file.c 21.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
/*
 *  linux/fs/nfs/file.c
 *
 *  Copyright (C) 1992  Rick Sladkey
 *
 *  Changes Copyright (C) 1994 by Florian La Roche
 *   - Do not copy data too often around in the kernel.
 *   - In nfs_file_read the return value of kmalloc wasn't checked.
 *   - Put in a better version of read look-ahead buffering. Original idea
 *     and implementation by Wai S Kok elekokws@ee.nus.sg.
 *
 *  Expire cache on write to a file by Wai S Kok (Oct 1994).
 *
 *  Total rewrite of read side for new NFS buffer cache.. Linus.
 *
 *  nfs regular file handling functions
 */

#include <linux/time.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/fcntl.h>
#include <linux/stat.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/pagemap.h>
A
Alexey Dobriyan 已提交
29
#include <linux/aio.h>
L
Linus Torvalds 已提交
30 31 32 33 34

#include <asm/uaccess.h>
#include <asm/system.h>

#include "delegation.h"
35
#include "internal.h"
C
Chuck Lever 已提交
36
#include "iostat.h"
D
David Howells 已提交
37
#include "fscache.h"
L
Linus Torvalds 已提交
38 39 40 41 42

#define NFSDBG_FACILITY		NFSDBG_FILE

static int nfs_file_open(struct inode *, struct file *);
static int nfs_file_release(struct inode *, struct file *);
43
static loff_t nfs_file_llseek(struct file *file, loff_t offset, int origin);
L
Linus Torvalds 已提交
44
static int  nfs_file_mmap(struct file *, struct vm_area_struct *);
45 46 47
static ssize_t nfs_file_splice_read(struct file *filp, loff_t *ppos,
					struct pipe_inode_info *pipe,
					size_t count, unsigned int flags);
48 49
static ssize_t nfs_file_read(struct kiocb *, const struct iovec *iov,
				unsigned long nr_segs, loff_t pos);
50 51 52
static ssize_t nfs_file_splice_write(struct pipe_inode_info *pipe,
					struct file *filp, loff_t *ppos,
					size_t count, unsigned int flags);
53 54
static ssize_t nfs_file_write(struct kiocb *, const struct iovec *iov,
				unsigned long nr_segs, loff_t pos);
55
static int  nfs_file_flush(struct file *, fl_owner_t id);
56
static int  nfs_file_fsync(struct file *, struct dentry *dentry, int datasync);
L
Linus Torvalds 已提交
57 58 59
static int nfs_check_flags(int flags);
static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl);
static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl);
J
J. Bruce Fields 已提交
60
static int nfs_setlease(struct file *file, long arg, struct file_lock **fl);
L
Linus Torvalds 已提交
61

62 63
static struct vm_operations_struct nfs_file_vm_ops;

64
const struct file_operations nfs_file_operations = {
65
	.llseek		= nfs_file_llseek,
L
Linus Torvalds 已提交
66 67
	.read		= do_sync_read,
	.write		= do_sync_write,
68 69
	.aio_read	= nfs_file_read,
	.aio_write	= nfs_file_write,
L
Linus Torvalds 已提交
70 71 72 73
	.mmap		= nfs_file_mmap,
	.open		= nfs_file_open,
	.flush		= nfs_file_flush,
	.release	= nfs_file_release,
74
	.fsync		= nfs_file_fsync,
L
Linus Torvalds 已提交
75 76
	.lock		= nfs_lock,
	.flock		= nfs_flock,
77
	.splice_read	= nfs_file_splice_read,
78
	.splice_write	= nfs_file_splice_write,
L
Linus Torvalds 已提交
79
	.check_flags	= nfs_check_flags,
J
J. Bruce Fields 已提交
80
	.setlease	= nfs_setlease,
L
Linus Torvalds 已提交
81 82
};

83
const struct inode_operations nfs_file_inode_operations = {
L
Linus Torvalds 已提交
84 85 86 87 88
	.permission	= nfs_permission,
	.getattr	= nfs_getattr,
	.setattr	= nfs_setattr,
};

