dir.c 58.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
/*
 *  linux/fs/nfs/dir.c
 *
 *  Copyright (C) 1992  Rick Sladkey
 *
 *  nfs directory handling functions
 *
 * 10 Apr 1996	Added silly rename for unlink	--okir
 * 28 Sep 1996	Improved directory cache --okir
 * 23 Aug 1997  Claus Heine claus@momo.math.rwth-aachen.de 
 *              Re-implemented silly rename for unlink, newly implemented
 *              silly rename for nfs_rename() following the suggestions
 *              of Olaf Kirch (okir) found in this file.
 *              Following Linus comments on my original hack, this version
 *              depends only on the dcache stuff and doesn't touch the inode
 *              layer (iput() and friends).
 *  6 Jun 1999	Cache readdir lookups in the page cache. -DaveM
 */

B
Bryan Schumaker 已提交
20
#include <linux/module.h>
L
Linus Torvalds 已提交
21 22 23 24 25 26 27 28 29 30 31 32
#include <linux/time.h>
#include <linux/errno.h>
#include <linux/stat.h>
#include <linux/fcntl.h>
#include <linux/string.h>
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/pagemap.h>
33
#include <linux/pagevec.h>
L
Linus Torvalds 已提交
34
#include <linux/namei.h>
35
#include <linux/mount.h>
36
#include <linux/swap.h>
A
Alexey Dobriyan 已提交
37
#include <linux/sched.h>
38
#include <linux/kmemleak.h>
39
#include <linux/xattr.h>
L
Linus Torvalds 已提交
40 41

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

46 47
#include "nfstrace.h"

L
Linus Torvalds 已提交
48 49 50
/* #define NFS_DEBUG_VERBOSE 1 */

static int nfs_opendir(struct inode *, struct file *);
51
static int nfs_closedir(struct inode *, struct file *);
A
Al Viro 已提交
52
static int nfs_readdir(struct file *, struct dir_context *);
53
static int nfs_fsync_dir(struct file *, loff_t, loff_t, int);
54
static loff_t nfs_llseek_dir(struct file *, loff_t, int);
55
static void nfs_readdir_clear_array(struct page*);
L
Linus Torvalds 已提交
56

57
const struct file_operations nfs_dir_operations = {
58
	.llseek		= nfs_llseek_dir,
L
Linus Torvalds 已提交
59
	.read		= generic_read_dir,
A
Al Viro 已提交
60
	.iterate	= nfs_readdir,
L
Linus Torvalds 已提交
61
	.open		= nfs_opendir,
62
	.release	= nfs_closedir,
L
Linus Torvalds 已提交
63 64 65
	.fsync		= nfs_fsync_dir,
};

66 67
const struct address_space_operations nfs_dir_aops = {
	.freepage = nfs_readdir_clear_array,
B
Bryan Schumaker 已提交
68 69
};

70
static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct inode *dir, struct rpc_cred *cred)
71 72 73 74
{
	struct nfs_open_dir_context *ctx;
	ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
	if (ctx != NULL) {
75
		ctx->duped = 0;
76
		ctx->attr_gencount = NFS_I(dir)->attr_gencount;
77
		ctx->dir_cookie = 0;
78
		ctx->dup_cookie = 0;
79
		ctx->cred = get_rpccred(cred);
80 81 82
		return ctx;
	}
	return  ERR_PTR(-ENOMEM);
83 84 85 86 87 88 89 90
}

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
91 92 93 94 95 96
/*
 * Open file
 */
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
97 98 99
	int res = 0;
	struct nfs_open_dir_context *ctx;
	struct rpc_cred *cred;
L
Linus Torvalds 已提交
100

101
	dfprintk(FILE, "NFS: open dir(%pD2)\n", filp);
102 103

	nfs_inc_stats(inode, NFSIOS_VFSOPEN);
C
Chuck Lever 已提交
104

105 106 107
	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return PTR_ERR(cred);
108
	ctx = alloc_nfs_open_dir_context(inode, cred);
109 110 111 112 113
	if (IS_ERR(ctx)) {
		res = PTR_ERR(ctx);
		goto out;
	}
	filp->private_data = ctx;
114 115 116 117 118 119 120
	if (filp->f_path.dentry == filp->f_path.mnt->mnt_root) {
		/* This is a mountpoint, so d_revalidate will never
		 * have been called, so we need to refresh the
		 * inode (for close-open consistency) ourselves.
		 */
		__nfs_revalidate_inode(NFS_SERVER(inode), inode);
	}
121 122
out:
	put_rpccred(cred);
L
Linus Torvalds 已提交
123 124 125
	return res;
}

126 127 128 129 130 131 132
static int
nfs_closedir(struct inode *inode, struct file *filp)
{
	put_nfs_open_dir_context(filp->private_data);
	return 0;
}

B
Bryan Schumaker 已提交
133 134 135 136
struct nfs_cache_array_entry {
	u64 cookie;
	u64 ino;
	struct qstr string;
137
	unsigned char d_type;
B
Bryan Schumaker 已提交
138 139 140
};

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

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

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

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

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

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

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

/*
 * the caller is responsible for freeing qstr.name
 * when called by nfs_readdir_add_to_array, the strings will be freed in
 * nfs_clear_readdir_array()
 */
static
207
int nfs_readdir_make_qstr(struct qstr *string, const char *name, unsigned int len)
B
Bryan Schumaker 已提交
208 209 210
{
	string->len = len;
	string->name = kmemdup(name, len, GFP_KERNEL);
211 212
	if (string->name == NULL)
		return -ENOMEM;
213 214 215 216 217
	/*
	 * Avoid a kmemleak false positive. The pointer to the name is stored
	 * in a page cache page which kmemleak does not scan.
	 */
	kmemleak_not_leak(string->name);
218 219
	string->hash = full_name_hash(name, len);
	return 0;
B
Bryan Schumaker 已提交
220 221 222 223 224 225
}

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

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

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

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

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

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

	if (diff < 0)
		goto out_eof;
	if (diff >= array->size) {
263
		if (array->eof_index >= 0)
B
Bryan Schumaker 已提交
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280
			goto out_eof;
		return -EAGAIN;
	}

	index = (unsigned int)diff;
	*desc->dir_cookie = array->array[index].cookie;
	desc->cache_entry_index = index;
	return 0;
out_eof:
	desc->eof = 1;
	return -EBADCOOKIE;
}

static
int nfs_readdir_search_for_cookie(struct nfs_cache_array *array, nfs_readdir_descriptor_t *desc)
{
	int i;
281
	loff_t new_pos;
B
Bryan Schumaker 已提交
282 283 284 285
	int status = -EAGAIN;

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

289
			new_pos = desc->current_index + i;
290 291 292 293
			if (ctx->attr_gencount != nfsi->attr_gencount
			    || (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))) {
				ctx->duped = 0;
				ctx->attr_gencount = nfsi->attr_gencount;
A
Al Viro 已提交
294
			} else if (new_pos < desc->ctx->pos) {
295 296 297
				if (ctx->duped > 0
				    && ctx->dup_cookie == *desc->dir_cookie) {
					if (printk_ratelimit()) {
298
						pr_notice("NFS: directory %pD2 contains a readdir loop."
299
								"Please contact your server vendor.  "
300
								"The file: %s has duplicate cookie %llu\n",
301
								desc->file,
302
								array->array[i].string.name,
303 304 305 306 307
								*desc->dir_cookie);
					}
					status = -ELOOP;
					goto out;
				}
308
				ctx->dup_cookie = *desc->dir_cookie;
309
				ctx->duped = -1;
310
			}
A
Al Viro 已提交
311
			desc->ctx->pos = new_pos;
B
Bryan Schumaker 已提交
312
			desc->cache_entry_index = i;
T
Trond Myklebust 已提交
313
			return 0;
B
Bryan Schumaker 已提交
314 315
		}
	}
T
Trond Myklebust 已提交
316
	if (array->eof_index >= 0) {
317
		status = -EBADCOOKIE;
318 319
		if (*desc->dir_cookie == array->last_cookie)
			desc->eof = 1;
320
	}
321
out:
B
Bryan Schumaker 已提交
322 323 324 325 326 327 328
	return status;
}

static
int nfs_readdir_search_array(nfs_readdir_descriptor_t *desc)
{
	struct nfs_cache_array *array;
T
Trond Myklebust 已提交
329
	int status;
B
Bryan Schumaker 已提交
330 331 332 333 334 335 336 337 338 339 340 341

