1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/super.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <linux/module.h>
9 #include <linux/init.h>
10 #include <linux/fs.h>
11 #include <linux/fs_context.h>
12 #include <linux/sched/mm.h>
13 #include <linux/statfs.h>
14 #include <linux/buffer_head.h>
15 #include <linux/kthread.h>
16 #include <linux/parser.h>
17 #include <linux/mount.h>
18 #include <linux/seq_file.h>
19 #include <linux/proc_fs.h>
20 #include <linux/random.h>
21 #include <linux/exportfs.h>
22 #include <linux/blkdev.h>
23 #include <linux/quotaops.h>
24 #include <linux/f2fs_fs.h>
25 #include <linux/sysfs.h>
26 #include <linux/quota.h>
27 #include <linux/unicode.h>
28 #include <linux/part_stat.h>
29 #include <linux/zstd.h>
30 #include <linux/lz4.h>
31 
32 #include "f2fs.h"
33 #include "node.h"
34 #include "segment.h"
35 #include "xattr.h"
36 #include "gc.h"
37 #include "iostat.h"
38 
39 #define CREATE_TRACE_POINTS
40 #include <trace/events/f2fs.h>
41 
42 static struct kmem_cache *f2fs_inode_cachep;
43 
44 #ifdef CONFIG_F2FS_FAULT_INJECTION
45 
46 const char *f2fs_fault_name[FAULT_MAX] = {
47 	[FAULT_KMALLOC]		= "kmalloc",
48 	[FAULT_KVMALLOC]	= "kvmalloc",
49 	[FAULT_PAGE_ALLOC]	= "page alloc",
50 	[FAULT_PAGE_GET]	= "page get",
51 	[FAULT_ALLOC_NID]	= "alloc nid",
52 	[FAULT_ORPHAN]		= "orphan",
53 	[FAULT_BLOCK]		= "no more block",
54 	[FAULT_DIR_DEPTH]	= "too big dir depth",
55 	[FAULT_EVICT_INODE]	= "evict_inode fail",
56 	[FAULT_TRUNCATE]	= "truncate fail",
57 	[FAULT_READ_IO]		= "read IO error",
58 	[FAULT_CHECKPOINT]	= "checkpoint error",
59 	[FAULT_DISCARD]		= "discard error",
60 	[FAULT_WRITE_IO]	= "write IO error",
61 	[FAULT_SLAB_ALLOC]	= "slab alloc",
62 	[FAULT_DQUOT_INIT]	= "dquot initialize",
63 	[FAULT_LOCK_OP]		= "lock_op",
64 };
65 
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned int rate,unsigned int type)66 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
67 							unsigned int type)
68 {
69 	struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
70 
71 	if (rate) {
72 		atomic_set(&ffi->inject_ops, 0);
73 		ffi->inject_rate = rate;
74 	}
75 
76 	if (type)
77 		ffi->inject_type = type;
78 
79 	if (!rate && !type)
80 		memset(ffi, 0, sizeof(struct f2fs_fault_info));
81 }
82 #endif
83 
84 /* f2fs-wide shrinker description */
85 static struct shrinker f2fs_shrinker_info = {
86 	.scan_objects = f2fs_shrink_scan,
87 	.count_objects = f2fs_shrink_count,
88 	.seeks = DEFAULT_SEEKS,
89 };
90 
91 enum {
92 	Opt_gc_background,
93 	Opt_disable_roll_forward,
94 	Opt_norecovery,
95 	Opt_discard,
96 	Opt_nodiscard,
97 	Opt_noheap,
98 	Opt_heap,
99 	Opt_user_xattr,
100 	Opt_nouser_xattr,
101 	Opt_acl,
102 	Opt_noacl,
103 	Opt_active_logs,
104 	Opt_disable_ext_identify,
105 	Opt_inline_xattr,
106 	Opt_noinline_xattr,
107 	Opt_inline_xattr_size,
108 	Opt_inline_data,
109 	Opt_inline_dentry,
110 	Opt_noinline_dentry,
111 	Opt_flush_merge,
112 	Opt_noflush_merge,
113 	Opt_nobarrier,
114 	Opt_fastboot,
115 	Opt_extent_cache,
116 	Opt_noextent_cache,
117 	Opt_noinline_data,
118 	Opt_data_flush,
119 	Opt_reserve_root,
120 	Opt_resgid,
121 	Opt_resuid,
122 	Opt_mode,
123 	Opt_io_size_bits,
124 	Opt_fault_injection,
125 	Opt_fault_type,
126 	Opt_lazytime,
127 	Opt_nolazytime,
128 	Opt_quota,
129 	Opt_noquota,
130 	Opt_usrquota,
131 	Opt_grpquota,
132 	Opt_prjquota,
133 	Opt_usrjquota,
134 	Opt_grpjquota,
135 	Opt_prjjquota,
136 	Opt_offusrjquota,
137 	Opt_offgrpjquota,
138 	Opt_offprjjquota,
139 	Opt_jqfmt_vfsold,
140 	Opt_jqfmt_vfsv0,
141 	Opt_jqfmt_vfsv1,
142 	Opt_alloc,
143 	Opt_fsync,
144 	Opt_test_dummy_encryption,
145 	Opt_inlinecrypt,
146 	Opt_checkpoint_disable,
147 	Opt_checkpoint_disable_cap,
148 	Opt_checkpoint_disable_cap_perc,
149 	Opt_checkpoint_enable,
150 	Opt_checkpoint_merge,
151 	Opt_nocheckpoint_merge,
152 	Opt_compress_algorithm,
153 	Opt_compress_log_size,
154 	Opt_compress_extension,
155 	Opt_nocompress_extension,
156 	Opt_compress_chksum,
157 	Opt_compress_mode,
158 	Opt_compress_cache,
159 	Opt_atgc,
160 	Opt_gc_merge,
161 	Opt_nogc_merge,
162 	Opt_discard_unit,
163 	Opt_memory_mode,
164 	Opt_err,
165 };
166 
167 static match_table_t f2fs_tokens = {
168 	{Opt_gc_background, "background_gc=%s"},
169 	{Opt_disable_roll_forward, "disable_roll_forward"},
170 	{Opt_norecovery, "norecovery"},
171 	{Opt_discard, "discard"},
172 	{Opt_nodiscard, "nodiscard"},
173 	{Opt_noheap, "no_heap"},
174 	{Opt_heap, "heap"},
175 	{Opt_user_xattr, "user_xattr"},
176 	{Opt_nouser_xattr, "nouser_xattr"},
177 	{Opt_acl, "acl"},
178 	{Opt_noacl, "noacl"},
179 	{Opt_active_logs, "active_logs=%u"},
180 	{Opt_disable_ext_identify, "disable_ext_identify"},
181 	{Opt_inline_xattr, "inline_xattr"},
182 	{Opt_noinline_xattr, "noinline_xattr"},
183 	{Opt_inline_xattr_size, "inline_xattr_size=%u"},
184 	{Opt_inline_data, "inline_data"},
185 	{Opt_inline_dentry, "inline_dentry"},
186 	{Opt_noinline_dentry, "noinline_dentry"},
187 	{Opt_flush_merge, "flush_merge"},
188 	{Opt_noflush_merge, "noflush_merge"},
189 	{Opt_nobarrier, "nobarrier"},
190 	{Opt_fastboot, "fastboot"},
191 	{Opt_extent_cache, "extent_cache"},
192 	{Opt_noextent_cache, "noextent_cache"},
193 	{Opt_noinline_data, "noinline_data"},
194 	{Opt_data_flush, "data_flush"},
195 	{Opt_reserve_root, "reserve_root=%u"},
196 	{Opt_resgid, "resgid=%u"},
197 	{Opt_resuid, "resuid=%u"},
198 	{Opt_mode, "mode=%s"},
199 	{Opt_io_size_bits, "io_bits=%u"},
200 	{Opt_fault_injection, "fault_injection=%u"},
201 	{Opt_fault_type, "fault_type=%u"},
202 	{Opt_lazytime, "lazytime"},
203 	{Opt_nolazytime, "nolazytime"},
204 	{Opt_quota, "quota"},
205 	{Opt_noquota, "noquota"},
206 	{Opt_usrquota, "usrquota"},
207 	{Opt_grpquota, "grpquota"},
208 	{Opt_prjquota, "prjquota"},
209 	{Opt_usrjquota, "usrjquota=%s"},
210 	{Opt_grpjquota, "grpjquota=%s"},
211 	{Opt_prjjquota, "prjjquota=%s"},
212 	{Opt_offusrjquota, "usrjquota="},
213 	{Opt_offgrpjquota, "grpjquota="},
214 	{Opt_offprjjquota, "prjjquota="},
215 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
216 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
217 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
218 	{Opt_alloc, "alloc_mode=%s"},
219 	{Opt_fsync, "fsync_mode=%s"},
220 	{Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
221 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
222 	{Opt_inlinecrypt, "inlinecrypt"},
223 	{Opt_checkpoint_disable, "checkpoint=disable"},
224 	{Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
225 	{Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
226 	{Opt_checkpoint_enable, "checkpoint=enable"},
227 	{Opt_checkpoint_merge, "checkpoint_merge"},
228 	{Opt_nocheckpoint_merge, "nocheckpoint_merge"},
229 	{Opt_compress_algorithm, "compress_algorithm=%s"},
230 	{Opt_compress_log_size, "compress_log_size=%u"},
231 	{Opt_compress_extension, "compress_extension=%s"},
232 	{Opt_nocompress_extension, "nocompress_extension=%s"},
233 	{Opt_compress_chksum, "compress_chksum"},
234 	{Opt_compress_mode, "compress_mode=%s"},
235 	{Opt_compress_cache, "compress_cache"},
236 	{Opt_atgc, "atgc"},
237 	{Opt_gc_merge, "gc_merge"},
238 	{Opt_nogc_merge, "nogc_merge"},
239 	{Opt_discard_unit, "discard_unit=%s"},
240 	{Opt_memory_mode, "memory=%s"},
241 	{Opt_err, NULL},
242 };
243 
f2fs_printk(struct f2fs_sb_info * sbi,const char * fmt,...)244 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
245 {
246 	struct va_format vaf;
247 	va_list args;
248 	int level;
249 
250 	va_start(args, fmt);
251 
252 	level = printk_get_level(fmt);
253 	vaf.fmt = printk_skip_level(fmt);
254 	vaf.va = &args;
255 	printk("%c%cF2FS-fs (%s): %pV\n",
256 	       KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
257 
258 	va_end(args);
259 }
260 
261 #if IS_ENABLED(CONFIG_UNICODE)
262 static const struct f2fs_sb_encodings {
263 	__u16 magic;
264 	char *name;
265 	unsigned int version;
266 } f2fs_sb_encoding_map[] = {
267 	{F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
268 };
269 
270 static const struct f2fs_sb_encodings *
f2fs_sb_read_encoding(const struct f2fs_super_block * sb)271 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
272 {
273 	__u16 magic = le16_to_cpu(sb->s_encoding);
274 	int i;
275 
276 	for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
277 		if (magic == f2fs_sb_encoding_map[i].magic)
278 			return &f2fs_sb_encoding_map[i];
279 
280 	return NULL;
281 }
282 
283 struct kmem_cache *f2fs_cf_name_slab;
f2fs_create_casefold_cache(void)284 static int __init f2fs_create_casefold_cache(void)
285 {
286 	f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
287 							F2FS_NAME_LEN);
288 	if (!f2fs_cf_name_slab)
289 		return -ENOMEM;
290 	return 0;
291 }
292 
f2fs_destroy_casefold_cache(void)293 static void f2fs_destroy_casefold_cache(void)
294 {
295 	kmem_cache_destroy(f2fs_cf_name_slab);
296 }
297 #else
f2fs_create_casefold_cache(void)298 static int __init f2fs_create_casefold_cache(void) { return 0; }
f2fs_destroy_casefold_cache(void)299 static void f2fs_destroy_casefold_cache(void) { }
300 #endif
301 
limit_reserve_root(struct f2fs_sb_info * sbi)302 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
303 {
304 	block_t limit = min((sbi->user_block_count >> 3),
305 			sbi->user_block_count - sbi->reserved_blocks);
306 
307 	/* limit is 12.5% */
308 	if (test_opt(sbi, RESERVE_ROOT) &&
309 			F2FS_OPTION(sbi).root_reserved_blocks > limit) {
310 		F2FS_OPTION(sbi).root_reserved_blocks = limit;
311 		f2fs_info(sbi, "Reduce reserved blocks for root = %u",
312 			  F2FS_OPTION(sbi).root_reserved_blocks);
313 	}
314 	if (!test_opt(sbi, RESERVE_ROOT) &&
315 		(!uid_eq(F2FS_OPTION(sbi).s_resuid,
316 				make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
317 		!gid_eq(F2FS_OPTION(sbi).s_resgid,
318 				make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
319 		f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
320 			  from_kuid_munged(&init_user_ns,
321 					   F2FS_OPTION(sbi).s_resuid),
322 			  from_kgid_munged(&init_user_ns,
323 					   F2FS_OPTION(sbi).s_resgid));
324 }
325 
adjust_reserved_segment(struct f2fs_sb_info * sbi)326 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
327 {
328 	unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
329 	unsigned int avg_vblocks;
330 	unsigned int wanted_reserved_segments;
331 	block_t avail_user_block_count;
332 
333 	if (!F2FS_IO_ALIGNED(sbi))
334 		return 0;
335 
336 	/* average valid block count in section in worst case */
337 	avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
338 
339 	/*
340 	 * we need enough free space when migrating one section in worst case
341 	 */
342 	wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
343 						reserved_segments(sbi);
344 	wanted_reserved_segments -= reserved_segments(sbi);
345 
346 	avail_user_block_count = sbi->user_block_count -
347 				sbi->current_reserved_blocks -
348 				F2FS_OPTION(sbi).root_reserved_blocks;
349 
350 	if (wanted_reserved_segments * sbi->blocks_per_seg >
351 					avail_user_block_count) {
352 		f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
353 			wanted_reserved_segments,
354 			avail_user_block_count >> sbi->log_blocks_per_seg);
355 		return -ENOSPC;
356 	}
357 
358 	SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
359 
360 	f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
361 			 wanted_reserved_segments);
362 
363 	return 0;
364 }
365 
adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)366 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
367 {
368 	if (!F2FS_OPTION(sbi).unusable_cap_perc)
369 		return;
370 
371 	if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
372 		F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
373 	else
374 		F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
375 					F2FS_OPTION(sbi).unusable_cap_perc;
376 
377 	f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
378 			F2FS_OPTION(sbi).unusable_cap,
379 			F2FS_OPTION(sbi).unusable_cap_perc);
380 }
381 
init_once(void * foo)382 static void init_once(void *foo)
383 {
384 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
385 
386 	inode_init_once(&fi->vfs_inode);
387 }
388 
389 #ifdef CONFIG_QUOTA
390 static const char * const quotatypes[] = INITQFNAMES;
391 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)392 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
393 							substring_t *args)
394 {
395 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
396 	char *qname;
397 	int ret = -EINVAL;
398 
399 	if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
400 		f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
401 		return -EINVAL;
402 	}
403 	if (f2fs_sb_has_quota_ino(sbi)) {
404 		f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
405 		return 0;
406 	}
407 
408 	qname = match_strdup(args);
409 	if (!qname) {
410 		f2fs_err(sbi, "Not enough memory for storing quotafile name");
411 		return -ENOMEM;
412 	}
413 	if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
414 		if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
415 			ret = 0;
416 		else
417 			f2fs_err(sbi, "%s quota file already specified",
418 				 QTYPE2NAME(qtype));
419 		goto errout;
420 	}
421 	if (strchr(qname, '/')) {
422 		f2fs_err(sbi, "quotafile must be on filesystem root");
423 		goto errout;
424 	}
425 	F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
426 	set_opt(sbi, QUOTA);
427 	return 0;
428 errout:
429 	kfree(qname);
430 	return ret;
431 }
432 
f2fs_clear_qf_name(struct super_block * sb,int qtype)433 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
434 {
435 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
436 
437 	if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
438 		f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
439 		return -EINVAL;
440 	}
441 	kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
442 	F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
443 	return 0;
444 }
445 
f2fs_check_quota_options(struct f2fs_sb_info * sbi)446 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
447 {
448 	/*
449 	 * We do the test below only for project quotas. 'usrquota' and
450 	 * 'grpquota' mount options are allowed even without quota feature
451 	 * to support legacy quotas in quota files.
452 	 */
453 	if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
454 		f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
455 		return -1;
456 	}
457 	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
458 			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
459 			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
460 		if (test_opt(sbi, USRQUOTA) &&
461 				F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
462 			clear_opt(sbi, USRQUOTA);
463 
464 		if (test_opt(sbi, GRPQUOTA) &&
465 				F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
466 			clear_opt(sbi, GRPQUOTA);
467 
468 		if (test_opt(sbi, PRJQUOTA) &&
469 				F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
470 			clear_opt(sbi, PRJQUOTA);
471 
472 		if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
473 				test_opt(sbi, PRJQUOTA)) {
474 			f2fs_err(sbi, "old and new quota format mixing");
475 			return -1;
476 		}
477 
478 		if (!F2FS_OPTION(sbi).s_jquota_fmt) {
479 			f2fs_err(sbi, "journaled quota format not specified");
480 			return -1;
481 		}
482 	}
483 
484 	if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
485 		f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
486 		F2FS_OPTION(sbi).s_jquota_fmt = 0;
487 	}
488 	return 0;
489 }
490 #endif
491 
f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)492 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
493 					  const char *opt,
494 					  const substring_t *arg,
495 					  bool is_remount)
496 {
497 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
498 	struct fs_parameter param = {
499 		.type = fs_value_is_string,
500 		.string = arg->from ? arg->from : "",
501 	};
502 	struct fscrypt_dummy_policy *policy =
503 		&F2FS_OPTION(sbi).dummy_enc_policy;
504 	int err;
505 
506 	if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
507 		f2fs_warn(sbi, "test_dummy_encryption option not supported");
508 		return -EINVAL;
509 	}
510 
511 	if (!f2fs_sb_has_encrypt(sbi)) {
512 		f2fs_err(sbi, "Encrypt feature is off");
513 		return -EINVAL;
514 	}
515 
516 	/*
517 	 * This mount option is just for testing, and it's not worthwhile to
518 	 * implement the extra complexity (e.g. RCU protection) that would be
519 	 * needed to allow it to be set or changed during remount.  We do allow
520 	 * it to be specified during remount, but only if there is no change.
521 	 */
522 	if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
523 		f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
524 		return -EINVAL;
525 	}
526 
527 	err = fscrypt_parse_test_dummy_encryption(&param, policy);
528 	if (err) {
529 		if (err == -EEXIST)
530 			f2fs_warn(sbi,
531 				  "Can't change test_dummy_encryption on remount");
532 		else if (err == -EINVAL)
533 			f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
534 				  opt);
535 		else
536 			f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
537 				  opt, err);
538 		return -EINVAL;
539 	}
540 	err = fscrypt_add_test_dummy_key(sb, policy);
541 	if (err) {
542 		f2fs_warn(sbi, "Error adding test dummy encryption key [%d]",
543 			  err);
544 		return err;
545 	}
546 	f2fs_warn(sbi, "Test dummy encryption mode enabled");
547 	return 0;
548 }
549 
550 #ifdef CONFIG_F2FS_FS_COMPRESSION
551 /*
552  * 1. The same extension name cannot not appear in both compress and non-compress extension
553  * at the same time.
554  * 2. If the compress extension specifies all files, the types specified by the non-compress
555  * extension will be treated as special cases and will not be compressed.
556  * 3. Don't allow the non-compress extension specifies all files.
