xref: /linux/fs/erofs/super.c (revision c2da8b3f914f83fb9089d26a692eb8f22146ddb9)
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 	ret = strscpy(sbi->volume_name, dsb->volume_name,
321 		      sizeof(dsb->volume_name));
322 	if (ret < 0) {	/* -E2BIG */
323 		erofs_err(sb, "bad volume name without NIL terminator");
324 		ret = -EFSCORRUPTED;
325 		goto out;
326 	}
327 
328 	/* parse on-disk compression configurations */
329 	ret = z_erofs_parse_cfgs(sb, dsb);
330 	if (ret < 0)
331 		goto out;
332 
333 	/* handle multiple devices */
334 	ret = erofs_scan_devices(sb, dsb);
335 
336 	if (erofs_is_fscache_mode(sb))
337 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
338 out:
339 	erofs_put_metabuf(&buf);
340 	return ret;
341 }
342 
erofs_default_options(struct erofs_sb_info * sbi)343 static void erofs_default_options(struct erofs_sb_info *sbi)
344 {
345 #ifdef CONFIG_EROFS_FS_ZIP
346 	sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
347 	sbi->opt.max_sync_decompress_pages = 3;
348 	sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
349 #endif
350 #ifdef CONFIG_EROFS_FS_XATTR
351 	set_opt(&sbi->opt, XATTR_USER);
352 #endif
353 #ifdef CONFIG_EROFS_FS_POSIX_ACL
354 	set_opt(&sbi->opt, POSIX_ACL);
355 #endif
356 }
357 
358 enum {
359 	Opt_user_xattr, Opt_acl, Opt_cache_strategy, Opt_dax, Opt_dax_enum,
360 	Opt_device, Opt_fsid, Opt_domain_id, Opt_directio,
361 	Opt_err
362 };
363 
364 static const struct constant_table erofs_param_cache_strategy[] = {
365 	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
366 	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
367 	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
368 	{}
369 };
370 
371 static const struct constant_table erofs_dax_param_enums[] = {
372 	{"always",	EROFS_MOUNT_DAX_ALWAYS},
373 	{"never",	EROFS_MOUNT_DAX_NEVER},
374 	{}
375 };
376 
377 static const struct fs_parameter_spec erofs_fs_parameters[] = {
378 	fsparam_flag_no("user_xattr",	Opt_user_xattr),
379 	fsparam_flag_no("acl",		Opt_acl),
380 	fsparam_enum("cache_strategy",	Opt_cache_strategy,
381 		     erofs_param_cache_strategy),
382 	fsparam_flag("dax",             Opt_dax),
383 	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
384 	fsparam_string("device",	Opt_device),
385 	fsparam_string("fsid",		Opt_fsid),
386 	fsparam_string("domain_id",	Opt_domain_id),
387 	fsparam_flag_no("directio",	Opt_directio),
388 	{}
389 };
390 
erofs_fc_set_dax_mode(struct fs_context * fc,unsigned int mode)391 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
392 {
393 #ifdef CONFIG_FS_DAX
394 	struct erofs_sb_info *sbi = fc->s_fs_info;
395 
396 	switch (mode) {
397 	case EROFS_MOUNT_DAX_ALWAYS:
398 		set_opt(&sbi->opt, DAX_ALWAYS);
399 		clear_opt(&sbi->opt, DAX_NEVER);
400 		return true;
401 	case EROFS_MOUNT_DAX_NEVER:
402 		set_opt(&sbi->opt, DAX_NEVER);
403 		clear_opt(&sbi->opt, DAX_ALWAYS);
404 		return true;
405 	default:
406 		DBG_BUGON(1);
407 		return false;
408 	}
409 #else
410 	errorfc(fc, "dax options not supported");
411 	return false;
412 #endif
413 }
414 
erofs_fc_parse_param(struct fs_context * fc,struct fs_parameter * param)415 static int erofs_fc_parse_param(struct fs_context *fc,
416 				struct fs_parameter *param)
417 {
