dir.c 59.3 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
/*
 *  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 50 51 52 53 54 55 56 57 58
static int nfs_readdir(struct file *, void *, filldir_t);
static struct dentry *nfs_lookup(struct inode *, struct dentry *, struct nameidata *);
static int nfs_create(struct inode *, struct dentry *, int, struct nameidata *);
static int nfs_mkdir(struct inode *, struct dentry *, int);
static int nfs_rmdir(struct inode *, struct dentry *);
static int nfs_unlink(struct inode *, struct dentry *);
static int nfs_symlink(struct inode *, struct dentry *, const char *);
static int nfs_link(struct dentry *, struct inode *, struct dentry *);
static int nfs_mknod(struct inode *, struct dentry *, int, dev_t);
static int nfs_rename(struct inode *, struct dentry *,
		      struct inode *, struct dentry *);
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 *);
115
static int nfs_open_create(struct inode *dir, struct dentry *dentry, int 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
207 208 209 210 211 212 213 214 215
};

struct nfs_cache_array {
	unsigned int size;
	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, KM_USER0);
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, KM_USER0);
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
471 472 473 474 475 476 477 478 479
		goto different;
	return 1;
different:
	return 0;
}

static
void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
{
480 481 482 483 484 485
	struct qstr filename = {
		.len = entry->len,
		.name = entry->name,
	};
	struct dentry *dentry;
	struct dentry *alias;
B
Bryan Schumaker 已提交
486 487 488
	struct inode *dir = parent->d_inode;
	struct inode *inode;

489 490 491 492 493 494 495
	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
496

497
	dentry = d_lookup(parent, &filename);
B
Bryan Schumaker 已提交
498 499 500 501 502 503 504 505 506 507 508
	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);
509 510 511
	if (dentry == NULL)
		return;

B
Bryan Schumaker 已提交
512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
	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
529 530
/* Perform conversion from xdr to cache array */
static
531
int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *entry,
532
				struct page **xdr_pages, struct page *page, unsigned int buflen)
L
Linus Torvalds 已提交
533
{
534
	struct xdr_stream stream;
535
	struct xdr_buf buf;
536
	struct page *scratch;
B
Bryan Schumaker 已提交
537
	struct nfs_cache_array *array;
538 539
	unsigned int count = 0;
	int status;
540

541 542 543
	scratch = alloc_page(GFP_KERNEL);
	if (scratch == NULL)
		return -ENOMEM;
544

545
	xdr_init_decode_pages(&stream, &buf, xdr_pages, buflen);
546
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
B
Bryan Schumaker 已提交
547 548 549

	do {
		status = xdr_decode(desc, entry, &stream);
550 551 552
		if (status != 0) {
			if (status == -EAGAIN)
				status = 0;
B
Bryan Schumaker 已提交
553
			break;
554
		}
B
Bryan Schumaker 已提交
555

556 557
		count++;

T
Trond Myklebust 已提交
558
		if (desc->plus != 0)
B
Bryan Schumaker 已提交
559
			nfs_prime_dcache(desc->file->f_path.dentry, entry);
560 561 562 563

		status = nfs_readdir_add_to_array(entry, page);
		if (status != 0)
			break;
B
Bryan Schumaker 已提交
564 565
	} while (!entry->eof);

T
Trond Myklebust 已提交
566
	if (count == 0 || (status == -EBADCOOKIE && entry->eof != 0)) {
B
Bryan Schumaker 已提交
567
		array = nfs_readdir_get_array(page);
568 569 570 571
		if (!IS_ERR(array)) {
			array->eof_index = array->size;
			status = 0;
			nfs_readdir_release_array(page);
572 573
		} else
			status = PTR_ERR(array);
L
Linus Torvalds 已提交
574
	}
575 576

	put_page(scratch);
577
	return status;
B
Bryan Schumaker 已提交
578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
}

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
600
int nfs_readdir_large_page(struct page **pages, unsigned int npages)
B
Bryan Schumaker 已提交
601 602 603 604 605 606 607 608 609
{
	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;
	}
610
	return 0;
B
Bryan Schumaker 已提交
611 612 613

out_freepages:
	nfs_readdir_free_pagearray(pages, i);
614
	return -ENOMEM;
L
Linus Torvalds 已提交
615 616
}

B
Bryan Schumaker 已提交
617 618
static
int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode)
619
{
B
Bryan Schumaker 已提交
620 621
	struct page *pages[NFS_MAX_READDIR_PAGES];
	void *pages_ptr = NULL;
B
Bryan Schumaker 已提交
622 623 624
	struct nfs_entry entry;
	struct file	*file = desc->file;
	struct nfs_cache_array *array;
625
	int status = -ENOMEM;
B
Bryan Schumaker 已提交
626
	unsigned int array_size = ARRAY_SIZE(pages);
B
Bryan Schumaker 已提交
627 628

	entry.prev_cookie = 0;
629
	entry.cookie = desc->last_cookie;
B
Bryan Schumaker 已提交
630 631 632
	entry.eof = 0;
	entry.fh = nfs_alloc_fhandle();
	entry.fattr = nfs_alloc_fattr();
633
	entry.server = NFS_SERVER(inode);
B
Bryan Schumaker 已提交
634 635
	if (entry.fh == NULL || entry.fattr == NULL)
		goto out;
636

B
Bryan Schumaker 已提交
637
	array = nfs_readdir_get_array(page);
638 639 640 641
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
642 643
	memset(array, 0, sizeof(struct nfs_cache_array));
	array->eof_index = -1;
644

