dir.c 56.8 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
/*
 *  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
 */

B
Bryan Schumaker 已提交
20
#include <linux/module.h>
L
Linus Torvalds 已提交
21 22 23 24 25 26 27 28 29 30 31 32
#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>
33
#include <linux/pagevec.h>
L
Linus Torvalds 已提交
34
#include <linux/namei.h>
35
#include <linux/mount.h>
A
Alexey Dobriyan 已提交
36
#include <linux/sched.h>
37
#include <linux/kmemleak.h>
38
#include <linux/xattr.h>
L
Linus Torvalds 已提交
39 40

#include "delegation.h"
C
Chuck Lever 已提交
41
#include "iostat.h"
42
#include "internal.h"
43
#include "fscache.h"
L
Linus Torvalds 已提交
44 45 46 47

/* #define NFS_DEBUG_VERBOSE 1 */

static int nfs_opendir(struct inode *, struct file *);
48
static int nfs_closedir(struct inode *, struct file *);
L
Linus Torvalds 已提交
49
static int nfs_readdir(struct file *, void *, filldir_t);
50
static int nfs_fsync_dir(struct file *, loff_t, loff_t, int);
51
static loff_t nfs_llseek_dir(struct file *, loff_t, int);
52
static void nfs_readdir_clear_array(struct page*);
L
Linus Torvalds 已提交
53

54
const struct file_operations nfs_dir_operations = {
55
	.llseek		= nfs_llseek_dir,
L
Linus Torvalds 已提交
56 57 58
	.read		= generic_read_dir,
	.readdir	= nfs_readdir,
	.open		= nfs_opendir,
59
	.release	= nfs_closedir,
L
Linus Torvalds 已提交
60 61 62
	.fsync		= nfs_fsync_dir,
};

63 64
const struct address_space_operations nfs_dir_aops = {
	.freepage = nfs_readdir_clear_array,
B
Bryan Schumaker 已提交
65 66
};

67
static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct inode *dir, struct rpc_cred *cred)
68 69 70 71
{
	struct nfs_open_dir_context *ctx;
	ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
	if (ctx != NULL) {
72
		ctx->duped = 0;
73
		ctx->attr_gencount = NFS_I(dir)->attr_gencount;
74
		ctx->dir_cookie = 0;
75
		ctx->dup_cookie = 0;
76
		ctx->cred = get_rpccred(cred);
77 78 79
		return ctx;
	}
	return  ERR_PTR(-ENOMEM);
80 81 82 83 84 85 86 87
}

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
88 89 90 91 92 93
/*
 * Open file
 */
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
94 95 96
	int res = 0;
	struct nfs_open_dir_context *ctx;
	struct rpc_cred *cred;
L
Linus Torvalds 已提交
97

C
Chuck Lever 已提交
98
	dfprintk(FILE, "NFS: open dir(%s/%s)\n",
99 100 101 102
			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
103

104 105 106
	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return PTR_ERR(cred);
107
	ctx = alloc_nfs_open_dir_context(inode, cred);
108 109 110 111 112
	if (IS_ERR(ctx)) {
		res = PTR_ERR(ctx);
		goto out;
	}
	filp->private_data = ctx;
113 114 115 116 117 118 119
	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);
	}
120 121
out:
	put_rpccred(cred);
L
Linus Torvalds 已提交
122 123 124
	return res;
}

125 126 127 128 129 130 131
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
132 133 134 135
struct nfs_cache_array_entry {
	u64 cookie;
	u64 ino;
	struct qstr string;
136
	unsigned char d_type;
B
Bryan Schumaker 已提交
137 138 139
};

struct nfs_cache_array {
140
	int size;
B
Bryan Schumaker 已提交
141 142 143 144 145
	int eof_index;
	u64 last_cookie;
	struct nfs_cache_array_entry array[0];
};

146
typedef int (*decode_dirent_t)(struct xdr_stream *, struct nfs_entry *, int);
L
Linus Torvalds 已提交
147 148 149 150
typedef struct {
	struct file	*file;
	struct page	*page;
	unsigned long	page_index;
151
	u64		*dir_cookie;
152
	u64		last_cookie;
153
	loff_t		current_index;
L
Linus Torvalds 已提交
154
	decode_dirent_t	decode;
B
Bryan Schumaker 已提交
155

156
	unsigned long	timestamp;
157
	unsigned long	gencount;
B
Bryan Schumaker 已提交
158 159 160
	unsigned int	cache_entry_index;
	unsigned int	plus:1;
	unsigned int	eof:1;
L
Linus Torvalds 已提交
161 162
} nfs_readdir_descriptor_t;

B
Bryan Schumaker 已提交
163 164
/*
 * The caller is responsible for calling nfs_readdir_release_array(page)
L
Linus Torvalds 已提交
165 166
 */
static
B
Bryan Schumaker 已提交
167 168
struct nfs_cache_array *nfs_readdir_get_array(struct page *page)
{
169
	void *ptr;
B
Bryan Schumaker 已提交
170 171
	if (page == NULL)
		return ERR_PTR(-EIO);
172 173 174 175
	ptr = kmap(page);
	if (ptr == NULL)
		return ERR_PTR(-ENOMEM);
	return ptr;
B
Bryan Schumaker 已提交
176 177 178 179 180 181 182 183 184 185 186 187
}

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

/*
 * we are freeing strings created by nfs_add_to_readdir_array()
 */
static
188
void nfs_readdir_clear_array(struct page *page)
B
Bryan Schumaker 已提交
189
{
190
	struct nfs_cache_array *array;
B
Bryan Schumaker 已提交
191
	int i;
192

193
	array = kmap_atomic(page);
B
Bryan Schumaker 已提交
194 195
	for (i = 0; i < array->size; i++)
		kfree(array->array[i].string.name);
196
	kunmap_atomic(array);
B
Bryan Schumaker 已提交
197 198 199 200 201 202 203 204
}

/*
 * 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
205
int nfs_readdir_make_qstr(struct qstr *string, const char *name, unsigned int len)
B
Bryan Schumaker 已提交
206 207 208
{
	string->len = len;
	string->name = kmemdup(name, len, GFP_KERNEL);
209 210
	if (string->name == NULL)
		return -ENOMEM;
211 212 213 214 215
	/*
	 * 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);
216 217
	string->hash = full_name_hash(name, len);
	return 0;
B
Bryan Schumaker 已提交
218 219 220 221 222 223
}

static
int nfs_readdir_add_to_array(struct nfs_entry *entry, struct page *page)
{
	struct nfs_cache_array *array = nfs_readdir_get_array(page);
224 225 226
	struct nfs_cache_array_entry *cache_entry;
	int ret;

B
Bryan Schumaker 已提交
227 228
	if (IS_ERR(array))
		return PTR_ERR(array);
229 230 231 232

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

	/* Check that this entry lies within the page bounds */
233
	ret = -ENOSPC;
234
	if ((char *)&cache_entry[1] - (char *)page_address(page) > PAGE_SIZE)
235
		goto out;
B
Bryan Schumaker 已提交
236

