write.c 42.6 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
/*
 * linux/fs/nfs/write.c
 *
 * Writing file data over NFS.
 *
 * We do it like this: When a (user) process wishes to write data to an
 * NFS file, a write request is allocated that contains the RPC task data
 * plus some info on the page to be written, and added to the inode's
 * write chain. If the process writes past the end of the page, an async
 * RPC call to write the page is scheduled immediately; otherwise, the call
 * is delayed for a few seconds.
 *
 * Just like readahead, no async I/O is performed if wsize < PAGE_SIZE.
 *
 * Write requests are kept on the inode's writeback list. Each entry in
 * that list references the page (portion) to be written. When the
 * cache timeout has expired, the RPC task is woken up, and tries to
 * lock the page. As soon as it manages to do so, the request is moved
 * from the writeback list to the writelock list.
 *
 * Note: we must make sure never to confuse the inode passed in the
 * write_page request with the one in page->inode. As far as I understand
 * it, these are different when doing a swap-out.
 *
 * To understand everything that goes on here and in the NFS read code,
 * one should be aware that a page is locked in exactly one of the following
 * cases:
 *
 *  -	A write request is in progress.
 *  -	A user process is in generic_file_write/nfs_update_page
 *  -	A user process is in generic_file_read
 *
 * Also note that because of the way pages are invalidated in
 * nfs_revalidate_inode, the following assertions hold:
 *
 *  -	If a page is dirty, there will be no read requests (a page will
 *	not be re-read unless invalidated by nfs_revalidate_inode).
 *  -	If the page is not uptodate, there will be no pending write
 *	requests, and no process will be in nfs_update_page.
 *
 * FIXME: Interaction with the vmscan routines is not optimal yet.
 * Either vmscan must be made nfs-savvy, or we need a different page
 * reclaim concept that supports something like FS-independent
 * buffer_heads with a b_ops-> field.
 *
 * 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>

#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/nfs_page.h>
60 61
#include <linux/backing-dev.h>

L
Linus Torvalds 已提交
62 63 64 65
#include <asm/uaccess.h>
#include <linux/smp_lock.h>

#include "delegation.h"
C
Chuck Lever 已提交
66
#include "iostat.h"
L
Linus Torvalds 已提交
67 68 69 70 71 72 73 74 75 76 77 78 79 80 81

#define NFSDBG_FACILITY		NFSDBG_PAGECACHE

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

/*
 * Local function declarations
 */
static struct nfs_page * nfs_update_request(struct nfs_open_context*,
					    struct inode *,
					    struct page *,
					    unsigned int, unsigned int);
static int nfs_wait_on_write_congestion(struct address_space *, int);
static int nfs_wait_on_requests(struct inode *, unsigned long, unsigned int);
T
Trond Myklebust 已提交
82
static long nfs_flush_mapping(struct address_space *mapping, struct writeback_control *wbc, int how);
83
static int nfs_wb_page_priority(struct inode *inode, struct page *page, int how);
84 85 86
static const struct rpc_call_ops nfs_write_partial_ops;
static const struct rpc_call_ops nfs_write_full_ops;
static const struct rpc_call_ops nfs_commit_ops;
L
Linus Torvalds 已提交
87 88

static kmem_cache_t *nfs_wdata_cachep;
89
static mempool_t *nfs_wdata_mempool;
L
Linus Torvalds 已提交
90 91 92 93
static mempool_t *nfs_commit_mempool;

static DECLARE_WAIT_QUEUE_HEAD(nfs_write_congestion);

94
struct nfs_write_data *nfs_commit_alloc(void)
L
Linus Torvalds 已提交
95 96
{
	struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, SLAB_NOFS);
97

L
Linus Torvalds 已提交
98 99 100 101 102 103 104
	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
	}
	return p;
}

105
void nfs_commit_rcu_free(struct rcu_head *head)
L
Linus Torvalds 已提交
106
{
107
	struct nfs_write_data *p = container_of(head, struct nfs_write_data, task.u.tk_rcu);
108 109
	if (p && (p->pagevec != &p->page_array[0]))
		kfree(p->pagevec);
L
Linus Torvalds 已提交
110 111 112
	mempool_free(p, nfs_commit_mempool);
}

113 114 115 116 117
void nfs_commit_free(struct nfs_write_data *wdata)
{
	call_rcu_bh(&wdata->task.u.tk_rcu, nfs_commit_rcu_free);
}

118
struct nfs_write_data *nfs_writedata_alloc(size_t len)
119
{
120
	unsigned int pagecount = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
121 122 123 124 125
	struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, SLAB_NOFS);

	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
126
		p->npages = pagecount;
127 128
		if (pagecount <= ARRAY_SIZE(p->page_array))
			p->pagevec = p->page_array;
129
		else {
130 131
			p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_NOFS);
			if (!p->pagevec) {
132 133 134 135 136 137 138 139
				mempool_free(p, nfs_wdata_mempool);
				p = NULL;
			}
		}
	}
	return p;
}

140
static void nfs_writedata_rcu_free(struct rcu_head *head)
141
{
142
	struct nfs_write_data *p = container_of(head, struct nfs_write_data, task.u.tk_rcu);
143 144 145 146 147
	if (p && (p->pagevec != &p->page_array[0]))
		kfree(p->pagevec);
	mempool_free(p, nfs_wdata_mempool);
}

148 149 150 151 152
static void nfs_writedata_free(struct nfs_write_data *wdata)
{
	call_rcu_bh(&wdata->task.u.tk_rcu, nfs_writedata_rcu_free);
}

153
void nfs_writedata_release(void *wdata)
L
Linus Torvalds 已提交
154 155 156 157
{
	nfs_writedata_free(wdata);
}

158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
static struct nfs_page *nfs_page_find_request_locked(struct page *page)
{
	struct nfs_page *req = NULL;

	if (PagePrivate(page)) {
		req = (struct nfs_page *)page_private(page);
		if (req != NULL)
			atomic_inc(&req->wb_count);
	}
	return req;
}

static struct nfs_page *nfs_page_find_request(struct page *page)
{
	struct nfs_page *req = NULL;
	spinlock_t *req_lock = &NFS_I(page->mapping->host)->req_lock;

	spin_lock(req_lock);
	req = nfs_page_find_request_locked(page);
	spin_unlock(req_lock);
	return req;
}

L
Linus Torvalds 已提交
181 182 183 184 185 186 187 188 189 190 191 192
/* Adjust the file length if we're writing beyond the end */
static void nfs_grow_file(struct page *page, unsigned int offset, unsigned int count)
{
	struct inode *inode = page->mapping->host;
	loff_t end, i_size = i_size_read(inode);
	unsigned long end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;

	if (i_size > 0 && page->index < end_index)
		return;
	end = ((loff_t)page->index << PAGE_CACHE_SHIFT) + ((loff_t)offset+count);
	if (i_size >= end)
		return;
C
Chuck Lever 已提交
193
	nfs_inc_stats(inode, NFSIOS_EXTENDWRITE);
L
Linus Torvalds 已提交
194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
	i_size_write(inode, end);
}

