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 ACCESS_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 break;
>> 553 case Opt_ssd_spread:
>> 554 btrfs_set_and_info(info, SSD_SPREAD,
>> 555 "use spread ssd allocation scheme");
>> 556 btrfs_set_opt(info->mount_opt, SSD);
>> 557 break;
>> 558 case Opt_nossd:
>> 559 btrfs_set_and_info(info, NOSSD,
>> 560 "not using ssd allocation scheme");
>> 561 btrfs_clear_opt(info->mount_opt, SSD);
>> 562 break;
>> 563 case Opt_barrier:
>> 564 btrfs_clear_and_info(info, NOBARRIER,
>> 565 "turning on barriers");
>> 566 break;
>> 567 case Opt_nobarrier:
>> 568 btrfs_set_and_info(info, NOBARRIER,
>> 569 "turning off barriers");
>> 570 break;
>> 571 case Opt_thread_pool:
>> 572 ret = match_int(&args[0], &intarg);
>> 573 if (ret) {
>> 574 goto out;
>> 575 } else if (intarg > 0) {
>> 576 info->thread_pool_size = intarg;
>> 577 } else {
>> 578 ret = -EINVAL;
>> 579 goto out;
>> 580 }
>> 581 break;
>> 582 case Opt_max_inline:
>> 583 num = match_strdup(&args[0]);
>> 584 if (num) {
>> 585 info->max_inline = memparse(num, NULL);
>> 586 kfree(num);
>> 587
>> 588 if (info->max_inline) {
>> 589 info->max_inline = min_t(u64,
>> 590 info->max_inline,
>> 591 info->sectorsize);
>> 592 }
>> 593 btrfs_info(info, "max_inline at %llu",
>> 594 info->max_inline);
>> 595 } else {
>> 596 ret = -ENOMEM;
>> 597 goto out;
>> 598 }
>> 599 break;
>> 600 case Opt_alloc_start:
>> 601 num = match_strdup(&args[0]);
>> 602 if (num) {
>> 603 mutex_lock(&info->chunk_mutex);
>> 604 info->alloc_start = memparse(num, NULL);
>> 605 mutex_unlock(&info->chunk_mutex);
>> 606 kfree(num);
>> 607 btrfs_info(info, "allocations start at %llu",
>> 608 info->alloc_start);
>> 609 } else {
>> 610 ret = -ENOMEM;
>> 611 goto out;
>> 612 }
>> 613 break;
>> 614 case Opt_acl:
430 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 615 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
431 fc->sb_flags |= SB_POS !! 616 info->sb->s_flags |= MS_POSIXACL;
>> 617 break;
432 #else 618 #else
433 btrfs_err(NULL, "suppo !! 619 btrfs_err(info, "support for ACL not compiled in!");
434 return -EINVAL; !! 620 ret = -EINVAL;
>> 621 goto out;
435 #endif 622 #endif
436 } !! 623 case Opt_noacl:
437 /* !! 624 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; 625 break;
492 default: !! 626 case Opt_notreelog:
493 btrfs_err(NULL, "unrec !! 627 btrfs_set_and_info(info, NOTREELOG,
494 param->key); !! 628 "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; 629 break;
519 default: !! 630 case Opt_treelog:
520 btrfs_err(NULL, "unrec !! 631 btrfs_clear_and_info(info, NOTREELOG,
521 param->key); !! 632 "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; 633 break;
562 case Opt_fatal_errors_bug: !! 634 case Opt_norecovery:
563 btrfs_clear_opt(ctx->m !! 635 case Opt_nologreplay:
>> 636 btrfs_set_and_info(info, NOLOGREPLAY,
>> 637 "disabling log replay at mount time");
564 break; 638 break;
565 default: !! 639 case Opt_flushoncommit:
566 btrfs_err(NULL, "unrec !! 640 btrfs_set_and_info(info, FLUSHONCOMMIT,
567 param->key); !! 641 "turning on flush-on-commit");
568 return -EINVAL; !! 642 break;
569 } !! 643 case Opt_noflushoncommit:
570 break; !! 644 btrfs_clear_and_info(info, FLUSHONCOMMIT,
571 case Opt_commit_interval: !! 645 "turning off flush-on-commit");
572 ctx->commit_interval = result. !! 646 break;
573 if (ctx->commit_interval == 0) !! 647 case Opt_ratio:
574 ctx->commit_interval = !! 648 ret = match_int(&args[0], &intarg);
575 break; !! 649 if (ret) {
576 case Opt_rescue: !! 650 goto out;
577 switch (result.uint_32) { !! 651 } else if (intarg >= 0) {
578 case Opt_rescue_usebackuproot: !! 652 info->metadata_ratio = intarg;
579 btrfs_set_opt(ctx->mou !! 653 btrfs_info(info, "metadata ratio %d",
580 break; !! 654 info->metadata_ratio);
581 case Opt_rescue_nologreplay: !! 655 } else {
582 btrfs_set_opt(ctx->mou !! 656 ret = -EINVAL;
583 break; !! 657 goto out;
584 case Opt_rescue_ignorebadroots !! 658 }
585 btrfs_set_opt(ctx->mou !! 659 break;
586 break; !! 660 case Opt_discard:
587 case Opt_rescue_ignoredatacsum !! 661 btrfs_set_and_info(info, DISCARD,
588 btrfs_set_opt(ctx->mou !! 662 "turning on discard");
589 break; !! 663 break;
590 case Opt_rescue_ignoremetacsum !! 664 case Opt_nodiscard:
591 btrfs_set_opt(ctx->mou !! 665 btrfs_clear_and_info(info, DISCARD,
592 break; !! 666 "turning off discard");
593 case Opt_rescue_ignoresuperfla !! 667 break;
594 btrfs_set_opt(ctx->mou !! 668 case Opt_space_cache:
595 break; !! 669 case Opt_space_cache_version:
596 case Opt_rescue_parameter_all: !! 670 if (token == Opt_space_cache ||
597 btrfs_set_opt(ctx->mou !! 671 strcmp(args[0].from, "v1") == 0) {
598 btrfs_set_opt(ctx->mou !! 672 btrfs_clear_opt(info->mount_opt,
599 btrfs_set_opt(ctx->mou !! 673 FREE_SPACE_TREE);
600 btrfs_set_opt(ctx->mou !! 674 btrfs_set_and_info(info, SPACE_CACHE,
601 btrfs_set_opt(ctx->mou !! 675 "enabling disk space caching");
>> 676 } else if (strcmp(args[0].from, "v2") == 0) {
>> 677 btrfs_clear_opt(info->mount_opt,
>> 678 SPACE_CACHE);
>> 679 btrfs_set_and_info(info, FREE_SPACE_TREE,
>> 680 "enabling free space tree");
>> 681 } else {
>> 682 ret = -EINVAL;
>> 683 goto out;
>> 684 }
>> 685 break;
>> 686 case Opt_rescan_uuid_tree:
>> 687 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
>> 688 break;
>> 689 case Opt_no_space_cache:
>> 690 if (btrfs_test_opt(info, SPACE_CACHE)) {
>> 691 btrfs_clear_and_info(info, SPACE_CACHE,
>> 692 "disabling disk space caching");
>> 693 }
>> 694 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
>> 695 btrfs_clear_and_info(info, FREE_SPACE_TREE,
>> 696 "disabling free space tree");
>> 697 }
>> 698 break;
>> 699 case Opt_inode_cache:
>> 700 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
>> 701 "enabling inode map caching");
>> 702 break;
>> 703 case Opt_noinode_cache:
>> 704 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
>> 705 "disabling inode map caching");
>> 706 break;
>> 707 case Opt_clear_cache:
>> 708 btrfs_set_and_info(info, CLEAR_CACHE,
>> 709 "force clearing of disk cache");
>> 710 break;
>> 711 case Opt_user_subvol_rm_allowed:
>> 712 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
>> 713 break;
>> 714 case Opt_enospc_debug:
>> 715 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
>> 716 break;
>> 717 case Opt_noenospc_debug:
>> 718 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
>> 719 break;
>> 720 case Opt_defrag:
>> 721 btrfs_set_and_info(info, AUTO_DEFRAG,
>> 722 "enabling auto defrag");
>> 723 break;
>> 724 case Opt_nodefrag:
>> 725 btrfs_clear_and_info(info, AUTO_DEFRAG,
>> 726 "disabling auto defrag");
>> 727 break;
>> 728 case Opt_recovery:
>> 729 btrfs_warn(info,
>> 730 "'recovery' is deprecated, use 'usebackuproot' instead");
>> 731 case Opt_usebackuproot:
>> 732 btrfs_info(info,
>> 733 "trying to use backup root at mount time");
>> 734 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
>> 735 break;
>> 736 case Opt_skip_balance:
>> 737 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
>> 738 break;
>> 739 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 740 case Opt_check_integrity_including_extent_data:
>> 741 btrfs_info(info,
>> 742 "enabling check integrity including extent data");
>> 743 btrfs_set_opt(info->mount_opt,
>> 744 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
>> 745 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
>> 746 break;
>> 747 case Opt_check_integrity:
>> 748 btrfs_info(info, "enabling check integrity");
>> 749 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
>> 750 break;
>> 751 case Opt_check_integrity_print_mask:
>> 752 ret = match_int(&args[0], &intarg);
>> 753 if (ret) {
>> 754 goto out;
>> 755 } else if (intarg >= 0) {
>> 756 info->check_integrity_print_mask = intarg;
>> 757 btrfs_info(info,
>> 758 "check_integrity_print_mask 0x%x",
>> 759 info->check_integrity_print_mask);
>> 760 } else {
>> 761 ret = -EINVAL;
>> 762 goto out;
>> 763 }
>> 764 break;
>> 765 #else
>> 766 case Opt_check_integrity_including_extent_data:
>> 767 case Opt_check_integrity:
>> 768 case Opt_check_integrity_print_mask:
>> 769 btrfs_err(info,
>> 770 "support for check_integrity* not compiled in!");
>> 771 ret = -EINVAL;
>> 772 goto out;
>> 773 #endif
>> 774 case Opt_fatal_errors:
>> 775 if (strcmp(args[0].from, "panic") == 0)
>> 776 btrfs_set_opt(info->mount_opt,
>> 777 PANIC_ON_FATAL_ERROR);
>> 778 else if (strcmp(args[0].from, "bug") == 0)
>> 779 btrfs_clear_opt(info->mount_opt,
>> 780 PANIC_ON_FATAL_ERROR);
>> 781 else {
>> 782 ret = -EINVAL;
>> 783 goto out;
>> 784 }
>> 785 break;
>> 786 case Opt_commit_interval:
>> 787 intarg = 0;
>> 788 ret = match_int(&args[0], &intarg);
>> 789 if (ret < 0) {
>> 790 btrfs_err(info, "invalid commit interval");
>> 791 ret = -EINVAL;
>> 792 goto out;
>> 793 }
>> 794 if (intarg > 0) {
>> 795 if (intarg > 300) {
>> 796 btrfs_warn(info,
>> 797 "excessive commit interval %d",
>> 798 intarg);
>> 799 }
>> 800 info->commit_interval = intarg;
>> 801 } else {
>> 802 btrfs_info(info,
>> 803 "using default commit interval %ds",
>> 804 BTRFS_DEFAULT_COMMIT_INTERVAL);
>> 805 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
>> 806 }
602 break; 807 break;
603 default: <<
604 btrfs_info(NULL, "unre <<
605 param->key) <<
606 return -EINVAL; <<
607 } <<
608 break; <<
609 #ifdef CONFIG_BTRFS_DEBUG 808 #ifdef CONFIG_BTRFS_DEBUG
610 case Opt_fragment: !! 809 case Opt_fragment_all:
611 switch (result.uint_32) { !! 810 btrfs_info(info, "fragmenting all space");
612 case Opt_fragment_parameter_al !! 811 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
613 btrfs_set_opt(ctx->mou !! 812 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
614 btrfs_set_opt(ctx->mou <<
615 break; 813 break;
616 case Opt_fragment_parameter_me !! 814 case Opt_fragment_metadata:
617 btrfs_set_opt(ctx->mou !! 815 btrfs_info(info, "fragmenting metadata");
>> 816 btrfs_set_opt(info->mount_opt,
>> 817 FRAGMENT_METADATA);
618 break; 818 break;
619 case Opt_fragment_parameter_da !! 819 case Opt_fragment_data:
620 btrfs_set_opt(ctx->mou !! 820 btrfs_info(info, "fragmenting data");
>> 821 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
621 break; 822 break;
>> 823 #endif
>> 824 case Opt_err:
>> 825 btrfs_info(info, "unrecognized mount option '%s'", p);
>> 826 ret = -EINVAL;
>> 827 goto out;
622 default: 828 default:
623 btrfs_info(NULL, "unre !! 829 break;
624 param->key) <<
625 return -EINVAL; <<
626 } 830 }
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 } 831 }
638 !! 832 check:
639 return 0; !! 833 /*
640 } !! 834 * Extra check for current option against current flag
641 !! 835 */
642 /* !! 836 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
643 * Some options only have meaning at mount tim !! 837 btrfs_err(info,
644 * remounts, or be displayed. Clear these at t !! 838 "nologreplay must be used with ro mount option");
645 * paths. !! 839 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 } 840 }
663 return false; !! 841 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 842 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