	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
342
	if (status == -EAGAIN) {
343
		desc->last_cookie = array->last_cookie;
344
		desc->current_index += array->size;
T
Trond Myklebust 已提交
345 346
		desc->page_index++;
	}
B
Bryan Schumaker 已提交
347 348 349 350 351 352 353
	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
354
int nfs_readdir_xdr_filler(struct page **pages, nfs_readdir_descriptor_t *desc,
B
Bryan Schumaker 已提交
355
			struct nfs_entry *entry, struct file *file, struct inode *inode)
L
Linus Torvalds 已提交
356
{
357 358
	struct nfs_open_dir_context *ctx = file->private_data;
	struct rpc_cred	*cred = ctx->cred;
359
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
360 361 362 363
	int		error;

 again:
	timestamp = jiffies;
364
	gencount = nfs_inc_attr_generation_counter();
B
Bryan Schumaker 已提交
365
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, entry->cookie, pages,
L
Linus Torvalds 已提交
366 367 368 369 370
					  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
371
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
372 373 374 375 376
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
377
	desc->timestamp = timestamp;
378
	desc->gencount = gencount;
B
Bryan Schumaker 已提交
379 380
error:
	return error;
L
Linus Torvalds 已提交
381 382
}

383 384
static int xdr_decode(nfs_readdir_descriptor_t *desc,
		      struct nfs_entry *entry, struct xdr_stream *xdr)
L
Linus Torvalds 已提交
385
{
386
	int error;
L
Linus Torvalds 已提交
387

388 389 390
	error = desc->decode(xdr, entry, desc->plus);
	if (error)
		return error;
B
Bryan Schumaker 已提交
391 392 393
	entry->fattr->time_start = desc->timestamp;
	entry->fattr->gencount = desc->gencount;
	return 0;
L
Linus Torvalds 已提交
394 395
}

B
Bryan Schumaker 已提交
396 397 398 399 400
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
401
	if (nfs_compare_fh(entry->fh, NFS_FH(dentry->d_inode)) != 0)
B
Bryan Schumaker 已提交
402 403 404 405 406 407
		goto different;
	return 1;
different:
	return 0;
}

408
static
A
Al Viro 已提交
409
bool nfs_use_readdirplus(struct inode *dir, struct dir_context *ctx)
410 411 412 413 414
{
	if (!nfs_server_capable(dir, NFS_CAP_READDIRPLUS))
		return false;
	if (test_and_clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(dir)->flags))
		return true;
A
Al Viro 已提交
415
	if (ctx->pos == 0)
416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
		return true;
	return false;
}

/*
 * This function is called by the lookup code to request the use of
 * readdirplus to accelerate any future lookups in the same
 * directory.
 */
static
void nfs_advise_use_readdirplus(struct inode *dir)
{
	set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(dir)->flags);
}

B
Bryan Schumaker 已提交
431 432 433
static
void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
{
434
	struct qstr filename = QSTR_INIT(entry->name, entry->len);
435 436
	struct dentry *dentry;
	struct dentry *alias;
B
Bryan Schumaker 已提交
437 438
	struct inode *dir = parent->d_inode;
	struct inode *inode;
439
	int status;
B
Bryan Schumaker 已提交
440

441 442 443 444 445 446 447
	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
448

449
	dentry = d_lookup(parent, &filename);
B
Bryan Schumaker 已提交
450 451
	if (dentry != NULL) {
		if (nfs_same_file(dentry, entry)) {
452
			nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
453 454 455
			status = nfs_refresh_inode(dentry->d_inode, entry->fattr);
			if (!status)
				nfs_setsecurity(dentry->d_inode, entry->fattr, entry->label);
B
Bryan Schumaker 已提交
456 457
			goto out;
		} else {
458 459
			if (d_invalidate(dentry) != 0)
				goto out;
B
Bryan Schumaker 已提交
460 461 462 463 464
			dput(dentry);
		}
	}

	dentry = d_alloc(parent, &filename);
465 466 467
	if (dentry == NULL)
		return;

468
	inode = nfs_fhget(dentry->d_sb, entry->fh, entry->fattr, entry->label);
B
Bryan Schumaker 已提交
469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
	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
485 486
/* Perform conversion from xdr to cache array */
static
487
int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *entry,
488
				struct page **xdr_pages, struct page *page, unsigned int buflen)
L
Linus Torvalds 已提交
489
{
490
	struct xdr_stream stream;
491
	struct xdr_buf buf;
492
	struct page *scratch;
B
Bryan Schumaker 已提交
493
	struct nfs_cache_array *array;
494 495
	unsigned int count = 0;
	int status;
496

497 498 499
	scratch = alloc_page(GFP_KERNEL);
	if (scratch == NULL)
		return -ENOMEM;
500

501
	xdr_init_decode_pages(&stream, &buf, xdr_pages, buflen);
502
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
B
Bryan Schumaker 已提交
503 504 505

	do {
		status = xdr_decode(desc, entry, &stream);
506 507 508
		if (status != 0) {
			if (status == -EAGAIN)
				status = 0;
B
Bryan Schumaker 已提交
509
			break;
510
		}
B
Bryan Schumaker 已提交
511

512 513
		count++;

T
Trond Myklebust 已提交
514
		if (desc->plus != 0)
B
Bryan Schumaker 已提交
515
			nfs_prime_dcache(desc->file->f_path.dentry, entry);
516 517 518 519

		status = nfs_readdir_add_to_array(entry, page);
		if (status != 0)
			break;
B
Bryan Schumaker 已提交
520 521
	} while (!entry->eof);

T
Trond Myklebust 已提交
522
	if (count == 0 || (status == -EBADCOOKIE && entry->eof != 0)) {
B
Bryan Schumaker 已提交
523
		array = nfs_readdir_get_array(page);
524 525 526 527
		if (!IS_ERR(array)) {
			array->eof_index = array->size;
			status = 0;
			nfs_readdir_release_array(page);
528 529
		} else
			status = PTR_ERR(array);
L
Linus Torvalds 已提交
530
	}
531 532

	put_page(scratch);
533
	return status;
B
Bryan Schumaker 已提交
534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555
}

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
556
int nfs_readdir_large_page(struct page **pages, unsigned int npages)
B
Bryan Schumaker 已提交
557 558 559 560 561 562 563 564 565
{
	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;
	}
566
	return 0;
B
Bryan Schumaker 已提交
567 568 569

out_freepages:
	nfs_readdir_free_pagearray(pages, i);
570
	return -ENOMEM;
L
Linus Torvalds 已提交
571 572
}

B
Bryan Schumaker 已提交
573 574
static
int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode)
575
{
B
Bryan Schumaker 已提交
576 577
	struct page *pages[NFS_MAX_READDIR_PAGES];
	void *pages_ptr = NULL;
B
Bryan Schumaker 已提交
578 579 580
	struct nfs_entry entry;
	struct file	*file = desc->file;
	struct nfs_cache_array *array;
581
	int status = -ENOMEM;
B
Bryan Schumaker 已提交
582
	unsigned int array_size = ARRAY_SIZE(pages);
B
Bryan Schumaker 已提交
583 584

	entry.prev_cookie = 0;
585
	entry.cookie = desc->last_cookie;
B
Bryan Schumaker 已提交
586 587 588
	entry.eof = 0;
	entry.fh = nfs_alloc_fhandle();
	entry.fattr = nfs_alloc_fattr();
589
	entry.server = NFS_SERVER(inode);
B
Bryan Schumaker 已提交
590 591
	if (entry.fh == NULL || entry.fattr == NULL)
		goto out;
592

593 594 595 596 597 598
	entry.label = nfs4_label_alloc(NFS_SERVER(inode), GFP_NOWAIT);
	if (IS_ERR(entry.label)) {
		status = PTR_ERR(entry.label);
		goto out;
	}

B
Bryan Schumaker 已提交
599
	array = nfs_readdir_get_array(page);
600 601
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
602
		goto out_label_free;
603
	}
B
Bryan Schumaker 已提交
604 605
	memset(array, 0, sizeof(struct nfs_cache_array));
	array->eof_index = -1;
606

607 608
	status = nfs_readdir_large_page(pages, array_size);
	if (status < 0)
B
Bryan Schumaker 已提交
609 610
		goto out_release_array;
	do {
611
		unsigned int pglen;
B
Bryan Schumaker 已提交
612
		status = nfs_readdir_xdr_filler(pages, desc, &entry, file, inode);
613

B
Bryan Schumaker 已提交
614
		if (status < 0)
615
			break;
616
		pglen = status;
617
		status = nfs_readdir_page_filler(desc, &entry, pages, page, pglen);
618 619 620 621 622 623
		if (status < 0) {
			if (status == -ENOSPC)
				status = 0;
			break;
		}
	} while (array->eof_index < 0);
B
Bryan Schumaker 已提交
624

