dir.c 59.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
/*
 *  linux/fs/nfs/dir.c
 *
 *  Copyright (C) 1992  Rick Sladkey
 *
 *  nfs directory handling functions
 *
 * 10 Apr 1996	Added silly rename for unlink	--okir
 * 28 Sep 1996	Improved directory cache --okir
 * 23 Aug 1997  Claus Heine claus@momo.math.rwth-aachen.de 
 *              Re-implemented silly rename for unlink, newly implemented
 *              silly rename for nfs_rename() following the suggestions
 *              of Olaf Kirch (okir) found in this file.
 *              Following Linus comments on my original hack, this version
 *              depends only on the dcache stuff and doesn't touch the inode
 *              layer (iput() and friends).
 *  6 Jun 1999	Cache readdir lookups in the page cache. -DaveM
 */

#include <linux/time.h>
#include <linux/errno.h>
#include <linux/stat.h>
#include <linux/fcntl.h>
#include <linux/string.h>
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/pagemap.h>
32
#include <linux/pagevec.h>
L
Linus Torvalds 已提交
33
#include <linux/namei.h>
34
#include <linux/mount.h>
A
Alexey Dobriyan 已提交
35
#include <linux/sched.h>
36
#include <linux/kmemleak.h>
37
#include <linux/xattr.h>
L
Linus Torvalds 已提交
38 39

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

/* #define NFS_DEBUG_VERBOSE 1 */

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

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

72
const struct inode_operations nfs_dir_inode_operations = {
L
Linus Torvalds 已提交
73 74 75 76 77 78 79 80 81 82 83 84 85 86
	.create		= nfs_create,
	.lookup		= nfs_lookup,
	.link		= nfs_link,
	.unlink		= nfs_unlink,
	.symlink	= nfs_symlink,
	.mkdir		= nfs_mkdir,
	.rmdir		= nfs_rmdir,
	.mknod		= nfs_mknod,
	.rename		= nfs_rename,
	.permission	= nfs_permission,
	.getattr	= nfs_getattr,
	.setattr	= nfs_setattr,
};

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

91
#ifdef CONFIG_NFS_V3
92
const struct inode_operations nfs3_dir_inode_operations = {
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111
	.create		= nfs_create,
	.lookup		= nfs_lookup,
	.link		= nfs_link,
	.unlink		= nfs_unlink,
	.symlink	= nfs_symlink,
	.mkdir		= nfs_mkdir,
	.rmdir		= nfs_rmdir,
	.mknod		= nfs_mknod,
	.rename		= nfs_rename,
	.permission	= nfs_permission,
	.getattr	= nfs_getattr,
	.setattr	= nfs_setattr,
	.listxattr	= nfs3_listxattr,
	.getxattr	= nfs3_getxattr,
	.setxattr	= nfs3_setxattr,
	.removexattr	= nfs3_removexattr,
};
#endif  /* CONFIG_NFS_V3 */

L
Linus Torvalds 已提交
112 113 114
#ifdef CONFIG_NFS_V4

static struct dentry *nfs_atomic_lookup(struct inode *, struct dentry *, struct nameidata *);
A
Al Viro 已提交
115
static int nfs_open_create(struct inode *dir, struct dentry *dentry, umode_t mode, struct nameidata *nd);
116
const struct inode_operations nfs4_dir_inode_operations = {
117
	.create		= nfs_open_create,
L
Linus Torvalds 已提交
118 119 120 121 122 123 124 125 126 127 128
	.lookup		= nfs_atomic_lookup,
	.link		= nfs_link,
	.unlink		= nfs_unlink,
	.symlink	= nfs_symlink,
	.mkdir		= nfs_mkdir,
	.rmdir		= nfs_rmdir,
	.mknod		= nfs_mknod,
	.rename		= nfs_rename,
	.permission	= nfs_permission,
	.getattr	= nfs_getattr,
	.setattr	= nfs_setattr,
129 130 131 132
	.getxattr	= generic_getxattr,
	.setxattr	= generic_setxattr,
	.listxattr	= generic_listxattr,
	.removexattr	= generic_removexattr,
L
Linus Torvalds 已提交
133 134 135 136
};

#endif /* CONFIG_NFS_V4 */

137
static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct inode *dir, struct rpc_cred *cred)
138 139 140 141
{
	struct nfs_open_dir_context *ctx;
	ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
	if (ctx != NULL) {
142
		ctx->duped = 0;
143
		ctx->attr_gencount = NFS_I(dir)->attr_gencount;
144
		ctx->dir_cookie = 0;
145
		ctx->dup_cookie = 0;
146
		ctx->cred = get_rpccred(cred);
147 148 149
		return ctx;
	}
	return  ERR_PTR(-ENOMEM);
150 151 152 153 154 155 156 157
}

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
158 159 160 161 162 163
/*
 * Open file
 */
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
164 165 166
	int res = 0;
	struct nfs_open_dir_context *ctx;
	struct rpc_cred *cred;
L
Linus Torvalds 已提交
167

C
Chuck Lever 已提交
168
	dfprintk(FILE, "NFS: open dir(%s/%s)\n",
169 170 171 172
			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
173

174 175 176
	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return PTR_ERR(cred);
177
	ctx = alloc_nfs_open_dir_context(inode, cred);
178 179 180 181 182
	if (IS_ERR(ctx)) {
		res = PTR_ERR(ctx);
		goto out;
	}
	filp->private_data = ctx;
183 184 185 186 187 188 189
	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);
	}
190 191
out:
	put_rpccred(cred);
L
Linus Torvalds 已提交
192 193 194
	return res;
}

195 196 197 198 199 200 201
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
202 203 204 205
struct nfs_cache_array_entry {
	u64 cookie;
	u64 ino;
	struct qstr string;
206
	unsigned char d_type;
B
Bryan Schumaker 已提交
207 208 209
};

struct nfs_cache_array {
210
	int size;
B
Bryan Schumaker 已提交
211 212 213 214 215
	int eof_index;
	u64 last_cookie;
	struct nfs_cache_array_entry array[0];
};

216
typedef int (*decode_dirent_t)(struct xdr_stream *, struct nfs_entry *, int);
L
Linus Torvalds 已提交
217 218 219 220
typedef struct {
	struct file	*file;
	struct page	*page;
	unsigned long	page_index;
221
	u64		*dir_cookie;
222
	u64		last_cookie;
223
	loff_t		current_index;
L
Linus Torvalds 已提交
224
	decode_dirent_t	decode;
B
Bryan Schumaker 已提交
225

226
	unsigned long	timestamp;
227
	unsigned long	gencount;
B
Bryan Schumaker 已提交
228 229 230
	unsigned int	cache_entry_index;
	unsigned int	plus:1;
	unsigned int	eof:1;
L
Linus Torvalds 已提交
231 232
} nfs_readdir_descriptor_t;

B
Bryan Schumaker 已提交
233 234
/*
 * The caller is responsible for calling nfs_readdir_release_array(page)
L
Linus Torvalds 已提交
235 236
 */
static
B
Bryan Schumaker 已提交
237 238
struct nfs_cache_array *nfs_readdir_get_array(struct page *page)
{
239
	void *ptr;
B
Bryan Schumaker 已提交
240 241
	if (page == NULL)
		return ERR_PTR(-EIO);
242 243 244 245
	ptr = kmap(page);
	if (ptr == NULL)
		return ERR_PTR(-ENOMEM);
	return ptr;
B
Bryan Schumaker 已提交
246 247 248 249 250 251 252 253 254 255 256 257
}

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

/*
 * we are freeing strings created by nfs_add_to_readdir_array()
 */
static
258
void nfs_readdir_clear_array(struct page *page)
B
Bryan Schumaker 已提交
259
{
260
	struct nfs_cache_array *array;
B
Bryan Schumaker 已提交
261
	int i;
262