89
#ifdef CONFIG_NFS_V3
90
const struct inode_operations nfs3_file_inode_operations = {
91 92 93 94 95 96 97 98 99 100
	.permission	= nfs_permission,
	.getattr	= nfs_getattr,
	.setattr	= nfs_setattr,
	.listxattr	= nfs3_listxattr,
	.getxattr	= nfs3_getxattr,
	.setxattr	= nfs3_setxattr,
	.removexattr	= nfs3_removexattr,
};
#endif  /* CONFIG_NFS_v3 */

L
Linus Torvalds 已提交
101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
/* Hack for future NFS swap support */
#ifndef IS_SWAPFILE
# define IS_SWAPFILE(inode)	(0)
#endif

static int nfs_check_flags(int flags)
{
	if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
		return -EINVAL;

	return 0;
}

/*
 * Open file
 */
static int
nfs_file_open(struct inode *inode, struct file *filp)
{
	int res;

C
Chuck Lever 已提交
122
	dprintk("NFS: open file(%s/%s)\n",
123 124 125
			filp->f_path.dentry->d_parent->d_name.name,
			filp->f_path.dentry->d_name.name);

L
Linus Torvalds 已提交
126 127 128 129
	res = nfs_check_flags(filp->f_flags);
	if (res)
		return res;

C
Chuck Lever 已提交
130
	nfs_inc_stats(inode, NFSIOS_VFSOPEN);
131
	res = nfs_open(inode, filp);
L
Linus Torvalds 已提交
132 133 134 135 136 137
	return res;
}

static int
nfs_file_release(struct inode *inode, struct file *filp)
{
C
Chuck Lever 已提交
138 139 140 141 142 143
	struct dentry *dentry = filp->f_path.dentry;

	dprintk("NFS: release(%s/%s)\n",
			dentry->d_parent->d_name.name,
			dentry->d_name.name);

C
Chuck Lever 已提交
144
	nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
145
	return nfs_release(inode, filp);
L
Linus Torvalds 已提交
146 147
}

148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
/**
 * nfs_revalidate_size - Revalidate the file size
 * @inode - pointer to inode struct
 * @file - pointer to struct file
 *
 * Revalidates the file length. This is basically a wrapper around
 * nfs_revalidate_inode() that takes into account the fact that we may
 * have cached writes (in which case we don't care about the server's
 * idea of what the file length is), or O_DIRECT (in which case we
 * shouldn't trust the cache).
 */
static int nfs_revalidate_file_size(struct inode *inode, struct file *filp)
{
	struct nfs_server *server = NFS_SERVER(inode);
	struct nfs_inode *nfsi = NFS_I(inode);

	if (server->flags & NFS_MOUNT_NOAC)
		goto force_reval;
	if (filp->f_flags & O_DIRECT)
		goto force_reval;
	if (nfsi->npages != 0)
		return 0;
170
	if (!(nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE) && !nfs_attribute_timeout(inode))
171
		return 0;
172 173 174 175 176 177
force_reval:
	return __nfs_revalidate_inode(server, inode);
}

static loff_t nfs_file_llseek(struct file *filp, loff_t offset, int origin)
{
A
Andi Kleen 已提交
178
	loff_t loff;
T
Trond Myklebust 已提交
179

C
Chuck Lever 已提交
180
	dprintk("NFS: llseek file(%s/%s, %lld, %d)\n",
181 182 183 184
			filp->f_path.dentry->d_parent->d_name.name,
			filp->f_path.dentry->d_name.name,
			offset, origin);

185
	/* origin == SEEK_END => we must revalidate the cached file length */
186
	if (origin == SEEK_END) {
187
		struct inode *inode = filp->f_mapping->host;
T
Trond Myklebust 已提交
188

189 190 191
		int retval = nfs_revalidate_file_size(inode, filp);
		if (retval < 0)
			return (loff_t)retval;
T
Trond Myklebust 已提交
192 193 194 195 196 197

		spin_lock(&inode->i_lock);
		loff = generic_file_llseek_unlocked(filp, offset, origin);
		spin_unlock(&inode->i_lock);
	} else
		loff = generic_file_llseek_unlocked(filp, offset, origin);
A
Andi Kleen 已提交
198
	return loff;
199 200
}

