]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - fs/f2fs/super.c
projects / thirdparty / kernel / stable.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: fix to do sanity check with current segment number
[thirdparty/kernel/stable.git] / fs / f2fs / super.c
1 /*
2 * fs/f2fs/super.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11 #include <linux/module.h>
12 #include <linux/init.h>
13 #include <linux/fs.h>
14 #include <linux/statfs.h>
15 #include <linux/buffer_head.h>
16 #include <linux/backing-dev.h>
17 #include <linux/kthread.h>
18 #include <linux/parser.h>
19 #include <linux/mount.h>
20 #include <linux/seq_file.h>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/quotaops.h>
26 #include <linux/f2fs_fs.h>
27 #include <linux/sysfs.h>
28 #include <linux/quota.h>
29
30 #include "f2fs.h"
31 #include "node.h"
32 #include "segment.h"
33 #include "xattr.h"
34 #include "gc.h"
35 #include "trace.h"
36
37 #define CREATE_TRACE_POINTS
38 #include <trace/events/f2fs.h>
39
40 static struct kmem_cache *f2fs_inode_cachep;
41
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
43
44 char *fault_name[FAULT_MAX] = {
45 [FAULT_KMALLOC] = "kmalloc",
46 [FAULT_PAGE_ALLOC] = "page alloc",
47 [FAULT_ALLOC_NID] = "alloc nid",
48 [FAULT_ORPHAN] = "orphan",
49 [FAULT_BLOCK] = "no more block",
50 [FAULT_DIR_DEPTH] = "too big dir depth",
51 [FAULT_EVICT_INODE] = "evict_inode fail",
52 [FAULT_TRUNCATE] = "truncate fail",
53 [FAULT_IO] = "IO error",
54 [FAULT_CHECKPOINT] = "checkpoint error",
55 };
56
57 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
58 unsigned int rate)
59 {
60 struct f2fs_fault_info *ffi = &sbi->fault_info;
61
62 if (rate) {
63 atomic_set(&ffi->inject_ops, 0);
64 ffi->inject_rate = rate;
65 ffi->inject_type = (1 << FAULT_MAX) - 1;
66 } else {
67 memset(ffi, 0, sizeof(struct f2fs_fault_info));
68 }
69 }
70 #endif
71
72 /* f2fs-wide shrinker description */
73 static struct shrinker f2fs_shrinker_info = {
74 .scan_objects = f2fs_shrink_scan,
75 .count_objects = f2fs_shrink_count,
76 .seeks = DEFAULT_SEEKS,
77 };
78
79 enum {
80 Opt_gc_background,
81 Opt_disable_roll_forward,
82 Opt_norecovery,
83 Opt_discard,
84 Opt_nodiscard,
85 Opt_noheap,
86 Opt_heap,
87 Opt_user_xattr,
88 Opt_nouser_xattr,
89 Opt_acl,
90 Opt_noacl,
91 Opt_active_logs,
92 Opt_disable_ext_identify,
93 Opt_inline_xattr,
94 Opt_noinline_xattr,
95 Opt_inline_data,
96 Opt_inline_dentry,
97 Opt_noinline_dentry,
98 Opt_flush_merge,
99 Opt_noflush_merge,
100 Opt_nobarrier,
101 Opt_fastboot,
102 Opt_extent_cache,
103 Opt_noextent_cache,
104 Opt_noinline_data,
105 Opt_data_flush,
106 Opt_mode,
107 Opt_io_size_bits,
108 Opt_fault_injection,
109 Opt_lazytime,
110 Opt_nolazytime,
111 Opt_quota,
112 Opt_noquota,
113 Opt_usrquota,
114 Opt_grpquota,
115 Opt_prjquota,
116 Opt_usrjquota,
117 Opt_grpjquota,
118 Opt_prjjquota,
119 Opt_offusrjquota,
120 Opt_offgrpjquota,
121 Opt_offprjjquota,
122 Opt_jqfmt_vfsold,
123 Opt_jqfmt_vfsv0,
124 Opt_jqfmt_vfsv1,
125 Opt_err,
126 };
127
128 static match_table_t f2fs_tokens = {
129 {Opt_gc_background, "background_gc=%s"},
130 {Opt_disable_roll_forward, "disable_roll_forward"},
131 {Opt_norecovery, "norecovery"},
132 {Opt_discard, "discard"},
133 {Opt_nodiscard, "nodiscard"},
134 {Opt_noheap, "no_heap"},
135 {Opt_heap, "heap"},
136 {Opt_user_xattr, "user_xattr"},
137 {Opt_nouser_xattr, "nouser_xattr"},
138 {Opt_acl, "acl"},
139 {Opt_noacl, "noacl"},
140 {Opt_active_logs, "active_logs=%u"},
141 {Opt_disable_ext_identify, "disable_ext_identify"},
142 {Opt_inline_xattr, "inline_xattr"},
143 {Opt_noinline_xattr, "noinline_xattr"},
144 {Opt_inline_data, "inline_data"},
145 {Opt_inline_dentry, "inline_dentry"},
146 {Opt_noinline_dentry, "noinline_dentry"},
147 {Opt_flush_merge, "flush_merge"},
148 {Opt_noflush_merge, "noflush_merge"},
149 {Opt_nobarrier, "nobarrier"},
150 {Opt_fastboot, "fastboot"},
151 {Opt_extent_cache, "extent_cache"},
152 {Opt_noextent_cache, "noextent_cache"},
153 {Opt_noinline_data, "noinline_data"},
154 {Opt_data_flush, "data_flush"},
155 {Opt_mode, "mode=%s"},
156 {Opt_io_size_bits, "io_bits=%u"},
157 {Opt_fault_injection, "fault_injection=%u"},
158 {Opt_lazytime, "lazytime"},
159 {Opt_nolazytime, "nolazytime"},
160 {Opt_quota, "quota"},
161 {Opt_noquota, "noquota"},
162 {Opt_usrquota, "usrquota"},
163 {Opt_grpquota, "grpquota"},
164 {Opt_prjquota, "prjquota"},
165 {Opt_usrjquota, "usrjquota=%s"},
166 {Opt_grpjquota, "grpjquota=%s"},
167 {Opt_prjjquota, "prjjquota=%s"},
168 {Opt_offusrjquota, "usrjquota="},
169 {Opt_offgrpjquota, "grpjquota="},
170 {Opt_offprjjquota, "prjjquota="},
171 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
172 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
173 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
174 {Opt_err, NULL},
175 };
176
177 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
178 {
179 struct va_format vaf;
180 va_list args;
181
182 va_start(args, fmt);
183 vaf.fmt = fmt;
184 vaf.va = &args;
185 printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
186 va_end(args);
187 }
188
189 static void init_once(void *foo)
190 {
191 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
192
193 inode_init_once(&fi->vfs_inode);
194 }
195
196 #ifdef CONFIG_QUOTA
197 static const char * const quotatypes[] = INITQFNAMES;
198 #define QTYPE2NAME(t) (quotatypes[t])
199 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
200 substring_t *args)
201 {
202 struct f2fs_sb_info *sbi = F2FS_SB(sb);
203 char *qname;
204 int ret = -EINVAL;
205
206 if (sb_any_quota_loaded(sb) && !sbi->s_qf_names[qtype]) {
207 f2fs_msg(sb, KERN_ERR,
208 "Cannot change journaled "
209 "quota options when quota turned on");
210 return -EINVAL;
211 }
212 qname = match_strdup(args);
213 if (!qname) {
214 f2fs_msg(sb, KERN_ERR,
215 "Not enough memory for storing quotafile name");
216 return -EINVAL;
217 }
218 if (sbi->s_qf_names[qtype]) {
219 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
220 ret = 0;
221 else
222 f2fs_msg(sb, KERN_ERR,
223 "%s quota file already specified",
224 QTYPE2NAME(qtype));
225 goto errout;
226 }
227 if (strchr(qname, '/')) {
228 f2fs_msg(sb, KERN_ERR,
229 "quotafile must be on filesystem root");
230 goto errout;
231 }
232 sbi->s_qf_names[qtype] = qname;
233 set_opt(sbi, QUOTA);
234 return 0;
235 errout:
236 kfree(qname);
237 return ret;
238 }
239
240 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
241 {
242 struct f2fs_sb_info *sbi = F2FS_SB(sb);
243
244 if (sb_any_quota_loaded(sb) && sbi->s_qf_names[qtype]) {
245 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
246 " when quota turned on");
247 return -EINVAL;
248 }
249 kfree(sbi->s_qf_names[qtype]);
250 sbi->s_qf_names[qtype] = NULL;
251 return 0;
252 }
253
254 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
255 {
256 /*
257 * We do the test below only for project quotas. 'usrquota' and
258 * 'grpquota' mount options are allowed even without quota feature
259 * to support legacy quotas in quota files.
260 */
261 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
262 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
263 "Cannot enable project quota enforcement.");
264 return -1;
265 }
266 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA] ||
267 sbi->s_qf_names[PRJQUOTA]) {
268 if (test_opt(sbi, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
269 clear_opt(sbi, USRQUOTA);
270
271 if (test_opt(sbi, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
272 clear_opt(sbi, GRPQUOTA);
273
274 if (test_opt(sbi, PRJQUOTA) && sbi->s_qf_names[PRJQUOTA])
275 clear_opt(sbi, PRJQUOTA);
276
277 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
278 test_opt(sbi, PRJQUOTA)) {
279 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
280 "format mixing");
281 return -1;
282 }
283
284 if (!sbi->s_jquota_fmt) {
285 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
286 "not specified");
287 return -1;
288 }
289 }
290 return 0;
291 }
292 #endif
293
294 static int parse_options(struct super_block *sb, char *options)
295 {
296 struct f2fs_sb_info *sbi = F2FS_SB(sb);
297 struct request_queue *q;
298 substring_t args[MAX_OPT_ARGS];
299 char *p, *name;
300 int arg = 0;
301 #ifdef CONFIG_QUOTA
302 int ret;
303 #endif
304
305 if (!options)
306 return 0;
307
308 while ((p = strsep(&options, ",")) != NULL) {
309 int token;
310 if (!*p)
311 continue;
312 /*
313 * Initialize args struct so we know whether arg was
314 * found; some options take optional arguments.
