]>
Commit | Line | Data |
---|---|---|
1 | // SPDX-License-Identifier: GPL-2.0 | |
2 | /* | |
3 | * fs/f2fs/super.c | |
4 | * | |
5 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
6 | * http://www.samsung.com/ | |
7 | */ | |
8 | #include <linux/module.h> | |
9 | #include <linux/init.h> | |
10 | #include <linux/fs.h> | |
11 | #include <linux/statfs.h> | |
12 | #include <linux/buffer_head.h> | |
13 | #include <linux/backing-dev.h> | |
14 | #include <linux/kthread.h> | |
15 | #include <linux/parser.h> | |
16 | #include <linux/mount.h> | |
17 | #include <linux/seq_file.h> | |
18 | #include <linux/proc_fs.h> | |
19 | #include <linux/random.h> | |
20 | #include <linux/exportfs.h> | |
21 | #include <linux/blkdev.h> | |
22 | #include <linux/quotaops.h> | |
23 | #include <linux/f2fs_fs.h> | |
24 | #include <linux/sysfs.h> | |
25 | #include <linux/quota.h> | |
26 | #include <linux/unicode.h> | |
27 | #include <linux/part_stat.h> | |
28 | ||
29 | #include "f2fs.h" | |
30 | #include "node.h" | |
31 | #include "segment.h" | |
32 | #include "xattr.h" | |
33 | #include "gc.h" | |
34 | #include "trace.h" | |
35 | ||
36 | #define CREATE_TRACE_POINTS | |
37 | #include <trace/events/f2fs.h> | |
38 | ||
39 | static struct kmem_cache *f2fs_inode_cachep; | |
40 | ||
41 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
42 | ||
43 | const char *f2fs_fault_name[FAULT_MAX] = { | |
44 | [FAULT_KMALLOC] = "kmalloc", | |
45 | [FAULT_KVMALLOC] = "kvmalloc", | |
46 | [FAULT_PAGE_ALLOC] = "page alloc", | |
47 | [FAULT_PAGE_GET] = "page get", | |
48 | [FAULT_ALLOC_BIO] = "alloc bio", | |
49 | [FAULT_ALLOC_NID] = "alloc nid", | |
50 | [FAULT_ORPHAN] = "orphan", | |
51 | [FAULT_BLOCK] = "no more block", | |
52 | [FAULT_DIR_DEPTH] = "too big dir depth", | |
53 | [FAULT_EVICT_INODE] = "evict_inode fail", | |
54 | [FAULT_TRUNCATE] = "truncate fail", | |
55 | [FAULT_READ_IO] = "read IO error", | |
56 | [FAULT_CHECKPOINT] = "checkpoint error", | |
57 | [FAULT_DISCARD] = "discard error", | |
58 | [FAULT_WRITE_IO] = "write IO error", | |
59 | }; | |
60 | ||
61 | void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, | |
62 | unsigned int type) | |
63 | { | |
64 | struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; | |
65 | ||
66 | if (rate) { | |
67 | atomic_set(&ffi->inject_ops, 0); | |
68 | ffi->inject_rate = rate; | |
69 | } | |
70 | ||
71 | if (type) | |
72 | ffi->inject_type = type; | |
73 | ||
74 | if (!rate && !type) | |
75 | memset(ffi, 0, sizeof(struct f2fs_fault_info)); | |
76 | } | |
77 | #endif | |
78 | ||
79 | /* f2fs-wide shrinker description */ | |
80 | static struct shrinker f2fs_shrinker_info = { | |
81 | .scan_objects = f2fs_shrink_scan, | |
82 | .count_objects = f2fs_shrink_count, | |
83 | .seeks = DEFAULT_SEEKS, | |
84 | }; | |
85 | ||
86 | enum { | |
87 | Opt_gc_background, | |
88 | Opt_disable_roll_forward, | |
89 | Opt_norecovery, | |
90 | Opt_discard, | |
91 | Opt_nodiscard, | |
92 | Opt_noheap, | |
93 | Opt_heap, | |
94 | Opt_user_xattr, | |
95 | Opt_nouser_xattr, | |
96 | Opt_acl, | |
97 | Opt_noacl, | |
98 | Opt_active_logs, | |
99 | Opt_disable_ext_identify, | |
100 | Opt_inline_xattr, | |
101 | Opt_noinline_xattr, | |
102 | Opt_inline_xattr_size, | |
103 | Opt_inline_data, | |
104 | Opt_inline_dentry, | |
105 | Opt_noinline_dentry, | |
106 | Opt_flush_merge, | |
107 | Opt_noflush_merge, | |
108 | Opt_nobarrier, | |
109 | Opt_fastboot, | |
110 | Opt_extent_cache, | |
111 | Opt_noextent_cache, | |
112 | Opt_noinline_data, | |
113 | Opt_data_flush, | |
114 | Opt_reserve_root, | |
115 | Opt_resgid, | |
116 | Opt_resuid, | |
117 | Opt_mode, | |
118 | Opt_io_size_bits, | |
119 | Opt_fault_injection, | |
120 | Opt_fault_type, | |
121 | Opt_lazytime, | |
122 | Opt_nolazytime, | |
123 | Opt_quota, | |
124 | Opt_noquota, | |
125 | Opt_usrquota, | |
126 | Opt_grpquota, | |
127 | Opt_prjquota, | |
128 | Opt_usrjquota, | |
129 | Opt_grpjquota, | |
130 | Opt_prjjquota, | |
131 | Opt_offusrjquota, | |
132 | Opt_offgrpjquota, | |
133 | Opt_offprjjquota, | |
134 | Opt_jqfmt_vfsold, | |
135 | Opt_jqfmt_vfsv0, | |
136 | Opt_jqfmt_vfsv1, | |
137 | Opt_whint, | |
138 | Opt_alloc, | |
139 | Opt_fsync, | |
140 | Opt_test_dummy_encryption, | |
141 | Opt_checkpoint_disable, | |
142 | Opt_checkpoint_disable_cap, | |
143 | Opt_checkpoint_disable_cap_perc, | |
144 | Opt_checkpoint_enable, | |
145 | Opt_compress_algorithm, | |
146 | Opt_compress_log_size, | |
147 | Opt_compress_extension, | |
148 | Opt_err, | |
149 | }; | |
150 | ||
151 | static match_table_t f2fs_tokens = { | |
152 | {Opt_gc_background, "background_gc=%s"}, | |
153 | {Opt_disable_roll_forward, "disable_roll_forward"}, | |
154 | {Opt_norecovery, "norecovery"}, | |
155 | {Opt_discard, "discard"}, | |
156 | {Opt_nodiscard, "nodiscard"}, | |
157 | {Opt_noheap, "no_heap"}, | |
158 | {Opt_heap, "heap"}, | |
159 | {Opt_user_xattr, "user_xattr"}, | |
160 | {Opt_nouser_xattr, "nouser_xattr"}, | |
161 | {Opt_acl, "acl"}, | |
162 | {Opt_noacl, "noacl"}, | |
163 | {Opt_active_logs, "active_logs=%u"}, | |
164 | {Opt_disable_ext_identify, "disable_ext_identify"}, | |
165 | {Opt_inline_xattr, "inline_xattr"}, | |
166 | {Opt_noinline_xattr, "noinline_xattr"}, | |
167 | {Opt_inline_xattr_size, "inline_xattr_size=%u"}, | |
168 | {Opt_inline_data, "inline_data"}, | |
169 | {Opt_inline_dentry, "inline_dentry"}, | |
170 | {Opt_noinline_dentry, "noinline_dentry"}, | |
171 | {Opt_flush_merge, "flush_merge"}, | |
172 | {Opt_noflush_merge, "noflush_merge"}, | |
173 | {Opt_nobarrier, "nobarrier"}, | |
174 | {Opt_fastboot, "fastboot"}, | |
175 | {Opt_extent_cache, "extent_cache"}, | |
176 | {Opt_noextent_cache, "noextent_cache"}, | |
177 | {Opt_noinline_data, "noinline_data"}, | |
178 | {Opt_data_flush, "data_flush"}, | |
179 | {Opt_reserve_root, "reserve_root=%u"}, | |
180 | {Opt_resgid, "resgid=%u"}, | |
181 | {Opt_resuid, "resuid=%u"}, | |
182 | {Opt_mode, "mode=%s"}, | |
183 | {Opt_io_size_bits, "io_bits=%u"}, | |
184 | {Opt_fault_injection, "fault_injection=%u"}, | |
185 | {Opt_fault_type, "fault_type=%u"}, | |
186 | {Opt_lazytime, "lazytime"}, | |
187 | {Opt_nolazytime, "nolazytime"}, | |
188 | {Opt_quota, "quota"}, | |
189 | {Opt_noquota, "noquota"}, | |
190 | {Opt_usrquota, "usrquota"}, | |
191 | {Opt_grpquota, "grpquota"}, | |
192 | {Opt_prjquota, "prjquota"}, | |
193 | {Opt_usrjquota, "usrjquota=%s"}, | |
194 | {Opt_grpjquota, "grpjquota=%s"}, | |
195 | {Opt_prjjquota, "prjjquota=%s"}, | |
196 | {Opt_offusrjquota, "usrjquota="}, | |
197 | {Opt_offgrpjquota, "grpjquota="}, | |
198 | {Opt_offprjjquota, "prjjquota="}, | |
199 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
200 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
201 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, | |
202 | {Opt_whint, "whint_mode=%s"}, | |
203 | {Opt_alloc, "alloc_mode=%s"}, | |
204 | {Opt_fsync, "fsync_mode=%s"}, | |
205 | {Opt_test_dummy_encryption, "test_dummy_encryption"}, | |
206 | {Opt_checkpoint_disable, "checkpoint=disable"}, | |
207 | {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, | |
208 | {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, | |
209 | {Opt_checkpoint_enable, "checkpoint=enable"}, | |
210 | {Opt_compress_algorithm, "compress_algorithm=%s"}, | |
211 | {Opt_compress_log_size, "compress_log_size=%u"}, | |
212 | {Opt_compress_extension, "compress_extension=%s"}, | |
213 | {Opt_err, NULL}, | |
214 | }; | |
215 | ||
216 | void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) | |
217 | { | |
218 | struct va_format vaf; | |
219 | va_list args; | |
220 | int level; | |
221 | ||
222 | va_start(args, fmt); | |
223 | ||
224 | level = printk_get_level(fmt); | |
225 | vaf.fmt = printk_skip_level(fmt); | |
226 | vaf.va = &args; | |
227 | printk("%c%cF2FS-fs (%s): %pV\n", | |
228 | KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); | |
229 | ||
230 | va_end(args); | |
231 | } | |
232 | ||
233 | #ifdef CONFIG_UNICODE | |
234 | static const struct f2fs_sb_encodings { | |
235 | __u16 magic; | |
236 | char *name; | |
237 | char *version; | |
238 | } f2fs_sb_encoding_map[] = { | |
239 | {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"}, | |
240 | }; | |
241 | ||
242 | static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb, | |
243 | const struct f2fs_sb_encodings **encoding, | |
244 | __u16 *flags) | |
245 | { | |
246 | __u16 magic = le16_to_cpu(sb->s_encoding); | |
247 | int i; | |
248 | ||
249 | for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) | |
250 | if (magic == f2fs_sb_encoding_map[i].magic) | |
251 | break; | |
252 | ||
253 | if (i >= ARRAY_SIZE(f2fs_sb_encoding_map)) | |
254 | return -EINVAL; | |
255 | ||
256 | *encoding = &f2fs_sb_encoding_map[i]; | |
257 | *flags = le16_to_cpu(sb->s_encoding_flags); | |
258 | ||
259 | return 0; | |
260 | } | |
261 | #endif | |
262 | ||
263 | static inline void limit_reserve_root(struct f2fs_sb_info *sbi) | |
264 | { | |
265 | block_t limit = min((sbi->user_block_count << 1) / 1000, | |
266 | sbi->user_block_count - sbi->reserved_blocks); | |
267 | ||
268 | /* limit is 0.2% */ | |
269 | if (test_opt(sbi, RESERVE_ROOT) && | |
270 | F2FS_OPTION(sbi).root_reserved_blocks > limit) { | |
271 | F2FS_OPTION(sbi).root_reserved_blocks = limit; | |
272 | f2fs_info(sbi, "Reduce reserved blocks for root = %u", | |
273 | F2FS_OPTION(sbi).root_reserved_blocks); | |
274 | } | |
275 | if (!test_opt(sbi, RESERVE_ROOT) && | |
276 | (!uid_eq(F2FS_OPTION(sbi).s_resuid, | |
277 | make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || | |
278 | !gid_eq(F2FS_OPTION(sbi).s_resgid, | |
279 | make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) | |
280 | f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", | |
281 | from_kuid_munged(&init_user_ns, | |
282 | F2FS_OPTION(sbi).s_resuid), | |
283 | from_kgid_munged(&init_user_ns, | |
284 | F2FS_OPTION(sbi).s_resgid)); | |
285 | } | |
286 | ||
287 | static void init_once(void *foo) | |
288 | { | |
289 | struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; | |
290 | ||
291 | inode_init_once(&fi->vfs_inode); | |
292 | } | |
293 | ||
294 | #ifdef CONFIG_QUOTA | |
295 | static const char * const quotatypes[] = INITQFNAMES; | |
296 | #define QTYPE2NAME(t) (quotatypes[t]) | |
297 | static int f2fs_set_qf_name(struct super_block *sb, int qtype, | |
298 | substring_t *args) | |
299 | { | |
300 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
301 | char *qname; | |
302 | int ret = -EINVAL; | |
303 | ||
304 | if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { | |
305 | f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); | |
306 | return -EINVAL; | |
307 | } | |
308 | if (f2fs_sb_has_quota_ino(sbi)) { | |
309 | f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); | |
310 | return 0; | |
311 | } | |
312 | ||
313 | qname = match_strdup(args); | |
314 | if (!qname) { | |
315 | f2fs_err(sbi, "Not enough memory for storing quotafile name"); | |
316 | return -ENOMEM; | |
317 | } | |
318 | if (F2FS_OPTION(sbi).s_qf_names[qtype]) { | |
319 | if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) | |
320 | ret = 0; | |
321 | else | |
322 | f2fs_err(sbi, "%s quota file already specified", | |
323 | QTYPE2NAME(qtype)); | |
324 | goto errout; | |
325 | } | |
326 | if (strchr(qname, '/')) { | |
327 | f2fs_err(sbi, "quotafile must be on filesystem root"); | |
328 | goto errout; | |
329 | } | |
330 | F2FS_OPTION(sbi).s_qf_names[qtype] = qname; | |
331 | set_opt(sbi, QUOTA); | |
332 | return 0; | |
333 | errout: | |
334 | kvfree(qname); | |
335 | return ret; | |
336 | } | |
337 | ||
338 | static int f2fs_clear_qf_name(struct super_block *sb, int qtype) | |
339 | { | |
340 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
341 | ||
342 | if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { | |
343 | f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); | |
344 | return -EINVAL; | |
345 | } | |
346 | kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]); | |
347 | F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; | |
348 | return 0; | |
349 | } | |
350 | ||
351 | static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) | |
352 | { | |
353 | /* | |
354 | * We do the test below only for project quotas. 'usrquota' and | |
355 | * 'grpquota' mount options are allowed even without quota feature | |
356 | * to support legacy quotas in quota files. | |
357 | */ | |
358 | if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { | |
359 | f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); | |
360 | return -1; | |
361 | } | |
362 | if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || | |
363 | F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || | |
364 | F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { | |
365 | if (test_opt(sbi, USRQUOTA) && | |
366 | F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) | |
367 | clear_opt(sbi, USRQUOTA); | |
368 | ||
369 | if (test_opt(sbi, GRPQUOTA) && | |
370 | F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) | |
371 | clear_opt(sbi, GRPQUOTA); | |
372 | ||
373 | if (test_opt(sbi, PRJQUOTA) && | |
374 | F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) | |
375 | clear_opt(sbi, PRJQUOTA); | |
376 | ||
377 | if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || | |
378 | test_opt(sbi, PRJQUOTA)) { | |
379 | f2fs_err(sbi, "old and new quota format mixing"); | |
380 | return -1; | |
381 | } | |
382 | ||
383 | if (!F2FS_OPTION(sbi).s_jquota_fmt) { | |
384 | f2fs_err(sbi, "journaled quota format not specified"); | |
385 | return -1; | |
386 | } | |
387 | } | |
388 | ||
389 | if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { | |
390 | f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); | |
391 | F2FS_OPTION(sbi).s_jquota_fmt = 0; | |
392 | } | |
393 | return 0; | |
394 | } | |
395 | #endif | |
396 | ||
397 | static int parse_options(struct super_block *sb, char *options) | |
398 | { | |
399 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
400 | substring_t args[MAX_OPT_ARGS]; | |
401 | unsigned char (*ext)[F2FS_EXTENSION_LEN]; | |
402 | char *p, *name; | |
403 | int arg = 0, ext_cnt; | |
404 | kuid_t uid; | |
405 | kgid_t gid; | |
406 | #ifdef CONFIG_QUOTA | |
407 | int ret; | |
408 | #endif | |
409 | ||
410 | if (!options) | |
411 | return 0; | |
412 | ||
413 | while ((p = strsep(&options, ",")) != NULL) { | |
414 | int token; | |
415 | if (!*p) | |
416 | continue; | |
417 | /* | |
418 | * Initialize args struct so we know whether arg was | |
419 | * found; some options take optional arguments. | |
420 | */ | |
421 | args[0].to = args[0].from = NULL; | |
422 | token = match_token(p, f2fs_tokens, args); | |
423 | ||
424 | switch (token) { | |
425 | case Opt_gc_background: | |
426 | name = match_strdup(&args[0]); | |
427 | ||
428 | if (!name) | |
429 | return -ENOMEM; | |
430 | if (!strcmp(name, "on")) { | |
431 | F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; | |
432 | } else if (!strcmp(name, "off")) { | |
433 | F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; | |
434 | } else if (!strcmp(name, "sync")) { | |
435 | F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; | |
436 | } else { | |
437 | kvfree(name); | |
438 | return -EINVAL; | |
439 | } | |
440 | kvfree(name); | |
441 | break; | |
442 | case Opt_disable_roll_forward: | |
443 | set_opt(sbi, DISABLE_ROLL_FORWARD); | |
444 | break; | |
445 | case Opt_norecovery: | |
446 | /* this option mounts f2fs with ro */ | |
447 | set_opt(sbi, NORECOVERY); | |
448 | if (!f2fs_readonly(sb)) | |
449 | return -EINVAL; | |
450 | break; | |
451 | case Opt_discard: | |
452 | set_opt(sbi, DISCARD); | |
453 | break; | |
454 | case Opt_nodiscard: | |
455 | if (f2fs_sb_has_blkzoned(sbi)) { | |
456 | f2fs_warn(sbi, "discard is required for zoned block devices"); | |
457 | return -EINVAL; | |
458 | } | |
459 | clear_opt(sbi, DISCARD); | |
460 | break; | |
461 | case Opt_noheap: | |
462 | set_opt(sbi, NOHEAP); | |
463 | break; | |
464 | case Opt_heap: | |
465 | clear_opt(sbi, NOHEAP); | |
466 | break; | |
467 | #ifdef CONFIG_F2FS_FS_XATTR | |
468 | case Opt_user_xattr: | |
469 | set_opt(sbi, XATTR_USER); | |
470 | break; | |
471 | case Opt_nouser_xattr: | |
472 | clear_opt(sbi, XATTR_USER); | |
473 | break; | |
474 | case Opt_inline_xattr: | |
475 | set_opt(sbi, INLINE_XATTR); | |
476 | break; | |
477 | case Opt_noinline_xattr: | |
478 | clear_opt(sbi, INLINE_XATTR); | |
479 | break; | |
480 | case Opt_inline_xattr_size: | |
481 | if (args->from && match_int(args, &arg)) | |
482 | return -EINVAL; | |
483 | set_opt(sbi, INLINE_XATTR_SIZE); | |
484 | F2FS_OPTION(sbi).inline_xattr_size = arg; | |
485 | break; | |
486 | #else | |
487 | case Opt_user_xattr: | |
488 | f2fs_info(sbi, "user_xattr options not supported"); | |
489 | break; | |
490 | case Opt_nouser_xattr: | |
491 | f2fs_info(sbi, "nouser_xattr options not supported"); | |
492 | break; | |
493 | case Opt_inline_xattr: | |
494 | f2fs_info(sbi, "inline_xattr options not supported"); | |
495 | break; | |
496 | case Opt_noinline_xattr: | |
497 | f2fs_info(sbi, "noinline_xattr options not supported"); | |
498 | break; | |
499 | #endif | |
500 | #ifdef CONFIG_F2FS_FS_POSIX_ACL | |
501 | case Opt_acl: | |
502 | set_opt(sbi, POSIX_ACL); | |
503 | break; | |
504 | case Opt_noacl: | |
505 | clear_opt(sbi, POSIX_ACL); | |
506 | break; | |
507 | #else | |
508 | case Opt_acl: | |
509 | f2fs_info(sbi, "acl options not supported"); | |
510 | break; | |
511 | case Opt_noacl: | |
