xfs_iops.c 31.6 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2 3
 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
 * All Rights Reserved.
L
Linus Torvalds 已提交
4
 *
5 6
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
L
Linus Torvalds 已提交
7 8
 * published by the Free Software Foundation.
 *
9 10 11 12
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
L
Linus Torvalds 已提交
13
 *
14 15 16
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
L
Linus Torvalds 已提交
17 18 19
 */
#include "xfs.h"
#include "xfs_fs.h"
20
#include "xfs_shared.h"
21 22 23
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
L
Linus Torvalds 已提交
24
#include "xfs_mount.h"
25
#include "xfs_da_format.h"
L
Linus Torvalds 已提交
26 27
#include "xfs_inode.h"
#include "xfs_bmap.h"
D
Dave Chinner 已提交
28
#include "xfs_bmap_util.h"
29 30
#include "xfs_acl.h"
#include "xfs_quota.h"
L
Linus Torvalds 已提交
31 32
#include "xfs_error.h"
#include "xfs_attr.h"
33
#include "xfs_trans.h"
C
Christoph Hellwig 已提交
34
#include "xfs_trace.h"
35
#include "xfs_icache.h"
D
Dave Chinner 已提交
36
#include "xfs_symlink.h"
37
#include "xfs_da_btree.h"
38
#include "xfs_dir2.h"
Z
Zhi Yong Wu 已提交
39
#include "xfs_trans_space.h"
40
#include "xfs_pnfs.h"
41
#include "xfs_iomap.h"
L
Linus Torvalds 已提交
42

43
#include <linux/capability.h>
L
Linus Torvalds 已提交
44
#include <linux/xattr.h>
45
#include <linux/posix_acl.h>
46
#include <linux/security.h>
47
#include <linux/iomap.h>
48
#include <linux/slab.h>
L
Linus Torvalds 已提交
49

50 51 52 53 54 55 56 57 58 59 60 61
/*
 * Directories have different lock order w.r.t. mmap_sem compared to regular
 * files. This is due to readdir potentially triggering page faults on a user
 * buffer inside filldir(), and this happens with the ilock on the directory
 * held. For regular files, the lock order is the other way around - the
 * mmap_sem is taken during the page fault, and then we lock the ilock to do
 * block mapping. Hence we need a different class for the directory ilock so
 * that lockdep can tell them apart.
 */
static struct lock_class_key xfs_nondir_ilock_class;
static struct lock_class_key xfs_dir_ilock_class;

62 63 64 65 66
static int
xfs_initxattrs(
	struct inode		*inode,
	const struct xattr	*xattr_array,
	void			*fs_info)
67
{
68 69 70
	const struct xattr	*xattr;
	struct xfs_inode	*ip = XFS_I(inode);
	int			error = 0;
71 72

	for (xattr = xattr_array; xattr->name != NULL; xattr++) {
D
Dave Chinner 已提交
73
		error = xfs_attr_set(ip, xattr->name, xattr->value,
74
				      xattr->value_len, ATTR_SECURE);
75 76 77 78 79 80
		if (error < 0)
			break;
	}
	return error;
}

81 82 83 84 85 86
/*
 * Hook in SELinux.  This is not quite correct yet, what we really need
 * here (as we do for default ACLs) is a mechanism by which creation of
 * these attrs can be journalled at inode creation time (along with the
 * inode, of course, such that log replay can't cause these to be lost).
 */
87

88
STATIC int
89
xfs_init_security(
90
	struct inode	*inode,
91 92
	struct inode	*dir,
	const struct qstr *qstr)
93
{
D
Dave Chinner 已提交
94
	return security_inode_init_security(inode, dir, qstr,
95
					     &xfs_initxattrs, NULL);
96 97
}

98 99
static void
xfs_dentry_to_name(
100 101 102 103 104 105 106 107 108 109
	struct xfs_name	*namep,
	struct dentry	*dentry)
{
	namep->name = dentry->d_name.name;
	namep->len = dentry->d_name.len;
	namep->type = XFS_DIR3_FT_UNKNOWN;
}

static int
xfs_dentry_mode_to_name(
110
	struct xfs_name	*namep,
111 112
	struct dentry	*dentry,
	int		mode)
113 114 115
{
	namep->name = dentry->d_name.name;
	namep->len = dentry->d_name.len;
116
	namep->type = xfs_mode_to_ftype(mode);
117 118 119 120 121

	if (unlikely(namep->type == XFS_DIR3_FT_UNKNOWN))
		return -EFSCORRUPTED;

	return 0;
122 123
}

124
STATIC void
125
xfs_cleanup_inode(
126
	struct inode	*dir,
127
	struct inode	*inode,
128
	struct dentry	*dentry)
129
{
130
	struct xfs_name	teardown;
131 132

	/* Oh, the horror.
133
	 * If we can't add the ACL or we fail in
134
	 * xfs_init_security we must back out.
135 136
	 * ENOSPC can hit here, among other things.
	 */
137
	xfs_dentry_to_name(&teardown, dentry);
138

139
	xfs_remove(XFS_I(dir), &teardown, XFS_I(inode));
140 141
}

L
Linus Torvalds 已提交
142
STATIC int
143
xfs_generic_create(
L
Linus Torvalds 已提交
144 145
	struct inode	*dir,
	struct dentry	*dentry,
A
Al Viro 已提交
146
	umode_t		mode,
147 148
	dev_t		rdev,
	bool		tmpfile)	/* unnamed file */
L
Linus Torvalds 已提交
149
{
C
Christoph Hellwig 已提交
150
	struct inode	*inode;
151
	struct xfs_inode *ip = NULL;
152
	struct posix_acl *default_acl, *acl;
153
	struct xfs_name	name;
L
Linus Torvalds 已提交
154 155 156 157 158 159
	int		error;

	/*
	 * Irix uses Missed'em'V split, but doesn't want to see
	 * the upper 5 bits of (14bit) major.
	 */
160 161 162 163 164 165 166
	if (S_ISCHR(mode) || S_ISBLK(mode)) {
		if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff))
			return -EINVAL;
		rdev = sysv_encode_dev(rdev);
	} else {
		rdev = 0;
	}
L
Linus Torvalds 已提交
167

168 169 170
	error = posix_acl_create(dir, &mode, &default_acl, &acl);
	if (error)
		return error;
L
Linus Torvalds 已提交
171

172 173 174 175 176
	/* Verify mode is valid also for tmpfile case */
	error = xfs_dentry_mode_to_name(&name, dentry, mode);
	if (unlikely(error))
		goto out_free_acl;

177 178 179 180 181
	if (!tmpfile) {
		error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip);
	} else {
		error = xfs_create_tmpfile(XFS_I(dir), dentry, mode, &ip);
	}
C
Christoph Hellwig 已提交
182 183
	if (unlikely(error))
		goto out_free_acl;
184

