xref: /linux/fs/erofs/super.c (revision 76a9701325d39d8602695b19c49a9d0828c897ca)
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 		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, false);
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 	if (erofs_sb_has_metabox(sbi)) {
323 		if (sbi->sb_size <= offsetof(struct erofs_super_block,
324 					     metabox_nid))
325 			return -EFSCORRUPTED;
326 		sbi->metabox_nid = le64_to_cpu(dsb->metabox_nid);
327 		if (sbi->metabox_nid & BIT_ULL(EROFS_DIRENT_NID_METABOX_BIT))
328 			return -EFSCORRUPTED;	/* self-loop detection */
329 	}
330 	sbi->inos = le64_to_cpu(dsb->inos);
331 
332 	sbi->epoch = (s64)le64_to_cpu(dsb->epoch);
333 	sbi->fixed_nsec = le32_to_cpu(dsb->fixed_nsec);
334 	super_set_uuid(sb, (void *)dsb->uuid, sizeof(dsb->uuid));
335 
336 	/* parse on-disk compression configurations */
337 	ret = z_erofs_parse_cfgs(sb, dsb);
338 	if (ret < 0)
339 		goto out;
340 
341 	/* handle multiple devices */
342 	ret = erofs_scan_devices(sb, dsb);
343 
344 	if (erofs_sb_has_48bit(sbi))
345 		erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
346 	if (erofs_sb_has_metabox(sbi))
347 		erofs_info(sb, "EXPERIMENTAL metadata compression support in use. Use at your own risk!");
348 	if (erofs_is_fscache_mode(sb))
349 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
350 out:
351 	erofs_put_metabuf(&buf);
352 	return ret;
353 }
354 
erofs_default_options(struct erofs_sb_info * sbi)355 static void erofs_default_options(struct erofs_sb_info *sbi)
356 {
357 #ifdef CONFIG_EROFS_FS_ZIP
358 	sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
359 	sbi->opt.max_sync_decompress_pages = 3;
360 	sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
361 #endif
362 #ifdef CONFIG_EROFS_FS_XATTR
363 	set_opt(&sbi->opt, XATTR_USER);
364 #endif
365 #ifdef CONFIG_EROFS_FS_POSIX_ACL
366 	set_opt(&sbi->opt, POSIX_ACL);
367 #endif
368 }
369 
370 enum {
371 	Opt_user_xattr, Opt_acl, Opt_cache_strategy, Opt_dax, Opt_dax_enum,
372 	Opt_device, Opt_fsid, Opt_domain_id, Opt_directio, Opt_fsoffset,
373 };
374 
375 static const struct constant_table erofs_param_cache_strategy[] = {
376 	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
377 	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
378 	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
379 	{}
380 };
381 
382 static const struct constant_table erofs_dax_param_enums[] = {
383 	{"always",	EROFS_MOUNT_DAX_ALWAYS},
384 	{"never",	EROFS_MOUNT_DAX_NEVER},
385 	{}
386 };
387 
388 static const struct fs_parameter_spec erofs_fs_parameters[] = {
389 	fsparam_flag_no("user_xattr",	Opt_user_xattr),
390 	fsparam_flag_no("acl",		Opt_acl),
391 	fsparam_enum("cache_strategy",	Opt_cache_strategy,
392 		     erofs_param_cache_strategy),
393 	fsparam_flag("dax",             Opt_dax),
394 	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
395 	fsparam_string("device",	Opt_device),
396 	fsparam_string("fsid",		Opt_fsid),
397 	fsparam_string("domain_id",	Opt_domain_id),
398 	fsparam_flag_no("directio",	Opt_directio),
399 	fsparam_u64("fsoffset",		Opt_fsoffset),
400 	{}
401 };
402 
erofs_fc_set_dax_mode(struct fs_context * fc,unsigned int mode)403 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
404 {
405 #ifdef CONFIG_FS_DAX
