dir.c 58.9 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>
36
#include <linux/kmemleak.h>
37
#include <linux/xattr.h>
L
Linus Torvalds 已提交
38 39

#include "delegation.h"
C
Chuck Lever 已提交
40
#include "iostat.h"
41
#include "internal.h"
42
#include "fscache.h"
L
Linus Torvalds 已提交
43 44 45 46

/* #define NFS_DEBUG_VERBOSE 1 */

static int nfs_opendir(struct inode *, struct file *);
47
static int nfs_closedir(struct inode *, struct file *);
L
Linus Torvalds 已提交
48 49
static int nfs_readdir(struct file *, void *, filldir_t);
static struct dentry *nfs_lookup(struct inode *, struct dentry *, struct nameidata *);
A
Al Viro 已提交
50
static int nfs_create(struct inode *, struct dentry *, umode_t, struct nameidata *);
51
static int nfs_mkdir(struct inode *, struct dentry *, umode_t);
L
Linus Torvalds 已提交
52 53 54 55
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 *);
A
Al Viro 已提交
56
static int nfs_mknod(struct inode *, struct dentry *, umode_t, dev_t);
L
Linus Torvalds 已提交
57 58
static int nfs_rename(struct inode *, struct dentry *,
		      struct inode *, struct dentry *);
59
static int nfs_fsync_dir(struct file *, loff_t, loff_t, int);
60
static loff_t nfs_llseek_dir(struct file *, loff_t, int);
61
static void nfs_readdir_clear_array(struct page*);
L
Linus Torvalds 已提交
62

63
const struct file_operations nfs_dir_operations = {
64
	.llseek		= nfs_llseek_dir,
L
Linus Torvalds 已提交
65 66 67
	.read		= generic_read_dir,
	.readdir	= nfs_readdir,
	.open		= nfs_opendir,
68
	.release	= nfs_closedir,
L
Linus Torvalds 已提交
69 70 71
	.fsync		= nfs_fsync_dir,
};

72
const struct inode_operations nfs_dir_inode_operations = {
L
Linus Torvalds 已提交
73 74 75 76 77 78 79 80 81 82 83 84 85 86
	.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,
};

87 88
const struct address_space_operations nfs_dir_aops = {
	.freepage = nfs_readdir_clear_array,
B
Bryan Schumaker 已提交
89 90
};

91
#ifdef CONFIG_NFS_V3
92
const struct inode_operations nfs3_dir_inode_operations = {
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111
	.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
112 113
#ifdef CONFIG_NFS_V4

114 115 116 117 118
static struct file *nfs_atomic_open(struct inode *, struct dentry *,
				    struct opendata *, unsigned, umode_t,
				    bool *);
static int nfs4_create(struct inode *dir, struct dentry *dentry,
		       umode_t mode, struct nameidata *nd);
119
const struct inode_operations nfs4_dir_inode_operations = {
120 121 122
	.create		= nfs4_create,
	.lookup		= nfs_lookup,
	.atomic_open	= nfs_atomic_open,
L
Linus Torvalds 已提交
123 124 125 126 127 128 129 130 131 132
	.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,
133 134 135 136
	.getxattr	= generic_getxattr,
	.setxattr	= generic_setxattr,
	.listxattr	= generic_listxattr,
	.removexattr	= generic_removexattr,
L
Linus Torvalds 已提交
137 138 139 140
};

#endif /* CONFIG_NFS_V4 */

141
static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct inode *dir, struct rpc_cred *cred)
142 143 144 145
{
	struct nfs_open_dir_context *ctx;
	ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
	if (ctx != NULL) {
146
		ctx->duped = 0;
147
		ctx->attr_gencount = NFS_I(dir)->attr_gencount;
148
		ctx->dir_cookie = 0;
149
		ctx->dup_cookie = 0;
150
		ctx->cred = get_rpccred(cred);
151 152 153
		return ctx;
	}
	return  ERR_PTR(-ENOMEM);
154 155 156 157 158 159 160 161
}

static void put_nfs_open_dir_context(struct nfs_open_dir_context *ctx)
{
	put_rpccred(ctx->cred);
	kfree(ctx);
}

L
Linus Torvalds 已提交
162 163 164 165 166 167
/*
 * Open file
 */
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
168 169 170
	int res = 0;
	struct nfs_open_dir_context *ctx;
	struct rpc_cred *cred;
L
Linus Torvalds 已提交
171

C
Chuck Lever 已提交
172
	dfprintk(FILE, "NFS: open dir(%s/%s)\n",
173 174 175 176
			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
177

178 179 180
	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return PTR_ERR(cred);
181
	ctx = alloc_nfs_open_dir_context(inode, cred);
182 183 184 185 186
	if (IS_ERR(ctx)) {
		res = PTR_ERR(ctx);
		goto out;
	}
	filp->private_data = ctx;
187 188 189 190 191 192 193
	if (filp->f_path.dentry == filp->f_path.mnt->mnt_root) {
		/* This is a mountpoint, so d_revalidate will never
		 * have been called, so we need to refresh the
		 * inode (for close-open consistency) ourselves.
		 */
		__nfs_revalidate_inode(NFS_SERVER(inode), inode);
	}
194 195
out:
	put_rpccred(cred);
L
Linus Torvalds 已提交
196 197 198
	return res;
}

199 200 201 202 203 204 205
static int
nfs_closedir(struct inode *inode, struct file *filp)
{
	put_nfs_open_dir_context(filp->private_data);
	return 0;
}

B
Bryan Schumaker 已提交
206 207 208 209
struct nfs_cache_array_entry {
	u64 cookie;
	u64 ino;
	struct qstr string;
210
	unsigned char d_type;
B
Bryan Schumaker 已提交
211 212 213
};

struct nfs_cache_array {
214
	int size;
B
Bryan Schumaker 已提交
215 216 217 218 219
	int eof_index;
	u64 last_cookie;
	struct nfs_cache_array_entry array[0];
};

220
typedef int (*decode_dirent_t)(struct xdr_stream *, struct nfs_entry *, int);
L
Linus Torvalds 已提交
221 222 223 224
typedef struct {
	struct file	*file;
	struct page	*page;
	unsigned long	page_index;
225
	u64		*dir_cookie;
226
	u64		last_cookie;
227
	loff_t		current_index;
L
Linus Torvalds 已提交
228
	decode_dirent_t	decode;
B
Bryan Schumaker 已提交
229

230
	unsigned long	timestamp;
231
	unsigned long	gencount;
B
Bryan Schumaker 已提交
232 233 234
	unsigned int	cache_entry_index;
	unsigned int	plus:1;
	unsigned int	eof:1;
L
Linus Torvalds 已提交
235 236
} nfs_readdir_descriptor_t;

B
Bryan Schumaker 已提交
237 238
/*
 * The caller is responsible for calling nfs_readdir_release_array(page)
L
Linus Torvalds 已提交
239 240
 */
static
B
Bryan Schumaker 已提交
241 242
struct nfs_cache_array *nfs_readdir_get_array(struct page *page)
{
243
	void *ptr;
B
Bryan Schumaker 已提交
244 245
	if (page == NULL)
		return ERR_PTR(-EIO);
246 247 248 249
	ptr = kmap(page);
	if (ptr == NULL)
		return ERR_PTR(-ENOMEM);
	return ptr;
B
Bryan Schumaker 已提交
250 251 252 253 254 255 256 257 258 259 260 261
}

static
void nfs_readdir_release_array(struct page *page)
{
	kunmap(page);
}

/*
 * we are freeing strings created by nfs_add_to_readdir_array()
 */
static
262
void nfs_readdir_clear_array(struct page *page)
B
Bryan Schumaker 已提交
263
{
264
	struct nfs_cache_array *array;
B
Bryan Schumaker 已提交
265
	int i;
266

