncmp(param->stri !! 625 enum btrfs_compression_type saved_compress_type;
367 ctx->compress_type = B !! 626 int saved_compress_level;
368 ctx->compress_level = !! 627 bool saved_compress_force;
369 btrfs_set_opt(ctx->mou !! 628 int no_compress = 0;
370 btrfs_clear_opt(ctx->m !! 629 const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
371 btrfs_clear_opt(ctx->m !! 630
372 } else if (strncmp(param->stri !! 631 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
373 ctx->compress_type = B !! 632 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
374 ctx->compress_level = !! 633 else if (btrfs_free_space_cache_v1_active(info)) {
375 btrfs_compress !! 634 if (btrfs_is_zoned(info)) {
376 !! 635 btrfs_info(info,
377 btrfs_set_opt(ctx->mou !! 636 "zoned: clearing existing space cache");
378 btrfs_clear_opt(ctx->m !! 637 btrfs_set_super_cache_generation(info->super_copy, 0);
379 btrfs_clear_opt(ctx->m <<
380 } else if (strncmp(param->stri <<
381 ctx->compress_level = <<
382 ctx->compress_type = 0 <<
383 btrfs_clear_opt(ctx->m <<
384 btrfs_clear_opt(ctx->m <<
385 } else { <<
386 btrfs_err(NULL, "unrec <<
387 param->strin <<
388 return -EINVAL; <<
389 } <<
390 break; <<
391 case Opt_ssd: <<
392 if (result.negated) { <<
393 btrfs_set_opt(ctx->mou <<
394 btrfs_clear_opt(ctx->m <<
395 btrfs_clear_opt(ctx->m <<
396 } else { <<
397 btrfs_set_opt(ctx->mou <<
398 btrfs_clear_opt(ctx->m <<
399 } <<
400 break; <<
401 case Opt_ssd_spread: <<
402 if (result.negated) { <<
403 btrfs_clear_opt(ctx->m <<
404 } else { 638 } else {
405 btrfs_set_opt(ctx->mou !! 639 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
406 btrfs_set_opt(ctx->mou <<
407 btrfs_clear_opt(ctx->m <<
408 } <<
409 break; <<
410 case Opt_barrier: <<
411 if (result.negated) <<
412 btrfs_set_opt(ctx->mou <<
413 else <<
414 btrfs_clear_opt(ctx->m <<
415 break; <<
416 case Opt_thread_pool: <<
417 if (result.uint_32 == 0) { <<
418 btrfs_err(NULL, "inval <<
419 return -EINVAL; <<
420 } 640 }
421 ctx->thread_pool_size = result !! 641 }
422 break; !! 642
423 case Opt_max_inline: !! 643 /*
424 ctx->max_inline = memparse(par !! 644 * Even the options are empty, we still need to do extra check
425 break; !! 645 * against new flags
426 case Opt_acl: !! 646 */
427 if (result.negated) { !! 647 if (!options)
428 fc->sb_flags &= ~SB_PO !! 648 goto check;
429 } else { !! 649
>> 650 while ((p = strsep(&options, ",")) != NULL) {
>> 651 int token;
>> 652 if (!*p)
>> 653 continue;
>> 654
>> 655 token = match_token(p, tokens, args);
>> 656 switch (token) {
>> 657 case Opt_degraded:
>> 658 btrfs_info(info, "allowing degraded mounts");
>> 659 btrfs_set_opt(info->mount_opt, DEGRADED);
>> 660 break;
>> 661 case Opt_subvol:
>> 662 case Opt_subvol_empty:
>> 663 case Opt_subvolid:
>> 664 case Opt_device:
>> 665 /*
>> 666 * These are parsed by btrfs_parse_subvol_options or
>> 667 * btrfs_parse_device_options and can be ignored here.
>> 668 */
>> 669 break;
>> 670 case Opt_nodatasum:
>> 671 btrfs_set_and_info(info, NODATASUM,
>> 672 "setting nodatasum");
>> 673 break;
>> 674 case Opt_datasum:
>> 675 if (btrfs_test_opt(info, NODATASUM)) {
>> 676 if (btrfs_test_opt(info, NODATACOW))
>> 677 btrfs_info(info,
>> 678 "setting datasum, datacow enabled");
>> 679 else
>> 680 btrfs_info(info, "setting datasum");
>> 681 }
>> 682 btrfs_clear_opt(info->mount_opt, NODATACOW);
>> 683 btrfs_clear_opt(info->mount_opt, NODATASUM);
>> 684 break;
>> 685 case Opt_nodatacow:
>> 686 if (!btrfs_test_opt(info, NODATACOW)) {
>> 687 if (!btrfs_test_opt(info, COMPRESS) ||
>> 688 !btrfs_test_opt(info, FORCE_COMPRESS)) {
>> 689 btrfs_info(info,
>> 690 "setting nodatacow, compression disabled");
>> 691 } else {
>> 692 btrfs_info(info, "setting nodatacow");
>> 693 }
>> 694 }
>> 695 btrfs_clear_opt(info->mount_opt, COMPRESS);
>> 696 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
>> 697 btrfs_set_opt(info->mount_opt, NODATACOW);
>> 698 btrfs_set_opt(info->mount_opt, NODATASUM);
>> 699 break;
>> 700 case Opt_datacow:
>> 701 btrfs_clear_and_info(info, NODATACOW,
>> 702 "setting datacow");
>> 703 break;
>> 704 case Opt_compress_force:
>> 705 case Opt_compress_force_type:
>> 706 compress_force = true;
>> 707 fallthrough;
>> 708 case Opt_compress:
>> 709 case Opt_compress_type:
>> 710 saved_compress_type = btrfs_test_opt(info,
>> 711 COMPRESS) ?
>> 712 info->compress_type : BTRFS_COMPRESS_NONE;
>> 713 saved_compress_force =
>> 714 btrfs_test_opt(info, FORCE_COMPRESS);
>> 715 saved_compress_level = info->compress_level;
>> 716 if (token == Opt_compress ||
>> 717 token == Opt_compress_force ||
>> 718 strncmp(args[0].from, "zlib", 4) == 0) {
>> 719 compress_type = "zlib";
>> 720
>> 721 info->compress_type = BTRFS_COMPRESS_ZLIB;
>> 722 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
>> 723 /*
>> 724 * args[0] contains uninitialized data since
>> 725 * for these tokens we don't expect any
>> 726 * parameter.
>> 727 */
>> 728 if (token != Opt_compress &&
>> 729 token != Opt_compress_force)
>> 730 info->compress_level =
>> 731 btrfs_compress_str2level(
>> 732 BTRFS_COMPRESS_ZLIB,
>> 733 args[0].from + 4);
>> 734 btrfs_set_opt(info->mount_opt, COMPRESS);
>> 735 btrfs_clear_opt(info->mount_opt, NODATACOW);
>> 736 btrfs_clear_opt(info->mount_opt, NODATASUM);
>> 737 no_compress = 0;
>> 738 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
>> 739 compress_type = "lzo";
>> 740 info->compress_type = BTRFS_COMPRESS_LZO;
>> 741 info->compress_level = 0;
>> 742 btrfs_set_opt(info->mount_opt, COMPRESS);
>> 743 btrfs_clear_opt(info->mount_opt, NODATACOW);
>> 744 btrfs_clear_opt(info->mount_opt, NODATASUM);
>> 745 btrfs_set_fs_incompat(info, COMPRESS_LZO);
>> 746 no_compress = 0;
>> 747 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
>> 748 compress_type = "zstd";
>> 749 info->compress_type = BTRFS_COMPRESS_ZSTD;
>> 750 info->compress_level =
>> 751 btrfs_compress_str2level(
>> 752 BTRFS_COMPRESS_ZSTD,
>> 753 args[0].from + 4);
>> 754 btrfs_set_opt(info->mount_opt, COMPRESS);
>> 755 btrfs_clear_opt(info->mount_opt, NODATACOW);
>> 756 btrfs_clear_opt(info->mount_opt, NODATASUM);
>> 757 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
>> 758 no_compress = 0;
>> 759 } else if (strncmp(args[0].from, "no", 2) == 0) {
>> 760 compress_type = "no";
>> 761 info->compress_level = 0;
>> 762 info->compress_type = 0;
>> 763 btrfs_clear_opt(info->mount_opt, COMPRESS);
>> 764 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
>> 765 compress_force = false;
>> 766 no_compress++;
>> 767 } else {
>> 768 btrfs_err(info, "unrecognized compression value %s",
>> 769 args[0].from);
>> 770 ret = -EINVAL;
>> 771 goto out;
>> 772 }
>> 773
>> 774 if (compress_force) {
>> 775 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
>> 776 } else {
>> 777 /*
>> 778 * If we remount from compress-force=xxx to
>> 779 * compress=xxx, we need clear FORCE_COMPRESS
>> 780 * flag, otherwise, there is no way for users
>> 781 * to disable forcible compression separately.
>> 782 */
>> 783 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
>> 784 }
>> 785 if (no_compress == 1) {
>> 786 btrfs_info(info, "use no compression");
>> 787 } else if ((info->compress_type != saved_compress_type) ||
>> 788 (compress_force != saved_compress_force) ||
>> 789 (info->compress_level != saved_compress_level)) {
>> 790 btrfs_info(info, "%s %s compression, level %d",
>> 791 (compress_force) ? "force" : "use",
>> 792 compress_type, info->compress_level);
>> 793 }
>> 794 compress_force = false;
>> 795 break;
>> 796 case Opt_ssd:
>> 797 btrfs_set_and_info(info, SSD,
>> 798 "enabling ssd optimizations");
>> 799 btrfs_clear_opt(info->mount_opt, NOSSD);
>> 800 break;
>> 801 case Opt_ssd_spread:
>> 802 btrfs_set_and_info(info, SSD,
>> 803 "enabling ssd optimizations");
>> 804 btrfs_set_and_info(info, SSD_SPREAD,
>> 805 "using spread ssd allocation scheme");
>> 806 btrfs_clear_opt(info->mount_opt, NOSSD);
>> 807 break;
>> 808 case Opt_nossd:
>> 809 btrfs_set_opt(info->mount_opt, NOSSD);
>> 810 btrfs_clear_and_info(info, SSD,
>> 811 "not using ssd optimizations");
>> 812 fallthrough;
>> 813 case Opt_nossd_spread:
>> 814 btrfs_clear_and_info(info, SSD_SPREAD,
>> 815 "not using spread ssd allocation scheme");
>> 816 break;
>> 817 case Opt_barrier:
>> 818 btrfs_clear_and_info(info, NOBARRIER,
>> 819 "turning on barriers");
>> 820 break;
>> 821 case Opt_nobarrier:
>> 822 btrfs_set_and_info(info, NOBARRIER,
>> 823 "turning off barriers");
>> 824 break;
>> 825 case Opt_thread_pool:
>> 826 ret = match_int(&args[0], &intarg);
>> 827 if (ret) {
>> 828 btrfs_err(info, "unrecognized thread_pool value %s",
>> 829 args[0].from);
>> 830 goto out;
>> 831 } else if (intarg == 0) {
>> 832 btrfs_err(info, "invalid value 0 for thread_pool");
>> 833 ret = -EINVAL;
>> 834 goto out;
>> 835 }
>> 836 info->thread_pool_size = intarg;
>> 837 break;
>> 838 case Opt_max_inline:
>> 839 num = match_strdup(&args[0]);
>> 840 if (num) {
>> 841 info->max_inline = memparse(num, NULL);
>> 842 kfree(num);
>> 843
>> 844 if (info->max_inline) {
>> 845 info->max_inline = min_t(u64,
>> 846 info->max_inline,
>> 847 info->sectorsize);
>> 848 }
>> 849 btrfs_info(info, "max_inline at %llu",
>> 850 info->max_inline);
>> 851 } else {
>> 852 ret = -ENOMEM;
>> 853 goto out;
>> 854 }
>> 855 break;
>> 856 case Opt_acl:
430 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 857 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
431 fc->sb_flags |= SB_POS !! 858 info->sb->s_flags |= SB_POSIXACL;
>> 859 break;
432 #else 860 #else
433 btrfs_err(NULL, "suppo !! 861 btrfs_err(info, "support for ACL not compiled in!");
434 return -EINVAL; !! 862 ret = -EINVAL;
>> 863 goto out;
435 #endif 864 #endif
436 } !! 865 case Opt_noacl:
437 /* !! 866 info->sb->s_flags &= ~SB_POSIXACL;
438 * VFS limits the ability to t !! 867 break;
439 * despite every file system a !! 868 case Opt_notreelog:
440 * an oversight since we all d !! 869 btrfs_set_and_info(info, NOTREELOG,
441 * remounting. So don't set t !! 870 "disabling tree log");
442 * btrfs_reconfigure and do th !! 871 break;
443 */ !! 872 case Opt_treelog:
444 if (fc->purpose != FS_CONTEXT_ !! 873 btrfs_clear_and_info(info, NOTREELOG,
445 fc->sb_flags_mask |= S !! 874 "enabling tree log");
446 break; !! 875 break;
447 case Opt_treelog: !! 876 case Opt_norecovery:
448 if (result.negated) !! 877 case Opt_nologreplay:
449 btrfs_set_opt(ctx->mou !! 878 btrfs_warn(info,
450 else <<
451 btrfs_clear_opt(ctx->m <<
452 break; <<
453 case Opt_nologreplay: <<
454 btrfs_warn(NULL, <<
455 "'nologreplay' is deprecated, 879 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
456 btrfs_set_opt(ctx->mount_opt, !! 880 btrfs_set_and_info(info, NOLOGREPLAY,
457 break; !! 881 "disabling log replay at mount time");
458 case Opt_norecovery: <<
459 btrfs_info(NULL, <<
460 "'norecovery' is for compatibility only, recom <<
461 btrfs_set_opt(ctx->mount_opt, <<
462 break; <<
463 case Opt_flushoncommit: <<
464 if (result.negated) <<
465 btrfs_clear_opt(ctx->m <<
466 else <<
467 btrfs_set_opt(ctx->mou <<
468 break; <<
469 case Opt_ratio: <<
470 ctx->metadata_ratio = result.u <<
471 break; <<
472 case Opt_discard: <<
473 if (result.negated) { <<
474 btrfs_clear_opt(ctx->m <<
475 btrfs_clear_opt(ctx->m <<
476 btrfs_set_opt(ctx->mou <<
477 } else { <<
478 btrfs_set_opt(ctx->mou <<
479 btrfs_clear_opt(ctx->m <<
480 } <<
481 break; <<
482 case Opt_discard_mode: <<
483 switch (result.uint_32) { <<
484 case Opt_discard_sync: <<
485 btrfs_clear_opt(ctx->m <<
486 btrfs_set_opt(ctx->mou <<
487 break; <<
488 case Opt_discard_async: <<
489 btrfs_clear_opt(ctx->m <<
490 btrfs_set_opt(ctx->mou <<
491 break; 882 break;
492 default: !! 883 case Opt_flushoncommit:
493 btrfs_err(NULL, "unrec !! 884 btrfs_set_and_info(info, FLUSHONCOMMIT,
494 param->key); !! 885 "turning on flush-on-commit");
495 return -EINVAL; <<
496 } <<
497 btrfs_clear_opt(ctx->mount_opt <<
498 break; <<
499 case Opt_space_cache: <<
500 if (result.negated) { <<
501 btrfs_set_opt(ctx->mou <<
502 btrfs_clear_opt(ctx->m <<
503 btrfs_clear_opt(ctx->m <<
504 } else { <<
505 btrfs_clear_opt(ctx->m <<
506 btrfs_set_opt(ctx->mou <<
507 } <<
508 break; <<
509 case Opt_space_cache_version: <<
510 switch (result.uint_32) { <<
511 case Opt_space_cache_v1: <<
512 btrfs_set_opt(ctx->mou <<
513 btrfs_clear_opt(ctx->m <<
514 break; <<
515 case Opt_space_cache_v2: <<
516 btrfs_clear_opt(ctx->m <<
517 btrfs_set_opt(ctx->mou <<
518 break; 886 break;
519 default: !! 887 case Opt_noflushoncommit:
520 btrfs_err(NULL, "unrec !! 888 btrfs_clear_and_info(info, FLUSHONCOMMIT,
521 param->key); !! 889 "turning off flush-on-commit");
522 return -EINVAL; <<
523 } <<
524 break; <<
525 case Opt_rescan_uuid_tree: <<
526 btrfs_set_opt(ctx->mount_opt, <<
527 break; <<
528 case Opt_clear_cache: <<
529 btrfs_set_opt(ctx->mount_opt, <<
530 break; <<
531 case Opt_user_subvol_rm_allowed: <<
532 btrfs_set_opt(ctx->mount_opt, <<
533 break; <<
534 case Opt_enospc_debug: <<
535 if (result.negated) <<
536 btrfs_clear_opt(ctx->m <<
537 else <<
538 btrfs_set_opt(ctx->mou <<
539 break; <<
540 case Opt_defrag: <<
541 if (result.negated) <<
542 btrfs_clear_opt(ctx->m <<
543 else <<
544 btrfs_set_opt(ctx->mou <<
545 break; <<
546 case Opt_usebackuproot: <<
547 btrfs_warn(NULL, <<
548 "'usebackuproot' is <<
549 btrfs_set_opt(ctx->mount_opt, <<
550 <<
551 /* If we're loading the backup <<
552 btrfs_set_opt(ctx->mount_opt, <<
553 break; <<
554 case Opt_skip_balance: <<
555 btrfs_set_opt(ctx->mount_opt, <<
556 break; <<
557 case Opt_fatal_errors: <<
558 switch (result.uint_32) { <<
559 case Opt_fatal_errors_panic: <<
560 btrfs_set_opt(ctx->mou <<
561 break; 890 break;
562 case Opt_fatal_errors_bug: !! 891 case Opt_ratio:
563 btrfs_clear_opt(ctx->m !! 892 ret = match_int(&args[0], &intarg);
>> 893 if (ret) {
>> 894 btrfs_err(info, "unrecognized metadata_ratio value %s",
>> 895 args[0].from);
>> 896 goto out;
>> 897 }
>> 898 info->metadata_ratio = intarg;
>> 899 btrfs_info(info, "metadata ratio %u",
>> 900 info->metadata_ratio);
564 break; 901 break;
565 default: !! 902 case Opt_discard:
566 btrfs_err(NULL, "unrec !! 903 case Opt_discard_mode:
567 param->key); !! 904 if (token == Opt_discard ||
568 return -EINVAL; !! 905 strcmp(args[0].from, "sync") == 0) {
569 } !! 906 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
570 break; !! 907 btrfs_set_and_info(info, DISCARD_SYNC,
571 case Opt_commit_interval: !! 908 "turning on sync discard");
572 ctx->commit_interval = result. !! 909 } else if (strcmp(args[0].from, "async") == 0) {
573 if (ctx->commit_interval == 0) !! 910 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
574 ctx->commit_interval = !! 911 btrfs_set_and_info(info, DISCARD_ASYNC,
575 break; !! 912 "turning on async discard");
576 case Opt_rescue: !! 913 } else {
577 switch (result.uint_32) { !! 914 btrfs_err(info, "unrecognized discard mode value %s",
578 case Opt_rescue_usebackuproot: !! 915 args[0].from);
579 btrfs_set_opt(ctx->mou !! 916 ret = -EINVAL;
580 break; !! 917 goto out;
581 case Opt_rescue_nologreplay: !! 918 }
582 btrfs_set_opt(ctx->mou !! 919 break;
583 break; !! 920 case Opt_nodiscard:
584 case Opt_rescue_ignorebadroots !! 921 btrfs_clear_and_info(info, DISCARD_SYNC,
585 btrfs_set_opt(ctx->mou !! 922 "turning off discard");
586 break; !! 923 btrfs_clear_and_info(info, DISCARD_ASYNC,
587 case Opt_rescue_ignoredatacsum !! 924 "turning off async discard");
588 btrfs_set_opt(ctx->mou !! 925 break;
589 break; !! 926 case Opt_space_cache:
590 case Opt_rescue_ignoremetacsum !! 927 case Opt_space_cache_version:
591 btrfs_set_opt(ctx->mou !! 928 /*
592 break; !! 929 * We already set FREE_SPACE_TREE above because we have
593 case Opt_rescue_ignoresuperfla !! 930 * compat_ro(FREE_SPACE_TREE) set, and we aren't going
594 btrfs_set_opt(ctx->mou !! 931 * to allow v1 to be set for extent tree v2, simply
595 break; !! 932 * ignore this setting if we're extent tree v2.