/* We can set the PG_uptodate flag if we see that a write request
 * covers the full page.
 */
static void nfs_mark_uptodate(struct page *page, unsigned int base, unsigned int count)
{
	loff_t end_offs;

	if (PageUptodate(page))
		return;
	if (base != 0)
		return;
	if (count == PAGE_CACHE_SIZE) {
		SetPageUptodate(page);
		return;
	}

	end_offs = i_size_read(page->mapping->host) - 1;
	if (end_offs < 0)
		return;
	/* Is this the last page? */
	if (page->index != (unsigned long)(end_offs >> PAGE_CACHE_SHIFT))
		return;
	/* This is the last page: set PG_uptodate if we cover the entire
	 * extent of the data, then zero the rest of the page.
	 */
	if (count == (unsigned int)(end_offs & (PAGE_CACHE_SIZE - 1)) + 1) {
		memclear_highpage_flush(page, count, PAGE_CACHE_SIZE - count);
		SetPageUptodate(page);
	}
}

/*
 * Write a page synchronously.
 * Offset is the data offset within the page.
 */
static int nfs_writepage_sync(struct nfs_open_context *ctx, struct inode *inode,
		struct page *page, unsigned int offset, unsigned int count,
		int how)
{
	unsigned int	wsize = NFS_SERVER(inode)->wsize;
	int		result, written = 0;
	struct nfs_write_data *wdata;

240
	wdata = nfs_writedata_alloc(wsize);
L
Linus Torvalds 已提交
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
	if (!wdata)
		return -ENOMEM;

	wdata->flags = how;
	wdata->cred = ctx->cred;
	wdata->inode = inode;
	wdata->args.fh = NFS_FH(inode);
	wdata->args.context = ctx;
	wdata->args.pages = &page;
	wdata->args.stable = NFS_FILE_SYNC;
	wdata->args.pgbase = offset;
	wdata->args.count = wsize;
	wdata->res.fattr = &wdata->fattr;
	wdata->res.verf = &wdata->verf;

	dprintk("NFS:      nfs_writepage_sync(%s/%Ld %d@%Ld)\n",
		inode->i_sb->s_id,
		(long long)NFS_FILEID(inode),
		count, (long long)(page_offset(page) + offset));

261
	set_page_writeback(page);
L
Linus Torvalds 已提交
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
	nfs_begin_data_update(inode);
	do {
		if (count < wsize)
			wdata->args.count = count;
		wdata->args.offset = page_offset(page) + wdata->args.pgbase;

		result = NFS_PROTO(inode)->write(wdata);

		if (result < 0) {
			/* Must mark the page invalid after I/O error */
			ClearPageUptodate(page);
			goto io_error;
		}
		if (result < wdata->args.count)
			printk(KERN_WARNING "NFS: short write, count=%u, result=%d\n",
					wdata->args.count, result);

		wdata->args.offset += result;
	        wdata->args.pgbase += result;
		written += result;
		count -= result;
C
Chuck Lever 已提交
283
		nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, result);
L
Linus Torvalds 已提交
284 285 286 287 288 289 290 291 292 293
	} while (count);
	/* Update file length */
	nfs_grow_file(page, offset, written);
	/* Set the PG_uptodate flag? */
	nfs_mark_uptodate(page, offset, written);

	if (PageError(page))
		ClearPageError(page);

io_error:
294
	nfs_end_data_update(inode);
295
	end_page_writeback(page);
296
	nfs_writedata_release(wdata);
L
Linus Torvalds 已提交
297 298 299 300 301 302 303 304 305 306
	return written ? written : result;
}

static int nfs_writepage_async(struct nfs_open_context *ctx,
		struct inode *inode, struct page *page,
		unsigned int offset, unsigned int count)
{
	struct nfs_page	*req;

	req = nfs_update_request(ctx, inode, page, offset, count);
307 308
	if (IS_ERR(req))
		return PTR_ERR(req);
L
Linus Torvalds 已提交
309 310 311 312 313
	/* Update file length */
	nfs_grow_file(page, offset, count);
	/* Set the PG_uptodate flag? */
	nfs_mark_uptodate(page, offset, count);
	nfs_unlock_request(req);
314
	return 0;
L
Linus Torvalds 已提交
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
}

static int wb_priority(struct writeback_control *wbc)
{
	if (wbc->for_reclaim)
		return FLUSH_HIGHPRI;
	if (wbc->for_kupdate)
		return FLUSH_LOWPRI;
	return 0;
}

/*
 * Write an mmapped page to the server.
 */
int nfs_writepage(struct page *page, struct writeback_control *wbc)
{
	struct nfs_open_context *ctx;
	struct inode *inode = page->mapping->host;
	unsigned long end_index;
	unsigned offset = PAGE_CACHE_SIZE;
	loff_t i_size = i_size_read(inode);
	int inode_referenced = 0;
	int priority = wb_priority(wbc);
	int err;

C
Chuck Lever 已提交
340 341 342
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
	nfs_add_stats(inode, NFSIOS_WRITEPAGES, 1);

L
Linus Torvalds 已提交
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
	/*
	 * Note: We need to ensure that we have a reference to the inode
	 *       if we are to do asynchronous writes. If not, waiting
	 *       in nfs_wait_on_request() may deadlock with clear_inode().
	 *
	 *       If igrab() fails here, then it is in any case safe to
	 *       call nfs_wb_page(), since there will be no pending writes.
	 */
	if (igrab(inode) != 0)
		inode_referenced = 1;
	end_index = i_size >> PAGE_CACHE_SHIFT;

	/* Ensure we've flushed out any previous writes */
	nfs_wb_page_priority(inode, page, priority);

	/* easy case */
	if (page->index < end_index)
		goto do_it;
	/* things got complicated... */
	offset = i_size & (PAGE_CACHE_SIZE-1);

	/* OK, are we completely out? */
	err = 0; /* potential race with truncate - ignore */
	if (page->index >= end_index+1 || !offset)
		goto out;
do_it:
369
	ctx = nfs_find_open_context(inode, NULL, FMODE_WRITE);
L
Linus Torvalds 已提交
370 371 372 373 374 375 376
	if (ctx == NULL) {
		err = -EBADF;
		goto out;
	}
	lock_kernel();
	if (!IS_SYNC(inode) && inode_referenced) {
		err = nfs_writepage_async(ctx, inode, page, 0, offset);
377
		if (!wbc->for_writepages)
T
Trond Myklebust 已提交
378
			nfs_flush_mapping(page->mapping, wbc, wb_priority(wbc));
L
Linus Torvalds 已提交
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
	} else {
		err = nfs_writepage_sync(ctx, inode, page, 0,
						offset, priority);
		if (err >= 0) {
			if (err != offset)
				redirty_page_for_writepage(wbc, page);
			err = 0;
		}
	}
	unlock_kernel();
	put_nfs_open_context(ctx);
out:
	unlock_page(page);
	if (inode_referenced)
		iput(inode);
	return err; 
}

