xref: /linux/fs/afs/inode.c (revision 55f3538c4923e9dfca132e99ebec370e8094afda)
1 /*
2  * Copyright (c) 2002 Red Hat, Inc. All rights reserved.
3  *
4  * This software may be freely redistributed under the terms of the
5  * GNU General Public License.
6  *
7  * You should have received a copy of the GNU General Public License
8  * along with this program; if not, write to the Free Software
9  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
10  *
11  * Authors: David Woodhouse <dwmw2@infradead.org>
12  *          David Howells <dhowells@redhat.com>
13  *
14  */
15 
16 #include <linux/kernel.h>
17 #include <linux/module.h>
18 #include <linux/init.h>
19 #include <linux/fs.h>
20 #include <linux/pagemap.h>
21 #include <linux/sched.h>
22 #include <linux/mount.h>
23 #include <linux/namei.h>
24 #include <linux/iversion.h>
25 #include "internal.h"
26 
27 static const struct inode_operations afs_symlink_inode_operations = {
28 	.get_link	= page_get_link,
29 	.listxattr	= afs_listxattr,
30 };
31 
32 /*
33  * map the AFS file status to the inode member variables
34  */
35 static int afs_inode_map_status(struct afs_vnode *vnode, struct key *key)
36 {
37 	struct inode *inode = AFS_VNODE_TO_I(vnode);
38 	bool changed;
39 
40 	_debug("FS: ft=%d lk=%d sz=%llu ver=%Lu mod=%hu",
41 	       vnode->status.type,
42 	       vnode->status.nlink,
43 	       (unsigned long long) vnode->status.size,
44 	       vnode->status.data_version,
45 	       vnode->status.mode);
46 
47 	read_seqlock_excl(&vnode->cb_lock);
48 
49 	switch (vnode->status.type) {
50 	case AFS_FTYPE_FILE:
51 		inode->i_mode	= S_IFREG | vnode->status.mode;
52 		inode->i_op	= &afs_file_inode_operations;
53 		inode->i_fop	= &afs_file_operations;
54 		break;
55 	case AFS_FTYPE_DIR:
56 		inode->i_mode	= S_IFDIR | vnode->status.mode;
57 		inode->i_op	= &afs_dir_inode_operations;
58 		inode->i_fop	= &afs_dir_file_operations;
59 		break;
60 	case AFS_FTYPE_SYMLINK:
61 		/* Symlinks with a mode of 0644 are actually mountpoints. */
62 		if ((vnode->status.mode & 0777) == 0644) {
63 			inode->i_flags |= S_AUTOMOUNT;
64 
65 			set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags);
66 
67 			inode->i_mode	= S_IFDIR | 0555;
68 			inode->i_op	= &afs_mntpt_inode_operations;
69 			inode->i_fop	= &afs_mntpt_file_operations;
70 		} else {
71 			inode->i_mode	= S_IFLNK | vnode->status.mode;
72 			inode->i_op	= &afs_symlink_inode_operations;
73 		}
74 		inode_nohighmem(inode);
75 		break;
76 	default:
77 		printk("kAFS: AFS vnode with undefined type\n");
78 		read_sequnlock_excl(&vnode->cb_lock);
79 		return -EBADMSG;
80 	}
81 
82 	changed = (vnode->status.size != inode->i_size);
83 
84 	set_nlink(inode, vnode->status.nlink);
85 	inode->i_uid		= vnode->status.owner;
86 	inode->i_gid            = vnode->status.group;
87 	inode->i_size		= vnode->status.size;
88 	inode->i_ctime.tv_sec	= vnode->status.mtime_client;
89 	inode->i_ctime.tv_nsec	= 0;
90 	inode->i_atime		= inode->i_mtime = inode->i_ctime;
91 	inode->i_blocks		= 0;
92 	inode->i_generation	= vnode->fid.unique;
93 	inode_set_iversion_raw(inode, vnode->status.data_version);
94 	inode->i_mapping->a_ops	= &afs_fs_aops;
95 
96 	read_sequnlock_excl(&vnode->cb_lock);
97 
98 #ifdef CONFIG_AFS_FSCACHE
99 	if (changed)
100 		fscache_attr_changed(vnode->cache);
101 #endif
102 	return 0;
103 }
104 
105 /*
106  * Fetch file status from the volume.
