dir.c 57.3 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
/*
 *  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>
36
#include <linux/swap.h>
A
Alexey Dobriyan 已提交
37
#include <linux/sched.h>
38
#include <linux/kmemleak.h>
39
#include <linux/xattr.h>
L
Linus Torvalds 已提交
40 41

#include "delegation.h"
C
Chuck Lever 已提交
42
#include "iostat.h"
43
#include "internal.h"
44
#include "fscache.h"
L
Linus Torvalds 已提交
45 46 47 48

/* #define NFS_DEBUG_VERBOSE 1 */

static int nfs_opendir(struct inode *, struct file *);
49
static int nfs_closedir(struct inode *, struct file *);
A
Al Viro 已提交
50
static int nfs_readdir(struct file *, struct dir_context *);
51
static int nfs_fsync_dir(struct file *, loff_t, loff_t, int);
52
static loff_t nfs_llseek_dir(struct file *, loff_t, int);
53
static void nfs_readdir_clear_array(struct page*);
L
Linus Torvalds 已提交
54

55
const struct file_operations nfs_dir_operations = {
56
	.llseek		= nfs_llseek_dir,
L
Linus Torvalds 已提交
57
	.read		= generic_read_dir,
A
Al Viro 已提交
58
	.iterate	= nfs_readdir,
L
Linus Torvalds 已提交
59
	.open		= nfs_opendir,
60
	.release	= nfs_closedir,
L
Linus Torvalds 已提交
61 62 63
	.fsync		= nfs_fsync_dir,
};

64 65
const struct address_space_operations nfs_dir_aops = {
	.freepage = nfs_readdir_clear_array,
B
Bryan Schumaker 已提交
66 67
};

68
static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct inode *dir, struct rpc_cred *cred)
69 70 71 72
{
	struct nfs_open_dir_context *ctx;
	ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
	if (ctx != NULL) {
73
		ctx->duped = 0;
74
		ctx->attr_gencount = NFS_I(dir)->attr_gencount;
75
		ctx->dir_cookie = 0;
76
		ctx->dup_cookie = 0;
77
		ctx->cred = get_rpccred(cred);
78 79 80
		return ctx;
	}
	return  ERR_PTR(-ENOMEM);
81 82 83 84 85 86 87 88
}

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
89 90 91 92 93 94
/*
 * Open file
 */
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
95 96 97
	int res = 0;
	struct nfs_open_dir_context *ctx;
	struct rpc_cred *cred;
L
Linus Torvalds 已提交
98

C
Chuck Lever 已提交
99
	dfprintk(FILE, "NFS: open dir(%s/%s)\n",
100 101 102 103
			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
104

105 106 107
	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return PTR_ERR(cred);
108
	ctx = alloc_nfs_open_dir_context(inode, cred);
109 110 111 112 113
	if (IS_ERR(ctx)) {
		res = PTR_ERR(ctx);
		goto out;
	}
	filp->private_data = ctx;
114 115 116 117 118 119 120
	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);
	}
121 122
out:
	put_rpccred(cred);
L
Linus Torvalds 已提交
123 124 125
	return res;
}

126 127 128 129 130 131 132
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
133 134 135 136
struct nfs_cache_array_entry {
	u64 cookie;
	u64 ino;
	struct qstr string;
137
	unsigned char d_type;
B
Bryan Schumaker 已提交
138 139 140
};

struct nfs_cache_array {
141
	int size;
B
Bryan Schumaker 已提交
142 143 144 145 146
	int eof_index;
	u64 last_cookie;
	struct nfs_cache_array_entry array[0];
};

147
typedef int (*decode_dirent_t)(struct xdr_stream *, struct nfs_entry *, int);
L
Linus Torvalds 已提交
148 149 150
typedef struct {
	struct file	*file;
	struct page	*page;
A
Al Viro 已提交
151
	struct dir_context *ctx;
L
Linus Torvalds 已提交
152
	unsigned long	page_index;
153
	u64		*dir_cookie;
154
	u64		last_cookie;
155
	loff_t		current_index;
L
Linus Torvalds 已提交
156
	decode_dirent_t	decode;
B
Bryan Schumaker 已提交
157

158
	unsigned long	timestamp;
159
	unsigned long	gencount;
B
Bryan Schumaker 已提交
160 161 162
	unsigned int	cache_entry_index;
	unsigned int	plus:1;
	unsigned int	eof:1;
L
Linus Torvalds 已提交
163 164
} nfs_readdir_descriptor_t;

B
Bryan Schumaker 已提交
165 166
/*
 * The caller is responsible for calling nfs_readdir_release_array(page)
L
Linus Torvalds 已提交
167 168
 */
static
B
Bryan Schumaker 已提交
169 170
struct nfs_cache_array *nfs_readdir_get_array(struct page *page)
{
171
	void *ptr;
B
Bryan Schumaker 已提交
172 173
	if (page == NULL)
		return ERR_PTR(-EIO);
174 175 176 177
	ptr = kmap(page);
	if (ptr == NULL)
		return ERR_PTR(-ENOMEM);
	return ptr;
B
Bryan Schumaker 已提交
178 179 180 181 182 183 184 185 186 187 188 189
}

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

/*
 * we are freeing strings created by nfs_add_to_readdir_array()
 */
static
190
void nfs_readdir_clear_array(struct page *page)
B
Bryan Schumaker 已提交
191
{
192
	struct nfs_cache_array *array;
B
Bryan Schumaker 已提交
193
	int i;
194

195
	array = kmap_atomic(page);
B
Bryan Schumaker 已提交
196 197
	for (i = 0; i < array->size; i++)
		kfree(array->array[i].string.name);
198
	kunmap_atomic(array);
B
Bryan Schumaker 已提交
199 200 201 202 203 204 205 206
}

/*
 * 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
207
int nfs_readdir_make_qstr(struct qstr *string, const char *name, unsigned int len)
B
Bryan Schumaker 已提交
208 209 210
{
	string->len = len;
	string->name = kmemdup(name, len, GFP_KERNEL);
211 212
	if (string->name == NULL)
		return -ENOMEM;
213 214 215 216 217
	/*
	 * 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);
218 219
	string->hash = full_name_hash(name, len);
	return 0;
B
Bryan Schumaker 已提交
220 221 222 223 224 225
}

static
int nfs_readdir_add_to_array(struct nfs_entry *entry, struct page *page)
{
	struct nfs_cache_array *array = nfs_readdir_get_array(page);
226 227 228
	struct nfs_cache_array_entry *cache_entry;
	int ret;

B
Bryan Schumaker 已提交
229 230
	if (IS_ERR(array))
		return PTR_ERR(array);
231 232 233 234

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

	/* Check that this entry lies within the page bounds */
235
	ret = -ENOSPC;
236
	if ((char *)&cache_entry[1] - (char *)page_address(page) > PAGE_SIZE)
237
		goto out;
B
Bryan Schumaker 已提交
238

