xref: /linux/fs/ocfs2/namei.c (revision ca55b2fef3a9373fcfc30f82fd26bc7fccbda732)
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * namei.c
5  *
6  * Create and rename file, directory, symlinks
7  *
8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
9  *
10  *  Portions of this code from linux/fs/ext3/dir.c
11  *
12  *  Copyright (C) 1992, 1993, 1994, 1995
13  *  Remy Card (card@masi.ibp.fr)
14  *  Laboratoire MASI - Institut Blaise pascal
15  *  Universite Pierre et Marie Curie (Paris VI)
16  *
17  *   from
18  *
19  *   linux/fs/minix/dir.c
20  *
21  *   Copyright (C) 1991, 1992 Linux Torvalds
22  *
23  * This program is free software; you can redistribute it and/or
24  * modify it under the terms of the GNU General Public
25  * License as published by the Free Software Foundation; either
26  * version 2 of the License, or (at your option) any later version.
27  *
28  * This program is distributed in the hope that it will be useful,
29  * but WITHOUT ANY WARRANTY; without even the implied warranty of
30  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
31  * General Public License for more details.
32  *
33  * You should have received a copy of the GNU General Public
34  * License along with this program; if not, write to the
35  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
36  * Boston, MA 021110-1307, USA.
37  */
38 
39 #include <linux/fs.h>
40 #include <linux/types.h>
41 #include <linux/slab.h>
42 #include <linux/highmem.h>
43 #include <linux/quotaops.h>
44 
45 #include <cluster/masklog.h>
46 
47 #include "ocfs2.h"
48 
49 #include "alloc.h"
50 #include "dcache.h"
51 #include "dir.h"
52 #include "dlmglue.h"
53 #include "extent_map.h"
54 #include "file.h"
55 #include "inode.h"
56 #include "journal.h"
57 #include "namei.h"
58 #include "suballoc.h"
59 #include "super.h"
60 #include "symlink.h"
61 #include "sysfile.h"
62 #include "uptodate.h"
63 #include "xattr.h"
64 #include "acl.h"
65 #include "ocfs2_trace.h"
66 
67 #include "buffer_head_io.h"
68 
69 static int ocfs2_mknod_locked(struct ocfs2_super *osb,
70 			      struct inode *dir,
71 			      struct inode *inode,
72 			      dev_t dev,
73 			      struct buffer_head **new_fe_bh,
74 			      struct buffer_head *parent_fe_bh,
75 			      handle_t *handle,
76 			      struct ocfs2_alloc_context *inode_ac);
77 
78 static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
79 				    struct inode **ret_orphan_dir,
80 				    u64 blkno,
81 				    char *name,
82 				    struct ocfs2_dir_lookup_result *lookup,
83 				    bool dio);
84 
85 static int ocfs2_orphan_add(struct ocfs2_super *osb,
86 			    handle_t *handle,
87 			    struct inode *inode,
88 			    struct buffer_head *fe_bh,
89 			    char *name,
90 			    struct ocfs2_dir_lookup_result *lookup,
91 			    struct inode *orphan_dir_inode,
92 			    bool dio);
93 
94 static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
95 				     handle_t *handle,
96 				     struct inode *inode,
97 				     const char *symname);
98 
99 static int ocfs2_double_lock(struct ocfs2_super *osb,
100 			     struct buffer_head **bh1,
101 			     struct inode *inode1,
102 			     struct buffer_head **bh2,
103 			     struct inode *inode2,
104 			     int rename);
105 
106 static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2);
107 /* An orphan dir name is an 8 byte value, printed as a hex string */
108 #define OCFS2_ORPHAN_NAMELEN ((int)(2 * sizeof(u64)))
109 #define OCFS2_DIO_ORPHAN_PREFIX "dio-"
110 #define OCFS2_DIO_ORPHAN_PREFIX_LEN 4
111 
112 static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry,
113 				   unsigned int flags)
114 {
115 	int status;
116 	u64 blkno;
117 	struct inode *inode = NULL;
118 	struct dentry *ret;
119 	struct ocfs2_inode_info *oi;
120 
121 	trace_ocfs2_lookup(dir, dentry, dentry->d_name.len,
122 			   dentry->d_name.name,
123 			   (unsigned long long)OCFS2_I(dir)->ip_blkno, 0);
124 
125 	if (dentry->d_name.len > OCFS2_MAX_FILENAME_LEN) {
126 		ret = ERR_PTR(-ENAMETOOLONG);
127 		goto bail;
128 	}
129 
130 	status = ocfs2_inode_lock_nested(dir, NULL, 0, OI_LS_PARENT);
131 	if (status < 0) {
132 		if (status != -ENOENT)
133 			mlog_errno(status);
134 		ret = ERR_PTR(status);
135 		goto bail;
136 	}
137 
138 	status = ocfs2_lookup_ino_from_name(dir, dentry->d_name.name,
139 					    dentry->d_name.len, &blkno);
140 	if (status < 0)
141 		goto bail_add;
142 
143 	inode = ocfs2_iget(OCFS2_SB(dir->i_sb), blkno, 0, 0);
144 	if (IS_ERR(inode)) {
145 		ret = ERR_PTR(-EACCES);
146 		goto bail_unlock;
147 	}
148 
149 	oi = OCFS2_I(inode);
150 	/* Clear any orphaned state... If we were able to look up the
151 	 * inode from a directory, it certainly can't be orphaned. We
152 	 * might have the bad state from a node which intended to
153 	 * orphan this inode but crashed before it could commit the
154 	 * unlink. */
155 	spin_lock(&oi->ip_lock);
156 	oi->ip_flags &= ~OCFS2_INODE_MAYBE_ORPHANED;
157 	spin_unlock(&oi->ip_lock);
158 
159 bail_add:
160 	ret = d_splice_alias(inode, dentry);
161 
162 	if (inode) {
163 		/*
164 		 * If d_splice_alias() finds a DCACHE_DISCONNECTED
165 		 * dentry, it will d_move() it on top of ourse. The
166 		 * return value will indicate this however, so in
167 		 * those cases, we switch them around for the locking
168 		 * code.
169 		 *
170 		 * NOTE: This dentry already has ->d_op set from
171 		 * ocfs2_get_parent() and ocfs2_get_dentry()
172 		 */
173 		if (!IS_ERR_OR_NULL(ret))
174 			dentry = ret;
175 
176 		status = ocfs2_dentry_attach_lock(dentry, inode,
177 						  OCFS2_I(dir)->ip_blkno);
178 		if (status) {
179 			mlog_errno(status);
180 			ret = ERR_PTR(status);
181 			goto bail_unlock;
182 		}
183 	} else
184 		ocfs2_dentry_attach_gen(dentry);
185 
186 bail_unlock:
187 	/* Don't drop the cluster lock until *after* the d_add --
188 	 * unlink on another node will message us to remove that
189 	 * dentry under this lock so otherwise we can race this with
190 	 * the downconvert thread and have a stale dentry. */
191 	ocfs2_inode_unlock(dir, 0);
192 
193 bail:
194 
195 	trace_ocfs2_lookup_ret(ret);
196 
197 	return ret;
198 }
199 
200 static struct inode *ocfs2_get_init_inode(struct inode *dir, umode_t mode)
201 {
202 	struct inode *inode;
203 	int status;
204 
205 	inode = new_inode(dir->i_sb);
206 	if (!inode) {
207 		mlog(ML_ERROR, "new_inode failed!\n");
208 		return ERR_PTR(-ENOMEM);
209 	}
210 
211 	/* populate as many fields early on as possible - many of
212 	 * these are used by the support functions here and in
213 	 * callers. */
214 	if (S_ISDIR(mode))
215 		set_nlink(inode, 2);
216 	inode_init_owner(inode, dir, mode);
217 	status = dquot_initialize(inode);
218 	if (status)
219 		return ERR_PTR(status);
220 
221 	return inode;
222 }
223 
224 static void ocfs2_cleanup_add_entry_failure(struct ocfs2_super *osb,
225 		struct dentry *dentry, struct inode *inode)
226 {
227 	struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
228 
229 	ocfs2_simple_drop_lockres(osb, &dl->dl_lockres);
230 	ocfs2_lock_res_free(&dl->dl_lockres);
231 	BUG_ON(dl->dl_count != 1);
232 	spin_lock(&dentry_attach_lock);
233 	dentry->d_fsdata = NULL;
234 	spin_unlock(&dentry_attach_lock);
235 	kfree(dl);
236 	iput(inode);
237 }
238 
239 static int ocfs2_mknod(struct inode *dir,
240 		       struct dentry *dentry,
241 		       umode_t mode,
242 		       dev_t dev)
243 {
244 	int status = 0;
245 	struct buffer_head *parent_fe_bh = NULL;
246 	handle_t *handle = NULL;
247 	struct ocfs2_super *osb;
248 	struct ocfs2_dinode *dirfe;
249 	struct buffer_head *new_fe_bh = NULL;
250 	struct inode *inode = NULL;
251 	struct ocfs2_alloc_context *inode_ac = NULL;
252 	struct ocfs2_alloc_context *data_ac = NULL;
253 	struct ocfs2_alloc_context *meta_ac = NULL;
254 	int want_clusters = 0;
255 	int want_meta = 0;
256 	int xattr_credits = 0;
257 	struct ocfs2_security_xattr_info si = {
258 		.enable = 1,
259 	};
260 	int did_quota_inode = 0;
261 	struct ocfs2_dir_lookup_result lookup = { NULL, };
262 	sigset_t oldset;
263 	int did_block_signals = 0;
264 	struct posix_acl *default_acl = NULL, *acl = NULL;
265 	struct ocfs2_dentry_lock *dl = NULL;
266 
267 	trace_ocfs2_mknod(dir, dentry, dentry->d_name.len, dentry->d_name.name,
268 			  (unsigned long long)OCFS2_I(dir)->ip_blkno,
269 			  (unsigned long)dev, mode);
270 
271 	status = dquot_initialize(dir);
272 	if (status) {
273 		mlog_errno(status);
274 		return status;
275 	}
276 
277 	/* get our super block */
278 	osb = OCFS2_SB(dir->i_sb);
279 
280 	status = ocfs2_inode_lock(dir, &parent_fe_bh, 1);
281 	if (status < 0) {
282 		if (status != -ENOENT)
283 			mlog_errno(status);
284 		return status;
285 	}
286 
287 	if (S_ISDIR(mode) && (dir->i_nlink >= ocfs2_link_max(osb))) {
288 		status = -EMLINK;
289 		goto leave;
290 	}
291 
292 	dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
293 	if (!ocfs2_read_links_count(dirfe)) {
294 		/* can't make a file in a deleted directory. */
295 		status = -ENOENT;
296 		goto leave;
297 	}
298 
299 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
300 					   dentry->d_name.len);
301 	if (status)
302 		goto leave;
303 
304 	/* get a spot inside the dir. */
305 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
306 					      dentry->d_name.name,
307 					      dentry->d_name.len, &lookup);
308 	if (status < 0) {
309 		mlog_errno(status);
310 		goto leave;
311 	}
312 
313 	/* reserve an inode spot */
314 	status = ocfs2_reserve_new_inode(osb, &inode_ac);
315 	if (status < 0) {
316 		if (status != -ENOSPC)
317 			mlog_errno(status);
318 		goto leave;
319 	}
320 
321 	inode = ocfs2_get_init_inode(dir, mode);
322 	if (IS_ERR(inode)) {
323 		status = PTR_ERR(inode);
324 		inode = NULL;
325 		mlog_errno(status);
326 		goto leave;
327 	}
328 
329 	/* get security xattr */
330 	status = ocfs2_init_security_get(inode, dir, &dentry->d_name, &si);
331 	if (status) {
332 		if (status == -EOPNOTSUPP)
333 			si.enable = 0;
334 		else {
335 			mlog_errno(status);
336 			goto leave;
337 		}
338 	}
339 
340 	/* calculate meta data/clusters for setting security and acl xattr */
341 	status = ocfs2_calc_xattr_init(dir, parent_fe_bh, mode,
342 				       &si, &want_clusters,
343 				       &xattr_credits, &want_meta);
344 	if (status < 0) {
345 		mlog_errno(status);
346 		goto leave;
347 	}
348 
349 	/* Reserve a cluster if creating an extent based directory. */
350 	if (S_ISDIR(mode) && !ocfs2_supports_inline_data(osb)) {
351 		want_clusters += 1;
352 
353 		/* Dir indexing requires extra space as well */
354 		if (ocfs2_supports_indexed_dirs(osb))
355 			want_meta++;
356 	}
357 
358 	status = ocfs2_reserve_new_metadata_blocks(osb, want_meta, &meta_ac);
359 	if (status < 0) {
360 		if (status != -ENOSPC)
361 			mlog_errno(status);
362 		goto leave;
363 	}
364 
365 	status = ocfs2_reserve_clusters(osb, want_clusters, &data_ac);
366 	if (status < 0) {
367 		if (status != -ENOSPC)
368 			mlog_errno(status);
369 		goto leave;
370 	}
371 
372 	status = posix_acl_create(dir, &mode, &default_acl, &acl);
373 	if (status) {
374 		mlog_errno(status);
375 		goto leave;
376 	}
377 
378 	handle = ocfs2_start_trans(osb, ocfs2_mknod_credits(osb->sb,
379 							    S_ISDIR(mode),
380 							    xattr_credits));
381 	if (IS_ERR(handle)) {
382 		status = PTR_ERR(handle);
383 		handle = NULL;
384 		mlog_errno(status);
385 		goto leave;
386 	}
387 
388 	/* Starting to change things, restart is no longer possible. */
389 	ocfs2_block_signals(&oldset);
390 	did_block_signals = 1;
391 
392 	status = dquot_alloc_inode(inode);
393 	if (status)
394 		goto leave;
395 	did_quota_inode = 1;
396 
397 	/* do the real work now. */
398 	status = ocfs2_mknod_locked(osb, dir, inode, dev,
399 				    &new_fe_bh, parent_fe_bh, handle,
400 				    inode_ac);
401 	if (status < 0) {
402 		mlog_errno(status);
403 		goto leave;
404 	}
405 
406 	if (S_ISDIR(mode)) {
407 		status = ocfs2_fill_new_dir(osb, handle, dir, inode,
408 					    new_fe_bh, data_ac, meta_ac);
409 		if (status < 0) {
410 			mlog_errno(status);
411 			goto leave;
412 		}
413 
414 		status = ocfs2_journal_access_di(handle, INODE_CACHE(dir),
415 						 parent_fe_bh,
416 						 OCFS2_JOURNAL_ACCESS_WRITE);
417 		if (status < 0) {
418 			mlog_errno(status);
419 			goto leave;
420 		}
421 		ocfs2_add_links_count(dirfe, 1);
422 		ocfs2_journal_dirty(handle, parent_fe_bh);
423 		inc_nlink(dir);
424 	}
425 
426 	if (default_acl) {
427 		status = ocfs2_set_acl(handle, inode, new_fe_bh,
428 				       ACL_TYPE_DEFAULT, default_acl,
429 				       meta_ac, data_ac);
430 	}
431 	if (!status && acl) {
432 		status = ocfs2_set_acl(handle, inode, new_fe_bh,
433 				       ACL_TYPE_ACCESS, acl,
434 				       meta_ac, data_ac);
435 	}
436 
437 	if (status < 0) {
438 		mlog_errno(status);
439 		goto leave;
440 	}
441 
442 	if (si.enable) {
443 		status = ocfs2_init_security_set(handle, inode, new_fe_bh, &si,
444 						 meta_ac, data_ac);
445 		if (status < 0) {
446 			mlog_errno(status);
447 			goto leave;
448 		}
449 	}
450 
451 	/*
452 	 * Do this before adding the entry to the directory. We add
453 	 * also set d_op after success so that ->d_iput() will cleanup
454 	 * the dentry lock even if ocfs2_add_entry() fails below.