596 case Opt_rescue_parameter_all: !! 933 */
597 btrfs_set_opt(ctx->mou !! 934 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
598 btrfs_set_opt(ctx->mou !! 935 break;
599 btrfs_set_opt(ctx->mou !! 936 if (token == Opt_space_cache ||
600 btrfs_set_opt(ctx->mou !! 937 strcmp(args[0].from, "v1") == 0) {
601 btrfs_set_opt(ctx->mou !! 938 btrfs_clear_opt(info->mount_opt,
>> 939 FREE_SPACE_TREE);
>> 940 btrfs_set_and_info(info, SPACE_CACHE,
>> 941 "enabling disk space caching");
>> 942 } else if (strcmp(args[0].from, "v2") == 0) {
>> 943 btrfs_clear_opt(info->mount_opt,
>> 944 SPACE_CACHE);
>> 945 btrfs_set_and_info(info, FREE_SPACE_TREE,
>> 946 "enabling free space tree");
>> 947 } else {
>> 948 btrfs_err(info, "unrecognized space_cache value %s",
>> 949 args[0].from);
>> 950 ret = -EINVAL;
>> 951 goto out;
>> 952 }
>> 953 break;
>> 954 case Opt_rescan_uuid_tree:
>> 955 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
>> 956 break;
>> 957 case Opt_no_space_cache:
>> 958 /*
>> 959 * We cannot operate without the free space tree with
>> 960 * extent tree v2, ignore this option.
>> 961 */
>> 962 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
>> 963 break;
>> 964 if (btrfs_test_opt(info, SPACE_CACHE)) {
>> 965 btrfs_clear_and_info(info, SPACE_CACHE,
>> 966 "disabling disk space caching");
>> 967 }
>> 968 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
>> 969 btrfs_clear_and_info(info, FREE_SPACE_TREE,
>> 970 "disabling free space tree");
>> 971 }
>> 972 break;
>> 973 case Opt_inode_cache:
>> 974 case Opt_noinode_cache:
>> 975 btrfs_warn(info,
>> 976 "the 'inode_cache' option is deprecated and has no effect since 5.11");
>> 977 break;
>> 978 case Opt_clear_cache:
>> 979 /*
>> 980 * We cannot clear the free space tree with extent tree
>> 981 * v2, ignore this option.
>> 982 */
>> 983 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
>> 984 break;
>> 985 btrfs_set_and_info(info, CLEAR_CACHE,
>> 986 "force clearing of disk cache");
>> 987 break;
>> 988 case Opt_user_subvol_rm_allowed:
>> 989 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
>> 990 break;
>> 991 case Opt_enospc_debug:
>> 992 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
>> 993 break;
>> 994 case Opt_noenospc_debug:
>> 995 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
>> 996 break;
>> 997 case Opt_defrag:
>> 998 btrfs_set_and_info(info, AUTO_DEFRAG,
>> 999 "enabling auto defrag");
>> 1000 break;
>> 1001 case Opt_nodefrag:
>> 1002 btrfs_clear_and_info(info, AUTO_DEFRAG,
>> 1003 "disabling auto defrag");
>> 1004 break;
>> 1005 case Opt_recovery:
>> 1006 case Opt_usebackuproot:
>> 1007 btrfs_warn(info,
>> 1008 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
>> 1009 token == Opt_recovery ? "recovery" :
>> 1010 "usebackuproot");
>> 1011 btrfs_info(info,
>> 1012 "trying to use backup root at mount time");
>> 1013 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
>> 1014 break;
>> 1015 case Opt_skip_balance:
>> 1016 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
>> 1017 break;
>> 1018 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 1019 case Opt_check_integrity_including_extent_data:
>> 1020 btrfs_info(info,
>> 1021 "enabling check integrity including extent data");
>> 1022 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
>> 1023 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
>> 1024 break;
>> 1025 case Opt_check_integrity:
>> 1026 btrfs_info(info, "enabling check integrity");
>> 1027 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
>> 1028 break;
>> 1029 case Opt_check_integrity_print_mask:
>> 1030 ret = match_int(&args[0], &intarg);
>> 1031 if (ret) {
>> 1032 btrfs_err(info,
>> 1033 "unrecognized check_integrity_print_mask value %s",
>> 1034 args[0].from);
>> 1035 goto out;
>> 1036 }
>> 1037 info->check_integrity_print_mask = intarg;
>> 1038 btrfs_info(info, "check_integrity_print_mask 0x%x",
>> 1039 info->check_integrity_print_mask);
>> 1040 break;
>> 1041 #else
>> 1042 case Opt_check_integrity_including_extent_data:
>> 1043 case Opt_check_integrity:
>> 1044 case Opt_check_integrity_print_mask:
>> 1045 btrfs_err(info,
>> 1046 "support for check_integrity* not compiled in!");
>> 1047 ret = -EINVAL;
>> 1048 goto out;
>> 1049 #endif
>> 1050 case Opt_fatal_errors:
>> 1051 if (strcmp(args[0].from, "panic") == 0) {
>> 1052 btrfs_set_opt(info->mount_opt,
>> 1053 PANIC_ON_FATAL_ERROR);
>> 1054 } else if (strcmp(args[0].from, "bug") == 0) {
>> 1055 btrfs_clear_opt(info->mount_opt,
>> 1056 PANIC_ON_FATAL_ERROR);
>> 1057 } else {
>> 1058 btrfs_err(info, "unrecognized fatal_errors value %s",
>> 1059 args[0].from);
>> 1060 ret = -EINVAL;
>> 1061 goto out;
>> 1062 }
>> 1063 break;
>> 1064 case Opt_commit_interval:
>> 1065 intarg = 0;
>> 1066 ret = match_int(&args[0], &intarg);
>> 1067 if (ret) {
>> 1068 btrfs_err(info, "unrecognized commit_interval value %s",
>> 1069 args[0].from);
>> 1070 ret = -EINVAL;
>> 1071 goto out;
>> 1072 }
>> 1073 if (intarg == 0) {
>> 1074 btrfs_info(info,
>> 1075 "using default commit interval %us",
>> 1076 BTRFS_DEFAULT_COMMIT_INTERVAL);
>> 1077 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
>> 1078 } else if (intarg > 300) {
>> 1079 btrfs_warn(info, "excessive commit interval %d",
>> 1080 intarg);
>> 1081 }
>> 1082 info->commit_interval = intarg;
>> 1083 break;
>> 1084 case Opt_rescue:
>> 1085 ret = parse_rescue_options(info, args[0].from);
>> 1086 if (ret < 0) {
>> 1087 btrfs_err(info, "unrecognized rescue value %s",
>> 1088 args[0].from);
>> 1089 goto out;
>> 1090 }
602 break; 1091 break;
603 default: <<
604 btrfs_info(NULL, "unre <<
605 param->key) <<
606 return -EINVAL; <<
607 } <<
608 break; <<
609 #ifdef CONFIG_BTRFS_DEBUG 1092 #ifdef CONFIG_BTRFS_DEBUG
610 case Opt_fragment: !! 1093 case Opt_fragment_all:
611 switch (result.uint_32) { !! 1094 btrfs_info(info, "fragmenting all space");
612 case Opt_fragment_parameter_al !! 1095 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
613 btrfs_set_opt(ctx->mou !! 1096 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
614 btrfs_set_opt(ctx->mou <<
615 break; 1097 break;
616 case Opt_fragment_parameter_me !! 1098 case Opt_fragment_metadata:
617 btrfs_set_opt(ctx->mou !! 1099 btrfs_info(info, "fragmenting metadata");
>> 1100 btrfs_set_opt(info->mount_opt,
>> 1101 FRAGMENT_METADATA);
618 break; 1102 break;
619 case Opt_fragment_parameter_da !! 1103 case Opt_fragment_data:
620 btrfs_set_opt(ctx->mou !! 1104 btrfs_info(info, "fragmenting data");
>> 1105 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
621 break; 1106 break;
622 default: <<
623 btrfs_info(NULL, "unre <<
624 param->key) <<
625 return -EINVAL; <<
626 } <<
627 break; <<
628 #endif 1107 #endif
629 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 1108 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
630 case Opt_ref_verify: !! 1109 case Opt_ref_verify:
631 btrfs_set_opt(ctx->mount_opt, !! 1110 btrfs_info(info, "doing ref verification");
632 break; !! 1111 btrfs_set_opt(info->mount_opt, REF_VERIFY);
>> 1112 break;
633 #endif 1113 #endif
634 default: !! 1114 case Opt_err:
635 btrfs_err(NULL, "unrecognized !! 1115 btrfs_err(info, "unrecognized mount option '%s'", p);
636 return -EINVAL; !! 1116 ret = -EINVAL;
>> 1117 goto out;
>> 1118 default:
>> 1119 break;
>> 1120 }
637 } 1121 }
>> 1122 check:
>> 1123 /* We're read-only, don't have to check. */
>> 1124 if (new_flags & SB_RDONLY)
>> 1125 goto out;
638 1126
639 return 0; !! 1127 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
>> 1128 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
>> 1129 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
>> 1130 ret = -EINVAL;
>> 1131 out:
>> 1132 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
>> 1133 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
>> 1134 !btrfs_test_opt(info, CLEAR_CACHE)) {
>> 1135 btrfs_err(info, "cannot disable free space tree");
>> 1136 ret = -EINVAL;
>> 1137
>> 1138 }
>> 1139 if (!ret)
>> 1140 ret = btrfs_check_mountopts_zoned(info);
>> 1141 if (!ret && !remounting) {
>> 1142 if (btrfs_test_opt(info, SPACE_CACHE))
>> 1143 btrfs_info(info, "disk space caching is enabled");
>> 1144 if (btrfs_test_opt(info, FREE_SPACE_TREE))
>> 1145 btrfs_info(info, "using free space tree");
>> 1146 }
>> 1147 return ret;
640 } 1148 }
641 1149
642 /* 1150 /*
643 * Some options only have meaning at mount tim !! 1151 * Parse mount options that are required early in the mount process.
644 * remounts, or be displayed. Clear these at t !! 1152 *
645 * paths. !! 1153 * All other options will be parsed on much later in the mount process and
>> 1154 * only when we need to allocate a new super block.
646 */ 1155 */
647 static void btrfs_clear_oneshot_options(struct !! 1156 static int btrfs_parse_device_options(const char *options, fmode_t flags,
>> 1157 void *holder)
648 { 1158 {
649 btrfs_clear_opt(fs_info->mount_opt, US !! 1159 substring_t args[MAX_OPT_ARGS];
650 btrfs_clear_opt(fs_info->mount_opt, CL !! 1160 char *device_name, *opts, *orig, *p;
651 btrfs_clear_opt(fs_info->mount_opt, NO !! 1161 struct btrfs_device *device = NULL;
652 } !! 1162 int error = 0;
653 1163
654 static bool check_ro_option(const struct btrfs !! 1164 lockdep_assert_held(&uuid_mutex);
655 unsigned long long <<
656 const char *opt_na <<
657 { <<
658 if (mount_opt & opt) { <<
659 btrfs_err(fs_info, "%s must be <<
660 opt_name); <<
661 return true; <<
662 } <<
663 return false; <<
664 } <<
665 1165
666 bool btrfs_check_options(const struct btrfs_fs !! 1166 if (!options)
667 unsigned long long *m !! 1167 return 0;
668 unsigned long flags) <<
669 { <<
670 bool ret = true; <<
671 <<
672 if (!(flags & SB_RDONLY) && <<
673 (check_ro_option(info, *mount_opt, <<
674 check_ro_option(info, *mount_opt, <<
675 check_ro_option(info, *mount_opt, <<
676 check_ro_option(info, *mount_opt, <<
677 check_ro_option(info, *mount_opt, <<
678 ret = false; <<
679 1168
680 if (btrfs_fs_compat_ro(info, FREE_SPAC !! 1169 /*
681 !btrfs_raw_test_opt(*mount_opt, FR !! 1170 * strsep changes the string, duplicate it because btrfs_parse_options
682 !btrfs_raw_test_opt(*mount_opt, CL !! 1171 * gets called later
683 btrfs_err(info, "cannot disabl !! 1172 */
684 ret = false; !! 1173 opts = kstrdup(options, GFP_KERNEL);
685 } !! 1174 if (!opts)
686 if (btrfs_fs_compat_ro(info, BLOCK_GRO !! 1175 return -ENOMEM;
687 !btrfs_raw_test_opt(*mount_opt, F !! 1176 orig = opts;
688 btrfs_err(info, "cannot disabl <<
689 ret = false; <<
690 } <<
691 1177
692 if (btrfs_check_mountopts_zoned(info, !! 1178 while ((p = strsep(&opts, ",")) != NULL) {
693 ret = false; !! 1179 int token;
694 1180
695 if (!test_bit(BTRFS_FS_STATE_REMOUNTIN !! 1181 if (!*p)
696 if (btrfs_raw_test_opt(*mount_ !! 1182 continue;
697 btrfs_info(info, "disk !! 1183
698 btrfs_warn(info, !! 1184 token = match_token(p, tokens, args);
699 "space cache v1 is being deprecated and will b !! 1185 if (token == Opt_device) {
>> 1186 device_name = match_strdup(&args[0]);
>> 1187 if (!device_name) {
>> 1188 error = -ENOMEM;
>> 1189 goto out;
>> 1190 }
>> 1191 device = btrfs_scan_one_device(device_name, flags,
>> 1192 holder);
>> 1193 kfree(device_name);
>> 1194 if (IS_ERR(device)) {
>> 1195 error = PTR_ERR(device);
>> 1196 goto out;
>> 1197 }
700 } 1198 }
701 if (btrfs_raw_test_opt(*mount_ <<
702 btrfs_info(info, "usin <<
703 } 1199 }
704 1200
705 return ret; !! 1201 out:
>> 1202 kfree(orig);
>> 1203 return error;
706 } 1204 }
707 1205
708 /* 1206 /*
709 * This is subtle, we only call this during op !! 1207 * Parse mount options that are related to subvolume id
710 * the mount options with the on-disk settings <<
711 * effect we would do this on mount and remoun <<
712 * only do this on the initial mount. <<
713 * 1208 *
714 * This isn't a change in behavior, because we !! 1209 * The value is later passed to mount_subvol()
715 * file system to set the current mount option <<
716 * options to disable these features and then <<
717 * settings, because mounting without these fe <<
718 * settings, so this being called on re-mount <<
719 */ 1210 */
720 void btrfs_set_free_space_cache_settings(struc !! 1211 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
>> 1212 u64 *subvol_objectid)
721 { 1213 {
722 if (fs_info->sectorsize < PAGE_SIZE) { !! 1214 substring_t args[MAX_OPT_ARGS];
723 btrfs_clear_opt(fs_info->mount !! 1215 char *opts, *orig, *p;
724 if (!btrfs_test_opt(fs_info, F !! 1216 int error = 0;
725 btrfs_info(fs_info, !! 1217 u64 subvolid;
726 "forcing fr !! 1218
727 fs_info->se !! 1219 if (!options)
728 btrfs_set_opt(fs_info- !! 1220 return 0;
729 } <<
730 } <<
731 1221
732 /* 1222 /*
733 * At this point our mount options are !! 1223 * strsep changes the string, duplicate it because
734 * these settings if we don't have any !! 1224 * btrfs_parse_device_options gets called later
735 */ 1225 */
736 if (btrfs_test_opt(fs_info, FREE_SPACE !! 1226 opts = kstrdup(options, GFP_KERNEL);
737 return; !! 1227 if (!opts)
738 !! 1228 return -ENOMEM;
739 if (btrfs_is_zoned(fs_info) && !! 1229 orig = opts;
740 btrfs_free_space_cache_v1_active(f <<
741 btrfs_info(fs_info, "zoned: cl <<
742 btrfs_set_super_cache_generati <<
743 return; <<
744 } <<
745 1230
746 if (btrfs_test_opt(fs_info, SPACE_CACH !! 1231 while ((p = strsep(&opts, ",")) != NULL) {
747 return; !! 1232 int token;
>> 1233 if (!*p)
>> 1234 continue;
748 1235
749 if (btrfs_test_opt(fs_info, NOSPACECAC !! 1236 token = match_token(p, tokens, args);
750 return; !! 1237 switch (token) {
>> 1238 case Opt_subvol:
>> 1239 kfree(*subvol_name);
>> 1240 *subvol_name = match_strdup(&args[0]);
>> 1241 if (!*subvol_name) {
>> 1242 error = -ENOMEM;
>> 1243 goto out;
>> 1244 }
>> 1245 break;
>> 1246 case Opt_subvolid:
>> 1247 error = match_u64(&args[0], &subvolid);
>> 1248 if (error)
>> 1249 goto out;
751 1250
752 /* !! 1251 /* we want the original fs_tree */
753 * At this point we don't have explici !! 1252 if (subvolid == 0)
754 * them ourselves based on the state o !! 1253 subvolid = BTRFS_FS_TREE_OBJECTID;
755 */ <<
756 if (btrfs_fs_compat_ro(fs_info, FREE_S <<
757 btrfs_set_opt(fs_info->mount_o <<
758 else if (btrfs_free_space_cache_v1_act <<
759 btrfs_set_opt(fs_info->mount_o <<
760 } <<
761 1254
762 static void set_device_specific_options(struct !! 1255 *subvol_objectid = subvolid;
763 { !! 1256 break;
764 if (!btrfs_test_opt(fs_info, NOSSD) && !! 1257 default:
765 !fs_info->fs_devices->rotating) !! 1258 break;
766 btrfs_set_opt(fs_info->mount_o !! 1259 }
>> 1260 }
767 1261
768 /* !! 1262 out:
769 * For devices supporting discard turn !! 1263 kfree(orig);
770 * unless it's already set or disabled !! 1264 return error;
771 * nodiscard for the same mount. <<
772 * <<
773 * The zoned mode piggy backs on the d <<
774 * resetting a zone. There is no reaso <<
775 * fast enough. So, do not enable asyn <<
776 */ <<
777 if (!(btrfs_test_opt(fs_info, DISCARD_ <<
778 btrfs_test_opt(fs_info, DISCARD_ <<
779 btrfs_test_opt(fs_info, NODISCAR <<
780 fs_info->fs_devices->discardable & <<
781 !btrfs_is_zoned(fs_info)) <<
782 btrfs_set_opt(fs_info->mount_o <<
783 } 1265 }
784 1266
785 char *btrfs_get_subvol_name_from_objectid(stru 1267 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
786 u64 1268 u64 subvol_objectid)
787 { 1269 {
788 struct btrfs_root *root = fs_info->tre 1270 struct btrfs_root *root = fs_info->tree_root;
789 struct btrfs_root *fs_root = NULL; 1271 struct btrfs_root *fs_root = NULL;
790 struct btrfs_root_ref *root_ref; 1272 struct btrfs_root_ref *root_ref;
791 struct btrfs_inode_ref *inode_ref; 1273 struct btrfs_inode_ref *inode_ref;
792 struct btrfs_key key; 1274 struct btrfs_key key;
793 struct btrfs_path *path = NULL; 1275 struct btrfs_path *path = NULL;
794 char *name = NULL, *ptr; 1276 char *name = NULL, *ptr;
795 u64 dirid; 1277 u64 dirid;
796 int len; 1278 int len;
797 int ret; 1279 int ret;
798 1280
799 path = btrfs_alloc_path(); 1281 path = btrfs_alloc_path();
800 if (!path) { 1282 if (!path) {
801 ret = -ENOMEM; 1283 ret = -ENOMEM;
802 goto err; 1284 goto err;
803 } 1285 }
804 1286
805 name = kmalloc(PATH_MAX, GFP_KERNEL); 1287 name = kmalloc(PATH_MAX, GFP_KERNEL);
806 if (!name) { 1288 if (!name) {
807 ret = -ENOMEM; 1289 ret = -ENOMEM;
808 goto err; 1290 goto err;
809 } 1291 }
810 ptr = name + PATH_MAX - 1; 1292 ptr = name + PATH_MAX - 1;
811 ptr[0] = '\0'; 1293 ptr[0] = '\0';
812 1294
813 /* 1295 /*
814 * Walk up the subvolume trees in the 1296 * Walk up the subvolume trees in the tree of tree roots by root
815 * backrefs until we hit the top-level 1297 * backrefs until we hit the top-level subvolume.