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

static
int nfs_readdir_search_for_pos(struct nfs_cache_array *array, nfs_readdir_descriptor_t *desc)
{
A
Al Viro 已提交
257
	loff_t diff = desc->ctx->pos - desc->current_index;
B
Bryan Schumaker 已提交
258 259 260 261 262
	unsigned int index;

	if (diff < 0)
		goto out_eof;
	if (diff >= array->size) {
263
		if (array->eof_index >= 0)
B
Bryan Schumaker 已提交
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280
			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;
281
	loff_t new_pos;
B
Bryan Schumaker 已提交
282 283 284 285
	int status = -EAGAIN;

	for (i = 0; i < array->size; i++) {
		if (array->array[i].cookie == *desc->dir_cookie) {
A
Al Viro 已提交
286
			struct nfs_inode *nfsi = NFS_I(file_inode(desc->file));
287 288
			struct nfs_open_dir_context *ctx = desc->file->private_data;

289
			new_pos = desc->current_index + i;
290 291 292 293
			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;
A
Al Viro 已提交
294
			} else if (new_pos < desc->ctx->pos) {
295 296 297 298 299
				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.  "
300
								"The file: %s has duplicate cookie %llu\n",
301 302
								desc->file->f_dentry->d_parent->d_name.name,
								desc->file->f_dentry->d_name.name,
303
								array->array[i].string.name,
304 305 306 307 308
								*desc->dir_cookie);
					}
					status = -ELOOP;
					goto out;
				}
309
				ctx->dup_cookie = *desc->dir_cookie;
310
				ctx->duped = -1;
311
			}
A
Al Viro 已提交
312
			desc->ctx->pos = new_pos;
B
Bryan Schumaker 已提交
313
			desc->cache_entry_index = i;
T
Trond Myklebust 已提交
314
			return 0;
B
Bryan Schumaker 已提交
315 316
		}
	}
T
Trond Myklebust 已提交
317
	if (array->eof_index >= 0) {
318
		status = -EBADCOOKIE;
319 320
		if (*desc->dir_cookie == array->last_cookie)
			desc->eof = 1;
321
	}
322
out:
B
Bryan Schumaker 已提交
323 324 325 326 327 328 329
	return status;
}

static
int nfs_readdir_search_array(nfs_readdir_descriptor_t *desc)
{
	struct nfs_cache_array *array;
T
Trond Myklebust 已提交
330
	int status;
B
Bryan Schumaker 已提交
331 332 333 334 335 336 337 338 339 340 341 342

	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
343
	if (status == -EAGAIN) {
344
		desc->last_cookie = array->last_cookie;
345
		desc->current_index += array->size;
T
Trond Myklebust 已提交
346 347
		desc->page_index++;
	}
B
Bryan Schumaker 已提交
348 349 350 351 352 353 354
	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
355
int nfs_readdir_xdr_filler(struct page **pages, nfs_readdir_descriptor_t *desc,
B
Bryan Schumaker 已提交
356
			struct nfs_entry *entry, struct file *file, struct inode *inode)
L
Linus Torvalds 已提交
357
{
358 359
	struct nfs_open_dir_context *ctx = file->private_data;
	struct rpc_cred	*cred = ctx->cred;
360
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
361 362 363 364
	int		error;

 again:
	timestamp = jiffies;
365
	gencount = nfs_inc_attr_generation_counter();
B
Bryan Schumaker 已提交
366
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, entry->cookie, pages,
L
Linus Torvalds 已提交
367 368 369 370 371
					  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
372
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
373 374 375 376 377
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
378
	desc->timestamp = timestamp;
379
	desc->gencount = gencount;
B
Bryan Schumaker 已提交
380 381
error:
	return error;
L
Linus Torvalds 已提交
382 383
}

384 385
static int xdr_decode(nfs_readdir_descriptor_t *desc,
		      struct nfs_entry *entry, struct xdr_stream *xdr)
L
Linus Torvalds 已提交
386
{
387
	int error;
L
Linus Torvalds 已提交
388

389 390 391
	error = desc->decode(xdr, entry, desc->plus);
	if (error)
		return error;
B
Bryan Schumaker 已提交
392 393 394
	entry->fattr->time_start = desc->timestamp;
	entry->fattr->gencount = desc->gencount;
	return 0;
L
Linus Torvalds 已提交
395 396
}

B
Bryan Schumaker 已提交
397 398 399 400 401
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
402
	if (nfs_compare_fh(entry->fh, NFS_FH(dentry->d_inode)) != 0)
B
Bryan Schumaker 已提交
403 404 405 406 407 408
		goto different;
	return 1;
different:
	return 0;
}

409
static
A
Al Viro 已提交
410
bool nfs_use_readdirplus(struct inode *dir, struct dir_context *ctx)
411 412 413 414 415
{
	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;
A
Al Viro 已提交
416
	if (ctx->pos == 0)
417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
		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
432 433 434
static
void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
{
435
	struct qstr filename = QSTR_INIT(entry->name, entry->len);
436 437
	struct dentry *dentry;
	struct dentry *alias;
B
Bryan Schumaker 已提交
438 439
	struct inode *dir = parent->d_inode;
	struct inode *inode;
440
	int status;
B
Bryan Schumaker 已提交
441

442 443 444 445 446 447 448
	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
449

450
	dentry = d_lookup(parent, &filename);
B
Bryan Schumaker 已提交
451 452
	if (dentry != NULL) {
		if (nfs_same_file(dentry, entry)) {
453
			nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
454 455 456
			status = nfs_refresh_inode(dentry->d_inode, entry->fattr);
			if (!status)
				nfs_setsecurity(dentry->d_inode, entry->fattr, entry->label);
B
Bryan Schumaker 已提交
457 458
			goto out;
		} else {
459 460
			if (d_invalidate(dentry) != 0)
				goto out;
B
Bryan Schumaker 已提交
461 462 463 464 465
			dput(dentry);
		}
	}

	dentry = d_alloc(parent, &filename);
466 467 468
	if (dentry == NULL)
		return;

469
	inode = nfs_fhget(dentry->d_sb, entry->fh, entry->fattr, entry->label);
B
Bryan Schumaker 已提交
470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485
	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
486 487
/* Perform conversion from xdr to cache array */
static
488
int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *entry,
489
				struct page **xdr_pages, struct page *page, unsigned int buflen)
L
Linus Torvalds 已提交
490
{
491
	struct xdr_stream stream;
492
	struct xdr_buf buf;
493
	struct page *scratch;
B
Bryan Schumaker 已提交
494
	struct nfs_cache_array *array;
495 496
	unsigned int count = 0;
	int status;
497

498 499 500
	scratch = alloc_page(GFP_KERNEL);
	if (scratch == NULL)
		return -ENOMEM;
501

502
	xdr_init_decode_pages(&stream, &buf, xdr_pages, buflen);
503
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
B
Bryan Schumaker 已提交
504 505 506

	do {
		status = xdr_decode(desc, entry, &stream);
507 508 509
		if (status != 0) {
			if (status == -EAGAIN)
				status = 0;
B
Bryan Schumaker 已提交
510
			break;
511
		}
B
Bryan Schumaker 已提交
512

513 514
		count++;

T
Trond Myklebust 已提交
515
		if (desc->plus != 0)
B
Bryan Schumaker 已提交
516
			nfs_prime_dcache(desc->file->f_path.dentry, entry);
517 518 519 520

		status = nfs_readdir_add_to_array(entry, page);
		if (status != 0)
			break;
B
Bryan Schumaker 已提交
521 522
	} while (!entry->eof);