315 */
316 args[0].to = args[0].from = NULL;
317 token = match_token(p, f2fs_tokens, args);
318
319 switch (token) {
320 case Opt_gc_background:
321 name = match_strdup(&args[0]);
322
323 if (!name)
324 return -ENOMEM;
325 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
326 set_opt(sbi, BG_GC);
327 clear_opt(sbi, FORCE_FG_GC);
328 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
329 clear_opt(sbi, BG_GC);
330 clear_opt(sbi, FORCE_FG_GC);
331 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
332 set_opt(sbi, BG_GC);
333 set_opt(sbi, FORCE_FG_GC);
334 } else {
335 kfree(name);
336 return -EINVAL;
337 }
338 kfree(name);
339 break;
340 case Opt_disable_roll_forward:
341 set_opt(sbi, DISABLE_ROLL_FORWARD);
342 break;
343 case Opt_norecovery:
344 /* this option mounts f2fs with ro */
345 set_opt(sbi, DISABLE_ROLL_FORWARD);
346 if (!f2fs_readonly(sb))
347 return -EINVAL;
348 break;
349 case Opt_discard:
350 q = bdev_get_queue(sb->s_bdev);
351 if (blk_queue_discard(q)) {
352 set_opt(sbi, DISCARD);
353 } else if (!f2fs_sb_mounted_blkzoned(sb)) {
354 f2fs_msg(sb, KERN_WARNING,
355 "mounting with \"discard\" option, but "
356 "the device does not support discard");
357 }
358 break;
359 case Opt_nodiscard:
360 if (f2fs_sb_mounted_blkzoned(sb)) {
361 f2fs_msg(sb, KERN_WARNING,
362 "discard is required for zoned block devices");
363 return -EINVAL;
364 }
365 clear_opt(sbi, DISCARD);
366 break;
367 case Opt_noheap:
368 set_opt(sbi, NOHEAP);
369 break;
370 case Opt_heap:
371 clear_opt(sbi, NOHEAP);
372 break;
373 #ifdef CONFIG_F2FS_FS_XATTR
374 case Opt_user_xattr:
375 set_opt(sbi, XATTR_USER);
376 break;
377 case Opt_nouser_xattr:
378 clear_opt(sbi, XATTR_USER);
379 break;
380 case Opt_inline_xattr:
381 set_opt(sbi, INLINE_XATTR);
382 break;
383 case Opt_noinline_xattr:
384 clear_opt(sbi, INLINE_XATTR);
385 break;
386 #else
387 case Opt_user_xattr:
388 f2fs_msg(sb, KERN_INFO,
389 "user_xattr options not supported");
390 break;
391 case Opt_nouser_xattr:
392 f2fs_msg(sb, KERN_INFO,
393 "nouser_xattr options not supported");
394 break;
395 case Opt_inline_xattr:
396 f2fs_msg(sb, KERN_INFO,
397 "inline_xattr options not supported");
398 break;
399 case Opt_noinline_xattr:
400 f2fs_msg(sb, KERN_INFO,
401 "noinline_xattr options not supported");
402 break;
403 #endif
404 #ifdef CONFIG_F2FS_FS_POSIX_ACL
405 case Opt_acl:
406 set_opt(sbi, POSIX_ACL);
407 break;
408 case Opt_noacl:
409 clear_opt(sbi, POSIX_ACL);
410 break;
411 #else
412 case Opt_acl:
413 f2fs_msg(sb, KERN_INFO, "acl options not supported");
414 break;
415 case Opt_noacl:
416 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
417 break;
418 #endif
419 case Opt_active_logs:
420 if (args->from && match_int(args, &arg))
421 return -EINVAL;
422 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
423 return -EINVAL;
424 sbi->active_logs = arg;
425 break;
426 case Opt_disable_ext_identify:
427 set_opt(sbi, DISABLE_EXT_IDENTIFY);
428 break;
429 case Opt_inline_data:
430 set_opt(sbi, INLINE_DATA);
431 break;
432 case Opt_inline_dentry:
433 set_opt(sbi, INLINE_DENTRY);
434 break;
435 case Opt_noinline_dentry:
436 clear_opt(sbi, INLINE_DENTRY);
437 break;
438 case Opt_flush_merge:
439 set_opt(sbi, FLUSH_MERGE);
440 break;
441 case Opt_noflush_merge:
442 clear_opt(sbi, FLUSH_MERGE);
443 break;
444 case Opt_nobarrier:
445 set_opt(sbi, NOBARRIER);
446 break;
447 case Opt_fastboot:
448 set_opt(sbi, FASTBOOT);
449 break;
450 case Opt_extent_cache:
451 set_opt(sbi, EXTENT_CACHE);
452 break;
453 case Opt_noextent_cache:
454 clear_opt(sbi, EXTENT_CACHE);
455 break;
456 case Opt_noinline_data:
457 clear_opt(sbi, INLINE_DATA);
458 break;
459 case Opt_data_flush:
460 set_opt(sbi, DATA_FLUSH);
461 break;
462 case Opt_mode:
463 name = match_strdup(&args[0]);
464
465 if (!name)
466 return -ENOMEM;
467 if (strlen(name) == 8 &&
468 !strncmp(name, "adaptive", 8)) {
469 if (f2fs_sb_mounted_blkzoned(sb)) {
470 f2fs_msg(sb, KERN_WARNING,
471 "adaptive mode is not allowed with "
472 "zoned block device feature");
473 kfree(name);
474 return -EINVAL;
475 }
476 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
477 } else if (strlen(name) == 3 &&
478 !strncmp(name, "lfs", 3)) {
479 set_opt_mode(sbi, F2FS_MOUNT_LFS);
480 } else {
481 kfree(name);
482 return -EINVAL;
483 }
484 kfree(name);
485 break;
486 case Opt_io_size_bits:
487 if (args->from && match_int(args, &arg))
488 return -EINVAL;
489 if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
490 f2fs_msg(sb, KERN_WARNING,
491 "Not support %d, larger than %d",
492 1 << arg, BIO_MAX_PAGES);
493 return -EINVAL;
494 }
495 sbi->write_io_size_bits = arg;
496 break;
497 case Opt_fault_injection:
498 if (args->from && match_int(args, &arg))
499 return -EINVAL;
500 #ifdef CONFIG_F2FS_FAULT_INJECTION
501 f2fs_build_fault_attr(sbi, arg);
502 set_opt(sbi, FAULT_INJECTION);
503 #else
504 f2fs_msg(sb, KERN_INFO,
505 "FAULT_INJECTION was not selected");
506 #endif
507 break;
508 case Opt_lazytime:
509 sb->s_flags |= MS_LAZYTIME;
510 break;
511 case Opt_nolazytime:
512 sb->s_flags &= ~MS_LAZYTIME;
513 break;
514 #ifdef CONFIG_QUOTA
515 case Opt_quota:
516 case Opt_usrquota:
517 set_opt(sbi, USRQUOTA);
518 break;
519 case Opt_grpquota:
520 set_opt(sbi, GRPQUOTA);
521 break;
522 case Opt_prjquota:
523 set_opt(sbi, PRJQUOTA);
524 break;
525 case Opt_usrjquota:
526 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
527 if (ret)
528 return ret;
529 break;
530 case Opt_grpjquota:
531 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
532 if (ret)
533 return ret;
534 break;
535 case Opt_prjjquota:
536 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
537 if (ret)
538 return ret;
539 break;
540 case Opt_offusrjquota:
541 ret = f2fs_clear_qf_name(sb, USRQUOTA);
542 if (ret)
543 return ret;
544 break;
545 case Opt_offgrpjquota:
546 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
547 if (ret)
548 return ret;
549 break;
550 case Opt_offprjjquota:
551 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
552 if (ret)
553 return ret;
554 break;
555 case Opt_jqfmt_vfsold:
556 sbi->s_jquota_fmt = QFMT_VFS_OLD;
557 break;
558 case Opt_jqfmt_vfsv0:
559 sbi->s_jquota_fmt = QFMT_VFS_V0;
560 break;
561 case Opt_jqfmt_vfsv1:
562 sbi->s_jquota_fmt = QFMT_VFS_V1;
563 break;
564 case Opt_noquota:
565 clear_opt(sbi, QUOTA);
566 clear_opt(sbi, USRQUOTA);
567 clear_opt(sbi, GRPQUOTA);
568 clear_opt(sbi, PRJQUOTA);
569 break;
570 #else
571 case Opt_quota:
572 case Opt_usrquota:
573 case Opt_grpquota:
574 case Opt_prjquota:
575 case Opt_usrjquota:
576 case Opt_grpjquota:
577 case Opt_prjjquota:
578 case Opt_offusrjquota:
579 case Opt_offgrpjquota:
580 case Opt_offprjjquota:
581 case Opt_jqfmt_vfsold:
582 case Opt_jqfmt_vfsv0:
583 case Opt_jqfmt_vfsv1:
584 case Opt_noquota:
585 f2fs_msg(sb, KERN_INFO,
586 "quota operations not supported");
587 break;
588 #endif
589 default:
590 f2fs_msg(sb, KERN_ERR,
591 "Unrecognized mount option \"%s\" or missing value",
592 p);
593 return -EINVAL;
594 }
595 }
596 #ifdef CONFIG_QUOTA
597 if (f2fs_check_quota_options(sbi))
598 return -EINVAL;
599 #endif
600
601 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
602 f2fs_msg(sb, KERN_ERR,
603 "Should set mode=lfs with %uKB-sized IO",
604 F2FS_IO_SIZE_KB(sbi));
605 return -EINVAL;
606 }
607 return 0;
608 }
609
610 static struct inode *f2fs_alloc_inode(struct super_block *sb)
611 {
612 struct f2fs_inode_info *fi;
613
614 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
615 if (!fi)
616 return NULL;
617
618 init_once((void *) fi);
619
620 /* Initialize f2fs-specific inode info */
621 fi->vfs_inode.i_version = 1;
622 atomic_set(&fi->dirty_pages, 0);
623 fi->i_current_depth = 1;
624 fi->i_advise = 0;
625 init_rwsem(&fi->i_sem);
626 INIT_LIST_HEAD(&fi->dirty_list);
627 INIT_LIST_HEAD(&fi->gdirty_list);
628 INIT_LIST_HEAD(&fi->inmem_pages);
629 mutex_init(&fi->inmem_lock);
630 init_rwsem(&fi->dio_rwsem[READ]);
631 init_rwsem(&fi->dio_rwsem[WRITE]);
632 init_rwsem(&fi->i_mmap_sem);
633 init_rwsem(&fi->i_xattr_sem);
634
635 #ifdef CONFIG_QUOTA
636 memset(&fi->i_dquot, 0, sizeof(fi->i_dquot));
637 fi->i_reserved_quota = 0;
638 #endif
639 /* Will be used by directory only */
640 fi->i_dir_level = F2FS_SB(sb)->dir_level;
641
642 return &fi->vfs_inode;
643 }
644
645 static int f2fs_drop_inode(struct inode *inode)
646 {
647 int ret;
648 /*
649 * This is to avoid a deadlock condition like below.
650 * writeback_single_inode(inode)
651 * - f2fs_write_data_page
652 * - f2fs_gc -> iput -> evict
653 * - inode_wait_for_writeback(inode)
654 */
655 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
656 if (!inode->i_nlink && !is_bad_inode(inode)) {
657 /* to avoid evict_inode call simultaneously */
658 atomic_inc(&inode->i_count);
659 spin_unlock(&inode->i_lock);
660
661 /* some remained atomic pages should discarded */
662 if (f2fs_is_atomic_file(inode))
663 drop_inmem_pages(inode);
664
665 /* should remain fi->extent_tree for writepage */
666 f2fs_destroy_extent_node(inode);
667
668 sb_start_intwrite(inode->i_sb);
669 f2fs_i_size_write(inode, 0);
670
671 if (F2FS_HAS_BLOCKS(inode))
672 f2fs_truncate(inode);
673
674 sb_end_intwrite(inode->i_sb);
675
676 fscrypt_put_encryption_info(inode, NULL);
677 spin_lock(&inode->i_lock);
678 atomic_dec(&inode->i_count);
679 }
680 trace_f2fs_drop_inode(inode, 0);
681 return 0;
682 }
683 ret = generic_drop_inode(inode);
684 trace_f2fs_drop_inode(inode, ret);
685 return ret;
686 }
687
688 int f2fs_inode_dirtied(struct inode *inode, bool sync)
689 {
690 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
691 int ret = 0;
692
693 spin_lock(&sbi->inode_lock[DIRTY_META]);
694 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
695 ret = 1;
696 } else {
697 set_inode_flag(inode, FI_DIRTY_INODE);
698 stat_inc_dirty_inode(sbi, DIRTY_META);
699 }
700 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
701 list_add_tail(&F2FS_I(inode)->gdirty_list,
702 &sbi->inode_list[DIRTY_META]);
703 inc_page_count(sbi, F2FS_DIRTY_IMETA);
704 }
705 spin_unlock(&sbi->inode_lock[DIRTY_META]);
706 return ret;
707 }
708
709 void f2fs_inode_synced(struct inode *inode)
710 {
711 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
712
713 spin_lock(&sbi->inode_lock[DIRTY_META]);
714 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
715 spin_unlock(&sbi->inode_lock[DIRTY_META]);
716 return;
717 }
718 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
719 list_del_init(&F2FS_I(inode)->gdirty_list);
720 dec_page_count(sbi, F2FS_DIRTY_IMETA);
721 }
722 clear_inode_flag(inode, FI_DIRTY_INODE);
723 clear_inode_flag(inode, FI_AUTO_RECOVER);
724 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
725 spin_unlock(&sbi->inode_lock[DIRTY_META]);
726 }
727
728 /*
729 * f2fs_dirty_inode() is called from __mark_inode_dirty()
730 *
731 * We should call set_dirty_inode to write the dirty inode through write_inode.