645 646
	status = nfs_readdir_large_page(pages, array_size);
	if (status < 0)
B
Bryan Schumaker 已提交
647 648
		goto out_release_array;
	do {
649
		unsigned int pglen;
B
Bryan Schumaker 已提交
650
		status = nfs_readdir_xdr_filler(pages, desc, &entry, file, inode);
651

B
Bryan Schumaker 已提交
652
		if (status < 0)
653
			break;
654
		pglen = status;
655
		status = nfs_readdir_page_filler(desc, &entry, pages, page, pglen);
656 657 658 659 660 661
		if (status < 0) {
			if (status == -ENOSPC)
				status = 0;
			break;
		}
	} while (array->eof_index < 0);
B
Bryan Schumaker 已提交
662

B
Bryan Schumaker 已提交
663
	nfs_readdir_free_large_page(pages_ptr, pages, array_size);
B
Bryan Schumaker 已提交
664 665 666 667 668
out_release_array:
	nfs_readdir_release_array(page);
out:
	nfs_free_fattr(entry.fattr);
	nfs_free_fhandle(entry.fh);
669 670 671 672
	return status;
}

/*
B
Bryan Schumaker 已提交
673 674 675 676
 * 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
677
 */
B
Bryan Schumaker 已提交
678 679
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page)
L
Linus Torvalds 已提交
680
{
681
	struct inode	*inode = desc->file->f_path.dentry->d_inode;
682
	int ret;
L
Linus Torvalds 已提交
683

684 685
	ret = nfs_readdir_xdr_to_array(desc, page, inode);
	if (ret < 0)
B
Bryan Schumaker 已提交
686 687
		goto error;
	SetPageUptodate(page);
L
Linus Torvalds 已提交
688

B
Bryan Schumaker 已提交
689 690 691
	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
692
	}
B
Bryan Schumaker 已提交
693 694 695 696
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
697
	return ret;
B
Bryan Schumaker 已提交
698
}
L
Linus Torvalds 已提交
699

B
Bryan Schumaker 已提交
700 701 702
static
void cache_page_release(nfs_readdir_descriptor_t *desc)
{
703 704
	if (!desc->page->mapping)
		nfs_readdir_clear_array(desc->page);
B
Bryan Schumaker 已提交
705 706 707 708 709 710 711
	page_cache_release(desc->page);
	desc->page = NULL;
}

static
struct page *get_cache_page(nfs_readdir_descriptor_t *desc)
{
712
	return read_cache_page(desc->file->f_path.dentry->d_inode->i_mapping,
B
Bryan Schumaker 已提交
713
			desc->page_index, (filler_t *)nfs_readdir_filler, desc);
L
Linus Torvalds 已提交
714 715 716
}

/*
B
Bryan Schumaker 已提交
717
 * Returns 0 if desc->dir_cookie was found on page desc->page_index
L
Linus Torvalds 已提交
718
 */
B
Bryan Schumaker 已提交
719 720 721 722 723 724 725 726 727 728
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
729 730
	if (res != 0)
		cache_page_release(desc);
B
Bryan Schumaker 已提交
731 732 733 734
	return res;
}

/* Search for desc->dir_cookie from the beginning of the page cache */
L
Linus Torvalds 已提交
735 736 737
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
738
	int res;
B
Bryan Schumaker 已提交
739

740
	if (desc->page_index == 0) {
741
		desc->current_index = 0;
742 743
		desc->last_cookie = 0;
	}
T
Trond Myklebust 已提交
744
	do {
B
Bryan Schumaker 已提交
745
		res = find_cache_page(desc);
T
Trond Myklebust 已提交
746
	} while (res == -EAGAIN);
L
Linus Torvalds 已提交
747 748 749 750 751 752 753 754 755 756 757
	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
758 759 760
	int i = 0;
	int res = 0;
	struct nfs_cache_array *array = NULL;
761 762
	struct nfs_open_dir_context *ctx = file->private_data;

B
Bryan Schumaker 已提交
763
	array = nfs_readdir_get_array(desc->page);
764 765 766 767
	if (IS_ERR(array)) {
		res = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
768 769

	for (i = desc->cache_entry_index; i < array->size; i++) {
770
		struct nfs_cache_array_entry *ent;
L
Linus Torvalds 已提交
771

772 773
		ent = &array->array[i];
		if (filldir(dirent, ent->string.name, ent->string.len,
774 775
		    file->f_pos, nfs_compat_user_ino64(ent->ino),
		    ent->d_type) < 0) {
776
			desc->eof = 1;
L
Linus Torvalds 已提交
777
			break;
778
		}
779
		file->f_pos++;
B
Bryan Schumaker 已提交
780 781 782 783
		if (i < (array->size-1))
			*desc->dir_cookie = array->array[i+1].cookie;
		else
			*desc->dir_cookie = array->last_cookie;
784 785
		if (ctx->duped != 0)
			ctx->duped = 1;
L
Linus Torvalds 已提交
786
	}
T
Trond Myklebust 已提交
787
	if (array->eof_index >= 0)
788
		desc->eof = 1;
B
Bryan Schumaker 已提交
789 790

	nfs_readdir_release_array(desc->page);
791
out:
B
Bryan Schumaker 已提交
792
	cache_page_release(desc);
C
Chuck Lever 已提交
793 794
	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
795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815
	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
816
	struct inode *inode = desc->file->f_path.dentry->d_inode;
817
	struct nfs_open_dir_context *ctx = desc->file->private_data;
L
Linus Torvalds 已提交
818