T
Trond Myklebust 已提交
523
	if (count == 0 || (status == -EBADCOOKIE && entry->eof != 0)) {
B
Bryan Schumaker 已提交
524
		array = nfs_readdir_get_array(page);
525 526 527 528
		if (!IS_ERR(array)) {
			array->eof_index = array->size;
			status = 0;
			nfs_readdir_release_array(page);
529 530
		} else
			status = PTR_ERR(array);
L
Linus Torvalds 已提交
531
	}
532 533

	put_page(scratch);
534
	return status;
B
Bryan Schumaker 已提交
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556
}

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
557
int nfs_readdir_large_page(struct page **pages, unsigned int npages)
B
Bryan Schumaker 已提交
558 559 560 561 562 563 564 565 566
{
	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;
	}
567
	return 0;
B
Bryan Schumaker 已提交
568 569 570

out_freepages:
	nfs_readdir_free_pagearray(pages, i);
571
	return -ENOMEM;
L
Linus Torvalds 已提交
572 573
}

B
Bryan Schumaker 已提交
574 575
static
int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode)
576
{
B
Bryan Schumaker 已提交
577 578
	struct page *pages[NFS_MAX_READDIR_PAGES];
	void *pages_ptr = NULL;
B
Bryan Schumaker 已提交
579 580 581
	struct nfs_entry entry;
	struct file	*file = desc->file;
	struct nfs_cache_array *array;
582
	int status = -ENOMEM;
B
Bryan Schumaker 已提交
583
	unsigned int array_size = ARRAY_SIZE(pages);
B
Bryan Schumaker 已提交
584 585

	entry.prev_cookie = 0;
586
	entry.cookie = desc->last_cookie;
B
Bryan Schumaker 已提交
587 588 589
	entry.eof = 0;
	entry.fh = nfs_alloc_fhandle();
	entry.fattr = nfs_alloc_fattr();
590
	entry.server = NFS_SERVER(inode);
B
Bryan Schumaker 已提交
591 592
	if (entry.fh == NULL || entry.fattr == NULL)
		goto out;
593

594 595 596 597 598 599
	entry.label = nfs4_label_alloc(NFS_SERVER(inode), GFP_NOWAIT);
	if (IS_ERR(entry.label)) {
		status = PTR_ERR(entry.label);
		goto out;
	}

B
Bryan Schumaker 已提交
600
	array = nfs_readdir_get_array(page);
601 602
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
603
		goto out_label_free;
604
	}
B
Bryan Schumaker 已提交
605 606
	memset(array, 0, sizeof(struct nfs_cache_array));
	array->eof_index = -1;
607

608 609
	status = nfs_readdir_large_page(pages, array_size);
	if (status < 0)
B
Bryan Schumaker 已提交
610 611
		goto out_release_array;
	do {
612
		unsigned int pglen;
B
Bryan Schumaker 已提交
613
		status = nfs_readdir_xdr_filler(pages, desc, &entry, file, inode);
614

B
Bryan Schumaker 已提交
615
		if (status < 0)
616
			break;
617
		pglen = status;
618
		status = nfs_readdir_page_filler(desc, &entry, pages, page, pglen);
619 620 621 622 623 624
		if (status < 0) {
			if (status == -ENOSPC)
				status = 0;
			break;
		}
	} while (array->eof_index < 0);
B
Bryan Schumaker 已提交
625

B
Bryan Schumaker 已提交
626
	nfs_readdir_free_large_page(pages_ptr, pages, array_size);
B
Bryan Schumaker 已提交
627 628
out_release_array:
	nfs_readdir_release_array(page);
629 630
out_label_free:
	nfs4_label_free(entry.label);
B
Bryan Schumaker 已提交
631 632 633
out:
	nfs_free_fattr(entry.fattr);
	nfs_free_fhandle(entry.fh);
634 635 636 637
	return status;
}

/*
B
Bryan Schumaker 已提交
638 639 640 641
 * 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
642
 */
B
Bryan Schumaker 已提交
643 644
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page)
L
Linus Torvalds 已提交
645
{
A
Al Viro 已提交
646
	struct inode	*inode = file_inode(desc->file);
647
	int ret;
L
Linus Torvalds 已提交
648

649 650
	ret = nfs_readdir_xdr_to_array(desc, page, inode);
	if (ret < 0)
B
Bryan Schumaker 已提交
651 652
		goto error;
	SetPageUptodate(page);
L
Linus Torvalds 已提交
653

B
Bryan Schumaker 已提交
654 655 656
	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
657
	}
B
Bryan Schumaker 已提交
658 659 660 661
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
662
	return ret;
B
Bryan Schumaker 已提交
663
}
L
Linus Torvalds 已提交
664

B
Bryan Schumaker 已提交
665 666 667
static
void cache_page_release(nfs_readdir_descriptor_t *desc)
{
668 669
	if (!desc->page->mapping)
		nfs_readdir_clear_array(desc->page);
B
Bryan Schumaker 已提交
670 671 672 673 674 675 676
	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
677
	return read_cache_page(file_inode(desc->file)->i_mapping,
B
Bryan Schumaker 已提交
678
			desc->page_index, (filler_t *)nfs_readdir_filler, desc);
L
Linus Torvalds 已提交
679 680 681
}

/*
B
Bryan Schumaker 已提交
682
 * Returns 0 if desc->dir_cookie was found on page desc->page_index
L
Linus Torvalds 已提交
683
 */
B
Bryan Schumaker 已提交
684 685 686 687 688 689 690 691 692 693
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
694 695
	if (res != 0)
		cache_page_release(desc);
B
Bryan Schumaker 已提交
696 697 698 699
	return res;
}

/* Search for desc->dir_cookie from the beginning of the page cache */
L
Linus Torvalds 已提交
700 701 702
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
703
	int res;
B
Bryan Schumaker 已提交
704

705
	if (desc->page_index == 0) {
706
		desc->current_index = 0;
707 708
		desc->last_cookie = 0;
	}
T
Trond Myklebust 已提交
709
	do {
B
Bryan Schumaker 已提交
710
		res = find_cache_page(desc);
T
Trond Myklebust 已提交
711
	} while (res == -EAGAIN);
L
Linus Torvalds 已提交
712 713 714 715 716 717 718
	return res;
}

/*
 * Once we've found the start of the dirent within a page: fill 'er up...
 */
static 
A
Al Viro 已提交
719
int nfs_do_filldir(nfs_readdir_descriptor_t *desc)
L
Linus Torvalds 已提交
720 721
{
	struct file	*file = desc->file;
B
Bryan Schumaker 已提交
722 723 724
	int i = 0;
	int res = 0;
	struct nfs_cache_array *array = NULL;
725 726
	struct nfs_open_dir_context *ctx = file->private_data;

B
Bryan Schumaker 已提交
727
	array = nfs_readdir_get_array(desc->page);
728 729 730 731
	if (IS_ERR(array)) {
		res = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
732 733

	for (i = desc->cache_entry_index; i < array->size; i++) {
734
		struct nfs_cache_array_entry *ent;
L
Linus Torvalds 已提交
735