732 */
733 static void f2fs_dirty_inode(struct inode *inode, int flags)
734 {
735 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
736
737 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
738 inode->i_ino == F2FS_META_INO(sbi))
739 return;
740
741 if (flags == I_DIRTY_TIME)
742 return;
743
744 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
745 clear_inode_flag(inode, FI_AUTO_RECOVER);
746
747 f2fs_inode_dirtied(inode, false);
748 }
749
750 static void f2fs_i_callback(struct rcu_head *head)
751 {
752 struct inode *inode = container_of(head, struct inode, i_rcu);
753 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
754 }
755
756 static void f2fs_destroy_inode(struct inode *inode)
757 {
758 call_rcu(&inode->i_rcu, f2fs_i_callback);
759 }
760
761 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
762 {
763 percpu_counter_destroy(&sbi->alloc_valid_block_count);
764 percpu_counter_destroy(&sbi->total_valid_inode_count);
765 }
766
767 static void destroy_device_list(struct f2fs_sb_info *sbi)
768 {
769 int i;
770
771 for (i = 0; i < sbi->s_ndevs; i++) {
772 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
773 #ifdef CONFIG_BLK_DEV_ZONED
774 kfree(FDEV(i).blkz_type);
775 #endif
776 }
777 kfree(sbi->devs);
778 }
779
780 static void f2fs_put_super(struct super_block *sb)
781 {
782 struct f2fs_sb_info *sbi = F2FS_SB(sb);
783 int i;
784
785 f2fs_quota_off_umount(sb);
786
787 /* prevent remaining shrinker jobs */
788 mutex_lock(&sbi->umount_mutex);
789
790 /*
791 * We don't need to do checkpoint when superblock is clean.
792 * But, the previous checkpoint was not done by umount, it needs to do
793 * clean checkpoint again.
794 */
795 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
796 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
797 struct cp_control cpc = {
798 .reason = CP_UMOUNT,
799 };
800 write_checkpoint(sbi, &cpc);
801 }
802
803 /* be sure to wait for any on-going discard commands */
804 f2fs_wait_discard_bios(sbi, true);
805
806 if (f2fs_discard_en(sbi) && !sbi->discard_blks) {
807 struct cp_control cpc = {
808 .reason = CP_UMOUNT | CP_TRIMMED,
809 };
810 write_checkpoint(sbi, &cpc);
811 }
812
813 /* write_checkpoint can update stat informaion */
814 f2fs_destroy_stats(sbi);
815
816 /*
817 * normally superblock is clean, so we need to release this.
818 * In addition, EIO will skip do checkpoint, we need this as well.
819 */
820 release_ino_entry(sbi, true);
821
822 f2fs_leave_shrinker(sbi);
823 mutex_unlock(&sbi->umount_mutex);
824
825 /* our cp_error case, we can wait for any writeback page */
826 f2fs_flush_merged_writes(sbi);
827
828 iput(sbi->node_inode);
829 iput(sbi->meta_inode);
830
831 /* destroy f2fs internal modules */
832 destroy_node_manager(sbi);
833 destroy_segment_manager(sbi);
834
835 kfree(sbi->ckpt);
836
837 f2fs_unregister_sysfs(sbi);
838
839 sb->s_fs_info = NULL;
840 if (sbi->s_chksum_driver)
841 crypto_free_shash(sbi->s_chksum_driver);
842 kfree(sbi->raw_super);
843
844 destroy_device_list(sbi);
845 mempool_destroy(sbi->write_io_dummy);
846 #ifdef CONFIG_QUOTA
847 for (i = 0; i < MAXQUOTAS; i++)
848 kfree(sbi->s_qf_names[i]);
849 #endif
850 destroy_percpu_info(sbi);
851 for (i = 0; i < NR_PAGE_TYPE; i++)
852 kfree(sbi->write_io[i]);
853 kfree(sbi);
854 }
855
856 int f2fs_sync_fs(struct super_block *sb, int sync)
857 {
858 struct f2fs_sb_info *sbi = F2FS_SB(sb);
859 int err = 0;
860
861 trace_f2fs_sync_fs(sb, sync);
862
863 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
864 return -EAGAIN;
865
866 if (sync) {
867 struct cp_control cpc;
868
869 cpc.reason = __get_cp_reason(sbi);
870
871 mutex_lock(&sbi->gc_mutex);
872 err = write_checkpoint(sbi, &cpc);
873 mutex_unlock(&sbi->gc_mutex);
874 }
875 f2fs_trace_ios(NULL, 1);
876
877 return err;
878 }
879
880 static int f2fs_freeze(struct super_block *sb)
881 {
882 if (f2fs_readonly(sb))
883 return 0;
884
885 /* IO error happened before */
886 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
887 return -EIO;
888
889 /* must be clean, since sync_filesystem() was already called */
890 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
891 return -EINVAL;
892 return 0;
893 }
894
895 static int f2fs_unfreeze(struct super_block *sb)
896 {
897 return 0;
898 }
899
900 #ifdef CONFIG_QUOTA
901 static int f2fs_statfs_project(struct super_block *sb,
902 kprojid_t projid, struct kstatfs *buf)
903 {
904 struct kqid qid;
905 struct dquot *dquot;
906 u64 limit;
907 u64 curblock;
908
909 qid = make_kqid_projid(projid);
910 dquot = dqget(sb, qid);
911 if (IS_ERR(dquot))
912 return PTR_ERR(dquot);
913 spin_lock(&dq_data_lock);
914
915 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
916 dquot->dq_dqb.dqb_bsoftlimit :
917 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
918 if (limit && buf->f_blocks > limit) {
919 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
920 buf->f_blocks = limit;
921 buf->f_bfree = buf->f_bavail =
922 (buf->f_blocks > curblock) ?
923 (buf->f_blocks - curblock) : 0;
924 }
925
926 limit = dquot->dq_dqb.dqb_isoftlimit ?
927 dquot->dq_dqb.dqb_isoftlimit :
928 dquot->dq_dqb.dqb_ihardlimit;
929 if (limit && buf->f_files > limit) {
930 buf->f_files = limit;
931 buf->f_ffree =
932 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
933 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
934 }
935
936 spin_unlock(&dq_data_lock);
937 dqput(dquot);
938 return 0;
939 }
940 #endif
941
942 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
943 {
944 struct super_block *sb = dentry->d_sb;
945 struct f2fs_sb_info *sbi = F2FS_SB(sb);
946 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
947 block_t total_count, user_block_count, start_count, ovp_count;
948 u64 avail_node_count;
949
950 total_count = le64_to_cpu(sbi->raw_super->block_count);
951 user_block_count = sbi->user_block_count;
952 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
953 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
954 buf->f_type = F2FS_SUPER_MAGIC;
955 buf->f_bsize = sbi->blocksize;
956
957 buf->f_blocks = total_count - start_count;
958 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
959 buf->f_bavail = user_block_count - valid_user_blocks(sbi) -
960 sbi->reserved_blocks;
961
962 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
963
964 if (avail_node_count > user_block_count) {
965 buf->f_files = user_block_count;
966 buf->f_ffree = buf->f_bavail;
967 } else {
968 buf->f_files = avail_node_count;
969 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
970 buf->f_bavail);
971 }
972
973 buf->f_namelen = F2FS_NAME_LEN;
974 buf->f_fsid.val[0] = (u32)id;
975 buf->f_fsid.val[1] = (u32)(id >> 32);
976
977 #ifdef CONFIG_QUOTA
978 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
979 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
980 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
981 }
982 #endif
983 return 0;
984 }
985
986 static inline void f2fs_show_quota_options(struct seq_file *seq,
987 struct super_block *sb)
988 {
989 #ifdef CONFIG_QUOTA
990 struct f2fs_sb_info *sbi = F2FS_SB(sb);
991
992 if (sbi->s_jquota_fmt) {
993 char *fmtname = "";
994
995 switch (sbi->s_jquota_fmt) {
996 case QFMT_VFS_OLD:
997 fmtname = "vfsold";
998 break;
999 case QFMT_VFS_V0:
1000 fmtname = "vfsv0";
1001 break;
1002 case QFMT_VFS_V1:
1003 fmtname = "vfsv1";
1004 break;
1005 }
1006 seq_printf(seq, ",jqfmt=%s", fmtname);
1007 }
1008
1009 if (sbi->s_qf_names[USRQUOTA])
1010 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1011
1012 if (sbi->s_qf_names[GRPQUOTA])
1013 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1014
1015 if (sbi->s_qf_names[PRJQUOTA])
1016 seq_show_option(seq, "prjjquota", sbi->s_qf_names[PRJQUOTA]);
1017 #endif
1018 }
1019
1020 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1021 {
1022 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1023
1024 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1025 if (test_opt(sbi, FORCE_FG_GC))
1026 seq_printf(seq, ",background_gc=%s", "sync");
1027 else
1028 seq_printf(seq, ",background_gc=%s", "on");
1029 } else {
1030 seq_printf(seq, ",background_gc=%s", "off");
1031 }
1032 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1033 seq_puts(seq, ",disable_roll_forward");
1034 if (test_opt(sbi, DISCARD))
1035 seq_puts(seq, ",discard");
1036 if (test_opt(sbi, NOHEAP))
1037 seq_puts(seq, ",no_heap");
1038 else
1039 seq_puts(seq, ",heap");
1040 #ifdef CONFIG_F2FS_FS_XATTR
1041 if (test_opt(sbi, XATTR_USER))
1042 seq_puts(seq, ",user_xattr");
1043 else
1044 seq_puts(seq, ",nouser_xattr");
1045 if (test_opt(sbi, INLINE_XATTR))
1046 seq_puts(seq, ",inline_xattr");
1047 else
1048 seq_puts(seq, ",noinline_xattr");
1049 #endif
1050 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1051 if (test_opt(sbi, POSIX_ACL))
1052 seq_puts(seq, ",acl");
1053 else
1054 seq_puts(seq, ",noacl");
1055 #endif
1056 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1057 seq_puts(seq, ",disable_ext_identify");
1058 if (test_opt(sbi, INLINE_DATA))
1059 seq_puts(seq, ",inline_data");
1060 else
1061 seq_puts(seq, ",noinline_data");
1062 if (test_opt(sbi, INLINE_DENTRY))
1063 seq_puts(seq, ",inline_dentry");
1064 else
1065 seq_puts(seq, ",noinline_dentry");
1066 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1067 seq_puts(seq, ",flush_merge");
1068 if (test_opt(sbi, NOBARRIER))
1069 seq_puts(seq, ",nobarrier");
1070 if (test_opt(sbi, FASTBOOT))
1071 seq_puts(seq, ",fastboot");
1072 if (test_opt(sbi, EXTENT_CACHE))
1073 seq_puts(seq, ",extent_cache");
1074 else
1075 seq_puts(seq, ",noextent_cache");
1076 if (test_opt(sbi, DATA_FLUSH))
1077 seq_puts(seq, ",data_flush");
1078
1079 seq_puts(seq, ",mode=");
1080 if (test_opt(sbi, ADAPTIVE))
1081 seq_puts(seq, "adaptive");
1082 else if (test_opt(sbi, LFS))
1083 seq_puts(seq, "lfs");
1084 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
1085 if (F2FS_IO_SIZE_BITS(sbi))
1086 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1087 #ifdef CONFIG_F2FS_FAULT_INJECTION
1088 if (test_opt(sbi, FAULT_INJECTION))
1089 seq_printf(seq, ",fault_injection=%u",
1090 sbi->fault_info.inject_rate);
1091 #endif
1092 #ifdef CONFIG_QUOTA
1093 if (test_opt(sbi, QUOTA))
1094 seq_puts(seq, ",quota");
1095 if (test_opt(sbi, USRQUOTA))
1096 seq_puts(seq, ",usrquota");
1097 if (test_opt(sbi, GRPQUOTA))
1098 seq_puts(seq, ",grpquota");
1099 if (test_opt(sbi, PRJQUOTA))
1100 seq_puts(seq, ",prjquota");
1101 #endif
1102 f2fs_show_quota_options(seq, sbi->sb);
1103
1104 return 0;
1105 }
1106
1107 static void default_options(struct f2fs_sb_info *sbi)
1108 {
1109 /* init some FS parameters */
1110 sbi->active_logs = NR_CURSEG_TYPE;
1111
1112 set_opt(sbi, BG_GC);
1113 set_opt(sbi, INLINE_XATTR);
1114 set_opt(sbi, INLINE_DATA);
1115 set_opt(sbi, INLINE_DENTRY);
1116 set_opt(sbi, EXTENT_CACHE);
1117 set_opt(sbi, NOHEAP);
1118 sbi->sb->s_flags |= MS_LAZYTIME;
1119 set_opt(sbi, FLUSH_MERGE);
1120 if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1121 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1122 set_opt(sbi, DISCARD);
1123 } else {
1124 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1125 }
1126
1127 #ifdef CONFIG_F2FS_FS_XATTR
1128 set_opt(sbi, XATTR_USER);
1129 #endif
1130 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1131 set_opt(sbi, POSIX_ACL);
1132 #endif
1133
1134 #ifdef CONFIG_F2FS_FAULT_INJECTION
1135 f2fs_build_fault_attr(sbi, 0);
1136 #endif
1137 }
1138
1139 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1140 {
1141 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1142 struct f2fs_mount_info org_mount_opt;
1143 unsigned long old_sb_flags;
1144 int err, active_logs;
1145 bool need_restart_gc = false;
1146 bool need_stop_gc = false;
1147 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1148 #ifdef CONFIG_F2FS_FAULT_INJECTION
1149 struct f2fs_fault_info ffi = sbi->fault_info;
1150 #endif
1151 #ifdef CONFIG_QUOTA
1152 int s_jquota_fmt;
1153 char *s_qf_names[MAXQUOTAS];
1154 int i, j;
1155 #endif
1156
1157 /*
1158 * Save the old mount options in case we
1159 * need to restore them.
