xfs_vnodeops.c 85.2 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
3
 * All Rights Reserved.
L
Linus Torvalds 已提交
4
 *
5 6
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
L
Linus Torvalds 已提交
7 8
 * published by the Free Software Foundation.
 *
9 10 11 12
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
L
Linus Torvalds 已提交
13
 *
14 15 16
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
L
Linus Torvalds 已提交
17
 */
18

L
Linus Torvalds 已提交
19
#include "xfs.h"
20
#include "xfs_fs.h"
L
Linus Torvalds 已提交
21
#include "xfs_types.h"
22
#include "xfs_bit.h"
L
Linus Torvalds 已提交
23
#include "xfs_log.h"
24
#include "xfs_inum.h"
L
Linus Torvalds 已提交
25 26 27 28 29 30
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
31
#include "xfs_da_btree.h"
L
Linus Torvalds 已提交
32
#include "xfs_bmap_btree.h"
33
#include "xfs_alloc_btree.h"
L
Linus Torvalds 已提交
34 35
#include "xfs_ialloc_btree.h"
#include "xfs_dir2_sf.h"
36
#include "xfs_attr_sf.h"
L
Linus Torvalds 已提交
37 38
#include "xfs_dinode.h"
#include "xfs_inode.h"
39 40 41 42 43
#include "xfs_inode_item.h"
#include "xfs_itable.h"
#include "xfs_btree.h"
#include "xfs_ialloc.h"
#include "xfs_alloc.h"
L
Linus Torvalds 已提交
44 45 46 47 48 49
#include "xfs_bmap.h"
#include "xfs_attr.h"
#include "xfs_rw.h"
#include "xfs_error.h"
#include "xfs_quota.h"
#include "xfs_utils.h"
50
#include "xfs_rtalloc.h"
L
Linus Torvalds 已提交
51 52
#include "xfs_trans_space.h"
#include "xfs_log_priv.h"
53
#include "xfs_filestream.h"
54
#include "xfs_vnodeops.h"
L
Linus Torvalds 已提交
55 56 57

int
xfs_setattr(
58 59
	struct xfs_inode	*ip,
	struct iattr		*iattr,
C
Christoph Hellwig 已提交
60
	int			flags)
L
Linus Torvalds 已提交
61
{
62
	xfs_mount_t		*mp = ip->i_mount;
63
	struct inode		*inode = VFS_I(ip);
64
	int			mask = iattr->ia_valid;
L
Linus Torvalds 已提交
65 66 67 68 69 70 71 72 73
	xfs_trans_t		*tp;
	int			code;
	uint			lock_flags;
	uint			commit_flags=0;
	uid_t			uid=0, iuid=0;
	gid_t			gid=0, igid=0;
	int			timeflags = 0;
	struct xfs_dquot	*udqp, *gdqp, *olddquot1, *olddquot2;
	int			file_owner;
74
	int			need_iolock = 1;
L
Linus Torvalds 已提交
75

76
	xfs_itrace_entry(ip);
L
Linus Torvalds 已提交
77

78
	if (mp->m_flags & XFS_MOUNT_RDONLY)
L
Linus Torvalds 已提交
79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94
		return XFS_ERROR(EROFS);

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

	olddquot1 = olddquot2 = NULL;
	udqp = gdqp = NULL;

	/*
	 * 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.
	 */
95
	if (XFS_IS_QUOTA_ON(mp) && (mask & (ATTR_UID|ATTR_GID))) {
L
Linus Torvalds 已提交
96 97
		uint	qflags = 0;

98 99
		if ((mask & ATTR_UID) && XFS_IS_UQUOTA_ON(mp)) {
			uid = iattr->ia_uid;
L
Linus Torvalds 已提交
100 101 102 103
			qflags |= XFS_QMOPT_UQUOTA;
		} else {
			uid = ip->i_d.di_uid;
		}
104 105
		if ((mask & ATTR_GID) && XFS_IS_GQUOTA_ON(mp)) {
			gid = iattr->ia_gid;
L
Linus Torvalds 已提交
106 107 108 109
			qflags |= XFS_QMOPT_GQUOTA;
		}  else {
			gid = ip->i_d.di_gid;
		}
110

L
Linus Torvalds 已提交
111 112 113 114 115 116 117
		/*
		 * 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);
118 119
		code = XFS_QM_DQVOPALLOC(mp, ip, uid, gid, ip->i_d.di_projid,
					 qflags, &udqp, &gdqp);
L
Linus Torvalds 已提交
120
		if (code)
121
			return code;
L
Linus Torvalds 已提交
122 123 124 125 126 127 128 129
	}

	/*
	 * For the other attributes, we acquire the inode lock and
	 * first do an error checking pass.
	 */
	tp = NULL;
	lock_flags = XFS_ILOCK_EXCL;
130
	if (flags & XFS_ATTR_NOLOCK)
131
		need_iolock = 0;
132 133
	if (!(mask & ATTR_SIZE)) {
		if ((mask != (ATTR_CTIME|ATTR_ATIME|ATTR_MTIME)) ||
L
Linus Torvalds 已提交
134 135 136 137 138 139 140 141 142 143 144
		    (mp->m_flags & XFS_MOUNT_WSYNC)) {
			tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
			commit_flags = 0;
			if ((code = xfs_trans_reserve(tp, 0,
						     XFS_ICHANGE_LOG_RES(mp), 0,
						     0, 0))) {
				lock_flags = 0;
				goto error_return;
			}
		}
	} else {
145
		if (DM_EVENT_ENABLED(ip, DM_EVENT_TRUNCATE) &&
146
		    !(flags & XFS_ATTR_DMI)) {
L
Linus Torvalds 已提交
147
			int dmflags = AT_DELAY_FLAG(flags) | DM_SEM_FLAG_WR;
148
			code = XFS_SEND_DATA(mp, DM_EVENT_TRUNCATE, ip,
149
				iattr->ia_size, 0, dmflags, NULL);
L
Linus Torvalds 已提交
150 151 152 153 154 155 156 157 158 159 160 161
			if (code) {
				lock_flags = 0;
				goto error_return;
			}
		}
		if (need_iolock)
			lock_flags |= XFS_IOLOCK_EXCL;
	}

	xfs_ilock(ip, lock_flags);

	/* boolean: are we the file owner? */
162
	file_owner = (current_fsuid() == ip->i_d.di_uid);
L
Linus Torvalds 已提交
163 164 165 166 167 168

	/*
	 * Change various properties of a file.
	 * Only the owner or users with CAP_FOWNER
	 * capability may do these things.
	 */
169
	if (mask & (ATTR_MODE|ATTR_UID|ATTR_GID)) {
L
Linus Torvalds 已提交
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192
		/*
		 * CAP_FOWNER overrides the following restrictions:
		 *
		 * The user ID of the calling process must be equal
		 * to the file owner ID, except in cases where the
		 * CAP_FSETID capability is applicable.
		 */
		if (!file_owner && !capable(CAP_FOWNER)) {
			code = XFS_ERROR(EPERM);
			goto error_return;
		}

		/*
		 * CAP_FSETID overrides the following restrictions:
		 *
		 * The effective user ID of the calling process shall match
		 * the file owner when setting the set-user-ID and
		 * set-group-ID bits on that file.
		 *
		 * The effective group ID or one of the supplementary group
		 * IDs of the calling process shall match the group owner of
		 * the file when setting the set-group-ID bit on that file
		 */
193
		if (mask & ATTR_MODE) {
L
Linus Torvalds 已提交
194 195
			mode_t m = 0;

196
			if ((iattr->ia_mode & S_ISUID) && !file_owner)
L
Linus Torvalds 已提交
197
				m |= S_ISUID;
198
			if ((iattr->ia_mode & S_ISGID) &&
L
Linus Torvalds 已提交
199 200 201 202
			    !in_group_p((gid_t)ip->i_d.di_gid))
				m |= S_ISGID;
#if 0
			/* Linux allows this, Irix doesn't. */
203
			if ((iattr->ia_mode & S_ISVTX) && !S_ISDIR(ip->i_d.di_mode))
L
Linus Torvalds 已提交
204 205 206
				m |= S_ISVTX;
#endif
			if (m && !capable(CAP_FSETID))
207
				iattr->ia_mode &= ~m;
L
Linus Torvalds 已提交
208 209 210 211 212 213
		}
	}

	/*
	 * Change file ownership.  Must be the owner or privileged.
	 */
214
	if (mask & (ATTR_UID|ATTR_GID)) {
L
Linus Torvalds 已提交
215 216 217 218 219 220 221 222
		/*
		 * 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.
		 */
		iuid = ip->i_d.di_uid;
		igid = ip->i_d.di_gid;
223 224
		gid = (mask & ATTR_GID) ? iattr->ia_gid : igid;
		uid = (mask & ATTR_UID) ? iattr->ia_uid : iuid;
L
Linus Torvalds 已提交
225 226 227 228 229 230 231 232 233 234 235

		/*
		 * CAP_CHOWN overrides the following restrictions:
		 *
		 * If _POSIX_CHOWN_RESTRICTED is defined, this capability
		 * shall override the restriction that a process cannot
		 * change the user ID of a file it owns and the restriction
		 * that the group ID supplied to the chown() function
		 * shall be equal to either the group ID or one of the
		 * supplementary group IDs of the calling process.
		 */
236 237
		if ((iuid != uid ||
		     (igid != gid && !in_group_p((gid_t)gid))) &&
L
Linus Torvalds 已提交
238 239 240 241 242
		    !capable(CAP_CHOWN)) {
			code = XFS_ERROR(EPERM);
			goto error_return;
		}
		/*
243
		 * Do a quota reservation only if uid/gid is actually
L
Linus Torvalds 已提交
244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259
		 * going to change.
		 */
		if ((XFS_IS_UQUOTA_ON(mp) && iuid != uid) ||
		    (XFS_IS_GQUOTA_ON(mp) && igid != gid)) {
			ASSERT(tp);
			code = XFS_QM_DQVOPCHOWNRESV(mp, tp, ip, udqp, gdqp,
						capable(CAP_FOWNER) ?
						XFS_QMOPT_FORCE_RES : 0);
			if (code)	/* out of quota */
				goto error_return;
		}
	}

	/*
	 * Truncate file.  Must have write permission and not be a directory.
	 */
260
	if (mask & ATTR_SIZE) {
L
Linus Torvalds 已提交
261
		/* Short circuit the truncate case for zero length files */
262 263
		if (iattr->ia_size == 0 &&
		    ip->i_size == 0 && ip->i_d.di_nextents == 0) {
L
Linus Torvalds 已提交
264 265
			xfs_iunlock(ip, XFS_ILOCK_EXCL);
			lock_flags &= ~XFS_ILOCK_EXCL;
266
			if (mask & ATTR_CTIME)
L
Linus Torvalds 已提交
267 268 269 270 271
				xfs_ichgtime(ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
			code = 0;
			goto error_return;
		}

272
		if (S_ISDIR(ip->i_d.di_mode)) {
L
Linus Torvalds 已提交
273 274
			code = XFS_ERROR(EISDIR);
			goto error_return;
275
		} else if (!S_ISREG(ip->i_d.di_mode)) {
L
Linus Torvalds 已提交
276 277 278 279 280 281 282 283 284 285 286 287 288
			code = XFS_ERROR(EINVAL);
			goto error_return;
		}
		/*
		 * Make sure that the dquots are attached to the inode.
		 */
		if ((code = XFS_QM_DQATTACH(mp, ip, XFS_QMOPT_ILOCKED)))
			goto error_return;
	}

	/*
	 * Change file access or modified times.
	 */
289
	if (mask & (ATTR_ATIME|ATTR_MTIME)) {
L
Linus Torvalds 已提交
290
		if (!file_owner) {
291
			if ((mask & (ATTR_MTIME_SET|ATTR_ATIME_SET)) &&
L
Linus Torvalds 已提交
292 293 294 295 296 297 298 299 300
			    !capable(CAP_FOWNER)) {
				code = XFS_ERROR(EPERM);
				goto error_return;
			}
		}
	}

	/*
	 * Now we can make the changes.  Before we join the inode
301
	 * to the transaction, if ATTR_SIZE is set then take care of
L
Linus Torvalds 已提交
302 303 304 305 306
	 * the part of the truncation that must be done without the
	 * inode lock.  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.
	 */
307
	if (mask & ATTR_SIZE) {
L
Linus Torvalds 已提交
308
		code = 0;
309
		if (iattr->ia_size > ip->i_size) {
310 311 312 313 314 315
			/*
			 * Do the first part of growing a file: zero any data
			 * in the last block that is beyond the old EOF.  We
			 * need to do this before the inode is joined to the
			 * transaction to modify the i_size.
			 */
316
			code = xfs_zero_eof(ip, iattr->ia_size, ip->i_size);
317
		}
L
Linus Torvalds 已提交
318
		xfs_iunlock(ip, XFS_ILOCK_EXCL);
319 320 321 322 323 324 325 326 327 328 329 330 331 332

		/*
		 * 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 problem.
		 *
		 * Only flush from the on disk size to the smaller of the in
		 * memory file size or the new size as that's the range we
		 * really care about here and prevents waiting for other data
		 * not within the range we care about here.
		 */
		if (!code &&
333 334
		    ip->i_size != ip->i_d.di_size &&
		    iattr->ia_size > ip->i_d.di_size) {
335
			code = xfs_flush_pages(ip,
336
					ip->i_d.di_size, iattr->ia_size,
337 338 339 340
					XFS_B_ASYNC, FI_NONE);
		}

		/* wait for all I/O to complete */
341
		vn_iowait(ip);
342

L
Linus Torvalds 已提交
343
		if (!code)
344
			code = xfs_itruncate_data(ip, iattr->ia_size);
L
Linus Torvalds 已提交
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372
		if (code) {
			ASSERT(tp == NULL);
			lock_flags &= ~XFS_ILOCK_EXCL;
			ASSERT(lock_flags == XFS_IOLOCK_EXCL);
			goto error_return;
		}
		tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_SIZE);
		if ((code = xfs_trans_reserve(tp, 0,
					     XFS_ITRUNCATE_LOG_RES(mp), 0,
					     XFS_TRANS_PERM_LOG_RES,
					     XFS_ITRUNCATE_LOG_COUNT))) {
			xfs_trans_cancel(tp, 0);
			if (need_iolock)
				xfs_iunlock(ip, XFS_IOLOCK_EXCL);
			return code;
		}
		commit_flags = XFS_TRANS_RELEASE_LOG_RES;
		xfs_ilock(ip, XFS_ILOCK_EXCL);
	}

	if (tp) {
		xfs_trans_ijoin(tp, ip, lock_flags);
		xfs_trans_ihold(tp, ip);
	}

	/*
	 * Truncate file.  Must have write permission and not be a directory.
	 */
373
	if (mask & ATTR_SIZE) {
374 375 376 377 378 379
		/*
		 * 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.
		 */
380
		if (iattr->ia_size != ip->i_size || (mask & ATTR_CTIME))
381 382
			timeflags |= XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;

383 384 385 386
		if (iattr->ia_size > ip->i_size) {
			ip->i_d.di_size = iattr->ia_size;
			ip->i_size = iattr->ia_size;
			if (!(flags & XFS_ATTR_DMI))
387 388
				xfs_ichgtime(ip, XFS_ICHGTIME_CHG);
			xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
389 390
		} else if (iattr->ia_size <= ip->i_size ||
			   (iattr->ia_size == 0 && ip->i_d.di_nextents)) {
L
Linus Torvalds 已提交
391 392 393 394 395
			/*
			 * signal a sync transaction unless
			 * we're truncating an already unlinked
			 * file on a wsync filesystem
			 */
396
			code = xfs_itruncate_finish(&tp, ip, iattr->ia_size,
L
Linus Torvalds 已提交
397 398 399 400
					    XFS_DATA_FORK,
					    ((ip->i_d.di_nlink != 0 ||
					      !(mp->m_flags & XFS_MOUNT_WSYNC))
					     ? 1 : 0));
401
			if (code)
L
Linus Torvalds 已提交
402
				goto abort_return;
403 404 405 406 407 408 409 410
			/*
			 * Truncated "down", so we're removing references
			 * to old data here - if we now delay flushing for
			 * a long time, we expose ourselves unduly to the
			 * notorious NULL files problem.  So, we mark this
			 * vnode and flush it when the file is closed, and
			 * do not wait the usual (long) time for writeout.
			 */
411
			xfs_iflags_set(ip, XFS_ITRUNCATED);
L
Linus Torvalds 已提交
412 413 414 415 416 417
		}
	}

