dir.c 60.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
/*
 *  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
	struct nfs_inode *nfsi = NFS_I(dir);
73 74 75
	struct nfs_open_dir_context *ctx;
	ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
	if (ctx != NULL) {
76
		ctx->duped = 0;
77
		ctx->attr_gencount = nfsi->attr_gencount;
78
		ctx->dir_cookie = 0;
79
		ctx->dup_cookie = 0;
80
		ctx->cred = get_rpccred(cred);
81 82 83
		spin_lock(&dir->i_lock);
		list_add(&ctx->list, &nfsi->open_files);
		spin_unlock(&dir->i_lock);
84 85 86
		return ctx;
	}
	return  ERR_PTR(-ENOMEM);
87 88
}

89
static void put_nfs_open_dir_context(struct inode *dir, struct nfs_open_dir_context *ctx)
90
{
91 92 93
	spin_lock(&dir->i_lock);
	list_del(&ctx->list);
	spin_unlock(&dir->i_lock);
94 95 96 97
	put_rpccred(ctx->cred);
	kfree(ctx);
}

L
Linus Torvalds 已提交
98 99 100 101 102 103
/*
 * Open file
 */
static int
nfs_opendir(struct inode *inode, struct file *filp)
{
104 105 106
	int res = 0;
	struct nfs_open_dir_context *ctx;
	struct rpc_cred *cred;
L
Linus Torvalds 已提交
107

108
	dfprintk(FILE, "NFS: open dir(%pD2)\n", filp);
109 110

	nfs_inc_stats(inode, NFSIOS_VFSOPEN);
C
Chuck Lever 已提交
111

112 113 114
	cred = rpc_lookup_cred();
	if (IS_ERR(cred))
		return PTR_ERR(cred);
115
	ctx = alloc_nfs_open_dir_context(inode, cred);
116 117 118 119 120
	if (IS_ERR(ctx)) {
		res = PTR_ERR(ctx);
		goto out;
	}
	filp->private_data = ctx;
121 122 123 124 125 126 127
	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);
	}
128 129
out:
	put_rpccred(cred);
L
Linus Torvalds 已提交
130 131 132
	return res;
}

133 134 135
static int
nfs_closedir(struct inode *inode, struct file *filp)
{
136
	put_nfs_open_dir_context(filp->f_path.dentry->d_inode, filp->private_data);
137 138 139
	return 0;
}

B
Bryan Schumaker 已提交
140 141 142 143
struct nfs_cache_array_entry {
	u64 cookie;
	u64 ino;
	struct qstr string;
144
	unsigned char d_type;
B
Bryan Schumaker 已提交
145 146 147
};

struct nfs_cache_array {
148
	int size;
B
Bryan Schumaker 已提交
149 150 151 152 153
	int eof_index;
	u64 last_cookie;
	struct nfs_cache_array_entry array[0];
};

154
typedef int (*decode_dirent_t)(struct xdr_stream *, struct nfs_entry *, int);
L
Linus Torvalds 已提交
155 156 157
typedef struct {
	struct file	*file;
	struct page	*page;
A
Al Viro 已提交
158
	struct dir_context *ctx;
L
Linus Torvalds 已提交
159
	unsigned long	page_index;
160
	u64		*dir_cookie;
161
	u64		last_cookie;
162
	loff_t		current_index;
L
Linus Torvalds 已提交
163
	decode_dirent_t	decode;
B
Bryan Schumaker 已提交
164

165
	unsigned long	timestamp;
166
	unsigned long	gencount;
B
Bryan Schumaker 已提交
167 168 169
	unsigned int	cache_entry_index;
	unsigned int	plus:1;
	unsigned int	eof:1;
L
Linus Torvalds 已提交
170 171
} nfs_readdir_descriptor_t;

B
Bryan Schumaker 已提交
172 173
/*
 * The caller is responsible for calling nfs_readdir_release_array(page)
L
Linus Torvalds 已提交
174 175
 */
static
B
Bryan Schumaker 已提交
176 177
struct nfs_cache_array *nfs_readdir_get_array(struct page *page)
{
178
	void *ptr;
B
Bryan Schumaker 已提交
179 180
	if (page == NULL)
		return ERR_PTR(-EIO);
181 182 183 184
	ptr = kmap(page);
	if (ptr == NULL)
		return ERR_PTR(-ENOMEM);
	return ptr;
B
Bryan Schumaker 已提交
185 186 187 188 189 190 191 192 193 194 195 196
}

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

/*
 * we are freeing strings created by nfs_add_to_readdir_array()
 */
static
197
void nfs_readdir_clear_array(struct page *page)
B
Bryan Schumaker 已提交
198
{
199
	struct nfs_cache_array *array;
B
Bryan Schumaker 已提交
200
	int i;
201

202
	array = kmap_atomic(page);
B
Bryan Schumaker 已提交
203 204
	for (i = 0; i < array->size; i++)
		kfree(array->array[i].string.name);
205
	kunmap_atomic(array);
B
Bryan Schumaker 已提交
206 207 208 209 210 211 212 213
}

/*
 * 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
214
int nfs_readdir_make_qstr(struct qstr *string, const char *name, unsigned int len)
B
Bryan Schumaker 已提交
215 216 217
{
	string->len = len;
	string->name = kmemdup(name, len, GFP_KERNEL);
218 219
	if (string->name == NULL)
		return -ENOMEM;
220 221 222 223 224
	/*
	 * 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);
225 226
	string->hash = full_name_hash(name, len);
	return 0;
B
Bryan Schumaker 已提交
227 228 229 230 231 232
}

static
int nfs_readdir_add_to_array(struct nfs_entry *entry, struct page *page)
{
	struct nfs_cache_array *array = nfs_readdir_get_array(page);
233 234 235
	struct nfs_cache_array_entry *cache_entry;
	int ret;

B
Bryan Schumaker 已提交
236 237
	if (IS_ERR(array))
		return PTR_ERR(array);
238 239 240 241

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

	/* Check that this entry lies within the page bounds */
242
	ret = -ENOSPC;
243
	if ((char *)&cache_entry[1] - (char *)page_address(page) > PAGE_SIZE)
244
		goto out;
B
Bryan Schumaker 已提交
245

246 247
	cache_entry->cookie = entry->prev_cookie;
	cache_entry->ino = entry->ino;
248
	cache_entry->d_type = entry->d_type;
249 250 251
	ret = nfs_readdir_make_qstr(&cache_entry->string, entry->name, entry->len);
	if (ret)
		goto out;
B
Bryan Schumaker 已提交
252
	array->last_cookie = entry->cookie;
253
	array->size++;
T
Trond Myklebust 已提交
254
	if (entry->eof != 0)
B
Bryan Schumaker 已提交
255
		array->eof_index = array->size;
256
out:
B
Bryan Schumaker 已提交
257
	nfs_readdir_release_array(page);
258
	return ret;
B
Bryan Schumaker 已提交
259 260 261 262 263
}

static
int nfs_readdir_search_for_pos(struct nfs_cache_array *array, nfs_readdir_descriptor_t *desc)
{
A
Al Viro 已提交
264
	loff_t diff = desc->ctx->pos - desc->current_index;
B
Bryan Schumaker 已提交
265 266 267 268 269
	unsigned int index;

	if (diff < 0)
		goto out_eof;
	if (diff >= array->size) {
270
		if (array->eof_index >= 0)
B
Bryan Schumaker 已提交
271 272 273 274 275 276 277 278 279 280 281 282 283
			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;
}

284 285 286 287 288 289 290 291 292
static bool
nfs_readdir_inode_mapping_valid(struct nfs_inode *nfsi)
{
	if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
		return false;
	smp_rmb();
	return !test_bit(NFS_INO_INVALIDATING, &nfsi->flags);
}

B
Bryan Schumaker 已提交
293 294 295 296
static
int nfs_readdir_search_for_cookie(struct nfs_cache_array *array, nfs_readdir_descriptor_t *desc)
{
	int i;
297
	loff_t new_pos;
B
Bryan Schumaker 已提交
298 299 300 301
	int status = -EAGAIN;

	for (i = 0; i < array->size; i++) {
		if (array->array[i].cookie == *desc->dir_cookie) {
A
Al Viro 已提交
302
			struct nfs_inode *nfsi = NFS_I(file_inode(desc->file));
303 304
			struct nfs_open_dir_context *ctx = desc->file->private_data;

305
			new_pos = desc->current_index + i;
306 307
			if (ctx->attr_gencount != nfsi->attr_gencount ||
			    !nfs_readdir_inode_mapping_valid(nfsi)) {
308 309
				ctx->duped = 0;
				ctx->attr_gencount = nfsi->attr_gencount;
A
Al Viro 已提交
310
			} else if (new_pos < desc->ctx->pos) {
311 312 313
				if (ctx->duped > 0
				    && ctx->dup_cookie == *desc->dir_cookie) {
					if (printk_ratelimit()) {
314
						pr_notice("NFS: directory %pD2 contains a readdir loop."
315
								"Please contact your server vendor.  "
316 317 318
								"The file: %.*s has duplicate cookie %llu\n",
								desc->file, array->array[i].string.len,
								array->array[i].string.name, *desc->dir_cookie);
319 320 321 322
					}
					status = -ELOOP;
					goto out;
				}
323
				ctx->dup_cookie = *desc->dir_cookie;
324
				ctx->duped = -1;
325
			}
A
Al Viro 已提交
326
			desc->ctx->pos = new_pos;
B
Bryan Schumaker 已提交
327
			desc->cache_entry_index = i;
T
Trond Myklebust 已提交
328
			return 0;
B
Bryan Schumaker 已提交
329 330
		}
	}
T
Trond Myklebust 已提交
331
	if (array->eof_index >= 0) {
332
		status = -EBADCOOKIE;
333 334
		if (*desc->dir_cookie == array->last_cookie)
			desc->eof = 1;
335
	}
336
out:
B
Bryan Schumaker 已提交
337 338 339 340 341 342 343
	return status;
}

static
int nfs_readdir_search_array(nfs_readdir_descriptor_t *desc)
{
	struct nfs_cache_array *array;
T
Trond Myklebust 已提交
344
	int status;
B
Bryan Schumaker 已提交
345 346 347 348 349 350 351 352 353 354 355 356

