1 // SPDX-License-Identifier: GPL-2.0 <<
2 /* 1 /*
3 * Copyright (C) 2007 Oracle. All rights rese 2 * Copyright (C) 2007 Oracle. All rights reserved.
>> 3 *
>> 4 * This program is free software; you can redistribute it and/or
>> 5 * modify it under the terms of the GNU General Public
>> 6 * License v2 as published by the Free Software Foundation.
>> 7 *
>> 8 * This program is distributed in the hope that it will be useful,
>> 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
>> 11 * General Public License for more details.
>> 12 *
>> 13 * You should have received a copy of the GNU General Public
>> 14 * License along with this program; if not, write to the
>> 15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
>> 16 * Boston, MA 021110-1307, USA.
4 */ 17 */
5 18
6 #include <linux/blkdev.h> 19 #include <linux/blkdev.h>
7 #include <linux/module.h> 20 #include <linux/module.h>
>> 21 #include <linux/buffer_head.h>
8 #include <linux/fs.h> 22 #include <linux/fs.h>
9 #include <linux/pagemap.h> 23 #include <linux/pagemap.h>
10 #include <linux/highmem.h> 24 #include <linux/highmem.h>
11 #include <linux/time.h> 25 #include <linux/time.h>
12 #include <linux/init.h> 26 #include <linux/init.h>
13 #include <linux/seq_file.h> 27 #include <linux/seq_file.h>
14 #include <linux/string.h> 28 #include <linux/string.h>
15 #include <linux/backing-dev.h> 29 #include <linux/backing-dev.h>
16 #include <linux/mount.h> 30 #include <linux/mount.h>
>> 31 #include <linux/mpage.h>
>> 32 #include <linux/swap.h>
17 #include <linux/writeback.h> 33 #include <linux/writeback.h>
18 #include <linux/statfs.h> 34 #include <linux/statfs.h>
19 #include <linux/compat.h> 35 #include <linux/compat.h>
20 #include <linux/parser.h> 36 #include <linux/parser.h>
21 #include <linux/ctype.h> 37 #include <linux/ctype.h>
22 #include <linux/namei.h> 38 #include <linux/namei.h>
23 #include <linux/miscdevice.h> 39 #include <linux/miscdevice.h>
24 #include <linux/magic.h> 40 #include <linux/magic.h>
25 #include <linux/slab.h> 41 #include <linux/slab.h>
>> 42 #include <linux/cleancache.h>
26 #include <linux/ratelimit.h> 43 #include <linux/ratelimit.h>
27 #include <linux/crc32c.h> <<
28 #include <linux/btrfs.h> 44 #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" 45 #include "delayed-inode.h"
34 #include "ctree.h" 46 #include "ctree.h"
35 #include "disk-io.h" 47 #include "disk-io.h"
36 #include "transaction.h" 48 #include "transaction.h"
37 #include "btrfs_inode.h" 49 #include "btrfs_inode.h"
38 #include "direct-io.h" !! 50 #include "print-tree.h"
>> 51 #include "hash.h"
39 #include "props.h" 52 #include "props.h"
40 #include "xattr.h" 53 #include "xattr.h"
41 #include "bio.h" !! 54 #include "volumes.h"
42 #include "export.h" 55 #include "export.h"
43 #include "compression.h" 56 #include "compression.h"
>> 57 #include "rcu-string.h"
44 #include "dev-replace.h" 58 #include "dev-replace.h"
45 #include "free-space-cache.h" 59 #include "free-space-cache.h"
46 #include "backref.h" 60 #include "backref.h"
47 #include "space-info.h" <<
48 #include "sysfs.h" <<
49 #include "zoned.h" <<
50 #include "tests/btrfs-tests.h" 61 #include "tests/btrfs-tests.h"
51 #include "block-group.h" !! 62
52 #include "discard.h" <<
53 #include "qgroup.h" 63 #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 64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h> 65 #include <trace/events/btrfs.h>
66 66
67 static const struct super_operations btrfs_sup 67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type; 68 static struct file_system_type btrfs_fs_type;
69 69
70 static void btrfs_put_super(struct super_block !! 70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
>> 71
>> 72 const char *btrfs_decode_error(int errno)
71 { 73 {
72 struct btrfs_fs_info *fs_info = btrfs_ !! 74 char *errstr = "unknown";
73 75
74 btrfs_info(fs_info, "last unmount of f !! 76 switch (errno) {
75 close_ctree(fs_info); !! 77 case -EIO:
>> 78 errstr = "IO failure";
>> 79 break;
>> 80 case -ENOMEM:
>> 81 errstr = "Out of memory";
>> 82 break;
>> 83 case -EROFS:
>> 84 errstr = "Readonly filesystem";
>> 85 break;
>> 86 case -EEXIST:
>> 87 errstr = "Object already exists";
>> 88 break;
>> 89 case -ENOSPC:
>> 90 errstr = "No space left";
>> 91 break;
>> 92 case -ENOENT:
>> 93 errstr = "No such entry";
>> 94 break;
>> 95 }
>> 96
>> 97 return errstr;
76 } 98 }
77 99
78 /* Store the mount options related information !! 100 /* btrfs handle error by forcing the filesystem readonly */
79 struct btrfs_fs_context { !! 101 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
80 char *subvol_name; !! 102 {
81 u64 subvol_objectid; !! 103 struct super_block *sb = fs_info->sb;
82 u64 max_inline; <<
83 u32 commit_interval; <<
84 u32 metadata_ratio; <<
85 u32 thread_pool_size; <<
86 unsigned long long mount_opt; <<
87 unsigned long compress_type:4; <<
88 unsigned int compress_level; <<
89 refcount_t refs; <<
90 }; <<
91 104
92 enum { !! 105 if (sb->s_flags & MS_RDONLY)
93 Opt_acl, !! 106 return;
94 Opt_clear_cache, <<
95 Opt_commit_interval, <<
96 Opt_compress, <<
97 Opt_compress_force, <<
98 Opt_compress_force_type, <<
99 Opt_compress_type, <<
100 Opt_degraded, <<
101 Opt_device, <<
102 Opt_fatal_errors, <<
103 Opt_flushoncommit, <<
104 Opt_max_inline, <<
105 Opt_barrier, <<
106 Opt_datacow, <<
107 Opt_datasum, <<
108 Opt_defrag, <<
109 Opt_discard, <<
110 Opt_discard_mode, <<
111 Opt_ratio, <<
112 Opt_rescan_uuid_tree, <<
113 Opt_skip_balance, <<
114 Opt_space_cache, <<
115 Opt_space_cache_version, <<
116 Opt_ssd, <<
117 Opt_ssd_spread, <<
118 Opt_subvol, <<
119 Opt_subvol_empty, <<
120 Opt_subvolid, <<
121 Opt_thread_pool, <<
122 Opt_treelog, <<
123 Opt_user_subvol_rm_allowed, <<
124 Opt_norecovery, <<
125 <<
126 /* Rescue options */ <<
127 Opt_rescue, <<
128 Opt_usebackuproot, <<
129 Opt_nologreplay, <<
130 107
131 /* Debugging options */ !! 108 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
132 Opt_enospc_debug, !! 109 sb->s_flags |= MS_RDONLY;
133 #ifdef CONFIG_BTRFS_DEBUG !! 110 btrfs_info(fs_info, "forced readonly");
134 Opt_fragment, Opt_fragment_data, Opt_f !! 111 /*
>> 112 * Note that a running device replace operation is not
>> 113 * canceled here although there is no way to update
>> 114 * the progress. It would add the risk of a deadlock,
>> 115 * therefore the canceling is omitted. The only penalty
>> 116 * is that some I/O remains active until the procedure
>> 117 * completes. The next time when the filesystem is
>> 118 * mounted writeable again, the device replace
>> 119 * operation continues.
>> 120 */
>> 121 }
>> 122 }
>> 123
>> 124 /*
>> 125 * __btrfs_handle_fs_error decodes expected errors from the caller and
>> 126 * invokes the approciate error response.
>> 127 */
>> 128 __cold
>> 129 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
>> 130 unsigned int line, int errno, const char *fmt, ...)
>> 131 {
>> 132 struct super_block *sb = fs_info->sb;
>> 133 #ifdef CONFIG_PRINTK
>> 134 const char *errstr;
135 #endif 135 #endif
136 #ifdef CONFIG_BTRFS_FS_REF_VERIFY !! 136
137 Opt_ref_verify, !! 137 /*
>> 138 * Special case: if the error is EROFS, and we're already
>> 139 * under MS_RDONLY, then it is safe here.
>> 140 */
>> 141 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
>> 142 return;
>> 143
>> 144 #ifdef CONFIG_PRINTK
>> 145 errstr = btrfs_decode_error(errno);
>> 146 if (fmt) {
>> 147 struct va_format vaf;
>> 148 va_list args;
>> 149
>> 150 va_start(args, fmt);
>> 151 vaf.fmt = fmt;
>> 152 vaf.va = &args;
>> 153
>> 154 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
>> 155 sb->s_id, function, line, errno, errstr, &vaf);
>> 156 va_end(args);
>> 157 } else {
>> 158 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
>> 159 sb->s_id, function, line, errno, errstr);
>> 160 }
138 #endif 161 #endif
139 Opt_err, <<
140 }; <<
141 162
142 enum { !! 163 /*
143 Opt_fatal_errors_panic, !! 164 * Today we only save the error info to memory. Long term we'll
144 Opt_fatal_errors_bug, !! 165 * also send it down to the disk
145 }; !! 166 */
>> 167 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
146 168
147 static const struct constant_table btrfs_param !! 169 /* Don't go through full error handling during mount */
148 { "panic", Opt_fatal_errors_panic }, !! 170 if (sb->s_flags & MS_BORN)
149 { "bug", Opt_fatal_errors_bug }, !! 171 btrfs_handle_error(fs_info);
150 {} !! 172 }
>> 173
>> 174 #ifdef CONFIG_PRINTK
>> 175 static const char * const logtypes[] = {
>> 176 "emergency",
>> 177 "alert",
>> 178 "critical",
>> 179 "error",
>> 180 "warning",
>> 181 "notice",
>> 182 "info",
>> 183 "debug",
151 }; 184 };
152 185
153 enum { <<
154 Opt_discard_sync, <<
155 Opt_discard_async, <<
156 }; <<
157 186
158 static const struct constant_table btrfs_param !! 187 /*
159 { "sync", Opt_discard_sync }, !! 188 * Use one ratelimit state per log level so that a flood of less important
160 { "async", Opt_discard_async }, !! 189 * messages doesn't cause more important ones to be dropped.
161 {} !! 190 */
>> 191 static struct ratelimit_state printk_limits[] = {
>> 192 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 193 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 194 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 195 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 196 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 197 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 198 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
>> 199 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
162 }; 200 };
163 201
164 enum { !! 202 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
165 Opt_space_cache_v1, !! 203 {
166 Opt_space_cache_v2, !! 204 struct super_block *sb = fs_info->sb;
167 }; !! 205 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
>> 206 struct va_format vaf;
>> 207 va_list args;
>> 208 int kern_level;
>> 209 const char *type = logtypes[4];
>> 210 struct ratelimit_state *ratelimit = &printk_limits[4];
168 211
169 static const struct constant_table btrfs_param !! 212 va_start(args, fmt);
170 { "v1", Opt_space_cache_v1 }, <<
171 { "v2", Opt_space_cache_v2 }, <<
172 {} <<
173 }; <<
174 213
175 enum { !! 214 while ((kern_level = printk_get_level(fmt)) != 0) {
176 Opt_rescue_usebackuproot, !! 215 size_t size = printk_skip_level(fmt) - fmt;
177 Opt_rescue_nologreplay, <<
178 Opt_rescue_ignorebadroots, <<
179 Opt_rescue_ignoredatacsums, <<
180 Opt_rescue_ignoremetacsums, <<
181 Opt_rescue_ignoresuperflags, <<
182 Opt_rescue_parameter_all, <<
183 }; <<
184 216
185 static const struct constant_table btrfs_param !! 217 if (kern_level >= '' && kern_level <= '7') {
186 { "usebackuproot", Opt_rescue_usebacku !! 218 memcpy(lvl, fmt, size);
187 { "nologreplay", Opt_rescue_nologrepla !! 219 lvl[size] = '\0';
188 { "ignorebadroots", Opt_rescue_ignoreb !! 220 type = logtypes[kern_level - ''];
189 { "ibadroots", Opt_rescue_ignorebadroo !! 221 ratelimit = &printk_limits[kern_level - ''];
190 { "ignoredatacsums", Opt_rescue_ignore !! 222 }
191 { "ignoremetacsums", Opt_rescue_ignore !! 223 fmt += size;
192 { "ignoresuperflags", Opt_rescue_ignor !! 224 }
193 { "idatacsums", Opt_rescue_ignoredatac <<
194 { "imetacsums", Opt_rescue_ignoremetac <<
195 { "isuperflags", Opt_rescue_ignoresupe <<
196 { "all", Opt_rescue_parameter_all }, <<
197 {} <<
198 }; <<
199 225
200 #ifdef CONFIG_BTRFS_DEBUG !! 226 vaf.fmt = fmt;
201 enum { !! 227 vaf.va = &args;
202 Opt_fragment_parameter_data, <<
203 Opt_fragment_parameter_metadata, <<
204 Opt_fragment_parameter_all, <<
205 }; <<
206 228
207 static const struct constant_table btrfs_param !! 229 if (__ratelimit(ratelimit))
208 { "data", Opt_fragment_parameter_data !! 230 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
209 { "metadata", Opt_fragment_parameter_m !! 231
210 { "all", Opt_fragment_parameter_all }, !! 232 va_end(args);
211 {} !! 233 }
212 }; <<
213 #endif 234 #endif
214 235
215 static const struct fs_parameter_spec btrfs_fs !! 236 /*
216 fsparam_flag_no("acl", Opt_acl), !! 237 * We only mark the transaction aborted and then set the file system read-only.
217 fsparam_flag_no("autodefrag", Opt_defr !! 238 * This will prevent new transactions from starting or trying to join this
218 fsparam_flag_no("barrier", Opt_barrier !! 239 * one.
219 fsparam_flag("clear_cache", Opt_clear_ !! 240 *
220 fsparam_u32("commit", Opt_commit_inter !! 241 * This means that error recovery at the call site is limited to freeing
221 fsparam_flag("compress", Opt_compress) !! 242 * any local memory allocations and passing the error code up without
222 fsparam_string("compress", Opt_compres !! 243 * further cleanup. The transaction should complete as it normally would
223 fsparam_flag("compress-force", Opt_com !! 244 * in the call path but will return -EIO.
224 fsparam_string("compress-force", Opt_c !! 245 *
225 fsparam_flag_no("datacow", Opt_datacow !! 246 * We'll complete the cleanup in btrfs_end_transaction and
226 fsparam_flag_no("datasum", Opt_datasum !! 247 * btrfs_commit_transaction.
227 fsparam_flag("degraded", Opt_degraded) !! 248 */
228 fsparam_string("device", Opt_device), !! 249 __cold
229 fsparam_flag_no("discard", Opt_discard !! 250 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
230 fsparam_enum("discard", Opt_discard_mo !! 251 const char *function,
231 fsparam_enum("fatal_errors", Opt_fatal !! 252 unsigned int line, int errno)
232 fsparam_flag_no("flushoncommit", Opt_f !! 253 {
233 fsparam_string("max_inline", Opt_max_i !! 254 struct btrfs_fs_info *fs_info = trans->fs_info;
234 fsparam_u32("metadata_ratio", Opt_rati !! 255
235 fsparam_flag("rescan_uuid_tree", Opt_r !! 256 trans->aborted = errno;
236 fsparam_flag("skip_balance", Opt_skip_ !! 257 /* Nothing used. The other threads that have joined this
237 fsparam_flag_no("space_cache", Opt_spa !! 258 * transaction may be able to continue. */
238 fsparam_enum("space_cache", Opt_space_ !! 259 if (!trans->dirty && list_empty(&trans->new_bgs)) {
239 fsparam_flag_no("ssd", Opt_ssd), !! 260 const char *errstr;
240 fsparam_flag_no("ssd_spread", Opt_ssd_ !! 261
241 fsparam_string("subvol", Opt_subvol), !! 262 errstr = btrfs_decode_error(errno);
242 fsparam_flag("subvol=", Opt_subvol_emp !! 263 btrfs_warn(fs_info,
243 fsparam_u64("subvolid", Opt_subvolid), !! 264 "%s:%d: Aborting unused transaction(%s).",
244 fsparam_u32("thread_pool", Opt_thread_ !! 265 function, line, errstr);
245 fsparam_flag_no("treelog", Opt_treelog !! 266 return;
246 fsparam_flag("user_subvol_rm_allowed", !! 267 }
247 !! 268 WRITE_ONCE(trans->transaction->aborted, errno);
248 /* Rescue options. */ !! 269 /* Wake up anybody who may be waiting on this transaction */
249 fsparam_enum("rescue", Opt_rescue, btr !! 270 wake_up(&fs_info->transaction_wait);
250 /* Deprecated, with alias rescue=nolog !! 271 wake_up(&fs_info->transaction_blocked_wait);
251 __fsparam(NULL, "nologreplay", Opt_nol !! 272 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
252 /* Deprecated, with alias rescue=useba !! 273 }
253 __fsparam(NULL, "usebackuproot", Opt_u !! 274 /*
254 /* For compatibility only, alias for " !! 275 * __btrfs_panic decodes unexpected, fatal errors from the caller,
255 fsparam_flag("norecovery", Opt_norecov !! 276 * issues an alert, and either panics or BUGs, depending on mount options.
>> 277 */
>> 278 __cold
>> 279 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
>> 280 unsigned int line, int errno, const char *fmt, ...)
