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_err(struct super_block * sb,const char * func,const char * fmt,...)21 void _erofs_err(struct super_block *sb, const char *func, const char *fmt, ...)
22 {
23 struct va_format vaf;
24 va_list args;
25
26 va_start(args, fmt);
27
28 vaf.fmt = fmt;
29 vaf.va = &args;
30
31 if (sb)
32 pr_err("(device %s): %s: %pV", sb->s_id, func, &vaf);
33 else
34 pr_err("%s: %pV", func, &vaf);
35 va_end(args);
36 }
37
_erofs_info(struct super_block * sb,const char * func,const char * fmt,...)38 void _erofs_info(struct super_block *sb, const char *func, const char *fmt, ...)
39 {
40 struct va_format vaf;
41 va_list args;
42
43 va_start(args, fmt);
44
45 vaf.fmt = fmt;
46 vaf.va = &args;
47
48 if (sb)
49 pr_info("(device %s): %pV", sb->s_id, &vaf);
50 else
51 pr_info("%pV", &vaf);
52 va_end(args);
53 }
54
erofs_superblock_csum_verify(struct super_block * sb,void * sbdata)55 static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
56 {
57 size_t len = 1 << EROFS_SB(sb)->blkszbits;
58 struct erofs_super_block *dsb;
59 u32 expected_crc, crc;
60
61 if (len > EROFS_SUPER_OFFSET)
62 len -= EROFS_SUPER_OFFSET;
63
64 dsb = kmemdup(sbdata + EROFS_SUPER_OFFSET, len, GFP_KERNEL);
65 if (!dsb)
66 return -ENOMEM;
67
68 expected_crc = le32_to_cpu(dsb->checksum);
69 dsb->checksum = 0;
70 /* to allow for x86 boot sectors and other oddities. */
71 crc = crc32c(~0, dsb, len);
72 kfree(dsb);
73
74 if (crc != expected_crc) {
75 erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
76 crc, expected_crc);
77 return -EBADMSG;
78 }
79 return 0;
80 }
81
erofs_inode_init_once(void * ptr)82 static void erofs_inode_init_once(void *ptr)
83 {
84 struct erofs_inode *vi = ptr;
85
86 inode_init_once(&vi->vfs_inode);
87 }
88
erofs_alloc_inode(struct super_block * sb)89 static struct inode *erofs_alloc_inode(struct super_block *sb)
90 {
91 struct erofs_inode *vi =
92 alloc_inode_sb(sb, erofs_inode_cachep, GFP_KERNEL);
93
94 if (!vi)
95 return NULL;
96
97 /* zero out everything except vfs_inode */
98 memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
99 return &vi->vfs_inode;
100 }
101
erofs_free_inode(struct inode * inode)102 static void erofs_free_inode(struct inode *inode)
103 {
104 struct erofs_inode *vi = EROFS_I(inode);
105
106 if (inode->i_op == &erofs_fast_symlink_iops)
107 kfree(inode->i_link);
108 kfree(vi->xattr_shared_xattrs);
109 kmem_cache_free(erofs_inode_cachep, vi);
110 }
111
112 /* 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)113 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
114 erofs_off_t *offset, int *lengthp)
115 {
116 u8 *buffer, *ptr;
117 int len, i, cnt;
118
119 *offset = round_up(*offset, 4);
120 ptr = erofs_bread(buf, *offset, EROFS_KMAP);
121 if (IS_ERR(ptr))
122 return ptr;
123
124 len = le16_to_cpu(*(__le16 *)ptr);
125 if (!len)
126 len = U16_MAX + 1;
127 buffer = kmalloc(len, GFP_KERNEL);
128 if (!buffer)
129 return ERR_PTR(-ENOMEM);
130 *offset += sizeof(__le16);
131 *lengthp = len;
132
133 for (i = 0; i < len; i += cnt) {
134 cnt = min_t(int, sb->s_blocksize - erofs_blkoff(sb, *offset),
135 len - i);
136 ptr = erofs_bread(buf, *offset, EROFS_KMAP);
137 if (IS_ERR(ptr)) {
138 kfree(buffer);
139 return ptr;
140 }
141 memcpy(buffer + i, ptr, cnt);
142 *offset += cnt;
143 }
144 return buffer;
145 }
146
147 #ifndef CONFIG_EROFS_FS_ZIP