736
		ent = &array->array[i];
A
Al Viro 已提交
737 738
		if (!dir_emit(desc->ctx, ent->string.name, ent->string.len,
		    nfs_compat_user_ino64(ent->ino), ent->d_type)) {
739
			desc->eof = 1;
L
Linus Torvalds 已提交
740
			break;
741
		}
A
Al Viro 已提交
742
		desc->ctx->pos++;
B
Bryan Schumaker 已提交
743 744 745 746
		if (i < (array->size-1))
			*desc->dir_cookie = array->array[i+1].cookie;
		else
			*desc->dir_cookie = array->last_cookie;
747 748
		if (ctx->duped != 0)
			ctx->duped = 1;
L
Linus Torvalds 已提交
749
	}
T
Trond Myklebust 已提交
750
	if (array->eof_index >= 0)
751
		desc->eof = 1;
B
Bryan Schumaker 已提交
752 753

	nfs_readdir_release_array(desc->page);
754
out:
B
Bryan Schumaker 已提交
755
	cache_page_release(desc);
C
Chuck Lever 已提交
756 757
	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
758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773
	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
A
Al Viro 已提交
774
int uncached_readdir(nfs_readdir_descriptor_t *desc)
L
Linus Torvalds 已提交
775 776 777
{
	struct page	*page = NULL;
	int		status;
A
Al Viro 已提交
778
	struct inode *inode = file_inode(desc->file);
779
	struct nfs_open_dir_context *ctx = desc->file->private_data;
L
Linus Torvalds 已提交
780

C
Chuck Lever 已提交
781 782
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
783 784 785 786 787 788

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
B
Bryan Schumaker 已提交
789

790
	desc->page_index = 0;
791
	desc->last_cookie = *desc->dir_cookie;
792
	desc->page = page;
793
	ctx->duped = 0;
794

795 796
	status = nfs_readdir_xdr_to_array(desc, page, inode);
	if (status < 0)
L
Linus Torvalds 已提交
797 798
		goto out_release;

A
Al Viro 已提交
799
	status = nfs_do_filldir(desc);
L
Linus Torvalds 已提交
800 801

 out:
C
Chuck Lever 已提交
802
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
803
			__func__, status);
L
Linus Torvalds 已提交
804 805
	return status;
 out_release:
B
Bryan Schumaker 已提交
806
	cache_page_release(desc);
L
Linus Torvalds 已提交
807 808 809
	goto out;
}

810 811 812
/* 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
813
 */
A
Al Viro 已提交
814
static int nfs_readdir(struct file *file, struct dir_context *ctx)
L
Linus Torvalds 已提交
815
{
A
Al Viro 已提交
816
	struct dentry	*dentry = file->f_path.dentry;
L
Linus Torvalds 已提交
817 818 819
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
A
Al Viro 已提交
820
	struct nfs_open_dir_context *dir_ctx = file->private_data;
821
	int res = 0;
L
Linus Torvalds 已提交
822

C
Chuck Lever 已提交
823
	dfprintk(FILE, "NFS: readdir(%s/%s) starting at cookie %llu\n",
C
Chuck Lever 已提交
824
			dentry->d_parent->d_name.name, dentry->d_name.name,
A
Al Viro 已提交
825
			(long long)ctx->pos);
C
Chuck Lever 已提交
826 827
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
828
	/*
A
Al Viro 已提交
829
	 * ctx->pos points to the dirent entry number.
830
	 * *desc->dir_cookie has the cookie for the next entry. We have
831 832
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
833 834 835
	 */
	memset(desc, 0, sizeof(*desc));

A
Al Viro 已提交
836 837
	desc->file = file;
	desc->ctx = ctx;
838
	desc->dir_cookie = &dir_ctx->dir_cookie;
L
Linus Torvalds 已提交
839
	desc->decode = NFS_PROTO(inode)->decode_dirent;
A
Al Viro 已提交
840
	desc->plus = nfs_use_readdirplus(inode, ctx) ? 1 : 0;
L
Linus Torvalds 已提交
841

T
Trond Myklebust 已提交
842
	nfs_block_sillyrename(dentry);
843 844
	if (ctx->pos == 0 || nfs_attribute_cache_expired(inode))
		res = nfs_revalidate_mapping(inode, file->f_mapping);
T
Trond Myklebust 已提交
845 846 847
	if (res < 0)
		goto out;

T
Trond Myklebust 已提交
848
	do {
L
Linus Torvalds 已提交
849
		res = readdir_search_pagecache(desc);
850

L
Linus Torvalds 已提交
851
		if (res == -EBADCOOKIE) {
852
			res = 0;
L
Linus Torvalds 已提交
853
			/* This means either end of directory */
B
Bryan Schumaker 已提交
854
			if (*desc->dir_cookie && desc->eof == 0) {
L
Linus Torvalds 已提交
855
				/* Or that the server has 'lost' a cookie */
A
Al Viro 已提交
856
				res = uncached_readdir(desc);
857
				if (res == 0)
L
Linus Torvalds 已提交
858 859 860 861 862
					continue;
			}
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
863
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
864
			nfs_zap_caches(inode);
865
			desc->page_index = 0;
L
Linus Torvalds 已提交
866
			desc->plus = 0;
B
Bryan Schumaker 已提交
867
			desc->eof = 0;
L
Linus Torvalds 已提交
868 869 870 871 872
			continue;
		}
		if (res < 0)
			break;

A
Al Viro 已提交
873
		res = nfs_do_filldir(desc);
874
		if (res < 0)
L
Linus Torvalds 已提交
875
			break;
T
Trond Myklebust 已提交
876
	} while (!desc->eof);
T
Trond Myklebust 已提交
877
out:
T
Trond Myklebust 已提交
878
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
879 880
	if (res > 0)
		res = 0;
881
	dfprintk(FILE, "NFS: readdir(%s/%s) returns %d\n",
C
Chuck Lever 已提交
882 883 884
			dentry->d_parent->d_name.name, dentry->d_name.name,
			res);
	return res;
L
Linus Torvalds 已提交
885 886
}

887
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int whence)
888
{
889 890
	struct dentry *dentry = filp->f_path.dentry;
	struct inode *inode = dentry->d_inode;
891
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
892

C
Chuck Lever 已提交
893
	dfprintk(FILE, "NFS: llseek dir(%s/%s, %lld, %d)\n",
894 895
			dentry->d_parent->d_name.name,
			dentry->d_name.name,
896
			offset, whence);
897 898

	mutex_lock(&inode->i_mutex);
899
	switch (whence) {
900 901 902 903 904 905 906 907 908 909 910
		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;
911
		dir_ctx->dir_cookie = 0;
912
		dir_ctx->duped = 0;
913 914
	}
out:
915
	mutex_unlock(&inode->i_mutex);
916 917 918
	return offset;
}

L
Linus Torvalds 已提交
919 920 921 922
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
923 924
static int nfs_fsync_dir(struct file *filp, loff_t start, loff_t end,
			 int datasync)
L
Linus Torvalds 已提交
925
{
926
	struct dentry *dentry = filp->f_path.dentry;
927
	struct inode *inode = dentry->d_inode;
928

C
Chuck Lever 已提交
929
	dfprintk(FILE, "NFS: fsync dir(%s/%s) datasync %d\n",
C
Chuck Lever 已提交
930 931 932
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);

933
	mutex_lock(&inode->i_mutex);
934
	nfs_inc_stats(dentry->d_inode, NFSIOS_VFSFSYNC);
935
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
936 937 938
	return 0;
}