185
	inode = VFS_I(ip);
186

187
	error = xfs_init_security(inode, dir, &dentry->d_name);
C
Christoph Hellwig 已提交
188 189 190
	if (unlikely(error))
		goto out_cleanup_inode;

191
#ifdef CONFIG_XFS_POSIX_ACL
C
Christoph Hellwig 已提交
192
	if (default_acl) {
D
Dave Chinner 已提交
193
		error = xfs_set_acl(inode, default_acl, ACL_TYPE_DEFAULT);
194
		if (error)
C
Christoph Hellwig 已提交
195
			goto out_cleanup_inode;
L
Linus Torvalds 已提交
196
	}
197
	if (acl) {
D
Dave Chinner 已提交
198
		error = xfs_set_acl(inode, acl, ACL_TYPE_ACCESS);
199 200 201 202
		if (error)
			goto out_cleanup_inode;
	}
#endif
L
Linus Torvalds 已提交
203

204 205
	xfs_setup_iops(ip);

206 207 208 209 210
	if (tmpfile)
		d_tmpfile(dentry, inode);
	else
		d_instantiate(dentry, inode);

211 212
	xfs_finish_inode_setup(ip);

213 214 215 216 217
 out_free_acl:
	if (default_acl)
		posix_acl_release(default_acl);
	if (acl)
		posix_acl_release(acl);
D
Dave Chinner 已提交
218
	return error;
C
Christoph Hellwig 已提交
219 220

 out_cleanup_inode:
221
	xfs_finish_inode_setup(ip);
222 223 224
	if (!tmpfile)
		xfs_cleanup_inode(dir, inode, dentry);
	iput(inode);
225
	goto out_free_acl;
L
Linus Torvalds 已提交
226 227
}

228 229 230 231 232 233 234 235 236 237
STATIC int
xfs_vn_mknod(
	struct inode	*dir,
	struct dentry	*dentry,
	umode_t		mode,
	dev_t		rdev)
{
	return xfs_generic_create(dir, dentry, mode, rdev, false);
}

L
Linus Torvalds 已提交
238
STATIC int
239
xfs_vn_create(
L
Linus Torvalds 已提交
240 241
	struct inode	*dir,
	struct dentry	*dentry,
A
Al Viro 已提交
242
	umode_t		mode,
A
Al Viro 已提交
243
	bool		flags)
L
Linus Torvalds 已提交
244
{
245
	return xfs_vn_mknod(dir, dentry, mode, 0);
L
Linus Torvalds 已提交
246 247 248
}

STATIC int
249
xfs_vn_mkdir(
L
Linus Torvalds 已提交
250 251
	struct inode	*dir,
	struct dentry	*dentry,
252
	umode_t		mode)
L
Linus Torvalds 已提交
253
{
254
	return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0);
L
Linus Torvalds 已提交
255 256 257
}

STATIC struct dentry *
258
xfs_vn_lookup(
L
Linus Torvalds 已提交
259 260
	struct inode	*dir,
	struct dentry	*dentry,
A
Al Viro 已提交
261
	unsigned int flags)
L
Linus Torvalds 已提交
262
{
263
	struct xfs_inode *cip;
264
	struct xfs_name	name;
L
Linus Torvalds 已提交
265 266 267 268 269
	int		error;

	if (dentry->d_name.len >= MAXNAMELEN)
		return ERR_PTR(-ENAMETOOLONG);

270
	xfs_dentry_to_name(&name, dentry);
271
	error = xfs_lookup(XFS_I(dir), &name, &cip, NULL);
272
	if (unlikely(error)) {
D
Dave Chinner 已提交
273 274
		if (unlikely(error != -ENOENT))
			return ERR_PTR(error);
L
Linus Torvalds 已提交
275 276 277 278
		d_add(dentry, NULL);
		return NULL;
	}

279
	return d_splice_alias(VFS_I(cip), dentry);
L
Linus Torvalds 已提交
280 281
}

282 283 284 285
STATIC struct dentry *
xfs_vn_ci_lookup(
	struct inode	*dir,
	struct dentry	*dentry,
A
Al Viro 已提交
286
	unsigned int flags)
287 288 289 290 291 292 293 294 295 296
{
	struct xfs_inode *ip;
	struct xfs_name	xname;
	struct xfs_name ci_name;
	struct qstr	dname;
	int		error;

	if (dentry->d_name.len >= MAXNAMELEN)
		return ERR_PTR(-ENAMETOOLONG);

297
	xfs_dentry_to_name(&xname, dentry);
298 299
	error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name);
	if (unlikely(error)) {
D
Dave Chinner 已提交
300 301
		if (unlikely(error != -ENOENT))
			return ERR_PTR(error);
302 303 304 305 306
		/*
		 * call d_add(dentry, NULL) here when d_drop_negative_children
		 * is called in xfs_vn_mknod (ie. allow negative dentries
		 * with CI filesystems).
		 */
307 308 309 310 311
		return NULL;
	}

	/* if exact match, just splice and exit */
	if (!ci_name.name)
312
		return d_splice_alias(VFS_I(ip), dentry);
313 314 315 316

	/* else case-insensitive match... */
	dname.name = ci_name.name;
	dname.len = ci_name.len;
317
	dentry = d_add_ci(dentry, VFS_I(ip), &dname);
318 319 320 321
	kmem_free(ci_name.name);
	return dentry;
}

L
Linus Torvalds 已提交
322
STATIC int
323
xfs_vn_link(
L
Linus Torvalds 已提交
324 325 326 327
	struct dentry	*old_dentry,
	struct inode	*dir,
	struct dentry	*dentry)
{
328
	struct inode	*inode = d_inode(old_dentry);
329
	struct xfs_name	name;
L
Linus Torvalds 已提交
330 331
	int		error;

332 333 334
	error = xfs_dentry_mode_to_name(&name, dentry, inode->i_mode);
	if (unlikely(error))
		return error;
L
Linus Torvalds 已提交
335

336
	error = xfs_link(XFS_I(dir), XFS_I(inode), &name);
337
	if (unlikely(error))
D
Dave Chinner 已提交
338
		return error;
339

A
Al Viro 已提交
340
	ihold(inode);
341 342
	d_instantiate(dentry, inode);
	return 0;
L
Linus Torvalds 已提交
343 344 345
}

STATIC int
346
xfs_vn_unlink(
L
Linus Torvalds 已提交
347 348 349
	struct inode	*dir,
	struct dentry	*dentry)
{
350
	struct xfs_name	name;
L
Linus Torvalds 已提交
351 352
	int		error;