512 | f2fs_info(sbi, "noacl options not supported"); | |
513 | break; | |
514 | #endif | |
515 | case Opt_active_logs: | |
516 | if (args->from && match_int(args, &arg)) | |
517 | return -EINVAL; | |
518 | if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE) | |
519 | return -EINVAL; | |
520 | F2FS_OPTION(sbi).active_logs = arg; | |
521 | break; | |
522 | case Opt_disable_ext_identify: | |
523 | set_opt(sbi, DISABLE_EXT_IDENTIFY); | |
524 | break; | |
525 | case Opt_inline_data: | |
526 | set_opt(sbi, INLINE_DATA); | |
527 | break; | |
528 | case Opt_inline_dentry: | |
529 | set_opt(sbi, INLINE_DENTRY); | |
530 | break; | |
531 | case Opt_noinline_dentry: | |
532 | clear_opt(sbi, INLINE_DENTRY); | |
533 | break; | |
534 | case Opt_flush_merge: | |
535 | set_opt(sbi, FLUSH_MERGE); | |
536 | break; | |
537 | case Opt_noflush_merge: | |
538 | clear_opt(sbi, FLUSH_MERGE); | |
539 | break; | |
540 | case Opt_nobarrier: | |
541 | set_opt(sbi, NOBARRIER); | |
542 | break; | |
543 | case Opt_fastboot: | |
544 | set_opt(sbi, FASTBOOT); | |
545 | break; | |
546 | case Opt_extent_cache: | |
547 | set_opt(sbi, EXTENT_CACHE); | |
548 | break; | |
549 | case Opt_noextent_cache: | |
550 | clear_opt(sbi, EXTENT_CACHE); | |
551 | break; | |
552 | case Opt_noinline_data: | |
553 | clear_opt(sbi, INLINE_DATA); | |
554 | break; | |
555 | case Opt_data_flush: | |
556 | set_opt(sbi, DATA_FLUSH); | |
557 | break; | |
558 | case Opt_reserve_root: | |
559 | if (args->from && match_int(args, &arg)) | |
560 | return -EINVAL; | |
561 | if (test_opt(sbi, RESERVE_ROOT)) { | |
562 | f2fs_info(sbi, "Preserve previous reserve_root=%u", | |
563 | F2FS_OPTION(sbi).root_reserved_blocks); | |
564 | } else { | |
565 | F2FS_OPTION(sbi).root_reserved_blocks = arg; | |
566 | set_opt(sbi, RESERVE_ROOT); | |
567 | } | |
568 | break; | |
569 | case Opt_resuid: | |
570 | if (args->from && match_int(args, &arg)) | |
571 | return -EINVAL; | |
572 | uid = make_kuid(current_user_ns(), arg); | |
573 | if (!uid_valid(uid)) { | |
574 | f2fs_err(sbi, "Invalid uid value %d", arg); | |
575 | return -EINVAL; | |
576 | } | |
577 | F2FS_OPTION(sbi).s_resuid = uid; | |
578 | break; | |
579 | case Opt_resgid: | |
580 | if (args->from && match_int(args, &arg)) | |
581 | return -EINVAL; | |
582 | gid = make_kgid(current_user_ns(), arg); | |
583 | if (!gid_valid(gid)) { | |
584 | f2fs_err(sbi, "Invalid gid value %d", arg); | |
585 | return -EINVAL; | |
586 | } | |
587 | F2FS_OPTION(sbi).s_resgid = gid; | |
588 | break; | |
589 | case Opt_mode: | |
590 | name = match_strdup(&args[0]); | |
591 | ||
592 | if (!name) | |
593 | return -ENOMEM; | |
594 | if (!strcmp(name, "adaptive")) { | |
595 | if (f2fs_sb_has_blkzoned(sbi)) { | |
596 | f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); | |
597 | kvfree(name); | |
598 | return -EINVAL; | |
599 | } | |
600 | F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; | |
601 | } else if (!strcmp(name, "lfs")) { | |
602 | F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; | |
603 | } else { | |
604 | kvfree(name); | |
605 | return -EINVAL; | |
606 | } | |
607 | kvfree(name); | |
608 | break; | |
609 | case Opt_io_size_bits: | |
610 | if (args->from && match_int(args, &arg)) | |
611 | return -EINVAL; | |
612 | if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) { | |
613 | f2fs_warn(sbi, "Not support %d, larger than %d", | |
614 | 1 << arg, BIO_MAX_PAGES); | |
615 | return -EINVAL; | |
616 | } | |
617 | F2FS_OPTION(sbi).write_io_size_bits = arg; | |
618 | break; | |
619 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
620 | case Opt_fault_injection: | |
621 | if (args->from && match_int(args, &arg)) | |
622 | return -EINVAL; | |
623 | f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); | |
624 | set_opt(sbi, FAULT_INJECTION); | |
625 | break; | |
626 | ||
627 | case Opt_fault_type: | |
628 | if (args->from && match_int(args, &arg)) | |
629 | return -EINVAL; | |
630 | f2fs_build_fault_attr(sbi, 0, arg); | |
631 | set_opt(sbi, FAULT_INJECTION); | |
632 | break; | |
633 | #else | |
634 | case Opt_fault_injection: | |
635 | f2fs_info(sbi, "fault_injection options not supported"); | |
636 | break; | |
637 | ||
638 | case Opt_fault_type: | |
639 | f2fs_info(sbi, "fault_type options not supported"); | |
640 | break; | |
641 | #endif | |
642 | case Opt_lazytime: | |
643 | sb->s_flags |= SB_LAZYTIME; | |
644 | break; | |
645 | case Opt_nolazytime: | |
646 | sb->s_flags &= ~SB_LAZYTIME; | |
647 | break; | |
648 | #ifdef CONFIG_QUOTA | |
649 | case Opt_quota: | |
650 | case Opt_usrquota: | |
651 | set_opt(sbi, USRQUOTA); | |
652 | break; | |
653 | case Opt_grpquota: | |
654 | set_opt(sbi, GRPQUOTA); | |
655 | break; | |
656 | case Opt_prjquota: | |
657 | set_opt(sbi, PRJQUOTA); | |
658 | break; | |
659 | case Opt_usrjquota: | |
660 | ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); | |
661 | if (ret) | |
662 | return ret; | |
663 | break; | |
664 | case Opt_grpjquota: | |
665 | ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); | |
666 | if (ret) | |
667 | return ret; | |
668 | break; | |
669 | case Opt_prjjquota: | |
670 | ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); | |
671 | if (ret) | |
672 | return ret; | |
673 | break; | |
674 | case Opt_offusrjquota: | |
675 | ret = f2fs_clear_qf_name(sb, USRQUOTA); | |
676 | if (ret) | |
677 | return ret; | |
678 | break; | |
679 | case Opt_offgrpjquota: | |
680 | ret = f2fs_clear_qf_name(sb, GRPQUOTA); | |
681 | if (ret) | |
682 | return ret; | |
683 | break; | |
684 | case Opt_offprjjquota: | |
685 | ret = f2fs_clear_qf_name(sb, PRJQUOTA); | |
686 | if (ret) | |
687 | return ret; | |
688 | break; | |
689 | case Opt_jqfmt_vfsold: | |
690 | F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; | |
691 | break; | |
692 | case Opt_jqfmt_vfsv0: | |
693 | F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; | |
694 | break; | |
695 | case Opt_jqfmt_vfsv1: | |
696 | F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; | |
697 | break; | |
698 | case Opt_noquota: | |
699 | clear_opt(sbi, QUOTA); | |
700 | clear_opt(sbi, USRQUOTA); | |
701 | clear_opt(sbi, GRPQUOTA); | |
702 | clear_opt(sbi, PRJQUOTA); | |
703 | break; | |
704 | #else | |
705 | case Opt_quota: | |
706 | case Opt_usrquota: | |
707 | case Opt_grpquota: | |
708 | case Opt_prjquota: | |
709 | case Opt_usrjquota: | |
710 | case Opt_grpjquota: | |
711 | case Opt_prjjquota: | |
712 | case Opt_offusrjquota: | |
713 | case Opt_offgrpjquota: | |
714 | case Opt_offprjjquota: | |
715 | case Opt_jqfmt_vfsold: | |
716 | case Opt_jqfmt_vfsv0: | |
717 | case Opt_jqfmt_vfsv1: | |
718 | case Opt_noquota: | |
719 | f2fs_info(sbi, "quota operations not supported"); | |
720 | break; | |
721 | #endif | |
722 | case Opt_whint: | |
723 | name = match_strdup(&args[0]); | |
724 | if (!name) | |
725 | return -ENOMEM; | |
726 | if (!strcmp(name, "user-based")) { | |
727 | F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; | |
728 | } else if (!strcmp(name, "off")) { | |
729 | F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; | |
730 | } else if (!strcmp(name, "fs-based")) { | |
731 | F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; | |
732 | } else { | |
733 | kvfree(name); | |
734 | return -EINVAL; | |
735 | } | |
736 | kvfree(name); | |
737 | break; | |
738 | case Opt_alloc: | |
739 | name = match_strdup(&args[0]); | |
740 | if (!name) | |
741 | return -ENOMEM; | |
742 | ||
743 | if (!strcmp(name, "default")) { | |
744 | F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; | |
745 | } else if (!strcmp(name, "reuse")) { | |
746 | F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; | |
747 | } else { | |
748 | kvfree(name); | |
749 | return -EINVAL; | |
750 | } | |
751 | kvfree(name); | |
752 | break; | |
753 | case Opt_fsync: | |
754 | name = match_strdup(&args[0]); | |
755 | if (!name) | |
756 | return -ENOMEM; | |
757 | if (!strcmp(name, "posix")) { | |
758 | F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; | |
759 | } else if (!strcmp(name, "strict")) { | |
760 | F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; | |
761 | } else if (!strcmp(name, "nobarrier")) { | |
762 | F2FS_OPTION(sbi).fsync_mode = | |
763 | FSYNC_MODE_NOBARRIER; | |
764 | } else { | |
765 | kvfree(name); | |
766 | return -EINVAL; | |
767 | } | |
768 | kvfree(name); | |
769 | break; | |
770 | case Opt_test_dummy_encryption: | |
771 | #ifdef CONFIG_FS_ENCRYPTION | |
772 | if (!f2fs_sb_has_encrypt(sbi)) { | |
773 | f2fs_err(sbi, "Encrypt feature is off"); | |
774 | return -EINVAL; | |
775 | } | |
776 | ||
777 | F2FS_OPTION(sbi).test_dummy_encryption = true; | |
778 | f2fs_info(sbi, "Test dummy encryption mode enabled"); | |
779 | #else | |
780 | f2fs_info(sbi, "Test dummy encryption mount option ignored"); | |
781 | #endif | |
782 | break; | |
783 | case Opt_checkpoint_disable_cap_perc: | |
784 | if (args->from && match_int(args, &arg)) | |
785 | return -EINVAL; | |
786 | if (arg < 0 || arg > 100) | |
787 | return -EINVAL; | |
788 | if (arg == 100) | |
789 | F2FS_OPTION(sbi).unusable_cap = | |
790 | sbi->user_block_count; | |
791 | else | |
792 | F2FS_OPTION(sbi).unusable_cap = | |
793 | (sbi->user_block_count / 100) * arg; | |
794 | set_opt(sbi, DISABLE_CHECKPOINT); | |
795 | break; | |
796 | case Opt_checkpoint_disable_cap: | |
797 | if (args->from && match_int(args, &arg)) | |
798 | return -EINVAL; | |
799 | F2FS_OPTION(sbi).unusable_cap = arg; | |
800 | set_opt(sbi, DISABLE_CHECKPOINT); | |
801 | break; | |
802 | case Opt_checkpoint_disable: | |
803 | set_opt(sbi, DISABLE_CHECKPOINT); | |
804 | break; | |
805 | case Opt_checkpoint_enable: | |
806 | clear_opt(sbi, DISABLE_CHECKPOINT); | |
807 | break; | |
808 | case Opt_compress_algorithm: | |
809 | if (!f2fs_sb_has_compression(sbi)) { | |
810 | f2fs_err(sbi, "Compression feature if off"); | |
811 | return -EINVAL; | |
812 | } | |
813 | name = match_strdup(&args[0]); | |
814 | if (!name) | |
815 | return -ENOMEM; | |
816 | if (!strcmp(name, "lzo")) { | |
817 | F2FS_OPTION(sbi).compress_algorithm = | |
818 | COMPRESS_LZO; | |
819 | } else if (!strcmp(name, "lz4")) { | |
820 | F2FS_OPTION(sbi).compress_algorithm = | |
821 | COMPRESS_LZ4; | |
822 | } else if (!strcmp(name, "zstd")) { | |
823 | F2FS_OPTION(sbi).compress_algorithm = | |
824 | COMPRESS_ZSTD; | |
825 | } else if (!strcmp(name, "lzo-rle")) { | |
826 | F2FS_OPTION(sbi).compress_algorithm = | |
827 | COMPRESS_LZORLE; | |
828 | } else { | |
829 | kfree(name); | |
830 | return -EINVAL; | |
831 | } | |
832 | kfree(name); | |
833 | break; | |
834 | case Opt_compress_log_size: | |
835 | if (!f2fs_sb_has_compression(sbi)) { | |
836 | f2fs_err(sbi, "Compression feature is off"); | |
837 | return -EINVAL; | |
838 | } | |
839 | if (args->from && match_int(args, &arg)) | |
840 | return -EINVAL; | |
841 | if (arg < MIN_COMPRESS_LOG_SIZE || | |
842 | arg > MAX_COMPRESS_LOG_SIZE) { | |
843 | f2fs_err(sbi, | |
844 | "Compress cluster log size is out of range"); | |
845 | return -EINVAL; | |
846 | } | |
847 | F2FS_OPTION(sbi).compress_log_size = arg; | |
848 | break; | |
849 | case Opt_compress_extension: | |
850 | if (!f2fs_sb_has_compression(sbi)) { | |
851 | f2fs_err(sbi, "Compression feature is off"); | |
852 | return -EINVAL; | |
853 | } | |
854 | name = match_strdup(&args[0]); | |
855 | if (!name) | |
856 | return -ENOMEM; | |
857 | ||
858 | ext = F2FS_OPTION(sbi).extensions; | |
859 | ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; | |
860 | ||
861 | if (strlen(name) >= F2FS_EXTENSION_LEN || | |
862 | ext_cnt >= COMPRESS_EXT_NUM) { | |
863 | f2fs_err(sbi, | |
864 | "invalid extension length/number"); | |
865 | kfree(name); | |
866 | return -EINVAL; | |
867 | } | |
868 | ||
869 | strcpy(ext[ext_cnt], name); | |
870 | F2FS_OPTION(sbi).compress_ext_cnt++; | |
871 | kfree(name); | |
872 | break; | |
873 | default: | |
874 | f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", | |
875 | p); | |
876 | return -EINVAL; | |
877 | } | |
878 | } | |
879 | #ifdef CONFIG_QUOTA | |
880 | if (f2fs_check_quota_options(sbi)) | |
881 | return -EINVAL; | |
882 | #else | |
883 | if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { | |
884 | f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); | |
885 | return -EINVAL; | |
886 | } | |
887 | if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { | |
888 | f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); | |
889 | return -EINVAL; | |
890 | } | |
891 | #endif | |
892 | #ifndef CONFIG_UNICODE | |
893 | if (f2fs_sb_has_casefold(sbi)) { | |
894 | f2fs_err(sbi, | |
895 | "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); | |
896 | return -EINVAL; | |
897 | } | |
898 | #endif | |
899 | ||
900 | if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { | |
901 | f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", | |
902 | F2FS_IO_SIZE_KB(sbi)); | |
903 | return -EINVAL; | |
904 | } | |
905 | ||
906 | if (test_opt(sbi, INLINE_XATTR_SIZE)) { | |
907 | int min_size, max_size; | |
908 | ||
909 | if (!f2fs_sb_has_extra_attr(sbi) || | |
910 | !f2fs_sb_has_flexible_inline_xattr(sbi)) { | |
911 | f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); | |
912 | return -EINVAL; | |
913 | } | |
914 | if (!test_opt(sbi, INLINE_XATTR)) { | |
915 | f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); | |
916 | return -EINVAL; | |
917 | } | |
918 | ||
919 | min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); | |
920 | max_size = MAX_INLINE_XATTR_SIZE; | |
921 | ||
922 | if (F2FS_OPTION(sbi).inline_xattr_size < min_size || | |
923 | F2FS_OPTION(sbi).inline_xattr_size > max_size) { | |
924 | f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", | |
925 | min_size, max_size); | |
926 | return -EINVAL; | |
927 | } | |
928 | } | |
929 | ||
930 | if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { | |
931 | f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n"); | |
932 | return -EINVAL; | |
933 | } | |
934 | ||
935 | /* Not pass down write hints if the number of active logs is lesser | |
936 | * than NR_CURSEG_TYPE. | |
937 | */ | |
938 | if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE) | |
939 | F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; | |
940 | return 0; | |
941 | } | |
942 | ||
943 | static struct inode *f2fs_alloc_inode(struct super_block *sb) | |
944 | { | |
945 | struct f2fs_inode_info *fi; | |
946 | ||
947 | fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); | |
948 | if (!fi) | |
949 | return NULL; | |
950 | ||
951 | init_once((void *) fi); | |
952 | ||
953 | /* Initialize f2fs-specific inode info */ | |
954 | atomic_set(&fi->dirty_pages, 0); | |
955 | init_rwsem(&fi->i_sem); | |
956 | spin_lock_init(&fi->i_size_lock); | |
957 | INIT_LIST_HEAD(&fi->dirty_list); | |
958 | INIT_LIST_HEAD(&fi->gdirty_list); | |
959 | INIT_LIST_HEAD(&fi->inmem_ilist); | |
960 | INIT_LIST_HEAD(&fi->inmem_pages); | |
961 | mutex_init(&fi->inmem_lock); | |
962 | init_rwsem(&fi->i_gc_rwsem[READ]); | |
963 | init_rwsem(&fi->i_gc_rwsem[WRITE]); | |
964 | init_rwsem(&fi->i_mmap_sem); | |
965 | init_rwsem(&fi->i_xattr_sem); | |
966 | ||
967 | /* Will be used by directory only */ | |
968 | fi->i_dir_level = F2FS_SB(sb)->dir_level; | |
969 | ||
970 | return &fi->vfs_inode; | |
971 | } | |
972 | ||
973 | static int f2fs_drop_inode(struct inode *inode) | |
974 | { | |
975 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
976 | int ret; | |
977 | ||
978 | /* | |
979 | * during filesystem shutdown, if checkpoint is disabled, | |
980 | * drop useless meta/node dirty pages. | |
981 | */ | |
982 | if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { | |
983 | if (inode->i_ino == F2FS_NODE_INO(sbi) || | |
984 | inode->i_ino == F2FS_META_INO(sbi)) { | |
985 | trace_f2fs_drop_inode(inode, 1); | |
986 | return 1; | |
987 | } | |
988 | } | |
989 | ||
990 | /* | |
991 | * This is to avoid a deadlock condition like below. | |
992 | * writeback_single_inode(inode) | |
993 | * - f2fs_write_data_page | |
994 | * - f2fs_gc -> iput -> evict | |
995 | * - inode_wait_for_writeback(inode) | |
996 | */ | |
997 | if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { | |
998 | if (!inode->i_nlink && !is_bad_inode(inode)) { | |
999 | /* to avoid evict_inode call simultaneously */ | |
1000 | atomic_inc(&inode->i_count); | |
1001 | spin_unlock(&inode->i_lock); | |
1002 | ||
1003 | /* some remained atomic pages should discarded */ | |
1004 | if (f2fs_is_atomic_file(inode)) | |
1005 | f2fs_drop_inmem_pages(inode); | |
1006 | ||
1007 | /* should remain fi->extent_tree for writepage */ | |
1008 | f2fs_destroy_extent_node(inode); | |
1009 | ||
1010 | sb_start_intwrite(inode->i_sb); | |
1011 | f2fs_i_size_write(inode, 0); | |
1012 | ||
1013 | f2fs_submit_merged_write_cond(F2FS_I_SB(inode), | |
1014 | inode, NULL, 0, DATA); | |
1015 | truncate_inode_pages_final(inode->i_mapping); | |
1016 | ||
1017 | if (F2FS_HAS_BLOCKS(inode)) | |
1018 | f2fs_truncate(inode); | |
1019 | ||
1020 | sb_end_intwrite(inode->i_sb); | |
1021 | ||
1022 | spin_lock(&inode->i_lock); | |
1023 | atomic_dec(&inode->i_count); | |
1024 | } | |
1025 | trace_f2fs_drop_inode(inode, 0); | |
1026 | return 0; | |
1027 | } | |
1028 | ret = generic_drop_inode(inode); | |
1029 | if (!ret) | |
1030 | ret = fscrypt_drop_inode(inode); | |
1031 | trace_f2fs_drop_inode(inode, ret); | |
1032 | return ret; | |
1033 | } | |
1034 | ||
1035 | int f2fs_inode_dirtied(struct inode *inode, bool sync) | |