	/*
	 * Change file access modes.
	 */
418
	if (mask & ATTR_MODE) {
L
Linus Torvalds 已提交
419
		ip->i_d.di_mode &= S_IFMT;
420
		ip->i_d.di_mode |= iattr->ia_mode & ~S_IFMT;
L
Linus Torvalds 已提交
421

422 423 424
		inode->i_mode &= S_IFMT;
		inode->i_mode |= iattr->ia_mode & ~S_IFMT;

L
Linus Torvalds 已提交
425 426 427 428 429 430 431
		xfs_trans_log_inode (tp, ip, XFS_ILOG_CORE);
		timeflags |= XFS_ICHGTIME_CHG;
	}

	/*
	 * Change file ownership.  Must be the owner or privileged.
	 */
432
	if (mask & (ATTR_UID|ATTR_GID)) {
L
Linus Torvalds 已提交
433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449
		/*
		 * 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()
		 */
		if ((ip->i_d.di_mode & (S_ISUID|S_ISGID)) &&
		    !capable(CAP_FSETID)) {
			ip->i_d.di_mode &= ~(S_ISUID|S_ISGID);
		}

		/*
		 * Change the ownerships and register quota modifications
		 * in the transaction.
		 */
		if (iuid != uid) {
			if (XFS_IS_UQUOTA_ON(mp)) {
450
				ASSERT(mask & ATTR_UID);
L
Linus Torvalds 已提交
451 452 453 454 455
				ASSERT(udqp);
				olddquot1 = XFS_QM_DQVOPCHOWN(mp, tp, ip,
							&ip->i_udquot, udqp);
			}
			ip->i_d.di_uid = uid;
456
			inode->i_uid = uid;
L
Linus Torvalds 已提交
457 458 459
		}
		if (igid != gid) {
			if (XFS_IS_GQUOTA_ON(mp)) {
460
				ASSERT(!XFS_IS_PQUOTA_ON(mp));
461
				ASSERT(mask & ATTR_GID);
L
Linus Torvalds 已提交
462 463 464 465 466
				ASSERT(gdqp);
				olddquot2 = XFS_QM_DQVOPCHOWN(mp, tp, ip,
							&ip->i_gdquot, gdqp);
			}
			ip->i_d.di_gid = gid;
467
			inode->i_gid = gid;
L
Linus Torvalds 已提交
468 469 470 471 472 473 474 475 476 477
		}

		xfs_trans_log_inode (tp, ip, XFS_ILOG_CORE);
		timeflags |= XFS_ICHGTIME_CHG;
	}


	/*
	 * Change file access or modified times.
	 */
478 479
	if (mask & (ATTR_ATIME|ATTR_MTIME)) {
		if (mask & ATTR_ATIME) {
480
			inode->i_atime = iattr->ia_atime;
481 482
			ip->i_d.di_atime.t_sec = iattr->ia_atime.tv_sec;
			ip->i_d.di_atime.t_nsec = iattr->ia_atime.tv_nsec;
L
Linus Torvalds 已提交
483 484
			ip->i_update_core = 1;
		}
485
		if (mask & ATTR_MTIME) {
486
			inode->i_mtime = iattr->ia_mtime;
487 488
			ip->i_d.di_mtime.t_sec = iattr->ia_mtime.tv_sec;
			ip->i_d.di_mtime.t_nsec = iattr->ia_mtime.tv_nsec;
L
Linus Torvalds 已提交
489 490 491
			timeflags &= ~XFS_ICHGTIME_MOD;
			timeflags |= XFS_ICHGTIME_CHG;
		}
492
		if (tp && (mask & (ATTR_MTIME_SET|ATTR_ATIME_SET)))
L
Linus Torvalds 已提交
493 494 495 496
			xfs_trans_log_inode (tp, ip, XFS_ILOG_CORE);
	}

	/*
497
	 * Change file inode change time only if ATTR_CTIME set
L
Linus Torvalds 已提交
498 499 500
	 * AND we have been called by a DMI function.
	 */

501
	if ((flags & XFS_ATTR_DMI) && (mask & ATTR_CTIME)) {
502
		inode->i_ctime = iattr->ia_ctime;
503 504
		ip->i_d.di_ctime.t_sec = iattr->ia_ctime.tv_sec;
		ip->i_d.di_ctime.t_nsec = iattr->ia_ctime.tv_nsec;
L
Linus Torvalds 已提交
505 506 507 508 509 510 511 512
		ip->i_update_core = 1;
		timeflags &= ~XFS_ICHGTIME_CHG;
	}

	/*
	 * Send out timestamp changes that need to be set to the
	 * current time.  Not done when called by a DMI function.
	 */
513
	if (timeflags && !(flags & XFS_ATTR_DMI))
L
Linus Torvalds 已提交
514 515 516 517 518 519 520 521
		xfs_ichgtime(ip, timeflags);

	XFS_STATS_INC(xs_ig_attrchg);

	/*
	 * If this is a synchronous mount, make sure that the
	 * transaction goes to disk before returning to the user.
	 * This is slightly sub-optimal in that truncates require
522
	 * two sync transactions instead of one for wsync filesystems.
L
Linus Torvalds 已提交
523 524 525 526 527 528 529 530 531 532
	 * One for the truncate and one for the timestamps since we
	 * don't want to change the timestamps unless we're sure the
	 * truncate worked.  Truncates are less than 1% of the laddis
	 * mix so this probably isn't worth the trouble to optimize.
	 */
	code = 0;
	if (tp) {
		if (mp->m_flags & XFS_MOUNT_WSYNC)
			xfs_trans_set_sync(tp);

533
		code = xfs_trans_commit(tp, commit_flags);
L
Linus Torvalds 已提交
534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
	}

	xfs_iunlock(ip, lock_flags);

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

	if (code) {
		return code;
	}

550
	if (DM_EVENT_ENABLED(ip, DM_EVENT_ATTRIBUTE) &&
551
	    !(flags & XFS_ATTR_DMI)) {
552
		(void) XFS_SEND_NAMESP(mp, DM_EVENT_ATTRIBUTE, ip, DM_RIGHT_NULL,
L
Linus Torvalds 已提交
553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572
					NULL, DM_RIGHT_NULL, NULL, NULL,
					0, 0, AT_DELAY_FLAG(flags));
	}
	return 0;

 abort_return:
	commit_flags |= XFS_TRANS_ABORT;
	/* FALLTHROUGH */
 error_return:
	XFS_QM_DQRELE(mp, udqp);
	XFS_QM_DQRELE(mp, gdqp);
	if (tp) {
		xfs_trans_cancel(tp, commit_flags);
	}
	if (lock_flags != 0) {
		xfs_iunlock(ip, lock_flags);
	}
	return code;
}

573 574 575 576 577 578 579
/*
 * The maximum pathlen is 1024 bytes. Since the minimum file system
 * blocksize is 512 bytes, we can get a max of 2 extents back from
 * bmapi.
 */
#define SYMLINK_MAPS 2

580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626
STATIC int
xfs_readlink_bmap(
	xfs_inode_t	*ip,
	char		*link)
{
	xfs_mount_t	*mp = ip->i_mount;
	int		pathlen = ip->i_d.di_size;
	int             nmaps = SYMLINK_MAPS;
	xfs_bmbt_irec_t mval[SYMLINK_MAPS];
	xfs_daddr_t	d;
	int		byte_cnt;
	int		n;
	xfs_buf_t	*bp;
	int		error = 0;

	error = xfs_bmapi(NULL, ip, 0, XFS_B_TO_FSB(mp, pathlen), 0, NULL, 0,
			mval, &nmaps, NULL, NULL);
	if (error)
		goto out;

	for (n = 0; n < nmaps; n++) {
		d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
		byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);

		bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0);
		error = XFS_BUF_GETERROR(bp);
		if (error) {
			xfs_ioerror_alert("xfs_readlink",
				  ip->i_mount, bp, XFS_BUF_ADDR(bp));
			xfs_buf_relse(bp);
			goto out;
		}
		if (pathlen < byte_cnt)
			byte_cnt = pathlen;
		pathlen -= byte_cnt;

		memcpy(link, XFS_BUF_PTR(bp), byte_cnt);
		xfs_buf_relse(bp);
	}

	link[ip->i_d.di_size] = '\0';
	error = 0;

 out:
	return error;
}

627
int
L
Linus Torvalds 已提交
628
xfs_readlink(
629
	xfs_inode_t     *ip,
630
	char		*link)
L
Linus Torvalds 已提交
631
{
632
	xfs_mount_t	*mp = ip->i_mount;
L
Linus Torvalds 已提交
633 634 635
	int		pathlen;
	int		error = 0;

636
	xfs_itrace_entry(ip);
L
Linus Torvalds 已提交
637 638 639 640 641 642 643

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

	xfs_ilock(ip, XFS_ILOCK_SHARED);

	ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFLNK);
644
	ASSERT(ip->i_d.di_size <= MAXPATHLEN);
L
Linus Torvalds 已提交
645

646 647 648
	pathlen = ip->i_d.di_size;
	if (!pathlen)
		goto out;
L
Linus Torvalds 已提交
649 650

	if (ip->i_df.if_flags & XFS_IFINLINE) {
651 652 653 654
		memcpy(link, ip->i_df.if_u1.if_data, pathlen);
		link[pathlen] = '\0';
	} else {
		error = xfs_readlink_bmap(ip, link);
L
Linus Torvalds 已提交
655 656
	}

657
 out:
L
Linus Torvalds 已提交
658 659 660 661 662 663 664
	xfs_iunlock(ip, XFS_ILOCK_SHARED);
	return error;
}

/*
 * xfs_fsync
 *
D
David Chinner 已提交
665 666 667 668
 * This is called to sync the inode and its data out to disk.  We need to hold
 * the I/O lock while flushing the data, and the inode lock while flushing the
 * inode.  The inode lock CANNOT be held while flushing the data, so acquire
 * after we're done with that.
L
Linus Torvalds 已提交
669
 */
670
int
L
Linus Torvalds 已提交
671
xfs_fsync(
D
David Chinner 已提交
672
	xfs_inode_t	*ip)
L
Linus Torvalds 已提交
673 674 675
{
	xfs_trans_t	*tp;
	int		error;
676
	int		log_flushed = 0, changed = 1;
L
Linus Torvalds 已提交
677

678
	xfs_itrace_entry(ip);
L
Linus Torvalds 已提交
679 680 681 682

	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
		return XFS_ERROR(EIO);

D
David Chinner 已提交
683
	/* capture size updates in I/O completion before writing the inode. */
684
	error = xfs_wait_on_pages(ip, 0, -1);
D
David Chinner 已提交
685 686
	if (error)
		return XFS_ERROR(error);
687

L
Linus Torvalds 已提交
688
	/*
D
David Chinner 已提交
689 690 691 692 693 694
	 * We always need to make sure that the required inode state is safe on
	 * disk.  The vnode might be clean but we still might need to force the
	 * log because of committed transactions that haven't hit the disk yet.
	 * Likewise, there could be unflushed non-transactional changes to the
	 * inode core that have to go to disk and this requires us to issue
	 * a synchronous transaction to capture these changes correctly.
L
Linus Torvalds 已提交
695
	 *
D
David Chinner 已提交
696 697 698
	 * This code relies on the assumption that if the update_* fields
	 * of the inode are clear and the inode is unpinned then it is clean
	 * and no action is required.
L
Linus Torvalds 已提交
699 700 701
	 */
	xfs_ilock(ip, XFS_ILOCK_SHARED);

D
David Chinner 已提交
702
	if (!(ip->i_update_size || ip->i_update_core)) {
L
Linus Torvalds 已提交
703
		/*
D
David Chinner 已提交
704 705 706 707 708 709
		 * Timestamps/size haven't changed since last inode flush or
		 * inode transaction commit.  That means either nothing got
		 * written or a transaction committed which caught the updates.
		 * If the latter happened and the transaction hasn't hit the
		 * disk yet, the inode will be still be pinned.  If it is,
		 * force the log.
L
Linus Torvalds 已提交
710 711 712 713 714
		 */

		xfs_iunlock(ip, XFS_ILOCK_SHARED);

		if (xfs_ipincount(ip)) {
D
David Chinner 已提交
715 716
			error = _xfs_log_force(ip->i_mount, (xfs_lsn_t)0,
				      XFS_LOG_FORCE | XFS_LOG_SYNC,
717 718 719
				      &log_flushed);
		} else {
			/*
D
David Chinner 已提交
720 721
			 * If the inode is not pinned and nothing has changed
			 * we don't need to flush the cache.
722 723
			 */
			changed = 0;
L
Linus Torvalds 已提交
724 725 726
		}
	} else	{
		/*
D
David Chinner 已提交
727 728
		 * Kick off a transaction to log the inode core to get the
		 * updates.  The sync transaction will also force the log.
L
Linus Torvalds 已提交
729 730 731
		 */
		xfs_iunlock(ip, XFS_ILOCK_SHARED);
		tp = xfs_trans_alloc(ip->i_mount, XFS_TRANS_FSYNC_TS);
D
David Chinner 已提交
732 733 734
		error = xfs_trans_reserve(tp, 0,
				XFS_FSYNC_TS_LOG_RES(ip->i_mount), 0, 0, 0);
		if (error) {
L
Linus Torvalds 已提交
735 736 737 738 739 740
			xfs_trans_cancel(tp, 0);
			return error;
		}
		xfs_ilock(ip, XFS_ILOCK_EXCL);

		/*
D
David Chinner 已提交
741 742 743 744 745 746 747
		 * Note - it's possible that we might have pushed ourselves out
		 * of the way during trans_reserve which would flush the inode.
		 * But there's no guarantee that the inode buffer has actually
		 * gone out yet (it's delwri).	Plus the buffer could be pinned
		 * anyway if it's part of an inode in another recent
		 * transaction.	 So we play it safe and fire off the
		 * transaction anyway.
L
Linus Torvalds 已提交
748 749 750 751
		 */
		xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);
		xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
D
David Chinner 已提交
752
		xfs_trans_set_sync(tp);
753
		error = _xfs_trans_commit(tp, 0, &log_flushed);
L
Linus Torvalds 已提交
754 755 756

		xfs_iunlock(ip, XFS_ILOCK_EXCL);
	}
757 758 759 760 761 762 763 764 765 766 767

	if ((ip->i_mount->m_flags & XFS_MOUNT_BARRIER) && changed) {
		/*
		 * If the log write didn't issue an ordered tag we need
		 * to flush the disk cache for the data device now.
		 */
		if (!log_flushed)
			xfs_blkdev_issue_flush(ip->i_mount->m_ddev_targp);

		/*
		 * If this inode is on the RT dev we need to flush that
768
		 * cache as well.
769
		 */
770
		if (XFS_IS_REALTIME_INODE(ip))
771 772 773
			xfs_blkdev_issue_flush(ip->i_mount->m_rtdev_targp);
	}

L
Linus Torvalds 已提交
774 775 776 777
	return error;
}

/*
778 779 780
 * This is called by xfs_inactive to free any blocks beyond eof
 * when the link count isn't zero and by xfs_dm_punch_hole() when
 * punching a hole to EOF.
L
Linus Torvalds 已提交
781
 */
782 783
int
xfs_free_eofblocks(
L
Linus Torvalds 已提交
784
	xfs_mount_t	*mp,
785 786
	xfs_inode_t	*ip,
	int		flags)
L
Linus Torvalds 已提交
787 788 789 790 791 792 793 794
{
	xfs_trans_t	*tp;
	int		error;
	xfs_fileoff_t	end_fsb;
	xfs_fileoff_t	last_fsb;
	xfs_filblks_t	map_len;
	int		nimaps;
	xfs_bmbt_irec_t	imap;
795
	int		use_iolock = (flags & XFS_FREE_EOF_LOCK);
L
Linus Torvalds 已提交
796 797 798 799 800

	/*
	 * Figure out if there are any blocks beyond the end
	 * of the file.  If not, then there is nothing to do.
	 */
801
	end_fsb = XFS_B_TO_FSB(mp, ((xfs_ufsize_t)ip->i_size));
L
Linus Torvalds 已提交
802 803 804
	last_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_MAXIOFFSET(mp));
	map_len = last_fsb - end_fsb;
	if (map_len <= 0)
805
		return 0;
L
Linus Torvalds 已提交
806 807 808

	nimaps = 1;
	xfs_ilock(ip, XFS_ILOCK_SHARED);
809
	error = xfs_bmapi(NULL, ip, end_fsb, map_len, 0,
810
			  NULL, 0, &imap, &nimaps, NULL, NULL);
L
Linus Torvalds 已提交
811 812 813
	xfs_iunlock(ip, XFS_ILOCK_SHARED);

	if (!error && (nimaps != 0) &&
814 815
	    (imap.br_startblock != HOLESTARTBLOCK ||
	     ip->i_delayed_blks)) {
L
Linus Torvalds 已提交
816 817 818 819
		/*
		 * Attach the dquots to the inode up front.
		 */
		if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
820
			return error;
L
Linus Torvalds 已提交
821 822 823 824 825 826 827 828 829 830 831 832 833 834 835