353
	xfs_dentry_to_name(&name, dentry);
L
Linus Torvalds 已提交
354

355
	error = xfs_remove(XFS_I(dir), &name, XFS_I(d_inode(dentry)));
356 357 358 359 360 361 362 363 364 365 366
	if (error)
		return error;

	/*
	 * With unlink, the VFS makes the dentry "negative": no inode,
	 * but still hashed. This is incompatible with case-insensitive
	 * mode, so invalidate (unhash) the dentry in CI-mode.
	 */
	if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb))
		d_invalidate(dentry);
	return 0;
L
Linus Torvalds 已提交
367 368 369
}

STATIC int
370
xfs_vn_symlink(
L
Linus Torvalds 已提交
371 372 373 374
	struct inode	*dir,
	struct dentry	*dentry,
	const char	*symname)
{
375 376
	struct inode	*inode;
	struct xfs_inode *cip = NULL;
377
	struct xfs_name	name;
L
Linus Torvalds 已提交
378
	int		error;
A
Al Viro 已提交
379
	umode_t		mode;
L
Linus Torvalds 已提交
380

381
	mode = S_IFLNK |
A
Al Viro 已提交
382
		(irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO);
383 384 385
	error = xfs_dentry_mode_to_name(&name, dentry, mode);
	if (unlikely(error))
		goto out;
L
Linus Torvalds 已提交
386

C
Christoph Hellwig 已提交
387
	error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip);
388 389 390
	if (unlikely(error))
		goto out;

391
	inode = VFS_I(cip);
392

393
	error = xfs_init_security(inode, dir, &dentry->d_name);
394 395 396
	if (unlikely(error))
		goto out_cleanup_inode;

397 398
	xfs_setup_iops(cip);

399
	d_instantiate(dentry, inode);
400
	xfs_finish_inode_setup(cip);
401 402 403
	return 0;

 out_cleanup_inode:
404
	xfs_finish_inode_setup(cip);
405
	xfs_cleanup_inode(dir, inode, dentry);
406
	iput(inode);
407
 out:
D
Dave Chinner 已提交
408
	return error;
L
Linus Torvalds 已提交
409 410 411
}

STATIC int
412
xfs_vn_rename(
L
Linus Torvalds 已提交
413 414 415
	struct inode	*odir,
	struct dentry	*odentry,
	struct inode	*ndir,
416 417
	struct dentry	*ndentry,
	unsigned int	flags)
L
Linus Torvalds 已提交
418
{
419
	struct inode	*new_inode = d_inode(ndentry);
420
	int		omode = 0;
421
	int		error;
422 423
	struct xfs_name	oname;
	struct xfs_name	nname;
L
Linus Torvalds 已提交
424

D
Dave Chinner 已提交
425
	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
426 427
		return -EINVAL;

428 429
	/* if we are exchanging files, we need to set i_mode of both files */
	if (flags & RENAME_EXCHANGE)
430
		omode = d_inode(ndentry)->i_mode;
431

432 433 434 435 436 437 438 439
	error = xfs_dentry_mode_to_name(&oname, odentry, omode);
	if (omode && unlikely(error))
		return error;

	error = xfs_dentry_mode_to_name(&nname, ndentry,
					d_inode(odentry)->i_mode);
	if (unlikely(error))
		return error;
440

441
	return xfs_rename(XFS_I(odir), &oname, XFS_I(d_inode(odentry)),
442
			  XFS_I(ndir), &nname,
443
			  new_inode ? XFS_I(new_inode) : NULL, flags);
L
Linus Torvalds 已提交
444 445 446 447 448 449 450
}

/*
 * careful here - this function can get called recursively, so
 * we need to be very careful about how much stack we use.
 * uio is kmalloced for this reason...
 */
451
STATIC const char *
452
xfs_vn_get_link(
L
Linus Torvalds 已提交
453
	struct dentry		*dentry,
454
	struct inode		*inode,
455
	struct delayed_call	*done)
L
Linus Torvalds 已提交
456 457
{
	char			*link;
458
	int			error = -ENOMEM;
L
Linus Torvalds 已提交
459

460 461 462
	if (!dentry)
		return ERR_PTR(-ECHILD);

463
	link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
464 465
	if (!link)
		goto out_err;
L
Linus Torvalds 已提交
466

467
	error = xfs_readlink(XFS_I(d_inode(dentry)), link);
468 469
	if (unlikely(error))
		goto out_kfree;
L
Linus Torvalds 已提交
470

471 472
	set_delayed_call(done, kfree_link, link);
	return link;
473 474 475 476

 out_kfree:
	kfree(link);
 out_err:
477
	return ERR_PTR(error);
L
Linus Torvalds 已提交
478 479
}

480 481 482 483 484 485 486 487 488 489
STATIC const char *
xfs_vn_get_link_inline(
	struct dentry		*dentry,
	struct inode		*inode,
	struct delayed_call	*done)
{
	ASSERT(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE);
	return XFS_I(inode)->i_df.if_u1.if_data;
}

L
Linus Torvalds 已提交
490
STATIC int
491
xfs_vn_getattr(
492 493 494
	struct vfsmount		*mnt,
	struct dentry		*dentry,
	struct kstat		*stat)
L
Linus Torvalds 已提交
495
{
496
	struct inode		*inode = d_inode(dentry);
497 498 499
	struct xfs_inode	*ip = XFS_I(inode);
	struct xfs_mount	*mp = ip->i_mount;

C
Christoph Hellwig 已提交
500
	trace_xfs_getattr(ip);
501 502

	if (XFS_FORCED_SHUTDOWN(mp))
E
Eric Sandeen 已提交
503
		return -EIO;
504 505 506

	stat->size = XFS_ISIZE(ip);
	stat->dev = inode->i_sb->s_dev;
D
Dave Chinner 已提交
507
	stat->mode = inode->i_mode;
508
	stat->nlink = inode->i_nlink;
509 510
	stat->uid = inode->i_uid;
	stat->gid = inode->i_gid;
511 512
	stat->ino = ip->i_ino;
	stat->atime = inode->i_atime;
513 514
	stat->mtime = inode->i_mtime;
	stat->ctime = inode->i_ctime;
515 516 517 518 519 520 521 522 523 524 525 526
	stat->blocks =
		XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);


	switch (inode->i_mode & S_IFMT) {
	case S_IFBLK:
	case S_IFCHR:
		stat->blksize = BLKDEV_IOSIZE;
		stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
				   sysv_minor(ip->i_df.if_u2.if_rdev));
		break;
	default:
527
		if (XFS_IS_REALTIME_INODE(ip)) {
528 529 530 531 532 533 534 535 536 537 538
			/*
			 * If the file blocks are being allocated from a
			 * realtime volume, then return the inode's realtime
			 * extent size or the realtime volume's extent size.
			 */
			stat->blksize =
				xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
		} else
			stat->blksize = xfs_preferred_iosize(mp);
		stat->rdev = 0;
		break;
539
	}
540 541

	return 0;
L
Linus Torvalds 已提交
542 543
}

544 545 546 547 548
static void
xfs_setattr_mode(
	struct xfs_inode	*ip,
	struct iattr		*iattr)
{
549 550
	struct inode		*inode = VFS_I(ip);
	umode_t			mode = iattr->ia_mode;
551 552 553 554 555 556 557