1160 */
1161 org_mount_opt = sbi->mount_opt;
1162 old_sb_flags = sb->s_flags;
1163 active_logs = sbi->active_logs;
1164
1165 #ifdef CONFIG_QUOTA
1166 s_jquota_fmt = sbi->s_jquota_fmt;
1167 for (i = 0; i < MAXQUOTAS; i++) {
1168 if (sbi->s_qf_names[i]) {
1169 s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
1170 GFP_KERNEL);
1171 if (!s_qf_names[i]) {
1172 for (j = 0; j < i; j++)
1173 kfree(s_qf_names[j]);
1174 return -ENOMEM;
1175 }
1176 } else {
1177 s_qf_names[i] = NULL;
1178 }
1179 }
1180 #endif
1181
1182 /* recover superblocks we couldn't write due to previous RO mount */
1183 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1184 err = f2fs_commit_super(sbi, false);
1185 f2fs_msg(sb, KERN_INFO,
1186 "Try to recover all the superblocks, ret: %d", err);
1187 if (!err)
1188 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1189 }
1190
1191 default_options(sbi);
1192
1193 /* parse mount options */
1194 err = parse_options(sb, data);
1195 if (err)
1196 goto restore_opts;
1197
1198 /*
1199 * Previous and new state of filesystem is RO,
1200 * so skip checking GC and FLUSH_MERGE conditions.
1201 */
1202 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1203 goto skip;
1204
1205 if (!f2fs_readonly(sb) && (*flags & MS_RDONLY)) {
1206 err = dquot_suspend(sb, -1);
1207 if (err < 0)
1208 goto restore_opts;
1209 } else {
1210 /* dquot_resume needs RW */
1211 sb->s_flags &= ~MS_RDONLY;
1212 dquot_resume(sb, -1);
1213 }
1214
1215 /* disallow enable/disable extent_cache dynamically */
1216 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1217 err = -EINVAL;
1218 f2fs_msg(sbi->sb, KERN_WARNING,
1219 "switch extent_cache option is not allowed");
1220 goto restore_opts;
1221 }
1222
1223 /*
1224 * We stop the GC thread if FS is mounted as RO
1225 * or if background_gc = off is passed in mount
1226 * option. Also sync the filesystem.
1227 */
1228 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1229 if (sbi->gc_thread) {
1230 stop_gc_thread(sbi);
1231 need_restart_gc = true;
1232 }
1233 } else if (!sbi->gc_thread) {
1234 err = start_gc_thread(sbi);
1235 if (err)
1236 goto restore_opts;
1237 need_stop_gc = true;
1238 }
1239
1240 if (*flags & MS_RDONLY) {
1241 writeback_inodes_sb(sb, WB_REASON_SYNC);
1242 sync_inodes_sb(sb);
1243
1244 set_sbi_flag(sbi, SBI_IS_DIRTY);
1245 set_sbi_flag(sbi, SBI_IS_CLOSE);
1246 f2fs_sync_fs(sb, 1);
1247 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1248 }
1249
1250 /*
1251 * We stop issue flush thread if FS is mounted as RO
1252 * or if flush_merge is not passed in mount option.
1253 */
1254 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1255 clear_opt(sbi, FLUSH_MERGE);
1256 destroy_flush_cmd_control(sbi, false);
1257 } else {
1258 err = create_flush_cmd_control(sbi);
1259 if (err)
1260 goto restore_gc;
1261 }
1262 skip:
1263 #ifdef CONFIG_QUOTA
1264 /* Release old quota file names */
1265 for (i = 0; i < MAXQUOTAS; i++)
1266 kfree(s_qf_names[i]);
1267 #endif
1268 /* Update the POSIXACL Flag */
1269 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1270 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1271
1272 return 0;
1273 restore_gc:
1274 if (need_restart_gc) {
1275 if (start_gc_thread(sbi))
1276 f2fs_msg(sbi->sb, KERN_WARNING,
1277 "background gc thread has stopped");
1278 } else if (need_stop_gc) {
1279 stop_gc_thread(sbi);
1280 }
1281 restore_opts:
1282 #ifdef CONFIG_QUOTA
1283 sbi->s_jquota_fmt = s_jquota_fmt;
1284 for (i = 0; i < MAXQUOTAS; i++) {
1285 kfree(sbi->s_qf_names[i]);
1286 sbi->s_qf_names[i] = s_qf_names[i];
1287 }
1288 #endif
1289 sbi->mount_opt = org_mount_opt;
1290 sbi->active_logs = active_logs;
1291 sb->s_flags = old_sb_flags;
1292 #ifdef CONFIG_F2FS_FAULT_INJECTION
1293 sbi->fault_info = ffi;
1294 #endif
1295 return err;
1296 }
1297
1298 #ifdef CONFIG_QUOTA
1299 /* Read data from quotafile */
1300 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1301 size_t len, loff_t off)
1302 {
1303 struct inode *inode = sb_dqopt(sb)->files[type];
1304 struct address_space *mapping = inode->i_mapping;
1305 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1306 int offset = off & (sb->s_blocksize - 1);
1307 int tocopy;
1308 size_t toread;
1309 loff_t i_size = i_size_read(inode);
1310 struct page *page;
1311 char *kaddr;
1312
1313 if (off > i_size)
1314 return 0;
1315
1316 if (off + len > i_size)
1317 len = i_size - off;
1318 toread = len;
1319 while (toread > 0) {
1320 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1321 repeat:
1322 page = read_mapping_page(mapping, blkidx, NULL);
1323 if (IS_ERR(page))
1324 return PTR_ERR(page);
1325
1326 lock_page(page);
1327
1328 if (unlikely(page->mapping != mapping)) {
1329 f2fs_put_page(page, 1);
1330 goto repeat;
1331 }
1332 if (unlikely(!PageUptodate(page))) {
1333 f2fs_put_page(page, 1);
1334 return -EIO;
1335 }
1336
1337 kaddr = kmap_atomic(page);
1338 memcpy(data, kaddr + offset, tocopy);
1339 kunmap_atomic(kaddr);
1340 f2fs_put_page(page, 1);
1341
1342 offset = 0;
1343 toread -= tocopy;
1344 data += tocopy;
1345 blkidx++;
1346 }
1347 return len;
1348 }
1349
1350 /* Write to quotafile */
1351 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1352 const char *data, size_t len, loff_t off)
1353 {
1354 struct inode *inode = sb_dqopt(sb)->files[type];
1355 struct address_space *mapping = inode->i_mapping;
1356 const struct address_space_operations *a_ops = mapping->a_ops;
1357 int offset = off & (sb->s_blocksize - 1);
1358 size_t towrite = len;
1359 struct page *page;
1360 char *kaddr;
1361 int err = 0;
1362 int tocopy;
1363
1364 while (towrite > 0) {
1365 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1366 towrite);
1367
1368 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1369 &page, NULL);
1370 if (unlikely(err))
1371 break;
1372
1373 kaddr = kmap_atomic(page);
1374 memcpy(kaddr + offset, data, tocopy);
1375 kunmap_atomic(kaddr);
1376 flush_dcache_page(page);
1377
1378 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1379 page, NULL);
1380 offset = 0;
1381 towrite -= tocopy;
1382 off += tocopy;
1383 data += tocopy;
1384 cond_resched();
1385 }
1386
1387 if (len == towrite)
1388 return 0;
1389 inode->i_version++;
1390 inode->i_mtime = inode->i_ctime = current_time(inode);
1391 f2fs_mark_inode_dirty_sync(inode, false);
1392 return len - towrite;
1393 }
1394
1395 static struct dquot **f2fs_get_dquots(struct inode *inode)
1396 {
1397 return F2FS_I(inode)->i_dquot;
1398 }
1399
1400 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1401 {
1402 return &F2FS_I(inode)->i_reserved_quota;
1403 }
1404
1405 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1406 {
1407 return dquot_quota_on_mount(sbi->sb, sbi->s_qf_names[type],
1408 sbi->s_jquota_fmt, type);
1409 }
1410
1411 void f2fs_enable_quota_files(struct f2fs_sb_info *sbi)
1412 {
1413 int i, ret;
1414
1415 for (i = 0; i < MAXQUOTAS; i++) {
1416 if (sbi->s_qf_names[i]) {
1417 ret = f2fs_quota_on_mount(sbi, i);
1418 if (ret < 0)
1419 f2fs_msg(sbi->sb, KERN_ERR,
1420 "Cannot turn on journaled "
1421 "quota: error %d", ret);
1422 }
1423 }
1424 }
1425
1426 static int f2fs_quota_sync(struct super_block *sb, int type)
1427 {
1428 struct quota_info *dqopt = sb_dqopt(sb);
1429 int cnt;
1430 int ret;
1431
1432 ret = dquot_writeback_dquots(sb, type);
1433 if (ret)
1434 return ret;
1435
1436 /*
1437 * Now when everything is written we can discard the pagecache so
1438 * that userspace sees the changes.
