write.c 37.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3
/*
 * linux/fs/nfs/write.c
 *
4
 * Write file data over NFS.
L
Linus Torvalds 已提交
5 6 7 8 9 10 11 12 13 14
 *
 * Copyright (C) 1996, 1997, Olaf Kirch <okir@monad.swb.de>
 */

#include <linux/types.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/file.h>
#include <linux/writeback.h>
15
#include <linux/swap.h>
L
Linus Torvalds 已提交
16 17 18 19 20

#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/nfs_page.h>
21 22
#include <linux/backing-dev.h>

L
Linus Torvalds 已提交
23 24 25
#include <asm/uaccess.h>

#include "delegation.h"
26
#include "internal.h"
C
Chuck Lever 已提交
27
#include "iostat.h"
L
Linus Torvalds 已提交
28 29 30 31 32 33 34 35 36 37 38 39

#define NFSDBG_FACILITY		NFSDBG_PAGECACHE

#define MIN_POOL_WRITE		(32)
#define MIN_POOL_COMMIT		(4)

/*
 * Local function declarations
 */
static struct nfs_page * nfs_update_request(struct nfs_open_context*,
					    struct page *,
					    unsigned int, unsigned int);
40 41
static void nfs_pageio_init_write(struct nfs_pageio_descriptor *desc,
				  struct inode *inode, int ioflags);
42 43 44
static const struct rpc_call_ops nfs_write_partial_ops;
static const struct rpc_call_ops nfs_write_full_ops;
static const struct rpc_call_ops nfs_commit_ops;
L
Linus Torvalds 已提交
45

46
static struct kmem_cache *nfs_wdata_cachep;
47
static mempool_t *nfs_wdata_mempool;
L
Linus Torvalds 已提交
48 49
static mempool_t *nfs_commit_mempool;

50
struct nfs_write_data *nfs_commit_alloc(void)
L
Linus Torvalds 已提交
51
{
C
Christoph Lameter 已提交
52
	struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, GFP_NOFS);
53

L
Linus Torvalds 已提交
54 55 56 57 58 59 60
	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
	}
	return p;
}

61
static void nfs_commit_rcu_free(struct rcu_head *head)
L
Linus Torvalds 已提交
62
{
63
	struct nfs_write_data *p = container_of(head, struct nfs_write_data, task.u.tk_rcu);
64 65
	if (p && (p->pagevec != &p->page_array[0]))
		kfree(p->pagevec);
L
Linus Torvalds 已提交
66 67 68
	mempool_free(p, nfs_commit_mempool);
}

69 70 71 72 73
void nfs_commit_free(struct nfs_write_data *wdata)
{
	call_rcu_bh(&wdata->task.u.tk_rcu, nfs_commit_rcu_free);
}

74
struct nfs_write_data *nfs_writedata_alloc(unsigned int pagecount)
75
{
C
Christoph Lameter 已提交
76
	struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, GFP_NOFS);
77 78 79 80

	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
81
		p->npages = pagecount;
82 83
		if (pagecount <= ARRAY_SIZE(p->page_array))
			p->pagevec = p->page_array;
84
		else {
85 86
			p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_NOFS);
			if (!p->pagevec) {
87 88 89 90 91 92 93 94
				mempool_free(p, nfs_wdata_mempool);
				p = NULL;
			}
		}
	}
	return p;
}

95
static void nfs_writedata_rcu_free(struct rcu_head *head)
96
{
97
	struct nfs_write_data *p = container_of(head, struct nfs_write_data, task.u.tk_rcu);
98 99 100 101 102
	if (p && (p->pagevec != &p->page_array[0]))
		kfree(p->pagevec);
	mempool_free(p, nfs_wdata_mempool);
}

103 104 105 106 107
static void nfs_writedata_free(struct nfs_write_data *wdata)
{
	call_rcu_bh(&wdata->task.u.tk_rcu, nfs_writedata_rcu_free);
}

108
void nfs_writedata_release(void *wdata)
L
Linus Torvalds 已提交
109 110 111 112
{
	nfs_writedata_free(wdata);
}

113 114 115 116 117 118 119
static struct nfs_page *nfs_page_find_request_locked(struct page *page)
{
	struct nfs_page *req = NULL;

	if (PagePrivate(page)) {
		req = (struct nfs_page *)page_private(page);
		if (req != NULL)
120
			kref_get(&req->wb_kref);
121 122 123 124 125 126
	}
	return req;
}

static struct nfs_page *nfs_page_find_request(struct page *page)
{
127
	struct inode *inode = page->mapping->host;
128 129
	struct nfs_page *req = NULL;

130
	spin_lock(&inode->i_lock);
131
	req = nfs_page_find_request_locked(page);
132
	spin_unlock(&inode->i_lock);
133 134 135
	return req;
}

L
Linus Torvalds 已提交
136 137 138 139 140
/* Adjust the file length if we're writing beyond the end */
static void nfs_grow_file(struct page *page, unsigned int offset, unsigned int count)
{
	struct inode *inode = page->mapping->host;
	loff_t end, i_size = i_size_read(inode);
141
	pgoff_t end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;
L
Linus Torvalds 已提交
142 143 144 145 146 147

	if (i_size > 0 && page->index < end_index)
		return;
	end = ((loff_t)page->index << PAGE_CACHE_SHIFT) + ((loff_t)offset+count);
	if (i_size >= end)
		return;
C
Chuck Lever 已提交
148
	nfs_inc_stats(inode, NFSIOS_EXTENDWRITE);
L
Linus Torvalds 已提交
149 150 151
	i_size_write(inode, end);
}

152 153 154 155 156 157 158
/* A writeback failed: mark the page as bad, and invalidate the page cache */
static void nfs_set_pageerror(struct page *page)
{
	SetPageError(page);
	nfs_zap_mapping(page->mapping->host, page->mapping);
}

L
Linus Torvalds 已提交
159 160 161 162 163 164 165 166 167
/* We can set the PG_uptodate flag if we see that a write request
 * covers the full page.
 */
static void nfs_mark_uptodate(struct page *page, unsigned int base, unsigned int count)
{
	if (PageUptodate(page))
		return;
	if (base != 0)
		return;
168
	if (count != nfs_page_length(page))
L
Linus Torvalds 已提交
169
		return;
170
	if (count != PAGE_CACHE_SIZE)
N
Nate Diller 已提交
171
		zero_user_page(page, count, PAGE_CACHE_SIZE - count, KM_USER0);
172
	SetPageUptodate(page);
L
Linus Torvalds 已提交
173 174
}

175
static int nfs_writepage_setup(struct nfs_open_context *ctx, struct page *page,
L
Linus Torvalds 已提交
176 177 178
		unsigned int offset, unsigned int count)
{
	struct nfs_page	*req;
179
	int ret;
L
Linus Torvalds 已提交
180