	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
357
	if (status == -EAGAIN) {
358
		desc->last_cookie = array->last_cookie;
359
		desc->current_index += array->size;
T
Trond Myklebust 已提交
360 361
		desc->page_index++;
	}
B
Bryan Schumaker 已提交
362 363 364 365 366 367 368
	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
369
int nfs_readdir_xdr_filler(struct page **pages, nfs_readdir_descriptor_t *desc,
B
Bryan Schumaker 已提交
370
			struct nfs_entry *entry, struct file *file, struct inode *inode)
L
Linus Torvalds 已提交
371
{
372 373
	struct nfs_open_dir_context *ctx = file->private_data;
	struct rpc_cred	*cred = ctx->cred;
374
	unsigned long	timestamp, gencount;
L
Linus Torvalds 已提交
375 376 377 378
	int		error;

 again:
	timestamp = jiffies;
379
	gencount = nfs_inc_attr_generation_counter();
B
Bryan Schumaker 已提交
380
	error = NFS_PROTO(inode)->readdir(file->f_path.dentry, cred, entry->cookie, pages,
L
Linus Torvalds 已提交
381 382 383 384 385
					  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
386
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
387 388 389 390 391
			desc->plus = 0;
			goto again;
		}
		goto error;
	}
392
	desc->timestamp = timestamp;
393
	desc->gencount = gencount;
B
Bryan Schumaker 已提交
394 395
error:
	return error;
L
Linus Torvalds 已提交
396 397
}

398 399
static int xdr_decode(nfs_readdir_descriptor_t *desc,
		      struct nfs_entry *entry, struct xdr_stream *xdr)
L
Linus Torvalds 已提交
400
{
401
	int error;
L
Linus Torvalds 已提交
402

403 404 405
	error = desc->decode(xdr, entry, desc->plus);
	if (error)
		return error;
B
Bryan Schumaker 已提交
406 407 408
	entry->fattr->time_start = desc->timestamp;
	entry->fattr->gencount = desc->gencount;
	return 0;
L
Linus Torvalds 已提交
409 410
}

B
Bryan Schumaker 已提交
411 412 413 414 415
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
416
	if (nfs_compare_fh(entry->fh, NFS_FH(dentry->d_inode)) != 0)
B
Bryan Schumaker 已提交
417 418 419 420 421 422
		goto different;
	return 1;
different:
	return 0;
}

423
static
A
Al Viro 已提交
424
bool nfs_use_readdirplus(struct inode *dir, struct dir_context *ctx)
425 426 427 428 429
{
	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
430
	if (ctx->pos == 0)
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445
		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);
}

446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
/*
 * This function is mainly for use by nfs_getattr().
 *
 * If this is an 'ls -l', we want to force use of readdirplus.
 * Do this by checking if there is an active file descriptor
 * and calling nfs_advise_use_readdirplus, then forcing a
 * cache flush.
 */
void nfs_force_use_readdirplus(struct inode *dir)
{
	if (!list_empty(&NFS_I(dir)->open_files)) {
		nfs_advise_use_readdirplus(dir);
		nfs_zap_mapping(dir, dir->i_mapping);
	}
}

B
Bryan Schumaker 已提交
462 463 464
static
void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
{
465
	struct qstr filename = QSTR_INIT(entry->name, entry->len);
466 467
	struct dentry *dentry;
	struct dentry *alias;
B
Bryan Schumaker 已提交
468 469
	struct inode *dir = parent->d_inode;
	struct inode *inode;
470
	int status;
B
Bryan Schumaker 已提交
471

472 473 474 475 476 477 478
	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
479

480
	dentry = d_lookup(parent, &filename);
B
Bryan Schumaker 已提交
481 482
	if (dentry != NULL) {
		if (nfs_same_file(dentry, entry)) {
483
			nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
484 485 486
			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
487 488
			goto out;
		} else {
489 490
			if (d_invalidate(dentry) != 0)
				goto out;
B
Bryan Schumaker 已提交
491 492 493 494 495
			dput(dentry);
		}
	}

	dentry = d_alloc(parent, &filename);
496 497 498
	if (dentry == NULL)
		return;

499
	inode = nfs_fhget(dentry->d_sb, entry->fh, entry->fattr, entry->label);
B
Bryan Schumaker 已提交
500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
	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
516 517
/* Perform conversion from xdr to cache array */
static
518
int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *entry,
519
				struct page **xdr_pages, struct page *page, unsigned int buflen)
L
Linus Torvalds 已提交
520
{
521
	struct xdr_stream stream;
522
	struct xdr_buf buf;
523
	struct page *scratch;
B
Bryan Schumaker 已提交
524
	struct nfs_cache_array *array;
525 526
	unsigned int count = 0;
	int status;
527

528 529 530
	scratch = alloc_page(GFP_KERNEL);
	if (scratch == NULL)
		return -ENOMEM;
531

532
	xdr_init_decode_pages(&stream, &buf, xdr_pages, buflen);
533
	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
B
Bryan Schumaker 已提交
534 535 536

	do {
		status = xdr_decode(desc, entry, &stream);
537 538 539
		if (status != 0) {
			if (status == -EAGAIN)
				status = 0;
B
Bryan Schumaker 已提交
540
			break;
541
		}
B
Bryan Schumaker 已提交
542

543 544
		count++;

T
Trond Myklebust 已提交
545
		if (desc->plus != 0)
B
Bryan Schumaker 已提交
546
			nfs_prime_dcache(desc->file->f_path.dentry, entry);
547 548 549 550

		status = nfs_readdir_add_to_array(entry, page);
		if (status != 0)
			break;
B
Bryan Schumaker 已提交
551 552
	} while (!entry->eof);

T
Trond Myklebust 已提交
553
	if (count == 0 || (status == -EBADCOOKIE && entry->eof != 0)) {
B
Bryan Schumaker 已提交
554
		array = nfs_readdir_get_array(page);
555 556 557 558
		if (!IS_ERR(array)) {
			array->eof_index = array->size;
			status = 0;
			nfs_readdir_release_array(page);
559 560
		} else
			status = PTR_ERR(array);
L
Linus Torvalds 已提交
561
	}
562 563

	put_page(scratch);
564
	return status;
B
Bryan Schumaker 已提交
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586
}

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
587
int nfs_readdir_large_page(struct page **pages, unsigned int npages)
B
Bryan Schumaker 已提交
588 589 590 591 592 593 594 595 596
{
	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;
	}
597
	return 0;
B
Bryan Schumaker 已提交
598 599 600

out_freepages:
	nfs_readdir_free_pagearray(pages, i);
601
	return -ENOMEM;
L
Linus Torvalds 已提交
602 603
}

B
Bryan Schumaker 已提交
604 605
static
int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode)
606
{
B
Bryan Schumaker 已提交
607 608
	struct page *pages[NFS_MAX_READDIR_PAGES];
	void *pages_ptr = NULL;
B
Bryan Schumaker 已提交
609 610 611
	struct nfs_entry entry;
	struct file	*file = desc->file;
	struct nfs_cache_array *array;
612
	int status = -ENOMEM;
B
Bryan Schumaker 已提交
613
	unsigned int array_size = ARRAY_SIZE(pages);
B
Bryan Schumaker 已提交
614 615

	entry.prev_cookie = 0;
616
	entry.cookie = desc->last_cookie;
B
Bryan Schumaker 已提交
617 618 619
	entry.eof = 0;
	entry.fh = nfs_alloc_fhandle();
	entry.fattr = nfs_alloc_fattr();
620
	entry.server = NFS_SERVER(inode);
B
Bryan Schumaker 已提交
621 622
	if (entry.fh == NULL || entry.fattr == NULL)
		goto out;
623

624 625 626 627 628 629
	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
630
	array = nfs_readdir_get_array(page);
631 632
	if (IS_ERR(array)) {
		status = PTR_ERR(array);
633
		goto out_label_free;
634
	}
B
Bryan Schumaker 已提交
635 636
	memset(array, 0, sizeof(struct nfs_cache_array));
	array->eof_index = -1;
637

638 639
	status = nfs_readdir_large_page(pages, array_size);
	if (status < 0)
B
Bryan Schumaker 已提交
640 641
		goto out_release_array;
	do {
642
		unsigned int pglen;
B
Bryan Schumaker 已提交
643
		status = nfs_readdir_xdr_filler(pages, desc, &entry, file, inode);
644

B
Bryan Schumaker 已提交
645
		if (status < 0)
646
			break;
647
		pglen = status;
648
		status = nfs_readdir_page_filler(desc, &entry, pages, page, pglen);
649 650 651 652 653 654
		if (status < 0) {
			if (status == -ENOSPC)
				status = 0;
			break;
		}
	} while (array->eof_index < 0);
B
Bryan Schumaker 已提交
655

B
Bryan Schumaker 已提交
656
	nfs_readdir_free_large_page(pages_ptr, pages, array_size);
B
Bryan Schumaker 已提交
657 658
out_release_array:
	nfs_readdir_release_array(page);
659 660
out_label_free:
	nfs4_label_free(entry.label);
B
Bryan Schumaker 已提交
661 662 663
out:
	nfs_free_fattr(entry.fattr);
	nfs_free_fhandle(entry.fh);
664 665 666 667
	return status;
}