C
Chuck Lever 已提交
819 820
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
821 822 823 824 825 826

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
B
Bryan Schumaker 已提交
827

828
	desc->page_index = 0;
829
	desc->last_cookie = *desc->dir_cookie;
830
	desc->page = page;
831
	ctx->duped = 0;
832

833 834
	status = nfs_readdir_xdr_to_array(desc, page, inode);
	if (status < 0)
L
Linus Torvalds 已提交
835 836 837 838 839
		goto out_release;

	status = nfs_do_filldir(desc, dirent, filldir);

 out:
C
Chuck Lever 已提交
840
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
841
			__func__, status);
L
Linus Torvalds 已提交
842 843
	return status;
 out_release:
B
Bryan Schumaker 已提交
844
	cache_page_release(desc);
L
Linus Torvalds 已提交
845 846 847
	goto out;
}

848 849 850
/* 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
851 852 853
 */
static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
854
	struct dentry	*dentry = filp->f_path.dentry;
L
Linus Torvalds 已提交
855 856 857
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
858
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
T
Trond Myklebust 已提交
859
	int res;
L
Linus Torvalds 已提交
860

C
Chuck Lever 已提交
861
	dfprintk(FILE, "NFS: readdir(%s/%s) starting at cookie %llu\n",
C
Chuck Lever 已提交
862 863
			dentry->d_parent->d_name.name, dentry->d_name.name,
			(long long)filp->f_pos);
C
Chuck Lever 已提交
864 865
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
866
	/*
867
	 * filp->f_pos points to the dirent entry number.
868
	 * *desc->dir_cookie has the cookie for the next entry. We have
869 870
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
871 872 873 874
	 */
	memset(desc, 0, sizeof(*desc));

	desc->file = filp;
875
	desc->dir_cookie = &dir_ctx->dir_cookie;
L
Linus Torvalds 已提交
876 877 878
	desc->decode = NFS_PROTO(inode)->decode_dirent;
	desc->plus = NFS_USE_READDIRPLUS(inode);

T
Trond Myklebust 已提交
879
	nfs_block_sillyrename(dentry);
880
	res = nfs_revalidate_mapping(inode, filp->f_mapping);
T
Trond Myklebust 已提交
881 882 883
	if (res < 0)
		goto out;

T
Trond Myklebust 已提交
884
	do {
L
Linus Torvalds 已提交
885
		res = readdir_search_pagecache(desc);
886

L
Linus Torvalds 已提交
887
		if (res == -EBADCOOKIE) {
888
			res = 0;
L
Linus Torvalds 已提交
889
			/* This means either end of directory */
B
Bryan Schumaker 已提交
890
			if (*desc->dir_cookie && desc->eof == 0) {
L
Linus Torvalds 已提交
891 892
				/* Or that the server has 'lost' a cookie */
				res = uncached_readdir(desc, dirent, filldir);
893
				if (res == 0)
L
Linus Torvalds 已提交
894 895 896 897 898
					continue;
			}
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
899
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
900
			nfs_zap_caches(inode);
901
			desc->page_index = 0;
L
Linus Torvalds 已提交
902
			desc->plus = 0;
B
Bryan Schumaker 已提交
903
			desc->eof = 0;
L
Linus Torvalds 已提交
904 905 906 907 908 909
			continue;
		}
		if (res < 0)
			break;

		res = nfs_do_filldir(desc, dirent, filldir);
910
		if (res < 0)
L
Linus Torvalds 已提交
911
			break;
T
Trond Myklebust 已提交
912
	} while (!desc->eof);
T
Trond Myklebust 已提交
913
out:
T
Trond Myklebust 已提交
914
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
915 916
	if (res > 0)
		res = 0;
917
	dfprintk(FILE, "NFS: readdir(%s/%s) returns %d\n",
C
Chuck Lever 已提交
918 919 920
			dentry->d_parent->d_name.name, dentry->d_name.name,
			res);
	return res;
L
Linus Torvalds 已提交
921 922
}

923
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
924
{
925 926
	struct dentry *dentry = filp->f_path.dentry;
	struct inode *inode = dentry->d_inode;
927
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
928

C
Chuck Lever 已提交
929
	dfprintk(FILE, "NFS: llseek dir(%s/%s, %lld, %d)\n",
930 931 932 933 934
			dentry->d_parent->d_name.name,
			dentry->d_name.name,
			offset, origin);

	mutex_lock(&inode->i_mutex);
935 936 937 938 939 940 941 942 943 944 945 946
	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;
947
		dir_ctx->dir_cookie = 0;
948
		dir_ctx->duped = 0;
949 950
	}
out:
951
	mutex_unlock(&inode->i_mutex);
952 953 954
	return offset;
}

L
Linus Torvalds 已提交
955 956 957 958
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
959 960
static int nfs_fsync_dir(struct file *filp, loff_t start, loff_t end,
			 int datasync)
L
Linus Torvalds 已提交
961
{
962
	struct dentry *dentry = filp->f_path.dentry;
963
	struct inode *inode = dentry->d_inode;
964

C
Chuck Lever 已提交
965
	dfprintk(FILE, "NFS: fsync dir(%s/%s) datasync %d\n",
C
Chuck Lever 已提交
966 967 968
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);

