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, true);
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 dif->file = file;
182 }
183
184 dif->blocks = le32_to_cpu(dis->blocks_lo);
185 dif->uniaddr = le32_to_cpu(dis->uniaddr_lo);
186 sbi->total_blocks += dif->blocks;
187 *pos += EROFS_DEVT_SLOT_SIZE;
188 return 0;
189 }
190
erofs_scan_devices(struct super_block * sb,struct erofs_super_block * dsb)191 static int erofs_scan_devices(struct super_block *sb,
192 struct erofs_super_block *dsb)
193 {
194 struct erofs_sb_info *sbi = EROFS_SB(sb);
195 unsigned int ondisk_extradevs;
196 erofs_off_t pos;
197 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
198 struct erofs_device_info *dif;
199 int id, err = 0;
200
201 sbi->total_blocks = sbi->dif0.blocks;
202 if (!erofs_sb_has_device_table(sbi))
203 ondisk_extradevs = 0;
204 else
205 ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
206
207 if (sbi->devs->extra_devices &&
208 ondisk_extradevs != sbi->devs->extra_devices) {
209 erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
210 ondisk_extradevs, sbi->devs->extra_devices);
211 return -EINVAL;
212 }
213 if (!ondisk_extradevs)
214 return 0;
215
216 if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
217 sbi->devs->flatdev = true;
218
219 sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
220 pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
221 down_read(&sbi->devs->rwsem);
222 if (sbi->devs->extra_devices) {
223 idr_for_each_entry(&sbi->devs->tree, dif, id) {
224 err = erofs_init_device(&buf, sb, dif, &pos);
225 if (err)
226 break;
227 }
228 } else {
229 for (id = 0; id < ondisk_extradevs; id++) {
230 dif = kzalloc(sizeof(*dif), GFP_KERNEL);
231 if (!dif) {
232 err = -ENOMEM;
233 break;
234 }
235
236 err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
237 if (err < 0) {
238 kfree(dif);
239 break;
240 }
241 ++sbi->devs->extra_devices;
242
243 err = erofs_init_device(&buf, sb, dif, &pos);
244 if (err)
245 break;
246 }
247 }
248 up_read(&sbi->devs->rwsem);
249 erofs_put_metabuf(&buf);
250 return err;
251 }
252
erofs_read_superblock(struct super_block * sb)253 static int erofs_read_superblock(struct super_block *sb)
254 {
255 struct erofs_sb_info *sbi = EROFS_SB(sb);
256 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
257 struct erofs_super_block *dsb;
258 void *data;
259 int ret;
260
261 data = erofs_read_metabuf(&buf, sb, 0, true);
262 if (IS_ERR(data)) {
263 erofs_err(sb, "cannot read erofs superblock");
264 return PTR_ERR(data);
265 }
266
267 dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
268 ret = -EINVAL;
269 if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
270 erofs_err(sb, "cannot find valid erofs superblock");
271 goto out;
272 }
273
274 sbi->blkszbits = dsb->blkszbits;
275 if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
276 erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
277 goto out;
278 }
279 if (dsb->dirblkbits) {
280 erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
281 goto out;
282 }
283
284 sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
285 if (erofs_sb_has_sb_chksum(sbi)) {
286 ret = erofs_superblock_csum_verify(sb, data);
287 if (ret)
288 goto out;
289 }
290
291 ret = -EINVAL;
292 sbi->feature_incompat = le32_to_cpu(dsb->feature_incompat);
293 if (sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT) {
294 erofs_err(sb, "unidentified incompatible feature %x, please upgrade kernel",
295 sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT);
