]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - fs/f2fs/super.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: dax: fix races between page faults and truncating pages
[mirror_ubuntu-jammy-kernel.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/f2fs_fs.h>
26 #include <linux/sysfs.h>
27
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "gc.h"
33 #include "trace.h"
34
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
37
38 static struct proc_dir_entry *f2fs_proc_root;
39 static struct kmem_cache *f2fs_inode_cachep;
40 static struct kset *f2fs_kset;
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_err,
112 };
113
114 static match_table_t f2fs_tokens = {
115 {Opt_gc_background, "background_gc=%s"},
116 {Opt_disable_roll_forward, "disable_roll_forward"},
117 {Opt_norecovery, "norecovery"},
118 {Opt_discard, "discard"},
119 {Opt_nodiscard, "nodiscard"},
120 {Opt_noheap, "no_heap"},
121 {Opt_heap, "heap"},
122 {Opt_user_xattr, "user_xattr"},
123 {Opt_nouser_xattr, "nouser_xattr"},
124 {Opt_acl, "acl"},
125 {Opt_noacl, "noacl"},
126 {Opt_active_logs, "active_logs=%u"},
127 {Opt_disable_ext_identify, "disable_ext_identify"},
128 {Opt_inline_xattr, "inline_xattr"},
129 {Opt_noinline_xattr, "noinline_xattr"},
130 {Opt_inline_data, "inline_data"},
131 {Opt_inline_dentry, "inline_dentry"},
132 {Opt_noinline_dentry, "noinline_dentry"},
133 {Opt_flush_merge, "flush_merge"},
134 {Opt_noflush_merge, "noflush_merge"},
135 {Opt_nobarrier, "nobarrier"},
136 {Opt_fastboot, "fastboot"},
137 {Opt_extent_cache, "extent_cache"},
138 {Opt_noextent_cache, "noextent_cache"},
139 {Opt_noinline_data, "noinline_data"},
140 {Opt_data_flush, "data_flush"},
141 {Opt_mode, "mode=%s"},
142 {Opt_io_size_bits, "io_bits=%u"},
143 {Opt_fault_injection, "fault_injection=%u"},
144 {Opt_lazytime, "lazytime"},
145 {Opt_nolazytime, "nolazytime"},
146 {Opt_err, NULL},
147 };
148
149 /* Sysfs support for f2fs */
150 enum {
151 GC_THREAD, /* struct f2fs_gc_thread */
152 SM_INFO, /* struct f2fs_sm_info */
153 DCC_INFO, /* struct discard_cmd_control */
154 NM_INFO, /* struct f2fs_nm_info */
155 F2FS_SBI, /* struct f2fs_sb_info */
156 #ifdef CONFIG_F2FS_FAULT_INJECTION
157 FAULT_INFO_RATE, /* struct f2fs_fault_info */
158 FAULT_INFO_TYPE, /* struct f2fs_fault_info */
159 #endif
160 };
161
162 struct f2fs_attr {
163 struct attribute attr;
164 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
165 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
166 const char *, size_t);
167 int struct_type;
168 int offset;
169 };
170
171 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
172 {
173 if (struct_type == GC_THREAD)
174 return (unsigned char *)sbi->gc_thread;
175 else if (struct_type == SM_INFO)
176 return (unsigned char *)SM_I(sbi);
177 else if (struct_type == DCC_INFO)
178 return (unsigned char *)SM_I(sbi)->dcc_info;
179 else if (struct_type == NM_INFO)
180 return (unsigned char *)NM_I(sbi);
181 else if (struct_type == F2FS_SBI)
182 return (unsigned char *)sbi;
183 #ifdef CONFIG_F2FS_FAULT_INJECTION
184 else if (struct_type == FAULT_INFO_RATE ||
185 struct_type == FAULT_INFO_TYPE)
186 return (unsigned char *)&sbi->fault_info;
187 #endif
188 return NULL;
189 }
190
191 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
192 struct f2fs_sb_info *sbi, char *buf)
193 {
194 struct super_block *sb = sbi->sb;
195
196 if (!sb->s_bdev->bd_part)
197 return snprintf(buf, PAGE_SIZE, "0\n");
198
199 return snprintf(buf, PAGE_SIZE, "%llu\n",
200 (unsigned long long)(sbi->kbytes_written +
201 BD_PART_WRITTEN(sbi)));
202 }
203
204 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
205 struct f2fs_sb_info *sbi, char *buf)
206 {
207 unsigned char *ptr = NULL;
208 unsigned int *ui;
209
210 ptr = __struct_ptr(sbi, a->struct_type);
211 if (!ptr)
212 return -EINVAL;
213
214 ui = (unsigned int *)(ptr + a->offset);
215
216 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
217 }
218
219 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
220 struct f2fs_sb_info *sbi,
221 const char *buf, size_t count)
222 {
223 unsigned char *ptr;
224 unsigned long t;
225 unsigned int *ui;
226 ssize_t ret;
227
228 ptr = __struct_ptr(sbi, a->struct_type);
229 if (!ptr)
230 return -EINVAL;
231
232 ui = (unsigned int *)(ptr + a->offset);
233
234 ret = kstrtoul(skip_spaces(buf), 0, &t);
235 if (ret < 0)
236 return ret;
237 #ifdef CONFIG_F2FS_FAULT_INJECTION
238 if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
239 return -EINVAL;
240 #endif
241 *ui = t;
242 return count;
243 }
244
245 static ssize_t f2fs_attr_show(struct kobject *kobj,
246 struct attribute *attr, char *buf)
247 {
248 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
249 s_kobj);
250 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
251
252 return a->show ? a->show(a, sbi, buf) : 0;
253 }
254
255 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
256 const char *buf, size_t len)
257 {
258 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
259 s_kobj);
260 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
261
262 return a->store ? a->store(a, sbi, buf, len) : 0;
263 }
264
265 static void f2fs_sb_release(struct kobject *kobj)
266 {
267 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
268 s_kobj);
269 complete(&sbi->s_kobj_unregister);
270 }
271
272 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
273 static struct f2fs_attr f2fs_attr_##_name = { \
274 .attr = {.name = __stringify(_name), .mode = _mode }, \
275 .show = _show, \
276 .store = _store, \
277 .struct_type = _struct_type, \
278 .offset = _offset \
279 }
280
281 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
282 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
283 f2fs_sbi_show, f2fs_sbi_store, \
284 offsetof(struct struct_name, elname))
285
286 #define F2FS_GENERAL_RO_ATTR(name) \
287 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
288
289 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
290 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
291 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
292 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
293 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
294 F2FS_RW_ATTR(DCC_INFO, discard_cmd_control, max_small_discards, max_discards);
295 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
296 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
297 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
298 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
299 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_hot_blocks, min_hot_blocks);
300 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
301 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
302 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
303 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
304 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
305 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
306 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
307 #ifdef CONFIG_F2FS_FAULT_INJECTION
308 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
309 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
310 #endif
311 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
312
313 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
314 static struct attribute *f2fs_attrs[] = {
315 ATTR_LIST(gc_min_sleep_time),
316 ATTR_LIST(gc_max_sleep_time),
317 ATTR_LIST(gc_no_gc_sleep_time),
318 ATTR_LIST(gc_idle),
319 ATTR_LIST(reclaim_segments),
320 ATTR_LIST(max_small_discards),
321 ATTR_LIST(batched_trim_sections),
322 ATTR_LIST(ipu_policy),
323 ATTR_LIST(min_ipu_util),
324 ATTR_LIST(min_fsync_blocks),
325 ATTR_LIST(min_hot_blocks),
326 ATTR_LIST(max_victim_search),
327 ATTR_LIST(dir_level),
328 ATTR_LIST(ram_thresh),
329 ATTR_LIST(ra_nid_pages),
330 ATTR_LIST(dirty_nats_ratio),
331 ATTR_LIST(cp_interval),
332 ATTR_LIST(idle_interval),
333 #ifdef CONFIG_F2FS_FAULT_INJECTION
334 ATTR_LIST(inject_rate),
335 ATTR_LIST(inject_type),
336 #endif
337 ATTR_LIST(lifetime_write_kbytes),
338 NULL,
339 };
340
341 static const struct sysfs_ops f2fs_attr_ops = {
342 .show = f2fs_attr_show,
343 .store = f2fs_attr_store,
344 };
345
346 static struct kobj_type f2fs_ktype = {
347 .default_attrs = f2fs_attrs,
348 .sysfs_ops = &f2fs_attr_ops,
349 .release = f2fs_sb_release,
350 };
351
352 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
353 {
354 struct va_format vaf;
355 va_list args;
356
357 va_start(args, fmt);
358 vaf.fmt = fmt;
359 vaf.va = &args;
360 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
361 va_end(args);
362 }
363
364 static void init_once(void *foo)
365 {
366 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
367
368 inode_init_once(&fi->vfs_inode);
369 }
370
371 static int parse_options(struct super_block *sb, char *options)
372 {
373 struct f2fs_sb_info *sbi = F2FS_SB(sb);
374 struct request_queue *q;
375 substring_t args[MAX_OPT_ARGS];
376 char *p, *name;
377 int arg = 0;
378
379 if (!options)
380 return 0;
381
382 while ((p = strsep(&options, ",")) != NULL) {
383 int token;
384 if (!*p)
385 continue;
386 /*
387 * Initialize args struct so we know whether arg was
388 * found; some options take optional arguments.