181 182 183 184 185 186 187 188 189 190 191
	for (;;) {
		req = nfs_update_request(ctx, page, offset, count);
		if (!IS_ERR(req))
			break;
		ret = PTR_ERR(req);
		if (ret != -EBUSY)
			return ret;
		ret = nfs_wb_page(page->mapping->host, page);
		if (ret != 0)
			return ret;
	}
L
Linus Torvalds 已提交
192 193 194
	/* Update file length */
	nfs_grow_file(page, offset, count);
	nfs_unlock_request(req);
195
	return 0;
L
Linus Torvalds 已提交
196 197 198 199 200
}

static int wb_priority(struct writeback_control *wbc)
{
	if (wbc->for_reclaim)
201
		return FLUSH_HIGHPRI | FLUSH_STABLE;
L
Linus Torvalds 已提交
202 203 204 205 206
	if (wbc->for_kupdate)
		return FLUSH_LOWPRI;
	return 0;
}

207 208 209 210 211 212 213 214 215 216
/*
 * NFS congestion control
 */

int nfs_congestion_kb;

#define NFS_CONGESTION_ON_THRESH 	(nfs_congestion_kb >> (PAGE_SHIFT-10))
#define NFS_CONGESTION_OFF_THRESH	\
	(NFS_CONGESTION_ON_THRESH - (NFS_CONGESTION_ON_THRESH >> 2))

217
static int nfs_set_page_writeback(struct page *page)
218
{
219 220 221
	int ret = test_set_page_writeback(page);

	if (!ret) {
222 223 224
		struct inode *inode = page->mapping->host;
		struct nfs_server *nfss = NFS_SERVER(inode);

225
		if (atomic_long_inc_return(&nfss->writeback) >
226 227 228
				NFS_CONGESTION_ON_THRESH)
			set_bdi_congested(&nfss->backing_dev_info, WRITE);
	}
229
	return ret;
230 231 232 233 234 235 236 237
}

static void nfs_end_page_writeback(struct page *page)
{
	struct inode *inode = page->mapping->host;
	struct nfs_server *nfss = NFS_SERVER(inode);

	end_page_writeback(page);
238
	if (atomic_long_dec_return(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH) {
239 240 241 242 243
		clear_bdi_congested(&nfss->backing_dev_info, WRITE);
		congestion_end(WRITE);
	}
}

244 245 246 247 248 249 250
/*
 * Find an associated nfs write request, and prepare to flush it out
 * Returns 1 if there was no write request, or if the request was
 * already tagged by nfs_set_page_dirty.Returns 0 if the request
 * was not tagged.
 * May also return an error if the user signalled nfs_wait_on_request().
 */
251 252
static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
				struct page *page)
253
{
254 255
	struct inode *inode = page->mapping->host;
	struct nfs_inode *nfsi = NFS_I(inode);
256 257 258
	struct nfs_page *req;
	int ret;

259
	spin_lock(&inode->i_lock);
260 261 262
	for(;;) {
		req = nfs_page_find_request_locked(page);
		if (req == NULL) {
263
			spin_unlock(&inode->i_lock);
264 265 266 267 268 269 270 271 272
			return 1;
		}
		if (nfs_lock_request_dontget(req))
			break;
		/* Note: If we hold the page lock, as is the case in nfs_writepage,
		 *	 then the call to nfs_lock_request_dontget() will always
		 *	 succeed provided that someone hasn't already marked the
		 *	 request as dirty (in which case we don't care).
		 */
273
		spin_unlock(&inode->i_lock);
274 275 276 277
		ret = nfs_wait_on_request(req);
		nfs_release_request(req);
		if (ret != 0)
			return ret;
278
		spin_lock(&inode->i_lock);
279
	}
280 281
	if (test_bit(PG_NEED_COMMIT, &req->wb_flags)) {
		/* This request is marked for commit */
282
		spin_unlock(&inode->i_lock);
283
		nfs_unlock_request(req);
284
		nfs_pageio_complete(pgio);
285 286
		return 1;
	}
287
	if (nfs_set_page_writeback(page) != 0) {
288
		spin_unlock(&inode->i_lock);
289 290 291
		BUG();
	}
	radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index,
292
			NFS_PAGE_TAG_LOCKED);
293
	ret = test_bit(PG_NEED_FLUSH, &req->wb_flags);
294
	spin_unlock(&inode->i_lock);
295
	nfs_pageio_add_request(pgio, req);
296 297 298
	return ret;
}

L
Linus Torvalds 已提交
299 300 301
/*
 * Write an mmapped page to the server.
 */
302
static int nfs_writepage_locked(struct page *page, struct writeback_control *wbc)
L
Linus Torvalds 已提交
303
{
304
	struct nfs_pageio_descriptor mypgio, *pgio;
L
Linus Torvalds 已提交
305 306
	struct nfs_open_context *ctx;
	struct inode *inode = page->mapping->host;
307
	unsigned offset;
308
	int err;
L
Linus Torvalds 已提交
309

C
Chuck Lever 已提交
310 311 312
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
	nfs_add_stats(inode, NFSIOS_WRITEPAGES, 1);

313 314 315 316 317 318 319
	if (wbc->for_writepages)
		pgio = wbc->fs_private;
	else {
		nfs_pageio_init_write(&mypgio, inode, wb_priority(wbc));
		pgio = &mypgio;
	}

320 321
	nfs_pageio_cond_complete(pgio, page->index);

322
	err = nfs_page_async_flush(pgio, page);
323 324 325
	if (err <= 0)
		goto out;
	err = 0;
326 327
	offset = nfs_page_length(page);
	if (!offset)
L
Linus Torvalds 已提交
328
		goto out;
329

330 331
	nfs_pageio_cond_complete(pgio, page->index);

332
	ctx = nfs_find_open_context(inode, NULL, FMODE_WRITE);
L
Linus Torvalds 已提交
333 334 335 336
	if (ctx == NULL) {
		err = -EBADF;
		goto out;
	}
T
Trond Myklebust 已提交
337
	err = nfs_writepage_setup(ctx, page, 0, offset);
L
Linus Torvalds 已提交
338
	put_nfs_open_context(ctx);
339 340
	if (err != 0)
		goto out;
341
	err = nfs_page_async_flush(pgio, page);
342 343
	if (err > 0)
		err = 0;
L
Linus Torvalds 已提交
344
out:
T
Trond Myklebust 已提交
345
	if (!wbc->for_writepages)
346
		nfs_pageio_complete(pgio);
347 348 349 350 351 352 353 354
	return err;
}

int nfs_writepage(struct page *page, struct writeback_control *wbc)
{
	int err;

	err = nfs_writepage_locked(page, wbc);
L
Linus Torvalds 已提交
355 356 357 358 359 360 361
	unlock_page(page);
	return err; 
}

int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
	struct inode *inode = mapping->host;
362
	struct nfs_pageio_descriptor pgio;
L
Linus Torvalds 已提交
363 364
	int err;

C
Chuck Lever 已提交
365 366
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);

367 368
	nfs_pageio_init_write(&pgio, inode, wb_priority(wbc));
	wbc->fs_private = &pgio;
L
Linus Torvalds 已提交
369
	err = generic_writepages(mapping, wbc);
370
	nfs_pageio_complete(&pgio);
L
Linus Torvalds 已提交
371 372
	if (err)
		return err;
373 374 375
	if (pgio.pg_error)
		return pgio.pg_error;
	return 0;
L
Linus Torvalds 已提交
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395
}

