write.c 52.9 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>
16
#include <linux/migrate.h>
L
Linus Torvalds 已提交
17 18 19 20 21

#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/nfs_page.h>
22
#include <linux/backing-dev.h>
23
#include <linux/export.h>
24

L
Linus Torvalds 已提交
25 26 27
#include <asm/uaccess.h>

#include "delegation.h"
28
#include "internal.h"
C
Chuck Lever 已提交
29
#include "iostat.h"
30
#include "nfs4_fs.h"
31
#include "fscache.h"
32
#include "pnfs.h"
L
Linus Torvalds 已提交
33

34 35
#include "nfstrace.h"

L
Linus Torvalds 已提交
36 37 38 39 40 41 42 43
#define NFSDBG_FACILITY		NFSDBG_PAGECACHE

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

/*
 * Local function declarations
 */
44
static void nfs_redirty_request(struct nfs_page *req);
45
static const struct rpc_call_ops nfs_commit_ops;
46
static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops;
47
static const struct nfs_commit_completion_ops nfs_commit_completion_ops;
48
static const struct nfs_rw_ops nfs_rw_write_ops;
49
static void nfs_clear_request_commit(struct nfs_page *req);
50 51
static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
				      struct inode *inode);
52 53 54
static struct nfs_page *
nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi,
						struct page *page);
L
Linus Torvalds 已提交
55

56
static struct kmem_cache *nfs_wdata_cachep;
57
static mempool_t *nfs_wdata_mempool;
58
static struct kmem_cache *nfs_cdata_cachep;
L
Linus Torvalds 已提交
59 60
static mempool_t *nfs_commit_mempool;

61
struct nfs_commit_data *nfs_commitdata_alloc(void)
L
Linus Torvalds 已提交
62
{
63
	struct nfs_commit_data *p = mempool_alloc(nfs_commit_mempool, GFP_NOIO);
64

L
Linus Torvalds 已提交
65 66 67 68 69 70
	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
	}
	return p;
}
71
EXPORT_SYMBOL_GPL(nfs_commitdata_alloc);
L
Linus Torvalds 已提交
72

73
void nfs_commit_free(struct nfs_commit_data *p)
L
Linus Torvalds 已提交
74 75 76
{
	mempool_free(p, nfs_commit_mempool);
}
77
EXPORT_SYMBOL_GPL(nfs_commit_free);
L
Linus Torvalds 已提交
78

79
static struct nfs_pgio_header *nfs_writehdr_alloc(void)
80
{
81
	struct nfs_pgio_header *p = mempool_alloc(nfs_wdata_mempool, GFP_NOIO);
82

83
	if (p)
84 85 86
		memset(p, 0, sizeof(*p));
	return p;
}
87

88
static void nfs_writehdr_free(struct nfs_pgio_header *hdr)
89
{
90
	mempool_free(hdr, nfs_wdata_mempool);
91
}
L
Linus Torvalds 已提交
92

93 94 95 96 97 98 99
static void nfs_context_set_write_error(struct nfs_open_context *ctx, int error)
{
	ctx->error = error;
	smp_wmb();
	set_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
}

100 101 102 103 104 105 106
/*
 * nfs_page_find_head_request_locked - find head request associated with @page
 *
 * must be called while holding the inode lock.
 *
 * returns matching head request with reference held, or NULL if not found.
 */
107
static struct nfs_page *
108
nfs_page_find_head_request_locked(struct nfs_inode *nfsi, struct page *page)
109 110 111
{
	struct nfs_page *req = NULL;

112
	if (PagePrivate(page))
113
		req = (struct nfs_page *)page_private(page);
114 115 116
	else if (unlikely(PageSwapCache(page)))
		req = nfs_page_search_commits_for_head_request_locked(nfsi,
			page);
117

118 119
	if (req) {
		WARN_ON_ONCE(req->wb_head != req);
120
		kref_get(&req->wb_kref);
121
	}
122

123 124 125
	return req;
}

126 127 128 129 130 131
/*
 * nfs_page_find_head_request - find head request associated with @page
 *
 * returns matching head request with reference held, or NULL if not found.
 */
static struct nfs_page *nfs_page_find_head_request(struct page *page)
132
{
133
	struct inode *inode = page_file_mapping(page)->host;
134 135
	struct nfs_page *req = NULL;

136
	spin_lock(&inode->i_lock);
137
	req = nfs_page_find_head_request_locked(NFS_I(inode), page);
138
	spin_unlock(&inode->i_lock);
139 140 141
	return req;
}

L
Linus Torvalds 已提交
142 143 144
/* 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)
{
145
	struct inode *inode = page_file_mapping(page)->host;
146 147
	loff_t end, i_size;
	pgoff_t end_index;
L
Linus Torvalds 已提交
148

149 150 151
	spin_lock(&inode->i_lock);
	i_size = i_size_read(inode);
	end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;
152
	if (i_size > 0 && page_file_index(page) < end_index)
153
		goto out;
154
	end = page_file_offset(page) + ((loff_t)offset+count);
L
Linus Torvalds 已提交
155
	if (i_size >= end)
156
		goto out;
L
Linus Torvalds 已提交
157
	i_size_write(inode, end);
158 159 160
	nfs_inc_stats(inode, NFSIOS_EXTENDWRITE);
out:
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
161 162
}

163 164 165
/* A writeback failed: mark the page as bad, and invalidate the page cache */
static void nfs_set_pageerror(struct page *page)
{
166
	nfs_zap_mapping(page_file_mapping(page)->host, page_file_mapping(page));
167 168
}

169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
/*
 * nfs_page_group_search_locked
 * @head - head request of page group
 * @page_offset - offset into page
 *
 * Search page group with head @head to find a request that contains the
 * page offset @page_offset.
 *
 * Returns a pointer to the first matching nfs request, or NULL if no
 * match is found.
 *
 * Must be called with the page group lock held
 */
static struct nfs_page *
nfs_page_group_search_locked(struct nfs_page *head, unsigned int page_offset)
{
	struct nfs_page *req;

	WARN_ON_ONCE(head != head->wb_head);
	WARN_ON_ONCE(!test_bit(PG_HEADLOCK, &head->wb_head->wb_flags));

	req = head;
	do {
		if (page_offset >= req->wb_pgbase &&
		    page_offset < (req->wb_pgbase + req->wb_bytes))
			return req;

		req = req->wb_this_page;
	} while (req != head);

	return NULL;
}

/*
 * nfs_page_group_covers_page
 * @head - head request of page group
 *
 * Return true if the page group with head @head covers the whole page,
 * returns false otherwise
 */
static bool nfs_page_group_covers_page(struct nfs_page *req)
{
	struct nfs_page *tmp;
	unsigned int pos = 0;
	unsigned int len = nfs_page_length(req->wb_page);

215
	nfs_page_group_lock(req, false);
216 217 218 219 220 221 222 223 224 225 226 227 228 229 230

	do {
		tmp = nfs_page_group_search_locked(req->wb_head, pos);
		if (tmp) {
			/* no way this should happen */
			WARN_ON_ONCE(tmp->wb_pgbase != pos);
			pos += tmp->wb_bytes - (pos - tmp->wb_pgbase);
		}
	} while (tmp && pos < len);

	nfs_page_group_unlock(req);
	WARN_ON_ONCE(pos > len);
	return pos == len;
}

L
Linus Torvalds 已提交
231 232 233
/* We can set the PG_uptodate flag if we see that a write request
 * covers the full page.
 */
234
static void nfs_mark_uptodate(struct nfs_page *req)
L
Linus Torvalds 已提交
235
{
236
	if (PageUptodate(req->wb_page))
L
Linus Torvalds 已提交
237
		return;
238
	if (!nfs_page_group_covers_page(req))
L
Linus Torvalds 已提交
239
		return;
240
	SetPageUptodate(req->wb_page);
L
Linus Torvalds 已提交
241 242 243 244
}

static int wb_priority(struct writeback_control *wbc)
{
245
	int ret = 0;
L
Linus Torvalds 已提交
246
	if (wbc->for_reclaim)
247
		return FLUSH_HIGHPRI | FLUSH_STABLE;
248 249
	if (wbc->sync_mode == WB_SYNC_ALL)
		ret = FLUSH_COND_STABLE;
250
	if (wbc->for_kupdate || wbc->for_background)
251 252
		ret |= FLUSH_LOWPRI;
	return ret;
L
Linus Torvalds 已提交
253 254
}

255 256 257 258 259 260 261 262 263 264
/*
 * 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))

265
static void nfs_set_page_writeback(struct page *page)
266
{
267
	struct nfs_server *nfss = NFS_SERVER(page_file_mapping(page)->host);
268 269
	int ret = test_set_page_writeback(page);

270
	WARN_ON_ONCE(ret != 0);
271

272 273 274 275
	if (atomic_long_inc_return(&nfss->writeback) >
			NFS_CONGESTION_ON_THRESH) {
		set_bdi_congested(&nfss->backing_dev_info,
					BLK_RW_ASYNC);
276 277 278
	}
}

279
static void nfs_end_page_writeback(struct nfs_page *req)
280
{
281
	struct inode *inode = page_file_mapping(req->wb_page)->host;
282 283
	struct nfs_server *nfss = NFS_SERVER(inode);

284 285 286 287
	if (!nfs_page_group_sync_on_bit(req, PG_WB_END))
		return;

	end_page_writeback(req->wb_page);
P
Peter Zijlstra 已提交
288
	if (atomic_long_dec_return(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH)
289
		clear_bdi_congested(&nfss->backing_dev_info, BLK_RW_ASYNC);
290 291
}

