~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/overlayfs/inode.c

Version: ~ [ linux-5.8-rc4 ] ~ [ linux-5.7.7 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.50 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.131 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.187 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.229 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.229 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *
  3  * Copyright (C) 2011 Novell Inc.
  4  *
  5  * This program is free software; you can redistribute it and/or modify it
  6  * under the terms of the GNU General Public License version 2 as published by
  7  * the Free Software Foundation.
  8  */
  9 
 10 #include <linux/fs.h>
 11 #include <linux/slab.h>
 12 #include <linux/cred.h>
 13 #include <linux/xattr.h>
 14 #include <linux/posix_acl.h>
 15 #include <linux/ratelimit.h>
 16 #include "overlayfs.h"
 17 
 18 
 19 int ovl_setattr(struct dentry *dentry, struct iattr *attr)
 20 {
 21         int err;
 22         struct dentry *upperdentry;
 23         const struct cred *old_cred;
 24 
 25         /*
 26          * Check for permissions before trying to copy-up.  This is redundant
 27          * since it will be rechecked later by ->setattr() on upper dentry.  But
 28          * without this, copy-up can be triggered by just about anybody.
 29          *
 30          * We don't initialize inode->size, which just means that
 31          * inode_newsize_ok() will always check against MAX_LFS_FILESIZE and not
 32          * check for a swapfile (which this won't be anyway).
 33          */
 34         err = setattr_prepare(dentry, attr);
 35         if (err)
 36                 return err;
 37 
 38         err = ovl_want_write(dentry);
 39         if (err)
 40                 goto out;
 41 
 42         err = ovl_copy_up(dentry);
 43         if (!err) {
 44                 upperdentry = ovl_dentry_upper(dentry);
 45 
 46                 if (attr->ia_valid & (ATTR_KILL_SUID|ATTR_KILL_SGID))
 47                         attr->ia_valid &= ~ATTR_MODE;
 48 
 49                 inode_lock(upperdentry->d_inode);
 50                 old_cred = ovl_override_creds(dentry->d_sb);
 51                 err = notify_change(upperdentry, attr, NULL);
 52                 revert_creds(old_cred);
 53                 if (!err)
 54                         ovl_copyattr(upperdentry->d_inode, dentry->d_inode);
 55                 inode_unlock(upperdentry->d_inode);
 56         }
 57         ovl_drop_write(dentry);
 58 out:
 59         return err;
 60 }
 61 
 62 static int ovl_map_dev_ino(struct dentry *dentry, struct kstat *stat,
 63                            struct ovl_layer *lower_layer)
 64 {
 65         bool samefs = ovl_same_sb(dentry->d_sb);
 66         unsigned int xinobits = ovl_xino_bits(dentry->d_sb);
 67 
 68         if (samefs) {
 69                 /*
 70                  * When all layers are on the same fs, all real inode
 71                  * number are unique, so we use the overlay st_dev,
 72                  * which is friendly to du -x.
 73                  */
 74                 stat->dev = dentry->d_sb->s_dev;
 75                 return 0;
 76         } else if (xinobits) {
 77                 unsigned int shift = 64 - xinobits;
 78                 /*
 79                  * All inode numbers of underlying fs should not be using the
 80                  * high xinobits, so we use high xinobits to partition the
 81                  * overlay st_ino address space. The high bits holds the fsid
 82                  * (upper fsid is 0). This way overlay inode numbers are unique
 83                  * and all inodes use overlay st_dev. Inode numbers are also
 84                  * persistent for a given layer configuration.
 85                  */
 86                 if (stat->ino >> shift) {
 87                         pr_warn_ratelimited("overlayfs: inode number too big (%pd2, ino=%llu, xinobits=%d)\n",
 88                                             dentry, stat->ino, xinobits);
 89                 } else {
 90                         if (lower_layer)
 91                                 stat->ino |= ((u64)lower_layer->fsid) << shift;
 92 
 93                         stat->dev = dentry->d_sb->s_dev;
 94                         return 0;
 95                 }
 96         }
 97 
 98         /* The inode could not be mapped to a unified st_ino address space */
 99         if (S_ISDIR(dentry->d_inode->i_mode)) {
100                 /*
101                  * Always use the overlay st_dev for directories, so 'find
102                  * -xdev' will scan the entire overlay mount and won't cross the
103                  * overlay mount boundaries.