681 !btrfs_raw_test_opt(*mount_opt, FR !! 843 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
682 !btrfs_raw_test_opt(*mount_opt, CL !! 844 !btrfs_test_opt(info, CLEAR_CACHE)) {
683 btrfs_err(info, "cannot disabl !! 845 btrfs_err(info, "cannot disable free space tree");
684 ret = false; !! 846 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 } <<
704 847
>> 848 }
>> 849 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
>> 850 btrfs_info(info, "disk space caching is enabled");
>> 851 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
>> 852 btrfs_info(info, "using free space tree");
>> 853 kfree(orig);
705 return ret; 854 return ret;
706 } 855 }
707 856
708 /* 857 /*
709 * This is subtle, we only call this during op !! 858 * 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 * 859 *
714 * This isn't a change in behavior, because we !! 860 * All other options will be parsed on much later in the mount process and
715 * file system to set the current mount option !! 861 * 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 */ 862 */
720 void btrfs_set_free_space_cache_settings(struc !! 863 static int btrfs_parse_early_options(const char *options, fmode_t flags,
721 { !! 864 void *holder, char **subvol_name, u64 *subvol_objectid,
722 if (fs_info->sectorsize < PAGE_SIZE) { !! 865 struct btrfs_fs_devices **fs_devices)
723 btrfs_clear_opt(fs_info->mount !! 866 {
724 if (!btrfs_test_opt(fs_info, F !! 867 substring_t args[MAX_OPT_ARGS];
725 btrfs_info(fs_info, !! 868 char *device_name, *opts, *orig, *p;
726 "forcing fr !! 869 char *num = NULL;
727 fs_info->se !! 870 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; <<
748 871
749 if (btrfs_test_opt(fs_info, NOSPACECAC !! 872 if (!options)
750 return; !! 873 return 0;
751 874
752 /* 875 /*
753 * At this point we don't have explici !! 876 * strsep changes the string, duplicate it because parse_options
754 * them ourselves based on the state o !! 877 * gets called twice
755 */ 878 */
756 if (btrfs_fs_compat_ro(fs_info, FREE_S !! 879 opts = kstrdup(options, GFP_KERNEL);
757 btrfs_set_opt(fs_info->mount_o !! 880 if (!opts)
758 else if (btrfs_free_space_cache_v1_act !! 881 return -ENOMEM;
759 btrfs_set_opt(fs_info->mount_o !! 882 orig = opts;
760 } <<
761 883
762 static void set_device_specific_options(struct !! 884 while ((p = strsep(&opts, ",")) != NULL) {
763 { !! 885 int token;
764 if (!btrfs_test_opt(fs_info, NOSSD) && !! 886 if (!*p)
765 !fs_info->fs_devices->rotating) !! 887 continue;
766 btrfs_set_opt(fs_info->mount_o <<
767 888
768 /* !! 889 token = match_token(p, tokens, args);
769 * For devices supporting discard turn !! 890 switch (token) {
770 * unless it's already set or disabled !! 891 case Opt_subvol:
771 * nodiscard for the same mount. !! 892 kfree(*subvol_name);
772 * !! 893 *subvol_name = match_strdup(&args[0]);
773 * The zoned mode piggy backs on the d !! 894 if (!*subvol_name) {
774 * resetting a zone. There is no reaso !! 895 error = -ENOMEM;
775 * fast enough. So, do not enable asyn !! 896 goto out;
776 */ !! 897 }
777 if (!(btrfs_test_opt(fs_info, DISCARD_ !! 898 break;
778 btrfs_test_opt(fs_info, DISCARD_ !! 899 case Opt_subvolid:
779 btrfs_test_opt(fs_info, NODISCAR !! 900 num = match_strdup(&args[0]);
780 fs_info->fs_devices->discardable & !! 901 if (num) {
781 !btrfs_is_zoned(fs_info)) !! 902 *subvol_objectid = memparse(num, NULL);
782 btrfs_set_opt(fs_info->mount_o !! 903 kfree(num);
>> 904 /* we want the original fs_tree */
>> 905 if (!*subvol_objectid)
>> 906 *subvol_objectid =
>> 907 BTRFS_FS_TREE_OBJECTID;
>> 908 } else {
>> 909 error = -EINVAL;
>> 910 goto out;
>> 911 }
>> 912 break;
>> 913 case Opt_subvolrootid:
>> 914 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
>> 915 break;
>> 916 case Opt_device:
>> 917 device_name = match_strdup(&args[0]);
>> 918 if (!device_name) {
>> 919 error = -ENOMEM;
>> 920 goto out;
>> 921 }
>> 922 error = btrfs_scan_one_device(device_name,
>> 923 flags, holder, fs_devices);
>> 924 kfree(device_name);
>> 925 if (error)
>> 926 goto out;
>> 927 break;
>> 928 default:
>> 929 break;
>> 930 }
>> 931 }
>> 932
>> 933 out:
>> 934 kfree(orig);
>> 935 return error;
783 } 936 }
784 937
785 char *btrfs_get_subvol_name_from_objectid(stru !! 938 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
786 u64 !! 939 u64 subvol_objectid)
787 { 940 {
788 struct btrfs_root *root = fs_info->tre 941 struct btrfs_root *root = fs_info->tree_root;
789 struct btrfs_root *fs_root = NULL; !! 942 struct btrfs_root *fs_root;
790 struct btrfs_root_ref *root_ref; 943 struct btrfs_root_ref *root_ref;
791 struct btrfs_inode_ref *inode_ref; 944 struct btrfs_inode_ref *inode_ref;
792 struct btrfs_key key; 945 struct btrfs_key key;
793 struct btrfs_path *path = NULL; 946 struct btrfs_path *path = NULL;
794 char *name = NULL, *ptr; 947 char *name = NULL, *ptr;
795 u64 dirid; 948 u64 dirid;
796 int len; 949 int len;
797 int ret; 950 int ret;
798 951
799 path = btrfs_alloc_path(); 952 path = btrfs_alloc_path();
800 if (!path) { 953 if (!path) {
801 ret = -ENOMEM; 954 ret = -ENOMEM;
802 goto err; 955 goto err;
803 } 956 }
>> 957 path->leave_spinning = 1;
804 958
805 name = kmalloc(PATH_MAX, GFP_KERNEL); !! 959 name = kmalloc(PATH_MAX, GFP_NOFS);
806 if (!name) { 960 if (!name) {
807 ret = -ENOMEM; 961 ret = -ENOMEM;
808 goto err; 962 goto err;
809 } 963 }
810 ptr = name + PATH_MAX - 1; 964 ptr = name + PATH_MAX - 1;
811 ptr[0] = '\0'; 965 ptr[0] = '\0';
812 966
813 /* 967 /*
814 * Walk up the subvolume trees in the 968 * Walk up the subvolume trees in the tree of tree roots by root
815 * backrefs until we hit the top-level 969 * backrefs until we hit the top-level subvolume.
816 */ 970 */
817 while (subvol_objectid != BTRFS_FS_TRE 971 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
818 key.objectid = subvol_objectid 972 key.objectid = subvol_objectid;
819 key.type = BTRFS_ROOT_BACKREF_ 973 key.type = BTRFS_ROOT_BACKREF_KEY;
820 key.offset = (u64)-1; 974 key.offset = (u64)-1;
821 975
822 ret = btrfs_search_backwards(r !! 976 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
823 if (ret < 0) { 977 if (ret < 0) {
824 goto err; 978 goto err;
825 } else if (ret > 0) { 979 } else if (ret > 0) {
826 ret = -ENOENT; !! 980 ret = btrfs_previous_item(root, path, subvol_objectid,
827 goto err; !! 981 BTRFS_ROOT_BACKREF_KEY);
>> 982 if (ret < 0) {
>> 983 goto err;
>> 984 } else if (ret > 0) {
>> 985 ret = -ENOENT;
>> 986 goto err;
>> 987 }
828 } 988 }
829 989
>> 990 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
830 subvol_objectid = key.offset; 991 subvol_objectid = key.offset;
831 992
832 root_ref = btrfs_item_ptr(path 993 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
833 stru 994 struct btrfs_root_ref);
834 len = btrfs_root_ref_name_len( 995 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
835 ptr -= len + 1; 996 ptr -= len + 1;
836 if (ptr < name) { 997 if (ptr < name) {
837 ret = -ENAMETOOLONG; 998 ret = -ENAMETOOLONG;
838 goto err; 999 goto err;
839 } 1000 }
840 read_extent_buffer(path->nodes 1001 read_extent_buffer(path->nodes[0], ptr + 1,
841 (unsigned l 1002 (unsigned long)(root_ref + 1), len);
842 ptr[0] = '/'; 1003 ptr[0] = '/';
843 dirid = btrfs_root_ref_dirid(p 1004 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
844 btrfs_release_path(path); 1005 btrfs_release_path(path);
845 1006
846 fs_root = btrfs_get_fs_root(fs !! 1007 key.objectid = subvol_objectid;
>> 1008 key.type = BTRFS_ROOT_ITEM_KEY;
>> 1009 key.offset = (u64)-1;
>> 1010 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
847 if (IS_ERR(fs_root)) { 1011 if (IS_ERR(fs_root)) {
848 ret = PTR_ERR(fs_root) 1012 ret = PTR_ERR(fs_root);
849 fs_root = NULL; <<
850 goto err; 1013 goto err;
851 } 1014 }
852 1015
853 /* 1016 /*
854 * Walk up the filesystem tree 1017 * Walk up the filesystem tree by inode refs until we hit the
855 * root directory. 1018 * root directory.
856 */ 1019 */
857 while (dirid != BTRFS_FIRST_FR 1020 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
858 key.objectid = dirid; 1021 key.objectid = dirid;
859 key.type = BTRFS_INODE 1022 key.type = BTRFS_INODE_REF_KEY;
860 key.offset = (u64)-1; 1023 key.offset = (u64)-1;
861 1024
862 ret = btrfs_search_bac !! 1025 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
863 if (ret < 0) { 1026 if (ret < 0) {
864 goto err; 1027 goto err;
865 } else if (ret > 0) { 1028 } else if (ret > 0) {
866 ret = -ENOENT; !! 1029 ret = btrfs_previous_item(fs_root, path, dirid,
867 goto err; !! 1030 BTRFS_INODE_REF_KEY);
>> 1031 if (ret < 0) {
>> 1032 goto err;
>> 1033 } else if (ret > 0) {
>> 1034 ret = -ENOENT;
>> 1035 goto err;
>> 1036 }
868 } 1037 }
869 1038
>> 1039 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
870 dirid = key.offset; 1040 dirid = key.offset;
871 1041
872 inode_ref = btrfs_item 1042 inode_ref = btrfs_item_ptr(path->nodes[0],
873 1043 path->slots[0],
874 1044 struct btrfs_inode_ref);
875 len = btrfs_inode_ref_ 1045 len = btrfs_inode_ref_name_len(path->nodes[0],
876 1046 inode_ref);
877 ptr -= len + 1; 1047 ptr -= len + 1;
878 if (ptr < name) { 1048 if (ptr < name) {
879 ret = -ENAMETO 1049 ret = -ENAMETOOLONG;
880 goto err; 1050 goto err;
881 } 1051 }
882 read_extent_buffer(pat 1052 read_extent_buffer(path->nodes[0], ptr + 1,
883 (un 1053 (unsigned long)(inode_ref + 1), len);
884 ptr[0] = '/'; 1054 ptr[0] = '/';
885 btrfs_release_path(pat 1055 btrfs_release_path(path);
886 } 1056 }
887 btrfs_put_root(fs_root); <<
888 fs_root = NULL; <<
889 } 1057 }
890 1058
891 btrfs_free_path(path); 1059 btrfs_free_path(path);
892 if (ptr == name + PATH_MAX - 1) { 1060 if (ptr == name + PATH_MAX - 1) {
893 name[0] = '/'; 1061 name[0] = '/';
894 name[1] = '\0'; 1062 name[1] = '\0';
895 } else { 1063 } else {
896 memmove(name, ptr, name + PATH 1064 memmove(name, ptr, name + PATH_MAX - ptr);
897 } 1065 }
898 return name; 1066 return name;
899 1067
900 err: 1068 err:
901 btrfs_put_root(fs_root); <<
902 btrfs_free_path(path); 1069 btrfs_free_path(path);
903 kfree(name); 1070 kfree(name);
904 return ERR_PTR(ret); 1071 return ERR_PTR(ret);
905 } 1072 }
906 1073
907 static int get_default_subvol_objectid(struct 1074 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
908 { 1075 {
909 struct btrfs_root *root = fs_info->tre 1076 struct btrfs_root *root = fs_info->tree_root;
910 struct btrfs_dir_item *di; 1077 struct btrfs_dir_item *di;
911 struct btrfs_path *path; 1078 struct btrfs_path *path;
912 struct btrfs_key location; 1079 struct btrfs_key location;
913 struct fscrypt_str name = FSTR_INIT("d <<
914 u64 dir_id; 1080 u64 dir_id;
915 1081
916 path = btrfs_alloc_path(); 1082 path = btrfs_alloc_path();
917 if (!path) 1083 if (!path)
918 return -ENOMEM; 1084 return -ENOMEM;
>> 1085 path->leave_spinning = 1;
919 1086
920 /* 1087 /*
921 * Find the "default" dir item which p 1088 * Find the "default" dir item which points to the root item that we
922 * will mount by default if we haven't 1089 * will mount by default if we haven't been given a specific subvolume
923 * to mount. 1090 * to mount.