969
	mutex_lock(&inode->i_mutex);
970
	nfs_inc_stats(dentry->d_inode, NFSIOS_VFSFSYNC);
971
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
972 973 974
	return 0;
}

975 976 977 978 979 980 981 982 983 984 985 986
/**
 * 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)
{
987
	NFS_I(dir)->cache_change_attribute++;
988 989
}

L
Linus Torvalds 已提交
990 991 992 993 994
/*
 * 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.
 */
995
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
996 997 998
{
	if (IS_ROOT(dentry))
		return 1;
999 1000
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
1001 1002 1003 1004 1005 1006 1007 1008
	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
1009 1010
}

1011 1012 1013 1014
/*
 * Return the intent data that applies to this particular path component
 *
 * Note that the current set of intents only apply to the very last
1015 1016
 * component of the path and none of them is set before that last
 * component.
1017
 */
1018 1019
static inline unsigned int nfs_lookup_check_intent(struct nameidata *nd,
						unsigned int mask)
1020 1021 1022 1023
{
	return nd->flags & mask;
}

1024 1025 1026 1027 1028 1029 1030 1031
/*
 * 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;
1032
	return nd && nfs_lookup_check_intent(nd, LOOKUP_EXCL);
1033 1034
}

1035 1036 1037 1038 1039 1040 1041 1042
/*
 * 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
1043 1044 1045 1046 1047
static inline
int nfs_lookup_verify_inode(struct inode *inode, struct nameidata *nd)
{
	struct nfs_server *server = NFS_SERVER(inode);

1048
	if (IS_AUTOMOUNT(inode))
1049
		return 0;
L
Linus Torvalds 已提交
1050 1051
	if (nd != NULL) {
		/* VFS wants an on-the-wire revalidation */
1052
		if (nd->flags & LOOKUP_REVAL)
L
Linus Torvalds 已提交
1053 1054
			goto out_force;
		/* This is an open(2) */
1055
		if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 &&
1056 1057 1058
				!(server->flags & NFS_MOUNT_NOCTO) &&
				(S_ISREG(inode->i_mode) ||
				 S_ISDIR(inode->i_mode)))
L
Linus Torvalds 已提交
1059
			goto out_force;
1060
		return 0;
L
Linus Torvalds 已提交
1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
	}
	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 */
1079
	if (nd != NULL && nfs_lookup_check_intent(nd, LOOKUP_CREATE) != 0)
L
Linus Torvalds 已提交
1080
		return 0;
1081 1082
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
	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.
 */
1097
static int nfs_lookup_revalidate(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
1098 1099 1100 1101
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
1102 1103
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1104 1105
	int error;

1106 1107 1108
	if (nd->flags & LOOKUP_RCU)
		return -ECHILD;

L
Linus Torvalds 已提交
1109 1110
	parent = dget_parent(dentry);
	dir = parent->d_inode;
C
Chuck Lever 已提交
1111
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
1112 1113 1114 1115 1116 1117 1118 1119 1120
	inode = dentry->d_inode;

	if (!inode) {
		if (nfs_neg_need_reval(dir, dentry, nd))
			goto out_bad;
		goto out_valid;
	}

	if (is_bad_inode(inode)) {
C
Chuck Lever 已提交
1121
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
1122
				__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1123
				dentry->d_name.name);
L
Linus Torvalds 已提交
1124 1125 1126
		goto out_bad;
	}

1127 1128 1129
	if (nfs_have_delegation(inode, FMODE_READ))
		goto out_set_verifier;

L
Linus Torvalds 已提交
1130
	/* Force a full look up iff the parent directory has changed */
1131
	if (!nfs_is_exclusive_create(dir, nd) && nfs_check_verifier(dir, dentry)) {
L
Linus Torvalds 已提交
1132 1133 1134 1135 1136 1137 1138 1139
		if (nfs_lookup_verify_inode(inode, nd))
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

1140 1141 1142 1143 1144 1145
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

1146
	error = NFS_PROTO(dir)->lookup(NFS_SERVER(dir)->client, dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1147 1148
	if (error)
		goto out_bad;
1149
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
1150
		goto out_bad;
1151
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
1152 1153
		goto out_bad;

1154 1155
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1156
out_set_verifier:
1157
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1158 1159
 out_valid:
	dput(parent);
C
Chuck Lever 已提交
1160
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n",
1161
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1162
			dentry->d_name.name);
L
Linus Torvalds 已提交
1163 1164 1165 1166
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
1167
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
1168 1169 1170 1171 1172 1173
	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;
1174 1175
		if (dentry->d_flags & DCACHE_DISCONNECTED)
			goto out_valid;
L
Linus Torvalds 已提交
1176 1177 1178
		shrink_dcache_parent(dentry);
	}
	d_drop(dentry);
1179 1180
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1181
	dput(parent);
C
Chuck Lever 已提交
1182
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
1183
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1184
			dentry->d_name.name);
L
Linus Torvalds 已提交
1185
	return 0;
1186 1187 1188 1189 1190 1191 1192 1193
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
1194 1195 1196 1197 1198
}

/*
 * This is called from dput() when d_count is going to 0.
 */
N
Nick Piggin 已提交
1199
static int nfs_dentry_delete(const struct dentry *dentry)
L
Linus Torvalds 已提交
1200 1201 1202 1203 1204
{
	dfprintk(VFS, "NFS: dentry_delete(%s/%s, %x)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		dentry->d_flags);

1205 1206 1207 1208
	/* 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
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
	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;

