dir.c 58.7 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 *, 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 */

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

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
157 158 159 160 161 162
/*
 * Open file
 */
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
163 164 165
	int res = 0;
	struct nfs_open_dir_context *ctx;
	struct rpc_cred *cred;
L
Linus Torvalds 已提交
166

C
Chuck Lever 已提交
167
	dfprintk(FILE, "NFS: open dir(%s/%s)\n",
168 169 170 171
			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
172

173 174 175 176 177 178 179 180 181
	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return PTR_ERR(cred);
	ctx = alloc_nfs_open_dir_context(cred);
	if (IS_ERR(ctx)) {
		res = PTR_ERR(ctx);
		goto out;
	}
	filp->private_data = ctx;
182 183 184 185 186 187 188
	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);
	}
189 190
out:
	put_rpccred(cred);
L
Linus Torvalds 已提交
191 192 193
	return res;
}

194 195 196 197 198 199 200
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
201 202 203 204
struct nfs_cache_array_entry {
	u64 cookie;
	u64 ino;
	struct qstr string;
205
	unsigned char d_type;
B
Bryan Schumaker 已提交
206 207 208 209 210 211 212 213 214
};

struct nfs_cache_array {
	unsigned int size;
	int eof_index;
	u64 last_cookie;
	struct nfs_cache_array_entry array[0];
};

215
typedef int (*decode_dirent_t)(struct xdr_stream *, struct nfs_entry *, int);
L
Linus Torvalds 已提交
216 217 218 219
typedef struct {
	struct file	*file;
	struct page	*page;
	unsigned long	page_index;
220
	u64		*dir_cookie;
221
	u64		last_cookie;
222
	loff_t		current_index;
L
Linus Torvalds 已提交
223
	decode_dirent_t	decode;
B
Bryan Schumaker 已提交
224

225
	unsigned long	timestamp;
226
	unsigned long	gencount;
B
Bryan Schumaker 已提交
227 228 229
	unsigned int	cache_entry_index;
	unsigned int	plus:1;
	unsigned int	eof:1;
L
Linus Torvalds 已提交
230 231
} nfs_readdir_descriptor_t;

B
Bryan Schumaker 已提交
232 233
/*
 * The caller is responsible for calling nfs_readdir_release_array(page)
L
Linus Torvalds 已提交
234 235
 */
static
B
Bryan Schumaker 已提交
236 237
struct nfs_cache_array *nfs_readdir_get_array(struct page *page)
{
238
	void *ptr;
B
Bryan Schumaker 已提交
239 240
	if (page == NULL)
		return ERR_PTR(-EIO);
241 242 243 244
	ptr = kmap(page);
	if (ptr == NULL)
		return ERR_PTR(-ENOMEM);
	return ptr;
B
Bryan Schumaker 已提交
245 246 247 248 249 250 251 252 253 254 255 256
}

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

/*
 * we are freeing strings created by nfs_add_to_readdir_array()
 */
static
257
void nfs_readdir_clear_array(struct page *page)
B
Bryan Schumaker 已提交
258
{
259
	struct nfs_cache_array *array;
B
Bryan Schumaker 已提交
260
	int i;
261

262
	array = kmap_atomic(page, KM_USER0);
B
Bryan Schumaker 已提交
263 264
	for (i = 0; i < array->size; i++)
		kfree(array->array[i].string.name);
265
	kunmap_atomic(array, KM_USER0);
B
Bryan Schumaker 已提交
266 267 268 269 270 271 272 273
}

/*
 * 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
274
int nfs_readdir_make_qstr(struct qstr *string, const char *name, unsigned int len)
B
Bryan Schumaker 已提交
275 276 277
{
	string->len = len;
	string->name = kmemdup(name, len, GFP_KERNEL);
278 279
	if (string->name == NULL)
		return -ENOMEM;
280 281 282 283 284
	/*
	 * 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);
285 286
	string->hash = full_name_hash(name, len);
	return 0;
B
Bryan Schumaker 已提交
287 288 289 290 291 292
}

static
int nfs_readdir_add_to_array(struct nfs_entry *entry, struct page *page)
{
	struct nfs_cache_array *array = nfs_readdir_get_array(page);
293 294 295
	struct nfs_cache_array_entry *cache_entry;
	int ret;

B
Bryan Schumaker 已提交
296 297
	if (IS_ERR(array))
		return PTR_ERR(array);
298 299 300 301

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

	/* Check that this entry lies within the page bounds */
302
	ret = -ENOSPC;
303
	if ((char *)&cache_entry[1] - (char *)page_address(page) > PAGE_SIZE)
304
		goto out;
B
Bryan Schumaker 已提交
305

306 307
	cache_entry->cookie = entry->prev_cookie;
	cache_entry->ino = entry->ino;
308
	cache_entry->d_type = entry->d_type;
309 310 311
	ret = nfs_readdir_make_qstr(&cache_entry->string, entry->name, entry->len);
	if (ret)
		goto out;
B
Bryan Schumaker 已提交
312
	array->last_cookie = entry->cookie;
313
	array->size++;
T
Trond Myklebust 已提交
314
	if (entry->eof != 0)
B
Bryan Schumaker 已提交
315
		array->eof_index = array->size;
316
out:
B
Bryan Schumaker 已提交
317
	nfs_readdir_release_array(page);
318
	return ret;
B
Bryan Schumaker 已提交
319 320 321 322 323 324 325
}

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;
326
	struct nfs_open_dir_context *ctx = desc->file->private_data;
B
Bryan Schumaker 已提交
327 328 329 330