201
/*
202
 * Helper for nfs_file_flush() and nfs_file_fsync()
203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
 *
 * Notice that it clears the NFS_CONTEXT_ERROR_WRITE before synching to
 * disk, but it retrieves and clears ctx->error after synching, despite
 * the two being set at the same time in nfs_context_set_write_error().
 * This is because the former is used to notify the _next_ call to
 * nfs_file_write() that a write error occured, and hence cause it to
 * fall back to doing a synchronous write.
 */
static int nfs_do_fsync(struct nfs_open_context *ctx, struct inode *inode)
{
	int have_error, status;
	int ret = 0;

	have_error = test_and_clear_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
	status = nfs_wb_all(inode);
	have_error |= test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
	if (have_error)
		ret = xchg(&ctx->error, 0);
	if (!ret)
		ret = status;
	return ret;
}

L
Linus Torvalds 已提交
226 227 228 229
/*
 * Flush all dirty pages, and check for write errors.
 */
static int
230
nfs_file_flush(struct file *file, fl_owner_t id)
L
Linus Torvalds 已提交
231
{
232
	struct nfs_open_context *ctx = nfs_file_open_context(file);
C
Chuck Lever 已提交
233 234
	struct dentry	*dentry = file->f_path.dentry;
	struct inode	*inode = dentry->d_inode;
L
Linus Torvalds 已提交
235

C
Chuck Lever 已提交
236 237 238
	dprintk("NFS: flush(%s/%s)\n",
			dentry->d_parent->d_name.name,
			dentry->d_name.name);
L
Linus Torvalds 已提交
239 240 241

	if ((file->f_mode & FMODE_WRITE) == 0)
		return 0;
C
Chuck Lever 已提交
242
	nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
243

T
Trond Myklebust 已提交
244 245
	/* Flush writes to the server and return any errors */
	return nfs_do_fsync(ctx, inode);
L
Linus Torvalds 已提交
246 247 248
}

static ssize_t
249 250
nfs_file_read(struct kiocb *iocb, const struct iovec *iov,
		unsigned long nr_segs, loff_t pos)
L
Linus Torvalds 已提交
251
{
252
	struct dentry * dentry = iocb->ki_filp->f_path.dentry;
L
Linus Torvalds 已提交
253 254
	struct inode * inode = dentry->d_inode;
	ssize_t result;
255
	size_t count = iov_length(iov, nr_segs);
L
Linus Torvalds 已提交
256 257

	if (iocb->ki_filp->f_flags & O_DIRECT)
258
		return nfs_file_direct_read(iocb, iov, nr_segs, pos);
L
Linus Torvalds 已提交
259

C
Chuck Lever 已提交
260
	dprintk("NFS: read(%s/%s, %lu@%lu)\n",
L
Linus Torvalds 已提交
261 262 263
		dentry->d_parent->d_name.name, dentry->d_name.name,
		(unsigned long) count, (unsigned long) pos);

264
	result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
C
Chuck Lever 已提交
265
	nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, count);
L
Linus Torvalds 已提交
266
	if (!result)
267
		result = generic_file_aio_read(iocb, iov, nr_segs, pos);
L
Linus Torvalds 已提交
268 269 270 271
	return result;
}

static ssize_t
272 273 274
nfs_file_splice_read(struct file *filp, loff_t *ppos,
		     struct pipe_inode_info *pipe, size_t count,
		     unsigned int flags)
L
Linus Torvalds 已提交
275
{
276
	struct dentry *dentry = filp->f_path.dentry;
L
Linus Torvalds 已提交
277 278 279
	struct inode *inode = dentry->d_inode;
	ssize_t res;

C
Chuck Lever 已提交
280
	dprintk("NFS: splice_read(%s/%s, %lu@%Lu)\n",
L
Linus Torvalds 已提交
281 282 283
		dentry->d_parent->d_name.name, dentry->d_name.name,
		(unsigned long) count, (unsigned long long) *ppos);