406 	struct erofs_sb_info *sbi = fc->s_fs_info;
407 
408 	switch (mode) {
409 	case EROFS_MOUNT_DAX_ALWAYS:
410 		set_opt(&sbi->opt, DAX_ALWAYS);
411 		clear_opt(&sbi->opt, DAX_NEVER);
412 		return true;
413 	case EROFS_MOUNT_DAX_NEVER:
414 		set_opt(&sbi->opt, DAX_NEVER);
415 		clear_opt(&sbi->opt, DAX_ALWAYS);
416 		return true;
417 	default:
418 		DBG_BUGON(1);
419 		return false;
420 	}
421 #else
422 	errorfc(fc, "dax options not supported");
423 	return false;
424 #endif
425 }
426 
erofs_fc_parse_param(struct fs_context * fc,struct fs_parameter * param)427 static int erofs_fc_parse_param(struct fs_context *fc,
428 				struct fs_parameter *param)
429 {
430 	struct erofs_sb_info *sbi = fc->s_fs_info;
431 	struct fs_parse_result result;
432 	struct erofs_device_info *dif;
433 	int opt, ret;
434 
435 	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
436 	if (opt < 0)
437 		return opt;
438 
439 	switch (opt) {
440 	case Opt_user_xattr:
441 #ifdef CONFIG_EROFS_FS_XATTR
442 		if (result.boolean)
443 			set_opt(&sbi->opt, XATTR_USER);
444 		else
445 			clear_opt(&sbi->opt, XATTR_USER);
446 #else
447 		errorfc(fc, "{,no}user_xattr options not supported");
448 #endif
449 		break;
450 	case Opt_acl:
451 #ifdef CONFIG_EROFS_FS_POSIX_ACL
452 		if (result.boolean)
453 			set_opt(&sbi->opt, POSIX_ACL);
454 		else
455 			clear_opt(&sbi->opt, POSIX_ACL);
456 #else
457 		errorfc(fc, "{,no}acl options not supported");
458 #endif
459 		break;
460 	case Opt_cache_strategy:
461 #ifdef CONFIG_EROFS_FS_ZIP
462 		sbi->opt.cache_strategy = result.uint_32;
463 #else
464 		errorfc(fc, "compression not supported, cache_strategy ignored");
465 #endif
466 		break;
467 	case Opt_dax:
468 		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
469 			return -EINVAL;
470 		break;
471 	case Opt_dax_enum:
472 		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
473 			return -EINVAL;
474 		break;
475 	case Opt_device:
476 		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
477 		if (!dif)
478 			return -ENOMEM;
479 		dif->path = kstrdup(param->string, GFP_KERNEL);
480 		if (!dif->path) {
481 			kfree(dif);
482 			return -ENOMEM;
483 		}
484 		down_write(&sbi->devs->rwsem);
485 		ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
486 		up_write(&sbi->devs->rwsem);
487 		if (ret < 0) {
488 			kfree(dif->path);
489 			kfree(dif);
490 			return ret;
491 		}
492 		++sbi->devs->extra_devices;
493 		break;
494 #ifdef CONFIG_EROFS_FS_ONDEMAND
495 	case Opt_fsid:
496 		kfree(sbi->fsid);
497 		sbi->fsid = kstrdup(param->string, GFP_KERNEL);
498 		if (!sbi->fsid)
499 			return -ENOMEM;
500 		break;
501 	case Opt_domain_id:
502 		kfree(sbi->domain_id);
503 		sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
504 		if (!sbi->domain_id)
505 			return -ENOMEM;
506 		break;
507 #else
508 	case Opt_fsid:
509 	case Opt_domain_id:
510 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
511 		break;
512 #endif
513 	case Opt_directio:
514 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
515 		if (result.boolean)
516 			set_opt(&sbi->opt, DIRECT_IO);
517 		else
518 			clear_opt(&sbi->opt, DIRECT_IO);
519 #else
520 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
521 #endif
522 		break;
523 	case Opt_fsoffset:
524 		sbi->dif0.fsoff = result.uint_64;
525 		break;
