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/cleancache.h>
26 #include <linux/ratelimit.h> 27 #include <linux/ratelimit.h>
27 #include <linux/crc32c.h> 28 #include <linux/crc32c.h>
28 #include <linux/btrfs.h> 29 #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" 30 #include "delayed-inode.h"
34 #include "ctree.h" 31 #include "ctree.h"
35 #include "disk-io.h" 32 #include "disk-io.h"
36 #include "transaction.h" 33 #include "transaction.h"
37 #include "btrfs_inode.h" 34 #include "btrfs_inode.h"
38 #include "direct-io.h" !! 35 #include "print-tree.h"
39 #include "props.h" 36 #include "props.h"
40 #include "xattr.h" 37 #include "xattr.h"
41 #include "bio.h" !! 38 #include "volumes.h"
42 #include "export.h" 39 #include "export.h"
43 #include "compression.h" 40 #include "compression.h"
>> 41 #include "rcu-string.h"
44 #include "dev-replace.h" 42 #include "dev-replace.h"
45 #include "free-space-cache.h" 43 #include "free-space-cache.h"
46 #include "backref.h" 44 #include "backref.h"
47 #include "space-info.h" 45 #include "space-info.h"
48 #include "sysfs.h" 46 #include "sysfs.h"
49 #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"
>> 50
53 #include "qgroup.h" 51 #include "qgroup.h"
54 #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 const char * __attribute_const__ btrfs_decode_error(int errno)
71 { 71 {
72 struct btrfs_fs_info *fs_info = btrfs_ !! 72 char *errstr = "unknown";
73 73
74 btrfs_info(fs_info, "last unmount of f !! 74 switch (errno) {
75 close_ctree(fs_info); !! 75 case -EIO:
>> 76 errstr = "IO failure";
>> 77 break;
>> 78 case -ENOMEM:
>> 79 errstr = "Out of memory";
>> 80 break;
>> 81 case -EROFS:
>> 82 errstr = "Readonly filesystem";
>> 83 break;
>> 84 case -EEXIST:
>> 85 errstr = "Object already exists";
>> 86 break;
>> 87 case -ENOSPC:
>> 88 errstr = "No space left";
>> 89 break;
>> 90 case -ENOENT:
>> 91 errstr = "No such entry";
>> 92 break;
>> 93 }
>> 94
>> 95 return errstr;
>> 96 }
>> 97
>> 98 /*
>> 99 * __btrfs_handle_fs_error decodes expected errors from the caller and
>> 100 * invokes the appropriate error response.
>> 101 */
>> 102 __cold
>> 103 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
>> 104 unsigned int line, int errno, const char *fmt, ...)
>> 105 {
>> 106 struct super_block *sb = fs_info->sb;
>> 107 #ifdef CONFIG_PRINTK
>> 108 const char *errstr;
>> 109 #endif
>> 110
>> 111 /*
>> 112 * Special case: if the error is EROFS, and we're already
>> 113 * under SB_RDONLY, then it is safe here.
>> 114 */
>> 115 if (errno == -EROFS && sb_rdonly(sb))
>> 116 return;
>> 117
>> 118 #ifdef CONFIG_PRINTK
>> 119 errstr = btrfs_decode_error(errno);
>> 120 if (fmt) {
>> 121 struct va_format vaf;
>> 122 va_list args;
>> 123
>> 124 va_start(args, fmt);
>> 125 vaf.fmt = fmt;
>> 126 vaf.va = &args;
>> 127
>> 128 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
>> 129 sb->s_id, function, line, errno, errstr, &vaf);
>> 130 va_end(args);
>> 131 } else {
>> 132 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
>> 133 sb->s_id, function, line, errno, errstr);
>> 134 }
>> 135 #endif
>> 136
>> 137 /*
>> 138 * Today we only save the error info to memory. Long term we'll
>> 139 * also send it down to the disk
>> 140 */
>> 141 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
>> 142
>> 143 /* Don't go through full error handling during mount */
>> 144 if (!(sb->s_flags & SB_BORN))
>> 145 return;
>> 146
>> 147 if (sb_rdonly(sb))
>> 148 return;
>> 149
>> 150 btrfs_discard_stop(fs_info);
>> 151
>> 152 /* btrfs handle error by forcing the filesystem readonly */
>> 153 sb->s_flags |= SB_RDONLY;
>> 154 btrfs_info(fs_info, "forced readonly");
>> 155 /*
>> 156 * Note that a running device replace operation is not canceled here
>> 157 * although there is no way to update the progress. It would add the
>> 158 * risk of a deadlock, therefore the canceling is omitted. The only
>> 159 * penalty is that some I/O remains active until the procedure
>> 160 * completes. The next time when the filesystem is mounted writable
>> 161 * again, the device replace operation continues.
>> 162 */
76 } 163 }
77 164
78 /* Store the mount options related information !! 165 #ifdef CONFIG_PRINTK
79 struct btrfs_fs_context { !! 166 static const char * const logtypes[] = {
80 char *subvol_name; !! 167 "emergency",
81 u64 subvol_objectid; !! 168 "alert",
82 u64 max_inline; !! 169 "critical",
83 u32 commit_interval; !! 170 "error",
84 u32 metadata_ratio; !! 171 "warning",
85 u32 thread_pool_size; !! 172 "notice",
86 unsigned long long mount_opt; !! 173 "info",
87 unsigned long compress_type:4; !! 174 "debug",
88 unsigned int compress_level; !! 175 };
89 refcount_t refs; !! 176
>> 177
>> 178 /*
>> 179 * Use one ratelimit state per log level so that a flood of less important
>> 180 * messages doesn't cause more important ones to be dropped.
>> 181 */
>> 182 static struct ratelimit_state printk_limits[] = {
>> 183 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 184 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 185 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 186 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 187 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 188 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 189 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 190 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
90 }; 191 };
91 192
>> 193 void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
>> 194 {
>> 195 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
>> 196 struct va_format vaf;
>> 197 va_list args;
>> 198 int kern_level;
>> 199 const char *type = logtypes[4];
>> 200 struct ratelimit_state *ratelimit = &printk_limits[4];
>> 201
>> 202 va_start(args, fmt);
>> 203
>> 204 while ((kern_level = printk_get_level(fmt)) != 0) {
>> 205 size_t size = printk_skip_level(fmt) - fmt;
>> 206
>> 207 if (kern_level >= '' && kern_level <= '7') {
>> 208 memcpy(lvl, fmt, size);
>> 209 lvl[size] = '\0';
>> 210 type = logtypes[kern_level - ''];
>> 211 ratelimit = &printk_limits[kern_level - ''];
>> 212 }
>> 213 fmt += size;
>> 214 }
>> 215
>> 216 vaf.fmt = fmt;
>> 217 vaf.va = &args;
>> 218
>> 219 if (__ratelimit(ratelimit))
>> 220 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
>> 221 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
>> 222
>> 223 va_end(args);
>> 224 }
>> 225 #endif
>> 226
>> 227 /*
>> 228 * We only mark the transaction aborted and then set the file system read-only.
>> 229 * This will prevent new transactions from starting or trying to join this
>> 230 * one.
>> 231 *
>> 232 * This means that error recovery at the call site is limited to freeing
>> 233 * any local memory allocations and passing the error code up without
>> 234 * further cleanup. The transaction should complete as it normally would
>> 235 * in the call path but will return -EIO.
>> 236 *
>> 237 * We'll complete the cleanup in btrfs_end_transaction and
>> 238 * btrfs_commit_transaction.
>> 239 */
>> 240 __cold
>> 241 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
>> 242 const char *function,
>> 243 unsigned int line, int errno)
>> 244 {
>> 245 struct btrfs_fs_info *fs_info = trans->fs_info;
>> 246
>> 247 trans->aborted = errno;
>> 248 /* Nothing used. The other threads that have joined this
>> 249 * transaction may be able to continue. */
>> 250 if (!trans->dirty && list_empty(&trans->new_bgs)) {
>> 251 const char *errstr;
>> 252
>> 253 errstr = btrfs_decode_error(errno);
>> 254 btrfs_warn(fs_info,
>> 255 "%s:%d: Aborting unused transaction(%s).",
>> 256 function, line, errstr);
>> 257 return;
>> 258 }
>> 259 WRITE_ONCE(trans->transaction->aborted, errno);
>> 260 /* Wake up anybody who may be waiting on this transaction */
>> 261 wake_up(&fs_info->transaction_wait);
>> 262 wake_up(&fs_info->transaction_blocked_wait);
>> 263 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
>> 264 }
>> 265 /*
>> 266 * __btrfs_panic decodes unexpected, fatal errors from the caller,
>> 267 * issues an alert, and either panics or BUGs, depending on mount options.
>> 268 */
>> 269 __cold
>> 270 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
>> 271 unsigned int line, int errno, const char *fmt, ...)
>> 272 {
>> 273 char *s_id = "<unknown>";
>> 274 const char *errstr;
>> 275 struct va_format vaf = { .fmt = fmt };
>> 276 va_list args;
>> 277
>> 278 if (fs_info)
>> 279 s_id = fs_info->sb->s_id;
>> 280
>> 281 va_start(args, fmt);
>> 282 vaf.va = &args;
>> 283
>> 284 errstr = btrfs_decode_error(errno);
>> 285 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
>> 286 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
>> 287 s_id, function, line, &vaf, errno, errstr);
>> 288
>> 289 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
>> 290 function, line, &vaf, errno, errstr);
>> 291 va_end(args);
>> 292 /* Caller calls BUG() */
>> 293 }
>> 294
>> 295 static void btrfs_put_super(struct super_block *sb)
>> 296 {
>> 297 close_ctree(btrfs_sb(sb));
>> 298 }
>> 299
92 enum { 300 enum {
93 Opt_acl, !! 301 Opt_acl, Opt_noacl,
94 Opt_clear_cache, 302 Opt_clear_cache,
95 Opt_commit_interval, 303 Opt_commit_interval,
96 Opt_compress, 304 Opt_compress,
97 Opt_compress_force, 305 Opt_compress_force,
98 Opt_compress_force_type, 306 Opt_compress_force_type,
99 Opt_compress_type, 307 Opt_compress_type,
100 Opt_degraded, 308 Opt_degraded,
101 Opt_device, 309 Opt_device,
102 Opt_fatal_errors, 310 Opt_fatal_errors,
103 Opt_flushoncommit, !! 311 Opt_flushoncommit, Opt_noflushoncommit,
>> 312 Opt_inode_cache, Opt_noinode_cache,
104 Opt_max_inline, 313 Opt_max_inline,
105 Opt_barrier, !! 314 Opt_barrier, Opt_nobarrier,
106 Opt_datacow, !! 315 Opt_datacow, Opt_nodatacow,
107 Opt_datasum, !! 316 Opt_datasum, Opt_nodatasum,
108 Opt_defrag, !! 317 Opt_defrag, Opt_nodefrag,
109 Opt_discard, !! 318 Opt_discard, Opt_nodiscard,
110 Opt_discard_mode, 319 Opt_discard_mode,
>> 320 Opt_nologreplay,
>> 321 Opt_norecovery,
111 Opt_ratio, 322 Opt_ratio,
112 Opt_rescan_uuid_tree, 323 Opt_rescan_uuid_tree,
113 Opt_skip_balance, 324 Opt_skip_balance,
114 Opt_space_cache, !! 325 Opt_space_cache, Opt_no_space_cache,
115 Opt_space_cache_version, 326 Opt_space_cache_version,
116 Opt_ssd, !! 327 Opt_ssd, Opt_nossd,
117 Opt_ssd_spread, !! 328 Opt_ssd_spread, Opt_nossd_spread,
118 Opt_subvol, 329 Opt_subvol,
119 Opt_subvol_empty, 330 Opt_subvol_empty,
120 Opt_subvolid, 331 Opt_subvolid,
121 Opt_thread_pool, 332 Opt_thread_pool,
122 Opt_treelog, !! 333 Opt_treelog, Opt_notreelog,
>> 334 Opt_usebackuproot,
123 Opt_user_subvol_rm_allowed, 335 Opt_user_subvol_rm_allowed,
124 Opt_norecovery, <<
125 336
126 /* Rescue options */ !! 337 /* Deprecated options */
127 Opt_rescue, !! 338 Opt_alloc_start,
128 Opt_usebackuproot, !! 339 Opt_recovery,
129 Opt_nologreplay, !! 340 Opt_subvolrootid,
130 341
131 /* Debugging options */ 342 /* Debugging options */
132 Opt_enospc_debug, !! 343 Opt_check_integrity,
>> 344 Opt_check_integrity_including_extent_data,
>> 345 Opt_check_integrity_print_mask,
>> 346 Opt_enospc_debug, Opt_noenospc_debug,
133 #ifdef CONFIG_BTRFS_DEBUG 347 #ifdef CONFIG_BTRFS_DEBUG
134 Opt_fragment, Opt_fragment_data, Opt_f !! 348 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
135 #endif 349 #endif
136 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 350 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
137 Opt_ref_verify, 351 Opt_ref_verify,
138 #endif 352 #endif
139 Opt_err, 353 Opt_err,
140 }; 354 };
141 355
142 enum { !! 356 static const match_table_t tokens = {
143 Opt_fatal_errors_panic, !! 357 {Opt_acl, "acl"},
144 Opt_fatal_errors_bug, !! 358 {Opt_noacl, "noacl"},
145 }; !! 359 {Opt_clear_cache, "clear_cache"},
146 !! 360 {Opt_commit_interval, "commit=%u"},
147 static const struct constant_table btrfs_param !! 361 {Opt_compress, "compress"},
148 { "panic", Opt_fatal_errors_panic }, !! 362 {Opt_compress_type, "compress=%s"},
149 { "bug", Opt_fatal_errors_bug }, !! 363 {Opt_compress_force, "compress-force"},
150 {} !! 364 {Opt_compress_force_type, "compress-force=%s"},
151 }; !! 365 {Opt_degraded, "degraded"},
152 !! 366 {Opt_device, "device=%s"},
153 enum { !! 367 {Opt_fatal_errors, "fatal_errors=%s"},
154 Opt_discard_sync, !! 368 {Opt_flushoncommit, "flushoncommit"},
155 Opt_discard_async, !! 369 {Opt_noflushoncommit, "noflushoncommit"},
156 }; !! 370 {Opt_inode_cache, "inode_cache"},
157 !! 371 {Opt_noinode_cache, "noinode_cache"},
158 static const struct constant_table btrfs_param !! 372 {Opt_max_inline, "max_inline=%s"},
159 { "sync", Opt_discard_sync }, !! 373 {Opt_barrier, "barrier"},
160 { "async", Opt_discard_async }, !! 374 {Opt_nobarrier, "nobarrier"},
161 {} !! 375 {Opt_datacow, "datacow"},
162 }; !! 376 {Opt_nodatacow, "nodatacow"},
163 !! 377 {Opt_datasum, "datasum"},
164 enum { !! 378 {Opt_nodatasum, "nodatasum"},
165 Opt_space_cache_v1, !! 379 {Opt_defrag, "autodefrag"},
166 Opt_space_cache_v2, !! 380 {Opt_nodefrag, "noautodefrag"},
167 }; !! 381 {Opt_discard, "discard"},
168 !! 382 {Opt_discard_mode, "discard=%s"},
169 static const struct constant_table btrfs_param !! 383 {Opt_nodiscard, "nodiscard"},
170 { "v1", Opt_space_cache_v1 }, !! 384 {Opt_nologreplay, "nologreplay"},
171 { "v2", Opt_space_cache_v2 }, !! 385 {Opt_norecovery, "norecovery"},
172 {} !! 386 {Opt_ratio, "metadata_ratio=%u"},
173 }; !! 387 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
174 !! 388 {Opt_skip_balance, "skip_balance"},
175 enum { !! 389 {Opt_space_cache, "space_cache"},
176 Opt_rescue_usebackuproot, !! 390 {Opt_no_space_cache, "nospace_cache"},
177 Opt_rescue_nologreplay, !! 391 {Opt_space_cache_version, "space_cache=%s"},
178 Opt_rescue_ignorebadroots, !! 392 {Opt_ssd, "ssd"},
179 Opt_rescue_ignoredatacsums, !! 393 {Opt_nossd, "nossd"},
180 Opt_rescue_ignoremetacsums, !! 394 {Opt_ssd_spread, "ssd_spread"},
181 Opt_rescue_ignoresuperflags, !! 395 {Opt_nossd_spread, "nossd_spread"},
182 Opt_rescue_parameter_all, !! 396 {Opt_subvol, "subvol=%s"},
183 }; !! 397 {Opt_subvol_empty, "subvol="},
184 !! 398 {Opt_subvolid, "subvolid=%s"},
185 static const struct constant_table btrfs_param !! 399 {Opt_thread_pool, "thread_pool=%u"},
186 { "usebackuproot", Opt_rescue_usebacku !! 400 {Opt_treelog, "treelog"},
187 { "nologreplay", Opt_rescue_nologrepla !! 401 {Opt_notreelog, "notreelog"},
188 { "ignorebadroots", Opt_rescue_ignoreb !! 402 {Opt_usebackuproot, "usebackuproot"},
189 { "ibadroots", Opt_rescue_ignorebadroo !! 403 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
190 { "ignoredatacsums", Opt_rescue_ignore !! 404
191 { "ignoremetacsums", Opt_rescue_ignore !! 405 /* Deprecated options */
192 { "ignoresuperflags", Opt_rescue_ignor !! 406 {Opt_alloc_start, "alloc_start=%s"},
193 { "idatacsums", Opt_rescue_ignoredatac !! 407 {Opt_recovery, "recovery"},
194 { "imetacsums", Opt_rescue_ignoremetac !! 408 {Opt_subvolrootid, "subvolrootid=%d"},
195 { "isuperflags", Opt_rescue_ignoresupe <<
196 { "all", Opt_rescue_parameter_all }, <<
197 {} <<
198 }; <<
199 409
>> 410 /* Debugging options */
>> 411 {Opt_check_integrity, "check_int"},
>> 412 {Opt_check_integrity_including_extent_data, "check_int_data"},
>> 413 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
>> 414 {Opt_enospc_debug, "enospc_debug"},
>> 415 {Opt_noenospc_debug, "noenospc_debug"},
200 #ifdef CONFIG_BTRFS_DEBUG 416 #ifdef CONFIG_BTRFS_DEBUG
201 enum { !! 417 {Opt_fragment_data, "fragment=data"},
202 Opt_fragment_parameter_data, !! 418 {Opt_fragment_metadata, "fragment=metadata"},
203 Opt_fragment_parameter_metadata, !! 419 {Opt_fragment_all, "fragment=all"},
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 <<
251 __fsparam(NULL, "nologreplay", Opt_nol <<
252 /* Deprecated, with alias rescue=useba <<
253 __fsparam(NULL, "usebackuproot", Opt_u <<
254 /* For compatibility only, alias for " <<
255 fsparam_flag("norecovery", Opt_norecov <<
256 <<
257 /* Debugging options. */ <<
258 fsparam_flag_no("enospc_debug", Opt_en <<
259 #ifdef CONFIG_BTRFS_DEBUG <<
260 fsparam_enum("fragment", Opt_fragment, <<
261 #endif 420 #endif
262 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 421 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
263 fsparam_flag("ref_verify", Opt_ref_ver !! 422 {Opt_ref_verify, "ref_verify"},
264 #endif 423 #endif
265 {} !! 424 {Opt_err, NULL},
266 }; 425 };
267 426
268 /* No support for restricting writes to btrfs !! 427 /*
269 static inline blk_mode_t btrfs_open_mode(struc !! 428 * Regular mount options parser. Everything that is needed only when
270 { !! 429 * reading in a new superblock is parsed here.