924 */ 1091 */
925 dir_id = btrfs_super_root_dir(fs_info- 1092 dir_id = btrfs_super_root_dir(fs_info->super_copy);
926 di = btrfs_lookup_dir_item(NULL, root, !! 1093 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
927 if (IS_ERR(di)) { 1094 if (IS_ERR(di)) {
928 btrfs_free_path(path); 1095 btrfs_free_path(path);
929 return PTR_ERR(di); 1096 return PTR_ERR(di);
930 } 1097 }
931 if (!di) { 1098 if (!di) {
932 /* 1099 /*
933 * Ok the default dir item isn 1100 * Ok the default dir item isn't there. This is weird since
934 * it's always been there, but 1101 * it's always been there, but don't freak out, just try and
935 * mount the top-level subvolu 1102 * mount the top-level subvolume.
936 */ 1103 */
937 btrfs_free_path(path); 1104 btrfs_free_path(path);
938 *objectid = BTRFS_FS_TREE_OBJE 1105 *objectid = BTRFS_FS_TREE_OBJECTID;
939 return 0; 1106 return 0;
940 } 1107 }
941 1108
942 btrfs_dir_item_key_to_cpu(path->nodes[ 1109 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
943 btrfs_free_path(path); 1110 btrfs_free_path(path);
944 *objectid = location.objectid; 1111 *objectid = location.objectid;
945 return 0; 1112 return 0;
946 } 1113 }
947 1114
948 static int btrfs_fill_super(struct super_block 1115 static int btrfs_fill_super(struct super_block *sb,
949 struct btrfs_fs_de 1116 struct btrfs_fs_devices *fs_devices,
950 void *data) !! 1117 void *data, int silent)
951 { 1118 {
952 struct inode *inode; 1119 struct inode *inode;
953 struct btrfs_fs_info *fs_info = btrfs_ 1120 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 1121 struct btrfs_key key;
954 int err; 1122 int err;
955 1123
956 sb->s_maxbytes = MAX_LFS_FILESIZE; 1124 sb->s_maxbytes = MAX_LFS_FILESIZE;
957 sb->s_magic = BTRFS_SUPER_MAGIC; 1125 sb->s_magic = BTRFS_SUPER_MAGIC;
958 sb->s_op = &btrfs_super_ops; 1126 sb->s_op = &btrfs_super_ops;
959 sb->s_d_op = &btrfs_dentry_operations; 1127 sb->s_d_op = &btrfs_dentry_operations;
960 sb->s_export_op = &btrfs_export_ops; 1128 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; 1129 sb->s_xattr = btrfs_xattr_handlers;
965 sb->s_time_gran = 1; 1130 sb->s_time_gran = 1;
>> 1131 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
>> 1132 sb->s_flags |= MS_POSIXACL;
>> 1133 #endif
>> 1134 sb->s_flags |= MS_I_VERSION;
966 sb->s_iflags |= SB_I_CGROUPWB; 1135 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 1136 err = open_ctree(sb, fs_devices, (char *)data);
975 if (err) { 1137 if (err) {
976 btrfs_err(fs_info, "open_ctree 1138 btrfs_err(fs_info, "open_ctree failed");
977 return err; 1139 return err;
978 } 1140 }
979 1141
980 inode = btrfs_iget(BTRFS_FIRST_FREE_OB !! 1142 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
>> 1143 key.type = BTRFS_INODE_ITEM_KEY;
>> 1144 key.offset = 0;
>> 1145 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
981 if (IS_ERR(inode)) { 1146 if (IS_ERR(inode)) {
982 err = PTR_ERR(inode); 1147 err = PTR_ERR(inode);
983 btrfs_handle_fs_error(fs_info, <<
984 goto fail_close; 1148 goto fail_close;
985 } 1149 }
986 1150
987 sb->s_root = d_make_root(inode); 1151 sb->s_root = d_make_root(inode);
988 if (!sb->s_root) { 1152 if (!sb->s_root) {
989 err = -ENOMEM; 1153 err = -ENOMEM;
990 goto fail_close; 1154 goto fail_close;
991 } 1155 }
992 1156
993 sb->s_flags |= SB_ACTIVE; !! 1157 save_mount_options(sb, data);
>> 1158 cleancache_init_fs(sb);
>> 1159 sb->s_flags |= MS_ACTIVE;
994 return 0; 1160 return 0;
995 1161
996 fail_close: 1162 fail_close:
997 close_ctree(fs_info); 1163 close_ctree(fs_info);
998 return err; 1164 return err;
999 } 1165 }
1000 1166
1001 int btrfs_sync_fs(struct super_block *sb, int 1167 int btrfs_sync_fs(struct super_block *sb, int wait)
1002 { 1168 {
1003 struct btrfs_trans_handle *trans; 1169 struct btrfs_trans_handle *trans;
1004 struct btrfs_fs_info *fs_info = btrfs 1170 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1005 struct btrfs_root *root = fs_info->tr 1171 struct btrfs_root *root = fs_info->tree_root;
1006 1172
1007 trace_btrfs_sync_fs(fs_info, wait); 1173 trace_btrfs_sync_fs(fs_info, wait);
1008 1174
1009 if (!wait) { 1175 if (!wait) {
1010 filemap_flush(fs_info->btree_ 1176 filemap_flush(fs_info->btree_inode->i_mapping);
1011 return 0; 1177 return 0;
1012 } 1178 }
1013 1179
1014 btrfs_wait_ordered_roots(fs_info, U64 !! 1180 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1015 1181
1016 trans = btrfs_attach_transaction_barr 1182 trans = btrfs_attach_transaction_barrier(root);
1017 if (IS_ERR(trans)) { 1183 if (IS_ERR(trans)) {
1018 /* no transaction, don't both 1184 /* no transaction, don't bother */
1019 if (PTR_ERR(trans) == -ENOENT 1185 if (PTR_ERR(trans) == -ENOENT) {
1020 /* 1186 /*
1021 * Exit unless we hav 1187 * Exit unless we have some pending changes
1022 * that need to go th 1188 * that need to go through commit
1023 */ 1189 */
1024 if (!test_bit(BTRFS_F !! 1190 if (fs_info->pending_changes == 0)
1025 &fs_inf <<
1026 return 0; 1191 return 0;
1027 /* 1192 /*
1028 * A non-blocking tes 1193 * A non-blocking test if the fs is frozen. We must not
1029 * start a new transa 1194 * start a new transaction here otherwise a deadlock
1030 * happens. The pendi 1195 * happens. The pending operations are delayed to the
1031 * next commit after 1196 * next commit after thawing.
1032 */ 1197 */
1033 if (sb_start_write_tr !! 1198 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1034 sb_end_write( !! 1199 __sb_end_write(sb, SB_FREEZE_WRITE);
1035 else 1200 else
1036 return 0; 1201 return 0;
1037 trans = btrfs_start_t 1202 trans = btrfs_start_transaction(root, 0);
1038 } 1203 }
1039 if (IS_ERR(trans)) 1204 if (IS_ERR(trans))
1040 return PTR_ERR(trans) 1205 return PTR_ERR(trans);
1041 } 1206 }
1042 return btrfs_commit_transaction(trans 1207 return btrfs_commit_transaction(trans);
1043 } 1208 }
1044 1209
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 1210 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1052 { 1211 {
1053 struct btrfs_fs_info *info = btrfs_sb 1212 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1054 const char *compress_type; !! 1213 char *compress_type;
1055 const char *subvol_name; <<
1056 bool printed = false; <<
1057 1214
1058 if (btrfs_test_opt(info, DEGRADED)) 1215 if (btrfs_test_opt(info, DEGRADED))
1059 seq_puts(seq, ",degraded"); 1216 seq_puts(seq, ",degraded");
1060 if (btrfs_test_opt(info, NODATASUM)) 1217 if (btrfs_test_opt(info, NODATASUM))
1061 seq_puts(seq, ",nodatasum"); 1218 seq_puts(seq, ",nodatasum");
1062 if (btrfs_test_opt(info, NODATACOW)) 1219 if (btrfs_test_opt(info, NODATACOW))
1063 seq_puts(seq, ",nodatacow"); 1220 seq_puts(seq, ",nodatacow");
1064 if (btrfs_test_opt(info, NOBARRIER)) 1221 if (btrfs_test_opt(info, NOBARRIER))
1065 seq_puts(seq, ",nobarrier"); 1222 seq_puts(seq, ",nobarrier");
1066 if (info->max_inline != BTRFS_DEFAULT 1223 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1067 seq_printf(seq, ",max_inline= 1224 seq_printf(seq, ",max_inline=%llu", info->max_inline);
>> 1225 if (info->alloc_start != 0)
>> 1226 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1068 if (info->thread_pool_size != min_t( 1227 if (info->thread_pool_size != min_t(unsigned long,
1069 1228 num_online_cpus() + 2, 8))
1070 seq_printf(seq, ",thread_pool !! 1229 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1071 if (btrfs_test_opt(info, COMPRESS)) { 1230 if (btrfs_test_opt(info, COMPRESS)) {
1072 compress_type = btrfs_compres !! 1231 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
>> 1232 compress_type = "zlib";
>> 1233 else
>> 1234 compress_type = "lzo";
1073 if (btrfs_test_opt(info, FORC 1235 if (btrfs_test_opt(info, FORCE_COMPRESS))
1074 seq_printf(seq, ",com 1236 seq_printf(seq, ",compress-force=%s", compress_type);
1075 else 1237 else
1076 seq_printf(seq, ",com 1238 seq_printf(seq, ",compress=%s", compress_type);
1077 if (info->compress_level) <<
1078 seq_printf(seq, ":%d" <<
1079 } 1239 }
1080 if (btrfs_test_opt(info, NOSSD)) 1240 if (btrfs_test_opt(info, NOSSD))
1081 seq_puts(seq, ",nossd"); 1241 seq_puts(seq, ",nossd");
1082 if (btrfs_test_opt(info, SSD_SPREAD)) 1242 if (btrfs_test_opt(info, SSD_SPREAD))
1083 seq_puts(seq, ",ssd_spread"); 1243 seq_puts(seq, ",ssd_spread");
1084 else if (btrfs_test_opt(info, SSD)) 1244 else if (btrfs_test_opt(info, SSD))
1085 seq_puts(seq, ",ssd"); 1245 seq_puts(seq, ",ssd");
1086 if (btrfs_test_opt(info, NOTREELOG)) 1246 if (btrfs_test_opt(info, NOTREELOG))
1087 seq_puts(seq, ",notreelog"); 1247 seq_puts(seq, ",notreelog");
1088 if (btrfs_test_opt(info, NOLOGREPLAY) 1248 if (btrfs_test_opt(info, NOLOGREPLAY))
1089 print_rescue_option(seq, "nol !! 1249 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 1250 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1101 seq_puts(seq, ",flushoncommit 1251 seq_puts(seq, ",flushoncommit");
1102 if (btrfs_test_opt(info, DISCARD_SYNC !! 1252 if (btrfs_test_opt(info, DISCARD))
1103 seq_puts(seq, ",discard"); 1253 seq_puts(seq, ",discard");
1104 if (btrfs_test_opt(info, DISCARD_ASYN !! 1254 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"); 1255 seq_puts(seq, ",noacl");
1108 if (btrfs_free_space_cache_v1_active( !! 1256 if (btrfs_test_opt(info, SPACE_CACHE))
1109 seq_puts(seq, ",space_cache") 1257 seq_puts(seq, ",space_cache");
1110 else if (btrfs_fs_compat_ro(info, FRE !! 1258 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1111 seq_puts(seq, ",space_cache=v 1259 seq_puts(seq, ",space_cache=v2");
1112 else 1260 else
1113 seq_puts(seq, ",nospace_cache 1261 seq_puts(seq, ",nospace_cache");
1114 if (btrfs_test_opt(info, RESCAN_UUID_ 1262 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1115 seq_puts(seq, ",rescan_uuid_t 1263 seq_puts(seq, ",rescan_uuid_tree");
1116 if (btrfs_test_opt(info, CLEAR_CACHE) 1264 if (btrfs_test_opt(info, CLEAR_CACHE))
1117 seq_puts(seq, ",clear_cache") 1265 seq_puts(seq, ",clear_cache");
1118 if (btrfs_test_opt(info, USER_SUBVOL_ 1266 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1119 seq_puts(seq, ",user_subvol_r 1267 seq_puts(seq, ",user_subvol_rm_allowed");
1120 if (btrfs_test_opt(info, ENOSPC_DEBUG 1268 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1121 seq_puts(seq, ",enospc_debug" 1269 seq_puts(seq, ",enospc_debug");
1122 if (btrfs_test_opt(info, AUTO_DEFRAG) 1270 if (btrfs_test_opt(info, AUTO_DEFRAG))
1123 seq_puts(seq, ",autodefrag"); 1271 seq_puts(seq, ",autodefrag");
>> 1272 if (btrfs_test_opt(info, INODE_MAP_CACHE))
>> 1273 seq_puts(seq, ",inode_cache");
1124 if (btrfs_test_opt(info, SKIP_BALANCE 1274 if (btrfs_test_opt(info, SKIP_BALANCE))
1125 seq_puts(seq, ",skip_balance" 1275 seq_puts(seq, ",skip_balance");
>> 1276 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
>> 1277 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
>> 1278 seq_puts(seq, ",check_int_data");
>> 1279 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