>> 281 {
>> 282 char *s_id = "<unknown>";
>> 283 const char *errstr;
>> 284 struct va_format vaf = { .fmt = fmt };
>> 285 va_list args;
>> 286
>> 287 if (fs_info)
>> 288 s_id = fs_info->sb->s_id;
>> 289
>> 290 va_start(args, fmt);
>> 291 vaf.va = &args;
>> 292
>> 293 errstr = btrfs_decode_error(errno);
>> 294 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
>> 295 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
>> 296 s_id, function, line, &vaf, errno, errstr);
>> 297
>> 298 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
>> 299 function, line, &vaf, errno, errstr);
>> 300 va_end(args);
>> 301 /* Caller calls BUG() */
>> 302 }
>> 303
>> 304 static void btrfs_put_super(struct super_block *sb)
>> 305 {
>> 306 close_ctree(btrfs_sb(sb));
>> 307 }
256 308
257 /* Debugging options. */ !! 309 enum {
258 fsparam_flag_no("enospc_debug", Opt_en !! 310 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
>> 311 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
>> 312 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
>> 313 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
>> 314 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
>> 315 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
>> 316 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
>> 317 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
>> 318 Opt_skip_balance, Opt_check_integrity,
>> 319 Opt_check_integrity_including_extent_data,
>> 320 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
>> 321 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
>> 322 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
>> 323 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
>> 324 Opt_nologreplay, Opt_norecovery,
259 #ifdef CONFIG_BTRFS_DEBUG 325 #ifdef CONFIG_BTRFS_DEBUG
260 fsparam_enum("fragment", Opt_fragment, !! 326 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
261 #endif 327 #endif
262 #ifdef CONFIG_BTRFS_FS_REF_VERIFY !! 328 Opt_err,
263 fsparam_flag("ref_verify", Opt_ref_ver !! 329 };
>> 330
>> 331 static const match_table_t tokens = {
>> 332 {Opt_degraded, "degraded"},
>> 333 {Opt_subvol, "subvol=%s"},
>> 334 {Opt_subvolid, "subvolid=%s"},
>> 335 {Opt_device, "device=%s"},
>> 336 {Opt_nodatasum, "nodatasum"},
>> 337 {Opt_datasum, "datasum"},
>> 338 {Opt_nodatacow, "nodatacow"},
>> 339 {Opt_datacow, "datacow"},
>> 340 {Opt_nobarrier, "nobarrier"},
>> 341 {Opt_barrier, "barrier"},
>> 342 {Opt_max_inline, "max_inline=%s"},
>> 343 {Opt_alloc_start, "alloc_start=%s"},
>> 344 {Opt_thread_pool, "thread_pool=%d"},
>> 345 {Opt_compress, "compress"},
>> 346 {Opt_compress_type, "compress=%s"},
>> 347 {Opt_compress_force, "compress-force"},
>> 348 {Opt_compress_force_type, "compress-force=%s"},
>> 349 {Opt_ssd, "ssd"},
>> 350 {Opt_ssd_spread, "ssd_spread"},
>> 351 {Opt_nossd, "nossd"},
>> 352 {Opt_acl, "acl"},
>> 353 {Opt_noacl, "noacl"},
>> 354 {Opt_notreelog, "notreelog"},
>> 355 {Opt_treelog, "treelog"},
>> 356 {Opt_nologreplay, "nologreplay"},
>> 357 {Opt_norecovery, "norecovery"},
>> 358 {Opt_flushoncommit, "flushoncommit"},
>> 359 {Opt_noflushoncommit, "noflushoncommit"},
>> 360 {Opt_ratio, "metadata_ratio=%d"},
>> 361 {Opt_discard, "discard"},
>> 362 {Opt_nodiscard, "nodiscard"},
>> 363 {Opt_space_cache, "space_cache"},
>> 364 {Opt_space_cache_version, "space_cache=%s"},
>> 365 {Opt_clear_cache, "clear_cache"},
>> 366 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
>> 367 {Opt_enospc_debug, "enospc_debug"},
>> 368 {Opt_noenospc_debug, "noenospc_debug"},
>> 369 {Opt_subvolrootid, "subvolrootid=%d"},
>> 370 {Opt_defrag, "autodefrag"},
>> 371 {Opt_nodefrag, "noautodefrag"},
>> 372 {Opt_inode_cache, "inode_cache"},
>> 373 {Opt_noinode_cache, "noinode_cache"},
>> 374 {Opt_no_space_cache, "nospace_cache"},
>> 375 {Opt_recovery, "recovery"}, /* deprecated */
>> 376 {Opt_usebackuproot, "usebackuproot"},
>> 377 {Opt_skip_balance, "skip_balance"},
>> 378 {Opt_check_integrity, "check_int"},
>> 379 {Opt_check_integrity_including_extent_data, "check_int_data"},
>> 380 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
>> 381 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
>> 382 {Opt_fatal_errors, "fatal_errors=%s"},
>> 383 {Opt_commit_interval, "commit=%d"},
>> 384 #ifdef CONFIG_BTRFS_DEBUG
>> 385 {Opt_fragment_data, "fragment=data"},
>> 386 {Opt_fragment_metadata, "fragment=metadata"},
>> 387 {Opt_fragment_all, "fragment=all"},
264 #endif 388 #endif
265 {} !! 389 {Opt_err, NULL},
266 }; 390 };
267 391
268 /* No support for restricting writes to btrfs !! 392 /*
269 static inline blk_mode_t btrfs_open_mode(struc !! 393 * Regular mount options parser. Everything that is needed only when
>> 394 * reading in a new superblock is parsed here.
>> 395 * XXX JDM: This needs to be cleaned up for remount.
>> 396 */
>> 397 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
>> 398 unsigned long new_flags)
270 { 399 {
271 return sb_open_mode(fc->sb_flags) & ~B !! 400 substring_t args[MAX_OPT_ARGS];
272 } !! 401 char *p, *num, *orig = NULL;
>> 402 u64 cache_gen;
>> 403 int intarg;
>> 404 int ret = 0;
>> 405 char *compress_type;
>> 406 bool compress_force = false;
>> 407 enum btrfs_compression_type saved_compress_type;
>> 408 bool saved_compress_force;
>> 409 int no_compress = 0;
273 410
274 static int btrfs_parse_param(struct fs_context !! 411 cache_gen = btrfs_super_cache_generation(info->super_copy);
275 { !! 412 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
276 struct btrfs_fs_context *ctx = fc->fs_ !! 413 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
277 struct fs_parse_result result; !! 414 else if (cache_gen)
278 int opt; !! 415 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
279 416
280 opt = fs_parse(fc, btrfs_fs_parameters !! 417 /*
281 if (opt < 0) !! 418 * Even the options are empty, we still need to do extra check
282 return opt; !! 419 * against new flags
>> 420 */
>> 421 if (!options)
>> 422 goto check;
283 423
284 switch (opt) { !! 424 /*
285 case Opt_degraded: !! 425 * strsep changes the string, duplicate it because parse_options
286 btrfs_set_opt(ctx->mount_opt, !! 426 * gets called twice
287 break; !! 427 */
288 case Opt_subvol_empty: !! 428 options = kstrdup(options, GFP_NOFS);
289 /* !! 429 if (!options)
290 * This exists because we used !! 430 return -ENOMEM;
291 * keeping it to maintain ABI. <<
292 * empty subvol= again"). <<
293 */ <<
294 break; <<
295 case Opt_subvol: <<
296 kfree(ctx->subvol_name); <<
297 ctx->subvol_name = kstrdup(par <<
298 if (!ctx->subvol_name) <<
299 return -ENOMEM; <<
300 break; <<
301 case Opt_subvolid: <<
302 ctx->subvol_objectid = result. <<
303 431
304 /* subvolid=0 means give me th !! 432 orig = options;
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 <<
312 mutex_lock(&uuid_mutex); <<
313 device = btrfs_scan_one_device <<
314 mutex_unlock(&uuid_mutex); <<
315 if (IS_ERR(device)) <<
316 return PTR_ERR(device) <<
317 break; <<
318 } <<
319 case Opt_datasum: <<
320 if (result.negated) { <<
321 btrfs_set_opt(ctx->mou <<
322 } else { <<
323 btrfs_clear_opt(ctx->m <<
324 btrfs_clear_opt(ctx->m <<
325 } <<
326 break; <<
327 case Opt_datacow: <<
328 if (result.negated) { <<
329 btrfs_clear_opt(ctx->m <<
330 btrfs_clear_opt(ctx->m <<
331 btrfs_set_opt(ctx->mou <<
332 btrfs_set_opt(ctx->mou <<
333 } else { <<
334 btrfs_clear_opt(ctx->m <<
335 } <<
336 break; <<
337 case Opt_compress_force: <<
338 case Opt_compress_force_type: <<
339 btrfs_set_opt(ctx->mount_opt, <<
340 fallthrough; <<
341 case Opt_compress: <<
342 case Opt_compress_type: <<
343 /* <<
344 * Provide the same semantics <<
345 * context, specifying the "co <<
346 * "force-compress" without th <<
347 * "compress-force=[no|none]" <<
348 */ <<
349 if (opt != Opt_compress_force <<
350 btrfs_clear_opt(ctx->m <<
351 433
352 if (opt == Opt_compress || opt !! 434 while ((p = strsep(&options, ",")) != NULL) {
353 ctx->compress_type = B !! 435 int token;
354 ctx->compress_level = !! 436 if (!*p)
355 btrfs_set_opt(ctx->mou !! 437 continue;
356 btrfs_clear_opt(ctx->m !! 438
357 btrfs_clear_opt(ctx->m !! 439 token = match_token(p, tokens, args);
358 } else if (strncmp(param->stri !! 440 switch (token) {
359 ctx->compress_type = B !! 441 case Opt_degraded:
360 ctx->compress_level = !! 442 btrfs_info(info, "allowing degraded mounts");
361 btrfs_compress !! 443 btrfs_set_opt(info->mount_opt, DEGRADED);
362 !! 444 break;
363 btrfs_set_opt(ctx->mou !! 445 case Opt_subvol:
364 btrfs_clear_opt(ctx->m !! 446 case Opt_subvolid:
365 btrfs_clear_opt(ctx->m !! 447 case Opt_subvolrootid:
366 } else if (strncmp(param->stri !! 448 case Opt_device:
367 ctx->compress_type = B !! 449 /*
368 ctx->compress_level = !! 450 * These are parsed by btrfs_parse_early_options
369 btrfs_set_opt(ctx->mou !! 451 * and can be happily ignored here.
370 btrfs_clear_opt(ctx->m !! 452 */
371 btrfs_clear_opt(ctx->m !! 453 break;
372 } else if (strncmp(param->stri !! 454 case Opt_nodatasum:
373 ctx->compress_type = B !! 455 btrfs_set_and_info(info, NODATASUM,
374 ctx->compress_level = !! 456 "setting nodatasum");
375 btrfs_compress !! 457 break;
376 !! 458 case Opt_datasum:
377 btrfs_set_opt(ctx->mou !! 459 if (btrfs_test_opt(info, NODATASUM)) {
378 btrfs_clear_opt(ctx->m !! 460 if (btrfs_test_opt(info, NODATACOW))
379 btrfs_clear_opt(ctx->m !! 461 btrfs_info(info,
380 } else if (strncmp(param->stri !! 462 "setting datasum, datacow enabled");
381 ctx->compress_level = !! 463 else
382 ctx->compress_type = 0 !! 464 btrfs_info(info, "setting datasum");
383 btrfs_clear_opt(ctx->m !! 465 }
384 btrfs_clear_opt(ctx->m !! 466 btrfs_clear_opt(info->mount_opt, NODATACOW);
385 } else { !! 467 btrfs_clear_opt(info->mount_opt, NODATASUM);
386 btrfs_err(NULL, "unrec !! 468 break;
387 param->strin !! 469 case Opt_nodatacow:
388 return -EINVAL; !! 470 if (!btrfs_test_opt(info, NODATACOW)) {
389 } !! 471 if (!btrfs_test_opt(info, COMPRESS) ||
390 break; !! 472 !btrfs_test_opt(info, FORCE_COMPRESS)) {
391 case Opt_ssd: !! 473 btrfs_info(info,
392 if (result.negated) { !! 474 "setting nodatacow, compression disabled");
393 btrfs_set_opt(ctx->mou !! 475 } else {
394 btrfs_clear_opt(ctx->m !! 476 btrfs_info(info, "setting nodatacow");
395 btrfs_clear_opt(ctx->m !! 477 }
396 } else { !! 478 }
397 btrfs_set_opt(ctx->mou !! 479 btrfs_clear_opt(info->mount_opt, COMPRESS);
398 btrfs_clear_opt(ctx->m !! 480 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
399 } !! 481 btrfs_set_opt(info->mount_opt, NODATACOW);
400 break; !! 482 btrfs_set_opt(info->mount_opt, NODATASUM);
401 case Opt_ssd_spread: !! 483 break;
402 if (result.negated) { !! 484 case Opt_datacow:
403 btrfs_clear_opt(ctx->m !! 485 btrfs_clear_and_info(info, NODATACOW,
404 } else { !! 486 "setting datacow");
405 btrfs_set_opt(ctx->mou !! 487 break;
406 btrfs_set_opt(ctx->mou !! 488 case Opt_compress_force:
407 btrfs_clear_opt(ctx->m !! 489 case Opt_compress_force_type:
408 } !! 490 compress_force = true;
409 break; !! 491 /* Fallthrough */
410 case Opt_barrier: !! 492 case Opt_compress:
411 if (result.negated) !! 493 case Opt_compress_type:
412 btrfs_set_opt(ctx->mou !! 494 saved_compress_type = btrfs_test_opt(info,
413 else !! 495 COMPRESS) ?
414 btrfs_clear_opt(ctx->m !! 496 info->compress_type : BTRFS_COMPRESS_NONE;
415 break; !! 497 saved_compress_force =
416 case Opt_thread_pool: !! 498 btrfs_test_opt(info, FORCE_COMPRESS);
417 if (result.uint_32 == 0) { !! 499 if (token == Opt_compress ||
418 btrfs_err(NULL, "inval !! 500 token == Opt_compress_force ||
419 return -EINVAL; !! 501 strcmp(args[0].from, "zlib") == 0) {
420 } !! 502 compress_type = "zlib";
421 ctx->thread_pool_size = result !! 503 info->compress_type = BTRFS_COMPRESS_ZLIB;
422 break; !! 504 btrfs_set_opt(info->mount_opt, COMPRESS);
423 case Opt_max_inline: !! 505 btrfs_clear_opt(info->mount_opt, NODATACOW);
424 ctx->max_inline = memparse(par !! 506 btrfs_clear_opt(info->mount_opt, NODATASUM);
425 break; !! 507 no_compress = 0;
426 case Opt_acl: !! 508 } else if (strcmp(args[0].from, "lzo") == 0) {
427 if (result.negated) { !! 509 compress_type = "lzo";
428 fc->sb_flags &= ~SB_PO !! 510 info->compress_type = BTRFS_COMPRESS_LZO;
429 } else { !! 511 btrfs_set_opt(info->mount_opt, COMPRESS);
>> 512 btrfs_clear_opt(info->mount_opt, NODATACOW);
>> 513 btrfs_clear_opt(info->mount_opt, NODATASUM);
>> 514 btrfs_set_fs_incompat(info, COMPRESS_LZO);
>> 515 no_compress = 0;
>> 516 } else if (strncmp(args[0].from, "no", 2) == 0) {
>> 517 compress_type = "no";
>> 518 btrfs_clear_opt(info->mount_opt, COMPRESS);
>> 519 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
>> 520 compress_force = false;
>> 521 no_compress++;
>> 522 } else {
>> 523 ret = -EINVAL;
>> 524 goto out;
>> 525 }
>> 526
>> 527 if (compress_force) {
>> 528 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
>> 529 } else {
>> 530 /*
>> 531 * If we remount from compress-force=xxx to
>> 532 * compress=xxx, we need clear FORCE_COMPRESS
>> 533 * flag, otherwise, there is no way for users
>> 534 * to disable forcible compression separately.
