]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - fs/f2fs/super.c
Merge tag 'for-linus-4.12b-rc0c-tag' of git://git.kernel.org/pub/scm/linux/kernel...
[mirror_ubuntu-bionic-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
628 /* Will be used by directory only */
629 fi->i_dir_level = F2FS_SB(sb)->dir_level;
630 return &fi->vfs_inode;
631 }
632
633 static int f2fs_drop_inode(struct inode *inode)
634 {
635 int ret;
636 /*
637 * This is to avoid a deadlock condition like below.
638 * writeback_single_inode(inode)
639 * - f2fs_write_data_page
640 * - f2fs_gc -> iput -> evict
641 * - inode_wait_for_writeback(inode)
642 */
643 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
644 if (!inode->i_nlink && !is_bad_inode(inode)) {
645 /* to avoid evict_inode call simultaneously */
646 atomic_inc(&inode->i_count);
647 spin_unlock(&inode->i_lock);
648
649 /* some remained atomic pages should discarded */
650 if (f2fs_is_atomic_file(inode))
651 drop_inmem_pages(inode);
652
653 /* should remain fi->extent_tree for writepage */
654 f2fs_destroy_extent_node(inode);
655
656 sb_start_intwrite(inode->i_sb);
657 f2fs_i_size_write(inode, 0);
658
659 if (F2FS_HAS_BLOCKS(inode))
660 f2fs_truncate(inode);
661
662 sb_end_intwrite(inode->i_sb);
663
664 fscrypt_put_encryption_info(inode, NULL);
665 spin_lock(&inode->i_lock);
666 atomic_dec(&inode->i_count);
667 }
668 trace_f2fs_drop_inode(inode, 0);
669 return 0;
670 }
671 ret = generic_drop_inode(inode);
672 trace_f2fs_drop_inode(inode, ret);
673 return ret;
674 }
675
676 int f2fs_inode_dirtied(struct inode *inode, bool sync)
677 {
678 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
679 int ret = 0;
680
681 spin_lock(&sbi->inode_lock[DIRTY_META]);
682 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
683 ret = 1;
684 } else {
685 set_inode_flag(inode, FI_DIRTY_INODE);
686 stat_inc_dirty_inode(sbi, DIRTY_META);
687 }
688 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
689 list_add_tail(&F2FS_I(inode)->gdirty_list,
690 &sbi->inode_list[DIRTY_META]);
691 inc_page_count(sbi, F2FS_DIRTY_IMETA);
692 }
693 spin_unlock(&sbi->inode_lock[DIRTY_META]);
694 return ret;
695 }
696
697 void f2fs_inode_synced(struct inode *inode)
698 {
699 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
700
701 spin_lock(&sbi->inode_lock[DIRTY_META]);
702 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
703 spin_unlock(&sbi->inode_lock[DIRTY_META]);
704 return;
705 }
706 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
707 list_del_init(&F2FS_I(inode)->gdirty_list);
708 dec_page_count(sbi, F2FS_DIRTY_IMETA);
709 }
710 clear_inode_flag(inode, FI_DIRTY_INODE);
711 clear_inode_flag(inode, FI_AUTO_RECOVER);
712 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
713 spin_unlock(&sbi->inode_lock[DIRTY_META]);
714 }
715
716 /*
717 * f2fs_dirty_inode() is called from __mark_inode_dirty()
718 *
719 * We should call set_dirty_inode to write the dirty inode through write_inode.
720 */
721 static void f2fs_dirty_inode(struct inode *inode, int flags)
722 {
723 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
724
725 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
726 inode->i_ino == F2FS_META_INO(sbi))
727 return;
728
729 if (flags == I_DIRTY_TIME)
730 return;
731
732 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
733 clear_inode_flag(inode, FI_AUTO_RECOVER);
734
735 f2fs_inode_dirtied(inode, false);
736 }
737
738 static void f2fs_i_callback(struct rcu_head *head)
739 {
740 struct inode *inode = container_of(head, struct inode, i_rcu);
741 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
742 }
743
744 static void f2fs_destroy_inode(struct inode *inode)
745 {
746 call_rcu(&inode->i_rcu, f2fs_i_callback);
747 }
748
749 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
750 {
751 percpu_counter_destroy(&sbi->alloc_valid_block_count);
752 percpu_counter_destroy(&sbi->total_valid_inode_count);
753 }
754
755 static void destroy_device_list(struct f2fs_sb_info *sbi)
756 {
757 int i;
758
759 for (i = 0; i < sbi->s_ndevs; i++) {
760 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
761 #ifdef CONFIG_BLK_DEV_ZONED
762 kfree(FDEV(i).blkz_type);
763 #endif
764 }
765 kfree(sbi->devs);
766 }
767
768 static void f2fs_put_super(struct super_block *sb)
769 {
770 struct f2fs_sb_info *sbi = F2FS_SB(sb);
771
772 if (sbi->s_proc) {
773 remove_proc_entry("segment_info", sbi->s_proc);
774 remove_proc_entry("segment_bits", sbi->s_proc);
775 remove_proc_entry(sb->s_id, f2fs_proc_root);
776 }
777 kobject_del(&sbi->s_kobj);
778
779 stop_gc_thread(sbi);
780
781 /* prevent remaining shrinker jobs */
782 mutex_lock(&sbi->umount_mutex);
783
784 /*
785 * We don't need to do checkpoint when superblock is clean.
786 * But, the previous checkpoint was not done by umount, it needs to do
787 * clean checkpoint again.
788 */
789 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
790 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
791 struct cp_control cpc = {
792 .reason = CP_UMOUNT,
793 };
794 write_checkpoint(sbi, &cpc);
795 }
796
797 /* be sure to wait for any on-going discard commands */
798 f2fs_wait_discard_bios(sbi);
799
800 if (!sbi->discard_blks) {
801 struct cp_control cpc = {
802 .reason = CP_UMOUNT | CP_TRIMMED,
803 };
804 write_checkpoint(sbi, &cpc);
805 }
806
807 /* write_checkpoint can update stat informaion */
808 f2fs_destroy_stats(sbi);
809
810 /*
811 * normally superblock is clean, so we need to release this.
812 * In addition, EIO will skip do checkpoint, we need this as well.