263
	array = kmap_atomic(page);
B
Bryan Schumaker 已提交
264 265
	for (i = 0; i < array->size; i++)
		kfree(array->array[i].string.name);
266
	kunmap_atomic(array);
B
Bryan Schumaker 已提交
267 268 269 270 271 272 273 274
}

/*
 * 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
275
int nfs_readdir_make_qstr(struct qstr *string, const char *name, unsigned int len)
B
Bryan Schumaker 已提交
276 277 278
{
	string->len = len;
	string->name = kmemdup(name, len, GFP_KERNEL);
279 280
	if (string->name == NULL)
		return -ENOMEM;
281 282 283 284 285
	/*
	 * 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);
286 287
	string->hash = full_name_hash(name, len);
	return 0;
B
Bryan Schumaker 已提交
288 289 290 291 292 293
}

static
int nfs_readdir_add_to_array(struct nfs_entry *entry, struct page *page)
{
	struct nfs_cache_array *array = nfs_readdir_get_array(page);
294 295 296
	struct nfs_cache_array_entry *cache_entry;
	int ret;

B
Bryan Schumaker 已提交
297 298
	if (IS_ERR(array))
		return PTR_ERR(array);
299 300 301 302

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

	/* Check that this entry lies within the page bounds */
303
	ret = -ENOSPC;
304
	if ((char *)&cache_entry[1] - (char *)page_address(page) > PAGE_SIZE)
305
		goto out;
B
Bryan Schumaker 已提交
306

307 308
	cache_entry->cookie = entry->prev_cookie;
	cache_entry->ino = entry->ino;
309
	cache_entry->d_type = entry->d_type;
310 311 312
	ret = nfs_readdir_make_qstr(&cache_entry->string, entry->name, entry->len);
	if (ret)
		goto out;
B
Bryan Schumaker 已提交
313
	array->last_cookie = entry->cookie;
314
	array->size++;
T
Trond Myklebust 已提交
315
	if (entry->eof != 0)
B
Bryan Schumaker 已提交
316
		array->eof_index = array->size;
317
out:
B
Bryan Schumaker 已提交
318
	nfs_readdir_release_array(page);
319
	return ret;
B
Bryan Schumaker 已提交
320 321 322 323 324 325 326 327 328 329 330
}

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) {
331
		if (array->eof_index >= 0)
B
Bryan Schumaker 已提交
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348
			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;
349
	loff_t new_pos;
B
Bryan Schumaker 已提交
350 351 352 353
	int status = -EAGAIN;

	for (i = 0; i < array->size; i++) {
		if (array->array[i].cookie == *desc->dir_cookie) {
354 355 356
			struct nfs_inode *nfsi = NFS_I(desc->file->f_path.dentry->d_inode);
			struct nfs_open_dir_context *ctx = desc->file->private_data;

357
			new_pos = desc->current_index + i;
358 359 360 361 362 363 364 365 366 367
			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.  "
368
								"The file: %s has duplicate cookie %llu\n",
369 370
								desc->file->f_dentry->d_parent->d_name.name,
								desc->file->f_dentry->d_name.name,
371
								array->array[i].string.name,
372 373 374 375 376
								*desc->dir_cookie);
					}
					status = -ELOOP;
					goto out;
				}
377
				ctx->dup_cookie = *desc->dir_cookie;
378
				ctx->duped = -1;
379 380
			}
			desc->file->f_pos = new_pos;
B
Bryan Schumaker 已提交
381
			desc->cache_entry_index = i;
T
Trond Myklebust 已提交
382
			return 0;
B
Bryan Schumaker 已提交
383 384
		}
	}
T
Trond Myklebust 已提交
385
	if (array->eof_index >= 0) {
386
		status = -EBADCOOKIE;
387 388
		if (*desc->dir_cookie == array->last_cookie)
			desc->eof = 1;
389
	}
390
out:
B
Bryan Schumaker 已提交
391 392 393 394 395 396 397
	return status;
}

static
int nfs_readdir_search_array(nfs_readdir_descriptor_t *desc)
{
	struct nfs_cache_array *array;
T
Trond Myklebust 已提交
398
	int status;
B
Bryan Schumaker 已提交
399 400 401 402 403 404 405 406 407 408 409 410

	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
411
	if (status == -EAGAIN) {
412
		desc->last_cookie = array->last_cookie;
413
		desc->current_index += array->size;
T
Trond Myklebust 已提交
414 415
		desc->page_index++;
	}
B
Bryan Schumaker 已提交
416 417 418 419 420 421 422
	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
423
int nfs_readdir_xdr_filler(struct page **pages, nfs_readdir_descriptor_t *desc,
B
Bryan Schumaker 已提交
424
			struct nfs_entry *entry, struct file *file, struct inode *inode)
L
Linus Torvalds 已提交
425
{
426 427
	struct nfs_open_dir_context *ctx = file->private_data;
	struct rpc_cred	*cred = ctx->cred;
428
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
429 430 431 432
	int		error;

 again:
	timestamp = jiffies;
433
	gencount = nfs_inc_attr_generation_counter();
B
Bryan Schumaker 已提交
434
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, entry->cookie, pages,
L
Linus Torvalds 已提交
435 436 437 438 439
					  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
440
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
441 442 443 444 445
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
446
	desc->timestamp = timestamp;
447
	desc->gencount = gencount;
B
Bryan Schumaker 已提交
448 449
error:
	return error;
L
Linus Torvalds 已提交
450 451
}

452 453
static int xdr_decode(nfs_readdir_descriptor_t *desc,
		      struct nfs_entry *entry, struct xdr_stream *xdr)
L
Linus Torvalds 已提交
454
{
455
	int error;
L
Linus Torvalds 已提交
456

457 458 459
	error = desc->decode(xdr, entry, desc->plus);
	if (error)
		return error;
B
Bryan Schumaker 已提交
460 461 462
	entry->fattr->time_start = desc->timestamp;
	entry->fattr->gencount = desc->gencount;
	return 0;
L
Linus Torvalds 已提交
463 464
}

B
Bryan Schumaker 已提交
465 466 467 468 469
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
470
	if (nfs_compare_fh(entry->fh, NFS_FH(dentry->d_inode)) != 0)
B
Bryan Schumaker 已提交
471 472 473 474 475 476
		goto different;
	return 1;
different:
	return 0;
}

477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499
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
500 501 502
static
void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
{
503
	struct qstr filename = QSTR_INIT(entry->name, entry->len);
504 505
	struct dentry *dentry;
	struct dentry *alias;
B
Bryan Schumaker 已提交
506 507 508
	struct inode *dir = parent->d_inode;
	struct inode *inode;

509 510 511 512 513 514 515
	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
516

517
	dentry = d_lookup(parent, &filename);
B
Bryan Schumaker 已提交
518 519 520 521 522 523 524 525 526 527 528
	if (dentry != NULL) {
		if (nfs_same_file(dentry, entry)) {
			nfs_refresh_inode(dentry->d_inode, entry->fattr);
			goto out;
		} else {
			d_drop(dentry);
			dput(dentry);
		}
	}

	dentry = d_alloc(parent, &filename);
529 530 531
	if (dentry == NULL)
		return;

B
Bryan Schumaker 已提交
532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548
	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
549 550
/* Perform conversion from xdr to cache array */
static
551
int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *entry,
552
				struct page **xdr_pages, struct page *page, unsigned int buflen)
L
Linus Torvalds 已提交
553
{
554
	struct xdr_stream stream;
555
	struct xdr_buf buf;
556
	struct page *scratch;
B
Bryan Schumaker 已提交
557
	struct nfs_cache_array *array;
558 559
	unsigned int count = 0;
	int status;
560

561 562 563
	scratch = alloc_page(GFP_KERNEL);
	if (scratch == NULL)
		return -ENOMEM;
564