	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
			goto out_eof;
		return -EAGAIN;
	}

	index = (unsigned int)diff;
	*desc->dir_cookie = array->array[index].cookie;
	desc->cache_entry_index = index;
339
	ctx->duped = 0;
B
Bryan Schumaker 已提交
340 341 342 343 344 345 346 347 348 349
	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;
350
	loff_t new_pos;
B
Bryan Schumaker 已提交
351
	int status = -EAGAIN;
352
	struct nfs_open_dir_context *ctx = desc->file->private_data;
B
Bryan Schumaker 已提交
353 354 355

	for (i = 0; i < array->size; i++) {
		if (array->array[i].cookie == *desc->dir_cookie) {
356 357 358 359 360 361
			new_pos = desc->current_index + i;
			if (new_pos < desc->file->f_pos) {
				ctx->dup_cookie = *desc->dir_cookie;
				ctx->duped = 1;
			}
			desc->file->f_pos = new_pos;
B
Bryan Schumaker 已提交
362
			desc->cache_entry_index = i;
T
Trond Myklebust 已提交
363
			return 0;
B
Bryan Schumaker 已提交
364 365
		}
	}
T
Trond Myklebust 已提交
366
	if (array->eof_index >= 0) {
367
		status = -EBADCOOKIE;
368 369
		if (*desc->dir_cookie == array->last_cookie)
			desc->eof = 1;
370
	}
B
Bryan Schumaker 已提交
371 372 373 374 375 376 377
	return status;
}

static
int nfs_readdir_search_array(nfs_readdir_descriptor_t *desc)
{
	struct nfs_cache_array *array;
T
Trond Myklebust 已提交
378
	int status;
B
Bryan Schumaker 已提交
379 380 381 382 383 384 385 386 387 388 389 390

	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
391
	if (status == -EAGAIN) {
392
		desc->last_cookie = array->last_cookie;
393
		desc->current_index += array->size;
T
Trond Myklebust 已提交
394 395
		desc->page_index++;
	}
B
Bryan Schumaker 已提交
396 397 398 399 400 401 402
	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
403
int nfs_readdir_xdr_filler(struct page **pages, nfs_readdir_descriptor_t *desc,
B
Bryan Schumaker 已提交
404
			struct nfs_entry *entry, struct file *file, struct inode *inode)
L
Linus Torvalds 已提交
405
{
406 407
	struct nfs_open_dir_context *ctx = file->private_data;
	struct rpc_cred	*cred = ctx->cred;
408
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
409 410 411 412
	int		error;

 again:
	timestamp = jiffies;
413
	gencount = nfs_inc_attr_generation_counter();
B
Bryan Schumaker 已提交
414
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, entry->cookie, pages,
L
Linus Torvalds 已提交
415 416 417 418 419
					  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
420
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
421 422 423 424 425
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
426
	desc->timestamp = timestamp;
427
	desc->gencount = gencount;
B
Bryan Schumaker 已提交
428 429
error:
	return error;
L
Linus Torvalds 已提交
430 431
}

432 433
static int xdr_decode(nfs_readdir_descriptor_t *desc,
		      struct nfs_entry *entry, struct xdr_stream *xdr)
L
Linus Torvalds 已提交
434
{
435
	int error;
L
Linus Torvalds 已提交
436

437 438 439
	error = desc->decode(xdr, entry, desc->plus);
	if (error)
		return error;
B
Bryan Schumaker 已提交
440 441 442
	entry->fattr->time_start = desc->timestamp;
	entry->fattr->gencount = desc->gencount;
	return 0;
L
Linus Torvalds 已提交
443 444
}

B
Bryan Schumaker 已提交
445 446 447 448 449
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
450
	if (nfs_compare_fh(entry->fh, NFS_FH(dentry->d_inode)) != 0)
B
Bryan Schumaker 已提交
451 452 453 454 455 456 457 458 459
		goto different;
	return 1;
different:
	return 0;
}

static
void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
{
460 461 462 463 464 465
	struct qstr filename = {
		.len = entry->len,
		.name = entry->name,
	};
	struct dentry *dentry;
	struct dentry *alias;
B
Bryan Schumaker 已提交
466 467 468
	struct inode *dir = parent->d_inode;
	struct inode *inode;

469 470 471 472 473 474 475
	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
476

477
	dentry = d_lookup(parent, &filename);
B
Bryan Schumaker 已提交
478 479 480 481 482 483 484 485 486 487 488
	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);
489 490 491
	if (dentry == NULL)
		return;

B
Bryan Schumaker 已提交
492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508
	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
509 510
/* Perform conversion from xdr to cache array */
static
511
int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *entry,
512
				struct page **xdr_pages, struct page *page, unsigned int buflen)
L
Linus Torvalds 已提交
513
{
514
	struct xdr_stream stream;
515 516 517 518 519 520 521
	struct xdr_buf buf = {
		.pages = xdr_pages,
		.page_len = buflen,
		.buflen = buflen,
		.len = buflen,
	};
	struct page *scratch;
B
Bryan Schumaker 已提交
522
	struct nfs_cache_array *array;
523 524
	unsigned int count = 0;
	int status;
525

526 527 528
	scratch = alloc_page(GFP_KERNEL);
	if (scratch == NULL)
		return -ENOMEM;
529

530 531
	xdr_init_decode(&stream, &buf, NULL);
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
B
Bryan Schumaker 已提交
532 533 534

	do {
		status = xdr_decode(desc, entry, &stream);
535 536 537
		if (status != 0) {
			if (status == -EAGAIN)
				status = 0;
B
Bryan Schumaker 已提交
538
			break;
539
		}
B
Bryan Schumaker 已提交
540

541 542
		count++;

T
Trond Myklebust 已提交
543
		if (desc->plus != 0)
B
Bryan Schumaker 已提交
544
			nfs_prime_dcache(desc->file->f_path.dentry, entry);
545 546 547 548

		status = nfs_readdir_add_to_array(entry, page);
		if (status != 0)
			break;
B
Bryan Schumaker 已提交
549 550
	} while (!entry->eof);

T
Trond Myklebust 已提交
551
	if (count == 0 || (status == -EBADCOOKIE && entry->eof != 0)) {
B
Bryan Schumaker 已提交
552
		array = nfs_readdir_get_array(page);
553 554 555 556
		if (!IS_ERR(array)) {
			array->eof_index = array->size;
			status = 0;
			nfs_readdir_release_array(page);
557 558
		} else
			status = PTR_ERR(array);
L
Linus Torvalds 已提交
559
	}
560 561

	put_page(scratch);
562
	return status;
B
Bryan Schumaker 已提交
563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
}

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
585
int nfs_readdir_large_page(struct page **pages, unsigned int npages)
B
Bryan Schumaker 已提交
586 587 588 589 590 591 592 593 594
{
	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;
	}
595
	return 0;
B
Bryan Schumaker 已提交
596 597 598

out_freepages:
	nfs_readdir_free_pagearray(pages, i);
599
	return -ENOMEM;
L
Linus Torvalds 已提交
600 601
}