/*
 * Note: causes nfs_update_request() to block on the assumption
 * 	 that the writeback is generated due to memory pressure.
 */
int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
	struct backing_dev_info *bdi = mapping->backing_dev_info;
	struct inode *inode = mapping->host;
	int err;

C
Chuck Lever 已提交
407 408
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);

L
Linus Torvalds 已提交
409 410 411 412 413 414 415 416
	err = generic_writepages(mapping, wbc);
	if (err)
		return err;
	while (test_and_set_bit(BDI_write_congested, &bdi->state) != 0) {
		if (wbc->nonblocking)
			return 0;
		nfs_wait_on_write_congestion(mapping, 0);
	}
T
Trond Myklebust 已提交
417
	err = nfs_flush_mapping(mapping, wbc, wb_priority(wbc));
L
Linus Torvalds 已提交
418 419
	if (err < 0)
		goto out;
C
Chuck Lever 已提交
420
	nfs_add_stats(inode, NFSIOS_WRITEPAGES, err);
L
Linus Torvalds 已提交
421 422 423 424 425
	if (!wbc->nonblocking && wbc->sync_mode == WB_SYNC_ALL) {
		err = nfs_wait_on_requests(inode, 0, 0);
		if (err < 0)
			goto out;
	}
426
	err = nfs_commit_inode(inode, wb_priority(wbc));
T
Trond Myklebust 已提交
427
	if (err > 0)
L
Linus Torvalds 已提交
428 429 430 431
		err = 0;
out:
	clear_bit(BDI_write_congested, &bdi->state);
	wake_up_all(&nfs_write_congestion);
432
	congestion_end(WRITE);
L
Linus Torvalds 已提交
433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453
	return err;
}

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

	error = radix_tree_insert(&nfsi->nfs_page_tree, req->wb_index, req);
	BUG_ON(error == -EEXIST);
	if (error)
		return error;
	if (!nfsi->npages) {
		igrab(inode);
		nfs_begin_data_update(inode);
		if (nfs_have_delegation(inode, FMODE_WRITE))
			nfsi->change_attr++;
	}
454
	SetPagePrivate(req->wb_page);
455
	set_page_private(req->wb_page, (unsigned long)req);
L
Linus Torvalds 已提交
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
	nfsi->npages++;
	atomic_inc(&req->wb_count);
	return 0;
}

/*
 * Insert a write request into an inode
 */
static void nfs_inode_remove_request(struct nfs_page *req)
{
	struct inode *inode = req->wb_context->dentry->d_inode;
	struct nfs_inode *nfsi = NFS_I(inode);

	BUG_ON (!NFS_WBACK_BUSY(req));

	spin_lock(&nfsi->req_lock);
472
	set_page_private(req->wb_page, 0);
473
	ClearPagePrivate(req->wb_page);
L
Linus Torvalds 已提交
474 475 476 477
	radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index);
	nfsi->npages--;
	if (!nfsi->npages) {
		spin_unlock(&nfsi->req_lock);
478
		nfs_end_data_update(inode);
L
Linus Torvalds 已提交
479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495
		iput(inode);
	} else
		spin_unlock(&nfsi->req_lock);
	nfs_clear_request(req);
	nfs_release_request(req);
}

/*
 * Add a request to the inode's dirty list.
 */
static void
nfs_mark_request_dirty(struct nfs_page *req)
{
	struct inode *inode = req->wb_context->dentry->d_inode;
	struct nfs_inode *nfsi = NFS_I(inode);

	spin_lock(&nfsi->req_lock);
496 497
	radix_tree_tag_set(&nfsi->nfs_page_tree,
			req->wb_index, NFS_PAGE_TAG_DIRTY);
L
Linus Torvalds 已提交
498 499 500
	nfs_list_add_request(req, &nfsi->dirty);
	nfsi->ndirty++;
	spin_unlock(&nfsi->req_lock);
501
	inc_zone_page_state(req->wb_page, NR_FILE_DIRTY);
L
Linus Torvalds 已提交
502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
	mark_inode_dirty(inode);
}

/*
 * Check if a request is dirty
 */
static inline int
nfs_dirty_request(struct nfs_page *req)
{
	struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode);
	return !list_empty(&req->wb_list) && req->wb_list_head == &nfsi->dirty;
}

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

	spin_lock(&nfsi->req_lock);
	nfs_list_add_request(req, &nfsi->commit);
	nfsi->ncommit++;
	spin_unlock(&nfsi->req_lock);
529
	inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
L
Linus Torvalds 已提交
530 531 532 533 534 535 536 537 538
	mark_inode_dirty(inode);
}
#endif

/*
 * Wait for a request to complete.
 *
 * Interruptible by signals only if mounted with intr flag.
 */
539
static int nfs_wait_on_requests_locked(struct inode *inode, unsigned long idx_start, unsigned int npages)
L
Linus Torvalds 已提交
540 541 542 543 544 545 546 547 548 549 550 551 552
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_page *req;
	unsigned long		idx_end, next;
	unsigned int		res = 0;
	int			error;

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

	next = idx_start;
553
	while (radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&req, next, 1, NFS_PAGE_TAG_WRITEBACK)) {
L
Linus Torvalds 已提交
554 555 556 557
		if (req->wb_index > idx_end)
			break;

		next = req->wb_index + 1;
558
		BUG_ON(!NFS_WBACK_BUSY(req));
L
Linus Torvalds 已提交
559 560 561 562 563

		atomic_inc(&req->wb_count);
		spin_unlock(&nfsi->req_lock);
		error = nfs_wait_on_request(req);
		nfs_release_request(req);
564
		spin_lock(&nfsi->req_lock);
L
Linus Torvalds 已提交
565 566 567 568 569 570 571
		if (error < 0)
			return error;
		res++;
	}
	return res;
}

572 573 574 575 576 577 578 579 580 581 582
static int nfs_wait_on_requests(struct inode *inode, unsigned long idx_start, unsigned int npages)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	int ret;

	spin_lock(&nfsi->req_lock);
	ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
	spin_unlock(&nfsi->req_lock);
	return ret;
}

T
Trond Myklebust 已提交
583
static void nfs_cancel_dirty_list(struct list_head *head)
584 585 586 587 588 589 590 591 592 593
{
	struct nfs_page *req;
	while(!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_inode_remove_request(req);
		nfs_clear_page_writeback(req);
	}
}