557  */
f2fs_test_compress_extension(struct f2fs_sb_info * sbi)558 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
559 {
560 	unsigned char (*ext)[F2FS_EXTENSION_LEN];
561 	unsigned char (*noext)[F2FS_EXTENSION_LEN];
562 	int ext_cnt, noext_cnt, index = 0, no_index = 0;
563 
564 	ext = F2FS_OPTION(sbi).extensions;
565 	ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
566 	noext = F2FS_OPTION(sbi).noextensions;
567 	noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
568 
569 	if (!noext_cnt)
570 		return 0;
571 
572 	for (no_index = 0; no_index < noext_cnt; no_index++) {
573 		if (!strcasecmp("*", noext[no_index])) {
574 			f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
575 			return -EINVAL;
576 		}
577 		for (index = 0; index < ext_cnt; index++) {
578 			if (!strcasecmp(ext[index], noext[no_index])) {
579 				f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
580 						ext[index]);
581 				return -EINVAL;
582 			}
583 		}
584 	}
585 	return 0;
586 }
587 
588 #ifdef CONFIG_F2FS_FS_LZ4
f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)589 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
590 {
591 #ifdef CONFIG_F2FS_FS_LZ4HC
592 	unsigned int level;
593 #endif
594 
595 	if (strlen(str) == 3) {
596 		F2FS_OPTION(sbi).compress_level = 0;
597 		return 0;
598 	}
599 
600 #ifdef CONFIG_F2FS_FS_LZ4HC
601 	str += 3;
602 
603 	if (str[0] != ':') {
604 		f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
605 		return -EINVAL;
606 	}
607 	if (kstrtouint(str + 1, 10, &level))
608 		return -EINVAL;
609 
610 	if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
611 		f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
612 		return -EINVAL;
613 	}
614 
615 	F2FS_OPTION(sbi).compress_level = level;
616 	return 0;
617 #else
618 	f2fs_info(sbi, "kernel doesn't support lz4hc compression");
619 	return -EINVAL;
620 #endif
621 }
622 #endif
623 
624 #ifdef CONFIG_F2FS_FS_ZSTD
f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)625 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
626 {
627 	unsigned int level;
628 	int len = 4;
629 
630 	if (strlen(str) == len) {
631 		F2FS_OPTION(sbi).compress_level = 0;
632 		return 0;
633 	}
634 
635 	str += len;
636 
637 	if (str[0] != ':') {
638 		f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
639 		return -EINVAL;
640 	}
641 	if (kstrtouint(str + 1, 10, &level))
642 		return -EINVAL;
643 
644 	if (!level || level > zstd_max_clevel()) {
645 		f2fs_info(sbi, "invalid zstd compress level: %d", level);
646 		return -EINVAL;
647 	}
648 
649 	F2FS_OPTION(sbi).compress_level = level;
650 	return 0;
651 }
652 #endif
653 #endif
654 
parse_options(struct super_block * sb,char * options,bool is_remount)655 static int parse_options(struct super_block *sb, char *options, bool is_remount)
656 {
657 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
658 	substring_t args[MAX_OPT_ARGS];
659 #ifdef CONFIG_F2FS_FS_COMPRESSION
660 	unsigned char (*ext)[F2FS_EXTENSION_LEN];
661 	unsigned char (*noext)[F2FS_EXTENSION_LEN];
662 	int ext_cnt, noext_cnt;
663 #endif
664 	char *p, *name;
665 	int arg = 0;
666 	kuid_t uid;
667 	kgid_t gid;
668 	int ret;
669 
670 	if (!options)
671 		goto default_check;
672 
673 	while ((p = strsep(&options, ",")) != NULL) {
674 		int token;
675 
676 		if (!*p)
677 			continue;
678 		/*
679 		 * Initialize args struct so we know whether arg was
680 		 * found; some options take optional arguments.
681 		 */
682 		args[0].to = args[0].from = NULL;
683 		token = match_token(p, f2fs_tokens, args);
684 
685 		switch (token) {
686 		case Opt_gc_background:
687 			name = match_strdup(&args[0]);
688 
689 			if (!name)
690 				return -ENOMEM;
691 			if (!strcmp(name, "on")) {
692 				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
693 			} else if (!strcmp(name, "off")) {
694 				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
695 			} else if (!strcmp(name, "sync")) {
696 				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
697 			} else {
698 				kfree(name);
699 				return -EINVAL;
700 			}
701 			kfree(name);
702 			break;
703 		case Opt_disable_roll_forward:
704 			set_opt(sbi, DISABLE_ROLL_FORWARD);
705 			break;
706 		case Opt_norecovery:
707 			/* this option mounts f2fs with ro */
708 			set_opt(sbi, NORECOVERY);
709 			if (!f2fs_readonly(sb))
710 				return -EINVAL;
711 			break;
712 		case Opt_discard:
713 			if (!f2fs_hw_support_discard(sbi)) {
714 				f2fs_warn(sbi, "device does not support discard");
715 				break;
716 			}
717 			set_opt(sbi, DISCARD);
718 			break;
719 		case Opt_nodiscard:
720 			if (f2fs_hw_should_discard(sbi)) {
721 				f2fs_warn(sbi, "discard is required for zoned block devices");
722 				return -EINVAL;
723 			}
724 			clear_opt(sbi, DISCARD);
725 			break;
726 		case Opt_noheap:
727 			set_opt(sbi, NOHEAP);
728 			break;
729 		case Opt_heap:
730 			clear_opt(sbi, NOHEAP);
731 			break;
732 #ifdef CONFIG_F2FS_FS_XATTR
733 		case Opt_user_xattr:
734 			set_opt(sbi, XATTR_USER);
735 			break;
736 		case Opt_nouser_xattr:
737 			clear_opt(sbi, XATTR_USER);
738 			break;
739 		case Opt_inline_xattr:
740 			set_opt(sbi, INLINE_XATTR);
741 			break;
742 		case Opt_noinline_xattr:
743 			clear_opt(sbi, INLINE_XATTR);
744 			break;
745 		case Opt_inline_xattr_size:
746 			if (args->from && match_int(args, &arg))
747 				return -EINVAL;
748 			set_opt(sbi, INLINE_XATTR_SIZE);
749 			F2FS_OPTION(sbi).inline_xattr_size = arg;
750 			break;
751 #else
752 		case Opt_user_xattr:
753 			f2fs_info(sbi, "user_xattr options not supported");
754 			break;
755 		case Opt_nouser_xattr:
756 			f2fs_info(sbi, "nouser_xattr options not supported");
757 			break;
758 		case Opt_inline_xattr:
759 			f2fs_info(sbi, "inline_xattr options not supported");
760 			break;
761 		case Opt_noinline_xattr:
762 			f2fs_info(sbi, "noinline_xattr options not supported");
763 			break;
764 #endif
765 #ifdef CONFIG_F2FS_FS_POSIX_ACL
766 		case Opt_acl:
767 			set_opt(sbi, POSIX_ACL);
768 			break;
769 		case Opt_noacl:
770 			clear_opt(sbi, POSIX_ACL);
771 			break;
772 #else
773 		case Opt_acl:
774 			f2fs_info(sbi, "acl options not supported");
775 			break;
776 		case Opt_noacl:
777 			f2fs_info(sbi, "noacl options not supported");
778 			break;
779 #endif
780 		case Opt_active_logs:
781 			if (args->from && match_int(args, &arg))
782 				return -EINVAL;
783 			if (arg != 2 && arg != 4 &&
784 				arg != NR_CURSEG_PERSIST_TYPE)
785 				return -EINVAL;
786 			F2FS_OPTION(sbi).active_logs = arg;
787 			break;
788 		case Opt_disable_ext_identify:
789 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
790 			break;
791 		case Opt_inline_data:
792 			set_opt(sbi, INLINE_DATA);
793 			break;
794 		case Opt_inline_dentry:
795 			set_opt(sbi, INLINE_DENTRY);
796 			break;
797 		case Opt_noinline_dentry:
798 			clear_opt(sbi, INLINE_DENTRY);
799 			break;
800 		case Opt_flush_merge:
801 			set_opt(sbi, FLUSH_MERGE);
802 			break;
803 		case Opt_noflush_merge:
804 			clear_opt(sbi, FLUSH_MERGE);
805 			break;
806 		case Opt_nobarrier:
807 			set_opt(sbi, NOBARRIER);
808 			break;
809 		case Opt_fastboot:
810 			set_opt(sbi, FASTBOOT);
811 			break;
812 		case Opt_extent_cache:
813 			set_opt(sbi, EXTENT_CACHE);
814 			break;
815 		case Opt_noextent_cache:
816 			clear_opt(sbi, EXTENT_CACHE);
817 			break;
818 		case Opt_noinline_data:
819 			clear_opt(sbi, INLINE_DATA);
820 			break;
821 		case Opt_data_flush:
822 			set_opt(sbi, DATA_FLUSH);
823 			break;
824 		case Opt_reserve_root:
825 			if (args->from && match_int(args, &arg))
826 				return -EINVAL;
827 			if (test_opt(sbi, RESERVE_ROOT)) {
828 				f2fs_info(sbi, "Preserve previous reserve_root=%u",
829 					  F2FS_OPTION(sbi).root_reserved_blocks);
830 			} else {
831 				F2FS_OPTION(sbi).root_reserved_blocks = arg;
832 				set_opt(sbi, RESERVE_ROOT);
833 			}
834 			break;
835 		case Opt_resuid:
836 			if (args->from && match_int(args, &arg))
837 				return -EINVAL;
838 			uid = make_kuid(current_user_ns(), arg);
839 			if (!uid_valid(uid)) {
840 				f2fs_err(sbi, "Invalid uid value %d", arg);
841 				return -EINVAL;
842 			}
843 			F2FS_OPTION(sbi).s_resuid = uid;
844 			break;
845 		case Opt_resgid:
846 			if (args->from && match_int(args, &arg))
847 				return -EINVAL;
848 			gid = make_kgid(current_user_ns(), arg);
849 			if (!gid_valid(gid)) {
850 				f2fs_err(sbi, "Invalid gid value %d", arg);
851 				return -EINVAL;
852 			}
853 			F2FS_OPTION(sbi).s_resgid = gid;
854 			break;
855 		case Opt_mode:
856 			name = match_strdup(&args[0]);
857 
858 			if (!name)
859 				return -ENOMEM;
860 			if (!strcmp(name, "adaptive")) {
861 				if (f2fs_sb_has_blkzoned(sbi)) {
862 					f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
863 					kfree(name);
864 					return -EINVAL;
865 				}
866 				F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
867 			} else if (!strcmp(name, "lfs")) {
868 				F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
869 			} else if (!strcmp(name, "fragment:segment")) {
870 				F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
871 			} else if (!strcmp(name, "fragment:block")) {
872 				F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
873 			} else {
874 				kfree(name);
875 				return -EINVAL;
876 			}
877 			kfree(name);
878 			break;
879 		case Opt_io_size_bits:
880 			if (args->from && match_int(args, &arg))
881 				return -EINVAL;
882 			if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
883 				f2fs_warn(sbi, "Not support %d, larger than %d",
884 					  1 << arg, BIO_MAX_VECS);
885 				return -EINVAL;
886 			}
887 			F2FS_OPTION(sbi).write_io_size_bits = arg;
888 			break;
889 #ifdef CONFIG_F2FS_FAULT_INJECTION
890 		case Opt_fault_injection:
891 			if (args->from && match_int(args, &arg))
892 				return -EINVAL;
893 			f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
894 			set_opt(sbi, FAULT_INJECTION);
895 			break;
896 
897 		case Opt_fault_type:
898 			if (args->from && match_int(args, &arg))
899 				return -EINVAL;
900 			f2fs_build_fault_attr(sbi, 0, arg);
901 			set_opt(sbi, FAULT_INJECTION);
902 			break;
903 #else
904 		case Opt_fault_injection:
905 			f2fs_info(sbi, "fault_injection options not supported");
906 			break;
907 
908 		case Opt_fault_type:
909 			f2fs_info(sbi, "fault_type options not supported");
910 			break;
911 #endif
912 		case Opt_lazytime:
913 			sb->s_flags |= SB_LAZYTIME;
914 			break;
915 		case Opt_nolazytime:
916 			sb->s_flags &= ~SB_LAZYTIME;
917 			break;
918 #ifdef CONFIG_QUOTA
919 		case Opt_quota:
920 		case Opt_usrquota:
921 			set_opt(sbi, USRQUOTA);
922 			break;
923 		case Opt_grpquota:
924 			set_opt(sbi, GRPQUOTA);
925 			break;
926 		case Opt_prjquota:
927 			set_opt(sbi, PRJQUOTA);
928 			break;
929 		case Opt_usrjquota:
930 			ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
931 			if (ret)
932 				return ret;
933 			break;
934 		case Opt_grpjquota:
935 			ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
936 			if (ret)
937 				return ret;
938 			break;
939 		case Opt_prjjquota:
940 			ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
941 			if (ret)
942 				return ret;
943 			break;
944 		case Opt_offusrjquota:
945 			ret = f2fs_clear_qf_name(sb, USRQUOTA);
946 			if (ret)
947 				return ret;
948 			break;
949 		case Opt_offgrpjquota:
950 			ret = f2fs_clear_qf_name(sb, GRPQUOTA);
951 			if (ret)
952 				return ret;
953 			break;
954 		case Opt_offprjjquota:
955 			ret = f2fs_clear_qf_name(sb, PRJQUOTA);
956 			if (ret)
957 				return ret;
958 			break;
959 		case Opt_jqfmt_vfsold:
960 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
961 			break;
962 		case Opt_jqfmt_vfsv0:
963 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
964 			break;
965 		case Opt_jqfmt_vfsv1:
966 			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
967 			break;
968 		case Opt_noquota:
969 			clear_opt(sbi, QUOTA);
970 			clear_opt(sbi, USRQUOTA);
971 			clear_opt(sbi, GRPQUOTA);
972 			clear_opt(sbi, PRJQUOTA);
973 			break;
974 #else
975 		case Opt_quota:
976 		case Opt_usrquota:
977 		case Opt_grpquota:
978 		case Opt_prjquota:
979 		case Opt_usrjquota:
980 		case Opt_grpjquota:
981 		case Opt_prjjquota:
982 		case Opt_offusrjquota:
983 		case Opt_offgrpjquota:
984 		case Opt_offprjjquota:
985 		case Opt_jqfmt_vfsold:
986 		case Opt_jqfmt_vfsv0:
987 		case Opt_jqfmt_vfsv1:
988 		case Opt_noquota:
989 			f2fs_info(sbi, "quota operations not supported");
990 			break;
991 #endif
992 		case Opt_alloc:
993 			name = match_strdup(&args[0]);
994 			if (!name)
995 				return -ENOMEM;
996 
997 			if (!strcmp(name, "default")) {
998 				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
999 			} else if (!strcmp(name, "reuse")) {
1000 				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1001 			} else {
1002 				kfree(name);
1003 				return -EINVAL;
1004 			}
1005 			kfree(name);
1006 			break;
1007 		case Opt_fsync:
1008 			name = match_strdup(&args[0]);
1009 			if (!name)
1010 				return -ENOMEM;
1011 			if (!strcmp(name, "posix")) {
1012 				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1013 			} else if (!strcmp(name, "strict")) {
1014 				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1015 			} else if (!strcmp(name, "nobarrier")) {
1016 				F2FS_OPTION(sbi).fsync_mode =
1017 							FSYNC_MODE_NOBARRIER;
1018 			} else {
1019 				kfree(name);
1020 				return -EINVAL;
1021 			}
1022 			kfree(name);
1023 			break;
1024 		case Opt_test_dummy_encryption:
1025 			ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1026 							     is_remount);
1027 			if (ret)
1028 				return ret;
1029 			break;
1030 		case Opt_inlinecrypt:
1031 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1032 			sb->s_flags |= SB_INLINECRYPT;
1033 #else
1034 			f2fs_info(sbi, "inline encryption not supported");
1035 #endif
1036 			break;
1037 		case Opt_checkpoint_disable_cap_perc:
1038 			if (args->from && match_int(args, &arg))
1039 				return -EINVAL;
1040 			if (arg < 0 || arg > 100)
1041 				return -EINVAL;
1042 			F2FS_OPTION(sbi).unusable_cap_perc = arg;
1043 			set_opt(sbi, DISABLE_CHECKPOINT);
1044 			break;
1045 		case Opt_checkpoint_disable_cap:
1046 			if (args->from && match_int(args, &arg))
1047 				return -EINVAL;
1048 			F2FS_OPTION(sbi).unusable_cap = arg;
1049 			set_opt(sbi, DISABLE_CHECKPOINT);
1050 			break;
1051 		case Opt_checkpoint_disable:
1052 			set_opt(sbi, DISABLE_CHECKPOINT);
1053 			break;
1054 		case Opt_checkpoint_enable:
1055 			clear_opt(sbi, DISABLE_CHECKPOINT);
1056 			break;
1057 		case Opt_checkpoint_merge:
1058 			set_opt(sbi, MERGE_CHECKPOINT);
1059 			break;
1060 		case Opt_nocheckpoint_merge:
1061 			clear_opt(sbi, MERGE_CHECKPOINT);
1062 			break;
1063 #ifdef CONFIG_F2FS_FS_COMPRESSION
1064 		case Opt_compress_algorithm:
1065 			if (!f2fs_sb_has_compression(sbi)) {
1066 				f2fs_info(sbi, "Image doesn't support compression");
1067 				break;
1068 			}
1069 			name = match_strdup(&args[0]);
1070 			if (!name)
1071 				return -ENOMEM;
1072 			if (!strcmp(name, "lzo")) {
1073 #ifdef CONFIG_F2FS_FS_LZO
1074 				F2FS_OPTION(sbi).compress_level = 0;
1075 				F2FS_OPTION(sbi).compress_algorithm =
1076 								COMPRESS_LZO;
1077 #else
1078 				f2fs_info(sbi, "kernel doesn't support lzo compression");
1079 #endif
1080 			} else if (!strncmp(name, "lz4", 3)) {
1081 #ifdef CONFIG_F2FS_FS_LZ4
1082 				ret = f2fs_set_lz4hc_level(sbi, name);
1083 				if (ret) {
1084 					kfree(name);
1085 					return -EINVAL;
1086 				}
1087 				F2FS_OPTION(sbi).compress_algorithm =
1088 								COMPRESS_LZ4;
1089 #else
1090 				f2fs_info(sbi, "kernel doesn't support lz4 compression");
1091 #endif
1092 			} else if (!strncmp(name, "zstd", 4)) {
1093 #ifdef CONFIG_F2FS_FS_ZSTD
1094 				ret = f2fs_set_zstd_level(sbi, name);
1095 				if (ret) {
1096 					kfree(name);
1097 					return -EINVAL;
1098 				}
1099 				F2FS_OPTION(sbi).compress_algorithm =
1100 								COMPRESS_ZSTD;
1101 #else
1102 				f2fs_info(sbi, "kernel doesn't support zstd compression");
1103 #endif
1104 			} else if (!strcmp(name, "lzo-rle")) {
1105 #ifdef CONFIG_F2FS_FS_LZORLE
1106 				F2FS_OPTION(sbi).compress_level = 0;
1107 				F2FS_OPTION(sbi).compress_algorithm =
1108 								COMPRESS_LZORLE;
1109 #else
1110 				f2fs_info(sbi, "kernel doesn't support lzorle compression");
1111 #endif
1112 			} else {
1113 				kfree(name);
1114 				return -EINVAL;
1115 			}
1116 			kfree(name);
1117 			break;
1118 		case Opt_compress_log_size:
1119 			if (!f2fs_sb_has_compression(sbi)) {
1120 				f2fs_info(sbi, "Image doesn't support compression");
1121 				break;
1122 			}
1123 			if (args->from && match_int(args, &arg))
1124 				return -EINVAL;
1125 			if (arg < MIN_COMPRESS_LOG_SIZE ||
1126 				arg > MAX_COMPRESS_LOG_SIZE) {
1127 				f2fs_err(sbi,
1128 					"Compress cluster log size is out of range");
1129 				return -EINVAL;
1130 			}
1131 			F2FS_OPTION(sbi).compress_log_size = arg;
1132 			break;
1133 		case Opt_compress_extension:
1134 			if (!f2fs_sb_has_compression(sbi)) {
1135 				f2fs_info(sbi, "Image doesn't support compression");
1136 				break;
1137 			}
1138 			name = match_strdup(&args[0]);
1139 			if (!name)
1140 				return -ENOMEM;
1141 
1142 			ext = F2FS_OPTION(sbi).extensions;
1143 			ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1144 
1145 			if (strlen(name) >= F2FS_EXTENSION_LEN ||
1146 				ext_cnt >= COMPRESS_EXT_NUM) {
1147 				f2fs_err(sbi,
1148 					"invalid extension length/number");
1149 				kfree(name);
1150 				return -EINVAL;
1151 			}
1152 
1153 			strcpy(ext[ext_cnt], name);
1154 			F2FS_OPTION(sbi).compress_ext_cnt++;
1155 			kfree(name);
1156 			break;
1157 		case Opt_nocompress_extension:
1158 			if (!f2fs_sb_has_compression(sbi)) {
1159 				f2fs_info(sbi, "Image doesn't support compression");
1160 				break;
1161 			}
1162 			name = match_strdup(&args[0]);
1163 			if (!name)
1164 				return -ENOMEM;
1165 
1166 			noext = F2FS_OPTION(sbi).noextensions;
1167 			noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1168 
1169 			if (strlen(name) >= F2FS_EXTENSION_LEN ||
1170 				noext_cnt >= COMPRESS_EXT_NUM) {
1171 				f2fs_err(sbi,
1172 					"invalid extension length/number");
1173 				kfree(name);
1174 				return -EINVAL;
1175 			}
1176 
1177 			strcpy(noext[noext_cnt], name);
1178 			F2FS_OPTION(sbi).nocompress_ext_cnt++;
1179 			kfree(name);
1180 			break;
1181 		case Opt_compress_chksum:
1182 			F2FS_OPTION(sbi).compress_chksum = true;
1183 			break;
1184 		case Opt_compress_mode:
1185 			name = match_strdup(&args[0]);
1186 			if (!name)
1187 				return -ENOMEM;
1188 			if (!strcmp(name, "fs")) {
1189 				F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1190 			} else if (!strcmp(name, "user")) {
1191 				F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1192 			} else {
1193 				kfree(name);
1194 				return -EINVAL;
1195 			}
1196 			kfree(name);
1197 			break;
1198 		case Opt_compress_cache:
1199 			set_opt(sbi, COMPRESS_CACHE);
1200 			break;
1201 #else
1202 		case Opt_compress_algorithm:
1203 		case Opt_compress_log_size:
1204 		case Opt_compress_extension:
1205 		case Opt_nocompress_extension:
1206 		case Opt_compress_chksum:
1207 		case Opt_compress_mode:
1208 		case Opt_compress_cache:
1209 			f2fs_info(sbi, "compression options not supported");
1210 			break;
1211 #endif
1212 		case Opt_atgc:
1213 			set_opt(sbi, ATGC);
1214 			break;
1215 		case Opt_gc_merge:
1216 			set_opt(sbi, GC_MERGE);
1217 			break;
1218 		case Opt_nogc_merge:
1219 			clear_opt(sbi, GC_MERGE);
1220 			break;
1221 		case Opt_discard_unit:
1222 			name = match_strdup(&args[0]);
1223 			if (!name)
1224 				return -ENOMEM;
1225 			if (!strcmp(name, "block")) {
1226 				F2FS_OPTION(sbi).discard_unit =
1227 						DISCARD_UNIT_BLOCK;
1228 			} else if (!strcmp(name, "segment")) {
1229 				F2FS_OPTION(sbi).discard_unit =
1230 						DISCARD_UNIT_SEGMENT;
1231 			} else if (!strcmp(name, "section")) {
1232 				F2FS_OPTION(sbi).discard_unit =
1233 						DISCARD_UNIT_SECTION;
1234 			} else {
1235 				kfree(name);
1236 				return -EINVAL;
1237 			}
1238 			kfree(name);
1239 			break;
1240 		case Opt_memory_mode:
1241 			name = match_strdup(&args[0]);
1242 			if (!name)
1243 				return -ENOMEM;
1244 			if (!strcmp(name, "normal")) {
1245 				F2FS_OPTION(sbi).memory_mode =
1246 						MEMORY_MODE_NORMAL;
1247 			} else if (!strcmp(name, "low")) {
1248 				F2FS_OPTION(sbi).memory_mode =
1249 						MEMORY_MODE_LOW;
1250 			} else {
1251 				kfree(name);
1252 				return -EINVAL;
1253 			}
1254 			kfree(name);
1255 			break;
1256 		default:
1257 			f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1258 				 p);
1259 			return -EINVAL;
1260 		}
1261 	}
1262 default_check:
1263 #ifdef CONFIG_QUOTA
1264 	if (f2fs_check_quota_options(sbi))
1265 		return -EINVAL;
1266 #else
1267 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1268 		f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1269 		return -EINVAL;
1270 	}
1271 	if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1272 		f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1273 		return -EINVAL;
1274 	}
1275 #endif
1276 #if !IS_ENABLED(CONFIG_UNICODE)
1277 	if (f2fs_sb_has_casefold(sbi)) {
1278 		f2fs_err(sbi,
1279 			"Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1280 		return -EINVAL;
1281 	}
1282 #endif
1283 	/*
1284 	 * The BLKZONED feature indicates that the drive was formatted with
1285 	 * zone alignment optimization. This is optional for host-aware
1286 	 * devices, but mandatory for host-managed zoned block devices.