>> 535 */
>> 536 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
>> 537 }
>> 538 if ((btrfs_test_opt(info, COMPRESS) &&
>> 539 (info->compress_type != saved_compress_type ||
>> 540 compress_force != saved_compress_force)) ||
>> 541 (!btrfs_test_opt(info, COMPRESS) &&
>> 542 no_compress == 1)) {
>> 543 btrfs_info(info, "%s %s compression",
>> 544 (compress_force) ? "force" : "use",
>> 545 compress_type);
>> 546 }
>> 547 compress_force = false;
>> 548 break;
>> 549 case Opt_ssd:
>> 550 btrfs_set_and_info(info, SSD,
>> 551 "use ssd allocation scheme");
>> 552 btrfs_clear_opt(info->mount_opt, NOSSD);
>> 553 break;
>> 554 case Opt_ssd_spread:
>> 555 btrfs_set_and_info(info, SSD_SPREAD,
>> 556 "use spread ssd allocation scheme");
>> 557 btrfs_set_opt(info->mount_opt, SSD);
>> 558 btrfs_clear_opt(info->mount_opt, NOSSD);
>> 559 break;
>> 560 case Opt_nossd:
>> 561 btrfs_set_and_info(info, NOSSD,
>> 562 "not using ssd allocation scheme");
>> 563 btrfs_clear_opt(info->mount_opt, SSD);
>> 564 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
>> 565 break;
>> 566 case Opt_barrier:
>> 567 btrfs_clear_and_info(info, NOBARRIER,
>> 568 "turning on barriers");
>> 569 break;
>> 570 case Opt_nobarrier:
>> 571 btrfs_set_and_info(info, NOBARRIER,
>> 572 "turning off barriers");
>> 573 break;
>> 574 case Opt_thread_pool:
>> 575 ret = match_int(&args[0], &intarg);
>> 576 if (ret) {
>> 577 goto out;
>> 578 } else if (intarg > 0) {
>> 579 info->thread_pool_size = intarg;
>> 580 } else {
>> 581 ret = -EINVAL;
>> 582 goto out;
>> 583 }
>> 584 break;
>> 585 case Opt_max_inline:
>> 586 num = match_strdup(&args[0]);
>> 587 if (num) {
>> 588 info->max_inline = memparse(num, NULL);
>> 589 kfree(num);
>> 590
>> 591 if (info->max_inline) {
>> 592 info->max_inline = min_t(u64,
>> 593 info->max_inline,
>> 594 info->sectorsize);
>> 595 }
>> 596 btrfs_info(info, "max_inline at %llu",
>> 597 info->max_inline);
>> 598 } else {
>> 599 ret = -ENOMEM;
>> 600 goto out;
>> 601 }
>> 602 break;
>> 603 case Opt_alloc_start:
>> 604 num = match_strdup(&args[0]);
>> 605 if (num) {
>> 606 mutex_lock(&info->chunk_mutex);
>> 607 info->alloc_start = memparse(num, NULL);
>> 608 mutex_unlock(&info->chunk_mutex);
>> 609 kfree(num);
>> 610 btrfs_info(info, "allocations start at %llu",
>> 611 info->alloc_start);
>> 612 } else {
>> 613 ret = -ENOMEM;
>> 614 goto out;
>> 615 }
>> 616 break;
>> 617 case Opt_acl:
430 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 618 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
431 fc->sb_flags |= SB_POS !! 619 info->sb->s_flags |= MS_POSIXACL;
>> 620 break;
432 #else 621 #else
433 btrfs_err(NULL, "suppo !! 622 btrfs_err(info, "support for ACL not compiled in!");
434 return -EINVAL; !! 623 ret = -EINVAL;
>> 624 goto out;
435 #endif 625 #endif
436 } !! 626 case Opt_noacl:
437 /* !! 627 info->sb->s_flags &= ~MS_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; 628 break;
492 default: !! 629 case Opt_notreelog:
493 btrfs_err(NULL, "unrec !! 630 btrfs_set_and_info(info, NOTREELOG,
494 param->key); !! 631 "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; 632 break;
519 default: !! 633 case Opt_treelog:
520 btrfs_err(NULL, "unrec !! 634 btrfs_clear_and_info(info, NOTREELOG,
521 param->key); !! 635 "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; 636 break;
562 case Opt_fatal_errors_bug: !! 637 case Opt_norecovery:
563 btrfs_clear_opt(ctx->m !! 638 case Opt_nologreplay:
>> 639 btrfs_set_and_info(info, NOLOGREPLAY,
>> 640 "disabling log replay at mount time");
564 break; 641 break;
565 default: !! 642 case Opt_flushoncommit:
566 btrfs_err(NULL, "unrec !! 643 btrfs_set_and_info(info, FLUSHONCOMMIT,
567 param->key); !! 644 "turning on flush-on-commit");
568 return -EINVAL; !! 645 break;
569 } !! 646 case Opt_noflushoncommit:
570 break; !! 647 btrfs_clear_and_info(info, FLUSHONCOMMIT,
571 case Opt_commit_interval: !! 648 "turning off flush-on-commit");
572 ctx->commit_interval = result. !! 649 break;
573 if (ctx->commit_interval == 0) !! 650 case Opt_ratio:
574 ctx->commit_interval = !! 651 ret = match_int(&args[0], &intarg);
575 break; !! 652 if (ret) {
576 case Opt_rescue: !! 653 goto out;
577 switch (result.uint_32) { !! 654 } else if (intarg >= 0) {
578 case Opt_rescue_usebackuproot: !! 655 info->metadata_ratio = intarg;
579 btrfs_set_opt(ctx->mou !! 656 btrfs_info(info, "metadata ratio %d",
580 break; !! 657 info->metadata_ratio);
581 case Opt_rescue_nologreplay: !! 658 } else {
582 btrfs_set_opt(ctx->mou !! 659 ret = -EINVAL;
583 break; !! 660 goto out;
584 case Opt_rescue_ignorebadroots !! 661 }
585 btrfs_set_opt(ctx->mou !! 662 break;
586 break; !! 663 case Opt_discard:
587 case Opt_rescue_ignoredatacsum !! 664 btrfs_set_and_info(info, DISCARD,
588 btrfs_set_opt(ctx->mou !! 665 "turning on discard");
589 break; !! 666 break;
590 case Opt_rescue_ignoremetacsum !! 667 case Opt_nodiscard:
591 btrfs_set_opt(ctx->mou !! 668 btrfs_clear_and_info(info, DISCARD,
592 break; !! 669 "turning off discard");
593 case Opt_rescue_ignoresuperfla !! 670 break;
594 btrfs_set_opt(ctx->mou !! 671 case Opt_space_cache:
595 break; !! 672 case Opt_space_cache_version:
596 case Opt_rescue_parameter_all: !! 673 if (token == Opt_space_cache ||
597 btrfs_set_opt(ctx->mou !! 674 strcmp(args[0].from, "v1") == 0) {
598 btrfs_set_opt(ctx->mou !! 675 btrfs_clear_opt(info->mount_opt,
599 btrfs_set_opt(ctx->mou !! 676 FREE_SPACE_TREE);
600 btrfs_set_opt(ctx->mou !! 677 btrfs_set_and_info(info, SPACE_CACHE,
601 btrfs_set_opt(ctx->mou !! 678 "enabling disk space caching");
>> 679 } else if (strcmp(args[0].from, "v2") == 0) {
>> 680 btrfs_clear_opt(info->mount_opt,
>> 681 SPACE_CACHE);
>> 682 btrfs_set_and_info(info, FREE_SPACE_TREE,
>> 683 "enabling free space tree");
>> 684 } else {
>> 685 ret = -EINVAL;
>> 686 goto out;
>> 687 }
>> 688 break;
>> 689 case Opt_rescan_uuid_tree:
>> 690 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
>> 691 break;
>> 692 case Opt_no_space_cache:
>> 693 if (btrfs_test_opt(info, SPACE_CACHE)) {
>> 694 btrfs_clear_and_info(info, SPACE_CACHE,
>> 695 "disabling disk space caching");
>> 696 }
>> 697 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
>> 698 btrfs_clear_and_info(info, FREE_SPACE_TREE,
>> 699 "disabling free space tree");
>> 700 }
>> 701 break;
>> 702 case Opt_inode_cache:
>> 703 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
>> 704 "enabling inode map caching");
>> 705 break;
>> 706 case Opt_noinode_cache:
>> 707 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
>> 708 "disabling inode map caching");
>> 709 break;
>> 710 case Opt_clear_cache:
>> 711 btrfs_set_and_info(info, CLEAR_CACHE,
>> 712 "force clearing of disk cache");
>> 713 break;
>> 714 case Opt_user_subvol_rm_allowed:
>> 715 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
>> 716 break;
>> 717 case Opt_enospc_debug:
>> 718 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
>> 719 break;
>> 720 case Opt_noenospc_debug:
>> 721 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
>> 722 break;
>> 723 case Opt_defrag:
>> 724 btrfs_set_and_info(info, AUTO_DEFRAG,
>> 725 "enabling auto defrag");
>> 726 break;
>> 727 case Opt_nodefrag:
>> 728 btrfs_clear_and_info(info, AUTO_DEFRAG,
>> 729 "disabling auto defrag");
>> 730 break;
>> 731 case Opt_recovery:
>> 732 btrfs_warn(info,
>> 733 "'recovery' is deprecated, use 'usebackuproot' instead");
>> 734 case Opt_usebackuproot:
>> 735 btrfs_info(info,
>> 736 "trying to use backup root at mount time");
>> 737 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
>> 738 break;
>> 739 case Opt_skip_balance:
>> 740 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
>> 741 break;
>> 742 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 743 case Opt_check_integrity_including_extent_data:
>> 744 btrfs_info(info,
>> 745 "enabling check integrity including extent data");
>> 746 btrfs_set_opt(info->mount_opt,
>> 747 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
>> 748 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
>> 749 break;
>> 750 case Opt_check_integrity:
>> 751 btrfs_info(info, "enabling check integrity");
>> 752 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
>> 753 break;
>> 754 case Opt_check_integrity_print_mask:
>> 755 ret = match_int(&args[0], &intarg);
>> 756 if (ret) {
>> 757 goto out;
>> 758 } else if (intarg >= 0) {
>> 759 info->check_integrity_print_mask = intarg;
>> 760 btrfs_info(info,
>> 761 "check_integrity_print_mask 0x%x",
>> 762 info->check_integrity_print_mask);
>> 763 } else {
>> 764 ret = -EINVAL;
>> 765 goto out;
>> 766 }
>> 767 break;
>> 768 #else
>> 769 case Opt_check_integrity_including_extent_data:
>> 770 case Opt_check_integrity:
>> 771 case Opt_check_integrity_print_mask:
>> 772 btrfs_err(info,
>> 773 "support for check_integrity* not compiled in!");
>> 774 ret = -EINVAL;
>> 775 goto out;
>> 776 #endif
>> 777 case Opt_fatal_errors:
>> 778 if (strcmp(args[0].from, "panic") == 0)
>> 779 btrfs_set_opt(info->mount_opt,
>> 780 PANIC_ON_FATAL_ERROR);
>> 781 else if (strcmp(args[0].from, "bug") == 0)
>> 782 btrfs_clear_opt(info->mount_opt,
>> 783 PANIC_ON_FATAL_ERROR);
>> 784 else {
>> 785 ret = -EINVAL;
>> 786 goto out;
>> 787 }
>> 788 break;
>> 789 case Opt_commit_interval:
>> 790 intarg = 0;
>> 791 ret = match_int(&args[0], &intarg);
>> 792 if (ret < 0) {
>> 793 btrfs_err(info, "invalid commit interval");
>> 794 ret = -EINVAL;
>> 795 goto out;
>> 796 }
>> 797 if (intarg > 0) {
>> 798 if (intarg > 300) {
>> 799 btrfs_warn(info,
>> 800 "excessive commit interval %d",
>> 801 intarg);
>> 802 }
>> 803 info->commit_interval = intarg;
>> 804 } else {
>> 805 btrfs_info(info,
>> 806 "using default commit interval %ds",
>> 807 BTRFS_DEFAULT_COMMIT_INTERVAL);
>> 808 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
>> 809 }
602 break; 810 break;
603 default: <<
604 btrfs_info(NULL, "unre <<
605 param->key) <<
606 return -EINVAL; <<
607 } <<
608 break; <<
609 #ifdef CONFIG_BTRFS_DEBUG 811 #ifdef CONFIG_BTRFS_DEBUG
610 case Opt_fragment: !! 812 case Opt_fragment_all:
611 switch (result.uint_32) { !! 813 btrfs_info(info, "fragmenting all space");
612 case Opt_fragment_parameter_al !! 814 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
613 btrfs_set_opt(ctx->mou !! 815 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
614 btrfs_set_opt(ctx->mou <<
615 break; 816 break;
616 case Opt_fragment_parameter_me !! 817 case Opt_fragment_metadata:
617 btrfs_set_opt(ctx->mou !! 818 btrfs_info(info, "fragmenting metadata");
>> 819 btrfs_set_opt(info->mount_opt,
>> 820 FRAGMENT_METADATA);
618 break; 821 break;
619 case Opt_fragment_parameter_da !! 822 case Opt_fragment_data:
620 btrfs_set_opt(ctx->mou !! 823 btrfs_info(info, "fragmenting data");
>> 824 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
621 break; 825 break;
>> 826 #endif
>> 827 case Opt_err:
>> 828 btrfs_info(info, "unrecognized mount option '%s'", p);
>> 829 ret = -EINVAL;
>> 830 goto out;
622 default: 831 default:
623 btrfs_info(NULL, "unre !! 832 break;
624 param->key) <<
625 return -EINVAL; <<
626 } 833 }
627 break; <<
628 #endif <<
629 #ifdef CONFIG_BTRFS_FS_REF_VERIFY <<
630 case Opt_ref_verify: <<
631 btrfs_set_opt(ctx->mount_opt, <<
632 break; <<
633 #endif <<
634 default: <<
635 btrfs_err(NULL, "unrecognized <<
636 return -EINVAL; <<
637 } 834 }
638 !! 835 check:
639 return 0; !! 836 /*
640 } !! 837 * Extra check for current option against current flag
641 !! 838 */
642 /* !! 839 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
643 * Some options only have meaning at mount tim !! 840 btrfs_err(info,
644 * remounts, or be displayed. Clear these at t !! 841 "nologreplay must be used with ro mount option");
645 * paths. !! 842 ret = -EINVAL;
646 */ <<
647 static void btrfs_clear_oneshot_options(struct <<
648 { <<
649 btrfs_clear_opt(fs_info->mount_opt, US <<
650 btrfs_clear_opt(fs_info->mount_opt, CL <<
651 btrfs_clear_opt(fs_info->mount_opt, NO <<
652 } <<
653 <<
654 static bool check_ro_option(const struct btrfs <<
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 } 843 }
663 return false; !! 844 out:
664 } <<
665 <<
666 bool btrfs_check_options(const struct btrfs_fs <<
667 unsigned long long *m <<
668 unsigned long flags) <<
669 { <<
670 bool ret = true; <<
671 <<
672 if (!(flags & SB_RDONLY) && <<
673 (check_ro_option(info, *mount_opt, <<
674 check_ro_option(info, *mount_opt, <<
675 check_ro_option(info, *mount_opt, <<
676 check_ro_option(info, *mount_opt, <<
677 check_ro_option(info, *mount_opt, <<
678 ret = false; <<
679 <<
680 if (btrfs_fs_compat_ro(info, FREE_SPAC 845 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
681 !btrfs_raw_test_opt(*mount_opt, FR !! 846 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
682 !btrfs_raw_test_opt(*mount_opt, CL !! 847 !btrfs_test_opt(info, CLEAR_CACHE)) {
683 btrfs_err(info, "cannot disabl !! 848 btrfs_err(info, "cannot disable free space tree");
684 ret = false; !! 849 ret = -EINVAL;
685 } <<
686 if (btrfs_fs_compat_ro(info, BLOCK_GRO <<
687 !btrfs_raw_test_opt(*mount_opt, F <<
688 btrfs_err(info, "cannot disabl <<
689 ret = false; <<
690 } <<
691 <<
692 if (btrfs_check_mountopts_zoned(info, <<
693 ret = false; <<
694 <<
695 if (!test_bit(BTRFS_FS_STATE_REMOUNTIN <<
696 if (btrfs_raw_test_opt(*mount_ <<
697 btrfs_info(info, "disk <<
698 btrfs_warn(info, <<
699 "space cache v1 is being deprecated and will b <<
700 } <<
701 if (btrfs_raw_test_opt(*mount_ <<
702 btrfs_info(info, "usin <<
703 } <<
704 850
>> 851 }
>> 852 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
>> 853 btrfs_info(info, "disk space caching is enabled");
>> 854 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
>> 855 btrfs_info(info, "using free space tree");
>> 856 kfree(orig);
705 return ret; 857 return ret;
706 } 858 }
707 859
708 /* 860 /*
709 * This is subtle, we only call this during op !! 861 * Parse mount options that are required early in the mount process.
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 * 862 *
714 * This isn't a change in behavior, because we !! 863 * All other options will be parsed on much later in the mount process and
715 * file system to set the current mount option !! 864 * only when we need to allocate a new super block.
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 */ 865 */
720 void btrfs_set_free_space_cache_settings(struc !! 866 static int btrfs_parse_early_options(const char *options, fmode_t flags,
721 { !! 867 void *holder, char **subvol_name, u64 *subvol_objectid,
722 if (fs_info->sectorsize < PAGE_SIZE) { !! 868 struct btrfs_fs_devices **fs_devices)
723 btrfs_clear_opt(fs_info->mount !! 869 {
724 if (!btrfs_test_opt(fs_info, F !! 870 substring_t args[MAX_OPT_ARGS];
725 btrfs_info(fs_info, !! 871 char *device_name, *opts, *orig, *p;
726 "forcing fr !! 872 char *num = NULL;
727 fs_info->se !! 873 int error = 0;
728 btrfs_set_opt(fs_info- <<
729 } <<
730 } <<
731 <<
732 /* <<
733 * At this point our mount options are <<
734 * these settings if we don't have any <<
735 */ <<
736 if (btrfs_test_opt(fs_info, FREE_SPACE <<
737 return; <<
738 <<
739 if (btrfs_is_zoned(fs_info) && <<
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 <<
746 if (btrfs_test_opt(fs_info, SPACE_CACH <<
747 return; <<
748 874
749 if (btrfs_test_opt(fs_info, NOSPACECAC !! 875 if (!options)
750 return; !! 876 return 0;
751 877
752 /* 878 /*
753 * At this point we don't have explici !! 879 * strsep changes the string, duplicate it because parse_options
754 * them ourselves based on the state o !! 880 * gets called twice
755 */ 881 */
756 if (btrfs_fs_compat_ro(fs_info, FREE_S !! 882 opts = kstrdup(options, GFP_KERNEL);
757 btrfs_set_opt(fs_info->mount_o !! 883 if (!opts)
758 else if (btrfs_free_space_cache_v1_act !! 884 return -ENOMEM;
759 btrfs_set_opt(fs_info->mount_o !! 885 orig = opts;
760 } <<
761 886
762 static void set_device_specific_options(struct !! 887 while ((p = strsep(&opts, ",")) != NULL) {
763 { !! 888 int token;
764 if (!btrfs_test_opt(fs_info, NOSSD) && !! 889 if (!*p)
765 !fs_info->fs_devices->rotating) !! 890 continue;
766 btrfs_set_opt(fs_info->mount_o <<
767 891
768 /* !! 892 token = match_token(p, tokens, args);
769 * For devices supporting discard turn !! 893 switch (token) {
770 * unless it's already set or disabled !! 894 case Opt_subvol:
771 * nodiscard for the same mount. !! 895 kfree(*subvol_name);
772 * !! 896 *subvol_name = match_strdup(&args[0]);
773 * The zoned mode piggy backs on the d !! 897 if (!*subvol_name) {
774 * resetting a zone. There is no reaso !! 898 error = -ENOMEM;
775 * fast enough. So, do not enable asyn !! 899 goto out;
776 */ !! 900 }
777 if (!(btrfs_test_opt(fs_info, DISCARD_ !! 901 break;
778 btrfs_test_opt(fs_info, DISCARD_ !! 902 case Opt_subvolid:
779 btrfs_test_opt(fs_info, NODISCAR !! 903 num = match_strdup(&args[0]);
780 fs_info->fs_devices->discardable & !! 904 if (num) {
781 !btrfs_is_zoned(fs_info)) !! 905 *subvol_objectid = memparse(num, NULL);
782 btrfs_set_opt(fs_info->mount_o !! 906 kfree(num);
>> 907 /* we want the original fs_tree */
>> 908 if (!*subvol_objectid)
>> 909 *subvol_objectid =
>> 910 BTRFS_FS_TREE_OBJECTID;
>> 911 } else {
>> 912 error = -EINVAL;
>> 913 goto out;
>> 914 }
>> 915 break;
>> 916 case Opt_subvolrootid:
>> 917 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
>> 918 break;
>> 919 case Opt_device:
>> 920 device_name = match_strdup(&args[0]);
>> 921 if (!device_name) {
>> 922 error = -ENOMEM;
>> 923 goto out;
>> 924 }
>> 925 error = btrfs_scan_one_device(device_name,
>> 926 flags, holder, fs_devices);
>> 927 kfree(device_name);
>> 928 if (error)
>> 929 goto out;
>> 930 break;
>> 931 default:
>> 932 break;
>> 933 }
>> 934 }
>> 935
>> 936 out:
>> 937 kfree(orig);
>> 938 return error;
783 } 939 }
784 940
785 char *btrfs_get_subvol_name_from_objectid(stru !! 941 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
786 u64 !! 942 u64 subvol_objectid)
787 { 943 {
788 struct btrfs_root *root = fs_info->tre 944 struct btrfs_root *root = fs_info->tree_root;
789 struct btrfs_root *fs_root = NULL; !! 945 struct btrfs_root *fs_root;
790 struct btrfs_root_ref *root_ref; 946 struct btrfs_root_ref *root_ref;
791 struct btrfs_inode_ref *inode_ref; 947 struct btrfs_inode_ref *inode_ref;
792 struct btrfs_key key; 948 struct btrfs_key key;
793 struct btrfs_path *path = NULL; 949 struct btrfs_path *path = NULL;
794 char *name = NULL, *ptr; 950 char *name = NULL, *ptr;
795 u64 dirid; 951 u64 dirid;
796 int len; 952 int len;
797 int ret; 953 int ret;
798 954
799 path = btrfs_alloc_path(); 955 path = btrfs_alloc_path();
800 if (!path) { 956 if (!path) {
801 ret = -ENOMEM; 957 ret = -ENOMEM;
802 goto err; 958 goto err;
803 } 959 }
>> 960 path->leave_spinning = 1;
804 961
805 name = kmalloc(PATH_MAX, GFP_KERNEL); !! 962 name = kmalloc(PATH_MAX, GFP_NOFS);
806 if (!name) { 963 if (!name) {
807 ret = -ENOMEM; 964 ret = -ENOMEM;
808 goto err; 965 goto err;
809 } 966 }
810 ptr = name + PATH_MAX - 1; 967 ptr = name + PATH_MAX - 1;
811 ptr[0] = '\0'; 968 ptr[0] = '\0';
812 969
813 /* 970 /*
814 * Walk up the subvolume trees in the 971 * Walk up the subvolume trees in the tree of tree roots by root
815 * backrefs until we hit the top-level 972 * backrefs until we hit the top-level subvolume.
816 */ 973 */
817 while (subvol_objectid != BTRFS_FS_TRE 974 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
818 key.objectid = subvol_objectid 975 key.objectid = subvol_objectid;
819 key.type = BTRFS_ROOT_BACKREF_ 976 key.type = BTRFS_ROOT_BACKREF_KEY;
820 key.offset = (u64)-1; 977 key.offset = (u64)-1;
821 978
822 ret = btrfs_search_backwards(r !! 979 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
823 if (ret < 0) { 980 if (ret < 0) {
824 goto err; 981 goto err;
825 } else if (ret > 0) { 982 } else if (ret > 0) {
826 ret = -ENOENT; !! 983 ret = btrfs_previous_item(root, path, subvol_objectid,
827 goto err; !! 984 BTRFS_ROOT_BACKREF_KEY);
>> 985 if (ret < 0) {
>> 986 goto err;
>> 987 } else if (ret > 0) {
>> 988 ret = -ENOENT;
>> 989 goto err;
>> 990 }
828 } 991 }
829 992
>> 993 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
830 subvol_objectid = key.offset; 994 subvol_objectid = key.offset;
831 995
832 root_ref = btrfs_item_ptr(path 996 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
833 stru 997 struct btrfs_root_ref);
834 len = btrfs_root_ref_name_len( 998 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
835 ptr -= len + 1; 999 ptr -= len + 1;
836 if (ptr < name) { 1000 if (ptr < name) {
837 ret = -ENAMETOOLONG; 1001 ret = -ENAMETOOLONG;
838 goto err; 1002 goto err;
839 } 1003 }
840 read_extent_buffer(path->nodes 1004 read_extent_buffer(path->nodes[0], ptr + 1,
841 (unsigned l 1005 (unsigned long)(root_ref + 1), len);
842 ptr[0] = '/'; 1006 ptr[0] = '/';
843 dirid = btrfs_root_ref_dirid(p 1007 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
844 btrfs_release_path(path); 1008 btrfs_release_path(path);
845 1009
846 fs_root = btrfs_get_fs_root(fs !! 1010 key.objectid = subvol_objectid;
>> 1011 key.type = BTRFS_ROOT_ITEM_KEY;
>> 1012 key.offset = (u64)-1;
>> 1013 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
847 if (IS_ERR(fs_root)) { 1014 if (IS_ERR(fs_root)) {
848 ret = PTR_ERR(fs_root) 1015 ret = PTR_ERR(fs_root);
849 fs_root = NULL; <<
850 goto err; 1016 goto err;
851 } 1017 }
852 1018
853 /* 1019 /*
854 * Walk up the filesystem tree 1020 * Walk up the filesystem tree by inode refs until we hit the
855 * root directory. 1021 * root directory.
