xfs_inode.c 111.3 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
3
 * 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 19
#include <linux/log2.h>

L
Linus Torvalds 已提交
20
#include "xfs.h"
21
#include "xfs_fs.h"
L
Linus Torvalds 已提交
22
#include "xfs_types.h"
23
#include "xfs_bit.h"
L
Linus Torvalds 已提交
24
#include "xfs_log.h"
25
#include "xfs_inum.h"
L
Linus Torvalds 已提交
26 27 28 29 30 31
#include "xfs_trans.h"
#include "xfs_trans_priv.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_mount.h"
#include "xfs_bmap_btree.h"
32
#include "xfs_alloc_btree.h"
L
Linus Torvalds 已提交
33
#include "xfs_ialloc_btree.h"
34
#include "xfs_attr_sf.h"
L
Linus Torvalds 已提交
35 36 37
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_buf_item.h"
38 39 40 41 42
#include "xfs_inode_item.h"
#include "xfs_btree.h"
#include "xfs_alloc.h"
#include "xfs_ialloc.h"
#include "xfs_bmap.h"
L
Linus Torvalds 已提交
43 44 45
#include "xfs_error.h"
#include "xfs_utils.h"
#include "xfs_quota.h"
46
#include "xfs_filestream.h"
47
#include "xfs_vnodeops.h"
C
Christoph Hellwig 已提交
48
#include "xfs_trace.h"
L
Linus Torvalds 已提交
49 50 51 52 53

kmem_zone_t *xfs_ifork_zone;
kmem_zone_t *xfs_inode_zone;

/*
54
 * Used in xfs_itruncate_extents().  This is the maximum number of extents
L
Linus Torvalds 已提交
55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70
 * freed from a file in a single transaction.
 */
#define	XFS_ITRUNC_MAX_EXTENTS	2

STATIC int xfs_iflush_int(xfs_inode_t *, xfs_buf_t *);
STATIC int xfs_iformat_local(xfs_inode_t *, xfs_dinode_t *, int, int);
STATIC int xfs_iformat_extents(xfs_inode_t *, xfs_dinode_t *, int);
STATIC int xfs_iformat_btree(xfs_inode_t *, xfs_dinode_t *, int);

#ifdef DEBUG
/*
 * Make sure that the extents in the given memory buffer
 * are valid.
 */
STATIC void
xfs_validate_extents(
71
	xfs_ifork_t		*ifp,
L
Linus Torvalds 已提交
72 73 74 75
	int			nrecs,
	xfs_exntfmt_t		fmt)
{
	xfs_bmbt_irec_t		irec;
76
	xfs_bmbt_rec_host_t	rec;
L
Linus Torvalds 已提交
77 78 79
	int			i;

	for (i = 0; i < nrecs; i++) {
80 81 82 83
		xfs_bmbt_rec_host_t *ep = xfs_iext_get_ext(ifp, i);
		rec.l0 = get_unaligned(&ep->l0);
		rec.l1 = get_unaligned(&ep->l1);
		xfs_bmbt_get_all(&rec, &irec);
L
Linus Torvalds 已提交
84 85 86 87 88
		if (fmt == XFS_EXTFMT_NOSTATE)
			ASSERT(irec.br_state == XFS_EXT_NORM);
	}
}
#else /* DEBUG */
89
#define xfs_validate_extents(ifp, nrecs, fmt)
L
Linus Torvalds 已提交
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111
#endif /* DEBUG */

/*
 * Check that none of the inode's in the buffer have a next
 * unlinked field of 0.
 */
#if defined(DEBUG)
void
xfs_inobp_check(
	xfs_mount_t	*mp,
	xfs_buf_t	*bp)
{
	int		i;
	int		j;
	xfs_dinode_t	*dip;

	j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog;

	for (i = 0; i < j; i++) {
		dip = (xfs_dinode_t *)xfs_buf_offset(bp,
					i * mp->m_sb.sb_inodesize);
		if (!dip->di_next_unlinked)  {
112 113
			xfs_alert(mp,
	"Detected bogus zero next_unlinked field in incore inode buffer 0x%p.",
L
Linus Torvalds 已提交
114 115 116 117 118 119 120
				bp);
			ASSERT(dip->di_next_unlinked);
		}
	}
}
#endif

121 122 123 124 125 126 127 128 129
/*
 * Find the buffer associated with the given inode map
 * We do basic validation checks on the buffer once it has been
 * retrieved from disk.
 */
STATIC int
xfs_imap_to_bp(
	xfs_mount_t	*mp,
	xfs_trans_t	*tp,
130
	struct xfs_imap	*imap,
131 132
	xfs_buf_t	**bpp,
	uint		buf_flags,
133
	uint		iget_flags)
134 135 136 137 138 139 140
{
	int		error;
	int		i;
	int		ni;
	xfs_buf_t	*bp;

	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
141
				   (int)imap->im_len, buf_flags, &bp);
142
	if (error) {
143
		if (error != EAGAIN) {
144 145 146
			xfs_warn(mp,
				"%s: xfs_trans_read_buf() returned error %d.",
				__func__, error);
147
		} else {
148
			ASSERT(buf_flags & XBF_TRYLOCK);
149
		}
150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168
		return error;
	}

	/*
	 * Validate the magic number and version of every inode in the buffer
	 * (if DEBUG kernel) or the first inode in the buffer, otherwise.
	 */
#ifdef DEBUG
	ni = BBTOB(imap->im_len) >> mp->m_sb.sb_inodelog;
#else	/* usual case */
	ni = 1;
#endif

	for (i = 0; i < ni; i++) {
		int		di_ok;
		xfs_dinode_t	*dip;

		dip = (xfs_dinode_t *)xfs_buf_offset(bp,
					(i << mp->m_sb.sb_inodelog));
169
		di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
C
Christoph Hellwig 已提交
170
			    XFS_DINODE_GOOD_VERSION(dip->di_version);
171 172 173
		if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
						XFS_ERRTAG_ITOBP_INOTOBP,
						XFS_RANDOM_ITOBP_INOTOBP))) {
174
			if (iget_flags & XFS_IGET_UNTRUSTED) {
175 176 177 178 179 180
				xfs_trans_brelse(tp, bp);
				return XFS_ERROR(EINVAL);
			}
			XFS_CORRUPTION_ERROR("xfs_imap_to_bp",
						XFS_ERRLEVEL_HIGH, mp, dip);
#ifdef DEBUG
181 182
			xfs_emerg(mp,
				"bad inode magic/vsn daddr %lld #%d (magic=%x)",
183
				(unsigned long long)imap->im_blkno, i,
C
Christoph Hellwig 已提交
184
				be16_to_cpu(dip->di_magic));
185
			ASSERT(0);
186 187 188 189 190 191 192 193 194 195 196
#endif
			xfs_trans_brelse(tp, bp);
			return XFS_ERROR(EFSCORRUPTED);
		}
	}

	xfs_inobp_check(mp, bp);
	*bpp = bp;
	return 0;
}

L
Linus Torvalds 已提交
197 198 199 200 201 202 203 204 205 206 207 208 209
/*
 * This routine is called to map an inode number within a file
 * system to the buffer containing the on-disk version of the
 * inode.  It returns a pointer to the buffer containing the
 * on-disk inode in the bpp parameter, and in the dip parameter
 * it returns a pointer to the on-disk inode within that buffer.
 *
 * If a non-zero error is returned, then the contents of bpp and
 * dipp are undefined.
 *
 * Use xfs_imap() to determine the size and location of the
 * buffer to read from disk.
 */
210
int
L
Linus Torvalds 已提交
211 212 213 214 215 216
xfs_inotobp(
	xfs_mount_t	*mp,
	xfs_trans_t	*tp,
	xfs_ino_t	ino,
	xfs_dinode_t	**dipp,
	xfs_buf_t	**bpp,
217 218
	int		*offset,
	uint		imap_flags)
L
Linus Torvalds 已提交
219
{
220
	struct xfs_imap	imap;
L
Linus Torvalds 已提交
221 222 223 224
	xfs_buf_t	*bp;
	int		error;

	imap.im_blkno = 0;
225
	error = xfs_imap(mp, tp, ino, &imap, imap_flags);
226
	if (error)
L
Linus Torvalds 已提交
227 228
		return error;

229
	error = xfs_imap_to_bp(mp, tp, &imap, &bp, XBF_LOCK, imap_flags);
230
	if (error)
L
Linus Torvalds 已提交
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249
		return error;

	*dipp = (xfs_dinode_t *)xfs_buf_offset(bp, imap.im_boffset);
	*bpp = bp;
	*offset = imap.im_boffset;
	return 0;
}


/*
 * This routine is called to map an inode to the buffer containing
 * the on-disk version of the inode.  It returns a pointer to the
 * buffer containing the on-disk inode in the bpp parameter, and in
 * the dip parameter it returns a pointer to the on-disk inode within
 * that buffer.
 *
 * If a non-zero error is returned, then the contents of bpp and
 * dipp are undefined.
 *
250 251 252 253
 * The inode is expected to already been mapped to its buffer and read
 * in once, thus we can use the mapping information stored in the inode
 * rather than calling xfs_imap().  This allows us to avoid the overhead
 * of looking at the inode btree for small block file systems
254
 * (see xfs_imap()).
L
Linus Torvalds 已提交
255 256 257 258 259 260 261 262
 */
int
xfs_itobp(
	xfs_mount_t	*mp,
	xfs_trans_t	*tp,
	xfs_inode_t	*ip,
	xfs_dinode_t	**dipp,
	xfs_buf_t	**bpp,
263
	uint		buf_flags)
L
Linus Torvalds 已提交
264 265 266 267
{
	xfs_buf_t	*bp;
	int		error;

268
	ASSERT(ip->i_imap.im_blkno != 0);
L
Linus Torvalds 已提交
269

270
	error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &bp, buf_flags, 0);
271
	if (error)
L
Linus Torvalds 已提交
272 273
		return error;

274
	if (!bp) {
275
		ASSERT(buf_flags & XBF_TRYLOCK);
276 277 278 279 280
		ASSERT(tp == NULL);
		*bpp = NULL;
		return EAGAIN;
	}

281
	*dipp = (xfs_dinode_t *)xfs_buf_offset(bp, ip->i_imap.im_boffset);
L
Linus Torvalds 已提交
282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307
	*bpp = bp;
	return 0;
}

/*
 * Move inode type and inode format specific information from the
 * on-disk inode to the in-core inode.  For fifos, devs, and sockets
 * this means set if_rdev to the proper value.  For files, directories,
 * and symlinks this means to bring in the in-line data or extent
 * pointers.  For a file in B-tree format, only the root is immediately
 * brought in-core.  The rest will be in-lined in if_extents when it
 * is first referenced (see xfs_iread_extents()).
 */
STATIC int
xfs_iformat(
	xfs_inode_t		*ip,
	xfs_dinode_t		*dip)
{
	xfs_attr_shortform_t	*atp;
	int			size;
	int			error;
	xfs_fsize_t             di_size;
	ip->i_df.if_ext_max =
		XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t);
	error = 0;

C
Christoph Hellwig 已提交
308 309 310
	if (unlikely(be32_to_cpu(dip->di_nextents) +
		     be16_to_cpu(dip->di_anextents) >
		     be64_to_cpu(dip->di_nblocks))) {
311
		xfs_warn(ip->i_mount,
312
			"corrupt dinode %Lu, extent total = %d, nblocks = %Lu.",
L
Linus Torvalds 已提交
313
			(unsigned long long)ip->i_ino,
C
Christoph Hellwig 已提交
314 315
			(int)(be32_to_cpu(dip->di_nextents) +
			      be16_to_cpu(dip->di_anextents)),
L
Linus Torvalds 已提交
316
			(unsigned long long)
C
Christoph Hellwig 已提交
317
				be64_to_cpu(dip->di_nblocks));
L
Linus Torvalds 已提交
318 319 320 321 322
		XFS_CORRUPTION_ERROR("xfs_iformat(1)", XFS_ERRLEVEL_LOW,
				     ip->i_mount, dip);
		return XFS_ERROR(EFSCORRUPTED);
	}

C
Christoph Hellwig 已提交
323
	if (unlikely(dip->di_forkoff > ip->i_mount->m_sb.sb_inodesize)) {
324
		xfs_warn(ip->i_mount, "corrupt dinode %Lu, forkoff = 0x%x.",
L
Linus Torvalds 已提交
325
			(unsigned long long)ip->i_ino,
C
Christoph Hellwig 已提交
326
			dip->di_forkoff);
L
Linus Torvalds 已提交
327 328 329 330 331
		XFS_CORRUPTION_ERROR("xfs_iformat(2)", XFS_ERRLEVEL_LOW,
				     ip->i_mount, dip);
		return XFS_ERROR(EFSCORRUPTED);
	}

332 333
	if (unlikely((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) &&
		     !ip->i_mount->m_rtdev_targp)) {
334
		xfs_warn(ip->i_mount,
335 336 337 338 339 340 341
			"corrupt dinode %Lu, has realtime flag set.",
			ip->i_ino);
		XFS_CORRUPTION_ERROR("xfs_iformat(realtime)",
				     XFS_ERRLEVEL_LOW, ip->i_mount, dip);
		return XFS_ERROR(EFSCORRUPTED);
	}

L
Linus Torvalds 已提交
342 343 344 345 346
	switch (ip->i_d.di_mode & S_IFMT) {
	case S_IFIFO:
	case S_IFCHR:
	case S_IFBLK:
	case S_IFSOCK:
C
Christoph Hellwig 已提交
347
		if (unlikely(dip->di_format != XFS_DINODE_FMT_DEV)) {
L
Linus Torvalds 已提交
348 349 350 351 352
			XFS_CORRUPTION_ERROR("xfs_iformat(3)", XFS_ERRLEVEL_LOW,
					      ip->i_mount, dip);
			return XFS_ERROR(EFSCORRUPTED);
		}
		ip->i_d.di_size = 0;
353
		ip->i_size = 0;
C
Christoph Hellwig 已提交
354
		ip->i_df.if_u2.if_rdev = xfs_dinode_get_rdev(dip);
L
Linus Torvalds 已提交
355 356 357 358 359
		break;

	case S_IFREG:
	case S_IFLNK:
	case S_IFDIR:
C
Christoph Hellwig 已提交
360
		switch (dip->di_format) {
L
Linus Torvalds 已提交
361 362 363 364
		case XFS_DINODE_FMT_LOCAL:
			/*
			 * no local regular files yet
			 */
365
			if (unlikely(S_ISREG(be16_to_cpu(dip->di_mode)))) {
366 367
				xfs_warn(ip->i_mount,
			"corrupt inode %Lu (local format for regular file).",
L
Linus Torvalds 已提交
368 369 370 371 372 373 374
					(unsigned long long) ip->i_ino);
				XFS_CORRUPTION_ERROR("xfs_iformat(4)",
						     XFS_ERRLEVEL_LOW,
						     ip->i_mount, dip);
				return XFS_ERROR(EFSCORRUPTED);
			}

C
Christoph Hellwig 已提交
375
			di_size = be64_to_cpu(dip->di_size);
L
Linus Torvalds 已提交
376
			if (unlikely(di_size > XFS_DFORK_DSIZE(dip, ip->i_mount))) {
377 378
				xfs_warn(ip->i_mount,
			"corrupt inode %Lu (bad size %Ld for local inode).",
L
Linus Torvalds 已提交
379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412
					(unsigned long long) ip->i_ino,
					(long long) di_size);
				XFS_CORRUPTION_ERROR("xfs_iformat(5)",
						     XFS_ERRLEVEL_LOW,
						     ip->i_mount, dip);
				return XFS_ERROR(EFSCORRUPTED);
			}

			size = (int)di_size;
			error = xfs_iformat_local(ip, dip, XFS_DATA_FORK, size);
			break;
		case XFS_DINODE_FMT_EXTENTS:
			error = xfs_iformat_extents(ip, dip, XFS_DATA_FORK);
			break;
		case XFS_DINODE_FMT_BTREE:
			error = xfs_iformat_btree(ip, dip, XFS_DATA_FORK);
			break;
		default:
			XFS_ERROR_REPORT("xfs_iformat(6)", XFS_ERRLEVEL_LOW,
					 ip->i_mount);
			return XFS_ERROR(EFSCORRUPTED);
		}
		break;

	default:
		XFS_ERROR_REPORT("xfs_iformat(7)", XFS_ERRLEVEL_LOW, ip->i_mount);
		return XFS_ERROR(EFSCORRUPTED);
	}
	if (error) {
		return error;
	}
	if (!XFS_DFORK_Q(dip))
		return 0;
	ASSERT(ip->i_afp == NULL);
413
	ip->i_afp = kmem_zone_zalloc(xfs_ifork_zone, KM_SLEEP | KM_NOFS);
L
Linus Torvalds 已提交
414 415
	ip->i_afp->if_ext_max =
		XFS_IFORK_ASIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t);
C
Christoph Hellwig 已提交
416
	switch (dip->di_aformat) {
L
Linus Torvalds 已提交
417 418
	case XFS_DINODE_FMT_LOCAL:
		atp = (xfs_attr_shortform_t *)XFS_DFORK_APTR(dip);
419
		size = be16_to_cpu(atp->hdr.totsize);
420 421

		if (unlikely(size < sizeof(struct xfs_attr_sf_hdr))) {
422 423
			xfs_warn(ip->i_mount,
				"corrupt inode %Lu (bad attr fork size %Ld).",
424 425 426 427 428 429 430 431
				(unsigned long long) ip->i_ino,
				(long long) size);
			XFS_CORRUPTION_ERROR("xfs_iformat(8)",
					     XFS_ERRLEVEL_LOW,
					     ip->i_mount, dip);
			return XFS_ERROR(EFSCORRUPTED);
		}

L
Linus Torvalds 已提交
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477
		error = xfs_iformat_local(ip, dip, XFS_ATTR_FORK, size);
		break;
	case XFS_DINODE_FMT_EXTENTS:
		error = xfs_iformat_extents(ip, dip, XFS_ATTR_FORK);
		break;
	case XFS_DINODE_FMT_BTREE:
		error = xfs_iformat_btree(ip, dip, XFS_ATTR_FORK);
		break;
	default:
		error = XFS_ERROR(EFSCORRUPTED);
		break;
	}
	if (error) {
		kmem_zone_free(xfs_ifork_zone, ip->i_afp);
		ip->i_afp = NULL;
		xfs_idestroy_fork(ip, XFS_DATA_FORK);
	}
	return error;
}

/*
 * The file is in-lined in the on-disk inode.
 * If it fits into if_inline_data, then copy
 * it there, otherwise allocate a buffer for it
 * and copy the data there.  Either way, set
 * if_data to point at the data.
 * If we allocate a buffer for the data, make
 * sure that its size is a multiple of 4 and
 * record the real size in i_real_bytes.
 */
STATIC int
xfs_iformat_local(
	xfs_inode_t	*ip,
	xfs_dinode_t	*dip,
	int		whichfork,
	int		size)
{
	xfs_ifork_t	*ifp;
	int		real_size;

	/*
	 * If the size is unreasonable, then something
	 * is wrong and we just bail out rather than crash in
	 * kmem_alloc() or memcpy() below.
	 */
	if (unlikely(size > XFS_DFORK_SIZE(dip, ip->i_mount, whichfork))) {
478 479
		xfs_warn(ip->i_mount,
	"corrupt inode %Lu (bad size %d for local fork, size = %d).",
L
Linus Torvalds 已提交
480 481 482 483 484 485 486 487 488 489 490 491 492 493
			(unsigned long long) ip->i_ino, size,
			XFS_DFORK_SIZE(dip, ip->i_mount, whichfork));
		XFS_CORRUPTION_ERROR("xfs_iformat_local", XFS_ERRLEVEL_LOW,
				     ip->i_mount, dip);
		return XFS_ERROR(EFSCORRUPTED);
	}
	ifp = XFS_IFORK_PTR(ip, whichfork);
	real_size = 0;
	if (size == 0)
		ifp->if_u1.if_data = NULL;
	else if (size <= sizeof(ifp->if_u2.if_inline_data))
		ifp->if_u1.if_data = ifp->if_u2.if_inline_data;
	else {
		real_size = roundup(size, 4);
494
		ifp->if_u1.if_data = kmem_alloc(real_size, KM_SLEEP | KM_NOFS);
L
Linus Torvalds 已提交
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
	}
	ifp->if_bytes = size;
	ifp->if_real_bytes = real_size;
	if (size)
		memcpy(ifp->if_u1.if_data, XFS_DFORK_PTR(dip, whichfork), size);
	ifp->if_flags &= ~XFS_IFEXTENTS;
	ifp->if_flags |= XFS_IFINLINE;
	return 0;
}

