dir.c 53.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
/*
 *  linux/fs/nfs/dir.c
 *
 *  Copyright (C) 1992  Rick Sladkey
 *
 *  nfs directory handling functions
 *
 * 10 Apr 1996	Added silly rename for unlink	--okir
 * 28 Sep 1996	Improved directory cache --okir
 * 23 Aug 1997  Claus Heine claus@momo.math.rwth-aachen.de 
 *              Re-implemented silly rename for unlink, newly implemented
 *              silly rename for nfs_rename() following the suggestions
 *              of Olaf Kirch (okir) found in this file.
 *              Following Linus comments on my original hack, this version
 *              depends only on the dcache stuff and doesn't touch the inode
 *              layer (iput() and friends).
 *  6 Jun 1999	Cache readdir lookups in the page cache. -DaveM
 */

#include <linux/time.h>
#include <linux/errno.h>
#include <linux/stat.h>
#include <linux/fcntl.h>
#include <linux/string.h>
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/pagemap.h>
32
#include <linux/pagevec.h>
L
Linus Torvalds 已提交
33
#include <linux/namei.h>
34
#include <linux/mount.h>
A
Alexey Dobriyan 已提交
35
#include <linux/sched.h>
L
Linus Torvalds 已提交
36

37
#include "nfs4_fs.h"
L
Linus Torvalds 已提交
38
#include "delegation.h"
C
Chuck Lever 已提交
39
#include "iostat.h"
40
#include "internal.h"
L
Linus Torvalds 已提交
41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56

/* #define NFS_DEBUG_VERBOSE 1 */

static int nfs_opendir(struct inode *, struct file *);
static int nfs_readdir(struct file *, void *, filldir_t);
static struct dentry *nfs_lookup(struct inode *, struct dentry *, struct nameidata *);
static int nfs_create(struct inode *, struct dentry *, int, struct nameidata *);
static int nfs_mkdir(struct inode *, struct dentry *, int);
static int nfs_rmdir(struct inode *, struct dentry *);
static int nfs_unlink(struct inode *, struct dentry *);
static int nfs_symlink(struct inode *, struct dentry *, const char *);
static int nfs_link(struct dentry *, struct inode *, struct dentry *);
static int nfs_mknod(struct inode *, struct dentry *, int, dev_t);
static int nfs_rename(struct inode *, struct dentry *,
		      struct inode *, struct dentry *);
static int nfs_fsync_dir(struct file *, struct dentry *, int);
57
static loff_t nfs_llseek_dir(struct file *, loff_t, int);
L
Linus Torvalds 已提交
58

59
const struct file_operations nfs_dir_operations = {
60
	.llseek		= nfs_llseek_dir,
L
Linus Torvalds 已提交
61 62 63 64 65 66 67
	.read		= generic_read_dir,
	.readdir	= nfs_readdir,
	.open		= nfs_opendir,
	.release	= nfs_release,
	.fsync		= nfs_fsync_dir,
};

68
const struct inode_operations nfs_dir_inode_operations = {
L
Linus Torvalds 已提交
69 70 71 72 73 74 75 76 77 78 79 80 81 82
	.create		= nfs_create,
	.lookup		= nfs_lookup,
	.link		= nfs_link,
	.unlink		= nfs_unlink,
	.symlink	= nfs_symlink,
	.mkdir		= nfs_mkdir,
	.rmdir		= nfs_rmdir,
	.mknod		= nfs_mknod,
	.rename		= nfs_rename,
	.permission	= nfs_permission,
	.getattr	= nfs_getattr,
	.setattr	= nfs_setattr,
};

83
#ifdef CONFIG_NFS_V3
84
const struct inode_operations nfs3_dir_inode_operations = {
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
	.create		= nfs_create,
	.lookup		= nfs_lookup,
	.link		= nfs_link,
	.unlink		= nfs_unlink,
	.symlink	= nfs_symlink,
	.mkdir		= nfs_mkdir,
	.rmdir		= nfs_rmdir,
	.mknod		= nfs_mknod,
	.rename		= nfs_rename,
	.permission	= nfs_permission,
	.getattr	= nfs_getattr,
	.setattr	= nfs_setattr,
	.listxattr	= nfs3_listxattr,
	.getxattr	= nfs3_getxattr,
	.setxattr	= nfs3_setxattr,
	.removexattr	= nfs3_removexattr,
};
#endif  /* CONFIG_NFS_V3 */

L
Linus Torvalds 已提交
104 105 106
#ifdef CONFIG_NFS_V4

static struct dentry *nfs_atomic_lookup(struct inode *, struct dentry *, struct nameidata *);
107
const struct inode_operations nfs4_dir_inode_operations = {
L
Linus Torvalds 已提交
108 109 110 111 112 113 114 115 116 117 118 119
	.create		= nfs_create,
	.lookup		= nfs_atomic_lookup,
	.link		= nfs_link,
	.unlink		= nfs_unlink,
	.symlink	= nfs_symlink,
	.mkdir		= nfs_mkdir,
	.rmdir		= nfs_rmdir,
	.mknod		= nfs_mknod,
	.rename		= nfs_rename,
	.permission	= nfs_permission,
	.getattr	= nfs_getattr,
	.setattr	= nfs_setattr,
120 121 122
	.getxattr       = nfs4_getxattr,
	.setxattr       = nfs4_setxattr,
	.listxattr      = nfs4_listxattr,
L
Linus Torvalds 已提交
123 124 125 126 127 128 129 130 131 132
};

#endif /* CONFIG_NFS_V4 */

/*
 * Open file
 */
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
133
	int res;
L
Linus Torvalds 已提交
134

C
Chuck Lever 已提交
135
	dfprintk(FILE, "NFS: open dir(%s/%s)\n",
136 137 138 139
			filp->f_path.dentry->d_parent->d_name.name,
			filp->f_path.dentry->d_name.name);

	nfs_inc_stats(inode, NFSIOS_VFSOPEN);
C
Chuck Lever 已提交
140

L
Linus Torvalds 已提交
141
	/* Call generic open code in order to cache credentials */
142
	res = nfs_open(inode, filp);
L
Linus Torvalds 已提交
143 144 145
	return res;
}

146
typedef __be32 * (*decode_dirent_t)(__be32 *, struct nfs_entry *, int);
L
Linus Torvalds 已提交
147 148 149 150
typedef struct {
	struct file	*file;
	struct page	*page;
	unsigned long	page_index;
151
	__be32		*ptr;
152 153
	u64		*dir_cookie;
	loff_t		current_index;
L
Linus Torvalds 已提交
154 155 156
	struct nfs_entry *entry;
	decode_dirent_t	decode;
	int		plus;
157
	unsigned long	timestamp;
158
	unsigned long	gencount;
159
	int		timestamp_valid;
L
Linus Torvalds 已提交
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
} nfs_readdir_descriptor_t;

/* Now we cache directories properly, by stuffing the dirent
 * data directly in the page cache.
 *
 * Inode invalidation due to refresh etc. takes care of
 * _everything_, no sloppy entry flushing logic, no extraneous
 * copying, network direct to page cache, the way it was meant
 * to be.
 *
 * NOTE: Dirent information verification is done always by the
 *	 page-in of the RPC reply, nowhere else, this simplies
 *	 things substantially.
 */
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page *page)
{
	struct file	*file = desc->file;
178
	struct inode	*inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
179
	struct rpc_cred	*cred = nfs_file_cred(file);
180
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
181 182
	int		error;

C
Chuck Lever 已提交
183
	dfprintk(DIRCACHE, "NFS: %s: reading cookie %Lu into page %lu\n",
184
			__func__, (long long)desc->entry->cookie,
C
Chuck Lever 已提交
185
			page->index);
L
Linus Torvalds 已提交
186 187 188

 again:
	timestamp = jiffies;
189
	gencount = nfs_inc_attr_generation_counter();
190
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, desc->entry->cookie, page,
L
Linus Torvalds 已提交
191 192 193 194 195
					  NFS_SERVER(inode)->dtsize, desc->plus);
	if (error < 0) {
		/* We requested READDIRPLUS, but the server doesn't grok it */
		if (error == -ENOTSUPP && desc->plus) {
			NFS_SERVER(inode)->caps &= ~NFS_CAP_READDIRPLUS;
B
Benny Halevy 已提交
196
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
197 198 199 200 201
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
202
	desc->timestamp = timestamp;
203
	desc->gencount = gencount;
204
	desc->timestamp_valid = 1;
L
Linus Torvalds 已提交
205 206 207
	SetPageUptodate(page);
	/* Ensure consistent page alignment of the data.
	 * Note: assumes we have exclusive access to this mapping either
208
	 *	 through inode->i_mutex or some other mechanism.
L
Linus Torvalds 已提交
209
	 */
210
	if (invalidate_inode_pages2_range(inode->i_mapping, page->index + 1, -1) < 0) {
211 212 213
		/* Should never happen */
		nfs_zap_mapping(inode, inode->i_mapping);
	}
L
Linus Torvalds 已提交
214 215 216 217 218 219 220 221 222 223
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
	return -EIO;
}

static inline
int dir_decode(nfs_readdir_descriptor_t *desc)
{
224
	__be32	*p = desc->ptr;
L
Linus Torvalds 已提交
225 226 227 228
	p = desc->decode(p, desc->entry, desc->plus);
	if (IS_ERR(p))
		return PTR_ERR(p);
	desc->ptr = p;
229
	if (desc->timestamp_valid) {
230
		desc->entry->fattr->time_start = desc->timestamp;
231 232
		desc->entry->fattr->gencount = desc->gencount;
	} else
233
		desc->entry->fattr->valid &= ~NFS_ATTR_FATTR;
L
Linus Torvalds 已提交
234 235 236 237 238 239 240 241 242 243 244 245 246 247
	return 0;
}

static inline
void dir_page_release(nfs_readdir_descriptor_t *desc)
{
	kunmap(desc->page);
	page_cache_release(desc->page);
	desc->page = NULL;
	desc->ptr = NULL;
}

