xfs_trans_buf.c 21.7 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
#include "xfs_types.h"
21
#include "xfs_bit.h"
L
Linus Torvalds 已提交
22
#include "xfs_log.h"
23
#include "xfs_inum.h"
L
Linus Torvalds 已提交
24 25 26 27
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_mount.h"
28 29 30 31 32 33
#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
34 35 36
#include "xfs_trans_priv.h"
#include "xfs_error.h"
#include "xfs_rw.h"
C
Christoph Hellwig 已提交
37
#include "xfs_trace.h"
L
Linus Torvalds 已提交
38

39 40 41 42 43 44 45 46 47 48 49
/*
 * 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)
{
50 51
	struct xfs_log_item_desc *lidp;
	struct xfs_buf_log_item	*blip;
L
Linus Torvalds 已提交
52

53
	len = BBTOB(len);
54 55 56
	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 &&
57
		    blip->bli_buf->b_target == target &&
58
		    XFS_BUF_ADDR(blip->bli_buf) == blkno &&
59
		    BBTOB(blip->bli_buf->b_length) == len)
60
			return blip->bli_buf;
61 62 63 64
	}

	return NULL;
}
L
Linus Torvalds 已提交
65

66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82
/*
 * 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;

83
	ASSERT(bp->b_transp == NULL);
84 85 86 87 88 89 90

	/*
	 * 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);
91
	bip = bp->b_fspriv;
92
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
93
	ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_CANCEL));
94 95 96 97 98 99 100 101 102 103 104 105
	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.
	 */
106
	xfs_trans_add_item(tp, &bip->bli_item);
107 108 109 110 111

	/*
	 * Initialize b_fsprivate2 so we can find it with incore_match()
	 * in xfs_trans_get_buf() and friends above.
	 */
112
	bp->b_transp = tp;
113 114 115 116 117 118 119 120 121 122 123

}

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
124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144

/*
 * 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;

	if (flags == 0)
D
Dave Chinner 已提交
145
		flags = XBF_MAPPED;
L
Linus Torvalds 已提交
146 147 148 149

	/*
	 * Default to a normal get_buf() call if the tp is NULL.
	 */
150
	if (tp == NULL)
151 152
		return xfs_buf_get(target_dev, blkno, len,
				   flags | XBF_DONT_BLOCK);
L
Linus Torvalds 已提交
153 154 155 156 157 158 159

	/*
	 * 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.
	 */
160
	bp = xfs_trans_buf_item_match(tp, target_dev, blkno, len);
L
Linus Torvalds 已提交
161
	if (bp != NULL) {
162
		ASSERT(xfs_buf_islocked(bp));
163 164 165 166
		if (XFS_FORCED_SHUTDOWN(tp->t_mountp)) {
			xfs_buf_stale(bp);
			XFS_BUF_DONE(bp);
		}
C
Christoph Hellwig 已提交
167

168
		ASSERT(bp->b_transp == tp);
169
		bip = bp->b_fspriv;
L
Linus Torvalds 已提交
170 171 172
		ASSERT(bip != NULL);
		ASSERT(atomic_read(&bip->bli_refcount) > 0);
		bip->bli_recur++;
C
Christoph Hellwig 已提交
173
		trace_xfs_trans_get_buf_recur(bip);
L
Linus Torvalds 已提交
174 175 176 177
		return (bp);
	}

	/*
178 179
	 * We always specify the XBF_DONT_BLOCK flag within a transaction
	 * so that get_buf does not try to push out a delayed write buffer
L
Linus Torvalds 已提交
180 181 182 183 184
	 * which might cause another transaction to take place (if the
	 * buffer was delayed alloc).  Such recursive transactions can
	 * easily deadlock with our current transaction as well as cause
	 * us to run out of stack space.
	 */
185
	bp = xfs_buf_get(target_dev, blkno, len, flags | XBF_DONT_BLOCK);
L
Linus Torvalds 已提交
186 187 188 189
	if (bp == NULL) {
		return NULL;
	}

190
	ASSERT(!bp->b_error);
L
Linus Torvalds 已提交
191

192 193
	_xfs_trans_bjoin(tp, bp, 1);
	trace_xfs_trans_get_buf(bp->b_fspriv);
L
Linus Torvalds 已提交
194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
	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;
228
	if (bp->b_transp == tp) {
229
		bip = bp->b_fspriv;
L
Linus Torvalds 已提交
230 231 232
		ASSERT(bip != NULL);
		ASSERT(atomic_read(&bip->bli_refcount) > 0);
		bip->bli_recur++;
C
Christoph Hellwig 已提交
233
		trace_xfs_trans_getsb_recur(bip);
L
Linus Torvalds 已提交
234 235 236 237
		return (bp);
	}

	bp = xfs_getsb(mp, flags);
238
	if (bp == NULL)
L
Linus Torvalds 已提交
239 240
		return NULL;