816 */ 1298 */
817 while (subvol_objectid != BTRFS_FS_TRE 1299 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
818 key.objectid = subvol_objectid 1300 key.objectid = subvol_objectid;
819 key.type = BTRFS_ROOT_BACKREF_ 1301 key.type = BTRFS_ROOT_BACKREF_KEY;
820 key.offset = (u64)-1; 1302 key.offset = (u64)-1;
821 1303
822 ret = btrfs_search_backwards(r 1304 ret = btrfs_search_backwards(root, &key, path);
823 if (ret < 0) { 1305 if (ret < 0) {
824 goto err; 1306 goto err;
825 } else if (ret > 0) { 1307 } else if (ret > 0) {
826 ret = -ENOENT; 1308 ret = -ENOENT;
827 goto err; 1309 goto err;
828 } 1310 }
829 1311
830 subvol_objectid = key.offset; 1312 subvol_objectid = key.offset;
831 1313
832 root_ref = btrfs_item_ptr(path 1314 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
833 stru 1315 struct btrfs_root_ref);
834 len = btrfs_root_ref_name_len( 1316 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
835 ptr -= len + 1; 1317 ptr -= len + 1;
836 if (ptr < name) { 1318 if (ptr < name) {
837 ret = -ENAMETOOLONG; 1319 ret = -ENAMETOOLONG;
838 goto err; 1320 goto err;
839 } 1321 }
840 read_extent_buffer(path->nodes 1322 read_extent_buffer(path->nodes[0], ptr + 1,
841 (unsigned l 1323 (unsigned long)(root_ref + 1), len);
842 ptr[0] = '/'; 1324 ptr[0] = '/';
843 dirid = btrfs_root_ref_dirid(p 1325 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
844 btrfs_release_path(path); 1326 btrfs_release_path(path);
845 1327
846 fs_root = btrfs_get_fs_root(fs 1328 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
847 if (IS_ERR(fs_root)) { 1329 if (IS_ERR(fs_root)) {
848 ret = PTR_ERR(fs_root) 1330 ret = PTR_ERR(fs_root);
849 fs_root = NULL; 1331 fs_root = NULL;
850 goto err; 1332 goto err;
851 } 1333 }
852 1334
853 /* 1335 /*
854 * Walk up the filesystem tree 1336 * Walk up the filesystem tree by inode refs until we hit the
855 * root directory. 1337 * root directory.
856 */ 1338 */
857 while (dirid != BTRFS_FIRST_FR 1339 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
858 key.objectid = dirid; 1340 key.objectid = dirid;
859 key.type = BTRFS_INODE 1341 key.type = BTRFS_INODE_REF_KEY;
860 key.offset = (u64)-1; 1342 key.offset = (u64)-1;
861 1343
862 ret = btrfs_search_bac 1344 ret = btrfs_search_backwards(fs_root, &key, path);
863 if (ret < 0) { 1345 if (ret < 0) {
864 goto err; 1346 goto err;
865 } else if (ret > 0) { 1347 } else if (ret > 0) {
866 ret = -ENOENT; 1348 ret = -ENOENT;
867 goto err; 1349 goto err;
868 } 1350 }
869 1351
870 dirid = key.offset; 1352 dirid = key.offset;
871 1353
872 inode_ref = btrfs_item 1354 inode_ref = btrfs_item_ptr(path->nodes[0],
873 1355 path->slots[0],
874 1356 struct btrfs_inode_ref);
875 len = btrfs_inode_ref_ 1357 len = btrfs_inode_ref_name_len(path->nodes[0],
876 1358 inode_ref);
877 ptr -= len + 1; 1359 ptr -= len + 1;
878 if (ptr < name) { 1360 if (ptr < name) {
879 ret = -ENAMETO 1361 ret = -ENAMETOOLONG;
880 goto err; 1362 goto err;
881 } 1363 }
882 read_extent_buffer(pat 1364 read_extent_buffer(path->nodes[0], ptr + 1,
883 (un 1365 (unsigned long)(inode_ref + 1), len);
884 ptr[0] = '/'; 1366 ptr[0] = '/';
885 btrfs_release_path(pat 1367 btrfs_release_path(path);
886 } 1368 }
887 btrfs_put_root(fs_root); 1369 btrfs_put_root(fs_root);
888 fs_root = NULL; 1370 fs_root = NULL;
889 } 1371 }
890 1372
891 btrfs_free_path(path); 1373 btrfs_free_path(path);
892 if (ptr == name + PATH_MAX - 1) { 1374 if (ptr == name + PATH_MAX - 1) {
893 name[0] = '/'; 1375 name[0] = '/';
894 name[1] = '\0'; 1376 name[1] = '\0';
895 } else { 1377 } else {
896 memmove(name, ptr, name + PATH 1378 memmove(name, ptr, name + PATH_MAX - ptr);
897 } 1379 }
898 return name; 1380 return name;
899 1381
900 err: 1382 err:
901 btrfs_put_root(fs_root); 1383 btrfs_put_root(fs_root);
902 btrfs_free_path(path); 1384 btrfs_free_path(path);
903 kfree(name); 1385 kfree(name);
904 return ERR_PTR(ret); 1386 return ERR_PTR(ret);
905 } 1387 }
906 1388
907 static int get_default_subvol_objectid(struct 1389 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
908 { 1390 {
909 struct btrfs_root *root = fs_info->tre 1391 struct btrfs_root *root = fs_info->tree_root;
910 struct btrfs_dir_item *di; 1392 struct btrfs_dir_item *di;
911 struct btrfs_path *path; 1393 struct btrfs_path *path;
912 struct btrfs_key location; 1394 struct btrfs_key location;
913 struct fscrypt_str name = FSTR_INIT("d <<
914 u64 dir_id; 1395 u64 dir_id;
915 1396
916 path = btrfs_alloc_path(); 1397 path = btrfs_alloc_path();
917 if (!path) 1398 if (!path)
918 return -ENOMEM; 1399 return -ENOMEM;
919 1400
920 /* 1401 /*
921 * Find the "default" dir item which p 1402 * Find the "default" dir item which points to the root item that we
922 * will mount by default if we haven't 1403 * will mount by default if we haven't been given a specific subvolume
923 * to mount. 1404 * to mount.
924 */ 1405 */
925 dir_id = btrfs_super_root_dir(fs_info- 1406 dir_id = btrfs_super_root_dir(fs_info->super_copy);
926 di = btrfs_lookup_dir_item(NULL, root, !! 1407 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
927 if (IS_ERR(di)) { 1408 if (IS_ERR(di)) {
928 btrfs_free_path(path); 1409 btrfs_free_path(path);
929 return PTR_ERR(di); 1410 return PTR_ERR(di);
930 } 1411 }
931 if (!di) { 1412 if (!di) {
932 /* 1413 /*
933 * Ok the default dir item isn 1414 * Ok the default dir item isn't there. This is weird since
934 * it's always been there, but 1415 * it's always been there, but don't freak out, just try and
935 * mount the top-level subvolu 1416 * mount the top-level subvolume.
936 */ 1417 */
937 btrfs_free_path(path); 1418 btrfs_free_path(path);
938 *objectid = BTRFS_FS_TREE_OBJE 1419 *objectid = BTRFS_FS_TREE_OBJECTID;
939 return 0; 1420 return 0;
940 } 1421 }
941 1422
942 btrfs_dir_item_key_to_cpu(path->nodes[ 1423 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
943 btrfs_free_path(path); 1424 btrfs_free_path(path);
944 *objectid = location.objectid; 1425 *objectid = location.objectid;
945 return 0; 1426 return 0;
946 } 1427 }
947 1428
948 static int btrfs_fill_super(struct super_block 1429 static int btrfs_fill_super(struct super_block *sb,
949 struct btrfs_fs_de 1430 struct btrfs_fs_devices *fs_devices,
950 void *data) 1431 void *data)
951 { 1432 {
952 struct inode *inode; 1433 struct inode *inode;
953 struct btrfs_fs_info *fs_info = btrfs_ 1434 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
954 int err; 1435 int err;
955 1436
956 sb->s_maxbytes = MAX_LFS_FILESIZE; 1437 sb->s_maxbytes = MAX_LFS_FILESIZE;
957 sb->s_magic = BTRFS_SUPER_MAGIC; 1438 sb->s_magic = BTRFS_SUPER_MAGIC;
958 sb->s_op = &btrfs_super_ops; 1439 sb->s_op = &btrfs_super_ops;
959 sb->s_d_op = &btrfs_dentry_operations; 1440 sb->s_d_op = &btrfs_dentry_operations;
960 sb->s_export_op = &btrfs_export_ops; 1441 sb->s_export_op = &btrfs_export_ops;
961 #ifdef CONFIG_FS_VERITY 1442 #ifdef CONFIG_FS_VERITY
962 sb->s_vop = &btrfs_verityops; 1443 sb->s_vop = &btrfs_verityops;
963 #endif 1444 #endif
964 sb->s_xattr = btrfs_xattr_handlers; 1445 sb->s_xattr = btrfs_xattr_handlers;
965 sb->s_time_gran = 1; 1446 sb->s_time_gran = 1;
>> 1447 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
>> 1448 sb->s_flags |= SB_POSIXACL;
>> 1449 #endif
>> 1450 sb->s_flags |= SB_I_VERSION;
966 sb->s_iflags |= SB_I_CGROUPWB; 1451 sb->s_iflags |= SB_I_CGROUPWB;
967 1452
968 err = super_setup_bdi(sb); 1453 err = super_setup_bdi(sb);
969 if (err) { 1454 if (err) {
970 btrfs_err(fs_info, "super_setu 1455 btrfs_err(fs_info, "super_setup_bdi failed");
971 return err; 1456 return err;
972 } 1457 }
973 1458
974 err = open_ctree(sb, fs_devices, (char 1459 err = open_ctree(sb, fs_devices, (char *)data);
975 if (err) { 1460 if (err) {
976 btrfs_err(fs_info, "open_ctree 1461 btrfs_err(fs_info, "open_ctree failed");
977 return err; 1462 return err;
978 } 1463 }
979 1464
980 inode = btrfs_iget(BTRFS_FIRST_FREE_OB !! 1465 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
981 if (IS_ERR(inode)) { 1466 if (IS_ERR(inode)) {
982 err = PTR_ERR(inode); 1467 err = PTR_ERR(inode);
983 btrfs_handle_fs_error(fs_info, <<
984 goto fail_close; 1468 goto fail_close;
985 } 1469 }
986 1470
987 sb->s_root = d_make_root(inode); 1471 sb->s_root = d_make_root(inode);
988 if (!sb->s_root) { 1472 if (!sb->s_root) {
989 err = -ENOMEM; 1473 err = -ENOMEM;
990 goto fail_close; 1474 goto fail_close;
991 } 1475 }
992 1476
993 sb->s_flags |= SB_ACTIVE; 1477 sb->s_flags |= SB_ACTIVE;
994 return 0; 1478 return 0;
995 1479
996 fail_close: 1480 fail_close:
997 close_ctree(fs_info); 1481 close_ctree(fs_info);
998 return err; 1482 return err;
999 } 1483 }
1000 1484
1001 int btrfs_sync_fs(struct super_block *sb, int 1485 int btrfs_sync_fs(struct super_block *sb, int wait)
1002 { 1486 {
1003 struct btrfs_trans_handle *trans; 1487 struct btrfs_trans_handle *trans;
1004 struct btrfs_fs_info *fs_info = btrfs 1488 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1005 struct btrfs_root *root = fs_info->tr 1489 struct btrfs_root *root = fs_info->tree_root;
1006 1490
1007 trace_btrfs_sync_fs(fs_info, wait); 1491 trace_btrfs_sync_fs(fs_info, wait);
1008 1492
1009 if (!wait) { 1493 if (!wait) {
1010 filemap_flush(fs_info->btree_ 1494 filemap_flush(fs_info->btree_inode->i_mapping);
1011 return 0; 1495 return 0;
1012 } 1496 }
1013 1497
1014 btrfs_wait_ordered_roots(fs_info, U64 !! 1498 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1015 1499
1016 trans = btrfs_attach_transaction_barr 1500 trans = btrfs_attach_transaction_barrier(root);
1017 if (IS_ERR(trans)) { 1501 if (IS_ERR(trans)) {
1018 /* no transaction, don't both 1502 /* no transaction, don't bother */
1019 if (PTR_ERR(trans) == -ENOENT 1503 if (PTR_ERR(trans) == -ENOENT) {
1020 /* 1504 /*
1021 * Exit unless we hav 1505 * Exit unless we have some pending changes
1022 * that need to go th 1506 * that need to go through commit
1023 */ 1507 */
1024 if (!test_bit(BTRFS_F !! 1508 if (fs_info->pending_changes == 0)
1025 &fs_inf <<
1026 return 0; 1509 return 0;
1027 /* 1510 /*
1028 * A non-blocking tes 1511 * A non-blocking test if the fs is frozen. We must not
1029 * start a new transa 1512 * start a new transaction here otherwise a deadlock
1030 * happens. The pendi 1513 * happens. The pending operations are delayed to the
1031 * next commit after 1514 * next commit after thawing.