526 	}
527 	return 0;
528 }
529 
erofs_encode_fh(struct inode * inode,u32 * fh,int * max_len,struct inode * parent)530 static int erofs_encode_fh(struct inode *inode, u32 *fh, int *max_len,
531 			   struct inode *parent)
532 {
533 	erofs_nid_t nid = EROFS_I(inode)->nid;
534 	int len = parent ? 6 : 3;
535 
536 	if (*max_len < len) {
537 		*max_len = len;
538 		return FILEID_INVALID;
539 	}
540 
541 	fh[0] = (u32)(nid >> 32);
542 	fh[1] = (u32)(nid & 0xffffffff);
543 	fh[2] = inode->i_generation;
544 
545 	if (parent) {
546 		nid = EROFS_I(parent)->nid;
547 
548 		fh[3] = (u32)(nid >> 32);
549 		fh[4] = (u32)(nid & 0xffffffff);
550 		fh[5] = parent->i_generation;
551 	}
552 
553 	*max_len = len;
554 	return parent ? FILEID_INO64_GEN_PARENT : FILEID_INO64_GEN;
555 }
556 
erofs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)557 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
558 		struct fid *fid, int fh_len, int fh_type)
559 {
560 	if ((fh_type != FILEID_INO64_GEN &&
561 	     fh_type != FILEID_INO64_GEN_PARENT) || fh_len < 3)
562 		return NULL;
563 
564 	return d_obtain_alias(erofs_iget(sb,
565 		((u64)fid->raw[0] << 32) | fid->raw[1]));
566 }
567 
erofs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)568 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
569 		struct fid *fid, int fh_len, int fh_type)
570 {
571 	if (fh_type != FILEID_INO64_GEN_PARENT || fh_len < 6)
572 		return NULL;
573 
574 	return d_obtain_alias(erofs_iget(sb,
575 		((u64)fid->raw[3] << 32) | fid->raw[4]));
576 }
577 
erofs_get_parent(struct dentry * child)578 static struct dentry *erofs_get_parent(struct dentry *child)
579 {
580 	erofs_nid_t nid;
581 	unsigned int d_type;
582 	int err;
583 
584 	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
585 	if (err)
586 		return ERR_PTR(err);
587 	return d_obtain_alias(erofs_iget(child->d_sb, nid));
588 }
589 
590 static const struct export_operations erofs_export_ops = {
591 	.encode_fh = erofs_encode_fh,
592 	.fh_to_dentry = erofs_fh_to_dentry,
593 	.fh_to_parent = erofs_fh_to_parent,
594 	.get_parent = erofs_get_parent,
595 };
596 
erofs_set_sysfs_name(struct super_block * sb)597 static void erofs_set_sysfs_name(struct super_block *sb)
598 {
599 	struct erofs_sb_info *sbi = EROFS_SB(sb);
600 
601 	if (sbi->domain_id)
602 		super_set_sysfs_name_generic(sb, "%s,%s", sbi->domain_id,
603 					     sbi->fsid);
604 	else if (sbi->fsid)
605 		super_set_sysfs_name_generic(sb, "%s", sbi->fsid);
606 	else if (erofs_is_fileio_mode(sbi))
607 		super_set_sysfs_name_generic(sb, "%s",
608 					     bdi_dev_name(sb->s_bdi));
609 	else
610 		super_set_sysfs_name_id(sb);
611 }
612 
erofs_fc_fill_super(struct super_block * sb,struct fs_context * fc)613 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
614 {
615 	struct inode *inode;
616 	struct erofs_sb_info *sbi = EROFS_SB(sb);
617 	int err;
618 
619 	sb->s_magic = EROFS_SUPER_MAGIC;
620 	sb->s_flags |= SB_RDONLY | SB_NOATIME;
621 	sb->s_maxbytes = MAX_LFS_FILESIZE;
622 	sb->s_op = &erofs_sops;
623 
624 	sbi->blkszbits = PAGE_SHIFT;
625 	if (!sb->s_bdev) {
626 		sb->s_blocksize = PAGE_SIZE;
627 		sb->s_blocksize_bits = PAGE_SHIFT;
628 
629 		if (erofs_is_fscache_mode(sb)) {
630 			err = erofs_fscache_register_fs(sb);
631 			if (err)
632 				return err;