389 */
390 args[0].to = args[0].from = NULL;
391 token = match_token(p, f2fs_tokens, args);
392
393 switch (token) {
394 case Opt_gc_background:
395 name = match_strdup(&args[0]);
396
397 if (!name)
398 return -ENOMEM;
399 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
400 set_opt(sbi, BG_GC);
401 clear_opt(sbi, FORCE_FG_GC);
402 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
403 clear_opt(sbi, BG_GC);
404 clear_opt(sbi, FORCE_FG_GC);
405 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
406 set_opt(sbi, BG_GC);
407 set_opt(sbi, FORCE_FG_GC);
408 } else {
409 kfree(name);
410 return -EINVAL;
411 }
412 kfree(name);
413 break;
414 case Opt_disable_roll_forward:
415 set_opt(sbi, DISABLE_ROLL_FORWARD);
416 break;
417 case Opt_norecovery:
418 /* this option mounts f2fs with ro */
419 set_opt(sbi, DISABLE_ROLL_FORWARD);
420 if (!f2fs_readonly(sb))
421 return -EINVAL;
422 break;
423 case Opt_discard:
424 q = bdev_get_queue(sb->s_bdev);
425 if (blk_queue_discard(q)) {
426 set_opt(sbi, DISCARD);
427 } else if (!f2fs_sb_mounted_blkzoned(sb)) {
428 f2fs_msg(sb, KERN_WARNING,
429 "mounting with \"discard\" option, but "
430 "the device does not support discard");
431 }
432 break;
433 case Opt_nodiscard:
434 if (f2fs_sb_mounted_blkzoned(sb)) {
435 f2fs_msg(sb, KERN_WARNING,
436 "discard is required for zoned block devices");
437 return -EINVAL;
438 }
439 clear_opt(sbi, DISCARD);
440 break;
441 case Opt_noheap:
442 set_opt(sbi, NOHEAP);
443 break;
444 case Opt_heap:
445 clear_opt(sbi, NOHEAP);
446 break;
447 #ifdef CONFIG_F2FS_FS_XATTR
448 case Opt_user_xattr:
449 set_opt(sbi, XATTR_USER);
450 break;
451 case Opt_nouser_xattr:
452 clear_opt(sbi, XATTR_USER);
453 break;
454 case Opt_inline_xattr:
455 set_opt(sbi, INLINE_XATTR);
456 break;
457 case Opt_noinline_xattr:
458 clear_opt(sbi, INLINE_XATTR);
459 break;
460 #else
461 case Opt_user_xattr:
462 f2fs_msg(sb, KERN_INFO,
463 "user_xattr options not supported");
464 break;
465 case Opt_nouser_xattr:
466 f2fs_msg(sb, KERN_INFO,
467 "nouser_xattr options not supported");
468 break;
469 case Opt_inline_xattr:
470 f2fs_msg(sb, KERN_INFO,
471 "inline_xattr options not supported");
472 break;
473 case Opt_noinline_xattr:
474 f2fs_msg(sb, KERN_INFO,
475 "noinline_xattr options not supported");
476 break;
477 #endif
478 #ifdef CONFIG_F2FS_FS_POSIX_ACL
479 case Opt_acl:
480 set_opt(sbi, POSIX_ACL);
481 break;
482 case Opt_noacl:
483 clear_opt(sbi, POSIX_ACL);
484 break;
485 #else
486 case Opt_acl:
487 f2fs_msg(sb, KERN_INFO, "acl options not supported");
488 break;
489 case Opt_noacl:
490 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
491 break;
492 #endif
493 case Opt_active_logs:
494 if (args->from && match_int(args, &arg))
495 return -EINVAL;
496 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
497 return -EINVAL;
498 sbi->active_logs = arg;
499 break;
500 case Opt_disable_ext_identify:
501 set_opt(sbi, DISABLE_EXT_IDENTIFY);
502 break;
503 case Opt_inline_data:
504 set_opt(sbi, INLINE_DATA);
505 break;
506 case Opt_inline_dentry:
507 set_opt(sbi, INLINE_DENTRY);
508 break;
509 case Opt_noinline_dentry:
510 clear_opt(sbi, INLINE_DENTRY);
511 break;
512 case Opt_flush_merge:
513 set_opt(sbi, FLUSH_MERGE);
514 break;
515 case Opt_noflush_merge:
516 clear_opt(sbi, FLUSH_MERGE);
517 break;
518 case Opt_nobarrier:
519 set_opt(sbi, NOBARRIER);
520 break;
521 case Opt_fastboot:
522 set_opt(sbi, FASTBOOT);
523 break;
524 case Opt_extent_cache:
525 set_opt(sbi, EXTENT_CACHE);
526 break;
527 case Opt_noextent_cache:
528 clear_opt(sbi, EXTENT_CACHE);
529 break;
530 case Opt_noinline_data:
531 clear_opt(sbi, INLINE_DATA);
532 break;
533 case Opt_data_flush:
534 set_opt(sbi, DATA_FLUSH);
535 break;
536 case Opt_mode:
537 name = match_strdup(&args[0]);
538
539 if (!name)
540 return -ENOMEM;
541 if (strlen(name) == 8 &&
542 !strncmp(name, "adaptive", 8)) {
543 if (f2fs_sb_mounted_blkzoned(sb)) {
544 f2fs_msg(sb, KERN_WARNING,
545 "adaptive mode is not allowed with "
546 "zoned block device feature");
547 kfree(name);
548 return -EINVAL;
549 }
550 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
551 } else if (strlen(name) == 3 &&
552 !strncmp(name, "lfs", 3)) {
553 set_opt_mode(sbi, F2FS_MOUNT_LFS);
554 } else {
555 kfree(name);
556 return -EINVAL;
557 }
558 kfree(name);
559 break;
560 case Opt_io_size_bits:
561 if (args->from && match_int(args, &arg))
562 return -EINVAL;
563 if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
564 f2fs_msg(sb, KERN_WARNING,
565 "Not support %d, larger than %d",
566 1 << arg, BIO_MAX_PAGES);
567 return -EINVAL;
568 }
569 sbi->write_io_size_bits = arg;
570 break;
571 case Opt_fault_injection:
572 if (args->from && match_int(args, &arg))
573 return -EINVAL;
574 #ifdef CONFIG_F2FS_FAULT_INJECTION
575 f2fs_build_fault_attr(sbi, arg);
576 set_opt(sbi, FAULT_INJECTION);
577 #else
578 f2fs_msg(sb, KERN_INFO,
579 "FAULT_INJECTION was not selected");
580 #endif
581 break;
582 case Opt_lazytime:
583 sb->s_flags |= MS_LAZYTIME;
584 break;
585 case Opt_nolazytime:
586 sb->s_flags &= ~MS_LAZYTIME;
587 break;
588 default:
589 f2fs_msg(sb, KERN_ERR,
590 "Unrecognized mount option \"%s\" or missing value",
591 p);
592 return -EINVAL;
593 }
594 }
595
596 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
597 f2fs_msg(sb, KERN_ERR,
598 "Should set mode=lfs with %uKB-sized IO",
599 F2FS_IO_SIZE_KB(sbi));
600 return -EINVAL;
601 }
602 return 0;
603 }
604
605 static struct inode *f2fs_alloc_inode(struct super_block *sb)
606 {
607 struct f2fs_inode_info *fi;
608
609 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
610 if (!fi)
611 return NULL;
612
613 init_once((void *) fi);
614
615 /* Initialize f2fs-specific inode info */
616 fi->vfs_inode.i_version = 1;
617 atomic_set(&fi->dirty_pages, 0);
618 fi->i_current_depth = 1;
619 fi->i_advise = 0;
620 init_rwsem(&fi->i_sem);
621 INIT_LIST_HEAD(&fi->dirty_list);
622 INIT_LIST_HEAD(&fi->gdirty_list);
623 INIT_LIST_HEAD(&fi->inmem_pages);
624 mutex_init(&fi->inmem_lock);
625 init_rwsem(&fi->dio_rwsem[READ]);
626 init_rwsem(&fi->dio_rwsem[WRITE]);
627 init_rwsem(&fi->i_mmap_sem);
628
629 /* Will be used by directory only */
630 fi->i_dir_level = F2FS_SB(sb)->dir_level;
631 return &fi->vfs_inode;
632 }
633
634 static int f2fs_drop_inode(struct inode *inode)
635 {
636 int ret;
637 /*
638 * This is to avoid a deadlock condition like below.