B
Bryan Schumaker 已提交
602 603
static
int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode)
604
{
B
Bryan Schumaker 已提交
605 606
	struct page *pages[NFS_MAX_READDIR_PAGES];
	void *pages_ptr = NULL;
B
Bryan Schumaker 已提交
607 608 609
	struct nfs_entry entry;
	struct file	*file = desc->file;
	struct nfs_cache_array *array;
610
	int status = -ENOMEM;
B
Bryan Schumaker 已提交
611
	unsigned int array_size = ARRAY_SIZE(pages);
B
Bryan Schumaker 已提交
612 613

	entry.prev_cookie = 0;
614
	entry.cookie = desc->last_cookie;
B
Bryan Schumaker 已提交
615 616 617
	entry.eof = 0;
	entry.fh = nfs_alloc_fhandle();
	entry.fattr = nfs_alloc_fattr();
618
	entry.server = NFS_SERVER(inode);
B
Bryan Schumaker 已提交
619 620
	if (entry.fh == NULL || entry.fattr == NULL)
		goto out;
621

B
Bryan Schumaker 已提交
622
	array = nfs_readdir_get_array(page);
623 624 625 626
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
627 628
	memset(array, 0, sizeof(struct nfs_cache_array));
	array->eof_index = -1;
629

630 631
	status = nfs_readdir_large_page(pages, array_size);
	if (status < 0)
B
Bryan Schumaker 已提交
632 633
		goto out_release_array;
	do {
634
		unsigned int pglen;
B
Bryan Schumaker 已提交
635
		status = nfs_readdir_xdr_filler(pages, desc, &entry, file, inode);
636

B
Bryan Schumaker 已提交
637
		if (status < 0)
638
			break;
639
		pglen = status;
640
		status = nfs_readdir_page_filler(desc, &entry, pages, page, pglen);
641 642 643 644 645 646
		if (status < 0) {
			if (status == -ENOSPC)
				status = 0;
			break;
		}
	} while (array->eof_index < 0);
B
Bryan Schumaker 已提交
647

B
Bryan Schumaker 已提交
648
	nfs_readdir_free_large_page(pages_ptr, pages, array_size);
B
Bryan Schumaker 已提交
649 650 651 652 653
out_release_array:
	nfs_readdir_release_array(page);
out:
	nfs_free_fattr(entry.fattr);
	nfs_free_fhandle(entry.fh);
654 655 656 657
	return status;
}

/*
B
Bryan Schumaker 已提交
658 659 660 661
 * 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
662
 */
B
Bryan Schumaker 已提交
663 664
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page)
L
Linus Torvalds 已提交
665
{
666
	struct inode	*inode = desc->file->f_path.dentry->d_inode;
667
	int ret;
L
Linus Torvalds 已提交
668

669 670
	ret = nfs_readdir_xdr_to_array(desc, page, inode);
	if (ret < 0)
B
Bryan Schumaker 已提交
671 672
		goto error;
	SetPageUptodate(page);
L
Linus Torvalds 已提交
673

B
Bryan Schumaker 已提交
674 675 676
	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
677
	}
B
Bryan Schumaker 已提交
678 679 680 681
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
682
	return ret;
B
Bryan Schumaker 已提交
683
}
L
Linus Torvalds 已提交
684

B
Bryan Schumaker 已提交
685 686 687
static
void cache_page_release(nfs_readdir_descriptor_t *desc)
{
688 689
	if (!desc->page->mapping)
		nfs_readdir_clear_array(desc->page);
B
Bryan Schumaker 已提交
690 691 692 693 694 695 696
	page_cache_release(desc->page);
	desc->page = NULL;
}

static
struct page *get_cache_page(nfs_readdir_descriptor_t *desc)
{
697
	return read_cache_page(desc->file->f_path.dentry->d_inode->i_mapping,
B
Bryan Schumaker 已提交
698
			desc->page_index, (filler_t *)nfs_readdir_filler, desc);
L
Linus Torvalds 已提交
699 700 701
}

/*
B
Bryan Schumaker 已提交
702
 * Returns 0 if desc->dir_cookie was found on page desc->page_index
L
Linus Torvalds 已提交
703
 */
B
Bryan Schumaker 已提交
704 705 706 707 708 709 710 711 712 713
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
714 715
	if (res != 0)
		cache_page_release(desc);
B
Bryan Schumaker 已提交
716 717 718 719
	return res;
}

/* Search for desc->dir_cookie from the beginning of the page cache */
L
Linus Torvalds 已提交
720 721 722
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
723
	int res;
B
Bryan Schumaker 已提交
724

725
	if (desc->page_index == 0) {
726
		desc->current_index = 0;
727 728
		desc->last_cookie = 0;
	}
T
Trond Myklebust 已提交
729
	do {
B
Bryan Schumaker 已提交
730
		res = find_cache_page(desc);
T
Trond Myklebust 已提交
731
	} while (res == -EAGAIN);
L
Linus Torvalds 已提交
732 733 734 735 736 737 738 739 740 741 742
	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
743 744 745
	int i = 0;
	int res = 0;
	struct nfs_cache_array *array = NULL;
746 747 748 749 750 751 752 753 754 755 756 757 758 759
	struct nfs_open_dir_context *ctx = file->private_data;

	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.  "
				"Offending cookie: %llu\n",
				file->f_dentry->d_parent->d_name.name,
				file->f_dentry->d_name.name,
				*desc->dir_cookie);
		}
		res = -ELOOP;
		goto out;
	}
B
Bryan Schumaker 已提交
760 761

	array = nfs_readdir_get_array(desc->page);
762 763 764 765
	if (IS_ERR(array)) {
		res = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
766 767

	for (i = desc->cache_entry_index; i < array->size; i++) {
768
		struct nfs_cache_array_entry *ent;
L
Linus Torvalds 已提交
769

