expfs.c 14.1 KB
Newer Older
1 2 3 4 5 6 7 8
/*
 * Copyright (C) Neil Brown 2002
 * Copyright (C) Christoph Hellwig 2007
 *
 * This file contains the code mapping from inodes to NFS file handles,
 * and for mapping back from file handles to dentries.
 *
 * For details on why we do all the strange and hairy things in here
9
 * take a look at Documentation/filesystems/nfs/Exporting.
10
 */
11
#include <linux/exportfs.h>
Linus Torvalds's avatar
Linus Torvalds committed
12 13 14
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/module.h>
15
#include <linux/mount.h>
Linus Torvalds's avatar
Linus Torvalds committed
16
#include <linux/namei.h>
17
#include <linux/sched.h>
18
#include <linux/cred.h>
Linus Torvalds's avatar
Linus Torvalds committed
19

20
#define dprintk(fmt, args...) do{}while(0)
Linus Torvalds's avatar
Linus Torvalds committed
21 22


23
static int get_name(const struct path *path, char *name, struct dentry *child);
24 25


26 27
static int exportfs_get_name(struct vfsmount *mnt, struct dentry *dir,
		char *name, struct dentry *child)
28
{
29
	const struct export_operations *nop = dir->d_sb->s_export_op;
30
	struct path path = {.mnt = mnt, .dentry = dir};
31 32 33 34

	if (nop->get_name)
		return nop->get_name(dir, name, child);
	else
35
		return get_name(&path, name, child);
36
}
Linus Torvalds's avatar
Linus Torvalds committed
37

38 39 40
/*
 * Check if the dentry or any of it's aliases is acceptable.
 */
41 42 43 44 45 46
static struct dentry *
find_acceptable_alias(struct dentry *result,
		int (*acceptable)(void *context, struct dentry *dentry),
		void *context)
{
	struct dentry *dentry, *toput = NULL;
47
	struct inode *inode;
48

49 50 51
	if (acceptable(context, result))
		return result;

52 53
	inode = result->d_inode;
	spin_lock(&inode->i_lock);
54
	hlist_for_each_entry(dentry, &inode->i_dentry, d_u.d_alias) {
55
		dget(dentry);
56
		spin_unlock(&inode->i_lock);
57 58 59 60 61 62
		if (toput)
			dput(toput);
		if (dentry != result && acceptable(context, dentry)) {
			dput(result);
			return dentry;
		}
63
		spin_lock(&inode->i_lock);
64 65
		toput = dentry;
	}
66
	spin_unlock(&inode->i_lock);
67 68 69 70 71 72

	if (toput)
		dput(toput);
	return NULL;
}

73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89
static bool dentry_connected(struct dentry *dentry)
{
	dget(dentry);
	while (dentry->d_flags & DCACHE_DISCONNECTED) {
		struct dentry *parent = dget_parent(dentry);

		dput(dentry);
		if (IS_ROOT(dentry)) {
			dput(parent);
			return false;
		}
		dentry = parent;
	}
	dput(dentry);
	return true;
}

90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107
static void clear_disconnected(struct dentry *dentry)
{
	dget(dentry);
	while (dentry->d_flags & DCACHE_DISCONNECTED) {
		struct dentry *parent = dget_parent(dentry);

		WARN_ON_ONCE(IS_ROOT(dentry));

		spin_lock(&dentry->d_lock);
		dentry->d_flags &= ~DCACHE_DISCONNECTED;
		spin_unlock(&dentry->d_lock);

		dput(dentry);
		dentry = parent;
	}
	dput(dentry);
}

108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127
/*
 * Reconnect a directory dentry with its parent.
 *
 * This can return a dentry, or NULL, or an error.
 *
 * In the first case the returned dentry is the parent of the given
 * dentry, and may itself need to be reconnected to its parent.
 *
 * In the NULL case, a concurrent VFS operation has either renamed or
 * removed this directory.  The concurrent operation has reconnected our
 * dentry, so we no longer need to.
 */
static struct dentry *reconnect_one(struct vfsmount *mnt,
		struct dentry *dentry, char *nbuf)
{
	struct dentry *parent;
	struct dentry *tmp;
	int err;