639 * writeback_single_inode(inode)
640 * - f2fs_write_data_page
641 * - f2fs_gc -> iput -> evict
642 * - inode_wait_for_writeback(inode)
643 */
644 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
645 if (!inode->i_nlink && !is_bad_inode(inode)) {
646 /* to avoid evict_inode call simultaneously */
647 atomic_inc(&inode->i_count);
648 spin_unlock(&inode->i_lock);
649
650 /* some remained atomic pages should discarded */
651 if (f2fs_is_atomic_file(inode))
652 drop_inmem_pages(inode);
653
654 /* should remain fi->extent_tree for writepage */
655 f2fs_destroy_extent_node(inode);
656
657 sb_start_intwrite(inode->i_sb);
658 f2fs_i_size_write(inode, 0);
659
660 if (F2FS_HAS_BLOCKS(inode))
661 f2fs_truncate(inode);
662
663 sb_end_intwrite(inode->i_sb);
664
665 fscrypt_put_encryption_info(inode, NULL);
666 spin_lock(&inode->i_lock);
667 atomic_dec(&inode->i_count);
668 }
669 trace_f2fs_drop_inode(inode, 0);
670 return 0;
671 }
672 ret = generic_drop_inode(inode);
673 trace_f2fs_drop_inode(inode, ret);
674 return ret;
675 }
676
677 int f2fs_inode_dirtied(struct inode *inode, bool sync)
678 {
679 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
680 int ret = 0;
681
682 spin_lock(&sbi->inode_lock[DIRTY_META]);
683 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
684 ret = 1;
685 } else {
686 set_inode_flag(inode, FI_DIRTY_INODE);
687 stat_inc_dirty_inode(sbi, DIRTY_META);
688 }
689 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
690 list_add_tail(&F2FS_I(inode)->gdirty_list,
691 &sbi->inode_list[DIRTY_META]);
692 inc_page_count(sbi, F2FS_DIRTY_IMETA);
693 }
694 spin_unlock(&sbi->inode_lock[DIRTY_META]);
695 return ret;
696 }
697
698 void f2fs_inode_synced(struct inode *inode)
699 {
700 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
701
702 spin_lock(&sbi->inode_lock[DIRTY_META]);
703 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
704 spin_unlock(&sbi->inode_lock[DIRTY_META]);
705 return;
706 }
707 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
708 list_del_init(&F2FS_I(inode)->gdirty_list);
709 dec_page_count(sbi, F2FS_DIRTY_IMETA);
710 }
711 clear_inode_flag(inode, FI_DIRTY_INODE);
712 clear_inode_flag(inode, FI_AUTO_RECOVER);
713 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
714 spin_unlock(&sbi->inode_lock[DIRTY_META]);
715 }
716
717 /*
718 * f2fs_dirty_inode() is called from __mark_inode_dirty()
719 *
720 * We should call set_dirty_inode to write the dirty inode through write_inode.
721 */
722 static void f2fs_dirty_inode(struct inode *inode, int flags)
723 {
724 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
725
726 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
727 inode->i_ino == F2FS_META_INO(sbi))
728 return;
729
730 if (flags == I_DIRTY_TIME)
731 return;
732
733 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
734 clear_inode_flag(inode, FI_AUTO_RECOVER);
735
736 f2fs_inode_dirtied(inode, false);
737 }
738
739 static void f2fs_i_callback(struct rcu_head *head)
740 {
741 struct inode *inode = container_of(head, struct inode, i_rcu);
742 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
743 }
744
745 static void f2fs_destroy_inode(struct inode *inode)
746 {
747 call_rcu(&inode->i_rcu, f2fs_i_callback);
748 }
749
750 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
751 {
752 percpu_counter_destroy(&sbi->alloc_valid_block_count);
753 percpu_counter_destroy(&sbi->total_valid_inode_count);
754 }
755
756 static void destroy_device_list(struct f2fs_sb_info *sbi)
757 {
758 int i;
759
760 for (i = 0; i < sbi->s_ndevs; i++) {
761 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
762 #ifdef CONFIG_BLK_DEV_ZONED
763 kfree(FDEV(i).blkz_type);
764 #endif
765 }
766 kfree(sbi->devs);
767 }
768
769 static void f2fs_put_super(struct super_block *sb)
770 {
771 struct f2fs_sb_info *sbi = F2FS_SB(sb);
772 int i;
773
774 if (sbi->s_proc) {
775 remove_proc_entry("segment_info", sbi->s_proc);
776 remove_proc_entry("segment_bits", sbi->s_proc);
777 remove_proc_entry(sb->s_id, f2fs_proc_root);
778 }
779 kobject_del(&sbi->s_kobj);
780
781 stop_gc_thread(sbi);
782
783 /* prevent remaining shrinker jobs */
784 mutex_lock(&sbi->umount_mutex);
785
786 /*
787 * We don't need to do checkpoint when superblock is clean.
788 * But, the previous checkpoint was not done by umount, it needs to do
789 * clean checkpoint again.
790 */
791 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
792 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
793 struct cp_control cpc = {
794 .reason = CP_UMOUNT,
795 };
796 write_checkpoint(sbi, &cpc);
797 }
798
799 /* be sure to wait for any on-going discard commands */
800 f2fs_wait_discard_bios(sbi);
801
802 if (!sbi->discard_blks) {
803 struct cp_control cpc = {
804 .reason = CP_UMOUNT | CP_TRIMMED,
805 };
806 write_checkpoint(sbi, &cpc);
807 }
808
809 /* write_checkpoint can update stat informaion */
810 f2fs_destroy_stats(sbi);
811
812 /*
813 * normally superblock is clean, so we need to release this.
814 * In addition, EIO will skip do checkpoint, we need this as well.
815 */
816 release_ino_entry(sbi, true);
817
818 f2fs_leave_shrinker(sbi);
819 mutex_unlock(&sbi->umount_mutex);
820
821 /* our cp_error case, we can wait for any writeback page */
822 f2fs_flush_merged_writes(sbi);
823
824 iput(sbi->node_inode);
825 iput(sbi->meta_inode);
826
827 /* destroy f2fs internal modules */
828 destroy_node_manager(sbi);
829 destroy_segment_manager(sbi);
830
831 kfree(sbi->ckpt);
832 kobject_put(&sbi->s_kobj);
833 wait_for_completion(&sbi->s_kobj_unregister);
834
835 sb->s_fs_info = NULL;
836 if (sbi->s_chksum_driver)
837 crypto_free_shash(sbi->s_chksum_driver);
838 kfree(sbi->raw_super);
839
840 destroy_device_list(sbi);
841 mempool_destroy(sbi->write_io_dummy);
842 destroy_percpu_info(sbi);
843 for (i = 0; i < NR_PAGE_TYPE; i++)
844 kfree(sbi->write_io[i]);
845 kfree(sbi);
846 }
847
848 int f2fs_sync_fs(struct super_block *sb, int sync)
849 {
850 struct f2fs_sb_info *sbi = F2FS_SB(sb);
851 int err = 0;
852
853 trace_f2fs_sync_fs(sb, sync);
854
855 if (sync) {
856 struct cp_control cpc;
857
858 cpc.reason = __get_cp_reason(sbi);
859
860 mutex_lock(&sbi->gc_mutex);
861 err = write_checkpoint(sbi, &cpc);
862 mutex_unlock(&sbi->gc_mutex);
863 }
864 f2fs_trace_ios(NULL, 1);
865
866 return err;
867 }
868
869 static int f2fs_freeze(struct super_block *sb)
870 {
871 if (f2fs_readonly(sb))
872 return 0;
873
874 /* IO error happened before */
875 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
876 return -EIO;
877
878 /* must be clean, since sync_filesystem() was already called */
879 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
880 return -EINVAL;
881 return 0;
882 }
883
884 static int f2fs_unfreeze(struct super_block *sb)
885 {
886 return 0;
887 }
888
889 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
890 {
891 struct super_block *sb = dentry->d_sb;
892 struct f2fs_sb_info *sbi = F2FS_SB(sb);
893 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
894 block_t total_count, user_block_count, start_count, ovp_count;
895
896 total_count = le64_to_cpu(sbi->raw_super->block_count);
897 user_block_count = sbi->user_block_count;
898 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
899 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
900 buf->f_type = F2FS_SUPER_MAGIC;
901 buf->f_bsize = sbi->blocksize;
902
903 buf->f_blocks = total_count - start_count;
904 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
905 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
906
907 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
908 buf->f_ffree = min(buf->f_files - valid_node_count(sbi),
909 buf->f_bavail);
910
911 buf->f_namelen = F2FS_NAME_LEN;
912 buf->f_fsid.val[0] = (u32)id;