107  */
108 int afs_fetch_status(struct afs_vnode *vnode, struct key *key)
109 {
110 	struct afs_fs_cursor fc;
111 	int ret;
112 
113 	_enter("%s,{%x:%u.%u,S=%lx}",
114 	       vnode->volume->name,
115 	       vnode->fid.vid, vnode->fid.vnode, vnode->fid.unique,
116 	       vnode->flags);
117 
118 	ret = -ERESTARTSYS;
119 	if (afs_begin_vnode_operation(&fc, vnode, key)) {
120 		while (afs_select_fileserver(&fc)) {
121 			fc.cb_break = vnode->cb_break + vnode->cb_s_break;
122 			afs_fs_fetch_file_status(&fc, NULL);
123 		}
124 
125 		afs_check_for_remote_deletion(&fc, fc.vnode);
126 		afs_vnode_commit_status(&fc, vnode, fc.cb_break);
127 		ret = afs_end_vnode_operation(&fc);
128 	}
129 
130 	_leave(" = %d", ret);
131 	return ret;
132 }
133 
134 /*
135  * iget5() comparator
136  */
137 int afs_iget5_test(struct inode *inode, void *opaque)
138 {
139 	struct afs_iget_data *data = opaque;
140 
141 	return inode->i_ino == data->fid.vnode &&
142 		inode->i_generation == data->fid.unique;
143 }
144 
145 /*
146  * iget5() comparator for inode created by autocell operations
147  *
148  * These pseudo inodes don't match anything.
149  */
150 static int afs_iget5_autocell_test(struct inode *inode, void *opaque)
151 {
152 	return 0;
153 }
154 
155 /*
156  * iget5() inode initialiser
157  */
158 static int afs_iget5_set(struct inode *inode, void *opaque)
159 {
160 	struct afs_iget_data *data = opaque;
161 	struct afs_vnode *vnode = AFS_FS_I(inode);
162 
163 	inode->i_ino = data->fid.vnode;
164 	inode->i_generation = data->fid.unique;
165 	vnode->fid = data->fid;
166 	vnode->volume = data->volume;
167 
168 	return 0;
169 }
170 
171 /*
172  * inode retrieval for autocell
173  */
174 struct inode *afs_iget_autocell(struct inode *dir, const char *dev_name,
175 				int namesz, struct key *key)
176 {
177 	struct afs_iget_data data;
178 	struct afs_super_info *as;
179 	struct afs_vnode *vnode;
180 	struct super_block *sb;
181 	struct inode *inode;
182 	static atomic_t afs_autocell_ino;
183 
184 	_enter("{%x:%u},%*.*s,",
185 	       AFS_FS_I(dir)->fid.vid, AFS_FS_I(dir)->fid.vnode,
186 	       namesz, namesz, dev_name ?: "");
187 
188 	sb = dir->i_sb;
189 	as = sb->s_fs_info;
190 	data.volume = as->volume;
191 	data.fid.vid = as->volume->vid;
192 	data.fid.unique = 0;
193 	data.fid.vnode = 0;
194 
195 	inode = iget5_locked(sb, atomic_inc_return(&afs_autocell_ino),
196 			     afs_iget5_autocell_test, afs_iget5_set,
197 			     &data);
198 	if (!inode) {
199 		_leave(" = -ENOMEM");
200 		return ERR_PTR(-ENOMEM);
201 	}
202 
203 	_debug("GOT INODE %p { ino=%lu, vl=%x, vn=%x, u=%x }",
204 	       inode, inode->i_ino, data.fid.vid, data.fid.vnode,
205 	       data.fid.unique);
206 
207 	vnode = AFS_FS_I(inode);
208 
209 	/* there shouldn't be an existing inode */
210 	BUG_ON(!(inode->i_state & I_NEW));
211 
212 	inode->i_size		= 0;
213 	inode->i_mode		= S_IFDIR | S_IRUGO | S_IXUGO;
214 	inode->i_op		= &afs_autocell_inode_operations;
215 	set_nlink(inode, 2);
216 	inode->i_uid		= GLOBAL_ROOT_UID;
217 	inode->i_gid		= GLOBAL_ROOT_GID;
218 	inode->i_ctime.tv_sec	= get_seconds();
219 	inode->i_ctime.tv_nsec	= 0;
220 	inode->i_atime		= inode->i_mtime = inode->i_ctime;
221 	inode->i_blocks		= 0;
222 	inode_set_iversion_raw(inode, 0);
223 	inode->i_generation	= 0;