565
	xdr_init_decode_pages(&stream, &buf, xdr_pages, buflen);
566
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
B
Bryan Schumaker 已提交
567 568 569

	do {
		status = xdr_decode(desc, entry, &stream);
570 571 572
		if (status != 0) {
			if (status == -EAGAIN)
				status = 0;
B
Bryan Schumaker 已提交
573
			break;
574
		}
B
Bryan Schumaker 已提交
575

576 577
		count++;

T
Trond Myklebust 已提交
578
		if (desc->plus != 0)
B
Bryan Schumaker 已提交
579
			nfs_prime_dcache(desc->file->f_path.dentry, entry);
580 581 582 583

		status = nfs_readdir_add_to_array(entry, page);
		if (status != 0)
			break;
B
Bryan Schumaker 已提交
584 585
	} while (!entry->eof);

T
Trond Myklebust 已提交
586
	if (count == 0 || (status == -EBADCOOKIE && entry->eof != 0)) {
B
Bryan Schumaker 已提交
587
		array = nfs_readdir_get_array(page);
588 589 590 591
		if (!IS_ERR(array)) {
			array->eof_index = array->size;
			status = 0;
			nfs_readdir_release_array(page);
592 593
		} else
			status = PTR_ERR(array);
L
Linus Torvalds 已提交
594
	}
595 596

	put_page(scratch);
597
	return status;
B
Bryan Schumaker 已提交
598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619
}

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
620
int nfs_readdir_large_page(struct page **pages, unsigned int npages)
B
Bryan Schumaker 已提交
621 622 623 624 625 626 627 628 629
{
	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;
	}
630
	return 0;
B
Bryan Schumaker 已提交
631 632 633

out_freepages:
	nfs_readdir_free_pagearray(pages, i);
634
	return -ENOMEM;
L
Linus Torvalds 已提交
635 636
}

B
Bryan Schumaker 已提交
637 638
static
int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode)
639
{
B
Bryan Schumaker 已提交
640 641
	struct page *pages[NFS_MAX_READDIR_PAGES];
	void *pages_ptr = NULL;
B
Bryan Schumaker 已提交
642 643 644
	struct nfs_entry entry;
	struct file	*file = desc->file;
	struct nfs_cache_array *array;
645
	int status = -ENOMEM;
B
Bryan Schumaker 已提交
646
	unsigned int array_size = ARRAY_SIZE(pages);
B
Bryan Schumaker 已提交
647 648

	entry.prev_cookie = 0;
649
	entry.cookie = desc->last_cookie;
B
Bryan Schumaker 已提交
650 651 652
	entry.eof = 0;
	entry.fh = nfs_alloc_fhandle();
	entry.fattr = nfs_alloc_fattr();
653
	entry.server = NFS_SERVER(inode);
B
Bryan Schumaker 已提交
654 655
	if (entry.fh == NULL || entry.fattr == NULL)
		goto out;
656

B
Bryan Schumaker 已提交
657
	array = nfs_readdir_get_array(page);
658 659 660 661
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
662 663
	memset(array, 0, sizeof(struct nfs_cache_array));
	array->eof_index = -1;
664

665 666
	status = nfs_readdir_large_page(pages, array_size);
	if (status < 0)
B
Bryan Schumaker 已提交
667 668
		goto out_release_array;
	do {
669
		unsigned int pglen;
B
Bryan Schumaker 已提交
670
		status = nfs_readdir_xdr_filler(pages, desc, &entry, file, inode);
671

B
Bryan Schumaker 已提交
672
		if (status < 0)
673
			break;
674
		pglen = status;
675
		status = nfs_readdir_page_filler(desc, &entry, pages, page, pglen);
676 677 678 679 680 681
		if (status < 0) {
			if (status == -ENOSPC)
				status = 0;
			break;
		}
	} while (array->eof_index < 0);
B
Bryan Schumaker 已提交
682

B
Bryan Schumaker 已提交
683
	nfs_readdir_free_large_page(pages_ptr, pages, array_size);
B
Bryan Schumaker 已提交
684 685 686 687 688
out_release_array:
	nfs_readdir_release_array(page);
out:
	nfs_free_fattr(entry.fattr);
	nfs_free_fhandle(entry.fh);
689 690 691 692
	return status;
}

/*
B
Bryan Schumaker 已提交
693 694 695 696
 * 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
697
 */
B
Bryan Schumaker 已提交
698 699
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page)
L
Linus Torvalds 已提交
700
{
701
	struct inode	*inode = desc->file->f_path.dentry->d_inode;
702
	int ret;
L
Linus Torvalds 已提交
703

704 705
	ret = nfs_readdir_xdr_to_array(desc, page, inode);
	if (ret < 0)
B
Bryan Schumaker 已提交
706 707
		goto error;
	SetPageUptodate(page);
L
Linus Torvalds 已提交
708

B
Bryan Schumaker 已提交
709 710 711
	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
712
	}
B
Bryan Schumaker 已提交
713 714 715 716
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
717
	return ret;
B
Bryan Schumaker 已提交
718
}
L
Linus Torvalds 已提交
719

B
Bryan Schumaker 已提交
720 721 722
static
void cache_page_release(nfs_readdir_descriptor_t *desc)
{
723 724
	if (!desc->page->mapping)
		nfs_readdir_clear_array(desc->page);
B
Bryan Schumaker 已提交
725 726 727 728 729 730 731
	page_cache_release(desc->page);
	desc->page = NULL;
}

static
struct page *get_cache_page(nfs_readdir_descriptor_t *desc)
{
732
	return read_cache_page(desc->file->f_path.dentry->d_inode->i_mapping,
B
Bryan Schumaker 已提交
733
			desc->page_index, (filler_t *)nfs_readdir_filler, desc);
L
Linus Torvalds 已提交
734 735 736
}

/*
B
Bryan Schumaker 已提交
737
 * Returns 0 if desc->dir_cookie was found on page desc->page_index
L
Linus Torvalds 已提交
738
 */
B
Bryan Schumaker 已提交
739 740 741 742 743 744 745 746 747 748
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
749 750
	if (res != 0)
		cache_page_release(desc);
B
Bryan Schumaker 已提交
751 752 753 754
	return res;
}

/* Search for desc->dir_cookie from the beginning of the page cache */
L
Linus Torvalds 已提交
755 756 757
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
758
	int res;
B
Bryan Schumaker 已提交
759

760
	if (desc->page_index == 0) {
761
		desc->current_index = 0;
762 763
		desc->last_cookie = 0;
	}
T
Trond Myklebust 已提交
764
	do {
B
Bryan Schumaker 已提交
765
		res = find_cache_page(desc);
T
Trond Myklebust 已提交
766
	} while (res == -EAGAIN);
L
Linus Torvalds 已提交
767 768 769 770 771 772 773 774 775 776 777
	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
778 779 780
	int i = 0;
	int res = 0;
	struct nfs_cache_array *array = NULL;
781 782
	struct nfs_open_dir_context *ctx = file->private_data;

B
Bryan Schumaker 已提交
783
	array = nfs_readdir_get_array(desc->page);
784 785 786 787
	if (IS_ERR(array)) {
		res = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
788 789

	for (i = desc->cache_entry_index; i < array->size; i++) {
790
		struct nfs_cache_array_entry *ent;
L
Linus Torvalds 已提交
791

