xfs_trans_buf.c 20.8 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2 3
 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
 * All Rights Reserved.
L
Linus Torvalds 已提交
4
 *
5 6
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
L
Linus Torvalds 已提交
7 8
 * published by the Free Software Foundation.
 *
9 10 11 12
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
L
Linus Torvalds 已提交
13
 *
14 15 16
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
L
Linus Torvalds 已提交
17 18
 */
#include "xfs.h"
19
#include "xfs_fs.h"
L
Linus Torvalds 已提交
20 21 22 23 24 25
#include "xfs_types.h"
#include "xfs_log.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_mount.h"
26 27 28 29 30 31
#include "xfs_bmap_btree.h"
#include "xfs_alloc_btree.h"
#include "xfs_ialloc_btree.h"
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_buf_item.h"
L
Linus Torvalds 已提交
32 33
#include "xfs_trans_priv.h"
#include "xfs_error.h"
C
Christoph Hellwig 已提交
34
#include "xfs_trace.h"
L
Linus Torvalds 已提交
35

36 37 38 39 40 41 42 43 44 45 46
/*
 * Check to see if a buffer matching the given parameters is already
 * a part of the given transaction.
 */
STATIC struct xfs_buf *
xfs_trans_buf_item_match(
	struct xfs_trans	*tp,
	struct xfs_buftarg	*target,
	xfs_daddr_t		blkno,
	int			len)
{
47 48
	struct xfs_log_item_desc *lidp;
	struct xfs_buf_log_item	*blip;
L
Linus Torvalds 已提交
49

50
	len = BBTOB(len);
51 52 53
	list_for_each_entry(lidp, &tp->t_items, lid_trans) {
		blip = (struct xfs_buf_log_item *)lidp->lid_item;
		if (blip->bli_item.li_type == XFS_LI_BUF &&
54
		    blip->bli_buf->b_target == target &&
55
		    XFS_BUF_ADDR(blip->bli_buf) == blkno &&
56
		    BBTOB(blip->bli_buf->b_length) == len)
57
			return blip->bli_buf;
58 59 60 61
	}

	return NULL;
}
L
Linus Torvalds 已提交
62

63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79
/*
 * Add the locked buffer to the transaction.
 *
 * The buffer must be locked, and it cannot be associated with any
 * transaction.
 *
 * If the buffer does not yet have a buf log item associated with it,
 * then allocate one for it.  Then add the buf item to the transaction.
 */
STATIC void
_xfs_trans_bjoin(
	struct xfs_trans	*tp,
	struct xfs_buf		*bp,
	int			reset_recur)
{
	struct xfs_buf_log_item	*bip;

80
	ASSERT(bp->b_transp == NULL);
81 82 83 84 85 86 87

	/*
	 * The xfs_buf_log_item pointer is stored in b_fsprivate.  If
	 * it doesn't have one yet, then allocate one and initialize it.
	 * The checks to see if one is there are in xfs_buf_item_init().
	 */
	xfs_buf_item_init(bp, tp->t_mountp);
88
	bip = bp->b_fspriv;
89
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
90
	ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_CANCEL));
91 92 93 94 95 96 97 98 99 100 101 102
	ASSERT(!(bip->bli_flags & XFS_BLI_LOGGED));
	if (reset_recur)
		bip->bli_recur = 0;

	/*
	 * Take a reference for this transaction on the buf item.
	 */
	atomic_inc(&bip->bli_refcount);

	/*
	 * Get a log_item_desc to point at the new item.
	 */
103
	xfs_trans_add_item(tp, &bip->bli_item);
104 105 106 107 108

	/*
	 * Initialize b_fsprivate2 so we can find it with incore_match()
	 * in xfs_trans_get_buf() and friends above.
	 */
109
	bp->b_transp = tp;
110 111 112 113 114 115 116 117 118 119 120