913 buf->f_fsid.val[1] = (u32)(id >> 32);
914
915 return 0;
916 }
917
918 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
919 {
920 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
921
922 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
923 if (test_opt(sbi, FORCE_FG_GC))
924 seq_printf(seq, ",background_gc=%s", "sync");
925 else
926 seq_printf(seq, ",background_gc=%s", "on");
927 } else {
928 seq_printf(seq, ",background_gc=%s", "off");
929 }
930 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
931 seq_puts(seq, ",disable_roll_forward");
932 if (test_opt(sbi, DISCARD))
933 seq_puts(seq, ",discard");
934 if (test_opt(sbi, NOHEAP))
935 seq_puts(seq, ",no_heap");
936 else
937 seq_puts(seq, ",heap");
938 #ifdef CONFIG_F2FS_FS_XATTR
939 if (test_opt(sbi, XATTR_USER))
940 seq_puts(seq, ",user_xattr");
941 else
942 seq_puts(seq, ",nouser_xattr");
943 if (test_opt(sbi, INLINE_XATTR))
944 seq_puts(seq, ",inline_xattr");
945 else
946 seq_puts(seq, ",noinline_xattr");
947 #endif
948 #ifdef CONFIG_F2FS_FS_POSIX_ACL
949 if (test_opt(sbi, POSIX_ACL))
950 seq_puts(seq, ",acl");
951 else
952 seq_puts(seq, ",noacl");
953 #endif
954 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
955 seq_puts(seq, ",disable_ext_identify");
956 if (test_opt(sbi, INLINE_DATA))
957 seq_puts(seq, ",inline_data");
958 else
959 seq_puts(seq, ",noinline_data");
960 if (test_opt(sbi, INLINE_DENTRY))
961 seq_puts(seq, ",inline_dentry");
962 else
963 seq_puts(seq, ",noinline_dentry");
964 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
965 seq_puts(seq, ",flush_merge");
966 if (test_opt(sbi, NOBARRIER))
967 seq_puts(seq, ",nobarrier");
968 if (test_opt(sbi, FASTBOOT))
969 seq_puts(seq, ",fastboot");
970 if (test_opt(sbi, EXTENT_CACHE))
971 seq_puts(seq, ",extent_cache");
972 else
973 seq_puts(seq, ",noextent_cache");
974 if (test_opt(sbi, DATA_FLUSH))
975 seq_puts(seq, ",data_flush");
976
977 seq_puts(seq, ",mode=");
978 if (test_opt(sbi, ADAPTIVE))
979 seq_puts(seq, "adaptive");
980 else if (test_opt(sbi, LFS))
981 seq_puts(seq, "lfs");
982 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
983 if (F2FS_IO_SIZE_BITS(sbi))
984 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
985 #ifdef CONFIG_F2FS_FAULT_INJECTION
986 if (test_opt(sbi, FAULT_INJECTION))
987 seq_puts(seq, ",fault_injection");
988 #endif
989
990 return 0;
991 }
992
993 static int segment_info_seq_show(struct seq_file *seq, void *offset)
994 {
995 struct super_block *sb = seq->private;
996 struct f2fs_sb_info *sbi = F2FS_SB(sb);
997 unsigned int total_segs =
998 le32_to_cpu(sbi->raw_super->segment_count_main);
999 int i;
1000
1001 seq_puts(seq, "format: segment_type|valid_blocks\n"
1002 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
1003
1004 for (i = 0; i < total_segs; i++) {
1005 struct seg_entry *se = get_seg_entry(sbi, i);
1006
1007 if ((i % 10) == 0)
1008 seq_printf(seq, "%-10d", i);
1009 seq_printf(seq, "%d|%-3u", se->type,
1010 get_valid_blocks(sbi, i, false));
1011 if ((i % 10) == 9 || i == (total_segs - 1))
1012 seq_putc(seq, '\n');
1013 else
1014 seq_putc(seq, ' ');
1015 }
1016
1017 return 0;
1018 }
1019
1020 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
1021 {
1022 struct super_block *sb = seq->private;
1023 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1024 unsigned int total_segs =
1025 le32_to_cpu(sbi->raw_super->segment_count_main);
1026 int i, j;
1027
1028 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
1029 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
1030
1031 for (i = 0; i < total_segs; i++) {
1032 struct seg_entry *se = get_seg_entry(sbi, i);
1033
1034 seq_printf(seq, "%-10d", i);
1035 seq_printf(seq, "%d|%-3u|", se->type,
1036 get_valid_blocks(sbi, i, false));
1037 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
1038 seq_printf(seq, " %.2x", se->cur_valid_map[j]);
1039 seq_putc(seq, '\n');
1040 }
1041 return 0;
1042 }
1043
1044 #define F2FS_PROC_FILE_DEF(_name) \
1045 static int _name##_open_fs(struct inode *inode, struct file *file) \
1046 { \
1047 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
1048 } \
1049 \
1050 static const struct file_operations f2fs_seq_##_name##_fops = { \
1051 .open = _name##_open_fs, \
1052 .read = seq_read, \
1053 .llseek = seq_lseek, \
1054 .release = single_release, \
1055 };
1056
1057 F2FS_PROC_FILE_DEF(segment_info);
1058 F2FS_PROC_FILE_DEF(segment_bits);
1059
1060 static void default_options(struct f2fs_sb_info *sbi)
1061 {
1062 /* init some FS parameters */
1063 sbi->active_logs = NR_CURSEG_TYPE;
1064
1065 set_opt(sbi, BG_GC);
1066 set_opt(sbi, INLINE_XATTR);
1067 set_opt(sbi, INLINE_DATA);
1068 set_opt(sbi, INLINE_DENTRY);
1069 set_opt(sbi, EXTENT_CACHE);
1070 set_opt(sbi, NOHEAP);
1071 sbi->sb->s_flags |= MS_LAZYTIME;
1072 set_opt(sbi, FLUSH_MERGE);
1073 if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1074 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1075 set_opt(sbi, DISCARD);
1076 } else {
1077 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1078 }
1079
1080 #ifdef CONFIG_F2FS_FS_XATTR
1081 set_opt(sbi, XATTR_USER);
1082 #endif
1083 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1084 set_opt(sbi, POSIX_ACL);
1085 #endif
1086
1087 #ifdef CONFIG_F2FS_FAULT_INJECTION
1088 f2fs_build_fault_attr(sbi, 0);
1089 #endif
1090 }
1091
1092 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1093 {
1094 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1095 struct f2fs_mount_info org_mount_opt;
1096 int err, active_logs;
1097 bool need_restart_gc = false;
1098 bool need_stop_gc = false;
1099 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1100 #ifdef CONFIG_F2FS_FAULT_INJECTION
1101 struct f2fs_fault_info ffi = sbi->fault_info;
1102 #endif
1103
1104 /*
1105 * Save the old mount options in case we
1106 * need to restore them.
1107 */
1108 org_mount_opt = sbi->mount_opt;
1109 active_logs = sbi->active_logs;
1110
1111 /* recover superblocks we couldn't write due to previous RO mount */
1112 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1113 err = f2fs_commit_super(sbi, false);
1114 f2fs_msg(sb, KERN_INFO,
1115 "Try to recover all the superblocks, ret: %d", err);
1116 if (!err)
1117 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1118 }
1119
1120 sbi->mount_opt.opt = 0;
1121 default_options(sbi);
1122
1123 /* parse mount options */
1124 err = parse_options(sb, data);
1125 if (err)
1126 goto restore_opts;
1127
1128 /*
1129 * Previous and new state of filesystem is RO,
1130 * so skip checking GC and FLUSH_MERGE conditions.
1131 */
1132 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1133 goto skip;
1134
1135 /* disallow enable/disable extent_cache dynamically */
1136 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1137 err = -EINVAL;
1138 f2fs_msg(sbi->sb, KERN_WARNING,
1139 "switch extent_cache option is not allowed");
1140 goto restore_opts;
1141 }
1142
1143 /*
1144 * We stop the GC thread if FS is mounted as RO
1145 * or if background_gc = off is passed in mount
1146 * option. Also sync the filesystem.
1147 */
1148 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1149 if (sbi->gc_thread) {
1150 stop_gc_thread(sbi);
1151 need_restart_gc = true;
1152 }
1153 } else if (!sbi->gc_thread) {
1154 err = start_gc_thread(sbi);
1155 if (err)
1156 goto restore_opts;
1157 need_stop_gc = true;
1158 }
1159
1160 if (*flags & MS_RDONLY) {
1161 writeback_inodes_sb(sb, WB_REASON_SYNC);
1162 sync_inodes_sb(sb);
1163
1164 set_sbi_flag(sbi, SBI_IS_DIRTY);
1165 set_sbi_flag(sbi, SBI_IS_CLOSE);
1166 f2fs_sync_fs(sb, 1);
1167 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1168 }
1169
1170 /*
1171 * We stop issue flush thread if FS is mounted as RO
1172 * or if flush_merge is not passed in mount option.