792 793
		ent = &array->array[i];
		if (filldir(dirent, ent->string.name, ent->string.len,
794 795
		    file->f_pos, nfs_compat_user_ino64(ent->ino),
		    ent->d_type) < 0) {
796
			desc->eof = 1;
L
Linus Torvalds 已提交
797
			break;
798
		}
799
		file->f_pos++;
B
Bryan Schumaker 已提交
800 801 802 803
		if (i < (array->size-1))
			*desc->dir_cookie = array->array[i+1].cookie;
		else
			*desc->dir_cookie = array->last_cookie;
804 805
		if (ctx->duped != 0)
			ctx->duped = 1;
L
Linus Torvalds 已提交
806
	}
T
Trond Myklebust 已提交
807
	if (array->eof_index >= 0)
808
		desc->eof = 1;
B
Bryan Schumaker 已提交
809 810

	nfs_readdir_release_array(desc->page);
811
out:
B
Bryan Schumaker 已提交
812
	cache_page_release(desc);
C
Chuck Lever 已提交
813 814
	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
815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835
	return res;
}

/*
 * If we cannot find a cookie in our cache, we suspect that this is
 * because it points to a deleted file, so we ask the server to return
 * whatever it thinks is the next entry. We then feed this to filldir.
 * If all goes well, we should then be able to find our way round the
 * cache on the next call to readdir_search_pagecache();
 *
 * NOTE: we cannot add the anonymous page to the pagecache because
 *	 the data it contains might not be page aligned. Besides,
 *	 we should already have a complete representation of the
 *	 directory in the page cache by the time we get here.
 */
static inline
int uncached_readdir(nfs_readdir_descriptor_t *desc, void *dirent,
		     filldir_t filldir)
{
	struct page	*page = NULL;
	int		status;
B
Bryan Schumaker 已提交
836
	struct inode *inode = desc->file->f_path.dentry->d_inode;
837
	struct nfs_open_dir_context *ctx = desc->file->private_data;
L
Linus Torvalds 已提交
838

C
Chuck Lever 已提交
839 840
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
841 842 843 844 845 846

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
B
Bryan Schumaker 已提交
847

848
	desc->page_index = 0;
849
	desc->last_cookie = *desc->dir_cookie;
850
	desc->page = page;
851
	ctx->duped = 0;
852

853 854
	status = nfs_readdir_xdr_to_array(desc, page, inode);
	if (status < 0)
L
Linus Torvalds 已提交
855 856 857 858 859
		goto out_release;

	status = nfs_do_filldir(desc, dirent, filldir);

 out:
C
Chuck Lever 已提交
860
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
861
			__func__, status);
L
Linus Torvalds 已提交
862 863
	return status;
 out_release:
B
Bryan Schumaker 已提交
864
	cache_page_release(desc);
L
Linus Torvalds 已提交
865 866 867
	goto out;
}

868 869 870
/* 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
871 872 873
 */
static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
874
	struct dentry	*dentry = filp->f_path.dentry;
L
Linus Torvalds 已提交
875 876 877
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
878
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
T
Trond Myklebust 已提交
879
	int res;
L
Linus Torvalds 已提交
880

C
Chuck Lever 已提交
881
	dfprintk(FILE, "NFS: readdir(%s/%s) starting at cookie %llu\n",
C
Chuck Lever 已提交
882 883
			dentry->d_parent->d_name.name, dentry->d_name.name,
			(long long)filp->f_pos);
C
Chuck Lever 已提交
884 885
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
886
	/*
887
	 * filp->f_pos points to the dirent entry number.
888
	 * *desc->dir_cookie has the cookie for the next entry. We have
889 890
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
891 892 893 894
	 */
	memset(desc, 0, sizeof(*desc));

	desc->file = filp;
895
	desc->dir_cookie = &dir_ctx->dir_cookie;
L
Linus Torvalds 已提交
896
	desc->decode = NFS_PROTO(inode)->decode_dirent;
897
	desc->plus = nfs_use_readdirplus(inode, filp) ? 1 : 0;
L
Linus Torvalds 已提交
898

T
Trond Myklebust 已提交
899
	nfs_block_sillyrename(dentry);
900
	res = nfs_revalidate_mapping(inode, filp->f_mapping);
T
Trond Myklebust 已提交
901 902 903
	if (res < 0)
		goto out;

T
Trond Myklebust 已提交
904
	do {
L
Linus Torvalds 已提交
905
		res = readdir_search_pagecache(desc);
906

L
Linus Torvalds 已提交
907
		if (res == -EBADCOOKIE) {
908
			res = 0;
L
Linus Torvalds 已提交
909
			/* This means either end of directory */
B
Bryan Schumaker 已提交
910
			if (*desc->dir_cookie && desc->eof == 0) {
L
Linus Torvalds 已提交
911 912
				/* Or that the server has 'lost' a cookie */
				res = uncached_readdir(desc, dirent, filldir);
913
				if (res == 0)
L
Linus Torvalds 已提交
914 915 916 917 918
					continue;
			}
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
919
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
920
			nfs_zap_caches(inode);
921
			desc->page_index = 0;
L
Linus Torvalds 已提交
922
			desc->plus = 0;
B
Bryan Schumaker 已提交
923
			desc->eof = 0;
L
Linus Torvalds 已提交
924 925 926 927 928 929
			continue;
		}
		if (res < 0)
			break;

		res = nfs_do_filldir(desc, dirent, filldir);
930
		if (res < 0)
L
Linus Torvalds 已提交
931
			break;
T
Trond Myklebust 已提交
932
	} while (!desc->eof);
T
Trond Myklebust 已提交
933
out:
T
Trond Myklebust 已提交
934
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
935 936
	if (res > 0)
		res = 0;
937
	dfprintk(FILE, "NFS: readdir(%s/%s) returns %d\n",
C
Chuck Lever 已提交
938 939 940
			dentry->d_parent->d_name.name, dentry->d_name.name,
			res);
	return res;
L
Linus Torvalds 已提交
941 942
}

943
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
944
{
945 946
	struct dentry *dentry = filp->f_path.dentry;
	struct inode *inode = dentry->d_inode;
947
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
948

C
Chuck Lever 已提交
949
	dfprintk(FILE, "NFS: llseek dir(%s/%s, %lld, %d)\n",
950 951 952 953 954
			dentry->d_parent->d_name.name,
			dentry->d_name.name,
			offset, origin);

	mutex_lock(&inode->i_mutex);
955 956 957 958 959 960 961 962 963 964 965 966
	switch (origin) {
		case 1:
			offset += filp->f_pos;
		case 0:
			if (offset >= 0)
				break;
		default:
			offset = -EINVAL;
			goto out;
	}
	if (offset != filp->f_pos) {
		filp->f_pos = offset;
967
		dir_ctx->dir_cookie = 0;
968
		dir_ctx->duped = 0;
969 970
	}
out:
971
	mutex_unlock(&inode->i_mutex);
972 973 974
	return offset;
}

L
Linus Torvalds 已提交
975 976 977 978
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
979 980
static int nfs_fsync_dir(struct file *filp, loff_t start, loff_t end,
			 int datasync)
L
Linus Torvalds 已提交
981
{
982
	struct dentry *dentry = filp->f_path.dentry;
983
	struct inode *inode = dentry->d_inode;
984

C
Chuck Lever 已提交
985
	dfprintk(FILE, "NFS: fsync dir(%s/%s) datasync %d\n",
C
Chuck Lever 已提交
986 987 988
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);

989
	mutex_lock(&inode->i_mutex);
990
	nfs_inc_stats(dentry->d_inode, NFSIOS_VFSFSYNC);
991
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
992 993 994
	return 0;
}

995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006
/**
 * 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)
{
1007
	NFS_I(dir)->cache_change_attribute++;
1008 1009
}

L
Linus Torvalds 已提交
1010 1011 1012 1013 1014
/*
 * 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.
 */
1015
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1016 1017 1018
{
	if (IS_ROOT(dentry))
		return 1;
1019 1020
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
1021 1022 1023 1024 1025 1026 1027 1028
	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
1029 1030
}