/*
B
Bryan Schumaker 已提交
668 669 670 671
 * 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
672
 */
B
Bryan Schumaker 已提交
673 674
static
int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page)
L
Linus Torvalds 已提交
675
{
A
Al Viro 已提交
676
	struct inode	*inode = file_inode(desc->file);
677
	int ret;
L
Linus Torvalds 已提交
678

679 680
	ret = nfs_readdir_xdr_to_array(desc, page, inode);
	if (ret < 0)
B
Bryan Schumaker 已提交
681 682
		goto error;
	SetPageUptodate(page);
L
Linus Torvalds 已提交
683

B
Bryan Schumaker 已提交
684 685 686
	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
687
	}
B
Bryan Schumaker 已提交
688 689 690 691
	unlock_page(page);
	return 0;
 error:
	unlock_page(page);
692
	return ret;
B
Bryan Schumaker 已提交
693
}
L
Linus Torvalds 已提交
694

B
Bryan Schumaker 已提交
695 696 697
static
void cache_page_release(nfs_readdir_descriptor_t *desc)
{
698 699
	if (!desc->page->mapping)
		nfs_readdir_clear_array(desc->page);
B
Bryan Schumaker 已提交
700 701 702 703 704 705 706
	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
707
	return read_cache_page(file_inode(desc->file)->i_mapping,
B
Bryan Schumaker 已提交
708
			desc->page_index, (filler_t *)nfs_readdir_filler, desc);
L
Linus Torvalds 已提交
709 710 711
}

/*
B
Bryan Schumaker 已提交
712
 * Returns 0 if desc->dir_cookie was found on page desc->page_index
L
Linus Torvalds 已提交
713
 */
B
Bryan Schumaker 已提交
714 715 716 717 718 719 720 721 722 723
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
724 725
	if (res != 0)
		cache_page_release(desc);
B
Bryan Schumaker 已提交
726 727 728 729
	return res;
}

/* Search for desc->dir_cookie from the beginning of the page cache */
L
Linus Torvalds 已提交
730 731 732
static inline
int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
{
733
	int res;
B
Bryan Schumaker 已提交
734

735
	if (desc->page_index == 0) {
736
		desc->current_index = 0;
737 738
		desc->last_cookie = 0;
	}
T
Trond Myklebust 已提交
739
	do {
B
Bryan Schumaker 已提交
740
		res = find_cache_page(desc);
T
Trond Myklebust 已提交
741
	} while (res == -EAGAIN);
L
Linus Torvalds 已提交
742 743 744 745 746 747 748
	return res;
}

/*
 * Once we've found the start of the dirent within a page: fill 'er up...
 */
static 
A
Al Viro 已提交
749
int nfs_do_filldir(nfs_readdir_descriptor_t *desc)
L
Linus Torvalds 已提交
750 751
{
	struct file	*file = desc->file;
B
Bryan Schumaker 已提交
752 753 754
	int i = 0;
	int res = 0;
	struct nfs_cache_array *array = NULL;
755 756
	struct nfs_open_dir_context *ctx = file->private_data;

B
Bryan Schumaker 已提交
757
	array = nfs_readdir_get_array(desc->page);
758 759 760 761
	if (IS_ERR(array)) {
		res = PTR_ERR(array);
		goto out;
	}
B
Bryan Schumaker 已提交
762 763

	for (i = desc->cache_entry_index; i < array->size; i++) {
764
		struct nfs_cache_array_entry *ent;
L
Linus Torvalds 已提交
765

766
		ent = &array->array[i];
A
Al Viro 已提交
767 768
		if (!dir_emit(desc->ctx, ent->string.name, ent->string.len,
		    nfs_compat_user_ino64(ent->ino), ent->d_type)) {
769
			desc->eof = 1;
L
Linus Torvalds 已提交
770
			break;
771
		}
A
Al Viro 已提交
772
		desc->ctx->pos++;
B
Bryan Schumaker 已提交
773 774 775 776
		if (i < (array->size-1))
			*desc->dir_cookie = array->array[i+1].cookie;
		else
			*desc->dir_cookie = array->last_cookie;
777 778
		if (ctx->duped != 0)
			ctx->duped = 1;
L
Linus Torvalds 已提交
779
	}
T
Trond Myklebust 已提交
780
	if (array->eof_index >= 0)
781
		desc->eof = 1;
B
Bryan Schumaker 已提交
782 783

	nfs_readdir_release_array(desc->page);
784
out:
B
Bryan Schumaker 已提交
785
	cache_page_release(desc);
C
Chuck Lever 已提交
786 787
	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
788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803
	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
804
int uncached_readdir(nfs_readdir_descriptor_t *desc)
L
Linus Torvalds 已提交
805 806 807
{
	struct page	*page = NULL;
	int		status;
A
Al Viro 已提交
808
	struct inode *inode = file_inode(desc->file);
809
	struct nfs_open_dir_context *ctx = desc->file->private_data;
L
Linus Torvalds 已提交
810

C
Chuck Lever 已提交
811 812
	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
			(unsigned long long)*desc->dir_cookie);
L
Linus Torvalds 已提交
813 814 815 816 817 818

	page = alloc_page(GFP_HIGHUSER);
	if (!page) {
		status = -ENOMEM;
		goto out;
	}
B
Bryan Schumaker 已提交
819

820
	desc->page_index = 0;
821
	desc->last_cookie = *desc->dir_cookie;
822
	desc->page = page;
823
	ctx->duped = 0;
824

825 826
	status = nfs_readdir_xdr_to_array(desc, page, inode);
	if (status < 0)
L
Linus Torvalds 已提交
827 828
		goto out_release;

A
Al Viro 已提交
829
	status = nfs_do_filldir(desc);
L
Linus Torvalds 已提交
830 831

 out:
C
Chuck Lever 已提交
832
	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
833
			__func__, status);
L
Linus Torvalds 已提交
834 835
	return status;
 out_release:
B
Bryan Schumaker 已提交
836
	cache_page_release(desc);
L
Linus Torvalds 已提交
837 838 839
	goto out;
}

840 841 842 843 844 845 846 847 848 849 850
static bool nfs_dir_mapping_need_revalidate(struct inode *dir)
{
	struct nfs_inode *nfsi = NFS_I(dir);

	if (nfs_attribute_cache_expired(dir))
		return true;
	if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
		return true;
	return false;
}

851 852 853
/* 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
854
 */
A
Al Viro 已提交
855
static int nfs_readdir(struct file *file, struct dir_context *ctx)
L
Linus Torvalds 已提交
856
{
A
Al Viro 已提交
857
	struct dentry	*dentry = file->f_path.dentry;
L
Linus Torvalds 已提交
858 859 860
	struct inode	*inode = dentry->d_inode;
	nfs_readdir_descriptor_t my_desc,
			*desc = &my_desc;
A
Al Viro 已提交
861
	struct nfs_open_dir_context *dir_ctx = file->private_data;
862
	int res = 0;
L
Linus Torvalds 已提交
863

864 865
	dfprintk(FILE, "NFS: readdir(%pD2) starting at cookie %llu\n",
			file, (long long)ctx->pos);
C
Chuck Lever 已提交
866 867
	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);

L
Linus Torvalds 已提交
868
	/*
A
Al Viro 已提交
869
	 * ctx->pos points to the dirent entry number.
870
	 * *desc->dir_cookie has the cookie for the next entry. We have
871 872
	 * to either find the entry with the appropriate number or
	 * revalidate the cookie.
L
Linus Torvalds 已提交
873 874 875
	 */
	memset(desc, 0, sizeof(*desc));

A
Al Viro 已提交
876 877
	desc->file = file;
	desc->ctx = ctx;
878
	desc->dir_cookie = &dir_ctx->dir_cookie;
L
Linus Torvalds 已提交
879
	desc->decode = NFS_PROTO(inode)->decode_dirent;
A
Al Viro 已提交
880
	desc->plus = nfs_use_readdirplus(inode, ctx) ? 1 : 0;
L
Linus Torvalds 已提交
881

T
Trond Myklebust 已提交
882
	nfs_block_sillyrename(dentry);
883
	if (ctx->pos == 0 || nfs_dir_mapping_need_revalidate(inode))
884
		res = nfs_revalidate_mapping(inode, file->f_mapping);
T
Trond Myklebust 已提交
885 886 887
	if (res < 0)
		goto out;

T
Trond Myklebust 已提交
888
	do {
L
Linus Torvalds 已提交
889
		res = readdir_search_pagecache(desc);
890

L
Linus Torvalds 已提交
891
		if (res == -EBADCOOKIE) {
892
			res = 0;
L
Linus Torvalds 已提交
893
			/* This means either end of directory */
B
Bryan Schumaker 已提交
894
			if (*desc->dir_cookie && desc->eof == 0) {
L
Linus Torvalds 已提交
895
				/* Or that the server has 'lost' a cookie */
A
Al Viro 已提交
896
				res = uncached_readdir(desc);
897
				if (res == 0)
L
Linus Torvalds 已提交
898 899 900 901 902
					continue;
			}
			break;
		}
		if (res == -ETOOSMALL && desc->plus) {
B
Benny Halevy 已提交
903
			clear_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
L
Linus Torvalds 已提交
904
			nfs_zap_caches(inode);
905
			desc->page_index = 0;
L
Linus Torvalds 已提交
906
			desc->plus = 0;
B
Bryan Schumaker 已提交
907
			desc->eof = 0;
L
Linus Torvalds 已提交
908 909 910 911 912
			continue;
		}
		if (res < 0)
			break;

A
Al Viro 已提交
913
		res = nfs_do_filldir(desc);
914
		if (res < 0)
L
Linus Torvalds 已提交
915
			break;
T
Trond Myklebust 已提交
916
	} while (!desc->eof);
T
Trond Myklebust 已提交
917
out:
T
Trond Myklebust 已提交
918
	nfs_unblock_sillyrename(dentry);
C
Chuck Lever 已提交
919 920
	if (res > 0)
		res = 0;
921
	dfprintk(FILE, "NFS: readdir(%pD2) returns %d\n", file, res);
C
Chuck Lever 已提交
922
	return res;
L
Linus Torvalds 已提交
923 924
}