418 	struct erofs_sb_info *sbi = fc->s_fs_info;
419 	struct fs_parse_result result;
420 	struct erofs_device_info *dif;
421 	int opt, ret;
422 
423 	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
424 	if (opt < 0)
425 		return opt;
426 
427 	switch (opt) {
428 	case Opt_user_xattr:
429 #ifdef CONFIG_EROFS_FS_XATTR
430 		if (result.boolean)
431 			set_opt(&sbi->opt, XATTR_USER);
432 		else
433 			clear_opt(&sbi->opt, XATTR_USER);
434 #else
435 		errorfc(fc, "{,no}user_xattr options not supported");
436 #endif
437 		break;
438 	case Opt_acl:
439 #ifdef CONFIG_EROFS_FS_POSIX_ACL
440 		if (result.boolean)
441 			set_opt(&sbi->opt, POSIX_ACL);
442 		else
443 			clear_opt(&sbi->opt, POSIX_ACL);
444 #else
445 		errorfc(fc, "{,no}acl options not supported");
446 #endif
447 		break;
448 	case Opt_cache_strategy:
449 #ifdef CONFIG_EROFS_FS_ZIP
450 		sbi->opt.cache_strategy = result.uint_32;
451 #else
452 		errorfc(fc, "compression not supported, cache_strategy ignored");
453 #endif
454 		break;
455 	case Opt_dax:
456 		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
457 			return -EINVAL;
458 		break;
459 	case Opt_dax_enum:
460 		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
461 			return -EINVAL;
462 		break;
463 	case Opt_device:
464 		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
465 		if (!dif)
466 			return -ENOMEM;
467 		dif->path = kstrdup(param->string, GFP_KERNEL);
468 		if (!dif->path) {
469 			kfree(dif);
470 			return -ENOMEM;
471 		}
472 		down_write(&sbi->devs->rwsem);
473 		ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
474 		up_write(&sbi->devs->rwsem);
475 		if (ret < 0) {
476 			kfree(dif->path);
477 			kfree(dif);
478 			return ret;
479 		}
480 		++sbi->devs->extra_devices;
481 		break;
482 #ifdef CONFIG_EROFS_FS_ONDEMAND
483 	case Opt_fsid:
484 		kfree(sbi->fsid);
485 		sbi->fsid = kstrdup(param->string, GFP_KERNEL);
486 		if (!sbi->fsid)
487 			return -ENOMEM;
488 		break;
489 	case Opt_domain_id:
490 		kfree(sbi->domain_id);
491 		sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
492 		if (!sbi->domain_id)
493 			return -ENOMEM;
494 		break;
495 #else
496 	case Opt_fsid:
497 	case Opt_domain_id:
498 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
499 		break;
500 #endif
501 	case Opt_directio:
502 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
503 		if (result.boolean)
504 			set_opt(&sbi->opt, DIRECT_IO);
505 		else
506 			clear_opt(&sbi->opt, DIRECT_IO);
507 #else
508 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
509 #endif
510 		break;
511 	}
512 	return 0;
513 }
514 
erofs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)515 static struct inode *erofs_nfs_get_inode(struct super_block *sb,
516 					 u64 ino, u32 generation)
517 {
518 	return erofs_iget(sb, ino);
519 }
520 
erofs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)521 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
522 		struct fid *fid, int fh_len, int fh_type)
523 {
524 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
525 				    erofs_nfs_get_inode);
526 }
527 
erofs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)528 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
529 		struct fid *fid, int fh_len, int fh_type)
530 {
531 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
