1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Copyright (C) 2017-2018 HUAWEI, Inc.
4 * https://www.huawei.com/
5 * Copyright (C) 2021, Alibaba Cloud
6 */
7 #include <linux/statfs.h>
8 #include <linux/seq_file.h>
9 #include <linux/crc32c.h>
10 #include <linux/fs_context.h>
11 #include <linux/fs_parser.h>
12 #include <linux/exportfs.h>
13 #include <linux/backing-dev.h>
14 #include "xattr.h"
15
16 #define CREATE_TRACE_POINTS
17 #include <trace/events/erofs.h>
18
19 static struct kmem_cache *erofs_inode_cachep __read_mostly;
20
_erofs_printk(struct super_block * sb,const char * fmt,...)21 void _erofs_printk(struct super_block *sb, const char *fmt, ...)
22 {
23 struct va_format vaf;
24 va_list args;
25 int level;
26
27 va_start(args, fmt);
28
29 level = printk_get_level(fmt);
30 vaf.fmt = printk_skip_level(fmt);
31 vaf.va = &args;
32 if (sb)
33 printk("%c%cerofs (device %s): %pV",
34 KERN_SOH_ASCII, level, sb->s_id, &vaf);
35 else
36 printk("%c%cerofs: %pV", KERN_SOH_ASCII, level, &vaf);
37 va_end(args);
38 }
39
erofs_superblock_csum_verify(struct super_block * sb,void * sbdata)40 static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
41 {
42 struct erofs_super_block *dsb = sbdata + EROFS_SUPER_OFFSET;
43 u32 len = 1 << EROFS_SB(sb)->blkszbits, crc;
44
45 if (len > EROFS_SUPER_OFFSET)
46 len -= EROFS_SUPER_OFFSET;
47 len -= offsetof(struct erofs_super_block, checksum) +
48 sizeof(dsb->checksum);
49
50 /* skip .magic(pre-verified) and .checksum(0) fields */
51 crc = crc32c(0x5045B54A, (&dsb->checksum) + 1, len);
52 if (crc == le32_to_cpu(dsb->checksum))
53 return 0;
54 erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
55 crc, le32_to_cpu(dsb->checksum));
56 return -EBADMSG;
57 }
58
erofs_inode_init_once(void * ptr)59 static void erofs_inode_init_once(void *ptr)
60 {
61 struct erofs_inode *vi = ptr;
62
63 inode_init_once(&vi->vfs_inode);
64 }
65
erofs_alloc_inode(struct super_block * sb)66 static struct inode *erofs_alloc_inode(struct super_block *sb)
67 {
68 struct erofs_inode *vi =
69 alloc_inode_sb(sb, erofs_inode_cachep, GFP_KERNEL);
70
71 if (!vi)
72 return NULL;
73
74 /* zero out everything except vfs_inode */
75 memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
76 return &vi->vfs_inode;
77 }
78
erofs_free_inode(struct inode * inode)79 static void erofs_free_inode(struct inode *inode)
80 {
81 struct erofs_inode *vi = EROFS_I(inode);
82
83 if (inode->i_op == &erofs_fast_symlink_iops)
84 kfree(inode->i_link);
85 kfree(vi->xattr_shared_xattrs);
86 kmem_cache_free(erofs_inode_cachep, vi);
87 }
88
89 /* read variable-sized metadata, offset will be aligned by 4-byte */
erofs_read_metadata(struct super_block * sb,struct erofs_buf * buf,erofs_off_t * offset,int * lengthp)90 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
91 erofs_off_t *offset, int *lengthp)
92 {
93 u8 *buffer, *ptr;
94 int len, i, cnt;
95
96 *offset = round_up(*offset, 4);
97 ptr = erofs_bread(buf, *offset, true);
98 if (IS_ERR(ptr))
99 return ptr;
100
101 len = le16_to_cpu(*(__le16 *)ptr);
102 if (!len)
103 len = U16_MAX + 1;
104 buffer = kmalloc(len, GFP_KERNEL);
105 if (!buffer)
106 return ERR_PTR(-ENOMEM);
107 *offset += sizeof(__le16);
108 *lengthp = len;
109
110 for (i = 0; i < len; i += cnt) {
111 cnt = min_t(int, sb->s_blocksize - erofs_blkoff(sb, *offset),
112 len - i);
113 ptr = erofs_bread(buf, *offset, true);
114 if (IS_ERR(ptr)) {
115 kfree(buffer);
116 return ptr;
117 }
118 memcpy(buffer + i, ptr, cnt);
119 *offset += cnt;
120 }
121 return buffer;
122 }
123
124 #ifndef CONFIG_EROFS_FS_ZIP
z_erofs_parse_cfgs(struct super_block * sb,struct erofs_super_block * dsb)125 static int z_erofs_parse_cfgs(struct super_block *sb,
126 struct erofs_super_block *dsb)
127 {
128 if (!dsb->u1.available_compr_algs)
129 return 0;
130
131 erofs_err(sb, "compression disabled, unable to mount compressed EROFS");
132 return -EOPNOTSUPP;
133 }
134 #endif
135
erofs_init_device(struct erofs_buf * buf,struct super_block * sb,struct erofs_device_info * dif,erofs_off_t * pos)136 static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
137 struct erofs_device_info *dif, erofs_off_t *pos)
138 {
139 struct erofs_sb_info *sbi = EROFS_SB(sb);
140 struct erofs_fscache *fscache;
141 struct erofs_deviceslot *dis;
142 struct file *file;
143
144 dis = erofs_read_metabuf(buf, sb, *pos, false);
145 if (IS_ERR(dis))
146 return PTR_ERR(dis);
147
148 if (!sbi->devs->flatdev && !dif->path) {
149 if (!dis->tag[0]) {
150 erofs_err(sb, "empty device tag @ pos %llu", *pos);
151 return -EINVAL;
152 }
153 dif->path = kmemdup_nul(dis->tag, sizeof(dis->tag), GFP_KERNEL);
154 if (!dif->path)
155 return -ENOMEM;
156 }
157
158 if (erofs_is_fscache_mode(sb)) {
159 fscache = erofs_fscache_register_cookie(sb, dif->path, 0);
160 if (IS_ERR(fscache))
161 return PTR_ERR(fscache);
162 dif->fscache = fscache;
163 } else if (!sbi->devs->flatdev) {
164 file = erofs_is_fileio_mode(sbi) ?