}

void
xfs_trans_bjoin(
	struct xfs_trans	*tp,
	struct xfs_buf		*bp)
{
	_xfs_trans_bjoin(tp, bp, 0);
	trace_xfs_trans_bjoin(bp->b_fspriv);
}
L
Linus Torvalds 已提交
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143

/*
 * Get and lock the buffer for the caller if it is not already
 * locked within the given transaction.  If it is already locked
 * within the transaction, just increment its lock recursion count
 * and return a pointer to it.
 *
 * If the transaction pointer is NULL, make this just a normal
 * get_buf() call.
 */
xfs_buf_t *
xfs_trans_get_buf(xfs_trans_t	*tp,
		  xfs_buftarg_t	*target_dev,
		  xfs_daddr_t	blkno,
		  int		len,
		  uint		flags)
{
	xfs_buf_t		*bp;
	xfs_buf_log_item_t	*bip;

	/*
	 * Default to a normal get_buf() call if the tp is NULL.
	 */
144
	if (tp == NULL)
D
Dave Chinner 已提交
145
		return xfs_buf_get(target_dev, blkno, len, flags);
L
Linus Torvalds 已提交
146 147 148 149 150 151 152

	/*
	 * If we find the buffer in the cache with this transaction
	 * pointer in its b_fsprivate2 field, then we know we already
	 * have it locked.  In this case we just increment the lock
	 * recursion count and return the buffer to the caller.
	 */
153
	bp = xfs_trans_buf_item_match(tp, target_dev, blkno, len);
L
Linus Torvalds 已提交
154
	if (bp != NULL) {
155
		ASSERT(xfs_buf_islocked(bp));
156 157 158 159
		if (XFS_FORCED_SHUTDOWN(tp->t_mountp)) {
			xfs_buf_stale(bp);
			XFS_BUF_DONE(bp);
		}
C
Christoph Hellwig 已提交
160

161
		ASSERT(bp->b_transp == tp);
162
		bip = bp->b_fspriv;
L
Linus Torvalds 已提交
163 164 165
		ASSERT(bip != NULL);
		ASSERT(atomic_read(&bip->bli_refcount) > 0);
		bip->bli_recur++;
C
Christoph Hellwig 已提交
166
		trace_xfs_trans_get_buf_recur(bip);
L
Linus Torvalds 已提交
167 168 169
		return (bp);
	}

D
Dave Chinner 已提交
170
	bp = xfs_buf_get(target_dev, blkno, len, flags);
L
Linus Torvalds 已提交
171 172 173 174
	if (bp == NULL) {
		return NULL;
	}

175
	ASSERT(!bp->b_error);
L
Linus Torvalds 已提交
176

177 178
	_xfs_trans_bjoin(tp, bp, 1);
	trace_xfs_trans_get_buf(bp->b_fspriv);
L
Linus Torvalds 已提交
179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212
	return (bp);
}

/*
 * Get and lock the superblock buffer of this file system for the
 * given transaction.
 *
 * We don't need to use incore_match() here, because the superblock
 * buffer is a private buffer which we keep a pointer to in the
 * mount structure.
 */
xfs_buf_t *
xfs_trans_getsb(xfs_trans_t	*tp,
		struct xfs_mount *mp,
		int		flags)
{
	xfs_buf_t		*bp;
	xfs_buf_log_item_t	*bip;

	/*
	 * Default to just trying to lock the superblock buffer
	 * if tp is NULL.
	 */
	if (tp == NULL) {
		return (xfs_getsb(mp, flags));
	}

	/*
	 * If the superblock buffer already has this transaction
	 * pointer in its b_fsprivate2 field, then we know we already
	 * have it locked.  In this case we just increment the lock
	 * recursion count and return the buffer to the caller.
	 */
	bp = mp->m_sb_bp;
213
	if (bp->b_transp == tp) {
214
		bip = bp->b_fspriv;
L
Linus Torvalds 已提交
215 216 217
		ASSERT(bip != NULL);
		ASSERT(atomic_read(&bip->bli_refcount) > 0);
		bip->bli_recur++;
C
Christoph Hellwig 已提交
218
		trace_xfs_trans_getsb_recur(bip);
L
Linus Torvalds 已提交
219 220 221 222
		return (bp);
	}

	bp = xfs_getsb(mp, flags);
223
	if (bp == NULL)
L
Linus Torvalds 已提交
224 225
		return NULL;

