write.c 51.9 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
241 242 243 244 245
}

static int wb_priority(struct writeback_control *wbc)
{
	if (wbc->for_reclaim)
246
		return FLUSH_HIGHPRI | FLUSH_STABLE;
247
	if (wbc->for_kupdate || wbc->for_background)
248 249
		return FLUSH_LOWPRI | FLUSH_COND_STABLE;
	return FLUSH_COND_STABLE;
L
Linus Torvalds 已提交
250 251
}

252 253 254 255 256 257 258 259 260 261
/*
 * 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))

262
static void nfs_set_page_writeback(struct page *page)
263
{
264
	struct nfs_server *nfss = NFS_SERVER(page_file_mapping(page)->host);
265 266
	int ret = test_set_page_writeback(page);

267
	WARN_ON_ONCE(ret != 0);
268

269 270 271 272
	if (atomic_long_inc_return(&nfss->writeback) >
			NFS_CONGESTION_ON_THRESH) {
		set_bdi_congested(&nfss->backing_dev_info,
					BLK_RW_ASYNC);
273 274 275
	}
}

276
static void nfs_end_page_writeback(struct nfs_page *req)
277
{
278
	struct inode *inode = page_file_mapping(req->wb_page)->host;
279 280
	struct nfs_server *nfss = NFS_SERVER(inode);

281 282 283 284
	if (!nfs_page_group_sync_on_bit(req, PG_WB_END))
		return;

	end_page_writeback(req->wb_page);
P
Peter Zijlstra 已提交
285
	if (atomic_long_dec_return(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH)
286
		clear_bdi_congested(&nfss->backing_dev_info, BLK_RW_ASYNC);
287 288
}

289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425

/* 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)
426
{
427
	struct inode *inode = page_file_mapping(page)->host;
428 429 430
	struct nfs_page *head, *subreq;
	struct nfs_page *destroy_list = NULL;
	unsigned int total_bytes;
431 432
	int ret;

433 434 435 436 437
try_again:
	total_bytes = 0;

	WARN_ON_ONCE(destroy_list);

438
	spin_lock(&inode->i_lock);
439 440 441 442 443 444 445 446 447

	/*
	 * 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) {
448
		spin_unlock(&inode->i_lock);
449 450 451
		return NULL;
	}

452 453
	/* holding inode lock, so always make a non-blocking call to try the
	 * page group lock */
454
	ret = nfs_page_group_lock(head, true);
455 456
	if (ret < 0) {
		spin_unlock(&inode->i_lock);
457 458 459 460 461 462 463

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

464
		nfs_release_request(head);
465
		return ERR_PTR(ret);
466
	}
467 468

	/* lock each request in the page group */
469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489
	subreq = head;
	do {
		/*
		 * Subrequests are always contiguous, non overlapping
		 * and in order. If not, it's a programming error.
		 */
		WARN_ON_ONCE(subreq->wb_offset !=
		     (head->wb_offset + total_bytes));

		/* keep track of how many bytes this group covers */
		total_bytes += subreq->wb_bytes;

		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;

490
			return ERR_PTR(ret);
491 492 493 494 495 496 497 498 499
		}

		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 {
500
		nfs_clear_request_commit(subreq);
501 502 503 504 505 506 507 508 509 510 511 512
		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;
513
	}
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529

	/*
	 * 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);

530
	/* drop lock to clean uprequests on destroy list */
531
	spin_unlock(&inode->i_lock);
532 533 534 535 536 537

	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;
538 539 540 541 542 543 544
}

/*
 * 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,
545
				struct page *page, bool nonblock)
546 547 548 549
{
	struct nfs_page *req;
	int ret = 0;

550
	req = nfs_lock_and_join_requests(page, nonblock);
551 552 553 554 555 556
	if (!req)
		goto out;
	ret = PTR_ERR(req);
	if (IS_ERR(req))
		goto out;

557 558
	nfs_set_page_writeback(page);
	WARN_ON_ONCE(test_bit(PG_CLEAN, &req->wb_flags));
559

560
	ret = 0;
561 562
	if (!nfs_pageio_add_request(pgio, req)) {
		nfs_redirty_request(req);
563
		ret = pgio->pg_error;
564
	}
565 566
out:
	return ret;
567 568
}

T
Trond Myklebust 已提交
569
static int nfs_do_writepage(struct page *page, struct writeback_control *wbc, struct nfs_pageio_descriptor *pgio)
L
Linus Torvalds 已提交
570
{
571
	struct inode *inode = page_file_mapping(page)->host;
572
	int ret;
L
Linus Torvalds 已提交
573

C
Chuck Lever 已提交
574 575 576
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
	nfs_add_stats(inode, NFSIOS_WRITEPAGES, 1);

577
	nfs_pageio_cond_complete(pgio, page_file_index(page));
578
	ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE);
579 580 581 582 583
	if (ret == -EAGAIN) {
		redirty_page_for_writepage(wbc, page);
		ret = 0;
	}
	return ret;
T
Trond Myklebust 已提交
584
}
585

T
Trond Myklebust 已提交
586 587 588 589 590 591 592
/*
 * 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;
593

594 595
	nfs_pageio_init_write(&pgio, page->mapping->host, wb_priority(wbc),
				false, &nfs_async_write_completion_ops);
T
Trond Myklebust 已提交
596 597 598 599 600 601 602
	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;
603 604 605 606
}

int nfs_writepage(struct page *page, struct writeback_control *wbc)
{
T
Trond Myklebust 已提交
607
	int ret;
608

T
Trond Myklebust 已提交
609
	ret = nfs_writepage_locked(page, wbc);
L
Linus Torvalds 已提交
610
	unlock_page(page);
T
Trond Myklebust 已提交
611 612 613 614 615 616 617 618 619 620
	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
621 622 623 624 625
}

int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
	struct inode *inode = mapping->host;
626
	unsigned long *bitlock = &NFS_I(inode)->flags;
627
	struct nfs_pageio_descriptor pgio;
L
Linus Torvalds 已提交
628 629
	int err;

630
	/* Stop dirtying of new pages while we sync */
631
	err = wait_on_bit_lock_action(bitlock, NFS_INO_FLUSHING,
632 633 634 635
			nfs_wait_bit_killable, TASK_KILLABLE);
	if (err)
		goto out_err;

C
Chuck Lever 已提交
636 637
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);