925
static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int whence)
926
{
927
	struct inode *inode = file_inode(filp);
928
	struct nfs_open_dir_context *dir_ctx = filp->private_data;
929

930 931
	dfprintk(FILE, "NFS: llseek dir(%pD2, %lld, %d)\n",
			filp, offset, whence);
932 933

	mutex_lock(&inode->i_mutex);
934
	switch (whence) {
935 936 937 938 939 940 941 942 943 944 945
		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;
946
		dir_ctx->dir_cookie = 0;
947
		dir_ctx->duped = 0;
948 949
	}
out:
950
	mutex_unlock(&inode->i_mutex);
951 952 953
	return offset;
}

L
Linus Torvalds 已提交
954 955 956 957
/*
 * All directory operations under NFS are synchronous, so fsync()
 * is a dummy operation.
 */
958 959
static int nfs_fsync_dir(struct file *filp, loff_t start, loff_t end,
			 int datasync)
L
Linus Torvalds 已提交
960
{
961
	struct inode *inode = file_inode(filp);
962

963
	dfprintk(FILE, "NFS: fsync dir(%pD2) datasync %d\n", filp, datasync);
C
Chuck Lever 已提交
964

965
	mutex_lock(&inode->i_mutex);
966
	nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
967
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
968 969 970
	return 0;
}

971 972 973 974 975 976 977 978 979 980 981 982
/**
 * 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)
{
983
	NFS_I(dir)->cache_change_attribute++;
984
}
B
Bryan Schumaker 已提交
985
EXPORT_SYMBOL_GPL(nfs_force_lookup_revalidate);
986

L
Linus Torvalds 已提交
987 988 989 990 991
/*
 * 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.
 */
992
static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
993 994 995
{
	if (IS_ROOT(dentry))
		return 1;
996 997
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
		return 0;
998 999 1000 1001 1002 1003 1004 1005
	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
1006 1007
}

1008 1009 1010 1011
/*
 * Use intent information to check whether or not we're going to do
 * an O_EXCL create using this path component.
 */
1012
static int nfs_is_exclusive_create(struct inode *dir, unsigned int flags)
1013 1014 1015
{
	if (NFS_PROTO(dir)->version == 2)
		return 0;
1016
	return flags & LOOKUP_EXCL;
1017 1018
}

1019 1020 1021 1022 1023 1024 1025 1026
/*
 * 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.
 *
 */
1027
static
1028
int nfs_lookup_verify_inode(struct inode *inode, unsigned int flags)
L
Linus Torvalds 已提交
1029 1030
{
	struct nfs_server *server = NFS_SERVER(inode);
1031
	int ret;
L
Linus Torvalds 已提交
1032

1033
	if (IS_AUTOMOUNT(inode))
1034
		return 0;
1035
	/* VFS wants an on-the-wire revalidation */
1036
	if (flags & LOOKUP_REVAL)
1037 1038
		goto out_force;
	/* This is an open(2) */
1039 1040
	if ((flags & LOOKUP_OPEN) && !(server->flags & NFS_MOUNT_NOCTO) &&
	    (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1041
		goto out_force;
1042 1043
out:
	return (inode->i_nlink == 0) ? -ENOENT : 0;
L
Linus Torvalds 已提交
1044
out_force:
1045 1046 1047 1048
	ret = __nfs_revalidate_inode(server, inode);
	if (ret != 0)
		return ret;
	goto out;
L
Linus Torvalds 已提交
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
}

/*
 * 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,
1060
		       unsigned int flags)
L
Linus Torvalds 已提交
1061 1062
{
	/* Don't revalidate a negative dentry if we're creating a new file */
1063
	if (flags & LOOKUP_CREATE)
L
Linus Torvalds 已提交
1064
		return 0;
1065 1066
	if (NFS_SERVER(dir)->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG)
		return 1;
L
Linus Torvalds 已提交
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
	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.
 */
1081
static int nfs_lookup_revalidate(struct dentry *dentry, unsigned int flags)
L
Linus Torvalds 已提交
1082 1083 1084 1085
{
	struct inode *dir;
	struct inode *inode;
	struct dentry *parent;
1086 1087
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
1088
	struct nfs4_label *label = NULL;
L
Linus Torvalds 已提交
1089 1090
	int error;

1091 1092 1093 1094 1095 1096 1097 1098 1099
	if (flags & LOOKUP_RCU) {
		parent = rcu_dereference(dentry->d_parent);
		dir = ACCESS_ONCE(parent->d_inode);
		if (!dir)
			return -ECHILD;
	} else {
		parent = dget_parent(dentry);
		dir = parent->d_inode;
	}
C
Chuck Lever 已提交
1100
	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
L
Linus Torvalds 已提交
1101 1102 1103
	inode = dentry->d_inode;

	if (!inode) {
1104 1105 1106
		if (flags & LOOKUP_RCU)
			return -ECHILD;

1107
		if (nfs_neg_need_reval(dir, dentry, flags))
L
Linus Torvalds 已提交
1108
			goto out_bad;
1109
		goto out_valid_noent;
L
Linus Torvalds 已提交
1110 1111 1112
	}

	if (is_bad_inode(inode)) {
1113 1114
		if (flags & LOOKUP_RCU)
			return -ECHILD;
1115 1116
		dfprintk(LOOKUPCACHE, "%s: %pd2 has dud inode\n",
				__func__, dentry);
L
Linus Torvalds 已提交
1117 1118 1119
		goto out_bad;
	}

1120
	if (NFS_PROTO(dir)->have_delegation(inode, FMODE_READ))
1121 1122
		goto out_set_verifier;

1123 1124 1125
	if (flags & LOOKUP_RCU)
		return -ECHILD;

L
Linus Torvalds 已提交
1126
	/* Force a full look up iff the parent directory has changed */
1127 1128
	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
1129 1130 1131 1132 1133 1134 1135
			goto out_zap_parent;
		goto out_valid;
	}

	if (NFS_STALE(inode))
		goto out_bad;

1136 1137 1138 1139 1140 1141
	error = -ENOMEM;
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out_error;

1142 1143 1144 1145
	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_NOWAIT);
	if (IS_ERR(label))
		goto out_error;

1146
	trace_nfs_lookup_revalidate_enter(dir, dentry, flags);
1147
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, label);
1148
	trace_nfs_lookup_revalidate_exit(dir, dentry, flags, error);
L
Linus Torvalds 已提交
1149 1150
	if (error)
		goto out_bad;
1151
	if (nfs_compare_fh(NFS_FH(inode), fhandle))
L
Linus Torvalds 已提交
1152
		goto out_bad;
1153
	if ((error = nfs_refresh_inode(inode, fattr)) != 0)
L
Linus Torvalds 已提交
1154 1155
		goto out_bad;

1156 1157
	nfs_setsecurity(inode, fattr, label);

1158 1159
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1160 1161
	nfs4_label_free(label);

1162
out_set_verifier:
1163
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1164
 out_valid:
1165 1166 1167
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);
 out_valid_noent:
1168 1169 1170 1171 1172
	if (flags & LOOKUP_RCU) {
		if (parent != rcu_dereference(dentry->d_parent))
			return -ECHILD;
	} else
		dput(parent);
1173 1174
	dfprintk(LOOKUPCACHE, "NFS: %s(%pd2) is valid\n",
			__func__, dentry);
L
Linus Torvalds 已提交
1175 1176 1177 1178
	return 1;
out_zap_parent:
	nfs_zap_caches(dir);
 out_bad:
1179
	WARN_ON(flags & LOOKUP_RCU);
A
Al Viro 已提交
1180 1181
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1182
	nfs4_label_free(label);
1183
	nfs_mark_for_revalidate(dir);
L
Linus Torvalds 已提交
1184 1185 1186
	if (inode && S_ISDIR(inode->i_mode)) {
		/* Purge readdir caches. */
		nfs_zap_caches(inode);
1187 1188 1189 1190 1191 1192 1193
		/*
		 * 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))
1194
			goto out_valid;
L
Linus Torvalds 已提交
1195
	}
1196 1197 1198 1199
	/* If we have submounts, don't unhash ! */
	if (check_submounts_and_drop(dentry) != 0)
		goto out_valid;

L
Linus Torvalds 已提交
1200
	dput(parent);
1201 1202
	dfprintk(LOOKUPCACHE, "NFS: %s(%pd2) is invalid\n",
			__func__, dentry);
L
Linus Torvalds 已提交
1203
	return 0;
1204
out_error:
1205
	WARN_ON(flags & LOOKUP_RCU);
1206 1207
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
1208
	nfs4_label_free(label);
1209
	dput(parent);
1210 1211
	dfprintk(LOOKUPCACHE, "NFS: %s(%pd2) lookup returned error %d\n",
			__func__, dentry, error);
1212
	return error;
L
Linus Torvalds 已提交
1213 1214
}

1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234
/*
 * 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) {
1235 1236
		dfprintk(LOOKUPCACHE, "%s: %pd2 has negative inode\n",
				__func__, dentry);
1237 1238 1239 1240
		return 1;
	}

	if (is_bad_inode(inode)) {
1241 1242
		dfprintk(LOOKUPCACHE, "%s: %pd2 has dud inode\n",
				__func__, dentry);
1243 1244 1245 1246 1247 1248 1249 1250 1251
		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
1252 1253 1254
/*
 * This is called from dput() when d_count is going to 0.
 */
N
Nick Piggin 已提交
1255
static int nfs_dentry_delete(const struct dentry *dentry)
L
Linus Torvalds 已提交
1256
{
1257 1258
	dfprintk(VFS, "NFS: dentry_delete(%pd2, %x)\n",
		dentry, dentry->d_flags);
L
Linus Torvalds 已提交
1259