226 227
	_xfs_trans_bjoin(tp, bp, 1);
	trace_xfs_trans_getsb(bp->b_fspriv);
L
Linus Torvalds 已提交
228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
	return (bp);
}

#ifdef DEBUG
xfs_buftarg_t *xfs_error_target;
int	xfs_do_error;
int	xfs_req_num;
int	xfs_error_mod = 33;
#endif

/*
 * Get and lock the buffer for the caller if it is not already
 * locked within the given transaction.  If it has not yet been
 * read in, read it from disk. If it is already locked
 * within the transaction and already read in, just increment its
 * lock recursion count and return a pointer to it.
 *
 * If the transaction pointer is NULL, make this just a normal
 * read_buf() call.
 */
int
xfs_trans_read_buf(
	xfs_mount_t	*mp,
	xfs_trans_t	*tp,
	xfs_buftarg_t	*target,
	xfs_daddr_t	blkno,
	int		len,
	uint		flags,
	xfs_buf_t	**bpp)
{
	xfs_buf_t		*bp;
	xfs_buf_log_item_t	*bip;
	int			error;

D
Dave Chinner 已提交
262 263
	*bpp = NULL;

L
Linus Torvalds 已提交
264 265 266 267
	/*
	 * Default to a normal get_buf() call if the tp is NULL.
	 */
	if (tp == NULL) {
D
Dave Chinner 已提交
268
		bp = xfs_buf_read(target, blkno, len, flags);
L
Linus Torvalds 已提交
269
		if (!bp)
270
			return (flags & XBF_TRYLOCK) ?
271
					EAGAIN : XFS_ERROR(ENOMEM);
L
Linus Torvalds 已提交
272

273 274
		if (bp->b_error) {
			error = bp->b_error;
275
			xfs_buf_ioerror_alert(bp, __func__);
D
Dave Chinner 已提交
276 277
			XFS_BUF_UNDONE(bp);
			xfs_buf_stale(bp);
L
Linus Torvalds 已提交
278 279 280 281
			xfs_buf_relse(bp);
			return error;
		}
#ifdef DEBUG
J
Julia Lawall 已提交
282
		if (xfs_do_error) {
L
Linus Torvalds 已提交
283 284 285
			if (xfs_error_target == target) {
				if (((xfs_req_num++) % xfs_error_mod) == 0) {
					xfs_buf_relse(bp);
286
					xfs_debug(mp, "Returning error!");
L
Linus Torvalds 已提交
287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
					return XFS_ERROR(EIO);
				}
			}
		}
#endif
		if (XFS_FORCED_SHUTDOWN(mp))
			goto shutdown_abort;
		*bpp = bp;
		return 0;
	}