1287 	 */
1288 #ifndef CONFIG_BLK_DEV_ZONED
1289 	if (f2fs_sb_has_blkzoned(sbi)) {
1290 		f2fs_err(sbi, "Zoned block device support is not enabled");
1291 		return -EINVAL;
1292 	}
1293 #endif
1294 	if (f2fs_sb_has_blkzoned(sbi)) {
1295 		if (F2FS_OPTION(sbi).discard_unit !=
1296 						DISCARD_UNIT_SECTION) {
1297 			f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1298 			F2FS_OPTION(sbi).discard_unit =
1299 					DISCARD_UNIT_SECTION;
1300 		}
1301 	}
1302 
1303 #ifdef CONFIG_F2FS_FS_COMPRESSION
1304 	if (f2fs_test_compress_extension(sbi)) {
1305 		f2fs_err(sbi, "invalid compress or nocompress extension");
1306 		return -EINVAL;
1307 	}
1308 #endif
1309 
1310 	if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1311 		f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1312 			 F2FS_IO_SIZE_KB(sbi));
1313 		return -EINVAL;
1314 	}
1315 
1316 	if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1317 		int min_size, max_size;
1318 
1319 		if (!f2fs_sb_has_extra_attr(sbi) ||
1320 			!f2fs_sb_has_flexible_inline_xattr(sbi)) {
1321 			f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1322 			return -EINVAL;
1323 		}
1324 		if (!test_opt(sbi, INLINE_XATTR)) {
1325 			f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1326 			return -EINVAL;
1327 		}
1328 
1329 		min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1330 		max_size = MAX_INLINE_XATTR_SIZE;
1331 
1332 		if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1333 				F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1334 			f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1335 				 min_size, max_size);
1336 			return -EINVAL;
1337 		}
1338 	}
1339 
1340 	if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1341 		f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1342 		return -EINVAL;
1343 	}
1344 
1345 	if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1346 		f2fs_err(sbi, "LFS not compatible with ATGC");
1347 		return -EINVAL;
1348 	}
1349 
1350 	if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1351 		f2fs_err(sbi, "Allow to mount readonly mode only");
1352 		return -EROFS;
1353 	}
1354 	return 0;
1355 }
1356 
f2fs_alloc_inode(struct super_block * sb)1357 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1358 {
1359 	struct f2fs_inode_info *fi;
1360 
1361 	if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) {
1362 		f2fs_show_injection_info(F2FS_SB(sb), FAULT_SLAB_ALLOC);
1363 		return NULL;
1364 	}
1365 
1366 	fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1367 	if (!fi)
1368 		return NULL;
1369 
1370 	init_once((void *) fi);
1371 
1372 	/* Initialize f2fs-specific inode info */
1373 	atomic_set(&fi->dirty_pages, 0);
1374 	atomic_set(&fi->i_compr_blocks, 0);
1375 	init_f2fs_rwsem(&fi->i_sem);
1376 	spin_lock_init(&fi->i_size_lock);
1377 	INIT_LIST_HEAD(&fi->dirty_list);
1378 	INIT_LIST_HEAD(&fi->gdirty_list);
1379 	init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1380 	init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1381 	init_f2fs_rwsem(&fi->i_xattr_sem);
1382 
1383 	/* Will be used by directory only */
1384 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
1385 
1386 	return &fi->vfs_inode;
1387 }
1388 
f2fs_drop_inode(struct inode * inode)1389 static int f2fs_drop_inode(struct inode *inode)
1390 {
1391 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1392 	int ret;
1393 
1394 	/*
1395 	 * during filesystem shutdown, if checkpoint is disabled,
1396 	 * drop useless meta/node dirty pages.
1397 	 */
1398 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1399 		if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1400 			inode->i_ino == F2FS_META_INO(sbi)) {
1401 			trace_f2fs_drop_inode(inode, 1);
1402 			return 1;
1403 		}
1404 	}
1405 
1406 	/*
1407 	 * This is to avoid a deadlock condition like below.
1408 	 * writeback_single_inode(inode)
1409 	 *  - f2fs_write_data_page
1410 	 *    - f2fs_gc -> iput -> evict
1411 	 *       - inode_wait_for_writeback(inode)
1412 	 */
1413 	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1414 		if (!inode->i_nlink && !is_bad_inode(inode)) {
1415 			/* to avoid evict_inode call simultaneously */
1416 			atomic_inc(&inode->i_count);
1417 			spin_unlock(&inode->i_lock);
1418 
1419 			f2fs_abort_atomic_write(inode, true);
1420 
1421 			/* should remain fi->extent_tree for writepage */
1422 			f2fs_destroy_extent_node(inode);
1423 
1424 			sb_start_intwrite(inode->i_sb);
1425 			f2fs_i_size_write(inode, 0);
1426 
1427 			f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1428 					inode, NULL, 0, DATA);
1429 			truncate_inode_pages_final(inode->i_mapping);
1430 
1431 			if (F2FS_HAS_BLOCKS(inode))
1432 				f2fs_truncate(inode);
1433 
1434 			sb_end_intwrite(inode->i_sb);
1435 
1436 			spin_lock(&inode->i_lock);
1437 			atomic_dec(&inode->i_count);
1438 		}
1439 		trace_f2fs_drop_inode(inode, 0);
1440 		return 0;
1441 	}
1442 	ret = generic_drop_inode(inode);
1443 	if (!ret)
1444 		ret = fscrypt_drop_inode(inode);
1445 	trace_f2fs_drop_inode(inode, ret);
1446 	return ret;
1447 }
1448 
f2fs_inode_dirtied(struct inode * inode,bool sync)1449 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1450 {
1451 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1452 	int ret = 0;
1453 
1454 	spin_lock(&sbi->inode_lock[DIRTY_META]);
1455 	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1456 		ret = 1;
1457 	} else {
1458 		set_inode_flag(inode, FI_DIRTY_INODE);
1459 		stat_inc_dirty_inode(sbi, DIRTY_META);
1460 	}
1461 	if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1462 		list_add_tail(&F2FS_I(inode)->gdirty_list,
1463 				&sbi->inode_list[DIRTY_META]);
1464 		inc_page_count(sbi, F2FS_DIRTY_IMETA);
1465 	}
1466 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
1467 	return ret;
1468 }
1469 
f2fs_inode_synced(struct inode * inode)1470 void f2fs_inode_synced(struct inode *inode)
1471 {
1472 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1473 
1474 	spin_lock(&sbi->inode_lock[DIRTY_META]);
1475 	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1476 		spin_unlock(&sbi->inode_lock[DIRTY_META]);
1477 		return;
1478 	}
1479 	if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1480 		list_del_init(&F2FS_I(inode)->gdirty_list);
1481 		dec_page_count(sbi, F2FS_DIRTY_IMETA);
1482 	}
1483 	clear_inode_flag(inode, FI_DIRTY_INODE);
1484 	clear_inode_flag(inode, FI_AUTO_RECOVER);
1485 	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1486 	spin_unlock(&sbi->inode_lock[DIRTY_META]);
1487 }
1488 
1489 /*
1490  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1491  *
1492  * We should call set_dirty_inode to write the dirty inode through write_inode.
1493  */
f2fs_dirty_inode(struct inode * inode,int flags)1494 static void f2fs_dirty_inode(struct inode *inode, int flags)
1495 {
1496 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1497 
1498 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1499 			inode->i_ino == F2FS_META_INO(sbi))
1500 		return;
1501 
1502 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1503 		clear_inode_flag(inode, FI_AUTO_RECOVER);
1504 
1505 	f2fs_inode_dirtied(inode, false);
1506 }
1507 
f2fs_free_inode(struct inode * inode)1508 static void f2fs_free_inode(struct inode *inode)
1509 {
1510 	fscrypt_free_inode(inode);
1511 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1512 }
1513 
destroy_percpu_info(struct f2fs_sb_info * sbi)1514 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1515 {
1516 	percpu_counter_destroy(&sbi->total_valid_inode_count);
1517 	percpu_counter_destroy(&sbi->rf_node_block_count);
1518 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
1519 }
1520 
destroy_device_list(struct f2fs_sb_info * sbi)1521 static void destroy_device_list(struct f2fs_sb_info *sbi)
1522 {
1523 	int i;
1524 
1525 	for (i = 0; i < sbi->s_ndevs; i++) {
1526 		blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1527 #ifdef CONFIG_BLK_DEV_ZONED
1528 		kvfree(FDEV(i).blkz_seq);
1529 #endif
1530 	}
1531 	kvfree(sbi->devs);
1532 }
1533 
f2fs_put_super(struct super_block * sb)1534 static void f2fs_put_super(struct super_block *sb)
1535 {
1536 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1537 	int i;
1538 	bool dropped;
1539 
1540 	/* unregister procfs/sysfs entries in advance to avoid race case */
1541 	f2fs_unregister_sysfs(sbi);
1542 
1543 	f2fs_quota_off_umount(sb);
1544 
1545 	/* prevent remaining shrinker jobs */
1546 	mutex_lock(&sbi->umount_mutex);
1547 
1548 	/*
1549 	 * flush all issued checkpoints and stop checkpoint issue thread.
1550 	 * after then, all checkpoints should be done by each process context.
1551 	 */
1552 	f2fs_stop_ckpt_thread(sbi);
1553 
1554 	/*
1555 	 * We don't need to do checkpoint when superblock is clean.
1556 	 * But, the previous checkpoint was not done by umount, it needs to do
1557 	 * clean checkpoint again.
1558 	 */
1559 	if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1560 			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1561 		struct cp_control cpc = {
1562 			.reason = CP_UMOUNT,
1563 		};
1564 		f2fs_write_checkpoint(sbi, &cpc);
1565 	}
1566 
1567 	/* be sure to wait for any on-going discard commands */
1568 	dropped = f2fs_issue_discard_timeout(sbi);
1569 
1570 	if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1571 					!sbi->discard_blks && !dropped) {
1572 		struct cp_control cpc = {
1573 			.reason = CP_UMOUNT | CP_TRIMMED,
1574 		};
1575 		f2fs_write_checkpoint(sbi, &cpc);
1576 	}
1577 
1578 	/*
1579 	 * normally superblock is clean, so we need to release this.
1580 	 * In addition, EIO will skip do checkpoint, we need this as well.
1581 	 */
1582 	f2fs_release_ino_entry(sbi, true);
1583 
1584 	f2fs_leave_shrinker(sbi);
1585 	mutex_unlock(&sbi->umount_mutex);
1586 
1587 	/* our cp_error case, we can wait for any writeback page */
1588 	f2fs_flush_merged_writes(sbi);
1589 
1590 	f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1591 
1592 	f2fs_bug_on(sbi, sbi->fsync_node_num);
1593 
1594 	f2fs_destroy_compress_inode(sbi);
1595 
1596 	iput(sbi->node_inode);
1597 	sbi->node_inode = NULL;
1598 
1599 	iput(sbi->meta_inode);
1600 	sbi->meta_inode = NULL;
1601 
1602 	/*
1603 	 * iput() can update stat information, if f2fs_write_checkpoint()
1604 	 * above failed with error.
1605 	 */
1606 	f2fs_destroy_stats(sbi);
1607 
1608 	/* destroy f2fs internal modules */
1609 	f2fs_destroy_node_manager(sbi);
1610 	f2fs_destroy_segment_manager(sbi);
1611 
1612 	f2fs_destroy_post_read_wq(sbi);
1613 
1614 	kvfree(sbi->ckpt);
1615 
1616 	sb->s_fs_info = NULL;
1617 	if (sbi->s_chksum_driver)
1618 		crypto_free_shash(sbi->s_chksum_driver);
1619 	kfree(sbi->raw_super);
1620 
1621 	destroy_device_list(sbi);
1622 	f2fs_destroy_page_array_cache(sbi);
1623 	f2fs_destroy_xattr_caches(sbi);
1624 	mempool_destroy(sbi->write_io_dummy);
1625 #ifdef CONFIG_QUOTA
1626 	for (i = 0; i < MAXQUOTAS; i++)
1627 		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1628 #endif
1629 	fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1630 	destroy_percpu_info(sbi);
1631 	f2fs_destroy_iostat(sbi);
1632 	for (i = 0; i < NR_PAGE_TYPE; i++)
1633 		kvfree(sbi->write_io[i]);
1634 #if IS_ENABLED(CONFIG_UNICODE)
1635 	utf8_unload(sb->s_encoding);
1636 #endif
1637 	kfree(sbi);
1638 }
1639 
f2fs_sync_fs(struct super_block * sb,int sync)1640 int f2fs_sync_fs(struct super_block *sb, int sync)
1641 {
1642 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1643 	int err = 0;
1644 
1645 	if (unlikely(f2fs_cp_error(sbi)))
1646 		return 0;
1647 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1648 		return 0;
1649 
1650 	trace_f2fs_sync_fs(sb, sync);
1651 
1652 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1653 		return -EAGAIN;
1654 
1655 	if (sync)
1656 		err = f2fs_issue_checkpoint(sbi);
1657 
1658 	return err;
1659 }
1660 
f2fs_freeze(struct super_block * sb)1661 static int f2fs_freeze(struct super_block *sb)
1662 {
1663 	if (f2fs_readonly(sb))
1664 		return 0;
1665 
1666 	/* IO error happened before */
1667 	if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1668 		return -EIO;
1669 
1670 	/* must be clean, since sync_filesystem() was already called */
1671 	if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1672 		return -EINVAL;
1673 
1674 	/* Let's flush checkpoints and stop the thread. */
1675 	f2fs_flush_ckpt_thread(F2FS_SB(sb));
1676 
1677 	/* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1678 	set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1679 	return 0;
1680 }
1681 
f2fs_unfreeze(struct super_block * sb)1682 static int f2fs_unfreeze(struct super_block *sb)
1683 {
1684 	clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1685 	return 0;
1686 }
1687 
1688 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1689 static int f2fs_statfs_project(struct super_block *sb,
1690 				kprojid_t projid, struct kstatfs *buf)
1691 {
1692 	struct kqid qid;
1693 	struct dquot *dquot;
1694 	u64 limit;
1695 	u64 curblock;
1696 
1697 	qid = make_kqid_projid(projid);
1698 	dquot = dqget(sb, qid);
1699 	if (IS_ERR(dquot))
1700 		return PTR_ERR(dquot);
1701 	spin_lock(&dquot->dq_dqb_lock);
1702 
1703 	limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1704 					dquot->dq_dqb.dqb_bhardlimit);
1705 	if (limit)
1706 		limit >>= sb->s_blocksize_bits;
1707 
1708 	if (limit && buf->f_blocks > limit) {
1709 		curblock = (dquot->dq_dqb.dqb_curspace +
1710 			    dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1711 		buf->f_blocks = limit;
1712 		buf->f_bfree = buf->f_bavail =
1713 			(buf->f_blocks > curblock) ?