1032 */ 1515 */
1033 if (sb_start_write_tr 1516 if (sb_start_write_trylock(sb))
1034 sb_end_write( 1517 sb_end_write(sb);
1035 else 1518 else
1036 return 0; 1519 return 0;
1037 trans = btrfs_start_t 1520 trans = btrfs_start_transaction(root, 0);
1038 } 1521 }
1039 if (IS_ERR(trans)) 1522 if (IS_ERR(trans))
1040 return PTR_ERR(trans) 1523 return PTR_ERR(trans);
1041 } 1524 }
1042 return btrfs_commit_transaction(trans 1525 return btrfs_commit_transaction(trans);
1043 } 1526 }
1044 1527
1045 static void print_rescue_option(struct seq_fi 1528 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1046 { 1529 {
1047 seq_printf(seq, "%s%s", (*printed) ? 1530 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1048 *printed = true; 1531 *printed = true;
1049 } 1532 }
1050 1533
1051 static int btrfs_show_options(struct seq_file 1534 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1052 { 1535 {
1053 struct btrfs_fs_info *info = btrfs_sb 1536 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1054 const char *compress_type; 1537 const char *compress_type;
1055 const char *subvol_name; 1538 const char *subvol_name;
1056 bool printed = false; 1539 bool printed = false;
1057 1540
1058 if (btrfs_test_opt(info, DEGRADED)) 1541 if (btrfs_test_opt(info, DEGRADED))
1059 seq_puts(seq, ",degraded"); 1542 seq_puts(seq, ",degraded");
1060 if (btrfs_test_opt(info, NODATASUM)) 1543 if (btrfs_test_opt(info, NODATASUM))
1061 seq_puts(seq, ",nodatasum"); 1544 seq_puts(seq, ",nodatasum");
1062 if (btrfs_test_opt(info, NODATACOW)) 1545 if (btrfs_test_opt(info, NODATACOW))
1063 seq_puts(seq, ",nodatacow"); 1546 seq_puts(seq, ",nodatacow");
1064 if (btrfs_test_opt(info, NOBARRIER)) 1547 if (btrfs_test_opt(info, NOBARRIER))
1065 seq_puts(seq, ",nobarrier"); 1548 seq_puts(seq, ",nobarrier");
1066 if (info->max_inline != BTRFS_DEFAULT 1549 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1067 seq_printf(seq, ",max_inline= 1550 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1068 if (info->thread_pool_size != min_t( 1551 if (info->thread_pool_size != min_t(unsigned long,
1069 1552 num_online_cpus() + 2, 8))
1070 seq_printf(seq, ",thread_pool 1553 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1071 if (btrfs_test_opt(info, COMPRESS)) { 1554 if (btrfs_test_opt(info, COMPRESS)) {
1072 compress_type = btrfs_compres 1555 compress_type = btrfs_compress_type2str(info->compress_type);
1073 if (btrfs_test_opt(info, FORC 1556 if (btrfs_test_opt(info, FORCE_COMPRESS))
1074 seq_printf(seq, ",com 1557 seq_printf(seq, ",compress-force=%s", compress_type);
1075 else 1558 else
1076 seq_printf(seq, ",com 1559 seq_printf(seq, ",compress=%s", compress_type);
1077 if (info->compress_level) 1560 if (info->compress_level)
1078 seq_printf(seq, ":%d" 1561 seq_printf(seq, ":%d", info->compress_level);
1079 } 1562 }
1080 if (btrfs_test_opt(info, NOSSD)) 1563 if (btrfs_test_opt(info, NOSSD))
1081 seq_puts(seq, ",nossd"); 1564 seq_puts(seq, ",nossd");
1082 if (btrfs_test_opt(info, SSD_SPREAD)) 1565 if (btrfs_test_opt(info, SSD_SPREAD))
1083 seq_puts(seq, ",ssd_spread"); 1566 seq_puts(seq, ",ssd_spread");
1084 else if (btrfs_test_opt(info, SSD)) 1567 else if (btrfs_test_opt(info, SSD))
1085 seq_puts(seq, ",ssd"); 1568 seq_puts(seq, ",ssd");
1086 if (btrfs_test_opt(info, NOTREELOG)) 1569 if (btrfs_test_opt(info, NOTREELOG))
1087 seq_puts(seq, ",notreelog"); 1570 seq_puts(seq, ",notreelog");
1088 if (btrfs_test_opt(info, NOLOGREPLAY) 1571 if (btrfs_test_opt(info, NOLOGREPLAY))
1089 print_rescue_option(seq, "nol 1572 print_rescue_option(seq, "nologreplay", &printed);
1090 if (btrfs_test_opt(info, USEBACKUPROO 1573 if (btrfs_test_opt(info, USEBACKUPROOT))
1091 print_rescue_option(seq, "use 1574 print_rescue_option(seq, "usebackuproot", &printed);
1092 if (btrfs_test_opt(info, IGNOREBADROO 1575 if (btrfs_test_opt(info, IGNOREBADROOTS))
1093 print_rescue_option(seq, "ign 1576 print_rescue_option(seq, "ignorebadroots", &printed);
1094 if (btrfs_test_opt(info, IGNOREDATACS 1577 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1095 print_rescue_option(seq, "ign 1578 print_rescue_option(seq, "ignoredatacsums", &printed);
1096 if (btrfs_test_opt(info, IGNOREMETACS <<
1097 print_rescue_option(seq, "ign <<
1098 if (btrfs_test_opt(info, IGNORESUPERF <<
1099 print_rescue_option(seq, "ign <<
1100 if (btrfs_test_opt(info, FLUSHONCOMMI 1579 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1101 seq_puts(seq, ",flushoncommit 1580 seq_puts(seq, ",flushoncommit");
1102 if (btrfs_test_opt(info, DISCARD_SYNC 1581 if (btrfs_test_opt(info, DISCARD_SYNC))
1103 seq_puts(seq, ",discard"); 1582 seq_puts(seq, ",discard");
1104 if (btrfs_test_opt(info, DISCARD_ASYN 1583 if (btrfs_test_opt(info, DISCARD_ASYNC))
1105 seq_puts(seq, ",discard=async 1584 seq_puts(seq, ",discard=async");
1106 if (!(info->sb->s_flags & SB_POSIXACL 1585 if (!(info->sb->s_flags & SB_POSIXACL))
1107 seq_puts(seq, ",noacl"); 1586 seq_puts(seq, ",noacl");
1108 if (btrfs_free_space_cache_v1_active( 1587 if (btrfs_free_space_cache_v1_active(info))
1109 seq_puts(seq, ",space_cache") 1588 seq_puts(seq, ",space_cache");
1110 else if (btrfs_fs_compat_ro(info, FRE 1589 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1111 seq_puts(seq, ",space_cache=v 1590 seq_puts(seq, ",space_cache=v2");
1112 else 1591 else
1113 seq_puts(seq, ",nospace_cache 1592 seq_puts(seq, ",nospace_cache");
1114 if (btrfs_test_opt(info, RESCAN_UUID_ 1593 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1115 seq_puts(seq, ",rescan_uuid_t 1594 seq_puts(seq, ",rescan_uuid_tree");
1116 if (btrfs_test_opt(info, CLEAR_CACHE) 1595 if (btrfs_test_opt(info, CLEAR_CACHE))
1117 seq_puts(seq, ",clear_cache") 1596 seq_puts(seq, ",clear_cache");
1118 if (btrfs_test_opt(info, USER_SUBVOL_ 1597 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1119 seq_puts(seq, ",user_subvol_r 1598 seq_puts(seq, ",user_subvol_rm_allowed");
1120 if (btrfs_test_opt(info, ENOSPC_DEBUG 1599 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1121 seq_puts(seq, ",enospc_debug" 1600 seq_puts(seq, ",enospc_debug");
1122 if (btrfs_test_opt(info, AUTO_DEFRAG) 1601 if (btrfs_test_opt(info, AUTO_DEFRAG))
1123 seq_puts(seq, ",autodefrag"); 1602 seq_puts(seq, ",autodefrag");
1124 if (btrfs_test_opt(info, SKIP_BALANCE 1603 if (btrfs_test_opt(info, SKIP_BALANCE))
1125 seq_puts(seq, ",skip_balance" 1604 seq_puts(seq, ",skip_balance");
>> 1605 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 1606 if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
>> 1607 seq_puts(seq, ",check_int_data");
>> 1608 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
>> 1609 seq_puts(seq, ",check_int");
>> 1610 if (info->check_integrity_print_mask)
>> 1611 seq_printf(seq, ",check_int_print_mask=%d",
>> 1612 info->check_integrity_print_mask);
>> 1613 #endif
1126 if (info->metadata_ratio) 1614 if (info->metadata_ratio)
1127 seq_printf(seq, ",metadata_ra 1615 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1128 if (btrfs_test_opt(info, PANIC_ON_FAT 1616 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1129 seq_puts(seq, ",fatal_errors= 1617 seq_puts(seq, ",fatal_errors=panic");
1130 if (info->commit_interval != BTRFS_DE 1618 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1131 seq_printf(seq, ",commit=%u", 1619 seq_printf(seq, ",commit=%u", info->commit_interval);
1132 #ifdef CONFIG_BTRFS_DEBUG 1620 #ifdef CONFIG_BTRFS_DEBUG
1133 if (btrfs_test_opt(info, FRAGMENT_DAT 1621 if (btrfs_test_opt(info, FRAGMENT_DATA))
1134 seq_puts(seq, ",fragment=data 1622 seq_puts(seq, ",fragment=data");
1135 if (btrfs_test_opt(info, FRAGMENT_MET 1623 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1136 seq_puts(seq, ",fragment=meta 1624 seq_puts(seq, ",fragment=metadata");
1137 #endif 1625 #endif
1138 if (btrfs_test_opt(info, REF_VERIFY)) 1626 if (btrfs_test_opt(info, REF_VERIFY))
1139 seq_puts(seq, ",ref_verify"); 1627 seq_puts(seq, ",ref_verify");
1140 seq_printf(seq, ",subvolid=%llu", btr !! 1628 seq_printf(seq, ",subvolid=%llu",
>> 1629 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1141 subvol_name = btrfs_get_subvol_name_f 1630 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1142 btrfs_root_id(BTRFS_I !! 1631 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1143 if (!IS_ERR(subvol_name)) { 1632 if (!IS_ERR(subvol_name)) {
1144 seq_puts(seq, ",subvol="); 1633 seq_puts(seq, ",subvol=");
1145 seq_escape(seq, subvol_name, 1634 seq_escape(seq, subvol_name, " \t\n\\");
1146 kfree(subvol_name); 1635 kfree(subvol_name);
1147 } 1636 }
1148 return 0; 1637 return 0;
1149 } 1638 }
1150 1639
>> 1640 static int btrfs_test_super(struct super_block *s, void *data)
>> 1641 {
>> 1642 struct btrfs_fs_info *p = data;
>> 1643 struct btrfs_fs_info *fs_info = btrfs_sb(s);
>> 1644
>> 1645 return fs_info->fs_devices == p->fs_devices;
>> 1646 }
>> 1647
>> 1648 static int btrfs_set_super(struct super_block *s, void *data)
>> 1649 {
>> 1650 int err = set_anon_super(s, data);
>> 1651 if (!err)
>> 1652 s->s_fs_info = data;
>> 1653 return err;
>> 1654 }
>> 1655
1151 /* 1656 /*
1152 * subvolumes are identified by ino 256 1657 * subvolumes are identified by ino 256
1153 */ 1658 */
1154 static inline int is_subvolume_inode(struct i 1659 static inline int is_subvolume_inode(struct inode *inode)
1155 { 1660 {
1156 if (inode && inode->i_ino == BTRFS_FI 1661 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1157 return 1; 1662 return 1;
1158 return 0; 1663 return 0;
1159 } 1664 }
1160 1665
1161 static struct dentry *mount_subvol(const char 1666 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1162 struct vfs 1667 struct vfsmount *mnt)
1163 { 1668 {
1164 struct dentry *root; 1669 struct dentry *root;
1165 int ret; 1670 int ret;
1166 1671
1167 if (!subvol_name) { 1672 if (!subvol_name) {
1168 if (!subvol_objectid) { 1673 if (!subvol_objectid) {
1169 ret = get_default_sub 1674 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1170 1675 &subvol_objectid);
1171 if (ret) { 1676 if (ret) {
1172 root = ERR_PT 1677 root = ERR_PTR(ret);
1173 goto out; 1678 goto out;
1174 } 1679 }
1175 } 1680 }
1176 subvol_name = btrfs_get_subvo 1681 subvol_name = btrfs_get_subvol_name_from_objectid(
1177 btrfs 1682 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1178 if (IS_ERR(subvol_name)) { 1683 if (IS_ERR(subvol_name)) {
1179 root = ERR_CAST(subvo 1684 root = ERR_CAST(subvol_name);
1180 subvol_name = NULL; 1685 subvol_name = NULL;
1181 goto out; 1686 goto out;
1182 } 1687 }
1183 1688
1184 } 1689 }
1185 1690
1186 root = mount_subtree(mnt, subvol_name 1691 root = mount_subtree(mnt, subvol_name);
1187 /* mount_subtree() drops our referenc 1692 /* mount_subtree() drops our reference on the vfsmount. */
1188 mnt = NULL; 1693 mnt = NULL;
1189 1694
1190 if (!IS_ERR(root)) { 1695 if (!IS_ERR(root)) {
1191 struct super_block *s = root- 1696 struct super_block *s = root->d_sb;
1192 struct btrfs_fs_info *fs_info 1697 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1193 struct inode *root_inode = d_ 1698 struct inode *root_inode = d_inode(root);
1194 u64 root_objectid = btrfs_roo !! 1699 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1195 1700
1196 ret = 0; 1701 ret = 0;
1197 if (!is_subvolume_inode(root_ 1702 if (!is_subvolume_inode(root_inode)) {
1198 btrfs_err(fs_info, "' 1703 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1199 subvol_name); 1704 subvol_name);
1200 ret = -EINVAL; 1705 ret = -EINVAL;
1201 } 1706 }
1202 if (subvol_objectid && root_o 1707 if (subvol_objectid && root_objectid != subvol_objectid) {
1203 /* 1708 /*
1204 * This will also cat 1709 * This will also catch a race condition where a
1205 * subvolume which wa 1710 * subvolume which was passed by ID is renamed and
1206 * another subvolume 1711 * another subvolume is renamed over the old location.
1207 */ 1712 */
1208 btrfs_err(fs_info, 1713 btrfs_err(fs_info,
1209 "subvol '%s 1714 "subvol '%s' does not match subvolid %llu",
1210 subvol_name 1715 subvol_name, subvol_objectid);
1211 ret = -EINVAL; 1716 ret = -EINVAL;
1212 } 1717 }
1213 if (ret) { 1718 if (ret) {
1214 dput(root); 1719 dput(root);
1215 root = ERR_PTR(ret); 1720 root = ERR_PTR(ret);
1216 deactivate_locked_sup 1721 deactivate_locked_super(s);
1217 } 1722 }
1218 } 1723 }
1219 1724
1220 out: 1725 out:
1221 mntput(mnt); 1726 mntput(mnt);
1222 kfree(subvol_name); 1727 kfree(subvol_name);
1223 return root; 1728 return root;
1224 } 1729 }
1225 1730
>> 1731 /*
>> 1732 * Find a superblock for the given device / mount point.
>> 1733 *
>> 1734 * Note: This is based on mount_bdev from fs/super.c with a few additions
>> 1735 * for multiple device setup. Make sure to keep it in sync.
>> 1736 */
>> 1737 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
>> 1738 int flags, const char *device_name, void *data)
>> 1739 {
>> 1740 struct block_device *bdev = NULL;
>> 1741 struct super_block *s;
>> 1742 struct btrfs_device *device = NULL;
>> 1743 struct btrfs_fs_devices *fs_devices = NULL;
>> 1744 struct btrfs_fs_info *fs_info = NULL;
>> 1745 void *new_sec_opts = NULL;
>> 1746 fmode_t mode = FMODE_READ;
>> 1747 int error = 0;
>> 1748
>> 1749 if (!(flags & SB_RDONLY))
>> 1750 mode |= FMODE_WRITE;
>> 1751
>> 1752 if (data) {
>> 1753 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
>> 1754 if (error)
>> 1755 return ERR_PTR(error);
>> 1756 }
>> 1757
>> 1758 /*
>> 1759 * Setup a dummy root and fs_info for test/set super. This is because
>> 1760 * we don't actually fill this stuff out until open_ctree, but we need
>> 1761 * then open_ctree will properly initialize the file system specific
>> 1762 * settings later. btrfs_init_fs_info initializes the static elements
>> 1763 * of the fs_info (locks and such) to make cleanup easier if we find a
>> 1764 * superblock with our given fs_devices later on at sget() time.
>> 1765 */
>> 1766 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
>> 1767 if (!fs_info) {
>> 1768 error = -ENOMEM;
>> 1769 goto error_sec_opts;
>> 1770 }
>> 1771 btrfs_init_fs_info(fs_info);
>> 1772
>> 1773 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
>> 1774 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
>> 1775 if (!fs_info->super_copy || !fs_info->super_for_commit) {
>> 1776 error = -ENOMEM;
>> 1777 goto error_fs_info;
>> 1778 }
>> 1779
>> 1780 mutex_lock(&uuid_mutex);
>> 1781 error = btrfs_parse_device_options(data, mode, fs_type);
>> 1782 if (error) {
>> 1783 mutex_unlock(&uuid_mutex);
>> 1784 goto error_fs_info;
>> 1785 }
>> 1786
>> 1787 device = btrfs_scan_one_device(device_name, mode, fs_type);
>> 1788 if (IS_ERR(device)) {
>> 1789 mutex_unlock(&uuid_mutex);
>> 1790 error = PTR_ERR(device);
>> 1791 goto error_fs_info;
>> 1792 }
>> 1793
>> 1794 fs_devices = device->fs_devices;
>> 1795 fs_info->fs_devices = fs_devices;
>> 1796
>> 1797 error = btrfs_open_devices(fs_devices, mode, fs_type);
>> 1798 mutex_unlock(&uuid_mutex);
>> 1799 if (error)
>> 1800 goto error_fs_info;
>> 1801
>> 1802 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
>> 1803 error = -EACCES;
>> 1804 goto error_close_devices;
>> 1805 }
>> 1806
>> 1807 bdev = fs_devices->latest_dev->bdev;
>> 1808 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
>> 1809 fs_info);
>> 1810 if (IS_ERR(s)) {
>> 1811 error = PTR_ERR(s);
>> 1812 goto error_close_devices;
>> 1813 }
>> 1814
>> 1815 if (s->s_root) {
>> 1816 btrfs_close_devices(fs_devices);
>> 1817 btrfs_free_fs_info(fs_info);
>> 1818 if ((flags ^ s->s_flags) & SB_RDONLY)
>> 1819 error = -EBUSY;
>> 1820 } else {
>> 1821 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
>> 1822 shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
>> 1823 s->s_id);
>> 1824 btrfs_sb(s)->bdev_holder = fs_type;
>> 1825 if (!strstr(crc32c_impl(), "generic"))
>> 1826 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
>> 1827 error = btrfs_fill_super(s, fs_devices, data);
>> 1828 }
>> 1829 if (!error)
>> 1830 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
>> 1831 security_free_mnt_opts(&new_sec_opts);
>> 1832 if (error) {
>> 1833 deactivate_locked_super(s);
>> 1834 return ERR_PTR(error);
>> 1835 }
>> 1836
>> 1837 return dget(s->s_root);
>> 1838
>> 1839 error_close_devices:
>> 1840 btrfs_close_devices(fs_devices);
>> 1841 error_fs_info:
>> 1842 btrfs_free_fs_info(fs_info);
>> 1843 error_sec_opts:
>> 1844 security_free_mnt_opts(&new_sec_opts);
>> 1845 return ERR_PTR(error);
>> 1846 }
>> 1847
>> 1848 /*
>> 1849 * Mount function which is called by VFS layer.
>> 1850 *
>> 1851 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
>> 1852 * which needs vfsmount* of device's root (/). This means device's root has to
>> 1853 * be mounted internally in any case.
>> 1854 *
>> 1855 * Operation flow:
>> 1856 * 1. Parse subvol id related options for later use in mount_subvol().
>> 1857 *
>> 1858 * 2. Mount device's root (/) by calling vfs_kern_mount().
>> 1859 *
>> 1860 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
>> 1861 * first place. In order to avoid calling btrfs_mount() again, we use
>> 1862 * different file_system_type which is not registered to VFS by
>> 1863 * register_filesystem() (btrfs_root_fs_type). As a result,
>> 1864 * btrfs_mount_root() is called. The return value will be used by
>> 1865 * mount_subtree() in mount_subvol().
>> 1866 *
>> 1867 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
>> 1868 * "btrfs subvolume set-default", mount_subvol() is called always.
>> 1869 */
>> 1870 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
>> 1871 const char *device_name, void *data)
>> 1872 {
>> 1873 struct vfsmount *mnt_root;
>> 1874 struct dentry *root;
>> 1875 char *subvol_name = NULL;
>> 1876 u64 subvol_objectid = 0;
>> 1877 int error = 0;
>> 1878
>> 1879 error = btrfs_parse_subvol_options(data, &subvol_name,
>> 1880 &subvol_objectid);
>> 1881 if (error) {
>> 1882 kfree(subvol_name);
>> 1883 return ERR_PTR(error);
>> 1884 }
>> 1885
>> 1886 /* mount device's root (/) */
>> 1887 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
>> 1888 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
>> 1889 if (flags & SB_RDONLY) {
>> 1890 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
>> 1891 flags & ~SB_RDONLY, device_name, data);
>> 1892 } else {
>> 1893 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
>> 1894 flags | SB_RDONLY, device_name, data);
>> 1895 if (IS_ERR(mnt_root)) {
>> 1896 root = ERR_CAST(mnt_root);
>> 1897 kfree(subvol_name);
>> 1898 goto out;
>> 1899 }
>> 1900
>> 1901 down_write(&mnt_root->mnt_sb->s_umount);
>> 1902 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
>> 1903 up_write(&mnt_root->mnt_sb->s_umount);
>> 1904 if (error < 0) {
>> 1905 root = ERR_PTR(error);
>> 1906 mntput(mnt_root);
>> 1907 kfree(subvol_name);
>> 1908 goto out;
>> 1909 }
>> 1910 }
>> 1911 }
>> 1912 if (IS_ERR(mnt_root)) {
>> 1913 root = ERR_CAST(mnt_root);
>> 1914 kfree(subvol_name);
>> 1915 goto out;
>> 1916 }
>> 1917
>> 1918 /* mount_subvol() will free subvol_name and mnt_root */
>> 1919 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
>> 1920
>> 1921 out:
>> 1922 return root;
>> 1923 }
>> 1924
1226 static void btrfs_resize_thread_pool(struct b 1925 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1227 u32 new_ 1926 u32 new_pool_size, u32 old_pool_size)
1228 { 1927 {
1229 if (new_pool_size == old_pool_size) 1928 if (new_pool_size == old_pool_size)
1230 return; 1929 return;
1231 1930
1232 fs_info->thread_pool_size = new_pool_ 1931 fs_info->thread_pool_size = new_pool_size;
1233 1932
1234 btrfs_info(fs_info, "resize thread po 1933 btrfs_info(fs_info, "resize thread pool %d -> %d",
1235 old_pool_size, new_pool_size); 1934 old_pool_size, new_pool_size);
1236 1935
1237 btrfs_workqueue_set_max(fs_info->work 1936 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
>> 1937 btrfs_workqueue_set_max(fs_info->hipri_workers, new_pool_size);
1238 btrfs_workqueue_set_max(fs_info->dela 1938 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1239 btrfs_workqueue_set_max(fs_info->cach 1939 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1240 workqueue_set_max_active(fs_info->end <<
1241 workqueue_set_max_active(fs_info->end <<
1242 btrfs_workqueue_set_max(fs_info->endi 1940 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1243 btrfs_workqueue_set_max(fs_info->endi 1941 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1244 btrfs_workqueue_set_max(fs_info->dela 1942 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1245 } 1943 }
1246 1944
1247 static inline void btrfs_remount_begin(struct 1945 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1248 unsign !! 1946 unsigned long old_opts, int flags)
1249 { 1947 {
1250 if (btrfs_raw_test_opt(old_opts, AUTO 1948 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1251 (!btrfs_raw_test_opt(fs_info->mou 1949 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1252 (flags & SB_RDONLY))) { 1950 (flags & SB_RDONLY))) {
1253 /* wait for any defraggers to 1951 /* wait for any defraggers to finish */
1254 wait_event(fs_info->transacti 1952 wait_event(fs_info->transaction_wait,
1255 (atomic_read(&fs_i 1953 (atomic_read(&fs_info->defrag_running) == 0));
1256 if (flags & SB_RDONLY) 1954 if (flags & SB_RDONLY)
1257 sync_filesystem(fs_in 1955 sync_filesystem(fs_info->sb);
1258 } 1956 }
1259 } 1957 }
1260 1958
1261 static inline void btrfs_remount_cleanup(stru 1959 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1262 unsi !! 1960 unsigned long old_opts)
1263 { 1961 {
1264 const bool cache_opt = btrfs_test_opt 1962 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1265 1963
1266 /* 1964 /*
1267 * We need to cleanup all defragable 1965 * We need to cleanup all defragable inodes if the autodefragment is
1268 * close or the filesystem is read on 1966 * close or the filesystem is read only.