B
Bryan Schumaker 已提交
625
	nfs_readdir_free_large_page(pages_ptr, pages, array_size);
B
Bryan Schumaker 已提交
626 627
out_release_array:
	nfs_readdir_release_array(page);
628 629
out_label_free:
	nfs4_label_free(entry.label);
B
Bryan Schumaker 已提交
630 631 632
out:
	nfs_free_fattr(entry.fattr);
	nfs_free_fhandle(entry.fh);
633 634 635 636
	return status;
}

/*
B
Bryan Schumaker 已提交
637 638 639 640
 * 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
641
 */
B
Bryan Schumaker 已提交
642 643
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page)
L
Linus Torvalds 已提交
644
{
A
Al Viro 已提交
645
	struct inode	*inode = file_inode(desc->file);
646
	int ret;
L
Linus Torvalds 已提交
647

648 649
	ret = nfs_readdir_xdr_to_array(desc, page, inode);
	if (ret < 0)
B
Bryan Schumaker 已提交
650 651
		goto error;
	SetPageUptodate(page);
L
Linus Torvalds 已提交
652

B
Bryan Schumaker 已提交
653 654 655
	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
656
	}
B
Bryan Schumaker 已提交
657 658 659 660
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
661
	return ret;
B
Bryan Schumaker 已提交
662
}
L
Linus Torvalds 已提交
663

B
Bryan Schumaker 已提交
664 665 666
static
void cache_page_release(nfs_readdir_descriptor_t *desc)
{
667 668
	if (!desc->page->mapping)
		nfs_readdir_clear_array(desc->page);
B
Bryan Schumaker 已提交
669 670 671 672 673 674 675
	page_cache_release(desc->page);
	desc->page = NULL;
}

static
struct page *get_cache_page(nfs_readdir_descriptor_t *desc)
{
A
Al Viro 已提交
676
	return read_cache_page(file_inode(desc->file)->i_mapping,
B
Bryan Schumaker 已提交
677
			desc->page_index, (filler_t *)nfs_readdir_filler, desc);
L
Linus Torvalds 已提交
678 679 680
}

/*
B
Bryan Schumaker 已提交
681
 * Returns 0 if desc->dir_cookie was found on page desc->page_index
L
Linus Torvalds 已提交
682
 */
B
Bryan Schumaker 已提交
683 684 685 686 687 688 689 690 691 692
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
693 694
	if (res != 0)
		cache_page_release(desc);
B
Bryan Schumaker 已提交
695 696 697 698
	return res;
}

/* Search for desc->dir_cookie from the beginning of the page cache */
L
Linus Torvalds 已提交
699 700 701
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
702
	int res;
B
Bryan Schumaker 已提交
703

704
	if (desc->page_index == 0) {
705
		desc->current_index = 0;
706 707
		desc->last_cookie = 0;
	}
T
Trond Myklebust 已提交
708
	do {
B
Bryan Schumaker 已提交
709
		res = find_cache_page(desc);
T
Trond Myklebust 已提交
710
	} while (res == -EAGAIN);
L
Linus Torvalds 已提交
711 712 713 714 715 716 717
	return res;
}

/*
 * Once we've found the start of the dirent within a page: fill 'er up...
 */
static 
A
Al Viro 已提交
718
int nfs_do_filldir(nfs_readdir_descriptor_t *desc)
L
Linus Torvalds 已提交
719 720
{
	struct file	*file = desc->file;
B
Bryan Schumaker 已提交
721 722 723
	int i = 0;
	int res = 0;
	struct nfs_cache_array *array = NULL;
724 725
	struct nfs_open_dir_context *ctx = file->private_data;

B
Bryan Schumaker 已提交
726
	array = nfs_readdir_get_array(desc->page);
727 728 729 730
	if (IS_ERR(array)) {
		res = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
731 732

	for (i = desc->cache_entry_index; i < array->size; i++) {
733
		struct nfs_cache_array_entry *ent;
L
Linus Torvalds 已提交
734

735
		ent = &array->array[i];
A
Al Viro 已提交
736 737
		if (!dir_emit(desc->ctx, ent->string.name, ent->string.len,
		    nfs_compat_user_ino64(ent->ino), ent->d_type)) {
738
			desc->eof = 1;
L
Linus Torvalds 已提交
739
			break;
740
		}
A
Al Viro 已提交
741
		desc->ctx->pos++;
B
Bryan Schumaker 已提交
742 743 744 745
		if (i < (array->size-1))
			*desc->dir_cookie = array->array[i+1].cookie;
		else
			*desc->dir_cookie = array->last_cookie;
746 747
		if (ctx->duped != 0)
			ctx->duped = 1;
L
Linus Torvalds 已提交
748
	}
T
Trond Myklebust 已提交
749
	if (array->eof_index >= 0)
750
		desc->eof = 1;
B
Bryan Schumaker 已提交
751 752

	nfs_readdir_release_array(desc->page);
753
out:
B
Bryan Schumaker 已提交
754
	cache_page_release(desc);
C
Chuck Lever 已提交
755 756
	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
757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772
	return res;
}

/*
 * If we cannot find a cookie in our cache, we suspect that this is
 * because it points to a deleted file, so we ask the server to return
 * whatever it thinks is the next entry. We then feed this to filldir.
 * If all goes well, we should then be able to find our way round the
 * cache on the next call to readdir_search_pagecache();
 *
 * NOTE: we cannot add the anonymous page to the pagecache because
 *	 the data it contains might not be page aligned. Besides,
 *	 we should already have a complete representation of the
 *	 directory in the page cache by the time we get here.
 */
static inline
A
Al Viro 已提交
773
int uncached_readdir(nfs_readdir_descriptor_t *desc)
L
Linus Torvalds 已提交
774 775 776
{
	struct page	*page = NULL;
	int		status;
A
Al Viro 已提交
777
	struct inode *inode = file_inode(desc->file);
778
	struct nfs_open_dir_context *ctx = desc->file->private_data;
L
Linus Torvalds 已提交
779

C
Chuck Lever 已提交
780 781
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
782 783 784 785 786 787

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
B
Bryan Schumaker 已提交
788

789
	desc->page_index = 0;
790
	desc->last_cookie = *desc->dir_cookie;
791
	desc->page = page;
792
	ctx->duped = 0;
793

794 795
	status = nfs_readdir_xdr_to_array(desc, page, inode);
	if (status < 0)
L
Linus Torvalds 已提交
796 797
		goto out_release;

A
Al Viro 已提交
798
	status = nfs_do_filldir(desc);
L
Linus Torvalds 已提交
799 800

 out:
C
Chuck Lever 已提交
801
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
802
			__func__, status);
L
Linus Torvalds 已提交
803 804
	return status;
 out_release:
B
Bryan Schumaker 已提交
805
	cache_page_release(desc);
L
Linus Torvalds 已提交
806 807 808
	goto out;
}

809 810 811
/* 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
812
 */
A
Al Viro 已提交
813
static int nfs_readdir(struct file *file, struct dir_context *ctx)
L
Linus Torvalds 已提交
814
{
A
Al Viro 已提交
815
	struct dentry	*dentry = file->f_path.dentry;
L
Linus Torvalds 已提交
816 817 818
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
A
Al Viro 已提交
819
	struct nfs_open_dir_context *dir_ctx = file->private_data;
820
	int res = 0;
L
Linus Torvalds 已提交
821

822 823
	dfprintk(FILE, "NFS: readdir(%pD2) starting at cookie %llu\n",
			file, (long long)ctx->pos);
C
Chuck Lever 已提交
824 825
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
826
	/*
A
Al Viro 已提交
827
	 * ctx->pos points to the dirent entry number.
828
	 * *desc->dir_cookie has the cookie for the next entry. We have
829 830
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
831 832 833
	 */
	memset(desc, 0, sizeof(*desc));

A
Al Viro 已提交
834 835
	desc->file = file;
	desc->ctx = ctx;
836
	desc->dir_cookie = &dir_ctx->dir_cookie;
L
Linus Torvalds 已提交
837
	desc->decode = NFS_PROTO(inode)->decode_dirent;
A
Al Viro 已提交
838
	desc->plus = nfs_use_readdirplus(inode, ctx) ? 1 : 0;
L
Linus Torvalds 已提交
839

T
Trond Myklebust 已提交
840
	nfs_block_sillyrename(dentry);
841 842
	if (ctx->pos == 0 || nfs_attribute_cache_expired(inode))
		res = nfs_revalidate_mapping(inode, file->f_mapping);
T
Trond Myklebust 已提交
843 844 845
	if (res < 0)
		goto out;