271 return sb_open_mode(fc->sb_flags) & ~B !! 430 * XXX JDM: This needs to be cleaned up for remount.
272 } !! 431 */
273 !! 432 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
274 static int btrfs_parse_param(struct fs_context !! 433 unsigned long new_flags)
275 { 434 {
276 struct btrfs_fs_context *ctx = fc->fs_ !! 435 substring_t args[MAX_OPT_ARGS];
277 struct fs_parse_result result; !! 436 char *p, *num;
278 int opt; !! 437 u64 cache_gen;
279 !! 438 int intarg;
280 opt = fs_parse(fc, btrfs_fs_parameters !! 439 int ret = 0;
281 if (opt < 0) !! 440 char *compress_type;
282 return opt; !! 441 bool compress_force = false;
283 !! 442 enum btrfs_compression_type saved_compress_type;
284 switch (opt) { !! 443 bool saved_compress_force;
285 case Opt_degraded: !! 444 int no_compress = 0;
286 btrfs_set_opt(ctx->mount_opt, !! 445
287 break; !! 446 cache_gen = btrfs_super_cache_generation(info->super_copy);
288 case Opt_subvol_empty: !! 447 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
289 /* !! 448 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
290 * This exists because we used !! 449 else if (cache_gen)
291 * keeping it to maintain ABI. !! 450 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
292 * empty subvol= again"). !! 451
293 */ !! 452 /*
294 break; !! 453 * Even the options are empty, we still need to do extra check
295 case Opt_subvol: !! 454 * against new flags
296 kfree(ctx->subvol_name); !! 455 */
297 ctx->subvol_name = kstrdup(par !! 456 if (!options)
298 if (!ctx->subvol_name) !! 457 goto check;
299 return -ENOMEM; !! 458
300 break; !! 459 while ((p = strsep(&options, ",")) != NULL) {
301 case Opt_subvolid: !! 460 int token;
302 ctx->subvol_objectid = result. !! 461 if (!*p)
303 !! 462 continue;
304 /* subvolid=0 means give me th <<
305 if (!ctx->subvol_objectid) <<
306 ctx->subvol_objectid = <<
307 break; <<
308 case Opt_device: { <<
309 struct btrfs_device *device; <<
310 blk_mode_t mode = btrfs_open_m <<
311 463
312 mutex_lock(&uuid_mutex); !! 464 token = match_token(p, tokens, args);
313 device = btrfs_scan_one_device !! 465 switch (token) {
314 mutex_unlock(&uuid_mutex); !! 466 case Opt_degraded:
315 if (IS_ERR(device)) !! 467 btrfs_info(info, "allowing degraded mounts");
316 return PTR_ERR(device) !! 468 btrfs_set_opt(info->mount_opt, DEGRADED);
317 break; !! 469 break;
318 } !! 470 case Opt_subvol:
319 case Opt_datasum: !! 471 case Opt_subvol_empty:
320 if (result.negated) { !! 472 case Opt_subvolid:
321 btrfs_set_opt(ctx->mou !! 473 case Opt_subvolrootid:
322 } else { !! 474 case Opt_device:
323 btrfs_clear_opt(ctx->m !! 475 /*
324 btrfs_clear_opt(ctx->m !! 476 * These are parsed by btrfs_parse_subvol_options or
325 } !! 477 * btrfs_parse_device_options and can be ignored here.
326 break; !! 478 */
327 case Opt_datacow: !! 479 break;
328 if (result.negated) { !! 480 case Opt_nodatasum:
329 btrfs_clear_opt(ctx->m !! 481 btrfs_set_and_info(info, NODATASUM,
330 btrfs_clear_opt(ctx->m !! 482 "setting nodatasum");
331 btrfs_set_opt(ctx->mou !! 483 break;
332 btrfs_set_opt(ctx->mou !! 484 case Opt_datasum:
333 } else { !! 485 if (btrfs_test_opt(info, NODATASUM)) {
334 btrfs_clear_opt(ctx->m !! 486 if (btrfs_test_opt(info, NODATACOW))
335 } !! 487 btrfs_info(info,
336 break; !! 488 "setting datasum, datacow enabled");
337 case Opt_compress_force: !! 489 else
338 case Opt_compress_force_type: !! 490 btrfs_info(info, "setting datasum");
339 btrfs_set_opt(ctx->mount_opt, !! 491 }
340 fallthrough; !! 492 btrfs_clear_opt(info->mount_opt, NODATACOW);
341 case Opt_compress: !! 493 btrfs_clear_opt(info->mount_opt, NODATASUM);
342 case Opt_compress_type: !! 494 break;
343 /* !! 495 case Opt_nodatacow:
344 * Provide the same semantics !! 496 if (!btrfs_test_opt(info, NODATACOW)) {
345 * context, specifying the "co !! 497 if (!btrfs_test_opt(info, COMPRESS) ||
346 * "force-compress" without th !! 498 !btrfs_test_opt(info, FORCE_COMPRESS)) {
347 * "compress-force=[no|none]" !! 499 btrfs_info(info,
348 */ !! 500 "setting nodatacow, compression disabled");
349 if (opt != Opt_compress_force !! 501 } else {
350 btrfs_clear_opt(ctx->m !! 502 btrfs_info(info, "setting nodatacow");
>> 503 }
>> 504 }
>> 505 btrfs_clear_opt(info->mount_opt, COMPRESS);
>> 506 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
>> 507 btrfs_set_opt(info->mount_opt, NODATACOW);
>> 508 btrfs_set_opt(info->mount_opt, NODATASUM);
>> 509 break;
>> 510 case Opt_datacow:
>> 511 btrfs_clear_and_info(info, NODATACOW,
>> 512 "setting datacow");
>> 513 break;
>> 514 case Opt_compress_force:
>> 515 case Opt_compress_force_type:
>> 516 compress_force = true;
>> 517 /* Fallthrough */
>> 518 case Opt_compress:
>> 519 case Opt_compress_type:
>> 520 saved_compress_type = btrfs_test_opt(info,
>> 521 COMPRESS) ?
>> 522 info->compress_type : BTRFS_COMPRESS_NONE;
>> 523 saved_compress_force =
>> 524 btrfs_test_opt(info, FORCE_COMPRESS);
>> 525 if (token == Opt_compress ||
>> 526 token == Opt_compress_force ||
>> 527 strncmp(args[0].from, "zlib", 4) == 0) {
>> 528 compress_type = "zlib";
>> 529
>> 530 info->compress_type = BTRFS_COMPRESS_ZLIB;
>> 531 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
>> 532 /*
>> 533 * args[0] contains uninitialized data since
>> 534 * for these tokens we don't expect any
>> 535 * parameter.
>> 536 */
>> 537 if (token != Opt_compress &&
>> 538 token != Opt_compress_force)
>> 539 info->compress_level =
>> 540 btrfs_compress_str2level(
>> 541 BTRFS_COMPRESS_ZLIB,
>> 542 args[0].from + 4);
>> 543 btrfs_set_opt(info->mount_opt, COMPRESS);
>> 544 btrfs_clear_opt(info->mount_opt, NODATACOW);
>> 545 btrfs_clear_opt(info->mount_opt, NODATASUM);
>> 546 no_compress = 0;
>> 547 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
>> 548 compress_type = "lzo";
>> 549 info->compress_type = BTRFS_COMPRESS_LZO;
>> 550 btrfs_set_opt(info->mount_opt, COMPRESS);
>> 551 btrfs_clear_opt(info->mount_opt, NODATACOW);
>> 552 btrfs_clear_opt(info->mount_opt, NODATASUM);
>> 553 btrfs_set_fs_incompat(info, COMPRESS_LZO);
>> 554 no_compress = 0;
>> 555 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
>> 556 compress_type = "zstd";
>> 557 info->compress_type = BTRFS_COMPRESS_ZSTD;
>> 558 info->compress_level =
>> 559 btrfs_compress_str2level(
>> 560 BTRFS_COMPRESS_ZSTD,
>> 561 args[0].from + 4);
>> 562 btrfs_set_opt(info->mount_opt, COMPRESS);
>> 563 btrfs_clear_opt(info->mount_opt, NODATACOW);
>> 564 btrfs_clear_opt(info->mount_opt, NODATASUM);
>> 565 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
>> 566 no_compress = 0;
>> 567 } else if (strncmp(args[0].from, "no", 2) == 0) {
>> 568 compress_type = "no";
>> 569 btrfs_clear_opt(info->mount_opt, COMPRESS);
>> 570 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
>> 571 compress_force = false;
>> 572 no_compress++;
>> 573 } else {
>> 574 ret = -EINVAL;
>> 575 goto out;
>> 576 }
351 577
352 if (opt == Opt_compress || opt !! 578 if (compress_force) {
353 ctx->compress_type = B !! 579 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
354 ctx->compress_level = !! 580 } else {
355 btrfs_set_opt(ctx->mou !! 581 /*
356 btrfs_clear_opt(ctx->m !! 582 * If we remount from compress-force=xxx to
357 btrfs_clear_opt(ctx->m !! 583 * compress=xxx, we need clear FORCE_COMPRESS
358 } else if (strncmp(param->stri !! 584 * flag, otherwise, there is no way for users
359 ctx->compress_type = B !! 585 * to disable forcible compression separately.
360 ctx->compress_level = !! 586 */
361 btrfs_compress !! 587 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
362 !! 588 }
363 btrfs_set_opt(ctx->mou !! 589 if ((btrfs_test_opt(info, COMPRESS) &&
364 btrfs_clear_opt(ctx->m !! 590 (info->compress_type != saved_compress_type ||
365 btrfs_clear_opt(ctx->m !! 591 compress_force != saved_compress_force)) ||
366 } else if (strncmp(param->stri !! 592 (!btrfs_test_opt(info, COMPRESS) &&
367 ctx->compress_type = B !! 593 no_compress == 1)) {
368 ctx->compress_level = !! 594 btrfs_info(info, "%s %s compression, level %d",
369 btrfs_set_opt(ctx->mou !! 595 (compress_force) ? "force" : "use",
370 btrfs_clear_opt(ctx->m !! 596 compress_type, info->compress_level);
371 btrfs_clear_opt(ctx->m !! 597 }
372 } else if (strncmp(param->stri !! 598 compress_force = false;
373 ctx->compress_type = B !! 599 break;
374 ctx->compress_level = !! 600 case Opt_ssd:
375 btrfs_compress !! 601 btrfs_set_and_info(info, SSD,
376 !! 602 "enabling ssd optimizations");
377 btrfs_set_opt(ctx->mou !! 603 btrfs_clear_opt(info->mount_opt, NOSSD);
378 btrfs_clear_opt(ctx->m !! 604 break;
379 btrfs_clear_opt(ctx->m !! 605 case Opt_ssd_spread:
380 } else if (strncmp(param->stri !! 606 btrfs_set_and_info(info, SSD,
381 ctx->compress_level = !! 607 "enabling ssd optimizations");
382 ctx->compress_type = 0 !! 608 btrfs_set_and_info(info, SSD_SPREAD,
383 btrfs_clear_opt(ctx->m !! 609 "using spread ssd allocation scheme");
384 btrfs_clear_opt(ctx->m !! 610 btrfs_clear_opt(info->mount_opt, NOSSD);
385 } else { !! 611 break;
386 btrfs_err(NULL, "unrec !! 612 case Opt_nossd:
387 param->strin !! 613 btrfs_set_opt(info->mount_opt, NOSSD);
388 return -EINVAL; !! 614 btrfs_clear_and_info(info, SSD,
389 } !! 615 "not using ssd optimizations");
390 break; !! 616 /* Fallthrough */
391 case Opt_ssd: !! 617 case Opt_nossd_spread:
392 if (result.negated) { !! 618 btrfs_clear_and_info(info, SSD_SPREAD,
393 btrfs_set_opt(ctx->mou !! 619 "not using spread ssd allocation scheme");
394 btrfs_clear_opt(ctx->m !! 620 break;
395 btrfs_clear_opt(ctx->m !! 621 case Opt_barrier:
396 } else { !! 622 btrfs_clear_and_info(info, NOBARRIER,
397 btrfs_set_opt(ctx->mou !! 623 "turning on barriers");
398 btrfs_clear_opt(ctx->m !! 624 break;
399 } !! 625 case Opt_nobarrier:
400 break; !! 626 btrfs_set_and_info(info, NOBARRIER,
401 case Opt_ssd_spread: !! 627 "turning off barriers");
402 if (result.negated) { !! 628 break;
403 btrfs_clear_opt(ctx->m !! 629 case Opt_thread_pool:
404 } else { !! 630 ret = match_int(&args[0], &intarg);
405 btrfs_set_opt(ctx->mou !! 631 if (ret) {
406 btrfs_set_opt(ctx->mou !! 632 goto out;
407 btrfs_clear_opt(ctx->m !! 633 } else if (intarg == 0) {
408 } !! 634 ret = -EINVAL;
409 break; !! 635 goto out;
410 case Opt_barrier: !! 636 }
411 if (result.negated) !! 637 info->thread_pool_size = intarg;
412 btrfs_set_opt(ctx->mou !! 638 break;
413 else !! 639 case Opt_max_inline:
414 btrfs_clear_opt(ctx->m !! 640 num = match_strdup(&args[0]);
415 break; !! 641 if (num) {
416 case Opt_thread_pool: !! 642 info->max_inline = memparse(num, NULL);
417 if (result.uint_32 == 0) { !! 643 kfree(num);
418 btrfs_err(NULL, "inval !! 644
419 return -EINVAL; !! 645 if (info->max_inline) {
420 } !! 646 info->max_inline = min_t(u64,
421 ctx->thread_pool_size = result !! 647 info->max_inline,
422 break; !! 648 info->sectorsize);
423 case Opt_max_inline: !! 649 }
424 ctx->max_inline = memparse(par !! 650 btrfs_info(info, "max_inline at %llu",
425 break; !! 651 info->max_inline);
426 case Opt_acl: !! 652 } else {
427 if (result.negated) { !! 653 ret = -ENOMEM;
428 fc->sb_flags &= ~SB_PO !! 654 goto out;
429 } else { !! 655 }
>> 656 break;
>> 657 case Opt_alloc_start:
>> 658 btrfs_info(info,
>> 659 "option alloc_start is obsolete, ignored");
>> 660 break;
>> 661 case Opt_acl:
430 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 662 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
431 fc->sb_flags |= SB_POS !! 663 info->sb->s_flags |= SB_POSIXACL;
>> 664 break;
432 #else 665 #else
433 btrfs_err(NULL, "suppo !! 666 btrfs_err(info, "support for ACL not compiled in!");
434 return -EINVAL; !! 667 ret = -EINVAL;
>> 668 goto out;
435 #endif 669 #endif
436 } !! 670 case Opt_noacl:
437 /* !! 671 info->sb->s_flags &= ~SB_POSIXACL;
438 * VFS limits the ability to t <<
439 * despite every file system a <<
440 * an oversight since we all d <<
441 * remounting. So don't set t <<
442 * btrfs_reconfigure and do th <<
443 */ <<
444 if (fc->purpose != FS_CONTEXT_ <<
445 fc->sb_flags_mask |= S <<
446 break; <<
447 case Opt_treelog: <<
448 if (result.negated) <<
449 btrfs_set_opt(ctx->mou <<
450 else <<
451 btrfs_clear_opt(ctx->m <<
452 break; <<
453 case Opt_nologreplay: <<
454 btrfs_warn(NULL, <<
455 "'nologreplay' is deprecated, <<
456 btrfs_set_opt(ctx->mount_opt, <<
457 break; <<
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; 672 break;
492 default: !! 673 case Opt_notreelog:
493 btrfs_err(NULL, "unrec !! 674 btrfs_set_and_info(info, NOTREELOG,
494 param->key); !! 675 "disabling tree log");
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; 676 break;
519 default: !! 677 case Opt_treelog:
520 btrfs_err(NULL, "unrec !! 678 btrfs_clear_and_info(info, NOTREELOG,
521 param->key); !! 679 "enabling tree log");
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; 680 break;
562 case Opt_fatal_errors_bug: !! 681 case Opt_norecovery:
563 btrfs_clear_opt(ctx->m !! 682 case Opt_nologreplay:
>> 683 btrfs_set_and_info(info, NOLOGREPLAY,
>> 684 "disabling log replay at mount time");
564 break; 685 break;
565 default: !! 686 case Opt_flushoncommit:
566 btrfs_err(NULL, "unrec !! 687 btrfs_set_and_info(info, FLUSHONCOMMIT,
567 param->key); !! 688 "turning on flush-on-commit");
568 return -EINVAL; !! 689 break;
569 } !! 690 case Opt_noflushoncommit:
570 break; !! 691 btrfs_clear_and_info(info, FLUSHONCOMMIT,
571 case Opt_commit_interval: !! 692 "turning off flush-on-commit");
572 ctx->commit_interval = result. !! 693 break;
573 if (ctx->commit_interval == 0) !! 694 case Opt_ratio:
574 ctx->commit_interval = !! 695 ret = match_int(&args[0], &intarg);
575 break; !! 696 if (ret)
576 case Opt_rescue: !! 697 goto out;
577 switch (result.uint_32) { !! 698 info->metadata_ratio = intarg;
578 case Opt_rescue_usebackuproot: !! 699 btrfs_info(info, "metadata ratio %u",
579 btrfs_set_opt(ctx->mou !! 700 info->metadata_ratio);
580 break; !! 701 break;
581 case Opt_rescue_nologreplay: !! 702 case Opt_discard:
582 btrfs_set_opt(ctx->mou !! 703 case Opt_discard_mode:
583 break; !! 704 if (token == Opt_discard ||
584 case Opt_rescue_ignorebadroots !! 705 strcmp(args[0].from, "sync") == 0) {
585 btrfs_set_opt(ctx->mou !! 706 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
586 break; !! 707 btrfs_set_and_info(info, DISCARD_SYNC,
587 case Opt_rescue_ignoredatacsum !! 708 "turning on sync discard");
588 btrfs_set_opt(ctx->mou !! 709 } else if (strcmp(args[0].from, "async") == 0) {
589 break; !! 710 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
590 case Opt_rescue_ignoremetacsum !! 711 btrfs_set_and_info(info, DISCARD_ASYNC,