1260 1261 1262 1263
	/* 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
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276
	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
1277
/* Ensure that we revalidate inode->i_nlink */
1278 1279 1280
static void nfs_drop_nlink(struct inode *inode)
{
	spin_lock(&inode->i_lock);
T
Trond Myklebust 已提交
1281 1282 1283 1284
	/* 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;
1285 1286 1287
	spin_unlock(&inode->i_lock);
}

L
Linus Torvalds 已提交
1288 1289 1290 1291 1292 1293
/*
 * 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)
{
1294 1295 1296 1297
	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
1298
	if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
1299
		nfs_complete_unlink(dentry, inode);
T
Trond Myklebust 已提交
1300
		nfs_drop_nlink(inode);
L
Linus Torvalds 已提交
1301 1302 1303 1304
	}
	iput(inode);
}

1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315
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
1316
const struct dentry_operations nfs_dentry_operations = {
L
Linus Torvalds 已提交
1317
	.d_revalidate	= nfs_lookup_revalidate,
1318
	.d_weak_revalidate	= nfs_weak_revalidate,
L
Linus Torvalds 已提交
1319 1320
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1321
	.d_automount	= nfs_d_automount,
1322
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1323
};
B
Bryan Schumaker 已提交
1324
EXPORT_SYMBOL_GPL(nfs_dentry_operations);
L
Linus Torvalds 已提交
1325

1326
struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
L
Linus Torvalds 已提交
1327 1328
{
	struct dentry *res;
T
Trond Myklebust 已提交
1329
	struct dentry *parent;
L
Linus Torvalds 已提交
1330
	struct inode *inode = NULL;
1331 1332
	struct nfs_fh *fhandle = NULL;
	struct nfs_fattr *fattr = NULL;
1333
	struct nfs4_label *label = NULL;
L
Linus Torvalds 已提交
1334 1335
	int error;

1336
	dfprintk(VFS, "NFS: lookup(%pd2)\n", dentry);
C
Chuck Lever 已提交
1337
	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
L
Linus Torvalds 已提交
1338 1339 1340 1341 1342

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

1343 1344 1345 1346
	/*
	 * If we're doing an exclusive create, optimize away the lookup
	 * but don't hash the dentry.
	 */
A
Al Viro 已提交
1347
	if (nfs_is_exclusive_create(dir, flags)) {
1348 1349
		d_instantiate(dentry, NULL);
		res = NULL;
1350
		goto out;
1351
	}
L
Linus Torvalds 已提交
1352

1353 1354 1355 1356 1357 1358
	res = ERR_PTR(-ENOMEM);
	fhandle = nfs_alloc_fhandle();
	fattr = nfs_alloc_fattr();
	if (fhandle == NULL || fattr == NULL)
		goto out;

1359 1360 1361 1362
	label = nfs4_label_alloc(NFS_SERVER(dir), GFP_NOWAIT);
	if (IS_ERR(label))
		goto out;

T
Trond Myklebust 已提交
1363 1364
	parent = dentry->d_parent;
	/* Protect against concurrent sillydeletes */
1365
	trace_nfs_lookup_enter(dir, dentry, flags);
T
Trond Myklebust 已提交
1366
	nfs_block_sillyrename(parent);
1367
	error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, label);
L
Linus Torvalds 已提交
1368 1369 1370 1371
	if (error == -ENOENT)
		goto no_entry;
	if (error < 0) {
		res = ERR_PTR(error);
T
Trond Myklebust 已提交
1372
		goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1373
	}
1374
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr, label);
N
Namhyung Kim 已提交
1375
	res = ERR_CAST(inode);
1376
	if (IS_ERR(res))
T
Trond Myklebust 已提交
1377
		goto out_unblock_sillyrename;
1378

1379 1380 1381
	/* Success: notify readdir to use READDIRPLUS */
	nfs_advise_use_readdirplus(dir);

L
Linus Torvalds 已提交
1382
no_entry:
1383
	res = d_materialise_unique(dentry, inode);
1384 1385
	if (res != NULL) {
		if (IS_ERR(res))
T
Trond Myklebust 已提交
1386
			goto out_unblock_sillyrename;
L
Linus Torvalds 已提交
1387
		dentry = res;
1388
	}
L
Linus Torvalds 已提交
1389
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
T
Trond Myklebust 已提交
1390 1391
out_unblock_sillyrename:
	nfs_unblock_sillyrename(parent);
1392
	trace_nfs_lookup_exit(dir, dentry, flags, error);
1393
	nfs4_label_free(label);
L
Linus Torvalds 已提交
1394
out:
1395 1396
	nfs_free_fattr(fattr);
	nfs_free_fhandle(fhandle);
L
Linus Torvalds 已提交
1397 1398
	return res;
}
B
Bryan Schumaker 已提交
1399
EXPORT_SYMBOL_GPL(nfs_lookup);
L
Linus Torvalds 已提交
1400

B
Bryan Schumaker 已提交
1401
#if IS_ENABLED(CONFIG_NFS_V4)
1402
static int nfs4_lookup_revalidate(struct dentry *, unsigned int);
L
Linus Torvalds 已提交
1403

A
Al Viro 已提交
1404
const struct dentry_operations nfs4_dentry_operations = {
1405
	.d_revalidate	= nfs4_lookup_revalidate,
L
Linus Torvalds 已提交
1406 1407
	.d_delete	= nfs_dentry_delete,
	.d_iput		= nfs_dentry_iput,
1408
	.d_automount	= nfs_d_automount,
1409
	.d_release	= nfs_d_release,
L
Linus Torvalds 已提交
1410
};
B
Bryan Schumaker 已提交
1411
EXPORT_SYMBOL_GPL(nfs4_dentry_operations);
L
Linus Torvalds 已提交
1412

1413 1414 1415 1416 1417 1418 1419 1420 1421 1422
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;
}

1423
static struct nfs_open_context *create_nfs_open_context(struct dentry *dentry, int open_flags)
1424
{
1425
	return alloc_nfs_open_context(dentry, flags_to_mode(open_flags));
1426 1427 1428 1429
}

static int do_open(struct inode *inode, struct file *filp)
{
1430
	nfs_fscache_open_file(inode, filp);
1431 1432 1433
	return 0;
}

A
Al Viro 已提交
1434 1435
static int nfs_finish_open(struct nfs_open_context *ctx,
			   struct dentry *dentry,
A
Al Viro 已提交
1436
			   struct file *file, unsigned open_flags,
A
Al Viro 已提交
1437
			   int *opened)
1438
{
1439 1440
	int err;

M
Miklos Szeredi 已提交
1441 1442 1443
	if ((open_flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
		*opened |= FILE_CREATED;

A
Al Viro 已提交
1444 1445
	err = finish_open(file, dentry, do_open, opened);
	if (err)
A
Al Viro 已提交
1446
		goto out;
A
Al Viro 已提交
1447
	nfs_file_set_open_context(file, ctx);
1448

1449
out:
A
Al Viro 已提交
1450
	return err;
1451 1452
}

1453 1454 1455
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
1456
{
1457
	struct nfs_open_context *ctx;
1458 1459
	struct dentry *res;
	struct iattr attr = { .ia_valid = ATTR_OPEN };
1460
	struct inode *inode;
1461
	unsigned int lookup_flags = 0;
1462
	int err;
L
Linus Torvalds 已提交
1463

1464 1465 1466
	/* Expect a negative dentry */
	BUG_ON(dentry->d_inode);

1467
	dfprintk(VFS, "NFS: atomic_open(%s/%lu), %pd\n",
1468
			dir->i_sb->s_id, dir->i_ino, dentry);
C
Chuck Lever 已提交
1469

1470 1471 1472 1473
	err = nfs_check_flags(open_flags);
	if (err)
		return err;

1474 1475 1476 1477 1478 1479 1480 1481
	/* 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
1482
			return -ENOENT;
1483
		}
1484
		lookup_flags = LOOKUP_OPEN|LOOKUP_DIRECTORY;
L
Linus Torvalds 已提交
1485
		goto no_open;
1486
	}
L
Linus Torvalds 已提交
1487

1488
	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
A
Al Viro 已提交
1489
		return -ENAMETOOLONG;
1490

1491
	if (open_flags & O_CREAT) {
1492
		attr.ia_valid |= ATTR_MODE;
1493 1494
		attr.ia_mode = mode & ~current_umask();
	}
1495 1496 1497
	if (open_flags & O_TRUNC) {
		attr.ia_valid |= ATTR_SIZE;
		attr.ia_size = 0;
1498 1499
	}

1500 1501 1502
	ctx = create_nfs_open_context(dentry, open_flags);
	err = PTR_ERR(ctx);
	if (IS_ERR(ctx))
A
Al Viro 已提交
1503
		goto out;
1504

1505
	trace_nfs_atomic_open_enter(dir, ctx, open_flags);
1506
	nfs_block_sillyrename(dentry->d_parent);
1507
	inode = NFS_PROTO(dir)->open_context(dir, ctx, open_flags, &attr, opened);
1508
	nfs_unblock_sillyrename(dentry->d_parent);
1509
	if (IS_ERR(inode)) {
1510
		err = PTR_ERR(inode);
1511
		trace_nfs_atomic_open_exit(dir, ctx, open_flags, err);
1512
		put_nfs_open_context(ctx);
1513 1514
		switch (err) {
		case -ENOENT:
1515
			d_drop(dentry);
1516 1517 1518 1519 1520 1521 1522
			d_add(dentry, NULL);
			break;
		case -EISDIR:
		case -ENOTDIR:
			goto no_open;
		case -ELOOP:
			if (!(open_flags & O_NOFOLLOW))
1523
				goto no_open;
1524
			break;
L
Linus Torvalds 已提交
1525
			/* case -EINVAL: */
1526 1527
		default:
			break;
L
Linus Torvalds 已提交
1528
		}
A
Al Viro 已提交
1529
		goto out;
1530
	}
1531