1269 */ 1967 */
1270 if (btrfs_raw_test_opt(old_opts, AUTO 1968 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1271 (!btrfs_raw_test_opt(fs_info->mou 1969 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1272 btrfs_cleanup_defrag_inodes(f 1970 btrfs_cleanup_defrag_inodes(fs_info);
1273 } 1971 }
1274 1972
1275 /* If we toggled discard async */ 1973 /* If we toggled discard async */
1276 if (!btrfs_raw_test_opt(old_opts, DIS 1974 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1277 btrfs_test_opt(fs_info, DISCARD_A 1975 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1278 btrfs_discard_resume(fs_info) 1976 btrfs_discard_resume(fs_info);
1279 else if (btrfs_raw_test_opt(old_opts, 1977 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1280 !btrfs_test_opt(fs_info, DIS 1978 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1281 btrfs_discard_cleanup(fs_info 1979 btrfs_discard_cleanup(fs_info);
1282 1980
1283 /* If we toggled space cache */ 1981 /* If we toggled space cache */
1284 if (cache_opt != btrfs_free_space_cac 1982 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1285 btrfs_set_free_space_cache_v1 1983 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1286 } 1984 }
1287 1985
1288 static int btrfs_remount_rw(struct btrfs_fs_i !! 1986 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1289 { 1987 {
>> 1988 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 1989 unsigned old_flags = sb->s_flags;
>> 1990 unsigned long old_opts = fs_info->mount_opt;
>> 1991 unsigned long old_compress_type = fs_info->compress_type;
>> 1992 u64 old_max_inline = fs_info->max_inline;
>> 1993 u32 old_thread_pool_size = fs_info->thread_pool_size;
>> 1994 u32 old_metadata_ratio = fs_info->metadata_ratio;
1290 int ret; 1995 int ret;
1291 1996
1292 if (BTRFS_FS_ERROR(fs_info)) { !! 1997 sync_filesystem(sb);
1293 btrfs_err(fs_info, !! 1998 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1294 "remounting read-wr <<
1295 return -EINVAL; <<
1296 } <<
1297 <<
1298 if (fs_info->fs_devices->rw_devices = <<
1299 return -EACCES; <<
1300 1999
1301 if (!btrfs_check_rw_degradable(fs_inf !! 2000 if (data) {
1302 btrfs_warn(fs_info, !! 2001 void *new_sec_opts = NULL;
1303 "too many missing <<
1304 return -EACCES; <<
1305 } <<
1306 2002
1307 if (btrfs_super_log_root(fs_info->sup !! 2003 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1308 btrfs_warn(fs_info, !! 2004 if (!ret)
1309 "mount required to !! 2005 ret = security_sb_remount(sb, new_sec_opts);
1310 return -EINVAL; !! 2006 security_free_mnt_opts(&new_sec_opts);
>> 2007 if (ret)
>> 2008 goto restore;
1311 } 2009 }
1312 2010
1313 /* !! 2011 ret = btrfs_parse_options(fs_info, data, *flags);
1314 * NOTE: when remounting with a chang <<
1315 * anywhere above this point, as we a <<
1316 * until we pass the above checks. <<
1317 */ <<
1318 ret = btrfs_start_pre_rw_mount(fs_inf <<
1319 if (ret) 2012 if (ret)
1320 return ret; !! 2013 goto restore;
1321 <<
1322 btrfs_clear_sb_rdonly(fs_info->sb); <<
1323 <<
1324 set_bit(BTRFS_FS_OPEN, &fs_info->flag <<
1325 <<
1326 /* <<
1327 * If we've gone from readonly -> rea <<
1328 * sync/async discard lists in the ri <<
1329 */ <<
1330 btrfs_discard_resume(fs_info); <<
1331 <<
1332 return 0; <<
1333 } <<
1334 <<
1335 static int btrfs_remount_ro(struct btrfs_fs_i <<
1336 { <<
1337 /* <<
1338 * This also happens on 'umount -rf' <<
1339 * filesystem is busy. <<
1340 */ <<
1341 cancel_work_sync(&fs_info->async_recl <<
1342 cancel_work_sync(&fs_info->async_data <<
1343 2014
1344 btrfs_discard_cleanup(fs_info); !! 2015 ret = btrfs_check_features(fs_info, !(*flags & SB_RDONLY));
>> 2016 if (ret < 0)
>> 2017 goto restore;
1345 2018
1346 /* Wait for the uuid_scan task to fin !! 2019 btrfs_remount_begin(fs_info, old_opts, *flags);
1347 down(&fs_info->uuid_tree_rescan_sem); !! 2020 btrfs_resize_thread_pool(fs_info,
1348 /* Avoid complains from lockdep et al !! 2021 fs_info->thread_pool_size, old_thread_pool_size);
1349 up(&fs_info->uuid_tree_rescan_sem); <<
1350 2022
1351 btrfs_set_sb_rdonly(fs_info->sb); !! 2023 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
>> 2024 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
>> 2025 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
>> 2026 btrfs_warn(fs_info,
>> 2027 "remount supports changing free space tree only from ro to rw");
>> 2028 /* Make sure free space cache options match the state on disk */
>> 2029 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
>> 2030 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
>> 2031 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
>> 2032 }
>> 2033 if (btrfs_free_space_cache_v1_active(fs_info)) {
>> 2034 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
>> 2035 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
>> 2036 }
>> 2037 }
1352 2038
1353 /* !! 2039 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1354 * Setting SB_RDONLY will put the cle !! 2040 goto out;
1355 * loop if it's already active. If i <<
1356 * unused block groups on disk until <<
1357 * unless we clean them up here. <<
1358 */ <<
1359 btrfs_delete_unused_bgs(fs_info); <<
1360 2041
1361 /* !! 2042 if (*flags & SB_RDONLY) {
1362 * The cleaner task could be already !! 2043 /*
1363 * BTRFS_FS_STATE_RO (and SB_RDONLY i !! 2044 * this also happens on 'umount -rf' or on shutdown, when
1364 * sure that after we finish the remo !! 2045 * the filesystem is busy.
1365 * btrfs_commit_super(), the cleaner !! 2046 */
1366 * - either because it was dropping a !! 2047 cancel_work_sync(&fs_info->async_reclaim_work);
1367 * or deleting an unused block grou !! 2048 cancel_work_sync(&fs_info->async_data_reclaim_work);
1368 * group from the list of unused bl <<
1369 * in the previous call to btrfs_de <<
1370 */ <<
1371 wait_on_bit(&fs_info->flags, BTRFS_FS <<
1372 2049
1373 /* !! 2050 btrfs_discard_cleanup(fs_info);
1374 * We've set the superblock to RO mod <<
1375 * cleaner task sleep without running <<
1376 * through all the delayed iputs here <<
1377 * without remounting RW we don't end <<
1378 * with a non-empty list of delayed i <<
1379 */ <<
1380 btrfs_run_delayed_iputs(fs_info); <<
1381 2051
1382 btrfs_dev_replace_suspend_for_unmount !! 2052 /* wait for the uuid_scan task to finish */
1383 btrfs_scrub_cancel(fs_info); !! 2053 down(&fs_info->uuid_tree_rescan_sem);
1384 btrfs_pause_balance(fs_info); !! 2054 /* avoid complains from lockdep et al. */
>> 2055 up(&fs_info->uuid_tree_rescan_sem);
1385 2056
1386 /* !! 2057 btrfs_set_sb_rdonly(sb);
1387 * Pause the qgroup rescan worker if <<
1388 * be still running after we are in R <<
1389 * we unmount, it might have left a t <<
1390 * the transaction and/or crash. <<
1391 */ <<
1392 btrfs_qgroup_wait_for_completion(fs_i <<
1393 2058
1394 return btrfs_commit_super(fs_info); !! 2059 /*
1395 } !! 2060 * Setting SB_RDONLY will put the cleaner thread to
>> 2061 * sleep at the next loop if it's already active.
>> 2062 * If it's already asleep, we'll leave unused block
>> 2063 * groups on disk until we're mounted read-write again
>> 2064 * unless we clean them up here.
>> 2065 */
>> 2066 btrfs_delete_unused_bgs(fs_info);
1396 2067
1397 static void btrfs_ctx_to_info(struct btrfs_fs !! 2068 /*
1398 { !! 2069 * The cleaner task could be already running before we set the
1399 fs_info->max_inline = ctx->max_inline !! 2070 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1400 fs_info->commit_interval = ctx->commi !! 2071 * We must make sure that after we finish the remount, i.e. after
1401 fs_info->metadata_ratio = ctx->metada !! 2072 * we call btrfs_commit_super(), the cleaner can no longer start
1402 fs_info->thread_pool_size = ctx->thre !! 2073 * a transaction - either because it was dropping a dead root,
1403 fs_info->mount_opt = ctx->mount_opt; !! 2074 * running delayed iputs or deleting an unused block group (the
1404 fs_info->compress_type = ctx->compres !! 2075 * cleaner picked a block group from the list of unused block
1405 fs_info->compress_level = ctx->compre !! 2076 * groups before we were able to in the previous call to
1406 } !! 2077 * btrfs_delete_unused_bgs()).
1407 !! 2078 */
1408 static void btrfs_info_to_ctx(struct btrfs_fs !! 2079 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1409 { !! 2080 TASK_UNINTERRUPTIBLE);
1410 ctx->max_inline = fs_info->max_inline <<
1411 ctx->commit_interval = fs_info->commi <<
1412 ctx->metadata_ratio = fs_info->metada <<
1413 ctx->thread_pool_size = fs_info->thre <<
1414 ctx->mount_opt = fs_info->mount_opt; <<
1415 ctx->compress_type = fs_info->compres <<
1416 ctx->compress_level = fs_info->compre <<
1417 } <<
1418 <<
1419 #define btrfs_info_if_set(fs_info, old_ctx, o <<
1420 do { <<
1421 if ((!old_ctx || !btrfs_raw_test_opt( <<
1422 btrfs_raw_test_opt(fs_info->mount <<
1423 btrfs_info(fs_info, fmt, ##ar <<
1424 } while (0) <<
1425 <<
1426 #define btrfs_info_if_unset(fs_info, old_ctx, <<
1427 do { <<
1428 if ((old_ctx && btrfs_raw_test_opt(ol <<
1429 !btrfs_raw_test_opt(fs_info->moun <<
1430 btrfs_info(fs_info, fmt, ##ar <<
1431 } while (0) <<
1432 <<
1433 static void btrfs_emit_options(struct btrfs_f <<
1434 struct btrfs_f <<
1435 { <<
1436 btrfs_info_if_set(info, old, NODATASU <<
1437 btrfs_info_if_set(info, old, DEGRADED <<
1438 btrfs_info_if_set(info, old, NODATASU <<
1439 btrfs_info_if_set(info, old, SSD, "en <<
1440 btrfs_info_if_set(info, old, SSD_SPRE <<
1441 btrfs_info_if_set(info, old, NOBARRIE <<
1442 btrfs_info_if_set(info, old, NOTREELO <<
1443 btrfs_info_if_set(info, old, NOLOGREP <<
1444 btrfs_info_if_set(info, old, FLUSHONC <<
1445 btrfs_info_if_set(info, old, DISCARD_ <<
1446 btrfs_info_if_set(info, old, DISCARD_ <<
1447 btrfs_info_if_set(info, old, FREE_SPA <<
1448 btrfs_info_if_set(info, old, SPACE_CA <<
1449 btrfs_info_if_set(info, old, CLEAR_CA <<
1450 btrfs_info_if_set(info, old, AUTO_DEF <<
1451 btrfs_info_if_set(info, old, FRAGMENT <<
1452 btrfs_info_if_set(info, old, FRAGMENT <<
1453 btrfs_info_if_set(info, old, REF_VERI <<
1454 btrfs_info_if_set(info, old, USEBACKU <<
1455 btrfs_info_if_set(info, old, IGNOREBA <<
1456 btrfs_info_if_set(info, old, IGNOREDA <<
1457 btrfs_info_if_set(info, old, IGNOREME <<
1458 btrfs_info_if_set(info, old, IGNORESU <<
1459 <<
1460 btrfs_info_if_unset(info, old, NODATA <<
1461 btrfs_info_if_unset(info, old, SSD, " <<
1462 btrfs_info_if_unset(info, old, SSD_SP <<
1463 btrfs_info_if_unset(info, old, NOBARR <<
1464 btrfs_info_if_unset(info, old, NOTREE <<
1465 btrfs_info_if_unset(info, old, SPACE_ <<
1466 btrfs_info_if_unset(info, old, FREE_S <<
1467 btrfs_info_if_unset(info, old, AUTO_D <<
1468 btrfs_info_if_unset(info, old, COMPRE <<
1469 <<
1470 /* Did the compression settings chang <<
1471 if (btrfs_test_opt(info, COMPRESS) && <<
1472 (!old || <<
1473 old->compress_type != info->comp <<
1474 old->compress_level != info->com <<
1475 (!btrfs_raw_test_opt(old->mount_ <<
1476 btrfs_raw_test_opt(info->mount_ <<
1477 const char *compress_type = b <<
1478 <<
1479 btrfs_info(info, "%s %s compr <<
1480 btrfs_test_opt(inf <<
1481 compress_type, inf <<
1482 } <<
1483 2081
1484 if (info->max_inline != BTRFS_DEFAULT !! 2082 /*
1485 btrfs_info(info, "max_inline !! 2083 * We've set the superblock to RO mode, so we might have made
1486 } !! 2084 * the cleaner task sleep without running all pending delayed
>> 2085 * iputs. Go through all the delayed iputs here, so that if an
>> 2086 * unmount happens without remounting RW we don't end up at
>> 2087 * finishing close_ctree() with a non-empty list of delayed
>> 2088 * iputs.
>> 2089 */
>> 2090 btrfs_run_delayed_iputs(fs_info);
1487 2091
1488 static int btrfs_reconfigure(struct fs_contex !! 2092 btrfs_dev_replace_suspend_for_unmount(fs_info);
1489 { !! 2093 btrfs_scrub_cancel(fs_info);
1490 struct super_block *sb = fc->root->d_ !! 2094 btrfs_pause_balance(fs_info);
1491 struct btrfs_fs_info *fs_info = btrfs <<
1492 struct btrfs_fs_context *ctx = fc->fs <<
1493 struct btrfs_fs_context old_ctx; <<
1494 int ret = 0; <<
1495 bool mount_reconfigure = (fc->s_fs_in <<
1496 2095
1497 btrfs_info_to_ctx(fs_info, &old_ctx); !! 2096 /*
>> 2097 * Pause the qgroup rescan worker if it is running. We don't want
>> 2098 * it to be still running after we are in RO mode, as after that,
>> 2099 * by the time we unmount, it might have left a transaction open,
>> 2100 * so we would leak the transaction and/or crash.
>> 2101 */
>> 2102 btrfs_qgroup_wait_for_completion(fs_info, false);
1498 2103
1499 /* !! 2104 ret = btrfs_commit_super(fs_info);
1500 * This is our "bind mount" trick, we !! 2105 if (ret)
1501 * anything other than mount a differ !! 2106 goto restore;
1502 * all of the mount options should be !! 2107 } else {
1503 */ !! 2108 if (BTRFS_FS_ERROR(fs_info)) {
1504 if (mount_reconfigure) !! 2109 btrfs_err(fs_info,
1505 ctx->mount_opt = old_ctx.moun !! 2110 "Remounting read-write after error is not allowed");
>> 2111 ret = -EINVAL;
>> 2112 goto restore;
>> 2113 }
>> 2114 if (fs_info->fs_devices->rw_devices == 0) {
>> 2115 ret = -EACCES;
>> 2116 goto restore;
>> 2117 }
1506 2118
1507 sync_filesystem(sb); !! 2119 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1508 set_bit(BTRFS_FS_STATE_REMOUNTING, &f !! 2120 btrfs_warn(fs_info,
>> 2121 "too many missing devices, writable remount is not allowed");
>> 2122 ret = -EACCES;
>> 2123 goto restore;
>> 2124 }
1509 2125
1510 if (!btrfs_check_options(fs_info, &ct !! 2126 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1511 return -EINVAL; !! 2127 btrfs_warn(fs_info,
>> 2128 "mount required to replay tree-log, cannot remount read-write");
>> 2129 ret = -EINVAL;
>> 2130 goto restore;
>> 2131 }
1512 2132
1513 ret = btrfs_check_features(fs_info, ! !! 2133 /*
1514 if (ret < 0) !! 2134 * NOTE: when remounting with a change that does writes, don't
1515 return ret; !! 2135 * put it anywhere above this point, as we are not sure to be
>> 2136 * safe to write until we pass the above checks.
>> 2137 */
>> 2138 ret = btrfs_start_pre_rw_mount(fs_info);
>> 2139 if (ret)
>> 2140 goto restore;
1516 2141
1517 btrfs_ctx_to_info(fs_info, ctx); !! 2142 btrfs_clear_sb_rdonly(sb);
1518 btrfs_remount_begin(fs_info, old_ctx. <<
1519 btrfs_resize_thread_pool(fs_info, fs_ <<
1520 old_ctx.thre <<
1521 2143
1522 if ((bool)btrfs_test_opt(fs_info, FRE !! 2144 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1523 (bool)btrfs_fs_compat_ro(fs_info, <<
1524 (!sb_rdonly(sb) || (fc->sb_flags <<
1525 btrfs_warn(fs_info, <<
1526 "remount supports changing fr <<
1527 /* Make sure free space cache <<
1528 if (btrfs_fs_compat_ro(fs_inf <<
1529 btrfs_set_opt(fs_info <<
1530 btrfs_clear_opt(fs_in <<
1531 } <<
1532 if (btrfs_free_space_cache_v1 <<
1533 btrfs_clear_opt(fs_in <<
1534 btrfs_set_opt(fs_info <<
1535 } <<
1536 } 2145 }
1537 !! 2146 out:
1538 ret = 0; <<
1539 if (!sb_rdonly(sb) && (fc->sb_flags & <<
1540 ret = btrfs_remount_ro(fs_inf <<
1541 else if (sb_rdonly(sb) && !(fc->sb_fl <<
1542 ret = btrfs_remount_rw(fs_inf <<
1543 if (ret) <<
1544 goto restore; <<
1545 <<
1546 /* 2147 /*
1547 * If we set the mask during the para !! 2148 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1548 * remount. Here we can set the mask !! 2149 * since the absence of the flag means it can be toggled off by remount.