591 btrfs_set_opt(ctx->mou !! 712 "turning on async discard");
592 break; !! 713 } else {
593 case Opt_rescue_ignoresuperfla !! 714 ret = -EINVAL;
594 btrfs_set_opt(ctx->mou !! 715 goto out;
595 break; !! 716 }
596 case Opt_rescue_parameter_all: !! 717 break;
597 btrfs_set_opt(ctx->mou !! 718 case Opt_nodiscard:
598 btrfs_set_opt(ctx->mou !! 719 btrfs_clear_and_info(info, DISCARD_SYNC,
599 btrfs_set_opt(ctx->mou !! 720 "turning off discard");
600 btrfs_set_opt(ctx->mou !! 721 btrfs_clear_and_info(info, DISCARD_ASYNC,
601 btrfs_set_opt(ctx->mou !! 722 "turning off async discard");
>> 723 break;
>> 724 case Opt_space_cache:
>> 725 case Opt_space_cache_version:
>> 726 if (token == Opt_space_cache ||
>> 727 strcmp(args[0].from, "v1") == 0) {
>> 728 btrfs_clear_opt(info->mount_opt,
>> 729 FREE_SPACE_TREE);
>> 730 btrfs_set_and_info(info, SPACE_CACHE,
>> 731 "enabling disk space caching");
>> 732 } else if (strcmp(args[0].from, "v2") == 0) {
>> 733 btrfs_clear_opt(info->mount_opt,
>> 734 SPACE_CACHE);
>> 735 btrfs_set_and_info(info, FREE_SPACE_TREE,
>> 736 "enabling free space tree");
>> 737 } else {
>> 738 ret = -EINVAL;
>> 739 goto out;
>> 740 }
>> 741 break;
>> 742 case Opt_rescan_uuid_tree:
>> 743 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
>> 744 break;
>> 745 case Opt_no_space_cache:
>> 746 if (btrfs_test_opt(info, SPACE_CACHE)) {
>> 747 btrfs_clear_and_info(info, SPACE_CACHE,
>> 748 "disabling disk space caching");
>> 749 }
>> 750 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
>> 751 btrfs_clear_and_info(info, FREE_SPACE_TREE,
>> 752 "disabling free space tree");
>> 753 }
>> 754 break;
>> 755 case Opt_inode_cache:
>> 756 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
>> 757 "enabling inode map caching");
>> 758 break;
>> 759 case Opt_noinode_cache:
>> 760 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
>> 761 "disabling inode map caching");
>> 762 break;
>> 763 case Opt_clear_cache:
>> 764 btrfs_set_and_info(info, CLEAR_CACHE,
>> 765 "force clearing of disk cache");
>> 766 break;
>> 767 case Opt_user_subvol_rm_allowed:
>> 768 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
>> 769 break;
>> 770 case Opt_enospc_debug:
>> 771 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
>> 772 break;
>> 773 case Opt_noenospc_debug:
>> 774 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
>> 775 break;
>> 776 case Opt_defrag:
>> 777 btrfs_set_and_info(info, AUTO_DEFRAG,
>> 778 "enabling auto defrag");
>> 779 break;
>> 780 case Opt_nodefrag:
>> 781 btrfs_clear_and_info(info, AUTO_DEFRAG,
>> 782 "disabling auto defrag");
>> 783 break;
>> 784 case Opt_recovery:
>> 785 btrfs_warn(info,
>> 786 "'recovery' is deprecated, use 'usebackuproot' instead");
>> 787 /* fall through */
>> 788 case Opt_usebackuproot:
>> 789 btrfs_info(info,
>> 790 "trying to use backup root at mount time");
>> 791 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
>> 792 break;
>> 793 case Opt_skip_balance:
>> 794 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
>> 795 break;
>> 796 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 797 case Opt_check_integrity_including_extent_data:
>> 798 btrfs_info(info,
>> 799 "enabling check integrity including extent data");
>> 800 btrfs_set_opt(info->mount_opt,
>> 801 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
>> 802 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
>> 803 break;
>> 804 case Opt_check_integrity:
>> 805 btrfs_info(info, "enabling check integrity");
>> 806 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
>> 807 break;
>> 808 case Opt_check_integrity_print_mask:
>> 809 ret = match_int(&args[0], &intarg);
>> 810 if (ret)
>> 811 goto out;
>> 812 info->check_integrity_print_mask = intarg;
>> 813 btrfs_info(info, "check_integrity_print_mask 0x%x",
>> 814 info->check_integrity_print_mask);
>> 815 break;
>> 816 #else
>> 817 case Opt_check_integrity_including_extent_data:
>> 818 case Opt_check_integrity:
>> 819 case Opt_check_integrity_print_mask:
>> 820 btrfs_err(info,
>> 821 "support for check_integrity* not compiled in!");
>> 822 ret = -EINVAL;
>> 823 goto out;
>> 824 #endif
>> 825 case Opt_fatal_errors:
>> 826 if (strcmp(args[0].from, "panic") == 0)
>> 827 btrfs_set_opt(info->mount_opt,
>> 828 PANIC_ON_FATAL_ERROR);
>> 829 else if (strcmp(args[0].from, "bug") == 0)
>> 830 btrfs_clear_opt(info->mount_opt,
>> 831 PANIC_ON_FATAL_ERROR);
>> 832 else {
>> 833 ret = -EINVAL;
>> 834 goto out;
>> 835 }
>> 836 break;
>> 837 case Opt_commit_interval:
>> 838 intarg = 0;
>> 839 ret = match_int(&args[0], &intarg);
>> 840 if (ret)
>> 841 goto out;
>> 842 if (intarg == 0) {
>> 843 btrfs_info(info,
>> 844 "using default commit interval %us",
>> 845 BTRFS_DEFAULT_COMMIT_INTERVAL);
>> 846 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
>> 847 } else if (intarg > 300) {
>> 848 btrfs_warn(info, "excessive commit interval %d",
>> 849 intarg);
>> 850 }
>> 851 info->commit_interval = intarg;
602 break; 852 break;
603 default: <<
604 btrfs_info(NULL, "unre <<
605 param->key) <<
606 return -EINVAL; <<
607 } <<
608 break; <<
609 #ifdef CONFIG_BTRFS_DEBUG 853 #ifdef CONFIG_BTRFS_DEBUG
610 case Opt_fragment: !! 854 case Opt_fragment_all:
611 switch (result.uint_32) { !! 855 btrfs_info(info, "fragmenting all space");
612 case Opt_fragment_parameter_al !! 856 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
613 btrfs_set_opt(ctx->mou !! 857 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
614 btrfs_set_opt(ctx->mou <<
615 break; 858 break;
616 case Opt_fragment_parameter_me !! 859 case Opt_fragment_metadata:
617 btrfs_set_opt(ctx->mou !! 860 btrfs_info(info, "fragmenting metadata");
>> 861 btrfs_set_opt(info->mount_opt,
>> 862 FRAGMENT_METADATA);
618 break; 863 break;
619 case Opt_fragment_parameter_da !! 864 case Opt_fragment_data:
620 btrfs_set_opt(ctx->mou !! 865 btrfs_info(info, "fragmenting data");
>> 866 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
621 break; 867 break;
622 default: <<
623 btrfs_info(NULL, "unre <<
624 param->key) <<
625 return -EINVAL; <<
626 } <<
627 break; <<
628 #endif 868 #endif
629 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 869 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
630 case Opt_ref_verify: !! 870 case Opt_ref_verify:
631 btrfs_set_opt(ctx->mount_opt, !! 871 btrfs_info(info, "doing ref verification");
632 break; !! 872 btrfs_set_opt(info->mount_opt, REF_VERIFY);
>> 873 break;
633 #endif 874 #endif
634 default: !! 875 case Opt_err:
635 btrfs_err(NULL, "unrecognized !! 876 btrfs_info(info, "unrecognized mount option '%s'", p);
636 return -EINVAL; !! 877 ret = -EINVAL;
>> 878 goto out;
>> 879 default:
>> 880 break;
>> 881 }
637 } 882 }
>> 883 check:
>> 884 /*
>> 885 * Extra check for current option against current flag
>> 886 */
>> 887 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
>> 888 btrfs_err(info,
>> 889 "nologreplay must be used with ro mount option");
>> 890 ret = -EINVAL;
>> 891 }
>> 892 out:
>> 893 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
>> 894 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
>> 895 !btrfs_test_opt(info, CLEAR_CACHE)) {
>> 896 btrfs_err(info, "cannot disable free space tree");
>> 897 ret = -EINVAL;
638 898
639 return 0; !! 899 }
>> 900 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
>> 901 btrfs_info(info, "disk space caching is enabled");
>> 902 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
>> 903 btrfs_info(info, "using free space tree");
>> 904 return ret;
640 } 905 }
641 906
642 /* 907 /*
643 * Some options only have meaning at mount tim !! 908 * Parse mount options that are required early in the mount process.
644 * remounts, or be displayed. Clear these at t !! 909 *
645 * paths. !! 910 * All other options will be parsed on much later in the mount process and
>> 911 * only when we need to allocate a new super block.
646 */ 912 */
647 static void btrfs_clear_oneshot_options(struct !! 913 static int btrfs_parse_device_options(const char *options, fmode_t flags,
>> 914 void *holder)
648 { 915 {
649 btrfs_clear_opt(fs_info->mount_opt, US !! 916 substring_t args[MAX_OPT_ARGS];
650 btrfs_clear_opt(fs_info->mount_opt, CL !! 917 char *device_name, *opts, *orig, *p;
651 btrfs_clear_opt(fs_info->mount_opt, NO !! 918 struct btrfs_device *device = NULL;
652 } !! 919 int error = 0;
653 920
654 static bool check_ro_option(const struct btrfs !! 921 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 } <<
665 922
666 bool btrfs_check_options(const struct btrfs_fs !! 923 if (!options)
667 unsigned long long *m !! 924 return 0;
668 unsigned long flags) !! 925
669 { !! 926 /*
670 bool ret = true; !! 927 * strsep changes the string, duplicate it because btrfs_parse_options
>> 928 * gets called later
>> 929 */
>> 930 opts = kstrdup(options, GFP_KERNEL);
>> 931 if (!opts)
>> 932 return -ENOMEM;
>> 933 orig = opts;
671 934
672 if (!(flags & SB_RDONLY) && !! 935 while ((p = strsep(&opts, ",")) != NULL) {
673 (check_ro_option(info, *mount_opt, !! 936 int token;
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; <<
679 937
680 if (btrfs_fs_compat_ro(info, FREE_SPAC !! 938 if (!*p)
681 !btrfs_raw_test_opt(*mount_opt, FR !! 939 continue;
682 !btrfs_raw_test_opt(*mount_opt, CL !! 940
683 btrfs_err(info, "cannot disabl !! 941 token = match_token(p, tokens, args);
684 ret = false; !! 942 if (token == Opt_device) {
685 } !! 943 device_name = match_strdup(&args[0]);
686 if (btrfs_fs_compat_ro(info, BLOCK_GRO !! 944 if (!device_name) {
687 !btrfs_raw_test_opt(*mount_opt, F !! 945 error = -ENOMEM;
688 btrfs_err(info, "cannot disabl !! 946 goto out;
689 ret = false; !! 947 }
690 } !! 948 device = btrfs_scan_one_device(device_name, flags,
691 !! 949 holder);
692 if (btrfs_check_mountopts_zoned(info, !! 950 kfree(device_name);
693 ret = false; !! 951 if (IS_ERR(device)) {
694 !! 952 error = PTR_ERR(device);
695 if (!test_bit(BTRFS_FS_STATE_REMOUNTIN !! 953 goto out;
696 if (btrfs_raw_test_opt(*mount_ !! 954 }
697 btrfs_info(info, "disk <<
698 btrfs_warn(info, <<
699 "space cache v1 is being deprecated and will b <<
700 } 955 }
701 if (btrfs_raw_test_opt(*mount_ <<
702 btrfs_info(info, "usin <<
703 } 956 }
704 957
705 return ret; !! 958 out:
>> 959 kfree(orig);
>> 960 return error;
706 } 961 }
707 962
708 /* 963 /*
709 * This is subtle, we only call this during op !! 964 * 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 * 965 *
714 * This isn't a change in behavior, because we !! 966 * 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 */ 967 */
720 void btrfs_set_free_space_cache_settings(struc !! 968 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
>> 969 u64 *subvol_objectid)
721 { 970 {
722 if (fs_info->sectorsize < PAGE_SIZE) { !! 971 substring_t args[MAX_OPT_ARGS];
723 btrfs_clear_opt(fs_info->mount !! 972 char *opts, *orig, *p;
724 if (!btrfs_test_opt(fs_info, F !! 973 int error = 0;
725 btrfs_info(fs_info, !! 974 u64 subvolid;
726 "forcing fr !! 975
727 fs_info->se !! 976 if (!options)
728 btrfs_set_opt(fs_info- !! 977 return 0;
729 } <<
730 } <<
731 978
732 /* 979 /*
733 * At this point our mount options are !! 980 * strsep changes the string, duplicate it because
734 * these settings if we don't have any !! 981 * btrfs_parse_device_options gets called later
735 */ 982 */
736 if (btrfs_test_opt(fs_info, FREE_SPACE !! 983 opts = kstrdup(options, GFP_KERNEL);
737 return; !! 984 if (!opts)
738 !! 985 return -ENOMEM;
739 if (btrfs_is_zoned(fs_info) && !! 986 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 } <<
745 987
746 if (btrfs_test_opt(fs_info, SPACE_CACH !! 988 while ((p = strsep(&opts, ",")) != NULL) {
747 return; !! 989 int token;
>> 990 if (!*p)
>> 991 continue;
748 992
749 if (btrfs_test_opt(fs_info, NOSPACECAC !! 993 token = match_token(p, tokens, args);
750 return; !! 994 switch (token) {
>> 995 case Opt_subvol:
>> 996 kfree(*subvol_name);
>> 997 *subvol_name = match_strdup(&args[0]);
>> 998 if (!*subvol_name) {
>> 999 error = -ENOMEM;
>> 1000 goto out;
>> 1001 }
>> 1002 break;
>> 1003 case Opt_subvolid:
>> 1004 error = match_u64(&args[0], &subvolid);
>> 1005 if (error)
>> 1006 goto out;
751 1007
752 /* !! 1008 /* we want the original fs_tree */
753 * At this point we don't have explici !! 1009 if (subvolid == 0)
754 * them ourselves based on the state o !! 1010 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 1011
762 static void set_device_specific_options(struct !! 1012 *subvol_objectid = subvolid;
763 { !! 1013 break;
764 if (!btrfs_test_opt(fs_info, NOSSD) && !! 1014 case Opt_subvolrootid:
765 !fs_info->fs_devices->rotating) !! 1015 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
766 btrfs_set_opt(fs_info->mount_o !! 1016 break;
>> 1017 default:
>> 1018 break;
>> 1019 }
>> 1020 }
767 1021
768 /* !! 1022 out:
769 * For devices supporting discard turn !! 1023 kfree(orig);
770 * unless it's already set or disabled !! 1024 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 } 1025 }
784 1026
785 char *btrfs_get_subvol_name_from_objectid(stru !! 1027 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
786 u64 !! 1028 u64 subvol_objectid)
787 { 1029 {
788 struct btrfs_root *root = fs_info->tre 1030 struct btrfs_root *root = fs_info->tree_root;
789 struct btrfs_root *fs_root = NULL; !! 1031 struct btrfs_root *fs_root;
790 struct btrfs_root_ref *root_ref; 1032 struct btrfs_root_ref *root_ref;
791 struct btrfs_inode_ref *inode_ref; 1033 struct btrfs_inode_ref *inode_ref;
792 struct btrfs_key key; 1034 struct btrfs_key key;
793 struct btrfs_path *path = NULL; 1035 struct btrfs_path *path = NULL;
794 char *name = NULL, *ptr; 1036 char *name = NULL, *ptr;
795 u64 dirid; 1037 u64 dirid;
796 int len; 1038 int len;
797 int ret; 1039 int ret;
798 1040
799 path = btrfs_alloc_path(); 1041 path = btrfs_alloc_path();
800 if (!path) { 1042 if (!path) {
801 ret = -ENOMEM; 1043 ret = -ENOMEM;
802 goto err; 1044 goto err;
803 } 1045 }
>> 1046 path->leave_spinning = 1;
804 1047
805 name = kmalloc(PATH_MAX, GFP_KERNEL); 1048 name = kmalloc(PATH_MAX, GFP_KERNEL);
806 if (!name) { 1049 if (!name) {
807 ret = -ENOMEM; 1050 ret = -ENOMEM;
808 goto err; 1051 goto err;
809 } 1052 }
810 ptr = name + PATH_MAX - 1; 1053 ptr = name + PATH_MAX - 1;
811 ptr[0] = '\0'; 1054 ptr[0] = '\0';
812 1055
813 /* 1056 /*
814 * Walk up the subvolume trees in the 1057 * Walk up the subvolume trees in the tree of tree roots by root
815 * backrefs until we hit the top-level 1058 * backrefs until we hit the top-level subvolume.