939 940 941 942 943 944 945 946 947 948 949 950
/**
 * 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)
{
951
	NFS_I(dir)->cache_change_attribute++;
952
}
B
Bryan Schumaker 已提交
953
EXPORT_SYMBOL_GPL(nfs_force_lookup_revalidate);
954

L
Linus Torvalds 已提交
955 956 957 958 959
/*
 * 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.
 */
960
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
961 962 963
{
	if (IS_ROOT(dentry))
		return 1;
964 965
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
966 967 968 969 970 971 972 973
	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
974 975
}

976 977 978 979
/*
 * Use intent information to check whether or not we're going to do
 * an O_EXCL create using this path component.
 */
980
static int nfs_is_exclusive_create(struct inode *dir, unsigned int flags)
981 982 983
{
	if (NFS_PROTO(dir)->version == 2)
		return 0;
984
	return flags & LOOKUP_EXCL;
985 986
}

987 988 989 990 991 992 993 994
/*
 * 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.
 *
 */
995
static
996
int nfs_lookup_verify_inode(struct inode *inode, unsigned int flags)
L
Linus Torvalds 已提交
997 998
{
	struct nfs_server *server = NFS_SERVER(inode);
999
	int ret;
L
Linus Torvalds 已提交
1000

1001
	if (IS_AUTOMOUNT(inode))
1002
		return 0;
1003
	/* VFS wants an on-the-wire revalidation */
1004
	if (flags & LOOKUP_REVAL)
1005 1006
		goto out_force;
	/* This is an open(2) */
1007 1008
	if ((flags & LOOKUP_OPEN) && !(server->flags & NFS_MOUNT_NOCTO) &&
	    (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1009
		goto out_force;
1010 1011
out:
	return (inode->i_nlink == 0) ? -ENOENT : 0;
L
Linus Torvalds 已提交
1012
out_force:
1013 1014 1015 1016
	ret = __nfs_revalidate_inode(server, inode);
	if (ret != 0)
		return ret;
	goto out;
L
Linus Torvalds 已提交
1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
}

/*
 * 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,
1028
		       unsigned int flags)
L
Linus Torvalds 已提交
1029 1030
{
	/* Don't revalidate a negative dentry if we're creating a new file */
1031
	if (flags & LOOKUP_CREATE)
L
Linus Torvalds 已提交
1032
		return 0;
1033 1034
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
	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.
 */
1049
static int nfs_lookup_revalidate(struct dentry *dentry, unsigned int flags)
L
Linus Torvalds 已提交
1050 1051 1052 1053
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
1054 1055
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
1056
	struct nfs4_label *label = NULL;
L
Linus Torvalds 已提交
1057 1058
	int error;

1059
	if (flags & LOOKUP_RCU)
1060 1061
		return -ECHILD;

L
Linus Torvalds 已提交
1062 1063
	parent = dget_parent(dentry);
	dir = parent->d_inode;
C
Chuck Lever 已提交
1064
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
1065 1066 1067
	inode = dentry->d_inode;

	if (!inode) {
1068
		if (nfs_neg_need_reval(dir, dentry, flags))
L
Linus Torvalds 已提交
1069
			goto out_bad;
1070
		goto out_valid_noent;
L
Linus Torvalds 已提交
1071 1072 1073
	}

	if (is_bad_inode(inode)) {
C
Chuck Lever 已提交
1074
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
1075
				__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1076
				dentry->d_name.name);
L
Linus Torvalds 已提交
1077 1078 1079
		goto out_bad;
	}

1080
	if (NFS_PROTO(dir)->have_delegation(inode, FMODE_READ))
1081 1082
		goto out_set_verifier;

L
Linus Torvalds 已提交
1083
	/* Force a full look up iff the parent directory has changed */
1084 1085
	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
1086 1087 1088 1089 1090 1091 1092
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

1093 1094 1095 1096 1097 1098
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

1099 1100 1101 1102
	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_NOWAIT);
	if (IS_ERR(label))
		goto out_error;

1103
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, label);
L
Linus Torvalds 已提交
1104 1105
	if (error)
		goto out_bad;
1106
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
1107
		goto out_bad;
1108
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
1109 1110
		goto out_bad;

1111 1112
	nfs_setsecurity(inode, fattr, label);

1113 1114
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1115 1116
	nfs4_label_free(label);

1117
out_set_verifier:
1118
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1119
 out_valid:
1120 1121 1122
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);
 out_valid_noent:
L
Linus Torvalds 已提交
1123
	dput(parent);
C
Chuck Lever 已提交
1124
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\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 1128 1129 1130
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
A
Al Viro 已提交
1131 1132
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1133
	nfs4_label_free(label);
1134
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
1135 1136 1137 1138 1139 1140
	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;
1141 1142
		if (dentry->d_flags & DCACHE_DISCONNECTED)
			goto out_valid;
L
Linus Torvalds 已提交
1143 1144 1145 1146
		shrink_dcache_parent(dentry);
	}
	d_drop(dentry);
	dput(parent);
C
Chuck Lever 已提交
1147
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
1148
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1149
			dentry->d_name.name);
L
Linus Torvalds 已提交
1150
	return 0;
1151 1152 1153
out_error:
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1154
	nfs4_label_free(label);
1155 1156 1157 1158 1159
	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
1160 1161
}

1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200
/*
 * 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
1201 1202 1203
/*
 * This is called from dput() when d_count is going to 0.
 */
N
Nick Piggin 已提交
1204
static int nfs_dentry_delete(const struct dentry *dentry)
L
Linus Torvalds 已提交
1205 1206 1207 1208 1209
{
	dfprintk(VFS, "NFS: dentry_delete(%s/%s, %x)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		dentry->d_flags);

1210 1211 1212 1213
	/* 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
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
	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
1227
/* Ensure that we revalidate inode->i_nlink */
1228 1229 1230
static void nfs_drop_nlink(struct inode *inode)
{
	spin_lock(&inode->i_lock);
T
Trond Myklebust 已提交
1231 1232 1233 1234
	/* 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;
1235 1236 1237
	spin_unlock(&inode->i_lock);
}

L
Linus Torvalds 已提交
1238 1239 1240 1241 1242 1243
/*
 * 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)
{
1244 1245 1246 1247
	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
1248
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
1249
		nfs_complete_unlink(dentry, inode);
T
Trond Myklebust 已提交
1250
		nfs_drop_nlink(inode);
L
Linus Torvalds 已提交
1251 1252 1253 1254
	}
	iput(inode);
}

1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265
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
1266
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
1267
	.d_revalidate	= nfs_lookup_revalidate,
1268
	.d_weak_revalidate	= nfs_weak_revalidate,
L
Linus Torvalds 已提交
1269 1270
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1271
	.d_automount	= nfs_d_automount,
1272
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1273
};
B
Bryan Schumaker 已提交
1274
EXPORT_SYMBOL_GPL(nfs_dentry_operations);
L
Linus Torvalds 已提交
1275

1276
struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
L
Linus Torvalds 已提交
1277 1278
{
	struct dentry *res;
T
Trond Myklebust 已提交
1279
	struct dentry *parent;
L
Linus Torvalds 已提交
1280
	struct inode *inode = NULL;
1281 1282
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
1283
	struct nfs4_label *label = NULL;
L
Linus Torvalds 已提交
1284 1285 1286 1287
	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
1288
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
1289 1290 1291 1292 1293

