buffer.c 88.7 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
/*
 *  linux/fs/buffer.c
 *
 *  Copyright (C) 1991, 1992, 2002  Linus Torvalds
 */

/*
 * Start bdflush() with kernel_thread not syscall - Paul Gortmaker, 12/95
 *
 * Removed a lot of unnecessary code and simplified things now that
 * the buffer cache isn't our primary cache - Andrew Tridgell 12/96
 *
 * Speed up hash, lru, and free list operations.  Use gfp() for allocating
 * hash table, use SLAB cache for buffer heads. SMP threading.  -DaveM
 *
 * Added 32k buffer block sizes - these are required older ARM systems. - RMK
 *
 * async buffer flushing, 1999 Andrea Arcangeli <andrea@suse.de>
 */

#include <linux/kernel.h>
#include <linux/syscalls.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/percpu.h>
#include <linux/slab.h>
27
#include <linux/capability.h>
L
Linus Torvalds 已提交
28 29 30 31
#include <linux/blkdev.h>
#include <linux/file.h>
#include <linux/quotaops.h>
#include <linux/highmem.h>
32
#include <linux/export.h>
L
Linus Torvalds 已提交
33 34 35 36
#include <linux/writeback.h>
#include <linux/hash.h>
#include <linux/suspend.h>
#include <linux/buffer_head.h>
37
#include <linux/task_io_accounting_ops.h>
L
Linus Torvalds 已提交
38 39 40 41 42
#include <linux/bio.h>
#include <linux/notifier.h>
#include <linux/cpu.h>
#include <linux/bitops.h>
#include <linux/mpage.h>
I
Ingo Molnar 已提交
43
#include <linux/bit_spinlock.h>
44
#include <trace/events/block.h>
L
Linus Torvalds 已提交
45 46 47 48 49

static int fsync_buffers_list(spinlock_t *lock, struct list_head *list);

#define BH_ENTRY(list) list_entry((list), struct buffer_head, b_assoc_buffers)

50
void init_buffer(struct buffer_head *bh, bh_end_io_t *handler, void *private)
L
Linus Torvalds 已提交
51 52 53 54
{
	bh->b_end_io = handler;
	bh->b_private = private;
}
55
EXPORT_SYMBOL(init_buffer);
L
Linus Torvalds 已提交
56

57 58
inline void touch_buffer(struct buffer_head *bh)
{
59
	trace_block_touch_buffer(bh);
60 61 62 63
	mark_page_accessed(bh->b_page);
}
EXPORT_SYMBOL(touch_buffer);

64
void __lock_buffer(struct buffer_head *bh)
L
Linus Torvalds 已提交
65
{
66
	wait_on_bit_lock_io(&bh->b_state, BH_Lock, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
67 68 69
}
EXPORT_SYMBOL(__lock_buffer);

70
void unlock_buffer(struct buffer_head *bh)
L
Linus Torvalds 已提交
71
{
N
Nick Piggin 已提交
72
	clear_bit_unlock(BH_Lock, &bh->b_state);
73
	smp_mb__after_atomic();
L
Linus Torvalds 已提交
74 75
	wake_up_bit(&bh->b_state, BH_Lock);
}
76
EXPORT_SYMBOL(unlock_buffer);
L
Linus Torvalds 已提交
77

78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111
/*
 * Returns if the page has dirty or writeback buffers. If all the buffers
 * are unlocked and clean then the PageDirty information is stale. If
 * any of the pages are locked, it is assumed they are locked for IO.
 */
void buffer_check_dirty_writeback(struct page *page,
				     bool *dirty, bool *writeback)
{
	struct buffer_head *head, *bh;
	*dirty = false;
	*writeback = false;

	BUG_ON(!PageLocked(page));

	if (!page_has_buffers(page))
		return;

	if (PageWriteback(page))
		*writeback = true;

	head = page_buffers(page);
	bh = head;
	do {
		if (buffer_locked(bh))
			*writeback = true;

		if (buffer_dirty(bh))
			*dirty = true;

		bh = bh->b_this_page;
	} while (bh != head);
}
EXPORT_SYMBOL(buffer_check_dirty_writeback);

L
Linus Torvalds 已提交
112 113 114 115 116 117 118
/*
 * Block until a buffer comes unlocked.  This doesn't stop it
 * from becoming locked again - you have to lock it yourself
 * if you want to preserve its state.
 */
void __wait_on_buffer(struct buffer_head * bh)
{
119
	wait_on_bit_io(&bh->b_state, BH_Lock, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
120
}
121
EXPORT_SYMBOL(__wait_on_buffer);
L
Linus Torvalds 已提交
122 123 124 125 126

static void
__clear_page_buffers(struct page *page)
{
	ClearPagePrivate(page);
H
Hugh Dickins 已提交
127
	set_page_private(page, 0);
L
Linus Torvalds 已提交
128 129 130
	page_cache_release(page);
}

131 132 133 134 135 136 137 138 139

static int quiet_error(struct buffer_head *bh)
{
	if (!test_bit(BH_Quiet, &bh->b_state) && printk_ratelimit())
		return 0;
	return 1;
}


L
Linus Torvalds 已提交
140 141 142 143 144 145 146 147 148
static void buffer_io_error(struct buffer_head *bh)
{
	char b[BDEVNAME_SIZE];
	printk(KERN_ERR "Buffer I/O error on device %s, logical block %Lu\n",
			bdevname(bh->b_bdev, b),
			(unsigned long long)bh->b_blocknr);
}

/*
149 150 151 152 153 154
 * End-of-IO handler helper function which does not touch the bh after
 * unlocking it.
 * Note: unlock_buffer() sort-of does touch the bh after unlocking it, but
 * a race there is benign: unlock_buffer() only use the bh's address for
 * hashing after unlocking the buffer, so it doesn't actually touch the bh
 * itself.
L
Linus Torvalds 已提交
155
 */
156
static void __end_buffer_read_notouch(struct buffer_head *bh, int uptodate)
L
Linus Torvalds 已提交
157 158 159 160 161 162 163 164
{
	if (uptodate) {
		set_buffer_uptodate(bh);
	} else {
		/* This happens, due to failed READA attempts. */
		clear_buffer_uptodate(bh);
	}
	unlock_buffer(bh);
165 166 167 168 169 170 171 172 173
}

/*
 * Default synchronous end-of-IO handler..  Just mark it up-to-date and
 * unlock the buffer. This is what ll_rw_block uses too.
 */
void end_buffer_read_sync(struct buffer_head *bh, int uptodate)
{
	__end_buffer_read_notouch(bh, uptodate);
L
Linus Torvalds 已提交
174 175
	put_bh(bh);
}
176
EXPORT_SYMBOL(end_buffer_read_sync);
L
Linus Torvalds 已提交
177 178 179 180 181 182 183 184

void end_buffer_write_sync(struct buffer_head *bh, int uptodate)
{
	char b[BDEVNAME_SIZE];

	if (uptodate) {
		set_buffer_uptodate(bh);
	} else {
185
		if (!quiet_error(bh)) {
L
Linus Torvalds 已提交
186 187 188 189 190 191 192 193 194 195 196
			buffer_io_error(bh);
			printk(KERN_WARNING "lost page write due to "
					"I/O error on %s\n",
				       bdevname(bh->b_bdev, b));
		}
		set_buffer_write_io_error(bh);
		clear_buffer_uptodate(bh);
	}
	unlock_buffer(bh);
	put_bh(bh);
}
197
EXPORT_SYMBOL(end_buffer_write_sync);
L
Linus Torvalds 已提交
198 199 200 201 202 203 204 205 206 207 208 209 210

/*
 * Various filesystems appear to want __find_get_block to be non-blocking.
 * But it's the page lock which protects the buffers.  To get around this,
 * we get exclusion from try_to_free_buffers with the blockdev mapping's
 * private_lock.
 *
 * Hack idea: for the blockdev mapping, i_bufferlist_lock contention
 * may be quite high.  This code could TryLock the page, and if that
 * succeeds, there is no need to take private_lock. (But if
 * private_lock is contended then so is mapping->tree_lock).
 */
static struct buffer_head *
211
__find_get_block_slow(struct block_device *bdev, sector_t block)
L
Linus Torvalds 已提交
212 213 214 215 216 217 218 219 220 221 222
{
	struct inode *bd_inode = bdev->bd_inode;
	struct address_space *bd_mapping = bd_inode->i_mapping;
	struct buffer_head *ret = NULL;
	pgoff_t index;
	struct buffer_head *bh;
	struct buffer_head *head;
	struct page *page;
	int all_mapped = 1;

	index = block >> (PAGE_CACHE_SHIFT - bd_inode->i_blkbits);
223
	page = find_get_page_flags(bd_mapping, index, FGP_ACCESSED);
L
Linus Torvalds 已提交
224 225 226 227 228 229 230 231 232
	if (!page)
		goto out;

	spin_lock(&bd_mapping->private_lock);
	if (!page_has_buffers(page))
		goto out_unlock;
	head = page_buffers(page);
	bh = head;
	do {
233 234 235
		if (!buffer_mapped(bh))
			all_mapped = 0;
		else if (bh->b_blocknr == block) {
L
Linus Torvalds 已提交
236 237 238 239 240 241 242 243 244 245 246 247 248
			ret = bh;
			get_bh(bh);
			goto out_unlock;
		}
		bh = bh->b_this_page;
	} while (bh != head);

	/* we might be here because some of the buffers on this page are
	 * not mapped.  This is due to various races between
	 * file io on the block device and getblk.  It gets dealt with
	 * elsewhere, don't buffer_error if we had some unmapped buffers
	 */
	if (all_mapped) {
249 250
		char b[BDEVNAME_SIZE];

L
Linus Torvalds 已提交
251 252
		printk("__find_get_block_slow() failed. "
			"block=%llu, b_blocknr=%llu\n",
253 254 255 256
			(unsigned long long)block,
			(unsigned long long)bh->b_blocknr);
		printk("b_state=0x%08lx, b_size=%zu\n",
			bh->b_state, bh->b_size);
257 258
		printk("device %s blocksize: %d\n", bdevname(bdev, b),
			1 << bd_inode->i_blkbits);
L
Linus Torvalds 已提交
259 260 261 262 263 264 265 266 267
	}
out_unlock:
	spin_unlock(&bd_mapping->private_lock);
	page_cache_release(page);
out:
	return ret;
}

/*
268
 * Kick the writeback threads then try to free up some ZONE_NORMAL memory.
L
Linus Torvalds 已提交
269 270 271
 */
static void free_more_memory(void)
{
272
	struct zone *zone;
273
	int nid;
L
Linus Torvalds 已提交
274

275
	wakeup_flusher_threads(1024, WB_REASON_FREE_MORE_MEM);
L
Linus Torvalds 已提交
276 277
	yield();

278
	for_each_online_node(nid) {
279 280 281 282
		(void)first_zones_zonelist(node_zonelist(nid, GFP_NOFS),
						gfp_zone(GFP_NOFS), NULL,
						&zone);
		if (zone)
283
			try_to_free_pages(node_zonelist(nid, GFP_NOFS), 0,
284
						GFP_NOFS, NULL);
L
Linus Torvalds 已提交
285 286 287 288 289 290 291 292 293 294
	}
}

/*
 * I/O completion handler for block_read_full_page() - pages
 * which come unlocked at the end of I/O.
 */
static void end_buffer_async_read(struct buffer_head *bh, int uptodate)
{
	unsigned long flags;
295
	struct buffer_head *first;
L
Linus Torvalds 已提交
296 297 298 299 300 301 302 303 304 305 306
	struct buffer_head *tmp;
	struct page *page;
	int page_uptodate = 1;

	BUG_ON(!buffer_async_read(bh));

	page = bh->b_page;
	if (uptodate) {
		set_buffer_uptodate(bh);
	} else {
		clear_buffer_uptodate(bh);
307
		if (!quiet_error(bh))
L
Linus Torvalds 已提交
308 309 310 311 312 313 314 315 316
			buffer_io_error(bh);
		SetPageError(page);
	}

	/*
	 * Be _very_ careful from here on. Bad things can happen if
	 * two buffer heads end IO at almost the same time and both
	 * decide that the page is now completely done.
	 */
317 318 319
	first = page_buffers(page);
	local_irq_save(flags);
	bit_spin_lock(BH_Uptodate_Lock, &first->b_state);
L
Linus Torvalds 已提交
320 321 322 323 324 325 326 327 328 329 330 331
	clear_buffer_async_read(bh);
	unlock_buffer(bh);
	tmp = bh;
	do {
		if (!buffer_uptodate(tmp))
			page_uptodate = 0;
		if (buffer_async_read(tmp)) {
			BUG_ON(!buffer_locked(tmp));
			goto still_busy;
		}
		tmp = tmp->b_this_page;
	} while (tmp != bh);
332 333
	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
	local_irq_restore(flags);
L
Linus Torvalds 已提交
334 335 336 337 338 339 340 341 342 343 344

	/*
	 * If none of the buffers had errors and they are all
	 * uptodate then we can set the page uptodate.
	 */
	if (page_uptodate && !PageError(page))
		SetPageUptodate(page);
	unlock_page(page);
	return;

still_busy:
345 346
	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
	local_irq_restore(flags);
L
Linus Torvalds 已提交
347 348 349 350 351 352 353
	return;
}

/*
 * Completion handler for block_write_full_page() - pages which are unlocked
 * during I/O, and which have PageWriteback cleared upon I/O completion.
 */
354
void end_buffer_async_write(struct buffer_head *bh, int uptodate)
L
Linus Torvalds 已提交
355 356 357
{
	char b[BDEVNAME_SIZE];
	unsigned long flags;
358
	struct buffer_head *first;
L
Linus Torvalds 已提交
359 360 361 362 363 364 365 366 367
	struct buffer_head *tmp;
	struct page *page;

	BUG_ON(!buffer_async_write(bh));

	page = bh->b_page;
	if (uptodate) {
		set_buffer_uptodate(bh);
	} else {
368
		if (!quiet_error(bh)) {
L
Linus Torvalds 已提交
369 370 371 372 373 374
			buffer_io_error(bh);
			printk(KERN_WARNING "lost page write due to "
					"I/O error on %s\n",
			       bdevname(bh->b_bdev, b));
		}
		set_bit(AS_EIO, &page->mapping->flags);
375
		set_buffer_write_io_error(bh);
L
Linus Torvalds 已提交
376 377 378 379
		clear_buffer_uptodate(bh);
		SetPageError(page);
	}

380 381 382 383
	first = page_buffers(page);
	local_irq_save(flags);
	bit_spin_lock(BH_Uptodate_Lock, &first->b_state);

L
Linus Torvalds 已提交
384 385 386 387 388 389 390 391 392 393
	clear_buffer_async_write(bh);
	unlock_buffer(bh);
	tmp = bh->b_this_page;
	while (tmp != bh) {
		if (buffer_async_write(tmp)) {
			BUG_ON(!buffer_locked(tmp));
			goto still_busy;
		}
		tmp = tmp->b_this_page;
	}
394 395
	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
	local_irq_restore(flags);
L
Linus Torvalds 已提交
396 397 398 399
	end_page_writeback(page);
	return;

still_busy:
400 401
	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
	local_irq_restore(flags);
L
Linus Torvalds 已提交
402 403
	return;
}
404
EXPORT_SYMBOL(end_buffer_async_write);
L
Linus Torvalds 已提交
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

/*
 * If a page's buffers are under async readin (end_buffer_async_read
 * completion) then there is a possibility that another thread of
 * control could lock one of the buffers after it has completed
 * but while some of the other buffers have not completed.  This
 * locked buffer would confuse end_buffer_async_read() into not unlocking
 * the page.  So the absence of BH_Async_Read tells end_buffer_async_read()
 * that this buffer is not under async I/O.
 *
 * The page comes unlocked when it has no locked buffer_async buffers
 * left.
 *
 * PageLocked prevents anyone starting new async I/O reads any of
 * the buffers.
 *
 * PageWriteback is used to prevent simultaneous writeout of the same
 * page.
 *
 * PageLocked prevents anyone from starting writeback of a page which is
 * under read I/O (PageWriteback is only ever set against a locked page).
 */
static void mark_buffer_async_read(struct buffer_head *bh)
{
	bh->b_end_io = end_buffer_async_read;
	set_buffer_async_read(bh);
}

