xref: /linux/fs/erofs/super.c (revision dfc0f6373094dd88e1eaf76c44f2ff01b65db851)
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 			if (!dif->dax_dev && test_opt(&sbi->opt, DAX_ALWAYS)) {
178 				erofs_info(sb, "DAX unsupported by %s. Turning off DAX.",
179 					   dif->path);
180 				clear_opt(&sbi->opt, DAX_ALWAYS);
181 			}
182 		} else if (!S_ISREG(file_inode(file)->i_mode)) {
183 			fput(file);
184 			return -EINVAL;
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 	if (!ondisk_extradevs) {
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 		return 0;
224 	}
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 	/* parse on-disk compression configurations */
347 	ret = z_erofs_parse_cfgs(sb, dsb);
348 	if (ret < 0)
349 		goto out;
350 
351 	ret = erofs_scan_devices(sb, dsb);
352 
353 	if (erofs_sb_has_48bit(sbi))
354 		erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
355 	if (erofs_sb_has_metabox(sbi))
356 		erofs_info(sb, "EXPERIMENTAL metadata compression support in use. Use at your own risk!");
357 	if (erofs_is_fscache_mode(sb))
358 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
359 out:
360 	erofs_put_metabuf(&buf);
361 	return ret;
362 }
363 
erofs_default_options(struct erofs_sb_info * sbi)364 static void erofs_default_options(struct erofs_sb_info *sbi)
365 {
366 #ifdef CONFIG_EROFS_FS_ZIP
367 	sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
368 	sbi->opt.max_sync_decompress_pages = 3;
369 	sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
370 #endif
371 #ifdef CONFIG_EROFS_FS_XATTR
372 	set_opt(&sbi->opt, XATTR_USER);
373 #endif
374 #ifdef CONFIG_EROFS_FS_POSIX_ACL
375 	set_opt(&sbi->opt, POSIX_ACL);
376 #endif
377 }
378 
379 enum {
380 	Opt_user_xattr, Opt_acl, Opt_cache_strategy, Opt_dax, Opt_dax_enum,
381 	Opt_device, Opt_fsid, Opt_domain_id, Opt_directio, Opt_fsoffset,
382 };
383 
384 static const struct constant_table erofs_param_cache_strategy[] = {
385 	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
386 	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
387 	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
388 	{}
389 };
390 
391 static const struct constant_table erofs_dax_param_enums[] = {
392 	{"always",	EROFS_MOUNT_DAX_ALWAYS},
393 	{"never",	EROFS_MOUNT_DAX_NEVER},
394 	{}
395 };
396 
397 static const struct fs_parameter_spec erofs_fs_parameters[] = {
398 	fsparam_flag_no("user_xattr",	Opt_user_xattr),
399 	fsparam_flag_no("acl",		Opt_acl),
400 	fsparam_enum("cache_strategy",	Opt_cache_strategy,
401 		     erofs_param_cache_strategy),
402 	fsparam_flag("dax",             Opt_dax),
403 	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
404 	fsparam_string("device",	Opt_device),
405 	fsparam_string("fsid",		Opt_fsid),
406 	fsparam_string("domain_id",	Opt_domain_id),
407 	fsparam_flag_no("directio",	Opt_directio),
408 	fsparam_u64("fsoffset",		Opt_fsoffset),
409 	{}
410 };
411 
erofs_fc_set_dax_mode(struct fs_context * fc,unsigned int mode)412 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
413 {
414 #ifdef CONFIG_FS_DAX
415 	struct erofs_sb_info *sbi = fc->s_fs_info;
416 
417 	switch (mode) {
418 	case EROFS_MOUNT_DAX_ALWAYS:
419 		set_opt(&sbi->opt, DAX_ALWAYS);
420 		clear_opt(&sbi->opt, DAX_NEVER);
421 		return true;
422 	case EROFS_MOUNT_DAX_NEVER:
423 		set_opt(&sbi->opt, DAX_NEVER);
424 		clear_opt(&sbi->opt, DAX_ALWAYS);
425 		return true;
426 	default:
427 		DBG_BUGON(1);
428 		return false;
429 	}
430 #else
431 	errorfc(fc, "dax options not supported");
432 	return false;
433 #endif
434 }
435 