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

1294 1295 1296 1297
	/*
	 * If we're doing an exclusive create, optimize away the lookup
	 * but don't hash the dentry.
	 */
A
Al Viro 已提交
1298
	if (nfs_is_exclusive_create(dir, flags)) {
1299 1300
		d_instantiate(dentry, NULL);
		res = NULL;
1301
		goto out;
1302
	}
L
Linus Torvalds 已提交
1303

1304 1305 1306 1307 1308 1309
	res = ERR_PTR(-ENOMEM);
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out;

1310 1311 1312 1313
	label = nfs4_label_alloc(NFS_SERVER(dir), GFP_NOWAIT);
	if (IS_ERR(label))
		goto out;

T
Trond Myklebust 已提交
1314 1315 1316
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
	nfs_block_sillyrename(parent);
1317
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, label);
L
Linus Torvalds 已提交
1318 1319 1320 1321
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
1322
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1323
	}
1324
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr, label);
N
Namhyung Kim 已提交
1325
	res = ERR_CAST(inode);
1326
	if (IS_ERR(res))
T
Trond Myklebust 已提交
1327
		goto out_unblock_sillyrename;
1328

1329 1330 1331
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);

L
Linus Torvalds 已提交
1332
no_entry:
1333
	res = d_materialise_unique(dentry, inode);
1334 1335
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
1336
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1337
		dentry = res;
1338
	}
L
Linus Torvalds 已提交
1339
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
1340 1341
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
1342
	nfs4_label_free(label);
L
Linus Torvalds 已提交
1343
out:
1344 1345
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1346 1347
	return res;
}
B
Bryan Schumaker 已提交
1348
EXPORT_SYMBOL_GPL(nfs_lookup);
L
Linus Torvalds 已提交
1349

B
Bryan Schumaker 已提交
1350
#if IS_ENABLED(CONFIG_NFS_V4)
1351
static int nfs4_lookup_revalidate(struct dentry *, unsigned int);
L
Linus Torvalds 已提交
1352

A
Al Viro 已提交
1353
const struct dentry_operations nfs4_dentry_operations = {
1354
	.d_revalidate	= nfs4_lookup_revalidate,
L
Linus Torvalds 已提交
1355 1356
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1357
	.d_automount	= nfs_d_automount,
1358
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1359
};
B
Bryan Schumaker 已提交
1360
EXPORT_SYMBOL_GPL(nfs4_dentry_operations);
L
Linus Torvalds 已提交
1361

1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
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;
}

1372
static struct nfs_open_context *create_nfs_open_context(struct dentry *dentry, int open_flags)
1373
{
1374
	return alloc_nfs_open_context(dentry, flags_to_mode(open_flags));
1375 1376 1377 1378 1379 1380 1381 1382
}

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
1383 1384
static int nfs_finish_open(struct nfs_open_context *ctx,
			   struct dentry *dentry,
A
Al Viro 已提交
1385
			   struct file *file, unsigned open_flags,
A
Al Viro 已提交
1386
			   int *opened)
1387
{
1388 1389
	int err;

A
Al Viro 已提交
1390 1391
	err = finish_open(file, dentry, do_open, opened);
	if (err)
A
Al Viro 已提交
1392
		goto out;
A
Al Viro 已提交
1393
	nfs_file_set_open_context(file, ctx);
1394

1395 1396
out:
	put_nfs_open_context(ctx);
A
Al Viro 已提交
1397
	return err;
1398 1399
}

1400 1401 1402
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
1403
{
1404
	struct nfs_open_context *ctx;
1405 1406
	struct dentry *res;
	struct iattr attr = { .ia_valid = ATTR_OPEN };
1407
	struct inode *inode;
1408
	int err;
L
Linus Torvalds 已提交
1409

1410 1411 1412 1413
	/* Expect a negative dentry */
	BUG_ON(dentry->d_inode);

	dfprintk(VFS, "NFS: atomic_open(%s/%ld), %s\n",
C
Chuck Lever 已提交
1414 1415
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);

1416 1417 1418 1419
	err = nfs_check_flags(open_flags);
	if (err)
		return err;

1420 1421 1422 1423 1424 1425 1426 1427
	/* 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
1428
			return -ENOENT;
1429
		}
L
Linus Torvalds 已提交
1430
		goto no_open;
1431
	}
L
Linus Torvalds 已提交
1432

1433
	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
A
Al Viro 已提交
1434
		return -ENAMETOOLONG;
1435

1436
	if (open_flags & O_CREAT) {
1437
		attr.ia_valid |= ATTR_MODE;
1438 1439
		attr.ia_mode = mode & ~current_umask();
	}
1440 1441 1442
	if (open_flags & O_TRUNC) {
		attr.ia_valid |= ATTR_SIZE;
		attr.ia_size = 0;
1443 1444
	}

1445 1446 1447
	ctx = create_nfs_open_context(dentry, open_flags);
	err = PTR_ERR(ctx);
	if (IS_ERR(ctx))
A
Al Viro 已提交
1448
		goto out;
1449

1450
	nfs_block_sillyrename(dentry->d_parent);
1451
	inode = NFS_PROTO(dir)->open_context(dir, ctx, open_flags, &attr);
1452
	nfs_unblock_sillyrename(dentry->d_parent);
1453
	if (IS_ERR(inode)) {
1454
		put_nfs_open_context(ctx);
1455 1456 1457
		err = PTR_ERR(inode);
		switch (err) {
		case -ENOENT:
1458
			d_drop(dentry);
1459 1460 1461 1462 1463 1464 1465
			d_add(dentry, NULL);
			break;
		case -EISDIR:
		case -ENOTDIR:
			goto no_open;
		case -ELOOP:
			if (!(open_flags & O_NOFOLLOW))
1466
				goto no_open;
1467
			break;
L
Linus Torvalds 已提交
1468
			/* case -EINVAL: */
1469 1470
		default:
			break;
L
Linus Torvalds 已提交
1471
		}
A
Al Viro 已提交
1472
		goto out;
1473
	}
1474

1475
	err = nfs_finish_open(ctx, ctx->dentry, file, open_flags, opened);
A
Al Viro 已提交
1476 1477
out:
	return err;
1478

L
Linus Torvalds 已提交
1479
no_open:
A
Al Viro 已提交
1480
	res = nfs_lookup(dir, dentry, 0);
1481 1482
	err = PTR_ERR(res);
	if (IS_ERR(res))
A
Al Viro 已提交
1483
		goto out;
1484

A
Al Viro 已提交
1485
	return finish_no_open(file, res);
L
Linus Torvalds 已提交
1486
}
B
Bryan Schumaker 已提交
1487
EXPORT_SYMBOL_GPL(nfs_atomic_open);
L
Linus Torvalds 已提交
1488

1489
static int nfs4_lookup_revalidate(struct dentry *dentry, unsigned int flags)
L
Linus Torvalds 已提交
1490 1491
{
	struct dentry *parent = NULL;
N
Nick Piggin 已提交
1492
	struct inode *inode;
L
Linus Torvalds 已提交
1493
	struct inode *dir;
1494
	int ret = 0;
L
Linus Torvalds 已提交
1495

1496
	if (flags & LOOKUP_RCU)
N
Nick Piggin 已提交
1497 1498
		return -ECHILD;