770 771
		ent = &array->array[i];
		if (filldir(dirent, ent->string.name, ent->string.len,
772 773
		    file->f_pos, nfs_compat_user_ino64(ent->ino),
		    ent->d_type) < 0) {
774
			desc->eof = 1;
L
Linus Torvalds 已提交
775
			break;
776
		}
777
		file->f_pos++;
B
Bryan Schumaker 已提交
778 779 780 781
		if (i < (array->size-1))
			*desc->dir_cookie = array->array[i+1].cookie;
		else
			*desc->dir_cookie = array->last_cookie;
L
Linus Torvalds 已提交
782
	}
T
Trond Myklebust 已提交
783
	if (array->eof_index >= 0)
784
		desc->eof = 1;
B
Bryan Schumaker 已提交
785 786

	nfs_readdir_release_array(desc->page);
787
out:
B
Bryan Schumaker 已提交
788
	cache_page_release(desc);
C
Chuck Lever 已提交
789 790
	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
791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811
	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
812
	struct inode *inode = desc->file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
813

C
Chuck Lever 已提交
814 815
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
816 817 818 819 820 821

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
B
Bryan Schumaker 已提交
822

823
	desc->page_index = 0;
824
	desc->last_cookie = *desc->dir_cookie;
825 826
	desc->page = page;

827 828
	status = nfs_readdir_xdr_to_array(desc, page, inode);
	if (status < 0)
L
Linus Torvalds 已提交
829 830 831 832 833
		goto out_release;

	status = nfs_do_filldir(desc, dirent, filldir);

 out:
C
Chuck Lever 已提交
834
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
835
			__func__, status);
L
Linus Torvalds 已提交
836 837
	return status;
 out_release:
B
Bryan Schumaker 已提交
838
	cache_page_release(desc);
L
Linus Torvalds 已提交
839 840 841
	goto out;
}

842 843 844
/* 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
845 846 847
 */
static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
848
	struct dentry	*dentry = filp->f_path.dentry;
L
Linus Torvalds 已提交
849 850 851
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
852
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
T
Trond Myklebust 已提交
853
	int res;
L
Linus Torvalds 已提交
854

C
Chuck Lever 已提交
855
	dfprintk(FILE, "NFS: readdir(%s/%s) starting at cookie %llu\n",
C
Chuck Lever 已提交
856 857
			dentry->d_parent->d_name.name, dentry->d_name.name,
			(long long)filp->f_pos);
C
Chuck Lever 已提交
858 859
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
860
	/*
861
	 * filp->f_pos points to the dirent entry number.
862
	 * *desc->dir_cookie has the cookie for the next entry. We have
863 864
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
865 866 867 868
	 */
	memset(desc, 0, sizeof(*desc));

	desc->file = filp;
869
	desc->dir_cookie = &dir_ctx->dir_cookie;
L
Linus Torvalds 已提交
870 871 872
	desc->decode = NFS_PROTO(inode)->decode_dirent;
	desc->plus = NFS_USE_READDIRPLUS(inode);

T
Trond Myklebust 已提交
873
	nfs_block_sillyrename(dentry);
874
	res = nfs_revalidate_mapping(inode, filp->f_mapping);
T
Trond Myklebust 已提交
875 876 877
	if (res < 0)
		goto out;

T
Trond Myklebust 已提交
878
	do {
L
Linus Torvalds 已提交
879
		res = readdir_search_pagecache(desc);
880

L
Linus Torvalds 已提交
881
		if (res == -EBADCOOKIE) {
882
			res = 0;
L
Linus Torvalds 已提交
883
			/* This means either end of directory */
B
Bryan Schumaker 已提交
884
			if (*desc->dir_cookie && desc->eof == 0) {
L
Linus Torvalds 已提交
885 886
				/* Or that the server has 'lost' a cookie */
				res = uncached_readdir(desc, dirent, filldir);
887
				if (res == 0)
L
Linus Torvalds 已提交
888 889 890 891 892
					continue;
			}
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
893
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
894
			nfs_zap_caches(inode);
895
			desc->page_index = 0;
L
Linus Torvalds 已提交
896
			desc->plus = 0;
B
Bryan Schumaker 已提交
897
			desc->eof = 0;
L
Linus Torvalds 已提交
898 899 900 901 902 903
			continue;
		}
		if (res < 0)
			break;

		res = nfs_do_filldir(desc, dirent, filldir);
904
		if (res < 0)
L
Linus Torvalds 已提交
905
			break;
T
Trond Myklebust 已提交
906
	} while (!desc->eof);
T
Trond Myklebust 已提交
907
out:
T
Trond Myklebust 已提交
908
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
909 910
	if (res > 0)
		res = 0;
911
	dfprintk(FILE, "NFS: readdir(%s/%s) returns %d\n",
C
Chuck Lever 已提交
912 913 914
			dentry->d_parent->d_name.name, dentry->d_name.name,
			res);
	return res;
L
Linus Torvalds 已提交
915 916
}

917
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
918
{
919 920
	struct dentry *dentry = filp->f_path.dentry;
	struct inode *inode = dentry->d_inode;
921
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
922

C
Chuck Lever 已提交
923
	dfprintk(FILE, "NFS: llseek dir(%s/%s, %lld, %d)\n",
924 925 926 927 928
			dentry->d_parent->d_name.name,
			dentry->d_name.name,
			offset, origin);

	mutex_lock(&inode->i_mutex);
929 930 931 932 933 934 935 936 937 938 939 940
	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;
941
		dir_ctx->dir_cookie = 0;
942
		dir_ctx->duped = 0;
943 944
	}
out:
945
	mutex_unlock(&inode->i_mutex);
946 947 948
	return offset;
}

L
Linus Torvalds 已提交
949 950 951 952
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
953
static int nfs_fsync_dir(struct file *filp, int datasync)
L
Linus Torvalds 已提交
954
{
955 956
	struct dentry *dentry = filp->f_path.dentry;

C
Chuck Lever 已提交
957
	dfprintk(FILE, "NFS: fsync dir(%s/%s) datasync %d\n",
C
Chuck Lever 已提交
958 959 960
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);

961
	nfs_inc_stats(dentry->d_inode, NFSIOS_VFSFSYNC);
L
Linus Torvalds 已提交
962 963 964
	return 0;
}

965 966 967 968 969 970 971 972 973 974 975 976
/**
 * 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)
{
977
	NFS_I(dir)->cache_change_attribute++;
978 979
}

L
Linus Torvalds 已提交
980 981 982 983 984
/*
 * 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.
 */
985
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
986 987 988
{
	if (IS_ROOT(dentry))
		return 1;
989 990
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
991 992 993 994 995 996 997 998
	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
999 1000
}