296 goto out;
297 }
298
299 sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
300 if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
301 erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
302 sbi->sb_size);
303 goto out;
304 }
305 sbi->dif0.blocks = le32_to_cpu(dsb->blocks_lo);
306 sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
307 #ifdef CONFIG_EROFS_FS_XATTR
308 sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
309 sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
310 sbi->xattr_prefix_count = dsb->xattr_prefix_count;
311 sbi->xattr_filter_reserved = dsb->xattr_filter_reserved;
312 #endif
313 sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
314 if (erofs_sb_has_48bit(sbi) && dsb->rootnid_8b) {
315 sbi->root_nid = le64_to_cpu(dsb->rootnid_8b);
316 sbi->dif0.blocks = (sbi->dif0.blocks << 32) |
317 le16_to_cpu(dsb->rb.blocks_hi);
318 } else {
319 sbi->root_nid = le16_to_cpu(dsb->rb.rootnid_2b);
320 }
321 sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
322 sbi->inos = le64_to_cpu(dsb->inos);
323
324 sbi->epoch = (s64)le64_to_cpu(dsb->epoch);
325 sbi->fixed_nsec = le32_to_cpu(dsb->fixed_nsec);
326 super_set_uuid(sb, (void *)dsb->uuid, sizeof(dsb->uuid));
327
328 /* parse on-disk compression configurations */
329 ret = z_erofs_parse_cfgs(sb, dsb);
330 if (ret < 0)
331 goto out;
332
333 /* handle multiple devices */
334 ret = erofs_scan_devices(sb, dsb);
335
336 if (erofs_sb_has_48bit(sbi))
337 erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
338 if (erofs_is_fscache_mode(sb))
339 erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
340 out:
341 erofs_put_metabuf(&buf);
342 return ret;
343 }
344
erofs_default_options(struct erofs_sb_info * sbi)345 static void erofs_default_options(struct erofs_sb_info *sbi)
346 {
347 #ifdef CONFIG_EROFS_FS_ZIP
348 sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
349 sbi->opt.max_sync_decompress_pages = 3;
350 sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
351 #endif
352 #ifdef CONFIG_EROFS_FS_XATTR
353 set_opt(&sbi->opt, XATTR_USER);
354 #endif
355 #ifdef CONFIG_EROFS_FS_POSIX_ACL
356 set_opt(&sbi->opt, POSIX_ACL);
357 #endif
358 }
359
360 enum {
361 Opt_user_xattr, Opt_acl, Opt_cache_strategy, Opt_dax, Opt_dax_enum,
362 Opt_device, Opt_fsid, Opt_domain_id, Opt_directio, Opt_fsoffset,
363 };
364
365 static const struct constant_table erofs_param_cache_strategy[] = {
366 {"disabled", EROFS_ZIP_CACHE_DISABLED},
367 {"readahead", EROFS_ZIP_CACHE_READAHEAD},
368 {"readaround", EROFS_ZIP_CACHE_READAROUND},
369 {}
370 };
371
372 static const struct constant_table erofs_dax_param_enums[] = {
373 {"always", EROFS_MOUNT_DAX_ALWAYS},
374 {"never", EROFS_MOUNT_DAX_NEVER},
375 {}
376 };
377
378 static const struct fs_parameter_spec erofs_fs_parameters[] = {
379 fsparam_flag_no("user_xattr", Opt_user_xattr),
380 fsparam_flag_no("acl", Opt_acl),
381 fsparam_enum("cache_strategy", Opt_cache_strategy,
382 erofs_param_cache_strategy),
383 fsparam_flag("dax", Opt_dax),
384 fsparam_enum("dax", Opt_dax_enum, erofs_dax_param_enums),
385 fsparam_string("device", Opt_device),
386 fsparam_string("fsid", Opt_fsid),
387 fsparam_string("domain_id", Opt_domain_id),
388 fsparam_flag_no("directio", Opt_directio),
389 fsparam_u64("fsoffset", Opt_fsoffset),
390 {}
391 };
392
erofs_fc_set_dax_mode(struct fs_context * fc,unsigned int mode)393 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
394 {
395 #ifdef CONFIG_FS_DAX