224 
225 	set_bit(AFS_VNODE_PSEUDODIR, &vnode->flags);
226 	set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags);
227 	inode->i_flags |= S_AUTOMOUNT | S_NOATIME;
228 	unlock_new_inode(inode);
229 	_leave(" = %p", inode);
230 	return inode;
231 }
232 
233 /*
234  * inode retrieval
235  */
236 struct inode *afs_iget(struct super_block *sb, struct key *key,
237 		       struct afs_fid *fid, struct afs_file_status *status,
238 		       struct afs_callback *cb, struct afs_cb_interest *cbi)
239 {
240 	struct afs_iget_data data = { .fid = *fid };
241 	struct afs_super_info *as;
242 	struct afs_vnode *vnode;
243 	struct inode *inode;
244 	int ret;
245 
246 	_enter(",{%x:%u.%u},,", fid->vid, fid->vnode, fid->unique);
247 
248 	as = sb->s_fs_info;
249 	data.volume = as->volume;
250 
251 	inode = iget5_locked(sb, fid->vnode, afs_iget5_test, afs_iget5_set,
252 			     &data);
253 	if (!inode) {
254 		_leave(" = -ENOMEM");
255 		return ERR_PTR(-ENOMEM);
256 	}
257 
258 	_debug("GOT INODE %p { vl=%x vn=%x, u=%x }",
259 	       inode, fid->vid, fid->vnode, fid->unique);
260 
261 	vnode = AFS_FS_I(inode);
262 
263 	/* deal with an existing inode */
264 	if (!(inode->i_state & I_NEW)) {
265 		_leave(" = %p", inode);
266 		return inode;
267 	}
268 
269 	if (!status) {
270 		/* it's a remotely extant inode */
271 		ret = afs_fetch_status(vnode, key);
272 		if (ret < 0)
273 			goto bad_inode;
274 	} else {
275 		/* it's an inode we just created */
276 		memcpy(&vnode->status, status, sizeof(vnode->status));
277 
278 		if (!cb) {
279 			/* it's a symlink we just created (the fileserver
280 			 * didn't give us a callback) */
281 			vnode->cb_version = 0;
282 			vnode->cb_type = 0;
283 			vnode->cb_expires_at = 0;
284 		} else {
285 			vnode->cb_version = cb->version;
286 			vnode->cb_type = cb->type;
287 			vnode->cb_expires_at = cb->expiry;
288 			vnode->cb_interest = afs_get_cb_interest(cbi);
289 			set_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
290 		}
291 
292 		vnode->cb_expires_at += ktime_get_real_seconds();
293 	}
294 
295 	/* set up caching before mapping the status, as map-status reads the
296 	 * first page of symlinks to see if they're really mountpoints */
297 	inode->i_size = vnode->status.size;
298 #ifdef CONFIG_AFS_FSCACHE
299 	vnode->cache = fscache_acquire_cookie(vnode->volume->cache,
300 					      &afs_vnode_cache_index_def,
301 					      vnode, true);
302 #endif
303 
304 	ret = afs_inode_map_status(vnode, key);
305 	if (ret < 0)
306 		goto bad_inode;
307 
308 	/* success */
309 	clear_bit(AFS_VNODE_UNSET, &vnode->flags);
310 	inode->i_flags |= S_NOATIME;
311 	unlock_new_inode(inode);
312 	_leave(" = %p [CB { v=%u t=%u }]", inode, vnode->cb_version, vnode->cb_type);
313 	return inode;
314 
315 	/* failure */
316 bad_inode:
317 #ifdef CONFIG_AFS_FSCACHE
318 	fscache_relinquish_cookie(vnode->cache, 0);
319 	vnode->cache = NULL;
320 #endif
321 	iget_failed(inode);
322 	_leave(" = %d [bad]", ret);
323 	return ERR_PTR(ret);
324 }
325 
326 /*
327  * mark the data attached to an inode as obsolete due to a write on the server
328  * - might also want to ditch all the outstanding writes and dirty pages
329  */
330 void afs_zap_data(struct afs_vnode *vnode)
331 {
332 	_enter("{%x:%u}", vnode->fid.vid, vnode->fid.vnode);