	parent = ERR_PTR(-EACCES);
Al Viro's avatar
Al Viro committed
128
	inode_lock(dentry->d_inode);
129 130
	if (mnt->mnt_sb->s_export_op->get_parent)
		parent = mnt->mnt_sb->s_export_op->get_parent(dentry);
Al Viro's avatar
Al Viro committed
131
	inode_unlock(dentry->d_inode);
132 133 134 135 136 137 138 139 140 141 142 143 144 145 146

	if (IS_ERR(parent)) {
		dprintk("%s: get_parent of %ld failed, err %d\n",
			__func__, dentry->d_inode->i_ino, PTR_ERR(parent));
		return parent;
	}

	dprintk("%s: find name of %lu in %lu\n", __func__,
		dentry->d_inode->i_ino, parent->d_inode->i_ino);
	err = exportfs_get_name(mnt, parent, nbuf, dentry);
	if (err == -ENOENT)
		goto out_reconnected;
	if (err)
		goto out_err;
	dprintk("%s: found name: %s\n", __func__, nbuf);
147
	tmp = lookup_one_len_unlocked(nbuf, parent, strlen(nbuf));
148 149 150 151 152
	if (IS_ERR(tmp)) {
		dprintk("%s: lookup failed: %d\n", __func__, PTR_ERR(tmp));
		goto out_err;
	}
	if (tmp != dentry) {
153 154 155 156 157 158
		/*
		 * Somebody has renamed it since exportfs_get_name();
		 * great, since it could've only been renamed if it
		 * got looked up and thus connected, and it would
		 * remain connected afterwards.  We are done.
		 */
159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
		dput(tmp);
		goto out_reconnected;
	}
	dput(tmp);
	if (IS_ROOT(dentry)) {
		err = -ESTALE;
		goto out_err;
	}
	return parent;

out_err:
	dput(parent);
	return ERR_PTR(err);
out_reconnected:
	dput(parent);
	/*
	 * Someone must have renamed our entry into another parent, in
	 * which case it has been reconnected by the rename.
	 *
	 * Or someone removed it entirely, in which case filehandle
	 * lookup will succeed but the directory is now IS_DEAD and
	 * subsequent operations on it will fail.
	 *
	 * Alternatively, maybe there was no race at all, and the
	 * filesystem is just corrupt and gave us a parent that doesn't
	 * actually contain any entry pointing to this inode.  So,
	 * double check that this worked and return -ESTALE if not:
	 */
	if (!dentry_connected(dentry))
		return ERR_PTR(-ESTALE);
	return NULL;
}

192 193
/*
 * Make sure target_dir is fully connected to the dentry tree.
Linus Torvalds's avatar
Linus Torvalds committed
194
 *
195 196 197 198 199 200 201 202 203 204 205 206 207
 * On successful return, DCACHE_DISCONNECTED will be cleared on
 * target_dir, and target_dir->d_parent->...->d_parent will reach the
 * root of the filesystem.
 *
 * Whenever DCACHE_DISCONNECTED is unset, target_dir is fully connected.
 * But the converse is not true: target_dir may have DCACHE_DISCONNECTED
 * set but already be connected.  In that case we'll verify the
 * connection to root and then clear the flag.
 *
 * Note that target_dir could be removed by a concurrent operation.  In
 * that case reconnect_path may still succeed with target_dir fully
 * connected, but further operations using the filehandle will fail when
 * necessary (due to S_DEAD being set on the directory).
Linus Torvalds's avatar
Linus Torvalds committed
208
 */
209
static int
210
reconnect_path(struct vfsmount *mnt, struct dentry *target_dir, char *nbuf)
Linus Torvalds's avatar
Linus Torvalds committed
211
{
212
	struct dentry *dentry, *parent;
Linus Torvalds's avatar
Linus Torvalds committed
213

214
	dentry = dget(target_dir);
Linus Torvalds's avatar
Linus Torvalds committed
215

216
	while (dentry->d_flags & DCACHE_DISCONNECTED) {
217
		BUG_ON(dentry == mnt->mnt_sb->s_root);
218