292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428

/* nfs_page_group_clear_bits
 *   @req - an nfs request
 * clears all page group related bits from @req
 */
static void
nfs_page_group_clear_bits(struct nfs_page *req)
{
	clear_bit(PG_TEARDOWN, &req->wb_flags);
	clear_bit(PG_UNLOCKPAGE, &req->wb_flags);
	clear_bit(PG_UPTODATE, &req->wb_flags);
	clear_bit(PG_WB_END, &req->wb_flags);
	clear_bit(PG_REMOVE, &req->wb_flags);
}


/*
 * nfs_unroll_locks_and_wait -  unlock all newly locked reqs and wait on @req
 *
 * this is a helper function for nfs_lock_and_join_requests
 *
 * @inode - inode associated with request page group, must be holding inode lock
 * @head  - head request of page group, must be holding head lock
 * @req   - request that couldn't lock and needs to wait on the req bit lock
 * @nonblock - if true, don't actually wait
 *
 * NOTE: this must be called holding page_group bit lock and inode spin lock
 *       and BOTH will be released before returning.
 *
 * returns 0 on success, < 0 on error.
 */
static int
nfs_unroll_locks_and_wait(struct inode *inode, struct nfs_page *head,
			  struct nfs_page *req, bool nonblock)
	__releases(&inode->i_lock)
{
	struct nfs_page *tmp;
	int ret;

	/* relinquish all the locks successfully grabbed this run */
	for (tmp = head ; tmp != req; tmp = tmp->wb_this_page)
		nfs_unlock_request(tmp);

	WARN_ON_ONCE(test_bit(PG_TEARDOWN, &req->wb_flags));

	/* grab a ref on the request that will be waited on */
	kref_get(&req->wb_kref);

	nfs_page_group_unlock(head);
	spin_unlock(&inode->i_lock);

	/* release ref from nfs_page_find_head_request_locked */
	nfs_release_request(head);

	if (!nonblock)
		ret = nfs_wait_on_request(req);
	else
		ret = -EAGAIN;
	nfs_release_request(req);

	return ret;
}

/*
 * nfs_destroy_unlinked_subrequests - destroy recently unlinked subrequests
 *
 * @destroy_list - request list (using wb_this_page) terminated by @old_head
 * @old_head - the old head of the list
 *
 * All subrequests must be locked and removed from all lists, so at this point
 * they are only "active" in this function, and possibly in nfs_wait_on_request
 * with a reference held by some other context.
 */
static void
nfs_destroy_unlinked_subrequests(struct nfs_page *destroy_list,
				 struct nfs_page *old_head)
{
	while (destroy_list) {
		struct nfs_page *subreq = destroy_list;

		destroy_list = (subreq->wb_this_page == old_head) ?
				   NULL : subreq->wb_this_page;

		WARN_ON_ONCE(old_head != subreq->wb_head);

		/* make sure old group is not used */
		subreq->wb_head = subreq;
		subreq->wb_this_page = subreq;

		/* subreq is now totally disconnected from page group or any
		 * write / commit lists. last chance to wake any waiters */
		nfs_unlock_request(subreq);

		if (!test_bit(PG_TEARDOWN, &subreq->wb_flags)) {
			/* release ref on old head request */
			nfs_release_request(old_head);

			nfs_page_group_clear_bits(subreq);

			/* release the PG_INODE_REF reference */
			if (test_and_clear_bit(PG_INODE_REF, &subreq->wb_flags))
				nfs_release_request(subreq);
			else
				WARN_ON_ONCE(1);
		} else {
			WARN_ON_ONCE(test_bit(PG_CLEAN, &subreq->wb_flags));
			/* zombie requests have already released the last
			 * reference and were waiting on the rest of the
			 * group to complete. Since it's no longer part of a
			 * group, simply free the request */
			nfs_page_group_clear_bits(subreq);
			nfs_free_request(subreq);
		}
	}
}

/*
 * nfs_lock_and_join_requests - join all subreqs to the head req and return
 *                              a locked reference, cancelling any pending
 *                              operations for this page.
 *
 * @page - the page used to lookup the "page group" of nfs_page structures
 * @nonblock - if true, don't block waiting for request locks
 *
 * This function joins all sub requests to the head request by first
 * locking all requests in the group, cancelling any pending operations
 * and finally updating the head request to cover the whole range covered by
 * the (former) group.  All subrequests are removed from any write or commit
 * lists, unlinked from the group and destroyed.
 *
 * Returns a locked, referenced pointer to the head request - which after
 * this call is guaranteed to be the only request associated with the page.
 * Returns NULL if no requests are found for @page, or a ERR_PTR if an
 * error was encountered.
 */
static struct nfs_page *
nfs_lock_and_join_requests(struct page *page, bool nonblock)
429
{
430
	struct inode *inode = page_file_mapping(page)->host;
431 432 433
	struct nfs_page *head, *subreq;
	struct nfs_page *destroy_list = NULL;
	unsigned int total_bytes;
434 435
	int ret;

436 437 438 439 440
try_again:
	total_bytes = 0;

	WARN_ON_ONCE(destroy_list);

441
	spin_lock(&inode->i_lock);
442 443 444 445 446 447 448 449 450

	/*
	 * A reference is taken only on the head request which acts as a
	 * reference to the whole page group - the group will not be destroyed
	 * until the head reference is released.
	 */
	head = nfs_page_find_head_request_locked(NFS_I(inode), page);

	if (!head) {
451
		spin_unlock(&inode->i_lock);
452 453 454
		return NULL;
	}

455 456
	/* holding inode lock, so always make a non-blocking call to try the
	 * page group lock */
457
	ret = nfs_page_group_lock(head, true);
458 459
	if (ret < 0) {
		spin_unlock(&inode->i_lock);
460 461 462 463 464 465 466

		if (!nonblock && ret == -EAGAIN) {
			nfs_page_group_lock_wait(head);
			nfs_release_request(head);
			goto try_again;
		}

467
		nfs_release_request(head);
468
		return ERR_PTR(ret);
469
	}
470 471

	/* lock each request in the page group */
472 473 474 475
	subreq = head;
	do {
		/*
		 * Subrequests are always contiguous, non overlapping
476
		 * and in order - but may be repeated (mirrored writes).
477
		 */
478 479 480 481 482 483 484 485 486 487
		if (subreq->wb_offset == (head->wb_offset + total_bytes)) {
			/* keep track of how many bytes this group covers */
			total_bytes += subreq->wb_bytes;
		} else if (WARN_ON_ONCE(subreq->wb_offset < head->wb_offset ||
			    ((subreq->wb_offset + subreq->wb_bytes) >
			     (head->wb_offset + total_bytes)))) {
			nfs_page_group_unlock(head);
			spin_unlock(&inode->i_lock);
			return ERR_PTR(-EIO);
		}
488 489 490 491 492 493 494 495 496 497

		if (!nfs_lock_request(subreq)) {
			/* releases page group bit lock and
			 * inode spin lock and all references */
			ret = nfs_unroll_locks_and_wait(inode, head,
				subreq, nonblock);

			if (ret == 0)
				goto try_again;

498
			return ERR_PTR(ret);
499 500 501 502 503 504 505 506 507
		}

		subreq = subreq->wb_this_page;
	} while (subreq != head);

	/* Now that all requests are locked, make sure they aren't on any list.
	 * Commit list removal accounting is done after locks are dropped */
	subreq = head;
	do {
508
		nfs_clear_request_commit(subreq);
509 510 511 512 513 514 515 516 517 518 519 520
		subreq = subreq->wb_this_page;
	} while (subreq != head);

	/* unlink subrequests from head, destroy them later */
	if (head->wb_this_page != head) {
		/* destroy list will be terminated by head */
		destroy_list = head->wb_this_page;
		head->wb_this_page = head;

		/* change head request to cover whole range that
		 * the former page group covered */
		head->wb_bytes = total_bytes;
521
	}
522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537

	/*
	 * prepare head request to be added to new pgio descriptor
	 */
	nfs_page_group_clear_bits(head);

	/*
	 * some part of the group was still on the inode list - otherwise
	 * the group wouldn't be involved in async write.
	 * grab a reference for the head request, iff it needs one.
	 */
	if (!test_and_set_bit(PG_INODE_REF, &head->wb_flags))
		kref_get(&head->wb_kref);

	nfs_page_group_unlock(head);

538
	/* drop lock to clean uprequests on destroy list */
539
	spin_unlock(&inode->i_lock);
540 541 542 543 544 545

	nfs_destroy_unlinked_subrequests(destroy_list, head);

	/* still holds ref on head from nfs_page_find_head_request_locked
	 * and still has lock on head from lock loop */
	return head;
546 547 548 549 550 551 552
}

/*
 * Find an associated nfs write request, and prepare to flush it out
 * May return an error if the user signalled nfs_wait_on_request().
 */
static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
553
				struct page *page, bool nonblock)
554 555 556 557
{
	struct nfs_page *req;
	int ret = 0;

558
	req = nfs_lock_and_join_requests(page, nonblock);
559 560 561 562 563 564
	if (!req)
		goto out;
	ret = PTR_ERR(req);
	if (IS_ERR(req))
		goto out;

565 566
	nfs_set_page_writeback(page);
	WARN_ON_ONCE(test_bit(PG_CLEAN, &req->wb_flags));
567

568
	ret = 0;
569 570
	if (!nfs_pageio_add_request(pgio, req)) {
		nfs_redirty_request(req);
571
		ret = pgio->pg_error;
572
	}
573 574
out:
	return ret;
575 576
}

T
Trond Myklebust 已提交
577
static int nfs_do_writepage(struct page *page, struct writeback_control *wbc, struct nfs_pageio_descriptor *pgio)
L
Linus Torvalds 已提交
578
{
579
	struct inode *inode = page_file_mapping(page)->host;
580
	int ret;
L
Linus Torvalds 已提交
581

C
Chuck Lever 已提交
582
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
583
	nfs_inc_stats(inode, NFSIOS_WRITEPAGES);
C
Chuck Lever 已提交
584