1439 */
1440 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1441 if (type != -1 && cnt != type)
1442 continue;
1443 if (!sb_has_quota_active(sb, cnt))
1444 continue;
1445
1446 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1447 if (ret)
1448 return ret;
1449
1450 inode_lock(dqopt->files[cnt]);
1451 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1452 inode_unlock(dqopt->files[cnt]);
1453 }
1454 return 0;
1455 }
1456
1457 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1458 const struct path *path)
1459 {
1460 struct inode *inode;
1461 int err;
1462
1463 err = f2fs_quota_sync(sb, type);
1464 if (err)
1465 return err;
1466
1467 err = dquot_quota_on(sb, type, format_id, path);
1468 if (err)
1469 return err;
1470
1471 inode = d_inode(path->dentry);
1472
1473 inode_lock(inode);
1474 F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1475 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1476 S_NOATIME | S_IMMUTABLE);
1477 inode_unlock(inode);
1478 f2fs_mark_inode_dirty_sync(inode, false);
1479
1480 return 0;
1481 }
1482
1483 static int f2fs_quota_off(struct super_block *sb, int type)
1484 {
1485 struct inode *inode = sb_dqopt(sb)->files[type];
1486 int err;
1487
1488 if (!inode || !igrab(inode))
1489 return dquot_quota_off(sb, type);
1490
1491 err = f2fs_quota_sync(sb, type);
1492 if (err)
1493 goto out_put;
1494
1495 err = dquot_quota_off(sb, type);
1496 if (err)
1497 goto out_put;
1498
1499 inode_lock(inode);
1500 F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1501 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1502 inode_unlock(inode);
1503 f2fs_mark_inode_dirty_sync(inode, false);
1504 out_put:
1505 iput(inode);
1506 return err;
1507 }
1508
1509 void f2fs_quota_off_umount(struct super_block *sb)
1510 {
1511 int type;
1512 int err;
1513
1514 for (type = 0; type < MAXQUOTAS; type++) {
1515 err = f2fs_quota_off(sb, type);
1516 if (err) {
1517 int ret = dquot_quota_off(sb, type);
1518
1519 f2fs_msg(sb, KERN_ERR,
1520 "Fail to turn off disk quota "
1521 "(type: %d, err: %d, ret:%d), Please "
1522 "run fsck to fix it.", type, err, ret);
1523 set_sbi_flag(F2FS_SB(sb), SBI_NEED_FSCK);
1524 }
1525 }
1526 }
1527
1528 int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1529 {
1530 *projid = F2FS_I(inode)->i_projid;
1531 return 0;
1532 }
1533
1534 static const struct dquot_operations f2fs_quota_operations = {
1535 .get_reserved_space = f2fs_get_reserved_space,
1536 .write_dquot = dquot_commit,
1537 .acquire_dquot = dquot_acquire,
1538 .release_dquot = dquot_release,
1539 .mark_dirty = dquot_mark_dquot_dirty,
1540 .write_info = dquot_commit_info,
1541 .alloc_dquot = dquot_alloc,
1542 .destroy_dquot = dquot_destroy,
1543 .get_projid = f2fs_get_projid,
1544 .get_next_id = dquot_get_next_id,
1545 };
1546
1547 static const struct quotactl_ops f2fs_quotactl_ops = {
1548 .quota_on = f2fs_quota_on,
1549 .quota_off = f2fs_quota_off,
1550 .quota_sync = f2fs_quota_sync,
1551 .get_state = dquot_get_state,
1552 .set_info = dquot_set_dqinfo,
1553 .get_dqblk = dquot_get_dqblk,
1554 .set_dqblk = dquot_set_dqblk,
1555 .get_nextdqblk = dquot_get_next_dqblk,
1556 };
1557 #else
1558 void f2fs_quota_off_umount(struct super_block *sb)
1559 {
1560 }
1561 #endif
1562
1563 static const struct super_operations f2fs_sops = {
1564 .alloc_inode = f2fs_alloc_inode,
1565 .drop_inode = f2fs_drop_inode,
1566 .destroy_inode = f2fs_destroy_inode,
1567 .write_inode = f2fs_write_inode,
1568 .dirty_inode = f2fs_dirty_inode,
1569 .show_options = f2fs_show_options,
1570 #ifdef CONFIG_QUOTA
1571 .quota_read = f2fs_quota_read,
1572 .quota_write = f2fs_quota_write,
1573 .get_dquots = f2fs_get_dquots,
1574 #endif
1575 .evict_inode = f2fs_evict_inode,
1576 .put_super = f2fs_put_super,
1577 .sync_fs = f2fs_sync_fs,
1578 .freeze_fs = f2fs_freeze,
1579 .unfreeze_fs = f2fs_unfreeze,
1580 .statfs = f2fs_statfs,
1581 .remount_fs = f2fs_remount,
1582 };
1583
1584 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1585 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1586 {
1587 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1588 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1589 ctx, len, NULL);
1590 }
1591
1592 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1593 void *fs_data)
1594 {
1595 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1596 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1597 ctx, len, fs_data, XATTR_CREATE);
1598 }
1599
1600 static unsigned f2fs_max_namelen(struct inode *inode)
1601 {
1602 return S_ISLNK(inode->i_mode) ?
1603 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1604 }
1605
1606 static const struct fscrypt_operations f2fs_cryptops = {
1607 .key_prefix = "f2fs:",
1608 .get_context = f2fs_get_context,
1609 .set_context = f2fs_set_context,
1610 .is_encrypted = f2fs_encrypted_inode,
1611 .empty_dir = f2fs_empty_dir,
1612 .max_namelen = f2fs_max_namelen,
1613 };
1614 #else
1615 static const struct fscrypt_operations f2fs_cryptops = {
1616 .is_encrypted = f2fs_encrypted_inode,
1617 };
1618 #endif
1619
1620 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1621 u64 ino, u32 generation)
1622 {
1623 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1624 struct inode *inode;
1625
1626 if (check_nid_range(sbi, ino))
1627 return ERR_PTR(-ESTALE);
1628
1629 /*
1630 * f2fs_iget isn't quite right if the inode is currently unallocated!
1631 * However f2fs_iget currently does appropriate checks to handle stale
1632 * inodes so everything is OK.
1633 */
1634 inode = f2fs_iget(sb, ino);
1635 if (IS_ERR(inode))
1636 return ERR_CAST(inode);
1637 if (unlikely(generation && inode->i_generation != generation)) {
1638 /* we didn't find the right inode.. */
1639 iput(inode);
1640 return ERR_PTR(-ESTALE);
1641 }
1642 return inode;
1643 }
1644
1645 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1646 int fh_len, int fh_type)
1647 {
1648 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1649 f2fs_nfs_get_inode);
1650 }
1651
1652 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1653 int fh_len, int fh_type)
1654 {
1655 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1656 f2fs_nfs_get_inode);
1657 }
1658
1659 static const struct export_operations f2fs_export_ops = {
1660 .fh_to_dentry = f2fs_fh_to_dentry,
1661 .fh_to_parent = f2fs_fh_to_parent,
1662 .get_parent = f2fs_get_parent,
1663 };
1664
1665 static loff_t max_file_blocks(void)
1666 {
1667 loff_t result = 0;
1668 loff_t leaf_count = ADDRS_PER_BLOCK;
1669
1670 /*
1671 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1672 * F2FS_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1673 * space in inode.i_addr, it will be more safe to reassign
1674 * result as zero.
1675 */
1676
1677 /* two direct node blocks */
1678 result += (leaf_count * 2);
1679
1680 /* two indirect node blocks */
1681 leaf_count *= NIDS_PER_BLOCK;
1682 result += (leaf_count * 2);
1683
1684 /* one double indirect node block */
1685 leaf_count *= NIDS_PER_BLOCK;
1686 result += leaf_count;
1687
1688 return result;
1689 }
1690
1691 static int __f2fs_commit_super(struct buffer_head *bh,
1692 struct f2fs_super_block *super)
1693 {
1694 lock_buffer(bh);
1695 if (super)
1696 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1697 set_buffer_uptodate(bh);
1698 set_buffer_dirty(bh);
1699 unlock_buffer(bh);
1700
1701 /* it's rare case, we can do fua all the time */
1702 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
1703 }
1704
1705 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1706 struct buffer_head *bh)
1707 {
1708 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1709 (bh->b_data + F2FS_SUPER_OFFSET);
1710 struct super_block *sb = sbi->sb;
1711 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1712 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1713 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1714 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1715 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1716 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1717 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1718 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1719 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1720 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1721 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1722 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1723 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1724 u64 main_end_blkaddr = main_blkaddr +
1725 (segment_count_main << log_blocks_per_seg);
1726 u64 seg_end_blkaddr = segment0_blkaddr +
1727 (segment_count << log_blocks_per_seg);
1728
1729 if (segment0_blkaddr != cp_blkaddr) {
1730 f2fs_msg(sb, KERN_INFO,
1731 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1732 segment0_blkaddr, cp_blkaddr);
1733 return true;
1734 }
1735
1736 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1737 sit_blkaddr) {
1738 f2fs_msg(sb, KERN_INFO,
1739 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1740 cp_blkaddr, sit_blkaddr,
1741 segment_count_ckpt << log_blocks_per_seg);
1742 return true;
1743 }
1744
1745 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1746 nat_blkaddr) {
1747 f2fs_msg(sb, KERN_INFO,
1748 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1749 sit_blkaddr, nat_blkaddr,
1750 segment_count_sit << log_blocks_per_seg);
1751 return true;
1752 }
1753
1754 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1755 ssa_blkaddr) {
1756 f2fs_msg(sb, KERN_INFO,
1757 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1758 nat_blkaddr, ssa_blkaddr,
1759 segment_count_nat << log_blocks_per_seg);
1760 return true;
1761 }
1762
1763 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1764 main_blkaddr) {
1765 f2fs_msg(sb, KERN_INFO,
1766 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1767 ssa_blkaddr, main_blkaddr,
1768 segment_count_ssa << log_blocks_per_seg);
1769 return true;
1770 }
1771
1772 if (main_end_blkaddr > seg_end_blkaddr) {
1773 f2fs_msg(sb, KERN_INFO,
1774 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1775 main_blkaddr,
1776 segment0_blkaddr +
1777 (segment_count << log_blocks_per_seg),
1778 segment_count_main << log_blocks_per_seg);
1779 return true;
1780 } else if (main_end_blkaddr < seg_end_blkaddr) {
1781 int err = 0;
1782 char *res;
1783
1784 /* fix in-memory information all the time */
1785 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1786 segment0_blkaddr) >> log_blocks_per_seg);
1787
1788 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1789 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1790 res = "internally";
1791 } else {
1792 err = __f2fs_commit_super(bh, NULL);
1793 res = err ? "failed" : "done";
1794 }
1795 f2fs_msg(sb, KERN_INFO,
1796 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1797 res, main_blkaddr,
1798 segment0_blkaddr +
1799 (segment_count << log_blocks_per_seg),
1800 segment_count_main << log_blocks_per_seg);
1801 if (err)
1802 return true;
1803 }
1804 return false;
1805 }
1806
1807 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1808 struct buffer_head *bh)
1809 {
1810 block_t segment_count, segs_per_sec, secs_per_zone;
1811 block_t total_sections, blocks_per_seg;
1812 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1813 (bh->b_data + F2FS_SUPER_OFFSET);
1814 struct super_block *sb = sbi->sb;
1815 unsigned int blocksize;
1816
1817 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1818 f2fs_msg(sb, KERN_INFO,
1819 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1820 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1821 return 1;
1822 }
1823
1824 /* Currently, support only 4KB page cache size */
1825 if (F2FS_BLKSIZE != PAGE_SIZE) {
1826 f2fs_msg(sb, KERN_INFO,
1827 "Invalid page_cache_size (%lu), supports only 4KB\n",