/*
 * Given a pointer to a buffer that has already been filled by a call
248
 * to readdir, find the next entry with cookie '*desc->dir_cookie'.
L
Linus Torvalds 已提交
249 250 251 252 253 254
 *
 * If the end of the buffer has been reached, return -EAGAIN, if not,
 * return the offset within the buffer of the next entry to be
 * read.
 */
static inline
255
int find_dirent(nfs_readdir_descriptor_t *desc)
L
Linus Torvalds 已提交
256 257 258 259 260 261
{
	struct nfs_entry *entry = desc->entry;
	int		loop_count = 0,
			status;

	while((status = dir_decode(desc)) == 0) {
C
Chuck Lever 已提交
262
		dfprintk(DIRCACHE, "NFS: %s: examining cookie %Lu\n",
263
				__func__, (unsigned long long)entry->cookie);
264
		if (entry->prev_cookie == *desc->dir_cookie)
L
Linus Torvalds 已提交
265 266 267 268 269 270 271 272 273 274
			break;
		if (loop_count++ > 200) {
			loop_count = 0;
			schedule();
		}
	}
	return status;
}

/*
275
 * Given a pointer to a buffer that has already been filled by a call
276
 * to readdir, find the entry at offset 'desc->file->f_pos'.
277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
 *
 * If the end of the buffer has been reached, return -EAGAIN, if not,
 * return the offset within the buffer of the next entry to be
 * read.
 */
static inline
int find_dirent_index(nfs_readdir_descriptor_t *desc)
{
	struct nfs_entry *entry = desc->entry;
	int		loop_count = 0,
			status;

	for(;;) {
		status = dir_decode(desc);
		if (status)
			break;

C
Chuck Lever 已提交
294 295
		dfprintk(DIRCACHE, "NFS: found cookie %Lu at index %Ld\n",
				(unsigned long long)entry->cookie, desc->current_index);
296

297 298
		if (desc->file->f_pos == desc->current_index) {
			*desc->dir_cookie = entry->cookie;
299 300 301 302 303 304 305 306 307 308 309 310 311 312
			break;
		}
		desc->current_index++;
		if (loop_count++ > 200) {
			loop_count = 0;
			schedule();
		}
	}
	return status;
}

/*
 * Find the given page, and call find_dirent() or find_dirent_index in
 * order to try to return the next entry.
L
Linus Torvalds 已提交
313 314 315 316
 */
static inline
int find_dirent_page(nfs_readdir_descriptor_t *desc)
{
317
	struct inode	*inode = desc->file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
318 319 320
	struct page	*page;
	int		status;

C
Chuck Lever 已提交
321
	dfprintk(DIRCACHE, "NFS: %s: searching page %ld for target %Lu\n",
322
			__func__, desc->page_index,
C
Chuck Lever 已提交
323
			(long long) *desc->dir_cookie);
L
Linus Torvalds 已提交
324

325 326 327 328
	/* If we find the page in the page_cache, we cannot be sure
	 * how fresh the data is, so we will ignore readdir_plus attributes.
	 */
	desc->timestamp_valid = 0;
L
Linus Torvalds 已提交
329 330 331 332 333 334 335 336 337 338
	page = read_cache_page(inode->i_mapping, desc->page_index,
			       (filler_t *)nfs_readdir_filler, desc);
	if (IS_ERR(page)) {
		status = PTR_ERR(page);
		goto out;
	}

	/* NOTE: Someone else may have changed the READDIRPLUS flag */
	desc->page = page;
	desc->ptr = kmap(page);		/* matching kunmap in nfs_do_filldir */
339
	if (*desc->dir_cookie != 0)
340 341 342
		status = find_dirent(desc);
	else
		status = find_dirent_index(desc);
L
Linus Torvalds 已提交
343 344 345
	if (status < 0)
		dir_page_release(desc);
 out:
346
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n", __func__, status);
L
Linus Torvalds 已提交
347 348 349 350 351 352 353
	return status;
}

/*
 * Recurse through the page cache pages, and return a
 * filled nfs_entry structure of the next directory entry if possible.
 *
354 355
 * The target for the search is '*desc->dir_cookie' if non-0,
 * 'desc->file->f_pos' otherwise
L
Linus Torvalds 已提交
356 357 358 359 360 361 362
 */
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
	int		loop_count = 0;
	int		res;

363
	/* Always search-by-index from the beginning of the cache */
364
	if (*desc->dir_cookie == 0) {
C
Chuck Lever 已提交
365 366
		dfprintk(DIRCACHE, "NFS: readdir_search_pagecache() searching for offset %Ld\n",
				(long long)desc->file->f_pos);
367 368 369 370
		desc->page_index = 0;
		desc->entry->cookie = desc->entry->prev_cookie = 0;
		desc->entry->eof = 0;
		desc->current_index = 0;
371
	} else
C
Chuck Lever 已提交
372 373
		dfprintk(DIRCACHE, "NFS: readdir_search_pagecache() searching for cookie %Lu\n",
				(unsigned long long)*desc->dir_cookie);
374

L
Linus Torvalds 已提交
375 376 377 378 379 380 381 382 383 384 385
	for (;;) {
		res = find_dirent_page(desc);
		if (res != -EAGAIN)
			break;
		/* Align to beginning of next page */
		desc->page_index ++;
		if (loop_count++ > 200) {
			loop_count = 0;
			schedule();
		}
	}
C
Chuck Lever 已提交
386

387
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n", __func__, res);
L
Linus Torvalds 已提交
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
	return res;
}

static inline unsigned int dt_type(struct inode *inode)
{
	return (inode->i_mode >> 12) & 15;
}

static struct dentry *nfs_readdir_lookup(nfs_readdir_descriptor_t *desc);

/*
 * Once we've found the start of the dirent within a page: fill 'er up...
 */
static 
int nfs_do_filldir(nfs_readdir_descriptor_t *desc, void *dirent,
		   filldir_t filldir)
{
	struct file	*file = desc->file;
	struct nfs_entry *entry = desc->entry;
	struct dentry	*dentry = NULL;
P
Peter Staubach 已提交
408
	u64		fileid;
L
Linus Torvalds 已提交
409 410 411
	int		loop_count = 0,
			res;

C
Chuck Lever 已提交
412 413
	dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling starting @ cookie %Lu\n",
			(unsigned long long)entry->cookie);
L
Linus Torvalds 已提交
414 415 416 417 418

	for(;;) {
		unsigned d_type = DT_UNKNOWN;
		/* Note: entry->prev_cookie contains the cookie for
		 *	 retrieving the current dirent on the server */
P
Peter Staubach 已提交
419
		fileid = entry->ino;
L
Linus Torvalds 已提交
420 421 422 423 424 425 426 427 428

		/* Get a dentry if we have one */
		if (dentry != NULL)
			dput(dentry);
		dentry = nfs_readdir_lookup(desc);

		/* Use readdirplus info */
		if (dentry != NULL && dentry->d_inode != NULL) {
			d_type = dt_type(dentry->d_inode);
P
Peter Staubach 已提交
429
			fileid = NFS_FILEID(dentry->d_inode);
L
Linus Torvalds 已提交
430 431 432
		}

		res = filldir(dirent, entry->name, entry->len, 
433 434
			      file->f_pos, nfs_compat_user_ino64(fileid),
			      d_type);
L
Linus Torvalds 已提交
435 436
		if (res < 0)
			break;
437
		file->f_pos++;
438
		*desc->dir_cookie = entry->cookie;
L
Linus Torvalds 已提交
439 440 441 442 443 444 445 446 447 448 449 450
		if (dir_decode(desc) != 0) {
			desc->page_index ++;
			break;
		}
		if (loop_count++ > 200) {
			loop_count = 0;
			schedule();
		}
	}
	dir_page_release(desc);
	if (dentry != NULL)
		dput(dentry);
C
Chuck Lever 已提交
451 452
	dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling ended @ cookie %Lu; returning = %d\n",
			(unsigned long long)*desc->dir_cookie, res);
L
Linus Torvalds 已提交
453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472
	return res;
}