1001 1002 1003 1004 1005 1006 1007
/*
 * Return the intent data that applies to this particular path component
 *
 * Note that the current set of intents only apply to the very last
 * component of the path.
 * We check for this using LOOKUP_CONTINUE and LOOKUP_PARENT.
 */
1008 1009
static inline unsigned int nfs_lookup_check_intent(struct nameidata *nd,
						unsigned int mask)
1010 1011 1012 1013 1014 1015
{
	if (nd->flags & (LOOKUP_CONTINUE|LOOKUP_PARENT))
		return 0;
	return nd->flags & mask;
}

1016 1017 1018 1019 1020 1021 1022 1023
/*
 * 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;
1024
	return nd && nfs_lookup_check_intent(nd, LOOKUP_EXCL);
1025 1026
}

1027 1028 1029 1030 1031 1032 1033 1034
/*
 * 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
1035 1036 1037 1038 1039
static inline
int nfs_lookup_verify_inode(struct inode *inode, struct nameidata *nd)
{
	struct nfs_server *server = NFS_SERVER(inode);

1040
	if (IS_AUTOMOUNT(inode))
1041
		return 0;
L
Linus Torvalds 已提交
1042 1043
	if (nd != NULL) {
		/* VFS wants an on-the-wire revalidation */
1044
		if (nd->flags & LOOKUP_REVAL)
L
Linus Torvalds 已提交
1045 1046
			goto out_force;
		/* This is an open(2) */
1047
		if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 &&
1048 1049 1050
				!(server->flags & NFS_MOUNT_NOCTO) &&
				(S_ISREG(inode->i_mode) ||
				 S_ISDIR(inode->i_mode)))
L
Linus Torvalds 已提交
1051
			goto out_force;
1052
		return 0;
L
Linus Torvalds 已提交
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
	}
	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 */
1071
	if (nd != NULL && nfs_lookup_check_intent(nd, LOOKUP_CREATE) != 0)
L
Linus Torvalds 已提交
1072
		return 0;
1073 1074
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088
	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.
 */
1089
static int nfs_lookup_revalidate(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
1090 1091 1092 1093
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
1094 1095
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1096 1097
	int error;

1098 1099 1100
	if (nd->flags & LOOKUP_RCU)
		return -ECHILD;

L
Linus Torvalds 已提交
1101 1102
	parent = dget_parent(dentry);
	dir = parent->d_inode;
C
Chuck Lever 已提交
1103
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
1104 1105 1106 1107 1108 1109 1110 1111 1112
	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
1113
		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
1114
				__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1115
				dentry->d_name.name);
L
Linus Torvalds 已提交
1116 1117 1118
		goto out_bad;
	}

1119 1120 1121
	if (nfs_have_delegation(inode, FMODE_READ))
		goto out_set_verifier;

L
Linus Torvalds 已提交
1122
	/* Force a full look up iff the parent directory has changed */
1123
	if (!nfs_is_exclusive_create(dir, nd) && nfs_check_verifier(dir, dentry)) {
L
Linus Torvalds 已提交
1124 1125 1126 1127 1128 1129 1130 1131
		if (nfs_lookup_verify_inode(inode, nd))
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

1132 1133 1134 1135 1136 1137
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

1138
	error = NFS_PROTO(dir)->lookup(NFS_SERVER(dir)->client, dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1139 1140
	if (error)
		goto out_bad;
1141
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
1142
		goto out_bad;
1143
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
1144 1145
		goto out_bad;

1146 1147
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1148
out_set_verifier:
1149
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1150 1151
 out_valid:
	dput(parent);
C
Chuck Lever 已提交
1152
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n",
1153
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1154
			dentry->d_name.name);
L
Linus Torvalds 已提交
1155 1156 1157 1158
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
1159
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
1160 1161 1162 1163 1164 1165
	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;
1166 1167
		if (dentry->d_flags & DCACHE_DISCONNECTED)
			goto out_valid;
L
Linus Torvalds 已提交
1168 1169 1170
		shrink_dcache_parent(dentry);
	}
	d_drop(dentry);
1171 1172
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1173
	dput(parent);
C
Chuck Lever 已提交
1174
	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
1175
			__func__, dentry->d_parent->d_name.name,
C
Chuck Lever 已提交
1176
			dentry->d_name.name);
L
Linus Torvalds 已提交
1177
	return 0;
1178 1179 1180 1181 1182 1183 1184 1185
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
1186 1187 1188 1189 1190
}

/*
 * This is called from dput() when d_count is going to 0.
 */
N
Nick Piggin 已提交
1191
static int nfs_dentry_delete(const struct dentry *dentry)
L
Linus Torvalds 已提交
1192 1193 1194 1195 1196
{
	dfprintk(VFS, "NFS: dentry_delete(%s/%s, %x)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		dentry->d_flags);

1197 1198 1199 1200
	/* 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
1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213
	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;

}

1214 1215 1216 1217 1218 1219 1220 1221
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
1222 1223 1224 1225 1226 1227
/*
 * 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)
{
1228 1229 1230 1231
	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
1232
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
1233
		drop_nlink(inode);
1234
		nfs_complete_unlink(dentry, inode);
L
Linus Torvalds 已提交
1235 1236 1237 1238
	}
	iput(inode);
}

1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
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
1250
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
1251 1252 1253
	.d_revalidate	= nfs_lookup_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1254
	.d_automount	= nfs_d_automount,
1255
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1256 1257 1258 1259 1260
};

static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
{
	struct dentry *res;
T
Trond Myklebust 已提交
1261
	struct dentry *parent;
L
Linus Torvalds 已提交
1262
	struct inode *inode = NULL;
1263 1264
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
L
Linus Torvalds 已提交
1265 1266 1267 1268
	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
1269
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
1270 1271 1272 1273 1274

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

1275 1276 1277 1278 1279 1280 1281
	/*
	 * 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;
1282
		goto out;
1283
	}
L
Linus Torvalds 已提交
1284

1285 1286 1287 1288 1289 1290
	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
1291 1292 1293
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
	nfs_block_sillyrename(parent);
1294
	error = NFS_PROTO(dir)->lookup(NFS_SERVER(dir)->client, dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1295 1296 1297 1298
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
1299
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1300
	}
1301
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
N
Namhyung Kim 已提交
1302
	res = ERR_CAST(inode);
1303
	if (IS_ERR(res))
T
Trond Myklebust 已提交
1304
		goto out_unblock_sillyrename;
1305