433 434
static void mark_buffer_async_write_endio(struct buffer_head *bh,
					  bh_end_io_t *handler)
L
Linus Torvalds 已提交
435
{
436
	bh->b_end_io = handler;
L
Linus Torvalds 已提交
437 438
	set_buffer_async_write(bh);
}
439 440 441 442 443

void mark_buffer_async_write(struct buffer_head *bh)
{
	mark_buffer_async_write_endio(bh, end_buffer_async_write);
}
L
Linus Torvalds 已提交
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
EXPORT_SYMBOL(mark_buffer_async_write);


/*
 * fs/buffer.c contains helper functions for buffer-backed address space's
 * fsync functions.  A common requirement for buffer-based filesystems is
 * that certain data from the backing blockdev needs to be written out for
 * a successful fsync().  For example, ext2 indirect blocks need to be
 * written back and waited upon before fsync() returns.
 *
 * The functions mark_buffer_inode_dirty(), fsync_inode_buffers(),
 * inode_has_buffers() and invalidate_inode_buffers() are provided for the
 * management of a list of dependent buffers at ->i_mapping->private_list.
 *
 * Locking is a little subtle: try_to_free_buffers() will remove buffers
 * from their controlling inode's queue when they are being freed.  But
 * try_to_free_buffers() will be operating against the *blockdev* mapping
 * at the time, not against the S_ISREG file which depends on those buffers.
 * So the locking for private_list is via the private_lock in the address_space
 * which backs the buffers.  Which is different from the address_space 
 * against which the buffers are listed.  So for a particular address_space,
 * mapping->private_lock does *not* protect mapping->private_list!  In fact,
 * mapping->private_list will always be protected by the backing blockdev's
 * ->private_lock.
 *
 * Which introduces a requirement: all buffers on an address_space's
 * ->private_list must be from the same address_space: the blockdev's.
 *
 * address_spaces which do not place buffers at ->private_list via these
 * utility functions are free to use private_lock and private_list for
 * whatever they want.  The only requirement is that list_empty(private_list)
 * be true at clear_inode() time.
 *
 * FIXME: clear_inode should not call invalidate_inode_buffers().  The
 * filesystems should do that.  invalidate_inode_buffers() should just go
 * BUG_ON(!list_empty).
 *
 * FIXME: mark_buffer_dirty_inode() is a data-plane operation.  It should
 * take an address_space, not an inode.  And it should be called
 * mark_buffer_dirty_fsync() to clearly define why those buffers are being
 * queued up.
 *
 * FIXME: mark_buffer_dirty_inode() doesn't need to add the buffer to the
 * list if it is already on a list.  Because if the buffer is on a list,
 * it *must* already be on the right one.  If not, the filesystem is being
 * silly.  This will save a ton of locking.  But first we have to ensure
 * that buffers are taken *off* the old inode's list when they are freed
 * (presumably in truncate).  That requires careful auditing of all
 * filesystems (do it inside bforget()).  It could also be done by bringing
 * b_inode back.
 */

/*
 * The buffer's backing address_space's private_lock must be held
 */
499
static void __remove_assoc_queue(struct buffer_head *bh)
L
Linus Torvalds 已提交
500 501
{
	list_del_init(&bh->b_assoc_buffers);
502 503 504 505
	WARN_ON(!bh->b_assoc_map);
	if (buffer_write_io_error(bh))
		set_bit(AS_EIO, &bh->b_assoc_map->flags);
	bh->b_assoc_map = NULL;
L
Linus Torvalds 已提交
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547
}

int inode_has_buffers(struct inode *inode)
{
	return !list_empty(&inode->i_data.private_list);
}

/*
 * osync is designed to support O_SYNC io.  It waits synchronously for
 * all already-submitted IO to complete, but does not queue any new
 * writes to the disk.
 *
 * To do O_SYNC writes, just queue the buffer writes with ll_rw_block as
 * you dirty the buffers, and then use osync_inode_buffers to wait for
 * completion.  Any other dirty buffers which are not yet queued for
 * write will not be flushed to disk by the osync.
 */
static int osync_buffers_list(spinlock_t *lock, struct list_head *list)
{
	struct buffer_head *bh;
	struct list_head *p;
	int err = 0;

	spin_lock(lock);
repeat:
	list_for_each_prev(p, list) {
		bh = BH_ENTRY(p);
		if (buffer_locked(bh)) {
			get_bh(bh);
			spin_unlock(lock);
			wait_on_buffer(bh);
			if (!buffer_uptodate(bh))
				err = -EIO;
			brelse(bh);
			spin_lock(lock);
			goto repeat;
		}
	}
	spin_unlock(lock);
	return err;
}

A
Al Viro 已提交
548
static void do_thaw_one(struct super_block *sb, void *unused)
549 550
{
	char b[BDEVNAME_SIZE];
A
Al Viro 已提交
551 552 553 554
	while (sb->s_bdev && !thaw_bdev(sb->s_bdev, sb))
		printk(KERN_WARNING "Emergency Thaw on %s\n",
		       bdevname(sb->s_bdev, b));
}
555

A
Al Viro 已提交
556 557 558
static void do_thaw_all(struct work_struct *work)
{
	iterate_supers(do_thaw_one, NULL);
559
	kfree(work);
560 561 562 563 564 565 566 567 568 569
	printk(KERN_WARNING "Emergency Thaw complete\n");
}

/**
 * emergency_thaw_all -- forcibly thaw every frozen filesystem
 *
 * Used for emergency unfreeze of all filesystems via SysRq
 */
void emergency_thaw_all(void)
{
570 571 572 573 574 575 576
	struct work_struct *work;

	work = kmalloc(sizeof(*work), GFP_ATOMIC);
	if (work) {
		INIT_WORK(work, do_thaw_all);
		schedule_work(work);
	}
577 578
}

L
Linus Torvalds 已提交
579
/**
580
 * sync_mapping_buffers - write out & wait upon a mapping's "associated" buffers
581
 * @mapping: the mapping which wants those buffers written
L
Linus Torvalds 已提交
582 583 584 585
 *
 * Starts I/O against the buffers at mapping->private_list, and waits upon
 * that I/O.
 *
586 587 588
 * Basically, this is a convenience function for fsync().
 * @mapping is a file or directory which needs those buffers to be written for
 * a successful fsync().
L
Linus Torvalds 已提交
589 590 591
 */
int sync_mapping_buffers(struct address_space *mapping)
{
592
	struct address_space *buffer_mapping = mapping->private_data;
L
Linus Torvalds 已提交
593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624

	if (buffer_mapping == NULL || list_empty(&mapping->private_list))
		return 0;

	return fsync_buffers_list(&buffer_mapping->private_lock,
					&mapping->private_list);
}
EXPORT_SYMBOL(sync_mapping_buffers);

/*
 * Called when we've recently written block `bblock', and it is known that
 * `bblock' was for a buffer_boundary() buffer.  This means that the block at
 * `bblock + 1' is probably a dirty indirect block.  Hunt it down and, if it's
 * dirty, schedule it for IO.  So that indirects merge nicely with their data.
 */
void write_boundary_block(struct block_device *bdev,
			sector_t bblock, unsigned blocksize)
{
	struct buffer_head *bh = __find_get_block(bdev, bblock + 1, blocksize);
	if (bh) {
		if (buffer_dirty(bh))
			ll_rw_block(WRITE, 1, &bh);
		put_bh(bh);
	}
}

void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode)
{
	struct address_space *mapping = inode->i_mapping;
	struct address_space *buffer_mapping = bh->b_page->mapping;

	mark_buffer_dirty(bh);
625 626
	if (!mapping->private_data) {
		mapping->private_data = buffer_mapping;
L
Linus Torvalds 已提交
627
	} else {
628
		BUG_ON(mapping->private_data != buffer_mapping);
L
Linus Torvalds 已提交
629
	}
630
	if (!bh->b_assoc_map) {
L
Linus Torvalds 已提交
631 632 633
		spin_lock(&buffer_mapping->private_lock);
		list_move_tail(&bh->b_assoc_buffers,
				&mapping->private_list);
634
		bh->b_assoc_map = mapping;
L
Linus Torvalds 已提交
635 636 637 638 639
		spin_unlock(&buffer_mapping->private_lock);
	}
}
EXPORT_SYMBOL(mark_buffer_dirty_inode);

640 641 642 643 644 645 646
/*
 * Mark the page dirty, and set it dirty in the radix tree, and mark the inode
 * dirty.
 *
 * If warn is true, then emit a warning if the page is not uptodate and has
 * not been truncated.
 */
647
static void __set_page_dirty(struct page *page,
648 649
		struct address_space *mapping, int warn)
{
650 651 652
	unsigned long flags;

	spin_lock_irqsave(&mapping->tree_lock, flags);
653 654
	if (page->mapping) {	/* Race with truncate? */
		WARN_ON_ONCE(warn && !PageUptodate(page));
655
		account_page_dirtied(page, mapping);
656 657 658
		radix_tree_tag_set(&mapping->page_tree,
				page_index(page), PAGECACHE_TAG_DIRTY);
	}
659
	spin_unlock_irqrestore(&mapping->tree_lock, flags);
660 661 662
	__mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
}

L
Linus Torvalds 已提交
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
/*
 * Add a page to the dirty page list.
 *
 * It is a sad fact of life that this function is called from several places
 * deeply under spinlocking.  It may not sleep.
 *
 * If the page has buffers, the uptodate buffers are set dirty, to preserve
 * dirty-state coherency between the page and the buffers.  It the page does
 * not have buffers then when they are later attached they will all be set
 * dirty.
 *
 * The buffers are dirtied before the page is dirtied.  There's a small race
 * window in which a writepage caller may see the page cleanness but not the
 * buffer dirtiness.  That's fine.  If this code were to set the page dirty
 * before the buffers, a concurrent writepage caller could clear the page dirty
 * bit, see a bunch of clean buffers and we'd end up with dirty buffers/clean
 * page on the dirty page list.
 *
 * We use private_lock to lock against try_to_free_buffers while using the
 * page's buffer list.  Also use this to protect against clean buffers being
 * added to the page after it was set dirty.
 *
 * FIXME: may need to call ->reservepage here as well.  That's rather up to the
 * address_space though.
 */
int __set_page_dirty_buffers(struct page *page)
{
690
	int newly_dirty;
691
	struct address_space *mapping = page_mapping(page);
692 693 694

	if (unlikely(!mapping))
		return !TestSetPageDirty(page);
L
Linus Torvalds 已提交
695 696 697 698 699 700 701 702 703 704 705

	spin_lock(&mapping->private_lock);
	if (page_has_buffers(page)) {
		struct buffer_head *head = page_buffers(page);
		struct buffer_head *bh = head;

		do {
			set_buffer_dirty(bh);
			bh = bh->b_this_page;
		} while (bh != head);
	}
706
	newly_dirty = !TestSetPageDirty(page);
L
Linus Torvalds 已提交
707 708
	spin_unlock(&mapping->private_lock);

709 710 711
	if (newly_dirty)
		__set_page_dirty(page, mapping, 1);
	return newly_dirty;
L
Linus Torvalds 已提交
712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737
}
EXPORT_SYMBOL(__set_page_dirty_buffers);

/*
 * Write out and wait upon a list of buffers.
 *
 * We have conflicting pressures: we want to make sure that all
 * initially dirty buffers get waited on, but that any subsequently
 * dirtied buffers don't.  After all, we don't want fsync to last
 * forever if somebody is actively writing to the file.
 *
 * Do this in two main stages: first we copy dirty buffers to a
 * temporary inode list, queueing the writes as we go.  Then we clean
 * up, waiting for those writes to complete.
 * 
 * During this second stage, any subsequent updates to the file may end
 * up refiling the buffer on the original inode's dirty list again, so
 * there is a chance we will end up with a buffer queued for write but
 * not yet completed on that list.  So, as a final cleanup we go through
 * the osync code to catch these locked, dirty buffers without requeuing
 * any newly dirty buffers for write.
 */
static int fsync_buffers_list(spinlock_t *lock, struct list_head *list)
{
	struct buffer_head *bh;
	struct list_head tmp;
J
Jens Axboe 已提交
738
	struct address_space *mapping;
L
Linus Torvalds 已提交
739
	int err = 0, err2;
J
Jens Axboe 已提交
740
	struct blk_plug plug;
L
Linus Torvalds 已提交
741 742

	INIT_LIST_HEAD(&tmp);
J
Jens Axboe 已提交
743
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
744 745 746 747

	spin_lock(lock);
	while (!list_empty(list)) {
		bh = BH_ENTRY(list->next);
748
		mapping = bh->b_assoc_map;
749
		__remove_assoc_queue(bh);
750 751 752
		/* Avoid race with mark_buffer_dirty_inode() which does
		 * a lockless check and we rely on seeing the dirty bit */
		smp_mb();
L
Linus Torvalds 已提交
753 754
		if (buffer_dirty(bh) || buffer_locked(bh)) {
			list_add(&bh->b_assoc_buffers, &tmp);
755
			bh->b_assoc_map = mapping;
L
Linus Torvalds 已提交
756 757 758 759 760
			if (buffer_dirty(bh)) {
				get_bh(bh);
				spin_unlock(lock);
				/*
				 * Ensure any pending I/O completes so that
C
Christoph Hellwig 已提交
761 762 763 764
				 * write_dirty_buffer() actually writes the
				 * current contents - it is a noop if I/O is
				 * still in flight on potentially older
				 * contents.
L
Linus Torvalds 已提交
765
				 */
J
Jens Axboe 已提交
766
				write_dirty_buffer(bh, WRITE_SYNC);
767 768 769 770 771 772 773

				/*
				 * Kick off IO for the previous mapping. Note
				 * that we will not run the very last mapping,
				 * wait_on_buffer() will do that for us
				 * through sync_buffer().
				 */
L
Linus Torvalds 已提交
774 775 776 777 778 779
				brelse(bh);
				spin_lock(lock);
			}
		}
	}

J
Jens Axboe 已提交
780 781 782 783
	spin_unlock(lock);
	blk_finish_plug(&plug);
	spin_lock(lock);

L
Linus Torvalds 已提交
784 785 786
	while (!list_empty(&tmp)) {
		bh = BH_ENTRY(tmp.prev);
		get_bh(bh);
787 788 789 790 791 792 793
		mapping = bh->b_assoc_map;
		__remove_assoc_queue(bh);
		/* Avoid race with mark_buffer_dirty_inode() which does
		 * a lockless check and we rely on seeing the dirty bit */
		smp_mb();
		if (buffer_dirty(bh)) {
			list_add(&bh->b_assoc_buffers,
794
				 &mapping->private_list);
795 796
			bh->b_assoc_map = mapping;
		}
L
Linus Torvalds 已提交
797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826
		spin_unlock(lock);
		wait_on_buffer(bh);
		if (!buffer_uptodate(bh))
			err = -EIO;
		brelse(bh);
		spin_lock(lock);
	}
	
	spin_unlock(lock);
	err2 = osync_buffers_list(lock, list);
	if (err)
		return err;
	else
		return err2;
}

/*
 * Invalidate any and all dirty buffers on a given inode.  We are
 * probably unmounting the fs, but that doesn't mean we have already
 * done a sync().  Just drop the buffers from the inode list.
 *
 * NOTE: we take the inode's blockdev's mapping's private_lock.  Which
 * assumes that all the buffers are against the blockdev.  Not true
 * for reiserfs.
 */
void invalidate_inode_buffers(struct inode *inode)
{
	if (inode_has_buffers(inode)) {
		struct address_space *mapping = &inode->i_data;
		struct list_head *list = &mapping->private_list;
827
		struct address_space *buffer_mapping = mapping->private_data;
L
Linus Torvalds 已提交
828 829 830 831 832 833 834

		spin_lock(&buffer_mapping->private_lock);
		while (!list_empty(list))
			__remove_assoc_queue(BH_ENTRY(list->next));
		spin_unlock(&buffer_mapping->private_lock);
	}
}
J
Jan Kara 已提交
835
EXPORT_SYMBOL(invalidate_inode_buffers);
L
Linus Torvalds 已提交
836 837 838 839 840 841 842 843 844 845 846 847 848 849

/*
 * Remove any clean buffers from the inode's buffer list.  This is called
 * when we're trying to free the inode itself.  Those buffers can pin it.
 *
 * Returns true if all buffers were removed.
 */
int remove_inode_buffers(struct inode *inode)
{
	int ret = 1;

	if (inode_has_buffers(inode)) {
		struct address_space *mapping = &inode->i_data;
		struct list_head *list = &mapping->private_list;
850
		struct address_space *buffer_mapping = mapping->private_data;
L
Linus Torvalds 已提交
851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944

		spin_lock(&buffer_mapping->private_lock);
		while (!list_empty(list)) {
			struct buffer_head *bh = BH_ENTRY(list->next);
			if (buffer_dirty(bh)) {
				ret = 0;
				break;
			}
			__remove_assoc_queue(bh);
		}
		spin_unlock(&buffer_mapping->private_lock);
	}
	return ret;
}

