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, EROFS_KMAP);
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, EROFS_KMAP);
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, EROFS_KMAP);
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 			return PTR_ERR(file);
170 
171 		if (!erofs_is_fileio_mode(sbi)) {
172 			dif->dax_dev = fs_dax_get_by_bdev(file_bdev(file),
173 					&dif->dax_part_off, NULL, NULL);
174 		} else if (!S_ISREG(file_inode(file)->i_mode)) {
175 			fput(file);
176 			return -EINVAL;
177 		}
178 		dif->file = file;
179 	}
180 
181 	dif->blocks = le32_to_cpu(dis->blocks);
182 	dif->mapped_blkaddr = le32_to_cpu(dis->mapped_blkaddr);
183 	sbi->total_blocks += dif->blocks;
184 	*pos += EROFS_DEVT_SLOT_SIZE;
185 	return 0;
186 }
187 
erofs_scan_devices(struct super_block * sb,struct erofs_super_block * dsb)188 static int erofs_scan_devices(struct super_block *sb,
189 			      struct erofs_super_block *dsb)
190 {
191 	struct erofs_sb_info *sbi = EROFS_SB(sb);
192 	unsigned int ondisk_extradevs;
193 	erofs_off_t pos;
194 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
195 	struct erofs_device_info *dif;
196 	int id, err = 0;
197 
198 	sbi->total_blocks = sbi->dif0.blocks;
199 	if (!erofs_sb_has_device_table(sbi))
200 		ondisk_extradevs = 0;
201 	else
202 		ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
203 
204 	if (sbi->devs->extra_devices &&
205 	    ondisk_extradevs != sbi->devs->extra_devices) {
206 		erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
207 			  ondisk_extradevs, sbi->devs->extra_devices);
208 		return -EINVAL;
209 	}
210 	if (!ondisk_extradevs)
211 		return 0;
212 
213 	if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
214 		sbi->devs->flatdev = true;
215 
216 	sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
217 	pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
218 	down_read(&sbi->devs->rwsem);
219 	if (sbi->devs->extra_devices) {
220 		idr_for_each_entry(&sbi->devs->tree, dif, id) {
221 			err = erofs_init_device(&buf, sb, dif, &pos);
222 			if (err)
223 				break;
224 		}
225 	} else {
226 		for (id = 0; id < ondisk_extradevs; id++) {
227 			dif = kzalloc(sizeof(*dif), GFP_KERNEL);
228 			if (!dif) {
229 				err = -ENOMEM;
230 				break;
231 			}
232 
233 			err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
234 			if (err < 0) {
235 				kfree(dif);
236 				break;
237 			}
238 			++sbi->devs->extra_devices;
239 
240 			err = erofs_init_device(&buf, sb, dif, &pos);
241 			if (err)
242 				break;
243 		}
244 	}
245 	up_read(&sbi->devs->rwsem);
246 	erofs_put_metabuf(&buf);
247 	return err;
248 }
249 
erofs_read_superblock(struct super_block * sb)250 static int erofs_read_superblock(struct super_block *sb)
251 {
252 	struct erofs_sb_info *sbi = EROFS_SB(sb);
253 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
254 	struct erofs_super_block *dsb;
255 	void *data;
256 	int ret;
257 
258 	data = erofs_read_metabuf(&buf, sb, 0, EROFS_KMAP);
259 	if (IS_ERR(data)) {
260 		erofs_err(sb, "cannot read erofs superblock");
261 		return PTR_ERR(data);
262 	}
263 
264 	dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
265 	ret = -EINVAL;
266 	if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
267 		erofs_err(sb, "cannot find valid erofs superblock");
268 		goto out;
269 	}
270 
271 	sbi->blkszbits  = dsb->blkszbits;
272 	if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
273 		erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
