jfs_metapage.c 20.1 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *   Copyright (C) International Business Machines Corp., 2000-2005
L
Linus Torvalds 已提交
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
 *   Portions Copyright (C) Christoph Hellwig, 2001-2002
 *
 *   This program is free software;  you can redistribute it and/or modify
 *   it under the terms of the GNU General Public License as published by
 *   the Free Software Foundation; either version 2 of the License, or 
 *   (at your option) any later version.
 * 
 *   This program is distributed in the hope that it will 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.
 *
 *   You should have received a copy of the GNU General Public License
 *   along with this program;  if not, write to the Free Software 
 *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 */

#include <linux/fs.h>
21 22
#include <linux/mm.h>
#include <linux/bio.h>
L
Linus Torvalds 已提交
23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
#include <linux/init.h>
#include <linux/buffer_head.h>
#include <linux/mempool.h>
#include "jfs_incore.h"
#include "jfs_superblock.h"
#include "jfs_filsys.h"
#include "jfs_metapage.h"
#include "jfs_txnmgr.h"
#include "jfs_debug.h"

#ifdef CONFIG_JFS_STATISTICS
static struct {
	uint	pagealloc;	/* # of page allocations */
	uint	pagefree;	/* # of page frees */
	uint	lockwait;	/* # of sleeping lock_metapage() calls */
} mpStat;
#endif

41 42
#define metapage_locked(mp) test_bit(META_locked, &(mp)->flag)
#define trylock_metapage(mp) test_and_set_bit(META_locked, &(mp)->flag)
L
Linus Torvalds 已提交
43 44 45 46 47 48 49

static inline void unlock_metapage(struct metapage *mp)
{
	clear_bit(META_locked, &mp->flag);
	wake_up(&mp->wait);
}

50
static inline void __lock_metapage(struct metapage *mp)
L
Linus Torvalds 已提交
51 52 53 54 55 56 57
{
	DECLARE_WAITQUEUE(wait, current);
	INCREMENT(mpStat.lockwait);
	add_wait_queue_exclusive(&mp->wait, &wait);
	do {
		set_current_state(TASK_UNINTERRUPTIBLE);
		if (metapage_locked(mp)) {
58
			unlock_page(mp->page);
L
Linus Torvalds 已提交
59
			schedule();
60
			lock_page(mp->page);
L
Linus Torvalds 已提交
61 62 63 64 65 66
		}
	} while (trylock_metapage(mp));
	__set_current_state(TASK_RUNNING);
	remove_wait_queue(&mp->wait, &wait);
}

67 68 69
/*
 * Must have mp->page locked
 */
L
Linus Torvalds 已提交
70 71 72 73 74 75 76 77 78 79
static inline void lock_metapage(struct metapage *mp)
{
	if (trylock_metapage(mp))
		__lock_metapage(mp);
}

#define METAPOOL_MIN_PAGES 32
static kmem_cache_t *metapage_cache;
static mempool_t *metapage_mempool;

80 81 82 83 84 85 86 87 88
#define MPS_PER_PAGE (PAGE_CACHE_SIZE >> L2PSIZE)

#if MPS_PER_PAGE > 1

struct meta_anchor {
	int mp_count;
	atomic_t io_count;
	struct metapage *mp[MPS_PER_PAGE];
};
H
Hugh Dickins 已提交
89
#define mp_anchor(page) ((struct meta_anchor *)page_private(page))
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106

static inline struct metapage *page_to_mp(struct page *page, uint offset)
{
	if (!PagePrivate(page))
		return NULL;
	return mp_anchor(page)->mp[offset >> L2PSIZE];
}

static inline int insert_metapage(struct page *page, struct metapage *mp)
{
	struct meta_anchor *a;
	int index;
	int l2mp_blocks;	/* log2 blocks per metapage */

	if (PagePrivate(page))
		a = mp_anchor(page);
	else {
E
Eric Sesterhenn 已提交
107
		a = kzalloc(sizeof(struct meta_anchor), GFP_NOFS);
108 109
		if (!a)
			return -ENOMEM;
H
Hugh Dickins 已提交
110
		set_page_private(page, (unsigned long)a);
111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137
		SetPagePrivate(page);
		kmap(page);
	}

	if (mp) {
		l2mp_blocks = L2PSIZE - page->mapping->host->i_blkbits;
		index = (mp->index >> l2mp_blocks) & (MPS_PER_PAGE - 1);
		a->mp_count++;
		a->mp[index] = mp;
	}

	return 0;
}

static inline void remove_metapage(struct page *page, struct metapage *mp)
{
	struct meta_anchor *a = mp_anchor(page);
	int l2mp_blocks = L2PSIZE - page->mapping->host->i_blkbits;
	int index;

	index = (mp->index >> l2mp_blocks) & (MPS_PER_PAGE - 1);