>> 1280 seq_puts(seq, ",check_int");
>> 1281 if (info->check_integrity_print_mask)
>> 1282 seq_printf(seq, ",check_int_print_mask=%d",
>> 1283 info->check_integrity_print_mask);
>> 1284 #endif
1126 if (info->metadata_ratio) 1285 if (info->metadata_ratio)
1127 seq_printf(seq, ",metadata_ra !! 1286 seq_printf(seq, ",metadata_ratio=%d",
>> 1287 info->metadata_ratio);
1128 if (btrfs_test_opt(info, PANIC_ON_FAT 1288 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1129 seq_puts(seq, ",fatal_errors= 1289 seq_puts(seq, ",fatal_errors=panic");
1130 if (info->commit_interval != BTRFS_DE 1290 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1131 seq_printf(seq, ",commit=%u", !! 1291 seq_printf(seq, ",commit=%d", info->commit_interval);
1132 #ifdef CONFIG_BTRFS_DEBUG 1292 #ifdef CONFIG_BTRFS_DEBUG
1133 if (btrfs_test_opt(info, FRAGMENT_DAT 1293 if (btrfs_test_opt(info, FRAGMENT_DATA))
1134 seq_puts(seq, ",fragment=data 1294 seq_puts(seq, ",fragment=data");
1135 if (btrfs_test_opt(info, FRAGMENT_MET 1295 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1136 seq_puts(seq, ",fragment=meta 1296 seq_puts(seq, ",fragment=metadata");
1137 #endif 1297 #endif
1138 if (btrfs_test_opt(info, REF_VERIFY)) !! 1298 seq_printf(seq, ",subvolid=%llu",
1139 seq_puts(seq, ",ref_verify"); !! 1299 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1140 seq_printf(seq, ",subvolid=%llu", btr !! 1300 seq_puts(seq, ",subvol=");
1141 subvol_name = btrfs_get_subvol_name_f !! 1301 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; 1302 return 0;
1149 } 1303 }
1150 1304
>> 1305 static int btrfs_test_super(struct super_block *s, void *data)
>> 1306 {
>> 1307 struct btrfs_fs_info *p = data;
>> 1308 struct btrfs_fs_info *fs_info = btrfs_sb(s);
>> 1309
>> 1310 return fs_info->fs_devices == p->fs_devices;
>> 1311 }
>> 1312
>> 1313 static int btrfs_set_super(struct super_block *s, void *data)
>> 1314 {
>> 1315 int err = set_anon_super(s, data);
>> 1316 if (!err)
>> 1317 s->s_fs_info = data;
>> 1318 return err;
>> 1319 }
>> 1320
1151 /* 1321 /*
1152 * subvolumes are identified by ino 256 1322 * subvolumes are identified by ino 256
1153 */ 1323 */
1154 static inline int is_subvolume_inode(struct i 1324 static inline int is_subvolume_inode(struct inode *inode)
1155 { 1325 {
1156 if (inode && inode->i_ino == BTRFS_FI 1326 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1157 return 1; 1327 return 1;
1158 return 0; 1328 return 0;
1159 } 1329 }
1160 1330
>> 1331 /*
>> 1332 * This will add subvolid=0 to the argument string while removing any subvol=
>> 1333 * and subvolid= arguments to make sure we get the top-level root for path
>> 1334 * walking to the subvol we want.
>> 1335 */
>> 1336 static char *setup_root_args(char *args)
>> 1337 {
>> 1338 char *buf, *dst, *sep;
>> 1339
>> 1340 if (!args)
>> 1341 return kstrdup("subvolid=0", GFP_NOFS);
>> 1342
>> 1343 /* The worst case is that we add ",subvolid=0" to the end. */
>> 1344 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
>> 1345 if (!buf)
>> 1346 return NULL;
>> 1347
>> 1348 while (1) {
>> 1349 sep = strchrnul(args, ',');
>> 1350 if (!strstarts(args, "subvol=") &&
>> 1351 !strstarts(args, "subvolid=")) {
>> 1352 memcpy(dst, args, sep - args);
>> 1353 dst += sep - args;
>> 1354 *dst++ = ',';
>> 1355 }
>> 1356 if (*sep)
>> 1357 args = sep + 1;
>> 1358 else
>> 1359 break;
>> 1360 }
>> 1361 strcpy(dst, "subvolid=0");
>> 1362
>> 1363 return buf;
>> 1364 }
>> 1365
1161 static struct dentry *mount_subvol(const char 1366 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1162 struct vfs !! 1367 int flags, const char *device_name,
>> 1368 char *data)
1163 { 1369 {
1164 struct dentry *root; 1370 struct dentry *root;
>> 1371 struct vfsmount *mnt = NULL;
>> 1372 char *newargs;
1165 int ret; 1373 int ret;
1166 1374
>> 1375 newargs = setup_root_args(data);
>> 1376 if (!newargs) {
>> 1377 root = ERR_PTR(-ENOMEM);
>> 1378 goto out;
>> 1379 }
>> 1380
>> 1381 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
>> 1382 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
>> 1383 if (flags & MS_RDONLY) {
>> 1384 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
>> 1385 device_name, newargs);
>> 1386 } else {
>> 1387 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
>> 1388 device_name, newargs);
>> 1389 if (IS_ERR(mnt)) {
>> 1390 root = ERR_CAST(mnt);
>> 1391 mnt = NULL;
>> 1392 goto out;
>> 1393 }
>> 1394
>> 1395 down_write(&mnt->mnt_sb->s_umount);
>> 1396 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
>> 1397 up_write(&mnt->mnt_sb->s_umount);
>> 1398 if (ret < 0) {
>> 1399 root = ERR_PTR(ret);
>> 1400 goto out;
>> 1401 }
>> 1402 }
>> 1403 }
>> 1404 if (IS_ERR(mnt)) {
>> 1405 root = ERR_CAST(mnt);
>> 1406 mnt = NULL;
>> 1407 goto out;
>> 1408 }
>> 1409
1167 if (!subvol_name) { 1410 if (!subvol_name) {
1168 if (!subvol_objectid) { 1411 if (!subvol_objectid) {
1169 ret = get_default_sub 1412 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1170 1413 &subvol_objectid);
1171 if (ret) { 1414 if (ret) {
1172 root = ERR_PT 1415 root = ERR_PTR(ret);
1173 goto out; 1416 goto out;
1174 } 1417 }
1175 } 1418 }
1176 subvol_name = btrfs_get_subvo !! 1419 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1177 btrfs !! 1420 subvol_objectid);
1178 if (IS_ERR(subvol_name)) { 1421 if (IS_ERR(subvol_name)) {
1179 root = ERR_CAST(subvo 1422 root = ERR_CAST(subvol_name);
1180 subvol_name = NULL; 1423 subvol_name = NULL;
1181 goto out; 1424 goto out;
1182 } 1425 }
1183 1426
1184 } 1427 }
1185 1428
1186 root = mount_subtree(mnt, subvol_name 1429 root = mount_subtree(mnt, subvol_name);
1187 /* mount_subtree() drops our referenc 1430 /* mount_subtree() drops our reference on the vfsmount. */
1188 mnt = NULL; 1431 mnt = NULL;
1189 1432
1190 if (!IS_ERR(root)) { 1433 if (!IS_ERR(root)) {
1191 struct super_block *s = root- 1434 struct super_block *s = root->d_sb;
1192 struct btrfs_fs_info *fs_info 1435 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1193 struct inode *root_inode = d_ 1436 struct inode *root_inode = d_inode(root);
1194 u64 root_objectid = btrfs_roo !! 1437 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1195 1438
1196 ret = 0; 1439 ret = 0;
1197 if (!is_subvolume_inode(root_ 1440 if (!is_subvolume_inode(root_inode)) {
1198 btrfs_err(fs_info, "' 1441 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1199 subvol_name); 1442 subvol_name);
1200 ret = -EINVAL; 1443 ret = -EINVAL;
1201 } 1444 }
1202 if (subvol_objectid && root_o 1445 if (subvol_objectid && root_objectid != subvol_objectid) {
1203 /* 1446 /*
1204 * This will also cat 1447 * This will also catch a race condition where a
1205 * subvolume which wa 1448 * subvolume which was passed by ID is renamed and
1206 * another subvolume 1449 * another subvolume is renamed over the old location.
1207 */ 1450 */
1208 btrfs_err(fs_info, 1451 btrfs_err(fs_info,
1209 "subvol '%s 1452 "subvol '%s' does not match subvolid %llu",
1210 subvol_name 1453 subvol_name, subvol_objectid);
1211 ret = -EINVAL; 1454 ret = -EINVAL;
1212 } 1455 }
1213 if (ret) { 1456 if (ret) {
1214 dput(root); 1457 dput(root);
1215 root = ERR_PTR(ret); 1458 root = ERR_PTR(ret);
1216 deactivate_locked_sup 1459 deactivate_locked_super(s);
1217 } 1460 }
1218 } 1461 }
1219 1462
1220 out: 1463 out:
1221 mntput(mnt); 1464 mntput(mnt);
>> 1465 kfree(newargs);
1222 kfree(subvol_name); 1466 kfree(subvol_name);
1223 return root; 1467 return root;
1224 } 1468 }
1225 1469
>> 1470 static int parse_security_options(char *orig_opts,
>> 1471 struct security_mnt_opts *sec_opts)
>> 1472 {
>> 1473 char *secdata = NULL;
>> 1474 int ret = 0;
>> 1475
>> 1476 secdata = alloc_secdata();
>> 1477 if (!secdata)
>> 1478 return -ENOMEM;
>> 1479 ret = security_sb_copy_data(orig_opts, secdata);
>> 1480 if (ret) {
>> 1481 free_secdata(secdata);
>> 1482 return ret;
>> 1483 }
>> 1484 ret = security_sb_parse_opts_str(secdata, sec_opts);
>> 1485 free_secdata(secdata);
>> 1486 return ret;
>> 1487 }
>> 1488
>> 1489 static int setup_security_options(struct btrfs_fs_info *fs_info,
>> 1490 struct super_block *sb,
>> 1491 struct security_mnt_opts *sec_opts)
>> 1492 {
>> 1493 int ret = 0;
>> 1494
>> 1495 /*
>> 1496 * Call security_sb_set_mnt_opts() to check whether new sec_opts
>> 1497 * is valid.
>> 1498 */
>> 1499 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
>> 1500 if (ret)
>> 1501 return ret;
>> 1502
>> 1503 #ifdef CONFIG_SECURITY
>> 1504 if (!fs_info->security_opts.num_mnt_opts) {
>> 1505 /* first time security setup, copy sec_opts to fs_info */
>> 1506 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
>> 1507 } else {
>> 1508 /*
>> 1509 * Since SELinux (the only one supporting security_mnt_opts)
>> 1510 * does NOT support changing context during remount/mount of
>> 1511 * the same sb, this must be the same or part of the same
>> 1512 * security options, just free it.
>> 1513 */
>> 1514 security_free_mnt_opts(sec_opts);
>> 1515 }
>> 1516 #endif
>> 1517 return ret;
>> 1518 }
>> 1519
>> 1520 /*
>> 1521 * Find a superblock for the given device / mount point.
>> 1522 *
>> 1523 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
>> 1524 * for multiple device setup. Make sure to keep it in sync.
>> 1525 */
>> 1526 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
>> 1527 const char *device_name, void *data)
>> 1528 {
>> 1529 struct block_device *bdev = NULL;
>> 1530 struct super_block *s;
>> 1531 struct btrfs_fs_devices *fs_devices = NULL;
>> 1532 struct btrfs_fs_info *fs_info = NULL;
>> 1533 struct security_mnt_opts new_sec_opts;
>> 1534 fmode_t mode = FMODE_READ;
>> 1535 char *subvol_name = NULL;
>> 1536 u64 subvol_objectid = 0;
>> 1537 int error = 0;
>> 1538
>> 1539 if (!(flags & MS_RDONLY))
>> 1540 mode |= FMODE_WRITE;
>> 1541
>> 1542 error = btrfs_parse_early_options(data, mode, fs_type,
>> 1543 &subvol_name, &subvol_objectid,
>> 1544 &fs_devices);
>> 1545 if (error) {
>> 1546 kfree(subvol_name);
>> 1547 return ERR_PTR(error);
>> 1548 }
>> 1549
>> 1550 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
>> 1551 /* mount_subvol() will free subvol_name. */
>> 1552 return mount_subvol(subvol_name, subvol_objectid, flags,
>> 1553 device_name, data);
>> 1554 }
>> 1555
>> 1556 security_init_mnt_opts(&new_sec_opts);
>> 1557 if (data) {
>> 1558 error = parse_security_options(data, &new_sec_opts);
>> 1559 if (error)
>> 1560 return ERR_PTR(error);
>> 1561 }
>> 1562
>> 1563 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
>> 1564 if (error)
>> 1565 goto error_sec_opts;
>> 1566
>> 1567 /*
>> 1568 * Setup a dummy root and fs_info for test/set super. This is because
>> 1569 * we don't actually fill this stuff out until open_ctree, but we need
>> 1570 * it for searching for existing supers, so this lets us do that and
>> 1571 * then open_ctree will properly initialize everything later.