813 */
814 release_ino_entry(sbi, true);
815
816 f2fs_leave_shrinker(sbi);
817 mutex_unlock(&sbi->umount_mutex);
818
819 /* our cp_error case, we can wait for any writeback page */
820 f2fs_flush_merged_bios(sbi);
821
822 iput(sbi->node_inode);
823 iput(sbi->meta_inode);
824
825 /* destroy f2fs internal modules */
826 destroy_node_manager(sbi);
827 destroy_segment_manager(sbi);
828
829 kfree(sbi->ckpt);
830 kobject_put(&sbi->s_kobj);
831 wait_for_completion(&sbi->s_kobj_unregister);
832
833 sb->s_fs_info = NULL;
834 if (sbi->s_chksum_driver)
835 crypto_free_shash(sbi->s_chksum_driver);
836 kfree(sbi->raw_super);
837
838 destroy_device_list(sbi);
839 mempool_destroy(sbi->write_io_dummy);
840 destroy_percpu_info(sbi);
841 kfree(sbi);
842 }
843
844 int f2fs_sync_fs(struct super_block *sb, int sync)
845 {
846 struct f2fs_sb_info *sbi = F2FS_SB(sb);
847 int err = 0;
848
849 trace_f2fs_sync_fs(sb, sync);
850
851 if (sync) {
852 struct cp_control cpc;
853
854 cpc.reason = __get_cp_reason(sbi);
855
856 mutex_lock(&sbi->gc_mutex);
857 err = write_checkpoint(sbi, &cpc);
858 mutex_unlock(&sbi->gc_mutex);
859 }
860 f2fs_trace_ios(NULL, 1);
861
862 return err;
863 }
864
865 static int f2fs_freeze(struct super_block *sb)
866 {
867 if (f2fs_readonly(sb))
868 return 0;
869
870 /* IO error happened before */
871 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
872 return -EIO;
873
874 /* must be clean, since sync_filesystem() was already called */
875 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
876 return -EINVAL;
877 return 0;
878 }
879
880 static int f2fs_unfreeze(struct super_block *sb)
881 {
882 return 0;
883 }
884
885 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
886 {
887 struct super_block *sb = dentry->d_sb;
888 struct f2fs_sb_info *sbi = F2FS_SB(sb);
889 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
890 block_t total_count, user_block_count, start_count, ovp_count;
891
892 total_count = le64_to_cpu(sbi->raw_super->block_count);
893 user_block_count = sbi->user_block_count;
894 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
895 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
896 buf->f_type = F2FS_SUPER_MAGIC;
897 buf->f_bsize = sbi->blocksize;
898
899 buf->f_blocks = total_count - start_count;
900 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
901 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
902
903 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
904 buf->f_ffree = min(buf->f_files - valid_node_count(sbi),
905 buf->f_bavail);
906
907 buf->f_namelen = F2FS_NAME_LEN;
908 buf->f_fsid.val[0] = (u32)id;
909 buf->f_fsid.val[1] = (u32)(id >> 32);
910
911 return 0;
912 }
913
914 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
915 {
916 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
917
918 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
919 if (test_opt(sbi, FORCE_FG_GC))
920 seq_printf(seq, ",background_gc=%s", "sync");
921 else
922 seq_printf(seq, ",background_gc=%s", "on");
923 } else {
924 seq_printf(seq, ",background_gc=%s", "off");
925 }
926 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
927 seq_puts(seq, ",disable_roll_forward");
928 if (test_opt(sbi, DISCARD))
929 seq_puts(seq, ",discard");
930 if (test_opt(sbi, NOHEAP))
931 seq_puts(seq, ",no_heap");
932 else
933 seq_puts(seq, ",heap");
934 #ifdef CONFIG_F2FS_FS_XATTR
935 if (test_opt(sbi, XATTR_USER))
936 seq_puts(seq, ",user_xattr");
937 else
938 seq_puts(seq, ",nouser_xattr");
939 if (test_opt(sbi, INLINE_XATTR))
940 seq_puts(seq, ",inline_xattr");
941 else
942 seq_puts(seq, ",noinline_xattr");
943 #endif
944 #ifdef CONFIG_F2FS_FS_POSIX_ACL
945 if (test_opt(sbi, POSIX_ACL))
946 seq_puts(seq, ",acl");
947 else
948 seq_puts(seq, ",noacl");
949 #endif
950 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
951 seq_puts(seq, ",disable_ext_identify");
952 if (test_opt(sbi, INLINE_DATA))
953 seq_puts(seq, ",inline_data");
954 else
955 seq_puts(seq, ",noinline_data");
956 if (test_opt(sbi, INLINE_DENTRY))
957 seq_puts(seq, ",inline_dentry");
958 else
959 seq_puts(seq, ",noinline_dentry");
960 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
961 seq_puts(seq, ",flush_merge");
962 if (test_opt(sbi, NOBARRIER))
963 seq_puts(seq, ",nobarrier");
964 if (test_opt(sbi, FASTBOOT))
965 seq_puts(seq, ",fastboot");
966 if (test_opt(sbi, EXTENT_CACHE))
967 seq_puts(seq, ",extent_cache");
968 else
969 seq_puts(seq, ",noextent_cache");
970 if (test_opt(sbi, DATA_FLUSH))
971 seq_puts(seq, ",data_flush");
972
973 seq_puts(seq, ",mode=");
974 if (test_opt(sbi, ADAPTIVE))
975 seq_puts(seq, "adaptive");
976 else if (test_opt(sbi, LFS))
977 seq_puts(seq, "lfs");
978 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
979 if (F2FS_IO_SIZE_BITS(sbi))
980 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
981 #ifdef CONFIG_F2FS_FAULT_INJECTION
982 if (test_opt(sbi, FAULT_INJECTION))
983 seq_puts(seq, ",fault_injection");
984 #endif
985
986 return 0;
987 }
988
989 static int segment_info_seq_show(struct seq_file *seq, void *offset)
990 {
991 struct super_block *sb = seq->private;
992 struct f2fs_sb_info *sbi = F2FS_SB(sb);
993 unsigned int total_segs =
994 le32_to_cpu(sbi->raw_super->segment_count_main);
995 int i;
996
997 seq_puts(seq, "format: segment_type|valid_blocks\n"
998 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
999
1000 for (i = 0; i < total_segs; i++) {
1001 struct seg_entry *se = get_seg_entry(sbi, i);
1002
1003 if ((i % 10) == 0)
1004 seq_printf(seq, "%-10d", i);
1005 seq_printf(seq, "%d|%-3u", se->type,
1006 get_valid_blocks(sbi, i, false));
1007 if ((i % 10) == 9 || i == (total_segs - 1))
1008 seq_putc(seq, '\n');
1009 else
1010 seq_putc(seq, ' ');
1011 }
1012
1013 return 0;
1014 }
1015
1016 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
1017 {
1018 struct super_block *sb = seq->private;
1019 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1020 unsigned int total_segs =
1021 le32_to_cpu(sbi->raw_super->segment_count_main);
1022 int i, j;
1023
1024 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
1025 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
1026
1027 for (i = 0; i < total_segs; i++) {