284
	res = nfs_revalidate_mapping(inode, filp->f_mapping);
L
Linus Torvalds 已提交
285
	if (!res)
286
		res = generic_file_splice_read(filp, ppos, pipe, count, flags);
L
Linus Torvalds 已提交
287 288 289 290 291 292
	return res;
}

static int
nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
{
293
	struct dentry *dentry = file->f_path.dentry;
L
Linus Torvalds 已提交
294 295 296
	struct inode *inode = dentry->d_inode;
	int	status;

C
Chuck Lever 已提交
297
	dprintk("NFS: mmap(%s/%s)\n",
L
Linus Torvalds 已提交
298 299
		dentry->d_parent->d_name.name, dentry->d_name.name);

300 301 302 303
	/* Note: generic_file_mmap() returns ENOSYS on nommu systems
	 *       so we call that before revalidating the mapping
	 */
	status = generic_file_mmap(file, vma);
304 305
	if (!status) {
		vma->vm_ops = &nfs_file_vm_ops;
306
		status = nfs_revalidate_mapping(inode, file->f_mapping);
307
	}
L
Linus Torvalds 已提交
308 309 310 311 312 313 314 315 316
	return status;
}

/*
 * Flush any dirty pages for this process, and check for write errors.
 * The return status from this call provides a reliable indication of
 * whether any write errors occurred for this process.
 */
static int
317
nfs_file_fsync(struct file *file, struct dentry *dentry, int datasync)
L
Linus Torvalds 已提交
318
{
319
	struct nfs_open_context *ctx = nfs_file_open_context(file);
L
Linus Torvalds 已提交
320 321
	struct inode *inode = dentry->d_inode;

C
Chuck Lever 已提交
322
	dprintk("NFS: fsync file(%s/%s) datasync %d\n",
323 324
			dentry->d_parent->d_name.name, dentry->d_name.name,
			datasync);
L
Linus Torvalds 已提交
325

C
Chuck Lever 已提交
326
	nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
327
	return nfs_do_fsync(ctx, inode);
L
Linus Torvalds 已提交
328 329 330
}

/*
N
Nick Piggin 已提交
331 332 333
 * This does the "real" work of the write. We must allocate and lock the
 * page to be sent back to the generic routine, which then copies the
 * data from user space.
L
Linus Torvalds 已提交
334 335 336 337
 *
 * If the writer ends up delaying the write, the writer needs to
 * increment the page use counts until he is done with the page.
 */
N
Nick Piggin 已提交
338 339 340
static int nfs_write_begin(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned flags,
			struct page **pagep, void **fsdata)
L
Linus Torvalds 已提交
341
{
N
Nick Piggin 已提交
342 343 344 345 346
	int ret;
	pgoff_t index;
	struct page *page;
	index = pos >> PAGE_CACHE_SHIFT;

347 348 349 350 351
	dfprintk(PAGECACHE, "NFS: write_begin(%s/%s(%ld), %u@%lld)\n",
		file->f_path.dentry->d_parent->d_name.name,
		file->f_path.dentry->d_name.name,
		mapping->host->i_ino, len, (long long) pos);

352 353 354 355 356 357 358 359 360
	/*
	 * Prevent starvation issues if someone is doing a consistency
	 * sync-to-disk
	 */
	ret = wait_on_bit(&NFS_I(mapping->host)->flags, NFS_INO_FLUSHING,
			nfs_wait_bit_killable, TASK_KILLABLE);
	if (ret)
		return ret;

361
	page = grab_cache_page_write_begin(mapping, index, flags);
N
Nick Piggin 已提交
362 363 364 365 366 367 368 369 370 371
	if (!page)
		return -ENOMEM;
	*pagep = page;

	ret = nfs_flush_incompatible(file, page);
	if (ret) {
		unlock_page(page);
		page_cache_release(page);
	}
	return ret;
L
Linus Torvalds 已提交
372 373
}

N
Nick Piggin 已提交
374 375 376
static int nfs_write_end(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned copied,
			struct page *page, void *fsdata)
L
Linus Torvalds 已提交
377
{
N
Nick Piggin 已提交
378 379
	unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
	int status;
L
Linus Torvalds 已提交
380

