bitmap.c 43.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
/*
 * bitmap.c two-level bitmap (C) Peter T. Breuer (ptb@ot.uc3m.es) 2003
 *
 * bitmap_create  - sets up the bitmap structure
 * bitmap_destroy - destroys the bitmap structure
 *
 * additions, Copyright (C) 2003-2004, Paul Clements, SteelEye Technology, Inc.:
 * - added disk storage for bitmap
 * - changes to allow various bitmap chunk sizes
 */

/*
 * Still to do:
 *
 * flush after percent set rather than just time based. (maybe both).
 * wait if count gets too high, wake when it drops to half.
 */

#include <linux/module.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/timer.h>
#include <linux/sched.h>
#include <linux/list.h>
#include <linux/file.h>
#include <linux/mount.h>
#include <linux/buffer_head.h>
#include <linux/raid/md.h>
#include <linux/raid/bitmap.h>

/* debug macros */

#define DEBUG 0

#if DEBUG
/* these are for debugging purposes only! */

/* define one and only one of these */
#define INJECT_FAULTS_1 0 /* cause bitmap_alloc_page to fail always */
#define INJECT_FAULTS_2 0 /* cause bitmap file to be kicked when first bit set*/
#define INJECT_FAULTS_3 0 /* treat bitmap file as kicked at init time */
#define INJECT_FAULTS_4 0 /* undef */
#define INJECT_FAULTS_5 0 /* undef */
#define INJECT_FAULTS_6 0

/* if these are defined, the driver will fail! debug only */
#define INJECT_FATAL_FAULT_1 0 /* fail kmalloc, causing bitmap_create to fail */
#define INJECT_FATAL_FAULT_2 0 /* undef */
#define INJECT_FATAL_FAULT_3 0 /* undef */
#endif

//#define DPRINTK PRINTK /* set this NULL to avoid verbose debug output */
#define DPRINTK(x...) do { } while(0)

#ifndef PRINTK
#  if DEBUG > 0
#    define PRINTK(x...) printk(KERN_DEBUG x)
#  else
#    define PRINTK(x...)
#  endif
#endif

static inline char * bmname(struct bitmap *bitmap)
{
	return bitmap->mddev ? mdname(bitmap->mddev) : "mdX";
}


/*
 * just a placeholder - calls kmalloc for bitmap pages
 */
static unsigned char *bitmap_alloc_page(struct bitmap *bitmap)
{
	unsigned char *page;

77
#ifdef INJECT_FAULTS_1
78 79 80 81 82 83 84
	page = NULL;
#else
	page = kmalloc(PAGE_SIZE, GFP_NOIO);
#endif
	if (!page)
		printk("%s: bitmap_alloc_page FAILED\n", bmname(bitmap));
	else
85
		PRINTK("%s: bitmap_alloc_page: allocated page at %p\n",
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
			bmname(bitmap), page);
	return page;
}

/*
 * for now just a placeholder -- just calls kfree for bitmap pages
 */
static void bitmap_free_page(struct bitmap *bitmap, unsigned char *page)
{
	PRINTK("%s: bitmap_free_page: free page %p\n", bmname(bitmap), page);
	kfree(page);
}

/*
 * check a page and, if necessary, allocate it (or hijack it if the alloc fails)
 *
 * 1) check to see if this page is allocated, if it's not then try to alloc
 * 2) if the alloc fails, set the page's hijacked flag so we'll use the
 *    page pointer directly as a counter
 *
 * if we find our page, we increment the page's refcount so that it stays
 * allocated while we're using it
 */
static int bitmap_checkpage(struct bitmap *bitmap, unsigned long page, int create)
{
	unsigned char *mappage;

	if (page >= bitmap->pages) {
114 115 116 117
		/* This can happen if bitmap_start_sync goes beyond
		 * End-of-device while looking for a whole page.
		 * It is harmless.
		 */
118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
		return -EINVAL;
	}


	if (bitmap->bp[page].hijacked) /* it's hijacked, don't try to alloc */
		return 0;

	if (bitmap->bp[page].map) /* page is already allocated, just return */
		return 0;

	if (!create)
		return -ENOENT;

	spin_unlock_irq(&bitmap->lock);

	/* this page has not been allocated yet */

	if ((mappage = bitmap_alloc_page(bitmap)) == NULL) {
		PRINTK("%s: bitmap map page allocation failed, hijacking\n",
			bmname(bitmap));
		/* failed - set the hijacked flag so that we can use the
		 * pointer as a counter */
		spin_lock_irq(&bitmap->lock);
		if (!bitmap->bp[page].map)
			bitmap->bp[page].hijacked = 1;
		goto out;
	}

	/* got a page */

	spin_lock_irq(&bitmap->lock);

	/* recheck the page */

	if (bitmap->bp[page].map || bitmap->bp[page].hijacked) {
		/* somebody beat us to getting the page */
		bitmap_free_page(bitmap, mappage);
		return 0;
	}

	/* no page was in place and we have one, so install it */

	memset(mappage, 0, PAGE_SIZE);
	bitmap->bp[page].map = mappage;
	bitmap->missing_pages--;
out:
	return 0;
}


/* if page is completely empty, put it back on the free list, or dealloc it */
/* if page was hijacked, unmark the flag so it might get alloced next time */
/* Note: lock should be held when calling this */
171
static void bitmap_checkfree(struct bitmap *bitmap, unsigned long page)
172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210
{
	char *ptr;

	if (bitmap->bp[page].count) /* page is still busy */
		return;

	/* page is no longer in use, it can be released */

	if (bitmap->bp[page].hijacked) { /* page was hijacked, undo this now */
		bitmap->bp[page].hijacked = 0;
		bitmap->bp[page].map = NULL;
		return;
	}

	/* normal case, free the page */

#if 0
/* actually ... let's not.  We will probably need the page again exactly when
 * memory is tight and we are flusing to disk
 */
	return;
#else
	ptr = bitmap->bp[page].map;
	bitmap->bp[page].map = NULL;
	bitmap->missing_pages++;
	bitmap_free_page(bitmap, ptr);
	return;
#endif
}


/*
 * bitmap file handling - read and write the bitmap file and its superblock
 */

/*
 * basic page I/O operations
 */

211
/* IO operations when bitmap is stored near all superblocks */
212 213 214
static struct page *read_sb_page(mddev_t *mddev, long offset,
				 struct page *page,
				 unsigned long index, int size)
215 216 217 218 219 220
{
	/* choose a good rdev and read the page from there */

	mdk_rdev_t *rdev;
	sector_t target;

221 222
	if (!page)
		page = alloc_page(GFP_KERNEL);
223 224 225
	if (!page)
		return ERR_PTR(-ENOMEM);

226
	list_for_each_entry(rdev, &mddev->disks, same_set) {
227 228
		if (! test_bit(In_sync, &rdev->flags)
		    || test_bit(Faulty, &rdev->flags))
229 230
			continue;

231
		target = rdev->sb_start + offset + index * (PAGE_SIZE/512);
232

233 234 235
		if (sync_page_io(rdev->bdev, target,
				 roundup(size, bdev_hardsect_size(rdev->bdev)),
				 page, READ)) {
236
			page->index = index;
237 238
			attach_page_buffers(page, NULL); /* so that free_buffer will
							  * quietly no-op */
239 240 241 242
			return page;
		}
	}
	return ERR_PTR(-EIO);
243 244 245