104                  *
105                  * If not all layers are on the same fs the pair {real st_ino;
106                  * overlay st_dev} is not unique, so use the non persistent
107                  * overlay st_ino for directories.
108                  */
109                 stat->dev = dentry->d_sb->s_dev;
110                 stat->ino = dentry->d_inode->i_ino;
111         } else if (lower_layer && lower_layer->fsid) {
112                 /*
113                  * For non-samefs setup, if we cannot map all layers st_ino
114                  * to a unified address space, we need to make sure that st_dev
115                  * is unique per lower fs. Upper layer uses real st_dev and
116                  * lower layers use the unique anonymous bdev assigned to the
117                  * lower fs.
118                  */
119                 stat->dev = lower_layer->fs->pseudo_dev;
120         }
121 
122         return 0;
123 }
124 
125 int ovl_getattr(const struct path *path, struct kstat *stat,
126                 u32 request_mask, unsigned int flags)
127 {
128         struct dentry *dentry = path->dentry;
129         enum ovl_path_type type;
130         struct path realpath;
131         const struct cred *old_cred;
132         bool is_dir = S_ISDIR(dentry->d_inode->i_mode);
133         bool samefs = ovl_same_sb(dentry->d_sb);
134         struct ovl_layer *lower_layer = NULL;
135         int err;
136 
137         type = ovl_path_real(dentry, &realpath);
138         old_cred = ovl_override_creds(dentry->d_sb);
139         err = vfs_getattr(&realpath, stat, request_mask, flags);
140         if (err)
141                 goto out;
142 
143         /*
144          * For non-dir or same fs, we use st_ino of the copy up origin.
145          * This guaranties constant st_dev/st_ino across copy up.
146          * With xino feature and non-samefs, we use st_ino of the copy up
147          * origin masked with high bits that represent the layer id.
148          *
149          * If lower filesystem supports NFS file handles, this also guaranties
150          * persistent st_ino across mount cycle.
151          */
152         if (!is_dir || samefs || ovl_xino_bits(dentry->d_sb)) {
153                 if (!OVL_TYPE_UPPER(type)) {
154                         lower_layer = ovl_layer_lower(dentry);
155                 } else if (OVL_TYPE_ORIGIN(type)) {
156                         struct kstat lowerstat;
157                         u32 lowermask = STATX_INO | (!is_dir ? STATX_NLINK : 0);
158 
159                         ovl_path_lower(dentry, &realpath);
160                         err = vfs_getattr(&realpath, &lowerstat,
161                                           lowermask, flags);
162                         if (err)
163                                 goto out;
164 
165                         /*
166                          * Lower hardlinks may be broken on copy up to different
167                          * upper files, so we cannot use the lower origin st_ino
168                          * for those different files, even for the same fs case.
169                          *
170                          * Similarly, several redirected dirs can point to the
171                          * same dir on a lower layer. With the "verify_lower"
172                          * feature, we do not use the lower origin st_ino, if
173                          * we haven't verified that this redirect is unique.
174                          *
175                          * With inodes index enabled, it is safe to use st_ino
176                          * of an indexed origin. The index validates that the
177                          * upper hardlink is not broken and that a redirected
178                          * dir is the only redirect to that origin.
179                          */
180                         if (ovl_test_flag(OVL_INDEX, d_inode(dentry)) ||
181                             (!ovl_verify_lower(dentry->d_sb) &&
182                              (is_dir || lowerstat.nlink == 1))) {
183                                 stat->ino = lowerstat.ino;
184                                 lower_layer = ovl_layer_lower(dentry);
185                         }
186                 }
187         }
188 
189         err = ovl_map_dev_ino(dentry, stat, lower_layer);
190         if (err)
191                 goto out;
192 
193         /*
194          * It's probably not worth it to count subdirs to get the
195          * correct link count.  nlink=1 seems to pacify 'find' and
196          * other utilities.