1828 PAGE_SIZE);
1829 return 1;
1830 }
1831
1832 /* Currently, support only 4KB block size */
1833 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1834 if (blocksize != F2FS_BLKSIZE) {
1835 f2fs_msg(sb, KERN_INFO,
1836 "Invalid blocksize (%u), supports only 4KB\n",
1837 blocksize);
1838 return 1;
1839 }
1840
1841 /* check log blocks per segment */
1842 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1843 f2fs_msg(sb, KERN_INFO,
1844 "Invalid log blocks per segment (%u)\n",
1845 le32_to_cpu(raw_super->log_blocks_per_seg));
1846 return 1;
1847 }
1848
1849 /* Currently, support 512/1024/2048/4096 bytes sector size */
1850 if (le32_to_cpu(raw_super->log_sectorsize) >
1851 F2FS_MAX_LOG_SECTOR_SIZE ||
1852 le32_to_cpu(raw_super->log_sectorsize) <
1853 F2FS_MIN_LOG_SECTOR_SIZE) {
1854 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1855 le32_to_cpu(raw_super->log_sectorsize));
1856 return 1;
1857 }
1858 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1859 le32_to_cpu(raw_super->log_sectorsize) !=
1860 F2FS_MAX_LOG_SECTOR_SIZE) {
1861 f2fs_msg(sb, KERN_INFO,
1862 "Invalid log sectors per block(%u) log sectorsize(%u)",
1863 le32_to_cpu(raw_super->log_sectors_per_block),
1864 le32_to_cpu(raw_super->log_sectorsize));
1865 return 1;
1866 }
1867
1868 segment_count = le32_to_cpu(raw_super->segment_count);
1869 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1870 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1871 total_sections = le32_to_cpu(raw_super->section_count);
1872
1873 /* blocks_per_seg should be 512, given the above check */
1874 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
1875
1876 if (segment_count > F2FS_MAX_SEGMENT ||
1877 segment_count < F2FS_MIN_SEGMENTS) {
1878 f2fs_msg(sb, KERN_INFO,
1879 "Invalid segment count (%u)",
1880 segment_count);
1881 return 1;
1882 }
1883
1884 if (total_sections > segment_count ||
1885 total_sections < F2FS_MIN_SEGMENTS ||
1886 segs_per_sec > segment_count || !segs_per_sec) {
1887 f2fs_msg(sb, KERN_INFO,
1888 "Invalid segment/section count (%u, %u x %u)",
1889 segment_count, total_sections, segs_per_sec);
1890 return 1;
1891 }
1892
1893 if ((segment_count / segs_per_sec) < total_sections) {
1894 f2fs_msg(sb, KERN_INFO,
1895 "Small segment_count (%u < %u * %u)",
1896 segment_count, segs_per_sec, total_sections);
1897 return 1;
1898 }
1899
1900 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
1901 f2fs_msg(sb, KERN_INFO,
1902 "Wrong segment_count / block_count (%u > %llu)",
1903 segment_count, le64_to_cpu(raw_super->block_count));
1904 return 1;
1905 }
1906
1907 if (secs_per_zone > total_sections || !secs_per_zone) {
1908 f2fs_msg(sb, KERN_INFO,
1909 "Wrong secs_per_zone / total_sections (%u, %u)",
1910 secs_per_zone, total_sections);
1911 return 1;
1912 }
1913 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION) {
1914 f2fs_msg(sb, KERN_INFO,
1915 "Corrupted extension count (%u > %u)",
1916 le32_to_cpu(raw_super->extension_count),
1917 F2FS_MAX_EXTENSION);
1918 return 1;
1919 }
1920
1921 if (le32_to_cpu(raw_super->cp_payload) >
1922 (blocks_per_seg - F2FS_CP_PACKS)) {
1923 f2fs_msg(sb, KERN_INFO,
1924 "Insane cp_payload (%u > %u)",
1925 le32_to_cpu(raw_super->cp_payload),
1926 blocks_per_seg - F2FS_CP_PACKS);
1927 return 1;
1928 }
1929
1930 /* check reserved ino info */
1931 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1932 le32_to_cpu(raw_super->meta_ino) != 2 ||
1933 le32_to_cpu(raw_super->root_ino) != 3) {
1934 f2fs_msg(sb, KERN_INFO,
1935 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1936 le32_to_cpu(raw_super->node_ino),
1937 le32_to_cpu(raw_super->meta_ino),
1938 le32_to_cpu(raw_super->root_ino));
1939 return 1;
1940 }
1941
1942 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1943 if (sanity_check_area_boundary(sbi, bh))
1944 return 1;
1945
1946 return 0;
1947 }
1948
1949 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1950 {
1951 unsigned int total, fsmeta;
1952 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1953 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1954 unsigned int ovp_segments, reserved_segments;
1955 unsigned int main_segs, blocks_per_seg;
1956 unsigned int sit_segs, nat_segs;
1957 unsigned int sit_bitmap_size, nat_bitmap_size;
1958 unsigned int log_blocks_per_seg;
1959 unsigned int segment_count_main;
1960 unsigned int cp_pack_start_sum, cp_payload;
1961 block_t user_block_count;
1962 int i, j;
1963
1964 total = le32_to_cpu(raw_super->segment_count);
1965 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1966 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
1967 fsmeta += sit_segs;
1968 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
1969 fsmeta += nat_segs;
1970 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1971 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1972
1973 if (unlikely(fsmeta >= total))
1974 return 1;
1975
1976 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1977 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1978
1979 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1980 ovp_segments == 0 || reserved_segments == 0)) {
1981 f2fs_msg(sbi->sb, KERN_ERR,
1982 "Wrong layout: check mkfs.f2fs version");
1983 return 1;
1984 }
1985
1986 user_block_count = le64_to_cpu(ckpt->user_block_count);
1987 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1988 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1989 if (!user_block_count || user_block_count >=
1990 segment_count_main << log_blocks_per_seg) {
1991 f2fs_msg(sbi->sb, KERN_ERR,
1992 "Wrong user_block_count: %u", user_block_count);
1993 return 1;
1994 }
1995
1996 main_segs = le32_to_cpu(raw_super->segment_count_main);
1997 blocks_per_seg = sbi->blocks_per_seg;
1998
1999 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2000 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2001 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2002 return 1;
2003 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2004 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2005 le32_to_cpu(ckpt->cur_node_segno[j])) {
2006 f2fs_msg(sbi->sb, KERN_ERR,
2007 "Node segment (%u, %u) has the same "
2008 "segno: %u", i, j,
2009 le32_to_cpu(ckpt->cur_node_segno[i]));
2010 return 1;
2011 }
2012 }
2013 }
2014 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2015 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2016 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2017 return 1;
2018 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2019 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2020 le32_to_cpu(ckpt->cur_data_segno[j])) {
2021 f2fs_msg(sbi->sb, KERN_ERR,
2022 "Data segment (%u, %u) has the same "
2023 "segno: %u", i, j,
2024 le32_to_cpu(ckpt->cur_data_segno[i]));
2025 return 1;
2026 }
2027 }
2028 }
2029 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2030 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2031 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2032 le32_to_cpu(ckpt->cur_data_segno[j])) {
2033 f2fs_msg(sbi->sb, KERN_ERR,
2034 "Data segment (%u) and Data segment (%u)"
2035 " has the same segno: %u", i, j,
2036 le32_to_cpu(ckpt->cur_node_segno[i]));
2037 return 1;
2038 }
2039 }
2040 }
2041
2042 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2043 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2044
2045 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2046 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2047 f2fs_msg(sbi->sb, KERN_ERR,
2048 "Wrong bitmap size: sit: %u, nat:%u",
2049 sit_bitmap_size, nat_bitmap_size);
2050 return 1;
2051 }
2052
2053 cp_pack_start_sum = __start_sum_addr(sbi);
2054 cp_payload = __cp_payload(sbi);
2055 if (cp_pack_start_sum < cp_payload + 1 ||
2056 cp_pack_start_sum > blocks_per_seg - 1 -
2057 NR_CURSEG_TYPE) {
2058 f2fs_msg(sbi->sb, KERN_ERR,
2059 "Wrong cp_pack_start_sum: %u",
2060 cp_pack_start_sum);
2061 return 1;
2062 }
2063
2064 if (unlikely(f2fs_cp_error(sbi))) {
2065 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2066 return 1;
2067 }
2068 return 0;
2069 }
2070
2071 static void init_sb_info(struct f2fs_sb_info *sbi)
2072 {
2073 struct f2fs_super_block *raw_super = sbi->raw_super;
2074 int i, j;
2075
2076 sbi->log_sectors_per_block =
2077 le32_to_cpu(raw_super->log_sectors_per_block);
2078 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2079 sbi->blocksize = 1 << sbi->log_blocksize;
2080 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2081 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2082 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2083 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2084 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2085 sbi->total_node_count =
2086 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2087 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2088 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2089 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2090 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2091 sbi->cur_victim_sec = NULL_SECNO;
2092 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2093
2094 sbi->dir_level = DEF_DIR_LEVEL;
2095 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2096 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2097 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2098
2099 for (i = 0; i < NR_COUNT_TYPE; i++)
2100 atomic_set(&sbi->nr_pages[i], 0);
2101
2102 atomic_set(&sbi->wb_sync_req, 0);
2103
2104 INIT_LIST_HEAD(&sbi->s_list);
2105 mutex_init(&sbi->umount_mutex);
2106 for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2107 for (j = HOT; j < NR_TEMP_TYPE; j++)
2108 mutex_init(&sbi->wio_mutex[i][j]);
2109 spin_lock_init(&sbi->cp_lock);
2110 }
2111
2112 static int init_percpu_info(struct f2fs_sb_info *sbi)
2113 {
2114 int err;
2115
2116 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2117 if (err)
2118 return err;
2119
2120 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2121 GFP_KERNEL);
2122 }
2123
2124 #ifdef CONFIG_BLK_DEV_ZONED
2125 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2126 {
2127 struct block_device *bdev = FDEV(devi).bdev;
2128 sector_t nr_sectors = bdev->bd_part->nr_sects;
2129 sector_t sector = 0;
2130 struct blk_zone *zones;
2131 unsigned int i, nr_zones;
2132 unsigned int n = 0;
2133 int err = -EIO;
2134
2135 if (!f2fs_sb_mounted_blkzoned(sbi->sb))
2136 return 0;
2137
2138 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2139 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2140 return -EINVAL;
2141 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2142 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2143 __ilog2_u32(sbi->blocks_per_blkz))
2144 return -EINVAL;
2145 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2146 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2147 sbi->log_blocks_per_blkz;
2148 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2149 FDEV(devi).nr_blkz++;
2150
2151 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
2152 if (!FDEV(devi).blkz_type)
2153 return -ENOMEM;
2154
2155 #define F2FS_REPORT_NR_ZONES 4096
2156
2157 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
2158 GFP_KERNEL);
2159 if (!zones)
2160 return -ENOMEM;
2161
2162 /* Get block zones type */
2163 while (zones && sector < nr_sectors) {
2164
2165 nr_zones = F2FS_REPORT_NR_ZONES;
2166 err = blkdev_report_zones(bdev, sector,
2167 zones, &nr_zones,
2168 GFP_KERNEL);
2169 if (err)
2170 break;
2171 if (!nr_zones) {
2172 err = -EIO;
2173 break;
2174 }
2175
2176 for (i = 0; i < nr_zones; i++) {
2177 FDEV(devi).blkz_type[n] = zones[i].type;
2178 sector += zones[i].len;
2179 n++;
2180 }
2181 }
2182
2183 kfree(zones);
2184
2185 return err;
2186 }
2187 #endif
2188
2189 /*
2190 * Read f2fs raw super block.