z_erofs_parse_cfgs(struct super_block * sb,struct erofs_super_block * dsb)148 static int z_erofs_parse_cfgs(struct super_block *sb,
149 struct erofs_super_block *dsb)
150 {
151 if (!dsb->u1.available_compr_algs)
152 return 0;
153
154 erofs_err(sb, "compression disabled, unable to mount compressed EROFS");
155 return -EOPNOTSUPP;
156 }
157 #endif
158
erofs_init_device(struct erofs_buf * buf,struct super_block * sb,struct erofs_device_info * dif,erofs_off_t * pos)159 static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
160 struct erofs_device_info *dif, erofs_off_t *pos)
161 {
162 struct erofs_sb_info *sbi = EROFS_SB(sb);
163 struct erofs_fscache *fscache;
164 struct erofs_deviceslot *dis;
165 struct file *file;
166
167 dis = erofs_read_metabuf(buf, sb, *pos, EROFS_KMAP);
168 if (IS_ERR(dis))
169 return PTR_ERR(dis);
170
171 if (!sbi->devs->flatdev && !dif->path) {
172 if (!dis->tag[0]) {
173 erofs_err(sb, "empty device tag @ pos %llu", *pos);
174 return -EINVAL;
175 }
176 dif->path = kmemdup_nul(dis->tag, sizeof(dis->tag), GFP_KERNEL);
177 if (!dif->path)
178 return -ENOMEM;
179 }
180
181 if (erofs_is_fscache_mode(sb)) {
182 fscache = erofs_fscache_register_cookie(sb, dif->path, 0);
183 if (IS_ERR(fscache))
184 return PTR_ERR(fscache);
185 dif->fscache = fscache;
186 } else if (!sbi->devs->flatdev) {
187 file = erofs_is_fileio_mode(sbi) ?
188 filp_open(dif->path, O_RDONLY | O_LARGEFILE, 0) :
189 bdev_file_open_by_path(dif->path,
190 BLK_OPEN_READ, sb->s_type, NULL);
191 if (IS_ERR(file))
192 return PTR_ERR(file);
193
194 dif->file = file;
195 if (!erofs_is_fileio_mode(sbi))
196 dif->dax_dev = fs_dax_get_by_bdev(file_bdev(file),
197 &dif->dax_part_off, NULL, NULL);
198 }
199
200 dif->blocks = le32_to_cpu(dis->blocks);
201 dif->mapped_blkaddr = le32_to_cpu(dis->mapped_blkaddr);
202 sbi->total_blocks += dif->blocks;
203 *pos += EROFS_DEVT_SLOT_SIZE;
204 return 0;
205 }
206
erofs_scan_devices(struct super_block * sb,struct erofs_super_block * dsb)207 static int erofs_scan_devices(struct super_block *sb,
208 struct erofs_super_block *dsb)
209 {
210 struct erofs_sb_info *sbi = EROFS_SB(sb);
211 unsigned int ondisk_extradevs;
212 erofs_off_t pos;
213 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
214 struct erofs_device_info *dif;
215 int id, err = 0;
216
217 sbi->total_blocks = sbi->primarydevice_blocks;
218 if (!erofs_sb_has_device_table(sbi))
219 ondisk_extradevs = 0;
220 else
221 ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
222
223 if (sbi->devs->extra_devices &&
224 ondisk_extradevs != sbi->devs->extra_devices) {
225 erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
226 ondisk_extradevs, sbi->devs->extra_devices);
227 return -EINVAL;
228 }
229 if (!ondisk_extradevs)
230 return 0;
231
232 if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
233 sbi->devs->flatdev = true;
234
235 sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
236 pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
237 down_read(&sbi->devs->rwsem);
238 if (sbi->devs->extra_devices) {
239 idr_for_each_entry(&sbi->devs->tree, dif, id) {
240 err = erofs_init_device(&buf, sb, dif, &pos);
241 if (err)
242 break;
243 }
244 } else {
245 for (id = 0; id < ondisk_extradevs; id++) {
246 dif = kzalloc(sizeof(*dif), GFP_KERNEL);
247 if (!dif) {
248 err = -ENOMEM;
249 break;
250 }
251
252 err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
253 if (err < 0) {