/*
 * If we cannot find a cookie in our cache, we suspect that this is
 * because it points to a deleted file, so we ask the server to return
 * whatever it thinks is the next entry. We then feed this to filldir.
 * If all goes well, we should then be able to find our way round the
 * cache on the next call to readdir_search_pagecache();
 *
 * NOTE: we cannot add the anonymous page to the pagecache because
 *	 the data it contains might not be page aligned. Besides,
 *	 we should already have a complete representation of the
 *	 directory in the page cache by the time we get here.
 */
static inline
int uncached_readdir(nfs_readdir_descriptor_t *desc, void *dirent,
		     filldir_t filldir)
{
	struct file	*file = desc->file;
473
	struct inode	*inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
474 475 476
	struct rpc_cred	*cred = nfs_file_cred(file);
	struct page	*page = NULL;
	int		status;
477
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
478

C
Chuck Lever 已提交
479 480
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
481 482 483 484 485 486

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
487
	timestamp = jiffies;
488
	gencount = nfs_inc_attr_generation_counter();
489 490
	status = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred,
						*desc->dir_cookie, page,
L
Linus Torvalds 已提交
491 492 493 494
						NFS_SERVER(inode)->dtsize,
						desc->plus);
	desc->page = page;
	desc->ptr = kmap(page);		/* matching kunmap in nfs_do_filldir */
495
	if (status >= 0) {
496
		desc->timestamp = timestamp;
497
		desc->gencount = gencount;
498
		desc->timestamp_valid = 1;
L
Linus Torvalds 已提交
499
		if ((status = dir_decode(desc)) == 0)
500
			desc->entry->prev_cookie = *desc->dir_cookie;
L
Linus Torvalds 已提交
501 502 503 504 505 506 507 508 509 510 511 512 513
	} else
		status = -EIO;
	if (status < 0)
		goto out_release;

	status = nfs_do_filldir(desc, dirent, filldir);

	/* Reset read descriptor so it searches the page cache from
	 * the start upon the next call to readdir_search_pagecache() */
	desc->page_index = 0;
	desc->entry->cookie = desc->entry->prev_cookie = 0;
	desc->entry->eof = 0;
 out:
C
Chuck Lever 已提交
514
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
515
			__func__, status);
L
Linus Torvalds 已提交
516 517 518 519 520 521
	return status;
 out_release:
	dir_page_release(desc);
	goto out;
}

522 523 524
/* The file offset position represents the dirent entry number.  A
   last cookie cache takes care of the common case of reading the
   whole directory.
L
Linus Torvalds 已提交
525 526 527
 */
static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
528
	struct dentry	*dentry = filp->f_path.dentry;
L
Linus Torvalds 已提交
529 530 531 532
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
	struct nfs_entry my_entry;
533
	int res = -ENOMEM;
L
Linus Torvalds 已提交
534

C
Chuck Lever 已提交
535
	dfprintk(FILE, "NFS: readdir(%s/%s) starting at cookie %llu\n",
C
Chuck Lever 已提交
536 537
			dentry->d_parent->d_name.name, dentry->d_name.name,
			(long long)filp->f_pos);
C
Chuck Lever 已提交
538 539
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
540
	/*
541
	 * filp->f_pos points to the dirent entry number.
542
	 * *desc->dir_cookie has the cookie for the next entry. We have
543 544
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
545 546 547 548
	 */
	memset(desc, 0, sizeof(*desc));

	desc->file = filp;
549
	desc->dir_cookie = &nfs_file_open_context(filp)->dir_cookie;
L
Linus Torvalds 已提交
550 551 552 553 554
	desc->decode = NFS_PROTO(inode)->decode_dirent;
	desc->plus = NFS_USE_READDIRPLUS(inode);

	my_entry.cookie = my_entry.prev_cookie = 0;
	my_entry.eof = 0;
555 556 557 558 559
	my_entry.fh = nfs_alloc_fhandle();
	my_entry.fattr = nfs_alloc_fattr();
	if (my_entry.fh == NULL || my_entry.fattr == NULL)
		goto out_alloc_failed;

L
Linus Torvalds 已提交
560 561
	desc->entry = &my_entry;

T
Trond Myklebust 已提交
562
	nfs_block_sillyrename(dentry);
563
	res = nfs_revalidate_mapping(inode, filp->f_mapping);
T
Trond Myklebust 已提交
564 565 566
	if (res < 0)
		goto out;

L
Linus Torvalds 已提交
567 568
	while(!desc->entry->eof) {
		res = readdir_search_pagecache(desc);
569

L
Linus Torvalds 已提交
570 571
		if (res == -EBADCOOKIE) {
			/* This means either end of directory */
572
			if (*desc->dir_cookie && desc->entry->cookie != *desc->dir_cookie) {
L
Linus Torvalds 已提交
573 574 575 576 577 578 579 580 581
				/* Or that the server has 'lost' a cookie */
				res = uncached_readdir(desc, dirent, filldir);
				if (res >= 0)
					continue;
			}
			res = 0;
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
582
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
583 584 585 586 587 588 589 590 591 592 593 594 595 596
			nfs_zap_caches(inode);
			desc->plus = 0;
			desc->entry->eof = 0;
			continue;
		}
		if (res < 0)
			break;

		res = nfs_do_filldir(desc, dirent, filldir);
		if (res < 0) {
			res = 0;
			break;
		}
	}
T
Trond Myklebust 已提交
597
out:
T
Trond Myklebust 已提交
598
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
599 600
	if (res > 0)
		res = 0;
601 602 603 604
out_alloc_failed:
	nfs_free_fattr(my_entry.fattr);
	nfs_free_fhandle(my_entry.fh);
	dfprintk(FILE, "NFS: readdir(%s/%s) returns %d\n",
C
Chuck Lever 已提交
605 606 607
			dentry->d_parent->d_name.name, dentry->d_name.name,
			res);
	return res;
L
Linus Torvalds 已提交
608 609
}

610
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
611
{
612 613 614
	struct dentry *dentry = filp->f_path.dentry;
	struct inode *inode = dentry->d_inode;

C
Chuck Lever 已提交
615
	dfprintk(FILE, "NFS: llseek dir(%s/%s, %lld, %d)\n",
616 617 618 619 620
			dentry->d_parent->d_name.name,
			dentry->d_name.name,
			offset, origin);

	mutex_lock(&inode->i_mutex);
621 622 623 624 625 626 627 628 629 630 631 632
	switch (origin) {
		case 1:
			offset += filp->f_pos;
		case 0:
			if (offset >= 0)
				break;
		default:
			offset = -EINVAL;
			goto out;
	}
	if (offset != filp->f_pos) {
		filp->f_pos = offset;
633
		nfs_file_open_context(filp)->dir_cookie = 0;
634 635
	}
out:
636
	mutex_unlock(&inode->i_mutex);
637 638 639
	return offset;
}

L
Linus Torvalds 已提交
640 641 642 643
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
644
static int nfs_fsync_dir(struct file *filp, struct dentry *dentry, int datasync)
L
Linus Torvalds 已提交
645
{
C
Chuck Lever 已提交
646
	dfprintk(FILE, "NFS: fsync dir(%s/%s) datasync %d\n",
C
Chuck Lever 已提交
647 648 649
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);

650
	nfs_inc_stats(dentry->d_inode, NFSIOS_VFSFSYNC);
L
Linus Torvalds 已提交
651 652 653
	return 0;
}

654 655 656 657 658 659 660 661 662 663 664 665
/**
 * nfs_force_lookup_revalidate - Mark the directory as having changed
 * @dir - pointer to directory inode
 *
 * This forces the revalidation code in nfs_lookup_revalidate() to do a
 * full lookup on all child dentries of 'dir' whenever a change occurs
 * on the server that might have invalidated our dcache.
 *
 * The caller should be holding dir->i_lock
 */
void nfs_force_lookup_revalidate(struct inode *dir)
{
666
	NFS_I(dir)->cache_change_attribute++;
667 668
}