274 		goto out;
275 	}
276 	if (dsb->dirblkbits) {
277 		erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
278 		goto out;
279 	}
280 
281 	sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
282 	if (erofs_sb_has_sb_chksum(sbi)) {
283 		ret = erofs_superblock_csum_verify(sb, data);
284 		if (ret)
285 			goto out;
286 	}
287 
288 	ret = -EINVAL;
289 	sbi->feature_incompat = le32_to_cpu(dsb->feature_incompat);
290 	if (sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT) {
291 		erofs_err(sb, "unidentified incompatible feature %x, please upgrade kernel",
292 			  sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT);
293 		goto out;
294 	}
295 
296 	sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
297 	if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
298 		erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
299 			  sbi->sb_size);
300 		goto out;
301 	}
302 	sbi->dif0.blocks = le32_to_cpu(dsb->blocks);
303 	sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
304 #ifdef CONFIG_EROFS_FS_XATTR
305 	sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
306 	sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
307 	sbi->xattr_prefix_count = dsb->xattr_prefix_count;
308 	sbi->xattr_filter_reserved = dsb->xattr_filter_reserved;
309 #endif
310 	sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
311 	sbi->root_nid = le16_to_cpu(dsb->root_nid);
312 	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
313 	sbi->inos = le64_to_cpu(dsb->inos);
314 
315 	sbi->build_time = le64_to_cpu(dsb->build_time);
316 	sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
317 
318 	super_set_uuid(sb, (void *)dsb->uuid, sizeof(dsb->uuid));
319 
320 	/* parse on-disk compression configurations */
321 	ret = z_erofs_parse_cfgs(sb, dsb);
322 	if (ret < 0)
323 		goto out;
324 
325 	/* handle multiple devices */
326 	ret = erofs_scan_devices(sb, dsb);
327 
328 	if (erofs_is_fscache_mode(sb))
329 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
330 out:
331 	erofs_put_metabuf(&buf);
332 	return ret;
333 }
334 
erofs_default_options(struct erofs_sb_info * sbi)335 static void erofs_default_options(struct erofs_sb_info *sbi)
336 {
337 #ifdef CONFIG_EROFS_FS_ZIP
338 	sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
339 	sbi->opt.max_sync_decompress_pages = 3;
340 	sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
341 #endif
342 #ifdef CONFIG_EROFS_FS_XATTR
343 	set_opt(&sbi->opt, XATTR_USER);
344 #endif
345 #ifdef CONFIG_EROFS_FS_POSIX_ACL
346 	set_opt(&sbi->opt, POSIX_ACL);
347 #endif
348 }
349 
350 enum {
351 	Opt_user_xattr, Opt_acl, Opt_cache_strategy, Opt_dax, Opt_dax_enum,
352 	Opt_device, Opt_fsid, Opt_domain_id, Opt_directio,
353 	Opt_err
354 };
355 
356 static const struct constant_table erofs_param_cache_strategy[] = {
357 	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
358 	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
359 	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
360 	{}
361 };
362 
363 static const struct constant_table erofs_dax_param_enums[] = {
364 	{"always",	EROFS_MOUNT_DAX_ALWAYS},
365 	{"never",	EROFS_MOUNT_DAX_NEVER},
366 	{}
367 };
368 
369 static const struct fs_parameter_spec erofs_fs_parameters[] = {
370 	fsparam_flag_no("user_xattr",	Opt_user_xattr),
371 	fsparam_flag_no("acl",		Opt_acl),
372 	fsparam_enum("cache_strategy",	Opt_cache_strategy,
373 		     erofs_param_cache_strategy),
374 	fsparam_flag("dax",             Opt_dax),