		/*
		 * There are blocks after the end of file.
		 * Free them up now by truncating the file to
		 * its current size.
		 */
		tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);

		/*
		 * Do the xfs_itruncate_start() call before
		 * reserving any log space because
		 * itruncate_start will call into the buffer
		 * cache and we can't
		 * do that within a transaction.
		 */
836 837
		if (use_iolock)
			xfs_ilock(ip, XFS_IOLOCK_EXCL);
838
		error = xfs_itruncate_start(ip, XFS_ITRUNC_DEFINITE,
839
				    ip->i_size);
840
		if (error) {
841
			xfs_trans_cancel(tp, 0);
842 843
			if (use_iolock)
				xfs_iunlock(ip, XFS_IOLOCK_EXCL);
844 845
			return error;
		}
L
Linus Torvalds 已提交
846 847 848 849 850 851 852 853 854

		error = xfs_trans_reserve(tp, 0,
					  XFS_ITRUNCATE_LOG_RES(mp),
					  0, XFS_TRANS_PERM_LOG_RES,
					  XFS_ITRUNCATE_LOG_COUNT);
		if (error) {
			ASSERT(XFS_FORCED_SHUTDOWN(mp));
			xfs_trans_cancel(tp, 0);
			xfs_iunlock(ip, XFS_IOLOCK_EXCL);
855
			return error;
L
Linus Torvalds 已提交
856 857 858 859 860 861 862 863 864
		}

		xfs_ilock(ip, XFS_ILOCK_EXCL);
		xfs_trans_ijoin(tp, ip,
				XFS_IOLOCK_EXCL |
				XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);

		error = xfs_itruncate_finish(&tp, ip,
865
					     ip->i_size,
L
Linus Torvalds 已提交
866 867 868 869 870 871 872 873 874 875 876 877
					     XFS_DATA_FORK,
					     0);
		/*
		 * If we get an error at this point we
		 * simply don't bother truncating the file.
		 */
		if (error) {
			xfs_trans_cancel(tp,
					 (XFS_TRANS_RELEASE_LOG_RES |
					  XFS_TRANS_ABORT));
		} else {
			error = xfs_trans_commit(tp,
878
						XFS_TRANS_RELEASE_LOG_RES);
L
Linus Torvalds 已提交
879
		}
880 881
		xfs_iunlock(ip, (use_iolock ? (XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)
					    : XFS_ILOCK_EXCL));
L
Linus Torvalds 已提交
882
	}
883
	return error;
L
Linus Torvalds 已提交
884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942
}

/*
 * Free a symlink that has blocks associated with it.
 */
STATIC int
xfs_inactive_symlink_rmt(
	xfs_inode_t	*ip,
	xfs_trans_t	**tpp)
{
	xfs_buf_t	*bp;
	int		committed;
	int		done;
	int		error;
	xfs_fsblock_t	first_block;
	xfs_bmap_free_t	free_list;
	int		i;
	xfs_mount_t	*mp;
	xfs_bmbt_irec_t	mval[SYMLINK_MAPS];
	int		nmaps;
	xfs_trans_t	*ntp;
	int		size;
	xfs_trans_t	*tp;

	tp = *tpp;
	mp = ip->i_mount;
	ASSERT(ip->i_d.di_size > XFS_IFORK_DSIZE(ip));
	/*
	 * We're freeing a symlink that has some
	 * blocks allocated to it.  Free the
	 * blocks here.  We know that we've got
	 * either 1 or 2 extents and that we can
	 * free them all in one bunmapi call.
	 */
	ASSERT(ip->i_d.di_nextents > 0 && ip->i_d.di_nextents <= 2);
	if ((error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0,
			XFS_TRANS_PERM_LOG_RES, XFS_ITRUNCATE_LOG_COUNT))) {
		ASSERT(XFS_FORCED_SHUTDOWN(mp));
		xfs_trans_cancel(tp, 0);
		*tpp = NULL;
		return error;
	}
	/*
	 * Lock the inode, fix the size, and join it to the transaction.
	 * Hold it so in the normal path, we still have it locked for
	 * the second transaction.  In the error paths we need it
	 * held so the cancel won't rele it, see below.
	 */
	xfs_ilock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
	size = (int)ip->i_d.di_size;
	ip->i_d.di_size = 0;
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
	xfs_trans_ihold(tp, ip);
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	/*
	 * Find the block(s) so we can inval and unmap them.
	 */
	done = 0;
	XFS_BMAP_INIT(&free_list, &first_block);
943
	nmaps = ARRAY_SIZE(mval);
L
Linus Torvalds 已提交
944 945
	if ((error = xfs_bmapi(tp, ip, 0, XFS_B_TO_FSB(mp, size),
			XFS_BMAPI_METADATA, &first_block, 0, mval, &nmaps,
946
			&free_list, NULL)))
L
Linus Torvalds 已提交
947 948 949 950 951 952 953 954 955 956 957 958 959 960
		goto error0;
	/*
	 * Invalidate the block(s).
	 */
	for (i = 0; i < nmaps; i++) {
		bp = xfs_trans_get_buf(tp, mp->m_ddev_targp,
			XFS_FSB_TO_DADDR(mp, mval[i].br_startblock),
			XFS_FSB_TO_BB(mp, mval[i].br_blockcount), 0);
		xfs_trans_binval(tp, bp);
	}
	/*
	 * Unmap the dead block(s) to the free_list.
	 */
	if ((error = xfs_bunmapi(tp, ip, 0, size, XFS_BMAPI_METADATA, nmaps,
961
			&first_block, &free_list, NULL, &done)))
L
Linus Torvalds 已提交
962 963 964 965 966
		goto error1;
	ASSERT(done);
	/*
	 * Commit the first transaction.  This logs the EFI and the inode.
	 */
967
	if ((error = xfs_bmap_finish(&tp, &free_list, &committed)))
L
Linus Torvalds 已提交
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987
		goto error1;
	/*
	 * The transaction must have been committed, since there were
	 * actually extents freed by xfs_bunmapi.  See xfs_bmap_finish.
	 * The new tp has the extent freeing and EFDs.
	 */
	ASSERT(committed);
	/*
	 * The first xact was committed, so add the inode to the new one.
	 * Mark it dirty so it will be logged and moved forward in the log as
	 * part of every commit.
	 */
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
	xfs_trans_ihold(tp, ip);
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	/*
	 * Get a new, empty transaction to return to our caller.
	 */
	ntp = xfs_trans_dup(tp);
	/*
988
	 * Commit the transaction containing extent freeing and EFDs.
L
Linus Torvalds 已提交
989 990 991 992
	 * If we get an error on the commit here or on the reserve below,
	 * we need to unlock the inode since the new transaction doesn't
	 * have the inode attached.
	 */
993
	error = xfs_trans_commit(tp, 0);
L
Linus Torvalds 已提交
994 995 996 997 998
	tp = ntp;
	if (error) {
		ASSERT(XFS_FORCED_SHUTDOWN(mp));
		goto error0;
	}
999 1000 1001 1002 1003 1004
	/*
	 * transaction commit worked ok so we can drop the extra ticket
	 * reference that we gained in xfs_trans_dup()
	 */
	xfs_log_ticket_put(tp->t_ticket);

L
Linus Torvalds 已提交
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 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 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
	/*
	 * Remove the memory for extent descriptions (just bookkeeping).
	 */
	if (ip->i_df.if_bytes)
		xfs_idata_realloc(ip, -ip->i_df.if_bytes, XFS_DATA_FORK);
	ASSERT(ip->i_df.if_bytes == 0);
	/*
	 * Put an itruncate log reservation in the new transaction
	 * for our caller.
	 */
	if ((error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0,
			XFS_TRANS_PERM_LOG_RES, XFS_ITRUNCATE_LOG_COUNT))) {
		ASSERT(XFS_FORCED_SHUTDOWN(mp));
		goto error0;
	}
	/*
	 * Return with the inode locked but not joined to the transaction.
	 */
	*tpp = tp;
	return 0;

 error1:
	xfs_bmap_cancel(&free_list);
 error0:
	/*
	 * Have to come here with the inode locked and either
	 * (held and in the transaction) or (not in the transaction).
	 * If the inode isn't held then cancel would iput it, but
	 * that's wrong since this is inactive and the vnode ref
	 * count is 0 already.
	 * Cancel won't do anything to the inode if held, but it still
	 * needs to be locked until the cancel is done, if it was
	 * joined to the transaction.
	 */
	xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
	xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
	*tpp = NULL;
	return error;

}

STATIC int
xfs_inactive_symlink_local(
	xfs_inode_t	*ip,
	xfs_trans_t	**tpp)
{
	int		error;

	ASSERT(ip->i_d.di_size <= XFS_IFORK_DSIZE(ip));
	/*
	 * We're freeing a symlink which fit into
	 * the inode.  Just free the memory used
	 * to hold the old symlink.
	 */
	error = xfs_trans_reserve(*tpp, 0,
				  XFS_ITRUNCATE_LOG_RES(ip->i_mount),
				  0, XFS_TRANS_PERM_LOG_RES,
				  XFS_ITRUNCATE_LOG_COUNT);

	if (error) {
		xfs_trans_cancel(*tpp, 0);
		*tpp = NULL;
1067
		return error;
L
Linus Torvalds 已提交
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
	}
	xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);

	/*
	 * Zero length symlinks _can_ exist.
	 */
	if (ip->i_df.if_bytes > 0) {
		xfs_idata_realloc(ip,
				  -(ip->i_df.if_bytes),
				  XFS_DATA_FORK);
		ASSERT(ip->i_df.if_bytes == 0);
	}
1080
	return 0;
L
Linus Torvalds 已提交
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
}

STATIC int
xfs_inactive_attrs(
	xfs_inode_t	*ip,
	xfs_trans_t	**tpp)
{
	xfs_trans_t	*tp;
	int		error;
	xfs_mount_t	*mp;

C
Christoph Hellwig 已提交
1092
	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
L
Linus Torvalds 已提交
1093 1094 1095
	tp = *tpp;
	mp = ip->i_mount;
	ASSERT(ip->i_d.di_forkoff != 0);
1096
	error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
L
Linus Torvalds 已提交
1097
	xfs_iunlock(ip, XFS_ILOCK_EXCL);
1098 1099
	if (error)
		goto error_unlock;
L
Linus Torvalds 已提交
1100 1101

	error = xfs_attr_inactive(ip);
1102 1103
	if (error)
		goto error_unlock;
L
Linus Torvalds 已提交
1104 1105 1106 1107 1108 1109

	tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);
	error = xfs_trans_reserve(tp, 0,
				  XFS_IFREE_LOG_RES(mp),
				  0, XFS_TRANS_PERM_LOG_RES,
				  XFS_INACTIVE_LOG_COUNT);
1110 1111
	if (error)
		goto error_cancel;
L
Linus Torvalds 已提交
1112 1113 1114 1115 1116 1117 1118 1119 1120

	xfs_ilock(ip, XFS_ILOCK_EXCL);
	xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
	xfs_trans_ihold(tp, ip);
	xfs_idestroy_fork(ip, XFS_ATTR_FORK);

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

	*tpp = tp;
1121
	return 0;
1122 1123 1124 1125 1126 1127 1128 1129

error_cancel:
	ASSERT(XFS_FORCED_SHUTDOWN(mp));
	xfs_trans_cancel(tp, 0);
error_unlock:
	*tpp = NULL;
	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	return error;
L
Linus Torvalds 已提交
1130 1131
}

1132
int
L
Linus Torvalds 已提交
1133
xfs_release(
1134
	xfs_inode_t	*ip)
L
Linus Torvalds 已提交
1135
{
1136
	xfs_mount_t	*mp = ip->i_mount;
L
Linus Torvalds 已提交
1137 1138
	int		error;

1139
	if (!S_ISREG(ip->i_d.di_mode) || (ip->i_d.di_mode == 0))
L
Linus Torvalds 已提交
1140 1141 1142
		return 0;

	/* If this is a read-only mount, don't do this (would generate I/O) */
1143
	if (mp->m_flags & XFS_MOUNT_RDONLY)
L
Linus Torvalds 已提交
1144 1145
		return 0;

1146
	if (!XFS_FORCED_SHUTDOWN(mp)) {
1147 1148
		int truncated;

1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
		/*
		 * If we are using filestreams, and we have an unlinked
		 * file that we are processing the last close on, then nothing
		 * will be able to reopen and write to this file. Purge this
		 * inode from the filestreams cache so that it doesn't delay
		 * teardown of the inode.
		 */
		if ((ip->i_d.di_nlink == 0) && xfs_inode_is_filestream(ip))
			xfs_filestream_deassociate(ip);

1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
		/*
		 * 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 noticable 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.
		 */
1169
		truncated = xfs_iflags_test_and_clear(ip, XFS_ITRUNCATED);
1170
		if (truncated && VN_DIRTY(VFS_I(ip)) && ip->i_delayed_blks > 0)
1171
			xfs_flush_pages(ip, 0, -1, XFS_B_ASYNC, FI_NONE);
1172 1173
	}

L
Linus Torvalds 已提交
1174 1175
	if (ip->i_d.di_nlink != 0) {
		if ((((ip->i_d.di_mode & S_IFMT) == S_IFREG) &&
1176
		     ((ip->i_size > 0) || (VN_CACHED(VFS_I(ip)) > 0 ||
1177
		       ip->i_delayed_blks > 0)) &&
L
Linus Torvalds 已提交
1178
		     (ip->i_df.if_flags & XFS_IFEXTENTS))  &&
1179 1180
		    (!(ip->i_d.di_flags &
				(XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)))) {
1181 1182
			error = xfs_free_eofblocks(mp, ip, XFS_FREE_EOF_LOCK);
			if (error)
1183
				return error;
L
Linus Torvalds 已提交
1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
		}
	}

	return 0;
}

/*
 * 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.
 */
1198
int
L
Linus Torvalds 已提交
1199
xfs_inactive(
1200
	xfs_inode_t	*ip)
L
Linus Torvalds 已提交
1201
{
1202
	xfs_bmap_free_t	free_list;
L
Linus Torvalds 已提交
1203 1204 1205 1206 1207 1208 1209
	xfs_fsblock_t	first_block;
	int		committed;
	xfs_trans_t	*tp;
	xfs_mount_t	*mp;
	int		error;
	int		truncate;

1210
	xfs_itrace_entry(ip);
L
Linus Torvalds 已提交
1211 1212 1213 1214 1215

	/*
	 * If the inode is already free, then there can be nothing
	 * to clean up here.
	 */
1216
	if (ip->i_d.di_mode == 0 || VN_BAD(VFS_I(ip))) {
L
Linus Torvalds 已提交
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
		ASSERT(ip->i_df.if_real_bytes == 0);
		ASSERT(ip->i_df.if_broot_bytes == 0);
		return VN_INACTIVE_CACHE;
	}

	/*
	 * Only do a truncate if it's a regular file with
	 * some actual space in it.  It's OK to look at the
	 * inode's fields without the lock because we're the
	 * only one with a reference to the inode.
	 */
	truncate = ((ip->i_d.di_nlink == 0) &&
1229 1230
	    ((ip->i_d.di_size != 0) || (ip->i_size != 0) ||
	     (ip->i_d.di_nextents > 0) || (ip->i_delayed_blks > 0)) &&
L
Linus Torvalds 已提交
1231 1232 1233 1234
	    ((ip->i_d.di_mode & S_IFMT) == S_IFREG));

	mp = ip->i_mount;

1235 1236
	if (ip->i_d.di_nlink == 0 && DM_EVENT_ENABLED(ip, DM_EVENT_DESTROY))
		XFS_SEND_DESTROY(mp, ip, DM_RIGHT_NULL);
L
Linus Torvalds 已提交
1237 1238 1239 1240

	error = 0;