381 382 383 384 385
	dfprintk(PAGECACHE, "NFS: write_end(%s/%s(%ld), %u@%lld)\n",
		file->f_path.dentry->d_parent->d_name.name,
		file->f_path.dentry->d_name.name,
		mapping->host->i_ino, len, (long long) pos);

386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
	/*
	 * Zero any uninitialised parts of the page, and then mark the page
	 * as up to date if it turns out that we're extending the file.
	 */
	if (!PageUptodate(page)) {
		unsigned pglen = nfs_page_length(page);
		unsigned end = offset + len;

		if (pglen == 0) {
			zero_user_segments(page, 0, offset,
					end, PAGE_CACHE_SIZE);
			SetPageUptodate(page);
		} else if (end >= pglen) {
			zero_user_segment(page, end, PAGE_CACHE_SIZE);
			if (offset == 0)
				SetPageUptodate(page);
		} else
			zero_user_segment(page, pglen, PAGE_CACHE_SIZE);
	}

N
Nick Piggin 已提交
406 407 408 409 410
	status = nfs_updatepage(file, page, offset, copied);

	unlock_page(page);
	page_cache_release(page);

C
Chuck Lever 已提交
411 412 413
	if (status < 0)
		return status;
	return copied;
L
Linus Torvalds 已提交
414 415
}

416 417 418 419
/*
 * Partially or wholly invalidate a page
 * - Release the private state associated with a page if undergoing complete
 *   page invalidation
D
David Howells 已提交
420
 * - Called if either PG_private or PG_fscache is set on the page
421 422
 * - Caller holds page lock
 */
423
static void nfs_invalidate_page(struct page *page, unsigned long offset)
424
{
425 426
	dfprintk(PAGECACHE, "NFS: invalidate_page(%p, %lu)\n", page, offset);

427 428
	if (offset != 0)
		return;
429
	/* Cancel any unstarted writes on this page */
430
	nfs_wb_page_cancel(page->mapping->host, page);
D
David Howells 已提交
431 432

	nfs_fscache_invalidate_page(page, page->mapping->host);
433 434
}

435 436
/*
 * Attempt to release the private state associated with a page
D
David Howells 已提交
437
 * - Called if either PG_private or PG_fscache is set on the page
438 439 440
 * - Caller holds page lock
 * - Return true (may release page) or false (may not)
 */
441 442
static int nfs_release_page(struct page *page, gfp_t gfp)
{
443 444
	dfprintk(PAGECACHE, "NFS: release_page(%p)\n", page);

445
	/* If PagePrivate() is set, then the page is not freeable */
D
David Howells 已提交
446 447 448
	if (PagePrivate(page))
		return 0;
	return nfs_fscache_release_page(page, gfp);
449 450
}

451 452 453 454
/*
 * Attempt to clear the private state associated with a page when an error
 * occurs that requires the cached contents of an inode to be written back or
 * destroyed
D
David Howells 已提交
455
 * - Called if either PG_private or fscache is set on the page
456 457 458
 * - Caller holds page lock
 * - Return 0 if successful, -error otherwise
 */
459 460
static int nfs_launder_page(struct page *page)
{
461
	struct inode *inode = page->mapping->host;
D
David Howells 已提交
462
	struct nfs_inode *nfsi = NFS_I(inode);
463 464 465 466

	dfprintk(PAGECACHE, "NFS: launder_page(%ld, %llu)\n",
		inode->i_ino, (long long)page_offset(page));

D
David Howells 已提交
467
	nfs_fscache_wait_on_page_write(nfsi, page);
468
	return nfs_wb_page(inode, page);
469 470
}

471
const struct address_space_operations nfs_file_aops = {
L
Linus Torvalds 已提交
472 473
	.readpage = nfs_readpage,
	.readpages = nfs_readpages,
T
Trond Myklebust 已提交
474
	.set_page_dirty = __set_page_dirty_nobuffers,
L
Linus Torvalds 已提交
475 476
	.writepage = nfs_writepage,
	.writepages = nfs_writepages,
N
Nick Piggin 已提交
477 478
	.write_begin = nfs_write_begin,
	.write_end = nfs_write_end,
479 480
	.invalidatepage = nfs_invalidate_page,
	.releasepage = nfs_release_page,
L
Linus Torvalds 已提交
481
	.direct_IO = nfs_direct_IO,
482
	.migratepage = nfs_migrate_page,
483
	.launder_page = nfs_launder_page,
L
Linus Torvalds 已提交
484 485
};