219 220 221 222 223 224
		if (IS_ROOT(dentry))
			parent = reconnect_one(mnt, dentry, nbuf);
		else
			parent = dget_parent(dentry);

		if (!parent)
225
			break;
226
		dput(dentry);
227 228 229
		if (IS_ERR(parent))
			return PTR_ERR(parent);
		dentry = parent;
Linus Torvalds's avatar
Linus Torvalds committed
230
	}
231
	dput(dentry);
232
	clear_disconnected(target_dir);
233 234 235
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
236
struct getdents_callback {
237
	struct dir_context ctx;
Linus Torvalds's avatar
Linus Torvalds committed
238 239
	char *name;		/* name that was found. It already points to a
				   buffer NAME_MAX+1 is size */
240
	u64 ino;		/* the inum we are looking for */
Linus Torvalds's avatar
Linus Torvalds committed
241 242 243 244 245 246 247 248
	int found;		/* inode matched? */
	int sequence;		/* sequence counter */
};

/*
 * A rather strange filldir function to capture
 * the name matching the specified inode number.
 */
249
static int filldir_one(struct dir_context *ctx, const char *name, int len,
250
			loff_t pos, u64 ino, unsigned int d_type)
Linus Torvalds's avatar
Linus Torvalds committed
251
{
252 253
	struct getdents_callback *buf =
		container_of(ctx, struct getdents_callback, ctx);
Linus Torvalds's avatar
Linus Torvalds committed
254 255 256
	int result = 0;

	buf->sequence++;
257
	if (buf->ino == ino && len <= NAME_MAX) {
Linus Torvalds's avatar
Linus Torvalds committed
258 259 260 261 262 263 264 265 266 267
		memcpy(buf->name, name, len);
		buf->name[len] = '\0';
		buf->found = 1;
		result = -1;
	}
	return result;
}

/**
 * get_name - default export_operations->get_name function
268
 * @path:   the directory in which to find a name
Linus Torvalds's avatar
Linus Torvalds committed
269 270 271 272 273 274
 * @name:   a pointer to a %NAME_MAX+1 char buffer to store the name
 * @child:  the dentry for the child directory.
 *
 * calls readdir on the parent until it finds an entry with
 * the same inode number as the child, and returns that.
 */
275
static int get_name(const struct path *path, char *name, struct dentry *child)
Linus Torvalds's avatar
Linus Torvalds committed
276
{
277
	const struct cred *cred = current_cred();
278
	struct inode *dir = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
279 280
	int error;
	struct file *file;
281 282 283 284 285
	struct kstat stat;
	struct path child_path = {
		.mnt = path->mnt,
		.dentry = child,
	};
Al Viro's avatar
Al Viro committed
286 287 288 289
	struct getdents_callback buffer = {
		.ctx.actor = filldir_one,
		.name = name,
	};
Linus Torvalds's avatar
Linus Torvalds committed
290 291 292 293 294 295 296

	error = -ENOTDIR;
	if (!dir || !S_ISDIR(dir->i_mode))
		goto out;
	error = -EINVAL;
	if (!dir->i_fop)
		goto out;
297 298 299 300 301 302
	/*
	 * inode->i_ino is unsigned long, kstat->ino is u64, so the
	 * former would be insufficient on 32-bit hosts when the
	 * filesystem supports 64-bit inode numbers.  So we need to
	 * actually call ->getattr, not just read i_ino:
	 */
303 304
	error = vfs_getattr_nosec(&child_path, &stat,
				  STATX_INO, AT_STATX_SYNC_AS_STAT);
305 306 307
	if (error)
		return error;
	buffer.ino = stat.ino;
Linus Torvalds's avatar
Linus Torvalds committed
308 309 310
	/*
	 * Open the directory ...
	 */
311
	file = dentry_open(path, O_RDONLY, cred);
Linus Torvalds's avatar
Linus Torvalds committed
312 313 314 315 316
	error = PTR_ERR(file);
	if (IS_ERR(file))
		goto out;

	error = -EINVAL;
317
	if (!file->f_op->iterate && !file->f_op->iterate_shared)
Linus Torvalds's avatar
Linus Torvalds committed
318 319 320 321 322 323
		goto out_close;

	buffer.sequence = 0;
	while (1) {
		int old_seq = buffer.sequence;