638 639
	nfs_pageio_init_write(&pgio, inode, wb_priority(wbc), false,
				&nfs_async_write_completion_ops);
T
Trond Myklebust 已提交
640
	err = write_cache_pages(mapping, wbc, nfs_writepages_callback, &pgio);
641
	nfs_pageio_complete(&pgio);
642 643

	clear_bit_unlock(NFS_INO_FLUSHING, bitlock);
644
	smp_mb__after_atomic();
645 646
	wake_up_bit(bitlock, NFS_INO_FLUSHING);

T
Trond Myklebust 已提交
647
	if (err < 0)
648 649 650 651
		goto out_err;
	err = pgio.pg_error;
	if (err < 0)
		goto out_err;
652
	return 0;
653 654
out_err:
	return err;
L
Linus Torvalds 已提交
655 656 657 658 659
}

/*
 * Insert a write request into an inode
 */
F
Fred Isaman 已提交
660
static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
L
Linus Torvalds 已提交
661 662
{
	struct nfs_inode *nfsi = NFS_I(inode);
663

664 665
	WARN_ON_ONCE(req->wb_this_page != req);

666
	/* Lock the request! */
667
	nfs_lock_request(req);
668 669

	spin_lock(&inode->i_lock);
670
	if (!nfsi->npages && NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
671
		inode->i_version++;
672 673 674 675 676 677 678 679 680
	/*
	 * 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);
	}
L
Linus Torvalds 已提交
681
	nfsi->npages++;
682 683 684
	/* this a head request for a page group - mark it as having an
	 * extra reference so sub groups can follow suit */
	WARN_ON(test_and_set_bit(PG_INODE_REF, &req->wb_flags));
685
	kref_get(&req->wb_kref);
686
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
687 688 689
}

/*
690
 * Remove a write request from an inode
L
Linus Torvalds 已提交
691 692 693
 */
static void nfs_inode_remove_request(struct nfs_page *req)
{
694
	struct inode *inode = req->wb_context->dentry->d_inode;
L
Linus Torvalds 已提交
695
	struct nfs_inode *nfsi = NFS_I(inode);
696
	struct nfs_page *head;
L
Linus Torvalds 已提交
697

698 699 700 701 702 703 704 705 706 707 708
	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);
			clear_bit(PG_MAPPED, &head->wb_flags);
		}
		nfsi->npages--;
		spin_unlock(&inode->i_lock);
709
	}
710 711 712

	if (test_and_clear_bit(PG_INODE_REF, &req->wb_flags))
		nfs_release_request(req);
713 714
	else
		WARN_ON_ONCE(1);
L
Linus Torvalds 已提交
715 716
}

717
static void
F
Fred 已提交
718
nfs_mark_request_dirty(struct nfs_page *req)
719 720 721 722
{
	__set_page_dirty_nobuffers(req->wb_page);
}

B
Bryan Schumaker 已提交
723
#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
/*
 * 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;
}

756 757 758
/**
 * nfs_request_add_commit_list - add request to a commit list
 * @req: pointer to a struct nfs_page
F
Fred Isaman 已提交
759 760
 * @dst: commit list head
 * @cinfo: holds list lock and accounting info
761
 *
F
Fred Isaman 已提交
762
 * This sets the PG_CLEAN bit, updates the cinfo count of
763 764 765
 * number of outstanding requests requiring a commit as well as
 * the MM page stats.
 *
F
Fred Isaman 已提交
766
 * The caller must _not_ hold the cinfo->lock, but must be
767
 * holding the nfs_page lock.
L
Linus Torvalds 已提交
768
 */
769
void
F
Fred Isaman 已提交
770 771
nfs_request_add_commit_list(struct nfs_page *req, struct list_head *dst,
			    struct nfs_commit_info *cinfo)
L
Linus Torvalds 已提交
772
{
773
	set_bit(PG_CLEAN, &(req)->wb_flags);
F
Fred Isaman 已提交
774 775 776 777
	spin_lock(cinfo->lock);
	nfs_list_add_request(req, dst);
	cinfo->mds->ncommit++;
	spin_unlock(cinfo->lock);
778 779
	if (!cinfo->dreq) {
		inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
780
		inc_bdi_stat(page_file_mapping(req->wb_page)->backing_dev_info,
781 782 783 784
			     BDI_RECLAIMABLE);
		__mark_inode_dirty(req->wb_context->dentry->d_inode,
				   I_DIRTY_DATASYNC);
	}
L
Linus Torvalds 已提交
785
}
786 787 788 789 790
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
791
 * @cinfo: holds list lock and accounting info
792
 *
F
Fred Isaman 已提交
793
 * This clears the PG_CLEAN bit, and updates the cinfo's count of
794 795 796
 * number of outstanding requests requiring a commit
 * It does not update the MM page stats.
 *
F
Fred Isaman 已提交
797
 * The caller _must_ hold the cinfo->lock and the nfs_page lock.
798 799
 */
void
F
Fred Isaman 已提交
800 801
nfs_request_remove_commit_list(struct nfs_page *req,
			       struct nfs_commit_info *cinfo)
802 803 804 805
{
	if (!test_and_clear_bit(PG_CLEAN, &(req)->wb_flags))
		return;
	nfs_list_remove_request(req);
F
Fred Isaman 已提交
806
	cinfo->mds->ncommit--;
807 808 809
}
EXPORT_SYMBOL_GPL(nfs_request_remove_commit_list);

F
Fred Isaman 已提交
810 811 812 813 814 815
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
816
	cinfo->dreq = NULL;
817
	cinfo->completion_ops = &nfs_commit_completion_ops;
F
Fred Isaman 已提交
818 819 820 821 822 823
}

void nfs_init_cinfo(struct nfs_commit_info *cinfo,
		    struct inode *inode,
		    struct nfs_direct_req *dreq)
{
824 825 826 827
	if (dreq)
		nfs_init_cinfo_from_dreq(cinfo, dreq);
	else
		nfs_init_cinfo_from_inode(cinfo, inode);
F
Fred Isaman 已提交
828 829
}
EXPORT_SYMBOL_GPL(nfs_init_cinfo);
830 831 832 833

/*
 * Add a request to the inode's commit list.
 */
834
void
F
Fred Isaman 已提交
835 836
nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg,
			struct nfs_commit_info *cinfo)
837
{
F
Fred Isaman 已提交
838
	if (pnfs_mark_request_commit(req, lseg, cinfo))
839
		return;
F
Fred Isaman 已提交
840
	nfs_request_add_commit_list(req, &cinfo->mds->list, cinfo);
841
}
842