1031 1032 1033 1034
/*
 * Return the intent data that applies to this particular path component
 *
 * Note that the current set of intents only apply to the very last
1035 1036
 * component of the path and none of them is set before that last
 * component.
1037
 */
1038 1039
static inline unsigned int nfs_lookup_check_intent(struct nameidata *nd,
						unsigned int mask)
1040 1041 1042 1043
{
	return nd->flags & mask;
}

1044 1045 1046 1047 1048 1049 1050 1051
/*
 * Use intent information to check whether or not we're going to do
 * an O_EXCL create using this path component.
 */
static int nfs_is_exclusive_create(struct inode *dir, struct nameidata *nd)
{
	if (NFS_PROTO(dir)->version == 2)
		return 0;
1052
	return nd && nfs_lookup_check_intent(nd, LOOKUP_EXCL);
1053 1054
}

1055 1056 1057 1058 1059 1060 1061 1062
/*
 * Inode and filehandle revalidation for lookups.
 *
 * We force revalidation in the cases where the VFS sets LOOKUP_REVAL,
 * or if the intent information indicates that we're about to open this
 * particular file and the "nocto" mount flag is not set.
 *
 */
L
Linus Torvalds 已提交
1063 1064 1065 1066 1067
static inline
int nfs_lookup_verify_inode(struct inode *inode, struct nameidata *nd)
{
	struct nfs_server *server = NFS_SERVER(inode);

1068
	if (IS_AUTOMOUNT(inode))
1069
		return 0;
L
Linus Torvalds 已提交
1070 1071
	if (nd != NULL) {
		/* VFS wants an on-the-wire revalidation */
1072
		if (nd->flags & LOOKUP_REVAL)
L
Linus Torvalds 已提交
1073 1074
			goto out_force;
		/* This is an open(2) */
1075
		if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 &&
1076 1077 1078
				!(server->flags & NFS_MOUNT_NOCTO) &&
				(S_ISREG(inode->i_mode) ||
				 S_ISDIR(inode->i_mode)))
L
Linus Torvalds 已提交
1079
			goto out_force;
1080
		return 0;
L
Linus Torvalds 已提交
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098
	}
	return nfs_revalidate_inode(server, inode);
out_force:
	return __nfs_revalidate_inode(server, inode);
}

/*
 * We judge how long we want to trust negative
 * dentries by looking at the parent inode mtime.
 *
 * If parent mtime has changed, we revalidate, else we wait for a
 * period corresponding to the parent's attribute cache timeout value.
 */
static inline
int nfs_neg_need_reval(struct inode *dir, struct dentry *dentry,
		       struct nameidata *nd)
{
	/* Don't revalidate a negative dentry if we're creating a new file */
1099
	if (nd != NULL && nfs_lookup_check_intent(nd, LOOKUP_CREATE) != 0)
L
Linus Torvalds 已提交
1100
		return 0;
1101 1102
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116
	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.
 */
1117
static int nfs_lookup_revalidate(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
1118 1119 1120 1121
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
1122 1123
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1124 1125
	int error;

1126 1127 1128
	if (nd->flags & LOOKUP_RCU)
		return -ECHILD;

L
Linus Torvalds 已提交
1129 1130
	parent = dget_parent(dentry);
	dir = parent->d_inode;
C
Chuck Lever 已提交
1131
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
1132 1133 1134 1135 1136
	inode = dentry->d_inode;

	if (!inode) {
		if (nfs_neg_need_reval(dir, dentry, nd))
			goto out_bad;
1137
		goto out_valid_noent;
L
Linus Torvalds 已提交
1138 1139 1140
	}

	if (is_bad_inode(inode)) {
C
Chuck Lever 已提交
1141
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
1142
				__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1143
				dentry->d_name.name);
L
Linus Torvalds 已提交
1144 1145 1146
		goto out_bad;
	}

1147 1148 1149
	if (nfs_have_delegation(inode, FMODE_READ))
		goto out_set_verifier;

L
Linus Torvalds 已提交
1150
	/* Force a full look up iff the parent directory has changed */
1151
	if (!nfs_is_exclusive_create(dir, nd) && nfs_check_verifier(dir, dentry)) {
L
Linus Torvalds 已提交
1152 1153 1154 1155 1156 1157 1158 1159
		if (nfs_lookup_verify_inode(inode, nd))
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

1160 1161 1162 1163 1164 1165
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

1166
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1167 1168
	if (error)
		goto out_bad;
1169
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
1170
		goto out_bad;
1171
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
1172 1173
		goto out_bad;

1174 1175
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1176
out_set_verifier:
1177
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1178
 out_valid:
1179 1180 1181
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);
 out_valid_noent:
L
Linus Torvalds 已提交
1182
	dput(parent);
C
Chuck Lever 已提交
1183
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n",
1184
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1185
			dentry->d_name.name);
L
Linus Torvalds 已提交
1186 1187 1188 1189
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
1190
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
1191 1192 1193 1194 1195 1196
	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;
1197 1198
		if (dentry->d_flags & DCACHE_DISCONNECTED)
			goto out_valid;
L
Linus Torvalds 已提交
1199 1200 1201
		shrink_dcache_parent(dentry);
	}
	d_drop(dentry);
1202 1203
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1204
	dput(parent);
C
Chuck Lever 已提交
1205
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
1206
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1207
			dentry->d_name.name);
L
Linus Torvalds 已提交
1208
	return 0;
1209 1210 1211 1212 1213 1214 1215 1216
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
1217 1218 1219 1220 1221
}

/*
 * This is called from dput() when d_count is going to 0.
 */
N
Nick Piggin 已提交
1222
static int nfs_dentry_delete(const struct dentry *dentry)
L
Linus Torvalds 已提交
1223 1224 1225 1226 1227
{
	dfprintk(VFS, "NFS: dentry_delete(%s/%s, %x)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		dentry->d_flags);

1228 1229 1230 1231
	/* 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
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244
	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;

}

1245 1246 1247 1248 1249 1250 1251 1252
static void nfs_drop_nlink(struct inode *inode)
{
	spin_lock(&inode->i_lock);
	if (inode->i_nlink > 0)
		drop_nlink(inode);
	spin_unlock(&inode->i_lock);
}

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

1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
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
1281
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
1282 1283 1284
	.d_revalidate	= nfs_lookup_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1285
	.d_automount	= nfs_d_automount,
1286
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1287 1288 1289 1290 1291
};

static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
{
	struct dentry *res;
T
Trond Myklebust 已提交
1292
	struct dentry *parent;
L
Linus Torvalds 已提交
1293
	struct inode *inode = NULL;
1294 1295
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1296 1297 1298 1299
	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
1300
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
1301 1302 1303 1304 1305

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

1306 1307 1308 1309 1310 1311 1312
	/*
	 * If we're doing an exclusive create, optimize away the lookup
	 * but don't hash the dentry.
	 */
	if (nfs_is_exclusive_create(dir, nd)) {
		d_instantiate(dentry, NULL);
		res = NULL;
1313
		goto out;
1314
	}
L
Linus Torvalds 已提交
1315

1316 1317 1318 1319 1320 1321
	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
1322 1323 1324
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
	nfs_block_sillyrename(parent);
1325
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1326 1327 1328 1329
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
1330
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1331
	}
1332
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
N
Namhyung Kim 已提交
1333
	res = ERR_CAST(inode);
1334
	if (IS_ERR(res))
T
Trond Myklebust 已提交
1335
		goto out_unblock_sillyrename;
1336

1337 1338 1339
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);

L
Linus Torvalds 已提交
1340
no_entry:
1341
	res = d_materialise_unique(dentry, inode);
1342 1343
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
1344
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1345
		dentry = res;
1346
	}
L
Linus Torvalds 已提交
1347
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
1348 1349
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
L
Linus Torvalds 已提交
1350
out:
1351 1352
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1353 1354 1355 1356 1357 1358
	return res;
}