856 */ 1022 */
857 while (dirid != BTRFS_FIRST_FR 1023 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
858 key.objectid = dirid; 1024 key.objectid = dirid;
859 key.type = BTRFS_INODE 1025 key.type = BTRFS_INODE_REF_KEY;
860 key.offset = (u64)-1; 1026 key.offset = (u64)-1;
861 1027
862 ret = btrfs_search_bac !! 1028 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
863 if (ret < 0) { 1029 if (ret < 0) {
864 goto err; 1030 goto err;
865 } else if (ret > 0) { 1031 } else if (ret > 0) {
866 ret = -ENOENT; !! 1032 ret = btrfs_previous_item(fs_root, path, dirid,
867 goto err; !! 1033 BTRFS_INODE_REF_KEY);
>> 1034 if (ret < 0) {
>> 1035 goto err;
>> 1036 } else if (ret > 0) {
>> 1037 ret = -ENOENT;
>> 1038 goto err;
>> 1039 }
868 } 1040 }
869 1041
>> 1042 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
870 dirid = key.offset; 1043 dirid = key.offset;
871 1044
872 inode_ref = btrfs_item 1045 inode_ref = btrfs_item_ptr(path->nodes[0],
873 1046 path->slots[0],
874 1047 struct btrfs_inode_ref);
875 len = btrfs_inode_ref_ 1048 len = btrfs_inode_ref_name_len(path->nodes[0],
876 1049 inode_ref);
877 ptr -= len + 1; 1050 ptr -= len + 1;
878 if (ptr < name) { 1051 if (ptr < name) {
879 ret = -ENAMETO 1052 ret = -ENAMETOOLONG;
880 goto err; 1053 goto err;
881 } 1054 }
882 read_extent_buffer(pat 1055 read_extent_buffer(path->nodes[0], ptr + 1,
883 (un 1056 (unsigned long)(inode_ref + 1), len);
884 ptr[0] = '/'; 1057 ptr[0] = '/';
885 btrfs_release_path(pat 1058 btrfs_release_path(path);
886 } 1059 }
887 btrfs_put_root(fs_root); <<
888 fs_root = NULL; <<
889 } 1060 }
890 1061
891 btrfs_free_path(path); 1062 btrfs_free_path(path);
892 if (ptr == name + PATH_MAX - 1) { 1063 if (ptr == name + PATH_MAX - 1) {
893 name[0] = '/'; 1064 name[0] = '/';
894 name[1] = '\0'; 1065 name[1] = '\0';
895 } else { 1066 } else {
896 memmove(name, ptr, name + PATH 1067 memmove(name, ptr, name + PATH_MAX - ptr);
897 } 1068 }
898 return name; 1069 return name;
899 1070
900 err: 1071 err:
901 btrfs_put_root(fs_root); <<
902 btrfs_free_path(path); 1072 btrfs_free_path(path);
903 kfree(name); 1073 kfree(name);
904 return ERR_PTR(ret); 1074 return ERR_PTR(ret);
905 } 1075 }
906 1076
907 static int get_default_subvol_objectid(struct 1077 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
908 { 1078 {
909 struct btrfs_root *root = fs_info->tre 1079 struct btrfs_root *root = fs_info->tree_root;
910 struct btrfs_dir_item *di; 1080 struct btrfs_dir_item *di;
911 struct btrfs_path *path; 1081 struct btrfs_path *path;
912 struct btrfs_key location; 1082 struct btrfs_key location;
913 struct fscrypt_str name = FSTR_INIT("d <<
914 u64 dir_id; 1083 u64 dir_id;
915 1084
916 path = btrfs_alloc_path(); 1085 path = btrfs_alloc_path();
917 if (!path) 1086 if (!path)
918 return -ENOMEM; 1087 return -ENOMEM;
>> 1088 path->leave_spinning = 1;
919 1089
920 /* 1090 /*
921 * Find the "default" dir item which p 1091 * Find the "default" dir item which points to the root item that we
922 * will mount by default if we haven't 1092 * will mount by default if we haven't been given a specific subvolume
923 * to mount. 1093 * to mount.
924 */ 1094 */
925 dir_id = btrfs_super_root_dir(fs_info- 1095 dir_id = btrfs_super_root_dir(fs_info->super_copy);
926 di = btrfs_lookup_dir_item(NULL, root, !! 1096 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
927 if (IS_ERR(di)) { 1097 if (IS_ERR(di)) {
928 btrfs_free_path(path); 1098 btrfs_free_path(path);
929 return PTR_ERR(di); 1099 return PTR_ERR(di);
930 } 1100 }
931 if (!di) { 1101 if (!di) {
932 /* 1102 /*
933 * Ok the default dir item isn 1103 * Ok the default dir item isn't there. This is weird since
934 * it's always been there, but 1104 * it's always been there, but don't freak out, just try and
935 * mount the top-level subvolu 1105 * mount the top-level subvolume.
936 */ 1106 */
937 btrfs_free_path(path); 1107 btrfs_free_path(path);
938 *objectid = BTRFS_FS_TREE_OBJE 1108 *objectid = BTRFS_FS_TREE_OBJECTID;
939 return 0; 1109 return 0;
940 } 1110 }
941 1111
942 btrfs_dir_item_key_to_cpu(path->nodes[ 1112 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
943 btrfs_free_path(path); 1113 btrfs_free_path(path);
944 *objectid = location.objectid; 1114 *objectid = location.objectid;
945 return 0; 1115 return 0;
946 } 1116 }
947 1117
948 static int btrfs_fill_super(struct super_block 1118 static int btrfs_fill_super(struct super_block *sb,
949 struct btrfs_fs_de 1119 struct btrfs_fs_devices *fs_devices,
950 void *data) 1120 void *data)
951 { 1121 {
952 struct inode *inode; 1122 struct inode *inode;
953 struct btrfs_fs_info *fs_info = btrfs_ 1123 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 1124 struct btrfs_key key;
954 int err; 1125 int err;
955 1126
956 sb->s_maxbytes = MAX_LFS_FILESIZE; 1127 sb->s_maxbytes = MAX_LFS_FILESIZE;
957 sb->s_magic = BTRFS_SUPER_MAGIC; 1128 sb->s_magic = BTRFS_SUPER_MAGIC;
958 sb->s_op = &btrfs_super_ops; 1129 sb->s_op = &btrfs_super_ops;
959 sb->s_d_op = &btrfs_dentry_operations; 1130 sb->s_d_op = &btrfs_dentry_operations;
960 sb->s_export_op = &btrfs_export_ops; 1131 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; 1132 sb->s_xattr = btrfs_xattr_handlers;
965 sb->s_time_gran = 1; 1133 sb->s_time_gran = 1;
>> 1134 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
>> 1135 sb->s_flags |= MS_POSIXACL;
>> 1136 #endif
>> 1137 sb->s_flags |= MS_I_VERSION;
966 sb->s_iflags |= SB_I_CGROUPWB; 1138 sb->s_iflags |= SB_I_CGROUPWB;
967 <<
968 err = super_setup_bdi(sb); <<
969 if (err) { <<
970 btrfs_err(fs_info, "super_setu <<
971 return err; <<
972 } <<
973 <<
974 err = open_ctree(sb, fs_devices, (char 1139 err = open_ctree(sb, fs_devices, (char *)data);
975 if (err) { 1140 if (err) {
976 btrfs_err(fs_info, "open_ctree 1141 btrfs_err(fs_info, "open_ctree failed");
977 return err; 1142 return err;
978 } 1143 }
979 1144
980 inode = btrfs_iget(BTRFS_FIRST_FREE_OB !! 1145 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
>> 1146 key.type = BTRFS_INODE_ITEM_KEY;
>> 1147 key.offset = 0;
>> 1148 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
981 if (IS_ERR(inode)) { 1149 if (IS_ERR(inode)) {
982 err = PTR_ERR(inode); 1150 err = PTR_ERR(inode);
983 btrfs_handle_fs_error(fs_info, <<
984 goto fail_close; 1151 goto fail_close;
985 } 1152 }
986 1153
987 sb->s_root = d_make_root(inode); 1154 sb->s_root = d_make_root(inode);
988 if (!sb->s_root) { 1155 if (!sb->s_root) {
989 err = -ENOMEM; 1156 err = -ENOMEM;
990 goto fail_close; 1157 goto fail_close;
991 } 1158 }
992 1159
993 sb->s_flags |= SB_ACTIVE; !! 1160 save_mount_options(sb, data);
>> 1161 cleancache_init_fs(sb);
>> 1162 sb->s_flags |= MS_ACTIVE;
994 return 0; 1163 return 0;
995 1164
996 fail_close: 1165 fail_close:
997 close_ctree(fs_info); 1166 close_ctree(fs_info);
998 return err; 1167 return err;
999 } 1168 }
1000 1169
1001 int btrfs_sync_fs(struct super_block *sb, int 1170 int btrfs_sync_fs(struct super_block *sb, int wait)
1002 { 1171 {
1003 struct btrfs_trans_handle *trans; 1172 struct btrfs_trans_handle *trans;
1004 struct btrfs_fs_info *fs_info = btrfs 1173 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1005 struct btrfs_root *root = fs_info->tr 1174 struct btrfs_root *root = fs_info->tree_root;
1006 1175
1007 trace_btrfs_sync_fs(fs_info, wait); 1176 trace_btrfs_sync_fs(fs_info, wait);
1008 1177
1009 if (!wait) { 1178 if (!wait) {
1010 filemap_flush(fs_info->btree_ 1179 filemap_flush(fs_info->btree_inode->i_mapping);
1011 return 0; 1180 return 0;
1012 } 1181 }
1013 1182
1014 btrfs_wait_ordered_roots(fs_info, U64 !! 1183 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1015 1184
1016 trans = btrfs_attach_transaction_barr 1185 trans = btrfs_attach_transaction_barrier(root);
1017 if (IS_ERR(trans)) { 1186 if (IS_ERR(trans)) {
1018 /* no transaction, don't both 1187 /* no transaction, don't bother */
1019 if (PTR_ERR(trans) == -ENOENT 1188 if (PTR_ERR(trans) == -ENOENT) {
1020 /* 1189 /*
1021 * Exit unless we hav 1190 * Exit unless we have some pending changes
1022 * that need to go th 1191 * that need to go through commit
1023 */ 1192 */
1024 if (!test_bit(BTRFS_F !! 1193 if (fs_info->pending_changes == 0)
1025 &fs_inf <<
1026 return 0; 1194 return 0;
1027 /* 1195 /*
1028 * A non-blocking tes 1196 * A non-blocking test if the fs is frozen. We must not
1029 * start a new transa 1197 * start a new transaction here otherwise a deadlock
1030 * happens. The pendi 1198 * happens. The pending operations are delayed to the
1031 * next commit after 1199 * next commit after thawing.
1032 */ 1200 */
1033 if (sb_start_write_tr !! 1201 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1034 sb_end_write( !! 1202 __sb_end_write(sb, SB_FREEZE_WRITE);
1035 else 1203 else
1036 return 0; 1204 return 0;
1037 trans = btrfs_start_t 1205 trans = btrfs_start_transaction(root, 0);
1038 } 1206 }
1039 if (IS_ERR(trans)) 1207 if (IS_ERR(trans))
1040 return PTR_ERR(trans) 1208 return PTR_ERR(trans);
1041 } 1209 }
1042 return btrfs_commit_transaction(trans 1210 return btrfs_commit_transaction(trans);
1043 } 1211 }
1044 1212
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 1213 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1052 { 1214 {
1053 struct btrfs_fs_info *info = btrfs_sb 1215 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1054 const char *compress_type; !! 1216 char *compress_type;
1055 const char *subvol_name; <<
1056 bool printed = false; <<
1057 1217
1058 if (btrfs_test_opt(info, DEGRADED)) 1218 if (btrfs_test_opt(info, DEGRADED))
1059 seq_puts(seq, ",degraded"); 1219 seq_puts(seq, ",degraded");
1060 if (btrfs_test_opt(info, NODATASUM)) 1220 if (btrfs_test_opt(info, NODATASUM))
1061 seq_puts(seq, ",nodatasum"); 1221 seq_puts(seq, ",nodatasum");
1062 if (btrfs_test_opt(info, NODATACOW)) 1222 if (btrfs_test_opt(info, NODATACOW))
1063 seq_puts(seq, ",nodatacow"); 1223 seq_puts(seq, ",nodatacow");
1064 if (btrfs_test_opt(info, NOBARRIER)) 1224 if (btrfs_test_opt(info, NOBARRIER))
1065 seq_puts(seq, ",nobarrier"); 1225 seq_puts(seq, ",nobarrier");
1066 if (info->max_inline != BTRFS_DEFAULT 1226 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1067 seq_printf(seq, ",max_inline= 1227 seq_printf(seq, ",max_inline=%llu", info->max_inline);
>> 1228 if (info->alloc_start != 0)
>> 1229 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1068 if (info->thread_pool_size != min_t( 1230 if (info->thread_pool_size != min_t(unsigned long,
1069 1231 num_online_cpus() + 2, 8))
1070 seq_printf(seq, ",thread_pool !! 1232 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1071 if (btrfs_test_opt(info, COMPRESS)) { 1233 if (btrfs_test_opt(info, COMPRESS)) {
1072 compress_type = btrfs_compres !! 1234 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
>> 1235 compress_type = "zlib";
>> 1236 else
>> 1237 compress_type = "lzo";
1073 if (btrfs_test_opt(info, FORC 1238 if (btrfs_test_opt(info, FORCE_COMPRESS))
1074 seq_printf(seq, ",com 1239 seq_printf(seq, ",compress-force=%s", compress_type);
1075 else 1240 else
1076 seq_printf(seq, ",com 1241 seq_printf(seq, ",compress=%s", compress_type);
1077 if (info->compress_level) <<
1078 seq_printf(seq, ":%d" <<
1079 } 1242 }
1080 if (btrfs_test_opt(info, NOSSD)) 1243 if (btrfs_test_opt(info, NOSSD))
1081 seq_puts(seq, ",nossd"); 1244 seq_puts(seq, ",nossd");
1082 if (btrfs_test_opt(info, SSD_SPREAD)) 1245 if (btrfs_test_opt(info, SSD_SPREAD))
1083 seq_puts(seq, ",ssd_spread"); 1246 seq_puts(seq, ",ssd_spread");
1084 else if (btrfs_test_opt(info, SSD)) 1247 else if (btrfs_test_opt(info, SSD))
1085 seq_puts(seq, ",ssd"); 1248 seq_puts(seq, ",ssd");
1086 if (btrfs_test_opt(info, NOTREELOG)) 1249 if (btrfs_test_opt(info, NOTREELOG))
1087 seq_puts(seq, ",notreelog"); 1250 seq_puts(seq, ",notreelog");
1088 if (btrfs_test_opt(info, NOLOGREPLAY) 1251 if (btrfs_test_opt(info, NOLOGREPLAY))
1089 print_rescue_option(seq, "nol !! 1252 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 1253 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1101 seq_puts(seq, ",flushoncommit 1254 seq_puts(seq, ",flushoncommit");
1102 if (btrfs_test_opt(info, DISCARD_SYNC !! 1255 if (btrfs_test_opt(info, DISCARD))
1103 seq_puts(seq, ",discard"); 1256 seq_puts(seq, ",discard");
1104 if (btrfs_test_opt(info, DISCARD_ASYN !! 1257 if (!(info->sb->s_flags & MS_POSIXACL))
1105 seq_puts(seq, ",discard=async <<
1106 if (!(info->sb->s_flags & SB_POSIXACL <<
1107 seq_puts(seq, ",noacl"); 1258 seq_puts(seq, ",noacl");
1108 if (btrfs_free_space_cache_v1_active( !! 1259 if (btrfs_test_opt(info, SPACE_CACHE))
1109 seq_puts(seq, ",space_cache") 1260 seq_puts(seq, ",space_cache");
1110 else if (btrfs_fs_compat_ro(info, FRE !! 1261 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1111 seq_puts(seq, ",space_cache=v 1262 seq_puts(seq, ",space_cache=v2");
1112 else 1263 else
1113 seq_puts(seq, ",nospace_cache 1264 seq_puts(seq, ",nospace_cache");
1114 if (btrfs_test_opt(info, RESCAN_UUID_ 1265 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1115 seq_puts(seq, ",rescan_uuid_t 1266 seq_puts(seq, ",rescan_uuid_tree");
1116 if (btrfs_test_opt(info, CLEAR_CACHE) 1267 if (btrfs_test_opt(info, CLEAR_CACHE))
1117 seq_puts(seq, ",clear_cache") 1268 seq_puts(seq, ",clear_cache");
1118 if (btrfs_test_opt(info, USER_SUBVOL_ 1269 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1119 seq_puts(seq, ",user_subvol_r 1270 seq_puts(seq, ",user_subvol_rm_allowed");
1120 if (btrfs_test_opt(info, ENOSPC_DEBUG 1271 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1121 seq_puts(seq, ",enospc_debug" 1272 seq_puts(seq, ",enospc_debug");
1122 if (btrfs_test_opt(info, AUTO_DEFRAG) 1273 if (btrfs_test_opt(info, AUTO_DEFRAG))
1123 seq_puts(seq, ",autodefrag"); 1274 seq_puts(seq, ",autodefrag");
>> 1275 if (btrfs_test_opt(info, INODE_MAP_CACHE))
>> 1276 seq_puts(seq, ",inode_cache");
1124 if (btrfs_test_opt(info, SKIP_BALANCE 1277 if (btrfs_test_opt(info, SKIP_BALANCE))
1125 seq_puts(seq, ",skip_balance" 1278 seq_puts(seq, ",skip_balance");
>> 1279 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 1280 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
>> 1281 seq_puts(seq, ",check_int_data");
>> 1282 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
>> 1283 seq_puts(seq, ",check_int");
>> 1284 if (info->check_integrity_print_mask)
>> 1285 seq_printf(seq, ",check_int_print_mask=%d",
>> 1286 info->check_integrity_print_mask);
>> 1287 #endif
1126 if (info->metadata_ratio) 1288 if (info->metadata_ratio)
1127 seq_printf(seq, ",metadata_ra !! 1289 seq_printf(seq, ",metadata_ratio=%d",
>> 1290 info->metadata_ratio);
1128 if (btrfs_test_opt(info, PANIC_ON_FAT 1291 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1129 seq_puts(seq, ",fatal_errors= 1292 seq_puts(seq, ",fatal_errors=panic");
1130 if (info->commit_interval != BTRFS_DE 1293 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1131 seq_printf(seq, ",commit=%u", !! 1294 seq_printf(seq, ",commit=%d", info->commit_interval);
1132 #ifdef CONFIG_BTRFS_DEBUG 1295 #ifdef CONFIG_BTRFS_DEBUG
1133 if (btrfs_test_opt(info, FRAGMENT_DAT 1296 if (btrfs_test_opt(info, FRAGMENT_DATA))
1134 seq_puts(seq, ",fragment=data 1297 seq_puts(seq, ",fragment=data");
1135 if (btrfs_test_opt(info, FRAGMENT_MET 1298 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1136 seq_puts(seq, ",fragment=meta 1299 seq_puts(seq, ",fragment=metadata");
1137 #endif 1300 #endif
1138 if (btrfs_test_opt(info, REF_VERIFY)) !! 1301 seq_printf(seq, ",subvolid=%llu",
1139 seq_puts(seq, ",ref_verify"); !! 1302 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1140 seq_printf(seq, ",subvolid=%llu", btr !! 1303 seq_puts(seq, ",subvol=");
1141 subvol_name = btrfs_get_subvol_name_f !! 1304 seq_dentry(seq, dentry, " \t\n\\");
1142 btrfs_root_id(BTRFS_I <<
1143 if (!IS_ERR(subvol_name)) { <<
1144 seq_puts(seq, ",subvol="); <<
1145 seq_escape(seq, subvol_name, <<
1146 kfree(subvol_name); <<
1147 } <<
1148 return 0; 1305 return 0;
1149 } 1306 }
1150 1307
>> 1308 static int btrfs_test_super(struct super_block *s, void *data)
>> 1309 {
>> 1310 struct btrfs_fs_info *p = data;
>> 1311 struct btrfs_fs_info *fs_info = btrfs_sb(s);
>> 1312
>> 1313 return fs_info->fs_devices == p->fs_devices;
>> 1314 }
>> 1315
>> 1316 static int btrfs_set_super(struct super_block *s, void *data)
>> 1317 {
>> 1318 int err = set_anon_super(s, data);
>> 1319 if (!err)
>> 1320 s->s_fs_info = data;
>> 1321 return err;
>> 1322 }
>> 1323
1151 /* 1324 /*
1152 * subvolumes are identified by ino 256 1325 * subvolumes are identified by ino 256
1153 */ 1326 */
1154 static inline int is_subvolume_inode(struct i 1327 static inline int is_subvolume_inode(struct inode *inode)
1155 { 1328 {
1156 if (inode && inode->i_ino == BTRFS_FI 1329 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1157 return 1; 1330 return 1;
1158 return 0; 1331 return 0;
1159 } 1332 }
1160 1333
>> 1334 /*
>> 1335 * This will add subvolid=0 to the argument string while removing any subvol=
>> 1336 * and subvolid= arguments to make sure we get the top-level root for path
>> 1337 * walking to the subvol we want.