/*
 * Create the appropriate buffers when given a page for data area and
 * the size of each buffer.. Use the bh->b_this_page linked list to
 * follow the buffers created.  Return NULL if unable to create more
 * buffers.
 *
 * The retry flag is used to differentiate async IO (paging, swapping)
 * which may not fail from ordinary buffer allocations.
 */
struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size,
		int retry)
{
	struct buffer_head *bh, *head;
	long offset;

try_again:
	head = NULL;
	offset = PAGE_SIZE;
	while ((offset -= size) >= 0) {
		bh = alloc_buffer_head(GFP_NOFS);
		if (!bh)
			goto no_grow;

		bh->b_this_page = head;
		bh->b_blocknr = -1;
		head = bh;

		bh->b_size = size;

		/* Link the buffer to its page */
		set_bh_page(bh, page, offset);
	}
	return head;
/*
 * In case anything failed, we just free everything we got.
 */
no_grow:
	if (head) {
		do {
			bh = head;
			head = head->b_this_page;
			free_buffer_head(bh);
		} while (head);
	}

	/*
	 * Return failure for non-async IO requests.  Async IO requests
	 * are not allowed to fail, so we have to wait until buffer heads
	 * become available.  But we don't want tasks sleeping with 
	 * partially complete buffers, so all were released above.
	 */
	if (!retry)
		return NULL;

	/* We're _really_ low on memory. Now we just
	 * wait for old buffer heads to become free due to
	 * finishing IO.  Since this is an async request and
	 * the reserve list is empty, we're sure there are 
	 * async buffer heads in use.
	 */
	free_more_memory();
	goto try_again;
}
EXPORT_SYMBOL_GPL(alloc_page_buffers);

static inline void
link_dev_buffers(struct page *page, struct buffer_head *head)
{
	struct buffer_head *bh, *tail;

	bh = head;
	do {
		tail = bh;
		bh = bh->b_this_page;
	} while (bh);
	tail->b_this_page = head;
	attach_page_buffers(page, head);
}

945 946 947 948 949 950 951 952 953 954 955 956
static sector_t blkdev_max_block(struct block_device *bdev, unsigned int size)
{
	sector_t retval = ~((sector_t)0);
	loff_t sz = i_size_read(bdev->bd_inode);

	if (sz) {
		unsigned int sizebits = blksize_bits(size);
		retval = (sz >> sizebits);
	}
	return retval;
}

L
Linus Torvalds 已提交
957 958 959
/*
 * Initialise the state of a blockdev page's buffers.
 */ 
960
static sector_t
L
Linus Torvalds 已提交
961 962 963 964 965 966
init_page_buffers(struct page *page, struct block_device *bdev,
			sector_t block, int size)
{
	struct buffer_head *head = page_buffers(page);
	struct buffer_head *bh = head;
	int uptodate = PageUptodate(page);
967
	sector_t end_block = blkdev_max_block(I_BDEV(bdev->bd_inode), size);
L
Linus Torvalds 已提交
968 969 970 971 972 973 974 975

	do {
		if (!buffer_mapped(bh)) {
			init_buffer(bh, NULL, NULL);
			bh->b_bdev = bdev;
			bh->b_blocknr = block;
			if (uptodate)
				set_buffer_uptodate(bh);
976 977
			if (block < end_block)
				set_buffer_mapped(bh);
L
Linus Torvalds 已提交
978 979 980 981
		}
		block++;
		bh = bh->b_this_page;
	} while (bh != head);
982 983 984 985 986

	/*
	 * Caller needs to validate requested block against end of device.
	 */
	return end_block;
L
Linus Torvalds 已提交
987 988 989 990 991
}

/*
 * Create the page-cache page that contains the requested block.
 *
992
 * This is used purely for blockdev mappings.
L
Linus Torvalds 已提交
993
 */
994
static int
L
Linus Torvalds 已提交
995
grow_dev_page(struct block_device *bdev, sector_t block,
996
		pgoff_t index, int size, int sizebits)
L
Linus Torvalds 已提交
997 998 999 1000
{
	struct inode *inode = bdev->bd_inode;
	struct page *page;
	struct buffer_head *bh;
1001 1002
	sector_t end_block;
	int ret = 0;		/* Will call free_more_memory() */
1003
	gfp_t gfp_mask;
L
Linus Torvalds 已提交
1004

1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
	gfp_mask = mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS;
	gfp_mask |= __GFP_MOVABLE;
	/*
	 * XXX: __getblk_slow() can not really deal with failure and
	 * will endlessly loop on improvised global reclaim.  Prefer
	 * looping in the allocator rather than here, at least that
	 * code knows what it's doing.
	 */
	gfp_mask |= __GFP_NOFAIL;

	page = find_or_create_page(inode->i_mapping, index, gfp_mask);
L
Linus Torvalds 已提交
1016
	if (!page)
1017
		return ret;
L
Linus Torvalds 已提交
1018

1019
	BUG_ON(!PageLocked(page));
L
Linus Torvalds 已提交
1020 1021 1022 1023

	if (page_has_buffers(page)) {
		bh = page_buffers(page);
		if (bh->b_size == size) {
1024
			end_block = init_page_buffers(page, bdev,
1025 1026
						(sector_t)index << sizebits,
						size);
1027
			goto done;
L
Linus Torvalds 已提交
1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
		}
		if (!try_to_free_buffers(page))
			goto failed;
	}

	/*
	 * Allocate some buffers for this page
	 */
	bh = alloc_page_buffers(page, size, 0);
	if (!bh)
		goto failed;

	/*
	 * Link the page to the buffers and initialise them.  Take the
	 * lock to be atomic wrt __find_get_block(), which does not
	 * run under the page lock.
	 */
	spin_lock(&inode->i_mapping->private_lock);
	link_dev_buffers(page, bh);
1047 1048
	end_block = init_page_buffers(page, bdev, (sector_t)index << sizebits,
			size);
L
Linus Torvalds 已提交
1049
	spin_unlock(&inode->i_mapping->private_lock);
1050 1051
done:
	ret = (block < end_block) ? 1 : -ENXIO;
L
Linus Torvalds 已提交
1052 1053 1054
failed:
	unlock_page(page);
	page_cache_release(page);
1055
	return ret;
L
Linus Torvalds 已提交
1056 1057 1058 1059 1060 1061
}

/*
 * Create buffers for the specified block device block's page.  If
 * that page was dirty, the buffers are set dirty also.
 */
1062
static int
L
Linus Torvalds 已提交
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074
grow_buffers(struct block_device *bdev, sector_t block, int size)
{
	pgoff_t index;
	int sizebits;

	sizebits = -1;
	do {
		sizebits++;
	} while ((size << sizebits) < PAGE_SIZE);

	index = block >> sizebits;

1075 1076 1077 1078 1079 1080 1081 1082 1083
	/*
	 * Check for a block which wants to lie outside our maximum possible
	 * pagecache index.  (this comparison is done using sector_t types).
	 */
	if (unlikely(index != block >> sizebits)) {
		char b[BDEVNAME_SIZE];

		printk(KERN_ERR "%s: requested out-of-range block %llu for "
			"device %s\n",
1084
			__func__, (unsigned long long)block,
1085 1086 1087
			bdevname(bdev, b));
		return -EIO;
	}
1088

L
Linus Torvalds 已提交
1089
	/* Create a page with the proper size buffers.. */
1090
	return grow_dev_page(bdev, block, index, size, sizebits);
L
Linus Torvalds 已提交
1091 1092
}

A
Adrian Bunk 已提交
1093
static struct buffer_head *
L
Linus Torvalds 已提交
1094 1095 1096
__getblk_slow(struct block_device *bdev, sector_t block, int size)
{
	/* Size must be multiple of hard sectorsize */
1097
	if (unlikely(size & (bdev_logical_block_size(bdev)-1) ||
L
Linus Torvalds 已提交
1098 1099 1100
			(size < 512 || size > PAGE_SIZE))) {
		printk(KERN_ERR "getblk(): invalid block size %d requested\n",
					size);
1101 1102
		printk(KERN_ERR "logical block size: %d\n",
					bdev_logical_block_size(bdev));
L
Linus Torvalds 已提交
1103 1104 1105 1106 1107

		dump_stack();
		return NULL;
	}

1108 1109 1110
	for (;;) {
		struct buffer_head *bh;
		int ret;
L
Linus Torvalds 已提交
1111 1112 1113 1114

		bh = __find_get_block(bdev, block, size);
		if (bh)
			return bh;
1115 1116 1117 1118 1119 1120

		ret = grow_buffers(bdev, block, size);
		if (ret < 0)
			return NULL;
		if (ret == 0)
			free_more_memory();
L
Linus Torvalds 已提交
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
	}
}

/*
 * The relationship between dirty buffers and dirty pages:
 *
 * Whenever a page has any dirty buffers, the page's dirty bit is set, and
 * the page is tagged dirty in its radix tree.
 *
 * At all times, the dirtiness of the buffers represents the dirtiness of
 * subsections of the page.  If the page has buffers, the page dirty bit is
 * merely a hint about the true dirty state.
 *
 * When a page is set dirty in its entirety, all its buffers are marked dirty
 * (if the page has buffers).
 *
 * When a buffer is marked dirty, its page is dirtied, but the page's other
 * buffers are not.
 *
 * Also.  When blockdev buffers are explicitly read with bread(), they
 * individually become uptodate.  But their backing page remains not
 * uptodate - even if all of its buffers are uptodate.  A subsequent
 * block_read_full_page() against that page will discover all the uptodate
 * buffers, will set the page uptodate and will perform no I/O.
 */

/**
 * mark_buffer_dirty - mark a buffer_head as needing writeout
1149
 * @bh: the buffer_head to mark dirty
L
Linus Torvalds 已提交
1150 1151 1152 1153 1154 1155 1156
 *
 * mark_buffer_dirty() will set the dirty bit against the buffer, then set its
 * backing page dirty, then tag the page as dirty in its address_space's radix
 * tree and then attach the address_space's inode to its superblock's dirty
 * inode list.
 *
 * mark_buffer_dirty() is atomic.  It takes bh->b_page->mapping->private_lock,
1157
 * mapping->tree_lock and mapping->host->i_lock.
L
Linus Torvalds 已提交
1158
 */
1159
void mark_buffer_dirty(struct buffer_head *bh)
L
Linus Torvalds 已提交
1160
{
1161
	WARN_ON_ONCE(!buffer_uptodate(bh));
1162

1163 1164
	trace_block_dirty_buffer(bh);

1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
	/*
	 * Very *carefully* optimize the it-is-already-dirty case.
	 *
	 * Don't let the final "is it dirty" escape to before we
	 * perhaps modified the buffer.
	 */
	if (buffer_dirty(bh)) {
		smp_mb();
		if (buffer_dirty(bh))
			return;
	}

1177 1178
	if (!test_set_buffer_dirty(bh)) {
		struct page *page = bh->b_page;
1179 1180 1181 1182 1183
		if (!TestSetPageDirty(page)) {
			struct address_space *mapping = page_mapping(page);
			if (mapping)
				__set_page_dirty(page, mapping, 0);
		}
1184
	}
L
Linus Torvalds 已提交
1185
}
1186
EXPORT_SYMBOL(mark_buffer_dirty);
L
Linus Torvalds 已提交
1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200

/*
 * Decrement a buffer_head's reference count.  If all buffers against a page
 * have zero reference count, are clean and unlocked, and if the page is clean
 * and unlocked then try_to_free_buffers() may strip the buffers from the page
 * in preparation for freeing it (sometimes, rarely, buffers are removed from
 * a page but it ends up not being freed, and buffers may later be reattached).
 */
void __brelse(struct buffer_head * buf)
{
	if (atomic_read(&buf->b_count)) {
		put_bh(buf);
		return;
	}
A
Arjan van de Ven 已提交
1201
	WARN(1, KERN_ERR "VFS: brelse: Trying to free free buffer\n");
L
Linus Torvalds 已提交
1202
}
1203
EXPORT_SYMBOL(__brelse);
L
Linus Torvalds 已提交
1204 1205 1206 1207 1208 1209 1210 1211

/*
 * bforget() is like brelse(), except it discards any
 * potentially dirty data.
 */
void __bforget(struct buffer_head *bh)
{
	clear_buffer_dirty(bh);
1212
	if (bh->b_assoc_map) {
L
Linus Torvalds 已提交
1213 1214 1215 1216
		struct address_space *buffer_mapping = bh->b_page->mapping;

		spin_lock(&buffer_mapping->private_lock);
		list_del_init(&bh->b_assoc_buffers);
1217
		bh->b_assoc_map = NULL;
L
Linus Torvalds 已提交
1218 1219 1220 1221
		spin_unlock(&buffer_mapping->private_lock);
	}
	__brelse(bh);
}
1222
EXPORT_SYMBOL(__bforget);
L
Linus Torvalds 已提交
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287

static struct buffer_head *__bread_slow(struct buffer_head *bh)
{
	lock_buffer(bh);
	if (buffer_uptodate(bh)) {
		unlock_buffer(bh);
		return bh;
	} else {
		get_bh(bh);
		bh->b_end_io = end_buffer_read_sync;
		submit_bh(READ, bh);
		wait_on_buffer(bh);
		if (buffer_uptodate(bh))
			return bh;
	}
	brelse(bh);
	return NULL;
}

/*
 * Per-cpu buffer LRU implementation.  To reduce the cost of __find_get_block().
 * The bhs[] array is sorted - newest buffer is at bhs[0].  Buffers have their
 * refcount elevated by one when they're in an LRU.  A buffer can only appear
 * once in a particular CPU's LRU.  A single buffer can be present in multiple
 * CPU's LRUs at the same time.
 *
 * This is a transparent caching front-end to sb_bread(), sb_getblk() and
 * sb_find_get_block().
 *
 * The LRUs themselves only need locking against invalidate_bh_lrus.  We use
 * a local interrupt disable for that.
 */

#define BH_LRU_SIZE	8

struct bh_lru {
	struct buffer_head *bhs[BH_LRU_SIZE];
};

static DEFINE_PER_CPU(struct bh_lru, bh_lrus) = {{ NULL }};

#ifdef CONFIG_SMP
#define bh_lru_lock()	local_irq_disable()
#define bh_lru_unlock()	local_irq_enable()
#else
#define bh_lru_lock()	preempt_disable()
#define bh_lru_unlock()	preempt_enable()
#endif

static inline void check_irqs_on(void)
{
#ifdef irqs_disabled
	BUG_ON(irqs_disabled());
#endif
}

/*
 * The LRU management algorithm is dopey-but-simple.  Sorry.
 */
static void bh_lru_install(struct buffer_head *bh)
{
	struct buffer_head *evictee = NULL;

	check_irqs_on();
	bh_lru_lock();
1288
	if (__this_cpu_read(bh_lrus.bhs[0]) != bh) {
L
Linus Torvalds 已提交
1289 1290 1291 1292 1293 1294 1295
		struct buffer_head *bhs[BH_LRU_SIZE];
		int in;
		int out = 0;

		get_bh(bh);
		bhs[out++] = bh;
		for (in = 0; in < BH_LRU_SIZE; in++) {
1296 1297
			struct buffer_head *bh2 =
				__this_cpu_read(bh_lrus.bhs[in]);
L
Linus Torvalds 已提交
1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311

			if (bh2 == bh) {
				__brelse(bh2);
			} else {
				if (out >= BH_LRU_SIZE) {
					BUG_ON(evictee != NULL);
					evictee = bh2;
				} else {
					bhs[out++] = bh2;
				}
			}
		}
		while (out < BH_LRU_SIZE)
			bhs[out++] = NULL;
1312
		memcpy(this_cpu_ptr(&bh_lrus.bhs), bhs, sizeof(bhs));
L
Linus Torvalds 已提交
1313 1314 1315 1316 1317 1318 1319 1320 1321 1322
	}
	bh_lru_unlock();

	if (evictee)
		__brelse(evictee);
}

/*
 * Look up the bh in this cpu's LRU.  If it's there, move it to the head.
 */
1323
static struct buffer_head *
1324
lookup_bh_lru(struct block_device *bdev, sector_t block, unsigned size)
L
Linus Torvalds 已提交
1325 1326
{
	struct buffer_head *ret = NULL;
1327
	unsigned int i;
L
Linus Torvalds 已提交
1328 1329 1330 1331

	check_irqs_on();
	bh_lru_lock();
	for (i = 0; i < BH_LRU_SIZE; i++) {
1332
		struct buffer_head *bh = __this_cpu_read(bh_lrus.bhs[i]);
L
Linus Torvalds 已提交
1333 1334 1335 1336 1337

		if (bh && bh->b_bdev == bdev &&
				bh->b_blocknr == block && bh->b_size == size) {
			if (i) {
				while (i) {
1338 1339
					__this_cpu_write(bh_lrus.bhs[i],
						__this_cpu_read(bh_lrus.bhs[i - 1]));
L
Linus Torvalds 已提交
1340 1341
					i--;
				}
1342
				__this_cpu_write(bh_lrus.bhs[0], bh);
L
Linus Torvalds 已提交
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358
			}
			get_bh(bh);
			ret = bh;
			break;
		}
	}
	bh_lru_unlock();
	return ret;
}