396 struct erofs_sb_info *sbi = fc->s_fs_info;
397
398 switch (mode) {
399 case EROFS_MOUNT_DAX_ALWAYS:
400 set_opt(&sbi->opt, DAX_ALWAYS);
401 clear_opt(&sbi->opt, DAX_NEVER);
402 return true;
403 case EROFS_MOUNT_DAX_NEVER:
404 set_opt(&sbi->opt, DAX_NEVER);
405 clear_opt(&sbi->opt, DAX_ALWAYS);
406 return true;
407 default:
408 DBG_BUGON(1);
409 return false;
410 }
411 #else
412 errorfc(fc, "dax options not supported");
413 return false;
414 #endif
415 }
416
erofs_fc_parse_param(struct fs_context * fc,struct fs_parameter * param)417 static int erofs_fc_parse_param(struct fs_context *fc,
418 struct fs_parameter *param)
419 {
420 struct erofs_sb_info *sbi = fc->s_fs_info;
421 struct fs_parse_result result;
422 struct erofs_device_info *dif;
423 int opt, ret;
424
425 opt = fs_parse(fc, erofs_fs_parameters, param, &result);
426 if (opt < 0)
427 return opt;
428
429 switch (opt) {
430 case Opt_user_xattr:
431 #ifdef CONFIG_EROFS_FS_XATTR
432 if (result.boolean)
433 set_opt(&sbi->opt, XATTR_USER);
434 else
435 clear_opt(&sbi->opt, XATTR_USER);
436 #else
437 errorfc(fc, "{,no}user_xattr options not supported");
438 #endif
439 break;
440 case Opt_acl:
441 #ifdef CONFIG_EROFS_FS_POSIX_ACL
442 if (result.boolean)
443 set_opt(&sbi->opt, POSIX_ACL);
444 else
445 clear_opt(&sbi->opt, POSIX_ACL);
446 #else
447 errorfc(fc, "{,no}acl options not supported");
448 #endif
449 break;
450 case Opt_cache_strategy:
451 #ifdef CONFIG_EROFS_FS_ZIP
452 sbi->opt.cache_strategy = result.uint_32;
453 #else
454 errorfc(fc, "compression not supported, cache_strategy ignored");
455 #endif
456 break;
457 case Opt_dax:
458 if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
459 return -EINVAL;
460 break;
461 case Opt_dax_enum:
462 if (!erofs_fc_set_dax_mode(fc, result.uint_32))
463 return -EINVAL;
464 break;
465 case Opt_device:
466 dif = kzalloc(sizeof(*dif), GFP_KERNEL);
467 if (!dif)
468 return -ENOMEM;
469 dif->path = kstrdup(param->string, GFP_KERNEL);
470 if (!dif->path) {
471 kfree(dif);
472 return -ENOMEM;
473 }
474 down_write(&sbi->devs->rwsem);
475 ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
476 up_write(&sbi->devs->rwsem);
477 if (ret < 0) {
478 kfree(dif->path);
479 kfree(dif);
480 return ret;
481 }
482 ++sbi->devs->extra_devices;
483 break;
484 #ifdef CONFIG_EROFS_FS_ONDEMAND
485 case Opt_fsid:
486 kfree(sbi->fsid);
487 sbi->fsid = kstrdup(param->string, GFP_KERNEL);
488 if (!sbi->fsid)
489 return -ENOMEM;
490 break;
491 case Opt_domain_id:
492 kfree(sbi->domain_id);
493 sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
494 if (!sbi->domain_id)
495 return -ENOMEM;
496 break;
497 #else
498 case Opt_fsid:
499 case Opt_domain_id:
500 errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
501 break;
502 #endif
503 case Opt_directio:
504 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
505 if (result.boolean)
506 set_opt(&sbi->opt, DIRECT_IO);
507 else
508 clear_opt(&sbi->opt, DIRECT_IO);
509 #else
510 errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
511 #endif
512 break;
513 case Opt_fsoffset:
514 sbi->dif0.fsoff = result.uint_64;
515 break;
516 }
517 return 0;
518 }
519
erofs_encode_fh(struct inode * inode,u32 * fh,int * max_len,struct inode * parent)520 static int erofs_encode_fh(struct inode *inode, u32 *fh, int *max_len,
521 struct inode *parent)
522 {
523 erofs_nid_t nid = EROFS_I(inode)->nid;
524 int len = parent ? 6 : 3;
525
526 if (*max_len < len) {
527 *max_len = len;
528 return FILEID_INVALID;
