xref: /openbmc/linux/fs/ocfs2/namei.c (revision e3b9f1e8)
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 #include <linux/iversion.h>
45 
46 #include <cluster/masklog.h>
47 
48 #include "ocfs2.h"
49 
50 #include "alloc.h"
51 #include "dcache.h"
52 #include "dir.h"
53 #include "dlmglue.h"
54 #include "extent_map.h"
55 #include "file.h"
56 #include "inode.h"
57 #include "journal.h"
58 #include "namei.h"
59 #include "suballoc.h"
60 #include "super.h"
61 #include "symlink.h"
62 #include "sysfile.h"
63 #include "uptodate.h"
64 #include "xattr.h"
65 #include "acl.h"
66 #include "ocfs2_trace.h"
67 
68 #include "buffer_head_io.h"
69 
70 static int ocfs2_mknod_locked(struct ocfs2_super *osb,
71 			      struct inode *dir,
72 			      struct inode *inode,
73 			      dev_t dev,
74 			      struct buffer_head **new_fe_bh,
75 			      struct buffer_head *parent_fe_bh,
76 			      handle_t *handle,
77 			      struct ocfs2_alloc_context *inode_ac);
78 
79 static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
80 				    struct inode **ret_orphan_dir,
81 				    u64 blkno,
82 				    char *name,
83 				    struct ocfs2_dir_lookup_result *lookup,
84 				    bool dio);
85 
86 static int ocfs2_orphan_add(struct ocfs2_super *osb,
87 			    handle_t *handle,
88 			    struct inode *inode,
89 			    struct buffer_head *fe_bh,
90 			    char *name,
91 			    struct ocfs2_dir_lookup_result *lookup,
92 			    struct inode *orphan_dir_inode,
93 			    bool dio);
94 
95 static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
96 				     handle_t *handle,
97 				     struct inode *inode,
98 				     const char *symname);
99 
100 static int ocfs2_double_lock(struct ocfs2_super *osb,
101 			     struct buffer_head **bh1,
102 			     struct inode *inode1,
103 			     struct buffer_head **bh2,
104 			     struct inode *inode2,
105 			     int rename);
106 
107 static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2);
108 /* An orphan dir name is an 8 byte value, printed as a hex string */
109 #define OCFS2_ORPHAN_NAMELEN ((int)(2 * sizeof(u64)))
110 
111 static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry,
112 				   unsigned int flags)
113 {
114 	int status;
115 	u64 blkno;
116 	struct inode *inode = NULL;
117 	struct dentry *ret;
118 	struct ocfs2_inode_info *oi;
119 
120 	trace_ocfs2_lookup(dir, dentry, dentry->d_name.len,
121 			   dentry->d_name.name,
122 			   (unsigned long long)OCFS2_I(dir)->ip_blkno, 0);
123 
124 	if (dentry->d_name.len > OCFS2_MAX_FILENAME_LEN) {
125 		ret = ERR_PTR(-ENAMETOOLONG);
126 		goto bail;
127 	}
128 
129 	status = ocfs2_inode_lock_nested(dir, NULL, 0, OI_LS_PARENT);
130 	if (status < 0) {
131 		if (status != -ENOENT)
132 			mlog_errno(status);
133 		ret = ERR_PTR(status);
134 		goto bail;
135 	}
136 
137 	status = ocfs2_lookup_ino_from_name(dir, dentry->d_name.name,
138 					    dentry->d_name.len, &blkno);
139 	if (status < 0)
140 		goto bail_add;
141 
142 	inode = ocfs2_iget(OCFS2_SB(dir->i_sb), blkno, 0, 0);
143 	if (IS_ERR(inode)) {
144 		ret = ERR_PTR(-EACCES);
145 		goto bail_unlock;
146 	}
147 
148 	oi = OCFS2_I(inode);
149 	/* Clear any orphaned state... If we were able to look up the
150 	 * inode from a directory, it certainly can't be orphaned. We
151 	 * might have the bad state from a node which intended to
152 	 * orphan this inode but crashed before it could commit the
153 	 * unlink. */
154 	spin_lock(&oi->ip_lock);
155 	oi->ip_flags &= ~OCFS2_INODE_MAYBE_ORPHANED;
156 	spin_unlock(&oi->ip_lock);
157 
158 bail_add:
159 	ret = d_splice_alias(inode, dentry);
160 
161 	if (inode) {
162 		/*
163 		 * If d_splice_alias() finds a DCACHE_DISCONNECTED
164 		 * dentry, it will d_move() it on top of ourse. The
165 		 * return value will indicate this however, so in
166 		 * those cases, we switch them around for the locking
167 		 * code.
