dir.c 53.3 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 32
/*
 *  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>
#include <linux/smp_lock.h>
33
#include <linux/pagevec.h>
L
Linus Torvalds 已提交
34
#include <linux/namei.h>
35
#include <linux/mount.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"
L
Linus Torvalds 已提交
40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56

#define NFS_PARANOIA 1
/* #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
135 136 137
	dfprintk(VFS, "NFS: opendir(%s/%ld)\n",
			inode->i_sb->s_id, inode->i_ino);

L
Linus Torvalds 已提交
138 139
	lock_kernel();
	/* Call generic open code in order to cache credentials */
140
	res = nfs_open(inode, filp);
L
Linus Torvalds 已提交
141 142 143 144
	unlock_kernel();
	return res;
}

145
typedef __be32 * (*decode_dirent_t)(__be32 *, struct nfs_entry *, int);
L
Linus Torvalds 已提交
146 147 148 149
typedef struct {
	struct file	*file;
	struct page	*page;
	unsigned long	page_index;
150
	__be32		*ptr;
151 152
	u64		*dir_cookie;
	loff_t		current_index;
L
Linus Torvalds 已提交
153 154 155 156
	struct nfs_entry *entry;
	decode_dirent_t	decode;
	int		plus;
	int		error;
157 158
	unsigned long	timestamp;
	int		timestamp_valid;
L
Linus Torvalds 已提交
159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
} 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;
177
	struct inode	*inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
178 179 180 181
	struct rpc_cred	*cred = nfs_file_cred(file);
	unsigned long	timestamp;
	int		error;

C
Chuck Lever 已提交
182 183 184
	dfprintk(DIRCACHE, "NFS: %s: reading cookie %Lu into page %lu\n",
			__FUNCTION__, (long long)desc->entry->cookie,
			page->index);
L
Linus Torvalds 已提交
185 186 187

 again:
	timestamp = jiffies;
188
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, desc->entry->cookie, page,
L
Linus Torvalds 已提交
189 190 191 192 193
					  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;
194
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
L
Linus Torvalds 已提交
195 196 197 198 199
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
200 201
	desc->timestamp = timestamp;
	desc->timestamp_valid = 1;
L
Linus Torvalds 已提交
202
	SetPageUptodate(page);
203
	spin_lock(&inode->i_lock);
204
	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATIME;
205
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
206 207
	/* 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
	 */
210 211 212 213
	if (page->index == 0 && invalidate_inode_pages2_range(inode->i_mapping, PAGE_CACHE_SIZE, -1) < 0) {
		/* Should never happen */
		nfs_zap_mapping(inode, inode->i_mapping);
	}
L
Linus Torvalds 已提交
214 215 216 217 218 219 220 221 222 223 224 225 226
	unlock_page(page);
	return 0;
 error:
	SetPageError(page);
	unlock_page(page);
	nfs_zap_caches(inode);
	desc->error = error;
	return -EIO;
}

static inline
int dir_decode(nfs_readdir_descriptor_t *desc)
{
227
	__be32	*p = desc->ptr;
L
Linus Torvalds 已提交
228 229 230 231
	p = desc->decode(p, desc->entry, desc->plus);
	if (IS_ERR(p))
		return PTR_ERR(p);
	desc->ptr = p;
232 233 234 235
	if (desc->timestamp_valid)
		desc->entry->fattr->time_start = desc->timestamp;
	else
		desc->entry->fattr->valid &= ~NFS_ATTR_FATTR;
L
Linus Torvalds 已提交
236 237 238 239 240 241 242 243 244 245 246 247 248 249
	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
250
 * to readdir, find the next entry with cookie '*desc->dir_cookie'.
L
Linus Torvalds 已提交
251 252 253 254 255 256
 *
 * 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
257
int find_dirent(nfs_readdir_descriptor_t *desc)
L
Linus Torvalds 已提交
258 259 260 261 262 263
{
	struct nfs_entry *entry = desc->entry;
	int		loop_count = 0,
			status;

	while((status = dir_decode(desc)) == 0) {
C
Chuck Lever 已提交
264 265
		dfprintk(DIRCACHE, "NFS: %s: examining cookie %Lu\n",
				__FUNCTION__, (unsigned long long)entry->cookie);
266
		if (entry->prev_cookie == *desc->dir_cookie)
L
Linus Torvalds 已提交
267 268 269 270 271 272 273 274 275 276
			break;
		if (loop_count++ > 200) {
			loop_count = 0;
			schedule();
		}
	}
	return status;
}

/*
277
 * Given a pointer to a buffer that has already been filled by a call
278
 * to readdir, find the entry at offset 'desc->file->f_pos'.
279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295
 *
 * 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
296 297
		dfprintk(DIRCACHE, "NFS: found cookie %Lu at index %Ld\n",
				(unsigned long long)entry->cookie, desc->current_index);
298

299 300
		if (desc->file->f_pos == desc->current_index) {
			*desc->dir_cookie = entry->cookie;
301 302 303 304 305 306 307 308 309 310 311 312 313 314
			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
315 316 317 318
 */
static inline
int find_dirent_page(nfs_readdir_descriptor_t *desc)
{
319
	struct inode	*inode = desc->file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
320 321 322
	struct page	*page;
	int		status;