633 		}
634 		err = super_setup_bdi(sb);
635 		if (err)
636 			return err;
637 	} else {
638 		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
639 			errorfc(fc, "failed to set initial blksize");
640 			return -EINVAL;
641 		}
642 
643 		sbi->dif0.dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
644 				&sbi->dif0.dax_part_off, NULL, NULL);
645 	}
646 
647 	err = erofs_read_superblock(sb);
648 	if (err)
649 		return err;
650 
651 	if (sb->s_blocksize_bits != sbi->blkszbits) {
652 		if (erofs_is_fscache_mode(sb)) {
653 			errorfc(fc, "unsupported blksize for fscache mode");
654 			return -EINVAL;
655 		}
656 
657 		if (erofs_is_fileio_mode(sbi)) {
658 			sb->s_blocksize = 1 << sbi->blkszbits;
659 			sb->s_blocksize_bits = sbi->blkszbits;
660 		} else if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
661 			errorfc(fc, "failed to set erofs blksize");
662 			return -EINVAL;
663 		}
664 	}
665 
666 	if (sbi->dif0.fsoff) {
667 		if (sbi->dif0.fsoff & (sb->s_blocksize - 1))
668 			return invalfc(fc, "fsoffset %llu is not aligned to block size %lu",
669 				       sbi->dif0.fsoff, sb->s_blocksize);
670 		if (erofs_is_fscache_mode(sb))
671 			return invalfc(fc, "cannot use fsoffset in fscache mode");
672 	}
673 
674 	if (test_opt(&sbi->opt, DAX_ALWAYS)) {
675 		if (!sbi->dif0.dax_dev) {
676 			errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
677 			clear_opt(&sbi->opt, DAX_ALWAYS);
678 		} else if (sbi->blkszbits != PAGE_SHIFT) {
679 			errorfc(fc, "unsupported blocksize for DAX");
680 			clear_opt(&sbi->opt, DAX_ALWAYS);
681 		}
682 	}
683 
684 	sb->s_time_gran = 1;
685 	sb->s_xattr = erofs_xattr_handlers;
686 	sb->s_export_op = &erofs_export_ops;
687 
688 	if (test_opt(&sbi->opt, POSIX_ACL))
689 		sb->s_flags |= SB_POSIXACL;
690 	else
691 		sb->s_flags &= ~SB_POSIXACL;
692 
693 	err = z_erofs_init_super(sb);
694 	if (err)
695 		return err;
696 
697 	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
698 		inode = erofs_iget(sb, sbi->packed_nid);
699 		if (IS_ERR(inode))
700 			return PTR_ERR(inode);
701 		sbi->packed_inode = inode;
702 	}
703 	if (erofs_sb_has_metabox(sbi)) {
704 		inode = erofs_iget(sb, sbi->metabox_nid);
705 		if (IS_ERR(inode))
706 			return PTR_ERR(inode);
707 		sbi->metabox_inode = inode;
708 	}
709 
710 	inode = erofs_iget(sb, sbi->root_nid);
711 	if (IS_ERR(inode))
712 		return PTR_ERR(inode);
713 
714 	if (!S_ISDIR(inode->i_mode)) {
715 		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
716 			  sbi->root_nid, inode->i_mode);
717 		iput(inode);
718 		return -EINVAL;
719 	}
720 	sb->s_root = d_make_root(inode);
721 	if (!sb->s_root)
722 		return -ENOMEM;
723 
724 	erofs_shrinker_register(sb);
725 	err = erofs_xattr_prefixes_init(sb);
726 	if (err)
727 		return err;
728 
729 	erofs_set_sysfs_name(sb);
730 	err = erofs_register_sysfs(sb);
731 	if (err)
732 		return err;
733 
734 	sbi->dir_ra_bytes = EROFS_DIR_RA_BYTES;
735 	erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
736 	return 0;
737 }
738 
erofs_fc_get_tree(struct fs_context * fc)739 static int erofs_fc_get_tree(struct fs_context *fc)
740 {
741 	struct erofs_sb_info *sbi = fc->s_fs_info;
742 	int ret;
743 
744 	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
745 		return get_tree_nodev(fc, erofs_fc_fill_super);