168 		 *
169 		 * NOTE: This dentry already has ->d_op set from
170 		 * ocfs2_get_parent() and ocfs2_get_dentry()
171 		 */
172 		if (!IS_ERR_OR_NULL(ret))
173 			dentry = ret;
174 
175 		status = ocfs2_dentry_attach_lock(dentry, inode,
176 						  OCFS2_I(dir)->ip_blkno);
177 		if (status) {
178 			mlog_errno(status);
179 			ret = ERR_PTR(status);
180 			goto bail_unlock;
181 		}
182 	} else
183 		ocfs2_dentry_attach_gen(dentry);
184 
185 bail_unlock:
186 	/* Don't drop the cluster lock until *after* the d_add --
187 	 * unlink on another node will message us to remove that
188 	 * dentry under this lock so otherwise we can race this with
189 	 * the downconvert thread and have a stale dentry. */
190 	ocfs2_inode_unlock(dir, 0);
191 
192 bail:
193 
194 	trace_ocfs2_lookup_ret(ret);
195 
196 	return ret;
197 }
198 
199 static struct inode *ocfs2_get_init_inode(struct inode *dir, umode_t mode)
200 {
201 	struct inode *inode;
202 	int status;
203 
204 	inode = new_inode(dir->i_sb);
205 	if (!inode) {
206 		mlog(ML_ERROR, "new_inode failed!\n");
207 		return ERR_PTR(-ENOMEM);
208 	}
209 
210 	/* populate as many fields early on as possible - many of
211 	 * these are used by the support functions here and in
212 	 * callers. */
213 	if (S_ISDIR(mode))
214 		set_nlink(inode, 2);
215 	inode_init_owner(inode, dir, mode);
216 	status = dquot_initialize(inode);
217 	if (status)
218 		return ERR_PTR(status);
219 
220 	return inode;
221 }
222 
223 static void ocfs2_cleanup_add_entry_failure(struct ocfs2_super *osb,
224 		struct dentry *dentry, struct inode *inode)
225 {
226 	struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
227 
228 	ocfs2_simple_drop_lockres(osb, &dl->dl_lockres);
229 	ocfs2_lock_res_free(&dl->dl_lockres);
230 	BUG_ON(dl->dl_count != 1);
231 	spin_lock(&dentry_attach_lock);
232 	dentry->d_fsdata = NULL;
233 	spin_unlock(&dentry_attach_lock);
234 	kfree(dl);
235 	iput(inode);
236 }
237 
238 static int ocfs2_mknod(struct inode *dir,
239 		       struct dentry *dentry,
240 		       umode_t mode,
241 		       dev_t dev)
242 {
243 	int status = 0;
244 	struct buffer_head *parent_fe_bh = NULL;
245 	handle_t *handle = NULL;
246 	struct ocfs2_super *osb;
247 	struct ocfs2_dinode *dirfe;
248 	struct buffer_head *new_fe_bh = NULL;
249 	struct inode *inode = NULL;
250 	struct ocfs2_alloc_context *inode_ac = NULL;
251 	struct ocfs2_alloc_context *data_ac = NULL;
252 	struct ocfs2_alloc_context *meta_ac = NULL;
253 	int want_clusters = 0;
254 	int want_meta = 0;
255 	int xattr_credits = 0;
256 	struct ocfs2_security_xattr_info si = {
257 		.enable = 1,
258 	};
259 	int did_quota_inode = 0;
260 	struct ocfs2_dir_lookup_result lookup = { NULL, };
261 	sigset_t oldset;
262 	int did_block_signals = 0;
263 	struct ocfs2_dentry_lock *dl = NULL;
264 
265 	trace_ocfs2_mknod(dir, dentry, dentry->d_name.len, dentry->d_name.name,
266 			  (unsigned long long)OCFS2_I(dir)->ip_blkno,
267 			  (unsigned long)dev, mode);
268 
269 	status = dquot_initialize(dir);
270 	if (status) {
271 		mlog_errno(status);
272 		return status;
273 	}
274 
275 	/* get our super block */
276 	osb = OCFS2_SB(dir->i_sb);
277 
278 	status = ocfs2_inode_lock(dir, &parent_fe_bh, 1);
279 	if (status < 0) {
280 		if (status != -ENOENT)
281 			mlog_errno(status);
282 		return status;
283 	}
284 
285 	if (S_ISDIR(mode) && (dir->i_nlink >= ocfs2_link_max(osb))) {
286 		status = -EMLINK;
287 		goto leave;
288 	}
289 
290 	dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
291 	if (!ocfs2_read_links_count(dirfe)) {
292 		/* can't make a file in a deleted directory. */
293 		status = -ENOENT;
294 		goto leave;
295 	}
296 
297 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
298 					   dentry->d_name.len);
299 	if (status)