165 filp_open(dif->path, O_RDONLY | O_LARGEFILE, 0) :
166 bdev_file_open_by_path(dif->path,
167 BLK_OPEN_READ, sb->s_type, NULL);
168 if (IS_ERR(file)) {
169 if (file == ERR_PTR(-ENOTBLK))
170 return -EINVAL;
171 return PTR_ERR(file);
172 }
173
174 if (!erofs_is_fileio_mode(sbi)) {
175 dif->dax_dev = fs_dax_get_by_bdev(file_bdev(file),
176 &dif->dax_part_off, NULL, NULL);
177 } else if (!S_ISREG(file_inode(file)->i_mode)) {
178 fput(file);
179 return -EINVAL;
180 }
181 if (!dif->dax_dev && test_opt(&sbi->opt, DAX_ALWAYS)) {
182 erofs_info(sb, "DAX unsupported by %s. Turning off DAX.",
183 dif->path);
184 clear_opt(&sbi->opt, DAX_ALWAYS);
185 }
186 dif->file = file;
187 }
188
189 dif->blocks = le32_to_cpu(dis->blocks_lo);
190 dif->uniaddr = le32_to_cpu(dis->uniaddr_lo);
191 sbi->total_blocks += dif->blocks;
192 *pos += EROFS_DEVT_SLOT_SIZE;
193 return 0;
194 }
195
erofs_scan_devices(struct super_block * sb,struct erofs_super_block * dsb)196 static int erofs_scan_devices(struct super_block *sb,
197 struct erofs_super_block *dsb)
198 {
199 struct erofs_sb_info *sbi = EROFS_SB(sb);
200 unsigned int ondisk_extradevs;
201 erofs_off_t pos;
202 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
203 struct erofs_device_info *dif;
204 int id, err = 0;
205
206 sbi->total_blocks = sbi->dif0.blocks;
207 if (!erofs_sb_has_device_table(sbi))
208 ondisk_extradevs = 0;
209 else
210 ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
211
212 if (sbi->devs->extra_devices &&
213 ondisk_extradevs != sbi->devs->extra_devices) {
214 erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
215 ondisk_extradevs, sbi->devs->extra_devices);
216 return -EINVAL;
217 }
218
219 if (test_opt(&sbi->opt, DAX_ALWAYS) && !sbi->dif0.dax_dev) {
220 erofs_info(sb, "DAX unsupported by block device. Turning off DAX.");
221 clear_opt(&sbi->opt, DAX_ALWAYS);
222 }
223 if (!ondisk_extradevs)
224 return 0;
225
226 if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
227 sbi->devs->flatdev = true;
228
229 sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
230 pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
231 down_read(&sbi->devs->rwsem);
232 if (sbi->devs->extra_devices) {
233 idr_for_each_entry(&sbi->devs->tree, dif, id) {
234 err = erofs_init_device(&buf, sb, dif, &pos);
235 if (err)
236 break;
237 }
238 } else {
239 for (id = 0; id < ondisk_extradevs; id++) {
240 dif = kzalloc(sizeof(*dif), GFP_KERNEL);
241 if (!dif) {
242 err = -ENOMEM;
243 break;
244 }
245
246 err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
247 if (err < 0) {
248 kfree(dif);
249 break;
250 }
251 ++sbi->devs->extra_devices;
252
253 err = erofs_init_device(&buf, sb, dif, &pos);
254 if (err)
255 break;
256 }
257 }
258 up_read(&sbi->devs->rwsem);
259 erofs_put_metabuf(&buf);
260 return err;
261 }
262
erofs_read_superblock(struct super_block * sb)263 static int erofs_read_superblock(struct super_block *sb)
264 {
265 struct erofs_sb_info *sbi = EROFS_SB(sb);
266 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
267 struct erofs_super_block *dsb;
268 void *data;
269 int ret;
270
271 data = erofs_read_metabuf(&buf, sb, 0, false);
272 if (IS_ERR(data)) {
273 erofs_err(sb, "cannot read erofs superblock");
274 return PTR_ERR(data);
275 }
276
277 dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
278 ret = -EINVAL;
279 if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
280 erofs_err(sb, "cannot find valid erofs superblock");
281 goto out;
282 }
283
284 sbi->blkszbits = dsb->blkszbits;
285 if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
286 erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
287 goto out;
288 }
289 if (dsb->dirblkbits) {
290 erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
291 goto out;
292 }
293