197          */
198         if (is_dir && OVL_TYPE_MERGE(type))
199                 stat->nlink = 1;
200 
201         /*
202          * Return the overlay inode nlinks for indexed upper inodes.
203          * Overlay inode nlink counts the union of the upper hardlinks
204          * and non-covered lower hardlinks. It does not include the upper
205          * index hardlink.
206          */
207         if (!is_dir && ovl_test_flag(OVL_INDEX, d_inode(dentry)))
208                 stat->nlink = dentry->d_inode->i_nlink;
209 
210 out:
211         revert_creds(old_cred);
212 
213         return err;
214 }
215 
216 int ovl_permission(struct inode *inode, int mask)
217 {
218         struct inode *upperinode = ovl_inode_upper(inode);
219         struct inode *realinode = upperinode ?: ovl_inode_lower(inode);
220         const struct cred *old_cred;
221         int err;
222 
223         /* Careful in RCU walk mode */
224         if (!realinode) {
225                 WARN_ON(!(mask & MAY_NOT_BLOCK));
226                 return -ECHILD;
227         }
228 
229         /*
230          * Check overlay inode with the creds of task and underlying inode
231          * with creds of mounter
232          */
233         err = generic_permission(inode, mask);
234         if (err)
235                 return err;
236 
237         old_cred = ovl_override_creds(inode->i_sb);
238         if (!upperinode &&
239             !special_file(realinode->i_mode) && mask & MAY_WRITE) {
240                 mask &= ~(MAY_WRITE | MAY_APPEND);
241                 /* Make sure mounter can read file for copy up later */
242                 mask |= MAY_READ;
243         }
244         err = inode_permission(realinode, mask);
245         revert_creds(old_cred);
246 
247         return err;
248 }
249 
250 static const char *ovl_get_link(struct dentry *dentry,
251                                 struct inode *inode,
252                                 struct delayed_call *done)
253 {
254         const struct cred *old_cred;
255         const char *p;
256 
257         if (!dentry)
258                 return ERR_PTR(-ECHILD);
259 
260         old_cred = ovl_override_creds(dentry->d_sb);
261         p = vfs_get_link(ovl_dentry_real(dentry), done);
262         revert_creds(old_cred);
263         return p;
264 }
265 
266 bool ovl_is_private_xattr(const char *name)
267 {
268         return strncmp(name, OVL_XATTR_PREFIX,
269                        sizeof(OVL_XATTR_PREFIX) - 1) == 0;
270 }
271 
272 int ovl_xattr_set(struct dentry *dentry, struct inode *inode, const char *name,
273                   const void *value, size_t size, int flags)
274 {
275         int err;
276         struct dentry *upperdentry = ovl_i_dentry_upper(inode);
277         struct dentry *realdentry = upperdentry ?: ovl_dentry_lower(dentry);
278         const struct cred *old_cred;
279 
280         err = ovl_want_write(dentry);
281         if (err)
282                 goto out;
283 
284         if (!value && !upperdentry) {
285                 err = vfs_getxattr(realdentry, name, NULL, 0);
286                 if (err < 0)
287                         goto out_drop_write;
288         }
289 
290         if (!upperdentry) {
291                 err = ovl_copy_up(dentry);
292                 if (err)
293                         goto out_drop_write;
294 
295                 realdentry = ovl_dentry_upper(dentry);
296         }
297 
298         old_cred = ovl_override_creds(dentry->d_sb);
299         if (value)
300                 err = vfs_setxattr(realdentry, name, value, size, flags);
301         else {
302                 WARN_ON(flags != XATTR_REPLACE);
303                 err = vfs_removexattr(realdentry, name);
304         }
305         revert_creds(old_cred);
306 
307 out_drop_write:
308         ovl_drop_write(dentry);
309 out:
310         return err;
311 }
312 
313 int ovl_xattr_get(struct dentry *dentry, struct inode *inode, const char *name,
314                   void *value, size_t size)
315 {
316         ssize_t res;
317         const struct cred *old_cred;
318         struct dentry *realdentry =
319                 ovl_i_dentry_upper(inode) ?: ovl_dentry_lower(dentry);
320 
321         old_cred = ovl_override_creds(dentry->d_sb);
322         res = vfs_getxattr(realdentry, name, value, size);
323         revert_creds(old_cred);
324         return res;
325 }
326 
327 static bool ovl_can_list(const char *s)
328 {
329         /* List all non-trusted xatts */
330         if (strncmp(s, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) != 0)
331                 return true;
