xfs_aops.c 35.2 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2 3
 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
 * All Rights Reserved.
L
Linus Torvalds 已提交
4
 *
5 6
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
L
Linus Torvalds 已提交
7 8
 * published by the Free Software Foundation.
 *
9 10 11 12
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
L
Linus Torvalds 已提交
13
 *
14 15 16
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
L
Linus Torvalds 已提交
17 18
 */
#include "xfs.h"
19
#include "xfs_bit.h"
L
Linus Torvalds 已提交
20
#include "xfs_log.h"
21
#include "xfs_inum.h"
L
Linus Torvalds 已提交
22
#include "xfs_sb.h"
23
#include "xfs_ag.h"
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31 32 33
#include "xfs_dir.h"
#include "xfs_dir2.h"
#include "xfs_trans.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_bmap_btree.h"
#include "xfs_alloc_btree.h"
#include "xfs_ialloc_btree.h"
#include "xfs_dir_sf.h"
#include "xfs_dir2_sf.h"
34
#include "xfs_attr_sf.h"
L
Linus Torvalds 已提交
35 36
#include "xfs_dinode.h"
#include "xfs_inode.h"
37 38
#include "xfs_alloc.h"
#include "xfs_btree.h"
L
Linus Torvalds 已提交
39 40 41 42
#include "xfs_error.h"
#include "xfs_rw.h"
#include "xfs_iomap.h"
#include <linux/mpage.h>
43
#include <linux/pagevec.h>
L
Linus Torvalds 已提交
44 45
#include <linux/writeback.h>

46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69
STATIC void
xfs_count_page_state(
	struct page		*page,
	int			*delalloc,
	int			*unmapped,
	int			*unwritten)
{
	struct buffer_head	*bh, *head;

	*delalloc = *unmapped = *unwritten = 0;

	bh = head = page_buffers(page);
	do {
		if (buffer_uptodate(bh) && !buffer_mapped(bh))
			(*unmapped) = 1;
		else if (buffer_unwritten(bh) && !buffer_delay(bh))
			clear_buffer_unwritten(bh);
		else if (buffer_unwritten(bh))
			(*unwritten) = 1;
		else if (buffer_delay(bh))
			(*delalloc) = 1;
	} while ((bh = bh->b_this_page) != head);
}

L
Linus Torvalds 已提交
70 71 72 73 74 75 76 77 78
#if defined(XFS_RW_TRACE)
void
xfs_page_trace(
	int		tag,
	struct inode	*inode,
	struct page	*page,
	int		mask)
{
	xfs_inode_t	*ip;
79
	vnode_t		*vp = vn_from_inode(inode);
L
Linus Torvalds 已提交
80
	loff_t		isize = i_size_read(inode);
81
	loff_t		offset = page_offset(page);
L
Linus Torvalds 已提交
82 83 84 85 86
	int		delalloc = -1, unmapped = -1, unwritten = -1;

	if (page_has_buffers(page))
		xfs_count_page_state(page, &delalloc, &unmapped, &unwritten);

87
	ip = xfs_vtoi(vp);
L
Linus Torvalds 已提交
88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105
	if (!ip->i_rwtrace)
		return;

	ktrace_enter(ip->i_rwtrace,
		(void *)((unsigned long)tag),
		(void *)ip,
		(void *)inode,
		(void *)page,
		(void *)((unsigned long)mask),
		(void *)((unsigned long)((ip->i_d.di_size >> 32) & 0xffffffff)),
		(void *)((unsigned long)(ip->i_d.di_size & 0xffffffff)),
		(void *)((unsigned long)((isize >> 32) & 0xffffffff)),
		(void *)((unsigned long)(isize & 0xffffffff)),
		(void *)((unsigned long)((offset >> 32) & 0xffffffff)),
		(void *)((unsigned long)(offset & 0xffffffff)),
		(void *)((unsigned long)delalloc),
		(void *)((unsigned long)unmapped),
		(void *)((unsigned long)unwritten),
106
		(void *)((unsigned long)current_pid()),
L
Linus Torvalds 已提交
107 108 109 110 111 112
		(void *)NULL);
}
#else
#define xfs_page_trace(tag, inode, page, mask)
#endif

113 114 115 116 117 118 119 120 121 122 123 124
/*
 * Schedule IO completion handling on a xfsdatad if this was
 * the final hold on this ioend.
 */
STATIC void
xfs_finish_ioend(
	xfs_ioend_t		*ioend)
{
	if (atomic_dec_and_test(&ioend->io_remaining))
		queue_work(xfsdatad_workqueue, &ioend->io_work);
}

125 126 127 128 129 130
/*
 * We're now finished for good with this ioend structure.
 * Update the page state via the associated buffer_heads,
 * release holds on the inode and bio, and finally free
 * up memory.  Do not use the ioend after this.
 */
131 132 133 134
STATIC void
xfs_destroy_ioend(
	xfs_ioend_t		*ioend)
{
135 136 137 138 139 140 141
	struct buffer_head	*bh, *next;

	for (bh = ioend->io_buffer_head; bh; bh = next) {
		next = bh->b_private;
		bh->b_end_io(bh, ioend->io_uptodate);
	}

142 143 144 145 146
	vn_iowake(ioend->io_vnode);
	mempool_free(ioend, xfs_ioend_pool);
}

/*
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
 * Buffered IO write completion for delayed allocate extents.
 * TODO: Update ondisk isize now that we know the file data
 * has been flushed (i.e. the notorious "NULL file" problem).
 */
STATIC void
xfs_end_bio_delalloc(
	void			*data)
{
	xfs_ioend_t		*ioend = data;

	xfs_destroy_ioend(ioend);
}

/*
 * Buffered IO write completion for regular, written extents.
 */
STATIC void
xfs_end_bio_written(
	void			*data)
{
	xfs_ioend_t		*ioend = data;

	xfs_destroy_ioend(ioend);
}

/*
 * IO write completion for unwritten extents.
 *
175
 * Issue transactions to convert a buffer range from unwritten
176
 * to written extents.
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
 */
STATIC void
xfs_end_bio_unwritten(
	void			*data)
{
	xfs_ioend_t		*ioend = data;
	vnode_t			*vp = ioend->io_vnode;
	xfs_off_t		offset = ioend->io_offset;
	size_t			size = ioend->io_size;
	int			error;

	if (ioend->io_uptodate)
		VOP_BMAP(vp, offset, size, BMAPI_UNWRITTEN, NULL, NULL, error);
	xfs_destroy_ioend(ioend);
}