746 
747 	ret = get_tree_bdev_flags(fc, erofs_fc_fill_super,
748 		IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) ?
749 			GET_TREE_BDEV_QUIET_LOOKUP : 0);
750 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
751 	if (ret == -ENOTBLK) {
752 		struct file *file;
753 
754 		if (!fc->source)
755 			return invalf(fc, "No source specified");
756 		file = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
757 		if (IS_ERR(file))
758 			return PTR_ERR(file);
759 		sbi->dif0.file = file;
760 
761 		if (S_ISREG(file_inode(sbi->dif0.file)->i_mode) &&
762 		    sbi->dif0.file->f_mapping->a_ops->read_folio)
763 			return get_tree_nodev(fc, erofs_fc_fill_super);
764 	}
765 #endif
766 	return ret;
767 }
768 
erofs_fc_reconfigure(struct fs_context * fc)769 static int erofs_fc_reconfigure(struct fs_context *fc)
770 {
771 	struct super_block *sb = fc->root->d_sb;
772 	struct erofs_sb_info *sbi = EROFS_SB(sb);
773 	struct erofs_sb_info *new_sbi = fc->s_fs_info;
774 
775 	DBG_BUGON(!sb_rdonly(sb));
776 
777 	if (new_sbi->fsid || new_sbi->domain_id)
778 		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
779 
780 	if (test_opt(&new_sbi->opt, POSIX_ACL))
781 		fc->sb_flags |= SB_POSIXACL;
782 	else
783 		fc->sb_flags &= ~SB_POSIXACL;
784 
785 	sbi->opt = new_sbi->opt;
786 
787 	fc->sb_flags |= SB_RDONLY;
788 	return 0;
789 }
790 
erofs_release_device_info(int id,void * ptr,void * data)791 static int erofs_release_device_info(int id, void *ptr, void *data)
792 {
793 	struct erofs_device_info *dif = ptr;
794 
795 	fs_put_dax(dif->dax_dev, NULL);
796 	if (dif->file)
797 		fput(dif->file);
798 	erofs_fscache_unregister_cookie(dif->fscache);
799 	dif->fscache = NULL;
800 	kfree(dif->path);
801 	kfree(dif);
802 	return 0;
803 }
804 
erofs_free_dev_context(struct erofs_dev_context * devs)805 static void erofs_free_dev_context(struct erofs_dev_context *devs)
806 {
807 	if (!devs)
808 		return;
809 	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
810 	idr_destroy(&devs->tree);
811 	kfree(devs);
812 }
813 
erofs_sb_free(struct erofs_sb_info * sbi)814 static void erofs_sb_free(struct erofs_sb_info *sbi)
815 {
816 	erofs_free_dev_context(sbi->devs);
817 	kfree(sbi->fsid);
818 	kfree(sbi->domain_id);
819 	if (sbi->dif0.file)
820 		fput(sbi->dif0.file);
821 	kfree(sbi);
822 }
823 
erofs_fc_free(struct fs_context * fc)824 static void erofs_fc_free(struct fs_context *fc)
825 {
826 	struct erofs_sb_info *sbi = fc->s_fs_info;
827 
828 	if (sbi) /* free here if an error occurs before transferring to sb */
829 		erofs_sb_free(sbi);
830 }
831 
832 static const struct fs_context_operations erofs_context_ops = {
833 	.parse_param	= erofs_fc_parse_param,
834 	.get_tree       = erofs_fc_get_tree,
835 	.reconfigure    = erofs_fc_reconfigure,
836 	.free		= erofs_fc_free,
837 };
838 
erofs_init_fs_context(struct fs_context * fc)839 static int erofs_init_fs_context(struct fs_context *fc)