529 }
530
531 fh[0] = (u32)(nid >> 32);
532 fh[1] = (u32)(nid & 0xffffffff);
533 fh[2] = inode->i_generation;
534
535 if (parent) {
536 nid = EROFS_I(parent)->nid;
537
538 fh[3] = (u32)(nid >> 32);
539 fh[4] = (u32)(nid & 0xffffffff);
540 fh[5] = parent->i_generation;
541 }
542
543 *max_len = len;
544 return parent ? FILEID_INO64_GEN_PARENT : FILEID_INO64_GEN;
545 }
546
erofs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)547 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
548 struct fid *fid, int fh_len, int fh_type)
549 {
550 if ((fh_type != FILEID_INO64_GEN &&
551 fh_type != FILEID_INO64_GEN_PARENT) || fh_len < 3)
552 return NULL;
553
554 return d_obtain_alias(erofs_iget(sb,
555 ((u64)fid->raw[0] << 32) | fid->raw[1]));
556 }
557
erofs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)558 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
559 struct fid *fid, int fh_len, int fh_type)
560 {
561 if (fh_type != FILEID_INO64_GEN_PARENT || fh_len < 6)
562 return NULL;
563
564 return d_obtain_alias(erofs_iget(sb,
565 ((u64)fid->raw[3] << 32) | fid->raw[4]));
566 }
567
erofs_get_parent(struct dentry * child)568 static struct dentry *erofs_get_parent(struct dentry *child)
569 {
570 erofs_nid_t nid;
571 unsigned int d_type;
572 int err;
573
574 err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
575 if (err)
576 return ERR_PTR(err);
577 return d_obtain_alias(erofs_iget(child->d_sb, nid));
578 }
579
580 static const struct export_operations erofs_export_ops = {
581 .encode_fh = erofs_encode_fh,
582 .fh_to_dentry = erofs_fh_to_dentry,
583 .fh_to_parent = erofs_fh_to_parent,
584 .get_parent = erofs_get_parent,
585 };
586
erofs_set_sysfs_name(struct super_block * sb)587 static void erofs_set_sysfs_name(struct super_block *sb)
588 {
589 struct erofs_sb_info *sbi = EROFS_SB(sb);
590
591 if (sbi->domain_id)
592 super_set_sysfs_name_generic(sb, "%s,%s", sbi->domain_id,
593 sbi->fsid);
594 else if (sbi->fsid)
595 super_set_sysfs_name_generic(sb, "%s", sbi->fsid);
596 else if (erofs_is_fileio_mode(sbi))
597 super_set_sysfs_name_generic(sb, "%s",
598 bdi_dev_name(sb->s_bdi));
599 else
600 super_set_sysfs_name_id(sb);
601 }
602
erofs_fc_fill_super(struct super_block * sb,struct fs_context * fc)603 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
604 {
605 struct inode *inode;
606 struct erofs_sb_info *sbi = EROFS_SB(sb);
607 int err;
608
609 sb->s_magic = EROFS_SUPER_MAGIC;
610 sb->s_flags |= SB_RDONLY | SB_NOATIME;
611 sb->s_maxbytes = MAX_LFS_FILESIZE;
612 sb->s_op = &erofs_sops;
613
614 sbi->blkszbits = PAGE_SHIFT;
615 if (!sb->s_bdev) {
616 sb->s_blocksize = PAGE_SIZE;
617 sb->s_blocksize_bits = PAGE_SHIFT;
618
619 if (erofs_is_fscache_mode(sb)) {
620 err = erofs_fscache_register_fs(sb);
621 if (err)
622 return err;
623 }
624 err = super_setup_bdi(sb);
625 if (err)
626 return err;
627 } else {
628 if (!sb_set_blocksize(sb, PAGE_SIZE)) {
629 errorfc(fc, "failed to set initial blksize");
630 return -EINVAL;
631 }
632
633 sbi->dif0.dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
634 &sbi->dif0.dax_part_off, NULL, NULL);
635 }
636
637 err = erofs_read_superblock(sb);
638 if (err)
639 return err;
640
641 if (sb->s_blocksize_bits != sbi->blkszbits) {
642 if (erofs_is_fscache_mode(sb)) {
643 errorfc(fc, "unsupported blksize for fscache mode");
644 return -EINVAL;
645 }
646
647 if (erofs_is_fileio_mode(sbi)) {
648 sb->s_blocksize = 1 << sbi->blkszbits;
649 sb->s_blocksize_bits = sbi->blkszbits;