	/* If this is a read-only mount, don't do this (would generate I/O) */
1241
	if (mp->m_flags & XFS_MOUNT_RDONLY)
L
Linus Torvalds 已提交
1242 1243 1244 1245
		goto out;

	if (ip->i_d.di_nlink != 0) {
		if ((((ip->i_d.di_mode & S_IFMT) == S_IFREG) &&
1246
                     ((ip->i_size > 0) || (VN_CACHED(VFS_I(ip)) > 0 ||
1247
                       ip->i_delayed_blks > 0)) &&
1248 1249 1250 1251
		      (ip->i_df.if_flags & XFS_IFEXTENTS) &&
		     (!(ip->i_d.di_flags &
				(XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)) ||
		      (ip->i_delayed_blks != 0)))) {
1252 1253
			error = xfs_free_eofblocks(mp, ip, XFS_FREE_EOF_LOCK);
			if (error)
1254
				return VN_INACTIVE_CACHE;
L
Linus Torvalds 已提交
1255 1256 1257 1258 1259 1260 1261
		}
		goto out;
	}

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

	if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
1262
		return VN_INACTIVE_CACHE;
L
Linus Torvalds 已提交
1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273

	tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);
	if (truncate) {
		/*
		 * Do the xfs_itruncate_start() call before
		 * reserving any log space because itruncate_start
		 * will call into the buffer cache and we can't
		 * do that within a transaction.
		 */
		xfs_ilock(ip, XFS_IOLOCK_EXCL);

1274 1275
		error = xfs_itruncate_start(ip, XFS_ITRUNC_DEFINITE, 0);
		if (error) {
1276
			xfs_trans_cancel(tp, 0);
1277 1278 1279
			xfs_iunlock(ip, XFS_IOLOCK_EXCL);
			return VN_INACTIVE_CACHE;
		}
L
Linus Torvalds 已提交
1280 1281 1282 1283 1284 1285 1286 1287 1288 1289

		error = xfs_trans_reserve(tp, 0,
					  XFS_ITRUNCATE_LOG_RES(mp),
					  0, XFS_TRANS_PERM_LOG_RES,
					  XFS_ITRUNCATE_LOG_COUNT);
		if (error) {
			/* Don't call itruncate_cleanup */
			ASSERT(XFS_FORCED_SHUTDOWN(mp));
			xfs_trans_cancel(tp, 0);
			xfs_iunlock(ip, XFS_IOLOCK_EXCL);
1290
			return VN_INACTIVE_CACHE;
L
Linus Torvalds 已提交
1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310
		}

		xfs_ilock(ip, XFS_ILOCK_EXCL);
		xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);

		/*
		 * normally, we have to run xfs_itruncate_finish sync.
		 * But if filesystem is wsync and we're in the inactive
		 * path, then we know that nlink == 0, and that the
		 * xaction that made nlink == 0 is permanently committed
		 * since xfs_remove runs as a synchronous transaction.
		 */
		error = xfs_itruncate_finish(&tp, ip, 0, XFS_DATA_FORK,
				(!(mp->m_flags & XFS_MOUNT_WSYNC) ? 1 : 0));

		if (error) {
			xfs_trans_cancel(tp,
				XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
			xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
1311
			return VN_INACTIVE_CACHE;
L
Linus Torvalds 已提交
1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
		}
	} else if ((ip->i_d.di_mode & S_IFMT) == S_IFLNK) {

		/*
		 * If we get an error while cleaning up a
		 * symlink we bail out.
		 */
		error = (ip->i_d.di_size > XFS_IFORK_DSIZE(ip)) ?
			xfs_inactive_symlink_rmt(ip, &tp) :
			xfs_inactive_symlink_local(ip, &tp);

		if (error) {
			ASSERT(tp == NULL);
1325
			return VN_INACTIVE_CACHE;
L
Linus Torvalds 已提交
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337
		}

		xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);
	} else {
		error = xfs_trans_reserve(tp, 0,
					  XFS_IFREE_LOG_RES(mp),
					  0, XFS_TRANS_PERM_LOG_RES,
					  XFS_INACTIVE_LOG_COUNT);
		if (error) {
			ASSERT(XFS_FORCED_SHUTDOWN(mp));
			xfs_trans_cancel(tp, 0);
1338
			return VN_INACTIVE_CACHE;
L
Linus Torvalds 已提交
1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
		}

		xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
		xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);
	}

	/*
	 * If there are attributes associated with the file
	 * then blow them away now.  The code calls a routine
	 * that recursively deconstructs the attribute fork.
	 * We need to just commit the current transaction
	 * because we can't use it for xfs_attr_inactive().
	 */
	if (ip->i_d.di_anextents > 0) {
		error = xfs_inactive_attrs(ip, &tp);
		/*
		 * If we got an error, the transaction is already
		 * cancelled, and the inode is unlocked. Just get out.
		 */
		 if (error)
1360
			 return VN_INACTIVE_CACHE;
L
Linus Torvalds 已提交
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
	} else if (ip->i_afp) {
		xfs_idestroy_fork(ip, XFS_ATTR_FORK);
	}

	/*
	 * Free the inode.
	 */
	XFS_BMAP_INIT(&free_list, &first_block);
	error = xfs_ifree(tp, ip, &free_list);
	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)) {
			cmn_err(CE_NOTE,
		"xfs_inactive:	xfs_ifree() returned an error = %d on %s",
				error, mp->m_fsname);
1380
			xfs_force_shutdown(mp, SHUTDOWN_META_IO_ERROR);
L
Linus Torvalds 已提交
1381 1382 1383 1384 1385 1386 1387 1388 1389
		}
		xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT);
	} else {
		/*
		 * Credit the quota account(s). The inode is gone.
		 */
		XFS_TRANS_MOD_DQUOT_BYINO(mp, tp, ip, XFS_TRANS_DQ_ICOUNT, -1);

		/*
1390 1391 1392
		 * 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.
L
Linus Torvalds 已提交
1393
		 */
1394 1395 1396 1397 1398 1399 1400 1401
		error = xfs_bmap_finish(&tp,  &free_list, &committed);
		if (error)
			xfs_fs_cmn_err(CE_NOTE, mp, "xfs_inactive: "
				"xfs_bmap_finish() returned error %d", error);
		error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
		if (error)
			xfs_fs_cmn_err(CE_NOTE, mp, "xfs_inactive: "
				"xfs_trans_commit() returned error %d", error);
L
Linus Torvalds 已提交
1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
	}
	/*
	 * Release the dquots held by inode, if any.
	 */
	XFS_QM_DQDETACH(mp, ip);

	xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);

 out:
	return VN_INACTIVE_CACHE;
}

1414 1415 1416 1417 1418 1419
/*
 * 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.
 */
1420
int
L
Linus Torvalds 已提交
1421
xfs_lookup(
1422
	xfs_inode_t		*dp,
1423
	struct xfs_name		*name,
1424 1425
	xfs_inode_t		**ipp,
	struct xfs_name		*ci_name)
L
Linus Torvalds 已提交
1426
{
C
Christoph Hellwig 已提交
1427
	xfs_ino_t		inum;
L
Linus Torvalds 已提交
1428 1429 1430
	int			error;
	uint			lock_mode;

1431
	xfs_itrace_entry(dp);
L
Linus Torvalds 已提交
1432 1433 1434 1435 1436

	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
		return XFS_ERROR(EIO);

	lock_mode = xfs_ilock_map_shared(dp);
1437
	error = xfs_dir_lookup(NULL, dp, name, &inum, ci_name);
L
Linus Torvalds 已提交
1438
	xfs_iunlock_map_shared(dp, lock_mode);
C
Christoph Hellwig 已提交
1439 1440 1441 1442 1443 1444

	if (error)
		goto out;

	error = xfs_iget(dp->i_mount, NULL, inum, 0, 0, ipp, 0);
	if (error)
1445
		goto out_free_name;
C
Christoph Hellwig 已提交
1446 1447 1448 1449

	xfs_itrace_ref(*ipp);
	return 0;

1450 1451 1452 1453
out_free_name:
	if (ci_name)
		kmem_free(ci_name->name);
out:
C
Christoph Hellwig 已提交
1454
	*ipp = NULL;
L
Linus Torvalds 已提交
1455 1456 1457
	return error;
}

1458
int
L
Linus Torvalds 已提交
1459
xfs_create(
1460
	xfs_inode_t		*dp,
1461
	struct xfs_name		*name,
1462 1463
	mode_t			mode,
	xfs_dev_t		rdev,
1464
	xfs_inode_t		**ipp,
L
Linus Torvalds 已提交
1465 1466
	cred_t			*credp)
{
1467
	xfs_mount_t		*mp = dp->i_mount;
1468
	xfs_inode_t		*ip;
L
Linus Torvalds 已提交
1469
	xfs_trans_t		*tp;
1470
	int			error;
L
Linus Torvalds 已提交
1471 1472
	xfs_bmap_free_t		free_list;
	xfs_fsblock_t		first_block;
1473
	boolean_t		unlock_dp_on_error = B_FALSE;
L
Linus Torvalds 已提交
1474 1475 1476 1477 1478 1479 1480
	int			dm_event_sent = 0;
	uint			cancel_flags;
	int			committed;
	xfs_prid_t		prid;
	struct xfs_dquot	*udqp, *gdqp;
	uint			resblks;

1481
	ASSERT(!*ipp);
1482
	xfs_itrace_entry(dp);
L
Linus Torvalds 已提交
1483

1484
	if (DM_EVENT_ENABLED(dp, DM_EVENT_CREATE)) {
L
Linus Torvalds 已提交
1485
		error = XFS_SEND_NAMESP(mp, DM_EVENT_CREATE,
1486
				dp, DM_RIGHT_NULL, NULL,
1487
				DM_RIGHT_NULL, name->name, NULL,
1488
				mode, 0, 0);
L
Linus Torvalds 已提交
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500

		if (error)
			return error;
		dm_event_sent = 1;
	}

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

	/* Return through std_return after this point. */

	udqp = gdqp = NULL;
1501 1502
	if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
		prid = dp->i_d.di_projid;
L
Linus Torvalds 已提交
1503 1504 1505 1506 1507 1508 1509
	else
		prid = (xfs_prid_t)dfltprid;

	/*
	 * Make sure that we have allocated dquot(s) on disk.
	 */
	error = XFS_QM_DQVOPALLOC(mp, dp,
1510
			current_fsuid(), current_fsgid(), prid,
L
Linus Torvalds 已提交
1511 1512 1513 1514 1515 1516 1517 1518
			XFS_QMOPT_QUOTALL|XFS_QMOPT_INHERIT, &udqp, &gdqp);
	if (error)
		goto std_return;

	ip = NULL;

	tp = xfs_trans_alloc(mp, XFS_TRANS_CREATE);
	cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
1519
	resblks = XFS_CREATE_SPACE_RES(mp, name->len);
L
Linus Torvalds 已提交
1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537
	/*
	 * 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.
	 */
	error = xfs_trans_reserve(tp, resblks, XFS_CREATE_LOG_RES(mp), 0,
			XFS_TRANS_PERM_LOG_RES, XFS_CREATE_LOG_COUNT);
	if (error == ENOSPC) {
		resblks = 0;
		error = xfs_trans_reserve(tp, 0, XFS_CREATE_LOG_RES(mp), 0,
				XFS_TRANS_PERM_LOG_RES, XFS_CREATE_LOG_COUNT);
	}
	if (error) {
		cancel_flags = 0;
		goto error_return;
	}

1538
	xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
1539
	unlock_dp_on_error = B_TRUE;
L
Linus Torvalds 已提交
1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551

	XFS_BMAP_INIT(&free_list, &first_block);

	ASSERT(ip == NULL);

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

1552 1553
	error = xfs_dir_canenter(tp, dp, name, resblks);
	if (error)
L
Linus Torvalds 已提交
1554
		goto error_return;
1555
	error = xfs_dir_ialloc(&tp, dp, mode, 1,
L
Linus Torvalds 已提交
1556 1557 1558 1559 1560 1561 1562
			rdev, credp, prid, resblks > 0,
			&ip, &committed);
	if (error) {
		if (error == ENOSPC)
			goto error_return;
		goto abort_return;
	}
1563
	xfs_itrace_ref(ip);
L
Linus Torvalds 已提交
1564 1565 1566 1567 1568 1569

	/*
	 * At this point, we've gotten a newly allocated inode.
	 * It is locked (and joined to the transaction).
	 */

C
Christoph Hellwig 已提交
1570
	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
L
Linus Torvalds 已提交
1571 1572

	/*
1573 1574 1575 1576 1577
	 * 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.
L
Linus Torvalds 已提交
1578
	 */
1579
	IHOLD(dp);
L
Linus Torvalds 已提交
1580
	xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
1581
	unlock_dp_on_error = B_FALSE;
L
Linus Torvalds 已提交
1582

1583
	error = xfs_dir_createname(tp, dp, name, ip->i_ino,
1584 1585
					&first_block, &free_list, resblks ?
					resblks - XFS_IALLOC_SPACE_RES(mp) : 0);
L
Linus Torvalds 已提交
1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
	if (error) {
		ASSERT(error != ENOSPC);
		goto abort_return;
	}
	xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);

	/*
	 * 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_DQVOPCREATE(mp, tp, ip, udqp, gdqp);

	/*
	 * xfs_trans_commit normally decrements the vnode ref count
	 * when it unlocks the inode. Since we want to return the
	 * vnode to the caller, we bump the vnode ref count now.
	 */
	IHOLD(ip);

1616
	error = xfs_bmap_finish(&tp, &free_list, &committed);
L
Linus Torvalds 已提交
1617 1618 1619 1620 1621
	if (error) {
		xfs_bmap_cancel(&free_list);
		goto abort_rele;
	}

1622
	error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
L
Linus Torvalds 已提交
1623 1624 1625 1626 1627 1628 1629 1630 1631
	if (error) {
		IRELE(ip);
		tp = NULL;
		goto error_return;
	}

	XFS_QM_DQRELE(mp, udqp);
	XFS_QM_DQRELE(mp, gdqp);

1632
	*ipp = ip;
L
Linus Torvalds 已提交
1633 1634 1635 1636

	/* Fallthrough to std_return with error = 0  */

std_return:
1637
	if ((*ipp || (error != 0 && dm_event_sent != 0)) &&
1638
	    DM_EVENT_ENABLED(dp, DM_EVENT_POSTCREATE)) {
L
Linus Torvalds 已提交
1639
		(void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTCREATE,
1640
			dp, DM_RIGHT_NULL,
1641
			*ipp ? ip : NULL,
1642
			DM_RIGHT_NULL, name->name, NULL,
1643
			mode, error, 0);
L
Linus Torvalds 已提交
1644 1645 1646 1647 1648 1649 1650
	}
	return error;

 abort_return:
	cancel_flags |= XFS_TRANS_ABORT;
	/* FALLTHROUGH */

1651
 error_return:
L
Linus Torvalds 已提交
1652 1653 1654 1655 1656 1657
	if (tp != NULL)
		xfs_trans_cancel(tp, cancel_flags);

	XFS_QM_DQRELE(mp, udqp);
	XFS_QM_DQRELE(mp, gdqp);

1658 1659 1660
	if (unlock_dp_on_error)
		xfs_iunlock(dp, XFS_ILOCK_EXCL);

L
Linus Torvalds 已提交
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
	goto std_return;

 abort_rele:
	/*
	 * Wait until after the current transaction is aborted to
	 * release the inode.  This prevents recursive transactions
	 * and deadlocks from xfs_inactive.
	 */
	cancel_flags |= XFS_TRANS_ABORT;
	xfs_trans_cancel(tp, cancel_flags);
	IRELE(ip);

	XFS_QM_DQRELE(mp, udqp);
	XFS_QM_DQRELE(mp, gdqp);

	goto std_return;
}

#ifdef DEBUG
int xfs_locked_n;
int xfs_small_retries;
int xfs_middle_retries;
int xfs_lots_retries;
int xfs_lock_delays;
#endif

1687 1688 1689 1690 1691 1692 1693 1694
/*
 * Bump the subclass so xfs_lock_inodes() acquires each lock with
 * a different value
 */
static inline int
xfs_lock_inumorder(int lock_mode, int subclass)
{
	if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL))
1695
		lock_mode |= (subclass + XFS_LOCK_INUMORDER) << XFS_IOLOCK_SHIFT;
1696
	if (lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL))
1697
		lock_mode |= (subclass + XFS_LOCK_INUMORDER) << XFS_ILOCK_SHIFT;
1698 1699 1700 1701

	return lock_mode;
}

L
Linus Torvalds 已提交
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723
/*
 * The following routine will lock n inodes in exclusive mode.
 * We assume the caller calls us with the inodes in i_ino order.
 *
 * 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.
 */
void
xfs_lock_inodes(
	xfs_inode_t	**ips,
	int		inodes,
	uint		lock_mode)
{
	int		attempts = 0, i, j, try_lock;
	xfs_log_item_t	*lp;