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

A
Al Viro 已提交
1359
const struct dentry_operations nfs4_dentry_operations = {
L
Linus Torvalds 已提交
1360 1361 1362
	.d_revalidate	= nfs_open_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1363
	.d_automount	= nfs_d_automount,
1364
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1365 1366
};

1367 1368 1369 1370
/*
 * Use intent information to determine whether we need to substitute
 * the NFSv4-style stateful OPEN for the LOOKUP call
 */
T
Trond Myklebust 已提交
1371
static int is_atomic_open(struct nameidata *nd)
L
Linus Torvalds 已提交
1372
{
1373
	if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_OPEN) == 0)
L
Linus Torvalds 已提交
1374 1375 1376 1377 1378
		return 0;
	/* NFS does not (yet) have a stateful open for directories */
	if (nd->flags & LOOKUP_DIRECTORY)
		return 0;
	/* Are we trying to write to a read only partition? */
1379
	if (__mnt_is_readonly(nd->path.mnt) &&
1380
	    (nd->intent.open.flags & (O_CREAT|O_TRUNC|O_ACCMODE)))
L
Linus Torvalds 已提交
1381 1382 1383 1384
		return 0;
	return 1;
}

1385 1386 1387 1388 1389 1390 1391 1392 1393 1394
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;
}

1395
static struct nfs_open_context *create_nfs_open_context(struct dentry *dentry, int open_flags)
1396
{
1397
	return alloc_nfs_open_context(dentry, flags_to_mode(open_flags));
1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412
}

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

static int nfs_intent_set_file(struct nameidata *nd, struct nfs_open_context *ctx)
{
	struct file *filp;
	int ret = 0;

	/* If the open_intent is for execute, we have an extra check to make */
	if (ctx->mode & FMODE_EXEC) {
1413
		ret = nfs_may_open(ctx->dentry->d_inode,
1414 1415 1416 1417 1418
				ctx->cred,
				nd->intent.open.flags);
		if (ret < 0)
			goto out;
	}
1419
	filp = lookup_instantiate_filp(nd, ctx->dentry, do_open);
1420 1421 1422 1423 1424 1425 1426 1427 1428
	if (IS_ERR(filp))
		ret = PTR_ERR(filp);
	else
		nfs_file_set_open_context(filp, ctx);
out:
	put_nfs_open_context(ctx);
	return ret;
}

L
Linus Torvalds 已提交
1429 1430
static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
{
1431 1432
	struct nfs_open_context *ctx;
	struct iattr attr;
L
Linus Torvalds 已提交
1433
	struct dentry *res = NULL;
1434
	struct inode *inode;
1435
	int open_flags;
1436
	int err;
L
Linus Torvalds 已提交
1437

C
Chuck Lever 已提交
1438 1439 1440
	dfprintk(VFS, "NFS: atomic_lookup(%s/%ld), %s\n",
			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);

L
Linus Torvalds 已提交
1441
	/* Check that we are indeed trying to open this file */
T
Trond Myklebust 已提交
1442
	if (!is_atomic_open(nd))
L
Linus Torvalds 已提交
1443 1444 1445 1446 1447 1448 1449
		goto no_open;

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

1450 1451
	/* Let vfs_create() deal with O_EXCL. Instantiate, but don't hash
	 * the dentry. */
1452
	if (nd->flags & LOOKUP_EXCL) {
1453
		d_instantiate(dentry, NULL);
1454 1455
		goto out;
	}
L
Linus Torvalds 已提交
1456

1457
	open_flags = nd->intent.open.flags;
1458
	attr.ia_valid = ATTR_OPEN;
1459 1460

	ctx = create_nfs_open_context(dentry, open_flags);
1461 1462 1463 1464 1465 1466
	res = ERR_CAST(ctx);
	if (IS_ERR(ctx))
		goto out;

	if (nd->flags & LOOKUP_CREATE) {
		attr.ia_mode = nd->intent.open.create_mode;
1467
		attr.ia_valid |= ATTR_MODE;
A
Aneesh Kumar K.V 已提交
1468
		attr.ia_mode &= ~current_umask();
1469
	} else
1470
		open_flags &= ~(O_EXCL | O_CREAT);
1471 1472 1473 1474

	if (open_flags & O_TRUNC) {
		attr.ia_valid |= ATTR_SIZE;
		attr.ia_size = 0;
1475 1476
	}

L
Linus Torvalds 已提交
1477
	/* Open the file on the server */
1478
	nfs_block_sillyrename(dentry->d_parent);
1479
	inode = NFS_PROTO(dir)->open_context(dir, ctx, open_flags, &attr);
1480 1481
	if (IS_ERR(inode)) {
		nfs_unblock_sillyrename(dentry->d_parent);
1482
		put_nfs_open_context(ctx);
1483
		switch (PTR_ERR(inode)) {
L
Linus Torvalds 已提交
1484 1485
			/* Make a negative dentry */
			case -ENOENT:
1486
				d_add(dentry, NULL);
1487 1488
				res = NULL;
				goto out;
L
Linus Torvalds 已提交
1489
			/* This turned out not to be a regular file */
1490
			case -EISDIR:
1491 1492
			case -ENOTDIR:
				goto no_open;
L
Linus Torvalds 已提交
1493 1494 1495 1496 1497
			case -ELOOP:
				if (!(nd->intent.open.flags & O_NOFOLLOW))
					goto no_open;
			/* case -EINVAL: */
			default:
1498
				res = ERR_CAST(inode);
L
Linus Torvalds 已提交
1499 1500
				goto out;
		}
1501
	}
1502
	res = d_add_unique(dentry, inode);
1503
	nfs_unblock_sillyrename(dentry->d_parent);
1504
	if (res != NULL) {
1505 1506
		dput(ctx->dentry);
		ctx->dentry = dget(res);
L
Linus Torvalds 已提交
1507
		dentry = res;
1508
	}
1509 1510 1511 1512 1513 1514
	err = nfs_intent_set_file(nd, ctx);
	if (err < 0) {
		if (res != NULL)
			dput(res);
		return ERR_PTR(err);
	}
L
Linus Torvalds 已提交
1515
out:
1516
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1517 1518 1519 1520 1521 1522 1523 1524
	return res;
no_open:
	return nfs_lookup(dir, dentry, nd);
}

static int nfs_open_revalidate(struct dentry *dentry, struct nameidata *nd)
{
	struct dentry *parent = NULL;
N
Nick Piggin 已提交
1525
	struct inode *inode;
L
Linus Torvalds 已提交
1526
	struct inode *dir;
1527
	struct nfs_open_context *ctx;
1528
	struct iattr attr;
L
Linus Torvalds 已提交
1529 1530
	int openflags, ret = 0;

N
Nick Piggin 已提交
1531 1532 1533 1534
	if (nd->flags & LOOKUP_RCU)
		return -ECHILD;

	inode = dentry->d_inode;
1535
	if (!is_atomic_open(nd) || d_mountpoint(dentry))
T
Trond Myklebust 已提交
1536
		goto no_open;
1537

L
Linus Torvalds 已提交
1538 1539
	parent = dget_parent(dentry);
	dir = parent->d_inode;
1540

L
Linus Torvalds 已提交
1541 1542 1543
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1544 1545 1546
	if (inode == NULL) {
		if (!nfs_neg_need_reval(dir, dentry, nd))
			ret = 1;
L
Linus Torvalds 已提交
1547
		goto out;
1548 1549
	}

L
Linus Torvalds 已提交
1550 1551
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
T
Trond Myklebust 已提交
1552
		goto no_open_dput;
L
Linus Torvalds 已提交
1553 1554 1555
	openflags = nd->intent.open.flags;
	/* We cannot do exclusive creation on a positive dentry */
	if ((openflags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL))
T
Trond Myklebust 已提交
1556
		goto no_open_dput;
1557 1558
	/* We can't create new files here */
	openflags &= ~(O_CREAT|O_EXCL);
L
Linus Torvalds 已提交
1559