300 		goto leave;
301 
302 	/* get a spot inside the dir. */
303 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
304 					      dentry->d_name.name,
305 					      dentry->d_name.len, &lookup);
306 	if (status < 0) {
307 		mlog_errno(status);
308 		goto leave;
309 	}
310 
311 	/* reserve an inode spot */
312 	status = ocfs2_reserve_new_inode(osb, &inode_ac);
313 	if (status < 0) {
314 		if (status != -ENOSPC)
315 			mlog_errno(status);
316 		goto leave;
317 	}
318 
319 	inode = ocfs2_get_init_inode(dir, mode);
320 	if (IS_ERR(inode)) {
321 		status = PTR_ERR(inode);
322 		inode = NULL;
323 		mlog_errno(status);
324 		goto leave;
325 	}
326 
327 	/* get security xattr */
328 	status = ocfs2_init_security_get(inode, dir, &dentry->d_name, &si);
329 	if (status) {
330 		if (status == -EOPNOTSUPP)
331 			si.enable = 0;
332 		else {
333 			mlog_errno(status);
334 			goto leave;
335 		}
336 	}
337 
338 	/* calculate meta data/clusters for setting security and acl xattr */
339 	status = ocfs2_calc_xattr_init(dir, parent_fe_bh, mode,
340 				       &si, &want_clusters,
341 				       &xattr_credits, &want_meta);
342 	if (status < 0) {
343 		mlog_errno(status);
344 		goto leave;
345 	}
346 
347 	/* Reserve a cluster if creating an extent based directory. */
348 	if (S_ISDIR(mode) && !ocfs2_supports_inline_data(osb)) {
349 		want_clusters += 1;
350 
351 		/* Dir indexing requires extra space as well */
352 		if (ocfs2_supports_indexed_dirs(osb))
353 			want_meta++;
354 	}
355 
356 	status = ocfs2_reserve_new_metadata_blocks(osb, want_meta, &meta_ac);
357 	if (status < 0) {
358 		if (status != -ENOSPC)
359 			mlog_errno(status);
360 		goto leave;
361 	}
362 
363 	status = ocfs2_reserve_clusters(osb, want_clusters, &data_ac);
364 	if (status < 0) {
365 		if (status != -ENOSPC)
366 			mlog_errno(status);
367 		goto leave;
368 	}
369 
370 	handle = ocfs2_start_trans(osb, ocfs2_mknod_credits(osb->sb,
371 							    S_ISDIR(mode),
372 							    xattr_credits));
373 	if (IS_ERR(handle)) {
374 		status = PTR_ERR(handle);
375 		handle = NULL;
376 		mlog_errno(status);
377 		goto leave;
378 	}
379 
380 	/* Starting to change things, restart is no longer possible. */
381 	ocfs2_block_signals(&oldset);
382 	did_block_signals = 1;
383 
384 	status = dquot_alloc_inode(inode);
385 	if (status)
386 		goto leave;
387 	did_quota_inode = 1;
388 
389 	/* do the real work now. */
390 	status = ocfs2_mknod_locked(osb, dir, inode, dev,
391 				    &new_fe_bh, parent_fe_bh, handle,
392 				    inode_ac);
393 	if (status < 0) {
394 		mlog_errno(status);
395 		goto leave;
396 	}
397 
398 	if (S_ISDIR(mode)) {
399 		status = ocfs2_fill_new_dir(osb, handle, dir, inode,
400 					    new_fe_bh, data_ac, meta_ac);
401 		if (status < 0) {
402 			mlog_errno(status);
403 			goto leave;
404 		}
405 
406 		status = ocfs2_journal_access_di(handle, INODE_CACHE(dir),
407 						 parent_fe_bh,
408 						 OCFS2_JOURNAL_ACCESS_WRITE);
409 		if (status < 0) {
410 			mlog_errno(status);
411 			goto leave;
412 		}
413 		ocfs2_add_links_count(dirfe, 1);
414 		ocfs2_journal_dirty(handle, parent_fe_bh);
415 		inc_nlink(dir);
416 	}
417 
418 	status = ocfs2_init_acl(handle, inode, dir, new_fe_bh, parent_fe_bh,
419 			 meta_ac, data_ac);
420 
421 	if (status < 0) {
422 		mlog_errno(status);
423 		goto leave;
424 	}
425 
426 	if (si.enable) {
427 		status = ocfs2_init_security_set(handle, inode, new_fe_bh, &si,
428 						 meta_ac, data_ac);
429 		if (status < 0) {
430 			mlog_errno(status);
431 			goto leave;
432 		}
433 	}
434 
435 	/*
436 	 * Do this before adding the entry to the directory. We add
437 	 * also set d_op after success so that ->d_iput() will cleanup
438 	 * the dentry lock even if ocfs2_add_entry() fails below.