333 
334 	/* nuke all the non-dirty pages that aren't locked, mapped or being
335 	 * written back in a regular file and completely discard the pages in a
336 	 * directory or symlink */
337 	if (S_ISREG(vnode->vfs_inode.i_mode))
338 		invalidate_remote_inode(&vnode->vfs_inode);
339 	else
340 		invalidate_inode_pages2(vnode->vfs_inode.i_mapping);
341 }
342 
343 /*
344  * validate a vnode/inode
345  * - there are several things we need to check
346  *   - parent dir data changes (rm, rmdir, rename, mkdir, create, link,
347  *     symlink)
348  *   - parent dir metadata changed (security changes)
349  *   - dentry data changed (write, truncate)
350  *   - dentry metadata changed (security changes)
351  */
352 int afs_validate(struct afs_vnode *vnode, struct key *key)
353 {
354 	time64_t now = ktime_get_real_seconds();
355 	bool valid = false;
356 	int ret;
357 
358 	_enter("{v={%x:%u} fl=%lx},%x",
359 	       vnode->fid.vid, vnode->fid.vnode, vnode->flags,
360 	       key_serial(key));
361 
362 	/* Quickly check the callback state.  Ideally, we'd use read_seqbegin
363 	 * here, but we have no way to pass the net namespace to the RCU
364 	 * cleanup for the server record.
365 	 */
366 	read_seqlock_excl(&vnode->cb_lock);
367 
368 	if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
369 		if (vnode->cb_s_break != vnode->cb_interest->server->cb_s_break) {
370 			vnode->cb_s_break = vnode->cb_interest->server->cb_s_break;
371 		} else if (!test_bit(AFS_VNODE_DIR_MODIFIED, &vnode->flags) &&
372 			   !test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags) &&
373 			   vnode->cb_expires_at - 10 > now) {
374 				valid = true;
375 		}
376 	} else if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
377 		valid = true;
378 	}
379 
380 	read_sequnlock_excl(&vnode->cb_lock);
381 
382 	if (test_bit(AFS_VNODE_DELETED, &vnode->flags))
383 		clear_nlink(&vnode->vfs_inode);
384 
385 	if (valid)
386 		goto valid;
387 
388 	mutex_lock(&vnode->validate_lock);
389 
390 	/* if the promise has expired, we need to check the server again to get
391 	 * a new promise - note that if the (parent) directory's metadata was
392 	 * changed then the security may be different and we may no longer have
393 	 * access */
394 	if (!test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
395 		_debug("not promised");
396 		ret = afs_fetch_status(vnode, key);
397 		if (ret < 0) {
398 			if (ret == -ENOENT) {
399 				set_bit(AFS_VNODE_DELETED, &vnode->flags);
400 				ret = -ESTALE;
401 			}
402 			goto error_unlock;
403 		}
404 		_debug("new promise [fl=%lx]", vnode->flags);
405 	}
406 
407 	if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
408 		_debug("file already deleted");
409 		ret = -ESTALE;
410 		goto error_unlock;
411 	}
412 
413 	/* if the vnode's data version number changed then its contents are
414 	 * different */
415 	if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
416 		afs_zap_data(vnode);
417 
418 	clear_bit(AFS_VNODE_DIR_MODIFIED, &vnode->flags);
419 	mutex_unlock(&vnode->validate_lock);
420 valid:
421 	_leave(" = 0");
422 	return 0;
423 
424 error_unlock:
425 	mutex_unlock(&vnode->validate_lock);
426 	_leave(" = %d", ret);
427 	return ret;
428 }
429 
430 /*
431  * read the attributes of an inode
432  */