T
Trond Myklebust 已提交
594 595 596 597 598 599
static void nfs_cancel_commit_list(struct list_head *head)
{
	struct nfs_page *req;

	while(!list_empty(head)) {
		req = nfs_list_entry(head->next);
600
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
T
Trond Myklebust 已提交
601 602
		nfs_list_remove_request(req);
		nfs_inode_remove_request(req);
603
		nfs_unlock_request(req);
T
Trond Myklebust 已提交
604 605 606
	}
}

L
Linus Torvalds 已提交
607 608 609 610 611 612 613 614 615 616 617 618 619 620 621
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
/*
 * nfs_scan_commit - Scan an inode for commit requests
 * @inode: NFS inode to scan
 * @dst: destination list
 * @idx_start: lower bound of page->index to scan.
 * @npages: idx_start + npages sets the upper bound to scan.
 *
 * Moves requests from the inode's 'commit' request list.
 * The requests are *not* checked to ensure that they form a contiguous set.
 */
static int
nfs_scan_commit(struct inode *inode, struct list_head *dst, unsigned long idx_start, unsigned int npages)
{
	struct nfs_inode *nfsi = NFS_I(inode);
622 623 624
	int res = 0;

	if (nfsi->ncommit != 0) {
625
		res = nfs_scan_list(nfsi, &nfsi->commit, dst, idx_start, npages);
626 627 628 629
		nfsi->ncommit -= res;
		if ((nfsi->ncommit == 0) != list_empty(&nfsi->commit))
			printk(KERN_ERR "NFS: desynchronized value of nfs_i.ncommit.\n");
	}
L
Linus Torvalds 已提交
630 631
	return res;
}
632 633 634 635 636
#else
static inline int nfs_scan_commit(struct inode *inode, struct list_head *dst, unsigned long idx_start, unsigned int npages)
{
	return 0;
}
L
Linus Torvalds 已提交
637 638 639 640 641 642 643 644 645 646 647 648
#endif

static int nfs_wait_on_write_congestion(struct address_space *mapping, int intr)
{
	struct backing_dev_info *bdi = mapping->backing_dev_info;
	DEFINE_WAIT(wait);
	int ret = 0;

	might_sleep();

	if (!bdi_write_congested(bdi))
		return 0;
C
Chuck Lever 已提交
649 650 651

	nfs_inc_stats(mapping->host, NFSIOS_CONGESTIONWAIT);

L
Linus Torvalds 已提交
652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699
	if (intr) {
		struct rpc_clnt *clnt = NFS_CLIENT(mapping->host);
		sigset_t oldset;

		rpc_clnt_sigmask(clnt, &oldset);
		prepare_to_wait(&nfs_write_congestion, &wait, TASK_INTERRUPTIBLE);
		if (bdi_write_congested(bdi)) {
			if (signalled())
				ret = -ERESTARTSYS;
			else
				schedule();
		}
		rpc_clnt_sigunmask(clnt, &oldset);
	} else {
		prepare_to_wait(&nfs_write_congestion, &wait, TASK_UNINTERRUPTIBLE);
		if (bdi_write_congested(bdi))
			schedule();
	}
	finish_wait(&nfs_write_congestion, &wait);
	return ret;
}


/*
 * Try to update any existing write request, or create one if there is none.
 * In order to match, the request's credentials must match those of
 * the calling process.
 *
 * Note: Should always be called with the Page Lock held!
 */
static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
		struct inode *inode, struct page *page,
		unsigned int offset, unsigned int bytes)
{
	struct nfs_server *server = NFS_SERVER(inode);
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_page		*req, *new = NULL;
	unsigned long		rqend, end;

	end = offset + bytes;

	if (nfs_wait_on_write_congestion(page->mapping, server->flags & NFS_MOUNT_INTR))
		return ERR_PTR(-ERESTARTSYS);
	for (;;) {
		/* Loop over all inode entries and see if we find
		 * A request for the page we wish to update
		 */
		spin_lock(&nfsi->req_lock);
700
		req = nfs_page_find_request_locked(page);
L
Linus Torvalds 已提交
701 702 703
		if (req) {
			if (!nfs_lock_request_dontget(req)) {
				int error;
704

L
Linus Torvalds 已提交
705 706 707
				spin_unlock(&nfsi->req_lock);
				error = nfs_wait_on_request(req);
				nfs_release_request(req);
N
Neil Brown 已提交
708 709 710
				if (error < 0) {
					if (new)
						nfs_release_request(new);
L
Linus Torvalds 已提交
711
					return ERR_PTR(error);
N
Neil Brown 已提交
712
				}
L
Linus Torvalds 已提交
713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781
				continue;
			}
			spin_unlock(&nfsi->req_lock);
			if (new)
				nfs_release_request(new);
			break;
		}

		if (new) {
			int error;
			nfs_lock_request_dontget(new);
			error = nfs_inode_add_request(inode, new);
			if (error) {
				spin_unlock(&nfsi->req_lock);
				nfs_unlock_request(new);
				return ERR_PTR(error);
			}
			spin_unlock(&nfsi->req_lock);
			nfs_mark_request_dirty(new);
			return new;
		}
		spin_unlock(&nfsi->req_lock);

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

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

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

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

	return req;
}

int nfs_flush_incompatible(struct file *file, struct page *page)
{
	struct nfs_open_context *ctx = (struct nfs_open_context *)file->private_data;
	struct nfs_page	*req;
	int		status = 0;
	/*
	 * 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.
	 */
782 783 784 785
	req = nfs_page_find_request(page);
	if (req != NULL) {
		int do_flush = req->wb_page != page || req->wb_context != ctx;

L
Linus Torvalds 已提交
786
		nfs_release_request(req);
787 788
		if (do_flush)
			status = nfs_wb_page(page->mapping->host, page);
L
Linus Torvalds 已提交
789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806
	}
	return (status < 0) ? status : 0;
}

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

C
Chuck Lever 已提交
807 808
	nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);

L
Linus Torvalds 已提交
809
	dprintk("NFS:      nfs_updatepage(%s/%s %d@%Ld)\n",
810 811 812
		file->f_dentry->d_parent->d_name.name,
		file->f_dentry->d_name.name, count,
		(long long)(page_offset(page) +offset));
L
Linus Torvalds 已提交
813 814 815 816 817 818 819 820 821 822 823 824 825 826 827

	if (IS_SYNC(inode)) {
		status = nfs_writepage_sync(ctx, inode, page, offset, count, 0);
		if (status > 0) {
			if (offset == 0 && status == PAGE_CACHE_SIZE)
				SetPageUptodate(page);
			return 0;
		}
		return status;
	}

	/* If we're not using byte range locks, and we know the page
	 * is entirely in cache, it may be more efficient to avoid
	 * fragmenting write requests.
	 */
828
	if (PageUptodate(page) && inode->i_flock == NULL && !(file->f_mode & O_SYNC)) {
L
Linus Torvalds 已提交
829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898
		loff_t end_offs = i_size_read(inode) - 1;
		unsigned long end_index = end_offs >> PAGE_CACHE_SHIFT;

		count += offset;
		offset = 0;
		if (unlikely(end_offs < 0)) {
			/* Do nothing */
		} else if (page->index == end_index) {
			unsigned int pglen;
			pglen = (unsigned int)(end_offs & (PAGE_CACHE_SIZE-1)) + 1;
			if (count < pglen)
				count = pglen;
		} else if (page->index < end_index)
			count = PAGE_CACHE_SIZE;
	}