332 
333         /* Never list trusted.overlay, list other trusted for superuser only */
334         return !ovl_is_private_xattr(s) && capable(CAP_SYS_ADMIN);
335 }
336 
337 ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size)
338 {
339         struct dentry *realdentry = ovl_dentry_real(dentry);
340         ssize_t res;
341         size_t len;
342         char *s;
343         const struct cred *old_cred;
344 
345         old_cred = ovl_override_creds(dentry->d_sb);
346         res = vfs_listxattr(realdentry, list, size);
347         revert_creds(old_cred);
348         if (res <= 0 || size == 0)
349                 return res;
350 
351         /* filter out private xattrs */
352         for (s = list, len = res; len;) {
353                 size_t slen = strnlen(s, len) + 1;
354 
355                 /* underlying fs providing us with an broken xattr list? */
356                 if (WARN_ON(slen > len))
357                         return -EIO;
358 
359                 len -= slen;
360                 if (!ovl_can_list(s)) {
361                         res -= slen;
362                         memmove(s, s + slen, len);
363                 } else {
364                         s += slen;
365                 }
366         }
367 
368         return res;
369 }
370 
371 struct posix_acl *ovl_get_acl(struct inode *inode, int type)
372 {
373         struct inode *realinode = ovl_inode_real(inode);
374         const struct cred *old_cred;
375         struct posix_acl *acl;
376 
377         if (!IS_ENABLED(CONFIG_FS_POSIX_ACL) || !IS_POSIXACL(realinode))
378                 return NULL;
379 
380         old_cred = ovl_override_creds(inode->i_sb);
381         acl = get_acl(realinode, type);
382         revert_creds(old_cred);
383 
384         return acl;
385 }
386 
387 static bool ovl_open_need_copy_up(struct dentry *dentry, int flags)
388 {
389         /* Copy up of disconnected dentry does not set upper alias */
390         if (ovl_dentry_upper(dentry) &&
391             (ovl_dentry_has_upper_alias(dentry) ||
392              (dentry->d_flags & DCACHE_DISCONNECTED)))
393                 return false;
394 
395         if (special_file(d_inode(dentry)->i_mode))
396                 return false;
397 
398         if (!(OPEN_FMODE(flags) & FMODE_WRITE) && !(flags & O_TRUNC))
399                 return false;
400 
401         return true;
402 }
403 
404 int ovl_open_maybe_copy_up(struct dentry *dentry, unsigned int file_flags)
405 {
406         int err = 0;
407 
408         if (ovl_open_need_copy_up(dentry, file_flags)) {
409                 err = ovl_want_write(dentry);
410                 if (!err) {
411                         err = ovl_copy_up_flags(dentry, file_flags);
412                         ovl_drop_write(dentry);
413                 }
414         }
415 
416         return err;
417 }
418 
419 int ovl_update_time(struct inode *inode, struct timespec64 *ts, int flags)
420 {
421         if (flags & S_ATIME) {
422                 struct ovl_fs *ofs = inode->i_sb->s_fs_info;
423                 struct path upperpath = {
424                         .mnt = ofs->upper_mnt,
425                         .dentry = ovl_upperdentry_dereference(OVL_I(inode)),
426                 };
427 
428                 if (upperpath.dentry) {
429                         touch_atime(&upperpath);
430                         inode->i_atime = d_inode(upperpath.dentry)->i_atime;
431                 }
432         }
433         return 0;
434 }
435 
436 static const struct inode_operations ovl_file_inode_operations = {
437         .setattr        = ovl_setattr,
438         .permission     = ovl_permission,
439         .getattr        = ovl_getattr,
440         .listxattr      = ovl_listxattr,
441         .get_acl        = ovl_get_acl,
442         .update_time    = ovl_update_time,
443 };
444 
445 static const struct inode_operations ovl_symlink_inode_operations = {
446         .setattr        = ovl_setattr,
447         .get_link       = ovl_get_link,
448         .getattr        = ovl_getattr,
449         .listxattr      = ovl_listxattr,
450         .update_time    = ovl_update_time,
451 };
452 
453 /*
454  * It is possible to stack overlayfs instance on top of another
455  * overlayfs instance as lower layer. We need to annonate the
456  * stackable i_mutex locks according to stack level of the super
457  * block instance. An overlayfs instance can never be in stack
458  * depth 0 (there is always a real fs below it).  An overlayfs
459  * inode lock will use the lockdep annotaion ovl_i_mutex_key[depth].