532 				    erofs_nfs_get_inode);
533 }
534 
erofs_get_parent(struct dentry * child)535 static struct dentry *erofs_get_parent(struct dentry *child)
536 {
537 	erofs_nid_t nid;
538 	unsigned int d_type;
539 	int err;
540 
541 	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
542 	if (err)
543 		return ERR_PTR(err);
544 	return d_obtain_alias(erofs_iget(child->d_sb, nid));
545 }
546 
547 static const struct export_operations erofs_export_ops = {
548 	.encode_fh = generic_encode_ino32_fh,
549 	.fh_to_dentry = erofs_fh_to_dentry,
550 	.fh_to_parent = erofs_fh_to_parent,
551 	.get_parent = erofs_get_parent,
552 };
553 
erofs_set_sysfs_name(struct super_block * sb)554 static void erofs_set_sysfs_name(struct super_block *sb)
555 {
556 	struct erofs_sb_info *sbi = EROFS_SB(sb);
557 
558 	if (sbi->domain_id)
559 		super_set_sysfs_name_generic(sb, "%s,%s", sbi->domain_id,
560 					     sbi->fsid);
561 	else if (sbi->fsid)
562 		super_set_sysfs_name_generic(sb, "%s", sbi->fsid);
563 	else if (erofs_is_fileio_mode(sbi))
564 		super_set_sysfs_name_generic(sb, "%s",
565 					     bdi_dev_name(sb->s_bdi));
566 	else
567 		super_set_sysfs_name_id(sb);
568 }
569 
erofs_fc_fill_super(struct super_block * sb,struct fs_context * fc)570 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
571 {
572 	struct inode *inode;
573 	struct erofs_sb_info *sbi = EROFS_SB(sb);
574 	int err;
575 
576 	sb->s_magic = EROFS_SUPER_MAGIC;
577 	sb->s_flags |= SB_RDONLY | SB_NOATIME;
578 	sb->s_maxbytes = MAX_LFS_FILESIZE;
579 	sb->s_op = &erofs_sops;
580 
581 	sbi->blkszbits = PAGE_SHIFT;
582 	if (!sb->s_bdev) {
583 		sb->s_blocksize = PAGE_SIZE;
584 		sb->s_blocksize_bits = PAGE_SHIFT;
585 
586 		if (erofs_is_fscache_mode(sb)) {
587 			err = erofs_fscache_register_fs(sb);
588 			if (err)
589 				return err;
590 		}
591 		err = super_setup_bdi(sb);
592 		if (err)
593 			return err;
594 	} else {
595 		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
596 			errorfc(fc, "failed to set initial blksize");
597 			return -EINVAL;
598 		}
599 
600 		sbi->dif0.dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
601 				&sbi->dif0.dax_part_off, NULL, NULL);
602 	}
603 
604 	err = erofs_read_superblock(sb);
605 	if (err)
606 		return err;
607 
608 	if (sb->s_blocksize_bits != sbi->blkszbits) {
609 		if (erofs_is_fscache_mode(sb)) {
610 			errorfc(fc, "unsupported blksize for fscache mode");
611 			return -EINVAL;
612 		}
613 
614 		if (erofs_is_fileio_mode(sbi)) {
615 			sb->s_blocksize = 1 << sbi->blkszbits;
616 			sb->s_blocksize_bits = sbi->blkszbits;
617 		} else if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
618 			errorfc(fc, "failed to set erofs blksize");
619 			return -EINVAL;
620 		}
621 	}
622 
623 	if (test_opt(&sbi->opt, DAX_ALWAYS)) {
624 		if (!sbi->dif0.dax_dev) {
625 			errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
626 			clear_opt(&sbi->opt, DAX_ALWAYS);
627 		} else if (sbi->blkszbits != PAGE_SHIFT) {
628 			errorfc(fc, "unsupported blocksize for DAX");
629 			clear_opt(&sbi->opt, DAX_ALWAYS);
630 		}
631 	}
632 
633 	sb->s_time_gran = 1;
634 	sb->s_xattr = erofs_xattr_handlers;
635 	sb->s_export_op = &erofs_export_ops;
636 
637 	if (test_opt(&sbi->opt, POSIX_ACL))