375 	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
376 	fsparam_string("device",	Opt_device),
377 	fsparam_string("fsid",		Opt_fsid),
378 	fsparam_string("domain_id",	Opt_domain_id),
379 	fsparam_flag_no("directio",	Opt_directio),
380 	{}
381 };
382 
erofs_fc_set_dax_mode(struct fs_context * fc,unsigned int mode)383 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
384 {
385 #ifdef CONFIG_FS_DAX
386 	struct erofs_sb_info *sbi = fc->s_fs_info;
387 
388 	switch (mode) {
389 	case EROFS_MOUNT_DAX_ALWAYS:
390 		set_opt(&sbi->opt, DAX_ALWAYS);
391 		clear_opt(&sbi->opt, DAX_NEVER);
392 		return true;
393 	case EROFS_MOUNT_DAX_NEVER:
394 		set_opt(&sbi->opt, DAX_NEVER);
395 		clear_opt(&sbi->opt, DAX_ALWAYS);
396 		return true;
397 	default:
398 		DBG_BUGON(1);
399 		return false;
400 	}
401 #else
402 	errorfc(fc, "dax options not supported");
403 	return false;
404 #endif
405 }
406 
erofs_fc_parse_param(struct fs_context * fc,struct fs_parameter * param)407 static int erofs_fc_parse_param(struct fs_context *fc,
408 				struct fs_parameter *param)
409 {
410 	struct erofs_sb_info *sbi = fc->s_fs_info;
411 	struct fs_parse_result result;
412 	struct erofs_device_info *dif;
413 	int opt, ret;
414 
415 	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
416 	if (opt < 0)
417 		return opt;
418 
419 	switch (opt) {
420 	case Opt_user_xattr:
421 #ifdef CONFIG_EROFS_FS_XATTR
422 		if (result.boolean)
423 			set_opt(&sbi->opt, XATTR_USER);
424 		else
425 			clear_opt(&sbi->opt, XATTR_USER);
426 #else
427 		errorfc(fc, "{,no}user_xattr options not supported");
428 #endif
429 		break;
430 	case Opt_acl:
431 #ifdef CONFIG_EROFS_FS_POSIX_ACL
432 		if (result.boolean)
433 			set_opt(&sbi->opt, POSIX_ACL);
434 		else
435 			clear_opt(&sbi->opt, POSIX_ACL);
436 #else
437 		errorfc(fc, "{,no}acl options not supported");
438 #endif
439 		break;
440 	case Opt_cache_strategy:
441 #ifdef CONFIG_EROFS_FS_ZIP
442 		sbi->opt.cache_strategy = result.uint_32;
443 #else
444 		errorfc(fc, "compression not supported, cache_strategy ignored");
445 #endif
446 		break;
447 	case Opt_dax:
448 		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
449 			return -EINVAL;
450 		break;
451 	case Opt_dax_enum:
452 		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
453 			return -EINVAL;
454 		break;
455 	case Opt_device:
456 		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
457 		if (!dif)
458 			return -ENOMEM;
459 		dif->path = kstrdup(param->string, GFP_KERNEL);
460 		if (!dif->path) {
461 			kfree(dif);
462 			return -ENOMEM;
463 		}
464 		down_write(&sbi->devs->rwsem);
465 		ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
466 		up_write(&sbi->devs->rwsem);
467 		if (ret < 0) {
468 			kfree(dif->path);
469 			kfree(dif);
470 			return ret;
471 		}
472 		++sbi->devs->extra_devices;
473 		break;
474 #ifdef CONFIG_EROFS_FS_ONDEMAND
475 	case Opt_fsid:
476 		kfree(sbi->fsid);
477 		sbi->fsid = kstrdup(param->string, GFP_KERNEL);
478 		if (!sbi->fsid)
479 			return -ENOMEM;
480 		break;
481 	case Opt_domain_id:
482 		kfree(sbi->domain_id);
483 		sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
484 		if (!sbi->domain_id)
485 			return -ENOMEM;
486 		break;
487 #else
488 	case Opt_fsid:
489 	case Opt_domain_id:
490 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
491 		break;
492 #endif
493 	case Opt_directio:
494 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
495 		if (result.boolean)