}

246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
static mdk_rdev_t *next_active_rdev(mdk_rdev_t *rdev, mddev_t *mddev)
{
	/* Iterate the disks of an mddev, using rcu to protect access to the
	 * linked list, and raising the refcount of devices we return to ensure
	 * they don't disappear while in use.
	 * As devices are only added or removed when raid_disk is < 0 and
	 * nr_pending is 0 and In_sync is clear, the entries we return will
	 * still be in the same position on the list when we re-enter
	 * list_for_each_continue_rcu.
	 */
	struct list_head *pos;
	rcu_read_lock();
	if (rdev == NULL)
		/* start at the beginning */
		pos = &mddev->disks;
	else {
		/* release the previous rdev and start from there. */
		rdev_dec_pending(rdev, mddev);
		pos = &rdev->same_set;
	}
	list_for_each_continue_rcu(pos, &mddev->disks) {
		rdev = list_entry(pos, mdk_rdev_t, same_set);
		if (rdev->raid_disk >= 0 &&
		    !test_bit(Faulty, &rdev->flags)) {
			/* this is a usable devices */
			atomic_inc(&rdev->nr_pending);
			rcu_read_unlock();
			return rdev;
		}
	}
	rcu_read_unlock();
	return NULL;
}

280
static int write_sb_page(struct bitmap *bitmap, struct page *page, int wait)
281
{
282
	mdk_rdev_t *rdev = NULL;
283
	mddev_t *mddev = bitmap->mddev;
284

285
	while ((rdev = next_active_rdev(rdev, mddev)) != NULL) {
286 287 288 289
			int size = PAGE_SIZE;
			if (page->index == bitmap->file_pages-1)
				size = roundup(bitmap->last_page_size,
					       bdev_hardsect_size(rdev->bdev));
290 291 292 293 294 295
			/* Just make sure we aren't corrupting data or
			 * metadata
			 */
			if (bitmap->offset < 0) {
				/* DATA  BITMAP METADATA  */
				if (bitmap->offset
296
				    + (long)(page->index * (PAGE_SIZE/512))
297 298
				    + size/512 > 0)
					/* bitmap runs in to metadata */
299
					goto bad_alignment;
300
				if (rdev->data_offset + mddev->size*2
301
				    > rdev->sb_start + bitmap->offset)
302
					/* data runs in to bitmap */
303
					goto bad_alignment;
304
			} else if (rdev->sb_start < rdev->data_offset) {
305
				/* METADATA BITMAP DATA */
306
				if (rdev->sb_start
307 308 309 310
				    + bitmap->offset
				    + page->index*(PAGE_SIZE/512) + size/512
				    > rdev->data_offset)
					/* bitmap runs in to data */
311
					goto bad_alignment;
312 313 314
			} else {
				/* DATA METADATA BITMAP - no problems */
			}
315
			md_super_write(mddev, rdev,
316
				       rdev->sb_start + bitmap->offset
317
				       + page->index * (PAGE_SIZE/512),
318
				       size,
319
				       page);
320
	}
321 322

	if (wait)
323
		md_super_wait(mddev);
324
	return 0;
325 326 327 328

 bad_alignment:
	rcu_read_unlock();
	return -EINVAL;
329 330
}

331
static void bitmap_file_kick(struct bitmap *bitmap);
332
/*
333
 * write out a page to a file
334
 */
335
static void write_page(struct bitmap *bitmap, struct page *page, int wait)
336
{
337
	struct buffer_head *bh;
338

339 340 341 342 343
	if (bitmap->file == NULL) {
		switch (write_sb_page(bitmap, page, wait)) {
		case -EINVAL:
			bitmap->flags |= BITMAP_WRITE_ERROR;
		}
344
	} else {
345

346
		bh = page_buffers(page);
347

348 349 350 351 352 353 354
		while (bh && bh->b_blocknr) {
			atomic_inc(&bitmap->pending_writes);
			set_buffer_locked(bh);
			set_buffer_mapped(bh);
			submit_bh(WRITE, bh);
			bh = bh->b_this_page;
		}
355

356 357 358 359
		if (wait) {
			wait_event(bitmap->write_wait,
				   atomic_read(&bitmap->pending_writes)==0);
		}
360
	}
361 362
	if (bitmap->flags & BITMAP_WRITE_ERROR)
		bitmap_file_kick(bitmap);
363 364 365 366 367 368
}

static void end_bitmap_write(struct buffer_head *bh, int uptodate)
{
	struct bitmap *bitmap = bh->b_private;
	unsigned long flags;
369

370 371 372 373
	if (!uptodate) {
		spin_lock_irqsave(&bitmap->lock, flags);
		bitmap->flags |= BITMAP_WRITE_ERROR;
		spin_unlock_irqrestore(&bitmap->lock, flags);
374
	}
375 376 377
	if (atomic_dec_and_test(&bitmap->pending_writes))
		wake_up(&bitmap->write_wait);
}
378

379 380 381 382 383 384 385 386 387 388 389
/* copied from buffer.c */
static void
__clear_page_buffers(struct page *page)
{
	ClearPagePrivate(page);
	set_page_private(page, 0);
	page_cache_release(page);
}
static void free_buffers(struct page *page)
{
	struct buffer_head *bh = page_buffers(page);
390

391 392 393 394
	while (bh) {
		struct buffer_head *next = bh->b_this_page;
		free_buffer_head(bh);
		bh = next;
395
	}
396 397
	__clear_page_buffers(page);
	put_page(page);
398 399
}

400 401 402 403 404 405 406
/* read a page from a file.
 * We both read the page, and attach buffers to the page to record the
 * address of each block (using bmap).  These addresses will be used
 * to write the block later, completely bypassing the filesystem.
 * This usage is similar to how swap files are handled, and allows us
 * to write to a file with no concerns of memory allocation failing.
 */
407
static struct page *read_page(struct file *file, unsigned long index,
408 409
			      struct bitmap *bitmap,
			      unsigned long count)
410 411
{
	struct page *page = NULL;
J
Josef Sipek 已提交
412
	struct inode *inode = file->f_path.dentry->d_inode;
413 414
	struct buffer_head *bh;
	sector_t block;
415

416 417
	PRINTK("read bitmap file (%dB @ %Lu)\n", (int)PAGE_SIZE,
			(unsigned long long)index << PAGE_SHIFT);
418

419 420 421
	page = alloc_page(GFP_KERNEL);
	if (!page)
		page = ERR_PTR(-ENOMEM);
422 423
	if (IS_ERR(page))
		goto out;
424 425 426

	bh = alloc_page_buffers(page, 1<<inode->i_blkbits, 0);
	if (!bh) {
427
		put_page(page);
428
		page = ERR_PTR(-ENOMEM);
429 430
		goto out;
	}
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
	attach_page_buffers(page, bh);
	block = index << (PAGE_SHIFT - inode->i_blkbits);
	while (bh) {
		if (count == 0)
			bh->b_blocknr = 0;
		else {
			bh->b_blocknr = bmap(inode, block);
			if (bh->b_blocknr == 0) {
				/* Cannot use this file! */
				free_buffers(page);
				page = ERR_PTR(-EINVAL);
				goto out;
			}
			bh->b_bdev = inode->i_sb->s_bdev;
			if (count < (1<<inode->i_blkbits))
				count = 0;
			else
				count -= (1<<inode->i_blkbits);

			bh->b_end_io = end_bitmap_write;
			bh->b_private = bitmap;
452 453 454 455
			atomic_inc(&bitmap->pending_writes);
			set_buffer_locked(bh);
			set_buffer_mapped(bh);
			submit_bh(READ, bh);
456 457 458 459 460
		}
		block++;
		bh = bh->b_this_page;
	}
	page->index = index;
461 462 463 464 465 466 467

	wait_event(bitmap->write_wait,
		   atomic_read(&bitmap->pending_writes)==0);
	if (bitmap->flags & BITMAP_WRITE_ERROR) {
		free_buffers(page);
		page = ERR_PTR(-EIO);
	}
468 469 470
out:
	if (IS_ERR(page))
		printk(KERN_ALERT "md: bitmap read error: (%dB @ %Lu): %ld\n",
471 472
			(int)PAGE_SIZE,
			(unsigned long long)index << PAGE_SHIFT,
473 474 475 476 477 478 479 480 481
			PTR_ERR(page));
	return page;
}