840 {
841 	struct erofs_sb_info *sbi;
842 
843 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
844 	if (!sbi)
845 		return -ENOMEM;
846 
847 	sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
848 	if (!sbi->devs) {
849 		kfree(sbi);
850 		return -ENOMEM;
851 	}
852 	fc->s_fs_info = sbi;
853 
854 	idr_init(&sbi->devs->tree);
855 	init_rwsem(&sbi->devs->rwsem);
856 	erofs_default_options(sbi);
857 	fc->ops = &erofs_context_ops;
858 	return 0;
859 }
860 
erofs_drop_internal_inodes(struct erofs_sb_info * sbi)861 static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
862 {
863 	iput(sbi->packed_inode);
864 	sbi->packed_inode = NULL;
865 	iput(sbi->metabox_inode);
866 	sbi->metabox_inode = NULL;
867 #ifdef CONFIG_EROFS_FS_ZIP
868 	iput(sbi->managed_cache);
869 	sbi->managed_cache = NULL;
870 #endif
871 }
872 
erofs_kill_sb(struct super_block * sb)873 static void erofs_kill_sb(struct super_block *sb)
874 {
875 	struct erofs_sb_info *sbi = EROFS_SB(sb);
876 
877 	if ((IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid) ||
878 	    sbi->dif0.file)
879 		kill_anon_super(sb);
880 	else
881 		kill_block_super(sb);
882 	erofs_drop_internal_inodes(sbi);
883 	fs_put_dax(sbi->dif0.dax_dev, NULL);
884 	erofs_fscache_unregister_fs(sb);
885 	erofs_sb_free(sbi);
886 	sb->s_fs_info = NULL;
887 }
888 
erofs_put_super(struct super_block * sb)889 static void erofs_put_super(struct super_block *sb)
890 {
891 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
892 
893 	erofs_unregister_sysfs(sb);
894 	erofs_shrinker_unregister(sb);
895 	erofs_xattr_prefixes_cleanup(sb);
896 	erofs_drop_internal_inodes(sbi);
897 	erofs_free_dev_context(sbi->devs);
898 	sbi->devs = NULL;
899 	erofs_fscache_unregister_fs(sb);
900 }
901 
902 static struct file_system_type erofs_fs_type = {
903 	.owner          = THIS_MODULE,
904 	.name           = "erofs",
905 	.init_fs_context = erofs_init_fs_context,
906 	.kill_sb        = erofs_kill_sb,
907 	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
908 };
909 MODULE_ALIAS_FS("erofs");
910 
erofs_module_init(void)911 static int __init erofs_module_init(void)
912 {
913 	int err;
914 
915 	erofs_check_ondisk_layout_definitions();
916 
917 	erofs_inode_cachep = kmem_cache_create("erofs_inode",
918 			sizeof(struct erofs_inode), 0,
919 			SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
920 			erofs_inode_init_once);
921 	if (!erofs_inode_cachep)
922 		return -ENOMEM;
923 
924 	err = erofs_init_shrinker();
925 	if (err)
926 		goto shrinker_err;
927 
928 	err = z_erofs_init_subsystem();
929 	if (err)
930 		goto zip_err;
931 
932 	err = erofs_init_sysfs();
933 	if (err)
934 		goto sysfs_err;
935 
936 	err = register_filesystem(&erofs_fs_type);
937 	if (err)
938 		goto fs_err;
939 
940 	return 0;
941 
942 fs_err:
943 	erofs_exit_sysfs();
944 sysfs_err:
945 	z_erofs_exit_subsystem();