>> 1338 */
>> 1339 static char *setup_root_args(char *args)
>> 1340 {
>> 1341 char *buf, *dst, *sep;
>> 1342
>> 1343 if (!args)
>> 1344 return kstrdup("subvolid=0", GFP_NOFS);
>> 1345
>> 1346 /* The worst case is that we add ",subvolid=0" to the end. */
>> 1347 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
>> 1348 if (!buf)
>> 1349 return NULL;
>> 1350
>> 1351 while (1) {
>> 1352 sep = strchrnul(args, ',');
>> 1353 if (!strstarts(args, "subvol=") &&
>> 1354 !strstarts(args, "subvolid=")) {
>> 1355 memcpy(dst, args, sep - args);
>> 1356 dst += sep - args;
>> 1357 *dst++ = ',';
>> 1358 }
>> 1359 if (*sep)
>> 1360 args = sep + 1;
>> 1361 else
>> 1362 break;
>> 1363 }
>> 1364 strcpy(dst, "subvolid=0");
>> 1365
>> 1366 return buf;
>> 1367 }
>> 1368
1161 static struct dentry *mount_subvol(const char 1369 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1162 struct vfs !! 1370 int flags, const char *device_name,
>> 1371 char *data)
1163 { 1372 {
1164 struct dentry *root; 1373 struct dentry *root;
>> 1374 struct vfsmount *mnt = NULL;
>> 1375 char *newargs;
1165 int ret; 1376 int ret;
1166 1377
>> 1378 newargs = setup_root_args(data);
>> 1379 if (!newargs) {
>> 1380 root = ERR_PTR(-ENOMEM);
>> 1381 goto out;
>> 1382 }
>> 1383
>> 1384 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
>> 1385 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
>> 1386 if (flags & MS_RDONLY) {
>> 1387 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
>> 1388 device_name, newargs);
>> 1389 } else {
>> 1390 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
>> 1391 device_name, newargs);
>> 1392 if (IS_ERR(mnt)) {
>> 1393 root = ERR_CAST(mnt);
>> 1394 mnt = NULL;
>> 1395 goto out;
>> 1396 }
>> 1397
>> 1398 down_write(&mnt->mnt_sb->s_umount);
>> 1399 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
>> 1400 up_write(&mnt->mnt_sb->s_umount);
>> 1401 if (ret < 0) {
>> 1402 root = ERR_PTR(ret);
>> 1403 goto out;
>> 1404 }
>> 1405 }
>> 1406 }
>> 1407 if (IS_ERR(mnt)) {
>> 1408 root = ERR_CAST(mnt);
>> 1409 mnt = NULL;
>> 1410 goto out;
>> 1411 }
>> 1412
1167 if (!subvol_name) { 1413 if (!subvol_name) {
1168 if (!subvol_objectid) { 1414 if (!subvol_objectid) {
1169 ret = get_default_sub 1415 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1170 1416 &subvol_objectid);
1171 if (ret) { 1417 if (ret) {
1172 root = ERR_PT 1418 root = ERR_PTR(ret);
1173 goto out; 1419 goto out;
1174 } 1420 }
1175 } 1421 }
1176 subvol_name = btrfs_get_subvo !! 1422 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1177 btrfs !! 1423 subvol_objectid);
1178 if (IS_ERR(subvol_name)) { 1424 if (IS_ERR(subvol_name)) {
1179 root = ERR_CAST(subvo 1425 root = ERR_CAST(subvol_name);
1180 subvol_name = NULL; 1426 subvol_name = NULL;
1181 goto out; 1427 goto out;
1182 } 1428 }
1183 1429
1184 } 1430 }
1185 1431
1186 root = mount_subtree(mnt, subvol_name 1432 root = mount_subtree(mnt, subvol_name);
1187 /* mount_subtree() drops our referenc 1433 /* mount_subtree() drops our reference on the vfsmount. */
1188 mnt = NULL; 1434 mnt = NULL;
1189 1435
1190 if (!IS_ERR(root)) { 1436 if (!IS_ERR(root)) {
1191 struct super_block *s = root- 1437 struct super_block *s = root->d_sb;
1192 struct btrfs_fs_info *fs_info 1438 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1193 struct inode *root_inode = d_ 1439 struct inode *root_inode = d_inode(root);
1194 u64 root_objectid = btrfs_roo !! 1440 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1195 1441
1196 ret = 0; 1442 ret = 0;
1197 if (!is_subvolume_inode(root_ 1443 if (!is_subvolume_inode(root_inode)) {
1198 btrfs_err(fs_info, "' 1444 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1199 subvol_name); 1445 subvol_name);
1200 ret = -EINVAL; 1446 ret = -EINVAL;
1201 } 1447 }
1202 if (subvol_objectid && root_o 1448 if (subvol_objectid && root_objectid != subvol_objectid) {
1203 /* 1449 /*
1204 * This will also cat 1450 * This will also catch a race condition where a
1205 * subvolume which wa 1451 * subvolume which was passed by ID is renamed and
1206 * another subvolume 1452 * another subvolume is renamed over the old location.
1207 */ 1453 */
1208 btrfs_err(fs_info, 1454 btrfs_err(fs_info,
1209 "subvol '%s 1455 "subvol '%s' does not match subvolid %llu",
1210 subvol_name 1456 subvol_name, subvol_objectid);
1211 ret = -EINVAL; 1457 ret = -EINVAL;
1212 } 1458 }
1213 if (ret) { 1459 if (ret) {
1214 dput(root); 1460 dput(root);
1215 root = ERR_PTR(ret); 1461 root = ERR_PTR(ret);
1216 deactivate_locked_sup 1462 deactivate_locked_super(s);
1217 } 1463 }
1218 } 1464 }
1219 1465
1220 out: 1466 out:
1221 mntput(mnt); 1467 mntput(mnt);
>> 1468 kfree(newargs);
1222 kfree(subvol_name); 1469 kfree(subvol_name);
1223 return root; 1470 return root;
1224 } 1471 }
1225 1472
>> 1473 static int parse_security_options(char *orig_opts,
>> 1474 struct security_mnt_opts *sec_opts)
>> 1475 {
>> 1476 char *secdata = NULL;
>> 1477 int ret = 0;
>> 1478
>> 1479 secdata = alloc_secdata();
>> 1480 if (!secdata)
>> 1481 return -ENOMEM;
>> 1482 ret = security_sb_copy_data(orig_opts, secdata);
>> 1483 if (ret) {
>> 1484 free_secdata(secdata);
>> 1485 return ret;
>> 1486 }
>> 1487 ret = security_sb_parse_opts_str(secdata, sec_opts);
>> 1488 free_secdata(secdata);
>> 1489 return ret;
>> 1490 }
>> 1491
>> 1492 static int setup_security_options(struct btrfs_fs_info *fs_info,
>> 1493 struct super_block *sb,
>> 1494 struct security_mnt_opts *sec_opts)
>> 1495 {
>> 1496 int ret = 0;
>> 1497
>> 1498 /*
>> 1499 * Call security_sb_set_mnt_opts() to check whether new sec_opts
>> 1500 * is valid.
>> 1501 */
>> 1502 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
>> 1503 if (ret)
>> 1504 return ret;
>> 1505
>> 1506 #ifdef CONFIG_SECURITY
>> 1507 if (!fs_info->security_opts.num_mnt_opts) {
>> 1508 /* first time security setup, copy sec_opts to fs_info */
>> 1509 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
>> 1510 } else {
>> 1511 /*
>> 1512 * Since SELinux (the only one supporting security_mnt_opts)
>> 1513 * does NOT support changing context during remount/mount of
>> 1514 * the same sb, this must be the same or part of the same
>> 1515 * security options, just free it.
>> 1516 */
>> 1517 security_free_mnt_opts(sec_opts);
>> 1518 }
>> 1519 #endif
>> 1520 return ret;
>> 1521 }
>> 1522
>> 1523 /*
>> 1524 * Find a superblock for the given device / mount point.
>> 1525 *
>> 1526 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
>> 1527 * for multiple device setup. Make sure to keep it in sync.
>> 1528 */
>> 1529 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
>> 1530 const char *device_name, void *data)
>> 1531 {
>> 1532 struct block_device *bdev = NULL;
>> 1533 struct super_block *s;
>> 1534 struct btrfs_fs_devices *fs_devices = NULL;
>> 1535 struct btrfs_fs_info *fs_info = NULL;
>> 1536 struct security_mnt_opts new_sec_opts;
>> 1537 fmode_t mode = FMODE_READ;
>> 1538 char *subvol_name = NULL;
>> 1539 u64 subvol_objectid = 0;
>> 1540 int error = 0;
>> 1541
>> 1542 if (!(flags & MS_RDONLY))
>> 1543 mode |= FMODE_WRITE;
>> 1544
>> 1545 error = btrfs_parse_early_options(data, mode, fs_type,
>> 1546 &subvol_name, &subvol_objectid,
>> 1547 &fs_devices);
>> 1548 if (error) {
>> 1549 kfree(subvol_name);
>> 1550 return ERR_PTR(error);
>> 1551 }
>> 1552
>> 1553 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
>> 1554 /* mount_subvol() will free subvol_name. */
>> 1555 return mount_subvol(subvol_name, subvol_objectid, flags,
>> 1556 device_name, data);
>> 1557 }
>> 1558
>> 1559 security_init_mnt_opts(&new_sec_opts);
>> 1560 if (data) {
>> 1561 error = parse_security_options(data, &new_sec_opts);
>> 1562 if (error)
>> 1563 return ERR_PTR(error);
>> 1564 }
>> 1565
>> 1566 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
>> 1567 if (error)
>> 1568 goto error_sec_opts;
>> 1569
>> 1570 /*
>> 1571 * Setup a dummy root and fs_info for test/set super. This is because
>> 1572 * we don't actually fill this stuff out until open_ctree, but we need
>> 1573 * it for searching for existing supers, so this lets us do that and
>> 1574 * then open_ctree will properly initialize everything later.
>> 1575 */
>> 1576 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
>> 1577 if (!fs_info) {
>> 1578 error = -ENOMEM;
>> 1579 goto error_sec_opts;
>> 1580 }
>> 1581
>> 1582 fs_info->fs_devices = fs_devices;
>> 1583
>> 1584 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
>> 1585 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
>> 1586 security_init_mnt_opts(&fs_info->security_opts);
>> 1587 if (!fs_info->super_copy || !fs_info->super_for_commit) {
>> 1588 error = -ENOMEM;
>> 1589 goto error_fs_info;
>> 1590 }
>> 1591
>> 1592 error = btrfs_open_devices(fs_devices, mode, fs_type);
>> 1593 if (error)
>> 1594 goto error_fs_info;
>> 1595
>> 1596 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
>> 1597 error = -EACCES;
>> 1598 goto error_close_devices;
>> 1599 }
>> 1600
>> 1601 bdev = fs_devices->latest_bdev;
>> 1602 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
>> 1603 fs_info);
>> 1604 if (IS_ERR(s)) {
>> 1605 error = PTR_ERR(s);
>> 1606 goto error_close_devices;
>> 1607 }
>> 1608
>> 1609 if (s->s_root) {
>> 1610 btrfs_close_devices(fs_devices);
>> 1611 free_fs_info(fs_info);
>> 1612 if ((flags ^ s->s_flags) & MS_RDONLY)
>> 1613 error = -EBUSY;
>> 1614 } else {
>> 1615 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
>> 1616 btrfs_sb(s)->bdev_holder = fs_type;
>> 1617 error = btrfs_fill_super(s, fs_devices, data);
>> 1618 }
>> 1619 if (error) {
>> 1620 deactivate_locked_super(s);
>> 1621 goto error_sec_opts;
>> 1622 }
>> 1623
>> 1624 fs_info = btrfs_sb(s);
>> 1625 error = setup_security_options(fs_info, s, &new_sec_opts);
>> 1626 if (error) {
>> 1627 deactivate_locked_super(s);
>> 1628 goto error_sec_opts;
>> 1629 }
>> 1630
>> 1631 return dget(s->s_root);
>> 1632
>> 1633 error_close_devices:
>> 1634 btrfs_close_devices(fs_devices);
>> 1635 error_fs_info:
>> 1636 free_fs_info(fs_info);
>> 1637 error_sec_opts:
>> 1638 security_free_mnt_opts(&new_sec_opts);
>> 1639 return ERR_PTR(error);
>> 1640 }
>> 1641
1226 static void btrfs_resize_thread_pool(struct b 1642 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1227 u32 new_ !! 1643 int new_pool_size, int old_pool_size)
1228 { 1644 {
1229 if (new_pool_size == old_pool_size) 1645 if (new_pool_size == old_pool_size)
1230 return; 1646 return;
1231 1647
1232 fs_info->thread_pool_size = new_pool_ 1648 fs_info->thread_pool_size = new_pool_size;
1233 1649
1234 btrfs_info(fs_info, "resize thread po 1650 btrfs_info(fs_info, "resize thread pool %d -> %d",
1235 old_pool_size, new_pool_size); 1651 old_pool_size, new_pool_size);
1236 1652
1237 btrfs_workqueue_set_max(fs_info->work 1653 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1238 btrfs_workqueue_set_max(fs_info->dela 1654 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
>> 1655 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1239 btrfs_workqueue_set_max(fs_info->cach 1656 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1240 workqueue_set_max_active(fs_info->end !! 1657 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1241 workqueue_set_max_active(fs_info->end !! 1658 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
>> 1659 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
>> 1660 new_pool_size);
1242 btrfs_workqueue_set_max(fs_info->endi 1661 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1243 btrfs_workqueue_set_max(fs_info->endi 1662 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1244 btrfs_workqueue_set_max(fs_info->dela 1663 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
>> 1664 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
>> 1665 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
>> 1666 new_pool_size);
>> 1667 }
>> 1668
>> 1669 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
>> 1670 {
>> 1671 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1245 } 1672 }
1246 1673
1247 static inline void btrfs_remount_begin(struct 1674 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1248 unsign !! 1675 unsigned long old_opts, int flags)
1249 { 1676 {
1250 if (btrfs_raw_test_opt(old_opts, AUTO 1677 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1251 (!btrfs_raw_test_opt(fs_info->mou 1678 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1252 (flags & SB_RDONLY))) { !! 1679 (flags & MS_RDONLY))) {
1253 /* wait for any defraggers to 1680 /* wait for any defraggers to finish */
1254 wait_event(fs_info->transacti 1681 wait_event(fs_info->transaction_wait,
1255 (atomic_read(&fs_i 1682 (atomic_read(&fs_info->defrag_running) == 0));
1256 if (flags & SB_RDONLY) !! 1683 if (flags & MS_RDONLY)
1257 sync_filesystem(fs_in 1684 sync_filesystem(fs_info->sb);
1258 } 1685 }
1259 } 1686 }
1260 1687
1261 static inline void btrfs_remount_cleanup(stru 1688 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1262 unsi !! 1689 unsigned long old_opts)
1263 { 1690 {
1264 const bool cache_opt = btrfs_test_opt <<
1265 <<
1266 /* 1691 /*
1267 * We need to cleanup all defragable 1692 * We need to cleanup all defragable inodes if the autodefragment is
1268 * close or the filesystem is read on 1693 * close or the filesystem is read only.