C
Chuck Lever 已提交
323 324 325
	dfprintk(DIRCACHE, "NFS: %s: searching page %ld for target %Lu\n",
			__FUNCTION__, desc->page_index,
			(long long) *desc->dir_cookie);
L
Linus Torvalds 已提交
326

327 328 329 330
	/* 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
331 332 333 334 335 336 337 338 339 340
	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 */
341
	if (*desc->dir_cookie != 0)
342 343 344
		status = find_dirent(desc);
	else
		status = find_dirent_index(desc);
L
Linus Torvalds 已提交
345 346 347
	if (status < 0)
		dir_page_release(desc);
 out:
C
Chuck Lever 已提交
348
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n", __FUNCTION__, status);
L
Linus Torvalds 已提交
349 350 351 352 353 354 355
	return status;
}

/*
 * Recurse through the page cache pages, and return a
 * filled nfs_entry structure of the next directory entry if possible.
 *
356 357
 * The target for the search is '*desc->dir_cookie' if non-0,
 * 'desc->file->f_pos' otherwise
L
Linus Torvalds 已提交
358 359 360 361 362 363 364
 */
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
	int		loop_count = 0;
	int		res;

365
	/* Always search-by-index from the beginning of the cache */
366
	if (*desc->dir_cookie == 0) {
C
Chuck Lever 已提交
367 368
		dfprintk(DIRCACHE, "NFS: readdir_search_pagecache() searching for offset %Ld\n",
				(long long)desc->file->f_pos);
369 370 371 372
		desc->page_index = 0;
		desc->entry->cookie = desc->entry->prev_cookie = 0;
		desc->entry->eof = 0;
		desc->current_index = 0;
373
	} else
C
Chuck Lever 已提交
374 375
		dfprintk(DIRCACHE, "NFS: readdir_search_pagecache() searching for cookie %Lu\n",
				(unsigned long long)*desc->dir_cookie);
376

L
Linus Torvalds 已提交
377 378 379 380 381 382 383 384 385 386 387
	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
388 389

	dfprintk(DIRCACHE, "NFS: %s: returns %d\n", __FUNCTION__, res);
L
Linus Torvalds 已提交
390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
	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;
	unsigned long	fileid;
	int		loop_count = 0,
			res;

C
Chuck Lever 已提交
414 415
	dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling starting @ cookie %Lu\n",
			(unsigned long long)entry->cookie);
L
Linus Torvalds 已提交
416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434

	for(;;) {
		unsigned d_type = DT_UNKNOWN;
		/* Note: entry->prev_cookie contains the cookie for
		 *	 retrieving the current dirent on the server */
		fileid = nfs_fileid_to_ino_t(entry->ino);

		/* 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);
			fileid = dentry->d_inode->i_ino;
		}

		res = filldir(dirent, entry->name, entry->len, 
435
			      file->f_pos, fileid, d_type);
L
Linus Torvalds 已提交
436 437
		if (res < 0)
			break;
438
		file->f_pos++;
439
		*desc->dir_cookie = entry->cookie;
L
Linus Torvalds 已提交
440 441 442 443 444 445 446 447 448 449 450 451
		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
452 453
	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
454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
	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;
474
	struct inode	*inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
475 476 477
	struct rpc_cred	*cred = nfs_file_cred(file);
	struct page	*page = NULL;
	int		status;
478
	unsigned long	timestamp;
L
Linus Torvalds 已提交
479

C
Chuck Lever 已提交
480 481
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
482 483 484 485 486 487

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
488
	timestamp = jiffies;
489
	desc->error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, *desc->dir_cookie,
L
Linus Torvalds 已提交
490 491 492
						page,
						NFS_SERVER(inode)->dtsize,
						desc->plus);
493
	spin_lock(&inode->i_lock);
494
	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATIME;
495
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
496 497 498
	desc->page = page;
	desc->ptr = kmap(page);		/* matching kunmap in nfs_do_filldir */
	if (desc->error >= 0) {
499 500
		desc->timestamp = timestamp;
		desc->timestamp_valid = 1;
L
Linus Torvalds 已提交
501
		if ((status = dir_decode(desc)) == 0)
502
			desc->entry->prev_cookie = *desc->dir_cookie;
L
Linus Torvalds 已提交
503 504 505 506 507 508 509 510 511 512 513 514 515
	} 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
516 517
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
			__FUNCTION__, status);
L
Linus Torvalds 已提交
518 519 520 521 522 523
	return status;
 out_release:
	dir_page_release(desc);
	goto out;
}

524 525 526
/* 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
527 528 529
 */
static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
530
	struct dentry	*dentry = filp->f_path.dentry;
L
Linus Torvalds 已提交
531 532 533 534 535 536 537 538
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
	struct nfs_entry my_entry;
	struct nfs_fh	 fh;
	struct nfs_fattr fattr;
	long		res;

C
Chuck Lever 已提交
539 540 541
	dfprintk(VFS, "NFS: readdir(%s/%s) starting at cookie %Lu\n",
			dentry->d_parent->d_name.name, dentry->d_name.name,
			(long long)filp->f_pos);
C
Chuck Lever 已提交
542 543
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
544 545
	lock_kernel();