254 kfree(dif);
255 break;
256 }
257 ++sbi->devs->extra_devices;
258
259 err = erofs_init_device(&buf, sb, dif, &pos);
260 if (err)
261 break;
262 }
263 }
264 up_read(&sbi->devs->rwsem);
265 erofs_put_metabuf(&buf);
266 return err;
267 }
268
erofs_read_superblock(struct super_block * sb)269 static int erofs_read_superblock(struct super_block *sb)
270 {
271 struct erofs_sb_info *sbi = EROFS_SB(sb);
272 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
273 struct erofs_super_block *dsb;
274 void *data;
275 int ret;
276
277 data = erofs_read_metabuf(&buf, sb, 0, EROFS_KMAP);
278 if (IS_ERR(data)) {
279 erofs_err(sb, "cannot read erofs superblock");
280 return PTR_ERR(data);
281 }
282
283 dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
284 ret = -EINVAL;
285 if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
286 erofs_err(sb, "cannot find valid erofs superblock");
287 goto out;
288 }
289
290 sbi->blkszbits = dsb->blkszbits;
291 if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
292 erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
293 goto out;
294 }
295 if (dsb->dirblkbits) {
296 erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
297 goto out;
298 }
299
300 sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
301 if (erofs_sb_has_sb_chksum(sbi)) {
302 ret = erofs_superblock_csum_verify(sb, data);
303 if (ret)
304 goto out;
305 }
306
307 ret = -EINVAL;
308 sbi->feature_incompat = le32_to_cpu(dsb->feature_incompat);
309 if (sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT) {
310 erofs_err(sb, "unidentified incompatible feature %x, please upgrade kernel",
311 sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT);
312 goto out;
313 }
314
315 sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
316 if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
317 erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
318 sbi->sb_size);
319 goto out;
320 }
321 sbi->primarydevice_blocks = le32_to_cpu(dsb->blocks);
322 sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
323 #ifdef CONFIG_EROFS_FS_XATTR
324 sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
325 sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
326 sbi->xattr_prefix_count = dsb->xattr_prefix_count;
327 sbi->xattr_filter_reserved = dsb->xattr_filter_reserved;
328 #endif
329 sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
330 sbi->root_nid = le16_to_cpu(dsb->root_nid);
331 sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
332 sbi->inos = le64_to_cpu(dsb->inos);
333
334 sbi->build_time = le64_to_cpu(dsb->build_time);
335 sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
336
337 super_set_uuid(sb, (void *)dsb->uuid, sizeof(dsb->uuid));
338
339 ret = strscpy(sbi->volume_name, dsb->volume_name,
340 sizeof(dsb->volume_name));
341 if (ret < 0) { /* -E2BIG */
342 erofs_err(sb, "bad volume name without NIL terminator");
343 ret = -EFSCORRUPTED;
344 goto out;
345 }
346
347 /* parse on-disk compression configurations */
348 ret = z_erofs_parse_cfgs(sb, dsb);
349 if (ret < 0)
350 goto out;
351
352 /* handle multiple devices */
353 ret = erofs_scan_devices(sb, dsb);
354
355 if (erofs_is_fscache_mode(sb))
356 erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
357 out:
358 erofs_put_metabuf(&buf);
359 return ret;
360 }
361
erofs_default_options(struct erofs_sb_info * sbi)362 static void erofs_default_options(struct erofs_sb_info *sbi)