	ASSERT(ips && (inodes >= 2)); /* we need at least two */

1724 1725
	try_lock = 0;
	i = 0;
L
Linus Torvalds 已提交
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762

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

		if (i && (ips[i] == ips[i-1]))	/* Already locked */
			continue;

		/*
		 * 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.
		 */

		if (!try_lock) {
			for (j = (i - 1); j >= 0 && !try_lock; j--) {
				lp = (xfs_log_item_t *)ips[j]->i_itemp;
				if (lp && (lp->li_flags & XFS_LI_IN_AIL)) {
					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.
		 */

		if (try_lock) {
			/* try_lock must be 0 if i is 0. */
			/*
			 * try_lock means we have an inode locked
			 * that is in the AIL.
			 */
			ASSERT(i != 0);
1763
			if (!xfs_ilock_nowait(ips[i], xfs_lock_inumorder(lock_mode, i))) {
L
Linus Torvalds 已提交
1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797
				attempts++;

				/*
				 * Unlock all previous guys and try again.
				 * xfs_iunlock will try to push the tail
				 * if the inode is in the AIL.
				 */

				for(j = i - 1; j >= 0; j--) {

					/*
					 * Check to see if we've already
					 * unlocked this one.
					 * Not the first one going back,
					 * and the inode ptr is the same.
					 */
					if ((j != (i - 1)) && ips[j] ==
								ips[j+1])
						continue;

					xfs_iunlock(ips[j], lock_mode);
				}

				if ((attempts % 5) == 0) {
					delay(1); /* Don't just spin the CPU */
#ifdef DEBUG
					xfs_lock_delays++;
#endif
				}
				i = 0;
				try_lock = 0;
				goto again;
			}
		} else {
1798
			xfs_ilock(ips[i], xfs_lock_inumorder(lock_mode, i));
L
Linus Torvalds 已提交
1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
		}
	}

#ifdef DEBUG
	if (attempts) {
		if (attempts < 5) xfs_small_retries++;
		else if (attempts < 100) xfs_middle_retries++;
		else xfs_lots_retries++;
	} else {
		xfs_locked_n++;
	}
#endif
}

1813 1814 1815 1816 1817 1818
/*
 * xfs_lock_two_inodes() can only be used to lock one type of lock
 * at a time - the iolock or the ilock, but not both at once. If
 * we lock both at once, lockdep will report false positives saying
 * we have violated locking orders.
 */
1819 1820 1821 1822 1823 1824 1825 1826 1827 1828
void
xfs_lock_two_inodes(
	xfs_inode_t		*ip0,
	xfs_inode_t		*ip1,
	uint			lock_mode)
{
	xfs_inode_t		*temp;
	int			attempts = 0;
	xfs_log_item_t		*lp;

1829 1830
	if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL))
		ASSERT((lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)) == 0);
1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
	ASSERT(ip0->i_ino != ip1->i_ino);

	if (ip0->i_ino > ip1->i_ino) {
		temp = ip0;
		ip0 = ip1;
		ip1 = temp;
	}

 again:
	xfs_ilock(ip0, xfs_lock_inumorder(lock_mode, 0));

	/*
	 * 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;
	if (lp && (lp->li_flags & XFS_LI_IN_AIL)) {
		if (!xfs_ilock_nowait(ip1, xfs_lock_inumorder(lock_mode, 1))) {
			xfs_iunlock(ip0, lock_mode);
			if ((++attempts % 5) == 0)
				delay(1); /* Don't just spin the CPU */
			goto again;
		}
	} else {
		xfs_ilock(ip1, xfs_lock_inumorder(lock_mode, 1));
	}
}

1860
int
L
Linus Torvalds 已提交
1861
xfs_remove(
1862
	xfs_inode_t             *dp,
1863 1864
	struct xfs_name		*name,
	xfs_inode_t		*ip)
L
Linus Torvalds 已提交
1865
{
1866
	xfs_mount_t		*mp = dp->i_mount;
L
Linus Torvalds 已提交
1867
	xfs_trans_t             *tp = NULL;
1868
	int			is_dir = S_ISDIR(ip->i_d.di_mode);
L
Linus Torvalds 已提交
1869 1870 1871 1872 1873 1874 1875
	int                     error = 0;
	xfs_bmap_free_t         free_list;
	xfs_fsblock_t           first_block;
	int			cancel_flags;
	int			committed;
	int			link_zero;
	uint			resblks;
1876
	uint			log_count;
L
Linus Torvalds 已提交
1877

1878
	xfs_itrace_entry(dp);
1879
	xfs_itrace_entry(ip);
L
Linus Torvalds 已提交
1880 1881 1882 1883

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

1884
	if (DM_EVENT_ENABLED(dp, DM_EVENT_REMOVE)) {
1885 1886 1887
		error = XFS_SEND_NAMESP(mp, DM_EVENT_REMOVE, dp, DM_RIGHT_NULL,
					NULL, DM_RIGHT_NULL, name->name, NULL,
					ip->i_d.di_mode, 0, 0);
L
Linus Torvalds 已提交
1888 1889 1890 1891 1892
		if (error)
			return error;
	}

	error = XFS_QM_DQATTACH(mp, dp, 0);
1893 1894 1895 1896 1897
	if (error)
		goto std_return;

	error = XFS_QM_DQATTACH(mp, ip, 0);
	if (error)
L
Linus Torvalds 已提交
1898 1899
		goto std_return;

1900 1901 1902 1903 1904 1905 1906
	if (is_dir) {
		tp = xfs_trans_alloc(mp, XFS_TRANS_RMDIR);
		log_count = XFS_DEFAULT_LOG_COUNT;
	} else {
		tp = xfs_trans_alloc(mp, XFS_TRANS_REMOVE);
		log_count = XFS_REMOVE_LOG_COUNT;
	}
L
Linus Torvalds 已提交
1907
	cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
1908

L
Linus Torvalds 已提交
1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919
	/*
	 * 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);
	error = xfs_trans_reserve(tp, resblks, XFS_REMOVE_LOG_RES(mp), 0,
1920
				  XFS_TRANS_PERM_LOG_RES, log_count);
L
Linus Torvalds 已提交
1921 1922 1923
	if (error == ENOSPC) {
		resblks = 0;
		error = xfs_trans_reserve(tp, 0, XFS_REMOVE_LOG_RES(mp), 0,
1924
					  XFS_TRANS_PERM_LOG_RES, log_count);
L
Linus Torvalds 已提交
1925 1926 1927
	}
	if (error) {
		ASSERT(error != ENOSPC);
1928 1929
		cancel_flags = 0;
		goto out_trans_cancel;
L
Linus Torvalds 已提交
1930 1931
	}

1932
	xfs_lock_two_inodes(dp, ip, XFS_ILOCK_EXCL);
L
Linus Torvalds 已提交
1933 1934 1935 1936 1937

	/*
	 * At this point, we've gotten both the directory and the entry
	 * inodes locked.
	 */
1938
	IHOLD(ip);
L
Linus Torvalds 已提交
1939
	xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
1940 1941 1942

	IHOLD(dp);
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
L
Linus Torvalds 已提交
1943

1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
	/*
	 * If we're removing a directory perform some additional validation.
	 */
	if (is_dir) {
		ASSERT(ip->i_d.di_nlink >= 2);
		if (ip->i_d.di_nlink != 2) {
			error = XFS_ERROR(ENOTEMPTY);
			goto out_trans_cancel;
		}
		if (!xfs_dir_isempty(ip)) {
			error = XFS_ERROR(ENOTEMPTY);
			goto out_trans_cancel;
		}
	}

L
Linus Torvalds 已提交
1959
	XFS_BMAP_INIT(&free_list, &first_block);
1960
	error = xfs_dir_removename(tp, dp, name, ip->i_ino,
1961
					&first_block, &free_list, resblks);
L
Linus Torvalds 已提交
1962 1963
	if (error) {
		ASSERT(error != ENOENT);
1964
		goto out_bmap_cancel;
L
Linus Torvalds 已提交
1965 1966 1967
	}
	xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);

1968 1969 1970 1971 1972 1973 1974 1975 1976
	if (is_dir) {
		/*
		 * Drop the link from ip's "..".
		 */
		error = xfs_droplink(tp, dp);
		if (error)
			goto out_bmap_cancel;

		/*
1977
		 * Drop the "." link from ip to self.
1978 1979 1980 1981 1982 1983 1984
		 */
		error = xfs_droplink(tp, ip);
		if (error)
			goto out_bmap_cancel;
	} else {
		/*
		 * When removing a non-directory we need to log the parent
1985 1986
		 * inode here.  For a directory this is done implicitly
		 * by the xfs_droplink call for the ".." entry.
1987 1988
		 */
		xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
L
Linus Torvalds 已提交
1989 1990
	}

1991
	/*
1992
	 * Drop the link from dp to ip.
1993 1994 1995 1996 1997 1998 1999
	 */
	error = xfs_droplink(tp, ip);
	if (error)
		goto out_bmap_cancel;

	/*
	 * Determine if this is the last link while
L
Linus Torvalds 已提交
2000 2001
	 * we are in the transaction.
	 */
2002
	link_zero = (ip->i_d.di_nlink == 0);
L
Linus Torvalds 已提交
2003 2004 2005 2006 2007 2008

	/*
	 * If this is a synchronous mount, make sure that the
	 * remove transaction goes to disk before returning to
	 * the user.
	 */
2009
	if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC))
L
Linus Torvalds 已提交
2010 2011
		xfs_trans_set_sync(tp);

2012
	error = xfs_bmap_finish(&tp, &free_list, &committed);
2013 2014
	if (error)
		goto out_bmap_cancel;
L
Linus Torvalds 已提交
2015

2016
	error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
2017
	if (error)
L
Linus Torvalds 已提交
2018 2019
		goto std_return;

2020 2021 2022 2023 2024 2025
	/*
	 * If we are using filestreams, kill the stream association.
	 * If the file is still open it may get a new one but that
	 * will get killed on last close in xfs_close() so we don't
	 * have to worry about that.
	 */
2026
	if (!is_dir && link_zero && xfs_inode_is_filestream(ip))
2027 2028
		xfs_filestream_deassociate(ip);

2029
	xfs_itrace_exit(ip);
2030
	xfs_itrace_exit(dp);
L
Linus Torvalds 已提交
2031 2032

 std_return:
2033
	if (DM_EVENT_ENABLED(dp, DM_EVENT_POSTREMOVE)) {
2034 2035 2036
		XFS_SEND_NAMESP(mp, DM_EVENT_POSTREMOVE, dp, DM_RIGHT_NULL,
				NULL, DM_RIGHT_NULL, name->name, NULL,
				ip->i_d.di_mode, error, 0);
L
Linus Torvalds 已提交
2037
	}
2038

L
Linus Torvalds 已提交
2039 2040
	return error;

2041
 out_bmap_cancel:
L
Linus Torvalds 已提交
2042 2043
	xfs_bmap_cancel(&free_list);
	cancel_flags |= XFS_TRANS_ABORT;
2044
 out_trans_cancel:
L
Linus Torvalds 已提交
2045 2046 2047 2048
	xfs_trans_cancel(tp, cancel_flags);
	goto std_return;
}

2049
int
L
Linus Torvalds 已提交
2050
xfs_link(
2051
	xfs_inode_t		*tdp,
2052
	xfs_inode_t		*sip,
2053
	struct xfs_name		*target_name)
L
Linus Torvalds 已提交
2054
{
2055
	xfs_mount_t		*mp = tdp->i_mount;
L
Linus Torvalds 已提交
2056 2057 2058 2059 2060 2061 2062 2063
	xfs_trans_t		*tp;
	int			error;
	xfs_bmap_free_t         free_list;
	xfs_fsblock_t           first_block;
	int			cancel_flags;
	int			committed;
	int			resblks;

2064
	xfs_itrace_entry(tdp);
2065
	xfs_itrace_entry(sip);
L
Linus Torvalds 已提交
2066

2067
	ASSERT(!S_ISDIR(sip->i_d.di_mode));
L
Linus Torvalds 已提交
2068 2069 2070 2071

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

2072
	if (DM_EVENT_ENABLED(tdp, DM_EVENT_LINK)) {
L
Linus Torvalds 已提交
2073
		error = XFS_SEND_NAMESP(mp, DM_EVENT_LINK,
2074 2075
					tdp, DM_RIGHT_NULL,
					sip, DM_RIGHT_NULL,
2076
					target_name->name, NULL, 0, 0, 0);
L
Linus Torvalds 已提交
2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
		if (error)
			return error;
	}

	/* Return through std_return after this point. */

	error = XFS_QM_DQATTACH(mp, sip, 0);
	if (!error && sip != tdp)
		error = XFS_QM_DQATTACH(mp, tdp, 0);
	if (error)
		goto std_return;

	tp = xfs_trans_alloc(mp, XFS_TRANS_LINK);
	cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
2091
	resblks = XFS_LINK_SPACE_RES(mp, target_name->len);
L
Linus Torvalds 已提交
2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103
	error = xfs_trans_reserve(tp, resblks, XFS_LINK_LOG_RES(mp), 0,
			XFS_TRANS_PERM_LOG_RES, XFS_LINK_LOG_COUNT);
	if (error == ENOSPC) {
		resblks = 0;
		error = xfs_trans_reserve(tp, 0, XFS_LINK_LOG_RES(mp), 0,
				XFS_TRANS_PERM_LOG_RES, XFS_LINK_LOG_COUNT);
	}
	if (error) {
		cancel_flags = 0;
		goto error_return;
	}

2104
	xfs_lock_two_inodes(sip, tdp, XFS_ILOCK_EXCL);
L
Linus Torvalds 已提交
2105 2106 2107 2108 2109 2110

	/*
	 * Increment vnode ref counts since xfs_trans_commit &
	 * xfs_trans_cancel will both unlock the inodes and
	 * decrement the associated ref counts.
	 */
2111 2112
	IHOLD(sip);
	IHOLD(tdp);
L
Linus Torvalds 已提交
2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123
	xfs_trans_ijoin(tp, sip, XFS_ILOCK_EXCL);
	xfs_trans_ijoin(tp, tdp, XFS_ILOCK_EXCL);

	/*
	 * If the source has too many links, we can't make any more to it.
	 */
	if (sip->i_d.di_nlink >= XFS_MAXLINK) {
		error = XFS_ERROR(EMLINK);
		goto error_return;
	}

2124 2125 2126 2127 2128 2129 2130
	/*
	 * 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) &&
		     (tdp->i_d.di_projid != sip->i_d.di_projid))) {
2131
		error = XFS_ERROR(EXDEV);
2132 2133 2134
		goto error_return;
	}

2135 2136
	error = xfs_dir_canenter(tp, tdp, target_name, resblks);
	if (error)
L
Linus Torvalds 已提交
2137 2138 2139 2140
		goto error_return;

	XFS_BMAP_INIT(&free_list, &first_block);

2141 2142
	error = xfs_dir_createname(tp, tdp, target_name, sip->i_ino,
					&first_block, &free_list, resblks);
L
Linus Torvalds 已提交
2143 2144 2145 2146 2147 2148
	if (error)
		goto abort_return;
	xfs_ichgtime(tdp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	xfs_trans_log_inode(tp, tdp, XFS_ILOG_CORE);

	error = xfs_bumplink(tp, sip);
2149
	if (error)
L
Linus Torvalds 已提交
2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160
		goto abort_return;

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

2161
	error = xfs_bmap_finish (&tp, &free_list, &committed);
L
Linus Torvalds 已提交
2162 2163 2164 2165 2166
	if (error) {
		xfs_bmap_cancel(&free_list);
		goto abort_return;
	}

2167
	error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
2168
	if (error)
L
Linus Torvalds 已提交
2169 2170 2171 2172
		goto std_return;

	/* Fall through to std_return with error = 0. */
std_return:
2173
	if (DM_EVENT_ENABLED(sip, DM_EVENT_POSTLINK)) {
L
Linus Torvalds 已提交
2174
		(void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTLINK,
2175 2176
				tdp, DM_RIGHT_NULL,
				sip, DM_RIGHT_NULL,
2177
				target_name->name, NULL, 0, error, 0);
L
Linus Torvalds 已提交
2178 2179 2180 2181 2182 2183
	}
	return error;

 abort_return:
	cancel_flags |= XFS_TRANS_ABORT;
	/* FALLTHROUGH */
2184

L
Linus Torvalds 已提交
2185 2186 2187 2188
 error_return:
	xfs_trans_cancel(tp, cancel_flags);
	goto std_return;
}
2189 2190