/*
 * bitmap file superblock operations
 */

/* update the event counter and sync the superblock to disk */
482
void bitmap_update_sb(struct bitmap *bitmap)
483 484 485 486 487
{
	bitmap_super_t *sb;
	unsigned long flags;

	if (!bitmap || !bitmap->mddev) /* no bitmap for this array */
488
		return;
489 490 491
	spin_lock_irqsave(&bitmap->lock, flags);
	if (!bitmap->sb_page) { /* no superblock */
		spin_unlock_irqrestore(&bitmap->lock, flags);
492
		return;
493 494
	}
	spin_unlock_irqrestore(&bitmap->lock, flags);
495
	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
496
	sb->events = cpu_to_le64(bitmap->mddev->events);
497 498 499 500 501
	if (bitmap->mddev->events < bitmap->events_cleared) {
		/* rocking back to read-only */
		bitmap->events_cleared = bitmap->mddev->events;
		sb->events_cleared = cpu_to_le64(bitmap->events_cleared);
	}
502
	kunmap_atomic(sb, KM_USER0);
503
	write_page(bitmap, bitmap->sb_page, 1);
504 505 506 507 508 509 510 511 512
}

/* print out the bitmap file superblock */
void bitmap_print_sb(struct bitmap *bitmap)
{
	bitmap_super_t *sb;

	if (!bitmap || !bitmap->sb_page)
		return;
513
	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
514
	printk(KERN_DEBUG "%s: bitmap file superblock:\n", bmname(bitmap));
515 516 517
	printk(KERN_DEBUG "         magic: %08x\n", le32_to_cpu(sb->magic));
	printk(KERN_DEBUG "       version: %d\n", le32_to_cpu(sb->version));
	printk(KERN_DEBUG "          uuid: %08x.%08x.%08x.%08x\n",
518 519 520 521
					*(__u32 *)(sb->uuid+0),
					*(__u32 *)(sb->uuid+4),
					*(__u32 *)(sb->uuid+8),
					*(__u32 *)(sb->uuid+12));
522
	printk(KERN_DEBUG "        events: %llu\n",
523
			(unsigned long long) le64_to_cpu(sb->events));
524
	printk(KERN_DEBUG "events cleared: %llu\n",
525
			(unsigned long long) le64_to_cpu(sb->events_cleared));
526 527 528 529 530
	printk(KERN_DEBUG "         state: %08x\n", le32_to_cpu(sb->state));
	printk(KERN_DEBUG "     chunksize: %d B\n", le32_to_cpu(sb->chunksize));
	printk(KERN_DEBUG "  daemon sleep: %ds\n", le32_to_cpu(sb->daemon_sleep));
	printk(KERN_DEBUG "     sync size: %llu KB\n",
			(unsigned long long)le64_to_cpu(sb->sync_size)/2);
531
	printk(KERN_DEBUG "max write behind: %d\n", le32_to_cpu(sb->write_behind));
532
	kunmap_atomic(sb, KM_USER0);
533 534 535 536 537 538 539
}

/* read the superblock from the bitmap file and initialize some bitmap fields */
static int bitmap_read_sb(struct bitmap *bitmap)
{
	char *reason = NULL;
	bitmap_super_t *sb;
540
	unsigned long chunksize, daemon_sleep, write_behind;
541 542 543 544
	unsigned long long events;
	int err = -EINVAL;

	/* page 0 is the superblock, read it... */
545 546 547 548 549 550
	if (bitmap->file) {
		loff_t isize = i_size_read(bitmap->file->f_mapping->host);
		int bytes = isize > PAGE_SIZE ? PAGE_SIZE : isize;

		bitmap->sb_page = read_page(bitmap->file, 0, bitmap, bytes);
	} else {
551 552 553
		bitmap->sb_page = read_sb_page(bitmap->mddev, bitmap->offset,
					       NULL,
					       0, sizeof(bitmap_super_t));
554
	}
555 556 557 558 559 560
	if (IS_ERR(bitmap->sb_page)) {
		err = PTR_ERR(bitmap->sb_page);
		bitmap->sb_page = NULL;
		return err;
	}

561
	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
562 563 564

	chunksize = le32_to_cpu(sb->chunksize);
	daemon_sleep = le32_to_cpu(sb->daemon_sleep);
565
	write_behind = le32_to_cpu(sb->write_behind);
566 567 568 569

	/* verify that the bitmap-specific fields are valid */
	if (sb->magic != cpu_to_le32(BITMAP_MAGIC))
		reason = "bad magic";
570 571
	else if (le32_to_cpu(sb->version) < BITMAP_MAJOR_LO ||
		 le32_to_cpu(sb->version) > BITMAP_MAJOR_HI)
572
		reason = "unrecognized superblock version";
573
	else if (chunksize < 512)
574
		reason = "bitmap chunksize too small";
575 576
	else if ((1 << ffz(~chunksize)) != chunksize)
		reason = "bitmap chunksize not a power of 2";
577 578
	else if (daemon_sleep < 1 || daemon_sleep > MAX_SCHEDULE_TIMEOUT / HZ)
		reason = "daemon sleep period out of range";
579 580
	else if (write_behind > COUNTER_MAX)
		reason = "write-behind limit out of range (0 - 16383)";
581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606
	if (reason) {
		printk(KERN_INFO "%s: invalid bitmap file superblock: %s\n",
			bmname(bitmap), reason);
		goto out;
	}

	/* keep the array size field of the bitmap superblock up to date */
	sb->sync_size = cpu_to_le64(bitmap->mddev->resync_max_sectors);

	if (!bitmap->mddev->persistent)
		goto success;

	/*
	 * if we have a persistent array superblock, compare the
	 * bitmap's UUID and event counter to the mddev's
	 */
	if (memcmp(sb->uuid, bitmap->mddev->uuid, 16)) {
		printk(KERN_INFO "%s: bitmap superblock UUID mismatch\n",
			bmname(bitmap));
		goto out;
	}
	events = le64_to_cpu(sb->events);
	if (events < bitmap->mddev->events) {
		printk(KERN_INFO "%s: bitmap file is out of date (%llu < %llu) "
			"-- forcing full recovery\n", bmname(bitmap), events,
			(unsigned long long) bitmap->mddev->events);
607
		sb->state |= cpu_to_le32(BITMAP_STALE);
608 609 610 611 612
	}
success:
	/* assign fields using values from superblock */
	bitmap->chunksize = chunksize;
	bitmap->daemon_sleep = daemon_sleep;
613
	bitmap->daemon_lastrun = jiffies;
614
	bitmap->max_write_behind = write_behind;
615
	bitmap->flags |= le32_to_cpu(sb->state);
616 617
	if (le32_to_cpu(sb->version) == BITMAP_MAJOR_HOSTENDIAN)
		bitmap->flags |= BITMAP_HOSTENDIAN;
618
	bitmap->events_cleared = le64_to_cpu(sb->events_cleared);
619
	if (sb->state & cpu_to_le32(BITMAP_STALE))
620
		bitmap->events_cleared = bitmap->mddev->events;
621 622
	err = 0;
out:
623
	kunmap_atomic(sb, KM_USER0);
624 625 626 627 628 629 630 631 632 633
	if (err)
		bitmap_print_sb(bitmap);
	return err;
}

enum bitmap_mask_op {
	MASK_SET,
	MASK_UNSET
};

634 635 636
/* record the state of the bitmap in the superblock.  Return the old value */
static int bitmap_mask_state(struct bitmap *bitmap, enum bitmap_state bits,
			     enum bitmap_mask_op op)
637 638 639
{
	bitmap_super_t *sb;
	unsigned long flags;
640
	int old;
641 642

	spin_lock_irqsave(&bitmap->lock, flags);
643
	if (!bitmap->sb_page) { /* can't set the state */
644
		spin_unlock_irqrestore(&bitmap->lock, flags);
645
		return 0;
646 647
	}
	spin_unlock_irqrestore(&bitmap->lock, flags);
648
	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
649
	old = le32_to_cpu(sb->state) & bits;
650
	switch (op) {
651
		case MASK_SET: sb->state |= cpu_to_le32(bits);
652
				break;
653
		case MASK_UNSET: sb->state &= cpu_to_le32(~bits);
654 655 656
				break;
		default: BUG();
	}
657
	kunmap_atomic(sb, KM_USER0);
658
	return old;
659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686
}