1549 * appropriately. <<
1550 */ 2150 */
1551 if ((fc->sb_flags & SB_POSIXACL) != ( !! 2151 *flags |= SB_I_VERSION;
1552 fc->sb_flags_mask |= SB_POSIX <<
1553 2152
1554 btrfs_emit_options(fs_info, &old_ctx) <<
1555 wake_up_process(fs_info->transaction_ 2153 wake_up_process(fs_info->transaction_kthread);
1556 btrfs_remount_cleanup(fs_info, old_ct !! 2154 btrfs_remount_cleanup(fs_info, old_opts);
1557 btrfs_clear_oneshot_options(fs_info); 2155 btrfs_clear_oneshot_options(fs_info);
1558 clear_bit(BTRFS_FS_STATE_REMOUNTING, 2156 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1559 2157
1560 return 0; 2158 return 0;
>> 2159
1561 restore: 2160 restore:
1562 btrfs_ctx_to_info(fs_info, &old_ctx); !! 2161 /* We've hit an error - don't reset SB_RDONLY */
1563 btrfs_remount_cleanup(fs_info, old_ct !! 2162 if (sb_rdonly(sb))
>> 2163 old_flags |= SB_RDONLY;
>> 2164 if (!(old_flags & SB_RDONLY))
>> 2165 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
>> 2166 sb->s_flags = old_flags;
>> 2167 fs_info->mount_opt = old_opts;
>> 2168 fs_info->compress_type = old_compress_type;
>> 2169 fs_info->max_inline = old_max_inline;
>> 2170 btrfs_resize_thread_pool(fs_info,
>> 2171 old_thread_pool_size, fs_info->thread_pool_size);
>> 2172 fs_info->metadata_ratio = old_metadata_ratio;
>> 2173 btrfs_remount_cleanup(fs_info, old_opts);
1564 clear_bit(BTRFS_FS_STATE_REMOUNTING, 2174 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
>> 2175
1565 return ret; 2176 return ret;
1566 } 2177 }
1567 2178
1568 /* Used to sort the devices by max_avail(desc 2179 /* Used to sort the devices by max_avail(descending sort) */
1569 static int btrfs_cmp_device_free_bytes(const 2180 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1570 { 2181 {
1571 const struct btrfs_device_info *dev_i 2182 const struct btrfs_device_info *dev_info1 = a;
1572 const struct btrfs_device_info *dev_i 2183 const struct btrfs_device_info *dev_info2 = b;
1573 2184
1574 if (dev_info1->max_avail > dev_info2- 2185 if (dev_info1->max_avail > dev_info2->max_avail)
1575 return -1; 2186 return -1;
1576 else if (dev_info1->max_avail < dev_i 2187 else if (dev_info1->max_avail < dev_info2->max_avail)
1577 return 1; 2188 return 1;
1578 return 0; 2189 return 0;
1579 } 2190 }
1580 2191
1581 /* 2192 /*
1582 * sort the devices by max_avail, in which ma 2193 * sort the devices by max_avail, in which max free extent size of each device
1583 * is stored.(Descending Sort) 2194 * is stored.(Descending Sort)
1584 */ 2195 */
1585 static inline void btrfs_descending_sort_devi 2196 static inline void btrfs_descending_sort_devices(
1586 struc 2197 struct btrfs_device_info *devices,
1587 size_ 2198 size_t nr_devices)
1588 { 2199 {
1589 sort(devices, nr_devices, sizeof(stru 2200 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1590 btrfs_cmp_device_free_bytes, NUL 2201 btrfs_cmp_device_free_bytes, NULL);
1591 } 2202 }
1592 2203
1593 /* 2204 /*
1594 * The helper to calc the free space on the d 2205 * The helper to calc the free space on the devices that can be used to store
1595 * file data. 2206 * file data.
1596 */ 2207 */
1597 static inline int btrfs_calc_avail_data_space 2208 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1598 2209 u64 *free_bytes)
1599 { 2210 {
1600 struct btrfs_device_info *devices_inf 2211 struct btrfs_device_info *devices_info;
1601 struct btrfs_fs_devices *fs_devices = 2212 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1602 struct btrfs_device *device; 2213 struct btrfs_device *device;
1603 u64 type; 2214 u64 type;
1604 u64 avail_space; 2215 u64 avail_space;
1605 u64 min_stripe_size; 2216 u64 min_stripe_size;
1606 int num_stripes = 1; 2217 int num_stripes = 1;
1607 int i = 0, nr_devices; 2218 int i = 0, nr_devices;
1608 const struct btrfs_raid_attr *rattr; 2219 const struct btrfs_raid_attr *rattr;
1609 2220
1610 /* 2221 /*
1611 * We aren't under the device list lo 2222 * We aren't under the device list lock, so this is racy-ish, but good
1612 * enough for our purposes. 2223 * enough for our purposes.
1613 */ 2224 */
1614 nr_devices = fs_info->fs_devices->ope 2225 nr_devices = fs_info->fs_devices->open_devices;
1615 if (!nr_devices) { 2226 if (!nr_devices) {
1616 smp_mb(); 2227 smp_mb();
1617 nr_devices = fs_info->fs_devi 2228 nr_devices = fs_info->fs_devices->open_devices;
1618 ASSERT(nr_devices); 2229 ASSERT(nr_devices);
1619 if (!nr_devices) { 2230 if (!nr_devices) {
1620 *free_bytes = 0; 2231 *free_bytes = 0;
1621 return 0; 2232 return 0;
1622 } 2233 }
1623 } 2234 }
1624 2235
1625 devices_info = kmalloc_array(nr_devic 2236 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1626 GFP_KERNEL); 2237 GFP_KERNEL);
1627 if (!devices_info) 2238 if (!devices_info)
1628 return -ENOMEM; 2239 return -ENOMEM;
1629 2240
1630 /* calc min stripe number for data sp 2241 /* calc min stripe number for data space allocation */
1631 type = btrfs_data_alloc_profile(fs_in 2242 type = btrfs_data_alloc_profile(fs_info);
1632 rattr = &btrfs_raid_array[btrfs_bg_fl 2243 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1633 2244
1634 if (type & BTRFS_BLOCK_GROUP_RAID0) 2245 if (type & BTRFS_BLOCK_GROUP_RAID0)
1635 num_stripes = nr_devices; 2246 num_stripes = nr_devices;
1636 else if (type & BTRFS_BLOCK_GROUP_RAI 2247 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1637 num_stripes = rattr->ncopies; 2248 num_stripes = rattr->ncopies;
1638 else if (type & BTRFS_BLOCK_GROUP_RAI 2249 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1639 num_stripes = 4; 2250 num_stripes = 4;
1640 2251
1641 /* Adjust for more than 1 stripe per 2252 /* Adjust for more than 1 stripe per device */
1642 min_stripe_size = rattr->dev_stripes 2253 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1643 2254
1644 rcu_read_lock(); 2255 rcu_read_lock();
1645 list_for_each_entry_rcu(device, &fs_d 2256 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1646 if (!test_bit(BTRFS_DEV_STATE 2257 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1647 2258 &device->dev_state) ||
1648 !device->bdev || 2259 !device->bdev ||
1649 test_bit(BTRFS_DEV_STATE_ 2260 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1650 continue; 2261 continue;
1651 2262
1652 if (i >= nr_devices) 2263 if (i >= nr_devices)
1653 break; 2264 break;
1654 2265
1655 avail_space = device->total_b 2266 avail_space = device->total_bytes - device->bytes_used;
1656 2267
1657 /* align with stripe_len */ 2268 /* align with stripe_len */
1658 avail_space = rounddown(avail 2269 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1659 2270
1660 /* 2271 /*
1661 * Ensure we have at least mi 2272 * Ensure we have at least min_stripe_size on top of the
1662 * reserved space on the devi 2273 * reserved space on the device.
1663 */ 2274 */
1664 if (avail_space <= BTRFS_DEVI 2275 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1665 continue; 2276 continue;
1666 2277
1667 avail_space -= BTRFS_DEVICE_R 2278 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1668 2279
1669 devices_info[i].dev = device; 2280 devices_info[i].dev = device;
1670 devices_info[i].max_avail = a 2281 devices_info[i].max_avail = avail_space;
1671 2282
1672 i++; 2283 i++;
1673 } 2284 }
1674 rcu_read_unlock(); 2285 rcu_read_unlock();
1675 2286
1676 nr_devices = i; 2287 nr_devices = i;
1677 2288
1678 btrfs_descending_sort_devices(devices 2289 btrfs_descending_sort_devices(devices_info, nr_devices);
1679 2290
1680 i = nr_devices - 1; 2291 i = nr_devices - 1;
1681 avail_space = 0; 2292 avail_space = 0;
1682 while (nr_devices >= rattr->devs_min) 2293 while (nr_devices >= rattr->devs_min) {
1683 num_stripes = min(num_stripes 2294 num_stripes = min(num_stripes, nr_devices);
1684 2295
1685 if (devices_info[i].max_avail 2296 if (devices_info[i].max_avail >= min_stripe_size) {
1686 int j; 2297 int j;
1687 u64 alloc_size; 2298 u64 alloc_size;
1688 2299
1689 avail_space += device 2300 avail_space += devices_info[i].max_avail * num_stripes;
1690 alloc_size = devices_ 2301 alloc_size = devices_info[i].max_avail;
1691 for (j = i + 1 - num_ 2302 for (j = i + 1 - num_stripes; j <= i; j++)
1692 devices_info[ 2303 devices_info[j].max_avail -= alloc_size;
1693 } 2304 }
1694 i--; 2305 i--;
1695 nr_devices--; 2306 nr_devices--;
1696 } 2307 }
1697 2308
1698 kfree(devices_info); 2309 kfree(devices_info);
1699 *free_bytes = avail_space; 2310 *free_bytes = avail_space;
1700 return 0; 2311 return 0;
1701 } 2312 }
1702 2313
1703 /* 2314 /*
1704 * Calculate numbers for 'df', pessimistic in 2315 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1705 * 2316 *
1706 * If there's a redundant raid level at DATA 2317 * If there's a redundant raid level at DATA block groups, use the respective
1707 * multiplier to scale the sizes. 2318 * multiplier to scale the sizes.
1708 * 2319 *
1709 * Unused device space usage is based on simu 2320 * Unused device space usage is based on simulating the chunk allocator
1710 * algorithm that respects the device sizes a 2321 * algorithm that respects the device sizes and order of allocations. This is
1711 * a close approximation of the actual use bu 2322 * a close approximation of the actual use but there are other factors that may
1712 * change the result (like a new metadata chu 2323 * change the result (like a new metadata chunk).
1713 * 2324 *
1714 * If metadata is exhausted, f_bavail will be 2325 * If metadata is exhausted, f_bavail will be 0.
1715 */ 2326 */
1716 static int btrfs_statfs(struct dentry *dentry 2327 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1717 { 2328 {
1718 struct btrfs_fs_info *fs_info = btrfs 2329 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1719 struct btrfs_super_block *disk_super 2330 struct btrfs_super_block *disk_super = fs_info->super_copy;
1720 struct btrfs_space_info *found; 2331 struct btrfs_space_info *found;
1721 u64 total_used = 0; 2332 u64 total_used = 0;
1722 u64 total_free_data = 0; 2333 u64 total_free_data = 0;
1723 u64 total_free_meta = 0; 2334 u64 total_free_meta = 0;
1724 u32 bits = fs_info->sectorsize_bits; 2335 u32 bits = fs_info->sectorsize_bits;
1725 __be32 *fsid = (__be32 *)fs_info->fs_ 2336 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1726 unsigned factor = 1; 2337 unsigned factor = 1;
1727 struct btrfs_block_rsv *block_rsv = & 2338 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1728 int ret; 2339 int ret;
1729 u64 thresh = 0; 2340 u64 thresh = 0;
1730 int mixed = 0; 2341 int mixed = 0;
1731 2342
1732 list_for_each_entry(found, &fs_info-> 2343 list_for_each_entry(found, &fs_info->space_info, list) {
1733 if (found->flags & BTRFS_BLOC 2344 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1734 int i; 2345 int i;
1735 2346
1736 total_free_data += fo 2347 total_free_data += found->disk_total - found->disk_used;
1737 total_free_data -= 2348 total_free_data -=
1738 btrfs_account 2349 btrfs_account_ro_block_groups_free_space(found);
1739 2350
1740 for (i = 0; i < BTRFS 2351 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1741 if (!list_emp 2352 if (!list_empty(&found->block_groups[i]))
1742 facto 2353 factor = btrfs_bg_type_to_factor(
1743 2354 btrfs_raid_array[i].bg_flag);
1744 } 2355 }
1745 } 2356 }
1746 2357
1747 /* 2358 /*
1748 * Metadata in mixed block gr !! 2359 * Metadata in mixed block goup profiles are accounted in data
1749 */ 2360 */
1750 if (!mixed && found->flags & 2361 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1751 if (found->flags & BT 2362 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1752 mixed = 1; 2363 mixed = 1;
1753 else 2364 else
1754 total_free_me 2365 total_free_meta += found->disk_total -
1755 found 2366 found->disk_used;
1756 } 2367 }
1757 2368
1758 total_used += found->disk_use 2369 total_used += found->disk_used;
1759 } 2370 }
1760 2371
1761 buf->f_blocks = div_u64(btrfs_super_t 2372 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1762 buf->f_blocks >>= bits; 2373 buf->f_blocks >>= bits;
1763 buf->f_bfree = buf->f_blocks - (div_u 2374 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1764 2375
1765 /* Account global block reserve as us 2376 /* Account global block reserve as used, it's in logical size already */
1766 spin_lock(&block_rsv->lock); 2377 spin_lock(&block_rsv->lock);
1767 /* Mixed block groups accounting is n 2378 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1768 if (buf->f_bfree >= block_rsv->size > 2379 if (buf->f_bfree >= block_rsv->size >> bits)
1769 buf->f_bfree -= block_rsv->si 2380 buf->f_bfree -= block_rsv->size >> bits;
1770 else 2381 else
1771 buf->f_bfree = 0; 2382 buf->f_bfree = 0;
1772 spin_unlock(&block_rsv->lock); 2383 spin_unlock(&block_rsv->lock);
1773 2384
1774 buf->f_bavail = div_u64(total_free_da 2385 buf->f_bavail = div_u64(total_free_data, factor);
1775 ret = btrfs_calc_avail_data_space(fs_ 2386 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1776 if (ret) 2387 if (ret)
1777 return ret; 2388 return ret;
1778 buf->f_bavail += div_u64(total_free_d 2389 buf->f_bavail += div_u64(total_free_data, factor);
1779 buf->f_bavail = buf->f_bavail >> bits 2390 buf->f_bavail = buf->f_bavail >> bits;
1780 2391
1781 /* 2392 /*
1782 * We calculate the remaining metadat 2393 * We calculate the remaining metadata space minus global reserve. If
1783 * this is (supposedly) smaller than 2394 * this is (supposedly) smaller than zero, there's no space. But this
1784 * does not hold in practice, the exh 2395 * does not hold in practice, the exhausted state happens where's still
1785 * some positive delta. So we apply s 2396 * some positive delta. So we apply some guesswork and compare the
1786 * delta to a 4M threshold. (Practic 2397 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1787 * 2398 *
1788 * We probably cannot calculate the e 2399 * We probably cannot calculate the exact threshold value because this
1789 * depends on the internal reservatio 2400 * depends on the internal reservations requested by various
1790 * operations, so some operations tha 2401 * operations, so some operations that consume a few metadata will
1791 * succeed even if the Avail is zero. 2402 * succeed even if the Avail is zero. But this is better than the other
1792 * way around. 2403 * way around.
1793 */ 2404 */
1794 thresh = SZ_4M; 2405 thresh = SZ_4M;
1795 2406
1796 /* 2407 /*
1797 * We only want to claim there's no a 2408 * We only want to claim there's no available space if we can no longer
1798 * allocate chunks for our metadata p 2409 * allocate chunks for our metadata profile and our global reserve will
1799 * not fit in the free metadata space 2410 * not fit in the free metadata space. If we aren't ->full then we
1800 * still can allocate chunks and thus 2411 * still can allocate chunks and thus are fine using the currently
1801 * calculated f_bavail. 2412 * calculated f_bavail.