1532
	err = nfs_finish_open(ctx, ctx->dentry, file, open_flags, opened);
1533
	trace_nfs_atomic_open_exit(dir, ctx, open_flags, err);
1534
	put_nfs_open_context(ctx);
A
Al Viro 已提交
1535 1536
out:
	return err;
1537

L
Linus Torvalds 已提交
1538
no_open:
1539
	res = nfs_lookup(dir, dentry, lookup_flags);
1540 1541
	err = PTR_ERR(res);
	if (IS_ERR(res))
A
Al Viro 已提交
1542
		goto out;
1543

A
Al Viro 已提交
1544
	return finish_no_open(file, res);
L
Linus Torvalds 已提交
1545
}
B
Bryan Schumaker 已提交
1546
EXPORT_SYMBOL_GPL(nfs_atomic_open);
L
Linus Torvalds 已提交
1547

1548
static int nfs4_lookup_revalidate(struct dentry *dentry, unsigned int flags)
L
Linus Torvalds 已提交
1549
{
N
Nick Piggin 已提交
1550
	struct inode *inode;
1551
	int ret = 0;
L
Linus Torvalds 已提交
1552

1553
	if (!(flags & LOOKUP_OPEN) || (flags & LOOKUP_DIRECTORY))
1554 1555
		goto no_open;
	if (d_mountpoint(dentry))
T
Trond Myklebust 已提交
1556
		goto no_open;
1557 1558
	if (NFS_SB(dentry->d_sb)->caps & NFS_CAP_ATOMIC_OPEN_V1)
		goto no_open;
1559

1560
	inode = dentry->d_inode;
1561

L
Linus Torvalds 已提交
1562 1563 1564
	/* We can't create new files in nfs_open_revalidate(), so we
	 * optimize away revalidation of negative dentries.
	 */
1565
	if (inode == NULL) {
1566 1567 1568
		struct dentry *parent;
		struct inode *dir;

1569 1570 1571
		if (flags & LOOKUP_RCU)
			return -ECHILD;

1572 1573
		parent = dget_parent(dentry);
		dir = parent->d_inode;
1574
		if (!nfs_neg_need_reval(dir, dentry, flags))
1575
			ret = 1;
1576
		dput(parent);
L
Linus Torvalds 已提交
1577
		goto out;
1578 1579
	}

L
Linus Torvalds 已提交
1580 1581
	/* NFS only supports OPEN on regular files */
	if (!S_ISREG(inode->i_mode))
1582
		goto no_open;
L
Linus Torvalds 已提交
1583
	/* We cannot do exclusive creation on a positive dentry */
1584
	if (flags & LOOKUP_EXCL)
1585
		goto no_open;
L
Linus Torvalds 已提交
1586

1587 1588
	/* Let f_op->open() actually open (and revalidate) the file */
	ret = 1;
1589

L
Linus Torvalds 已提交
1590 1591
out:
	return ret;
1592

T
Trond Myklebust 已提交
1593
no_open:
1594
	return nfs_lookup_revalidate(dentry, flags);
1595 1596
}

L
Linus Torvalds 已提交
1597 1598 1599 1600 1601 1602
#endif /* CONFIG_NFSV4 */

/*
 * Code common to create, mkdir, and mknod.
 */
int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
1603 1604
				struct nfs_fattr *fattr,
				struct nfs4_label *label)
L
Linus Torvalds 已提交
1605
{
1606 1607
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
L
Linus Torvalds 已提交
1608 1609 1610
	struct inode *inode;
	int error = -EACCES;

1611 1612
	d_drop(dentry);

L
Linus Torvalds 已提交
1613 1614
	/* We may have been initialized further down */
	if (dentry->d_inode)
1615
		goto out;
L
Linus Torvalds 已提交
1616
	if (fhandle->size == 0) {
1617
		error = NFS_PROTO(dir)->lookup(dir, &dentry->d_name, fhandle, fattr, NULL);
L
Linus Torvalds 已提交
1618
		if (error)
1619
			goto out_error;
L
Linus Torvalds 已提交
1620
	}
1621
	nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
L
Linus Torvalds 已提交
1622 1623
	if (!(fattr->valid & NFS_ATTR_FATTR)) {
		struct nfs_server *server = NFS_SB(dentry->d_sb);
1624
		error = server->nfs_client->rpc_ops->getattr(server, fhandle, fattr, NULL);
L
Linus Torvalds 已提交
1625
		if (error < 0)
1626
			goto out_error;
L
Linus Torvalds 已提交
1627
	}
1628
	inode = nfs_fhget(dentry->d_sb, fhandle, fattr, label);
1629 1630
	error = PTR_ERR(inode);
	if (IS_ERR(inode))
1631 1632 1633 1634
		goto out_error;
	d_add(dentry, inode);
out:
	dput(parent);
L
Linus Torvalds 已提交
1635
	return 0;
1636 1637 1638 1639
out_error:
	nfs_mark_for_revalidate(dir);
	dput(parent);
	return error;
L
Linus Torvalds 已提交
1640
}
B
Bryan Schumaker 已提交
1641
EXPORT_SYMBOL_GPL(nfs_instantiate);
L
Linus Torvalds 已提交
1642 1643 1644 1645 1646 1647 1648

/*
 * 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.
 */
1649
int nfs_create(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
1650
		umode_t mode, bool excl)
L
Linus Torvalds 已提交
1651 1652
{
	struct iattr attr;
A
Al Viro 已提交
1653
	int open_flags = excl ? O_CREAT | O_EXCL : O_CREAT;
L
Linus Torvalds 已提交
1654 1655
	int error;

1656
	dfprintk(VFS, "NFS: create(%s/%lu), %pd\n",
1657
			dir->i_sb->s_id, dir->i_ino, dentry);
L
Linus Torvalds 已提交
1658 1659 1660 1661

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

1662
	trace_nfs_create_enter(dir, dentry, open_flags);
1663
	error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags);
1664
	trace_nfs_create_exit(dir, dentry, open_flags, error);
L
Linus Torvalds 已提交
1665 1666 1667 1668 1669 1670 1671
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1672
EXPORT_SYMBOL_GPL(nfs_create);
L
Linus Torvalds 已提交
1673 1674 1675 1676

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1677
int
A
Al Viro 已提交
1678
nfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t rdev)
L
Linus Torvalds 已提交
1679 1680 1681 1682
{
	struct iattr attr;
	int status;

1683
	dfprintk(VFS, "NFS: mknod(%s/%lu), %pd\n",
1684
			dir->i_sb->s_id, dir->i_ino, dentry);
L
Linus Torvalds 已提交
1685 1686 1687 1688 1689 1690 1691

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

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

1692
	trace_nfs_mknod_enter(dir, dentry);
L
Linus Torvalds 已提交
1693
	status = NFS_PROTO(dir)->mknod(dir, dentry, &attr, rdev);
1694
	trace_nfs_mknod_exit(dir, dentry, status);
L
Linus Torvalds 已提交
1695 1696 1697 1698 1699 1700 1701
	if (status != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return status;
}
B
Bryan Schumaker 已提交
1702
EXPORT_SYMBOL_GPL(nfs_mknod);
L
Linus Torvalds 已提交
1703 1704 1705 1706

/*
 * See comments for nfs_proc_create regarding failed operations.
 */
1707
int nfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
1708 1709 1710 1711
{
	struct iattr attr;
	int error;

1712
	dfprintk(VFS, "NFS: mkdir(%s/%lu), %pd\n",
1713
			dir->i_sb->s_id, dir->i_ino, dentry);
L
Linus Torvalds 已提交
1714 1715 1716 1717

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

1718
	trace_nfs_mkdir_enter(dir, dentry);
L
Linus Torvalds 已提交
1719
	error = NFS_PROTO(dir)->mkdir(dir, dentry, &attr);
1720
	trace_nfs_mkdir_exit(dir, dentry, error);
L
Linus Torvalds 已提交
1721 1722 1723 1724 1725 1726 1727
	if (error != 0)
		goto out_err;
	return 0;
out_err:
	d_drop(dentry);
	return error;
}
B
Bryan Schumaker 已提交
1728
EXPORT_SYMBOL_GPL(nfs_mkdir);
L
Linus Torvalds 已提交
1729

1730 1731 1732 1733 1734 1735
static void nfs_dentry_handle_enoent(struct dentry *dentry)
{
	if (dentry->d_inode != NULL && !d_unhashed(dentry))
		d_delete(dentry);
}

1736
int nfs_rmdir(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1737 1738 1739
{
	int error;

1740
	dfprintk(VFS, "NFS: rmdir(%s/%lu), %pd\n",
1741
			dir->i_sb->s_id, dir->i_ino, dentry);
L
Linus Torvalds 已提交
1742

1743
	trace_nfs_rmdir_enter(dir, dentry);
1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756
	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);
1757
	trace_nfs_rmdir_exit(dir, dentry, error);
L
Linus Torvalds 已提交
1758 1759 1760

	return error;
}
B
Bryan Schumaker 已提交
1761
EXPORT_SYMBOL_GPL(nfs_rmdir);
L
Linus Torvalds 已提交
1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775

/*
 * 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;
		
1776
	dfprintk(VFS, "NFS: safe_remove(%pd2)\n", dentry);
L
Linus Torvalds 已提交
1777 1778 1779 1780 1781 1782 1783