/*
 * Perform a pagecache lookup for the matching buffer.  If it's there, refresh
 * it in the LRU and mark it as accessed.  If it is not present then return
 * NULL
 */
struct buffer_head *
1359
__find_get_block(struct block_device *bdev, sector_t block, unsigned size)
L
Linus Torvalds 已提交
1360 1361 1362 1363
{
	struct buffer_head *bh = lookup_bh_lru(bdev, block, size);

	if (bh == NULL) {
1364
		/* __find_get_block_slow will mark the page accessed */
1365
		bh = __find_get_block_slow(bdev, block);
L
Linus Torvalds 已提交
1366 1367
		if (bh)
			bh_lru_install(bh);
1368
	} else
L
Linus Torvalds 已提交
1369
		touch_buffer(bh);
1370

L
Linus Torvalds 已提交
1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
	return bh;
}
EXPORT_SYMBOL(__find_get_block);

/*
 * __getblk will locate (and, if necessary, create) the buffer_head
 * which corresponds to the passed block_device, block and size. The
 * returned buffer has its reference count incremented.
 *
 * __getblk() will lock up the machine if grow_dev_page's try_to_free_buffers()
 * attempt is failing.  FIXME, perhaps?
 */
struct buffer_head *
1384
__getblk(struct block_device *bdev, sector_t block, unsigned size)
L
Linus Torvalds 已提交
1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
{
	struct buffer_head *bh = __find_get_block(bdev, block, size);

	might_sleep();
	if (bh == NULL)
		bh = __getblk_slow(bdev, block, size);
	return bh;
}
EXPORT_SYMBOL(__getblk);

/*
 * Do async read-ahead on a buffer..
 */
1398
void __breadahead(struct block_device *bdev, sector_t block, unsigned size)
L
Linus Torvalds 已提交
1399 1400
{
	struct buffer_head *bh = __getblk(bdev, block, size);
A
Andrew Morton 已提交
1401 1402 1403 1404
	if (likely(bh)) {
		ll_rw_block(READA, 1, &bh);
		brelse(bh);
	}
L
Linus Torvalds 已提交
1405 1406 1407 1408 1409
}
EXPORT_SYMBOL(__breadahead);

/**
 *  __bread() - reads a specified block and returns the bh
1410
 *  @bdev: the block_device to read from
L
Linus Torvalds 已提交
1411 1412 1413 1414 1415 1416 1417
 *  @block: number of block
 *  @size: size (in bytes) to read
 * 
 *  Reads a specified block, and returns buffer head that contains it.
 *  It returns NULL if the block was unreadable.
 */
struct buffer_head *
1418
__bread(struct block_device *bdev, sector_t block, unsigned size)
L
Linus Torvalds 已提交
1419 1420 1421
{
	struct buffer_head *bh = __getblk(bdev, block, size);

A
Andrew Morton 已提交
1422
	if (likely(bh) && !buffer_uptodate(bh))
L
Linus Torvalds 已提交
1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443
		bh = __bread_slow(bh);
	return bh;
}
EXPORT_SYMBOL(__bread);

/*
 * invalidate_bh_lrus() is called rarely - but not only at unmount.
 * This doesn't race because it runs in each cpu either in irq
 * or with preempt disabled.
 */
static void invalidate_bh_lru(void *arg)
{
	struct bh_lru *b = &get_cpu_var(bh_lrus);
	int i;

	for (i = 0; i < BH_LRU_SIZE; i++) {
		brelse(b->bhs[i]);
		b->bhs[i] = NULL;
	}
	put_cpu_var(bh_lrus);
}
1444 1445 1446 1447 1448

static bool has_bh_in_lru(int cpu, void *dummy)
{
	struct bh_lru *b = per_cpu_ptr(&bh_lrus, cpu);
	int i;
L
Linus Torvalds 已提交
1449
	
1450 1451 1452 1453 1454 1455 1456 1457
	for (i = 0; i < BH_LRU_SIZE; i++) {
		if (b->bhs[i])
			return 1;
	}

	return 0;
}

P
Peter Zijlstra 已提交
1458
void invalidate_bh_lrus(void)
L
Linus Torvalds 已提交
1459
{
1460
	on_each_cpu_cond(has_bh_in_lru, invalidate_bh_lru, NULL, 1, GFP_KERNEL);
L
Linus Torvalds 已提交
1461
}
N
Nick Piggin 已提交
1462
EXPORT_SYMBOL_GPL(invalidate_bh_lrus);
L
Linus Torvalds 已提交
1463 1464 1465 1466 1467

void set_bh_page(struct buffer_head *bh,
		struct page *page, unsigned long offset)
{
	bh->b_page = page;
1468
	BUG_ON(offset >= PAGE_SIZE);
L
Linus Torvalds 已提交
1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481
	if (PageHighMem(page))
		/*
		 * This catches illegal uses and preserves the offset:
		 */
		bh->b_data = (char *)(0 + offset);
	else
		bh->b_data = page_address(page) + offset;
}
EXPORT_SYMBOL(set_bh_page);

/*
 * Called when truncating a buffer on a page completely.
 */
1482 1483 1484 1485 1486 1487

/* Bits that are cleared during an invalidate */
#define BUFFER_FLAGS_DISCARD \
	(1 << BH_Mapped | 1 << BH_New | 1 << BH_Req | \
	 1 << BH_Delay | 1 << BH_Unwritten)

1488
static void discard_buffer(struct buffer_head * bh)
L
Linus Torvalds 已提交
1489
{
1490 1491
	unsigned long b_state, b_state_old;

L
Linus Torvalds 已提交
1492 1493 1494
	lock_buffer(bh);
	clear_buffer_dirty(bh);
	bh->b_bdev = NULL;
1495 1496 1497 1498 1499 1500 1501 1502
	b_state = bh->b_state;
	for (;;) {
		b_state_old = cmpxchg(&bh->b_state, b_state,
				      (b_state & ~BUFFER_FLAGS_DISCARD));
		if (b_state_old == b_state)
			break;
		b_state = b_state_old;
	}
L
Linus Torvalds 已提交
1503 1504 1505 1506
	unlock_buffer(bh);
}

/**
1507
 * block_invalidatepage - invalidate part or all of a buffer-backed page
L
Linus Torvalds 已提交
1508 1509
 *
 * @page: the page which is affected
1510 1511
 * @offset: start of the range to invalidate
 * @length: length of the range to invalidate
L
Linus Torvalds 已提交
1512 1513
 *
 * block_invalidatepage() is called when all or part of the page has become
1514
 * invalidated by a truncate operation.
L
Linus Torvalds 已提交
1515 1516 1517 1518 1519 1520 1521
 *
 * block_invalidatepage() does not have to release all buffers, but it must
 * ensure that no dirty buffer is left outside @offset and that no I/O
 * is underway against any of the blocks which are outside the truncation
 * point.  Because the caller is about to free (and possibly reuse) those
 * blocks on-disk.
 */
1522 1523
void block_invalidatepage(struct page *page, unsigned int offset,
			  unsigned int length)
L
Linus Torvalds 已提交
1524 1525 1526
{
	struct buffer_head *head, *bh, *next;
	unsigned int curr_off = 0;
1527
	unsigned int stop = length + offset;
L
Linus Torvalds 已提交
1528 1529 1530 1531 1532

	BUG_ON(!PageLocked(page));
	if (!page_has_buffers(page))
		goto out;

1533 1534 1535 1536 1537
	/*
	 * Check for overflow
	 */
	BUG_ON(stop > PAGE_CACHE_SIZE || stop < length);

L
Linus Torvalds 已提交
1538 1539 1540 1541 1542 1543
	head = page_buffers(page);
	bh = head;
	do {
		unsigned int next_off = curr_off + bh->b_size;
		next = bh->b_this_page;

1544 1545 1546 1547 1548 1549
		/*
		 * Are we still fully in range ?
		 */
		if (next_off > stop)
			goto out;

L
Linus Torvalds 已提交
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
		/*
		 * is this block fully invalidated?
		 */
		if (offset <= curr_off)
			discard_buffer(bh);
		curr_off = next_off;
		bh = next;
	} while (bh != head);

	/*
	 * We release buffers only if the entire page is being invalidated.
	 * The get_block cached value has been unconditionally invalidated,
	 * so real IO is not possible anymore.
	 */
	if (offset == 0)
1565
		try_to_release_page(page, 0);
L
Linus Torvalds 已提交
1566
out:
1567
	return;
L
Linus Torvalds 已提交
1568 1569 1570
}
EXPORT_SYMBOL(block_invalidatepage);

1571

L
Linus Torvalds 已提交
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 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628
/*
 * We attach and possibly dirty the buffers atomically wrt
 * __set_page_dirty_buffers() via private_lock.  try_to_free_buffers
 * is already excluded via the page lock.
 */
void create_empty_buffers(struct page *page,
			unsigned long blocksize, unsigned long b_state)
{
	struct buffer_head *bh, *head, *tail;

	head = alloc_page_buffers(page, blocksize, 1);
	bh = head;
	do {
		bh->b_state |= b_state;
		tail = bh;
		bh = bh->b_this_page;
	} while (bh);
	tail->b_this_page = head;

	spin_lock(&page->mapping->private_lock);
	if (PageUptodate(page) || PageDirty(page)) {
		bh = head;
		do {
			if (PageDirty(page))
				set_buffer_dirty(bh);
			if (PageUptodate(page))
				set_buffer_uptodate(bh);
			bh = bh->b_this_page;
		} while (bh != head);
	}
	attach_page_buffers(page, head);
	spin_unlock(&page->mapping->private_lock);
}
EXPORT_SYMBOL(create_empty_buffers);

/*
 * We are taking a block for data and we don't want any output from any
 * buffer-cache aliases starting from return from that function and
 * until the moment when something will explicitly mark the buffer
 * dirty (hopefully that will not happen until we will free that block ;-)
 * We don't even need to mark it not-uptodate - nobody can expect
 * anything from a newly allocated buffer anyway. We used to used
 * unmap_buffer() for such invalidation, but that was wrong. We definitely
 * don't want to mark the alias unmapped, for example - it would confuse
 * anyone who might pick it with bread() afterwards...
 *
 * Also..  Note that bforget() doesn't lock the buffer.  So there can
 * be writeout I/O going on against recently-freed buffers.  We don't
 * wait on that I/O in bforget() - it's more efficient to wait on the I/O
 * only if we really need to.  That happens here.
 */
void unmap_underlying_metadata(struct block_device *bdev, sector_t block)
{
	struct buffer_head *old_bh;

	might_sleep();

1629
	old_bh = __find_get_block_slow(bdev, block);
L
Linus Torvalds 已提交
1630 1631 1632 1633 1634 1635 1636 1637 1638
	if (old_bh) {
		clear_buffer_dirty(old_bh);
		wait_on_buffer(old_bh);
		clear_buffer_req(old_bh);
		__brelse(old_bh);
	}
}
EXPORT_SYMBOL(unmap_underlying_metadata);

1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
/*
 * Size is a power-of-two in the range 512..PAGE_SIZE,
 * and the case we care about most is PAGE_SIZE.
 *
 * So this *could* possibly be written with those
 * constraints in mind (relevant mostly if some
 * architecture has a slow bit-scan instruction)
 */
static inline int block_size_bits(unsigned int blocksize)
{
	return ilog2(blocksize);
}

static struct buffer_head *create_page_buffers(struct page *page, struct inode *inode, unsigned int b_state)
{
	BUG_ON(!PageLocked(page));

	if (!page_has_buffers(page))
		create_empty_buffers(page, 1 << ACCESS_ONCE(inode->i_blkbits), b_state);
	return page_buffers(page);
}

L
Linus Torvalds 已提交
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
/*
 * NOTE! All mapped/uptodate combinations are valid:
 *
 *	Mapped	Uptodate	Meaning
 *
 *	No	No		"unknown" - must do get_block()
 *	No	Yes		"hole" - zero-filled
 *	Yes	No		"allocated" - allocated on disk, not read in
 *	Yes	Yes		"valid" - allocated and up-to-date in memory.
 *
 * "Dirty" is valid only with the last case (mapped+uptodate).
 */

/*
 * While block_write_full_page is writing back the dirty buffers under
 * the page lock, whoever dirtied the buffers may decide to clean them
 * again at any time.  We handle that by only looking at the buffer
 * state inside lock_buffer().
 *
 * If block_write_full_page() is called for regular writeback
 * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a
 * locked buffer.   This only can happen if someone has written the buffer
 * directly, with submit_bh().  At the address_space level PageWriteback
 * prevents this contention from occurring.
1685 1686
 *
 * If block_write_full_page() is called with wbc->sync_mode ==
J
Jens Axboe 已提交
1687 1688
 * WB_SYNC_ALL, the writes are posted using WRITE_SYNC; this
 * causes the writes to be flagged as synchronous writes.
L
Linus Torvalds 已提交
1689 1690
 */
static int __block_write_full_page(struct inode *inode, struct page *page,
1691 1692
			get_block_t *get_block, struct writeback_control *wbc,
			bh_end_io_t *handler)
L
Linus Torvalds 已提交
1693 1694 1695 1696
{
	int err;
	sector_t block;
	sector_t last_block;
1697
	struct buffer_head *bh, *head;
1698
	unsigned int blocksize, bbits;
L
Linus Torvalds 已提交
1699
	int nr_underway = 0;
1700
	int write_op = (wbc->sync_mode == WB_SYNC_ALL ?
J
Jens Axboe 已提交
1701
			WRITE_SYNC : WRITE);
L
Linus Torvalds 已提交
1702

1703
	head = create_page_buffers(page, inode,
L
Linus Torvalds 已提交
1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
					(1 << BH_Dirty)|(1 << BH_Uptodate));

	/*
	 * Be very careful.  We have no exclusion from __set_page_dirty_buffers
	 * here, and the (potentially unmapped) buffers may become dirty at
	 * any time.  If a buffer becomes dirty here after we've inspected it
	 * then we just miss that fact, and the page stays dirty.
	 *
	 * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
	 * handle that here by just cleaning them.
	 */

	bh = head;
1717 1718 1719 1720 1721
	blocksize = bh->b_size;
	bbits = block_size_bits(blocksize);

	block = (sector_t)page->index << (PAGE_CACHE_SHIFT - bbits);
	last_block = (i_size_read(inode) - 1) >> bbits;
L
Linus Torvalds 已提交
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738