1036 | { | |
1037 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
1038 | int ret = 0; | |
1039 | ||
1040 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
1041 | if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { | |
1042 | ret = 1; | |
1043 | } else { | |
1044 | set_inode_flag(inode, FI_DIRTY_INODE); | |
1045 | stat_inc_dirty_inode(sbi, DIRTY_META); | |
1046 | } | |
1047 | if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { | |
1048 | list_add_tail(&F2FS_I(inode)->gdirty_list, | |
1049 | &sbi->inode_list[DIRTY_META]); | |
1050 | inc_page_count(sbi, F2FS_DIRTY_IMETA); | |
1051 | } | |
1052 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1053 | return ret; | |
1054 | } | |
1055 | ||
1056 | void f2fs_inode_synced(struct inode *inode) | |
1057 | { | |
1058 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
1059 | ||
1060 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
1061 | if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { | |
1062 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1063 | return; | |
1064 | } | |
1065 | if (!list_empty(&F2FS_I(inode)->gdirty_list)) { | |
1066 | list_del_init(&F2FS_I(inode)->gdirty_list); | |
1067 | dec_page_count(sbi, F2FS_DIRTY_IMETA); | |
1068 | } | |
1069 | clear_inode_flag(inode, FI_DIRTY_INODE); | |
1070 | clear_inode_flag(inode, FI_AUTO_RECOVER); | |
1071 | stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); | |
1072 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
1073 | } | |
1074 | ||
1075 | /* | |
1076 | * f2fs_dirty_inode() is called from __mark_inode_dirty() | |
1077 | * | |
1078 | * We should call set_dirty_inode to write the dirty inode through write_inode. | |
1079 | */ | |
1080 | static void f2fs_dirty_inode(struct inode *inode, int flags) | |
1081 | { | |
1082 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
1083 | ||
1084 | if (inode->i_ino == F2FS_NODE_INO(sbi) || | |
1085 | inode->i_ino == F2FS_META_INO(sbi)) | |
1086 | return; | |
1087 | ||
1088 | if (flags == I_DIRTY_TIME) | |
1089 | return; | |
1090 | ||
1091 | if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) | |
1092 | clear_inode_flag(inode, FI_AUTO_RECOVER); | |
1093 | ||
1094 | f2fs_inode_dirtied(inode, false); | |
1095 | } | |
1096 | ||
1097 | static void f2fs_free_inode(struct inode *inode) | |
1098 | { | |
1099 | fscrypt_free_inode(inode); | |
1100 | kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); | |
1101 | } | |
1102 | ||
1103 | static void destroy_percpu_info(struct f2fs_sb_info *sbi) | |
1104 | { | |
1105 | percpu_counter_destroy(&sbi->alloc_valid_block_count); | |
1106 | percpu_counter_destroy(&sbi->total_valid_inode_count); | |
1107 | } | |
1108 | ||
1109 | static void destroy_device_list(struct f2fs_sb_info *sbi) | |
1110 | { | |
1111 | int i; | |
1112 | ||
1113 | for (i = 0; i < sbi->s_ndevs; i++) { | |
1114 | blkdev_put(FDEV(i).bdev, FMODE_EXCL); | |
1115 | #ifdef CONFIG_BLK_DEV_ZONED | |
1116 | kvfree(FDEV(i).blkz_seq); | |
1117 | #endif | |
1118 | } | |
1119 | kvfree(sbi->devs); | |
1120 | } | |
1121 | ||
1122 | static void f2fs_put_super(struct super_block *sb) | |
1123 | { | |
1124 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1125 | int i; | |
1126 | bool dropped; | |
1127 | ||
1128 | f2fs_quota_off_umount(sb); | |
1129 | ||
1130 | /* prevent remaining shrinker jobs */ | |
1131 | mutex_lock(&sbi->umount_mutex); | |
1132 | ||
1133 | /* | |
1134 | * We don't need to do checkpoint when superblock is clean. | |
1135 | * But, the previous checkpoint was not done by umount, it needs to do | |
1136 | * clean checkpoint again. | |
1137 | */ | |
1138 | if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || | |
1139 | !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { | |
1140 | struct cp_control cpc = { | |
1141 | .reason = CP_UMOUNT, | |
1142 | }; | |
1143 | f2fs_write_checkpoint(sbi, &cpc); | |
1144 | } | |
1145 | ||
1146 | /* be sure to wait for any on-going discard commands */ | |
1147 | dropped = f2fs_issue_discard_timeout(sbi); | |
1148 | ||
1149 | if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && | |
1150 | !sbi->discard_blks && !dropped) { | |
1151 | struct cp_control cpc = { | |
1152 | .reason = CP_UMOUNT | CP_TRIMMED, | |
1153 | }; | |
1154 | f2fs_write_checkpoint(sbi, &cpc); | |
1155 | } | |
1156 | ||
1157 | /* | |
1158 | * normally superblock is clean, so we need to release this. | |
1159 | * In addition, EIO will skip do checkpoint, we need this as well. | |
1160 | */ | |
1161 | f2fs_release_ino_entry(sbi, true); | |
1162 | ||
1163 | f2fs_leave_shrinker(sbi); | |
1164 | mutex_unlock(&sbi->umount_mutex); | |
1165 | ||
1166 | /* our cp_error case, we can wait for any writeback page */ | |
1167 | f2fs_flush_merged_writes(sbi); | |
1168 | ||
1169 | f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); | |
1170 | ||
1171 | f2fs_bug_on(sbi, sbi->fsync_node_num); | |
1172 | ||
1173 | iput(sbi->node_inode); | |
1174 | sbi->node_inode = NULL; | |
1175 | ||
1176 | iput(sbi->meta_inode); | |
1177 | sbi->meta_inode = NULL; | |
1178 | ||
1179 | /* | |
1180 | * iput() can update stat information, if f2fs_write_checkpoint() | |
1181 | * above failed with error. | |
1182 | */ | |
1183 | f2fs_destroy_stats(sbi); | |
1184 | ||
1185 | /* destroy f2fs internal modules */ | |
1186 | f2fs_destroy_node_manager(sbi); | |
1187 | f2fs_destroy_segment_manager(sbi); | |
1188 | ||
1189 | f2fs_destroy_post_read_wq(sbi); | |
1190 | ||
1191 | kvfree(sbi->ckpt); | |
1192 | ||
1193 | f2fs_unregister_sysfs(sbi); | |
1194 | ||
1195 | sb->s_fs_info = NULL; | |
1196 | if (sbi->s_chksum_driver) | |
1197 | crypto_free_shash(sbi->s_chksum_driver); | |
1198 | kvfree(sbi->raw_super); | |
1199 | ||
1200 | destroy_device_list(sbi); | |
1201 | f2fs_destroy_xattr_caches(sbi); | |
1202 | mempool_destroy(sbi->write_io_dummy); | |
1203 | #ifdef CONFIG_QUOTA | |
1204 | for (i = 0; i < MAXQUOTAS; i++) | |
1205 | kvfree(F2FS_OPTION(sbi).s_qf_names[i]); | |
1206 | #endif | |
1207 | destroy_percpu_info(sbi); | |
1208 | for (i = 0; i < NR_PAGE_TYPE; i++) | |
1209 | kvfree(sbi->write_io[i]); | |
1210 | #ifdef CONFIG_UNICODE | |
1211 | utf8_unload(sbi->s_encoding); | |
1212 | #endif | |
1213 | kvfree(sbi); | |
1214 | } | |
1215 | ||
1216 | int f2fs_sync_fs(struct super_block *sb, int sync) | |
1217 | { | |
1218 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1219 | int err = 0; | |
1220 | ||
1221 | if (unlikely(f2fs_cp_error(sbi))) | |
1222 | return 0; | |
1223 | if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) | |
1224 | return 0; | |
1225 | ||
1226 | trace_f2fs_sync_fs(sb, sync); | |
1227 | ||
1228 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1229 | return -EAGAIN; | |
1230 | ||
1231 | if (sync) { | |
1232 | struct cp_control cpc; | |
1233 | ||
1234 | cpc.reason = __get_cp_reason(sbi); | |
1235 | ||
1236 | down_write(&sbi->gc_lock); | |
1237 | err = f2fs_write_checkpoint(sbi, &cpc); | |
1238 | up_write(&sbi->gc_lock); | |
1239 | } | |
1240 | f2fs_trace_ios(NULL, 1); | |
1241 | ||
1242 | return err; | |
1243 | } | |
1244 | ||
1245 | static int f2fs_freeze(struct super_block *sb) | |
1246 | { | |
1247 | if (f2fs_readonly(sb)) | |
1248 | return 0; | |
1249 | ||
1250 | /* IO error happened before */ | |
1251 | if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) | |
1252 | return -EIO; | |
1253 | ||
1254 | /* must be clean, since sync_filesystem() was already called */ | |
1255 | if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) | |
1256 | return -EINVAL; | |
1257 | return 0; | |
1258 | } | |
1259 | ||
1260 | static int f2fs_unfreeze(struct super_block *sb) | |
1261 | { | |
1262 | return 0; | |
1263 | } | |
1264 | ||
1265 | #ifdef CONFIG_QUOTA | |
1266 | static int f2fs_statfs_project(struct super_block *sb, | |
1267 | kprojid_t projid, struct kstatfs *buf) | |
1268 | { | |
1269 | struct kqid qid; | |
1270 | struct dquot *dquot; | |
1271 | u64 limit; | |
1272 | u64 curblock; | |
1273 | ||
1274 | qid = make_kqid_projid(projid); | |
1275 | dquot = dqget(sb, qid); | |
1276 | if (IS_ERR(dquot)) | |
1277 | return PTR_ERR(dquot); | |
1278 | spin_lock(&dquot->dq_dqb_lock); | |
1279 | ||
1280 | limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, | |
1281 | dquot->dq_dqb.dqb_bhardlimit); | |
1282 | if (limit) | |
1283 | limit >>= sb->s_blocksize_bits; | |
1284 | ||
1285 | if (limit && buf->f_blocks > limit) { | |
1286 | curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits; | |
1287 | buf->f_blocks = limit; | |
1288 | buf->f_bfree = buf->f_bavail = | |
1289 | (buf->f_blocks > curblock) ? | |
1290 | (buf->f_blocks - curblock) : 0; | |
1291 | } | |
1292 | ||
1293 | limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, | |
1294 | dquot->dq_dqb.dqb_ihardlimit); | |
1295 | ||
1296 | if (limit && buf->f_files > limit) { | |
1297 | buf->f_files = limit; | |
1298 | buf->f_ffree = | |
1299 | (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? | |
1300 | (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; | |
1301 | } | |
1302 | ||
1303 | spin_unlock(&dquot->dq_dqb_lock); | |
1304 | dqput(dquot); | |
1305 | return 0; | |
1306 | } | |
1307 | #endif | |
1308 | ||
1309 | static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
1310 | { | |
1311 | struct super_block *sb = dentry->d_sb; | |
1312 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1313 | u64 id = huge_encode_dev(sb->s_bdev->bd_dev); | |
1314 | block_t total_count, user_block_count, start_count; | |
1315 | u64 avail_node_count; | |
1316 | ||
1317 | total_count = le64_to_cpu(sbi->raw_super->block_count); | |
1318 | user_block_count = sbi->user_block_count; | |
1319 | start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); | |
1320 | buf->f_type = F2FS_SUPER_MAGIC; | |
1321 | buf->f_bsize = sbi->blocksize; | |
1322 | ||
1323 | buf->f_blocks = total_count - start_count; | |
1324 | buf->f_bfree = user_block_count - valid_user_blocks(sbi) - | |
1325 | sbi->current_reserved_blocks; | |
1326 | ||
1327 | spin_lock(&sbi->stat_lock); | |
1328 | if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) | |
1329 | buf->f_bfree = 0; | |
1330 | else | |
1331 | buf->f_bfree -= sbi->unusable_block_count; | |
1332 | spin_unlock(&sbi->stat_lock); | |
1333 | ||
1334 | if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) | |
1335 | buf->f_bavail = buf->f_bfree - | |
1336 | F2FS_OPTION(sbi).root_reserved_blocks; | |
1337 | else | |
1338 | buf->f_bavail = 0; | |
1339 | ||
1340 | avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; | |
1341 | ||
1342 | if (avail_node_count > user_block_count) { | |
1343 | buf->f_files = user_block_count; | |
1344 | buf->f_ffree = buf->f_bavail; | |
1345 | } else { | |
1346 | buf->f_files = avail_node_count; | |
1347 | buf->f_ffree = min(avail_node_count - valid_node_count(sbi), | |
1348 | buf->f_bavail); | |
1349 | } | |
1350 | ||
1351 | buf->f_namelen = F2FS_NAME_LEN; | |
1352 | buf->f_fsid.val[0] = (u32)id; | |
1353 | buf->f_fsid.val[1] = (u32)(id >> 32); | |
1354 | ||
1355 | #ifdef CONFIG_QUOTA | |
1356 | if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && | |
1357 | sb_has_quota_limits_enabled(sb, PRJQUOTA)) { | |
1358 | f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); | |
1359 | } | |
1360 | #endif | |
1361 | return 0; | |
1362 | } | |
1363 | ||
1364 | static inline void f2fs_show_quota_options(struct seq_file *seq, | |
1365 | struct super_block *sb) | |
1366 | { | |
1367 | #ifdef CONFIG_QUOTA | |
1368 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1369 | ||
1370 | if (F2FS_OPTION(sbi).s_jquota_fmt) { | |
1371 | char *fmtname = ""; | |
1372 | ||
1373 | switch (F2FS_OPTION(sbi).s_jquota_fmt) { | |
1374 | case QFMT_VFS_OLD: | |
1375 | fmtname = "vfsold"; | |
1376 | break; | |
1377 | case QFMT_VFS_V0: | |
1378 | fmtname = "vfsv0"; | |
1379 | break; | |
1380 | case QFMT_VFS_V1: | |
1381 | fmtname = "vfsv1"; | |
1382 | break; | |
1383 | } | |
1384 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
1385 | } | |
1386 | ||
1387 | if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) | |
1388 | seq_show_option(seq, "usrjquota", | |
1389 | F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); | |
1390 | ||
1391 | if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) | |
1392 | seq_show_option(seq, "grpjquota", | |
1393 | F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); | |
1394 | ||
1395 | if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) | |
1396 | seq_show_option(seq, "prjjquota", | |
1397 | F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); | |
1398 | #endif | |
1399 | } | |
1400 | ||
1401 | static inline void f2fs_show_compress_options(struct seq_file *seq, | |
1402 | struct super_block *sb) | |
1403 | { | |
1404 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1405 | char *algtype = ""; | |
1406 | int i; | |
1407 | ||
1408 | if (!f2fs_sb_has_compression(sbi)) | |
1409 | return; | |
1410 | ||
1411 | switch (F2FS_OPTION(sbi).compress_algorithm) { | |
1412 | case COMPRESS_LZO: | |
1413 | algtype = "lzo"; | |
1414 | break; | |
1415 | case COMPRESS_LZ4: | |
1416 | algtype = "lz4"; | |
1417 | break; | |
1418 | case COMPRESS_ZSTD: | |
1419 | algtype = "zstd"; | |
1420 | break; | |
1421 | case COMPRESS_LZORLE: | |
1422 | algtype = "lzo-rle"; | |
1423 | break; | |
1424 | } | |
1425 | seq_printf(seq, ",compress_algorithm=%s", algtype); | |
1426 | ||
1427 | seq_printf(seq, ",compress_log_size=%u", | |
1428 | F2FS_OPTION(sbi).compress_log_size); | |
1429 | ||
1430 | for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { | |
1431 | seq_printf(seq, ",compress_extension=%s", | |
1432 | F2FS_OPTION(sbi).extensions[i]); | |
1433 | } | |
1434 | } | |
1435 | ||
1436 | static int f2fs_show_options(struct seq_file *seq, struct dentry *root) | |
1437 | { | |
1438 | struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); | |
1439 | ||
1440 | if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) | |
1441 | seq_printf(seq, ",background_gc=%s", "sync"); | |
1442 | else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) | |
1443 | seq_printf(seq, ",background_gc=%s", "on"); | |
1444 | else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) | |
1445 | seq_printf(seq, ",background_gc=%s", "off"); | |
1446 | ||
1447 | if (test_opt(sbi, DISABLE_ROLL_FORWARD)) | |
1448 | seq_puts(seq, ",disable_roll_forward"); | |
1449 | if (test_opt(sbi, NORECOVERY)) | |
1450 | seq_puts(seq, ",norecovery"); | |
1451 | if (test_opt(sbi, DISCARD)) | |
1452 | seq_puts(seq, ",discard"); | |
1453 | else | |
1454 | seq_puts(seq, ",nodiscard"); | |
1455 | if (test_opt(sbi, NOHEAP)) | |
1456 | seq_puts(seq, ",no_heap"); | |
1457 | else | |
1458 | seq_puts(seq, ",heap"); | |
1459 | #ifdef CONFIG_F2FS_FS_XATTR | |
1460 | if (test_opt(sbi, XATTR_USER)) | |
1461 | seq_puts(seq, ",user_xattr"); | |
1462 | else | |
1463 | seq_puts(seq, ",nouser_xattr"); | |
1464 | if (test_opt(sbi, INLINE_XATTR)) | |
1465 | seq_puts(seq, ",inline_xattr"); | |
1466 | else | |
1467 | seq_puts(seq, ",noinline_xattr"); | |
1468 | if (test_opt(sbi, INLINE_XATTR_SIZE)) | |
1469 | seq_printf(seq, ",inline_xattr_size=%u", | |
1470 | F2FS_OPTION(sbi).inline_xattr_size); | |
1471 | #endif | |
1472 | #ifdef CONFIG_F2FS_FS_POSIX_ACL | |
1473 | if (test_opt(sbi, POSIX_ACL)) | |
1474 | seq_puts(seq, ",acl"); | |
1475 | else | |
1476 | seq_puts(seq, ",noacl"); | |
1477 | #endif | |
1478 | if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) | |
1479 | seq_puts(seq, ",disable_ext_identify"); | |
1480 | if (test_opt(sbi, INLINE_DATA)) | |
1481 | seq_puts(seq, ",inline_data"); | |
1482 | else | |
1483 | seq_puts(seq, ",noinline_data"); | |
1484 | if (test_opt(sbi, INLINE_DENTRY)) | |
1485 | seq_puts(seq, ",inline_dentry"); | |
1486 | else | |
1487 | seq_puts(seq, ",noinline_dentry"); | |
1488 | if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) | |
1489 | seq_puts(seq, ",flush_merge"); | |
1490 | if (test_opt(sbi, NOBARRIER)) | |
1491 | seq_puts(seq, ",nobarrier"); | |
1492 | if (test_opt(sbi, FASTBOOT)) | |
1493 | seq_puts(seq, ",fastboot"); | |
1494 | if (test_opt(sbi, EXTENT_CACHE)) | |
1495 | seq_puts(seq, ",extent_cache"); | |
1496 | else | |
1497 | seq_puts(seq, ",noextent_cache"); | |
1498 | if (test_opt(sbi, DATA_FLUSH)) | |
1499 | seq_puts(seq, ",data_flush"); | |
1500 | ||
1501 | seq_puts(seq, ",mode="); | |
1502 | if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) | |
1503 | seq_puts(seq, "adaptive"); | |
1504 | else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) | |
1505 | seq_puts(seq, "lfs"); | |
1506 | seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); | |
1507 | if (test_opt(sbi, RESERVE_ROOT)) | |
1508 | seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", | |
1509 | F2FS_OPTION(sbi).root_reserved_blocks, | |
1510 | from_kuid_munged(&init_user_ns, | |
1511 | F2FS_OPTION(sbi).s_resuid), | |
1512 | from_kgid_munged(&init_user_ns, | |
1513 | F2FS_OPTION(sbi).s_resgid)); | |
1514 | if (F2FS_IO_SIZE_BITS(sbi)) | |
1515 | seq_printf(seq, ",io_bits=%u", | |
1516 | F2FS_OPTION(sbi).write_io_size_bits); | |
1517 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
1518 | if (test_opt(sbi, FAULT_INJECTION)) { | |
1519 | seq_printf(seq, ",fault_injection=%u", | |
1520 | F2FS_OPTION(sbi).fault_info.inject_rate); | |
1521 | seq_printf(seq, ",fault_type=%u", | |
1522 | F2FS_OPTION(sbi).fault_info.inject_type); | |
1523 | } | |
1524 | #endif | |
1525 | #ifdef CONFIG_QUOTA | |