455 	 */
456 	status = ocfs2_dentry_attach_lock(dentry, inode,
457 					  OCFS2_I(dir)->ip_blkno);
458 	if (status) {
459 		mlog_errno(status);
460 		goto leave;
461 	}
462 
463 	dl = dentry->d_fsdata;
464 
465 	status = ocfs2_add_entry(handle, dentry, inode,
466 				 OCFS2_I(inode)->ip_blkno, parent_fe_bh,
467 				 &lookup);
468 	if (status < 0) {
469 		mlog_errno(status);
470 		goto leave;
471 	}
472 
473 	insert_inode_hash(inode);
474 	d_instantiate(dentry, inode);
475 	status = 0;
476 leave:
477 	if (default_acl)
478 		posix_acl_release(default_acl);
479 	if (acl)
480 		posix_acl_release(acl);
481 	if (status < 0 && did_quota_inode)
482 		dquot_free_inode(inode);
483 	if (handle)
484 		ocfs2_commit_trans(osb, handle);
485 
486 	ocfs2_inode_unlock(dir, 1);
487 	if (did_block_signals)
488 		ocfs2_unblock_signals(&oldset);
489 
490 	brelse(new_fe_bh);
491 	brelse(parent_fe_bh);
492 	kfree(si.value);
493 
494 	ocfs2_free_dir_lookup_result(&lookup);
495 
496 	if (inode_ac)
497 		ocfs2_free_alloc_context(inode_ac);
498 
499 	if (data_ac)
500 		ocfs2_free_alloc_context(data_ac);
501 
502 	if (meta_ac)
503 		ocfs2_free_alloc_context(meta_ac);
504 
505 	/*
506 	 * We should call iput after the i_mutex of the bitmap been
507 	 * unlocked in ocfs2_free_alloc_context, or the
508 	 * ocfs2_delete_inode will mutex_lock again.
509 	 */
510 	if ((status < 0) && inode) {
511 		if (dl)
512 			ocfs2_cleanup_add_entry_failure(osb, dentry, inode);
513 
514 		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR;
515 		clear_nlink(inode);
516 		iput(inode);
517 	}
518 
519 	if (status)
520 		mlog_errno(status);
521 
522 	return status;
523 }
524 
525 static int __ocfs2_mknod_locked(struct inode *dir,
526 				struct inode *inode,
527 				dev_t dev,
528 				struct buffer_head **new_fe_bh,
529 				struct buffer_head *parent_fe_bh,
530 				handle_t *handle,
531 				struct ocfs2_alloc_context *inode_ac,
532 				u64 fe_blkno, u64 suballoc_loc, u16 suballoc_bit)
533 {
534 	int status = 0;
535 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
536 	struct ocfs2_dinode *fe = NULL;
537 	struct ocfs2_extent_list *fel;
538 	u16 feat;
539 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
540 
541 	*new_fe_bh = NULL;
542 
543 	/* populate as many fields early on as possible - many of
544 	 * these are used by the support functions here and in
545 	 * callers. */
546 	inode->i_ino = ino_from_blkno(osb->sb, fe_blkno);
547 	OCFS2_I(inode)->ip_blkno = fe_blkno;
548 	spin_lock(&osb->osb_lock);
549 	inode->i_generation = osb->s_next_generation++;
550 	spin_unlock(&osb->osb_lock);
551 
552 	*new_fe_bh = sb_getblk(osb->sb, fe_blkno);
553 	if (!*new_fe_bh) {
554 		status = -ENOMEM;
555 		mlog_errno(status);
556 		goto leave;
557 	}
558 	ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), *new_fe_bh);
559 
560 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
561 					 *new_fe_bh,
562 					 OCFS2_JOURNAL_ACCESS_CREATE);
563 	if (status < 0) {
564 		mlog_errno(status);
565 		goto leave;
566 	}
567 
568 	fe = (struct ocfs2_dinode *) (*new_fe_bh)->b_data;
569 	memset(fe, 0, osb->sb->s_blocksize);
570 
571 	fe->i_generation = cpu_to_le32(inode->i_generation);
572 	fe->i_fs_generation = cpu_to_le32(osb->fs_generation);
573 	fe->i_blkno = cpu_to_le64(fe_blkno);
574 	fe->i_suballoc_loc = cpu_to_le64(suballoc_loc);
575 	fe->i_suballoc_bit = cpu_to_le16(suballoc_bit);
576 	fe->i_suballoc_slot = cpu_to_le16(inode_ac->ac_alloc_slot);
577 	fe->i_uid = cpu_to_le32(i_uid_read(inode));
578 	fe->i_gid = cpu_to_le32(i_gid_read(inode));
579 	fe->i_mode = cpu_to_le16(inode->i_mode);
580 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
581 		fe->id1.dev1.i_rdev = cpu_to_le64(huge_encode_dev(dev));
582 
583 	ocfs2_set_links_count(fe, inode->i_nlink);
584 
585 	fe->i_last_eb_blk = 0;
586 	strcpy(fe->i_signature, OCFS2_INODE_SIGNATURE);
587 	fe->i_flags |= cpu_to_le32(OCFS2_VALID_FL);
588 	fe->i_atime = fe->i_ctime = fe->i_mtime =
589 		cpu_to_le64(CURRENT_TIME.tv_sec);
590 	fe->i_mtime_nsec = fe->i_ctime_nsec = fe->i_atime_nsec =
591 		cpu_to_le32(CURRENT_TIME.tv_nsec);
592 	fe->i_dtime = 0;
593 
594 	/*
595 	 * If supported, directories start with inline data. If inline
596 	 * isn't supported, but indexing is, we start them as indexed.
597 	 */
598 	feat = le16_to_cpu(fe->i_dyn_features);
599 	if (S_ISDIR(inode->i_mode) && ocfs2_supports_inline_data(osb)) {
600 		fe->i_dyn_features = cpu_to_le16(feat | OCFS2_INLINE_DATA_FL);
601 
602 		fe->id2.i_data.id_count = cpu_to_le16(
603 				ocfs2_max_inline_data_with_xattr(osb->sb, fe));
604 	} else {
605 		fel = &fe->id2.i_list;
606 		fel->l_tree_depth = 0;
607 		fel->l_next_free_rec = 0;
608 		fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb));
609 	}
610 
611 	ocfs2_journal_dirty(handle, *new_fe_bh);
612 
613 	ocfs2_populate_inode(inode, fe, 1);
614 	ocfs2_ci_set_new(osb, INODE_CACHE(inode));
615 	if (!ocfs2_mount_local(osb)) {
616 		status = ocfs2_create_new_inode_locks(inode);
617 		if (status < 0)
618 			mlog_errno(status);
619 	}
620 
621 	oi->i_sync_tid = handle->h_transaction->t_tid;
622 	oi->i_datasync_tid = handle->h_transaction->t_tid;
623 
624 leave:
625 	if (status < 0) {
626 		if (*new_fe_bh) {
627 			brelse(*new_fe_bh);
628 			*new_fe_bh = NULL;
629 		}
630 	}
631 
632 	if (status)
633 		mlog_errno(status);
634 	return status;
635 }
636 
637 static int ocfs2_mknod_locked(struct ocfs2_super *osb,
638 			      struct inode *dir,
639 			      struct inode *inode,
640 			      dev_t dev,
641 			      struct buffer_head **new_fe_bh,
642 			      struct buffer_head *parent_fe_bh,
643 			      handle_t *handle,
644 			      struct ocfs2_alloc_context *inode_ac)
645 {
646 	int status = 0;
647 	u64 suballoc_loc, fe_blkno = 0;
648 	u16 suballoc_bit;
649 
650 	*new_fe_bh = NULL;
651 
652 	status = ocfs2_claim_new_inode(handle, dir, parent_fe_bh,
653 				       inode_ac, &suballoc_loc,
654 				       &suballoc_bit, &fe_blkno);
655 	if (status < 0) {
656 		mlog_errno(status);
657 		return status;
658 	}
659 
660 	return __ocfs2_mknod_locked(dir, inode, dev, new_fe_bh,
661 				    parent_fe_bh, handle, inode_ac,
662 				    fe_blkno, suballoc_loc, suballoc_bit);
663 }
664 
665 static int ocfs2_mkdir(struct inode *dir,
666 		       struct dentry *dentry,
667 		       umode_t mode)
668 {
669 	int ret;
670 
671 	trace_ocfs2_mkdir(dir, dentry, dentry->d_name.len, dentry->d_name.name,
672 			  OCFS2_I(dir)->ip_blkno, mode);
673 	ret = ocfs2_mknod(dir, dentry, mode | S_IFDIR, 0);
674 	if (ret)
675 		mlog_errno(ret);
676 
677 	return ret;
678 }
679 
680 static int ocfs2_create(struct inode *dir,
681 			struct dentry *dentry,
682 			umode_t mode,
683 			bool excl)
684 {
685 	int ret;
686 
687 	trace_ocfs2_create(dir, dentry, dentry->d_name.len, dentry->d_name.name,
688 			   (unsigned long long)OCFS2_I(dir)->ip_blkno, mode);
689 	ret = ocfs2_mknod(dir, dentry, mode | S_IFREG, 0);
690 	if (ret)
691 		mlog_errno(ret);
692 
693 	return ret;
694 }
695 
696 static int ocfs2_link(struct dentry *old_dentry,
697 		      struct inode *dir,
698 		      struct dentry *dentry)
699 {
700 	handle_t *handle;
701 	struct inode *inode = d_inode(old_dentry);
702 	struct inode *old_dir = d_inode(old_dentry->d_parent);
703 	int err;
704 	struct buffer_head *fe_bh = NULL;
705 	struct buffer_head *old_dir_bh = NULL;
706 	struct buffer_head *parent_fe_bh = NULL;
707 	struct ocfs2_dinode *fe = NULL;
708 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
709 	struct ocfs2_dir_lookup_result lookup = { NULL, };
710 	sigset_t oldset;
711 	u64 old_de_ino;
712 
713 	trace_ocfs2_link((unsigned long long)OCFS2_I(inode)->ip_blkno,
714 			 old_dentry->d_name.len, old_dentry->d_name.name,
715 			 dentry->d_name.len, dentry->d_name.name);
716 
717 	if (S_ISDIR(inode->i_mode))
718 		return -EPERM;
719 
720 	err = dquot_initialize(dir);
721 	if (err) {
722 		mlog_errno(err);
723 		return err;
724 	}
725 
726 	err = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
727 			&parent_fe_bh, dir, 0);
728 	if (err < 0) {
729 		if (err != -ENOENT)
730 			mlog_errno(err);
731 		return err;
732 	}
733 
734 	/* make sure both dirs have bhs
735 	 * get an extra ref on old_dir_bh if old==new */
736 	if (!parent_fe_bh) {
737 		if (old_dir_bh) {
738 			parent_fe_bh = old_dir_bh;
739 			get_bh(parent_fe_bh);
740 		} else {
741 			mlog(ML_ERROR, "%s: no old_dir_bh!\n", osb->uuid_str);
742 			err = -EIO;
743 			goto out;
744 		}
745 	}
746 
747 	if (!dir->i_nlink) {
748 		err = -ENOENT;
749 		goto out;
750 	}
751 
752 	err = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name,
753 			old_dentry->d_name.len, &old_de_ino);
754 	if (err) {
755 		err = -ENOENT;
756 		goto out;
757 	}
758 
759 	/*
760 	 * Check whether another node removed the source inode while we
761 	 * were in the vfs.