1560
	ctx = create_nfs_open_context(dentry, openflags);
1561 1562 1563
	ret = PTR_ERR(ctx);
	if (IS_ERR(ctx))
		goto out;
1564

1565
	attr.ia_valid = ATTR_OPEN;
1566 1567 1568 1569 1570 1571
	if (openflags & O_TRUNC) {
		attr.ia_valid |= ATTR_SIZE;
		attr.ia_size = 0;
		nfs_wb_all(inode);
	}

L
Linus Torvalds 已提交
1572
	/*
1573
	 * Note: we're not holding inode->i_mutex and so may be racing with
L
Linus Torvalds 已提交
1574 1575 1576
	 * operations that change the directory. We therefore save the
	 * change attribute *before* we do the RPC call.
	 */
1577
	inode = NFS_PROTO(dir)->open_context(dir, ctx, openflags, &attr);
1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591
	if (IS_ERR(inode)) {
		ret = PTR_ERR(inode);
		switch (ret) {
		case -EPERM:
		case -EACCES:
		case -EDQUOT:
		case -ENOSPC:
		case -EROFS:
			goto out_put_ctx;
		default:
			goto out_drop;
		}
	}
	iput(inode);
1592
	if (inode != dentry->d_inode)
1593
		goto out_drop;
1594 1595 1596 1597 1598

	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	ret = nfs_intent_set_file(nd, ctx);
	if (ret >= 0)
		ret = 1;
L
Linus Torvalds 已提交
1599 1600 1601
out:
	dput(parent);
	return ret;
1602 1603 1604 1605 1606 1607 1608
out_drop:
	d_drop(dentry);
	ret = 0;
out_put_ctx:
	put_nfs_open_context(ctx);
	goto out;

T
Trond Myklebust 已提交
1609
no_open_dput:
L
Linus Torvalds 已提交
1610
	dput(parent);
T
Trond Myklebust 已提交
1611
no_open:
L
Linus Torvalds 已提交
1612 1613
	return nfs_lookup_revalidate(dentry, nd);
}
1614

A
Al Viro 已提交
1615 1616
static int nfs_open_create(struct inode *dir, struct dentry *dentry,
		umode_t mode, struct nameidata *nd)
1617 1618 1619 1620
{
	struct nfs_open_context *ctx = NULL;
	struct iattr attr;
	int error;
1621
	int open_flags = O_CREAT|O_EXCL;
1622 1623 1624 1625 1626 1627 1628

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

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

1629
	if (nd)
1630 1631
		open_flags = nd->intent.open.flags;

A
Al Viro 已提交
1632 1633 1634 1635
	ctx = create_nfs_open_context(dentry, open_flags);
	error = PTR_ERR(ctx);
	if (IS_ERR(ctx))
		goto out_err_drop;
1636 1637 1638 1639

	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, ctx);
	if (error != 0)
		goto out_put_ctx;
1640
	if (nd) {
1641 1642 1643
		error = nfs_intent_set_file(nd, ctx);
		if (error < 0)
			goto out_err;
A
Al Viro 已提交
1644 1645
	} else {
		put_nfs_open_context(ctx);
1646
	}
1647 1648
	return 0;
out_put_ctx:
A
Al Viro 已提交
1649
	put_nfs_open_context(ctx);
1650
out_err_drop:
1651
	d_drop(dentry);
1652
out_err:
1653 1654 1655
	return error;
}

L
Linus Torvalds 已提交
1656 1657 1658 1659 1660 1661 1662 1663
#endif /* CONFIG_NFSV4 */

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
				struct nfs_fattr *fattr)
{
1664 1665
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1666 1667 1668
	struct inode *inode;
	int error = -EACCES;

1669 1670
	d_drop(dentry);

L
Linus Torvalds 已提交
1671 1672
	/* We may have been initialized further down */
	if (dentry->d_inode)
1673
		goto out;
L
Linus Torvalds 已提交
1674
	if (fhandle->size == 0) {
1675
		error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1676
		if (error)
1677
			goto out_error;
L
Linus Torvalds 已提交
1678
	}
1679
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1680 1681
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1682
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr);
L
Linus Torvalds 已提交
1683
		if (error < 0)
1684
			goto out_error;
L
Linus Torvalds 已提交
1685 1686
	}
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
1687 1688
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1689 1690 1691 1692
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1693
	return 0;
1694 1695 1696 1697
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1698 1699 1700 1701 1702 1703 1704 1705
}

/*
 * Following a failed create operation, we drop the dentry rather
 * than retain a negative dentry. This avoids a problem in the event
 * that the operation succeeded on the server, but an error in the
 * reply path made it appear to have failed.
 */
A
Al Viro 已提交
1706 1707
static int nfs_create(struct inode *dir, struct dentry *dentry,
		umode_t mode, struct nameidata *nd)
L
Linus Torvalds 已提交
1708 1709 1710
{
	struct iattr attr;
	int error;
1711
	int open_flags = O_CREAT|O_EXCL;
L
Linus Torvalds 已提交
1712

C
Chuck Lever 已提交
1713 1714
	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
1715 1716 1717 1718

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

1719
	if (nd)
1720 1721 1722
		open_flags = nd->intent.open.flags;

	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, NULL);
L
Linus Torvalds 已提交
1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
static int
A
Al Viro 已提交
1735
nfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t rdev)
L
Linus Torvalds 已提交
1736 1737 1738 1739
{
	struct iattr attr;
	int status;

C
Chuck Lever 已提交
1740 1741
	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
1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760

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

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

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

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1761
static int nfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
1762 1763 1764 1765
{
	struct iattr attr;
	int error;

C
Chuck Lever 已提交
1766 1767
	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
1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780

	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;
}

1781 1782 1783 1784 1785 1786
static void nfs_dentry_handle_enoent(struct dentry *dentry)
{
	if (dentry->d_inode != NULL && !d_unhashed(dentry))
		d_delete(dentry);
}

L
Linus Torvalds 已提交
1787 1788 1789 1790
static int nfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	int error;

C
Chuck Lever 已提交
1791 1792
	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
1793 1794 1795 1796

	error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
	/* Ensure the VFS deletes this inode */
	if (error == 0 && dentry->d_inode != NULL)
1797
		clear_nlink(dentry->d_inode);
1798 1799
	else if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826

	return error;
}

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

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

	if (inode != NULL) {
1827
		nfs_inode_return_delegation(inode);
L
Linus Torvalds 已提交
1828 1829 1830
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		/* The VFS may want to delete this inode */
		if (error == 0)
1831
			nfs_drop_nlink(inode);
1832
		nfs_mark_for_revalidate(inode);
L
Linus Torvalds 已提交
1833 1834
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1835 1836
	if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854
out:
	return error;
}

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

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

	spin_lock(&dentry->d_lock);
N
Nick Piggin 已提交
1855
	if (dentry->d_count > 1) {
L
Linus Torvalds 已提交
1856
		spin_unlock(&dentry->d_lock);
T
Trond Myklebust 已提交
1857 1858
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1859 1860 1861 1862 1863 1864 1865 1866 1867
		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);
1868
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1869 1870 1871 1872 1873 1874
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
	return error;
}

1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890
/*
 * To create a symbolic link, most file systems instantiate a new inode,
 * add a page to it containing the path, then write it out to the disk
 * using prepare_write/commit_write.
 *
 * Unfortunately the NFS client can't create the in-core inode first
 * because it needs a file handle to create an in-core inode (see
 * fs/nfs/inode.c:nfs_fhget).  We only have a file handle *after* the
 * symlink request has completed on the server.
 *
 * So instead we allocate a raw page, copy the symname into it, then do
 * the SYMLINK request with the page as the buffer.  If it succeeds, we
 * now have a new file handle and can instantiate an in-core NFS inode
 * and move the raw page into its mapping.
 */
static int nfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
L
Linus Torvalds 已提交
1891
{
1892 1893 1894
	struct pagevec lru_pvec;
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1895
	struct iattr attr;
1896
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1897 1898 1899 1900 1901
	int error;