}

1222 1223 1224 1225 1226 1227 1228 1229
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
1230 1231 1232 1233 1234 1235
/*
 * 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)
{
1236 1237 1238 1239
	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
1240
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
1241
		drop_nlink(inode);
1242
		nfs_complete_unlink(dentry, inode);
L
Linus Torvalds 已提交
1243 1244 1245 1246
	}
	iput(inode);
}

1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257
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
1258
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
1259 1260 1261
	.d_revalidate	= nfs_lookup_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1262
	.d_automount	= nfs_d_automount,
1263
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1264 1265 1266 1267 1268
};

static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
{
	struct dentry *res;
T
Trond Myklebust 已提交
1269
	struct dentry *parent;
L
Linus Torvalds 已提交
1270
	struct inode *inode = NULL;
1271 1272
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1273 1274 1275 1276
	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
1277
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
1278 1279 1280 1281 1282

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

1283 1284 1285 1286 1287 1288 1289
	/*
	 * 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;
1290
		goto out;
1291
	}
L
Linus Torvalds 已提交
1292

1293 1294 1295 1296 1297 1298
	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
1299 1300 1301
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
	nfs_block_sillyrename(parent);
1302
	error = NFS_PROTO(dir)->lookup(NFS_SERVER(dir)->client, dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1303 1304 1305 1306
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
1307
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1308
	}
1309
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
N
Namhyung Kim 已提交
1310
	res = ERR_CAST(inode);
1311
	if (IS_ERR(res))
T
Trond Myklebust 已提交
1312
		goto out_unblock_sillyrename;
1313

L
Linus Torvalds 已提交
1314
no_entry:
1315
	res = d_materialise_unique(dentry, inode);
1316 1317
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
1318
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1319
		dentry = res;
1320
	}
L
Linus Torvalds 已提交
1321
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
1322 1323
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
L
Linus Torvalds 已提交
1324
out:
1325 1326
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1327 1328 1329 1330 1331 1332
	return res;
}

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

A
Al Viro 已提交
1333
const struct dentry_operations nfs4_dentry_operations = {
L
Linus Torvalds 已提交
1334 1335 1336
	.d_revalidate	= nfs_open_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1337
	.d_automount	= nfs_d_automount,
1338
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1339 1340
};

1341 1342 1343 1344
/*
 * 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
1345
static int is_atomic_open(struct nameidata *nd)
L
Linus Torvalds 已提交
1346
{
1347
	if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_OPEN) == 0)
L
Linus Torvalds 已提交
1348 1349 1350 1351 1352
		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? */
1353
	if (__mnt_is_readonly(nd->path.mnt) &&
1354
	    (nd->intent.open.flags & (O_CREAT|O_TRUNC|O_ACCMODE)))
L
Linus Torvalds 已提交
1355 1356 1357 1358
		return 0;
	return 1;
}

1359 1360 1361 1362 1363 1364 1365 1366 1367 1368
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;
}

1369
static struct nfs_open_context *create_nfs_open_context(struct dentry *dentry, int open_flags)
1370 1371 1372
{
	struct nfs_open_context *ctx;
	struct rpc_cred *cred;
1373
	fmode_t fmode = flags_to_mode(open_flags);
1374 1375 1376 1377

	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return ERR_CAST(cred);
1378
	ctx = alloc_nfs_open_context(dentry, cred, fmode);
1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
	put_rpccred(cred);
	if (ctx == NULL)
		return ERR_PTR(-ENOMEM);
	return ctx;
}

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) {
1398
		ret = nfs_may_open(ctx->dentry->d_inode,
1399 1400 1401 1402 1403
				ctx->cred,
				nd->intent.open.flags);
		if (ret < 0)
			goto out;
	}
1404
	filp = lookup_instantiate_filp(nd, ctx->dentry, do_open);
1405 1406 1407 1408 1409 1410 1411 1412 1413
	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
1414 1415
static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
{
1416 1417
	struct nfs_open_context *ctx;
	struct iattr attr;
L
Linus Torvalds 已提交
1418
	struct dentry *res = NULL;
1419
	struct inode *inode;
1420
	int open_flags;
1421
	int err;
L
Linus Torvalds 已提交
1422

C
Chuck Lever 已提交
1423 1424 1425
	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
1426
	/* Check that we are indeed trying to open this file */
T
Trond Myklebust 已提交
1427
	if (!is_atomic_open(nd))
L
Linus Torvalds 已提交
1428 1429 1430 1431 1432 1433 1434
		goto no_open;

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

1435 1436
	/* Let vfs_create() deal with O_EXCL. Instantiate, but don't hash
	 * the dentry. */
1437
	if (nd->flags & LOOKUP_EXCL) {
1438
		d_instantiate(dentry, NULL);
1439 1440
		goto out;
	}
L
Linus Torvalds 已提交
1441

1442 1443 1444
	open_flags = nd->intent.open.flags;

	ctx = create_nfs_open_context(dentry, open_flags);
1445 1446 1447 1448 1449 1450 1451
	res = ERR_CAST(ctx);
	if (IS_ERR(ctx))
		goto out;

	if (nd->flags & LOOKUP_CREATE) {
		attr.ia_mode = nd->intent.open.create_mode;
		attr.ia_valid = ATTR_MODE;
A
Aneesh Kumar K.V 已提交
1452
		attr.ia_mode &= ~current_umask();
1453
	} else {
1454
		open_flags &= ~(O_EXCL | O_CREAT);
1455 1456 1457
		attr.ia_valid = 0;
	}