>> 1572 */
>> 1573 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
>> 1574 if (!fs_info) {
>> 1575 error = -ENOMEM;
>> 1576 goto error_sec_opts;
>> 1577 }
>> 1578
>> 1579 fs_info->fs_devices = fs_devices;
>> 1580
>> 1581 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
>> 1582 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
>> 1583 security_init_mnt_opts(&fs_info->security_opts);
>> 1584 if (!fs_info->super_copy || !fs_info->super_for_commit) {
>> 1585 error = -ENOMEM;
>> 1586 goto error_fs_info;
>> 1587 }
>> 1588
>> 1589 error = btrfs_open_devices(fs_devices, mode, fs_type);
>> 1590 if (error)
>> 1591 goto error_fs_info;
>> 1592
>> 1593 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
>> 1594 error = -EACCES;
>> 1595 goto error_close_devices;
>> 1596 }
>> 1597
>> 1598 bdev = fs_devices->latest_bdev;
>> 1599 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
>> 1600 fs_info);
>> 1601 if (IS_ERR(s)) {
>> 1602 error = PTR_ERR(s);
>> 1603 goto error_close_devices;
>> 1604 }
>> 1605
>> 1606 if (s->s_root) {
>> 1607 btrfs_close_devices(fs_devices);
>> 1608 free_fs_info(fs_info);
>> 1609 if ((flags ^ s->s_flags) & MS_RDONLY)
>> 1610 error = -EBUSY;
>> 1611 } else {
>> 1612 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
>> 1613 btrfs_sb(s)->bdev_holder = fs_type;
>> 1614 error = btrfs_fill_super(s, fs_devices, data,
>> 1615 flags & MS_SILENT ? 1 : 0);
>> 1616 }
>> 1617 if (error) {
>> 1618 deactivate_locked_super(s);
>> 1619 goto error_sec_opts;
>> 1620 }
>> 1621
>> 1622 fs_info = btrfs_sb(s);
>> 1623 error = setup_security_options(fs_info, s, &new_sec_opts);
>> 1624 if (error) {
>> 1625 deactivate_locked_super(s);
>> 1626 goto error_sec_opts;
>> 1627 }
>> 1628
>> 1629 return dget(s->s_root);
>> 1630
>> 1631 error_close_devices:
>> 1632 btrfs_close_devices(fs_devices);
>> 1633 error_fs_info:
>> 1634 free_fs_info(fs_info);
>> 1635 error_sec_opts:
>> 1636 security_free_mnt_opts(&new_sec_opts);
>> 1637 return ERR_PTR(error);
>> 1638 }
>> 1639
1226 static void btrfs_resize_thread_pool(struct b 1640 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1227 u32 new_ !! 1641 int new_pool_size, int old_pool_size)
1228 { 1642 {
1229 if (new_pool_size == old_pool_size) 1643 if (new_pool_size == old_pool_size)
1230 return; 1644 return;
1231 1645
1232 fs_info->thread_pool_size = new_pool_ 1646 fs_info->thread_pool_size = new_pool_size;
1233 1647
1234 btrfs_info(fs_info, "resize thread po 1648 btrfs_info(fs_info, "resize thread pool %d -> %d",
1235 old_pool_size, new_pool_size); 1649 old_pool_size, new_pool_size);
1236 1650
1237 btrfs_workqueue_set_max(fs_info->work 1651 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1238 btrfs_workqueue_set_max(fs_info->dela 1652 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
>> 1653 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1239 btrfs_workqueue_set_max(fs_info->cach 1654 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1240 workqueue_set_max_active(fs_info->end !! 1655 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1241 workqueue_set_max_active(fs_info->end !! 1656 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
>> 1657 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
>> 1658 new_pool_size);
1242 btrfs_workqueue_set_max(fs_info->endi 1659 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1243 btrfs_workqueue_set_max(fs_info->endi 1660 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1244 btrfs_workqueue_set_max(fs_info->dela 1661 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
>> 1662 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
>> 1663 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
>> 1664 new_pool_size);
>> 1665 }
>> 1666
>> 1667 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
>> 1668 {
>> 1669 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1245 } 1670 }
1246 1671
1247 static inline void btrfs_remount_begin(struct 1672 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1248 unsign !! 1673 unsigned long old_opts, int flags)
1249 { 1674 {
1250 if (btrfs_raw_test_opt(old_opts, AUTO 1675 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1251 (!btrfs_raw_test_opt(fs_info->mou 1676 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1252 (flags & SB_RDONLY))) { !! 1677 (flags & MS_RDONLY))) {
1253 /* wait for any defraggers to 1678 /* wait for any defraggers to finish */
1254 wait_event(fs_info->transacti 1679 wait_event(fs_info->transaction_wait,
1255 (atomic_read(&fs_i 1680 (atomic_read(&fs_info->defrag_running) == 0));
1256 if (flags & SB_RDONLY) !! 1681 if (flags & MS_RDONLY)
1257 sync_filesystem(fs_in 1682 sync_filesystem(fs_info->sb);
1258 } 1683 }
1259 } 1684 }
1260 1685
1261 static inline void btrfs_remount_cleanup(stru 1686 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1262 unsi !! 1687 unsigned long old_opts)
1263 { 1688 {
1264 const bool cache_opt = btrfs_test_opt <<
1265 <<
1266 /* 1689 /*
1267 * We need to cleanup all defragable 1690 * We need to cleanup all defragable inodes if the autodefragment is
1268 * close or the filesystem is read on 1691 * close or the filesystem is read only.
1269 */ 1692 */
1270 if (btrfs_raw_test_opt(old_opts, AUTO 1693 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1271 (!btrfs_raw_test_opt(fs_info->mou !! 1694 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
>> 1695 (fs_info->sb->s_flags & MS_RDONLY))) {
1272 btrfs_cleanup_defrag_inodes(f 1696 btrfs_cleanup_defrag_inodes(fs_info);
1273 } 1697 }
1274 1698
1275 /* If we toggled discard async */ !! 1699 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 } 1700 }
1287 1701
1288 static int btrfs_remount_rw(struct btrfs_fs_i !! 1702 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1289 { 1703 {
>> 1704 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 1705 struct btrfs_root *root = fs_info->tree_root;
>> 1706 unsigned old_flags = sb->s_flags;
>> 1707 unsigned long old_opts = fs_info->mount_opt;
>> 1708 unsigned long old_compress_type = fs_info->compress_type;
>> 1709 u64 old_max_inline = fs_info->max_inline;
>> 1710 u64 old_alloc_start = fs_info->alloc_start;
>> 1711 int old_thread_pool_size = fs_info->thread_pool_size;
>> 1712 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1290 int ret; 1713 int ret;
1291 1714
1292 if (BTRFS_FS_ERROR(fs_info)) { !! 1715 sync_filesystem(sb);
1293 btrfs_err(fs_info, !! 1716 btrfs_remount_prepare(fs_info);
1294 "remounting read-wr <<
1295 return -EINVAL; <<
1296 } <<
1297 1717
1298 if (fs_info->fs_devices->rw_devices = !! 1718 if (data) {
1299 return -EACCES; !! 1719 struct security_mnt_opts new_sec_opts;
1300 1720
1301 if (!btrfs_check_rw_degradable(fs_inf !! 1721 security_init_mnt_opts(&new_sec_opts);
1302 btrfs_warn(fs_info, !! 1722 ret = parse_security_options(data, &new_sec_opts);
1303 "too many missing !! 1723 if (ret)
1304 return -EACCES; !! 1724 goto restore;
>> 1725 ret = setup_security_options(fs_info, sb,
>> 1726 &new_sec_opts);
>> 1727 if (ret) {
>> 1728 security_free_mnt_opts(&new_sec_opts);
>> 1729 goto restore;
>> 1730 }
1305 } 1731 }
1306 1732
1307 if (btrfs_super_log_root(fs_info->sup !! 1733 ret = btrfs_parse_options(fs_info, data, *flags);
1308 btrfs_warn(fs_info, !! 1734 if (ret) {
1309 "mount required to !! 1735 ret = -EINVAL;
1310 return -EINVAL; !! 1736 goto restore;
1311 } 1737 }
1312 1738
1313 /* !! 1739 btrfs_remount_begin(fs_info, old_opts, *flags);
1314 * NOTE: when remounting with a chang !! 1740 btrfs_resize_thread_pool(fs_info,
1315 * anywhere above this point, as we a !! 1741 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 1742
1346 /* Wait for the uuid_scan task to fin !! 1743 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1347 down(&fs_info->uuid_tree_rescan_sem); !! 1744 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 1745
1386 /* !! 1746 if (*flags & MS_RDONLY) {
1387 * Pause the qgroup rescan worker if !! 1747 /*
1388 * be still running after we are in R !! 1748 * this also happens on 'umount -rf' or on shutdown, when
1389 * we unmount, it might have left a t !! 1749 * the filesystem is busy.
1390 * the transaction and/or crash. !! 1750 */
1391 */ !! 1751 cancel_work_sync(&fs_info->async_reclaim_work);
1392 btrfs_qgroup_wait_for_completion(fs_i <<
1393 1752
1394 return btrfs_commit_super(fs_info); !! 1753 /* wait for the uuid_scan task to finish */
1395 } !! 1754 down(&fs_info->uuid_tree_rescan_sem);
>> 1755 /* avoid complains from lockdep et al. */
>> 1756 up(&fs_info->uuid_tree_rescan_sem);
1396 1757
1397 static void btrfs_ctx_to_info(struct btrfs_fs !! 1758 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 1759
1484 if (info->max_inline != BTRFS_DEFAULT !! 1760 /*
1485 btrfs_info(info, "max_inline !! 1761 * Setting MS_RDONLY will put the cleaner thread to
1486 } !! 1762 * sleep at the next loop if it's already active.
>> 1763 * If it's already asleep, we'll leave unused block
>> 1764 * groups on disk until we're mounted read-write again
>> 1765 * unless we clean them up here.