762 	 */
763 	if (old_de_ino != OCFS2_I(inode)->ip_blkno) {
764 		err = -ENOENT;
765 		goto out;
766 	}
767 
768 	err = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
769 					dentry->d_name.len);
770 	if (err)
771 		goto out;
772 
773 	err = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
774 					   dentry->d_name.name,
775 					   dentry->d_name.len, &lookup);
776 	if (err < 0) {
777 		mlog_errno(err);
778 		goto out;
779 	}
780 
781 	err = ocfs2_inode_lock(inode, &fe_bh, 1);
782 	if (err < 0) {
783 		if (err != -ENOENT)
784 			mlog_errno(err);
785 		goto out;
786 	}
787 
788 	fe = (struct ocfs2_dinode *) fe_bh->b_data;
789 	if (ocfs2_read_links_count(fe) >= ocfs2_link_max(osb)) {
790 		err = -EMLINK;
791 		goto out_unlock_inode;
792 	}
793 
794 	handle = ocfs2_start_trans(osb, ocfs2_link_credits(osb->sb));
795 	if (IS_ERR(handle)) {
796 		err = PTR_ERR(handle);
797 		handle = NULL;
798 		mlog_errno(err);
799 		goto out_unlock_inode;
800 	}
801 
802 	/* Starting to change things, restart is no longer possible. */
803 	ocfs2_block_signals(&oldset);
804 
805 	err = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
806 				      OCFS2_JOURNAL_ACCESS_WRITE);
807 	if (err < 0) {
808 		mlog_errno(err);
809 		goto out_commit;
810 	}
811 
812 	inc_nlink(inode);
813 	inode->i_ctime = CURRENT_TIME;
814 	ocfs2_set_links_count(fe, inode->i_nlink);
815 	fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
816 	fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
817 	ocfs2_journal_dirty(handle, fe_bh);
818 
819 	err = ocfs2_add_entry(handle, dentry, inode,
820 			      OCFS2_I(inode)->ip_blkno,
821 			      parent_fe_bh, &lookup);
822 	if (err) {
823 		ocfs2_add_links_count(fe, -1);
824 		drop_nlink(inode);
825 		mlog_errno(err);
826 		goto out_commit;
827 	}
828 
829 	err = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
830 	if (err) {
831 		mlog_errno(err);
832 		goto out_commit;
833 	}
834 
835 	ihold(inode);
836 	d_instantiate(dentry, inode);
837 
838 out_commit:
839 	ocfs2_commit_trans(osb, handle);
840 	ocfs2_unblock_signals(&oldset);
841 out_unlock_inode:
842 	ocfs2_inode_unlock(inode, 1);
843 
844 out:
845 	ocfs2_double_unlock(old_dir, dir);
846 
847 	brelse(fe_bh);
848 	brelse(parent_fe_bh);
849 	brelse(old_dir_bh);
850 
851 	ocfs2_free_dir_lookup_result(&lookup);
852 
853 	if (err)
854 		mlog_errno(err);
855 
856 	return err;
857 }
858 
859 /*
860  * Takes and drops an exclusive lock on the given dentry. This will
861  * force other nodes to drop it.
862  */
863 static int ocfs2_remote_dentry_delete(struct dentry *dentry)
864 {
865 	int ret;
866 
867 	ret = ocfs2_dentry_lock(dentry, 1);
868 	if (ret)
869 		mlog_errno(ret);
870 	else
871 		ocfs2_dentry_unlock(dentry, 1);
872 
873 	return ret;
874 }
875 
876 static inline int ocfs2_inode_is_unlinkable(struct inode *inode)
877 {
878 	if (S_ISDIR(inode->i_mode)) {
879 		if (inode->i_nlink == 2)
880 			return 1;
881 		return 0;
882 	}
883 
884 	if (inode->i_nlink == 1)
885 		return 1;
886 	return 0;
887 }
888 
889 static int ocfs2_unlink(struct inode *dir,
890 			struct dentry *dentry)
891 {
892 	int status;
893 	int child_locked = 0;
894 	bool is_unlinkable = false;
895 	struct inode *inode = d_inode(dentry);
896 	struct inode *orphan_dir = NULL;
897 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
898 	u64 blkno;
899 	struct ocfs2_dinode *fe = NULL;
900 	struct buffer_head *fe_bh = NULL;
901 	struct buffer_head *parent_node_bh = NULL;
902 	handle_t *handle = NULL;
903 	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
904 	struct ocfs2_dir_lookup_result lookup = { NULL, };
905 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
906 
907 	trace_ocfs2_unlink(dir, dentry, dentry->d_name.len,
908 			   dentry->d_name.name,
909 			   (unsigned long long)OCFS2_I(dir)->ip_blkno,
910 			   (unsigned long long)OCFS2_I(inode)->ip_blkno);
911 
912 	status = dquot_initialize(dir);
913 	if (status) {
914 		mlog_errno(status);
915 		return status;
916 	}
917 
918 	BUG_ON(d_inode(dentry->d_parent) != dir);
919 
920 	if (inode == osb->root_inode)
921 		return -EPERM;
922 
923 	status = ocfs2_inode_lock_nested(dir, &parent_node_bh, 1,
924 					 OI_LS_PARENT);
925 	if (status < 0) {
926 		if (status != -ENOENT)
927 			mlog_errno(status);
928 		return status;
929 	}
930 
931 	status = ocfs2_find_files_on_disk(dentry->d_name.name,
932 					  dentry->d_name.len, &blkno, dir,
933 					  &lookup);
934 	if (status < 0) {
935 		if (status != -ENOENT)
936 			mlog_errno(status);
937 		goto leave;
938 	}
939 
940 	if (OCFS2_I(inode)->ip_blkno != blkno) {
941 		status = -ENOENT;
942 
943 		trace_ocfs2_unlink_noent(
944 				(unsigned long long)OCFS2_I(inode)->ip_blkno,
945 				(unsigned long long)blkno,
946 				OCFS2_I(inode)->ip_flags);
947 		goto leave;
948 	}
949 
950 	status = ocfs2_inode_lock(inode, &fe_bh, 1);
951 	if (status < 0) {
952 		if (status != -ENOENT)
953 			mlog_errno(status);
954 		goto leave;
955 	}
956 	child_locked = 1;
957 
958 	if (S_ISDIR(inode->i_mode)) {
959 		if (inode->i_nlink != 2 || !ocfs2_empty_dir(inode)) {
960 			status = -ENOTEMPTY;
961 			goto leave;
962 		}
963 	}
964 
965 	status = ocfs2_remote_dentry_delete(dentry);
966 	if (status < 0) {
967 		/* This remote delete should succeed under all normal
968 		 * circumstances. */
969 		mlog_errno(status);
970 		goto leave;
971 	}
972 
973 	if (ocfs2_inode_is_unlinkable(inode)) {
974 		status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
975 						  OCFS2_I(inode)->ip_blkno,
976 						  orphan_name, &orphan_insert,
977 						  false);
978 		if (status < 0) {
979 			mlog_errno(status);
980 			goto leave;
981 		}
982 		is_unlinkable = true;
983 	}
984 
985 	handle = ocfs2_start_trans(osb, ocfs2_unlink_credits(osb->sb));
986 	if (IS_ERR(handle)) {
987 		status = PTR_ERR(handle);
988 		handle = NULL;
989 		mlog_errno(status);
990 		goto leave;
991 	}
992 
993 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
994 					 OCFS2_JOURNAL_ACCESS_WRITE);
995 	if (status < 0) {
996 		mlog_errno(status);
997 		goto leave;
998 	}
999 
1000 	fe = (struct ocfs2_dinode *) fe_bh->b_data;
1001 
1002 	/* delete the name from the parent dir */
1003 	status = ocfs2_delete_entry(handle, dir, &lookup);
1004 	if (status < 0) {
1005 		mlog_errno(status);
1006 		goto leave;
1007 	}
1008 
1009 	if (S_ISDIR(inode->i_mode))
1010 		drop_nlink(inode);
1011 	drop_nlink(inode);
1012 	ocfs2_set_links_count(fe, inode->i_nlink);
1013 	ocfs2_journal_dirty(handle, fe_bh);
1014 
1015 	dir->i_ctime = dir->i_mtime = CURRENT_TIME;
1016 	if (S_ISDIR(inode->i_mode))
1017 		drop_nlink(dir);
1018 
1019 	status = ocfs2_mark_inode_dirty(handle, dir, parent_node_bh);
1020 	if (status < 0) {
1021 		mlog_errno(status);
1022 		if (S_ISDIR(inode->i_mode))
1023 			inc_nlink(dir);
1024 		goto leave;
1025 	}
1026 
1027 	if (is_unlinkable) {
1028 		status = ocfs2_orphan_add(osb, handle, inode, fe_bh,
1029 				orphan_name, &orphan_insert, orphan_dir, false);
1030 		if (status < 0)
1031 			mlog_errno(status);
1032 	}
1033 
1034 leave:
1035 	if (handle)
1036 		ocfs2_commit_trans(osb, handle);
1037 
1038 	if (orphan_dir) {
1039 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
1040 		ocfs2_inode_unlock(orphan_dir, 1);
1041 		mutex_unlock(&orphan_dir->i_mutex);
1042 		iput(orphan_dir);
1043 	}
1044 
1045 	if (child_locked)
1046 		ocfs2_inode_unlock(inode, 1);
1047 
1048 	ocfs2_inode_unlock(dir, 1);
1049 
1050 	brelse(fe_bh);
1051 	brelse(parent_node_bh);
1052 
1053 	ocfs2_free_dir_lookup_result(&orphan_insert);
1054 	ocfs2_free_dir_lookup_result(&lookup);
1055 
1056 	if (status && (status != -ENOTEMPTY) && (status != -ENOENT))
1057 		mlog_errno(status);
1058 
1059 	return status;
1060 }
1061 
1062 static int ocfs2_check_if_ancestor(struct ocfs2_super *osb,
1063 		u64 src_inode_no, u64 dest_inode_no)
1064 {
1065 	int ret = 0, i = 0;
1066 	u64 parent_inode_no = 0;
1067 	u64 child_inode_no = src_inode_no;
1068 	struct inode *child_inode;
1069 
1070 #define MAX_LOOKUP_TIMES 32
1071 	while (1) {
1072 		child_inode = ocfs2_iget(osb, child_inode_no, 0, 0);
1073 		if (IS_ERR(child_inode)) {
1074 			ret = PTR_ERR(child_inode);
1075 			break;
1076 		}
1077 
1078 		ret = ocfs2_inode_lock(child_inode, NULL, 0);
1079 		if (ret < 0) {
1080 			iput(child_inode);
1081 			if (ret != -ENOENT)
1082 				mlog_errno(ret);
1083 			break;
1084 		}
1085 
1086 		ret = ocfs2_lookup_ino_from_name(child_inode, "..", 2,
1087 				&parent_inode_no);
1088 		ocfs2_inode_unlock(child_inode, 0);
1089 		iput(child_inode);
1090 		if (ret < 0) {
1091 			ret = -ENOENT;
1092 			break;
1093 		}
1094 
1095 		if (parent_inode_no == dest_inode_no) {
1096 			ret = 1;
1097 			break;
1098 		}
1099 
1100 		if (parent_inode_no == osb->root_inode->i_ino) {
1101 			ret = 0;
1102 			break;
1103 		}
1104 
1105 		child_inode_no = parent_inode_no;
1106 
1107 		if (++i >= MAX_LOOKUP_TIMES) {
1108 			mlog(ML_NOTICE, "max lookup times reached, filesystem "
1109 					"may have nested directories, "
1110 					"src inode: %llu, dest inode: %llu.\n",
1111 					(unsigned long long)src_inode_no,
1112 					(unsigned long long)dest_inode_no);
1113 			ret = 0;
1114 			break;
1115 		}
1116 	}
1117 
1118 	return ret;
1119 }
1120 
1121 /*
1122  * The only place this should be used is rename and link!
1123  * if they have the same id, then the 1st one is the only one locked.
1124  */
1125 static int ocfs2_double_lock(struct ocfs2_super *osb,
1126 			     struct buffer_head **bh1,
1127 			     struct inode *inode1,
1128 			     struct buffer_head **bh2,
1129 			     struct inode *inode2,
1130 			     int rename)
1131 {
1132 	int status;
1133 	int inode1_is_ancestor, inode2_is_ancestor;
1134 	struct ocfs2_inode_info *oi1 = OCFS2_I(inode1);
1135 	struct ocfs2_inode_info *oi2 = OCFS2_I(inode2);
1136 
1137 	trace_ocfs2_double_lock((unsigned long long)oi1->ip_blkno,
1138 				(unsigned long long)oi2->ip_blkno);
1139 
1140 	if (*bh1)
1141 		*bh1 = NULL;
1142 	if (*bh2)
1143 		*bh2 = NULL;
1144 
1145 	/* we always want to lock the one with the lower lockid first.