650 } else if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
651 errorfc(fc, "failed to set erofs blksize");
652 return -EINVAL;
653 }
654 }
655
656 if (sbi->dif0.fsoff) {
657 if (sbi->dif0.fsoff & (sb->s_blocksize - 1))
658 return invalfc(fc, "fsoffset %llu is not aligned to block size %lu",
659 sbi->dif0.fsoff, sb->s_blocksize);
660 if (erofs_is_fscache_mode(sb))
661 return invalfc(fc, "cannot use fsoffset in fscache mode");
662 }
663
664 if (test_opt(&sbi->opt, DAX_ALWAYS)) {
665 if (!sbi->dif0.dax_dev) {
666 errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
667 clear_opt(&sbi->opt, DAX_ALWAYS);
668 } else if (sbi->blkszbits != PAGE_SHIFT) {
669 errorfc(fc, "unsupported blocksize for DAX");
670 clear_opt(&sbi->opt, DAX_ALWAYS);
671 }
672 }
673
674 sb->s_time_gran = 1;
675 sb->s_xattr = erofs_xattr_handlers;
676 sb->s_export_op = &erofs_export_ops;
677
678 if (test_opt(&sbi->opt, POSIX_ACL))
679 sb->s_flags |= SB_POSIXACL;
680 else
681 sb->s_flags &= ~SB_POSIXACL;
682
683 err = z_erofs_init_super(sb);
684 if (err)
685 return err;
686
687 if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
688 inode = erofs_iget(sb, sbi->packed_nid);
689 if (IS_ERR(inode))
690 return PTR_ERR(inode);
691 sbi->packed_inode = inode;
692 }
693
694 inode = erofs_iget(sb, sbi->root_nid);
695 if (IS_ERR(inode))
696 return PTR_ERR(inode);
697
698 if (!S_ISDIR(inode->i_mode)) {
699 erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
700 sbi->root_nid, inode->i_mode);
701 iput(inode);
702 return -EINVAL;
703 }
704 sb->s_root = d_make_root(inode);
705 if (!sb->s_root)
706 return -ENOMEM;
707
708 erofs_shrinker_register(sb);
709 err = erofs_xattr_prefixes_init(sb);
710 if (err)
711 return err;
712
713 erofs_set_sysfs_name(sb);
714 err = erofs_register_sysfs(sb);
715 if (err)
716 return err;
717
718 erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
719 return 0;
720 }
721
erofs_fc_get_tree(struct fs_context * fc)722 static int erofs_fc_get_tree(struct fs_context *fc)
723 {
724 struct erofs_sb_info *sbi = fc->s_fs_info;
725 int ret;
726
727 if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
728 return get_tree_nodev(fc, erofs_fc_fill_super);
729
730 ret = get_tree_bdev_flags(fc, erofs_fc_fill_super,
731 IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) ?
732 GET_TREE_BDEV_QUIET_LOOKUP : 0);
733 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
734 if (ret == -ENOTBLK) {
735 struct file *file;
736
737 if (!fc->source)
738 return invalf(fc, "No source specified");
739 file = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
740 if (IS_ERR(file))
741 return PTR_ERR(file);
742 sbi->dif0.file = file;
743
744 if (S_ISREG(file_inode(sbi->dif0.file)->i_mode) &&
745 sbi->dif0.file->f_mapping->a_ops->read_folio)
746 return get_tree_nodev(fc, erofs_fc_fill_super);
747 }
748 #endif
749 return ret;
750 }
751
erofs_fc_reconfigure(struct fs_context * fc)752 static int erofs_fc_reconfigure(struct fs_context *fc)
753 {
754 struct super_block *sb = fc->root->d_sb;
755 struct erofs_sb_info *sbi = EROFS_SB(sb);
756 struct erofs_sb_info *new_sbi = fc->s_fs_info;
757
758 DBG_BUGON(!sb_rdonly(sb));
759
760 if (new_sbi->fsid || new_sbi->domain_id)
761 erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
762
763 if (test_opt(&new_sbi->opt, POSIX_ACL))
764 fc->sb_flags |= SB_POSIXACL;
765 else
766 fc->sb_flags &= ~SB_POSIXACL;
767
768 sbi->opt = new_sbi->opt;
769
770 fc->sb_flags |= SB_RDONLY;
771 return 0;
772 }
773