460  *
461  * For example, here is a snip from /proc/lockdep_chains after
462  * dir_iterate of nested overlayfs:
463  *
464  * [...] &ovl_i_mutex_dir_key[depth]   (stack_depth=2)
465  * [...] &ovl_i_mutex_dir_key[depth]#2 (stack_depth=1)
466  * [...] &type->i_mutex_dir_key        (stack_depth=0)
467  */
468 #define OVL_MAX_NESTING FILESYSTEM_MAX_STACK_DEPTH
469 
470 static inline void ovl_lockdep_annotate_inode_mutex_key(struct inode *inode)
471 {
472 #ifdef CONFIG_LOCKDEP
473         static struct lock_class_key ovl_i_mutex_key[OVL_MAX_NESTING];
474         static struct lock_class_key ovl_i_mutex_dir_key[OVL_MAX_NESTING];
475         static struct lock_class_key ovl_i_lock_key[OVL_MAX_NESTING];
476 
477         int depth = inode->i_sb->s_stack_depth - 1;
478 
479         if (WARN_ON_ONCE(depth < 0 || depth >= OVL_MAX_NESTING))
480                 depth = 0;
481 
482         if (S_ISDIR(inode->i_mode))
483                 lockdep_set_class(&inode->i_rwsem, &ovl_i_mutex_dir_key[depth]);
484         else
485                 lockdep_set_class(&inode->i_rwsem, &ovl_i_mutex_key[depth]);
486 
487         lockdep_set_class(&OVL_I(inode)->lock, &ovl_i_lock_key[depth]);
488 #endif
489 }
490 
491 static void ovl_fill_inode(struct inode *inode, umode_t mode, dev_t rdev,
492                            unsigned long ino, int fsid)
493 {
494         int xinobits = ovl_xino_bits(inode->i_sb);
495 
496         /*
497          * When NFS export is enabled and d_ino is consistent with st_ino
498          * (samefs or i_ino has enough bits to encode layer), set the same
499          * value used for d_ino to i_ino, because nfsd readdirplus compares
500          * d_ino values to i_ino values of child entries. When called from
501          * ovl_new_inode(), ino arg is 0, so i_ino will be updated to real
502          * upper inode i_ino on ovl_inode_init() or ovl_inode_update().