585
	nfs_pageio_cond_complete(pgio, page_file_index(page));
586
	ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE);
587 588 589 590 591
	if (ret == -EAGAIN) {
		redirty_page_for_writepage(wbc, page);
		ret = 0;
	}
	return ret;
T
Trond Myklebust 已提交
592
}
593

T
Trond Myklebust 已提交
594 595 596 597 598 599 600
/*
 * Write an mmapped page to the server.
 */
static int nfs_writepage_locked(struct page *page, struct writeback_control *wbc)
{
	struct nfs_pageio_descriptor pgio;
	int err;
601

602 603
	nfs_pageio_init_write(&pgio, page->mapping->host, wb_priority(wbc),
				false, &nfs_async_write_completion_ops);
T
Trond Myklebust 已提交
604 605 606 607 608 609 610
	err = nfs_do_writepage(page, wbc, &pgio);
	nfs_pageio_complete(&pgio);
	if (err < 0)
		return err;
	if (pgio.pg_error < 0)
		return pgio.pg_error;
	return 0;
611 612 613 614
}

int nfs_writepage(struct page *page, struct writeback_control *wbc)
{
T
Trond Myklebust 已提交
615
	int ret;
616

T
Trond Myklebust 已提交
617
	ret = nfs_writepage_locked(page, wbc);
L
Linus Torvalds 已提交
618
	unlock_page(page);
T
Trond Myklebust 已提交
619 620 621 622 623 624 625 626 627 628
	return ret;
}

static int nfs_writepages_callback(struct page *page, struct writeback_control *wbc, void *data)
{
	int ret;

	ret = nfs_do_writepage(page, wbc, data);
	unlock_page(page);
	return ret;
L
Linus Torvalds 已提交
629 630 631 632 633
}

int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
	struct inode *inode = mapping->host;
634
	unsigned long *bitlock = &NFS_I(inode)->flags;
635
	struct nfs_pageio_descriptor pgio;
L
Linus Torvalds 已提交
636 637
	int err;

638
	/* Stop dirtying of new pages while we sync */
639
	err = wait_on_bit_lock_action(bitlock, NFS_INO_FLUSHING,
640 641 642 643
			nfs_wait_bit_killable, TASK_KILLABLE);
	if (err)
		goto out_err;

C
Chuck Lever 已提交
644 645
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);

646 647
	nfs_pageio_init_write(&pgio, inode, wb_priority(wbc), false,
				&nfs_async_write_completion_ops);
T
Trond Myklebust 已提交
648
	err = write_cache_pages(mapping, wbc, nfs_writepages_callback, &pgio);
649
	nfs_pageio_complete(&pgio);
650 651

	clear_bit_unlock(NFS_INO_FLUSHING, bitlock);
652
	smp_mb__after_atomic();
653 654
	wake_up_bit(bitlock, NFS_INO_FLUSHING);

T
Trond Myklebust 已提交
655
	if (err < 0)
656 657 658 659
		goto out_err;
	err = pgio.pg_error;
	if (err < 0)
		goto out_err;
660
	return 0;
661 662
out_err:
	return err;
L
Linus Torvalds 已提交
663 664 665 666 667
}

/*
 * Insert a write request into an inode
 */
F
Fred Isaman 已提交
668
static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
L
Linus Torvalds 已提交
669 670
{
	struct nfs_inode *nfsi = NFS_I(inode);
671

672 673
	WARN_ON_ONCE(req->wb_this_page != req);

674
	/* Lock the request! */
675
	nfs_lock_request(req);
676 677

	spin_lock(&inode->i_lock);
678 679
	if (!nfsi->nrequests &&
	    NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
680
		inode->i_version++;
681 682 683 684 685 686 687 688 689
	/*
	 * Swap-space should not get truncated. Hence no need to plug the race
	 * with invalidate/truncate.
	 */
	if (likely(!PageSwapCache(req->wb_page))) {
		set_bit(PG_MAPPED, &req->wb_flags);
		SetPagePrivate(req->wb_page);
		set_page_private(req->wb_page, (unsigned long)req);
	}
690
	nfsi->nrequests++;
691
	/* this a head request for a page group - mark it as having an
692 693 694
	 * extra reference so sub groups can follow suit.
	 * This flag also informs pgio layer when to bump nrequests when
	 * adding subrequests. */
695
	WARN_ON(test_and_set_bit(PG_INODE_REF, &req->wb_flags));
696
	kref_get(&req->wb_kref);
697
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
698 699 700
}

/*
701
 * Remove a write request from an inode
L
Linus Torvalds 已提交
702 703 704
 */
static void nfs_inode_remove_request(struct nfs_page *req)
{
705
	struct inode *inode = req->wb_context->dentry->d_inode;
L
Linus Torvalds 已提交
706
	struct nfs_inode *nfsi = NFS_I(inode);
707
	struct nfs_page *head;
L
Linus Torvalds 已提交
708

709 710 711 712 713 714 715
	if (nfs_page_group_sync_on_bit(req, PG_REMOVE)) {
		head = req->wb_head;

		spin_lock(&inode->i_lock);
		if (likely(!PageSwapCache(head->wb_page))) {
			set_page_private(head->wb_page, 0);
			ClearPagePrivate(head->wb_page);
716 717
			smp_mb__after_atomic();
			wake_up_page(head->wb_page, PG_private);
718 719
			clear_bit(PG_MAPPED, &head->wb_flags);
		}
720 721 722 723 724
		nfsi->nrequests--;
		spin_unlock(&inode->i_lock);
	} else {
		spin_lock(&inode->i_lock);
		nfsi->nrequests--;
725
		spin_unlock(&inode->i_lock);
726
	}
727 728 729

	if (test_and_clear_bit(PG_INODE_REF, &req->wb_flags))
		nfs_release_request(req);
L
Linus Torvalds 已提交
730 731
}

732
static void
F
Fred 已提交
733
nfs_mark_request_dirty(struct nfs_page *req)
734 735 736 737
{
	__set_page_dirty_nobuffers(req->wb_page);
}

738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
/*
 * nfs_page_search_commits_for_head_request_locked
 *
 * Search through commit lists on @inode for the head request for @page.
 * Must be called while holding the inode (which is cinfo) lock.
 *
 * Returns the head request if found, or NULL if not found.
 */
static struct nfs_page *
nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi,
						struct page *page)
{
	struct nfs_page *freq, *t;
	struct nfs_commit_info cinfo;
	struct inode *inode = &nfsi->vfs_inode;

	nfs_init_cinfo_from_inode(&cinfo, inode);

	/* search through pnfs commit lists */
	freq = pnfs_search_commit_reqs(inode, &cinfo, page);
	if (freq)
		return freq->wb_head;

	/* Linearly search the commit list for the correct request */
	list_for_each_entry_safe(freq, t, &cinfo.mds->list, wb_list) {
		if (freq->wb_page == page)
			return freq->wb_head;
	}

	return NULL;
}

770 771 772
/**
 * nfs_request_add_commit_list - add request to a commit list
 * @req: pointer to a struct nfs_page
F
Fred Isaman 已提交
773 774
 * @dst: commit list head
 * @cinfo: holds list lock and accounting info
775
 *
F
Fred Isaman 已提交
776
 * This sets the PG_CLEAN bit, updates the cinfo count of
777 778 779
 * number of outstanding requests requiring a commit as well as
 * the MM page stats.
 *
F
Fred Isaman 已提交
780
 * The caller must _not_ hold the cinfo->lock, but must be
781
 * holding the nfs_page lock.
L
Linus Torvalds 已提交
782
 */
783
void
F
Fred Isaman 已提交
784 785
nfs_request_add_commit_list(struct nfs_page *req, struct list_head *dst,
			    struct nfs_commit_info *cinfo)
L
Linus Torvalds 已提交
786
{
787
	set_bit(PG_CLEAN, &(req)->wb_flags);
F
Fred Isaman 已提交
788 789 790 791
	spin_lock(cinfo->lock);
	nfs_list_add_request(req, dst);
	cinfo->mds->ncommit++;
	spin_unlock(cinfo->lock);
792 793
	if (!cinfo->dreq)
		nfs_mark_page_unstable(req->wb_page);
L
Linus Torvalds 已提交
794
}
795 796 797 798 799
EXPORT_SYMBOL_GPL(nfs_request_add_commit_list);

/**
 * nfs_request_remove_commit_list - Remove request from a commit list
 * @req: pointer to a nfs_page
F
Fred Isaman 已提交
800
 * @cinfo: holds list lock and accounting info
801
 *
F
Fred Isaman 已提交
802
 * This clears the PG_CLEAN bit, and updates the cinfo's count of
803 804 805
 * number of outstanding requests requiring a commit
 * It does not update the MM page stats.
 *
F
Fred Isaman 已提交
806
 * The caller _must_ hold the cinfo->lock and the nfs_page lock.
807 808
 */
void
F
Fred Isaman 已提交
809 810
nfs_request_remove_commit_list(struct nfs_page *req,
			       struct nfs_commit_info *cinfo)
811 812 813 814
{
	if (!test_and_clear_bit(PG_CLEAN, &(req)->wb_flags))
		return;
	nfs_list_remove_request(req);
F
Fred Isaman 已提交
815
	cinfo->mds->ncommit--;
816 817 818
}
EXPORT_SYMBOL_GPL(nfs_request_remove_commit_list);