	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));

	inode->i_mode &= S_IFMT;
	inode->i_mode |= mode & ~S_IFMT;
}

558
void
C
Christoph Hellwig 已提交
559 560 561 562 563 564 565 566
xfs_setattr_time(
	struct xfs_inode	*ip,
	struct iattr		*iattr)
{
	struct inode		*inode = VFS_I(ip);

	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));

567
	if (iattr->ia_valid & ATTR_ATIME)
C
Christoph Hellwig 已提交
568
		inode->i_atime = iattr->ia_atime;
569
	if (iattr->ia_valid & ATTR_CTIME)
C
Christoph Hellwig 已提交
570
		inode->i_ctime = iattr->ia_ctime;
571
	if (iattr->ia_valid & ATTR_MTIME)
C
Christoph Hellwig 已提交
572 573 574
		inode->i_mtime = iattr->ia_mtime;
}

575 576 577 578 579
static int
xfs_vn_change_ok(
	struct dentry	*dentry,
	struct iattr	*iattr)
{
580
	struct xfs_mount	*mp = XFS_I(d_inode(dentry))->i_mount;
581 582 583 584 585 586 587

	if (mp->m_flags & XFS_MOUNT_RDONLY)
		return -EROFS;

	if (XFS_FORCED_SHUTDOWN(mp))
		return -EIO;

588
	return setattr_prepare(dentry, iattr);
589 590 591 592 593 594
}

/*
 * Set non-size attributes of an inode.
 *
 * Caution: The caller of this function is responsible for calling
595
 * setattr_prepare() or otherwise verifying the change is fine.
596
 */
C
Christoph Hellwig 已提交
597 598 599 600 601 602 603 604 605 606 607
int
xfs_setattr_nonsize(
	struct xfs_inode	*ip,
	struct iattr		*iattr,
	int			flags)
{
	xfs_mount_t		*mp = ip->i_mount;
	struct inode		*inode = VFS_I(ip);
	int			mask = iattr->ia_valid;
	xfs_trans_t		*tp;
	int			error;
608 609
	kuid_t			uid = GLOBAL_ROOT_UID, iuid = GLOBAL_ROOT_UID;
	kgid_t			gid = GLOBAL_ROOT_GID, igid = GLOBAL_ROOT_GID;
C
Christoph Hellwig 已提交
610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629
	struct xfs_dquot	*udqp = NULL, *gdqp = NULL;
	struct xfs_dquot	*olddquot1 = NULL, *olddquot2 = NULL;

	ASSERT((mask & ATTR_SIZE) == 0);

	/*
	 * If disk quotas is on, we make sure that the dquots do exist on disk,
	 * before we start any other transactions. Trying to do this later
	 * is messy. We don't care to take a readlock to look at the ids
	 * in inode here, because we can't hold it across the trans_reserve.
	 * If the IDs do change before we take the ilock, we're covered
	 * because the i_*dquot fields will get updated anyway.
	 */
	if (XFS_IS_QUOTA_ON(mp) && (mask & (ATTR_UID|ATTR_GID))) {
		uint	qflags = 0;

		if ((mask & ATTR_UID) && XFS_IS_UQUOTA_ON(mp)) {
			uid = iattr->ia_uid;
			qflags |= XFS_QMOPT_UQUOTA;
		} else {
630
			uid = inode->i_uid;
C
Christoph Hellwig 已提交
631 632 633 634 635
		}
		if ((mask & ATTR_GID) && XFS_IS_GQUOTA_ON(mp)) {
			gid = iattr->ia_gid;
			qflags |= XFS_QMOPT_GQUOTA;
		}  else {
636
			gid = inode->i_gid;
C
Christoph Hellwig 已提交
637 638 639 640 641 642 643 644 645
		}

		/*
		 * We take a reference when we initialize udqp and gdqp,
		 * so it is important that we never blindly double trip on
		 * the same variable. See xfs_create() for an example.
		 */
		ASSERT(udqp == NULL);
		ASSERT(gdqp == NULL);
646 647 648 649
		error = xfs_qm_vop_dqalloc(ip, xfs_kuid_to_uid(uid),
					   xfs_kgid_to_gid(gid),
					   xfs_get_projid(ip),
					   qflags, &udqp, &gdqp, NULL);
C
Christoph Hellwig 已提交
650 651 652 653
		if (error)
			return error;
	}

654
	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
C
Christoph Hellwig 已提交
655
	if (error)
656
		goto out_dqrele;
C
Christoph Hellwig 已提交
657 658

	xfs_ilock(ip, XFS_ILOCK_EXCL);
659
	xfs_trans_ijoin(tp, ip, 0);
C
Christoph Hellwig 已提交
660 661 662 663 664 665 666 667 668 669 670

	/*
	 * Change file ownership.  Must be the owner or privileged.
	 */
	if (mask & (ATTR_UID|ATTR_GID)) {
		/*
		 * These IDs could have changed since we last looked at them.
		 * But, we're assured that if the ownership did change
		 * while we didn't have the inode locked, inode's dquot(s)
		 * would have changed also.
		 */
671 672
		iuid = inode->i_uid;
		igid = inode->i_gid;
C
Christoph Hellwig 已提交
673 674 675 676 677 678 679 680
		gid = (mask & ATTR_GID) ? iattr->ia_gid : igid;
		uid = (mask & ATTR_UID) ? iattr->ia_uid : iuid;

		/*
		 * Do a quota reservation only if uid/gid is actually
		 * going to change.
		 */
		if (XFS_IS_QUOTA_RUNNING(mp) &&
681 682
		    ((XFS_IS_UQUOTA_ON(mp) && !uid_eq(iuid, uid)) ||
		     (XFS_IS_GQUOTA_ON(mp) && !gid_eq(igid, gid)))) {
C
Christoph Hellwig 已提交
683 684
			ASSERT(tp);
			error = xfs_qm_vop_chown_reserve(tp, ip, udqp, gdqp,
685
						NULL, capable(CAP_FOWNER) ?
C
Christoph Hellwig 已提交
686 687
						XFS_QMOPT_FORCE_RES : 0);
			if (error)	/* out of quota */
688
				goto out_cancel;
C
Christoph Hellwig 已提交
689 690 691 692 693 694 695 696 697 698 699 700 701
		}
	}