1714 			 (buf->f_blocks - curblock) : 0;
1715 	}
1716 
1717 	limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1718 					dquot->dq_dqb.dqb_ihardlimit);
1719 
1720 	if (limit && buf->f_files > limit) {
1721 		buf->f_files = limit;
1722 		buf->f_ffree =
1723 			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1724 			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1725 	}
1726 
1727 	spin_unlock(&dquot->dq_dqb_lock);
1728 	dqput(dquot);
1729 	return 0;
1730 }
1731 #endif
1732 
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1733 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1734 {
1735 	struct super_block *sb = dentry->d_sb;
1736 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1737 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1738 	block_t total_count, user_block_count, start_count;
1739 	u64 avail_node_count;
1740 	unsigned int total_valid_node_count;
1741 
1742 	total_count = le64_to_cpu(sbi->raw_super->block_count);
1743 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1744 	buf->f_type = F2FS_SUPER_MAGIC;
1745 	buf->f_bsize = sbi->blocksize;
1746 
1747 	buf->f_blocks = total_count - start_count;
1748 
1749 	spin_lock(&sbi->stat_lock);
1750 
1751 	user_block_count = sbi->user_block_count;
1752 	total_valid_node_count = valid_node_count(sbi);
1753 	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1754 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1755 						sbi->current_reserved_blocks;
1756 
1757 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1758 		buf->f_bfree = 0;
1759 	else
1760 		buf->f_bfree -= sbi->unusable_block_count;
1761 	spin_unlock(&sbi->stat_lock);
1762 
1763 	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1764 		buf->f_bavail = buf->f_bfree -
1765 				F2FS_OPTION(sbi).root_reserved_blocks;
1766 	else
1767 		buf->f_bavail = 0;
1768 
1769 	if (avail_node_count > user_block_count) {
1770 		buf->f_files = user_block_count;
1771 		buf->f_ffree = buf->f_bavail;
1772 	} else {
1773 		buf->f_files = avail_node_count;
1774 		buf->f_ffree = min(avail_node_count - total_valid_node_count,
1775 					buf->f_bavail);
1776 	}
1777 
1778 	buf->f_namelen = F2FS_NAME_LEN;
1779 	buf->f_fsid    = u64_to_fsid(id);
1780 
1781 #ifdef CONFIG_QUOTA
1782 	if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1783 			sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1784 		f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1785 	}
1786 #endif
1787 	return 0;
1788 }
1789 
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1790 static inline void f2fs_show_quota_options(struct seq_file *seq,
1791 					   struct super_block *sb)
1792 {
1793 #ifdef CONFIG_QUOTA
1794 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1795 
1796 	if (F2FS_OPTION(sbi).s_jquota_fmt) {
1797 		char *fmtname = "";
1798 
1799 		switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1800 		case QFMT_VFS_OLD:
1801 			fmtname = "vfsold";
1802 			break;
1803 		case QFMT_VFS_V0:
1804 			fmtname = "vfsv0";
1805 			break;
1806 		case QFMT_VFS_V1:
1807 			fmtname = "vfsv1";
1808 			break;
1809 		}
1810 		seq_printf(seq, ",jqfmt=%s", fmtname);
1811 	}
1812 
1813 	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1814 		seq_show_option(seq, "usrjquota",
1815 			F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1816 
1817 	if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1818 		seq_show_option(seq, "grpjquota",
1819 			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1820 
1821 	if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1822 		seq_show_option(seq, "prjjquota",
1823 			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1824 #endif
1825 }
1826 
1827 #ifdef CONFIG_F2FS_FS_COMPRESSION
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1828 static inline void f2fs_show_compress_options(struct seq_file *seq,
1829 							struct super_block *sb)
1830 {
1831 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1832 	char *algtype = "";
1833 	int i;
1834 
1835 	if (!f2fs_sb_has_compression(sbi))
1836 		return;
1837 
1838 	switch (F2FS_OPTION(sbi).compress_algorithm) {
1839 	case COMPRESS_LZO:
1840 		algtype = "lzo";
1841 		break;
1842 	case COMPRESS_LZ4:
1843 		algtype = "lz4";
1844 		break;
1845 	case COMPRESS_ZSTD:
1846 		algtype = "zstd";
1847 		break;
1848 	case COMPRESS_LZORLE:
1849 		algtype = "lzo-rle";
1850 		break;
1851 	}
1852 	seq_printf(seq, ",compress_algorithm=%s", algtype);
1853 
1854 	if (F2FS_OPTION(sbi).compress_level)
1855 		seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1856 
1857 	seq_printf(seq, ",compress_log_size=%u",
1858 			F2FS_OPTION(sbi).compress_log_size);
1859 
1860 	for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1861 		seq_printf(seq, ",compress_extension=%s",
1862 			F2FS_OPTION(sbi).extensions[i]);
1863 	}
1864 
1865 	for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1866 		seq_printf(seq, ",nocompress_extension=%s",
1867 			F2FS_OPTION(sbi).noextensions[i]);
1868 	}
1869 
1870 	if (F2FS_OPTION(sbi).compress_chksum)
1871 		seq_puts(seq, ",compress_chksum");
1872 
1873 	if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1874 		seq_printf(seq, ",compress_mode=%s", "fs");
1875 	else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1876 		seq_printf(seq, ",compress_mode=%s", "user");
1877 
1878 	if (test_opt(sbi, COMPRESS_CACHE))
1879 		seq_puts(seq, ",compress_cache");
1880 }
1881 #endif
1882 
f2fs_show_options(struct seq_file * seq,struct dentry * root)1883 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1884 {
1885 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1886 
1887 	if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1888 		seq_printf(seq, ",background_gc=%s", "sync");
1889 	else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1890 		seq_printf(seq, ",background_gc=%s", "on");
1891 	else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1892 		seq_printf(seq, ",background_gc=%s", "off");
1893 
1894 	if (test_opt(sbi, GC_MERGE))
1895 		seq_puts(seq, ",gc_merge");
1896 
1897 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1898 		seq_puts(seq, ",disable_roll_forward");
1899 	if (test_opt(sbi, NORECOVERY))
1900 		seq_puts(seq, ",norecovery");
1901 	if (test_opt(sbi, DISCARD))
1902 		seq_puts(seq, ",discard");
1903 	else
1904 		seq_puts(seq, ",nodiscard");
1905 	if (test_opt(sbi, NOHEAP))
1906 		seq_puts(seq, ",no_heap");
1907 	else
1908 		seq_puts(seq, ",heap");
1909 #ifdef CONFIG_F2FS_FS_XATTR
1910 	if (test_opt(sbi, XATTR_USER))
1911 		seq_puts(seq, ",user_xattr");
1912 	else
1913 		seq_puts(seq, ",nouser_xattr");
1914 	if (test_opt(sbi, INLINE_XATTR))
1915 		seq_puts(seq, ",inline_xattr");
1916 	else
1917 		seq_puts(seq, ",noinline_xattr");
1918 	if (test_opt(sbi, INLINE_XATTR_SIZE))
1919 		seq_printf(seq, ",inline_xattr_size=%u",
1920 					F2FS_OPTION(sbi).inline_xattr_size);
1921 #endif
1922 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1923 	if (test_opt(sbi, POSIX_ACL))
1924 		seq_puts(seq, ",acl");
1925 	else
1926 		seq_puts(seq, ",noacl");
1927 #endif
1928 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1929 		seq_puts(seq, ",disable_ext_identify");
1930 	if (test_opt(sbi, INLINE_DATA))
1931 		seq_puts(seq, ",inline_data");
1932 	else
1933 		seq_puts(seq, ",noinline_data");
1934 	if (test_opt(sbi, INLINE_DENTRY))
1935 		seq_puts(seq, ",inline_dentry");
1936 	else
1937 		seq_puts(seq, ",noinline_dentry");
1938 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1939 		seq_puts(seq, ",flush_merge");
1940 	if (test_opt(sbi, NOBARRIER))
1941 		seq_puts(seq, ",nobarrier");
1942 	if (test_opt(sbi, FASTBOOT))
1943 		seq_puts(seq, ",fastboot");
1944 	if (test_opt(sbi, EXTENT_CACHE))
1945 		seq_puts(seq, ",extent_cache");
1946 	else
1947 		seq_puts(seq, ",noextent_cache");
1948 	if (test_opt(sbi, DATA_FLUSH))
1949 		seq_puts(seq, ",data_flush");
1950 
1951 	seq_puts(seq, ",mode=");
1952 	if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1953 		seq_puts(seq, "adaptive");
1954 	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1955 		seq_puts(seq, "lfs");
1956 	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1957 		seq_puts(seq, "fragment:segment");
1958 	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1959 		seq_puts(seq, "fragment:block");
1960 	seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1961 	if (test_opt(sbi, RESERVE_ROOT))
1962 		seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1963 				F2FS_OPTION(sbi).root_reserved_blocks,
1964 				from_kuid_munged(&init_user_ns,
1965 					F2FS_OPTION(sbi).s_resuid),
1966 				from_kgid_munged(&init_user_ns,
1967 					F2FS_OPTION(sbi).s_resgid));
1968 	if (F2FS_IO_SIZE_BITS(sbi))
1969 		seq_printf(seq, ",io_bits=%u",
1970 				F2FS_OPTION(sbi).write_io_size_bits);
1971 #ifdef CONFIG_F2FS_FAULT_INJECTION
1972 	if (test_opt(sbi, FAULT_INJECTION)) {
1973 		seq_printf(seq, ",fault_injection=%u",
1974 				F2FS_OPTION(sbi).fault_info.inject_rate);
1975 		seq_printf(seq, ",fault_type=%u",
1976 				F2FS_OPTION(sbi).fault_info.inject_type);
1977 	}
1978 #endif
1979 #ifdef CONFIG_QUOTA
1980 	if (test_opt(sbi, QUOTA))
1981 		seq_puts(seq, ",quota");
1982 	if (test_opt(sbi, USRQUOTA))
1983 		seq_puts(seq, ",usrquota");
1984 	if (test_opt(sbi, GRPQUOTA))
1985 		seq_puts(seq, ",grpquota");
1986 	if (test_opt(sbi, PRJQUOTA))
1987 		seq_puts(seq, ",prjquota");
1988 #endif
1989 	f2fs_show_quota_options(seq, sbi->sb);
1990 
1991 	fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1992 
1993 	if (sbi->sb->s_flags & SB_INLINECRYPT)
1994 		seq_puts(seq, ",inlinecrypt");
1995 
1996 	if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1997 		seq_printf(seq, ",alloc_mode=%s", "default");
1998 	else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1999 		seq_printf(seq, ",alloc_mode=%s", "reuse");
2000 
2001 	if (test_opt(sbi, DISABLE_CHECKPOINT))
2002 		seq_printf(seq, ",checkpoint=disable:%u",
2003 				F2FS_OPTION(sbi).unusable_cap);
2004 	if (test_opt(sbi, MERGE_CHECKPOINT))
2005 		seq_puts(seq, ",checkpoint_merge");
2006 	else
2007 		seq_puts(seq, ",nocheckpoint_merge");
2008 	if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2009 		seq_printf(seq, ",fsync_mode=%s", "posix");
2010 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2011 		seq_printf(seq, ",fsync_mode=%s", "strict");
2012 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2013 		seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2014 
2015 #ifdef CONFIG_F2FS_FS_COMPRESSION
2016 	f2fs_show_compress_options(seq, sbi->sb);
2017 #endif
2018 
2019 	if (test_opt(sbi, ATGC))
2020 		seq_puts(seq, ",atgc");
2021 
2022 	if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2023 		seq_printf(seq, ",discard_unit=%s", "block");
2024 	else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2025 		seq_printf(seq, ",discard_unit=%s", "segment");
2026 	else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2027 		seq_printf(seq, ",discard_unit=%s", "section");
2028 
2029 	if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2030 		seq_printf(seq, ",memory=%s", "normal");
2031 	else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2032 		seq_printf(seq, ",memory=%s", "low");
2033 
2034 	return 0;
2035 }
2036 
default_options(struct f2fs_sb_info * sbi)2037 static void default_options(struct f2fs_sb_info *sbi)
2038 {
2039 	/* init some FS parameters */
2040 	if (f2fs_sb_has_readonly(sbi))
2041 		F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2042 	else
2043 		F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2044 
2045 	F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2046 	F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2047 	F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2048 	F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2049 	F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2050 	F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2051 	F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2052 	F2FS_OPTION(sbi).compress_ext_cnt = 0;
2053 	F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2054 	F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2055 	F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2056 
2057 	sbi->sb->s_flags &= ~SB_INLINECRYPT;
2058 
2059 	set_opt(sbi, INLINE_XATTR);
2060 	set_opt(sbi, INLINE_DATA);
2061 	set_opt(sbi, INLINE_DENTRY);
2062 	set_opt(sbi, EXTENT_CACHE);
2063 	set_opt(sbi, NOHEAP);
2064 	clear_opt(sbi, DISABLE_CHECKPOINT);
2065 	set_opt(sbi, MERGE_CHECKPOINT);
2066 	F2FS_OPTION(sbi).unusable_cap = 0;
2067 	sbi->sb->s_flags |= SB_LAZYTIME;
2068 	set_opt(sbi, FLUSH_MERGE);
2069 	if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2070 		set_opt(sbi, DISCARD);
2071 	if (f2fs_sb_has_blkzoned(sbi)) {
2072 		F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2073 		F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2074 	} else {
2075 		F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2076 		F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2077 	}
2078 
2079 #ifdef CONFIG_F2FS_FS_XATTR
2080 	set_opt(sbi, XATTR_USER);
2081 #endif
2082 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2083 	set_opt(sbi, POSIX_ACL);
2084 #endif
2085 
2086 	f2fs_build_fault_attr(sbi, 0, 0);
2087 }
2088 
2089 #ifdef CONFIG_QUOTA
2090 static int f2fs_enable_quotas(struct super_block *sb);
2091 #endif
2092 
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2093 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2094 {
2095 	unsigned int s_flags = sbi->sb->s_flags;
2096 	struct cp_control cpc;
2097 	unsigned int gc_mode = sbi->gc_mode;
2098 	int err = 0;
2099 	int ret;
2100 	block_t unusable;
2101 
2102 	if (s_flags & SB_RDONLY) {
2103 		f2fs_err(sbi, "checkpoint=disable on readonly fs");
2104 		return -EINVAL;
2105 	}
2106 	sbi->sb->s_flags |= SB_ACTIVE;
2107 
2108 	/* check if we need more GC first */
2109 	unusable = f2fs_get_unusable_blocks(sbi);
2110 	if (!f2fs_disable_cp_again(sbi, unusable))
2111 		goto skip_gc;
2112 
2113 	f2fs_update_time(sbi, DISABLE_TIME);
2114 
2115 	sbi->gc_mode = GC_URGENT_HIGH;
2116 
2117 	while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2118 		struct f2fs_gc_control gc_control = {
2119 			.victim_segno = NULL_SEGNO,
2120 			.init_gc_type = FG_GC,
2121 			.should_migrate_blocks = false,
2122 			.err_gc_skipped = true,
2123 			.nr_free_secs = 1 };
2124 
2125 		f2fs_down_write(&sbi->gc_lock);
2126 		err = f2fs_gc(sbi, &gc_control);
2127 		if (err == -ENODATA) {
2128 			err = 0;
2129 			break;
2130 		}
2131 		if (err && err != -EAGAIN)
2132 			break;
2133 	}
2134 
2135 	ret = sync_filesystem(sbi->sb);
2136 	if (ret || err) {
2137 		err = ret ? ret : err;
2138 		goto restore_flag;
2139 	}
2140 
2141 	unusable = f2fs_get_unusable_blocks(sbi);
2142 	if (f2fs_disable_cp_again(sbi, unusable)) {
2143 		err = -EAGAIN;
2144 		goto restore_flag;
2145 	}
2146 
2147 skip_gc:
2148 	f2fs_down_write(&sbi->gc_lock);
2149 	cpc.reason = CP_PAUSE;
2150 	set_sbi_flag(sbi, SBI_CP_DISABLED);
2151 	err = f2fs_write_checkpoint(sbi, &cpc);
2152 	if (err)
2153 		goto out_unlock;
2154 
2155 	spin_lock(&sbi->stat_lock);
2156 	sbi->unusable_block_count = unusable;
2157 	spin_unlock(&sbi->stat_lock);
2158 
2159 out_unlock:
2160 	f2fs_up_write(&sbi->gc_lock);
2161 restore_flag:
2162 	sbi->gc_mode = gc_mode;
2163 	sbi->sb->s_flags = s_flags;	/* Restore SB_RDONLY status */
2164 	return err;
2165 }
2166 
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2167 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2168 {
2169 	int retry = DEFAULT_RETRY_IO_COUNT;
2170 
2171 	/* we should flush all the data to keep data consistency */
2172 	do {
2173 		sync_inodes_sb(sbi->sb);
2174 		f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2175 	} while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2176 
2177 	if (unlikely(retry < 0))
2178 		f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2179 
2180 	f2fs_down_write(&sbi->gc_lock);
2181 	f2fs_dirty_to_prefree(sbi);
2182 
2183 	clear_sbi_flag(sbi, SBI_CP_DISABLED);
2184 	set_sbi_flag(sbi, SBI_IS_DIRTY);
2185 	f2fs_up_write(&sbi->gc_lock);
2186 
2187 	f2fs_sync_fs(sbi->sb, 1);
2188 
2189 	/* Let's ensure there's no pending checkpoint anymore */
2190 	f2fs_flush_ckpt_thread(sbi);
2191 }
2192 
f2fs_remount(struct super_block * sb,int * flags,char * data)2193 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2194 {
2195 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2196 	struct f2fs_mount_info org_mount_opt;
2197 	unsigned long old_sb_flags;
2198 	int err;
2199 	bool need_restart_gc = false, need_stop_gc = false;
2200 	bool need_restart_ckpt = false, need_stop_ckpt = false;
2201 	bool need_restart_flush = false, need_stop_flush = false;
2202 	bool need_restart_discard = false, need_stop_discard = false;
2203 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
2204 	bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2205 	bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2206 	bool no_atgc = !test_opt(sbi, ATGC);
2207 	bool no_discard = !test_opt(sbi, DISCARD);
2208 	bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2209 	bool block_unit_discard = f2fs_block_unit_discard(sbi);
2210 	struct discard_cmd_control *dcc;
2211 #ifdef CONFIG_QUOTA
2212 	int i, j;
2213 #endif
2214 
2215 	/*
2216 	 * Save the old mount options in case we
2217 	 * need to restore them.
2218 	 */
2219 	org_mount_opt = sbi->mount_opt;
2220 	old_sb_flags = sb->s_flags;
2221 
2222 #ifdef CONFIG_QUOTA
2223 	org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2224 	for (i = 0; i < MAXQUOTAS; i++) {
2225 		if (F2FS_OPTION(sbi).s_qf_names[i]) {
2226 			org_mount_opt.s_qf_names[i] =
2227 				kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2228 				GFP_KERNEL);
2229 			if (!org_mount_opt.s_qf_names[i]) {
2230 				for (j = 0; j < i; j++)
2231 					kfree(org_mount_opt.s_qf_names[j]);
2232 				return -ENOMEM;
2233 			}
2234 		} else {
2235 			org_mount_opt.s_qf_names[i] = NULL;
2236 		}
2237 	}
2238 #endif
2239 
2240 	/* recover superblocks we couldn't write due to previous RO mount */
2241 	if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2242 		err = f2fs_commit_super(sbi, false);
2243 		f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2244 			  err);
2245 		if (!err)
2246 			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2247 	}
2248 
2249 	default_options(sbi);
2250 
2251 	/* parse mount options */
2252 	err = parse_options(sb, data, true);
2253 	if (err)
2254 		goto restore_opts;
2255 
2256 	/*
2257 	 * Previous and new state of filesystem is RO,
2258 	 * so skip checking GC and FLUSH_MERGE conditions.
2259 	 */
2260 	if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2261 		goto skip;
2262 
2263 	if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2264 		err = -EROFS;
2265 		goto restore_opts;
2266 	}
2267 
2268 #ifdef CONFIG_QUOTA
2269 	if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2270 		err = dquot_suspend(sb, -1);
2271 		if (err < 0)
2272 			goto restore_opts;
2273 	} else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2274 		/* dquot_resume needs RW */
2275 		sb->s_flags &= ~SB_RDONLY;
2276 		if (sb_any_quota_suspended(sb)) {
2277 			dquot_resume(sb, -1);
2278 		} else if (f2fs_sb_has_quota_ino(sbi)) {
2279 			err = f2fs_enable_quotas(sb);
2280 			if (err)
2281 				goto restore_opts;
2282 		}
2283 	}
2284 #endif
2285 	/* disallow enable atgc dynamically */
2286 	if (no_atgc == !!test_opt(sbi, ATGC)) {
2287 		err = -EINVAL;
2288 		f2fs_warn(sbi, "switch atgc option is not allowed");
2289 		goto restore_opts;
2290 	}
2291 
2292 	/* disallow enable/disable extent_cache dynamically */
2293 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
2294 		err = -EINVAL;
2295 		f2fs_warn(sbi, "switch extent_cache option is not allowed");
2296 		goto restore_opts;
2297 	}
2298 
2299 	if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2300 		err = -EINVAL;
2301 		f2fs_warn(sbi, "switch io_bits option is not allowed");
2302 		goto restore_opts;
2303 	}
2304 
2305 	if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2306 		err = -EINVAL;
2307 		f2fs_warn(sbi, "switch compress_cache option is not allowed");
2308 		goto restore_opts;
2309 	}
2310 
2311 	if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2312 		err = -EINVAL;
2313 		f2fs_warn(sbi, "switch discard_unit option is not allowed");
2314 		goto restore_opts;
2315 	}
2316 
2317 	if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2318 		err = -EINVAL;
2319 		f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2320 		goto restore_opts;
2321 	}
2322 
2323 	/*
2324 	 * We stop the GC thread if FS is mounted as RO
2325 	 * or if background_gc = off is passed in mount
2326 	 * option. Also sync the filesystem.