F
Fred Isaman 已提交
819 820 821 822 823 824
static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
				      struct inode *inode)
{
	cinfo->lock = &inode->i_lock;
	cinfo->mds = &NFS_I(inode)->commit_info;
	cinfo->ds = pnfs_get_ds_info(inode);
F
Fred Isaman 已提交
825
	cinfo->dreq = NULL;
826
	cinfo->completion_ops = &nfs_commit_completion_ops;
F
Fred Isaman 已提交
827 828 829 830 831 832
}

void nfs_init_cinfo(struct nfs_commit_info *cinfo,
		    struct inode *inode,
		    struct nfs_direct_req *dreq)
{
833 834 835 836
	if (dreq)
		nfs_init_cinfo_from_dreq(cinfo, dreq);
	else
		nfs_init_cinfo_from_inode(cinfo, inode);
F
Fred Isaman 已提交
837 838
}
EXPORT_SYMBOL_GPL(nfs_init_cinfo);
839 840 841 842

/*
 * Add a request to the inode's commit list.
 */
843
void
F
Fred Isaman 已提交
844
nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg,
845
			struct nfs_commit_info *cinfo, u32 ds_commit_idx)
846
{
847
	if (pnfs_mark_request_commit(req, lseg, cinfo, ds_commit_idx))
848
		return;
F
Fred Isaman 已提交
849
	nfs_request_add_commit_list(req, &cinfo->mds->list, cinfo);
850
}
851

F
Fred Isaman 已提交
852 853 854 855
static void
nfs_clear_page_commit(struct page *page)
{
	dec_zone_page_state(page, NR_UNSTABLE_NFS);
856
	dec_bdi_stat(inode_to_bdi(page_file_mapping(page)->host), BDI_RECLAIMABLE);
F
Fred Isaman 已提交
857 858
}

859
/* Called holding inode (/cinfo) lock */
860
static void
861 862
nfs_clear_request_commit(struct nfs_page *req)
{
863 864
	if (test_bit(PG_CLEAN, &req->wb_flags)) {
		struct inode *inode = req->wb_context->dentry->d_inode;
F
Fred Isaman 已提交
865
		struct nfs_commit_info cinfo;
866

F
Fred Isaman 已提交
867 868 869
		nfs_init_cinfo_from_inode(&cinfo, inode);
		if (!pnfs_clear_request_commit(req, &cinfo)) {
			nfs_request_remove_commit_list(req, &cinfo);
870
		}
F
Fred Isaman 已提交
871
		nfs_clear_page_commit(req->wb_page);
872 873 874
	}
}

875
int nfs_write_need_commit(struct nfs_pgio_header *hdr)
876
{
877
	if (hdr->verf.committed == NFS_DATA_SYNC)
878
		return hdr->lseg == NULL;
879
	return hdr->verf.committed != NFS_FILE_SYNC;
880 881
}

882
static void nfs_write_completion(struct nfs_pgio_header *hdr)
883
{
F
Fred Isaman 已提交
884
	struct nfs_commit_info cinfo;
885 886 887 888
	unsigned long bytes = 0;

	if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
		goto out;
F
Fred Isaman 已提交
889
	nfs_init_cinfo_from_inode(&cinfo, hdr->inode);
890 891 892 893 894 895 896
	while (!list_empty(&hdr->pages)) {
		struct nfs_page *req = nfs_list_entry(hdr->pages.next);

		bytes += req->wb_bytes;
		nfs_list_remove_request(req);
		if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) &&
		    (hdr->good_bytes < bytes)) {
897
			nfs_set_pageerror(req->wb_page);
898 899 900
			nfs_context_set_write_error(req->wb_context, hdr->error);
			goto remove_req;
		}
901
		if (nfs_write_need_commit(hdr)) {
902
			memcpy(&req->wb_verf, &hdr->verf.verifier, sizeof(req->wb_verf));
903
			nfs_mark_request_commit(req, hdr->lseg, &cinfo,
904
				hdr->pgio_mirror_idx);
905 906 907 908 909
			goto next;
		}
remove_req:
		nfs_inode_remove_request(req);
next:
910
		nfs_unlock_request(req);
911
		nfs_end_page_writeback(req);
912
		nfs_release_request(req);
913 914 915
	}
out:
	hdr->release(hdr);
916
}
L
Linus Torvalds 已提交
917

918
unsigned long
F
Fred Isaman 已提交
919
nfs_reqs_to_commit(struct nfs_commit_info *cinfo)
920
{
F
Fred Isaman 已提交
921
	return cinfo->mds->ncommit;
F
Fred Isaman 已提交
922 923
}

F
Fred Isaman 已提交
924
/* cinfo->lock held by caller */
925
int
F
Fred Isaman 已提交
926 927
nfs_scan_commit_list(struct list_head *src, struct list_head *dst,
		     struct nfs_commit_info *cinfo, int max)
F
Fred Isaman 已提交
928 929 930 931 932
{
	struct nfs_page *req, *tmp;
	int ret = 0;

	list_for_each_entry_safe(req, tmp, src, wb_list) {
933 934
		if (!nfs_lock_request(req))
			continue;
935
		kref_get(&req->wb_kref);
F
Fred Isaman 已提交
936
		if (cond_resched_lock(cinfo->lock))
937
			list_safe_reset_next(req, tmp, wb_list);
F
Fred Isaman 已提交
938
		nfs_request_remove_commit_list(req, cinfo);
939 940
		nfs_list_add_request(req, dst);
		ret++;
941
		if ((ret == max) && !cinfo->dreq)
942
			break;
F
Fred Isaman 已提交
943 944
	}
	return ret;
945 946
}

L
Linus Torvalds 已提交
947 948 949
/*
 * nfs_scan_commit - Scan an inode for commit requests
 * @inode: NFS inode to scan
F
Fred Isaman 已提交
950 951
 * @dst: mds destination list
 * @cinfo: mds and ds lists of reqs ready to commit
L
Linus Torvalds 已提交
952 953 954 955
 *
 * Moves requests from the inode's 'commit' request list.
 * The requests are *not* checked to ensure that they form a contiguous set.
 */
956
int
F
Fred Isaman 已提交
957 958
nfs_scan_commit(struct inode *inode, struct list_head *dst,
		struct nfs_commit_info *cinfo)
L
Linus Torvalds 已提交
959
{
F
Fred Isaman 已提交
960
	int ret = 0;
961

F
Fred Isaman 已提交
962 963
	spin_lock(cinfo->lock);
	if (cinfo->mds->ncommit > 0) {
964
		const int max = INT_MAX;
F
Fred Isaman 已提交
965

F
Fred Isaman 已提交
966 967 968
		ret = nfs_scan_commit_list(&cinfo->mds->list, dst,
					   cinfo, max);
		ret += pnfs_scan_commit_lists(inode, cinfo, max - ret);
F
Fred Isaman 已提交
969
	}
F
Fred Isaman 已提交
970
	spin_unlock(cinfo->lock);
971
	return ret;
L
Linus Torvalds 已提交
972
}
F
Fred Isaman 已提交
973

L
Linus Torvalds 已提交
974
/*
975 976
 * Search for an existing write request, and attempt to update
 * it to reflect a new dirty region on a given page.
L
Linus Torvalds 已提交
977
 *
978 979
 * If the attempt fails, then the existing request is flushed out
 * to disk.
L
Linus Torvalds 已提交
980
 */
981 982 983 984
static struct nfs_page *nfs_try_to_update_request(struct inode *inode,
		struct page *page,
		unsigned int offset,
		unsigned int bytes)
L
Linus Torvalds 已提交
985
{
986 987 988 989 990 991 992
	struct nfs_page *req;
	unsigned int rqend;
	unsigned int end;
	int error;

	if (!PagePrivate(page))
		return NULL;
L
Linus Torvalds 已提交
993 994

	end = offset + bytes;
995
	spin_lock(&inode->i_lock);
L
Linus Torvalds 已提交
996 997

	for (;;) {
998
		req = nfs_page_find_head_request_locked(NFS_I(inode), page);
999 1000 1001
		if (req == NULL)
			goto out_unlock;

1002 1003 1004 1005
		/* should be handled by nfs_flush_incompatible */
		WARN_ON_ONCE(req->wb_head != req);
		WARN_ON_ONCE(req->wb_this_page != req);

1006 1007 1008 1009 1010 1011 1012
		rqend = req->wb_offset + req->wb_bytes;
		/*
		 * Tell the caller to flush out the request if
		 * the offsets are non-contiguous.
		 * Note: nfs_flush_incompatible() will already
		 * have flushed out requests having wrong owners.
		 */
1013
		if (offset > rqend
1014 1015 1016
		    || end < req->wb_offset)
			goto out_flushme;

1017
		if (nfs_lock_request(req))
L
Linus Torvalds 已提交
1018 1019
			break;

1020
		/* The request is locked, so wait and then retry */
1021
		spin_unlock(&inode->i_lock);
1022 1023 1024 1025 1026
		error = nfs_wait_on_request(req);
		nfs_release_request(req);
		if (error != 0)
			goto out_err;
		spin_lock(&inode->i_lock);
L
Linus Torvalds 已提交
1027 1028 1029 1030 1031 1032 1033 1034 1035
	}

	/* Okay, the request matches. Update the region */
	if (offset < req->wb_offset) {
		req->wb_offset = offset;
		req->wb_pgbase = offset;
	}
	if (end > rqend)
		req->wb_bytes = end - req->wb_offset;
1036 1037 1038
	else
		req->wb_bytes = rqend - req->wb_offset;
out_unlock:
1039 1040
	if (req)
		nfs_clear_request_commit(req);
1041
	spin_unlock(&inode->i_lock);
1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
	return req;
out_flushme:
	spin_unlock(&inode->i_lock);
	nfs_release_request(req);
	error = nfs_wb_page(inode, page);
out_err:
	return ERR_PTR(error);
}

/*
 * Try to update an existing write request, or create one if there is none.
 *
 * Note: Should always be called with the Page Lock held to prevent races
 * if we have to add a new request. Also assumes that the caller has
 * already called nfs_flush_incompatible() if necessary.
 */
static struct nfs_page * nfs_setup_write_request(struct nfs_open_context* ctx,
		struct page *page, unsigned int offset, unsigned int bytes)
{
1061
	struct inode *inode = page_file_mapping(page)->host;
1062
	struct nfs_page	*req;
L
Linus Torvalds 已提交
1063