1526 | if (test_opt(sbi, QUOTA)) | |
1527 | seq_puts(seq, ",quota"); | |
1528 | if (test_opt(sbi, USRQUOTA)) | |
1529 | seq_puts(seq, ",usrquota"); | |
1530 | if (test_opt(sbi, GRPQUOTA)) | |
1531 | seq_puts(seq, ",grpquota"); | |
1532 | if (test_opt(sbi, PRJQUOTA)) | |
1533 | seq_puts(seq, ",prjquota"); | |
1534 | #endif | |
1535 | f2fs_show_quota_options(seq, sbi->sb); | |
1536 | if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) | |
1537 | seq_printf(seq, ",whint_mode=%s", "user-based"); | |
1538 | else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) | |
1539 | seq_printf(seq, ",whint_mode=%s", "fs-based"); | |
1540 | #ifdef CONFIG_FS_ENCRYPTION | |
1541 | if (F2FS_OPTION(sbi).test_dummy_encryption) | |
1542 | seq_puts(seq, ",test_dummy_encryption"); | |
1543 | #endif | |
1544 | ||
1545 | if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) | |
1546 | seq_printf(seq, ",alloc_mode=%s", "default"); | |
1547 | else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) | |
1548 | seq_printf(seq, ",alloc_mode=%s", "reuse"); | |
1549 | ||
1550 | if (test_opt(sbi, DISABLE_CHECKPOINT)) | |
1551 | seq_printf(seq, ",checkpoint=disable:%u", | |
1552 | F2FS_OPTION(sbi).unusable_cap); | |
1553 | if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) | |
1554 | seq_printf(seq, ",fsync_mode=%s", "posix"); | |
1555 | else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) | |
1556 | seq_printf(seq, ",fsync_mode=%s", "strict"); | |
1557 | else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) | |
1558 | seq_printf(seq, ",fsync_mode=%s", "nobarrier"); | |
1559 | ||
1560 | f2fs_show_compress_options(seq, sbi->sb); | |
1561 | return 0; | |
1562 | } | |
1563 | ||
1564 | static void default_options(struct f2fs_sb_info *sbi) | |
1565 | { | |
1566 | /* init some FS parameters */ | |
1567 | F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE; | |
1568 | F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; | |
1569 | F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; | |
1570 | F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; | |
1571 | F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; | |
1572 | F2FS_OPTION(sbi).test_dummy_encryption = false; | |
1573 | F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); | |
1574 | F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); | |
1575 | F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; | |
1576 | F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; | |
1577 | F2FS_OPTION(sbi).compress_ext_cnt = 0; | |
1578 | F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; | |
1579 | ||
1580 | set_opt(sbi, INLINE_XATTR); | |
1581 | set_opt(sbi, INLINE_DATA); | |
1582 | set_opt(sbi, INLINE_DENTRY); | |
1583 | set_opt(sbi, EXTENT_CACHE); | |
1584 | set_opt(sbi, NOHEAP); | |
1585 | clear_opt(sbi, DISABLE_CHECKPOINT); | |
1586 | F2FS_OPTION(sbi).unusable_cap = 0; | |
1587 | sbi->sb->s_flags |= SB_LAZYTIME; | |
1588 | set_opt(sbi, FLUSH_MERGE); | |
1589 | set_opt(sbi, DISCARD); | |
1590 | if (f2fs_sb_has_blkzoned(sbi)) | |
1591 | F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; | |
1592 | else | |
1593 | F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; | |
1594 | ||
1595 | #ifdef CONFIG_F2FS_FS_XATTR | |
1596 | set_opt(sbi, XATTR_USER); | |
1597 | #endif | |
1598 | #ifdef CONFIG_F2FS_FS_POSIX_ACL | |
1599 | set_opt(sbi, POSIX_ACL); | |
1600 | #endif | |
1601 | ||
1602 | f2fs_build_fault_attr(sbi, 0, 0); | |
1603 | } | |
1604 | ||
1605 | #ifdef CONFIG_QUOTA | |
1606 | static int f2fs_enable_quotas(struct super_block *sb); | |
1607 | #endif | |
1608 | ||
1609 | static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) | |
1610 | { | |
1611 | unsigned int s_flags = sbi->sb->s_flags; | |
1612 | struct cp_control cpc; | |
1613 | int err = 0; | |
1614 | int ret; | |
1615 | block_t unusable; | |
1616 | ||
1617 | if (s_flags & SB_RDONLY) { | |
1618 | f2fs_err(sbi, "checkpoint=disable on readonly fs"); | |
1619 | return -EINVAL; | |
1620 | } | |
1621 | sbi->sb->s_flags |= SB_ACTIVE; | |
1622 | ||
1623 | f2fs_update_time(sbi, DISABLE_TIME); | |
1624 | ||
1625 | while (!f2fs_time_over(sbi, DISABLE_TIME)) { | |
1626 | down_write(&sbi->gc_lock); | |
1627 | err = f2fs_gc(sbi, true, false, NULL_SEGNO); | |
1628 | if (err == -ENODATA) { | |
1629 | err = 0; | |
1630 | break; | |
1631 | } | |
1632 | if (err && err != -EAGAIN) | |
1633 | break; | |
1634 | } | |
1635 | ||
1636 | ret = sync_filesystem(sbi->sb); | |
1637 | if (ret || err) { | |
1638 | err = ret ? ret: err; | |
1639 | goto restore_flag; | |
1640 | } | |
1641 | ||
1642 | unusable = f2fs_get_unusable_blocks(sbi); | |
1643 | if (f2fs_disable_cp_again(sbi, unusable)) { | |
1644 | err = -EAGAIN; | |
1645 | goto restore_flag; | |
1646 | } | |
1647 | ||
1648 | down_write(&sbi->gc_lock); | |
1649 | cpc.reason = CP_PAUSE; | |
1650 | set_sbi_flag(sbi, SBI_CP_DISABLED); | |
1651 | err = f2fs_write_checkpoint(sbi, &cpc); | |
1652 | if (err) | |
1653 | goto out_unlock; | |
1654 | ||
1655 | spin_lock(&sbi->stat_lock); | |
1656 | sbi->unusable_block_count = unusable; | |
1657 | spin_unlock(&sbi->stat_lock); | |
1658 | ||
1659 | out_unlock: | |
1660 | up_write(&sbi->gc_lock); | |
1661 | restore_flag: | |
1662 | sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ | |
1663 | return err; | |
1664 | } | |
1665 | ||
1666 | static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) | |
1667 | { | |
1668 | down_write(&sbi->gc_lock); | |
1669 | f2fs_dirty_to_prefree(sbi); | |
1670 | ||
1671 | clear_sbi_flag(sbi, SBI_CP_DISABLED); | |
1672 | set_sbi_flag(sbi, SBI_IS_DIRTY); | |
1673 | up_write(&sbi->gc_lock); | |
1674 | ||
1675 | f2fs_sync_fs(sbi->sb, 1); | |
1676 | } | |
1677 | ||
1678 | static int f2fs_remount(struct super_block *sb, int *flags, char *data) | |
1679 | { | |
1680 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1681 | struct f2fs_mount_info org_mount_opt; | |
1682 | unsigned long old_sb_flags; | |
1683 | int err; | |
1684 | bool need_restart_gc = false; | |
1685 | bool need_stop_gc = false; | |
1686 | bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); | |
1687 | bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT); | |
1688 | bool no_io_align = !F2FS_IO_ALIGNED(sbi); | |
1689 | bool checkpoint_changed; | |
1690 | #ifdef CONFIG_QUOTA | |
1691 | int i, j; | |
1692 | #endif | |
1693 | ||
1694 | /* | |
1695 | * Save the old mount options in case we | |
1696 | * need to restore them. | |
1697 | */ | |
1698 | org_mount_opt = sbi->mount_opt; | |
1699 | old_sb_flags = sb->s_flags; | |
1700 | ||
1701 | #ifdef CONFIG_QUOTA | |
1702 | org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; | |
1703 | for (i = 0; i < MAXQUOTAS; i++) { | |
1704 | if (F2FS_OPTION(sbi).s_qf_names[i]) { | |
1705 | org_mount_opt.s_qf_names[i] = | |
1706 | kstrdup(F2FS_OPTION(sbi).s_qf_names[i], | |
1707 | GFP_KERNEL); | |
1708 | if (!org_mount_opt.s_qf_names[i]) { | |
1709 | for (j = 0; j < i; j++) | |
1710 | kvfree(org_mount_opt.s_qf_names[j]); | |
1711 | return -ENOMEM; | |
1712 | } | |
1713 | } else { | |
1714 | org_mount_opt.s_qf_names[i] = NULL; | |
1715 | } | |
1716 | } | |
1717 | #endif | |
1718 | ||
1719 | /* recover superblocks we couldn't write due to previous RO mount */ | |
1720 | if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { | |
1721 | err = f2fs_commit_super(sbi, false); | |
1722 | f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", | |
1723 | err); | |
1724 | if (!err) | |
1725 | clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); | |
1726 | } | |
1727 | ||
1728 | default_options(sbi); | |
1729 | ||
1730 | /* parse mount options */ | |
1731 | err = parse_options(sb, data); | |
1732 | if (err) | |
1733 | goto restore_opts; | |
1734 | checkpoint_changed = | |
1735 | disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT); | |
1736 | ||
1737 | /* | |
1738 | * Previous and new state of filesystem is RO, | |
1739 | * so skip checking GC and FLUSH_MERGE conditions. | |
1740 | */ | |
1741 | if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) | |
1742 | goto skip; | |
1743 | ||
1744 | #ifdef CONFIG_QUOTA | |
1745 | if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { | |
1746 | err = dquot_suspend(sb, -1); | |
1747 | if (err < 0) | |
1748 | goto restore_opts; | |
1749 | } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { | |
1750 | /* dquot_resume needs RW */ | |
1751 | sb->s_flags &= ~SB_RDONLY; | |
1752 | if (sb_any_quota_suspended(sb)) { | |
1753 | dquot_resume(sb, -1); | |
1754 | } else if (f2fs_sb_has_quota_ino(sbi)) { | |
1755 | err = f2fs_enable_quotas(sb); | |
1756 | if (err) | |
1757 | goto restore_opts; | |
1758 | } | |
1759 | } | |
1760 | #endif | |
1761 | /* disallow enable/disable extent_cache dynamically */ | |
1762 | if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { | |
1763 | err = -EINVAL; | |
1764 | f2fs_warn(sbi, "switch extent_cache option is not allowed"); | |
1765 | goto restore_opts; | |
1766 | } | |
1767 | ||
1768 | if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { | |
1769 | err = -EINVAL; | |
1770 | f2fs_warn(sbi, "switch io_bits option is not allowed"); | |
1771 | goto restore_opts; | |
1772 | } | |
1773 | ||
1774 | if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { | |
1775 | err = -EINVAL; | |
1776 | f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); | |
1777 | goto restore_opts; | |
1778 | } | |
1779 | ||
1780 | /* | |
1781 | * We stop the GC thread if FS is mounted as RO | |
1782 | * or if background_gc = off is passed in mount | |
1783 | * option. Also sync the filesystem. | |
1784 | */ | |
1785 | if ((*flags & SB_RDONLY) || | |
1786 | F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) { | |
1787 | if (sbi->gc_thread) { | |
1788 | f2fs_stop_gc_thread(sbi); | |
1789 | need_restart_gc = true; | |
1790 | } | |
1791 | } else if (!sbi->gc_thread) { | |
1792 | err = f2fs_start_gc_thread(sbi); | |
1793 | if (err) | |
1794 | goto restore_opts; | |
1795 | need_stop_gc = true; | |
1796 | } | |
1797 | ||
1798 | if (*flags & SB_RDONLY || | |
1799 | F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { | |
1800 | writeback_inodes_sb(sb, WB_REASON_SYNC); | |
1801 | sync_inodes_sb(sb); | |
1802 | ||
1803 | set_sbi_flag(sbi, SBI_IS_DIRTY); | |
1804 | set_sbi_flag(sbi, SBI_IS_CLOSE); | |
1805 | f2fs_sync_fs(sb, 1); | |
1806 | clear_sbi_flag(sbi, SBI_IS_CLOSE); | |
1807 | } | |
1808 | ||
1809 | if (checkpoint_changed) { | |
1810 | if (test_opt(sbi, DISABLE_CHECKPOINT)) { | |
1811 | err = f2fs_disable_checkpoint(sbi); | |
1812 | if (err) | |
1813 | goto restore_gc; | |
1814 | } else { | |
1815 | f2fs_enable_checkpoint(sbi); | |
1816 | } | |
1817 | } | |
1818 | ||
1819 | /* | |
1820 | * We stop issue flush thread if FS is mounted as RO | |
1821 | * or if flush_merge is not passed in mount option. | |
1822 | */ | |
1823 | if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { | |
1824 | clear_opt(sbi, FLUSH_MERGE); | |
1825 | f2fs_destroy_flush_cmd_control(sbi, false); | |
1826 | } else { | |
1827 | err = f2fs_create_flush_cmd_control(sbi); | |
1828 | if (err) | |
1829 | goto restore_gc; | |
1830 | } | |
1831 | skip: | |
1832 | #ifdef CONFIG_QUOTA | |
1833 | /* Release old quota file names */ | |
1834 | for (i = 0; i < MAXQUOTAS; i++) | |
1835 | kvfree(org_mount_opt.s_qf_names[i]); | |
1836 | #endif | |
1837 | /* Update the POSIXACL Flag */ | |
1838 | sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | | |
1839 | (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); | |
1840 | ||
1841 | limit_reserve_root(sbi); | |
1842 | *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); | |
1843 | return 0; | |
1844 | restore_gc: | |
1845 | if (need_restart_gc) { | |
1846 | if (f2fs_start_gc_thread(sbi)) | |
1847 | f2fs_warn(sbi, "background gc thread has stopped"); | |
1848 | } else if (need_stop_gc) { | |
1849 | f2fs_stop_gc_thread(sbi); | |
1850 | } | |
1851 | restore_opts: | |
1852 | #ifdef CONFIG_QUOTA | |
1853 | F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; | |
1854 | for (i = 0; i < MAXQUOTAS; i++) { | |
1855 | kvfree(F2FS_OPTION(sbi).s_qf_names[i]); | |
1856 | F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; | |
1857 | } | |
1858 | #endif | |
1859 | sbi->mount_opt = org_mount_opt; | |
1860 | sb->s_flags = old_sb_flags; | |
1861 | return err; | |
1862 | } | |
1863 | ||
1864 | #ifdef CONFIG_QUOTA | |
1865 | /* Read data from quotafile */ | |
1866 | static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, | |
1867 | size_t len, loff_t off) | |
1868 | { | |
1869 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
1870 | struct address_space *mapping = inode->i_mapping; | |
1871 | block_t blkidx = F2FS_BYTES_TO_BLK(off); | |
1872 | int offset = off & (sb->s_blocksize - 1); | |
1873 | int tocopy; | |
1874 | size_t toread; | |
1875 | loff_t i_size = i_size_read(inode); | |
1876 | struct page *page; | |
1877 | char *kaddr; | |
1878 | ||
1879 | if (off > i_size) | |
1880 | return 0; | |
1881 | ||
1882 | if (off + len > i_size) | |
1883 | len = i_size - off; | |
1884 | toread = len; | |
1885 | while (toread > 0) { | |
1886 | tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); | |
1887 | repeat: | |
1888 | page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); | |
1889 | if (IS_ERR(page)) { | |
1890 | if (PTR_ERR(page) == -ENOMEM) { | |
1891 | congestion_wait(BLK_RW_ASYNC, | |
1892 | DEFAULT_IO_TIMEOUT); | |
1893 | goto repeat; | |
1894 | } | |
1895 | set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); | |
1896 | return PTR_ERR(page); | |
1897 | } | |
1898 | ||
1899 | lock_page(page); | |
1900 | ||
1901 | if (unlikely(page->mapping != mapping)) { | |
1902 | f2fs_put_page(page, 1); | |
1903 | goto repeat; | |
1904 | } | |
1905 | if (unlikely(!PageUptodate(page))) { | |
1906 | f2fs_put_page(page, 1); | |
1907 | set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); | |
1908 | return -EIO; | |
1909 | } | |
1910 | ||
1911 | kaddr = kmap_atomic(page); | |
1912 | memcpy(data, kaddr + offset, tocopy); | |
1913 | kunmap_atomic(kaddr); | |
1914 | f2fs_put_page(page, 1); | |
1915 | ||
1916 | offset = 0; | |
1917 | toread -= tocopy; | |
1918 | data += tocopy; | |
1919 | blkidx++; | |
1920 | } | |
1921 | return len; | |
1922 | } | |
1923 | ||
1924 | /* Write to quotafile */ | |
1925 | static ssize_t f2fs_quota_write(struct super_block *sb, int type, | |
1926 | const char *data, size_t len, loff_t off) | |
1927 | { | |
1928 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
1929 | struct address_space *mapping = inode->i_mapping; | |
1930 | const struct address_space_operations *a_ops = mapping->a_ops; | |
1931 | int offset = off & (sb->s_blocksize - 1); | |
1932 | size_t towrite = len; | |
1933 | struct page *page; | |
1934 | void *fsdata = NULL; | |
1935 | char *kaddr; | |
1936 | int err = 0; | |
1937 | int tocopy; | |
1938 | ||
1939 | while (towrite > 0) { | |
1940 | tocopy = min_t(unsigned long, sb->s_blocksize - offset, | |
1941 | towrite); | |
1942 | retry: | |
1943 | err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, | |
1944 | &page, &fsdata); | |
1945 | if (unlikely(err)) { | |
1946 | if (err == -ENOMEM) { | |
1947 | congestion_wait(BLK_RW_ASYNC, | |
1948 | DEFAULT_IO_TIMEOUT); | |
1949 | goto retry; | |
1950 | } | |
1951 | set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); | |
1952 | break; | |
1953 | } | |
1954 | ||
1955 | kaddr = kmap_atomic(page); | |
1956 | memcpy(kaddr + offset, data, tocopy); | |
1957 | kunmap_atomic(kaddr); | |
1958 | flush_dcache_page(page); | |
1959 | ||
1960 | a_ops->write_end(NULL, mapping, off, tocopy, tocopy, | |
1961 | page, fsdata); | |
1962 | offset = 0; | |
1963 | towrite -= tocopy; | |
1964 | off += tocopy; | |
1965 | data += tocopy; | |
1966 | cond_resched(); | |
1967 | } | |
1968 | ||
1969 | if (len == towrite) | |
1970 | return err; | |
1971 | inode->i_mtime = inode->i_ctime = current_time(inode); | |
1972 | f2fs_mark_inode_dirty_sync(inode, false); | |
1973 | return len - towrite; | |
1974 | } | |
1975 | ||
1976 | static struct dquot **f2fs_get_dquots(struct inode *inode) | |
1977 | { | |
1978 | return F2FS_I(inode)->i_dquot; | |
1979 | } | |
1980 | ||
1981 | static qsize_t *f2fs_get_reserved_space(struct inode *inode) | |
1982 | { | |
1983 | return &F2FS_I(inode)->i_reserved_quota; | |
1984 | } | |
1985 | ||
1986 | static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) | |
1987 | { | |
1988 | if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { | |
1989 | f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); | |
1990 | return 0; | |
1991 | } | |
1992 | ||
1993 | return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], | |
1994 | F2FS_OPTION(sbi).s_jquota_fmt, type); | |
1995 | } | |
1996 | ||
1997 | int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) | |
1998 | { | |
1999 | int enabled = 0; | |
2000 | int i, err; | |
2001 | ||
2002 | if (f2fs_sb_has_quota_ino(sbi) && rdonly) { | |
2003 | err = f2fs_enable_quotas(sbi->sb); | |
2004 | if (err) { | |
2005 | f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); | |
2006 | return 0; | |
2007 | } | |
2008 | return 1; | |
2009 | } | |
2010 | ||
2011 | for (i = 0; i < MAXQUOTAS; i++) { | |
2012 | if (F2FS_OPTION(sbi).s_qf_names[i]) { | |
2013 | err = f2fs_quota_on_mount(sbi, i); | |
2014 | if (!err) { | |
2015 | enabled = 1; | |
2016 | continue; | |
2017 | } | |
2018 | f2fs_err(sbi, "Cannot turn on quotas: %d on %d", | |
2019 | err, i); | |
2020 | } | |
2021 | } | |