816 */ 1059 */
817 while (subvol_objectid != BTRFS_FS_TRE 1060 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
818 key.objectid = subvol_objectid 1061 key.objectid = subvol_objectid;
819 key.type = BTRFS_ROOT_BACKREF_ 1062 key.type = BTRFS_ROOT_BACKREF_KEY;
820 key.offset = (u64)-1; 1063 key.offset = (u64)-1;
821 1064
822 ret = btrfs_search_backwards(r !! 1065 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
823 if (ret < 0) { 1066 if (ret < 0) {
824 goto err; 1067 goto err;
825 } else if (ret > 0) { 1068 } else if (ret > 0) {
826 ret = -ENOENT; !! 1069 ret = btrfs_previous_item(root, path, subvol_objectid,
827 goto err; !! 1070 BTRFS_ROOT_BACKREF_KEY);
>> 1071 if (ret < 0) {
>> 1072 goto err;
>> 1073 } else if (ret > 0) {
>> 1074 ret = -ENOENT;
>> 1075 goto err;
>> 1076 }
828 } 1077 }
829 1078
>> 1079 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
830 subvol_objectid = key.offset; 1080 subvol_objectid = key.offset;
831 1081
832 root_ref = btrfs_item_ptr(path 1082 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
833 stru 1083 struct btrfs_root_ref);
834 len = btrfs_root_ref_name_len( 1084 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
835 ptr -= len + 1; 1085 ptr -= len + 1;
836 if (ptr < name) { 1086 if (ptr < name) {
837 ret = -ENAMETOOLONG; 1087 ret = -ENAMETOOLONG;
838 goto err; 1088 goto err;
839 } 1089 }
840 read_extent_buffer(path->nodes 1090 read_extent_buffer(path->nodes[0], ptr + 1,
841 (unsigned l 1091 (unsigned long)(root_ref + 1), len);
842 ptr[0] = '/'; 1092 ptr[0] = '/';
843 dirid = btrfs_root_ref_dirid(p 1093 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
844 btrfs_release_path(path); 1094 btrfs_release_path(path);
845 1095
846 fs_root = btrfs_get_fs_root(fs !! 1096 key.objectid = subvol_objectid;
>> 1097 key.type = BTRFS_ROOT_ITEM_KEY;
>> 1098 key.offset = (u64)-1;
>> 1099 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
847 if (IS_ERR(fs_root)) { 1100 if (IS_ERR(fs_root)) {
848 ret = PTR_ERR(fs_root) 1101 ret = PTR_ERR(fs_root);
849 fs_root = NULL; <<
850 goto err; 1102 goto err;
851 } 1103 }
852 1104
853 /* 1105 /*
854 * Walk up the filesystem tree 1106 * Walk up the filesystem tree by inode refs until we hit the
855 * root directory. 1107 * root directory.
856 */ 1108 */
857 while (dirid != BTRFS_FIRST_FR 1109 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
858 key.objectid = dirid; 1110 key.objectid = dirid;
859 key.type = BTRFS_INODE 1111 key.type = BTRFS_INODE_REF_KEY;
860 key.offset = (u64)-1; 1112 key.offset = (u64)-1;
861 1113
862 ret = btrfs_search_bac !! 1114 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
863 if (ret < 0) { 1115 if (ret < 0) {
864 goto err; 1116 goto err;
865 } else if (ret > 0) { 1117 } else if (ret > 0) {
866 ret = -ENOENT; !! 1118 ret = btrfs_previous_item(fs_root, path, dirid,
867 goto err; !! 1119 BTRFS_INODE_REF_KEY);
>> 1120 if (ret < 0) {
>> 1121 goto err;
>> 1122 } else if (ret > 0) {
>> 1123 ret = -ENOENT;
>> 1124 goto err;
>> 1125 }
868 } 1126 }
869 1127
>> 1128 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
870 dirid = key.offset; 1129 dirid = key.offset;
871 1130
872 inode_ref = btrfs_item 1131 inode_ref = btrfs_item_ptr(path->nodes[0],
873 1132 path->slots[0],
874 1133 struct btrfs_inode_ref);
875 len = btrfs_inode_ref_ 1134 len = btrfs_inode_ref_name_len(path->nodes[0],
876 1135 inode_ref);
877 ptr -= len + 1; 1136 ptr -= len + 1;
878 if (ptr < name) { 1137 if (ptr < name) {
879 ret = -ENAMETO 1138 ret = -ENAMETOOLONG;
880 goto err; 1139 goto err;
881 } 1140 }
882 read_extent_buffer(pat 1141 read_extent_buffer(path->nodes[0], ptr + 1,
883 (un 1142 (unsigned long)(inode_ref + 1), len);
884 ptr[0] = '/'; 1143 ptr[0] = '/';
885 btrfs_release_path(pat 1144 btrfs_release_path(path);
886 } 1145 }
887 btrfs_put_root(fs_root); <<
888 fs_root = NULL; <<
889 } 1146 }
890 1147
891 btrfs_free_path(path); 1148 btrfs_free_path(path);
892 if (ptr == name + PATH_MAX - 1) { 1149 if (ptr == name + PATH_MAX - 1) {
893 name[0] = '/'; 1150 name[0] = '/';
894 name[1] = '\0'; 1151 name[1] = '\0';
895 } else { 1152 } else {
896 memmove(name, ptr, name + PATH 1153 memmove(name, ptr, name + PATH_MAX - ptr);
897 } 1154 }
898 return name; 1155 return name;
899 1156
900 err: 1157 err:
901 btrfs_put_root(fs_root); <<
902 btrfs_free_path(path); 1158 btrfs_free_path(path);
903 kfree(name); 1159 kfree(name);
904 return ERR_PTR(ret); 1160 return ERR_PTR(ret);
905 } 1161 }
906 1162
907 static int get_default_subvol_objectid(struct 1163 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
908 { 1164 {
909 struct btrfs_root *root = fs_info->tre 1165 struct btrfs_root *root = fs_info->tree_root;
910 struct btrfs_dir_item *di; 1166 struct btrfs_dir_item *di;
911 struct btrfs_path *path; 1167 struct btrfs_path *path;
912 struct btrfs_key location; 1168 struct btrfs_key location;
913 struct fscrypt_str name = FSTR_INIT("d <<
914 u64 dir_id; 1169 u64 dir_id;
915 1170
916 path = btrfs_alloc_path(); 1171 path = btrfs_alloc_path();
917 if (!path) 1172 if (!path)
918 return -ENOMEM; 1173 return -ENOMEM;
>> 1174 path->leave_spinning = 1;
919 1175
920 /* 1176 /*
921 * Find the "default" dir item which p 1177 * Find the "default" dir item which points to the root item that we
922 * will mount by default if we haven't 1178 * will mount by default if we haven't been given a specific subvolume
923 * to mount. 1179 * to mount.
924 */ 1180 */
925 dir_id = btrfs_super_root_dir(fs_info- 1181 dir_id = btrfs_super_root_dir(fs_info->super_copy);
926 di = btrfs_lookup_dir_item(NULL, root, !! 1182 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
927 if (IS_ERR(di)) { 1183 if (IS_ERR(di)) {
928 btrfs_free_path(path); 1184 btrfs_free_path(path);
929 return PTR_ERR(di); 1185 return PTR_ERR(di);
930 } 1186 }
931 if (!di) { 1187 if (!di) {
932 /* 1188 /*
933 * Ok the default dir item isn 1189 * Ok the default dir item isn't there. This is weird since
934 * it's always been there, but 1190 * it's always been there, but don't freak out, just try and
935 * mount the top-level subvolu 1191 * mount the top-level subvolume.
936 */ 1192 */
937 btrfs_free_path(path); 1193 btrfs_free_path(path);
938 *objectid = BTRFS_FS_TREE_OBJE 1194 *objectid = BTRFS_FS_TREE_OBJECTID;
939 return 0; 1195 return 0;
940 } 1196 }
941 1197
942 btrfs_dir_item_key_to_cpu(path->nodes[ 1198 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
943 btrfs_free_path(path); 1199 btrfs_free_path(path);
944 *objectid = location.objectid; 1200 *objectid = location.objectid;
945 return 0; 1201 return 0;
946 } 1202 }
947 1203
948 static int btrfs_fill_super(struct super_block 1204 static int btrfs_fill_super(struct super_block *sb,
949 struct btrfs_fs_de 1205 struct btrfs_fs_devices *fs_devices,
950 void *data) 1206 void *data)
951 { 1207 {
952 struct inode *inode; 1208 struct inode *inode;
953 struct btrfs_fs_info *fs_info = btrfs_ 1209 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 1210 struct btrfs_key key;
954 int err; 1211 int err;
955 1212
956 sb->s_maxbytes = MAX_LFS_FILESIZE; 1213 sb->s_maxbytes = MAX_LFS_FILESIZE;
957 sb->s_magic = BTRFS_SUPER_MAGIC; 1214 sb->s_magic = BTRFS_SUPER_MAGIC;
958 sb->s_op = &btrfs_super_ops; 1215 sb->s_op = &btrfs_super_ops;
959 sb->s_d_op = &btrfs_dentry_operations; 1216 sb->s_d_op = &btrfs_dentry_operations;
960 sb->s_export_op = &btrfs_export_ops; 1217 sb->s_export_op = &btrfs_export_ops;
961 #ifdef CONFIG_FS_VERITY <<
962 sb->s_vop = &btrfs_verityops; <<
963 #endif <<
964 sb->s_xattr = btrfs_xattr_handlers; 1218 sb->s_xattr = btrfs_xattr_handlers;
965 sb->s_time_gran = 1; 1219 sb->s_time_gran = 1;
>> 1220 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
>> 1221 sb->s_flags |= SB_POSIXACL;
>> 1222 #endif
>> 1223 sb->s_flags |= SB_I_VERSION;
966 sb->s_iflags |= SB_I_CGROUPWB; 1224 sb->s_iflags |= SB_I_CGROUPWB;
967 1225
968 err = super_setup_bdi(sb); 1226 err = super_setup_bdi(sb);
969 if (err) { 1227 if (err) {
970 btrfs_err(fs_info, "super_setu 1228 btrfs_err(fs_info, "super_setup_bdi failed");
971 return err; 1229 return err;
972 } 1230 }
973 1231
974 err = open_ctree(sb, fs_devices, (char 1232 err = open_ctree(sb, fs_devices, (char *)data);
975 if (err) { 1233 if (err) {
976 btrfs_err(fs_info, "open_ctree 1234 btrfs_err(fs_info, "open_ctree failed");
977 return err; 1235 return err;
978 } 1236 }
979 1237
980 inode = btrfs_iget(BTRFS_FIRST_FREE_OB !! 1238 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
>> 1239 key.type = BTRFS_INODE_ITEM_KEY;
>> 1240 key.offset = 0;
>> 1241 inode = btrfs_iget(sb, &key, fs_info->fs_root);
981 if (IS_ERR(inode)) { 1242 if (IS_ERR(inode)) {
982 err = PTR_ERR(inode); 1243 err = PTR_ERR(inode);
983 btrfs_handle_fs_error(fs_info, <<
984 goto fail_close; 1244 goto fail_close;
985 } 1245 }
986 1246
987 sb->s_root = d_make_root(inode); 1247 sb->s_root = d_make_root(inode);
988 if (!sb->s_root) { 1248 if (!sb->s_root) {
989 err = -ENOMEM; 1249 err = -ENOMEM;
990 goto fail_close; 1250 goto fail_close;
991 } 1251 }
992 1252
>> 1253 cleancache_init_fs(sb);
993 sb->s_flags |= SB_ACTIVE; 1254 sb->s_flags |= SB_ACTIVE;
994 return 0; 1255 return 0;
995 1256
996 fail_close: 1257 fail_close:
997 close_ctree(fs_info); 1258 close_ctree(fs_info);
998 return err; 1259 return err;
999 } 1260 }
1000 1261
1001 int btrfs_sync_fs(struct super_block *sb, int 1262 int btrfs_sync_fs(struct super_block *sb, int wait)
1002 { 1263 {
1003 struct btrfs_trans_handle *trans; 1264 struct btrfs_trans_handle *trans;
1004 struct btrfs_fs_info *fs_info = btrfs 1265 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1005 struct btrfs_root *root = fs_info->tr 1266 struct btrfs_root *root = fs_info->tree_root;
1006 1267
1007 trace_btrfs_sync_fs(fs_info, wait); 1268 trace_btrfs_sync_fs(fs_info, wait);
1008 1269
1009 if (!wait) { 1270 if (!wait) {
1010 filemap_flush(fs_info->btree_ 1271 filemap_flush(fs_info->btree_inode->i_mapping);
1011 return 0; 1272 return 0;
1012 } 1273 }
1013 1274
1014 btrfs_wait_ordered_roots(fs_info, U64 !! 1275 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1015 1276
1016 trans = btrfs_attach_transaction_barr 1277 trans = btrfs_attach_transaction_barrier(root);
1017 if (IS_ERR(trans)) { 1278 if (IS_ERR(trans)) {
1018 /* no transaction, don't both 1279 /* no transaction, don't bother */
1019 if (PTR_ERR(trans) == -ENOENT 1280 if (PTR_ERR(trans) == -ENOENT) {
1020 /* 1281 /*
1021 * Exit unless we hav 1282 * Exit unless we have some pending changes
1022 * that need to go th 1283 * that need to go through commit
1023 */ 1284 */
1024 if (!test_bit(BTRFS_F !! 1285 if (fs_info->pending_changes == 0)
1025 &fs_inf <<
1026 return 0; 1286 return 0;
1027 /* 1287 /*
1028 * A non-blocking tes 1288 * A non-blocking test if the fs is frozen. We must not
1029 * start a new transa 1289 * start a new transaction here otherwise a deadlock
1030 * happens. The pendi 1290 * happens. The pending operations are delayed to the
1031 * next commit after 1291 * next commit after thawing.