1499
	if (!(flags & LOOKUP_OPEN) || (flags & LOOKUP_DIRECTORY))
1500 1501
		goto no_open;
	if (d_mountpoint(dentry))
T
Trond Myklebust 已提交
1502
		goto no_open;
1503 1504
	if (NFS_SB(dentry->d_sb)->caps & NFS_CAP_ATOMIC_OPEN_V1)
		goto no_open;
1505

1506
	inode = dentry->d_inode;
L
Linus Torvalds 已提交
1507 1508
	parent = dget_parent(dentry);
	dir = parent->d_inode;
1509

L
Linus Torvalds 已提交
1510 1511 1512
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1513
	if (inode == NULL) {
1514
		if (!nfs_neg_need_reval(dir, dentry, flags))
1515
			ret = 1;
L
Linus Torvalds 已提交
1516
		goto out;
1517 1518
	}

L
Linus Torvalds 已提交
1519 1520
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
T
Trond Myklebust 已提交
1521
		goto no_open_dput;
L
Linus Torvalds 已提交
1522
	/* We cannot do exclusive creation on a positive dentry */
1523
	if (flags & LOOKUP_EXCL)
T
Trond Myklebust 已提交
1524
		goto no_open_dput;
L
Linus Torvalds 已提交
1525

1526 1527
	/* Let f_op->open() actually open (and revalidate) the file */
	ret = 1;
1528

L
Linus Torvalds 已提交
1529 1530 1531
out:
	dput(parent);
	return ret;
1532

T
Trond Myklebust 已提交
1533
no_open_dput:
L
Linus Torvalds 已提交
1534
	dput(parent);
T
Trond Myklebust 已提交
1535
no_open:
1536
	return nfs_lookup_revalidate(dentry, flags);
1537 1538
}

L
Linus Torvalds 已提交
1539 1540 1541 1542 1543 1544
#endif /* CONFIG_NFSV4 */

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
1545 1546
				struct nfs_fattr *fattr,
				struct nfs4_label *label)
L
Linus Torvalds 已提交
1547
{
1548 1549
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1550 1551 1552
	struct inode *inode;
	int error = -EACCES;

1553 1554
	d_drop(dentry);

L
Linus Torvalds 已提交
1555 1556
	/* We may have been initialized further down */
	if (dentry->d_inode)
1557
		goto out;
L
Linus Torvalds 已提交
1558
	if (fhandle->size == 0) {
1559
		error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, NULL);
L
Linus Torvalds 已提交
1560
		if (error)
1561
			goto out_error;
L
Linus Torvalds 已提交
1562
	}
1563
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1564 1565
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1566
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr, NULL);
L
Linus Torvalds 已提交
1567
		if (error < 0)
1568
			goto out_error;
L
Linus Torvalds 已提交
1569
	}
1570
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr, label);
1571 1572
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1573 1574 1575 1576
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1577
	return 0;
1578 1579 1580 1581
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1582
}
B
Bryan Schumaker 已提交
1583
EXPORT_SYMBOL_GPL(nfs_instantiate);
L
Linus Torvalds 已提交
1584 1585 1586 1587 1588 1589 1590

/*
 * 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.
 */
1591
int nfs_create(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
1592
		umode_t mode, bool excl)
L
Linus Torvalds 已提交
1593 1594
{
	struct iattr attr;
A
Al Viro 已提交
1595
	int open_flags = excl ? O_CREAT | O_EXCL : O_CREAT;
L
Linus Torvalds 已提交
1596 1597
	int error;

C
Chuck Lever 已提交
1598 1599
	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
1600 1601 1602 1603

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

1604
	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags);
L
Linus Torvalds 已提交
1605 1606 1607 1608 1609 1610 1611
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1612
EXPORT_SYMBOL_GPL(nfs_create);
L
Linus Torvalds 已提交
1613 1614 1615 1616

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1617
int
A
Al Viro 已提交
1618
nfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t rdev)
L
Linus Torvalds 已提交
1619 1620 1621 1622
{
	struct iattr attr;
	int status;

C
Chuck Lever 已提交
1623 1624
	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
1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639

	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
1640
EXPORT_SYMBOL_GPL(nfs_mknod);
L
Linus Torvalds 已提交
1641 1642 1643 1644

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1645
int nfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
1646 1647 1648 1649
{
	struct iattr attr;
	int error;

C
Chuck Lever 已提交
1650 1651
	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
1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663

	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
1664
EXPORT_SYMBOL_GPL(nfs_mkdir);
L
Linus Torvalds 已提交
1665

1666 1667 1668 1669 1670 1671
static void nfs_dentry_handle_enoent(struct dentry *dentry)
{
	if (dentry->d_inode != NULL && !d_unhashed(dentry))
		d_delete(dentry);
}

1672
int nfs_rmdir(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1673 1674 1675
{
	int error;

C
Chuck Lever 已提交
1676 1677
	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
1678 1679 1680 1681

	error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
	/* Ensure the VFS deletes this inode */
	if (error == 0 && dentry->d_inode != NULL)
1682
		clear_nlink(dentry->d_inode);
1683 1684
	else if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1685 1686 1687

	return error;
}
B
Bryan Schumaker 已提交
1688
EXPORT_SYMBOL_GPL(nfs_rmdir);
L
Linus Torvalds 已提交
1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712

/*
 * 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) {
1713
		NFS_PROTO(inode)->return_delegation(inode);
L
Linus Torvalds 已提交
1714 1715
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		if (error == 0)
1716
			nfs_drop_nlink(inode);
L
Linus Torvalds 已提交
1717 1718
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1719 1720
	if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1721 1722 1723 1724 1725 1726 1727 1728 1729
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.
 */
1730
int nfs_unlink(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1731 1732 1733 1734 1735 1736 1737 1738
{
	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);
A
Al Viro 已提交
1739
	if (d_count(dentry) > 1) {
L
Linus Torvalds 已提交
1740
		spin_unlock(&dentry->d_lock);
T
Trond Myklebust 已提交
1741 1742
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1743 1744 1745 1746 1747 1748 1749 1750 1751
		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);
1752
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1753 1754 1755 1756 1757
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1758
EXPORT_SYMBOL_GPL(nfs_unlink);
L
Linus Torvalds 已提交
1759

1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774
/*
 * 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.
 */
1775
int nfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
L
Linus Torvalds 已提交
1776
{
1777 1778
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1779
	struct iattr attr;
1780
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1781 1782 1783 1784 1785
	int error;

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

1786 1787
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1788

1789 1790
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1791

1792
	page = alloc_page(GFP_HIGHUSER);
1793
	if (!page)
1794 1795
		return -ENOMEM;

1796
	kaddr = kmap_atomic(page);
1797 1798 1799
	memcpy(kaddr, symname, pathlen);
	if (pathlen < PAGE_SIZE)
		memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen);
1800
	kunmap_atomic(kaddr);
1801

1802
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1803 1804 1805 1806
	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
1807
		d_drop(dentry);
1808 1809 1810 1811 1812 1813 1814 1815
		__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.
	 */
1816
	if (!add_to_page_cache_lru(page, dentry->d_inode->i_mapping, 0,
1817 1818 1819 1820 1821 1822 1823
							GFP_KERNEL)) {
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1824
}
B
Bryan Schumaker 已提交
1825
EXPORT_SYMBOL_GPL(nfs_symlink);
L
Linus Torvalds 已提交
1826