496 			set_opt(&sbi->opt, DIRECT_IO);
497 		else
498 			clear_opt(&sbi->opt, DIRECT_IO);
499 #else
500 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
501 #endif
502 		break;
503 	}
504 	return 0;
505 }
506 
erofs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)507 static struct inode *erofs_nfs_get_inode(struct super_block *sb,
508 					 u64 ino, u32 generation)
509 {
510 	return erofs_iget(sb, ino);
511 }
512 
erofs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)513 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
514 		struct fid *fid, int fh_len, int fh_type)
515 {
516 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
517 				    erofs_nfs_get_inode);
518 }
519 
erofs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)520 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
521 		struct fid *fid, int fh_len, int fh_type)
522 {
523 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
524 				    erofs_nfs_get_inode);
525 }
526 
erofs_get_parent(struct dentry * child)527 static struct dentry *erofs_get_parent(struct dentry *child)
528 {
529 	erofs_nid_t nid;
530 	unsigned int d_type;
531 	int err;
532 
533 	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
534 	if (err)
535 		return ERR_PTR(err);
536 	return d_obtain_alias(erofs_iget(child->d_sb, nid));
537 }
538 
539 static const struct export_operations erofs_export_ops = {
540 	.encode_fh = generic_encode_ino32_fh,
541 	.fh_to_dentry = erofs_fh_to_dentry,
542 	.fh_to_parent = erofs_fh_to_parent,
543 	.get_parent = erofs_get_parent,
544 };
545 
erofs_set_sysfs_name(struct super_block * sb)546 static void erofs_set_sysfs_name(struct super_block *sb)
547 {
548 	struct erofs_sb_info *sbi = EROFS_SB(sb);
549 
550 	if (sbi->domain_id)
551 		super_set_sysfs_name_generic(sb, "%s,%s", sbi->domain_id,
552 					     sbi->fsid);
553 	else if (sbi->fsid)
554 		super_set_sysfs_name_generic(sb, "%s", sbi->fsid);
555 	else if (erofs_is_fileio_mode(sbi))
556 		super_set_sysfs_name_generic(sb, "%s",
557 					     bdi_dev_name(sb->s_bdi));
558 	else
559 		super_set_sysfs_name_id(sb);
560 }
561 
erofs_fc_fill_super(struct super_block * sb,struct fs_context * fc)562 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
563 {
564 	struct inode *inode;
565 	struct erofs_sb_info *sbi = EROFS_SB(sb);
566 	int err;
567 
568 	sb->s_magic = EROFS_SUPER_MAGIC;
569 	sb->s_flags |= SB_RDONLY | SB_NOATIME;
570 	sb->s_maxbytes = MAX_LFS_FILESIZE;
571 	sb->s_op = &erofs_sops;
572 
573 	sbi->blkszbits = PAGE_SHIFT;
574 	if (!sb->s_bdev) {
575 		sb->s_blocksize = PAGE_SIZE;
576 		sb->s_blocksize_bits = PAGE_SHIFT;
577 
578 		if (erofs_is_fscache_mode(sb)) {
579 			err = erofs_fscache_register_fs(sb);
580 			if (err)
581 				return err;
582 		}
583 		err = super_setup_bdi(sb);
584 		if (err)
585 			return err;
586 	} else {
587 		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
588 			errorfc(fc, "failed to set initial blksize");
589 			return -EINVAL;
590 		}
591 
592 		sbi->dif0.dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
593 				&sbi->dif0.dax_part_off, NULL, NULL);
594 	}
595 
596 	err = erofs_read_superblock(sb);
597 	if (err)
598 		return err;
599 
600 	if (sb->s_blocksize_bits != sbi->blkszbits) {
601 		if (erofs_is_fscache_mode(sb)) {
602 			errorfc(fc, "unsupported blksize for fscache mode");
603 			return -EINVAL;
604 		}
605 
606 		if (erofs_is_fileio_mode(sbi)) {
607 			sb->s_blocksize = 1 << sbi->blkszbits;
608 			sb->s_blocksize_bits = sbi->blkszbits;