638 		sb->s_flags |= SB_POSIXACL;
639 	else
640 		sb->s_flags &= ~SB_POSIXACL;
641 
642 #ifdef CONFIG_EROFS_FS_ZIP
643 	xa_init(&sbi->managed_pslots);
644 #endif
645 
646 	inode = erofs_iget(sb, sbi->root_nid);
647 	if (IS_ERR(inode))
648 		return PTR_ERR(inode);
649 
650 	if (!S_ISDIR(inode->i_mode)) {
651 		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
652 			  sbi->root_nid, inode->i_mode);
653 		iput(inode);
654 		return -EINVAL;
655 	}
656 
657 	sb->s_root = d_make_root(inode);
658 	if (!sb->s_root)
659 		return -ENOMEM;
660 
661 	erofs_shrinker_register(sb);
662 	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
663 		sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
664 		if (IS_ERR(sbi->packed_inode)) {
665 			err = PTR_ERR(sbi->packed_inode);
666 			sbi->packed_inode = NULL;
667 			return err;
668 		}
669 	}
670 	err = erofs_init_managed_cache(sb);
671 	if (err)
672 		return err;
673 
674 	err = erofs_xattr_prefixes_init(sb);
675 	if (err)
676 		return err;
677 
678 	erofs_set_sysfs_name(sb);
679 	err = erofs_register_sysfs(sb);
680 	if (err)
681 		return err;
682 
683 	erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
684 	return 0;
685 }
686 
erofs_fc_get_tree(struct fs_context * fc)687 static int erofs_fc_get_tree(struct fs_context *fc)
688 {
689 	struct erofs_sb_info *sbi = fc->s_fs_info;
690 	int ret;
691 
692 	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
693 		return get_tree_nodev(fc, erofs_fc_fill_super);
694 
695 	ret = get_tree_bdev_flags(fc, erofs_fc_fill_super,
696 		IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) ?
697 			GET_TREE_BDEV_QUIET_LOOKUP : 0);
698 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
699 	if (ret == -ENOTBLK) {
700 		struct file *file;
701 
702 		if (!fc->source)
703 			return invalf(fc, "No source specified");
704 		file = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
705 		if (IS_ERR(file))
706 			return PTR_ERR(file);
707 		sbi->dif0.file = file;
708 
709 		if (S_ISREG(file_inode(sbi->dif0.file)->i_mode) &&
710 		    sbi->dif0.file->f_mapping->a_ops->read_folio)
711 			return get_tree_nodev(fc, erofs_fc_fill_super);
712 	}
713 #endif
714 	return ret;
715 }
716 
erofs_fc_reconfigure(struct fs_context * fc)717 static int erofs_fc_reconfigure(struct fs_context *fc)
718 {
719 	struct super_block *sb = fc->root->d_sb;
720 	struct erofs_sb_info *sbi = EROFS_SB(sb);
721 	struct erofs_sb_info *new_sbi = fc->s_fs_info;
722 
723 	DBG_BUGON(!sb_rdonly(sb));
724 
725 	if (new_sbi->fsid || new_sbi->domain_id)
726 		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
727 
728 	if (test_opt(&new_sbi->opt, POSIX_ACL))
729 		fc->sb_flags |= SB_POSIXACL;
730 	else
731 		fc->sb_flags &= ~SB_POSIXACL;
732 
733 	sbi->opt = new_sbi->opt;
734 
735 	fc->sb_flags |= SB_RDONLY;
736 	return 0;
737 }
738 
erofs_release_device_info(int id,void * ptr,void * data)739 static int erofs_release_device_info(int id, void *ptr, void *data)
740 {
741 	struct erofs_device_info *dif = ptr;
742 
743 	fs_put_dax(dif->dax_dev, NULL);
744 	if (dif->file)
745 		fput(dif->file);
746 	erofs_fscache_unregister_cookie(dif->fscache);
747 	dif->fscache = NULL;
748 	kfree(dif->path);
749 	kfree(dif);
750 	return 0;
751 }
752 