486 487 488 489 490
/*
 * Notification that a PTE pointing to an NFS page is about to be made
 * writable, implying that someone is about to modify the page through a
 * shared-writable mapping
 */
491
static int nfs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
492
{
493
	struct page *page = vmf->page;
494
	struct file *filp = vma->vm_file;
495
	struct dentry *dentry = filp->f_path.dentry;
496 497
	unsigned pagelen;
	int ret = -EINVAL;
N
Nick Piggin 已提交
498
	struct address_space *mapping;
499

500 501 502 503 504
	dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%s/%s(%ld), offset %lld)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		filp->f_mapping->host->i_ino,
		(long long)page_offset(page));

D
David Howells 已提交
505 506 507
	/* make sure the cache has finished storing the page */
	nfs_fscache_wait_on_page_write(NFS_I(dentry->d_inode), page);

508
	lock_page(page);
N
Nick Piggin 已提交
509
	mapping = page->mapping;
510
	if (mapping != dentry->d_inode->i_mapping)
511 512 513
		goto out_unlock;

	ret = 0;
514
	pagelen = nfs_page_length(page);
515 516
	if (pagelen == 0)
		goto out_unlock;
N
Nick Piggin 已提交
517

518 519 520 521 522 523
	ret = nfs_flush_incompatible(filp, page);
	if (ret != 0)
		goto out_unlock;

	ret = nfs_updatepage(filp, page, 0, pagelen);
out_unlock:
524 525
	if (!ret)
		return VM_FAULT_LOCKED;
526
	unlock_page(page);
527
	return VM_FAULT_SIGBUS;
528 529 530 531 532 533 534
}

static struct vm_operations_struct nfs_file_vm_ops = {
	.fault = filemap_fault,
	.page_mkwrite = nfs_vm_page_mkwrite,
};

535 536 537 538 539 540
static int nfs_need_sync_write(struct file *filp, struct inode *inode)
{
	struct nfs_open_context *ctx;

	if (IS_SYNC(inode) || (filp->f_flags & O_SYNC))
		return 1;
541
	ctx = nfs_file_open_context(filp);
542 543 544 545 546
	if (test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags))
		return 1;
	return 0;
}

547 548
static ssize_t nfs_file_write(struct kiocb *iocb, const struct iovec *iov,
				unsigned long nr_segs, loff_t pos)
L
Linus Torvalds 已提交
549
{
550
	struct dentry * dentry = iocb->ki_filp->f_path.dentry;
L
Linus Torvalds 已提交
551 552
	struct inode * inode = dentry->d_inode;
	ssize_t result;
553
	size_t count = iov_length(iov, nr_segs);
L
Linus Torvalds 已提交
554 555

	if (iocb->ki_filp->f_flags & O_DIRECT)
556
		return nfs_file_direct_write(iocb, iov, nr_segs, pos);
L
Linus Torvalds 已提交
557

C
Chuck Lever 已提交
558
	dprintk("NFS: write(%s/%s, %lu@%Ld)\n",
L
Linus Torvalds 已提交
559
		dentry->d_parent->d_name.name, dentry->d_name.name,
C
Chuck Lever 已提交
560
		(unsigned long) count, (long long) pos);
L
Linus Torvalds 已提交
561 562 563 564

	result = -EBUSY;
	if (IS_SWAPFILE(inode))
		goto out_swapfile;
565 566 567 568 569 570 571
	/*
	 * O_APPEND implies that we must revalidate the file length.
	 */
	if (iocb->ki_filp->f_flags & O_APPEND) {
		result = nfs_revalidate_file_size(inode, iocb->ki_filp);
		if (result)
			goto out;
572
	}
L
Linus Torvalds 已提交
573 574 575 576 577

	result = count;
	if (!count)
		goto out;