1173 */
1174 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1175 clear_opt(sbi, FLUSH_MERGE);
1176 destroy_flush_cmd_control(sbi, false);
1177 } else {
1178 err = create_flush_cmd_control(sbi);
1179 if (err)
1180 goto restore_gc;
1181 }
1182 skip:
1183 /* Update the POSIXACL Flag */
1184 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1185 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1186
1187 return 0;
1188 restore_gc:
1189 if (need_restart_gc) {
1190 if (start_gc_thread(sbi))
1191 f2fs_msg(sbi->sb, KERN_WARNING,
1192 "background gc thread has stopped");
1193 } else if (need_stop_gc) {
1194 stop_gc_thread(sbi);
1195 }
1196 restore_opts:
1197 sbi->mount_opt = org_mount_opt;
1198 sbi->active_logs = active_logs;
1199 #ifdef CONFIG_F2FS_FAULT_INJECTION
1200 sbi->fault_info = ffi;
1201 #endif
1202 return err;
1203 }
1204
1205 static struct super_operations f2fs_sops = {
1206 .alloc_inode = f2fs_alloc_inode,
1207 .drop_inode = f2fs_drop_inode,
1208 .destroy_inode = f2fs_destroy_inode,
1209 .write_inode = f2fs_write_inode,
1210 .dirty_inode = f2fs_dirty_inode,
1211 .show_options = f2fs_show_options,
1212 .evict_inode = f2fs_evict_inode,
1213 .put_super = f2fs_put_super,
1214 .sync_fs = f2fs_sync_fs,
1215 .freeze_fs = f2fs_freeze,
1216 .unfreeze_fs = f2fs_unfreeze,
1217 .statfs = f2fs_statfs,
1218 .remount_fs = f2fs_remount,
1219 };
1220
1221 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1222 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1223 {
1224 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1225 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1226 ctx, len, NULL);
1227 }
1228
1229 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1230 void *fs_data)
1231 {
1232 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1233 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1234 ctx, len, fs_data, XATTR_CREATE);
1235 }
1236
1237 static unsigned f2fs_max_namelen(struct inode *inode)
1238 {
1239 return S_ISLNK(inode->i_mode) ?
1240 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1241 }
1242
1243 static const struct fscrypt_operations f2fs_cryptops = {
1244 .key_prefix = "f2fs:",
1245 .get_context = f2fs_get_context,
1246 .set_context = f2fs_set_context,
1247 .is_encrypted = f2fs_encrypted_inode,
1248 .empty_dir = f2fs_empty_dir,
1249 .max_namelen = f2fs_max_namelen,
1250 };
1251 #else
1252 static const struct fscrypt_operations f2fs_cryptops = {
1253 .is_encrypted = f2fs_encrypted_inode,
1254 };
1255 #endif
1256
1257 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1258 u64 ino, u32 generation)
1259 {
1260 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1261 struct inode *inode;
1262
1263 if (check_nid_range(sbi, ino))
1264 return ERR_PTR(-ESTALE);
1265
1266 /*
1267 * f2fs_iget isn't quite right if the inode is currently unallocated!
1268 * However f2fs_iget currently does appropriate checks to handle stale
1269 * inodes so everything is OK.
1270 */
1271 inode = f2fs_iget(sb, ino);
1272 if (IS_ERR(inode))
1273 return ERR_CAST(inode);
1274 if (unlikely(generation && inode->i_generation != generation)) {
1275 /* we didn't find the right inode.. */
1276 iput(inode);
1277 return ERR_PTR(-ESTALE);
1278 }
1279 return inode;
1280 }
1281
1282 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1283 int fh_len, int fh_type)
1284 {
1285 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1286 f2fs_nfs_get_inode);
1287 }
1288
1289 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1290 int fh_len, int fh_type)
1291 {
1292 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1293 f2fs_nfs_get_inode);
1294 }
1295
1296 static const struct export_operations f2fs_export_ops = {
1297 .fh_to_dentry = f2fs_fh_to_dentry,
1298 .fh_to_parent = f2fs_fh_to_parent,
1299 .get_parent = f2fs_get_parent,
1300 };
1301
1302 static loff_t max_file_blocks(void)
1303 {
1304 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1305 loff_t leaf_count = ADDRS_PER_BLOCK;
1306
1307 /* two direct node blocks */
1308 result += (leaf_count * 2);
1309
1310 /* two indirect node blocks */
1311 leaf_count *= NIDS_PER_BLOCK;
1312 result += (leaf_count * 2);
1313
1314 /* one double indirect node block */
1315 leaf_count *= NIDS_PER_BLOCK;
1316 result += leaf_count;
1317
1318 return result;
1319 }
1320
1321 static int __f2fs_commit_super(struct buffer_head *bh,
1322 struct f2fs_super_block *super)
1323 {
1324 lock_buffer(bh);
1325 if (super)
1326 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1327 set_buffer_uptodate(bh);
1328 set_buffer_dirty(bh);
1329 unlock_buffer(bh);
1330
1331 /* it's rare case, we can do fua all the time */
1332 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
1333 }
1334
1335 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1336 struct buffer_head *bh)
1337 {
1338 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1339 (bh->b_data + F2FS_SUPER_OFFSET);
1340 struct super_block *sb = sbi->sb;
1341 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1342 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1343 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1344 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1345 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1346 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1347 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1348 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1349 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1350 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1351 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1352 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1353 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1354 u64 main_end_blkaddr = main_blkaddr +
1355 (segment_count_main << log_blocks_per_seg);
1356 u64 seg_end_blkaddr = segment0_blkaddr +
1357 (segment_count << log_blocks_per_seg);
1358
1359 if (segment0_blkaddr != cp_blkaddr) {
1360 f2fs_msg(sb, KERN_INFO,
1361 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1362 segment0_blkaddr, cp_blkaddr);
1363 return true;
1364 }
1365
1366 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1367 sit_blkaddr) {
1368 f2fs_msg(sb, KERN_INFO,
1369 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1370 cp_blkaddr, sit_blkaddr,
1371 segment_count_ckpt << log_blocks_per_seg);
1372 return true;
1373 }
1374
1375 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1376 nat_blkaddr) {
1377 f2fs_msg(sb, KERN_INFO,
1378 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1379 sit_blkaddr, nat_blkaddr,
1380 segment_count_sit << log_blocks_per_seg);
1381 return true;
1382 }
1383
1384 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1385 ssa_blkaddr) {
1386 f2fs_msg(sb, KERN_INFO,
1387 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1388 nat_blkaddr, ssa_blkaddr,
1389 segment_count_nat << log_blocks_per_seg);
1390 return true;
1391 }
1392
1393 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1394 main_blkaddr) {
1395 f2fs_msg(sb, KERN_INFO,
1396 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1397 ssa_blkaddr, main_blkaddr,
1398 segment_count_ssa << log_blocks_per_seg);
1399 return true;
1400 }
1401
1402 if (main_end_blkaddr > seg_end_blkaddr) {
1403 f2fs_msg(sb, KERN_INFO,
1404 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1405 main_blkaddr,
1406 segment0_blkaddr +
1407 (segment_count << log_blocks_per_seg),
1408 segment_count_main << log_blocks_per_seg);
1409 return true;
1410 } else if (main_end_blkaddr < seg_end_blkaddr) {
1411 int err = 0;
1412 char *res;
1413
1414 /* fix in-memory information all the time */
1415 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1416 segment0_blkaddr) >> log_blocks_per_seg);
1417
1418 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1419 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1420 res = "internally";
1421 } else {
1422 err = __f2fs_commit_super(bh, NULL);
1423 res = err ? "failed" : "done";
1424 }
1425 f2fs_msg(sb, KERN_INFO,
1426 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1427 res, main_blkaddr,
1428 segment0_blkaddr +
1429 (segment_count << log_blocks_per_seg),
1430 segment_count_main << log_blocks_per_seg);
1431 if (err)
1432 return true;
1433 }
1434 return false;
1435 }
1436
1437 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1438 struct buffer_head *bh)
1439 {
1440 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1441 (bh->b_data + F2FS_SUPER_OFFSET);
1442 struct super_block *sb = sbi->sb;
1443 unsigned int blocksize;
1444
1445 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1446 f2fs_msg(sb, KERN_INFO,
1447 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1448 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1449 return 1;
1450 }
1451
1452 /* Currently, support only 4KB page cache size */
1453 if (F2FS_BLKSIZE != PAGE_SIZE) {
1454 f2fs_msg(sb, KERN_INFO,
1455 "Invalid page_cache_size (%lu), supports only 4KB\n",
1456 PAGE_SIZE);
1457 return 1;
1458 }
1459
1460 /* Currently, support only 4KB block size */
1461 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1462 if (blocksize != F2FS_BLKSIZE) {
1463 f2fs_msg(sb, KERN_INFO,
1464 "Invalid blocksize (%u), supports only 4KB\n",
1465 blocksize);
1466 return 1;
1467 }
1468
1469 /* check log blocks per segment */
1470 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1471 f2fs_msg(sb, KERN_INFO,
1472 "Invalid log blocks per segment (%u)\n",
1473 le32_to_cpu(raw_super->log_blocks_per_seg));
1474 return 1;
1475 }
1476
1477 /* Currently, support 512/1024/2048/4096 bytes sector size */
1478 if (le32_to_cpu(raw_super->log_sectorsize) >