1269 */ 1694 */
1270 if (btrfs_raw_test_opt(old_opts, AUTO 1695 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1271 (!btrfs_raw_test_opt(fs_info->mou !! 1696 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
>> 1697 (fs_info->sb->s_flags & MS_RDONLY))) {
1272 btrfs_cleanup_defrag_inodes(f 1698 btrfs_cleanup_defrag_inodes(fs_info);
1273 } 1699 }
1274 1700
1275 /* If we toggled discard async */ !! 1701 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1276 if (!btrfs_raw_test_opt(old_opts, DIS <<
1277 btrfs_test_opt(fs_info, DISCARD_A <<
1278 btrfs_discard_resume(fs_info) <<
1279 else if (btrfs_raw_test_opt(old_opts, <<
1280 !btrfs_test_opt(fs_info, DIS <<
1281 btrfs_discard_cleanup(fs_info <<
1282 <<
1283 /* If we toggled space cache */ <<
1284 if (cache_opt != btrfs_free_space_cac <<
1285 btrfs_set_free_space_cache_v1 <<
1286 } 1702 }
1287 1703
1288 static int btrfs_remount_rw(struct btrfs_fs_i !! 1704 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1289 { 1705 {
>> 1706 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 1707 struct btrfs_root *root = fs_info->tree_root;
>> 1708 unsigned old_flags = sb->s_flags;
>> 1709 unsigned long old_opts = fs_info->mount_opt;
>> 1710 unsigned long old_compress_type = fs_info->compress_type;
>> 1711 u64 old_max_inline = fs_info->max_inline;
>> 1712 u64 old_alloc_start = fs_info->alloc_start;
>> 1713 int old_thread_pool_size = fs_info->thread_pool_size;
>> 1714 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1290 int ret; 1715 int ret;
1291 1716
1292 if (BTRFS_FS_ERROR(fs_info)) { !! 1717 sync_filesystem(sb);
1293 btrfs_err(fs_info, !! 1718 btrfs_remount_prepare(fs_info);
1294 "remounting read-wr <<
1295 return -EINVAL; <<
1296 } <<
1297 1719
1298 if (fs_info->fs_devices->rw_devices = !! 1720 if (data) {
1299 return -EACCES; !! 1721 struct security_mnt_opts new_sec_opts;
1300 1722
1301 if (!btrfs_check_rw_degradable(fs_inf !! 1723 security_init_mnt_opts(&new_sec_opts);
1302 btrfs_warn(fs_info, !! 1724 ret = parse_security_options(data, &new_sec_opts);
1303 "too many missing !! 1725 if (ret)
1304 return -EACCES; !! 1726 goto restore;
>> 1727 ret = setup_security_options(fs_info, sb,
>> 1728 &new_sec_opts);
>> 1729 if (ret) {
>> 1730 security_free_mnt_opts(&new_sec_opts);
>> 1731 goto restore;
>> 1732 }
1305 } 1733 }
1306 1734
1307 if (btrfs_super_log_root(fs_info->sup !! 1735 ret = btrfs_parse_options(fs_info, data, *flags);
1308 btrfs_warn(fs_info, !! 1736 if (ret) {
1309 "mount required to !! 1737 ret = -EINVAL;
1310 return -EINVAL; !! 1738 goto restore;
1311 } 1739 }
1312 1740
1313 /* !! 1741 btrfs_remount_begin(fs_info, old_opts, *flags);
1314 * NOTE: when remounting with a chang !! 1742 btrfs_resize_thread_pool(fs_info,
1315 * anywhere above this point, as we a !! 1743 fs_info->thread_pool_size, old_thread_pool_size);
1316 * until we pass the above checks. <<
1317 */ <<
1318 ret = btrfs_start_pre_rw_mount(fs_inf <<
1319 if (ret) <<
1320 return ret; <<
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 1744
1346 /* Wait for the uuid_scan task to fin !! 1745 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1347 down(&fs_info->uuid_tree_rescan_sem); !! 1746 goto out;
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 <<
1361 /* <<
1362 * The cleaner task could be already <<
1363 * BTRFS_FS_STATE_RO (and SB_RDONLY i <<
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 <<
1373 /* <<
1374 * We've set the superblock to RO mod <<
1375 * cleaner task sleep without running <<
1376 * through all the delayed iputs here <<
1377 * without remounting RW we don't end <<
1378 * with a non-empty list of delayed i <<
1379 */ <<
1380 btrfs_run_delayed_iputs(fs_info); <<
1381 <<
1382 btrfs_dev_replace_suspend_for_unmount <<
1383 btrfs_scrub_cancel(fs_info); <<
1384 btrfs_pause_balance(fs_info); <<
1385 1747
1386 /* !! 1748 if (*flags & MS_RDONLY) {
1387 * Pause the qgroup rescan worker if !! 1749 /*
1388 * be still running after we are in R !! 1750 * this also happens on 'umount -rf' or on shutdown, when
1389 * we unmount, it might have left a t !! 1751 * the filesystem is busy.
1390 * the transaction and/or crash. !! 1752 */
1391 */ !! 1753 cancel_work_sync(&fs_info->async_reclaim_work);
1392 btrfs_qgroup_wait_for_completion(fs_i <<
1393 1754
1394 return btrfs_commit_super(fs_info); !! 1755 /* wait for the uuid_scan task to finish */
1395 } !! 1756 down(&fs_info->uuid_tree_rescan_sem);
>> 1757 /* avoid complains from lockdep et al. */
>> 1758 up(&fs_info->uuid_tree_rescan_sem);
1396 1759
1397 static void btrfs_ctx_to_info(struct btrfs_fs !! 1760 sb->s_flags |= MS_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 1761
1484 if (info->max_inline != BTRFS_DEFAULT !! 1762 /*
1485 btrfs_info(info, "max_inline !! 1763 * Setting MS_RDONLY will put the cleaner thread to
1486 } !! 1764 * sleep at the next loop if it's already active.
>> 1765 * If it's already asleep, we'll leave unused block
>> 1766 * groups on disk until we're mounted read-write again
>> 1767 * unless we clean them up here.
>> 1768 */
>> 1769 btrfs_delete_unused_bgs(fs_info);
1487 1770
1488 static int btrfs_reconfigure(struct fs_contex !! 1771 btrfs_dev_replace_suspend_for_unmount(fs_info);
1489 { !! 1772 btrfs_scrub_cancel(fs_info);
1490 struct super_block *sb = fc->root->d_ !! 1773 btrfs_pause_balance(fs_info);
1491 struct btrfs_fs_info *fs_info = btrfs !! 1774
1492 struct btrfs_fs_context *ctx = fc->fs !! 1775 ret = btrfs_commit_super(fs_info);
1493 struct btrfs_fs_context old_ctx; !! 1776 if (ret)
1494 int ret = 0; !! 1777 goto restore;
1495 bool mount_reconfigure = (fc->s_fs_in !! 1778 } else {
>> 1779 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
>> 1780 btrfs_err(fs_info,
>> 1781 "Remounting read-write after error is not allowed");
>> 1782 ret = -EINVAL;
>> 1783 goto restore;
>> 1784 }
>> 1785 if (fs_info->fs_devices->rw_devices == 0) {
>> 1786 ret = -EACCES;
>> 1787 goto restore;
>> 1788 }
1496 1789
1497 btrfs_info_to_ctx(fs_info, &old_ctx); !! 1790 if (fs_info->fs_devices->missing_devices >
>> 1791 fs_info->num_tolerated_disk_barrier_failures &&
>> 1792 !(*flags & MS_RDONLY)) {
>> 1793 btrfs_warn(fs_info,
>> 1794 "too many missing devices, writeable remount is not allowed");
>> 1795 ret = -EACCES;
>> 1796 goto restore;
>> 1797 }
1498 1798
1499 /* !! 1799 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1500 * This is our "bind mount" trick, we !! 1800 ret = -EINVAL;
1501 * anything other than mount a differ !! 1801 goto restore;
1502 * all of the mount options should be !! 1802 }
1503 */ <<
1504 if (mount_reconfigure) <<
1505 ctx->mount_opt = old_ctx.moun <<
1506 1803
1507 sync_filesystem(sb); !! 1804 ret = btrfs_cleanup_fs_roots(fs_info);
1508 set_bit(BTRFS_FS_STATE_REMOUNTING, &f !! 1805 if (ret)
>> 1806 goto restore;
1509 1807
1510 if (!btrfs_check_options(fs_info, &ct !! 1808 /* recover relocation */
1511 return -EINVAL; !! 1809 mutex_lock(&fs_info->cleaner_mutex);
>> 1810 ret = btrfs_recover_relocation(root);
>> 1811 mutex_unlock(&fs_info->cleaner_mutex);
>> 1812 if (ret)
>> 1813 goto restore;
1512 1814
1513 ret = btrfs_check_features(fs_info, ! !! 1815 ret = btrfs_resume_balance_async(fs_info);
1514 if (ret < 0) !! 1816 if (ret)
1515 return ret; !! 1817 goto restore;
1516 1818
1517 btrfs_ctx_to_info(fs_info, ctx); !! 1819 ret = btrfs_resume_dev_replace_async(fs_info);
1518 btrfs_remount_begin(fs_info, old_ctx. !! 1820 if (ret) {
1519 btrfs_resize_thread_pool(fs_info, fs_ !! 1821 btrfs_warn(fs_info, "failed to resume dev_replace");
1520 old_ctx.thre !! 1822 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 } 1823 }
1536 } <<
1537 1824
1538 ret = 0; !! 1825 if (!fs_info->uuid_root) {
1539 if (!sb_rdonly(sb) && (fc->sb_flags & !! 1826 btrfs_info(fs_info, "creating UUID tree");
1540 ret = btrfs_remount_ro(fs_inf !! 1827 ret = btrfs_create_uuid_tree(fs_info);
1541 else if (sb_rdonly(sb) && !(fc->sb_fl !! 1828 if (ret) {
1542 ret = btrfs_remount_rw(fs_inf !! 1829 btrfs_warn(fs_info,
1543 if (ret) !! 1830 "failed to create the UUID tree %d",
1544 goto restore; !! 1831 ret);
1545 !! 1832 goto restore;
1546 /* !! 1833 }
1547 * If we set the mask during the para !! 1834 }
1548 * remount. Here we can set the mask !! 1835 sb->s_flags &= ~MS_RDONLY;
1549 * appropriately. <<
1550 */ <<
1551 if ((fc->sb_flags & SB_POSIXACL) != ( <<
1552 fc->sb_flags_mask |= SB_POSIX <<
1553 1836
1554 btrfs_emit_options(fs_info, &old_ctx) !! 1837 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
>> 1838 }
>> 1839 out:
1555 wake_up_process(fs_info->transaction_ 1840 wake_up_process(fs_info->transaction_kthread);
1556 btrfs_remount_cleanup(fs_info, old_ct !! 1841 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; 1842 return 0;
>> 1843
1561 restore: 1844 restore:
1562 btrfs_ctx_to_info(fs_info, &old_ctx); !! 1845 /* We've hit an error - don't reset MS_RDONLY */
1563 btrfs_remount_cleanup(fs_info, old_ct !! 1846 if (sb->s_flags & MS_RDONLY)
1564 clear_bit(BTRFS_FS_STATE_REMOUNTING, !! 1847 old_flags |= MS_RDONLY;
>> 1848 sb->s_flags = old_flags;
>> 1849 fs_info->mount_opt = old_opts;
>> 1850 fs_info->compress_type = old_compress_type;
>> 1851 fs_info->max_inline = old_max_inline;
>> 1852 mutex_lock(&fs_info->chunk_mutex);
>> 1853 fs_info->alloc_start = old_alloc_start;
>> 1854 mutex_unlock(&fs_info->chunk_mutex);
>> 1855 btrfs_resize_thread_pool(fs_info,
>> 1856 old_thread_pool_size, fs_info->thread_pool_size);
>> 1857 fs_info->metadata_ratio = old_metadata_ratio;
>> 1858 btrfs_remount_cleanup(fs_info, old_opts);
1565 return ret; 1859 return ret;
1566 } 1860 }
1567 1861
1568 /* Used to sort the devices by max_avail(desc 1862 /* Used to sort the devices by max_avail(descending sort) */
1569 static int btrfs_cmp_device_free_bytes(const !! 1863 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
>> 1864 const void *dev_info2)
1570 { 1865 {
1571 const struct btrfs_device_info *dev_i !! 1866 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1572 const struct btrfs_device_info *dev_i !! 1867 ((struct btrfs_device_info *)dev_info2)->max_avail)
1573 <<
1574 if (dev_info1->max_avail > dev_info2- <<
1575 return -1; 1868 return -1;
1576 else if (dev_info1->max_avail < dev_i !! 1869 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
>> 1870 ((struct btrfs_device_info *)dev_info2)->max_avail)
1577 return 1; 1871 return 1;
>> 1872 else
1578 return 0; 1873 return 0;
1579 } 1874 }
1580 1875
1581 /* 1876 /*
1582 * sort the devices by max_avail, in which ma 1877 * sort the devices by max_avail, in which max free extent size of each device
1583 * is stored.(Descending Sort) 1878 * is stored.(Descending Sort)
1584 */ 1879 */
1585 static inline void btrfs_descending_sort_devi 1880 static inline void btrfs_descending_sort_devices(
1586 struc 1881 struct btrfs_device_info *devices,
1587 size_ 1882 size_t nr_devices)
1588 { 1883 {
1589 sort(devices, nr_devices, sizeof(stru 1884 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1590 btrfs_cmp_device_free_bytes, NUL 1885 btrfs_cmp_device_free_bytes, NULL);
1591 } 1886 }
1592 1887
1593 /* 1888 /*
1594 * The helper to calc the free space on the d 1889 * The helper to calc the free space on the devices that can be used to store
1595 * file data. 1890 * file data.
1596 */ 1891 */
1597 static inline int btrfs_calc_avail_data_space !! 1892 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1598 !! 1893 u64 *free_bytes)
1599 { 1894 {
>> 1895 struct btrfs_root *root = fs_info->tree_root;
1600 struct btrfs_device_info *devices_inf 1896 struct btrfs_device_info *devices_info;
1601 struct btrfs_fs_devices *fs_devices = 1897 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1602 struct btrfs_device *device; 1898 struct btrfs_device *device;
>> 1899 u64 skip_space;
1603 u64 type; 1900 u64 type;
1604 u64 avail_space; 1901 u64 avail_space;
>> 1902 u64 used_space;
1605 u64 min_stripe_size; 1903 u64 min_stripe_size;
1606 int num_stripes = 1; !! 1904 int min_stripes = 1, num_stripes = 1;
1607 int i = 0, nr_devices; 1905 int i = 0, nr_devices;
1608 const struct btrfs_raid_attr *rattr; !! 1906 int ret;
1609 1907
1610 /* 1908 /*
1611 * We aren't under the device list lo 1909 * We aren't under the device list lock, so this is racy-ish, but good
1612 * enough for our purposes. 1910 * enough for our purposes.
1613 */ 1911 */
1614 nr_devices = fs_info->fs_devices->ope 1912 nr_devices = fs_info->fs_devices->open_devices;
1615 if (!nr_devices) { 1913 if (!nr_devices) {
1616 smp_mb(); 1914 smp_mb();
1617 nr_devices = fs_info->fs_devi 1915 nr_devices = fs_info->fs_devices->open_devices;
1618 ASSERT(nr_devices); 1916 ASSERT(nr_devices);
1619 if (!nr_devices) { 1917 if (!nr_devices) {
1620 *free_bytes = 0; 1918 *free_bytes = 0;
1621 return 0; 1919 return 0;
1622 } 1920 }
1623 } 1921 }
1624 1922
1625 devices_info = kmalloc_array(nr_devic 1923 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1626 GFP_KERNEL); !! 1924 GFP_NOFS);
1627 if (!devices_info) 1925 if (!devices_info)
1628 return -ENOMEM; 1926 return -ENOMEM;
1629 1927
1630 /* calc min stripe number for data sp 1928 /* calc min stripe number for data space allocation */
1631 type = btrfs_data_alloc_profile(fs_in !! 1929 type = btrfs_get_alloc_profile(root, 1);
1632 rattr = &btrfs_raid_array[btrfs_bg_fl !! 1930 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1633 !! 1931 min_stripes = 2;
1634 if (type & BTRFS_BLOCK_GROUP_RAID0) <<
1635 num_stripes = nr_devices; 1932 num_stripes = nr_devices;
1636 else if (type & BTRFS_BLOCK_GROUP_RAI !! 1933 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1637 num_stripes = rattr->ncopies; !! 1934 min_stripes = 2;
1638 else if (type & BTRFS_BLOCK_GROUP_RAI !! 1935 num_stripes = 2;
>> 1936 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
>> 1937 min_stripes = 4;
1639 num_stripes = 4; 1938 num_stripes = 4;
>> 1939 }
1640 1940
1641 /* Adjust for more than 1 stripe per !! 1941 if (type & BTRFS_BLOCK_GROUP_DUP)
1642 min_stripe_size = rattr->dev_stripes !! 1942 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
>> 1943 else
>> 1944 min_stripe_size = BTRFS_STRIPE_LEN;
1643 1945
>> 1946 if (fs_info->alloc_start)
>> 1947 mutex_lock(&fs_devices->device_list_mutex);
1644 rcu_read_lock(); 1948 rcu_read_lock();
1645 list_for_each_entry_rcu(device, &fs_d 1949 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1646 if (!test_bit(BTRFS_DEV_STATE !! 1950 if (!device->in_fs_metadata || !device->bdev ||
1647 !! 1951 device->is_tgtdev_for_dev_replace)
1648 !device->bdev || <<
1649 test_bit(BTRFS_DEV_STATE_ <<
1650 continue; 1952 continue;
1651 1953
1652 if (i >= nr_devices) 1954 if (i >= nr_devices)
1653 break; 1955 break;
1654 1956
1655 avail_space = device->total_b 1957 avail_space = device->total_bytes - device->bytes_used;
1656 1958
1657 /* align with stripe_len */ 1959 /* align with stripe_len */
1658 avail_space = rounddown(avail !! 1960 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
>> 1961 avail_space *= BTRFS_STRIPE_LEN;
1659 1962
1660 /* 1963 /*
1661 * Ensure we have at least mi !! 1964 * In order to avoid overwriting the superblock on the drive,
1662 * reserved space on the devi !! 1965 * btrfs starts at an offset of at least 1MB when doing chunk
>> 1966 * allocation.
1663 */ 1967 */
1664 if (avail_space <= BTRFS_DEVI !! 1968 skip_space = SZ_1M;
1665 continue; !! 1969
>> 1970 /* user can set the offset in fs_info->alloc_start. */
>> 1971 if (fs_info->alloc_start &&
>> 1972 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
>> 1973 device->total_bytes) {
>> 1974 rcu_read_unlock();
>> 1975 skip_space = max(fs_info->alloc_start, skip_space);
1666 1976
1667 avail_space -= BTRFS_DEVICE_R !! 1977 /*
>> 1978 * btrfs can not use the free space in
>> 1979 * [0, skip_space - 1], we must subtract it from the
>> 1980 * total. In order to implement it, we account the used
>> 1981 * space in this range first.
>> 1982 */
>> 1983 ret = btrfs_account_dev_extents_size(device, 0,
>> 1984 skip_space - 1,
>> 1985 &used_space);
>> 1986 if (ret) {
>> 1987 kfree(devices_info);
>> 1988 mutex_unlock(&fs_devices->device_list_mutex);
>> 1989 return ret;
>> 1990 }
>> 1991
>> 1992 rcu_read_lock();
>> 1993
>> 1994 /* calc the free space in [0, skip_space - 1] */
>> 1995 skip_space -= used_space;
>> 1996 }
>> 1997
>> 1998 /*
>> 1999 * we can use the free space in [0, skip_space - 1], subtract
>> 2000 * it from the total.
>> 2001 */
>> 2002 if (avail_space && avail_space >= skip_space)
>> 2003 avail_space -= skip_space;
>> 2004 else
>> 2005 avail_space = 0;
>> 2006
>> 2007 if (avail_space < min_stripe_size)
>> 2008 continue;
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();
>> 2016 if (fs_info->alloc_start)
>> 2017 mutex_unlock(&fs_devices->device_list_mutex);
1675 2018
1676 nr_devices = i; 2019 nr_devices = i;
1677 2020
1678 btrfs_descending_sort_devices(devices 2021 btrfs_descending_sort_devices(devices_info, nr_devices);
1679 2022
1680 i = nr_devices - 1; 2023 i = nr_devices - 1;
1681 avail_space = 0; 2024 avail_space = 0;
1682 while (nr_devices >= rattr->devs_min) !! 2025 while (nr_devices >= min_stripes) {
1683 num_stripes = min(num_stripes !! 2026 if (num_stripes > nr_devices)
>> 2027 num_stripes = nr_devices;
1684 2028
1685 if (devices_info[i].max_avail 2029 if (devices_info[i].max_avail >= min_stripe_size) {
1686 int j; 2030 int j;
1687 u64 alloc_size; 2031 u64 alloc_size;
1688 2032
1689 avail_space += device 2033 avail_space += devices_info[i].max_avail * num_stripes;
1690 alloc_size = devices_ 2034 alloc_size = devices_info[i].max_avail;
1691 for (j = i + 1 - num_ 2035 for (j = i + 1 - num_stripes; j <= i; j++)
1692 devices_info[ 2036 devices_info[j].max_avail -= alloc_size;
1693 } 2037 }
1694 i--; 2038 i--;
1695 nr_devices--; 2039 nr_devices--;
1696 } 2040 }
1697 2041
1698 kfree(devices_info); 2042 kfree(devices_info);
1699 *free_bytes = avail_space; 2043 *free_bytes = avail_space;
1700 return 0; 2044 return 0;
1701 } 2045 }
1702 2046
1703 /* 2047 /*
1704 * Calculate numbers for 'df', pessimistic in 2048 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1705 * 2049 *
1706 * If there's a redundant raid level at DATA 2050 * If there's a redundant raid level at DATA block groups, use the respective
1707 * multiplier to scale the sizes. 2051 * multiplier to scale the sizes.