1028 struct seg_entry *se = get_seg_entry(sbi, i);
1029
1030 seq_printf(seq, "%-10d", i);
1031 seq_printf(seq, "%d|%-3u|", se->type,
1032 get_valid_blocks(sbi, i, false));
1033 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
1034 seq_printf(seq, " %.2x", se->cur_valid_map[j]);
1035 seq_putc(seq, '\n');
1036 }
1037 return 0;
1038 }
1039
1040 #define F2FS_PROC_FILE_DEF(_name) \
1041 static int _name##_open_fs(struct inode *inode, struct file *file) \
1042 { \
1043 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
1044 } \
1045 \
1046 static const struct file_operations f2fs_seq_##_name##_fops = { \
1047 .open = _name##_open_fs, \
1048 .read = seq_read, \
1049 .llseek = seq_lseek, \
1050 .release = single_release, \
1051 };
1052
1053 F2FS_PROC_FILE_DEF(segment_info);
1054 F2FS_PROC_FILE_DEF(segment_bits);
1055
1056 static void default_options(struct f2fs_sb_info *sbi)
1057 {
1058 /* init some FS parameters */
1059 sbi->active_logs = NR_CURSEG_TYPE;
1060
1061 set_opt(sbi, BG_GC);
1062 set_opt(sbi, INLINE_XATTR);
1063 set_opt(sbi, INLINE_DATA);
1064 set_opt(sbi, INLINE_DENTRY);
1065 set_opt(sbi, EXTENT_CACHE);
1066 set_opt(sbi, NOHEAP);
1067 sbi->sb->s_flags |= MS_LAZYTIME;
1068 set_opt(sbi, FLUSH_MERGE);
1069 if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1070 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1071 set_opt(sbi, DISCARD);
1072 } else {
1073 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1074 }
1075
1076 #ifdef CONFIG_F2FS_FS_XATTR
1077 set_opt(sbi, XATTR_USER);
1078 #endif
1079 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1080 set_opt(sbi, POSIX_ACL);
1081 #endif
1082
1083 #ifdef CONFIG_F2FS_FAULT_INJECTION
1084 f2fs_build_fault_attr(sbi, 0);
1085 #endif
1086 }
1087
1088 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1089 {
1090 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1091 struct f2fs_mount_info org_mount_opt;
1092 int err, active_logs;
1093 bool need_restart_gc = false;
1094 bool need_stop_gc = false;
1095 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1096 #ifdef CONFIG_F2FS_FAULT_INJECTION
1097 struct f2fs_fault_info ffi = sbi->fault_info;
1098 #endif
1099
1100 /*
1101 * Save the old mount options in case we
1102 * need to restore them.
1103 */
1104 org_mount_opt = sbi->mount_opt;
1105 active_logs = sbi->active_logs;
1106
1107 /* recover superblocks we couldn't write due to previous RO mount */
1108 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1109 err = f2fs_commit_super(sbi, false);
1110 f2fs_msg(sb, KERN_INFO,
1111 "Try to recover all the superblocks, ret: %d", err);
1112 if (!err)
1113 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1114 }
1115
1116 sbi->mount_opt.opt = 0;
1117 default_options(sbi);
1118
1119 /* parse mount options */
1120 err = parse_options(sb, data);
1121 if (err)
1122 goto restore_opts;
1123
1124 /*
1125 * Previous and new state of filesystem is RO,
1126 * so skip checking GC and FLUSH_MERGE conditions.
1127 */
1128 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1129 goto skip;
1130
1131 /* disallow enable/disable extent_cache dynamically */
1132 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1133 err = -EINVAL;
1134 f2fs_msg(sbi->sb, KERN_WARNING,
1135 "switch extent_cache option is not allowed");
1136 goto restore_opts;
1137 }
1138
1139 /*
1140 * We stop the GC thread if FS is mounted as RO
1141 * or if background_gc = off is passed in mount
1142 * option. Also sync the filesystem.
1143 */
1144 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1145 if (sbi->gc_thread) {
1146 stop_gc_thread(sbi);
1147 need_restart_gc = true;
1148 }
1149 } else if (!sbi->gc_thread) {
1150 err = start_gc_thread(sbi);
1151 if (err)
1152 goto restore_opts;
1153 need_stop_gc = true;
1154 }
1155
1156 if (*flags & MS_RDONLY) {
1157 writeback_inodes_sb(sb, WB_REASON_SYNC);
1158 sync_inodes_sb(sb);
1159
1160 set_sbi_flag(sbi, SBI_IS_DIRTY);
1161 set_sbi_flag(sbi, SBI_IS_CLOSE);
1162 f2fs_sync_fs(sb, 1);
1163 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1164 }
1165
1166 /*
1167 * We stop issue flush thread if FS is mounted as RO
1168 * or if flush_merge is not passed in mount option.
1169 */
1170 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1171 clear_opt(sbi, FLUSH_MERGE);
1172 destroy_flush_cmd_control(sbi, false);
1173 } else {
1174 err = create_flush_cmd_control(sbi);
1175 if (err)
1176 goto restore_gc;
1177 }
1178 skip:
1179 /* Update the POSIXACL Flag */
1180 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1181 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1182
1183 return 0;
1184 restore_gc:
1185 if (need_restart_gc) {
1186 if (start_gc_thread(sbi))
1187 f2fs_msg(sbi->sb, KERN_WARNING,
1188 "background gc thread has stopped");
1189 } else if (need_stop_gc) {
1190 stop_gc_thread(sbi);
1191 }
1192 restore_opts:
1193 sbi->mount_opt = org_mount_opt;
1194 sbi->active_logs = active_logs;
1195 #ifdef CONFIG_F2FS_FAULT_INJECTION
1196 sbi->fault_info = ffi;
1197 #endif
1198 return err;
1199 }
1200
1201 static struct super_operations f2fs_sops = {
1202 .alloc_inode = f2fs_alloc_inode,
1203 .drop_inode = f2fs_drop_inode,
1204 .destroy_inode = f2fs_destroy_inode,
1205 .write_inode = f2fs_write_inode,
1206 .dirty_inode = f2fs_dirty_inode,
1207 .show_options = f2fs_show_options,
1208 .evict_inode = f2fs_evict_inode,
1209 .put_super = f2fs_put_super,
1210 .sync_fs = f2fs_sync_fs,
1211 .freeze_fs = f2fs_freeze,
1212 .unfreeze_fs = f2fs_unfreeze,
1213 .statfs = f2fs_statfs,
1214 .remount_fs = f2fs_remount,
1215 };
1216
1217 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1218 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1219 {
1220 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1221 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1222 ctx, len, NULL);
1223 }
1224
1225 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1226 void *fs_data)
1227 {
1228 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1229 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1230 ctx, len, fs_data, XATTR_CREATE);
1231 }
1232
1233 static unsigned f2fs_max_namelen(struct inode *inode)
1234 {
1235 return S_ISLNK(inode->i_mode) ?