237 238
	cache_entry->cookie = entry->prev_cookie;
	cache_entry->ino = entry->ino;
239
	cache_entry->d_type = entry->d_type;
240 241 242
	ret = nfs_readdir_make_qstr(&cache_entry->string, entry->name, entry->len);
	if (ret)
		goto out;
B
Bryan Schumaker 已提交
243
	array->last_cookie = entry->cookie;
244
	array->size++;
T
Trond Myklebust 已提交
245
	if (entry->eof != 0)
B
Bryan Schumaker 已提交
246
		array->eof_index = array->size;
247
out:
B
Bryan Schumaker 已提交
248
	nfs_readdir_release_array(page);
249
	return ret;
B
Bryan Schumaker 已提交
250 251 252 253 254 255 256 257 258 259 260
}

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) {
261
		if (array->eof_index >= 0)
B
Bryan Schumaker 已提交
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
			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;
279
	loff_t new_pos;
B
Bryan Schumaker 已提交
280 281 282 283
	int status = -EAGAIN;

	for (i = 0; i < array->size; i++) {
		if (array->array[i].cookie == *desc->dir_cookie) {
A
Al Viro 已提交
284
			struct nfs_inode *nfsi = NFS_I(file_inode(desc->file));
285 286
			struct nfs_open_dir_context *ctx = desc->file->private_data;

287
			new_pos = desc->current_index + i;
288 289 290 291 292 293 294 295 296 297
			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.  "
298
								"The file: %s has duplicate cookie %llu\n",
299 300
								desc->file->f_dentry->d_parent->d_name.name,
								desc->file->f_dentry->d_name.name,
301
								array->array[i].string.name,
302 303 304 305 306
								*desc->dir_cookie);
					}
					status = -ELOOP;
					goto out;
				}
307
				ctx->dup_cookie = *desc->dir_cookie;
308
				ctx->duped = -1;
309 310
			}
			desc->file->f_pos = new_pos;
B
Bryan Schumaker 已提交
311
			desc->cache_entry_index = i;
T
Trond Myklebust 已提交
312
			return 0;
B
Bryan Schumaker 已提交
313 314
		}
	}
T
Trond Myklebust 已提交
315
	if (array->eof_index >= 0) {
316
		status = -EBADCOOKIE;
317 318
		if (*desc->dir_cookie == array->last_cookie)
			desc->eof = 1;
319
	}
320
out:
B
Bryan Schumaker 已提交
321 322 323 324 325 326 327
	return status;
}

static
int nfs_readdir_search_array(nfs_readdir_descriptor_t *desc)
{
	struct nfs_cache_array *array;
T
Trond Myklebust 已提交
328
	int status;
B
Bryan Schumaker 已提交
329 330 331 332 333 334 335 336 337 338 339 340

	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
341
	if (status == -EAGAIN) {
342
		desc->last_cookie = array->last_cookie;
343
		desc->current_index += array->size;
T
Trond Myklebust 已提交
344 345
		desc->page_index++;
	}
B
Bryan Schumaker 已提交
346 347 348 349 350 351 352
	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
353
int nfs_readdir_xdr_filler(struct page **pages, nfs_readdir_descriptor_t *desc,
B
Bryan Schumaker 已提交
354
			struct nfs_entry *entry, struct file *file, struct inode *inode)
L
Linus Torvalds 已提交
355
{
356 357
	struct nfs_open_dir_context *ctx = file->private_data;
	struct rpc_cred	*cred = ctx->cred;
358
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
359 360 361 362
	int		error;

 again:
	timestamp = jiffies;
363
	gencount = nfs_inc_attr_generation_counter();
B
Bryan Schumaker 已提交
364
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, entry->cookie, pages,
L
Linus Torvalds 已提交
365 366 367 368 369
					  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
370
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
371 372 373 374 375
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
376
	desc->timestamp = timestamp;
377
	desc->gencount = gencount;
B
Bryan Schumaker 已提交
378 379
error:
	return error;
L
Linus Torvalds 已提交
380 381
}

382 383
static int xdr_decode(nfs_readdir_descriptor_t *desc,
		      struct nfs_entry *entry, struct xdr_stream *xdr)
L
Linus Torvalds 已提交
384
{
385
	int error;
L
Linus Torvalds 已提交
386

387 388 389
	error = desc->decode(xdr, entry, desc->plus);
	if (error)
		return error;
B
Bryan Schumaker 已提交
390 391 392
	entry->fattr->time_start = desc->timestamp;
	entry->fattr->gencount = desc->gencount;
	return 0;
L
Linus Torvalds 已提交
393 394
}

B
Bryan Schumaker 已提交
395 396 397 398 399
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
400
	if (nfs_compare_fh(entry->fh, NFS_FH(dentry->d_inode)) != 0)
B
Bryan Schumaker 已提交
401 402 403 404 405 406
		goto different;
	return 1;
different:
	return 0;
}

407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429
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
430 431 432
static
void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
{
433
	struct qstr filename = QSTR_INIT(entry->name, entry->len);
434 435
	struct dentry *dentry;
	struct dentry *alias;
B
Bryan Schumaker 已提交
436 437 438
	struct inode *dir = parent->d_inode;
	struct inode *inode;

439 440 441 442 443 444 445
	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
446

447
	dentry = d_lookup(parent, &filename);
B
Bryan Schumaker 已提交
448 449 450 451 452
	if (dentry != NULL) {
		if (nfs_same_file(dentry, entry)) {
			nfs_refresh_inode(dentry->d_inode, entry->fattr);
			goto out;
		} else {
453 454
			if (d_invalidate(dentry) != 0)
				goto out;
B
Bryan Schumaker 已提交
455 456 457 458 459
			dput(dentry);
		}
	}

	dentry = d_alloc(parent, &filename);
460 461 462
	if (dentry == NULL)
		return;

B
Bryan Schumaker 已提交
463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
	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
480 481
/* Perform conversion from xdr to cache array */
static
482
int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *entry,
483
				struct page **xdr_pages, struct page *page, unsigned int buflen)
L
Linus Torvalds 已提交
484
{
485
	struct xdr_stream stream;
486
	struct xdr_buf buf;
487
	struct page *scratch;
B
Bryan Schumaker 已提交
488
	struct nfs_cache_array *array;
489 490
	unsigned int count = 0;
	int status;
491

492 493 494
	scratch = alloc_page(GFP_KERNEL);
	if (scratch == NULL)
		return -ENOMEM;
495

496
	xdr_init_decode_pages(&stream, &buf, xdr_pages, buflen);
497
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
B
Bryan Schumaker 已提交
498 499 500

	do {
		status = xdr_decode(desc, entry, &stream);
501 502 503
		if (status != 0) {
			if (status == -EAGAIN)
				status = 0;
B
Bryan Schumaker 已提交
504
			break;
505
		}
B
Bryan Schumaker 已提交
506

507 508
		count++;

T
Trond Myklebust 已提交
509
		if (desc->plus != 0)
B
Bryan Schumaker 已提交
510
			nfs_prime_dcache(desc->file->f_path.dentry, entry);
511 512 513 514

		status = nfs_readdir_add_to_array(entry, page);
		if (status != 0)
			break;
B
Bryan Schumaker 已提交
515 516
	} while (!entry->eof);

T
Trond Myklebust 已提交
517
	if (count == 0 || (status == -EBADCOOKIE && entry->eof != 0)) {
B
Bryan Schumaker 已提交
518
		array = nfs_readdir_get_array(page);
519 520 521 522
		if (!IS_ERR(array)) {
			array->eof_index = array->size;
			status = 0;
			nfs_readdir_release_array(page);
523 524
		} else
			status = PTR_ERR(array);
L
Linus Torvalds 已提交
525
	}
526 527

	put_page(scratch);
528
	return status;
B
Bryan Schumaker 已提交
529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550
}