	/*
	 * Try to find an NFS request corresponding to this page
	 * and update it.
	 * If the existing request cannot be updated, we must flush
	 * it out now.
	 */
	do {
		req = nfs_update_request(ctx, inode, page, offset, count);
		status = (IS_ERR(req)) ? PTR_ERR(req) : 0;
		if (status != -EBUSY)
			break;
		/* Request could not be updated. Flush it out and try again */
		status = nfs_wb_page(inode, page);
	} while (status >= 0);
	if (status < 0)
		goto done;

	status = 0;

	/* Update file length */
	nfs_grow_file(page, offset, count);
	/* Set the PG_uptodate flag? */
	nfs_mark_uptodate(page, req->wb_pgbase, req->wb_bytes);
	nfs_unlock_request(req);
done:
        dprintk("NFS:      nfs_updatepage returns %d (isize %Ld)\n",
			status, (long long)i_size_read(inode));
	if (status < 0)
		ClearPageUptodate(page);
	return status;
}

static void nfs_writepage_release(struct nfs_page *req)
{
	end_page_writeback(req->wb_page);

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
	if (!PageError(req->wb_page)) {
		if (NFS_NEED_RESCHED(req)) {
			nfs_mark_request_dirty(req);
			goto out;
		} else if (NFS_NEED_COMMIT(req)) {
			nfs_mark_request_commit(req);
			goto out;
		}
	}
	nfs_inode_remove_request(req);

out:
	nfs_clear_commit(req);
	nfs_clear_reschedule(req);
#else
	nfs_inode_remove_request(req);
#endif
899
	nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917
}

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

/*
 * Set up the argument/result storage required for the RPC call.
 */
static void nfs_write_rpcsetup(struct nfs_page *req,
		struct nfs_write_data *data,
918
		const struct rpc_call_ops *call_ops,
L
Linus Torvalds 已提交
919 920 921 922
		unsigned int count, unsigned int offset,
		int how)
{
	struct inode		*inode;
923
	int flags;
L
Linus Torvalds 已提交
924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941

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

	data->req = req;
	data->inode = inode = req->wb_context->dentry->d_inode;
	data->cred = req->wb_context->cred;

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

	data->res.fattr   = &data->fattr;
	data->res.count   = count;
	data->res.verf    = &data->verf;
942
	nfs_fattr_init(&data->fattr);
L
Linus Torvalds 已提交
943

944 945 946
	/* Set up the initial task struct.  */
	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, call_ops, data);
L
Linus Torvalds 已提交
947 948 949 950 951 952
	NFS_PROTO(inode)->write_setup(data, how);

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

	dprintk("NFS: %4d initiated write call (req %s/%Ld, %u bytes @ offset %Lu)\n",
953
		data->task.tk_pid,
L
Linus Torvalds 已提交
954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973
		inode->i_sb->s_id,
		(long long)NFS_FILEID(inode),
		count,
		(unsigned long long)data->args.offset);
}

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

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

/*
 * Generate multiple small requests to write out a single
 * contiguous dirty area on one page.
 */
T
Trond Myklebust 已提交
974
static int nfs_flush_multi(struct inode *inode, struct list_head *head, int how)
L
Linus Torvalds 已提交
975 976 977 978
{
	struct nfs_page *req = nfs_list_entry(head->next);
	struct page *page = req->wb_page;
	struct nfs_write_data *data;
979 980
	size_t wsize = NFS_SERVER(inode)->wsize, nbytes;
	unsigned int offset;
L
Linus Torvalds 已提交
981 982 983 984 985 986
	int requests = 0;
	LIST_HEAD(list);

	nfs_list_remove_request(req);

	nbytes = req->wb_bytes;
987 988 989 990
	do {
		size_t len = min(nbytes, wsize);

		data = nfs_writedata_alloc(len);
L
Linus Torvalds 已提交
991 992 993 994
		if (!data)
			goto out_bad;
		list_add(&data->pages, &list);
		requests++;
995 996
		nbytes -= len;
	} while (nbytes != 0);
L
Linus Torvalds 已提交
997 998 999
	atomic_set(&req->wb_complete, requests);

	ClearPageError(page);
1000
	set_page_writeback(page);
L
Linus Torvalds 已提交
1001 1002 1003 1004 1005 1006 1007 1008 1009
	offset = 0;
	nbytes = req->wb_bytes;
	do {
		data = list_entry(list.next, struct nfs_write_data, pages);
		list_del_init(&data->pages);

		data->pagevec[0] = page;

		if (nbytes > wsize) {
1010 1011
			nfs_write_rpcsetup(req, data, &nfs_write_partial_ops,
					wsize, offset, how);
L
Linus Torvalds 已提交
1012 1013 1014
			offset += wsize;
			nbytes -= wsize;
		} else {
1015 1016
			nfs_write_rpcsetup(req, data, &nfs_write_partial_ops,
					nbytes, offset, how);
L
Linus Torvalds 已提交
1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
			nbytes = 0;
		}
		nfs_execute_write(data);
	} while (nbytes != 0);

	return 0;

out_bad:
	while (!list_empty(&list)) {
		data = list_entry(list.next, struct nfs_write_data, pages);
		list_del(&data->pages);
1028
		nfs_writedata_release(data);
L
Linus Torvalds 已提交
1029 1030
	}
	nfs_mark_request_dirty(req);
1031
	nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
	return -ENOMEM;
}

/*
 * Create an RPC task for the given write request and kick it.
 * The page must have been locked by the caller.
 *
 * It may happen that the page we're passed is not marked dirty.
 * This is the case if nfs_updatepage detects a conflicting request
 * that has been written but not committed.
 */
T
Trond Myklebust 已提交
1043
static int nfs_flush_one(struct inode *inode, struct list_head *head, int how)
L
Linus Torvalds 已提交
1044 1045 1046 1047 1048 1049
{
	struct nfs_page		*req;
	struct page		**pages;
	struct nfs_write_data	*data;
	unsigned int		count;

1050
	data = nfs_writedata_alloc(NFS_SERVER(inode)->wsize);
L
Linus Torvalds 已提交
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
	if (!data)
		goto out_bad;

	pages = data->pagevec;
	count = 0;
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_list_add_request(req, &data->pages);
		ClearPageError(req->wb_page);
1061
		set_page_writeback(req->wb_page);
L
Linus Torvalds 已提交
1062 1063 1064 1065 1066 1067
		*pages++ = req->wb_page;
		count += req->wb_bytes;
	}
	req = nfs_list_entry(data->pages.next);