>> 1766 */
>> 1767 btrfs_delete_unused_bgs(fs_info);
1487 1768
1488 static int btrfs_reconfigure(struct fs_contex !! 1769 btrfs_dev_replace_suspend_for_unmount(fs_info);
1489 { !! 1770 btrfs_scrub_cancel(fs_info);
1490 struct super_block *sb = fc->root->d_ !! 1771 btrfs_pause_balance(fs_info);
1491 struct btrfs_fs_info *fs_info = btrfs !! 1772
1492 struct btrfs_fs_context *ctx = fc->fs !! 1773 ret = btrfs_commit_super(fs_info);
1493 struct btrfs_fs_context old_ctx; !! 1774 if (ret)
1494 int ret = 0; !! 1775 goto restore;
1495 bool mount_reconfigure = (fc->s_fs_in !! 1776 } else {
>> 1777 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
>> 1778 btrfs_err(fs_info,
>> 1779 "Remounting read-write after error is not allowed");
>> 1780 ret = -EINVAL;
>> 1781 goto restore;
>> 1782 }
>> 1783 if (fs_info->fs_devices->rw_devices == 0) {
>> 1784 ret = -EACCES;
>> 1785 goto restore;
>> 1786 }
1496 1787
1497 btrfs_info_to_ctx(fs_info, &old_ctx); !! 1788 if (fs_info->fs_devices->missing_devices >
>> 1789 fs_info->num_tolerated_disk_barrier_failures &&
>> 1790 !(*flags & MS_RDONLY)) {
>> 1791 btrfs_warn(fs_info,
>> 1792 "too many missing devices, writeable remount is not allowed");
>> 1793 ret = -EACCES;
>> 1794 goto restore;
>> 1795 }
1498 1796
1499 /* !! 1797 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1500 * This is our "bind mount" trick, we !! 1798 ret = -EINVAL;
1501 * anything other than mount a differ !! 1799 goto restore;
1502 * all of the mount options should be !! 1800 }
1503 */ <<
1504 if (mount_reconfigure) <<
1505 ctx->mount_opt = old_ctx.moun <<
1506 1801
1507 sync_filesystem(sb); !! 1802 ret = btrfs_cleanup_fs_roots(fs_info);
1508 set_bit(BTRFS_FS_STATE_REMOUNTING, &f !! 1803 if (ret)
>> 1804 goto restore;
1509 1805
1510 if (!btrfs_check_options(fs_info, &ct !! 1806 /* recover relocation */
1511 return -EINVAL; !! 1807 mutex_lock(&fs_info->cleaner_mutex);
>> 1808 ret = btrfs_recover_relocation(root);
>> 1809 mutex_unlock(&fs_info->cleaner_mutex);
>> 1810 if (ret)
>> 1811 goto restore;
1512 1812
1513 ret = btrfs_check_features(fs_info, ! !! 1813 ret = btrfs_resume_balance_async(fs_info);
1514 if (ret < 0) !! 1814 if (ret)
1515 return ret; !! 1815 goto restore;
1516 1816
1517 btrfs_ctx_to_info(fs_info, ctx); !! 1817 ret = btrfs_resume_dev_replace_async(fs_info);
1518 btrfs_remount_begin(fs_info, old_ctx. !! 1818 if (ret) {
1519 btrfs_resize_thread_pool(fs_info, fs_ !! 1819 btrfs_warn(fs_info, "failed to resume dev_replace");
1520 old_ctx.thre !! 1820 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 } 1821 }
1536 } <<
1537 1822
1538 ret = 0; !! 1823 if (!fs_info->uuid_root) {
1539 if (!sb_rdonly(sb) && (fc->sb_flags & !! 1824 btrfs_info(fs_info, "creating UUID tree");
1540 ret = btrfs_remount_ro(fs_inf !! 1825 ret = btrfs_create_uuid_tree(fs_info);
1541 else if (sb_rdonly(sb) && !(fc->sb_fl !! 1826 if (ret) {
1542 ret = btrfs_remount_rw(fs_inf !! 1827 btrfs_warn(fs_info,
1543 if (ret) !! 1828 "failed to create the UUID tree %d",
1544 goto restore; !! 1829 ret);
1545 !! 1830 goto restore;
1546 /* !! 1831 }
1547 * If we set the mask during the para !! 1832 }
1548 * remount. Here we can set the mask !! 1833 sb->s_flags &= ~MS_RDONLY;
1549 * appropriately. <<
1550 */ <<
1551 if ((fc->sb_flags & SB_POSIXACL) != ( <<
1552 fc->sb_flags_mask |= SB_POSIX <<
1553 1834
1554 btrfs_emit_options(fs_info, &old_ctx) !! 1835 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
>> 1836 }
>> 1837 out:
1555 wake_up_process(fs_info->transaction_ 1838 wake_up_process(fs_info->transaction_kthread);
1556 btrfs_remount_cleanup(fs_info, old_ct !! 1839 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; 1840 return 0;
>> 1841
1561 restore: 1842 restore:
1562 btrfs_ctx_to_info(fs_info, &old_ctx); !! 1843 /* We've hit an error - don't reset MS_RDONLY */
1563 btrfs_remount_cleanup(fs_info, old_ct !! 1844 if (sb->s_flags & MS_RDONLY)
1564 clear_bit(BTRFS_FS_STATE_REMOUNTING, !! 1845 old_flags |= MS_RDONLY;
>> 1846 sb->s_flags = old_flags;
>> 1847 fs_info->mount_opt = old_opts;
>> 1848 fs_info->compress_type = old_compress_type;
>> 1849 fs_info->max_inline = old_max_inline;
>> 1850 mutex_lock(&fs_info->chunk_mutex);
>> 1851 fs_info->alloc_start = old_alloc_start;
>> 1852 mutex_unlock(&fs_info->chunk_mutex);
>> 1853 btrfs_resize_thread_pool(fs_info,
>> 1854 old_thread_pool_size, fs_info->thread_pool_size);
>> 1855 fs_info->metadata_ratio = old_metadata_ratio;
>> 1856 btrfs_remount_cleanup(fs_info, old_opts);
1565 return ret; 1857 return ret;
1566 } 1858 }
1567 1859
1568 /* Used to sort the devices by max_avail(desc 1860 /* Used to sort the devices by max_avail(descending sort) */
1569 static int btrfs_cmp_device_free_bytes(const !! 1861 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
>> 1862 const void *dev_info2)
1570 { 1863 {
1571 const struct btrfs_device_info *dev_i !! 1864 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1572 const struct btrfs_device_info *dev_i !! 1865 ((struct btrfs_device_info *)dev_info2)->max_avail)
1573 <<
1574 if (dev_info1->max_avail > dev_info2- <<
1575 return -1; 1866 return -1;
1576 else if (dev_info1->max_avail < dev_i !! 1867 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
>> 1868 ((struct btrfs_device_info *)dev_info2)->max_avail)
1577 return 1; 1869 return 1;
>> 1870 else
1578 return 0; 1871 return 0;
1579 } 1872 }
1580 1873
1581 /* 1874 /*
1582 * sort the devices by max_avail, in which ma 1875 * sort the devices by max_avail, in which max free extent size of each device
1583 * is stored.(Descending Sort) 1876 * is stored.(Descending Sort)
1584 */ 1877 */
1585 static inline void btrfs_descending_sort_devi 1878 static inline void btrfs_descending_sort_devices(
1586 struc 1879 struct btrfs_device_info *devices,
1587 size_ 1880 size_t nr_devices)
1588 { 1881 {
1589 sort(devices, nr_devices, sizeof(stru 1882 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1590 btrfs_cmp_device_free_bytes, NUL 1883 btrfs_cmp_device_free_bytes, NULL);
1591 } 1884 }
1592 1885
1593 /* 1886 /*
1594 * The helper to calc the free space on the d 1887 * The helper to calc the free space on the devices that can be used to store
1595 * file data. 1888 * file data.
1596 */ 1889 */
1597 static inline int btrfs_calc_avail_data_space !! 1890 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1598 !! 1891 u64 *free_bytes)
1599 { 1892 {
>> 1893 struct btrfs_root *root = fs_info->tree_root;
1600 struct btrfs_device_info *devices_inf 1894 struct btrfs_device_info *devices_info;
1601 struct btrfs_fs_devices *fs_devices = 1895 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1602 struct btrfs_device *device; 1896 struct btrfs_device *device;
>> 1897 u64 skip_space;
1603 u64 type; 1898 u64 type;
1604 u64 avail_space; 1899 u64 avail_space;
>> 1900 u64 used_space;
1605 u64 min_stripe_size; 1901 u64 min_stripe_size;
1606 int num_stripes = 1; !! 1902 int min_stripes = 1, num_stripes = 1;
1607 int i = 0, nr_devices; 1903 int i = 0, nr_devices;
1608 const struct btrfs_raid_attr *rattr; !! 1904 int ret;
1609 1905
1610 /* 1906 /*
1611 * We aren't under the device list lo 1907 * We aren't under the device list lock, so this is racy-ish, but good
1612 * enough for our purposes. 1908 * enough for our purposes.
1613 */ 1909 */
1614 nr_devices = fs_info->fs_devices->ope 1910 nr_devices = fs_info->fs_devices->open_devices;
1615 if (!nr_devices) { 1911 if (!nr_devices) {
1616 smp_mb(); 1912 smp_mb();
1617 nr_devices = fs_info->fs_devi 1913 nr_devices = fs_info->fs_devices->open_devices;
1618 ASSERT(nr_devices); 1914 ASSERT(nr_devices);
1619 if (!nr_devices) { 1915 if (!nr_devices) {
1620 *free_bytes = 0; 1916 *free_bytes = 0;
1621 return 0; 1917 return 0;
1622 } 1918 }
1623 } 1919 }
1624 1920
1625 devices_info = kmalloc_array(nr_devic 1921 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1626 GFP_KERNEL); !! 1922 GFP_NOFS);
1627 if (!devices_info) 1923 if (!devices_info)
1628 return -ENOMEM; 1924 return -ENOMEM;
1629 1925
1630 /* calc min stripe number for data sp 1926 /* calc min stripe number for data space allocation */
1631 type = btrfs_data_alloc_profile(fs_in !! 1927 type = btrfs_get_alloc_profile(root, 1);
1632 rattr = &btrfs_raid_array[btrfs_bg_fl !! 1928 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1633 !! 1929 min_stripes = 2;
1634 if (type & BTRFS_BLOCK_GROUP_RAID0) <<
1635 num_stripes = nr_devices; 1930 num_stripes = nr_devices;
1636 else if (type & BTRFS_BLOCK_GROUP_RAI !! 1931 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1637 num_stripes = rattr->ncopies; !! 1932 min_stripes = 2;
1638 else if (type & BTRFS_BLOCK_GROUP_RAI !! 1933 num_stripes = 2;
>> 1934 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
>> 1935 min_stripes = 4;
1639 num_stripes = 4; 1936 num_stripes = 4;
>> 1937 }
1640 1938
1641 /* Adjust for more than 1 stripe per !! 1939 if (type & BTRFS_BLOCK_GROUP_DUP)
1642 min_stripe_size = rattr->dev_stripes !! 1940 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
>> 1941 else
>> 1942 min_stripe_size = BTRFS_STRIPE_LEN;
1643 1943
>> 1944 if (fs_info->alloc_start)
>> 1945 mutex_lock(&fs_devices->device_list_mutex);
1644 rcu_read_lock(); 1946 rcu_read_lock();
1645 list_for_each_entry_rcu(device, &fs_d 1947 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1646 if (!test_bit(BTRFS_DEV_STATE !! 1948 if (!device->in_fs_metadata || !device->bdev ||
1647 !! 1949 device->is_tgtdev_for_dev_replace)
1648 !device->bdev || <<
1649 test_bit(BTRFS_DEV_STATE_ <<
1650 continue; 1950 continue;
1651 1951
1652 if (i >= nr_devices) 1952 if (i >= nr_devices)
1653 break; 1953 break;
1654 1954
1655 avail_space = device->total_b 1955 avail_space = device->total_bytes - device->bytes_used;
1656 1956
1657 /* align with stripe_len */ 1957 /* align with stripe_len */
1658 avail_space = rounddown(avail !! 1958 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
>> 1959 avail_space *= BTRFS_STRIPE_LEN;
1659 1960
1660 /* 1961 /*
1661 * Ensure we have at least mi !! 1962 * In order to avoid overwriting the superblock on the drive,
1662 * reserved space on the devi !! 1963 * btrfs starts at an offset of at least 1MB when doing chunk
>> 1964 * allocation.
1663 */ 1965 */
1664 if (avail_space <= BTRFS_DEVI !! 1966 skip_space = SZ_1M;
1665 continue; !! 1967
>> 1968 /* user can set the offset in fs_info->alloc_start. */
>> 1969 if (fs_info->alloc_start &&
>> 1970 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
>> 1971 device->total_bytes) {
>> 1972 rcu_read_unlock();
>> 1973 skip_space = max(fs_info->alloc_start, skip_space);
1666 1974
1667 avail_space -= BTRFS_DEVICE_R !! 1975 /*
>> 1976 * btrfs can not use the free space in
>> 1977 * [0, skip_space - 1], we must subtract it from the
>> 1978 * total. In order to implement it, we account the used
>> 1979 * space in this range first.
>> 1980 */
>> 1981 ret = btrfs_account_dev_extents_size(device, 0,
>> 1982 skip_space - 1,
>> 1983 &used_space);
>> 1984 if (ret) {
>> 1985 kfree(devices_info);
>> 1986 mutex_unlock(&fs_devices->device_list_mutex);
>> 1987 return ret;
>> 1988 }
>> 1989
>> 1990 rcu_read_lock();
>> 1991
>> 1992 /* calc the free space in [0, skip_space - 1] */
>> 1993 skip_space -= used_space;
>> 1994 }
>> 1995
>> 1996 /*
>> 1997 * we can use the free space in [0, skip_space - 1], subtract
>> 1998 * it from the total.
>> 1999 */
>> 2000 if (avail_space && avail_space >= skip_space)
>> 2001 avail_space -= skip_space;
>> 2002 else
>> 2003 avail_space = 0;
>> 2004
>> 2005 if (avail_space < min_stripe_size)
>> 2006 continue;
1668 2007
1669 devices_info[i].dev = device; 2008 devices_info[i].dev = device;
1670 devices_info[i].max_avail = a 2009 devices_info[i].max_avail = avail_space;
1671 2010
1672 i++; 2011 i++;
1673 } 2012 }
1674 rcu_read_unlock(); 2013 rcu_read_unlock();
>> 2014 if (fs_info->alloc_start)
>> 2015 mutex_unlock(&fs_devices->device_list_mutex);
1675 2016
1676 nr_devices = i; 2017 nr_devices = i;
1677 2018
1678 btrfs_descending_sort_devices(devices 2019 btrfs_descending_sort_devices(devices_info, nr_devices);
1679 2020
1680 i = nr_devices - 1; 2021 i = nr_devices - 1;
1681 avail_space = 0; 2022 avail_space = 0;
1682 while (nr_devices >= rattr->devs_min) !! 2023 while (nr_devices >= min_stripes) {
1683 num_stripes = min(num_stripes !! 2024 if (num_stripes > nr_devices)
>> 2025 num_stripes = nr_devices;
1684 2026
1685 if (devices_info[i].max_avail 2027 if (devices_info[i].max_avail >= min_stripe_size) {
1686 int j; 2028 int j;
1687 u64 alloc_size; 2029 u64 alloc_size;
1688 2030
1689 avail_space += device 2031 avail_space += devices_info[i].max_avail * num_stripes;
1690 alloc_size = devices_ 2032 alloc_size = devices_info[i].max_avail;
1691 for (j = i + 1 - num_ 2033 for (j = i + 1 - num_stripes; j <= i; j++)
1692 devices_info[ 2034 devices_info[j].max_avail -= alloc_size;
1693 } 2035 }
1694 i--; 2036 i--;
1695 nr_devices--; 2037 nr_devices--;
1696 } 2038 }
1697 2039
1698 kfree(devices_info); 2040 kfree(devices_info);
1699 *free_bytes = avail_space; 2041 *free_bytes = avail_space;
1700 return 0; 2042 return 0;
1701 } 2043 }
1702 2044
1703 /* 2045 /*
1704 * Calculate numbers for 'df', pessimistic in 2046 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1705 * 2047 *
1706 * If there's a redundant raid level at DATA 2048 * If there's a redundant raid level at DATA block groups, use the respective
1707 * multiplier to scale the sizes. 2049 * multiplier to scale the sizes.