erofs_fc_parse_param(struct fs_context * fc,struct fs_parameter * param)436 static int erofs_fc_parse_param(struct fs_context *fc,
437 				struct fs_parameter *param)
438 {
439 	struct erofs_sb_info *sbi = fc->s_fs_info;
440 	struct fs_parse_result result;
441 	struct erofs_device_info *dif;
442 	int opt, ret;
443 
444 	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
445 	if (opt < 0)
446 		return opt;
447 
448 	switch (opt) {
449 	case Opt_user_xattr:
450 #ifdef CONFIG_EROFS_FS_XATTR
451 		if (result.boolean)
452 			set_opt(&sbi->opt, XATTR_USER);
453 		else
454 			clear_opt(&sbi->opt, XATTR_USER);
455 #else
456 		errorfc(fc, "{,no}user_xattr options not supported");
457 #endif
458 		break;
459 	case Opt_acl:
460 #ifdef CONFIG_EROFS_FS_POSIX_ACL
461 		if (result.boolean)
462 			set_opt(&sbi->opt, POSIX_ACL);
463 		else
464 			clear_opt(&sbi->opt, POSIX_ACL);
465 #else
466 		errorfc(fc, "{,no}acl options not supported");
467 #endif
468 		break;
469 	case Opt_cache_strategy:
470 #ifdef CONFIG_EROFS_FS_ZIP
471 		sbi->opt.cache_strategy = result.uint_32;
472 #else
473 		errorfc(fc, "compression not supported, cache_strategy ignored");
474 #endif
475 		break;
476 	case Opt_dax:
477 		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
478 			return -EINVAL;
479 		break;
480 	case Opt_dax_enum:
481 		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
482 			return -EINVAL;
483 		break;
484 	case Opt_device:
485 		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
486 		if (!dif)
487 			return -ENOMEM;
488 		dif->path = kstrdup(param->string, GFP_KERNEL);
489 		if (!dif->path) {
490 			kfree(dif);
491 			return -ENOMEM;
492 		}
493 		down_write(&sbi->devs->rwsem);
494 		ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
495 		up_write(&sbi->devs->rwsem);
496 		if (ret < 0) {
497 			kfree(dif->path);
498 			kfree(dif);
499 			return ret;
500 		}
501 		++sbi->devs->extra_devices;
502 		break;
503 #ifdef CONFIG_EROFS_FS_ONDEMAND
504 	case Opt_fsid:
505 		kfree(sbi->fsid);
506 		sbi->fsid = kstrdup(param->string, GFP_KERNEL);
507 		if (!sbi->fsid)
508 			return -ENOMEM;
509 		break;
510 	case Opt_domain_id:
511 		kfree(sbi->domain_id);
512 		sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
513 		if (!sbi->domain_id)
514 			return -ENOMEM;
515 		break;
516 #else
517 	case Opt_fsid:
518 	case Opt_domain_id:
519 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
520 		break;
521 #endif
522 	case Opt_directio:
523 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
524 		if (result.boolean)
525 			set_opt(&sbi->opt, DIRECT_IO);
526 		else
527 			clear_opt(&sbi->opt, DIRECT_IO);
528 #else
529 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
530 #endif
531 		break;
532 	case Opt_fsoffset:
533 		sbi->dif0.fsoff = result.uint_64;
534 		break;
535 	}
536 	return 0;
537 }
538 
erofs_encode_fh(struct inode * inode,u32 * fh,int * max_len,struct inode * parent)539 static int erofs_encode_fh(struct inode *inode, u32 *fh, int *max_len,
540 			   struct inode *parent)
541 {
542 	erofs_nid_t nid = EROFS_I(inode)->nid;
543 	int len = parent ? 6 : 3;
544 
545 	if (*max_len < len) {
546 		*max_len = len;
547 		return FILEID_INVALID;
548 	}
549 
550 	fh[0] = (u32)(nid >> 32);
551 	fh[1] = (u32)(nid & 0xffffffff);
552 	fh[2] = inode->i_generation;
553 
554 	if (parent) {
555 		nid = EROFS_I(parent)->nid;
556 
557 		fh[3] = (u32)(nid >> 32);
558 		fh[4] = (u32)(nid & 0xffffffff);
559 		fh[5] = parent->i_generation;
560 	}
561 
562 	*max_len = len;
563 	return parent ? FILEID_INO64_GEN_PARENT : FILEID_INO64_GEN;
564 }
565 