433 int afs_getattr(const struct path *path, struct kstat *stat,
434 		u32 request_mask, unsigned int query_flags)
435 {
436 	struct inode *inode = d_inode(path->dentry);
437 	struct afs_vnode *vnode = AFS_FS_I(inode);
438 	int seq = 0;
439 
440 	_enter("{ ino=%lu v=%u }", inode->i_ino, inode->i_generation);
441 
442 	do {
443 		read_seqbegin_or_lock(&vnode->cb_lock, &seq);
444 		generic_fillattr(inode, stat);
445 	} while (need_seqretry(&vnode->cb_lock, seq));
446 
447 	done_seqretry(&vnode->cb_lock, seq);
448 	return 0;
449 }
450 
451 /*
452  * discard an AFS inode
453  */
454 int afs_drop_inode(struct inode *inode)
455 {
456 	_enter("");
457 
458 	if (test_bit(AFS_VNODE_PSEUDODIR, &AFS_FS_I(inode)->flags))
459 		return generic_delete_inode(inode);
460 	else
461 		return generic_drop_inode(inode);
462 }
463 
464 /*
465  * clear an AFS inode
466  */
467 void afs_evict_inode(struct inode *inode)
468 {
469 	struct afs_vnode *vnode;
470 
471 	vnode = AFS_FS_I(inode);
472 
473 	_enter("{%x:%u.%d}",
474 	       vnode->fid.vid,
475 	       vnode->fid.vnode,
476 	       vnode->fid.unique);
477 
478 	_debug("CLEAR INODE %p", inode);
479 
480 	ASSERTCMP(inode->i_ino, ==, vnode->fid.vnode);
481 
482 	truncate_inode_pages_final(&inode->i_data);
483 	clear_inode(inode);
484 
485 	if (vnode->cb_interest) {
486 		afs_put_cb_interest(afs_i2net(inode), vnode->cb_interest);
487 		vnode->cb_interest = NULL;
488 	}
489 
490 	while (!list_empty(&vnode->wb_keys)) {
491 		struct afs_wb_key *wbk = list_entry(vnode->wb_keys.next,
492 						    struct afs_wb_key, vnode_link);
493 		list_del(&wbk->vnode_link);
494 		afs_put_wb_key(wbk);
495 	}
496 
497 #ifdef CONFIG_AFS_FSCACHE
498 	fscache_relinquish_cookie(vnode->cache, 0);
499 	vnode->cache = NULL;
500 #endif
501 
502 	afs_put_permits(vnode->permit_cache);
503 	_leave("");
504 }
505 
506 /*
507  * set the attributes of an inode
508  */
509 int afs_setattr(struct dentry *dentry, struct iattr *attr)
510 {
511 	struct afs_fs_cursor fc;
512 	struct afs_vnode *vnode = AFS_FS_I(d_inode(dentry));
513 	struct key *key;
514 	int ret;
515 
516 	_enter("{%x:%u},{n=%pd},%x",
517 	       vnode->fid.vid, vnode->fid.vnode, dentry,
518 	       attr->ia_valid);
519 
520 	if (!(attr->ia_valid & (ATTR_SIZE | ATTR_MODE | ATTR_UID | ATTR_GID |
521 				ATTR_MTIME))) {
522 		_leave(" = 0 [unsupported]");
523 		return 0;
524 	}
525 
526 	/* flush any dirty data outstanding on a regular file */
527 	if (S_ISREG(vnode->vfs_inode.i_mode))
528 		filemap_write_and_wait(vnode->vfs_inode.i_mapping);
529 
530 	if (attr->ia_valid & ATTR_FILE) {
531 		key = afs_file_key(attr->ia_file);
532 	} else {
533 		key = afs_request_key(vnode->volume->cell);
534 		if (IS_ERR(key)) {
535 			ret = PTR_ERR(key);
536 			goto error;
537 		}
538 	}
539 
540 	ret = -ERESTARTSYS;
541 	if (afs_begin_vnode_operation(&fc, vnode, key)) {
542 		while (afs_select_fileserver(&fc)) {
543 			fc.cb_break = vnode->cb_break + vnode->cb_s_break;
544 			afs_fs_setattr(&fc, attr);
545 		}
546 
547 		afs_check_for_remote_deletion(&fc, fc.vnode);
548 		afs_vnode_commit_status(&fc, vnode, fc.cb_break);
549 		ret = afs_end_vnode_operation(&fc);
550 	}
551 
552 	if (!(attr->ia_valid & ATTR_FILE))
553 		key_put(key);
554 
555 error:
556 	_leave(" = %d", ret);
557 	return ret;
558 }
559