546
	res = nfs_revalidate_mapping_nolock(inode, filp->f_mapping);
L
Linus Torvalds 已提交
547 548 549 550 551 552
	if (res < 0) {
		unlock_kernel();
		return res;
	}

	/*
553
	 * filp->f_pos points to the dirent entry number.
554
	 * *desc->dir_cookie has the cookie for the next entry. We have
555 556
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
557 558 559 560
	 */
	memset(desc, 0, sizeof(*desc));

	desc->file = filp;
561
	desc->dir_cookie = &((struct nfs_open_context *)filp->private_data)->dir_cookie;
L
Linus Torvalds 已提交
562 563 564 565 566 567 568
	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;
	my_entry.fh = &fh;
	my_entry.fattr = &fattr;
569
	nfs_fattr_init(&fattr);
L
Linus Torvalds 已提交
570 571 572 573
	desc->entry = &my_entry;

	while(!desc->entry->eof) {
		res = readdir_search_pagecache(desc);
574

L
Linus Torvalds 已提交
575 576
		if (res == -EBADCOOKIE) {
			/* This means either end of directory */
577
			if (*desc->dir_cookie && desc->entry->cookie != *desc->dir_cookie) {
L
Linus Torvalds 已提交
578 579 580 581 582 583 584 585 586
				/* 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) {
587
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
L
Linus Torvalds 已提交
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
			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;
		}
	}
	unlock_kernel();
C
Chuck Lever 已提交
603 604 605 606 607 608
	if (res > 0)
		res = 0;
	dfprintk(VFS, "NFS: readdir(%s/%s) returns %ld\n",
			dentry->d_parent->d_name.name, dentry->d_name.name,
			res);
	return res;
L
Linus Torvalds 已提交
609 610
}

611 612
loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
{
613
	mutex_lock(&filp->f_path.dentry->d_inode->i_mutex);
614 615 616 617 618 619 620 621 622 623 624 625 626 627 628
	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;
		((struct nfs_open_context *)filp->private_data)->dir_cookie = 0;
	}
out:
629
	mutex_unlock(&filp->f_path.dentry->d_inode->i_mutex);
630 631 632
	return offset;
}

L
Linus Torvalds 已提交
633 634 635 636 637 638
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
int nfs_fsync_dir(struct file *filp, struct dentry *dentry, int datasync)
{
C
Chuck Lever 已提交
639 640 641 642
	dfprintk(VFS, "NFS: fsync_dir(%s/%s) datasync %d\n",
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);

L
Linus Torvalds 已提交
643 644 645 646 647 648 649 650
	return 0;
}

/*
 * 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.
 */
651
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
652
{
653 654
	unsigned long verf;

L
Linus Torvalds 已提交
655 656
	if (IS_ROOT(dentry))
		return 1;
657
	verf = (unsigned long)dentry->d_fsdata;
658
	if (nfs_caches_unstable(dir)
659
			|| verf != NFS_I(dir)->cache_change_attribute)
L
Linus Torvalds 已提交
660
		return 0;
661
	return 1;
L
Linus Torvalds 已提交
662 663 664 665 666 667 668
}

static inline void nfs_set_verifier(struct dentry * dentry, unsigned long verf)
{
	dentry->d_fsdata = (void *)verf;
}

669 670
static void nfs_refresh_verifier(struct dentry * dentry, unsigned long verf)
{
671
	nfs_set_verifier(dentry, verf);
672 673
}

L
Linus Torvalds 已提交
674 675 676 677 678 679 680 681 682
/*
 * Whenever an NFS operation succeeds, we know that the dentry
 * is valid, so we update the revalidation timestamp.
 */
static inline void nfs_renew_times(struct dentry * dentry)
{
	dentry->d_time = jiffies;
}

683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704
/*
 * 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;
}

/*
 * 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
705 706 707 708 709 710 711
static inline
int nfs_lookup_verify_inode(struct inode *inode, struct nameidata *nd)
{
	struct nfs_server *server = NFS_SERVER(inode);

	if (nd != NULL) {
		/* VFS wants an on-the-wire revalidation */
712
		if (nd->flags & LOOKUP_REVAL)
L
Linus Torvalds 已提交
713 714
			goto out_force;
		/* This is an open(2) */
715
		if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 &&
716 717 718
				!(server->flags & NFS_MOUNT_NOCTO) &&
				(S_ISREG(inode->i_mode) ||
				 S_ISDIR(inode->i_mode)))
L
Linus Torvalds 已提交
719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737
			goto out_force;
	}
	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 */
738
	if (nd != NULL && nfs_lookup_check_intent(nd, LOOKUP_CREATE) != 0)
L
Linus Torvalds 已提交
739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766
		return 0;
	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;
	int error;
	struct nfs_fh fhandle;
	struct nfs_fattr fattr;
	unsigned long verifier;

	parent = dget_parent(dentry);
	lock_kernel();
	dir = parent->d_inode;
C
Chuck Lever 已提交
767
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
768 769
	inode = dentry->d_inode;