	/*
	 * If we find the buffer in the cache with this transaction
	 * pointer in its b_fsprivate2 field, then we know we already
	 * have it locked.  If it is already read in we just increment
	 * the lock recursion count and return the buffer to the caller.
	 * If the buffer is not yet read in, then we read it in, increment
	 * the lock recursion count, and return it to the caller.
	 */
306
	bp = xfs_trans_buf_item_match(tp, target, blkno, len);
L
Linus Torvalds 已提交
307
	if (bp != NULL) {
308
		ASSERT(xfs_buf_islocked(bp));
309
		ASSERT(bp->b_transp == tp);
310
		ASSERT(bp->b_fspriv != NULL);
311
		ASSERT(!bp->b_error);
L
Linus Torvalds 已提交
312
		if (!(XFS_BUF_ISDONE(bp))) {
C
Christoph Hellwig 已提交
313
			trace_xfs_trans_read_buf_io(bp, _RET_IP_);
L
Linus Torvalds 已提交
314 315 316
			ASSERT(!XFS_BUF_ISASYNC(bp));
			XFS_BUF_READ(bp);
			xfsbdstrat(tp->t_mountp, bp);
C
Christoph Hellwig 已提交
317
			error = xfs_buf_iowait(bp);
318
			if (error) {
319
				xfs_buf_ioerror_alert(bp, __func__);
L
Linus Torvalds 已提交
320 321
				xfs_buf_relse(bp);
				/*
322 323 324
				 * We can gracefully recover from most read
				 * errors. Ones we can't are those that happen
				 * after the transaction's already dirty.
L
Linus Torvalds 已提交
325 326 327
				 */
				if (tp->t_flags & XFS_TRANS_DIRTY)
					xfs_force_shutdown(tp->t_mountp,
328
							SHUTDOWN_META_IO_ERROR);
L
Linus Torvalds 已提交
329 330 331 332 333 334 335 336
				return error;
			}
		}
		/*
		 * We never locked this buf ourselves, so we shouldn't
		 * brelse it either. Just get out.
		 */
		if (XFS_FORCED_SHUTDOWN(mp)) {
C
Christoph Hellwig 已提交
337
			trace_xfs_trans_read_buf_shut(bp, _RET_IP_);
L
Linus Torvalds 已提交
338 339 340 341 342
			*bpp = NULL;
			return XFS_ERROR(EIO);
		}


343
		bip = bp->b_fspriv;
L
Linus Torvalds 已提交
344 345 346
		bip->bli_recur++;

		ASSERT(atomic_read(&bip->bli_refcount) > 0);
C
Christoph Hellwig 已提交
347
		trace_xfs_trans_read_buf_recur(bip);
L
Linus Torvalds 已提交
348 349 350 351
		*bpp = bp;
		return 0;
	}

D
Dave Chinner 已提交
352
	bp = xfs_buf_read(target, blkno, len, flags);
L
Linus Torvalds 已提交
353 354
	if (bp == NULL) {
		*bpp = NULL;
355 356
		return (flags & XBF_TRYLOCK) ?
					0 : XFS_ERROR(ENOMEM);
L
Linus Torvalds 已提交
357
	}
358 359
	if (bp->b_error) {
		error = bp->b_error;
360 361
		xfs_buf_stale(bp);
		XFS_BUF_DONE(bp);
362
		xfs_buf_ioerror_alert(bp, __func__);
L
Linus Torvalds 已提交
363
		if (tp->t_flags & XFS_TRANS_DIRTY)
364
			xfs_force_shutdown(tp->t_mountp, SHUTDOWN_META_IO_ERROR);
L
Linus Torvalds 已提交
365 366 367 368 369 370 371 372
		xfs_buf_relse(bp);
		return error;
	}
#ifdef DEBUG
	if (xfs_do_error && !(tp->t_flags & XFS_TRANS_DIRTY)) {
		if (xfs_error_target == target) {
			if (((xfs_req_num++) % xfs_error_mod) == 0) {
				xfs_force_shutdown(tp->t_mountp,
373
						   SHUTDOWN_META_IO_ERROR);
L
Linus Torvalds 已提交
374
				xfs_buf_relse(bp);
375
				xfs_debug(mp, "Returning trans error!");
L
Linus Torvalds 已提交
376 377 378 379 380 381 382 383
				return XFS_ERROR(EIO);
			}
		}
	}
#endif
	if (XFS_FORCED_SHUTDOWN(mp))
		goto shutdown_abort;

384 385
	_xfs_trans_bjoin(tp, bp, 1);
	trace_xfs_trans_read_buf(bp->b_fspriv);
L
Linus Torvalds 已提交
386 387 388 389 390