/*
 * Allocate and initialise an IO completion structure.
 * We need to track unwritten extent write completion here initially.
 * We'll need to extend this for updating the ondisk inode size later
 * (vs. incore size).
 */
STATIC xfs_ioend_t *
xfs_alloc_ioend(
201 202
	struct inode		*inode,
	unsigned int		type)
203 204 205 206 207 208 209 210 211 212 213 214
{
	xfs_ioend_t		*ioend;

	ioend = mempool_alloc(xfs_ioend_pool, GFP_NOFS);

	/*
	 * Set the count to 1 initially, which will prevent an I/O
	 * completion callback from happening before we have started
	 * all the I/O from calling the completion routine too early.
	 */
	atomic_set(&ioend->io_remaining, 1);
	ioend->io_uptodate = 1; /* cleared if any I/O fails */
215 216
	ioend->io_list = NULL;
	ioend->io_type = type;
217
	ioend->io_vnode = vn_from_inode(inode);
218
	ioend->io_buffer_head = NULL;
219
	ioend->io_buffer_tail = NULL;
220 221 222 223
	atomic_inc(&ioend->io_vnode->v_iocount);
	ioend->io_offset = 0;
	ioend->io_size = 0;

224 225 226 227 228 229
	if (type == IOMAP_UNWRITTEN)
		INIT_WORK(&ioend->io_work, xfs_end_bio_unwritten, ioend);
	else if (type == IOMAP_DELAY)
		INIT_WORK(&ioend->io_work, xfs_end_bio_delalloc, ioend);
	else
		INIT_WORK(&ioend->io_work, xfs_end_bio_written, ioend);
230 231 232 233

	return ioend;
}

L
Linus Torvalds 已提交
234 235 236 237 238 239 240 241
STATIC int
xfs_map_blocks(
	struct inode		*inode,
	loff_t			offset,
	ssize_t			count,
	xfs_iomap_t		*mapp,
	int			flags)
{
242
	vnode_t			*vp = vn_from_inode(inode);
L
Linus Torvalds 已提交
243 244 245 246 247 248 249 250
	int			error, nmaps = 1;

	VOP_BMAP(vp, offset, count, flags, mapp, &nmaps, error);
	if (!error && (flags & (BMAPI_WRITE|BMAPI_ALLOCATE)))
		VMODIFY(vp);
	return -error;
}

251 252
STATIC inline int
xfs_iomap_valid(
L
Linus Torvalds 已提交
253
	xfs_iomap_t		*iomapp,
254
	loff_t			offset)
L
Linus Torvalds 已提交
255
{
256 257
	return offset >= iomapp->iomap_offset &&
		offset < iomapp->iomap_offset + iomapp->iomap_bsize;
L
Linus Torvalds 已提交
258 259
}

260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360
/*
 * BIO completion handler for buffered IO.
 */
STATIC int
xfs_end_bio(
	struct bio		*bio,
	unsigned int		bytes_done,
	int			error)
{
	xfs_ioend_t		*ioend = bio->bi_private;

	if (bio->bi_size)
		return 1;

	ASSERT(ioend);
	ASSERT(atomic_read(&bio->bi_cnt) >= 1);

	/* Toss bio and pass work off to an xfsdatad thread */
	if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
		ioend->io_uptodate = 0;
	bio->bi_private = NULL;
	bio->bi_end_io = NULL;

	bio_put(bio);
	xfs_finish_ioend(ioend);
	return 0;
}

STATIC void
xfs_submit_ioend_bio(
	xfs_ioend_t	*ioend,
	struct bio	*bio)
{
	atomic_inc(&ioend->io_remaining);

	bio->bi_private = ioend;
	bio->bi_end_io = xfs_end_bio;

	submit_bio(WRITE, bio);
	ASSERT(!bio_flagged(bio, BIO_EOPNOTSUPP));
	bio_put(bio);
}

STATIC struct bio *
xfs_alloc_ioend_bio(
	struct buffer_head	*bh)
{
	struct bio		*bio;
	int			nvecs = bio_get_nr_vecs(bh->b_bdev);

	do {
		bio = bio_alloc(GFP_NOIO, nvecs);
		nvecs >>= 1;
	} while (!bio);

	ASSERT(bio->bi_private == NULL);
	bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9);
	bio->bi_bdev = bh->b_bdev;
	bio_get(bio);
	return bio;
}

STATIC void
xfs_start_buffer_writeback(
	struct buffer_head	*bh)
{
	ASSERT(buffer_mapped(bh));
	ASSERT(buffer_locked(bh));
	ASSERT(!buffer_delay(bh));
	ASSERT(!buffer_unwritten(bh));

	mark_buffer_async_write(bh);
	set_buffer_uptodate(bh);
	clear_buffer_dirty(bh);
}

STATIC void
xfs_start_page_writeback(
	struct page		*page,
	struct writeback_control *wbc,
	int			clear_dirty,
	int			buffers)
{
	ASSERT(PageLocked(page));
	ASSERT(!PageWriteback(page));
	set_page_writeback(page);
	if (clear_dirty)
		clear_page_dirty(page);
	unlock_page(page);
	if (!buffers) {
		end_page_writeback(page);
		wbc->pages_skipped++;	/* We didn't write this page */
	}
}

static inline int bio_add_buffer(struct bio *bio, struct buffer_head *bh)
{
	return bio_add_page(bio, bh->b_page, bh->b_size, bh_offset(bh));
}

/*
361 362 363 364 365 366 367 368 369 370 371 372 373 374
 * Submit all of the bios for all of the ioends we have saved up, covering the
 * initial writepage page and also any probed pages.
 *
 * Because we may have multiple ioends spanning a page, we need to start
 * writeback on all the buffers before we submit them for I/O. If we mark the
 * buffers as we got, then we can end up with a page that only has buffers
 * marked async write and I/O complete on can occur before we mark the other
 * buffers async write.
 *
 * The end result of this is that we trip a bug in end_page_writeback() because
 * we call it twice for the one page as the code in end_buffer_async_write()
 * assumes that all buffers on the page are started at the same time.
 *
 * The fix is two passes across the ioend list - one to start writeback on the
375
 * buffer_heads, and then submit them for I/O on the second pass.
376 377 378 379 380
 */
STATIC void
xfs_submit_ioend(
	xfs_ioend_t		*ioend)
{
381
	xfs_ioend_t		*head = ioend;
382 383 384 385 386
	xfs_ioend_t		*next;
	struct buffer_head	*bh;
	struct bio		*bio;
	sector_t		lastblock = 0;

387 388 389 390 391 392 393 394 395 396
	/* Pass 1 - start writeback */
	do {
		next = ioend->io_list;
		for (bh = ioend->io_buffer_head; bh; bh = bh->b_private) {
			xfs_start_buffer_writeback(bh);
		}
	} while ((ioend = next) != NULL);