1708 * 2050 *
1709 * Unused device space usage is based on simu 2051 * Unused device space usage is based on simulating the chunk allocator
1710 * algorithm that respects the device sizes a !! 2052 * algorithm that respects the device sizes, order of allocations and the
1711 * a close approximation of the actual use bu !! 2053 * 'alloc_start' value, this is a close approximation of the actual use but
1712 * change the result (like a new metadata chu !! 2054 * there are other factors that may change the result (like a new metadata
>> 2055 * chunk).
1713 * 2056 *
1714 * If metadata is exhausted, f_bavail will be 2057 * If metadata is exhausted, f_bavail will be 0.
1715 */ 2058 */
1716 static int btrfs_statfs(struct dentry *dentry 2059 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1717 { 2060 {
1718 struct btrfs_fs_info *fs_info = btrfs 2061 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1719 struct btrfs_super_block *disk_super 2062 struct btrfs_super_block *disk_super = fs_info->super_copy;
>> 2063 struct list_head *head = &fs_info->space_info;
1720 struct btrfs_space_info *found; 2064 struct btrfs_space_info *found;
1721 u64 total_used = 0; 2065 u64 total_used = 0;
1722 u64 total_free_data = 0; 2066 u64 total_free_data = 0;
1723 u64 total_free_meta = 0; 2067 u64 total_free_meta = 0;
1724 u32 bits = fs_info->sectorsize_bits; !! 2068 int bits = dentry->d_sb->s_blocksize_bits;
1725 __be32 *fsid = (__be32 *)fs_info->fs_ !! 2069 __be32 *fsid = (__be32 *)fs_info->fsid;
1726 unsigned factor = 1; 2070 unsigned factor = 1;
1727 struct btrfs_block_rsv *block_rsv = & 2071 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1728 int ret; 2072 int ret;
1729 u64 thresh = 0; 2073 u64 thresh = 0;
1730 int mixed = 0; 2074 int mixed = 0;
1731 2075
1732 list_for_each_entry(found, &fs_info-> !! 2076 rcu_read_lock();
>> 2077 list_for_each_entry_rcu(found, head, list) {
1733 if (found->flags & BTRFS_BLOC 2078 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1734 int i; 2079 int i;
1735 2080
1736 total_free_data += fo 2081 total_free_data += found->disk_total - found->disk_used;
1737 total_free_data -= 2082 total_free_data -=
1738 btrfs_account 2083 btrfs_account_ro_block_groups_free_space(found);
1739 2084
1740 for (i = 0; i < BTRFS 2085 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1741 if (!list_emp !! 2086 if (!list_empty(&found->block_groups[i])) {
1742 facto !! 2087 switch (i) {
1743 !! 2088 case BTRFS_RAID_DUP:
>> 2089 case BTRFS_RAID_RAID1:
>> 2090 case BTRFS_RAID_RAID10:
>> 2091 factor = 2;
>> 2092 }
>> 2093 }
1744 } 2094 }
1745 } 2095 }
1746 2096
1747 /* 2097 /*
1748 * Metadata in mixed block gr !! 2098 * Metadata in mixed block goup profiles are accounted in data
1749 */ 2099 */
1750 if (!mixed && found->flags & 2100 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1751 if (found->flags & BT 2101 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1752 mixed = 1; 2102 mixed = 1;
1753 else 2103 else
1754 total_free_me 2104 total_free_meta += found->disk_total -
1755 found 2105 found->disk_used;
1756 } 2106 }
1757 2107
1758 total_used += found->disk_use 2108 total_used += found->disk_used;
1759 } 2109 }
1760 2110
>> 2111 rcu_read_unlock();
>> 2112
1761 buf->f_blocks = div_u64(btrfs_super_t 2113 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1762 buf->f_blocks >>= bits; 2114 buf->f_blocks >>= bits;
1763 buf->f_bfree = buf->f_blocks - (div_u 2115 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1764 2116
1765 /* Account global block reserve as us 2117 /* Account global block reserve as used, it's in logical size already */
1766 spin_lock(&block_rsv->lock); 2118 spin_lock(&block_rsv->lock);
1767 /* Mixed block groups accounting is n 2119 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1768 if (buf->f_bfree >= block_rsv->size > 2120 if (buf->f_bfree >= block_rsv->size >> bits)
1769 buf->f_bfree -= block_rsv->si 2121 buf->f_bfree -= block_rsv->size >> bits;
1770 else 2122 else
1771 buf->f_bfree = 0; 2123 buf->f_bfree = 0;
1772 spin_unlock(&block_rsv->lock); 2124 spin_unlock(&block_rsv->lock);
1773 2125
1774 buf->f_bavail = div_u64(total_free_da 2126 buf->f_bavail = div_u64(total_free_data, factor);
1775 ret = btrfs_calc_avail_data_space(fs_ 2127 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1776 if (ret) 2128 if (ret)
1777 return ret; 2129 return ret;
1778 buf->f_bavail += div_u64(total_free_d 2130 buf->f_bavail += div_u64(total_free_data, factor);
1779 buf->f_bavail = buf->f_bavail >> bits 2131 buf->f_bavail = buf->f_bavail >> bits;
1780 2132
1781 /* 2133 /*
1782 * We calculate the remaining metadat 2134 * We calculate the remaining metadata space minus global reserve. If
1783 * this is (supposedly) smaller than 2135 * this is (supposedly) smaller than zero, there's no space. But this
1784 * does not hold in practice, the exh 2136 * does not hold in practice, the exhausted state happens where's still
1785 * some positive delta. So we apply s 2137 * some positive delta. So we apply some guesswork and compare the
1786 * delta to a 4M threshold. (Practic 2138 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1787 * 2139 *
1788 * We probably cannot calculate the e 2140 * We probably cannot calculate the exact threshold value because this
1789 * depends on the internal reservatio 2141 * depends on the internal reservations requested by various
1790 * operations, so some operations tha 2142 * operations, so some operations that consume a few metadata will
1791 * succeed even if the Avail is zero. 2143 * succeed even if the Avail is zero. But this is better than the other
1792 * way around. 2144 * way around.
1793 */ 2145 */
1794 thresh = SZ_4M; !! 2146 thresh = 4 * 1024 * 1024;
1795 2147
1796 /* !! 2148 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; 2149 buf->f_bavail = 0;
1806 2150
1807 buf->f_type = BTRFS_SUPER_MAGIC; 2151 buf->f_type = BTRFS_SUPER_MAGIC;
1808 buf->f_bsize = fs_info->sectorsize; !! 2152 buf->f_bsize = dentry->d_sb->s_blocksize;
1809 buf->f_namelen = BTRFS_NAME_LEN; 2153 buf->f_namelen = BTRFS_NAME_LEN;
1810 2154
1811 /* We treat it as constant endianness 2155 /* We treat it as constant endianness (it doesn't matter _which_)
1812 because we want the fsid to come o 2156 because we want the fsid to come out the same whether mounted
1813 on a big-endian or little-endian h 2157 on a big-endian or little-endian host */
1814 buf->f_fsid.val[0] = be32_to_cpu(fsid 2158 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1815 buf->f_fsid.val[1] = be32_to_cpu(fsid 2159 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1816 /* Mask in the root object ID too, to 2160 /* Mask in the root object ID too, to disambiguate subvols */
1817 buf->f_fsid.val[0] ^= btrfs_root_id(B !! 2161 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
1818 buf->f_fsid.val[1] ^= btrfs_root_id(B !! 2162 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
1819 2163
1820 return 0; 2164 return 0;
1821 } 2165 }
1822 2166
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 2167 static void btrfs_kill_super(struct super_block *sb)
2112 { 2168 {
2113 struct btrfs_fs_info *fs_info = btrfs 2169 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2114 kill_anon_super(sb); 2170 kill_anon_super(sb);
2115 btrfs_free_fs_info(fs_info); !! 2171 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 } 2172 }
2187 2173
2188 static struct file_system_type btrfs_fs_type 2174 static struct file_system_type btrfs_fs_type = {
2189 .owner = THIS_MODULE !! 2175 .owner = THIS_MODULE,
2190 .name = "btrfs", !! 2176 .name = "btrfs",
2191 .init_fs_context = btrfs_init_ !! 2177 .mount = btrfs_mount,
2192 .parameters = btrfs_fs_pa !! 2178 .kill_sb = btrfs_kill_super,
2193 .kill_sb = btrfs_kill_ !! 2179 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2194 .fs_flags = FS_REQUIRES !! 2180 };
2195 }; <<
2196 <<
2197 MODULE_ALIAS_FS("btrfs"); 2181 MODULE_ALIAS_FS("btrfs");
2198 2182
2199 static int btrfs_control_open(struct inode *i 2183 static int btrfs_control_open(struct inode *inode, struct file *file)
2200 { 2184 {
2201 /* 2185 /*
2202 * The control file's private_data is 2186 * The control file's private_data is used to hold the
2203 * transaction when it is started and 2187 * transaction when it is started and is used to keep
2204 * track of whether a transaction is 2188 * track of whether a transaction is already in progress.
2205 */ 2189 */
2206 file->private_data = NULL; 2190 file->private_data = NULL;
2207 return 0; 2191 return 0;
2208 } 2192 }
2209 2193
2210 /* 2194 /*
2211 * Used by /dev/btrfs-control for devices ioc !! 2195 * used by btrfsctl to scan devices when no FS is mounted
2212 */ 2196 */
2213 static long btrfs_control_ioctl(struct file * 2197 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2214 unsigned long 2198 unsigned long arg)
2215 { 2199 {
2216 struct btrfs_ioctl_vol_args *vol; 2200 struct btrfs_ioctl_vol_args *vol;
2217 struct btrfs_device *device = NULL; !! 2201 struct btrfs_fs_devices *fs_devices;
2218 dev_t devt = 0; <<
2219 int ret = -ENOTTY; 2202 int ret = -ENOTTY;
2220 2203
2221 if (!capable(CAP_SYS_ADMIN)) 2204 if (!capable(CAP_SYS_ADMIN))
2222 return -EPERM; 2205 return -EPERM;
2223 2206
2224 vol = memdup_user((void __user *)arg, 2207 vol = memdup_user((void __user *)arg, sizeof(*vol));
2225 if (IS_ERR(vol)) 2208 if (IS_ERR(vol))
2226 return PTR_ERR(vol); 2209 return PTR_ERR(vol);
2227 ret = btrfs_check_ioctl_vol_args_path <<
2228 if (ret < 0) <<
2229 goto out; <<
2230 2210
2231 switch (cmd) { 2211 switch (cmd) {
2232 case BTRFS_IOC_SCAN_DEV: 2212 case BTRFS_IOC_SCAN_DEV:
2233 mutex_lock(&uuid_mutex); !! 2213 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2234 /* !! 2214 &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; 2215 break;
2250 case BTRFS_IOC_DEVICES_READY: 2216 case BTRFS_IOC_DEVICES_READY:
2251 mutex_lock(&uuid_mutex); !! 2217 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2252 /* !! 2218 &btrfs_fs_type, &fs_devices);
2253 * Scanning outside of mount !! 2219 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; 2220 break;
2261 } !! 2221 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; 2222 break;
2266 case BTRFS_IOC_GET_SUPPORTED_FEATURES 2223 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2267 ret = btrfs_ioctl_get_support 2224 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2268 break; 2225 break;
2269 } 2226 }
2270 2227
2271 out: <<
2272 kfree(vol); 2228 kfree(vol);
2273 return ret; 2229 return ret;
2274 } 2230 }
2275 2231
2276 static int btrfs_freeze(struct super_block *s 2232 static int btrfs_freeze(struct super_block *sb)
2277 { 2233 {
>> 2234 struct btrfs_trans_handle *trans;
2278 struct btrfs_fs_info *fs_info = btrfs 2235 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
>> 2236 struct btrfs_root *root = fs_info->tree_root;
2279 2237
2280 set_bit(BTRFS_FS_FROZEN, &fs_info->fl !! 2238 fs_info->fs_frozen = 1;
2281 /* 2239 /*
2282 * We don't need a barrier here, we'l 2240 * We don't need a barrier here, we'll wait for any transaction that
2283 * could be in progress on other thre 2241 * could be in progress on other threads (and do delayed iputs that
2284 * we want to avoid on a frozen files 2242 * we want to avoid on a frozen filesystem), or do the commit
2285 * ourselves. 2243 * ourselves.