erofs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)566 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
567 		struct fid *fid, int fh_len, int fh_type)
568 {
569 	if ((fh_type != FILEID_INO64_GEN &&
570 	     fh_type != FILEID_INO64_GEN_PARENT) || fh_len < 3)
571 		return NULL;
572 
573 	return d_obtain_alias(erofs_iget(sb,
574 		((u64)fid->raw[0] << 32) | fid->raw[1]));
575 }
576 
erofs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)577 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
578 		struct fid *fid, int fh_len, int fh_type)
579 {
580 	if (fh_type != FILEID_INO64_GEN_PARENT || fh_len < 6)
581 		return NULL;
582 
583 	return d_obtain_alias(erofs_iget(sb,
584 		((u64)fid->raw[3] << 32) | fid->raw[4]));
585 }
586 
erofs_get_parent(struct dentry * child)587 static struct dentry *erofs_get_parent(struct dentry *child)
588 {
589 	erofs_nid_t nid;
590 	unsigned int d_type;
591 	int err;
592 
593 	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
594 	if (err)
595 		return ERR_PTR(err);
596 	return d_obtain_alias(erofs_iget(child->d_sb, nid));
597 }
598 
599 static const struct export_operations erofs_export_ops = {
600 	.encode_fh = erofs_encode_fh,
601 	.fh_to_dentry = erofs_fh_to_dentry,
602 	.fh_to_parent = erofs_fh_to_parent,
603 	.get_parent = erofs_get_parent,
604 };
605 
erofs_set_sysfs_name(struct super_block * sb)606 static void erofs_set_sysfs_name(struct super_block *sb)
607 {
608 	struct erofs_sb_info *sbi = EROFS_SB(sb);
609 
610 	if (sbi->domain_id)
611 		super_set_sysfs_name_generic(sb, "%s,%s", sbi->domain_id,
612 					     sbi->fsid);
613 	else if (sbi->fsid)
614 		super_set_sysfs_name_generic(sb, "%s", sbi->fsid);
615 	else if (erofs_is_fileio_mode(sbi))
616 		super_set_sysfs_name_generic(sb, "%s",
617 					     bdi_dev_name(sb->s_bdi));
618 	else
619 		super_set_sysfs_name_id(sb);
620 }
621 
erofs_fc_fill_super(struct super_block * sb,struct fs_context * fc)622 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
623 {
624 	struct inode *inode;
625 	struct erofs_sb_info *sbi = EROFS_SB(sb);
626 	int err;
627 
628 	sb->s_magic = EROFS_SUPER_MAGIC;
629 	sb->s_flags |= SB_RDONLY | SB_NOATIME;
630 	sb->s_maxbytes = MAX_LFS_FILESIZE;
631 	sb->s_op = &erofs_sops;
632 
633 	sbi->blkszbits = PAGE_SHIFT;
634 	if (!sb->s_bdev) {
635 		sb->s_blocksize = PAGE_SIZE;
636 		sb->s_blocksize_bits = PAGE_SHIFT;
637 
638 		if (erofs_is_fscache_mode(sb)) {
639 			err = erofs_fscache_register_fs(sb);
640 			if (err)
641 				return err;
642 		}
643 		err = super_setup_bdi(sb);
644 		if (err)
645 			return err;
646 	} else {
647 		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
648 			errorfc(fc, "failed to set initial blksize");
649 			return -EINVAL;
650 		}
651 
652 		sbi->dif0.dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
653 				&sbi->dif0.dax_part_off, NULL, NULL);
654 	}
655 
656 	err = erofs_read_superblock(sb);
657 	if (err)
658 		return err;
659 
660 	if (sb->s_blocksize_bits != sbi->blkszbits) {
661 		if (erofs_is_fscache_mode(sb)) {
662 			errorfc(fc, "unsupported blksize for fscache mode");
663 			return -EINVAL;
664 		}
665 
666 		if (erofs_is_fileio_mode(sbi)) {
667 			sb->s_blocksize = 1 << sbi->blkszbits;
668 			sb->s_blocksize_bits = sbi->blkszbits;
669 		} else if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
670 			errorfc(fc, "failed to set erofs blksize");
671 			return -EINVAL;
672 		}
673 	}
674 
675 	if (sbi->dif0.fsoff) {
676 		if (sbi->dif0.fsoff & (sb->s_blocksize - 1))
677 			return invalfc(fc, "fsoffset %llu is not aligned to block size %lu",