T
Trond Myklebust 已提交
846
	do {
L
Linus Torvalds 已提交
847
		res = readdir_search_pagecache(desc);
848

L
Linus Torvalds 已提交
849
		if (res == -EBADCOOKIE) {
850
			res = 0;
L
Linus Torvalds 已提交
851
			/* This means either end of directory */
B
Bryan Schumaker 已提交
852
			if (*desc->dir_cookie && desc->eof == 0) {
L
Linus Torvalds 已提交
853
				/* Or that the server has 'lost' a cookie */
A
Al Viro 已提交
854
				res = uncached_readdir(desc);
855
				if (res == 0)
L
Linus Torvalds 已提交
856 857 858 859 860
					continue;
			}
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
861
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
862
			nfs_zap_caches(inode);
863
			desc->page_index = 0;
L
Linus Torvalds 已提交
864
			desc->plus = 0;
B
Bryan Schumaker 已提交
865
			desc->eof = 0;
L
Linus Torvalds 已提交
866 867 868 869 870
			continue;
		}
		if (res < 0)
			break;

A
Al Viro 已提交
871
		res = nfs_do_filldir(desc);
872
		if (res < 0)
L
Linus Torvalds 已提交
873
			break;
T
Trond Myklebust 已提交
874
	} while (!desc->eof);
T
Trond Myklebust 已提交
875
out:
T
Trond Myklebust 已提交
876
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
877 878
	if (res > 0)
		res = 0;
879
	dfprintk(FILE, "NFS: readdir(%pD2) returns %d\n", file, res);
C
Chuck Lever 已提交
880
	return res;
L
Linus Torvalds 已提交
881 882
}

883
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int whence)
884
{
885
	struct inode *inode = file_inode(filp);
886
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
887

888 889
	dfprintk(FILE, "NFS: llseek dir(%pD2, %lld, %d)\n",
			filp, offset, whence);
890 891

	mutex_lock(&inode->i_mutex);
892
	switch (whence) {
893 894 895 896 897 898 899 900 901 902 903
		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;
904
		dir_ctx->dir_cookie = 0;
905
		dir_ctx->duped = 0;
906 907
	}
out:
908
	mutex_unlock(&inode->i_mutex);
909 910 911
	return offset;
}

L
Linus Torvalds 已提交
912 913 914 915
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
916 917
static int nfs_fsync_dir(struct file *filp, loff_t start, loff_t end,
			 int datasync)
L
Linus Torvalds 已提交
918
{
919
	struct inode *inode = file_inode(filp);
920

921
	dfprintk(FILE, "NFS: fsync dir(%pD2) datasync %d\n", filp, datasync);
C
Chuck Lever 已提交
922

923
	mutex_lock(&inode->i_mutex);
924
	nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
925
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
926 927 928
	return 0;
}

929 930 931 932 933 934 935 936 937 938 939 940
/**
 * 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)
{
941
	NFS_I(dir)->cache_change_attribute++;
942
}
B
Bryan Schumaker 已提交
943
EXPORT_SYMBOL_GPL(nfs_force_lookup_revalidate);
944

L
Linus Torvalds 已提交
945 946 947 948 949
/*
 * 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.
 */
950
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
951 952 953
{
	if (IS_ROOT(dentry))
		return 1;
954 955
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
956 957 958 959 960 961 962 963
	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
964 965
}

966 967 968 969
/*
 * Use intent information to check whether or not we're going to do
 * an O_EXCL create using this path component.
 */
970
static int nfs_is_exclusive_create(struct inode *dir, unsigned int flags)
971 972 973
{
	if (NFS_PROTO(dir)->version == 2)
		return 0;
974
	return flags & LOOKUP_EXCL;
975 976
}

977 978 979 980 981 982 983 984
/*
 * 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.
 *
 */
985
static
986
int nfs_lookup_verify_inode(struct inode *inode, unsigned int flags)
L
Linus Torvalds 已提交
987 988
{
	struct nfs_server *server = NFS_SERVER(inode);
989
	int ret;
L
Linus Torvalds 已提交
990

991
	if (IS_AUTOMOUNT(inode))
992
		return 0;
993
	/* VFS wants an on-the-wire revalidation */
994
	if (flags & LOOKUP_REVAL)
995 996
		goto out_force;
	/* This is an open(2) */
997 998
	if ((flags & LOOKUP_OPEN) && !(server->flags & NFS_MOUNT_NOCTO) &&
	    (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
999
		goto out_force;
1000 1001
out:
	return (inode->i_nlink == 0) ? -ENOENT : 0;
L
Linus Torvalds 已提交
1002
out_force:
1003 1004 1005 1006
	ret = __nfs_revalidate_inode(server, inode);
	if (ret != 0)
		return ret;
	goto out;
L
Linus Torvalds 已提交
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
}

/*
 * 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,
1018
		       unsigned int flags)
L
Linus Torvalds 已提交
1019 1020
{
	/* Don't revalidate a negative dentry if we're creating a new file */
1021
	if (flags & LOOKUP_CREATE)
L
Linus Torvalds 已提交
1022
		return 0;
1023 1024
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
	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.
 */
1039
static int nfs_lookup_revalidate(struct dentry *dentry, unsigned int flags)
L
Linus Torvalds 已提交
1040 1041 1042 1043
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
1044 1045
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
1046
	struct nfs4_label *label = NULL;
L
Linus Torvalds 已提交
1047 1048
	int error;

1049
	if (flags & LOOKUP_RCU)
1050 1051
		return -ECHILD;

L
Linus Torvalds 已提交
1052 1053
	parent = dget_parent(dentry);
	dir = parent->d_inode;
C
Chuck Lever 已提交
1054
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
1055 1056 1057
	inode = dentry->d_inode;

	if (!inode) {
1058
		if (nfs_neg_need_reval(dir, dentry, flags))
L
Linus Torvalds 已提交
1059
			goto out_bad;
1060
		goto out_valid_noent;
L
Linus Torvalds 已提交
1061 1062 1063
	}

	if (is_bad_inode(inode)) {
1064 1065
		dfprintk(LOOKUPCACHE, "%s: %pd2 has dud inode\n",
				__func__, dentry);
L
Linus Torvalds 已提交
1066 1067 1068
		goto out_bad;
	}

1069
	if (NFS_PROTO(dir)->have_delegation(inode, FMODE_READ))
1070 1071
		goto out_set_verifier;

L
Linus Torvalds 已提交
1072
	/* Force a full look up iff the parent directory has changed */
1073 1074
	if (!nfs_is_exclusive_create(dir, flags) && nfs_check_verifier(dir, dentry)) {
		if (nfs_lookup_verify_inode(inode, flags))
L
Linus Torvalds 已提交
1075 1076 1077 1078 1079 1080 1081
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

1082 1083 1084 1085 1086 1087
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

1088 1089 1090 1091
	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_NOWAIT);
	if (IS_ERR(label))
		goto out_error;

1092
	trace_nfs_lookup_revalidate_enter(dir, dentry, flags);
1093
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, label);
1094
	trace_nfs_lookup_revalidate_exit(dir, dentry, flags, error);
L
Linus Torvalds 已提交
1095 1096
	if (error)
		goto out_bad;
1097
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
1098
		goto out_bad;
1099
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
1100 1101
		goto out_bad;

1102 1103
	nfs_setsecurity(inode, fattr, label);

1104 1105
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1106 1107
	nfs4_label_free(label);

1108
out_set_verifier:
1109
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1110
 out_valid:
1111 1112 1113
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);
 out_valid_noent:
L
Linus Torvalds 已提交
1114
	dput(parent);
1115 1116
	dfprintk(LOOKUPCACHE, "NFS: %s(%pd2) is valid\n",
			__func__, dentry);
L
Linus Torvalds 已提交
1117 1118 1119 1120
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
A
Al Viro 已提交
1121 1122
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1123
	nfs4_label_free(label);
1124
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
1125 1126 1127
	if (inode && S_ISDIR(inode->i_mode)) {
		/* Purge readdir caches. */
		nfs_zap_caches(inode);
1128 1129 1130 1131 1132 1133 1134
		/*
		 * We can't d_drop the root of a disconnected tree:
		 * its d_hash is on the s_anon list and d_drop() would hide
		 * it from shrink_dcache_for_unmount(), leading to busy
		 * inodes on unmount and further oopses.
		 */
		if (IS_ROOT(dentry))
1135
			goto out_valid;
L
Linus Torvalds 已提交
1136
	}
1137 1138 1139 1140
	/* If we have submounts, don't unhash ! */
	if (check_submounts_and_drop(dentry) != 0)
		goto out_valid;