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

1902 1903
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1904

1905 1906
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1907

1908
	page = alloc_page(GFP_HIGHUSER);
1909
	if (!page)
1910 1911
		return -ENOMEM;

1912
	kaddr = kmap_atomic(page);
1913 1914 1915
	memcpy(kaddr, symname, pathlen);
	if (pathlen < PAGE_SIZE)
		memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen);
1916
	kunmap_atomic(kaddr);
1917

1918
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1919 1920 1921 1922
	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
1923
		d_drop(dentry);
1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934
		__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)) {
1935
		pagevec_add(&lru_pvec, page);
1936
		pagevec_lru_add_file(&lru_pvec);
1937 1938 1939 1940 1941 1942
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954
}

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

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

1955 1956
	nfs_inode_return_delegation(inode);

1957
	d_drop(dentry);
L
Linus Torvalds 已提交
1958
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1959
	if (error == 0) {
A
Al Viro 已提交
1960
		ihold(inode);
1961
		d_add(dentry, inode);
1962
	}
L
Linus Torvalds 已提交
1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
	return error;
}

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

	dfprintk(VFS, "NFS: rename(%s/%s -> %s/%s, ct=%d)\n",
		 old_dentry->d_parent->d_name.name, old_dentry->d_name.name,
		 new_dentry->d_parent->d_name.name, new_dentry->d_name.name,
N
Nick Piggin 已提交
2001
		 new_dentry->d_count);
L
Linus Torvalds 已提交
2002 2003

	/*
M
Miklos Szeredi 已提交
2004 2005 2006 2007
	 * 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
2008
	 */
2009 2010 2011 2012 2013 2014 2015 2016 2017
	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
2018

N
Nick Piggin 已提交
2019
		if (new_dentry->d_count > 2) {
2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
			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);
2030
			if (err)
2031
				goto out;
2032 2033

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
2034
			rehash = NULL;
2035
			new_inode = NULL;
2036
		}
2037
	}
L
Linus Torvalds 已提交
2038

2039
	nfs_inode_return_delegation(old_inode);
2040
	if (new_inode != NULL)
2041
		nfs_inode_return_delegation(new_inode);
L
Linus Torvalds 已提交
2042 2043 2044

	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
2045
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
2046 2047 2048 2049
out:
	if (rehash)
		d_rehash(rehash);
	if (!error) {
2050 2051
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
2052
		d_move(old_dentry, new_dentry);
2053 2054
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
2055 2056
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
2057 2058 2059 2060 2061 2062 2063

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

2064 2065 2066 2067
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

2068 2069 2070 2071
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
2072 2073 2074
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
2075 2076
}

2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087
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);
	}
}

2088 2089
int nfs_access_cache_shrinker(struct shrinker *shrink,
			      struct shrink_control *sc)
2090 2091
{
	LIST_HEAD(head);
2092
	struct nfs_inode *nfsi, *next;
2093
	struct nfs_access_entry *cache;
2094 2095
	int nr_to_scan = sc->nr_to_scan;
	gfp_t gfp_mask = sc->gfp_mask;
2096

2097 2098
	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
		return (nr_to_scan == 0) ? 0 : -1;
2099

2100
	spin_lock(&nfs_access_lru_lock);
2101
	list_for_each_entry_safe(nfsi, next, &nfs_access_lru_list, access_cache_inode_lru) {
2102 2103 2104 2105
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
2106
		inode = &nfsi->vfs_inode;
2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119
		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);
2120
			smp_mb__before_clear_bit();
2121
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
2122
			smp_mb__after_clear_bit();
2123
		}
2124
		spin_unlock(&inode->i_lock);
2125 2126
	}
	spin_unlock(&nfs_access_lru_lock);
2127
	nfs_access_free_list(&head);
2128 2129 2130
	return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure;
}

2131
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
2132
{
2133
	struct rb_root *root_node = &nfsi->access_cache;
2134
	struct rb_node *n;
2135 2136 2137 2138 2139 2140
	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);
2141
		list_move(&entry->lru, head);
2142 2143
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
2144 2145
}

2146
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
2147
{
2148 2149 2150 2151
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
2152
	/* Remove from global LRU init */
2153 2154
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
2155 2156
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

2157
	spin_lock(&inode->i_lock);
2158 2159 2160 2161
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
2162
}
L
Linus Torvalds 已提交
2163

2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177
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
2178
	}
2179 2180 2181
	return NULL;
}

2182
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
2183 2184 2185 2186 2187
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

2188
	spin_lock(&inode->i_lock);
2189 2190 2191 2192 2193
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
2194
	if (!nfs_have_delegated_attributes(inode) &&
2195
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
2196 2197 2198 2199
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
2200
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
2201 2202 2203 2204 2205 2206
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
2207
	list_del(&cache->lru);
2208 2209 2210 2211
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
2212 2213
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
2214 2215 2216 2217 2218
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
2219 2220
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238
	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);
2239
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
2240
	spin_unlock(&inode->i_lock);
2241 2242 2243
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
2244 2245
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
2246 2247 2248 2249
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

2250
static void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
2251 2252 2253 2254 2255
{
	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
2256
	cache->jiffies = set->jiffies;
2257
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
2258
	cache->mask = set->mask;
2259 2260

	nfs_access_add_rbtree(inode, cache);
2261 2262 2263 2264 2265 2266 2267

	/* 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 */
2268
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
2269
		spin_lock(&nfs_access_lru_lock);
2270 2271 2272
		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);
2273 2274
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290
}

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);
2291 2292 2293 2294 2295 2296
	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
2297
		return status;
2298
	}
L
Linus Torvalds 已提交
2299 2300
	nfs_access_add_cache(inode, &cache);
out:
2301
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2302 2303 2304 2305
		return 0;
	return -EACCES;
}

2306 2307 2308 2309
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

2310
	if ((openflags & O_ACCMODE) != O_WRONLY)
2311
		mask |= MAY_READ;
2312
	if ((openflags & O_ACCMODE) != O_RDONLY)
2313
		mask |= MAY_WRITE;
2314
	if (openflags & __FMODE_EXEC)
2315 2316 2317 2318 2319 2320 2321 2322 2323
		mask |= MAY_EXEC;
	return mask;
}

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

2324
int nfs_permission(struct inode *inode, int mask)
L
Linus Torvalds 已提交
2325 2326 2327 2328
{
	struct rpc_cred *cred;
	int res = 0;

2329
	if (mask & MAY_NOT_BLOCK)
2330 2331
		return -ECHILD;

C
Chuck Lever 已提交
2332 2333
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

2334
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2335 2336
		goto out;
	/* Is this sys_access() ? */
E
Eric Paris 已提交
2337
	if (mask & (MAY_ACCESS | MAY_CHDIR))
L
Linus Torvalds 已提交
2338 2339 2340 2341 2342 2343 2344 2345
		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)
2346 2347
					&& (mask & MAY_OPEN)
					&& !(mask & MAY_EXEC))
L
Linus Torvalds 已提交
2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362
				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;

2363
	cred = rpc_lookup_cred();
L
Linus Torvalds 已提交
2364 2365 2366 2367 2368 2369
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
out:
2370 2371 2372
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

C
Chuck Lever 已提交
2373 2374
	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
2375 2376 2377 2378
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
2379
		res = generic_permission(inode, mask);
C
Chuck Lever 已提交
2380
	goto out;
L
Linus Torvalds 已提交
2381 2382 2383 2384 2385 2386 2387 2388
}

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