erofs_release_device_info(int id,void * ptr,void * data)774 static int erofs_release_device_info(int id, void *ptr, void *data)
775 {
776 struct erofs_device_info *dif = ptr;
777
778 fs_put_dax(dif->dax_dev, NULL);
779 if (dif->file)
780 fput(dif->file);
781 erofs_fscache_unregister_cookie(dif->fscache);
782 dif->fscache = NULL;
783 kfree(dif->path);
784 kfree(dif);
785 return 0;
786 }
787
erofs_free_dev_context(struct erofs_dev_context * devs)788 static void erofs_free_dev_context(struct erofs_dev_context *devs)
789 {
790 if (!devs)
791 return;
792 idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
793 idr_destroy(&devs->tree);
794 kfree(devs);
795 }
796
erofs_sb_free(struct erofs_sb_info * sbi)797 static void erofs_sb_free(struct erofs_sb_info *sbi)
798 {
799 erofs_free_dev_context(sbi->devs);
800 kfree(sbi->fsid);
801 kfree(sbi->domain_id);
802 if (sbi->dif0.file)
803 fput(sbi->dif0.file);
804 kfree(sbi);
805 }
806
erofs_fc_free(struct fs_context * fc)807 static void erofs_fc_free(struct fs_context *fc)
808 {
809 struct erofs_sb_info *sbi = fc->s_fs_info;
810
811 if (sbi) /* free here if an error occurs before transferring to sb */
812 erofs_sb_free(sbi);
813 }
814
815 static const struct fs_context_operations erofs_context_ops = {
816 .parse_param = erofs_fc_parse_param,
817 .get_tree = erofs_fc_get_tree,
818 .reconfigure = erofs_fc_reconfigure,
819 .free = erofs_fc_free,
820 };
821
erofs_init_fs_context(struct fs_context * fc)822 static int erofs_init_fs_context(struct fs_context *fc)
823 {
824 struct erofs_sb_info *sbi;
825
826 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
827 if (!sbi)
828 return -ENOMEM;
829
830 sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
831 if (!sbi->devs) {
832 kfree(sbi);
833 return -ENOMEM;
834 }
835 fc->s_fs_info = sbi;
836
837 idr_init(&sbi->devs->tree);
838 init_rwsem(&sbi->devs->rwsem);
839 erofs_default_options(sbi);
840 fc->ops = &erofs_context_ops;
841 return 0;
842 }
843
erofs_drop_internal_inodes(struct erofs_sb_info * sbi)844 static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
845 {
846 iput(sbi->packed_inode);
847 sbi->packed_inode = NULL;
848 #ifdef CONFIG_EROFS_FS_ZIP
849 iput(sbi->managed_cache);
850 sbi->managed_cache = NULL;
851 #endif
852 }
853
erofs_kill_sb(struct super_block * sb)854 static void erofs_kill_sb(struct super_block *sb)
855 {
856 struct erofs_sb_info *sbi = EROFS_SB(sb);
857
858 if ((IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid) ||
859 sbi->dif0.file)
860 kill_anon_super(sb);
861 else
862 kill_block_super(sb);
863 erofs_drop_internal_inodes(sbi);
864 fs_put_dax(sbi->dif0.dax_dev, NULL);
865 erofs_fscache_unregister_fs(sb);
866 erofs_sb_free(sbi);
867 sb->s_fs_info = NULL;
868 }
869
erofs_put_super(struct super_block * sb)870 static void erofs_put_super(struct super_block *sb)
871 {
872 struct erofs_sb_info *const sbi = EROFS_SB(sb);
873
874 erofs_unregister_sysfs(sb);
875 erofs_shrinker_unregister(sb);
876 erofs_xattr_prefixes_cleanup(sb);
877 erofs_drop_internal_inodes(sbi);
878 erofs_free_dev_context(sbi->devs);
879 sbi->devs = NULL;
880 erofs_fscache_unregister_fs(sb);
881 }
882
883 static struct file_system_type erofs_fs_type = {
884 .owner = THIS_MODULE,
885 .name = "erofs",
886 .init_fs_context = erofs_init_fs_context,
887 .kill_sb = erofs_kill_sb,
888 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
889 };
890 MODULE_ALIAS_FS("erofs");
891
erofs_module_init(void)892 static int __init erofs_module_init(void)