F
Fred Isaman 已提交
843 844 845 846
static void
nfs_clear_page_commit(struct page *page)
{
	dec_zone_page_state(page, NR_UNSTABLE_NFS);
847
	dec_bdi_stat(page_file_mapping(page)->backing_dev_info, BDI_RECLAIMABLE);
F
Fred Isaman 已提交
848 849
}

850
/* Called holding inode (/cinfo) lock */
851
static void
852 853
nfs_clear_request_commit(struct nfs_page *req)
{
854 855
	if (test_bit(PG_CLEAN, &req->wb_flags)) {
		struct inode *inode = req->wb_context->dentry->d_inode;
F
Fred Isaman 已提交
856
		struct nfs_commit_info cinfo;
857

F
Fred Isaman 已提交
858 859 860
		nfs_init_cinfo_from_inode(&cinfo, inode);
		if (!pnfs_clear_request_commit(req, &cinfo)) {
			nfs_request_remove_commit_list(req, &cinfo);
861
		}
F
Fred Isaman 已提交
862
		nfs_clear_page_commit(req->wb_page);
863 864 865
	}
}

866
int nfs_write_need_commit(struct nfs_pgio_header *hdr)
867
{
868
	if (hdr->verf.committed == NFS_DATA_SYNC)
869
		return hdr->lseg == NULL;
870
	return hdr->verf.committed != NFS_FILE_SYNC;
871 872 873
}

#else
874 875 876 877 878 879 880
static struct nfs_page *
nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi,
						struct page *page)
{
	return NULL;
}

881 882 883 884 885 886 887 888 889 890 891
static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
				      struct inode *inode)
{
}

void nfs_init_cinfo(struct nfs_commit_info *cinfo,
		    struct inode *inode,
		    struct nfs_direct_req *dreq)
{
}

892
void
F
Fred Isaman 已提交
893 894
nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg,
			struct nfs_commit_info *cinfo)
895 896 897
{
}

898
static void
899 900 901 902
nfs_clear_request_commit(struct nfs_page *req)
{
}

903
int nfs_write_need_commit(struct nfs_pgio_header *hdr)
904 905 906 907
{
	return 0;
}

908 909
#endif

910
static void nfs_write_completion(struct nfs_pgio_header *hdr)
911
{
F
Fred Isaman 已提交
912
	struct nfs_commit_info cinfo;
913 914 915 916
	unsigned long bytes = 0;

	if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
		goto out;
F
Fred Isaman 已提交
917
	nfs_init_cinfo_from_inode(&cinfo, hdr->inode);
918 919 920 921 922 923 924
	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)) {
925
			nfs_set_pageerror(req->wb_page);
926 927 928
			nfs_context_set_write_error(req->wb_context, hdr->error);
			goto remove_req;
		}
929
		if (nfs_write_need_commit(hdr)) {
930
			memcpy(&req->wb_verf, &hdr->verf.verifier, sizeof(req->wb_verf));
F
Fred Isaman 已提交
931
			nfs_mark_request_commit(req, hdr->lseg, &cinfo);
932 933 934 935 936
			goto next;
		}
remove_req:
		nfs_inode_remove_request(req);
next:
937
		nfs_unlock_request(req);
938
		nfs_end_page_writeback(req);
939
		nfs_release_request(req);
940 941 942
	}
out:
	hdr->release(hdr);
943
}
L
Linus Torvalds 已提交
944

B
Bryan Schumaker 已提交
945
#if  IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
946
unsigned long
F
Fred Isaman 已提交
947
nfs_reqs_to_commit(struct nfs_commit_info *cinfo)
948
{
F
Fred Isaman 已提交
949
	return cinfo->mds->ncommit;
F
Fred Isaman 已提交
950 951
}

F
Fred Isaman 已提交
952
/* cinfo->lock held by caller */
953
int
F
Fred Isaman 已提交
954 955
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
956 957 958 959 960
{
	struct nfs_page *req, *tmp;
	int ret = 0;

	list_for_each_entry_safe(req, tmp, src, wb_list) {
961 962
		if (!nfs_lock_request(req))
			continue;
963
		kref_get(&req->wb_kref);
F
Fred Isaman 已提交
964
		if (cond_resched_lock(cinfo->lock))
965
			list_safe_reset_next(req, tmp, wb_list);
F
Fred Isaman 已提交
966
		nfs_request_remove_commit_list(req, cinfo);
967 968
		nfs_list_add_request(req, dst);
		ret++;
969
		if ((ret == max) && !cinfo->dreq)
970
			break;
F
Fred Isaman 已提交
971 972
	}
	return ret;
973 974
}

L
Linus Torvalds 已提交
975 976 977
/*
 * nfs_scan_commit - Scan an inode for commit requests
 * @inode: NFS inode to scan
F
Fred Isaman 已提交
978 979
 * @dst: mds destination list
 * @cinfo: mds and ds lists of reqs ready to commit
L
Linus Torvalds 已提交
980 981 982 983
 *
 * Moves requests from the inode's 'commit' request list.
 * The requests are *not* checked to ensure that they form a contiguous set.
 */
984
int
F
Fred Isaman 已提交
985 986
nfs_scan_commit(struct inode *inode, struct list_head *dst,
		struct nfs_commit_info *cinfo)
L
Linus Torvalds 已提交
987
{
F
Fred Isaman 已提交
988
	int ret = 0;
989

F
Fred Isaman 已提交
990 991
	spin_lock(cinfo->lock);
	if (cinfo->mds->ncommit > 0) {
992
		const int max = INT_MAX;
F
Fred Isaman 已提交
993

F
Fred Isaman 已提交
994 995 996
		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
997
	}
F
Fred Isaman 已提交
998
	spin_unlock(cinfo->lock);
999
	return ret;
L
Linus Torvalds 已提交
1000
}
F
Fred Isaman 已提交
1001

1002
#else
1003
unsigned long nfs_reqs_to_commit(struct nfs_commit_info *cinfo)
1004 1005 1006 1007
{
	return 0;
}

1008 1009
int nfs_scan_commit(struct inode *inode, struct list_head *dst,
		    struct nfs_commit_info *cinfo)
1010 1011 1012
{
	return 0;
}
L
Linus Torvalds 已提交
1013 1014 1015
#endif