	/*
	 * Get all the dirty buffers mapped to disk addresses and
	 * handle any aliases from the underlying blockdev's mapping.
	 */
	do {
		if (block > last_block) {
			/*
			 * mapped buffers outside i_size will occur, because
			 * this page can be outside i_size when there is a
			 * truncate in progress.
			 */
			/*
			 * The buffer was zeroed by block_write_full_page()
			 */
			clear_buffer_dirty(bh);
			set_buffer_uptodate(bh);
1739 1740
		} else if ((!buffer_mapped(bh) || buffer_delay(bh)) &&
			   buffer_dirty(bh)) {
1741
			WARN_ON(bh->b_size != blocksize);
L
Linus Torvalds 已提交
1742 1743 1744
			err = get_block(inode, block, bh, 1);
			if (err)
				goto recover;
1745
			clear_buffer_delay(bh);
L
Linus Torvalds 已提交
1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762
			if (buffer_new(bh)) {
				/* blockdev mappings never come here */
				clear_buffer_new(bh);
				unmap_underlying_metadata(bh->b_bdev,
							bh->b_blocknr);
			}
		}
		bh = bh->b_this_page;
		block++;
	} while (bh != head);

	do {
		if (!buffer_mapped(bh))
			continue;
		/*
		 * If it's a fully non-blocking write attempt and we cannot
		 * lock the buffer then redirty the page.  Note that this can
1763 1764 1765
		 * potentially cause a busy-wait loop from writeback threads
		 * and kswapd activity, but those code paths have their own
		 * higher-level throttling.
L
Linus Torvalds 已提交
1766
		 */
1767
		if (wbc->sync_mode != WB_SYNC_NONE) {
L
Linus Torvalds 已提交
1768
			lock_buffer(bh);
N
Nick Piggin 已提交
1769
		} else if (!trylock_buffer(bh)) {
L
Linus Torvalds 已提交
1770 1771 1772 1773
			redirty_page_for_writepage(wbc, page);
			continue;
		}
		if (test_clear_buffer_dirty(bh)) {
1774
			mark_buffer_async_write_endio(bh, handler);
L
Linus Torvalds 已提交
1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789
		} else {
			unlock_buffer(bh);
		}
	} while ((bh = bh->b_this_page) != head);

	/*
	 * The page and its buffers are protected by PageWriteback(), so we can
	 * drop the bh refcounts early.
	 */
	BUG_ON(PageWriteback(page));
	set_page_writeback(page);

	do {
		struct buffer_head *next = bh->b_this_page;
		if (buffer_async_write(bh)) {
1790
			submit_bh(write_op, bh);
L
Linus Torvalds 已提交
1791 1792 1793 1794
			nr_underway++;
		}
		bh = next;
	} while (bh != head);
1795
	unlock_page(page);
L
Linus Torvalds 已提交
1796 1797 1798 1799 1800 1801 1802 1803 1804 1805

	err = 0;
done:
	if (nr_underway == 0) {
		/*
		 * The page was marked dirty, but the buffers were
		 * clean.  Someone wrote them back by hand with
		 * ll_rw_block/submit_bh.  A rare case.
		 */
		end_page_writeback(page);
1806

L
Linus Torvalds 已提交
1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
		/*
		 * The page and buffer_heads can be released at any time from
		 * here on.
		 */
	}
	return err;

recover:
	/*
	 * ENOSPC, or some other error.  We may already have added some
	 * blocks to the file, so we need to write these out to avoid
	 * exposing stale data.
	 * The page is currently locked and not marked for writeback
	 */
	bh = head;
	/* Recovery: lock and submit the mapped buffers */
	do {
1824 1825
		if (buffer_mapped(bh) && buffer_dirty(bh) &&
		    !buffer_delay(bh)) {
L
Linus Torvalds 已提交
1826
			lock_buffer(bh);
1827
			mark_buffer_async_write_endio(bh, handler);
L
Linus Torvalds 已提交
1828 1829 1830 1831 1832 1833 1834 1835 1836 1837
		} else {
			/*
			 * The buffer may have been set dirty during
			 * attachment to a dirty page.
			 */
			clear_buffer_dirty(bh);
		}
	} while ((bh = bh->b_this_page) != head);
	SetPageError(page);
	BUG_ON(PageWriteback(page));
1838
	mapping_set_error(page->mapping, err);
L
Linus Torvalds 已提交
1839 1840 1841 1842 1843
	set_page_writeback(page);
	do {
		struct buffer_head *next = bh->b_this_page;
		if (buffer_async_write(bh)) {
			clear_buffer_dirty(bh);
1844
			submit_bh(write_op, bh);
L
Linus Torvalds 已提交
1845 1846 1847 1848
			nr_underway++;
		}
		bh = next;
	} while (bh != head);
1849
	unlock_page(page);
L
Linus Torvalds 已提交
1850 1851 1852
	goto done;
}

1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879
/*
 * If a page has any new buffers, zero them out here, and mark them uptodate
 * and dirty so they'll be written out (in order to prevent uninitialised
 * block data from leaking). And clear the new bit.
 */
void page_zero_new_buffers(struct page *page, unsigned from, unsigned to)
{
	unsigned int block_start, block_end;
	struct buffer_head *head, *bh;

	BUG_ON(!PageLocked(page));
	if (!page_has_buffers(page))
		return;

	bh = head = page_buffers(page);
	block_start = 0;
	do {
		block_end = block_start + bh->b_size;

		if (buffer_new(bh)) {
			if (block_end > from && block_start < to) {
				if (!PageUptodate(page)) {
					unsigned start, size;

					start = max(from, block_start);
					size = min(to, block_end) - start;

1880
					zero_user(page, start, size);
1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894
					set_buffer_uptodate(bh);
				}

				clear_buffer_new(bh);
				mark_buffer_dirty(bh);
			}
		}

		block_start = block_end;
		bh = bh->b_this_page;
	} while (bh != head);
}
EXPORT_SYMBOL(page_zero_new_buffers);

C
Christoph Hellwig 已提交
1895
int __block_write_begin(struct page *page, loff_t pos, unsigned len,
1896
		get_block_t *get_block)
L
Linus Torvalds 已提交
1897
{
C
Christoph Hellwig 已提交
1898 1899
	unsigned from = pos & (PAGE_CACHE_SIZE - 1);
	unsigned to = from + len;
1900
	struct inode *inode = page->mapping->host;
L
Linus Torvalds 已提交
1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
	unsigned block_start, block_end;
	sector_t block;
	int err = 0;
	unsigned blocksize, bbits;
	struct buffer_head *bh, *head, *wait[2], **wait_bh=wait;

	BUG_ON(!PageLocked(page));
	BUG_ON(from > PAGE_CACHE_SIZE);
	BUG_ON(to > PAGE_CACHE_SIZE);
	BUG_ON(from > to);

1912 1913 1914
	head = create_page_buffers(page, inode, 0);
	blocksize = head->b_size;
	bbits = block_size_bits(blocksize);
L
Linus Torvalds 已提交
1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930

	block = (sector_t)page->index << (PAGE_CACHE_SHIFT - bbits);

	for(bh = head, block_start = 0; bh != head || !block_start;
	    block++, block_start=block_end, bh = bh->b_this_page) {
		block_end = block_start + blocksize;
		if (block_end <= from || block_start >= to) {
			if (PageUptodate(page)) {
				if (!buffer_uptodate(bh))
					set_buffer_uptodate(bh);
			}
			continue;
		}
		if (buffer_new(bh))
			clear_buffer_new(bh);
		if (!buffer_mapped(bh)) {
1931
			WARN_ON(bh->b_size != blocksize);
L
Linus Torvalds 已提交
1932 1933
			err = get_block(inode, block, bh, 1);
			if (err)
1934
				break;
L
Linus Torvalds 已提交
1935 1936 1937 1938
			if (buffer_new(bh)) {
				unmap_underlying_metadata(bh->b_bdev,
							bh->b_blocknr);
				if (PageUptodate(page)) {
N
Nick Piggin 已提交
1939
					clear_buffer_new(bh);
L
Linus Torvalds 已提交
1940
					set_buffer_uptodate(bh);
N
Nick Piggin 已提交
1941
					mark_buffer_dirty(bh);
L
Linus Torvalds 已提交
1942 1943
					continue;
				}
1944 1945 1946 1947
				if (block_end > to || block_start < from)
					zero_user_segments(page,
						to, block_end,
						block_start, from);
L
Linus Torvalds 已提交
1948 1949 1950 1951 1952 1953 1954 1955 1956
				continue;
			}
		}
		if (PageUptodate(page)) {
			if (!buffer_uptodate(bh))
				set_buffer_uptodate(bh);
			continue; 
		}
		if (!buffer_uptodate(bh) && !buffer_delay(bh) &&
1957
		    !buffer_unwritten(bh) &&
L
Linus Torvalds 已提交
1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968
		     (block_start < from || block_end > to)) {
			ll_rw_block(READ, 1, &bh);
			*wait_bh++=bh;
		}
	}
	/*
	 * If we issued read requests - let them complete.
	 */
	while(wait_bh > wait) {
		wait_on_buffer(*--wait_bh);
		if (!buffer_uptodate(*wait_bh))
1969
			err = -EIO;
L
Linus Torvalds 已提交
1970
	}
1971
	if (unlikely(err))
1972
		page_zero_new_buffers(page, from, to);
L
Linus Torvalds 已提交
1973 1974
	return err;
}
C
Christoph Hellwig 已提交
1975
EXPORT_SYMBOL(__block_write_begin);
L
Linus Torvalds 已提交
1976 1977 1978 1979 1980 1981 1982 1983 1984

static int __block_commit_write(struct inode *inode, struct page *page,
		unsigned from, unsigned to)
{
	unsigned block_start, block_end;
	int partial = 0;
	unsigned blocksize;
	struct buffer_head *bh, *head;

1985 1986
	bh = head = page_buffers(page);
	blocksize = bh->b_size;
L
Linus Torvalds 已提交
1987

1988 1989
	block_start = 0;
	do {
L
Linus Torvalds 已提交
1990 1991 1992 1993 1994 1995 1996 1997
		block_end = block_start + blocksize;
		if (block_end <= from || block_start >= to) {
			if (!buffer_uptodate(bh))
				partial = 1;
		} else {
			set_buffer_uptodate(bh);
			mark_buffer_dirty(bh);
		}
1998
		clear_buffer_new(bh);
1999 2000 2001 2002

		block_start = block_end;
		bh = bh->b_this_page;
	} while (bh != head);
L
Linus Torvalds 已提交
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

	/*
	 * If this is a partial write which happened to make all buffers
	 * uptodate then we can optimize away a bogus readpage() for
	 * the next read(). Here we 'discover' whether the page went
	 * uptodate as a result of this (potentially partial) write.
	 */
	if (!partial)
		SetPageUptodate(page);
	return 0;
}

2015
/*
2016 2017 2018
 * block_write_begin takes care of the basic task of block allocation and
 * bringing partial write blocks uptodate first.
 *
2019
 * The filesystem needs to handle block truncation upon failure.
2020
 */
2021 2022
int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len,
		unsigned flags, struct page **pagep, get_block_t *get_block)
2023
{
2024
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
2025
	struct page *page;
2026
	int status;
2027

2028 2029 2030
	page = grab_cache_page_write_begin(mapping, index, flags);
	if (!page)
		return -ENOMEM;
2031

2032
	status = __block_write_begin(page, pos, len, get_block);
2033
	if (unlikely(status)) {
2034 2035 2036
		unlock_page(page);
		page_cache_release(page);
		page = NULL;
2037 2038
	}

2039
	*pagep = page;
2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
	return status;
}
EXPORT_SYMBOL(block_write_begin);

int block_write_end(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned copied,
			struct page *page, void *fsdata)
{
	struct inode *inode = mapping->host;
	unsigned start;

	start = pos & (PAGE_CACHE_SIZE - 1);

	if (unlikely(copied < len)) {
		/*
		 * The buffers that were written will now be uptodate, so we
		 * don't have to worry about a readpage reading them and
		 * overwriting a partial write. However if we have encountered
		 * a short write and only partially written into a buffer, it
		 * will not be marked uptodate, so a readpage might come in and
		 * destroy our partial write.
		 *
		 * Do the simplest thing, and just treat any short write to a
		 * non uptodate page as a zero-length write, and force the
		 * caller to redo the whole thing.
		 */
		if (!PageUptodate(page))
			copied = 0;

		page_zero_new_buffers(page, start+copied, start+len);
	}
	flush_dcache_page(page);

	/* This could be a short (even 0-length) commit */
	__block_commit_write(inode, page, start, start+copied);

	return copied;
}
EXPORT_SYMBOL(block_write_end);

int generic_write_end(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned copied,
			struct page *page, void *fsdata)
{
	struct inode *inode = mapping->host;
2085
	int i_size_changed = 0;
2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097

	copied = block_write_end(file, mapping, pos, len, copied, page, fsdata);

	/*
	 * No need to use i_size_read() here, the i_size
	 * cannot change under us because we hold i_mutex.
	 *
	 * But it's important to update i_size while still holding page lock:
	 * page writeout could otherwise come in and zero beyond i_size.
	 */
	if (pos+copied > inode->i_size) {
		i_size_write(inode, pos+copied);
2098
		i_size_changed = 1;
2099 2100 2101 2102 2103
	}

	unlock_page(page);
	page_cache_release(page);

2104 2105 2106 2107 2108 2109 2110 2111 2112
	/*
	 * Don't mark the inode dirty under page lock. First, it unnecessarily
	 * makes the holding time of page lock longer. Second, it forces lock
	 * ordering of page lock and transaction start for journaling
	 * filesystems.
	 */
	if (i_size_changed)
		mark_inode_dirty(inode);

2113 2114 2115 2116
	return copied;
}
EXPORT_SYMBOL(generic_write_end);

2117 2118 2119 2120 2121 2122 2123
/*
 * block_is_partially_uptodate checks whether buffers within a page are
 * uptodate or not.
 *
 * Returns true if all buffers which correspond to a file portion
 * we want to read are uptodate.
 */
2124 2125
int block_is_partially_uptodate(struct page *page, unsigned long from,
					unsigned long count)
2126 2127 2128 2129 2130 2131 2132 2133 2134
{
	unsigned block_start, block_end, blocksize;
	unsigned to;
	struct buffer_head *bh, *head;
	int ret = 1;

	if (!page_has_buffers(page))
		return 0;

2135 2136
	head = page_buffers(page);
	blocksize = head->b_size;
2137
	to = min_t(unsigned, PAGE_CACHE_SIZE - from, count);
2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161
	to = from + to;
	if (from < blocksize && to > PAGE_CACHE_SIZE - blocksize)
		return 0;

	bh = head;
	block_start = 0;
	do {
		block_end = block_start + blocksize;
		if (block_end > from && block_start < to) {
			if (!buffer_uptodate(bh)) {
				ret = 0;
				break;
			}
			if (block_end >= to)
				break;
		}
		block_start = block_end;
		bh = bh->b_this_page;
	} while (bh != head);

	return ret;
}
EXPORT_SYMBOL(block_is_partially_uptodate);

L
Linus Torvalds 已提交
2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173
/*
 * Generic "read page" function for block devices that have the normal
 * get_block functionality. This is most of the block device filesystems.
 * Reads the page asynchronously --- the unlock_buffer() and
 * set/clear_buffer_uptodate() functions propagate buffer state into the
 * page struct once IO has completed.
 */
int block_read_full_page(struct page *page, get_block_t *get_block)
{
	struct inode *inode = page->mapping->host;
	sector_t iblock, lblock;
	struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
2174
	unsigned int blocksize, bbits;
L
Linus Torvalds 已提交
2175 2176 2177
	int nr, i;
	int fully_mapped = 1;

2178 2179 2180
	head = create_page_buffers(page, inode, 0);
	blocksize = head->b_size;
	bbits = block_size_bits(blocksize);
L
Linus Torvalds 已提交
2181

2182 2183
	iblock = (sector_t)page->index << (PAGE_CACHE_SHIFT - bbits);
	lblock = (i_size_read(inode)+blocksize-1) >> bbits;
L
Linus Torvalds 已提交
2184 2185 2186 2187 2188 2189 2190 2191 2192
	bh = head;
	nr = 0;
	i = 0;

	do {
		if (buffer_uptodate(bh))
			continue;

		if (!buffer_mapped(bh)) {
2193 2194
			int err = 0;

L
Linus Torvalds 已提交
2195 2196
			fully_mapped = 0;
			if (iblock < lblock) {
2197
				WARN_ON(bh->b_size != blocksize);
2198 2199
				err = get_block(inode, iblock, bh, 0);
				if (err)
L
Linus Torvalds 已提交
2200 2201 2202
					SetPageError(page);
			}
			if (!buffer_mapped(bh)) {
2203
				zero_user(page, i * blocksize, blocksize);
2204 2205
				if (!err)
					set_buffer_uptodate(bh);
L
Linus Torvalds 已提交
2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252
				continue;
			}
			/*
			 * get_block() might have updated the buffer
			 * synchronously
			 */
			if (buffer_uptodate(bh))
				continue;
		}
		arr[nr++] = bh;
	} while (i++, iblock++, (bh = bh->b_this_page) != head);

	if (fully_mapped)
		SetPageMappedToDisk(page);

	if (!nr) {
		/*
		 * All buffers are uptodate - we can set the page uptodate
		 * as well. But not if get_block() returned an error.
		 */
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
		return 0;
	}