1064 1065 1066
	req = nfs_try_to_update_request(inode, page, offset, bytes);
	if (req != NULL)
		goto out;
1067
	req = nfs_create_request(ctx, page, NULL, offset, bytes);
1068 1069
	if (IS_ERR(req))
		goto out;
F
Fred Isaman 已提交
1070
	nfs_inode_add_request(inode, req);
1071
out:
T
Trond Myklebust 已提交
1072
	return req;
L
Linus Torvalds 已提交
1073 1074
}

1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
static int nfs_writepage_setup(struct nfs_open_context *ctx, struct page *page,
		unsigned int offset, unsigned int count)
{
	struct nfs_page	*req;

	req = nfs_setup_write_request(ctx, page, offset, count);
	if (IS_ERR(req))
		return PTR_ERR(req);
	/* Update file length */
	nfs_grow_file(page, offset, count);
1085
	nfs_mark_uptodate(req);
1086
	nfs_mark_request_dirty(req);
1087
	nfs_unlock_and_release_request(req);
1088 1089 1090
	return 0;
}

L
Linus Torvalds 已提交
1091 1092
int nfs_flush_incompatible(struct file *file, struct page *page)
{
1093
	struct nfs_open_context *ctx = nfs_file_open_context(file);
1094
	struct nfs_lock_context *l_ctx;
1095
	struct file_lock_context *flctx = file_inode(file)->i_flctx;
L
Linus Torvalds 已提交
1096
	struct nfs_page	*req;
T
Trond Myklebust 已提交
1097
	int do_flush, status;
L
Linus Torvalds 已提交
1098 1099 1100 1101 1102 1103 1104 1105
	/*
	 * 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
1106
	do {
1107
		req = nfs_page_find_head_request(page);
T
Trond Myklebust 已提交
1108 1109
		if (req == NULL)
			return 0;
1110 1111
		l_ctx = req->wb_lock_context;
		do_flush = req->wb_page != page || req->wb_context != ctx;
1112 1113
		/* for now, flush if more than 1 request in page_group */
		do_flush |= req->wb_this_page != req;
1114 1115 1116
		if (l_ctx && flctx &&
		    !(list_empty_careful(&flctx->flc_posix) &&
		      list_empty_careful(&flctx->flc_flock))) {
1117 1118 1119
			do_flush |= l_ctx->lockowner.l_owner != current->files
				|| l_ctx->lockowner.l_pid != current->tgid;
		}
L
Linus Torvalds 已提交
1120
		nfs_release_request(req);
T
Trond Myklebust 已提交
1121 1122
		if (!do_flush)
			return 0;
1123
		status = nfs_wb_page(page_file_mapping(page)->host, page);
T
Trond Myklebust 已提交
1124 1125
	} while (status == 0);
	return status;
L
Linus Torvalds 已提交
1126 1127
}

1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
/*
 * Avoid buffered writes when a open context credential's key would
 * expire soon.
 *
 * Returns -EACCES if the key will expire within RPC_KEY_EXPIRE_FAIL.
 *
 * Return 0 and set a credential flag which triggers the inode to flush
 * and performs  NFS_FILE_SYNC writes if the key will expired within
 * RPC_KEY_EXPIRE_TIMEO.
 */
int
nfs_key_timeout_notify(struct file *filp, struct inode *inode)
{
	struct nfs_open_context *ctx = nfs_file_open_context(filp);
	struct rpc_auth *auth = NFS_SERVER(inode)->client->cl_auth;

	return rpcauth_key_timeout_notify(auth, ctx->cred);
}

/*
 * Test if the open context credential key is marked to expire soon.
 */
bool nfs_ctx_key_to_expire(struct nfs_open_context *ctx)
{
	return rpcauth_cred_key_to_expire(ctx->cred);
}

1155 1156 1157 1158 1159
/*
 * If the page cache is marked as unsafe or invalid, then we can't rely on
 * the PageUptodate() flag. In this case, we will need to turn off
 * write optimisations that depend on the page contents being correct.
 */
1160
static bool nfs_write_pageuptodate(struct page *page, struct inode *inode)
1161
{
1162 1163
	struct nfs_inode *nfsi = NFS_I(inode);

1164 1165
	if (nfs_have_delegated_attributes(inode))
		goto out;
1166
	if (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)
1167
		return false;
1168
	smp_rmb();
1169
	if (test_bit(NFS_INO_INVALIDATING, &nfsi->flags))
1170 1171
		return false;
out:
1172 1173
	if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
		return false;
1174
	return PageUptodate(page) != 0;
1175 1176
}

1177 1178 1179 1180 1181 1182 1183
static bool
is_whole_file_wrlock(struct file_lock *fl)
{
	return fl->fl_start == 0 && fl->fl_end == OFFSET_MAX &&
			fl->fl_type == F_WRLCK;
}

1184 1185 1186 1187 1188
/* If we know the page is up to date, and we're not using byte range locks (or
 * if we have the whole file locked for writing), it may be more efficient to
 * extend the write to cover the entire page in order to avoid fragmentation
 * inefficiencies.
 *
1189 1190
 * If the file is opened for synchronous writes then we can just skip the rest
 * of the checks.
1191 1192 1193
 */
static int nfs_can_extend_write(struct file *file, struct page *page, struct inode *inode)
{
1194 1195 1196 1197
	int ret;
	struct file_lock_context *flctx = inode->i_flctx;
	struct file_lock *fl;

1198 1199
	if (file->f_flags & O_DSYNC)
		return 0;
1200 1201
	if (!nfs_write_pageuptodate(page, inode))
		return 0;
1202 1203
	if (NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
		return 1;
1204 1205
	if (!flctx || (list_empty_careful(&flctx->flc_flock) &&
		       list_empty_careful(&flctx->flc_posix)))
1206 1207 1208 1209
		return 0;

	/* Check to see if there are whole file write locks */
	ret = 0;
1210
	spin_lock(&flctx->flc_lock);
1211 1212 1213 1214 1215 1216
	if (!list_empty(&flctx->flc_posix)) {
		fl = list_first_entry(&flctx->flc_posix, struct file_lock,
					fl_list);
		if (is_whole_file_wrlock(fl))
			ret = 1;
	} else if (!list_empty(&flctx->flc_flock)) {
1217 1218 1219 1220 1221
		fl = list_first_entry(&flctx->flc_flock, struct file_lock,
					fl_list);
		if (fl->fl_type == F_WRLCK)
			ret = 1;
	}
1222
	spin_unlock(&flctx->flc_lock);
1223
	return ret;
1224 1225
}

L
Linus Torvalds 已提交
1226 1227 1228 1229 1230 1231 1232 1233 1234
/*
 * 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)
{
1235
	struct nfs_open_context *ctx = nfs_file_open_context(file);
1236
	struct inode	*inode = page_file_mapping(page)->host;
L
Linus Torvalds 已提交
1237 1238
	int		status = 0;

C
Chuck Lever 已提交
1239 1240
	nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);

1241 1242
	dprintk("NFS:       nfs_updatepage(%pD2 %d@%lld)\n",
		file, count, (long long)(page_file_offset(page) + offset));
L
Linus Torvalds 已提交
1243

1244
	if (nfs_can_extend_write(file, page, inode)) {
1245
		count = max(count + offset, nfs_page_length(page));
L
Linus Torvalds 已提交
1246 1247 1248
		offset = 0;
	}

1249
	status = nfs_writepage_setup(ctx, page, offset, count);
1250 1251
	if (status < 0)
		nfs_set_pageerror(page);
1252 1253
	else
		__set_page_dirty_nobuffers(page);
L
Linus Torvalds 已提交
1254

1255
	dprintk("NFS:       nfs_updatepage returns %d (isize %lld)\n",
L
Linus Torvalds 已提交
1256 1257 1258 1259
			status, (long long)i_size_read(inode));
	return status;
}

1260
static int flush_task_priority(int how)
L
Linus Torvalds 已提交
1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
{
	switch (how & (FLUSH_HIGHPRI|FLUSH_LOWPRI)) {
		case FLUSH_HIGHPRI:
			return RPC_PRIORITY_HIGH;
		case FLUSH_LOWPRI:
			return RPC_PRIORITY_LOW;
	}
	return RPC_PRIORITY_NORMAL;
}

1271 1272
static void nfs_initiate_write(struct nfs_pgio_header *hdr,
			       struct rpc_message *msg,
1273
			       const struct nfs_rpc_ops *rpc_ops,
1274
			       struct rpc_task_setup *task_setup_data, int how)
L
Linus Torvalds 已提交
1275
{
1276
	int priority = flush_task_priority(how);
1277

1278
	task_setup_data->priority = priority;
1279
	rpc_ops->write_setup(hdr, msg);
1280

1281
	nfs4_state_protect_write(NFS_SERVER(hdr->inode)->nfs_client,
1282
				 &task_setup_data->rpc_client, msg, hdr);
1283 1284
}

F
Fred 已提交
1285 1286 1287 1288 1289 1290 1291
/* If a nfs_flush_* function fails, it should remove reqs from @head and
 * call this on each, which will prepare them to be retried on next
 * writeback using standard nfs.
 */
static void nfs_redirty_request(struct nfs_page *req)
{
	nfs_mark_request_dirty(req);
1292
	nfs_unlock_request(req);
1293
	nfs_end_page_writeback(req);
1294
	nfs_release_request(req);
F
Fred 已提交
1295 1296
}

1297
static void nfs_async_write_error(struct list_head *head)
1298 1299 1300 1301 1302 1303 1304 1305 1306 1307
{
	struct nfs_page	*req;

	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_redirty_request(req);
	}
}

1308 1309 1310 1311 1312
static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops = {
	.error_cleanup = nfs_async_write_error,
	.completion = nfs_write_completion,
};