/*
 * The file consists of a set of extents all
 * of which fit into the on-disk inode.
 * If there are few enough extents to fit into
 * the if_inline_ext, then copy them there.
 * Otherwise allocate a buffer for them and copy
 * them into it.  Either way, set if_extents
 * to point at the extents.
 */
STATIC int
xfs_iformat_extents(
	xfs_inode_t	*ip,
	xfs_dinode_t	*dip,
	int		whichfork)
{
520
	xfs_bmbt_rec_t	*dp;
L
Linus Torvalds 已提交
521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
	xfs_ifork_t	*ifp;
	int		nex;
	int		size;
	int		i;

	ifp = XFS_IFORK_PTR(ip, whichfork);
	nex = XFS_DFORK_NEXTENTS(dip, whichfork);
	size = nex * (uint)sizeof(xfs_bmbt_rec_t);

	/*
	 * If the number of extents is unreasonable, then something
	 * is wrong and we just bail out rather than crash in
	 * kmem_alloc() or memcpy() below.
	 */
	if (unlikely(size < 0 || size > XFS_DFORK_SIZE(dip, ip->i_mount, whichfork))) {
536
		xfs_warn(ip->i_mount, "corrupt inode %Lu ((a)extents = %d).",
L
Linus Torvalds 已提交
537 538 539 540 541 542
			(unsigned long long) ip->i_ino, nex);
		XFS_CORRUPTION_ERROR("xfs_iformat_extents(1)", XFS_ERRLEVEL_LOW,
				     ip->i_mount, dip);
		return XFS_ERROR(EFSCORRUPTED);
	}

543
	ifp->if_real_bytes = 0;
L
Linus Torvalds 已提交
544 545 546 547
	if (nex == 0)
		ifp->if_u1.if_extents = NULL;
	else if (nex <= XFS_INLINE_EXTS)
		ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext;
548 549 550
	else
		xfs_iext_add(ifp, 0, nex);

L
Linus Torvalds 已提交
551 552 553
	ifp->if_bytes = size;
	if (size) {
		dp = (xfs_bmbt_rec_t *) XFS_DFORK_PTR(dip, whichfork);
554
		xfs_validate_extents(ifp, nex, XFS_EXTFMT_INODE(ip));
555
		for (i = 0; i < nex; i++, dp++) {
556
			xfs_bmbt_rec_host_t *ep = xfs_iext_get_ext(ifp, i);
557 558
			ep->l0 = get_unaligned_be64(&dp->l0);
			ep->l1 = get_unaligned_be64(&dp->l1);
L
Linus Torvalds 已提交
559
		}
560
		XFS_BMAP_TRACE_EXLIST(ip, nex, whichfork);
L
Linus Torvalds 已提交
561 562 563
		if (whichfork != XFS_DATA_FORK ||
			XFS_EXTFMT_INODE(ip) == XFS_EXTFMT_NOSTATE)
				if (unlikely(xfs_check_nostate_extents(
564
				    ifp, 0, nex))) {
L
Linus Torvalds 已提交
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597
					XFS_ERROR_REPORT("xfs_iformat_extents(2)",
							 XFS_ERRLEVEL_LOW,
							 ip->i_mount);
					return XFS_ERROR(EFSCORRUPTED);
				}
	}
	ifp->if_flags |= XFS_IFEXTENTS;
	return 0;
}

/*
 * The file has too many extents to fit into
 * the inode, so they are in B-tree format.
 * Allocate a buffer for the root of the B-tree
 * and copy the root into it.  The i_extents
 * field will remain NULL until all of the
 * extents are read in (when they are needed).
 */
STATIC int
xfs_iformat_btree(
	xfs_inode_t		*ip,
	xfs_dinode_t		*dip,
	int			whichfork)
{
	xfs_bmdr_block_t	*dfp;
	xfs_ifork_t		*ifp;
	/* REFERENCED */
	int			nrecs;
	int			size;

	ifp = XFS_IFORK_PTR(ip, whichfork);
	dfp = (xfs_bmdr_block_t *)XFS_DFORK_PTR(dip, whichfork);
	size = XFS_BMAP_BROOT_SPACE(dfp);
598
	nrecs = be16_to_cpu(dfp->bb_numrecs);
L
Linus Torvalds 已提交
599 600 601 602 603 604 605 606 607 608 609 610

	/*
	 * blow out if -- fork has less extents than can fit in
	 * fork (fork shouldn't be a btree format), root btree
	 * block has more records than can fit into the fork,
	 * or the number of extents is greater than the number of
	 * blocks.
	 */
	if (unlikely(XFS_IFORK_NEXTENTS(ip, whichfork) <= ifp->if_ext_max
	    || XFS_BMDR_SPACE_CALC(nrecs) >
			XFS_DFORK_SIZE(dip, ip->i_mount, whichfork)
	    || XFS_IFORK_NEXTENTS(ip, whichfork) > ip->i_d.di_nblocks)) {
611
		xfs_warn(ip->i_mount, "corrupt inode %Lu (btree).",
L
Linus Torvalds 已提交
612
			(unsigned long long) ip->i_ino);
613 614
		XFS_CORRUPTION_ERROR("xfs_iformat_btree", XFS_ERRLEVEL_LOW,
				 ip->i_mount, dip);
L
Linus Torvalds 已提交
615 616 617 618
		return XFS_ERROR(EFSCORRUPTED);
	}

	ifp->if_broot_bytes = size;
619
	ifp->if_broot = kmem_alloc(size, KM_SLEEP | KM_NOFS);
L
Linus Torvalds 已提交
620 621 622 623 624
	ASSERT(ifp->if_broot != NULL);
	/*
	 * Copy and convert from the on-disk structure
	 * to the in-memory structure.
	 */
625 626 627
	xfs_bmdr_to_bmbt(ip->i_mount, dfp,
			 XFS_DFORK_SIZE(dip, ip->i_mount, whichfork),
			 ifp->if_broot, size);
L
Linus Torvalds 已提交
628 629 630 631 632 633
	ifp->if_flags &= ~XFS_IFEXTENTS;
	ifp->if_flags |= XFS_IFBROOT;

	return 0;
}

634
STATIC void
635 636
xfs_dinode_from_disk(
	xfs_icdinode_t		*to,
C
Christoph Hellwig 已提交
637
	xfs_dinode_t		*from)
L
Linus Torvalds 已提交
638
{
639 640 641 642 643 644 645 646
	to->di_magic = be16_to_cpu(from->di_magic);
	to->di_mode = be16_to_cpu(from->di_mode);
	to->di_version = from ->di_version;
	to->di_format = from->di_format;
	to->di_onlink = be16_to_cpu(from->di_onlink);
	to->di_uid = be32_to_cpu(from->di_uid);
	to->di_gid = be32_to_cpu(from->di_gid);
	to->di_nlink = be32_to_cpu(from->di_nlink);
647 648
	to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
	to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
	memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
	to->di_flushiter = be16_to_cpu(from->di_flushiter);
	to->di_atime.t_sec = be32_to_cpu(from->di_atime.t_sec);
	to->di_atime.t_nsec = be32_to_cpu(from->di_atime.t_nsec);
	to->di_mtime.t_sec = be32_to_cpu(from->di_mtime.t_sec);
	to->di_mtime.t_nsec = be32_to_cpu(from->di_mtime.t_nsec);
	to->di_ctime.t_sec = be32_to_cpu(from->di_ctime.t_sec);
	to->di_ctime.t_nsec = be32_to_cpu(from->di_ctime.t_nsec);
	to->di_size = be64_to_cpu(from->di_size);
	to->di_nblocks = be64_to_cpu(from->di_nblocks);
	to->di_extsize = be32_to_cpu(from->di_extsize);
	to->di_nextents = be32_to_cpu(from->di_nextents);
	to->di_anextents = be16_to_cpu(from->di_anextents);
	to->di_forkoff = from->di_forkoff;
	to->di_aformat	= from->di_aformat;
	to->di_dmevmask	= be32_to_cpu(from->di_dmevmask);
	to->di_dmstate	= be16_to_cpu(from->di_dmstate);
	to->di_flags	= be16_to_cpu(from->di_flags);
	to->di_gen	= be32_to_cpu(from->di_gen);
}

void
xfs_dinode_to_disk(
C
Christoph Hellwig 已提交
672
	xfs_dinode_t		*to,
673 674 675 676 677 678 679 680 681 682
	xfs_icdinode_t		*from)
{
	to->di_magic = cpu_to_be16(from->di_magic);
	to->di_mode = cpu_to_be16(from->di_mode);
	to->di_version = from ->di_version;
	to->di_format = from->di_format;
	to->di_onlink = cpu_to_be16(from->di_onlink);
	to->di_uid = cpu_to_be32(from->di_uid);
	to->di_gid = cpu_to_be32(from->di_gid);
	to->di_nlink = cpu_to_be32(from->di_nlink);
683 684
	to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
	to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703
	memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
	to->di_flushiter = cpu_to_be16(from->di_flushiter);
	to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
	to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
	to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
	to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
	to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
	to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
	to->di_size = cpu_to_be64(from->di_size);
	to->di_nblocks = cpu_to_be64(from->di_nblocks);
	to->di_extsize = cpu_to_be32(from->di_extsize);
	to->di_nextents = cpu_to_be32(from->di_nextents);
	to->di_anextents = cpu_to_be16(from->di_anextents);
	to->di_forkoff = from->di_forkoff;
	to->di_aformat = from->di_aformat;
	to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
	to->di_dmstate = cpu_to_be16(from->di_dmstate);
	to->di_flags = cpu_to_be16(from->di_flags);
	to->di_gen = cpu_to_be32(from->di_gen);
L
Linus Torvalds 已提交
704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732
}

STATIC uint
_xfs_dic2xflags(
	__uint16_t		di_flags)
{
	uint			flags = 0;

	if (di_flags & XFS_DIFLAG_ANY) {
		if (di_flags & XFS_DIFLAG_REALTIME)
			flags |= XFS_XFLAG_REALTIME;
		if (di_flags & XFS_DIFLAG_PREALLOC)
			flags |= XFS_XFLAG_PREALLOC;
		if (di_flags & XFS_DIFLAG_IMMUTABLE)
			flags |= XFS_XFLAG_IMMUTABLE;
		if (di_flags & XFS_DIFLAG_APPEND)
			flags |= XFS_XFLAG_APPEND;
		if (di_flags & XFS_DIFLAG_SYNC)
			flags |= XFS_XFLAG_SYNC;
		if (di_flags & XFS_DIFLAG_NOATIME)
			flags |= XFS_XFLAG_NOATIME;
		if (di_flags & XFS_DIFLAG_NODUMP)
			flags |= XFS_XFLAG_NODUMP;
		if (di_flags & XFS_DIFLAG_RTINHERIT)
			flags |= XFS_XFLAG_RTINHERIT;
		if (di_flags & XFS_DIFLAG_PROJINHERIT)
			flags |= XFS_XFLAG_PROJINHERIT;
		if (di_flags & XFS_DIFLAG_NOSYMLINKS)
			flags |= XFS_XFLAG_NOSYMLINKS;
733 734 735 736
		if (di_flags & XFS_DIFLAG_EXTSIZE)
			flags |= XFS_XFLAG_EXTSIZE;
		if (di_flags & XFS_DIFLAG_EXTSZINHERIT)
			flags |= XFS_XFLAG_EXTSZINHERIT;
737 738
		if (di_flags & XFS_DIFLAG_NODEFRAG)
			flags |= XFS_XFLAG_NODEFRAG;
739 740
		if (di_flags & XFS_DIFLAG_FILESTREAM)
			flags |= XFS_XFLAG_FILESTREAM;
L
Linus Torvalds 已提交
741 742 743 744 745 746 747 748 749
	}

	return flags;
}

uint
xfs_ip2xflags(
	xfs_inode_t		*ip)
{
750
	xfs_icdinode_t		*dic = &ip->i_d;
L
Linus Torvalds 已提交
751

752
	return _xfs_dic2xflags(dic->di_flags) |
753
				(XFS_IFORK_Q(ip) ? XFS_XFLAG_HASATTR : 0);
L
Linus Torvalds 已提交
754 755 756 757
}

uint
xfs_dic2xflags(
758
	xfs_dinode_t		*dip)
L
Linus Torvalds 已提交
759
{
C
Christoph Hellwig 已提交
760
	return _xfs_dic2xflags(be16_to_cpu(dip->di_flags)) |
761
				(XFS_DFORK_Q(dip) ? XFS_XFLAG_HASATTR : 0);
L
Linus Torvalds 已提交
762 763
}

764
/*
765
 * Read the disk inode attributes into the in-core inode structure.
L
Linus Torvalds 已提交
766 767 768 769 770
 */
int
xfs_iread(
	xfs_mount_t	*mp,
	xfs_trans_t	*tp,
771 772
	xfs_inode_t	*ip,
	uint		iget_flags)
L
Linus Torvalds 已提交
773 774 775 776 777 778
{
	xfs_buf_t	*bp;
	xfs_dinode_t	*dip;
	int		error;

	/*
779
	 * Fill in the location information in the in-core inode.
L
Linus Torvalds 已提交
780
	 */
781
	error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
782
	if (error)
783
		return error;
784 785

	/*
786
	 * Get pointers to the on-disk inode and the buffer containing it.
787
	 */
788
	error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &bp,
789
			       XBF_LOCK, iget_flags);
790
	if (error)
791
		return error;
792
	dip = (xfs_dinode_t *)xfs_buf_offset(bp, ip->i_imap.im_boffset);
L
Linus Torvalds 已提交
793 794 795 796 797

	/*
	 * If we got something that isn't an inode it means someone
	 * (nfs or dmi) has a stale handle.
	 */
798
	if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC)) {
L
Linus Torvalds 已提交
799
#ifdef DEBUG
800 801 802
		xfs_alert(mp,
			"%s: dip->di_magic (0x%x) != XFS_DINODE_MAGIC (0x%x)",
			__func__, be16_to_cpu(dip->di_magic), XFS_DINODE_MAGIC);
L
Linus Torvalds 已提交
803
#endif /* DEBUG */
804 805
		error = XFS_ERROR(EINVAL);
		goto out_brelse;
L
Linus Torvalds 已提交
806 807 808 809 810 811 812 813 814
	}

	/*
	 * If the on-disk inode is already linked to a directory
	 * entry, copy all of the inode into the in-core inode.
	 * xfs_iformat() handles copying in the inode format
	 * specific information.
	 * Otherwise, just get the truly permanent information.
	 */
C
Christoph Hellwig 已提交
815 816
	if (dip->di_mode) {
		xfs_dinode_from_disk(&ip->i_d, dip);
L
Linus Torvalds 已提交
817 818 819
		error = xfs_iformat(ip, dip);
		if (error)  {
#ifdef DEBUG
820 821
			xfs_alert(mp, "%s: xfs_iformat() returned error %d",
				__func__, error);
L
Linus Torvalds 已提交
822
#endif /* DEBUG */
823
			goto out_brelse;
L
Linus Torvalds 已提交
824 825
		}
	} else {
C
Christoph Hellwig 已提交
826 827 828 829
		ip->i_d.di_magic = be16_to_cpu(dip->di_magic);
		ip->i_d.di_version = dip->di_version;
		ip->i_d.di_gen = be32_to_cpu(dip->di_gen);
		ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
L
Linus Torvalds 已提交
830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856
		/*
		 * Make sure to pull in the mode here as well in
		 * case the inode is released without being used.
		 * This ensures that xfs_inactive() will see that
		 * the inode is already free and not try to mess
		 * with the uninitialized part of it.
		 */
		ip->i_d.di_mode = 0;
		/*
		 * Initialize the per-fork minima and maxima for a new
		 * inode here.  xfs_iformat will do it for old inodes.
		 */
		ip->i_df.if_ext_max =
			XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t);
	}

	/*
	 * The inode format changed when we moved the link count and
	 * made it 32 bits long.  If this is an old format inode,
	 * convert it in memory to look like a new one.  If it gets
	 * flushed to disk we will convert back before flushing or
	 * logging it.  We zero out the new projid field and the old link
	 * count field.  We'll handle clearing the pad field (the remains
	 * of the old uuid field) when we actually convert the inode to
	 * the new format. We don't change the version number so that we
	 * can distinguish this from a real new format inode.
	 */
857
	if (ip->i_d.di_version == 1) {
L
Linus Torvalds 已提交
858 859
		ip->i_d.di_nlink = ip->i_d.di_onlink;
		ip->i_d.di_onlink = 0;
860
		xfs_set_projid(ip, 0);
L
Linus Torvalds 已提交
861 862 863
	}

	ip->i_delayed_blks = 0;
864
	ip->i_size = ip->i_d.di_size;
L
Linus Torvalds 已提交
865 866 867 868 869 870

	/*
	 * Mark the buffer containing the inode as something to keep
	 * around for a while.  This helps to keep recently accessed
	 * meta-data in-core longer.
	 */
871
	xfs_buf_set_ref(bp, XFS_INO_REF);
L
Linus Torvalds 已提交
872 873 874 875 876 877 878 879 880 881 882 883 884 885

	/*
	 * Use xfs_trans_brelse() to release the buffer containing the
	 * on-disk inode, because it was acquired with xfs_trans_read_buf()
	 * in xfs_itobp() above.  If tp is NULL, this is just a normal
	 * brelse().  If we're within a transaction, then xfs_trans_brelse()
	 * will only release the buffer if it is not dirty within the
	 * transaction.  It will be OK to release the buffer in this case,
	 * because inodes on disk are never destroyed and we will be
	 * locking the new in-core inode before putting it in the hash
	 * table where other processes can find it.  Thus we don't have
	 * to worry about the inode being changed just because we released
	 * the buffer.
	 */
886 887 888
 out_brelse:
	xfs_trans_brelse(tp, bp);
	return error;
L
Linus Torvalds 已提交
889 890 891 892 893 894 895 896 897 898 899 900 901 902
}

/*
 * Read in extents from a btree-format inode.
 * Allocate and fill in if_extents.  Real work is done in xfs_bmap.c.
 */
int
xfs_iread_extents(
	xfs_trans_t	*tp,
	xfs_inode_t	*ip,
	int		whichfork)
{
	int		error;
	xfs_ifork_t	*ifp;
903
	xfs_extnum_t	nextents;
L
Linus Torvalds 已提交
904 905 906 907 908 909

	if (unlikely(XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE)) {
		XFS_ERROR_REPORT("xfs_iread_extents", XFS_ERRLEVEL_LOW,
				 ip->i_mount);
		return XFS_ERROR(EFSCORRUPTED);
	}
910
	nextents = XFS_IFORK_NEXTENTS(ip, whichfork);
L
Linus Torvalds 已提交
911
	ifp = XFS_IFORK_PTR(ip, whichfork);
912

L
Linus Torvalds 已提交
913 914 915
	/*
	 * We know that the size is valid (it's checked in iformat_btree)
	 */
916
	ifp->if_bytes = ifp->if_real_bytes = 0;
L
Linus Torvalds 已提交
917
	ifp->if_flags |= XFS_IFEXTENTS;
918
	xfs_iext_add(ifp, 0, nextents);
L
Linus Torvalds 已提交
919 920
	error = xfs_bmap_read_extents(tp, ip, whichfork);
	if (error) {
921
		xfs_iext_destroy(ifp);
L
Linus Torvalds 已提交
922 923 924
		ifp->if_flags &= ~XFS_IFEXTENTS;
		return error;
	}
925
	xfs_validate_extents(ifp, nextents, XFS_EXTFMT_INODE(ip));
L
Linus Torvalds 已提交
926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953
	return 0;
}