946 zip_err:
947 	erofs_exit_shrinker();
948 shrinker_err:
949 	kmem_cache_destroy(erofs_inode_cachep);
950 	return err;
951 }
952 
erofs_module_exit(void)953 static void __exit erofs_module_exit(void)
954 {
955 	unregister_filesystem(&erofs_fs_type);
956 
957 	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
958 	rcu_barrier();
959 
960 	erofs_exit_sysfs();
961 	z_erofs_exit_subsystem();
962 	erofs_exit_shrinker();
963 	kmem_cache_destroy(erofs_inode_cachep);
964 }
965 
erofs_statfs(struct dentry * dentry,struct kstatfs * buf)966 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
967 {
968 	struct super_block *sb = dentry->d_sb;
969 	struct erofs_sb_info *sbi = EROFS_SB(sb);
970 
971 	buf->f_type = sb->s_magic;
972 	buf->f_bsize = sb->s_blocksize;
973 	buf->f_blocks = sbi->total_blocks;
974 	buf->f_bfree = buf->f_bavail = 0;
975 	buf->f_files = ULLONG_MAX;
976 	buf->f_ffree = ULLONG_MAX - sbi->inos;
977 	buf->f_namelen = EROFS_NAME_LEN;
978 
979 	if (uuid_is_null(&sb->s_uuid))
980 		buf->f_fsid = u64_to_fsid(!sb->s_bdev ? 0 :
981 				huge_encode_dev(sb->s_bdev->bd_dev));
982 	else
983 		buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
984 	return 0;
985 }
986 
erofs_show_options(struct seq_file * seq,struct dentry * root)987 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
988 {
989 	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
990 	struct erofs_mount_opts *opt = &sbi->opt;
991 
992 	if (IS_ENABLED(CONFIG_EROFS_FS_XATTR))
993 		seq_puts(seq, test_opt(opt, XATTR_USER) ?
994 				",user_xattr" : ",nouser_xattr");
995 	if (IS_ENABLED(CONFIG_EROFS_FS_POSIX_ACL))
996 		seq_puts(seq, test_opt(opt, POSIX_ACL) ? ",acl" : ",noacl");
997 	if (IS_ENABLED(CONFIG_EROFS_FS_ZIP))
998 		seq_printf(seq, ",cache_strategy=%s",
999 			  erofs_param_cache_strategy[opt->cache_strategy].name);
1000 	if (test_opt(opt, DAX_ALWAYS))
1001 		seq_puts(seq, ",dax=always");
1002 	if (test_opt(opt, DAX_NEVER))
1003 		seq_puts(seq, ",dax=never");
1004 	if (erofs_is_fileio_mode(sbi) && test_opt(opt, DIRECT_IO))
1005 		seq_puts(seq, ",directio");
1006 #ifdef CONFIG_EROFS_FS_ONDEMAND
1007 	if (sbi->fsid)
1008 		seq_printf(seq, ",fsid=%s", sbi->fsid);
1009 	if (sbi->domain_id)
1010 		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
1011 #endif
1012 	if (sbi->dif0.fsoff)
1013 		seq_printf(seq, ",fsoffset=%llu", sbi->dif0.fsoff);
1014 	return 0;
1015 }
1016 
1017 const struct super_operations erofs_sops = {
1018 	.put_super = erofs_put_super,
1019 	.alloc_inode = erofs_alloc_inode,
1020 	.free_inode = erofs_free_inode,
1021 	.statfs = erofs_statfs,
1022 	.show_options = erofs_show_options,
1023 };
1024 
1025 module_init(erofs_module_init);
1026 module_exit(erofs_module_exit);
1027 
1028 MODULE_DESCRIPTION("Enhanced ROM File System");
1029 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
1030 MODULE_LICENSE("GPL");
1031