2327 	 */
2328 	if ((*flags & SB_RDONLY) ||
2329 			(F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2330 			!test_opt(sbi, GC_MERGE))) {
2331 		if (sbi->gc_thread) {
2332 			f2fs_stop_gc_thread(sbi);
2333 			need_restart_gc = true;
2334 		}
2335 	} else if (!sbi->gc_thread) {
2336 		err = f2fs_start_gc_thread(sbi);
2337 		if (err)
2338 			goto restore_opts;
2339 		need_stop_gc = true;
2340 	}
2341 
2342 	if (*flags & SB_RDONLY) {
2343 		sync_inodes_sb(sb);
2344 
2345 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2346 		set_sbi_flag(sbi, SBI_IS_CLOSE);
2347 		f2fs_sync_fs(sb, 1);
2348 		clear_sbi_flag(sbi, SBI_IS_CLOSE);
2349 	}
2350 
2351 	if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2352 			!test_opt(sbi, MERGE_CHECKPOINT)) {
2353 		f2fs_stop_ckpt_thread(sbi);
2354 		need_restart_ckpt = true;
2355 	} else {
2356 		/* Flush if the prevous checkpoint, if exists. */
2357 		f2fs_flush_ckpt_thread(sbi);
2358 
2359 		err = f2fs_start_ckpt_thread(sbi);
2360 		if (err) {
2361 			f2fs_err(sbi,
2362 			    "Failed to start F2FS issue_checkpoint_thread (%d)",
2363 			    err);
2364 			goto restore_gc;
2365 		}
2366 		need_stop_ckpt = true;
2367 	}
2368 
2369 	/*
2370 	 * We stop issue flush thread if FS is mounted as RO
2371 	 * or if flush_merge is not passed in mount option.
2372 	 */
2373 	if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2374 		clear_opt(sbi, FLUSH_MERGE);
2375 		f2fs_destroy_flush_cmd_control(sbi, false);
2376 		need_restart_flush = true;
2377 	} else {
2378 		err = f2fs_create_flush_cmd_control(sbi);
2379 		if (err)
2380 			goto restore_ckpt;
2381 		need_stop_flush = true;
2382 	}
2383 
2384 	if (no_discard == !!test_opt(sbi, DISCARD)) {
2385 		if (test_opt(sbi, DISCARD)) {
2386 			err = f2fs_start_discard_thread(sbi);
2387 			if (err)
2388 				goto restore_flush;
2389 			need_stop_discard = true;
2390 		} else {
2391 			dcc = SM_I(sbi)->dcc_info;
2392 			f2fs_stop_discard_thread(sbi);
2393 			if (atomic_read(&dcc->discard_cmd_cnt))
2394 				f2fs_issue_discard_timeout(sbi);
2395 			need_restart_discard = true;
2396 		}
2397 	}
2398 
2399 	if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2400 		if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2401 			err = f2fs_disable_checkpoint(sbi);
2402 			if (err)
2403 				goto restore_discard;
2404 		} else {
2405 			f2fs_enable_checkpoint(sbi);
2406 		}
2407 	}
2408 
2409 skip:
2410 #ifdef CONFIG_QUOTA
2411 	/* Release old quota file names */
2412 	for (i = 0; i < MAXQUOTAS; i++)
2413 		kfree(org_mount_opt.s_qf_names[i]);
2414 #endif
2415 	/* Update the POSIXACL Flag */
2416 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2417 		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2418 
2419 	limit_reserve_root(sbi);
2420 	adjust_unusable_cap_perc(sbi);
2421 	*flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2422 	return 0;
2423 restore_discard:
2424 	if (need_restart_discard) {
2425 		if (f2fs_start_discard_thread(sbi))
2426 			f2fs_warn(sbi, "discard has been stopped");
2427 	} else if (need_stop_discard) {
2428 		f2fs_stop_discard_thread(sbi);
2429 	}
2430 restore_flush:
2431 	if (need_restart_flush) {
2432 		if (f2fs_create_flush_cmd_control(sbi))
2433 			f2fs_warn(sbi, "background flush thread has stopped");
2434 	} else if (need_stop_flush) {
2435 		clear_opt(sbi, FLUSH_MERGE);
2436 		f2fs_destroy_flush_cmd_control(sbi, false);
2437 	}
2438 restore_ckpt:
2439 	if (need_restart_ckpt) {
2440 		if (f2fs_start_ckpt_thread(sbi))
2441 			f2fs_warn(sbi, "background ckpt thread has stopped");
2442 	} else if (need_stop_ckpt) {
2443 		f2fs_stop_ckpt_thread(sbi);
2444 	}
2445 restore_gc:
2446 	if (need_restart_gc) {
2447 		if (f2fs_start_gc_thread(sbi))
2448 			f2fs_warn(sbi, "background gc thread has stopped");
2449 	} else if (need_stop_gc) {
2450 		f2fs_stop_gc_thread(sbi);
2451 	}
2452 restore_opts:
2453 #ifdef CONFIG_QUOTA
2454 	F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2455 	for (i = 0; i < MAXQUOTAS; i++) {
2456 		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2457 		F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2458 	}
2459 #endif
2460 	sbi->mount_opt = org_mount_opt;
2461 	sb->s_flags = old_sb_flags;
2462 	return err;
2463 }
2464 
2465 #ifdef CONFIG_QUOTA
2466 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2467 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2468 			       size_t len, loff_t off)
2469 {
2470 	struct inode *inode = sb_dqopt(sb)->files[type];
2471 	struct address_space *mapping = inode->i_mapping;
2472 	block_t blkidx = F2FS_BYTES_TO_BLK(off);
2473 	int offset = off & (sb->s_blocksize - 1);
2474 	int tocopy;
2475 	size_t toread;
2476 	loff_t i_size = i_size_read(inode);
2477 	struct page *page;
2478 
2479 	if (off > i_size)
2480 		return 0;
2481 
2482 	if (off + len > i_size)
2483 		len = i_size - off;
2484 	toread = len;
2485 	while (toread > 0) {
2486 		tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2487 repeat:
2488 		page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2489 		if (IS_ERR(page)) {
2490 			if (PTR_ERR(page) == -ENOMEM) {
2491 				memalloc_retry_wait(GFP_NOFS);
2492 				goto repeat;
2493 			}
2494 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2495 			return PTR_ERR(page);
2496 		}
2497 
2498 		lock_page(page);
2499 
2500 		if (unlikely(page->mapping != mapping)) {
2501 			f2fs_put_page(page, 1);
2502 			goto repeat;
2503 		}
2504 		if (unlikely(!PageUptodate(page))) {
2505 			f2fs_put_page(page, 1);
2506 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2507 			return -EIO;
2508 		}
2509 
2510 		memcpy_from_page(data, page, offset, tocopy);
2511 		f2fs_put_page(page, 1);
2512 
2513 		offset = 0;
2514 		toread -= tocopy;
2515 		data += tocopy;
2516 		blkidx++;
2517 	}
2518 	return len;
2519 }
2520 
2521 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2522 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2523 				const char *data, size_t len, loff_t off)
2524 {
2525 	struct inode *inode = sb_dqopt(sb)->files[type];
2526 	struct address_space *mapping = inode->i_mapping;
2527 	const struct address_space_operations *a_ops = mapping->a_ops;
2528 	int offset = off & (sb->s_blocksize - 1);
2529 	size_t towrite = len;
2530 	struct page *page;
2531 	void *fsdata = NULL;
2532 	int err = 0;
2533 	int tocopy;
2534 
2535 	while (towrite > 0) {
2536 		tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2537 								towrite);
2538 retry:
2539 		err = a_ops->write_begin(NULL, mapping, off, tocopy,
2540 							&page, &fsdata);
2541 		if (unlikely(err)) {
2542 			if (err == -ENOMEM) {
2543 				f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2544 				goto retry;
2545 			}
2546 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2547 			break;
2548 		}
2549 
2550 		memcpy_to_page(page, offset, data, tocopy);
2551 
2552 		a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2553 						page, fsdata);
2554 		offset = 0;
2555 		towrite -= tocopy;
2556 		off += tocopy;
2557 		data += tocopy;
2558 		cond_resched();
2559 	}
2560 
2561 	if (len == towrite)
2562 		return err;
2563 	inode->i_mtime = inode->i_ctime = current_time(inode);
2564 	f2fs_mark_inode_dirty_sync(inode, false);
2565 	return len - towrite;
2566 }
2567 
f2fs_dquot_initialize(struct inode * inode)2568 int f2fs_dquot_initialize(struct inode *inode)
2569 {
2570 	if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2571 		f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2572 		return -ESRCH;
2573 	}
2574 
2575 	return dquot_initialize(inode);
2576 }
2577 
f2fs_get_dquots(struct inode * inode)2578 static struct dquot **f2fs_get_dquots(struct inode *inode)
2579 {
2580 	return F2FS_I(inode)->i_dquot;
2581 }
2582 
f2fs_get_reserved_space(struct inode * inode)2583 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2584 {
2585 	return &F2FS_I(inode)->i_reserved_quota;
2586 }
2587 
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2588 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2589 {
2590 	if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2591 		f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2592 		return 0;
2593 	}
2594 
2595 	return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2596 					F2FS_OPTION(sbi).s_jquota_fmt, type);
2597 }
2598 
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2599 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2600 {
2601 	int enabled = 0;
2602 	int i, err;
2603 
2604 	if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2605 		err = f2fs_enable_quotas(sbi->sb);
2606 		if (err) {
2607 			f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2608 			return 0;
2609 		}
2610 		return 1;
2611 	}
2612 
2613 	for (i = 0; i < MAXQUOTAS; i++) {
2614 		if (F2FS_OPTION(sbi).s_qf_names[i]) {
2615 			err = f2fs_quota_on_mount(sbi, i);
2616 			if (!err) {
2617 				enabled = 1;
2618 				continue;
2619 			}
2620 			f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2621 				 err, i);
2622 		}
2623 	}
2624 	return enabled;
2625 }
2626 
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2627 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2628 			     unsigned int flags)
2629 {
2630 	struct inode *qf_inode;
2631 	unsigned long qf_inum;
2632 	int err;
2633 
2634 	BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2635 
2636 	qf_inum = f2fs_qf_ino(sb, type);
2637 	if (!qf_inum)
2638 		return -EPERM;
2639 
2640 	qf_inode = f2fs_iget(sb, qf_inum);
2641 	if (IS_ERR(qf_inode)) {
2642 		f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2643 		return PTR_ERR(qf_inode);
2644 	}
2645 
2646 	/* Don't account quota for quota files to avoid recursion */
2647 	qf_inode->i_flags |= S_NOQUOTA;
2648 	err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2649 	iput(qf_inode);
2650 	return err;
2651 }
2652 
f2fs_enable_quotas(struct super_block * sb)2653 static int f2fs_enable_quotas(struct super_block *sb)
2654 {
2655 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2656 	int type, err = 0;
2657 	unsigned long qf_inum;
2658 	bool quota_mopt[MAXQUOTAS] = {
2659 		test_opt(sbi, USRQUOTA),
2660 		test_opt(sbi, GRPQUOTA),
2661 		test_opt(sbi, PRJQUOTA),
2662 	};
2663 
2664 	if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2665 		f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2666 		return 0;
2667 	}
2668 
2669 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2670 
2671 	for (type = 0; type < MAXQUOTAS; type++) {
2672 		qf_inum = f2fs_qf_ino(sb, type);
2673 		if (qf_inum) {
2674 			err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2675 				DQUOT_USAGE_ENABLED |
2676 				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2677 			if (err) {
2678 				f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2679 					 type, err);
2680 				for (type--; type >= 0; type--)
2681 					dquot_quota_off(sb, type);
2682 				set_sbi_flag(F2FS_SB(sb),
2683 						SBI_QUOTA_NEED_REPAIR);
2684 				return err;
2685 			}
2686 		}
2687 	}
2688 	return 0;
2689 }
2690 
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2691 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2692 {
2693 	struct quota_info *dqopt = sb_dqopt(sbi->sb);
2694 	struct address_space *mapping = dqopt->files[type]->i_mapping;
2695 	int ret = 0;
2696 
2697 	ret = dquot_writeback_dquots(sbi->sb, type);
2698 	if (ret)
2699 		goto out;
2700 
2701 	ret = filemap_fdatawrite(mapping);
2702 	if (ret)
2703 		goto out;
2704 
2705 	/* if we are using journalled quota */
2706 	if (is_journalled_quota(sbi))
2707 		goto out;
2708 
2709 	ret = filemap_fdatawait(mapping);
2710 
2711 	truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2712 out:
2713 	if (ret)
2714 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2715 	return ret;
2716 }
2717 
f2fs_quota_sync(struct super_block * sb,int type)2718 int f2fs_quota_sync(struct super_block *sb, int type)
2719 {
2720 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2721 	struct quota_info *dqopt = sb_dqopt(sb);
2722 	int cnt;
2723 	int ret = 0;
2724 
2725 	/*
2726 	 * Now when everything is written we can discard the pagecache so
2727 	 * that userspace sees the changes.
2728 	 */
2729 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2730 
2731 		if (type != -1 && cnt != type)
2732 			continue;
2733 
2734 		if (!sb_has_quota_active(sb, cnt))
2735 			continue;
2736 
2737 		if (!f2fs_sb_has_quota_ino(sbi))
2738 			inode_lock(dqopt->files[cnt]);
2739 
2740 		/*
2741 		 * do_quotactl
2742 		 *  f2fs_quota_sync
2743 		 *  f2fs_down_read(quota_sem)
2744 		 *  dquot_writeback_dquots()
2745 		 *  f2fs_dquot_commit
2746 		 *			      block_operation
2747 		 *			      f2fs_down_read(quota_sem)
2748 		 */
2749 		f2fs_lock_op(sbi);
2750 		f2fs_down_read(&sbi->quota_sem);
2751 
2752 		ret = f2fs_quota_sync_file(sbi, cnt);
2753 
2754 		f2fs_up_read(&sbi->quota_sem);
2755 		f2fs_unlock_op(sbi);
2756 
2757 		if (!f2fs_sb_has_quota_ino(sbi))
2758 			inode_unlock(dqopt->files[cnt]);
2759 
2760 		if (ret)
2761 			break;
2762 	}
2763 	return ret;
2764 }
2765 
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2766 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2767 							const struct path *path)
2768 {
2769 	struct inode *inode;
2770 	int err;
2771 
2772 	/* if quota sysfile exists, deny enabling quota with specific file */
2773 	if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2774 		f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2775 		return -EBUSY;
2776 	}
2777 
2778 	err = f2fs_quota_sync(sb, type);
2779 	if (err)
2780 		return err;
2781 
2782 	err = dquot_quota_on(sb, type, format_id, path);
2783 	if (err)
2784 		return err;
2785 
2786 	inode = d_inode(path->dentry);
2787 
2788 	inode_lock(inode);
2789 	F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2790 	f2fs_set_inode_flags(inode);
2791 	inode_unlock(inode);
2792 	f2fs_mark_inode_dirty_sync(inode, false);
2793 
2794 	return 0;
2795 }
2796 
__f2fs_quota_off(struct super_block * sb,int type)2797 static int __f2fs_quota_off(struct super_block *sb, int type)
2798 {
2799 	struct inode *inode = sb_dqopt(sb)->files[type];
2800 	int err;
2801 
2802 	if (!inode || !igrab(inode))
2803 		return dquot_quota_off(sb, type);
2804 
2805 	err = f2fs_quota_sync(sb, type);
2806 	if (err)
2807 		goto out_put;
2808 
2809 	err = dquot_quota_off(sb, type);
2810 	if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2811 		goto out_put;
2812 
2813 	inode_lock(inode);
2814 	F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2815 	f2fs_set_inode_flags(inode);
2816 	inode_unlock(inode);
2817 	f2fs_mark_inode_dirty_sync(inode, false);
2818 out_put:
2819 	iput(inode);
2820 	return err;
2821 }
2822 
f2fs_quota_off(struct super_block * sb,int type)2823 static int f2fs_quota_off(struct super_block *sb, int type)
2824 {
2825 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2826 	int err;
2827 
2828 	err = __f2fs_quota_off(sb, type);
2829 
2830 	/*
2831 	 * quotactl can shutdown journalled quota, result in inconsistence
2832 	 * between quota record and fs data by following updates, tag the
2833 	 * flag to let fsck be aware of it.
2834 	 */
2835 	if (is_journalled_quota(sbi))
2836 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2837 	return err;
2838 }
2839 
f2fs_quota_off_umount(struct super_block * sb)2840 void f2fs_quota_off_umount(struct super_block *sb)
2841 {
2842 	int type;
2843 	int err;
2844 
2845 	for (type = 0; type < MAXQUOTAS; type++) {
2846 		err = __f2fs_quota_off(sb, type);
2847 		if (err) {
2848 			int ret = dquot_quota_off(sb, type);
2849 
2850 			f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2851 				 type, err, ret);
2852 			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2853 		}
2854 	}
2855 	/*
2856 	 * In case of checkpoint=disable, we must flush quota blocks.
2857 	 * This can cause NULL exception for node_inode in end_io, since
2858 	 * put_super already dropped it.
2859 	 */
2860 	sync_filesystem(sb);
2861 }
2862 
f2fs_truncate_quota_inode_pages(struct super_block * sb)2863 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2864 {
2865 	struct quota_info *dqopt = sb_dqopt(sb);
2866 	int type;
2867 
2868 	for (type = 0; type < MAXQUOTAS; type++) {
2869 		if (!dqopt->files[type])
2870 			continue;
2871 		f2fs_inode_synced(dqopt->files[type]);
2872 	}
2873 }
2874 
f2fs_dquot_commit(struct dquot * dquot)2875 static int f2fs_dquot_commit(struct dquot *dquot)
2876 {
2877 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2878 	int ret;
2879 
2880 	f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2881 	ret = dquot_commit(dquot);
2882 	if (ret < 0)
2883 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2884 	f2fs_up_read(&sbi->quota_sem);
2885 	return ret;
2886 }
2887 
f2fs_dquot_acquire(struct dquot * dquot)2888 static int f2fs_dquot_acquire(struct dquot *dquot)
2889 {
2890 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2891 	int ret;
2892 
2893 	f2fs_down_read(&sbi->quota_sem);
2894 	ret = dquot_acquire(dquot);
2895 	if (ret < 0)
2896 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2897 	f2fs_up_read(&sbi->quota_sem);
2898 	return ret;
2899 }
2900 
f2fs_dquot_release(struct dquot * dquot)2901 static int f2fs_dquot_release(struct dquot *dquot)
2902 {
2903 	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2904 	int ret = dquot_release(dquot);
2905 
2906 	if (ret < 0)
2907 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2908 	return ret;
2909 }
2910 
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)2911 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2912 {
2913 	struct super_block *sb = dquot->dq_sb;
2914 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2915 	int ret = dquot_mark_dquot_dirty(dquot);
2916 
2917 	/* if we are using journalled quota */
2918 	if (is_journalled_quota(sbi))
2919 		set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2920 
2921 	return ret;
2922 }
2923 
f2fs_dquot_commit_info(struct super_block * sb,int type)2924 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2925 {
2926 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2927 	int ret = dquot_commit_info(sb, type);
2928 
2929 	if (ret < 0)
2930 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2931 	return ret;
2932 }
2933 
f2fs_get_projid(struct inode * inode,kprojid_t * projid)2934 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2935 {
2936 	*projid = F2FS_I(inode)->i_projid;
2937 	return 0;
2938 }
2939 
2940 static const struct dquot_operations f2fs_quota_operations = {
2941 	.get_reserved_space = f2fs_get_reserved_space,
2942 	.write_dquot	= f2fs_dquot_commit,
2943 	.acquire_dquot	= f2fs_dquot_acquire,
2944 	.release_dquot	= f2fs_dquot_release,
2945 	.mark_dirty	= f2fs_dquot_mark_dquot_dirty,
2946 	.write_info	= f2fs_dquot_commit_info,
2947 	.alloc_dquot	= dquot_alloc,
2948 	.destroy_dquot	= dquot_destroy,
2949 	.get_projid	= f2fs_get_projid,
2950 	.get_next_id	= dquot_get_next_id,
2951 };
2952 
2953 static const struct quotactl_ops f2fs_quotactl_ops = {
2954 	.quota_on	= f2fs_quota_on,
2955 	.quota_off	= f2fs_quota_off,
2956 	.quota_sync	= f2fs_quota_sync,
2957 	.get_state	= dquot_get_state,
2958 	.set_info	= dquot_set_dqinfo,
2959 	.get_dqblk	= dquot_get_dqblk,
2960 	.set_dqblk	= dquot_set_dqblk,
2961 	.get_nextdqblk	= dquot_get_next_dqblk,
2962 };
2963 #else
f2fs_dquot_initialize(struct inode * inode)2964 int f2fs_dquot_initialize(struct inode *inode)
2965 {
2966 	return 0;
2967 }
2968 
f2fs_quota_sync(struct super_block * sb,int type)2969 int f2fs_quota_sync(struct super_block *sb, int type)
2970 {
2971 	return 0;
2972 }
2973 
f2fs_quota_off_umount(struct super_block * sb)2974 void f2fs_quota_off_umount(struct super_block *sb)
2975 {
2976 }
2977 #endif
2978 
2979 static const struct super_operations f2fs_sops = {
2980 	.alloc_inode	= f2fs_alloc_inode,
2981 	.free_inode	= f2fs_free_inode,
2982 	.drop_inode	= f2fs_drop_inode,
2983 	.write_inode	= f2fs_write_inode,
2984 	.dirty_inode	= f2fs_dirty_inode,
2985 	.show_options	= f2fs_show_options,
2986 #ifdef CONFIG_QUOTA
2987 	.quota_read	= f2fs_quota_read,
2988 	.quota_write	= f2fs_quota_write,
2989 	.get_dquots	= f2fs_get_dquots,
2990 #endif
2991 	.evict_inode	= f2fs_evict_inode,
2992 	.put_super	= f2fs_put_super,
2993 	.sync_fs	= f2fs_sync_fs,
2994 	.freeze_fs	= f2fs_freeze,
2995 	.unfreeze_fs	= f2fs_unfreeze,
2996 	.statfs		= f2fs_statfs,
2997 	.remount_fs	= f2fs_remount,
2998 };
2999 
3000 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)3001 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3002 {
3003 	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3004 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3005 				ctx, len, NULL);
3006 }
3007 
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3008 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3009 							void *fs_data)
3010 {
3011 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3012 
3013 	/*
3014 	 * Encrypting the root directory is not allowed because fsck
3015 	 * expects lost+found directory to exist and remain unencrypted
3016 	 * if LOST_FOUND feature is enabled.