/*
 * Allocate an inode on disk and return a copy of its in-core version.
 * The in-core inode is locked exclusively.  Set mode, nlink, and rdev
 * appropriately within the inode.  The uid and gid for the inode are
 * set according to the contents of the given cred structure.
 *
 * Use xfs_dialloc() to allocate the on-disk inode. If xfs_dialloc()
 * has a free inode available, call xfs_iget()
 * to obtain the in-core version of the allocated inode.  Finally,
 * fill in the inode and log its initial contents.  In this case,
 * ialloc_context would be set to NULL and call_again set to false.
 *
 * If xfs_dialloc() does not have an available inode,
 * it will replenish its supply by doing an allocation. Since we can
 * only do one allocation within a transaction without deadlocks, we
 * must commit the current transaction before returning the inode itself.
 * In this case, therefore, we will set call_again to true and return.
 * The caller should then commit the current transaction, start a new
 * transaction, and call xfs_ialloc() again to actually get the inode.
 *
 * To ensure that some other process does not grab the inode that
 * was allocated during the first call to xfs_ialloc(), this routine
 * also returns the [locked] bp pointing to the head of the freelist
 * as ialloc_context.  The caller should hold this buffer across
 * the commit and pass it back into this routine on the second call.
D
David Chinner 已提交
954 955 956 957 958
 *
 * If we are allocating quota inodes, we do not have a parent inode
 * to attach to or associate with (i.e. pip == NULL) because they
 * are not linked into the directory structure - they are attached
 * directly to the superblock - and so have no parent.
L
Linus Torvalds 已提交
959 960 961 962 963
 */
int
xfs_ialloc(
	xfs_trans_t	*tp,
	xfs_inode_t	*pip,
A
Al Viro 已提交
964
	umode_t		mode,
965
	xfs_nlink_t	nlink,
L
Linus Torvalds 已提交
966
	xfs_dev_t	rdev,
967
	prid_t		prid,
L
Linus Torvalds 已提交
968 969 970 971 972 973 974 975 976
	int		okalloc,
	xfs_buf_t	**ialloc_context,
	boolean_t	*call_again,
	xfs_inode_t	**ipp)
{
	xfs_ino_t	ino;
	xfs_inode_t	*ip;
	uint		flags;
	int		error;
977
	timespec_t	tv;
978
	int		filestreams = 0;
L
Linus Torvalds 已提交
979 980 981 982 983

	/*
	 * Call the space management code to pick
	 * the on-disk inode to be allocated.
	 */
D
David Chinner 已提交
984
	error = xfs_dialloc(tp, pip ? pip->i_ino : 0, mode, okalloc,
L
Linus Torvalds 已提交
985
			    ialloc_context, call_again, &ino);
986
	if (error)
L
Linus Torvalds 已提交
987 988 989 990 991 992 993 994 995 996 997 998
		return error;
	if (*call_again || ino == NULLFSINO) {
		*ipp = NULL;
		return 0;
	}
	ASSERT(*ialloc_context == NULL);

	/*
	 * Get the in-core inode with the lock held exclusively.
	 * This is because we're setting fields here we need
	 * to prevent others from looking at until we're done.
	 */
999 1000
	error = xfs_iget(tp->t_mountp, tp, ino, XFS_IGET_CREATE,
			 XFS_ILOCK_EXCL, &ip);
1001
	if (error)
L
Linus Torvalds 已提交
1002 1003 1004
		return error;
	ASSERT(ip != NULL);

A
Al Viro 已提交
1005
	ip->i_d.di_mode = mode;
L
Linus Torvalds 已提交
1006 1007 1008
	ip->i_d.di_onlink = 0;
	ip->i_d.di_nlink = nlink;
	ASSERT(ip->i_d.di_nlink == nlink);
1009 1010
	ip->i_d.di_uid = current_fsuid();
	ip->i_d.di_gid = current_fsgid();
1011
	xfs_set_projid(ip, prid);
L
Linus Torvalds 已提交
1012 1013 1014 1015 1016 1017 1018 1019
	memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));

	/*
	 * If the superblock version is up to where we support new format
	 * inodes and this is currently an old format inode, then change
	 * the inode version number now.  This way we only do the conversion
	 * here rather than here and in the flush/logging code.
	 */
1020
	if (xfs_sb_version_hasnlink(&tp->t_mountp->m_sb) &&
1021 1022
	    ip->i_d.di_version == 1) {
		ip->i_d.di_version = 2;
L
Linus Torvalds 已提交
1023 1024 1025 1026 1027 1028 1029 1030 1031
		/*
		 * We've already zeroed the old link count, the projid field,
		 * and the pad field.
		 */
	}

	/*
	 * Project ids won't be stored on disk if we are using a version 1 inode.
	 */
1032
	if ((prid != 0) && (ip->i_d.di_version == 1))
L
Linus Torvalds 已提交
1033 1034
		xfs_bump_ino_vers2(tp, ip);

1035
	if (pip && XFS_INHERIT_GID(pip)) {
L
Linus Torvalds 已提交
1036
		ip->i_d.di_gid = pip->i_d.di_gid;
1037
		if ((pip->i_d.di_mode & S_ISGID) && S_ISDIR(mode)) {
L
Linus Torvalds 已提交
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
			ip->i_d.di_mode |= S_ISGID;
		}
	}

	/*
	 * If the group ID of the new file does not match the effective group
	 * ID or one of the supplementary group IDs, the S_ISGID bit is cleared
	 * (and only if the irix_sgid_inherit compatibility variable is set).
	 */
	if ((irix_sgid_inherit) &&
	    (ip->i_d.di_mode & S_ISGID) &&
	    (!in_group_p((gid_t)ip->i_d.di_gid))) {
		ip->i_d.di_mode &= ~S_ISGID;
	}

	ip->i_d.di_size = 0;
1054
	ip->i_size = 0;
L
Linus Torvalds 已提交
1055 1056
	ip->i_d.di_nextents = 0;
	ASSERT(ip->i_d.di_nblocks == 0);
1057 1058 1059 1060 1061 1062 1063

	nanotime(&tv);
	ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
	ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
	ip->i_d.di_atime = ip->i_d.di_mtime;
	ip->i_d.di_ctime = ip->i_d.di_mtime;

L
Linus Torvalds 已提交
1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
	/*
	 * di_gen will have been taken care of in xfs_iread.
	 */
	ip->i_d.di_extsize = 0;
	ip->i_d.di_dmevmask = 0;
	ip->i_d.di_dmstate = 0;
	ip->i_d.di_flags = 0;
	flags = XFS_ILOG_CORE;
	switch (mode & S_IFMT) {
	case S_IFIFO:
	case S_IFCHR:
	case S_IFBLK:
	case S_IFSOCK:
		ip->i_d.di_format = XFS_DINODE_FMT_DEV;
		ip->i_df.if_u2.if_rdev = rdev;
		ip->i_df.if_flags = 0;
		flags |= XFS_ILOG_DEV;
		break;
	case S_IFREG:
1083 1084 1085 1086 1087 1088
		/*
		 * we can't set up filestreams until after the VFS inode
		 * is set up properly.
		 */
		if (pip && xfs_inode_is_filestream(pip))
			filestreams = 1;
1089
		/* fall through */
L
Linus Torvalds 已提交
1090
	case S_IFDIR:
D
David Chinner 已提交
1091
		if (pip && (pip->i_d.di_flags & XFS_DIFLAG_ANY)) {
1092 1093
			uint	di_flags = 0;

1094
			if (S_ISDIR(mode)) {
1095 1096
				if (pip->i_d.di_flags & XFS_DIFLAG_RTINHERIT)
					di_flags |= XFS_DIFLAG_RTINHERIT;
1097 1098 1099 1100
				if (pip->i_d.di_flags & XFS_DIFLAG_EXTSZINHERIT) {
					di_flags |= XFS_DIFLAG_EXTSZINHERIT;
					ip->i_d.di_extsize = pip->i_d.di_extsize;
				}
1101
			} else if (S_ISREG(mode)) {
C
Christoph Hellwig 已提交
1102
				if (pip->i_d.di_flags & XFS_DIFLAG_RTINHERIT)
1103
					di_flags |= XFS_DIFLAG_REALTIME;
1104 1105 1106 1107
				if (pip->i_d.di_flags & XFS_DIFLAG_EXTSZINHERIT) {
					di_flags |= XFS_DIFLAG_EXTSIZE;
					ip->i_d.di_extsize = pip->i_d.di_extsize;
				}
L
Linus Torvalds 已提交
1108 1109 1110
			}
			if ((pip->i_d.di_flags & XFS_DIFLAG_NOATIME) &&
			    xfs_inherit_noatime)
1111
				di_flags |= XFS_DIFLAG_NOATIME;
L
Linus Torvalds 已提交
1112 1113
			if ((pip->i_d.di_flags & XFS_DIFLAG_NODUMP) &&
			    xfs_inherit_nodump)
1114
				di_flags |= XFS_DIFLAG_NODUMP;
L
Linus Torvalds 已提交
1115 1116
			if ((pip->i_d.di_flags & XFS_DIFLAG_SYNC) &&
			    xfs_inherit_sync)
1117
				di_flags |= XFS_DIFLAG_SYNC;
L
Linus Torvalds 已提交
1118 1119
			if ((pip->i_d.di_flags & XFS_DIFLAG_NOSYMLINKS) &&
			    xfs_inherit_nosymlinks)
1120 1121 1122
				di_flags |= XFS_DIFLAG_NOSYMLINKS;
			if (pip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
				di_flags |= XFS_DIFLAG_PROJINHERIT;
1123 1124 1125
			if ((pip->i_d.di_flags & XFS_DIFLAG_NODEFRAG) &&
			    xfs_inherit_nodefrag)
				di_flags |= XFS_DIFLAG_NODEFRAG;
1126 1127
			if (pip->i_d.di_flags & XFS_DIFLAG_FILESTREAM)
				di_flags |= XFS_DIFLAG_FILESTREAM;
1128
			ip->i_d.di_flags |= di_flags;
L
Linus Torvalds 已提交
1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
		}
		/* FALLTHROUGH */
	case S_IFLNK:
		ip->i_d.di_format = XFS_DINODE_FMT_EXTENTS;
		ip->i_df.if_flags = XFS_IFEXTENTS;
		ip->i_df.if_bytes = ip->i_df.if_real_bytes = 0;
		ip->i_df.if_u1.if_extents = NULL;
		break;
	default:
		ASSERT(0);
	}
	/*
	 * Attribute fork settings for new inode.
	 */
	ip->i_d.di_aformat = XFS_DINODE_FMT_EXTENTS;
	ip->i_d.di_anextents = 0;

	/*
	 * Log the new values stuffed into the inode.
	 */
1149
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
L
Linus Torvalds 已提交
1150 1151
	xfs_trans_log_inode(tp, ip, flags);

1152
	/* now that we have an i_mode we can setup inode ops and unlock */
1153
	xfs_setup_inode(ip);
L
Linus Torvalds 已提交
1154

1155 1156 1157 1158 1159 1160 1161 1162 1163
	/* now we have set up the vfs inode we can associate the filestream */
	if (filestreams) {
		error = xfs_filestream_associate(pip, ip);
		if (error < 0)
			return -error;
		if (!error)
			xfs_iflags_set(ip, XFS_IFILESTREAM);
	}

L
Linus Torvalds 已提交
1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174
	*ipp = ip;
	return 0;
}

/*
 * Check to make sure that there are no blocks allocated to the
 * file beyond the size of the file.  We don't check this for
 * files with fixed size extents or real time extents, but we
 * at least do it for regular files.
 */
#ifdef DEBUG
1175
STATIC void
L
Linus Torvalds 已提交
1176
xfs_isize_check(
1177 1178
	struct xfs_inode	*ip,
	xfs_fsize_t		isize)
L
Linus Torvalds 已提交
1179
{
1180 1181 1182 1183
	struct xfs_mount	*mp = ip->i_mount;
	xfs_fileoff_t		map_first;
	int			nimaps;
	xfs_bmbt_irec_t		imaps[2];
D
Dave Chinner 已提交
1184
	int			error;
L
Linus Torvalds 已提交
1185

1186
	if (!S_ISREG(ip->i_d.di_mode))
L
Linus Torvalds 已提交
1187 1188
		return;

1189 1190 1191 1192
	if (XFS_IS_REALTIME_INODE(ip))
		return;

	if (ip->i_d.di_flags & XFS_DIFLAG_EXTSIZE)
L
Linus Torvalds 已提交
1193 1194 1195 1196 1197 1198 1199 1200
		return;

	nimaps = 2;
	map_first = XFS_B_TO_FSB(mp, (xfs_ufsize_t)isize);
	/*
	 * The filesystem could be shutting down, so bmapi may return
	 * an error.
	 */
D
Dave Chinner 已提交
1201
	error = xfs_bmapi_read(ip, map_first,
L
Linus Torvalds 已提交
1202
			 (XFS_B_TO_FSB(mp,
D
Dave Chinner 已提交
1203 1204 1205 1206
			       (xfs_ufsize_t)XFS_MAXIOFFSET(mp)) - map_first),
			 imaps, &nimaps, XFS_BMAPI_ENTIRE);
	if (error)
		return;
L
Linus Torvalds 已提交
1207 1208 1209
	ASSERT(nimaps == 1);
	ASSERT(imaps[0].br_startblock == HOLESTARTBLOCK);
}
1210 1211
#else	/* DEBUG */
#define xfs_isize_check(ip, isize)
L
Linus Torvalds 已提交
1212 1213 1214
#endif	/* DEBUG */

/*
1215 1216 1217
 * Free up the underlying blocks past new_size.  The new size must be smaller
 * than the current size.  This routine can be used both for the attribute and
 * data fork, and does not modify the inode size, which is left to the caller.
L
Linus Torvalds 已提交
1218
 *
1219 1220 1221 1222 1223 1224 1225 1226 1227
 * The transaction passed to this routine must have made a permanent log
 * reservation of at least XFS_ITRUNCATE_LOG_RES.  This routine may commit the
 * given transaction and start new ones, so make sure everything involved in
 * the transaction is tidy before calling here.  Some transaction will be
 * returned to the caller to be committed.  The incoming transaction must
 * already include the inode, and both inode locks must be held exclusively.
 * The inode must also be "held" within the transaction.  On return the inode
 * will be "held" within the returned transaction.  This routine does NOT
 * require any disk space to be reserved for it within the transaction.
L
Linus Torvalds 已提交
1228
 *
1229 1230 1231 1232 1233
 * If we get an error, we must return with the inode locked and linked into the
 * current transaction. This keeps things simple for the higher level code,
 * because it always knows that the inode is locked and held in the transaction
 * that returns to it whether errors occur or not.  We don't mark the inode
 * dirty on error so that transactions can be easily aborted if possible.
L
Linus Torvalds 已提交
1234 1235
 */
int
1236 1237 1238 1239 1240
xfs_itruncate_extents(
	struct xfs_trans	**tpp,
	struct xfs_inode	*ip,
	int			whichfork,
	xfs_fsize_t		new_size)
L
Linus Torvalds 已提交
1241
{
1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
	struct xfs_mount	*mp = ip->i_mount;
	struct xfs_trans	*tp = *tpp;
	struct xfs_trans	*ntp;
	xfs_bmap_free_t		free_list;
	xfs_fsblock_t		first_block;
	xfs_fileoff_t		first_unmap_block;
	xfs_fileoff_t		last_block;
	xfs_filblks_t		unmap_len;
	int			committed;
	int			error = 0;
	int			done = 0;
L
Linus Torvalds 已提交
1253

C
Christoph Hellwig 已提交
1254
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_IOLOCK_EXCL));
1255 1256
	ASSERT(new_size <= ip->i_size);
	ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
L
Linus Torvalds 已提交
1257
	ASSERT(ip->i_itemp != NULL);
1258
	ASSERT(ip->i_itemp->ili_lock_flags == 0);
1259
	ASSERT(!XFS_NOT_DQATTACHED(mp, ip));
L
Linus Torvalds 已提交
1260 1261 1262 1263 1264 1265 1266 1267 1268 1269

	/*
	 * Since it is possible for space to become allocated beyond
	 * the end of the file (in a crash where the space is allocated
	 * but the inode size is not yet updated), simply remove any
	 * blocks which show up between the new EOF and the maximum
	 * possible file size.  If the first block to be removed is
	 * beyond the maximum file size (ie it is the same as last_block),
	 * then there is nothing to do.
	 */
1270
	first_unmap_block = XFS_B_TO_FSB(mp, (xfs_ufsize_t)new_size);
L
Linus Torvalds 已提交
1271
	last_block = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_MAXIOFFSET(mp));
1272 1273 1274 1275 1276
	if (first_unmap_block == last_block)
		return 0;

	ASSERT(first_unmap_block < last_block);
	unmap_len = last_block - first_unmap_block + 1;
L
Linus Torvalds 已提交
1277
	while (!done) {
1278
		xfs_bmap_init(&free_list, &first_block);
1279
		error = xfs_bunmapi(tp, ip,
1280
				    first_unmap_block, unmap_len,
1281
				    xfs_bmapi_aflag(whichfork),
L
Linus Torvalds 已提交
1282
				    XFS_ITRUNC_MAX_EXTENTS,
1283
				    &first_block, &free_list,
1284
				    &done);
1285 1286
		if (error)
			goto out_bmap_cancel;
L
Linus Torvalds 已提交
1287 1288 1289 1290 1291

		/*
		 * Duplicate the transaction that has the permanent
		 * reservation and commit the old transaction.
		 */
1292
		error = xfs_bmap_finish(&tp, &free_list, &committed);
1293
		if (committed)
1294
			xfs_trans_ijoin(tp, ip, 0);
1295 1296
		if (error)
			goto out_bmap_cancel;
L
Linus Torvalds 已提交
1297 1298 1299

		if (committed) {
			/*
1300
			 * Mark the inode dirty so it will be logged and
1301
			 * moved forward in the log as part of every commit.
L
Linus Torvalds 已提交
1302
			 */
1303
			xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
L
Linus Torvalds 已提交
1304
		}
1305

1306 1307 1308
		ntp = xfs_trans_dup(tp);
		error = xfs_trans_commit(tp, 0);
		tp = ntp;
1309

1310
		xfs_trans_ijoin(tp, ip, 0);
1311

1312
		if (error)
1313 1314
			goto out;

1315
		/*
1316
		 * Transaction commit worked ok so we can drop the extra ticket
1317 1318
		 * reference that we gained in xfs_trans_dup()
		 */
1319 1320
		xfs_log_ticket_put(tp->t_ticket);
		error = xfs_trans_reserve(tp, 0,
1321 1322 1323 1324
					XFS_ITRUNCATE_LOG_RES(mp), 0,
					XFS_TRANS_PERM_LOG_RES,
					XFS_ITRUNCATE_LOG_COUNT);
		if (error)
1325
			goto out;
L
Linus Torvalds 已提交
1326
	}
1327 1328 1329 1330 1331

out:
	*tpp = tp;
	return error;
out_bmap_cancel:
L
Linus Torvalds 已提交
1332
	/*
1333 1334 1335
	 * If the bunmapi call encounters an error, return to the caller where
	 * the transaction can be properly aborted.  We just need to make sure
	 * we're not holding any resources that we were not when we came in.
L
Linus Torvalds 已提交
1336
	 */
1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363
	xfs_bmap_cancel(&free_list);
	goto out;
}

int
xfs_itruncate_data(
	struct xfs_trans	**tpp,
	struct xfs_inode	*ip,
	xfs_fsize_t		new_size)
{
	int			error;

	trace_xfs_itruncate_data_start(ip, new_size);

	/*
	 * The first thing we do is set the size to new_size permanently on
	 * disk.  This way we don't have to worry about anyone ever being able
	 * to look at the data being freed even in the face of a crash.
	 * What we're getting around here is the case where we free a block, it
	 * is allocated to another file, it is written to, and then we crash.
	 * If the new data gets written to the file but the log buffers
	 * containing the free and reallocation don't, then we'd end up with
	 * garbage in the blocks being freed.  As long as we make the new_size
	 * permanent before actually freeing any blocks it doesn't matter if
	 * they get written to.
	 */
	if (ip->i_d.di_nextents > 0) {
1364
		/*
1365 1366 1367 1368 1369 1370
		 * If we are not changing the file size then do not update
		 * the on-disk file size - we may be called from
		 * xfs_inactive_free_eofblocks().  If we update the on-disk
		 * file size and then the system crashes before the contents
		 * of the file are flushed to disk then the files may be
		 * full of holes (ie NULL files bug).
1371 1372 1373 1374
		 */
		if (ip->i_size != new_size) {
			ip->i_d.di_size = new_size;
			ip->i_size = new_size;
1375
			xfs_trans_log_inode(*tpp, ip, XFS_ILOG_CORE);
1376
		}
L
Linus Torvalds 已提交
1377
	}
1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405

	error = xfs_itruncate_extents(tpp, ip, XFS_DATA_FORK, new_size);
	if (error)
		return error;

	/*
	 * If we are not changing the file size then do not update the on-disk
	 * file size - we may be called from xfs_inactive_free_eofblocks().
	 * If we update the on-disk file size and then the system crashes
	 * before the contents of the file are flushed to disk then the files
	 * may be full of holes (ie NULL files bug).
	 */
	xfs_isize_check(ip, new_size);
	if (ip->i_size != new_size) {
		ip->i_d.di_size = new_size;
		ip->i_size = new_size;
	}

	ASSERT(new_size != 0 || ip->i_delayed_blks == 0);
	ASSERT(new_size != 0 || ip->i_d.di_nextents == 0);