L
Linus Torvalds 已提交
1306
no_entry:
1307
	res = d_materialise_unique(dentry, inode);
1308 1309
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
1310
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1311
		dentry = res;
1312
	}
L
Linus Torvalds 已提交
1313
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
1314 1315
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
L
Linus Torvalds 已提交
1316
out:
1317 1318
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1319 1320 1321 1322 1323 1324
	return res;
}

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

A
Al Viro 已提交
1325
const struct dentry_operations nfs4_dentry_operations = {
L
Linus Torvalds 已提交
1326 1327 1328
	.d_revalidate	= nfs_open_revalidate,
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1329
	.d_automount	= nfs_d_automount,
1330
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1331 1332
};

1333 1334 1335 1336
/*
 * 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
1337
static int is_atomic_open(struct nameidata *nd)
L
Linus Torvalds 已提交
1338
{
1339
	if (nd == NULL || nfs_lookup_check_intent(nd, LOOKUP_OPEN) == 0)
L
Linus Torvalds 已提交
1340 1341 1342 1343 1344
		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? */
1345 1346
	if (__mnt_is_readonly(nd->path.mnt) &&
	    (nd->intent.open.flags & (O_CREAT|O_TRUNC|FMODE_WRITE)))
L
Linus Torvalds 已提交
1347 1348 1349 1350
		return 0;
	return 1;
}

1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
static struct nfs_open_context *nameidata_to_nfs_open_context(struct dentry *dentry, struct nameidata *nd)
{
	struct path path = {
		.mnt = nd->path.mnt,
		.dentry = dentry,
	};
	struct nfs_open_context *ctx;
	struct rpc_cred *cred;
	fmode_t fmode = nd->intent.open.flags & (FMODE_READ | FMODE_WRITE | FMODE_EXEC);

	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return ERR_CAST(cred);
	ctx = alloc_nfs_open_context(&path, cred, fmode);
	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) {
		ret = nfs_may_open(ctx->path.dentry->d_inode,
				ctx->cred,
				nd->intent.open.flags);
		if (ret < 0)
			goto out;
	}
	filp = lookup_instantiate_filp(nd, ctx->path.dentry, do_open);
	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
1400 1401
static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
{
1402 1403
	struct nfs_open_context *ctx;
	struct iattr attr;
L
Linus Torvalds 已提交
1404
	struct dentry *res = NULL;
1405
	struct inode *inode;
1406
	int open_flags;
1407
	int err;
L
Linus Torvalds 已提交
1408

C
Chuck Lever 已提交
1409 1410 1411
	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
1412
	/* Check that we are indeed trying to open this file */
T
Trond Myklebust 已提交
1413
	if (!is_atomic_open(nd))
L
Linus Torvalds 已提交
1414 1415 1416 1417 1418 1419 1420
		goto no_open;

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

1421 1422
	/* Let vfs_create() deal with O_EXCL. Instantiate, but don't hash
	 * the dentry. */
1423
	if (nd->flags & LOOKUP_EXCL) {
1424
		d_instantiate(dentry, NULL);
1425 1426
		goto out;
	}
L
Linus Torvalds 已提交
1427

1428 1429 1430 1431 1432 1433 1434 1435 1436
	ctx = nameidata_to_nfs_open_context(dentry, nd);
	res = ERR_CAST(ctx);
	if (IS_ERR(ctx))
		goto out;

	open_flags = nd->intent.open.flags;
	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
1437
		attr.ia_mode &= ~current_umask();
1438
	} else {
1439
		open_flags &= ~(O_EXCL | O_CREAT);
1440 1441 1442
		attr.ia_valid = 0;
	}

L
Linus Torvalds 已提交
1443
	/* Open the file on the server */
1444
	nfs_block_sillyrename(dentry->d_parent);
1445
	inode = NFS_PROTO(dir)->open_context(dir, ctx, open_flags, &attr);
1446 1447
	if (IS_ERR(inode)) {
		nfs_unblock_sillyrename(dentry->d_parent);
1448
		put_nfs_open_context(ctx);
1449
		switch (PTR_ERR(inode)) {
L
Linus Torvalds 已提交
1450 1451
			/* Make a negative dentry */
			case -ENOENT:
1452
				d_add(dentry, NULL);
1453 1454
				res = NULL;
				goto out;
L
Linus Torvalds 已提交
1455
			/* This turned out not to be a regular file */
1456 1457
			case -ENOTDIR:
				goto no_open;
L
Linus Torvalds 已提交
1458 1459 1460
			case -ELOOP:
				if (!(nd->intent.open.flags & O_NOFOLLOW))
					goto no_open;
1461
			/* case -EISDIR: */
L
Linus Torvalds 已提交
1462 1463
			/* case -EINVAL: */
			default:
1464
				res = ERR_CAST(inode);
L
Linus Torvalds 已提交
1465 1466
				goto out;
		}
1467
	}
1468
	res = d_add_unique(dentry, inode);
1469
	nfs_unblock_sillyrename(dentry->d_parent);
1470 1471 1472
	if (res != NULL) {
		dput(ctx->path.dentry);
		ctx->path.dentry = dget(res);
L
Linus Torvalds 已提交
1473
		dentry = res;
1474
	}
1475 1476 1477 1478 1479 1480
	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
1481
out:
1482
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1483 1484 1485 1486 1487 1488 1489 1490
	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
1491
	struct inode *inode;
L
Linus Torvalds 已提交
1492
	struct inode *dir;
1493
	struct nfs_open_context *ctx;
L
Linus Torvalds 已提交
1494 1495
	int openflags, ret = 0;

N
Nick Piggin 已提交
1496 1497 1498 1499
	if (nd->flags & LOOKUP_RCU)
		return -ECHILD;

	inode = dentry->d_inode;
1500
	if (!is_atomic_open(nd) || d_mountpoint(dentry))
T
Trond Myklebust 已提交
1501
		goto no_open;
1502

L
Linus Torvalds 已提交
1503 1504
	parent = dget_parent(dentry);
	dir = parent->d_inode;
1505