L
Linus Torvalds 已提交
669 670 671 672 673
/*
 * A check for whether or not the parent directory has changed.
 * In the case it has, we assume that the dentries are untrustworthy
 * and may need to be looked up again.
 */
674
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
675 676 677
{
	if (IS_ROOT(dentry))
		return 1;
678 679
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
680 681 682 683 684 685 686 687
	if (!nfs_verify_change_attribute(dir, dentry->d_time))
		return 0;
	/* Revalidate nfsi->cache_change_attribute before we declare a match */
	if (nfs_revalidate_inode(NFS_SERVER(dir), dir) < 0)
		return 0;
	if (!nfs_verify_change_attribute(dir, dentry->d_time))
		return 0;
	return 1;
L
Linus Torvalds 已提交
688 689
}

690 691 692 693 694 695 696 697 698 699 700 701 702 703
/*
 * Return the intent data that applies to this particular path component
 *
 * Note that the current set of intents only apply to the very last
 * component of the path.
 * We check for this using LOOKUP_CONTINUE and LOOKUP_PARENT.
 */
static inline unsigned int nfs_lookup_check_intent(struct nameidata *nd, unsigned int mask)
{
	if (nd->flags & (LOOKUP_CONTINUE|LOOKUP_PARENT))
		return 0;
	return nd->flags & mask;
}

704 705 706 707 708 709 710 711
/*
 * Use intent information to check whether or not we're going to do
 * an O_EXCL create using this path component.
 */
static int nfs_is_exclusive_create(struct inode *dir, struct nameidata *nd)
{
	if (NFS_PROTO(dir)->version == 2)
		return 0;
712
	return nd && nfs_lookup_check_intent(nd, LOOKUP_EXCL);
713 714
}

715 716 717 718 719 720 721 722
/*
 * Inode and filehandle revalidation for lookups.
 *
 * We force revalidation in the cases where the VFS sets LOOKUP_REVAL,
 * or if the intent information indicates that we're about to open this
 * particular file and the "nocto" mount flag is not set.
 *
 */
L
Linus Torvalds 已提交
723 724 725 726 727
static inline
int nfs_lookup_verify_inode(struct inode *inode, struct nameidata *nd)
{
	struct nfs_server *server = NFS_SERVER(inode);

728 729
	if (test_bit(NFS_INO_MOUNTPOINT, &NFS_I(inode)->flags))
		return 0;
L
Linus Torvalds 已提交
730 731
	if (nd != NULL) {
		/* VFS wants an on-the-wire revalidation */
732
		if (nd->flags & LOOKUP_REVAL)
L
Linus Torvalds 已提交
733 734
			goto out_force;
		/* This is an open(2) */
735
		if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 &&
736 737 738
				!(server->flags & NFS_MOUNT_NOCTO) &&
				(S_ISREG(inode->i_mode) ||
				 S_ISDIR(inode->i_mode)))
L
Linus Torvalds 已提交
739
			goto out_force;
740
		return 0;
L
Linus Torvalds 已提交
741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758
	}
	return nfs_revalidate_inode(server, inode);
out_force:
	return __nfs_revalidate_inode(server, inode);
}

/*
 * We judge how long we want to trust negative
 * dentries by looking at the parent inode mtime.
 *
 * If parent mtime has changed, we revalidate, else we wait for a
 * period corresponding to the parent's attribute cache timeout value.
 */
static inline
int nfs_neg_need_reval(struct inode *dir, struct dentry *dentry,
		       struct nameidata *nd)
{
	/* Don't revalidate a negative dentry if we're creating a new file */
759
	if (nd != NULL && nfs_lookup_check_intent(nd, LOOKUP_CREATE) != 0)
L
Linus Torvalds 已提交
760
		return 0;
761 762
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781
	return !nfs_check_verifier(dir, dentry);
}

/*
 * This is called every time the dcache has a lookup hit,
 * and we should check whether we can really trust that
 * lookup.
 *
 * NOTE! The hit can be a negative hit too, don't assume
 * we have an inode!
 *
 * If the parent directory is seen to have changed, we throw out the
 * cached dentry and do a new lookup.
 */
static int nfs_lookup_revalidate(struct dentry * dentry, struct nameidata *nd)
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
782 783
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
784 785 786 787
	int error;

	parent = dget_parent(dentry);
	dir = parent->d_inode;
C
Chuck Lever 已提交
788
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
789 790 791 792 793 794 795 796 797
	inode = dentry->d_inode;

	if (!inode) {
		if (nfs_neg_need_reval(dir, dentry, nd))
			goto out_bad;
		goto out_valid;
	}

	if (is_bad_inode(inode)) {
C
Chuck Lever 已提交
798
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
799
				__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
800
				dentry->d_name.name);
L
Linus Torvalds 已提交
801 802 803
		goto out_bad;
	}

804 805 806
	if (nfs_have_delegation(inode, FMODE_READ))
		goto out_set_verifier;

L
Linus Torvalds 已提交
807
	/* Force a full look up iff the parent directory has changed */
808
	if (!nfs_is_exclusive_create(dir, nd) && nfs_check_verifier(dir, dentry)) {
L
Linus Torvalds 已提交
809 810 811 812 813 814 815 816
		if (nfs_lookup_verify_inode(inode, nd))
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

817 818 819 820 821 822 823
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
824 825
	if (error)
		goto out_bad;
826
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
827
		goto out_bad;
828
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
829 830
		goto out_bad;

831 832
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
833
out_set_verifier:
834
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
835 836
 out_valid:
	dput(parent);
C
Chuck Lever 已提交
837
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n",
838
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
839
			dentry->d_name.name);
L
Linus Torvalds 已提交
840 841 842 843
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
844
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
845 846 847 848 849 850
	if (inode && S_ISDIR(inode->i_mode)) {
		/* Purge readdir caches. */
		nfs_zap_caches(inode);
		/* If we have submounts, don't unhash ! */
		if (have_submounts(dentry))
			goto out_valid;
851 852
		if (dentry->d_flags & DCACHE_DISCONNECTED)
			goto out_valid;
L
Linus Torvalds 已提交
853 854 855
		shrink_dcache_parent(dentry);
	}
	d_drop(dentry);
856 857
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
858
	dput(parent);
C
Chuck Lever 已提交
859
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
860
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
861
			dentry->d_name.name);
L
Linus Torvalds 已提交
862
	return 0;
863 864 865 866 867 868 869 870
out_error:
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
	dput(parent);
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) lookup returned error %d\n",
			__func__, dentry->d_parent->d_name.name,
			dentry->d_name.name, error);
	return error;
L
Linus Torvalds 已提交
871 872 873 874 875 876 877 878 879 880 881
}

/*
 * This is called from dput() when d_count is going to 0.
 */
static int nfs_dentry_delete(struct dentry *dentry)
{
	dfprintk(VFS, "NFS: dentry_delete(%s/%s, %x)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		dentry->d_flags);

882 883 884 885
	/* Unhash any dentry with a stale inode */
	if (dentry->d_inode != NULL && NFS_STALE(dentry->d_inode))
		return 1;

L
Linus Torvalds 已提交
886 887 888 889 890 891 892 893 894 895 896 897 898
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
		/* Unhash it, so that ->d_iput() would be called */
		return 1;
	}
	if (!(dentry->d_sb->s_flags & MS_ACTIVE)) {
		/* Unhash it, so that ancestors of killed async unlink
		 * files will be cleaned up during umount */
		return 1;
	}
	return 0;

}

899 900 901 902 903 904 905 906
static void nfs_drop_nlink(struct inode *inode)
{
	spin_lock(&inode->i_lock);
	if (inode->i_nlink > 0)
		drop_nlink(inode);
	spin_unlock(&inode->i_lock);
}

L
Linus Torvalds 已提交
907 908 909 910 911 912
/*
 * Called when the dentry loses inode.
 * We use it to clean up silly-renamed files.
 */
static void nfs_dentry_iput(struct dentry *dentry, struct inode *inode)
{
913 914 915 916
	if (S_ISDIR(inode->i_mode))
		/* drop any readdir cache as it could easily be old */
		NFS_I(inode)->cache_validity |= NFS_INO_INVALID_DATA;

L
Linus Torvalds 已提交
917
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
918
		drop_nlink(inode);
919
		nfs_complete_unlink(dentry, inode);
L
Linus Torvalds 已提交
920 921 922 923
	}
	iput(inode);
}

A
Al Viro 已提交
924
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
925 926 927 928 929 930 931 932
	.d_revalidate	= nfs_lookup_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
};

static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
{
	struct dentry *res;
T
Trond Myklebust 已提交
933
	struct dentry *parent;
L
Linus Torvalds 已提交
934
	struct inode *inode = NULL;
935 936
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
937 938 939 940
	int error;

	dfprintk(VFS, "NFS: lookup(%s/%s)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name);
C
Chuck Lever 已提交
941
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
942 943 944 945 946 947 948

	res = ERR_PTR(-ENAMETOOLONG);
	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
		goto out;

	dentry->d_op = NFS_PROTO(dir)->dentry_ops;