	/*
	 * Always re-log the inode so that our permanent transaction can keep
	 * on rolling it forward in the log.
	 */
	xfs_trans_log_inode(*tpp, ip, XFS_ILOG_CORE);

	trace_xfs_itruncate_data_end(ip, new_size);
L
Linus Torvalds 已提交
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 1436 1437
	return 0;
}

/*
 * This is called when the inode's link count goes to 0.
 * We place the on-disk inode on a list in the AGI.  It
 * will be pulled from this list when the inode is freed.
 */
int
xfs_iunlink(
	xfs_trans_t	*tp,
	xfs_inode_t	*ip)
{
	xfs_mount_t	*mp;
	xfs_agi_t	*agi;
	xfs_dinode_t	*dip;
	xfs_buf_t	*agibp;
	xfs_buf_t	*ibp;
	xfs_agino_t	agino;
	short		bucket_index;
	int		offset;
	int		error;

	ASSERT(ip->i_d.di_nlink == 0);
	ASSERT(ip->i_d.di_mode != 0);

	mp = tp->t_mountp;

	/*
	 * Get the agi buffer first.  It ensures lock ordering
	 * on the list.
	 */
1438
	error = xfs_read_agi(mp, tp, XFS_INO_TO_AGNO(mp, ip->i_ino), &agibp);
1439
	if (error)
L
Linus Torvalds 已提交
1440 1441
		return error;
	agi = XFS_BUF_TO_AGI(agibp);
1442

L
Linus Torvalds 已提交
1443 1444 1445 1446 1447 1448 1449 1450
	/*
	 * Get the index into the agi hash table for the
	 * list this inode will go on.
	 */
	agino = XFS_INO_TO_AGINO(mp, ip->i_ino);
	ASSERT(agino != 0);
	bucket_index = agino % XFS_AGI_UNLINKED_BUCKETS;
	ASSERT(agi->agi_unlinked[bucket_index]);
1451
	ASSERT(be32_to_cpu(agi->agi_unlinked[bucket_index]) != agino);
L
Linus Torvalds 已提交
1452

1453
	if (agi->agi_unlinked[bucket_index] != cpu_to_be32(NULLAGINO)) {
L
Linus Torvalds 已提交
1454 1455 1456 1457 1458 1459
		/*
		 * There is already another inode in the bucket we need
		 * to add ourselves to.  Add us at the front of the list.
		 * Here we put the head pointer into our next pointer,
		 * and then we fall through to point the head at us.
		 */
1460
		error = xfs_itobp(mp, tp, ip, &dip, &ibp, XBF_LOCK);
1461 1462 1463
		if (error)
			return error;

1464
		ASSERT(dip->di_next_unlinked == cpu_to_be32(NULLAGINO));
L
Linus Torvalds 已提交
1465
		dip->di_next_unlinked = agi->agi_unlinked[bucket_index];
1466
		offset = ip->i_imap.im_boffset +
L
Linus Torvalds 已提交
1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477
			offsetof(xfs_dinode_t, di_next_unlinked);
		xfs_trans_inode_buf(tp, ibp);
		xfs_trans_log_buf(tp, ibp, offset,
				  (offset + sizeof(xfs_agino_t) - 1));
		xfs_inobp_check(mp, ibp);
	}

	/*
	 * Point the bucket head pointer at the inode being inserted.
	 */
	ASSERT(agino != 0);
1478
	agi->agi_unlinked[bucket_index] = cpu_to_be32(agino);
L
Linus Torvalds 已提交
1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503
	offset = offsetof(xfs_agi_t, agi_unlinked) +
		(sizeof(xfs_agino_t) * bucket_index);
	xfs_trans_log_buf(tp, agibp, offset,
			  (offset + sizeof(xfs_agino_t) - 1));
	return 0;
}

/*
 * Pull the on-disk inode from the AGI unlinked list.
 */
STATIC int
xfs_iunlink_remove(
	xfs_trans_t	*tp,
	xfs_inode_t	*ip)
{
	xfs_ino_t	next_ino;
	xfs_mount_t	*mp;
	xfs_agi_t	*agi;
	xfs_dinode_t	*dip;
	xfs_buf_t	*agibp;
	xfs_buf_t	*ibp;
	xfs_agnumber_t	agno;
	xfs_agino_t	agino;
	xfs_agino_t	next_agino;
	xfs_buf_t	*last_ibp;
1504
	xfs_dinode_t	*last_dip = NULL;
L
Linus Torvalds 已提交
1505
	short		bucket_index;
1506
	int		offset, last_offset = 0;
L
Linus Torvalds 已提交
1507 1508 1509 1510 1511 1512 1513 1514 1515
	int		error;

	mp = tp->t_mountp;
	agno = XFS_INO_TO_AGNO(mp, ip->i_ino);

	/*
	 * Get the agi buffer first.  It ensures lock ordering
	 * on the list.
	 */
1516 1517
	error = xfs_read_agi(mp, tp, agno, &agibp);
	if (error)
L
Linus Torvalds 已提交
1518
		return error;
1519

L
Linus Torvalds 已提交
1520
	agi = XFS_BUF_TO_AGI(agibp);
1521

L
Linus Torvalds 已提交
1522 1523 1524 1525 1526 1527 1528
	/*
	 * Get the index into the agi hash table for the
	 * list this inode will go on.
	 */
	agino = XFS_INO_TO_AGINO(mp, ip->i_ino);
	ASSERT(agino != 0);
	bucket_index = agino % XFS_AGI_UNLINKED_BUCKETS;
1529
	ASSERT(agi->agi_unlinked[bucket_index] != cpu_to_be32(NULLAGINO));
L
Linus Torvalds 已提交
1530 1531
	ASSERT(agi->agi_unlinked[bucket_index]);

1532
	if (be32_to_cpu(agi->agi_unlinked[bucket_index]) == agino) {
L
Linus Torvalds 已提交
1533 1534 1535 1536 1537 1538 1539 1540
		/*
		 * We're at the head of the list.  Get the inode's
		 * on-disk buffer to see if there is anyone after us
		 * on the list.  Only modify our next pointer if it
		 * is not already NULLAGINO.  This saves us the overhead
		 * of dealing with the buffer when there is no need to
		 * change it.
		 */
1541
		error = xfs_itobp(mp, tp, ip, &dip, &ibp, XBF_LOCK);
L
Linus Torvalds 已提交
1542
		if (error) {
1543 1544
			xfs_warn(mp, "%s: xfs_itobp() returned error %d.",
				__func__, error);
L
Linus Torvalds 已提交
1545 1546
			return error;
		}
1547
		next_agino = be32_to_cpu(dip->di_next_unlinked);
L
Linus Torvalds 已提交
1548 1549
		ASSERT(next_agino != 0);
		if (next_agino != NULLAGINO) {
1550
			dip->di_next_unlinked = cpu_to_be32(NULLAGINO);
1551
			offset = ip->i_imap.im_boffset +
L
Linus Torvalds 已提交
1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
				offsetof(xfs_dinode_t, di_next_unlinked);
			xfs_trans_inode_buf(tp, ibp);
			xfs_trans_log_buf(tp, ibp, offset,
					  (offset + sizeof(xfs_agino_t) - 1));
			xfs_inobp_check(mp, ibp);
		} else {
			xfs_trans_brelse(tp, ibp);
		}
		/*
		 * Point the bucket head pointer at the next inode.
		 */
		ASSERT(next_agino != 0);
		ASSERT(next_agino != agino);
1565
		agi->agi_unlinked[bucket_index] = cpu_to_be32(next_agino);
L
Linus Torvalds 已提交
1566 1567 1568 1569 1570 1571 1572 1573
		offset = offsetof(xfs_agi_t, agi_unlinked) +
			(sizeof(xfs_agino_t) * bucket_index);
		xfs_trans_log_buf(tp, agibp, offset,
				  (offset + sizeof(xfs_agino_t) - 1));
	} else {
		/*
		 * We need to search the list for the inode being freed.
		 */
1574
		next_agino = be32_to_cpu(agi->agi_unlinked[bucket_index]);
L
Linus Torvalds 已提交
1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586
		last_ibp = NULL;
		while (next_agino != agino) {
			/*
			 * If the last inode wasn't the one pointing to
			 * us, then release its buffer since we're not
			 * going to do anything with it.
			 */
			if (last_ibp != NULL) {
				xfs_trans_brelse(tp, last_ibp);
			}
			next_ino = XFS_AGINO_TO_INO(mp, agno, next_agino);
			error = xfs_inotobp(mp, tp, next_ino, &last_dip,
1587
					    &last_ibp, &last_offset, 0);
L
Linus Torvalds 已提交
1588
			if (error) {
1589 1590 1591
				xfs_warn(mp,
					"%s: xfs_inotobp() returned error %d.",
					__func__, error);
L
Linus Torvalds 已提交
1592 1593
				return error;
			}
1594
			next_agino = be32_to_cpu(last_dip->di_next_unlinked);
L
Linus Torvalds 已提交
1595 1596 1597 1598 1599 1600 1601
			ASSERT(next_agino != NULLAGINO);
			ASSERT(next_agino != 0);
		}
		/*
		 * Now last_ibp points to the buffer previous to us on
		 * the unlinked list.  Pull us from the list.
		 */
1602
		error = xfs_itobp(mp, tp, ip, &dip, &ibp, XBF_LOCK);
L
Linus Torvalds 已提交
1603
		if (error) {
1604 1605
			xfs_warn(mp, "%s: xfs_itobp(2) returned error %d.",
				__func__, error);
L
Linus Torvalds 已提交
1606 1607
			return error;
		}
1608
		next_agino = be32_to_cpu(dip->di_next_unlinked);
L
Linus Torvalds 已提交
1609 1610 1611
		ASSERT(next_agino != 0);
		ASSERT(next_agino != agino);
		if (next_agino != NULLAGINO) {
1612
			dip->di_next_unlinked = cpu_to_be32(NULLAGINO);
1613
			offset = ip->i_imap.im_boffset +
L
Linus Torvalds 已提交
1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
				offsetof(xfs_dinode_t, di_next_unlinked);
			xfs_trans_inode_buf(tp, ibp);
			xfs_trans_log_buf(tp, ibp, offset,
					  (offset + sizeof(xfs_agino_t) - 1));
			xfs_inobp_check(mp, ibp);
		} else {
			xfs_trans_brelse(tp, ibp);
		}
		/*
		 * Point the previous inode on the list to the next inode.
		 */
1625
		last_dip->di_next_unlinked = cpu_to_be32(next_agino);
L
Linus Torvalds 已提交
1626 1627 1628 1629 1630 1631 1632 1633 1634 1635
		ASSERT(next_agino != 0);
		offset = last_offset + offsetof(xfs_dinode_t, di_next_unlinked);
		xfs_trans_inode_buf(tp, last_ibp);
		xfs_trans_log_buf(tp, last_ibp, offset,
				  (offset + sizeof(xfs_agino_t) - 1));
		xfs_inobp_check(mp, last_ibp);
	}
	return 0;
}

1636 1637 1638 1639 1640
/*
 * A big issue when freeing the inode cluster is is that we _cannot_ skip any
 * inodes that are in memory - they all must be marked stale and attached to
 * the cluster buffer.
 */
1641
STATIC int
L
Linus Torvalds 已提交
1642 1643 1644 1645 1646 1647 1648 1649 1650
xfs_ifree_cluster(
	xfs_inode_t	*free_ip,
	xfs_trans_t	*tp,
	xfs_ino_t	inum)
{
	xfs_mount_t		*mp = free_ip->i_mount;
	int			blks_per_cluster;
	int			nbufs;
	int			ninodes;
1651
	int			i, j;
L
Linus Torvalds 已提交
1652 1653
	xfs_daddr_t		blkno;
	xfs_buf_t		*bp;
1654
	xfs_inode_t		*ip;
L
Linus Torvalds 已提交
1655 1656
	xfs_inode_log_item_t	*iip;
	xfs_log_item_t		*lip;
D
Dave Chinner 已提交
1657
	struct xfs_perag	*pag;
L
Linus Torvalds 已提交
1658

D
Dave Chinner 已提交
1659
	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, inum));
L
Linus Torvalds 已提交
1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674
	if (mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp)) {
		blks_per_cluster = 1;
		ninodes = mp->m_sb.sb_inopblock;
		nbufs = XFS_IALLOC_BLOCKS(mp);
	} else {
		blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) /
					mp->m_sb.sb_blocksize;
		ninodes = blks_per_cluster * mp->m_sb.sb_inopblock;
		nbufs = XFS_IALLOC_BLOCKS(mp) / blks_per_cluster;
	}

	for (j = 0; j < nbufs; j++, inum += ninodes) {
		blkno = XFS_AGB_TO_DADDR(mp, XFS_INO_TO_AGNO(mp, inum),
					 XFS_INO_TO_AGBNO(mp, inum));

1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
		/*
		 * We obtain and lock the backing buffer first in the process
		 * here, as we have to ensure that any dirty inode that we
		 * can't get the flush lock on is attached to the buffer.
		 * If we scan the in-memory inodes first, then buffer IO can
		 * complete before we get a lock on it, and hence we may fail
		 * to mark all the active inodes on the buffer stale.
		 */
		bp = xfs_trans_get_buf(tp, mp->m_ddev_targp, blkno,
					mp->m_bsize * blks_per_cluster,
					XBF_LOCK);

1687 1688
		if (!bp)
			return ENOMEM;
1689 1690 1691
		/*
		 * Walk the inodes already attached to the buffer and mark them
		 * stale. These will all have the flush locks held, so an
1692 1693 1694
		 * in-memory inode walk can't lock them. By marking them all
		 * stale first, we will not attempt to lock them in the loop
		 * below as the XFS_ISTALE flag will be set.
1695
		 */
1696
		lip = bp->b_fspriv;
1697 1698 1699 1700
		while (lip) {
			if (lip->li_type == XFS_LI_INODE) {
				iip = (xfs_inode_log_item_t *)lip;
				ASSERT(iip->ili_logged == 1);
1701
				lip->li_cb = xfs_istale_done;
1702 1703 1704 1705 1706 1707 1708
				xfs_trans_ail_copy_lsn(mp->m_ail,
							&iip->ili_flush_lsn,
							&iip->ili_item.li_lsn);
				xfs_iflags_set(iip->ili_inode, XFS_ISTALE);
			}
			lip = lip->li_bio_list;
		}
L
Linus Torvalds 已提交
1709

1710

L
Linus Torvalds 已提交
1711
		/*
1712 1713 1714 1715
		 * For each inode in memory attempt to add it to the inode
		 * buffer and set it up for being staled on buffer IO
		 * completion.  This is safe as we've locked out tail pushing
		 * and flushing by locking the buffer.
L
Linus Torvalds 已提交
1716
		 *
1717 1718 1719
		 * We have already marked every inode that was part of a
		 * transaction stale above, which means there is no point in
		 * even trying to lock them.
L
Linus Torvalds 已提交
1720 1721
		 */
		for (i = 0; i < ninodes; i++) {
1722
retry:
1723
			rcu_read_lock();
1724 1725
			ip = radix_tree_lookup(&pag->pag_ici_root,
					XFS_INO_TO_AGINO(mp, (inum + i)));
L
Linus Torvalds 已提交
1726

1727 1728 1729
			/* Inode not in memory, nothing to do */
			if (!ip) {
				rcu_read_unlock();
L
Linus Torvalds 已提交
1730 1731 1732
				continue;
			}

1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748
			/*
			 * because this is an RCU protected lookup, we could
			 * find a recently freed or even reallocated inode
			 * during the lookup. We need to check under the
			 * i_flags_lock for a valid inode here. Skip it if it
			 * is not valid, the wrong inode or stale.
			 */
			spin_lock(&ip->i_flags_lock);
			if (ip->i_ino != inum + i ||
			    __xfs_iflags_test(ip, XFS_ISTALE)) {
				spin_unlock(&ip->i_flags_lock);
				rcu_read_unlock();
				continue;
			}
			spin_unlock(&ip->i_flags_lock);

1749 1750 1751 1752 1753 1754 1755
			/*
			 * Don't try to lock/unlock the current inode, but we
			 * _cannot_ skip the other inodes that we did not find
			 * in the list attached to the buffer and are not
			 * already marked stale. If we can't lock it, back off
			 * and retry.
			 */
1756 1757
			if (ip != free_ip &&
			    !xfs_ilock_nowait(ip, XFS_ILOCK_EXCL)) {
1758
				rcu_read_unlock();
1759 1760
				delay(1);
				goto retry;
L
Linus Torvalds 已提交
1761
			}
1762
			rcu_read_unlock();
L
Linus Torvalds 已提交
1763

1764
			xfs_iflock(ip);
1765
			xfs_iflags_set(ip, XFS_ISTALE);
L
Linus Torvalds 已提交
1766

1767 1768 1769 1770
			/*
			 * we don't need to attach clean inodes or those only
			 * with unlogged changes (which we throw away, anyway).
			 */
L
Linus Torvalds 已提交
1771
			iip = ip->i_itemp;
1772
			if (!iip || xfs_inode_clean(ip)) {
1773
				ASSERT(ip != free_ip);
L
Linus Torvalds 已提交
1774 1775 1776 1777 1778 1779 1780 1781 1782
				ip->i_update_core = 0;
				xfs_ifunlock(ip);
				xfs_iunlock(ip, XFS_ILOCK_EXCL);
				continue;
			}

			iip->ili_last_fields = iip->ili_format.ilf_fields;
			iip->ili_format.ilf_fields = 0;
			iip->ili_logged = 1;
1783 1784
			xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn,
						&iip->ili_item.li_lsn);
L
Linus Torvalds 已提交
1785

1786 1787
			xfs_buf_attach_iodone(bp, xfs_istale_done,
						  &iip->ili_item);
1788 1789

			if (ip != free_ip)
L
Linus Torvalds 已提交
1790 1791 1792
				xfs_iunlock(ip, XFS_ILOCK_EXCL);
		}

1793
		xfs_trans_stale_inode_buf(tp, bp);
L
Linus Torvalds 已提交
1794 1795 1796
		xfs_trans_binval(tp, bp);
	}

D
Dave Chinner 已提交
1797
	xfs_perag_put(pag);
1798
	return 0;
L
Linus Torvalds 已提交
1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819
}

/*
 * This is called to return an inode to the inode free list.
 * The inode should already be truncated to 0 length and have
 * no pages associated with it.  This routine also assumes that
 * the inode is already a part of the transaction.
 *
 * The on-disk copy of the inode will have been added to the list
 * of unlinked inodes in the AGI. We need to remove the inode from
 * that list atomically with respect to freeing it here.
 */
int
xfs_ifree(
	xfs_trans_t	*tp,
	xfs_inode_t	*ip,
	xfs_bmap_free_t	*flist)
{
	int			error;
	int			delete;
	xfs_ino_t		first_ino;
1820 1821
	xfs_dinode_t    	*dip;
	xfs_buf_t       	*ibp;
L
Linus Torvalds 已提交
1822

C
Christoph Hellwig 已提交
1823
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
L
Linus Torvalds 已提交
1824 1825 1826
	ASSERT(ip->i_d.di_nlink == 0);
	ASSERT(ip->i_d.di_nextents == 0);
	ASSERT(ip->i_d.di_anextents == 0);
1827
	ASSERT((ip->i_d.di_size == 0 && ip->i_size == 0) ||
1828
	       (!S_ISREG(ip->i_d.di_mode)));
L
Linus Torvalds 已提交
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
	ASSERT(ip->i_d.di_nblocks == 0);

	/*
	 * Pull the on-disk inode from the AGI unlinked list.
	 */
	error = xfs_iunlink_remove(tp, ip);
	if (error != 0) {
		return error;
	}

	error = xfs_difree(tp, ip->i_ino, flist, &delete, &first_ino);
	if (error != 0) {
		return error;
	}
	ip->i_d.di_mode = 0;		/* mark incore inode as free */
	ip->i_d.di_flags = 0;
	ip->i_d.di_dmevmask = 0;
	ip->i_d.di_forkoff = 0;		/* mark the attr fork not in use */
	ip->i_df.if_ext_max =
		XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t);
	ip->i_d.di_format = XFS_DINODE_FMT_EXTENTS;
	ip->i_d.di_aformat = XFS_DINODE_FMT_EXTENTS;
	/*
	 * Bump the generation count so no one will be confused
	 * by reincarnations of this inode.
	 */
	ip->i_d.di_gen++;
1856

L
Linus Torvalds 已提交
1857 1858
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

1859
	error = xfs_itobp(ip->i_mount, tp, ip, &dip, &ibp, XBF_LOCK);
1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874
	if (error)
		return error;