2022 | return enabled; | |
2023 | } | |
2024 | ||
2025 | static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, | |
2026 | unsigned int flags) | |
2027 | { | |
2028 | struct inode *qf_inode; | |
2029 | unsigned long qf_inum; | |
2030 | int err; | |
2031 | ||
2032 | BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); | |
2033 | ||
2034 | qf_inum = f2fs_qf_ino(sb, type); | |
2035 | if (!qf_inum) | |
2036 | return -EPERM; | |
2037 | ||
2038 | qf_inode = f2fs_iget(sb, qf_inum); | |
2039 | if (IS_ERR(qf_inode)) { | |
2040 | f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); | |
2041 | return PTR_ERR(qf_inode); | |
2042 | } | |
2043 | ||
2044 | /* Don't account quota for quota files to avoid recursion */ | |
2045 | qf_inode->i_flags |= S_NOQUOTA; | |
2046 | err = dquot_load_quota_inode(qf_inode, type, format_id, flags); | |
2047 | iput(qf_inode); | |
2048 | return err; | |
2049 | } | |
2050 | ||
2051 | static int f2fs_enable_quotas(struct super_block *sb) | |
2052 | { | |
2053 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
2054 | int type, err = 0; | |
2055 | unsigned long qf_inum; | |
2056 | bool quota_mopt[MAXQUOTAS] = { | |
2057 | test_opt(sbi, USRQUOTA), | |
2058 | test_opt(sbi, GRPQUOTA), | |
2059 | test_opt(sbi, PRJQUOTA), | |
2060 | }; | |
2061 | ||
2062 | if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { | |
2063 | f2fs_err(sbi, "quota file may be corrupted, skip loading it"); | |
2064 | return 0; | |
2065 | } | |
2066 | ||
2067 | sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; | |
2068 | ||
2069 | for (type = 0; type < MAXQUOTAS; type++) { | |
2070 | qf_inum = f2fs_qf_ino(sb, type); | |
2071 | if (qf_inum) { | |
2072 | err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, | |
2073 | DQUOT_USAGE_ENABLED | | |
2074 | (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); | |
2075 | if (err) { | |
2076 | f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", | |
2077 | type, err); | |
2078 | for (type--; type >= 0; type--) | |
2079 | dquot_quota_off(sb, type); | |
2080 | set_sbi_flag(F2FS_SB(sb), | |
2081 | SBI_QUOTA_NEED_REPAIR); | |
2082 | return err; | |
2083 | } | |
2084 | } | |
2085 | } | |
2086 | return 0; | |
2087 | } | |
2088 | ||
2089 | int f2fs_quota_sync(struct super_block *sb, int type) | |
2090 | { | |
2091 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
2092 | struct quota_info *dqopt = sb_dqopt(sb); | |
2093 | int cnt; | |
2094 | int ret; | |
2095 | ||
2096 | /* | |
2097 | * do_quotactl | |
2098 | * f2fs_quota_sync | |
2099 | * down_read(quota_sem) | |
2100 | * dquot_writeback_dquots() | |
2101 | * f2fs_dquot_commit | |
2102 | * block_operation | |
2103 | * down_read(quota_sem) | |
2104 | */ | |
2105 | f2fs_lock_op(sbi); | |
2106 | ||
2107 | down_read(&sbi->quota_sem); | |
2108 | ret = dquot_writeback_dquots(sb, type); | |
2109 | if (ret) | |
2110 | goto out; | |
2111 | ||
2112 | /* | |
2113 | * Now when everything is written we can discard the pagecache so | |
2114 | * that userspace sees the changes. | |
2115 | */ | |
2116 | for (cnt = 0; cnt < MAXQUOTAS; cnt++) { | |
2117 | struct address_space *mapping; | |
2118 | ||
2119 | if (type != -1 && cnt != type) | |
2120 | continue; | |
2121 | if (!sb_has_quota_active(sb, cnt)) | |
2122 | continue; | |
2123 | ||
2124 | mapping = dqopt->files[cnt]->i_mapping; | |
2125 | ||
2126 | ret = filemap_fdatawrite(mapping); | |
2127 | if (ret) | |
2128 | goto out; | |
2129 | ||
2130 | /* if we are using journalled quota */ | |
2131 | if (is_journalled_quota(sbi)) | |
2132 | continue; | |
2133 | ||
2134 | ret = filemap_fdatawait(mapping); | |
2135 | if (ret) | |
2136 | set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); | |
2137 | ||
2138 | inode_lock(dqopt->files[cnt]); | |
2139 | truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); | |
2140 | inode_unlock(dqopt->files[cnt]); | |
2141 | } | |
2142 | out: | |
2143 | if (ret) | |
2144 | set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); | |
2145 | up_read(&sbi->quota_sem); | |
2146 | f2fs_unlock_op(sbi); | |
2147 | return ret; | |
2148 | } | |
2149 | ||
2150 | static int f2fs_quota_on(struct super_block *sb, int type, int format_id, | |
2151 | const struct path *path) | |
2152 | { | |
2153 | struct inode *inode; | |
2154 | int err; | |
2155 | ||
2156 | /* if quota sysfile exists, deny enabling quota with specific file */ | |
2157 | if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { | |
2158 | f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); | |
2159 | return -EBUSY; | |
2160 | } | |
2161 | ||
2162 | err = f2fs_quota_sync(sb, type); | |
2163 | if (err) | |
2164 | return err; | |
2165 | ||
2166 | err = dquot_quota_on(sb, type, format_id, path); | |
2167 | if (err) | |
2168 | return err; | |
2169 | ||
2170 | inode = d_inode(path->dentry); | |
2171 | ||
2172 | inode_lock(inode); | |
2173 | F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; | |
2174 | f2fs_set_inode_flags(inode); | |
2175 | inode_unlock(inode); | |
2176 | f2fs_mark_inode_dirty_sync(inode, false); | |
2177 | ||
2178 | return 0; | |
2179 | } | |
2180 | ||
2181 | static int __f2fs_quota_off(struct super_block *sb, int type) | |
2182 | { | |
2183 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
2184 | int err; | |
2185 | ||
2186 | if (!inode || !igrab(inode)) | |
2187 | return dquot_quota_off(sb, type); | |
2188 | ||
2189 | err = f2fs_quota_sync(sb, type); | |
2190 | if (err) | |
2191 | goto out_put; | |
2192 | ||
2193 | err = dquot_quota_off(sb, type); | |
2194 | if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) | |
2195 | goto out_put; | |
2196 | ||
2197 | inode_lock(inode); | |
2198 | F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); | |
2199 | f2fs_set_inode_flags(inode); | |
2200 | inode_unlock(inode); | |
2201 | f2fs_mark_inode_dirty_sync(inode, false); | |
2202 | out_put: | |
2203 | iput(inode); | |
2204 | return err; | |
2205 | } | |
2206 | ||
2207 | static int f2fs_quota_off(struct super_block *sb, int type) | |
2208 | { | |
2209 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
2210 | int err; | |
2211 | ||
2212 | err = __f2fs_quota_off(sb, type); | |
2213 | ||
2214 | /* | |
2215 | * quotactl can shutdown journalled quota, result in inconsistence | |
2216 | * between quota record and fs data by following updates, tag the | |
2217 | * flag to let fsck be aware of it. | |
2218 | */ | |
2219 | if (is_journalled_quota(sbi)) | |
2220 | set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); | |
2221 | return err; | |
2222 | } | |
2223 | ||
2224 | void f2fs_quota_off_umount(struct super_block *sb) | |
2225 | { | |
2226 | int type; | |
2227 | int err; | |
2228 | ||
2229 | for (type = 0; type < MAXQUOTAS; type++) { | |
2230 | err = __f2fs_quota_off(sb, type); | |
2231 | if (err) { | |
2232 | int ret = dquot_quota_off(sb, type); | |
2233 | ||
2234 | f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", | |
2235 | type, err, ret); | |
2236 | set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); | |
2237 | } | |
2238 | } | |
2239 | /* | |
2240 | * In case of checkpoint=disable, we must flush quota blocks. | |
2241 | * This can cause NULL exception for node_inode in end_io, since | |
2242 | * put_super already dropped it. | |
2243 | */ | |
2244 | sync_filesystem(sb); | |
2245 | } | |
2246 | ||
2247 | static void f2fs_truncate_quota_inode_pages(struct super_block *sb) | |
2248 | { | |
2249 | struct quota_info *dqopt = sb_dqopt(sb); | |
2250 | int type; | |
2251 | ||
2252 | for (type = 0; type < MAXQUOTAS; type++) { | |
2253 | if (!dqopt->files[type]) | |
2254 | continue; | |
2255 | f2fs_inode_synced(dqopt->files[type]); | |
2256 | } | |
2257 | } | |
2258 | ||
2259 | static int f2fs_dquot_commit(struct dquot *dquot) | |
2260 | { | |
2261 | struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); | |
2262 | int ret; | |
2263 | ||
2264 | down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); | |
2265 | ret = dquot_commit(dquot); | |
2266 | if (ret < 0) | |
2267 | set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); | |
2268 | up_read(&sbi->quota_sem); | |
2269 | return ret; | |
2270 | } | |
2271 | ||
2272 | static int f2fs_dquot_acquire(struct dquot *dquot) | |
2273 | { | |
2274 | struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); | |
2275 | int ret; | |
2276 | ||
2277 | down_read(&sbi->quota_sem); | |
2278 | ret = dquot_acquire(dquot); | |
2279 | if (ret < 0) | |
2280 | set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); | |
2281 | up_read(&sbi->quota_sem); | |
2282 | return ret; | |
2283 | } | |
2284 | ||
2285 | static int f2fs_dquot_release(struct dquot *dquot) | |
2286 | { | |
2287 | struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); | |
2288 | int ret = dquot_release(dquot); | |
2289 | ||
2290 | if (ret < 0) | |
2291 | set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); | |
2292 | return ret; | |
2293 | } | |
2294 | ||
2295 | static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) | |
2296 | { | |
2297 | struct super_block *sb = dquot->dq_sb; | |
2298 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
2299 | int ret = dquot_mark_dquot_dirty(dquot); | |
2300 | ||
2301 | /* if we are using journalled quota */ | |
2302 | if (is_journalled_quota(sbi)) | |
2303 | set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); | |
2304 | ||
2305 | return ret; | |
2306 | } | |
2307 | ||
2308 | static int f2fs_dquot_commit_info(struct super_block *sb, int type) | |
2309 | { | |
2310 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
2311 | int ret = dquot_commit_info(sb, type); | |
2312 | ||
2313 | if (ret < 0) | |
2314 | set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); | |
2315 | return ret; | |
2316 | } | |
2317 | ||
2318 | static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) | |
2319 | { | |
2320 | *projid = F2FS_I(inode)->i_projid; | |
2321 | return 0; | |
2322 | } | |
2323 | ||
2324 | static const struct dquot_operations f2fs_quota_operations = { | |
2325 | .get_reserved_space = f2fs_get_reserved_space, | |
2326 | .write_dquot = f2fs_dquot_commit, | |
2327 | .acquire_dquot = f2fs_dquot_acquire, | |
2328 | .release_dquot = f2fs_dquot_release, | |
2329 | .mark_dirty = f2fs_dquot_mark_dquot_dirty, | |
2330 | .write_info = f2fs_dquot_commit_info, | |
2331 | .alloc_dquot = dquot_alloc, | |
2332 | .destroy_dquot = dquot_destroy, | |
2333 | .get_projid = f2fs_get_projid, | |
2334 | .get_next_id = dquot_get_next_id, | |
2335 | }; | |
2336 | ||
2337 | static const struct quotactl_ops f2fs_quotactl_ops = { | |
2338 | .quota_on = f2fs_quota_on, | |
2339 | .quota_off = f2fs_quota_off, | |
2340 | .quota_sync = f2fs_quota_sync, | |
2341 | .get_state = dquot_get_state, | |
2342 | .set_info = dquot_set_dqinfo, | |
2343 | .get_dqblk = dquot_get_dqblk, | |
2344 | .set_dqblk = dquot_set_dqblk, | |
2345 | .get_nextdqblk = dquot_get_next_dqblk, | |
2346 | }; | |
2347 | #else | |
2348 | int f2fs_quota_sync(struct super_block *sb, int type) | |
2349 | { | |
2350 | return 0; | |
2351 | } | |
2352 | ||
2353 | void f2fs_quota_off_umount(struct super_block *sb) | |
2354 | { | |
2355 | } | |
2356 | #endif | |
2357 | ||
2358 | static const struct super_operations f2fs_sops = { | |
2359 | .alloc_inode = f2fs_alloc_inode, | |
2360 | .free_inode = f2fs_free_inode, | |
2361 | .drop_inode = f2fs_drop_inode, | |
2362 | .write_inode = f2fs_write_inode, | |
2363 | .dirty_inode = f2fs_dirty_inode, | |
2364 | .show_options = f2fs_show_options, | |
2365 | #ifdef CONFIG_QUOTA | |
2366 | .quota_read = f2fs_quota_read, | |
2367 | .quota_write = f2fs_quota_write, | |
2368 | .get_dquots = f2fs_get_dquots, | |
2369 | #endif | |
2370 | .evict_inode = f2fs_evict_inode, | |
2371 | .put_super = f2fs_put_super, | |
2372 | .sync_fs = f2fs_sync_fs, | |
2373 | .freeze_fs = f2fs_freeze, | |
2374 | .unfreeze_fs = f2fs_unfreeze, | |
2375 | .statfs = f2fs_statfs, | |
2376 | .remount_fs = f2fs_remount, | |
2377 | }; | |
2378 | ||
2379 | #ifdef CONFIG_FS_ENCRYPTION | |
2380 | static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) | |
2381 | { | |
2382 | return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, | |
2383 | F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, | |
2384 | ctx, len, NULL); | |
2385 | } | |
2386 | ||
2387 | static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, | |
2388 | void *fs_data) | |
2389 | { | |
2390 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
2391 | ||
2392 | /* | |
2393 | * Encrypting the root directory is not allowed because fsck | |
2394 | * expects lost+found directory to exist and remain unencrypted | |
2395 | * if LOST_FOUND feature is enabled. | |
2396 | * | |
2397 | */ | |
2398 | if (f2fs_sb_has_lost_found(sbi) && | |
2399 | inode->i_ino == F2FS_ROOT_INO(sbi)) | |
2400 | return -EPERM; | |
2401 | ||
2402 | return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, | |
2403 | F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, | |
2404 | ctx, len, fs_data, XATTR_CREATE); | |
2405 | } | |
2406 | ||
2407 | static bool f2fs_dummy_context(struct inode *inode) | |
2408 | { | |
2409 | return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode)); | |
2410 | } | |
2411 | ||
2412 | static bool f2fs_has_stable_inodes(struct super_block *sb) | |
2413 | { | |
2414 | return true; | |
2415 | } | |
2416 | ||
2417 | static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, | |
2418 | int *ino_bits_ret, int *lblk_bits_ret) | |
2419 | { | |
2420 | *ino_bits_ret = 8 * sizeof(nid_t); | |
2421 | *lblk_bits_ret = 8 * sizeof(block_t); | |
2422 | } | |
2423 | ||
2424 | static const struct fscrypt_operations f2fs_cryptops = { | |
2425 | .key_prefix = "f2fs:", | |
2426 | .get_context = f2fs_get_context, | |
2427 | .set_context = f2fs_set_context, | |
2428 | .dummy_context = f2fs_dummy_context, | |
2429 | .empty_dir = f2fs_empty_dir, | |
2430 | .max_namelen = F2FS_NAME_LEN, | |
2431 | .has_stable_inodes = f2fs_has_stable_inodes, | |
2432 | .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, | |
2433 | }; | |
2434 | #endif | |
2435 | ||
2436 | static struct inode *f2fs_nfs_get_inode(struct super_block *sb, | |
2437 | u64 ino, u32 generation) | |
2438 | { | |
2439 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
2440 | struct inode *inode; | |
2441 | ||
2442 | if (f2fs_check_nid_range(sbi, ino)) | |
2443 | return ERR_PTR(-ESTALE); | |
2444 | ||
2445 | /* | |
2446 | * f2fs_iget isn't quite right if the inode is currently unallocated! | |
2447 | * However f2fs_iget currently does appropriate checks to handle stale | |
2448 | * inodes so everything is OK. | |
2449 | */ | |
2450 | inode = f2fs_iget(sb, ino); | |
2451 | if (IS_ERR(inode)) | |
2452 | return ERR_CAST(inode); | |
2453 | if (unlikely(generation && inode->i_generation != generation)) { | |
2454 | /* we didn't find the right inode.. */ | |
2455 | iput(inode); | |
2456 | return ERR_PTR(-ESTALE); | |
2457 | } | |
2458 | return inode; | |
2459 | } | |
2460 | ||
2461 | static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
2462 | int fh_len, int fh_type) | |
2463 | { | |
2464 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
2465 | f2fs_nfs_get_inode); | |
2466 | } | |
2467 | ||
2468 | static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, | |
2469 | int fh_len, int fh_type) | |
2470 | { | |
2471 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
2472 | f2fs_nfs_get_inode); | |
2473 | } | |
2474 | ||
2475 | static const struct export_operations f2fs_export_ops = { | |
2476 | .fh_to_dentry = f2fs_fh_to_dentry, | |
2477 | .fh_to_parent = f2fs_fh_to_parent, | |
2478 | .get_parent = f2fs_get_parent, | |
2479 | }; | |
2480 | ||
2481 | static loff_t max_file_blocks(void) | |
2482 | { | |
2483 | loff_t result = 0; | |
2484 | loff_t leaf_count = DEF_ADDRS_PER_BLOCK; | |
2485 | ||
2486 | /* | |
2487 | * note: previously, result is equal to (DEF_ADDRS_PER_INODE - | |
2488 | * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more | |
2489 | * space in inode.i_addr, it will be more safe to reassign | |
2490 | * result as zero. | |
2491 | */ | |
2492 | ||
2493 | /* two direct node blocks */ | |
2494 | result += (leaf_count * 2); | |
2495 | ||
2496 | /* two indirect node blocks */ | |
2497 | leaf_count *= NIDS_PER_BLOCK; | |
2498 | result += (leaf_count * 2); | |
2499 | ||
2500 | /* one double indirect node block */ | |
2501 | leaf_count *= NIDS_PER_BLOCK; | |
2502 | result += leaf_count; | |
2503 | ||
2504 | return result; | |
2505 | } | |
2506 | ||
2507 | static int __f2fs_commit_super(struct buffer_head *bh, | |
2508 | struct f2fs_super_block *super) | |
2509 | { | |
2510 | lock_buffer(bh); | |
2511 | if (super) | |
2512 | memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); | |
2513 | set_buffer_dirty(bh); | |
2514 | unlock_buffer(bh); | |
2515 | ||
2516 | /* it's rare case, we can do fua all the time */ | |
2517 | return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); | |
2518 | } | |
2519 | ||
2520 | static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, | |
2521 | struct buffer_head *bh) | |
2522 | { | |
2523 | struct f2fs_super_block *raw_super = (struct f2fs_super_block *) | |
2524 | (bh->b_data + F2FS_SUPER_OFFSET); | |
2525 | struct super_block *sb = sbi->sb; | |
2526 | u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); | |
2527 | u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); | |
2528 | u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); | |