949 950 951 952 953 954 955
	/*
	 * If we're doing an exclusive create, optimize away the lookup
	 * but don't hash the dentry.
	 */
	if (nfs_is_exclusive_create(dir, nd)) {
		d_instantiate(dentry, NULL);
		res = NULL;
956
		goto out;
957
	}
L
Linus Torvalds 已提交
958

959 960 961 962 963 964
	res = ERR_PTR(-ENOMEM);
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out;

T
Trond Myklebust 已提交
965 966 967
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
	nfs_block_sillyrename(parent);
968
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
969 970 971 972
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
973
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
974
	}
975
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
976 977
	res = (struct dentry *)inode;
	if (IS_ERR(res))
T
Trond Myklebust 已提交
978
		goto out_unblock_sillyrename;
979

L
Linus Torvalds 已提交
980
no_entry:
981
	res = d_materialise_unique(dentry, inode);
982 983
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
984
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
985
		dentry = res;
986
	}
L
Linus Torvalds 已提交
987
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
988 989
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
L
Linus Torvalds 已提交
990
out:
991 992
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
993 994 995 996 997 998
	return res;
}

#ifdef CONFIG_NFS_V4
static int nfs_open_revalidate(struct dentry *, struct nameidata *);

A
Al Viro 已提交
999
const struct dentry_operations nfs4_dentry_operations = {
L
Linus Torvalds 已提交
1000 1001 1002 1003 1004
	.d_revalidate	= nfs_open_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
};

1005 1006 1007 1008
/*
 * Use intent information to determine whether we need to substitute
 * the NFSv4-style stateful OPEN for the LOOKUP call
 */
T
Trond Myklebust 已提交
1009
static int is_atomic_open(struct nameidata *nd)
L
Linus Torvalds 已提交
1010
{
1011
	if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_OPEN) == 0)
L
Linus Torvalds 已提交
1012 1013 1014 1015 1016
		return 0;
	/* NFS does not (yet) have a stateful open for directories */
	if (nd->flags & LOOKUP_DIRECTORY)
		return 0;
	/* Are we trying to write to a read only partition? */
1017 1018
	if (__mnt_is_readonly(nd->path.mnt) &&
	    (nd->intent.open.flags & (O_CREAT|O_TRUNC|FMODE_WRITE)))
L
Linus Torvalds 已提交
1019 1020 1021 1022 1023 1024 1025 1026 1027
		return 0;
	return 1;
}

static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
{
	struct dentry *res = NULL;
	int error;

C
Chuck Lever 已提交
1028 1029 1030
	dfprintk(VFS, "NFS: atomic_lookup(%s/%ld), %s\n",
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);

L
Linus Torvalds 已提交
1031
	/* Check that we are indeed trying to open this file */
T
Trond Myklebust 已提交
1032
	if (!is_atomic_open(nd))
L
Linus Torvalds 已提交
1033 1034 1035 1036 1037 1038 1039 1040
		goto no_open;

	if (dentry->d_name.len > NFS_SERVER(dir)->namelen) {
		res = ERR_PTR(-ENAMETOOLONG);
		goto out;
	}
	dentry->d_op = NFS_PROTO(dir)->dentry_ops;

1041 1042
	/* Let vfs_create() deal with O_EXCL. Instantiate, but don't hash
	 * the dentry. */
1043
	if (nd->flags & LOOKUP_EXCL) {
1044
		d_instantiate(dentry, NULL);
1045 1046
		goto out;
	}
L
Linus Torvalds 已提交
1047 1048

	/* Open the file on the server */
1049
	res = nfs4_atomic_open(dir, dentry, nd);
1050 1051
	if (IS_ERR(res)) {
		error = PTR_ERR(res);
L
Linus Torvalds 已提交
1052 1053 1054
		switch (error) {
			/* Make a negative dentry */
			case -ENOENT:
1055 1056
				res = NULL;
				goto out;
L
Linus Torvalds 已提交
1057
			/* This turned out not to be a regular file */
1058
			case -EISDIR:
1059 1060
			case -ENOTDIR:
				goto no_open;
L
Linus Torvalds 已提交
1061 1062 1063 1064 1065 1066 1067
			case -ELOOP:
				if (!(nd->intent.open.flags & O_NOFOLLOW))
					goto no_open;
			/* case -EINVAL: */
			default:
				goto out;
		}
1068
	} else if (res != NULL)
L
Linus Torvalds 已提交
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
		dentry = res;
out:
	return res;
no_open:
	return nfs_lookup(dir, dentry, nd);
}

static int nfs_open_revalidate(struct dentry *dentry, struct nameidata *nd)
{
	struct dentry *parent = NULL;
	struct inode *inode = dentry->d_inode;
	struct inode *dir;
	int openflags, ret = 0;

1083
	if (!is_atomic_open(nd) || d_mountpoint(dentry))
T
Trond Myklebust 已提交
1084
		goto no_open;
L
Linus Torvalds 已提交
1085 1086 1087 1088 1089
	parent = dget_parent(dentry);
	dir = parent->d_inode;
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1090 1091 1092
	if (inode == NULL) {
		if (!nfs_neg_need_reval(dir, dentry, nd))
			ret = 1;
L
Linus Torvalds 已提交
1093
		goto out;
1094 1095
	}

L
Linus Torvalds 已提交
1096 1097
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
T
Trond Myklebust 已提交
1098
		goto no_open_dput;
L
Linus Torvalds 已提交
1099 1100 1101
	openflags = nd->intent.open.flags;
	/* We cannot do exclusive creation on a positive dentry */
	if ((openflags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL))
T
Trond Myklebust 已提交
1102
		goto no_open_dput;
L
Linus Torvalds 已提交
1103 1104 1105 1106
	/* We can't create new files, or truncate existing ones here */
	openflags &= ~(O_CREAT|O_TRUNC);

	/*
1107
	 * Note: we're not holding inode->i_mutex and so may be racing with
L
Linus Torvalds 已提交
1108 1109 1110
	 * operations that change the directory. We therefore save the
	 * change attribute *before* we do the RPC call.
	 */
1111
	ret = nfs4_open_revalidate(dir, dentry, openflags, nd);
L
Linus Torvalds 已提交
1112 1113 1114 1115 1116
out:
	dput(parent);
	if (!ret)
		d_drop(dentry);
	return ret;
T
Trond Myklebust 已提交
1117
no_open_dput:
L
Linus Torvalds 已提交
1118
	dput(parent);
T
Trond Myklebust 已提交
1119
no_open:
L
Linus Torvalds 已提交
1120 1121 1122 1123 1124 1125
	return nfs_lookup_revalidate(dentry, nd);
}
#endif /* CONFIG_NFSV4 */

static struct dentry *nfs_readdir_lookup(nfs_readdir_descriptor_t *desc)
{
1126
	struct dentry *parent = desc->file->f_path.dentry;
L
Linus Torvalds 已提交
1127 1128 1129 1130 1131 1132 1133 1134
	struct inode *dir = parent->d_inode;
	struct nfs_entry *entry = desc->entry;
	struct dentry *dentry, *alias;
	struct qstr name = {
		.name = entry->name,
		.len = entry->len,
	};
	struct inode *inode;
1135
	unsigned long verf = nfs_save_change_attribute(dir);
L
Linus Torvalds 已提交
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145

	switch (name.len) {
		case 2:
			if (name.name[0] == '.' && name.name[1] == '.')
				return dget_parent(parent);
			break;
		case 1:
			if (name.name[0] == '.')
				return dget(parent);
	}
1146 1147 1148 1149 1150 1151 1152 1153

	spin_lock(&dir->i_lock);
	if (NFS_I(dir)->cache_validity & NFS_INO_INVALID_DATA) {
		spin_unlock(&dir->i_lock);
		return NULL;
	}
	spin_unlock(&dir->i_lock);

L
Linus Torvalds 已提交
1154 1155
	name.hash = full_name_hash(name.name, name.len);
	dentry = d_lookup(parent, &name);
1156
	if (dentry != NULL) {
1157 1158 1159 1160 1161 1162 1163 1164 1165 1166
		/* Is this a positive dentry that matches the readdir info? */
		if (dentry->d_inode != NULL &&
				(NFS_FILEID(dentry->d_inode) == entry->ino ||
				d_mountpoint(dentry))) {
			if (!desc->plus || entry->fh->size == 0)
				return dentry;
			if (nfs_compare_fh(NFS_FH(dentry->d_inode),
						entry->fh) == 0)
				goto out_renew;
		}
1167 1168 1169 1170
		/* No, so d_drop to allow one to be created */
		d_drop(dentry);
		dput(dentry);
	}
L
Linus Torvalds 已提交
1171 1172
	if (!desc->plus || !(entry->fattr->valid & NFS_ATTR_FATTR))
		return NULL;
1173 1174
	if (name.len > NFS_SERVER(dir)->namelen)
		return NULL;
1175
	/* Note: caller is already holding the dir->i_mutex! */
L
Linus Torvalds 已提交
1176 1177 1178 1179 1180
	dentry = d_alloc(parent, &name);
	if (dentry == NULL)
		return NULL;
	dentry->d_op = NFS_PROTO(dir)->dentry_ops;
	inode = nfs_fhget(dentry->d_sb, entry->fh, entry->fattr);
1181
	if (IS_ERR(inode)) {
L
Linus Torvalds 已提交
1182 1183 1184
		dput(dentry);
		return NULL;
	}
1185 1186

	alias = d_materialise_unique(dentry, inode);
L
Linus Torvalds 已提交
1187 1188
	if (alias != NULL) {
		dput(dentry);
1189 1190
		if (IS_ERR(alias))
			return NULL;
L
Linus Torvalds 已提交
1191 1192
		dentry = alias;
	}
1193