294 sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
295 if (erofs_sb_has_sb_chksum(sbi)) {
296 ret = erofs_superblock_csum_verify(sb, data);
297 if (ret)
298 goto out;
299 }
300
301 ret = -EINVAL;
302 sbi->feature_incompat = le32_to_cpu(dsb->feature_incompat);
303 if (sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT) {
304 erofs_err(sb, "unidentified incompatible feature %x, please upgrade kernel",
305 sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT);
306 goto out;
307 }
308
309 sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
310 if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
311 erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
312 sbi->sb_size);
313 goto out;
314 }
315 sbi->dif0.blocks = le32_to_cpu(dsb->blocks_lo);
316 sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
317 #ifdef CONFIG_EROFS_FS_XATTR
318 sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
319 sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
320 sbi->xattr_prefix_count = dsb->xattr_prefix_count;
321 sbi->xattr_filter_reserved = dsb->xattr_filter_reserved;
322 #endif
323 sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
324 if (erofs_sb_has_48bit(sbi) && dsb->rootnid_8b) {
325 sbi->root_nid = le64_to_cpu(dsb->rootnid_8b);
326 sbi->dif0.blocks = sbi->dif0.blocks |
327 ((u64)le16_to_cpu(dsb->rb.blocks_hi) << 32);
328 } else {
329 sbi->root_nid = le16_to_cpu(dsb->rb.rootnid_2b);
330 }
331 sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
332 if (erofs_sb_has_metabox(sbi)) {
333 if (sbi->sb_size <= offsetof(struct erofs_super_block,
334 metabox_nid))
335 return -EFSCORRUPTED;
336 sbi->metabox_nid = le64_to_cpu(dsb->metabox_nid);
337 if (sbi->metabox_nid & BIT_ULL(EROFS_DIRENT_NID_METABOX_BIT))
338 return -EFSCORRUPTED; /* self-loop detection */
339 }
340 sbi->inos = le64_to_cpu(dsb->inos);
341
342 sbi->epoch = (s64)le64_to_cpu(dsb->epoch);
343 sbi->fixed_nsec = le32_to_cpu(dsb->fixed_nsec);
344 super_set_uuid(sb, (void *)dsb->uuid, sizeof(dsb->uuid));
345
346 if (dsb->volume_name[0]) {
347 sbi->volume_name = kstrndup(dsb->volume_name,
348 sizeof(dsb->volume_name), GFP_KERNEL);
349 if (!sbi->volume_name)
350 return -ENOMEM;
351 }
352
353 /* parse on-disk compression configurations */
354 ret = z_erofs_parse_cfgs(sb, dsb);
355 if (ret < 0)
356 goto out;
357
358 ret = erofs_scan_devices(sb, dsb);
359
360 if (erofs_sb_has_48bit(sbi))
361 erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
362 if (erofs_sb_has_metabox(sbi))
363 erofs_info(sb, "EXPERIMENTAL metadata compression support in use. Use at your own risk!");
364 if (erofs_is_fscache_mode(sb))
365 erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
366 out:
367 erofs_put_metabuf(&buf);
368 return ret;
369 }
370
erofs_default_options(struct erofs_sb_info * sbi)371 static void erofs_default_options(struct erofs_sb_info *sbi)
372 {
373 #ifdef CONFIG_EROFS_FS_ZIP
374 sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
375 sbi->opt.max_sync_decompress_pages = 3;
376 sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
377 #endif
378 #ifdef CONFIG_EROFS_FS_XATTR
379 set_opt(&sbi->opt, XATTR_USER);
380 #endif
381 #ifdef CONFIG_EROFS_FS_POSIX_ACL
382 set_opt(&sbi->opt, POSIX_ACL);
383 #endif
384 }
385
386 enum {
387 Opt_user_xattr, Opt_acl, Opt_cache_strategy, Opt_dax, Opt_dax_enum,
388 Opt_device, Opt_fsid, Opt_domain_id, Opt_directio, Opt_fsoffset,
389 };
390
391 static const struct constant_table erofs_param_cache_strategy[] = {
392 {"disabled", EROFS_ZIP_CACHE_DISABLED},
393 {"readahead", EROFS_ZIP_CACHE_READAHEAD},
394 {"readaround", EROFS_ZIP_CACHE_READAROUND},
395 {}
396 };
397
398 static const struct constant_table erofs_dax_param_enums[] = {
399 {"always", EROFS_MOUNT_DAX_ALWAYS},
400 {"never", EROFS_MOUNT_DAX_NEVER},