	/* Set up the argument struct */
1068
	nfs_write_rpcsetup(req, data, &nfs_write_full_ops, count, 0, how);
L
Linus Torvalds 已提交
1069 1070 1071 1072 1073 1074 1075 1076

	nfs_execute_write(data);
	return 0;
 out_bad:
	while (!list_empty(head)) {
		struct nfs_page *req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_dirty(req);
1077
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1078 1079 1080 1081
	}
	return -ENOMEM;
}

T
Trond Myklebust 已提交
1082
static int nfs_flush_list(struct inode *inode, struct list_head *head, int npages, int how)
L
Linus Torvalds 已提交
1083 1084
{
	LIST_HEAD(one_request);
T
Trond Myklebust 已提交
1085 1086 1087 1088 1089
	int (*flush_one)(struct inode *, struct list_head *, int);
	struct nfs_page	*req;
	int wpages = NFS_SERVER(inode)->wpages;
	int wsize = NFS_SERVER(inode)->wsize;
	int error;
L
Linus Torvalds 已提交
1090

T
Trond Myklebust 已提交
1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
	flush_one = nfs_flush_one;
	if (wsize < PAGE_CACHE_SIZE)
		flush_one = nfs_flush_multi;
	/* For single writes, FLUSH_STABLE is more efficient */
	if (npages <= wpages && npages == NFS_I(inode)->npages
			&& nfs_list_entry(head->next)->wb_bytes <= wsize)
		how |= FLUSH_STABLE;

	do {
		nfs_coalesce_requests(head, &one_request, wpages);
L
Linus Torvalds 已提交
1101
		req = nfs_list_entry(one_request.next);
T
Trond Myklebust 已提交
1102
		error = flush_one(inode, &one_request, how);
L
Linus Torvalds 已提交
1103
		if (error < 0)
T
Trond Myklebust 已提交
1104 1105 1106 1107
			goto out_err;
	} while (!list_empty(head));
	return 0;
out_err:
L
Linus Torvalds 已提交
1108 1109 1110 1111
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_dirty(req);
1112
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1113 1114 1115 1116 1117 1118 1119
	}
	return error;
}

/*
 * Handle a write reply that flushed part of a page.
 */
1120
static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1121
{
1122
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
1123 1124 1125 1126 1127 1128 1129 1130 1131
	struct nfs_page		*req = data->req;
	struct page		*page = req->wb_page;

	dprintk("NFS: write (%s/%Ld %d@%Ld)",
		req->wb_context->dentry->d_inode->i_sb->s_id,
		(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
		req->wb_bytes,
		(long long)req_offset(req));

1132 1133 1134 1135
	if (nfs_writeback_done(task, data) != 0)
		return;

	if (task->tk_status < 0) {
L
Linus Torvalds 已提交
1136 1137
		ClearPageUptodate(page);
		SetPageError(page);
1138 1139
		req->wb_context->error = task->tk_status;
		dprintk(", error = %d\n", task->tk_status);
L
Linus Torvalds 已提交
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159
	} else {
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
		if (data->verf.committed < NFS_FILE_SYNC) {
			if (!NFS_NEED_COMMIT(req)) {
				nfs_defer_commit(req);
				memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
				dprintk(" defer commit\n");
			} else if (memcmp(&req->wb_verf, &data->verf, sizeof(req->wb_verf))) {
				nfs_defer_reschedule(req);
				dprintk(" server reboot detected\n");
			}
		} else
#endif
			dprintk(" OK\n");
	}

	if (atomic_dec_and_test(&req->wb_complete))
		nfs_writepage_release(req);
}

1160 1161 1162 1163 1164
static const struct rpc_call_ops nfs_write_partial_ops = {
	.rpc_call_done = nfs_writeback_done_partial,
	.rpc_release = nfs_writedata_release,
};

L
Linus Torvalds 已提交
1165 1166 1167 1168 1169 1170 1171
/*
 * Handle a write reply that flushes a whole page.
 *
 * FIXME: There is an inherent race with invalidate_inode_pages and
 *	  writebacks since the page->count is kept > 1 for as long
 *	  as the page has a write request pending.
 */
1172
static void nfs_writeback_done_full(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1173
{
1174
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
1175 1176 1177
	struct nfs_page		*req;
	struct page		*page;

1178 1179 1180
	if (nfs_writeback_done(task, data) != 0)
		return;

L
Linus Torvalds 已提交
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192
	/* Update attributes as result of writeback. */
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
		page = req->wb_page;

		dprintk("NFS: write (%s/%Ld %d@%Ld)",
			req->wb_context->dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
			req->wb_bytes,
			(long long)req_offset(req));

1193
		if (task->tk_status < 0) {
L
Linus Torvalds 已提交
1194 1195
			ClearPageUptodate(page);
			SetPageError(page);
1196
			req->wb_context->error = task->tk_status;
L
Linus Torvalds 已提交
1197 1198
			end_page_writeback(page);
			nfs_inode_remove_request(req);
1199
			dprintk(", error = %d\n", task->tk_status);
L
Linus Torvalds 已提交
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
			goto next;
		}
		end_page_writeback(page);

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
		if (data->args.stable != NFS_UNSTABLE || data->verf.committed == NFS_FILE_SYNC) {
			nfs_inode_remove_request(req);
			dprintk(" OK\n");
			goto next;
		}
		memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
		nfs_mark_request_commit(req);
		dprintk(" marked for commit\n");
#else
		nfs_inode_remove_request(req);
#endif
	next:
1217
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1218 1219 1220
	}
}

1221 1222 1223 1224 1225 1226
static const struct rpc_call_ops nfs_write_full_ops = {
	.rpc_call_done = nfs_writeback_done_full,
	.rpc_release = nfs_writedata_release,
};


L
Linus Torvalds 已提交
1227 1228 1229
/*
 * This function is called when the WRITE call is complete.
 */
1230
int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
L
Linus Torvalds 已提交
1231 1232 1233
{
	struct nfs_writeargs	*argp = &data->args;
	struct nfs_writeres	*resp = &data->res;
1234
	int status;
L
Linus Torvalds 已提交
1235 1236 1237 1238

	dprintk("NFS: %4d nfs_writeback_done (status %d)\n",
		task->tk_pid, task->tk_status);

1239 1240 1241 1242 1243 1244 1245
	/*
	 * ->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.
	 */
1246 1247 1248
	status = NFS_PROTO(data->inode)->write_done(task, data);
	if (status != 0)
		return status;
C
Chuck Lever 已提交
1249 1250
	nfs_add_stats(data->inode, NFSIOS_SERVERWRITTENBYTES, resp->count);