770 771 772 773
	/* Revalidate parent directory attribute cache */
	if (nfs_revalidate_inode(NFS_SERVER(dir), dir) < 0)
		goto out_zap_parent;

L
Linus Torvalds 已提交
774 775 776 777 778 779 780
	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
781 782 783
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
				__FUNCTION__, dentry->d_parent->d_name.name,
				dentry->d_name.name);
L
Linus Torvalds 已提交
784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806
		goto out_bad;
	}

	/* Force a full look up iff the parent directory has changed */
	if (nfs_check_verifier(dir, dentry)) {
		if (nfs_lookup_verify_inode(inode, nd))
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

	verifier = nfs_save_change_attribute(dir);
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, &fhandle, &fattr);
	if (error)
		goto out_bad;
	if (nfs_compare_fh(NFS_FH(inode), &fhandle))
		goto out_bad;
	if ((error = nfs_refresh_inode(inode, &fattr)) != 0)
		goto out_bad;

	nfs_renew_times(dentry);
807
	nfs_refresh_verifier(dentry, verifier);
L
Linus Torvalds 已提交
808 809 810
 out_valid:
	unlock_kernel();
	dput(parent);
C
Chuck Lever 已提交
811 812 813
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n",
			__FUNCTION__, dentry->d_parent->d_name.name,
			dentry->d_name.name);
L
Linus Torvalds 已提交
814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
	NFS_CACHEINV(dir);
	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;
		shrink_dcache_parent(dentry);
	}
	d_drop(dentry);
	unlock_kernel();
	dput(parent);
C
Chuck Lever 已提交
830 831 832
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
			__FUNCTION__, dentry->d_parent->d_name.name,
			dentry->d_name.name);
L
Linus Torvalds 已提交
833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863
	return 0;
}

/*
 * 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);

	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;

}

/*
 * 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)
{
864
	nfs_inode_return_delegation(inode);
865 866 867 868
	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
869 870
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
		lock_kernel();
871
		drop_nlink(inode);
L
Linus Torvalds 已提交
872 873 874 875 876 877 878 879 880 881 882 883 884 885
		nfs_complete_unlink(dentry);
		unlock_kernel();
	}
	/* When creating a negative dentry, we want to renew d_time */
	nfs_renew_times(dentry);
	iput(inode);
}

struct dentry_operations nfs_dentry_operations = {
	.d_revalidate	= nfs_lookup_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
};

886 887 888 889
/*
 * Use intent information to check whether or not we're going to do
 * an O_EXCL create using this path component.
 */
L
Linus Torvalds 已提交
890 891 892 893 894
static inline
int nfs_is_exclusive_create(struct inode *dir, struct nameidata *nd)
{
	if (NFS_PROTO(dir)->version == 2)
		return 0;
895
	if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_CREATE) == 0)
L
Linus Torvalds 已提交
896 897 898 899
		return 0;
	return (nd->intent.open.flags & O_EXCL) != 0;
}

900 901
static inline int nfs_reval_fsid(struct vfsmount *mnt, struct inode *dir,
				 struct nfs_fh *fh, struct nfs_fattr *fattr)
902 903 904 905 906
{
	struct nfs_server *server = NFS_SERVER(dir);

	if (!nfs_fsid_equal(&server->fsid, &fattr->fsid))
		/* Revalidate fsid on root dir */
907
		return __nfs_revalidate_inode(server, mnt->mnt_root->d_inode);
908 909 910
	return 0;
}

L
Linus Torvalds 已提交
911 912 913 914 915 916 917 918 919 920
static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
{
	struct dentry *res;
	struct inode *inode = NULL;
	int error;
	struct nfs_fh fhandle;
	struct nfs_fattr fattr;

	dfprintk(VFS, "NFS: lookup(%s/%s)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name);
C
Chuck Lever 已提交
921
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
922 923 924 925 926 927 928 929 930 931

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

	res = ERR_PTR(-ENOMEM);
	dentry->d_op = NFS_PROTO(dir)->dentry_ops;

	lock_kernel();

932 933 934 935 936 937 938 939 940
	/*
	 * 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;
		goto out_unlock;
	}
L
Linus Torvalds 已提交
941 942 943 944 945 946 947 948

	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, &fhandle, &fattr);
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
		goto out_unlock;
	}
949
	error = nfs_reval_fsid(nd->mnt, dir, &fhandle, &fattr);
950 951 952 953
	if (error < 0) {
		res = ERR_PTR(error);
		goto out_unlock;
	}
L
Linus Torvalds 已提交
954
	inode = nfs_fhget(dentry->d_sb, &fhandle, &fattr);
955 956
	res = (struct dentry *)inode;
	if (IS_ERR(res))
L
Linus Torvalds 已提交
957
		goto out_unlock;
958

L
Linus Torvalds 已提交
959
no_entry:
960
	res = d_materialise_unique(dentry, inode);
961
	if (res != NULL) {
962
		struct dentry *parent;
963 964
		if (IS_ERR(res))
			goto out_unlock;
965 966 967 968 969
		/* Was a directory renamed! */
		parent = dget_parent(res);
		if (!IS_ROOT(parent))
			nfs_mark_for_revalidate(parent->d_inode);
		dput(parent);
L
Linus Torvalds 已提交
970
		dentry = res;
971
	}
L
Linus Torvalds 已提交
972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988
	nfs_renew_times(dentry);
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
out_unlock:
	unlock_kernel();
out:
	return res;
}