2191
int
L
Linus Torvalds 已提交
2192
xfs_mkdir(
2193
	xfs_inode_t             *dp,
2194
	struct xfs_name		*dir_name,
2195
	mode_t			mode,
2196
	xfs_inode_t		**ipp,
L
Linus Torvalds 已提交
2197 2198
	cred_t			*credp)
{
2199
	xfs_mount_t		*mp = dp->i_mount;
L
Linus Torvalds 已提交
2200 2201 2202 2203 2204 2205 2206
	xfs_inode_t		*cdp;	/* inode of created dir */
	xfs_trans_t		*tp;
	int			cancel_flags;
	int			error;
	int			committed;
	xfs_bmap_free_t         free_list;
	xfs_fsblock_t           first_block;
2207
	boolean_t		unlock_dp_on_error = B_FALSE;
L
Linus Torvalds 已提交
2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218
	boolean_t		created = B_FALSE;
	int			dm_event_sent = 0;
	xfs_prid_t		prid;
	struct xfs_dquot	*udqp, *gdqp;
	uint			resblks;

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

	tp = NULL;

2219
	if (DM_EVENT_ENABLED(dp, DM_EVENT_CREATE)) {
L
Linus Torvalds 已提交
2220
		error = XFS_SEND_NAMESP(mp, DM_EVENT_CREATE,
2221
					dp, DM_RIGHT_NULL, NULL,
2222
					DM_RIGHT_NULL, dir_name->name, NULL,
2223
					mode, 0, 0);
L
Linus Torvalds 已提交
2224 2225 2226 2227 2228 2229 2230
		if (error)
			return error;
		dm_event_sent = 1;
	}

	/* Return through std_return after this point. */

2231
	xfs_itrace_entry(dp);
L
Linus Torvalds 已提交
2232 2233 2234

	mp = dp->i_mount;
	udqp = gdqp = NULL;
2235 2236
	if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
		prid = dp->i_d.di_projid;
L
Linus Torvalds 已提交
2237 2238 2239 2240 2241 2242 2243
	else
		prid = (xfs_prid_t)dfltprid;

	/*
	 * Make sure that we have allocated dquot(s) on disk.
	 */
	error = XFS_QM_DQVOPALLOC(mp, dp,
2244
			current_fsuid(), current_fsgid(), prid,
L
Linus Torvalds 已提交
2245 2246 2247 2248 2249 2250
			XFS_QMOPT_QUOTALL | XFS_QMOPT_INHERIT, &udqp, &gdqp);
	if (error)
		goto std_return;

	tp = xfs_trans_alloc(mp, XFS_TRANS_MKDIR);
	cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
2251
	resblks = XFS_MKDIR_SPACE_RES(mp, dir_name->len);
L
Linus Torvalds 已提交
2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264
	error = xfs_trans_reserve(tp, resblks, XFS_MKDIR_LOG_RES(mp), 0,
				  XFS_TRANS_PERM_LOG_RES, XFS_MKDIR_LOG_COUNT);
	if (error == ENOSPC) {
		resblks = 0;
		error = xfs_trans_reserve(tp, 0, XFS_MKDIR_LOG_RES(mp), 0,
					  XFS_TRANS_PERM_LOG_RES,
					  XFS_MKDIR_LOG_COUNT);
	}
	if (error) {
		cancel_flags = 0;
		goto error_return;
	}

2265
	xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
2266
	unlock_dp_on_error = B_TRUE;
L
Linus Torvalds 已提交
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282

	/*
	 * Check for directory link count overflow.
	 */
	if (dp->i_d.di_nlink >= XFS_MAXLINK) {
		error = XFS_ERROR(EMLINK);
		goto error_return;
	}

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

2283 2284
	error = xfs_dir_canenter(tp, dp, dir_name, resblks);
	if (error)
L
Linus Torvalds 已提交
2285 2286 2287 2288
		goto error_return;
	/*
	 * create the directory inode.
	 */
2289
	error = xfs_dir_ialloc(&tp, dp, mode, 2,
L
Linus Torvalds 已提交
2290 2291 2292 2293 2294 2295 2296
			0, credp, prid, resblks > 0,
		&cdp, NULL);
	if (error) {
		if (error == ENOSPC)
			goto error_return;
		goto abort_return;
	}
2297
	xfs_itrace_ref(cdp);
L
Linus Torvalds 已提交
2298 2299 2300 2301 2302

	/*
	 * Now we add the directory inode to the transaction.
	 * We waited until now since xfs_dir_ialloc might start
	 * a new transaction.  Had we joined the transaction
2303 2304 2305
	 * earlier, the locks might have gotten released. An error
	 * from here on will result in the transaction cancel
	 * unlocking dp so don't do it explicitly in the error path.
L
Linus Torvalds 已提交
2306
	 */
2307
	IHOLD(dp);
L
Linus Torvalds 已提交
2308
	xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
2309
	unlock_dp_on_error = B_FALSE;
L
Linus Torvalds 已提交
2310 2311 2312

	XFS_BMAP_INIT(&free_list, &first_block);

2313 2314 2315
	error = xfs_dir_createname(tp, dp, dir_name, cdp->i_ino,
					&first_block, &free_list, resblks ?
					resblks - XFS_IALLOC_SPACE_RES(mp) : 0);
L
Linus Torvalds 已提交
2316 2317 2318 2319 2320 2321
	if (error) {
		ASSERT(error != ENOSPC);
		goto error1;
	}
	xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);

2322 2323
	error = xfs_dir_init(tp, cdp, dp);
	if (error)
L
Linus Torvalds 已提交
2324 2325 2326
		goto error2;

	error = xfs_bumplink(tp, dp);
2327
	if (error)
L
Linus Torvalds 已提交
2328 2329 2330 2331
		goto error2;

	created = B_TRUE;

2332
	*ipp = cdp;
L
Linus Torvalds 已提交
2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348
	IHOLD(cdp);

	/*
	 * Attach the dquots to the new inode and modify the icount incore.
	 */
	XFS_QM_DQVOPCREATE(mp, tp, cdp, udqp, gdqp);

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

2349
	error = xfs_bmap_finish(&tp, &free_list, &committed);
L
Linus Torvalds 已提交
2350 2351 2352 2353 2354
	if (error) {
		IRELE(cdp);
		goto error2;
	}

2355
	error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
L
Linus Torvalds 已提交
2356 2357 2358 2359 2360 2361 2362 2363 2364 2365
	XFS_QM_DQRELE(mp, udqp);
	XFS_QM_DQRELE(mp, gdqp);
	if (error) {
		IRELE(cdp);
	}

	/* Fall through to std_return with error = 0 or errno from
	 * xfs_trans_commit. */

std_return:
2366
	if ((created || (error != 0 && dm_event_sent != 0)) &&
2367
	    DM_EVENT_ENABLED(dp, DM_EVENT_POSTCREATE)) {
L
Linus Torvalds 已提交
2368
		(void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTCREATE,
2369 2370
					dp, DM_RIGHT_NULL,
					created ? cdp : NULL,
L
Linus Torvalds 已提交
2371
					DM_RIGHT_NULL,
2372
					dir_name->name, NULL,
2373
					mode, error, 0);
L
Linus Torvalds 已提交
2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386
	}
	return error;

 error2:
 error1:
	xfs_bmap_cancel(&free_list);
 abort_return:
	cancel_flags |= XFS_TRANS_ABORT;
 error_return:
	xfs_trans_cancel(tp, cancel_flags);
	XFS_QM_DQRELE(mp, udqp);
	XFS_QM_DQRELE(mp, gdqp);

2387
	if (unlock_dp_on_error)
L
Linus Torvalds 已提交
2388 2389 2390 2391 2392
		xfs_iunlock(dp, XFS_ILOCK_EXCL);

	goto std_return;
}

2393
int
L
Linus Torvalds 已提交
2394
xfs_symlink(
2395
	xfs_inode_t		*dp,
2396 2397
	struct xfs_name		*link_name,
	const char		*target_path,
2398
	mode_t			mode,
2399
	xfs_inode_t		**ipp,
L
Linus Torvalds 已提交
2400 2401
	cred_t			*credp)
{
2402
	xfs_mount_t		*mp = dp->i_mount;
L
Linus Torvalds 已提交
2403 2404 2405 2406 2407 2408
	xfs_trans_t		*tp;
	xfs_inode_t		*ip;
	int			error;
	int			pathlen;
	xfs_bmap_free_t		free_list;
	xfs_fsblock_t		first_block;
2409
	boolean_t		unlock_dp_on_error = B_FALSE;
L
Linus Torvalds 已提交
2410 2411 2412 2413 2414 2415 2416
	uint			cancel_flags;
	int			committed;
	xfs_fileoff_t		first_fsb;
	xfs_filblks_t		fs_blocks;
	int			nmaps;
	xfs_bmbt_irec_t		mval[SYMLINK_MAPS];
	xfs_daddr_t		d;
2417
	const char		*cur_chunk;
L
Linus Torvalds 已提交
2418 2419 2420 2421 2422 2423 2424
	int			byte_cnt;
	int			n;
	xfs_buf_t		*bp;
	xfs_prid_t		prid;
	struct xfs_dquot	*udqp, *gdqp;
	uint			resblks;

2425
	*ipp = NULL;
L
Linus Torvalds 已提交
2426 2427 2428 2429
	error = 0;
	ip = NULL;
	tp = NULL;

2430
	xfs_itrace_entry(dp);
L
Linus Torvalds 已提交
2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

	/*
	 * Check component lengths of the target path name.
	 */
	pathlen = strlen(target_path);
	if (pathlen >= MAXPATHLEN)      /* total string too long */
		return XFS_ERROR(ENAMETOOLONG);

2442
	if (DM_EVENT_ENABLED(dp, DM_EVENT_SYMLINK)) {
2443
		error = XFS_SEND_NAMESP(mp, DM_EVENT_SYMLINK, dp,
L
Linus Torvalds 已提交
2444
					DM_RIGHT_NULL, NULL, DM_RIGHT_NULL,
2445
					link_name->name, target_path, 0, 0, 0);
L
Linus Torvalds 已提交
2446 2447 2448 2449 2450 2451 2452
		if (error)
			return error;
	}

	/* Return through std_return after this point. */

	udqp = gdqp = NULL;
2453 2454
	if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
		prid = dp->i_d.di_projid;
L
Linus Torvalds 已提交
2455 2456 2457 2458 2459 2460 2461
	else
		prid = (xfs_prid_t)dfltprid;

	/*
	 * Make sure that we have allocated dquot(s) on disk.
	 */
	error = XFS_QM_DQVOPALLOC(mp, dp,
2462
			current_fsuid(), current_fsgid(), prid,
L
Linus Torvalds 已提交
2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476
			XFS_QMOPT_QUOTALL | XFS_QMOPT_INHERIT, &udqp, &gdqp);
	if (error)
		goto std_return;

	tp = xfs_trans_alloc(mp, XFS_TRANS_SYMLINK);
	cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
	/*
	 * The symlink will fit into the inode data fork?
	 * There can't be any attributes so we get the whole variable part.
	 */
	if (pathlen <= XFS_LITINO(mp))
		fs_blocks = 0;
	else
		fs_blocks = XFS_B_TO_FSB(mp, pathlen);
2477
	resblks = XFS_SYMLINK_SPACE_RES(mp, link_name->len, fs_blocks);
L
Linus Torvalds 已提交
2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489
	error = xfs_trans_reserve(tp, resblks, XFS_SYMLINK_LOG_RES(mp), 0,
			XFS_TRANS_PERM_LOG_RES, XFS_SYMLINK_LOG_COUNT);
	if (error == ENOSPC && fs_blocks == 0) {
		resblks = 0;
		error = xfs_trans_reserve(tp, 0, XFS_SYMLINK_LOG_RES(mp), 0,
				XFS_TRANS_PERM_LOG_RES, XFS_SYMLINK_LOG_COUNT);
	}
	if (error) {
		cancel_flags = 0;
		goto error_return;
	}

2490
	xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
2491
	unlock_dp_on_error = B_TRUE;
L
Linus Torvalds 已提交
2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510

	/*
	 * Check whether the directory allows new symlinks or not.
	 */
	if (dp->i_d.di_flags & XFS_DIFLAG_NOSYMLINKS) {
		error = XFS_ERROR(EPERM);
		goto error_return;
	}

	/*
	 * Reserve disk quota : blocks and inode.
	 */
	error = XFS_TRANS_RESERVE_QUOTA(mp, tp, udqp, gdqp, resblks, 1, 0);
	if (error)
		goto error_return;

	/*
	 * Check for ability to enter directory entry, if no space reserved.
	 */
2511 2512
	error = xfs_dir_canenter(tp, dp, link_name, resblks);
	if (error)
L
Linus Torvalds 已提交
2513 2514 2515 2516 2517 2518 2519 2520 2521 2522
		goto error_return;
	/*
	 * Initialize the bmap freelist prior to calling either
	 * bmapi or the directory create code.
	 */
	XFS_BMAP_INIT(&free_list, &first_block);

	/*
	 * Allocate an inode for the symlink.
	 */
2523
	error = xfs_dir_ialloc(&tp, dp, S_IFLNK | (mode & ~S_IFMT),
L
Linus Torvalds 已提交
2524 2525 2526 2527 2528 2529
			       1, 0, credp, prid, resblks > 0, &ip, NULL);
	if (error) {
		if (error == ENOSPC)
			goto error_return;
		goto error1;
	}
2530
	xfs_itrace_ref(ip);
L
Linus Torvalds 已提交
2531

2532 2533 2534 2535 2536
	/*
	 * An error after we've joined dp to the transaction will result in the
	 * transaction cancel unlocking dp so don't do it explicitly in the
	 * error path.
	 */
2537
	IHOLD(dp);
L
Linus Torvalds 已提交
2538
	xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
2539
	unlock_dp_on_error = B_FALSE;
L
Linus Torvalds 已提交
2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571

	/*
	 * Also attach the dquot(s) to it, if applicable.
	 */
	XFS_QM_DQVOPCREATE(mp, tp, ip, udqp, gdqp);

	if (resblks)
		resblks -= XFS_IALLOC_SPACE_RES(mp);
	/*
	 * If the symlink will fit into the inode, write it inline.
	 */
	if (pathlen <= XFS_IFORK_DSIZE(ip)) {
		xfs_idata_realloc(ip, pathlen, XFS_DATA_FORK);
		memcpy(ip->i_df.if_u1.if_data, target_path, pathlen);
		ip->i_d.di_size = pathlen;

		/*
		 * The inode was initially created in extent format.
		 */
		ip->i_df.if_flags &= ~(XFS_IFEXTENTS | XFS_IFBROOT);
		ip->i_df.if_flags |= XFS_IFINLINE;

		ip->i_d.di_format = XFS_DINODE_FMT_LOCAL;
		xfs_trans_log_inode(tp, ip, XFS_ILOG_DDATA | XFS_ILOG_CORE);

	} else {
		first_fsb = 0;
		nmaps = SYMLINK_MAPS;

		error = xfs_bmapi(tp, ip, first_fsb, fs_blocks,
				  XFS_BMAPI_WRITE | XFS_BMAPI_METADATA,
				  &first_block, resblks, mval, &nmaps,
2572
				  &free_list, NULL);
L
Linus Torvalds 已提交
2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603
		if (error) {
			goto error1;
		}

		if (resblks)
			resblks -= fs_blocks;
		ip->i_d.di_size = pathlen;
		xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

		cur_chunk = target_path;
		for (n = 0; n < nmaps; n++) {
			d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
			byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);
			bp = xfs_trans_get_buf(tp, mp->m_ddev_targp, d,
					       BTOBB(byte_cnt), 0);
			ASSERT(bp && !XFS_BUF_GETERROR(bp));
			if (pathlen < byte_cnt) {
				byte_cnt = pathlen;
			}
			pathlen -= byte_cnt;

			memcpy(XFS_BUF_PTR(bp), cur_chunk, byte_cnt);
			cur_chunk += byte_cnt;

			xfs_trans_log_buf(tp, bp, 0, byte_cnt - 1);
		}
	}

	/*
	 * Create the directory entry for the symlink.
	 */
2604 2605
	error = xfs_dir_createname(tp, dp, link_name, ip->i_ino,
					&first_block, &free_list, resblks);
2606
	if (error)
L
Linus Torvalds 已提交
2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626
		goto error1;
	xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);

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

	/*
	 * xfs_trans_commit normally decrements the vnode ref count
	 * when it unlocks the inode. Since we want to return the
	 * vnode to the caller, we bump the vnode ref count now.
	 */
	IHOLD(ip);