439 	 */
440 	status = ocfs2_dentry_attach_lock(dentry, inode,
441 					  OCFS2_I(dir)->ip_blkno);
442 	if (status) {
443 		mlog_errno(status);
444 		goto leave;
445 	}
446 
447 	dl = dentry->d_fsdata;
448 
449 	status = ocfs2_add_entry(handle, dentry, inode,
450 				 OCFS2_I(inode)->ip_blkno, parent_fe_bh,
451 				 &lookup);
452 	if (status < 0) {
453 		mlog_errno(status);
454 		goto leave;
455 	}
456 
457 	insert_inode_hash(inode);
458 	d_instantiate(dentry, inode);
459 	status = 0;
460 leave:
461 	if (status < 0 && did_quota_inode)
462 		dquot_free_inode(inode);
463 	if (handle)
464 		ocfs2_commit_trans(osb, handle);
465 
466 	ocfs2_inode_unlock(dir, 1);
467 	if (did_block_signals)
468 		ocfs2_unblock_signals(&oldset);
469 
470 	brelse(new_fe_bh);
471 	brelse(parent_fe_bh);
472 	kfree(si.value);
473 
474 	ocfs2_free_dir_lookup_result(&lookup);
475 
476 	if (inode_ac)
477 		ocfs2_free_alloc_context(inode_ac);
478 
479 	if (data_ac)
480 		ocfs2_free_alloc_context(data_ac);
481 
482 	if (meta_ac)
483 		ocfs2_free_alloc_context(meta_ac);
484 
485 	/*
486 	 * We should call iput after the i_mutex of the bitmap been
487 	 * unlocked in ocfs2_free_alloc_context, or the
488 	 * ocfs2_delete_inode will mutex_lock again.
489 	 */
490 	if ((status < 0) && inode) {
491 		if (dl)
492 			ocfs2_cleanup_add_entry_failure(osb, dentry, inode);
493 
494 		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR;
495 		clear_nlink(inode);
496 		iput(inode);
497 	}
498 
499 	if (status)
500 		mlog_errno(status);
501 
502 	return status;
503 }
504 
505 static int __ocfs2_mknod_locked(struct inode *dir,
506 				struct inode *inode,
507 				dev_t dev,
508 				struct buffer_head **new_fe_bh,
509 				struct buffer_head *parent_fe_bh,
510 				handle_t *handle,
511 				struct ocfs2_alloc_context *inode_ac,
512 				u64 fe_blkno, u64 suballoc_loc, u16 suballoc_bit)
513 {
514 	int status = 0;
515 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
516 	struct ocfs2_dinode *fe = NULL;
517 	struct ocfs2_extent_list *fel;
518 	u16 feat;
519 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
520 	struct timespec64 ts;
521 
522 	*new_fe_bh = NULL;
523 
524 	/* populate as many fields early on as possible - many of
525 	 * these are used by the support functions here and in
526 	 * callers. */
527 	inode->i_ino = ino_from_blkno(osb->sb, fe_blkno);
528 	OCFS2_I(inode)->ip_blkno = fe_blkno;
529 	spin_lock(&osb->osb_lock);
530 	inode->i_generation = osb->s_next_generation++;
531 	spin_unlock(&osb->osb_lock);
532 
533 	*new_fe_bh = sb_getblk(osb->sb, fe_blkno);
534 	if (!*new_fe_bh) {
535 		status = -ENOMEM;
536 		mlog_errno(status);
537 		goto leave;
538 	}
539 	ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), *new_fe_bh);
540 
541 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
542 					 *new_fe_bh,
543 					 OCFS2_JOURNAL_ACCESS_CREATE);
544 	if (status < 0) {
545 		mlog_errno(status);
546 		goto leave;
547 	}
548 
549 	fe = (struct ocfs2_dinode *) (*new_fe_bh)->b_data;