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

1784
	trace_nfs_remove_enter(dir, dentry);
L
Linus Torvalds 已提交
1785
	if (inode != NULL) {
1786
		NFS_PROTO(inode)->return_delegation(inode);
L
Linus Torvalds 已提交
1787 1788
		error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
		if (error == 0)
1789
			nfs_drop_nlink(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);
1794
	trace_nfs_remove_exit(dir, dentry, error);
L
Linus Torvalds 已提交
1795 1796 1797 1798 1799 1800 1801 1802 1803
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.
 */
1804
int nfs_unlink(struct inode *dir, struct dentry *dentry)
L
Linus Torvalds 已提交
1805 1806 1807 1808
{
	int error;
	int need_rehash = 0;

1809
	dfprintk(VFS, "NFS: unlink(%s/%lu, %pd)\n", dir->i_sb->s_id,
1810
		dir->i_ino, dentry);
L
Linus Torvalds 已提交
1811

1812
	trace_nfs_unlink_enter(dir, dentry);
L
Linus Torvalds 已提交
1813
	spin_lock(&dentry->d_lock);
A
Al Viro 已提交
1814
	if (d_count(dentry) > 1) {
L
Linus Torvalds 已提交
1815
		spin_unlock(&dentry->d_lock);
T
Trond Myklebust 已提交
1816 1817
		/* Start asynchronous writeout of the inode */
		write_inode_now(dentry->d_inode, 0);
L
Linus Torvalds 已提交
1818
		error = nfs_sillyrename(dir, dentry);
1819
		goto out;
L
Linus Torvalds 已提交
1820 1821 1822 1823 1824 1825 1826
	}
	if (!d_unhashed(dentry)) {
		__d_drop(dentry);
		need_rehash = 1;
	}
	spin_unlock(&dentry->d_lock);
	error = nfs_safe_remove(dentry);
1827
	if (!error || error == -ENOENT) {
L
Linus Torvalds 已提交
1828 1829 1830
		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
	} else if (need_rehash)
		d_rehash(dentry);
1831 1832
out:
	trace_nfs_unlink_exit(dir, dentry, error);
L
Linus Torvalds 已提交
1833 1834
	return error;
}
B
Bryan Schumaker 已提交
1835
EXPORT_SYMBOL_GPL(nfs_unlink);
L
Linus Torvalds 已提交
1836

1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851
/*
 * 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.
 */
1852
int nfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
L
Linus Torvalds 已提交
1853
{
1854 1855
	struct page *page;
	char *kaddr;
L
Linus Torvalds 已提交
1856
	struct iattr attr;
1857
	unsigned int pathlen = strlen(symname);
L
Linus Torvalds 已提交
1858 1859
	int error;

1860
	dfprintk(VFS, "NFS: symlink(%s/%lu, %pd, %s)\n", dir->i_sb->s_id,
1861
		dir->i_ino, dentry, symname);
L
Linus Torvalds 已提交
1862

1863 1864
	if (pathlen > PAGE_SIZE)
		return -ENAMETOOLONG;
L
Linus Torvalds 已提交
1865

1866 1867
	attr.ia_mode = S_IFLNK | S_IRWXUGO;
	attr.ia_valid = ATTR_MODE;
L
Linus Torvalds 已提交
1868

1869
	page = alloc_page(GFP_HIGHUSER);
1870
	if (!page)
1871 1872
		return -ENOMEM;

1873
	kaddr = kmap_atomic(page);
1874 1875 1876
	memcpy(kaddr, symname, pathlen);
	if (pathlen < PAGE_SIZE)
		memset(kaddr + pathlen, 0, PAGE_SIZE - pathlen);
1877
	kunmap_atomic(kaddr);
1878

1879
	trace_nfs_symlink_enter(dir, dentry);
1880
	error = NFS_PROTO(dir)->symlink(dir, dentry, page, pathlen, &attr);
1881
	trace_nfs_symlink_exit(dir, dentry, error);
1882
	if (error != 0) {
1883
		dfprintk(VFS, "NFS: symlink(%s/%lu, %pd, %s) error %d\n",
1884
			dir->i_sb->s_id, dir->i_ino,
1885
			dentry, symname, error);
L
Linus Torvalds 已提交
1886
		d_drop(dentry);
1887 1888 1889 1890 1891 1892 1893 1894
		__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.
	 */
1895
	if (!add_to_page_cache_lru(page, dentry->d_inode->i_mapping, 0,
1896 1897 1898
							GFP_KERNEL)) {
		SetPageUptodate(page);
		unlock_page(page);
1899 1900 1901 1902 1903
		/*
		 * add_to_page_cache_lru() grabs an extra page refcount.
		 * Drop it here to avoid leaking this page later.
		 */
		page_cache_release(page);
1904 1905 1906 1907
	} else
		__free_page(page);

	return 0;
L
Linus Torvalds 已提交
1908
}
B
Bryan Schumaker 已提交
1909
EXPORT_SYMBOL_GPL(nfs_symlink);
L
Linus Torvalds 已提交
1910

1911
int
L
Linus Torvalds 已提交
1912 1913 1914 1915 1916
nfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = old_dentry->d_inode;
	int error;

1917 1918
	dfprintk(VFS, "NFS: link(%pd2 -> %pd2)\n",
		old_dentry, dentry);
L
Linus Torvalds 已提交
1919

1920
	trace_nfs_link_enter(inode, dir, dentry);
1921
	NFS_PROTO(inode)->return_delegation(inode);
1922

1923
	d_drop(dentry);
L
Linus Torvalds 已提交
1924
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
1925
	if (error == 0) {
A
Al Viro 已提交
1926
		ihold(inode);
1927
		d_add(dentry, inode);
1928
	}
1929
	trace_nfs_link_exit(inode, dir, dentry, error);
L
Linus Torvalds 已提交
1930 1931
	return error;
}
B
Bryan Schumaker 已提交
1932
EXPORT_SYMBOL_GPL(nfs_link);
L
Linus Torvalds 已提交
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

/*
 * 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.
 */
1958
int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
L
Linus Torvalds 已提交
1959 1960 1961 1962 1963
		      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;
1964
	struct rpc_task *task;
L
Linus Torvalds 已提交
1965 1966
	int error = -EBUSY;

1967 1968
	dfprintk(VFS, "NFS: rename(%pd2 -> %pd2, ct=%d)\n",
		 old_dentry, new_dentry,
A
Al Viro 已提交
1969
		 d_count(new_dentry));
L
Linus Torvalds 已提交
1970

1971
	trace_nfs_rename_enter(old_dir, old_dentry, new_dir, new_dentry);
L
Linus Torvalds 已提交
1972
	/*
M
Miklos Szeredi 已提交
1973 1974 1975 1976
	 * 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
1977
	 */
1978 1979 1980 1981 1982 1983 1984 1985 1986
	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
1987

A
Al Viro 已提交
1988
		if (d_count(new_dentry) > 2) {
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
			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);
1999
			if (err)
2000
				goto out;
2001 2002

			new_dentry = dentry;
O
OGAWA Hirofumi 已提交
2003
			rehash = NULL;
2004
			new_inode = NULL;
2005
		}
2006
	}
L
Linus Torvalds 已提交
2007

2008
	NFS_PROTO(old_inode)->return_delegation(old_inode);
2009
	if (new_inode != NULL)
2010
		NFS_PROTO(new_inode)->return_delegation(new_inode);
L
Linus Torvalds 已提交
2011

2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
	task = nfs_async_rename(old_dir, new_dir, old_dentry, new_dentry, NULL);
	if (IS_ERR(task)) {
		error = PTR_ERR(task);
		goto out;
	}

	error = rpc_wait_for_completion_task(task);
	if (error == 0)
		error = task->tk_status;
	rpc_put_task(task);
2022
	nfs_mark_for_revalidate(old_inode);
L
Linus Torvalds 已提交
2023 2024 2025
out:
	if (rehash)
		d_rehash(rehash);
2026 2027
	trace_nfs_rename_exit(old_dir, old_dentry,
			new_dir, new_dentry, error);
L
Linus Torvalds 已提交
2028
	if (!error) {
2029 2030
		if (new_inode != NULL)
			nfs_drop_nlink(new_inode);
2031
		d_move(old_dentry, new_dentry);
2032 2033
		nfs_set_verifier(new_dentry,
					nfs_save_change_attribute(new_dir));
2034 2035
	} else if (error == -ENOENT)
		nfs_dentry_handle_enoent(old_dentry);
L
Linus Torvalds 已提交
2036 2037 2038 2039 2040 2041

	/* new dentry created? */
	if (dentry)
		dput(dentry);
	return error;
}
B
Bryan Schumaker 已提交
2042
EXPORT_SYMBOL_GPL(nfs_rename);
L
Linus Torvalds 已提交
2043

2044 2045 2046 2047
static DEFINE_SPINLOCK(nfs_access_lru_lock);
static LIST_HEAD(nfs_access_lru_list);
static atomic_long_t nfs_access_nr_entries;

2048 2049 2050 2051
static unsigned long nfs_access_max_cachesize = ULONG_MAX;
module_param(nfs_access_max_cachesize, ulong, 0644);
MODULE_PARM_DESC(nfs_access_max_cachesize, "NFS access maximum total cache length");

2052 2053 2054 2055
static void nfs_access_free_entry(struct nfs_access_entry *entry)
{
	put_rpccred(entry->cred);
	kfree(entry);
2056
	smp_mb__before_atomic();
2057
	atomic_long_dec(&nfs_access_nr_entries);
2058
	smp_mb__after_atomic();
2059 2060
}

2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071
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);
	}
}

2072 2073
static unsigned long
nfs_do_access_cache_scan(unsigned int nr_to_scan)
2074 2075
{
	LIST_HEAD(head);
2076
	struct nfs_inode *nfsi, *next;
2077
	struct nfs_access_entry *cache;
2078
	long freed = 0;
2079