241 242
	_xfs_trans_bjoin(tp, bp, 1);
	trace_xfs_trans_getsb(bp->b_fspriv);
L
Linus Torvalds 已提交
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
	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;

	if (flags == 0)
D
Dave Chinner 已提交
278
		flags = XBF_MAPPED;
L
Linus Torvalds 已提交
279 280 281 282 283

	/*
	 * Default to a normal get_buf() call if the tp is NULL.
	 */
	if (tp == NULL) {
284
		bp = xfs_buf_read(target, blkno, len, flags | XBF_DONT_BLOCK);
L
Linus Torvalds 已提交
285
		if (!bp)
286
			return (flags & XBF_TRYLOCK) ?
287
					EAGAIN : XFS_ERROR(ENOMEM);
L
Linus Torvalds 已提交
288

289 290
		if (bp->b_error) {
			error = bp->b_error;
291
			xfs_buf_ioerror_alert(bp, __func__);
L
Linus Torvalds 已提交
292 293 294 295
			xfs_buf_relse(bp);
			return error;
		}
#ifdef DEBUG
J
Julia Lawall 已提交
296
		if (xfs_do_error) {
L
Linus Torvalds 已提交
297 298 299
			if (xfs_error_target == target) {
				if (((xfs_req_num++) % xfs_error_mod) == 0) {
					xfs_buf_relse(bp);
300
					xfs_debug(mp, "Returning error!");
L
Linus Torvalds 已提交
301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319
					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.
	 */
320
	bp = xfs_trans_buf_item_match(tp, target, blkno, len);
L
Linus Torvalds 已提交
321
	if (bp != NULL) {
322
		ASSERT(xfs_buf_islocked(bp));
323
		ASSERT(bp->b_transp == tp);
324
		ASSERT(bp->b_fspriv != NULL);
325
		ASSERT(!bp->b_error);
L
Linus Torvalds 已提交
326
		if (!(XFS_BUF_ISDONE(bp))) {
C
Christoph Hellwig 已提交
327
			trace_xfs_trans_read_buf_io(bp, _RET_IP_);
L
Linus Torvalds 已提交
328 329 330
			ASSERT(!XFS_BUF_ISASYNC(bp));
			XFS_BUF_READ(bp);
			xfsbdstrat(tp->t_mountp, bp);
C
Christoph Hellwig 已提交
331
			error = xfs_buf_iowait(bp);
332
			if (error) {
333
				xfs_buf_ioerror_alert(bp, __func__);
L
Linus Torvalds 已提交
334 335
				xfs_buf_relse(bp);
				/*
336 337 338
				 * 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
339 340 341
				 */
				if (tp->t_flags & XFS_TRANS_DIRTY)
					xfs_force_shutdown(tp->t_mountp,
342
							SHUTDOWN_META_IO_ERROR);
L
Linus Torvalds 已提交
343 344 345 346 347 348 349 350
				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
351
			trace_xfs_trans_read_buf_shut(bp, _RET_IP_);
L
Linus Torvalds 已提交
352 353 354 355 356
			*bpp = NULL;
			return XFS_ERROR(EIO);
		}


357
		bip = bp->b_fspriv;
L
Linus Torvalds 已提交
358 359 360
		bip->bli_recur++;

		ASSERT(atomic_read(&bip->bli_refcount) > 0);
C
Christoph Hellwig 已提交
361
		trace_xfs_trans_read_buf_recur(bip);
L
Linus Torvalds 已提交
362 363 364 365 366
		*bpp = bp;
		return 0;
	}