	*bpp = bp;
	return 0;

shutdown_abort:
C
Christoph Hellwig 已提交
391
	trace_xfs_trans_read_buf_shut(bp, _RET_IP_);
L
Linus Torvalds 已提交
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422
	xfs_buf_relse(bp);
	*bpp = NULL;
	return XFS_ERROR(EIO);
}


/*
 * Release the buffer bp which was previously acquired with one of the
 * xfs_trans_... buffer allocation routines if the buffer has not
 * been modified within this transaction.  If the buffer is modified
 * within this transaction, do decrement the recursion count but do
 * not release the buffer even if the count goes to 0.  If the buffer is not
 * modified within the transaction, decrement the recursion count and
 * release the buffer if the recursion count goes to 0.
 *
 * If the buffer is to be released and it was not modified before
 * this transaction began, then free the buf_log_item associated with it.
 *
 * If the transaction pointer is NULL, make this just a normal
 * brelse() call.
 */
void
xfs_trans_brelse(xfs_trans_t	*tp,
		 xfs_buf_t	*bp)
{
	xfs_buf_log_item_t	*bip;

	/*
	 * Default to a normal brelse() call if the tp is NULL.
	 */
	if (tp == NULL) {
423
		ASSERT(bp->b_transp == NULL);
L
Linus Torvalds 已提交
424 425 426 427
		xfs_buf_relse(bp);
		return;
	}

428
	ASSERT(bp->b_transp == tp);
429
	bip = bp->b_fspriv;
L
Linus Torvalds 已提交
430 431
	ASSERT(bip->bli_item.li_type == XFS_LI_BUF);
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
432
	ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_CANCEL));
L
Linus Torvalds 已提交
433 434
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

C
Christoph Hellwig 已提交
435 436
	trace_xfs_trans_brelse(bip);

L
Linus Torvalds 已提交
437 438 439 440 441 442 443 444 445 446 447 448 449
	/*
	 * If the release is just for a recursive lock,
	 * then decrement the count and return.
	 */
	if (bip->bli_recur > 0) {
		bip->bli_recur--;
		return;
	}

	/*
	 * If the buffer is dirty within this transaction, we can't
	 * release it until we commit.
	 */
450
	if (bip->bli_item.li_desc->lid_flags & XFS_LID_DIRTY)
L
Linus Torvalds 已提交
451 452 453 454 455 456 457 458
		return;

	/*
	 * If the buffer has been invalidated, then we can't release
	 * it until the transaction commits to disk unless it is re-dirtied
	 * as part of this transaction.  This prevents us from pulling
	 * the item from the AIL before we should.
	 */
C
Christoph Hellwig 已提交
459
	if (bip->bli_flags & XFS_BLI_STALE)
L
Linus Torvalds 已提交
460 461 462 463 464 465 466
		return;

	ASSERT(!(bip->bli_flags & XFS_BLI_LOGGED));

	/*
	 * Free up the log item descriptor tracking the released item.
	 */
467
	xfs_trans_del_item(&bip->bli_item);
L
Linus Torvalds 已提交
468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499

	/*
	 * Clear the hold flag in the buf log item if it is set.
	 * We wouldn't want the next user of the buffer to
	 * get confused.
	 */
	if (bip->bli_flags & XFS_BLI_HOLD) {
		bip->bli_flags &= ~XFS_BLI_HOLD;
	}

	/*
	 * Drop our reference to the buf log item.
	 */
	atomic_dec(&bip->bli_refcount);

	/*
	 * If the buf item is not tracking data in the log, then
	 * we must free it before releasing the buffer back to the
	 * free pool.  Before releasing the buffer to the free pool,
	 * clear the transaction pointer in b_fsprivate2 to dissolve
	 * its relation to this transaction.
	 */
	if (!xfs_buf_item_dirty(bip)) {
/***
		ASSERT(bp->b_pincount == 0);
***/
		ASSERT(atomic_read(&bip->bli_refcount) == 0);
		ASSERT(!(bip->bli_item.li_flags & XFS_LI_IN_AIL));
		ASSERT(!(bip->bli_flags & XFS_BLI_INODE_ALLOC_BUF));
		xfs_buf_item_relse(bp);
	}