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
551
int nfs_readdir_large_page(struct page **pages, unsigned int npages)
B
Bryan Schumaker 已提交
552 553 554 555 556 557 558 559 560
{
	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;
	}
561
	return 0;
B
Bryan Schumaker 已提交
562 563 564

out_freepages:
	nfs_readdir_free_pagearray(pages, i);
565
	return -ENOMEM;
L
Linus Torvalds 已提交
566 567
}

B
Bryan Schumaker 已提交
568 569
static
int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode)
570
{
B
Bryan Schumaker 已提交
571 572
	struct page *pages[NFS_MAX_READDIR_PAGES];
	void *pages_ptr = NULL;
B
Bryan Schumaker 已提交
573 574 575
	struct nfs_entry entry;
	struct file	*file = desc->file;
	struct nfs_cache_array *array;
576
	int status = -ENOMEM;
B
Bryan Schumaker 已提交
577
	unsigned int array_size = ARRAY_SIZE(pages);
B
Bryan Schumaker 已提交
578 579

	entry.prev_cookie = 0;
580
	entry.cookie = desc->last_cookie;
B
Bryan Schumaker 已提交
581 582 583
	entry.eof = 0;
	entry.fh = nfs_alloc_fhandle();
	entry.fattr = nfs_alloc_fattr();
584
	entry.server = NFS_SERVER(inode);
B
Bryan Schumaker 已提交
585 586
	if (entry.fh == NULL || entry.fattr == NULL)
		goto out;
587

B
Bryan Schumaker 已提交
588
	array = nfs_readdir_get_array(page);
589 590 591 592
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
593 594
	memset(array, 0, sizeof(struct nfs_cache_array));
	array->eof_index = -1;
595

596 597
	status = nfs_readdir_large_page(pages, array_size);
	if (status < 0)
B
Bryan Schumaker 已提交
598 599
		goto out_release_array;
	do {
600
		unsigned int pglen;
B
Bryan Schumaker 已提交
601
		status = nfs_readdir_xdr_filler(pages, desc, &entry, file, inode);
602

B
Bryan Schumaker 已提交
603
		if (status < 0)
604
			break;
605
		pglen = status;
606
		status = nfs_readdir_page_filler(desc, &entry, pages, page, pglen);
607 608 609 610 611 612
		if (status < 0) {
			if (status == -ENOSPC)
				status = 0;
			break;
		}
	} while (array->eof_index < 0);
B
Bryan Schumaker 已提交
613

B
Bryan Schumaker 已提交
614
	nfs_readdir_free_large_page(pages_ptr, pages, array_size);
B
Bryan Schumaker 已提交
615 616 617 618 619
out_release_array:
	nfs_readdir_release_array(page);
out:
	nfs_free_fattr(entry.fattr);
	nfs_free_fhandle(entry.fh);
620 621 622 623
	return status;
}

/*
B
Bryan Schumaker 已提交
624 625 626 627
 * 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
628
 */
B
Bryan Schumaker 已提交
629 630
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page)
L
Linus Torvalds 已提交
631
{
A
Al Viro 已提交
632
	struct inode	*inode = file_inode(desc->file);
633
	int ret;
L
Linus Torvalds 已提交
634

635 636
	ret = nfs_readdir_xdr_to_array(desc, page, inode);
	if (ret < 0)
B
Bryan Schumaker 已提交
637 638
		goto error;
	SetPageUptodate(page);
L
Linus Torvalds 已提交
639

B
Bryan Schumaker 已提交
640 641 642
	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
643
	}
B
Bryan Schumaker 已提交
644 645 646 647
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
648
	return ret;
B
Bryan Schumaker 已提交
649
}
L
Linus Torvalds 已提交
650

B
Bryan Schumaker 已提交
651 652 653
static
void cache_page_release(nfs_readdir_descriptor_t *desc)
{
654 655
	if (!desc->page->mapping)
		nfs_readdir_clear_array(desc->page);
B
Bryan Schumaker 已提交
656 657 658 659 660 661 662
	page_cache_release(desc->page);
	desc->page = NULL;
}

static
struct page *get_cache_page(nfs_readdir_descriptor_t *desc)
{
A
Al Viro 已提交
663
	return read_cache_page(file_inode(desc->file)->i_mapping,
B
Bryan Schumaker 已提交
664
			desc->page_index, (filler_t *)nfs_readdir_filler, desc);
L
Linus Torvalds 已提交
665 666 667
}

/*
B
Bryan Schumaker 已提交
668
 * Returns 0 if desc->dir_cookie was found on page desc->page_index
L
Linus Torvalds 已提交
669
 */
B
Bryan Schumaker 已提交
670 671 672 673 674 675 676 677 678 679
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
680 681
	if (res != 0)
		cache_page_release(desc);
B
Bryan Schumaker 已提交
682 683 684 685
	return res;
}

/* Search for desc->dir_cookie from the beginning of the page cache */
L
Linus Torvalds 已提交
686 687 688
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
689
	int res;
B
Bryan Schumaker 已提交
690

691
	if (desc->page_index == 0) {
692
		desc->current_index = 0;
693 694
		desc->last_cookie = 0;
	}
T
Trond Myklebust 已提交
695
	do {
B
Bryan Schumaker 已提交
696
		res = find_cache_page(desc);
T
Trond Myklebust 已提交
697
	} while (res == -EAGAIN);
L
Linus Torvalds 已提交
698 699 700 701 702 703 704 705 706 707 708
	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
709 710 711
	int i = 0;
	int res = 0;
	struct nfs_cache_array *array = NULL;
712 713
	struct nfs_open_dir_context *ctx = file->private_data;

B
Bryan Schumaker 已提交
714
	array = nfs_readdir_get_array(desc->page);
715 716 717 718
	if (IS_ERR(array)) {
		res = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
719 720

	for (i = desc->cache_entry_index; i < array->size; i++) {
721
		struct nfs_cache_array_entry *ent;
L
Linus Torvalds 已提交
722

723 724
		ent = &array->array[i];
		if (filldir(dirent, ent->string.name, ent->string.len,
725 726
		    file->f_pos, nfs_compat_user_ino64(ent->ino),
		    ent->d_type) < 0) {
727
			desc->eof = 1;
L
Linus Torvalds 已提交
728
			break;
729
		}
730
		file->f_pos++;
B
Bryan Schumaker 已提交
731 732 733 734
		if (i < (array->size-1))
			*desc->dir_cookie = array->array[i+1].cookie;
		else
			*desc->dir_cookie = array->last_cookie;
735 736
		if (ctx->duped != 0)
			ctx->duped = 1;
L
Linus Torvalds 已提交
737
	}
T
Trond Myklebust 已提交
738
	if (array->eof_index >= 0)
739
		desc->eof = 1;
B
Bryan Schumaker 已提交
740 741

	nfs_readdir_release_array(desc->page);
742
out:
B
Bryan Schumaker 已提交
743
	cache_page_release(desc);
C
Chuck Lever 已提交
744 745
	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
746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766
	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;
A
Al Viro 已提交
767
	struct inode *inode = file_inode(desc->file);
768
	struct nfs_open_dir_context *ctx = desc->file->private_data;
L
Linus Torvalds 已提交
769