401 {}
402 };
403
404 static const struct fs_parameter_spec erofs_fs_parameters[] = {
405 fsparam_flag_no("user_xattr", Opt_user_xattr),
406 fsparam_flag_no("acl", Opt_acl),
407 fsparam_enum("cache_strategy", Opt_cache_strategy,
408 erofs_param_cache_strategy),
409 fsparam_flag("dax", Opt_dax),
410 fsparam_enum("dax", Opt_dax_enum, erofs_dax_param_enums),
411 fsparam_string("device", Opt_device),
412 fsparam_string("fsid", Opt_fsid),
413 fsparam_string("domain_id", Opt_domain_id),
414 fsparam_flag_no("directio", Opt_directio),
415 fsparam_u64("fsoffset", Opt_fsoffset),
416 {}
417 };
418
erofs_fc_set_dax_mode(struct fs_context * fc,unsigned int mode)419 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
420 {
421 #ifdef CONFIG_FS_DAX
422 struct erofs_sb_info *sbi = fc->s_fs_info;
423
424 switch (mode) {
425 case EROFS_MOUNT_DAX_ALWAYS:
426 set_opt(&sbi->opt, DAX_ALWAYS);
427 clear_opt(&sbi->opt, DAX_NEVER);
428 return true;
429 case EROFS_MOUNT_DAX_NEVER:
430 set_opt(&sbi->opt, DAX_NEVER);
431 clear_opt(&sbi->opt, DAX_ALWAYS);
432 return true;
433 default:
434 DBG_BUGON(1);
435 return false;
436 }
437 #else
438 errorfc(fc, "dax options not supported");
439 return false;
440 #endif
441 }
442
erofs_fc_parse_param(struct fs_context * fc,struct fs_parameter * param)443 static int erofs_fc_parse_param(struct fs_context *fc,
444 struct fs_parameter *param)
445 {
446 struct erofs_sb_info *sbi = fc->s_fs_info;
447 struct fs_parse_result result;
448 struct erofs_device_info *dif;
449 int opt, ret;
450
451 opt = fs_parse(fc, erofs_fs_parameters, param, &result);
452 if (opt < 0)
453 return opt;
454
455 switch (opt) {
456 case Opt_user_xattr:
457 #ifdef CONFIG_EROFS_FS_XATTR
458 if (result.boolean)
459 set_opt(&sbi->opt, XATTR_USER);
460 else
461 clear_opt(&sbi->opt, XATTR_USER);
462 #else
463 errorfc(fc, "{,no}user_xattr options not supported");
464 #endif
465 break;
466 case Opt_acl:
467 #ifdef CONFIG_EROFS_FS_POSIX_ACL
468 if (result.boolean)
469 set_opt(&sbi->opt, POSIX_ACL);
470 else
471 clear_opt(&sbi->opt, POSIX_ACL);
472 #else
473 errorfc(fc, "{,no}acl options not supported");
474 #endif
475 break;
476 case Opt_cache_strategy:
477 #ifdef CONFIG_EROFS_FS_ZIP
478 sbi->opt.cache_strategy = result.uint_32;
479 #else
480 errorfc(fc, "compression not supported, cache_strategy ignored");
481 #endif
482 break;
483 case Opt_dax:
484 if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
485 return -EINVAL;
486 break;
487 case Opt_dax_enum:
488 if (!erofs_fc_set_dax_mode(fc, result.uint_32))
489 return -EINVAL;
490 break;
491 case Opt_device:
492 dif = kzalloc(sizeof(*dif), GFP_KERNEL);
493 if (!dif)
494 return -ENOMEM;
495 dif->path = kstrdup(param->string, GFP_KERNEL);
496 if (!dif->path) {
497 kfree(dif);
498 return -ENOMEM;
499 }
500 down_write(&sbi->devs->rwsem);
501 ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
502 up_write(&sbi->devs->rwsem);
503 if (ret < 0) {
504 kfree(dif->path);
505 kfree(dif);
506 return ret;
507 }
508 ++sbi->devs->extra_devices;
509 break;
510 #ifdef CONFIG_EROFS_FS_ONDEMAND
511 case Opt_fsid:
512 kfree(sbi->fsid);
513 sbi->fsid = kstrdup(param->string, GFP_KERNEL);
514 if (!sbi->fsid)
515 return -ENOMEM;
516 break;
517 case Opt_domain_id:
518 kfree(sbi->domain_id);
519 sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
520 if (!sbi->domain_id)
521 return -ENOMEM;
522 break;
523 #else
524 case Opt_fsid:
525 case Opt_domain_id:
526 errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
527 break;
528 #endif
529 case Opt_directio:
530 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
531 if (result.boolean)
532 set_opt(&sbi->opt, DIRECT_IO);
533 else
534 clear_opt(&sbi->opt, DIRECT_IO);
535 #else
536 errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
537 #endif