L
Linus Torvalds 已提交
1141
	dput(parent);
1142 1143
	dfprintk(LOOKUPCACHE, "NFS: %s(%pd2) is invalid\n",
			__func__, dentry);
L
Linus Torvalds 已提交
1144
	return 0;
1145 1146 1147
out_error:
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1148
	nfs4_label_free(label);
1149
	dput(parent);
1150 1151
	dfprintk(LOOKUPCACHE, "NFS: %s(%pd2) lookup returned error %d\n",
			__func__, dentry, error);
1152
	return error;
L
Linus Torvalds 已提交
1153 1154
}

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

	/*
	 * I believe we can only get a negative dentry here in the case of a
	 * procfs-style symlink. Just assume it's correct for now, but we may
	 * eventually need to do something more here.
	 */
	if (!inode) {
1175 1176
		dfprintk(LOOKUPCACHE, "%s: %pd2 has negative inode\n",
				__func__, dentry);
1177 1178 1179 1180
		return 1;
	}

	if (is_bad_inode(inode)) {
1181 1182
		dfprintk(LOOKUPCACHE, "%s: %pd2 has dud inode\n",
				__func__, dentry);
1183 1184 1185 1186 1187 1188 1189 1190 1191
		return 0;
	}

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

L
Linus Torvalds 已提交
1192 1193 1194
/*
 * This is called from dput() when d_count is going to 0.
 */
N
Nick Piggin 已提交
1195
static int nfs_dentry_delete(const struct dentry *dentry)
L
Linus Torvalds 已提交
1196
{
1197 1198
	dfprintk(VFS, "NFS: dentry_delete(%pd2, %x)\n",
		dentry, dentry->d_flags);
L
Linus Torvalds 已提交
1199

1200 1201 1202 1203
	/* 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
1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
		/* Unhash it, so that ->d_iput() would be called */
		return 1;
	}
	if (!(dentry->d_sb->s_flags & MS_ACTIVE)) {
		/* Unhash it, so that ancestors of killed async unlink
		 * files will be cleaned up during umount */
		return 1;
	}
	return 0;

}

T
Trond Myklebust 已提交
1217
/* Ensure that we revalidate inode->i_nlink */
1218 1219 1220
static void nfs_drop_nlink(struct inode *inode)
{
	spin_lock(&inode->i_lock);
T
Trond Myklebust 已提交
1221 1222 1223 1224
	/* drop the inode if we're reasonably sure this is the last link */
	if (inode->i_nlink == 1)
		clear_nlink(inode);
	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
1225 1226 1227
	spin_unlock(&inode->i_lock);
}

L
Linus Torvalds 已提交
1228 1229 1230 1231 1232 1233
/*
 * 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)
{
1234 1235 1236 1237
	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
1238
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
1239
		nfs_complete_unlink(dentry, inode);
T
Trond Myklebust 已提交
1240
		nfs_drop_nlink(inode);
L
Linus Torvalds 已提交
1241 1242 1243 1244
	}
	iput(inode);
}

1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255
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
1256
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
1257
	.d_revalidate	= nfs_lookup_revalidate,
1258
	.d_weak_revalidate	= nfs_weak_revalidate,
L
Linus Torvalds 已提交
1259 1260
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1261
	.d_automount	= nfs_d_automount,
1262
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1263
};
B
Bryan Schumaker 已提交
1264
EXPORT_SYMBOL_GPL(nfs_dentry_operations);
L
Linus Torvalds 已提交
1265

1266
struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
L
Linus Torvalds 已提交
1267 1268
{
	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;
1273
	struct nfs4_label *label = NULL;
L
Linus Torvalds 已提交
1274 1275
	int error;

1276
	dfprintk(VFS, "NFS: lookup(%pd2)\n", dentry);
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
	/*
	 * If we're doing an exclusive create, optimize away the lookup
	 * but don't hash the dentry.
	 */
A
Al Viro 已提交
1287
	if (nfs_is_exclusive_create(dir, flags)) {
1288 1289
		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;

1299 1300 1301 1302
	label = nfs4_label_alloc(NFS_SERVER(dir), GFP_NOWAIT);
	if (IS_ERR(label))
		goto out;

T
Trond Myklebust 已提交
1303 1304
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
1305
	trace_nfs_lookup_enter(dir, dentry, flags);
T
Trond Myklebust 已提交
1306
	nfs_block_sillyrename(parent);
1307
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, label);
L
Linus Torvalds 已提交
1308 1309 1310 1311
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
1312
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1313
	}
1314
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr, label);
N
Namhyung Kim 已提交
1315
	res = ERR_CAST(inode);
1316
	if (IS_ERR(res))
T
Trond Myklebust 已提交
1317
		goto out_unblock_sillyrename;
1318

1319 1320 1321
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);

L
Linus Torvalds 已提交
1322
no_entry:
1323
	res = d_materialise_unique(dentry, inode);
1324 1325
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
1326
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1327
		dentry = res;
1328
	}
L
Linus Torvalds 已提交
1329
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
1330 1331
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
1332
	trace_nfs_lookup_exit(dir, dentry, flags, error);
1333
	nfs4_label_free(label);
L
Linus Torvalds 已提交
1334
out:
1335 1336
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1337 1338
	return res;
}
B
Bryan Schumaker 已提交
1339
EXPORT_SYMBOL_GPL(nfs_lookup);
L
Linus Torvalds 已提交
1340

B
Bryan Schumaker 已提交
1341
#if IS_ENABLED(CONFIG_NFS_V4)
1342
static int nfs4_lookup_revalidate(struct dentry *, unsigned int);
L
Linus Torvalds 已提交
1343

A
Al Viro 已提交
1344
const struct dentry_operations nfs4_dentry_operations = {
1345
	.d_revalidate	= nfs4_lookup_revalidate,
L
Linus Torvalds 已提交
1346 1347
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1348
	.d_automount	= nfs_d_automount,
1349
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1350
};
B
Bryan Schumaker 已提交
1351
EXPORT_SYMBOL_GPL(nfs4_dentry_operations);
L
Linus Torvalds 已提交
1352

1353 1354 1355 1356 1357 1358 1359 1360 1361 1362
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;
}

1363
static struct nfs_open_context *create_nfs_open_context(struct dentry *dentry, int open_flags)
1364
{
1365
	return alloc_nfs_open_context(dentry, flags_to_mode(open_flags));
1366 1367 1368 1369
}

static int do_open(struct inode *inode, struct file *filp)
{
1370
	nfs_fscache_open_file(inode, filp);
1371 1372 1373
	return 0;
}

A
Al Viro 已提交
1374 1375
static int nfs_finish_open(struct nfs_open_context *ctx,
			   struct dentry *dentry,
A
Al Viro 已提交
1376
			   struct file *file, unsigned open_flags,
A
Al Viro 已提交
1377
			   int *opened)
1378
{
1379 1380
	int err;

M
Miklos Szeredi 已提交
1381 1382 1383
	if ((open_flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
		*opened |= FILE_CREATED;

A
Al Viro 已提交
1384 1385
	err = finish_open(file, dentry, do_open, opened);
	if (err)
A
Al Viro 已提交
1386
		goto out;
A
Al Viro 已提交
1387
	nfs_file_set_open_context(file, ctx);
1388

1389
out:
A
Al Viro 已提交
1390
	return err;
1391 1392
}

1393 1394 1395
int nfs_atomic_open(struct inode *dir, struct dentry *dentry,
		    struct file *file, unsigned open_flags,
		    umode_t mode, int *opened)
L
Linus Torvalds 已提交
1396
{
1397
	struct nfs_open_context *ctx;
1398 1399
	struct dentry *res;
	struct iattr attr = { .ia_valid = ATTR_OPEN };
1400
	struct inode *inode;
1401
	unsigned int lookup_flags = 0;
1402
	int err;
L
Linus Torvalds 已提交
1403

1404 1405 1406
	/* Expect a negative dentry */
	BUG_ON(dentry->d_inode);

1407 1408
	dfprintk(VFS, "NFS: atomic_open(%s/%ld), %pd\n",
			dir->i_sb->s_id, dir->i_ino, dentry);
C
Chuck Lever 已提交
1409

1410 1411 1412 1413
	err = nfs_check_flags(open_flags);
	if (err)
		return err;