/*
1016 1017
 * 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
1018
 *
1019 1020
 * If the attempt fails, then the existing request is flushed out
 * to disk.
L
Linus Torvalds 已提交
1021
 */
1022 1023 1024 1025
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
1026
{
1027 1028 1029 1030 1031 1032 1033
	struct nfs_page *req;
	unsigned int rqend;
	unsigned int end;
	int error;

	if (!PagePrivate(page))
		return NULL;
L
Linus Torvalds 已提交
1034 1035

	end = offset + bytes;
1036
	spin_lock(&inode->i_lock);
L
Linus Torvalds 已提交
1037 1038

	for (;;) {
1039
		req = nfs_page_find_head_request_locked(NFS_I(inode), page);
1040 1041 1042
		if (req == NULL)
			goto out_unlock;

1043 1044 1045 1046
		/* should be handled by nfs_flush_incompatible */
		WARN_ON_ONCE(req->wb_head != req);
		WARN_ON_ONCE(req->wb_this_page != req);

1047 1048 1049 1050 1051 1052 1053
		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.
		 */
1054
		if (offset > rqend
1055 1056 1057
		    || end < req->wb_offset)
			goto out_flushme;

1058
		if (nfs_lock_request(req))
L
Linus Torvalds 已提交
1059 1060
			break;

1061
		/* The request is locked, so wait and then retry */
1062
		spin_unlock(&inode->i_lock);
1063 1064 1065 1066 1067
		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
1068 1069 1070 1071 1072 1073 1074 1075 1076
	}

	/* 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;
1077 1078 1079
	else
		req->wb_bytes = rqend - req->wb_offset;
out_unlock:
1080 1081
	if (req)
		nfs_clear_request_commit(req);
1082
	spin_unlock(&inode->i_lock);
1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
	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)
{
1102
	struct inode *inode = page_file_mapping(page)->host;
1103
	struct nfs_page	*req;
L
Linus Torvalds 已提交
1104

1105 1106 1107
	req = nfs_try_to_update_request(inode, page, offset, bytes);
	if (req != NULL)
		goto out;
1108
	req = nfs_create_request(ctx, page, NULL, offset, bytes);
1109 1110
	if (IS_ERR(req))
		goto out;
F
Fred Isaman 已提交
1111
	nfs_inode_add_request(inode, req);
1112
out:
T
Trond Myklebust 已提交
1113
	return req;
L
Linus Torvalds 已提交
1114 1115
}

1116 1117 1118 1119 1120 1121 1122 1123 1124 1125
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);
1126
	nfs_mark_uptodate(req);
1127
	nfs_mark_request_dirty(req);
1128
	nfs_unlock_and_release_request(req);
1129 1130 1131
	return 0;
}

L
Linus Torvalds 已提交
1132 1133
int nfs_flush_incompatible(struct file *file, struct page *page)
{
1134
	struct nfs_open_context *ctx = nfs_file_open_context(file);
1135
	struct nfs_lock_context *l_ctx;
L
Linus Torvalds 已提交
1136
	struct nfs_page	*req;
T
Trond Myklebust 已提交
1137
	int do_flush, status;
L
Linus Torvalds 已提交
1138 1139 1140 1141 1142 1143 1144 1145
	/*
	 * 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
1146
	do {
1147
		req = nfs_page_find_head_request(page);
T
Trond Myklebust 已提交
1148 1149
		if (req == NULL)
			return 0;
1150 1151
		l_ctx = req->wb_lock_context;
		do_flush = req->wb_page != page || req->wb_context != ctx;
1152 1153
		/* for now, flush if more than 1 request in page_group */
		do_flush |= req->wb_this_page != req;
1154
		if (l_ctx && ctx->dentry->d_inode->i_flock != NULL) {
1155 1156 1157
			do_flush |= l_ctx->lockowner.l_owner != current->files
				|| l_ctx->lockowner.l_pid != current->tgid;
		}
L
Linus Torvalds 已提交
1158
		nfs_release_request(req);
T
Trond Myklebust 已提交
1159 1160
		if (!do_flush)
			return 0;
1161
		status = nfs_wb_page(page_file_mapping(page)->host, page);
T
Trond Myklebust 已提交
1162 1163
	} while (status == 0);
	return status;
L
Linus Torvalds 已提交
1164 1165
}

1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192
/*
 * 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);
}

1193 1194 1195 1196 1197
/*
 * 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.
 */
1198
static bool nfs_write_pageuptodate(struct page *page, struct inode *inode)
1199
{
1200 1201
	struct nfs_inode *nfsi = NFS_I(inode);

1202 1203
	if (nfs_have_delegated_attributes(inode))
		goto out;
1204
	if (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)
1205
		return false;
1206
	smp_rmb();
1207
	if (test_bit(NFS_INO_INVALIDATING, &nfsi->flags))
1208 1209
		return false;
out:
1210 1211
	if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
		return false;
1212
	return PageUptodate(page) != 0;
1213 1214
}

1215 1216 1217 1218 1219
/* 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.
 *
1220 1221
 * If the file is opened for synchronous writes then we can just skip the rest
 * of the checks.
1222 1223 1224 1225 1226
 */
static int nfs_can_extend_write(struct file *file, struct page *page, struct inode *inode)
{
	if (file->f_flags & O_DSYNC)
		return 0;
1227 1228
	if (!nfs_write_pageuptodate(page, inode))
		return 0;
1229 1230
	if (NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
		return 1;
1231
	if (inode->i_flock == NULL || (inode->i_flock->fl_start == 0 &&
1232
			inode->i_flock->fl_end == OFFSET_MAX &&
1233
			inode->i_flock->fl_type != F_RDLCK))
1234 1235 1236 1237
		return 1;
	return 0;
}

L
Linus Torvalds 已提交
1238 1239 1240 1241 1242 1243 1244 1245 1246
/*
 * 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)
{
1247
	struct nfs_open_context *ctx = nfs_file_open_context(file);
1248
	struct inode	*inode = page_file_mapping(page)->host;
L
Linus Torvalds 已提交
1249 1250
	int		status = 0;

C
Chuck Lever 已提交
1251 1252
	nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);

1253 1254
	dprintk("NFS:       nfs_updatepage(%pD2 %d@%lld)\n",
		file, count, (long long)(page_file_offset(page) + offset));
L
Linus Torvalds 已提交
1255

1256
	if (nfs_can_extend_write(file, page, inode)) {
1257
		count = max(count + offset, nfs_page_length(page));
L
Linus Torvalds 已提交
1258 1259 1260
		offset = 0;
	}