1146 	 * and if they are nested, we lock ancestor first */
1147 	if (oi1->ip_blkno != oi2->ip_blkno) {
1148 		inode1_is_ancestor = ocfs2_check_if_ancestor(osb, oi2->ip_blkno,
1149 				oi1->ip_blkno);
1150 		if (inode1_is_ancestor < 0) {
1151 			status = inode1_is_ancestor;
1152 			goto bail;
1153 		}
1154 
1155 		inode2_is_ancestor = ocfs2_check_if_ancestor(osb, oi1->ip_blkno,
1156 				oi2->ip_blkno);
1157 		if (inode2_is_ancestor < 0) {
1158 			status = inode2_is_ancestor;
1159 			goto bail;
1160 		}
1161 
1162 		if ((inode1_is_ancestor == 1) ||
1163 				(oi1->ip_blkno < oi2->ip_blkno &&
1164 				inode2_is_ancestor == 0)) {
1165 			/* switch id1 and id2 around */
1166 			swap(bh2, bh1);
1167 			swap(inode2, inode1);
1168 		}
1169 		/* lock id2 */
1170 		status = ocfs2_inode_lock_nested(inode2, bh2, 1,
1171 				rename == 1 ? OI_LS_RENAME1 : OI_LS_PARENT);
1172 		if (status < 0) {
1173 			if (status != -ENOENT)
1174 				mlog_errno(status);
1175 			goto bail;
1176 		}
1177 	}
1178 
1179 	/* lock id1 */
1180 	status = ocfs2_inode_lock_nested(inode1, bh1, 1,
1181 			rename == 1 ?  OI_LS_RENAME2 : OI_LS_PARENT);
1182 	if (status < 0) {
1183 		/*
1184 		 * An error return must mean that no cluster locks
1185 		 * were held on function exit.
1186 		 */
1187 		if (oi1->ip_blkno != oi2->ip_blkno) {
1188 			ocfs2_inode_unlock(inode2, 1);
1189 			brelse(*bh2);
1190 			*bh2 = NULL;
1191 		}
1192 
1193 		if (status != -ENOENT)
1194 			mlog_errno(status);
1195 	}
1196 
1197 	trace_ocfs2_double_lock_end(
1198 			(unsigned long long)OCFS2_I(inode1)->ip_blkno,
1199 			(unsigned long long)OCFS2_I(inode2)->ip_blkno);
1200 
1201 bail:
1202 	if (status)
1203 		mlog_errno(status);
1204 	return status;
1205 }
1206 
1207 static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2)
1208 {
1209 	ocfs2_inode_unlock(inode1, 1);
1210 
1211 	if (inode1 != inode2)
1212 		ocfs2_inode_unlock(inode2, 1);
1213 }
1214 
1215 static int ocfs2_rename(struct inode *old_dir,
1216 			struct dentry *old_dentry,
1217 			struct inode *new_dir,
1218 			struct dentry *new_dentry)
1219 {
1220 	int status = 0, rename_lock = 0, parents_locked = 0, target_exists = 0;
1221 	int old_child_locked = 0, new_child_locked = 0, update_dot_dot = 0;
1222 	struct inode *old_inode = d_inode(old_dentry);
1223 	struct inode *new_inode = d_inode(new_dentry);
1224 	struct inode *orphan_dir = NULL;
1225 	struct ocfs2_dinode *newfe = NULL;
1226 	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
1227 	struct buffer_head *newfe_bh = NULL;
1228 	struct buffer_head *old_inode_bh = NULL;
1229 	struct ocfs2_super *osb = NULL;
1230 	u64 newfe_blkno, old_de_ino;
1231 	handle_t *handle = NULL;
1232 	struct buffer_head *old_dir_bh = NULL;
1233 	struct buffer_head *new_dir_bh = NULL;
1234 	u32 old_dir_nlink = old_dir->i_nlink;
1235 	struct ocfs2_dinode *old_di;
1236 	struct ocfs2_dir_lookup_result old_inode_dot_dot_res = { NULL, };
1237 	struct ocfs2_dir_lookup_result target_lookup_res = { NULL, };
1238 	struct ocfs2_dir_lookup_result old_entry_lookup = { NULL, };
1239 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
1240 	struct ocfs2_dir_lookup_result target_insert = { NULL, };
1241 	bool should_add_orphan = false;
1242 
1243 	/* At some point it might be nice to break this function up a
1244 	 * bit. */
1245 
1246 	trace_ocfs2_rename(old_dir, old_dentry, new_dir, new_dentry,
1247 			   old_dentry->d_name.len, old_dentry->d_name.name,
1248 			   new_dentry->d_name.len, new_dentry->d_name.name);
1249 
1250 	status = dquot_initialize(old_dir);
1251 	if (status) {
1252 		mlog_errno(status);
1253 		goto bail;
1254 	}
1255 	status = dquot_initialize(new_dir);
1256 	if (status) {
1257 		mlog_errno(status);
1258 		goto bail;
1259 	}
1260 
1261 	osb = OCFS2_SB(old_dir->i_sb);
1262 
1263 	if (new_inode) {
1264 		if (!igrab(new_inode))
1265 			BUG();
1266 	}
1267 
1268 	/* Assume a directory hierarchy thusly:
1269 	 * a/b/c
1270 	 * a/d
1271 	 * a,b,c, and d are all directories.
1272 	 *
1273 	 * from cwd of 'a' on both nodes:
1274 	 * node1: mv b/c d
1275 	 * node2: mv d   b/c
1276 	 *
1277 	 * And that's why, just like the VFS, we need a file system
1278 	 * rename lock. */
1279 	if (old_dir != new_dir && S_ISDIR(old_inode->i_mode)) {
1280 		status = ocfs2_rename_lock(osb);
1281 		if (status < 0) {
1282 			mlog_errno(status);
1283 			goto bail;
1284 		}
1285 		rename_lock = 1;
1286 
1287 		/* here we cannot guarantee the inodes haven't just been
1288 		 * changed, so check if they are nested again */
1289 		status = ocfs2_check_if_ancestor(osb, new_dir->i_ino,
1290 				old_inode->i_ino);
1291 		if (status < 0) {
1292 			mlog_errno(status);
1293 			goto bail;
1294 		} else if (status == 1) {
1295 			status = -EPERM;
1296 			trace_ocfs2_rename_not_permitted(
1297 					(unsigned long long)old_inode->i_ino,
1298 					(unsigned long long)new_dir->i_ino);
1299 			goto bail;
1300 		}
1301 	}
1302 
1303 	/* if old and new are the same, this'll just do one lock. */
1304 	status = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
1305 				   &new_dir_bh, new_dir, 1);
1306 	if (status < 0) {
1307 		mlog_errno(status);
1308 		goto bail;
1309 	}
1310 	parents_locked = 1;
1311 
1312 	if (!new_dir->i_nlink) {
1313 		status = -EACCES;
1314 		goto bail;
1315 	}
1316 
1317 	/* make sure both dirs have bhs
1318 	 * get an extra ref on old_dir_bh if old==new */
1319 	if (!new_dir_bh) {
1320 		if (old_dir_bh) {
1321 			new_dir_bh = old_dir_bh;
1322 			get_bh(new_dir_bh);
1323 		} else {
1324 			mlog(ML_ERROR, "no old_dir_bh!\n");
1325 			status = -EIO;
1326 			goto bail;
1327 		}
1328 	}
1329 
1330 	/*
1331 	 * Aside from allowing a meta data update, the locking here
1332 	 * also ensures that the downconvert thread on other nodes
1333 	 * won't have to concurrently downconvert the inode and the
1334 	 * dentry locks.
1335 	 */
1336 	status = ocfs2_inode_lock_nested(old_inode, &old_inode_bh, 1,
1337 					 OI_LS_PARENT);
1338 	if (status < 0) {
1339 		if (status != -ENOENT)
1340 			mlog_errno(status);
1341 		goto bail;
1342 	}
1343 	old_child_locked = 1;
1344 
1345 	status = ocfs2_remote_dentry_delete(old_dentry);
1346 	if (status < 0) {
1347 		mlog_errno(status);
1348 		goto bail;
1349 	}
1350 
1351 	if (S_ISDIR(old_inode->i_mode)) {
1352 		u64 old_inode_parent;
1353 
1354 		update_dot_dot = 1;
1355 		status = ocfs2_find_files_on_disk("..", 2, &old_inode_parent,
1356 						  old_inode,
1357 						  &old_inode_dot_dot_res);
1358 		if (status) {
1359 			status = -EIO;
1360 			goto bail;
1361 		}
1362 
1363 		if (old_inode_parent != OCFS2_I(old_dir)->ip_blkno) {
1364 			status = -EIO;
1365 			goto bail;
1366 		}
1367 
1368 		if (!new_inode && new_dir != old_dir &&
1369 		    new_dir->i_nlink >= ocfs2_link_max(osb)) {
1370 			status = -EMLINK;
1371 			goto bail;
1372 		}
1373 	}
1374 
1375 	status = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name,
1376 					    old_dentry->d_name.len,
1377 					    &old_de_ino);
1378 	if (status) {
1379 		status = -ENOENT;
1380 		goto bail;
1381 	}
1382 
1383 	/*
1384 	 *  Check for inode number is _not_ due to possible IO errors.
1385 	 *  We might rmdir the source, keep it as pwd of some process
1386 	 *  and merrily kill the link to whatever was created under the
1387 	 *  same name. Goodbye sticky bit ;-<
1388 	 */
1389 	if (old_de_ino != OCFS2_I(old_inode)->ip_blkno) {
1390 		status = -ENOENT;
1391 		goto bail;
1392 	}
1393 
1394 	/* check if the target already exists (in which case we need
1395 	 * to delete it */
1396 	status = ocfs2_find_files_on_disk(new_dentry->d_name.name,
1397 					  new_dentry->d_name.len,
1398 					  &newfe_blkno, new_dir,
1399 					  &target_lookup_res);
1400 	/* The only error we allow here is -ENOENT because the new
1401 	 * file not existing is perfectly valid. */
1402 	if ((status < 0) && (status != -ENOENT)) {
1403 		/* If we cannot find the file specified we should just */
1404 		/* return the error... */
1405 		mlog_errno(status);
1406 		goto bail;
1407 	}
1408 	if (status == 0)
1409 		target_exists = 1;
1410 
1411 	if (!target_exists && new_inode) {
1412 		/*
1413 		 * Target was unlinked by another node while we were
1414 		 * waiting to get to ocfs2_rename(). There isn't
1415 		 * anything we can do here to help the situation, so
1416 		 * bubble up the appropriate error.
1417 		 */
1418 		status = -ENOENT;
1419 		goto bail;
1420 	}
1421 
1422 	/* In case we need to overwrite an existing file, we blow it
1423 	 * away first */
1424 	if (target_exists) {
1425 		/* VFS didn't think there existed an inode here, but
1426 		 * someone else in the cluster must have raced our
1427 		 * rename to create one. Today we error cleanly, in
1428 		 * the future we should consider calling iget to build
1429 		 * a new struct inode for this entry. */
1430 		if (!new_inode) {
1431 			status = -EACCES;
1432 
1433 			trace_ocfs2_rename_target_exists(new_dentry->d_name.len,
1434 						new_dentry->d_name.name);
1435 			goto bail;
1436 		}
1437 
1438 		if (OCFS2_I(new_inode)->ip_blkno != newfe_blkno) {
1439 			status = -EACCES;
1440 
1441 			trace_ocfs2_rename_disagree(
1442 			     (unsigned long long)OCFS2_I(new_inode)->ip_blkno,
1443 			     (unsigned long long)newfe_blkno,
1444 			     OCFS2_I(new_inode)->ip_flags);
1445 			goto bail;
1446 		}
1447 
1448 		status = ocfs2_inode_lock(new_inode, &newfe_bh, 1);
1449 		if (status < 0) {
1450 			if (status != -ENOENT)
1451 				mlog_errno(status);
1452 			goto bail;
1453 		}
1454 		new_child_locked = 1;
1455 
1456 		status = ocfs2_remote_dentry_delete(new_dentry);
1457 		if (status < 0) {
1458 			mlog_errno(status);
1459 			goto bail;
1460 		}
1461 
1462 		newfe = (struct ocfs2_dinode *) newfe_bh->b_data;
1463 
1464 		trace_ocfs2_rename_over_existing(
1465 		     (unsigned long long)newfe_blkno, newfe_bh, newfe_bh ?