/*
 * Insert a write request into an inode
 */
static int nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	int error;

	error = radix_tree_insert(&nfsi->nfs_page_tree, req->wb_index, req);
	BUG_ON(error == -EEXIST);
	if (error)
		return error;
	if (!nfsi->npages) {
		igrab(inode);
		nfs_begin_data_update(inode);
		if (nfs_have_delegation(inode, FMODE_WRITE))
			nfsi->change_attr++;
	}
396
	SetPagePrivate(req->wb_page);
397
	set_page_private(req->wb_page, (unsigned long)req);
398 399
	if (PageDirty(req->wb_page))
		set_bit(PG_NEED_FLUSH, &req->wb_flags);
L
Linus Torvalds 已提交
400
	nfsi->npages++;
401
	kref_get(&req->wb_kref);
L
Linus Torvalds 已提交
402 403 404 405
	return 0;
}

/*
406
 * Remove a write request from an inode
L
Linus Torvalds 已提交
407 408 409
 */
static void nfs_inode_remove_request(struct nfs_page *req)
{
410
	struct inode *inode = req->wb_context->path.dentry->d_inode;
L
Linus Torvalds 已提交
411 412 413 414
	struct nfs_inode *nfsi = NFS_I(inode);

	BUG_ON (!NFS_WBACK_BUSY(req));

415
	spin_lock(&inode->i_lock);
416
	set_page_private(req->wb_page, 0);
417
	ClearPagePrivate(req->wb_page);
L
Linus Torvalds 已提交
418
	radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index);
419 420
	if (test_and_clear_bit(PG_NEED_FLUSH, &req->wb_flags))
		__set_page_dirty_nobuffers(req->wb_page);
L
Linus Torvalds 已提交
421 422
	nfsi->npages--;
	if (!nfsi->npages) {
423
		spin_unlock(&inode->i_lock);
424
		nfs_end_data_update(inode);
L
Linus Torvalds 已提交
425 426
		iput(inode);
	} else
427
		spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
428 429 430 431
	nfs_clear_request(req);
	nfs_release_request(req);
}

432 433 434 435 436 437
static void
nfs_redirty_request(struct nfs_page *req)
{
	__set_page_dirty_nobuffers(req->wb_page);
}

L
Linus Torvalds 已提交
438 439 440 441 442 443
/*
 * Check if a request is dirty
 */
static inline int
nfs_dirty_request(struct nfs_page *req)
{
444 445
	struct page *page = req->wb_page;

446
	if (page == NULL || test_bit(PG_NEED_COMMIT, &req->wb_flags))
447 448
		return 0;
	return !PageWriteback(req->wb_page);
L
Linus Torvalds 已提交
449 450 451 452 453 454 455 456 457
}

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
/*
 * Add a request to the inode's commit list.
 */
static void
nfs_mark_request_commit(struct nfs_page *req)
{
458
	struct inode *inode = req->wb_context->path.dentry->d_inode;
L
Linus Torvalds 已提交
459 460
	struct nfs_inode *nfsi = NFS_I(inode);

461
	spin_lock(&inode->i_lock);
L
Linus Torvalds 已提交
462
	nfsi->ncommit++;
463
	set_bit(PG_NEED_COMMIT, &(req)->wb_flags);
464 465 466
	radix_tree_tag_set(&nfsi->nfs_page_tree,
			req->wb_index,
			NFS_PAGE_TAG_COMMIT);
467
	spin_unlock(&inode->i_lock);
468
	inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
469
	__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
L
Linus Torvalds 已提交
470
}
471 472 473 474 475 476 477 478 479 480

static inline
int nfs_write_need_commit(struct nfs_write_data *data)
{
	return data->verf.committed != NFS_FILE_SYNC;
}

static inline
int nfs_reschedule_unstable_write(struct nfs_page *req)
{
481
	if (test_bit(PG_NEED_COMMIT, &req->wb_flags)) {
482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507
		nfs_mark_request_commit(req);
		return 1;
	}
	if (test_and_clear_bit(PG_NEED_RESCHED, &req->wb_flags)) {
		nfs_redirty_request(req);
		return 1;
	}
	return 0;
}
#else
static inline void
nfs_mark_request_commit(struct nfs_page *req)
{
}

static inline
int nfs_write_need_commit(struct nfs_write_data *data)
{
	return 0;
}

static inline
int nfs_reschedule_unstable_write(struct nfs_page *req)
{
	return 0;
}
L
Linus Torvalds 已提交
508 509 510 511 512 513 514
#endif

/*
 * Wait for a request to complete.
 *
 * Interruptible by signals only if mounted with intr flag.
 */
515
static int nfs_wait_on_requests_locked(struct inode *inode, pgoff_t idx_start, unsigned int npages)
L
Linus Torvalds 已提交
516 517 518
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_page *req;
519
	pgoff_t idx_end, next;
L
Linus Torvalds 已提交
520 521 522 523 524 525 526 527 528
	unsigned int		res = 0;
	int			error;

	if (npages == 0)
		idx_end = ~0;
	else
		idx_end = idx_start + npages - 1;

	next = idx_start;
529
	while (radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&req, next, 1, NFS_PAGE_TAG_LOCKED)) {
L
Linus Torvalds 已提交
530 531 532 533
		if (req->wb_index > idx_end)
			break;

		next = req->wb_index + 1;
534
		BUG_ON(!NFS_WBACK_BUSY(req));
L
Linus Torvalds 已提交
535

536
		kref_get(&req->wb_kref);
537
		spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
538 539
		error = nfs_wait_on_request(req);
		nfs_release_request(req);
540
		spin_lock(&inode->i_lock);
L
Linus Torvalds 已提交
541 542 543 544 545 546 547
		if (error < 0)
			return error;
		res++;
	}
	return res;
}

T
Trond Myklebust 已提交
548 549 550 551 552 553
static void nfs_cancel_commit_list(struct list_head *head)
{
	struct nfs_page *req;

	while(!list_empty(head)) {
		req = nfs_list_entry(head->next);
554
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
T
Trond Myklebust 已提交
555
		nfs_list_remove_request(req);
556
		clear_bit(PG_NEED_COMMIT, &(req)->wb_flags);
T
Trond Myklebust 已提交
557
		nfs_inode_remove_request(req);
558
		nfs_unlock_request(req);
T
Trond Myklebust 已提交
559 560 561
	}
}

L
Linus Torvalds 已提交
562 563 564 565 566 567 568 569 570 571 572 573
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
/*
 * nfs_scan_commit - Scan an inode for commit requests
 * @inode: NFS inode to scan
 * @dst: destination list
 * @idx_start: lower bound of page->index to scan.
 * @npages: idx_start + npages sets the upper bound to scan.
 *
 * Moves requests from the inode's 'commit' request list.
 * The requests are *not* checked to ensure that they form a contiguous set.
 */
static int
574
nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, unsigned int npages)
L
Linus Torvalds 已提交
575 576
{
	struct nfs_inode *nfsi = NFS_I(inode);
577 578 579
	int res = 0;

	if (nfsi->ncommit != 0) {
580 581
		res = nfs_scan_list(nfsi, dst, idx_start, npages,
				NFS_PAGE_TAG_COMMIT);
582 583
		nfsi->ncommit -= res;
	}
L
Linus Torvalds 已提交
584 585
	return res;
}
586
#else
587
static inline int nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, unsigned int npages)
588 589 590
{
	return 0;
}
L
Linus Torvalds 已提交
591 592 593 594 595 596 597 598 599 600
#endif