363 {
364 #ifdef CONFIG_EROFS_FS_ZIP
365 sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
366 sbi->opt.max_sync_decompress_pages = 3;
367 sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
368 #endif
369 #ifdef CONFIG_EROFS_FS_XATTR
370 set_opt(&sbi->opt, XATTR_USER);
371 #endif
372 #ifdef CONFIG_EROFS_FS_POSIX_ACL
373 set_opt(&sbi->opt, POSIX_ACL);
374 #endif
375 }
376
377 enum {
378 Opt_user_xattr,
379 Opt_acl,
380 Opt_cache_strategy,
381 Opt_dax,
382 Opt_dax_enum,
383 Opt_device,
384 Opt_fsid,
385 Opt_domain_id,
386 Opt_err
387 };
388
389 static const struct constant_table erofs_param_cache_strategy[] = {
390 {"disabled", EROFS_ZIP_CACHE_DISABLED},
391 {"readahead", EROFS_ZIP_CACHE_READAHEAD},
392 {"readaround", EROFS_ZIP_CACHE_READAROUND},
393 {}
394 };
395
396 static const struct constant_table erofs_dax_param_enums[] = {
397 {"always", EROFS_MOUNT_DAX_ALWAYS},
398 {"never", EROFS_MOUNT_DAX_NEVER},
399 {}
400 };
401
402 static const struct fs_parameter_spec erofs_fs_parameters[] = {
403 fsparam_flag_no("user_xattr", Opt_user_xattr),
404 fsparam_flag_no("acl", Opt_acl),
405 fsparam_enum("cache_strategy", Opt_cache_strategy,
406 erofs_param_cache_strategy),
407 fsparam_flag("dax", Opt_dax),
408 fsparam_enum("dax", Opt_dax_enum, erofs_dax_param_enums),
409 fsparam_string("device", Opt_device),
410 fsparam_string("fsid", Opt_fsid),
411 fsparam_string("domain_id", Opt_domain_id),
412 {}
413 };
414
erofs_fc_set_dax_mode(struct fs_context * fc,unsigned int mode)415 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
416 {
417 #ifdef CONFIG_FS_DAX
418 struct erofs_sb_info *sbi = fc->s_fs_info;
419
420 switch (mode) {
421 case EROFS_MOUNT_DAX_ALWAYS:
422 set_opt(&sbi->opt, DAX_ALWAYS);
423 clear_opt(&sbi->opt, DAX_NEVER);
424 return true;
425 case EROFS_MOUNT_DAX_NEVER:
426 set_opt(&sbi->opt, DAX_NEVER);
427 clear_opt(&sbi->opt, DAX_ALWAYS);
428 return true;
429 default:
430 DBG_BUGON(1);
431 return false;
432 }
433 #else
434 errorfc(fc, "dax options not supported");
435 return false;
436 #endif
437 }
438
erofs_fc_parse_param(struct fs_context * fc,struct fs_parameter * param)439 static int erofs_fc_parse_param(struct fs_context *fc,
440 struct fs_parameter *param)
441 {
442 struct erofs_sb_info *sbi = fc->s_fs_info;
443 struct fs_parse_result result;
444 struct erofs_device_info *dif;
445 int opt, ret;
446
447 opt = fs_parse(fc, erofs_fs_parameters, param, &result);
448 if (opt < 0)
449 return opt;
450
451 switch (opt) {
452 case Opt_user_xattr:
453 #ifdef CONFIG_EROFS_FS_XATTR
454 if (result.boolean)
455 set_opt(&sbi->opt, XATTR_USER);
456 else
457 clear_opt(&sbi->opt, XATTR_USER);
458 #else
459 errorfc(fc, "{,no}user_xattr options not supported");
460 #endif
461 break;
462 case Opt_acl:
463 #ifdef CONFIG_EROFS_FS_POSIX_ACL
464 if (result.boolean)
465 set_opt(&sbi->opt, POSIX_ACL);
466 else
467 clear_opt(&sbi->opt, POSIX_ACL);
468 #else
469 errorfc(fc, "{,no}acl options not supported");
470 #endif
471 break;
472 case Opt_cache_strategy:
473 #ifdef CONFIG_EROFS_FS_ZIP
474 sbi->opt.cache_strategy = result.uint_32;
475 #else
476 errorfc(fc, "compression not supported, cache_strategy ignored");
477 #endif
478 break;
479 case Opt_dax:
480 if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
481 return -EINVAL;
482 break;
483 case Opt_dax_enum:
484 if (!erofs_fc_set_dax_mode(fc, result.uint_32))