1194
out_renew:
1195
	nfs_set_verifier(dentry, verf);
1196
	return dentry;
L
Linus Torvalds 已提交
1197 1198 1199 1200 1201 1202 1203 1204
}

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
				struct nfs_fattr *fattr)
{
1205 1206
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1207 1208 1209
	struct inode *inode;
	int error = -EACCES;

1210 1211
	d_drop(dentry);

L
Linus Torvalds 已提交
1212 1213
	/* We may have been initialized further down */
	if (dentry->d_inode)
1214
		goto out;
L
Linus Torvalds 已提交
1215 1216 1217
	if (fhandle->size == 0) {
		error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
		if (error)
1218
			goto out_error;
L
Linus Torvalds 已提交
1219
	}
1220
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1221 1222
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1223
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr);
L
Linus Torvalds 已提交
1224
		if (error < 0)
1225
			goto out_error;
L
Linus Torvalds 已提交
1226 1227
	}
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
1228 1229
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1230 1231 1232 1233
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1234
	return 0;
1235 1236 1237 1238
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253
}

/*
 * Following a failed create operation, we drop the dentry rather
 * than retain a negative dentry. This avoids a problem in the event
 * that the operation succeeded on the server, but an error in the
 * reply path made it appear to have failed.
 */
static int nfs_create(struct inode *dir, struct dentry *dentry, int mode,
		struct nameidata *nd)
{
	struct iattr attr;
	int error;
	int open_flags = 0;

C
Chuck Lever 已提交
1254 1255
	dfprintk(VFS, "NFS: create(%s/%ld), %s\n",
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
L
Linus Torvalds 已提交
1256 1257 1258 1259

	attr.ia_mode = mode;
	attr.ia_valid = ATTR_MODE;

1260
	if ((nd->flags & LOOKUP_CREATE) != 0)
L
Linus Torvalds 已提交
1261 1262
		open_flags = nd->intent.open.flags;

1263
	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, nd);
L
Linus Torvalds 已提交
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
static int
nfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t rdev)
{
	struct iattr attr;
	int status;

C
Chuck Lever 已提交
1281 1282
	dfprintk(VFS, "NFS: mknod(%s/%ld), %s\n",
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
L
Linus Torvalds 已提交
1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306

	if (!new_valid_dev(rdev))
		return -EINVAL;

	attr.ia_mode = mode;
	attr.ia_valid = ATTR_MODE;

	status = NFS_PROTO(dir)->mknod(dir, dentry, &attr, rdev);
	if (status != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return status;
}

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
static int nfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
	struct iattr attr;
	int error;

C
Chuck Lever 已提交
1307 1308
	dfprintk(VFS, "NFS: mkdir(%s/%ld), %s\n",
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
L
Linus Torvalds 已提交
1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321

	attr.ia_valid = ATTR_MODE;
	attr.ia_mode = mode | S_IFDIR;

	error = NFS_PROTO(dir)->mkdir(dir, dentry, &attr);
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}

1322 1323 1324 1325 1326 1327
static void nfs_dentry_handle_enoent(struct dentry *dentry)
{
	if (dentry->d_inode != NULL && !d_unhashed(dentry))
		d_delete(dentry);
}

L
Linus Torvalds 已提交
1328 1329 1330 1331
static int nfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	int error;

C
Chuck Lever 已提交
1332 1333
	dfprintk(VFS, "NFS: rmdir(%s/%ld), %s\n",
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
L
Linus Torvalds 已提交
1334 1335 1336 1337

	error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
	/* Ensure the VFS deletes this inode */
	if (error == 0 && dentry->d_inode != NULL)
1338
		clear_nlink(dentry->d_inode);
1339 1340
	else if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1341 1342 1343 1344 1345 1346 1347

	return error;
}

static int nfs_sillyrename(struct inode *dir, struct dentry *dentry)
{
	static unsigned int sillycounter;
P
Peter Staubach 已提交
1348
	const int      fileidsize  = sizeof(NFS_FILEID(dentry->d_inode))*2;
L
Linus Torvalds 已提交
1349
	const int      countersize = sizeof(sillycounter)*2;
P
Peter Staubach 已提交
1350
	const int      slen        = sizeof(".nfs")+fileidsize+countersize-1;
L
Linus Torvalds 已提交
1351 1352 1353 1354 1355 1356 1357 1358
	char           silly[slen+1];
	struct qstr    qsilly;
	struct dentry *sdentry;
	int            error = -EIO;

	dfprintk(VFS, "NFS: silly-rename(%s/%s, ct=%d)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name, 
		atomic_read(&dentry->d_count));
C
Chuck Lever 已提交
1359
	nfs_inc_stats(dir, NFSIOS_SILLYRENAME);
L
Linus Torvalds 已提交
1360 1361 1362 1363 1364 1365 1366 1367

	/*
	 * We don't allow a dentry to be silly-renamed twice.
	 */
	error = -EBUSY;
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
		goto out;

P
Peter Staubach 已提交
1368 1369 1370
	sprintf(silly, ".nfs%*.*Lx",
		fileidsize, fileidsize,
		(unsigned long long)NFS_FILEID(dentry->d_inode));
L
Linus Torvalds 已提交
1371

1372 1373 1374
	/* Return delegation in anticipation of the rename */
	nfs_inode_return_delegation(dentry->d_inode);

L
Linus Torvalds 已提交
1375 1376 1377 1378 1379 1380 1381 1382
	sdentry = NULL;
	do {
		char *suffix = silly + slen - countersize;

		dput(sdentry);
		sillycounter++;
		sprintf(suffix, "%*.*x", countersize, countersize, sillycounter);

C
Chuck Lever 已提交
1383 1384
		dfprintk(VFS, "NFS: trying to rename %s to %s\n",
				dentry->d_name.name, silly);
L
Linus Torvalds 已提交
1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
		
		sdentry = lookup_one_len(silly, dentry->d_parent, slen);
		/*
		 * N.B. Better to return EBUSY here ... it could be
		 * dangerous to delete the file while it's in use.
		 */
		if (IS_ERR(sdentry))
			goto out;
	} while(sdentry->d_inode != NULL); /* need negative lookup */

	qsilly.name = silly;
	qsilly.len  = strlen(silly);
	if (dentry->d_inode) {
		error = NFS_PROTO(dir)->rename(dir, &dentry->d_name,
				dir, &qsilly);
1400
		nfs_mark_for_revalidate(dentry->d_inode);
L
Linus Torvalds 已提交
1401 1402 1403 1404 1405 1406
	} else
		error = NFS_PROTO(dir)->rename(dir, &dentry->d_name,
				dir, &qsilly);
	if (!error) {
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
		d_move(dentry, sdentry);
1407
		error = nfs_async_unlink(dir, dentry);
L
Linus Torvalds 已提交
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437
 		/* If we return 0 we don't unlink */
	}
	dput(sdentry);
out:
	return error;
}

/*
 * Remove a file after making sure there are no pending writes,
 * and after checking that the file has only one user. 
 *
 * We invalidate the attribute cache and free the inode prior to the operation
 * to avoid possible races if the server reuses the inode.
 */
static int nfs_safe_remove(struct dentry *dentry)
{
	struct inode *dir = dentry->d_parent->d_inode;
	struct inode *inode = dentry->d_inode;
	int error = -EBUSY;
		
	dfprintk(VFS, "NFS: safe_remove(%s/%s)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name);

	/* If the dentry was sillyrenamed, we simply call d_delete() */
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
		error = 0;
		goto out;
	}

	if (inode != NULL) {
1438
		nfs_inode_return_delegation(inode);
L
Linus Torvalds 已提交
1439 1440 1441
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		/* The VFS may want to delete this inode */
		if (error == 0)
1442
			nfs_drop_nlink(inode);
1443
		nfs_mark_for_revalidate(inode);
L
Linus Torvalds 已提交
1444 1445
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1446 1447
	if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469
out:
	return error;
}