	BUG_ON(a->mp[index] != mp);

	a->mp[index] = NULL;
	if (--a->mp_count == 0) {
		kfree(a);
H
Hugh Dickins 已提交
138
		set_page_private(page, 0);
139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157
		ClearPagePrivate(page);
		kunmap(page);
	}
}

static inline void inc_io(struct page *page)
{
	atomic_inc(&mp_anchor(page)->io_count);
}

static inline void dec_io(struct page *page, void (*handler) (struct page *))
{
	if (atomic_dec_and_test(&mp_anchor(page)->io_count))
		handler(page);
}

#else
static inline struct metapage *page_to_mp(struct page *page, uint offset)
{
H
Hugh Dickins 已提交
158
	return PagePrivate(page) ? (struct metapage *)page_private(page) : NULL;
159 160 161 162 163
}

static inline int insert_metapage(struct page *page, struct metapage *mp)
{
	if (mp) {
H
Hugh Dickins 已提交
164
		set_page_private(page, (unsigned long)mp);
165 166 167 168 169 170 171 172
		SetPagePrivate(page);
		kmap(page);
	}
	return 0;
}

static inline void remove_metapage(struct page *page, struct metapage *mp)
{
H
Hugh Dickins 已提交
173
	set_page_private(page, 0);
174 175 176 177 178 179 180 181 182
	ClearPagePrivate(page);
	kunmap(page);
}

#define inc_io(page) do {} while(0)
#define dec_io(page, handler) handler(page)

#endif

L
Linus Torvalds 已提交
183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199
static void init_once(void *foo, kmem_cache_t *cachep, unsigned long flags)
{
	struct metapage *mp = (struct metapage *)foo;

	if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
	    SLAB_CTOR_CONSTRUCTOR) {
		mp->lid = 0;
		mp->lsn = 0;
		mp->flag = 0;
		mp->data = NULL;
		mp->clsn = 0;
		mp->log = NULL;
		set_bit(META_free, &mp->flag);
		init_waitqueue_head(&mp->wait);
	}
}

A
Al Viro 已提交
200
static inline struct metapage *alloc_metapage(gfp_t gfp_mask)
L
Linus Torvalds 已提交
201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222
{
	return mempool_alloc(metapage_mempool, gfp_mask);
}

static inline void free_metapage(struct metapage *mp)
{
	mp->flag = 0;
	set_bit(META_free, &mp->flag);

	mempool_free(mp, metapage_mempool);
}

int __init metapage_init(void)
{
	/*
	 * Allocate the metapage structures
	 */
	metapage_cache = kmem_cache_create("jfs_mp", sizeof(struct metapage),
					   0, 0, init_once, NULL);
	if (metapage_cache == NULL)
		return -ENOMEM;

223 224
	metapage_mempool = mempool_create_slab_pool(METAPOOL_MIN_PAGES,
						    metapage_cache);
L
Linus Torvalds 已提交
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239

	if (metapage_mempool == NULL) {
		kmem_cache_destroy(metapage_cache);
		return -ENOMEM;
	}

	return 0;
}

void metapage_exit(void)
{
	mempool_destroy(metapage_mempool);
	kmem_cache_destroy(metapage_cache);
}

240 241 242 243 244 245 246 247 248 249
static inline void drop_metapage(struct page *page, struct metapage *mp)
{
	if (mp->count || mp->nohomeok || test_bit(META_dirty, &mp->flag) ||
	    test_bit(META_io, &mp->flag))
		return;
	remove_metapage(page, mp);
	INCREMENT(mpStat.pagefree);
	free_metapage(mp);
}