L
Linus Torvalds 已提交
1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
	if (resp->verf->committed < argp->stable && task->tk_status >= 0) {
		/* We tried a write call, but the server did not
		 * commit data to stable storage even though we
		 * requested it.
		 * Note: There is a known bug in Tru64 < 5.0 in which
		 *	 the server reports NFS_DATA_SYNC, but performs
		 *	 NFS_FILE_SYNC. We therefore implement this checking
		 *	 as a dprintk() in order to avoid filling syslog.
		 */
		static unsigned long    complain;

		if (time_before(complain, jiffies)) {
			dprintk("NFS: faulty NFS server %s:"
				" (committed = %d) != (stable = %d)\n",
1266
				NFS_SERVER(data->inode)->nfs_client->cl_hostname,
L
Linus Torvalds 已提交
1267 1268 1269 1270 1271 1272 1273 1274 1275
				resp->verf->committed, argp->stable);
			complain = jiffies + 300 * HZ;
		}
	}
#endif
	/* Is this a short write? */
	if (task->tk_status >= 0 && resp->count < argp->count) {
		static unsigned long    complain;

C
Chuck Lever 已提交
1276 1277
		nfs_inc_stats(data->inode, NFSIOS_SHORTWRITE);

L
Linus Torvalds 已提交
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292
		/* Has the server at least made some progress? */
		if (resp->count != 0) {
			/* Was this an NFSv2 write or an NFSv3 stable write? */
			if (resp->verf->committed != NFS_UNSTABLE) {
				/* Resend from where the server left off */
				argp->offset += resp->count;
				argp->pgbase += resp->count;
				argp->count -= resp->count;
			} else {
				/* Resend as a stable write in order to avoid
				 * headaches in the case of a server crash.
				 */
				argp->stable = NFS_FILE_SYNC;
			}
			rpc_restart_call(task);
1293
			return -EAGAIN;
L
Linus Torvalds 已提交
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
		}
		if (time_before(complain, jiffies)) {
			printk(KERN_WARNING
			       "NFS: Server wrote zero bytes, expected %u.\n",
					argp->count);
			complain = jiffies + 300 * HZ;
		}
		/* Can't do anything about it except throw an error. */
		task->tk_status = -EIO;
	}
1304
	return 0;
L
Linus Torvalds 已提交
1305 1306 1307 1308
}


#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
1309
void nfs_commit_release(void *wdata)
L
Linus Torvalds 已提交
1310 1311 1312 1313 1314 1315 1316 1317
{
	nfs_commit_free(wdata);
}

/*
 * Set up the argument/result storage required for the RPC call.
 */
static void nfs_commit_rpcsetup(struct list_head *head,
1318 1319
		struct nfs_write_data *data,
		int how)
L
Linus Torvalds 已提交
1320
{
1321
	struct nfs_page		*first;
L
Linus Torvalds 已提交
1322
	struct inode		*inode;
1323
	int flags;
L
Linus Torvalds 已提交
1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335

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

	list_splice_init(head, &data->pages);
	first = nfs_list_entry(data->pages.next);
	inode = first->wb_context->dentry->d_inode;

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

	data->args.fh     = NFS_FH(data->inode);
1336 1337 1338 1339
	/* Note: we always request a commit of the entire inode */
	data->args.offset = 0;
	data->args.count  = 0;
	data->res.count   = 0;
L
Linus Torvalds 已提交
1340 1341
	data->res.fattr   = &data->fattr;
	data->res.verf    = &data->verf;
1342
	nfs_fattr_init(&data->fattr);
1343 1344 1345 1346

	/* Set up the initial task struct.  */
	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, &nfs_commit_ops, data);
L
Linus Torvalds 已提交
1347 1348 1349 1350 1351
	NFS_PROTO(inode)->commit_setup(data, how);

	data->task.tk_priority = flush_task_priority(how);
	data->task.tk_cookie = (unsigned long)inode;
	
1352
	dprintk("NFS: %4d initiated commit call\n", data->task.tk_pid);
L
Linus Torvalds 已提交
1353 1354 1355 1356 1357 1358
}

/*
 * Commit dirty pages
 */
static int
1359
nfs_commit_list(struct inode *inode, struct list_head *head, int how)
L
Linus Torvalds 已提交
1360 1361 1362 1363
{
	struct nfs_write_data	*data;
	struct nfs_page         *req;

1364
	data = nfs_commit_alloc();
L
Linus Torvalds 已提交
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378

	if (!data)
		goto out_bad;

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

	nfs_execute_write(data);
	return 0;
 out_bad:
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_commit(req);
T
Trond Myklebust 已提交
1379
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
1380
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1381 1382 1383 1384 1385 1386 1387
	}
	return -ENOMEM;
}

/*
 * COMMIT call returned
 */
1388
static void nfs_commit_done(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1389
{
1390
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
1391 1392 1393 1394 1395
	struct nfs_page		*req;

        dprintk("NFS: %4d nfs_commit_done (status %d)\n",
                                task->tk_pid, task->tk_status);

1396 1397 1398 1399
	/* Call the NFS version-specific code */
	if (NFS_PROTO(data->inode)->commit_done(task, data) != 0)
		return;

L
Linus Torvalds 已提交
1400 1401 1402
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
1403
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
L
Linus Torvalds 已提交
1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428

		dprintk("NFS: commit (%s/%Ld %d@%Ld)",
			req->wb_context->dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
			req->wb_bytes,
			(long long)req_offset(req));
		if (task->tk_status < 0) {
			req->wb_context->error = task->tk_status;
			nfs_inode_remove_request(req);
			dprintk(", error = %d\n", task->tk_status);
			goto next;
		}

		/* Okay, COMMIT succeeded, apparently. Check the verifier
		 * returned by the server against all stored verfs. */
		if (!memcmp(req->wb_verf.verifier, data->verf.verifier, sizeof(data->verf.verifier))) {
			/* We have a match */
			nfs_inode_remove_request(req);
			dprintk(" OK\n");
			goto next;
		}
		/* We have a mismatch. Write the page again */
		dprintk(" mismatch\n");
		nfs_mark_request_dirty(req);
	next:
1429
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1430 1431
	}
}
1432 1433 1434 1435 1436

static const struct rpc_call_ops nfs_commit_ops = {
	.rpc_call_done = nfs_commit_done,
	.rpc_release = nfs_commit_release,
};
1437 1438 1439 1440 1441
#else
static inline int nfs_commit_list(struct inode *inode, struct list_head *head, int how)
{
	return 0;
}
L
Linus Torvalds 已提交
1442 1443
#endif