324
		error = iterate_dir(file, &buffer.ctx);
325 326 327 328
		if (buffer.found) {
			error = 0;
			break;
		}
Linus Torvalds's avatar
Linus Torvalds committed
329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345

		if (error < 0)
			break;

		error = -ENOENT;
		if (old_seq == buffer.sequence)
			break;
	}

out_close:
	fput(file);
out:
	return error;
}

/**
 * export_encode_fh - default export_operations->encode_fh function
Al Viro's avatar
Al Viro committed
346
 * @inode:   the object to encode
347
 * @fid:     where to store the file handle fragment
Linus Torvalds's avatar
Linus Torvalds committed
348
 * @max_len: maximum length to store there
Al Viro's avatar
Al Viro committed
349
 * @parent:  parent directory inode, if wanted
Linus Torvalds's avatar
Linus Torvalds committed
350 351 352 353 354 355
 *
 * This default encode_fh function assumes that the 32 inode number
 * is suitable for locating an inode, and that the generation number
 * can be used to check that it is still valid.  It places them in the
 * filehandle fragment where export_decode_fh expects to find them.
 */
Al Viro's avatar
Al Viro committed
356 357
static int export_encode_fh(struct inode *inode, struct fid *fid,
		int *max_len, struct inode *parent)
Linus Torvalds's avatar
Linus Torvalds committed
358 359
{
	int len = *max_len;
Christoph Hellwig's avatar
Christoph Hellwig committed
360
	int type = FILEID_INO32_GEN;
361

Al Viro's avatar
Al Viro committed
362
	if (parent && (len < 4)) {
363
		*max_len = 4;
364
		return FILEID_INVALID;
365 366
	} else if (len < 2) {
		*max_len = 2;
367
		return FILEID_INVALID;
368
	}
Linus Torvalds's avatar
Linus Torvalds committed
369 370

	len = 2;
Christoph Hellwig's avatar
Christoph Hellwig committed
371 372
	fid->i32.ino = inode->i_ino;
	fid->i32.gen = inode->i_generation;
Al Viro's avatar
Al Viro committed
373
	if (parent) {
Christoph Hellwig's avatar
Christoph Hellwig committed
374 375
		fid->i32.parent_ino = parent->i_ino;
		fid->i32.parent_gen = parent->i_generation;
Linus Torvalds's avatar
Linus Torvalds committed
376
		len = 4;
Christoph Hellwig's avatar
Christoph Hellwig committed
377
		type = FILEID_INO32_GEN_PARENT;
Linus Torvalds's avatar
Linus Torvalds committed
378 379 380 381 382
	}
	*max_len = len;
	return type;
}

383 384 385 386 387 388 389 390 391 392 393 394
int exportfs_encode_inode_fh(struct inode *inode, struct fid *fid,
			     int *max_len, struct inode *parent)
{
	const struct export_operations *nop = inode->i_sb->s_export_op;

	if (nop && nop->encode_fh)
		return nop->encode_fh(inode, fid->raw, max_len, parent);

	return export_encode_fh(inode, fid, max_len, parent);
}
EXPORT_SYMBOL_GPL(exportfs_encode_inode_fh);

Christoph Hellwig's avatar
Christoph Hellwig committed
395
int exportfs_encode_fh(struct dentry *dentry, struct fid *fid, int *max_len,
396 397
		int connectable)
{
398
	int error;
Al Viro's avatar
Al Viro committed
399 400
	struct dentry *p = NULL;
	struct inode *inode = dentry->d_inode, *parent = NULL;
401

Al Viro's avatar
Al Viro committed
402 403 404 405 406 407 408 409
	if (connectable && !S_ISDIR(inode->i_mode)) {
		p = dget_parent(dentry);
		/*
		 * note that while p might've ceased to be our parent already,
		 * it's still pinned by and still positive.
		 */
		parent = p->d_inode;
	}
410 411

	error = exportfs_encode_inode_fh(inode, fid, max_len, parent);
Al Viro's avatar
Al Viro committed
412
	dput(p);
413 414

	return error;
415 416 417
}
EXPORT_SYMBOL_GPL(exportfs_encode_fh);

Christoph Hellwig's avatar
Christoph Hellwig committed
418 419 420
struct dentry *exportfs_decode_fh(struct vfsmount *mnt, struct fid *fid,
		int fh_len, int fileid_type,
		int (*acceptable)(void *, struct dentry *), void *context)
421
{
422
	const struct export_operations *nop = mnt->mnt_sb->s_export_op;
423
	struct dentry *result, *alias;
424
	char nbuf[NAME_MAX+1];
425
	int err;
426