2191 * Because we have two copies of super block, so read both of them
2192 * to get the first valid one. If any one of them is broken, we pass
2193 * them recovery flag back to the caller.
2194 */
2195 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2196 struct f2fs_super_block **raw_super,
2197 int *valid_super_block, int *recovery)
2198 {
2199 struct super_block *sb = sbi->sb;
2200 int block;
2201 struct buffer_head *bh;
2202 struct f2fs_super_block *super;
2203 int err = 0;
2204
2205 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2206 if (!super)
2207 return -ENOMEM;
2208
2209 for (block = 0; block < 2; block++) {
2210 bh = sb_bread(sb, block);
2211 if (!bh) {
2212 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2213 block + 1);
2214 err = -EIO;
2215 continue;
2216 }
2217
2218 /* sanity checking of raw super */
2219 if (sanity_check_raw_super(sbi, bh)) {
2220 f2fs_msg(sb, KERN_ERR,
2221 "Can't find valid F2FS filesystem in %dth superblock",
2222 block + 1);
2223 err = -EINVAL;
2224 brelse(bh);
2225 continue;
2226 }
2227
2228 if (!*raw_super) {
2229 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2230 sizeof(*super));
2231 *valid_super_block = block;
2232 *raw_super = super;
2233 }
2234 brelse(bh);
2235 }
2236
2237 /* Fail to read any one of the superblocks*/
2238 if (err < 0)
2239 *recovery = 1;
2240
2241 /* No valid superblock */
2242 if (!*raw_super)
2243 kfree(super);
2244 else
2245 err = 0;
2246
2247 return err;
2248 }
2249
2250 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2251 {
2252 struct buffer_head *bh;
2253 int err;
2254
2255 if ((recover && f2fs_readonly(sbi->sb)) ||
2256 bdev_read_only(sbi->sb->s_bdev)) {
2257 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2258 return -EROFS;
2259 }
2260
2261 /* write back-up superblock first */
2262 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
2263 if (!bh)
2264 return -EIO;
2265 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2266 brelse(bh);
2267
2268 /* if we are in recovery path, skip writing valid superblock */
2269 if (recover || err)
2270 return err;
2271
2272 /* write current valid superblock */
2273 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
2274 if (!bh)
2275 return -EIO;
2276 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2277 brelse(bh);
2278 return err;
2279 }
2280
2281 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2282 {
2283 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2284 unsigned int max_devices = MAX_DEVICES;
2285 int i;
2286
2287 /* Initialize single device information */
2288 if (!RDEV(0).path[0]) {
2289 if (!bdev_is_zoned(sbi->sb->s_bdev))
2290 return 0;
2291 max_devices = 1;
2292 }
2293
2294 /*
2295 * Initialize multiple devices information, or single
2296 * zoned block device information.
2297 */
2298 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
2299 GFP_KERNEL);
2300 if (!sbi->devs)
2301 return -ENOMEM;
2302
2303 for (i = 0; i < max_devices; i++) {
2304
2305 if (i > 0 && !RDEV(i).path[0])
2306 break;
2307
2308 if (max_devices == 1) {
2309 /* Single zoned block device mount */
2310 FDEV(0).bdev =
2311 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2312 sbi->sb->s_mode, sbi->sb->s_type);
2313 } else {
2314 /* Multi-device mount */
2315 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2316 FDEV(i).total_segments =
2317 le32_to_cpu(RDEV(i).total_segments);
2318 if (i == 0) {
2319 FDEV(i).start_blk = 0;
2320 FDEV(i).end_blk = FDEV(i).start_blk +
2321 (FDEV(i).total_segments <<
2322 sbi->log_blocks_per_seg) - 1 +
2323 le32_to_cpu(raw_super->segment0_blkaddr);
2324 } else {
2325 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2326 FDEV(i).end_blk = FDEV(i).start_blk +
2327 (FDEV(i).total_segments <<
2328 sbi->log_blocks_per_seg) - 1;
2329 }
2330 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2331 sbi->sb->s_mode, sbi->sb->s_type);
2332 }
2333 if (IS_ERR(FDEV(i).bdev))
2334 return PTR_ERR(FDEV(i).bdev);
2335
2336 /* to release errored devices */
2337 sbi->s_ndevs = i + 1;
2338
2339 #ifdef CONFIG_BLK_DEV_ZONED
2340 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2341 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
2342 f2fs_msg(sbi->sb, KERN_ERR,
2343 "Zoned block device feature not enabled\n");
2344 return -EINVAL;
2345 }
2346 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2347 if (init_blkz_info(sbi, i)) {
2348 f2fs_msg(sbi->sb, KERN_ERR,
2349 "Failed to initialize F2FS blkzone information");
2350 return -EINVAL;
2351 }
2352 if (max_devices == 1)
2353 break;
2354 f2fs_msg(sbi->sb, KERN_INFO,
2355 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2356 i, FDEV(i).path,
2357 FDEV(i).total_segments,
2358 FDEV(i).start_blk, FDEV(i).end_blk,
2359 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2360 "Host-aware" : "Host-managed");
2361 continue;
2362 }
2363 #endif
2364 f2fs_msg(sbi->sb, KERN_INFO,
2365 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2366 i, FDEV(i).path,
2367 FDEV(i).total_segments,
2368 FDEV(i).start_blk, FDEV(i).end_blk);
2369 }
2370 f2fs_msg(sbi->sb, KERN_INFO,
2371 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2372 return 0;
2373 }
2374
2375 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2376 {
2377 struct f2fs_sb_info *sbi;
2378 struct f2fs_super_block *raw_super;
2379 struct inode *root;
2380 int err;
2381 bool retry = true, need_fsck = false;
2382 char *options = NULL;
2383 int recovery, i, valid_super_block;
2384 struct curseg_info *seg_i;
2385
2386 try_onemore:
2387 err = -EINVAL;
2388 raw_super = NULL;
2389 valid_super_block = -1;
2390 recovery = 0;
2391
2392 /* allocate memory for f2fs-specific super block info */
2393 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2394 if (!sbi)
2395 return -ENOMEM;
2396
2397 sbi->sb = sb;
2398
2399 /* Load the checksum driver */
2400 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2401 if (IS_ERR(sbi->s_chksum_driver)) {
2402 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2403 err = PTR_ERR(sbi->s_chksum_driver);
2404 sbi->s_chksum_driver = NULL;
2405 goto free_sbi;
2406 }
2407
2408 /* set a block size */
2409 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2410 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2411 goto free_sbi;
2412 }
2413
2414 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2415 &recovery);
2416 if (err)
2417 goto free_sbi;
2418
2419 sb->s_fs_info = sbi;
2420 sbi->raw_super = raw_super;
2421
2422 /* precompute checksum seed for metadata */
2423 if (f2fs_sb_has_inode_chksum(sb))
2424 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2425 sizeof(raw_super->uuid));
2426
2427 /*
2428 * The BLKZONED feature indicates that the drive was formatted with
2429 * zone alignment optimization. This is optional for host-aware
2430 * devices, but mandatory for host-managed zoned block devices.
2431 */
2432 #ifndef CONFIG_BLK_DEV_ZONED
2433 if (f2fs_sb_mounted_blkzoned(sb)) {
2434 f2fs_msg(sb, KERN_ERR,
2435 "Zoned block device support is not enabled\n");
2436 err = -EOPNOTSUPP;
2437 goto free_sb_buf;
2438 }
2439 #endif
2440 default_options(sbi);
2441 /* parse mount options */
2442 options = kstrdup((const char *)data, GFP_KERNEL);
2443 if (data && !options) {
2444 err = -ENOMEM;
2445 goto free_sb_buf;
2446 }
2447
2448 err = parse_options(sb, options);
2449 if (err)
2450 goto free_options;
2451
2452 sbi->max_file_blocks = max_file_blocks();
2453 sb->s_maxbytes = sbi->max_file_blocks <<
2454 le32_to_cpu(raw_super->log_blocksize);
2455 sb->s_max_links = F2FS_LINK_MAX;
2456 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2457
2458 #ifdef CONFIG_QUOTA
2459 sb->dq_op = &f2fs_quota_operations;
2460 sb->s_qcop = &f2fs_quotactl_ops;
2461 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2462 #endif
2463
2464 sb->s_op = &f2fs_sops;
2465 sb->s_cop = &f2fs_cryptops;
2466 sb->s_xattr = f2fs_xattr_handlers;
2467 sb->s_export_op = &f2fs_export_ops;
2468 sb->s_magic = F2FS_SUPER_MAGIC;
2469 sb->s_time_gran = 1;
2470 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2471 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
2472 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2473
2474 /* init f2fs-specific super block info */
2475 sbi->valid_super_block = valid_super_block;
2476 mutex_init(&sbi->gc_mutex);
2477 mutex_init(&sbi->cp_mutex);
2478 init_rwsem(&sbi->node_write);
2479 init_rwsem(&sbi->node_change);
2480
2481 /* disallow all the data/node/meta page writes */
2482 set_sbi_flag(sbi, SBI_POR_DOING);
2483 spin_lock_init(&sbi->stat_lock);
2484
2485 /* init iostat info */
2486 spin_lock_init(&sbi->iostat_lock);
2487 sbi->iostat_enable = false;
2488
2489 for (i = 0; i < NR_PAGE_TYPE; i++) {
2490 int n = (i == META) ? 1: NR_TEMP_TYPE;
2491 int j;
2492
2493 sbi->write_io[i] = kmalloc(n * sizeof(struct f2fs_bio_info),
2494 GFP_KERNEL);
2495 if (!sbi->write_io[i]) {
2496 err = -ENOMEM;
2497 goto free_options;
2498 }
2499
2500 for (j = HOT; j < n; j++) {
2501 init_rwsem(&sbi->write_io[i][j].io_rwsem);
2502 sbi->write_io[i][j].sbi = sbi;
2503 sbi->write_io[i][j].bio = NULL;
2504 spin_lock_init(&sbi->write_io[i][j].io_lock);
2505 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2506 }
2507 }
2508
2509 init_rwsem(&sbi->cp_rwsem);
2510 init_waitqueue_head(&sbi->cp_wait);
2511 init_sb_info(sbi);
2512
2513 err = init_percpu_info(sbi);
2514 if (err)
2515 goto free_options;
2516
2517 if (F2FS_IO_SIZE(sbi) > 1) {
2518 sbi->write_io_dummy =
2519 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2520 if (!sbi->write_io_dummy) {
2521 err = -ENOMEM;
2522 goto free_options;
2523 }
2524 }
2525
2526 /* get an inode for meta space */
2527 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2528 if (IS_ERR(sbi->meta_inode)) {
2529 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2530 err = PTR_ERR(sbi->meta_inode);