/*
 * general bitmap file operations
 */

/* calculate the index of the page that contains this bit */
static inline unsigned long file_page_index(unsigned long chunk)
{
	return CHUNK_BIT_OFFSET(chunk) >> PAGE_BIT_SHIFT;
}

/* calculate the (bit) offset of this bit within a page */
static inline unsigned long file_page_offset(unsigned long chunk)
{
	return CHUNK_BIT_OFFSET(chunk) & (PAGE_BITS - 1);
}

/*
 * return a pointer to the page in the filemap that contains the given bit
 *
 * this lookup is complicated by the fact that the bitmap sb might be exactly
 * 1 page (e.g., x86) or less than 1 page -- so the bitmap might start on page
 * 0 or page 1
 */
static inline struct page *filemap_get_page(struct bitmap *bitmap,
					unsigned long chunk)
{
687
	if (file_page_index(chunk) >= bitmap->file_pages) return NULL;
688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711
	return bitmap->filemap[file_page_index(chunk) - file_page_index(0)];
}


static void bitmap_file_unmap(struct bitmap *bitmap)
{
	struct page **map, *sb_page;
	unsigned long *attr;
	int pages;
	unsigned long flags;

	spin_lock_irqsave(&bitmap->lock, flags);
	map = bitmap->filemap;
	bitmap->filemap = NULL;
	attr = bitmap->filemap_attr;
	bitmap->filemap_attr = NULL;
	pages = bitmap->file_pages;
	bitmap->file_pages = 0;
	sb_page = bitmap->sb_page;
	bitmap->sb_page = NULL;
	spin_unlock_irqrestore(&bitmap->lock, flags);

	while (pages--)
		if (map[pages]->index != 0) /* 0 is sb_page, release it below */
712
			free_buffers(map[pages]);
713 714 715
	kfree(map);
	kfree(attr);

716 717
	if (sb_page)
		free_buffers(sb_page);
718 719 720 721 722 723 724 725 726 727 728 729
}

static void bitmap_file_put(struct bitmap *bitmap)
{
	struct file *file;
	unsigned long flags;

	spin_lock_irqsave(&bitmap->lock, flags);
	file = bitmap->file;
	bitmap->file = NULL;
	spin_unlock_irqrestore(&bitmap->lock, flags);

730 731 732
	if (file)
		wait_event(bitmap->write_wait,
			   atomic_read(&bitmap->pending_writes)==0);
733 734
	bitmap_file_unmap(bitmap);

735
	if (file) {
J
Josef Sipek 已提交
736
		struct inode *inode = file->f_path.dentry->d_inode;
737
		invalidate_mapping_pages(inode->i_mapping, 0, -1);
738
		fput(file);
739
	}
740 741 742 743 744 745 746 747 748 749 750 751
}


/*
 * bitmap_file_kick - if an error occurs while manipulating the bitmap file
 * then it is no longer reliable, so we stop using it and we mark the file
 * as failed in the superblock
 */
static void bitmap_file_kick(struct bitmap *bitmap)
{
	char *path, *ptr = NULL;

752 753
	if (bitmap_mask_state(bitmap, BITMAP_STALE, MASK_SET) == 0) {
		bitmap_update_sb(bitmap);
754

755 756 757
		if (bitmap->file) {
			path = kmalloc(PAGE_SIZE, GFP_KERNEL);
			if (path)
C
Christoph Hellwig 已提交
758 759 760
				ptr = d_path(&bitmap->file->f_path, path,
					     PAGE_SIZE);

761

762 763
			printk(KERN_ALERT
			      "%s: kicking failed bitmap file %s from array!\n",
C
Christoph Hellwig 已提交
764
			      bmname(bitmap), IS_ERR(ptr) ? "" : ptr);
765

766 767 768 769 770
			kfree(path);
		} else
			printk(KERN_ALERT
			       "%s: disabling internal bitmap due to errors\n",
			       bmname(bitmap));
771
	}
772 773 774 775 776 777 778

	bitmap_file_put(bitmap);

	return;
}

enum bitmap_page_attr {
779 780 781
	BITMAP_PAGE_DIRTY = 0, // there are set bits that need to be synced
	BITMAP_PAGE_CLEAN = 1, // there are bits that might need to be cleared
	BITMAP_PAGE_NEEDWRITE=2, // there are cleared bits that need to be synced
782 783 784 785 786
};

static inline void set_page_attr(struct bitmap *bitmap, struct page *page,
				enum bitmap_page_attr attr)
{
787
	__set_bit((page->index<<2) + attr, bitmap->filemap_attr);
788 789 790 791 792
}

static inline void clear_page_attr(struct bitmap *bitmap, struct page *page,
				enum bitmap_page_attr attr)
{
793
	__clear_bit((page->index<<2) + attr, bitmap->filemap_attr);
794 795
}

796 797
static inline unsigned long test_page_attr(struct bitmap *bitmap, struct page *page,
					   enum bitmap_page_attr attr)
798
{
799
	return test_bit((page->index<<2) + attr, bitmap->filemap_attr);
800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815
}

/*
 * bitmap_file_set_bit -- called before performing a write to the md device
 * to set (and eventually sync) a particular bit in the bitmap file
 *
 * we set the bit immediately, then we record the page number so that
 * when an unplug occurs, we can flush the dirty pages out to disk
 */
static void bitmap_file_set_bit(struct bitmap *bitmap, sector_t block)
{
	unsigned long bit;
	struct page *page;
	void *kaddr;
	unsigned long chunk = block >> CHUNK_BLOCK_SHIFT(bitmap);

816
	if (!bitmap->filemap) {
817 818 819 820
		return;
	}

	page = filemap_get_page(bitmap, chunk);
821
	if (!page) return;
822 823 824 825
	bit = file_page_offset(chunk);

 	/* set the bit */
	kaddr = kmap_atomic(page, KM_USER0);
826 827 828 829
	if (bitmap->flags & BITMAP_HOSTENDIAN)
		set_bit(bit, kaddr);
	else
		ext2_set_bit(bit, kaddr);
830 831 832 833 834 835 836 837 838 839 840
	kunmap_atomic(kaddr, KM_USER0);
	PRINTK("set file bit %lu page %lu\n", bit, page->index);

	/* record page number so it gets flushed to disk when unplug occurs */
	set_page_attr(bitmap, page, BITMAP_PAGE_DIRTY);

}

/* this gets called when the md device is ready to unplug its underlying
 * (slave) device queues -- before we let any writes go down, we need to
 * sync the dirty pages of the bitmap file to disk */
841
void bitmap_unplug(struct bitmap *bitmap)
842
{
843 844
	unsigned long i, flags;
	int dirty, need_write;
845 846 847 848
	struct page *page;
	int wait = 0;

	if (!bitmap)
849
		return;
850 851 852 853 854

	/* look at each page to see if there are any set bits that need to be
	 * flushed out to disk */
	for (i = 0; i < bitmap->file_pages; i++) {
		spin_lock_irqsave(&bitmap->lock, flags);
855
		if (!bitmap->filemap) {
856
			spin_unlock_irqrestore(&bitmap->lock, flags);
857
			return;
858 859
		}
		page = bitmap->filemap[i];
860 861
		dirty = test_page_attr(bitmap, page, BITMAP_PAGE_DIRTY);
		need_write = test_page_attr(bitmap, page, BITMAP_PAGE_NEEDWRITE);
862 863
		clear_page_attr(bitmap, page, BITMAP_PAGE_DIRTY);
		clear_page_attr(bitmap, page, BITMAP_PAGE_NEEDWRITE);
864
		if (dirty)
865 866 867
			wait = 1;
		spin_unlock_irqrestore(&bitmap->lock, flags);