L
Linus Torvalds 已提交
250
/*
251
 * Metapage address space operations
L
Linus Torvalds 已提交
252
 */
253 254 255

static sector_t metapage_get_blocks(struct inode *inode, sector_t lblock,
				    unsigned int *len)
L
Linus Torvalds 已提交
256
{
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
	int rc = 0;
	int xflag;
	s64 xaddr;
	sector_t file_blocks = (inode->i_size + inode->i_blksize - 1) >>
			       inode->i_blkbits;

	if (lblock >= file_blocks)
		return 0;
	if (lblock + *len > file_blocks)
		*len = file_blocks - lblock;

	if (inode->i_ino) {
		rc = xtLookup(inode, (s64)lblock, *len, &xflag, &xaddr, len, 0);
		if ((rc == 0) && *len)
			lblock = (sector_t)xaddr;
		else
			lblock = 0;
	} /* else no mapping */

	return lblock;
L
Linus Torvalds 已提交
277 278
}

279
static void last_read_complete(struct page *page)
L
Linus Torvalds 已提交
280
{
281 282 283 284 285 286 287 288 289 290 291 292
	if (!PageError(page))
		SetPageUptodate(page);
	unlock_page(page);
}

static int metapage_read_end_io(struct bio *bio, unsigned int bytes_done,
				int err)
{
	struct page *page = bio->bi_private;

	if (bio->bi_size)
		return 1;
L
Linus Torvalds 已提交
293

294 295 296
	if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
		printk(KERN_ERR "metapage_read_end_io: I/O error\n");
		SetPageError(page);
L
Linus Torvalds 已提交
297 298
	}

299 300 301 302
	dec_io(page, last_read_complete);
	bio_put(bio);

	return 0;
L
Linus Torvalds 已提交
303 304
}

305
static void remove_from_logsync(struct metapage *mp)
L
Linus Torvalds 已提交
306
{
307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
	struct jfs_log *log = mp->log;
	unsigned long flags;
/*
 * This can race.  Recheck that log hasn't been set to null, and after
 * acquiring logsync lock, recheck lsn
 */
	if (!log)
		return;

	LOGSYNC_LOCK(log, flags);
	if (mp->lsn) {
		mp->log = NULL;
		mp->lsn = 0;
		mp->clsn = 0;
		log->count--;
		list_del(&mp->synclist);
	}
	LOGSYNC_UNLOCK(log, flags);
}
L
Linus Torvalds 已提交
326

327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
static void last_write_complete(struct page *page)
{
	struct metapage *mp;
	unsigned int offset;

	for (offset = 0; offset < PAGE_CACHE_SIZE; offset += PSIZE) {
		mp = page_to_mp(page, offset);
		if (mp && test_bit(META_io, &mp->flag)) {
			if (mp->lsn)
				remove_from_logsync(mp);
			clear_bit(META_io, &mp->flag);
		}
		/*
		 * I'd like to call drop_metapage here, but I don't think it's
		 * safe unless I have the page locked
		 */
	}
	end_page_writeback(page);
L
Linus Torvalds 已提交
345 346
}

347 348
static int metapage_write_end_io(struct bio *bio, unsigned int bytes_done,
				 int err)
L
Linus Torvalds 已提交
349
{
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
	struct page *page = bio->bi_private;

	BUG_ON(!PagePrivate(page));

	if (bio->bi_size)
		return 1;

	if (! test_bit(BIO_UPTODATE, &bio->bi_flags)) {
		printk(KERN_ERR "metapage_write_end_io: I/O error\n");
		SetPageError(page);
	}
	dec_io(page, last_write_complete);
	bio_put(bio);
	return 0;
}

static int metapage_writepage(struct page *page, struct writeback_control *wbc)
{
	struct bio *bio = NULL;
	unsigned int block_offset;	/* block offset of mp within page */
	struct inode *inode = page->mapping->host;
	unsigned int blocks_per_mp = JFS_SBI(inode->i_sb)->nbperpage;
	unsigned int len;
	unsigned int xlen;
	struct metapage *mp;
	int redirty = 0;
	sector_t lblock;
	sector_t pblock;
	sector_t next_block = 0;
	sector_t page_start;
	unsigned long bio_bytes = 0;
	unsigned long bio_offset = 0;
	unsigned int offset;

	page_start = (sector_t)page->index <<
		     (PAGE_CACHE_SHIFT - inode->i_blkbits);
	BUG_ON(!PageLocked(page));
	BUG_ON(PageWriteback(page));

	for (offset = 0; offset < PAGE_CACHE_SIZE; offset += PSIZE) {
		mp = page_to_mp(page, offset);

		if (!mp || !test_bit(META_dirty, &mp->flag))
			continue;