L
Linus Torvalds 已提交
1458
	/* Open the file on the server */
1459
	nfs_block_sillyrename(dentry->d_parent);
1460
	inode = NFS_PROTO(dir)->open_context(dir, ctx, open_flags, &attr);
1461 1462
	if (IS_ERR(inode)) {
		nfs_unblock_sillyrename(dentry->d_parent);
1463
		put_nfs_open_context(ctx);
1464
		switch (PTR_ERR(inode)) {
L
Linus Torvalds 已提交
1465 1466
			/* Make a negative dentry */
			case -ENOENT:
1467
				d_add(dentry, NULL);
1468 1469
				res = NULL;
				goto out;
L
Linus Torvalds 已提交
1470
			/* This turned out not to be a regular file */
1471 1472
			case -ENOTDIR:
				goto no_open;
L
Linus Torvalds 已提交
1473 1474 1475
			case -ELOOP:
				if (!(nd->intent.open.flags & O_NOFOLLOW))
					goto no_open;
1476
			/* case -EISDIR: */
L
Linus Torvalds 已提交
1477 1478
			/* case -EINVAL: */
			default:
1479
				res = ERR_CAST(inode);
L
Linus Torvalds 已提交
1480 1481
				goto out;
		}
1482
	}
1483
	res = d_add_unique(dentry, inode);
1484
	nfs_unblock_sillyrename(dentry->d_parent);
1485
	if (res != NULL) {
1486 1487
		dput(ctx->dentry);
		ctx->dentry = dget(res);
L
Linus Torvalds 已提交
1488
		dentry = res;
1489
	}
1490 1491 1492 1493 1494 1495
	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
1496
out:
1497
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1498 1499 1500 1501 1502 1503 1504 1505
	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
1506
	struct inode *inode;
L
Linus Torvalds 已提交
1507
	struct inode *dir;
1508
	struct nfs_open_context *ctx;
L
Linus Torvalds 已提交
1509 1510
	int openflags, ret = 0;

N
Nick Piggin 已提交
1511 1512 1513 1514
	if (nd->flags & LOOKUP_RCU)
		return -ECHILD;

	inode = dentry->d_inode;
1515
	if (!is_atomic_open(nd) || d_mountpoint(dentry))
T
Trond Myklebust 已提交
1516
		goto no_open;
1517

L
Linus Torvalds 已提交
1518 1519
	parent = dget_parent(dentry);
	dir = parent->d_inode;
1520

L
Linus Torvalds 已提交
1521 1522 1523
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1524 1525 1526
	if (inode == NULL) {
		if (!nfs_neg_need_reval(dir, dentry, nd))
			ret = 1;
L
Linus Torvalds 已提交
1527
		goto out;
1528 1529
	}

L
Linus Torvalds 已提交
1530 1531
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
T
Trond Myklebust 已提交
1532
		goto no_open_dput;
L
Linus Torvalds 已提交
1533 1534 1535
	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
1536
		goto no_open_dput;
L
Linus Torvalds 已提交
1537
	/* We can't create new files, or truncate existing ones here */
1538
	openflags &= ~(O_CREAT|O_EXCL|O_TRUNC);
L
Linus Torvalds 已提交
1539

1540
	ctx = create_nfs_open_context(dentry, openflags);
1541 1542 1543
	ret = PTR_ERR(ctx);
	if (IS_ERR(ctx))
		goto out;
L
Linus Torvalds 已提交
1544
	/*
1545
	 * Note: we're not holding inode->i_mutex and so may be racing with
L
Linus Torvalds 已提交
1546 1547 1548
	 * operations that change the directory. We therefore save the
	 * change attribute *before* we do the RPC call.
	 */
1549
	inode = NFS_PROTO(dir)->open_context(dir, ctx, openflags, NULL);
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563
	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);
1564
	if (inode != dentry->d_inode)
1565
		goto out_drop;
1566 1567 1568 1569 1570

	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
1571 1572 1573
out:
	dput(parent);
	return ret;
1574 1575 1576 1577 1578 1579 1580
out_drop:
	d_drop(dentry);
	ret = 0;
out_put_ctx:
	put_nfs_open_context(ctx);
	goto out;

T
Trond Myklebust 已提交
1581
no_open_dput:
L
Linus Torvalds 已提交
1582
	dput(parent);
T
Trond Myklebust 已提交
1583
no_open:
L
Linus Torvalds 已提交
1584 1585
	return nfs_lookup_revalidate(dentry, nd);
}
1586 1587 1588 1589 1590 1591 1592

static int nfs_open_create(struct inode *dir, struct dentry *dentry, int mode,
		struct nameidata *nd)
{
	struct nfs_open_context *ctx = NULL;
	struct iattr attr;
	int error;
1593
	int open_flags = O_CREAT|O_EXCL;
1594 1595 1596 1597 1598 1599 1600

	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;

1601
	if (nd)
1602 1603
		open_flags = nd->intent.open.flags;

A
Al Viro 已提交
1604 1605 1606 1607
	ctx = create_nfs_open_context(dentry, open_flags);
	error = PTR_ERR(ctx);
	if (IS_ERR(ctx))
		goto out_err_drop;
1608 1609 1610 1611