868
		if (dirty | need_write)
869
			write_page(bitmap, page, 0);
870 871
	}
	if (wait) { /* if any writes were performed, we need to wait on them */
872
		if (bitmap->file)
873 874
			wait_event(bitmap->write_wait,
				   atomic_read(&bitmap->pending_writes)==0);
875
		else
876
			md_super_wait(bitmap->mddev);
877
	}
878 879
	if (bitmap->flags & BITMAP_WRITE_ERROR)
		bitmap_file_kick(bitmap);
880 881
}

882
static void bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int needed);
883 884 885 886 887 888 889
/* * bitmap_init_from_disk -- called at bitmap_create time to initialize
 * the in-memory bitmap from the on-disk bitmap -- also, sets up the
 * memory mapping of the bitmap file
 * Special cases:
 *   if there's no bitmap file, or if the bitmap file had been
 *   previously kicked from the array, we mark all the bits as
 *   1's in order to cause a full resync.
890 891 892
 *
 * We ignore all bits for sectors that end earlier than 'start'.
 * This is used when reading an out-of-date bitmap...
893
 */
894
static int bitmap_init_from_disk(struct bitmap *bitmap, sector_t start)
895 896 897 898 899
{
	unsigned long i, chunks, index, oldindex, bit;
	struct page *page = NULL, *oldpage = NULL;
	unsigned long num_pages, bit_cnt = 0;
	struct file *file;
900
	unsigned long bytes, offset;
901 902
	int outofdate;
	int ret = -ENOSPC;
903
	void *paddr;
904 905 906 907

	chunks = bitmap->chunks;
	file = bitmap->file;

908
	BUG_ON(!file && !bitmap->offset);
909

910
#ifdef INJECT_FAULTS_3
911 912 913 914 915 916 917 918 919
	outofdate = 1;
#else
	outofdate = bitmap->flags & BITMAP_STALE;
#endif
	if (outofdate)
		printk(KERN_INFO "%s: bitmap file is out of date, doing full "
			"recovery\n", bmname(bitmap));

	bytes = (chunks + 7) / 8;
920

921
	num_pages = (bytes + sizeof(bitmap_super_t) + PAGE_SIZE - 1) / PAGE_SIZE;
922

923
	if (file && i_size_read(file->f_mapping->host) < bytes + sizeof(bitmap_super_t)) {
924 925 926 927
		printk(KERN_INFO "%s: bitmap file too short %lu < %lu\n",
			bmname(bitmap),
			(unsigned long) i_size_read(file->f_mapping->host),
			bytes + sizeof(bitmap_super_t));
928
		goto err;
929
	}
930 931 932

	ret = -ENOMEM;

933
	bitmap->filemap = kmalloc(sizeof(struct page *) * num_pages, GFP_KERNEL);
934
	if (!bitmap->filemap)
935
		goto err;
936

937 938
	/* We need 4 bits per page, rounded up to a multiple of sizeof(unsigned long) */
	bitmap->filemap_attr = kzalloc(
939
		roundup( DIV_ROUND_UP(num_pages*4, 8), sizeof(unsigned long)),
940
		GFP_KERNEL);
941
	if (!bitmap->filemap_attr)
942
		goto err;
943 944 945 946

	oldindex = ~0L;

	for (i = 0; i < chunks; i++) {
947
		int b;
948 949 950
		index = file_page_index(i);
		bit = file_page_offset(i);
		if (index != oldindex) { /* this is a new page, read it in */
951
			int count;
952
			/* unmap the old page, we're done with it */
953
			if (index == num_pages-1)
954 955
				count = bytes + sizeof(bitmap_super_t)
					- index * PAGE_SIZE;
956 957
			else
				count = PAGE_SIZE;
958 959 960 961 962 963 964 965
			if (index == 0) {
				/*
				 * if we're here then the superblock page
				 * contains some bits (PAGE_SIZE != sizeof sb)
				 * we've already read it in, so just use it
				 */
				page = bitmap->sb_page;
				offset = sizeof(bitmap_super_t);
N
NeilBrown 已提交
966 967 968 969 970
				if (!file)
					read_sb_page(bitmap->mddev,
						     bitmap->offset,
						     page,
						     index, count);
971
			} else if (file) {
972
				page = read_page(file, index, bitmap, count);
973 974
				offset = 0;
			} else {
975 976 977
				page = read_sb_page(bitmap->mddev, bitmap->offset,
						    NULL,
						    index, count);
978 979
				offset = 0;
			}
980 981
			if (IS_ERR(page)) { /* read error */
				ret = PTR_ERR(page);
982
				goto err;
983 984
			}

985 986 987 988 989 990 991 992
			oldindex = index;
			oldpage = page;

			if (outofdate) {
				/*
				 * if bitmap is out of date, dirty the
			 	 * whole page and write it out
				 */
993 994
				paddr = kmap_atomic(page, KM_USER0);
				memset(paddr + offset, 0xff,
995
				       PAGE_SIZE - offset);
996
				kunmap_atomic(paddr, KM_USER0);
997 998 999 1000
				write_page(bitmap, page, 1);

				ret = -EIO;
				if (bitmap->flags & BITMAP_WRITE_ERROR) {
1001
					/* release, page not in filemap yet */
1002
					put_page(page);
1003
					goto err;
1004 1005 1006 1007
				}
			}

			bitmap->filemap[bitmap->file_pages++] = page;
1008
			bitmap->last_page_size = count;
1009
		}
1010
		paddr = kmap_atomic(page, KM_USER0);
1011
		if (bitmap->flags & BITMAP_HOSTENDIAN)
1012
			b = test_bit(bit, paddr);
1013
		else
1014 1015
			b = ext2_test_bit(bit, paddr);
		kunmap_atomic(paddr, KM_USER0);
1016
		if (b) {
1017
			/* if the disk bit is set, set the memory bit */
1018 1019 1020
			bitmap_set_memory_bits(bitmap, i << CHUNK_BLOCK_SHIFT(bitmap),
					       ((i+1) << (CHUNK_BLOCK_SHIFT(bitmap)) >= start)
				);
1021
			bit_cnt++;
1022
			set_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
		}
	}

 	/* everything went OK */
	ret = 0;
	bitmap_mask_state(bitmap, BITMAP_STALE, MASK_UNSET);

	if (bit_cnt) { /* Kick recovery if any bits were set */
		set_bit(MD_RECOVERY_NEEDED, &bitmap->mddev->recovery);
		md_wakeup_thread(bitmap->mddev->thread);
	}

	printk(KERN_INFO "%s: bitmap initialized from disk: "
1036 1037 1038 1039
		"read %lu/%lu pages, set %lu bits\n",
		bmname(bitmap), bitmap->file_pages, num_pages, bit_cnt);

	return 0;
1040

1041 1042 1043
 err:
	printk(KERN_INFO "%s: bitmap initialisation failed: %d\n",
	       bmname(bitmap), ret);
1044 1045 1046
	return ret;
}

1047 1048 1049 1050 1051
void bitmap_write_all(struct bitmap *bitmap)
{
	/* We don't actually write all bitmap blocks here,
	 * just flag them as needing to be written
	 */
1052
	int i;
1053

1054 1055 1056
	for (i=0; i < bitmap->file_pages; i++)
		set_page_attr(bitmap, bitmap->filemap[i],
			      BITMAP_PAGE_NEEDWRITE);
1057 1058
}