	/* Pass 2 - submit I/O */
	ioend = head;
397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
	do {
		next = ioend->io_list;
		bio = NULL;

		for (bh = ioend->io_buffer_head; bh; bh = bh->b_private) {

			if (!bio) {
 retry:
				bio = xfs_alloc_ioend_bio(bh);
			} else if (bh->b_blocknr != lastblock + 1) {
				xfs_submit_ioend_bio(ioend, bio);
				goto retry;
			}

			if (bio_add_buffer(bio, bh) != bh->b_size) {
				xfs_submit_ioend_bio(ioend, bio);
				goto retry;
			}

			lastblock = bh->b_blocknr;
		}
		if (bio)
			xfs_submit_ioend_bio(ioend, bio);
		xfs_finish_ioend(ioend);
	} while ((ioend = next) != NULL);
}

/*
 * Cancel submission of all buffer_heads so far in this endio.
 * Toss the endio too.  Only ever called for the initial page
 * in a writepage request, so only ever one page.
 */
STATIC void
xfs_cancel_ioend(
	xfs_ioend_t		*ioend)
{
	xfs_ioend_t		*next;
	struct buffer_head	*bh, *next_bh;

	do {
		next = ioend->io_list;
		bh = ioend->io_buffer_head;
		do {
			next_bh = bh->b_private;
			clear_buffer_async_write(bh);
			unlock_buffer(bh);
		} while ((bh = next_bh) != NULL);

		vn_iowake(ioend->io_vnode);
		mempool_free(ioend, xfs_ioend_pool);
	} while ((ioend = next) != NULL);
}

/*
 * Test to see if we've been building up a completion structure for
 * earlier buffers -- if so, we try to append to this ioend if we
 * can, otherwise we finish off any current ioend and start another.
 * Return true if we've finished the given ioend.
 */
STATIC void
xfs_add_to_ioend(
	struct inode		*inode,
	struct buffer_head	*bh,
460
	xfs_off_t		offset,
461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485
	unsigned int		type,
	xfs_ioend_t		**result,
	int			need_ioend)
{
	xfs_ioend_t		*ioend = *result;

	if (!ioend || need_ioend || type != ioend->io_type) {
		xfs_ioend_t	*previous = *result;

		ioend = xfs_alloc_ioend(inode, type);
		ioend->io_offset = offset;
		ioend->io_buffer_head = bh;
		ioend->io_buffer_tail = bh;
		if (previous)
			previous->io_list = ioend;
		*result = ioend;
	} else {
		ioend->io_buffer_tail->b_private = bh;
		ioend->io_buffer_tail = bh;
	}

	bh->b_private = NULL;
	ioend->io_size += bh->b_size;
}

486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505
STATIC void
xfs_map_buffer(
	struct buffer_head	*bh,
	xfs_iomap_t		*mp,
	xfs_off_t		offset,
	uint			block_bits)
{
	sector_t		bn;

	ASSERT(mp->iomap_bn != IOMAP_DADDR_NULL);

	bn = (mp->iomap_bn >> (block_bits - BBSHIFT)) +
	      ((offset - mp->iomap_offset) >> block_bits);

	ASSERT(bn || (mp->iomap_flags & IOMAP_REALTIME));

	bh->b_blocknr = bn;
	set_buffer_mapped(bh);
}

L
Linus Torvalds 已提交
506 507 508
STATIC void
xfs_map_at_offset(
	struct buffer_head	*bh,
509
	loff_t			offset,
L
Linus Torvalds 已提交
510
	int			block_bits,
511
	xfs_iomap_t		*iomapp)
L
Linus Torvalds 已提交
512 513 514 515 516
{
	ASSERT(!(iomapp->iomap_flags & IOMAP_HOLE));
	ASSERT(!(iomapp->iomap_flags & IOMAP_DELAY));

	lock_buffer(bh);
517
	xfs_map_buffer(bh, iomapp, offset, block_bits);
518
	bh->b_bdev = iomapp->iomap_target->bt_bdev;
L
Linus Torvalds 已提交
519 520
	set_buffer_mapped(bh);
	clear_buffer_delay(bh);
521
	clear_buffer_unwritten(bh);
L
Linus Torvalds 已提交
522 523 524
}

/*
525
 * Look for a page at index that is suitable for clustering.
L
Linus Torvalds 已提交
526 527
 */
STATIC unsigned int
528
xfs_probe_page(
529
	struct page		*page,
530 531
	unsigned int		pg_offset,
	int			mapped)
L
Linus Torvalds 已提交
532 533 534 535
{
	int			ret = 0;

	if (PageWriteback(page))
536
		return 0;
L
Linus Torvalds 已提交
537 538 539 540 541 542 543

	if (page->mapping && PageDirty(page)) {
		if (page_has_buffers(page)) {
			struct buffer_head	*bh, *head;

			bh = head = page_buffers(page);
			do {
544 545 546
				if (!buffer_uptodate(bh))
					break;
				if (mapped != buffer_mapped(bh))
L
Linus Torvalds 已提交
547 548 549 550 551 552
					break;
				ret += bh->b_size;
				if (ret >= pg_offset)
					break;
			} while ((bh = bh->b_this_page) != head);
		} else
553
			ret = mapped ? 0 : PAGE_CACHE_SIZE;
L
Linus Torvalds 已提交
554 555 556 557 558
	}

	return ret;
}

559
STATIC size_t
560
xfs_probe_cluster(
L
Linus Torvalds 已提交
561 562 563
	struct inode		*inode,
	struct page		*startpage,
	struct buffer_head	*bh,
564 565
	struct buffer_head	*head,
	int			mapped)
L
Linus Torvalds 已提交
566
{
567
	struct pagevec		pvec;
L
Linus Torvalds 已提交
568
	pgoff_t			tindex, tlast, tloff;
569 570
	size_t			total = 0;
	int			done = 0, i;
L
Linus Torvalds 已提交
571 572 573

	/* First sum forwards in this page */
	do {
574
		if (!buffer_uptodate(bh) || (mapped != buffer_mapped(bh)))
575
			return total;
L
Linus Torvalds 已提交
576 577 578
		total += bh->b_size;
	} while ((bh = bh->b_this_page) != head);

579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
	/* if we reached the end of the page, sum forwards in following pages */
	tlast = i_size_read(inode) >> PAGE_CACHE_SHIFT;
	tindex = startpage->index + 1;

	/* Prune this back to avoid pathological behavior */
	tloff = min(tlast, startpage->index + 64);

	pagevec_init(&pvec, 0);
	while (!done && tindex <= tloff) {
		unsigned len = min_t(pgoff_t, PAGEVEC_SIZE, tlast - tindex + 1);

		if (!pagevec_lookup(&pvec, inode->i_mapping, tindex, len))
			break;