1032 */ 1292 */
1033 if (sb_start_write_tr 1293 if (sb_start_write_trylock(sb))
1034 sb_end_write( 1294 sb_end_write(sb);
1035 else 1295 else
1036 return 0; 1296 return 0;
1037 trans = btrfs_start_t 1297 trans = btrfs_start_transaction(root, 0);
1038 } 1298 }
1039 if (IS_ERR(trans)) 1299 if (IS_ERR(trans))
1040 return PTR_ERR(trans) 1300 return PTR_ERR(trans);
1041 } 1301 }
1042 return btrfs_commit_transaction(trans 1302 return btrfs_commit_transaction(trans);
1043 } 1303 }
1044 1304
1045 static void print_rescue_option(struct seq_fi <<
1046 { <<
1047 seq_printf(seq, "%s%s", (*printed) ? <<
1048 *printed = true; <<
1049 } <<
1050 <<
1051 static int btrfs_show_options(struct seq_file 1305 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1052 { 1306 {
1053 struct btrfs_fs_info *info = btrfs_sb 1307 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1054 const char *compress_type; 1308 const char *compress_type;
1055 const char *subvol_name; <<
1056 bool printed = false; <<
1057 1309
1058 if (btrfs_test_opt(info, DEGRADED)) 1310 if (btrfs_test_opt(info, DEGRADED))
1059 seq_puts(seq, ",degraded"); 1311 seq_puts(seq, ",degraded");
1060 if (btrfs_test_opt(info, NODATASUM)) 1312 if (btrfs_test_opt(info, NODATASUM))
1061 seq_puts(seq, ",nodatasum"); 1313 seq_puts(seq, ",nodatasum");
1062 if (btrfs_test_opt(info, NODATACOW)) 1314 if (btrfs_test_opt(info, NODATACOW))
1063 seq_puts(seq, ",nodatacow"); 1315 seq_puts(seq, ",nodatacow");
1064 if (btrfs_test_opt(info, NOBARRIER)) 1316 if (btrfs_test_opt(info, NOBARRIER))
1065 seq_puts(seq, ",nobarrier"); 1317 seq_puts(seq, ",nobarrier");
1066 if (info->max_inline != BTRFS_DEFAULT 1318 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1067 seq_printf(seq, ",max_inline= 1319 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1068 if (info->thread_pool_size != min_t( 1320 if (info->thread_pool_size != min_t(unsigned long,
1069 1321 num_online_cpus() + 2, 8))
1070 seq_printf(seq, ",thread_pool 1322 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1071 if (btrfs_test_opt(info, COMPRESS)) { 1323 if (btrfs_test_opt(info, COMPRESS)) {
1072 compress_type = btrfs_compres 1324 compress_type = btrfs_compress_type2str(info->compress_type);
1073 if (btrfs_test_opt(info, FORC 1325 if (btrfs_test_opt(info, FORCE_COMPRESS))
1074 seq_printf(seq, ",com 1326 seq_printf(seq, ",compress-force=%s", compress_type);
1075 else 1327 else
1076 seq_printf(seq, ",com 1328 seq_printf(seq, ",compress=%s", compress_type);
1077 if (info->compress_level) 1329 if (info->compress_level)
1078 seq_printf(seq, ":%d" 1330 seq_printf(seq, ":%d", info->compress_level);
1079 } 1331 }
1080 if (btrfs_test_opt(info, NOSSD)) 1332 if (btrfs_test_opt(info, NOSSD))
1081 seq_puts(seq, ",nossd"); 1333 seq_puts(seq, ",nossd");
1082 if (btrfs_test_opt(info, SSD_SPREAD)) 1334 if (btrfs_test_opt(info, SSD_SPREAD))
1083 seq_puts(seq, ",ssd_spread"); 1335 seq_puts(seq, ",ssd_spread");
1084 else if (btrfs_test_opt(info, SSD)) 1336 else if (btrfs_test_opt(info, SSD))
1085 seq_puts(seq, ",ssd"); 1337 seq_puts(seq, ",ssd");
1086 if (btrfs_test_opt(info, NOTREELOG)) 1338 if (btrfs_test_opt(info, NOTREELOG))
1087 seq_puts(seq, ",notreelog"); 1339 seq_puts(seq, ",notreelog");
1088 if (btrfs_test_opt(info, NOLOGREPLAY) 1340 if (btrfs_test_opt(info, NOLOGREPLAY))
1089 print_rescue_option(seq, "nol !! 1341 seq_puts(seq, ",nologreplay");
1090 if (btrfs_test_opt(info, USEBACKUPROO <<
1091 print_rescue_option(seq, "use <<
1092 if (btrfs_test_opt(info, IGNOREBADROO <<
1093 print_rescue_option(seq, "ign <<
1094 if (btrfs_test_opt(info, IGNOREDATACS <<
1095 print_rescue_option(seq, "ign <<
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 1342 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1101 seq_puts(seq, ",flushoncommit 1343 seq_puts(seq, ",flushoncommit");
1102 if (btrfs_test_opt(info, DISCARD_SYNC 1344 if (btrfs_test_opt(info, DISCARD_SYNC))
1103 seq_puts(seq, ",discard"); 1345 seq_puts(seq, ",discard");
1104 if (btrfs_test_opt(info, DISCARD_ASYN 1346 if (btrfs_test_opt(info, DISCARD_ASYNC))
1105 seq_puts(seq, ",discard=async 1347 seq_puts(seq, ",discard=async");
1106 if (!(info->sb->s_flags & SB_POSIXACL 1348 if (!(info->sb->s_flags & SB_POSIXACL))
1107 seq_puts(seq, ",noacl"); 1349 seq_puts(seq, ",noacl");
1108 if (btrfs_free_space_cache_v1_active( !! 1350 if (btrfs_test_opt(info, SPACE_CACHE))
1109 seq_puts(seq, ",space_cache") 1351 seq_puts(seq, ",space_cache");
1110 else if (btrfs_fs_compat_ro(info, FRE !! 1352 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1111 seq_puts(seq, ",space_cache=v 1353 seq_puts(seq, ",space_cache=v2");
1112 else 1354 else
1113 seq_puts(seq, ",nospace_cache 1355 seq_puts(seq, ",nospace_cache");
1114 if (btrfs_test_opt(info, RESCAN_UUID_ 1356 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1115 seq_puts(seq, ",rescan_uuid_t 1357 seq_puts(seq, ",rescan_uuid_tree");
1116 if (btrfs_test_opt(info, CLEAR_CACHE) 1358 if (btrfs_test_opt(info, CLEAR_CACHE))
1117 seq_puts(seq, ",clear_cache") 1359 seq_puts(seq, ",clear_cache");
1118 if (btrfs_test_opt(info, USER_SUBVOL_ 1360 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1119 seq_puts(seq, ",user_subvol_r 1361 seq_puts(seq, ",user_subvol_rm_allowed");
1120 if (btrfs_test_opt(info, ENOSPC_DEBUG 1362 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1121 seq_puts(seq, ",enospc_debug" 1363 seq_puts(seq, ",enospc_debug");
1122 if (btrfs_test_opt(info, AUTO_DEFRAG) 1364 if (btrfs_test_opt(info, AUTO_DEFRAG))
1123 seq_puts(seq, ",autodefrag"); 1365 seq_puts(seq, ",autodefrag");
>> 1366 if (btrfs_test_opt(info, INODE_MAP_CACHE))
>> 1367 seq_puts(seq, ",inode_cache");
1124 if (btrfs_test_opt(info, SKIP_BALANCE 1368 if (btrfs_test_opt(info, SKIP_BALANCE))
1125 seq_puts(seq, ",skip_balance" 1369 seq_puts(seq, ",skip_balance");
>> 1370 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 1371 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
>> 1372 seq_puts(seq, ",check_int_data");
>> 1373 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
>> 1374 seq_puts(seq, ",check_int");
>> 1375 if (info->check_integrity_print_mask)
>> 1376 seq_printf(seq, ",check_int_print_mask=%d",
>> 1377 info->check_integrity_print_mask);
>> 1378 #endif
1126 if (info->metadata_ratio) 1379 if (info->metadata_ratio)
1127 seq_printf(seq, ",metadata_ra 1380 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1128 if (btrfs_test_opt(info, PANIC_ON_FAT 1381 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1129 seq_puts(seq, ",fatal_errors= 1382 seq_puts(seq, ",fatal_errors=panic");
1130 if (info->commit_interval != BTRFS_DE 1383 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1131 seq_printf(seq, ",commit=%u", 1384 seq_printf(seq, ",commit=%u", info->commit_interval);
1132 #ifdef CONFIG_BTRFS_DEBUG 1385 #ifdef CONFIG_BTRFS_DEBUG
1133 if (btrfs_test_opt(info, FRAGMENT_DAT 1386 if (btrfs_test_opt(info, FRAGMENT_DATA))
1134 seq_puts(seq, ",fragment=data 1387 seq_puts(seq, ",fragment=data");
1135 if (btrfs_test_opt(info, FRAGMENT_MET 1388 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1136 seq_puts(seq, ",fragment=meta 1389 seq_puts(seq, ",fragment=metadata");
1137 #endif 1390 #endif
1138 if (btrfs_test_opt(info, REF_VERIFY)) 1391 if (btrfs_test_opt(info, REF_VERIFY))
1139 seq_puts(seq, ",ref_verify"); 1392 seq_puts(seq, ",ref_verify");
1140 seq_printf(seq, ",subvolid=%llu", btr !! 1393 seq_printf(seq, ",subvolid=%llu",
1141 subvol_name = btrfs_get_subvol_name_f !! 1394 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1142 btrfs_root_id(BTRFS_I !! 1395 seq_puts(seq, ",subvol=");
1143 if (!IS_ERR(subvol_name)) { !! 1396 seq_dentry(seq, dentry, " \t\n\\");
1144 seq_puts(seq, ",subvol="); <<
1145 seq_escape(seq, subvol_name, <<
1146 kfree(subvol_name); <<
1147 } <<
1148 return 0; 1397 return 0;
1149 } 1398 }
1150 1399
>> 1400 static int btrfs_test_super(struct super_block *s, void *data)
>> 1401 {
>> 1402 struct btrfs_fs_info *p = data;
>> 1403 struct btrfs_fs_info *fs_info = btrfs_sb(s);
>> 1404
>> 1405 return fs_info->fs_devices == p->fs_devices;
>> 1406 }
>> 1407
>> 1408 static int btrfs_set_super(struct super_block *s, void *data)
>> 1409 {
>> 1410 int err = set_anon_super(s, data);
>> 1411 if (!err)
>> 1412 s->s_fs_info = data;
>> 1413 return err;
>> 1414 }
>> 1415
1151 /* 1416 /*
1152 * subvolumes are identified by ino 256 1417 * subvolumes are identified by ino 256
1153 */ 1418 */
1154 static inline int is_subvolume_inode(struct i 1419 static inline int is_subvolume_inode(struct inode *inode)
1155 { 1420 {
1156 if (inode && inode->i_ino == BTRFS_FI 1421 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1157 return 1; 1422 return 1;
1158 return 0; 1423 return 0;
1159 } 1424 }
1160 1425
1161 static struct dentry *mount_subvol(const char 1426 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1162 struct vfs 1427 struct vfsmount *mnt)
1163 { 1428 {
1164 struct dentry *root; 1429 struct dentry *root;
1165 int ret; 1430 int ret;
1166 1431
1167 if (!subvol_name) { 1432 if (!subvol_name) {
1168 if (!subvol_objectid) { 1433 if (!subvol_objectid) {
1169 ret = get_default_sub 1434 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1170 1435 &subvol_objectid);
1171 if (ret) { 1436 if (ret) {
1172 root = ERR_PT 1437 root = ERR_PTR(ret);
1173 goto out; 1438 goto out;
1174 } 1439 }
1175 } 1440 }
1176 subvol_name = btrfs_get_subvo !! 1441 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1177 btrfs !! 1442 subvol_objectid);
1178 if (IS_ERR(subvol_name)) { 1443 if (IS_ERR(subvol_name)) {
1179 root = ERR_CAST(subvo 1444 root = ERR_CAST(subvol_name);
1180 subvol_name = NULL; 1445 subvol_name = NULL;
1181 goto out; 1446 goto out;
1182 } 1447 }
1183 1448
1184 } 1449 }
1185 1450
1186 root = mount_subtree(mnt, subvol_name 1451 root = mount_subtree(mnt, subvol_name);
1187 /* mount_subtree() drops our referenc 1452 /* mount_subtree() drops our reference on the vfsmount. */
1188 mnt = NULL; 1453 mnt = NULL;
1189 1454
1190 if (!IS_ERR(root)) { 1455 if (!IS_ERR(root)) {
1191 struct super_block *s = root- 1456 struct super_block *s = root->d_sb;
1192 struct btrfs_fs_info *fs_info 1457 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1193 struct inode *root_inode = d_ 1458 struct inode *root_inode = d_inode(root);
1194 u64 root_objectid = btrfs_roo !! 1459 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1195 1460
1196 ret = 0; 1461 ret = 0;
1197 if (!is_subvolume_inode(root_ 1462 if (!is_subvolume_inode(root_inode)) {
1198 btrfs_err(fs_info, "' 1463 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1199 subvol_name); 1464 subvol_name);
1200 ret = -EINVAL; 1465 ret = -EINVAL;
1201 } 1466 }
1202 if (subvol_objectid && root_o 1467 if (subvol_objectid && root_objectid != subvol_objectid) {
1203 /* 1468 /*
1204 * This will also cat 1469 * This will also catch a race condition where a
1205 * subvolume which wa 1470 * subvolume which was passed by ID is renamed and
1206 * another subvolume 1471 * another subvolume is renamed over the old location.
1207 */ 1472 */
1208 btrfs_err(fs_info, 1473 btrfs_err(fs_info,
1209 "subvol '%s 1474 "subvol '%s' does not match subvolid %llu",
1210 subvol_name 1475 subvol_name, subvol_objectid);
1211 ret = -EINVAL; 1476 ret = -EINVAL;
1212 } 1477 }
1213 if (ret) { 1478 if (ret) {
1214 dput(root); 1479 dput(root);
1215 root = ERR_PTR(ret); 1480 root = ERR_PTR(ret);
1216 deactivate_locked_sup 1481 deactivate_locked_super(s);
1217 } 1482 }
1218 } 1483 }
1219 1484
1220 out: 1485 out:
1221 mntput(mnt); 1486 mntput(mnt);
1222 kfree(subvol_name); 1487 kfree(subvol_name);
1223 return root; 1488 return root;
1224 } 1489 }
1225 1490
>> 1491 /*
>> 1492 * Find a superblock for the given device / mount point.
>> 1493 *
>> 1494 * Note: This is based on mount_bdev from fs/super.c with a few additions
>> 1495 * for multiple device setup. Make sure to keep it in sync.
>> 1496 */
>> 1497 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
>> 1498 int flags, const char *device_name, void *data)
>> 1499 {
>> 1500 struct block_device *bdev = NULL;
>> 1501 struct super_block *s;
>> 1502 struct btrfs_device *device = NULL;
>> 1503 struct btrfs_fs_devices *fs_devices = NULL;
>> 1504 struct btrfs_fs_info *fs_info = NULL;
>> 1505 void *new_sec_opts = NULL;
>> 1506 fmode_t mode = FMODE_READ;
>> 1507 int error = 0;
>> 1508
>> 1509 if (!(flags & SB_RDONLY))
>> 1510 mode |= FMODE_WRITE;
>> 1511
>> 1512 if (data) {
>> 1513 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
>> 1514 if (error)
>> 1515 return ERR_PTR(error);
>> 1516 }
>> 1517
>> 1518 /*
>> 1519 * Setup a dummy root and fs_info for test/set super. This is because
>> 1520 * we don't actually fill this stuff out until open_ctree, but we need
>> 1521 * it for searching for existing supers, so this lets us do that and
>> 1522 * then open_ctree will properly initialize everything later.
>> 1523 */
>> 1524 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
>> 1525 if (!fs_info) {
>> 1526 error = -ENOMEM;
>> 1527 goto error_sec_opts;
>> 1528 }
>> 1529
>> 1530 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
>> 1531 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
>> 1532 if (!fs_info->super_copy || !fs_info->super_for_commit) {
>> 1533 error = -ENOMEM;
>> 1534 goto error_fs_info;
>> 1535 }
>> 1536
>> 1537 mutex_lock(&uuid_mutex);
>> 1538 error = btrfs_parse_device_options(data, mode, fs_type);
>> 1539 if (error) {
>> 1540 mutex_unlock(&uuid_mutex);
>> 1541 goto error_fs_info;
>> 1542 }
>> 1543
>> 1544 device = btrfs_scan_one_device(device_name, mode, fs_type);
>> 1545 if (IS_ERR(device)) {
>> 1546 mutex_unlock(&uuid_mutex);
>> 1547 error = PTR_ERR(device);
>> 1548 goto error_fs_info;
>> 1549 }
>> 1550
>> 1551 fs_devices = device->fs_devices;
>> 1552 fs_info->fs_devices = fs_devices;
>> 1553
>> 1554 error = btrfs_open_devices(fs_devices, mode, fs_type);
>> 1555 mutex_unlock(&uuid_mutex);
>> 1556 if (error)
>> 1557 goto error_fs_info;
>> 1558
>> 1559 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
>> 1560 error = -EACCES;
>> 1561 goto error_close_devices;
>> 1562 }
>> 1563
>> 1564 bdev = fs_devices->latest_bdev;
>> 1565 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
>> 1566 fs_info);
>> 1567 if (IS_ERR(s)) {
>> 1568 error = PTR_ERR(s);
>> 1569 goto error_close_devices;
>> 1570 }
>> 1571
>> 1572 if (s->s_root) {
>> 1573 btrfs_close_devices(fs_devices);
>> 1574 free_fs_info(fs_info);
>> 1575 if ((flags ^ s->s_flags) & SB_RDONLY)
>> 1576 error = -EBUSY;
>> 1577 } else {
>> 1578 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
>> 1579 btrfs_sb(s)->bdev_holder = fs_type;
>> 1580 if (!strstr(crc32c_impl(), "generic"))
>> 1581 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
>> 1582 error = btrfs_fill_super(s, fs_devices, data);
>> 1583 }
>> 1584 if (!error)
>> 1585 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
>> 1586 security_free_mnt_opts(&new_sec_opts);
>> 1587 if (error) {
>> 1588 deactivate_locked_super(s);
>> 1589 return ERR_PTR(error);
>> 1590 }
>> 1591
>> 1592 return dget(s->s_root);
>> 1593
>> 1594 error_close_devices:
>> 1595 btrfs_close_devices(fs_devices);
>> 1596 error_fs_info:
>> 1597 free_fs_info(fs_info);
>> 1598 error_sec_opts:
>> 1599 security_free_mnt_opts(&new_sec_opts);
>> 1600 return ERR_PTR(error);
>> 1601 }
>> 1602
>> 1603 /*
>> 1604 * Mount function which is called by VFS layer.
>> 1605 *
>> 1606 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
>> 1607 * which needs vfsmount* of device's root (/). This means device's root has to
>> 1608 * be mounted internally in any case.
>> 1609 *
>> 1610 * Operation flow:
>> 1611 * 1. Parse subvol id related options for later use in mount_subvol().
>> 1612 *
>> 1613 * 2. Mount device's root (/) by calling vfs_kern_mount().
>> 1614 *
>> 1615 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
>> 1616 * first place. In order to avoid calling btrfs_mount() again, we use
>> 1617 * different file_system_type which is not registered to VFS by
>> 1618 * register_filesystem() (btrfs_root_fs_type). As a result,
>> 1619 * btrfs_mount_root() is called. The return value will be used by
>> 1620 * mount_subtree() in mount_subvol().
>> 1621 *
>> 1622 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
>> 1623 * "btrfs subvolume set-default", mount_subvol() is called always.