	/* Stage two: lock the buffers */
	for (i = 0; i < nr; i++) {
		bh = arr[i];
		lock_buffer(bh);
		mark_buffer_async_read(bh);
	}

	/*
	 * Stage 3: start the IO.  Check for uptodateness
	 * inside the buffer lock in case another process reading
	 * the underlying blockdev brought it uptodate (the sct fix).
	 */
	for (i = 0; i < nr; i++) {
		bh = arr[i];
		if (buffer_uptodate(bh))
			end_buffer_async_read(bh, 1);
		else
			submit_bh(READ, bh);
	}
	return 0;
}
2253
EXPORT_SYMBOL(block_read_full_page);
L
Linus Torvalds 已提交
2254 2255

/* utility function for filesystems that need to do work on expanding
N
Nick Piggin 已提交
2256
 * truncates.  Uses filesystem pagecache writes to allow the filesystem to
L
Linus Torvalds 已提交
2257 2258
 * deal with the hole.  
 */
N
Nick Piggin 已提交
2259
int generic_cont_expand_simple(struct inode *inode, loff_t size)
L
Linus Torvalds 已提交
2260 2261 2262
{
	struct address_space *mapping = inode->i_mapping;
	struct page *page;
N
Nick Piggin 已提交
2263
	void *fsdata;
L
Linus Torvalds 已提交
2264 2265
	int err;

N
npiggin@suse.de 已提交
2266 2267
	err = inode_newsize_ok(inode, size);
	if (err)
L
Linus Torvalds 已提交
2268 2269
		goto out;

N
Nick Piggin 已提交
2270 2271 2272 2273
	err = pagecache_write_begin(NULL, mapping, size, 0,
				AOP_FLAG_UNINTERRUPTIBLE|AOP_FLAG_CONT_EXPAND,
				&page, &fsdata);
	if (err)
2274 2275
		goto out;

N
Nick Piggin 已提交
2276 2277
	err = pagecache_write_end(NULL, mapping, size, 0, 0, page, fsdata);
	BUG_ON(err > 0);
2278

L
Linus Torvalds 已提交
2279 2280 2281
out:
	return err;
}
2282
EXPORT_SYMBOL(generic_cont_expand_simple);
L
Linus Torvalds 已提交
2283

2284 2285
static int cont_expand_zero(struct file *file, struct address_space *mapping,
			    loff_t pos, loff_t *bytes)
L
Linus Torvalds 已提交
2286 2287 2288
{
	struct inode *inode = mapping->host;
	unsigned blocksize = 1 << inode->i_blkbits;
N
Nick Piggin 已提交
2289 2290 2291 2292 2293 2294
	struct page *page;
	void *fsdata;
	pgoff_t index, curidx;
	loff_t curpos;
	unsigned zerofrom, offset, len;
	int err = 0;
L
Linus Torvalds 已提交
2295

N
Nick Piggin 已提交
2296 2297 2298 2299 2300
	index = pos >> PAGE_CACHE_SHIFT;
	offset = pos & ~PAGE_CACHE_MASK;

	while (index > (curidx = (curpos = *bytes)>>PAGE_CACHE_SHIFT)) {
		zerofrom = curpos & ~PAGE_CACHE_MASK;
L
Linus Torvalds 已提交
2301 2302 2303 2304
		if (zerofrom & (blocksize-1)) {
			*bytes |= (blocksize-1);
			(*bytes)++;
		}
N
Nick Piggin 已提交
2305
		len = PAGE_CACHE_SIZE - zerofrom;
L
Linus Torvalds 已提交
2306

N
Nick Piggin 已提交
2307 2308 2309 2310 2311
		err = pagecache_write_begin(file, mapping, curpos, len,
						AOP_FLAG_UNINTERRUPTIBLE,
						&page, &fsdata);
		if (err)
			goto out;
2312
		zero_user(page, zerofrom, len);
N
Nick Piggin 已提交
2313 2314 2315 2316 2317 2318
		err = pagecache_write_end(file, mapping, curpos, len, len,
						page, fsdata);
		if (err < 0)
			goto out;
		BUG_ON(err != len);
		err = 0;
2319 2320

		balance_dirty_pages_ratelimited(mapping);
N
Nick Piggin 已提交
2321
	}
L
Linus Torvalds 已提交
2322

N
Nick Piggin 已提交
2323 2324 2325
	/* page covers the boundary, find the boundary offset */
	if (index == curidx) {
		zerofrom = curpos & ~PAGE_CACHE_MASK;
L
Linus Torvalds 已提交
2326
		/* if we will expand the thing last block will be filled */
N
Nick Piggin 已提交
2327 2328 2329 2330
		if (offset <= zerofrom) {
			goto out;
		}
		if (zerofrom & (blocksize-1)) {
L
Linus Torvalds 已提交
2331 2332 2333
			*bytes |= (blocksize-1);
			(*bytes)++;
		}
N
Nick Piggin 已提交
2334
		len = offset - zerofrom;
L
Linus Torvalds 已提交
2335

N
Nick Piggin 已提交
2336 2337 2338 2339 2340
		err = pagecache_write_begin(file, mapping, curpos, len,
						AOP_FLAG_UNINTERRUPTIBLE,
						&page, &fsdata);
		if (err)
			goto out;
2341
		zero_user(page, zerofrom, len);
N
Nick Piggin 已提交
2342 2343 2344 2345 2346 2347
		err = pagecache_write_end(file, mapping, curpos, len, len,
						page, fsdata);
		if (err < 0)
			goto out;
		BUG_ON(err != len);
		err = 0;
L
Linus Torvalds 已提交
2348
	}
N
Nick Piggin 已提交
2349 2350 2351 2352 2353 2354 2355 2356
out:
	return err;
}

/*
 * For moronic filesystems that do not allow holes in file.
 * We may have to extend the file.
 */
2357
int cont_write_begin(struct file *file, struct address_space *mapping,
N
Nick Piggin 已提交
2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368
			loff_t pos, unsigned len, unsigned flags,
			struct page **pagep, void **fsdata,
			get_block_t *get_block, loff_t *bytes)
{
	struct inode *inode = mapping->host;
	unsigned blocksize = 1 << inode->i_blkbits;
	unsigned zerofrom;
	int err;

	err = cont_expand_zero(file, mapping, pos, bytes);
	if (err)
2369
		return err;
N
Nick Piggin 已提交
2370 2371 2372 2373 2374

	zerofrom = *bytes & ~PAGE_CACHE_MASK;
	if (pos+len > *bytes && zerofrom & (blocksize-1)) {
		*bytes |= (blocksize-1);
		(*bytes)++;
L
Linus Torvalds 已提交
2375 2376
	}

2377
	return block_write_begin(mapping, pos, len, flags, pagep, get_block);
L
Linus Torvalds 已提交
2378
}
2379
EXPORT_SYMBOL(cont_write_begin);
L
Linus Torvalds 已提交
2380 2381 2382 2383 2384 2385 2386

int block_commit_write(struct page *page, unsigned from, unsigned to)
{
	struct inode *inode = page->mapping->host;
	__block_commit_write(inode,page,from,to);
	return 0;
}
2387
EXPORT_SYMBOL(block_commit_write);
L
Linus Torvalds 已提交
2388

2389 2390 2391 2392 2393 2394 2395 2396 2397 2398
/*
 * block_page_mkwrite() is not allowed to change the file size as it gets
 * called from a page fault handler when a page is first dirtied. Hence we must
 * be careful to check for EOF conditions here. We set the page up correctly
 * for a written page which means we get ENOSPC checking when writing into
 * holes and correct delalloc and unwritten extent mapping on filesystems that
 * support these features.
 *
 * We are not allowed to take the i_mutex here so we have to play games to
 * protect against truncate races as the page could now be beyond EOF.  Because
2399
 * truncate writes the inode size before removing pages, once we have the
2400 2401 2402
 * page lock we can determine safely if the page is beyond EOF. If it is not
 * beyond EOF, then the page is guaranteed safe against truncation until we
 * unlock the page.
2403
 *
2404 2405
 * Direct callers of this function should protect against filesystem freezing
 * using sb_start_write() - sb_end_write() functions.
2406
 */
2407 2408
int __block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
			 get_block_t get_block)
2409
{
2410
	struct page *page = vmf->page;
A
Al Viro 已提交
2411
	struct inode *inode = file_inode(vma->vm_file);
2412 2413
	unsigned long end;
	loff_t size;
2414
	int ret;
2415 2416 2417 2418

	lock_page(page);
	size = i_size_read(inode);
	if ((page->mapping != inode->i_mapping) ||
N
Nick Piggin 已提交
2419
	    (page_offset(page) > size)) {
2420 2421 2422
		/* We overload EFAULT to mean page got truncated */
		ret = -EFAULT;
		goto out_unlock;
2423 2424 2425 2426 2427 2428 2429 2430
	}

	/* page is wholly or partially inside EOF */
	if (((page->index + 1) << PAGE_CACHE_SHIFT) > size)
		end = size & ~PAGE_CACHE_MASK;
	else
		end = PAGE_CACHE_SIZE;

C
Christoph Hellwig 已提交
2431
	ret = __block_write_begin(page, 0, end, get_block);
2432 2433 2434
	if (!ret)
		ret = block_commit_write(page, 0, end);

2435 2436
	if (unlikely(ret < 0))
		goto out_unlock;
2437
	set_page_dirty(page);
2438
	wait_for_stable_page(page);
2439 2440 2441
	return 0;
out_unlock:
	unlock_page(page);
2442
	return ret;
2443 2444 2445 2446 2447 2448
}
EXPORT_SYMBOL(__block_page_mkwrite);

int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
		   get_block_t get_block)
{
2449
	int ret;
A
Al Viro 已提交
2450
	struct super_block *sb = file_inode(vma->vm_file)->i_sb;
2451

2452
	sb_start_pagefault(sb);
2453 2454 2455 2456 2457 2458 2459

	/*
	 * Update file times before taking page lock. We may end up failing the
	 * fault so this update may be superfluous but who really cares...
	 */
	file_update_time(vma->vm_file);

2460
	ret = __block_page_mkwrite(vma, vmf, get_block);
2461
	sb_end_pagefault(sb);
2462
	return block_page_mkwrite_return(ret);
2463
}
2464
EXPORT_SYMBOL(block_page_mkwrite);
L
Linus Torvalds 已提交
2465 2466

/*
N
Nick Piggin 已提交
2467
 * nobh_write_begin()'s prereads are special: the buffer_heads are freed
L
Linus Torvalds 已提交
2468 2469 2470 2471 2472
 * immediately, while under the page lock.  So it needs a special end_io
 * handler which does not touch the bh after unlocking it.
 */
static void end_buffer_read_nobh(struct buffer_head *bh, int uptodate)
{
2473
	__end_buffer_read_notouch(bh, uptodate);
L
Linus Torvalds 已提交
2474 2475
}

N
Nick Piggin 已提交
2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499
/*
 * Attach the singly-linked list of buffers created by nobh_write_begin, to
 * the page (converting it to circular linked list and taking care of page
 * dirty races).
 */
static void attach_nobh_buffers(struct page *page, struct buffer_head *head)
{
	struct buffer_head *bh;

	BUG_ON(!PageLocked(page));

	spin_lock(&page->mapping->private_lock);
	bh = head;
	do {
		if (PageDirty(page))
			set_buffer_dirty(bh);
		if (!bh->b_this_page)
			bh->b_this_page = head;
		bh = bh->b_this_page;
	} while (bh != head);
	attach_page_buffers(page, head);
	spin_unlock(&page->mapping->private_lock);
}

L
Linus Torvalds 已提交
2500
/*
2501 2502
 * On entry, the page is fully not uptodate.
 * On exit the page is fully uptodate in the areas outside (from,to)
2503
 * The filesystem needs to handle block truncation upon failure.
L
Linus Torvalds 已提交
2504
 */
2505
int nobh_write_begin(struct address_space *mapping,
N
Nick Piggin 已提交
2506 2507
			loff_t pos, unsigned len, unsigned flags,
			struct page **pagep, void **fsdata,
L
Linus Torvalds 已提交
2508 2509
			get_block_t *get_block)
{
N
Nick Piggin 已提交
2510
	struct inode *inode = mapping->host;
L
Linus Torvalds 已提交
2511 2512
	const unsigned blkbits = inode->i_blkbits;
	const unsigned blocksize = 1 << blkbits;
N
Nick Piggin 已提交
2513
	struct buffer_head *head, *bh;
N
Nick Piggin 已提交
2514 2515 2516
	struct page *page;
	pgoff_t index;
	unsigned from, to;
L
Linus Torvalds 已提交
2517
	unsigned block_in_page;
N
Nick Piggin 已提交
2518
	unsigned block_start, block_end;
L
Linus Torvalds 已提交
2519 2520 2521 2522 2523
	sector_t block_in_file;
	int nr_reads = 0;
	int ret = 0;
	int is_mapped_to_disk = 1;

N
Nick Piggin 已提交
2524 2525 2526 2527
	index = pos >> PAGE_CACHE_SHIFT;
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;

2528
	page = grab_cache_page_write_begin(mapping, index, flags);
N
Nick Piggin 已提交
2529 2530 2531 2532 2533 2534
	if (!page)
		return -ENOMEM;
	*pagep = page;
	*fsdata = NULL;

	if (page_has_buffers(page)) {
2535 2536 2537 2538
		ret = __block_write_begin(page, pos, len, get_block);
		if (unlikely(ret))
			goto out_release;
		return ret;
N
Nick Piggin 已提交
2539
	}
N
Nick Piggin 已提交
2540

L
Linus Torvalds 已提交
2541 2542 2543
	if (PageMappedToDisk(page))
		return 0;

N
Nick Piggin 已提交
2544 2545 2546 2547 2548 2549 2550 2551 2552 2553
	/*
	 * Allocate buffers so that we can keep track of state, and potentially
	 * attach them to the page if an error occurs. In the common case of
	 * no error, they will just be freed again without ever being attached
	 * to the page (which is all OK, because we're under the page lock).
	 *
	 * Be careful: the buffer linked list is a NULL terminated one, rather
	 * than the circular one we're used to.
	 */
	head = alloc_page_buffers(page, blocksize, 0);
N
Nick Piggin 已提交
2554 2555 2556 2557
	if (!head) {
		ret = -ENOMEM;
		goto out_release;
	}
N
Nick Piggin 已提交
2558

L
Linus Torvalds 已提交
2559 2560 2561 2562 2563 2564 2565
	block_in_file = (sector_t)page->index << (PAGE_CACHE_SHIFT - blkbits);

	/*
	 * We loop across all blocks in the page, whether or not they are
	 * part of the affected region.  This is so we can discover if the
	 * page is fully mapped-to-disk.
	 */
N
Nick Piggin 已提交
2566
	for (block_start = 0, block_in_page = 0, bh = head;
L
Linus Torvalds 已提交
2567
		  block_start < PAGE_CACHE_SIZE;
N
Nick Piggin 已提交
2568
		  block_in_page++, block_start += blocksize, bh = bh->b_this_page) {
L
Linus Torvalds 已提交
2569 2570
		int create;

N
Nick Piggin 已提交
2571 2572
		block_end = block_start + blocksize;
		bh->b_state = 0;
L
Linus Torvalds 已提交
2573 2574 2575 2576
		create = 1;
		if (block_start >= to)
			create = 0;
		ret = get_block(inode, block_in_file + block_in_page,
N
Nick Piggin 已提交
2577
					bh, create);
L
Linus Torvalds 已提交
2578 2579
		if (ret)
			goto failed;
N
Nick Piggin 已提交
2580
		if (!buffer_mapped(bh))
L
Linus Torvalds 已提交
2581
			is_mapped_to_disk = 0;
N
Nick Piggin 已提交
2582 2583 2584 2585
		if (buffer_new(bh))
			unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
		if (PageUptodate(page)) {
			set_buffer_uptodate(bh);
L
Linus Torvalds 已提交
2586
			continue;
N
Nick Piggin 已提交
2587 2588
		}
		if (buffer_new(bh) || !buffer_mapped(bh)) {
2589 2590
			zero_user_segments(page, block_start, from,
							to, block_end);
L
Linus Torvalds 已提交
2591 2592
			continue;
		}
N
Nick Piggin 已提交
2593
		if (buffer_uptodate(bh))
L
Linus Torvalds 已提交
2594 2595
			continue;	/* reiserfs does this */
		if (block_start < from || block_end > to) {
N
Nick Piggin 已提交
2596 2597 2598 2599
			lock_buffer(bh);
			bh->b_end_io = end_buffer_read_nobh;
			submit_bh(READ, bh);
			nr_reads++;
L
Linus Torvalds 已提交
2600 2601 2602 2603 2604 2605 2606 2607 2608
		}
	}

	if (nr_reads) {
		/*
		 * The page is locked, so these buffers are protected from
		 * any VM or truncate activity.  Hence we don't need to care
		 * for the buffer_head refcounts.
		 */
N
Nick Piggin 已提交
2609
		for (bh = head; bh; bh = bh->b_this_page) {
L
Linus Torvalds 已提交
2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620
			wait_on_buffer(bh);
			if (!buffer_uptodate(bh))
				ret = -EIO;
		}
		if (ret)
			goto failed;
	}

	if (is_mapped_to_disk)
		SetPageMappedToDisk(page);