500
	bp->b_transp = NULL;
L
Linus Torvalds 已提交
501 502 503 504 505 506 507 508 509 510 511 512 513
	xfs_buf_relse(bp);
}

/*
 * Mark the buffer as not needing to be unlocked when the buf item's
 * IOP_UNLOCK() routine is called.  The buffer must already be locked
 * and associated with the given transaction.
 */
/* ARGSUSED */
void
xfs_trans_bhold(xfs_trans_t	*tp,
		xfs_buf_t	*bp)
{
514
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
515

516
	ASSERT(bp->b_transp == tp);
517
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
518
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
519
	ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_CANCEL));
L
Linus Torvalds 已提交
520
	ASSERT(atomic_read(&bip->bli_refcount) > 0);
521

L
Linus Torvalds 已提交
522
	bip->bli_flags |= XFS_BLI_HOLD;
C
Christoph Hellwig 已提交
523
	trace_xfs_trans_bhold(bip);
L
Linus Torvalds 已提交
524 525
}

526 527 528 529 530 531 532 533
/*
 * Cancel the previous buffer hold request made on this buffer
 * for this transaction.
 */
void
xfs_trans_bhold_release(xfs_trans_t	*tp,
			xfs_buf_t	*bp)
{
534
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
535

536
	ASSERT(bp->b_transp == tp);
537
	ASSERT(bip != NULL);
538
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
539
	ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_CANCEL));
540 541
	ASSERT(atomic_read(&bip->bli_refcount) > 0);
	ASSERT(bip->bli_flags & XFS_BLI_HOLD);
C
Christoph Hellwig 已提交
542

543
	bip->bli_flags &= ~XFS_BLI_HOLD;
C
Christoph Hellwig 已提交
544
	trace_xfs_trans_bhold_release(bip);
545 546
}

L
Linus Torvalds 已提交
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561
/*
 * This is called to mark bytes first through last inclusive of the given
 * buffer as needing to be logged when the transaction is committed.
 * The buffer must already be associated with the given transaction.
 *
 * First and last are numbers relative to the beginning of this buffer,
 * so the first byte in the buffer is numbered 0 regardless of the
 * value of b_blkno.
 */
void
xfs_trans_log_buf(xfs_trans_t	*tp,
		  xfs_buf_t	*bp,
		  uint		first,
		  uint		last)
{
562
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
563

564
	ASSERT(bp->b_transp == tp);
565
	ASSERT(bip != NULL);
566
	ASSERT(first <= last && last < BBTOB(bp->b_length));
567 568
	ASSERT(bp->b_iodone == NULL ||
	       bp->b_iodone == xfs_buf_iodone_callbacks);
L
Linus Torvalds 已提交
569 570 571 572 573 574 575 576 577 578 579 580 581 582

	/*
	 * Mark the buffer as needing to be written out eventually,
	 * and set its iodone function to remove the buffer's buf log
	 * item from the AIL and free it when the buffer is flushed
	 * to disk.  See xfs_buf_attach_iodone() for more details
	 * on li_cb and xfs_buf_iodone_callbacks().
	 * If we end up aborting this transaction, we trap this buffer
	 * inside the b_bdstrat callback so that this won't get written to
	 * disk.
	 */
	XFS_BUF_DONE(bp);

	ASSERT(atomic_read(&bip->bli_refcount) > 0);
583
	bp->b_iodone = xfs_buf_iodone_callbacks;
584
	bip->bli_item.li_cb = xfs_buf_iodone;
L
Linus Torvalds 已提交
585

C
Christoph Hellwig 已提交
586 587
	trace_xfs_trans_log_buf(bip);