267
	array = kmap_atomic(page);
B
Bryan Schumaker 已提交
268 269
	for (i = 0; i < array->size; i++)
		kfree(array->array[i].string.name);
270
	kunmap_atomic(array);
B
Bryan Schumaker 已提交
271 272 273 274 275 276 277 278
}

/*
 * the caller is responsible for freeing qstr.name
 * when called by nfs_readdir_add_to_array, the strings will be freed in
 * nfs_clear_readdir_array()
 */
static
279
int nfs_readdir_make_qstr(struct qstr *string, const char *name, unsigned int len)
B
Bryan Schumaker 已提交
280 281 282
{
	string->len = len;
	string->name = kmemdup(name, len, GFP_KERNEL);
283 284
	if (string->name == NULL)
		return -ENOMEM;
285 286 287 288 289
	/*
	 * Avoid a kmemleak false positive. The pointer to the name is stored
	 * in a page cache page which kmemleak does not scan.
	 */
	kmemleak_not_leak(string->name);
290 291
	string->hash = full_name_hash(name, len);
	return 0;
B
Bryan Schumaker 已提交
292 293 294 295 296 297
}

static
int nfs_readdir_add_to_array(struct nfs_entry *entry, struct page *page)
{
	struct nfs_cache_array *array = nfs_readdir_get_array(page);
298 299 300
	struct nfs_cache_array_entry *cache_entry;
	int ret;

B
Bryan Schumaker 已提交
301 302
	if (IS_ERR(array))
		return PTR_ERR(array);
303 304 305 306

	cache_entry = &array->array[array->size];

	/* Check that this entry lies within the page bounds */
307
	ret = -ENOSPC;
308
	if ((char *)&cache_entry[1] - (char *)page_address(page) > PAGE_SIZE)
309
		goto out;
B
Bryan Schumaker 已提交
310

311 312
	cache_entry->cookie = entry->prev_cookie;
	cache_entry->ino = entry->ino;
313
	cache_entry->d_type = entry->d_type;
314 315 316
	ret = nfs_readdir_make_qstr(&cache_entry->string, entry->name, entry->len);
	if (ret)
		goto out;
B
Bryan Schumaker 已提交
317
	array->last_cookie = entry->cookie;
318
	array->size++;
T
Trond Myklebust 已提交
319
	if (entry->eof != 0)
B
Bryan Schumaker 已提交
320
		array->eof_index = array->size;
321
out:
B
Bryan Schumaker 已提交
322
	nfs_readdir_release_array(page);
323
	return ret;
B
Bryan Schumaker 已提交
324 325 326 327 328 329 330 331 332 333 334
}

static
int nfs_readdir_search_for_pos(struct nfs_cache_array *array, nfs_readdir_descriptor_t *desc)
{
	loff_t diff = desc->file->f_pos - desc->current_index;
	unsigned int index;

	if (diff < 0)
		goto out_eof;
	if (diff >= array->size) {
335
		if (array->eof_index >= 0)
B
Bryan Schumaker 已提交
336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352
			goto out_eof;
		return -EAGAIN;
	}

	index = (unsigned int)diff;
	*desc->dir_cookie = array->array[index].cookie;
	desc->cache_entry_index = index;
	return 0;
out_eof:
	desc->eof = 1;
	return -EBADCOOKIE;
}

static
int nfs_readdir_search_for_cookie(struct nfs_cache_array *array, nfs_readdir_descriptor_t *desc)
{
	int i;
353
	loff_t new_pos;
B
Bryan Schumaker 已提交
354 355 356 357
	int status = -EAGAIN;

	for (i = 0; i < array->size; i++) {
		if (array->array[i].cookie == *desc->dir_cookie) {
358 359 360
			struct nfs_inode *nfsi = NFS_I(desc->file->f_path.dentry->d_inode);
			struct nfs_open_dir_context *ctx = desc->file->private_data;

361
			new_pos = desc->current_index + i;
362 363 364 365 366 367 368 369 370 371
			if (ctx->attr_gencount != nfsi->attr_gencount
			    || (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))) {
				ctx->duped = 0;
				ctx->attr_gencount = nfsi->attr_gencount;
			} else if (new_pos < desc->file->f_pos) {
				if (ctx->duped > 0
				    && ctx->dup_cookie == *desc->dir_cookie) {
					if (printk_ratelimit()) {
						pr_notice("NFS: directory %s/%s contains a readdir loop."
								"Please contact your server vendor.  "
372
								"The file: %s has duplicate cookie %llu\n",
373 374
								desc->file->f_dentry->d_parent->d_name.name,
								desc->file->f_dentry->d_name.name,
375
								array->array[i].string.name,
376 377 378 379 380
								*desc->dir_cookie);
					}
					status = -ELOOP;
					goto out;
				}
381
				ctx->dup_cookie = *desc->dir_cookie;
382
				ctx->duped = -1;
383 384
			}
			desc->file->f_pos = new_pos;
B
Bryan Schumaker 已提交
385
			desc->cache_entry_index = i;
T
Trond Myklebust 已提交
386
			return 0;
B
Bryan Schumaker 已提交
387 388
		}
	}
T
Trond Myklebust 已提交
389
	if (array->eof_index >= 0) {
390
		status = -EBADCOOKIE;
391 392
		if (*desc->dir_cookie == array->last_cookie)
			desc->eof = 1;
393
	}
394
out:
B
Bryan Schumaker 已提交
395 396 397 398 399 400 401
	return status;
}

static
int nfs_readdir_search_array(nfs_readdir_descriptor_t *desc)
{
	struct nfs_cache_array *array;
T
Trond Myklebust 已提交
402
	int status;
B
Bryan Schumaker 已提交
403 404 405 406 407 408 409 410 411 412 413 414

	array = nfs_readdir_get_array(desc->page);
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
		goto out;
	}

	if (*desc->dir_cookie == 0)
		status = nfs_readdir_search_for_pos(array, desc);
	else
		status = nfs_readdir_search_for_cookie(array, desc);

T
Trond Myklebust 已提交
415
	if (status == -EAGAIN) {
416
		desc->last_cookie = array->last_cookie;
417
		desc->current_index += array->size;
T
Trond Myklebust 已提交
418 419
		desc->page_index++;
	}
B
Bryan Schumaker 已提交
420 421 422 423 424 425 426
	nfs_readdir_release_array(desc->page);
out:
	return status;
}

/* Fill a page with xdr information before transferring to the cache page */
static
B
Bryan Schumaker 已提交
427
int nfs_readdir_xdr_filler(struct page **pages, nfs_readdir_descriptor_t *desc,
B
Bryan Schumaker 已提交
428
			struct nfs_entry *entry, struct file *file, struct inode *inode)
L
Linus Torvalds 已提交
429
{
430 431
	struct nfs_open_dir_context *ctx = file->private_data;
	struct rpc_cred	*cred = ctx->cred;
432
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
433 434 435 436
	int		error;

 again:
	timestamp = jiffies;
437
	gencount = nfs_inc_attr_generation_counter();
B
Bryan Schumaker 已提交
438
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, entry->cookie, pages,
L
Linus Torvalds 已提交
439 440 441 442 443
					  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
444
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
445 446 447 448 449
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
450
	desc->timestamp = timestamp;
451
	desc->gencount = gencount;
B
Bryan Schumaker 已提交
452 453
error:
	return error;
L
Linus Torvalds 已提交
454 455
}

456 457
static int xdr_decode(nfs_readdir_descriptor_t *desc,
		      struct nfs_entry *entry, struct xdr_stream *xdr)
L
Linus Torvalds 已提交
458
{
459
	int error;
L
Linus Torvalds 已提交
460