	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, ctx);
	if (error != 0)
		goto out_put_ctx;
1612
	if (nd) {
1613 1614 1615
		error = nfs_intent_set_file(nd, ctx);
		if (error < 0)
			goto out_err;
A
Al Viro 已提交
1616 1617
	} else {
		put_nfs_open_context(ctx);
1618
	}
1619 1620
	return 0;
out_put_ctx:
A
Al Viro 已提交
1621
	put_nfs_open_context(ctx);
1622
out_err_drop:
1623
	d_drop(dentry);
1624
out_err:
1625 1626 1627
	return error;
}

L
Linus Torvalds 已提交
1628 1629 1630 1631 1632 1633 1634 1635
#endif /* CONFIG_NFSV4 */

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
				struct nfs_fattr *fattr)
{
1636 1637
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1638 1639 1640
	struct inode *inode;
	int error = -EACCES;

1641 1642
	d_drop(dentry);

L
Linus Torvalds 已提交
1643 1644
	/* We may have been initialized further down */
	if (dentry->d_inode)
1645
		goto out;
L
Linus Torvalds 已提交
1646
	if (fhandle->size == 0) {
1647
		error = NFS_PROTO(dir)->lookup(NFS_SERVER(dir)->client, dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1648
		if (error)
1649
			goto out_error;
L
Linus Torvalds 已提交
1650
	}
1651
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1652 1653
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1654
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr);
L
Linus Torvalds 已提交
1655
		if (error < 0)
1656
			goto out_error;
L
Linus Torvalds 已提交
1657 1658
	}
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
1659 1660
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1661 1662 1663 1664
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1665
	return 0;
1666 1667 1668 1669
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682
}

/*
 * Following a failed create operation, we drop the dentry rather
 * than retain a negative dentry. This avoids a problem in the event
 * that the operation succeeded on the server, but an error in the
 * reply path made it appear to have failed.
 */
static int nfs_create(struct inode *dir, struct dentry *dentry, int mode,
		struct nameidata *nd)
{
	struct iattr attr;
	int error;
1683
	int open_flags = O_CREAT|O_EXCL;
L
Linus Torvalds 已提交
1684

C
Chuck Lever 已提交
1685 1686
	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
1687 1688 1689 1690

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

1691
	if (nd)
1692 1693 1694
		open_flags = nd->intent.open.flags;

	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, NULL);
L
Linus Torvalds 已提交
1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711
	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
nfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t rdev)
{
	struct iattr attr;
	int status;

C
Chuck Lever 已提交
1712 1713
	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
1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737

	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.
 */
static int nfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
	struct iattr attr;
	int error;

C
Chuck Lever 已提交
1738 1739
	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
1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752

	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;
}

1753 1754 1755 1756 1757 1758
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
1759 1760 1761 1762
static int nfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	int error;

C
Chuck Lever 已提交
1763 1764
	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
1765 1766 1767 1768

	error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
	/* Ensure the VFS deletes this inode */
	if (error == 0 && dentry->d_inode != NULL)
1769
		clear_nlink(dentry->d_inode);
1770 1771
	else if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798

	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) {
1799
		nfs_inode_return_delegation(inode);
L
Linus Torvalds 已提交
1800 1801 1802
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		/* The VFS may want to delete this inode */
		if (error == 0)
1803
			nfs_drop_nlink(inode);
1804
		nfs_mark_for_revalidate(inode);
L
Linus Torvalds 已提交
1805 1806
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1807 1808
	if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826
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
1827
	if (dentry->d_count > 1) {
L
Linus Torvalds 已提交
1828
		spin_unlock(&dentry->d_lock);
T
Trond Myklebust 已提交
1829 1830
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1831 1832 1833 1834 1835 1836 1837 1838 1839
		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);
1840
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1841 1842 1843 1844 1845 1846
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
	return error;
}

1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862
/*
 * 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
1863
{
1864 1865 1866
	struct pagevec lru_pvec;
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1867
	struct iattr attr;
1868
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1869 1870 1871 1872 1873
	int error;

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

1874 1875
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1876

1877 1878
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1879

1880
	page = alloc_page(GFP_HIGHUSER);
1881
	if (!page)
1882 1883 1884 1885 1886 1887 1888 1889
		return -ENOMEM;

	kaddr = kmap_atomic(page, KM_USER0);
	memcpy(kaddr, symname, pathlen);
	if (pathlen < PAGE_SIZE)
		memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen);
	kunmap_atomic(kaddr, KM_USER0);

1890
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1891 1892 1893 1894
	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
1895
		d_drop(dentry);
1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906
		__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)) {
1907
		pagevec_add(&lru_pvec, page);
1908
		pagevec_lru_add_file(&lru_pvec);
1909 1910 1911 1912 1913 1914
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
}

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);

1927 1928
	nfs_inode_return_delegation(inode);

1929
	d_drop(dentry);
L
Linus Torvalds 已提交
1930
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1931
	if (error == 0) {
A
Al Viro 已提交
1932
		ihold(inode);
1933
		d_add(dentry, inode);
1934
	}
L
Linus Torvalds 已提交
1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
	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
1973
		 new_dentry->d_count);
L
Linus Torvalds 已提交
1974 1975

	/*
M
Miklos Szeredi 已提交
1976 1977 1978 1979
	 * 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
1980
	 */
1981 1982 1983 1984 1985 1986 1987 1988 1989
	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
1990

N
Nick Piggin 已提交
1991
		if (new_dentry->d_count > 2) {
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
			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);
2002
			if (err)
2003
				goto out;
2004 2005

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
2006
			rehash = NULL;
2007
			new_inode = NULL;
2008
		}
2009
	}
L
Linus Torvalds 已提交
2010