2627
	error = xfs_bmap_finish(&tp, &free_list, &committed);
L
Linus Torvalds 已提交
2628 2629 2630
	if (error) {
		goto error2;
	}
2631
	error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
L
Linus Torvalds 已提交
2632 2633 2634 2635 2636 2637
	XFS_QM_DQRELE(mp, udqp);
	XFS_QM_DQRELE(mp, gdqp);

	/* Fall through to std_return with error = 0 or errno from
	 * xfs_trans_commit	*/
std_return:
2638
	if (DM_EVENT_ENABLED(dp, DM_EVENT_POSTSYMLINK)) {
L
Linus Torvalds 已提交
2639
		(void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTSYMLINK,
2640 2641
					dp, DM_RIGHT_NULL,
					error ? NULL : ip,
2642 2643
					DM_RIGHT_NULL, link_name->name,
					target_path, 0, error, 0);
L
Linus Torvalds 已提交
2644 2645
	}

2646 2647
	if (!error)
		*ipp = ip;
L
Linus Torvalds 已提交
2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659
	return error;

 error2:
	IRELE(ip);
 error1:
	xfs_bmap_cancel(&free_list);
	cancel_flags |= XFS_TRANS_ABORT;
 error_return:
	xfs_trans_cancel(tp, cancel_flags);
	XFS_QM_DQRELE(mp, udqp);
	XFS_QM_DQRELE(mp, gdqp);

2660
	if (unlock_dp_on_error)
L
Linus Torvalds 已提交
2661 2662 2663 2664 2665
		xfs_iunlock(dp, XFS_ILOCK_EXCL);

	goto std_return;
}

2666
int
L
Linus Torvalds 已提交
2667
xfs_inode_flush(
2668
	xfs_inode_t	*ip,
L
Linus Torvalds 已提交
2669 2670
	int		flags)
{
2671
	xfs_mount_t	*mp = ip->i_mount;
L
Linus Torvalds 已提交
2672 2673 2674 2675 2676 2677 2678 2679 2680
	int		error = 0;

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

	/*
	 * Bypass inodes which have already been cleaned by
	 * the inode flush clustering code inside xfs_iflush
	 */
2681
	if (xfs_inode_clean(ip))
L
Linus Torvalds 已提交
2682 2683 2684 2685 2686 2687 2688 2689 2690
		return 0;

	/*
	 * We make this non-blocking if the inode is contended,
	 * return EAGAIN to indicate to the caller that they
	 * did not succeed. This prevents the flush path from
	 * blocking on inodes inside another operation right
	 * now, they get caught later by xfs_sync.
	 */
2691 2692 2693 2694 2695 2696
	if (flags & FLUSH_SYNC) {
		xfs_ilock(ip, XFS_ILOCK_SHARED);
		xfs_iflock(ip);
	} else if (xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) {
		if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip)) {
			xfs_iunlock(ip, XFS_ILOCK_SHARED);
L
Linus Torvalds 已提交
2697 2698
			return EAGAIN;
		}
2699 2700
	} else {
		return EAGAIN;
L
Linus Torvalds 已提交
2701 2702
	}

2703 2704 2705 2706
	error = xfs_iflush(ip, (flags & FLUSH_SYNC) ? XFS_IFLUSH_SYNC
						    : XFS_IFLUSH_ASYNC_NOBLOCK);
	xfs_iunlock(ip, XFS_ILOCK_SHARED);

L
Linus Torvalds 已提交
2707 2708 2709
	return error;
}

2710

L
Linus Torvalds 已提交
2711
int
2712 2713
xfs_set_dmattrs(
	xfs_inode_t     *ip,
L
Linus Torvalds 已提交
2714
	u_int		evmask,
2715
	u_int16_t	state)
L
Linus Torvalds 已提交
2716
{
2717
	xfs_mount_t	*mp = ip->i_mount;
L
Linus Torvalds 已提交
2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735
	xfs_trans_t	*tp;
	int		error;

	if (!capable(CAP_SYS_ADMIN))
		return XFS_ERROR(EPERM);

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

	tp = xfs_trans_alloc(mp, XFS_TRANS_SET_DMATTRS);
	error = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES (mp), 0, 0, 0);
	if (error) {
		xfs_trans_cancel(tp, 0);
		return error;
	}
	xfs_ilock(ip, XFS_ILOCK_EXCL);
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);

C
Christoph Hellwig 已提交
2736 2737
	ip->i_d.di_dmevmask = evmask;
	ip->i_d.di_dmstate  = state;
L
Linus Torvalds 已提交
2738 2739 2740

	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	IHOLD(ip);
2741
	error = xfs_trans_commit(tp, 0);
L
Linus Torvalds 已提交
2742 2743 2744 2745

	return error;
}

2746
int
L
Linus Torvalds 已提交
2747
xfs_reclaim(
2748
	xfs_inode_t	*ip)
L
Linus Torvalds 已提交
2749 2750
{

2751
	xfs_itrace_entry(ip);
L
Linus Torvalds 已提交
2752

2753
	ASSERT(!VN_MAPPED(VFS_I(ip)));
L
Linus Torvalds 已提交
2754 2755

	/* bad inode, get out here ASAP */
2756
	if (VN_BAD(VFS_I(ip))) {
L
Linus Torvalds 已提交
2757 2758 2759 2760
		xfs_ireclaim(ip);
		return 0;
	}

2761
	vn_iowait(ip);
L
Linus Torvalds 已提交
2762

2763
	ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
L
Linus Torvalds 已提交
2764

2765 2766 2767 2768 2769
	/*
	 * Make sure the atime in the XFS inode is correct before freeing the
	 * Linux inode.
	 */
	xfs_synchronize_atime(ip);
L
Linus Torvalds 已提交
2770

2771 2772 2773 2774 2775 2776 2777 2778 2779 2780
	/*
	 * If we have nothing to flush with this inode then complete the
	 * teardown now, otherwise break the link between the xfs inode and the
	 * linux inode and clean up the xfs inode later. This avoids flushing
	 * the inode to disk during the delete operation itself.
	 *
	 * When breaking the link, we need to set the XFS_IRECLAIMABLE flag
	 * first to ensure that xfs_iunpin() will never see an xfs inode
	 * that has a linux inode being reclaimed. Synchronisation is provided
	 * by the i_flags_lock.
L
Linus Torvalds 已提交
2781 2782 2783 2784
	 */
	if (!ip->i_update_core && (ip->i_itemp == NULL)) {
		xfs_ilock(ip, XFS_ILOCK_EXCL);
		xfs_iflock(ip);
2785
		xfs_iflags_set(ip, XFS_IRECLAIMABLE);
2786
		return xfs_reclaim_inode(ip, 1, XFS_IFLUSH_DELWRI_ELSE_SYNC);
L
Linus Torvalds 已提交
2787
	}
D
David Chinner 已提交
2788
	xfs_inode_set_reclaim_tag(ip);
L
Linus Torvalds 已提交
2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810
	return 0;
}

/*
 * xfs_alloc_file_space()
 *      This routine allocates disk space for the given file.
 *
 *	If alloc_type == 0, this request is for an ALLOCSP type
 *	request which will change the file size.  In this case, no
 *	DMAPI event will be generated by the call.  A TRUNCATE event
 *	will be generated later by xfs_setattr.
 *
 *	If alloc_type != 0, this request is for a RESVSP type
 *	request, and a DMAPI DM_EVENT_WRITE will be generated if the
 *	lower block boundary byte address is less than the file's
 *	length.
 *
 * RETURNS:
 *       0 on success
 *      errno on error
 *
 */
2811
STATIC int
L
Linus Torvalds 已提交
2812 2813 2814 2815 2816 2817 2818
xfs_alloc_file_space(
	xfs_inode_t		*ip,
	xfs_off_t		offset,
	xfs_off_t		len,
	int			alloc_type,
	int			attr_flags)
{
2819 2820
	xfs_mount_t		*mp = ip->i_mount;
	xfs_off_t		count;
L
Linus Torvalds 已提交
2821 2822
	xfs_filblks_t		allocated_fsb;
	xfs_filblks_t		allocatesize_fsb;
2823 2824
	xfs_extlen_t		extsz, temp;
	xfs_fileoff_t		startoffset_fsb;
L
Linus Torvalds 已提交
2825
	xfs_fsblock_t		firstfsb;
2826 2827 2828
	int			nimaps;
	int			bmapi_flag;
	int			quota_flag;
L
Linus Torvalds 已提交
2829 2830
	int			rt;
	xfs_trans_t		*tp;
2831 2832 2833 2834 2835
	xfs_bmbt_irec_t		imaps[1], *imapp;
	xfs_bmap_free_t		free_list;
	uint			qblocks, resblks, resrtextents;
	int			committed;
	int			error;
L
Linus Torvalds 已提交
2836

2837
	xfs_itrace_entry(ip);
L
Linus Torvalds 已提交
2838 2839 2840 2841 2842 2843 2844 2845 2846 2847

	if (XFS_FORCED_SHUTDOWN(mp))
		return XFS_ERROR(EIO);

	if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
		return error;

	if (len <= 0)
		return XFS_ERROR(EINVAL);

2848 2849 2850
	rt = XFS_IS_REALTIME_INODE(ip);
	extsz = xfs_get_extsz_hint(ip);

L
Linus Torvalds 已提交
2851 2852
	count = len;
	imapp = &imaps[0];
2853 2854
	nimaps = 1;
	bmapi_flag = XFS_BMAPI_WRITE | (alloc_type ? XFS_BMAPI_PREALLOC : 0);
L
Linus Torvalds 已提交
2855 2856 2857 2858
	startoffset_fsb	= XFS_B_TO_FSBT(mp, offset);
	allocatesize_fsb = XFS_B_TO_FSB(mp, count);

	/*	Generate a DMAPI event if needed.	*/
2859
	if (alloc_type != 0 && offset < ip->i_size &&
2860
			(attr_flags & XFS_ATTR_DMI) == 0  &&
2861
			DM_EVENT_ENABLED(ip, DM_EVENT_WRITE)) {
L
Linus Torvalds 已提交
2862 2863 2864
		xfs_off_t           end_dmi_offset;

		end_dmi_offset = offset+len;
2865 2866
		if (end_dmi_offset > ip->i_size)
			end_dmi_offset = ip->i_size;
2867 2868
		error = XFS_SEND_DATA(mp, DM_EVENT_WRITE, ip, offset,
				      end_dmi_offset - offset, 0, NULL);
L
Linus Torvalds 已提交
2869
		if (error)
2870
			return error;
L
Linus Torvalds 已提交
2871 2872 2873
	}

	/*
2874
	 * Allocate file space until done or until there is an error
L
Linus Torvalds 已提交
2875 2876 2877
	 */
retry:
	while (allocatesize_fsb && !error) {
2878 2879
		xfs_fileoff_t	s, e;

L
Linus Torvalds 已提交
2880
		/*
2881
		 * Determine space reservations for data/realtime.
L
Linus Torvalds 已提交
2882
		 */
2883
		if (unlikely(extsz)) {
L
Linus Torvalds 已提交
2884
			s = startoffset_fsb;
2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901
			do_div(s, extsz);
			s *= extsz;
			e = startoffset_fsb + allocatesize_fsb;
			if ((temp = do_mod(startoffset_fsb, extsz)))
				e += temp;
			if ((temp = do_mod(e, extsz)))
				e += extsz - temp;
		} else {
			s = 0;
			e = allocatesize_fsb;
		}

		if (unlikely(rt)) {
			resrtextents = qblocks = (uint)(e - s);
			resrtextents /= mp->m_sb.sb_rextsize;
			resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0);
			quota_flag = XFS_QMOPT_RES_RTBLKS;
L
Linus Torvalds 已提交
2902
		} else {
2903 2904 2905 2906
			resrtextents = 0;
			resblks = qblocks = \
				XFS_DIOSTRAT_SPACE_RES(mp, (uint)(e - s));
			quota_flag = XFS_QMOPT_RES_REGBLKS;
L
Linus Torvalds 已提交
2907 2908 2909
		}

		/*
2910
		 * Allocate and setup the transaction.
L
Linus Torvalds 已提交
2911 2912
		 */
		tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT);
2913 2914
		error = xfs_trans_reserve(tp, resblks,
					  XFS_WRITE_LOG_RES(mp), resrtextents,
L
Linus Torvalds 已提交
2915 2916 2917
					  XFS_TRANS_PERM_LOG_RES,
					  XFS_WRITE_LOG_COUNT);
		/*
2918
		 * Check for running out of space
L
Linus Torvalds 已提交
2919 2920 2921 2922 2923 2924 2925 2926 2927 2928
		 */
		if (error) {
			/*
			 * Free the transaction structure.
			 */
			ASSERT(error == ENOSPC || XFS_FORCED_SHUTDOWN(mp));
			xfs_trans_cancel(tp, 0);
			break;
		}
		xfs_ilock(ip, XFS_ILOCK_EXCL);
2929 2930
		error = XFS_TRANS_RESERVE_QUOTA_NBLKS(mp, tp, ip,
						      qblocks, 0, quota_flag);
L
Linus Torvalds 已提交
2931 2932 2933 2934 2935 2936 2937
		if (error)
			goto error1;

		xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);

		/*
2938
		 * Issue the xfs_bmapi() call to allocate the blocks
L
Linus Torvalds 已提交
2939 2940
		 */
		XFS_BMAP_INIT(&free_list, &firstfsb);
2941
		error = xfs_bmapi(tp, ip, startoffset_fsb,
2942 2943
				  allocatesize_fsb, bmapi_flag,
				  &firstfsb, 0, imapp, &nimaps,
2944
				  &free_list, NULL);
L
Linus Torvalds 已提交
2945 2946 2947 2948 2949
		if (error) {
			goto error0;
		}

		/*
2950
		 * Complete the transaction
L
Linus Torvalds 已提交
2951
		 */
2952
		error = xfs_bmap_finish(&tp, &free_list, &committed);
L
Linus Torvalds 已提交
2953 2954 2955 2956
		if (error) {
			goto error0;
		}

2957
		error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
L
Linus Torvalds 已提交
2958 2959 2960 2961 2962 2963 2964
		xfs_iunlock(ip, XFS_ILOCK_EXCL);
		if (error) {
			break;
		}

		allocated_fsb = imapp->br_blockcount;

2965
		if (nimaps == 0) {
L
Linus Torvalds 已提交
2966 2967 2968 2969 2970 2971 2972 2973
			error = XFS_ERROR(ENOSPC);
			break;
		}

		startoffset_fsb += allocated_fsb;
		allocatesize_fsb -= allocated_fsb;
	}
dmapi_enospc_check:
2974
	if (error == ENOSPC && (attr_flags & XFS_ATTR_DMI) == 0 &&
2975
	    DM_EVENT_ENABLED(ip, DM_EVENT_NOSPACE)) {
L
Linus Torvalds 已提交
2976
		error = XFS_SEND_NAMESP(mp, DM_EVENT_NOSPACE,
2977 2978
				ip, DM_RIGHT_NULL,
				ip, DM_RIGHT_NULL,
L
Linus Torvalds 已提交
2979 2980 2981 2982 2983 2984 2985 2986
				NULL, NULL, 0, 0, 0); /* Delay flag intentionally unused */
		if (error == 0)
			goto retry;	/* Maybe DMAPI app. has made space */
		/* else fall through with error from XFS_SEND_DATA */
	}

	return error;

2987
error0:	/* Cancel bmap, unlock inode, unreserve quota blocks, cancel trans */
L
Linus Torvalds 已提交
2988
	xfs_bmap_cancel(&free_list);
2989 2990 2991
	XFS_TRANS_UNRESERVE_QUOTA_NBLKS(mp, tp, ip, qblocks, 0, quota_flag);

error1:	/* Just cancel transaction */
L
Linus Torvalds 已提交
2992 2993 2994 2995 2996 2997 2998 2999
	xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
	xfs_iunlock(ip, XFS_ILOCK_EXCL);
	goto dmapi_enospc_check;
}

/*
 * Zero file bytes between startoff and endoff inclusive.
 * The iolock is held exclusive and no blocks are buffered.
3000 3001 3002 3003 3004 3005 3006
 *
 * This function is used by xfs_free_file_space() to zero
 * partial blocks when the range to free is not block aligned.
 * When unreserving space with boundaries that are not block
 * aligned we round up the start and round down the end
 * boundaries and then use this function to zero the parts of
 * the blocks that got dropped during the rounding.
L
Linus Torvalds 已提交
3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022
 */
STATIC int
xfs_zero_remaining_bytes(
	xfs_inode_t		*ip,
	xfs_off_t		startoff,
	xfs_off_t		endoff)
{
	xfs_bmbt_irec_t		imap;
	xfs_fileoff_t		offset_fsb;
	xfs_off_t		lastoffset;
	xfs_off_t		offset;
	xfs_buf_t		*bp;
	xfs_mount_t		*mp = ip->i_mount;
	int			nimap;
	int			error = 0;