/*  We do silly rename. In case sillyrename() returns -EBUSY, the inode
 *  belongs to an active ".nfs..." file and we return -EBUSY.
 *
 *  If sillyrename() returns 0, we do nothing, otherwise we unlink.
 */
static int nfs_unlink(struct inode *dir, struct dentry *dentry)
{
	int error;
	int need_rehash = 0;

	dfprintk(VFS, "NFS: unlink(%s/%ld, %s)\n", dir->i_sb->s_id,
		dir->i_ino, dentry->d_name.name);

	spin_lock(&dcache_lock);
	spin_lock(&dentry->d_lock);
	if (atomic_read(&dentry->d_count) > 1) {
		spin_unlock(&dentry->d_lock);
		spin_unlock(&dcache_lock);
T
Trond Myklebust 已提交
1470 1471
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1472 1473 1474 1475 1476 1477 1478 1479 1480 1481
		error = nfs_sillyrename(dir, dentry);
		return error;
	}
	if (!d_unhashed(dentry)) {
		__d_drop(dentry);
		need_rehash = 1;
	}
	spin_unlock(&dentry->d_lock);
	spin_unlock(&dcache_lock);
	error = nfs_safe_remove(dentry);
1482
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1483 1484 1485 1486 1487 1488
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
	return error;
}

1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
/*
 * To create a symbolic link, most file systems instantiate a new inode,
 * add a page to it containing the path, then write it out to the disk
 * using prepare_write/commit_write.
 *
 * Unfortunately the NFS client can't create the in-core inode first
 * because it needs a file handle to create an in-core inode (see
 * fs/nfs/inode.c:nfs_fhget).  We only have a file handle *after* the
 * symlink request has completed on the server.
 *
 * So instead we allocate a raw page, copy the symname into it, then do
 * the SYMLINK request with the page as the buffer.  If it succeeds, we
 * now have a new file handle and can instantiate an in-core NFS inode
 * and move the raw page into its mapping.
 */
static int nfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
L
Linus Torvalds 已提交
1505
{
1506 1507 1508
	struct pagevec lru_pvec;
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1509
	struct iattr attr;
1510
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1511 1512 1513 1514 1515
	int error;

	dfprintk(VFS, "NFS: symlink(%s/%ld, %s, %s)\n", dir->i_sb->s_id,
		dir->i_ino, dentry->d_name.name, symname);

1516 1517
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1518

1519 1520
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1521

1522
	page = alloc_page(GFP_HIGHUSER);
1523
	if (!page)
1524 1525 1526 1527 1528 1529 1530 1531
		return -ENOMEM;

	kaddr = kmap_atomic(page, KM_USER0);
	memcpy(kaddr, symname, pathlen);
	if (pathlen < PAGE_SIZE)
		memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen);
	kunmap_atomic(kaddr, KM_USER0);

1532
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1533 1534 1535 1536
	if (error != 0) {
		dfprintk(VFS, "NFS: symlink(%s/%ld, %s, %s) error %d\n",
			dir->i_sb->s_id, dir->i_ino,
			dentry->d_name.name, symname, error);
L
Linus Torvalds 已提交
1537
		d_drop(dentry);
1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548
		__free_page(page);
		return error;
	}

	/*
	 * No big deal if we can't add this page to the page cache here.
	 * READLINK will get the missing page from the server if needed.
	 */
	pagevec_init(&lru_pvec, 0);
	if (!add_to_page_cache(page, dentry->d_inode->i_mapping, 0,
							GFP_KERNEL)) {
1549
		pagevec_add(&lru_pvec, page);
1550
		pagevec_lru_add_file(&lru_pvec);
1551 1552 1553 1554 1555 1556
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
}

static int 
nfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = old_dentry->d_inode;
	int error;

	dfprintk(VFS, "NFS: link(%s/%s -> %s/%s)\n",
		old_dentry->d_parent->d_name.name, old_dentry->d_name.name,
		dentry->d_parent->d_name.name, dentry->d_name.name);

1569 1570
	nfs_inode_return_delegation(inode);

1571
	d_drop(dentry);
L
Linus Torvalds 已提交
1572
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1573 1574
	if (error == 0) {
		atomic_inc(&inode->i_count);
1575
		d_add(dentry, inode);
1576
	}
L
Linus Torvalds 已提交
1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
	return error;
}

/*
 * RENAME
 * FIXME: Some nfsds, like the Linux user space nfsd, may generate a
 * different file handle for the same inode after a rename (e.g. when
 * moving to a different directory). A fail-safe method to do so would
 * be to look up old_dir/old_name, create a link to new_dir/new_name and
 * rename the old file using the sillyrename stuff. This way, the original
 * file in old_dir will go away when the last process iput()s the inode.
 *
 * FIXED.
 * 
 * It actually works quite well. One needs to have the possibility for
 * at least one ".nfs..." file in each directory the file ever gets
 * moved or linked to which happens automagically with the new
 * implementation that only depends on the dcache stuff instead of
 * using the inode layer
 *
 * Unfortunately, things are a little more complicated than indicated
 * above. For a cross-directory move, we want to make sure we can get
 * rid of the old inode after the operation.  This means there must be
 * no pending writes (if it's a file), and the use count must be 1.
 * If these conditions are met, we can drop the dentries before doing
 * the rename.
 */
static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
		      struct inode *new_dir, struct dentry *new_dentry)
{
	struct inode *old_inode = old_dentry->d_inode;
	struct inode *new_inode = new_dentry->d_inode;
	struct dentry *dentry = NULL, *rehash = NULL;
	int error = -EBUSY;

	dfprintk(VFS, "NFS: rename(%s/%s -> %s/%s, ct=%d)\n",
		 old_dentry->d_parent->d_name.name, old_dentry->d_name.name,
		 new_dentry->d_parent->d_name.name, new_dentry->d_name.name,
		 atomic_read(&new_dentry->d_count));

	/*
M
Miklos Szeredi 已提交
1618 1619 1620 1621
	 * For non-directories, check whether the target is busy and if so,
	 * make a copy of the dentry and then do a silly-rename. If the
	 * silly-rename succeeds, the copied dentry is hashed and becomes
	 * the new target.
L
Linus Torvalds 已提交
1622
	 */
1623 1624 1625 1626 1627 1628 1629 1630 1631
	if (new_inode && !S_ISDIR(new_inode->i_mode)) {
		/*
		 * To prevent any new references to the target during the
		 * rename, we unhash the dentry in advance.
		 */
		if (!d_unhashed(new_dentry)) {
			d_drop(new_dentry);
			rehash = new_dentry;
		}
L
Linus Torvalds 已提交
1632

1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643
		if (atomic_read(&new_dentry->d_count) > 2) {
			int err;

			/* copy the target dentry's name */
			dentry = d_alloc(new_dentry->d_parent,
					 &new_dentry->d_name);
			if (!dentry)
				goto out;

			/* silly-rename the existing target ... */
			err = nfs_sillyrename(new_dir, new_dentry);
1644
			if (err)
1645
				goto out;
1646 1647

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
1648
			rehash = NULL;
1649
			new_inode = NULL;
1650
		}
1651
	}
L
Linus Torvalds 已提交
1652 1653 1654 1655 1656

	/*
	 * ... prune child dentries and writebacks if needed.
	 */
	if (atomic_read(&old_dentry->d_count) > 1) {
1657 1658
		if (S_ISREG(old_inode->i_mode))
			nfs_wb_all(old_inode);
L
Linus Torvalds 已提交
1659 1660
		shrink_dcache_parent(old_dentry);
	}
1661
	nfs_inode_return_delegation(old_inode);
L
Linus Torvalds 已提交
1662

1663
	if (new_inode != NULL)
1664
		nfs_inode_return_delegation(new_inode);
L
Linus Torvalds 已提交
1665 1666 1667

	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
1668
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
1669 1670 1671 1672
out:
	if (rehash)
		d_rehash(rehash);
	if (!error) {
1673 1674
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
1675
		d_move(old_dentry, new_dentry);
1676 1677
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
1678 1679
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
1680 1681 1682 1683 1684 1685 1686

	/* new dentry created? */
	if (dentry)
		dput(dentry);
	return error;
}

1687 1688 1689 1690
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

1691 1692 1693 1694
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
1695 1696 1697
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
1698 1699
}

1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710
static void nfs_access_free_list(struct list_head *head)
{
	struct nfs_access_entry *cache;

	while (!list_empty(head)) {
		cache = list_entry(head->next, struct nfs_access_entry, lru);
		list_del(&cache->lru);
		nfs_access_free_entry(cache);
	}
}

1711 1712 1713 1714 1715 1716
int nfs_access_cache_shrinker(int nr_to_scan, gfp_t gfp_mask)
{
	LIST_HEAD(head);
	struct nfs_inode *nfsi;
	struct nfs_access_entry *cache;