503          */
504         if (inode->i_sb->s_export_op &&
505             (ovl_same_sb(inode->i_sb) || xinobits)) {
506                 inode->i_ino = ino;
507                 if (xinobits && fsid && !(ino >> (64 - xinobits)))
508                         inode->i_ino |= (unsigned long)fsid << (64 - xinobits);
509         } else {
510                 inode->i_ino = get_next_ino();
511         }
512         inode->i_mode = mode;
513         inode->i_flags |= S_NOCMTIME;
514 #ifdef CONFIG_FS_POSIX_ACL
515         inode->i_acl = inode->i_default_acl = ACL_DONT_CACHE;
516 #endif
517 
518         ovl_lockdep_annotate_inode_mutex_key(inode);
519 
520         switch (mode & S_IFMT) {
521         case S_IFREG:
522                 inode->i_op = &ovl_file_inode_operations;
523                 break;
524 
525         case S_IFDIR:
526                 inode->i_op = &ovl_dir_inode_operations;
527                 inode->i_fop = &ovl_dir_operations;
528                 break;
529 
530         case S_IFLNK:
531                 inode->i_op = &ovl_symlink_inode_operations;
532                 break;
533 
534         default:
535                 inode->i_op = &ovl_file_inode_operations;
536                 init_special_inode(inode, mode, rdev);
537                 break;
538         }
539 }
540 
541 /*
542  * With inodes index enabled, an overlay inode nlink counts the union of upper
543  * hardlinks and non-covered lower hardlinks. During the lifetime of a non-pure
544  * upper inode, the following nlink modifying operations can happen:
545  *
546  * 1. Lower hardlink copy up
547  * 2. Upper hardlink created, unlinked or renamed over
548  * 3. Lower hardlink whiteout or renamed over
549  *
550  * For the first, copy up case, the union nlink does not change, whether the
551  * operation succeeds or fails, but the upper inode nlink may change.
552  * Therefore, before copy up, we store the union nlink value relative to the
553  * lower inode nlink in the index inode xattr trusted.overlay.nlink.
554  *
555  * For the second, upper hardlink case, the union nlink should be incremented
556  * or decremented IFF the operation succeeds, aligned with nlink change of the
557  * upper inode. Therefore, before link/unlink/rename, we store the union nlink
558  * value relative to the upper inode nlink in the index inode.
559  *
560  * For the last, lower cover up case, we simplify things by preceding the
561  * whiteout or cover up with copy up. This makes sure that there is an index
562  * upper inode where the nlink xattr can be stored before the copied up upper
563  * entry is unlink.
564  */
565 #define OVL_NLINK_ADD_UPPER     (1 << 0)
566 
567 /*
568  * On-disk format for indexed nlink:
569  *
570  * nlink relative to the upper inode - "U[+-]NUM"
571  * nlink relative to the lower inode - "L[+-]NUM"
572  */
573 
574 static int ovl_set_nlink_common(struct dentry *dentry,
575                                 struct dentry *realdentry, const char *format)
576 {
577         struct inode *inode = d_inode(dentry);
578         struct inode *realinode = d_inode(realdentry);
579         char buf[13];
580         int len;
581 
582         len = snprintf(buf, sizeof(buf), format,
583                        (int) (inode->i_nlink - realinode->i_nlink));
584 
585         if (WARN_ON(len >= sizeof(buf)))
586                 return -EIO;
587 
588         return ovl_do_setxattr(ovl_dentry_upper(dentry),
589                                OVL_XATTR_NLINK, buf, len, 0);
590 }
591 
592 int ovl_set_nlink_upper(struct dentry *dentry)
593 {
594         return ovl_set_nlink_common(dentry, ovl_dentry_upper(dentry), "U%+i");
595 }
596 
597 int ovl_set_nlink_lower(struct dentry *dentry)
598 {
599         return ovl_set_nlink_common(dentry, ovl_dentry_lower(dentry), "L%+i");
600 }
601 
602 unsigned int ovl_get_nlink(struct dentry *lowerdentry,
603                            struct dentry *upperdentry,
604                            unsigned int fallback)
605 {