N
Nick Piggin 已提交
2621
	*fsdata = head; /* to be released by nobh_write_end */
N
Nick Piggin 已提交
2622

L
Linus Torvalds 已提交
2623 2624 2625
	return 0;

failed:
N
Nick Piggin 已提交
2626
	BUG_ON(!ret);
L
Linus Torvalds 已提交
2627
	/*
N
Nick Piggin 已提交
2628 2629 2630 2631 2632
	 * Error recovery is a bit difficult. We need to zero out blocks that
	 * were newly allocated, and dirty them to ensure they get written out.
	 * Buffers need to be attached to the page at this point, otherwise
	 * the handling of potential IO errors during writeout would be hard
	 * (could try doing synchronous writeout, but what if that fails too?)
L
Linus Torvalds 已提交
2633
	 */
N
Nick Piggin 已提交
2634 2635
	attach_nobh_buffers(page, head);
	page_zero_new_buffers(page, from, to);
N
Nick Piggin 已提交
2636

N
Nick Piggin 已提交
2637 2638 2639 2640
out_release:
	unlock_page(page);
	page_cache_release(page);
	*pagep = NULL;
N
Nick Piggin 已提交
2641

2642 2643
	return ret;
}
N
Nick Piggin 已提交
2644
EXPORT_SYMBOL(nobh_write_begin);
L
Linus Torvalds 已提交
2645

N
Nick Piggin 已提交
2646 2647 2648
int nobh_write_end(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned copied,
			struct page *page, void *fsdata)
L
Linus Torvalds 已提交
2649 2650
{
	struct inode *inode = page->mapping->host;
N
Nick Piggin 已提交
2651
	struct buffer_head *head = fsdata;
N
Nick Piggin 已提交
2652
	struct buffer_head *bh;
2653
	BUG_ON(fsdata != NULL && page_has_buffers(page));
L
Linus Torvalds 已提交
2654

2655
	if (unlikely(copied < len) && head)
2656 2657 2658 2659
		attach_nobh_buffers(page, head);
	if (page_has_buffers(page))
		return generic_write_end(file, mapping, pos, len,
					copied, page, fsdata);
N
Nick Piggin 已提交
2660

N
Nick Piggin 已提交
2661
	SetPageUptodate(page);
L
Linus Torvalds 已提交
2662
	set_page_dirty(page);
N
Nick Piggin 已提交
2663 2664
	if (pos+copied > inode->i_size) {
		i_size_write(inode, pos+copied);
L
Linus Torvalds 已提交
2665 2666
		mark_inode_dirty(inode);
	}
N
Nick Piggin 已提交
2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677

	unlock_page(page);
	page_cache_release(page);

	while (head) {
		bh = head;
		head = head->b_this_page;
		free_buffer_head(bh);
	}

	return copied;
L
Linus Torvalds 已提交
2678
}
N
Nick Piggin 已提交
2679
EXPORT_SYMBOL(nobh_write_end);
L
Linus Torvalds 已提交
2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722

/*
 * nobh_writepage() - based on block_full_write_page() except
 * that it tries to operate without attaching bufferheads to
 * the page.
 */
int nobh_writepage(struct page *page, get_block_t *get_block,
			struct writeback_control *wbc)
{
	struct inode * const inode = page->mapping->host;
	loff_t i_size = i_size_read(inode);
	const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
	unsigned offset;
	int ret;

	/* Is the page fully inside i_size? */
	if (page->index < end_index)
		goto out;

	/* Is the page fully outside i_size? (truncate in progress) */
	offset = i_size & (PAGE_CACHE_SIZE-1);
	if (page->index >= end_index+1 || !offset) {
		/*
		 * The page may have dirty, unmapped buffers.  For example,
		 * they may have been added in ext3_writepage().  Make them
		 * freeable here, so the page does not leak.
		 */
#if 0
		/* Not really sure about this  - do we need this ? */
		if (page->mapping->a_ops->invalidatepage)
			page->mapping->a_ops->invalidatepage(page, offset);
#endif
		unlock_page(page);
		return 0; /* don't care */
	}

	/*
	 * The page straddles i_size.  It must be zeroed out on each and every
	 * writepage invocation because it may be mmapped.  "A file is mapped
	 * in multiples of the page size.  For a file that is not a multiple of
	 * the  page size, the remaining memory is zeroed when mapped, and
	 * writes to that region are not written out to the file."
	 */
2723
	zero_user_segment(page, offset, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
2724 2725 2726
out:
	ret = mpage_writepage(page, get_block, wbc);
	if (ret == -EAGAIN)
2727 2728
		ret = __block_write_full_page(inode, page, get_block, wbc,
					      end_buffer_async_write);
L
Linus Torvalds 已提交
2729 2730 2731 2732
	return ret;
}
EXPORT_SYMBOL(nobh_writepage);

N
Nick Piggin 已提交
2733 2734
int nobh_truncate_page(struct address_space *mapping,
			loff_t from, get_block_t *get_block)
L
Linus Torvalds 已提交
2735 2736 2737
{
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
N
Nick Piggin 已提交
2738 2739 2740 2741
	unsigned blocksize;
	sector_t iblock;
	unsigned length, pos;
	struct inode *inode = mapping->host;
L
Linus Torvalds 已提交
2742
	struct page *page;
N
Nick Piggin 已提交
2743 2744
	struct buffer_head map_bh;
	int err;
L
Linus Torvalds 已提交
2745

N
Nick Piggin 已提交
2746 2747 2748 2749 2750 2751 2752 2753 2754
	blocksize = 1 << inode->i_blkbits;
	length = offset & (blocksize - 1);

	/* Block boundary? Nothing to do */
	if (!length)
		return 0;

	length = blocksize - length;
	iblock = (sector_t)index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
L
Linus Torvalds 已提交
2755 2756

	page = grab_cache_page(mapping, index);
N
Nick Piggin 已提交
2757
	err = -ENOMEM;
L
Linus Torvalds 已提交
2758 2759 2760
	if (!page)
		goto out;

N
Nick Piggin 已提交
2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774
	if (page_has_buffers(page)) {
has_buffers:
		unlock_page(page);
		page_cache_release(page);
		return block_truncate_page(mapping, from, get_block);
	}

	/* Find the buffer that contains "offset" */
	pos = blocksize;
	while (offset >= pos) {
		iblock++;
		pos += blocksize;
	}

2775 2776
	map_bh.b_size = blocksize;
	map_bh.b_state = 0;
N
Nick Piggin 已提交
2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797
	err = get_block(inode, iblock, &map_bh, 0);
	if (err)
		goto unlock;
	/* unmapped? It's a hole - nothing to do */
	if (!buffer_mapped(&map_bh))
		goto unlock;

	/* Ok, it's mapped. Make sure it's up-to-date */
	if (!PageUptodate(page)) {
		err = mapping->a_ops->readpage(NULL, page);
		if (err) {
			page_cache_release(page);
			goto out;
		}
		lock_page(page);
		if (!PageUptodate(page)) {
			err = -EIO;
			goto unlock;
		}
		if (page_has_buffers(page))
			goto has_buffers;
L
Linus Torvalds 已提交
2798
	}
2799
	zero_user(page, offset, length);
N
Nick Piggin 已提交
2800 2801 2802 2803
	set_page_dirty(page);
	err = 0;

unlock:
L
Linus Torvalds 已提交
2804 2805 2806
	unlock_page(page);
	page_cache_release(page);
out:
N
Nick Piggin 已提交
2807
	return err;
L
Linus Torvalds 已提交
2808 2809 2810 2811 2812 2813 2814 2815 2816
}
EXPORT_SYMBOL(nobh_truncate_page);

int block_truncate_page(struct address_space *mapping,
			loff_t from, get_block_t *get_block)
{
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	unsigned blocksize;
2817
	sector_t iblock;
L
Linus Torvalds 已提交
2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831
	unsigned length, pos;
	struct inode *inode = mapping->host;
	struct page *page;
	struct buffer_head *bh;
	int err;

	blocksize = 1 << inode->i_blkbits;
	length = offset & (blocksize - 1);

	/* Block boundary? Nothing to do */
	if (!length)
		return 0;

	length = blocksize - length;
2832
	iblock = (sector_t)index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
L
Linus Torvalds 已提交
2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852
	
	page = grab_cache_page(mapping, index);
	err = -ENOMEM;
	if (!page)
		goto out;

	if (!page_has_buffers(page))
		create_empty_buffers(page, blocksize, 0);

	/* Find the buffer that contains "offset" */
	bh = page_buffers(page);
	pos = blocksize;
	while (offset >= pos) {
		bh = bh->b_this_page;
		iblock++;
		pos += blocksize;
	}

	err = 0;
	if (!buffer_mapped(bh)) {
2853
		WARN_ON(bh->b_size != blocksize);
L
Linus Torvalds 已提交
2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865
		err = get_block(inode, iblock, bh, 0);
		if (err)
			goto unlock;
		/* unmapped? It's a hole - nothing to do */
		if (!buffer_mapped(bh))
			goto unlock;
	}

	/* Ok, it's mapped. Make sure it's up-to-date */
	if (PageUptodate(page))
		set_buffer_uptodate(bh);

2866
	if (!buffer_uptodate(bh) && !buffer_delay(bh) && !buffer_unwritten(bh)) {
L
Linus Torvalds 已提交
2867 2868 2869 2870 2871 2872 2873 2874
		err = -EIO;
		ll_rw_block(READ, 1, &bh);
		wait_on_buffer(bh);
		/* Uhhuh. Read error. Complain and punt. */
		if (!buffer_uptodate(bh))
			goto unlock;
	}

2875
	zero_user(page, offset, length);
L
Linus Torvalds 已提交
2876 2877 2878 2879 2880 2881 2882 2883 2884
	mark_buffer_dirty(bh);
	err = 0;

unlock:
	unlock_page(page);
	page_cache_release(page);
out:
	return err;
}
2885
EXPORT_SYMBOL(block_truncate_page);
L
Linus Torvalds 已提交
2886 2887 2888 2889

/*
 * The generic ->writepage function for buffer-backed address_spaces
 */
2890 2891
int block_write_full_page(struct page *page, get_block_t *get_block,
			struct writeback_control *wbc)
L
Linus Torvalds 已提交
2892 2893 2894 2895 2896 2897 2898 2899
{
	struct inode * const inode = page->mapping->host;
	loff_t i_size = i_size_read(inode);
	const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
	unsigned offset;

	/* Is the page fully inside i_size? */
	if (page->index < end_index)
2900
		return __block_write_full_page(inode, page, get_block, wbc,
2901
					       end_buffer_async_write);
L
Linus Torvalds 已提交
2902 2903 2904 2905 2906 2907 2908 2909 2910

	/* Is the page fully outside i_size? (truncate in progress) */
	offset = i_size & (PAGE_CACHE_SIZE-1);
	if (page->index >= end_index+1 || !offset) {
		/*
		 * The page may have dirty, unmapped buffers.  For example,
		 * they may have been added in ext3_writepage().  Make them
		 * freeable here, so the page does not leak.
		 */
2911
		do_invalidatepage(page, 0, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
2912 2913 2914 2915 2916 2917
		unlock_page(page);
		return 0; /* don't care */
	}

	/*
	 * The page straddles i_size.  It must be zeroed out on each and every
2918
	 * writepage invocation because it may be mmapped.  "A file is mapped
L
Linus Torvalds 已提交
2919 2920 2921 2922
	 * in multiples of the page size.  For a file that is not a multiple of
	 * the  page size, the remaining memory is zeroed when mapped, and
	 * writes to that region are not written out to the file."
	 */
2923
	zero_user_segment(page, offset, PAGE_CACHE_SIZE);
2924 2925
	return __block_write_full_page(inode, page, get_block, wbc,
							end_buffer_async_write);
2926
}
2927
EXPORT_SYMBOL(block_write_full_page);
2928

L
Linus Torvalds 已提交
2929 2930 2931 2932 2933 2934 2935
sector_t generic_block_bmap(struct address_space *mapping, sector_t block,
			    get_block_t *get_block)
{
	struct buffer_head tmp;
	struct inode *inode = mapping->host;
	tmp.b_state = 0;
	tmp.b_blocknr = 0;
2936
	tmp.b_size = 1 << inode->i_blkbits;
L
Linus Torvalds 已提交
2937 2938 2939
	get_block(inode, block, &tmp, 0);
	return tmp.b_blocknr;
}
2940
EXPORT_SYMBOL(generic_block_bmap);
L
Linus Torvalds 已提交
2941

2942
static void end_bio_bh_io_sync(struct bio *bio, int err)
L
Linus Torvalds 已提交
2943 2944 2945 2946 2947 2948 2949
{
	struct buffer_head *bh = bio->bi_private;

	if (err == -EOPNOTSUPP) {
		set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
	}

2950 2951 2952
	if (unlikely (test_bit(BIO_QUIET,&bio->bi_flags)))
		set_bit(BH_Quiet, &bh->b_state);

L
Linus Torvalds 已提交
2953 2954 2955 2956
	bh->b_end_io(bh, test_bit(BIO_UPTODATE, &bio->bi_flags));
	bio_put(bio);
}

2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982
/*
 * This allows us to do IO even on the odd last sectors
 * of a device, even if the bh block size is some multiple
 * of the physical sector size.
 *
 * We'll just truncate the bio to the size of the device,
 * and clear the end of the buffer head manually.
 *
 * Truly out-of-range accesses will turn into actual IO
 * errors, this only handles the "we need to be able to
 * do IO at the final sector" case.
 */
static void guard_bh_eod(int rw, struct bio *bio, struct buffer_head *bh)
{
	sector_t maxsector;
	unsigned bytes;

	maxsector = i_size_read(bio->bi_bdev->bd_inode) >> 9;
	if (!maxsector)
		return;

	/*
	 * If the *whole* IO is past the end of the device,
	 * let it through, and the IO layer will turn it into
	 * an EIO.
	 */
2983
	if (unlikely(bio->bi_iter.bi_sector >= maxsector))
2984 2985
		return;

2986 2987
	maxsector -= bio->bi_iter.bi_sector;
	bytes = bio->bi_iter.bi_size;
2988 2989 2990 2991 2992 2993 2994
	if (likely((bytes >> 9) <= maxsector))
		return;

	/* Uhhuh. We've got a bh that straddles the device size! */
	bytes = maxsector << 9;

	/* Truncate the bio.. */
2995
	bio->bi_iter.bi_size = bytes;
2996 2997 2998
	bio->bi_io_vec[0].bv_len = bytes;

	/* ..and clear the end of the buffer for reads */
2999
	if ((rw & RW_MASK) == READ) {
3000 3001 3002
		void *kaddr = kmap_atomic(bh->b_page);
		memset(kaddr + bh_offset(bh) + bytes, 0, bh->b_size - bytes);
		kunmap_atomic(kaddr);
3003
		flush_dcache_page(bh->b_page);
3004 3005 3006
	}
}

3007
int _submit_bh(int rw, struct buffer_head *bh, unsigned long bio_flags)
L
Linus Torvalds 已提交
3008 3009 3010 3011 3012 3013 3014
{
	struct bio *bio;
	int ret = 0;

	BUG_ON(!buffer_locked(bh));
	BUG_ON(!buffer_mapped(bh));
	BUG_ON(!bh->b_end_io);
3015 3016
	BUG_ON(buffer_delay(bh));
	BUG_ON(buffer_unwritten(bh));
L
Linus Torvalds 已提交
3017 3018