678 				       sbi->dif0.fsoff, sb->s_blocksize);
679 		if (erofs_is_fscache_mode(sb))
680 			return invalfc(fc, "cannot use fsoffset in fscache mode");
681 	}
682 
683 	if (test_opt(&sbi->opt, DAX_ALWAYS) && sbi->blkszbits != PAGE_SHIFT) {
684 		erofs_info(sb, "unsupported blocksize for DAX");
685 		clear_opt(&sbi->opt, DAX_ALWAYS);
686 	}
687 
688 	sb->s_time_gran = 1;
689 	sb->s_xattr = erofs_xattr_handlers;
690 	sb->s_export_op = &erofs_export_ops;
691 
692 	if (test_opt(&sbi->opt, POSIX_ACL))
693 		sb->s_flags |= SB_POSIXACL;
694 	else
695 		sb->s_flags &= ~SB_POSIXACL;
696 
697 	err = z_erofs_init_super(sb);
698 	if (err)
699 		return err;
700 
701 	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
702 		inode = erofs_iget(sb, sbi->packed_nid);
703 		if (IS_ERR(inode))
704 			return PTR_ERR(inode);
705 		sbi->packed_inode = inode;
706 	}
707 	if (erofs_sb_has_metabox(sbi)) {
708 		inode = erofs_iget(sb, sbi->metabox_nid);
709 		if (IS_ERR(inode))
710 			return PTR_ERR(inode);
711 		sbi->metabox_inode = inode;
712 	}
713 
714 	inode = erofs_iget(sb, sbi->root_nid);
715 	if (IS_ERR(inode))
716 		return PTR_ERR(inode);
717 
718 	if (!S_ISDIR(inode->i_mode)) {
719 		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
720 			  sbi->root_nid, inode->i_mode);
721 		iput(inode);
722 		return -EINVAL;
723 	}
724 	sb->s_root = d_make_root(inode);
725 	if (!sb->s_root)
726 		return -ENOMEM;
727 
728 	erofs_shrinker_register(sb);
729 	err = erofs_xattr_prefixes_init(sb);
730 	if (err)
731 		return err;
732 
733 	erofs_set_sysfs_name(sb);
734 	err = erofs_register_sysfs(sb);
735 	if (err)
736 		return err;
737 
738 	sbi->dir_ra_bytes = EROFS_DIR_RA_BYTES;
739 	erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
740 	return 0;
741 }
742 
erofs_fc_get_tree(struct fs_context * fc)743 static int erofs_fc_get_tree(struct fs_context *fc)
744 {
745 	struct erofs_sb_info *sbi = fc->s_fs_info;
746 	int ret;
747 
748 	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
749 		return get_tree_nodev(fc, erofs_fc_fill_super);
750 
751 	ret = get_tree_bdev_flags(fc, erofs_fc_fill_super,
752 		IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) ?
753 			GET_TREE_BDEV_QUIET_LOOKUP : 0);
754 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
755 	if (ret == -ENOTBLK) {
756 		struct file *file;
757 
758 		if (!fc->source)
759 			return invalf(fc, "No source specified");
760 		file = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
761 		if (IS_ERR(file))
762 			return PTR_ERR(file);
763 		sbi->dif0.file = file;
764 
765 		if (S_ISREG(file_inode(sbi->dif0.file)->i_mode) &&
766 		    sbi->dif0.file->f_mapping->a_ops->read_folio)
767 			return get_tree_nodev(fc, erofs_fc_fill_super);
768 	}
769 #endif
770 	return ret;
771 }
772 
erofs_fc_reconfigure(struct fs_context * fc)773 static int erofs_fc_reconfigure(struct fs_context *fc)
774 {
775 	struct super_block *sb = fc->root->d_sb;
776 	struct erofs_sb_info *sbi = EROFS_SB(sb);
777 	struct erofs_sb_info *new_sbi = fc->s_fs_info;
778 
779 	DBG_BUGON(!sb_rdonly(sb));
780 
781 	if (new_sbi->fsid || new_sbi->domain_id)
782 		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
783 
784 	if (test_opt(&new_sbi->opt, POSIX_ACL))
785 		fc->sb_flags |= SB_POSIXACL;
786 	else
787 		fc->sb_flags &= ~SB_POSIXACL;
788 
789 	sbi->opt = new_sbi->opt;
790 
791 	fc->sb_flags |= SB_RDONLY;
792 	return 0;
793 }
794 
erofs_release_device_info(int id,void * ptr,void * data)795 static int erofs_release_device_info(int id, void *ptr, void *data)
796 {
797 	struct erofs_device_info *dif = ptr;