550 	memset(fe, 0, osb->sb->s_blocksize);
551 
552 	fe->i_generation = cpu_to_le32(inode->i_generation);
553 	fe->i_fs_generation = cpu_to_le32(osb->fs_generation);
554 	fe->i_blkno = cpu_to_le64(fe_blkno);
555 	fe->i_suballoc_loc = cpu_to_le64(suballoc_loc);
556 	fe->i_suballoc_bit = cpu_to_le16(suballoc_bit);
557 	fe->i_suballoc_slot = cpu_to_le16(inode_ac->ac_alloc_slot);
558 	fe->i_uid = cpu_to_le32(i_uid_read(inode));
559 	fe->i_gid = cpu_to_le32(i_gid_read(inode));
560 	fe->i_mode = cpu_to_le16(inode->i_mode);
561 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
562 		fe->id1.dev1.i_rdev = cpu_to_le64(huge_encode_dev(dev));
563 
564 	ocfs2_set_links_count(fe, inode->i_nlink);
565 
566 	fe->i_last_eb_blk = 0;
567 	strcpy(fe->i_signature, OCFS2_INODE_SIGNATURE);
568 	fe->i_flags |= cpu_to_le32(OCFS2_VALID_FL);
569 	ktime_get_real_ts64(&ts);
570 	fe->i_atime = fe->i_ctime = fe->i_mtime =
571 		cpu_to_le64(ts.tv_sec);
572 	fe->i_mtime_nsec = fe->i_ctime_nsec = fe->i_atime_nsec =
573 		cpu_to_le32(ts.tv_nsec);
574 	fe->i_dtime = 0;
575 
576 	/*
577 	 * If supported, directories start with inline data. If inline
578 	 * isn't supported, but indexing is, we start them as indexed.
579 	 */
580 	feat = le16_to_cpu(fe->i_dyn_features);
581 	if (S_ISDIR(inode->i_mode) && ocfs2_supports_inline_data(osb)) {
582 		fe->i_dyn_features = cpu_to_le16(feat | OCFS2_INLINE_DATA_FL);
583 
584 		fe->id2.i_data.id_count = cpu_to_le16(
585 				ocfs2_max_inline_data_with_xattr(osb->sb, fe));
586 	} else {
587 		fel = &fe->id2.i_list;
588 		fel->l_tree_depth = 0;
589 		fel->l_next_free_rec = 0;
590 		fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb));
591 	}
592 
593 	ocfs2_journal_dirty(handle, *new_fe_bh);
594 
595 	ocfs2_populate_inode(inode, fe, 1);
596 	ocfs2_ci_set_new(osb, INODE_CACHE(inode));
597 	if (!ocfs2_mount_local(osb)) {
598 		status = ocfs2_create_new_inode_locks(inode);
599 		if (status < 0)
600 			mlog_errno(status);
601 	}
602 
603 	oi->i_sync_tid = handle->h_transaction->t_tid;
604 	oi->i_datasync_tid = handle->h_transaction->t_tid;
605 
606 leave:
607 	if (status < 0) {
608 		if (*new_fe_bh) {
609 			brelse(*new_fe_bh);
610 			*new_fe_bh = NULL;
611 		}
612 	}
613 
614 	if (status)
615 		mlog_errno(status);
616 	return status;
617 }
618 
619 static int ocfs2_mknod_locked(struct ocfs2_super *osb,
620 			      struct inode *dir,
621 			      struct inode *inode,
622 			      dev_t dev,
623 			      struct buffer_head **new_fe_bh,
624 			      struct buffer_head *parent_fe_bh,
625 			      handle_t *handle,
626 			      struct ocfs2_alloc_context *inode_ac)
627 {
628 	int status = 0;
629 	u64 suballoc_loc, fe_blkno = 0;
630 	u16 suballoc_bit;
631 
632 	*new_fe_bh = NULL;
633 
634 	status = ocfs2_claim_new_inode(handle, dir, parent_fe_bh,
635 				       inode_ac, &suballoc_loc,
636 				       &suballoc_bit, &fe_blkno);
637 	if (status < 0) {
638 		mlog_errno(status);
639 		return status;
640 	}
641 
642 	status = __ocfs2_mknod_locked(dir, inode, dev, new_fe_bh,