>> 1624 */
>> 1625 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
>> 1626 const char *device_name, void *data)
>> 1627 {
>> 1628 struct vfsmount *mnt_root;
>> 1629 struct dentry *root;
>> 1630 char *subvol_name = NULL;
>> 1631 u64 subvol_objectid = 0;
>> 1632 int error = 0;
>> 1633
>> 1634 error = btrfs_parse_subvol_options(data, &subvol_name,
>> 1635 &subvol_objectid);
>> 1636 if (error) {
>> 1637 kfree(subvol_name);
>> 1638 return ERR_PTR(error);
>> 1639 }
>> 1640
>> 1641 /* mount device's root (/) */
>> 1642 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
>> 1643 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
>> 1644 if (flags & SB_RDONLY) {
>> 1645 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
>> 1646 flags & ~SB_RDONLY, device_name, data);
>> 1647 } else {
>> 1648 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
>> 1649 flags | SB_RDONLY, device_name, data);
>> 1650 if (IS_ERR(mnt_root)) {
>> 1651 root = ERR_CAST(mnt_root);
>> 1652 kfree(subvol_name);
>> 1653 goto out;
>> 1654 }
>> 1655
>> 1656 down_write(&mnt_root->mnt_sb->s_umount);
>> 1657 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
>> 1658 up_write(&mnt_root->mnt_sb->s_umount);
>> 1659 if (error < 0) {
>> 1660 root = ERR_PTR(error);
>> 1661 mntput(mnt_root);
>> 1662 kfree(subvol_name);
>> 1663 goto out;
>> 1664 }
>> 1665 }
>> 1666 }
>> 1667 if (IS_ERR(mnt_root)) {
>> 1668 root = ERR_CAST(mnt_root);
>> 1669 kfree(subvol_name);
>> 1670 goto out;
>> 1671 }
>> 1672
>> 1673 /* mount_subvol() will free subvol_name and mnt_root */
>> 1674 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
>> 1675
>> 1676 out:
>> 1677 return root;
>> 1678 }
>> 1679
1226 static void btrfs_resize_thread_pool(struct b 1680 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1227 u32 new_ 1681 u32 new_pool_size, u32 old_pool_size)
1228 { 1682 {
1229 if (new_pool_size == old_pool_size) 1683 if (new_pool_size == old_pool_size)
1230 return; 1684 return;
1231 1685
1232 fs_info->thread_pool_size = new_pool_ 1686 fs_info->thread_pool_size = new_pool_size;
1233 1687
1234 btrfs_info(fs_info, "resize thread po 1688 btrfs_info(fs_info, "resize thread pool %d -> %d",
1235 old_pool_size, new_pool_size); 1689 old_pool_size, new_pool_size);
1236 1690
1237 btrfs_workqueue_set_max(fs_info->work 1691 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1238 btrfs_workqueue_set_max(fs_info->dela 1692 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1239 btrfs_workqueue_set_max(fs_info->cach 1693 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1240 workqueue_set_max_active(fs_info->end !! 1694 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1241 workqueue_set_max_active(fs_info->end !! 1695 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
>> 1696 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
>> 1697 new_pool_size);
1242 btrfs_workqueue_set_max(fs_info->endi 1698 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1243 btrfs_workqueue_set_max(fs_info->endi 1699 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1244 btrfs_workqueue_set_max(fs_info->dela 1700 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
>> 1701 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
>> 1702 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
>> 1703 new_pool_size);
>> 1704 }
>> 1705
>> 1706 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
>> 1707 {
>> 1708 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1245 } 1709 }
1246 1710
1247 static inline void btrfs_remount_begin(struct 1711 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1248 unsign !! 1712 unsigned long old_opts, int flags)
1249 { 1713 {
1250 if (btrfs_raw_test_opt(old_opts, AUTO 1714 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1251 (!btrfs_raw_test_opt(fs_info->mou 1715 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1252 (flags & SB_RDONLY))) { 1716 (flags & SB_RDONLY))) {
1253 /* wait for any defraggers to 1717 /* wait for any defraggers to finish */
1254 wait_event(fs_info->transacti 1718 wait_event(fs_info->transaction_wait,
1255 (atomic_read(&fs_i 1719 (atomic_read(&fs_info->defrag_running) == 0));
1256 if (flags & SB_RDONLY) 1720 if (flags & SB_RDONLY)
1257 sync_filesystem(fs_in 1721 sync_filesystem(fs_info->sb);
1258 } 1722 }
1259 } 1723 }
1260 1724
1261 static inline void btrfs_remount_cleanup(stru 1725 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1262 unsi !! 1726 unsigned long old_opts)
1263 { 1727 {
1264 const bool cache_opt = btrfs_test_opt <<
1265 <<
1266 /* 1728 /*
1267 * We need to cleanup all defragable 1729 * We need to cleanup all defragable inodes if the autodefragment is
1268 * close or the filesystem is read on 1730 * close or the filesystem is read only.
1269 */ 1731 */
1270 if (btrfs_raw_test_opt(old_opts, AUTO 1732 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1271 (!btrfs_raw_test_opt(fs_info->mou 1733 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1272 btrfs_cleanup_defrag_inodes(f 1734 btrfs_cleanup_defrag_inodes(fs_info);
1273 } 1735 }
1274 1736
1275 /* If we toggled discard async */ 1737 /* If we toggled discard async */
1276 if (!btrfs_raw_test_opt(old_opts, DIS 1738 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1277 btrfs_test_opt(fs_info, DISCARD_A 1739 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1278 btrfs_discard_resume(fs_info) 1740 btrfs_discard_resume(fs_info);
1279 else if (btrfs_raw_test_opt(old_opts, 1741 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1280 !btrfs_test_opt(fs_info, DIS 1742 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1281 btrfs_discard_cleanup(fs_info 1743 btrfs_discard_cleanup(fs_info);
1282 1744
1283 /* If we toggled space cache */ !! 1745 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1284 if (cache_opt != btrfs_free_space_cac <<
1285 btrfs_set_free_space_cache_v1 <<
1286 } 1746 }
1287 1747
1288 static int btrfs_remount_rw(struct btrfs_fs_i !! 1748 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1289 { 1749 {
>> 1750 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 1751 struct btrfs_root *root = fs_info->tree_root;
>> 1752 unsigned old_flags = sb->s_flags;
>> 1753 unsigned long old_opts = fs_info->mount_opt;
>> 1754 unsigned long old_compress_type = fs_info->compress_type;
>> 1755 u64 old_max_inline = fs_info->max_inline;
>> 1756 u32 old_thread_pool_size = fs_info->thread_pool_size;
>> 1757 u32 old_metadata_ratio = fs_info->metadata_ratio;
1290 int ret; 1758 int ret;
1291 1759
1292 if (BTRFS_FS_ERROR(fs_info)) { !! 1760 sync_filesystem(sb);
1293 btrfs_err(fs_info, !! 1761 btrfs_remount_prepare(fs_info);
1294 "remounting read-wr <<
1295 return -EINVAL; <<
1296 } <<
1297 <<
1298 if (fs_info->fs_devices->rw_devices = <<
1299 return -EACCES; <<
1300 1762
1301 if (!btrfs_check_rw_degradable(fs_inf !! 1763 if (data) {
1302 btrfs_warn(fs_info, !! 1764 void *new_sec_opts = NULL;
1303 "too many missing <<
1304 return -EACCES; <<
1305 } <<
1306 1765
1307 if (btrfs_super_log_root(fs_info->sup !! 1766 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1308 btrfs_warn(fs_info, !! 1767 if (!ret)
1309 "mount required to !! 1768 ret = security_sb_remount(sb, new_sec_opts);
1310 return -EINVAL; !! 1769 security_free_mnt_opts(&new_sec_opts);
>> 1770 if (ret)
>> 1771 goto restore;
1311 } 1772 }
1312 1773
1313 /* !! 1774 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) 1775 if (ret)
1320 return ret; !! 1776 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 <<
1344 btrfs_discard_cleanup(fs_info); <<
1345 <<
1346 /* Wait for the uuid_scan task to fin <<
1347 down(&fs_info->uuid_tree_rescan_sem); <<
1348 /* Avoid complains from lockdep et al <<
1349 up(&fs_info->uuid_tree_rescan_sem); <<
1350 <<
1351 btrfs_set_sb_rdonly(fs_info->sb); <<
1352 <<
1353 /* <<
1354 * Setting SB_RDONLY will put the cle <<
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 1777
1361 /* !! 1778 btrfs_remount_begin(fs_info, old_opts, *flags);
1362 * The cleaner task could be already !! 1779 btrfs_resize_thread_pool(fs_info,
1363 * BTRFS_FS_STATE_RO (and SB_RDONLY i !! 1780 fs_info->thread_pool_size, old_thread_pool_size);
1364 * sure that after we finish the remo <<
1365 * btrfs_commit_super(), the cleaner <<
1366 * - either because it was dropping a <<
1367 * or deleting an unused block grou <<
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 1781
1373 /* !! 1782 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1374 * We've set the superblock to RO mod !! 1783 goto out;
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 1784
1382 btrfs_dev_replace_suspend_for_unmount !! 1785 if (*flags & SB_RDONLY) {
1383 btrfs_scrub_cancel(fs_info); !! 1786 /*
1384 btrfs_pause_balance(fs_info); !! 1787 * this also happens on 'umount -rf' or on shutdown, when
>> 1788 * the filesystem is busy.
>> 1789 */
>> 1790 cancel_work_sync(&fs_info->async_reclaim_work);
1385 1791
1386 /* !! 1792 btrfs_discard_cleanup(fs_info);
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 1793
1394 return btrfs_commit_super(fs_info); !! 1794 /* wait for the uuid_scan task to finish */
1395 } !! 1795 down(&fs_info->uuid_tree_rescan_sem);
>> 1796 /* avoid complains from lockdep et al. */
>> 1797 up(&fs_info->uuid_tree_rescan_sem);
1396 1798
1397 static void btrfs_ctx_to_info(struct btrfs_fs !! 1799 sb->s_flags |= SB_RDONLY;
1398 { <<
1399 fs_info->max_inline = ctx->max_inline <<
1400 fs_info->commit_interval = ctx->commi <<
1401 fs_info->metadata_ratio = ctx->metada <<
1402 fs_info->thread_pool_size = ctx->thre <<
1403 fs_info->mount_opt = ctx->mount_opt; <<
1404 fs_info->compress_type = ctx->compres <<
1405 fs_info->compress_level = ctx->compre <<
1406 } <<
1407 <<
1408 static void btrfs_info_to_ctx(struct btrfs_fs <<
1409 { <<
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 1800
1484 if (info->max_inline != BTRFS_DEFAULT !! 1801 /*
1485 btrfs_info(info, "max_inline !! 1802 * Setting SB_RDONLY will put the cleaner thread to
1486 } !! 1803 * sleep at the next loop if it's already active.
>> 1804 * If it's already asleep, we'll leave unused block
>> 1805 * groups on disk until we're mounted read-write again
>> 1806 * unless we clean them up here.
>> 1807 */
>> 1808 btrfs_delete_unused_bgs(fs_info);
1487 1809
1488 static int btrfs_reconfigure(struct fs_contex !! 1810 btrfs_dev_replace_suspend_for_unmount(fs_info);
1489 { !! 1811 btrfs_scrub_cancel(fs_info);
1490 struct super_block *sb = fc->root->d_ !! 1812 btrfs_pause_balance(fs_info);
1491 struct btrfs_fs_info *fs_info = btrfs !! 1813
1492 struct btrfs_fs_context *ctx = fc->fs !! 1814 ret = btrfs_commit_super(fs_info);
1493 struct btrfs_fs_context old_ctx; !! 1815 if (ret)
1494 int ret = 0; !! 1816 goto restore;
1495 bool mount_reconfigure = (fc->s_fs_in !! 1817 } else {
>> 1818 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
>> 1819 btrfs_err(fs_info,
>> 1820 "Remounting read-write after error is not allowed");
>> 1821 ret = -EINVAL;
>> 1822 goto restore;
>> 1823 }
>> 1824 if (fs_info->fs_devices->rw_devices == 0) {
>> 1825 ret = -EACCES;
>> 1826 goto restore;
>> 1827 }
1496 1828
1497 btrfs_info_to_ctx(fs_info, &old_ctx); !! 1829 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
>> 1830 btrfs_warn(fs_info,
>> 1831 "too many missing devices, writable remount is not allowed");
>> 1832 ret = -EACCES;
>> 1833 goto restore;
>> 1834 }
1498 1835
1499 /* !! 1836 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1500 * This is our "bind mount" trick, we !! 1837 btrfs_warn(fs_info,
1501 * anything other than mount a differ !! 1838 "mount required to replay tree-log, cannot remount read-write");
1502 * all of the mount options should be !! 1839 ret = -EINVAL;
1503 */ !! 1840 goto restore;
1504 if (mount_reconfigure) !! 1841 }
1505 ctx->mount_opt = old_ctx.moun <<
1506 1842
1507 sync_filesystem(sb); !! 1843 ret = btrfs_cleanup_fs_roots(fs_info);
1508 set_bit(BTRFS_FS_STATE_REMOUNTING, &f !! 1844 if (ret)
>> 1845 goto restore;
1509 1846
1510 if (!btrfs_check_options(fs_info, &ct !! 1847 /* recover relocation */
1511 return -EINVAL; !! 1848 mutex_lock(&fs_info->cleaner_mutex);
>> 1849 ret = btrfs_recover_relocation(root);
>> 1850 mutex_unlock(&fs_info->cleaner_mutex);
>> 1851 if (ret)
>> 1852 goto restore;
1512 1853
1513 ret = btrfs_check_features(fs_info, ! !! 1854 ret = btrfs_resume_balance_async(fs_info);
1514 if (ret < 0) !! 1855 if (ret)
1515 return ret; !! 1856 goto restore;
1516 1857
1517 btrfs_ctx_to_info(fs_info, ctx); !! 1858 ret = btrfs_resume_dev_replace_async(fs_info);
1518 btrfs_remount_begin(fs_info, old_ctx. !! 1859 if (ret) {
1519 btrfs_resize_thread_pool(fs_info, fs_ !! 1860 btrfs_warn(fs_info, "failed to resume dev_replace");
1520 old_ctx.thre !! 1861 goto restore;
1521 <<
1522 if ((bool)btrfs_test_opt(fs_info, FRE <<
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 } 1862 }
1536 } <<
1537 1863
1538 ret = 0; !! 1864 btrfs_qgroup_rescan_resume(fs_info);
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 1865
1546 /* !! 1866 if (!fs_info->uuid_root) {
1547 * If we set the mask during the para !! 1867 btrfs_info(fs_info, "creating UUID tree");
1548 * remount. Here we can set the mask !! 1868 ret = btrfs_create_uuid_tree(fs_info);
1549 * appropriately. !! 1869 if (ret) {
1550 */ !! 1870 btrfs_warn(fs_info,
1551 if ((fc->sb_flags & SB_POSIXACL) != ( !! 1871 "failed to create the UUID tree %d",
1552 fc->sb_flags_mask |= SB_POSIX !! 1872 ret);
>> 1873 goto restore;
>> 1874 }
>> 1875 }
>> 1876 sb->s_flags &= ~SB_RDONLY;
1553 1877
1554 btrfs_emit_options(fs_info, &old_ctx) !! 1878 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
>> 1879 }
>> 1880 out:
1555 wake_up_process(fs_info->transaction_ 1881 wake_up_process(fs_info->transaction_kthread);
1556 btrfs_remount_cleanup(fs_info, old_ct !! 1882 btrfs_remount_cleanup(fs_info, old_opts);
1557 btrfs_clear_oneshot_options(fs_info); <<
1558 clear_bit(BTRFS_FS_STATE_REMOUNTING, <<
1559 <<
1560 return 0; 1883 return 0;
>> 1884
1561 restore: 1885 restore:
1562 btrfs_ctx_to_info(fs_info, &old_ctx); !! 1886 /* We've hit an error - don't reset SB_RDONLY */
1563 btrfs_remount_cleanup(fs_info, old_ct !! 1887 if (sb_rdonly(sb))
1564 clear_bit(BTRFS_FS_STATE_REMOUNTING, !! 1888 old_flags |= SB_RDONLY;
>> 1889 sb->s_flags = old_flags;
>> 1890 fs_info->mount_opt = old_opts;
>> 1891 fs_info->compress_type = old_compress_type;
>> 1892 fs_info->max_inline = old_max_inline;
>> 1893 btrfs_resize_thread_pool(fs_info,
>> 1894 old_thread_pool_size, fs_info->thread_pool_size);
>> 1895 fs_info->metadata_ratio = old_metadata_ratio;
>> 1896 btrfs_remount_cleanup(fs_info, old_opts);
1565 return ret; 1897 return ret;
1566 } 1898 }
1567 1899
1568 /* Used to sort the devices by max_avail(desc 1900 /* Used to sort the devices by max_avail(descending sort) */
1569 static int btrfs_cmp_device_free_bytes(const !! 1901 static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
>> 1902 const void *dev_info2)
1570 { 1903 {
1571 const struct btrfs_device_info *dev_i !! 1904 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1572 const struct btrfs_device_info *dev_i !! 1905 ((struct btrfs_device_info *)dev_info2)->max_avail)
1573 <<
1574 if (dev_info1->max_avail > dev_info2- <<
1575 return -1; 1906 return -1;
1576 else if (dev_info1->max_avail < dev_i !! 1907 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
>> 1908 ((struct btrfs_device_info *)dev_info2)->max_avail)
1577 return 1; 1909 return 1;
>> 1910 else
1578 return 0; 1911 return 0;
1579 } 1912 }
1580 1913
1581 /* 1914 /*
1582 * sort the devices by max_avail, in which ma 1915 * sort the devices by max_avail, in which max free extent size of each device
1583 * is stored.(Descending Sort) 1916 * is stored.(Descending Sort)
1584 */ 1917 */
1585 static inline void btrfs_descending_sort_devi 1918 static inline void btrfs_descending_sort_devices(
1586 struc 1919 struct btrfs_device_info *devices,
1587 size_ 1920 size_t nr_devices)
1588 { 1921 {
1589 sort(devices, nr_devices, sizeof(stru 1922 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1590 btrfs_cmp_device_free_bytes, NUL 1923 btrfs_cmp_device_free_bytes, NULL);
1591 } 1924 }
1592 1925
1593 /* 1926 /*
1594 * The helper to calc the free space on the d 1927 * The helper to calc the free space on the devices that can be used to store
1595 * file data. 1928 * file data.
1596 */ 1929 */
1597 static inline int btrfs_calc_avail_data_space 1930 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1598 1931 u64 *free_bytes)
1599 { 1932 {
1600 struct btrfs_device_info *devices_inf 1933 struct btrfs_device_info *devices_info;
1601 struct btrfs_fs_devices *fs_devices = 1934 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1602 struct btrfs_device *device; 1935 struct btrfs_device *device;
1603 u64 type; 1936 u64 type;
1604 u64 avail_space; 1937 u64 avail_space;
1605 u64 min_stripe_size; 1938 u64 min_stripe_size;
1606 int num_stripes = 1; 1939 int num_stripes = 1;
1607 int i = 0, nr_devices; 1940 int i = 0, nr_devices;
1608 const struct btrfs_raid_attr *rattr; 1941 const struct btrfs_raid_attr *rattr;
1609 1942
1610 /* 1943 /*
1611 * We aren't under the device list lo 1944 * We aren't under the device list lock, so this is racy-ish, but good
1612 * enough for our purposes. 1945 * enough for our purposes.