1236 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1237 }
1238
1239 static const struct fscrypt_operations f2fs_cryptops = {
1240 .key_prefix = "f2fs:",
1241 .get_context = f2fs_get_context,
1242 .set_context = f2fs_set_context,
1243 .is_encrypted = f2fs_encrypted_inode,
1244 .empty_dir = f2fs_empty_dir,
1245 .max_namelen = f2fs_max_namelen,
1246 };
1247 #else
1248 static const struct fscrypt_operations f2fs_cryptops = {
1249 .is_encrypted = f2fs_encrypted_inode,
1250 };
1251 #endif
1252
1253 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1254 u64 ino, u32 generation)
1255 {
1256 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1257 struct inode *inode;
1258
1259 if (check_nid_range(sbi, ino))
1260 return ERR_PTR(-ESTALE);
1261
1262 /*
1263 * f2fs_iget isn't quite right if the inode is currently unallocated!
1264 * However f2fs_iget currently does appropriate checks to handle stale
1265 * inodes so everything is OK.
1266 */
1267 inode = f2fs_iget(sb, ino);
1268 if (IS_ERR(inode))
1269 return ERR_CAST(inode);
1270 if (unlikely(generation && inode->i_generation != generation)) {
1271 /* we didn't find the right inode.. */
1272 iput(inode);
1273 return ERR_PTR(-ESTALE);
1274 }
1275 return inode;
1276 }
1277
1278 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1279 int fh_len, int fh_type)
1280 {
1281 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1282 f2fs_nfs_get_inode);
1283 }
1284
1285 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1286 int fh_len, int fh_type)
1287 {
1288 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1289 f2fs_nfs_get_inode);
1290 }
1291
1292 static const struct export_operations f2fs_export_ops = {
1293 .fh_to_dentry = f2fs_fh_to_dentry,
1294 .fh_to_parent = f2fs_fh_to_parent,
1295 .get_parent = f2fs_get_parent,
1296 };
1297
1298 static loff_t max_file_blocks(void)
1299 {
1300 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1301 loff_t leaf_count = ADDRS_PER_BLOCK;
1302
1303 /* two direct node blocks */
1304 result += (leaf_count * 2);
1305
1306 /* two indirect node blocks */
1307 leaf_count *= NIDS_PER_BLOCK;
1308 result += (leaf_count * 2);
1309
1310 /* one double indirect node block */
1311 leaf_count *= NIDS_PER_BLOCK;
1312 result += leaf_count;
1313
1314 return result;
1315 }
1316
1317 static int __f2fs_commit_super(struct buffer_head *bh,
1318 struct f2fs_super_block *super)
1319 {
1320 lock_buffer(bh);
1321 if (super)
1322 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1323 set_buffer_uptodate(bh);
1324 set_buffer_dirty(bh);
1325 unlock_buffer(bh);
1326
1327 /* it's rare case, we can do fua all the time */
1328 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
1329 }
1330
1331 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1332 struct buffer_head *bh)
1333 {
1334 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1335 (bh->b_data + F2FS_SUPER_OFFSET);
1336 struct super_block *sb = sbi->sb;
1337 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1338 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1339 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1340 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1341 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1342 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1343 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1344 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1345 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1346 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1347 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1348 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1349 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1350 u64 main_end_blkaddr = main_blkaddr +
1351 (segment_count_main << log_blocks_per_seg);
1352 u64 seg_end_blkaddr = segment0_blkaddr +
1353 (segment_count << log_blocks_per_seg);
1354
1355 if (segment0_blkaddr != cp_blkaddr) {
1356 f2fs_msg(sb, KERN_INFO,
1357 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1358 segment0_blkaddr, cp_blkaddr);
1359 return true;
1360 }
1361
1362 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1363 sit_blkaddr) {
1364 f2fs_msg(sb, KERN_INFO,
1365 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1366 cp_blkaddr, sit_blkaddr,
1367 segment_count_ckpt << log_blocks_per_seg);
1368 return true;
1369 }
1370
1371 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1372 nat_blkaddr) {
1373 f2fs_msg(sb, KERN_INFO,
1374 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1375 sit_blkaddr, nat_blkaddr,
1376 segment_count_sit << log_blocks_per_seg);
1377 return true;
1378 }
1379
1380 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1381 ssa_blkaddr) {
1382 f2fs_msg(sb, KERN_INFO,
1383 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1384 nat_blkaddr, ssa_blkaddr,
1385 segment_count_nat << log_blocks_per_seg);
1386 return true;
1387 }
1388
1389 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1390 main_blkaddr) {
1391 f2fs_msg(sb, KERN_INFO,
1392 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1393 ssa_blkaddr, main_blkaddr,
1394 segment_count_ssa << log_blocks_per_seg);
1395 return true;
1396 }
1397
1398 if (main_end_blkaddr > seg_end_blkaddr) {
1399 f2fs_msg(sb, KERN_INFO,
1400 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1401 main_blkaddr,
1402 segment0_blkaddr +
1403 (segment_count << log_blocks_per_seg),
1404 segment_count_main << log_blocks_per_seg);
1405 return true;
1406 } else if (main_end_blkaddr < seg_end_blkaddr) {
1407 int err = 0;
1408 char *res;
1409
1410 /* fix in-memory information all the time */
1411 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1412 segment0_blkaddr) >> log_blocks_per_seg);
1413
1414 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1415 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1416 res = "internally";
1417 } else {
1418 err = __f2fs_commit_super(bh, NULL);
1419 res = err ? "failed" : "done";
1420 }
1421 f2fs_msg(sb, KERN_INFO,
1422 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1423 res, main_blkaddr,
1424 segment0_blkaddr +
1425 (segment_count << log_blocks_per_seg),
1426 segment_count_main << log_blocks_per_seg);
1427 if (err)
1428 return true;
1429 }
1430 return false;
1431 }
1432
1433 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1434 struct buffer_head *bh)
1435 {
1436 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1437 (bh->b_data + F2FS_SUPER_OFFSET);
1438 struct super_block *sb = sbi->sb;
1439 unsigned int blocksize;
1440
1441 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1442 f2fs_msg(sb, KERN_INFO,
1443 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1444 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1445 return 1;
1446 }
1447
1448 /* Currently, support only 4KB page cache size */
1449 if (F2FS_BLKSIZE != PAGE_SIZE) {
1450 f2fs_msg(sb, KERN_INFO,
1451 "Invalid page_cache_size (%lu), supports only 4KB\n",
1452 PAGE_SIZE);
1453 return 1;
1454 }
1455