1802 */ 2413 */
1803 if (!mixed && block_rsv->space_info-> 2414 if (!mixed && block_rsv->space_info->full &&
1804 (total_free_meta < thresh || tota !! 2415 total_free_meta - thresh < block_rsv->size)
1805 buf->f_bavail = 0; 2416 buf->f_bavail = 0;
1806 2417
1807 buf->f_type = BTRFS_SUPER_MAGIC; 2418 buf->f_type = BTRFS_SUPER_MAGIC;
1808 buf->f_bsize = fs_info->sectorsize; !! 2419 buf->f_bsize = dentry->d_sb->s_blocksize;
1809 buf->f_namelen = BTRFS_NAME_LEN; 2420 buf->f_namelen = BTRFS_NAME_LEN;
1810 2421
1811 /* We treat it as constant endianness 2422 /* We treat it as constant endianness (it doesn't matter _which_)
1812 because we want the fsid to come o 2423 because we want the fsid to come out the same whether mounted
1813 on a big-endian or little-endian h 2424 on a big-endian or little-endian host */
1814 buf->f_fsid.val[0] = be32_to_cpu(fsid 2425 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1815 buf->f_fsid.val[1] = be32_to_cpu(fsid 2426 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1816 /* Mask in the root object ID too, to 2427 /* Mask in the root object ID too, to disambiguate subvols */
1817 buf->f_fsid.val[0] ^= btrfs_root_id(B !! 2428 buf->f_fsid.val[0] ^=
1818 buf->f_fsid.val[1] ^= btrfs_root_id(B !! 2429 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
>> 2430 buf->f_fsid.val[1] ^=
>> 2431 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
1819 2432
1820 return 0; 2433 return 0;
1821 } 2434 }
1822 2435
1823 static int btrfs_fc_test_super(struct super_b <<
1824 { <<
1825 struct btrfs_fs_info *p = fc->s_fs_in <<
1826 struct btrfs_fs_info *fs_info = btrfs <<
1827 <<
1828 return fs_info->fs_devices == p->fs_d <<
1829 } <<
1830 <<
1831 static int btrfs_get_tree_super(struct fs_con <<
1832 { <<
1833 struct btrfs_fs_info *fs_info = fc->s <<
1834 struct btrfs_fs_context *ctx = fc->fs <<
1835 struct btrfs_fs_devices *fs_devices = <<
1836 struct block_device *bdev; <<
1837 struct btrfs_device *device; <<
1838 struct super_block *sb; <<
1839 blk_mode_t mode = btrfs_open_mode(fc) <<
1840 int ret; <<
1841 <<
1842 btrfs_ctx_to_info(fs_info, ctx); <<
1843 mutex_lock(&uuid_mutex); <<
1844 <<
1845 /* <<
1846 * With 'true' passed to btrfs_scan_o <<
1847 * either a valid device or an error. <<
1848 */ <<
1849 device = btrfs_scan_one_device(fc->so <<
1850 ASSERT(device != NULL); <<
1851 if (IS_ERR(device)) { <<
1852 mutex_unlock(&uuid_mutex); <<
1853 return PTR_ERR(device); <<
1854 } <<
1855 <<
1856 fs_devices = device->fs_devices; <<
1857 fs_info->fs_devices = fs_devices; <<
1858 <<
1859 ret = btrfs_open_devices(fs_devices, <<
1860 mutex_unlock(&uuid_mutex); <<
1861 if (ret) <<
1862 return ret; <<
1863 <<
1864 if (!(fc->sb_flags & SB_RDONLY) && fs <<
1865 ret = -EACCES; <<
1866 goto error; <<
1867 } <<
1868 <<
1869 bdev = fs_devices->latest_dev->bdev; <<
1870 <<
1871 /* <<
1872 * From now on the error handling is <<
1873 * <<
1874 * If successful, this will transfer <<
1875 * and fc->s_fs_info will be NULL. H <<
1876 * super, we'll still have fc->s_fs_i <<
1877 * completely out it'll be cleaned up <<
1878 * otherwise it's tied to the lifetim <<
1879 */ <<
1880 sb = sget_fc(fc, btrfs_fc_test_super, <<
1881 if (IS_ERR(sb)) { <<
1882 ret = PTR_ERR(sb); <<
1883 goto error; <<
1884 } <<
1885 <<
1886 set_device_specific_options(fs_info); <<
1887 <<
1888 if (sb->s_root) { <<
1889 btrfs_close_devices(fs_device <<
1890 if ((fc->sb_flags ^ sb->s_fla <<
1891 ret = -EBUSY; <<
1892 } else { <<
1893 snprintf(sb->s_id, sizeof(sb- <<
1894 shrinker_debugfs_rename(sb->s <<
1895 btrfs_sb(sb)->bdev_holder = & <<
1896 ret = btrfs_fill_super(sb, fs <<
1897 } <<
1898 <<
1899 if (ret) { <<
1900 deactivate_locked_super(sb); <<
1901 return ret; <<
1902 } <<
1903 <<
1904 btrfs_clear_oneshot_options(fs_info); <<
1905 <<
1906 fc->root = dget(sb->s_root); <<
1907 return 0; <<
1908 <<
1909 error: <<
1910 btrfs_close_devices(fs_devices); <<
1911 return ret; <<
1912 } <<
1913 <<
1914 /* <<
1915 * Ever since commit 0723a0473fb4 ("btrfs: al <<
1916 * with different ro/rw options") the followi <<
1917 * <<
1918 * (i) mount /dev/sda3 -o subvol=foo,r <<
1919 * (ii) mount /dev/sda3 -o subvol=bar,r <<
1920 * <<
1921 * which looks nice and innocent but is actua <<
1922 * a long comment. <<
1923 * <<
1924 * On another filesystem a subvolume mount is <<
1925 * <<
1926 * (iii) # create rw superblock + initia <<
1927 * mount -t xfs /dev/sdb /opt/ <<
1928 * <<
1929 * # create ro bind mount <<
1930 * mount --bind -o ro /opt/foo /mn <<
1931 * <<
1932 * # unmount initial mount <<
1933 * umount /opt <<
1934 * <<
1935 * Of course, there's some special subvolume <<
1936 * sb->s_root dentry is really swapped after <<
1937 * it's very close and will help us understan <<
1938 * <<
1939 * The old mount API didn't cleanly distingui <<
1940 * and a superblock being made ro. The only <<
1941 * either object was by passing ms_rdonly. If <<
1942 * mount(2) such as: <<
1943 * <<
1944 * mount("/dev/sdb", "/mnt", "xfs", ms_r <<
1945 * <<
1946 * the MS_RDONLY flag being specified had two <<
1947 * <<
1948 * (1) MNT_READONLY was raised -> the resulti <<
1949 * @mnt->mnt_flags |= MNT_READONLY raised <<
1950 * <<
1951 * (2) MS_RDONLY was passed to the filesystem <<
1952 * made the superblock ro. Note, how SB_R <<
1953 * ms_rdonly and is raised whenever MS_RD <<
1954 * <<
1955 * Creating a subtree mount via (iii) ends up <<
1956 * subtree mounted ro. <<
1957 * <<
1958 * But consider the effect on the old mount A <<
1959 * which combines the distinct step in (iii) <<
1960 * <<
1961 * By issuing (i) both the mount and the supe <<
1962 * is issued the superblock is ro and thus ev <<
1963 * rw it wouldn't help. Hence, btrfs needed t <<
1964 * to rw for (ii) which it did using an inter <<
1965 * <<
1966 * IOW, subvolume mounting was inherently com <<
1967 * MS_RDONLY in mount(2). Note, this ambiguit <<
1968 * "ro" to MS_RDONLY. IOW, in both (i) and (i <<
1969 * passed by mount(8) to mount(2). <<
1970 * <<
1971 * Enter the new mount API. The new mount API <<
1972 * and making a superblock ro. <<
1973 * <<
1974 * (3) To turn a mount ro the MOUNT_ATTR_ONLY <<
1975 * fsmount() or mount_setattr() this is a <<
1976 * specific mount or mount tree that is n <<
1977 * <<
1978 * (4) To turn a superblock ro the "ro" flag <<
1979 * fsconfig(FSCONFIG_SET_FLAG, "ro"). Thi <<
1980 * in fc->sb_flags. <<
1981 * <<
1982 * But, currently the util-linux mount comman <<
1983 * API and is still setting fsconfig(FSCONFIG <<
1984 * btrfs or not, setting the whole super bloc <<
1985 * work with different options work we need t <<
1986 */ <<
1987 static struct vfsmount *btrfs_reconfigure_for <<
1988 { <<
1989 struct vfsmount *mnt; <<
1990 int ret; <<
1991 const bool ro2rw = !(fc->sb_flags & S <<
1992 <<
1993 /* <<
1994 * We got an EBUSY because our SB_RDO <<
1995 * super block, so invert our setting <<
1996 * can get our vfsmount. <<
1997 */ <<
1998 if (ro2rw) <<
1999 fc->sb_flags |= SB_RDONLY; <<
2000 else <<
2001 fc->sb_flags &= ~SB_RDONLY; <<
2002 <<
2003 mnt = fc_mount(fc); <<
2004 if (IS_ERR(mnt)) <<
2005 return mnt; <<
2006 <<
2007 if (!ro2rw) <<
2008 return mnt; <<
2009 <<
2010 /* We need to convert to rw, call rec <<
2011 fc->sb_flags &= ~SB_RDONLY; <<
2012 down_write(&mnt->mnt_sb->s_umount); <<
2013 ret = btrfs_reconfigure(fc); <<
2014 up_write(&mnt->mnt_sb->s_umount); <<
2015 if (ret) { <<
2016 mntput(mnt); <<
2017 return ERR_PTR(ret); <<
2018 } <<
2019 return mnt; <<
2020 } <<
2021 <<
2022 static int btrfs_get_tree_subvol(struct fs_co <<
2023 { <<
2024 struct btrfs_fs_info *fs_info = NULL; <<
2025 struct btrfs_fs_context *ctx = fc->fs <<
2026 struct fs_context *dup_fc; <<
2027 struct dentry *dentry; <<
2028 struct vfsmount *mnt; <<
2029 <<
2030 /* <<
2031 * Setup a dummy root and fs_info for <<
2032 * we don't actually fill this stuff <<
2033 * then open_ctree will properly init <<
2034 * settings later. btrfs_init_fs_inf <<
2035 * of the fs_info (locks and such) to <<
2036 * superblock with our given fs_devic <<
2037 */ <<
2038 fs_info = kvzalloc(sizeof(struct btrf <<
2039 if (!fs_info) <<
2040 return -ENOMEM; <<
2041 <<
2042 fs_info->super_copy = kzalloc(BTRFS_S <<
2043 fs_info->super_for_commit = kzalloc(B <<
2044 if (!fs_info->super_copy || !fs_info- <<
2045 btrfs_free_fs_info(fs_info); <<
2046 return -ENOMEM; <<
2047 } <<
2048 btrfs_init_fs_info(fs_info); <<
2049 <<
2050 dup_fc = vfs_dup_fs_context(fc); <<
2051 if (IS_ERR(dup_fc)) { <<
2052 btrfs_free_fs_info(fs_info); <<
2053 return PTR_ERR(dup_fc); <<
2054 } <<
2055 <<
2056 /* <<
2057 * When we do the sget_fc this gets t <<
2058 * need to set it on the dup_fc as th <<
2059 */ <<
2060 dup_fc->s_fs_info = fs_info; <<
2061 <<
2062 /* <<
2063 * We'll do the security settings in <<
2064 * loop, they were duplicated into du <<
2065 * here. <<
2066 */ <<
2067 security_free_mnt_opts(&fc->security) <<
2068 fc->security = NULL; <<
2069 <<
2070 mnt = fc_mount(dup_fc); <<
2071 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) <<
2072 mnt = btrfs_reconfigure_for_m <<
2073 put_fs_context(dup_fc); <<
2074 if (IS_ERR(mnt)) <<
2075 return PTR_ERR(mnt); <<
2076 <<
2077 /* <<
2078 * This free's ->subvol_name, because <<
2079 * allocate a buffer to hold the subv <<
2080 * reference to it here. <<
2081 */ <<
2082 dentry = mount_subvol(ctx->subvol_nam <<
2083 ctx->subvol_name = NULL; <<
2084 if (IS_ERR(dentry)) <<
2085 return PTR_ERR(dentry); <<
2086 <<
2087 fc->root = dentry; <<
2088 return 0; <<
2089 } <<
2090 <<
2091 static int btrfs_get_tree(struct fs_context * <<
2092 { <<
2093 /* <<
2094 * Since we use mount_subtree to moun <<
2095 * have to do mounts in two steps. <<
2096 * <<
2097 * First pass through we call btrfs_g <<
2098 * wrapper around fc_mount() to call <<
2099 * we'll call btrfs_get_tree_super(). <<
2100 * everything to open the devices and <<
2101 * with a fully constructed vfsmount <<
2102 * from there we can do our mount_sub <<
2103 * whichever subvol we're mounting an <<
2104 * appropriate dentry for the subvol. <<
2105 */ <<
2106 if (fc->s_fs_info) <<
2107 return btrfs_get_tree_super(f <<
2108 return btrfs_get_tree_subvol(fc); <<
2109 } <<
2110 <<
2111 static void btrfs_kill_super(struct super_blo 2436 static void btrfs_kill_super(struct super_block *sb)
2112 { 2437 {
2113 struct btrfs_fs_info *fs_info = btrfs 2438 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2114 kill_anon_super(sb); 2439 kill_anon_super(sb);
2115 btrfs_free_fs_info(fs_info); 2440 btrfs_free_fs_info(fs_info);
2116 } 2441 }
2117 2442
2118 static void btrfs_free_fs_context(struct fs_c !! 2443 static struct file_system_type btrfs_fs_type = {
2119 { !! 2444 .owner = THIS_MODULE,
2120 struct btrfs_fs_context *ctx = fc->fs !! 2445 .name = "btrfs",
2121 struct btrfs_fs_info *fs_info = fc->s !! 2446 .mount = btrfs_mount,
2122 !! 2447 .kill_sb = btrfs_kill_super,
2123 if (fs_info) !! 2448 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2124 btrfs_free_fs_info(fs_info); <<
2125 <<
2126 if (ctx && refcount_dec_and_test(&ctx <<
2127 kfree(ctx->subvol_name); <<
2128 kfree(ctx); <<
2129 } <<
2130 } <<
2131 <<
2132 static int btrfs_dup_fs_context(struct fs_con <<
2133 { <<
2134 struct btrfs_fs_context *ctx = src_fc <<
2135 <<
2136 /* <<
2137 * Give a ref to our ctx to this dup, <<
2138 * our original fc so we can have the <<
2139 * <<
2140 * We unset ->source in the original <<
2141 * mounting, and then once we free th <<
2142 * need to make sure we're only point <<
2143 */ <<
2144 refcount_inc(&ctx->refs); <<
2145 fc->fs_private = ctx; <<
2146 fc->source = src_fc->source; <<
2147 src_fc->source = NULL; <<
2148 return 0; <<
2149 } <<
2150 <<
2151 static const struct fs_context_operations btr <<
2152 .parse_param = btrfs_parse_param, <<
2153 .reconfigure = btrfs_reconfigure, <<
2154 .get_tree = btrfs_get_tree, <<
2155 .dup = btrfs_dup_fs_contex <<
2156 .free = btrfs_free_fs_conte <<
2157 }; 2449 };
2158 2450
2159 static int btrfs_init_fs_context(struct fs_co !! 2451 static struct file_system_type btrfs_root_fs_type = {
2160 { !! 2452 .owner = THIS_MODULE,
2161 struct btrfs_fs_context *ctx; !! 2453 .name = "btrfs",
2162 !! 2454 .mount = btrfs_mount_root,
2163 ctx = kzalloc(sizeof(struct btrfs_fs_ !! 2455 .kill_sb = btrfs_kill_super,
2164 if (!ctx) !! 2456 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2165 return -ENOMEM; !! 2457 };
2166 <<
2167 refcount_set(&ctx->refs, 1); <<
2168 fc->fs_private = ctx; <<
2169 fc->ops = &btrfs_fs_context_ops; <<
2170 <<
2171 if (fc->purpose == FS_CONTEXT_FOR_REC <<
2172 btrfs_info_to_ctx(btrfs_sb(fc <<
2173 } else { <<
2174 ctx->thread_pool_size = <<
2175 min_t(unsigned long, <<
2176 ctx->max_inline = BTRFS_DEFAU <<
2177 ctx->commit_interval = BTRFS_ <<
2178 } <<
2179 <<
2180 #ifdef CONFIG_BTRFS_FS_POSIX_ACL <<
2181 fc->sb_flags |= SB_POSIXACL; <<
2182 #endif <<
2183 fc->sb_flags |= SB_I_VERSION; <<
2184 <<
2185 return 0; <<
2186 } <<
2187 <<
2188 static struct file_system_type btrfs_fs_type <<
2189 .owner = THIS_MODULE <<
2190 .name = "btrfs", <<
2191 .init_fs_context = btrfs_init_ <<
2192 .parameters = btrfs_fs_pa <<
2193 .kill_sb = btrfs_kill_ <<
2194 .fs_flags = FS_REQUIRES <<
2195 }; <<
2196 2458
2197 MODULE_ALIAS_FS("btrfs"); 2459 MODULE_ALIAS_FS("btrfs");
2198 2460
2199 static int btrfs_control_open(struct inode *i 2461 static int btrfs_control_open(struct inode *inode, struct file *file)
2200 { 2462 {
2201 /* 2463 /*
2202 * The control file's private_data is 2464 * The control file's private_data is used to hold the
2203 * transaction when it is started and 2465 * transaction when it is started and is used to keep
2204 * track of whether a transaction is 2466 * track of whether a transaction is already in progress.
2205 */ 2467 */
2206 file->private_data = NULL; 2468 file->private_data = NULL;
2207 return 0; 2469 return 0;
2208 } 2470 }
2209 2471
2210 /* 2472 /*
2211 * Used by /dev/btrfs-control for devices ioc 2473 * Used by /dev/btrfs-control for devices ioctls.
2212 */ 2474 */
2213 static long btrfs_control_ioctl(struct file * 2475 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2214 unsigned long 2476 unsigned long arg)
2215 { 2477 {
2216 struct btrfs_ioctl_vol_args *vol; 2478 struct btrfs_ioctl_vol_args *vol;
2217 struct btrfs_device *device = NULL; 2479 struct btrfs_device *device = NULL;
2218 dev_t devt = 0; 2480 dev_t devt = 0;
2219 int ret = -ENOTTY; 2481 int ret = -ENOTTY;
2220 2482
2221 if (!capable(CAP_SYS_ADMIN)) 2483 if (!capable(CAP_SYS_ADMIN))
2222 return -EPERM; 2484 return -EPERM;
2223 2485
2224 vol = memdup_user((void __user *)arg, 2486 vol = memdup_user((void __user *)arg, sizeof(*vol));
2225 if (IS_ERR(vol)) 2487 if (IS_ERR(vol))
2226 return PTR_ERR(vol); 2488 return PTR_ERR(vol);
2227 ret = btrfs_check_ioctl_vol_args_path !! 2489 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2228 if (ret < 0) <<
2229 goto out; <<
2230 2490
2231 switch (cmd) { 2491 switch (cmd) {
2232 case BTRFS_IOC_SCAN_DEV: 2492 case BTRFS_IOC_SCAN_DEV:
2233 mutex_lock(&uuid_mutex); 2493 mutex_lock(&uuid_mutex);
2234 /* !! 2494 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2235 * Scanning outside of mount !! 2495 &btrfs_root_fs_type);
2236 * into 0 error code. <<
2237 */ <<
2238 device = btrfs_scan_one_devic <<
2239 ret = PTR_ERR_OR_ZERO(device) 2496 ret = PTR_ERR_OR_ZERO(device);
2240 mutex_unlock(&uuid_mutex); 2497 mutex_unlock(&uuid_mutex);
2241 break; 2498 break;
2242 case BTRFS_IOC_FORGET_DEV: 2499 case BTRFS_IOC_FORGET_DEV:
2243 if (vol->name[0] != 0) { 2500 if (vol->name[0] != 0) {
2244 ret = lookup_bdev(vol 2501 ret = lookup_bdev(vol->name, &devt);
2245 if (ret) 2502 if (ret)
2246 break; 2503 break;
2247 } 2504 }
2248 ret = btrfs_forget_devices(de 2505 ret = btrfs_forget_devices(devt);
2249 break; 2506 break;
2250 case BTRFS_IOC_DEVICES_READY: 2507 case BTRFS_IOC_DEVICES_READY:
2251 mutex_lock(&uuid_mutex); 2508 mutex_lock(&uuid_mutex);
2252 /* !! 2509 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2253 * Scanning outside of mount !! 2510 &btrfs_root_fs_type);
2254 * into 0 error code. !! 2511 if (IS_ERR(device)) {
2255 */ <<
2256 device = btrfs_scan_one_devic <<
2257 if (IS_ERR_OR_NULL(device)) { <<
2258 mutex_unlock(&uuid_mu 2512 mutex_unlock(&uuid_mutex);
2259 ret = PTR_ERR(device) 2513 ret = PTR_ERR(device);
2260 break; 2514 break;
2261 } 2515 }
2262 ret = !(device->fs_devices->n 2516 ret = !(device->fs_devices->num_devices ==
2263 device->fs_devices->t 2517 device->fs_devices->total_devices);
2264 mutex_unlock(&uuid_mutex); 2518 mutex_unlock(&uuid_mutex);
2265 break; 2519 break;
2266 case BTRFS_IOC_GET_SUPPORTED_FEATURES 2520 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2267 ret = btrfs_ioctl_get_support 2521 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2268 break; 2522 break;
2269 } 2523 }
2270 2524
2271 out: <<
2272 kfree(vol); 2525 kfree(vol);
2273 return ret; 2526 return ret;
2274 } 2527 }
2275 2528
2276 static int btrfs_freeze(struct super_block *s 2529 static int btrfs_freeze(struct super_block *sb)
2277 { 2530 {
>> 2531 struct btrfs_trans_handle *trans;
2278 struct btrfs_fs_info *fs_info = btrfs 2532 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 2533 struct btrfs_root *root = fs_info->tree_root;
2279 2534
2280 set_bit(BTRFS_FS_FROZEN, &fs_info->fl 2535 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2281 /* 2536 /*
2282 * We don't need a barrier here, we'l 2537 * We don't need a barrier here, we'll wait for any transaction that
2283 * could be in progress on other thre 2538 * could be in progress on other threads (and do delayed iputs that
2284 * we want to avoid on a frozen files 2539 * we want to avoid on a frozen filesystem), or do the commit
2285 * ourselves. 2540 * ourselves.