        /*
	* Clear the on-disk di_mode. This is to prevent xfs_bulkstat
	* from picking up this inode when it is reclaimed (its incore state
	* initialzed but not flushed to disk yet). The in-core di_mode is
	* already cleared  and a corresponding transaction logged.
	* The hack here just synchronizes the in-core to on-disk
	* di_mode value in advance before the actual inode sync to disk.
	* This is OK because the inode is already unlinked and would never
	* change its di_mode again for this inode generation.
	* This is a temporary hack that would require a proper fix
	* in the future.
	*/
C
Christoph Hellwig 已提交
1875
	dip->di_mode = 0;
1876

L
Linus Torvalds 已提交
1877
	if (delete) {
1878
		error = xfs_ifree_cluster(ip, tp, first_ino);
L
Linus Torvalds 已提交
1879 1880
	}

1881
	return error;
L
Linus Torvalds 已提交
1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907
}

/*
 * Reallocate the space for if_broot based on the number of records
 * being added or deleted as indicated in rec_diff.  Move the records
 * and pointers in if_broot to fit the new size.  When shrinking this
 * will eliminate holes between the records and pointers created by
 * the caller.  When growing this will create holes to be filled in
 * by the caller.
 *
 * The caller must not request to add more records than would fit in
 * the on-disk inode root.  If the if_broot is currently NULL, then
 * if we adding records one will be allocated.  The caller must also
 * not request that the number of records go below zero, although
 * it can go to zero.
 *
 * ip -- the inode whose if_broot area is changing
 * ext_diff -- the change in the number of records, positive or negative,
 *	 requested for the if_broot array.
 */
void
xfs_iroot_realloc(
	xfs_inode_t		*ip,
	int			rec_diff,
	int			whichfork)
{
1908
	struct xfs_mount	*mp = ip->i_mount;
L
Linus Torvalds 已提交
1909 1910
	int			cur_max;
	xfs_ifork_t		*ifp;
1911
	struct xfs_btree_block	*new_broot;
L
Linus Torvalds 已提交
1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
	int			new_max;
	size_t			new_size;
	char			*np;
	char			*op;

	/*
	 * Handle the degenerate case quietly.
	 */
	if (rec_diff == 0) {
		return;
	}

	ifp = XFS_IFORK_PTR(ip, whichfork);
	if (rec_diff > 0) {
		/*
		 * If there wasn't any memory allocated before, just
		 * allocate it now and get out.
		 */
		if (ifp->if_broot_bytes == 0) {
			new_size = (size_t)XFS_BMAP_BROOT_SPACE_CALC(rec_diff);
1932
			ifp->if_broot = kmem_alloc(new_size, KM_SLEEP | KM_NOFS);
L
Linus Torvalds 已提交
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
			ifp->if_broot_bytes = (int)new_size;
			return;
		}

		/*
		 * If there is already an existing if_broot, then we need
		 * to realloc() it and shift the pointers to their new
		 * location.  The records don't change location because
		 * they are kept butted up against the btree block header.
		 */
1943
		cur_max = xfs_bmbt_maxrecs(mp, ifp->if_broot_bytes, 0);
L
Linus Torvalds 已提交
1944 1945
		new_max = cur_max + rec_diff;
		new_size = (size_t)XFS_BMAP_BROOT_SPACE_CALC(new_max);
1946
		ifp->if_broot = kmem_realloc(ifp->if_broot, new_size,
L
Linus Torvalds 已提交
1947
				(size_t)XFS_BMAP_BROOT_SPACE_CALC(cur_max), /* old size */
1948
				KM_SLEEP | KM_NOFS);
1949 1950 1951 1952
		op = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, ifp->if_broot, 1,
						     ifp->if_broot_bytes);
		np = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, ifp->if_broot, 1,
						     (int)new_size);
L
Linus Torvalds 已提交
1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
		ifp->if_broot_bytes = (int)new_size;
		ASSERT(ifp->if_broot_bytes <=
			XFS_IFORK_SIZE(ip, whichfork) + XFS_BROOT_SIZE_ADJ);
		memmove(np, op, cur_max * (uint)sizeof(xfs_dfsbno_t));
		return;
	}

	/*
	 * rec_diff is less than 0.  In this case, we are shrinking the
	 * if_broot buffer.  It must already exist.  If we go to zero
	 * records, just get rid of the root and clear the status bit.
	 */
	ASSERT((ifp->if_broot != NULL) && (ifp->if_broot_bytes > 0));
1966
	cur_max = xfs_bmbt_maxrecs(mp, ifp->if_broot_bytes, 0);
L
Linus Torvalds 已提交
1967 1968 1969 1970 1971 1972 1973
	new_max = cur_max + rec_diff;
	ASSERT(new_max >= 0);
	if (new_max > 0)
		new_size = (size_t)XFS_BMAP_BROOT_SPACE_CALC(new_max);
	else
		new_size = 0;
	if (new_size > 0) {
1974
		new_broot = kmem_alloc(new_size, KM_SLEEP | KM_NOFS);
L
Linus Torvalds 已提交
1975 1976 1977
		/*
		 * First copy over the btree block header.
		 */
1978
		memcpy(new_broot, ifp->if_broot, XFS_BTREE_LBLOCK_LEN);
L
Linus Torvalds 已提交
1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
	} else {
		new_broot = NULL;
		ifp->if_flags &= ~XFS_IFBROOT;
	}

	/*
	 * Only copy the records and pointers if there are any.
	 */
	if (new_max > 0) {
		/*
		 * First copy the records.
		 */
1991 1992
		op = (char *)XFS_BMBT_REC_ADDR(mp, ifp->if_broot, 1);
		np = (char *)XFS_BMBT_REC_ADDR(mp, new_broot, 1);
L
Linus Torvalds 已提交
1993 1994 1995 1996 1997
		memcpy(np, op, new_max * (uint)sizeof(xfs_bmbt_rec_t));

		/*
		 * Then copy the pointers.
		 */
1998
		op = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, ifp->if_broot, 1,
L
Linus Torvalds 已提交
1999
						     ifp->if_broot_bytes);
2000
		np = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, new_broot, 1,
L
Linus Torvalds 已提交
2001 2002 2003
						     (int)new_size);
		memcpy(np, op, new_max * (uint)sizeof(xfs_dfsbno_t));
	}
2004
	kmem_free(ifp->if_broot);
L
Linus Torvalds 已提交
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047
	ifp->if_broot = new_broot;
	ifp->if_broot_bytes = (int)new_size;
	ASSERT(ifp->if_broot_bytes <=
		XFS_IFORK_SIZE(ip, whichfork) + XFS_BROOT_SIZE_ADJ);
	return;
}


/*
 * This is called when the amount of space needed for if_data
 * is increased or decreased.  The change in size is indicated by
 * the number of bytes that need to be added or deleted in the
 * byte_diff parameter.
 *
 * If the amount of space needed has decreased below the size of the
 * inline buffer, then switch to using the inline buffer.  Otherwise,
 * use kmem_realloc() or kmem_alloc() to adjust the size of the buffer
 * to what is needed.
 *
 * ip -- the inode whose if_data area is changing
 * byte_diff -- the change in the number of bytes, positive or negative,
 *	 requested for the if_data array.
 */
void
xfs_idata_realloc(
	xfs_inode_t	*ip,
	int		byte_diff,
	int		whichfork)
{
	xfs_ifork_t	*ifp;
	int		new_size;
	int		real_size;

	if (byte_diff == 0) {
		return;
	}

	ifp = XFS_IFORK_PTR(ip, whichfork);
	new_size = (int)ifp->if_bytes + byte_diff;
	ASSERT(new_size >= 0);

	if (new_size == 0) {
		if (ifp->if_u1.if_data != ifp->if_u2.if_inline_data) {
2048
			kmem_free(ifp->if_u1.if_data);
L
Linus Torvalds 已提交
2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062
		}
		ifp->if_u1.if_data = NULL;
		real_size = 0;
	} else if (new_size <= sizeof(ifp->if_u2.if_inline_data)) {
		/*
		 * If the valid extents/data can fit in if_inline_ext/data,
		 * copy them from the malloc'd vector and free it.
		 */
		if (ifp->if_u1.if_data == NULL) {
			ifp->if_u1.if_data = ifp->if_u2.if_inline_data;
		} else if (ifp->if_u1.if_data != ifp->if_u2.if_inline_data) {
			ASSERT(ifp->if_real_bytes != 0);
			memcpy(ifp->if_u2.if_inline_data, ifp->if_u1.if_data,
			      new_size);
2063
			kmem_free(ifp->if_u1.if_data);
L
Linus Torvalds 已提交
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077
			ifp->if_u1.if_data = ifp->if_u2.if_inline_data;
		}
		real_size = 0;
	} else {
		/*
		 * Stuck with malloc/realloc.
		 * For inline data, the underlying buffer must be
		 * a multiple of 4 bytes in size so that it can be
		 * logged and stay on word boundaries.  We enforce
		 * that here.
		 */
		real_size = roundup(new_size, 4);
		if (ifp->if_u1.if_data == NULL) {
			ASSERT(ifp->if_real_bytes == 0);
2078 2079
			ifp->if_u1.if_data = kmem_alloc(real_size,
							KM_SLEEP | KM_NOFS);
L
Linus Torvalds 已提交
2080 2081 2082 2083 2084 2085 2086 2087 2088 2089
		} else if (ifp->if_u1.if_data != ifp->if_u2.if_inline_data) {
			/*
			 * Only do the realloc if the underlying size
			 * is really changing.
			 */
			if (ifp->if_real_bytes != real_size) {
				ifp->if_u1.if_data =
					kmem_realloc(ifp->if_u1.if_data,
							real_size,
							ifp->if_real_bytes,
2090
							KM_SLEEP | KM_NOFS);
L
Linus Torvalds 已提交
2091 2092 2093
			}
		} else {
			ASSERT(ifp->if_real_bytes == 0);
2094 2095
			ifp->if_u1.if_data = kmem_alloc(real_size,
							KM_SLEEP | KM_NOFS);
L
Linus Torvalds 已提交
2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113
			memcpy(ifp->if_u1.if_data, ifp->if_u2.if_inline_data,
				ifp->if_bytes);
		}
	}
	ifp->if_real_bytes = real_size;
	ifp->if_bytes = new_size;
	ASSERT(ifp->if_bytes <= XFS_IFORK_SIZE(ip, whichfork));
}

void
xfs_idestroy_fork(
	xfs_inode_t	*ip,
	int		whichfork)
{
	xfs_ifork_t	*ifp;

	ifp = XFS_IFORK_PTR(ip, whichfork);
	if (ifp->if_broot != NULL) {
2114
		kmem_free(ifp->if_broot);
L
Linus Torvalds 已提交
2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127
		ifp->if_broot = NULL;
	}

	/*
	 * If the format is local, then we can't have an extents
	 * array so just look for an inline data array.  If we're
	 * not local then we may or may not have an extents list,
	 * so check and free it up if we do.
	 */
	if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_LOCAL) {
		if ((ifp->if_u1.if_data != ifp->if_u2.if_inline_data) &&
		    (ifp->if_u1.if_data != NULL)) {
			ASSERT(ifp->if_real_bytes != 0);
2128
			kmem_free(ifp->if_u1.if_data);
L
Linus Torvalds 已提交
2129 2130 2131 2132
			ifp->if_u1.if_data = NULL;
			ifp->if_real_bytes = 0;
		}
	} else if ((ifp->if_flags & XFS_IFEXTENTS) &&
2133 2134 2135
		   ((ifp->if_flags & XFS_IFEXTIREC) ||
		    ((ifp->if_u1.if_extents != NULL) &&
		     (ifp->if_u1.if_extents != ifp->if_u2.if_inline_ext)))) {
L
Linus Torvalds 已提交
2136
		ASSERT(ifp->if_real_bytes != 0);
2137
		xfs_iext_destroy(ifp);
L
Linus Torvalds 已提交
2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148
	}
	ASSERT(ifp->if_u1.if_extents == NULL ||
	       ifp->if_u1.if_extents == ifp->if_u2.if_inline_ext);
	ASSERT(ifp->if_real_bytes == 0);
	if (whichfork == XFS_ATTR_FORK) {
		kmem_zone_free(xfs_ifork_zone, ip->i_afp);
		ip->i_afp = NULL;
	}
}

/*
2149 2150 2151
 * This is called to unpin an inode.  The caller must have the inode locked
 * in at least shared mode so that the buffer cannot be subsequently pinned
 * once someone is waiting for it to be unpinned.
L
Linus Torvalds 已提交
2152
 */
2153 2154 2155
static void
xfs_iunpin_nowait(
	struct xfs_inode	*ip)
L
Linus Torvalds 已提交
2156
{
C
Christoph Hellwig 已提交
2157
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
L
Linus Torvalds 已提交
2158

2159 2160
	trace_xfs_inode_unpin_nowait(ip, _RET_IP_);

2161
	/* Give the log a push to start the unpinning I/O */
2162
	xfs_log_force_lsn(ip->i_mount, ip->i_itemp->ili_last_lsn, 0);
2163

2164
}
L
Linus Torvalds 已提交
2165

2166
void
2167
xfs_iunpin_wait(
2168
	struct xfs_inode	*ip)
2169
{
2170 2171 2172 2173
	if (xfs_ipincount(ip)) {
		xfs_iunpin_nowait(ip);
		wait_event(ip->i_ipin_wait, (xfs_ipincount(ip) == 0));
	}
L
Linus Torvalds 已提交
2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
}

/*
 * xfs_iextents_copy()
 *
 * This is called to copy the REAL extents (as opposed to the delayed
 * allocation extents) from the inode into the given buffer.  It
 * returns the number of bytes copied into the buffer.
 *
 * If there are no delayed allocation extents, then we can just
 * memcpy() the extents into the buffer.  Otherwise, we need to
 * examine each extent in turn and skip those which are delayed.
 */
int
xfs_iextents_copy(
	xfs_inode_t		*ip,
2190
	xfs_bmbt_rec_t		*dp,
L
Linus Torvalds 已提交
2191 2192 2193 2194 2195 2196 2197 2198 2199
	int			whichfork)
{
	int			copied;
	int			i;
	xfs_ifork_t		*ifp;
	int			nrecs;
	xfs_fsblock_t		start_block;

	ifp = XFS_IFORK_PTR(ip, whichfork);
C
Christoph Hellwig 已提交
2200
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
L
Linus Torvalds 已提交
2201 2202 2203
	ASSERT(ifp->if_bytes > 0);

	nrecs = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
2204
	XFS_BMAP_TRACE_EXLIST(ip, nrecs, whichfork);
L
Linus Torvalds 已提交
2205 2206 2207 2208 2209 2210 2211 2212 2213 2214
	ASSERT(nrecs > 0);

	/*
	 * There are some delayed allocation extents in the
	 * inode, so copy the extents one at a time and skip
	 * the delayed ones.  There must be at least one
	 * non-delayed extent.
	 */
	copied = 0;
	for (i = 0; i < nrecs; i++) {
2215
		xfs_bmbt_rec_host_t *ep = xfs_iext_get_ext(ifp, i);
L
Linus Torvalds 已提交
2216
		start_block = xfs_bmbt_get_startblock(ep);
2217
		if (isnullstartblock(start_block)) {
L
Linus Torvalds 已提交
2218 2219 2220 2221 2222 2223 2224
			/*
			 * It's a delayed allocation extent, so skip it.
			 */
			continue;
		}

		/* Translate to on disk format */
2225 2226
		put_unaligned(cpu_to_be64(ep->l0), &dp->l0);
		put_unaligned(cpu_to_be64(ep->l1), &dp->l1);
2227
		dp++;
L
Linus Torvalds 已提交
2228 2229 2230
		copied++;
	}
	ASSERT(copied != 0);
2231
	xfs_validate_extents(ifp, copied, XFS_EXTFMT_INODE(ip));
L
Linus Torvalds 已提交
2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246

	return (copied * (uint)sizeof(xfs_bmbt_rec_t));
}

/*
 * Each of the following cases stores data into the same region
 * of the on-disk inode, so only one of them can be valid at
 * any given time. While it is possible to have conflicting formats
 * and log flags, e.g. having XFS_ILOG_?DATA set when the fork is
 * in EXTENTS format, this can only happen when the fork has
 * changed formats after being modified but before being flushed.
 * In these cases, the format always takes precedence, because the
 * format indicates the current state of the fork.
 */
/*ARGSUSED*/
2247
STATIC void
L
Linus Torvalds 已提交
2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267
xfs_iflush_fork(
	xfs_inode_t		*ip,
	xfs_dinode_t		*dip,
	xfs_inode_log_item_t	*iip,
	int			whichfork,
	xfs_buf_t		*bp)
{
	char			*cp;
	xfs_ifork_t		*ifp;
	xfs_mount_t		*mp;
#ifdef XFS_TRANS_DEBUG
	int			first;
#endif
	static const short	brootflag[2] =
		{ XFS_ILOG_DBROOT, XFS_ILOG_ABROOT };
	static const short	dataflag[2] =
		{ XFS_ILOG_DDATA, XFS_ILOG_ADATA };
	static const short	extflag[2] =
		{ XFS_ILOG_DEXT, XFS_ILOG_AEXT };

2268 2269
	if (!iip)
		return;
L
Linus Torvalds 已提交
2270 2271 2272 2273 2274
	ifp = XFS_IFORK_PTR(ip, whichfork);
	/*
	 * This can happen if we gave up in iformat in an error path,
	 * for the attribute fork.
	 */
2275
	if (!ifp) {
L
Linus Torvalds 已提交
2276
		ASSERT(whichfork == XFS_ATTR_FORK);
2277
		return;
L
Linus Torvalds 已提交
2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295
	}
	cp = XFS_DFORK_PTR(dip, whichfork);
	mp = ip->i_mount;
	switch (XFS_IFORK_FORMAT(ip, whichfork)) {
	case XFS_DINODE_FMT_LOCAL:
		if ((iip->ili_format.ilf_fields & dataflag[whichfork]) &&
		    (ifp->if_bytes > 0)) {
			ASSERT(ifp->if_u1.if_data != NULL);
			ASSERT(ifp->if_bytes <= XFS_IFORK_SIZE(ip, whichfork));
			memcpy(cp, ifp->if_u1.if_data, ifp->if_bytes);
		}
		break;

	case XFS_DINODE_FMT_EXTENTS:
		ASSERT((ifp->if_flags & XFS_IFEXTENTS) ||
		       !(iip->ili_format.ilf_fields & extflag[whichfork]));
		if ((iip->ili_format.ilf_fields & extflag[whichfork]) &&
		    (ifp->if_bytes > 0)) {
2296
			ASSERT(xfs_iext_get_ext(ifp, 0));
L
Linus Torvalds 已提交
2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
			ASSERT(XFS_IFORK_NEXTENTS(ip, whichfork) > 0);
			(void)xfs_iextents_copy(ip, (xfs_bmbt_rec_t *)cp,
				whichfork);
		}
		break;

	case XFS_DINODE_FMT_BTREE:
		if ((iip->ili_format.ilf_fields & brootflag[whichfork]) &&
		    (ifp->if_broot_bytes > 0)) {
			ASSERT(ifp->if_broot != NULL);
			ASSERT(ifp->if_broot_bytes <=
			       (XFS_IFORK_SIZE(ip, whichfork) +
				XFS_BROOT_SIZE_ADJ));
2310
			xfs_bmbt_to_bmdr(mp, ifp->if_broot, ifp->if_broot_bytes,
L
Linus Torvalds 已提交
2311 2312 2313 2314 2315 2316 2317 2318
				(xfs_bmdr_block_t *)cp,
				XFS_DFORK_SIZE(dip, mp, whichfork));
		}
		break;

	case XFS_DINODE_FMT_DEV:
		if (iip->ili_format.ilf_fields & XFS_ILOG_DEV) {
			ASSERT(whichfork == XFS_DATA_FORK);
C
Christoph Hellwig 已提交
2319
			xfs_dinode_put_rdev(dip, ip->i_df.if_u2.if_rdev);
L
Linus Torvalds 已提交
2320 2321 2322 2323 2324 2325
		}
		break;

	case XFS_DINODE_FMT_UUID:
		if (iip->ili_format.ilf_fields & XFS_ILOG_UUID) {
			ASSERT(whichfork == XFS_DATA_FORK);
C
Christoph Hellwig 已提交
2326 2327 2328
			memcpy(XFS_DFORK_DPTR(dip),
			       &ip->i_df.if_u2.if_uuid,
			       sizeof(uuid_t));
L
Linus Torvalds 已提交
2329 2330 2331 2332 2333 2334 2335 2336 2337
		}
		break;

	default:
		ASSERT(0);
		break;
	}
}

2338 2339 2340 2341 2342 2343
STATIC int
xfs_iflush_cluster(
	xfs_inode_t	*ip,
	xfs_buf_t	*bp)
{
	xfs_mount_t		*mp = ip->i_mount;
D
Dave Chinner 已提交
2344
	struct xfs_perag	*pag;
2345
	unsigned long		first_index, mask;
2346
	unsigned long		inodes_per_cluster;
2347 2348 2349 2350 2351 2352 2353 2354
	int			ilist_size;
	xfs_inode_t		**ilist;
	xfs_inode_t		*iq;
	int			nr_found;
	int			clcount = 0;
	int			bufwasdelwri;
	int			i;