		if (mp->nohomeok && !test_bit(META_forcewrite, &mp->flag)) {
			redirty = 1;
397 398 399 400 401 402
			/*
			 * Make sure this page isn't blocked indefinitely.
			 * If the journal isn't undergoing I/O, push it
			 */
			if (mp->log && !(mp->log->cflag & logGC_PAGEOUT))
				jfs_flush_journal(mp->log, 0);
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526
			continue;
		}

		clear_bit(META_dirty, &mp->flag);
		block_offset = offset >> inode->i_blkbits;
		lblock = page_start + block_offset;
		if (bio) {
			if (xlen && lblock == next_block) {
				/* Contiguous, in memory & on disk */
				len = min(xlen, blocks_per_mp);
				xlen -= len;
				bio_bytes += len << inode->i_blkbits;
				set_bit(META_io, &mp->flag);
				continue;
			}
			/* Not contiguous */
			if (bio_add_page(bio, page, bio_bytes, bio_offset) <
			    bio_bytes)
				goto add_failed;
			/*
			 * Increment counter before submitting i/o to keep
			 * count from hitting zero before we're through
			 */
			inc_io(page);
			if (!bio->bi_size)
				goto dump_bio;
			submit_bio(WRITE, bio);
			bio = NULL;
		} else {
			set_page_writeback(page);
			inc_io(page);
		}
		xlen = (PAGE_CACHE_SIZE - offset) >> inode->i_blkbits;
		pblock = metapage_get_blocks(inode, lblock, &xlen);
		if (!pblock) {
			/* Need better error handling */
			printk(KERN_ERR "JFS: metapage_get_blocks failed\n");
			dec_io(page, last_write_complete);
			continue;
		}
		set_bit(META_io, &mp->flag);
		len = min(xlen, (uint) JFS_SBI(inode->i_sb)->nbperpage);

		bio = bio_alloc(GFP_NOFS, 1);
		bio->bi_bdev = inode->i_sb->s_bdev;
		bio->bi_sector = pblock << (inode->i_blkbits - 9);
		bio->bi_end_io = metapage_write_end_io;
		bio->bi_private = page;

		/* Don't call bio_add_page yet, we may add to this vec */
		bio_offset = offset;
		bio_bytes = len << inode->i_blkbits;

		xlen -= len;
		next_block = lblock + len;
	}
	if (bio) {
		if (bio_add_page(bio, page, bio_bytes, bio_offset) < bio_bytes)
				goto add_failed;
		if (!bio->bi_size)
			goto dump_bio;
		
		submit_bio(WRITE, bio);
	}
	if (redirty)
		redirty_page_for_writepage(wbc, page);

	unlock_page(page);

	return 0;
add_failed:
	/* We should never reach here, since we're only adding one vec */
	printk(KERN_ERR "JFS: bio_add_page failed unexpectedly\n");
	goto skip;
dump_bio:
	dump_mem("bio", bio, sizeof(*bio));
skip:
	bio_put(bio);
	unlock_page(page);
	dec_io(page, last_write_complete);

	return -EIO;
}

static int metapage_readpage(struct file *fp, struct page *page)
{
	struct inode *inode = page->mapping->host;
	struct bio *bio = NULL;
	unsigned int block_offset;
	unsigned int blocks_per_page = PAGE_CACHE_SIZE >> inode->i_blkbits;
	sector_t page_start;	/* address of page in fs blocks */
	sector_t pblock;
	unsigned int xlen;
	unsigned int len;
	unsigned int offset;

	BUG_ON(!PageLocked(page));
	page_start = (sector_t)page->index <<
		     (PAGE_CACHE_SHIFT - inode->i_blkbits);

	block_offset = 0;
	while (block_offset < blocks_per_page) {
		xlen = blocks_per_page - block_offset;
		pblock = metapage_get_blocks(inode, page_start + block_offset,
					     &xlen);
		if (pblock) {
			if (!PagePrivate(page))
				insert_metapage(page, NULL);
			inc_io(page);
			if (bio)
				submit_bio(READ, bio);

			bio = bio_alloc(GFP_NOFS, 1);
			bio->bi_bdev = inode->i_sb->s_bdev;
			bio->bi_sector = pblock << (inode->i_blkbits - 9);
			bio->bi_end_io = metapage_read_end_io;
			bio->bi_private = page;
			len = xlen << inode->i_blkbits;
			offset = block_offset << inode->i_blkbits;
			if (bio_add_page(bio, page, len, offset) < len)
				goto add_failed;
			block_offset += xlen;
		} else
			block_offset++;
L
Linus Torvalds 已提交
527
	}
528 529 530 531 532 533
	if (bio)
		submit_bio(READ, bio);
	else
		unlock_page(page);

	return 0;
L
Linus Torvalds 已提交
534

535 536 537 538 539
add_failed:
	printk(KERN_ERR "JFS: bio_add_page failed unexpectedly\n");
	bio_put(bio);
	dec_io(page, last_read_complete);
	return -EIO;
L
Linus Torvalds 已提交
540 541
}