2286 */ 2541 */
2287 return btrfs_commit_current_transacti !! 2542 trans = btrfs_attach_transaction_barrier(root);
>> 2543 if (IS_ERR(trans)) {
>> 2544 /* no transaction, don't bother */
>> 2545 if (PTR_ERR(trans) == -ENOENT)
>> 2546 return 0;
>> 2547 return PTR_ERR(trans);
>> 2548 }
>> 2549 return btrfs_commit_transaction(trans);
2288 } 2550 }
2289 2551
2290 static int check_dev_super(struct btrfs_devic 2552 static int check_dev_super(struct btrfs_device *dev)
2291 { 2553 {
2292 struct btrfs_fs_info *fs_info = dev-> 2554 struct btrfs_fs_info *fs_info = dev->fs_info;
2293 struct btrfs_super_block *sb; 2555 struct btrfs_super_block *sb;
2294 u64 last_trans; <<
2295 u16 csum_type; 2556 u16 csum_type;
2296 int ret = 0; 2557 int ret = 0;
2297 2558
2298 /* This should be called with fs stil 2559 /* This should be called with fs still frozen. */
2299 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_ 2560 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2300 2561
2301 /* Missing dev, no need to check. */ 2562 /* Missing dev, no need to check. */
2302 if (!dev->bdev) 2563 if (!dev->bdev)
2303 return 0; 2564 return 0;
2304 2565
2305 /* Only need to check the primary sup 2566 /* Only need to check the primary super block. */
2306 sb = btrfs_read_dev_one_super(dev->bd 2567 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2307 if (IS_ERR(sb)) 2568 if (IS_ERR(sb))
2308 return PTR_ERR(sb); 2569 return PTR_ERR(sb);
2309 2570
2310 /* Verify the checksum. */ 2571 /* Verify the checksum. */
2311 csum_type = btrfs_super_csum_type(sb) 2572 csum_type = btrfs_super_csum_type(sb);
2312 if (csum_type != btrfs_super_csum_typ 2573 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2313 btrfs_err(fs_info, "csum type 2574 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2314 csum_type, btrfs_su 2575 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2315 ret = -EUCLEAN; 2576 ret = -EUCLEAN;
2316 goto out; 2577 goto out;
2317 } 2578 }
2318 2579
2319 if (btrfs_check_super_csum(fs_info, s 2580 if (btrfs_check_super_csum(fs_info, sb)) {
2320 btrfs_err(fs_info, "csum for 2581 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2321 ret = -EUCLEAN; 2582 ret = -EUCLEAN;
2322 goto out; 2583 goto out;
2323 } 2584 }
2324 2585
2325 /* Btrfs_validate_super() includes fs 2586 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2326 ret = btrfs_validate_super(fs_info, s 2587 ret = btrfs_validate_super(fs_info, sb, 0);
2327 if (ret < 0) 2588 if (ret < 0)
2328 goto out; 2589 goto out;
2329 2590
2330 last_trans = btrfs_get_last_trans_com !! 2591 if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2331 if (btrfs_super_generation(sb) != las <<
2332 btrfs_err(fs_info, "transid m 2592 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2333 btrfs_super_generat !! 2593 btrfs_super_generation(sb),
>> 2594 fs_info->last_trans_committed);
2334 ret = -EUCLEAN; 2595 ret = -EUCLEAN;
2335 goto out; 2596 goto out;
2336 } 2597 }
2337 out: 2598 out:
2338 btrfs_release_disk_super(sb); 2599 btrfs_release_disk_super(sb);
2339 return ret; 2600 return ret;
2340 } 2601 }
2341 2602
2342 static int btrfs_unfreeze(struct super_block 2603 static int btrfs_unfreeze(struct super_block *sb)
2343 { 2604 {
2344 struct btrfs_fs_info *fs_info = btrfs 2605 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2345 struct btrfs_device *device; 2606 struct btrfs_device *device;
2346 int ret = 0; 2607 int ret = 0;
2347 2608
2348 /* 2609 /*
2349 * Make sure the fs is not changed by 2610 * Make sure the fs is not changed by accident (like hibernation then
2350 * modified by other OS). 2611 * modified by other OS).
2351 * If we found anything wrong, we mar 2612 * If we found anything wrong, we mark the fs error immediately.
2352 * 2613 *
2353 * And since the fs is frozen, no one 2614 * And since the fs is frozen, no one can modify the fs yet, thus
2354 * we don't need to hold device_list_ 2615 * we don't need to hold device_list_mutex.
2355 */ 2616 */
2356 list_for_each_entry(device, &fs_info- 2617 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2357 ret = check_dev_super(device) 2618 ret = check_dev_super(device);
2358 if (ret < 0) { 2619 if (ret < 0) {
2359 btrfs_handle_fs_error 2620 btrfs_handle_fs_error(fs_info, ret,
2360 "super block 2621 "super block on devid %llu got modified unexpectedly",
2361 device->devid 2622 device->devid);
2362 break; 2623 break;
2363 } 2624 }
2364 } 2625 }
2365 clear_bit(BTRFS_FS_FROZEN, &fs_info-> 2626 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2366 2627
2367 /* 2628 /*
2368 * We still return 0, to allow VFS la 2629 * We still return 0, to allow VFS layer to unfreeze the fs even the
2369 * above checks failed. Since the fs 2630 * above checks failed. Since the fs is either fine or read-only, we're
2370 * safe to continue, without causing 2631 * safe to continue, without causing further damage.
2371 */ 2632 */
2372 return 0; 2633 return 0;
2373 } 2634 }
2374 2635
2375 static int btrfs_show_devname(struct seq_file 2636 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2376 { 2637 {
2377 struct btrfs_fs_info *fs_info = btrfs 2638 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2378 2639
2379 /* 2640 /*
2380 * There should be always a valid poi 2641 * There should be always a valid pointer in latest_dev, it may be stale
2381 * for a short moment in case it's be 2642 * for a short moment in case it's being deleted but still valid until
2382 * the end of RCU grace period. 2643 * the end of RCU grace period.
2383 */ 2644 */
2384 rcu_read_lock(); 2645 rcu_read_lock();
2385 seq_escape(m, btrfs_dev_name(fs_info- !! 2646 seq_escape(m, rcu_str_deref(fs_info->fs_devices->latest_dev->name), " \t\n\\");
2386 rcu_read_unlock(); 2647 rcu_read_unlock();
2387 2648
2388 return 0; 2649 return 0;
2389 } 2650 }
2390 2651
2391 static long btrfs_nr_cached_objects(struct su <<
2392 { <<
2393 struct btrfs_fs_info *fs_info = btrfs <<
2394 const s64 nr = percpu_counter_sum_pos <<
2395 <<
2396 trace_btrfs_extent_map_shrinker_count <<
2397 <<
2398 /* <<
2399 * Only report the real number for DE <<
2400 * serious performance degradation ca <<
2401 */ <<
2402 if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) <<
2403 return nr; <<
2404 return 0; <<
2405 } <<
2406 <<
2407 static long btrfs_free_cached_objects(struct <<
2408 { <<
2409 const long nr_to_scan = min_t(unsigne <<
2410 struct btrfs_fs_info *fs_info = btrfs <<
2411 <<
2412 /* <<
2413 * We may be called from any task try <<
2414 * want to slow it down with scanning <<
2415 * also cause heavy lock contention i <<
2416 * here. Therefore only allow kswapd <<
2417 */ <<
2418 if (!current_is_kswapd()) <<
2419 return 0; <<
2420 <<
2421 return btrfs_free_extent_maps(fs_info <<
2422 } <<
2423 <<
2424 static const struct super_operations btrfs_su 2652 static const struct super_operations btrfs_super_ops = {
2425 .drop_inode = btrfs_drop_inode, 2653 .drop_inode = btrfs_drop_inode,
2426 .evict_inode = btrfs_evict_inode, 2654 .evict_inode = btrfs_evict_inode,
2427 .put_super = btrfs_put_super, 2655 .put_super = btrfs_put_super,
2428 .sync_fs = btrfs_sync_fs, 2656 .sync_fs = btrfs_sync_fs,
2429 .show_options = btrfs_show_options, 2657 .show_options = btrfs_show_options,
2430 .show_devname = btrfs_show_devname, 2658 .show_devname = btrfs_show_devname,
2431 .alloc_inode = btrfs_alloc_inode, 2659 .alloc_inode = btrfs_alloc_inode,
2432 .destroy_inode = btrfs_destroy_inode 2660 .destroy_inode = btrfs_destroy_inode,
2433 .free_inode = btrfs_free_inode, 2661 .free_inode = btrfs_free_inode,
2434 .statfs = btrfs_statfs, 2662 .statfs = btrfs_statfs,
>> 2663 .remount_fs = btrfs_remount,
2435 .freeze_fs = btrfs_freeze, 2664 .freeze_fs = btrfs_freeze,
2436 .unfreeze_fs = btrfs_unfreeze, 2665 .unfreeze_fs = btrfs_unfreeze,
2437 .nr_cached_objects = btrfs_nr_cached_ <<
2438 .free_cached_objects = btrfs_free_cac <<
2439 }; 2666 };
2440 2667
2441 static const struct file_operations btrfs_ctl 2668 static const struct file_operations btrfs_ctl_fops = {
2442 .open = btrfs_control_open, 2669 .open = btrfs_control_open,
2443 .unlocked_ioctl = btrfs_control_ioct 2670 .unlocked_ioctl = btrfs_control_ioctl,
2444 .compat_ioctl = compat_ptr_ioctl, 2671 .compat_ioctl = compat_ptr_ioctl,
2445 .owner = THIS_MODULE, 2672 .owner = THIS_MODULE,
2446 .llseek = noop_llseek, 2673 .llseek = noop_llseek,
2447 }; 2674 };
2448 2675
2449 static struct miscdevice btrfs_misc = { 2676 static struct miscdevice btrfs_misc = {
2450 .minor = BTRFS_MINOR, 2677 .minor = BTRFS_MINOR,
2451 .name = "btrfs-control", 2678 .name = "btrfs-control",
2452 .fops = &btrfs_ctl_fops 2679 .fops = &btrfs_ctl_fops
2453 }; 2680 };
2454 2681
2455 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 2682 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2456 MODULE_ALIAS("devname:btrfs-control"); 2683 MODULE_ALIAS("devname:btrfs-control");
2457 2684
2458 static int __init btrfs_interface_init(void) 2685 static int __init btrfs_interface_init(void)
2459 { 2686 {
2460 return misc_register(&btrfs_misc); 2687 return misc_register(&btrfs_misc);
2461 } 2688 }
2462 2689
2463 static __cold void btrfs_interface_exit(void) 2690 static __cold void btrfs_interface_exit(void)
2464 { 2691 {
2465 misc_deregister(&btrfs_misc); 2692 misc_deregister(&btrfs_misc);
2466 } 2693 }
2467 2694
2468 static int __init btrfs_print_mod_info(void) !! 2695 static void __init btrfs_print_mod_info(void)
2469 { 2696 {
2470 static const char options[] = "" 2697 static const char options[] = ""
2471 #ifdef CONFIG_BTRFS_DEBUG 2698 #ifdef CONFIG_BTRFS_DEBUG
2472 ", debug=on" 2699 ", debug=on"
2473 #endif 2700 #endif
2474 #ifdef CONFIG_BTRFS_ASSERT 2701 #ifdef CONFIG_BTRFS_ASSERT
2475 ", assert=on" 2702 ", assert=on"
2476 #endif 2703 #endif
>> 2704 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 2705 ", integrity-checker=on"
>> 2706 #endif
2477 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 2707 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2478 ", ref-verify=on" 2708 ", ref-verify=on"
2479 #endif 2709 #endif
2480 #ifdef CONFIG_BLK_DEV_ZONED 2710 #ifdef CONFIG_BLK_DEV_ZONED
2481 ", zoned=yes" 2711 ", zoned=yes"
2482 #else 2712 #else
2483 ", zoned=no" 2713 ", zoned=no"
2484 #endif 2714 #endif
2485 #ifdef CONFIG_FS_VERITY 2715 #ifdef CONFIG_FS_VERITY
2486 ", fsverity=yes" 2716 ", fsverity=yes"
2487 #else 2717 #else
2488 ", fsverity=no" 2718 ", fsverity=no"
2489 #endif 2719 #endif
2490 ; 2720 ;
2491 pr_info("Btrfs loaded%s\n", options); !! 2721 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2492 return 0; <<
2493 } 2722 }
2494 2723
2495 static int register_btrfs(void) !! 2724 static int __init init_btrfs_fs(void)
2496 { 2725 {
2497 return register_filesystem(&btrfs_fs_ !! 2726 int err;
2498 } <<
2499 2727
2500 static void unregister_btrfs(void) !! 2728 btrfs_props_init();
2501 { <<
2502 unregister_filesystem(&btrfs_fs_type) <<
2503 } <<
2504 2729
2505 /* Helper structure for long init/exit functi !! 2730 err = btrfs_init_sysfs();
2506 struct init_sequence { !! 2731 if (err)
2507 int (*init_func)(void); !! 2732 return err;
2508 /* Can be NULL if the init_func doesn <<
2509 void (*exit_func)(void); <<
2510 }; <<
2511 2733
2512 static const struct init_sequence mod_init_se !! 2734 btrfs_init_compress();
2513 { <<
2514 .init_func = btrfs_props_init <<
2515 .exit_func = NULL, <<
2516 }, { <<
2517 .init_func = btrfs_init_sysfs <<
2518 .exit_func = btrfs_exit_sysfs <<
2519 }, { <<
2520 .init_func = btrfs_init_compr <<
2521 .exit_func = btrfs_exit_compr <<
2522 }, { <<
2523 .init_func = btrfs_init_cache <<
2524 .exit_func = btrfs_destroy_ca <<
2525 }, { <<
2526 .init_func = btrfs_init_dio, <<
2527 .exit_func = btrfs_destroy_di <<
2528 }, { <<
2529 .init_func = btrfs_transactio <<
2530 .exit_func = btrfs_transactio <<
2531 }, { <<
2532 .init_func = btrfs_ctree_init <<
2533 .exit_func = btrfs_ctree_exit <<
2534 }, { <<
2535 .init_func = btrfs_free_space <<
2536 .exit_func = btrfs_free_space <<
2537 }, { <<
2538 .init_func = extent_state_ini <<
2539 .exit_func = extent_state_fre <<
2540 }, { <<
2541 .init_func = extent_buffer_in <<
2542 .exit_func = extent_buffer_fr <<
2543 }, { <<
2544 .init_func = btrfs_bioset_ini <<
2545 .exit_func = btrfs_bioset_exi <<
2546 }, { <<
2547 .init_func = extent_map_init, <<
2548 .exit_func = extent_map_exit, <<
2549 }, { <<
2550 .init_func = ordered_data_ini <<
2551 .exit_func = ordered_data_exi <<
2552 }, { <<
2553 .init_func = btrfs_delayed_in <<
2554 .exit_func = btrfs_delayed_in <<
2555 }, { <<
2556 .init_func = btrfs_auto_defra <<
2557 .exit_func = btrfs_auto_defra <<
2558 }, { <<
2559 .init_func = btrfs_delayed_re <<
2560 .exit_func = btrfs_delayed_re <<
2561 }, { <<
2562 .init_func = btrfs_prelim_ref <<
2563 .exit_func = btrfs_prelim_ref <<
2564 }, { <<
2565 .init_func = btrfs_interface_ <<
2566 .exit_func = btrfs_interface_ <<
2567 }, { <<
2568 .init_func = btrfs_print_mod_ <<
2569 .exit_func = NULL, <<
2570 }, { <<
2571 .init_func = btrfs_run_sanity <<
2572 .exit_func = NULL, <<
2573 }, { <<
2574 .init_func = register_btrfs, <<
2575 .exit_func = unregister_btrfs <<
2576 } <<
2577 }; <<
2578 2735
2579 static bool mod_init_result[ARRAY_SIZE(mod_in !! 2736 err = btrfs_init_cachep();
>> 2737 if (err)
>> 2738 goto free_compress;
2580 2739
2581 static __always_inline void btrfs_exit_btrfs_ !! 2740 err = extent_io_init();
2582 { !! 2741 if (err)
2583 int i; !! 2742 goto free_cachep;
2584 2743
2585 for (i = ARRAY_SIZE(mod_init_seq) - 1 !! 2744 err = extent_state_cache_init();
2586 if (!mod_init_result[i]) !! 2745 if (err)
2587 continue; !! 2746 goto free_extent_io;
2588 if (mod_init_seq[i].exit_func !! 2747
2589 mod_init_seq[i].exit_ !! 2748 err = extent_map_init();
2590 mod_init_result[i] = false; !! 2749 if (err)
2591 } !! 2750 goto free_extent_state_cache;
>> 2751
>> 2752 err = ordered_data_init();
>> 2753 if (err)
>> 2754 goto free_extent_map;
>> 2755
>> 2756 err = btrfs_delayed_inode_init();
>> 2757 if (err)
>> 2758 goto free_ordered_data;
>> 2759
>> 2760 err = btrfs_auto_defrag_init();
>> 2761 if (err)
>> 2762 goto free_delayed_inode;
>> 2763
>> 2764 err = btrfs_delayed_ref_init();
>> 2765 if (err)
>> 2766 goto free_auto_defrag;
>> 2767
>> 2768 err = btrfs_prelim_ref_init();
>> 2769 if (err)
>> 2770 goto free_delayed_ref;
>> 2771
>> 2772 err = btrfs_interface_init();
>> 2773 if (err)
>> 2774 goto free_prelim_ref;
>> 2775
>> 2776 btrfs_print_mod_info();
>> 2777
>> 2778 err = btrfs_run_sanity_tests();
>> 2779 if (err)
>> 2780 goto unregister_ioctl;
>> 2781
>> 2782 err = register_filesystem(&btrfs_fs_type);
>> 2783 if (err)
>> 2784 goto unregister_ioctl;
>> 2785
>> 2786 return 0;
>> 2787
>> 2788 unregister_ioctl:
>> 2789 btrfs_interface_exit();
>> 2790 free_prelim_ref:
>> 2791 btrfs_prelim_ref_exit();
>> 2792 free_delayed_ref:
>> 2793 btrfs_delayed_ref_exit();
>> 2794 free_auto_defrag:
>> 2795 btrfs_auto_defrag_exit();
>> 2796 free_delayed_inode:
>> 2797 btrfs_delayed_inode_exit();
>> 2798 free_ordered_data:
>> 2799 ordered_data_exit();
>> 2800 free_extent_map:
>> 2801 extent_map_exit();
>> 2802 free_extent_state_cache:
>> 2803 extent_state_cache_exit();
>> 2804 free_extent_io:
>> 2805 extent_io_exit();
>> 2806 free_cachep:
>> 2807 btrfs_destroy_cachep();
>> 2808 free_compress:
>> 2809 btrfs_exit_compress();
>> 2810 btrfs_exit_sysfs();
>> 2811
>> 2812 return err;
2592 } 2813 }
2593 2814
2594 static void __exit exit_btrfs_fs(void) 2815 static void __exit exit_btrfs_fs(void)
2595 { 2816 {
2596 btrfs_exit_btrfs_fs(); !! 2817 btrfs_destroy_cachep();
>> 2818 btrfs_delayed_ref_exit();
>> 2819 btrfs_auto_defrag_exit();
>> 2820 btrfs_delayed_inode_exit();
>> 2821 btrfs_prelim_ref_exit();
>> 2822 ordered_data_exit();
>> 2823 extent_map_exit();
>> 2824 extent_state_cache_exit();
>> 2825 extent_io_exit();
>> 2826 btrfs_interface_exit();
>> 2827 unregister_filesystem(&btrfs_fs_type);
>> 2828 btrfs_exit_sysfs();
2597 btrfs_cleanup_fs_uuids(); 2829 btrfs_cleanup_fs_uuids();
2598 } !! 2830 btrfs_exit_compress();
2599 <<
2600 static int __init init_btrfs_fs(void) <<
2601 { <<
2602 int ret; <<
2603 int i; <<
2604 <<
2605 for (i = 0; i < ARRAY_SIZE(mod_init_s <<
2606 ASSERT(!mod_init_result[i]); <<
2607 ret = mod_init_seq[i].init_fu <<
2608 if (ret < 0) { <<
2609 btrfs_exit_btrfs_fs() <<
2610 return ret; <<
2611 } <<
2612 mod_init_result[i] = true; <<
2613 } <<
2614 return 0; <<
2615 } 2831 }
2616 2832
2617 late_initcall(init_btrfs_fs); 2833 late_initcall(init_btrfs_fs);
2618 module_exit(exit_btrfs_fs) 2834 module_exit(exit_btrfs_fs)
2619 2835
2620 MODULE_DESCRIPTION("B-Tree File System (BTRFS <<
2621 MODULE_LICENSE("GPL"); 2836 MODULE_LICENSE("GPL");
2622 MODULE_SOFTDEP("pre: crc32c"); 2837 MODULE_SOFTDEP("pre: crc32c");
2623 MODULE_SOFTDEP("pre: xxhash64"); 2838 MODULE_SOFTDEP("pre: xxhash64");
2624 MODULE_SOFTDEP("pre: sha256"); 2839 MODULE_SOFTDEP("pre: sha256");
2625 MODULE_SOFTDEP("pre: blake2b-256"); 2840 MODULE_SOFTDEP("pre: blake2b-256");
2626 2841
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.