/*
 * Try to update any existing write request, or create one if there is none.
 * In order to match, the request's credentials must match those of
 * the calling process.
 *
 * Note: Should always be called with the Page Lock held!
 */
static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
601
		struct page *page, unsigned int offset, unsigned int bytes)
L
Linus Torvalds 已提交
602
{
603 604
	struct address_space *mapping = page->mapping;
	struct inode *inode = mapping->host;
L
Linus Torvalds 已提交
605
	struct nfs_page		*req, *new = NULL;
606
	pgoff_t		rqend, end;
L
Linus Torvalds 已提交
607 608 609 610 611 612 613

	end = offset + bytes;

	for (;;) {
		/* Loop over all inode entries and see if we find
		 * A request for the page we wish to update
		 */
614
		spin_lock(&inode->i_lock);
615
		req = nfs_page_find_request_locked(page);
L
Linus Torvalds 已提交
616 617 618
		if (req) {
			if (!nfs_lock_request_dontget(req)) {
				int error;
619

620
				spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
621 622
				error = nfs_wait_on_request(req);
				nfs_release_request(req);
N
Neil Brown 已提交
623 624 625
				if (error < 0) {
					if (new)
						nfs_release_request(new);
L
Linus Torvalds 已提交
626
					return ERR_PTR(error);
N
Neil Brown 已提交
627
				}
L
Linus Torvalds 已提交
628 629
				continue;
			}
630
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
631 632 633 634 635 636 637 638 639 640
			if (new)
				nfs_release_request(new);
			break;
		}

		if (new) {
			int error;
			nfs_lock_request_dontget(new);
			error = nfs_inode_add_request(inode, new);
			if (error) {
641
				spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
642 643 644
				nfs_unlock_request(new);
				return ERR_PTR(error);
			}
645
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
646 647
			return new;
		}
648
		spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686

		new = nfs_create_request(ctx, inode, page, offset, bytes);
		if (IS_ERR(new))
			return new;
	}

	/* We have a request for our page.
	 * If the creds don't match, or the
	 * page addresses don't match,
	 * tell the caller to wait on the conflicting
	 * request.
	 */
	rqend = req->wb_offset + req->wb_bytes;
	if (req->wb_context != ctx
	    || req->wb_page != page
	    || !nfs_dirty_request(req)
	    || offset > rqend || end < req->wb_offset) {
		nfs_unlock_request(req);
		return ERR_PTR(-EBUSY);
	}

	/* Okay, the request matches. Update the region */
	if (offset < req->wb_offset) {
		req->wb_offset = offset;
		req->wb_pgbase = offset;
		req->wb_bytes = rqend - req->wb_offset;
	}

	if (end > rqend)
		req->wb_bytes = end - req->wb_offset;

	return req;
}

int nfs_flush_incompatible(struct file *file, struct page *page)
{
	struct nfs_open_context *ctx = (struct nfs_open_context *)file->private_data;
	struct nfs_page	*req;
T
Trond Myklebust 已提交
687
	int do_flush, status;
L
Linus Torvalds 已提交
688 689 690 691 692 693 694 695
	/*
	 * Look for a request corresponding to this page. If there
	 * is one, and it belongs to another file, we flush it out
	 * before we try to copy anything into the page. Do this
	 * due to the lack of an ACCESS-type call in NFSv2.
	 * Also do the same if we find a request from an existing
	 * dropped page.
	 */
T
Trond Myklebust 已提交
696 697 698 699 700
	do {
		req = nfs_page_find_request(page);
		if (req == NULL)
			return 0;
		do_flush = req->wb_page != page || req->wb_context != ctx
701
			|| !nfs_dirty_request(req);
L
Linus Torvalds 已提交
702
		nfs_release_request(req);
T
Trond Myklebust 已提交
703 704 705 706 707
		if (!do_flush)
			return 0;
		status = nfs_wb_page(page->mapping->host, page);
	} while (status == 0);
	return status;
L
Linus Torvalds 已提交
708 709 710 711 712 713 714 715 716 717 718 719 720 721 722
}

/*
 * Update and possibly write a cached page of an NFS file.
 *
 * XXX: Keep an eye on generic_file_read to make sure it doesn't do bad
 * things with a page scheduled for an RPC call (e.g. invalidate it).
 */
int nfs_updatepage(struct file *file, struct page *page,
		unsigned int offset, unsigned int count)
{
	struct nfs_open_context *ctx = (struct nfs_open_context *)file->private_data;
	struct inode	*inode = page->mapping->host;
	int		status = 0;

C
Chuck Lever 已提交
723 724
	nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);

L
Linus Torvalds 已提交
725
	dprintk("NFS:      nfs_updatepage(%s/%s %d@%Ld)\n",
726 727
		file->f_path.dentry->d_parent->d_name.name,
		file->f_path.dentry->d_name.name, count,
728
		(long long)(page_offset(page) +offset));
L
Linus Torvalds 已提交
729 730 731 732 733

	/* If we're not using byte range locks, and we know the page
	 * is entirely in cache, it may be more efficient to avoid
	 * fragmenting write requests.
	 */
734
	if (PageUptodate(page) && inode->i_flock == NULL && !(file->f_mode & O_SYNC)) {
735
		count = max(count + offset, nfs_page_length(page));
L
Linus Torvalds 已提交
736 737 738
		offset = 0;
	}

739
	status = nfs_writepage_setup(ctx, page, offset, count);
740
	__set_page_dirty_nobuffers(page);
L
Linus Torvalds 已提交
741 742 743 744

        dprintk("NFS:      nfs_updatepage returns %d (isize %Ld)\n",
			status, (long long)i_size_read(inode));
	if (status < 0)
745
		nfs_set_pageerror(page);
L
Linus Torvalds 已提交
746 747 748 749 750 751
	return status;
}

static void nfs_writepage_release(struct nfs_page *req)
{

752 753 754 755 756 757
	if (PageError(req->wb_page)) {
		nfs_end_page_writeback(req->wb_page);
		nfs_inode_remove_request(req);
	} else if (!nfs_reschedule_unstable_write(req)) {
		/* Set the PG_uptodate flag */
		nfs_mark_uptodate(req->wb_page, req->wb_pgbase, req->wb_bytes);
758 759 760 761
		nfs_end_page_writeback(req->wb_page);
		nfs_inode_remove_request(req);
	} else
		nfs_end_page_writeback(req->wb_page);
762
	nfs_clear_page_tag_locked(req);
L
Linus Torvalds 已提交
763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780
}

static inline int flush_task_priority(int how)
{
	switch (how & (FLUSH_HIGHPRI|FLUSH_LOWPRI)) {
		case FLUSH_HIGHPRI:
			return RPC_PRIORITY_HIGH;
		case FLUSH_LOWPRI:
			return RPC_PRIORITY_LOW;
	}
	return RPC_PRIORITY_NORMAL;
}

/*
 * Set up the argument/result storage required for the RPC call.
 */
static void nfs_write_rpcsetup(struct nfs_page *req,
		struct nfs_write_data *data,
781
		const struct rpc_call_ops *call_ops,
L
Linus Torvalds 已提交
782 783 784 785
		unsigned int count, unsigned int offset,
		int how)
{
	struct inode		*inode;
786
	int flags;
L
Linus Torvalds 已提交
787 788 789 790 791