538 break;
539 case Opt_fsoffset:
540 sbi->dif0.fsoff = result.uint_64;
541 break;
542 }
543 return 0;
544 }
545
erofs_encode_fh(struct inode * inode,u32 * fh,int * max_len,struct inode * parent)546 static int erofs_encode_fh(struct inode *inode, u32 *fh, int *max_len,
547 struct inode *parent)
548 {
549 erofs_nid_t nid = EROFS_I(inode)->nid;
550 int len = parent ? 6 : 3;
551
552 if (*max_len < len) {
553 *max_len = len;
554 return FILEID_INVALID;
555 }
556
557 fh[0] = (u32)(nid >> 32);
558 fh[1] = (u32)(nid & 0xffffffff);
559 fh[2] = inode->i_generation;
560
561 if (parent) {
562 nid = EROFS_I(parent)->nid;
563
564 fh[3] = (u32)(nid >> 32);
565 fh[4] = (u32)(nid & 0xffffffff);
566 fh[5] = parent->i_generation;
567 }
568
569 *max_len = len;
570 return parent ? FILEID_INO64_GEN_PARENT : FILEID_INO64_GEN;
571 }
572
erofs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)573 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
574 struct fid *fid, int fh_len, int fh_type)
575 {
576 if ((fh_type != FILEID_INO64_GEN &&
577 fh_type != FILEID_INO64_GEN_PARENT) || fh_len < 3)
578 return NULL;
579
580 return d_obtain_alias(erofs_iget(sb,
581 ((u64)fid->raw[0] << 32) | fid->raw[1]));
582 }
583
erofs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)584 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
585 struct fid *fid, int fh_len, int fh_type)
586 {
587 if (fh_type != FILEID_INO64_GEN_PARENT || fh_len < 6)
588 return NULL;
589
590 return d_obtain_alias(erofs_iget(sb,
591 ((u64)fid->raw[3] << 32) | fid->raw[4]));
592 }
593
erofs_get_parent(struct dentry * child)594 static struct dentry *erofs_get_parent(struct dentry *child)
595 {
596 erofs_nid_t nid;
597 unsigned int d_type;
598 int err;
599
600 err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
601 if (err)
602 return ERR_PTR(err);
603 return d_obtain_alias(erofs_iget(child->d_sb, nid));
604 }
605
606 static const struct export_operations erofs_export_ops = {
607 .encode_fh = erofs_encode_fh,
608 .fh_to_dentry = erofs_fh_to_dentry,
609 .fh_to_parent = erofs_fh_to_parent,
610 .get_parent = erofs_get_parent,
611 };
612
erofs_set_sysfs_name(struct super_block * sb)613 static void erofs_set_sysfs_name(struct super_block *sb)
614 {
615 struct erofs_sb_info *sbi = EROFS_SB(sb);
616
617 if (sbi->domain_id)
618 super_set_sysfs_name_generic(sb, "%s,%s", sbi->domain_id,
619 sbi->fsid);
620 else if (sbi->fsid)
621 super_set_sysfs_name_generic(sb, "%s", sbi->fsid);
622 else if (erofs_is_fileio_mode(sbi))
623 super_set_sysfs_name_generic(sb, "%s",
624 bdi_dev_name(sb->s_bdi));
625 else
626 super_set_sysfs_name_id(sb);
627 }
628
erofs_fc_fill_super(struct super_block * sb,struct fs_context * fc)629 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
630 {
631 struct inode *inode;
632 struct erofs_sb_info *sbi = EROFS_SB(sb);
633 int err;
634
635 sb->s_magic = EROFS_SUPER_MAGIC;
636 sb->s_flags |= SB_RDONLY | SB_NOATIME;
637 sb->s_maxbytes = MAX_LFS_FILESIZE;
638 sb->s_op = &erofs_sops;
639
640 sbi->blkszbits = PAGE_SHIFT;
641 if (!sb->s_bdev) {
642 /*
643 * (File-backed mounts) EROFS claims it's safe to nest other
644 * fs contexts (including its own) due to self-controlled RO
645 * accesses/contexts and no side-effect changes that need to
646 * context save & restore so it can reuse the current thread
647 * context.
648 * However, we still need to prevent kernel stack overflow due
649 * to filesystem nesting: just ensure that s_stack_depth is 0
650 * to disallow mounting EROFS on stacked filesystems.
651 * Note: s_stack_depth is not incremented here for now, since
652 * EROFS is the only fs supporting file-backed mounts for now.
653 * It MUST change if another fs plans to support them, which
654 * may also require adjusting FILESYSTEM_MAX_STACK_DEPTH.