643 				    parent_fe_bh, handle, inode_ac,
644 				    fe_blkno, suballoc_loc, suballoc_bit);
645 	if (status < 0) {
646 		u64 bg_blkno = ocfs2_which_suballoc_group(fe_blkno, suballoc_bit);
647 		int tmp = ocfs2_free_suballoc_bits(handle, inode_ac->ac_inode,
648 				inode_ac->ac_bh, suballoc_bit, bg_blkno, 1);
649 		if (tmp)
650 			mlog_errno(tmp);
651 	}
652 
653 	return status;
654 }
655 
656 static int ocfs2_mkdir(struct inode *dir,
657 		       struct dentry *dentry,
658 		       umode_t mode)
659 {
660 	int ret;
661 
662 	trace_ocfs2_mkdir(dir, dentry, dentry->d_name.len, dentry->d_name.name,
663 			  OCFS2_I(dir)->ip_blkno, mode);
664 	ret = ocfs2_mknod(dir, dentry, mode | S_IFDIR, 0);
665 	if (ret)
666 		mlog_errno(ret);
667 
668 	return ret;
669 }
670 
671 static int ocfs2_create(struct inode *dir,
672 			struct dentry *dentry,
673 			umode_t mode,
674 			bool excl)
675 {
676 	int ret;
677 
678 	trace_ocfs2_create(dir, dentry, dentry->d_name.len, dentry->d_name.name,
679 			   (unsigned long long)OCFS2_I(dir)->ip_blkno, mode);
680 	ret = ocfs2_mknod(dir, dentry, mode | S_IFREG, 0);
681 	if (ret)
682 		mlog_errno(ret);
683 
684 	return ret;
685 }
686 
687 static int ocfs2_link(struct dentry *old_dentry,
688 		      struct inode *dir,
689 		      struct dentry *dentry)
690 {
691 	handle_t *handle;
692 	struct inode *inode = d_inode(old_dentry);
693 	struct inode *old_dir = d_inode(old_dentry->d_parent);
694 	int err;
695 	struct buffer_head *fe_bh = NULL;
696 	struct buffer_head *old_dir_bh = NULL;
697 	struct buffer_head *parent_fe_bh = NULL;
698 	struct ocfs2_dinode *fe = NULL;
699 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
700 	struct ocfs2_dir_lookup_result lookup = { NULL, };
701 	sigset_t oldset;
702 	u64 old_de_ino;
703 
704 	trace_ocfs2_link((unsigned long long)OCFS2_I(inode)->ip_blkno,
705 			 old_dentry->d_name.len, old_dentry->d_name.name,
706 			 dentry->d_name.len, dentry->d_name.name);
707 
708 	if (S_ISDIR(inode->i_mode))
709 		return -EPERM;
710 
711 	err = dquot_initialize(dir);
712 	if (err) {
713 		mlog_errno(err);
714 		return err;
715 	}
716 
717 	err = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
718 			&parent_fe_bh, dir, 0);
719 	if (err < 0) {
720 		if (err != -ENOENT)
721 			mlog_errno(err);
722 		return err;
723 	}
724 
725 	/* make sure both dirs have bhs
726 	 * get an extra ref on old_dir_bh if old==new */
727 	if (!parent_fe_bh) {
728 		if (old_dir_bh) {
729 			parent_fe_bh = old_dir_bh;
730 			get_bh(parent_fe_bh);
731 		} else {
732 			mlog(ML_ERROR, "%s: no old_dir_bh!\n", osb->uuid_str);
733 			err = -EIO;
734 			goto out;
735 		}
736 	}
737 
738 	if (!dir->i_nlink) {
739 		err = -ENOENT;
740 		goto out;
741 	}
742 
743 	err = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name,
744 			old_dentry->d_name.len, &old_de_ino);
745 	if (err) {
746 		err = -ENOENT;
747 		goto out;
748 	}
749 
750 	/*
751 	 * Check whether another node removed the source inode while we
752 	 * were in the vfs.