485 return -EINVAL;
486 break;
487 case Opt_device:
488 dif = kzalloc(sizeof(*dif), GFP_KERNEL);
489 if (!dif)
490 return -ENOMEM;
491 dif->path = kstrdup(param->string, GFP_KERNEL);
492 if (!dif->path) {
493 kfree(dif);
494 return -ENOMEM;
495 }
496 down_write(&sbi->devs->rwsem);
497 ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
498 up_write(&sbi->devs->rwsem);
499 if (ret < 0) {
500 kfree(dif->path);
501 kfree(dif);
502 return ret;
503 }
504 ++sbi->devs->extra_devices;
505 break;
506 #ifdef CONFIG_EROFS_FS_ONDEMAND
507 case Opt_fsid:
508 kfree(sbi->fsid);
509 sbi->fsid = kstrdup(param->string, GFP_KERNEL);
510 if (!sbi->fsid)
511 return -ENOMEM;
512 break;
513 case Opt_domain_id:
514 kfree(sbi->domain_id);
515 sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
516 if (!sbi->domain_id)
517 return -ENOMEM;
518 break;
519 #else
520 case Opt_fsid:
521 case Opt_domain_id:
522 errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
523 break;
524 #endif
525 default:
526 return -ENOPARAM;
527 }
528 return 0;
529 }
530
erofs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)531 static struct inode *erofs_nfs_get_inode(struct super_block *sb,
532 u64 ino, u32 generation)
533 {
534 return erofs_iget(sb, ino);
535 }
536
erofs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)537 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
538 struct fid *fid, int fh_len, int fh_type)
539 {
540 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
541 erofs_nfs_get_inode);
542 }
543
erofs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)544 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
545 struct fid *fid, int fh_len, int fh_type)
546 {
547 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
548 erofs_nfs_get_inode);
549 }
550
erofs_get_parent(struct dentry * child)551 static struct dentry *erofs_get_parent(struct dentry *child)
552 {
553 erofs_nid_t nid;
554 unsigned int d_type;
555 int err;
556
557 err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
558 if (err)
559 return ERR_PTR(err);
560 return d_obtain_alias(erofs_iget(child->d_sb, nid));
561 }
562
563 static const struct export_operations erofs_export_ops = {
564 .encode_fh = generic_encode_ino32_fh,
565 .fh_to_dentry = erofs_fh_to_dentry,
566 .fh_to_parent = erofs_fh_to_parent,
567 .get_parent = erofs_get_parent,
568 };
569
erofs_set_sysfs_name(struct super_block * sb)570 static void erofs_set_sysfs_name(struct super_block *sb)
571 {
572 struct erofs_sb_info *sbi = EROFS_SB(sb);
573
574 if (sbi->domain_id)
575 super_set_sysfs_name_generic(sb, "%s,%s", sbi->domain_id,
576 sbi->fsid);
577 else if (sbi->fsid)
578 super_set_sysfs_name_generic(sb, "%s", sbi->fsid);
579 else if (erofs_is_fileio_mode(sbi))
580 super_set_sysfs_name_generic(sb, "%s",
581 bdi_dev_name(sb->s_bdi));
582 else
583 super_set_sysfs_name_id(sb);
584 }
585
erofs_fc_fill_super(struct super_block * sb,struct fs_context * fc)586 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
587 {
588 struct inode *inode;
589 struct erofs_sb_info *sbi = EROFS_SB(sb);
590 int err;
591
592 sb->s_magic = EROFS_SUPER_MAGIC;
593 sb->s_flags |= SB_RDONLY | SB_NOATIME;
594 sb->s_maxbytes = MAX_LFS_FILESIZE;
595 sb->s_op = &erofs_sops;
596
597 sbi->blkszbits = PAGE_SHIFT;
598 if (!sb->s_bdev) {
599 sb->s_blocksize = PAGE_SIZE;
600 sb->s_blocksize_bits = PAGE_SHIFT;
601
602 if (erofs_is_fscache_mode(sb)) {