	/*
367 368
	 * We always specify the XBF_DONT_BLOCK flag within a transaction
	 * so that get_buf does not try to push out a delayed write buffer
L
Linus Torvalds 已提交
369 370 371 372 373
	 * which might cause another transaction to take place (if the
	 * buffer was delayed alloc).  Such recursive transactions can
	 * easily deadlock with our current transaction as well as cause
	 * us to run out of stack space.
	 */
374
	bp = xfs_buf_read(target, blkno, len, flags | XBF_DONT_BLOCK);
L
Linus Torvalds 已提交
375 376
	if (bp == NULL) {
		*bpp = NULL;
377 378
		return (flags & XBF_TRYLOCK) ?
					0 : XFS_ERROR(ENOMEM);
L
Linus Torvalds 已提交
379
	}
380 381
	if (bp->b_error) {
		error = bp->b_error;
382 383
		xfs_buf_stale(bp);
		XFS_BUF_DONE(bp);
384
		xfs_buf_ioerror_alert(bp, __func__);
L
Linus Torvalds 已提交
385
		if (tp->t_flags & XFS_TRANS_DIRTY)
386
			xfs_force_shutdown(tp->t_mountp, SHUTDOWN_META_IO_ERROR);
L
Linus Torvalds 已提交
387 388 389 390 391 392 393 394
		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,
395
						   SHUTDOWN_META_IO_ERROR);
L
Linus Torvalds 已提交
396
				xfs_buf_relse(bp);
397
				xfs_debug(mp, "Returning trans error!");
L
Linus Torvalds 已提交
398 399 400 401 402 403 404 405
				return XFS_ERROR(EIO);
			}
		}
	}
#endif
	if (XFS_FORCED_SHUTDOWN(mp))
		goto shutdown_abort;

406 407
	_xfs_trans_bjoin(tp, bp, 1);
	trace_xfs_trans_read_buf(bp->b_fspriv);
L
Linus Torvalds 已提交
408 409 410 411 412

	*bpp = bp;
	return 0;

shutdown_abort:
C
Christoph Hellwig 已提交
413
	trace_xfs_trans_read_buf_shut(bp, _RET_IP_);
L
Linus Torvalds 已提交
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444
	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) {
445
		ASSERT(bp->b_transp == NULL);
L
Linus Torvalds 已提交
446 447 448 449
		xfs_buf_relse(bp);
		return;
	}

450
	ASSERT(bp->b_transp == tp);
451
	bip = bp->b_fspriv;
L
Linus Torvalds 已提交
452 453
	ASSERT(bip->bli_item.li_type == XFS_LI_BUF);
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
454
	ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_CANCEL));
L
Linus Torvalds 已提交
455 456
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

C
Christoph Hellwig 已提交
457 458
	trace_xfs_trans_brelse(bip);

L
Linus Torvalds 已提交
459 460 461 462 463 464 465 466 467 468 469 470 471
	/*
	 * 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.
	 */
472
	if (bip->bli_item.li_desc->lid_flags & XFS_LID_DIRTY)
L
Linus Torvalds 已提交
473 474 475 476 477 478 479 480
		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
481
	if (bip->bli_flags & XFS_BLI_STALE)
L
Linus Torvalds 已提交
482 483 484 485 486 487 488
		return;

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

	/*
	 * Free up the log item descriptor tracking the released item.
	 */
489
	xfs_trans_del_item(&bip->bli_item);
L
Linus Torvalds 已提交
490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521

	/*
	 * 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);
	}

522
	bp->b_transp = NULL;
L
Linus Torvalds 已提交
523 524 525 526 527 528 529 530 531 532 533 534 535
	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)
{
536
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
537

538
	ASSERT(bp->b_transp == tp);
539
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
540
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
541
	ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_CANCEL));
L
Linus Torvalds 已提交
542
	ASSERT(atomic_read(&bip->bli_refcount) > 0);
543

L
Linus Torvalds 已提交
544
	bip->bli_flags |= XFS_BLI_HOLD;
C
Christoph Hellwig 已提交
545
	trace_xfs_trans_bhold(bip);
L
Linus Torvalds 已提交
546 547
}

548 549 550 551 552 553 554 555
/*
 * 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)
{
556
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
557

558
	ASSERT(bp->b_transp == tp);
559
	ASSERT(bip != NULL);
560
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
561
	ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_CANCEL));
562 563
	ASSERT(atomic_read(&bip->bli_refcount) > 0);
	ASSERT(bip->bli_flags & XFS_BLI_HOLD);
C
Christoph Hellwig 已提交
564

565
	bip->bli_flags &= ~XFS_BLI_HOLD;
C
Christoph Hellwig 已提交
566
	trace_xfs_trans_bhold_release(bip);
567 568
}

L
Linus Torvalds 已提交
569 570 571 572 573 574 575 576 577 578 579 580 581 582 583
/*
 * 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)
{
584
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
585

586
	ASSERT(bp->b_transp == tp);
587
	ASSERT(bip != NULL);
588
	ASSERT(first <= last && last < BBTOB(bp->b_length));
589 590
	ASSERT(bp->b_iodone == NULL ||
	       bp->b_iodone == xfs_buf_iodone_callbacks);
L
Linus Torvalds 已提交
591 592 593 594 595 596 597 598 599 600 601 602 603 604