753 	 */
754 	if (old_de_ino != OCFS2_I(inode)->ip_blkno) {
755 		err = -ENOENT;
756 		goto out;
757 	}
758 
759 	err = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
760 					dentry->d_name.len);
761 	if (err)
762 		goto out;
763 
764 	err = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
765 					   dentry->d_name.name,
766 					   dentry->d_name.len, &lookup);
767 	if (err < 0) {
768 		mlog_errno(err);
769 		goto out;
770 	}
771 
772 	err = ocfs2_inode_lock(inode, &fe_bh, 1);
773 	if (err < 0) {
774 		if (err != -ENOENT)
775 			mlog_errno(err);
776 		goto out;
777 	}
778 
779 	fe = (struct ocfs2_dinode *) fe_bh->b_data;
780 	if (ocfs2_read_links_count(fe) >= ocfs2_link_max(osb)) {
781 		err = -EMLINK;
782 		goto out_unlock_inode;
783 	}
784 
785 	handle = ocfs2_start_trans(osb, ocfs2_link_credits(osb->sb));
786 	if (IS_ERR(handle)) {
787 		err = PTR_ERR(handle);
788 		handle = NULL;
789 		mlog_errno(err);
790 		goto out_unlock_inode;
791 	}
792 
793 	/* Starting to change things, restart is no longer possible. */
794 	ocfs2_block_signals(&oldset);
795 
796 	err = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
797 				      OCFS2_JOURNAL_ACCESS_WRITE);
798 	if (err < 0) {
799 		mlog_errno(err);
800 		goto out_commit;
801 	}
802 
803 	inc_nlink(inode);
804 	inode->i_ctime = current_time(inode);
805 	ocfs2_set_links_count(fe, inode->i_nlink);
806 	fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
807 	fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
808 	ocfs2_journal_dirty(handle, fe_bh);
809 
810 	err = ocfs2_add_entry(handle, dentry, inode,
811 			      OCFS2_I(inode)->ip_blkno,
812 			      parent_fe_bh, &lookup);
813 	if (err) {
814 		ocfs2_add_links_count(fe, -1);
815 		drop_nlink(inode);
816 		mlog_errno(err);
817 		goto out_commit;
818 	}
819 
820 	err = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
821 	if (err) {
822 		mlog_errno(err);
823 		goto out_commit;
824 	}
825 
826 	ihold(inode);
827 	d_instantiate(dentry, inode);
828 
829 out_commit:
830 	ocfs2_commit_trans(osb, handle);
831 	ocfs2_unblock_signals(&oldset);
832 out_unlock_inode:
833 	ocfs2_inode_unlock(inode, 1);
834 
835 out:
836 	ocfs2_double_unlock(old_dir, dir);
837 
838 	brelse(fe_bh);
839 	brelse(parent_fe_bh);
840 	brelse(old_dir_bh);
841 
842 	ocfs2_free_dir_lookup_result(&lookup);
843 
844 	if (err)
845 		mlog_errno(err);
846 
847 	return err;
848 }
849 
850 /*
851  * Takes and drops an exclusive lock on the given dentry. This will
852  * force other nodes to drop it.
853  */
854 static int ocfs2_remote_dentry_delete(struct dentry *dentry)
855 {
856 	int ret;
857 
858 	ret = ocfs2_dentry_lock(dentry, 1);
859 	if (ret)
860 		mlog_errno(ret);
861 	else
862 		ocfs2_dentry_unlock(dentry, 1);
863 
864 	return ret;
865 }
866 
867 static inline int ocfs2_inode_is_unlinkable(struct inode *inode)
868 {
869 	if (S_ISDIR(inode->i_mode)) {
870 		if (inode->i_nlink == 2)
871 			return 1;
872 		return 0;
873 	}
874 
875 	if (inode->i_nlink == 1)
876 		return 1;
877 	return 0;
878 }
879 
880 static int ocfs2_unlink(struct inode *dir,
881 			struct dentry *dentry)
882 {
883 	int status;
884 	int child_locked = 0;
885 	bool is_unlinkable = false;
886 	struct inode *inode = d_inode(dentry);
887 	struct inode *orphan_dir = NULL;
888 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
889 	u64 blkno;
890 	struct ocfs2_dinode *fe = NULL;
891 	struct buffer_head *fe_bh = NULL;
892 	struct buffer_head *parent_node_bh = NULL;
893 	handle_t *handle = NULL;
894 	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