1414 1415 1416 1417 1418 1419 1420 1421
	/* NFS only supports OPEN on regular files */
	if ((open_flags & O_DIRECTORY)) {
		if (!d_unhashed(dentry)) {
			/*
			 * Hashed negative dentry with O_DIRECTORY: dentry was
			 * revalidated and is fine, no need to perform lookup
			 * again
			 */
A
Al Viro 已提交
1422
			return -ENOENT;
1423
		}
1424
		lookup_flags = LOOKUP_OPEN|LOOKUP_DIRECTORY;
L
Linus Torvalds 已提交
1425
		goto no_open;
1426
	}
L
Linus Torvalds 已提交
1427

1428
	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
A
Al Viro 已提交
1429
		return -ENAMETOOLONG;
1430

1431
	if (open_flags & O_CREAT) {
1432
		attr.ia_valid |= ATTR_MODE;
1433 1434
		attr.ia_mode = mode & ~current_umask();
	}
1435 1436 1437
	if (open_flags & O_TRUNC) {
		attr.ia_valid |= ATTR_SIZE;
		attr.ia_size = 0;
1438 1439
	}

1440 1441 1442
	ctx = create_nfs_open_context(dentry, open_flags);
	err = PTR_ERR(ctx);
	if (IS_ERR(ctx))
A
Al Viro 已提交
1443
		goto out;
1444

1445
	trace_nfs_atomic_open_enter(dir, ctx, open_flags);
1446
	nfs_block_sillyrename(dentry->d_parent);
1447
	inode = NFS_PROTO(dir)->open_context(dir, ctx, open_flags, &attr, opened);
1448
	nfs_unblock_sillyrename(dentry->d_parent);
1449
	if (IS_ERR(inode)) {
1450
		err = PTR_ERR(inode);
1451
		trace_nfs_atomic_open_exit(dir, ctx, open_flags, err);
1452
		put_nfs_open_context(ctx);
1453 1454
		switch (err) {
		case -ENOENT:
1455
			d_drop(dentry);
1456 1457 1458 1459 1460 1461 1462
			d_add(dentry, NULL);
			break;
		case -EISDIR:
		case -ENOTDIR:
			goto no_open;
		case -ELOOP:
			if (!(open_flags & O_NOFOLLOW))
1463
				goto no_open;
1464
			break;
L
Linus Torvalds 已提交
1465
			/* case -EINVAL: */
1466 1467
		default:
			break;
L
Linus Torvalds 已提交
1468
		}
A
Al Viro 已提交
1469
		goto out;
1470
	}
1471

1472
	err = nfs_finish_open(ctx, ctx->dentry, file, open_flags, opened);
1473
	trace_nfs_atomic_open_exit(dir, ctx, open_flags, err);
1474
	put_nfs_open_context(ctx);
A
Al Viro 已提交
1475 1476
out:
	return err;
1477

L
Linus Torvalds 已提交
1478
no_open:
1479
	res = nfs_lookup(dir, dentry, lookup_flags);
1480 1481
	err = PTR_ERR(res);
	if (IS_ERR(res))
A
Al Viro 已提交
1482
		goto out;
1483

A
Al Viro 已提交
1484
	return finish_no_open(file, res);
L
Linus Torvalds 已提交
1485
}
B
Bryan Schumaker 已提交
1486
EXPORT_SYMBOL_GPL(nfs_atomic_open);
L
Linus Torvalds 已提交
1487

1488
static int nfs4_lookup_revalidate(struct dentry *dentry, unsigned int flags)
L
Linus Torvalds 已提交
1489 1490
{
	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
	int ret = 0;
L
Linus Torvalds 已提交
1494

1495
	if (flags & LOOKUP_RCU)
N
Nick Piggin 已提交
1496 1497
		return -ECHILD;

1498
	if (!(flags & LOOKUP_OPEN) || (flags & LOOKUP_DIRECTORY))
1499 1500
		goto no_open;
	if (d_mountpoint(dentry))
T
Trond Myklebust 已提交
1501
		goto no_open;
1502 1503
	if (NFS_SB(dentry->d_sb)->caps & NFS_CAP_ATOMIC_OPEN_V1)
		goto no_open;
1504

1505
	inode = dentry->d_inode;
L
Linus Torvalds 已提交
1506 1507
	parent = dget_parent(dentry);
	dir = parent->d_inode;
1508

L
Linus Torvalds 已提交
1509 1510 1511
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1512
	if (inode == NULL) {
1513
		if (!nfs_neg_need_reval(dir, dentry, flags))
1514
			ret = 1;
L
Linus Torvalds 已提交
1515
		goto out;
1516 1517
	}

L
Linus Torvalds 已提交
1518 1519
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
T
Trond Myklebust 已提交
1520
		goto no_open_dput;
L
Linus Torvalds 已提交
1521
	/* We cannot do exclusive creation on a positive dentry */
1522
	if (flags & LOOKUP_EXCL)
T
Trond Myklebust 已提交
1523
		goto no_open_dput;
L
Linus Torvalds 已提交
1524

1525 1526
	/* Let f_op->open() actually open (and revalidate) the file */
	ret = 1;
1527

L
Linus Torvalds 已提交
1528 1529 1530
out:
	dput(parent);
	return ret;
1531

T
Trond Myklebust 已提交
1532
no_open_dput:
L
Linus Torvalds 已提交
1533
	dput(parent);
T
Trond Myklebust 已提交
1534
no_open:
1535
	return nfs_lookup_revalidate(dentry, flags);
1536 1537
}

L
Linus Torvalds 已提交
1538 1539 1540 1541 1542 1543
#endif /* CONFIG_NFSV4 */

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
1544 1545
				struct nfs_fattr *fattr,
				struct nfs4_label *label)
L
Linus Torvalds 已提交
1546
{
1547 1548
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1549 1550 1551
	struct inode *inode;
	int error = -EACCES;

1552 1553
	d_drop(dentry);

L
Linus Torvalds 已提交
1554 1555
	/* We may have been initialized further down */
	if (dentry->d_inode)
1556
		goto out;
L
Linus Torvalds 已提交
1557
	if (fhandle->size == 0) {
1558
		error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, NULL);
L
Linus Torvalds 已提交
1559
		if (error)
1560
			goto out_error;
L
Linus Torvalds 已提交
1561
	}
1562
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1563 1564
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1565
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr, NULL);
L
Linus Torvalds 已提交
1566
		if (error < 0)
1567
			goto out_error;
L
Linus Torvalds 已提交
1568
	}
1569
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr, label);
1570 1571
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1572 1573 1574 1575
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1576
	return 0;
1577 1578 1579 1580
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1581
}
B
Bryan Schumaker 已提交
1582
EXPORT_SYMBOL_GPL(nfs_instantiate);
L
Linus Torvalds 已提交
1583 1584 1585 1586 1587 1588 1589

/*
 * 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.
 */
1590
int nfs_create(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
1591
		umode_t mode, bool excl)
L
Linus Torvalds 已提交
1592 1593
{
	struct iattr attr;
A
Al Viro 已提交
1594
	int open_flags = excl ? O_CREAT | O_EXCL : O_CREAT;
L
Linus Torvalds 已提交
1595 1596
	int error;

1597 1598
	dfprintk(VFS, "NFS: create(%s/%ld), %pd\n",
			dir->i_sb->s_id, dir->i_ino, dentry);
L
Linus Torvalds 已提交
1599 1600 1601 1602

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

1603
	trace_nfs_create_enter(dir, dentry, open_flags);
1604
	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags);
1605
	trace_nfs_create_exit(dir, dentry, open_flags, error);
L
Linus Torvalds 已提交
1606 1607 1608 1609 1610 1611 1612
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1613
EXPORT_SYMBOL_GPL(nfs_create);
L
Linus Torvalds 已提交
1614 1615 1616 1617

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1618
int
A
Al Viro 已提交
1619
nfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t rdev)
L
Linus Torvalds 已提交
1620 1621 1622 1623
{
	struct iattr attr;
	int status;

1624 1625
	dfprintk(VFS, "NFS: mknod(%s/%ld), %pd\n",
			dir->i_sb->s_id, dir->i_ino, dentry);
L
Linus Torvalds 已提交
1626 1627 1628 1629 1630 1631 1632

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

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

1633
	trace_nfs_mknod_enter(dir, dentry);
L
Linus Torvalds 已提交
1634
	status = NFS_PROTO(dir)->mknod(dir, dentry, &attr, rdev);
1635
	trace_nfs_mknod_exit(dir, dentry, status);
L
Linus Torvalds 已提交
1636 1637 1638 1639 1640 1641 1642
	if (status != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return status;
}
B
Bryan Schumaker 已提交
1643
EXPORT_SYMBOL_GPL(nfs_mknod);
L
Linus Torvalds 已提交
1644 1645 1646 1647