893 {
894 int err;
895
896 erofs_check_ondisk_layout_definitions();
897
898 erofs_inode_cachep = kmem_cache_create("erofs_inode",
899 sizeof(struct erofs_inode), 0,
900 SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
901 erofs_inode_init_once);
902 if (!erofs_inode_cachep)
903 return -ENOMEM;
904
905 err = erofs_init_shrinker();
906 if (err)
907 goto shrinker_err;
908
909 err = z_erofs_init_subsystem();
910 if (err)
911 goto zip_err;
912
913 err = erofs_init_sysfs();
914 if (err)
915 goto sysfs_err;
916
917 err = register_filesystem(&erofs_fs_type);
918 if (err)
919 goto fs_err;
920
921 return 0;
922
923 fs_err:
924 erofs_exit_sysfs();
925 sysfs_err:
926 z_erofs_exit_subsystem();
927 zip_err:
928 erofs_exit_shrinker();
929 shrinker_err:
930 kmem_cache_destroy(erofs_inode_cachep);
931 return err;
932 }
933
erofs_module_exit(void)934 static void __exit erofs_module_exit(void)
935 {
936 unregister_filesystem(&erofs_fs_type);
937
938 /* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
939 rcu_barrier();
940
941 erofs_exit_sysfs();
942 z_erofs_exit_subsystem();
943 erofs_exit_shrinker();
944 kmem_cache_destroy(erofs_inode_cachep);
945 }
946
erofs_statfs(struct dentry * dentry,struct kstatfs * buf)947 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
948 {
949 struct super_block *sb = dentry->d_sb;
950 struct erofs_sb_info *sbi = EROFS_SB(sb);
951
952 buf->f_type = sb->s_magic;
953 buf->f_bsize = sb->s_blocksize;
954 buf->f_blocks = sbi->total_blocks;
955 buf->f_bfree = buf->f_bavail = 0;
956 buf->f_files = ULLONG_MAX;
957 buf->f_ffree = ULLONG_MAX - sbi->inos;
958 buf->f_namelen = EROFS_NAME_LEN;
959
960 if (uuid_is_null(&sb->s_uuid))
961 buf->f_fsid = u64_to_fsid(!sb->s_bdev ? 0 :
962 huge_encode_dev(sb->s_bdev->bd_dev));
963 else
964 buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
965 return 0;
966 }
967
erofs_show_options(struct seq_file * seq,struct dentry * root)968 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
969 {
970 struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
971 struct erofs_mount_opts *opt = &sbi->opt;
972
973 if (IS_ENABLED(CONFIG_EROFS_FS_XATTR))
974 seq_puts(seq, test_opt(opt, XATTR_USER) ?
975 ",user_xattr" : ",nouser_xattr");
976 if (IS_ENABLED(CONFIG_EROFS_FS_POSIX_ACL))
977 seq_puts(seq, test_opt(opt, POSIX_ACL) ? ",acl" : ",noacl");
978 if (IS_ENABLED(CONFIG_EROFS_FS_ZIP))
979 seq_printf(seq, ",cache_strategy=%s",
980 erofs_param_cache_strategy[opt->cache_strategy].name);
981 if (test_opt(opt, DAX_ALWAYS))
982 seq_puts(seq, ",dax=always");
983 if (test_opt(opt, DAX_NEVER))
984 seq_puts(seq, ",dax=never");
985 if (erofs_is_fileio_mode(sbi) && test_opt(opt, DIRECT_IO))
986 seq_puts(seq, ",directio");
987 #ifdef CONFIG_EROFS_FS_ONDEMAND
988 if (sbi->fsid)
989 seq_printf(seq, ",fsid=%s", sbi->fsid);
990 if (sbi->domain_id)
991 seq_printf(seq, ",domain_id=%s", sbi->domain_id);
992 #endif
993 if (sbi->dif0.fsoff)
994 seq_printf(seq, ",fsoffset=%llu", sbi->dif0.fsoff);
995 return 0;
996 }
997
998 const struct super_operations erofs_sops = {
999 .put_super = erofs_put_super,
1000 .alloc_inode = erofs_alloc_inode,
1001 .free_inode = erofs_free_inode,
1002 .statfs = erofs_statfs,
1003 .show_options = erofs_show_options,
1004 };
1005
1006 module_init(erofs_module_init);
1007 module_exit(erofs_module_exit);
1008
1009 MODULE_DESCRIPTION("Enhanced ROM File System");
1010 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
1011 MODULE_LICENSE("GPL");
1012