D
Dave Chinner 已提交
2355
	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
2356

2357 2358
	inodes_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog;
	ilist_size = inodes_per_cluster * sizeof(xfs_inode_t *);
2359
	ilist = kmem_alloc(ilist_size, KM_MAYFAIL|KM_NOFS);
2360
	if (!ilist)
2361
		goto out_put;
2362 2363 2364

	mask = ~(((XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog)) - 1);
	first_index = XFS_INO_TO_AGINO(mp, ip->i_ino) & mask;
2365
	rcu_read_lock();
2366 2367
	/* really need a gang lookup range call here */
	nr_found = radix_tree_gang_lookup(&pag->pag_ici_root, (void**)ilist,
2368
					first_index, inodes_per_cluster);
2369 2370 2371 2372 2373 2374 2375
	if (nr_found == 0)
		goto out_free;

	for (i = 0; i < nr_found; i++) {
		iq = ilist[i];
		if (iq == ip)
			continue;
2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390

		/*
		 * because this is an RCU protected lookup, we could find a
		 * recently freed or even reallocated inode during the lookup.
		 * We need to check under the i_flags_lock for a valid inode
		 * here. Skip it if it is not valid or the wrong inode.
		 */
		spin_lock(&ip->i_flags_lock);
		if (!ip->i_ino ||
		    (XFS_INO_TO_AGINO(mp, iq->i_ino) & mask) != first_index) {
			spin_unlock(&ip->i_flags_lock);
			continue;
		}
		spin_unlock(&ip->i_flags_lock);

2391 2392 2393 2394 2395
		/*
		 * Do an un-protected check to see if the inode is dirty and
		 * is a candidate for flushing.  These checks will be repeated
		 * later after the appropriate locks are acquired.
		 */
2396
		if (xfs_inode_clean(iq) && xfs_ipincount(iq) == 0)
2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419
			continue;

		/*
		 * Try to get locks.  If any are unavailable or it is pinned,
		 * then this inode cannot be flushed and is skipped.
		 */

		if (!xfs_ilock_nowait(iq, XFS_ILOCK_SHARED))
			continue;
		if (!xfs_iflock_nowait(iq)) {
			xfs_iunlock(iq, XFS_ILOCK_SHARED);
			continue;
		}
		if (xfs_ipincount(iq)) {
			xfs_ifunlock(iq);
			xfs_iunlock(iq, XFS_ILOCK_SHARED);
			continue;
		}

		/*
		 * arriving here means that this inode can be flushed.  First
		 * re-check that it's dirty before flushing.
		 */
2420 2421
		if (!xfs_inode_clean(iq)) {
			int	error;
2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439
			error = xfs_iflush_int(iq, bp);
			if (error) {
				xfs_iunlock(iq, XFS_ILOCK_SHARED);
				goto cluster_corrupt_out;
			}
			clcount++;
		} else {
			xfs_ifunlock(iq);
		}
		xfs_iunlock(iq, XFS_ILOCK_SHARED);
	}

	if (clcount) {
		XFS_STATS_INC(xs_icluster_flushcnt);
		XFS_STATS_ADD(xs_icluster_flushinode, clcount);
	}

out_free:
2440
	rcu_read_unlock();
2441
	kmem_free(ilist);
2442 2443
out_put:
	xfs_perag_put(pag);
2444 2445 2446 2447 2448 2449 2450 2451
	return 0;


cluster_corrupt_out:
	/*
	 * Corruption detected in the clustering loop.  Invalidate the
	 * inode buffer and shut down the filesystem.
	 */
2452
	rcu_read_unlock();
2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469
	/*
	 * Clean up the buffer.  If it was B_DELWRI, just release it --
	 * brelse can handle it with no problems.  If not, shut down the
	 * filesystem before releasing the buffer.
	 */
	bufwasdelwri = XFS_BUF_ISDELAYWRITE(bp);
	if (bufwasdelwri)
		xfs_buf_relse(bp);

	xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);

	if (!bufwasdelwri) {
		/*
		 * Just like incore_relse: if we have b_iodone functions,
		 * mark the buffer as an error and call them.  Otherwise
		 * mark it as stale and brelse.
		 */
2470
		if (bp->b_iodone) {
2471
			XFS_BUF_UNDONE(bp);
2472
			xfs_buf_stale(bp);
2473
			xfs_buf_ioerror(bp, EIO);
C
Christoph Hellwig 已提交
2474
			xfs_buf_ioend(bp, 0);
2475
		} else {
2476
			xfs_buf_stale(bp);
2477 2478 2479 2480 2481 2482 2483 2484
			xfs_buf_relse(bp);
		}
	}

	/*
	 * Unlocks the flush lock
	 */
	xfs_iflush_abort(iq);
2485
	kmem_free(ilist);
2486
	xfs_perag_put(pag);
2487 2488 2489
	return XFS_ERROR(EFSCORRUPTED);
}

L
Linus Torvalds 已提交
2490 2491 2492
/*
 * xfs_iflush() will write a modified inode's changes out to the
 * inode's on disk home.  The caller must have the inode lock held
2493 2494
 * in at least shared mode and the inode flush completion must be
 * active as well.  The inode lock will still be held upon return from
L
Linus Torvalds 已提交
2495
 * the call and the caller is free to unlock it.
2496
 * The inode flush will be completed when the inode reaches the disk.
L
Linus Torvalds 已提交
2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511
 * The flags indicate how the inode's buffer should be written out.
 */
int
xfs_iflush(
	xfs_inode_t		*ip,
	uint			flags)
{
	xfs_inode_log_item_t	*iip;
	xfs_buf_t		*bp;
	xfs_dinode_t		*dip;
	xfs_mount_t		*mp;
	int			error;

	XFS_STATS_INC(xs_iflush_count);

C
Christoph Hellwig 已提交
2512
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
2513
	ASSERT(!completion_done(&ip->i_flush));
L
Linus Torvalds 已提交
2514 2515 2516 2517 2518 2519 2520
	ASSERT(ip->i_d.di_format != XFS_DINODE_FMT_BTREE ||
	       ip->i_d.di_nextents > ip->i_df.if_ext_max);

	iip = ip->i_itemp;
	mp = ip->i_mount;

	/*
2521
	 * We can't flush the inode until it is unpinned, so wait for it if we
L
Lucas De Marchi 已提交
2522
	 * are allowed to block.  We know no one new can pin it, because we are
2523 2524 2525 2526 2527 2528 2529
	 * holding the inode lock shared and you need to hold it exclusively to
	 * pin the inode.
	 *
	 * If we are not allowed to block, force the log out asynchronously so
	 * that when we come back the inode will be unpinned. If other inodes
	 * in the same cluster are dirty, they will probably write the inode
	 * out for us if they occur after the log force completes.
L
Linus Torvalds 已提交
2530
	 */
2531
	if (!(flags & SYNC_WAIT) && xfs_ipincount(ip)) {
2532 2533 2534 2535
		xfs_iunpin_nowait(ip);
		xfs_ifunlock(ip);
		return EAGAIN;
	}
L
Linus Torvalds 已提交
2536 2537
	xfs_iunpin_wait(ip);

2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550
	/*
	 * For stale inodes we cannot rely on the backing buffer remaining
	 * stale in cache for the remaining life of the stale inode and so
	 * xfs_itobp() below may give us a buffer that no longer contains
	 * inodes below. We have to check this after ensuring the inode is
	 * unpinned so that it is safe to reclaim the stale inode after the
	 * flush call.
	 */
	if (xfs_iflags_test(ip, XFS_ISTALE)) {
		xfs_ifunlock(ip);
		return 0;
	}

L
Linus Torvalds 已提交
2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563
	/*
	 * This may have been unpinned because the filesystem is shutting
	 * down forcibly. If that's the case we must not write this inode
	 * to disk, because the log record didn't make it to disk!
	 */
	if (XFS_FORCED_SHUTDOWN(mp)) {
		ip->i_update_core = 0;
		if (iip)
			iip->ili_format.ilf_fields = 0;
		xfs_ifunlock(ip);
		return XFS_ERROR(EIO);
	}

2564 2565 2566
	/*
	 * Get the buffer containing the on-disk inode.
	 */
2567
	error = xfs_itobp(mp, NULL, ip, &dip, &bp,
2568
				(flags & SYNC_TRYLOCK) ? XBF_TRYLOCK : XBF_LOCK);
2569 2570 2571 2572 2573
	if (error || !bp) {
		xfs_ifunlock(ip);
		return error;
	}

L
Linus Torvalds 已提交
2574 2575 2576 2577
	/*
	 * First flush out the inode that xfs_iflush was called with.
	 */
	error = xfs_iflush_int(ip, bp);
2578
	if (error)
L
Linus Torvalds 已提交
2579 2580
		goto corrupt_out;

2581 2582 2583 2584
	/*
	 * If the buffer is pinned then push on the log now so we won't
	 * get stuck waiting in the write for too long.
	 */
2585
	if (xfs_buf_ispinned(bp))
2586
		xfs_log_force(mp, 0);
2587

L
Linus Torvalds 已提交
2588 2589 2590 2591
	/*
	 * inode clustering:
	 * see if other inodes can be gathered into this write
	 */
2592 2593 2594
	error = xfs_iflush_cluster(ip, bp);
	if (error)
		goto cluster_corrupt_out;
L
Linus Torvalds 已提交
2595

2596
	if (flags & SYNC_WAIT)
2597 2598
		error = xfs_bwrite(bp);
	else
2599
		xfs_buf_delwri_queue(bp);
2600 2601

	xfs_buf_relse(bp);
L
Linus Torvalds 已提交
2602 2603 2604 2605
	return error;

corrupt_out:
	xfs_buf_relse(bp);
2606
	xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
L
Linus Torvalds 已提交
2607 2608 2609 2610
cluster_corrupt_out:
	/*
	 * Unlocks the flush lock
	 */
2611
	xfs_iflush_abort(ip);
L
Linus Torvalds 已提交
2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627
	return XFS_ERROR(EFSCORRUPTED);
}


STATIC int
xfs_iflush_int(
	xfs_inode_t		*ip,
	xfs_buf_t		*bp)
{
	xfs_inode_log_item_t	*iip;
	xfs_dinode_t		*dip;
	xfs_mount_t		*mp;
#ifdef XFS_TRANS_DEBUG
	int			first;
#endif

C
Christoph Hellwig 已提交
2628
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
2629
	ASSERT(!completion_done(&ip->i_flush));
L
Linus Torvalds 已提交
2630 2631 2632 2633 2634 2635 2636
	ASSERT(ip->i_d.di_format != XFS_DINODE_FMT_BTREE ||
	       ip->i_d.di_nextents > ip->i_df.if_ext_max);

	iip = ip->i_itemp;
	mp = ip->i_mount;

	/* set *dip = inode's place in the buffer */
2637
	dip = (xfs_dinode_t *)xfs_buf_offset(bp, ip->i_imap.im_boffset);
L
Linus Torvalds 已提交
2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653

	/*
	 * Clear i_update_core before copying out the data.
	 * This is for coordination with our timestamp updates
	 * that don't hold the inode lock. They will always
	 * update the timestamps BEFORE setting i_update_core,
	 * so if we clear i_update_core after they set it we
	 * are guaranteed to see their updates to the timestamps.
	 * I believe that this depends on strongly ordered memory
	 * semantics, but we have that.  We use the SYNCHRONIZE
	 * macro to make sure that the compiler does not reorder
	 * the i_update_core access below the data copy below.
	 */
	ip->i_update_core = 0;
	SYNCHRONIZE();

2654
	/*
2655
	 * Make sure to get the latest timestamps from the Linux inode.
2656
	 */
2657
	xfs_synchronize_times(ip);
2658

2659
	if (XFS_TEST_ERROR(dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC),
L
Linus Torvalds 已提交
2660
			       mp, XFS_ERRTAG_IFLUSH_1, XFS_RANDOM_IFLUSH_1)) {
2661 2662 2663
		xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
			"%s: Bad inode %Lu magic number 0x%x, ptr 0x%p",
			__func__, ip->i_ino, be16_to_cpu(dip->di_magic), dip);
L
Linus Torvalds 已提交
2664 2665 2666 2667
		goto corrupt_out;
	}
	if (XFS_TEST_ERROR(ip->i_d.di_magic != XFS_DINODE_MAGIC,
				mp, XFS_ERRTAG_IFLUSH_2, XFS_RANDOM_IFLUSH_2)) {
2668 2669 2670
		xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
			"%s: Bad inode %Lu, ptr 0x%p, magic number 0x%x",
			__func__, ip->i_ino, ip, ip->i_d.di_magic);
L
Linus Torvalds 已提交
2671 2672
		goto corrupt_out;
	}
2673
	if (S_ISREG(ip->i_d.di_mode)) {
L
Linus Torvalds 已提交
2674 2675 2676 2677
		if (XFS_TEST_ERROR(
		    (ip->i_d.di_format != XFS_DINODE_FMT_EXTENTS) &&
		    (ip->i_d.di_format != XFS_DINODE_FMT_BTREE),
		    mp, XFS_ERRTAG_IFLUSH_3, XFS_RANDOM_IFLUSH_3)) {
2678 2679 2680
			xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
				"%s: Bad regular inode %Lu, ptr 0x%p",
				__func__, ip->i_ino, ip);
L
Linus Torvalds 已提交
2681 2682
			goto corrupt_out;
		}
2683
	} else if (S_ISDIR(ip->i_d.di_mode)) {
L
Linus Torvalds 已提交
2684 2685 2686 2687 2688
		if (XFS_TEST_ERROR(
		    (ip->i_d.di_format != XFS_DINODE_FMT_EXTENTS) &&
		    (ip->i_d.di_format != XFS_DINODE_FMT_BTREE) &&
		    (ip->i_d.di_format != XFS_DINODE_FMT_LOCAL),
		    mp, XFS_ERRTAG_IFLUSH_4, XFS_RANDOM_IFLUSH_4)) {
2689 2690 2691
			xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
				"%s: Bad directory inode %Lu, ptr 0x%p",
				__func__, ip->i_ino, ip);
L
Linus Torvalds 已提交
2692 2693 2694 2695 2696 2697
			goto corrupt_out;
		}
	}
	if (XFS_TEST_ERROR(ip->i_d.di_nextents + ip->i_d.di_anextents >
				ip->i_d.di_nblocks, mp, XFS_ERRTAG_IFLUSH_5,
				XFS_RANDOM_IFLUSH_5)) {
2698 2699 2700 2701
		xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
			"%s: detected corrupt incore inode %Lu, "
			"total extents = %d, nblocks = %Ld, ptr 0x%p",
			__func__, ip->i_ino,
L
Linus Torvalds 已提交
2702
			ip->i_d.di_nextents + ip->i_d.di_anextents,
2703
			ip->i_d.di_nblocks, ip);
L
Linus Torvalds 已提交
2704 2705 2706 2707
		goto corrupt_out;
	}
	if (XFS_TEST_ERROR(ip->i_d.di_forkoff > mp->m_sb.sb_inodesize,
				mp, XFS_ERRTAG_IFLUSH_6, XFS_RANDOM_IFLUSH_6)) {
2708 2709 2710
		xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
			"%s: bad inode %Lu, forkoff 0x%x, ptr 0x%p",
			__func__, ip->i_ino, ip->i_d.di_forkoff, ip);
L
Linus Torvalds 已提交
2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725
		goto corrupt_out;
	}
	/*
	 * bump the flush iteration count, used to detect flushes which
	 * postdate a log record during recovery.
	 */

	ip->i_d.di_flushiter++;

	/*
	 * Copy the dirty parts of the inode into the on-disk
	 * inode.  We always copy out the core of the inode,
	 * because if the inode is dirty at all the core must
	 * be.
	 */
C
Christoph Hellwig 已提交
2726
	xfs_dinode_to_disk(dip, &ip->i_d);
L
Linus Torvalds 已提交
2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737

	/* Wrap, we never let the log put out DI_MAX_FLUSH */
	if (ip->i_d.di_flushiter == DI_MAX_FLUSH)
		ip->i_d.di_flushiter = 0;

	/*
	 * If this is really an old format inode and the superblock version
	 * has not been updated to support only new format inodes, then
	 * convert back to the old inode format.  If the superblock version
	 * has been updated, then make the conversion permanent.
	 */
2738 2739
	ASSERT(ip->i_d.di_version == 1 || xfs_sb_version_hasnlink(&mp->m_sb));
	if (ip->i_d.di_version == 1) {
2740
		if (!xfs_sb_version_hasnlink(&mp->m_sb)) {
L
Linus Torvalds 已提交
2741 2742 2743 2744
			/*
			 * Convert it back.
			 */
			ASSERT(ip->i_d.di_nlink <= XFS_MAXLINK_1);
C
Christoph Hellwig 已提交
2745
			dip->di_onlink = cpu_to_be16(ip->i_d.di_nlink);
L
Linus Torvalds 已提交
2746 2747 2748 2749 2750 2751
		} else {
			/*
			 * The superblock version has already been bumped,
			 * so just make the conversion to the new inode
			 * format permanent.
			 */
2752 2753
			ip->i_d.di_version = 2;
			dip->di_version = 2;
L
Linus Torvalds 已提交
2754
			ip->i_d.di_onlink = 0;
C
Christoph Hellwig 已提交
2755
			dip->di_onlink = 0;
L
Linus Torvalds 已提交
2756
			memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));
C
Christoph Hellwig 已提交
2757 2758
			memset(&(dip->di_pad[0]), 0,
			      sizeof(dip->di_pad));
2759
			ASSERT(xfs_get_projid(ip) == 0);
L
Linus Torvalds 已提交
2760 2761 2762
		}
	}

2763 2764 2765
	xfs_iflush_fork(ip, dip, iip, XFS_DATA_FORK, bp);
	if (XFS_IFORK_Q(ip))
		xfs_iflush_fork(ip, dip, iip, XFS_ATTR_FORK, bp);
L
Linus Torvalds 已提交
2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800
	xfs_inobp_check(mp, bp);

	/*
	 * We've recorded everything logged in the inode, so we'd
	 * like to clear the ilf_fields bits so we don't log and
	 * flush things unnecessarily.  However, we can't stop
	 * logging all this information until the data we've copied
	 * into the disk buffer is written to disk.  If we did we might
	 * overwrite the copy of the inode in the log with all the
	 * data after re-logging only part of it, and in the face of
	 * a crash we wouldn't have all the data we need to recover.
	 *
	 * What we do is move the bits to the ili_last_fields field.
	 * When logging the inode, these bits are moved back to the
	 * ilf_fields field.  In the xfs_iflush_done() routine we
	 * clear ili_last_fields, since we know that the information
	 * those bits represent is permanently on disk.  As long as
	 * the flush completes before the inode is logged again, then
	 * both ilf_fields and ili_last_fields will be cleared.
	 *
	 * We can play with the ilf_fields bits here, because the inode
	 * lock must be held exclusively in order to set bits there
	 * and the flush lock protects the ili_last_fields bits.
	 * Set ili_logged so the flush done
	 * routine can tell whether or not to look in the AIL.
	 * Also, store the current LSN of the inode so that we can tell
	 * whether the item has moved in the AIL from xfs_iflush_done().
	 * In order to read the lsn we need the AIL lock, because
	 * it is a 64 bit value that cannot be read atomically.
	 */
	if (iip != NULL && iip->ili_format.ilf_fields != 0) {
		iip->ili_last_fields = iip->ili_format.ilf_fields;
		iip->ili_format.ilf_fields = 0;
		iip->ili_logged = 1;

2801 2802
		xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn,
					&iip->ili_item.li_lsn);
L
Linus Torvalds 已提交
2803 2804 2805 2806 2807 2808 2809

		/*
		 * Attach the function xfs_iflush_done to the inode's
		 * buffer.  This will remove the inode from the AIL
		 * and unlock the inode's flush lock when the inode is
		 * completely written to disk.
		 */
2810
		xfs_buf_attach_iodone(bp, xfs_iflush_done, &iip->ili_item);
L
Linus Torvalds 已提交
2811