3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
	/*
	 * Avoid doing I/O beyond eof - it's not necessary
	 * since nothing can read beyond eof.  The space will
	 * be zeroed when the file is extended anyway.
	 */
	if (startoff >= ip->i_size)
		return 0;

	if (endoff > ip->i_size)
		endoff = ip->i_size;

L
Linus Torvalds 已提交
3034
	bp = xfs_buf_get_noaddr(mp->m_sb.sb_blocksize,
3035
				XFS_IS_REALTIME_INODE(ip) ?
L
Linus Torvalds 已提交
3036 3037 3038 3039 3040
				mp->m_rtdev_targp : mp->m_ddev_targp);

	for (offset = startoff; offset <= endoff; offset = lastoffset + 1) {
		offset_fsb = XFS_B_TO_FSBT(mp, offset);
		nimap = 1;
3041
		error = xfs_bmapi(NULL, ip, offset_fsb, 1, 0,
3042
			NULL, 0, &imap, &nimap, NULL, NULL);
L
Linus Torvalds 已提交
3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059
		if (error || nimap < 1)
			break;
		ASSERT(imap.br_blockcount >= 1);
		ASSERT(imap.br_startoff == offset_fsb);
		lastoffset = XFS_FSB_TO_B(mp, imap.br_startoff + 1) - 1;
		if (lastoffset > endoff)
			lastoffset = endoff;
		if (imap.br_startblock == HOLESTARTBLOCK)
			continue;
		ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
		if (imap.br_state == XFS_EXT_UNWRITTEN)
			continue;
		XFS_BUF_UNDONE(bp);
		XFS_BUF_UNWRITE(bp);
		XFS_BUF_READ(bp);
		XFS_BUF_SET_ADDR(bp, XFS_FSB_TO_DB(ip, imap.br_startblock));
		xfsbdstrat(mp, bp);
3060 3061
		error = xfs_iowait(bp);
		if (error) {
L
Linus Torvalds 已提交
3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072
			xfs_ioerror_alert("xfs_zero_remaining_bytes(read)",
					  mp, bp, XFS_BUF_ADDR(bp));
			break;
		}
		memset(XFS_BUF_PTR(bp) +
			(offset - XFS_FSB_TO_B(mp, imap.br_startoff)),
		      0, lastoffset - offset + 1);
		XFS_BUF_UNDONE(bp);
		XFS_BUF_UNREAD(bp);
		XFS_BUF_WRITE(bp);
		xfsbdstrat(mp, bp);
3073 3074
		error = xfs_iowait(bp);
		if (error) {
L
Linus Torvalds 已提交
3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115
			xfs_ioerror_alert("xfs_zero_remaining_bytes(write)",
					  mp, bp, XFS_BUF_ADDR(bp));
			break;
		}
	}
	xfs_buf_free(bp);
	return error;
}

/*
 * xfs_free_file_space()
 *      This routine frees disk space for the given file.
 *
 *	This routine is only called by xfs_change_file_space
 *	for an UNRESVSP type call.
 *
 * RETURNS:
 *       0 on success
 *      errno on error
 *
 */
STATIC int
xfs_free_file_space(
	xfs_inode_t		*ip,
	xfs_off_t		offset,
	xfs_off_t		len,
	int			attr_flags)
{
	int			committed;
	int			done;
	xfs_off_t		end_dmi_offset;
	xfs_fileoff_t		endoffset_fsb;
	int			error;
	xfs_fsblock_t		firstfsb;
	xfs_bmap_free_t		free_list;
	xfs_bmbt_irec_t		imap;
	xfs_off_t		ioffset;
	xfs_extlen_t		mod=0;
	xfs_mount_t		*mp;
	int			nimap;
	uint			resblks;
3116
	uint			rounding;
L
Linus Torvalds 已提交
3117 3118 3119
	int			rt;
	xfs_fileoff_t		startoffset_fsb;
	xfs_trans_t		*tp;
3120
	int			need_iolock = 1;
L
Linus Torvalds 已提交
3121 3122 3123

	mp = ip->i_mount;

3124
	xfs_itrace_entry(ip);
3125

L
Linus Torvalds 已提交
3126 3127 3128 3129 3130 3131
	if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
		return error;

	error = 0;
	if (len <= 0)	/* if nothing being freed */
		return error;
3132
	rt = XFS_IS_REALTIME_INODE(ip);
L
Linus Torvalds 已提交
3133 3134 3135 3136
	startoffset_fsb	= XFS_B_TO_FSB(mp, offset);
	end_dmi_offset = offset + len;
	endoffset_fsb = XFS_B_TO_FSBT(mp, end_dmi_offset);

3137
	if (offset < ip->i_size && (attr_flags & XFS_ATTR_DMI) == 0 &&
3138
	    DM_EVENT_ENABLED(ip, DM_EVENT_WRITE)) {
3139 3140
		if (end_dmi_offset > ip->i_size)
			end_dmi_offset = ip->i_size;
3141
		error = XFS_SEND_DATA(mp, DM_EVENT_WRITE, ip,
L
Linus Torvalds 已提交
3142 3143 3144
				offset, end_dmi_offset - offset,
				AT_DELAY_FLAG(attr_flags), NULL);
		if (error)
3145
			return error;
L
Linus Torvalds 已提交
3146 3147
	}

3148
	if (attr_flags & XFS_ATTR_NOLOCK)
3149
		need_iolock = 0;
3150
	if (need_iolock) {
L
Linus Torvalds 已提交
3151
		xfs_ilock(ip, XFS_IOLOCK_EXCL);
3152
		vn_iowait(ip);	/* wait for the completion of any pending DIOs */
3153
	}
3154

3155
	rounding = max_t(uint, 1 << mp->m_sb.sb_blocklog, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
3156
	ioffset = offset & ~(rounding - 1);
3157

3158
	if (VN_CACHED(VFS_I(ip)) != 0) {
3159 3160
		xfs_inval_cached_trace(ip, ioffset, -1, ioffset, -1);
		error = xfs_flushinval_pages(ip, ioffset, -1, FI_REMAPF_LOCKED);
3161 3162
		if (error)
			goto out_unlock_iolock;
3163 3164
	}

L
Linus Torvalds 已提交
3165 3166 3167 3168 3169 3170
	/*
	 * Need to zero the stuff we're not freeing, on disk.
	 * If its a realtime file & can't use unwritten extents then we
	 * actually need to zero the extent edges.  Otherwise xfs_bunmapi
	 * will take care of it for us.
	 */
3171
	if (rt && !xfs_sb_version_hasextflgbit(&mp->m_sb)) {
L
Linus Torvalds 已提交
3172
		nimap = 1;
3173
		error = xfs_bmapi(NULL, ip, startoffset_fsb,
3174
			1, 0, NULL, 0, &imap, &nimap, NULL, NULL);
L
Linus Torvalds 已提交
3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187
		if (error)
			goto out_unlock_iolock;
		ASSERT(nimap == 0 || nimap == 1);
		if (nimap && imap.br_startblock != HOLESTARTBLOCK) {
			xfs_daddr_t	block;

			ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
			block = imap.br_startblock;
			mod = do_div(block, mp->m_sb.sb_rextsize);
			if (mod)
				startoffset_fsb += mp->m_sb.sb_rextsize - mod;
		}
		nimap = 1;
3188
		error = xfs_bmapi(NULL, ip, endoffset_fsb - 1,
3189
			1, 0, NULL, 0, &imap, &nimap, NULL, NULL);
L
Linus Torvalds 已提交
3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225
		if (error)
			goto out_unlock_iolock;
		ASSERT(nimap == 0 || nimap == 1);
		if (nimap && imap.br_startblock != HOLESTARTBLOCK) {
			ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
			mod++;
			if (mod && (mod != mp->m_sb.sb_rextsize))
				endoffset_fsb -= mod;
		}
	}
	if ((done = (endoffset_fsb <= startoffset_fsb)))
		/*
		 * One contiguous piece to clear
		 */
		error = xfs_zero_remaining_bytes(ip, offset, offset + len - 1);
	else {
		/*
		 * Some full blocks, possibly two pieces to clear
		 */
		if (offset < XFS_FSB_TO_B(mp, startoffset_fsb))
			error = xfs_zero_remaining_bytes(ip, offset,
				XFS_FSB_TO_B(mp, startoffset_fsb) - 1);
		if (!error &&
		    XFS_FSB_TO_B(mp, endoffset_fsb) < offset + len)
			error = xfs_zero_remaining_bytes(ip,
				XFS_FSB_TO_B(mp, endoffset_fsb),
				offset + len - 1);
	}

	/*
	 * free file space until done or until there is an error
	 */
	resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0);
	while (!error && !done) {

		/*
3226 3227 3228
		 * allocate and setup the transaction. Allow this
		 * transaction to dip into the reserve blocks to ensure
		 * the freeing of the space succeeds at ENOSPC.
L
Linus Torvalds 已提交
3229 3230
		 */
		tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT);
3231
		tp->t_flags |= XFS_TRANS_RESERVE;
L
Linus Torvalds 已提交
3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251
		error = xfs_trans_reserve(tp,
					  resblks,
					  XFS_WRITE_LOG_RES(mp),
					  0,
					  XFS_TRANS_PERM_LOG_RES,
					  XFS_WRITE_LOG_COUNT);

		/*
		 * check for running out of space
		 */
		if (error) {
			/*
			 * Free the transaction structure.
			 */
			ASSERT(error == ENOSPC || XFS_FORCED_SHUTDOWN(mp));
			xfs_trans_cancel(tp, 0);
			break;
		}
		xfs_ilock(ip, XFS_ILOCK_EXCL);
		error = XFS_TRANS_RESERVE_QUOTA(mp, tp,
3252 3253
				ip->i_udquot, ip->i_gdquot, resblks, 0,
				XFS_QMOPT_RES_REGBLKS);
L
Linus Torvalds 已提交
3254 3255 3256 3257 3258 3259 3260 3261 3262 3263
		if (error)
			goto error1;

		xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);

		/*
		 * issue the bunmapi() call to free the blocks
		 */
		XFS_BMAP_INIT(&free_list, &firstfsb);
3264
		error = xfs_bunmapi(tp, ip, startoffset_fsb,
L
Linus Torvalds 已提交
3265
				  endoffset_fsb - startoffset_fsb,
3266
				  0, 2, &firstfsb, &free_list, NULL, &done);
L
Linus Torvalds 已提交
3267 3268 3269 3270 3271 3272 3273
		if (error) {
			goto error0;
		}

		/*
		 * complete the transaction
		 */
3274
		error = xfs_bmap_finish(&tp, &free_list, &committed);
L
Linus Torvalds 已提交
3275 3276 3277 3278
		if (error) {
			goto error0;
		}

3279
		error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
L
Linus Torvalds 已提交
3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309
		xfs_iunlock(ip, XFS_ILOCK_EXCL);
	}

 out_unlock_iolock:
	if (need_iolock)
		xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	return error;

 error0:
	xfs_bmap_cancel(&free_list);
 error1:
	xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
	xfs_iunlock(ip, need_iolock ? (XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL) :
		    XFS_ILOCK_EXCL);
	return error;
}

/*
 * xfs_change_file_space()
 *      This routine allocates or frees disk space for the given file.
 *      The user specified parameters are checked for alignment and size
 *      limitations.
 *
 * RETURNS:
 *       0 on success
 *      errno on error
 *
 */
int
xfs_change_file_space(
3310
	xfs_inode_t	*ip,
L
Linus Torvalds 已提交
3311 3312 3313 3314 3315
	int		cmd,
	xfs_flock64_t	*bf,
	xfs_off_t	offset,
	int		attr_flags)
{
3316
	xfs_mount_t	*mp = ip->i_mount;
L
Linus Torvalds 已提交
3317 3318 3319 3320 3321 3322 3323
	int		clrprealloc;
	int		error;
	xfs_fsize_t	fsize;
	int		setprealloc;
	xfs_off_t	startoffset;
	xfs_off_t	llen;
	xfs_trans_t	*tp;
3324
	struct iattr	iattr;
L
Linus Torvalds 已提交
3325

3326
	xfs_itrace_entry(ip);
L
Linus Torvalds 已提交
3327

3328
	if (!S_ISREG(ip->i_d.di_mode))
L
Linus Torvalds 已提交
3329 3330 3331 3332 3333 3334 3335 3336 3337
		return XFS_ERROR(EINVAL);

	switch (bf->l_whence) {
	case 0: /*SEEK_SET*/
		break;
	case 1: /*SEEK_CUR*/
		bf->l_start += offset;
		break;
	case 2: /*SEEK_END*/
3338
		bf->l_start += ip->i_size;
L
Linus Torvalds 已提交
3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354
		break;
	default:
		return XFS_ERROR(EINVAL);
	}

	llen = bf->l_len > 0 ? bf->l_len - 1 : bf->l_len;

	if (   (bf->l_start < 0)
	    || (bf->l_start > XFS_MAXIOFFSET(mp))
	    || (bf->l_start + llen < 0)
	    || (bf->l_start + llen > XFS_MAXIOFFSET(mp)))
		return XFS_ERROR(EINVAL);

	bf->l_whence = 0;

	startoffset = bf->l_start;
3355
	fsize = ip->i_size;
L
Linus Torvalds 已提交
3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397

	/*
	 * XFS_IOC_RESVSP and XFS_IOC_UNRESVSP will reserve or unreserve
	 * file space.
	 * These calls do NOT zero the data space allocated to the file,
	 * nor do they change the file size.
	 *
	 * XFS_IOC_ALLOCSP and XFS_IOC_FREESP will allocate and free file
	 * space.
	 * These calls cause the new file data to be zeroed and the file
	 * size to be changed.
	 */
	setprealloc = clrprealloc = 0;

	switch (cmd) {
	case XFS_IOC_RESVSP:
	case XFS_IOC_RESVSP64:
		error = xfs_alloc_file_space(ip, startoffset, bf->l_len,
								1, attr_flags);
		if (error)
			return error;
		setprealloc = 1;
		break;

	case XFS_IOC_UNRESVSP:
	case XFS_IOC_UNRESVSP64:
		if ((error = xfs_free_file_space(ip, startoffset, bf->l_len,
								attr_flags)))
			return error;
		break;

	case XFS_IOC_ALLOCSP:
	case XFS_IOC_ALLOCSP64:
	case XFS_IOC_FREESP:
	case XFS_IOC_FREESP64:
		if (startoffset > fsize) {
			error = xfs_alloc_file_space(ip, fsize,
					startoffset - fsize, 0, attr_flags);
			if (error)
				break;
		}

3398 3399
		iattr.ia_valid = ATTR_SIZE;
		iattr.ia_size = startoffset;
L
Linus Torvalds 已提交
3400

C
Christoph Hellwig 已提交
3401
		error = xfs_setattr(ip, &iattr, attr_flags);
L
Linus Torvalds 已提交
3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430

		if (error)
			return error;

		clrprealloc = 1;
		break;

	default:
		ASSERT(0);
		return XFS_ERROR(EINVAL);
	}

	/*
	 * update the inode timestamp, mode, and prealloc flag bits
	 */
	tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID);

	if ((error = xfs_trans_reserve(tp, 0, XFS_WRITEID_LOG_RES(mp),
				      0, 0, 0))) {
		/* ASSERT(0); */
		xfs_trans_cancel(tp, 0);
		return error;
	}

	xfs_ilock(ip, XFS_ILOCK_EXCL);

	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	xfs_trans_ihold(tp, ip);

3431
	if ((attr_flags & XFS_ATTR_DMI) == 0) {
L
Linus Torvalds 已提交
3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453
		ip->i_d.di_mode &= ~S_ISUID;

		/*
		 * Note that we don't have to worry about mandatory
		 * file locking being disabled here because we only
		 * clear the S_ISGID bit if the Group execute bit is
		 * on, but if it was on then mandatory locking wouldn't
		 * have been enabled.
		 */
		if (ip->i_d.di_mode & S_IXGRP)
			ip->i_d.di_mode &= ~S_ISGID;

		xfs_ichgtime(ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	}
	if (setprealloc)
		ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC;
	else if (clrprealloc)
		ip->i_d.di_flags &= ~XFS_DIFLAG_PREALLOC;

	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	xfs_trans_set_sync(tp);

3454
	error = xfs_trans_commit(tp, 0);
L
Linus Torvalds 已提交
3455 3456 3457 3458 3459

	xfs_iunlock(ip, XFS_ILOCK_EXCL);

	return error;
}