603 err = erofs_fscache_register_fs(sb);
604 if (err)
605 return err;
606 }
607 err = super_setup_bdi(sb);
608 if (err)
609 return err;
610 } else {
611 if (!sb_set_blocksize(sb, PAGE_SIZE)) {
612 errorfc(fc, "failed to set initial blksize");
613 return -EINVAL;
614 }
615
616 sbi->dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
617 &sbi->dax_part_off,
618 NULL, NULL);
619 }
620
621 err = erofs_read_superblock(sb);
622 if (err)
623 return err;
624
625 if (sb->s_blocksize_bits != sbi->blkszbits) {
626 if (erofs_is_fscache_mode(sb)) {
627 errorfc(fc, "unsupported blksize for fscache mode");
628 return -EINVAL;
629 }
630 if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
631 errorfc(fc, "failed to set erofs blksize");
632 return -EINVAL;
633 }
634 }
635
636 if (test_opt(&sbi->opt, DAX_ALWAYS)) {
637 if (!sbi->dax_dev) {
638 errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
639 clear_opt(&sbi->opt, DAX_ALWAYS);
640 } else if (sbi->blkszbits != PAGE_SHIFT) {
641 errorfc(fc, "unsupported blocksize for DAX");
642 clear_opt(&sbi->opt, DAX_ALWAYS);
643 }
644 }
645
646 sb->s_time_gran = 1;
647 sb->s_xattr = erofs_xattr_handlers;
648 sb->s_export_op = &erofs_export_ops;
649
650 if (test_opt(&sbi->opt, POSIX_ACL))
651 sb->s_flags |= SB_POSIXACL;
652 else
653 sb->s_flags &= ~SB_POSIXACL;
654
655 #ifdef CONFIG_EROFS_FS_ZIP
656 xa_init(&sbi->managed_pslots);
657 #endif
658
659 inode = erofs_iget(sb, sbi->root_nid);
660 if (IS_ERR(inode))
661 return PTR_ERR(inode);
662
663 if (!S_ISDIR(inode->i_mode)) {
664 erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
665 sbi->root_nid, inode->i_mode);
666 iput(inode);
667 return -EINVAL;
668 }
669
670 sb->s_root = d_make_root(inode);
671 if (!sb->s_root)
672 return -ENOMEM;
673
674 erofs_shrinker_register(sb);
675 if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
676 sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
677 if (IS_ERR(sbi->packed_inode)) {
678 err = PTR_ERR(sbi->packed_inode);
679 sbi->packed_inode = NULL;
680 return err;
681 }
682 }
683 err = erofs_init_managed_cache(sb);
684 if (err)
685 return err;
686
687 err = erofs_xattr_prefixes_init(sb);
688 if (err)
689 return err;
690
691 erofs_set_sysfs_name(sb);
692 err = erofs_register_sysfs(sb);
693 if (err)
694 return err;
695
696 erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
697 return 0;
698 }
699
erofs_fc_get_tree(struct fs_context * fc)700 static int erofs_fc_get_tree(struct fs_context *fc)
701 {
702 struct erofs_sb_info *sbi = fc->s_fs_info;
703 int ret;
704
705 if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
706 return get_tree_nodev(fc, erofs_fc_fill_super);
707
708 ret = get_tree_bdev(fc, erofs_fc_fill_super);
709 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
710 if (ret == -ENOTBLK) {
711 if (!fc->source)
712 return invalf(fc, "No source specified");
713 sbi->fdev = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
714 if (IS_ERR(sbi->fdev))
715 return PTR_ERR(sbi->fdev);
716
717 return get_tree_nodev(fc, erofs_fc_fill_super);
718 }
719 #endif
720 return ret;
721 }
722
erofs_fc_reconfigure(struct fs_context * fc)723 static int erofs_fc_reconfigure(struct fs_context *fc)
724 {
725 struct super_block *sb = fc->root->d_sb;
726 struct erofs_sb_info *sbi = EROFS_SB(sb);
727 struct erofs_sb_info *new_sbi = fc->s_fs_info;
728
729 DBG_BUGON(!sb_rdonly(sb));
730