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

struct dentry_operations nfs4_dentry_operations = {
	.d_revalidate	= nfs_open_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
};

989 990 991 992
/*
 * Use intent information to determine whether we need to substitute
 * the NFSv4-style stateful OPEN for the LOOKUP call
 */
L
Linus Torvalds 已提交
993 994
static int is_atomic_open(struct inode *dir, struct nameidata *nd)
{
995
	if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_OPEN) == 0)
L
Linus Torvalds 已提交
996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
		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? */
	if (IS_RDONLY(dir) && (nd->intent.open.flags & (O_CREAT|O_TRUNC|FMODE_WRITE)))
		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
1011 1012 1013
	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
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
	/* Check that we are indeed trying to open this file */
	if (!is_atomic_open(dir, nd))
		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;

	/* Let vfs_create() deal with O_EXCL */
1025 1026 1027 1028
	if (nd->intent.open.flags & O_EXCL) {
		d_add(dentry, NULL);
		goto out;
	}
L
Linus Torvalds 已提交
1029 1030 1031 1032 1033 1034 1035

	/* Open the file on the server */
	lock_kernel();
	/* Revalidate parent directory attribute cache */
	error = nfs_revalidate_inode(NFS_SERVER(dir), dir);
	if (error < 0) {
		res = ERR_PTR(error);
1036
		unlock_kernel();
L
Linus Torvalds 已提交
1037 1038 1039 1040 1041
		goto out;
	}

	if (nd->intent.open.flags & O_CREAT) {
		nfs_begin_data_update(dir);
1042
		res = nfs4_atomic_open(dir, dentry, nd);
L
Linus Torvalds 已提交
1043 1044
		nfs_end_data_update(dir);
	} else
1045
		res = nfs4_atomic_open(dir, dentry, nd);
L
Linus Torvalds 已提交
1046
	unlock_kernel();
1047 1048
	if (IS_ERR(res)) {
		error = PTR_ERR(res);
L
Linus Torvalds 已提交
1049 1050 1051
		switch (error) {
			/* Make a negative dentry */
			case -ENOENT:
1052 1053
				res = NULL;
				goto out;
L
Linus Torvalds 已提交
1054
			/* This turned out not to be a regular file */
1055 1056 1057
			case -EISDIR:
			case -ENOTDIR:
				goto no_open;
L
Linus Torvalds 已提交
1058 1059 1060 1061 1062 1063 1064
			case -ELOOP:
				if (!(nd->intent.open.flags & O_NOFOLLOW))
					goto no_open;
			/* case -EINVAL: */
			default:
				goto out;
		}
1065
	} else if (res != NULL)
L
Linus Torvalds 已提交
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102
		dentry = res;
	nfs_renew_times(dentry);
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
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;
	unsigned long verifier;
	int openflags, ret = 0;

	parent = dget_parent(dentry);
	dir = parent->d_inode;
	if (!is_atomic_open(dir, nd))
		goto no_open;
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
	if (inode == NULL)
		goto out;
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
		goto no_open;
	openflags = nd->intent.open.flags;
	/* We cannot do exclusive creation on a positive dentry */
	if ((openflags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL))
		goto no_open;
	/* We can't create new files, or truncate existing ones here */
	openflags &= ~(O_CREAT|O_TRUNC);

	/*
1103
	 * Note: we're not holding inode->i_mutex and so may be racing with
L
Linus Torvalds 已提交
1104 1105 1106 1107 1108
	 * operations that change the directory. We therefore save the
	 * change attribute *before* we do the RPC call.
	 */
	lock_kernel();
	verifier = nfs_save_change_attribute(dir);
1109
	ret = nfs4_open_revalidate(dir, dentry, openflags, nd);
L
Linus Torvalds 已提交
1110
	if (!ret)
1111
		nfs_refresh_verifier(dentry, verifier);
L
Linus Torvalds 已提交
1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
	unlock_kernel();
out:
	dput(parent);
	if (!ret)
		d_drop(dentry);
	return ret;
no_open:
	dput(parent);
	if (inode != NULL && nfs_have_delegation(inode, FMODE_READ))
		return 1;
	return nfs_lookup_revalidate(dentry, nd);
}
#endif /* CONFIG_NFSV4 */

static struct dentry *nfs_readdir_lookup(nfs_readdir_descriptor_t *desc)
{
1128
	struct dentry *parent = desc->file->f_path.dentry;
L
Linus Torvalds 已提交
1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
	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;