2529 | u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); | |
2530 | u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); | |
2531 | u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); | |
2532 | u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); | |
2533 | u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); | |
2534 | u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); | |
2535 | u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); | |
2536 | u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); | |
2537 | u32 segment_count = le32_to_cpu(raw_super->segment_count); | |
2538 | u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); | |
2539 | u64 main_end_blkaddr = main_blkaddr + | |
2540 | (segment_count_main << log_blocks_per_seg); | |
2541 | u64 seg_end_blkaddr = segment0_blkaddr + | |
2542 | (segment_count << log_blocks_per_seg); | |
2543 | ||
2544 | if (segment0_blkaddr != cp_blkaddr) { | |
2545 | f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", | |
2546 | segment0_blkaddr, cp_blkaddr); | |
2547 | return true; | |
2548 | } | |
2549 | ||
2550 | if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != | |
2551 | sit_blkaddr) { | |
2552 | f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", | |
2553 | cp_blkaddr, sit_blkaddr, | |
2554 | segment_count_ckpt << log_blocks_per_seg); | |
2555 | return true; | |
2556 | } | |
2557 | ||
2558 | if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != | |
2559 | nat_blkaddr) { | |
2560 | f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", | |
2561 | sit_blkaddr, nat_blkaddr, | |
2562 | segment_count_sit << log_blocks_per_seg); | |
2563 | return true; | |
2564 | } | |
2565 | ||
2566 | if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != | |
2567 | ssa_blkaddr) { | |
2568 | f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", | |
2569 | nat_blkaddr, ssa_blkaddr, | |
2570 | segment_count_nat << log_blocks_per_seg); | |
2571 | return true; | |
2572 | } | |
2573 | ||
2574 | if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != | |
2575 | main_blkaddr) { | |
2576 | f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", | |
2577 | ssa_blkaddr, main_blkaddr, | |
2578 | segment_count_ssa << log_blocks_per_seg); | |
2579 | return true; | |
2580 | } | |
2581 | ||
2582 | if (main_end_blkaddr > seg_end_blkaddr) { | |
2583 | f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)", | |
2584 | main_blkaddr, | |
2585 | segment0_blkaddr + | |
2586 | (segment_count << log_blocks_per_seg), | |
2587 | segment_count_main << log_blocks_per_seg); | |
2588 | return true; | |
2589 | } else if (main_end_blkaddr < seg_end_blkaddr) { | |
2590 | int err = 0; | |
2591 | char *res; | |
2592 | ||
2593 | /* fix in-memory information all the time */ | |
2594 | raw_super->segment_count = cpu_to_le32((main_end_blkaddr - | |
2595 | segment0_blkaddr) >> log_blocks_per_seg); | |
2596 | ||
2597 | if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { | |
2598 | set_sbi_flag(sbi, SBI_NEED_SB_WRITE); | |
2599 | res = "internally"; | |
2600 | } else { | |
2601 | err = __f2fs_commit_super(bh, NULL); | |
2602 | res = err ? "failed" : "done"; | |
2603 | } | |
2604 | f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)", | |
2605 | res, main_blkaddr, | |
2606 | segment0_blkaddr + | |
2607 | (segment_count << log_blocks_per_seg), | |
2608 | segment_count_main << log_blocks_per_seg); | |
2609 | if (err) | |
2610 | return true; | |
2611 | } | |
2612 | return false; | |
2613 | } | |
2614 | ||
2615 | static int sanity_check_raw_super(struct f2fs_sb_info *sbi, | |
2616 | struct buffer_head *bh) | |
2617 | { | |
2618 | block_t segment_count, segs_per_sec, secs_per_zone; | |
2619 | block_t total_sections, blocks_per_seg; | |
2620 | struct f2fs_super_block *raw_super = (struct f2fs_super_block *) | |
2621 | (bh->b_data + F2FS_SUPER_OFFSET); | |
2622 | unsigned int blocksize; | |
2623 | size_t crc_offset = 0; | |
2624 | __u32 crc = 0; | |
2625 | ||
2626 | if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { | |
2627 | f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", | |
2628 | F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); | |
2629 | return -EINVAL; | |
2630 | } | |
2631 | ||
2632 | /* Check checksum_offset and crc in superblock */ | |
2633 | if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { | |
2634 | crc_offset = le32_to_cpu(raw_super->checksum_offset); | |
2635 | if (crc_offset != | |
2636 | offsetof(struct f2fs_super_block, crc)) { | |
2637 | f2fs_info(sbi, "Invalid SB checksum offset: %zu", | |
2638 | crc_offset); | |
2639 | return -EFSCORRUPTED; | |
2640 | } | |
2641 | crc = le32_to_cpu(raw_super->crc); | |
2642 | if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { | |
2643 | f2fs_info(sbi, "Invalid SB checksum value: %u", crc); | |
2644 | return -EFSCORRUPTED; | |
2645 | } | |
2646 | } | |
2647 | ||
2648 | /* Currently, support only 4KB page cache size */ | |
2649 | if (F2FS_BLKSIZE != PAGE_SIZE) { | |
2650 | f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB", | |
2651 | PAGE_SIZE); | |
2652 | return -EFSCORRUPTED; | |
2653 | } | |
2654 | ||
2655 | /* Currently, support only 4KB block size */ | |
2656 | blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); | |
2657 | if (blocksize != F2FS_BLKSIZE) { | |
2658 | f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB", | |
2659 | blocksize); | |
2660 | return -EFSCORRUPTED; | |
2661 | } | |
2662 | ||
2663 | /* check log blocks per segment */ | |
2664 | if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { | |
2665 | f2fs_info(sbi, "Invalid log blocks per segment (%u)", | |
2666 | le32_to_cpu(raw_super->log_blocks_per_seg)); | |
2667 | return -EFSCORRUPTED; | |
2668 | } | |
2669 | ||
2670 | /* Currently, support 512/1024/2048/4096 bytes sector size */ | |
2671 | if (le32_to_cpu(raw_super->log_sectorsize) > | |
2672 | F2FS_MAX_LOG_SECTOR_SIZE || | |
2673 | le32_to_cpu(raw_super->log_sectorsize) < | |
2674 | F2FS_MIN_LOG_SECTOR_SIZE) { | |
2675 | f2fs_info(sbi, "Invalid log sectorsize (%u)", | |
2676 | le32_to_cpu(raw_super->log_sectorsize)); | |
2677 | return -EFSCORRUPTED; | |
2678 | } | |
2679 | if (le32_to_cpu(raw_super->log_sectors_per_block) + | |
2680 | le32_to_cpu(raw_super->log_sectorsize) != | |
2681 | F2FS_MAX_LOG_SECTOR_SIZE) { | |
2682 | f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", | |
2683 | le32_to_cpu(raw_super->log_sectors_per_block), | |
2684 | le32_to_cpu(raw_super->log_sectorsize)); | |
2685 | return -EFSCORRUPTED; | |
2686 | } | |
2687 | ||
2688 | segment_count = le32_to_cpu(raw_super->segment_count); | |
2689 | segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); | |
2690 | secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); | |
2691 | total_sections = le32_to_cpu(raw_super->section_count); | |
2692 | ||
2693 | /* blocks_per_seg should be 512, given the above check */ | |
2694 | blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); | |
2695 | ||
2696 | if (segment_count > F2FS_MAX_SEGMENT || | |
2697 | segment_count < F2FS_MIN_SEGMENTS) { | |
2698 | f2fs_info(sbi, "Invalid segment count (%u)", segment_count); | |
2699 | return -EFSCORRUPTED; | |
2700 | } | |
2701 | ||
2702 | if (total_sections > segment_count || | |
2703 | total_sections < F2FS_MIN_SEGMENTS || | |
2704 | segs_per_sec > segment_count || !segs_per_sec) { | |
2705 | f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", | |
2706 | segment_count, total_sections, segs_per_sec); | |
2707 | return -EFSCORRUPTED; | |
2708 | } | |
2709 | ||
2710 | if ((segment_count / segs_per_sec) < total_sections) { | |
2711 | f2fs_info(sbi, "Small segment_count (%u < %u * %u)", | |
2712 | segment_count, segs_per_sec, total_sections); | |
2713 | return -EFSCORRUPTED; | |
2714 | } | |
2715 | ||
2716 | if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { | |
2717 | f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", | |
2718 | segment_count, le64_to_cpu(raw_super->block_count)); | |
2719 | return -EFSCORRUPTED; | |
2720 | } | |
2721 | ||
2722 | if (RDEV(0).path[0]) { | |
2723 | block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); | |
2724 | int i = 1; | |
2725 | ||
2726 | while (i < MAX_DEVICES && RDEV(i).path[0]) { | |
2727 | dev_seg_count += le32_to_cpu(RDEV(i).total_segments); | |
2728 | i++; | |
2729 | } | |
2730 | if (segment_count != dev_seg_count) { | |
2731 | f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", | |
2732 | segment_count, dev_seg_count); | |
2733 | return -EFSCORRUPTED; | |
2734 | } | |
2735 | } | |
2736 | ||
2737 | if (secs_per_zone > total_sections || !secs_per_zone) { | |
2738 | f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", | |
2739 | secs_per_zone, total_sections); | |
2740 | return -EFSCORRUPTED; | |
2741 | } | |
2742 | if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || | |
2743 | raw_super->hot_ext_count > F2FS_MAX_EXTENSION || | |
2744 | (le32_to_cpu(raw_super->extension_count) + | |
2745 | raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { | |
2746 | f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", | |
2747 | le32_to_cpu(raw_super->extension_count), | |
2748 | raw_super->hot_ext_count, | |
2749 | F2FS_MAX_EXTENSION); | |
2750 | return -EFSCORRUPTED; | |
2751 | } | |
2752 | ||
2753 | if (le32_to_cpu(raw_super->cp_payload) > | |
2754 | (blocks_per_seg - F2FS_CP_PACKS)) { | |
2755 | f2fs_info(sbi, "Insane cp_payload (%u > %u)", | |
2756 | le32_to_cpu(raw_super->cp_payload), | |
2757 | blocks_per_seg - F2FS_CP_PACKS); | |
2758 | return -EFSCORRUPTED; | |
2759 | } | |
2760 | ||
2761 | /* check reserved ino info */ | |
2762 | if (le32_to_cpu(raw_super->node_ino) != 1 || | |
2763 | le32_to_cpu(raw_super->meta_ino) != 2 || | |
2764 | le32_to_cpu(raw_super->root_ino) != 3) { | |
2765 | f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", | |
2766 | le32_to_cpu(raw_super->node_ino), | |
2767 | le32_to_cpu(raw_super->meta_ino), | |
2768 | le32_to_cpu(raw_super->root_ino)); | |
2769 | return -EFSCORRUPTED; | |
2770 | } | |
2771 | ||
2772 | /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ | |
2773 | if (sanity_check_area_boundary(sbi, bh)) | |
2774 | return -EFSCORRUPTED; | |
2775 | ||
2776 | return 0; | |
2777 | } | |
2778 | ||
2779 | int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) | |
2780 | { | |
2781 | unsigned int total, fsmeta; | |
2782 | struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); | |
2783 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
2784 | unsigned int ovp_segments, reserved_segments; | |
2785 | unsigned int main_segs, blocks_per_seg; | |
2786 | unsigned int sit_segs, nat_segs; | |
2787 | unsigned int sit_bitmap_size, nat_bitmap_size; | |
2788 | unsigned int log_blocks_per_seg; | |
2789 | unsigned int segment_count_main; | |
2790 | unsigned int cp_pack_start_sum, cp_payload; | |
2791 | block_t user_block_count, valid_user_blocks; | |
2792 | block_t avail_node_count, valid_node_count; | |
2793 | int i, j; | |
2794 | ||
2795 | total = le32_to_cpu(raw_super->segment_count); | |
2796 | fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); | |
2797 | sit_segs = le32_to_cpu(raw_super->segment_count_sit); | |
2798 | fsmeta += sit_segs; | |
2799 | nat_segs = le32_to_cpu(raw_super->segment_count_nat); | |
2800 | fsmeta += nat_segs; | |
2801 | fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); | |
2802 | fsmeta += le32_to_cpu(raw_super->segment_count_ssa); | |
2803 | ||
2804 | if (unlikely(fsmeta >= total)) | |
2805 | return 1; | |
2806 | ||
2807 | ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); | |
2808 | reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); | |
2809 | ||
2810 | if (unlikely(fsmeta < F2FS_MIN_SEGMENTS || | |
2811 | ovp_segments == 0 || reserved_segments == 0)) { | |
2812 | f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); | |
2813 | return 1; | |
2814 | } | |
2815 | ||
2816 | user_block_count = le64_to_cpu(ckpt->user_block_count); | |
2817 | segment_count_main = le32_to_cpu(raw_super->segment_count_main); | |
2818 | log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); | |
2819 | if (!user_block_count || user_block_count >= | |
2820 | segment_count_main << log_blocks_per_seg) { | |
2821 | f2fs_err(sbi, "Wrong user_block_count: %u", | |
2822 | user_block_count); | |
2823 | return 1; | |
2824 | } | |
2825 | ||
2826 | valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); | |
2827 | if (valid_user_blocks > user_block_count) { | |
2828 | f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", | |
2829 | valid_user_blocks, user_block_count); | |
2830 | return 1; | |
2831 | } | |
2832 | ||
2833 | valid_node_count = le32_to_cpu(ckpt->valid_node_count); | |
2834 | avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; | |
2835 | if (valid_node_count > avail_node_count) { | |
2836 | f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", | |
2837 | valid_node_count, avail_node_count); | |
2838 | return 1; | |
2839 | } | |
2840 | ||
2841 | main_segs = le32_to_cpu(raw_super->segment_count_main); | |
2842 | blocks_per_seg = sbi->blocks_per_seg; | |
2843 | ||
2844 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { | |
2845 | if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || | |
2846 | le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) | |
2847 | return 1; | |
2848 | for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { | |
2849 | if (le32_to_cpu(ckpt->cur_node_segno[i]) == | |
2850 | le32_to_cpu(ckpt->cur_node_segno[j])) { | |
2851 | f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", | |
2852 | i, j, | |
2853 | le32_to_cpu(ckpt->cur_node_segno[i])); | |
2854 | return 1; | |
2855 | } | |
2856 | } | |
2857 | } | |
2858 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { | |
2859 | if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || | |
2860 | le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) | |
2861 | return 1; | |
2862 | for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { | |
2863 | if (le32_to_cpu(ckpt->cur_data_segno[i]) == | |
2864 | le32_to_cpu(ckpt->cur_data_segno[j])) { | |
2865 | f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", | |
2866 | i, j, | |
2867 | le32_to_cpu(ckpt->cur_data_segno[i])); | |
2868 | return 1; | |
2869 | } | |
2870 | } | |
2871 | } | |
2872 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { | |
2873 | for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { | |
2874 | if (le32_to_cpu(ckpt->cur_node_segno[i]) == | |
2875 | le32_to_cpu(ckpt->cur_data_segno[j])) { | |
2876 | f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", | |
2877 | i, j, | |
2878 | le32_to_cpu(ckpt->cur_node_segno[i])); | |
2879 | return 1; | |
2880 | } | |
2881 | } | |
2882 | } | |
2883 | ||
2884 | sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); | |
2885 | nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); | |
2886 | ||
2887 | if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || | |
2888 | nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { | |
2889 | f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", | |
2890 | sit_bitmap_size, nat_bitmap_size); | |
2891 | return 1; | |
2892 | } | |
2893 | ||
2894 | cp_pack_start_sum = __start_sum_addr(sbi); | |
2895 | cp_payload = __cp_payload(sbi); | |
2896 | if (cp_pack_start_sum < cp_payload + 1 || | |
2897 | cp_pack_start_sum > blocks_per_seg - 1 - | |
2898 | NR_CURSEG_TYPE) { | |
2899 | f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", | |
2900 | cp_pack_start_sum); | |
2901 | return 1; | |
2902 | } | |
2903 | ||
2904 | if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && | |
2905 | le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { | |
2906 | f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " | |
2907 | "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " | |
2908 | "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", | |
2909 | le32_to_cpu(ckpt->checksum_offset)); | |
2910 | return 1; | |
2911 | } | |
2912 | ||
2913 | if (unlikely(f2fs_cp_error(sbi))) { | |
2914 | f2fs_err(sbi, "A bug case: need to run fsck"); | |
2915 | return 1; | |
2916 | } | |
2917 | return 0; | |
2918 | } | |
2919 | ||
2920 | static void init_sb_info(struct f2fs_sb_info *sbi) | |
2921 | { | |
2922 | struct f2fs_super_block *raw_super = sbi->raw_super; | |
2923 | int i; | |
2924 | ||
2925 | sbi->log_sectors_per_block = | |
2926 | le32_to_cpu(raw_super->log_sectors_per_block); | |
2927 | sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); | |
2928 | sbi->blocksize = 1 << sbi->log_blocksize; | |
2929 | sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); | |
2930 | sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; | |
2931 | sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); | |
2932 | sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); | |
2933 | sbi->total_sections = le32_to_cpu(raw_super->section_count); | |
2934 | sbi->total_node_count = | |
2935 | (le32_to_cpu(raw_super->segment_count_nat) / 2) | |
2936 | * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; | |
2937 | sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); | |
2938 | sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); | |
2939 | sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); | |
2940 | sbi->cur_victim_sec = NULL_SECNO; | |
2941 | sbi->next_victim_seg[BG_GC] = NULL_SEGNO; | |
2942 | sbi->next_victim_seg[FG_GC] = NULL_SEGNO; | |
2943 | sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; | |