2286 */ 2244 */
2287 return btrfs_commit_current_transacti !! 2245 trans = btrfs_attach_transaction_barrier(root);
2288 } !! 2246 if (IS_ERR(trans)) {
2289 !! 2247 /* no transaction, don't bother */
2290 static int check_dev_super(struct btrfs_devic !! 2248 if (PTR_ERR(trans) == -ENOENT)
2291 { !! 2249 return 0;
2292 struct btrfs_fs_info *fs_info = dev-> !! 2250 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 } 2251 }
2337 out: !! 2252 return btrfs_commit_transaction(trans);
2338 btrfs_release_disk_super(sb); <<
2339 return ret; <<
2340 } 2253 }
2341 2254
2342 static int btrfs_unfreeze(struct super_block 2255 static int btrfs_unfreeze(struct super_block *sb)
2343 { 2256 {
2344 struct btrfs_fs_info *fs_info = btrfs !! 2257 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; 2258 return 0;
2373 } 2259 }
2374 2260
2375 static int btrfs_show_devname(struct seq_file 2261 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2376 { 2262 {
2377 struct btrfs_fs_info *fs_info = btrfs 2263 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2378 !! 2264 struct btrfs_fs_devices *cur_devices;
2379 /* !! 2265 struct btrfs_device *dev, *first_dev = NULL;
2380 * There should be always a valid poi !! 2266 struct list_head *head;
2381 * for a short moment in case it's be !! 2267 struct rcu_string *name;
2382 * the end of RCU grace period. !! 2268
2383 */ !! 2269 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2384 rcu_read_lock(); !! 2270 cur_devices = fs_info->fs_devices;
2385 seq_escape(m, btrfs_dev_name(fs_info- !! 2271 while (cur_devices) {
2386 rcu_read_unlock(); !! 2272 head = &cur_devices->devices;
2387 !! 2273 list_for_each_entry(dev, head, dev_list) {
2388 return 0; !! 2274 if (dev->missing)
2389 } !! 2275 continue;
2390 !! 2276 if (!dev->name)
2391 static long btrfs_nr_cached_objects(struct su !! 2277 continue;
2392 { !! 2278 if (!first_dev || dev->devid < first_dev->devid)
2393 struct btrfs_fs_info *fs_info = btrfs !! 2279 first_dev = dev;
2394 const s64 nr = percpu_counter_sum_pos !! 2280 }
2395 !! 2281 cur_devices = cur_devices->seed;
2396 trace_btrfs_extent_map_shrinker_count !! 2282 }
2397 !! 2283
2398 /* !! 2284 if (first_dev) {
2399 * Only report the real number for DE !! 2285 rcu_read_lock();
2400 * serious performance degradation ca !! 2286 name = rcu_dereference(first_dev->name);
2401 */ !! 2287 seq_escape(m, name->str, " \t\n\\");
2402 if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) !! 2288 rcu_read_unlock();
2403 return nr; !! 2289 } else {
>> 2290 WARN_ON(1);
>> 2291 }
>> 2292 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2404 return 0; 2293 return 0;
2405 } 2294 }
2406 2295
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 2296 static const struct super_operations btrfs_super_ops = {
2425 .drop_inode = btrfs_drop_inode, 2297 .drop_inode = btrfs_drop_inode,
2426 .evict_inode = btrfs_evict_inode, 2298 .evict_inode = btrfs_evict_inode,
2427 .put_super = btrfs_put_super, 2299 .put_super = btrfs_put_super,
2428 .sync_fs = btrfs_sync_fs, 2300 .sync_fs = btrfs_sync_fs,
2429 .show_options = btrfs_show_options, 2301 .show_options = btrfs_show_options,
2430 .show_devname = btrfs_show_devname, 2302 .show_devname = btrfs_show_devname,
>> 2303 .write_inode = btrfs_write_inode,
2431 .alloc_inode = btrfs_alloc_inode, 2304 .alloc_inode = btrfs_alloc_inode,
2432 .destroy_inode = btrfs_destroy_inode 2305 .destroy_inode = btrfs_destroy_inode,
2433 .free_inode = btrfs_free_inode, <<
2434 .statfs = btrfs_statfs, 2306 .statfs = btrfs_statfs,
>> 2307 .remount_fs = btrfs_remount,
2435 .freeze_fs = btrfs_freeze, 2308 .freeze_fs = btrfs_freeze,
2436 .unfreeze_fs = btrfs_unfreeze, 2309 .unfreeze_fs = btrfs_unfreeze,
2437 .nr_cached_objects = btrfs_nr_cached_ <<
2438 .free_cached_objects = btrfs_free_cac <<
2439 }; 2310 };
2440 2311
2441 static const struct file_operations btrfs_ctl 2312 static const struct file_operations btrfs_ctl_fops = {
2442 .open = btrfs_control_open, 2313 .open = btrfs_control_open,
2443 .unlocked_ioctl = btrfs_control_ioct 2314 .unlocked_ioctl = btrfs_control_ioctl,
2444 .compat_ioctl = compat_ptr_ioctl, !! 2315 .compat_ioctl = btrfs_control_ioctl,
2445 .owner = THIS_MODULE, 2316 .owner = THIS_MODULE,
2446 .llseek = noop_llseek, 2317 .llseek = noop_llseek,
2447 }; 2318 };
2448 2319
2449 static struct miscdevice btrfs_misc = { 2320 static struct miscdevice btrfs_misc = {
2450 .minor = BTRFS_MINOR, 2321 .minor = BTRFS_MINOR,
2451 .name = "btrfs-control", 2322 .name = "btrfs-control",
2452 .fops = &btrfs_ctl_fops 2323 .fops = &btrfs_ctl_fops
2453 }; 2324 };
2454 2325
2455 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 2326 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2456 MODULE_ALIAS("devname:btrfs-control"); 2327 MODULE_ALIAS("devname:btrfs-control");
2457 2328
2458 static int __init btrfs_interface_init(void) !! 2329 static int btrfs_interface_init(void)
2459 { 2330 {
2460 return misc_register(&btrfs_misc); 2331 return misc_register(&btrfs_misc);
2461 } 2332 }
2462 2333
2463 static __cold void btrfs_interface_exit(void) !! 2334 static void btrfs_interface_exit(void)
2464 { 2335 {
2465 misc_deregister(&btrfs_misc); 2336 misc_deregister(&btrfs_misc);
2466 } 2337 }
2467 2338
2468 static int __init btrfs_print_mod_info(void) !! 2339 static void btrfs_print_mod_info(void)
2469 { 2340 {
2470 static const char options[] = "" !! 2341 pr_info("Btrfs loaded, crc32c=%s"
2471 #ifdef CONFIG_BTRFS_DEBUG 2342 #ifdef CONFIG_BTRFS_DEBUG
2472 ", debug=on" 2343 ", debug=on"
2473 #endif 2344 #endif
2474 #ifdef CONFIG_BTRFS_ASSERT 2345 #ifdef CONFIG_BTRFS_ASSERT
2475 ", assert=on" 2346 ", assert=on"
2476 #endif 2347 #endif
2477 #ifdef CONFIG_BTRFS_FS_REF_VERIFY !! 2348 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2478 ", ref-verify=on" !! 2349 ", integrity-checker=on"
2479 #endif 2350 #endif
2480 #ifdef CONFIG_BLK_DEV_ZONED !! 2351 "\n",
2481 ", zoned=yes" !! 2352 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 } 2353 }
2494 2354
2495 static int register_btrfs(void) !! 2355 static int __init init_btrfs_fs(void)
2496 { 2356 {
2497 return register_filesystem(&btrfs_fs_ !! 2357 int err;
2498 } <<
2499 2358
2500 static void unregister_btrfs(void) !! 2359 err = btrfs_hash_init();
2501 { !! 2360 if (err)
2502 unregister_filesystem(&btrfs_fs_type) !! 2361 return err;
2503 } <<
2504 2362
2505 /* Helper structure for long init/exit functi !! 2363 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 2364
2512 static const struct init_sequence mod_init_se !! 2365 err = btrfs_init_sysfs();
2513 { !! 2366 if (err)
2514 .init_func = btrfs_props_init !! 2367 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 2368
2579 static bool mod_init_result[ARRAY_SIZE(mod_in !! 2369 btrfs_init_compress();
2580 2370
2581 static __always_inline void btrfs_exit_btrfs_ !! 2371 err = btrfs_init_cachep();
2582 { !! 2372 if (err)
2583 int i; !! 2373 goto free_compress;
2584 2374
2585 for (i = ARRAY_SIZE(mod_init_seq) - 1 !! 2375 err = extent_io_init();
2586 if (!mod_init_result[i]) !! 2376 if (err)
2587 continue; !! 2377 goto free_cachep;
2588 if (mod_init_seq[i].exit_func !! 2378
2589 mod_init_seq[i].exit_ !! 2379 err = extent_map_init();
2590 mod_init_result[i] = false; !! 2380 if (err)
2591 } !! 2381 goto free_extent_io;
>> 2382
>> 2383 err = ordered_data_init();
>> 2384 if (err)
>> 2385 goto free_extent_map;
>> 2386
>> 2387 err = btrfs_delayed_inode_init();
>> 2388 if (err)
>> 2389 goto free_ordered_data;
>> 2390
>> 2391 err = btrfs_auto_defrag_init();
>> 2392 if (err)
>> 2393 goto free_delayed_inode;
>> 2394
>> 2395 err = btrfs_delayed_ref_init();
>> 2396 if (err)
>> 2397 goto free_auto_defrag;
>> 2398
>> 2399 err = btrfs_prelim_ref_init();
>> 2400 if (err)
>> 2401 goto free_delayed_ref;
>> 2402
>> 2403 err = btrfs_end_io_wq_init();
>> 2404 if (err)
>> 2405 goto free_prelim_ref;
>> 2406
>> 2407 err = btrfs_interface_init();
>> 2408 if (err)
>> 2409 goto free_end_io_wq;
>> 2410
>> 2411 btrfs_init_lockdep();
>> 2412
>> 2413 btrfs_print_mod_info();
>> 2414
>> 2415 err = btrfs_run_sanity_tests();
>> 2416 if (err)
>> 2417 goto unregister_ioctl;
>> 2418
>> 2419 err = register_filesystem(&btrfs_fs_type);
>> 2420 if (err)
>> 2421 goto unregister_ioctl;
>> 2422
>> 2423 return 0;
>> 2424
>> 2425 unregister_ioctl:
>> 2426 btrfs_interface_exit();
>> 2427 free_end_io_wq:
>> 2428 btrfs_end_io_wq_exit();
>> 2429 free_prelim_ref:
>> 2430 btrfs_prelim_ref_exit();
>> 2431 free_delayed_ref:
>> 2432 btrfs_delayed_ref_exit();
>> 2433 free_auto_defrag:
>> 2434 btrfs_auto_defrag_exit();
>> 2435 free_delayed_inode:
>> 2436 btrfs_delayed_inode_exit();
>> 2437 free_ordered_data:
>> 2438 ordered_data_exit();
>> 2439 free_extent_map:
>> 2440 extent_map_exit();
>> 2441 free_extent_io:
>> 2442 extent_io_exit();
>> 2443 free_cachep:
>> 2444 btrfs_destroy_cachep();
>> 2445 free_compress:
>> 2446 btrfs_exit_compress();
>> 2447 btrfs_exit_sysfs();
>> 2448 free_hash:
>> 2449 btrfs_hash_exit();
>> 2450 return err;
2592 } 2451 }
2593 2452
2594 static void __exit exit_btrfs_fs(void) 2453 static void __exit exit_btrfs_fs(void)
2595 { 2454 {
2596 btrfs_exit_btrfs_fs(); !! 2455 btrfs_destroy_cachep();
>> 2456 btrfs_delayed_ref_exit();
>> 2457 btrfs_auto_defrag_exit();
>> 2458 btrfs_delayed_inode_exit();
>> 2459 btrfs_prelim_ref_exit();
>> 2460 ordered_data_exit();
>> 2461 extent_map_exit();
>> 2462 extent_io_exit();
>> 2463 btrfs_interface_exit();
>> 2464 btrfs_end_io_wq_exit();
>> 2465 unregister_filesystem(&btrfs_fs_type);
>> 2466 btrfs_exit_sysfs();
2597 btrfs_cleanup_fs_uuids(); 2467 btrfs_cleanup_fs_uuids();
2598 } !! 2468 btrfs_exit_compress();
2599 !! 2469 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 } 2470 }
2616 2471
2617 late_initcall(init_btrfs_fs); 2472 late_initcall(init_btrfs_fs);
2618 module_exit(exit_btrfs_fs) 2473 module_exit(exit_btrfs_fs)
2619 2474
2620 MODULE_DESCRIPTION("B-Tree File System (BTRFS <<
2621 MODULE_LICENSE("GPL"); 2475 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 2476
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.