erofs_free_dev_context(struct erofs_dev_context * devs)753 static void erofs_free_dev_context(struct erofs_dev_context *devs)
754 {
755 	if (!devs)
756 		return;
757 	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
758 	idr_destroy(&devs->tree);
759 	kfree(devs);
760 }
761 
erofs_sb_free(struct erofs_sb_info * sbi)762 static void erofs_sb_free(struct erofs_sb_info *sbi)
763 {
764 	erofs_free_dev_context(sbi->devs);
765 	kfree(sbi->fsid);
766 	kfree(sbi->domain_id);
767 	if (sbi->dif0.file)
768 		fput(sbi->dif0.file);
769 	kfree(sbi);
770 }
771 
erofs_fc_free(struct fs_context * fc)772 static void erofs_fc_free(struct fs_context *fc)
773 {
774 	struct erofs_sb_info *sbi = fc->s_fs_info;
775 
776 	if (sbi) /* free here if an error occurs before transferring to sb */
777 		erofs_sb_free(sbi);
778 }
779 
780 static const struct fs_context_operations erofs_context_ops = {
781 	.parse_param	= erofs_fc_parse_param,
782 	.get_tree       = erofs_fc_get_tree,
783 	.reconfigure    = erofs_fc_reconfigure,
784 	.free		= erofs_fc_free,
785 };
786 
erofs_init_fs_context(struct fs_context * fc)787 static int erofs_init_fs_context(struct fs_context *fc)
788 {
789 	struct erofs_sb_info *sbi;
790 
791 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
792 	if (!sbi)
793 		return -ENOMEM;
794 
795 	sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
796 	if (!sbi->devs) {
797 		kfree(sbi);
798 		return -ENOMEM;
799 	}
800 	fc->s_fs_info = sbi;
801 
802 	idr_init(&sbi->devs->tree);
803 	init_rwsem(&sbi->devs->rwsem);
804 	erofs_default_options(sbi);
805 	fc->ops = &erofs_context_ops;
806 	return 0;
807 }
808 
erofs_kill_sb(struct super_block * sb)809 static void erofs_kill_sb(struct super_block *sb)
810 {
811 	struct erofs_sb_info *sbi = EROFS_SB(sb);
812 
813 	if ((IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid) ||
814 	    sbi->dif0.file)
815 		kill_anon_super(sb);
816 	else
817 		kill_block_super(sb);
818 	fs_put_dax(sbi->dif0.dax_dev, NULL);
819 	erofs_fscache_unregister_fs(sb);
820 	erofs_sb_free(sbi);
821 	sb->s_fs_info = NULL;
822 }
823 
erofs_put_super(struct super_block * sb)824 static void erofs_put_super(struct super_block *sb)
825 {
826 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
827 
828 	DBG_BUGON(!sbi);
829 
830 	erofs_unregister_sysfs(sb);
831 	erofs_shrinker_unregister(sb);
832 	erofs_xattr_prefixes_cleanup(sb);
833 #ifdef CONFIG_EROFS_FS_ZIP
834 	iput(sbi->managed_cache);
835 	sbi->managed_cache = NULL;
836 #endif
837 	iput(sbi->packed_inode);
838 	sbi->packed_inode = NULL;
839 	erofs_free_dev_context(sbi->devs);
840 	sbi->devs = NULL;
841 	erofs_fscache_unregister_fs(sb);
842 }
843 
844 static struct file_system_type erofs_fs_type = {
845 	.owner          = THIS_MODULE,
846 	.name           = "erofs",
847 	.init_fs_context = erofs_init_fs_context,
848 	.kill_sb        = erofs_kill_sb,
849 	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
850 };
851 MODULE_ALIAS_FS("erofs");
852 
erofs_module_init(void)853 static int __init erofs_module_init(void)
854 {
855 	int err;
856 
857 	erofs_check_ondisk_layout_definitions();
858 
859 	erofs_inode_cachep = kmem_cache_create("erofs_inode",
860 			sizeof(struct erofs_inode), 0,
861 			SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