1466 		     (unsigned long long)newfe_bh->b_blocknr : 0ULL);
1467 
1468 		if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) {
1469 			status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
1470 						OCFS2_I(new_inode)->ip_blkno,
1471 						orphan_name, &orphan_insert,
1472 						false);
1473 			if (status < 0) {
1474 				mlog_errno(status);
1475 				goto bail;
1476 			}
1477 			should_add_orphan = true;
1478 		}
1479 	} else {
1480 		BUG_ON(d_inode(new_dentry->d_parent) != new_dir);
1481 
1482 		status = ocfs2_check_dir_for_entry(new_dir,
1483 						   new_dentry->d_name.name,
1484 						   new_dentry->d_name.len);
1485 		if (status)
1486 			goto bail;
1487 
1488 		status = ocfs2_prepare_dir_for_insert(osb, new_dir, new_dir_bh,
1489 						      new_dentry->d_name.name,
1490 						      new_dentry->d_name.len,
1491 						      &target_insert);
1492 		if (status < 0) {
1493 			mlog_errno(status);
1494 			goto bail;
1495 		}
1496 	}
1497 
1498 	handle = ocfs2_start_trans(osb, ocfs2_rename_credits(osb->sb));
1499 	if (IS_ERR(handle)) {
1500 		status = PTR_ERR(handle);
1501 		handle = NULL;
1502 		mlog_errno(status);
1503 		goto bail;
1504 	}
1505 
1506 	if (target_exists) {
1507 		if (S_ISDIR(new_inode->i_mode)) {
1508 			if (new_inode->i_nlink != 2 ||
1509 			    !ocfs2_empty_dir(new_inode)) {
1510 				status = -ENOTEMPTY;
1511 				goto bail;
1512 			}
1513 		}
1514 		status = ocfs2_journal_access_di(handle, INODE_CACHE(new_inode),
1515 						 newfe_bh,
1516 						 OCFS2_JOURNAL_ACCESS_WRITE);
1517 		if (status < 0) {
1518 			mlog_errno(status);
1519 			goto bail;
1520 		}
1521 
1522 		/* change the dirent to point to the correct inode */
1523 		status = ocfs2_update_entry(new_dir, handle, &target_lookup_res,
1524 					    old_inode);
1525 		if (status < 0) {
1526 			mlog_errno(status);
1527 			goto bail;
1528 		}
1529 		new_dir->i_version++;
1530 
1531 		if (S_ISDIR(new_inode->i_mode))
1532 			ocfs2_set_links_count(newfe, 0);
1533 		else
1534 			ocfs2_add_links_count(newfe, -1);
1535 		ocfs2_journal_dirty(handle, newfe_bh);
1536 		if (should_add_orphan) {
1537 			status = ocfs2_orphan_add(osb, handle, new_inode,
1538 					newfe_bh, orphan_name,
1539 					&orphan_insert, orphan_dir, false);
1540 			if (status < 0) {
1541 				mlog_errno(status);
1542 				goto bail;
1543 			}
1544 		}
1545 	} else {
1546 		/* if the name was not found in new_dir, add it now */
1547 		status = ocfs2_add_entry(handle, new_dentry, old_inode,
1548 					 OCFS2_I(old_inode)->ip_blkno,
1549 					 new_dir_bh, &target_insert);
1550 	}
1551 
1552 	old_inode->i_ctime = CURRENT_TIME;
1553 	mark_inode_dirty(old_inode);
1554 
1555 	status = ocfs2_journal_access_di(handle, INODE_CACHE(old_inode),
1556 					 old_inode_bh,
1557 					 OCFS2_JOURNAL_ACCESS_WRITE);
1558 	if (status >= 0) {
1559 		old_di = (struct ocfs2_dinode *) old_inode_bh->b_data;
1560 
1561 		old_di->i_ctime = cpu_to_le64(old_inode->i_ctime.tv_sec);
1562 		old_di->i_ctime_nsec = cpu_to_le32(old_inode->i_ctime.tv_nsec);
1563 		ocfs2_journal_dirty(handle, old_inode_bh);
1564 	} else
1565 		mlog_errno(status);
1566 
1567 	/*
1568 	 * Now that the name has been added to new_dir, remove the old name.
1569 	 *
1570 	 * We don't keep any directory entry context around until now
1571 	 * because the insert might have changed the type of directory
1572 	 * we're dealing with.
1573 	 */
1574 	status = ocfs2_find_entry(old_dentry->d_name.name,
1575 				  old_dentry->d_name.len, old_dir,
1576 				  &old_entry_lookup);
1577 	if (status) {
1578 		if (!is_journal_aborted(osb->journal->j_journal)) {
1579 			ocfs2_error(osb->sb, "new entry %.*s is added, but old entry %.*s "
1580 					"is not deleted.",
1581 					new_dentry->d_name.len, new_dentry->d_name.name,
1582 					old_dentry->d_name.len, old_dentry->d_name.name);
1583 		}
1584 		goto bail;
1585 	}
1586 
1587 	status = ocfs2_delete_entry(handle, old_dir, &old_entry_lookup);
1588 	if (status < 0) {
1589 		mlog_errno(status);
1590 		if (!is_journal_aborted(osb->journal->j_journal)) {
1591 			ocfs2_error(osb->sb, "new entry %.*s is added, but old entry %.*s "
1592 					"is not deleted.",
1593 					new_dentry->d_name.len, new_dentry->d_name.name,
1594 					old_dentry->d_name.len, old_dentry->d_name.name);
1595 		}
1596 		goto bail;
1597 	}
1598 
1599 	if (new_inode) {
1600 		drop_nlink(new_inode);
1601 		new_inode->i_ctime = CURRENT_TIME;
1602 	}
1603 	old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME;
1604 
1605 	if (update_dot_dot) {
1606 		status = ocfs2_update_entry(old_inode, handle,
1607 					    &old_inode_dot_dot_res, new_dir);
1608 		drop_nlink(old_dir);
1609 		if (new_inode) {
1610 			drop_nlink(new_inode);
1611 		} else {
1612 			inc_nlink(new_dir);
1613 			mark_inode_dirty(new_dir);
1614 		}
1615 	}
1616 	mark_inode_dirty(old_dir);
1617 	ocfs2_mark_inode_dirty(handle, old_dir, old_dir_bh);
1618 	if (new_inode) {
1619 		mark_inode_dirty(new_inode);
1620 		ocfs2_mark_inode_dirty(handle, new_inode, newfe_bh);
1621 	}
1622 
1623 	if (old_dir != new_dir) {
1624 		/* Keep the same times on both directories.*/
1625 		new_dir->i_ctime = new_dir->i_mtime = old_dir->i_ctime;
1626 
1627 		/*
1628 		 * This will also pick up the i_nlink change from the
1629 		 * block above.
1630 		 */
1631 		ocfs2_mark_inode_dirty(handle, new_dir, new_dir_bh);
1632 	}
1633 
1634 	if (old_dir_nlink != old_dir->i_nlink) {
1635 		if (!old_dir_bh) {
1636 			mlog(ML_ERROR, "need to change nlink for old dir "
1637 			     "%llu from %d to %d but bh is NULL!\n",
1638 			     (unsigned long long)OCFS2_I(old_dir)->ip_blkno,
1639 			     (int)old_dir_nlink, old_dir->i_nlink);
1640 		} else {
1641 			struct ocfs2_dinode *fe;
1642 			status = ocfs2_journal_access_di(handle,
1643 							 INODE_CACHE(old_dir),
1644 							 old_dir_bh,
1645 							 OCFS2_JOURNAL_ACCESS_WRITE);
1646 			fe = (struct ocfs2_dinode *) old_dir_bh->b_data;
1647 			ocfs2_set_links_count(fe, old_dir->i_nlink);
1648 			ocfs2_journal_dirty(handle, old_dir_bh);
1649 		}
1650 	}
1651 	ocfs2_dentry_move(old_dentry, new_dentry, old_dir, new_dir);
1652 	status = 0;
1653 bail:
1654 	if (handle)
1655 		ocfs2_commit_trans(osb, handle);
1656 
1657 	if (orphan_dir) {
1658 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
1659 		ocfs2_inode_unlock(orphan_dir, 1);
1660 		mutex_unlock(&orphan_dir->i_mutex);
1661 		iput(orphan_dir);
1662 	}
1663 
1664 	if (new_child_locked)
1665 		ocfs2_inode_unlock(new_inode, 1);
1666 
1667 	if (old_child_locked)
1668 		ocfs2_inode_unlock(old_inode, 1);
1669 
1670 	if (parents_locked)
1671 		ocfs2_double_unlock(old_dir, new_dir);
1672 
1673 	if (rename_lock)
1674 		ocfs2_rename_unlock(osb);
1675 
1676 	if (new_inode)
1677 		sync_mapping_buffers(old_inode->i_mapping);
1678 
1679 	if (new_inode)
1680 		iput(new_inode);
1681 
1682 	ocfs2_free_dir_lookup_result(&target_lookup_res);
1683 	ocfs2_free_dir_lookup_result(&old_entry_lookup);
1684 	ocfs2_free_dir_lookup_result(&old_inode_dot_dot_res);
1685 	ocfs2_free_dir_lookup_result(&orphan_insert);
1686 	ocfs2_free_dir_lookup_result(&target_insert);
1687 
1688 	brelse(newfe_bh);
1689 	brelse(old_inode_bh);
1690 	brelse(old_dir_bh);
1691 	brelse(new_dir_bh);
1692 
1693 	if (status)
1694 		mlog_errno(status);
1695 
1696 	return status;
1697 }
1698 
1699 /*
1700  * we expect i_size = strlen(symname). Copy symname into the file
1701  * data, including the null terminator.
1702  */
1703 static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
1704 				     handle_t *handle,
1705 				     struct inode *inode,
1706 				     const char *symname)
1707 {
1708 	struct buffer_head **bhs = NULL;
1709 	const char *c;
1710 	struct super_block *sb = osb->sb;
1711 	u64 p_blkno, p_blocks;
1712 	int virtual, blocks, status, i, bytes_left;
1713 
1714 	bytes_left = i_size_read(inode) + 1;
1715 	/* we can't trust i_blocks because we're actually going to
1716 	 * write i_size + 1 bytes. */
1717 	blocks = (bytes_left + sb->s_blocksize - 1) >> sb->s_blocksize_bits;
1718 
1719 	trace_ocfs2_create_symlink_data((unsigned long long)inode->i_blocks,
1720 					i_size_read(inode), blocks);
1721 
1722 	/* Sanity check -- make sure we're going to fit. */
1723 	if (bytes_left >
1724 	    ocfs2_clusters_to_bytes(sb, OCFS2_I(inode)->ip_clusters)) {
1725 		status = -EIO;
1726 		mlog_errno(status);
1727 		goto bail;
1728 	}
1729 
1730 	bhs = kcalloc(blocks, sizeof(struct buffer_head *), GFP_KERNEL);
1731 	if (!bhs) {
1732 		status = -ENOMEM;
1733 		mlog_errno(status);
1734 		goto bail;
1735 	}
1736 
1737 	status = ocfs2_extent_map_get_blocks(inode, 0, &p_blkno, &p_blocks,
1738 					     NULL);
1739 	if (status < 0) {
1740 		mlog_errno(status);
1741 		goto bail;
1742 	}
1743 
1744 	/* links can never be larger than one cluster so we know this
1745 	 * is all going to be contiguous, but do a sanity check
1746 	 * anyway. */
1747 	if ((p_blocks << sb->s_blocksize_bits) < bytes_left) {
1748 		status = -EIO;
1749 		mlog_errno(status);
1750 		goto bail;
1751 	}
1752 
1753 	virtual = 0;
1754 	while(bytes_left > 0) {
1755 		c = &symname[virtual * sb->s_blocksize];
1756 
1757 		bhs[virtual] = sb_getblk(sb, p_blkno);
1758 		if (!bhs[virtual]) {
1759 			status = -ENOMEM;
1760 			mlog_errno(status);
1761 			goto bail;
1762 		}
1763 		ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode),
1764 					      bhs[virtual]);
1765 
1766 		status = ocfs2_journal_access(handle, INODE_CACHE(inode),
1767 					      bhs[virtual],
1768 					      OCFS2_JOURNAL_ACCESS_CREATE);
1769 		if (status < 0) {
1770 			mlog_errno(status);
1771 			goto bail;
1772 		}
1773 
1774 		memset(bhs[virtual]->b_data, 0, sb->s_blocksize);
1775 
1776 		memcpy(bhs[virtual]->b_data, c,
1777 		       (bytes_left > sb->s_blocksize) ? sb->s_blocksize :
1778 		       bytes_left);
1779 
1780 		ocfs2_journal_dirty(handle, bhs[virtual]);
1781 
1782 		virtual++;
1783 		p_blkno++;
1784 		bytes_left -= sb->s_blocksize;
1785 	}
1786 
1787 	status = 0;
1788 bail:
1789 
1790 	if (bhs) {
1791 		for(i = 0; i < blocks; i++)
1792 			brelse(bhs[i]);
1793 		kfree(bhs);
1794 	}
1795 
1796 	if (status)
1797 		mlog_errno(status);
1798 	return status;
1799 }
1800 
1801 static int ocfs2_symlink(struct inode *dir,
1802 			 struct dentry *dentry,
1803 			 const char *symname)
1804 {
1805 	int status, l, credits;
1806 	u64 newsize;
1807 	struct ocfs2_super *osb = NULL;
1808 	struct inode *inode = NULL;
1809 	struct super_block *sb;
1810 	struct buffer_head *new_fe_bh = NULL;
1811 	struct buffer_head *parent_fe_bh = NULL;
1812 	struct ocfs2_dinode *fe = NULL;
1813 	struct ocfs2_dinode *dirfe;
1814 	handle_t *handle = NULL;
1815 	struct ocfs2_alloc_context *inode_ac = NULL;
1816 	struct ocfs2_alloc_context *data_ac = NULL;
1817 	struct ocfs2_alloc_context *xattr_ac = NULL;
1818 	int want_clusters = 0;
1819 	int xattr_credits = 0;
1820 	struct ocfs2_security_xattr_info si = {
1821 		.enable = 1,
1822 	};
1823 	int did_quota = 0, did_quota_inode = 0;
1824 	struct ocfs2_dir_lookup_result lookup = { NULL, };
1825 	sigset_t oldset;
1826 	int did_block_signals = 0;
1827 	struct ocfs2_dentry_lock *dl = NULL;
1828 
1829 	trace_ocfs2_symlink_begin(dir, dentry, symname,
1830 				  dentry->d_name.len, dentry->d_name.name);
1831 
1832 	status = dquot_initialize(dir);
1833 	if (status) {
1834 		mlog_errno(status);
1835 		goto bail;
1836 	}
1837 
1838 	sb = dir->i_sb;