798 
799 	fs_put_dax(dif->dax_dev, NULL);
800 	if (dif->file)
801 		fput(dif->file);
802 	erofs_fscache_unregister_cookie(dif->fscache);
803 	dif->fscache = NULL;
804 	kfree(dif->path);
805 	kfree(dif);
806 	return 0;
807 }
808 
erofs_free_dev_context(struct erofs_dev_context * devs)809 static void erofs_free_dev_context(struct erofs_dev_context *devs)
810 {
811 	if (!devs)
812 		return;
813 	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
814 	idr_destroy(&devs->tree);
815 	kfree(devs);
816 }
817 
erofs_sb_free(struct erofs_sb_info * sbi)818 static void erofs_sb_free(struct erofs_sb_info *sbi)
819 {
820 	erofs_free_dev_context(sbi->devs);
821 	kfree(sbi->fsid);
822 	kfree(sbi->domain_id);
823 	if (sbi->dif0.file)
824 		fput(sbi->dif0.file);
825 	kfree(sbi);
826 }
827 
erofs_fc_free(struct fs_context * fc)828 static void erofs_fc_free(struct fs_context *fc)
829 {
830 	struct erofs_sb_info *sbi = fc->s_fs_info;
831 
832 	if (sbi) /* free here if an error occurs before transferring to sb */
833 		erofs_sb_free(sbi);
834 }
835 
836 static const struct fs_context_operations erofs_context_ops = {
837 	.parse_param	= erofs_fc_parse_param,
838 	.get_tree       = erofs_fc_get_tree,
839 	.reconfigure    = erofs_fc_reconfigure,
840 	.free		= erofs_fc_free,
841 };
842 
erofs_init_fs_context(struct fs_context * fc)843 static int erofs_init_fs_context(struct fs_context *fc)
844 {
845 	struct erofs_sb_info *sbi;
846 
847 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
848 	if (!sbi)
849 		return -ENOMEM;
850 
851 	sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
852 	if (!sbi->devs) {
853 		kfree(sbi);
854 		return -ENOMEM;
855 	}
856 	fc->s_fs_info = sbi;
857 
858 	idr_init(&sbi->devs->tree);
859 	init_rwsem(&sbi->devs->rwsem);
860 	erofs_default_options(sbi);
861 	fc->ops = &erofs_context_ops;
862 	return 0;
863 }
864 
erofs_drop_internal_inodes(struct erofs_sb_info * sbi)865 static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
866 {
867 	iput(sbi->packed_inode);
868 	sbi->packed_inode = NULL;
869 	iput(sbi->metabox_inode);
870 	sbi->metabox_inode = NULL;
871 #ifdef CONFIG_EROFS_FS_ZIP
872 	iput(sbi->managed_cache);
873 	sbi->managed_cache = NULL;
874 #endif
875 }
876 
erofs_kill_sb(struct super_block * sb)877 static void erofs_kill_sb(struct super_block *sb)
878 {
879 	struct erofs_sb_info *sbi = EROFS_SB(sb);
880 
881 	if ((IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid) ||
882 	    sbi->dif0.file)
883 		kill_anon_super(sb);
884 	else
885 		kill_block_super(sb);
886 	erofs_drop_internal_inodes(sbi);
887 	fs_put_dax(sbi->dif0.dax_dev, NULL);
888 	erofs_fscache_unregister_fs(sb);
889 	erofs_sb_free(sbi);
890 	sb->s_fs_info = NULL;
891 }
892 
erofs_put_super(struct super_block * sb)893 static void erofs_put_super(struct super_block *sb)
894 {
895 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
896 
897 	erofs_unregister_sysfs(sb);
898 	erofs_shrinker_unregister(sb);
899 	erofs_xattr_prefixes_cleanup(sb);
900 	erofs_drop_internal_inodes(sbi);
901 	erofs_free_dev_context(sbi->devs);
902 	sbi->devs = NULL;
903 	erofs_fscache_unregister_fs(sb);
904 }
905 
906 static struct file_system_type erofs_fs_type = {
907 	.owner          = THIS_MODULE,
908 	.name           = "erofs",
909 	.init_fs_context = erofs_init_fs_context,
910 	.kill_sb        = erofs_kill_sb,
911 	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
912 };
913 MODULE_ALIAS_FS("erofs");
914 
erofs_module_init(void)915 static int __init erofs_module_init(void)
916 {
917 	int err;
918 