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1648
int nfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
1649 1650 1651 1652
{
	struct iattr attr;
	int error;

1653 1654
	dfprintk(VFS, "NFS: mkdir(%s/%ld), %pd\n",
			dir->i_sb->s_id, dir->i_ino, dentry);
L
Linus Torvalds 已提交
1655 1656 1657 1658

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

1659
	trace_nfs_mkdir_enter(dir, dentry);
L
Linus Torvalds 已提交
1660
	error = NFS_PROTO(dir)->mkdir(dir, dentry, &attr);
1661
	trace_nfs_mkdir_exit(dir, dentry, error);
L
Linus Torvalds 已提交
1662 1663 1664 1665 1666 1667 1668
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1669
EXPORT_SYMBOL_GPL(nfs_mkdir);
L
Linus Torvalds 已提交
1670

1671 1672 1673 1674 1675 1676
static void nfs_dentry_handle_enoent(struct dentry *dentry)
{
	if (dentry->d_inode != NULL && !d_unhashed(dentry))
		d_delete(dentry);
}

1677
int nfs_rmdir(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1678 1679 1680
{
	int error;

1681 1682
	dfprintk(VFS, "NFS: rmdir(%s/%ld), %pd\n",
			dir->i_sb->s_id, dir->i_ino, dentry);
L
Linus Torvalds 已提交
1683

1684
	trace_nfs_rmdir_enter(dir, dentry);
1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
	if (dentry->d_inode) {
		nfs_wait_on_sillyrename(dentry);
		error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
		/* Ensure the VFS deletes this inode */
		switch (error) {
		case 0:
			clear_nlink(dentry->d_inode);
			break;
		case -ENOENT:
			nfs_dentry_handle_enoent(dentry);
		}
	} else
		error = NFS_PROTO(dir)->rmdir(dir, &dentry->d_name);
1698
	trace_nfs_rmdir_exit(dir, dentry, error);
L
Linus Torvalds 已提交
1699 1700 1701

	return error;
}
B
Bryan Schumaker 已提交
1702
EXPORT_SYMBOL_GPL(nfs_rmdir);
L
Linus Torvalds 已提交
1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716

/*
 * 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;
		
1717
	dfprintk(VFS, "NFS: safe_remove(%pd2)\n", dentry);
L
Linus Torvalds 已提交
1718 1719 1720 1721 1722 1723 1724

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

1725
	trace_nfs_remove_enter(dir, dentry);
L
Linus Torvalds 已提交
1726
	if (inode != NULL) {
1727
		NFS_PROTO(inode)->return_delegation(inode);
L
Linus Torvalds 已提交
1728 1729
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		if (error == 0)
1730
			nfs_drop_nlink(inode);
L
Linus Torvalds 已提交
1731 1732
	} else
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
1733 1734
	if (error == -ENOENT)
		nfs_dentry_handle_enoent(dentry);
1735
	trace_nfs_remove_exit(dir, dentry, error);
L
Linus Torvalds 已提交
1736 1737 1738 1739 1740 1741 1742 1743 1744
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.
 */
1745
int nfs_unlink(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1746 1747 1748 1749
{
	int error;
	int need_rehash = 0;

1750 1751
	dfprintk(VFS, "NFS: unlink(%s/%ld, %pd)\n", dir->i_sb->s_id,
		dir->i_ino, dentry);
L
Linus Torvalds 已提交
1752

1753
	trace_nfs_unlink_enter(dir, dentry);
L
Linus Torvalds 已提交
1754
	spin_lock(&dentry->d_lock);
A
Al Viro 已提交
1755
	if (d_count(dentry) > 1) {
L
Linus Torvalds 已提交
1756
		spin_unlock(&dentry->d_lock);
T
Trond Myklebust 已提交
1757 1758
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1759
		error = nfs_sillyrename(dir, dentry);
1760
		goto out;
L
Linus Torvalds 已提交
1761 1762 1763 1764 1765 1766 1767
	}
	if (!d_unhashed(dentry)) {
		__d_drop(dentry);
		need_rehash = 1;
	}
	spin_unlock(&dentry->d_lock);
	error = nfs_safe_remove(dentry);
1768
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1769 1770 1771
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
1772 1773
out:
	trace_nfs_unlink_exit(dir, dentry, error);
L
Linus Torvalds 已提交
1774 1775
	return error;
}
B
Bryan Schumaker 已提交
1776
EXPORT_SYMBOL_GPL(nfs_unlink);
L
Linus Torvalds 已提交
1777

1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
/*
 * 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.
 */
1793
int nfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
L
Linus Torvalds 已提交
1794
{
1795 1796
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1797
	struct iattr attr;
1798
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1799 1800
	int error;

1801 1802
	dfprintk(VFS, "NFS: symlink(%s/%ld, %pd, %s)\n", dir->i_sb->s_id,
		dir->i_ino, dentry, symname);
L
Linus Torvalds 已提交
1803

1804 1805
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1806

1807 1808
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1809

1810
	page = alloc_page(GFP_HIGHUSER);
1811
	if (!page)
1812 1813
		return -ENOMEM;

1814
	kaddr = kmap_atomic(page);
1815 1816 1817
	memcpy(kaddr, symname, pathlen);
	if (pathlen < PAGE_SIZE)
		memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen);
1818
	kunmap_atomic(kaddr);
1819

1820
	trace_nfs_symlink_enter(dir, dentry);
1821
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1822
	trace_nfs_symlink_exit(dir, dentry, error);
1823
	if (error != 0) {
1824
		dfprintk(VFS, "NFS: symlink(%s/%ld, %pd, %s) error %d\n",
1825
			dir->i_sb->s_id, dir->i_ino,
1826
			dentry, symname, error);
L
Linus Torvalds 已提交
1827
		d_drop(dentry);
1828 1829 1830 1831 1832 1833 1834 1835
		__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.
	 */
1836
	if (!add_to_page_cache_lru(page, dentry->d_inode->i_mapping, 0,
1837 1838 1839 1840 1841 1842 1843
							GFP_KERNEL)) {
		SetPageUptodate(page);
		unlock_page(page);
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1844
}
B
Bryan Schumaker 已提交
1845
EXPORT_SYMBOL_GPL(nfs_symlink);
L
Linus Torvalds 已提交
1846

1847
int
L
Linus Torvalds 已提交
1848 1849 1850 1851 1852
nfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = old_dentry->d_inode;
	int error;

1853 1854
	dfprintk(VFS, "NFS: link(%pd2 -> %pd2)\n",
		old_dentry, dentry);
L
Linus Torvalds 已提交
1855

1856
	trace_nfs_link_enter(inode, dir, dentry);
1857
	NFS_PROTO(inode)->return_delegation(inode);
1858

1859
	d_drop(dentry);
L
Linus Torvalds 已提交
1860
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1861
	if (error == 0) {
A
Al Viro 已提交
1862
		ihold(inode);
1863
		d_add(dentry, inode);
1864
	}
1865
	trace_nfs_link_exit(inode, dir, dentry, error);
L
Linus Torvalds 已提交
1866 1867
	return error;
}
B
Bryan Schumaker 已提交
1868
EXPORT_SYMBOL_GPL(nfs_link);
L
Linus Torvalds 已提交
1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893

/*
 * 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.
 */
1894
int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
L
Linus Torvalds 已提交
1895 1896 1897 1898 1899 1900 1901
		      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;

1902 1903
	dfprintk(VFS, "NFS: rename(%pd2 -> %pd2, ct=%d)\n",
		 old_dentry, new_dentry,
A
Al Viro 已提交
1904
		 d_count(new_dentry));
L
Linus Torvalds 已提交
1905

1906
	trace_nfs_rename_enter(old_dir, old_dentry, new_dir, new_dentry);
L
Linus Torvalds 已提交
1907
	/*
M
Miklos Szeredi 已提交
1908 1909 1910 1911
	 * 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
1912
	 */
1913 1914 1915 1916 1917 1918 1919 1920 1921
	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
1922

A
Al Viro 已提交
1923
		if (d_count(new_dentry) > 2) {
1924 1925 1926 1927 1928 1929 1930 1931 1932 1933
			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);
1934
			if (err)
1935
				goto out;
1936 1937

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
1938
			rehash = NULL;
1939
			new_inode = NULL;
1940
		}
1941
	}
L
Linus Torvalds 已提交
1942