A
Al Viro 已提交
542
static int metapage_releasepage(struct page *page, gfp_t gfp_mask)
543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580
{
	struct metapage *mp;
	int busy = 0;
	unsigned int offset;

	for (offset = 0; offset < PAGE_CACHE_SIZE; offset += PSIZE) {
		mp = page_to_mp(page, offset);

		if (!mp)
			continue;

		jfs_info("metapage_releasepage: mp = 0x%p", mp);
		if (mp->count || mp->nohomeok) {
			jfs_info("count = %ld, nohomeok = %d", mp->count,
				 mp->nohomeok);
			busy = 1;
			continue;
		}
		wait_on_page_writeback(page);
		//WARN_ON(test_bit(META_dirty, &mp->flag));
		if (test_bit(META_dirty, &mp->flag)) {
			dump_mem("dirty mp in metapage_releasepage", mp,
				 sizeof(struct metapage));
			dump_mem("page", page, sizeof(struct page));
			dump_stack();
		}
		if (mp->lsn)
			remove_from_logsync(mp);
		remove_metapage(page, mp);
		INCREMENT(mpStat.pagefree);
		free_metapage(mp);
	}
	if (busy)
		return -1;

	return 0;
}

581
static void metapage_invalidatepage(struct page *page, unsigned long offset)
582 583 584
{
	BUG_ON(offset);

585
	BUG_ON(PageWriteback(page));
586

587
	metapage_releasepage(page, 0);
588 589 590 591 592 593 594 595 596 597 598
}

struct address_space_operations jfs_metapage_aops = {
	.readpage	= metapage_readpage,
	.writepage	= metapage_writepage,
	.sync_page	= block_sync_page,
	.releasepage	= metapage_releasepage,
	.invalidatepage	= metapage_invalidatepage,
	.set_page_dirty	= __set_page_dirty_nobuffers,
};

L
Linus Torvalds 已提交
599 600 601 602 603 604 605
struct metapage *__get_metapage(struct inode *inode, unsigned long lblock,
				unsigned int size, int absolute,
				unsigned long new)
{
	int l2BlocksPerPage;
	int l2bsize;
	struct address_space *mapping;
606 607
	struct metapage *mp = NULL;
	struct page *page;
L
Linus Torvalds 已提交
608 609 610
	unsigned long page_index;
	unsigned long page_offset;

611 612 613 614 615 616 617 618 619 620 621 622 623
	jfs_info("__get_metapage: ino = %ld, lblock = 0x%lx, abs=%d",
		 inode->i_ino, lblock, absolute);

	l2bsize = inode->i_blkbits;
	l2BlocksPerPage = PAGE_CACHE_SHIFT - l2bsize;
	page_index = lblock >> l2BlocksPerPage;
	page_offset = (lblock - (page_index << l2BlocksPerPage)) << l2bsize;
	if ((page_offset + size) > PAGE_CACHE_SIZE) {
		jfs_err("MetaData crosses page boundary!!");
		jfs_err("lblock = %lx, size  = %d", lblock, size);
		dump_stack();
		return NULL;
	}
L
Linus Torvalds 已提交
624
	if (absolute)
625
		mapping = JFS_SBI(inode->i_sb)->direct_inode->i_mapping;
L
Linus Torvalds 已提交
626 627 628 629 630 631 632 633 634 635 636
	else {
		/*
		 * If an nfs client tries to read an inode that is larger
		 * than any existing inodes, we may try to read past the
		 * end of the inode map
		 */
		if ((lblock << inode->i_blkbits) >= inode->i_size)
			return NULL;
		mapping = inode->i_mapping;
	}