1479 F2FS_MAX_LOG_SECTOR_SIZE ||
1480 le32_to_cpu(raw_super->log_sectorsize) <
1481 F2FS_MIN_LOG_SECTOR_SIZE) {
1482 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1483 le32_to_cpu(raw_super->log_sectorsize));
1484 return 1;
1485 }
1486 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1487 le32_to_cpu(raw_super->log_sectorsize) !=
1488 F2FS_MAX_LOG_SECTOR_SIZE) {
1489 f2fs_msg(sb, KERN_INFO,
1490 "Invalid log sectors per block(%u) log sectorsize(%u)",
1491 le32_to_cpu(raw_super->log_sectors_per_block),
1492 le32_to_cpu(raw_super->log_sectorsize));
1493 return 1;
1494 }
1495
1496 /* check reserved ino info */
1497 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1498 le32_to_cpu(raw_super->meta_ino) != 2 ||
1499 le32_to_cpu(raw_super->root_ino) != 3) {
1500 f2fs_msg(sb, KERN_INFO,
1501 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1502 le32_to_cpu(raw_super->node_ino),
1503 le32_to_cpu(raw_super->meta_ino),
1504 le32_to_cpu(raw_super->root_ino));
1505 return 1;
1506 }
1507
1508 if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) {
1509 f2fs_msg(sb, KERN_INFO,
1510 "Invalid segment count (%u)",
1511 le32_to_cpu(raw_super->segment_count));
1512 return 1;
1513 }
1514
1515 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1516 if (sanity_check_area_boundary(sbi, bh))
1517 return 1;
1518
1519 return 0;
1520 }
1521
1522 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1523 {
1524 unsigned int total, fsmeta;
1525 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1526 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1527 unsigned int ovp_segments, reserved_segments;
1528 unsigned int main_segs, blocks_per_seg;
1529 int i;
1530
1531 total = le32_to_cpu(raw_super->segment_count);
1532 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1533 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1534 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1535 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1536 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1537
1538 if (unlikely(fsmeta >= total))
1539 return 1;
1540
1541 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1542 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1543
1544 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1545 ovp_segments == 0 || reserved_segments == 0)) {
1546 f2fs_msg(sbi->sb, KERN_ERR,
1547 "Wrong layout: check mkfs.f2fs version");
1548 return 1;
1549 }
1550
1551 main_segs = le32_to_cpu(raw_super->segment_count_main);
1552 blocks_per_seg = sbi->blocks_per_seg;
1553
1554 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
1555 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
1556 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
1557 return 1;
1558 }
1559 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
1560 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
1561 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
1562 return 1;
1563 }
1564
1565 if (unlikely(f2fs_cp_error(sbi))) {
1566 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1567 return 1;
1568 }
1569 return 0;
1570 }
1571
1572 static void init_sb_info(struct f2fs_sb_info *sbi)
1573 {
1574 struct f2fs_super_block *raw_super = sbi->raw_super;
1575 int i, j;
1576
1577 sbi->log_sectors_per_block =
1578 le32_to_cpu(raw_super->log_sectors_per_block);
1579 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1580 sbi->blocksize = 1 << sbi->log_blocksize;
1581 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1582 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1583 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1584 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1585 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1586 sbi->total_node_count =
1587 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1588 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1589 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1590 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1591 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1592 sbi->cur_victim_sec = NULL_SECNO;
1593 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1594
1595 sbi->dir_level = DEF_DIR_LEVEL;
1596 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1597 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1598 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1599
1600 for (i = 0; i < NR_COUNT_TYPE; i++)
1601 atomic_set(&sbi->nr_pages[i], 0);
1602
1603 atomic_set(&sbi->wb_sync_req, 0);
1604
1605 INIT_LIST_HEAD(&sbi->s_list);
1606 mutex_init(&sbi->umount_mutex);
1607 for (i = 0; i < NR_PAGE_TYPE - 1; i++)
1608 for (j = HOT; j < NR_TEMP_TYPE; j++)
1609 mutex_init(&sbi->wio_mutex[i][j]);
1610 spin_lock_init(&sbi->cp_lock);
1611 }
1612
1613 static int init_percpu_info(struct f2fs_sb_info *sbi)
1614 {
1615 int err;
1616
1617 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1618 if (err)
1619 return err;
1620
1621 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1622 GFP_KERNEL);
1623 }
1624
1625 #ifdef CONFIG_BLK_DEV_ZONED
1626 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
1627 {
1628 struct block_device *bdev = FDEV(devi).bdev;
1629 sector_t nr_sectors = bdev->bd_part->nr_sects;
1630 sector_t sector = 0;
1631 struct blk_zone *zones;
1632 unsigned int i, nr_zones;
1633 unsigned int n = 0;
1634 int err = -EIO;
1635
1636 if (!f2fs_sb_mounted_blkzoned(sbi->sb))
1637 return 0;
1638
1639 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
1640 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
1641 return -EINVAL;
1642 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
1643 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
1644 __ilog2_u32(sbi->blocks_per_blkz))
1645 return -EINVAL;
1646 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
1647 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
1648 sbi->log_blocks_per_blkz;
1649 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
1650 FDEV(devi).nr_blkz++;
1651
1652 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
1653 if (!FDEV(devi).blkz_type)
1654 return -ENOMEM;
1655
1656 #define F2FS_REPORT_NR_ZONES 4096
1657
1658 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
1659 GFP_KERNEL);
1660 if (!zones)
1661 return -ENOMEM;
1662
1663 /* Get block zones type */
1664 while (zones && sector < nr_sectors) {
1665
1666 nr_zones = F2FS_REPORT_NR_ZONES;
1667 err = blkdev_report_zones(bdev, sector,
1668 zones, &nr_zones,
1669 GFP_KERNEL);
1670 if (err)
1671 break;
1672 if (!nr_zones) {
1673 err = -EIO;
1674 break;
1675 }
1676
1677 for (i = 0; i < nr_zones; i++) {
1678 FDEV(devi).blkz_type[n] = zones[i].type;
1679 sector += zones[i].len;
1680 n++;
1681 }
1682 }
1683
1684 kfree(zones);
1685
1686 return err;
1687 }
1688 #endif
1689
1690 /*
1691 * Read f2fs raw super block.
1692 * Because we have two copies of super block, so read both of them
1693 * to get the first valid one. If any one of them is broken, we pass
1694 * them recovery flag back to the caller.
1695 */
1696 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1697 struct f2fs_super_block **raw_super,
1698 int *valid_super_block, int *recovery)
1699 {
1700 struct super_block *sb = sbi->sb;
1701 int block;
1702 struct buffer_head *bh;
1703 struct f2fs_super_block *super;
1704 int err = 0;
1705
1706 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1707 if (!super)
1708 return -ENOMEM;
1709
1710 for (block = 0; block < 2; block++) {
1711 bh = sb_bread(sb, block);
1712 if (!bh) {
1713 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1714 block + 1);
1715 err = -EIO;
1716 continue;
1717 }
1718
1719 /* sanity checking of raw super */
1720 if (sanity_check_raw_super(sbi, bh)) {
1721 f2fs_msg(sb, KERN_ERR,
1722 "Can't find valid F2FS filesystem in %dth superblock",
1723 block + 1);
1724 err = -EINVAL;
1725 brelse(bh);
1726 continue;
1727 }
1728
1729 if (!*raw_super) {
1730 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1731 sizeof(*super));
1732 *valid_super_block = block;
1733 *raw_super = super;
1734 }
1735 brelse(bh);
1736 }
1737
1738 /* Fail to read any one of the superblocks*/
1739 if (err < 0)
1740 *recovery = 1;
1741
1742 /* No valid superblock */
1743 if (!*raw_super)
1744 kfree(super);
1745 else
1746 err = 0;
1747
1748 return err;
1749 }
1750
1751 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1752 {
1753 struct buffer_head *bh;
1754 int err;
1755
1756 if ((recover && f2fs_readonly(sbi->sb)) ||
1757 bdev_read_only(sbi->sb->s_bdev)) {
1758 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1759 return -EROFS;
1760 }
1761
1762 /* write back-up superblock first */
1763 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1764 if (!bh)
1765 return -EIO;
1766 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1767 brelse(bh);
1768
1769 /* if we are in recovery path, skip writing valid superblock */
1770 if (recover || err)
1771 return err;
1772
1773 /* write current valid superblock */
1774 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1775 if (!bh)
1776 return -EIO;
1777 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1778 brelse(bh);
1779 return err;
1780 }
1781
1782 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
1783 {
1784 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1785 unsigned int max_devices = MAX_DEVICES;
1786 int i;
1787
1788 /* Initialize single device information */
1789 if (!RDEV(0).path[0]) {
1790 if (!bdev_is_zoned(sbi->sb->s_bdev))
1791 return 0;
1792 max_devices = 1;
1793 }
1794
1795 /*
1796 * Initialize multiple devices information, or single
1797 * zoned block device information.