609 		} else if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
610 			errorfc(fc, "failed to set erofs blksize");
611 			return -EINVAL;
612 		}
613 	}
614 
615 	if (test_opt(&sbi->opt, DAX_ALWAYS)) {
616 		if (!sbi->dif0.dax_dev) {
617 			errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
618 			clear_opt(&sbi->opt, DAX_ALWAYS);
619 		} else if (sbi->blkszbits != PAGE_SHIFT) {
620 			errorfc(fc, "unsupported blocksize for DAX");
621 			clear_opt(&sbi->opt, DAX_ALWAYS);
622 		}
623 	}
624 
625 	sb->s_time_gran = 1;
626 	sb->s_xattr = erofs_xattr_handlers;
627 	sb->s_export_op = &erofs_export_ops;
628 
629 	if (test_opt(&sbi->opt, POSIX_ACL))
630 		sb->s_flags |= SB_POSIXACL;
631 	else
632 		sb->s_flags &= ~SB_POSIXACL;
633 
634 #ifdef CONFIG_EROFS_FS_ZIP
635 	xa_init(&sbi->managed_pslots);
636 #endif
637 
638 	inode = erofs_iget(sb, sbi->root_nid);
639 	if (IS_ERR(inode))
640 		return PTR_ERR(inode);
641 
642 	if (!S_ISDIR(inode->i_mode)) {
643 		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
644 			  sbi->root_nid, inode->i_mode);
645 		iput(inode);
646 		return -EINVAL;
647 	}
648 
649 	sb->s_root = d_make_root(inode);
650 	if (!sb->s_root)
651 		return -ENOMEM;
652 
653 	erofs_shrinker_register(sb);
654 	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
655 		sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
656 		if (IS_ERR(sbi->packed_inode)) {
657 			err = PTR_ERR(sbi->packed_inode);
658 			sbi->packed_inode = NULL;
659 			return err;
660 		}
661 	}
662 	err = erofs_init_managed_cache(sb);
663 	if (err)
664 		return err;
665 
666 	err = erofs_xattr_prefixes_init(sb);
667 	if (err)
668 		return err;
669 
670 	erofs_set_sysfs_name(sb);
671 	err = erofs_register_sysfs(sb);
672 	if (err)
673 		return err;
674 
675 	erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
676 	return 0;
677 }
678 
erofs_fc_get_tree(struct fs_context * fc)679 static int erofs_fc_get_tree(struct fs_context *fc)
680 {
681 	struct erofs_sb_info *sbi = fc->s_fs_info;
682 	int ret;
683 
684 	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
685 		return get_tree_nodev(fc, erofs_fc_fill_super);
686 
687 	ret = get_tree_bdev_flags(fc, erofs_fc_fill_super,
688 		IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) ?
689 			GET_TREE_BDEV_QUIET_LOOKUP : 0);
690 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
691 	if (ret == -ENOTBLK) {
692 		struct file *file;
693 
694 		if (!fc->source)
695 			return invalf(fc, "No source specified");
696 		file = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
697 		if (IS_ERR(file))
698 			return PTR_ERR(file);
699 		sbi->dif0.file = file;
700 
701 		if (S_ISREG(file_inode(sbi->dif0.file)->i_mode) &&
702 		    sbi->dif0.file->f_mapping->a_ops->read_folio)
703 			return get_tree_nodev(fc, erofs_fc_fill_super);
704 	}
705 #endif
706 	return ret;
707 }
708 
erofs_fc_reconfigure(struct fs_context * fc)709 static int erofs_fc_reconfigure(struct fs_context *fc)
710 {
711 	struct super_block *sb = fc->root->d_sb;
712 	struct erofs_sb_info *sbi = EROFS_SB(sb);
713 	struct erofs_sb_info *new_sbi = fc->s_fs_info;
714 
715 	DBG_BUGON(!sb_rdonly(sb));
716 
717 	if (new_sbi->fsid || new_sbi->domain_id)
718 		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
719 
720 	if (test_opt(&new_sbi->opt, POSIX_ACL))
721 		fc->sb_flags |= SB_POSIXACL;
722 	else
723 		fc->sb_flags &= ~SB_POSIXACL;
724 
725 	sbi->opt = new_sbi->opt;
726 
727 	fc->sb_flags |= SB_RDONLY;