1613 */ 1946 */
1614 nr_devices = fs_info->fs_devices->ope 1947 nr_devices = fs_info->fs_devices->open_devices;
1615 if (!nr_devices) { 1948 if (!nr_devices) {
1616 smp_mb(); 1949 smp_mb();
1617 nr_devices = fs_info->fs_devi 1950 nr_devices = fs_info->fs_devices->open_devices;
1618 ASSERT(nr_devices); 1951 ASSERT(nr_devices);
1619 if (!nr_devices) { 1952 if (!nr_devices) {
1620 *free_bytes = 0; 1953 *free_bytes = 0;
1621 return 0; 1954 return 0;
1622 } 1955 }
1623 } 1956 }
1624 1957
1625 devices_info = kmalloc_array(nr_devic 1958 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1626 GFP_KERNEL); 1959 GFP_KERNEL);
1627 if (!devices_info) 1960 if (!devices_info)
1628 return -ENOMEM; 1961 return -ENOMEM;
1629 1962
1630 /* calc min stripe number for data sp 1963 /* calc min stripe number for data space allocation */
1631 type = btrfs_data_alloc_profile(fs_in 1964 type = btrfs_data_alloc_profile(fs_info);
1632 rattr = &btrfs_raid_array[btrfs_bg_fl 1965 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1633 1966
1634 if (type & BTRFS_BLOCK_GROUP_RAID0) 1967 if (type & BTRFS_BLOCK_GROUP_RAID0)
1635 num_stripes = nr_devices; 1968 num_stripes = nr_devices;
1636 else if (type & BTRFS_BLOCK_GROUP_RAI !! 1969 else if (type & BTRFS_BLOCK_GROUP_RAID1)
1637 num_stripes = rattr->ncopies; !! 1970 num_stripes = 2;
>> 1971 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
>> 1972 num_stripes = 3;
>> 1973 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
>> 1974 num_stripes = 4;
1638 else if (type & BTRFS_BLOCK_GROUP_RAI 1975 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1639 num_stripes = 4; 1976 num_stripes = 4;
1640 1977
1641 /* Adjust for more than 1 stripe per 1978 /* Adjust for more than 1 stripe per device */
1642 min_stripe_size = rattr->dev_stripes 1979 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1643 1980
1644 rcu_read_lock(); 1981 rcu_read_lock();
1645 list_for_each_entry_rcu(device, &fs_d 1982 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1646 if (!test_bit(BTRFS_DEV_STATE 1983 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1647 1984 &device->dev_state) ||
1648 !device->bdev || 1985 !device->bdev ||
1649 test_bit(BTRFS_DEV_STATE_ 1986 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1650 continue; 1987 continue;
1651 1988
1652 if (i >= nr_devices) 1989 if (i >= nr_devices)
1653 break; 1990 break;
1654 1991
1655 avail_space = device->total_b 1992 avail_space = device->total_bytes - device->bytes_used;
1656 1993
1657 /* align with stripe_len */ 1994 /* align with stripe_len */
1658 avail_space = rounddown(avail 1995 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1659 1996
1660 /* 1997 /*
1661 * Ensure we have at least mi !! 1998 * In order to avoid overwriting the superblock on the drive,
1662 * reserved space on the devi !! 1999 * btrfs starts at an offset of at least 1MB when doing chunk
>> 2000 * allocation.
>> 2001 *
>> 2002 * This ensures we have at least min_stripe_size free space
>> 2003 * after excluding 1MB.
1663 */ 2004 */
1664 if (avail_space <= BTRFS_DEVI !! 2005 if (avail_space <= SZ_1M + min_stripe_size)
1665 continue; 2006 continue;
1666 2007
1667 avail_space -= BTRFS_DEVICE_R !! 2008 avail_space -= SZ_1M;
1668 2009
1669 devices_info[i].dev = device; 2010 devices_info[i].dev = device;
1670 devices_info[i].max_avail = a 2011 devices_info[i].max_avail = avail_space;
1671 2012
1672 i++; 2013 i++;
1673 } 2014 }
1674 rcu_read_unlock(); 2015 rcu_read_unlock();
1675 2016
1676 nr_devices = i; 2017 nr_devices = i;
1677 2018
1678 btrfs_descending_sort_devices(devices 2019 btrfs_descending_sort_devices(devices_info, nr_devices);
1679 2020
1680 i = nr_devices - 1; 2021 i = nr_devices - 1;
1681 avail_space = 0; 2022 avail_space = 0;
1682 while (nr_devices >= rattr->devs_min) 2023 while (nr_devices >= rattr->devs_min) {
1683 num_stripes = min(num_stripes 2024 num_stripes = min(num_stripes, nr_devices);
1684 2025
1685 if (devices_info[i].max_avail 2026 if (devices_info[i].max_avail >= min_stripe_size) {
1686 int j; 2027 int j;
1687 u64 alloc_size; 2028 u64 alloc_size;
1688 2029
1689 avail_space += device 2030 avail_space += devices_info[i].max_avail * num_stripes;
1690 alloc_size = devices_ 2031 alloc_size = devices_info[i].max_avail;
1691 for (j = i + 1 - num_ 2032 for (j = i + 1 - num_stripes; j <= i; j++)
1692 devices_info[ 2033 devices_info[j].max_avail -= alloc_size;
1693 } 2034 }
1694 i--; 2035 i--;
1695 nr_devices--; 2036 nr_devices--;
1696 } 2037 }
1697 2038
1698 kfree(devices_info); 2039 kfree(devices_info);
1699 *free_bytes = avail_space; 2040 *free_bytes = avail_space;
1700 return 0; 2041 return 0;
1701 } 2042 }
1702 2043
1703 /* 2044 /*
1704 * Calculate numbers for 'df', pessimistic in 2045 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1705 * 2046 *
1706 * If there's a redundant raid level at DATA 2047 * If there's a redundant raid level at DATA block groups, use the respective
1707 * multiplier to scale the sizes. 2048 * multiplier to scale the sizes.
1708 * 2049 *
1709 * Unused device space usage is based on simu 2050 * Unused device space usage is based on simulating the chunk allocator
1710 * algorithm that respects the device sizes a 2051 * algorithm that respects the device sizes and order of allocations. This is
1711 * a close approximation of the actual use bu 2052 * a close approximation of the actual use but there are other factors that may
1712 * change the result (like a new metadata chu 2053 * change the result (like a new metadata chunk).
1713 * 2054 *
1714 * If metadata is exhausted, f_bavail will be 2055 * If metadata is exhausted, f_bavail will be 0.
1715 */ 2056 */
1716 static int btrfs_statfs(struct dentry *dentry 2057 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1717 { 2058 {
1718 struct btrfs_fs_info *fs_info = btrfs 2059 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1719 struct btrfs_super_block *disk_super 2060 struct btrfs_super_block *disk_super = fs_info->super_copy;
1720 struct btrfs_space_info *found; 2061 struct btrfs_space_info *found;
1721 u64 total_used = 0; 2062 u64 total_used = 0;
1722 u64 total_free_data = 0; 2063 u64 total_free_data = 0;
1723 u64 total_free_meta = 0; 2064 u64 total_free_meta = 0;
1724 u32 bits = fs_info->sectorsize_bits; !! 2065 int bits = dentry->d_sb->s_blocksize_bits;
1725 __be32 *fsid = (__be32 *)fs_info->fs_ 2066 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1726 unsigned factor = 1; 2067 unsigned factor = 1;
1727 struct btrfs_block_rsv *block_rsv = & 2068 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1728 int ret; 2069 int ret;
1729 u64 thresh = 0; 2070 u64 thresh = 0;
1730 int mixed = 0; 2071 int mixed = 0;
1731 2072
1732 list_for_each_entry(found, &fs_info-> !! 2073 rcu_read_lock();
>> 2074 list_for_each_entry_rcu(found, &fs_info->space_info, list) {
1733 if (found->flags & BTRFS_BLOC 2075 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1734 int i; 2076 int i;
1735 2077
1736 total_free_data += fo 2078 total_free_data += found->disk_total - found->disk_used;
1737 total_free_data -= 2079 total_free_data -=
1738 btrfs_account 2080 btrfs_account_ro_block_groups_free_space(found);
1739 2081
1740 for (i = 0; i < BTRFS 2082 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1741 if (!list_emp 2083 if (!list_empty(&found->block_groups[i]))
1742 facto 2084 factor = btrfs_bg_type_to_factor(
1743 2085 btrfs_raid_array[i].bg_flag);
1744 } 2086 }
1745 } 2087 }
1746 2088
1747 /* 2089 /*
1748 * Metadata in mixed block gr !! 2090 * Metadata in mixed block goup profiles are accounted in data
1749 */ 2091 */
1750 if (!mixed && found->flags & 2092 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1751 if (found->flags & BT 2093 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1752 mixed = 1; 2094 mixed = 1;
1753 else 2095 else
1754 total_free_me 2096 total_free_meta += found->disk_total -
1755 found 2097 found->disk_used;
1756 } 2098 }
1757 2099
1758 total_used += found->disk_use 2100 total_used += found->disk_used;
1759 } 2101 }
1760 2102
>> 2103 rcu_read_unlock();
>> 2104
1761 buf->f_blocks = div_u64(btrfs_super_t 2105 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1762 buf->f_blocks >>= bits; 2106 buf->f_blocks >>= bits;
1763 buf->f_bfree = buf->f_blocks - (div_u 2107 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1764 2108
1765 /* Account global block reserve as us 2109 /* Account global block reserve as used, it's in logical size already */
1766 spin_lock(&block_rsv->lock); 2110 spin_lock(&block_rsv->lock);
1767 /* Mixed block groups accounting is n 2111 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1768 if (buf->f_bfree >= block_rsv->size > 2112 if (buf->f_bfree >= block_rsv->size >> bits)
1769 buf->f_bfree -= block_rsv->si 2113 buf->f_bfree -= block_rsv->size >> bits;
1770 else 2114 else
1771 buf->f_bfree = 0; 2115 buf->f_bfree = 0;
1772 spin_unlock(&block_rsv->lock); 2116 spin_unlock(&block_rsv->lock);
1773 2117
1774 buf->f_bavail = div_u64(total_free_da 2118 buf->f_bavail = div_u64(total_free_data, factor);
1775 ret = btrfs_calc_avail_data_space(fs_ 2119 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1776 if (ret) 2120 if (ret)
1777 return ret; 2121 return ret;
1778 buf->f_bavail += div_u64(total_free_d 2122 buf->f_bavail += div_u64(total_free_data, factor);
1779 buf->f_bavail = buf->f_bavail >> bits 2123 buf->f_bavail = buf->f_bavail >> bits;
1780 2124
1781 /* 2125 /*
1782 * We calculate the remaining metadat 2126 * We calculate the remaining metadata space minus global reserve. If
1783 * this is (supposedly) smaller than 2127 * this is (supposedly) smaller than zero, there's no space. But this
1784 * does not hold in practice, the exh 2128 * does not hold in practice, the exhausted state happens where's still
1785 * some positive delta. So we apply s 2129 * some positive delta. So we apply some guesswork and compare the
1786 * delta to a 4M threshold. (Practic 2130 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1787 * 2131 *
1788 * We probably cannot calculate the e 2132 * We probably cannot calculate the exact threshold value because this
1789 * depends on the internal reservatio 2133 * depends on the internal reservations requested by various
1790 * operations, so some operations tha 2134 * operations, so some operations that consume a few metadata will
1791 * succeed even if the Avail is zero. 2135 * succeed even if the Avail is zero. But this is better than the other
1792 * way around. 2136 * way around.
1793 */ 2137 */
1794 thresh = SZ_4M; 2138 thresh = SZ_4M;
1795 2139
1796 /* 2140 /*
1797 * We only want to claim there's no a 2141 * We only want to claim there's no available space if we can no longer
1798 * allocate chunks for our metadata p 2142 * allocate chunks for our metadata profile and our global reserve will
1799 * not fit in the free metadata space 2143 * not fit in the free metadata space. If we aren't ->full then we
1800 * still can allocate chunks and thus 2144 * still can allocate chunks and thus are fine using the currently
1801 * calculated f_bavail. 2145 * calculated f_bavail.
1802 */ 2146 */
1803 if (!mixed && block_rsv->space_info-> 2147 if (!mixed && block_rsv->space_info->full &&
1804 (total_free_meta < thresh || tota !! 2148 total_free_meta - thresh < block_rsv->size)
1805 buf->f_bavail = 0; 2149 buf->f_bavail = 0;
1806 2150
1807 buf->f_type = BTRFS_SUPER_MAGIC; 2151 buf->f_type = BTRFS_SUPER_MAGIC;
1808 buf->f_bsize = fs_info->sectorsize; !! 2152 buf->f_bsize = dentry->d_sb->s_blocksize;
1809 buf->f_namelen = BTRFS_NAME_LEN; 2153 buf->f_namelen = BTRFS_NAME_LEN;
1810 2154
1811 /* We treat it as constant endianness 2155 /* We treat it as constant endianness (it doesn't matter _which_)
1812 because we want the fsid to come o 2156 because we want the fsid to come out the same whether mounted
1813 on a big-endian or little-endian h 2157 on a big-endian or little-endian host */
1814 buf->f_fsid.val[0] = be32_to_cpu(fsid 2158 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 2159 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1816 /* Mask in the root object ID too, to 2160 /* Mask in the root object ID too, to disambiguate subvols */
1817 buf->f_fsid.val[0] ^= btrfs_root_id(B !! 2161 buf->f_fsid.val[0] ^=
1818 buf->f_fsid.val[1] ^= btrfs_root_id(B !! 2162 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
1819 !! 2163 buf->f_fsid.val[1] ^=
1820 return 0; !! 2164 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
1821 } <<
1822 <<
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 2165
1906 fc->root = dget(sb->s_root); <<
1907 return 0; 2166 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 } 2167 }
2110 2168
2111 static void btrfs_kill_super(struct super_blo 2169 static void btrfs_kill_super(struct super_block *sb)
2112 { 2170 {
2113 struct btrfs_fs_info *fs_info = btrfs 2171 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2114 kill_anon_super(sb); 2172 kill_anon_super(sb);
2115 btrfs_free_fs_info(fs_info); !! 2173 free_fs_info(fs_info);
2116 } 2174 }
2117 2175
2118 static void btrfs_free_fs_context(struct fs_c !! 2176 static struct file_system_type btrfs_fs_type = {
2119 { !! 2177 .owner = THIS_MODULE,
2120 struct btrfs_fs_context *ctx = fc->fs !! 2178 .name = "btrfs",
2121 struct btrfs_fs_info *fs_info = fc->s !! 2179 .mount = btrfs_mount,
2122 !! 2180 .kill_sb = btrfs_kill_super,
2123 if (fs_info) !! 2181 .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 }; 2182 };
2158 2183
2159 static int btrfs_init_fs_context(struct fs_co !! 2184 static struct file_system_type btrfs_root_fs_type = {
2160 { !! 2185 .owner = THIS_MODULE,
2161 struct btrfs_fs_context *ctx; !! 2186 .name = "btrfs",
2162 !! 2187 .mount = btrfs_mount_root,
2163 ctx = kzalloc(sizeof(struct btrfs_fs_ !! 2188 .kill_sb = btrfs_kill_super,
2164 if (!ctx) !! 2189 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2165 return -ENOMEM; !! 2190 };
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 2191
2197 MODULE_ALIAS_FS("btrfs"); 2192 MODULE_ALIAS_FS("btrfs");
2198 2193
2199 static int btrfs_control_open(struct inode *i 2194 static int btrfs_control_open(struct inode *inode, struct file *file)
2200 { 2195 {
2201 /* 2196 /*
2202 * The control file's private_data is 2197 * The control file's private_data is used to hold the
2203 * transaction when it is started and 2198 * transaction when it is started and is used to keep
2204 * track of whether a transaction is 2199 * track of whether a transaction is already in progress.
2205 */ 2200 */
2206 file->private_data = NULL; 2201 file->private_data = NULL;
2207 return 0; 2202 return 0;
2208 } 2203 }
2209 2204
2210 /* 2205 /*
2211 * Used by /dev/btrfs-control for devices ioc !! 2206 * used by btrfsctl to scan devices when no FS is mounted
2212 */ 2207 */
2213 static long btrfs_control_ioctl(struct file * 2208 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2214 unsigned long 2209 unsigned long arg)
2215 { 2210 {
2216 struct btrfs_ioctl_vol_args *vol; 2211 struct btrfs_ioctl_vol_args *vol;
2217 struct btrfs_device *device = NULL; 2212 struct btrfs_device *device = NULL;
2218 dev_t devt = 0; <<
2219 int ret = -ENOTTY; 2213 int ret = -ENOTTY;
2220 2214
2221 if (!capable(CAP_SYS_ADMIN)) 2215 if (!capable(CAP_SYS_ADMIN))
2222 return -EPERM; 2216 return -EPERM;
2223 2217
2224 vol = memdup_user((void __user *)arg, 2218 vol = memdup_user((void __user *)arg, sizeof(*vol));
2225 if (IS_ERR(vol)) 2219 if (IS_ERR(vol))
2226 return PTR_ERR(vol); 2220 return PTR_ERR(vol);
2227 ret = btrfs_check_ioctl_vol_args_path !! 2221 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2228 if (ret < 0) <<
2229 goto out; <<
2230 2222
2231 switch (cmd) { 2223 switch (cmd) {
2232 case BTRFS_IOC_SCAN_DEV: 2224 case BTRFS_IOC_SCAN_DEV:
2233 mutex_lock(&uuid_mutex); 2225 mutex_lock(&uuid_mutex);
2234 /* !! 2226 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2235 * Scanning outside of mount !! 2227 &btrfs_root_fs_type);
2236 * into 0 error code. <<
2237 */ <<
2238 device = btrfs_scan_one_devic <<
2239 ret = PTR_ERR_OR_ZERO(device) 2228 ret = PTR_ERR_OR_ZERO(device);
2240 mutex_unlock(&uuid_mutex); 2229 mutex_unlock(&uuid_mutex);
2241 break; 2230 break;
2242 case BTRFS_IOC_FORGET_DEV: 2231 case BTRFS_IOC_FORGET_DEV:
2243 if (vol->name[0] != 0) { !! 2232 ret = btrfs_forget_devices(vol->name);
2244 ret = lookup_bdev(vol <<
2245 if (ret) <<
2246 break; <<
2247 } <<
2248 ret = btrfs_forget_devices(de <<
2249 break; 2233 break;
2250 case BTRFS_IOC_DEVICES_READY: 2234 case BTRFS_IOC_DEVICES_READY:
2251 mutex_lock(&uuid_mutex); 2235 mutex_lock(&uuid_mutex);
2252 /* !! 2236 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2253 * Scanning outside of mount !! 2237 &btrfs_root_fs_type);
2254 * into 0 error code. !! 2238 if (IS_ERR(device)) {
2255 */ <<
2256 device = btrfs_scan_one_devic <<
2257 if (IS_ERR_OR_NULL(device)) { <<
2258 mutex_unlock(&uuid_mu 2239 mutex_unlock(&uuid_mutex);
2259 ret = PTR_ERR(device) 2240 ret = PTR_ERR(device);
2260 break; 2241 break;
2261 } 2242 }
2262 ret = !(device->fs_devices->n 2243 ret = !(device->fs_devices->num_devices ==
2263 device->fs_devices->t 2244 device->fs_devices->total_devices);
2264 mutex_unlock(&uuid_mutex); 2245 mutex_unlock(&uuid_mutex);
2265 break; 2246 break;
2266 case BTRFS_IOC_GET_SUPPORTED_FEATURES 2247 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2267 ret = btrfs_ioctl_get_support 2248 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2268 break; 2249 break;
2269 } 2250 }
2270 2251
2271 out: <<
2272 kfree(vol); 2252 kfree(vol);
2273 return ret; 2253 return ret;
2274 } 2254 }
2275 2255
2276 static int btrfs_freeze(struct super_block *s 2256 static int btrfs_freeze(struct super_block *sb)
2277 { 2257 {
>> 2258 struct btrfs_trans_handle *trans;
2278 struct btrfs_fs_info *fs_info = btrfs 2259 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 2260 struct btrfs_root *root = fs_info->tree_root;
2279 2261
2280 set_bit(BTRFS_FS_FROZEN, &fs_info->fl 2262 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2281 /* 2263 /*
2282 * We don't need a barrier here, we'l 2264 * We don't need a barrier here, we'll wait for any transaction that
2283 * could be in progress on other thre 2265 * could be in progress on other threads (and do delayed iputs that
2284 * we want to avoid on a frozen files 2266 * we want to avoid on a frozen filesystem), or do the commit
2285 * ourselves. 2267 * ourselves.