1717 1718
	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
		return (nr_to_scan == 0) ? 0 : -1;
1719
restart:
1720
	spin_lock(&nfs_access_lru_lock);
1721
	list_for_each_entry(nfsi, &nfs_access_lru_list, access_cache_inode_lru) {
1722
		struct rw_semaphore *s_umount;
1723 1724 1725 1726
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
1727 1728 1729
		s_umount = &nfsi->vfs_inode.i_sb->s_umount;
		if (!down_read_trylock(s_umount))
			continue;
1730
		inode = igrab(&nfsi->vfs_inode);
1731 1732
		if (inode == NULL) {
			up_read(s_umount);
1733
			continue;
1734
		}
1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750
		spin_lock(&inode->i_lock);
		if (list_empty(&nfsi->access_cache_entry_lru))
			goto remove_lru_entry;
		cache = list_entry(nfsi->access_cache_entry_lru.next,
				struct nfs_access_entry, lru);
		list_move(&cache->lru, &head);
		rb_erase(&cache->rb_node, &nfsi->access_cache);
		if (!list_empty(&nfsi->access_cache_entry_lru))
			list_move_tail(&nfsi->access_cache_inode_lru,
					&nfs_access_lru_list);
		else {
remove_lru_entry:
			list_del_init(&nfsi->access_cache_inode_lru);
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
		}
		spin_unlock(&inode->i_lock);
1751
		spin_unlock(&nfs_access_lru_lock);
1752
		iput(inode);
1753
		up_read(s_umount);
1754 1755 1756
		goto restart;
	}
	spin_unlock(&nfs_access_lru_lock);
1757
	nfs_access_free_list(&head);
1758 1759 1760
	return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure;
}

1761
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
1762
{
1763
	struct rb_root *root_node = &nfsi->access_cache;
1764
	struct rb_node *n;
1765 1766 1767 1768 1769 1770
	struct nfs_access_entry *entry;

	/* Unhook entries from the cache */
	while ((n = rb_first(root_node)) != NULL) {
		entry = rb_entry(n, struct nfs_access_entry, rb_node);
		rb_erase(n, root_node);
1771
		list_move(&entry->lru, head);
1772 1773
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
1774 1775
}

1776
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
1777
{
1778 1779 1780 1781
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
1782
	/* Remove from global LRU init */
1783 1784
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
1785 1786
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

1787
	spin_lock(&inode->i_lock);
1788 1789 1790 1791
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
1792
}
L
Linus Torvalds 已提交
1793

1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807
static struct nfs_access_entry *nfs_access_search_rbtree(struct inode *inode, struct rpc_cred *cred)
{
	struct rb_node *n = NFS_I(inode)->access_cache.rb_node;
	struct nfs_access_entry *entry;

	while (n != NULL) {
		entry = rb_entry(n, struct nfs_access_entry, rb_node);

		if (cred < entry->cred)
			n = n->rb_left;
		else if (cred > entry->cred)
			n = n->rb_right;
		else
			return entry;
L
Linus Torvalds 已提交
1808
	}
1809 1810 1811
	return NULL;
}

1812
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
1813 1814 1815 1816 1817
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

1818
	spin_lock(&inode->i_lock);
1819 1820 1821 1822 1823
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
1824
	if (!nfs_have_delegated_attributes(inode) &&
1825
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
1826 1827 1828 1829
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
1830
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
1831 1832 1833 1834 1835 1836
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
1837
	list_del(&cache->lru);
1838 1839 1840 1841
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
1842 1843
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
1844 1845 1846 1847 1848
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
1849 1850
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
	struct rb_node **p = &root_node->rb_node;
	struct rb_node *parent = NULL;
	struct nfs_access_entry *entry;

	spin_lock(&inode->i_lock);
	while (*p != NULL) {
		parent = *p;
		entry = rb_entry(parent, struct nfs_access_entry, rb_node);

		if (set->cred < entry->cred)
			p = &parent->rb_left;
		else if (set->cred > entry->cred)
			p = &parent->rb_right;
		else
			goto found;
	}
	rb_link_node(&set->rb_node, parent, p);
	rb_insert_color(&set->rb_node, root_node);
1869
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
1870
	spin_unlock(&inode->i_lock);
1871 1872 1873
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
1874 1875
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
1876 1877 1878 1879
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

1880
static void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
1881 1882 1883 1884 1885
{
	struct nfs_access_entry *cache = kmalloc(sizeof(*cache), GFP_KERNEL);
	if (cache == NULL)
		return;
	RB_CLEAR_NODE(&cache->rb_node);
L
Linus Torvalds 已提交
1886
	cache->jiffies = set->jiffies;
1887
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
1888
	cache->mask = set->mask;
1889 1890

	nfs_access_add_rbtree(inode, cache);
1891 1892 1893 1894 1895 1896 1897

	/* Update accounting */
	smp_mb__before_atomic_inc();
	atomic_long_inc(&nfs_access_nr_entries);
	smp_mb__after_atomic_inc();

	/* Add inode to global LRU list */
1898
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
1899
		spin_lock(&nfs_access_lru_lock);
1900 1901 1902
		if (!test_and_set_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
			list_add_tail(&NFS_I(inode)->access_cache_inode_lru,
					&nfs_access_lru_list);
1903 1904
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
}

static int nfs_do_access(struct inode *inode, struct rpc_cred *cred, int mask)
{
	struct nfs_access_entry cache;
	int status;

	status = nfs_access_get_cached(inode, cred, &cache);
	if (status == 0)
		goto out;

	/* Be clever: ask server to check for all possible rights */
	cache.mask = MAY_EXEC | MAY_WRITE | MAY_READ;
	cache.cred = cred;
	cache.jiffies = jiffies;
	status = NFS_PROTO(inode)->access(inode, &cache);
1921 1922 1923 1924 1925 1926
	if (status != 0) {
		if (status == -ESTALE) {
			nfs_zap_caches(inode);
			if (!S_ISDIR(inode->i_mode))
				set_bit(NFS_INO_STALE, &NFS_I(inode)->flags);
		}
L
Linus Torvalds 已提交
1927
		return status;
1928
	}
L
Linus Torvalds 已提交
1929 1930
	nfs_access_add_cache(inode, &cache);
out:
1931
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
1932 1933 1934 1935
		return 0;
	return -EACCES;
}

1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

	if (openflags & FMODE_READ)
		mask |= MAY_READ;
	if (openflags & FMODE_WRITE)
		mask |= MAY_WRITE;
	if (openflags & FMODE_EXEC)
		mask |= MAY_EXEC;
	return mask;
}

int nfs_may_open(struct inode *inode, struct rpc_cred *cred, int openflags)
{
	return nfs_do_access(inode, cred, nfs_open_permission_mask(openflags));
}

1954
int nfs_permission(struct inode *inode, int mask)
L
Linus Torvalds 已提交
1955 1956 1957 1958
{
	struct rpc_cred *cred;
	int res = 0;

C
Chuck Lever 已提交
1959 1960
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

1961
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
1962 1963
		goto out;
	/* Is this sys_access() ? */
1964
	if (mask & MAY_ACCESS)
L
Linus Torvalds 已提交
1965 1966 1967 1968 1969 1970 1971 1972
		goto force_lookup;

	switch (inode->i_mode & S_IFMT) {
		case S_IFLNK:
			goto out;
		case S_IFREG:
			/* NFSv4 has atomic_open... */
			if (nfs_server_capable(inode, NFS_CAP_ATOMIC_OPEN)
1973 1974
					&& (mask & MAY_OPEN)
					&& !(mask & MAY_EXEC))
L
Linus Torvalds 已提交
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989
				goto out;
			break;
		case S_IFDIR:
			/*
			 * Optimize away all write operations, since the server
			 * will check permissions when we perform the op.
			 */
			if ((mask & MAY_WRITE) && !(mask & MAY_READ))
				goto out;
	}

force_lookup:
	if (!NFS_PROTO(inode)->access)
		goto out_notsup;

1990
	cred = rpc_lookup_cred();
L
Linus Torvalds 已提交
1991 1992 1993 1994 1995 1996
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
out:
1997 1998 1999
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

C
Chuck Lever 已提交
2000 2001
	dfprintk(VFS, "NFS: permission(%s/%ld), mask=0x%x, res=%d\n",
		inode->i_sb->s_id, inode->i_ino, mask, res);
L
Linus Torvalds 已提交
2002 2003 2004 2005 2006
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
		res = generic_permission(inode, mask, NULL);
C
Chuck Lever 已提交
2007
	goto out;
L
Linus Torvalds 已提交
2008 2009 2010 2011 2012 2013 2014 2015
}

/*
 * Local variables:
 *  version-control: t
 *  kept-new-versions: 5
 * End:
 */