C
Chuck Lever 已提交
578
	nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, count);
579
	result = generic_file_aio_write(iocb, iov, nr_segs, pos);
T
Trond Myklebust 已提交
580
	/* Return error values for O_SYNC and IS_SYNC() */
581
	if (result >= 0 && nfs_need_sync_write(iocb->ki_filp, inode)) {
582
		int err = nfs_do_fsync(nfs_file_open_context(iocb->ki_filp), inode);
T
Trond Myklebust 已提交
583 584 585
		if (err < 0)
			result = err;
	}
L
Linus Torvalds 已提交
586 587 588 589 590 591 592 593
out:
	return result;

out_swapfile:
	printk(KERN_INFO "NFS: attempt to write to active swap file!\n");
	goto out;
}

594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
static ssize_t nfs_file_splice_write(struct pipe_inode_info *pipe,
				     struct file *filp, loff_t *ppos,
				     size_t count, unsigned int flags)
{
	struct dentry *dentry = filp->f_path.dentry;
	struct inode *inode = dentry->d_inode;
	ssize_t ret;

	dprintk("NFS splice_write(%s/%s, %lu@%llu)\n",
		dentry->d_parent->d_name.name, dentry->d_name.name,
		(unsigned long) count, (unsigned long long) *ppos);

	/*
	 * The combination of splice and an O_APPEND destination is disallowed.
	 */

	nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, count);

	ret = generic_file_splice_write(pipe, filp, ppos, count, flags);
	if (ret >= 0 && nfs_need_sync_write(filp, inode)) {
		int err = nfs_do_fsync(nfs_file_open_context(filp), inode);
		if (err < 0)
			ret = err;
	}
	return ret;
}

L
Linus Torvalds 已提交
621 622 623 624 625
static int do_getlk(struct file *filp, int cmd, struct file_lock *fl)
{
	struct inode *inode = filp->f_mapping->host;
	int status = 0;

626
	/* Try local locking first */
627 628 629
	posix_test_lock(filp, fl);
	if (fl->fl_type != F_UNLCK) {
		/* found a conflict */
630
		goto out;
L
Linus Torvalds 已提交
631
	}
632 633 634 635 636 637 638 639 640

	if (nfs_have_delegation(inode, FMODE_READ))
		goto out_noconflict;

	if (NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)
		goto out_noconflict;

	status = NFS_PROTO(inode)->lock(filp, cmd, fl);
out:
L
Linus Torvalds 已提交
641
	return status;
642 643 644
out_noconflict:
	fl->fl_type = F_UNLCK;
	goto out;
L
Linus Torvalds 已提交
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
}

static int do_vfs_lock(struct file *file, struct file_lock *fl)
{
	int res = 0;
	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
		case FL_POSIX:
			res = posix_lock_file_wait(file, fl);
			break;
		case FL_FLOCK:
			res = flock_lock_file_wait(file, fl);
			break;
		default:
			BUG();
	}
	if (res < 0)
661 662
		dprintk(KERN_WARNING "%s: VFS is out of sync with lock manager"
			" - error %d!\n",
663
				__func__, res);
L
Linus Torvalds 已提交
664 665 666 667 668 669 670 671 672 673 674 675
	return res;
}

static int do_unlk(struct file *filp, int cmd, struct file_lock *fl)
{
	struct inode *inode = filp->f_mapping->host;
	int status;

	/*
	 * Flush all pending writes before doing anything
	 * with locks..
	 */
T
Trond Myklebust 已提交
676
	nfs_sync_mapping(filp->f_mapping);
L
Linus Torvalds 已提交
677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698

	/* NOTE: special case
	 * 	If we're signalled while cleaning up locks on process exit, we
	 * 	still need to complete the unlock.
	 */
	/* Use local locking if mounted with "-onolock" */
	if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM))
		status = NFS_PROTO(inode)->lock(filp, cmd, fl);
	else
		status = do_vfs_lock(filp, fl);
	return status;
}

static int do_setlk(struct file *filp, int cmd, struct file_lock *fl)
{
	struct inode *inode = filp->f_mapping->host;
	int status;

	/*
	 * Flush all pending writes before doing anything
	 * with locks..
	 */
T
Trond Myklebust 已提交
699 700
	status = nfs_sync_mapping(filp->f_mapping);
	if (status != 0)
L
Linus Torvalds 已提交
701 702 703
		goto out;