2531 goto free_io_dummy;
2532 }
2533
2534 err = get_valid_checkpoint(sbi);
2535 if (err) {
2536 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2537 goto free_meta_inode;
2538 }
2539
2540 /* Initialize device list */
2541 err = f2fs_scan_devices(sbi);
2542 if (err) {
2543 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2544 goto free_devices;
2545 }
2546
2547 sbi->total_valid_node_count =
2548 le32_to_cpu(sbi->ckpt->valid_node_count);
2549 percpu_counter_set(&sbi->total_valid_inode_count,
2550 le32_to_cpu(sbi->ckpt->valid_inode_count));
2551 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2552 sbi->total_valid_block_count =
2553 le64_to_cpu(sbi->ckpt->valid_block_count);
2554 sbi->last_valid_block_count = sbi->total_valid_block_count;
2555 sbi->reserved_blocks = 0;
2556
2557 for (i = 0; i < NR_INODE_TYPE; i++) {
2558 INIT_LIST_HEAD(&sbi->inode_list[i]);
2559 spin_lock_init(&sbi->inode_lock[i]);
2560 }
2561
2562 init_extent_cache_info(sbi);
2563
2564 init_ino_entry_info(sbi);
2565
2566 /* setup f2fs internal modules */
2567 err = build_segment_manager(sbi);
2568 if (err) {
2569 f2fs_msg(sb, KERN_ERR,
2570 "Failed to initialize F2FS segment manager");
2571 goto free_sm;
2572 }
2573 err = build_node_manager(sbi);
2574 if (err) {
2575 f2fs_msg(sb, KERN_ERR,
2576 "Failed to initialize F2FS node manager");
2577 goto free_nm;
2578 }
2579
2580 /* For write statistics */
2581 if (sb->s_bdev->bd_part)
2582 sbi->sectors_written_start =
2583 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2584
2585 /* Read accumulated write IO statistics if exists */
2586 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2587 if (__exist_node_summaries(sbi))
2588 sbi->kbytes_written =
2589 le64_to_cpu(seg_i->journal->info.kbytes_written);
2590
2591 build_gc_manager(sbi);
2592
2593 /* get an inode for node space */
2594 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2595 if (IS_ERR(sbi->node_inode)) {
2596 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2597 err = PTR_ERR(sbi->node_inode);
2598 goto free_nm;
2599 }
2600
2601 f2fs_join_shrinker(sbi);
2602
2603 err = f2fs_build_stats(sbi);
2604 if (err)
2605 goto free_nm;
2606
2607 /* read root inode and dentry */
2608 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2609 if (IS_ERR(root)) {
2610 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2611 err = PTR_ERR(root);
2612 goto free_node_inode;
2613 }
2614 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2615 iput(root);
2616 err = -EINVAL;
2617 goto free_node_inode;
2618 }
2619
2620 sb->s_root = d_make_root(root); /* allocate root dentry */
2621 if (!sb->s_root) {
2622 err = -ENOMEM;
2623 goto free_root_inode;
2624 }
2625
2626 err = f2fs_register_sysfs(sbi);
2627 if (err)
2628 goto free_root_inode;
2629
2630 /* if there are nt orphan nodes free them */
2631 err = recover_orphan_inodes(sbi);
2632 if (err)
2633 goto free_sysfs;
2634
2635 /* recover fsynced data */
2636 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2637 /*
2638 * mount should be failed, when device has readonly mode, and
2639 * previous checkpoint was not done by clean system shutdown.
2640 */
2641 if (bdev_read_only(sb->s_bdev) &&
2642 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2643 err = -EROFS;
2644 goto free_meta;
2645 }
2646
2647 if (need_fsck)
2648 set_sbi_flag(sbi, SBI_NEED_FSCK);
2649
2650 if (!retry)
2651 goto skip_recovery;
2652
2653 err = recover_fsync_data(sbi, false);
2654 if (err < 0) {
2655 need_fsck = true;
2656 f2fs_msg(sb, KERN_ERR,
2657 "Cannot recover all fsync data errno=%d", err);
2658 goto free_meta;
2659 }
2660 } else {
2661 err = recover_fsync_data(sbi, true);
2662
2663 if (!f2fs_readonly(sb) && err > 0) {
2664 err = -EINVAL;
2665 f2fs_msg(sb, KERN_ERR,
2666 "Need to recover fsync data");
2667 goto free_sysfs;
2668 }
2669 }
2670 skip_recovery:
2671 /* recover_fsync_data() cleared this already */
2672 clear_sbi_flag(sbi, SBI_POR_DOING);
2673
2674 /*
2675 * If filesystem is not mounted as read-only then
2676 * do start the gc_thread.
2677 */
2678 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2679 /* After POR, we can run background GC thread.*/
2680 err = start_gc_thread(sbi);
2681 if (err)
2682 goto free_meta;
2683 }
2684 kfree(options);
2685
2686 /* recover broken superblock */
2687 if (recovery) {
2688 err = f2fs_commit_super(sbi, true);
2689 f2fs_msg(sb, KERN_INFO,
2690 "Try to recover %dth superblock, ret: %d",
2691 sbi->valid_super_block ? 1 : 2, err);
2692 }
2693
2694 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2695 cur_cp_version(F2FS_CKPT(sbi)));
2696 f2fs_update_time(sbi, CP_TIME);
2697 f2fs_update_time(sbi, REQ_TIME);
2698 return 0;
2699
2700 free_meta:
2701 f2fs_sync_inode_meta(sbi);
2702 /*
2703 * Some dirty meta pages can be produced by recover_orphan_inodes()
2704 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2705 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2706 * falls into an infinite loop in sync_meta_pages().
2707 */
2708 truncate_inode_pages_final(META_MAPPING(sbi));
2709 free_sysfs:
2710 f2fs_unregister_sysfs(sbi);
2711 free_root_inode:
2712 dput(sb->s_root);
2713 sb->s_root = NULL;
2714 free_node_inode:
2715 truncate_inode_pages_final(NODE_MAPPING(sbi));
2716 mutex_lock(&sbi->umount_mutex);
2717 release_ino_entry(sbi, true);
2718 f2fs_leave_shrinker(sbi);
2719 iput(sbi->node_inode);
2720 mutex_unlock(&sbi->umount_mutex);
2721 f2fs_destroy_stats(sbi);
2722 free_nm:
2723 destroy_node_manager(sbi);
2724 free_sm:
2725 destroy_segment_manager(sbi);
2726 free_devices:
2727 destroy_device_list(sbi);
2728 kfree(sbi->ckpt);
2729 free_meta_inode:
2730 make_bad_inode(sbi->meta_inode);
2731 iput(sbi->meta_inode);
2732 free_io_dummy:
2733 mempool_destroy(sbi->write_io_dummy);
2734 free_options:
2735 for (i = 0; i < NR_PAGE_TYPE; i++)
2736 kfree(sbi->write_io[i]);
2737 destroy_percpu_info(sbi);
2738 #ifdef CONFIG_QUOTA
2739 for (i = 0; i < MAXQUOTAS; i++)
2740 kfree(sbi->s_qf_names[i]);
2741 #endif
2742 kfree(options);
2743 free_sb_buf:
2744 kfree(raw_super);
2745 free_sbi:
2746 if (sbi->s_chksum_driver)
2747 crypto_free_shash(sbi->s_chksum_driver);
2748 kfree(sbi);
2749
2750 /* give only one another chance */
2751 if (retry) {
2752 retry = false;
2753 shrink_dcache_sb(sb);
2754 goto try_onemore;
2755 }
2756 return err;
2757 }
2758
2759 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2760 const char *dev_name, void *data)
2761 {
2762 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2763 }
2764
2765 static void kill_f2fs_super(struct super_block *sb)
2766 {
2767 if (sb->s_root) {
2768 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2769 stop_gc_thread(F2FS_SB(sb));
2770 stop_discard_thread(F2FS_SB(sb));
2771 }
2772 kill_block_super(sb);
2773 }
2774
2775 static struct file_system_type f2fs_fs_type = {
2776 .owner = THIS_MODULE,
2777 .name = "f2fs",
2778 .mount = f2fs_mount,
2779 .kill_sb = kill_f2fs_super,
2780 .fs_flags = FS_REQUIRES_DEV,
2781 };
2782 MODULE_ALIAS_FS("f2fs");
2783
2784 static int __init init_inodecache(void)
2785 {
2786 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2787 sizeof(struct f2fs_inode_info), 0,
2788 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2789 if (!f2fs_inode_cachep)
2790 return -ENOMEM;
2791 return 0;
2792 }
2793
2794 static void destroy_inodecache(void)
2795 {
2796 /*
2797 * Make sure all delayed rcu free inodes are flushed before we
2798 * destroy cache.
2799 */
2800 rcu_barrier();
2801 kmem_cache_destroy(f2fs_inode_cachep);
2802 }
2803
2804 static int __init init_f2fs_fs(void)
2805 {
2806 int err;
2807
2808 if (PAGE_SIZE != F2FS_BLKSIZE) {
2809 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
2810 PAGE_SIZE, F2FS_BLKSIZE);
2811 return -EINVAL;
2812 }
2813
2814 f2fs_build_trace_ios();
2815
2816 err = init_inodecache();
2817 if (err)
2818 goto fail;
2819 err = create_node_manager_caches();
2820 if (err)
2821 goto free_inodecache;
2822 err = create_segment_manager_caches();
2823 if (err)
2824 goto free_node_manager_caches;
2825 err = create_checkpoint_caches();
2826 if (err)
2827 goto free_segment_manager_caches;
2828 err = create_extent_cache();
2829 if (err)
2830 goto free_checkpoint_caches;
2831 err = f2fs_init_sysfs();
2832 if (err)
2833 goto free_extent_cache;
2834 err = register_shrinker(&f2fs_shrinker_info);
2835 if (err)
2836 goto free_sysfs;
2837 err = register_filesystem(&f2fs_fs_type);
2838 if (err)
2839 goto free_shrinker;
2840 err = f2fs_create_root_stats();
2841 if (err)
2842 goto free_filesystem;
2843 return 0;
2844
2845 free_filesystem:
2846 unregister_filesystem(&f2fs_fs_type);
2847 free_shrinker:
2848 unregister_shrinker(&f2fs_shrinker_info);
2849 free_sysfs:
2850 f2fs_exit_sysfs();
2851 free_extent_cache:
2852 destroy_extent_cache();
2853 free_checkpoint_caches:
2854 destroy_checkpoint_caches();
2855 free_segment_manager_caches:
2856 destroy_segment_manager_caches();
2857 free_node_manager_caches:
2858 destroy_node_manager_caches();
2859 free_inodecache:
2860 destroy_inodecache();
2861 fail:
2862 return err;
2863 }
2864
2865 static void __exit exit_f2fs_fs(void)
2866 {
2867 f2fs_destroy_root_stats();
2868 unregister_filesystem(&f2fs_fs_type);
2869 unregister_shrinker(&f2fs_shrinker_info);
2870 f2fs_exit_sysfs();
2871 destroy_extent_cache();
2872 destroy_checkpoint_caches();
2873 destroy_segment_manager_caches();
2874 destroy_node_manager_caches();
2875 destroy_inodecache();
2876 f2fs_destroy_trace_ios();
2877 }
2878
2879 module_init(init_f2fs_fs)
2880 module_exit(exit_f2fs_fs)
2881
2882 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2883 MODULE_DESCRIPTION("Flash Friendly File System");
2884 MODULE_LICENSE("GPL");
2885
Mirror https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git