919 	erofs_check_ondisk_layout_definitions();
920 
921 	erofs_inode_cachep = kmem_cache_create("erofs_inode",
922 			sizeof(struct erofs_inode), 0,
923 			SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
924 			erofs_inode_init_once);
925 	if (!erofs_inode_cachep)
926 		return -ENOMEM;
927 
928 	err = erofs_init_shrinker();
929 	if (err)
930 		goto shrinker_err;
931 
932 	err = z_erofs_init_subsystem();
933 	if (err)
934 		goto zip_err;
935 
936 	err = erofs_init_sysfs();
937 	if (err)
938 		goto sysfs_err;
939 
940 	err = register_filesystem(&erofs_fs_type);
941 	if (err)
942 		goto fs_err;
943 
944 	return 0;
945 
946 fs_err:
947 	erofs_exit_sysfs();
948 sysfs_err:
949 	z_erofs_exit_subsystem();
950 zip_err:
951 	erofs_exit_shrinker();
952 shrinker_err:
953 	kmem_cache_destroy(erofs_inode_cachep);
954 	return err;
955 }
956 
erofs_module_exit(void)957 static void __exit erofs_module_exit(void)
958 {
959 	unregister_filesystem(&erofs_fs_type);
960 
961 	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
962 	rcu_barrier();
963 
964 	erofs_exit_sysfs();
965 	z_erofs_exit_subsystem();
966 	erofs_exit_shrinker();
967 	kmem_cache_destroy(erofs_inode_cachep);
968 }
969 
erofs_statfs(struct dentry * dentry,struct kstatfs * buf)970 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
971 {
972 	struct super_block *sb = dentry->d_sb;
973 	struct erofs_sb_info *sbi = EROFS_SB(sb);
974 
975 	buf->f_type = sb->s_magic;
976 	buf->f_bsize = sb->s_blocksize;
977 	buf->f_blocks = sbi->total_blocks;
978 	buf->f_bfree = buf->f_bavail = 0;
979 	buf->f_files = ULLONG_MAX;
980 	buf->f_ffree = ULLONG_MAX - sbi->inos;
981 	buf->f_namelen = EROFS_NAME_LEN;
982 
983 	if (uuid_is_null(&sb->s_uuid))
984 		buf->f_fsid = u64_to_fsid(!sb->s_bdev ? 0 :
985 				huge_encode_dev(sb->s_bdev->bd_dev));
986 	else
987 		buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
988 	return 0;
989 }
990 
erofs_show_options(struct seq_file * seq,struct dentry * root)991 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
992 {
993 	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
994 	struct erofs_mount_opts *opt = &sbi->opt;
995 
996 	if (IS_ENABLED(CONFIG_EROFS_FS_XATTR))
997 		seq_puts(seq, test_opt(opt, XATTR_USER) ?
998 				",user_xattr" : ",nouser_xattr");
999 	if (IS_ENABLED(CONFIG_EROFS_FS_POSIX_ACL))
1000 		seq_puts(seq, test_opt(opt, POSIX_ACL) ? ",acl" : ",noacl");
1001 	if (IS_ENABLED(CONFIG_EROFS_FS_ZIP))
1002 		seq_printf(seq, ",cache_strategy=%s",
1003 			  erofs_param_cache_strategy[opt->cache_strategy].name);
1004 	if (test_opt(opt, DAX_ALWAYS))
1005 		seq_puts(seq, ",dax=always");
1006 	if (test_opt(opt, DAX_NEVER))
1007 		seq_puts(seq, ",dax=never");
1008 	if (erofs_is_fileio_mode(sbi) && test_opt(opt, DIRECT_IO))
1009 		seq_puts(seq, ",directio");
1010 #ifdef CONFIG_EROFS_FS_ONDEMAND
1011 	if (sbi->fsid)
1012 		seq_printf(seq, ",fsid=%s", sbi->fsid);
1013 	if (sbi->domain_id)
1014 		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
1015 #endif
1016 	if (sbi->dif0.fsoff)
1017 		seq_printf(seq, ",fsoffset=%llu", sbi->dif0.fsoff);
1018 	return 0;
1019 }
1020 
1021 const struct super_operations erofs_sops = {
1022 	.put_super = erofs_put_super,
1023 	.alloc_inode = erofs_alloc_inode,
1024 	.free_inode = erofs_free_inode,
1025 	.statfs = erofs_statfs,
1026 	.show_options = erofs_show_options,
1027 };
1028 
1029 module_init(erofs_module_init);
1030 module_exit(erofs_module_exit);
1031 
1032 MODULE_DESCRIPTION("Enhanced ROM File System");
1033 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
1034 MODULE_LICENSE("GPL");
1035