	/*
	 * Change file ownership.  Must be the owner or privileged.
	 */
	if (mask & (ATTR_UID|ATTR_GID)) {
		/*
		 * CAP_FSETID overrides the following restrictions:
		 *
		 * The set-user-ID and set-group-ID bits of a file will be
		 * cleared upon successful return from chown()
		 */
D
Dave Chinner 已提交
702
		if ((inode->i_mode & (S_ISUID|S_ISGID)) &&
C
Christoph Hellwig 已提交
703
		    !capable(CAP_FSETID))
D
Dave Chinner 已提交
704
			inode->i_mode &= ~(S_ISUID|S_ISGID);
C
Christoph Hellwig 已提交
705 706 707 708 709

		/*
		 * Change the ownerships and register quota modifications
		 * in the transaction.
		 */
710
		if (!uid_eq(iuid, uid)) {
C
Christoph Hellwig 已提交
711 712 713 714 715 716
			if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_UQUOTA_ON(mp)) {
				ASSERT(mask & ATTR_UID);
				ASSERT(udqp);
				olddquot1 = xfs_qm_vop_chown(tp, ip,
							&ip->i_udquot, udqp);
			}
717
			ip->i_d.di_uid = xfs_kuid_to_uid(uid);
C
Christoph Hellwig 已提交
718 719
			inode->i_uid = uid;
		}
720
		if (!gid_eq(igid, gid)) {
C
Christoph Hellwig 已提交
721
			if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_GQUOTA_ON(mp)) {
722 723
				ASSERT(xfs_sb_version_has_pquotino(&mp->m_sb) ||
				       !XFS_IS_PQUOTA_ON(mp));
C
Christoph Hellwig 已提交
724 725 726 727 728
				ASSERT(mask & ATTR_GID);
				ASSERT(gdqp);
				olddquot2 = xfs_qm_vop_chown(tp, ip,
							&ip->i_gdquot, gdqp);
			}
729
			ip->i_d.di_gid = xfs_kgid_to_gid(gid);
C
Christoph Hellwig 已提交
730 731 732 733
			inode->i_gid = gid;
		}
	}

734
	if (mask & ATTR_MODE)
735
		xfs_setattr_mode(ip, iattr);
C
Christoph Hellwig 已提交
736 737
	if (mask & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME))
		xfs_setattr_time(ip, iattr);
C
Christoph Hellwig 已提交
738 739 740

	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

741
	XFS_STATS_INC(mp, xs_ig_attrchg);
C
Christoph Hellwig 已提交
742 743 744

	if (mp->m_flags & XFS_MOUNT_WSYNC)
		xfs_trans_set_sync(tp);
745
	error = xfs_trans_commit(tp);
C
Christoph Hellwig 已提交
746 747 748 749 750 751 752 753 754 755 756 757

	xfs_iunlock(ip, XFS_ILOCK_EXCL);

	/*
	 * Release any dquot(s) the inode had kept before chown.
	 */
	xfs_qm_dqrele(olddquot1);
	xfs_qm_dqrele(olddquot2);
	xfs_qm_dqrele(udqp);
	xfs_qm_dqrele(gdqp);

	if (error)
E
Eric Sandeen 已提交
758
		return error;
C
Christoph Hellwig 已提交
759 760 761 762 763 764 765 766 767

	/*
	 * XXX(hch): Updating the ACL entries is not atomic vs the i_mode
	 * 	     update.  We could avoid this with linked transactions
	 * 	     and passing down the transaction pointer all the way
	 *	     to attr_set.  No previous user of the generic
	 * 	     Posix ACL code seems to care about this issue either.
	 */
	if ((mask & ATTR_MODE) && !(flags & XFS_ATTR_NOACL)) {
D
Dave Chinner 已提交
768
		error = posix_acl_chmod(inode, inode->i_mode);
C
Christoph Hellwig 已提交
769
		if (error)
E
Eric Sandeen 已提交
770
			return error;
C
Christoph Hellwig 已提交
771 772 773 774
	}

	return 0;

775
out_cancel:
776
	xfs_trans_cancel(tp);
777
out_dqrele:
C
Christoph Hellwig 已提交
778 779 780 781 782
	xfs_qm_dqrele(udqp);
	xfs_qm_dqrele(gdqp);
	return error;
}

783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798
int
xfs_vn_setattr_nonsize(
	struct dentry		*dentry,
	struct iattr		*iattr)
{
	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
	int error;

	trace_xfs_setattr(ip);

	error = xfs_vn_change_ok(dentry, iattr);
	if (error)
		return error;
	return xfs_setattr_nonsize(ip, iattr, 0);
}

C
Christoph Hellwig 已提交
799 800
/*
 * Truncate file.  Must have write permission and not be a directory.
801 802
 *
 * Caution: The caller of this function is responsible for calling
803
 * setattr_prepare() or otherwise verifying the change is fine.
C
Christoph Hellwig 已提交
804 805 806 807
 */
int
xfs_setattr_size(
	struct xfs_inode	*ip,
808
	struct iattr		*iattr)
C
Christoph Hellwig 已提交
809 810 811
{
	struct xfs_mount	*mp = ip->i_mount;
	struct inode		*inode = VFS_I(ip);
812
	xfs_off_t		oldsize, newsize;
C
Christoph Hellwig 已提交
813 814
	struct xfs_trans	*tp;
	int			error;
815
	uint			lock_flags = 0;
816
	bool			did_zeroing = false;
C
Christoph Hellwig 已提交
817

818
	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
819
	ASSERT(xfs_isilocked(ip, XFS_MMAPLOCK_EXCL));
D
Dave Chinner 已提交
820
	ASSERT(S_ISREG(inode->i_mode));
821 822
	ASSERT((iattr->ia_valid & (ATTR_UID|ATTR_GID|ATTR_ATIME|ATTR_ATIME_SET|
		ATTR_MTIME_SET|ATTR_KILL_PRIV|ATTR_TIMES_SET)) == 0);
C
Christoph Hellwig 已提交
823

824
	oldsize = inode->i_size;
825 826
	newsize = iattr->ia_size;

C
Christoph Hellwig 已提交
827 828 829
	/*
	 * Short circuit the truncate case for zero length files.
	 */
830
	if (newsize == 0 && oldsize == 0 && ip->i_d.di_nextents == 0) {
831
		if (!(iattr->ia_valid & (ATTR_CTIME|ATTR_MTIME)))
832
			return 0;
833 834 835 836 837 838

		/*
		 * Use the regular setattr path to update the timestamps.
		 */
		iattr->ia_valid &= ~ATTR_SIZE;
		return xfs_setattr_nonsize(ip, iattr, 0);
C
Christoph Hellwig 已提交
839 840 841 842 843
	}