1456 /* Currently, support only 4KB block size */
1457 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1458 if (blocksize != F2FS_BLKSIZE) {
1459 f2fs_msg(sb, KERN_INFO,
1460 "Invalid blocksize (%u), supports only 4KB\n",
1461 blocksize);
1462 return 1;
1463 }
1464
1465 /* check log blocks per segment */
1466 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1467 f2fs_msg(sb, KERN_INFO,
1468 "Invalid log blocks per segment (%u)\n",
1469 le32_to_cpu(raw_super->log_blocks_per_seg));
1470 return 1;
1471 }
1472
1473 /* Currently, support 512/1024/2048/4096 bytes sector size */
1474 if (le32_to_cpu(raw_super->log_sectorsize) >
1475 F2FS_MAX_LOG_SECTOR_SIZE ||
1476 le32_to_cpu(raw_super->log_sectorsize) <
1477 F2FS_MIN_LOG_SECTOR_SIZE) {
1478 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1479 le32_to_cpu(raw_super->log_sectorsize));
1480 return 1;
1481 }
1482 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1483 le32_to_cpu(raw_super->log_sectorsize) !=
1484 F2FS_MAX_LOG_SECTOR_SIZE) {
1485 f2fs_msg(sb, KERN_INFO,
1486 "Invalid log sectors per block(%u) log sectorsize(%u)",
1487 le32_to_cpu(raw_super->log_sectors_per_block),
1488 le32_to_cpu(raw_super->log_sectorsize));
1489 return 1;
1490 }
1491
1492 /* check reserved ino info */
1493 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1494 le32_to_cpu(raw_super->meta_ino) != 2 ||
1495 le32_to_cpu(raw_super->root_ino) != 3) {
1496 f2fs_msg(sb, KERN_INFO,
1497 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1498 le32_to_cpu(raw_super->node_ino),
1499 le32_to_cpu(raw_super->meta_ino),
1500 le32_to_cpu(raw_super->root_ino));
1501 return 1;
1502 }
1503
1504 if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) {
1505 f2fs_msg(sb, KERN_INFO,
1506 "Invalid segment count (%u)",
1507 le32_to_cpu(raw_super->segment_count));
1508 return 1;
1509 }
1510
1511 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1512 if (sanity_check_area_boundary(sbi, bh))
1513 return 1;
1514
1515 return 0;
1516 }
1517
1518 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1519 {
1520 unsigned int total, fsmeta;
1521 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1522 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1523 unsigned int ovp_segments, reserved_segments;
1524
1525 total = le32_to_cpu(raw_super->segment_count);
1526 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1527 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1528 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1529 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1530 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1531
1532 if (unlikely(fsmeta >= total))
1533 return 1;
1534
1535 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1536 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1537
1538 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1539 ovp_segments == 0 || reserved_segments == 0)) {
1540 f2fs_msg(sbi->sb, KERN_ERR,
1541 "Wrong layout: check mkfs.f2fs version");
1542 return 1;
1543 }
1544
1545 if (unlikely(f2fs_cp_error(sbi))) {
1546 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1547 return 1;
1548 }
1549 return 0;
1550 }
1551
1552 static void init_sb_info(struct f2fs_sb_info *sbi)
1553 {
1554 struct f2fs_super_block *raw_super = sbi->raw_super;
1555 int i;
1556
1557 sbi->log_sectors_per_block =
1558 le32_to_cpu(raw_super->log_sectors_per_block);
1559 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1560 sbi->blocksize = 1 << sbi->log_blocksize;
1561 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1562 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1563 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1564 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1565 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1566 sbi->total_node_count =
1567 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1568 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1569 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1570 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1571 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1572 sbi->cur_victim_sec = NULL_SECNO;
1573 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1574
1575 sbi->dir_level = DEF_DIR_LEVEL;
1576 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1577 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1578 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1579
1580 for (i = 0; i < NR_COUNT_TYPE; i++)
1581 atomic_set(&sbi->nr_pages[i], 0);
1582
1583 atomic_set(&sbi->wb_sync_req, 0);
1584
1585 INIT_LIST_HEAD(&sbi->s_list);
1586 mutex_init(&sbi->umount_mutex);
1587 mutex_init(&sbi->wio_mutex[NODE]);
1588 mutex_init(&sbi->wio_mutex[DATA]);
1589 spin_lock_init(&sbi->cp_lock);
1590 }
1591
1592 static int init_percpu_info(struct f2fs_sb_info *sbi)
1593 {
1594 int err;
1595
1596 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1597 if (err)
1598 return err;
1599
1600 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1601 GFP_KERNEL);
1602 }
1603
1604 #ifdef CONFIG_BLK_DEV_ZONED
1605 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
1606 {
1607 struct block_device *bdev = FDEV(devi).bdev;
1608 sector_t nr_sectors = bdev->bd_part->nr_sects;
1609 sector_t sector = 0;
1610 struct blk_zone *zones;
1611 unsigned int i, nr_zones;
1612 unsigned int n = 0;
1613 int err = -EIO;
1614
1615 if (!f2fs_sb_mounted_blkzoned(sbi->sb))
1616 return 0;
1617
1618 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
1619 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
1620 return -EINVAL;
1621 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
1622 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
1623 __ilog2_u32(sbi->blocks_per_blkz))
1624 return -EINVAL;
1625 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
1626 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
1627 sbi->log_blocks_per_blkz;
1628 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
1629 FDEV(devi).nr_blkz++;
1630
1631 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
1632 if (!FDEV(devi).blkz_type)
1633 return -ENOMEM;
1634
1635 #define F2FS_REPORT_NR_ZONES 4096
1636
1637 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
1638 GFP_KERNEL);
1639 if (!zones)
1640 return -ENOMEM;
1641
1642 /* Get block zones type */
1643 while (zones && sector < nr_sectors) {
1644
1645 nr_zones = F2FS_REPORT_NR_ZONES;
1646 err = blkdev_report_zones(bdev, sector,
1647 zones, &nr_zones,
1648 GFP_KERNEL);
1649 if (err)
1650 break;
1651 if (!nr_zones) {
1652 err = -EIO;
1653 break;
1654 }
1655
1656 for (i = 0; i < nr_zones; i++) {
1657 FDEV(devi).blkz_type[n] = zones[i].type;
1658 sector += zones[i].len;
1659 n++;
1660 }
1661 }
1662
1663 kfree(zones);
1664
1665 return err;
1666 }
1667 #endif
1668
1669 /*
1670 * Read f2fs raw super block.
1671 * Because we have two copies of super block, so read both of them
1672 * to get the first valid one. If any one of them is broken, we pass
1673 * them recovery flag back to the caller.