1261
	status = nfs_writepage_setup(ctx, page, offset, count);
1262 1263
	if (status < 0)
		nfs_set_pageerror(page);
1264 1265
	else
		__set_page_dirty_nobuffers(page);
L
Linus Torvalds 已提交
1266

1267
	dprintk("NFS:       nfs_updatepage returns %d (isize %lld)\n",
L
Linus Torvalds 已提交
1268 1269 1270 1271
			status, (long long)i_size_read(inode));
	return status;
}

1272
static int flush_task_priority(int how)
L
Linus Torvalds 已提交
1273 1274 1275 1276 1277 1278 1279 1280 1281 1282
{
	switch (how & (FLUSH_HIGHPRI|FLUSH_LOWPRI)) {
		case FLUSH_HIGHPRI:
			return RPC_PRIORITY_HIGH;
		case FLUSH_LOWPRI:
			return RPC_PRIORITY_LOW;
	}
	return RPC_PRIORITY_NORMAL;
}

1283 1284
static void nfs_initiate_write(struct nfs_pgio_header *hdr,
			       struct rpc_message *msg,
1285
			       struct rpc_task_setup *task_setup_data, int how)
L
Linus Torvalds 已提交
1286
{
1287
	struct inode *inode = hdr->inode;
1288
	int priority = flush_task_priority(how);
1289

1290
	task_setup_data->priority = priority;
1291
	NFS_PROTO(inode)->write_setup(hdr, msg);
1292

1293
	nfs4_state_protect_write(NFS_SERVER(inode)->nfs_client,
1294
				 &task_setup_data->rpc_client, msg, hdr);
1295 1296
}

F
Fred 已提交
1297 1298 1299 1300 1301 1302 1303
/* 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);
1304
	nfs_unlock_request(req);
1305
	nfs_end_page_writeback(req);
1306
	nfs_release_request(req);
F
Fred 已提交
1307 1308
}

1309
static void nfs_async_write_error(struct list_head *head)
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319
{
	struct nfs_page	*req;

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

1320 1321 1322 1323 1324
static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops = {
	.error_cleanup = nfs_async_write_error,
	.completion = nfs_write_completion,
};

1325
void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio,
1326
			       struct inode *inode, int ioflags, bool force_mds,
1327
			       const struct nfs_pgio_completion_ops *compl_ops)
L
Linus Torvalds 已提交
1328
{
1329
	struct nfs_server *server = NFS_SERVER(inode);
1330
	const struct nfs_pageio_ops *pg_ops = &nfs_pgio_rw_ops;
1331 1332 1333 1334 1335

#ifdef CONFIG_NFS_V4_1
	if (server->pnfs_curr_ld && !force_mds)
		pg_ops = server->pnfs_curr_ld->pg_write_ops;
#endif
1336 1337
	nfs_pageio_init(pgio, inode, pg_ops, compl_ops, &nfs_rw_write_ops,
			server->wsize, ioflags);
1338
}
B
Bryan Schumaker 已提交
1339
EXPORT_SYMBOL_GPL(nfs_pageio_init_write);
L
Linus Torvalds 已提交
1340

1341 1342
void nfs_pageio_reset_write_mds(struct nfs_pageio_descriptor *pgio)
{
1343
	pgio->pg_ops = &nfs_pgio_rw_ops;
1344 1345
	pgio->pg_bsize = NFS_SERVER(pgio->pg_inode)->wsize;
}
1346
EXPORT_SYMBOL_GPL(nfs_pageio_reset_write_mds);
1347

L
Linus Torvalds 已提交
1348

1349 1350 1351 1352 1353 1354 1355
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);
}

1356
static void nfs_writeback_release_common(struct nfs_pgio_header *hdr)
1357
{
1358
	/* do nothing! */
L
Linus Torvalds 已提交
1359 1360
}

1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
/*
 * 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;
1372

1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384
	/*
	 * 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;
}
1385

L
Linus Torvalds 已提交
1386 1387 1388
/*
 * This function is called when the WRITE call is complete.
 */
1389 1390
static int nfs_writeback_done(struct rpc_task *task,
			      struct nfs_pgio_header *hdr,
1391
			      struct inode *inode)
L
Linus Torvalds 已提交
1392
{
1393
	int status;
L
Linus Torvalds 已提交
1394

1395 1396 1397 1398 1399 1400 1401
	/*
	 * ->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.
	 */
1402
	status = NFS_PROTO(inode)->write_done(task, hdr);
1403
	if (status != 0)
1404
		return status;
1405
	nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, hdr->res.count);
C
Chuck Lever 已提交
1406

B
Bryan Schumaker 已提交
1407
#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
1408 1409
	if (hdr->res.verf->committed < hdr->args.stable &&
	    task->tk_status >= 0) {
L
Linus Torvalds 已提交
1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
		/* 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;

1420
		/* Note this will print the MDS for a DS write */
L
Linus Torvalds 已提交
1421
		if (time_before(complain, jiffies)) {
1422
			dprintk("NFS:       faulty NFS server %s:"
L
Linus Torvalds 已提交
1423
				" (committed = %d) != (stable = %d)\n",
1424
				NFS_SERVER(inode)->nfs_client->cl_hostname,
1425
				hdr->res.verf->committed, hdr->args.stable);
L
Linus Torvalds 已提交
1426 1427 1428 1429
			complain = jiffies + 300 * HZ;
		}
	}
#endif
1430 1431 1432 1433

	/* Deal with the suid/sgid bit corner case */
	if (nfs_should_remove_suid(inode))
		nfs_mark_for_revalidate(inode);
1434 1435 1436 1437 1438 1439
	return 0;
}

/*
 * This function is called when the WRITE call is complete.
 */
1440 1441
static void nfs_writeback_result(struct rpc_task *task,
				 struct nfs_pgio_header *hdr)
1442
{
1443 1444
	struct nfs_pgio_args	*argp = &hdr->args;
	struct nfs_pgio_res	*resp = &hdr->res;
1445 1446

	if (resp->count < argp->count) {
L
Linus Torvalds 已提交
1447 1448
		static unsigned long    complain;

1449
		/* This a short write! */
1450
		nfs_inc_stats(hdr->inode, NFSIOS_SHORTWRITE);
C
Chuck Lever 已提交
1451