		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];
			size_t pg_offset, len = 0;

			if (tindex == tlast) {
				pg_offset =
				    i_size_read(inode) & (PAGE_CACHE_SIZE - 1);
600 601
				if (!pg_offset) {
					done = 1;
602
					break;
603
				}
604 605 606 607
			} else
				pg_offset = PAGE_CACHE_SIZE;

			if (page->index == tindex && !TestSetPageLocked(page)) {
608
				len = xfs_probe_page(page, pg_offset, mapped);
609 610 611 612 613 614 615 616
				unlock_page(page);
			}

			if (!len) {
				done = 1;
				break;
			}

L
Linus Torvalds 已提交
617
			total += len;
618
			tindex++;
L
Linus Torvalds 已提交
619
		}
620 621 622

		pagevec_release(&pvec);
		cond_resched();
L
Linus Torvalds 已提交
623
	}
624

L
Linus Torvalds 已提交
625 626 627 628
	return total;
}

/*
629 630
 * Test if a given page is suitable for writing as part of an unwritten
 * or delayed allocate extent.
L
Linus Torvalds 已提交
631
 */
632 633 634
STATIC int
xfs_is_delayed_page(
	struct page		*page,
635
	unsigned int		type)
L
Linus Torvalds 已提交
636 637
{
	if (PageWriteback(page))
638
		return 0;
L
Linus Torvalds 已提交
639 640 641 642 643 644 645

	if (page->mapping && page_has_buffers(page)) {
		struct buffer_head	*bh, *head;
		int			acceptable = 0;

		bh = head = page_buffers(page);
		do {
646 647 648 649
			if (buffer_unwritten(bh))
				acceptable = (type == IOMAP_UNWRITTEN);
			else if (buffer_delay(bh))
				acceptable = (type == IOMAP_DELAY);
650
			else if (buffer_dirty(bh) && buffer_mapped(bh))
651
				acceptable = (type == 0);
652
			else
L
Linus Torvalds 已提交
653 654 655 656
				break;
		} while ((bh = bh->b_this_page) != head);

		if (acceptable)
657
			return 1;
L
Linus Torvalds 已提交
658 659
	}

660
	return 0;
L
Linus Torvalds 已提交
661 662 663 664 665 666 667 668
}

/*
 * Allocate & map buffers for page given the extent map. Write it out.
 * except for the original page of a writepage, this is called on
 * delalloc/unwritten pages only, for the original page it is possible
 * that the page has no mapping at all.
 */
669
STATIC int
L
Linus Torvalds 已提交
670 671 672
xfs_convert_page(
	struct inode		*inode,
	struct page		*page,
673
	loff_t			tindex,
674
	xfs_iomap_t		*mp,
675
	xfs_ioend_t		**ioendp,
L
Linus Torvalds 已提交
676 677 678 679
	struct writeback_control *wbc,
	int			startio,
	int			all_bh)
{
680
	struct buffer_head	*bh, *head;
681 682
	xfs_off_t		end_offset;
	unsigned long		p_offset;
683
	unsigned int		type;
L
Linus Torvalds 已提交
684
	int			bbits = inode->i_blkbits;
685
	int			len, page_dirty;
686
	int			count = 0, done = 0, uptodate = 1;
687
 	xfs_off_t		offset = page_offset(page);
L
Linus Torvalds 已提交
688

689 690 691 692 693 694 695 696 697 698 699
	if (page->index != tindex)
		goto fail;
	if (TestSetPageLocked(page))
		goto fail;
	if (PageWriteback(page))
		goto fail_unlock_page;
	if (page->mapping != inode->i_mapping)
		goto fail_unlock_page;
	if (!xfs_is_delayed_page(page, (*ioendp)->io_type))
		goto fail_unlock_page;

700 701
	/*
	 * page_dirty is initially a count of buffers on the page before
702
	 * EOF and is decremented as we move each into a cleanable state.
703 704 705 706 707 708 709 710 711
	 *
	 * Derivation:
	 *
	 * End offset is the highest offset that this page should represent.
	 * If we are on the last page, (end_offset & (PAGE_CACHE_SIZE - 1))
	 * will evaluate non-zero and be less than PAGE_CACHE_SIZE and
	 * hence give us the correct page_dirty count. On any other page,
	 * it will be zero and in that case we need page_dirty to be the
	 * count of buffers on the page.
712
	 */
713 714 715 716
	end_offset = min_t(unsigned long long,
			(xfs_off_t)(page->index + 1) << PAGE_CACHE_SHIFT,
			i_size_read(inode));

717
	len = 1 << inode->i_blkbits;
718 719 720 721
	p_offset = min_t(unsigned long, end_offset & (PAGE_CACHE_SIZE - 1),
					PAGE_CACHE_SIZE);
	p_offset = p_offset ? roundup(p_offset, len) : PAGE_CACHE_SIZE;
	page_dirty = p_offset / len;
722

L
Linus Torvalds 已提交
723 724
	bh = head = page_buffers(page);
	do {
725
		if (offset >= end_offset)
L
Linus Torvalds 已提交
726
			break;
727 728 729 730
		if (!buffer_uptodate(bh))
			uptodate = 0;
		if (!(PageUptodate(page) || buffer_uptodate(bh))) {
			done = 1;
L
Linus Torvalds 已提交
731
			continue;
732 733
		}

734 735 736 737 738 739 740
		if (buffer_unwritten(bh) || buffer_delay(bh)) {
			if (buffer_unwritten(bh))
				type = IOMAP_UNWRITTEN;
			else
				type = IOMAP_DELAY;

			if (!xfs_iomap_valid(mp, offset)) {
741
				done = 1;
742 743 744 745 746 747 748 749
				continue;
			}

			ASSERT(!(mp->iomap_flags & IOMAP_HOLE));
			ASSERT(!(mp->iomap_flags & IOMAP_DELAY));

			xfs_map_at_offset(bh, offset, bbits, mp);
			if (startio) {
750
				xfs_add_to_ioend(inode, bh, offset,
751 752 753 754 755 756 757 758 759 760 761
						type, ioendp, done);
			} else {
				set_buffer_dirty(bh);
				unlock_buffer(bh);
				mark_buffer_dirty(bh);
			}
			page_dirty--;
			count++;
		} else {
			type = 0;
			if (buffer_mapped(bh) && all_bh && startio) {
L
Linus Torvalds 已提交
762
				lock_buffer(bh);
763
				xfs_add_to_ioend(inode, bh, offset,
764 765
						type, ioendp, done);
				count++;
766
				page_dirty--;
767 768
			} else {
				done = 1;
L
Linus Torvalds 已提交
769 770
			}
		}
771
	} while (offset += len, (bh = bh->b_this_page) != head);
L
Linus Torvalds 已提交
772