	/* Use local locking if mounted with "-onolock" */
704
	if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM))
L
Linus Torvalds 已提交
705
		status = NFS_PROTO(inode)->lock(filp, cmd, fl);
706
	else
L
Linus Torvalds 已提交
707 708 709 710 711 712 713
		status = do_vfs_lock(filp, fl);
	if (status < 0)
		goto out;
	/*
	 * Make sure we clear the cache whenever we try to get the lock.
	 * This makes locking act as a cache coherency point.
	 */
T
Trond Myklebust 已提交
714
	nfs_sync_mapping(filp->f_mapping);
715 716
	if (!nfs_have_delegation(inode, FMODE_READ))
		nfs_zap_caches(inode);
L
Linus Torvalds 已提交
717 718 719 720 721 722 723 724 725
out:
	return status;
}

/*
 * Lock a (portion of) a file
 */
static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
{
C
Chuck Lever 已提交
726
	struct inode *inode = filp->f_mapping->host;
727
	int ret = -ENOLCK;
L
Linus Torvalds 已提交
728

C
Chuck Lever 已提交
729 730 731
	dprintk("NFS: lock(%s/%s, t=%x, fl=%x, r=%lld:%lld)\n",
			filp->f_path.dentry->d_parent->d_name.name,
			filp->f_path.dentry->d_name.name,
L
Linus Torvalds 已提交
732 733
			fl->fl_type, fl->fl_flags,
			(long long)fl->fl_start, (long long)fl->fl_end);
C
Chuck Lever 已提交
734

C
Chuck Lever 已提交
735
	nfs_inc_stats(inode, NFSIOS_VFSLOCK);
L
Linus Torvalds 已提交
736 737

	/* No mandatory locks over NFS */
738
	if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
739 740 741 742 743 744 745
		goto out_err;

	if (NFS_PROTO(inode)->lock_check_bounds != NULL) {
		ret = NFS_PROTO(inode)->lock_check_bounds(fl);
		if (ret < 0)
			goto out_err;
	}
L
Linus Torvalds 已提交
746 747

	if (IS_GETLK(cmd))
748 749 750 751 752 753 754
		ret = do_getlk(filp, cmd, fl);
	else if (fl->fl_type == F_UNLCK)
		ret = do_unlk(filp, cmd, fl);
	else
		ret = do_setlk(filp, cmd, fl);
out_err:
	return ret;
L
Linus Torvalds 已提交
755 756 757 758 759 760 761
}

/*
 * Lock a (portion of) a file
 */
static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
{
C
Chuck Lever 已提交
762 763 764
	dprintk("NFS: flock(%s/%s, t=%x, fl=%x)\n",
			filp->f_path.dentry->d_parent->d_name.name,
			filp->f_path.dentry->d_name.name,
L
Linus Torvalds 已提交
765 766 767 768 769 770 771 772 773 774 775 776 777 778
			fl->fl_type, fl->fl_flags);

	if (!(fl->fl_flags & FL_FLOCK))
		return -ENOLCK;

	/* We're simulating flock() locks using posix locks on the server */
	fl->fl_owner = (fl_owner_t)filp;
	fl->fl_start = 0;
	fl->fl_end = OFFSET_MAX;

	if (fl->fl_type == F_UNLCK)
		return do_unlk(filp, cmd, fl);
	return do_setlk(filp, cmd, fl);
}
J
J. Bruce Fields 已提交
779

C
Chuck Lever 已提交
780 781 782 783
/*
 * There is no protocol support for leases, so we have no way to implement
 * them correctly in the face of opens by other clients.
 */
J
J. Bruce Fields 已提交
784 785
static int nfs_setlease(struct file *file, long arg, struct file_lock **fl)
{
C
Chuck Lever 已提交
786 787 788 789
	dprintk("NFS: setlease(%s/%s, arg=%ld)\n",
			file->f_path.dentry->d_parent->d_name.name,
			file->f_path.dentry->d_name.name, arg);

J
J. Bruce Fields 已提交
790 791
	return -EINVAL;
}