728 	return 0;
729 }
730 
erofs_release_device_info(int id,void * ptr,void * data)731 static int erofs_release_device_info(int id, void *ptr, void *data)
732 {
733 	struct erofs_device_info *dif = ptr;
734 
735 	fs_put_dax(dif->dax_dev, NULL);
736 	if (dif->file)
737 		fput(dif->file);
738 	erofs_fscache_unregister_cookie(dif->fscache);
739 	dif->fscache = NULL;
740 	kfree(dif->path);
741 	kfree(dif);
742 	return 0;
743 }
744 
erofs_free_dev_context(struct erofs_dev_context * devs)745 static void erofs_free_dev_context(struct erofs_dev_context *devs)
746 {
747 	if (!devs)
748 		return;
749 	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
750 	idr_destroy(&devs->tree);
751 	kfree(devs);
752 }
753 
erofs_sb_free(struct erofs_sb_info * sbi)754 static void erofs_sb_free(struct erofs_sb_info *sbi)
755 {
756 	erofs_free_dev_context(sbi->devs);
757 	kfree(sbi->fsid);
758 	kfree(sbi->domain_id);
759 	if (sbi->dif0.file)
760 		fput(sbi->dif0.file);
761 	kfree(sbi);
762 }
763 
erofs_fc_free(struct fs_context * fc)764 static void erofs_fc_free(struct fs_context *fc)
765 {
766 	struct erofs_sb_info *sbi = fc->s_fs_info;
767 
768 	if (sbi) /* free here if an error occurs before transferring to sb */
769 		erofs_sb_free(sbi);
770 }
771 
772 static const struct fs_context_operations erofs_context_ops = {
773 	.parse_param	= erofs_fc_parse_param,
774 	.get_tree       = erofs_fc_get_tree,
775 	.reconfigure    = erofs_fc_reconfigure,
776 	.free		= erofs_fc_free,
777 };
778 
erofs_init_fs_context(struct fs_context * fc)779 static int erofs_init_fs_context(struct fs_context *fc)
780 {
781 	struct erofs_sb_info *sbi;
782 
783 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
784 	if (!sbi)
785 		return -ENOMEM;
786 
787 	sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
788 	if (!sbi->devs) {
789 		kfree(sbi);
790 		return -ENOMEM;
791 	}
792 	fc->s_fs_info = sbi;
793 
794 	idr_init(&sbi->devs->tree);
795 	init_rwsem(&sbi->devs->rwsem);
796 	erofs_default_options(sbi);
797 	fc->ops = &erofs_context_ops;
798 	return 0;
799 }
800 
erofs_kill_sb(struct super_block * sb)801 static void erofs_kill_sb(struct super_block *sb)
802 {
803 	struct erofs_sb_info *sbi = EROFS_SB(sb);
804 
805 	if ((IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid) ||
806 	    sbi->dif0.file)
807 		kill_anon_super(sb);
808 	else
809 		kill_block_super(sb);
810 	fs_put_dax(sbi->dif0.dax_dev, NULL);
811 	erofs_fscache_unregister_fs(sb);
812 	erofs_sb_free(sbi);
813 	sb->s_fs_info = NULL;
814 }
815 
erofs_put_super(struct super_block * sb)816 static void erofs_put_super(struct super_block *sb)
817 {
818 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
819 
820 	DBG_BUGON(!sbi);
821 
822 	erofs_unregister_sysfs(sb);
823 	erofs_shrinker_unregister(sb);
824 	erofs_xattr_prefixes_cleanup(sb);
825 #ifdef CONFIG_EROFS_FS_ZIP
826 	iput(sbi->managed_cache);
827 	sbi->managed_cache = NULL;
828 #endif
829 	iput(sbi->packed_inode);
830 	sbi->packed_inode = NULL;
831 	erofs_free_dev_context(sbi->devs);
832 	sbi->devs = NULL;
833 	erofs_fscache_unregister_fs(sb);
834 }
835 
836 static struct file_system_type erofs_fs_type = {
837 	.owner          = THIS_MODULE,
838 	.name           = "erofs",
839 	.init_fs_context = erofs_init_fs_context,
840 	.kill_sb        = erofs_kill_sb,
841 	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
842 };
843 MODULE_ALIAS_FS("erofs");
844 
erofs_module_init(void)845 static int __init erofs_module_init(void)
846 {
847 	int err;