1313
void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio,
1314
			       struct inode *inode, int ioflags, bool force_mds,
1315
			       const struct nfs_pgio_completion_ops *compl_ops)
L
Linus Torvalds 已提交
1316
{
1317
	struct nfs_server *server = NFS_SERVER(inode);
1318
	const struct nfs_pageio_ops *pg_ops = &nfs_pgio_rw_ops;
1319 1320 1321 1322 1323

#ifdef CONFIG_NFS_V4_1
	if (server->pnfs_curr_ld && !force_mds)
		pg_ops = server->pnfs_curr_ld->pg_write_ops;
#endif
1324 1325
	nfs_pageio_init(pgio, inode, pg_ops, compl_ops, &nfs_rw_write_ops,
			server->wsize, ioflags);
1326
}
B
Bryan Schumaker 已提交
1327
EXPORT_SYMBOL_GPL(nfs_pageio_init_write);
L
Linus Torvalds 已提交
1328

1329 1330
void nfs_pageio_reset_write_mds(struct nfs_pageio_descriptor *pgio)
{
1331 1332
	struct nfs_pgio_mirror *mirror;

1333
	pgio->pg_ops = &nfs_pgio_rw_ops;
1334 1335 1336 1337 1338

	nfs_pageio_stop_mirroring(pgio);

	mirror = &pgio->pg_mirrors[0];
	mirror->pg_bsize = NFS_SERVER(pgio->pg_inode)->wsize;
1339
}
1340
EXPORT_SYMBOL_GPL(nfs_pageio_reset_write_mds);
1341

L
Linus Torvalds 已提交
1342

1343 1344 1345 1346 1347 1348 1349
void nfs_commit_prepare(struct rpc_task *task, void *calldata)
{
	struct nfs_commit_data *data = calldata;

	NFS_PROTO(data->inode)->commit_rpc_prepare(task, data);
}

1350
static void nfs_writeback_release_common(struct nfs_pgio_header *hdr)
1351
{
1352
	/* do nothing! */
L
Linus Torvalds 已提交
1353 1354
}

1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365
/*
 * Special version of should_remove_suid() that ignores capabilities.
 */
static int nfs_should_remove_suid(const struct inode *inode)
{
	umode_t mode = inode->i_mode;
	int kill = 0;

	/* suid always must be killed */
	if (unlikely(mode & S_ISUID))
		kill = ATTR_KILL_SUID;
1366

1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
	/*
	 * sgid without any exec bits is just a mandatory locking mark; leave
	 * it alone.  If some exec bits are set, it's a real sgid; kill it.
	 */
	if (unlikely((mode & S_ISGID) && (mode & S_IXGRP)))
		kill |= ATTR_KILL_SGID;

	if (unlikely(kill && S_ISREG(mode)))
		return kill;

	return 0;
}
1379

1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409
static void nfs_writeback_check_extend(struct nfs_pgio_header *hdr,
		struct nfs_fattr *fattr)
{
	struct nfs_pgio_args *argp = &hdr->args;
	struct nfs_pgio_res *resp = &hdr->res;

	if (!(fattr->valid & NFS_ATTR_FATTR_SIZE))
		return;
	if (argp->offset + resp->count != fattr->size)
		return;
	if (nfs_size_to_loff_t(fattr->size) < i_size_read(hdr->inode))
		return;
	/* Set attribute barrier */
	nfs_fattr_set_barrier(fattr);
}

void nfs_writeback_update_inode(struct nfs_pgio_header *hdr)
{
	struct nfs_fattr *fattr = hdr->res.fattr;
	struct inode *inode = hdr->inode;

	if (fattr == NULL)
		return;
	spin_lock(&inode->i_lock);
	nfs_writeback_check_extend(hdr, fattr);
	nfs_post_op_update_inode_force_wcc_locked(inode, fattr);
	spin_unlock(&inode->i_lock);
}
EXPORT_SYMBOL_GPL(nfs_writeback_update_inode);

L
Linus Torvalds 已提交
1410 1411 1412
/*
 * This function is called when the WRITE call is complete.
 */
1413 1414
static int nfs_writeback_done(struct rpc_task *task,
			      struct nfs_pgio_header *hdr,
1415
			      struct inode *inode)
L
Linus Torvalds 已提交
1416
{
1417
	int status;
L
Linus Torvalds 已提交
1418

1419 1420 1421 1422 1423 1424 1425
	/*
	 * ->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.
	 */
1426
	status = NFS_PROTO(inode)->write_done(task, hdr);
1427
	if (status != 0)
1428
		return status;
1429
	nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, hdr->res.count);
C
Chuck Lever 已提交
1430

1431 1432
	if (hdr->res.verf->committed < hdr->args.stable &&
	    task->tk_status >= 0) {
L
Linus Torvalds 已提交
1433 1434 1435 1436 1437 1438 1439 1440 1441 1442
		/* 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;

1443
		/* Note this will print the MDS for a DS write */
L
Linus Torvalds 已提交
1444
		if (time_before(complain, jiffies)) {
1445
			dprintk("NFS:       faulty NFS server %s:"
L
Linus Torvalds 已提交
1446
				" (committed = %d) != (stable = %d)\n",
1447
				NFS_SERVER(inode)->nfs_client->cl_hostname,
1448
				hdr->res.verf->committed, hdr->args.stable);
L
Linus Torvalds 已提交
1449 1450 1451
			complain = jiffies + 300 * HZ;
		}
	}
1452 1453 1454 1455

	/* Deal with the suid/sgid bit corner case */
	if (nfs_should_remove_suid(inode))
		nfs_mark_for_revalidate(inode);
1456 1457 1458 1459 1460 1461
	return 0;
}

/*
 * This function is called when the WRITE call is complete.
 */
1462 1463
static void nfs_writeback_result(struct rpc_task *task,
				 struct nfs_pgio_header *hdr)
1464
{
1465 1466
	struct nfs_pgio_args	*argp = &hdr->args;
	struct nfs_pgio_res	*resp = &hdr->res;
1467 1468

	if (resp->count < argp->count) {
L
Linus Torvalds 已提交
1469 1470
		static unsigned long    complain;

1471
		/* This a short write! */
1472
		nfs_inc_stats(hdr->inode, NFSIOS_SHORTWRITE);
C
Chuck Lever 已提交
1473

L
Linus Torvalds 已提交
1474
		/* Has the server at least made some progress? */
1475 1476 1477 1478 1479 1480
		if (resp->count == 0) {
			if (time_before(complain, jiffies)) {
				printk(KERN_WARNING
				       "NFS: Server wrote zero bytes, expected %u.\n",
				       argp->count);
				complain = jiffies + 300 * HZ;
L
Linus Torvalds 已提交
1481
			}
1482
			nfs_set_pgio_error(hdr, -EIO, argp->offset);
1483
			task->tk_status = -EIO;
1484
			return;
L
Linus Torvalds 已提交
1485
		}
1486 1487 1488
		/* Was this an NFSv2 write or an NFSv3 stable write? */
		if (resp->verf->committed != NFS_UNSTABLE) {
			/* Resend from where the server left off */
1489
			hdr->mds_offset += resp->count;
1490 1491 1492 1493 1494 1495 1496 1497
			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;
L
Linus Torvalds 已提交
1498
		}
1499
		rpc_restart_call_prepare(task);
L
Linus Torvalds 已提交
1500 1501 1502 1503
	}
}


1504 1505
static int nfs_commit_set_lock(struct nfs_inode *nfsi, int may_wait)
{
1506 1507
	int ret;

1508 1509
	if (!test_and_set_bit(NFS_INO_COMMIT, &nfsi->flags))
		return 1;
1510 1511 1512 1513 1514 1515 1516
	if (!may_wait)
		return 0;
	ret = out_of_line_wait_on_bit_lock(&nfsi->flags,
				NFS_INO_COMMIT,
				nfs_wait_bit_killable,
				TASK_KILLABLE);
	return (ret < 0) ? ret : 1;
1517 1518
}

1519
static void nfs_commit_clear_lock(struct nfs_inode *nfsi)
1520 1521
{
	clear_bit(NFS_INO_COMMIT, &nfsi->flags);
1522
	smp_mb__after_atomic();
1523 1524 1525
	wake_up_bit(&nfsi->flags, NFS_INO_COMMIT);
}

1526
void nfs_commitdata_release(struct nfs_commit_data *data)
L
Linus Torvalds 已提交
1527
{
1528 1529
	put_nfs_open_context(data->context);
	nfs_commit_free(data);
L
Linus Torvalds 已提交
1530
}
1531
EXPORT_SYMBOL_GPL(nfs_commitdata_release);
L
Linus Torvalds 已提交
1532

1533
int nfs_initiate_commit(struct rpc_clnt *clnt, struct nfs_commit_data *data,
1534
			const struct nfs_rpc_ops *nfs_ops,
1535
			const struct rpc_call_ops *call_ops,
1536
			int how, int flags)
L
Linus Torvalds 已提交
1537
{
1538
	struct rpc_task *task;
1539
	int priority = flush_task_priority(how);
1540 1541 1542
	struct rpc_message msg = {
		.rpc_argp = &data->args,
		.rpc_resp = &data->res,
1543
		.rpc_cred = data->cred,
1544
	};
1545
	struct rpc_task_setup task_setup_data = {
1546
		.task = &data->task,
1547
		.rpc_client = clnt,
1548
		.rpc_message = &msg,
1549
		.callback_ops = call_ops,
1550
		.callback_data = data,
1551
		.workqueue = nfsiod_workqueue,
1552
		.flags = RPC_TASK_ASYNC | flags,
1553
		.priority = priority,
1554
	};
1555
	/* Set up the initial task struct.  */
1556
	nfs_ops->commit_setup(data, &msg);
1557 1558 1559

	dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);