	/*
	 * Make sure that the dquots are attached to the inode.
	 */
844
	error = xfs_qm_dqattach(ip, 0);
C
Christoph Hellwig 已提交
845
	if (error)
846
		return error;
C
Christoph Hellwig 已提交
847

848 849 850 851 852
	/*
	 * Wait for all direct I/O to complete.
	 */
	inode_dio_wait(inode);

C
Christoph Hellwig 已提交
853
	/*
854 855 856 857 858
	 * File data changes must be complete before we start the transaction to
	 * modify the inode.  This needs to be done before joining the inode to
	 * the transaction because the inode cannot be unlocked once it is a
	 * part of the transaction.
	 *
859 860 861
	 * Start with zeroing any data beyond EOF that we may expose on file
	 * extension, or zeroing out the rest of the block on a downward
	 * truncate.
C
Christoph Hellwig 已提交
862
	 */
863
	if (newsize > oldsize) {
864
		error = xfs_zero_eof(ip, newsize, oldsize, &did_zeroing);
865
	} else {
866 867
		error = iomap_truncate_page(inode, newsize, &did_zeroing,
				&xfs_iomap_ops);
C
Christoph Hellwig 已提交
868 869
	}

870 871 872
	if (error)
		return error;

C
Christoph Hellwig 已提交
873 874 875 876 877
	/*
	 * We are going to log the inode size change in this transaction so
	 * any previous writes that are beyond the on disk EOF and the new
	 * EOF that have not been written out need to be written here.  If we
	 * do not write the data out, we expose ourselves to the null files
878 879
	 * problem. Note that this includes any block zeroing we did above;
	 * otherwise those blocks may not be zeroed after a crash.
C
Christoph Hellwig 已提交
880
	 */
881 882
	if (did_zeroing ||
	    (newsize > ip->i_d.di_size && oldsize != ip->i_d.di_size)) {
D
Dave Chinner 已提交
883
		error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping,
D
Dave Chinner 已提交
884
						      ip->i_d.di_size, newsize);
C
Christoph Hellwig 已提交
885
		if (error)
886
			return error;
C
Christoph Hellwig 已提交
887 888
	}

889
	/*
890 891 892 893 894 895
	 * We've already locked out new page faults, so now we can safely remove
	 * pages from the page cache knowing they won't get refaulted until we
	 * drop the XFS_MMAP_EXCL lock after the extent manipulations are
	 * complete. The truncate_setsize() call also cleans partial EOF page
	 * PTEs on extending truncates and hence ensures sub-page block size
	 * filesystems are correctly handled, too.
896
	 *
897 898 899
	 * We have to do all the page cache truncate work outside the
	 * transaction context as the "lock" order is page lock->log space
	 * reservation as defined by extent allocation in the writeback path.
900
	 * Hence a truncate can fail with ENOMEM from xfs_trans_alloc(), but
901 902 903 904
	 * having already truncated the in-memory version of the file (i.e. made
	 * user visible changes). There's not much we can do about this, except
	 * to hope that the caller sees ENOMEM and retries the truncate
	 * operation.
905 906
	 */
	truncate_setsize(inode, newsize);
C
Christoph Hellwig 已提交
907

908
	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp);
C
Christoph Hellwig 已提交
909
	if (error)
910
		return error;
C
Christoph Hellwig 已提交
911 912 913

	lock_flags |= XFS_ILOCK_EXCL;
	xfs_ilock(ip, XFS_ILOCK_EXCL);
914
	xfs_trans_ijoin(tp, ip, 0);
C
Christoph Hellwig 已提交
915 916 917 918 919 920 921 922 923 924 925

	/*
	 * Only change the c/mtime if we are changing the size or we are
	 * explicitly asked to change it.  This handles the semantic difference
	 * between truncate() and ftruncate() as implemented in the VFS.
	 *
	 * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a
	 * special case where we need to update the times despite not having
	 * these flags set.  For all other operations the VFS set these flags
	 * explicitly if it wants a timestamp update.
	 */
926 927
	if (newsize != oldsize &&
	    !(iattr->ia_valid & (ATTR_CTIME | ATTR_MTIME))) {
C
Christoph Hellwig 已提交
928
		iattr->ia_ctime = iattr->ia_mtime =
929
			current_time(inode);
930
		iattr->ia_valid |= ATTR_CTIME | ATTR_MTIME;
C
Christoph Hellwig 已提交
931 932
	}

933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949
	/*
	 * The first thing we do is set the size to new_size permanently on
	 * disk.  This way we don't have to worry about anyone ever being able
	 * to look at the data being freed even in the face of a crash.
	 * What we're getting around here is the case where we free a block, it
	 * is allocated to another file, it is written to, and then we crash.
	 * If the new data gets written to the file but the log buffers
	 * containing the free and reallocation don't, then we'd end up with
	 * garbage in the blocks being freed.  As long as we make the new size
	 * permanent before actually freeing any blocks it doesn't matter if
	 * they get written to.
	 */
	ip->i_d.di_size = newsize;
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

	if (newsize <= oldsize) {
		error = xfs_itruncate_extents(&tp, ip, XFS_DATA_FORK, newsize);
C
Christoph Hellwig 已提交
950
		if (error)
951
			goto out_trans_cancel;
C
Christoph Hellwig 已提交
952 953 954 955 956 957 958 959 960

		/*
		 * Truncated "down", so we're removing references to old data
		 * here - if we delay flushing for a long time, we expose
		 * ourselves unduly to the notorious NULL files problem.  So,
		 * we mark this inode and flush it when the file is closed,
		 * and do not wait the usual (long) time for writeout.
		 */
		xfs_iflags_set(ip, XFS_ITRUNCATED);
961 962 963

		/* A truncate down always removes post-EOF blocks. */
		xfs_inode_clear_eofblocks_tag(ip);
C
Christoph Hellwig 已提交
964 965
	}

966
	if (iattr->ia_valid & ATTR_MODE)
967
		xfs_setattr_mode(ip, iattr);
968
	if (iattr->ia_valid & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME))
C
Christoph Hellwig 已提交
969
		xfs_setattr_time(ip, iattr);
C
Christoph Hellwig 已提交
970 971 972

	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

973
	XFS_STATS_INC(mp, xs_ig_attrchg);
C
Christoph Hellwig 已提交
974 975 976 977

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

978
	error = xfs_trans_commit(tp);
C
Christoph Hellwig 已提交
979 980 981 982 983 984
out_unlock:
	if (lock_flags)
		xfs_iunlock(ip, lock_flags);
	return error;

out_trans_cancel:
985
	xfs_trans_cancel(tp);
C
Christoph Hellwig 已提交
986 987 988
	goto out_unlock;
}

989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
int
xfs_vn_setattr_size(
	struct dentry		*dentry,
	struct iattr		*iattr)
{
	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
	int error;

	trace_xfs_setattr(ip);

	error = xfs_vn_change_ok(dentry, iattr);
	if (error)
		return error;
	return xfs_setattr_size(ip, iattr);
}

L
Linus Torvalds 已提交
1005
STATIC int
1006
xfs_vn_setattr(
1007 1008
	struct dentry		*dentry,
	struct iattr		*iattr)
L
Linus Torvalds 已提交
1009
{
1010 1011 1012
	int			error;

	if (iattr->ia_valid & ATTR_SIZE) {
1013 1014
		struct xfs_inode	*ip = XFS_I(d_inode(dentry));
		uint			iolock = XFS_IOLOCK_EXCL;
1015

1016 1017 1018
		error = xfs_break_layouts(d_inode(dentry), &iolock);
		if (error)
			return error;
1019

1020
		xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
1021
		error = xfs_vn_setattr_size(dentry, iattr);
1022
		xfs_iunlock(ip, XFS_MMAPLOCK_EXCL);
1023
	} else {
1024
		error = xfs_vn_setattr_nonsize(dentry, iattr);
1025 1026
	}

D
Dave Chinner 已提交
1027
	return error;
L
Linus Torvalds 已提交
1028 1029
}

C
Christoph Hellwig 已提交
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
STATIC int
xfs_vn_update_time(
	struct inode		*inode,
	struct timespec		*now,
	int			flags)
{
	struct xfs_inode	*ip = XFS_I(inode);
	struct xfs_mount	*mp = ip->i_mount;
	struct xfs_trans	*tp;
	int			error;

	trace_xfs_update_time(ip);

1043 1044
	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp);
	if (error)
D
Dave Chinner 已提交
1045
		return error;
C
Christoph Hellwig 已提交
1046 1047

	xfs_ilock(ip, XFS_ILOCK_EXCL);
1048
	if (flags & S_CTIME)
C
Christoph Hellwig 已提交
1049
		inode->i_ctime = *now;
1050
	if (flags & S_MTIME)
C
Christoph Hellwig 已提交
1051
		inode->i_mtime = *now;
1052
	if (flags & S_ATIME)
C
Christoph Hellwig 已提交
1053
		inode->i_atime = *now;
1054

C
Christoph Hellwig 已提交
1055 1056
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP);
1057
	return xfs_trans_commit(tp);
C
Christoph Hellwig 已提交
1058 1059
}

E
Eric Sandeen 已提交
1060 1061 1062 1063 1064 1065 1066 1067 1068
STATIC int
xfs_vn_fiemap(
	struct inode		*inode,
	struct fiemap_extent_info *fieinfo,
	u64			start,
	u64			length)
{
	int			error;

1069
	xfs_ilock(XFS_I(inode), XFS_IOLOCK_SHARED);
1070 1071 1072 1073 1074 1075 1076 1077
	if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
		fieinfo->fi_flags &= ~FIEMAP_FLAG_XATTR;
		error = iomap_fiemap(inode, fieinfo, start, length,
				&xfs_xattr_iomap_ops);
	} else {
		error = iomap_fiemap(inode, fieinfo, start, length,
				&xfs_iomap_ops);
	}
1078
	xfs_iunlock(XFS_I(inode), XFS_IOLOCK_SHARED);
E
Eric Sandeen 已提交
1079

1080
	return error;
E
Eric Sandeen 已提交
1081 1082
}

Z
Zhi Yong Wu 已提交
1083 1084 1085 1086 1087 1088
STATIC int
xfs_vn_tmpfile(
	struct inode	*dir,
	struct dentry	*dentry,
	umode_t		mode)
{
1089
	return xfs_generic_create(dir, dentry, mode, 0, true);
Z
Zhi Yong Wu 已提交
1090 1091
}

1092
static const struct inode_operations xfs_inode_operations = {
1093
	.get_acl		= xfs_get_acl,
1094
	.set_acl		= xfs_set_acl,
1095 1096 1097
	.getattr		= xfs_vn_getattr,
	.setattr		= xfs_vn_setattr,
	.listxattr		= xfs_vn_listxattr,
E
Eric Sandeen 已提交
1098
	.fiemap			= xfs_vn_fiemap,
C
Christoph Hellwig 已提交
1099
	.update_time		= xfs_vn_update_time,
L
Linus Torvalds 已提交
1100 1101
};

1102
static const struct inode_operations xfs_dir_inode_operations = {
1103 1104 1105 1106 1107 1108
	.create			= xfs_vn_create,
	.lookup			= xfs_vn_lookup,
	.link			= xfs_vn_link,
	.unlink			= xfs_vn_unlink,
	.symlink		= xfs_vn_symlink,
	.mkdir			= xfs_vn_mkdir,
1109 1110 1111 1112 1113 1114 1115
	/*
	 * Yes, XFS uses the same method for rmdir and unlink.
	 *
	 * There are some subtile differences deeper in the code,
	 * but we use S_ISDIR to check for those.
	 */
	.rmdir			= xfs_vn_unlink,
1116
	.mknod			= xfs_vn_mknod,
1117
	.rename			= xfs_vn_rename,
1118
	.get_acl		= xfs_get_acl,
1119
	.set_acl		= xfs_set_acl,
1120 1121 1122
	.getattr		= xfs_vn_getattr,
	.setattr		= xfs_vn_setattr,
	.listxattr		= xfs_vn_listxattr,
C
Christoph Hellwig 已提交
1123
	.update_time		= xfs_vn_update_time,
Z
Zhi Yong Wu 已提交
1124
	.tmpfile		= xfs_vn_tmpfile,
L
Linus Torvalds 已提交
1125 1126
};

1127
static const struct inode_operations xfs_dir_ci_inode_operations = {
1128 1129 1130 1131 1132 1133
	.create			= xfs_vn_create,
	.lookup			= xfs_vn_ci_lookup,
	.link			= xfs_vn_link,
	.unlink			= xfs_vn_unlink,
	.symlink		= xfs_vn_symlink,
	.mkdir			= xfs_vn_mkdir,
1134 1135 1136 1137 1138 1139 1140
	/*
	 * Yes, XFS uses the same method for rmdir and unlink.
	 *
	 * There are some subtile differences deeper in the code,
	 * but we use S_ISDIR to check for those.
	 */
	.rmdir			= xfs_vn_unlink,
1141
	.mknod			= xfs_vn_mknod,
1142
	.rename			= xfs_vn_rename,
1143
	.get_acl		= xfs_get_acl,
1144
	.set_acl		= xfs_set_acl,
1145 1146 1147
	.getattr		= xfs_vn_getattr,
	.setattr		= xfs_vn_setattr,
	.listxattr		= xfs_vn_listxattr,
C
Christoph Hellwig 已提交
1148
	.update_time		= xfs_vn_update_time,
Z
Zhi Yong Wu 已提交
1149
	.tmpfile		= xfs_vn_tmpfile,
1150 1151
};