606         int nlink_diff;
607         int nlink;
608         char buf[13];
609         int err;
610 
611         if (!lowerdentry || !upperdentry || d_inode(lowerdentry)->i_nlink == 1)
612                 return fallback;
613 
614         err = vfs_getxattr(upperdentry, OVL_XATTR_NLINK, &buf, sizeof(buf) - 1);
615         if (err < 0)
616                 goto fail;
617 
618         buf[err] = '\0';
619         if ((buf[0] != 'L' && buf[0] != 'U') ||
620             (buf[1] != '+' && buf[1] != '-'))
621                 goto fail;
622 
623         err = kstrtoint(buf + 1, 10, &nlink_diff);
624         if (err < 0)
625                 goto fail;
626 
627         nlink = d_inode(buf[0] == 'L' ? lowerdentry : upperdentry)->i_nlink;
628         nlink += nlink_diff;
629 
630         if (nlink <= 0)
631                 goto fail;
632 
633         return nlink;
634 
635 fail:
636         pr_warn_ratelimited("overlayfs: failed to get index nlink (%pd2, err=%i)\n",
637                             upperdentry, err);
638         return fallback;
639 }
640 
641 struct inode *ovl_new_inode(struct super_block *sb, umode_t mode, dev_t rdev)
642 {
643         struct inode *inode;
644 
645         inode = new_inode(sb);
646         if (inode)
647                 ovl_fill_inode(inode, mode, rdev, 0, 0);
648 
649         return inode;
650 }
651 
652 static int ovl_inode_test(struct inode *inode, void *data)
653 {
654         return inode->i_private == data;
655 }
656 
657 static int ovl_inode_set(struct inode *inode, void *data)
658 {
659         inode->i_private = data;
660         return 0;
661 }
662 
663 static bool ovl_verify_inode(struct inode *inode, struct dentry *lowerdentry,
664                              struct dentry *upperdentry, bool strict)
665 {
666         /*
667          * For directories, @strict verify from lookup path performs consistency
668          * checks, so NULL lower/upper in dentry must match NULL lower/upper in
669          * inode. Non @strict verify from NFS handle decode path passes NULL for
670          * 'unknown' lower/upper.
671          */
672         if (S_ISDIR(inode->i_mode) && strict) {
673                 /* Real lower dir moved to upper layer under us? */
674                 if (!lowerdentry && ovl_inode_lower(inode))
675                         return false;
676 
677                 /* Lookup of an uncovered redirect origin? */
678                 if (!upperdentry && ovl_inode_upper(inode))
679                         return false;
680         }
681 
682         /*
683          * Allow non-NULL lower inode in ovl_inode even if lowerdentry is NULL.
684          * This happens when finding a copied up overlay inode for a renamed
685          * or hardlinked overlay dentry and lower dentry cannot be followed
686          * by origin because lower fs does not support file handles.
687          */
688         if (lowerdentry && ovl_inode_lower(inode) != d_inode(lowerdentry))
689                 return false;
690 
691         /*
692          * Allow non-NULL __upperdentry in inode even if upperdentry is NULL.
693          * This happens when finding a lower alias for a copied up hard link.
694          */
695         if (upperdentry && ovl_inode_upper(inode) != d_inode(upperdentry))
696                 return false;
697 
698         return true;
699 }
700 
701 struct inode *ovl_lookup_inode(struct super_block *sb, struct dentry *real,
702                                bool is_upper)
703 {
704         struct inode *inode, *key = d_inode(real);
705 
706         inode = ilookup5(sb, (unsigned long) key, ovl_inode_test, key);
707         if (!inode)
708                 return NULL;
709 
710         if (!ovl_verify_inode(inode, is_upper ? NULL : real,
711                               is_upper ? real : NULL, false)) {
712                 iput(inode);
713                 return ERR_PTR(-ESTALE);
714         }
715 
716         return inode;
717 }
718 
719 /*
720  * Does overlay inode need to be hashed by lower inode?