731 if (new_sbi->fsid || new_sbi->domain_id)
732 erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
733
734 if (test_opt(&new_sbi->opt, POSIX_ACL))
735 fc->sb_flags |= SB_POSIXACL;
736 else
737 fc->sb_flags &= ~SB_POSIXACL;
738
739 sbi->opt = new_sbi->opt;
740
741 fc->sb_flags |= SB_RDONLY;
742 return 0;
743 }
744
erofs_release_device_info(int id,void * ptr,void * data)745 static int erofs_release_device_info(int id, void *ptr, void *data)
746 {
747 struct erofs_device_info *dif = ptr;
748
749 fs_put_dax(dif->dax_dev, NULL);
750 if (dif->file)
751 fput(dif->file);
752 erofs_fscache_unregister_cookie(dif->fscache);
753 dif->fscache = NULL;
754 kfree(dif->path);
755 kfree(dif);
756 return 0;
757 }
758
erofs_free_dev_context(struct erofs_dev_context * devs)759 static void erofs_free_dev_context(struct erofs_dev_context *devs)
760 {
761 if (!devs)
762 return;
763 idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
764 idr_destroy(&devs->tree);
765 kfree(devs);
766 }
767
erofs_fc_free(struct fs_context * fc)768 static void erofs_fc_free(struct fs_context *fc)
769 {
770 struct erofs_sb_info *sbi = fc->s_fs_info;
771
772 if (!sbi)
773 return;
774
775 erofs_free_dev_context(sbi->devs);
776 kfree(sbi->fsid);
777 kfree(sbi->domain_id);
778 kfree(sbi);
779 }
780
781 static const struct fs_context_operations erofs_context_ops = {
782 .parse_param = erofs_fc_parse_param,
783 .get_tree = erofs_fc_get_tree,
784 .reconfigure = erofs_fc_reconfigure,
785 .free = erofs_fc_free,
786 };
787
erofs_init_fs_context(struct fs_context * fc)788 static int erofs_init_fs_context(struct fs_context *fc)
789 {
790 struct erofs_sb_info *sbi;
791
792 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
793 if (!sbi)
794 return -ENOMEM;
795
796 sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
797 if (!sbi->devs) {
798 kfree(sbi);
799 return -ENOMEM;
800 }
801 fc->s_fs_info = sbi;
802
803 idr_init(&sbi->devs->tree);
804 init_rwsem(&sbi->devs->rwsem);
805 erofs_default_options(sbi);
806 fc->ops = &erofs_context_ops;
807 return 0;
808 }
809
erofs_kill_sb(struct super_block * sb)810 static void erofs_kill_sb(struct super_block *sb)
811 {
812 struct erofs_sb_info *sbi = EROFS_SB(sb);
813
814 if ((IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid) || sbi->fdev)
815 kill_anon_super(sb);
816 else
817 kill_block_super(sb);
818
819 erofs_free_dev_context(sbi->devs);
820 fs_put_dax(sbi->dax_dev, NULL);
821 erofs_fscache_unregister_fs(sb);
822 kfree(sbi->fsid);
823 kfree(sbi->domain_id);
824 if (sbi->fdev)
825 fput(sbi->fdev);
826 kfree(sbi);
827 sb->s_fs_info = NULL;
828 }
829
erofs_put_super(struct super_block * sb)830 static void erofs_put_super(struct super_block *sb)
831 {
832 struct erofs_sb_info *const sbi = EROFS_SB(sb);
833
834 DBG_BUGON(!sbi);
835
836 erofs_unregister_sysfs(sb);
837 erofs_shrinker_unregister(sb);
838 erofs_xattr_prefixes_cleanup(sb);
839 #ifdef CONFIG_EROFS_FS_ZIP
840 iput(sbi->managed_cache);
841 sbi->managed_cache = NULL;
842 #endif
843 iput(sbi->packed_inode);
844 sbi->packed_inode = NULL;
845 erofs_free_dev_context(sbi->devs);
846 sbi->devs = NULL;
847 erofs_fscache_unregister_fs(sb);
848 }
849
850 static struct file_system_type erofs_fs_type = {
851 .owner = THIS_MODULE,
852 .name = "erofs",
853 .init_fs_context = erofs_init_fs_context,
854 .kill_sb = erofs_kill_sb,
855 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