461 462 463
	error = desc->decode(xdr, entry, desc->plus);
	if (error)
		return error;
B
Bryan Schumaker 已提交
464 465 466
	entry->fattr->time_start = desc->timestamp;
	entry->fattr->gencount = desc->gencount;
	return 0;
L
Linus Torvalds 已提交
467 468
}

B
Bryan Schumaker 已提交
469 470 471 472 473
static
int nfs_same_file(struct dentry *dentry, struct nfs_entry *entry)
{
	if (dentry->d_inode == NULL)
		goto different;
T
Trond Myklebust 已提交
474
	if (nfs_compare_fh(entry->fh, NFS_FH(dentry->d_inode)) != 0)
B
Bryan Schumaker 已提交
475 476 477 478 479 480
		goto different;
	return 1;
different:
	return 0;
}

481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503
static
bool nfs_use_readdirplus(struct inode *dir, struct file *filp)
{
	if (!nfs_server_capable(dir, NFS_CAP_READDIRPLUS))
		return false;
	if (test_and_clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(dir)->flags))
		return true;
	if (filp->f_pos == 0)
		return true;
	return false;
}

/*
 * This function is called by the lookup code to request the use of
 * readdirplus to accelerate any future lookups in the same
 * directory.
 */
static
void nfs_advise_use_readdirplus(struct inode *dir)
{
	set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(dir)->flags);
}

B
Bryan Schumaker 已提交
504 505 506
static
void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
{
507
	struct qstr filename = QSTR_INIT(entry->name, entry->len);
508 509
	struct dentry *dentry;
	struct dentry *alias;
B
Bryan Schumaker 已提交
510 511 512
	struct inode *dir = parent->d_inode;
	struct inode *inode;

513 514 515 516 517 518 519
	if (filename.name[0] == '.') {
		if (filename.len == 1)
			return;
		if (filename.len == 2 && filename.name[1] == '.')
			return;
	}
	filename.hash = full_name_hash(filename.name, filename.len);
B
Bryan Schumaker 已提交
520

521
	dentry = d_lookup(parent, &filename);
B
Bryan Schumaker 已提交
522 523 524 525 526 527 528 529 530 531 532
	if (dentry != NULL) {
		if (nfs_same_file(dentry, entry)) {
			nfs_refresh_inode(dentry->d_inode, entry->fattr);
			goto out;
		} else {
			d_drop(dentry);
			dput(dentry);
		}
	}

	dentry = d_alloc(parent, &filename);
533 534 535
	if (dentry == NULL)
		return;

B
Bryan Schumaker 已提交
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552
	inode = nfs_fhget(dentry->d_sb, entry->fh, entry->fattr);
	if (IS_ERR(inode))
		goto out;

	alias = d_materialise_unique(dentry, inode);
	if (IS_ERR(alias))
		goto out;
	else if (alias) {
		nfs_set_verifier(alias, nfs_save_change_attribute(dir));
		dput(alias);
	} else
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));

out:
	dput(dentry);
}

B
Bryan Schumaker 已提交
553 554
/* Perform conversion from xdr to cache array */
static
555
int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *entry,
556
				struct page **xdr_pages, struct page *page, unsigned int buflen)
L
Linus Torvalds 已提交
557
{
558
	struct xdr_stream stream;
559
	struct xdr_buf buf;
560
	struct page *scratch;
B
Bryan Schumaker 已提交
561
	struct nfs_cache_array *array;
562 563
	unsigned int count = 0;
	int status;
564

565 566 567
	scratch = alloc_page(GFP_KERNEL);
	if (scratch == NULL)
		return -ENOMEM;
568

569
	xdr_init_decode_pages(&stream, &buf, xdr_pages, buflen);
570
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
B
Bryan Schumaker 已提交
571 572 573

	do {
		status = xdr_decode(desc, entry, &stream);
574 575 576
		if (status != 0) {
			if (status == -EAGAIN)
				status = 0;
B
Bryan Schumaker 已提交
577
			break;
578
		}
B
Bryan Schumaker 已提交
579

580 581
		count++;

T
Trond Myklebust 已提交
582
		if (desc->plus != 0)
B
Bryan Schumaker 已提交
583
			nfs_prime_dcache(desc->file->f_path.dentry, entry);
584 585 586 587

		status = nfs_readdir_add_to_array(entry, page);
		if (status != 0)
			break;
B
Bryan Schumaker 已提交
588 589
	} while (!entry->eof);

T
Trond Myklebust 已提交
590
	if (count == 0 || (status == -EBADCOOKIE && entry->eof != 0)) {
B
Bryan Schumaker 已提交
591
		array = nfs_readdir_get_array(page);
592 593 594 595
		if (!IS_ERR(array)) {
			array->eof_index = array->size;
			status = 0;
			nfs_readdir_release_array(page);
596 597
		} else
			status = PTR_ERR(array);
L
Linus Torvalds 已提交
598
	}
599 600

	put_page(scratch);
601
	return status;
B
Bryan Schumaker 已提交
602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623
}

static
void nfs_readdir_free_pagearray(struct page **pages, unsigned int npages)
{
	unsigned int i;
	for (i = 0; i < npages; i++)
		put_page(pages[i]);
}

static
void nfs_readdir_free_large_page(void *ptr, struct page **pages,
		unsigned int npages)
{
	nfs_readdir_free_pagearray(pages, npages);
}

/*
 * nfs_readdir_large_page will allocate pages that must be freed with a call
 * to nfs_readdir_free_large_page
 */
static
624
int nfs_readdir_large_page(struct page **pages, unsigned int npages)
B
Bryan Schumaker 已提交
625 626 627 628 629 630 631 632 633
{
	unsigned int i;

	for (i = 0; i < npages; i++) {
		struct page *page = alloc_page(GFP_KERNEL);
		if (page == NULL)
			goto out_freepages;
		pages[i] = page;
	}
634
	return 0;
B
Bryan Schumaker 已提交
635 636 637

out_freepages:
	nfs_readdir_free_pagearray(pages, i);
638
	return -ENOMEM;
L
Linus Torvalds 已提交
639 640
}

B
Bryan Schumaker 已提交
641 642
static
int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode)
643
{
B
Bryan Schumaker 已提交
644 645
	struct page *pages[NFS_MAX_READDIR_PAGES];
	void *pages_ptr = NULL;
B
Bryan Schumaker 已提交
646 647 648
	struct nfs_entry entry;
	struct file	*file = desc->file;
	struct nfs_cache_array *array;
649
	int status = -ENOMEM;
B
Bryan Schumaker 已提交
650
	unsigned int array_size = ARRAY_SIZE(pages);
B
Bryan Schumaker 已提交
651 652

	entry.prev_cookie = 0;
653
	entry.cookie = desc->last_cookie;
B
Bryan Schumaker 已提交
654 655 656
	entry.eof = 0;
	entry.fh = nfs_alloc_fhandle();
	entry.fattr = nfs_alloc_fattr();
657
	entry.server = NFS_SERVER(inode);
B
Bryan Schumaker 已提交
658 659
	if (entry.fh == NULL || entry.fattr == NULL)
		goto out;
660

B
Bryan Schumaker 已提交
661
	array = nfs_readdir_get_array(page);
662 663 664 665
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
666 667
	memset(array, 0, sizeof(struct nfs_cache_array));
	array->eof_index = -1;
668

669 670
	status = nfs_readdir_large_page(pages, array_size);
	if (status < 0)
B
Bryan Schumaker 已提交
671 672
		goto out_release_array;
	do {
673
		unsigned int pglen;
B
Bryan Schumaker 已提交
674
		status = nfs_readdir_xdr_filler(pages, desc, &entry, file, inode);
675

B
Bryan Schumaker 已提交
676
		if (status < 0)
677
			break;
678
		pglen = status;
679
		status = nfs_readdir_page_filler(desc, &entry, pages, page, pglen);
680 681 682 683 684 685
		if (status < 0) {
			if (status == -ENOSPC)
				status = 0;
			break;
		}
	} while (array->eof_index < 0);
B
Bryan Schumaker 已提交
686

B
Bryan Schumaker 已提交
687
	nfs_readdir_free_large_page(pages_ptr, pages, array_size);
B
Bryan Schumaker 已提交
688 689 690 691 692
out_release_array:
	nfs_readdir_release_array(page);
out:
	nfs_free_fattr(entry.fattr);
	nfs_free_fhandle(entry.fh);
693 694 695 696
	return status;
}

/*
B
Bryan Schumaker 已提交
697 698 699 700
 * Now we cache directories properly, by converting xdr information
 * to an array that can be used for lookups later.  This results in
 * fewer cache pages, since we can store more information on each page.
 * We only need to convert from xdr once so future lookups are much simpler
L
Linus Torvalds 已提交
701
 */
B
Bryan Schumaker 已提交
702 703
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page)
L
Linus Torvalds 已提交
704
{
705
	struct inode	*inode = desc->file->f_path.dentry->d_inode;
706
	int ret;
L
Linus Torvalds 已提交
707

708 709
	ret = nfs_readdir_xdr_to_array(desc, page, inode);
	if (ret < 0)
B
Bryan Schumaker 已提交
710 711
		goto error;
	SetPageUptodate(page);
L
Linus Torvalds 已提交
712

B
Bryan Schumaker 已提交
713 714 715
	if (invalidate_inode_pages2_range(inode->i_mapping, page->index + 1, -1) < 0) {
		/* Should never happen */
		nfs_zap_mapping(inode, inode->i_mapping);
L
Linus Torvalds 已提交
716
	}
B
Bryan Schumaker 已提交
717 718 719 720
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
721
	return ret;
B
Bryan Schumaker 已提交
722
}
L
Linus Torvalds 已提交
723

B
Bryan Schumaker 已提交
724 725 726
static
void cache_page_release(nfs_readdir_descriptor_t *desc)
{
727 728
	if (!desc->page->mapping)
		nfs_readdir_clear_array(desc->page);
B
Bryan Schumaker 已提交
729 730 731 732 733 734 735
	page_cache_release(desc->page);
	desc->page = NULL;
}

static
struct page *get_cache_page(nfs_readdir_descriptor_t *desc)
{
736
	return read_cache_page(desc->file->f_path.dentry->d_inode->i_mapping,
B
Bryan Schumaker 已提交
737
			desc->page_index, (filler_t *)nfs_readdir_filler, desc);
L
Linus Torvalds 已提交
738 739 740
}

/*
B
Bryan Schumaker 已提交
741
 * Returns 0 if desc->dir_cookie was found on page desc->page_index
L
Linus Torvalds 已提交
742
 */
B
Bryan Schumaker 已提交
743 744 745 746 747 748 749 750 751 752
static
int find_cache_page(nfs_readdir_descriptor_t *desc)
{
	int res;

	desc->page = get_cache_page(desc);
	if (IS_ERR(desc->page))
		return PTR_ERR(desc->page);

	res = nfs_readdir_search_array(desc);
T
Trond Myklebust 已提交
753 754
	if (res != 0)
		cache_page_release(desc);
B
Bryan Schumaker 已提交
755 756 757 758
	return res;
}

/* Search for desc->dir_cookie from the beginning of the page cache */
L
Linus Torvalds 已提交
759 760 761
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
762
	int res;
B
Bryan Schumaker 已提交
763

764
	if (desc->page_index == 0) {
765
		desc->current_index = 0;
766 767
		desc->last_cookie = 0;
	}
T
Trond Myklebust 已提交
768
	do {
B
Bryan Schumaker 已提交
769
		res = find_cache_page(desc);
T
Trond Myklebust 已提交
770
	} while (res == -EAGAIN);
L
Linus Torvalds 已提交
771 772 773 774 775 776 777 778 779 780 781
	return res;
}

/*
 * 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;
B
Bryan Schumaker 已提交
782 783 784
	int i = 0;
	int res = 0;
	struct nfs_cache_array *array = NULL;
785 786
	struct nfs_open_dir_context *ctx = file->private_data;

B
Bryan Schumaker 已提交
787
	array = nfs_readdir_get_array(desc->page);
788 789 790 791
	if (IS_ERR(array)) {
		res = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
792 793

	for (i = desc->cache_entry_index; i < array->size; i++) {
794
		struct nfs_cache_array_entry *ent;
L
Linus Torvalds 已提交
795

796 797
		ent = &array->array[i];
		if (filldir(dirent, ent->string.name, ent->string.len,
798 799
		    file->f_pos, nfs_compat_user_ino64(ent->ino),
		    ent->d_type) < 0) {
800
			desc->eof = 1;
L
Linus Torvalds 已提交
801
			break;
802
		}
803
		file->f_pos++;
B
Bryan Schumaker 已提交
804 805 806 807
		if (i < (array->size-1))
			*desc->dir_cookie = array->array[i+1].cookie;
		else
			*desc->dir_cookie = array->last_cookie;
808 809
		if (ctx->duped != 0)
			ctx->duped = 1;
L
Linus Torvalds 已提交
810
	}
T
Trond Myklebust 已提交
811
	if (array->eof_index >= 0)
812
		desc->eof = 1;
B
Bryan Schumaker 已提交
813 814

	nfs_readdir_release_array(desc->page);
815
out:
B
Bryan Schumaker 已提交
816
	cache_page_release(desc);
C
Chuck Lever 已提交
817 818
	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
819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839
	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 page	*page = NULL;
	int		status;
B
Bryan Schumaker 已提交
840
	struct inode *inode = desc->file->f_path.dentry->d_inode;
841
	struct nfs_open_dir_context *ctx = desc->file->private_data;
L
Linus Torvalds 已提交
842

C
Chuck Lever 已提交
843 844
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
845 846 847 848 849 850

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
B
Bryan Schumaker 已提交
851

852
	desc->page_index = 0;
853
	desc->last_cookie = *desc->dir_cookie;
854
	desc->page = page;
855
	ctx->duped = 0;
856

857 858
	status = nfs_readdir_xdr_to_array(desc, page, inode);
	if (status < 0)
L
Linus Torvalds 已提交
859 860 861 862 863
		goto out_release;

	status = nfs_do_filldir(desc, dirent, filldir);

 out:
C
Chuck Lever 已提交
864
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
865
			__func__, status);
L
Linus Torvalds 已提交
866 867
	return status;
 out_release:
B
Bryan Schumaker 已提交
868
	cache_page_release(desc);
L
Linus Torvalds 已提交
869 870 871
	goto out;
}

872 873 874
/* 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
875 876 877
 */
static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
878
	struct dentry	*dentry = filp->f_path.dentry;
L
Linus Torvalds 已提交
879 880 881
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
882
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
T
Trond Myklebust 已提交
883
	int res;
L
Linus Torvalds 已提交
884

C
Chuck Lever 已提交
885
	dfprintk(FILE, "NFS: readdir(%s/%s) starting at cookie %llu\n",
C
Chuck Lever 已提交
886 887
			dentry->d_parent->d_name.name, dentry->d_name.name,
			(long long)filp->f_pos);
C
Chuck Lever 已提交
888 889
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
890
	/*
891
	 * filp->f_pos points to the dirent entry number.
892
	 * *desc->dir_cookie has the cookie for the next entry. We have
893 894
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
895 896 897 898
	 */
	memset(desc, 0, sizeof(*desc));

	desc->file = filp;
899
	desc->dir_cookie = &dir_ctx->dir_cookie;
L
Linus Torvalds 已提交
900
	desc->decode = NFS_PROTO(inode)->decode_dirent;
901
	desc->plus = nfs_use_readdirplus(inode, filp) ? 1 : 0;
L
Linus Torvalds 已提交
902

T
Trond Myklebust 已提交
903
	nfs_block_sillyrename(dentry);
904
	res = nfs_revalidate_mapping(inode, filp->f_mapping);
T
Trond Myklebust 已提交
905 906 907
	if (res < 0)
		goto out;

T
Trond Myklebust 已提交
908
	do {
L
Linus Torvalds 已提交
909
		res = readdir_search_pagecache(desc);
910

L
Linus Torvalds 已提交
911
		if (res == -EBADCOOKIE) {
912
			res = 0;
L
Linus Torvalds 已提交
913
			/* This means either end of directory */
B
Bryan Schumaker 已提交
914
			if (*desc->dir_cookie && desc->eof == 0) {
L
Linus Torvalds 已提交
915 916
				/* Or that the server has 'lost' a cookie */
				res = uncached_readdir(desc, dirent, filldir);
917
				if (res == 0)
L
Linus Torvalds 已提交
918 919 920 921 922
					continue;
			}
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
923
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
924
			nfs_zap_caches(inode);
925
			desc->page_index = 0;
L
Linus Torvalds 已提交
926
			desc->plus = 0;
B
Bryan Schumaker 已提交
927
			desc->eof = 0;
L
Linus Torvalds 已提交
928 929 930 931 932 933
			continue;
		}
		if (res < 0)
			break;

		res = nfs_do_filldir(desc, dirent, filldir);
934
		if (res < 0)
L
Linus Torvalds 已提交
935
			break;
T
Trond Myklebust 已提交
936
	} while (!desc->eof);
T
Trond Myklebust 已提交
937
out:
T
Trond Myklebust 已提交
938
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
939 940
	if (res > 0)
		res = 0;
941
	dfprintk(FILE, "NFS: readdir(%s/%s) returns %d\n",
C
Chuck Lever 已提交
942 943 944
			dentry->d_parent->d_name.name, dentry->d_name.name,
			res);
	return res;
L
Linus Torvalds 已提交
945 946
}

947
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
948
{
949 950
	struct dentry *dentry = filp->f_path.dentry;
	struct inode *inode = dentry->d_inode;
951
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
952

C
Chuck Lever 已提交
953
	dfprintk(FILE, "NFS: llseek dir(%s/%s, %lld, %d)\n",
954 955 956 957 958
			dentry->d_parent->d_name.name,
			dentry->d_name.name,
			offset, origin);

	mutex_lock(&inode->i_mutex);
959 960 961 962 963 964 965 966 967 968 969 970
	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;
971
		dir_ctx->dir_cookie = 0;
972
		dir_ctx->duped = 0;
973 974
	}
out:
975
	mutex_unlock(&inode->i_mutex);
976 977 978
	return offset;
}

L
Linus Torvalds 已提交
979 980 981 982
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
983 984
static int nfs_fsync_dir(struct file *filp, loff_t start, loff_t end,
			 int datasync)
L
Linus Torvalds 已提交
985
{
986
	struct dentry *dentry = filp->f_path.dentry;
987
	struct inode *inode = dentry->d_inode;
988

C
Chuck Lever 已提交
989
	dfprintk(FILE, "NFS: fsync dir(%s/%s) datasync %d\n",
C
Chuck Lever 已提交
990 991 992
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);

993
	mutex_lock(&inode->i_mutex);
994
	nfs_inc_stats(dentry->d_inode, NFSIOS_VFSFSYNC);
995
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
996 997 998
	return 0;
}

999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
/**
 * 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)
{
1011
	NFS_I(dir)->cache_change_attribute++;
1012 1013
}

L
Linus Torvalds 已提交
1014 1015 1016 1017 1018
/*
 * 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.
 */
1019
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1020 1021 1022
{
	if (IS_ROOT(dentry))
		return 1;
1023 1024
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
1025 1026 1027 1028 1029 1030 1031 1032
	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
1033 1034
}

1035 1036 1037 1038
/*
 * Return the intent data that applies to this particular path component
 *
 * Note that the current set of intents only apply to the very last
1039 1040
 * component of the path and none of them is set before that last
 * component.
1041
 */
1042 1043
static inline unsigned int nfs_lookup_check_intent(struct nameidata *nd,
						unsigned int mask)
1044 1045 1046 1047
{
	return nd->flags & mask;
}

1048 1049 1050 1051 1052 1053 1054 1055
/*
 * 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;
1056
	return nd && nfs_lookup_check_intent(nd, LOOKUP_EXCL);
1057 1058
}

1059 1060 1061 1062 1063 1064 1065 1066
/*
 * 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
1067 1068 1069 1070 1071
static inline
int nfs_lookup_verify_inode(struct inode *inode, struct nameidata *nd)
{
	struct nfs_server *server = NFS_SERVER(inode);

1072
	if (IS_AUTOMOUNT(inode))
1073
		return 0;
L
Linus Torvalds 已提交
1074 1075
	if (nd != NULL) {
		/* VFS wants an on-the-wire revalidation */
1076
		if (nd->flags & LOOKUP_REVAL)
L
Linus Torvalds 已提交
1077 1078
			goto out_force;
		/* This is an open(2) */
1079
		if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 &&
1080 1081 1082
				!(server->flags & NFS_MOUNT_NOCTO) &&
				(S_ISREG(inode->i_mode) ||
				 S_ISDIR(inode->i_mode)))
L
Linus Torvalds 已提交
1083
			goto out_force;
1084
		return 0;
L
Linus Torvalds 已提交
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102
	}
	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 */
1103
	if (nd != NULL && nfs_lookup_check_intent(nd, LOOKUP_CREATE) != 0)
L
Linus Torvalds 已提交
1104
		return 0;
1105 1106
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120
	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.
 */
1121
static int nfs_lookup_revalidate(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
1122 1123 1124 1125
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
1126 1127
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1128 1129
	int error;

1130 1131 1132
	if (nd->flags & LOOKUP_RCU)
		return -ECHILD;

L
Linus Torvalds 已提交
1133 1134
	parent = dget_parent(dentry);
	dir = parent->d_inode;
C
Chuck Lever 已提交
1135
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
1136 1137 1138 1139 1140
	inode = dentry->d_inode;

	if (!inode) {
		if (nfs_neg_need_reval(dir, dentry, nd))
			goto out_bad;
1141
		goto out_valid_noent;
L
Linus Torvalds 已提交
1142 1143 1144
	}

	if (is_bad_inode(inode)) {
C
Chuck Lever 已提交
1145
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
1146
				__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1147
				dentry->d_name.name);
L
Linus Torvalds 已提交
1148 1149 1150
		goto out_bad;
	}

1151 1152 1153
	if (nfs_have_delegation(inode, FMODE_READ))
		goto out_set_verifier;

L
Linus Torvalds 已提交
1154
	/* Force a full look up iff the parent directory has changed */
1155
	if (!nfs_is_exclusive_create(dir, nd) && nfs_check_verifier(dir, dentry)) {
L
Linus Torvalds 已提交
1156 1157 1158 1159 1160 1161 1162 1163
		if (nfs_lookup_verify_inode(inode, nd))
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

1164 1165 1166 1167 1168 1169
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

1170
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1171 1172
	if (error)
		goto out_bad;
1173
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
1174
		goto out_bad;
1175
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
1176 1177
		goto out_bad;

1178 1179
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1180
out_set_verifier:
1181
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1182
 out_valid:
1183 1184 1185
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);
 out_valid_noent:
L
Linus Torvalds 已提交
1186
	dput(parent);
C
Chuck Lever 已提交
1187
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n",
1188
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1189
			dentry->d_name.name);
L
Linus Torvalds 已提交
1190 1191 1192 1193
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
1194
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
1195 1196 1197 1198 1199 1200
	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;
1201 1202
		if (dentry->d_flags & DCACHE_DISCONNECTED)
			goto out_valid;
L
Linus Torvalds 已提交
1203 1204 1205
		shrink_dcache_parent(dentry);
	}
	d_drop(dentry);
1206 1207
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1208
	dput(parent);
C
Chuck Lever 已提交
1209
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
1210
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1211
			dentry->d_name.name);
L
Linus Torvalds 已提交
1212
	return 0;
1213 1214 1215 1216 1217 1218 1219 1220
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
1221 1222 1223 1224 1225
}

/*
 * This is called from dput() when d_count is going to 0.
 */
N
Nick Piggin 已提交
1226
static int nfs_dentry_delete(const struct dentry *dentry)
L
Linus Torvalds 已提交
1227 1228 1229 1230 1231
{
	dfprintk(VFS, "NFS: dentry_delete(%s/%s, %x)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		dentry->d_flags);

1232 1233 1234 1235
	/* 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
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
	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;

}

1249 1250 1251 1252 1253 1254 1255 1256
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
1257 1258 1259 1260 1261 1262
/*
 * 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)
{
1263 1264 1265 1266
	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
1267
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
1268
		drop_nlink(inode);
1269
		nfs_complete_unlink(dentry, inode);
L
Linus Torvalds 已提交
1270 1271 1272 1273
	}
	iput(inode);
}

1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
static void nfs_d_release(struct dentry *dentry)
{
	/* free cached devname value, if it survived that far */
	if (unlikely(dentry->d_fsdata)) {
		if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
			WARN_ON(1);
		else
			kfree(dentry->d_fsdata);
	}
}

A
Al Viro 已提交
1285
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
1286 1287 1288
	.d_revalidate	= nfs_lookup_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1289
	.d_automount	= nfs_d_automount,
1290
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1291 1292 1293 1294 1295
};

static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
{
	struct dentry *res;
T
Trond Myklebust 已提交
1296
	struct dentry *parent;
L
Linus Torvalds 已提交
1297
	struct inode *inode = NULL;
1298 1299
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1300 1301 1302 1303
	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
1304
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
1305 1306 1307 1308 1309

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

1310 1311 1312 1313 1314 1315 1316
	/*
	 * 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;
1317
		goto out;
1318
	}
L
Linus Torvalds 已提交
1319

1320 1321 1322 1323 1324 1325
	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
1326 1327 1328
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
	nfs_block_sillyrename(parent);
1329
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1330 1331 1332 1333
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
1334
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1335
	}
1336
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
N
Namhyung Kim 已提交
1337
	res = ERR_CAST(inode);
1338
	if (IS_ERR(res))
T
Trond Myklebust 已提交
1339
		goto out_unblock_sillyrename;
1340

1341 1342 1343
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);

L
Linus Torvalds 已提交
1344
no_entry:
1345
	res = d_materialise_unique(dentry, inode);
1346 1347
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
1348
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1349
		dentry = res;
1350
	}
L
Linus Torvalds 已提交
1351
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
1352 1353
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
L
Linus Torvalds 已提交
1354
out:
1355 1356
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1357 1358 1359 1360
	return res;
}

#ifdef CONFIG_NFS_V4
1361
static int nfs4_lookup_revalidate(struct dentry *, struct nameidata *);
L
Linus Torvalds 已提交
1362

A
Al Viro 已提交
1363
const struct dentry_operations nfs4_dentry_operations = {
1364
	.d_revalidate	= nfs4_lookup_revalidate,
L
Linus Torvalds 已提交
1365 1366
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1367
	.d_automount	= nfs_d_automount,
1368
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1369 1370
};

1371 1372 1373 1374
/*
 * 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
1375
static int is_atomic_open(struct nameidata *nd)
L
Linus Torvalds 已提交
1376
{
1377
	if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_OPEN) == 0)
L
Linus Torvalds 已提交
1378 1379 1380 1381 1382
		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? */
1383
	if (__mnt_is_readonly(nd->path.mnt) &&
1384
	    (nd->intent.open.flags & (O_CREAT|O_TRUNC|O_ACCMODE)))
L
Linus Torvalds 已提交
1385 1386 1387 1388
		return 0;
	return 1;
}

1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
static fmode_t flags_to_mode(int flags)
{
	fmode_t res = (__force fmode_t)flags & FMODE_EXEC;
	if ((flags & O_ACCMODE) != O_WRONLY)
		res |= FMODE_READ;
	if ((flags & O_ACCMODE) != O_RDONLY)
		res |= FMODE_WRITE;
	return res;
}

1399
static struct nfs_open_context *create_nfs_open_context(struct dentry *dentry, int open_flags)
1400
{
1401
	return alloc_nfs_open_context(dentry, flags_to_mode(open_flags));
1402 1403 1404 1405 1406 1407 1408 1409
}

static int do_open(struct inode *inode, struct file *filp)
{
	nfs_fscache_set_inode_cookie(inode, filp);
	return 0;
}

1410 1411 1412
static struct file *nfs_finish_open(struct nfs_open_context *ctx,
				    struct dentry *dentry,
				    struct opendata *od, unsigned open_flags)
1413 1414
{
	struct file *filp;
1415 1416 1417 1418 1419 1420
	int err;

	if (ctx->dentry != dentry) {
		dput(ctx->dentry);
		ctx->dentry = dget(dentry);
	}
1421 1422 1423

	/* If the open_intent is for execute, we have an extra check to make */
	if (ctx->mode & FMODE_EXEC) {
1424 1425 1426
		err = nfs_may_open(dentry->d_inode, ctx->cred, open_flags);
		if (err < 0) {
			filp = ERR_PTR(err);
1427
			goto out;
1428
		}
1429
	}
1430 1431 1432

	filp = finish_open(od, dentry, do_open);
	if (!IS_ERR(filp))
1433
		nfs_file_set_open_context(filp, ctx);
1434

1435 1436
out:
	put_nfs_open_context(ctx);
1437
	return filp;
1438 1439
}

1440 1441 1442
static struct file *nfs_atomic_open(struct inode *dir, struct dentry *dentry,
				    struct opendata *od, unsigned open_flags,
				    umode_t mode, bool *created)
L
Linus Torvalds 已提交
1443
{
1444
	struct nfs_open_context *ctx;
1445 1446
	struct dentry *res;
	struct iattr attr = { .ia_valid = ATTR_OPEN };
1447
	struct inode *inode;
1448
	struct file *filp;
1449
	int err;
L
Linus Torvalds 已提交
1450

1451 1452 1453 1454
	/* Expect a negative dentry */
	BUG_ON(dentry->d_inode);

	dfprintk(VFS, "NFS: atomic_open(%s/%ld), %s\n",
C
Chuck Lever 已提交
1455 1456
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);

1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467
	/* NFS only supports OPEN on regular files */
	if ((open_flags & O_DIRECTORY)) {
		err = -ENOENT;
		if (!d_unhashed(dentry)) {
			/*
			 * Hashed negative dentry with O_DIRECTORY: dentry was
			 * revalidated and is fine, no need to perform lookup
			 * again
			 */
			goto out_err;
		}
L
Linus Torvalds 已提交
1468
		goto no_open;
1469
	}
L
Linus Torvalds 已提交
1470

1471 1472 1473
	err = -ENAMETOOLONG;
	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
		goto out_err;
1474

1475
	if (open_flags & O_CREAT) {
1476
		attr.ia_valid |= ATTR_MODE;
1477 1478
		attr.ia_mode = mode & ~current_umask();
	}
1479 1480 1481
	if (open_flags & O_TRUNC) {
		attr.ia_valid |= ATTR_SIZE;
		attr.ia_size = 0;
1482 1483
	}

1484 1485 1486 1487 1488
	ctx = create_nfs_open_context(dentry, open_flags);
	err = PTR_ERR(ctx);
	if (IS_ERR(ctx))
		goto out_err;

1489
	nfs_block_sillyrename(dentry->d_parent);
1490
	inode = NFS_PROTO(dir)->open_context(dir, ctx, open_flags, &attr);
1491
	d_drop(dentry);
1492 1493
	if (IS_ERR(inode)) {
		nfs_unblock_sillyrename(dentry->d_parent);
1494
		put_nfs_open_context(ctx);
1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
		err = PTR_ERR(inode);
		switch (err) {
		case -ENOENT:
			d_add(dentry, NULL);
			break;
		case -EISDIR:
		case -ENOTDIR:
			goto no_open;
		case -ELOOP:
			if (!(open_flags & O_NOFOLLOW))
1505
				goto no_open;
1506
			break;
L
Linus Torvalds 已提交
1507
			/* case -EINVAL: */
1508 1509
		default:
			break;
L
Linus Torvalds 已提交
1510
		}
1511
		goto out_err;
1512
	}
1513
	res = d_add_unique(dentry, inode);
1514
	if (res != NULL)
L
Linus Torvalds 已提交
1515
		dentry = res;
1516 1517

	nfs_unblock_sillyrename(dentry->d_parent);
1518
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
1519 1520 1521 1522 1523 1524 1525 1526 1527

	filp = nfs_finish_open(ctx, dentry, od, open_flags);

	dput(res);
	return filp;

out_err:
	return ERR_PTR(err);

L
Linus Torvalds 已提交
1528
no_open:
1529 1530 1531 1532 1533 1534 1535
	res = nfs_lookup(dir, dentry, NULL);
	err = PTR_ERR(res);
	if (IS_ERR(res))
		goto out_err;

	finish_no_open(od, res);
	return NULL;
L
Linus Torvalds 已提交
1536 1537
}

1538
static int nfs4_lookup_revalidate(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
1539 1540
{
	struct dentry *parent = NULL;
N
Nick Piggin 已提交
1541
	struct inode *inode;
L
Linus Torvalds 已提交
1542 1543 1544
	struct inode *dir;
	int openflags, ret = 0;

N
Nick Piggin 已提交
1545 1546 1547 1548
	if (nd->flags & LOOKUP_RCU)
		return -ECHILD;

	inode = dentry->d_inode;
1549
	if (!is_atomic_open(nd) || d_mountpoint(dentry))
T
Trond Myklebust 已提交
1550
		goto no_open;
1551

L
Linus Torvalds 已提交
1552 1553
	parent = dget_parent(dentry);
	dir = parent->d_inode;
1554

L
Linus Torvalds 已提交
1555 1556 1557
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1558 1559 1560
	if (inode == NULL) {
		if (!nfs_neg_need_reval(dir, dentry, nd))
			ret = 1;
L
Linus Torvalds 已提交
1561
		goto out;
1562 1563
	}

L
Linus Torvalds 已提交
1564 1565
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
T
Trond Myklebust 已提交
1566
		goto no_open_dput;
L
Linus Torvalds 已提交
1567 1568 1569
	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
1570
		goto no_open_dput;
L
Linus Torvalds 已提交
1571

1572 1573
	/* Let f_op->open() actually open (and revalidate) the file */
	ret = 1;
1574

L
Linus Torvalds 已提交
1575 1576 1577
out:
	dput(parent);
	return ret;
1578

T
Trond Myklebust 已提交
1579
no_open_dput:
L
Linus Torvalds 已提交
1580
	dput(parent);
T
Trond Myklebust 已提交
1581
no_open:
L
Linus Torvalds 已提交
1582 1583
	return nfs_lookup_revalidate(dentry, nd);
}
1584

1585 1586
static int nfs4_create(struct inode *dir, struct dentry *dentry,
		       umode_t mode, struct nameidata *nd)
1587 1588 1589 1590
{
	struct nfs_open_context *ctx = NULL;
	struct iattr attr;
	int error;
1591
	int open_flags = O_CREAT|O_EXCL;
1592 1593 1594 1595 1596 1597 1598

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

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

1599
	if (nd)
1600 1601
		open_flags = nd->intent.open.flags;

A
Al Viro 已提交
1602 1603 1604 1605
	ctx = create_nfs_open_context(dentry, open_flags);
	error = PTR_ERR(ctx);
	if (IS_ERR(ctx))
		goto out_err_drop;
1606 1607 1608 1609

	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, ctx);
	if (error != 0)
		goto out_put_ctx;
1610 1611 1612

	put_nfs_open_context(ctx);

1613 1614
	return 0;
out_put_ctx:
A
Al Viro 已提交
1615
	put_nfs_open_context(ctx);
1616
out_err_drop:
1617 1618 1619 1620
	d_drop(dentry);
	return error;
}

L
Linus Torvalds 已提交
1621 1622 1623 1624 1625 1626 1627 1628
#endif /* CONFIG_NFSV4 */

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
				struct nfs_fattr *fattr)
{
1629 1630
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1631 1632 1633
	struct inode *inode;
	int error = -EACCES;

1634 1635
	d_drop(dentry);

L
Linus Torvalds 已提交
1636 1637
	/* We may have been initialized further down */
	if (dentry->d_inode)
1638
		goto out;
L
Linus Torvalds 已提交
1639
	if (fhandle->size == 0) {
1640
		error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1641
		if (error)
1642
			goto out_error;
L
Linus Torvalds 已提交
1643
	}
1644
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1645 1646
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1647
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr);
L
Linus Torvalds 已提交
1648
		if (error < 0)
1649
			goto out_error;
L
Linus Torvalds 已提交
1650 1651
	}
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
1652 1653
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1654 1655 1656 1657
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1658
	return 0;
1659 1660 1661 1662
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1663 1664 1665 1666 1667 1668 1669 1670
}

/*
 * 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.
 */
A
Al Viro 已提交
1671 1672
static int nfs_create(struct inode *dir, struct dentry *dentry,
		umode_t mode, struct nameidata *nd)
L
Linus Torvalds 已提交
1673 1674 1675
{
	struct iattr attr;
	int error;
1676
	int open_flags = O_CREAT|O_EXCL;
L
Linus Torvalds 已提交
1677

C
Chuck Lever 已提交
1678 1679
	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
1680 1681 1682 1683

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

1684
	if (nd)
1685 1686 1687
		open_flags = nd->intent.open.flags;

	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, NULL);
L
Linus Torvalds 已提交
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699
	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
A
Al Viro 已提交
1700
nfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t rdev)
L
Linus Torvalds 已提交
1701 1702 1703 1704
{
	struct iattr attr;
	int status;

C
Chuck Lever 已提交
1705 1706
	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
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725

	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.
 */
1726
static int nfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
1727 1728 1729 1730
{
	struct iattr attr;
	int error;

C
Chuck Lever 已提交
1731 1732
	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
1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745

	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;
}

1746 1747 1748 1749 1750 1751
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
1752 1753 1754 1755
static int nfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	int error;

C
Chuck Lever 已提交
1756 1757
	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
1758 1759 1760 1761

	error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
	/* Ensure the VFS deletes this inode */
	if (error == 0 && dentry->d_inode != NULL)
1762
		clear_nlink(dentry->d_inode);
1763 1764
	else if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
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

	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) {
1792
		nfs_inode_return_delegation(inode);
L
Linus Torvalds 已提交
1793 1794 1795
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		/* The VFS may want to delete this inode */
		if (error == 0)
1796
			nfs_drop_nlink(inode);
1797
		nfs_mark_for_revalidate(inode);
L
Linus Torvalds 已提交
1798 1799
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1800 1801
	if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819
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(&dentry->d_lock);
N
Nick Piggin 已提交
1820
	if (dentry->d_count > 1) {
L
Linus Torvalds 已提交
1821
		spin_unlock(&dentry->d_lock);
T
Trond Myklebust 已提交
1822 1823
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1824 1825 1826 1827 1828 1829 1830 1831 1832
		error = nfs_sillyrename(dir, dentry);
		return error;
	}
	if (!d_unhashed(dentry)) {
		__d_drop(dentry);
		need_rehash = 1;
	}
	spin_unlock(&dentry->d_lock);
	error = nfs_safe_remove(dentry);
1833
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1834 1835 1836 1837 1838 1839
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
	return error;
}

1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
/*
 * 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
1856
{
1857 1858 1859
	struct pagevec lru_pvec;
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1860
	struct iattr attr;
1861
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1862 1863 1864 1865 1866
	int error;

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

1867 1868
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1869

1870 1871
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1872

1873
	page = alloc_page(GFP_HIGHUSER);
1874
	if (!page)
1875 1876
		return -ENOMEM;

1877
	kaddr = kmap_atomic(page);
1878 1879 1880
	memcpy(kaddr, symname, pathlen);
	if (pathlen < PAGE_SIZE)
		memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen);
1881
	kunmap_atomic(kaddr);
1882

1883
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1884 1885 1886 1887
	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
1888
		d_drop(dentry);
1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899
		__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)) {
1900
		pagevec_add(&lru_pvec, page);
1901
		pagevec_lru_add_file(&lru_pvec);
1902 1903 1904 1905 1906 1907
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919
}

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);

1920 1921
	nfs_inode_return_delegation(inode);

1922
	d_drop(dentry);
L
Linus Torvalds 已提交
1923
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1924
	if (error == 0) {
A
Al Viro 已提交
1925
		ihold(inode);
1926
		d_add(dentry, inode);
1927
	}
L
Linus Torvalds 已提交
1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
	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,
N
Nick Piggin 已提交
1966
		 new_dentry->d_count);
L
Linus Torvalds 已提交
1967 1968

	/*
M
Miklos Szeredi 已提交
1969 1970 1971 1972
	 * 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
1973
	 */
1974 1975 1976 1977 1978 1979 1980 1981 1982
	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
1983

N
Nick Piggin 已提交
1984
		if (new_dentry->d_count > 2) {
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
			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);
1995
			if (err)
1996
				goto out;
1997 1998

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
1999
			rehash = NULL;
2000
			new_inode = NULL;
2001
		}
2002
	}
L
Linus Torvalds 已提交
2003

2004
	nfs_inode_return_delegation(old_inode);
2005
	if (new_inode != NULL)
2006
		nfs_inode_return_delegation(new_inode);
L
Linus Torvalds 已提交
2007 2008 2009

	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
2010
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
2011 2012 2013 2014
out:
	if (rehash)
		d_rehash(rehash);
	if (!error) {
2015 2016
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
2017
		d_move(old_dentry, new_dentry);
2018 2019
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
2020 2021
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
2022 2023 2024 2025 2026 2027 2028

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

2029 2030 2031 2032
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

2033 2034 2035 2036
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
2037 2038 2039
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
2040 2041
}

2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052
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);
	}
}

2053 2054
int nfs_access_cache_shrinker(struct shrinker *shrink,
			      struct shrink_control *sc)
2055 2056
{
	LIST_HEAD(head);
2057
	struct nfs_inode *nfsi, *next;
2058
	struct nfs_access_entry *cache;
2059 2060
	int nr_to_scan = sc->nr_to_scan;
	gfp_t gfp_mask = sc->gfp_mask;
2061

2062 2063
	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
		return (nr_to_scan == 0) ? 0 : -1;
2064

2065
	spin_lock(&nfs_access_lru_lock);
2066
	list_for_each_entry_safe(nfsi, next, &nfs_access_lru_list, access_cache_inode_lru) {
2067 2068 2069 2070
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
2071
		inode = &nfsi->vfs_inode;
2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
		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);
2085
			smp_mb__before_clear_bit();
2086
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
2087
			smp_mb__after_clear_bit();
2088
		}
2089
		spin_unlock(&inode->i_lock);
2090 2091
	}
	spin_unlock(&nfs_access_lru_lock);
2092
	nfs_access_free_list(&head);
2093 2094 2095
	return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure;
}

2096
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
2097
{
2098
	struct rb_root *root_node = &nfsi->access_cache;
2099
	struct rb_node *n;
2100 2101 2102 2103 2104 2105
	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);
2106
		list_move(&entry->lru, head);
2107 2108
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
2109 2110
}

2111
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
2112
{
2113 2114 2115 2116
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
2117
	/* Remove from global LRU init */
2118 2119
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
2120 2121
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

2122
	spin_lock(&inode->i_lock);
2123 2124 2125 2126
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
2127
}
L
Linus Torvalds 已提交
2128

2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142
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
2143
	}
2144 2145 2146
	return NULL;
}

2147
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
2148 2149 2150 2151 2152
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

2153
	spin_lock(&inode->i_lock);
2154 2155 2156 2157 2158
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
2159
	if (!nfs_have_delegated_attributes(inode) &&
2160
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
2161 2162 2163 2164
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
2165
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
2166 2167 2168 2169 2170 2171
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
2172
	list_del(&cache->lru);
2173 2174 2175 2176
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
2177 2178
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
2179 2180 2181 2182 2183
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
2184 2185
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203
	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);
2204
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
2205
	spin_unlock(&inode->i_lock);
2206 2207 2208
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
2209 2210
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
2211 2212 2213 2214
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

2215
static void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
2216 2217 2218 2219 2220
{
	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
2221
	cache->jiffies = set->jiffies;
2222
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
2223
	cache->mask = set->mask;
2224 2225

	nfs_access_add_rbtree(inode, cache);
2226 2227 2228 2229 2230 2231 2232

	/* 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 */
2233
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
2234
		spin_lock(&nfs_access_lru_lock);
2235 2236 2237
		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);
2238 2239
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255
}

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);
2256 2257 2258 2259 2260 2261
	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
2262
		return status;
2263
	}
L
Linus Torvalds 已提交
2264 2265
	nfs_access_add_cache(inode, &cache);
out:
2266
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2267 2268 2269 2270
		return 0;
	return -EACCES;
}

2271 2272 2273 2274
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

2275
	if ((openflags & O_ACCMODE) != O_WRONLY)
2276
		mask |= MAY_READ;
2277
	if ((openflags & O_ACCMODE) != O_RDONLY)
2278
		mask |= MAY_WRITE;
2279
	if (openflags & __FMODE_EXEC)
2280 2281 2282 2283 2284 2285 2286 2287 2288
		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));
}

2289
int nfs_permission(struct inode *inode, int mask)
L
Linus Torvalds 已提交
2290 2291 2292 2293
{
	struct rpc_cred *cred;
	int res = 0;

2294
	if (mask & MAY_NOT_BLOCK)
2295 2296
		return -ECHILD;

C
Chuck Lever 已提交
2297 2298
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

2299
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2300 2301
		goto out;
	/* Is this sys_access() ? */
E
Eric Paris 已提交
2302
	if (mask & (MAY_ACCESS | MAY_CHDIR))
L
Linus Torvalds 已提交
2303 2304 2305 2306 2307 2308 2309 2310
		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)
2311 2312
					&& (mask & MAY_OPEN)
					&& !(mask & MAY_EXEC))
L
Linus Torvalds 已提交
2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327
				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;

2328
	cred = rpc_lookup_cred();
L
Linus Torvalds 已提交
2329 2330 2331 2332 2333 2334
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
out:
2335 2336 2337
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

C
Chuck Lever 已提交
2338 2339
	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
2340 2341 2342 2343
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
2344
		res = generic_permission(inode, mask);
C
Chuck Lever 已提交
2345
	goto out;
L
Linus Torvalds 已提交
2346 2347 2348 2349 2350 2351 2352 2353
}

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