773 774 775 776
	if (uptodate && bh == head)
		SetPageUptodate(page);

	if (startio) {
777 778 779 780
		if (count) {
			struct backing_dev_info *bdi;

			bdi = inode->i_mapping->backing_dev_info;
781
			wbc->nr_to_write--;
782 783 784
			if (bdi_write_congested(bdi)) {
				wbc->encountered_congestion = 1;
				done = 1;
785
			} else if (wbc->nr_to_write <= 0) {
786 787 788
				done = 1;
			}
		}
789
		xfs_start_page_writeback(page, wbc, !page_dirty, count);
L
Linus Torvalds 已提交
790
	}
791 792

	return done;
793 794 795 796
 fail_unlock_page:
	unlock_page(page);
 fail:
	return 1;
L
Linus Torvalds 已提交
797 798 799 800 801 802 803 804 805 806 807
}

/*
 * Convert & write out a cluster of pages in the same extent as defined
 * by mp and following the start page.
 */
STATIC void
xfs_cluster_write(
	struct inode		*inode,
	pgoff_t			tindex,
	xfs_iomap_t		*iomapp,
808
	xfs_ioend_t		**ioendp,
L
Linus Torvalds 已提交
809 810 811 812 813
	struct writeback_control *wbc,
	int			startio,
	int			all_bh,
	pgoff_t			tlast)
{
814 815
	struct pagevec		pvec;
	int			done = 0, i;
L
Linus Torvalds 已提交
816

817 818 819 820 821
	pagevec_init(&pvec, 0);
	while (!done && tindex <= tlast) {
		unsigned len = min_t(pgoff_t, PAGEVEC_SIZE, tlast - tindex + 1);

		if (!pagevec_lookup(&pvec, inode->i_mapping, tindex, len))
L
Linus Torvalds 已提交
822
			break;
823 824 825 826 827 828 829 830 831 832

		for (i = 0; i < pagevec_count(&pvec); i++) {
			done = xfs_convert_page(inode, pvec.pages[i], tindex++,
					iomapp, ioendp, wbc, startio, all_bh);
			if (done)
				break;
		}

		pagevec_release(&pvec);
		cond_resched();
L
Linus Torvalds 已提交
833 834 835 836 837 838 839 840 841 842 843 844
	}
}

/*
 * Calling this without startio set means we are being asked to make a dirty
 * page ready for freeing it's buffers.  When called with startio set then
 * we are coming from writepage.
 *
 * When called with startio set it is important that we write the WHOLE
 * page if possible.
 * The bh->b_state's cannot know if any of the blocks or which block for
 * that matter are dirty due to mmap writes, and therefore bh uptodate is
845
 * only valid if the page itself isn't completely uptodate.  Some layers
L
Linus Torvalds 已提交
846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862
 * may clear the page dirty flag prior to calling write page, under the
 * assumption the entire page will be written out; by not writing out the
 * whole page the page can be reused before all valid dirty data is
 * written out.  Note: in the case of a page that has been dirty'd by
 * mapwrite and but partially setup by block_prepare_write the
 * bh->b_states's will not agree and only ones setup by BPW/BCW will have
 * valid state, thus the whole page must be written out thing.
 */

STATIC int
xfs_page_state_convert(
	struct inode	*inode,
	struct page	*page,
	struct writeback_control *wbc,
	int		startio,
	int		unmapped) /* also implies page uptodate */
{
863
	struct buffer_head	*bh, *head;
864
	xfs_iomap_t		iomap;
865
	xfs_ioend_t		*ioend = NULL, *iohead = NULL;
L
Linus Torvalds 已提交
866 867
	loff_t			offset;
	unsigned long           p_offset = 0;
868
	unsigned int		type;
L
Linus Torvalds 已提交
869 870
	__uint64_t              end_offset;
	pgoff_t                 end_index, last_index, tlast;
871 872
	ssize_t			size, len;
	int			flags, err, iomap_valid = 0, uptodate = 1;
873 874
	int			page_dirty, count = 0;
	int			trylock = 0;
875
	int			all_bh = unmapped;
L
Linus Torvalds 已提交
876

877 878 879 880
	if (startio) {
		if (wbc->sync_mode == WB_SYNC_NONE && wbc->nonblocking)
			trylock |= BMAPI_TRYLOCK;
	}
881

L
Linus Torvalds 已提交
882 883 884 885 886 887 888
	/* Is this page beyond the end of the file? */
	offset = i_size_read(inode);
	end_index = offset >> PAGE_CACHE_SHIFT;
	last_index = (offset - 1) >> PAGE_CACHE_SHIFT;
	if (page->index >= end_index) {
		if ((page->index >= end_index + 1) ||
		    !(i_size_read(inode) & (PAGE_CACHE_SIZE - 1))) {
889 890 891
			if (startio)
				unlock_page(page);
			return 0;
L
Linus Torvalds 已提交
892 893 894 895
		}
	}

	/*
896
	 * page_dirty is initially a count of buffers on the page before
897
	 * EOF and is decremented as we move each into a cleanable state.
898 899 900 901 902 903 904 905 906 907 908 909
	 *
	 * Derivation:
	 *
	 * End offset is the highest offset that this page should represent.
	 * If we are on the last page, (end_offset & (PAGE_CACHE_SIZE - 1))
	 * will evaluate non-zero and be less than PAGE_CACHE_SIZE and
	 * hence give us the correct page_dirty count. On any other page,
	 * it will be zero and in that case we need page_dirty to be the
	 * count of buffers on the page.
 	 */
	end_offset = min_t(unsigned long long,
			(xfs_off_t)(page->index + 1) << PAGE_CACHE_SHIFT, offset);
910
	len = 1 << inode->i_blkbits;
911 912 913
	p_offset = min_t(unsigned long, end_offset & (PAGE_CACHE_SIZE - 1),
					PAGE_CACHE_SIZE);
	p_offset = p_offset ? roundup(p_offset, len) : PAGE_CACHE_SIZE;
914 915 916
	page_dirty = p_offset / len;

	bh = head = page_buffers(page);
917
	offset = page_offset(page);
918 919
	flags = -1;
	type = 0;
920 921

	/* TODO: cleanup count and page_dirty */
L
Linus Torvalds 已提交
922 923 924 925 926 927

	do {
		if (offset >= end_offset)
			break;
		if (!buffer_uptodate(bh))
			uptodate = 0;
928
		if (!(PageUptodate(page) || buffer_uptodate(bh)) && !startio) {
929 930 931 932 933
			/*
			 * the iomap is actually still valid, but the ioend
			 * isn't.  shouldn't happen too often.
			 */
			iomap_valid = 0;
L
Linus Torvalds 已提交
934
			continue;
935
		}
L
Linus Torvalds 已提交
936

937 938
		if (iomap_valid)
			iomap_valid = xfs_iomap_valid(&iomap, offset);
L
Linus Torvalds 已提交
939 940 941 942