C
Chuck Lever 已提交
770 771
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
772 773 774 775 776 777

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
B
Bryan Schumaker 已提交
778

779
	desc->page_index = 0;
780
	desc->last_cookie = *desc->dir_cookie;
781
	desc->page = page;
782
	ctx->duped = 0;
783

784 785
	status = nfs_readdir_xdr_to_array(desc, page, inode);
	if (status < 0)
L
Linus Torvalds 已提交
786 787 788 789 790
		goto out_release;

	status = nfs_do_filldir(desc, dirent, filldir);

 out:
C
Chuck Lever 已提交
791
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
792
			__func__, status);
L
Linus Torvalds 已提交
793 794
	return status;
 out_release:
B
Bryan Schumaker 已提交
795
	cache_page_release(desc);
L
Linus Torvalds 已提交
796 797 798
	goto out;
}

799 800 801
/* 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
802 803 804
 */
static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
805
	struct dentry	*dentry = filp->f_path.dentry;
L
Linus Torvalds 已提交
806 807 808
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
809
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
T
Trond Myklebust 已提交
810
	int res;
L
Linus Torvalds 已提交
811

C
Chuck Lever 已提交
812
	dfprintk(FILE, "NFS: readdir(%s/%s) starting at cookie %llu\n",
C
Chuck Lever 已提交
813 814
			dentry->d_parent->d_name.name, dentry->d_name.name,
			(long long)filp->f_pos);
C
Chuck Lever 已提交
815 816
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
817
	/*
818
	 * filp->f_pos points to the dirent entry number.
819
	 * *desc->dir_cookie has the cookie for the next entry. We have
820 821
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
822 823 824 825
	 */
	memset(desc, 0, sizeof(*desc));

	desc->file = filp;
826
	desc->dir_cookie = &dir_ctx->dir_cookie;
L
Linus Torvalds 已提交
827
	desc->decode = NFS_PROTO(inode)->decode_dirent;
828
	desc->plus = nfs_use_readdirplus(inode, filp) ? 1 : 0;
L
Linus Torvalds 已提交
829

T
Trond Myklebust 已提交
830
	nfs_block_sillyrename(dentry);
831
	res = nfs_revalidate_mapping(inode, filp->f_mapping);
T
Trond Myklebust 已提交
832 833 834
	if (res < 0)
		goto out;

T
Trond Myklebust 已提交
835
	do {
L
Linus Torvalds 已提交
836
		res = readdir_search_pagecache(desc);
837

L
Linus Torvalds 已提交
838
		if (res == -EBADCOOKIE) {
839
			res = 0;
L
Linus Torvalds 已提交
840
			/* This means either end of directory */
B
Bryan Schumaker 已提交
841
			if (*desc->dir_cookie && desc->eof == 0) {
L
Linus Torvalds 已提交
842 843
				/* Or that the server has 'lost' a cookie */
				res = uncached_readdir(desc, dirent, filldir);
844
				if (res == 0)
L
Linus Torvalds 已提交
845 846 847 848 849
					continue;
			}
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
850
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
851
			nfs_zap_caches(inode);
852
			desc->page_index = 0;
L
Linus Torvalds 已提交
853
			desc->plus = 0;
B
Bryan Schumaker 已提交
854
			desc->eof = 0;
L
Linus Torvalds 已提交
855 856 857 858 859 860
			continue;
		}
		if (res < 0)
			break;

		res = nfs_do_filldir(desc, dirent, filldir);
861
		if (res < 0)
L
Linus Torvalds 已提交
862
			break;
T
Trond Myklebust 已提交
863
	} while (!desc->eof);
T
Trond Myklebust 已提交
864
out:
T
Trond Myklebust 已提交
865
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
866 867
	if (res > 0)
		res = 0;
868
	dfprintk(FILE, "NFS: readdir(%s/%s) returns %d\n",
C
Chuck Lever 已提交
869 870 871
			dentry->d_parent->d_name.name, dentry->d_name.name,
			res);
	return res;
L
Linus Torvalds 已提交
872 873
}

874
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int whence)
875
{
876 877
	struct dentry *dentry = filp->f_path.dentry;
	struct inode *inode = dentry->d_inode;
878
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
879

C
Chuck Lever 已提交
880
	dfprintk(FILE, "NFS: llseek dir(%s/%s, %lld, %d)\n",
881 882
			dentry->d_parent->d_name.name,
			dentry->d_name.name,
883
			offset, whence);
884 885

	mutex_lock(&inode->i_mutex);
886
	switch (whence) {
887 888 889 890 891 892 893 894 895 896 897
		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;
898
		dir_ctx->dir_cookie = 0;
899
		dir_ctx->duped = 0;
900 901
	}
out:
902
	mutex_unlock(&inode->i_mutex);
903 904 905
	return offset;
}

L
Linus Torvalds 已提交
906 907 908 909
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
910 911
static int nfs_fsync_dir(struct file *filp, loff_t start, loff_t end,
			 int datasync)
L
Linus Torvalds 已提交
912
{
913
	struct dentry *dentry = filp->f_path.dentry;
914
	struct inode *inode = dentry->d_inode;
915

C
Chuck Lever 已提交
916
	dfprintk(FILE, "NFS: fsync dir(%s/%s) datasync %d\n",
C
Chuck Lever 已提交
917 918 919
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);

920
	mutex_lock(&inode->i_mutex);
921
	nfs_inc_stats(dentry->d_inode, NFSIOS_VFSFSYNC);
922
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
923 924 925
	return 0;
}

926 927 928 929 930 931 932 933 934 935 936 937
/**
 * 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)
{
938
	NFS_I(dir)->cache_change_attribute++;
939
}
B
Bryan Schumaker 已提交
940
EXPORT_SYMBOL_GPL(nfs_force_lookup_revalidate);
941

L
Linus Torvalds 已提交
942 943 944 945 946
/*
 * 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.
 */
947
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
948 949 950
{
	if (IS_ROOT(dentry))
		return 1;
951 952
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
953 954 955 956 957 958 959 960
	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
961 962
}

963 964 965 966
/*
 * Use intent information to check whether or not we're going to do
 * an O_EXCL create using this path component.
 */
967
static int nfs_is_exclusive_create(struct inode *dir, unsigned int flags)
968 969 970
{
	if (NFS_PROTO(dir)->version == 2)
		return 0;
971
	return flags & LOOKUP_EXCL;
972 973
}

974 975 976 977 978 979 980 981
/*
 * 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.
 *
 */
982
static
983
int nfs_lookup_verify_inode(struct inode *inode, unsigned int flags)
L
Linus Torvalds 已提交
984 985
{
	struct nfs_server *server = NFS_SERVER(inode);
986
	int ret;
L
Linus Torvalds 已提交
987

988
	if (IS_AUTOMOUNT(inode))
989
		return 0;
990
	/* VFS wants an on-the-wire revalidation */
991
	if (flags & LOOKUP_REVAL)
992 993
		goto out_force;
	/* This is an open(2) */
994 995
	if ((flags & LOOKUP_OPEN) && !(server->flags & NFS_MOUNT_NOCTO) &&
	    (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
996
		goto out_force;
997 998
out:
	return (inode->i_nlink == 0) ? -ENOENT : 0;
L
Linus Torvalds 已提交
999
out_force:
1000 1001 1002 1003
	ret = __nfs_revalidate_inode(server, inode);
	if (ret != 0)
		return ret;
	goto out;
L
Linus Torvalds 已提交
1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
}