721  */
722 static bool ovl_hash_bylower(struct super_block *sb, struct dentry *upper,
723                              struct dentry *lower, struct dentry *index)
724 {
725         struct ovl_fs *ofs = sb->s_fs_info;
726 
727         /* No, if pure upper */
728         if (!lower)
729                 return false;
730 
731         /* Yes, if already indexed */
732         if (index)
733                 return true;
734 
735         /* Yes, if won't be copied up */
736         if (!ofs->upper_mnt)
737                 return true;
738 
739         /* No, if lower hardlink is or will be broken on copy up */
740         if ((upper || !ovl_indexdir(sb)) &&
741             !d_is_dir(lower) && d_inode(lower)->i_nlink > 1)
742                 return false;
743 
744         /* No, if non-indexed upper with NFS export */
745         if (sb->s_export_op && upper)
746                 return false;
747 
748         /* Otherwise, hash by lower inode for fsnotify */
749         return true;
750 }
751 
752 static struct inode *ovl_iget5(struct super_block *sb, struct inode *newinode,
753                                struct inode *key)
754 {
755         return newinode ? inode_insert5(newinode, (unsigned long) key,
756                                          ovl_inode_test, ovl_inode_set, key) :
757                           iget5_locked(sb, (unsigned long) key,
758                                        ovl_inode_test, ovl_inode_set, key);
759 }
760 
761 struct inode *ovl_get_inode(struct super_block *sb,
762                             struct ovl_inode_params *oip)
763 {
764         struct dentry *upperdentry = oip->upperdentry;
765         struct ovl_path *lowerpath = oip->lowerpath;
766         struct inode *realinode = upperdentry ? d_inode(upperdentry) : NULL;
767         struct inode *inode;
768         struct dentry *lowerdentry = lowerpath ? lowerpath->dentry : NULL;
769         bool bylower = ovl_hash_bylower(sb, upperdentry, lowerdentry,
770                                         oip->index);
771         int fsid = bylower ? oip->lowerpath->layer->fsid : 0;
772         bool is_dir;
773         unsigned long ino = 0;
774 
775         if (!realinode)
776                 realinode = d_inode(lowerdentry);
777 
778         /*
779          * Copy up origin (lower) may exist for non-indexed upper, but we must
780          * not use lower as hash key if this is a broken hardlink.
781          */
782         is_dir = S_ISDIR(realinode->i_mode);
783         if (upperdentry || bylower) {
784                 struct inode *key = d_inode(bylower ? lowerdentry :
785                                                       upperdentry);
786                 unsigned int nlink = is_dir ? 1 : realinode->i_nlink;
787 
788                 inode = ovl_iget5(sb, oip->newinode, key);
789                 if (!inode)
790                         goto out_nomem;
791                 if (!(inode->i_state & I_NEW)) {
792                         /*
793                          * Verify that the underlying files stored in the inode
794                          * match those in the dentry.
795                          */
796                         if (!ovl_verify_inode(inode, lowerdentry, upperdentry,
797                                               true)) {
798                                 iput(inode);
799                                 inode = ERR_PTR(-ESTALE);
800                                 goto out;
801                         }
802 
803                         dput(upperdentry);
804                         goto out;
805                 }
806 
807                 /* Recalculate nlink for non-dir due to indexing */
808                 if (!is_dir)
809                         nlink = ovl_get_nlink(lowerdentry, upperdentry, nlink);
810                 set_nlink(inode, nlink);
811                 ino = key->i_ino;
812         } else {
813                 /* Lower hardlink that will be broken on copy up */
814                 inode = new_inode(sb);
815                 if (!inode)
816                         goto out_nomem;
817         }
818         ovl_fill_inode(inode, realinode->i_mode, realinode->i_rdev, ino, fsid);
819         ovl_inode_init(inode, upperdentry, lowerdentry);
820 
821         if (upperdentry && ovl_is_impuredir(upperdentry))
822                 ovl_set_flag(OVL_IMPURE, inode);
823 
824         if (oip->index)
825                 ovl_set_flag(OVL_INDEX, inode);
826 
827         /* Check for non-merge dir that may have whiteouts */
828         if (is_dir) {
829                 if (((upperdentry && lowerdentry) || oip->numlower > 1) ||
830                     ovl_check_origin_xattr(upperdentry ?: lowerdentry)) {
831                         ovl_set_flag(OVL_WHITEOUTS, inode);
832                 }
833         }
834 
835         if (inode->i_state & I_NEW)
836                 unlock_new_inode(inode);
837 out:
838         return inode;
839 
840 out_nomem:
841         inode = ERR_PTR(-ENOMEM);
842         goto out;
843 }
844 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp