xfs_inode.c 99.9 KB
Newer Older
D
Dave Chinner 已提交
1
// SPDX-License-Identifier: GPL-2.0
L
Linus Torvalds 已提交
2
/*
3
 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
4
 * All Rights Reserved.
L
Linus Torvalds 已提交
5
 */
6
#include <linux/log2.h>
J
Jeff Layton 已提交
7
#include <linux/iversion.h>
8

L
Linus Torvalds 已提交
9
#include "xfs.h"
10
#include "xfs_fs.h"
11
#include "xfs_shared.h"
12 13 14
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
L
Linus Torvalds 已提交
15 16
#include "xfs_sb.h"
#include "xfs_mount.h"
17
#include "xfs_defer.h"
18
#include "xfs_inode.h"
19
#include "xfs_da_format.h"
D
Dave Chinner 已提交
20 21
#include "xfs_da_btree.h"
#include "xfs_dir2.h"
22
#include "xfs_attr_sf.h"
D
Dave Chinner 已提交
23
#include "xfs_attr.h"
24 25
#include "xfs_trans_space.h"
#include "xfs_trans.h"
L
Linus Torvalds 已提交
26
#include "xfs_buf_item.h"
27 28 29
#include "xfs_inode_item.h"
#include "xfs_ialloc.h"
#include "xfs_bmap.h"
D
Dave Chinner 已提交
30
#include "xfs_bmap_util.h"
31
#include "xfs_errortag.h"
L
Linus Torvalds 已提交
32 33
#include "xfs_error.h"
#include "xfs_quota.h"
34
#include "xfs_filestream.h"
35
#include "xfs_cksum.h"
C
Christoph Hellwig 已提交
36
#include "xfs_trace.h"
D
Dave Chinner 已提交
37
#include "xfs_icache.h"
D
Dave Chinner 已提交
38
#include "xfs_symlink.h"
39 40
#include "xfs_trans_priv.h"
#include "xfs_log.h"
41
#include "xfs_bmap_btree.h"
42
#include "xfs_reflink.h"
43
#include "xfs_dir2_priv.h"
L
Linus Torvalds 已提交
44 45 46 47

kmem_zone_t *xfs_inode_zone;

/*
48
 * Used in xfs_itruncate_extents().  This is the maximum number of extents
L
Linus Torvalds 已提交
49 50 51 52
 * freed from a file in a single transaction.
 */
#define	XFS_ITRUNC_MAX_EXTENTS	2

53 54 55
STATIC int xfs_iflush_int(struct xfs_inode *, struct xfs_buf *);
STATIC int xfs_iunlink(struct xfs_trans *, struct xfs_inode *);
STATIC int xfs_iunlink_remove(struct xfs_trans *, struct xfs_inode *);
56

57 58 59 60 61 62 63 64 65 66 67 68 69 70
/*
 * helper function to extract extent size hint from inode
 */
xfs_extlen_t
xfs_get_extsz_hint(
	struct xfs_inode	*ip)
{
	if ((ip->i_d.di_flags & XFS_DIFLAG_EXTSIZE) && ip->i_d.di_extsize)
		return ip->i_d.di_extsize;
	if (XFS_IS_REALTIME_INODE(ip))
		return ip->i_mount->m_sb.sb_rextsize;
	return 0;
}

71 72 73
/*
 * Helper function to extract CoW extent size hint from inode.
 * Between the extent size hint and the CoW extent size hint, we
74 75
 * return the greater of the two.  If the value is zero (automatic),
 * use the default size.
76 77 78 79 80 81 82 83 84 85 86 87
 */
xfs_extlen_t
xfs_get_cowextsz_hint(
	struct xfs_inode	*ip)
{
	xfs_extlen_t		a, b;

	a = 0;
	if (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE)
		a = ip->i_d.di_cowextsize;
	b = xfs_get_extsz_hint(ip);

88 89 90 91
	a = max(a, b);
	if (a == 0)
		return XFS_DEFAULT_COWEXTSZ_HINT;
	return a;
92 93
}

94
/*
95 96 97 98 99 100 101 102 103 104
 * These two are wrapper routines around the xfs_ilock() routine used to
 * centralize some grungy code.  They are used in places that wish to lock the
 * inode solely for reading the extents.  The reason these places can't just
 * call xfs_ilock(ip, XFS_ILOCK_SHARED) is that the inode lock also guards to
 * bringing in of the extents from disk for a file in b-tree format.  If the
 * inode is in b-tree format, then we need to lock the inode exclusively until
 * the extents are read in.  Locking it exclusively all the time would limit
 * our parallelism unnecessarily, though.  What we do instead is check to see
 * if the extents have been read in yet, and only lock the inode exclusively
 * if they have not.
105
 *
106
 * The functions return a value which should be given to the corresponding
107
 * xfs_iunlock() call.
108 109
 */
uint
110 111
xfs_ilock_data_map_shared(
	struct xfs_inode	*ip)
112
{
113
	uint			lock_mode = XFS_ILOCK_SHARED;
114

115 116
	if (ip->i_d.di_format == XFS_DINODE_FMT_BTREE &&
	    (ip->i_df.if_flags & XFS_IFEXTENTS) == 0)
117 118 119 120 121
		lock_mode = XFS_ILOCK_EXCL;
	xfs_ilock(ip, lock_mode);
	return lock_mode;
}

122 123 124
uint
xfs_ilock_attr_map_shared(
	struct xfs_inode	*ip)
125
{
126 127 128 129 130 131 132
	uint			lock_mode = XFS_ILOCK_SHARED;

	if (ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE &&
	    (ip->i_afp->if_flags & XFS_IFEXTENTS) == 0)
		lock_mode = XFS_ILOCK_EXCL;
	xfs_ilock(ip, lock_mode);
	return lock_mode;
133 134 135
}

/*
136 137 138
 * In addition to i_rwsem in the VFS inode, the xfs inode contains 2
 * multi-reader locks: i_mmap_lock and the i_lock.  This routine allows
 * various combinations of the locks to be obtained.
139
 *
D
Dave Chinner 已提交
140 141
 * The 3 locks should always be ordered so that the IO lock is obtained first,
 * the mmap lock second and the ilock last in order to prevent deadlock.
142
 *
D
Dave Chinner 已提交
143 144
 * Basic locking order:
 *
145
 * i_rwsem -> i_mmap_lock -> page_lock -> i_ilock
D
Dave Chinner 已提交
146 147 148
 *
 * mmap_sem locking order:
 *
149
 * i_rwsem -> page lock -> mmap_sem
D
Dave Chinner 已提交
150 151 152 153 154 155
 * mmap_sem -> i_mmap_lock -> page_lock
 *
 * The difference in mmap_sem locking order mean that we cannot hold the
 * i_mmap_lock over syscall based read(2)/write(2) based IO. These IO paths can
 * fault in pages during copy in/out (for buffered IO) or require the mmap_sem
 * in get_user_pages() to map the user pages into the kernel address space for
156
 * direct IO. Similarly the i_rwsem cannot be taken inside a page fault because
D
Dave Chinner 已提交
157 158 159
 * page faults already hold the mmap_sem.
 *
 * Hence to serialise fully against both syscall and mmap based IO, we need to
160
 * take both the i_rwsem and the i_mmap_lock. These locks should *only* be both
D
Dave Chinner 已提交
161 162 163
 * taken in places where we need to invalidate the page cache in a race
 * free manner (e.g. truncate, hole punch and other extent manipulation
 * functions).
164 165 166 167 168 169 170 171 172 173 174 175 176 177 178
 */
void
xfs_ilock(
	xfs_inode_t		*ip,
	uint			lock_flags)
{
	trace_xfs_ilock(ip, lock_flags, _RET_IP_);

	/*
	 * You can't set both SHARED and EXCL for the same lock,
	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
	 */
	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
D
Dave Chinner 已提交
179 180
	ASSERT((lock_flags & (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)) !=
	       (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL));
181 182
	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
183
	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_SUBCLASS_MASK)) == 0);
184

185 186 187 188 189 190 191
	if (lock_flags & XFS_IOLOCK_EXCL) {
		down_write_nested(&VFS_I(ip)->i_rwsem,
				  XFS_IOLOCK_DEP(lock_flags));
	} else if (lock_flags & XFS_IOLOCK_SHARED) {
		down_read_nested(&VFS_I(ip)->i_rwsem,
				 XFS_IOLOCK_DEP(lock_flags));
	}
192

D
Dave Chinner 已提交
193 194 195 196 197
	if (lock_flags & XFS_MMAPLOCK_EXCL)
		mrupdate_nested(&ip->i_mmaplock, XFS_MMAPLOCK_DEP(lock_flags));
	else if (lock_flags & XFS_MMAPLOCK_SHARED)
		mraccess_nested(&ip->i_mmaplock, XFS_MMAPLOCK_DEP(lock_flags));

198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
	if (lock_flags & XFS_ILOCK_EXCL)
		mrupdate_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags));
	else if (lock_flags & XFS_ILOCK_SHARED)
		mraccess_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags));
}

/*
 * This is just like xfs_ilock(), except that the caller
 * is guaranteed not to sleep.  It returns 1 if it gets
 * the requested locks and 0 otherwise.  If the IO lock is
 * obtained but the inode lock cannot be, then the IO lock
 * is dropped before returning.
 *
 * ip -- the inode being locked
 * lock_flags -- this parameter indicates the inode's locks to be
 *       to be locked.  See the comment for xfs_ilock() for a list
 *	 of valid values.
 */
int
xfs_ilock_nowait(
	xfs_inode_t		*ip,
	uint			lock_flags)
{
	trace_xfs_ilock_nowait(ip, lock_flags, _RET_IP_);

	/*
	 * You can't set both SHARED and EXCL for the same lock,
	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
	 */
	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
D
Dave Chinner 已提交
230 231
	ASSERT((lock_flags & (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)) !=
	       (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL));
232 233
	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
234
	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_SUBCLASS_MASK)) == 0);
235 236

	if (lock_flags & XFS_IOLOCK_EXCL) {
237
		if (!down_write_trylock(&VFS_I(ip)->i_rwsem))
238 239
			goto out;
	} else if (lock_flags & XFS_IOLOCK_SHARED) {
240
		if (!down_read_trylock(&VFS_I(ip)->i_rwsem))
241 242
			goto out;
	}
D
Dave Chinner 已提交
243 244 245 246 247 248 249 250 251

	if (lock_flags & XFS_MMAPLOCK_EXCL) {
		if (!mrtryupdate(&ip->i_mmaplock))
			goto out_undo_iolock;
	} else if (lock_flags & XFS_MMAPLOCK_SHARED) {
		if (!mrtryaccess(&ip->i_mmaplock))
			goto out_undo_iolock;
	}

252 253
	if (lock_flags & XFS_ILOCK_EXCL) {
		if (!mrtryupdate(&ip->i_lock))
D
Dave Chinner 已提交
254
			goto out_undo_mmaplock;
255 256
	} else if (lock_flags & XFS_ILOCK_SHARED) {
		if (!mrtryaccess(&ip->i_lock))
D
Dave Chinner 已提交
257
			goto out_undo_mmaplock;
258 259 260
	}
	return 1;

D
Dave Chinner 已提交
261 262 263 264 265 266
out_undo_mmaplock:
	if (lock_flags & XFS_MMAPLOCK_EXCL)
		mrunlock_excl(&ip->i_mmaplock);
	else if (lock_flags & XFS_MMAPLOCK_SHARED)
		mrunlock_shared(&ip->i_mmaplock);
out_undo_iolock:
267
	if (lock_flags & XFS_IOLOCK_EXCL)
268
		up_write(&VFS_I(ip)->i_rwsem);
269
	else if (lock_flags & XFS_IOLOCK_SHARED)
270
		up_read(&VFS_I(ip)->i_rwsem);
D
Dave Chinner 已提交
271
out:
272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298
	return 0;
}

/*
 * xfs_iunlock() is used to drop the inode locks acquired with
 * xfs_ilock() and xfs_ilock_nowait().  The caller must pass
 * in the flags given to xfs_ilock() or xfs_ilock_nowait() so
 * that we know which locks to drop.
 *
 * ip -- the inode being unlocked
 * lock_flags -- this parameter indicates the inode's locks to be
 *       to be unlocked.  See the comment for xfs_ilock() for a list
 *	 of valid values for this parameter.
 *
 */
void
xfs_iunlock(
	xfs_inode_t		*ip,
	uint			lock_flags)
{
	/*
	 * You can't set both SHARED and EXCL for the same lock,
	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
	 */
	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
D
Dave Chinner 已提交
299 300
	ASSERT((lock_flags & (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)) !=
	       (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL));
301 302
	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
303
	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_SUBCLASS_MASK)) == 0);
304 305 306
	ASSERT(lock_flags != 0);

	if (lock_flags & XFS_IOLOCK_EXCL)
307
		up_write(&VFS_I(ip)->i_rwsem);
308
	else if (lock_flags & XFS_IOLOCK_SHARED)
309
		up_read(&VFS_I(ip)->i_rwsem);
310

D
Dave Chinner 已提交
311 312 313 314 315
	if (lock_flags & XFS_MMAPLOCK_EXCL)
		mrunlock_excl(&ip->i_mmaplock);
	else if (lock_flags & XFS_MMAPLOCK_SHARED)
		mrunlock_shared(&ip->i_mmaplock);

316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
	if (lock_flags & XFS_ILOCK_EXCL)
		mrunlock_excl(&ip->i_lock);
	else if (lock_flags & XFS_ILOCK_SHARED)
		mrunlock_shared(&ip->i_lock);

	trace_xfs_iunlock(ip, lock_flags, _RET_IP_);
}

/*
 * give up write locks.  the i/o lock cannot be held nested
 * if it is being demoted.
 */
void
xfs_ilock_demote(
	xfs_inode_t		*ip,
	uint			lock_flags)
{
D
Dave Chinner 已提交
333 334 335
	ASSERT(lock_flags & (XFS_IOLOCK_EXCL|XFS_MMAPLOCK_EXCL|XFS_ILOCK_EXCL));
	ASSERT((lock_flags &
		~(XFS_IOLOCK_EXCL|XFS_MMAPLOCK_EXCL|XFS_ILOCK_EXCL)) == 0);
336 337 338

	if (lock_flags & XFS_ILOCK_EXCL)
		mrdemote(&ip->i_lock);
D
Dave Chinner 已提交
339 340
	if (lock_flags & XFS_MMAPLOCK_EXCL)
		mrdemote(&ip->i_mmaplock);
341
	if (lock_flags & XFS_IOLOCK_EXCL)
342
		downgrade_write(&VFS_I(ip)->i_rwsem);
343 344 345 346

	trace_xfs_ilock_demote(ip, lock_flags, _RET_IP_);
}

D
Dave Chinner 已提交
347
#if defined(DEBUG) || defined(XFS_WARN)
348 349 350 351 352 353 354 355 356 357 358
int
xfs_isilocked(
	xfs_inode_t		*ip,
	uint			lock_flags)
{
	if (lock_flags & (XFS_ILOCK_EXCL|XFS_ILOCK_SHARED)) {
		if (!(lock_flags & XFS_ILOCK_SHARED))
			return !!ip->i_lock.mr_writer;
		return rwsem_is_locked(&ip->i_lock.mr_lock);
	}

D
Dave Chinner 已提交
359 360 361 362 363 364
	if (lock_flags & (XFS_MMAPLOCK_EXCL|XFS_MMAPLOCK_SHARED)) {
		if (!(lock_flags & XFS_MMAPLOCK_SHARED))
			return !!ip->i_mmaplock.mr_writer;
		return rwsem_is_locked(&ip->i_mmaplock.mr_lock);
	}

365 366
	if (lock_flags & (XFS_IOLOCK_EXCL|XFS_IOLOCK_SHARED)) {
		if (!(lock_flags & XFS_IOLOCK_SHARED))
367 368 369
			return !debug_locks ||
				lockdep_is_held_type(&VFS_I(ip)->i_rwsem, 0);
		return rwsem_is_locked(&VFS_I(ip)->i_rwsem);
370 371 372 373 374 375 376
	}

	ASSERT(0);
	return 0;
}
#endif

377 378 379 380 381 382 383
/*
 * xfs_lockdep_subclass_ok() is only used in an ASSERT, so is only called when
 * DEBUG or XFS_WARN is set. And MAX_LOCKDEP_SUBCLASSES is then only defined
 * when CONFIG_LOCKDEP is set. Hence the complex define below to avoid build
 * errors and warnings.
 */
#if (defined(DEBUG) || defined(XFS_WARN)) && defined(CONFIG_LOCKDEP)
384 385 386 387 388 389 390 391 392 393
static bool
xfs_lockdep_subclass_ok(
	int subclass)
{
	return subclass < MAX_LOCKDEP_SUBCLASSES;
}
#else
#define xfs_lockdep_subclass_ok(subclass)	(true)
#endif

D
Dave Chinner 已提交
394
/*
D
Dave Chinner 已提交
395
 * Bump the subclass so xfs_lock_inodes() acquires each lock with a different
396 397 398
 * value. This can be called for any type of inode lock combination, including
 * parent locking. Care must be taken to ensure we don't overrun the subclass
 * storage fields in the class mask we build.
D
Dave Chinner 已提交
399 400 401 402
 */
static inline int
xfs_lock_inumorder(int lock_mode, int subclass)
{
403 404 405 406
	int	class = 0;

	ASSERT(!(lock_mode & (XFS_ILOCK_PARENT | XFS_ILOCK_RTBITMAP |
			      XFS_ILOCK_RTSUM)));
407
	ASSERT(xfs_lockdep_subclass_ok(subclass));
408

D
Dave Chinner 已提交
409
	if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL)) {
410 411
		ASSERT(subclass <= XFS_IOLOCK_MAX_SUBCLASS);
		class += subclass << XFS_IOLOCK_SHIFT;
D
Dave Chinner 已提交
412 413 414
	}

	if (lock_mode & (XFS_MMAPLOCK_SHARED|XFS_MMAPLOCK_EXCL)) {
415 416
		ASSERT(subclass <= XFS_MMAPLOCK_MAX_SUBCLASS);
		class += subclass << XFS_MMAPLOCK_SHIFT;
D
Dave Chinner 已提交
417 418
	}

419 420 421 422
	if (lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)) {
		ASSERT(subclass <= XFS_ILOCK_MAX_SUBCLASS);
		class += subclass << XFS_ILOCK_SHIFT;
	}
D
Dave Chinner 已提交
423

424
	return (lock_mode & ~XFS_LOCK_SUBCLASS_MASK) | class;
D
Dave Chinner 已提交
425 426 427
}

/*
428 429
 * The following routine will lock n inodes in exclusive mode.  We assume the
 * caller calls us with the inodes in i_ino order.
D
Dave Chinner 已提交
430
 *
431 432 433 434 435
 * We need to detect deadlock where an inode that we lock is in the AIL and we
 * start waiting for another inode that is locked by a thread in a long running
 * transaction (such as truncate). This can result in deadlock since the long
 * running trans might need to wait for the inode we just locked in order to
 * push the tail and free space in the log.
436 437 438 439 440
 *
 * xfs_lock_inodes() can only be used to lock one type of lock at a time -
 * the iolock, the mmaplock or the ilock, but not more than one at a time. If we
 * lock more than one at a time, lockdep will report false positives saying we
 * have violated locking orders.
D
Dave Chinner 已提交
441
 */
442
static void
D
Dave Chinner 已提交
443 444 445 446 447 448 449 450
xfs_lock_inodes(
	xfs_inode_t	**ips,
	int		inodes,
	uint		lock_mode)
{
	int		attempts = 0, i, j, try_lock;
	xfs_log_item_t	*lp;

451 452 453 454 455 456 457
	/*
	 * Currently supports between 2 and 5 inodes with exclusive locking.  We
	 * support an arbitrary depth of locking here, but absolute limits on
	 * inodes depend on the the type of locking and the limits placed by
	 * lockdep annotations in xfs_lock_inumorder.  These are all checked by
	 * the asserts.
	 */
458
	ASSERT(ips && inodes >= 2 && inodes <= 5);
459 460 461 462 463 464 465 466 467 468 469 470 471
	ASSERT(lock_mode & (XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL |
			    XFS_ILOCK_EXCL));
	ASSERT(!(lock_mode & (XFS_IOLOCK_SHARED | XFS_MMAPLOCK_SHARED |
			      XFS_ILOCK_SHARED)));
	ASSERT(!(lock_mode & XFS_MMAPLOCK_EXCL) ||
		inodes <= XFS_MMAPLOCK_MAX_SUBCLASS + 1);
	ASSERT(!(lock_mode & XFS_ILOCK_EXCL) ||
		inodes <= XFS_ILOCK_MAX_SUBCLASS + 1);

	if (lock_mode & XFS_IOLOCK_EXCL) {
		ASSERT(!(lock_mode & (XFS_MMAPLOCK_EXCL | XFS_ILOCK_EXCL)));
	} else if (lock_mode & XFS_MMAPLOCK_EXCL)
		ASSERT(!(lock_mode & XFS_ILOCK_EXCL));
D
Dave Chinner 已提交
472 473 474 475 476 477 478

	try_lock = 0;
	i = 0;
again:
	for (; i < inodes; i++) {
		ASSERT(ips[i]);

479
		if (i && (ips[i] == ips[i - 1]))	/* Already locked */
D
Dave Chinner 已提交
480 481 482
			continue;

		/*
483 484
		 * If try_lock is not set yet, make sure all locked inodes are
		 * not in the AIL.  If any are, set try_lock to be used later.
D
Dave Chinner 已提交
485 486 487 488
		 */
		if (!try_lock) {
			for (j = (i - 1); j >= 0 && !try_lock; j--) {
				lp = (xfs_log_item_t *)ips[j]->i_itemp;
D
Dave Chinner 已提交
489
				if (lp && test_bit(XFS_LI_IN_AIL, &lp->li_flags))
D
Dave Chinner 已提交
490 491 492 493 494 495 496 497 498 499
					try_lock++;
			}
		}

		/*
		 * If any of the previous locks we have locked is in the AIL,
		 * we must TRY to get the second and subsequent locks. If
		 * we can't get any, we must release all we have
		 * and try again.
		 */
500 501 502 503 504 505 506 507 508
		if (!try_lock) {
			xfs_ilock(ips[i], xfs_lock_inumorder(lock_mode, i));
			continue;
		}

		/* try_lock means we have an inode locked that is in the AIL. */
		ASSERT(i != 0);
		if (xfs_ilock_nowait(ips[i], xfs_lock_inumorder(lock_mode, i)))
			continue;
D
Dave Chinner 已提交
509

510 511 512 513 514 515
		/*
		 * Unlock all previous guys and try again.  xfs_iunlock will try
		 * to push the tail if the inode is in the AIL.
		 */
		attempts++;
		for (j = i - 1; j >= 0; j--) {
D
Dave Chinner 已提交
516
			/*
517 518 519
			 * Check to see if we've already unlocked this one.  Not
			 * the first one going back, and the inode ptr is the
			 * same.
D
Dave Chinner 已提交
520
			 */
521 522
			if (j != (i - 1) && ips[j] == ips[j + 1])
				continue;
D
Dave Chinner 已提交
523

524 525
			xfs_iunlock(ips[j], lock_mode);
		}
D
Dave Chinner 已提交
526

527 528
		if ((attempts % 5) == 0) {
			delay(1); /* Don't just spin the CPU */
D
Dave Chinner 已提交
529
		}
530 531 532
		i = 0;
		try_lock = 0;
		goto again;
D
Dave Chinner 已提交
533 534 535 536
	}
}

/*
D
Dave Chinner 已提交
537
 * xfs_lock_two_inodes() can only be used to lock one type of lock at a time -
538 539 540 541 542
 * the mmaplock or the ilock, but not more than one type at a time. If we lock
 * more than one at a time, lockdep will report false positives saying we have
 * violated locking orders.  The iolock must be double-locked separately since
 * we use i_rwsem for that.  We now support taking one lock EXCL and the other
 * SHARED.
D
Dave Chinner 已提交
543 544 545
 */
void
xfs_lock_two_inodes(
546 547 548 549
	struct xfs_inode	*ip0,
	uint			ip0_mode,
	struct xfs_inode	*ip1,
	uint			ip1_mode)
D
Dave Chinner 已提交
550
{
551 552
	struct xfs_inode	*temp;
	uint			mode_temp;
D
Dave Chinner 已提交
553 554 555
	int			attempts = 0;
	xfs_log_item_t		*lp;

556 557 558 559 560 561 562 563 564 565 566 567
	ASSERT(hweight32(ip0_mode) == 1);
	ASSERT(hweight32(ip1_mode) == 1);
	ASSERT(!(ip0_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL)));
	ASSERT(!(ip1_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL)));
	ASSERT(!(ip0_mode & (XFS_MMAPLOCK_SHARED|XFS_MMAPLOCK_EXCL)) ||
	       !(ip0_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)));
	ASSERT(!(ip1_mode & (XFS_MMAPLOCK_SHARED|XFS_MMAPLOCK_EXCL)) ||
	       !(ip1_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)));
	ASSERT(!(ip1_mode & (XFS_MMAPLOCK_SHARED|XFS_MMAPLOCK_EXCL)) ||
	       !(ip0_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)));
	ASSERT(!(ip0_mode & (XFS_MMAPLOCK_SHARED|XFS_MMAPLOCK_EXCL)) ||
	       !(ip1_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)));
D
Dave Chinner 已提交
568

D
Dave Chinner 已提交
569 570 571 572 573 574
	ASSERT(ip0->i_ino != ip1->i_ino);

	if (ip0->i_ino > ip1->i_ino) {
		temp = ip0;
		ip0 = ip1;
		ip1 = temp;
575 576 577
		mode_temp = ip0_mode;
		ip0_mode = ip1_mode;
		ip1_mode = mode_temp;
D
Dave Chinner 已提交
578 579 580
	}

 again:
581
	xfs_ilock(ip0, xfs_lock_inumorder(ip0_mode, 0));
D
Dave Chinner 已提交
582 583 584 585 586 587 588

	/*
	 * If the first lock we have locked is in the AIL, we must TRY to get
	 * the second lock. If we can't get it, we must release the first one
	 * and try again.
	 */
	lp = (xfs_log_item_t *)ip0->i_itemp;
D
Dave Chinner 已提交
589
	if (lp && test_bit(XFS_LI_IN_AIL, &lp->li_flags)) {
590 591
		if (!xfs_ilock_nowait(ip1, xfs_lock_inumorder(ip1_mode, 1))) {
			xfs_iunlock(ip0, ip0_mode);
D
Dave Chinner 已提交
592 593 594 595 596
			if ((++attempts % 5) == 0)
				delay(1); /* Don't just spin the CPU */
			goto again;
		}
	} else {
597
		xfs_ilock(ip1, xfs_lock_inumorder(ip1_mode, 1));
D
Dave Chinner 已提交
598 599 600
	}
}

601 602 603 604 605 606 607 608
void
__xfs_iflock(
	struct xfs_inode	*ip)
{
	wait_queue_head_t *wq = bit_waitqueue(&ip->i_flags, __XFS_IFLOCK_BIT);
	DEFINE_WAIT_BIT(wait, &ip->i_flags, __XFS_IFLOCK_BIT);

	do {
609
		prepare_to_wait_exclusive(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
610 611 612 613
		if (xfs_isiflocked(ip))
			io_schedule();
	} while (!xfs_iflock_nowait(ip));

614
	finish_wait(wq, &wait.wq_entry);
615 616
}

L
Linus Torvalds 已提交
617 618
STATIC uint
_xfs_dic2xflags(
619
	uint16_t		di_flags,
620 621
	uint64_t		di_flags2,
	bool			has_attr)
L
Linus Torvalds 已提交
622 623 624 625 626
{
	uint			flags = 0;

	if (di_flags & XFS_DIFLAG_ANY) {
		if (di_flags & XFS_DIFLAG_REALTIME)
627
			flags |= FS_XFLAG_REALTIME;
L
Linus Torvalds 已提交
628
		if (di_flags & XFS_DIFLAG_PREALLOC)
629
			flags |= FS_XFLAG_PREALLOC;
L
Linus Torvalds 已提交
630
		if (di_flags & XFS_DIFLAG_IMMUTABLE)
631
			flags |= FS_XFLAG_IMMUTABLE;
L
Linus Torvalds 已提交
632
		if (di_flags & XFS_DIFLAG_APPEND)
633
			flags |= FS_XFLAG_APPEND;
L
Linus Torvalds 已提交
634
		if (di_flags & XFS_DIFLAG_SYNC)
635
			flags |= FS_XFLAG_SYNC;
L
Linus Torvalds 已提交
636
		if (di_flags & XFS_DIFLAG_NOATIME)
637
			flags |= FS_XFLAG_NOATIME;
L
Linus Torvalds 已提交
638
		if (di_flags & XFS_DIFLAG_NODUMP)
639
			flags |= FS_XFLAG_NODUMP;
L
Linus Torvalds 已提交
640
		if (di_flags & XFS_DIFLAG_RTINHERIT)
641
			flags |= FS_XFLAG_RTINHERIT;
L
Linus Torvalds 已提交
642
		if (di_flags & XFS_DIFLAG_PROJINHERIT)
643
			flags |= FS_XFLAG_PROJINHERIT;
L
Linus Torvalds 已提交
644
		if (di_flags & XFS_DIFLAG_NOSYMLINKS)
645
			flags |= FS_XFLAG_NOSYMLINKS;
646
		if (di_flags & XFS_DIFLAG_EXTSIZE)
647
			flags |= FS_XFLAG_EXTSIZE;
648
		if (di_flags & XFS_DIFLAG_EXTSZINHERIT)
649
			flags |= FS_XFLAG_EXTSZINHERIT;
650
		if (di_flags & XFS_DIFLAG_NODEFRAG)
651
			flags |= FS_XFLAG_NODEFRAG;
652
		if (di_flags & XFS_DIFLAG_FILESTREAM)
653
			flags |= FS_XFLAG_FILESTREAM;
L
Linus Torvalds 已提交
654 655
	}

656 657 658
	if (di_flags2 & XFS_DIFLAG2_ANY) {
		if (di_flags2 & XFS_DIFLAG2_DAX)
			flags |= FS_XFLAG_DAX;
659 660
		if (di_flags2 & XFS_DIFLAG2_COWEXTSIZE)
			flags |= FS_XFLAG_COWEXTSIZE;
661 662 663 664 665
	}

	if (has_attr)
		flags |= FS_XFLAG_HASATTR;

L
Linus Torvalds 已提交
666 667 668 669 670
	return flags;
}

uint
xfs_ip2xflags(
671
	struct xfs_inode	*ip)
L
Linus Torvalds 已提交
672
{
673
	struct xfs_icdinode	*dic = &ip->i_d;
L
Linus Torvalds 已提交
674

675
	return _xfs_dic2xflags(dic->di_flags, dic->di_flags2, XFS_IFORK_Q(ip));
L
Linus Torvalds 已提交
676 677
}

D
Dave Chinner 已提交
678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
/*
 * Lookups up an inode from "name". If ci_name is not NULL, then a CI match
 * is allowed, otherwise it has to be an exact match. If a CI match is found,
 * ci_name->name will point to a the actual name (caller must free) or
 * will be set to NULL if an exact match is found.
 */
int
xfs_lookup(
	xfs_inode_t		*dp,
	struct xfs_name		*name,
	xfs_inode_t		**ipp,
	struct xfs_name		*ci_name)
{
	xfs_ino_t		inum;
	int			error;

	trace_xfs_lookup(dp, name);

	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
D
Dave Chinner 已提交
697
		return -EIO;
D
Dave Chinner 已提交
698 699 700

	error = xfs_dir_lookup(NULL, dp, name, &inum, ci_name);
	if (error)
701
		goto out_unlock;
D
Dave Chinner 已提交
702 703 704 705 706 707 708 709 710 711

	error = xfs_iget(dp->i_mount, NULL, inum, 0, 0, ipp);
	if (error)
		goto out_free_name;

	return 0;

out_free_name:
	if (ci_name)
		kmem_free(ci_name->name);
712
out_unlock:
D
Dave Chinner 已提交
713 714 715 716
	*ipp = NULL;
	return error;
}

L
Linus Torvalds 已提交
717 718 719 720 721 722 723
/*
 * 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()
C
Carlos Maiolino 已提交
724 725 726 727
 * 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.
L
Linus Torvalds 已提交
728
 *
C
Carlos Maiolino 已提交
729 730 731 732 733
 * 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 ialloc_context and return.
L
Linus Torvalds 已提交
734 735 736 737 738 739 740 741
 * 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
742 743 744 745 746
 *
 * 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
747
 */
748
static int
L
Linus Torvalds 已提交
749 750 751
xfs_ialloc(
	xfs_trans_t	*tp,
	xfs_inode_t	*pip,
A
Al Viro 已提交
752
	umode_t		mode,
753
	xfs_nlink_t	nlink,
C
Christoph Hellwig 已提交
754
	dev_t		rdev,
755
	prid_t		prid,
L
Linus Torvalds 已提交
756 757 758
	xfs_buf_t	**ialloc_context,
	xfs_inode_t	**ipp)
{
759
	struct xfs_mount *mp = tp->t_mountp;
L
Linus Torvalds 已提交
760 761 762 763
	xfs_ino_t	ino;
	xfs_inode_t	*ip;
	uint		flags;
	int		error;
D
Dave Chinner 已提交
764
	struct timespec	tv;
765
	struct inode	*inode;
L
Linus Torvalds 已提交
766 767 768 769 770

	/*
	 * Call the space management code to pick
	 * the on-disk inode to be allocated.
	 */
771
	error = xfs_dialloc(tp, pip ? pip->i_ino : 0, mode,
772
			    ialloc_context, &ino);
773
	if (error)
L
Linus Torvalds 已提交
774
		return error;
775
	if (*ialloc_context || ino == NULLFSINO) {
L
Linus Torvalds 已提交
776 777 778 779 780
		*ipp = NULL;
		return 0;
	}
	ASSERT(*ialloc_context == NULL);

781 782 783 784 785 786 787 788 789 790 791 792
	/*
	 * Protect against obviously corrupt allocation btree records. Later
	 * xfs_iget checks will catch re-allocation of other active in-memory
	 * and on-disk inodes. If we don't catch reallocating the parent inode
	 * here we will deadlock in xfs_iget() so we have to do these checks
	 * first.
	 */
	if ((pip && ino == pip->i_ino) || !xfs_verify_dir_ino(mp, ino)) {
		xfs_alert(mp, "Allocated a known in-use inode 0x%llx!", ino);
		return -EFSCORRUPTED;
	}

L
Linus Torvalds 已提交
793 794 795 796 797
	/*
	 * 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.
	 */
798
	error = xfs_iget(mp, tp, ino, XFS_IGET_CREATE,
799
			 XFS_ILOCK_EXCL, &ip);
800
	if (error)
L
Linus Torvalds 已提交
801 802
		return error;
	ASSERT(ip != NULL);
803
	inode = VFS_I(ip);
L
Linus Torvalds 已提交
804

805 806 807 808 809 810 811 812
	/*
	 * We always convert v1 inodes to v2 now - we only support filesystems
	 * with >= v2 inode capability, so there is no reason for ever leaving
	 * an inode in v1 format.
	 */
	if (ip->i_d.di_version == 1)
		ip->i_d.di_version = 2;

D
Dave Chinner 已提交
813
	inode->i_mode = mode;
814
	set_nlink(inode, nlink);
815 816
	ip->i_d.di_uid = xfs_kuid_to_uid(current_fsuid());
	ip->i_d.di_gid = xfs_kgid_to_gid(current_fsgid());
C
Christoph Hellwig 已提交
817
	inode->i_rdev = rdev;
818
	xfs_set_projid(ip, prid);
L
Linus Torvalds 已提交
819

820
	if (pip && XFS_INHERIT_GID(pip)) {
L
Linus Torvalds 已提交
821
		ip->i_d.di_gid = pip->i_d.di_gid;
D
Dave Chinner 已提交
822 823
		if ((VFS_I(pip)->i_mode & S_ISGID) && S_ISDIR(mode))
			inode->i_mode |= S_ISGID;
L
Linus Torvalds 已提交
824 825 826 827 828 829 830 831
	}

	/*
	 * 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) &&
D
Dave Chinner 已提交
832 833 834
	    (inode->i_mode & S_ISGID) &&
	    (!in_group_p(xfs_gid_to_kgid(ip->i_d.di_gid))))
		inode->i_mode &= ~S_ISGID;
L
Linus Torvalds 已提交
835 836 837 838

	ip->i_d.di_size = 0;
	ip->i_d.di_nextents = 0;
	ASSERT(ip->i_d.di_nblocks == 0);
839

840
	tv = current_time(inode);
841 842 843
	inode->i_mtime = tv;
	inode->i_atime = tv;
	inode->i_ctime = tv;
844

L
Linus Torvalds 已提交
845 846 847 848
	ip->i_d.di_extsize = 0;
	ip->i_d.di_dmevmask = 0;
	ip->i_d.di_dmstate = 0;
	ip->i_d.di_flags = 0;
849 850

	if (ip->i_d.di_version == 3) {
J
Jeff Layton 已提交
851
		inode_set_iversion(inode, 1);
852
		ip->i_d.di_flags2 = 0;
853
		ip->i_d.di_cowextsize = 0;
854 855
		ip->i_d.di_crtime.t_sec = (int32_t)tv.tv_sec;
		ip->i_d.di_crtime.t_nsec = (int32_t)tv.tv_nsec;
856 857 858
	}


L
Linus Torvalds 已提交
859 860 861 862 863 864 865 866 867 868 869 870
	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_flags = 0;
		flags |= XFS_ILOG_DEV;
		break;
	case S_IFREG:
	case S_IFDIR:
D
David Chinner 已提交
871
		if (pip && (pip->i_d.di_flags & XFS_DIFLAG_ANY)) {
872
			uint		di_flags = 0;
873

874
			if (S_ISDIR(mode)) {
875 876
				if (pip->i_d.di_flags & XFS_DIFLAG_RTINHERIT)
					di_flags |= XFS_DIFLAG_RTINHERIT;
877 878 879 880
				if (pip->i_d.di_flags & XFS_DIFLAG_EXTSZINHERIT) {
					di_flags |= XFS_DIFLAG_EXTSZINHERIT;
					ip->i_d.di_extsize = pip->i_d.di_extsize;
				}
881 882
				if (pip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
					di_flags |= XFS_DIFLAG_PROJINHERIT;
883
			} else if (S_ISREG(mode)) {
C
Christoph Hellwig 已提交
884
				if (pip->i_d.di_flags & XFS_DIFLAG_RTINHERIT)
885
					di_flags |= XFS_DIFLAG_REALTIME;
886 887 888 889
				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
890 891 892
			}
			if ((pip->i_d.di_flags & XFS_DIFLAG_NOATIME) &&
			    xfs_inherit_noatime)
893
				di_flags |= XFS_DIFLAG_NOATIME;
L
Linus Torvalds 已提交
894 895
			if ((pip->i_d.di_flags & XFS_DIFLAG_NODUMP) &&
			    xfs_inherit_nodump)
896
				di_flags |= XFS_DIFLAG_NODUMP;
L
Linus Torvalds 已提交
897 898
			if ((pip->i_d.di_flags & XFS_DIFLAG_SYNC) &&
			    xfs_inherit_sync)
899
				di_flags |= XFS_DIFLAG_SYNC;
L
Linus Torvalds 已提交
900 901
			if ((pip->i_d.di_flags & XFS_DIFLAG_NOSYMLINKS) &&
			    xfs_inherit_nosymlinks)
902
				di_flags |= XFS_DIFLAG_NOSYMLINKS;
903 904 905
			if ((pip->i_d.di_flags & XFS_DIFLAG_NODEFRAG) &&
			    xfs_inherit_nodefrag)
				di_flags |= XFS_DIFLAG_NODEFRAG;
906 907
			if (pip->i_d.di_flags & XFS_DIFLAG_FILESTREAM)
				di_flags |= XFS_DIFLAG_FILESTREAM;
908

909
			ip->i_d.di_flags |= di_flags;
L
Linus Torvalds 已提交
910
		}
911 912 913 914
		if (pip &&
		    (pip->i_d.di_flags2 & XFS_DIFLAG2_ANY) &&
		    pip->i_d.di_version == 3 &&
		    ip->i_d.di_version == 3) {
915 916
			uint64_t	di_flags2 = 0;

917
			if (pip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE) {
918
				di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
919 920
				ip->i_d.di_cowextsize = pip->i_d.di_cowextsize;
			}
921 922 923 924
			if (pip->i_d.di_flags2 & XFS_DIFLAG2_DAX)
				di_flags2 |= XFS_DIFLAG2_DAX;

			ip->i_d.di_flags2 |= di_flags2;
925
		}
L
Linus Torvalds 已提交
926 927 928 929 930
		/* 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;
931
		ip->i_df.if_u1.if_root = NULL;
L
Linus Torvalds 已提交
932 933 934 935 936 937 938 939 940 941 942 943 944
		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.
	 */
945
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
L
Linus Torvalds 已提交
946 947
	xfs_trans_log_inode(tp, ip, flags);

948
	/* now that we have an i_mode we can setup the inode structure */
949
	xfs_setup_inode(ip);
L
Linus Torvalds 已提交
950 951 952 953 954

	*ipp = ip;
	return 0;
}

D
Dave Chinner 已提交
955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972
/*
 * Allocates a new inode from disk and return a pointer to the
 * incore copy. This routine will internally commit the current
 * transaction and allocate a new one if the Space Manager needed
 * to do an allocation to replenish the inode free-list.
 *
 * This routine is designed to be called from xfs_create and
 * xfs_create_dir.
 *
 */
int
xfs_dir_ialloc(
	xfs_trans_t	**tpp,		/* input: current transaction;
					   output: may be a new transaction. */
	xfs_inode_t	*dp,		/* directory within whose allocate
					   the inode. */
	umode_t		mode,
	xfs_nlink_t	nlink,
C
Christoph Hellwig 已提交
973
	dev_t		rdev,
D
Dave Chinner 已提交
974
	prid_t		prid,		/* project id */
975
	xfs_inode_t	**ipp)		/* pointer to inode; it will be
D
Dave Chinner 已提交
976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002
					   locked. */
{
	xfs_trans_t	*tp;
	xfs_inode_t	*ip;
	xfs_buf_t	*ialloc_context = NULL;
	int		code;
	void		*dqinfo;
	uint		tflags;

	tp = *tpp;
	ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);

	/*
	 * xfs_ialloc will return a pointer to an incore inode if
	 * the Space Manager has an available inode on the free
	 * list. Otherwise, it will do an allocation and replenish
	 * the freelist.  Since we can only do one allocation per
	 * transaction without deadlocks, we will need to commit the
	 * current transaction and start a new one.  We will then
	 * need to call xfs_ialloc again to get the inode.
	 *
	 * If xfs_ialloc did an allocation to replenish the freelist,
	 * it returns the bp containing the head of the freelist as
	 * ialloc_context. We will hold a lock on it across the
	 * transaction commit so that no other process can steal
	 * the inode(s) that we've just allocated.
	 */
1003 1004
	code = xfs_ialloc(tp, dp, mode, nlink, rdev, prid, &ialloc_context,
			&ip);
D
Dave Chinner 已提交
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016

	/*
	 * Return an error if we were unable to allocate a new inode.
	 * This should only happen if we run out of space on disk or
	 * encounter a disk error.
	 */
	if (code) {
		*ipp = NULL;
		return code;
	}
	if (!ialloc_context && !ip) {
		*ipp = NULL;
D
Dave Chinner 已提交
1017
		return -ENOSPC;
D
Dave Chinner 已提交
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
	}

	/*
	 * If the AGI buffer is non-NULL, then we were unable to get an
	 * inode in one operation.  We need to commit the current
	 * transaction and call xfs_ialloc() again.  It is guaranteed
	 * to succeed the second time.
	 */
	if (ialloc_context) {
		/*
		 * Normally, xfs_trans_commit releases all the locks.
		 * We call bhold to hang on to the ialloc_context across
		 * the commit.  Holding this buffer prevents any other
		 * processes from doing any allocations in this
		 * allocation group.
		 */
		xfs_trans_bhold(tp, ialloc_context);

		/*
		 * We want the quota changes to be associated with the next
		 * transaction, NOT this one. So, detach the dqinfo from this
		 * and attach it to the next transaction.
		 */
		dqinfo = NULL;
		tflags = 0;
		if (tp->t_dqinfo) {
			dqinfo = (void *)tp->t_dqinfo;
			tp->t_dqinfo = NULL;
			tflags = tp->t_flags & XFS_TRANS_DQ_DIRTY;
			tp->t_flags &= ~(XFS_TRANS_DQ_DIRTY);
		}

C
Christoph Hellwig 已提交
1050
		code = xfs_trans_roll(&tp);
1051

D
Dave Chinner 已提交
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
		/*
		 * Re-attach the quota info that we detached from prev trx.
		 */
		if (dqinfo) {
			tp->t_dqinfo = dqinfo;
			tp->t_flags |= tflags;
		}

		if (code) {
			xfs_buf_relse(ialloc_context);
1062
			*tpp = tp;
D
Dave Chinner 已提交
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
			*ipp = NULL;
			return code;
		}
		xfs_trans_bjoin(tp, ialloc_context);

		/*
		 * Call ialloc again. Since we've locked out all
		 * other allocations in this allocation group,
		 * this call should always succeed.
		 */
		code = xfs_ialloc(tp, dp, mode, nlink, rdev, prid,
1074
				  &ialloc_context, &ip);
D
Dave Chinner 已提交
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095

		/*
		 * If we get an error at this point, return to the caller
		 * so that the current transaction can be aborted.
		 */
		if (code) {
			*tpp = tp;
			*ipp = NULL;
			return code;
		}
		ASSERT(!ialloc_context && ip);

	}

	*ipp = ip;
	*tpp = tp;

	return 0;
}

/*
1096 1097 1098
 * Decrement the link count on an inode & log the change.  If this causes the
 * link count to go to zero, move the inode to AGI unlinked list so that it can
 * be freed when the last active reference goes away via xfs_inactive().
D
Dave Chinner 已提交
1099
 */
1100
static int			/* error */
D
Dave Chinner 已提交
1101 1102 1103 1104 1105 1106 1107 1108 1109
xfs_droplink(
	xfs_trans_t *tp,
	xfs_inode_t *ip)
{
	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);

	drop_nlink(VFS_I(ip));
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

1110 1111 1112 1113
	if (VFS_I(ip)->i_nlink)
		return 0;

	return xfs_iunlink(tp, ip);
D
Dave Chinner 已提交
1114 1115 1116 1117 1118
}

/*
 * Increment the link count on an inode & log the change.
 */
1119
static int
D
Dave Chinner 已提交
1120 1121 1122 1123 1124 1125
xfs_bumplink(
	xfs_trans_t *tp,
	xfs_inode_t *ip)
{
	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);

1126
	ASSERT(ip->i_d.di_version > 1);
D
Dave Chinner 已提交
1127 1128 1129 1130 1131
	inc_nlink(VFS_I(ip));
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	return 0;
}

D
Dave Chinner 已提交
1132 1133 1134 1135 1136
int
xfs_create(
	xfs_inode_t		*dp,
	struct xfs_name		*name,
	umode_t			mode,
C
Christoph Hellwig 已提交
1137
	dev_t			rdev,
D
Dave Chinner 已提交
1138 1139 1140 1141 1142 1143 1144
	xfs_inode_t		**ipp)
{
	int			is_dir = S_ISDIR(mode);
	struct xfs_mount	*mp = dp->i_mount;
	struct xfs_inode	*ip = NULL;
	struct xfs_trans	*tp = NULL;
	int			error;
1145
	struct xfs_defer_ops	dfops;
D
Dave Chinner 已提交
1146 1147 1148 1149 1150 1151
	xfs_fsblock_t		first_block;
	bool                    unlock_dp_on_error = false;
	prid_t			prid;
	struct xfs_dquot	*udqp = NULL;
	struct xfs_dquot	*gdqp = NULL;
	struct xfs_dquot	*pdqp = NULL;
1152
	struct xfs_trans_res	*tres;
D
Dave Chinner 已提交
1153 1154 1155 1156 1157
	uint			resblks;

	trace_xfs_create(dp, name);

	if (XFS_FORCED_SHUTDOWN(mp))
D
Dave Chinner 已提交
1158
		return -EIO;
D
Dave Chinner 已提交
1159

1160
	prid = xfs_get_initial_prid(dp);
D
Dave Chinner 已提交
1161 1162 1163 1164

	/*
	 * Make sure that we have allocated dquot(s) on disk.
	 */
1165 1166
	error = xfs_qm_vop_dqalloc(dp, xfs_kuid_to_uid(current_fsuid()),
					xfs_kgid_to_gid(current_fsgid()), prid,
D
Dave Chinner 已提交
1167 1168 1169 1170 1171 1172 1173
					XFS_QMOPT_QUOTALL | XFS_QMOPT_INHERIT,
					&udqp, &gdqp, &pdqp);
	if (error)
		return error;

	if (is_dir) {
		resblks = XFS_MKDIR_SPACE_RES(mp, name->len);
1174
		tres = &M_RES(mp)->tr_mkdir;
D
Dave Chinner 已提交
1175 1176
	} else {
		resblks = XFS_CREATE_SPACE_RES(mp, name->len);
1177
		tres = &M_RES(mp)->tr_create;
D
Dave Chinner 已提交
1178 1179 1180 1181 1182 1183 1184 1185
	}

	/*
	 * Initially assume that the file does not exist and
	 * reserve the resources for that case.  If that is not
	 * the case we'll drop the one we have and get a more
	 * appropriate transaction later.
	 */
1186
	error = xfs_trans_alloc(mp, tres, resblks, 0, 0, &tp);
D
Dave Chinner 已提交
1187
	if (error == -ENOSPC) {
D
Dave Chinner 已提交
1188 1189
		/* flush outstanding delalloc blocks and retry */
		xfs_flush_inodes(mp);
1190
		error = xfs_trans_alloc(mp, tres, resblks, 0, 0, &tp);
D
Dave Chinner 已提交
1191
	}
1192
	if (error)
1193
		goto out_release_inode;
D
Dave Chinner 已提交
1194

1195
	xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
D
Dave Chinner 已提交
1196 1197
	unlock_dp_on_error = true;

1198
	xfs_defer_init(&dfops, &first_block);
1199
	tp->t_agfl_dfops = &dfops;
D
Dave Chinner 已提交
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213

	/*
	 * Reserve disk quota and the inode.
	 */
	error = xfs_trans_reserve_quota(tp, mp, udqp, gdqp,
						pdqp, resblks, 1, 0);
	if (error)
		goto out_trans_cancel;

	/*
	 * A newly created regular or special file just has one directory
	 * entry pointing to them, but a directory also the "." entry
	 * pointing to itself.
	 */
1214
	error = xfs_dir_ialloc(&tp, dp, mode, is_dir ? 2 : 1, rdev, prid, &ip);
1215
	if (error)
1216
		goto out_trans_cancel;
D
Dave Chinner 已提交
1217 1218 1219 1220 1221 1222 1223 1224

	/*
	 * Now we join the directory inode to the transaction.  We do not do it
	 * earlier because xfs_dir_ialloc might commit the previous transaction
	 * (and release all the locks).  An error from here on will result in
	 * the transaction cancel unlocking dp so don't do it explicitly in the
	 * error path.
	 */
1225
	xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
1226 1227 1228
	unlock_dp_on_error = false;

	error = xfs_dir_createname(tp, dp, name, ip->i_ino,
1229
					&first_block, &dfops, resblks ?
D
Dave Chinner 已提交
1230 1231
					resblks - XFS_IALLOC_SPACE_RES(mp) : 0);
	if (error) {
D
Dave Chinner 已提交
1232
		ASSERT(error != -ENOSPC);
1233
		goto out_trans_cancel;
D
Dave Chinner 已提交
1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
	}
	xfs_trans_ichgtime(tp, dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);

	if (is_dir) {
		error = xfs_dir_init(tp, ip, dp);
		if (error)
			goto out_bmap_cancel;

		error = xfs_bumplink(tp, dp);
		if (error)
			goto out_bmap_cancel;
	}

	/*
	 * If this is a synchronous mount, make sure that the
	 * create transaction goes to disk before returning to
	 * the user.
	 */
	if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC))
		xfs_trans_set_sync(tp);

	/*
	 * Attach the dquot(s) to the inodes and modify them incore.
	 * These ids of the inode couldn't have changed since the new
	 * inode has been locked ever since it was created.
	 */
	xfs_qm_vop_create_dqattach(tp, ip, udqp, gdqp, pdqp);

1263
	error = xfs_defer_finish(&tp, &dfops);
D
Dave Chinner 已提交
1264 1265 1266
	if (error)
		goto out_bmap_cancel;

1267
	error = xfs_trans_commit(tp);
D
Dave Chinner 已提交
1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
	if (error)
		goto out_release_inode;

	xfs_qm_dqrele(udqp);
	xfs_qm_dqrele(gdqp);
	xfs_qm_dqrele(pdqp);

	*ipp = ip;
	return 0;

 out_bmap_cancel:
1279
	xfs_defer_cancel(&dfops);
D
Dave Chinner 已提交
1280
 out_trans_cancel:
1281
	xfs_trans_cancel(tp);
D
Dave Chinner 已提交
1282 1283
 out_release_inode:
	/*
1284 1285 1286
	 * Wait until after the current transaction is aborted to finish the
	 * setup of the inode and release the inode.  This prevents recursive
	 * transactions and deadlocks from xfs_inactive.
D
Dave Chinner 已提交
1287
	 */
1288 1289
	if (ip) {
		xfs_finish_inode_setup(ip);
D
Dave Chinner 已提交
1290
		IRELE(ip);
1291
	}
D
Dave Chinner 已提交
1292 1293 1294 1295 1296 1297

	xfs_qm_dqrele(udqp);
	xfs_qm_dqrele(gdqp);
	xfs_qm_dqrele(pdqp);

	if (unlock_dp_on_error)
1298
		xfs_iunlock(dp, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
1299 1300 1301
	return error;
}

Z
Zhi Yong Wu 已提交
1302 1303 1304
int
xfs_create_tmpfile(
	struct xfs_inode	*dp,
1305 1306
	umode_t			mode,
	struct xfs_inode	**ipp)
Z
Zhi Yong Wu 已提交
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319
{
	struct xfs_mount	*mp = dp->i_mount;
	struct xfs_inode	*ip = NULL;
	struct xfs_trans	*tp = NULL;
	int			error;
	prid_t                  prid;
	struct xfs_dquot	*udqp = NULL;
	struct xfs_dquot	*gdqp = NULL;
	struct xfs_dquot	*pdqp = NULL;
	struct xfs_trans_res	*tres;
	uint			resblks;

	if (XFS_FORCED_SHUTDOWN(mp))
D
Dave Chinner 已提交
1320
		return -EIO;
Z
Zhi Yong Wu 已提交
1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335

	prid = xfs_get_initial_prid(dp);

	/*
	 * Make sure that we have allocated dquot(s) on disk.
	 */
	error = xfs_qm_vop_dqalloc(dp, xfs_kuid_to_uid(current_fsuid()),
				xfs_kgid_to_gid(current_fsgid()), prid,
				XFS_QMOPT_QUOTALL | XFS_QMOPT_INHERIT,
				&udqp, &gdqp, &pdqp);
	if (error)
		return error;

	resblks = XFS_IALLOC_SPACE_RES(mp);
	tres = &M_RES(mp)->tr_create_tmpfile;
1336 1337

	error = xfs_trans_alloc(mp, tres, resblks, 0, 0, &tp);
1338
	if (error)
1339
		goto out_release_inode;
Z
Zhi Yong Wu 已提交
1340 1341 1342 1343 1344 1345

	error = xfs_trans_reserve_quota(tp, mp, udqp, gdqp,
						pdqp, resblks, 1, 0);
	if (error)
		goto out_trans_cancel;

1346
	error = xfs_dir_ialloc(&tp, dp, mode, 1, 0, prid, &ip);
1347
	if (error)
1348
		goto out_trans_cancel;
Z
Zhi Yong Wu 已提交
1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361

	if (mp->m_flags & XFS_MOUNT_WSYNC)
		xfs_trans_set_sync(tp);

	/*
	 * Attach the dquot(s) to the inodes and modify them incore.
	 * These ids of the inode couldn't have changed since the new
	 * inode has been locked ever since it was created.
	 */
	xfs_qm_vop_create_dqattach(tp, ip, udqp, gdqp, pdqp);

	error = xfs_iunlink(tp, ip);
	if (error)
1362
		goto out_trans_cancel;
Z
Zhi Yong Wu 已提交
1363

1364
	error = xfs_trans_commit(tp);
Z
Zhi Yong Wu 已提交
1365 1366 1367 1368 1369 1370 1371
	if (error)
		goto out_release_inode;

	xfs_qm_dqrele(udqp);
	xfs_qm_dqrele(gdqp);
	xfs_qm_dqrele(pdqp);

1372
	*ipp = ip;
Z
Zhi Yong Wu 已提交
1373 1374 1375
	return 0;

 out_trans_cancel:
1376
	xfs_trans_cancel(tp);
Z
Zhi Yong Wu 已提交
1377 1378
 out_release_inode:
	/*
1379 1380 1381
	 * Wait until after the current transaction is aborted to finish the
	 * setup of the inode and release the inode.  This prevents recursive
	 * transactions and deadlocks from xfs_inactive.
Z
Zhi Yong Wu 已提交
1382
	 */
1383 1384
	if (ip) {
		xfs_finish_inode_setup(ip);
Z
Zhi Yong Wu 已提交
1385
		IRELE(ip);
1386
	}
Z
Zhi Yong Wu 已提交
1387 1388 1389 1390 1391 1392 1393 1394

	xfs_qm_dqrele(udqp);
	xfs_qm_dqrele(gdqp);
	xfs_qm_dqrele(pdqp);

	return error;
}

D
Dave Chinner 已提交
1395 1396 1397 1398 1399 1400 1401 1402 1403
int
xfs_link(
	xfs_inode_t		*tdp,
	xfs_inode_t		*sip,
	struct xfs_name		*target_name)
{
	xfs_mount_t		*mp = tdp->i_mount;
	xfs_trans_t		*tp;
	int			error;
1404
	struct xfs_defer_ops	dfops;
D
Dave Chinner 已提交
1405 1406 1407 1408 1409
	xfs_fsblock_t           first_block;
	int			resblks;

	trace_xfs_link(tdp, target_name);

D
Dave Chinner 已提交
1410
	ASSERT(!S_ISDIR(VFS_I(sip)->i_mode));
D
Dave Chinner 已提交
1411 1412

	if (XFS_FORCED_SHUTDOWN(mp))
D
Dave Chinner 已提交
1413
		return -EIO;
D
Dave Chinner 已提交
1414

1415
	error = xfs_qm_dqattach(sip);
D
Dave Chinner 已提交
1416 1417 1418
	if (error)
		goto std_return;

1419
	error = xfs_qm_dqattach(tdp);
D
Dave Chinner 已提交
1420 1421 1422 1423
	if (error)
		goto std_return;

	resblks = XFS_LINK_SPACE_RES(mp, target_name->len);
1424
	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_link, resblks, 0, 0, &tp);
D
Dave Chinner 已提交
1425
	if (error == -ENOSPC) {
D
Dave Chinner 已提交
1426
		resblks = 0;
1427
		error = xfs_trans_alloc(mp, &M_RES(mp)->tr_link, 0, 0, 0, &tp);
D
Dave Chinner 已提交
1428
	}
1429
	if (error)
1430
		goto std_return;
D
Dave Chinner 已提交
1431

1432
	xfs_lock_two_inodes(sip, XFS_ILOCK_EXCL, tdp, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
1433 1434

	xfs_trans_ijoin(tp, sip, XFS_ILOCK_EXCL);
1435
	xfs_trans_ijoin(tp, tdp, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
1436 1437 1438 1439 1440 1441 1442 1443

	/*
	 * If we are using project inheritance, we only allow hard link
	 * creation in our tree when the project IDs are the same; else
	 * the tree quota mechanism could be circumvented.
	 */
	if (unlikely((tdp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
		     (xfs_get_projid(tdp) != xfs_get_projid(sip)))) {
D
Dave Chinner 已提交
1444
		error = -EXDEV;
D
Dave Chinner 已提交
1445 1446 1447
		goto error_return;
	}

1448 1449 1450 1451 1452
	if (!resblks) {
		error = xfs_dir_canenter(tp, tdp, target_name);
		if (error)
			goto error_return;
	}
D
Dave Chinner 已提交
1453

1454
	xfs_defer_init(&dfops, &first_block);
1455
	tp->t_agfl_dfops = &dfops;
D
Dave Chinner 已提交
1456

1457 1458 1459 1460
	/*
	 * Handle initial link state of O_TMPFILE inode
	 */
	if (VFS_I(sip)->i_nlink == 0) {
1461 1462
		error = xfs_iunlink_remove(tp, sip);
		if (error)
1463
			goto error_return;
1464 1465
	}

D
Dave Chinner 已提交
1466
	error = xfs_dir_createname(tp, tdp, target_name, sip->i_ino,
1467
					&first_block, &dfops, resblks);
D
Dave Chinner 已提交
1468
	if (error)
1469
		goto error_return;
D
Dave Chinner 已提交
1470 1471 1472 1473 1474
	xfs_trans_ichgtime(tp, tdp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	xfs_trans_log_inode(tp, tdp, XFS_ILOG_CORE);

	error = xfs_bumplink(tp, sip);
	if (error)
1475
		goto error_return;
D
Dave Chinner 已提交
1476 1477 1478 1479 1480 1481

	/*
	 * If this is a synchronous mount, make sure that the
	 * link transaction goes to disk before returning to
	 * the user.
	 */
1482
	if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC))
D
Dave Chinner 已提交
1483 1484
		xfs_trans_set_sync(tp);

1485
	error = xfs_defer_finish(&tp, &dfops);
D
Dave Chinner 已提交
1486
	if (error) {
1487
		xfs_defer_cancel(&dfops);
1488
		goto error_return;
D
Dave Chinner 已提交
1489 1490
	}

1491
	return xfs_trans_commit(tp);
D
Dave Chinner 已提交
1492 1493

 error_return:
1494
	xfs_trans_cancel(tp);
D
Dave Chinner 已提交
1495 1496 1497 1498
 std_return:
	return error;
}

1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
/* Clear the reflink flag and the cowblocks tag if possible. */
static void
xfs_itruncate_clear_reflink_flags(
	struct xfs_inode	*ip)
{
	struct xfs_ifork	*dfork;
	struct xfs_ifork	*cfork;

	if (!xfs_is_reflink_inode(ip))
		return;
	dfork = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
	cfork = XFS_IFORK_PTR(ip, XFS_COW_FORK);
	if (dfork->if_bytes == 0 && cfork->if_bytes == 0)
		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
	if (cfork->if_bytes == 0)
		xfs_inode_clear_cowblocks_tag(ip);
}

L
Linus Torvalds 已提交
1517
/*
1518 1519 1520
 * 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
1521
 *
1522 1523 1524 1525 1526 1527 1528 1529 1530
 * 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
1531
 *
1532 1533 1534 1535 1536
 * 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
1537 1538
 */
int
B
Brian Foster 已提交
1539
xfs_itruncate_extents_flags(
1540 1541 1542
	struct xfs_trans	**tpp,
	struct xfs_inode	*ip,
	int			whichfork,
1543
	xfs_fsize_t		new_size,
B
Brian Foster 已提交
1544
	int			flags)
L
Linus Torvalds 已提交
1545
{
1546 1547
	struct xfs_mount	*mp = ip->i_mount;
	struct xfs_trans	*tp = *tpp;
1548
	struct xfs_defer_ops	dfops;
1549 1550 1551 1552 1553 1554
	xfs_fsblock_t		first_block;
	xfs_fileoff_t		first_unmap_block;
	xfs_fileoff_t		last_block;
	xfs_filblks_t		unmap_len;
	int			error = 0;
	int			done = 0;
L
Linus Torvalds 已提交
1555

1556 1557 1558
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
	ASSERT(!atomic_read(&VFS_I(ip)->i_count) ||
	       xfs_isilocked(ip, XFS_IOLOCK_EXCL));
1559
	ASSERT(new_size <= XFS_ISIZE(ip));
1560
	ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
L
Linus Torvalds 已提交
1561
	ASSERT(ip->i_itemp != NULL);
1562
	ASSERT(ip->i_itemp->ili_lock_flags == 0);
1563
	ASSERT(!XFS_NOT_DQATTACHED(mp, ip));
L
Linus Torvalds 已提交
1564

1565 1566
	trace_xfs_itruncate_extents_start(ip, new_size);

B
Brian Foster 已提交
1567
	flags |= xfs_bmapi_aflag(whichfork);
1568

L
Linus Torvalds 已提交
1569 1570 1571 1572 1573 1574 1575 1576 1577
	/*
	 * 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.
	 */
1578
	first_unmap_block = XFS_B_TO_FSB(mp, (xfs_ufsize_t)new_size);
1579
	last_block = XFS_B_TO_FSB(mp, mp->m_super->s_maxbytes);
1580 1581 1582 1583 1584
	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
1585
	while (!done) {
1586
		xfs_defer_init(&dfops, &first_block);
1587 1588 1589
		error = xfs_bunmapi(tp, ip, first_unmap_block, unmap_len, flags,
				    XFS_ITRUNC_MAX_EXTENTS, &first_block,
				    &dfops, &done);
1590 1591
		if (error)
			goto out_bmap_cancel;
L
Linus Torvalds 已提交
1592 1593 1594 1595 1596

		/*
		 * Duplicate the transaction that has the permanent
		 * reservation and commit the old transaction.
		 */
1597 1598
		xfs_defer_ijoin(&dfops, ip);
		error = xfs_defer_finish(&tp, &dfops);
1599 1600
		if (error)
			goto out_bmap_cancel;
L
Linus Torvalds 已提交
1601

C
Christoph Hellwig 已提交
1602
		error = xfs_trans_roll_inode(&tp, ip);
1603
		if (error)
1604
			goto out;
L
Linus Torvalds 已提交
1605
	}
1606

1607 1608 1609 1610 1611 1612
	if (whichfork == XFS_DATA_FORK) {
		/* Remove all pending CoW reservations. */
		error = xfs_reflink_cancel_cow_blocks(ip, &tp,
				first_unmap_block, last_block, true);
		if (error)
			goto out;
1613

1614 1615
		xfs_itruncate_clear_reflink_flags(ip);
	}
1616

1617 1618 1619 1620 1621 1622 1623 1624
	/*
	 * Always re-log the inode so that our permanent transaction can keep
	 * on rolling it forward in the log.
	 */
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

	trace_xfs_itruncate_extents_end(ip, new_size);

1625 1626 1627 1628
out:
	*tpp = tp;
	return error;
out_bmap_cancel:
L
Linus Torvalds 已提交
1629
	/*
1630 1631 1632
	 * 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
1633
	 */
1634
	xfs_defer_cancel(&dfops);
1635 1636 1637
	goto out;
}

D
Dave Chinner 已提交
1638 1639 1640 1641 1642 1643 1644
int
xfs_release(
	xfs_inode_t	*ip)
{
	xfs_mount_t	*mp = ip->i_mount;
	int		error;

D
Dave Chinner 已提交
1645
	if (!S_ISREG(VFS_I(ip)->i_mode) || (VFS_I(ip)->i_mode == 0))
D
Dave Chinner 已提交
1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
		return 0;

	/* If this is a read-only mount, don't do this (would generate I/O) */
	if (mp->m_flags & XFS_MOUNT_RDONLY)
		return 0;

	if (!XFS_FORCED_SHUTDOWN(mp)) {
		int truncated;

		/*
		 * If we previously truncated this file and removed old data
		 * in the process, we want to initiate "early" writeout on
		 * the last close.  This is an attempt to combat the notorious
		 * NULL files problem which is particularly noticeable from a
		 * truncate down, buffered (re-)write (delalloc), followed by
		 * a crash.  What we are effectively doing here is
		 * significantly reducing the time window where we'd otherwise
		 * be exposed to that problem.
		 */
		truncated = xfs_iflags_test_and_clear(ip, XFS_ITRUNCATED);
		if (truncated) {
			xfs_iflags_clear(ip, XFS_IDIRTY_RELEASE);
D
Dave Chinner 已提交
1668
			if (ip->i_delayed_blks > 0) {
D
Dave Chinner 已提交
1669
				error = filemap_flush(VFS_I(ip)->i_mapping);
D
Dave Chinner 已提交
1670 1671 1672 1673 1674 1675
				if (error)
					return error;
			}
		}
	}

1676
	if (VFS_I(ip)->i_nlink == 0)
D
Dave Chinner 已提交
1677 1678 1679 1680
		return 0;

	if (xfs_can_free_eofblocks(ip, false)) {

1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696
		/*
		 * Check if the inode is being opened, written and closed
		 * frequently and we have delayed allocation blocks outstanding
		 * (e.g. streaming writes from the NFS server), truncating the
		 * blocks past EOF will cause fragmentation to occur.
		 *
		 * In this case don't do the truncation, but we have to be
		 * careful how we detect this case. Blocks beyond EOF show up as
		 * i_delayed_blks even when the inode is clean, so we need to
		 * truncate them away first before checking for a dirty release.
		 * Hence on the first dirty close we will still remove the
		 * speculative allocation, but after that we will leave it in
		 * place.
		 */
		if (xfs_iflags_test(ip, XFS_IDIRTY_RELEASE))
			return 0;
D
Dave Chinner 已提交
1697 1698 1699
		/*
		 * If we can't get the iolock just skip truncating the blocks
		 * past EOF because we could deadlock with the mmap_sem
1700
		 * otherwise. We'll get another chance to drop them once the
D
Dave Chinner 已提交
1701 1702 1703
		 * last reference to the inode is dropped, so we'll never leak
		 * blocks permanently.
		 */
1704 1705 1706 1707 1708 1709
		if (xfs_ilock_nowait(ip, XFS_IOLOCK_EXCL)) {
			error = xfs_free_eofblocks(ip);
			xfs_iunlock(ip, XFS_IOLOCK_EXCL);
			if (error)
				return error;
		}
D
Dave Chinner 已提交
1710 1711 1712 1713 1714 1715 1716 1717

		/* delalloc blocks after truncation means it really is dirty */
		if (ip->i_delayed_blks)
			xfs_iflags_set(ip, XFS_IDIRTY_RELEASE);
	}
	return 0;
}

1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
/*
 * xfs_inactive_truncate
 *
 * Called to perform a truncate when an inode becomes unlinked.
 */
STATIC int
xfs_inactive_truncate(
	struct xfs_inode *ip)
{
	struct xfs_mount	*mp = ip->i_mount;
	struct xfs_trans	*tp;
	int			error;

1731
	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp);
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742
	if (error) {
		ASSERT(XFS_FORCED_SHUTDOWN(mp));
		return error;
	}

	xfs_ilock(ip, XFS_ILOCK_EXCL);
	xfs_trans_ijoin(tp, ip, 0);

	/*
	 * Log the inode size first to prevent stale data exposure in the event
	 * of a system crash before the truncate completes. See the related
1743
	 * comment in xfs_vn_setattr_size() for details.
1744 1745 1746 1747 1748 1749 1750 1751 1752 1753
	 */
	ip->i_d.di_size = 0;
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

	error = xfs_itruncate_extents(&tp, ip, XFS_DATA_FORK, 0);
	if (error)
		goto error_trans_cancel;

	ASSERT(ip->i_d.di_nextents == 0);

1754
	error = xfs_trans_commit(tp);
1755 1756 1757 1758 1759 1760 1761
	if (error)
		goto error_unlock;

	xfs_iunlock(ip, XFS_ILOCK_EXCL);
	return 0;

error_trans_cancel:
1762
	xfs_trans_cancel(tp);
1763 1764 1765 1766 1767
error_unlock:
	xfs_iunlock(ip, XFS_ILOCK_EXCL);
	return error;
}

1768 1769 1770 1771 1772 1773 1774 1775 1776
/*
 * xfs_inactive_ifree()
 *
 * Perform the inode free when an inode is unlinked.
 */
STATIC int
xfs_inactive_ifree(
	struct xfs_inode *ip)
{
1777
	struct xfs_defer_ops	dfops;
1778 1779 1780 1781 1782
	xfs_fsblock_t		first_block;
	struct xfs_mount	*mp = ip->i_mount;
	struct xfs_trans	*tp;
	int			error;

1783
	/*
1784 1785 1786 1787 1788
	 * We try to use a per-AG reservation for any block needed by the finobt
	 * tree, but as the finobt feature predates the per-AG reservation
	 * support a degraded file system might not have enough space for the
	 * reservation at mount time.  In that case try to dip into the reserved
	 * pool and pray.
1789 1790 1791 1792 1793
	 *
	 * Send a warning if the reservation does happen to fail, as the inode
	 * now remains allocated and sits on the unlinked list until the fs is
	 * repaired.
	 */
1794 1795 1796 1797 1798 1799 1800
	if (unlikely(mp->m_inotbt_nores)) {
		error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ifree,
				XFS_IFREE_SPACE_RES(mp), 0, XFS_TRANS_RESERVE,
				&tp);
	} else {
		error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ifree, 0, 0, 0, &tp);
	}
1801
	if (error) {
D
Dave Chinner 已提交
1802
		if (error == -ENOSPC) {
1803 1804 1805 1806 1807 1808
			xfs_warn_ratelimited(mp,
			"Failed to remove inode(s) from unlinked list. "
			"Please free space, unmount and run xfs_repair.");
		} else {
			ASSERT(XFS_FORCED_SHUTDOWN(mp));
		}
1809 1810 1811 1812 1813 1814
		return error;
	}

	xfs_ilock(ip, XFS_ILOCK_EXCL);
	xfs_trans_ijoin(tp, ip, 0);

1815
	xfs_defer_init(&dfops, &first_block);
1816
	tp->t_agfl_dfops = &dfops;
1817
	error = xfs_ifree(tp, ip, &dfops);
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828
	if (error) {
		/*
		 * If we fail to free the inode, shut down.  The cancel
		 * might do that, we need to make sure.  Otherwise the
		 * inode might be lost for a long time or forever.
		 */
		if (!XFS_FORCED_SHUTDOWN(mp)) {
			xfs_notice(mp, "%s: xfs_ifree returned error %d",
				__func__, error);
			xfs_force_shutdown(mp, SHUTDOWN_META_IO_ERROR);
		}
1829
		xfs_trans_cancel(tp);
1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
		xfs_iunlock(ip, XFS_ILOCK_EXCL);
		return error;
	}

	/*
	 * Credit the quota account(s). The inode is gone.
	 */
	xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_ICOUNT, -1);

	/*
1840 1841
	 * Just ignore errors at this point.  There is nothing we can do except
	 * to try to keep going. Make sure it's not a silent error.
1842
	 */
1843
	error = xfs_defer_finish(&tp, &dfops);
1844
	if (error) {
1845
		xfs_notice(mp, "%s: xfs_defer_finish returned error %d",
1846
			__func__, error);
1847
		xfs_defer_cancel(&dfops);
1848
	}
1849
	error = xfs_trans_commit(tp);
1850 1851 1852 1853 1854 1855 1856 1857
	if (error)
		xfs_notice(mp, "%s: xfs_trans_commit returned error %d",
			__func__, error);

	xfs_iunlock(ip, XFS_ILOCK_EXCL);
	return 0;
}

D
Dave Chinner 已提交
1858 1859 1860 1861 1862 1863 1864 1865
/*
 * xfs_inactive
 *
 * This is called when the vnode reference count for the vnode
 * goes to zero.  If the file has been unlinked, then it must
 * now be truncated.  Also, we clear all of the read-ahead state
 * kept for the inode here since the file is now closed.
 */
1866
void
D
Dave Chinner 已提交
1867 1868 1869
xfs_inactive(
	xfs_inode_t	*ip)
{
1870
	struct xfs_mount	*mp;
1871
	struct xfs_ifork	*cow_ifp = XFS_IFORK_PTR(ip, XFS_COW_FORK);
1872 1873
	int			error;
	int			truncate = 0;
D
Dave Chinner 已提交
1874 1875 1876 1877 1878

	/*
	 * If the inode is already free, then there can be nothing
	 * to clean up here.
	 */
D
Dave Chinner 已提交
1879
	if (VFS_I(ip)->i_mode == 0) {
D
Dave Chinner 已提交
1880 1881
		ASSERT(ip->i_df.if_real_bytes == 0);
		ASSERT(ip->i_df.if_broot_bytes == 0);
1882
		return;
D
Dave Chinner 已提交
1883 1884 1885
	}

	mp = ip->i_mount;
1886
	ASSERT(!xfs_iflags_test(ip, XFS_IRECOVERY));
D
Dave Chinner 已提交
1887 1888 1889

	/* If this is a read-only mount, don't do this (would generate I/O) */
	if (mp->m_flags & XFS_MOUNT_RDONLY)
1890
		return;
D
Dave Chinner 已提交
1891

1892 1893 1894 1895
	/* Try to clean out the cow blocks if there are any. */
	if (xfs_is_reflink_inode(ip) && cow_ifp->if_bytes > 0)
		xfs_reflink_cancel_cow_range(ip, 0, NULLFILEOFF, true);

1896
	if (VFS_I(ip)->i_nlink != 0) {
D
Dave Chinner 已提交
1897 1898 1899 1900
		/*
		 * force is true because we are evicting an inode from the
		 * cache. Post-eof blocks must be freed, lest we end up with
		 * broken free space accounting.
1901 1902 1903 1904
		 *
		 * Note: don't bother with iolock here since lockdep complains
		 * about acquiring it in reclaim context. We have the only
		 * reference to the inode at this point anyways.
D
Dave Chinner 已提交
1905
		 */
1906
		if (xfs_can_free_eofblocks(ip, true))
1907
			xfs_free_eofblocks(ip);
1908 1909

		return;
D
Dave Chinner 已提交
1910 1911
	}

D
Dave Chinner 已提交
1912
	if (S_ISREG(VFS_I(ip)->i_mode) &&
D
Dave Chinner 已提交
1913 1914 1915 1916
	    (ip->i_d.di_size != 0 || XFS_ISIZE(ip) != 0 ||
	     ip->i_d.di_nextents > 0 || ip->i_delayed_blks > 0))
		truncate = 1;

1917
	error = xfs_qm_dqattach(ip);
D
Dave Chinner 已提交
1918
	if (error)
1919
		return;
D
Dave Chinner 已提交
1920

D
Dave Chinner 已提交
1921
	if (S_ISLNK(VFS_I(ip)->i_mode))
1922
		error = xfs_inactive_symlink(ip);
1923 1924 1925
	else if (truncate)
		error = xfs_inactive_truncate(ip);
	if (error)
1926
		return;
D
Dave Chinner 已提交
1927 1928 1929 1930

	/*
	 * If there are attributes associated with the file then blow them away
	 * now.  The code calls a routine that recursively deconstructs the
1931
	 * attribute fork. If also blows away the in-core attribute fork.
D
Dave Chinner 已提交
1932
	 */
1933
	if (XFS_IFORK_Q(ip)) {
D
Dave Chinner 已提交
1934 1935
		error = xfs_attr_inactive(ip);
		if (error)
1936
			return;
D
Dave Chinner 已提交
1937 1938
	}

1939
	ASSERT(!ip->i_afp);
D
Dave Chinner 已提交
1940
	ASSERT(ip->i_d.di_anextents == 0);
1941
	ASSERT(ip->i_d.di_forkoff == 0);
D
Dave Chinner 已提交
1942 1943 1944 1945

	/*
	 * Free the inode.
	 */
1946 1947
	error = xfs_inactive_ifree(ip);
	if (error)
1948
		return;
D
Dave Chinner 已提交
1949 1950 1951 1952 1953 1954 1955

	/*
	 * Release the dquots held by inode, if any.
	 */
	xfs_qm_dqdetach(ip);
}

L
Linus Torvalds 已提交
1956
/*
1957 1958 1959 1960 1961 1962 1963 1964
 * This is called when the inode's link count goes to 0 or we are creating a
 * tmpfile via O_TMPFILE. In the case of a tmpfile, @ignore_linkcount will be
 * set to true as the link count is dropped to zero by the VFS after we've
 * created the file successfully, so we have to add it to the unlinked list
 * while the link count is non-zero.
 *
 * We place the on-disk inode on a list in the AGI.  It will be pulled from this
 * list when the inode is freed.
L
Linus Torvalds 已提交
1965
 */
1966
STATIC int
L
Linus Torvalds 已提交
1967
xfs_iunlink(
1968 1969
	struct xfs_trans *tp,
	struct xfs_inode *ip)
L
Linus Torvalds 已提交
1970
{
1971
	xfs_mount_t	*mp = tp->t_mountp;
L
Linus Torvalds 已提交
1972 1973 1974 1975 1976 1977 1978 1979 1980
	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;

D
Dave Chinner 已提交
1981
	ASSERT(VFS_I(ip)->i_mode != 0);
L
Linus Torvalds 已提交
1982 1983 1984 1985 1986

	/*
	 * Get the agi buffer first.  It ensures lock ordering
	 * on the list.
	 */
1987
	error = xfs_read_agi(mp, tp, XFS_INO_TO_AGNO(mp, ip->i_ino), &agibp);
1988
	if (error)
L
Linus Torvalds 已提交
1989 1990
		return error;
	agi = XFS_BUF_TO_AGI(agibp);
1991

L
Linus Torvalds 已提交
1992 1993 1994 1995 1996 1997 1998 1999
	/*
	 * 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]);
2000
	ASSERT(be32_to_cpu(agi->agi_unlinked[bucket_index]) != agino);
L
Linus Torvalds 已提交
2001

2002
	if (agi->agi_unlinked[bucket_index] != cpu_to_be32(NULLAGINO)) {
L
Linus Torvalds 已提交
2003 2004 2005 2006 2007 2008
		/*
		 * 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.
		 */
2009 2010
		error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &ibp,
				       0, 0);
2011 2012 2013
		if (error)
			return error;

2014
		ASSERT(dip->di_next_unlinked == cpu_to_be32(NULLAGINO));
L
Linus Torvalds 已提交
2015
		dip->di_next_unlinked = agi->agi_unlinked[bucket_index];
2016
		offset = ip->i_imap.im_boffset +
L
Linus Torvalds 已提交
2017
			offsetof(xfs_dinode_t, di_next_unlinked);
2018 2019 2020 2021

		/* need to recalc the inode CRC if appropriate */
		xfs_dinode_calc_crc(mp, dip);

L
Linus Torvalds 已提交
2022 2023 2024 2025 2026 2027 2028 2029 2030 2031
		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);
2032
	agi->agi_unlinked[bucket_index] = cpu_to_be32(agino);
L
Linus Torvalds 已提交
2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057
	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;
2058
	xfs_dinode_t	*last_dip = NULL;
L
Linus Torvalds 已提交
2059
	short		bucket_index;
2060
	int		offset, last_offset = 0;
L
Linus Torvalds 已提交
2061 2062 2063 2064 2065 2066 2067 2068 2069
	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.
	 */
2070 2071
	error = xfs_read_agi(mp, tp, agno, &agibp);
	if (error)
L
Linus Torvalds 已提交
2072
		return error;
2073

L
Linus Torvalds 已提交
2074
	agi = XFS_BUF_TO_AGI(agibp);
2075

L
Linus Torvalds 已提交
2076 2077 2078 2079 2080
	/*
	 * 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);
2081 2082
	if (!xfs_verify_agino(mp, agno, agino))
		return -EFSCORRUPTED;
L
Linus Torvalds 已提交
2083
	bucket_index = agino % XFS_AGI_UNLINKED_BUCKETS;
2084 2085 2086 2087 2088 2089
	if (!xfs_verify_agino(mp, agno,
			be32_to_cpu(agi->agi_unlinked[bucket_index]))) {
		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
				agi, sizeof(*agi));
		return -EFSCORRUPTED;
	}
L
Linus Torvalds 已提交
2090

2091
	if (be32_to_cpu(agi->agi_unlinked[bucket_index]) == agino) {
L
Linus Torvalds 已提交
2092
		/*
2093 2094 2095 2096 2097
		 * 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.
L
Linus Torvalds 已提交
2098
		 */
2099 2100
		error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &ibp,
				       0, 0);
L
Linus Torvalds 已提交
2101
		if (error) {
2102
			xfs_warn(mp, "%s: xfs_imap_to_bp returned error %d.",
2103
				__func__, error);
L
Linus Torvalds 已提交
2104 2105
			return error;
		}
2106
		next_agino = be32_to_cpu(dip->di_next_unlinked);
L
Linus Torvalds 已提交
2107 2108
		ASSERT(next_agino != 0);
		if (next_agino != NULLAGINO) {
2109
			dip->di_next_unlinked = cpu_to_be32(NULLAGINO);
2110
			offset = ip->i_imap.im_boffset +
L
Linus Torvalds 已提交
2111
				offsetof(xfs_dinode_t, di_next_unlinked);
2112 2113 2114 2115

			/* need to recalc the inode CRC if appropriate */
			xfs_dinode_calc_crc(mp, dip);

L
Linus Torvalds 已提交
2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127
			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);
2128
		agi->agi_unlinked[bucket_index] = cpu_to_be32(next_agino);
L
Linus Torvalds 已提交
2129 2130 2131 2132 2133 2134 2135 2136
		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.
		 */
2137
		next_agino = be32_to_cpu(agi->agi_unlinked[bucket_index]);
L
Linus Torvalds 已提交
2138 2139
		last_ibp = NULL;
		while (next_agino != agino) {
C
Christoph Hellwig 已提交
2140 2141 2142
			struct xfs_imap	imap;

			if (last_ibp)
L
Linus Torvalds 已提交
2143
				xfs_trans_brelse(tp, last_ibp);
C
Christoph Hellwig 已提交
2144 2145

			imap.im_blkno = 0;
L
Linus Torvalds 已提交
2146
			next_ino = XFS_AGINO_TO_INO(mp, agno, next_agino);
C
Christoph Hellwig 已提交
2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157

			error = xfs_imap(mp, tp, next_ino, &imap, 0);
			if (error) {
				xfs_warn(mp,
	"%s: xfs_imap returned error %d.",
					 __func__, error);
				return error;
			}

			error = xfs_imap_to_bp(mp, tp, &imap, &last_dip,
					       &last_ibp, 0, 0);
L
Linus Torvalds 已提交
2158
			if (error) {
2159
				xfs_warn(mp,
C
Christoph Hellwig 已提交
2160
	"%s: xfs_imap_to_bp returned error %d.",
2161
					__func__, error);
L
Linus Torvalds 已提交
2162 2163
				return error;
			}
C
Christoph Hellwig 已提交
2164 2165

			last_offset = imap.im_boffset;
2166
			next_agino = be32_to_cpu(last_dip->di_next_unlinked);
2167 2168 2169 2170 2171 2172
			if (!xfs_verify_agino(mp, agno, next_agino)) {
				XFS_CORRUPTION_ERROR(__func__,
						XFS_ERRLEVEL_LOW, mp,
						last_dip, sizeof(*last_dip));
				return -EFSCORRUPTED;
			}
L
Linus Torvalds 已提交
2173
		}
2174

L
Linus Torvalds 已提交
2175
		/*
2176 2177
		 * Now last_ibp points to the buffer previous to us on the
		 * unlinked list.  Pull us from the list.
L
Linus Torvalds 已提交
2178
		 */
2179 2180
		error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &ibp,
				       0, 0);
L
Linus Torvalds 已提交
2181
		if (error) {
2182
			xfs_warn(mp, "%s: xfs_imap_to_bp(2) returned error %d.",
2183
				__func__, error);
L
Linus Torvalds 已提交
2184 2185
			return error;
		}
2186
		next_agino = be32_to_cpu(dip->di_next_unlinked);
L
Linus Torvalds 已提交
2187 2188 2189
		ASSERT(next_agino != 0);
		ASSERT(next_agino != agino);
		if (next_agino != NULLAGINO) {
2190
			dip->di_next_unlinked = cpu_to_be32(NULLAGINO);
2191
			offset = ip->i_imap.im_boffset +
L
Linus Torvalds 已提交
2192
				offsetof(xfs_dinode_t, di_next_unlinked);
2193 2194 2195 2196

			/* need to recalc the inode CRC if appropriate */
			xfs_dinode_calc_crc(mp, dip);

L
Linus Torvalds 已提交
2197 2198 2199 2200 2201 2202 2203 2204 2205 2206
			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.
		 */
2207
		last_dip->di_next_unlinked = cpu_to_be32(next_agino);
L
Linus Torvalds 已提交
2208 2209
		ASSERT(next_agino != 0);
		offset = last_offset + offsetof(xfs_dinode_t, di_next_unlinked);
2210 2211 2212 2213

		/* need to recalc the inode CRC if appropriate */
		xfs_dinode_calc_crc(mp, last_dip);

L
Linus Torvalds 已提交
2214 2215 2216 2217 2218 2219 2220 2221
		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;
}

2222
/*
2223
 * A big issue when freeing the inode cluster is that we _cannot_ skip any
2224 2225 2226
 * inodes that are in memory - they all must be marked stale and attached to
 * the cluster buffer.
 */
2227
STATIC int
L
Linus Torvalds 已提交
2228
xfs_ifree_cluster(
2229 2230 2231
	xfs_inode_t		*free_ip,
	xfs_trans_t		*tp,
	struct xfs_icluster	*xic)
L
Linus Torvalds 已提交
2232 2233 2234
{
	xfs_mount_t		*mp = free_ip->i_mount;
	int			blks_per_cluster;
2235
	int			inodes_per_cluster;
L
Linus Torvalds 已提交
2236
	int			nbufs;
2237
	int			i, j;
2238
	int			ioffset;
L
Linus Torvalds 已提交
2239 2240
	xfs_daddr_t		blkno;
	xfs_buf_t		*bp;
2241
	xfs_inode_t		*ip;
L
Linus Torvalds 已提交
2242
	xfs_inode_log_item_t	*iip;
2243
	struct xfs_log_item	*lip;
D
Dave Chinner 已提交
2244
	struct xfs_perag	*pag;
2245
	xfs_ino_t		inum;
L
Linus Torvalds 已提交
2246

2247
	inum = xic->first_ino;
D
Dave Chinner 已提交
2248
	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, inum));
2249 2250 2251
	blks_per_cluster = xfs_icluster_size_fsb(mp);
	inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog;
	nbufs = mp->m_ialloc_blks / blks_per_cluster;
L
Linus Torvalds 已提交
2252

2253
	for (j = 0; j < nbufs; j++, inum += inodes_per_cluster) {
2254 2255 2256 2257 2258
		/*
		 * The allocation bitmap tells us which inodes of the chunk were
		 * physically allocated. Skip the cluster if an inode falls into
		 * a sparse region.
		 */
2259 2260
		ioffset = inum - xic->first_ino;
		if ((xic->alloc & XFS_INOBT_MASK(ioffset)) == 0) {
2261
			ASSERT(ioffset % inodes_per_cluster == 0);
2262 2263 2264
			continue;
		}

L
Linus Torvalds 已提交
2265 2266 2267
		blkno = XFS_AGB_TO_DADDR(mp, XFS_INO_TO_AGNO(mp, inum),
					 XFS_INO_TO_AGBNO(mp, inum));

2268 2269 2270 2271 2272 2273 2274 2275 2276
		/*
		 * 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,
2277 2278
					mp->m_bsize * blks_per_cluster,
					XBF_UNMAPPED);
2279

2280
		if (!bp)
D
Dave Chinner 已提交
2281
			return -ENOMEM;
2282 2283 2284 2285 2286 2287 2288 2289 2290 2291

		/*
		 * This buffer may not have been correctly initialised as we
		 * didn't read it from disk. That's not important because we are
		 * only using to mark the buffer as stale in the log, and to
		 * attach stale cached inodes on it. That means it will never be
		 * dispatched for IO. If it is, we want to know about it, and we
		 * want it to fail. We can acheive this by adding a write
		 * verifier to the buffer.
		 */
2292
		 bp->b_ops = &xfs_inode_buf_ops;
2293

2294 2295 2296
		/*
		 * Walk the inodes already attached to the buffer and mark them
		 * stale. These will all have the flush locks held, so an
2297 2298 2299
		 * 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.
2300
		 */
2301
		list_for_each_entry(lip, &bp->b_li_list, li_bio_list) {
2302 2303 2304
			if (lip->li_type == XFS_LI_INODE) {
				iip = (xfs_inode_log_item_t *)lip;
				ASSERT(iip->ili_logged == 1);
2305
				lip->li_cb = xfs_istale_done;
2306 2307 2308 2309 2310 2311
				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);
			}
		}
L
Linus Torvalds 已提交
2312

2313

L
Linus Torvalds 已提交
2314
		/*
2315 2316 2317 2318
		 * 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
2319
		 *
2320 2321 2322
		 * 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
2323
		 */
2324
		for (i = 0; i < inodes_per_cluster; i++) {
2325
retry:
2326
			rcu_read_lock();
2327 2328
			ip = radix_tree_lookup(&pag->pag_ici_root,
					XFS_INO_TO_AGINO(mp, (inum + i)));
L
Linus Torvalds 已提交
2329

2330 2331 2332
			/* Inode not in memory, nothing to do */
			if (!ip) {
				rcu_read_unlock();
L
Linus Torvalds 已提交
2333 2334 2335
				continue;
			}

2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351
			/*
			 * 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);

2352 2353 2354 2355 2356 2357 2358
			/*
			 * 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.
			 */
2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
			if (ip != free_ip) {
				if (!xfs_ilock_nowait(ip, XFS_ILOCK_EXCL)) {
					rcu_read_unlock();
					delay(1);
					goto retry;
				}

				/*
				 * Check the inode number again in case we're
				 * racing with freeing in xfs_reclaim_inode().
				 * See the comments in that function for more
				 * information as to why the initial check is
				 * not sufficient.
				 */
				if (ip->i_ino != inum + i) {
					xfs_iunlock(ip, XFS_ILOCK_EXCL);
2375
					rcu_read_unlock();
2376 2377
					continue;
				}
L
Linus Torvalds 已提交
2378
			}
2379
			rcu_read_unlock();
L
Linus Torvalds 已提交
2380

2381
			xfs_iflock(ip);
2382
			xfs_iflags_set(ip, XFS_ISTALE);
L
Linus Torvalds 已提交
2383

2384 2385 2386 2387
			/*
			 * 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
2388
			iip = ip->i_itemp;
2389
			if (!iip || xfs_inode_clean(ip)) {
2390
				ASSERT(ip != free_ip);
L
Linus Torvalds 已提交
2391 2392 2393 2394 2395
				xfs_ifunlock(ip);
				xfs_iunlock(ip, XFS_ILOCK_EXCL);
				continue;
			}

2396 2397
			iip->ili_last_fields = iip->ili_fields;
			iip->ili_fields = 0;
2398
			iip->ili_fsync_fields = 0;
L
Linus Torvalds 已提交
2399
			iip->ili_logged = 1;
2400 2401
			xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn,
						&iip->ili_item.li_lsn);
L
Linus Torvalds 已提交
2402

2403 2404
			xfs_buf_attach_iodone(bp, xfs_istale_done,
						  &iip->ili_item);
2405 2406

			if (ip != free_ip)
L
Linus Torvalds 已提交
2407 2408 2409
				xfs_iunlock(ip, XFS_ILOCK_EXCL);
		}

2410
		xfs_trans_stale_inode_buf(tp, bp);
L
Linus Torvalds 已提交
2411 2412 2413
		xfs_trans_binval(tp, bp);
	}

D
Dave Chinner 已提交
2414
	xfs_perag_put(pag);
2415
	return 0;
L
Linus Torvalds 已提交
2416 2417
}

2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435
/*
 * Free any local-format buffers sitting around before we reset to
 * extents format.
 */
static inline void
xfs_ifree_local_data(
	struct xfs_inode	*ip,
	int			whichfork)
{
	struct xfs_ifork	*ifp;

	if (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_LOCAL)
		return;

	ifp = XFS_IFORK_PTR(ip, whichfork);
	xfs_idata_realloc(ip, -ifp->if_bytes, whichfork);
}

L
Linus Torvalds 已提交
2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449
/*
 * 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,
2450
	struct xfs_defer_ops	*dfops)
L
Linus Torvalds 已提交
2451 2452
{
	int			error;
2453
	struct xfs_icluster	xic = { 0 };
L
Linus Torvalds 已提交
2454

C
Christoph Hellwig 已提交
2455
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
2456
	ASSERT(VFS_I(ip)->i_nlink == 0);
L
Linus Torvalds 已提交
2457 2458
	ASSERT(ip->i_d.di_nextents == 0);
	ASSERT(ip->i_d.di_anextents == 0);
D
Dave Chinner 已提交
2459
	ASSERT(ip->i_d.di_size == 0 || !S_ISREG(VFS_I(ip)->i_mode));
L
Linus Torvalds 已提交
2460 2461 2462 2463 2464 2465
	ASSERT(ip->i_d.di_nblocks == 0);

	/*
	 * Pull the on-disk inode from the AGI unlinked list.
	 */
	error = xfs_iunlink_remove(tp, ip);
2466
	if (error)
L
Linus Torvalds 已提交
2467 2468
		return error;

2469
	error = xfs_difree(tp, ip->i_ino, dfops, &xic);
2470
	if (error)
L
Linus Torvalds 已提交
2471
		return error;
2472

2473 2474 2475
	xfs_ifree_local_data(ip, XFS_DATA_FORK);
	xfs_ifree_local_data(ip, XFS_ATTR_FORK);

D
Dave Chinner 已提交
2476
	VFS_I(ip)->i_mode = 0;		/* mark incore inode as free */
L
Linus Torvalds 已提交
2477
	ip->i_d.di_flags = 0;
2478
	ip->i_d.di_flags2 = 0;
L
Linus Torvalds 已提交
2479 2480 2481 2482
	ip->i_d.di_dmevmask = 0;
	ip->i_d.di_forkoff = 0;		/* mark the attr fork not in use */
	ip->i_d.di_format = XFS_DINODE_FMT_EXTENTS;
	ip->i_d.di_aformat = XFS_DINODE_FMT_EXTENTS;
2483 2484 2485 2486

	/* Don't attempt to replay owner changes for a deleted inode */
	ip->i_itemp->ili_fields &= ~(XFS_ILOG_AOWNER|XFS_ILOG_DOWNER);

L
Linus Torvalds 已提交
2487 2488 2489 2490
	/*
	 * Bump the generation count so no one will be confused
	 * by reincarnations of this inode.
	 */
2491
	VFS_I(ip)->i_generation++;
L
Linus Torvalds 已提交
2492 2493
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

2494 2495
	if (xic.deleted)
		error = xfs_ifree_cluster(ip, tp, &xic);
L
Linus Torvalds 已提交
2496

2497
	return error;
L
Linus Torvalds 已提交
2498 2499 2500
}

/*
2501 2502 2503
 * 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
2504
 */
2505
static void
2506
xfs_iunpin(
2507
	struct xfs_inode	*ip)
L
Linus Torvalds 已提交
2508
{
C
Christoph Hellwig 已提交
2509
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
L
Linus Torvalds 已提交
2510

2511 2512
	trace_xfs_inode_unpin_nowait(ip, _RET_IP_);

2513
	/* Give the log a push to start the unpinning I/O */
2514
	xfs_log_force_lsn(ip->i_mount, ip->i_itemp->ili_last_lsn, 0, NULL);
2515

2516
}
L
Linus Torvalds 已提交
2517

2518 2519 2520 2521 2522 2523 2524 2525 2526 2527
static void
__xfs_iunpin_wait(
	struct xfs_inode	*ip)
{
	wait_queue_head_t *wq = bit_waitqueue(&ip->i_flags, __XFS_IPINNED_BIT);
	DEFINE_WAIT_BIT(wait, &ip->i_flags, __XFS_IPINNED_BIT);

	xfs_iunpin(ip);

	do {
2528
		prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
2529 2530 2531
		if (xfs_ipincount(ip))
			io_schedule();
	} while (xfs_ipincount(ip));
2532
	finish_wait(wq, &wait.wq_entry);
2533 2534
}

2535
void
2536
xfs_iunpin_wait(
2537
	struct xfs_inode	*ip)
2538
{
2539 2540
	if (xfs_ipincount(ip))
		__xfs_iunpin_wait(ip);
L
Linus Torvalds 已提交
2541 2542
}

2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564
/*
 * Removing an inode from the namespace involves removing the directory entry
 * and dropping the link count on the inode. Removing the directory entry can
 * result in locking an AGF (directory blocks were freed) and removing a link
 * count can result in placing the inode on an unlinked list which results in
 * locking an AGI.
 *
 * The big problem here is that we have an ordering constraint on AGF and AGI
 * locking - inode allocation locks the AGI, then can allocate a new extent for
 * new inodes, locking the AGF after the AGI. Similarly, freeing the inode
 * removes the inode from the unlinked list, requiring that we lock the AGI
 * first, and then freeing the inode can result in an inode chunk being freed
 * and hence freeing disk space requiring that we lock an AGF.
 *
 * Hence the ordering that is imposed by other parts of the code is AGI before
 * AGF. This means we cannot remove the directory entry before we drop the inode
 * reference count and put it on the unlinked list as this results in a lock
 * order of AGF then AGI, and this can deadlock against inode allocation and
 * freeing. Therefore we must drop the link counts before we remove the
 * directory entry.
 *
 * This is still safe from a transactional point of view - it is not until we
2565
 * get to xfs_defer_finish() that we have the possibility of multiple
2566 2567 2568 2569
 * transactions in this operation. Hence as long as we remove the directory
 * entry and drop the link count in the first transaction of the remove
 * operation, there are no transactional constraints on the ordering here.
 */
D
Dave Chinner 已提交
2570 2571 2572 2573 2574 2575 2576 2577
int
xfs_remove(
	xfs_inode_t             *dp,
	struct xfs_name		*name,
	xfs_inode_t		*ip)
{
	xfs_mount_t		*mp = dp->i_mount;
	xfs_trans_t             *tp = NULL;
D
Dave Chinner 已提交
2578
	int			is_dir = S_ISDIR(VFS_I(ip)->i_mode);
D
Dave Chinner 已提交
2579
	int                     error = 0;
2580
	struct xfs_defer_ops	dfops;
D
Dave Chinner 已提交
2581 2582 2583 2584 2585 2586
	xfs_fsblock_t           first_block;
	uint			resblks;

	trace_xfs_remove(dp, name);

	if (XFS_FORCED_SHUTDOWN(mp))
D
Dave Chinner 已提交
2587
		return -EIO;
D
Dave Chinner 已提交
2588

2589
	error = xfs_qm_dqattach(dp);
D
Dave Chinner 已提交
2590 2591 2592
	if (error)
		goto std_return;

2593
	error = xfs_qm_dqattach(ip);
D
Dave Chinner 已提交
2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606
	if (error)
		goto std_return;

	/*
	 * We try to get the real space reservation first,
	 * allowing for directory btree deletion(s) implying
	 * possible bmap insert(s).  If we can't get the space
	 * reservation then we use 0 instead, and avoid the bmap
	 * btree insert(s) in the directory code by, if the bmap
	 * insert tries to happen, instead trimming the LAST
	 * block from the directory.
	 */
	resblks = XFS_REMOVE_SPACE_RES(mp);
2607
	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_remove, resblks, 0, 0, &tp);
D
Dave Chinner 已提交
2608
	if (error == -ENOSPC) {
D
Dave Chinner 已提交
2609
		resblks = 0;
2610 2611
		error = xfs_trans_alloc(mp, &M_RES(mp)->tr_remove, 0, 0, 0,
				&tp);
D
Dave Chinner 已提交
2612 2613
	}
	if (error) {
D
Dave Chinner 已提交
2614
		ASSERT(error != -ENOSPC);
2615
		goto std_return;
D
Dave Chinner 已提交
2616 2617
	}

2618
	xfs_lock_two_inodes(dp, XFS_ILOCK_EXCL, ip, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
2619

2620
	xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
2621 2622 2623 2624 2625 2626
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);

	/*
	 * If we're removing a directory perform some additional validation.
	 */
	if (is_dir) {
2627 2628
		ASSERT(VFS_I(ip)->i_nlink >= 2);
		if (VFS_I(ip)->i_nlink != 2) {
D
Dave Chinner 已提交
2629
			error = -ENOTEMPTY;
D
Dave Chinner 已提交
2630 2631 2632
			goto out_trans_cancel;
		}
		if (!xfs_dir_isempty(ip)) {
D
Dave Chinner 已提交
2633
			error = -ENOTEMPTY;
D
Dave Chinner 已提交
2634 2635 2636
			goto out_trans_cancel;
		}

2637
		/* Drop the link from ip's "..".  */
D
Dave Chinner 已提交
2638 2639
		error = xfs_droplink(tp, dp);
		if (error)
2640
			goto out_trans_cancel;
D
Dave Chinner 已提交
2641

2642
		/* Drop the "." link from ip to self.  */
D
Dave Chinner 已提交
2643 2644
		error = xfs_droplink(tp, ip);
		if (error)
2645
			goto out_trans_cancel;
D
Dave Chinner 已提交
2646 2647 2648 2649 2650 2651 2652 2653
	} else {
		/*
		 * When removing a non-directory we need to log the parent
		 * inode here.  For a directory this is done implicitly
		 * by the xfs_droplink call for the ".." entry.
		 */
		xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
	}
2654
	xfs_trans_ichgtime(tp, dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
D
Dave Chinner 已提交
2655

2656
	/* Drop the link from dp to ip. */
D
Dave Chinner 已提交
2657 2658
	error = xfs_droplink(tp, ip);
	if (error)
2659
		goto out_trans_cancel;
D
Dave Chinner 已提交
2660

2661
	xfs_defer_init(&dfops, &first_block);
2662
	tp->t_agfl_dfops = &dfops;
2663
	error = xfs_dir_removename(tp, dp, name, ip->i_ino,
2664
					&first_block, &dfops, resblks);
2665
	if (error) {
D
Dave Chinner 已提交
2666
		ASSERT(error != -ENOENT);
2667 2668 2669
		goto out_bmap_cancel;
	}

D
Dave Chinner 已提交
2670 2671 2672 2673 2674 2675 2676 2677
	/*
	 * If this is a synchronous mount, make sure that the
	 * remove transaction goes to disk before returning to
	 * the user.
	 */
	if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC))
		xfs_trans_set_sync(tp);

2678
	error = xfs_defer_finish(&tp, &dfops);
D
Dave Chinner 已提交
2679 2680 2681
	if (error)
		goto out_bmap_cancel;

2682
	error = xfs_trans_commit(tp);
D
Dave Chinner 已提交
2683 2684 2685
	if (error)
		goto std_return;

2686
	if (is_dir && xfs_inode_is_filestream(ip))
D
Dave Chinner 已提交
2687 2688 2689 2690 2691
		xfs_filestream_deassociate(ip);

	return 0;

 out_bmap_cancel:
2692
	xfs_defer_cancel(&dfops);
D
Dave Chinner 已提交
2693
 out_trans_cancel:
2694
	xfs_trans_cancel(tp);
D
Dave Chinner 已提交
2695 2696 2697 2698
 std_return:
	return error;
}

D
Dave Chinner 已提交
2699 2700 2701
/*
 * Enter all inodes for a rename transaction into a sorted array.
 */
2702
#define __XFS_SORT_INODES	5
D
Dave Chinner 已提交
2703 2704
STATIC void
xfs_sort_for_rename(
2705 2706 2707 2708 2709 2710 2711
	struct xfs_inode	*dp1,	/* in: old (source) directory inode */
	struct xfs_inode	*dp2,	/* in: new (target) directory inode */
	struct xfs_inode	*ip1,	/* in: inode of old entry */
	struct xfs_inode	*ip2,	/* in: inode of new entry */
	struct xfs_inode	*wip,	/* in: whiteout inode */
	struct xfs_inode	**i_tab,/* out: sorted array of inodes */
	int			*num_inodes)  /* in/out: inodes in array */
D
Dave Chinner 已提交
2712 2713 2714
{
	int			i, j;

2715 2716 2717
	ASSERT(*num_inodes == __XFS_SORT_INODES);
	memset(i_tab, 0, *num_inodes * sizeof(struct xfs_inode *));

D
Dave Chinner 已提交
2718 2719 2720 2721 2722 2723 2724
	/*
	 * i_tab contains a list of pointers to inodes.  We initialize
	 * the table here & we'll sort it.  We will then use it to
	 * order the acquisition of the inode locks.
	 *
	 * Note that the table may contain duplicates.  e.g., dp1 == dp2.
	 */
2725 2726 2727 2728 2729 2730 2731 2732 2733
	i = 0;
	i_tab[i++] = dp1;
	i_tab[i++] = dp2;
	i_tab[i++] = ip1;
	if (ip2)
		i_tab[i++] = ip2;
	if (wip)
		i_tab[i++] = wip;
	*num_inodes = i;
D
Dave Chinner 已提交
2734 2735 2736

	/*
	 * Sort the elements via bubble sort.  (Remember, there are at
2737
	 * most 5 elements to sort, so this is adequate.)
D
Dave Chinner 已提交
2738 2739 2740 2741
	 */
	for (i = 0; i < *num_inodes; i++) {
		for (j = 1; j < *num_inodes; j++) {
			if (i_tab[j]->i_ino < i_tab[j-1]->i_ino) {
2742
				struct xfs_inode *temp = i_tab[j];
D
Dave Chinner 已提交
2743 2744 2745 2746 2747 2748 2749
				i_tab[j] = i_tab[j-1];
				i_tab[j-1] = temp;
			}
		}
	}
}

2750 2751 2752
static int
xfs_finish_rename(
	struct xfs_trans	*tp,
2753
	struct xfs_defer_ops	*dfops)
2754 2755 2756 2757 2758 2759 2760 2761 2762 2763
{
	int			error;

	/*
	 * If this is a synchronous mount, make sure that the rename transaction
	 * goes to disk before returning to the user.
	 */
	if (tp->t_mountp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC))
		xfs_trans_set_sync(tp);

2764
	error = xfs_defer_finish(&tp, dfops);
2765
	if (error) {
2766
		xfs_defer_cancel(dfops);
2767
		xfs_trans_cancel(tp);
2768 2769 2770
		return error;
	}

2771
	return xfs_trans_commit(tp);
2772 2773
}

2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787
/*
 * xfs_cross_rename()
 *
 * responsible for handling RENAME_EXCHANGE flag in renameat2() sytemcall
 */
STATIC int
xfs_cross_rename(
	struct xfs_trans	*tp,
	struct xfs_inode	*dp1,
	struct xfs_name		*name1,
	struct xfs_inode	*ip1,
	struct xfs_inode	*dp2,
	struct xfs_name		*name2,
	struct xfs_inode	*ip2,
2788
	struct xfs_defer_ops	*dfops,
2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799
	xfs_fsblock_t		*first_block,
	int			spaceres)
{
	int		error = 0;
	int		ip1_flags = 0;
	int		ip2_flags = 0;
	int		dp2_flags = 0;

	/* Swap inode number for dirent in first parent */
	error = xfs_dir_replace(tp, dp1, name1,
				ip2->i_ino,
2800
				first_block, dfops, spaceres);
2801
	if (error)
2802
		goto out_trans_abort;
2803 2804 2805 2806

	/* Swap inode number for dirent in second parent */
	error = xfs_dir_replace(tp, dp2, name2,
				ip1->i_ino,
2807
				first_block, dfops, spaceres);
2808
	if (error)
2809
		goto out_trans_abort;
2810 2811 2812 2813 2814 2815 2816 2817 2818

	/*
	 * If we're renaming one or more directories across different parents,
	 * update the respective ".." entries (and link counts) to match the new
	 * parents.
	 */
	if (dp1 != dp2) {
		dp2_flags = XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;

D
Dave Chinner 已提交
2819
		if (S_ISDIR(VFS_I(ip2)->i_mode)) {
2820 2821
			error = xfs_dir_replace(tp, ip2, &xfs_name_dotdot,
						dp1->i_ino, first_block,
2822
						dfops, spaceres);
2823
			if (error)
2824
				goto out_trans_abort;
2825 2826

			/* transfer ip2 ".." reference to dp1 */
D
Dave Chinner 已提交
2827
			if (!S_ISDIR(VFS_I(ip1)->i_mode)) {
2828 2829
				error = xfs_droplink(tp, dp2);
				if (error)
2830
					goto out_trans_abort;
2831 2832
				error = xfs_bumplink(tp, dp1);
				if (error)
2833
					goto out_trans_abort;
2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845
			}

			/*
			 * Although ip1 isn't changed here, userspace needs
			 * to be warned about the change, so that applications
			 * relying on it (like backup ones), will properly
			 * notify the change
			 */
			ip1_flags |= XFS_ICHGTIME_CHG;
			ip2_flags |= XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;
		}

D
Dave Chinner 已提交
2846
		if (S_ISDIR(VFS_I(ip1)->i_mode)) {
2847 2848
			error = xfs_dir_replace(tp, ip1, &xfs_name_dotdot,
						dp2->i_ino, first_block,
2849
						dfops, spaceres);
2850
			if (error)
2851
				goto out_trans_abort;
2852 2853

			/* transfer ip1 ".." reference to dp2 */
D
Dave Chinner 已提交
2854
			if (!S_ISDIR(VFS_I(ip2)->i_mode)) {
2855 2856
				error = xfs_droplink(tp, dp1);
				if (error)
2857
					goto out_trans_abort;
2858 2859
				error = xfs_bumplink(tp, dp2);
				if (error)
2860
					goto out_trans_abort;
2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887
			}

			/*
			 * Although ip2 isn't changed here, userspace needs
			 * to be warned about the change, so that applications
			 * relying on it (like backup ones), will properly
			 * notify the change
			 */
			ip1_flags |= XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;
			ip2_flags |= XFS_ICHGTIME_CHG;
		}
	}

	if (ip1_flags) {
		xfs_trans_ichgtime(tp, ip1, ip1_flags);
		xfs_trans_log_inode(tp, ip1, XFS_ILOG_CORE);
	}
	if (ip2_flags) {
		xfs_trans_ichgtime(tp, ip2, ip2_flags);
		xfs_trans_log_inode(tp, ip2, XFS_ILOG_CORE);
	}
	if (dp2_flags) {
		xfs_trans_ichgtime(tp, dp2, dp2_flags);
		xfs_trans_log_inode(tp, dp2, XFS_ILOG_CORE);
	}
	xfs_trans_ichgtime(tp, dp1, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	xfs_trans_log_inode(tp, dp1, XFS_ILOG_CORE);
2888
	return xfs_finish_rename(tp, dfops);
2889 2890

out_trans_abort:
2891
	xfs_defer_cancel(dfops);
2892
	xfs_trans_cancel(tp);
2893 2894 2895
	return error;
}

D
Dave Chinner 已提交
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911
/*
 * xfs_rename_alloc_whiteout()
 *
 * Return a referenced, unlinked, unlocked inode that that can be used as a
 * whiteout in a rename transaction. We use a tmpfile inode here so that if we
 * crash between allocating the inode and linking it into the rename transaction
 * recovery will free the inode and we won't leak it.
 */
static int
xfs_rename_alloc_whiteout(
	struct xfs_inode	*dp,
	struct xfs_inode	**wip)
{
	struct xfs_inode	*tmpfile;
	int			error;

2912
	error = xfs_create_tmpfile(dp, S_IFCHR | WHITEOUT_MODE, &tmpfile);
D
Dave Chinner 已提交
2913 2914 2915
	if (error)
		return error;

2916 2917 2918 2919 2920 2921 2922
	/*
	 * Prepare the tmpfile inode as if it were created through the VFS.
	 * Otherwise, the link increment paths will complain about nlink 0->1.
	 * Drop the link count as done by d_tmpfile(), complete the inode setup
	 * and flag it as linkable.
	 */
	drop_nlink(VFS_I(tmpfile));
2923
	xfs_setup_iops(tmpfile);
D
Dave Chinner 已提交
2924 2925 2926 2927 2928 2929 2930
	xfs_finish_inode_setup(tmpfile);
	VFS_I(tmpfile)->i_state |= I_LINKABLE;

	*wip = tmpfile;
	return 0;
}

D
Dave Chinner 已提交
2931 2932 2933 2934 2935
/*
 * xfs_rename
 */
int
xfs_rename(
D
Dave Chinner 已提交
2936 2937 2938 2939 2940 2941 2942
	struct xfs_inode	*src_dp,
	struct xfs_name		*src_name,
	struct xfs_inode	*src_ip,
	struct xfs_inode	*target_dp,
	struct xfs_name		*target_name,
	struct xfs_inode	*target_ip,
	unsigned int		flags)
D
Dave Chinner 已提交
2943
{
D
Dave Chinner 已提交
2944 2945
	struct xfs_mount	*mp = src_dp->i_mount;
	struct xfs_trans	*tp;
2946
	struct xfs_defer_ops	dfops;
D
Dave Chinner 已提交
2947 2948 2949 2950
	xfs_fsblock_t		first_block;
	struct xfs_inode	*wip = NULL;		/* whiteout inode */
	struct xfs_inode	*inodes[__XFS_SORT_INODES];
	int			num_inodes = __XFS_SORT_INODES;
2951
	bool			new_parent = (src_dp != target_dp);
D
Dave Chinner 已提交
2952
	bool			src_is_directory = S_ISDIR(VFS_I(src_ip)->i_mode);
D
Dave Chinner 已提交
2953 2954
	int			spaceres;
	int			error;
D
Dave Chinner 已提交
2955 2956 2957

	trace_xfs_rename(src_dp, target_dp, src_name, target_name);

2958 2959 2960
	if ((flags & RENAME_EXCHANGE) && !target_ip)
		return -EINVAL;

D
Dave Chinner 已提交
2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974
	/*
	 * If we are doing a whiteout operation, allocate the whiteout inode
	 * we will be placing at the target and ensure the type is set
	 * appropriately.
	 */
	if (flags & RENAME_WHITEOUT) {
		ASSERT(!(flags & (RENAME_NOREPLACE | RENAME_EXCHANGE)));
		error = xfs_rename_alloc_whiteout(target_dp, &wip);
		if (error)
			return error;

		/* setup target dirent info as whiteout */
		src_name->type = XFS_DIR3_FT_CHRDEV;
	}
D
Dave Chinner 已提交
2975

D
Dave Chinner 已提交
2976
	xfs_sort_for_rename(src_dp, target_dp, src_ip, target_ip, wip,
D
Dave Chinner 已提交
2977 2978 2979
				inodes, &num_inodes);

	spaceres = XFS_RENAME_SPACE_RES(mp, target_name->len);
2980
	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_rename, spaceres, 0, 0, &tp);
D
Dave Chinner 已提交
2981
	if (error == -ENOSPC) {
D
Dave Chinner 已提交
2982
		spaceres = 0;
2983 2984
		error = xfs_trans_alloc(mp, &M_RES(mp)->tr_rename, 0, 0, 0,
				&tp);
D
Dave Chinner 已提交
2985
	}
2986
	if (error)
2987
		goto out_release_wip;
D
Dave Chinner 已提交
2988 2989 2990 2991 2992

	/*
	 * Attach the dquots to the inodes
	 */
	error = xfs_qm_vop_rename_dqattach(inodes);
2993 2994
	if (error)
		goto out_trans_cancel;
D
Dave Chinner 已提交
2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008

	/*
	 * Lock all the participating inodes. Depending upon whether
	 * the target_name exists in the target directory, and
	 * whether the target directory is the same as the source
	 * directory, we can lock from 2 to 4 inodes.
	 */
	xfs_lock_inodes(inodes, num_inodes, XFS_ILOCK_EXCL);

	/*
	 * Join all the inodes to the transaction. From this point on,
	 * we can rely on either trans_commit or trans_cancel to unlock
	 * them.
	 */
3009
	xfs_trans_ijoin(tp, src_dp, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
3010
	if (new_parent)
3011
		xfs_trans_ijoin(tp, target_dp, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
3012 3013 3014
	xfs_trans_ijoin(tp, src_ip, XFS_ILOCK_EXCL);
	if (target_ip)
		xfs_trans_ijoin(tp, target_ip, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
3015 3016
	if (wip)
		xfs_trans_ijoin(tp, wip, XFS_ILOCK_EXCL);
D
Dave Chinner 已提交
3017 3018 3019 3020 3021 3022 3023 3024

	/*
	 * If we are using project inheritance, we only allow renames
	 * into our tree when the project IDs are the same; else the
	 * tree quota mechanism would be circumvented.
	 */
	if (unlikely((target_dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
		     (xfs_get_projid(target_dp) != xfs_get_projid(src_ip)))) {
D
Dave Chinner 已提交
3025
		error = -EXDEV;
3026
		goto out_trans_cancel;
D
Dave Chinner 已提交
3027 3028
	}

3029
	xfs_defer_init(&dfops, &first_block);
3030
	tp->t_agfl_dfops = &dfops;
3031

3032 3033 3034 3035
	/* RENAME_EXCHANGE is unique from here on. */
	if (flags & RENAME_EXCHANGE)
		return xfs_cross_rename(tp, src_dp, src_name, src_ip,
					target_dp, target_name, target_ip,
3036
					&dfops, &first_block, spaceres);
3037

D
Dave Chinner 已提交
3038 3039 3040 3041 3042 3043 3044 3045
	/*
	 * Set up the target.
	 */
	if (target_ip == NULL) {
		/*
		 * If there's no space reservation, check the entry will
		 * fit before actually inserting it.
		 */
3046 3047 3048
		if (!spaceres) {
			error = xfs_dir_canenter(tp, target_dp, target_name);
			if (error)
3049
				goto out_trans_cancel;
3050
		}
D
Dave Chinner 已提交
3051 3052 3053 3054 3055 3056 3057
		/*
		 * If target does not exist and the rename crosses
		 * directories, adjust the target directory link count
		 * to account for the ".." reference from the new entry.
		 */
		error = xfs_dir_createname(tp, target_dp, target_name,
						src_ip->i_ino, &first_block,
3058
						&dfops, spaceres);
D
Dave Chinner 已提交
3059
		if (error)
3060
			goto out_bmap_cancel;
D
Dave Chinner 已提交
3061 3062 3063 3064 3065 3066 3067

		xfs_trans_ichgtime(tp, target_dp,
					XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);

		if (new_parent && src_is_directory) {
			error = xfs_bumplink(tp, target_dp);
			if (error)
3068
				goto out_bmap_cancel;
D
Dave Chinner 已提交
3069 3070 3071 3072 3073 3074 3075
		}
	} else { /* target_ip != NULL */
		/*
		 * If target exists and it's a directory, check that both
		 * target and source are directories and that target can be
		 * destroyed, or that neither is a directory.
		 */
D
Dave Chinner 已提交
3076
		if (S_ISDIR(VFS_I(target_ip)->i_mode)) {
D
Dave Chinner 已提交
3077 3078 3079 3080
			/*
			 * Make sure target dir is empty.
			 */
			if (!(xfs_dir_isempty(target_ip)) ||
3081
			    (VFS_I(target_ip)->i_nlink > 2)) {
D
Dave Chinner 已提交
3082
				error = -EEXIST;
3083
				goto out_trans_cancel;
D
Dave Chinner 已提交
3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097
			}
		}

		/*
		 * Link the source inode under the target name.
		 * If the source inode is a directory and we are moving
		 * it across directories, its ".." entry will be
		 * inconsistent until we replace that down below.
		 *
		 * In case there is already an entry with the same
		 * name at the destination directory, remove it first.
		 */
		error = xfs_dir_replace(tp, target_dp, target_name,
					src_ip->i_ino,
3098
					&first_block, &dfops, spaceres);
D
Dave Chinner 已提交
3099
		if (error)
3100
			goto out_bmap_cancel;
D
Dave Chinner 已提交
3101 3102 3103 3104 3105 3106 3107 3108 3109 3110

		xfs_trans_ichgtime(tp, target_dp,
					XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);

		/*
		 * Decrement the link count on the target since the target
		 * dir no longer points to it.
		 */
		error = xfs_droplink(tp, target_ip);
		if (error)
3111
			goto out_bmap_cancel;
D
Dave Chinner 已提交
3112 3113 3114 3115 3116 3117 3118

		if (src_is_directory) {
			/*
			 * Drop the link from the old "." entry.
			 */
			error = xfs_droplink(tp, target_ip);
			if (error)
3119
				goto out_bmap_cancel;
D
Dave Chinner 已提交
3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132
		}
	} /* target_ip != NULL */

	/*
	 * Remove the source.
	 */
	if (new_parent && src_is_directory) {
		/*
		 * Rewrite the ".." entry to point to the new
		 * directory.
		 */
		error = xfs_dir_replace(tp, src_ip, &xfs_name_dotdot,
					target_dp->i_ino,
3133
					&first_block, &dfops, spaceres);
D
Dave Chinner 已提交
3134
		ASSERT(error != -EEXIST);
D
Dave Chinner 已提交
3135
		if (error)
3136
			goto out_bmap_cancel;
D
Dave Chinner 已提交
3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161
	}

	/*
	 * We always want to hit the ctime on the source inode.
	 *
	 * This isn't strictly required by the standards since the source
	 * inode isn't really being changed, but old unix file systems did
	 * it and some incremental backup programs won't work without it.
	 */
	xfs_trans_ichgtime(tp, src_ip, XFS_ICHGTIME_CHG);
	xfs_trans_log_inode(tp, src_ip, XFS_ILOG_CORE);

	/*
	 * Adjust the link count on src_dp.  This is necessary when
	 * renaming a directory, either within one parent when
	 * the target existed, or across two parent directories.
	 */
	if (src_is_directory && (new_parent || target_ip != NULL)) {

		/*
		 * Decrement link count on src_directory since the
		 * entry that's moved no longer points to it.
		 */
		error = xfs_droplink(tp, src_dp);
		if (error)
3162
			goto out_bmap_cancel;
D
Dave Chinner 已提交
3163 3164
	}

D
Dave Chinner 已提交
3165 3166 3167 3168 3169 3170 3171
	/*
	 * For whiteouts, we only need to update the source dirent with the
	 * inode number of the whiteout inode rather than removing it
	 * altogether.
	 */
	if (wip) {
		error = xfs_dir_replace(tp, src_dp, src_name, wip->i_ino,
3172
					&first_block, &dfops, spaceres);
D
Dave Chinner 已提交
3173 3174
	} else
		error = xfs_dir_removename(tp, src_dp, src_name, src_ip->i_ino,
3175
					   &first_block, &dfops, spaceres);
D
Dave Chinner 已提交
3176
	if (error)
3177
		goto out_bmap_cancel;
D
Dave Chinner 已提交
3178 3179

	/*
D
Dave Chinner 已提交
3180 3181 3182 3183 3184 3185
	 * For whiteouts, we need to bump the link count on the whiteout inode.
	 * This means that failures all the way up to this point leave the inode
	 * on the unlinked list and so cleanup is a simple matter of dropping
	 * the remaining reference to it. If we fail here after bumping the link
	 * count, we're shutting down the filesystem so we'll never see the
	 * intermediate state on disk.
D
Dave Chinner 已提交
3186
	 */
D
Dave Chinner 已提交
3187
	if (wip) {
3188
		ASSERT(VFS_I(wip)->i_nlink == 0);
D
Dave Chinner 已提交
3189 3190
		error = xfs_bumplink(tp, wip);
		if (error)
3191
			goto out_bmap_cancel;
D
Dave Chinner 已提交
3192 3193
		error = xfs_iunlink_remove(tp, wip);
		if (error)
3194
			goto out_bmap_cancel;
D
Dave Chinner 已提交
3195
		xfs_trans_log_inode(tp, wip, XFS_ILOG_CORE);
D
Dave Chinner 已提交
3196

D
Dave Chinner 已提交
3197 3198 3199 3200 3201 3202
		/*
		 * Now we have a real link, clear the "I'm a tmpfile" state
		 * flag from the inode so it doesn't accidentally get misused in
		 * future.
		 */
		VFS_I(wip)->i_state &= ~I_LINKABLE;
D
Dave Chinner 已提交
3203 3204 3205 3206 3207 3208 3209
	}

	xfs_trans_ichgtime(tp, src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE);
	if (new_parent)
		xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE);

3210
	error = xfs_finish_rename(tp, &dfops);
D
Dave Chinner 已提交
3211 3212 3213
	if (wip)
		IRELE(wip);
	return error;
D
Dave Chinner 已提交
3214

3215
out_bmap_cancel:
3216
	xfs_defer_cancel(&dfops);
3217
out_trans_cancel:
3218
	xfs_trans_cancel(tp);
3219
out_release_wip:
D
Dave Chinner 已提交
3220 3221
	if (wip)
		IRELE(wip);
D
Dave Chinner 已提交
3222 3223 3224
	return error;
}

3225 3226
STATIC int
xfs_iflush_cluster(
3227 3228
	struct xfs_inode	*ip,
	struct xfs_buf		*bp)
L
Linus Torvalds 已提交
3229
{
3230
	struct xfs_mount	*mp = ip->i_mount;
3231 3232 3233
	struct xfs_perag	*pag;
	unsigned long		first_index, mask;
	unsigned long		inodes_per_cluster;
3234 3235 3236
	int			cilist_size;
	struct xfs_inode	**cilist;
	struct xfs_inode	*cip;
3237 3238 3239
	int			nr_found;
	int			clcount = 0;
	int			bufwasdelwri;
L
Linus Torvalds 已提交
3240 3241
	int			i;

3242
	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
L
Linus Torvalds 已提交
3243

3244
	inodes_per_cluster = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog;
3245 3246 3247
	cilist_size = inodes_per_cluster * sizeof(xfs_inode_t *);
	cilist = kmem_alloc(cilist_size, KM_MAYFAIL|KM_NOFS);
	if (!cilist)
3248
		goto out_put;
L
Linus Torvalds 已提交
3249

3250
	mask = ~(((mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog)) - 1);
3251 3252 3253
	first_index = XFS_INO_TO_AGINO(mp, ip->i_ino) & mask;
	rcu_read_lock();
	/* really need a gang lookup range call here */
3254
	nr_found = radix_tree_gang_lookup(&pag->pag_ici_root, (void**)cilist,
3255 3256 3257 3258 3259
					first_index, inodes_per_cluster);
	if (nr_found == 0)
		goto out_free;

	for (i = 0; i < nr_found; i++) {
3260 3261
		cip = cilist[i];
		if (cip == ip)
3262
			continue;
3263 3264 3265 3266 3267 3268 3269

		/*
		 * 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.
		 */
3270 3271 3272 3273
		spin_lock(&cip->i_flags_lock);
		if (!cip->i_ino ||
		    __xfs_iflags_test(cip, XFS_ISTALE)) {
			spin_unlock(&cip->i_flags_lock);
3274 3275
			continue;
		}
3276 3277 3278 3279 3280 3281

		/*
		 * Once we fall off the end of the cluster, no point checking
		 * any more inodes in the list because they will also all be
		 * outside the cluster.
		 */
3282 3283
		if ((XFS_INO_TO_AGINO(mp, cip->i_ino) & mask) != first_index) {
			spin_unlock(&cip->i_flags_lock);
3284 3285
			break;
		}
3286
		spin_unlock(&cip->i_flags_lock);
3287

3288 3289 3290 3291 3292
		/*
		 * 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.
		 */
3293
		if (xfs_inode_clean(cip) && xfs_ipincount(cip) == 0)
3294 3295 3296 3297 3298 3299 3300
			continue;

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

3301
		if (!xfs_ilock_nowait(cip, XFS_ILOCK_SHARED))
3302
			continue;
3303 3304
		if (!xfs_iflock_nowait(cip)) {
			xfs_iunlock(cip, XFS_ILOCK_SHARED);
3305 3306
			continue;
		}
3307 3308 3309
		if (xfs_ipincount(cip)) {
			xfs_ifunlock(cip);
			xfs_iunlock(cip, XFS_ILOCK_SHARED);
3310 3311 3312
			continue;
		}

3313 3314 3315 3316 3317 3318 3319

		/*
		 * Check the inode number again, just to be certain we are not
		 * racing with freeing in xfs_reclaim_inode(). See the comments
		 * in that function for more information as to why the initial
		 * check is not sufficient.
		 */
3320 3321 3322
		if (!cip->i_ino) {
			xfs_ifunlock(cip);
			xfs_iunlock(cip, XFS_ILOCK_SHARED);
3323 3324 3325 3326 3327 3328 3329
			continue;
		}

		/*
		 * arriving here means that this inode can be flushed.  First
		 * re-check that it's dirty before flushing.
		 */
3330
		if (!xfs_inode_clean(cip)) {
3331
			int	error;
3332
			error = xfs_iflush_int(cip, bp);
3333
			if (error) {
3334
				xfs_iunlock(cip, XFS_ILOCK_SHARED);
3335 3336 3337 3338
				goto cluster_corrupt_out;
			}
			clcount++;
		} else {
3339
			xfs_ifunlock(cip);
3340
		}
3341
		xfs_iunlock(cip, XFS_ILOCK_SHARED);
3342 3343 3344
	}

	if (clcount) {
3345 3346
		XFS_STATS_INC(mp, xs_icluster_flushcnt);
		XFS_STATS_ADD(mp, xs_icluster_flushinode, clcount);
3347 3348 3349
	}

out_free:
3350
	rcu_read_unlock();
3351
	kmem_free(cilist);
3352 3353
out_put:
	xfs_perag_put(pag);
3354 3355 3356 3357 3358 3359 3360 3361
	return 0;


cluster_corrupt_out:
	/*
	 * Corruption detected in the clustering loop.  Invalidate the
	 * inode buffer and shut down the filesystem.
	 */
3362
	rcu_read_unlock();
3363
	/*
3364
	 * Clean up the buffer.  If it was delwri, just release it --
3365 3366 3367
	 * brelse can handle it with no problems.  If not, shut down the
	 * filesystem before releasing the buffer.
	 */
3368
	bufwasdelwri = (bp->b_flags & _XBF_DELWRI_Q);
3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379
	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.
		 */
3380
		if (bp->b_iodone) {
3381
			bp->b_flags &= ~XBF_DONE;
3382
			xfs_buf_stale(bp);
D
Dave Chinner 已提交
3383
			xfs_buf_ioerror(bp, -EIO);
3384
			xfs_buf_ioend(bp);
3385
		} else {
3386
			xfs_buf_stale(bp);
3387 3388 3389 3390 3391 3392 3393
			xfs_buf_relse(bp);
		}
	}

	/*
	 * Unlocks the flush lock
	 */
3394 3395
	xfs_iflush_abort(cip, false);
	kmem_free(cilist);
3396
	xfs_perag_put(pag);
D
Dave Chinner 已提交
3397
	return -EFSCORRUPTED;
3398 3399
}

L
Linus Torvalds 已提交
3400
/*
3401 3402 3403 3404 3405 3406 3407
 * Flush dirty inode metadata into the backing buffer.
 *
 * The caller must have the inode lock and the inode flush lock held.  The
 * inode lock will still be held upon return to the caller, and the inode
 * flush lock will be released after the inode has reached the disk.
 *
 * The caller must write out the buffer returned in *bpp and release it.
L
Linus Torvalds 已提交
3408 3409 3410
 */
int
xfs_iflush(
3411 3412
	struct xfs_inode	*ip,
	struct xfs_buf		**bpp)
L
Linus Torvalds 已提交
3413
{
3414
	struct xfs_mount	*mp = ip->i_mount;
3415
	struct xfs_buf		*bp = NULL;
3416
	struct xfs_dinode	*dip;
L
Linus Torvalds 已提交
3417 3418
	int			error;

3419
	XFS_STATS_INC(mp, xs_iflush_count);
L
Linus Torvalds 已提交
3420

C
Christoph Hellwig 已提交
3421
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
3422
	ASSERT(xfs_isiflocked(ip));
L
Linus Torvalds 已提交
3423
	ASSERT(ip->i_d.di_format != XFS_DINODE_FMT_BTREE ||
3424
	       ip->i_d.di_nextents > XFS_IFORK_MAXEXT(ip, XFS_DATA_FORK));
L
Linus Torvalds 已提交
3425

3426
	*bpp = NULL;
L
Linus Torvalds 已提交
3427 3428 3429

	xfs_iunpin_wait(ip);

3430 3431 3432
	/*
	 * For stale inodes we cannot rely on the backing buffer remaining
	 * stale in cache for the remaining life of the stale inode and so
3433
	 * xfs_imap_to_bp() below may give us a buffer that no longer contains
3434 3435 3436 3437 3438 3439 3440 3441 3442
	 * 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
3443 3444 3445
	/*
	 * This may have been unpinned because the filesystem is shutting
	 * down forcibly. If that's the case we must not write this inode
3446 3447 3448 3449
	 * to disk, because the log record didn't make it to disk.
	 *
	 * We also have to remove the log item from the AIL in this case,
	 * as we wait for an empty AIL as part of the unmount process.
L
Linus Torvalds 已提交
3450 3451
	 */
	if (XFS_FORCED_SHUTDOWN(mp)) {
D
Dave Chinner 已提交
3452
		error = -EIO;
3453
		goto abort_out;
L
Linus Torvalds 已提交
3454 3455
	}

3456
	/*
3457 3458 3459 3460 3461 3462 3463
	 * Get the buffer containing the on-disk inode. We are doing a try-lock
	 * operation here, so we may get  an EAGAIN error. In that case, we
	 * simply want to return with the inode still dirty.
	 *
	 * If we get any other error, we effectively have a corruption situation
	 * and we cannot flush the inode, so we treat it the same as failing
	 * xfs_iflush_int().
3464
	 */
3465 3466
	error = xfs_imap_to_bp(mp, NULL, &ip->i_imap, &dip, &bp, XBF_TRYLOCK,
			       0);
3467
	if (error == -EAGAIN) {
3468 3469 3470
		xfs_ifunlock(ip);
		return error;
	}
3471 3472
	if (error)
		goto corrupt_out;
3473

L
Linus Torvalds 已提交
3474 3475 3476 3477
	/*
	 * First flush out the inode that xfs_iflush was called with.
	 */
	error = xfs_iflush_int(ip, bp);
3478
	if (error)
L
Linus Torvalds 已提交
3479 3480
		goto corrupt_out;

3481 3482 3483 3484
	/*
	 * If the buffer is pinned then push on the log now so we won't
	 * get stuck waiting in the write for too long.
	 */
3485
	if (xfs_buf_ispinned(bp))
3486
		xfs_log_force(mp, 0);
3487

L
Linus Torvalds 已提交
3488 3489 3490 3491
	/*
	 * inode clustering:
	 * see if other inodes can be gathered into this write
	 */
3492 3493 3494
	error = xfs_iflush_cluster(ip, bp);
	if (error)
		goto cluster_corrupt_out;
L
Linus Torvalds 已提交
3495

3496 3497
	*bpp = bp;
	return 0;
L
Linus Torvalds 已提交
3498 3499

corrupt_out:
3500 3501
	if (bp)
		xfs_buf_relse(bp);
3502
	xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
L
Linus Torvalds 已提交
3503
cluster_corrupt_out:
D
Dave Chinner 已提交
3504
	error = -EFSCORRUPTED;
3505
abort_out:
L
Linus Torvalds 已提交
3506 3507 3508
	/*
	 * Unlocks the flush lock
	 */
3509
	xfs_iflush_abort(ip, false);
3510
	return error;
L
Linus Torvalds 已提交
3511 3512
}

3513 3514 3515 3516 3517 3518 3519 3520
/*
 * If there are inline format data / attr forks attached to this inode,
 * make sure they're not corrupt.
 */
bool
xfs_inode_verify_forks(
	struct xfs_inode	*ip)
{
3521
	struct xfs_ifork	*ifp;
3522 3523 3524 3525
	xfs_failaddr_t		fa;

	fa = xfs_ifork_verify_data(ip, &xfs_default_ifork_ops);
	if (fa) {
3526 3527 3528
		ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
		xfs_inode_verifier_error(ip, -EFSCORRUPTED, "data fork",
				ifp->if_u1.if_data, ifp->if_bytes, fa);
3529 3530 3531 3532 3533
		return false;
	}

	fa = xfs_ifork_verify_attr(ip, &xfs_default_ifork_ops);
	if (fa) {
3534 3535 3536 3537
		ifp = XFS_IFORK_PTR(ip, XFS_ATTR_FORK);
		xfs_inode_verifier_error(ip, -EFSCORRUPTED, "attr fork",
				ifp ? ifp->if_u1.if_data : NULL,
				ifp ? ifp->if_bytes : 0, fa);
3538 3539 3540 3541 3542
		return false;
	}
	return true;
}

L
Linus Torvalds 已提交
3543 3544
STATIC int
xfs_iflush_int(
3545 3546
	struct xfs_inode	*ip,
	struct xfs_buf		*bp)
L
Linus Torvalds 已提交
3547
{
3548 3549 3550
	struct xfs_inode_log_item *iip = ip->i_itemp;
	struct xfs_dinode	*dip;
	struct xfs_mount	*mp = ip->i_mount;
L
Linus Torvalds 已提交
3551

C
Christoph Hellwig 已提交
3552
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
3553
	ASSERT(xfs_isiflocked(ip));
L
Linus Torvalds 已提交
3554
	ASSERT(ip->i_d.di_format != XFS_DINODE_FMT_BTREE ||
3555
	       ip->i_d.di_nextents > XFS_IFORK_MAXEXT(ip, XFS_DATA_FORK));
3556
	ASSERT(iip != NULL && iip->ili_fields != 0);
3557
	ASSERT(ip->i_d.di_version > 1);
L
Linus Torvalds 已提交
3558 3559

	/* set *dip = inode's place in the buffer */
3560
	dip = xfs_buf_offset(bp, ip->i_imap.im_boffset);
L
Linus Torvalds 已提交
3561

3562
	if (XFS_TEST_ERROR(dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC),
3563
			       mp, XFS_ERRTAG_IFLUSH_1)) {
3564
		xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
3565
			"%s: Bad inode %Lu magic number 0x%x, ptr "PTR_FMT,
3566
			__func__, ip->i_ino, be16_to_cpu(dip->di_magic), dip);
L
Linus Torvalds 已提交
3567 3568
		goto corrupt_out;
	}
D
Dave Chinner 已提交
3569
	if (S_ISREG(VFS_I(ip)->i_mode)) {
L
Linus Torvalds 已提交
3570 3571 3572
		if (XFS_TEST_ERROR(
		    (ip->i_d.di_format != XFS_DINODE_FMT_EXTENTS) &&
		    (ip->i_d.di_format != XFS_DINODE_FMT_BTREE),
3573
		    mp, XFS_ERRTAG_IFLUSH_3)) {
3574
			xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
3575
				"%s: Bad regular inode %Lu, ptr "PTR_FMT,
3576
				__func__, ip->i_ino, ip);
L
Linus Torvalds 已提交
3577 3578
			goto corrupt_out;
		}
D
Dave Chinner 已提交
3579
	} else if (S_ISDIR(VFS_I(ip)->i_mode)) {
L
Linus Torvalds 已提交
3580 3581 3582 3583
		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),
3584
		    mp, XFS_ERRTAG_IFLUSH_4)) {
3585
			xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
3586
				"%s: Bad directory inode %Lu, ptr "PTR_FMT,
3587
				__func__, ip->i_ino, ip);
L
Linus Torvalds 已提交
3588 3589 3590 3591
			goto corrupt_out;
		}
	}
	if (XFS_TEST_ERROR(ip->i_d.di_nextents + ip->i_d.di_anextents >
3592
				ip->i_d.di_nblocks, mp, XFS_ERRTAG_IFLUSH_5)) {
3593 3594
		xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
			"%s: detected corrupt incore inode %Lu, "
3595
			"total extents = %d, nblocks = %Ld, ptr "PTR_FMT,
3596
			__func__, ip->i_ino,
L
Linus Torvalds 已提交
3597
			ip->i_d.di_nextents + ip->i_d.di_anextents,
3598
			ip->i_d.di_nblocks, ip);
L
Linus Torvalds 已提交
3599 3600 3601
		goto corrupt_out;
	}
	if (XFS_TEST_ERROR(ip->i_d.di_forkoff > mp->m_sb.sb_inodesize,
3602
				mp, XFS_ERRTAG_IFLUSH_6)) {
3603
		xfs_alert_tag(mp, XFS_PTAG_IFLUSH,
3604
			"%s: bad inode %Lu, forkoff 0x%x, ptr "PTR_FMT,
3605
			__func__, ip->i_ino, ip->i_d.di_forkoff, ip);
L
Linus Torvalds 已提交
3606 3607
		goto corrupt_out;
	}
3608

L
Linus Torvalds 已提交
3609
	/*
3610
	 * Inode item log recovery for v2 inodes are dependent on the
3611 3612 3613 3614 3615 3616
	 * di_flushiter count for correct sequencing. We bump the flush
	 * iteration count so we can detect flushes which postdate a log record
	 * during recovery. This is redundant as we now log every change and
	 * hence this can't happen but we need to still do it to ensure
	 * backwards compatibility with old kernels that predate logging all
	 * inode changes.
L
Linus Torvalds 已提交
3617
	 */
3618 3619
	if (ip->i_d.di_version < 3)
		ip->i_d.di_flushiter++;
L
Linus Torvalds 已提交
3620

3621 3622
	/* Check the inline fork data before we write out. */
	if (!xfs_inode_verify_forks(ip))
3623 3624
		goto corrupt_out;

L
Linus Torvalds 已提交
3625
	/*
3626 3627 3628
	 * 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.
L
Linus Torvalds 已提交
3629
	 */
3630
	xfs_inode_to_disk(ip, dip, iip->ili_item.li_lsn);
L
Linus Torvalds 已提交
3631 3632 3633 3634 3635

	/* 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;

3636 3637 3638
	xfs_iflush_fork(ip, dip, iip, XFS_DATA_FORK);
	if (XFS_IFORK_Q(ip))
		xfs_iflush_fork(ip, dip, iip, XFS_ATTR_FORK);
L
Linus Torvalds 已提交
3639 3640 3641
	xfs_inobp_check(mp, bp);

	/*
3642 3643 3644 3645 3646 3647 3648
	 * We've recorded everything logged in the inode, so we'd like to clear
	 * the ili_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.
L
Linus Torvalds 已提交
3649
	 *
3650 3651 3652 3653 3654 3655
	 * What we do is move the bits to the ili_last_fields field.  When
	 * logging the inode, these bits are moved back to the ili_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 ili_fields and ili_last_fields will be cleared.
L
Linus Torvalds 已提交
3656
	 *
3657 3658 3659 3660 3661 3662 3663 3664
	 * We can play with the ili_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.
L
Linus Torvalds 已提交
3665
	 */
3666 3667
	iip->ili_last_fields = iip->ili_fields;
	iip->ili_fields = 0;
3668
	iip->ili_fsync_fields = 0;
3669
	iip->ili_logged = 1;
L
Linus Torvalds 已提交
3670

3671 3672
	xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn,
				&iip->ili_item.li_lsn);
L
Linus Torvalds 已提交
3673

3674 3675 3676 3677 3678 3679 3680
	/*
	 * 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.
	 */
	xfs_buf_attach_iodone(bp, xfs_iflush_done, &iip->ili_item);
L
Linus Torvalds 已提交
3681

3682 3683
	/* generate the checksum. */
	xfs_dinode_calc_crc(mp, dip);
L
Linus Torvalds 已提交
3684

3685
	ASSERT(!list_empty(&bp->b_li_list));
3686
	ASSERT(bp->b_iodone != NULL);
L
Linus Torvalds 已提交
3687 3688 3689
	return 0;

corrupt_out:
D
Dave Chinner 已提交
3690
	return -EFSCORRUPTED;
L
Linus Torvalds 已提交
3691
}