3017 	 *
3018 	 */
3019 	if (f2fs_sb_has_lost_found(sbi) &&
3020 			inode->i_ino == F2FS_ROOT_INO(sbi))
3021 		return -EPERM;
3022 
3023 	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3024 				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3025 				ctx, len, fs_data, XATTR_CREATE);
3026 }
3027 
f2fs_get_dummy_policy(struct super_block * sb)3028 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3029 {
3030 	return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3031 }
3032 
f2fs_has_stable_inodes(struct super_block * sb)3033 static bool f2fs_has_stable_inodes(struct super_block *sb)
3034 {
3035 	return true;
3036 }
3037 
f2fs_get_ino_and_lblk_bits(struct super_block * sb,int * ino_bits_ret,int * lblk_bits_ret)3038 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3039 				       int *ino_bits_ret, int *lblk_bits_ret)
3040 {
3041 	*ino_bits_ret = 8 * sizeof(nid_t);
3042 	*lblk_bits_ret = 8 * sizeof(block_t);
3043 }
3044 
f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3045 static struct block_device **f2fs_get_devices(struct super_block *sb,
3046 					      unsigned int *num_devs)
3047 {
3048 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
3049 	struct block_device **devs;
3050 	int i;
3051 
3052 	if (!f2fs_is_multi_device(sbi))
3053 		return NULL;
3054 
3055 	devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3056 	if (!devs)
3057 		return ERR_PTR(-ENOMEM);
3058 
3059 	for (i = 0; i < sbi->s_ndevs; i++)
3060 		devs[i] = FDEV(i).bdev;
3061 	*num_devs = sbi->s_ndevs;
3062 	return devs;
3063 }
3064 
3065 static const struct fscrypt_operations f2fs_cryptops = {
3066 	.key_prefix		= "f2fs:",
3067 	.get_context		= f2fs_get_context,
3068 	.set_context		= f2fs_set_context,
3069 	.get_dummy_policy	= f2fs_get_dummy_policy,
3070 	.empty_dir		= f2fs_empty_dir,
3071 	.has_stable_inodes	= f2fs_has_stable_inodes,
3072 	.get_ino_and_lblk_bits	= f2fs_get_ino_and_lblk_bits,
3073 	.get_devices		= f2fs_get_devices,
3074 };
3075 #endif
3076 
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3077 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3078 		u64 ino, u32 generation)
3079 {
3080 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
3081 	struct inode *inode;
3082 
3083 	if (f2fs_check_nid_range(sbi, ino))
3084 		return ERR_PTR(-ESTALE);
3085 
3086 	/*
3087 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
3088 	 * However f2fs_iget currently does appropriate checks to handle stale
3089 	 * inodes so everything is OK.
3090 	 */
3091 	inode = f2fs_iget(sb, ino);
3092 	if (IS_ERR(inode))
3093 		return ERR_CAST(inode);
3094 	if (unlikely(generation && inode->i_generation != generation)) {
3095 		/* we didn't find the right inode.. */
3096 		iput(inode);
3097 		return ERR_PTR(-ESTALE);
3098 	}
3099 	return inode;
3100 }
3101 
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3102 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3103 		int fh_len, int fh_type)
3104 {
3105 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3106 				    f2fs_nfs_get_inode);
3107 }
3108 
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3109 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3110 		int fh_len, int fh_type)
3111 {
3112 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3113 				    f2fs_nfs_get_inode);
3114 }
3115 
3116 static const struct export_operations f2fs_export_ops = {
3117 	.fh_to_dentry = f2fs_fh_to_dentry,
3118 	.fh_to_parent = f2fs_fh_to_parent,
3119 	.get_parent = f2fs_get_parent,
3120 };
3121 
max_file_blocks(struct inode * inode)3122 loff_t max_file_blocks(struct inode *inode)
3123 {
3124 	loff_t result = 0;
3125 	loff_t leaf_count;
3126 
3127 	/*
3128 	 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3129 	 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3130 	 * space in inode.i_addr, it will be more safe to reassign
3131 	 * result as zero.
3132 	 */
3133 
3134 	if (inode && f2fs_compressed_file(inode))
3135 		leaf_count = ADDRS_PER_BLOCK(inode);
3136 	else
3137 		leaf_count = DEF_ADDRS_PER_BLOCK;
3138 
3139 	/* two direct node blocks */
3140 	result += (leaf_count * 2);
3141 
3142 	/* two indirect node blocks */
3143 	leaf_count *= NIDS_PER_BLOCK;
3144 	result += (leaf_count * 2);
3145 
3146 	/* one double indirect node block */
3147 	leaf_count *= NIDS_PER_BLOCK;
3148 	result += leaf_count;
3149 
3150 	return result;
3151 }
3152 
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)3153 static int __f2fs_commit_super(struct buffer_head *bh,
3154 			struct f2fs_super_block *super)
3155 {
3156 	lock_buffer(bh);
3157 	if (super)
3158 		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3159 	set_buffer_dirty(bh);
3160 	unlock_buffer(bh);
3161 
3162 	/* it's rare case, we can do fua all the time */
3163 	return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3164 }
3165 
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)3166 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3167 					struct buffer_head *bh)
3168 {
3169 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3170 					(bh->b_data + F2FS_SUPER_OFFSET);
3171 	struct super_block *sb = sbi->sb;
3172 	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3173 	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3174 	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3175 	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3176 	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3177 	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3178 	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3179 	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3180 	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3181 	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3182 	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3183 	u32 segment_count = le32_to_cpu(raw_super->segment_count);
3184 	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3185 	u64 main_end_blkaddr = main_blkaddr +
3186 				(segment_count_main << log_blocks_per_seg);
3187 	u64 seg_end_blkaddr = segment0_blkaddr +
3188 				(segment_count << log_blocks_per_seg);
3189 
3190 	if (segment0_blkaddr != cp_blkaddr) {
3191 		f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3192 			  segment0_blkaddr, cp_blkaddr);
3193 		return true;
3194 	}
3195 
3196 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3197 							sit_blkaddr) {
3198 		f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3199 			  cp_blkaddr, sit_blkaddr,
3200 			  segment_count_ckpt << log_blocks_per_seg);
3201 		return true;
3202 	}
3203 
3204 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3205 							nat_blkaddr) {
3206 		f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3207 			  sit_blkaddr, nat_blkaddr,
3208 			  segment_count_sit << log_blocks_per_seg);
3209 		return true;
3210 	}
3211 
3212 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3213 							ssa_blkaddr) {
3214 		f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3215 			  nat_blkaddr, ssa_blkaddr,
3216 			  segment_count_nat << log_blocks_per_seg);
3217 		return true;
3218 	}
3219 
3220 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3221 							main_blkaddr) {
3222 		f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3223 			  ssa_blkaddr, main_blkaddr,
3224 			  segment_count_ssa << log_blocks_per_seg);
3225 		return true;
3226 	}
3227 
3228 	if (main_end_blkaddr > seg_end_blkaddr) {
3229 		f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3230 			  main_blkaddr, seg_end_blkaddr,
3231 			  segment_count_main << log_blocks_per_seg);
3232 		return true;
3233 	} else if (main_end_blkaddr < seg_end_blkaddr) {
3234 		int err = 0;
3235 		char *res;
3236 
3237 		/* fix in-memory information all the time */
3238 		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3239 				segment0_blkaddr) >> log_blocks_per_seg);
3240 
3241 		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3242 			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3243 			res = "internally";
3244 		} else {
3245 			err = __f2fs_commit_super(bh, NULL);
3246 			res = err ? "failed" : "done";
3247 		}
3248 		f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3249 			  res, main_blkaddr, seg_end_blkaddr,
3250 			  segment_count_main << log_blocks_per_seg);
3251 		if (err)
3252 			return true;
3253 	}
3254 	return false;
3255 }
3256 
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)3257 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3258 				struct buffer_head *bh)
3259 {
3260 	block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3261 	block_t total_sections, blocks_per_seg;
3262 	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3263 					(bh->b_data + F2FS_SUPER_OFFSET);
3264 	size_t crc_offset = 0;
3265 	__u32 crc = 0;
3266 
3267 	if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3268 		f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3269 			  F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3270 		return -EINVAL;
3271 	}
3272 
3273 	/* Check checksum_offset and crc in superblock */
3274 	if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3275 		crc_offset = le32_to_cpu(raw_super->checksum_offset);
3276 		if (crc_offset !=
3277 			offsetof(struct f2fs_super_block, crc)) {
3278 			f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3279 				  crc_offset);
3280 			return -EFSCORRUPTED;
3281 		}
3282 		crc = le32_to_cpu(raw_super->crc);
3283 		if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3284 			f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3285 			return -EFSCORRUPTED;
3286 		}
3287 	}
3288 
3289 	/* Currently, support only 4KB block size */
3290 	if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3291 		f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3292 			  le32_to_cpu(raw_super->log_blocksize),
3293 			  F2FS_BLKSIZE_BITS);
3294 		return -EFSCORRUPTED;
3295 	}
3296 
3297 	/* check log blocks per segment */
3298 	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3299 		f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3300 			  le32_to_cpu(raw_super->log_blocks_per_seg));
3301 		return -EFSCORRUPTED;
3302 	}
3303 
3304 	/* Currently, support 512/1024/2048/4096 bytes sector size */
3305 	if (le32_to_cpu(raw_super->log_sectorsize) >
3306 				F2FS_MAX_LOG_SECTOR_SIZE ||
3307 		le32_to_cpu(raw_super->log_sectorsize) <
3308 				F2FS_MIN_LOG_SECTOR_SIZE) {
3309 		f2fs_info(sbi, "Invalid log sectorsize (%u)",
3310 			  le32_to_cpu(raw_super->log_sectorsize));
3311 		return -EFSCORRUPTED;
3312 	}
3313 	if (le32_to_cpu(raw_super->log_sectors_per_block) +
3314 		le32_to_cpu(raw_super->log_sectorsize) !=
3315 			F2FS_MAX_LOG_SECTOR_SIZE) {
3316 		f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3317 			  le32_to_cpu(raw_super->log_sectors_per_block),
3318 			  le32_to_cpu(raw_super->log_sectorsize));
3319 		return -EFSCORRUPTED;
3320 	}
3321 
3322 	segment_count = le32_to_cpu(raw_super->segment_count);
3323 	segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3324 	segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3325 	secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3326 	total_sections = le32_to_cpu(raw_super->section_count);
3327 
3328 	/* blocks_per_seg should be 512, given the above check */
3329 	blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3330 
3331 	if (segment_count > F2FS_MAX_SEGMENT ||
3332 				segment_count < F2FS_MIN_SEGMENTS) {
3333 		f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3334 		return -EFSCORRUPTED;
3335 	}
3336 
3337 	if (total_sections > segment_count_main || total_sections < 1 ||
3338 			segs_per_sec > segment_count || !segs_per_sec) {
3339 		f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3340 			  segment_count, total_sections, segs_per_sec);
3341 		return -EFSCORRUPTED;
3342 	}
3343 
3344 	if (segment_count_main != total_sections * segs_per_sec) {
3345 		f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3346 			  segment_count_main, total_sections, segs_per_sec);
3347 		return -EFSCORRUPTED;
3348 	}
3349 
3350 	if ((segment_count / segs_per_sec) < total_sections) {
3351 		f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3352 			  segment_count, segs_per_sec, total_sections);
3353 		return -EFSCORRUPTED;
3354 	}
3355 
3356 	if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3357 		f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3358 			  segment_count, le64_to_cpu(raw_super->block_count));
3359 		return -EFSCORRUPTED;
3360 	}
3361 
3362 	if (RDEV(0).path[0]) {
3363 		block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3364 		int i = 1;
3365 
3366 		while (i < MAX_DEVICES && RDEV(i).path[0]) {
3367 			dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3368 			i++;
3369 		}
3370 		if (segment_count != dev_seg_count) {
3371 			f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3372 					segment_count, dev_seg_count);
3373 			return -EFSCORRUPTED;
3374 		}
3375 	} else {
3376 		if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3377 					!bdev_is_zoned(sbi->sb->s_bdev)) {
3378 			f2fs_info(sbi, "Zoned block device path is missing");
3379 			return -EFSCORRUPTED;
3380 		}
3381 	}
3382 
3383 	if (secs_per_zone > total_sections || !secs_per_zone) {
3384 		f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3385 			  secs_per_zone, total_sections);
3386 		return -EFSCORRUPTED;
3387 	}
3388 	if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3389 			raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3390 			(le32_to_cpu(raw_super->extension_count) +
3391 			raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3392 		f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3393 			  le32_to_cpu(raw_super->extension_count),
3394 			  raw_super->hot_ext_count,
3395 			  F2FS_MAX_EXTENSION);
3396 		return -EFSCORRUPTED;
3397 	}
3398 
3399 	if (le32_to_cpu(raw_super->cp_payload) >=
3400 				(blocks_per_seg - F2FS_CP_PACKS -
3401 				NR_CURSEG_PERSIST_TYPE)) {
3402 		f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3403 			  le32_to_cpu(raw_super->cp_payload),
3404 			  blocks_per_seg - F2FS_CP_PACKS -
3405 			  NR_CURSEG_PERSIST_TYPE);
3406 		return -EFSCORRUPTED;
3407 	}
3408 
3409 	/* check reserved ino info */
3410 	if (le32_to_cpu(raw_super->node_ino) != 1 ||
3411 		le32_to_cpu(raw_super->meta_ino) != 2 ||
3412 		le32_to_cpu(raw_super->root_ino) != 3) {
3413 		f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3414 			  le32_to_cpu(raw_super->node_ino),
3415 			  le32_to_cpu(raw_super->meta_ino),
3416 			  le32_to_cpu(raw_super->root_ino));
3417 		return -EFSCORRUPTED;
3418 	}
3419 
3420 	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3421 	if (sanity_check_area_boundary(sbi, bh))
3422 		return -EFSCORRUPTED;
3423 
3424 	return 0;
3425 }
3426 
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3427 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3428 {
3429 	unsigned int total, fsmeta;
3430 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3431 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3432 	unsigned int ovp_segments, reserved_segments;
3433 	unsigned int main_segs, blocks_per_seg;
3434 	unsigned int sit_segs, nat_segs;
3435 	unsigned int sit_bitmap_size, nat_bitmap_size;
3436 	unsigned int log_blocks_per_seg;
3437 	unsigned int segment_count_main;
3438 	unsigned int cp_pack_start_sum, cp_payload;
3439 	block_t user_block_count, valid_user_blocks;
3440 	block_t avail_node_count, valid_node_count;
3441 	unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3442 	int i, j;
3443 
3444 	total = le32_to_cpu(raw_super->segment_count);
3445 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3446 	sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3447 	fsmeta += sit_segs;
3448 	nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3449 	fsmeta += nat_segs;
3450 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3451 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3452 
3453 	if (unlikely(fsmeta >= total))
3454 		return 1;
3455 
3456 	ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3457 	reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3458 
3459 	if (!f2fs_sb_has_readonly(sbi) &&
3460 			unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3461 			ovp_segments == 0 || reserved_segments == 0)) {
3462 		f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3463 		return 1;
3464 	}
3465 	user_block_count = le64_to_cpu(ckpt->user_block_count);
3466 	segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3467 			(f2fs_sb_has_readonly(sbi) ? 1 : 0);
3468 	log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3469 	if (!user_block_count || user_block_count >=
3470 			segment_count_main << log_blocks_per_seg) {
3471 		f2fs_err(sbi, "Wrong user_block_count: %u",
3472 			 user_block_count);
3473 		return 1;
3474 	}
3475 
3476 	valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3477 	if (valid_user_blocks > user_block_count) {
3478 		f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3479 			 valid_user_blocks, user_block_count);
3480 		return 1;
3481 	}
3482 
3483 	valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3484 	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3485 	if (valid_node_count > avail_node_count) {
3486 		f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3487 			 valid_node_count, avail_node_count);
3488 		return 1;
3489 	}
3490 
3491 	main_segs = le32_to_cpu(raw_super->segment_count_main);
3492 	blocks_per_seg = sbi->blocks_per_seg;
3493 
3494 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3495 		if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3496 			le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3497 			return 1;
3498 
3499 		if (f2fs_sb_has_readonly(sbi))
3500 			goto check_data;
3501 
3502 		for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3503 			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3504 				le32_to_cpu(ckpt->cur_node_segno[j])) {
3505 				f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3506 					 i, j,
3507 					 le32_to_cpu(ckpt->cur_node_segno[i]));
3508 				return 1;
3509 			}
3510 		}
3511 	}
3512 check_data:
3513 	for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3514 		if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3515 			le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3516 			return 1;
3517 
3518 		if (f2fs_sb_has_readonly(sbi))
3519 			goto skip_cross;
3520 
3521 		for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3522 			if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3523 				le32_to_cpu(ckpt->cur_data_segno[j])) {
3524 				f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3525 					 i, j,
3526 					 le32_to_cpu(ckpt->cur_data_segno[i]));
3527 				return 1;
3528 			}
3529 		}
3530 	}
3531 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3532 		for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3533 			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3534 				le32_to_cpu(ckpt->cur_data_segno[j])) {
3535 				f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3536 					 i, j,
3537 					 le32_to_cpu(ckpt->cur_node_segno[i]));
3538 				return 1;
3539 			}
3540 		}
3541 	}
3542 skip_cross:
3543 	sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3544 	nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3545 
3546 	if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3547 		nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3548 		f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3549 			 sit_bitmap_size, nat_bitmap_size);
3550 		return 1;
3551 	}
3552 
3553 	cp_pack_start_sum = __start_sum_addr(sbi);
3554 	cp_payload = __cp_payload(sbi);
3555 	if (cp_pack_start_sum < cp_payload + 1 ||
3556 		cp_pack_start_sum > blocks_per_seg - 1 -
3557 			NR_CURSEG_PERSIST_TYPE) {
3558 		f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3559 			 cp_pack_start_sum);
3560 		return 1;
3561 	}
3562 
3563 	if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3564 		le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3565 		f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3566 			  "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3567 			  "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3568 			  le32_to_cpu(ckpt->checksum_offset));
3569 		return 1;
3570 	}
3571 
3572 	nat_blocks = nat_segs << log_blocks_per_seg;
3573 	nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3574 	nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3575 	if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3576 		(cp_payload + F2FS_CP_PACKS +
3577 		NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3578 		f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3579 			  cp_payload, nat_bits_blocks);
3580 		return 1;
3581 	}
3582 
3583 	if (unlikely(f2fs_cp_error(sbi))) {
3584 		f2fs_err(sbi, "A bug case: need to run fsck");
3585 		return 1;
3586 	}
3587 	return 0;
3588 }
3589 
init_sb_info(struct f2fs_sb_info * sbi)3590 static void init_sb_info(struct f2fs_sb_info *sbi)
3591 {
3592 	struct f2fs_super_block *raw_super = sbi->raw_super;
3593 	int i;
3594 
3595 	sbi->log_sectors_per_block =
3596 		le32_to_cpu(raw_super->log_sectors_per_block);
3597 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3598 	sbi->blocksize = 1 << sbi->log_blocksize;
3599 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3600 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3601 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3602 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3603 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
3604 	sbi->total_node_count =
3605 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
3606 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3607 	F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3608 	F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3609 	F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3610 	sbi->cur_victim_sec = NULL_SECNO;
3611 	sbi->gc_mode = GC_NORMAL;
3612 	sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3613 	sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3614 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3615 	sbi->migration_granularity = sbi->segs_per_sec;
3616 	sbi->seq_file_ra_mul = MIN_RA_MUL;
3617 	sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3618 	sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3619 	spin_lock_init(&sbi->gc_urgent_high_lock);
3620 	atomic64_set(&sbi->current_atomic_write, 0);
3621 
3622 	sbi->dir_level = DEF_DIR_LEVEL;
3623 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3624 	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3625 	sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3626 	sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3627 	sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3628 	sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3629 				DEF_UMOUNT_DISCARD_TIMEOUT;
3630 	clear_sbi_flag(sbi, SBI_NEED_FSCK);
3631 
3632 	for (i = 0; i < NR_COUNT_TYPE; i++)
3633 		atomic_set(&sbi->nr_pages[i], 0);
3634 
3635 	for (i = 0; i < META; i++)
3636 		atomic_set(&sbi->wb_sync_req[i], 0);
3637 
3638 	INIT_LIST_HEAD(&sbi->s_list);
3639 	mutex_init(&sbi->umount_mutex);
3640 	init_f2fs_rwsem(&sbi->io_order_lock);
3641 	spin_lock_init(&sbi->cp_lock);
3642 
3643 	sbi->dirty_device = 0;
3644 	spin_lock_init(&sbi->dev_lock);
3645 
3646 	init_f2fs_rwsem(&sbi->sb_lock);
3647 	init_f2fs_rwsem(&sbi->pin_sem);
3648 }
3649 
init_percpu_info(struct f2fs_sb_info * sbi)3650 static int init_percpu_info(struct f2fs_sb_info *sbi)
3651 {
3652 	int err;
3653 
3654 	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3655 	if (err)
3656 		return err;
3657 
3658 	err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3659 	if (err)
3660 		goto err_valid_block;
3661 
3662 	err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3663 								GFP_KERNEL);
3664 	if (err)
3665 		goto err_node_block;
3666 	return 0;
3667 
3668 err_node_block:
3669 	percpu_counter_destroy(&sbi->rf_node_block_count);
3670 err_valid_block:
3671 	percpu_counter_destroy(&sbi->alloc_valid_block_count);
3672 	return err;
3673 }
3674 
3675 #ifdef CONFIG_BLK_DEV_ZONED
3676 
3677 struct f2fs_report_zones_args {
3678 	struct f2fs_sb_info *sbi;
3679 	struct f2fs_dev_info *dev;
3680 };
3681 
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3682 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3683 			      void *data)
3684 {
3685 	struct f2fs_report_zones_args *rz_args = data;
3686 	block_t unusable_blocks = (zone->len - zone->capacity) >>
3687 					F2FS_LOG_SECTORS_PER_BLOCK;
3688 
3689 	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3690 		return 0;
3691 
3692 	set_bit(idx, rz_args->dev->blkz_seq);
3693 	if (!rz_args->sbi->unusable_blocks_per_sec) {
3694 		rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3695 		return 0;
3696 	}
3697 	if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3698 		f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3699 		return -EINVAL;
3700 	}
3701 	return 0;
3702 }
3703 
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3704 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3705 {
3706 	struct block_device *bdev = FDEV(devi).bdev;
3707 	sector_t nr_sectors = bdev_nr_sectors(bdev);
3708 	struct f2fs_report_zones_args rep_zone_arg;
3709 	u64 zone_sectors;
3710 	int ret;
3711 
3712 	if (!f2fs_sb_has_blkzoned(sbi))
3713 		return 0;
3714 
3715 	zone_sectors = bdev_zone_sectors(bdev);
3716 	if (!is_power_of_2(zone_sectors)) {
3717 		f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n");
3718 		return -EINVAL;
3719 	}
3720 
3721 	if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3722 				SECTOR_TO_BLOCK(zone_sectors))
3723 		return -EINVAL;
3724 	sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3725 	if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3726 				__ilog2_u32(sbi->blocks_per_blkz))
3727 		return -EINVAL;
3728 	sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3729 	FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3730 					sbi->log_blocks_per_blkz;
3731 	if (nr_sectors & (zone_sectors - 1))
3732 		FDEV(devi).nr_blkz++;
3733 
3734 	FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3735 					BITS_TO_LONGS(FDEV(devi).nr_blkz)
3736 					* sizeof(unsigned long),
3737 					GFP_KERNEL);
3738 	if (!FDEV(devi).blkz_seq)
3739 		return -ENOMEM;
3740 
3741 	rep_zone_arg.sbi = sbi;
3742 	rep_zone_arg.dev = &FDEV(devi);
3743 
3744 	ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3745 				  &rep_zone_arg);
3746 	if (ret < 0)
3747 		return ret;
3748 	return 0;
3749 }
3750 #endif
3751 
3752 /*
3753  * Read f2fs raw super block.