1798 */
1799 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
1800 GFP_KERNEL);
1801 if (!sbi->devs)
1802 return -ENOMEM;
1803
1804 for (i = 0; i < max_devices; i++) {
1805
1806 if (i > 0 && !RDEV(i).path[0])
1807 break;
1808
1809 if (max_devices == 1) {
1810 /* Single zoned block device mount */
1811 FDEV(0).bdev =
1812 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
1813 sbi->sb->s_mode, sbi->sb->s_type);
1814 } else {
1815 /* Multi-device mount */
1816 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
1817 FDEV(i).total_segments =
1818 le32_to_cpu(RDEV(i).total_segments);
1819 if (i == 0) {
1820 FDEV(i).start_blk = 0;
1821 FDEV(i).end_blk = FDEV(i).start_blk +
1822 (FDEV(i).total_segments <<
1823 sbi->log_blocks_per_seg) - 1 +
1824 le32_to_cpu(raw_super->segment0_blkaddr);
1825 } else {
1826 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
1827 FDEV(i).end_blk = FDEV(i).start_blk +
1828 (FDEV(i).total_segments <<
1829 sbi->log_blocks_per_seg) - 1;
1830 }
1831 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
1832 sbi->sb->s_mode, sbi->sb->s_type);
1833 }
1834 if (IS_ERR(FDEV(i).bdev))
1835 return PTR_ERR(FDEV(i).bdev);
1836
1837 /* to release errored devices */
1838 sbi->s_ndevs = i + 1;
1839
1840 #ifdef CONFIG_BLK_DEV_ZONED
1841 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
1842 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
1843 f2fs_msg(sbi->sb, KERN_ERR,
1844 "Zoned block device feature not enabled\n");
1845 return -EINVAL;
1846 }
1847 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
1848 if (init_blkz_info(sbi, i)) {
1849 f2fs_msg(sbi->sb, KERN_ERR,
1850 "Failed to initialize F2FS blkzone information");
1851 return -EINVAL;
1852 }
1853 if (max_devices == 1)
1854 break;
1855 f2fs_msg(sbi->sb, KERN_INFO,
1856 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
1857 i, FDEV(i).path,
1858 FDEV(i).total_segments,
1859 FDEV(i).start_blk, FDEV(i).end_blk,
1860 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
1861 "Host-aware" : "Host-managed");
1862 continue;
1863 }
1864 #endif
1865 f2fs_msg(sbi->sb, KERN_INFO,
1866 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
1867 i, FDEV(i).path,
1868 FDEV(i).total_segments,
1869 FDEV(i).start_blk, FDEV(i).end_blk);
1870 }
1871 f2fs_msg(sbi->sb, KERN_INFO,
1872 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
1873 return 0;
1874 }
1875
1876 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1877 {
1878 struct f2fs_sb_info *sbi;
1879 struct f2fs_super_block *raw_super;
1880 struct inode *root;
1881 int err;
1882 bool retry = true, need_fsck = false;
1883 char *options = NULL;
1884 int recovery, i, valid_super_block;
1885 struct curseg_info *seg_i;
1886
1887 try_onemore:
1888 err = -EINVAL;
1889 raw_super = NULL;
1890 valid_super_block = -1;
1891 recovery = 0;
1892
1893 /* allocate memory for f2fs-specific super block info */
1894 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1895 if (!sbi)
1896 return -ENOMEM;
1897
1898 sbi->sb = sb;
1899
1900 /* Load the checksum driver */
1901 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1902 if (IS_ERR(sbi->s_chksum_driver)) {
1903 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1904 err = PTR_ERR(sbi->s_chksum_driver);
1905 sbi->s_chksum_driver = NULL;
1906 goto free_sbi;
1907 }
1908
1909 /* set a block size */
1910 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1911 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1912 goto free_sbi;
1913 }
1914
1915 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1916 &recovery);
1917 if (err)
1918 goto free_sbi;
1919
1920 sb->s_fs_info = sbi;
1921 sbi->raw_super = raw_super;
1922
1923 /*
1924 * The BLKZONED feature indicates that the drive was formatted with
1925 * zone alignment optimization. This is optional for host-aware
1926 * devices, but mandatory for host-managed zoned block devices.
1927 */
1928 #ifndef CONFIG_BLK_DEV_ZONED
1929 if (f2fs_sb_mounted_blkzoned(sb)) {
1930 f2fs_msg(sb, KERN_ERR,
1931 "Zoned block device support is not enabled\n");
1932 goto free_sb_buf;
1933 }
1934 #endif
1935 default_options(sbi);
1936 /* parse mount options */
1937 options = kstrdup((const char *)data, GFP_KERNEL);
1938 if (data && !options) {
1939 err = -ENOMEM;
1940 goto free_sb_buf;
1941 }
1942
1943 err = parse_options(sb, options);
1944 if (err)
1945 goto free_options;
1946
1947 sbi->max_file_blocks = max_file_blocks();
1948 sb->s_maxbytes = sbi->max_file_blocks <<
1949 le32_to_cpu(raw_super->log_blocksize);
1950 sb->s_max_links = F2FS_LINK_MAX;
1951 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1952
1953 sb->s_op = &f2fs_sops;
1954 sb->s_cop = &f2fs_cryptops;
1955 sb->s_xattr = f2fs_xattr_handlers;
1956 sb->s_export_op = &f2fs_export_ops;
1957 sb->s_magic = F2FS_SUPER_MAGIC;
1958 sb->s_time_gran = 1;
1959 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1960 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1961 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1962
1963 /* init f2fs-specific super block info */
1964 sbi->valid_super_block = valid_super_block;
1965 mutex_init(&sbi->gc_mutex);
1966 mutex_init(&sbi->cp_mutex);
1967 init_rwsem(&sbi->node_write);
1968 init_rwsem(&sbi->node_change);
1969
1970 /* disallow all the data/node/meta page writes */
1971 set_sbi_flag(sbi, SBI_POR_DOING);
1972 spin_lock_init(&sbi->stat_lock);
1973
1974 for (i = 0; i < NR_PAGE_TYPE; i++) {
1975 int n = (i == META) ? 1: NR_TEMP_TYPE;
1976 int j;
1977
1978 sbi->write_io[i] = kmalloc(n * sizeof(struct f2fs_bio_info),
1979 GFP_KERNEL);
1980 if (!sbi->write_io[i])
1981 goto free_options;
1982
1983 for (j = HOT; j < n; j++) {
1984 init_rwsem(&sbi->write_io[i][j].io_rwsem);
1985 sbi->write_io[i][j].sbi = sbi;
1986 sbi->write_io[i][j].bio = NULL;
1987 spin_lock_init(&sbi->write_io[i][j].io_lock);
1988 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
1989 }
1990 }
1991
1992 init_rwsem(&sbi->cp_rwsem);
1993 init_waitqueue_head(&sbi->cp_wait);
1994 init_sb_info(sbi);
1995
1996 err = init_percpu_info(sbi);
1997 if (err)
1998 goto free_options;
1999
2000 if (F2FS_IO_SIZE(sbi) > 1) {
2001 sbi->write_io_dummy =
2002 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2003 if (!sbi->write_io_dummy)
2004 goto free_options;
2005 }
2006
2007 /* get an inode for meta space */
2008 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2009 if (IS_ERR(sbi->meta_inode)) {
2010 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2011 err = PTR_ERR(sbi->meta_inode);
2012 goto free_io_dummy;
2013 }
2014
2015 err = get_valid_checkpoint(sbi);
2016 if (err) {
2017 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2018 goto free_meta_inode;
2019 }
2020
2021 /* Initialize device list */
2022 err = f2fs_scan_devices(sbi);
2023 if (err) {
2024 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2025 goto free_devices;
2026 }
2027
2028 sbi->total_valid_node_count =
2029 le32_to_cpu(sbi->ckpt->valid_node_count);
2030 percpu_counter_set(&sbi->total_valid_inode_count,
2031 le32_to_cpu(sbi->ckpt->valid_inode_count));
2032 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2033 sbi->total_valid_block_count =
2034 le64_to_cpu(sbi->ckpt->valid_block_count);
2035 sbi->last_valid_block_count = sbi->total_valid_block_count;
2036
2037 for (i = 0; i < NR_INODE_TYPE; i++) {
2038 INIT_LIST_HEAD(&sbi->inode_list[i]);
2039 spin_lock_init(&sbi->inode_lock[i]);
2040 }
2041
2042 init_extent_cache_info(sbi);
2043
2044 init_ino_entry_info(sbi);
2045
2046 /* setup f2fs internal modules */
2047 err = build_segment_manager(sbi);
2048 if (err) {
2049 f2fs_msg(sb, KERN_ERR,
2050 "Failed to initialize F2FS segment manager");
2051 goto free_sm;
2052 }
2053 err = build_node_manager(sbi);
2054 if (err) {
2055 f2fs_msg(sb, KERN_ERR,
2056 "Failed to initialize F2FS node manager");
2057 goto free_nm;
2058 }
2059
2060 /* For write statistics */
2061 if (sb->s_bdev->bd_part)
2062 sbi->sectors_written_start =
2063 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2064
2065 /* Read accumulated write IO statistics if exists */
2066 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2067 if (__exist_node_summaries(sbi))
2068 sbi->kbytes_written =
2069 le64_to_cpu(seg_i->journal->info.kbytes_written);
2070
2071 build_gc_manager(sbi);
2072
2073 /* get an inode for node space */
2074 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2075 if (IS_ERR(sbi->node_inode)) {
2076 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2077 err = PTR_ERR(sbi->node_inode);
2078 goto free_nm;
2079 }
2080
2081 f2fs_join_shrinker(sbi);
2082
2083 err = f2fs_build_stats(sbi);
2084 if (err)
2085 goto free_nm;
2086
2087 /* if there are nt orphan nodes free them */
2088 err = recover_orphan_inodes(sbi);
2089 if (err)
2090 goto free_node_inode;
2091
2092 /* read root inode and dentry */
2093 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2094 if (IS_ERR(root)) {
2095 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2096 err = PTR_ERR(root);
2097 goto free_node_inode;
2098 }
2099 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2100 iput(root);
2101 err = -EINVAL;
2102 goto free_node_inode;
2103 }
2104
2105 sb->s_root = d_make_root(root); /* allocate root dentry */
2106 if (!sb->s_root) {
2107 err = -ENOMEM;
2108 goto free_root_inode;
2109 }
2110
2111 if (f2fs_proc_root)
2112 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
2113
2114 if (sbi->s_proc) {
2115 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
2116 &f2fs_seq_segment_info_fops, sb);
2117 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
2118 &f2fs_seq_segment_bits_fops, sb);
2119 }
2120
2121 sbi->s_kobj.kset = f2fs_kset;
2122 init_completion(&sbi->s_kobj_unregister);
2123 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
2124 "%s", sb->s_id);
2125 if (err)
2126 goto free_proc;
2127
2128 /* recover fsynced data */
2129 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2130 /*
2131 * mount should be failed, when device has readonly mode, and
2132 * previous checkpoint was not done by clean system shutdown.