1152
static const struct inode_operations xfs_symlink_inode_operations = {
1153
	.get_link		= xfs_vn_get_link,
1154 1155 1156
	.getattr		= xfs_vn_getattr,
	.setattr		= xfs_vn_setattr,
	.listxattr		= xfs_vn_listxattr,
C
Christoph Hellwig 已提交
1157
	.update_time		= xfs_vn_update_time,
L
Linus Torvalds 已提交
1158
};
1159

1160 1161 1162 1163 1164 1165 1166 1167
static const struct inode_operations xfs_inline_symlink_inode_operations = {
	.get_link		= xfs_vn_get_link_inline,
	.getattr		= xfs_vn_getattr,
	.setattr		= xfs_vn_setattr,
	.listxattr		= xfs_vn_listxattr,
	.update_time		= xfs_vn_update_time,
};

1168 1169 1170 1171 1172
STATIC void
xfs_diflags_to_iflags(
	struct inode		*inode,
	struct xfs_inode	*ip)
{
D
Dave Chinner 已提交
1173 1174 1175 1176 1177 1178
	uint16_t		flags = ip->i_d.di_flags;

	inode->i_flags &= ~(S_IMMUTABLE | S_APPEND | S_SYNC |
			    S_NOATIME | S_DAX);

	if (flags & XFS_DIFLAG_IMMUTABLE)
1179
		inode->i_flags |= S_IMMUTABLE;
D
Dave Chinner 已提交
1180
	if (flags & XFS_DIFLAG_APPEND)
1181
		inode->i_flags |= S_APPEND;
D
Dave Chinner 已提交
1182
	if (flags & XFS_DIFLAG_SYNC)
1183
		inode->i_flags |= S_SYNC;
D
Dave Chinner 已提交
1184
	if (flags & XFS_DIFLAG_NOATIME)
1185
		inode->i_flags |= S_NOATIME;
1186
	if (S_ISREG(inode->i_mode) &&
1187
	    ip->i_mount->m_sb.sb_blocksize == PAGE_SIZE &&
1188
	    !xfs_is_reflink_inode(ip) &&
1189 1190
	    (ip->i_mount->m_flags & XFS_MOUNT_DAX ||
	     ip->i_d.di_flags2 & XFS_DIFLAG2_DAX))
D
Dave Chinner 已提交
1191
		inode->i_flags |= S_DAX;
1192 1193 1194
}

/*
1195
 * Initialize the Linux inode.
1196
 *
1197 1198 1199 1200
 * When reading existing inodes from disk this is called directly from xfs_iget,
 * when creating a new inode it is called from xfs_ialloc after setting up the
 * inode. These callers have different criteria for clearing XFS_INEW, so leave
 * it up to the caller to deal with unlocking the inode appropriately.
1201 1202 1203 1204 1205
 */
void
xfs_setup_inode(
	struct xfs_inode	*ip)
{
1206
	struct inode		*inode = &ip->i_vnode;
1207
	gfp_t			gfp_mask;
1208 1209

	inode->i_ino = ip->i_ino;
C
Christoph Hellwig 已提交
1210
	inode->i_state = I_NEW;
1211 1212

	inode_sb_list_add(inode);
C
Christoph Hellwig 已提交
1213 1214
	/* make the inode look hashed for the writeback code */
	hlist_add_fake(&inode->i_hash);
1215

1216 1217
	inode->i_uid    = xfs_uid_to_kuid(ip->i_d.di_uid);
	inode->i_gid    = xfs_gid_to_kgid(ip->i_d.di_gid);
1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233

	switch (inode->i_mode & S_IFMT) {
	case S_IFBLK:
	case S_IFCHR:
		inode->i_rdev =
			MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
			      sysv_minor(ip->i_df.if_u2.if_rdev));
		break;
	default:
		inode->i_rdev = 0;
		break;
	}

	i_size_write(inode, ip->i_d.di_size);
	xfs_diflags_to_iflags(inode, ip);

1234
	if (S_ISDIR(inode->i_mode)) {
1235
		lockdep_set_class(&ip->i_lock.mr_lock, &xfs_dir_ilock_class);
1236
		ip->d_ops = ip->i_mount->m_dir_inode_ops;
1237 1238 1239
	} else {
		ip->d_ops = ip->i_mount->m_nondir_inode_ops;
		lockdep_set_class(&ip->i_lock.mr_lock, &xfs_nondir_ilock_class);
1240 1241
	}

1242 1243 1244 1245 1246 1247 1248 1249
	/*
	 * Ensure all page cache allocations are done from GFP_NOFS context to
	 * prevent direct reclaim recursion back into the filesystem and blowing
	 * stacks or deadlocking.
	 */
	gfp_mask = mapping_gfp_mask(inode->i_mapping);
	mapping_set_gfp_mask(inode->i_mapping, (gfp_mask & ~(__GFP_FS)));

1250 1251 1252 1253 1254 1255
	/*
	 * If there is no attribute fork no ACL can exist on this inode,
	 * and it can't have any file capabilities attached to it either.
	 */
	if (!XFS_IFORK_Q(ip)) {
		inode_has_no_xattr(inode);
1256
		cache_no_acl(inode);
1257
	}
1258
}
1259 1260 1261 1262 1263 1264 1265

void
xfs_setup_iops(
	struct xfs_inode	*ip)
{
	struct inode		*inode = &ip->i_vnode;

1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279
	switch (inode->i_mode & S_IFMT) {
	case S_IFREG:
		inode->i_op = &xfs_inode_operations;
		inode->i_fop = &xfs_file_operations;
		inode->i_mapping->a_ops = &xfs_address_space_operations;
		break;
	case S_IFDIR:
		if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb))
			inode->i_op = &xfs_dir_ci_inode_operations;
		else
			inode->i_op = &xfs_dir_inode_operations;
		inode->i_fop = &xfs_dir_file_operations;
		break;
	case S_IFLNK:
1280 1281 1282 1283
		if (ip->i_df.if_flags & XFS_IFINLINE)
			inode->i_op = &xfs_inline_symlink_inode_operations;
		else
			inode->i_op = &xfs_symlink_inode_operations;
1284 1285 1286 1287 1288 1289 1290
		break;
	default:
		inode->i_op = &xfs_inode_operations;
		init_special_inode(inode, inode->i_mode, inode->i_rdev);
		break;
	}
}