1674 */
1675 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1676 struct f2fs_super_block **raw_super,
1677 int *valid_super_block, int *recovery)
1678 {
1679 struct super_block *sb = sbi->sb;
1680 int block;
1681 struct buffer_head *bh;
1682 struct f2fs_super_block *super;
1683 int err = 0;
1684
1685 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1686 if (!super)
1687 return -ENOMEM;
1688
1689 for (block = 0; block < 2; block++) {
1690 bh = sb_bread(sb, block);
1691 if (!bh) {
1692 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1693 block + 1);
1694 err = -EIO;
1695 continue;
1696 }
1697
1698 /* sanity checking of raw super */
1699 if (sanity_check_raw_super(sbi, bh)) {
1700 f2fs_msg(sb, KERN_ERR,
1701 "Can't find valid F2FS filesystem in %dth superblock",
1702 block + 1);
1703 err = -EINVAL;
1704 brelse(bh);
1705 continue;
1706 }
1707
1708 if (!*raw_super) {
1709 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1710 sizeof(*super));
1711 *valid_super_block = block;
1712 *raw_super = super;
1713 }
1714 brelse(bh);
1715 }
1716
1717 /* Fail to read any one of the superblocks*/
1718 if (err < 0)
1719 *recovery = 1;
1720
1721 /* No valid superblock */
1722 if (!*raw_super)
1723 kfree(super);
1724 else
1725 err = 0;
1726
1727 return err;
1728 }
1729
1730 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1731 {
1732 struct buffer_head *bh;
1733 int err;
1734
1735 if ((recover && f2fs_readonly(sbi->sb)) ||
1736 bdev_read_only(sbi->sb->s_bdev)) {
1737 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1738 return -EROFS;
1739 }
1740
1741 /* write back-up superblock first */
1742 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1743 if (!bh)
1744 return -EIO;
1745 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1746 brelse(bh);
1747
1748 /* if we are in recovery path, skip writing valid superblock */
1749 if (recover || err)
1750 return err;
1751
1752 /* write current valid superblock */
1753 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1754 if (!bh)
1755 return -EIO;
1756 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1757 brelse(bh);
1758 return err;
1759 }
1760
1761 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
1762 {
1763 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1764 unsigned int max_devices = MAX_DEVICES;
1765 int i;
1766
1767 /* Initialize single device information */
1768 if (!RDEV(0).path[0]) {
1769 if (!bdev_is_zoned(sbi->sb->s_bdev))
1770 return 0;
1771 max_devices = 1;
1772 }
1773
1774 /*
1775 * Initialize multiple devices information, or single
1776 * zoned block device information.
1777 */
1778 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
1779 GFP_KERNEL);
1780 if (!sbi->devs)
1781 return -ENOMEM;
1782
1783 for (i = 0; i < max_devices; i++) {
1784
1785 if (i > 0 && !RDEV(i).path[0])
1786 break;
1787
1788 if (max_devices == 1) {
1789 /* Single zoned block device mount */
1790 FDEV(0).bdev =
1791 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
1792 sbi->sb->s_mode, sbi->sb->s_type);
1793 } else {
1794 /* Multi-device mount */
1795 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
1796 FDEV(i).total_segments =
1797 le32_to_cpu(RDEV(i).total_segments);
1798 if (i == 0) {
1799 FDEV(i).start_blk = 0;
1800 FDEV(i).end_blk = FDEV(i).start_blk +
1801 (FDEV(i).total_segments <<
1802 sbi->log_blocks_per_seg) - 1 +
1803 le32_to_cpu(raw_super->segment0_blkaddr);
1804 } else {
1805 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
1806 FDEV(i).end_blk = FDEV(i).start_blk +
1807 (FDEV(i).total_segments <<
1808 sbi->log_blocks_per_seg) - 1;
1809 }
1810 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
1811 sbi->sb->s_mode, sbi->sb->s_type);
1812 }
1813 if (IS_ERR(FDEV(i).bdev))
1814 return PTR_ERR(FDEV(i).bdev);
1815
1816 /* to release errored devices */
1817 sbi->s_ndevs = i + 1;
1818
1819 #ifdef CONFIG_BLK_DEV_ZONED
1820 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
1821 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
1822 f2fs_msg(sbi->sb, KERN_ERR,
1823 "Zoned block device feature not enabled\n");
1824 return -EINVAL;
1825 }
1826 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
1827 if (init_blkz_info(sbi, i)) {
1828 f2fs_msg(sbi->sb, KERN_ERR,
1829 "Failed to initialize F2FS blkzone information");
1830 return -EINVAL;
1831 }
1832 if (max_devices == 1)
1833 break;
1834 f2fs_msg(sbi->sb, KERN_INFO,
1835 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
1836 i, FDEV(i).path,
1837 FDEV(i).total_segments,
1838 FDEV(i).start_blk, FDEV(i).end_blk,
1839 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
1840 "Host-aware" : "Host-managed");
1841 continue;
1842 }
1843 #endif
1844 f2fs_msg(sbi->sb, KERN_INFO,
1845 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
1846 i, FDEV(i).path,
1847 FDEV(i).total_segments,
1848 FDEV(i).start_blk, FDEV(i).end_blk);
1849 }
1850 f2fs_msg(sbi->sb, KERN_INFO,
1851 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
1852 return 0;
1853 }
1854
1855 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1856 {
1857 struct f2fs_sb_info *sbi;
1858 struct f2fs_super_block *raw_super;
1859 struct inode *root;
1860 int err;
1861 bool retry = true, need_fsck = false;
1862 char *options = NULL;
1863 int recovery, i, valid_super_block;
1864 struct curseg_info *seg_i;
1865
1866 try_onemore:
1867 err = -EINVAL;
1868 raw_super = NULL;
1869 valid_super_block = -1;
1870 recovery = 0;
1871
1872 /* allocate memory for f2fs-specific super block info */
1873 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1874 if (!sbi)
1875 return -ENOMEM;
1876
1877 sbi->sb = sb;
1878
1879 /* Load the checksum driver */
1880 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1881 if (IS_ERR(sbi->s_chksum_driver)) {
1882 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1883 err = PTR_ERR(sbi->s_chksum_driver);
1884 sbi->s_chksum_driver = NULL;
1885 goto free_sbi;
1886 }
1887
1888 /* set a block size */
1889 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1890 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1891 goto free_sbi;
1892 }
1893
1894 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1895 &recovery);
1896 if (err)
1897 goto free_sbi;
1898
1899 sb->s_fs_info = sbi;
1900 sbi->raw_super = raw_super;
1901
1902 /*
1903 * The BLKZONED feature indicates that the drive was formatted with
1904 * zone alignment optimization. This is optional for host-aware
1905 * devices, but mandatory for host-managed zoned block devices.