	/* Set up the RPC argument and reply structs
	 * NB: take care not to mess about with data->commit et al. */

	data->req = req;
792
	data->inode = inode = req->wb_context->path.dentry->d_inode;
L
Linus Torvalds 已提交
793 794 795 796 797 798 799 800 801 802 803 804
	data->cred = req->wb_context->cred;

	data->args.fh     = NFS_FH(inode);
	data->args.offset = req_offset(req) + offset;
	data->args.pgbase = req->wb_pgbase + offset;
	data->args.pages  = data->pagevec;
	data->args.count  = count;
	data->args.context = req->wb_context;

	data->res.fattr   = &data->fattr;
	data->res.count   = count;
	data->res.verf    = &data->verf;
805
	nfs_fattr_init(&data->fattr);
L
Linus Torvalds 已提交
806

807 808 809
	/* Set up the initial task struct.  */
	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, call_ops, data);
L
Linus Torvalds 已提交
810 811 812 813 814
	NFS_PROTO(inode)->write_setup(data, how);

	data->task.tk_priority = flush_task_priority(how);
	data->task.tk_cookie = (unsigned long)inode;

C
Chuck Lever 已提交
815 816
	dprintk("NFS: %5u initiated write call "
		"(req %s/%Ld, %u bytes @ offset %Lu)\n",
817
		data->task.tk_pid,
L
Linus Torvalds 已提交
818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837
		inode->i_sb->s_id,
		(long long)NFS_FILEID(inode),
		count,
		(unsigned long long)data->args.offset);
}

static void nfs_execute_write(struct nfs_write_data *data)
{
	struct rpc_clnt *clnt = NFS_CLIENT(data->inode);
	sigset_t oldset;

	rpc_clnt_sigmask(clnt, &oldset);
	rpc_execute(&data->task);
	rpc_clnt_sigunmask(clnt, &oldset);
}

/*
 * Generate multiple small requests to write out a single
 * contiguous dirty area on one page.
 */
838
static int nfs_flush_multi(struct inode *inode, struct list_head *head, unsigned int npages, size_t count, int how)
L
Linus Torvalds 已提交
839 840 841 842
{
	struct nfs_page *req = nfs_list_entry(head->next);
	struct page *page = req->wb_page;
	struct nfs_write_data *data;
843 844
	size_t wsize = NFS_SERVER(inode)->wsize, nbytes;
	unsigned int offset;
L
Linus Torvalds 已提交
845 846 847 848 849
	int requests = 0;
	LIST_HEAD(list);

	nfs_list_remove_request(req);

850
	nbytes = count;
851 852 853
	do {
		size_t len = min(nbytes, wsize);

854
		data = nfs_writedata_alloc(1);
L
Linus Torvalds 已提交
855 856 857 858
		if (!data)
			goto out_bad;
		list_add(&data->pages, &list);
		requests++;
859 860
		nbytes -= len;
	} while (nbytes != 0);
L
Linus Torvalds 已提交
861 862 863 864
	atomic_set(&req->wb_complete, requests);

	ClearPageError(page);
	offset = 0;
865
	nbytes = count;
L
Linus Torvalds 已提交
866 867 868 869 870 871
	do {
		data = list_entry(list.next, struct nfs_write_data, pages);
		list_del_init(&data->pages);

		data->pagevec[0] = page;

872 873 874 875 876 877
		if (nbytes < wsize)
			wsize = nbytes;
		nfs_write_rpcsetup(req, data, &nfs_write_partial_ops,
				   wsize, offset, how);
		offset += wsize;
		nbytes -= wsize;
L
Linus Torvalds 已提交
878 879 880 881 882 883 884 885 886
		nfs_execute_write(data);
	} while (nbytes != 0);

	return 0;

out_bad:
	while (!list_empty(&list)) {
		data = list_entry(list.next, struct nfs_write_data, pages);
		list_del(&data->pages);
887
		nfs_writedata_release(data);
L
Linus Torvalds 已提交
888
	}
889
	nfs_redirty_request(req);
890
	nfs_end_page_writeback(req->wb_page);
891
	nfs_clear_page_tag_locked(req);
L
Linus Torvalds 已提交
892 893 894 895 896 897 898 899 900 901 902
	return -ENOMEM;
}

/*
 * Create an RPC task for the given write request and kick it.
 * The page must have been locked by the caller.
 *
 * It may happen that the page we're passed is not marked dirty.
 * This is the case if nfs_updatepage detects a conflicting request
 * that has been written but not committed.
 */
903
static int nfs_flush_one(struct inode *inode, struct list_head *head, unsigned int npages, size_t count, int how)
L
Linus Torvalds 已提交
904 905 906 907 908
{
	struct nfs_page		*req;
	struct page		**pages;
	struct nfs_write_data	*data;

909
	data = nfs_writedata_alloc(npages);
L
Linus Torvalds 已提交
910 911 912 913 914 915 916 917 918 919 920 921 922 923
	if (!data)
		goto out_bad;

	pages = data->pagevec;
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_list_add_request(req, &data->pages);
		ClearPageError(req->wb_page);
		*pages++ = req->wb_page;
	}
	req = nfs_list_entry(data->pages.next);

	/* Set up the argument struct */
924
	nfs_write_rpcsetup(req, data, &nfs_write_full_ops, count, 0, how);
L
Linus Torvalds 已提交
925 926 927 928 929

	nfs_execute_write(data);
	return 0;
 out_bad:
	while (!list_empty(head)) {
930
		req = nfs_list_entry(head->next);
L
Linus Torvalds 已提交
931
		nfs_list_remove_request(req);
932
		nfs_redirty_request(req);
933
		nfs_end_page_writeback(req->wb_page);
934
		nfs_clear_page_tag_locked(req);
L
Linus Torvalds 已提交
935 936 937 938
	}
	return -ENOMEM;
}

939 940
static void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio,
				  struct inode *inode, int ioflags)
L
Linus Torvalds 已提交
941
{
T
Trond Myklebust 已提交
942
	int wsize = NFS_SERVER(inode)->wsize;
L
Linus Torvalds 已提交
943

944
	if (wsize < PAGE_CACHE_SIZE)
945
		nfs_pageio_init(pgio, inode, nfs_flush_multi, wsize, ioflags);
946
	else
947
		nfs_pageio_init(pgio, inode, nfs_flush_one, wsize, ioflags);
L
Linus Torvalds 已提交
948 949 950 951 952
}

/*
 * Handle a write reply that flushed part of a page.
 */
953
static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
954
{
955
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
956 957 958 959
	struct nfs_page		*req = data->req;
	struct page		*page = req->wb_page;

	dprintk("NFS: write (%s/%Ld %d@%Ld)",
960 961
		req->wb_context->path.dentry->d_inode->i_sb->s_id,
		(long long)NFS_FILEID(req->wb_context->path.dentry->d_inode),
L
Linus Torvalds 已提交
962 963 964
		req->wb_bytes,
		(long long)req_offset(req));