2812
		ASSERT(bp->b_fspriv != NULL);
2813
		ASSERT(bp->b_iodone != NULL);
L
Linus Torvalds 已提交
2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837
	} else {
		/*
		 * We're flushing an inode which is not in the AIL and has
		 * not been logged but has i_update_core set.  For this
		 * case we can use a B_DELWRI flush and immediately drop
		 * the inode flush lock because we can avoid the whole
		 * AIL state thing.  It's OK to drop the flush lock now,
		 * because we've already locked the buffer and to do anything
		 * you really need both.
		 */
		if (iip != NULL) {
			ASSERT(iip->ili_logged == 0);
			ASSERT(iip->ili_last_fields == 0);
			ASSERT((iip->ili_item.li_flags & XFS_LI_IN_AIL) == 0);
		}
		xfs_ifunlock(ip);
	}

	return 0;

corrupt_out:
	return XFS_ERROR(EFSCORRUPTED);
}

2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858
void
xfs_promote_inode(
	struct xfs_inode	*ip)
{
	struct xfs_buf		*bp;

	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));

	bp = xfs_incore(ip->i_mount->m_ddev_targp, ip->i_imap.im_blkno,
			ip->i_imap.im_len, XBF_TRYLOCK);
	if (!bp)
		return;

	if (XFS_BUF_ISDELAYWRITE(bp)) {
		xfs_buf_delwri_promote(bp);
		wake_up_process(ip->i_mount->m_ddev_targp->bt_task);
	}

	xfs_buf_relse(bp);
}

2859 2860 2861
/*
 * Return a pointer to the extent record at file index idx.
 */
2862
xfs_bmbt_rec_host_t *
2863 2864 2865 2866 2867
xfs_iext_get_ext(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_extnum_t	idx)		/* index of target extent */
{
	ASSERT(idx >= 0);
2868 2869
	ASSERT(idx < ifp->if_bytes / sizeof(xfs_bmbt_rec_t));

2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
	if ((ifp->if_flags & XFS_IFEXTIREC) && (idx == 0)) {
		return ifp->if_u1.if_ext_irec->er_extbuf;
	} else if (ifp->if_flags & XFS_IFEXTIREC) {
		xfs_ext_irec_t	*erp;		/* irec pointer */
		int		erp_idx = 0;	/* irec index */
		xfs_extnum_t	page_idx = idx;	/* ext index in target list */

		erp = xfs_iext_idx_to_irec(ifp, &page_idx, &erp_idx, 0);
		return &erp->er_extbuf[page_idx];
	} else if (ifp->if_bytes) {
2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891
		return &ifp->if_u1.if_extents[idx];
	} else {
		return NULL;
	}
}

/*
 * Insert new item(s) into the extent records for incore inode
 * fork 'ifp'.  'count' new items are inserted at index 'idx'.
 */
void
xfs_iext_insert(
2892
	xfs_inode_t	*ip,		/* incore inode pointer */
2893 2894
	xfs_extnum_t	idx,		/* starting index of new items */
	xfs_extnum_t	count,		/* number of inserted items */
2895 2896
	xfs_bmbt_irec_t	*new,		/* items to insert */
	int		state)		/* type of extent conversion */
2897
{
2898
	xfs_ifork_t	*ifp = (state & BMAP_ATTRFORK) ? ip->i_afp : &ip->i_df;
2899 2900
	xfs_extnum_t	i;		/* extent record index */

C
Christoph Hellwig 已提交
2901 2902
	trace_xfs_iext_insert(ip, idx, new, state, _RET_IP_);

2903 2904
	ASSERT(ifp->if_flags & XFS_IFEXTENTS);
	xfs_iext_add(ifp, idx, count);
2905 2906
	for (i = idx; i < idx + count; i++, new++)
		xfs_bmbt_set_all(xfs_iext_get_ext(ifp, i), new);
2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924
}

/*
 * This is called when the amount of space required for incore file
 * extents needs to be increased. The ext_diff parameter stores the
 * number of new extents being added and the idx parameter contains
 * the extent index where the new extents will be added. If the new
 * extents are being appended, then we just need to (re)allocate and
 * initialize the space. Otherwise, if the new extents are being
 * inserted into the middle of the existing entries, a bit more work
 * is required to make room for the new extents to be inserted. The
 * caller is responsible for filling in the new extent entries upon
 * return.
 */
void
xfs_iext_add(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_extnum_t	idx,		/* index to begin adding exts */
2925
	int		ext_diff)	/* number of extents to add */
2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955
{
	int		byte_diff;	/* new bytes being added */
	int		new_size;	/* size of extents after adding */
	xfs_extnum_t	nextents;	/* number of extents in file */

	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
	ASSERT((idx >= 0) && (idx <= nextents));
	byte_diff = ext_diff * sizeof(xfs_bmbt_rec_t);
	new_size = ifp->if_bytes + byte_diff;
	/*
	 * If the new number of extents (nextents + ext_diff)
	 * fits inside the inode, then continue to use the inline
	 * extent buffer.
	 */
	if (nextents + ext_diff <= XFS_INLINE_EXTS) {
		if (idx < nextents) {
			memmove(&ifp->if_u2.if_inline_ext[idx + ext_diff],
				&ifp->if_u2.if_inline_ext[idx],
				(nextents - idx) * sizeof(xfs_bmbt_rec_t));
			memset(&ifp->if_u2.if_inline_ext[idx], 0, byte_diff);
		}
		ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext;
		ifp->if_real_bytes = 0;
	}
	/*
	 * Otherwise use a linear (direct) extent list.
	 * If the extents are currently inside the inode,
	 * xfs_iext_realloc_direct will switch us from
	 * inline to direct extent allocation mode.
	 */
2956
	else if (nextents + ext_diff <= XFS_LINEAR_EXTS) {
2957 2958 2959 2960 2961 2962 2963 2964
		xfs_iext_realloc_direct(ifp, new_size);
		if (idx < nextents) {
			memmove(&ifp->if_u1.if_extents[idx + ext_diff],
				&ifp->if_u1.if_extents[idx],
				(nextents - idx) * sizeof(xfs_bmbt_rec_t));
			memset(&ifp->if_u1.if_extents[idx], 0, byte_diff);
		}
	}
2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014
	/* Indirection array */
	else {
		xfs_ext_irec_t	*erp;
		int		erp_idx = 0;
		int		page_idx = idx;

		ASSERT(nextents + ext_diff > XFS_LINEAR_EXTS);
		if (ifp->if_flags & XFS_IFEXTIREC) {
			erp = xfs_iext_idx_to_irec(ifp, &page_idx, &erp_idx, 1);
		} else {
			xfs_iext_irec_init(ifp);
			ASSERT(ifp->if_flags & XFS_IFEXTIREC);
			erp = ifp->if_u1.if_ext_irec;
		}
		/* Extents fit in target extent page */
		if (erp && erp->er_extcount + ext_diff <= XFS_LINEAR_EXTS) {
			if (page_idx < erp->er_extcount) {
				memmove(&erp->er_extbuf[page_idx + ext_diff],
					&erp->er_extbuf[page_idx],
					(erp->er_extcount - page_idx) *
					sizeof(xfs_bmbt_rec_t));
				memset(&erp->er_extbuf[page_idx], 0, byte_diff);
			}
			erp->er_extcount += ext_diff;
			xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, ext_diff);
		}
		/* Insert a new extent page */
		else if (erp) {
			xfs_iext_add_indirect_multi(ifp,
				erp_idx, page_idx, ext_diff);
		}
		/*
		 * If extent(s) are being appended to the last page in
		 * the indirection array and the new extent(s) don't fit
		 * in the page, then erp is NULL and erp_idx is set to
		 * the next index needed in the indirection array.
		 */
		else {
			int	count = ext_diff;

			while (count) {
				erp = xfs_iext_irec_new(ifp, erp_idx);
				erp->er_extcount = count;
				count -= MIN(count, (int)XFS_LINEAR_EXTS);
				if (count) {
					erp_idx++;
				}
			}
		}
	}
3015 3016 3017
	ifp->if_bytes = new_size;
}

3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058
/*
 * This is called when incore extents are being added to the indirection
 * array and the new extents do not fit in the target extent list. The
 * erp_idx parameter contains the irec index for the target extent list
 * in the indirection array, and the idx parameter contains the extent
 * index within the list. The number of extents being added is stored
 * in the count parameter.
 *
 *    |-------|   |-------|
 *    |       |   |       |    idx - number of extents before idx
 *    |  idx  |   | count |
 *    |       |   |       |    count - number of extents being inserted at idx
 *    |-------|   |-------|
 *    | count |   | nex2  |    nex2 - number of extents after idx + count
 *    |-------|   |-------|
 */
void
xfs_iext_add_indirect_multi(
	xfs_ifork_t	*ifp,			/* inode fork pointer */
	int		erp_idx,		/* target extent irec index */
	xfs_extnum_t	idx,			/* index within target list */
	int		count)			/* new extents being added */
{
	int		byte_diff;		/* new bytes being added */
	xfs_ext_irec_t	*erp;			/* pointer to irec entry */
	xfs_extnum_t	ext_diff;		/* number of extents to add */
	xfs_extnum_t	ext_cnt;		/* new extents still needed */
	xfs_extnum_t	nex2;			/* extents after idx + count */
	xfs_bmbt_rec_t	*nex2_ep = NULL;	/* temp list for nex2 extents */
	int		nlists;			/* number of irec's (lists) */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	erp = &ifp->if_u1.if_ext_irec[erp_idx];
	nex2 = erp->er_extcount - idx;
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;

	/*
	 * Save second part of target extent list
	 * (all extents past */
	if (nex2) {
		byte_diff = nex2 * sizeof(xfs_bmbt_rec_t);
3059
		nex2_ep = (xfs_bmbt_rec_t *) kmem_alloc(byte_diff, KM_NOFS);
3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124
		memmove(nex2_ep, &erp->er_extbuf[idx], byte_diff);
		erp->er_extcount -= nex2;
		xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, -nex2);
		memset(&erp->er_extbuf[idx], 0, byte_diff);
	}

	/*
	 * Add the new extents to the end of the target
	 * list, then allocate new irec record(s) and
	 * extent buffer(s) as needed to store the rest
	 * of the new extents.
	 */
	ext_cnt = count;
	ext_diff = MIN(ext_cnt, (int)XFS_LINEAR_EXTS - erp->er_extcount);
	if (ext_diff) {
		erp->er_extcount += ext_diff;
		xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, ext_diff);
		ext_cnt -= ext_diff;
	}
	while (ext_cnt) {
		erp_idx++;
		erp = xfs_iext_irec_new(ifp, erp_idx);
		ext_diff = MIN(ext_cnt, (int)XFS_LINEAR_EXTS);
		erp->er_extcount = ext_diff;
		xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, ext_diff);
		ext_cnt -= ext_diff;
	}

	/* Add nex2 extents back to indirection array */
	if (nex2) {
		xfs_extnum_t	ext_avail;
		int		i;

		byte_diff = nex2 * sizeof(xfs_bmbt_rec_t);
		ext_avail = XFS_LINEAR_EXTS - erp->er_extcount;
		i = 0;
		/*
		 * If nex2 extents fit in the current page, append
		 * nex2_ep after the new extents.
		 */
		if (nex2 <= ext_avail) {
			i = erp->er_extcount;
		}
		/*
		 * Otherwise, check if space is available in the
		 * next page.
		 */
		else if ((erp_idx < nlists - 1) &&
			 (nex2 <= (ext_avail = XFS_LINEAR_EXTS -
			  ifp->if_u1.if_ext_irec[erp_idx+1].er_extcount))) {
			erp_idx++;
			erp++;
			/* Create a hole for nex2 extents */
			memmove(&erp->er_extbuf[nex2], erp->er_extbuf,
				erp->er_extcount * sizeof(xfs_bmbt_rec_t));
		}
		/*
		 * Final choice, create a new extent page for
		 * nex2 extents.
		 */
		else {
			erp_idx++;
			erp = xfs_iext_irec_new(ifp, erp_idx);
		}
		memmove(&erp->er_extbuf[i], nex2_ep, byte_diff);
3125
		kmem_free(nex2_ep);
3126 3127 3128 3129 3130
		erp->er_extcount += nex2;
		xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, nex2);
	}
}

3131 3132 3133 3134 3135
/*
 * This is called when the amount of space required for incore file
 * extents needs to be decreased. The ext_diff parameter stores the
 * number of extents to be removed and the idx parameter contains
 * the extent index where the extents will be removed from.
3136 3137 3138 3139 3140
 *
 * If the amount of space needed has decreased below the linear
 * limit, XFS_IEXT_BUFSZ, then switch to using the contiguous
 * extent array.  Otherwise, use kmem_realloc() to adjust the
 * size to what is needed.
3141 3142 3143
 */
void
xfs_iext_remove(
3144
	xfs_inode_t	*ip,		/* incore inode pointer */
3145
	xfs_extnum_t	idx,		/* index to begin removing exts */
3146 3147
	int		ext_diff,	/* number of extents to remove */
	int		state)		/* type of extent conversion */
3148
{
3149
	xfs_ifork_t	*ifp = (state & BMAP_ATTRFORK) ? ip->i_afp : &ip->i_df;
3150 3151 3152
	xfs_extnum_t	nextents;	/* number of extents in file */
	int		new_size;	/* size of extents after removal */

C
Christoph Hellwig 已提交
3153 3154
	trace_xfs_iext_remove(ip, idx, state, _RET_IP_);

3155 3156 3157 3158 3159 3160
	ASSERT(ext_diff > 0);
	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
	new_size = (nextents - ext_diff) * sizeof(xfs_bmbt_rec_t);

	if (new_size == 0) {
		xfs_iext_destroy(ifp);
3161 3162
	} else if (ifp->if_flags & XFS_IFEXTIREC) {
		xfs_iext_remove_indirect(ifp, idx, ext_diff);
3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182
	} else if (ifp->if_real_bytes) {
		xfs_iext_remove_direct(ifp, idx, ext_diff);
	} else {
		xfs_iext_remove_inline(ifp, idx, ext_diff);
	}
	ifp->if_bytes = new_size;
}

/*
 * This removes ext_diff extents from the inline buffer, beginning
 * at extent index idx.
 */
void
xfs_iext_remove_inline(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_extnum_t	idx,		/* index to begin removing exts */
	int		ext_diff)	/* number of extents to remove */
{
	int		nextents;	/* number of extents in file */

3183
	ASSERT(!(ifp->if_flags & XFS_IFEXTIREC));
3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220
	ASSERT(idx < XFS_INLINE_EXTS);
	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
	ASSERT(((nextents - ext_diff) > 0) &&
		(nextents - ext_diff) < XFS_INLINE_EXTS);

	if (idx + ext_diff < nextents) {
		memmove(&ifp->if_u2.if_inline_ext[idx],
			&ifp->if_u2.if_inline_ext[idx + ext_diff],
			(nextents - (idx + ext_diff)) *
			 sizeof(xfs_bmbt_rec_t));
		memset(&ifp->if_u2.if_inline_ext[nextents - ext_diff],
			0, ext_diff * sizeof(xfs_bmbt_rec_t));
	} else {
		memset(&ifp->if_u2.if_inline_ext[idx], 0,
			ext_diff * sizeof(xfs_bmbt_rec_t));
	}
}

/*
 * This removes ext_diff extents from a linear (direct) extent list,
 * beginning at extent index idx. If the extents are being removed
 * from the end of the list (ie. truncate) then we just need to re-
 * allocate the list to remove the extra space. Otherwise, if the
 * extents are being removed from the middle of the existing extent
 * entries, then we first need to move the extent records beginning
 * at idx + ext_diff up in the list to overwrite the records being
 * removed, then remove the extra space via kmem_realloc.
 */
void
xfs_iext_remove_direct(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_extnum_t	idx,		/* index to begin removing exts */
	int		ext_diff)	/* number of extents to remove */
{
	xfs_extnum_t	nextents;	/* number of extents in file */
	int		new_size;	/* size of extents after removal */

3221
	ASSERT(!(ifp->if_flags & XFS_IFEXTIREC));
3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248
	new_size = ifp->if_bytes -
		(ext_diff * sizeof(xfs_bmbt_rec_t));
	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);

	if (new_size == 0) {
		xfs_iext_destroy(ifp);
		return;
	}
	/* Move extents up in the list (if needed) */
	if (idx + ext_diff < nextents) {
		memmove(&ifp->if_u1.if_extents[idx],
			&ifp->if_u1.if_extents[idx + ext_diff],
			(nextents - (idx + ext_diff)) *
			 sizeof(xfs_bmbt_rec_t));
	}
	memset(&ifp->if_u1.if_extents[nextents - ext_diff],
		0, ext_diff * sizeof(xfs_bmbt_rec_t));
	/*
	 * Reallocate the direct extent list. If the extents
	 * will fit inside the inode then xfs_iext_realloc_direct
	 * will switch from direct to inline extent allocation
	 * mode for us.
	 */
	xfs_iext_realloc_direct(ifp, new_size);
	ifp->if_bytes = new_size;
}

3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324
/*
 * This is called when incore extents are being removed from the
 * indirection array and the extents being removed span multiple extent
 * buffers. The idx parameter contains the file extent index where we
 * want to begin removing extents, and the count parameter contains
 * how many extents need to be removed.
 *
 *    |-------|   |-------|
 *    | nex1  |   |       |    nex1 - number of extents before idx
 *    |-------|   | count |
 *    |       |   |       |    count - number of extents being removed at idx
 *    | count |   |-------|
 *    |       |   | nex2  |    nex2 - number of extents after idx + count
 *    |-------|   |-------|
 */
void
xfs_iext_remove_indirect(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_extnum_t	idx,		/* index to begin removing extents */
	int		count)		/* number of extents to remove */
{
	xfs_ext_irec_t	*erp;		/* indirection array pointer */
	int		erp_idx = 0;	/* indirection array index */
	xfs_extnum_t	ext_cnt;	/* extents left to remove */
	xfs_extnum_t	ext_diff;	/* extents to remove in current list */
	xfs_extnum_t	nex1;		/* number of extents before idx */
	xfs_extnum_t	nex2;		/* extents after idx + count */
	int		page_idx = idx;	/* index in target extent list */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	erp = xfs_iext_idx_to_irec(ifp,  &page_idx, &erp_idx, 0);
	ASSERT(erp != NULL);
	nex1 = page_idx;
	ext_cnt = count;
	while (ext_cnt) {
		nex2 = MAX((erp->er_extcount - (nex1 + ext_cnt)), 0);
		ext_diff = MIN(ext_cnt, (erp->er_extcount - nex1));
		/*
		 * Check for deletion of entire list;
		 * xfs_iext_irec_remove() updates extent offsets.
		 */
		if (ext_diff == erp->er_extcount) {
			xfs_iext_irec_remove(ifp, erp_idx);
			ext_cnt -= ext_diff;
			nex1 = 0;
			if (ext_cnt) {
				ASSERT(erp_idx < ifp->if_real_bytes /
					XFS_IEXT_BUFSZ);
				erp = &ifp->if_u1.if_ext_irec[erp_idx];
				nex1 = 0;
				continue;
			} else {
				break;
			}
		}
		/* Move extents up (if needed) */
		if (nex2) {
			memmove(&erp->er_extbuf[nex1],
				&erp->er_extbuf[nex1 + ext_diff],
				nex2 * sizeof(xfs_bmbt_rec_t));
		}
		/* Zero out rest of page */
		memset(&erp->er_extbuf[nex1 + nex2], 0, (XFS_IEXT_BUFSZ -
			((nex1 + nex2) * sizeof(xfs_bmbt_rec_t))));
		/* Update remaining counters */
		erp->er_extcount -= ext_diff;
		xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, -ext_diff);
		ext_cnt -= ext_diff;
		nex1 = 0;
		erp_idx++;
		erp++;
	}
	ifp->if_bytes -= count * sizeof(xfs_bmbt_rec_t);
	xfs_iext_irec_compact(ifp);
}

3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336
/*
 * Create, destroy, or resize a linear (direct) block of extents.
 */
void
xfs_iext_realloc_direct(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	int		new_size)	/* new size of extents */
{
	int		rnew_size;	/* real new size of extents */

	rnew_size = new_size;

3337 3338 3339 3340
	ASSERT(!(ifp->if_flags & XFS_IFEXTIREC) ||
		((new_size >= 0) && (new_size <= XFS_IEXT_BUFSZ) &&
		 (new_size != ifp->if_real_bytes)));