1906 */
1907 #ifndef CONFIG_BLK_DEV_ZONED
1908 if (f2fs_sb_mounted_blkzoned(sb)) {
1909 f2fs_msg(sb, KERN_ERR,
1910 "Zoned block device support is not enabled\n");
1911 goto free_sb_buf;
1912 }
1913 #endif
1914 default_options(sbi);
1915 /* parse mount options */
1916 options = kstrdup((const char *)data, GFP_KERNEL);
1917 if (data && !options) {
1918 err = -ENOMEM;
1919 goto free_sb_buf;
1920 }
1921
1922 err = parse_options(sb, options);
1923 if (err)
1924 goto free_options;
1925
1926 sbi->max_file_blocks = max_file_blocks();
1927 sb->s_maxbytes = sbi->max_file_blocks <<
1928 le32_to_cpu(raw_super->log_blocksize);
1929 sb->s_max_links = F2FS_LINK_MAX;
1930 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1931
1932 sb->s_op = &f2fs_sops;
1933 sb->s_cop = &f2fs_cryptops;
1934 sb->s_xattr = f2fs_xattr_handlers;
1935 sb->s_export_op = &f2fs_export_ops;
1936 sb->s_magic = F2FS_SUPER_MAGIC;
1937 sb->s_time_gran = 1;
1938 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1939 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1940 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1941
1942 /* init f2fs-specific super block info */
1943 sbi->valid_super_block = valid_super_block;
1944 mutex_init(&sbi->gc_mutex);
1945 mutex_init(&sbi->cp_mutex);
1946 init_rwsem(&sbi->node_write);
1947 init_rwsem(&sbi->node_change);
1948
1949 /* disallow all the data/node/meta page writes */
1950 set_sbi_flag(sbi, SBI_POR_DOING);
1951 spin_lock_init(&sbi->stat_lock);
1952
1953 init_rwsem(&sbi->read_io.io_rwsem);
1954 sbi->read_io.sbi = sbi;
1955 sbi->read_io.bio = NULL;
1956 for (i = 0; i < NR_PAGE_TYPE; i++) {
1957 init_rwsem(&sbi->write_io[i].io_rwsem);
1958 sbi->write_io[i].sbi = sbi;
1959 sbi->write_io[i].bio = NULL;
1960 }
1961
1962 init_rwsem(&sbi->cp_rwsem);
1963 init_waitqueue_head(&sbi->cp_wait);
1964 init_sb_info(sbi);
1965
1966 err = init_percpu_info(sbi);
1967 if (err)
1968 goto free_options;
1969
1970 if (F2FS_IO_SIZE(sbi) > 1) {
1971 sbi->write_io_dummy =
1972 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
1973 if (!sbi->write_io_dummy)
1974 goto free_options;
1975 }
1976
1977 /* get an inode for meta space */
1978 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1979 if (IS_ERR(sbi->meta_inode)) {
1980 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1981 err = PTR_ERR(sbi->meta_inode);
1982 goto free_io_dummy;
1983 }
1984
1985 err = get_valid_checkpoint(sbi);
1986 if (err) {
1987 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1988 goto free_meta_inode;
1989 }
1990
1991 /* Initialize device list */
1992 err = f2fs_scan_devices(sbi);
1993 if (err) {
1994 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
1995 goto free_devices;
1996 }
1997
1998 sbi->total_valid_node_count =
1999 le32_to_cpu(sbi->ckpt->valid_node_count);
2000 percpu_counter_set(&sbi->total_valid_inode_count,
2001 le32_to_cpu(sbi->ckpt->valid_inode_count));
2002 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2003 sbi->total_valid_block_count =
2004 le64_to_cpu(sbi->ckpt->valid_block_count);
2005 sbi->last_valid_block_count = sbi->total_valid_block_count;
2006
2007 for (i = 0; i < NR_INODE_TYPE; i++) {
2008 INIT_LIST_HEAD(&sbi->inode_list[i]);
2009 spin_lock_init(&sbi->inode_lock[i]);
2010 }
2011
2012 init_extent_cache_info(sbi);
2013
2014 init_ino_entry_info(sbi);
2015
2016 /* setup f2fs internal modules */
2017 err = build_segment_manager(sbi);
2018 if (err) {
2019 f2fs_msg(sb, KERN_ERR,
2020 "Failed to initialize F2FS segment manager");
2021 goto free_sm;
2022 }
2023 err = build_node_manager(sbi);
2024 if (err) {
2025 f2fs_msg(sb, KERN_ERR,
2026 "Failed to initialize F2FS node manager");
2027 goto free_nm;
2028 }
2029
2030 /* For write statistics */
2031 if (sb->s_bdev->bd_part)
2032 sbi->sectors_written_start =
2033 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2034
2035 /* Read accumulated write IO statistics if exists */
2036 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2037 if (__exist_node_summaries(sbi))
2038 sbi->kbytes_written =
2039 le64_to_cpu(seg_i->journal->info.kbytes_written);
2040
2041 build_gc_manager(sbi);
2042
2043 /* get an inode for node space */
2044 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2045 if (IS_ERR(sbi->node_inode)) {
2046 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2047 err = PTR_ERR(sbi->node_inode);
2048 goto free_nm;
2049 }
2050
2051 f2fs_join_shrinker(sbi);
2052
2053 err = f2fs_build_stats(sbi);
2054 if (err)
2055 goto free_nm;
2056
2057 /* if there are nt orphan nodes free them */
2058 err = recover_orphan_inodes(sbi);
2059 if (err)
2060 goto free_node_inode;
2061
2062 /* read root inode and dentry */
2063 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2064 if (IS_ERR(root)) {
2065 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2066 err = PTR_ERR(root);
2067 goto free_node_inode;
2068 }
2069 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2070 iput(root);
2071 err = -EINVAL;
2072 goto free_node_inode;
2073 }
2074
2075 sb->s_root = d_make_root(root); /* allocate root dentry */
2076 if (!sb->s_root) {
2077 err = -ENOMEM;
2078 goto free_root_inode;
2079 }
2080
2081 if (f2fs_proc_root)
2082 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
2083
2084 if (sbi->s_proc) {
2085 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
2086 &f2fs_seq_segment_info_fops, sb);
2087 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
2088 &f2fs_seq_segment_bits_fops, sb);
2089 }
2090
2091 sbi->s_kobj.kset = f2fs_kset;
2092 init_completion(&sbi->s_kobj_unregister);
2093 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
2094 "%s", sb->s_id);
2095 if (err)
2096 goto free_proc;
2097
2098 /* recover fsynced data */
2099 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2100 /*
2101 * mount should be failed, when device has readonly mode, and
2102 * previous checkpoint was not done by clean system shutdown.
2103 */
2104 if (bdev_read_only(sb->s_bdev) &&
2105 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2106 err = -EROFS;
2107 goto free_kobj;
2108 }
2109
2110 if (need_fsck)
2111 set_sbi_flag(sbi, SBI_NEED_FSCK);
2112
2113 if (!retry)
2114 goto skip_recovery;
2115
2116 err = recover_fsync_data(sbi, false);
2117 if (err < 0) {
2118 need_fsck = true;
2119 f2fs_msg(sb, KERN_ERR,
2120 "Cannot recover all fsync data errno=%d", err);
2121 goto free_kobj;
2122 }
2123 } else {
2124 err = recover_fsync_data(sbi, true);
2125
2126 if (!f2fs_readonly(sb) && err > 0) {
2127 err = -EINVAL;
2128 f2fs_msg(sb, KERN_ERR,
2129 "Need to recover fsync data");
2130 goto free_kobj;
2131 }
2132 }
2133 skip_recovery:
2134 /* recover_fsync_data() cleared this already */
2135 clear_sbi_flag(sbi, SBI_POR_DOING);
2136
2137 /*
2138 * If filesystem is not mounted as read-only then
2139 * do start the gc_thread.