965 966 967 968
	if (nfs_writeback_done(task, data) != 0)
		return;

	if (task->tk_status < 0) {
969
		nfs_set_pageerror(page);
970 971
		req->wb_context->error = task->tk_status;
		dprintk(", error = %d\n", task->tk_status);
972
		goto out;
L
Linus Torvalds 已提交
973 974
	}

975
	if (nfs_write_need_commit(data)) {
976
		struct inode *inode = page->mapping->host;
977

978
		spin_lock(&inode->i_lock);
979 980 981 982 983 984 985 986 987 988
		if (test_bit(PG_NEED_RESCHED, &req->wb_flags)) {
			/* Do nothing we need to resend the writes */
		} else if (!test_and_set_bit(PG_NEED_COMMIT, &req->wb_flags)) {
			memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
			dprintk(" defer commit\n");
		} else if (memcmp(&req->wb_verf, &data->verf, sizeof(req->wb_verf))) {
			set_bit(PG_NEED_RESCHED, &req->wb_flags);
			clear_bit(PG_NEED_COMMIT, &req->wb_flags);
			dprintk(" server reboot detected\n");
		}
989
		spin_unlock(&inode->i_lock);
990 991 992 993
	} else
		dprintk(" OK\n");

out:
L
Linus Torvalds 已提交
994 995 996 997
	if (atomic_dec_and_test(&req->wb_complete))
		nfs_writepage_release(req);
}

998 999 1000 1001 1002
static const struct rpc_call_ops nfs_write_partial_ops = {
	.rpc_call_done = nfs_writeback_done_partial,
	.rpc_release = nfs_writedata_release,
};

L
Linus Torvalds 已提交
1003 1004 1005 1006 1007 1008 1009
/*
 * Handle a write reply that flushes a whole page.
 *
 * FIXME: There is an inherent race with invalidate_inode_pages and
 *	  writebacks since the page->count is kept > 1 for as long
 *	  as the page has a write request pending.
 */
1010
static void nfs_writeback_done_full(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1011
{
1012
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
1013 1014 1015
	struct nfs_page		*req;
	struct page		*page;

1016 1017 1018
	if (nfs_writeback_done(task, data) != 0)
		return;

L
Linus Torvalds 已提交
1019 1020 1021 1022 1023 1024 1025
	/* Update attributes as result of writeback. */
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
		page = req->wb_page;

		dprintk("NFS: write (%s/%Ld %d@%Ld)",
1026 1027
			req->wb_context->path.dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->path.dentry->d_inode),
L
Linus Torvalds 已提交
1028 1029 1030
			req->wb_bytes,
			(long long)req_offset(req));

1031
		if (task->tk_status < 0) {
1032
			nfs_set_pageerror(page);
1033 1034
			req->wb_context->error = task->tk_status;
			dprintk(", error = %d\n", task->tk_status);
1035
			goto remove_request;
L
Linus Torvalds 已提交
1036 1037
		}

1038 1039 1040 1041 1042
		if (nfs_write_need_commit(data)) {
			memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
			nfs_mark_request_commit(req);
			nfs_end_page_writeback(page);
			dprintk(" marked for commit\n");
L
Linus Torvalds 已提交
1043 1044
			goto next;
		}
1045 1046
		/* Set the PG_uptodate flag? */
		nfs_mark_uptodate(page, req->wb_pgbase, req->wb_bytes);
1047 1048 1049
		dprintk(" OK\n");
remove_request:
		nfs_end_page_writeback(page);
L
Linus Torvalds 已提交
1050 1051
		nfs_inode_remove_request(req);
	next:
1052
		nfs_clear_page_tag_locked(req);
L
Linus Torvalds 已提交
1053 1054 1055
	}
}

1056 1057 1058 1059 1060 1061
static const struct rpc_call_ops nfs_write_full_ops = {
	.rpc_call_done = nfs_writeback_done_full,
	.rpc_release = nfs_writedata_release,
};


L
Linus Torvalds 已提交
1062 1063 1064
/*
 * This function is called when the WRITE call is complete.
 */
1065
int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
L
Linus Torvalds 已提交
1066 1067 1068
{
	struct nfs_writeargs	*argp = &data->args;
	struct nfs_writeres	*resp = &data->res;
1069
	int status;
L
Linus Torvalds 已提交
1070

C
Chuck Lever 已提交
1071
	dprintk("NFS: %5u nfs_writeback_done (status %d)\n",
L
Linus Torvalds 已提交
1072 1073
		task->tk_pid, task->tk_status);

1074 1075 1076 1077 1078 1079 1080
	/*
	 * ->write_done will attempt to use post-op attributes to detect
	 * conflicting writes by other clients.  A strict interpretation
	 * of close-to-open would allow us to continue caching even if
	 * another writer had changed the file, but some applications
	 * depend on tighter cache coherency when writing.
	 */
1081 1082 1083
	status = NFS_PROTO(data->inode)->write_done(task, data);
	if (status != 0)
		return status;
C
Chuck Lever 已提交
1084 1085
	nfs_add_stats(data->inode, NFSIOS_SERVERWRITTENBYTES, resp->count);

L
Linus Torvalds 已提交
1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
	if (resp->verf->committed < argp->stable && task->tk_status >= 0) {
		/* We tried a write call, but the server did not
		 * commit data to stable storage even though we
		 * requested it.
		 * Note: There is a known bug in Tru64 < 5.0 in which
		 *	 the server reports NFS_DATA_SYNC, but performs
		 *	 NFS_FILE_SYNC. We therefore implement this checking
		 *	 as a dprintk() in order to avoid filling syslog.
		 */
		static unsigned long    complain;

		if (time_before(complain, jiffies)) {
			dprintk("NFS: faulty NFS server %s:"
				" (committed = %d) != (stable = %d)\n",
1101
				NFS_SERVER(data->inode)->nfs_client->cl_hostname,
L
Linus Torvalds 已提交
1102 1103 1104 1105 1106 1107 1108 1109 1110
				resp->verf->committed, argp->stable);
			complain = jiffies + 300 * HZ;
		}
	}
#endif
	/* Is this a short write? */
	if (task->tk_status >= 0 && resp->count < argp->count) {
		static unsigned long    complain;

C
Chuck Lever 已提交
1111 1112
		nfs_inc_stats(data->inode, NFSIOS_SHORTWRITE);

L
Linus Torvalds 已提交
1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
		/* Has the server at least made some progress? */
		if (resp->count != 0) {
			/* Was this an NFSv2 write or an NFSv3 stable write? */
			if (resp->verf->committed != NFS_UNSTABLE) {
				/* Resend from where the server left off */
				argp->offset += resp->count;
				argp->pgbase += resp->count;
				argp->count -= resp->count;
			} else {
				/* Resend as a stable write in order to avoid
				 * headaches in the case of a server crash.
				 */
				argp->stable = NFS_FILE_SYNC;
			}
			rpc_restart_call(task);
1128
			return -EAGAIN;
L
Linus Torvalds 已提交
1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
		}
		if (time_before(complain, jiffies)) {
			printk(KERN_WARNING
			       "NFS: Server wrote zero bytes, expected %u.\n",
					argp->count);
			complain = jiffies + 300 * HZ;
		}
		/* Can't do anything about it except throw an error. */
		task->tk_status = -EIO;
	}
1139
	return 0;
L
Linus Torvalds 已提交
1140 1141 1142 1143
}