L
Linus Torvalds 已提交
1452
		/* Has the server at least made some progress? */
1453 1454 1455 1456 1457 1458
		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
1459
			}
1460
			nfs_set_pgio_error(hdr, -EIO, argp->offset);
1461
			task->tk_status = -EIO;
1462
			return;
L
Linus Torvalds 已提交
1463
		}
1464 1465 1466
		/* Was this an NFSv2 write or an NFSv3 stable write? */
		if (resp->verf->committed != NFS_UNSTABLE) {
			/* Resend from where the server left off */
1467
			hdr->mds_offset += resp->count;
1468 1469 1470 1471 1472 1473 1474 1475
			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
1476
		}
1477
		rpc_restart_call_prepare(task);
L
Linus Torvalds 已提交
1478 1479 1480 1481
	}
}


B
Bryan Schumaker 已提交
1482
#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
1483 1484
static int nfs_commit_set_lock(struct nfs_inode *nfsi, int may_wait)
{
1485 1486
	int ret;

1487 1488
	if (!test_and_set_bit(NFS_INO_COMMIT, &nfsi->flags))
		return 1;
1489 1490 1491 1492 1493 1494 1495
	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;
1496 1497
}

1498
static void nfs_commit_clear_lock(struct nfs_inode *nfsi)
1499 1500
{
	clear_bit(NFS_INO_COMMIT, &nfsi->flags);
1501
	smp_mb__after_atomic();
1502 1503 1504
	wake_up_bit(&nfsi->flags, NFS_INO_COMMIT);
}

1505
void nfs_commitdata_release(struct nfs_commit_data *data)
L
Linus Torvalds 已提交
1506
{
1507 1508
	put_nfs_open_context(data->context);
	nfs_commit_free(data);
L
Linus Torvalds 已提交
1509
}
1510
EXPORT_SYMBOL_GPL(nfs_commitdata_release);
L
Linus Torvalds 已提交
1511

1512
int nfs_initiate_commit(struct rpc_clnt *clnt, struct nfs_commit_data *data,
1513
			const struct rpc_call_ops *call_ops,
1514
			int how, int flags)
L
Linus Torvalds 已提交
1515
{
1516
	struct rpc_task *task;
1517
	int priority = flush_task_priority(how);
1518 1519 1520
	struct rpc_message msg = {
		.rpc_argp = &data->args,
		.rpc_resp = &data->res,
1521
		.rpc_cred = data->cred,
1522
	};
1523
	struct rpc_task_setup task_setup_data = {
1524
		.task = &data->task,
1525
		.rpc_client = clnt,
1526
		.rpc_message = &msg,
1527
		.callback_ops = call_ops,
1528
		.callback_data = data,
1529
		.workqueue = nfsiod_workqueue,
1530
		.flags = RPC_TASK_ASYNC | flags,
1531
		.priority = priority,
1532
	};
1533 1534 1535 1536 1537
	/* Set up the initial task struct.  */
	NFS_PROTO(data->inode)->commit_setup(data, &msg);

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

1538 1539 1540
	nfs4_state_protect(NFS_SERVER(data->inode)->nfs_client,
		NFS_SP4_MACH_CRED_COMMIT, &task_setup_data.rpc_client, &msg);

1541 1542 1543 1544 1545 1546 1547 1548
	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;
}
1549
EXPORT_SYMBOL_GPL(nfs_initiate_commit);
1550

1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562
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;
}

1563 1564 1565
/*
 * Set up the argument/result storage required for the RPC call.
 */
1566
void nfs_init_commit(struct nfs_commit_data *data,
1567 1568 1569
		     struct list_head *head,
		     struct pnfs_layout_segment *lseg,
		     struct nfs_commit_info *cinfo)
1570 1571
{
	struct nfs_page *first = nfs_list_entry(head->next);
1572
	struct inode *inode = first->wb_context->dentry->d_inode;
L
Linus Torvalds 已提交
1573 1574 1575 1576 1577 1578 1579

	/* 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;
1580
	data->cred	  = first->wb_context->cred;
1581
	data->lseg	  = lseg; /* reference transferred */
1582 1583 1584
	/* only set lwb for pnfs commit */
	if (lseg)
		data->lwb = nfs_get_lwb(&data->pages);
1585
	data->mds_ops     = &nfs_commit_ops;
1586
	data->completion_ops = cinfo->completion_ops;
F
Fred Isaman 已提交
1587
	data->dreq	  = cinfo->dreq;
L
Linus Torvalds 已提交
1588 1589

	data->args.fh     = NFS_FH(data->inode);
1590 1591 1592
	/* Note: we always request a commit of the entire inode */
	data->args.offset = 0;
	data->args.count  = 0;
1593
	data->context     = get_nfs_open_context(first->wb_context);
L
Linus Torvalds 已提交
1594 1595
	data->res.fattr   = &data->fattr;
	data->res.verf    = &data->verf;
1596
	nfs_fattr_init(&data->fattr);
L
Linus Torvalds 已提交
1597
}
1598
EXPORT_SYMBOL_GPL(nfs_init_commit);
L
Linus Torvalds 已提交
1599

1600
void nfs_retry_commit(struct list_head *page_list,
F
Fred Isaman 已提交
1601 1602
		      struct pnfs_layout_segment *lseg,
		      struct nfs_commit_info *cinfo)
1603 1604 1605 1606 1607 1608
{
	struct nfs_page *req;

	while (!list_empty(page_list)) {
		req = nfs_list_entry(page_list->next);
		nfs_list_remove_request(req);
F
Fred Isaman 已提交
1609
		nfs_mark_request_commit(req, lseg, cinfo);
1610 1611
		if (!cinfo->dreq) {
			dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
1612
			dec_bdi_stat(page_file_mapping(req->wb_page)->backing_dev_info,
1613 1614
				     BDI_RECLAIMABLE);
		}
1615
		nfs_unlock_and_release_request(req);
1616 1617
	}
}
1618
EXPORT_SYMBOL_GPL(nfs_retry_commit);
1619

L
Linus Torvalds 已提交
1620 1621 1622 1623
/*
 * Commit dirty pages
 */
static int
F
Fred Isaman 已提交
1624 1625
nfs_commit_list(struct inode *inode, struct list_head *head, int how,
		struct nfs_commit_info *cinfo)
L
Linus Torvalds 已提交
1626
{
1627
	struct nfs_commit_data	*data;
L
Linus Torvalds 已提交
1628

1629
	data = nfs_commitdata_alloc();
L
Linus Torvalds 已提交
1630 1631 1632 1633 1634

	if (!data)
		goto out_bad;