2944 | sbi->migration_granularity = sbi->segs_per_sec; | |
2945 | ||
2946 | sbi->dir_level = DEF_DIR_LEVEL; | |
2947 | sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; | |
2948 | sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; | |
2949 | sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; | |
2950 | sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; | |
2951 | sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; | |
2952 | sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = | |
2953 | DEF_UMOUNT_DISCARD_TIMEOUT; | |
2954 | clear_sbi_flag(sbi, SBI_NEED_FSCK); | |
2955 | ||
2956 | for (i = 0; i < NR_COUNT_TYPE; i++) | |
2957 | atomic_set(&sbi->nr_pages[i], 0); | |
2958 | ||
2959 | for (i = 0; i < META; i++) | |
2960 | atomic_set(&sbi->wb_sync_req[i], 0); | |
2961 | ||
2962 | INIT_LIST_HEAD(&sbi->s_list); | |
2963 | mutex_init(&sbi->umount_mutex); | |
2964 | init_rwsem(&sbi->io_order_lock); | |
2965 | spin_lock_init(&sbi->cp_lock); | |
2966 | ||
2967 | sbi->dirty_device = 0; | |
2968 | spin_lock_init(&sbi->dev_lock); | |
2969 | ||
2970 | init_rwsem(&sbi->sb_lock); | |
2971 | init_rwsem(&sbi->pin_sem); | |
2972 | } | |
2973 | ||
2974 | static int init_percpu_info(struct f2fs_sb_info *sbi) | |
2975 | { | |
2976 | int err; | |
2977 | ||
2978 | err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); | |
2979 | if (err) | |
2980 | return err; | |
2981 | ||
2982 | err = percpu_counter_init(&sbi->total_valid_inode_count, 0, | |
2983 | GFP_KERNEL); | |
2984 | if (err) | |
2985 | percpu_counter_destroy(&sbi->alloc_valid_block_count); | |
2986 | ||
2987 | return err; | |
2988 | } | |
2989 | ||
2990 | #ifdef CONFIG_BLK_DEV_ZONED | |
2991 | static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, | |
2992 | void *data) | |
2993 | { | |
2994 | struct f2fs_dev_info *dev = data; | |
2995 | ||
2996 | if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL) | |
2997 | set_bit(idx, dev->blkz_seq); | |
2998 | return 0; | |
2999 | } | |
3000 | ||
3001 | static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) | |
3002 | { | |
3003 | struct block_device *bdev = FDEV(devi).bdev; | |
3004 | sector_t nr_sectors = bdev->bd_part->nr_sects; | |
3005 | int ret; | |
3006 | ||
3007 | if (!f2fs_sb_has_blkzoned(sbi)) | |
3008 | return 0; | |
3009 | ||
3010 | if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != | |
3011 | SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) | |
3012 | return -EINVAL; | |
3013 | sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); | |
3014 | if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != | |
3015 | __ilog2_u32(sbi->blocks_per_blkz)) | |
3016 | return -EINVAL; | |
3017 | sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); | |
3018 | FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> | |
3019 | sbi->log_blocks_per_blkz; | |
3020 | if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) | |
3021 | FDEV(devi).nr_blkz++; | |
3022 | ||
3023 | FDEV(devi).blkz_seq = f2fs_kzalloc(sbi, | |
3024 | BITS_TO_LONGS(FDEV(devi).nr_blkz) | |
3025 | * sizeof(unsigned long), | |
3026 | GFP_KERNEL); | |
3027 | if (!FDEV(devi).blkz_seq) | |
3028 | return -ENOMEM; | |
3029 | ||
3030 | /* Get block zones type */ | |
3031 | ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, | |
3032 | &FDEV(devi)); | |
3033 | if (ret < 0) | |
3034 | return ret; | |
3035 | ||
3036 | return 0; | |
3037 | } | |
3038 | #endif | |
3039 | ||
3040 | /* | |
3041 | * Read f2fs raw super block. | |
3042 | * Because we have two copies of super block, so read both of them | |
3043 | * to get the first valid one. If any one of them is broken, we pass | |
3044 | * them recovery flag back to the caller. | |
3045 | */ | |
3046 | static int read_raw_super_block(struct f2fs_sb_info *sbi, | |
3047 | struct f2fs_super_block **raw_super, | |
3048 | int *valid_super_block, int *recovery) | |
3049 | { | |
3050 | struct super_block *sb = sbi->sb; | |
3051 | int block; | |
3052 | struct buffer_head *bh; | |
3053 | struct f2fs_super_block *super; | |
3054 | int err = 0; | |
3055 | ||
3056 | super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); | |
3057 | if (!super) | |
3058 | return -ENOMEM; | |
3059 | ||
3060 | for (block = 0; block < 2; block++) { | |
3061 | bh = sb_bread(sb, block); | |
3062 | if (!bh) { | |
3063 | f2fs_err(sbi, "Unable to read %dth superblock", | |
3064 | block + 1); | |
3065 | err = -EIO; | |
3066 | *recovery = 1; | |
3067 | continue; | |
3068 | } | |
3069 | ||
3070 | /* sanity checking of raw super */ | |
3071 | err = sanity_check_raw_super(sbi, bh); | |
3072 | if (err) { | |
3073 | f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", | |
3074 | block + 1); | |
3075 | brelse(bh); | |
3076 | *recovery = 1; | |
3077 | continue; | |
3078 | } | |
3079 | ||
3080 | if (!*raw_super) { | |
3081 | memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, | |
3082 | sizeof(*super)); | |
3083 | *valid_super_block = block; | |
3084 | *raw_super = super; | |
3085 | } | |
3086 | brelse(bh); | |
3087 | } | |
3088 | ||
3089 | /* No valid superblock */ | |
3090 | if (!*raw_super) | |
3091 | kvfree(super); | |
3092 | else | |
3093 | err = 0; | |
3094 | ||
3095 | return err; | |
3096 | } | |
3097 | ||
3098 | int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) | |
3099 | { | |
3100 | struct buffer_head *bh; | |
3101 | __u32 crc = 0; | |
3102 | int err; | |
3103 | ||
3104 | if ((recover && f2fs_readonly(sbi->sb)) || | |
3105 | bdev_read_only(sbi->sb->s_bdev)) { | |
3106 | set_sbi_flag(sbi, SBI_NEED_SB_WRITE); | |
3107 | return -EROFS; | |
3108 | } | |
3109 | ||
3110 | /* we should update superblock crc here */ | |
3111 | if (!recover && f2fs_sb_has_sb_chksum(sbi)) { | |
3112 | crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), | |
3113 | offsetof(struct f2fs_super_block, crc)); | |
3114 | F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); | |
3115 | } | |
3116 | ||
3117 | /* write back-up superblock first */ | |
3118 | bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); | |
3119 | if (!bh) | |
3120 | return -EIO; | |
3121 | err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); | |
3122 | brelse(bh); | |
3123 | ||
3124 | /* if we are in recovery path, skip writing valid superblock */ | |
3125 | if (recover || err) | |
3126 | return err; | |
3127 | ||
3128 | /* write current valid superblock */ | |
3129 | bh = sb_bread(sbi->sb, sbi->valid_super_block); | |
3130 | if (!bh) | |
3131 | return -EIO; | |
3132 | err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); | |
3133 | brelse(bh); | |
3134 | return err; | |
3135 | } | |
3136 | ||
3137 | static int f2fs_scan_devices(struct f2fs_sb_info *sbi) | |
3138 | { | |
3139 | struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); | |
3140 | unsigned int max_devices = MAX_DEVICES; | |
3141 | int i; | |
3142 | ||
3143 | /* Initialize single device information */ | |
3144 | if (!RDEV(0).path[0]) { | |
3145 | if (!bdev_is_zoned(sbi->sb->s_bdev)) | |
3146 | return 0; | |
3147 | max_devices = 1; | |
3148 | } | |
3149 | ||
3150 | /* | |
3151 | * Initialize multiple devices information, or single | |
3152 | * zoned block device information. | |
3153 | */ | |
3154 | sbi->devs = f2fs_kzalloc(sbi, | |
3155 | array_size(max_devices, | |
3156 | sizeof(struct f2fs_dev_info)), | |
3157 | GFP_KERNEL); | |
3158 | if (!sbi->devs) | |
3159 | return -ENOMEM; | |
3160 | ||
3161 | for (i = 0; i < max_devices; i++) { | |
3162 | ||
3163 | if (i > 0 && !RDEV(i).path[0]) | |
3164 | break; | |
3165 | ||
3166 | if (max_devices == 1) { | |
3167 | /* Single zoned block device mount */ | |
3168 | FDEV(0).bdev = | |
3169 | blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, | |
3170 | sbi->sb->s_mode, sbi->sb->s_type); | |
3171 | } else { | |
3172 | /* Multi-device mount */ | |
3173 | memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); | |
3174 | FDEV(i).total_segments = | |
3175 | le32_to_cpu(RDEV(i).total_segments); | |
3176 | if (i == 0) { | |
3177 | FDEV(i).start_blk = 0; | |
3178 | FDEV(i).end_blk = FDEV(i).start_blk + | |
3179 | (FDEV(i).total_segments << | |
3180 | sbi->log_blocks_per_seg) - 1 + | |
3181 | le32_to_cpu(raw_super->segment0_blkaddr); | |
3182 | } else { | |
3183 | FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; | |
3184 | FDEV(i).end_blk = FDEV(i).start_blk + | |
3185 | (FDEV(i).total_segments << | |
3186 | sbi->log_blocks_per_seg) - 1; | |
3187 | } | |
3188 | FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, | |
3189 | sbi->sb->s_mode, sbi->sb->s_type); | |
3190 | } | |
3191 | if (IS_ERR(FDEV(i).bdev)) | |
3192 | return PTR_ERR(FDEV(i).bdev); | |
3193 | ||
3194 | /* to release errored devices */ | |
3195 | sbi->s_ndevs = i + 1; | |
3196 | ||
3197 | #ifdef CONFIG_BLK_DEV_ZONED | |
3198 | if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && | |
3199 | !f2fs_sb_has_blkzoned(sbi)) { | |
3200 | f2fs_err(sbi, "Zoned block device feature not enabled\n"); | |
3201 | return -EINVAL; | |
3202 | } | |
3203 | if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { | |
3204 | if (init_blkz_info(sbi, i)) { | |
3205 | f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); | |
3206 | return -EINVAL; | |
3207 | } | |
3208 | if (max_devices == 1) | |
3209 | break; | |
3210 | f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", | |
3211 | i, FDEV(i).path, | |
3212 | FDEV(i).total_segments, | |
3213 | FDEV(i).start_blk, FDEV(i).end_blk, | |
3214 | bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? | |
3215 | "Host-aware" : "Host-managed"); | |
3216 | continue; | |
3217 | } | |
3218 | #endif | |
3219 | f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", | |
3220 | i, FDEV(i).path, | |
3221 | FDEV(i).total_segments, | |
3222 | FDEV(i).start_blk, FDEV(i).end_blk); | |
3223 | } | |
3224 | f2fs_info(sbi, | |
3225 | "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); | |
3226 | return 0; | |
3227 | } | |
3228 | ||
3229 | static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) | |
3230 | { | |
3231 | #ifdef CONFIG_UNICODE | |
3232 | if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) { | |
3233 | const struct f2fs_sb_encodings *encoding_info; | |
3234 | struct unicode_map *encoding; | |
3235 | __u16 encoding_flags; | |
3236 | ||
3237 | if (f2fs_sb_has_encrypt(sbi)) { | |
3238 | f2fs_err(sbi, | |
3239 | "Can't mount with encoding and encryption"); | |
3240 | return -EINVAL; | |
3241 | } | |
3242 | ||
3243 | if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, | |
3244 | &encoding_flags)) { | |
3245 | f2fs_err(sbi, | |
3246 | "Encoding requested by superblock is unknown"); | |
3247 | return -EINVAL; | |
3248 | } | |
3249 | ||
3250 | encoding = utf8_load(encoding_info->version); | |
3251 | if (IS_ERR(encoding)) { | |
3252 | f2fs_err(sbi, | |
3253 | "can't mount with superblock charset: %s-%s " | |
3254 | "not supported by the kernel. flags: 0x%x.", | |
3255 | encoding_info->name, encoding_info->version, | |
3256 | encoding_flags); | |
3257 | return PTR_ERR(encoding); | |
3258 | } | |
3259 | f2fs_info(sbi, "Using encoding defined by superblock: " | |
3260 | "%s-%s with flags 0x%hx", encoding_info->name, | |
3261 | encoding_info->version?:"\b", encoding_flags); | |
3262 | ||
3263 | sbi->s_encoding = encoding; | |
3264 | sbi->s_encoding_flags = encoding_flags; | |
3265 | sbi->sb->s_d_op = &f2fs_dentry_ops; | |
3266 | } | |
3267 | #else | |
3268 | if (f2fs_sb_has_casefold(sbi)) { | |
3269 | f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); | |
3270 | return -EINVAL; | |
3271 | } | |
3272 | #endif | |
3273 | return 0; | |
3274 | } | |
3275 | ||
3276 | static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) | |
3277 | { | |
3278 | struct f2fs_sm_info *sm_i = SM_I(sbi); | |
3279 | ||
3280 | /* adjust parameters according to the volume size */ | |
3281 | if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { | |
3282 | F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; | |
3283 | sm_i->dcc_info->discard_granularity = 1; | |
3284 | sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; | |
3285 | } | |
3286 | ||
3287 | sbi->readdir_ra = 1; | |
3288 | } | |
3289 | ||
3290 | static int f2fs_fill_super(struct super_block *sb, void *data, int silent) | |
3291 | { | |
3292 | struct f2fs_sb_info *sbi; | |
3293 | struct f2fs_super_block *raw_super; | |
3294 | struct inode *root; | |
3295 | int err; | |
3296 | bool skip_recovery = false, need_fsck = false; | |
3297 | char *options = NULL; | |
3298 | int recovery, i, valid_super_block; | |
3299 | struct curseg_info *seg_i; | |
3300 | int retry_cnt = 1; | |
3301 | ||
3302 | try_onemore: | |
3303 | err = -EINVAL; | |
3304 | raw_super = NULL; | |
3305 | valid_super_block = -1; | |
3306 | recovery = 0; | |
3307 | ||
3308 | /* allocate memory for f2fs-specific super block info */ | |
3309 | sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); | |
3310 | if (!sbi) | |
3311 | return -ENOMEM; | |
3312 | ||
3313 | sbi->sb = sb; | |
3314 | ||
3315 | /* Load the checksum driver */ | |
3316 | sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); | |
3317 | if (IS_ERR(sbi->s_chksum_driver)) { | |
3318 | f2fs_err(sbi, "Cannot load crc32 driver."); | |
3319 | err = PTR_ERR(sbi->s_chksum_driver); | |
3320 | sbi->s_chksum_driver = NULL; | |
3321 | goto free_sbi; | |
3322 | } | |
3323 | ||
3324 | /* set a block size */ | |
3325 | if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { | |
3326 | f2fs_err(sbi, "unable to set blocksize"); | |
3327 | goto free_sbi; | |
3328 | } | |
3329 | ||
3330 | err = read_raw_super_block(sbi, &raw_super, &valid_super_block, | |
3331 | &recovery); | |
3332 | if (err) | |
3333 | goto free_sbi; | |
3334 | ||
3335 | sb->s_fs_info = sbi; | |
3336 | sbi->raw_super = raw_super; | |
3337 | ||
3338 | /* precompute checksum seed for metadata */ | |
3339 | if (f2fs_sb_has_inode_chksum(sbi)) | |
3340 | sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, | |
3341 | sizeof(raw_super->uuid)); | |
3342 | ||
3343 | /* | |
3344 | * The BLKZONED feature indicates that the drive was formatted with | |
3345 | * zone alignment optimization. This is optional for host-aware | |
3346 | * devices, but mandatory for host-managed zoned block devices. | |
3347 | */ | |
3348 | #ifndef CONFIG_BLK_DEV_ZONED | |
3349 | if (f2fs_sb_has_blkzoned(sbi)) { | |
3350 | f2fs_err(sbi, "Zoned block device support is not enabled"); | |
3351 | err = -EOPNOTSUPP; | |
3352 | goto free_sb_buf; | |
3353 | } | |
3354 | #endif | |
3355 | default_options(sbi); | |
3356 | /* parse mount options */ | |
3357 | options = kstrdup((const char *)data, GFP_KERNEL); | |
3358 | if (data && !options) { | |
3359 | err = -ENOMEM; | |
3360 | goto free_sb_buf; | |
3361 | } | |
3362 | ||
3363 | err = parse_options(sb, options); | |
3364 | if (err) | |
3365 | goto free_options; | |
3366 | ||
3367 | sbi->max_file_blocks = max_file_blocks(); | |
3368 | sb->s_maxbytes = sbi->max_file_blocks << | |
3369 | le32_to_cpu(raw_super->log_blocksize); | |
3370 | sb->s_max_links = F2FS_LINK_MAX; | |
3371 | ||
3372 | err = f2fs_setup_casefold(sbi); | |
3373 | if (err) | |
3374 | goto free_options; | |
3375 | ||
3376 | #ifdef CONFIG_QUOTA | |
3377 | sb->dq_op = &f2fs_quota_operations; | |
3378 | sb->s_qcop = &f2fs_quotactl_ops; | |
3379 | sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; | |
3380 | ||
3381 | if (f2fs_sb_has_quota_ino(sbi)) { | |
3382 | for (i = 0; i < MAXQUOTAS; i++) { | |
3383 | if (f2fs_qf_ino(sbi->sb, i)) | |
3384 | sbi->nquota_files++; | |
3385 | } | |
3386 | } | |
3387 | #endif | |
3388 | ||
3389 | sb->s_op = &f2fs_sops; | |
3390 | #ifdef CONFIG_FS_ENCRYPTION | |
3391 | sb->s_cop = &f2fs_cryptops; | |
3392 | #endif | |
3393 | #ifdef CONFIG_FS_VERITY | |
3394 | sb->s_vop = &f2fs_verityops; | |
3395 | #endif | |
3396 | sb->s_xattr = f2fs_xattr_handlers; | |
3397 | sb->s_export_op = &f2fs_export_ops; | |
3398 | sb->s_magic = F2FS_SUPER_MAGIC; | |
3399 | sb->s_time_gran = 1; | |
3400 | sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | | |
3401 | (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); | |
3402 | memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); | |
3403 | sb->s_iflags |= SB_I_CGROUPWB; | |
3404 | ||
3405 | /* init f2fs-specific super block info */ | |
3406 | sbi->valid_super_block = valid_super_block; | |
3407 | init_rwsem(&sbi->gc_lock); | |
3408 | mutex_init(&sbi->writepages); | |
3409 | mutex_init(&sbi->cp_mutex); | |
3410 | mutex_init(&sbi->resize_mutex); | |
3411 | init_rwsem(&sbi->node_write); | |
3412 | init_rwsem(&sbi->node_change); | |
3413 | ||
3414 | /* disallow all the data/node/meta page writes */ | |
3415 | set_sbi_flag(sbi, SBI_POR_DOING); | |
3416 | spin_lock_init(&sbi->stat_lock); | |
3417 | ||
3418 | /* init iostat info */ | |
3419 | spin_lock_init(&sbi->iostat_lock); | |
3420 | sbi->iostat_enable = false; | |
3421 | sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS; | |
3422 | ||
3423 | for (i = 0; i < NR_PAGE_TYPE; i++) { | |
3424 | int n = (i == META) ? 1: NR_TEMP_TYPE; | |
3425 | int j; | |
3426 | ||
3427 | sbi->write_io[i] = | |
3428 | f2fs_kmalloc(sbi, | |
3429 | array_size(n, | |
3430 | sizeof(struct f2fs_bio_info)), | |
3431 | GFP_KERNEL); | |
3432 | if (!sbi->write_io[i]) { | |
3433 | err = -ENOMEM; | |
3434 | goto free_bio_info; | |
3435 | } | |
3436 | ||
3437 | for (j = HOT; j < n; j++) { | |
3438 | init_rwsem(&sbi->write_io[i][j].io_rwsem); | |
3439 | sbi->write_io[i][j].sbi = sbi; | |