856 };
857 MODULE_ALIAS_FS("erofs");
858
erofs_module_init(void)859 static int __init erofs_module_init(void)
860 {
861 int err;
862
863 erofs_check_ondisk_layout_definitions();
864
865 erofs_inode_cachep = kmem_cache_create("erofs_inode",
866 sizeof(struct erofs_inode), 0,
867 SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
868 erofs_inode_init_once);
869 if (!erofs_inode_cachep)
870 return -ENOMEM;
871
872 err = erofs_init_shrinker();
873 if (err)
874 goto shrinker_err;
875
876 err = z_erofs_init_subsystem();
877 if (err)
878 goto zip_err;
879
880 err = erofs_init_sysfs();
881 if (err)
882 goto sysfs_err;
883
884 err = register_filesystem(&erofs_fs_type);
885 if (err)
886 goto fs_err;
887
888 return 0;
889
890 fs_err:
891 erofs_exit_sysfs();
892 sysfs_err:
893 z_erofs_exit_subsystem();
894 zip_err:
895 erofs_exit_shrinker();
896 shrinker_err:
897 kmem_cache_destroy(erofs_inode_cachep);
898 return err;
899 }
900
erofs_module_exit(void)901 static void __exit erofs_module_exit(void)
902 {
903 unregister_filesystem(&erofs_fs_type);
904
905 /* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
906 rcu_barrier();
907
908 erofs_exit_sysfs();
909 z_erofs_exit_subsystem();
910 erofs_exit_shrinker();
911 kmem_cache_destroy(erofs_inode_cachep);
912 }
913
erofs_statfs(struct dentry * dentry,struct kstatfs * buf)914 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
915 {
916 struct super_block *sb = dentry->d_sb;
917 struct erofs_sb_info *sbi = EROFS_SB(sb);
918
919 buf->f_type = sb->s_magic;
920 buf->f_bsize = sb->s_blocksize;
921 buf->f_blocks = sbi->total_blocks;
922 buf->f_bfree = buf->f_bavail = 0;
923 buf->f_files = ULLONG_MAX;
924 buf->f_ffree = ULLONG_MAX - sbi->inos;
925 buf->f_namelen = EROFS_NAME_LEN;
926
927 if (uuid_is_null(&sb->s_uuid))
928 buf->f_fsid = u64_to_fsid(!sb->s_bdev ? 0 :
929 huge_encode_dev(sb->s_bdev->bd_dev));
930 else
931 buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
932 return 0;
933 }
934
erofs_show_options(struct seq_file * seq,struct dentry * root)935 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
936 {
937 struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
938 struct erofs_mount_opts *opt = &sbi->opt;
939
940 if (IS_ENABLED(CONFIG_EROFS_FS_XATTR))
941 seq_puts(seq, test_opt(opt, XATTR_USER) ?
942 ",user_xattr" : ",nouser_xattr");
943 if (IS_ENABLED(CONFIG_EROFS_FS_POSIX_ACL))
944 seq_puts(seq, test_opt(opt, POSIX_ACL) ? ",acl" : ",noacl");
945 if (IS_ENABLED(CONFIG_EROFS_FS_ZIP))
946 seq_printf(seq, ",cache_strategy=%s",
947 erofs_param_cache_strategy[opt->cache_strategy].name);
948 if (test_opt(opt, DAX_ALWAYS))
949 seq_puts(seq, ",dax=always");
950 if (test_opt(opt, DAX_NEVER))
951 seq_puts(seq, ",dax=never");
952 #ifdef CONFIG_EROFS_FS_ONDEMAND
953 if (sbi->fsid)
954 seq_printf(seq, ",fsid=%s", sbi->fsid);
955 if (sbi->domain_id)
956 seq_printf(seq, ",domain_id=%s", sbi->domain_id);
957 #endif
958 return 0;
959 }
960
961 const struct super_operations erofs_sops = {
962 .put_super = erofs_put_super,
963 .alloc_inode = erofs_alloc_inode,
964 .free_inode = erofs_free_inode,
965 .statfs = erofs_statfs,
966 .show_options = erofs_show_options,
967 };
968
969 module_init(erofs_module_init);
970 module_exit(erofs_module_exit);
971
972 MODULE_DESCRIPTION("Enhanced ROM File System");
973 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
974 MODULE_LICENSE("GPL");
975