#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
1144
void nfs_commit_release(void *wdata)
L
Linus Torvalds 已提交
1145 1146 1147 1148 1149 1150 1151 1152
{
	nfs_commit_free(wdata);
}

/*
 * Set up the argument/result storage required for the RPC call.
 */
static void nfs_commit_rpcsetup(struct list_head *head,
1153 1154
		struct nfs_write_data *data,
		int how)
L
Linus Torvalds 已提交
1155
{
1156
	struct nfs_page		*first;
L
Linus Torvalds 已提交
1157
	struct inode		*inode;
1158
	int flags;
L
Linus Torvalds 已提交
1159 1160 1161 1162 1163 1164

	/* Set up the RPC argument and reply structs
	 * NB: take care not to mess about with data->commit et al. */

	list_splice_init(head, &data->pages);
	first = nfs_list_entry(data->pages.next);
1165
	inode = first->wb_context->path.dentry->d_inode;
L
Linus Torvalds 已提交
1166 1167 1168 1169 1170

	data->inode	  = inode;
	data->cred	  = first->wb_context->cred;

	data->args.fh     = NFS_FH(data->inode);
1171 1172 1173 1174
	/* Note: we always request a commit of the entire inode */
	data->args.offset = 0;
	data->args.count  = 0;
	data->res.count   = 0;
L
Linus Torvalds 已提交
1175 1176
	data->res.fattr   = &data->fattr;
	data->res.verf    = &data->verf;
1177
	nfs_fattr_init(&data->fattr);
1178 1179 1180 1181

	/* Set up the initial task struct.  */
	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, &nfs_commit_ops, data);
L
Linus Torvalds 已提交
1182 1183 1184 1185 1186
	NFS_PROTO(inode)->commit_setup(data, how);

	data->task.tk_priority = flush_task_priority(how);
	data->task.tk_cookie = (unsigned long)inode;
	
C
Chuck Lever 已提交
1187
	dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
L
Linus Torvalds 已提交
1188 1189 1190 1191 1192 1193
}

/*
 * Commit dirty pages
 */
static int
1194
nfs_commit_list(struct inode *inode, struct list_head *head, int how)
L
Linus Torvalds 已提交
1195 1196 1197 1198
{
	struct nfs_write_data	*data;
	struct nfs_page         *req;

1199
	data = nfs_commit_alloc();
L
Linus Torvalds 已提交
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213

	if (!data)
		goto out_bad;

	/* Set up the argument struct */
	nfs_commit_rpcsetup(head, data, how);

	nfs_execute_write(data);
	return 0;
 out_bad:
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_commit(req);
T
Trond Myklebust 已提交
1214
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
1215
		nfs_clear_page_tag_locked(req);
L
Linus Torvalds 已提交
1216 1217 1218 1219 1220 1221 1222
	}
	return -ENOMEM;
}

/*
 * COMMIT call returned
 */
1223
static void nfs_commit_done(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1224
{
1225
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
1226 1227
	struct nfs_page		*req;

C
Chuck Lever 已提交
1228
        dprintk("NFS: %5u nfs_commit_done (status %d)\n",
L
Linus Torvalds 已提交
1229 1230
                                task->tk_pid, task->tk_status);

1231 1232 1233 1234
	/* Call the NFS version-specific code */
	if (NFS_PROTO(data->inode)->commit_done(task, data) != 0)
		return;

L
Linus Torvalds 已提交
1235 1236 1237
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
1238
		clear_bit(PG_NEED_COMMIT, &(req)->wb_flags);
1239
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
L
Linus Torvalds 已提交
1240 1241

		dprintk("NFS: commit (%s/%Ld %d@%Ld)",
1242 1243
			req->wb_context->path.dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->path.dentry->d_inode),
L
Linus Torvalds 已提交
1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
			req->wb_bytes,
			(long long)req_offset(req));
		if (task->tk_status < 0) {
			req->wb_context->error = task->tk_status;
			nfs_inode_remove_request(req);
			dprintk(", error = %d\n", task->tk_status);
			goto next;
		}

		/* Okay, COMMIT succeeded, apparently. Check the verifier
		 * returned by the server against all stored verfs. */
		if (!memcmp(req->wb_verf.verifier, data->verf.verifier, sizeof(data->verf.verifier))) {
			/* We have a match */
1257 1258 1259
			/* Set the PG_uptodate flag */
			nfs_mark_uptodate(req->wb_page, req->wb_pgbase,
					req->wb_bytes);
L
Linus Torvalds 已提交
1260 1261 1262 1263 1264 1265
			nfs_inode_remove_request(req);
			dprintk(" OK\n");
			goto next;
		}
		/* We have a mismatch. Write the page again */
		dprintk(" mismatch\n");
1266
		nfs_redirty_request(req);
L
Linus Torvalds 已提交
1267
	next:
1268
		nfs_clear_page_tag_locked(req);
L
Linus Torvalds 已提交
1269 1270
	}
}
1271 1272 1273 1274 1275

static const struct rpc_call_ops nfs_commit_ops = {
	.rpc_call_done = nfs_commit_done,
	.rpc_release = nfs_commit_release,
};
L
Linus Torvalds 已提交
1276

1277
int nfs_commit_inode(struct inode *inode, int how)
L
Linus Torvalds 已提交
1278 1279
{
	LIST_HEAD(head);
T
Trond Myklebust 已提交
1280
	int res;
L
Linus Torvalds 已提交
1281

1282
	spin_lock(&inode->i_lock);
1283
	res = nfs_scan_commit(inode, &head, 0, 0);
1284
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1285
	if (res) {
T
Trond Myklebust 已提交
1286
		int error = nfs_commit_list(inode, &head, how);
1287 1288 1289
		if (error < 0)
			return error;
	}
L
Linus Torvalds 已提交
1290 1291
	return res;
}
1292 1293 1294 1295 1296
#else
static inline int nfs_commit_list(struct inode *inode, struct list_head *head, int how)
{
	return 0;
}
L
Linus Torvalds 已提交
1297 1298
#endif

1299
long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_control *wbc, int how)
L
Linus Torvalds 已提交
1300
{
1301
	struct inode *inode = mapping->host;
1302
	pgoff_t idx_start, idx_end;
1303
	unsigned int npages = 0;
1304
	LIST_HEAD(head);
1305
	int nocommit = how & FLUSH_NOCOMMIT;
T
Trond Myklebust 已提交
1306
	long pages, ret;
L
Linus Torvalds 已提交
1307