637 638 639 640 641 642 643 644 645 646
	if (new && (PSIZE == PAGE_CACHE_SIZE)) {
		page = grab_cache_page(mapping, page_index);
		if (!page) {
			jfs_err("grab_cache_page failed!");
			return NULL;
		}
		SetPageUptodate(page);
	} else {
		page = read_cache_page(mapping, page_index,
			    (filler_t *)mapping->a_ops->readpage, NULL);
D
Dave Kleikamp 已提交
647
		if (IS_ERR(page) || !PageUptodate(page)) {
648 649 650 651 652 653 654
			jfs_err("read_cache_page failed!");
			return NULL;
		}
		lock_page(page);
	}

	mp = page_to_mp(page, page_offset);
L
Linus Torvalds 已提交
655
	if (mp) {
656 657 658 659 660 661 662
		if (mp->logical_size != size) {
			jfs_error(inode->i_sb,
				  "__get_metapage: mp->logical_size != size");
			jfs_err("logical_size = %d, size = %d",
				mp->logical_size, size);
			dump_stack();
			goto unlock; 
L
Linus Torvalds 已提交
663 664 665 666 667 668 669 670
		}
		mp->count++;
		lock_metapage(mp);
		if (test_bit(META_discard, &mp->flag)) {
			if (!new) {
				jfs_error(inode->i_sb,
					  "__get_metapage: using a "
					  "discarded metapage");
671 672
				discard_metapage(mp);
				goto unlock; 
L
Linus Torvalds 已提交
673 674 675 676
			}
			clear_bit(META_discard, &mp->flag);
		}
	} else {
677 678 679
		INCREMENT(mpStat.pagealloc);
		mp = alloc_metapage(GFP_NOFS);
		mp->page = page;
L
Linus Torvalds 已提交
680 681 682
		mp->flag = 0;
		mp->xflag = COMMIT_PAGE;
		mp->count = 1;
683
		mp->nohomeok = 0;
L
Linus Torvalds 已提交
684
		mp->logical_size = size;
685 686 687 688 689
		mp->data = page_address(page) + page_offset;
		mp->index = lblock;
		if (unlikely(insert_metapage(page, mp))) {
			free_metapage(mp);
			goto unlock;
L
Linus Torvalds 已提交
690
		}
691
		lock_metapage(mp);
L
Linus Torvalds 已提交
692 693
	}

694 695
	if (new) {
		jfs_info("zeroing mp = 0x%p", mp);
L
Linus Torvalds 已提交
696
		memset(mp->data, 0, PSIZE);
697
	}
L
Linus Torvalds 已提交
698

699 700
	unlock_page(page);
	jfs_info("__get_metapage: returning = 0x%p data = 0x%p", mp, mp->data);
L
Linus Torvalds 已提交
701 702
	return mp;

703 704
unlock:
	unlock_page(page);
L
Linus Torvalds 已提交
705 706 707
	return NULL;
}

708
void grab_metapage(struct metapage * mp)
L
Linus Torvalds 已提交
709
{
710 711 712
	jfs_info("grab_metapage: mp = 0x%p", mp);
	page_cache_get(mp->page);
	lock_page(mp->page);
L
Linus Torvalds 已提交
713
	mp->count++;
714 715
	lock_metapage(mp);
	unlock_page(mp->page);
L
Linus Torvalds 已提交
716 717
}

718
void force_metapage(struct metapage *mp)
L
Linus Torvalds 已提交
719
{
720 721 722 723 724 725 726 727 728 729 730
	struct page *page = mp->page;
	jfs_info("force_metapage: mp = 0x%p", mp);
	set_bit(META_forcewrite, &mp->flag);
	clear_bit(META_sync, &mp->flag);
	page_cache_get(page);
	lock_page(page);
	set_page_dirty(page);
	write_one_page(page, 1);
	clear_bit(META_forcewrite, &mp->flag);
	page_cache_release(page);
}
L
Linus Torvalds 已提交
731

732
void hold_metapage(struct metapage *mp)
733
{
L
Linus Torvalds 已提交
734
	lock_page(mp->page);
735 736
}

737
void put_metapage(struct metapage *mp)
738 739 740
{
	if (mp->count || mp->nohomeok) {
		/* Someone else will release this */
L
Linus Torvalds 已提交
741 742 743
		unlock_page(mp->page);
		return;
	}
744 745 746
	page_cache_get(mp->page);
	mp->count++;
	lock_metapage(mp);
L
Linus Torvalds 已提交
747
	unlock_page(mp->page);
748
	release_metapage(mp);
L
Linus Torvalds 已提交
749 750 751 752
}

void release_metapage(struct metapage * mp)
{
753
	struct page *page = mp->page;
L
Linus Torvalds 已提交
754 755
	jfs_info("release_metapage: mp = 0x%p, flag = 0x%lx", mp, mp->flag);