/*
 * 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,
1015
		       unsigned int flags)
L
Linus Torvalds 已提交
1016 1017
{
	/* Don't revalidate a negative dentry if we're creating a new file */
1018
	if (flags & LOOKUP_CREATE)
L
Linus Torvalds 已提交
1019
		return 0;
1020 1021
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
	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.
 */
1036
static int nfs_lookup_revalidate(struct dentry *dentry, unsigned int flags)
L
Linus Torvalds 已提交
1037 1038 1039 1040
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
1041 1042
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1043 1044
	int error;

1045
	if (flags & LOOKUP_RCU)
1046 1047
		return -ECHILD;

L
Linus Torvalds 已提交
1048 1049
	parent = dget_parent(dentry);
	dir = parent->d_inode;
C
Chuck Lever 已提交
1050
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
1051 1052 1053
	inode = dentry->d_inode;

	if (!inode) {
1054
		if (nfs_neg_need_reval(dir, dentry, flags))
L
Linus Torvalds 已提交
1055
			goto out_bad;
1056
		goto out_valid_noent;
L
Linus Torvalds 已提交
1057 1058 1059
	}

	if (is_bad_inode(inode)) {
C
Chuck Lever 已提交
1060
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
1061
				__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1062
				dentry->d_name.name);
L
Linus Torvalds 已提交
1063 1064 1065
		goto out_bad;
	}

1066
	if (NFS_PROTO(dir)->have_delegation(inode, FMODE_READ))
1067 1068
		goto out_set_verifier;

L
Linus Torvalds 已提交
1069
	/* Force a full look up iff the parent directory has changed */
1070 1071
	if (!nfs_is_exclusive_create(dir, flags) && nfs_check_verifier(dir, dentry)) {
		if (nfs_lookup_verify_inode(inode, flags))
L
Linus Torvalds 已提交
1072 1073 1074 1075 1076 1077 1078
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

1079 1080 1081 1082 1083 1084
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

1085
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1086 1087
	if (error)
		goto out_bad;
1088
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
1089
		goto out_bad;
1090
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
1091 1092
		goto out_bad;

1093 1094
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1095
out_set_verifier:
1096
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1097
 out_valid:
1098 1099 1100
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);
 out_valid_noent:
L
Linus Torvalds 已提交
1101
	dput(parent);
C
Chuck Lever 已提交
1102
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n",
1103
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1104
			dentry->d_name.name);
L
Linus Torvalds 已提交
1105 1106 1107 1108
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
A
Al Viro 已提交
1109 1110
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1111
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
1112 1113 1114 1115 1116 1117
	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;
1118 1119
		if (dentry->d_flags & DCACHE_DISCONNECTED)
			goto out_valid;
L
Linus Torvalds 已提交
1120 1121 1122 1123
		shrink_dcache_parent(dentry);
	}
	d_drop(dentry);
	dput(parent);
C
Chuck Lever 已提交
1124
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
1125
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1126
			dentry->d_name.name);
L
Linus Torvalds 已提交
1127
	return 0;
1128 1129 1130 1131 1132 1133 1134 1135
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
1136 1137
}

1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
/*
 * A weaker form of d_revalidate for revalidating just the dentry->d_inode
 * when we don't really care about the dentry name. This is called when a
 * pathwalk ends on a dentry that was not found via a normal lookup in the
 * parent dir (e.g.: ".", "..", procfs symlinks or mountpoint traversals).
 *
 * In this situation, we just want to verify that the inode itself is OK
 * since the dentry might have changed on the server.
 */
static int nfs_weak_revalidate(struct dentry *dentry, unsigned int flags)
{
	int error;
	struct inode *inode = dentry->d_inode;

	/*
	 * I believe we can only get a negative dentry here in the case of a
	 * procfs-style symlink. Just assume it's correct for now, but we may
	 * eventually need to do something more here.
	 */
	if (!inode) {
		dfprintk(LOOKUPCACHE, "%s: %s/%s has negative inode\n",
				__func__, dentry->d_parent->d_name.name,
				dentry->d_name.name);
		return 1;
	}

	if (is_bad_inode(inode)) {
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
				__func__, dentry->d_parent->d_name.name,
				dentry->d_name.name);
		return 0;
	}

	error = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	dfprintk(LOOKUPCACHE, "NFS: %s: inode %lu is %s\n",
			__func__, inode->i_ino, error ? "invalid" : "valid");
	return !error;
}

L
Linus Torvalds 已提交
1177 1178 1179
/*
 * This is called from dput() when d_count is going to 0.
 */
N
Nick Piggin 已提交
1180
static int nfs_dentry_delete(const struct dentry *dentry)
L
Linus Torvalds 已提交
1181 1182 1183 1184 1185
{
	dfprintk(VFS, "NFS: dentry_delete(%s/%s, %x)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		dentry->d_flags);

1186 1187 1188 1189
	/* 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
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
	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;

}

T
Trond Myklebust 已提交
1203
/* Ensure that we revalidate inode->i_nlink */
1204 1205 1206
static void nfs_drop_nlink(struct inode *inode)
{
	spin_lock(&inode->i_lock);
T
Trond Myklebust 已提交
1207 1208 1209 1210
	/* drop the inode if we're reasonably sure this is the last link */
	if (inode->i_nlink == 1)
		clear_nlink(inode);
	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
1211 1212 1213
	spin_unlock(&inode->i_lock);
}

L
Linus Torvalds 已提交
1214 1215 1216 1217 1218 1219
/*
 * 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)
{
1220 1221 1222 1223
	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
1224
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
1225
		nfs_complete_unlink(dentry, inode);
T
Trond Myklebust 已提交
1226
		nfs_drop_nlink(inode);
L
Linus Torvalds 已提交
1227 1228 1229 1230
	}
	iput(inode);
}

1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
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
1242
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
1243
	.d_revalidate	= nfs_lookup_revalidate,
1244
	.d_weak_revalidate	= nfs_weak_revalidate,
L
Linus Torvalds 已提交
1245 1246
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1247
	.d_automount	= nfs_d_automount,
1248
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1249
};
B
Bryan Schumaker 已提交
1250
EXPORT_SYMBOL_GPL(nfs_dentry_operations);
L
Linus Torvalds 已提交
1251

1252
struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
L
Linus Torvalds 已提交
1253 1254
{
	struct dentry *res;
T
Trond Myklebust 已提交
1255
	struct dentry *parent;
L
Linus Torvalds 已提交
1256
	struct inode *inode = NULL;
1257 1258
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1259 1260 1261 1262
	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
1263
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
1264 1265 1266 1267 1268

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

1269 1270 1271 1272
	/*
	 * If we're doing an exclusive create, optimize away the lookup
	 * but don't hash the dentry.
	 */
A
Al Viro 已提交
1273
	if (nfs_is_exclusive_create(dir, flags)) {
1274 1275
		d_instantiate(dentry, NULL);
		res = NULL;
1276
		goto out;
1277
	}
L
Linus Torvalds 已提交
1278

1279 1280 1281 1282 1283 1284
	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
1285 1286 1287
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
	nfs_block_sillyrename(parent);
1288
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1289 1290 1291 1292
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
1293
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1294
	}
1295
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
N
Namhyung Kim 已提交
1296
	res = ERR_CAST(inode);
1297
	if (IS_ERR(res))
T
Trond Myklebust 已提交
1298
		goto out_unblock_sillyrename;
1299