2080
	spin_lock(&nfs_access_lru_lock);
2081
	list_for_each_entry_safe(nfsi, next, &nfs_access_lru_list, access_cache_inode_lru) {
2082 2083 2084 2085
		struct inode *inode;

		if (nr_to_scan-- == 0)
			break;
2086
		inode = &nfsi->vfs_inode;
2087 2088 2089 2090 2091 2092 2093
		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);
2094
		freed++;
2095 2096 2097 2098 2099 2100
		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);
2101
			smp_mb__before_atomic();
2102
			clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
2103
			smp_mb__after_atomic();
2104
		}
2105
		spin_unlock(&inode->i_lock);
2106 2107
	}
	spin_unlock(&nfs_access_lru_lock);
2108
	nfs_access_free_list(&head);
2109 2110 2111
	return freed;
}

2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123
unsigned long
nfs_access_cache_scan(struct shrinker *shrink, struct shrink_control *sc)
{
	int nr_to_scan = sc->nr_to_scan;
	gfp_t gfp_mask = sc->gfp_mask;

	if ((gfp_mask & GFP_KERNEL) != GFP_KERNEL)
		return SHRINK_STOP;
	return nfs_do_access_cache_scan(nr_to_scan);
}


2124 2125 2126
unsigned long
nfs_access_cache_count(struct shrinker *shrink, struct shrink_control *sc)
{
2127
	return vfs_pressure_ratio(atomic_long_read(&nfs_access_nr_entries));
2128 2129
}

2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145
static void
nfs_access_cache_enforce_limit(void)
{
	long nr_entries = atomic_long_read(&nfs_access_nr_entries);
	unsigned long diff;
	unsigned int nr_to_scan;

	if (nr_entries < 0 || nr_entries <= nfs_access_max_cachesize)
		return;
	nr_to_scan = 100;
	diff = nr_entries - nfs_access_max_cachesize;
	if (diff < nr_to_scan)
		nr_to_scan = diff;
	nfs_do_access_cache_scan(nr_to_scan);
}

2146
static void __nfs_access_zap_cache(struct nfs_inode *nfsi, struct list_head *head)
L
Linus Torvalds 已提交
2147
{
2148
	struct rb_root *root_node = &nfsi->access_cache;
2149
	struct rb_node *n;
2150 2151 2152 2153 2154 2155
	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);
2156
		list_move(&entry->lru, head);
2157 2158
	}
	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
L
Linus Torvalds 已提交
2159 2160
}

2161
void nfs_access_zap_cache(struct inode *inode)
L
Linus Torvalds 已提交
2162
{
2163 2164 2165 2166
	LIST_HEAD(head);

	if (test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags) == 0)
		return;
2167
	/* Remove from global LRU init */
2168 2169
	spin_lock(&nfs_access_lru_lock);
	if (test_and_clear_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags))
2170 2171
		list_del_init(&NFS_I(inode)->access_cache_inode_lru);

2172
	spin_lock(&inode->i_lock);
2173 2174 2175 2176
	__nfs_access_zap_cache(NFS_I(inode), &head);
	spin_unlock(&inode->i_lock);
	spin_unlock(&nfs_access_lru_lock);
	nfs_access_free_list(&head);
2177
}
B
Bryan Schumaker 已提交
2178
EXPORT_SYMBOL_GPL(nfs_access_zap_cache);
L
Linus Torvalds 已提交
2179

2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193
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
2194
	}
2195 2196 2197
	return NULL;
}

2198
static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
2199 2200 2201 2202 2203
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_access_entry *cache;
	int err = -ENOENT;

2204
	spin_lock(&inode->i_lock);
2205 2206 2207 2208 2209
	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
		goto out_zap;
	cache = nfs_access_search_rbtree(inode, cred);
	if (cache == NULL)
		goto out;
2210
	if (!nfs_have_delegated_attributes(inode) &&
2211
	    !time_in_range_open(jiffies, cache->jiffies, cache->jiffies + nfsi->attrtimeo))
2212 2213 2214 2215
		goto out_stale;
	res->jiffies = cache->jiffies;
	res->cred = cache->cred;
	res->mask = cache->mask;
2216
	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
2217 2218 2219 2220 2221 2222
	err = 0;
out:
	spin_unlock(&inode->i_lock);
	return err;
out_stale:
	rb_erase(&cache->rb_node, &nfsi->access_cache);
2223
	list_del(&cache->lru);
2224 2225 2226 2227
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(cache);
	return -ENOENT;
out_zap:
2228 2229
	spin_unlock(&inode->i_lock);
	nfs_access_zap_cache(inode);
2230 2231 2232 2233 2234
	return -ENOENT;
}

static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
{
2235 2236
	struct nfs_inode *nfsi = NFS_I(inode);
	struct rb_root *root_node = &nfsi->access_cache;
2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254
	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);
2255
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
2256
	spin_unlock(&inode->i_lock);
2257 2258 2259
	return;
found:
	rb_replace_node(parent, &set->rb_node, root_node);
2260 2261
	list_add_tail(&set->lru, &nfsi->access_cache_entry_lru);
	list_del(&entry->lru);
2262 2263 2264 2265
	spin_unlock(&inode->i_lock);
	nfs_access_free_entry(entry);
}

2266
void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
2267 2268 2269 2270 2271
{
	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
2272
	cache->jiffies = set->jiffies;
2273
	cache->cred = get_rpccred(set->cred);
L
Linus Torvalds 已提交
2274
	cache->mask = set->mask;
2275 2276

	nfs_access_add_rbtree(inode, cache);
2277 2278

	/* Update accounting */
2279
	smp_mb__before_atomic();
2280
	atomic_long_inc(&nfs_access_nr_entries);
2281
	smp_mb__after_atomic();
2282 2283

	/* Add inode to global LRU list */
2284
	if (!test_bit(NFS_INO_ACL_LRU_SET, &NFS_I(inode)->flags)) {
2285
		spin_lock(&nfs_access_lru_lock);
2286 2287 2288
		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);
2289 2290
		spin_unlock(&nfs_access_lru_lock);
	}
2291
	nfs_access_cache_enforce_limit();
L
Linus Torvalds 已提交
2292
}
2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
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
2307 2308 2309 2310 2311 2312

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

2313 2314
	trace_nfs_access_enter(inode);

L
Linus Torvalds 已提交
2315 2316
	status = nfs_access_get_cached(inode, cred, &cache);
	if (status == 0)
2317
		goto out_cached;
L
Linus Torvalds 已提交
2318

2319 2320 2321 2322
	status = -ECHILD;
	if (mask & MAY_NOT_BLOCK)
		goto out;

L
Linus Torvalds 已提交
2323 2324 2325 2326 2327
	/* 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);
2328 2329 2330 2331 2332 2333
	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);
		}
2334
		goto out;
2335
	}
L
Linus Torvalds 已提交
2336
	nfs_access_add_cache(inode, &cache);
2337 2338 2339
out_cached:
	if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) != 0)
		status = -EACCES;
L
Linus Torvalds 已提交
2340
out:
2341 2342
	trace_nfs_access_exit(inode, status);
	return status;
L
Linus Torvalds 已提交
2343 2344
}

2345 2346 2347 2348
static int nfs_open_permission_mask(int openflags)
{
	int mask = 0;

2349 2350 2351 2352 2353 2354 2355 2356 2357 2358
	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;
	}

2359 2360 2361 2362 2363 2364 2365
	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
2366
EXPORT_SYMBOL_GPL(nfs_may_open);
2367

2368
int nfs_permission(struct inode *inode, int mask)
L
Linus Torvalds 已提交
2369 2370 2371 2372
{
	struct rpc_cred *cred;
	int res = 0;

C
Chuck Lever 已提交
2373 2374
	nfs_inc_stats(inode, NFSIOS_VFSACCESS);

2375
	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
L
Linus Torvalds 已提交
2376 2377
		goto out;
	/* Is this sys_access() ? */
E
Eric Paris 已提交
2378
	if (mask & (MAY_ACCESS | MAY_CHDIR))
L
Linus Torvalds 已提交
2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398
		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;

2399 2400 2401 2402 2403 2404
	/* Always try fast lookups first */
	rcu_read_lock();
	cred = rpc_lookup_cred_nonblock();
	if (!IS_ERR(cred))
		res = nfs_do_access(inode, cred, mask|MAY_NOT_BLOCK);
	else
L
Linus Torvalds 已提交
2405
		res = PTR_ERR(cred);
2406 2407 2408 2409 2410 2411 2412 2413 2414 2415
	rcu_read_unlock();
	if (res == -ECHILD && !(mask & MAY_NOT_BLOCK)) {
		/* Fast lookup failed, try the slow way */
		cred = rpc_lookup_cred();
		if (!IS_ERR(cred)) {
			res = nfs_do_access(inode, cred, mask);
			put_rpccred(cred);
		} else
			res = PTR_ERR(cred);
	}
L
Linus Torvalds 已提交
2416
out:
2417 2418 2419
	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
		res = -EACCES;

2420
	dfprintk(VFS, "NFS: permission(%s/%lu), mask=0x%x, res=%d\n",
C
Chuck Lever 已提交
2421
		inode->i_sb->s_id, inode->i_ino, mask, res);
L
Linus Torvalds 已提交
2422 2423
	return res;
out_notsup:
2424 2425 2426
	if (mask & MAY_NOT_BLOCK)
		return -ECHILD;

L
Linus Torvalds 已提交
2427 2428
	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
	if (res == 0)
2429
		res = generic_permission(inode, mask);
C
Chuck Lever 已提交
2430
	goto out;
L
Linus Torvalds 已提交
2431
}
B
Bryan Schumaker 已提交
2432
EXPORT_SYMBOL_GPL(nfs_permission);
L
Linus Torvalds 已提交
2433 2434 2435 2436 2437 2438 2439

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