1308 1309 1310 1311 1312 1313 1314
	/* FIXME */
	if (wbc->range_cyclic)
		idx_start = 0;
	else {
		idx_start = wbc->range_start >> PAGE_CACHE_SHIFT;
		idx_end = wbc->range_end >> PAGE_CACHE_SHIFT;
		if (idx_end > idx_start) {
1315
			pgoff_t l_npages = 1 + idx_end - idx_start;
1316 1317
			npages = l_npages;
			if (sizeof(npages) != sizeof(l_npages) &&
1318
					(pgoff_t)npages != l_npages)
1319 1320 1321
				npages = 0;
		}
	}
1322
	how &= ~FLUSH_NOCOMMIT;
1323
	spin_lock(&inode->i_lock);
L
Linus Torvalds 已提交
1324
	do {
1325 1326
		ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
		if (ret != 0)
1327
			continue;
1328 1329
		if (nocommit)
			break;
1330
		pages = nfs_scan_commit(inode, &head, idx_start, npages);
1331
		if (pages == 0)
1332
			break;
1333
		if (how & FLUSH_INVALIDATE) {
1334
			spin_unlock(&inode->i_lock);
T
Trond Myklebust 已提交
1335
			nfs_cancel_commit_list(&head);
1336
			ret = pages;
1337
			spin_lock(&inode->i_lock);
1338 1339 1340
			continue;
		}
		pages += nfs_scan_commit(inode, &head, 0, 0);
1341
		spin_unlock(&inode->i_lock);
1342
		ret = nfs_commit_list(inode, &head, how);
1343 1344
		spin_lock(&inode->i_lock);

1345
	} while (ret >= 0);
1346
	spin_unlock(&inode->i_lock);
1347
	return ret;
L
Linus Torvalds 已提交
1348 1349
}

1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
/*
 * flush the inode to disk.
 */
int nfs_wb_all(struct inode *inode)
{
	struct address_space *mapping = inode->i_mapping;
	struct writeback_control wbc = {
		.bdi = mapping->backing_dev_info,
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = LONG_MAX,
1360
		.for_writepages = 1,
1361 1362 1363 1364
		.range_cyclic = 1,
	};
	int ret;

1365
	ret = nfs_writepages(mapping, &wbc);
1366 1367
	if (ret < 0)
		goto out;
1368 1369 1370
	ret = nfs_sync_mapping_wait(mapping, &wbc, 0);
	if (ret >= 0)
		return 0;
1371
out:
1372
	__mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
	return ret;
}

int nfs_sync_mapping_range(struct address_space *mapping, loff_t range_start, loff_t range_end, int how)
{
	struct writeback_control wbc = {
		.bdi = mapping->backing_dev_info,
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = LONG_MAX,
		.range_start = range_start,
		.range_end = range_end,
1384
		.for_writepages = 1,
1385 1386 1387
	};
	int ret;

1388 1389 1390
	ret = nfs_writepages(mapping, &wbc);
	if (ret < 0)
		goto out;
1391 1392 1393
	ret = nfs_sync_mapping_wait(mapping, &wbc, how);
	if (ret >= 0)
		return 0;
1394
out:
1395
	__mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
1396 1397 1398
	return ret;
}

1399
int nfs_wb_page_priority(struct inode *inode, struct page *page, int how)
1400 1401 1402
{
	loff_t range_start = page_offset(page);
	loff_t range_end = range_start + (loff_t)(PAGE_CACHE_SIZE - 1);
1403 1404 1405 1406 1407 1408 1409 1410
	struct writeback_control wbc = {
		.bdi = page->mapping->backing_dev_info,
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = LONG_MAX,
		.range_start = range_start,
		.range_end = range_end,
	};
	int ret;
1411

1412
	BUG_ON(!PageLocked(page));
1413
	if (clear_page_dirty_for_io(page)) {
1414 1415 1416 1417
		ret = nfs_writepage_locked(page, &wbc);
		if (ret < 0)
			goto out;
	}
1418 1419
	if (!PagePrivate(page))
		return 0;
1420 1421 1422 1423
	ret = nfs_sync_mapping_wait(page->mapping, &wbc, how);
	if (ret >= 0)
		return 0;
out:
1424
	__mark_inode_dirty(inode, I_DIRTY_PAGES);
1425
	return ret;
1426 1427 1428 1429 1430 1431 1432
}

/*
 * Write back all requests on one page - we do this before reading it.
 */
int nfs_wb_page(struct inode *inode, struct page* page)
{
1433
	return nfs_wb_page_priority(inode, page, FLUSH_STABLE);
1434 1435
}

T
Trond Myklebust 已提交
1436 1437
int nfs_set_page_dirty(struct page *page)
{
T
Trond Myklebust 已提交
1438 1439
	struct address_space *mapping = page->mapping;
	struct inode *inode;
T
Trond Myklebust 已提交
1440
	struct nfs_page *req;
1441
	int ret;
T
Trond Myklebust 已提交
1442

T
Trond Myklebust 已提交
1443 1444 1445 1446 1447
	if (!mapping)
		goto out_raced;
	inode = mapping->host;
	if (!inode)
		goto out_raced;
1448
	spin_lock(&inode->i_lock);
1449
	req = nfs_page_find_request_locked(page);
T
Trond Myklebust 已提交
1450 1451
	if (req != NULL) {
		/* Mark any existing write requests for flushing */
1452
		ret = !test_and_set_bit(PG_NEED_FLUSH, &req->wb_flags);
1453
		spin_unlock(&inode->i_lock);
T
Trond Myklebust 已提交
1454
		nfs_release_request(req);
1455
		return ret;
T
Trond Myklebust 已提交
1456
	}
1457
	ret = __set_page_dirty_nobuffers(page);
1458
	spin_unlock(&inode->i_lock);
1459
	return ret;
T
Trond Myklebust 已提交
1460 1461
out_raced:
	return !TestSetPageDirty(page);
T
Trond Myklebust 已提交
1462 1463
}

1464

D
David Howells 已提交
1465
int __init nfs_init_writepagecache(void)
L
Linus Torvalds 已提交
1466 1467 1468 1469 1470 1471 1472 1473
{
	nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
					     sizeof(struct nfs_write_data),
					     0, SLAB_HWCACHE_ALIGN,
					     NULL, NULL);
	if (nfs_wdata_cachep == NULL)
		return -ENOMEM;

1474 1475
	nfs_wdata_mempool = mempool_create_slab_pool(MIN_POOL_WRITE,
						     nfs_wdata_cachep);
L
Linus Torvalds 已提交
1476 1477 1478
	if (nfs_wdata_mempool == NULL)
		return -ENOMEM;

1479 1480
	nfs_commit_mempool = mempool_create_slab_pool(MIN_POOL_COMMIT,
						      nfs_wdata_cachep);
L
Linus Torvalds 已提交
1481 1482 1483
	if (nfs_commit_mempool == NULL)
		return -ENOMEM;

1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503
	/*
	 * NFS congestion size, scale with available memory.
	 *
	 *  64MB:    8192k
	 * 128MB:   11585k
	 * 256MB:   16384k
	 * 512MB:   23170k
	 *   1GB:   32768k
	 *   2GB:   46340k
	 *   4GB:   65536k
	 *   8GB:   92681k
	 *  16GB:  131072k
	 *
	 * This allows larger machines to have larger/more transfers.
	 * Limit the default to 256M
	 */
	nfs_congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
	if (nfs_congestion_kb > 256*1024)
		nfs_congestion_kb = 256*1024;

L
Linus Torvalds 已提交
1504 1505 1506
	return 0;
}

1507
void nfs_destroy_writepagecache(void)
L
Linus Torvalds 已提交
1508 1509 1510
{
	mempool_destroy(nfs_commit_mempool);
	mempool_destroy(nfs_wdata_mempool);
1511
	kmem_cache_destroy(nfs_wdata_cachep);
L
Linus Torvalds 已提交
1512 1513
}