1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079

static void bitmap_count_page(struct bitmap *bitmap, sector_t offset, int inc)
{
	sector_t chunk = offset >> CHUNK_BLOCK_SHIFT(bitmap);
	unsigned long page = chunk >> PAGE_COUNTER_SHIFT;
	bitmap->bp[page].count += inc;
/*
	if (page == 0) printk("count page 0, offset %llu: %d gives %d\n",
			      (unsigned long long)offset, inc, bitmap->bp[page].count);
*/
	bitmap_checkfree(bitmap, page);
}
static bitmap_counter_t *bitmap_get_counter(struct bitmap *bitmap,
					    sector_t offset, int *blocks,
					    int create);

/*
 * bitmap daemon -- periodically wakes up to clean bits and flush pages
 *			out to disk
 */

1080
void bitmap_daemon_work(struct bitmap *bitmap)
1081
{
1082
	unsigned long j;
1083 1084 1085
	unsigned long flags;
	struct page *page = NULL, *lastpage = NULL;
	int blocks;
1086
	void *paddr;
1087 1088

	if (bitmap == NULL)
1089
		return;
1090
	if (time_before(jiffies, bitmap->daemon_lastrun + bitmap->daemon_sleep*HZ))
1091 1092
		goto done;

1093
	bitmap->daemon_lastrun = jiffies;
1094 1095 1096 1097 1098
	if (bitmap->allclean) {
		bitmap->mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
		return;
	}
	bitmap->allclean = 1;
1099 1100 1101 1102

	for (j = 0; j < bitmap->chunks; j++) {
		bitmap_counter_t *bmc;
		spin_lock_irqsave(&bitmap->lock, flags);
1103
		if (!bitmap->filemap) {
1104 1105 1106 1107 1108 1109 1110 1111
			/* error or shutdown */
			spin_unlock_irqrestore(&bitmap->lock, flags);
			break;
		}

		page = filemap_get_page(bitmap, j);

		if (page != lastpage) {
1112
			/* skip this page unless it's marked as needing cleaning */
1113 1114 1115
			if (!test_page_attr(bitmap, page, BITMAP_PAGE_CLEAN)) {
				int need_write = test_page_attr(bitmap, page,
								BITMAP_PAGE_NEEDWRITE);
1116
				if (need_write)
1117
					clear_page_attr(bitmap, page, BITMAP_PAGE_NEEDWRITE);
1118

1119
				spin_unlock_irqrestore(&bitmap->lock, flags);
1120
				if (need_write) {
1121
					write_page(bitmap, page, 0);
1122 1123
					bitmap->allclean = 0;
				}
1124 1125 1126
				continue;
			}

1127 1128
			/* grab the new page, sync and release the old */
			if (lastpage != NULL) {
1129
				if (test_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE)) {
1130 1131
					clear_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE);
					spin_unlock_irqrestore(&bitmap->lock, flags);
1132
					write_page(bitmap, lastpage, 0);
1133 1134 1135 1136 1137 1138 1139
				} else {
					set_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE);
					spin_unlock_irqrestore(&bitmap->lock, flags);
				}
			} else
				spin_unlock_irqrestore(&bitmap->lock, flags);
			lastpage = page;
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152

			/* We are possibly going to clear some bits, so make
			 * sure that events_cleared is up-to-date.
			 */
			if (bitmap->need_sync) {
				bitmap_super_t *sb;
				bitmap->need_sync = 0;
				sb = kmap_atomic(bitmap->sb_page, KM_USER0);
				sb->events_cleared =
					cpu_to_le64(bitmap->events_cleared);
				kunmap_atomic(sb, KM_USER0);
				write_page(bitmap, bitmap->sb_page, 1);
			}
1153 1154 1155 1156 1157 1158 1159 1160 1161
			spin_lock_irqsave(&bitmap->lock, flags);
			clear_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
		}
		bmc = bitmap_get_counter(bitmap, j << CHUNK_BLOCK_SHIFT(bitmap),
					&blocks, 0);
		if (bmc) {
/*
  if (j < 100) printk("bitmap: j=%lu, *bmc = 0x%x\n", j, *bmc);
*/
1162 1163 1164
			if (*bmc)
				bitmap->allclean = 0;

1165 1166 1167 1168 1169 1170 1171 1172 1173 1174
			if (*bmc == 2) {
				*bmc=1; /* maybe clear the bit next time */
				set_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
			} else if (*bmc == 1) {
				/* we can clear the bit */
				*bmc = 0;
				bitmap_count_page(bitmap, j << CHUNK_BLOCK_SHIFT(bitmap),
						  -1);

				/* clear the bit */
1175
				paddr = kmap_atomic(page, KM_USER0);
1176
				if (bitmap->flags & BITMAP_HOSTENDIAN)
1177
					clear_bit(file_page_offset(j), paddr);
1178
				else
1179 1180
					ext2_clear_bit(file_page_offset(j), paddr);
				kunmap_atomic(paddr, KM_USER0);
1181 1182 1183 1184 1185 1186 1187 1188
			}
		}
		spin_unlock_irqrestore(&bitmap->lock, flags);
	}

	/* now sync the final page */
	if (lastpage != NULL) {
		spin_lock_irqsave(&bitmap->lock, flags);
1189
		if (test_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE)) {
1190 1191
			clear_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE);
			spin_unlock_irqrestore(&bitmap->lock, flags);
1192
			write_page(bitmap, lastpage, 0);
1193 1194 1195 1196 1197 1198
		} else {
			set_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE);
			spin_unlock_irqrestore(&bitmap->lock, flags);
		}
	}

1199
 done:
1200 1201
	if (bitmap->allclean == 0)
		bitmap->mddev->thread->timeout = bitmap->daemon_sleep * HZ;
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240
}

static bitmap_counter_t *bitmap_get_counter(struct bitmap *bitmap,
					    sector_t offset, int *blocks,
					    int create)
{
	/* If 'create', we might release the lock and reclaim it.
	 * The lock must have been taken with interrupts enabled.
	 * If !create, we don't release the lock.
	 */
	sector_t chunk = offset >> CHUNK_BLOCK_SHIFT(bitmap);
	unsigned long page = chunk >> PAGE_COUNTER_SHIFT;
	unsigned long pageoff = (chunk & PAGE_COUNTER_MASK) << COUNTER_BYTE_SHIFT;
	sector_t csize;

	if (bitmap_checkpage(bitmap, page, create) < 0) {
		csize = ((sector_t)1) << (CHUNK_BLOCK_SHIFT(bitmap));
		*blocks = csize - (offset & (csize- 1));
		return NULL;
	}
	/* now locked ... */

	if (bitmap->bp[page].hijacked) { /* hijacked pointer */
		/* should we use the first or second counter field
		 * of the hijacked pointer? */
		int hi = (pageoff > PAGE_COUNTER_MASK);
		csize = ((sector_t)1) << (CHUNK_BLOCK_SHIFT(bitmap) +
					  PAGE_COUNTER_SHIFT - 1);
		*blocks = csize - (offset & (csize- 1));
		return  &((bitmap_counter_t *)
			  &bitmap->bp[page].map)[hi];
	} else { /* page is allocated */
		csize = ((sector_t)1) << (CHUNK_BLOCK_SHIFT(bitmap));
		*blocks = csize - (offset & (csize- 1));
		return (bitmap_counter_t *)
			&(bitmap->bp[page].map[pageoff]);
	}
}

1241
int bitmap_startwrite(struct bitmap *bitmap, sector_t offset, unsigned long sectors, int behind)
1242 1243
{
	if (!bitmap) return 0;
1244 1245 1246 1247 1248 1249 1250

	if (behind) {
		atomic_inc(&bitmap->behind_writes);
		PRINTK(KERN_DEBUG "inc write-behind count %d/%d\n",
		  atomic_read(&bitmap->behind_writes), bitmap->max_write_behind);
	}

1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261
	while (sectors) {
		int blocks;
		bitmap_counter_t *bmc;

		spin_lock_irq(&bitmap->lock);
		bmc = bitmap_get_counter(bitmap, offset, &blocks, 1);
		if (!bmc) {
			spin_unlock_irq(&bitmap->lock);
			return 0;
		}