1560 1561 1562
	nfs4_state_protect(NFS_SERVER(data->inode)->nfs_client,
		NFS_SP4_MACH_CRED_COMMIT, &task_setup_data.rpc_client, &msg);

1563 1564 1565 1566 1567 1568 1569 1570
	task = rpc_run_task(&task_setup_data);
	if (IS_ERR(task))
		return PTR_ERR(task);
	if (how & FLUSH_SYNC)
		rpc_wait_for_completion_task(task);
	rpc_put_task(task);
	return 0;
}
1571
EXPORT_SYMBOL_GPL(nfs_initiate_commit);
1572

1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584
static loff_t nfs_get_lwb(struct list_head *head)
{
	loff_t lwb = 0;
	struct nfs_page *req;

	list_for_each_entry(req, head, wb_list)
		if (lwb < (req_offset(req) + req->wb_bytes))
			lwb = req_offset(req) + req->wb_bytes;

	return lwb;
}

1585 1586 1587
/*
 * Set up the argument/result storage required for the RPC call.
 */
1588
void nfs_init_commit(struct nfs_commit_data *data,
1589 1590 1591
		     struct list_head *head,
		     struct pnfs_layout_segment *lseg,
		     struct nfs_commit_info *cinfo)
1592 1593
{
	struct nfs_page *first = nfs_list_entry(head->next);
1594
	struct inode *inode = first->wb_context->dentry->d_inode;
L
Linus Torvalds 已提交
1595 1596 1597 1598 1599 1600 1601

	/* 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);

	data->inode	  = inode;
1602
	data->cred	  = first->wb_context->cred;
1603
	data->lseg	  = lseg; /* reference transferred */
1604 1605 1606
	/* only set lwb for pnfs commit */
	if (lseg)
		data->lwb = nfs_get_lwb(&data->pages);
1607
	data->mds_ops     = &nfs_commit_ops;
1608
	data->completion_ops = cinfo->completion_ops;
F
Fred Isaman 已提交
1609
	data->dreq	  = cinfo->dreq;
L
Linus Torvalds 已提交
1610 1611

	data->args.fh     = NFS_FH(data->inode);
1612 1613 1614
	/* Note: we always request a commit of the entire inode */
	data->args.offset = 0;
	data->args.count  = 0;
1615
	data->context     = get_nfs_open_context(first->wb_context);
L
Linus Torvalds 已提交
1616 1617
	data->res.fattr   = &data->fattr;
	data->res.verf    = &data->verf;
1618
	nfs_fattr_init(&data->fattr);
L
Linus Torvalds 已提交
1619
}
1620
EXPORT_SYMBOL_GPL(nfs_init_commit);
L
Linus Torvalds 已提交
1621

1622
void nfs_retry_commit(struct list_head *page_list,
F
Fred Isaman 已提交
1623
		      struct pnfs_layout_segment *lseg,
1624 1625
		      struct nfs_commit_info *cinfo,
		      u32 ds_commit_idx)
1626 1627 1628 1629 1630 1631
{
	struct nfs_page *req;

	while (!list_empty(page_list)) {
		req = nfs_list_entry(page_list->next);
		nfs_list_remove_request(req);
1632
		nfs_mark_request_commit(req, lseg, cinfo, ds_commit_idx);
1633 1634
		if (!cinfo->dreq)
			nfs_clear_page_commit(req->wb_page);
1635
		nfs_unlock_and_release_request(req);
1636 1637
	}
}
1638
EXPORT_SYMBOL_GPL(nfs_retry_commit);
1639

L
Linus Torvalds 已提交
1640 1641 1642 1643
/*
 * Commit dirty pages
 */
static int
F
Fred Isaman 已提交
1644 1645
nfs_commit_list(struct inode *inode, struct list_head *head, int how,
		struct nfs_commit_info *cinfo)
L
Linus Torvalds 已提交
1646
{
1647
	struct nfs_commit_data	*data;
L
Linus Torvalds 已提交
1648

1649
	data = nfs_commitdata_alloc();
L
Linus Torvalds 已提交
1650 1651 1652 1653 1654

	if (!data)
		goto out_bad;

	/* Set up the argument struct */
1655 1656
	nfs_init_commit(data, head, NULL, cinfo);
	atomic_inc(&cinfo->mds->rpcs_out);
1657 1658
	return nfs_initiate_commit(NFS_CLIENT(inode), data, NFS_PROTO(inode),
				   data->mds_ops, how, 0);
L
Linus Torvalds 已提交
1659
 out_bad:
1660
	nfs_retry_commit(head, NULL, cinfo, 0);
1661
	cinfo->completion_ops->error_cleanup(NFS_I(inode));
L
Linus Torvalds 已提交
1662 1663 1664 1665 1666 1667
	return -ENOMEM;
}

/*
 * COMMIT call returned
 */
1668
static void nfs_commit_done(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1669
{
1670
	struct nfs_commit_data	*data = calldata;
L
Linus Torvalds 已提交
1671

C
Chuck Lever 已提交
1672
        dprintk("NFS: %5u nfs_commit_done (status %d)\n",
L
Linus Torvalds 已提交
1673 1674
                                task->tk_pid, task->tk_status);

1675
	/* Call the NFS version-specific code */
1676
	NFS_PROTO(data->inode)->commit_done(task, data);
1677 1678
}

1679
static void nfs_commit_release_pages(struct nfs_commit_data *data)
1680
{
1681
	struct nfs_page	*req;
1682
	int status = data->task.tk_status;
1683
	struct nfs_commit_info cinfo;
1684
	struct nfs_server *nfss;
1685

L
Linus Torvalds 已提交
1686 1687 1688
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
F
Fred Isaman 已提交
1689
		nfs_clear_page_commit(req->wb_page);
L
Linus Torvalds 已提交
1690

1691
		dprintk("NFS:       commit (%s/%llu %d@%lld)",
1692
			req->wb_context->dentry->d_sb->s_id,
1693
			(unsigned long long)NFS_FILEID(req->wb_context->dentry->d_inode),
L
Linus Torvalds 已提交
1694 1695
			req->wb_bytes,
			(long long)req_offset(req));
1696 1697
		if (status < 0) {
			nfs_context_set_write_error(req->wb_context, status);
L
Linus Torvalds 已提交
1698
			nfs_inode_remove_request(req);
1699
			dprintk(", error = %d\n", status);
L
Linus Torvalds 已提交
1700 1701 1702 1703 1704
			goto next;
		}

		/* Okay, COMMIT succeeded, apparently. Check the verifier
		 * returned by the server against all stored verfs. */
1705
		if (!memcmp(&req->wb_verf, &data->verf.verifier, sizeof(req->wb_verf))) {
L
Linus Torvalds 已提交
1706 1707 1708 1709 1710 1711 1712
			/* We have a match */
			nfs_inode_remove_request(req);
			dprintk(" OK\n");
			goto next;
		}
		/* We have a mismatch. Write the page again */
		dprintk(" mismatch\n");
F
Fred 已提交
1713
		nfs_mark_request_dirty(req);
1714
		set_bit(NFS_CONTEXT_RESEND_WRITES, &req->wb_context->flags);
L
Linus Torvalds 已提交
1715
	next:
1716
		nfs_unlock_and_release_request(req);
L
Linus Torvalds 已提交
1717
	}
1718 1719 1720 1721
	nfss = NFS_SERVER(data->inode);
	if (atomic_long_read(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH)
		clear_bdi_congested(&nfss->backing_dev_info, BLK_RW_ASYNC);

1722 1723 1724
	nfs_init_cinfo(&cinfo, data->inode, data->dreq);
	if (atomic_dec_and_test(&cinfo.mds->rpcs_out))
		nfs_commit_clear_lock(NFS_I(data->inode));
1725 1726 1727 1728
}

static void nfs_commit_release(void *calldata)
{
1729
	struct nfs_commit_data *data = calldata;
1730

1731
	data->completion_ops->completion(data);
1732
	nfs_commitdata_release(calldata);
L
Linus Torvalds 已提交
1733
}
1734 1735

static const struct rpc_call_ops nfs_commit_ops = {
1736
	.rpc_call_prepare = nfs_commit_prepare,
1737 1738 1739
	.rpc_call_done = nfs_commit_done,
	.rpc_release = nfs_commit_release,
};
L
Linus Torvalds 已提交
1740

1741 1742 1743 1744 1745
static const struct nfs_commit_completion_ops nfs_commit_completion_ops = {
	.completion = nfs_commit_release_pages,
	.error_cleanup = nfs_commit_clear_lock,
};

1746 1747
int nfs_generic_commit_list(struct inode *inode, struct list_head *head,
			    int how, struct nfs_commit_info *cinfo)
1748 1749 1750
{
	int status;

F
Fred Isaman 已提交
1751
	status = pnfs_commit_list(inode, head, how, cinfo);
1752
	if (status == PNFS_NOT_ATTEMPTED)
F
Fred Isaman 已提交
1753
		status = nfs_commit_list(inode, head, how, cinfo);
1754 1755 1756
	return status;
}

1757
int nfs_commit_inode(struct inode *inode, int how)
L
Linus Torvalds 已提交
1758 1759
{
	LIST_HEAD(head);
F
Fred Isaman 已提交
1760
	struct nfs_commit_info cinfo;
1761
	int may_wait = how & FLUSH_SYNC;
1762
	int res;
L
Linus Torvalds 已提交
1763

1764 1765
	res = nfs_commit_set_lock(NFS_I(inode), may_wait);
	if (res <= 0)
1766
		goto out_mark_dirty;
F
Fred Isaman 已提交
1767 1768
	nfs_init_cinfo_from_inode(&cinfo, inode);
	res = nfs_scan_commit(inode, &head, &cinfo);
L
Linus Torvalds 已提交
1769
	if (res) {
1770 1771
		int error;