	/*
	 * 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);
605
	bp->b_iodone = xfs_buf_iodone_callbacks;
606
	bip->bli_item.li_cb = xfs_buf_iodone;
L
Linus Torvalds 已提交
607

C
Christoph Hellwig 已提交
608 609
	trace_xfs_trans_log_buf(bip);

L
Linus Torvalds 已提交
610 611 612 613 614 615 616 617 618 619
	/*
	 * 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);
620
		bip->bli_format.blf_flags &= ~XFS_BLF_CANCEL;
L
Linus Torvalds 已提交
621 622 623
	}

	tp->t_flags |= XFS_TRANS_DIRTY;
624
	bip->bli_item.li_desc->lid_flags |= XFS_LID_DIRTY;
L
Linus Torvalds 已提交
625 626 627 628 629 630
	bip->bli_flags |= XFS_BLI_LOGGED;
	xfs_buf_item_log(bip, first, last);
}


/*
631 632 633 634 635 636
 * 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
637
 *
638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657
 * 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
658 659 660 661 662 663
 */
void
xfs_trans_binval(
	xfs_trans_t	*tp,
	xfs_buf_t	*bp)
{
664
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
665

666
	ASSERT(bp->b_transp == tp);
667
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
668 669
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

C
Christoph Hellwig 已提交
670 671
	trace_xfs_trans_binval(bip);

L
Linus Torvalds 已提交
672 673 674 675 676 677 678
	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)));
679 680
		ASSERT(!(bip->bli_format.blf_flags & XFS_BLF_INODE_BUF));
		ASSERT(bip->bli_format.blf_flags & XFS_BLF_CANCEL);
681
		ASSERT(bip->bli_item.li_desc->lid_flags & XFS_LID_DIRTY);
L
Linus Torvalds 已提交
682 683 684 685
		ASSERT(tp->t_flags & XFS_TRANS_DIRTY);
		return;
	}

686
	xfs_buf_stale(bp);
687

L
Linus Torvalds 已提交
688
	bip->bli_flags |= XFS_BLI_STALE;
689
	bip->bli_flags &= ~(XFS_BLI_INODE_BUF | XFS_BLI_LOGGED | XFS_BLI_DIRTY);
690 691
	bip->bli_format.blf_flags &= ~XFS_BLF_INODE_BUF;
	bip->bli_format.blf_flags |= XFS_BLF_CANCEL;
L
Linus Torvalds 已提交
692 693
	memset((char *)(bip->bli_format.blf_data_map), 0,
	      (bip->bli_format.blf_map_size * sizeof(uint)));
694
	bip->bli_item.li_desc->lid_flags |= XFS_LID_DIRTY;
L
Linus Torvalds 已提交
695 696 697 698
	tp->t_flags |= XFS_TRANS_DIRTY;
}

/*
699 700 701 702 703
 * 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
704
 *
705 706 707
 * 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
708 709 710 711 712 713
 */
void
xfs_trans_inode_buf(
	xfs_trans_t	*tp,
	xfs_buf_t	*bp)
{
714
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
715

716
	ASSERT(bp->b_transp == tp);
717
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
718 719
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

720
	bip->bli_flags |= XFS_BLI_INODE_BUF;
L
Linus Torvalds 已提交
721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736
}

/*
 * 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)
{
737
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
738

739
	ASSERT(bp->b_transp == tp);
740
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
741 742 743
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

	bip->bli_flags |= XFS_BLI_STALE_INODE;
744
	bip->bli_item.li_cb = xfs_buf_iodone;
L
Linus Torvalds 已提交
745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760
}

/*
 * 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)
{
761
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
762

763
	ASSERT(bp->b_transp == tp);
764
	ASSERT(bip != NULL);
L
Linus Torvalds 已提交
765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787
	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)
{
788
	xfs_buf_log_item_t	*bip = bp->b_fspriv;
L
Linus Torvalds 已提交
789

790
	ASSERT(bp->b_transp == tp);
791
	ASSERT(bip != NULL);
792 793 794
	ASSERT(type == XFS_BLF_UDQUOT_BUF ||
	       type == XFS_BLF_PDQUOT_BUF ||
	       type == XFS_BLF_GDQUOT_BUF);
L
Linus Torvalds 已提交
795 796 797 798
	ASSERT(atomic_read(&bip->bli_refcount) > 0);

	bip->bli_format.blf_flags |= type;
}