3754  * Because we have two copies of super block, so read both of them
3755  * to get the first valid one. If any one of them is broken, we pass
3756  * them recovery flag back to the caller.
3757  */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3758 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3759 			struct f2fs_super_block **raw_super,
3760 			int *valid_super_block, int *recovery)
3761 {
3762 	struct super_block *sb = sbi->sb;
3763 	int block;
3764 	struct buffer_head *bh;
3765 	struct f2fs_super_block *super;
3766 	int err = 0;
3767 
3768 	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3769 	if (!super)
3770 		return -ENOMEM;
3771 
3772 	for (block = 0; block < 2; block++) {
3773 		bh = sb_bread(sb, block);
3774 		if (!bh) {
3775 			f2fs_err(sbi, "Unable to read %dth superblock",
3776 				 block + 1);
3777 			err = -EIO;
3778 			*recovery = 1;
3779 			continue;
3780 		}
3781 
3782 		/* sanity checking of raw super */
3783 		err = sanity_check_raw_super(sbi, bh);
3784 		if (err) {
3785 			f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3786 				 block + 1);
3787 			brelse(bh);
3788 			*recovery = 1;
3789 			continue;
3790 		}
3791 
3792 		if (!*raw_super) {
3793 			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3794 							sizeof(*super));
3795 			*valid_super_block = block;
3796 			*raw_super = super;
3797 		}
3798 		brelse(bh);
3799 	}
3800 
3801 	/* No valid superblock */
3802 	if (!*raw_super)
3803 		kfree(super);
3804 	else
3805 		err = 0;
3806 
3807 	return err;
3808 }
3809 
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)3810 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3811 {
3812 	struct buffer_head *bh;
3813 	__u32 crc = 0;
3814 	int err;
3815 
3816 	if ((recover && f2fs_readonly(sbi->sb)) ||
3817 				bdev_read_only(sbi->sb->s_bdev)) {
3818 		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3819 		return -EROFS;
3820 	}
3821 
3822 	/* we should update superblock crc here */
3823 	if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3824 		crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3825 				offsetof(struct f2fs_super_block, crc));
3826 		F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3827 	}
3828 
3829 	/* write back-up superblock first */
3830 	bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3831 	if (!bh)
3832 		return -EIO;
3833 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3834 	brelse(bh);
3835 
3836 	/* if we are in recovery path, skip writing valid superblock */
3837 	if (recover || err)
3838 		return err;
3839 
3840 	/* write current valid superblock */
3841 	bh = sb_bread(sbi->sb, sbi->valid_super_block);
3842 	if (!bh)
3843 		return -EIO;
3844 	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3845 	brelse(bh);
3846 	return err;
3847 }
3848 
f2fs_handle_stop(struct f2fs_sb_info * sbi,unsigned char reason)3849 void f2fs_handle_stop(struct f2fs_sb_info *sbi, unsigned char reason)
3850 {
3851 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3852 	int err;
3853 
3854 	f2fs_down_write(&sbi->sb_lock);
3855 
3856 	if (raw_super->s_stop_reason[reason] < ((1 << BITS_PER_BYTE) - 1))
3857 		raw_super->s_stop_reason[reason]++;
3858 
3859 	err = f2fs_commit_super(sbi, false);
3860 	if (err)
3861 		f2fs_err(sbi, "f2fs_commit_super fails to record reason:%u err:%d",
3862 								reason, err);
3863 	f2fs_up_write(&sbi->sb_lock);
3864 }
3865 
f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)3866 static void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
3867 {
3868 	spin_lock(&sbi->error_lock);
3869 	if (!test_bit(flag, (unsigned long *)sbi->errors)) {
3870 		set_bit(flag, (unsigned long *)sbi->errors);
3871 		sbi->error_dirty = true;
3872 	}
3873 	spin_unlock(&sbi->error_lock);
3874 }
3875 
f2fs_update_errors(struct f2fs_sb_info * sbi)3876 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
3877 {
3878 	bool need_update = false;
3879 
3880 	spin_lock(&sbi->error_lock);
3881 	if (sbi->error_dirty) {
3882 		memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
3883 							MAX_F2FS_ERRORS);
3884 		sbi->error_dirty = false;
3885 		need_update = true;
3886 	}
3887 	spin_unlock(&sbi->error_lock);
3888 
3889 	return need_update;
3890 }
3891 
f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)3892 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
3893 {
3894 	int err;
3895 
3896 	f2fs_save_errors(sbi, error);
3897 
3898 	f2fs_down_write(&sbi->sb_lock);
3899 
3900 	if (!f2fs_update_errors(sbi))
3901 		goto out_unlock;
3902 
3903 	err = f2fs_commit_super(sbi, false);
3904 	if (err)
3905 		f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
3906 								error, err);
3907 out_unlock:
3908 	f2fs_up_write(&sbi->sb_lock);
3909 }
3910 
f2fs_scan_devices(struct f2fs_sb_info * sbi)3911 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3912 {
3913 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3914 	unsigned int max_devices = MAX_DEVICES;
3915 	unsigned int logical_blksize;
3916 	int i;
3917 
3918 	/* Initialize single device information */
3919 	if (!RDEV(0).path[0]) {
3920 		if (!bdev_is_zoned(sbi->sb->s_bdev))
3921 			return 0;
3922 		max_devices = 1;
3923 	}
3924 
3925 	/*
3926 	 * Initialize multiple devices information, or single
3927 	 * zoned block device information.
3928 	 */
3929 	sbi->devs = f2fs_kzalloc(sbi,
3930 				 array_size(max_devices,
3931 					    sizeof(struct f2fs_dev_info)),
3932 				 GFP_KERNEL);
3933 	if (!sbi->devs)
3934 		return -ENOMEM;
3935 
3936 	logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3937 	sbi->aligned_blksize = true;
3938 
3939 	for (i = 0; i < max_devices; i++) {
3940 
3941 		if (i > 0 && !RDEV(i).path[0])
3942 			break;
3943 
3944 		if (max_devices == 1) {
3945 			/* Single zoned block device mount */
3946 			FDEV(0).bdev =
3947 				blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3948 					sbi->sb->s_mode, sbi->sb->s_type);
3949 		} else {
3950 			/* Multi-device mount */
3951 			memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3952 			FDEV(i).total_segments =
3953 				le32_to_cpu(RDEV(i).total_segments);
3954 			if (i == 0) {
3955 				FDEV(i).start_blk = 0;
3956 				FDEV(i).end_blk = FDEV(i).start_blk +
3957 				    (FDEV(i).total_segments <<
3958 				    sbi->log_blocks_per_seg) - 1 +
3959 				    le32_to_cpu(raw_super->segment0_blkaddr);
3960 			} else {
3961 				FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3962 				FDEV(i).end_blk = FDEV(i).start_blk +
3963 					(FDEV(i).total_segments <<
3964 					sbi->log_blocks_per_seg) - 1;
3965 			}
3966 			FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3967 					sbi->sb->s_mode, sbi->sb->s_type);
3968 		}
3969 		if (IS_ERR(FDEV(i).bdev))
3970 			return PTR_ERR(FDEV(i).bdev);
3971 
3972 		/* to release errored devices */
3973 		sbi->s_ndevs = i + 1;
3974 
3975 		if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
3976 			sbi->aligned_blksize = false;
3977 
3978 #ifdef CONFIG_BLK_DEV_ZONED
3979 		if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3980 				!f2fs_sb_has_blkzoned(sbi)) {
3981 			f2fs_err(sbi, "Zoned block device feature not enabled");
3982 			return -EINVAL;
3983 		}
3984 		if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3985 			if (init_blkz_info(sbi, i)) {
3986 				f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3987 				return -EINVAL;
3988 			}
3989 			if (max_devices == 1)
3990 				break;
3991 			f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3992 				  i, FDEV(i).path,
3993 				  FDEV(i).total_segments,
3994 				  FDEV(i).start_blk, FDEV(i).end_blk,
3995 				  bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3996 				  "Host-aware" : "Host-managed");
3997 			continue;
3998 		}
3999 #endif
4000 		f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4001 			  i, FDEV(i).path,
4002 			  FDEV(i).total_segments,
4003 			  FDEV(i).start_blk, FDEV(i).end_blk);
4004 	}
4005 	f2fs_info(sbi,
4006 		  "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
4007 	return 0;
4008 }
4009 
f2fs_setup_casefold(struct f2fs_sb_info * sbi)4010 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4011 {
4012 #if IS_ENABLED(CONFIG_UNICODE)
4013 	if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4014 		const struct f2fs_sb_encodings *encoding_info;
4015 		struct unicode_map *encoding;
4016 		__u16 encoding_flags;
4017 
4018 		encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4019 		if (!encoding_info) {
4020 			f2fs_err(sbi,
4021 				 "Encoding requested by superblock is unknown");
4022 			return -EINVAL;
4023 		}
4024 
4025 		encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4026 		encoding = utf8_load(encoding_info->version);
4027 		if (IS_ERR(encoding)) {
4028 			f2fs_err(sbi,
4029 				 "can't mount with superblock charset: %s-%u.%u.%u "
4030 				 "not supported by the kernel. flags: 0x%x.",
4031 				 encoding_info->name,
4032 				 unicode_major(encoding_info->version),
4033 				 unicode_minor(encoding_info->version),
4034 				 unicode_rev(encoding_info->version),
4035 				 encoding_flags);
4036 			return PTR_ERR(encoding);
4037 		}
4038 		f2fs_info(sbi, "Using encoding defined by superblock: "
4039 			 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4040 			 unicode_major(encoding_info->version),
4041 			 unicode_minor(encoding_info->version),
4042 			 unicode_rev(encoding_info->version),
4043 			 encoding_flags);
4044 
4045 		sbi->sb->s_encoding = encoding;
4046 		sbi->sb->s_encoding_flags = encoding_flags;
4047 	}
4048 #else
4049 	if (f2fs_sb_has_casefold(sbi)) {
4050 		f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4051 		return -EINVAL;
4052 	}
4053 #endif
4054 	return 0;
4055 }
4056 
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4057 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4058 {
4059 	struct f2fs_sm_info *sm_i = SM_I(sbi);
4060 
4061 	/* adjust parameters according to the volume size */
4062 	if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
4063 		F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
4064 		if (f2fs_block_unit_discard(sbi))
4065 			sm_i->dcc_info->discard_granularity = 1;
4066 		sm_i->ipu_policy = 1 << F2FS_IPU_FORCE |
4067 					1 << F2FS_IPU_HONOR_OPU_WRITE;
4068 	}
4069 
4070 	sbi->readdir_ra = 1;
4071 }
4072 
f2fs_fill_super(struct super_block * sb,void * data,int silent)4073 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4074 {
4075 	struct f2fs_sb_info *sbi;
4076 	struct f2fs_super_block *raw_super;
4077 	struct inode *root;
4078 	int err;
4079 	bool skip_recovery = false, need_fsck = false;
4080 	char *options = NULL;
4081 	int recovery, i, valid_super_block;
4082 	struct curseg_info *seg_i;
4083 	int retry_cnt = 1;
4084 
4085 try_onemore:
4086 	err = -EINVAL;
4087 	raw_super = NULL;
4088 	valid_super_block = -1;
4089 	recovery = 0;
4090 
4091 	/* allocate memory for f2fs-specific super block info */
4092 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4093 	if (!sbi)
4094 		return -ENOMEM;
4095 
4096 	sbi->sb = sb;
4097 
4098 	/* Load the checksum driver */
4099 	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4100 	if (IS_ERR(sbi->s_chksum_driver)) {
4101 		f2fs_err(sbi, "Cannot load crc32 driver.");
4102 		err = PTR_ERR(sbi->s_chksum_driver);
4103 		sbi->s_chksum_driver = NULL;
4104 		goto free_sbi;
4105 	}
4106 
4107 	/* set a block size */
4108 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4109 		f2fs_err(sbi, "unable to set blocksize");
4110 		goto free_sbi;
4111 	}
4112 
4113 	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4114 								&recovery);
4115 	if (err)
4116 		goto free_sbi;
4117 
4118 	sb->s_fs_info = sbi;
4119 	sbi->raw_super = raw_super;
4120 
4121 	/* precompute checksum seed for metadata */
4122 	if (f2fs_sb_has_inode_chksum(sbi))
4123 		sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4124 						sizeof(raw_super->uuid));
4125 
4126 	default_options(sbi);
4127 	/* parse mount options */
4128 	options = kstrdup((const char *)data, GFP_KERNEL);
4129 	if (data && !options) {
4130 		err = -ENOMEM;
4131 		goto free_sb_buf;
4132 	}
4133 
4134 	err = parse_options(sb, options, false);
4135 	if (err)
4136 		goto free_options;
4137 
4138 	sb->s_maxbytes = max_file_blocks(NULL) <<
4139 				le32_to_cpu(raw_super->log_blocksize);
4140 	sb->s_max_links = F2FS_LINK_MAX;
4141 
4142 	err = f2fs_setup_casefold(sbi);
4143 	if (err)
4144 		goto free_options;
4145 
4146 #ifdef CONFIG_QUOTA
4147 	sb->dq_op = &f2fs_quota_operations;
4148 	sb->s_qcop = &f2fs_quotactl_ops;
4149 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4150 
4151 	if (f2fs_sb_has_quota_ino(sbi)) {
4152 		for (i = 0; i < MAXQUOTAS; i++) {
4153 			if (f2fs_qf_ino(sbi->sb, i))
4154 				sbi->nquota_files++;
4155 		}
4156 	}
4157 #endif
4158 
4159 	sb->s_op = &f2fs_sops;
4160 #ifdef CONFIG_FS_ENCRYPTION
4161 	sb->s_cop = &f2fs_cryptops;
4162 #endif
4163 #ifdef CONFIG_FS_VERITY
4164 	sb->s_vop = &f2fs_verityops;
4165 #endif
4166 	sb->s_xattr = f2fs_xattr_handlers;
4167 	sb->s_export_op = &f2fs_export_ops;
4168 	sb->s_magic = F2FS_SUPER_MAGIC;
4169 	sb->s_time_gran = 1;
4170 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4171 		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4172 	memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4173 	sb->s_iflags |= SB_I_CGROUPWB;
4174 
4175 	/* init f2fs-specific super block info */
4176 	sbi->valid_super_block = valid_super_block;
4177 	init_f2fs_rwsem(&sbi->gc_lock);
4178 	mutex_init(&sbi->writepages);
4179 	init_f2fs_rwsem(&sbi->cp_global_sem);
4180 	init_f2fs_rwsem(&sbi->node_write);
4181 	init_f2fs_rwsem(&sbi->node_change);
4182 
4183 	/* disallow all the data/node/meta page writes */
4184 	set_sbi_flag(sbi, SBI_POR_DOING);
4185 	spin_lock_init(&sbi->stat_lock);
4186 
4187 	err = f2fs_init_write_merge_io(sbi);
4188 	if (err)
4189 		goto free_bio_info;
4190 
4191 	spin_lock_init(&sbi->error_lock);
4192 	memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4193 
4194 	init_f2fs_rwsem(&sbi->cp_rwsem);
4195 	init_f2fs_rwsem(&sbi->quota_sem);
4196 	init_waitqueue_head(&sbi->cp_wait);
4197 	init_sb_info(sbi);
4198 
4199 	err = f2fs_init_iostat(sbi);
4200 	if (err)
4201 		goto free_bio_info;
4202 
4203 	err = init_percpu_info(sbi);
4204 	if (err)
4205 		goto free_iostat;
4206 
4207 	if (F2FS_IO_ALIGNED(sbi)) {
4208 		sbi->write_io_dummy =
4209 			mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4210 		if (!sbi->write_io_dummy) {
4211 			err = -ENOMEM;
4212 			goto free_percpu;
4213 		}
4214 	}
4215 
4216 	/* init per sbi slab cache */
4217 	err = f2fs_init_xattr_caches(sbi);
4218 	if (err)
4219 		goto free_io_dummy;
4220 	err = f2fs_init_page_array_cache(sbi);
4221 	if (err)
4222 		goto free_xattr_cache;
4223 
4224 	/* get an inode for meta space */
4225 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4226 	if (IS_ERR(sbi->meta_inode)) {
4227 		f2fs_err(sbi, "Failed to read F2FS meta data inode");
4228 		err = PTR_ERR(sbi->meta_inode);
4229 		goto free_page_array_cache;
4230 	}
4231 
4232 	err = f2fs_get_valid_checkpoint(sbi);
4233 	if (err) {
4234 		f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4235 		goto free_meta_inode;
4236 	}
4237 
4238 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4239 		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4240 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4241 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4242 		sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4243 	}
4244 
4245 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4246 		set_sbi_flag(sbi, SBI_NEED_FSCK);
4247 
4248 	/* Initialize device list */
4249 	err = f2fs_scan_devices(sbi);
4250 	if (err) {
4251 		f2fs_err(sbi, "Failed to find devices");
4252 		goto free_devices;
4253 	}
4254 
4255 	err = f2fs_init_post_read_wq(sbi);
4256 	if (err) {