1708 * 2052 *
1709 * Unused device space usage is based on simu 2053 * Unused device space usage is based on simulating the chunk allocator
1710 * algorithm that respects the device sizes a !! 2054 * algorithm that respects the device sizes, order of allocations and the
1711 * a close approximation of the actual use bu !! 2055 * 'alloc_start' value, this is a close approximation of the actual use but
1712 * change the result (like a new metadata chu !! 2056 * there are other factors that may change the result (like a new metadata
>> 2057 * chunk).
1713 * 2058 *
1714 * If metadata is exhausted, f_bavail will be 2059 * If metadata is exhausted, f_bavail will be 0.
1715 */ 2060 */
1716 static int btrfs_statfs(struct dentry *dentry 2061 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1717 { 2062 {
1718 struct btrfs_fs_info *fs_info = btrfs 2063 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1719 struct btrfs_super_block *disk_super 2064 struct btrfs_super_block *disk_super = fs_info->super_copy;
>> 2065 struct list_head *head = &fs_info->space_info;
1720 struct btrfs_space_info *found; 2066 struct btrfs_space_info *found;
1721 u64 total_used = 0; 2067 u64 total_used = 0;
1722 u64 total_free_data = 0; 2068 u64 total_free_data = 0;
1723 u64 total_free_meta = 0; 2069 u64 total_free_meta = 0;
1724 u32 bits = fs_info->sectorsize_bits; !! 2070 int bits = dentry->d_sb->s_blocksize_bits;
1725 __be32 *fsid = (__be32 *)fs_info->fs_ !! 2071 __be32 *fsid = (__be32 *)fs_info->fsid;
1726 unsigned factor = 1; 2072 unsigned factor = 1;
1727 struct btrfs_block_rsv *block_rsv = & 2073 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1728 int ret; 2074 int ret;
1729 u64 thresh = 0; 2075 u64 thresh = 0;
1730 int mixed = 0; 2076 int mixed = 0;
1731 2077
1732 list_for_each_entry(found, &fs_info-> !! 2078 rcu_read_lock();
>> 2079 list_for_each_entry_rcu(found, head, list) {
1733 if (found->flags & BTRFS_BLOC 2080 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1734 int i; 2081 int i;
1735 2082
1736 total_free_data += fo 2083 total_free_data += found->disk_total - found->disk_used;
1737 total_free_data -= 2084 total_free_data -=
1738 btrfs_account 2085 btrfs_account_ro_block_groups_free_space(found);
1739 2086
1740 for (i = 0; i < BTRFS 2087 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1741 if (!list_emp !! 2088 if (!list_empty(&found->block_groups[i])) {
1742 facto !! 2089 switch (i) {
1743 !! 2090 case BTRFS_RAID_DUP:
>> 2091 case BTRFS_RAID_RAID1:
>> 2092 case BTRFS_RAID_RAID10:
>> 2093 factor = 2;
>> 2094 }
>> 2095 }
1744 } 2096 }
1745 } 2097 }
1746 2098
1747 /* 2099 /*
1748 * Metadata in mixed block gr !! 2100 * Metadata in mixed block goup profiles are accounted in data
1749 */ 2101 */
1750 if (!mixed && found->flags & 2102 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1751 if (found->flags & BT 2103 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1752 mixed = 1; 2104 mixed = 1;
1753 else 2105 else
1754 total_free_me 2106 total_free_meta += found->disk_total -
1755 found 2107 found->disk_used;
1756 } 2108 }
1757 2109
1758 total_used += found->disk_use 2110 total_used += found->disk_used;
1759 } 2111 }
1760 2112
>> 2113 rcu_read_unlock();
>> 2114
1761 buf->f_blocks = div_u64(btrfs_super_t 2115 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1762 buf->f_blocks >>= bits; 2116 buf->f_blocks >>= bits;
1763 buf->f_bfree = buf->f_blocks - (div_u 2117 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1764 2118
1765 /* Account global block reserve as us 2119 /* Account global block reserve as used, it's in logical size already */
1766 spin_lock(&block_rsv->lock); 2120 spin_lock(&block_rsv->lock);
1767 /* Mixed block groups accounting is n 2121 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1768 if (buf->f_bfree >= block_rsv->size > 2122 if (buf->f_bfree >= block_rsv->size >> bits)
1769 buf->f_bfree -= block_rsv->si 2123 buf->f_bfree -= block_rsv->size >> bits;
1770 else 2124 else
1771 buf->f_bfree = 0; 2125 buf->f_bfree = 0;
1772 spin_unlock(&block_rsv->lock); 2126 spin_unlock(&block_rsv->lock);
1773 2127
1774 buf->f_bavail = div_u64(total_free_da 2128 buf->f_bavail = div_u64(total_free_data, factor);
1775 ret = btrfs_calc_avail_data_space(fs_ 2129 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1776 if (ret) 2130 if (ret)
1777 return ret; 2131 return ret;
1778 buf->f_bavail += div_u64(total_free_d 2132 buf->f_bavail += div_u64(total_free_data, factor);
1779 buf->f_bavail = buf->f_bavail >> bits 2133 buf->f_bavail = buf->f_bavail >> bits;
1780 2134
1781 /* 2135 /*
1782 * We calculate the remaining metadat 2136 * We calculate the remaining metadata space minus global reserve. If
1783 * this is (supposedly) smaller than 2137 * this is (supposedly) smaller than zero, there's no space. But this
1784 * does not hold in practice, the exh 2138 * does not hold in practice, the exhausted state happens where's still
1785 * some positive delta. So we apply s 2139 * some positive delta. So we apply some guesswork and compare the
1786 * delta to a 4M threshold. (Practic 2140 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1787 * 2141 *
1788 * We probably cannot calculate the e 2142 * We probably cannot calculate the exact threshold value because this
1789 * depends on the internal reservatio 2143 * depends on the internal reservations requested by various
1790 * operations, so some operations tha 2144 * operations, so some operations that consume a few metadata will
1791 * succeed even if the Avail is zero. 2145 * succeed even if the Avail is zero. But this is better than the other
1792 * way around. 2146 * way around.
1793 */ 2147 */
1794 thresh = SZ_4M; !! 2148 thresh = 4 * 1024 * 1024;
1795 2149
1796 /* !! 2150 if (!mixed && total_free_meta - thresh < block_rsv->size)
1797 * We only want to claim there's no a <<
1798 * allocate chunks for our metadata p <<
1799 * not fit in the free metadata space <<
1800 * still can allocate chunks and thus <<
1801 * calculated f_bavail. <<
1802 */ <<
1803 if (!mixed && block_rsv->space_info-> <<
1804 (total_free_meta < thresh || tota <<
1805 buf->f_bavail = 0; 2151 buf->f_bavail = 0;
1806 2152
1807 buf->f_type = BTRFS_SUPER_MAGIC; 2153 buf->f_type = BTRFS_SUPER_MAGIC;
1808 buf->f_bsize = fs_info->sectorsize; !! 2154 buf->f_bsize = dentry->d_sb->s_blocksize;
1809 buf->f_namelen = BTRFS_NAME_LEN; 2155 buf->f_namelen = BTRFS_NAME_LEN;
1810 2156
1811 /* We treat it as constant endianness 2157 /* We treat it as constant endianness (it doesn't matter _which_)
1812 because we want the fsid to come o 2158 because we want the fsid to come out the same whether mounted
1813 on a big-endian or little-endian h 2159 on a big-endian or little-endian host */
1814 buf->f_fsid.val[0] = be32_to_cpu(fsid 2160 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 2161 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1816 /* Mask in the root object ID too, to 2162 /* Mask in the root object ID too, to disambiguate subvols */
1817 buf->f_fsid.val[0] ^= btrfs_root_id(B !! 2163 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
1818 buf->f_fsid.val[1] ^= btrfs_root_id(B !! 2164 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
1819 2165
1820 return 0; 2166 return 0;
1821 } 2167 }
1822 2168
1823 static int btrfs_fc_test_super(struct super_b <<
1824 { <<
1825 struct btrfs_fs_info *p = fc->s_fs_in <<
1826 struct btrfs_fs_info *fs_info = btrfs <<
1827 <<
1828 return fs_info->fs_devices == p->fs_d <<
1829 } <<
1830 <<
1831 static int btrfs_get_tree_super(struct fs_con <<
1832 { <<
1833 struct btrfs_fs_info *fs_info = fc->s <<
1834 struct btrfs_fs_context *ctx = fc->fs <<
1835 struct btrfs_fs_devices *fs_devices = <<
1836 struct block_device *bdev; <<
1837 struct btrfs_device *device; <<
1838 struct super_block *sb; <<
1839 blk_mode_t mode = btrfs_open_mode(fc) <<
1840 int ret; <<
1841 <<
1842 btrfs_ctx_to_info(fs_info, ctx); <<
1843 mutex_lock(&uuid_mutex); <<
1844 <<
1845 /* <<
1846 * With 'true' passed to btrfs_scan_o <<
1847 * either a valid device or an error. <<
1848 */ <<
1849 device = btrfs_scan_one_device(fc->so <<
1850 ASSERT(device != NULL); <<
1851 if (IS_ERR(device)) { <<
1852 mutex_unlock(&uuid_mutex); <<
1853 return PTR_ERR(device); <<
1854 } <<
1855 <<
1856 fs_devices = device->fs_devices; <<
1857 fs_info->fs_devices = fs_devices; <<
1858 <<
1859 ret = btrfs_open_devices(fs_devices, <<
1860 mutex_unlock(&uuid_mutex); <<
1861 if (ret) <<
1862 return ret; <<
1863 <<
1864 if (!(fc->sb_flags & SB_RDONLY) && fs <<
1865 ret = -EACCES; <<
1866 goto error; <<
1867 } <<
1868 <<
1869 bdev = fs_devices->latest_dev->bdev; <<
1870 <<
1871 /* <<
1872 * From now on the error handling is <<
1873 * <<
1874 * If successful, this will transfer <<
1875 * and fc->s_fs_info will be NULL. H <<
1876 * super, we'll still have fc->s_fs_i <<
1877 * completely out it'll be cleaned up <<
1878 * otherwise it's tied to the lifetim <<
1879 */ <<
1880 sb = sget_fc(fc, btrfs_fc_test_super, <<
1881 if (IS_ERR(sb)) { <<
1882 ret = PTR_ERR(sb); <<
1883 goto error; <<
1884 } <<
1885 <<
1886 set_device_specific_options(fs_info); <<
1887 <<
1888 if (sb->s_root) { <<
1889 btrfs_close_devices(fs_device <<
1890 if ((fc->sb_flags ^ sb->s_fla <<
1891 ret = -EBUSY; <<
1892 } else { <<
1893 snprintf(sb->s_id, sizeof(sb- <<
1894 shrinker_debugfs_rename(sb->s <<
1895 btrfs_sb(sb)->bdev_holder = & <<
1896 ret = btrfs_fill_super(sb, fs <<
1897 } <<
1898 <<
1899 if (ret) { <<
1900 deactivate_locked_super(sb); <<
1901 return ret; <<
1902 } <<
1903 <<
1904 btrfs_clear_oneshot_options(fs_info); <<
1905 <<
1906 fc->root = dget(sb->s_root); <<
1907 return 0; <<
1908 <<
1909 error: <<
1910 btrfs_close_devices(fs_devices); <<
1911 return ret; <<
1912 } <<
1913 <<
1914 /* <<
1915 * Ever since commit 0723a0473fb4 ("btrfs: al <<
1916 * with different ro/rw options") the followi <<
1917 * <<
1918 * (i) mount /dev/sda3 -o subvol=foo,r <<
1919 * (ii) mount /dev/sda3 -o subvol=bar,r <<
1920 * <<
1921 * which looks nice and innocent but is actua <<
1922 * a long comment. <<
1923 * <<
1924 * On another filesystem a subvolume mount is <<
1925 * <<
1926 * (iii) # create rw superblock + initia <<
1927 * mount -t xfs /dev/sdb /opt/ <<
1928 * <<
1929 * # create ro bind mount <<
1930 * mount --bind -o ro /opt/foo /mn <<
1931 * <<
1932 * # unmount initial mount <<
1933 * umount /opt <<
1934 * <<
1935 * Of course, there's some special subvolume <<
1936 * sb->s_root dentry is really swapped after <<
1937 * it's very close and will help us understan <<
1938 * <<
1939 * The old mount API didn't cleanly distingui <<
1940 * and a superblock being made ro. The only <<
1941 * either object was by passing ms_rdonly. If <<
1942 * mount(2) such as: <<
1943 * <<
1944 * mount("/dev/sdb", "/mnt", "xfs", ms_r <<
1945 * <<
1946 * the MS_RDONLY flag being specified had two <<
1947 * <<
1948 * (1) MNT_READONLY was raised -> the resulti <<
1949 * @mnt->mnt_flags |= MNT_READONLY raised <<
1950 * <<
1951 * (2) MS_RDONLY was passed to the filesystem <<
1952 * made the superblock ro. Note, how SB_R <<
1953 * ms_rdonly and is raised whenever MS_RD <<
1954 * <<
1955 * Creating a subtree mount via (iii) ends up <<
1956 * subtree mounted ro. <<
1957 * <<
1958 * But consider the effect on the old mount A <<
1959 * which combines the distinct step in (iii) <<
1960 * <<
1961 * By issuing (i) both the mount and the supe <<
1962 * is issued the superblock is ro and thus ev <<
1963 * rw it wouldn't help. Hence, btrfs needed t <<
1964 * to rw for (ii) which it did using an inter <<
1965 * <<
1966 * IOW, subvolume mounting was inherently com <<
1967 * MS_RDONLY in mount(2). Note, this ambiguit <<
1968 * "ro" to MS_RDONLY. IOW, in both (i) and (i <<
1969 * passed by mount(8) to mount(2). <<
1970 * <<
1971 * Enter the new mount API. The new mount API <<
1972 * and making a superblock ro. <<
1973 * <<
1974 * (3) To turn a mount ro the MOUNT_ATTR_ONLY <<
1975 * fsmount() or mount_setattr() this is a <<
1976 * specific mount or mount tree that is n <<
1977 * <<
1978 * (4) To turn a superblock ro the "ro" flag <<
1979 * fsconfig(FSCONFIG_SET_FLAG, "ro"). Thi <<
1980 * in fc->sb_flags. <<
1981 * <<
1982 * But, currently the util-linux mount comman <<
1983 * API and is still setting fsconfig(FSCONFIG <<
1984 * btrfs or not, setting the whole super bloc <<
1985 * work with different options work we need t <<
1986 */ <<
1987 static struct vfsmount *btrfs_reconfigure_for <<
1988 { <<
1989 struct vfsmount *mnt; <<
1990 int ret; <<
1991 const bool ro2rw = !(fc->sb_flags & S <<
1992 <<
1993 /* <<
1994 * We got an EBUSY because our SB_RDO <<
1995 * super block, so invert our setting <<
1996 * can get our vfsmount. <<
1997 */ <<
1998 if (ro2rw) <<
1999 fc->sb_flags |= SB_RDONLY; <<
2000 else <<
2001 fc->sb_flags &= ~SB_RDONLY; <<
2002 <<
2003 mnt = fc_mount(fc); <<
2004 if (IS_ERR(mnt)) <<
2005 return mnt; <<
2006 <<
2007 if (!ro2rw) <<
2008 return mnt; <<
2009 <<
2010 /* We need to convert to rw, call rec <<
2011 fc->sb_flags &= ~SB_RDONLY; <<
2012 down_write(&mnt->mnt_sb->s_umount); <<
2013 ret = btrfs_reconfigure(fc); <<
2014 up_write(&mnt->mnt_sb->s_umount); <<
2015 if (ret) { <<
2016 mntput(mnt); <<
2017 return ERR_PTR(ret); <<
2018 } <<
2019 return mnt; <<
2020 } <<
2021 <<
2022 static int btrfs_get_tree_subvol(struct fs_co <<
2023 { <<
2024 struct btrfs_fs_info *fs_info = NULL; <<
2025 struct btrfs_fs_context *ctx = fc->fs <<
2026 struct fs_context *dup_fc; <<
2027 struct dentry *dentry; <<
2028 struct vfsmount *mnt; <<
2029 <<
2030 /* <<
2031 * Setup a dummy root and fs_info for <<
2032 * we don't actually fill this stuff <<
2033 * then open_ctree will properly init <<
2034 * settings later. btrfs_init_fs_inf <<
2035 * of the fs_info (locks and such) to <<
2036 * superblock with our given fs_devic <<
2037 */ <<
2038 fs_info = kvzalloc(sizeof(struct btrf <<
2039 if (!fs_info) <<
2040 return -ENOMEM; <<
2041 <<
2042 fs_info->super_copy = kzalloc(BTRFS_S <<
2043 fs_info->super_for_commit = kzalloc(B <<
2044 if (!fs_info->super_copy || !fs_info- <<
2045 btrfs_free_fs_info(fs_info); <<
2046 return -ENOMEM; <<
2047 } <<
2048 btrfs_init_fs_info(fs_info); <<
2049 <<
2050 dup_fc = vfs_dup_fs_context(fc); <<
2051 if (IS_ERR(dup_fc)) { <<
2052 btrfs_free_fs_info(fs_info); <<
2053 return PTR_ERR(dup_fc); <<
2054 } <<
2055 <<
2056 /* <<
2057 * When we do the sget_fc this gets t <<
2058 * need to set it on the dup_fc as th <<
2059 */ <<
2060 dup_fc->s_fs_info = fs_info; <<
2061 <<
2062 /* <<
2063 * We'll do the security settings in <<
2064 * loop, they were duplicated into du <<
2065 * here. <<
2066 */ <<
2067 security_free_mnt_opts(&fc->security) <<
2068 fc->security = NULL; <<
2069 <<
2070 mnt = fc_mount(dup_fc); <<
2071 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) <<
2072 mnt = btrfs_reconfigure_for_m <<
2073 put_fs_context(dup_fc); <<
2074 if (IS_ERR(mnt)) <<
2075 return PTR_ERR(mnt); <<
2076 <<
2077 /* <<
2078 * This free's ->subvol_name, because <<
2079 * allocate a buffer to hold the subv <<
2080 * reference to it here. <<
2081 */ <<
2082 dentry = mount_subvol(ctx->subvol_nam <<
2083 ctx->subvol_name = NULL; <<
2084 if (IS_ERR(dentry)) <<
2085 return PTR_ERR(dentry); <<
2086 <<
2087 fc->root = dentry; <<
2088 return 0; <<
2089 } <<
2090 <<
2091 static int btrfs_get_tree(struct fs_context * <<
2092 { <<
2093 /* <<
2094 * Since we use mount_subtree to moun <<
2095 * have to do mounts in two steps. <<
2096 * <<
2097 * First pass through we call btrfs_g <<
2098 * wrapper around fc_mount() to call <<
2099 * we'll call btrfs_get_tree_super(). <<
2100 * everything to open the devices and <<
2101 * with a fully constructed vfsmount <<
2102 * from there we can do our mount_sub <<
2103 * whichever subvol we're mounting an <<
2104 * appropriate dentry for the subvol. <<
2105 */ <<
2106 if (fc->s_fs_info) <<
2107 return btrfs_get_tree_super(f <<
2108 return btrfs_get_tree_subvol(fc); <<
2109 } <<
2110 <<
2111 static void btrfs_kill_super(struct super_blo 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 } <<
2117 <<
2118 static void btrfs_free_fs_context(struct fs_c <<
2119 { <<
2120 struct btrfs_fs_context *ctx = fc->fs <<
2121 struct btrfs_fs_info *fs_info = fc->s <<
2122 <<
2123 if (fs_info) <<
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 }; <<
2158 <<
2159 static int btrfs_init_fs_context(struct fs_co <<
2160 { <<
2161 struct btrfs_fs_context *ctx; <<
2162 <<
2163 ctx = kzalloc(sizeof(struct btrfs_fs_ <<
2164 if (!ctx) <<
2165 return -ENOMEM; <<
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 } 2174 }
2187 2175
2188 static struct file_system_type btrfs_fs_type 2176 static struct file_system_type btrfs_fs_type = {
2189 .owner = THIS_MODULE !! 2177 .owner = THIS_MODULE,
2190 .name = "btrfs", !! 2178 .name = "btrfs",
2191 .init_fs_context = btrfs_init_ !! 2179 .mount = btrfs_mount,
2192 .parameters = btrfs_fs_pa !! 2180 .kill_sb = btrfs_kill_super,
2193 .kill_sb = btrfs_kill_ !! 2181 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2194 .fs_flags = FS_REQUIRES !! 2182 };
2195 }; <<
2196 <<
2197 MODULE_ALIAS_FS("btrfs"); 2183 MODULE_ALIAS_FS("btrfs");
2198 2184
2199 static int btrfs_control_open(struct inode *i 2185 static int btrfs_control_open(struct inode *inode, struct file *file)
2200 { 2186 {
2201 /* 2187 /*
2202 * The control file's private_data is 2188 * The control file's private_data is used to hold the
2203 * transaction when it is started and 2189 * transaction when it is started and is used to keep
2204 * track of whether a transaction is 2190 * track of whether a transaction is already in progress.