		/*
		 * First case, map an unwritten extent and prepare for
		 * extent state conversion transaction on completion.
943
		 *
L
Linus Torvalds 已提交
944 945
		 * Second case, allocate space for a delalloc buffer.
		 * We can return EAGAIN here in the release page case.
946 947 948 949 950 951 952
		 *
		 * Third case, an unmapped buffer was found, and we are
		 * in a path where we need to write the whole page out.
 		 */
		if (buffer_unwritten(bh) || buffer_delay(bh) ||
		    ((buffer_uptodate(bh) || PageUptodate(page)) &&
		     !buffer_mapped(bh) && (unmapped || startio))) {
953 954 955 956 957 958
		     	/*
			 * Make sure we don't use a read-only iomap
			 */
		     	if (flags == BMAPI_READ)
				iomap_valid = 0;

959 960
			if (buffer_unwritten(bh)) {
				type = IOMAP_UNWRITTEN;
961
				flags = BMAPI_WRITE | BMAPI_IGNSTATE;
962
			} else if (buffer_delay(bh)) {
963
				type = IOMAP_DELAY;
964
				flags = BMAPI_ALLOCATE | trylock;
965
			} else {
966
				type = IOMAP_NEW;
967
				flags = BMAPI_WRITE | BMAPI_MMAP;
968 969
			}

970
			if (!iomap_valid) {
971 972 973
				if (type == IOMAP_NEW) {
					size = xfs_probe_cluster(inode,
							page, bh, head, 0);
974 975 976 977 978 979
				} else {
					size = len;
				}

				err = xfs_map_blocks(inode, offset, size,
						&iomap, flags);
980
				if (err)
L
Linus Torvalds 已提交
981
					goto error;
982
				iomap_valid = xfs_iomap_valid(&iomap, offset);
L
Linus Torvalds 已提交
983
			}
984 985 986
			if (iomap_valid) {
				xfs_map_at_offset(bh, offset,
						inode->i_blkbits, &iomap);
L
Linus Torvalds 已提交
987
				if (startio) {
988
					xfs_add_to_ioend(inode, bh, offset,
989 990
							type, &ioend,
							!iomap_valid);
L
Linus Torvalds 已提交
991 992 993 994 995 996
				} else {
					set_buffer_dirty(bh);
					unlock_buffer(bh);
					mark_buffer_dirty(bh);
				}
				page_dirty--;
997
				count++;
L
Linus Torvalds 已提交
998
			}
999
		} else if (buffer_uptodate(bh) && startio) {
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
			/*
			 * we got here because the buffer is already mapped.
			 * That means it must already have extents allocated
			 * underneath it. Map the extent by reading it.
			 */
			if (!iomap_valid || type != 0) {
				flags = BMAPI_READ;
				size = xfs_probe_cluster(inode, page, bh,
								head, 1);
				err = xfs_map_blocks(inode, offset, size,
						&iomap, flags);
				if (err)
					goto error;
				iomap_valid = xfs_iomap_valid(&iomap, offset);
			}
1015

1016
			type = 0;
1017 1018
			if (!test_and_set_bit(BH_Lock, &bh->b_state)) {
				ASSERT(buffer_mapped(bh));
1019 1020
				if (iomap_valid)
					all_bh = 1;
1021
				xfs_add_to_ioend(inode, bh, offset, type,
1022 1023 1024
						&ioend, !iomap_valid);
				page_dirty--;
				count++;
1025
			} else {
1026
				iomap_valid = 0;
L
Linus Torvalds 已提交
1027
			}
1028 1029 1030
		} else if ((buffer_uptodate(bh) || PageUptodate(page)) &&
			   (unmapped || startio)) {
			iomap_valid = 0;
L
Linus Torvalds 已提交
1031
		}
1032 1033 1034 1035 1036

		if (!iohead)
			iohead = ioend;

	} while (offset += len, ((bh = bh->b_this_page) != head));
L
Linus Torvalds 已提交
1037 1038 1039 1040

	if (uptodate && bh == head)
		SetPageUptodate(page);

1041 1042
	if (startio)
		xfs_start_page_writeback(page, wbc, 1, count);
L
Linus Torvalds 已提交
1043

1044 1045
	if (ioend && iomap_valid) {
		offset = (iomap.iomap_offset + iomap.iomap_bsize - 1) >>
L
Linus Torvalds 已提交
1046
					PAGE_CACHE_SHIFT;
1047
		tlast = min_t(pgoff_t, offset, last_index);
1048
		xfs_cluster_write(inode, page->index + 1, &iomap, &ioend,
1049
					wbc, startio, all_bh, tlast);
L
Linus Torvalds 已提交
1050 1051
	}

1052 1053 1054
	if (iohead)
		xfs_submit_ioend(iohead);

L
Linus Torvalds 已提交
1055 1056 1057
	return page_dirty;

error:
1058 1059
	if (iohead)
		xfs_cancel_ioend(iohead);
L
Linus Torvalds 已提交
1060 1061 1062 1063 1064 1065 1066

	/*
	 * If it's delalloc and we have nowhere to put it,
	 * throw it away, unless the lower layers told
	 * us to try again.
	 */
	if (err != -EAGAIN) {
1067
		if (!unmapped)
L
Linus Torvalds 已提交
1068 1069 1070 1071 1072 1073
			block_invalidatepage(page, 0);
		ClearPageUptodate(page);
	}
	return err;
}

1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
/*
 * writepage: Called from one of two places:
 *
 * 1. we are flushing a delalloc buffer head.
 *
 * 2. we are writing out a dirty page. Typically the page dirty
 *    state is cleared before we get here. In this case is it
 *    conceivable we have no buffer heads.
 *
 * For delalloc space on the page we need to allocate space and
 * flush it. For unmapped buffer heads on the page we should
 * allocate space if the page is uptodate. For any other dirty
 * buffer heads on the page we should flush them.
 *
 * If we detect that a transaction would be required to flush
 * the page, we have to check the process flags first, if we
 * are already in a transaction or disk I/O during allocations
 * is off, we need to fail the writepage and redirty the page.
 */

STATIC int
1095
xfs_vm_writepage(
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180
	struct page		*page,
	struct writeback_control *wbc)
{
	int			error;
	int			need_trans;
	int			delalloc, unmapped, unwritten;
	struct inode		*inode = page->mapping->host;

	xfs_page_trace(XFS_WRITEPAGE_ENTER, inode, page, 0);

	/*
	 * We need a transaction if:
	 *  1. There are delalloc buffers on the page
	 *  2. The page is uptodate and we have unmapped buffers
	 *  3. The page is uptodate and we have no buffers
	 *  4. There are unwritten buffers on the page
	 */

	if (!page_has_buffers(page)) {
		unmapped = 1;
		need_trans = 1;
	} else {
		xfs_count_page_state(page, &delalloc, &unmapped, &unwritten);
		if (!PageUptodate(page))
			unmapped = 0;
		need_trans = delalloc + unmapped + unwritten;
	}