756 757 758 759
	BUG_ON(!page);

	lock_page(page);
	unlock_metapage(mp);
L
Linus Torvalds 已提交
760 761

	assert(mp->count);
762 763 764
	if (--mp->count || mp->nohomeok) {
		unlock_page(page);
		page_cache_release(page);
L
Linus Torvalds 已提交
765 766 767
		return;
	}

768 769
	if (test_bit(META_dirty, &mp->flag)) {
		set_page_dirty(page);
L
Linus Torvalds 已提交
770 771
		if (test_bit(META_sync, &mp->flag)) {
			clear_bit(META_sync, &mp->flag);
772 773
			write_one_page(page, 1);
			lock_page(page); /* write_one_page unlocks the page */
L
Linus Torvalds 已提交
774
		}
775 776
	} else if (mp->lsn)	/* discard_metapage doesn't remove it */
		remove_from_logsync(mp);
L
Linus Torvalds 已提交
777

778 779 780 781 782 783
#if MPS_PER_PAGE == 1
	/*
	 * If we know this is the only thing in the page, we can throw
	 * the page out of the page cache.  If pages are larger, we
	 * don't want to do this.
	 */
L
Linus Torvalds 已提交
784

785 786 787 788
	/* Retest mp->count since we may have released page lock */
	if (test_bit(META_discard, &mp->flag) && !mp->count) {
		clear_page_dirty(page);
		ClearPageUptodate(page);
L
Linus Torvalds 已提交
789
	}
790 791 792 793 794 795
#else
	/* Try to keep metapages from using up too much memory */
	drop_metapage(page, mp);
#endif
	unlock_page(page);
	page_cache_release(page);
L
Linus Torvalds 已提交
796 797 798 799
}

void __invalidate_metapages(struct inode *ip, s64 addr, int len)
{
800
	sector_t lblock;
L
Linus Torvalds 已提交
801
	int l2BlocksPerPage = PAGE_CACHE_SHIFT - ip->i_blkbits;
802
	int BlocksPerPage = 1 << l2BlocksPerPage;
L
Linus Torvalds 已提交
803
	/* All callers are interested in block device's mapping */
804 805
	struct address_space *mapping =
		JFS_SBI(ip->i_sb)->direct_inode->i_mapping;
L
Linus Torvalds 已提交
806 807
	struct metapage *mp;
	struct page *page;
808
	unsigned int offset;
L
Linus Torvalds 已提交
809 810

	/*
811
	 * Mark metapages to discard.  They will eventually be
L
Linus Torvalds 已提交
812 813
	 * released, but should not be written.
	 */
814 815 816 817 818 819 820 821 822 823 824 825 826
	for (lblock = addr & ~(BlocksPerPage - 1); lblock < addr + len;
	     lblock += BlocksPerPage) {
		page = find_lock_page(mapping, lblock >> l2BlocksPerPage);
		if (!page)
			continue;
		for (offset = 0; offset < PAGE_CACHE_SIZE; offset += PSIZE) {
			mp = page_to_mp(page, offset);
			if (!mp)
				continue;
			if (mp->index < addr)
				continue;
			if (mp->index >= addr + len)
				break;
L
Linus Torvalds 已提交
827 828 829

			clear_bit(META_dirty, &mp->flag);
			set_bit(META_discard, &mp->flag);
830 831
			if (mp->lsn)
				remove_from_logsync(mp);
L
Linus Torvalds 已提交
832
		}
833 834
		unlock_page(page);
		page_cache_release(page);
L
Linus Torvalds 已提交
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
	}
}

#ifdef CONFIG_JFS_STATISTICS
int jfs_mpstat_read(char *buffer, char **start, off_t offset, int length,
		    int *eof, void *data)
{
	int len = 0;
	off_t begin;

	len += sprintf(buffer,
		       "JFS Metapage statistics\n"
		       "=======================\n"
		       "page allocations = %d\n"
		       "page frees = %d\n"
		       "lock waits = %d\n",
		       mpStat.pagealloc,
		       mpStat.pagefree,
		       mpStat.lockwait);

	begin = offset;
	*start = buffer + begin;
	len -= begin;

	if (len > length)
		len = length;
	else
		*eof = 1;

	if (len < 0)
		len = 0;

	return len;
}
#endif