848 
849 	erofs_check_ondisk_layout_definitions();
850 
851 	erofs_inode_cachep = kmem_cache_create("erofs_inode",
852 			sizeof(struct erofs_inode), 0,
853 			SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
854 			erofs_inode_init_once);
855 	if (!erofs_inode_cachep)
856 		return -ENOMEM;
857 
858 	err = erofs_init_shrinker();
859 	if (err)
860 		goto shrinker_err;
861 
862 	err = z_erofs_init_subsystem();
863 	if (err)
864 		goto zip_err;
865 
866 	err = erofs_init_sysfs();
867 	if (err)
868 		goto sysfs_err;
869 
870 	err = register_filesystem(&erofs_fs_type);
871 	if (err)
872 		goto fs_err;
873 
874 	return 0;
875 
876 fs_err:
877 	erofs_exit_sysfs();
878 sysfs_err:
879 	z_erofs_exit_subsystem();
880 zip_err:
881 	erofs_exit_shrinker();
882 shrinker_err:
883 	kmem_cache_destroy(erofs_inode_cachep);
884 	return err;
885 }
886 
erofs_module_exit(void)887 static void __exit erofs_module_exit(void)
888 {
889 	unregister_filesystem(&erofs_fs_type);
890 
891 	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
892 	rcu_barrier();
893 
894 	erofs_exit_sysfs();
895 	z_erofs_exit_subsystem();
896 	erofs_exit_shrinker();
897 	kmem_cache_destroy(erofs_inode_cachep);
898 }
899 
erofs_statfs(struct dentry * dentry,struct kstatfs * buf)900 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
901 {
902 	struct super_block *sb = dentry->d_sb;
903 	struct erofs_sb_info *sbi = EROFS_SB(sb);
904 
905 	buf->f_type = sb->s_magic;
906 	buf->f_bsize = sb->s_blocksize;
907 	buf->f_blocks = sbi->total_blocks;
908 	buf->f_bfree = buf->f_bavail = 0;
909 	buf->f_files = ULLONG_MAX;
910 	buf->f_ffree = ULLONG_MAX - sbi->inos;
911 	buf->f_namelen = EROFS_NAME_LEN;
912 
913 	if (uuid_is_null(&sb->s_uuid))
914 		buf->f_fsid = u64_to_fsid(!sb->s_bdev ? 0 :
915 				huge_encode_dev(sb->s_bdev->bd_dev));
916 	else
917 		buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
918 	return 0;
919 }
920 
erofs_show_options(struct seq_file * seq,struct dentry * root)921 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
922 {
923 	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
924 	struct erofs_mount_opts *opt = &sbi->opt;
925 
926 	if (IS_ENABLED(CONFIG_EROFS_FS_XATTR))
927 		seq_puts(seq, test_opt(opt, XATTR_USER) ?
928 				",user_xattr" : ",nouser_xattr");
929 	if (IS_ENABLED(CONFIG_EROFS_FS_POSIX_ACL))
930 		seq_puts(seq, test_opt(opt, POSIX_ACL) ? ",acl" : ",noacl");
931 	if (IS_ENABLED(CONFIG_EROFS_FS_ZIP))
932 		seq_printf(seq, ",cache_strategy=%s",
933 			  erofs_param_cache_strategy[opt->cache_strategy].name);
934 	if (test_opt(opt, DAX_ALWAYS))
935 		seq_puts(seq, ",dax=always");
936 	if (test_opt(opt, DAX_NEVER))
937 		seq_puts(seq, ",dax=never");
938 	if (erofs_is_fileio_mode(sbi) && test_opt(opt, DIRECT_IO))
939 		seq_puts(seq, ",directio");
940 #ifdef CONFIG_EROFS_FS_ONDEMAND
941 	if (sbi->fsid)
942 		seq_printf(seq, ",fsid=%s", sbi->fsid);
943 	if (sbi->domain_id)
944 		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
945 #endif
946 	return 0;
947 }
948 
949 const struct super_operations erofs_sops = {
950 	.put_super = erofs_put_super,
951 	.alloc_inode = erofs_alloc_inode,
952 	.free_inode = erofs_free_inode,
953 	.statfs = erofs_statfs,
954 	.show_options = erofs_show_options,
955 };
956 
957 module_init(erofs_module_init);
958 module_exit(erofs_module_exit);
959 
960 MODULE_DESCRIPTION("Enhanced ROM File System");
961 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
962 MODULE_LICENSE("GPL");
963