1262 1263 1264 1265 1266 1267 1268 1269 1270
		if (unlikely((*bmc & COUNTER_MAX) == COUNTER_MAX)) {
			DEFINE_WAIT(__wait);
			/* note that it is safe to do the prepare_to_wait
			 * after the test as long as we do it before dropping
			 * the spinlock.
			 */
			prepare_to_wait(&bitmap->overflow_wait, &__wait,
					TASK_UNINTERRUPTIBLE);
			spin_unlock_irq(&bitmap->lock);
1271
			blk_unplug(bitmap->mddev->queue);
1272 1273 1274 1275 1276
			schedule();
			finish_wait(&bitmap->overflow_wait, &__wait);
			continue;
		}

1277 1278 1279 1280
		switch(*bmc) {
		case 0:
			bitmap_file_set_bit(bitmap, offset);
			bitmap_count_page(bitmap,offset, 1);
1281
			blk_plug_device_unlocked(bitmap->mddev->queue);
1282 1283 1284 1285
			/* fall through */
		case 1:
			*bmc = 2;
		}
1286

1287 1288 1289 1290 1291 1292 1293 1294 1295
		(*bmc)++;

		spin_unlock_irq(&bitmap->lock);

		offset += blocks;
		if (sectors > blocks)
			sectors -= blocks;
		else sectors = 0;
	}
1296
	bitmap->allclean = 0;
1297 1298 1299 1300
	return 0;
}

void bitmap_endwrite(struct bitmap *bitmap, sector_t offset, unsigned long sectors,
1301
		     int success, int behind)
1302 1303
{
	if (!bitmap) return;
1304 1305 1306 1307 1308
	if (behind) {
		atomic_dec(&bitmap->behind_writes);
		PRINTK(KERN_DEBUG "dec write-behind count %d/%d\n",
		  atomic_read(&bitmap->behind_writes), bitmap->max_write_behind);
	}
1309 1310 1311
	if (bitmap->mddev->degraded)
		/* Never clear bits or update events_cleared when degraded */
		success = 0;
1312

1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
	while (sectors) {
		int blocks;
		unsigned long flags;
		bitmap_counter_t *bmc;

		spin_lock_irqsave(&bitmap->lock, flags);
		bmc = bitmap_get_counter(bitmap, offset, &blocks, 0);
		if (!bmc) {
			spin_unlock_irqrestore(&bitmap->lock, flags);
			return;
		}

1325 1326 1327 1328 1329 1330
		if (success &&
		    bitmap->events_cleared < bitmap->mddev->events) {
			bitmap->events_cleared = bitmap->mddev->events;
			bitmap->need_sync = 1;
		}

1331 1332 1333
		if (!success && ! (*bmc & NEEDED_MASK))
			*bmc |= NEEDED_MASK;

1334 1335 1336
		if ((*bmc & COUNTER_MAX) == COUNTER_MAX)
			wake_up(&bitmap->overflow_wait);

1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
		(*bmc)--;
		if (*bmc <= 2) {
			set_page_attr(bitmap,
				      filemap_get_page(bitmap, offset >> CHUNK_BLOCK_SHIFT(bitmap)),
				      BITMAP_PAGE_CLEAN);
		}
		spin_unlock_irqrestore(&bitmap->lock, flags);
		offset += blocks;
		if (sectors > blocks)
			sectors -= blocks;
		else sectors = 0;
	}
}

1351 1352
static int __bitmap_start_sync(struct bitmap *bitmap, sector_t offset, int *blocks,
			       int degraded)
1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368
{
	bitmap_counter_t *bmc;
	int rv;
	if (bitmap == NULL) {/* FIXME or bitmap set as 'failed' */
		*blocks = 1024;
		return 1; /* always resync if no bitmap */
	}
	spin_lock_irq(&bitmap->lock);
	bmc = bitmap_get_counter(bitmap, offset, blocks, 0);
	rv = 0;
	if (bmc) {
		/* locked */
		if (RESYNC(*bmc))
			rv = 1;
		else if (NEEDED(*bmc)) {
			rv = 1;
1369 1370 1371 1372
			if (!degraded) { /* don't set/clear bits if degraded */
				*bmc |= RESYNC_MASK;
				*bmc &= ~NEEDED_MASK;
			}
1373 1374 1375
		}
	}
	spin_unlock_irq(&bitmap->lock);
1376
	bitmap->allclean = 0;
1377 1378 1379
	return rv;
}

1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402
int bitmap_start_sync(struct bitmap *bitmap, sector_t offset, int *blocks,
		      int degraded)
{
	/* bitmap_start_sync must always report on multiples of whole
	 * pages, otherwise resync (which is very PAGE_SIZE based) will
	 * get confused.
	 * So call __bitmap_start_sync repeatedly (if needed) until
	 * At least PAGE_SIZE>>9 blocks are covered.
	 * Return the 'or' of the result.
	 */
	int rv = 0;
	int blocks1;

	*blocks = 0;
	while (*blocks < (PAGE_SIZE>>9)) {
		rv |= __bitmap_start_sync(bitmap, offset,
					  &blocks1, degraded);
		offset += blocks1;
		*blocks += blocks1;
	}
	return rv;
}

1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
void bitmap_end_sync(struct bitmap *bitmap, sector_t offset, int *blocks, int aborted)
{
	bitmap_counter_t *bmc;
	unsigned long flags;
/*
	if (offset == 0) printk("bitmap_end_sync 0 (%d)\n", aborted);
*/	if (bitmap == NULL) {
		*blocks = 1024;
		return;
	}
	spin_lock_irqsave(&bitmap->lock, flags);
	bmc = bitmap_get_counter(bitmap, offset, blocks, 0);
	if (bmc == NULL)
		goto unlock;
	/* locked */
/*
	if (offset == 0) printk("bitmap_end sync found 0x%x, blocks %d\n", *bmc, *blocks);
*/
	if (RESYNC(*bmc)) {
		*bmc &= ~RESYNC_MASK;

		if (!NEEDED(*bmc) && aborted)
			*bmc |= NEEDED_MASK;
		else {
			if (*bmc <= 2) {
				set_page_attr(bitmap,
					      filemap_get_page(bitmap, offset >> CHUNK_BLOCK_SHIFT(bitmap)),
					      BITMAP_PAGE_CLEAN);
			}
		}
	}
 unlock:
	spin_unlock_irqrestore(&bitmap->lock, flags);
1436
	bitmap->allclean = 0;
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
}

void bitmap_close_sync(struct bitmap *bitmap)
{
	/* Sync has finished, and any bitmap chunks that weren't synced
	 * properly have been aborted.  It remains to us to clear the
	 * RESYNC bit wherever it is still on
	 */
	sector_t sector = 0;
	int blocks;
N
NeilBrown 已提交
1447 1448
	if (!bitmap)
		return;
1449 1450
	while (sector < bitmap->mddev->resync_max_sectors) {
		bitmap_end_sync(bitmap, sector, &blocks, 0);
N
NeilBrown 已提交
1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476
		sector += blocks;
	}
}

void bitmap_cond_end_sync(struct bitmap *bitmap, sector_t sector)
{
	sector_t s = 0;
	int blocks;

	if (!bitmap)
		return;
	if (sector == 0) {
		bitmap->last_end_sync = jiffies;
		return;
	}
	if (time_before(jiffies, (bitmap->last_end_sync
				  + bitmap->daemon_sleep * HZ)))
		return;
	wait_event(bitmap->mddev->recovery_wait,
		   atomic_read(&bitmap->mddev->recovery_active) == 0);

	sector &= ~((1ULL << CHUNK_BLOCK_SHIFT(bitmap)) - 1);
	s = 0;
	while (s < sector && s < bitmap->mddev->resync_max_sectors) {
		bitmap_end_sync(bitmap, s, &blocks, 0);
		s += blocks;
1477
	}
N
NeilBrown 已提交
1478
	bitmap->last_end_sync = jiffies;
1479 1480
}