1827
int
L
Linus Torvalds 已提交
1828 1829 1830 1831 1832 1833 1834 1835 1836
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);

1837
	NFS_PROTO(inode)->return_delegation(inode);
1838

1839
	d_drop(dentry);
L
Linus Torvalds 已提交
1840
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1841
	if (error == 0) {
A
Al Viro 已提交
1842
		ihold(inode);
1843
		d_add(dentry, inode);
1844
	}
L
Linus Torvalds 已提交
1845 1846
	return error;
}
B
Bryan Schumaker 已提交
1847
EXPORT_SYMBOL_GPL(nfs_link);
L
Linus Torvalds 已提交
1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872

/*
 * 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.
 */
1873
int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
L
Linus Torvalds 已提交
1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
		      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,
A
Al Viro 已提交
1884
		 d_count(new_dentry));
L
Linus Torvalds 已提交
1885 1886

	/*
M
Miklos Szeredi 已提交
1887 1888 1889 1890
	 * 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
1891
	 */
1892 1893 1894 1895 1896 1897 1898 1899 1900
	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
1901

A
Al Viro 已提交
1902
		if (d_count(new_dentry) > 2) {
1903 1904 1905 1906 1907 1908 1909 1910 1911 1912
			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);
1913
			if (err)
1914
				goto out;
1915 1916

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
1917
			rehash = NULL;
1918
			new_inode = NULL;
1919
		}
1920
	}
L
Linus Torvalds 已提交
1921

1922
	NFS_PROTO(old_inode)->return_delegation(old_inode);
1923
	if (new_inode != NULL)
1924
		NFS_PROTO(new_inode)->return_delegation(new_inode);
L
Linus Torvalds 已提交
1925 1926 1927

	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
1928
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
1929 1930 1931 1932
out:
	if (rehash)
		d_rehash(rehash);
	if (!error) {
1933 1934
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
1935
		d_move(old_dentry, new_dentry);
1936 1937
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
1938 1939
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
1940 1941 1942 1943 1944 1945

	/* new dentry created? */
	if (dentry)
		dput(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1946
EXPORT_SYMBOL_GPL(nfs_rename);
L
Linus Torvalds 已提交
1947

1948 1949 1950 1951
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

1952 1953 1954 1955
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
1956 1957 1958
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
1959 1960
}

1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971
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);
	}
}

1972 1973
int nfs_access_cache_shrinker(struct shrinker *shrink,
			      struct shrink_control *sc)
1974 1975
{
	LIST_HEAD(head);
1976
	struct nfs_inode *nfsi, *next;
1977
	struct nfs_access_entry *cache;
1978 1979
	int nr_to_scan = sc->nr_to_scan;
	gfp_t gfp_mask = sc->gfp_mask;
1980

1981 1982
	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
		return (nr_to_scan == 0) ? 0 : -1;
1983

1984
	spin_lock(&nfs_access_lru_lock);
1985
	list_for_each_entry_safe(nfsi, next, &nfs_access_lru_list, access_cache_inode_lru) {
1986 1987 1988 1989
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
1990
		inode = &nfsi->vfs_inode;
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
		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);
2004
			smp_mb__before_clear_bit();
2005
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
2006
			smp_mb__after_clear_bit();
2007
		}
2008
		spin_unlock(&inode->i_lock);
2009 2010
	}
	spin_unlock(&nfs_access_lru_lock);
2011
	nfs_access_free_list(&head);
2012 2013 2014
	return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure;
}

2015
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
2016
{
2017
	struct rb_root *root_node = &nfsi->access_cache;
2018
	struct rb_node *n;
2019 2020 2021 2022 2023 2024
	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);
2025
		list_move(&entry->lru, head);
2026 2027
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
2028 2029
}

2030
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
2031
{
2032 2033 2034 2035
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
2036
	/* Remove from global LRU init */
2037 2038
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
2039 2040
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

2041
	spin_lock(&inode->i_lock);
2042 2043 2044 2045
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
2046
}
B
Bryan Schumaker 已提交
2047
EXPORT_SYMBOL_GPL(nfs_access_zap_cache);
L
Linus Torvalds 已提交
2048

2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062
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
2063
	}
2064 2065 2066
	return NULL;
}

2067
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
2068 2069 2070 2071 2072
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

2073
	spin_lock(&inode->i_lock);
2074 2075 2076 2077 2078
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
2079
	if (!nfs_have_delegated_attributes(inode) &&
2080
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
2081 2082 2083 2084
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
2085
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
2086 2087 2088 2089 2090 2091
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
2092
	list_del(&cache->lru);
2093 2094 2095 2096
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
2097 2098
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
2099 2100 2101 2102 2103
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
2104 2105
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123
	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);
2124
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
2125
	spin_unlock(&inode->i_lock);
2126 2127 2128
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
2129 2130
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
2131 2132 2133 2134
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

2135
void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
2136 2137 2138 2139 2140
{
	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
2141
	cache->jiffies = set->jiffies;
2142
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
2143
	cache->mask = set->mask;
2144 2145

	nfs_access_add_rbtree(inode, cache);
2146 2147 2148 2149 2150 2151 2152

	/* 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 */
2153
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
2154
		spin_lock(&nfs_access_lru_lock);
2155 2156 2157
		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);
2158 2159
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
2160
}
2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174
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
2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189

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);
2190 2191 2192 2193 2194 2195
	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
2196
		return status;
2197
	}
L
Linus Torvalds 已提交
2198 2199
	nfs_access_add_cache(inode, &cache);
out:
2200
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2201 2202 2203 2204
		return 0;
	return -EACCES;
}

2205 2206 2207 2208
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

2209 2210 2211 2212 2213 2214 2215 2216 2217 2218
	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;
	}

2219 2220 2221 2222 2223 2224 2225
	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
2226
EXPORT_SYMBOL_GPL(nfs_may_open);
2227

2228
int nfs_permission(struct inode *inode, int mask)
L
Linus Torvalds 已提交
2229 2230 2231 2232
{
	struct rpc_cred *cred;
	int res = 0;

2233
	if (mask & MAY_NOT_BLOCK)
2234 2235
		return -ECHILD;

C
Chuck Lever 已提交
2236 2237
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

2238
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2239 2240
		goto out;
	/* Is this sys_access() ? */
E
Eric Paris 已提交
2241
	if (mask & (MAY_ACCESS | MAY_CHDIR))
L
Linus Torvalds 已提交
2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261
		goto force_lookup;

	switch (inode->i_mode & S_IFMT) {
		case S_IFLNK:
			goto out;
		case S_IFREG:
			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;

2262
	cred = rpc_lookup_cred();
L
Linus Torvalds 已提交
2263 2264 2265 2266 2267 2268
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
out:
2269 2270 2271
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

C
Chuck Lever 已提交
2272 2273
	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
2274 2275 2276 2277
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
2278
		res = generic_permission(inode, mask);
C
Chuck Lever 已提交
2279
	goto out;
L
Linus Torvalds 已提交
2280
}
B
Bryan Schumaker 已提交
2281
EXPORT_SYMBOL_GPL(nfs_permission);
L
Linus Torvalds 已提交
2282 2283 2284 2285 2286 2287 2288

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