655 */
656 if (erofs_is_fileio_mode(sbi)) {
657 inode = file_inode(sbi->dif0.file);
658 if ((inode->i_sb->s_op == &erofs_sops &&
659 !inode->i_sb->s_bdev) ||
660 inode->i_sb->s_stack_depth) {
661 erofs_err(sb, "file-backed mounts cannot be applied to stacked fses");
662 return -ENOTBLK;
663 }
664 }
665 sb->s_blocksize = PAGE_SIZE;
666 sb->s_blocksize_bits = PAGE_SHIFT;
667
668 if (erofs_is_fscache_mode(sb)) {
669 err = erofs_fscache_register_fs(sb);
670 if (err)
671 return err;
672 }
673 err = super_setup_bdi(sb);
674 if (err)
675 return err;
676 } else {
677 if (!sb_set_blocksize(sb, PAGE_SIZE)) {
678 errorfc(fc, "failed to set initial blksize");
679 return -EINVAL;
680 }
681
682 sbi->dif0.dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
683 &sbi->dif0.dax_part_off, NULL, NULL);
684 }
685
686 err = erofs_read_superblock(sb);
687 if (err)
688 return err;
689
690 if (sb->s_blocksize_bits != sbi->blkszbits) {
691 if (erofs_is_fscache_mode(sb)) {
692 errorfc(fc, "unsupported blksize for fscache mode");
693 return -EINVAL;
694 }
695
696 if (erofs_is_fileio_mode(sbi)) {
697 sb->s_blocksize = 1 << sbi->blkszbits;
698 sb->s_blocksize_bits = sbi->blkszbits;
699 } else if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
700 errorfc(fc, "failed to set erofs blksize");
701 return -EINVAL;
702 }
703 }
704
705 if (sbi->dif0.fsoff) {
706 if (sbi->dif0.fsoff & (sb->s_blocksize - 1))
707 return invalfc(fc, "fsoffset %llu is not aligned to block size %lu",
708 sbi->dif0.fsoff, sb->s_blocksize);
709 if (erofs_is_fscache_mode(sb))
710 return invalfc(fc, "cannot use fsoffset in fscache mode");
711 }
712
713 if (test_opt(&sbi->opt, DAX_ALWAYS) && sbi->blkszbits != PAGE_SHIFT) {
714 erofs_info(sb, "unsupported blocksize for DAX");
715 clear_opt(&sbi->opt, DAX_ALWAYS);
716 }
717
718 sb->s_time_gran = 1;
719 sb->s_xattr = erofs_xattr_handlers;
720 sb->s_export_op = &erofs_export_ops;
721
722 if (test_opt(&sbi->opt, POSIX_ACL))
723 sb->s_flags |= SB_POSIXACL;
724 else
725 sb->s_flags &= ~SB_POSIXACL;
726
727 err = z_erofs_init_super(sb);
728 if (err)
729 return err;
730
731 if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
732 inode = erofs_iget(sb, sbi->packed_nid);
733 if (IS_ERR(inode))
734 return PTR_ERR(inode);
735 sbi->packed_inode = inode;
736 }
737 if (erofs_sb_has_metabox(sbi)) {
738 inode = erofs_iget(sb, sbi->metabox_nid);
739 if (IS_ERR(inode))
740 return PTR_ERR(inode);
741 sbi->metabox_inode = inode;
742 }
743
744 inode = erofs_iget(sb, sbi->root_nid);
745 if (IS_ERR(inode))
746 return PTR_ERR(inode);
747
748 if (!S_ISDIR(inode->i_mode)) {
749 erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
750 sbi->root_nid, inode->i_mode);
751 iput(inode);
752 return -EINVAL;
753 }
754 sb->s_root = d_make_root(inode);
755 if (!sb->s_root)
756 return -ENOMEM;
757
758 erofs_shrinker_register(sb);
759 err = erofs_xattr_prefixes_init(sb);
760 if (err)
761 return err;
762
763 erofs_set_sysfs_name(sb);
764 err = erofs_register_sysfs(sb);
765 if (err)
766 return err;
767
768 sbi->dir_ra_bytes = EROFS_DIR_RA_BYTES;
769 erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
770 return 0;
771 }
772
erofs_fc_get_tree(struct fs_context * fc)773 static int erofs_fc_get_tree(struct fs_context *fc)
774 {
775 struct erofs_sb_info *sbi = fc->s_fs_info;
776 int ret;
777
778 if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
779 return get_tree_nodev(fc, erofs_fc_fill_super);