3440 | sbi->write_io[i][j].bio = NULL; | |
3441 | spin_lock_init(&sbi->write_io[i][j].io_lock); | |
3442 | INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); | |
3443 | INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); | |
3444 | init_rwsem(&sbi->write_io[i][j].bio_list_lock); | |
3445 | } | |
3446 | } | |
3447 | ||
3448 | init_rwsem(&sbi->cp_rwsem); | |
3449 | init_rwsem(&sbi->quota_sem); | |
3450 | init_waitqueue_head(&sbi->cp_wait); | |
3451 | init_sb_info(sbi); | |
3452 | ||
3453 | err = init_percpu_info(sbi); | |
3454 | if (err) | |
3455 | goto free_bio_info; | |
3456 | ||
3457 | if (F2FS_IO_ALIGNED(sbi)) { | |
3458 | sbi->write_io_dummy = | |
3459 | mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); | |
3460 | if (!sbi->write_io_dummy) { | |
3461 | err = -ENOMEM; | |
3462 | goto free_percpu; | |
3463 | } | |
3464 | } | |
3465 | ||
3466 | /* init per sbi slab cache */ | |
3467 | err = f2fs_init_xattr_caches(sbi); | |
3468 | if (err) | |
3469 | goto free_io_dummy; | |
3470 | ||
3471 | /* get an inode for meta space */ | |
3472 | sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); | |
3473 | if (IS_ERR(sbi->meta_inode)) { | |
3474 | f2fs_err(sbi, "Failed to read F2FS meta data inode"); | |
3475 | err = PTR_ERR(sbi->meta_inode); | |
3476 | goto free_xattr_cache; | |
3477 | } | |
3478 | ||
3479 | err = f2fs_get_valid_checkpoint(sbi); | |
3480 | if (err) { | |
3481 | f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); | |
3482 | goto free_meta_inode; | |
3483 | } | |
3484 | ||
3485 | if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) | |
3486 | set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); | |
3487 | if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { | |
3488 | set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); | |
3489 | sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; | |
3490 | } | |
3491 | ||
3492 | if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) | |
3493 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
3494 | ||
3495 | /* Initialize device list */ | |
3496 | err = f2fs_scan_devices(sbi); | |
3497 | if (err) { | |
3498 | f2fs_err(sbi, "Failed to find devices"); | |
3499 | goto free_devices; | |
3500 | } | |
3501 | ||
3502 | err = f2fs_init_post_read_wq(sbi); | |
3503 | if (err) { | |
3504 | f2fs_err(sbi, "Failed to initialize post read workqueue"); | |
3505 | goto free_devices; | |
3506 | } | |
3507 | ||
3508 | sbi->total_valid_node_count = | |
3509 | le32_to_cpu(sbi->ckpt->valid_node_count); | |
3510 | percpu_counter_set(&sbi->total_valid_inode_count, | |
3511 | le32_to_cpu(sbi->ckpt->valid_inode_count)); | |
3512 | sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); | |
3513 | sbi->total_valid_block_count = | |
3514 | le64_to_cpu(sbi->ckpt->valid_block_count); | |
3515 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
3516 | sbi->reserved_blocks = 0; | |
3517 | sbi->current_reserved_blocks = 0; | |
3518 | limit_reserve_root(sbi); | |
3519 | ||
3520 | for (i = 0; i < NR_INODE_TYPE; i++) { | |
3521 | INIT_LIST_HEAD(&sbi->inode_list[i]); | |
3522 | spin_lock_init(&sbi->inode_lock[i]); | |
3523 | } | |
3524 | mutex_init(&sbi->flush_lock); | |
3525 | ||
3526 | f2fs_init_extent_cache_info(sbi); | |
3527 | ||
3528 | f2fs_init_ino_entry_info(sbi); | |
3529 | ||
3530 | f2fs_init_fsync_node_info(sbi); | |
3531 | ||
3532 | /* setup f2fs internal modules */ | |
3533 | err = f2fs_build_segment_manager(sbi); | |
3534 | if (err) { | |
3535 | f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", | |
3536 | err); | |
3537 | goto free_sm; | |
3538 | } | |
3539 | err = f2fs_build_node_manager(sbi); | |
3540 | if (err) { | |
3541 | f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", | |
3542 | err); | |
3543 | goto free_nm; | |
3544 | } | |
3545 | ||
3546 | /* For write statistics */ | |
3547 | if (sb->s_bdev->bd_part) | |
3548 | sbi->sectors_written_start = | |
3549 | (u64)part_stat_read(sb->s_bdev->bd_part, | |
3550 | sectors[STAT_WRITE]); | |
3551 | ||
3552 | /* Read accumulated write IO statistics if exists */ | |
3553 | seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
3554 | if (__exist_node_summaries(sbi)) | |
3555 | sbi->kbytes_written = | |
3556 | le64_to_cpu(seg_i->journal->info.kbytes_written); | |
3557 | ||
3558 | f2fs_build_gc_manager(sbi); | |
3559 | ||
3560 | err = f2fs_build_stats(sbi); | |
3561 | if (err) | |
3562 | goto free_nm; | |
3563 | ||
3564 | /* get an inode for node space */ | |
3565 | sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); | |
3566 | if (IS_ERR(sbi->node_inode)) { | |
3567 | f2fs_err(sbi, "Failed to read node inode"); | |
3568 | err = PTR_ERR(sbi->node_inode); | |
3569 | goto free_stats; | |
3570 | } | |
3571 | ||
3572 | /* read root inode and dentry */ | |
3573 | root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); | |
3574 | if (IS_ERR(root)) { | |
3575 | f2fs_err(sbi, "Failed to read root inode"); | |
3576 | err = PTR_ERR(root); | |
3577 | goto free_node_inode; | |
3578 | } | |
3579 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || | |
3580 | !root->i_size || !root->i_nlink) { | |
3581 | iput(root); | |
3582 | err = -EINVAL; | |
3583 | goto free_node_inode; | |
3584 | } | |
3585 | ||
3586 | sb->s_root = d_make_root(root); /* allocate root dentry */ | |
3587 | if (!sb->s_root) { | |
3588 | err = -ENOMEM; | |
3589 | goto free_node_inode; | |
3590 | } | |
3591 | ||
3592 | err = f2fs_register_sysfs(sbi); | |
3593 | if (err) | |
3594 | goto free_root_inode; | |
3595 | ||
3596 | #ifdef CONFIG_QUOTA | |
3597 | /* Enable quota usage during mount */ | |
3598 | if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { | |
3599 | err = f2fs_enable_quotas(sb); | |
3600 | if (err) | |
3601 | f2fs_err(sbi, "Cannot turn on quotas: error %d", err); | |
3602 | } | |
3603 | #endif | |
3604 | /* if there are any orphan inodes, free them */ | |
3605 | err = f2fs_recover_orphan_inodes(sbi); | |
3606 | if (err) | |
3607 | goto free_meta; | |
3608 | ||
3609 | if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) | |
3610 | goto reset_checkpoint; | |
3611 | ||
3612 | /* recover fsynced data */ | |
3613 | if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && | |
3614 | !test_opt(sbi, NORECOVERY)) { | |
3615 | /* | |
3616 | * mount should be failed, when device has readonly mode, and | |
3617 | * previous checkpoint was not done by clean system shutdown. | |
3618 | */ | |
3619 | if (f2fs_hw_is_readonly(sbi)) { | |
3620 | if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { | |
3621 | err = -EROFS; | |
3622 | f2fs_err(sbi, "Need to recover fsync data, but write access unavailable"); | |
3623 | goto free_meta; | |
3624 | } | |
3625 | f2fs_info(sbi, "write access unavailable, skipping recovery"); | |
3626 | goto reset_checkpoint; | |
3627 | } | |
3628 | ||
3629 | if (need_fsck) | |
3630 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
3631 | ||
3632 | if (skip_recovery) | |
3633 | goto reset_checkpoint; | |
3634 | ||
3635 | err = f2fs_recover_fsync_data(sbi, false); | |
3636 | if (err < 0) { | |
3637 | if (err != -ENOMEM) | |
3638 | skip_recovery = true; | |
3639 | need_fsck = true; | |
3640 | f2fs_err(sbi, "Cannot recover all fsync data errno=%d", | |
3641 | err); | |
3642 | goto free_meta; | |
3643 | } | |
3644 | } else { | |
3645 | err = f2fs_recover_fsync_data(sbi, true); | |
3646 | ||
3647 | if (!f2fs_readonly(sb) && err > 0) { | |
3648 | err = -EINVAL; | |
3649 | f2fs_err(sbi, "Need to recover fsync data"); | |
3650 | goto free_meta; | |
3651 | } | |
3652 | } | |
3653 | ||
3654 | /* | |
3655 | * If the f2fs is not readonly and fsync data recovery succeeds, | |
3656 | * check zoned block devices' write pointer consistency. | |
3657 | */ | |
3658 | if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { | |
3659 | err = f2fs_check_write_pointer(sbi); | |
3660 | if (err) | |
3661 | goto free_meta; | |
3662 | } | |
3663 | ||
3664 | reset_checkpoint: | |
3665 | /* f2fs_recover_fsync_data() cleared this already */ | |
3666 | clear_sbi_flag(sbi, SBI_POR_DOING); | |
3667 | ||
3668 | if (test_opt(sbi, DISABLE_CHECKPOINT)) { | |
3669 | err = f2fs_disable_checkpoint(sbi); | |
3670 | if (err) | |
3671 | goto sync_free_meta; | |
3672 | } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { | |
3673 | f2fs_enable_checkpoint(sbi); | |
3674 | } | |
3675 | ||
3676 | /* | |
3677 | * If filesystem is not mounted as read-only then | |
3678 | * do start the gc_thread. | |
3679 | */ | |
3680 | if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) { | |
3681 | /* After POR, we can run background GC thread.*/ | |
3682 | err = f2fs_start_gc_thread(sbi); | |
3683 | if (err) | |
3684 | goto sync_free_meta; | |
3685 | } | |
3686 | kvfree(options); | |
3687 | ||
3688 | /* recover broken superblock */ | |
3689 | if (recovery) { | |
3690 | err = f2fs_commit_super(sbi, true); | |
3691 | f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", | |
3692 | sbi->valid_super_block ? 1 : 2, err); | |
3693 | } | |
3694 | ||
3695 | f2fs_join_shrinker(sbi); | |
3696 | ||
3697 | f2fs_tuning_parameters(sbi); | |
3698 | ||
3699 | f2fs_notice(sbi, "Mounted with checkpoint version = %llx", | |
3700 | cur_cp_version(F2FS_CKPT(sbi))); | |
3701 | f2fs_update_time(sbi, CP_TIME); | |
3702 | f2fs_update_time(sbi, REQ_TIME); | |
3703 | clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); | |
3704 | return 0; | |
3705 | ||
3706 | sync_free_meta: | |
3707 | /* safe to flush all the data */ | |
3708 | sync_filesystem(sbi->sb); | |
3709 | retry_cnt = 0; | |
3710 | ||
3711 | free_meta: | |
3712 | #ifdef CONFIG_QUOTA | |
3713 | f2fs_truncate_quota_inode_pages(sb); | |
3714 | if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) | |
3715 | f2fs_quota_off_umount(sbi->sb); | |
3716 | #endif | |
3717 | /* | |
3718 | * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() | |
3719 | * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() | |
3720 | * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which | |
3721 | * falls into an infinite loop in f2fs_sync_meta_pages(). | |
3722 | */ | |
3723 | truncate_inode_pages_final(META_MAPPING(sbi)); | |
3724 | /* evict some inodes being cached by GC */ | |
3725 | evict_inodes(sb); | |
3726 | f2fs_unregister_sysfs(sbi); | |
3727 | free_root_inode: | |
3728 | dput(sb->s_root); | |
3729 | sb->s_root = NULL; | |
3730 | free_node_inode: | |
3731 | f2fs_release_ino_entry(sbi, true); | |
3732 | truncate_inode_pages_final(NODE_MAPPING(sbi)); | |
3733 | iput(sbi->node_inode); | |
3734 | sbi->node_inode = NULL; | |
3735 | free_stats: | |
3736 | f2fs_destroy_stats(sbi); | |
3737 | free_nm: | |
3738 | f2fs_destroy_node_manager(sbi); | |
3739 | free_sm: | |
3740 | f2fs_destroy_segment_manager(sbi); | |
3741 | f2fs_destroy_post_read_wq(sbi); | |
3742 | free_devices: | |
3743 | destroy_device_list(sbi); | |
3744 | kvfree(sbi->ckpt); | |
3745 | free_meta_inode: | |
3746 | make_bad_inode(sbi->meta_inode); | |
3747 | iput(sbi->meta_inode); | |
3748 | sbi->meta_inode = NULL; | |
3749 | free_xattr_cache: | |
3750 | f2fs_destroy_xattr_caches(sbi); | |
3751 | free_io_dummy: | |
3752 | mempool_destroy(sbi->write_io_dummy); | |
3753 | free_percpu: | |
3754 | destroy_percpu_info(sbi); | |
3755 | free_bio_info: | |
3756 | for (i = 0; i < NR_PAGE_TYPE; i++) | |
3757 | kvfree(sbi->write_io[i]); | |
3758 | ||
3759 | #ifdef CONFIG_UNICODE | |
3760 | utf8_unload(sbi->s_encoding); | |
3761 | #endif | |
3762 | free_options: | |
3763 | #ifdef CONFIG_QUOTA | |
3764 | for (i = 0; i < MAXQUOTAS; i++) | |
3765 | kvfree(F2FS_OPTION(sbi).s_qf_names[i]); | |
3766 | #endif | |
3767 | kvfree(options); | |
3768 | free_sb_buf: | |
3769 | kvfree(raw_super); | |
3770 | free_sbi: | |
3771 | if (sbi->s_chksum_driver) | |
3772 | crypto_free_shash(sbi->s_chksum_driver); | |
3773 | kvfree(sbi); | |
3774 | ||
3775 | /* give only one another chance */ | |
3776 | if (retry_cnt > 0 && skip_recovery) { | |
3777 | retry_cnt--; | |
3778 | shrink_dcache_sb(sb); | |
3779 | goto try_onemore; | |
3780 | } | |
3781 | return err; | |
3782 | } | |
3783 | ||
3784 | static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, | |
3785 | const char *dev_name, void *data) | |
3786 | { | |
3787 | return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); | |
3788 | } | |
3789 | ||
3790 | static void kill_f2fs_super(struct super_block *sb) | |
3791 | { | |
3792 | if (sb->s_root) { | |
3793 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
3794 | ||
3795 | set_sbi_flag(sbi, SBI_IS_CLOSE); | |
3796 | f2fs_stop_gc_thread(sbi); | |
3797 | f2fs_stop_discard_thread(sbi); | |
3798 | ||
3799 | if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || | |
3800 | !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { | |
3801 | struct cp_control cpc = { | |
3802 | .reason = CP_UMOUNT, | |
3803 | }; | |
3804 | f2fs_write_checkpoint(sbi, &cpc); | |
3805 | } | |
3806 | ||
3807 | if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) | |
3808 | sb->s_flags &= ~SB_RDONLY; | |
3809 | } | |
3810 | kill_block_super(sb); | |
3811 | } | |
3812 | ||
3813 | static struct file_system_type f2fs_fs_type = { | |
3814 | .owner = THIS_MODULE, | |
3815 | .name = "f2fs", | |
3816 | .mount = f2fs_mount, | |
3817 | .kill_sb = kill_f2fs_super, | |
3818 | .fs_flags = FS_REQUIRES_DEV, | |
3819 | }; | |
3820 | MODULE_ALIAS_FS("f2fs"); | |
3821 | ||
3822 | static int __init init_inodecache(void) | |
3823 | { | |
3824 | f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", | |
3825 | sizeof(struct f2fs_inode_info), 0, | |
3826 | SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); | |
3827 | if (!f2fs_inode_cachep) | |
3828 | return -ENOMEM; | |
3829 | return 0; | |
3830 | } | |
3831 | ||
3832 | static void destroy_inodecache(void) | |
3833 | { | |
3834 | /* | |
3835 | * Make sure all delayed rcu free inodes are flushed before we | |
3836 | * destroy cache. | |
3837 | */ | |
3838 | rcu_barrier(); | |
3839 | kmem_cache_destroy(f2fs_inode_cachep); | |
3840 | } | |
3841 | ||
3842 | static int __init init_f2fs_fs(void) | |
3843 | { | |
3844 | int err; | |
3845 | ||
3846 | if (PAGE_SIZE != F2FS_BLKSIZE) { | |
3847 | printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", | |
3848 | PAGE_SIZE, F2FS_BLKSIZE); | |
3849 | return -EINVAL; | |
3850 | } | |
3851 | ||
3852 | f2fs_build_trace_ios(); | |
3853 | ||
3854 | err = init_inodecache(); | |
3855 | if (err) | |
3856 | goto fail; | |
3857 | err = f2fs_create_node_manager_caches(); | |
3858 | if (err) | |
3859 | goto free_inodecache; | |
3860 | err = f2fs_create_segment_manager_caches(); | |
3861 | if (err) | |
3862 | goto free_node_manager_caches; | |
3863 | err = f2fs_create_checkpoint_caches(); | |
3864 | if (err) | |
3865 | goto free_segment_manager_caches; | |
3866 | err = f2fs_create_extent_cache(); | |
3867 | if (err) | |
3868 | goto free_checkpoint_caches; | |
3869 | err = f2fs_init_sysfs(); | |
3870 | if (err) | |
3871 | goto free_extent_cache; | |
3872 | err = register_shrinker(&f2fs_shrinker_info); | |
3873 | if (err) | |
3874 | goto free_sysfs; | |
3875 | err = register_filesystem(&f2fs_fs_type); | |
3876 | if (err) | |
3877 | goto free_shrinker; | |
3878 | f2fs_create_root_stats(); | |
3879 | err = f2fs_init_post_read_processing(); | |
3880 | if (err) | |
3881 | goto free_root_stats; | |
3882 | err = f2fs_init_bio_entry_cache(); | |
3883 | if (err) | |
3884 | goto free_post_read; | |
3885 | err = f2fs_init_bioset(); | |
3886 | if (err) | |
3887 | goto free_bio_enrty_cache; | |
3888 | err = f2fs_init_compress_mempool(); | |
3889 | if (err) | |
3890 | goto free_bioset; | |
3891 | return 0; | |
3892 | free_bioset: | |
3893 | f2fs_destroy_bioset(); | |
3894 | free_bio_enrty_cache: | |
3895 | f2fs_destroy_bio_entry_cache(); | |
3896 | free_post_read: | |
3897 | f2fs_destroy_post_read_processing(); | |
3898 | free_root_stats: | |
3899 | f2fs_destroy_root_stats(); | |
3900 | unregister_filesystem(&f2fs_fs_type); | |
3901 | free_shrinker: | |
3902 | unregister_shrinker(&f2fs_shrinker_info); | |
3903 | free_sysfs: | |
3904 | f2fs_exit_sysfs(); | |
3905 | free_extent_cache: | |
3906 | f2fs_destroy_extent_cache(); | |
3907 | free_checkpoint_caches: | |
3908 | f2fs_destroy_checkpoint_caches(); | |
3909 | free_segment_manager_caches: | |
3910 | f2fs_destroy_segment_manager_caches(); | |
3911 | free_node_manager_caches: | |
3912 | f2fs_destroy_node_manager_caches(); | |
3913 | free_inodecache: | |
3914 | destroy_inodecache(); | |
3915 | fail: | |
3916 | return err; | |
3917 | } | |
3918 | ||
3919 | static void __exit exit_f2fs_fs(void) | |
3920 | { | |
3921 | f2fs_destroy_compress_mempool(); | |
3922 | f2fs_destroy_bioset(); | |
3923 | f2fs_destroy_bio_entry_cache(); | |
3924 | f2fs_destroy_post_read_processing(); | |
3925 | f2fs_destroy_root_stats(); | |
3926 | unregister_filesystem(&f2fs_fs_type); | |
3927 | unregister_shrinker(&f2fs_shrinker_info); | |
3928 | f2fs_exit_sysfs(); | |
3929 | f2fs_destroy_extent_cache(); | |
3930 | f2fs_destroy_checkpoint_caches(); | |
3931 | f2fs_destroy_segment_manager_caches(); | |
3932 | f2fs_destroy_node_manager_caches(); | |
3933 | destroy_inodecache(); | |
3934 | f2fs_destroy_trace_ios(); | |
3935 | } | |
3936 | ||
3937 | module_init(init_f2fs_fs) | |
3938 | module_exit(exit_f2fs_fs) | |
3939 | ||
3940 | MODULE_AUTHOR("Samsung Electronics's Praesto Team"); | |
3941 | MODULE_DESCRIPTION("Flash Friendly File System"); | |
3942 | MODULE_LICENSE("GPL"); | |
3943 |