1300 1301 1302
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);

L
Linus Torvalds 已提交
1303
no_entry:
1304
	res = d_materialise_unique(dentry, inode);
1305 1306
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
1307
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1308
		dentry = res;
1309
	}
L
Linus Torvalds 已提交
1310
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
1311 1312
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
L
Linus Torvalds 已提交
1313
out:
1314 1315
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1316 1317
	return res;
}
B
Bryan Schumaker 已提交
1318
EXPORT_SYMBOL_GPL(nfs_lookup);
L
Linus Torvalds 已提交
1319

B
Bryan Schumaker 已提交
1320
#if IS_ENABLED(CONFIG_NFS_V4)
1321
static int nfs4_lookup_revalidate(struct dentry *, unsigned int);
L
Linus Torvalds 已提交
1322

A
Al Viro 已提交
1323
const struct dentry_operations nfs4_dentry_operations = {
1324
	.d_revalidate	= nfs4_lookup_revalidate,
L
Linus Torvalds 已提交
1325 1326
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1327
	.d_automount	= nfs_d_automount,
1328
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1329
};
B
Bryan Schumaker 已提交
1330
EXPORT_SYMBOL_GPL(nfs4_dentry_operations);
L
Linus Torvalds 已提交
1331

1332 1333 1334 1335 1336 1337 1338 1339 1340 1341
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;
}

1342
static struct nfs_open_context *create_nfs_open_context(struct dentry *dentry, int open_flags)
1343
{
1344
	return alloc_nfs_open_context(dentry, flags_to_mode(open_flags));
1345 1346 1347 1348 1349 1350 1351 1352
}

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

A
Al Viro 已提交
1353 1354
static int nfs_finish_open(struct nfs_open_context *ctx,
			   struct dentry *dentry,
A
Al Viro 已提交
1355
			   struct file *file, unsigned open_flags,
A
Al Viro 已提交
1356
			   int *opened)
1357
{
1358 1359
	int err;

A
Al Viro 已提交
1360 1361
	err = finish_open(file, dentry, do_open, opened);
	if (err)
A
Al Viro 已提交
1362
		goto out;
A
Al Viro 已提交
1363
	nfs_file_set_open_context(file, ctx);
1364

1365 1366
out:
	put_nfs_open_context(ctx);
A
Al Viro 已提交
1367
	return err;
1368 1369
}

1370 1371 1372
int nfs_atomic_open(struct inode *dir, struct dentry *dentry,
		    struct file *file, unsigned open_flags,
		    umode_t mode, int *opened)
L
Linus Torvalds 已提交
1373
{
1374
	struct nfs_open_context *ctx;
1375 1376
	struct dentry *res;
	struct iattr attr = { .ia_valid = ATTR_OPEN };
1377
	struct inode *inode;
1378
	int err;
L
Linus Torvalds 已提交
1379

1380 1381 1382 1383
	/* Expect a negative dentry */
	BUG_ON(dentry->d_inode);

	dfprintk(VFS, "NFS: atomic_open(%s/%ld), %s\n",
C
Chuck Lever 已提交
1384 1385
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);

1386 1387 1388 1389 1390 1391 1392 1393
	/* NFS only supports OPEN on regular files */
	if ((open_flags & O_DIRECTORY)) {
		if (!d_unhashed(dentry)) {
			/*
			 * Hashed negative dentry with O_DIRECTORY: dentry was
			 * revalidated and is fine, no need to perform lookup
			 * again
			 */
A
Al Viro 已提交
1394
			return -ENOENT;
1395
		}
L
Linus Torvalds 已提交
1396
		goto no_open;
1397
	}
L
Linus Torvalds 已提交
1398

1399
	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
A
Al Viro 已提交
1400
		return -ENAMETOOLONG;
1401

1402
	if (open_flags & O_CREAT) {
1403
		attr.ia_valid |= ATTR_MODE;
1404 1405
		attr.ia_mode = mode & ~current_umask();
	}
1406 1407 1408
	if (open_flags & O_TRUNC) {
		attr.ia_valid |= ATTR_SIZE;
		attr.ia_size = 0;
1409 1410
	}

1411 1412 1413
	ctx = create_nfs_open_context(dentry, open_flags);
	err = PTR_ERR(ctx);
	if (IS_ERR(ctx))
A
Al Viro 已提交
1414
		goto out;
1415

1416
	nfs_block_sillyrename(dentry->d_parent);
1417
	inode = NFS_PROTO(dir)->open_context(dir, ctx, open_flags, &attr);
1418
	nfs_unblock_sillyrename(dentry->d_parent);
1419
	if (IS_ERR(inode)) {
1420
		put_nfs_open_context(ctx);
1421 1422 1423
		err = PTR_ERR(inode);
		switch (err) {
		case -ENOENT:
1424
			d_drop(dentry);
1425 1426 1427 1428 1429 1430 1431
			d_add(dentry, NULL);
			break;
		case -EISDIR:
		case -ENOTDIR:
			goto no_open;
		case -ELOOP:
			if (!(open_flags & O_NOFOLLOW))
1432
				goto no_open;
1433
			break;
L
Linus Torvalds 已提交
1434
			/* case -EINVAL: */
1435 1436
		default:
			break;
L
Linus Torvalds 已提交
1437
		}
A
Al Viro 已提交
1438
		goto out;
1439
	}
1440

1441
	err = nfs_finish_open(ctx, ctx->dentry, file, open_flags, opened);
A
Al Viro 已提交
1442 1443
out:
	return err;
1444

L
Linus Torvalds 已提交
1445
no_open:
A
Al Viro 已提交
1446
	res = nfs_lookup(dir, dentry, 0);
1447 1448
	err = PTR_ERR(res);
	if (IS_ERR(res))
A
Al Viro 已提交
1449
		goto out;
1450

A
Al Viro 已提交
1451
	return finish_no_open(file, res);
L
Linus Torvalds 已提交
1452
}
B
Bryan Schumaker 已提交
1453
EXPORT_SYMBOL_GPL(nfs_atomic_open);
L
Linus Torvalds 已提交
1454

1455
static int nfs4_lookup_revalidate(struct dentry *dentry, unsigned int flags)
L
Linus Torvalds 已提交
1456 1457
{
	struct dentry *parent = NULL;
N
Nick Piggin 已提交
1458
	struct inode *inode;
L
Linus Torvalds 已提交
1459
	struct inode *dir;
1460
	int ret = 0;
L
Linus Torvalds 已提交
1461

1462
	if (flags & LOOKUP_RCU)
N
Nick Piggin 已提交
1463 1464
		return -ECHILD;

1465
	if (!(flags & LOOKUP_OPEN) || (flags & LOOKUP_DIRECTORY))
1466 1467
		goto no_open;
	if (d_mountpoint(dentry))
T
Trond Myklebust 已提交
1468
		goto no_open;
1469 1470
	if (NFS_SB(dentry->d_sb)->caps & NFS_CAP_ATOMIC_OPEN_V1)
		goto no_open;
1471

1472
	inode = dentry->d_inode;
L
Linus Torvalds 已提交
1473 1474
	parent = dget_parent(dentry);
	dir = parent->d_inode;
1475