	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);
	}
	name.hash = full_name_hash(name.name, name.len);
	dentry = d_lookup(parent, &name);
1149
	if (dentry != NULL) {
1150 1151 1152 1153 1154 1155 1156 1157 1158 1159
		/* 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;
		}
1160 1161 1162 1163
		/* No, so d_drop to allow one to be created */
		d_drop(dentry);
		dput(dentry);
	}
L
Linus Torvalds 已提交
1164 1165
	if (!desc->plus || !(entry->fattr->valid & NFS_ATTR_FATTR))
		return NULL;
1166
	/* Note: caller is already holding the dir->i_mutex! */
L
Linus Torvalds 已提交
1167 1168 1169 1170 1171
	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);
1172
	if (IS_ERR(inode)) {
L
Linus Torvalds 已提交
1173 1174 1175
		dput(dentry);
		return NULL;
	}
1176 1177

	alias = d_materialise_unique(dentry, inode);
L
Linus Torvalds 已提交
1178 1179
	if (alias != NULL) {
		dput(dentry);
1180 1181
		if (IS_ERR(alias))
			return NULL;
L
Linus Torvalds 已提交
1182 1183
		dentry = alias;
	}
1184

L
Linus Torvalds 已提交
1185 1186 1187
	nfs_renew_times(dentry);
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	return dentry;
1188 1189 1190 1191
out_renew:
	nfs_renew_times(dentry);
	nfs_refresh_verifier(dentry, nfs_save_change_attribute(dir));
	return dentry;
L
Linus Torvalds 已提交
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209
}

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
				struct nfs_fattr *fattr)
{
	struct inode *inode;
	int error = -EACCES;

	/* We may have been initialized further down */
	if (dentry->d_inode)
		return 0;
	if (fhandle->size == 0) {
		struct inode *dir = dentry->d_parent->d_inode;
		error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
		if (error)
1210
			return error;
L
Linus Torvalds 已提交
1211 1212 1213
	}
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1214
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr);
L
Linus Torvalds 已提交
1215
		if (error < 0)
1216
			return error;
L
Linus Torvalds 已提交
1217 1218
	}
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
1219 1220
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1221
		return error;
L
Linus Torvalds 已提交
1222
	d_instantiate(dentry, inode);
1223 1224
	if (d_unhashed(dentry))
		d_rehash(dentry);
L
Linus Torvalds 已提交
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240
	return 0;
}

/*
 * 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
1241 1242
	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
1243 1244 1245 1246 1247 1248 1249 1250 1251

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

	if (nd && (nd->flags & LOOKUP_CREATE))
		open_flags = nd->intent.open.flags;

	lock_kernel();
	nfs_begin_data_update(dir);
1252
	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, nd);
L
Linus Torvalds 已提交
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
	nfs_end_data_update(dir);
	if (error != 0)
		goto out_err;
	nfs_renew_times(dentry);
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	unlock_kernel();
	return 0;
out_err:
	unlock_kernel();
	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
1275 1276
	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
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307

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

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

	lock_kernel();
	nfs_begin_data_update(dir);
	status = NFS_PROTO(dir)->mknod(dir, dentry, &attr, rdev);
	nfs_end_data_update(dir);
	if (status != 0)
		goto out_err;
	nfs_renew_times(dentry);
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	unlock_kernel();
	return 0;
out_err:
	unlock_kernel();
	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
1308 1309
	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
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333

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

	lock_kernel();
	nfs_begin_data_update(dir);
	error = NFS_PROTO(dir)->mkdir(dir, dentry, &attr);
	nfs_end_data_update(dir);
	if (error != 0)
		goto out_err;
	nfs_renew_times(dentry);
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	unlock_kernel();
	return 0;
out_err:
	d_drop(dentry);
	unlock_kernel();
	return error;
}

static int nfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	int error;

C
Chuck Lever 已提交
1334 1335
	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
1336 1337 1338 1339 1340 1341

	lock_kernel();
	nfs_begin_data_update(dir);
	error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
	/* Ensure the VFS deletes this inode */
	if (error == 0 && dentry->d_inode != NULL)
1342
		clear_nlink(dentry->d_inode);
L
Linus Torvalds 已提交
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362
	nfs_end_data_update(dir);
	unlock_kernel();

	return error;
}

static int nfs_sillyrename(struct inode *dir, struct dentry *dentry)
{
	static unsigned int sillycounter;
	const int      i_inosize  = sizeof(dir->i_ino)*2;
	const int      countersize = sizeof(sillycounter)*2;
	const int      slen       = sizeof(".nfs") + i_inosize + countersize - 1;
	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
1363
	nfs_inc_stats(dir, NFSIOS_SILLYRENAME);
L
Linus Torvalds 已提交
1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379

#ifdef NFS_PARANOIA
if (!dentry->d_inode)
printk("NFS: silly-renaming %s/%s, negative dentry??\n",
dentry->d_parent->d_name.name, dentry->d_name.name);
#endif
	/*
	 * We don't allow a dentry to be silly-renamed twice.
	 */
	error = -EBUSY;
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
		goto out;

	sprintf(silly, ".nfs%*.*lx",
		i_inosize, i_inosize, dentry->d_inode->i_ino);

1380 1381 1382
	/* Return delegation in anticipation of the rename */
	nfs_inode_return_delegation(dentry->d_inode);

L
Linus Torvalds 已提交
1383 1384 1385 1386 1387 1388 1389 1390
	sdentry = NULL;
	do {
		char *suffix = silly + slen - countersize;

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

C
Chuck Lever 已提交
1391 1392
		dfprintk(VFS, "NFS: trying to rename %s to %s\n",
				dentry->d_name.name, silly);
L
Linus Torvalds 已提交
1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409
		