2133 */
2134 if (bdev_read_only(sb->s_bdev) &&
2135 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2136 err = -EROFS;
2137 goto free_kobj;
2138 }
2139
2140 if (need_fsck)
2141 set_sbi_flag(sbi, SBI_NEED_FSCK);
2142
2143 if (!retry)
2144 goto skip_recovery;
2145
2146 err = recover_fsync_data(sbi, false);
2147 if (err < 0) {
2148 need_fsck = true;
2149 f2fs_msg(sb, KERN_ERR,
2150 "Cannot recover all fsync data errno=%d", err);
2151 goto free_kobj;
2152 }
2153 } else {
2154 err = recover_fsync_data(sbi, true);
2155
2156 if (!f2fs_readonly(sb) && err > 0) {
2157 err = -EINVAL;
2158 f2fs_msg(sb, KERN_ERR,
2159 "Need to recover fsync data");
2160 goto free_kobj;
2161 }
2162 }
2163 skip_recovery:
2164 /* recover_fsync_data() cleared this already */
2165 clear_sbi_flag(sbi, SBI_POR_DOING);
2166
2167 /*
2168 * If filesystem is not mounted as read-only then
2169 * do start the gc_thread.
2170 */
2171 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2172 /* After POR, we can run background GC thread.*/
2173 err = start_gc_thread(sbi);
2174 if (err)
2175 goto free_kobj;
2176 }
2177 kfree(options);
2178
2179 /* recover broken superblock */
2180 if (recovery) {
2181 err = f2fs_commit_super(sbi, true);
2182 f2fs_msg(sb, KERN_INFO,
2183 "Try to recover %dth superblock, ret: %d",
2184 sbi->valid_super_block ? 1 : 2, err);
2185 }
2186
2187 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2188 cur_cp_version(F2FS_CKPT(sbi)));
2189 f2fs_update_time(sbi, CP_TIME);
2190 f2fs_update_time(sbi, REQ_TIME);
2191 return 0;
2192
2193 free_kobj:
2194 f2fs_sync_inode_meta(sbi);
2195 kobject_del(&sbi->s_kobj);
2196 kobject_put(&sbi->s_kobj);
2197 wait_for_completion(&sbi->s_kobj_unregister);
2198 free_proc:
2199 if (sbi->s_proc) {
2200 remove_proc_entry("segment_info", sbi->s_proc);
2201 remove_proc_entry("segment_bits", sbi->s_proc);
2202 remove_proc_entry(sb->s_id, f2fs_proc_root);
2203 }
2204 free_root_inode:
2205 dput(sb->s_root);
2206 sb->s_root = NULL;
2207 free_node_inode:
2208 truncate_inode_pages_final(NODE_MAPPING(sbi));
2209 mutex_lock(&sbi->umount_mutex);
2210 release_ino_entry(sbi, true);
2211 f2fs_leave_shrinker(sbi);
2212 /*
2213 * Some dirty meta pages can be produced by recover_orphan_inodes()
2214 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2215 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2216 * falls into an infinite loop in sync_meta_pages().
2217 */
2218 truncate_inode_pages_final(META_MAPPING(sbi));
2219 iput(sbi->node_inode);
2220 mutex_unlock(&sbi->umount_mutex);
2221 f2fs_destroy_stats(sbi);
2222 free_nm:
2223 destroy_node_manager(sbi);
2224 free_sm:
2225 destroy_segment_manager(sbi);
2226 free_devices:
2227 destroy_device_list(sbi);
2228 kfree(sbi->ckpt);
2229 free_meta_inode:
2230 make_bad_inode(sbi->meta_inode);
2231 iput(sbi->meta_inode);
2232 free_io_dummy:
2233 mempool_destroy(sbi->write_io_dummy);
2234 free_options:
2235 for (i = 0; i < NR_PAGE_TYPE; i++)
2236 kfree(sbi->write_io[i]);
2237 destroy_percpu_info(sbi);
2238 kfree(options);
2239 free_sb_buf:
2240 kfree(raw_super);
2241 free_sbi:
2242 if (sbi->s_chksum_driver)
2243 crypto_free_shash(sbi->s_chksum_driver);
2244 kfree(sbi);
2245
2246 /* give only one another chance */
2247 if (retry) {
2248 retry = false;
2249 shrink_dcache_sb(sb);
2250 goto try_onemore;
2251 }
2252 return err;
2253 }
2254
2255 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2256 const char *dev_name, void *data)
2257 {
2258 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2259 }
2260
2261 static void kill_f2fs_super(struct super_block *sb)
2262 {
2263 if (sb->s_root)
2264 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2265 kill_block_super(sb);
2266 }
2267
2268 static struct file_system_type f2fs_fs_type = {
2269 .owner = THIS_MODULE,
2270 .name = "f2fs",
2271 .mount = f2fs_mount,
2272 .kill_sb = kill_f2fs_super,
2273 .fs_flags = FS_REQUIRES_DEV,
2274 };
2275 MODULE_ALIAS_FS("f2fs");
2276
2277 static int __init init_inodecache(void)
2278 {
2279 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2280 sizeof(struct f2fs_inode_info), 0,
2281 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2282 if (!f2fs_inode_cachep)
2283 return -ENOMEM;
2284 return 0;
2285 }
2286
2287 static void destroy_inodecache(void)
2288 {
2289 /*
2290 * Make sure all delayed rcu free inodes are flushed before we
2291 * destroy cache.
2292 */
2293 rcu_barrier();
2294 kmem_cache_destroy(f2fs_inode_cachep);
2295 }
2296
2297 static int __init init_f2fs_fs(void)
2298 {
2299 int err;
2300
2301 f2fs_build_trace_ios();
2302
2303 err = init_inodecache();
2304 if (err)
2305 goto fail;
2306 err = create_node_manager_caches();
2307 if (err)
2308 goto free_inodecache;
2309 err = create_segment_manager_caches();
2310 if (err)
2311 goto free_node_manager_caches;
2312 err = create_checkpoint_caches();
2313 if (err)
2314 goto free_segment_manager_caches;
2315 err = create_extent_cache();
2316 if (err)
2317 goto free_checkpoint_caches;
2318 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
2319 if (!f2fs_kset) {
2320 err = -ENOMEM;
2321 goto free_extent_cache;
2322 }
2323 err = register_shrinker(&f2fs_shrinker_info);
2324 if (err)
2325 goto free_kset;
2326
2327 err = register_filesystem(&f2fs_fs_type);
2328 if (err)
2329 goto free_shrinker;
2330 err = f2fs_create_root_stats();
2331 if (err)
2332 goto free_filesystem;
2333 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2334 return 0;
2335
2336 free_filesystem:
2337 unregister_filesystem(&f2fs_fs_type);
2338 free_shrinker:
2339 unregister_shrinker(&f2fs_shrinker_info);
2340 free_kset:
2341 kset_unregister(f2fs_kset);
2342 free_extent_cache:
2343 destroy_extent_cache();
2344 free_checkpoint_caches:
2345 destroy_checkpoint_caches();
2346 free_segment_manager_caches:
2347 destroy_segment_manager_caches();
2348 free_node_manager_caches:
2349 destroy_node_manager_caches();
2350 free_inodecache:
2351 destroy_inodecache();
2352 fail:
2353 return err;
2354 }
2355
2356 static void __exit exit_f2fs_fs(void)
2357 {
2358 remove_proc_entry("fs/f2fs", NULL);
2359 f2fs_destroy_root_stats();
2360 unregister_filesystem(&f2fs_fs_type);
2361 unregister_shrinker(&f2fs_shrinker_info);
2362 kset_unregister(f2fs_kset);
2363 destroy_extent_cache();
2364 destroy_checkpoint_caches();
2365 destroy_segment_manager_caches();
2366 destroy_node_manager_caches();
2367 destroy_inodecache();
2368 f2fs_destroy_trace_ios();
2369 }
2370
2371 module_init(init_f2fs_fs)
2372 module_exit(exit_f2fs_fs)
2373
2374 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2375 MODULE_DESCRIPTION("Flash Friendly File System");
2376 MODULE_LICENSE("GPL");
2377
Ubuntu Jammy Linux kernel mirror