L
Linus Torvalds 已提交
1506 1507 1508
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1509 1510 1511
	if (inode == NULL) {
		if (!nfs_neg_need_reval(dir, dentry, nd))
			ret = 1;
L
Linus Torvalds 已提交
1512
		goto out;
1513 1514
	}

L
Linus Torvalds 已提交
1515 1516
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
T
Trond Myklebust 已提交
1517
		goto no_open_dput;
L
Linus Torvalds 已提交
1518 1519 1520
	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
1521
		goto no_open_dput;
L
Linus Torvalds 已提交
1522
	/* We can't create new files, or truncate existing ones here */
1523
	openflags &= ~(O_CREAT|O_EXCL|O_TRUNC);
L
Linus Torvalds 已提交
1524

1525 1526 1527 1528
	ctx = nameidata_to_nfs_open_context(dentry, nd);
	ret = PTR_ERR(ctx);
	if (IS_ERR(ctx))
		goto out;
L
Linus Torvalds 已提交
1529
	/*
1530
	 * Note: we're not holding inode->i_mutex and so may be racing with
L
Linus Torvalds 已提交
1531 1532 1533
	 * operations that change the directory. We therefore save the
	 * change attribute *before* we do the RPC call.
	 */
1534
	inode = NFS_PROTO(dir)->open_context(dir, ctx, openflags, NULL);
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548
	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);
1549
	if (inode != dentry->d_inode)
1550
		goto out_drop;
1551 1552 1553 1554 1555

	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
1556 1557 1558
out:
	dput(parent);
	return ret;
1559 1560 1561 1562 1563 1564 1565
out_drop:
	d_drop(dentry);
	ret = 0;
out_put_ctx:
	put_nfs_open_context(ctx);
	goto out;

T
Trond Myklebust 已提交
1566
no_open_dput:
L
Linus Torvalds 已提交
1567
	dput(parent);
T
Trond Myklebust 已提交
1568
no_open:
L
Linus Torvalds 已提交
1569 1570
	return nfs_lookup_revalidate(dentry, nd);
}
1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591

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;
	int open_flags = 0;

	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;

	if ((nd->flags & LOOKUP_CREATE) != 0) {
		open_flags = nd->intent.open.flags;

		ctx = nameidata_to_nfs_open_context(dentry, nd);
		error = PTR_ERR(ctx);
		if (IS_ERR(ctx))
1592
			goto out_err_drop;
1593 1594 1595 1596 1597
	}

	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, ctx);
	if (error != 0)
		goto out_put_ctx;
1598 1599 1600 1601 1602
	if (ctx != NULL) {
		error = nfs_intent_set_file(nd, ctx);
		if (error < 0)
			goto out_err;
	}
1603 1604 1605 1606
	return 0;
out_put_ctx:
	if (ctx != NULL)
		put_nfs_open_context(ctx);
1607
out_err_drop:
1608
	d_drop(dentry);
1609
out_err:
1610 1611 1612
	return error;
}

L
Linus Torvalds 已提交
1613 1614 1615 1616 1617 1618 1619 1620
#endif /* CONFIG_NFSV4 */

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
				struct nfs_fattr *fattr)
{
1621 1622
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1623 1624 1625
	struct inode *inode;
	int error = -EACCES;

1626 1627
	d_drop(dentry);

L
Linus Torvalds 已提交
1628 1629
	/* We may have been initialized further down */
	if (dentry->d_inode)
1630
		goto out;
L
Linus Torvalds 已提交
1631
	if (fhandle->size == 0) {
1632
		error = NFS_PROTO(dir)->lookup(NFS_SERVER(dir)->client, dir, &dentry->d_name, fhandle, fattr);
L
Linus Torvalds 已提交
1633
		if (error)
1634
			goto out_error;
L
Linus Torvalds 已提交
1635
	}
1636
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1637 1638
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1639
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr);
L
Linus Torvalds 已提交
1640
		if (error < 0)
1641
			goto out_error;
L
Linus Torvalds 已提交
1642 1643
	}
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
1644 1645
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1646 1647 1648 1649
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1650
	return 0;
1651 1652 1653 1654
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
}

/*
 * 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;
1668
	int open_flags = 0;
L
Linus Torvalds 已提交
1669

C
Chuck Lever 已提交
1670 1671
	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
1672 1673 1674 1675

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

1676 1677 1678 1679
	if ((nd->flags & LOOKUP_CREATE) != 0)
		open_flags = nd->intent.open.flags;

	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, NULL);
L
Linus Torvalds 已提交
1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696
	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
1697 1698
	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
1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722

	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
1723 1724
	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
1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737

	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;
}

1738 1739 1740 1741 1742 1743
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
1744 1745 1746 1747
static int nfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	int error;

C
Chuck Lever 已提交
1748 1749
	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
1750 1751 1752 1753

	error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
	/* Ensure the VFS deletes this inode */
	if (error == 0 && dentry->d_inode != NULL)
1754
		clear_nlink(dentry->d_inode);
1755 1756
	else if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783

	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) {
1784
		nfs_inode_return_delegation(inode);
L
Linus Torvalds 已提交
1785 1786 1787
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		/* The VFS may want to delete this inode */
		if (error == 0)
1788
			nfs_drop_nlink(inode);
1789
		nfs_mark_for_revalidate(inode);
L
Linus Torvalds 已提交
1790 1791
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1792 1793
	if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
L
Linus Torvalds 已提交
1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811
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
1812
	if (dentry->d_count > 1) {
L
Linus Torvalds 已提交
1813
		spin_unlock(&dentry->d_lock);
T
Trond Myklebust 已提交
1814 1815
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1816 1817 1818 1819 1820 1821 1822 1823 1824
		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);
1825
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1826 1827 1828 1829 1830 1831
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
	return error;
}

1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847
/*
 * 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
1848
{
1849 1850 1851
	struct pagevec lru_pvec;
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1852
	struct iattr attr;
1853
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1854 1855 1856 1857 1858
	int error;

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

1859 1860
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1861

1862 1863
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1864

1865
	page = alloc_page(GFP_HIGHUSER);
1866
	if (!page)
1867 1868 1869 1870 1871 1872 1873 1874
		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);

1875
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1876 1877 1878 1879
	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
1880
		d_drop(dentry);
1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891
		__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)) {
1892
		pagevec_add(&lru_pvec, page);
1893
		pagevec_lru_add_file(&lru_pvec);
1894 1895 1896 1897 1898 1899
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
}