2140 */
2141 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2142 /* After POR, we can run background GC thread.*/
2143 err = start_gc_thread(sbi);
2144 if (err)
2145 goto free_kobj;
2146 }
2147 kfree(options);
2148
2149 /* recover broken superblock */
2150 if (recovery) {
2151 err = f2fs_commit_super(sbi, true);
2152 f2fs_msg(sb, KERN_INFO,
2153 "Try to recover %dth superblock, ret: %d",
2154 sbi->valid_super_block ? 1 : 2, err);
2155 }
2156
2157 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2158 cur_cp_version(F2FS_CKPT(sbi)));
2159 f2fs_update_time(sbi, CP_TIME);
2160 f2fs_update_time(sbi, REQ_TIME);
2161 return 0;
2162
2163 free_kobj:
2164 f2fs_sync_inode_meta(sbi);
2165 kobject_del(&sbi->s_kobj);
2166 kobject_put(&sbi->s_kobj);
2167 wait_for_completion(&sbi->s_kobj_unregister);
2168 free_proc:
2169 if (sbi->s_proc) {
2170 remove_proc_entry("segment_info", sbi->s_proc);
2171 remove_proc_entry("segment_bits", sbi->s_proc);
2172 remove_proc_entry(sb->s_id, f2fs_proc_root);
2173 }
2174 free_root_inode:
2175 dput(sb->s_root);
2176 sb->s_root = NULL;
2177 free_node_inode:
2178 truncate_inode_pages_final(NODE_MAPPING(sbi));
2179 mutex_lock(&sbi->umount_mutex);
2180 release_ino_entry(sbi, true);
2181 f2fs_leave_shrinker(sbi);
2182 /*
2183 * Some dirty meta pages can be produced by recover_orphan_inodes()
2184 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2185 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2186 * falls into an infinite loop in sync_meta_pages().
2187 */
2188 truncate_inode_pages_final(META_MAPPING(sbi));
2189 iput(sbi->node_inode);
2190 mutex_unlock(&sbi->umount_mutex);
2191 f2fs_destroy_stats(sbi);
2192 free_nm:
2193 destroy_node_manager(sbi);
2194 free_sm:
2195 destroy_segment_manager(sbi);
2196 free_devices:
2197 destroy_device_list(sbi);
2198 kfree(sbi->ckpt);
2199 free_meta_inode:
2200 make_bad_inode(sbi->meta_inode);
2201 iput(sbi->meta_inode);
2202 free_io_dummy:
2203 mempool_destroy(sbi->write_io_dummy);
2204 free_options:
2205 destroy_percpu_info(sbi);
2206 kfree(options);
2207 free_sb_buf:
2208 kfree(raw_super);
2209 free_sbi:
2210 if (sbi->s_chksum_driver)
2211 crypto_free_shash(sbi->s_chksum_driver);
2212 kfree(sbi);
2213
2214 /* give only one another chance */
2215 if (retry) {
2216 retry = false;
2217 shrink_dcache_sb(sb);
2218 goto try_onemore;
2219 }
2220 return err;
2221 }
2222
2223 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2224 const char *dev_name, void *data)
2225 {
2226 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2227 }
2228
2229 static void kill_f2fs_super(struct super_block *sb)
2230 {
2231 if (sb->s_root)
2232 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2233 kill_block_super(sb);
2234 }
2235
2236 static struct file_system_type f2fs_fs_type = {
2237 .owner = THIS_MODULE,
2238 .name = "f2fs",
2239 .mount = f2fs_mount,
2240 .kill_sb = kill_f2fs_super,
2241 .fs_flags = FS_REQUIRES_DEV,
2242 };
2243 MODULE_ALIAS_FS("f2fs");
2244
2245 static int __init init_inodecache(void)
2246 {
2247 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2248 sizeof(struct f2fs_inode_info), 0,
2249 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2250 if (!f2fs_inode_cachep)
2251 return -ENOMEM;
2252 return 0;
2253 }
2254
2255 static void destroy_inodecache(void)
2256 {
2257 /*
2258 * Make sure all delayed rcu free inodes are flushed before we
2259 * destroy cache.
2260 */
2261 rcu_barrier();
2262 kmem_cache_destroy(f2fs_inode_cachep);
2263 }
2264
2265 static int __init init_f2fs_fs(void)
2266 {
2267 int err;
2268
2269 f2fs_build_trace_ios();
2270
2271 err = init_inodecache();
2272 if (err)
2273 goto fail;
2274 err = create_node_manager_caches();
2275 if (err)
2276 goto free_inodecache;
2277 err = create_segment_manager_caches();
2278 if (err)
2279 goto free_node_manager_caches;
2280 err = create_checkpoint_caches();
2281 if (err)
2282 goto free_segment_manager_caches;
2283 err = create_extent_cache();
2284 if (err)
2285 goto free_checkpoint_caches;
2286 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
2287 if (!f2fs_kset) {
2288 err = -ENOMEM;
2289 goto free_extent_cache;
2290 }
2291 err = register_shrinker(&f2fs_shrinker_info);
2292 if (err)
2293 goto free_kset;
2294
2295 err = register_filesystem(&f2fs_fs_type);
2296 if (err)
2297 goto free_shrinker;
2298 err = f2fs_create_root_stats();
2299 if (err)
2300 goto free_filesystem;
2301 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2302 return 0;
2303
2304 free_filesystem:
2305 unregister_filesystem(&f2fs_fs_type);
2306 free_shrinker:
2307 unregister_shrinker(&f2fs_shrinker_info);
2308 free_kset:
2309 kset_unregister(f2fs_kset);
2310 free_extent_cache:
2311 destroy_extent_cache();
2312 free_checkpoint_caches:
2313 destroy_checkpoint_caches();
2314 free_segment_manager_caches:
2315 destroy_segment_manager_caches();
2316 free_node_manager_caches:
2317 destroy_node_manager_caches();
2318 free_inodecache:
2319 destroy_inodecache();
2320 fail:
2321 return err;
2322 }
2323
2324 static void __exit exit_f2fs_fs(void)
2325 {
2326 remove_proc_entry("fs/f2fs", NULL);
2327 f2fs_destroy_root_stats();
2328 unregister_filesystem(&f2fs_fs_type);
2329 unregister_shrinker(&f2fs_shrinker_info);
2330 kset_unregister(f2fs_kset);
2331 destroy_extent_cache();
2332 destroy_checkpoint_caches();
2333 destroy_segment_manager_caches();
2334 destroy_node_manager_caches();
2335 destroy_inodecache();
2336 f2fs_destroy_trace_ios();
2337 }
2338
2339 module_init(init_f2fs_fs)
2340 module_exit(exit_f2fs_fs)
2341
2342 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2343 MODULE_DESCRIPTION("Flash Friendly File System");
2344 MODULE_LICENSE("GPL");
2345