	/*
	 * If we need a transaction and the process flags say
	 * we are already in a transaction, or no IO is allowed
	 * then mark the page dirty again and leave the page
	 * as is.
	 */
	if (PFLAGS_TEST_FSTRANS() && need_trans)
		goto out_fail;

	/*
	 * Delay hooking up buffer heads until we have
	 * made our go/no-go decision.
	 */
	if (!page_has_buffers(page))
		create_empty_buffers(page, 1 << inode->i_blkbits, 0);

	/*
	 * Convert delayed allocate, unwritten or unmapped space
	 * to real space and flush out to disk.
	 */
	error = xfs_page_state_convert(inode, page, wbc, 1, unmapped);
	if (error == -EAGAIN)
		goto out_fail;
	if (unlikely(error < 0))
		goto out_unlock;

	return 0;

out_fail:
	redirty_page_for_writepage(wbc, page);
	unlock_page(page);
	return 0;
out_unlock:
	unlock_page(page);
	return error;
}

/*
 * Called to move a page into cleanable state - and from there
 * to be released. Possibly the page is already clean. We always
 * have buffer heads in this call.
 *
 * Returns 0 if the page is ok to release, 1 otherwise.
 *
 * Possible scenarios are:
 *
 * 1. We are being called to release a page which has been written
 *    to via regular I/O. buffer heads will be dirty and possibly
 *    delalloc. If no delalloc buffer heads in this case then we
 *    can just return zero.
 *
 * 2. We are called to release a page which has been written via
 *    mmap, all we need to do is ensure there is no delalloc
 *    state in the buffer heads, if not we can let the caller
 *    free them and we should come back later via writepage.
 */
STATIC int
1181
xfs_vm_releasepage(
1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193
	struct page		*page,
	gfp_t			gfp_mask)
{
	struct inode		*inode = page->mapping->host;
	int			dirty, delalloc, unmapped, unwritten;
	struct writeback_control wbc = {
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = 1,
	};

	xfs_page_trace(XFS_RELEASEPAGE_ENTER, inode, page, gfp_mask);

1194 1195 1196
	if (!page_has_buffers(page))
		return 0;

1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224
	xfs_count_page_state(page, &delalloc, &unmapped, &unwritten);
	if (!delalloc && !unwritten)
		goto free_buffers;

	if (!(gfp_mask & __GFP_FS))
		return 0;

	/* If we are already inside a transaction or the thread cannot
	 * do I/O, we cannot release this page.
	 */
	if (PFLAGS_TEST_FSTRANS())
		return 0;

	/*
	 * Convert delalloc space to real space, do not flush the
	 * data out to disk, that will be done by the caller.
	 * Never need to allocate space here - we will always
	 * come back to writepage in that case.
	 */
	dirty = xfs_page_state_convert(inode, page, &wbc, 0, 0);
	if (dirty == 0 && !unwritten)
		goto free_buffers;
	return 0;

free_buffers:
	return try_to_free_buffers(page);
}

L
Linus Torvalds 已提交
1225
STATIC int
1226
__xfs_get_blocks(
L
Linus Torvalds 已提交
1227 1228 1229 1230 1231 1232 1233
	struct inode		*inode,
	sector_t		iblock,
	struct buffer_head	*bh_result,
	int			create,
	int			direct,
	bmapi_flags_t		flags)
{
1234
	vnode_t			*vp = vn_from_inode(inode);
L
Linus Torvalds 已提交
1235
	xfs_iomap_t		iomap;
1236 1237
	xfs_off_t		offset;
	ssize_t			size;
1238
	int			niomap = 1;
L
Linus Torvalds 已提交
1239 1240
	int			error;

1241
	offset = (xfs_off_t)iblock << inode->i_blkbits;
1242 1243
	ASSERT(bh_result->b_size >= (1 << inode->i_blkbits));
	size = bh_result->b_size;
L
Linus Torvalds 已提交
1244
	VOP_BMAP(vp, offset, size,
1245
		create ? flags : BMAPI_READ, &iomap, &niomap, error);
L
Linus Torvalds 已提交
1246 1247
	if (error)
		return -error;
1248
	if (niomap == 0)
L
Linus Torvalds 已提交
1249 1250 1251
		return 0;

	if (iomap.iomap_bn != IOMAP_DADDR_NULL) {
1252 1253
		/*
		 * For unwritten extents do not report a disk address on
L
Linus Torvalds 已提交
1254 1255 1256
		 * the read case (treat as if we're reading into a hole).
		 */
		if (create || !(iomap.iomap_flags & IOMAP_UNWRITTEN)) {
1257 1258
			xfs_map_buffer(bh_result, &iomap, offset,
				       inode->i_blkbits);
L
Linus Torvalds 已提交
1259 1260 1261 1262 1263 1264 1265 1266 1267
		}
		if (create && (iomap.iomap_flags & IOMAP_UNWRITTEN)) {
			if (direct)
				bh_result->b_private = inode;
			set_buffer_unwritten(bh_result);
			set_buffer_delay(bh_result);
		}
	}

1268 1269 1270 1271
	/*
	 * If this is a realtime file, data may be on a different device.
	 * to that pointed to from the buffer_head b_bdev currently.
	 */
1272
	bh_result->b_bdev = iomap.iomap_target->bt_bdev;
L
Linus Torvalds 已提交
1273

1274 1275 1276 1277
	/*
	 * If we previously allocated a block out beyond eof and we are
	 * now coming back to use it then we will need to flag it as new
	 * even if it has a disk address.
L
Linus Torvalds 已提交
1278 1279 1280
	 */
	if (create &&
	    ((!buffer_mapped(bh_result) && !buffer_uptodate(bh_result)) ||
1281
	     (offset >= i_size_read(inode)) || (iomap.iomap_flags & IOMAP_NEW)))
L
Linus Torvalds 已提交
1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292
		set_buffer_new(bh_result);

	if (iomap.iomap_flags & IOMAP_DELAY) {
		BUG_ON(direct);
		if (create) {
			set_buffer_uptodate(bh_result);
			set_buffer_mapped(bh_result);
			set_buffer_delay(bh_result);
		}
	}