3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353
	/* Free extent records */
	if (new_size == 0) {
		xfs_iext_destroy(ifp);
	}
	/* Resize direct extent list and zero any new bytes */
	else if (ifp->if_real_bytes) {
		/* Check if extents will fit inside the inode */
		if (new_size <= XFS_INLINE_EXTS * sizeof(xfs_bmbt_rec_t)) {
			xfs_iext_direct_to_inline(ifp, new_size /
				(uint)sizeof(xfs_bmbt_rec_t));
			ifp->if_bytes = new_size;
			return;
		}
3354
		if (!is_power_of_2(new_size)){
3355
			rnew_size = roundup_pow_of_two(new_size);
3356 3357
		}
		if (rnew_size != ifp->if_real_bytes) {
3358
			ifp->if_u1.if_extents =
3359 3360
				kmem_realloc(ifp->if_u1.if_extents,
						rnew_size,
3361
						ifp->if_real_bytes, KM_NOFS);
3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375
		}
		if (rnew_size > ifp->if_real_bytes) {
			memset(&ifp->if_u1.if_extents[ifp->if_bytes /
				(uint)sizeof(xfs_bmbt_rec_t)], 0,
				rnew_size - ifp->if_real_bytes);
		}
	}
	/*
	 * Switch from the inline extent buffer to a direct
	 * extent list. Be sure to include the inline extent
	 * bytes in new_size.
	 */
	else {
		new_size += ifp->if_bytes;
3376
		if (!is_power_of_2(new_size)) {
3377
			rnew_size = roundup_pow_of_two(new_size);
3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401
		}
		xfs_iext_inline_to_direct(ifp, rnew_size);
	}
	ifp->if_real_bytes = rnew_size;
	ifp->if_bytes = new_size;
}

/*
 * Switch from linear (direct) extent records to inline buffer.
 */
void
xfs_iext_direct_to_inline(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_extnum_t	nextents)	/* number of extents in file */
{
	ASSERT(ifp->if_flags & XFS_IFEXTENTS);
	ASSERT(nextents <= XFS_INLINE_EXTS);
	/*
	 * The inline buffer was zeroed when we switched
	 * from inline to direct extent allocation mode,
	 * so we don't need to clear it here.
	 */
	memcpy(ifp->if_u2.if_inline_ext, ifp->if_u1.if_extents,
		nextents * sizeof(xfs_bmbt_rec_t));
3402
	kmem_free(ifp->if_u1.if_extents);
3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419
	ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext;
	ifp->if_real_bytes = 0;
}

/*
 * Switch from inline buffer to linear (direct) extent records.
 * new_size should already be rounded up to the next power of 2
 * by the caller (when appropriate), so use new_size as it is.
 * However, since new_size may be rounded up, we can't update
 * if_bytes here. It is the caller's responsibility to update
 * if_bytes upon return.
 */
void
xfs_iext_inline_to_direct(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	int		new_size)	/* number of extents in file */
{
3420
	ifp->if_u1.if_extents = kmem_alloc(new_size, KM_NOFS);
3421 3422 3423 3424 3425 3426 3427 3428 3429 3430
	memset(ifp->if_u1.if_extents, 0, new_size);
	if (ifp->if_bytes) {
		memcpy(ifp->if_u1.if_extents, ifp->if_u2.if_inline_ext,
			ifp->if_bytes);
		memset(ifp->if_u2.if_inline_ext, 0, XFS_INLINE_EXTS *
			sizeof(xfs_bmbt_rec_t));
	}
	ifp->if_real_bytes = new_size;
}

3431 3432 3433
/*
 * Resize an extent indirection array to new_size bytes.
 */
3434
STATIC void
3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451
xfs_iext_realloc_indirect(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	int		new_size)	/* new indirection array size */
{
	int		nlists;		/* number of irec's (ex lists) */
	int		size;		/* current indirection array size */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
	size = nlists * sizeof(xfs_ext_irec_t);
	ASSERT(ifp->if_real_bytes);
	ASSERT((new_size >= 0) && (new_size != size));
	if (new_size == 0) {
		xfs_iext_destroy(ifp);
	} else {
		ifp->if_u1.if_ext_irec = (xfs_ext_irec_t *)
			kmem_realloc(ifp->if_u1.if_ext_irec,
3452
				new_size, size, KM_NOFS);
3453 3454 3455 3456 3457 3458
	}
}

/*
 * Switch from indirection array to linear (direct) extent allocations.
 */
3459
STATIC void
3460 3461 3462
xfs_iext_indirect_to_direct(
	 xfs_ifork_t	*ifp)		/* inode fork pointer */
{
3463
	xfs_bmbt_rec_host_t *ep;	/* extent record pointer */
3464 3465 3466 3467 3468 3469 3470 3471
	xfs_extnum_t	nextents;	/* number of extents in file */
	int		size;		/* size of file extents */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
	ASSERT(nextents <= XFS_LINEAR_EXTS);
	size = nextents * sizeof(xfs_bmbt_rec_t);

3472
	xfs_iext_irec_compact_pages(ifp);
3473 3474 3475
	ASSERT(ifp->if_real_bytes == XFS_IEXT_BUFSZ);

	ep = ifp->if_u1.if_ext_irec->er_extbuf;
3476
	kmem_free(ifp->if_u1.if_ext_irec);
3477 3478 3479 3480 3481 3482 3483 3484
	ifp->if_flags &= ~XFS_IFEXTIREC;
	ifp->if_u1.if_extents = ep;
	ifp->if_bytes = size;
	if (nextents < XFS_LINEAR_EXTS) {
		xfs_iext_realloc_direct(ifp, size);
	}
}

3485 3486 3487 3488 3489 3490 3491
/*
 * Free incore file extents.
 */
void
xfs_iext_destroy(
	xfs_ifork_t	*ifp)		/* inode fork pointer */
{
3492 3493 3494 3495 3496 3497 3498 3499 3500 3501
	if (ifp->if_flags & XFS_IFEXTIREC) {
		int	erp_idx;
		int	nlists;

		nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
		for (erp_idx = nlists - 1; erp_idx >= 0 ; erp_idx--) {
			xfs_iext_irec_remove(ifp, erp_idx);
		}
		ifp->if_flags &= ~XFS_IFEXTIREC;
	} else if (ifp->if_real_bytes) {
3502
		kmem_free(ifp->if_u1.if_extents);
3503 3504 3505 3506 3507 3508 3509 3510
	} else if (ifp->if_bytes) {
		memset(ifp->if_u2.if_inline_ext, 0, XFS_INLINE_EXTS *
			sizeof(xfs_bmbt_rec_t));
	}
	ifp->if_u1.if_extents = NULL;
	ifp->if_real_bytes = 0;
	ifp->if_bytes = 0;
}
3511

3512 3513 3514
/*
 * Return a pointer to the extent record for file system block bno.
 */
3515
xfs_bmbt_rec_host_t *			/* pointer to found extent record */
3516 3517 3518 3519 3520
xfs_iext_bno_to_ext(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_fileoff_t	bno,		/* block number to search for */
	xfs_extnum_t	*idxp)		/* index of target extent */
{
3521
	xfs_bmbt_rec_host_t *base;	/* pointer to first extent */
3522
	xfs_filblks_t	blockcount = 0;	/* number of blocks in extent */
3523
	xfs_bmbt_rec_host_t *ep = NULL;	/* pointer to target extent */
3524
	xfs_ext_irec_t	*erp = NULL;	/* indirection array pointer */
3525
	int		high;		/* upper boundary in search */
3526
	xfs_extnum_t	idx = 0;	/* index of target extent */
3527
	int		low;		/* lower boundary in search */
3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580
	xfs_extnum_t	nextents;	/* number of file extents */
	xfs_fileoff_t	startoff = 0;	/* start offset of extent */

	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
	if (nextents == 0) {
		*idxp = 0;
		return NULL;
	}
	low = 0;
	if (ifp->if_flags & XFS_IFEXTIREC) {
		/* Find target extent list */
		int	erp_idx = 0;
		erp = xfs_iext_bno_to_irec(ifp, bno, &erp_idx);
		base = erp->er_extbuf;
		high = erp->er_extcount - 1;
	} else {
		base = ifp->if_u1.if_extents;
		high = nextents - 1;
	}
	/* Binary search extent records */
	while (low <= high) {
		idx = (low + high) >> 1;
		ep = base + idx;
		startoff = xfs_bmbt_get_startoff(ep);
		blockcount = xfs_bmbt_get_blockcount(ep);
		if (bno < startoff) {
			high = idx - 1;
		} else if (bno >= startoff + blockcount) {
			low = idx + 1;
		} else {
			/* Convert back to file-based extent index */
			if (ifp->if_flags & XFS_IFEXTIREC) {
				idx += erp->er_extoff;
			}
			*idxp = idx;
			return ep;
		}
	}
	/* Convert back to file-based extent index */
	if (ifp->if_flags & XFS_IFEXTIREC) {
		idx += erp->er_extoff;
	}
	if (bno >= startoff + blockcount) {
		if (++idx == nextents) {
			ep = NULL;
		} else {
			ep = xfs_iext_get_ext(ifp, idx);
		}
	}
	*idxp = idx;
	return ep;
}

3581 3582 3583 3584 3585
/*
 * Return a pointer to the indirection array entry containing the
 * extent record for filesystem block bno. Store the index of the
 * target irec in *erp_idxp.
 */
3586
xfs_ext_irec_t *			/* pointer to found extent record */
3587 3588 3589 3590 3591 3592 3593
xfs_iext_bno_to_irec(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_fileoff_t	bno,		/* block number to search for */
	int		*erp_idxp)	/* irec index of target ext list */
{
	xfs_ext_irec_t	*erp = NULL;	/* indirection array pointer */
	xfs_ext_irec_t	*erp_next;	/* next indirection array entry */
3594
	int		erp_idx;	/* indirection array index */
3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642
	int		nlists;		/* number of extent irec's (lists) */
	int		high;		/* binary search upper limit */
	int		low;		/* binary search lower limit */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
	erp_idx = 0;
	low = 0;
	high = nlists - 1;
	while (low <= high) {
		erp_idx = (low + high) >> 1;
		erp = &ifp->if_u1.if_ext_irec[erp_idx];
		erp_next = erp_idx < nlists - 1 ? erp + 1 : NULL;
		if (bno < xfs_bmbt_get_startoff(erp->er_extbuf)) {
			high = erp_idx - 1;
		} else if (erp_next && bno >=
			   xfs_bmbt_get_startoff(erp_next->er_extbuf)) {
			low = erp_idx + 1;
		} else {
			break;
		}
	}
	*erp_idxp = erp_idx;
	return erp;
}

/*
 * Return a pointer to the indirection array entry containing the
 * extent record at file extent index *idxp. Store the index of the
 * target irec in *erp_idxp and store the page index of the target
 * extent record in *idxp.
 */
xfs_ext_irec_t *
xfs_iext_idx_to_irec(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	xfs_extnum_t	*idxp,		/* extent index (file -> page) */
	int		*erp_idxp,	/* pointer to target irec */
	int		realloc)	/* new bytes were just added */
{
	xfs_ext_irec_t	*prev;		/* pointer to previous irec */
	xfs_ext_irec_t	*erp = NULL;	/* pointer to current irec */
	int		erp_idx;	/* indirection array index */
	int		nlists;		/* number of irec's (ex lists) */
	int		high;		/* binary search upper limit */
	int		low;		/* binary search lower limit */
	xfs_extnum_t	page_idx = *idxp; /* extent index in target list */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
3643 3644 3645 3646
	ASSERT(page_idx >= 0);
	ASSERT(page_idx <= ifp->if_bytes / sizeof(xfs_bmbt_rec_t));
	ASSERT(page_idx < ifp->if_bytes / sizeof(xfs_bmbt_rec_t) || realloc);

3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
	erp_idx = 0;
	low = 0;
	high = nlists - 1;

	/* Binary search extent irec's */
	while (low <= high) {
		erp_idx = (low + high) >> 1;
		erp = &ifp->if_u1.if_ext_irec[erp_idx];
		prev = erp_idx > 0 ? erp - 1 : NULL;
		if (page_idx < erp->er_extoff || (page_idx == erp->er_extoff &&
		     realloc && prev && prev->er_extcount < XFS_LINEAR_EXTS)) {
			high = erp_idx - 1;
		} else if (page_idx > erp->er_extoff + erp->er_extcount ||
			   (page_idx == erp->er_extoff + erp->er_extcount &&
			    !realloc)) {
			low = erp_idx + 1;
		} else if (page_idx == erp->er_extoff + erp->er_extcount &&
			   erp->er_extcount == XFS_LINEAR_EXTS) {
			ASSERT(realloc);
			page_idx = 0;
			erp_idx++;
			erp = erp_idx < nlists ? erp + 1 : NULL;
			break;
		} else {
			page_idx -= erp->er_extoff;
			break;
		}
	}
	*idxp = page_idx;
	*erp_idxp = erp_idx;
	return(erp);
}

/*
 * Allocate and initialize an indirection array once the space needed
 * for incore extents increases above XFS_IEXT_BUFSZ.
 */
void
xfs_iext_irec_init(
	xfs_ifork_t	*ifp)		/* inode fork pointer */
{
	xfs_ext_irec_t	*erp;		/* indirection array pointer */
	xfs_extnum_t	nextents;	/* number of extents in file */

	ASSERT(!(ifp->if_flags & XFS_IFEXTIREC));
	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
	ASSERT(nextents <= XFS_LINEAR_EXTS);

3696
	erp = kmem_alloc(sizeof(xfs_ext_irec_t), KM_NOFS);
3697 3698

	if (nextents == 0) {
3699
		ifp->if_u1.if_extents = kmem_alloc(XFS_IEXT_BUFSZ, KM_NOFS);
3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746
	} else if (!ifp->if_real_bytes) {
		xfs_iext_inline_to_direct(ifp, XFS_IEXT_BUFSZ);
	} else if (ifp->if_real_bytes < XFS_IEXT_BUFSZ) {
		xfs_iext_realloc_direct(ifp, XFS_IEXT_BUFSZ);
	}
	erp->er_extbuf = ifp->if_u1.if_extents;
	erp->er_extcount = nextents;
	erp->er_extoff = 0;

	ifp->if_flags |= XFS_IFEXTIREC;
	ifp->if_real_bytes = XFS_IEXT_BUFSZ;
	ifp->if_bytes = nextents * sizeof(xfs_bmbt_rec_t);
	ifp->if_u1.if_ext_irec = erp;

	return;
}

/*
 * Allocate and initialize a new entry in the indirection array.
 */
xfs_ext_irec_t *
xfs_iext_irec_new(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	int		erp_idx)	/* index for new irec */
{
	xfs_ext_irec_t	*erp;		/* indirection array pointer */
	int		i;		/* loop counter */
	int		nlists;		/* number of irec's (ex lists) */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;

	/* Resize indirection array */
	xfs_iext_realloc_indirect(ifp, ++nlists *
				  sizeof(xfs_ext_irec_t));
	/*
	 * Move records down in the array so the
	 * new page can use erp_idx.
	 */
	erp = ifp->if_u1.if_ext_irec;
	for (i = nlists - 1; i > erp_idx; i--) {
		memmove(&erp[i], &erp[i-1], sizeof(xfs_ext_irec_t));
	}
	ASSERT(i == erp_idx);

	/* Initialize new extent record */
	erp = ifp->if_u1.if_ext_irec;
3747
	erp[erp_idx].er_extbuf = kmem_alloc(XFS_IEXT_BUFSZ, KM_NOFS);
3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773
	ifp->if_real_bytes = nlists * XFS_IEXT_BUFSZ;
	memset(erp[erp_idx].er_extbuf, 0, XFS_IEXT_BUFSZ);
	erp[erp_idx].er_extcount = 0;
	erp[erp_idx].er_extoff = erp_idx > 0 ?
		erp[erp_idx-1].er_extoff + erp[erp_idx-1].er_extcount : 0;
	return (&erp[erp_idx]);
}

/*
 * Remove a record from the indirection array.
 */
void
xfs_iext_irec_remove(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	int		erp_idx)	/* irec index to remove */
{
	xfs_ext_irec_t	*erp;		/* indirection array pointer */
	int		i;		/* loop counter */
	int		nlists;		/* number of irec's (ex lists) */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
	erp = &ifp->if_u1.if_ext_irec[erp_idx];
	if (erp->er_extbuf) {
		xfs_iext_irec_update_extoffs(ifp, erp_idx + 1,
			-erp->er_extcount);
3774
		kmem_free(erp->er_extbuf);
3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791
	}
	/* Compact extent records */
	erp = ifp->if_u1.if_ext_irec;
	for (i = erp_idx; i < nlists - 1; i++) {
		memmove(&erp[i], &erp[i+1], sizeof(xfs_ext_irec_t));
	}
	/*
	 * Manually free the last extent record from the indirection
	 * array.  A call to xfs_iext_realloc_indirect() with a size
	 * of zero would result in a call to xfs_iext_destroy() which
	 * would in turn call this function again, creating a nasty
	 * infinite loop.
	 */
	if (--nlists) {
		xfs_iext_realloc_indirect(ifp,
			nlists * sizeof(xfs_ext_irec_t));
	} else {
3792
		kmem_free(ifp->if_u1.if_ext_irec);
3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804
	}
	ifp->if_real_bytes = nlists * XFS_IEXT_BUFSZ;
}

/*
 * This is called to clean up large amounts of unused memory allocated
 * by the indirection array.  Before compacting anything though, verify
 * that the indirection array is still needed and switch back to the
 * linear extent list (or even the inline buffer) if possible.  The
 * compaction policy is as follows:
 *
 *    Full Compaction: Extents fit into a single page (or inline buffer)
3805
 * Partial Compaction: Extents occupy less than 50% of allocated space
3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848
 *      No Compaction: Extents occupy at least 50% of allocated space
 */
void
xfs_iext_irec_compact(
	xfs_ifork_t	*ifp)		/* inode fork pointer */
{
	xfs_extnum_t	nextents;	/* number of extents in file */
	int		nlists;		/* number of irec's (ex lists) */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);

	if (nextents == 0) {
		xfs_iext_destroy(ifp);
	} else if (nextents <= XFS_INLINE_EXTS) {
		xfs_iext_indirect_to_direct(ifp);
		xfs_iext_direct_to_inline(ifp, nextents);
	} else if (nextents <= XFS_LINEAR_EXTS) {
		xfs_iext_indirect_to_direct(ifp);
	} else if (nextents < (nlists * XFS_LINEAR_EXTS) >> 1) {
		xfs_iext_irec_compact_pages(ifp);
	}
}

/*
 * Combine extents from neighboring extent pages.
 */
void
xfs_iext_irec_compact_pages(
	xfs_ifork_t	*ifp)		/* inode fork pointer */
{
	xfs_ext_irec_t	*erp, *erp_next;/* pointers to irec entries */
	int		erp_idx = 0;	/* indirection array index */
	int		nlists;		/* number of irec's (ex lists) */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
	while (erp_idx < nlists - 1) {
		erp = &ifp->if_u1.if_ext_irec[erp_idx];
		erp_next = erp + 1;
		if (erp_next->er_extcount <=
		    (XFS_LINEAR_EXTS - erp->er_extcount)) {
3849
			memcpy(&erp->er_extbuf[erp->er_extcount],
3850 3851 3852 3853 3854 3855 3856 3857
				erp_next->er_extbuf, erp_next->er_extcount *
				sizeof(xfs_bmbt_rec_t));
			erp->er_extcount += erp_next->er_extcount;
			/*
			 * Free page before removing extent record
			 * so er_extoffs don't get modified in
			 * xfs_iext_irec_remove.
			 */
3858
			kmem_free(erp_next->er_extbuf);
3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889
			erp_next->er_extbuf = NULL;
			xfs_iext_irec_remove(ifp, erp_idx + 1);
			nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
		} else {
			erp_idx++;
		}
	}
}

/*
 * This is called to update the er_extoff field in the indirection
 * array when extents have been added or removed from one of the
 * extent lists. erp_idx contains the irec index to begin updating
 * at and ext_diff contains the number of extents that were added
 * or removed.
 */
void
xfs_iext_irec_update_extoffs(
	xfs_ifork_t	*ifp,		/* inode fork pointer */
	int		erp_idx,	/* irec index to update */
	int		ext_diff)	/* number of new extents */
{
	int		i;		/* loop counter */
	int		nlists;		/* number of irec's (ex lists */

	ASSERT(ifp->if_flags & XFS_IFEXTIREC);
	nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
	for (i = erp_idx; i < nlists; i++) {
		ifp->if_u1.if_ext_irec[i].er_extoff += ext_diff;
	}
}