L
Linus Torvalds 已提交
588 589 590 591 592 593 594 595 596 597
	/*
	 * If we invalidated the buffer within this transaction, then
	 * cancel the invalidation now that we're dirtying the buffer
	 * again.  There are no races with the code in xfs_buf_item_unpin(),
	 * because we have a reference to the buffer this entire time.
	 */
	if (bip->bli_flags & XFS_BLI_STALE) {
		bip->bli_flags &= ~XFS_BLI_STALE;
		ASSERT(XFS_BUF_ISSTALE(bp));
		XFS_BUF_UNSTALE(bp);
598
		bip->bli_format.blf_flags &= ~XFS_BLF_CANCEL;
L
Linus Torvalds 已提交
599 600 601
	}

	tp->t_flags |= XFS_TRANS_DIRTY;
602
	bip->bli_item.li_desc->lid_flags |= XFS_LID_DIRTY;
L
Linus Torvalds 已提交
603 604 605 606 607 608
	bip->bli_flags |= XFS_BLI_LOGGED;
	xfs_buf_item_log(bip, first, last);
}


/*
609 610 611 612 613 614
 * Invalidate a buffer that is being used within a transaction.
 *
 * Typically this is because the blocks in the buffer are being freed, so we
 * need to prevent it from being written out when we're done.  Allowing it
 * to be written again might overwrite data in the free blocks if they are
 * reallocated to a file.
L
Linus Torvalds 已提交
615
 *
616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635
 * We prevent the buffer from being written out by marking it stale.  We can't
 * get rid of the buf log item at this point because the buffer may still be
 * pinned by another transaction.  If that is the case, then we'll wait until
 * the buffer is committed to disk for the last time (we can tell by the ref
 * count) and free it in xfs_buf_item_unpin().  Until that happens we will
 * keep the buffer locked so that the buffer and buf log item are not reused.
 *
 * We also set the XFS_BLF_CANCEL flag in the buf log format structure and log
 * the buf item.  This will be used at recovery time to determine that copies
 * of the buffer in the log before this should not be replayed.
 *
 * We mark the item descriptor and the transaction dirty so that we'll hold
 * the buffer until after the commit.
 *
 * Since we're invalidating the buffer, we also clear the state about which
 * parts of the buffer have been logged.  We also clear the flag indicating
 * that this is an inode buffer since the data in the buffer will no longer
 * be valid.
 *
 * We set the stale bit in the buffer as well since we're getting rid of it.
L
Linus Torvalds 已提交
636 637 638 639 640 641
 */
void
xfs_trans_binval(
	xfs_trans_t	*tp,
	xfs_buf_t	*bp)
{
642
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
643

644
	ASSERT(bp->b_transp == tp);
645
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
646 647
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

C
Christoph Hellwig 已提交
648 649
	trace_xfs_trans_binval(bip);

L
Linus Torvalds 已提交
650 651 652 653 654 655 656
	if (bip->bli_flags & XFS_BLI_STALE) {
		/*
		 * If the buffer is already invalidated, then
		 * just return.
		 */
		ASSERT(XFS_BUF_ISSTALE(bp));
		ASSERT(!(bip->bli_flags & (XFS_BLI_LOGGED | XFS_BLI_DIRTY)));
657 658
		ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_INODE_BUF));
		ASSERT(bip->bli_format.blf_flags & XFS_BLF_CANCEL);
659
		ASSERT(bip->bli_item.li_desc->lid_flags & XFS_LID_DIRTY);
L
Linus Torvalds 已提交
660 661 662 663
		ASSERT(tp->t_flags & XFS_TRANS_DIRTY);
		return;
	}

664
	xfs_buf_stale(bp);
665

L
Linus Torvalds 已提交
666
	bip->bli_flags |= XFS_BLI_STALE;
667
	bip->bli_flags &= ~(XFS_BLI_INODE_BUF | XFS_BLI_LOGGED | XFS_BLI_DIRTY);
668 669
	bip->bli_format.blf_flags &= ~XFS_BLF_INODE_BUF;
	bip->bli_format.blf_flags |= XFS_BLF_CANCEL;
L
Linus Torvalds 已提交
670 671
	memset((char *)(bip->bli_format.blf_data_map), 0,
	      (bip->bli_format.blf_map_size * sizeof(uint)));
672
	bip->bli_item.li_desc->lid_flags |= XFS_LID_DIRTY;
L
Linus Torvalds 已提交
673 674 675 676
	tp->t_flags |= XFS_TRANS_DIRTY;
}