1293
	if (direct || size > (1 << inode->i_blkbits)) {
1294 1295
		ASSERT(iomap.iomap_bsize - iomap.iomap_delta > 0);
		offset = min_t(xfs_off_t,
1296 1297
				iomap.iomap_bsize - iomap.iomap_delta, size);
		bh_result->b_size = (ssize_t)min_t(xfs_off_t, LONG_MAX, offset);
L
Linus Torvalds 已提交
1298 1299 1300 1301 1302 1303
	}

	return 0;
}

int
1304
xfs_get_blocks(
L
Linus Torvalds 已提交
1305 1306 1307 1308 1309
	struct inode		*inode,
	sector_t		iblock,
	struct buffer_head	*bh_result,
	int			create)
{
1310
	return __xfs_get_blocks(inode, iblock,
1311
				bh_result, create, 0, BMAPI_WRITE);
L
Linus Torvalds 已提交
1312 1313 1314
}

STATIC int
1315
xfs_get_blocks_direct(
L
Linus Torvalds 已提交
1316 1317 1318 1319 1320
	struct inode		*inode,
	sector_t		iblock,
	struct buffer_head	*bh_result,
	int			create)
{
1321
	return __xfs_get_blocks(inode, iblock,
1322
				bh_result, create, 1, BMAPI_WRITE|BMAPI_DIRECT);
L
Linus Torvalds 已提交
1323 1324
}

1325
STATIC void
1326
xfs_end_io_direct(
1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
	struct kiocb	*iocb,
	loff_t		offset,
	ssize_t		size,
	void		*private)
{
	xfs_ioend_t	*ioend = iocb->private;

	/*
	 * Non-NULL private data means we need to issue a transaction to
	 * convert a range from unwritten to written extents.  This needs
1337
	 * to happen from process context but aio+dio I/O completion
1338
	 * happens from irq context so we need to defer it to a workqueue.
1339
	 * This is not necessary for synchronous direct I/O, but we do
1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355
	 * it anyway to keep the code uniform and simpler.
	 *
	 * The core direct I/O code might be changed to always call the
	 * completion handler in the future, in which case all this can
	 * go away.
	 */
	if (private && size > 0) {
		ioend->io_offset = offset;
		ioend->io_size = size;
		xfs_finish_ioend(ioend);
	} else {
		ASSERT(size >= 0);
		xfs_destroy_ioend(ioend);
	}

	/*
1356
	 * blockdev_direct_IO can return an error even after the I/O
1357 1358 1359 1360 1361 1362
	 * completion handler was called.  Thus we need to protect
	 * against double-freeing.
	 */
	iocb->private = NULL;
}

L
Linus Torvalds 已提交
1363
STATIC ssize_t
1364
xfs_vm_direct_IO(
L
Linus Torvalds 已提交
1365 1366 1367 1368 1369 1370 1371 1372
	int			rw,
	struct kiocb		*iocb,
	const struct iovec	*iov,
	loff_t			offset,
	unsigned long		nr_segs)
{
	struct file	*file = iocb->ki_filp;
	struct inode	*inode = file->f_mapping->host;
1373
	vnode_t		*vp = vn_from_inode(inode);
L
Linus Torvalds 已提交
1374 1375 1376
	xfs_iomap_t	iomap;
	int		maps = 1;
	int		error;
1377
	ssize_t		ret;
L
Linus Torvalds 已提交
1378 1379 1380 1381 1382

	VOP_BMAP(vp, offset, 0, BMAPI_DEVICE, &iomap, &maps, error);
	if (error)
		return -error;

1383
	iocb->private = xfs_alloc_ioend(inode, IOMAP_UNWRITTEN);
1384 1385

	ret = blockdev_direct_IO_own_locking(rw, iocb, inode,
1386
		iomap.iomap_target->bt_bdev,
L
Linus Torvalds 已提交
1387
		iov, offset, nr_segs,
1388 1389
		xfs_get_blocks_direct,
		xfs_end_io_direct);
1390 1391 1392 1393

	if (unlikely(ret <= 0 && iocb->private))
		xfs_destroy_ioend(iocb->private);
	return ret;
L
Linus Torvalds 已提交
1394 1395
}

1396
STATIC int
1397
xfs_vm_prepare_write(
1398 1399 1400 1401 1402
	struct file		*file,
	struct page		*page,
	unsigned int		from,
	unsigned int		to)
{
1403
	return block_prepare_write(page, from, to, xfs_get_blocks);
1404
}
L
Linus Torvalds 已提交
1405 1406

STATIC sector_t
1407
xfs_vm_bmap(
L
Linus Torvalds 已提交
1408 1409 1410 1411
	struct address_space	*mapping,
	sector_t		block)
{
	struct inode		*inode = (struct inode *)mapping->host;
1412
	vnode_t			*vp = vn_from_inode(inode);
L
Linus Torvalds 已提交
1413 1414
	int			error;

1415
	vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address);
L
Linus Torvalds 已提交
1416 1417 1418 1419

	VOP_RWLOCK(vp, VRWLOCK_READ);
	VOP_FLUSH_PAGES(vp, (xfs_off_t)0, -1, 0, FI_REMAPF, error);
	VOP_RWUNLOCK(vp, VRWLOCK_READ);
1420
	return generic_block_bmap(mapping, block, xfs_get_blocks);
L
Linus Torvalds 已提交
1421 1422 1423
}

STATIC int
1424
xfs_vm_readpage(
L
Linus Torvalds 已提交
1425 1426 1427
	struct file		*unused,
	struct page		*page)
{
1428
	return mpage_readpage(page, xfs_get_blocks);
L
Linus Torvalds 已提交
1429 1430 1431
}

STATIC int
1432
xfs_vm_readpages(
L
Linus Torvalds 已提交
1433 1434 1435 1436 1437
	struct file		*unused,
	struct address_space	*mapping,
	struct list_head	*pages,
	unsigned		nr_pages)
{
1438
	return mpage_readpages(mapping, pages, nr_pages, xfs_get_blocks);
L
Linus Torvalds 已提交
1439 1440
}

1441
STATIC void
1442
xfs_vm_invalidatepage(
1443 1444 1445 1446 1447
	struct page		*page,
	unsigned long		offset)
{
	xfs_page_trace(XFS_INVALIDPAGE_ENTER,
			page->mapping->host, page, offset);
1448
	block_invalidatepage(page, offset);
1449 1450
}

1451 1452 1453 1454
struct address_space_operations xfs_address_space_operations = {
	.readpage		= xfs_vm_readpage,
	.readpages		= xfs_vm_readpages,
	.writepage		= xfs_vm_writepage,
L
Linus Torvalds 已提交
1455
	.sync_page		= block_sync_page,
1456 1457
	.releasepage		= xfs_vm_releasepage,
	.invalidatepage		= xfs_vm_invalidatepage,
1458
	.prepare_write		= xfs_vm_prepare_write,
L
Linus Torvalds 已提交
1459
	.commit_write		= generic_commit_write,
1460 1461
	.bmap			= xfs_vm_bmap,
	.direct_IO		= xfs_vm_direct_IO,
1462
	.migratepage		= buffer_migrate_page,
L
Linus Torvalds 已提交
1463
};