862 			erofs_inode_init_once);
863 	if (!erofs_inode_cachep)
864 		return -ENOMEM;
865 
866 	err = erofs_init_shrinker();
867 	if (err)
868 		goto shrinker_err;
869 
870 	err = z_erofs_init_subsystem();
871 	if (err)
872 		goto zip_err;
873 
874 	err = erofs_init_sysfs();
875 	if (err)
876 		goto sysfs_err;
877 
878 	err = register_filesystem(&erofs_fs_type);
879 	if (err)
880 		goto fs_err;
881 
882 	return 0;
883 
884 fs_err:
885 	erofs_exit_sysfs();
886 sysfs_err:
887 	z_erofs_exit_subsystem();
888 zip_err:
889 	erofs_exit_shrinker();
890 shrinker_err:
891 	kmem_cache_destroy(erofs_inode_cachep);
892 	return err;
893 }
894 
erofs_module_exit(void)895 static void __exit erofs_module_exit(void)
896 {
897 	unregister_filesystem(&erofs_fs_type);
898 
899 	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
900 	rcu_barrier();
901 
902 	erofs_exit_sysfs();
903 	z_erofs_exit_subsystem();
904 	erofs_exit_shrinker();
905 	kmem_cache_destroy(erofs_inode_cachep);
906 }
907 
erofs_statfs(struct dentry * dentry,struct kstatfs * buf)908 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
909 {
910 	struct super_block *sb = dentry->d_sb;
911 	struct erofs_sb_info *sbi = EROFS_SB(sb);
912 
913 	buf->f_type = sb->s_magic;
914 	buf->f_bsize = sb->s_blocksize;
915 	buf->f_blocks = sbi->total_blocks;
916 	buf->f_bfree = buf->f_bavail = 0;
917 	buf->f_files = ULLONG_MAX;
918 	buf->f_ffree = ULLONG_MAX - sbi->inos;
919 	buf->f_namelen = EROFS_NAME_LEN;
920 
921 	if (uuid_is_null(&sb->s_uuid))
922 		buf->f_fsid = u64_to_fsid(!sb->s_bdev ? 0 :
923 				huge_encode_dev(sb->s_bdev->bd_dev));
924 	else
925 		buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
926 	return 0;
927 }
928 
erofs_show_options(struct seq_file * seq,struct dentry * root)929 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
930 {
931 	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
932 	struct erofs_mount_opts *opt = &sbi->opt;
933 
934 	if (IS_ENABLED(CONFIG_EROFS_FS_XATTR))
935 		seq_puts(seq, test_opt(opt, XATTR_USER) ?
936 				",user_xattr" : ",nouser_xattr");
937 	if (IS_ENABLED(CONFIG_EROFS_FS_POSIX_ACL))
938 		seq_puts(seq, test_opt(opt, POSIX_ACL) ? ",acl" : ",noacl");
939 	if (IS_ENABLED(CONFIG_EROFS_FS_ZIP))
940 		seq_printf(seq, ",cache_strategy=%s",
941 			  erofs_param_cache_strategy[opt->cache_strategy].name);
942 	if (test_opt(opt, DAX_ALWAYS))
943 		seq_puts(seq, ",dax=always");
944 	if (test_opt(opt, DAX_NEVER))
945 		seq_puts(seq, ",dax=never");
946 	if (erofs_is_fileio_mode(sbi) && test_opt(opt, DIRECT_IO))
947 		seq_puts(seq, ",directio");
948 #ifdef CONFIG_EROFS_FS_ONDEMAND
949 	if (sbi->fsid)
950 		seq_printf(seq, ",fsid=%s", sbi->fsid);
951 	if (sbi->domain_id)
952 		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
953 #endif
954 	return 0;
955 }
956 
957 const struct super_operations erofs_sops = {
958 	.put_super = erofs_put_super,
959 	.alloc_inode = erofs_alloc_inode,
960 	.free_inode = erofs_free_inode,
961 	.statfs = erofs_statfs,
962 	.show_options = erofs_show_options,
963 };
964 
965 module_init(erofs_module_init);
966 module_exit(erofs_module_exit);
967 
968 MODULE_DESCRIPTION("Enhanced ROM File System");
969 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
970 MODULE_LICENSE("GPL");
971