780
781 ret = get_tree_bdev_flags(fc, erofs_fc_fill_super,
782 IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) ?
783 GET_TREE_BDEV_QUIET_LOOKUP : 0);
784 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
785 if (ret == -ENOTBLK) {
786 struct file *file;
787
788 if (!fc->source)
789 return invalf(fc, "No source specified");
790 file = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
791 if (IS_ERR(file))
792 return PTR_ERR(file);
793 sbi->dif0.file = file;
794
795 if (S_ISREG(file_inode(sbi->dif0.file)->i_mode) &&
796 sbi->dif0.file->f_mapping->a_ops->read_folio)
797 return get_tree_nodev(fc, erofs_fc_fill_super);
798 }
799 #endif
800 return ret;
801 }
802
erofs_fc_reconfigure(struct fs_context * fc)803 static int erofs_fc_reconfigure(struct fs_context *fc)
804 {
805 struct super_block *sb = fc->root->d_sb;
806 struct erofs_sb_info *sbi = EROFS_SB(sb);
807 struct erofs_sb_info *new_sbi = fc->s_fs_info;
808
809 DBG_BUGON(!sb_rdonly(sb));
810
811 if (new_sbi->fsid || new_sbi->domain_id)
812 erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
813
814 if (test_opt(&new_sbi->opt, POSIX_ACL))
815 fc->sb_flags |= SB_POSIXACL;
816 else
817 fc->sb_flags &= ~SB_POSIXACL;
818
819 sbi->opt = new_sbi->opt;
820
821 fc->sb_flags |= SB_RDONLY;
822 return 0;
823 }
824
erofs_release_device_info(int id,void * ptr,void * data)825 static int erofs_release_device_info(int id, void *ptr, void *data)
826 {
827 struct erofs_device_info *dif = ptr;
828
829 fs_put_dax(dif->dax_dev, NULL);
830 if (dif->file)
831 fput(dif->file);
832 erofs_fscache_unregister_cookie(dif->fscache);
833 dif->fscache = NULL;
834 kfree(dif->path);
835 kfree(dif);
836 return 0;
837 }
838
erofs_free_dev_context(struct erofs_dev_context * devs)839 static void erofs_free_dev_context(struct erofs_dev_context *devs)
840 {
841 if (!devs)
842 return;
843 idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
844 idr_destroy(&devs->tree);
845 kfree(devs);
846 }
847
erofs_sb_free(struct erofs_sb_info * sbi)848 static void erofs_sb_free(struct erofs_sb_info *sbi)
849 {
850 erofs_free_dev_context(sbi->devs);
851 kfree(sbi->fsid);
852 kfree(sbi->domain_id);
853 if (sbi->dif0.file)
854 fput(sbi->dif0.file);
855 kfree(sbi->volume_name);
856 kfree(sbi);
857 }
858
erofs_fc_free(struct fs_context * fc)859 static void erofs_fc_free(struct fs_context *fc)
860 {
861 struct erofs_sb_info *sbi = fc->s_fs_info;
862
863 if (sbi) /* free here if an error occurs before transferring to sb */
864 erofs_sb_free(sbi);
865 }
866
867 static const struct fs_context_operations erofs_context_ops = {
868 .parse_param = erofs_fc_parse_param,
869 .get_tree = erofs_fc_get_tree,
870 .reconfigure = erofs_fc_reconfigure,
871 .free = erofs_fc_free,
872 };
873
erofs_init_fs_context(struct fs_context * fc)874 static int erofs_init_fs_context(struct fs_context *fc)
875 {
876 struct erofs_sb_info *sbi;
877
878 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
879 if (!sbi)
880 return -ENOMEM;
881
882 sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
883 if (!sbi->devs) {
884 kfree(sbi);
885 return -ENOMEM;
886 }
887 fc->s_fs_info = sbi;
888
889 idr_init(&sbi->devs->tree);
890 init_rwsem(&sbi->devs->rwsem);
891 erofs_default_options(sbi);
892 fc->ops = &erofs_context_ops;
893 return 0;
894 }
895
erofs_drop_internal_inodes(struct erofs_sb_info * sbi)896 static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
897 {
898 iput(sbi->packed_inode);
899 sbi->packed_inode = NULL;
900 iput(sbi->metabox_inode);
901 sbi->metabox_inode = NULL;
902 #ifdef CONFIG_EROFS_FS_ZIP
903 iput(sbi->managed_cache);
904 sbi->managed_cache = NULL;
905 #endif
906 }
907
erofs_kill_sb(struct super_block * sb)908 static void erofs_kill_sb(struct super_block *sb)
909 {
910 struct erofs_sb_info *sbi = EROFS_SB(sb);
911
912 if ((IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid) ||
913 sbi->dif0.file)
914 kill_anon_super(sb);
915 else
916 kill_block_super(sb);
917 erofs_drop_internal_inodes(sbi);
918 fs_put_dax(sbi->dif0.dax_dev, NULL);
919 erofs_fscache_unregister_fs(sb);
920 erofs_sb_free(sbi);
921 sb->s_fs_info = NULL;
922 }
923
erofs_put_super(struct super_block * sb)924 static void erofs_put_super(struct super_block *sb)
925 {
926 struct erofs_sb_info *const sbi = EROFS_SB(sb);
927
928 erofs_unregister_sysfs(sb);
929 erofs_shrinker_unregister(sb);