L
Linus Torvalds 已提交
1476 1477 1478
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1479
	if (inode == NULL) {
1480
		if (!nfs_neg_need_reval(dir, dentry, flags))
1481
			ret = 1;
L
Linus Torvalds 已提交
1482
		goto out;
1483 1484
	}

L
Linus Torvalds 已提交
1485 1486
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
T
Trond Myklebust 已提交
1487
		goto no_open_dput;
L
Linus Torvalds 已提交
1488
	/* We cannot do exclusive creation on a positive dentry */
1489
	if (flags & LOOKUP_EXCL)
T
Trond Myklebust 已提交
1490
		goto no_open_dput;
L
Linus Torvalds 已提交
1491

1492 1493
	/* Let f_op->open() actually open (and revalidate) the file */
	ret = 1;
1494

L
Linus Torvalds 已提交
1495 1496 1497
out:
	dput(parent);
	return ret;
1498

T
Trond Myklebust 已提交
1499
no_open_dput:
L
Linus Torvalds 已提交
1500
	dput(parent);
T
Trond Myklebust 已提交
1501
no_open:
1502
	return nfs_lookup_revalidate(dentry, flags);
1503 1504
}

L
Linus Torvalds 已提交
1505 1506 1507 1508 1509 1510 1511 1512
#endif /* CONFIG_NFSV4 */

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
				struct nfs_fattr *fattr)
{
1513 1514
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1515 1516 1517
	struct inode *inode;
	int error = -EACCES;

1518 1519
	d_drop(dentry);

L
Linus Torvalds 已提交
1520 1521
	/* We may have been initialized further down */
	if (dentry->d_inode)
1522
		goto out;
L
Linus Torvalds 已提交
1523
	if (fhandle->size == 0) {
1524
		error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1525
		if (error)
1526
			goto out_error;
L
Linus Torvalds 已提交
1527
	}
1528
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1529 1530
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1531
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr);
L
Linus Torvalds 已提交
1532
		if (error < 0)
1533
			goto out_error;
L
Linus Torvalds 已提交
1534 1535
	}
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
1536 1537
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1538 1539 1540 1541
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1542
	return 0;
1543 1544 1545 1546
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1547
}
B
Bryan Schumaker 已提交
1548
EXPORT_SYMBOL_GPL(nfs_instantiate);
L
Linus Torvalds 已提交
1549 1550 1551 1552 1553 1554 1555

/*
 * 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.
 */
1556
int nfs_create(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
1557
		umode_t mode, bool excl)
L
Linus Torvalds 已提交
1558 1559
{
	struct iattr attr;
A
Al Viro 已提交
1560
	int open_flags = excl ? O_CREAT | O_EXCL : O_CREAT;
L
Linus Torvalds 已提交
1561 1562
	int error;

C
Chuck Lever 已提交
1563 1564
	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
1565 1566 1567 1568

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

1569
	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags);
L
Linus Torvalds 已提交
1570 1571 1572 1573 1574 1575 1576
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1577
EXPORT_SYMBOL_GPL(nfs_create);
L
Linus Torvalds 已提交
1578 1579 1580 1581

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1582
int
A
Al Viro 已提交
1583
nfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t rdev)
L
Linus Torvalds 已提交
1584 1585 1586 1587
{
	struct iattr attr;
	int status;

C
Chuck Lever 已提交
1588 1589
	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
1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604

	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;
}
B
Bryan Schumaker 已提交
1605
EXPORT_SYMBOL_GPL(nfs_mknod);
L
Linus Torvalds 已提交
1606 1607 1608 1609

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1610
int nfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
1611 1612 1613 1614
{
	struct iattr attr;
	int error;

C
Chuck Lever 已提交
1615 1616
	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
1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628

	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;
}
B
Bryan Schumaker 已提交
1629
EXPORT_SYMBOL_GPL(nfs_mkdir);
L
Linus Torvalds 已提交
1630

1631 1632 1633 1634 1635 1636
static void nfs_dentry_handle_enoent(struct dentry *dentry)
{
	if (dentry->d_inode != NULL && !d_unhashed(dentry))
		d_delete(dentry);
}

1637
int nfs_rmdir(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1638 1639 1640
{
	int error;

C
Chuck Lever 已提交
1641 1642
	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
1643 1644 1645 1646

	error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
	/* Ensure the VFS deletes this inode */
	if (error == 0 && dentry->d_inode != NULL)
1647
		clear_nlink(dentry->d_inode);
1648 1649
	else if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1650 1651 1652

	return error;
}
B
Bryan Schumaker 已提交
1653
EXPORT_SYMBOL_GPL(nfs_rmdir);
L
Linus Torvalds 已提交
1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677

/*
 * 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) {
1678
		NFS_PROTO(inode)->return_delegation(inode);
L
Linus Torvalds 已提交
1679 1680
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		if (error == 0)
1681
			nfs_drop_nlink(inode);
L
Linus Torvalds 已提交
1682 1683
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1684 1685
	if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1686 1687 1688 1689 1690 1691 1692 1693 1694
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.
 */
1695
int nfs_unlink(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1696 1697 1698 1699 1700 1701 1702 1703
{
	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
1704
	if (dentry->d_count > 1) {
L
Linus Torvalds 已提交
1705
		spin_unlock(&dentry->d_lock);
T
Trond Myklebust 已提交
1706 1707
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1708 1709 1710 1711 1712 1713 1714 1715 1716
		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);
1717
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1718 1719 1720 1721 1722
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1723
EXPORT_SYMBOL_GPL(nfs_unlink);
L
Linus Torvalds 已提交
1724

1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739
/*
 * 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.
 */
1740
int nfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
L
Linus Torvalds 已提交
1741
{
1742 1743 1744
	struct pagevec lru_pvec;
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1745
	struct iattr attr;
1746
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1747 1748 1749 1750 1751
	int error;

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

1752 1753
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1754

1755 1756
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1757

1758
	page = alloc_page(GFP_HIGHUSER);
1759
	if (!page)
1760 1761
		return -ENOMEM;

1762
	kaddr = kmap_atomic(page);
1763 1764 1765
	memcpy(kaddr, symname, pathlen);
	if (pathlen < PAGE_SIZE)
		memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen);
1766
	kunmap_atomic(kaddr);
1767

1768
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1769 1770 1771 1772
	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
1773
		d_drop(dentry);
1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784
		__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)) {
1785
		pagevec_add(&lru_pvec, page);
1786
		pagevec_lru_add_file(&lru_pvec);
1787 1788 1789 1790 1791 1792
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1793
}
B
Bryan Schumaker 已提交
1794
EXPORT_SYMBOL_GPL(nfs_symlink);
L
Linus Torvalds 已提交
1795

1796
int
L
Linus Torvalds 已提交
1797 1798 1799 1800 1801 1802 1803 1804 1805
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);

1806
	NFS_PROTO(inode)->return_delegation(inode);
1807

1808
	d_drop(dentry);
L
Linus Torvalds 已提交
1809
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1810
	if (error == 0) {
A
Al Viro 已提交
1811
		ihold(inode);
1812
		d_add(dentry, inode);
1813
	}
L
Linus Torvalds 已提交
1814 1815
	return error;
}
B
Bryan Schumaker 已提交
1816
EXPORT_SYMBOL_GPL(nfs_link);
L
Linus Torvalds 已提交
1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841