1839 	osb = OCFS2_SB(sb);
1840 
1841 	l = strlen(symname) + 1;
1842 
1843 	credits = ocfs2_calc_symlink_credits(sb);
1844 
1845 	/* lock the parent directory */
1846 	status = ocfs2_inode_lock(dir, &parent_fe_bh, 1);
1847 	if (status < 0) {
1848 		if (status != -ENOENT)
1849 			mlog_errno(status);
1850 		return status;
1851 	}
1852 
1853 	dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
1854 	if (!ocfs2_read_links_count(dirfe)) {
1855 		/* can't make a file in a deleted directory. */
1856 		status = -ENOENT;
1857 		goto bail;
1858 	}
1859 
1860 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
1861 					   dentry->d_name.len);
1862 	if (status)
1863 		goto bail;
1864 
1865 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
1866 					      dentry->d_name.name,
1867 					      dentry->d_name.len, &lookup);
1868 	if (status < 0) {
1869 		mlog_errno(status);
1870 		goto bail;
1871 	}
1872 
1873 	status = ocfs2_reserve_new_inode(osb, &inode_ac);
1874 	if (status < 0) {
1875 		if (status != -ENOSPC)
1876 			mlog_errno(status);
1877 		goto bail;
1878 	}
1879 
1880 	inode = ocfs2_get_init_inode(dir, S_IFLNK | S_IRWXUGO);
1881 	if (IS_ERR(inode)) {
1882 		status = PTR_ERR(inode);
1883 		inode = NULL;
1884 		mlog_errno(status);
1885 		goto bail;
1886 	}
1887 
1888 	/* get security xattr */
1889 	status = ocfs2_init_security_get(inode, dir, &dentry->d_name, &si);
1890 	if (status) {
1891 		if (status == -EOPNOTSUPP)
1892 			si.enable = 0;
1893 		else {
1894 			mlog_errno(status);
1895 			goto bail;
1896 		}
1897 	}
1898 
1899 	/* calculate meta data/clusters for setting security xattr */
1900 	if (si.enable) {
1901 		status = ocfs2_calc_security_init(dir, &si, &want_clusters,
1902 						  &xattr_credits, &xattr_ac);
1903 		if (status < 0) {
1904 			mlog_errno(status);
1905 			goto bail;
1906 		}
1907 	}
1908 
1909 	/* don't reserve bitmap space for fast symlinks. */
1910 	if (l > ocfs2_fast_symlink_chars(sb))
1911 		want_clusters += 1;
1912 
1913 	status = ocfs2_reserve_clusters(osb, want_clusters, &data_ac);
1914 	if (status < 0) {
1915 		if (status != -ENOSPC)
1916 			mlog_errno(status);
1917 		goto bail;
1918 	}
1919 
1920 	handle = ocfs2_start_trans(osb, credits + xattr_credits);
1921 	if (IS_ERR(handle)) {
1922 		status = PTR_ERR(handle);
1923 		handle = NULL;
1924 		mlog_errno(status);
1925 		goto bail;
1926 	}
1927 
1928 	/* Starting to change things, restart is no longer possible. */
1929 	ocfs2_block_signals(&oldset);
1930 	did_block_signals = 1;
1931 
1932 	status = dquot_alloc_inode(inode);
1933 	if (status)
1934 		goto bail;
1935 	did_quota_inode = 1;
1936 
1937 	trace_ocfs2_symlink_create(dir, dentry, dentry->d_name.len,
1938 				   dentry->d_name.name,
1939 				   (unsigned long long)OCFS2_I(dir)->ip_blkno,
1940 				   inode->i_mode);
1941 
1942 	status = ocfs2_mknod_locked(osb, dir, inode,
1943 				    0, &new_fe_bh, parent_fe_bh, handle,
1944 				    inode_ac);
1945 	if (status < 0) {
1946 		mlog_errno(status);
1947 		goto bail;
1948 	}
1949 
1950 	fe = (struct ocfs2_dinode *) new_fe_bh->b_data;
1951 	inode->i_rdev = 0;
1952 	newsize = l - 1;
1953 	inode->i_op = &ocfs2_symlink_inode_operations;
1954 	if (l > ocfs2_fast_symlink_chars(sb)) {
1955 		u32 offset = 0;
1956 
1957 		status = dquot_alloc_space_nodirty(inode,
1958 		    ocfs2_clusters_to_bytes(osb->sb, 1));
1959 		if (status)
1960 			goto bail;
1961 		did_quota = 1;
1962 		inode->i_mapping->a_ops = &ocfs2_aops;
1963 		status = ocfs2_add_inode_data(osb, inode, &offset, 1, 0,
1964 					      new_fe_bh,
1965 					      handle, data_ac, NULL,
1966 					      NULL);
1967 		if (status < 0) {
1968 			if (status != -ENOSPC && status != -EINTR) {
1969 				mlog(ML_ERROR,
1970 				     "Failed to extend file to %llu\n",
1971 				     (unsigned long long)newsize);
1972 				mlog_errno(status);
1973 				status = -ENOSPC;
1974 			}
1975 			goto bail;
1976 		}
1977 		i_size_write(inode, newsize);
1978 		inode->i_blocks = ocfs2_inode_sector_count(inode);
1979 	} else {
1980 		inode->i_mapping->a_ops = &ocfs2_fast_symlink_aops;
1981 		memcpy((char *) fe->id2.i_symlink, symname, l);
1982 		i_size_write(inode, newsize);
1983 		inode->i_blocks = 0;
1984 	}
1985 
1986 	status = ocfs2_mark_inode_dirty(handle, inode, new_fe_bh);
1987 	if (status < 0) {
1988 		mlog_errno(status);
1989 		goto bail;
1990 	}
1991 
1992 	if (!ocfs2_inode_is_fast_symlink(inode)) {
1993 		status = ocfs2_create_symlink_data(osb, handle, inode,
1994 						   symname);
1995 		if (status < 0) {
1996 			mlog_errno(status);
1997 			goto bail;
1998 		}
1999 	}
2000 
2001 	if (si.enable) {
2002 		status = ocfs2_init_security_set(handle, inode, new_fe_bh, &si,
2003 						 xattr_ac, data_ac);
2004 		if (status < 0) {
2005 			mlog_errno(status);
2006 			goto bail;
2007 		}
2008 	}
2009 
2010 	/*
2011 	 * Do this before adding the entry to the directory. We add
2012 	 * also set d_op after success so that ->d_iput() will cleanup
2013 	 * the dentry lock even if ocfs2_add_entry() fails below.
2014 	 */
2015 	status = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
2016 	if (status) {
2017 		mlog_errno(status);
2018 		goto bail;
2019 	}
2020 
2021 	dl = dentry->d_fsdata;
2022 
2023 	status = ocfs2_add_entry(handle, dentry, inode,
2024 				 le64_to_cpu(fe->i_blkno), parent_fe_bh,
2025 				 &lookup);
2026 	if (status < 0) {
2027 		mlog_errno(status);
2028 		goto bail;
2029 	}
2030 
2031 	insert_inode_hash(inode);
2032 	d_instantiate(dentry, inode);
2033 bail:
2034 	if (status < 0 && did_quota)
2035 		dquot_free_space_nodirty(inode,
2036 					ocfs2_clusters_to_bytes(osb->sb, 1));
2037 	if (status < 0 && did_quota_inode)
2038 		dquot_free_inode(inode);
2039 	if (handle)
2040 		ocfs2_commit_trans(osb, handle);
2041 
2042 	ocfs2_inode_unlock(dir, 1);
2043 	if (did_block_signals)
2044 		ocfs2_unblock_signals(&oldset);
2045 
2046 	brelse(new_fe_bh);
2047 	brelse(parent_fe_bh);
2048 	kfree(si.value);
2049 	ocfs2_free_dir_lookup_result(&lookup);
2050 	if (inode_ac)
2051 		ocfs2_free_alloc_context(inode_ac);
2052 	if (data_ac)
2053 		ocfs2_free_alloc_context(data_ac);
2054 	if (xattr_ac)
2055 		ocfs2_free_alloc_context(xattr_ac);
2056 	if ((status < 0) && inode) {
2057 		if (dl)
2058 			ocfs2_cleanup_add_entry_failure(osb, dentry, inode);
2059 
2060 		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR;
2061 		clear_nlink(inode);
2062 		iput(inode);
2063 	}
2064 
2065 	if (status)
2066 		mlog_errno(status);
2067 
2068 	return status;
2069 }
2070 
2071 static int ocfs2_blkno_stringify(u64 blkno, char *name)
2072 {
2073 	int status, namelen;
2074 
2075 	namelen = snprintf(name, OCFS2_ORPHAN_NAMELEN + 1, "%016llx",
2076 			   (long long)blkno);
2077 	if (namelen <= 0) {
2078 		if (namelen)
2079 			status = namelen;
2080 		else
2081 			status = -EINVAL;
2082 		mlog_errno(status);
2083 		goto bail;
2084 	}
2085 	if (namelen != OCFS2_ORPHAN_NAMELEN) {
2086 		status = -EINVAL;
2087 		mlog_errno(status);
2088 		goto bail;
2089 	}
2090 
2091 	trace_ocfs2_blkno_stringify(blkno, name, namelen);
2092 
2093 	status = 0;
2094 bail:
2095 	if (status < 0)
2096 		mlog_errno(status);
2097 	return status;
2098 }
2099 
2100 static int ocfs2_lookup_lock_orphan_dir(struct ocfs2_super *osb,
2101 					struct inode **ret_orphan_dir,
2102 					struct buffer_head **ret_orphan_dir_bh)
2103 {
2104 	struct inode *orphan_dir_inode;
2105 	struct buffer_head *orphan_dir_bh = NULL;
2106 	int ret = 0;
2107 
2108 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2109 						       ORPHAN_DIR_SYSTEM_INODE,
2110 						       osb->slot_num);
2111 	if (!orphan_dir_inode) {
2112 		ret = -ENOENT;
2113 		mlog_errno(ret);
2114 		return ret;
2115 	}
2116 
2117 	mutex_lock(&orphan_dir_inode->i_mutex);
2118 
2119 	ret = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2120 	if (ret < 0) {
2121 		mutex_unlock(&orphan_dir_inode->i_mutex);
2122 		iput(orphan_dir_inode);
2123 
2124 		mlog_errno(ret);
2125 		return ret;
2126 	}
2127 
2128 	*ret_orphan_dir = orphan_dir_inode;
2129 	*ret_orphan_dir_bh = orphan_dir_bh;
2130 
2131 	return 0;
2132 }
2133 
2134 static int __ocfs2_prepare_orphan_dir(struct inode *orphan_dir_inode,
2135 				      struct buffer_head *orphan_dir_bh,
2136 				      u64 blkno,
2137 				      char *name,
2138 				      struct ocfs2_dir_lookup_result *lookup,
2139 				      bool dio)
2140 {
2141 	int ret;
2142 	struct ocfs2_super *osb = OCFS2_SB(orphan_dir_inode->i_sb);
2143 	int namelen = dio ?
2144 			(OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN) :
2145 			OCFS2_ORPHAN_NAMELEN;
2146 
2147 	if (dio) {
2148 		ret = snprintf(name, OCFS2_DIO_ORPHAN_PREFIX_LEN + 1, "%s",
2149 				OCFS2_DIO_ORPHAN_PREFIX);
2150 		if (ret != OCFS2_DIO_ORPHAN_PREFIX_LEN) {
2151 			ret = -EINVAL;
2152 			mlog_errno(ret);
2153 			return ret;
2154 		}
2155 
2156 		ret = ocfs2_blkno_stringify(blkno,
2157 				name + OCFS2_DIO_ORPHAN_PREFIX_LEN);
2158 	} else
2159 		ret = ocfs2_blkno_stringify(blkno, name);
2160 	if (ret < 0) {
2161 		mlog_errno(ret);
2162 		return ret;
2163 	}
2164 
2165 	ret = ocfs2_prepare_dir_for_insert(osb, orphan_dir_inode,
2166 					   orphan_dir_bh, name,
2167 					   namelen, lookup);
2168 	if (ret < 0) {
2169 		mlog_errno(ret);
2170 		return ret;
2171 	}
2172 
2173 	return 0;
2174 }
2175 
2176 /**
2177  * ocfs2_prepare_orphan_dir() - Prepare an orphan directory for
2178  * insertion of an orphan.
2179  * @osb: ocfs2 file system
2180  * @ret_orphan_dir: Orphan dir inode - returned locked!
2181  * @blkno: Actual block number of the inode to be inserted into orphan dir.
2182  * @lookup: dir lookup result, to be passed back into functions like
2183  *          ocfs2_orphan_add
2184  *
2185  * Returns zero on success and the ret_orphan_dir, name and lookup
2186  * fields will be populated.
2187  *
2188  * Returns non-zero on failure.
2189  */
2190 static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
2191 				    struct inode **ret_orphan_dir,
2192 				    u64 blkno,
2193 				    char *name,
2194 				    struct ocfs2_dir_lookup_result *lookup,
2195 				    bool dio)
2196 {
2197 	struct inode *orphan_dir_inode = NULL;
2198 	struct buffer_head *orphan_dir_bh = NULL;
2199 	int ret = 0;
2200 
2201 	ret = ocfs2_lookup_lock_orphan_dir(osb, &orphan_dir_inode,
2202 					   &orphan_dir_bh);
2203 	if (ret < 0) {
2204 		mlog_errno(ret);
2205 		return ret;
2206 	}
2207 
2208 	ret = __ocfs2_prepare_orphan_dir(orphan_dir_inode, orphan_dir_bh,
2209 					 blkno, name, lookup, dio);
2210 	if (ret < 0) {
2211 		mlog_errno(ret);
2212 		goto out;
2213 	}
2214 
2215 	*ret_orphan_dir = orphan_dir_inode;
2216 
2217 out:
2218 	brelse(orphan_dir_bh);
2219 
2220 	if (ret) {
2221 		ocfs2_inode_unlock(orphan_dir_inode, 1);
2222 		mutex_unlock(&orphan_dir_inode->i_mutex);
2223 		iput(orphan_dir_inode);
2224 	}
2225 
2226 	if (ret)
2227 		mlog_errno(ret);
2228 	return ret;
2229 }
2230 
2231 static int ocfs2_orphan_add(struct ocfs2_super *osb,
2232 			    handle_t *handle,
2233 			    struct inode *inode,
2234 			    struct buffer_head *fe_bh,
2235 			    char *name,
2236 			    struct ocfs2_dir_lookup_result *lookup,
2237 			    struct inode *orphan_dir_inode,
2238 			    bool dio)
2239 {
2240 	struct buffer_head *orphan_dir_bh = NULL;
2241 	int status = 0;
2242 	struct ocfs2_dinode *orphan_fe;
2243 	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data;
2244 	int namelen = dio ?