1943
	NFS_PROTO(old_inode)->return_delegation(old_inode);
1944
	if (new_inode != NULL)
1945
		NFS_PROTO(new_inode)->return_delegation(new_inode);
L
Linus Torvalds 已提交
1946 1947 1948

	error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
					   new_dir, &new_dentry->d_name);
1949
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
1950 1951 1952
out:
	if (rehash)
		d_rehash(rehash);
1953 1954
	trace_nfs_rename_exit(old_dir, old_dentry,
			new_dir, new_dentry, error);
L
Linus Torvalds 已提交
1955
	if (!error) {
1956 1957
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
1958
		d_move(old_dentry, new_dentry);
1959 1960
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
1961 1962
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
1963 1964 1965 1966 1967 1968

	/* new dentry created? */
	if (dentry)
		dput(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1969
EXPORT_SYMBOL_GPL(nfs_rename);
L
Linus Torvalds 已提交
1970

1971 1972 1973 1974
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

1975 1976 1977 1978
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
1979 1980 1981
	smp_mb__before_atomic_dec();
	atomic_long_dec(&nfs_access_nr_entries);
	smp_mb__after_atomic_dec();
1982 1983
}

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
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);
	}
}

1995 1996
unsigned long
nfs_access_cache_scan(struct shrinker *shrink, struct shrink_control *sc)
1997 1998
{
	LIST_HEAD(head);
1999
	struct nfs_inode *nfsi, *next;
2000
	struct nfs_access_entry *cache;
2001 2002
	int nr_to_scan = sc->nr_to_scan;
	gfp_t gfp_mask = sc->gfp_mask;
2003
	long freed = 0;
2004

2005
	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
2006
		return SHRINK_STOP;
2007

2008
	spin_lock(&nfs_access_lru_lock);
2009
	list_for_each_entry_safe(nfsi, next, &nfs_access_lru_list, access_cache_inode_lru) {
2010 2011 2012 2013
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
2014
		inode = &nfsi->vfs_inode;
2015 2016 2017 2018 2019 2020 2021
		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);
2022
		freed++;
2023 2024 2025 2026 2027 2028
		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);
2029
			smp_mb__before_clear_bit();
2030
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
2031
			smp_mb__after_clear_bit();
2032
		}
2033
		spin_unlock(&inode->i_lock);
2034 2035
	}
	spin_unlock(&nfs_access_lru_lock);
2036
	nfs_access_free_list(&head);
2037 2038 2039 2040 2041 2042
	return freed;
}

unsigned long
nfs_access_cache_count(struct shrinker *shrink, struct shrink_control *sc)
{
2043
	return vfs_pressure_ratio(atomic_long_read(&nfs_access_nr_entries));
2044 2045
}

2046
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
2047
{
2048
	struct rb_root *root_node = &nfsi->access_cache;
2049
	struct rb_node *n;
2050 2051 2052 2053 2054 2055
	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);
2056
		list_move(&entry->lru, head);
2057 2058
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
2059 2060
}

2061
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
2062
{
2063 2064 2065 2066
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
2067
	/* Remove from global LRU init */
2068 2069
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
2070 2071
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

2072
	spin_lock(&inode->i_lock);
2073 2074 2075 2076
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
2077
}
B
Bryan Schumaker 已提交
2078
EXPORT_SYMBOL_GPL(nfs_access_zap_cache);
L
Linus Torvalds 已提交
2079

2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093
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
2094
	}
2095 2096 2097
	return NULL;
}

2098
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
2099 2100 2101 2102 2103
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

2104
	spin_lock(&inode->i_lock);
2105 2106 2107 2108 2109
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
2110
	if (!nfs_have_delegated_attributes(inode) &&
2111
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
2112 2113 2114 2115
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
2116
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
2117 2118 2119 2120 2121 2122
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
2123
	list_del(&cache->lru);
2124 2125 2126 2127
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
2128 2129
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
2130 2131 2132 2133 2134
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
2135 2136
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154
	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);
2155
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
2156
	spin_unlock(&inode->i_lock);
2157 2158 2159
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
2160 2161
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
2162 2163 2164 2165
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

2166
void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
2167 2168 2169 2170 2171
{
	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
2172
	cache->jiffies = set->jiffies;
2173
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
2174
	cache->mask = set->mask;
2175 2176

	nfs_access_add_rbtree(inode, cache);
2177 2178 2179 2180 2181 2182 2183

	/* 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 */
2184
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
2185
		spin_lock(&nfs_access_lru_lock);
2186 2187 2188
		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);
2189 2190
		spin_unlock(&nfs_access_lru_lock);
	}
L
Linus Torvalds 已提交
2191
}
2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205
EXPORT_SYMBOL_GPL(nfs_access_add_cache);

void nfs_access_set_mask(struct nfs_access_entry *entry, u32 access_result)
{
	entry->mask = 0;
	if (access_result & NFS4_ACCESS_READ)
		entry->mask |= MAY_READ;
	if (access_result &
	    (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
		entry->mask |= MAY_WRITE;
	if (access_result & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
		entry->mask |= MAY_EXEC;
}
EXPORT_SYMBOL_GPL(nfs_access_set_mask);
L
Linus Torvalds 已提交
2206 2207 2208 2209 2210 2211

static int nfs_do_access(struct inode *inode, struct rpc_cred *cred, int mask)
{
	struct nfs_access_entry cache;
	int status;

2212 2213
	trace_nfs_access_enter(inode);

L
Linus Torvalds 已提交
2214 2215
	status = nfs_access_get_cached(inode, cred, &cache);
	if (status == 0)
2216
		goto out_cached;
L
Linus Torvalds 已提交
2217 2218 2219 2220 2221 2222

	/* 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);
2223 2224 2225 2226 2227 2228
	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);
		}
2229
		goto out;
2230
	}
L
Linus Torvalds 已提交
2231
	nfs_access_add_cache(inode, &cache);
2232 2233 2234
out_cached:
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) != 0)
		status = -EACCES;
L
Linus Torvalds 已提交
2235
out:
2236 2237
	trace_nfs_access_exit(inode, status);
	return status;
L
Linus Torvalds 已提交
2238 2239
}

2240 2241 2242 2243
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

2244 2245 2246 2247 2248 2249 2250 2251 2252 2253
	if (openflags & __FMODE_EXEC) {
		/* ONLY check exec rights */
		mask = MAY_EXEC;
	} else {
		if ((openflags & O_ACCMODE) != O_WRONLY)
			mask |= MAY_READ;
		if ((openflags & O_ACCMODE) != O_RDONLY)
			mask |= MAY_WRITE;
	}

2254 2255 2256 2257 2258 2259 2260
	return mask;
}

int nfs_may_open(struct inode *inode, struct rpc_cred *cred, int openflags)
{
	return nfs_do_access(inode, cred, nfs_open_permission_mask(openflags));
}
B
Bryan Schumaker 已提交
2261
EXPORT_SYMBOL_GPL(nfs_may_open);
2262

2263
int nfs_permission(struct inode *inode, int mask)
L
Linus Torvalds 已提交
2264 2265 2266 2267
{
	struct rpc_cred *cred;
	int res = 0;

2268
	if (mask & MAY_NOT_BLOCK)
2269 2270
		return -ECHILD;

C
Chuck Lever 已提交
2271 2272
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

2273
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2274 2275
		goto out;
	/* Is this sys_access() ? */
E
Eric Paris 已提交
2276
	if (mask & (MAY_ACCESS | MAY_CHDIR))
L
Linus Torvalds 已提交
2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296
		goto force_lookup;

	switch (inode->i_mode & S_IFMT) {
		case S_IFLNK:
			goto out;
		case S_IFREG:
			break;
		case S_IFDIR:
			/*
			 * Optimize away all write operations, since the server
			 * will check permissions when we perform the op.
			 */
			if ((mask & MAY_WRITE) && !(mask & MAY_READ))
				goto out;
	}

force_lookup:
	if (!NFS_PROTO(inode)->access)
		goto out_notsup;

2297
	cred = rpc_lookup_cred();
L
Linus Torvalds 已提交
2298 2299 2300 2301 2302 2303
	if (!IS_ERR(cred)) {
		res = nfs_do_access(inode, cred, mask);
		put_rpccred(cred);
	} else
		res = PTR_ERR(cred);
out:
2304 2305 2306
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

C
Chuck Lever 已提交
2307 2308
	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
2309 2310 2311 2312
	return res;
out_notsup:
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
2313
		res = generic_permission(inode, mask);
C
Chuck Lever 已提交
2314
	goto out;
L
Linus Torvalds 已提交
2315
}
B
Bryan Schumaker 已提交
2316
EXPORT_SYMBOL_GPL(nfs_permission);
L
Linus Torvalds 已提交
2317 2318 2319 2320 2321 2322 2323

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