		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);
	nfs_begin_data_update(dir);
	if (dentry->d_inode) {
		nfs_begin_data_update(dentry->d_inode);
		error = NFS_PROTO(dir)->rename(dir, &dentry->d_name,
				dir, &qsilly);
1410
		nfs_mark_for_revalidate(dentry->d_inode);
L
Linus Torvalds 已提交
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 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451
		nfs_end_data_update(dentry->d_inode);
	} else
		error = NFS_PROTO(dir)->rename(dir, &dentry->d_name,
				dir, &qsilly);
	nfs_end_data_update(dir);
	if (!error) {
		nfs_renew_times(dentry);
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
		d_move(dentry, sdentry);
		error = nfs_async_unlink(dentry);
 		/* 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;
	}

	nfs_begin_data_update(dir);
	if (inode != NULL) {
1452
		nfs_inode_return_delegation(inode);
L
Linus Torvalds 已提交
1453 1454 1455 1456
		nfs_begin_data_update(inode);
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		/* The VFS may want to delete this inode */
		if (error == 0)
1457
			drop_nlink(inode);
1458
		nfs_mark_for_revalidate(inode);
L
Linus Torvalds 已提交
1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485
		nfs_end_data_update(inode);
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
	nfs_end_data_update(dir);
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);

	lock_kernel();
	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
1486 1487
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507
		error = nfs_sillyrename(dir, dentry);
		unlock_kernel();
		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);
	if (!error) {
		nfs_renew_times(dentry);
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
	unlock_kernel();
	return error;
}

1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523
/*
 * 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
1524
{
1525 1526 1527
	struct pagevec lru_pvec;
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1528
	struct iattr attr;
1529
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1530 1531 1532 1533 1534
	int error;

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

1535 1536
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1537

1538 1539
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1540 1541

	lock_kernel();
1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554

	page = alloc_page(GFP_KERNEL);
	if (!page) {
		unlock_kernel();
		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);

L
Linus Torvalds 已提交
1555
	nfs_begin_data_update(dir);
1556
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
L
Linus Torvalds 已提交
1557
	nfs_end_data_update(dir);
1558 1559 1560 1561
	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
1562
		d_drop(dentry);
1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574
		__free_page(page);
		unlock_kernel();
		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)) {
1575 1576
		pagevec_add(&lru_pvec, page);
		pagevec_lru_add(&lru_pvec);
1577 1578 1579 1580 1581
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

L
Linus Torvalds 已提交
1582
	unlock_kernel();
1583
	return 0;
L
Linus Torvalds 已提交
1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599
}

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);

	lock_kernel();
	nfs_begin_data_update(dir);
	nfs_begin_data_update(inode);
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1600 1601 1602 1603
	if (error == 0) {
		atomic_inc(&inode->i_count);
		d_instantiate(dentry, inode);
	}
L
Linus Torvalds 已提交
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666
	nfs_end_data_update(inode);
	nfs_end_data_update(dir);
	unlock_kernel();
	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;

	/*
	 * To prevent any new references to the target during the rename,
	 * we unhash the dentry and free the inode in advance.
	 */
	lock_kernel();
	if (!d_unhashed(new_dentry)) {
		d_drop(new_dentry);
		rehash = new_dentry;
	}

	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));

	/*
	 * First check whether the target is busy ... we can't
	 * safely do _any_ rename if the target is in use.
	 *
	 * For files, 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.
	 */
	if (!new_inode)
		goto go_ahead;
1667 1668 1669 1670 1671
	if (S_ISDIR(new_inode->i_mode)) {
		error = -EISDIR;
		if (!S_ISDIR(old_inode->i_mode))
			goto out;
	} else if (atomic_read(&new_dentry->d_count) > 2) {
L
Linus Torvalds 已提交
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695
		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);
		if (!err) {
			new_dentry = rehash = dentry;
			new_inode = NULL;
			/* instantiate the replacement target */
			d_instantiate(new_dentry, NULL);
		} else if (atomic_read(&new_dentry->d_count) > 1) {
		/* dentry still busy? */
#ifdef NFS_PARANOIA
			printk("nfs_rename: target %s/%s busy, d_count=%d\n",
			       new_dentry->d_parent->d_name.name,
			       new_dentry->d_name.name,
			       atomic_read(&new_dentry->d_count));
#endif
			goto out;
		}
T
Trond Myklebust 已提交
1696
	} else
1697
		drop_nlink(new_inode);
L
Linus Torvalds 已提交
1698 1699 1700 1701 1702 1703

go_ahead:
	/*
	 * ... prune child dentries and writebacks if needed.
	 */
	if (atomic_read(&old_dentry->d_count) > 1) {
1704 1705
		if (S_ISREG(old_inode->i_mode))
			nfs_wb_all(old_inode);
L
Linus Torvalds 已提交
1706 1707
		shrink_dcache_parent(old_dentry);
	}
1708
	nfs_inode_return_delegation(old_inode);
L
Linus Torvalds 已提交
1709