/*
 * 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.
 */
1842
int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
L
Linus Torvalds 已提交
1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
		      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
1853
		 new_dentry->d_count);
L
Linus Torvalds 已提交
1854 1855

	/*
M
Miklos Szeredi 已提交
1856 1857 1858 1859
	 * 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
1860
	 */
1861 1862 1863 1864 1865 1866 1867 1868 1869
	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
1870

N
Nick Piggin 已提交
1871
		if (new_dentry->d_count > 2) {
1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
			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);
1882
			if (err)
1883
				goto out;
1884 1885

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
1886
			rehash = NULL;
1887
			new_inode = NULL;
1888
		}
1889
	}
L
Linus Torvalds 已提交
1890

1891
	NFS_PROTO(old_inode)->return_delegation(old_inode);
1892
	if (new_inode != NULL)
1893
		NFS_PROTO(new_inode)->return_delegation(new_inode);
L
Linus Torvalds 已提交
1894 1895 1896

	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
1897
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
1898 1899 1900 1901
out:
	if (rehash)
		d_rehash(rehash);
	if (!error) {
1902 1903
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
1904
		d_move(old_dentry, new_dentry);
1905 1906
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
1907 1908
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
1909 1910 1911 1912 1913 1914

	/* new dentry created? */
	if (dentry)
		dput(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1915
EXPORT_SYMBOL_GPL(nfs_rename);
L
Linus Torvalds 已提交
1916

1917 1918 1919 1920
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

1921 1922 1923 1924
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
1925 1926 1927
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
1928 1929
}

1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940
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);
	}
}

1941 1942
int nfs_access_cache_shrinker(struct shrinker *shrink,
			      struct shrink_control *sc)
1943 1944
{
	LIST_HEAD(head);
1945
	struct nfs_inode *nfsi, *next;
1946
	struct nfs_access_entry *cache;
1947 1948
	int nr_to_scan = sc->nr_to_scan;
	gfp_t gfp_mask = sc->gfp_mask;
1949

1950 1951
	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
		return (nr_to_scan == 0) ? 0 : -1;
1952

1953
	spin_lock(&nfs_access_lru_lock);
1954
	list_for_each_entry_safe(nfsi, next, &nfs_access_lru_list, access_cache_inode_lru) {
1955 1956 1957 1958
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
1959
		inode = &nfsi->vfs_inode;
1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
		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);
1973
			smp_mb__before_clear_bit();
1974
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
1975
			smp_mb__after_clear_bit();
1976
		}
1977
		spin_unlock(&inode->i_lock);
1978 1979
	}
	spin_unlock(&nfs_access_lru_lock);
1980
	nfs_access_free_list(&head);
1981 1982 1983
	return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure;
}

1984
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
1985
{
1986
	struct rb_root *root_node = &nfsi->access_cache;
1987
	struct rb_node *n;
1988 1989 1990 1991 1992 1993
	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);
1994
		list_move(&entry->lru, head);
1995 1996
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
1997 1998
}

1999
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
2000
{
2001 2002 2003 2004
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
2005
	/* Remove from global LRU init */
2006 2007
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
2008 2009
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

2010
	spin_lock(&inode->i_lock);
2011 2012 2013 2014
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
2015
}
B
Bryan Schumaker 已提交
2016
EXPORT_SYMBOL_GPL(nfs_access_zap_cache);
L
Linus Torvalds 已提交
2017

2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031
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
2032
	}
2033 2034 2035
	return NULL;
}

2036
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
2037 2038 2039 2040 2041
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

2042
	spin_lock(&inode->i_lock);
2043 2044 2045 2046 2047
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
2048
	if (!nfs_have_delegated_attributes(inode) &&
2049
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
2050 2051 2052 2053
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
2054
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
2055 2056 2057 2058 2059 2060
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
2061
	list_del(&cache->lru);
2062 2063 2064 2065
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
2066 2067
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
2068 2069 2070 2071 2072
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
2073 2074
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092
	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);
2093
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
2094
	spin_unlock(&inode->i_lock);
2095 2096 2097
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
2098 2099
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
2100 2101 2102 2103
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

2104
void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
2105 2106 2107 2108 2109
{
	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
2110
	cache->jiffies = set->jiffies;
2111
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
2112
	cache->mask = set->mask;
2113 2114

	nfs_access_add_rbtree(inode, cache);
2115 2116 2117 2118 2119 2120 2121

	/* 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 */
2122
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
2123
		spin_lock(&nfs_access_lru_lock);
2124 2125 2126
		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);
2127 2128
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
2129
}
2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143
EXPORT_SYMBOL_GPL(nfs_access_add_cache);

void nfs_access_set_mask(struct nfs_access_entry *entry, u32 access_result)
{
	entry->mask = 0;
	if (access_result & NFS4_ACCESS_READ)
		entry->mask |= MAY_READ;
	if (access_result &
	    (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
		entry->mask |= MAY_WRITE;
	if (access_result & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
		entry->mask |= MAY_EXEC;
}
EXPORT_SYMBOL_GPL(nfs_access_set_mask);
L
Linus Torvalds 已提交
2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158

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);
2159 2160 2161 2162 2163 2164
	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
2165
		return status;
2166
	}
L
Linus Torvalds 已提交
2167 2168
	nfs_access_add_cache(inode, &cache);
out:
2169
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2170 2171 2172 2173
		return 0;
	return -EACCES;
}

2174 2175 2176 2177
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

2178 2179 2180 2181 2182 2183 2184 2185 2186 2187
	if (openflags & __FMODE_EXEC) {
		/* ONLY check exec rights */
		mask = MAY_EXEC;
	} else {
		if ((openflags & O_ACCMODE) != O_WRONLY)
			mask |= MAY_READ;
		if ((openflags & O_ACCMODE) != O_RDONLY)
			mask |= MAY_WRITE;
	}

2188 2189 2190 2191 2192 2193 2194
	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));
}
B
Bryan Schumaker 已提交
2195
EXPORT_SYMBOL_GPL(nfs_may_open);
2196

2197
int nfs_permission(struct inode *inode, int mask)
L
Linus Torvalds 已提交
2198 2199 2200 2201
{
	struct rpc_cred *cred;
	int res = 0;

2202
	if (mask & MAY_NOT_BLOCK)
2203 2204
		return -ECHILD;

C
Chuck Lever 已提交
2205 2206
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

2207
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2208 2209
		goto out;
	/* Is this sys_access() ? */
E
Eric Paris 已提交
2210
	if (mask & (MAY_ACCESS | MAY_CHDIR))
L
Linus Torvalds 已提交
2211 2212 2213 2214 2215 2216 2217 2218
		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)
2219 2220
					&& (mask & MAY_OPEN)
					&& !(mask & MAY_EXEC))
L
Linus Torvalds 已提交
2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235
				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;

2236
	cred = rpc_lookup_cred();
L
Linus Torvalds 已提交
2237 2238 2239 2240 2241 2242
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
out:
2243 2244 2245
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

C
Chuck Lever 已提交
2246 2247
	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
2248 2249 2250 2251
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
2252
		res = generic_permission(inode, mask);
C
Chuck Lever 已提交
2253
	goto out;
L
Linus Torvalds 已提交
2254
}
B
Bryan Schumaker 已提交
2255
EXPORT_SYMBOL_GPL(nfs_permission);
L
Linus Torvalds 已提交
2256 2257 2258 2259 2260 2261 2262

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