2245 			(OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN) :
2246 			OCFS2_ORPHAN_NAMELEN;
2247 
2248 	trace_ocfs2_orphan_add_begin(
2249 				(unsigned long long)OCFS2_I(inode)->ip_blkno);
2250 
2251 	status = ocfs2_read_inode_block(orphan_dir_inode, &orphan_dir_bh);
2252 	if (status < 0) {
2253 		mlog_errno(status);
2254 		goto leave;
2255 	}
2256 
2257 	status = ocfs2_journal_access_di(handle,
2258 					 INODE_CACHE(orphan_dir_inode),
2259 					 orphan_dir_bh,
2260 					 OCFS2_JOURNAL_ACCESS_WRITE);
2261 	if (status < 0) {
2262 		mlog_errno(status);
2263 		goto leave;
2264 	}
2265 
2266 	/*
2267 	 * We're going to journal the change of i_flags and i_orphaned_slot.
2268 	 * It's safe anyway, though some callers may duplicate the journaling.
2269 	 * Journaling within the func just make the logic look more
2270 	 * straightforward.
2271 	 */
2272 	status = ocfs2_journal_access_di(handle,
2273 					 INODE_CACHE(inode),
2274 					 fe_bh,
2275 					 OCFS2_JOURNAL_ACCESS_WRITE);
2276 	if (status < 0) {
2277 		mlog_errno(status);
2278 		goto leave;
2279 	}
2280 
2281 	/* we're a cluster, and nlink can change on disk from
2282 	 * underneath us... */
2283 	orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data;
2284 	if (S_ISDIR(inode->i_mode))
2285 		ocfs2_add_links_count(orphan_fe, 1);
2286 	set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2287 	ocfs2_journal_dirty(handle, orphan_dir_bh);
2288 
2289 	status = __ocfs2_add_entry(handle, orphan_dir_inode, name,
2290 				   namelen, inode,
2291 				   OCFS2_I(inode)->ip_blkno,
2292 				   orphan_dir_bh, lookup);
2293 	if (status < 0) {
2294 		mlog_errno(status);
2295 		goto rollback;
2296 	}
2297 
2298 	if (dio) {
2299 		/* Update flag OCFS2_DIO_ORPHANED_FL and record the orphan
2300 		 * slot.
2301 		 */
2302 		fe->i_flags |= cpu_to_le32(OCFS2_DIO_ORPHANED_FL);
2303 		fe->i_dio_orphaned_slot = cpu_to_le16(osb->slot_num);
2304 	} else {
2305 		fe->i_flags |= cpu_to_le32(OCFS2_ORPHANED_FL);
2306 		OCFS2_I(inode)->ip_flags &= ~OCFS2_INODE_SKIP_ORPHAN_DIR;
2307 
2308 		/* Record which orphan dir our inode now resides
2309 		 * in. delete_inode will use this to determine which orphan
2310 		 * dir to lock. */
2311 		fe->i_orphaned_slot = cpu_to_le16(osb->slot_num);
2312 	}
2313 
2314 	ocfs2_journal_dirty(handle, fe_bh);
2315 
2316 	trace_ocfs2_orphan_add_end((unsigned long long)OCFS2_I(inode)->ip_blkno,
2317 				   osb->slot_num);
2318 
2319 rollback:
2320 	if (status < 0) {
2321 		if (S_ISDIR(inode->i_mode))
2322 			ocfs2_add_links_count(orphan_fe, -1);
2323 		set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2324 	}
2325 
2326 leave:
2327 	brelse(orphan_dir_bh);
2328 
2329 	return status;
2330 }
2331 
2332 /* unlike orphan_add, we expect the orphan dir to already be locked here. */
2333 int ocfs2_orphan_del(struct ocfs2_super *osb,
2334 		     handle_t *handle,
2335 		     struct inode *orphan_dir_inode,
2336 		     struct inode *inode,
2337 		     struct buffer_head *orphan_dir_bh,
2338 		     bool dio)
2339 {
2340 	const int namelen = OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN;
2341 	char name[namelen + 1];
2342 	struct ocfs2_dinode *orphan_fe;
2343 	int status = 0;
2344 	struct ocfs2_dir_lookup_result lookup = { NULL, };
2345 
2346 	if (dio) {
2347 		status = snprintf(name, OCFS2_DIO_ORPHAN_PREFIX_LEN + 1, "%s",
2348 				OCFS2_DIO_ORPHAN_PREFIX);
2349 		if (status != OCFS2_DIO_ORPHAN_PREFIX_LEN) {
2350 			status = -EINVAL;
2351 			mlog_errno(status);
2352 			return status;
2353 		}
2354 
2355 		status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno,
2356 				name + OCFS2_DIO_ORPHAN_PREFIX_LEN);
2357 	} else
2358 		status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name);
2359 	if (status < 0) {
2360 		mlog_errno(status);
2361 		goto leave;
2362 	}
2363 
2364 	trace_ocfs2_orphan_del(
2365 	     (unsigned long long)OCFS2_I(orphan_dir_inode)->ip_blkno,
2366 	     name, strlen(name));
2367 
2368 	/* find it's spot in the orphan directory */
2369 	status = ocfs2_find_entry(name, strlen(name), orphan_dir_inode,
2370 				  &lookup);
2371 	if (status) {
2372 		mlog_errno(status);
2373 		goto leave;
2374 	}
2375 
2376 	/* remove it from the orphan directory */
2377 	status = ocfs2_delete_entry(handle, orphan_dir_inode, &lookup);
2378 	if (status < 0) {
2379 		mlog_errno(status);
2380 		goto leave;
2381 	}
2382 
2383 	status = ocfs2_journal_access_di(handle,
2384 					 INODE_CACHE(orphan_dir_inode),
2385 					 orphan_dir_bh,
2386 					 OCFS2_JOURNAL_ACCESS_WRITE);
2387 	if (status < 0) {
2388 		mlog_errno(status);
2389 		goto leave;
2390 	}
2391 
2392 	/* do the i_nlink dance! :) */
2393 	orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data;
2394 	if (S_ISDIR(inode->i_mode))
2395 		ocfs2_add_links_count(orphan_fe, -1);
2396 	set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2397 	ocfs2_journal_dirty(handle, orphan_dir_bh);
2398 
2399 leave:
2400 	ocfs2_free_dir_lookup_result(&lookup);
2401 
2402 	if (status)
2403 		mlog_errno(status);
2404 	return status;
2405 }
2406 
2407 /**
2408  * ocfs2_prep_new_orphaned_file() - Prepare the orphan dir to receive a newly
2409  * allocated file. This is different from the typical 'add to orphan dir'
2410  * operation in that the inode does not yet exist. This is a problem because
2411  * the orphan dir stringifies the inode block number to come up with it's
2412  * dirent. Obviously if the inode does not yet exist we have a chicken and egg
2413  * problem. This function works around it by calling deeper into the orphan
2414  * and suballoc code than other callers. Use this only by necessity.
2415  * @dir: The directory which this inode will ultimately wind up under - not the
2416  * orphan dir!
2417  * @dir_bh: buffer_head the @dir inode block
2418  * @orphan_name: string of length (CFS2_ORPHAN_NAMELEN + 1). Will be filled
2419  * with the string to be used for orphan dirent. Pass back to the orphan dir
2420  * code.
2421  * @ret_orphan_dir: orphan dir inode returned to be passed back into orphan
2422  * dir code.
2423  * @ret_di_blkno: block number where the new inode will be allocated.
2424  * @orphan_insert: Dir insert context to be passed back into orphan dir code.
2425  * @ret_inode_ac: Inode alloc context to be passed back to the allocator.
2426  *
2427  * Returns zero on success and the ret_orphan_dir, name and lookup
2428  * fields will be populated.
2429  *
2430  * Returns non-zero on failure.
2431  */
2432 static int ocfs2_prep_new_orphaned_file(struct inode *dir,
2433 					struct buffer_head *dir_bh,
2434 					char *orphan_name,
2435 					struct inode **ret_orphan_dir,
2436 					u64 *ret_di_blkno,
2437 					struct ocfs2_dir_lookup_result *orphan_insert,
2438 					struct ocfs2_alloc_context **ret_inode_ac)
2439 {
2440 	int ret;
2441 	u64 di_blkno;
2442 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
2443 	struct inode *orphan_dir = NULL;
2444 	struct buffer_head *orphan_dir_bh = NULL;
2445 	struct ocfs2_alloc_context *inode_ac = NULL;
2446 
2447 	ret = ocfs2_lookup_lock_orphan_dir(osb, &orphan_dir, &orphan_dir_bh);
2448 	if (ret < 0) {
2449 		mlog_errno(ret);
2450 		return ret;
2451 	}
2452 
2453 	/* reserve an inode spot */
2454 	ret = ocfs2_reserve_new_inode(osb, &inode_ac);
2455 	if (ret < 0) {
2456 		if (ret != -ENOSPC)
2457 			mlog_errno(ret);
2458 		goto out;
2459 	}
2460 
2461 	ret = ocfs2_find_new_inode_loc(dir, dir_bh, inode_ac,
2462 				       &di_blkno);
2463 	if (ret) {
2464 		mlog_errno(ret);
2465 		goto out;
2466 	}
2467 
2468 	ret = __ocfs2_prepare_orphan_dir(orphan_dir, orphan_dir_bh,
2469 					 di_blkno, orphan_name, orphan_insert,
2470 					 false);
2471 	if (ret < 0) {
2472 		mlog_errno(ret);
2473 		goto out;
2474 	}
2475 
2476 out:
2477 	if (ret == 0) {
2478 		*ret_orphan_dir = orphan_dir;
2479 		*ret_di_blkno = di_blkno;
2480 		*ret_inode_ac = inode_ac;
2481 		/*
2482 		 * orphan_name and orphan_insert are already up to
2483 		 * date via prepare_orphan_dir
2484 		 */
2485 	} else {
2486 		/* Unroll reserve_new_inode* */
2487 		if (inode_ac)
2488 			ocfs2_free_alloc_context(inode_ac);
2489 
2490 		/* Unroll orphan dir locking */
2491 		mutex_unlock(&orphan_dir->i_mutex);
2492 		ocfs2_inode_unlock(orphan_dir, 1);
2493 		iput(orphan_dir);
2494 	}
2495 
2496 	brelse(orphan_dir_bh);
2497 
2498 	return ret;
2499 }
2500 
2501 int ocfs2_create_inode_in_orphan(struct inode *dir,
2502 				 int mode,
2503 				 struct inode **new_inode)
2504 {
2505 	int status, did_quota_inode = 0;
2506 	struct inode *inode = NULL;
2507 	struct inode *orphan_dir = NULL;
2508 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
2509 	struct ocfs2_dinode *di = NULL;
2510 	handle_t *handle = NULL;
2511 	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
2512 	struct buffer_head *parent_di_bh = NULL;
2513 	struct buffer_head *new_di_bh = NULL;
2514 	struct ocfs2_alloc_context *inode_ac = NULL;
2515 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
2516 	u64 uninitialized_var(di_blkno), suballoc_loc;
2517 	u16 suballoc_bit;
2518 
2519 	status = ocfs2_inode_lock(dir, &parent_di_bh, 1);
2520 	if (status < 0) {
2521 		if (status != -ENOENT)
2522 			mlog_errno(status);
2523 		return status;
2524 	}
2525 
2526 	status = ocfs2_prep_new_orphaned_file(dir, parent_di_bh,
2527 					      orphan_name, &orphan_dir,
2528 					      &di_blkno, &orphan_insert, &inode_ac);
2529 	if (status < 0) {
2530 		if (status != -ENOSPC)
2531 			mlog_errno(status);
2532 		goto leave;
2533 	}
2534 
2535 	inode = ocfs2_get_init_inode(dir, mode);
2536 	if (IS_ERR(inode)) {
2537 		status = PTR_ERR(inode);
2538 		inode = NULL;
2539 		mlog_errno(status);
2540 		goto leave;
2541 	}
2542 
2543 	handle = ocfs2_start_trans(osb, ocfs2_mknod_credits(osb->sb, 0, 0));
2544 	if (IS_ERR(handle)) {
2545 		status = PTR_ERR(handle);
2546 		handle = NULL;
2547 		mlog_errno(status);
2548 		goto leave;
2549 	}
2550 
2551 	status = dquot_alloc_inode(inode);
2552 	if (status)
2553 		goto leave;
2554 	did_quota_inode = 1;
2555 
2556 	status = ocfs2_claim_new_inode_at_loc(handle, dir, inode_ac,
2557 					      &suballoc_loc,
2558 					      &suballoc_bit, di_blkno);
2559 	if (status < 0) {
2560 		mlog_errno(status);
2561 		goto leave;
2562 	}
2563 
2564 	clear_nlink(inode);
2565 	/* do the real work now. */
2566 	status = __ocfs2_mknod_locked(dir, inode,
2567 				      0, &new_di_bh, parent_di_bh, handle,
2568 				      inode_ac, di_blkno, suballoc_loc,
2569 				      suballoc_bit);
2570 	if (status < 0) {
2571 		mlog_errno(status);
2572 		goto leave;
2573 	}
2574 
2575 	di = (struct ocfs2_dinode *)new_di_bh->b_data;
2576 	status = ocfs2_orphan_add(osb, handle, inode, new_di_bh, orphan_name,
2577 				  &orphan_insert, orphan_dir, false);
2578 	if (status < 0) {
2579 		mlog_errno(status);
2580 		goto leave;
2581 	}
2582 
2583 	/* get open lock so that only nodes can't remove it from orphan dir. */
2584 	status = ocfs2_open_lock(inode);
2585 	if (status < 0)
2586 		mlog_errno(status);
2587 
2588 	insert_inode_hash(inode);
2589 leave:
2590 	if (status < 0 && did_quota_inode)
2591 		dquot_free_inode(inode);
2592 	if (handle)
2593 		ocfs2_commit_trans(osb, handle);
2594 
2595 	if (orphan_dir) {
2596 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
2597 		ocfs2_inode_unlock(orphan_dir, 1);
2598 		mutex_unlock(&orphan_dir->i_mutex);
2599 		iput(orphan_dir);
2600 	}
2601 
2602 	if ((status < 0) && inode) {
2603 		clear_nlink(inode);
2604 		iput(inode);
2605 	}
2606 
2607 	if (inode_ac)
2608 		ocfs2_free_alloc_context(inode_ac);
2609 
2610 	brelse(new_di_bh);
2611 
2612 	if (!status)
2613 		*new_inode = inode;
2614 
2615 	ocfs2_free_dir_lookup_result(&orphan_insert);
2616 
2617 	ocfs2_inode_unlock(dir, 1);
2618 	brelse(parent_di_bh);
2619 	return status;
2620 }
2621 
2622 int ocfs2_add_inode_to_orphan(struct ocfs2_super *osb,
2623 	struct inode *inode)
2624 {
2625 	char orphan_name[OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN + 1];
2626 	struct inode *orphan_dir_inode = NULL;
2627 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
2628 	struct buffer_head *di_bh = NULL;
2629 	int status = 0;
2630 	handle_t *handle = NULL;
2631 	struct ocfs2_dinode *di = NULL;
2632 
2633 	status = ocfs2_inode_lock(inode, &di_bh, 1);
2634 	if (status < 0) {
2635 		mlog_errno(status);
2636 		goto bail;
2637 	}
2638 
2639 	di = (struct ocfs2_dinode *) di_bh->b_data;
2640 	/*
2641 	 * Another append dio crashed?