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);

1912 1913
	nfs_inode_return_delegation(inode);

1914
	d_drop(dentry);
L
Linus Torvalds 已提交
1915
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1916
	if (error == 0) {
A
Al Viro 已提交
1917
		ihold(inode);
1918
		d_add(dentry, inode);
1919
	}
L
Linus Torvalds 已提交
1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957
	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
1958
		 new_dentry->d_count);
L
Linus Torvalds 已提交
1959 1960

	/*
M
Miklos Szeredi 已提交
1961 1962 1963 1964
	 * 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
1965
	 */
1966 1967 1968 1969 1970 1971 1972 1973 1974
	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
1975

N
Nick Piggin 已提交
1976
		if (new_dentry->d_count > 2) {
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
			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);
1987
			if (err)
1988
				goto out;
1989 1990

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
1991
			rehash = NULL;
1992
			new_inode = NULL;
1993
		}
1994
	}
L
Linus Torvalds 已提交
1995

1996
	nfs_inode_return_delegation(old_inode);
1997
	if (new_inode != NULL)
1998
		nfs_inode_return_delegation(new_inode);
L
Linus Torvalds 已提交
1999 2000 2001

	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
2002
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
2003 2004 2005 2006
out:
	if (rehash)
		d_rehash(rehash);
	if (!error) {
2007 2008
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
2009
		d_move(old_dentry, new_dentry);
2010 2011
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
2012 2013
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
2014 2015 2016 2017 2018 2019 2020

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

2021 2022 2023 2024
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

2025 2026 2027 2028
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
2029 2030 2031
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
2032 2033
}

2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044
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);
	}
}

2045
int nfs_access_cache_shrinker(struct shrinker *shrink, int nr_to_scan, gfp_t gfp_mask)
2046 2047
{
	LIST_HEAD(head);
2048
	struct nfs_inode *nfsi, *next;
2049 2050
	struct nfs_access_entry *cache;

2051 2052
	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
		return (nr_to_scan == 0) ? 0 : -1;
2053

2054
	spin_lock(&nfs_access_lru_lock);
2055
	list_for_each_entry_safe(nfsi, next, &nfs_access_lru_list, access_cache_inode_lru) {
2056 2057 2058 2059
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
2060
		inode = &nfsi->vfs_inode;
2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073
		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);
2074
			smp_mb__before_clear_bit();
2075
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
2076
			smp_mb__after_clear_bit();
2077
		}
2078
		spin_unlock(&inode->i_lock);
2079 2080
	}
	spin_unlock(&nfs_access_lru_lock);
2081
	nfs_access_free_list(&head);
2082 2083 2084
	return (atomic_long_read(&nfs_access_nr_entries) / 100) * sysctl_vfs_cache_pressure;
}

2085
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
2086
{
2087
	struct rb_root *root_node = &nfsi->access_cache;
2088
	struct rb_node *n;
2089 2090 2091 2092 2093 2094
	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);
2095
		list_move(&entry->lru, head);
2096 2097
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
2098 2099
}

2100
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
2101
{
2102 2103 2104 2105
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
2106
	/* Remove from global LRU init */
2107 2108
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
2109 2110
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

2111
	spin_lock(&inode->i_lock);
2112 2113 2114 2115
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
2116
}
L
Linus Torvalds 已提交
2117

2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131
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
2132
	}
2133 2134 2135
	return NULL;
}

2136
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
2137 2138 2139 2140 2141
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

2142
	spin_lock(&inode->i_lock);
2143 2144 2145 2146 2147
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
2148
	if (!nfs_have_delegated_attributes(inode) &&
2149
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
2150 2151 2152 2153
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
2154
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
2155 2156 2157 2158 2159 2160
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
2161
	list_del(&cache->lru);
2162 2163 2164 2165
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
2166 2167
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
2168 2169 2170 2171 2172
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
2173 2174
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192
	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);
2193
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
2194
	spin_unlock(&inode->i_lock);
2195 2196 2197
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
2198 2199
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
2200 2201 2202 2203
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

2204
static void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
2205 2206 2207 2208 2209
{
	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
2210
	cache->jiffies = set->jiffies;
2211
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
2212
	cache->mask = set->mask;
2213 2214

	nfs_access_add_rbtree(inode, cache);
2215 2216 2217 2218 2219 2220 2221

	/* 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 */
2222
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
2223
		spin_lock(&nfs_access_lru_lock);
2224 2225 2226
		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);
2227 2228
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244
}

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);
2245 2246 2247 2248 2249 2250
	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
2251
		return status;
2252
	}
L
Linus Torvalds 已提交
2253 2254
	nfs_access_add_cache(inode, &cache);
out:
2255
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2256 2257 2258 2259
		return 0;
	return -EACCES;
}

2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

	if (openflags & FMODE_READ)
		mask |= MAY_READ;
	if (openflags & FMODE_WRITE)
		mask |= MAY_WRITE;
	if (openflags & FMODE_EXEC)
		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));
}

2278
int nfs_permission(struct inode *inode, int mask, unsigned int flags)
L
Linus Torvalds 已提交
2279 2280 2281 2282
{
	struct rpc_cred *cred;
	int res = 0;

2283 2284 2285
	if (flags & IPERM_FLAG_RCU)
		return -ECHILD;

C
Chuck Lever 已提交
2286 2287
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

2288
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2289 2290
		goto out;
	/* Is this sys_access() ? */
E
Eric Paris 已提交
2291
	if (mask & (MAY_ACCESS | MAY_CHDIR))
L
Linus Torvalds 已提交
2292 2293 2294 2295 2296 2297 2298 2299
		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)
2300 2301
					&& (mask & MAY_OPEN)
					&& !(mask & MAY_EXEC))
L
Linus Torvalds 已提交
2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316
				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;

2317
	cred = rpc_lookup_cred();
L
Linus Torvalds 已提交
2318 2319 2320 2321 2322 2323
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
out:
2324 2325 2326
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

C
Chuck Lever 已提交
2327 2328
	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
2329 2330 2331 2332
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
2333
		res = generic_permission(inode, mask, flags, NULL);
C
Chuck Lever 已提交
2334
	goto out;
L
Linus Torvalds 已提交
2335 2336 2337 2338 2339 2340 2341 2342
}

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