2011
	nfs_inode_return_delegation(old_inode);
2012
	if (new_inode != NULL)
2013
		nfs_inode_return_delegation(new_inode);
L
Linus Torvalds 已提交
2014 2015 2016

	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
2017
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
2018 2019 2020 2021
out:
	if (rehash)
		d_rehash(rehash);
	if (!error) {
2022 2023
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
2024
		d_move(old_dentry, new_dentry);
2025 2026
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
2027 2028
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
2029 2030 2031 2032 2033 2034 2035

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

2036 2037 2038 2039
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

2040 2041 2042 2043
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
2044 2045 2046
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
2047 2048
}

2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
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);
	}
}

2060 2061
int nfs_access_cache_shrinker(struct shrinker *shrink,
			      struct shrink_control *sc)
2062 2063
{
	LIST_HEAD(head);
2064
	struct nfs_inode *nfsi, *next;
2065
	struct nfs_access_entry *cache;
2066 2067
	int nr_to_scan = sc->nr_to_scan;
	gfp_t gfp_mask = sc->gfp_mask;
2068

2069 2070
	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
		return (nr_to_scan == 0) ? 0 : -1;
2071

2072
	spin_lock(&nfs_access_lru_lock);
2073
	list_for_each_entry_safe(nfsi, next, &nfs_access_lru_list, access_cache_inode_lru) {
2074 2075 2076 2077
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
2078
		inode = &nfsi->vfs_inode;
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091
		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);
2092
			smp_mb__before_clear_bit();
2093
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
2094
			smp_mb__after_clear_bit();
2095
		}
2096
		spin_unlock(&inode->i_lock);
2097 2098
	}
	spin_unlock(&nfs_access_lru_lock);
2099
	nfs_access_free_list(&head);
2100 2101 2102
	return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure;
}

2103
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
2104
{
2105
	struct rb_root *root_node = &nfsi->access_cache;
2106
	struct rb_node *n;
2107 2108 2109 2110 2111 2112
	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);
2113
		list_move(&entry->lru, head);
2114 2115
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
2116 2117
}

2118
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
2119
{
2120 2121 2122 2123
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
2124
	/* Remove from global LRU init */
2125 2126
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
2127 2128
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

2129
	spin_lock(&inode->i_lock);
2130 2131 2132 2133
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
2134
}
L
Linus Torvalds 已提交
2135

2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149
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
2150
	}
2151 2152 2153
	return NULL;
}

2154
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
2155 2156 2157 2158 2159
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

2160
	spin_lock(&inode->i_lock);
2161 2162 2163 2164 2165
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
2166
	if (!nfs_have_delegated_attributes(inode) &&
2167
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
2168 2169 2170 2171
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
2172
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
2173 2174 2175 2176 2177 2178
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
2179
	list_del(&cache->lru);
2180 2181 2182 2183
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
2184 2185
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
2186 2187 2188 2189 2190
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
2191 2192
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210
	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);
2211
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
2212
	spin_unlock(&inode->i_lock);
2213 2214 2215
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
2216 2217
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
2218 2219 2220 2221
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

2222
static void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
2223 2224 2225 2226 2227
{
	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
2228
	cache->jiffies = set->jiffies;
2229
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
2230
	cache->mask = set->mask;
2231 2232

	nfs_access_add_rbtree(inode, cache);
2233 2234 2235 2236 2237 2238 2239

	/* 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 */
2240
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
2241
		spin_lock(&nfs_access_lru_lock);
2242 2243 2244
		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);
2245 2246
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
}

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);
2263 2264 2265 2266 2267 2268
	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
2269
		return status;
2270
	}
L
Linus Torvalds 已提交
2271 2272
	nfs_access_add_cache(inode, &cache);
out:
2273
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2274 2275 2276 2277
		return 0;
	return -EACCES;
}

2278 2279 2280 2281
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

2282
	if ((openflags & O_ACCMODE) != O_WRONLY)
2283
		mask |= MAY_READ;
2284
	if ((openflags & O_ACCMODE) != O_RDONLY)
2285
		mask |= MAY_WRITE;
2286
	if (openflags & __FMODE_EXEC)
2287 2288 2289 2290 2291 2292 2293 2294 2295
		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));
}

2296
int nfs_permission(struct inode *inode, int mask)
L
Linus Torvalds 已提交
2297 2298 2299 2300
{
	struct rpc_cred *cred;
	int res = 0;

2301
	if (mask & MAY_NOT_BLOCK)
2302 2303
		return -ECHILD;

C
Chuck Lever 已提交
2304 2305
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

2306
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2307 2308
		goto out;
	/* Is this sys_access() ? */
E
Eric Paris 已提交
2309
	if (mask & (MAY_ACCESS | MAY_CHDIR))
L
Linus Torvalds 已提交
2310 2311 2312 2313 2314 2315 2316 2317
		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)
2318 2319
					&& (mask & MAY_OPEN)
					&& !(mask & MAY_EXEC))
L
Linus Torvalds 已提交
2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334
				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;

2335
	cred = rpc_lookup_cred();
L
Linus Torvalds 已提交
2336 2337 2338 2339 2340 2341
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
out:
2342 2343 2344
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

C
Chuck Lever 已提交
2345 2346
	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
2347 2348 2349 2350
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
2351
		res = generic_permission(inode, mask);
C
Chuck Lever 已提交
2352
	goto out;
L
Linus Torvalds 已提交
2353 2354 2355 2356 2357 2358 2359 2360
}

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