2642 	 * If so, manually recover it first.
2643 	 */
2644 	if (unlikely(di->i_flags & cpu_to_le32(OCFS2_DIO_ORPHANED_FL))) {
2645 		status = ocfs2_truncate_file(inode, di_bh, i_size_read(inode));
2646 		if (status < 0) {
2647 			if (status != -ENOSPC)
2648 				mlog_errno(status);
2649 			goto bail_unlock_inode;
2650 		}
2651 
2652 		status = ocfs2_del_inode_from_orphan(osb, inode, di_bh, 0, 0);
2653 		if (status < 0) {
2654 			mlog_errno(status);
2655 			goto bail_unlock_inode;
2656 		}
2657 	}
2658 
2659 	status = ocfs2_prepare_orphan_dir(osb, &orphan_dir_inode,
2660 			OCFS2_I(inode)->ip_blkno,
2661 			orphan_name,
2662 			&orphan_insert,
2663 			true);
2664 	if (status < 0) {
2665 		mlog_errno(status);
2666 		goto bail_unlock_inode;
2667 	}
2668 
2669 	handle = ocfs2_start_trans(osb,
2670 			OCFS2_INODE_ADD_TO_ORPHAN_CREDITS);
2671 	if (IS_ERR(handle)) {
2672 		status = PTR_ERR(handle);
2673 		goto bail_unlock_orphan;
2674 	}
2675 
2676 	status = ocfs2_orphan_add(osb, handle, inode, di_bh, orphan_name,
2677 			&orphan_insert, orphan_dir_inode, true);
2678 	if (status)
2679 		mlog_errno(status);
2680 
2681 	ocfs2_commit_trans(osb, handle);
2682 
2683 bail_unlock_orphan:
2684 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2685 	mutex_unlock(&orphan_dir_inode->i_mutex);
2686 	iput(orphan_dir_inode);
2687 
2688 	ocfs2_free_dir_lookup_result(&orphan_insert);
2689 
2690 bail_unlock_inode:
2691 	ocfs2_inode_unlock(inode, 1);
2692 	brelse(di_bh);
2693 
2694 bail:
2695 	return status;
2696 }
2697 
2698 int ocfs2_del_inode_from_orphan(struct ocfs2_super *osb,
2699 		struct inode *inode, struct buffer_head *di_bh,
2700 		int update_isize, loff_t end)
2701 {
2702 	struct inode *orphan_dir_inode = NULL;
2703 	struct buffer_head *orphan_dir_bh = NULL;
2704 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
2705 	handle_t *handle = NULL;
2706 	int status = 0;
2707 
2708 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2709 			ORPHAN_DIR_SYSTEM_INODE,
2710 			le16_to_cpu(di->i_dio_orphaned_slot));
2711 	if (!orphan_dir_inode) {
2712 		status = -ENOENT;
2713 		mlog_errno(status);
2714 		goto bail;
2715 	}
2716 
2717 	mutex_lock(&orphan_dir_inode->i_mutex);
2718 	status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2719 	if (status < 0) {
2720 		mutex_unlock(&orphan_dir_inode->i_mutex);
2721 		iput(orphan_dir_inode);
2722 		mlog_errno(status);
2723 		goto bail;
2724 	}
2725 
2726 	handle = ocfs2_start_trans(osb,
2727 			OCFS2_INODE_DEL_FROM_ORPHAN_CREDITS);
2728 	if (IS_ERR(handle)) {
2729 		status = PTR_ERR(handle);
2730 		goto bail_unlock_orphan;
2731 	}
2732 
2733 	BUG_ON(!(di->i_flags & cpu_to_le32(OCFS2_DIO_ORPHANED_FL)));
2734 
2735 	status = ocfs2_orphan_del(osb, handle, orphan_dir_inode,
2736 				inode, orphan_dir_bh, true);
2737 	if (status < 0) {
2738 		mlog_errno(status);
2739 		goto bail_commit;
2740 	}
2741 
2742 	status = ocfs2_journal_access_di(handle,
2743 			INODE_CACHE(inode),
2744 			di_bh,
2745 			OCFS2_JOURNAL_ACCESS_WRITE);
2746 	if (status < 0) {
2747 		mlog_errno(status);
2748 		goto bail_commit;
2749 	}
2750 
2751 	di->i_flags &= ~cpu_to_le32(OCFS2_DIO_ORPHANED_FL);
2752 	di->i_dio_orphaned_slot = 0;
2753 
2754 	if (update_isize) {
2755 		status = ocfs2_set_inode_size(handle, inode, di_bh, end);
2756 		if (status)
2757 			mlog_errno(status);
2758 	} else
2759 		ocfs2_journal_dirty(handle, di_bh);
2760 
2761 bail_commit:
2762 	ocfs2_commit_trans(osb, handle);
2763 
2764 bail_unlock_orphan:
2765 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2766 	mutex_unlock(&orphan_dir_inode->i_mutex);
2767 	brelse(orphan_dir_bh);
2768 	iput(orphan_dir_inode);
2769 
2770 bail:
2771 	return status;
2772 }
2773 
2774 int ocfs2_mv_orphaned_inode_to_new(struct inode *dir,
2775 				   struct inode *inode,
2776 				   struct dentry *dentry)
2777 {
2778 	int status = 0;
2779 	struct buffer_head *parent_di_bh = NULL;
2780 	handle_t *handle = NULL;
2781 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
2782 	struct ocfs2_dinode *dir_di, *di;
2783 	struct inode *orphan_dir_inode = NULL;
2784 	struct buffer_head *orphan_dir_bh = NULL;
2785 	struct buffer_head *di_bh = NULL;
2786 	struct ocfs2_dir_lookup_result lookup = { NULL, };
2787 
2788 	trace_ocfs2_mv_orphaned_inode_to_new(dir, dentry,
2789 				dentry->d_name.len, dentry->d_name.name,
2790 				(unsigned long long)OCFS2_I(dir)->ip_blkno,
2791 				(unsigned long long)OCFS2_I(inode)->ip_blkno);
2792 
2793 	status = ocfs2_inode_lock(dir, &parent_di_bh, 1);
2794 	if (status < 0) {
2795 		if (status != -ENOENT)
2796 			mlog_errno(status);
2797 		return status;
2798 	}
2799 
2800 	dir_di = (struct ocfs2_dinode *) parent_di_bh->b_data;
2801 	if (!dir_di->i_links_count) {
2802 		/* can't make a file in a deleted directory. */
2803 		status = -ENOENT;
2804 		goto leave;
2805 	}
2806 
2807 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
2808 					   dentry->d_name.len);
2809 	if (status)
2810 		goto leave;
2811 
2812 	/* get a spot inside the dir. */
2813 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_di_bh,
2814 					      dentry->d_name.name,
2815 					      dentry->d_name.len, &lookup);
2816 	if (status < 0) {
2817 		mlog_errno(status);
2818 		goto leave;
2819 	}
2820 
2821 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2822 						       ORPHAN_DIR_SYSTEM_INODE,
2823 						       osb->slot_num);
2824 	if (!orphan_dir_inode) {
2825 		status = -ENOENT;
2826 		mlog_errno(status);
2827 		goto leave;
2828 	}
2829 
2830 	mutex_lock(&orphan_dir_inode->i_mutex);
2831 
2832 	status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2833 	if (status < 0) {
2834 		mlog_errno(status);
2835 		mutex_unlock(&orphan_dir_inode->i_mutex);
2836 		iput(orphan_dir_inode);
2837 		goto leave;
2838 	}
2839 
2840 	status = ocfs2_read_inode_block(inode, &di_bh);
2841 	if (status < 0) {
2842 		mlog_errno(status);
2843 		goto orphan_unlock;
2844 	}
2845 
2846 	handle = ocfs2_start_trans(osb, ocfs2_rename_credits(osb->sb));
2847 	if (IS_ERR(handle)) {
2848 		status = PTR_ERR(handle);
2849 		handle = NULL;
2850 		mlog_errno(status);
2851 		goto orphan_unlock;
2852 	}
2853 
2854 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
2855 					 di_bh, OCFS2_JOURNAL_ACCESS_WRITE);
2856 	if (status < 0) {
2857 		mlog_errno(status);
2858 		goto out_commit;
2859 	}
2860 
2861 	status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode,
2862 				  orphan_dir_bh, false);
2863 	if (status < 0) {
2864 		mlog_errno(status);
2865 		goto out_commit;
2866 	}
2867 
2868 	di = (struct ocfs2_dinode *)di_bh->b_data;
2869 	di->i_flags &= ~cpu_to_le32(OCFS2_ORPHANED_FL);
2870 	di->i_orphaned_slot = 0;
2871 	set_nlink(inode, 1);
2872 	ocfs2_set_links_count(di, inode->i_nlink);
2873 	ocfs2_update_inode_fsync_trans(handle, inode, 1);
2874 	ocfs2_journal_dirty(handle, di_bh);
2875 
2876 	status = ocfs2_add_entry(handle, dentry, inode,
2877 				 OCFS2_I(inode)->ip_blkno, parent_di_bh,
2878 				 &lookup);
2879 	if (status < 0) {
2880 		mlog_errno(status);
2881 		goto out_commit;
2882 	}
2883 
2884 	status = ocfs2_dentry_attach_lock(dentry, inode,
2885 					  OCFS2_I(dir)->ip_blkno);
2886 	if (status) {
2887 		mlog_errno(status);
2888 		goto out_commit;
2889 	}
2890 
2891 	d_instantiate(dentry, inode);
2892 	status = 0;
2893 out_commit:
2894 	ocfs2_commit_trans(osb, handle);
2895 orphan_unlock:
2896 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2897 	mutex_unlock(&orphan_dir_inode->i_mutex);
2898 	iput(orphan_dir_inode);
2899 leave:
2900 
2901 	ocfs2_inode_unlock(dir, 1);
2902 
2903 	brelse(di_bh);
2904 	brelse(parent_di_bh);
2905 	brelse(orphan_dir_bh);
2906 
2907 	ocfs2_free_dir_lookup_result(&lookup);
2908 
2909 	if (status)
2910 		mlog_errno(status);
2911 
2912 	return status;
2913 }
2914 
2915 const struct inode_operations ocfs2_dir_iops = {
2916 	.create		= ocfs2_create,
2917 	.lookup		= ocfs2_lookup,
2918 	.link		= ocfs2_link,
2919 	.unlink		= ocfs2_unlink,
2920 	.rmdir		= ocfs2_unlink,
2921 	.symlink	= ocfs2_symlink,
2922 	.mkdir		= ocfs2_mkdir,
2923 	.mknod		= ocfs2_mknod,
2924 	.rename		= ocfs2_rename,
2925 	.setattr	= ocfs2_setattr,
2926 	.getattr	= ocfs2_getattr,
2927 	.permission	= ocfs2_permission,
2928 	.setxattr	= generic_setxattr,
2929 	.getxattr	= generic_getxattr,
2930 	.listxattr	= ocfs2_listxattr,
2931 	.removexattr	= generic_removexattr,
2932 	.fiemap         = ocfs2_fiemap,
2933 	.get_acl	= ocfs2_iop_get_acl,
2934 	.set_acl	= ocfs2_iop_set_acl,
2935 };
2936