	/*
3019
	 * Only clear out a write error when rewriting
L
Linus Torvalds 已提交
3020
	 */
3021
	if (test_set_buffer_req(bh) && (rw & WRITE))
L
Linus Torvalds 已提交
3022 3023 3024 3025 3026 3027 3028 3029
		clear_buffer_write_io_error(bh);

	/*
	 * from here on down, it's all bio -- do the initial mapping,
	 * submit_bio -> generic_make_request may further map this bio around
	 */
	bio = bio_alloc(GFP_NOIO, 1);

3030
	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
L
Linus Torvalds 已提交
3031 3032 3033 3034 3035 3036
	bio->bi_bdev = bh->b_bdev;
	bio->bi_io_vec[0].bv_page = bh->b_page;
	bio->bi_io_vec[0].bv_len = bh->b_size;
	bio->bi_io_vec[0].bv_offset = bh_offset(bh);

	bio->bi_vcnt = 1;
3037
	bio->bi_iter.bi_size = bh->b_size;
L
Linus Torvalds 已提交
3038 3039 3040

	bio->bi_end_io = end_bio_bh_io_sync;
	bio->bi_private = bh;
3041
	bio->bi_flags |= bio_flags;
L
Linus Torvalds 已提交
3042

3043 3044 3045
	/* Take care of bh's that straddle the end of the device */
	guard_bh_eod(rw, bio, bh);

3046 3047 3048 3049 3050
	if (buffer_meta(bh))
		rw |= REQ_META;
	if (buffer_prio(bh))
		rw |= REQ_PRIO;

L
Linus Torvalds 已提交
3051 3052 3053 3054 3055 3056 3057 3058 3059
	bio_get(bio);
	submit_bio(rw, bio);

	if (bio_flagged(bio, BIO_EOPNOTSUPP))
		ret = -EOPNOTSUPP;

	bio_put(bio);
	return ret;
}
3060 3061 3062 3063 3064 3065
EXPORT_SYMBOL_GPL(_submit_bh);

int submit_bh(int rw, struct buffer_head *bh)
{
	return _submit_bh(rw, bh, 0);
}
3066
EXPORT_SYMBOL(submit_bh);
L
Linus Torvalds 已提交
3067 3068 3069

/**
 * ll_rw_block: low-level access to block devices (DEPRECATED)
C
Christoph Hellwig 已提交
3070
 * @rw: whether to %READ or %WRITE or maybe %READA (readahead)
L
Linus Torvalds 已提交
3071 3072 3073
 * @nr: number of &struct buffer_heads in the array
 * @bhs: array of pointers to &struct buffer_head
 *
3074 3075
 * ll_rw_block() takes an array of pointers to &struct buffer_heads, and
 * requests an I/O operation on them, either a %READ or a %WRITE.  The third
C
Christoph Hellwig 已提交
3076 3077
 * %READA option is described in the documentation for generic_make_request()
 * which ll_rw_block() calls.
L
Linus Torvalds 已提交
3078 3079
 *
 * This function drops any buffer that it cannot get a lock on (with the
C
Christoph Hellwig 已提交
3080 3081 3082 3083 3084
 * BH_Lock state bit), any buffer that appears to be clean when doing a write
 * request, and any buffer that appears to be up-to-date when doing read
 * request.  Further it marks as clean buffers that are processed for
 * writing (the buffer cache won't assume that they are actually clean
 * until the buffer gets unlocked).
L
Linus Torvalds 已提交
3085 3086
 *
 * ll_rw_block sets b_end_io to simple completion handler that marks
3087
 * the buffer up-to-date (if appropriate), unlocks the buffer and wakes
L
Linus Torvalds 已提交
3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099
 * any waiters. 
 *
 * All of the buffers must be for the same device, and must also be a
 * multiple of the current approved size for the device.
 */
void ll_rw_block(int rw, int nr, struct buffer_head *bhs[])
{
	int i;

	for (i = 0; i < nr; i++) {
		struct buffer_head *bh = bhs[i];

C
Christoph Hellwig 已提交
3100
		if (!trylock_buffer(bh))
L
Linus Torvalds 已提交
3101
			continue;
C
Christoph Hellwig 已提交
3102
		if (rw == WRITE) {
L
Linus Torvalds 已提交
3103
			if (test_clear_buffer_dirty(bh)) {
3104
				bh->b_end_io = end_buffer_write_sync;
3105
				get_bh(bh);
C
Christoph Hellwig 已提交
3106
				submit_bh(WRITE, bh);
L
Linus Torvalds 已提交
3107 3108 3109 3110
				continue;
			}
		} else {
			if (!buffer_uptodate(bh)) {
3111
				bh->b_end_io = end_buffer_read_sync;
3112
				get_bh(bh);
L
Linus Torvalds 已提交
3113 3114 3115 3116 3117 3118 3119
				submit_bh(rw, bh);
				continue;
			}
		}
		unlock_buffer(bh);
	}
}
3120
EXPORT_SYMBOL(ll_rw_block);
L
Linus Torvalds 已提交
3121

C
Christoph Hellwig 已提交
3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134
void write_dirty_buffer(struct buffer_head *bh, int rw)
{
	lock_buffer(bh);
	if (!test_clear_buffer_dirty(bh)) {
		unlock_buffer(bh);
		return;
	}
	bh->b_end_io = end_buffer_write_sync;
	get_bh(bh);
	submit_bh(rw, bh);
}
EXPORT_SYMBOL(write_dirty_buffer);

L
Linus Torvalds 已提交
3135 3136 3137 3138 3139
/*
 * For a data-integrity writeout, we need to wait upon any in-progress I/O
 * and then start new I/O and then wait upon it.  The caller must have a ref on
 * the buffer_head.
 */
C
Christoph Hellwig 已提交
3140
int __sync_dirty_buffer(struct buffer_head *bh, int rw)
L
Linus Torvalds 已提交
3141 3142 3143 3144 3145 3146 3147 3148
{
	int ret = 0;

	WARN_ON(atomic_read(&bh->b_count) < 1);
	lock_buffer(bh);
	if (test_clear_buffer_dirty(bh)) {
		get_bh(bh);
		bh->b_end_io = end_buffer_write_sync;
C
Christoph Hellwig 已提交
3149
		ret = submit_bh(rw, bh);
L
Linus Torvalds 已提交
3150 3151 3152 3153 3154 3155 3156 3157
		wait_on_buffer(bh);
		if (!ret && !buffer_uptodate(bh))
			ret = -EIO;
	} else {
		unlock_buffer(bh);
	}
	return ret;
}
C
Christoph Hellwig 已提交
3158 3159 3160 3161 3162 3163
EXPORT_SYMBOL(__sync_dirty_buffer);

int sync_dirty_buffer(struct buffer_head *bh)
{
	return __sync_dirty_buffer(bh, WRITE_SYNC);
}
3164
EXPORT_SYMBOL(sync_dirty_buffer);
L
Linus Torvalds 已提交
3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199

/*
 * try_to_free_buffers() checks if all the buffers on this particular page
 * are unused, and releases them if so.
 *
 * Exclusion against try_to_free_buffers may be obtained by either
 * locking the page or by holding its mapping's private_lock.
 *
 * If the page is dirty but all the buffers are clean then we need to
 * be sure to mark the page clean as well.  This is because the page
 * may be against a block device, and a later reattachment of buffers
 * to a dirty page will set *all* buffers dirty.  Which would corrupt
 * filesystem data on the same device.
 *
 * The same applies to regular filesystem pages: if all the buffers are
 * clean then we set the page clean and proceed.  To do that, we require
 * total exclusion from __set_page_dirty_buffers().  That is obtained with
 * private_lock.
 *
 * try_to_free_buffers() is non-blocking.
 */
static inline int buffer_busy(struct buffer_head *bh)
{
	return atomic_read(&bh->b_count) |
		(bh->b_state & ((1 << BH_Dirty) | (1 << BH_Lock)));
}

static int
drop_buffers(struct page *page, struct buffer_head **buffers_to_free)
{
	struct buffer_head *head = page_buffers(page);
	struct buffer_head *bh;

	bh = head;
	do {
A
akpm@osdl.org 已提交
3200
		if (buffer_write_io_error(bh) && page->mapping)
L
Linus Torvalds 已提交
3201 3202 3203 3204 3205 3206 3207 3208 3209
			set_bit(AS_EIO, &page->mapping->flags);
		if (buffer_busy(bh))
			goto failed;
		bh = bh->b_this_page;
	} while (bh != head);

	do {
		struct buffer_head *next = bh->b_this_page;

3210
		if (bh->b_assoc_map)
L
Linus Torvalds 已提交
3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227
			__remove_assoc_queue(bh);
		bh = next;
	} while (bh != head);
	*buffers_to_free = head;
	__clear_page_buffers(page);
	return 1;
failed:
	return 0;
}

int try_to_free_buffers(struct page *page)
{
	struct address_space * const mapping = page->mapping;
	struct buffer_head *buffers_to_free = NULL;
	int ret = 0;

	BUG_ON(!PageLocked(page));
3228
	if (PageWriteback(page))
L
Linus Torvalds 已提交
3229 3230 3231 3232 3233 3234 3235 3236 3237
		return 0;

	if (mapping == NULL) {		/* can this still happen? */
		ret = drop_buffers(page, &buffers_to_free);
		goto out;
	}

	spin_lock(&mapping->private_lock);
	ret = drop_buffers(page, &buffers_to_free);
3238 3239 3240 3241 3242 3243 3244 3245 3246 3247

	/*
	 * If the filesystem writes its buffers by hand (eg ext3)
	 * then we can have clean buffers against a dirty page.  We
	 * clean the page here; otherwise the VM will never notice
	 * that the filesystem did any IO at all.
	 *
	 * Also, during truncate, discard_buffer will have marked all
	 * the page's buffers clean.  We discover that here and clean
	 * the page also.
3248 3249 3250 3251
	 *
	 * private_lock must be held over this entire operation in order
	 * to synchronise against __set_page_dirty_buffers and prevent the
	 * dirty bit from being lost.
3252 3253 3254
	 */
	if (ret)
		cancel_dirty_page(page, PAGE_CACHE_SIZE);
3255
	spin_unlock(&mapping->private_lock);
L
Linus Torvalds 已提交
3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274
out:
	if (buffers_to_free) {
		struct buffer_head *bh = buffers_to_free;

		do {
			struct buffer_head *next = bh->b_this_page;
			free_buffer_head(bh);
			bh = next;
		} while (bh != buffers_to_free);
	}
	return ret;
}
EXPORT_SYMBOL(try_to_free_buffers);

/*
 * There are no bdflush tunables left.  But distributions are
 * still running obsolete flush daemons, so we terminate them here.
 *
 * Use of bdflush() is deprecated and will be removed in a future kernel.
3275
 * The `flush-X' kernel threads fully replace bdflush daemons and this call.
L
Linus Torvalds 已提交
3276
 */
3277
SYSCALL_DEFINE2(bdflush, int, func, long, data)
L
Linus Torvalds 已提交
3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299
{
	static int msg_count;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

	if (msg_count < 5) {
		msg_count++;
		printk(KERN_INFO
			"warning: process `%s' used the obsolete bdflush"
			" system call\n", current->comm);
		printk(KERN_INFO "Fix your initscripts?\n");
	}

	if (func == 1)
		do_exit(0);
	return 0;
}

/*
 * Buffer-head allocation
 */
3300
static struct kmem_cache *bh_cachep __read_mostly;
L
Linus Torvalds 已提交
3301 3302 3303 3304 3305

/*
 * Once the number of bh's in the machine exceeds this level, we start
 * stripping them in writeback.
 */
3306
static unsigned long max_buffer_heads;
L
Linus Torvalds 已提交
3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321

int buffer_heads_over_limit;

struct bh_accounting {
	int nr;			/* Number of live bh's */
	int ratelimit;		/* Limit cacheline bouncing */
};

static DEFINE_PER_CPU(struct bh_accounting, bh_accounting) = {0, 0};

static void recalc_bh_state(void)
{
	int i;
	int tot = 0;

3322
	if (__this_cpu_inc_return(bh_accounting.ratelimit) - 1 < 4096)
L
Linus Torvalds 已提交
3323
		return;
3324
	__this_cpu_write(bh_accounting.ratelimit, 0);
3325
	for_each_online_cpu(i)
L
Linus Torvalds 已提交
3326 3327 3328
		tot += per_cpu(bh_accounting, i).nr;
	buffer_heads_over_limit = (tot > max_buffer_heads);
}
3329

A
Al Viro 已提交
3330
struct buffer_head *alloc_buffer_head(gfp_t gfp_flags)
L
Linus Torvalds 已提交
3331
{
3332
	struct buffer_head *ret = kmem_cache_zalloc(bh_cachep, gfp_flags);
L
Linus Torvalds 已提交
3333
	if (ret) {
C
Christoph Lameter 已提交
3334
		INIT_LIST_HEAD(&ret->b_assoc_buffers);
3335 3336
		preempt_disable();
		__this_cpu_inc(bh_accounting.nr);
L
Linus Torvalds 已提交
3337
		recalc_bh_state();
3338
		preempt_enable();
L
Linus Torvalds 已提交
3339 3340 3341 3342 3343 3344 3345 3346 3347
	}
	return ret;
}
EXPORT_SYMBOL(alloc_buffer_head);

void free_buffer_head(struct buffer_head *bh)
{
	BUG_ON(!list_empty(&bh->b_assoc_buffers));
	kmem_cache_free(bh_cachep, bh);
3348 3349
	preempt_disable();
	__this_cpu_dec(bh_accounting.nr);
L
Linus Torvalds 已提交
3350
	recalc_bh_state();
3351
	preempt_enable();
L
Linus Torvalds 已提交
3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363
}
EXPORT_SYMBOL(free_buffer_head);

static void buffer_exit_cpu(int cpu)
{
	int i;
	struct bh_lru *b = &per_cpu(bh_lrus, cpu);

	for (i = 0; i < BH_LRU_SIZE; i++) {
		brelse(b->bhs[i]);
		b->bhs[i] = NULL;
	}
3364
	this_cpu_add(bh_accounting.nr, per_cpu(bh_accounting, cpu).nr);
3365
	per_cpu(bh_accounting, cpu).nr = 0;
L
Linus Torvalds 已提交
3366 3367 3368 3369 3370
}

static int buffer_cpu_notify(struct notifier_block *self,
			      unsigned long action, void *hcpu)
{
3371
	if (action == CPU_DEAD || action == CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
3372 3373 3374 3375
		buffer_exit_cpu((unsigned long)hcpu);
	return NOTIFY_OK;
}

3376
/**
3377
 * bh_uptodate_or_lock - Test whether the buffer is uptodate
3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395
 * @bh: struct buffer_head
 *
 * Return true if the buffer is up-to-date and false,
 * with the buffer locked, if not.
 */
int bh_uptodate_or_lock(struct buffer_head *bh)
{
	if (!buffer_uptodate(bh)) {
		lock_buffer(bh);
		if (!buffer_uptodate(bh))
			return 0;
		unlock_buffer(bh);
	}
	return 1;
}
EXPORT_SYMBOL(bh_uptodate_or_lock);

/**
3396
 * bh_submit_read - Submit a locked buffer for reading
3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419
 * @bh: struct buffer_head
 *
 * Returns zero on success and -EIO on error.
 */
int bh_submit_read(struct buffer_head *bh)
{
	BUG_ON(!buffer_locked(bh));

	if (buffer_uptodate(bh)) {
		unlock_buffer(bh);
		return 0;
	}

	get_bh(bh);
	bh->b_end_io = end_buffer_read_sync;
	submit_bh(READ, bh);
	wait_on_buffer(bh);
	if (buffer_uptodate(bh))
		return 0;
	return -EIO;
}
EXPORT_SYMBOL(bh_submit_read);

L
Linus Torvalds 已提交
3420 3421
void __init buffer_init(void)
{
3422
	unsigned long nrpages;
L
Linus Torvalds 已提交
3423

3424 3425 3426 3427
	bh_cachep = kmem_cache_create("buffer_head",
			sizeof(struct buffer_head), 0,
				(SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
				SLAB_MEM_SPREAD),
3428
				NULL);
L
Linus Torvalds 已提交
3429 3430 3431 3432 3433 3434 3435 3436

	/*
	 * Limit the bh occupancy to 10% of ZONE_NORMAL
	 */
	nrpages = (nr_free_buffer_pages() * 10) / 100;
	max_buffer_heads = nrpages * (PAGE_SIZE / sizeof(struct buffer_head));
	hotcpu_notifier(buffer_cpu_notify, 0);
}