427 428 429
	/*
	 * Try to get any dentry for the given file handle from the filesystem.
	 */
430 431
	if (!nop || !nop->fh_to_dentry)
		return ERR_PTR(-ESTALE);
432
	result = nop->fh_to_dentry(mnt->mnt_sb, fid, fh_len, fileid_type);
433 434 435 436
	if (PTR_ERR(result) == -ENOMEM)
		return ERR_CAST(result);
	if (IS_ERR_OR_NULL(result))
		return ERR_PTR(-ESTALE);
437

438 439 440 441 442 443 444 445 446
	/*
	 * If no acceptance criteria was specified by caller, a disconnected
	 * dentry is also accepatable. Callers may use this mode to query if
	 * file handle is stale or to get a reference to an inode without
	 * risking the high overhead caused by directory reconnect.
	 */
	if (!acceptable)
		return result;

447
	if (d_is_dir(result)) {
448 449 450 451 452 453 454 455 456
		/*
		 * This request is for a directory.
		 *
		 * On the positive side there is only one dentry for each
		 * directory inode.  On the negative side this implies that we
		 * to ensure our dentry is connected all the way up to the
		 * filesystem root.
		 */
		if (result->d_flags & DCACHE_DISCONNECTED) {
457
			err = reconnect_path(mnt, result, nbuf);
458 459 460 461 462 463 464 465 466 467
			if (err)
				goto err_result;
		}

		if (!acceptable(context, result)) {
			err = -EACCES;
			goto err_result;
		}

		return result;
468
	} else {
469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496
		/*
		 * It's not a directory.  Life is a little more complicated.
		 */
		struct dentry *target_dir, *nresult;

		/*
		 * See if either the dentry we just got from the filesystem
		 * or any alias for it is acceptable.  This is always true
		 * if this filesystem is exported without the subtreecheck
		 * option.  If the filesystem is exported with the subtree
		 * check option there's a fair chance we need to look at
		 * the parent directory in the file handle and make sure
		 * it's connected to the filesystem root.
		 */
		alias = find_acceptable_alias(result, acceptable, context);
		if (alias)
			return alias;

		/*
		 * Try to extract a dentry for the parent directory from the
		 * file handle.  If this fails we'll have to give up.
		 */
		err = -ESTALE;
		if (!nop->fh_to_parent)
			goto err_result;

		target_dir = nop->fh_to_parent(mnt->mnt_sb, fid,
				fh_len, fileid_type);
497 498
		if (!target_dir)
			goto err_result;
499 500 501 502 503 504 505 506 507
		err = PTR_ERR(target_dir);
		if (IS_ERR(target_dir))
			goto err_result;

		/*
		 * And as usual we need to make sure the parent directory is
		 * connected to the filesystem root.  The VFS really doesn't
		 * like disconnected directories..
		 */
508
		err = reconnect_path(mnt, target_dir, nbuf);
509 510 511 512 513 514 515 516 517 518
		if (err) {
			dput(target_dir);
			goto err_result;
		}

		/*
		 * Now that we've got both a well-connected parent and a
		 * dentry for the inode we're after, make sure that our
		 * inode is actually connected to the parent.
		 */
519
		err = exportfs_get_name(mnt, target_dir, nbuf, result);
520
		if (!err) {
Al Viro's avatar
Al Viro committed
521
			inode_lock(target_dir->d_inode);
522 523
			nresult = lookup_one_len(nbuf, target_dir,
						 strlen(nbuf));
Al Viro's avatar
Al Viro committed
524
			inode_unlock(target_dir->d_inode);
525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550
			if (!IS_ERR(nresult)) {
				if (nresult->d_inode) {
					dput(result);
					result = nresult;
				} else
					dput(nresult);
			}
		}

		/*
		 * At this point we are done with the parent, but it's pinned
		 * by the child dentry anyway.
		 */
		dput(target_dir);

		/*
		 * And finally make sure the dentry is actually acceptable
		 * to NFSD.
		 */
		alias = find_acceptable_alias(result, acceptable, context);
		if (!alias) {
			err = -EACCES;
			goto err_result;
		}

		return alias;
551 552
	}

553 554
 err_result:
	dput(result);
555 556
	if (err != -ENOMEM)
		err = -ESTALE;
557
	return ERR_PTR(err);
558 559 560
}
EXPORT_SYMBOL_GPL(exportfs_decode_fh);

Linus Torvalds's avatar
Linus Torvalds committed
561
MODULE_LICENSE("GPL");