/*
677 678 679 680 681
 * This call is used to indicate that the buffer contains on-disk inodes which
 * must be handled specially during recovery.  They require special handling
 * because only the di_next_unlinked from the inodes in the buffer should be
 * recovered.  The rest of the data in the buffer is logged via the inodes
 * themselves.
L
Linus Torvalds 已提交
682
 *
683 684 685
 * All we do is set the XFS_BLI_INODE_BUF flag in the items flags so it can be
 * transferred to the buffer's log format structure so that we'll know what to
 * do at recovery time.
L
Linus Torvalds 已提交
686 687 688 689 690 691
 */
void
xfs_trans_inode_buf(
	xfs_trans_t	*tp,
	xfs_buf_t	*bp)
{
692
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
693

694
	ASSERT(bp->b_transp == tp);
695
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
696 697
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

698
	bip->bli_flags |= XFS_BLI_INODE_BUF;
L
Linus Torvalds 已提交
699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714
}

/*
 * This call is used to indicate that the buffer is going to
 * be staled and was an inode buffer. This means it gets
 * special processing during unpin - where any inodes 
 * associated with the buffer should be removed from ail.
 * There is also special processing during recovery,
 * any replay of the inodes in the buffer needs to be
 * prevented as the buffer may have been reused.
 */
void
xfs_trans_stale_inode_buf(
	xfs_trans_t	*tp,
	xfs_buf_t	*bp)
{
715
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
716

717
	ASSERT(bp->b_transp == tp);
718
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
719 720 721
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

	bip->bli_flags |= XFS_BLI_STALE_INODE;
722
	bip->bli_item.li_cb = xfs_buf_iodone;
L
Linus Torvalds 已提交
723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738
}

/*
 * Mark the buffer as being one which contains newly allocated
 * inodes.  We need to make sure that even if this buffer is
 * relogged as an 'inode buf' we still recover all of the inode
 * images in the face of a crash.  This works in coordination with
 * xfs_buf_item_committed() to ensure that the buffer remains in the
 * AIL at its original location even after it has been relogged.
 */
/* ARGSUSED */
void
xfs_trans_inode_alloc_buf(
	xfs_trans_t	*tp,
	xfs_buf_t	*bp)
{
739
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
740

741
	ASSERT(bp->b_transp == tp);
742
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

	bip->bli_flags |= XFS_BLI_INODE_ALLOC_BUF;
}


/*
 * Similar to xfs_trans_inode_buf(), this marks the buffer as a cluster of
 * dquots. However, unlike in inode buffer recovery, dquot buffers get
 * recovered in their entirety. (Hence, no XFS_BLI_DQUOT_ALLOC_BUF flag).
 * The only thing that makes dquot buffers different from regular
 * buffers is that we must not replay dquot bufs when recovering
 * if a _corresponding_ quotaoff has happened. We also have to distinguish
 * between usr dquot bufs and grp dquot bufs, because usr and grp quotas
 * can be turned off independently.
 */
/* ARGSUSED */
void
xfs_trans_dquot_buf(
	xfs_trans_t	*tp,
	xfs_buf_t	*bp,
	uint		type)
{
766
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
767

768
	ASSERT(bp->b_transp == tp);
769
	ASSERT(bip != NULL);
770 771 772
	ASSERT(type == XFS_BLF_UDQUOT_BUF ||
	       type == XFS_BLF_PDQUOT_BUF ||
	       type == XFS_BLF_GDQUOT_BUF);
L
Linus Torvalds 已提交
773 774 775 776
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

	bip->bli_format.blf_flags |= type;
}