F
Fred Isaman 已提交
1772
		error = nfs_generic_commit_list(inode, &head, how, &cinfo);
1773 1774
		if (error < 0)
			return error;
1775
		if (!may_wait)
1776
			goto out_mark_dirty;
1777
		error = wait_on_bit_action(&NFS_I(inode)->flags,
1778 1779 1780 1781 1782
				NFS_INO_COMMIT,
				nfs_wait_bit_killable,
				TASK_KILLABLE);
		if (error < 0)
			return error;
1783 1784
	} else
		nfs_commit_clear_lock(NFS_I(inode));
1785 1786 1787 1788 1789 1790 1791 1792
	return res;
	/* Note: If we exit without ensuring that the commit is complete,
	 * we must mark the inode as dirty. Otherwise, future calls to
	 * sync_inode() with the WB_SYNC_ALL flag set will fail to ensure
	 * that the data is on the disk.
	 */
out_mark_dirty:
	__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
L
Linus Torvalds 已提交
1793 1794
	return res;
}
1795 1796 1797

static int nfs_commit_unstable_pages(struct inode *inode, struct writeback_control *wbc)
{
1798 1799 1800
	struct nfs_inode *nfsi = NFS_I(inode);
	int flags = FLUSH_SYNC;
	int ret = 0;
1801

1802
	/* no commits means nothing needs to be done */
F
Fred Isaman 已提交
1803
	if (!nfsi->commit_info.ncommit)
1804 1805
		return ret;

1806 1807 1808 1809
	if (wbc->sync_mode == WB_SYNC_NONE) {
		/* Don't commit yet if this is a non-blocking flush and there
		 * are a lot of outstanding writes for this mapping.
		 */
1810
		if (nfsi->commit_info.ncommit <= (nfsi->nrequests >> 1))
1811
			goto out_mark_dirty;
1812

1813
		/* don't wait for the COMMIT response */
1814
		flags = 0;
1815 1816
	}

1817 1818 1819 1820 1821 1822 1823 1824
	ret = nfs_commit_inode(inode, flags);
	if (ret >= 0) {
		if (wbc->sync_mode == WB_SYNC_NONE) {
			if (ret < wbc->nr_to_write)
				wbc->nr_to_write -= ret;
			else
				wbc->nr_to_write = 0;
		}
1825
		return 0;
1826 1827
	}
out_mark_dirty:
1828 1829 1830
	__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
	return ret;
}
L
Linus Torvalds 已提交
1831

1832 1833
int nfs_write_inode(struct inode *inode, struct writeback_control *wbc)
{
1834 1835
	return nfs_commit_unstable_pages(inode, wbc);
}
B
Bryan Schumaker 已提交
1836
EXPORT_SYMBOL_GPL(nfs_write_inode);
1837

1838 1839 1840 1841
/*
 * flush the inode to disk.
 */
int nfs_wb_all(struct inode *inode)
T
Trond Myklebust 已提交
1842 1843
{
	struct writeback_control wbc = {
1844
		.sync_mode = WB_SYNC_ALL,
T
Trond Myklebust 已提交
1845
		.nr_to_write = LONG_MAX,
1846 1847
		.range_start = 0,
		.range_end = LLONG_MAX,
T
Trond Myklebust 已提交
1848
	};
1849 1850 1851 1852 1853
	int ret;

	trace_nfs_writeback_inode_enter(inode);

	ret = sync_inode(inode, &wbc);
T
Trond Myklebust 已提交
1854

1855 1856
	trace_nfs_writeback_inode_exit(inode, ret);
	return ret;
1857
}
B
Bryan Schumaker 已提交
1858
EXPORT_SYMBOL_GPL(nfs_wb_all);
1859

1860 1861 1862 1863 1864
int nfs_wb_page_cancel(struct inode *inode, struct page *page)
{
	struct nfs_page *req;
	int ret = 0;

1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884
	wait_on_page_writeback(page);

	/* blocking call to cancel all requests and join to a single (head)
	 * request */
	req = nfs_lock_and_join_requests(page, false);

	if (IS_ERR(req)) {
		ret = PTR_ERR(req);
	} else if (req) {
		/* all requests from this page have been cancelled by
		 * nfs_lock_and_join_requests, so just remove the head
		 * request from the inode / page_private pointer and
		 * release it */
		nfs_inode_remove_request(req);
		/*
		 * In case nfs_inode_remove_request has marked the
		 * page as being dirty
		 */
		cancel_dirty_page(page, PAGE_CACHE_SIZE);
		nfs_unlock_and_release_request(req);
1885
	}
1886

1887 1888 1889
	return ret;
}

T
Trond Myklebust 已提交
1890 1891 1892 1893
/*
 * Write back all requests on one page - we do this before reading it.
 */
int nfs_wb_page(struct inode *inode, struct page *page)
1894
{
1895
	loff_t range_start = page_file_offset(page);
1896
	loff_t range_end = range_start + (loff_t)(PAGE_CACHE_SIZE - 1);
1897 1898
	struct writeback_control wbc = {
		.sync_mode = WB_SYNC_ALL,
T
Trond Myklebust 已提交
1899
		.nr_to_write = 0,
1900 1901 1902 1903
		.range_start = range_start,
		.range_end = range_end,
	};
	int ret;
1904

1905 1906
	trace_nfs_writeback_page_enter(inode);

1907
	for (;;) {
1908
		wait_on_page_writeback(page);
1909 1910 1911 1912
		if (clear_page_dirty_for_io(page)) {
			ret = nfs_writepage_locked(page, &wbc);
			if (ret < 0)
				goto out_error;
1913
			continue;
T
Trond Myklebust 已提交
1914
		}
1915
		ret = 0;
1916 1917 1918
		if (!PagePrivate(page))
			break;
		ret = nfs_commit_inode(inode, FLUSH_SYNC);
1919
		if (ret < 0)
1920
			goto out_error;
T
Trond Myklebust 已提交
1921
	}
1922
out_error:
1923
	trace_nfs_writeback_page_exit(inode, ret);
1924
	return ret;
1925 1926
}

1927 1928
#ifdef CONFIG_MIGRATION
int nfs_migrate_page(struct address_space *mapping, struct page *newpage,
1929
		struct page *page, enum migrate_mode mode)
1930
{
1931 1932 1933 1934 1935 1936 1937 1938 1939 1940
	/*
	 * If PagePrivate is set, then the page is currently associated with
	 * an in-progress read or write request. Don't try to migrate it.
	 *
	 * FIXME: we could do this in principle, but we'll need a way to ensure
	 *        that we can safely release the inode reference while holding
	 *        the page lock.
	 */
	if (PagePrivate(page))
		return -EBUSY;
1941

1942 1943
	if (!nfs_fscache_release_page(page, GFP_KERNEL))
		return -EBUSY;
1944

1945
	return migrate_page(mapping, newpage, page, mode);
1946 1947 1948
}
#endif

D
David Howells 已提交
1949
int __init nfs_init_writepagecache(void)
L
Linus Torvalds 已提交
1950 1951
{
	nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
1952
					     sizeof(struct nfs_pgio_header),
L
Linus Torvalds 已提交
1953
					     0, SLAB_HWCACHE_ALIGN,
1954
					     NULL);
L
Linus Torvalds 已提交
1955 1956 1957
	if (nfs_wdata_cachep == NULL)
		return -ENOMEM;

1958 1959
	nfs_wdata_mempool = mempool_create_slab_pool(MIN_POOL_WRITE,
						     nfs_wdata_cachep);
L
Linus Torvalds 已提交
1960
	if (nfs_wdata_mempool == NULL)
1961
		goto out_destroy_write_cache;
L
Linus Torvalds 已提交
1962

1963 1964 1965 1966 1967
	nfs_cdata_cachep = kmem_cache_create("nfs_commit_data",
					     sizeof(struct nfs_commit_data),
					     0, SLAB_HWCACHE_ALIGN,
					     NULL);
	if (nfs_cdata_cachep == NULL)
1968
		goto out_destroy_write_mempool;
1969

1970
	nfs_commit_mempool = mempool_create_slab_pool(MIN_POOL_COMMIT,
1971
						      nfs_cdata_cachep);
L
Linus Torvalds 已提交
1972
	if (nfs_commit_mempool == NULL)
1973
		goto out_destroy_commit_cache;
L
Linus Torvalds 已提交
1974

1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
	/*
	 * 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
1995
	return 0;
1996 1997 1998 1999 2000 2001 2002 2003

out_destroy_commit_cache:
	kmem_cache_destroy(nfs_cdata_cachep);
out_destroy_write_mempool:
	mempool_destroy(nfs_wdata_mempool);
out_destroy_write_cache:
	kmem_cache_destroy(nfs_wdata_cachep);
	return -ENOMEM;
L
Linus Torvalds 已提交
2004 2005
}

2006
void nfs_destroy_writepagecache(void)
L
Linus Torvalds 已提交
2007 2008
{
	mempool_destroy(nfs_commit_mempool);
2009
	kmem_cache_destroy(nfs_cdata_cachep);
L
Linus Torvalds 已提交
2010
	mempool_destroy(nfs_wdata_mempool);
2011
	kmem_cache_destroy(nfs_wdata_cachep);
L
Linus Torvalds 已提交
2012 2013
}

2014
static const struct nfs_rw_ops nfs_rw_write_ops = {
2015
	.rw_mode		= FMODE_WRITE,
2016 2017
	.rw_alloc_header	= nfs_writehdr_alloc,
	.rw_free_header		= nfs_writehdr_free,
2018
	.rw_release		= nfs_writeback_release_common,
2019 2020
	.rw_done		= nfs_writeback_done,
	.rw_result		= nfs_writeback_result,
2021
	.rw_initiate		= nfs_initiate_write,
2022
};