2205 */ 2191 */
2206 file->private_data = NULL; 2192 file->private_data = NULL;
2207 return 0; 2193 return 0;
2208 } 2194 }
2209 2195
2210 /* 2196 /*
2211 * Used by /dev/btrfs-control for devices ioc !! 2197 * used by btrfsctl to scan devices when no FS is mounted
2212 */ 2198 */
2213 static long btrfs_control_ioctl(struct file * 2199 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2214 unsigned long 2200 unsigned long arg)
2215 { 2201 {
2216 struct btrfs_ioctl_vol_args *vol; 2202 struct btrfs_ioctl_vol_args *vol;
2217 struct btrfs_device *device = NULL; !! 2203 struct btrfs_fs_devices *fs_devices;
2218 dev_t devt = 0; <<
2219 int ret = -ENOTTY; 2204 int ret = -ENOTTY;
2220 2205
2221 if (!capable(CAP_SYS_ADMIN)) 2206 if (!capable(CAP_SYS_ADMIN))
2222 return -EPERM; 2207 return -EPERM;
2223 2208
2224 vol = memdup_user((void __user *)arg, 2209 vol = memdup_user((void __user *)arg, sizeof(*vol));
2225 if (IS_ERR(vol)) 2210 if (IS_ERR(vol))
2226 return PTR_ERR(vol); 2211 return PTR_ERR(vol);
2227 ret = btrfs_check_ioctl_vol_args_path <<
2228 if (ret < 0) <<
2229 goto out; <<
2230 2212
2231 switch (cmd) { 2213 switch (cmd) {
2232 case BTRFS_IOC_SCAN_DEV: 2214 case BTRFS_IOC_SCAN_DEV:
2233 mutex_lock(&uuid_mutex); !! 2215 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2234 /* !! 2216 &btrfs_fs_type, &fs_devices);
2235 * Scanning outside of mount <<
2236 * into 0 error code. <<
2237 */ <<
2238 device = btrfs_scan_one_devic <<
2239 ret = PTR_ERR_OR_ZERO(device) <<
2240 mutex_unlock(&uuid_mutex); <<
2241 break; <<
2242 case BTRFS_IOC_FORGET_DEV: <<
2243 if (vol->name[0] != 0) { <<
2244 ret = lookup_bdev(vol <<
2245 if (ret) <<
2246 break; <<
2247 } <<
2248 ret = btrfs_forget_devices(de <<
2249 break; 2217 break;
2250 case BTRFS_IOC_DEVICES_READY: 2218 case BTRFS_IOC_DEVICES_READY:
2251 mutex_lock(&uuid_mutex); !! 2219 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2252 /* !! 2220 &btrfs_fs_type, &fs_devices);
2253 * Scanning outside of mount !! 2221 if (ret)
2254 * into 0 error code. <<
2255 */ <<
2256 device = btrfs_scan_one_devic <<
2257 if (IS_ERR_OR_NULL(device)) { <<
2258 mutex_unlock(&uuid_mu <<
2259 ret = PTR_ERR(device) <<
2260 break; 2222 break;
2261 } !! 2223 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2262 ret = !(device->fs_devices->n <<
2263 device->fs_devices->t <<
2264 mutex_unlock(&uuid_mutex); <<
2265 break; 2224 break;
2266 case BTRFS_IOC_GET_SUPPORTED_FEATURES 2225 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2267 ret = btrfs_ioctl_get_support 2226 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2268 break; 2227 break;
2269 } 2228 }
2270 2229
2271 out: <<
2272 kfree(vol); 2230 kfree(vol);
2273 return ret; 2231 return ret;
2274 } 2232 }
2275 2233
2276 static int btrfs_freeze(struct super_block *s 2234 static int btrfs_freeze(struct super_block *sb)
2277 { 2235 {
>> 2236 struct btrfs_trans_handle *trans;
2278 struct btrfs_fs_info *fs_info = btrfs 2237 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 2238 struct btrfs_root *root = fs_info->tree_root;
2279 2239
2280 set_bit(BTRFS_FS_FROZEN, &fs_info->fl !! 2240 fs_info->fs_frozen = 1;
2281 /* 2241 /*
2282 * We don't need a barrier here, we'l 2242 * We don't need a barrier here, we'll wait for any transaction that
2283 * could be in progress on other thre 2243 * could be in progress on other threads (and do delayed iputs that
2284 * we want to avoid on a frozen files 2244 * we want to avoid on a frozen filesystem), or do the commit
2285 * ourselves. 2245 * ourselves.
2286 */ 2246 */
2287 return btrfs_commit_current_transacti !! 2247 trans = btrfs_attach_transaction_barrier(root);
2288 } !! 2248 if (IS_ERR(trans)) {
2289 !! 2249 /* no transaction, don't bother */
2290 static int check_dev_super(struct btrfs_devic !! 2250 if (PTR_ERR(trans) == -ENOENT)
2291 { !! 2251 return 0;
2292 struct btrfs_fs_info *fs_info = dev-> !! 2252 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 } 2253 }
2337 out: !! 2254 return btrfs_commit_transaction(trans);
2338 btrfs_release_disk_super(sb); <<
2339 return ret; <<
2340 } 2255 }
2341 2256
2342 static int btrfs_unfreeze(struct super_block 2257 static int btrfs_unfreeze(struct super_block *sb)
2343 { 2258 {
2344 struct btrfs_fs_info *fs_info = btrfs !! 2259 btrfs_sb(sb)->fs_frozen = 0;
2345 struct btrfs_device *device; <<
2346 int ret = 0; <<
2347 <<
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-> <<
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; 2260 return 0;
2373 } 2261 }
2374 2262
2375 static int btrfs_show_devname(struct seq_file 2263 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2376 { 2264 {
2377 struct btrfs_fs_info *fs_info = btrfs 2265 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2378 !! 2266 struct btrfs_fs_devices *cur_devices;
2379 /* !! 2267 struct btrfs_device *dev, *first_dev = NULL;
2380 * There should be always a valid poi !! 2268 struct list_head *head;
2381 * for a short moment in case it's be !! 2269 struct rcu_string *name;
2382 * the end of RCU grace period. !! 2270
2383 */ !! 2271 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2384 rcu_read_lock(); !! 2272 cur_devices = fs_info->fs_devices;
2385 seq_escape(m, btrfs_dev_name(fs_info- !! 2273 while (cur_devices) {
2386 rcu_read_unlock(); !! 2274 head = &cur_devices->devices;
2387 !! 2275 list_for_each_entry(dev, head, dev_list) {
2388 return 0; !! 2276 if (dev->missing)
2389 } !! 2277 continue;
2390 !! 2278 if (!dev->name)
2391 static long btrfs_nr_cached_objects(struct su !! 2279 continue;
2392 { !! 2280 if (!first_dev || dev->devid < first_dev->devid)
2393 struct btrfs_fs_info *fs_info = btrfs !! 2281 first_dev = dev;
2394 const s64 nr = percpu_counter_sum_pos !! 2282 }
2395 !! 2283 cur_devices = cur_devices->seed;
2396 trace_btrfs_extent_map_shrinker_count !! 2284 }
2397 !! 2285
2398 /* !! 2286 if (first_dev) {
2399 * Only report the real number for DE !! 2287 rcu_read_lock();
2400 * serious performance degradation ca !! 2288 name = rcu_dereference(first_dev->name);
2401 */ !! 2289 seq_escape(m, name->str, " \t\n\\");
2402 if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) !! 2290 rcu_read_unlock();
2403 return nr; !! 2291 } else {
>> 2292 WARN_ON(1);
>> 2293 }
>> 2294 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2404 return 0; 2295 return 0;
2405 } 2296 }
2406 2297
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 2298 static const struct super_operations btrfs_super_ops = {
2425 .drop_inode = btrfs_drop_inode, 2299 .drop_inode = btrfs_drop_inode,
2426 .evict_inode = btrfs_evict_inode, 2300 .evict_inode = btrfs_evict_inode,
2427 .put_super = btrfs_put_super, 2301 .put_super = btrfs_put_super,
2428 .sync_fs = btrfs_sync_fs, 2302 .sync_fs = btrfs_sync_fs,
2429 .show_options = btrfs_show_options, 2303 .show_options = btrfs_show_options,
2430 .show_devname = btrfs_show_devname, 2304 .show_devname = btrfs_show_devname,
>> 2305 .write_inode = btrfs_write_inode,
2431 .alloc_inode = btrfs_alloc_inode, 2306 .alloc_inode = btrfs_alloc_inode,
2432 .destroy_inode = btrfs_destroy_inode 2307 .destroy_inode = btrfs_destroy_inode,
2433 .free_inode = btrfs_free_inode, <<
2434 .statfs = btrfs_statfs, 2308 .statfs = btrfs_statfs,
>> 2309 .remount_fs = btrfs_remount,
2435 .freeze_fs = btrfs_freeze, 2310 .freeze_fs = btrfs_freeze,
2436 .unfreeze_fs = btrfs_unfreeze, 2311 .unfreeze_fs = btrfs_unfreeze,
2437 .nr_cached_objects = btrfs_nr_cached_ <<
2438 .free_cached_objects = btrfs_free_cac <<
2439 }; 2312 };
2440 2313
2441 static const struct file_operations btrfs_ctl 2314 static const struct file_operations btrfs_ctl_fops = {
2442 .open = btrfs_control_open, 2315 .open = btrfs_control_open,
2443 .unlocked_ioctl = btrfs_control_ioct 2316 .unlocked_ioctl = btrfs_control_ioctl,
2444 .compat_ioctl = compat_ptr_ioctl, !! 2317 .compat_ioctl = btrfs_control_ioctl,
2445 .owner = THIS_MODULE, 2318 .owner = THIS_MODULE,
2446 .llseek = noop_llseek, 2319 .llseek = noop_llseek,
2447 }; 2320 };
2448 2321
2449 static struct miscdevice btrfs_misc = { 2322 static struct miscdevice btrfs_misc = {
2450 .minor = BTRFS_MINOR, 2323 .minor = BTRFS_MINOR,
2451 .name = "btrfs-control", 2324 .name = "btrfs-control",
2452 .fops = &btrfs_ctl_fops 2325 .fops = &btrfs_ctl_fops
2453 }; 2326 };
2454 2327
2455 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 2328 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2456 MODULE_ALIAS("devname:btrfs-control"); 2329 MODULE_ALIAS("devname:btrfs-control");
2457 2330
2458 static int __init btrfs_interface_init(void) !! 2331 static int btrfs_interface_init(void)
2459 { 2332 {
2460 return misc_register(&btrfs_misc); 2333 return misc_register(&btrfs_misc);
2461 } 2334 }
2462 2335
2463 static __cold void btrfs_interface_exit(void) !! 2336 static void btrfs_interface_exit(void)
2464 { 2337 {
2465 misc_deregister(&btrfs_misc); 2338 misc_deregister(&btrfs_misc);
2466 } 2339 }
2467 2340
2468 static int __init btrfs_print_mod_info(void) !! 2341 static void btrfs_print_mod_info(void)
2469 { 2342 {
2470 static const char options[] = "" !! 2343 pr_info("Btrfs loaded, crc32c=%s"
2471 #ifdef CONFIG_BTRFS_DEBUG 2344 #ifdef CONFIG_BTRFS_DEBUG
2472 ", debug=on" 2345 ", debug=on"
2473 #endif 2346 #endif
2474 #ifdef CONFIG_BTRFS_ASSERT 2347 #ifdef CONFIG_BTRFS_ASSERT
2475 ", assert=on" 2348 ", assert=on"
2476 #endif 2349 #endif
2477 #ifdef CONFIG_BTRFS_FS_REF_VERIFY !! 2350 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2478 ", ref-verify=on" !! 2351 ", integrity-checker=on"
2479 #endif 2352 #endif
2480 #ifdef CONFIG_BLK_DEV_ZONED !! 2353 "\n",
2481 ", zoned=yes" !! 2354 btrfs_crc32c_impl());
2482 #else <<
2483 ", zoned=no" <<
2484 #endif <<
2485 #ifdef CONFIG_FS_VERITY <<
2486 ", fsverity=yes" <<
2487 #else <<
2488 ", fsverity=no" <<
2489 #endif <<
2490 ; <<
2491 pr_info("Btrfs loaded%s\n", options); <<
2492 return 0; <<
2493 } 2355 }
2494 2356
2495 static int register_btrfs(void) !! 2357 static int __init init_btrfs_fs(void)
2496 { 2358 {
2497 return register_filesystem(&btrfs_fs_ !! 2359 int err;
2498 } <<
2499 2360
2500 static void unregister_btrfs(void) !! 2361 err = btrfs_hash_init();
2501 { !! 2362 if (err)
2502 unregister_filesystem(&btrfs_fs_type) !! 2363 return err;
2503 } <<
2504 2364
2505 /* Helper structure for long init/exit functi !! 2365 btrfs_props_init();
2506 struct init_sequence { <<
2507 int (*init_func)(void); <<
2508 /* Can be NULL if the init_func doesn <<
2509 void (*exit_func)(void); <<
2510 }; <<
2511 2366
2512 static const struct init_sequence mod_init_se !! 2367 err = btrfs_init_sysfs();
2513 { !! 2368 if (err)
2514 .init_func = btrfs_props_init !! 2369 goto free_hash;
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 2370
2579 static bool mod_init_result[ARRAY_SIZE(mod_in !! 2371 btrfs_init_compress();
2580 2372
2581 static __always_inline void btrfs_exit_btrfs_ !! 2373 err = btrfs_init_cachep();
2582 { !! 2374 if (err)
2583 int i; !! 2375 goto free_compress;
2584 2376
2585 for (i = ARRAY_SIZE(mod_init_seq) - 1 !! 2377 err = extent_io_init();
2586 if (!mod_init_result[i]) !! 2378 if (err)
2587 continue; !! 2379 goto free_cachep;
2588 if (mod_init_seq[i].exit_func !! 2380
2589 mod_init_seq[i].exit_ !! 2381 err = extent_map_init();
2590 mod_init_result[i] = false; !! 2382 if (err)
2591 } !! 2383 goto free_extent_io;
>> 2384
>> 2385 err = ordered_data_init();
>> 2386 if (err)
>> 2387 goto free_extent_map;
>> 2388
>> 2389 err = btrfs_delayed_inode_init();
>> 2390 if (err)
>> 2391 goto free_ordered_data;
>> 2392
>> 2393 err = btrfs_auto_defrag_init();
>> 2394 if (err)
>> 2395 goto free_delayed_inode;
>> 2396
>> 2397 err = btrfs_delayed_ref_init();
>> 2398 if (err)
>> 2399 goto free_auto_defrag;
>> 2400
>> 2401 err = btrfs_prelim_ref_init();
>> 2402 if (err)
>> 2403 goto free_delayed_ref;
>> 2404
>> 2405 err = btrfs_end_io_wq_init();
>> 2406 if (err)
>> 2407 goto free_prelim_ref;
>> 2408
>> 2409 err = btrfs_interface_init();
>> 2410 if (err)
>> 2411 goto free_end_io_wq;
>> 2412
>> 2413 btrfs_init_lockdep();
>> 2414
>> 2415 btrfs_print_mod_info();
>> 2416
>> 2417 err = btrfs_run_sanity_tests();
>> 2418 if (err)
>> 2419 goto unregister_ioctl;
>> 2420
>> 2421 err = register_filesystem(&btrfs_fs_type);
>> 2422 if (err)
>> 2423 goto unregister_ioctl;
>> 2424
>> 2425 return 0;
>> 2426
>> 2427 unregister_ioctl:
>> 2428 btrfs_interface_exit();
>> 2429 free_end_io_wq:
>> 2430 btrfs_end_io_wq_exit();
>> 2431 free_prelim_ref:
>> 2432 btrfs_prelim_ref_exit();
>> 2433 free_delayed_ref:
>> 2434 btrfs_delayed_ref_exit();
>> 2435 free_auto_defrag:
>> 2436 btrfs_auto_defrag_exit();
>> 2437 free_delayed_inode:
>> 2438 btrfs_delayed_inode_exit();
>> 2439 free_ordered_data:
>> 2440 ordered_data_exit();
>> 2441 free_extent_map:
>> 2442 extent_map_exit();
>> 2443 free_extent_io:
>> 2444 extent_io_exit();
>> 2445 free_cachep:
>> 2446 btrfs_destroy_cachep();
>> 2447 free_compress:
>> 2448 btrfs_exit_compress();
>> 2449 btrfs_exit_sysfs();
>> 2450 free_hash:
>> 2451 btrfs_hash_exit();
>> 2452 return err;
2592 } 2453 }
2593 2454
2594 static void __exit exit_btrfs_fs(void) 2455 static void __exit exit_btrfs_fs(void)
2595 { 2456 {
2596 btrfs_exit_btrfs_fs(); !! 2457 btrfs_destroy_cachep();
>> 2458 btrfs_delayed_ref_exit();
>> 2459 btrfs_auto_defrag_exit();
>> 2460 btrfs_delayed_inode_exit();
>> 2461 btrfs_prelim_ref_exit();
>> 2462 ordered_data_exit();
>> 2463 extent_map_exit();
>> 2464 extent_io_exit();
>> 2465 btrfs_interface_exit();
>> 2466 btrfs_end_io_wq_exit();
>> 2467 unregister_filesystem(&btrfs_fs_type);
>> 2468 btrfs_exit_sysfs();
2597 btrfs_cleanup_fs_uuids(); 2469 btrfs_cleanup_fs_uuids();
2598 } !! 2470 btrfs_exit_compress();
2599 !! 2471 btrfs_hash_exit();
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 } 2472 }
2616 2473
2617 late_initcall(init_btrfs_fs); 2474 late_initcall(init_btrfs_fs);
2618 module_exit(exit_btrfs_fs) 2475 module_exit(exit_btrfs_fs)
2619 2476
2620 MODULE_DESCRIPTION("B-Tree File System (BTRFS <<
2621 MODULE_LICENSE("GPL"); 2477 MODULE_LICENSE("GPL");
2622 MODULE_SOFTDEP("pre: crc32c"); <<
2623 MODULE_SOFTDEP("pre: xxhash64"); <<
2624 MODULE_SOFTDEP("pre: sha256"); <<
2625 MODULE_SOFTDEP("pre: blake2b-256"); <<
2626 2478
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.