	/* Set up the argument struct */
1635 1636
	nfs_init_commit(data, head, NULL, cinfo);
	atomic_inc(&cinfo->mds->rpcs_out);
1637 1638
	return nfs_initiate_commit(NFS_CLIENT(inode), data, data->mds_ops,
				   how, 0);
L
Linus Torvalds 已提交
1639
 out_bad:
F
Fred Isaman 已提交
1640
	nfs_retry_commit(head, NULL, cinfo);
1641
	cinfo->completion_ops->error_cleanup(NFS_I(inode));
L
Linus Torvalds 已提交
1642 1643 1644 1645 1646 1647
	return -ENOMEM;
}

/*
 * COMMIT call returned
 */
1648
static void nfs_commit_done(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1649
{
1650
	struct nfs_commit_data	*data = calldata;
L
Linus Torvalds 已提交
1651

C
Chuck Lever 已提交
1652
        dprintk("NFS: %5u nfs_commit_done (status %d)\n",
L
Linus Torvalds 已提交
1653 1654
                                task->tk_pid, task->tk_status);

1655
	/* Call the NFS version-specific code */
1656
	NFS_PROTO(data->inode)->commit_done(task, data);
1657 1658
}

1659
static void nfs_commit_release_pages(struct nfs_commit_data *data)
1660
{
1661
	struct nfs_page	*req;
1662
	int status = data->task.tk_status;
1663
	struct nfs_commit_info cinfo;
1664

L
Linus Torvalds 已提交
1665 1666 1667
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
F
Fred Isaman 已提交
1668
		nfs_clear_page_commit(req->wb_page);
L
Linus Torvalds 已提交
1669

1670
		dprintk("NFS:       commit (%s/%llu %d@%lld)",
1671
			req->wb_context->dentry->d_sb->s_id,
1672
			(unsigned long long)NFS_FILEID(req->wb_context->dentry->d_inode),
L
Linus Torvalds 已提交
1673 1674
			req->wb_bytes,
			(long long)req_offset(req));
1675 1676
		if (status < 0) {
			nfs_context_set_write_error(req->wb_context, status);
L
Linus Torvalds 已提交
1677
			nfs_inode_remove_request(req);
1678
			dprintk(", error = %d\n", status);
L
Linus Torvalds 已提交
1679 1680 1681 1682 1683
			goto next;
		}

		/* Okay, COMMIT succeeded, apparently. Check the verifier
		 * returned by the server against all stored verfs. */
1684
		if (!memcmp(&req->wb_verf, &data->verf.verifier, sizeof(req->wb_verf))) {
L
Linus Torvalds 已提交
1685 1686 1687 1688 1689 1690 1691
			/* 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
1692
		nfs_mark_request_dirty(req);
1693
		set_bit(NFS_CONTEXT_RESEND_WRITES, &req->wb_context->flags);
L
Linus Torvalds 已提交
1694
	next:
1695
		nfs_unlock_and_release_request(req);
L
Linus Torvalds 已提交
1696
	}
1697 1698 1699
	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));
1700 1701 1702 1703
}

static void nfs_commit_release(void *calldata)
{
1704
	struct nfs_commit_data *data = calldata;
1705

1706
	data->completion_ops->completion(data);
1707
	nfs_commitdata_release(calldata);
L
Linus Torvalds 已提交
1708
}
1709 1710

static const struct rpc_call_ops nfs_commit_ops = {
1711
	.rpc_call_prepare = nfs_commit_prepare,
1712 1713 1714
	.rpc_call_done = nfs_commit_done,
	.rpc_release = nfs_commit_release,
};
L
Linus Torvalds 已提交
1715

1716 1717 1718 1719 1720
static const struct nfs_commit_completion_ops nfs_commit_completion_ops = {
	.completion = nfs_commit_release_pages,
	.error_cleanup = nfs_commit_clear_lock,
};

1721 1722
int nfs_generic_commit_list(struct inode *inode, struct list_head *head,
			    int how, struct nfs_commit_info *cinfo)
1723 1724 1725
{
	int status;

F
Fred Isaman 已提交
1726
	status = pnfs_commit_list(inode, head, how, cinfo);
1727
	if (status == PNFS_NOT_ATTEMPTED)
F
Fred Isaman 已提交
1728
		status = nfs_commit_list(inode, head, how, cinfo);
1729 1730 1731
	return status;
}

1732
int nfs_commit_inode(struct inode *inode, int how)
L
Linus Torvalds 已提交
1733 1734
{
	LIST_HEAD(head);
F
Fred Isaman 已提交
1735
	struct nfs_commit_info cinfo;
1736
	int may_wait = how & FLUSH_SYNC;
1737
	int res;
L
Linus Torvalds 已提交
1738

1739 1740
	res = nfs_commit_set_lock(NFS_I(inode), may_wait);
	if (res <= 0)
1741
		goto out_mark_dirty;
F
Fred Isaman 已提交
1742 1743
	nfs_init_cinfo_from_inode(&cinfo, inode);
	res = nfs_scan_commit(inode, &head, &cinfo);
L
Linus Torvalds 已提交
1744
	if (res) {
1745 1746
		int error;

F
Fred Isaman 已提交
1747
		error = nfs_generic_commit_list(inode, &head, how, &cinfo);
1748 1749
		if (error < 0)
			return error;
1750
		if (!may_wait)
1751
			goto out_mark_dirty;
1752
		error = wait_on_bit_action(&NFS_I(inode)->flags,
1753 1754 1755 1756 1757
				NFS_INO_COMMIT,
				nfs_wait_bit_killable,
				TASK_KILLABLE);
		if (error < 0)
			return error;
1758 1759
	} else
		nfs_commit_clear_lock(NFS_I(inode));
1760 1761 1762 1763 1764 1765 1766 1767
	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
1768 1769
	return res;
}
1770 1771 1772

static int nfs_commit_unstable_pages(struct inode *inode, struct writeback_control *wbc)
{
1773 1774 1775
	struct nfs_inode *nfsi = NFS_I(inode);
	int flags = FLUSH_SYNC;
	int ret = 0;
1776

1777
	/* no commits means nothing needs to be done */
F
Fred Isaman 已提交
1778
	if (!nfsi->commit_info.ncommit)
1779 1780
		return ret;

1781 1782 1783 1784
	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.
		 */
F
Fred Isaman 已提交
1785
		if (nfsi->commit_info.ncommit <= (nfsi->npages >> 1))
1786
			goto out_mark_dirty;
1787