4257 		f2fs_err(sbi, "Failed to initialize post read workqueue");
4258 		goto free_devices;
4259 	}
4260 
4261 	sbi->total_valid_node_count =
4262 				le32_to_cpu(sbi->ckpt->valid_node_count);
4263 	percpu_counter_set(&sbi->total_valid_inode_count,
4264 				le32_to_cpu(sbi->ckpt->valid_inode_count));
4265 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4266 	sbi->total_valid_block_count =
4267 				le64_to_cpu(sbi->ckpt->valid_block_count);
4268 	sbi->last_valid_block_count = sbi->total_valid_block_count;
4269 	sbi->reserved_blocks = 0;
4270 	sbi->current_reserved_blocks = 0;
4271 	limit_reserve_root(sbi);
4272 	adjust_unusable_cap_perc(sbi);
4273 
4274 	for (i = 0; i < NR_INODE_TYPE; i++) {
4275 		INIT_LIST_HEAD(&sbi->inode_list[i]);
4276 		spin_lock_init(&sbi->inode_lock[i]);
4277 	}
4278 	mutex_init(&sbi->flush_lock);
4279 
4280 	f2fs_init_extent_cache_info(sbi);
4281 
4282 	f2fs_init_ino_entry_info(sbi);
4283 
4284 	f2fs_init_fsync_node_info(sbi);
4285 
4286 	/* setup checkpoint request control and start checkpoint issue thread */
4287 	f2fs_init_ckpt_req_control(sbi);
4288 	if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4289 			test_opt(sbi, MERGE_CHECKPOINT)) {
4290 		err = f2fs_start_ckpt_thread(sbi);
4291 		if (err) {
4292 			f2fs_err(sbi,
4293 			    "Failed to start F2FS issue_checkpoint_thread (%d)",
4294 			    err);
4295 			goto stop_ckpt_thread;
4296 		}
4297 	}
4298 
4299 	/* setup f2fs internal modules */
4300 	err = f2fs_build_segment_manager(sbi);
4301 	if (err) {
4302 		f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4303 			 err);
4304 		goto free_sm;
4305 	}
4306 	err = f2fs_build_node_manager(sbi);
4307 	if (err) {
4308 		f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4309 			 err);
4310 		goto free_nm;
4311 	}
4312 
4313 	err = adjust_reserved_segment(sbi);
4314 	if (err)
4315 		goto free_nm;
4316 
4317 	/* For write statistics */
4318 	sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4319 
4320 	/* Read accumulated write IO statistics if exists */
4321 	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4322 	if (__exist_node_summaries(sbi))
4323 		sbi->kbytes_written =
4324 			le64_to_cpu(seg_i->journal->info.kbytes_written);
4325 
4326 	f2fs_build_gc_manager(sbi);
4327 
4328 	err = f2fs_build_stats(sbi);
4329 	if (err)
4330 		goto free_nm;
4331 
4332 	/* get an inode for node space */
4333 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4334 	if (IS_ERR(sbi->node_inode)) {
4335 		f2fs_err(sbi, "Failed to read node inode");
4336 		err = PTR_ERR(sbi->node_inode);
4337 		goto free_stats;
4338 	}
4339 
4340 	/* read root inode and dentry */
4341 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4342 	if (IS_ERR(root)) {
4343 		f2fs_err(sbi, "Failed to read root inode");
4344 		err = PTR_ERR(root);
4345 		goto free_node_inode;
4346 	}
4347 	if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4348 			!root->i_size || !root->i_nlink) {
4349 		iput(root);
4350 		err = -EINVAL;
4351 		goto free_node_inode;
4352 	}
4353 
4354 	sb->s_root = d_make_root(root); /* allocate root dentry */
4355 	if (!sb->s_root) {
4356 		err = -ENOMEM;
4357 		goto free_node_inode;
4358 	}
4359 
4360 	err = f2fs_init_compress_inode(sbi);
4361 	if (err)
4362 		goto free_root_inode;
4363 
4364 	err = f2fs_register_sysfs(sbi);
4365 	if (err)
4366 		goto free_compress_inode;
4367 
4368 #ifdef CONFIG_QUOTA
4369 	/* Enable quota usage during mount */
4370 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4371 		err = f2fs_enable_quotas(sb);
4372 		if (err)
4373 			f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4374 	}
4375 #endif
4376 	/* if there are any orphan inodes, free them */
4377 	err = f2fs_recover_orphan_inodes(sbi);
4378 	if (err)
4379 		goto free_meta;
4380 
4381 	if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4382 		goto reset_checkpoint;
4383 
4384 	/* recover fsynced data */
4385 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4386 			!test_opt(sbi, NORECOVERY)) {
4387 		/*
4388 		 * mount should be failed, when device has readonly mode, and
4389 		 * previous checkpoint was not done by clean system shutdown.
4390 		 */
4391 		if (f2fs_hw_is_readonly(sbi)) {
4392 			if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4393 				err = f2fs_recover_fsync_data(sbi, true);
4394 				if (err > 0) {
4395 					err = -EROFS;
4396 					f2fs_err(sbi, "Need to recover fsync data, but "
4397 						"write access unavailable, please try "
4398 						"mount w/ disable_roll_forward or norecovery");
4399 				}
4400 				if (err < 0)
4401 					goto free_meta;
4402 			}
4403 			f2fs_info(sbi, "write access unavailable, skipping recovery");
4404 			goto reset_checkpoint;
4405 		}
4406 
4407 		if (need_fsck)
4408 			set_sbi_flag(sbi, SBI_NEED_FSCK);
4409 
4410 		if (skip_recovery)
4411 			goto reset_checkpoint;
4412 
4413 		err = f2fs_recover_fsync_data(sbi, false);
4414 		if (err < 0) {
4415 			if (err != -ENOMEM)
4416 				skip_recovery = true;
4417 			need_fsck = true;
4418 			f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4419 				 err);
4420 			goto free_meta;
4421 		}
4422 	} else {
4423 		err = f2fs_recover_fsync_data(sbi, true);
4424 
4425 		if (!f2fs_readonly(sb) && err > 0) {
4426 			err = -EINVAL;
4427 			f2fs_err(sbi, "Need to recover fsync data");
4428 			goto free_meta;
4429 		}
4430 	}
4431 
4432 	/*
4433 	 * If the f2fs is not readonly and fsync data recovery succeeds,
4434 	 * check zoned block devices' write pointer consistency.
4435 	 */
4436 	if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4437 		err = f2fs_check_write_pointer(sbi);
4438 		if (err)
4439 			goto free_meta;
4440 	}
4441 
4442 reset_checkpoint:
4443 	f2fs_init_inmem_curseg(sbi);
4444 
4445 	/* f2fs_recover_fsync_data() cleared this already */
4446 	clear_sbi_flag(sbi, SBI_POR_DOING);
4447 
4448 	if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4449 		err = f2fs_disable_checkpoint(sbi);
4450 		if (err)
4451 			goto sync_free_meta;
4452 	} else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4453 		f2fs_enable_checkpoint(sbi);
4454 	}
4455 
4456 	/*
4457 	 * If filesystem is not mounted as read-only then
4458 	 * do start the gc_thread.
4459 	 */
4460 	if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4461 		test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4462 		/* After POR, we can run background GC thread.*/
4463 		err = f2fs_start_gc_thread(sbi);
4464 		if (err)
4465 			goto sync_free_meta;
4466 	}
4467 	kvfree(options);
4468 
4469 	/* recover broken superblock */
4470 	if (recovery) {
4471 		err = f2fs_commit_super(sbi, true);
4472 		f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4473 			  sbi->valid_super_block ? 1 : 2, err);
4474 	}
4475 
4476 	f2fs_join_shrinker(sbi);
4477 
4478 	f2fs_tuning_parameters(sbi);
4479 
4480 	f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4481 		    cur_cp_version(F2FS_CKPT(sbi)));
4482 	f2fs_update_time(sbi, CP_TIME);
4483 	f2fs_update_time(sbi, REQ_TIME);
4484 	clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4485 	return 0;
4486 
4487 sync_free_meta:
4488 	/* safe to flush all the data */
4489 	sync_filesystem(sbi->sb);
4490 	retry_cnt = 0;
4491 
4492 free_meta:
4493 #ifdef CONFIG_QUOTA
4494 	f2fs_truncate_quota_inode_pages(sb);
4495 	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4496 		f2fs_quota_off_umount(sbi->sb);
4497 #endif
4498 	/*
4499 	 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4500 	 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4501 	 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4502 	 * falls into an infinite loop in f2fs_sync_meta_pages().
4503 	 */
4504 	truncate_inode_pages_final(META_MAPPING(sbi));
4505 	/* evict some inodes being cached by GC */
4506 	evict_inodes(sb);
4507 	f2fs_unregister_sysfs(sbi);
4508 free_compress_inode:
4509 	f2fs_destroy_compress_inode(sbi);
4510 free_root_inode:
4511 	dput(sb->s_root);
4512 	sb->s_root = NULL;
4513 free_node_inode:
4514 	f2fs_release_ino_entry(sbi, true);
4515 	truncate_inode_pages_final(NODE_MAPPING(sbi));
4516 	iput(sbi->node_inode);
4517 	sbi->node_inode = NULL;
4518 free_stats:
4519 	f2fs_destroy_stats(sbi);
4520 free_nm:
4521 	/* stop discard thread before destroying node manager */
4522 	f2fs_stop_discard_thread(sbi);
4523 	f2fs_destroy_node_manager(sbi);
4524 free_sm:
4525 	f2fs_destroy_segment_manager(sbi);
4526 stop_ckpt_thread:
4527 	f2fs_stop_ckpt_thread(sbi);
4528 	f2fs_destroy_post_read_wq(sbi);
4529 free_devices:
4530 	destroy_device_list(sbi);
4531 	kvfree(sbi->ckpt);
4532 free_meta_inode:
4533 	make_bad_inode(sbi->meta_inode);
4534 	iput(sbi->meta_inode);
4535 	sbi->meta_inode = NULL;
4536 free_page_array_cache:
4537 	f2fs_destroy_page_array_cache(sbi);
4538 free_xattr_cache:
4539 	f2fs_destroy_xattr_caches(sbi);
4540 free_io_dummy:
4541 	mempool_destroy(sbi->write_io_dummy);
4542 free_percpu:
4543 	destroy_percpu_info(sbi);
4544 free_iostat:
4545 	f2fs_destroy_iostat(sbi);
4546 free_bio_info:
4547 	for (i = 0; i < NR_PAGE_TYPE; i++)
4548 		kvfree(sbi->write_io[i]);
4549 
4550 #if IS_ENABLED(CONFIG_UNICODE)
4551 	utf8_unload(sb->s_encoding);
4552 	sb->s_encoding = NULL;
4553 #endif
4554 free_options:
4555 #ifdef CONFIG_QUOTA
4556 	for (i = 0; i < MAXQUOTAS; i++)
4557 		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4558 #endif
4559 	fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4560 	kvfree(options);
4561 free_sb_buf:
4562 	kfree(raw_super);
4563 free_sbi:
4564 	if (sbi->s_chksum_driver)
4565 		crypto_free_shash(sbi->s_chksum_driver);
4566 	kfree(sbi);
4567 
4568 	/* give only one another chance */
4569 	if (retry_cnt > 0 && skip_recovery) {
4570 		retry_cnt--;
4571 		shrink_dcache_sb(sb);
4572 		goto try_onemore;
4573 	}
4574 	return err;
4575 }
4576 
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4577 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4578 			const char *dev_name, void *data)
4579 {
4580 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4581 }
4582 
kill_f2fs_super(struct super_block * sb)4583 static void kill_f2fs_super(struct super_block *sb)
4584 {
4585 	if (sb->s_root) {
4586 		struct f2fs_sb_info *sbi = F2FS_SB(sb);
4587 
4588 		set_sbi_flag(sbi, SBI_IS_CLOSE);
4589 		f2fs_stop_gc_thread(sbi);
4590 		f2fs_stop_discard_thread(sbi);
4591 
4592 #ifdef CONFIG_F2FS_FS_COMPRESSION
4593 		/*
4594 		 * latter evict_inode() can bypass checking and invalidating
4595 		 * compress inode cache.
4596 		 */
4597 		if (test_opt(sbi, COMPRESS_CACHE))
4598 			truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4599 #endif
4600 
4601 		if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4602 				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4603 			struct cp_control cpc = {
4604 				.reason = CP_UMOUNT,
4605 			};
4606 			f2fs_write_checkpoint(sbi, &cpc);
4607 		}
4608 
4609 		if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4610 			sb->s_flags &= ~SB_RDONLY;
4611 	}
4612 	kill_block_super(sb);
4613 }
4614 
4615 static struct file_system_type f2fs_fs_type = {
4616 	.owner		= THIS_MODULE,
4617 	.name		= "f2fs",
4618 	.mount		= f2fs_mount,
4619 	.kill_sb	= kill_f2fs_super,
4620 	.fs_flags	= FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4621 };
4622 MODULE_ALIAS_FS("f2fs");
4623 
init_inodecache(void)4624 static int __init init_inodecache(void)
4625 {
4626 	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4627 			sizeof(struct f2fs_inode_info), 0,
4628 			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4629 	if (!f2fs_inode_cachep)
4630 		return -ENOMEM;
4631 	return 0;
4632 }
4633 
destroy_inodecache(void)4634 static void destroy_inodecache(void)
4635 {
4636 	/*
4637 	 * Make sure all delayed rcu free inodes are flushed before we
4638 	 * destroy cache.
4639 	 */
4640 	rcu_barrier();
4641 	kmem_cache_destroy(f2fs_inode_cachep);
4642 }
4643 
init_f2fs_fs(void)4644 static int __init init_f2fs_fs(void)
4645 {
4646 	int err;
4647 
4648 	if (PAGE_SIZE != F2FS_BLKSIZE) {
4649 		printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4650 				PAGE_SIZE, F2FS_BLKSIZE);
4651 		return -EINVAL;
4652 	}
4653 
4654 	err = init_inodecache();
4655 	if (err)
4656 		goto fail;
4657 	err = f2fs_create_node_manager_caches();
4658 	if (err)
4659 		goto free_inodecache;
4660 	err = f2fs_create_segment_manager_caches();
4661 	if (err)
4662 		goto free_node_manager_caches;
4663 	err = f2fs_create_checkpoint_caches();
4664 	if (err)
4665 		goto free_segment_manager_caches;
4666 	err = f2fs_create_recovery_cache();
4667 	if (err)
4668 		goto free_checkpoint_caches;
4669 	err = f2fs_create_extent_cache();
4670 	if (err)
4671 		goto free_recovery_cache;
4672 	err = f2fs_create_garbage_collection_cache();
4673 	if (err)
4674 		goto free_extent_cache;
4675 	err = f2fs_init_sysfs();
4676 	if (err)
4677 		goto free_garbage_collection_cache;
4678 	err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4679 	if (err)
4680 		goto free_sysfs;
4681 	err = register_filesystem(&f2fs_fs_type);
4682 	if (err)
4683 		goto free_shrinker;
4684 	f2fs_create_root_stats();
4685 	err = f2fs_init_post_read_processing();
4686 	if (err)
4687 		goto free_root_stats;
4688 	err = f2fs_init_iostat_processing();
4689 	if (err)
4690 		goto free_post_read;
4691 	err = f2fs_init_bio_entry_cache();
4692 	if (err)
4693 		goto free_iostat;
4694 	err = f2fs_init_bioset();
4695 	if (err)
4696 		goto free_bio_enrty_cache;
4697 	err = f2fs_init_compress_mempool();
4698 	if (err)
4699 		goto free_bioset;
4700 	err = f2fs_init_compress_cache();
4701 	if (err)
4702 		goto free_compress_mempool;
4703 	err = f2fs_create_casefold_cache();
4704 	if (err)
4705 		goto free_compress_cache;
4706 	return 0;
4707 free_compress_cache:
4708 	f2fs_destroy_compress_cache();
4709 free_compress_mempool:
4710 	f2fs_destroy_compress_mempool();
4711 free_bioset:
4712 	f2fs_destroy_bioset();
4713 free_bio_enrty_cache:
4714 	f2fs_destroy_bio_entry_cache();
4715 free_iostat:
4716 	f2fs_destroy_iostat_processing();
4717 free_post_read:
4718 	f2fs_destroy_post_read_processing();
4719 free_root_stats:
4720 	f2fs_destroy_root_stats();
4721 	unregister_filesystem(&f2fs_fs_type);
4722 free_shrinker:
4723 	unregister_shrinker(&f2fs_shrinker_info);
4724 free_sysfs:
4725 	f2fs_exit_sysfs();
4726 free_garbage_collection_cache:
4727 	f2fs_destroy_garbage_collection_cache();
4728 free_extent_cache:
4729 	f2fs_destroy_extent_cache();
4730 free_recovery_cache:
4731 	f2fs_destroy_recovery_cache();
4732 free_checkpoint_caches:
4733 	f2fs_destroy_checkpoint_caches();
4734 free_segment_manager_caches:
4735 	f2fs_destroy_segment_manager_caches();
4736 free_node_manager_caches:
4737 	f2fs_destroy_node_manager_caches();
4738 free_inodecache:
4739 	destroy_inodecache();
4740 fail:
4741 	return err;
4742 }
4743 
exit_f2fs_fs(void)4744 static void __exit exit_f2fs_fs(void)
4745 {
4746 	f2fs_destroy_casefold_cache();
4747 	f2fs_destroy_compress_cache();
4748 	f2fs_destroy_compress_mempool();
4749 	f2fs_destroy_bioset();
4750 	f2fs_destroy_bio_entry_cache();
4751 	f2fs_destroy_iostat_processing();
4752 	f2fs_destroy_post_read_processing();
4753 	f2fs_destroy_root_stats();
4754 	unregister_filesystem(&f2fs_fs_type);
4755 	unregister_shrinker(&f2fs_shrinker_info);
4756 	f2fs_exit_sysfs();
4757 	f2fs_destroy_garbage_collection_cache();
4758 	f2fs_destroy_extent_cache();
4759 	f2fs_destroy_recovery_cache();
4760 	f2fs_destroy_checkpoint_caches();
4761 	f2fs_destroy_segment_manager_caches();
4762 	f2fs_destroy_node_manager_caches();
4763 	destroy_inodecache();
4764 }
4765 
4766 module_init(init_f2fs_fs)
4767 module_exit(exit_f2fs_fs)
4768 
4769 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4770 MODULE_DESCRIPTION("Flash Friendly File System");
4771 MODULE_LICENSE("GPL");
4772 MODULE_SOFTDEP("pre: crc32");
4773 
4774