1481
static void bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int needed)
1482 1483
{
	/* For each chunk covered by any of these sectors, set the
1484
	 * counter to 1 and set resync_needed.  They should all
1485 1486
	 * be 0 at this point
	 */
1487 1488 1489 1490 1491 1492

	int secs;
	bitmap_counter_t *bmc;
	spin_lock_irq(&bitmap->lock);
	bmc = bitmap_get_counter(bitmap, offset, &secs, 1);
	if (!bmc) {
1493
		spin_unlock_irq(&bitmap->lock);
1494
		return;
1495
	}
1496 1497
	if (! *bmc) {
		struct page *page;
1498
		*bmc = 1 | (needed?NEEDED_MASK:0);
1499 1500 1501 1502 1503
		bitmap_count_page(bitmap, offset, 1);
		page = filemap_get_page(bitmap, offset >> CHUNK_BLOCK_SHIFT(bitmap));
		set_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
	}
	spin_unlock_irq(&bitmap->lock);
1504
	bitmap->allclean = 0;
1505 1506
}

1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518
/* dirty the memory and file bits for bitmap chunks "s" to "e" */
void bitmap_dirty_bits(struct bitmap *bitmap, unsigned long s, unsigned long e)
{
	unsigned long chunk;

	for (chunk = s; chunk <= e; chunk++) {
		sector_t sec = chunk << CHUNK_BLOCK_SHIFT(bitmap);
		bitmap_set_memory_bits(bitmap, sec, 1);
		bitmap_file_set_bit(bitmap, sec);
	}
}

1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
/*
 * flush out any pending updates
 */
void bitmap_flush(mddev_t *mddev)
{
	struct bitmap *bitmap = mddev->bitmap;
	int sleep;

	if (!bitmap) /* there was no bitmap */
		return;

	/* run the daemon_work three time to ensure everything is flushed
	 * that can be
	 */
	sleep = bitmap->daemon_sleep;
	bitmap->daemon_sleep = 0;
	bitmap_daemon_work(bitmap);
	bitmap_daemon_work(bitmap);
	bitmap_daemon_work(bitmap);
	bitmap->daemon_sleep = sleep;
	bitmap_update_sb(bitmap);
}

1542 1543 1544
/*
 * free memory that was allocated
 */
1545
static void bitmap_free(struct bitmap *bitmap)
1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567
{
	unsigned long k, pages;
	struct bitmap_page *bp;

	if (!bitmap) /* there was no bitmap */
		return;

	/* release the bitmap file and kill the daemon */
	bitmap_file_put(bitmap);

	bp = bitmap->bp;
	pages = bitmap->pages;

	/* free all allocated memory */

	if (bp) /* deallocate the page memory */
		for (k = 0; k < pages; k++)
			if (bp[k].map && !bp[k].hijacked)
				kfree(bp[k].map);
	kfree(bp);
	kfree(bitmap);
}
1568 1569 1570 1571 1572 1573 1574 1575
void bitmap_destroy(mddev_t *mddev)
{
	struct bitmap *bitmap = mddev->bitmap;

	if (!bitmap) /* there was no bitmap */
		return;

	mddev->bitmap = NULL; /* disconnect from the md device */
1576 1577
	if (mddev->thread)
		mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
1578 1579 1580

	bitmap_free(bitmap);
}
1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593

/*
 * initialize the bitmap structure
 * if this returns an error, bitmap_destroy must be called to do clean up
 */
int bitmap_create(mddev_t *mddev)
{
	struct bitmap *bitmap;
	unsigned long blocks = mddev->resync_max_sectors;
	unsigned long chunks;
	unsigned long pages;
	struct file *file = mddev->bitmap_file;
	int err;
1594
	sector_t start;
1595

A
Alexey Dobriyan 已提交
1596
	BUILD_BUG_ON(sizeof(bitmap_super_t) != 256);
1597

1598
	if (!file && !mddev->bitmap_offset) /* bitmap disabled, nothing to do */
1599 1600
		return 0;

1601 1602
	BUG_ON(file && mddev->bitmap_offset);

1603
	bitmap = kzalloc(sizeof(*bitmap), GFP_KERNEL);
1604 1605 1606 1607
	if (!bitmap)
		return -ENOMEM;

	spin_lock_init(&bitmap->lock);
1608 1609
	atomic_set(&bitmap->pending_writes, 0);
	init_waitqueue_head(&bitmap->write_wait);
1610
	init_waitqueue_head(&bitmap->overflow_wait);
1611

1612 1613 1614
	bitmap->mddev = mddev;

	bitmap->file = file;
1615
	bitmap->offset = mddev->bitmap_offset;
1616 1617
	if (file) {
		get_file(file);
M
Mark Fasheh 已提交
1618 1619 1620 1621
		do_sync_mapping_range(file->f_mapping, 0, LLONG_MAX,
				      SYNC_FILE_RANGE_WAIT_BEFORE |
				      SYNC_FILE_RANGE_WRITE |
				      SYNC_FILE_RANGE_WAIT_AFTER);
1622
	}
1623 1624 1625
	/* read superblock from bitmap file (this sets bitmap->chunksize) */
	err = bitmap_read_sb(bitmap);
	if (err)
1626
		goto error;
1627

1628
	bitmap->chunkshift = ffz(~bitmap->chunksize);
1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643

	/* now that chunksize and chunkshift are set, we can use these macros */
 	chunks = (blocks + CHUNK_BLOCK_RATIO(bitmap) - 1) /
			CHUNK_BLOCK_RATIO(bitmap);
 	pages = (chunks + PAGE_COUNTER_RATIO - 1) / PAGE_COUNTER_RATIO;

	BUG_ON(!pages);

	bitmap->chunks = chunks;
	bitmap->pages = pages;
	bitmap->missing_pages = pages;
	bitmap->counter_bits = COUNTER_BITS;

	bitmap->syncchunk = ~0UL;

1644
#ifdef INJECT_FATAL_FAULT_1
1645 1646
	bitmap->bp = NULL;
#else
1647
	bitmap->bp = kzalloc(pages * sizeof(*bitmap->bp), GFP_KERNEL);
1648
#endif
1649
	err = -ENOMEM;
1650
	if (!bitmap->bp)
1651
		goto error;
1652 1653 1654

	/* now that we have some pages available, initialize the in-memory
	 * bitmap from the on-disk bitmap */
1655 1656 1657 1658 1659 1660
	start = 0;
	if (mddev->degraded == 0
	    || bitmap->events_cleared == mddev->events)
		/* no need to keep dirty bits to optimise a re-add of a missing device */
		start = mddev->recovery_cp;
	err = bitmap_init_from_disk(bitmap, start);
1661

1662
	if (err)
1663
		goto error;
1664 1665 1666 1667

	printk(KERN_INFO "created bitmap (%lu pages) for device %s\n",
		pages, bmname(bitmap));

1668 1669
	mddev->bitmap = bitmap;

1670 1671
	mddev->thread->timeout = bitmap->daemon_sleep * HZ;

1672 1673 1674
	bitmap_update_sb(bitmap);

	return (bitmap->flags & BITMAP_WRITE_ERROR) ? -EIO : 0;
1675 1676 1677 1678

 error:
	bitmap_free(bitmap);
	return err;
1679 1680 1681 1682 1683 1684 1685 1686 1687
}

/* the bitmap API -- for raid personalities */
EXPORT_SYMBOL(bitmap_startwrite);
EXPORT_SYMBOL(bitmap_endwrite);
EXPORT_SYMBOL(bitmap_start_sync);
EXPORT_SYMBOL(bitmap_end_sync);
EXPORT_SYMBOL(bitmap_unplug);
EXPORT_SYMBOL(bitmap_close_sync);
N
NeilBrown 已提交
1688
EXPORT_SYMBOL(bitmap_cond_end_sync);