1788
		/* don't wait for the COMMIT response */
1789
		flags = 0;
1790 1791
	}

1792 1793 1794 1795 1796 1797 1798 1799
	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;
		}
1800
		return 0;
1801 1802
	}
out_mark_dirty:
1803 1804 1805
	__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
	return ret;
}
1806
#else
1807 1808 1809 1810
static int nfs_commit_unstable_pages(struct inode *inode, struct writeback_control *wbc)
{
	return 0;
}
L
Linus Torvalds 已提交
1811 1812
#endif

1813 1814
int nfs_write_inode(struct inode *inode, struct writeback_control *wbc)
{
1815 1816
	return nfs_commit_unstable_pages(inode, wbc);
}
B
Bryan Schumaker 已提交
1817
EXPORT_SYMBOL_GPL(nfs_write_inode);
1818

1819 1820 1821 1822
/*
 * flush the inode to disk.
 */
int nfs_wb_all(struct inode *inode)
T
Trond Myklebust 已提交
1823 1824
{
	struct writeback_control wbc = {
1825
		.sync_mode = WB_SYNC_ALL,
T
Trond Myklebust 已提交
1826
		.nr_to_write = LONG_MAX,
1827 1828
		.range_start = 0,
		.range_end = LLONG_MAX,
T
Trond Myklebust 已提交
1829
	};
1830 1831 1832 1833 1834
	int ret;

	trace_nfs_writeback_inode_enter(inode);

	ret = sync_inode(inode, &wbc);
T
Trond Myklebust 已提交
1835

1836 1837
	trace_nfs_writeback_inode_exit(inode, ret);
	return ret;
1838
}
B
Bryan Schumaker 已提交
1839
EXPORT_SYMBOL_GPL(nfs_wb_all);
1840

1841 1842 1843 1844 1845
int nfs_wb_page_cancel(struct inode *inode, struct page *page)
{
	struct nfs_page *req;
	int ret = 0;

1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865
	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);
1866
	}
1867

1868 1869 1870
	return ret;
}

T
Trond Myklebust 已提交
1871 1872 1873 1874
/*
 * Write back all requests on one page - we do this before reading it.
 */
int nfs_wb_page(struct inode *inode, struct page *page)
1875
{
1876
	loff_t range_start = page_file_offset(page);
1877
	loff_t range_end = range_start + (loff_t)(PAGE_CACHE_SIZE - 1);
1878 1879
	struct writeback_control wbc = {
		.sync_mode = WB_SYNC_ALL,
T
Trond Myklebust 已提交
1880
		.nr_to_write = 0,
1881 1882 1883 1884
		.range_start = range_start,
		.range_end = range_end,
	};
	int ret;
1885

1886 1887
	trace_nfs_writeback_page_enter(inode);

1888
	for (;;) {
1889
		wait_on_page_writeback(page);
1890 1891 1892 1893
		if (clear_page_dirty_for_io(page)) {
			ret = nfs_writepage_locked(page, &wbc);
			if (ret < 0)
				goto out_error;
1894
			continue;
T
Trond Myklebust 已提交
1895
		}
1896
		ret = 0;
1897 1898 1899
		if (!PagePrivate(page))
			break;
		ret = nfs_commit_inode(inode, FLUSH_SYNC);
1900
		if (ret < 0)
1901
			goto out_error;
T
Trond Myklebust 已提交
1902
	}
1903
out_error:
1904
	trace_nfs_writeback_page_exit(inode, ret);
1905
	return ret;
1906 1907
}

1908 1909
#ifdef CONFIG_MIGRATION
int nfs_migrate_page(struct address_space *mapping, struct page *newpage,
1910
		struct page *page, enum migrate_mode mode)
1911
{
1912 1913 1914 1915 1916 1917 1918 1919 1920 1921
	/*
	 * 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;
1922

1923 1924
	if (!nfs_fscache_release_page(page, GFP_KERNEL))
		return -EBUSY;
1925

1926
	return migrate_page(mapping, newpage, page, mode);
1927 1928 1929
}
#endif

D
David Howells 已提交
1930
int __init nfs_init_writepagecache(void)
L
Linus Torvalds 已提交
1931 1932
{
	nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
1933
					     sizeof(struct nfs_pgio_header),
L
Linus Torvalds 已提交
1934
					     0, SLAB_HWCACHE_ALIGN,
1935
					     NULL);
L
Linus Torvalds 已提交
1936 1937 1938
	if (nfs_wdata_cachep == NULL)
		return -ENOMEM;

1939 1940
	nfs_wdata_mempool = mempool_create_slab_pool(MIN_POOL_WRITE,
						     nfs_wdata_cachep);
L
Linus Torvalds 已提交
1941
	if (nfs_wdata_mempool == NULL)
1942
		goto out_destroy_write_cache;
L
Linus Torvalds 已提交
1943

1944 1945 1946 1947 1948
	nfs_cdata_cachep = kmem_cache_create("nfs_commit_data",
					     sizeof(struct nfs_commit_data),
					     0, SLAB_HWCACHE_ALIGN,
					     NULL);
	if (nfs_cdata_cachep == NULL)
1949
		goto out_destroy_write_mempool;
1950

1951
	nfs_commit_mempool = mempool_create_slab_pool(MIN_POOL_COMMIT,
1952
						      nfs_cdata_cachep);
L
Linus Torvalds 已提交
1953
	if (nfs_commit_mempool == NULL)
1954
		goto out_destroy_commit_cache;
L
Linus Torvalds 已提交
1955

1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
	/*
	 * 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
1976
	return 0;
1977 1978 1979 1980 1981 1982 1983 1984

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
1985 1986
}

1987
void nfs_destroy_writepagecache(void)
L
Linus Torvalds 已提交
1988 1989
{
	mempool_destroy(nfs_commit_mempool);
1990
	kmem_cache_destroy(nfs_cdata_cachep);
L
Linus Torvalds 已提交
1991
	mempool_destroy(nfs_wdata_mempool);
1992
	kmem_cache_destroy(nfs_wdata_cachep);
L
Linus Torvalds 已提交
1993 1994
}

1995
static const struct nfs_rw_ops nfs_rw_write_ops = {
1996
	.rw_mode		= FMODE_WRITE,
1997 1998
	.rw_alloc_header	= nfs_writehdr_alloc,
	.rw_free_header		= nfs_writehdr_free,
1999
	.rw_release		= nfs_writeback_release_common,
2000 2001
	.rw_done		= nfs_writeback_done,
	.rw_result		= nfs_writeback_result,
2002
	.rw_initiate		= nfs_initiate_write,
2003
};