T
Trond Myklebust 已提交
1444
static long nfs_flush_mapping(struct address_space *mapping, struct writeback_control *wbc, int how)
L
Linus Torvalds 已提交
1445
{
T
Trond Myklebust 已提交
1446
	struct nfs_inode *nfsi = NFS_I(mapping->host);
L
Linus Torvalds 已提交
1447
	LIST_HEAD(head);
T
Trond Myklebust 已提交
1448
	long res;
L
Linus Torvalds 已提交
1449 1450

	spin_lock(&nfsi->req_lock);
T
Trond Myklebust 已提交
1451
	res = nfs_scan_dirty(mapping, wbc, &head);
L
Linus Torvalds 已提交
1452
	spin_unlock(&nfsi->req_lock);
1453
	if (res) {
T
Trond Myklebust 已提交
1454
		int error = nfs_flush_list(mapping->host, &head, res, how);
T
Trond Myklebust 已提交
1455 1456
		if (error < 0)
			return error;
1457
	}
L
Linus Torvalds 已提交
1458 1459 1460 1461
	return res;
}

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
1462
int nfs_commit_inode(struct inode *inode, int how)
L
Linus Torvalds 已提交
1463 1464 1465
{
	struct nfs_inode *nfsi = NFS_I(inode);
	LIST_HEAD(head);
T
Trond Myklebust 已提交
1466
	int res;
L
Linus Torvalds 已提交
1467 1468

	spin_lock(&nfsi->req_lock);
1469 1470
	res = nfs_scan_commit(inode, &head, 0, 0);
	spin_unlock(&nfsi->req_lock);
L
Linus Torvalds 已提交
1471
	if (res) {
T
Trond Myklebust 已提交
1472
		int error = nfs_commit_list(inode, &head, how);
1473 1474 1475
		if (error < 0)
			return error;
	}
L
Linus Torvalds 已提交
1476 1477 1478 1479
	return res;
}
#endif

1480
long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_control *wbc, int how)
L
Linus Torvalds 已提交
1481
{
1482
	struct inode *inode = mapping->host;
1483
	struct nfs_inode *nfsi = NFS_I(inode);
1484 1485
	unsigned long idx_start, idx_end;
	unsigned int npages = 0;
1486
	LIST_HEAD(head);
1487
	int nocommit = how & FLUSH_NOCOMMIT;
T
Trond Myklebust 已提交
1488
	long pages, ret;
L
Linus Torvalds 已提交
1489

1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503
	/* FIXME */
	if (wbc->range_cyclic)
		idx_start = 0;
	else {
		idx_start = wbc->range_start >> PAGE_CACHE_SHIFT;
		idx_end = wbc->range_end >> PAGE_CACHE_SHIFT;
		if (idx_end > idx_start) {
			unsigned long l_npages = 1 + idx_end - idx_start;
			npages = l_npages;
			if (sizeof(npages) != sizeof(l_npages) &&
					(unsigned long)npages != l_npages)
				npages = 0;
		}
	}
1504 1505
	how &= ~FLUSH_NOCOMMIT;
	spin_lock(&nfsi->req_lock);
L
Linus Torvalds 已提交
1506
	do {
1507
		wbc->pages_skipped = 0;
1508 1509
		ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
		if (ret != 0)
1510
			continue;
1511
		pages = nfs_scan_dirty(mapping, wbc, &head);
1512 1513
		if (pages != 0) {
			spin_unlock(&nfsi->req_lock);
1514
			if (how & FLUSH_INVALIDATE) {
T
Trond Myklebust 已提交
1515
				nfs_cancel_dirty_list(&head);
1516 1517
				ret = pages;
			} else
1518
				ret = nfs_flush_list(inode, &head, pages, how);
1519 1520 1521
			spin_lock(&nfsi->req_lock);
			continue;
		}
1522 1523
		if (wbc->pages_skipped != 0)
			continue;
1524 1525
		if (nocommit)
			break;
1526
		pages = nfs_scan_commit(inode, &head, idx_start, npages);
1527 1528 1529
		if (pages == 0) {
			if (wbc->pages_skipped != 0)
				continue;
1530
			break;
1531
		}
1532 1533
		if (how & FLUSH_INVALIDATE) {
			spin_unlock(&nfsi->req_lock);
T
Trond Myklebust 已提交
1534
			nfs_cancel_commit_list(&head);
1535
			ret = pages;
1536 1537 1538 1539
			spin_lock(&nfsi->req_lock);
			continue;
		}
		pages += nfs_scan_commit(inode, &head, 0, 0);
1540 1541 1542 1543 1544 1545
		spin_unlock(&nfsi->req_lock);
		ret = nfs_commit_list(inode, &head, how);
		spin_lock(&nfsi->req_lock);
	} while (ret >= 0);
	spin_unlock(&nfsi->req_lock);
	return ret;
L
Linus Torvalds 已提交
1546 1547
}

1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
/*
 * flush the inode to disk.
 */
int nfs_wb_all(struct inode *inode)
{
	struct address_space *mapping = inode->i_mapping;
	struct writeback_control wbc = {
		.bdi = mapping->backing_dev_info,
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = LONG_MAX,
		.range_cyclic = 1,
	};
	int ret;

	ret = nfs_sync_mapping_wait(mapping, &wbc, 0);
	if (ret >= 0)
		return 0;
	return ret;
}

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

	ret = nfs_sync_mapping_wait(mapping, &wbc, how);
	if (ret >= 0)
		return 0;
	return ret;
}

static int nfs_wb_page_priority(struct inode *inode, struct page *page, int how)
{
	loff_t range_start = page_offset(page);
	loff_t range_end = range_start + (loff_t)(PAGE_CACHE_SIZE - 1);

	return nfs_sync_mapping_range(inode->i_mapping, range_start, range_end, how | FLUSH_STABLE);
}

/*
 * Write back all requests on one page - we do this before reading it.
 */
int nfs_wb_page(struct inode *inode, struct page* page)
{
	return nfs_wb_page_priority(inode, page, 0);
}


D
David Howells 已提交
1602
int __init nfs_init_writepagecache(void)
L
Linus Torvalds 已提交
1603 1604 1605 1606 1607 1608 1609 1610
{
	nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
					     sizeof(struct nfs_write_data),
					     0, SLAB_HWCACHE_ALIGN,
					     NULL, NULL);
	if (nfs_wdata_cachep == NULL)
		return -ENOMEM;

1611 1612
	nfs_wdata_mempool = mempool_create_slab_pool(MIN_POOL_WRITE,
						     nfs_wdata_cachep);
L
Linus Torvalds 已提交
1613 1614 1615
	if (nfs_wdata_mempool == NULL)
		return -ENOMEM;

1616 1617
	nfs_commit_mempool = mempool_create_slab_pool(MIN_POOL_COMMIT,
						      nfs_wdata_cachep);
L
Linus Torvalds 已提交
1618 1619 1620 1621 1622 1623
	if (nfs_commit_mempool == NULL)
		return -ENOMEM;

	return 0;
}

1624
void nfs_destroy_writepagecache(void)
L
Linus Torvalds 已提交
1625 1626 1627
{
	mempool_destroy(nfs_commit_mempool);
	mempool_destroy(nfs_wdata_mempool);
1628
	kmem_cache_destroy(nfs_wdata_cachep);
L
Linus Torvalds 已提交
1629 1630
}