930 erofs_xattr_prefixes_cleanup(sb);
931 erofs_drop_internal_inodes(sbi);
932 erofs_free_dev_context(sbi->devs);
933 sbi->devs = NULL;
934 erofs_fscache_unregister_fs(sb);
935 }
936
937 static struct file_system_type erofs_fs_type = {
938 .owner = THIS_MODULE,
939 .name = "erofs",
940 .init_fs_context = erofs_init_fs_context,
941 .kill_sb = erofs_kill_sb,
942 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
943 };
944 MODULE_ALIAS_FS("erofs");
945
erofs_module_init(void)946 static int __init erofs_module_init(void)
947 {
948 int err;
949
950 erofs_check_ondisk_layout_definitions();
951
952 erofs_inode_cachep = kmem_cache_create("erofs_inode",
953 sizeof(struct erofs_inode), 0,
954 SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
955 erofs_inode_init_once);
956 if (!erofs_inode_cachep)
957 return -ENOMEM;
958
959 err = erofs_init_shrinker();
960 if (err)
961 goto shrinker_err;
962
963 err = z_erofs_init_subsystem();
964 if (err)
965 goto zip_err;
966
967 err = erofs_init_sysfs();
968 if (err)
969 goto sysfs_err;
970
971 err = register_filesystem(&erofs_fs_type);
972 if (err)
973 goto fs_err;
974
975 return 0;
976
977 fs_err:
978 erofs_exit_sysfs();
979 sysfs_err:
980 z_erofs_exit_subsystem();
981 zip_err:
982 erofs_exit_shrinker();
983 shrinker_err:
984 kmem_cache_destroy(erofs_inode_cachep);
985 return err;
986 }
987
erofs_module_exit(void)988 static void __exit erofs_module_exit(void)
989 {
990 unregister_filesystem(&erofs_fs_type);
991
992 /* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
993 rcu_barrier();
994
995 erofs_exit_sysfs();
996 z_erofs_exit_subsystem();
997 erofs_exit_shrinker();
998 kmem_cache_destroy(erofs_inode_cachep);
999 }
1000
erofs_statfs(struct dentry * dentry,struct kstatfs * buf)1001 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
1002 {
1003 struct super_block *sb = dentry->d_sb;
1004 struct erofs_sb_info *sbi = EROFS_SB(sb);
1005
1006 buf->f_type = sb->s_magic;
1007 buf->f_bsize = sb->s_blocksize;
1008 buf->f_blocks = sbi->total_blocks;
1009 buf->f_bfree = buf->f_bavail = 0;
1010 buf->f_files = ULLONG_MAX;
1011 buf->f_ffree = ULLONG_MAX - sbi->inos;
1012 buf->f_namelen = EROFS_NAME_LEN;
1013
1014 if (uuid_is_null(&sb->s_uuid))
1015 buf->f_fsid = u64_to_fsid(!sb->s_bdev ? 0 :
1016 huge_encode_dev(sb->s_bdev->bd_dev));
1017 else
1018 buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
1019 return 0;
1020 }
1021
erofs_show_options(struct seq_file * seq,struct dentry * root)1022 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
1023 {
1024 struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
1025 struct erofs_mount_opts *opt = &sbi->opt;
1026
1027 if (IS_ENABLED(CONFIG_EROFS_FS_XATTR))
1028 seq_puts(seq, test_opt(opt, XATTR_USER) ?
1029 ",user_xattr" : ",nouser_xattr");
1030 if (IS_ENABLED(CONFIG_EROFS_FS_POSIX_ACL))
1031 seq_puts(seq, test_opt(opt, POSIX_ACL) ? ",acl" : ",noacl");
1032 if (IS_ENABLED(CONFIG_EROFS_FS_ZIP))
1033 seq_printf(seq, ",cache_strategy=%s",
1034 erofs_param_cache_strategy[opt->cache_strategy].name);
1035 if (test_opt(opt, DAX_ALWAYS))
1036 seq_puts(seq, ",dax=always");
1037 if (test_opt(opt, DAX_NEVER))
1038 seq_puts(seq, ",dax=never");
1039 if (erofs_is_fileio_mode(sbi) && test_opt(opt, DIRECT_IO))
1040 seq_puts(seq, ",directio");
1041 #ifdef CONFIG_EROFS_FS_ONDEMAND
1042 if (sbi->fsid)
1043 seq_printf(seq, ",fsid=%s", sbi->fsid);
1044 if (sbi->domain_id)
1045 seq_printf(seq, ",domain_id=%s", sbi->domain_id);
1046 #endif
1047 if (sbi->dif0.fsoff)
1048 seq_printf(seq, ",fsoffset=%llu", sbi->dif0.fsoff);
1049 return 0;
1050 }
1051
erofs_evict_inode(struct inode * inode)1052 static void erofs_evict_inode(struct inode *inode)
1053 {
1054 #ifdef CONFIG_FS_DAX
1055 if (IS_DAX(inode))
1056 dax_break_layout_final(inode);
1057 #endif
1058
1059 truncate_inode_pages_final(&inode->i_data);
1060 clear_inode(inode);
1061 }
1062
1063 const struct super_operations erofs_sops = {
1064 .put_super = erofs_put_super,
1065 .alloc_inode = erofs_alloc_inode,
1066 .free_inode = erofs_free_inode,
1067 .evict_inode = erofs_evict_inode,
1068 .statfs = erofs_statfs,
1069 .show_options = erofs_show_options,
1070 };
1071
1072 module_init(erofs_module_init);
1073 module_exit(erofs_module_exit);
1074
1075 MODULE_DESCRIPTION("Enhanced ROM File System");
1076 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
1077 MODULE_LICENSE("GPL");
1078