1710 1711
	if (new_inode != NULL) {
		nfs_inode_return_delegation(new_inode);
L
Linus Torvalds 已提交
1712
		d_delete(new_dentry);
1713
	}
L
Linus Torvalds 已提交
1714 1715 1716 1717 1718 1719

	nfs_begin_data_update(old_dir);
	nfs_begin_data_update(new_dir);
	nfs_begin_data_update(old_inode);
	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
1720
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
1721 1722 1723 1724 1725 1726 1727
	nfs_end_data_update(old_inode);
	nfs_end_data_update(new_dir);
	nfs_end_data_update(old_dir);
out:
	if (rehash)
		d_rehash(rehash);
	if (!error) {
1728
		d_move(old_dentry, new_dentry);
L
Linus Torvalds 已提交
1729
		nfs_renew_times(new_dentry);
1730
		nfs_refresh_verifier(new_dentry, nfs_save_change_attribute(new_dir));
L
Linus Torvalds 已提交
1731 1732 1733 1734 1735 1736 1737 1738 1739
	}

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

1740 1741 1742 1743
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

1744 1745 1746 1747
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
1748 1749 1750
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
1751 1752
}

1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
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;

	spin_lock(&nfs_access_lru_lock);
restart:
	list_for_each_entry(nfsi, &nfs_access_lru_list, access_cache_inode_lru) {
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
		inode = igrab(&nfsi->vfs_inode);
		if (inode == NULL)
			continue;
		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);
		iput(inode);
		goto restart;
	}
	spin_unlock(&nfs_access_lru_lock);
	while (!list_empty(&head)) {
		cache = list_entry(head.next, struct nfs_access_entry, lru);
		list_del(&cache->lru);
		nfs_access_free_entry(cache);
	}
	return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure;
}

1797
static void __nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
1798
{
1799
	struct nfs_inode *nfsi = NFS_I(inode);
1800 1801 1802 1803 1804 1805 1806 1807
	struct rb_root *root_node = &nfsi->access_cache;
	struct rb_node *n, *dispose = NULL;
	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);
1808
		list_del(&entry->lru);
1809 1810 1811 1812 1813
		n->rb_left = dispose;
		dispose = n;
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1814

1815 1816 1817 1818 1819 1820
	/* Now kill them all! */
	while (dispose != NULL) {
		n = dispose;
		dispose = n->rb_left;
		nfs_access_free_entry(rb_entry(n, struct nfs_access_entry, rb_node));
	}
L
Linus Torvalds 已提交
1821 1822
}

1823
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
1824
{
1825 1826 1827 1828 1829 1830 1831
	/* Remove from global LRU init */
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_FLAGS(inode))) {
		spin_lock(&nfs_access_lru_lock);
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);
		spin_unlock(&nfs_access_lru_lock);
	}

1832 1833 1834 1835
	spin_lock(&inode->i_lock);
	/* This will release the spinlock */
	__nfs_access_zap_cache(inode);
}
L
Linus Torvalds 已提交
1836

1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850
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
1851
	}
1852 1853 1854 1855 1856 1857 1858 1859 1860
	return NULL;
}

int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

1861
	spin_lock(&inode->i_lock);
1862 1863 1864 1865 1866 1867 1868 1869 1870 1871
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
	if (time_after(jiffies, cache->jiffies + NFS_ATTRTIMEO(inode)))
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
1872
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
1873 1874 1875 1876 1877 1878
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
1879
	list_del(&cache->lru);
1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
	/* This will release the spinlock */
	__nfs_access_zap_cache(inode);
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
1891 1892
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
	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);
1911
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
1912
	spin_unlock(&inode->i_lock);
1913 1914 1915
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
1916 1917
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
{
	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
1928
	cache->jiffies = set->jiffies;
1929
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
1930
	cache->mask = set->mask;
1931 1932

	nfs_access_add_rbtree(inode, cache);
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944

	/* 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 */
	if (!test_and_set_bit(NFS_INO_ACL_LRU_SET, &NFS_FLAGS(inode))) {
		spin_lock(&nfs_access_lru_lock);
		list_add_tail(&NFS_I(inode)->access_cache_inode_lru, &nfs_access_lru_list);
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974
}

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);
	if (status != 0)
		return status;
	nfs_access_add_cache(inode, &cache);
out:
	if ((cache.mask & mask) == mask)
		return 0;
	return -EACCES;
}

int nfs_permission(struct inode *inode, int mask, struct nameidata *nd)
{
	struct rpc_cred *cred;
	int res = 0;

C
Chuck Lever 已提交
1975 1976
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

L
Linus Torvalds 已提交
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
	if (mask == 0)
		goto out;
	/* Is this sys_access() ? */
	if (nd != NULL && (nd->flags & LOOKUP_ACCESS))
		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)
					&& nd != NULL
					&& (nd->flags & LOOKUP_OPEN))
				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:
	lock_kernel();

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

	cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
	unlock_kernel();
out:
C
Chuck Lever 已提交
2016 2017
	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
2018 2019 2020 2021 2022 2023
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
		res = generic_permission(inode, mask, NULL);
	unlock_kernel();
C
Chuck Lever 已提交
2024
	goto out;
L
Linus Torvalds 已提交
2025 2026 2027 2028 2029 2030 2031 2032
}

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