2286 */ 2268 */
2287 return btrfs_commit_current_transacti !! 2269 trans = btrfs_attach_transaction_barrier(root);
2288 } !! 2270 if (IS_ERR(trans)) {
2289 !! 2271 /* no transaction, don't bother */
2290 static int check_dev_super(struct btrfs_devic !! 2272 if (PTR_ERR(trans) == -ENOENT)
2291 { !! 2273 return 0;
2292 struct btrfs_fs_info *fs_info = dev-> !! 2274 return PTR_ERR(trans);
2293 struct btrfs_super_block *sb; <<
2294 u64 last_trans; <<
2295 u16 csum_type; <<
2296 int ret = 0; <<
2297 <<
2298 /* This should be called with fs stil <<
2299 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_ <<
2300 <<
2301 /* Missing dev, no need to check. */ <<
2302 if (!dev->bdev) <<
2303 return 0; <<
2304 <<
2305 /* Only need to check the primary sup <<
2306 sb = btrfs_read_dev_one_super(dev->bd <<
2307 if (IS_ERR(sb)) <<
2308 return PTR_ERR(sb); <<
2309 <<
2310 /* Verify the checksum. */ <<
2311 csum_type = btrfs_super_csum_type(sb) <<
2312 if (csum_type != btrfs_super_csum_typ <<
2313 btrfs_err(fs_info, "csum type <<
2314 csum_type, btrfs_su <<
2315 ret = -EUCLEAN; <<
2316 goto out; <<
2317 } <<
2318 <<
2319 if (btrfs_check_super_csum(fs_info, s <<
2320 btrfs_err(fs_info, "csum for <<
2321 ret = -EUCLEAN; <<
2322 goto out; <<
2323 } <<
2324 <<
2325 /* Btrfs_validate_super() includes fs <<
2326 ret = btrfs_validate_super(fs_info, s <<
2327 if (ret < 0) <<
2328 goto out; <<
2329 <<
2330 last_trans = btrfs_get_last_trans_com <<
2331 if (btrfs_super_generation(sb) != las <<
2332 btrfs_err(fs_info, "transid m <<
2333 btrfs_super_generat <<
2334 ret = -EUCLEAN; <<
2335 goto out; <<
2336 } 2275 }
2337 out: !! 2276 return btrfs_commit_transaction(trans);
2338 btrfs_release_disk_super(sb); <<
2339 return ret; <<
2340 } 2277 }
2341 2278
2342 static int btrfs_unfreeze(struct super_block 2279 static int btrfs_unfreeze(struct super_block *sb)
2343 { 2280 {
2344 struct btrfs_fs_info *fs_info = btrfs 2281 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2345 struct btrfs_device *device; <<
2346 int ret = 0; <<
2347 2282
2348 /* <<
2349 * Make sure the fs is not changed by <<
2350 * modified by other OS). <<
2351 * If we found anything wrong, we mar <<
2352 * <<
2353 * And since the fs is frozen, no one <<
2354 * we don't need to hold device_list_ <<
2355 */ <<
2356 list_for_each_entry(device, &fs_info- <<
2357 ret = check_dev_super(device) <<
2358 if (ret < 0) { <<
2359 btrfs_handle_fs_error <<
2360 "super block <<
2361 device->devid <<
2362 break; <<
2363 } <<
2364 } <<
2365 clear_bit(BTRFS_FS_FROZEN, &fs_info-> 2283 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2366 <<
2367 /* <<
2368 * We still return 0, to allow VFS la <<
2369 * above checks failed. Since the fs <<
2370 * safe to continue, without causing <<
2371 */ <<
2372 return 0; 2284 return 0;
2373 } 2285 }
2374 2286
2375 static int btrfs_show_devname(struct seq_file 2287 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2376 { 2288 {
2377 struct btrfs_fs_info *fs_info = btrfs 2289 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
>> 2290 struct btrfs_fs_devices *cur_devices;
>> 2291 struct btrfs_device *dev, *first_dev = NULL;
>> 2292 struct list_head *head;
2378 2293
2379 /* 2294 /*
2380 * There should be always a valid poi !! 2295 * Lightweight locking of the devices. We should not need
2381 * for a short moment in case it's be !! 2296 * device_list_mutex here as we only read the device data and the list
2382 * the end of RCU grace period. !! 2297 * is protected by RCU. Even if a device is deleted during the list
>> 2298 * traversals, we'll get valid data, the freeing callback will wait at
>> 2299 * least until the rcu_read_unlock.
2383 */ 2300 */
2384 rcu_read_lock(); 2301 rcu_read_lock();
2385 seq_escape(m, btrfs_dev_name(fs_info- !! 2302 cur_devices = fs_info->fs_devices;
2386 rcu_read_unlock(); !! 2303 while (cur_devices) {
2387 !! 2304 head = &cur_devices->devices;
2388 return 0; !! 2305 list_for_each_entry_rcu(dev, head, dev_list) {
2389 } !! 2306 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2390 !! 2307 continue;
2391 static long btrfs_nr_cached_objects(struct su !! 2308 if (!dev->name)
2392 { !! 2309 continue;
2393 struct btrfs_fs_info *fs_info = btrfs !! 2310 if (!first_dev || dev->devid < first_dev->devid)
2394 const s64 nr = percpu_counter_sum_pos !! 2311 first_dev = dev;
2395 !! 2312 }
2396 trace_btrfs_extent_map_shrinker_count !! 2313 cur_devices = cur_devices->seed;
>> 2314 }
2397 2315
2398 /* !! 2316 if (first_dev)
2399 * Only report the real number for DE !! 2317 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2400 * serious performance degradation ca !! 2318 else
2401 */ !! 2319 WARN_ON(1);
2402 if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) !! 2320 rcu_read_unlock();
2403 return nr; <<
2404 return 0; 2321 return 0;
2405 } 2322 }
2406 2323
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 2324 static const struct super_operations btrfs_super_ops = {
2425 .drop_inode = btrfs_drop_inode, 2325 .drop_inode = btrfs_drop_inode,
2426 .evict_inode = btrfs_evict_inode, 2326 .evict_inode = btrfs_evict_inode,
2427 .put_super = btrfs_put_super, 2327 .put_super = btrfs_put_super,
2428 .sync_fs = btrfs_sync_fs, 2328 .sync_fs = btrfs_sync_fs,
2429 .show_options = btrfs_show_options, 2329 .show_options = btrfs_show_options,
2430 .show_devname = btrfs_show_devname, 2330 .show_devname = btrfs_show_devname,
2431 .alloc_inode = btrfs_alloc_inode, 2331 .alloc_inode = btrfs_alloc_inode,
2432 .destroy_inode = btrfs_destroy_inode 2332 .destroy_inode = btrfs_destroy_inode,
2433 .free_inode = btrfs_free_inode, 2333 .free_inode = btrfs_free_inode,
2434 .statfs = btrfs_statfs, 2334 .statfs = btrfs_statfs,
>> 2335 .remount_fs = btrfs_remount,
2435 .freeze_fs = btrfs_freeze, 2336 .freeze_fs = btrfs_freeze,
2436 .unfreeze_fs = btrfs_unfreeze, 2337 .unfreeze_fs = btrfs_unfreeze,
2437 .nr_cached_objects = btrfs_nr_cached_ <<
2438 .free_cached_objects = btrfs_free_cac <<
2439 }; 2338 };
2440 2339
2441 static const struct file_operations btrfs_ctl 2340 static const struct file_operations btrfs_ctl_fops = {
2442 .open = btrfs_control_open, 2341 .open = btrfs_control_open,
2443 .unlocked_ioctl = btrfs_control_ioct 2342 .unlocked_ioctl = btrfs_control_ioctl,
2444 .compat_ioctl = compat_ptr_ioctl, 2343 .compat_ioctl = compat_ptr_ioctl,
2445 .owner = THIS_MODULE, 2344 .owner = THIS_MODULE,
2446 .llseek = noop_llseek, 2345 .llseek = noop_llseek,
2447 }; 2346 };
2448 2347
2449 static struct miscdevice btrfs_misc = { 2348 static struct miscdevice btrfs_misc = {
2450 .minor = BTRFS_MINOR, 2349 .minor = BTRFS_MINOR,
2451 .name = "btrfs-control", 2350 .name = "btrfs-control",
2452 .fops = &btrfs_ctl_fops 2351 .fops = &btrfs_ctl_fops
2453 }; 2352 };
2454 2353
2455 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 2354 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2456 MODULE_ALIAS("devname:btrfs-control"); 2355 MODULE_ALIAS("devname:btrfs-control");
2457 2356
2458 static int __init btrfs_interface_init(void) 2357 static int __init btrfs_interface_init(void)
2459 { 2358 {
2460 return misc_register(&btrfs_misc); 2359 return misc_register(&btrfs_misc);
2461 } 2360 }
2462 2361
2463 static __cold void btrfs_interface_exit(void) 2362 static __cold void btrfs_interface_exit(void)
2464 { 2363 {
2465 misc_deregister(&btrfs_misc); 2364 misc_deregister(&btrfs_misc);
2466 } 2365 }
2467 2366
2468 static int __init btrfs_print_mod_info(void) !! 2367 static void __init btrfs_print_mod_info(void)
2469 { 2368 {
2470 static const char options[] = "" 2369 static const char options[] = ""
2471 #ifdef CONFIG_BTRFS_DEBUG 2370 #ifdef CONFIG_BTRFS_DEBUG
2472 ", debug=on" 2371 ", debug=on"
2473 #endif 2372 #endif
2474 #ifdef CONFIG_BTRFS_ASSERT 2373 #ifdef CONFIG_BTRFS_ASSERT
2475 ", assert=on" 2374 ", assert=on"
2476 #endif 2375 #endif
>> 2376 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 2377 ", integrity-checker=on"
>> 2378 #endif
2477 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 2379 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2478 ", ref-verify=on" 2380 ", ref-verify=on"
2479 #endif 2381 #endif
2480 #ifdef CONFIG_BLK_DEV_ZONED <<
2481 ", zoned=yes" <<
2482 #else <<
2483 ", zoned=no" <<
2484 #endif <<
2485 #ifdef CONFIG_FS_VERITY <<
2486 ", fsverity=yes" <<
2487 #else <<
2488 ", fsverity=no" <<
2489 #endif <<
2490 ; 2382 ;
2491 pr_info("Btrfs loaded%s\n", options); !! 2383 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2492 return 0; <<
2493 } 2384 }
2494 2385
2495 static int register_btrfs(void) !! 2386 static int __init init_btrfs_fs(void)
2496 { 2387 {
2497 return register_filesystem(&btrfs_fs_ !! 2388 int err;
2498 } <<
2499 2389
2500 static void unregister_btrfs(void) !! 2390 btrfs_props_init();
2501 { <<
2502 unregister_filesystem(&btrfs_fs_type) <<
2503 } <<
2504 2391
2505 /* Helper structure for long init/exit functi !! 2392 err = btrfs_init_sysfs();
2506 struct init_sequence { !! 2393 if (err)
2507 int (*init_func)(void); !! 2394 return err;
2508 /* Can be NULL if the init_func doesn <<
2509 void (*exit_func)(void); <<
2510 }; <<
2511 2395
2512 static const struct init_sequence mod_init_se !! 2396 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 2397
2579 static bool mod_init_result[ARRAY_SIZE(mod_in !! 2398 err = btrfs_init_cachep();
>> 2399 if (err)
>> 2400 goto free_compress;
2580 2401
2581 static __always_inline void btrfs_exit_btrfs_ !! 2402 err = extent_io_init();
2582 { !! 2403 if (err)
2583 int i; !! 2404 goto free_cachep;
2584 2405
2585 for (i = ARRAY_SIZE(mod_init_seq) - 1 !! 2406 err = extent_state_cache_init();
2586 if (!mod_init_result[i]) !! 2407 if (err)
2587 continue; !! 2408 goto free_extent_io;
2588 if (mod_init_seq[i].exit_func !! 2409
2589 mod_init_seq[i].exit_ !! 2410 err = extent_map_init();
2590 mod_init_result[i] = false; !! 2411 if (err)
2591 } !! 2412 goto free_extent_state_cache;
>> 2413
>> 2414 err = ordered_data_init();
>> 2415 if (err)
>> 2416 goto free_extent_map;
>> 2417
>> 2418 err = btrfs_delayed_inode_init();
>> 2419 if (err)
>> 2420 goto free_ordered_data;
>> 2421
>> 2422 err = btrfs_auto_defrag_init();
>> 2423 if (err)
>> 2424 goto free_delayed_inode;
>> 2425
>> 2426 err = btrfs_delayed_ref_init();
>> 2427 if (err)
>> 2428 goto free_auto_defrag;
>> 2429
>> 2430 err = btrfs_prelim_ref_init();
>> 2431 if (err)
>> 2432 goto free_delayed_ref;
>> 2433
>> 2434 err = btrfs_end_io_wq_init();
>> 2435 if (err)
>> 2436 goto free_prelim_ref;
>> 2437
>> 2438 err = btrfs_interface_init();
>> 2439 if (err)
>> 2440 goto free_end_io_wq;
>> 2441
>> 2442 btrfs_init_lockdep();
>> 2443
>> 2444 btrfs_print_mod_info();
>> 2445
>> 2446 err = btrfs_run_sanity_tests();
>> 2447 if (err)
>> 2448 goto unregister_ioctl;
>> 2449
>> 2450 err = register_filesystem(&btrfs_fs_type);
>> 2451 if (err)
>> 2452 goto unregister_ioctl;
>> 2453
>> 2454 return 0;
>> 2455
>> 2456 unregister_ioctl:
>> 2457 btrfs_interface_exit();
>> 2458 free_end_io_wq:
>> 2459 btrfs_end_io_wq_exit();
>> 2460 free_prelim_ref:
>> 2461 btrfs_prelim_ref_exit();
>> 2462 free_delayed_ref:
>> 2463 btrfs_delayed_ref_exit();
>> 2464 free_auto_defrag:
>> 2465 btrfs_auto_defrag_exit();
>> 2466 free_delayed_inode:
>> 2467 btrfs_delayed_inode_exit();
>> 2468 free_ordered_data:
>> 2469 ordered_data_exit();
>> 2470 free_extent_map:
>> 2471 extent_map_exit();
>> 2472 free_extent_state_cache:
>> 2473 extent_state_cache_exit();
>> 2474 free_extent_io:
>> 2475 extent_io_exit();
>> 2476 free_cachep:
>> 2477 btrfs_destroy_cachep();
>> 2478 free_compress:
>> 2479 btrfs_exit_compress();
>> 2480 btrfs_exit_sysfs();
>> 2481
>> 2482 return err;
2592 } 2483 }
2593 2484
2594 static void __exit exit_btrfs_fs(void) 2485 static void __exit exit_btrfs_fs(void)
2595 { 2486 {
2596 btrfs_exit_btrfs_fs(); !! 2487 btrfs_destroy_cachep();
>> 2488 btrfs_delayed_ref_exit();
>> 2489 btrfs_auto_defrag_exit();
>> 2490 btrfs_delayed_inode_exit();
>> 2491 btrfs_prelim_ref_exit();
>> 2492 ordered_data_exit();
>> 2493 extent_map_exit();
>> 2494 extent_state_cache_exit();
>> 2495 extent_io_exit();
>> 2496 btrfs_interface_exit();
>> 2497 btrfs_end_io_wq_exit();
>> 2498 unregister_filesystem(&btrfs_fs_type);
>> 2499 btrfs_exit_sysfs();
2597 btrfs_cleanup_fs_uuids(); 2500 btrfs_cleanup_fs_uuids();
2598 } !! 2501 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 } 2502 }
2616 2503
2617 late_initcall(init_btrfs_fs); 2504 late_initcall(init_btrfs_fs);
2618 module_exit(exit_btrfs_fs) 2505 module_exit(exit_btrfs_fs)
2619 2506
2620 MODULE_DESCRIPTION("B-Tree File System (BTRFS <<
2621 MODULE_LICENSE("GPL"); 2507 MODULE_LICENSE("GPL");
2622 MODULE_SOFTDEP("pre: crc32c"); 2508 MODULE_SOFTDEP("pre: crc32c");
2623 MODULE_SOFTDEP("pre: xxhash64"); 2509 MODULE_SOFTDEP("pre: xxhash64");
2624 MODULE_SOFTDEP("pre: sha256"); 2510 MODULE_SOFTDEP("pre: sha256");
2625 MODULE_SOFTDEP("pre: blake2b-256"); 2511 MODULE_SOFTDEP("pre: blake2b-256");
2626 2512
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.