895 	struct ocfs2_dir_lookup_result lookup = { NULL, };
896 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
897 
898 	trace_ocfs2_unlink(dir, dentry, dentry->d_name.len,
899 			   dentry->d_name.name,
900 			   (unsigned long long)OCFS2_I(dir)->ip_blkno,
901 			   (unsigned long long)OCFS2_I(inode)->ip_blkno);
902 
903 	status = dquot_initialize(dir);
904 	if (status) {
905 		mlog_errno(status);
906 		return status;
907 	}
908 
909 	BUG_ON(d_inode(dentry->d_parent) != dir);
910 
911 	if (inode == osb->root_inode)
912 		return -EPERM;
913 
914 	status = ocfs2_inode_lock_nested(dir, &parent_node_bh, 1,
915 					 OI_LS_PARENT);
916 	if (status < 0) {
917 		if (status != -ENOENT)
918 			mlog_errno(status);
919 		return status;
920 	}
921 
922 	status = ocfs2_find_files_on_disk(dentry->d_name.name,
923 					  dentry->d_name.len, &blkno, dir,
924 					  &lookup);
925 	if (status < 0) {
926 		if (status != -ENOENT)
927 			mlog_errno(status);
928 		goto leave;
929 	}
930 
931 	if (OCFS2_I(inode)->ip_blkno != blkno) {
932 		status = -ENOENT;
933 
934 		trace_ocfs2_unlink_noent(
935 				(unsigned long long)OCFS2_I(inode)->ip_blkno,
936 				(unsigned long long)blkno,
937 				OCFS2_I(inode)->ip_flags);
938 		goto leave;
939 	}
940 
941 	status = ocfs2_inode_lock(inode, &fe_bh, 1);
942 	if (status < 0) {
943 		if (status != -ENOENT)
944 			mlog_errno(status);
945 		goto leave;
946 	}
947 	child_locked = 1;
948 
949 	if (S_ISDIR(inode->i_mode)) {
950 		if (inode->i_nlink != 2 || !ocfs2_empty_dir(inode)) {
951 			status = -ENOTEMPTY;
952 			goto leave;
953 		}
954 	}
955 
956 	status = ocfs2_remote_dentry_delete(dentry);
957 	if (status < 0) {
958 		/* This remote delete should succeed under all normal
959 		 * circumstances. */
960 		mlog_errno(status);
961 		goto leave;
962 	}
963 
964 	if (ocfs2_inode_is_unlinkable(inode)) {
965 		status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
966 						  OCFS2_I(inode)->ip_blkno,
967 						  orphan_name, &orphan_insert,
968 						  false);
969 		if (status < 0) {
970 			mlog_errno(status);
971 			goto leave;
972 		}
973 		is_unlinkable = true;
974 	}
975 
976 	handle = ocfs2_start_trans(osb, ocfs2_unlink_credits(osb->sb));
977 	if (IS_ERR(handle)) {
978 		status = PTR_ERR(handle);
979 		handle = NULL;
980 		mlog_errno(status);
981 		goto leave;
982 	}
983 
984 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
985 					 OCFS2_JOURNAL_ACCESS_WRITE);
986 	if (status < 0) {
987 		mlog_errno(status);
988 		goto leave;
989 	}
990 
991 	fe = (struct ocfs2_dinode *) fe_bh->b_data;
992 
993 	/* delete the name from the parent dir */
994 	status = ocfs2_delete_entry(handle, dir, &lookup);
995 	if (status < 0) {
996 		mlog_errno(status);
997 		goto leave;
998 	}
999 
1000 	if (S_ISDIR(inode->i_mode))
1001 		drop_nlink(inode);
1002 	drop_nlink(inode);
1003 	ocfs2_set_links_count(fe, inode->i_nlink);
1004 	ocfs2_journal_dirty(handle, fe_bh);
1005 
1006 	dir->i_ctime = dir->i_mtime = current_time(dir);
1007 	if (S_ISDIR(inode->i_mode))
1008 		drop_nlink(dir);
1009 
1010 	status = ocfs2_mark_inode_dirty(handle, dir, parent_node_bh);
1011 	if (status < 0) {
1012 		mlog_errno(status);
1013 		if (S_ISDIR(inode->i_mode))
1014 			inc_nlink(dir);
1015 		goto leave;
1016 	}
1017 
1018 	if (is_unlinkable) {
1019 		status = ocfs2_orphan_add(osb, handle, inode, fe_bh,
1020 				orphan_name, &orphan_insert, orphan_dir, false);
1021 		if (status < 0)
1022 			mlog_errno(status);
1023 	}
1024 
1025 leave:
1026 	if (handle)
1027 		ocfs2_commit_trans(osb, handle);
1028 
1029 	if (orphan_dir) {
1030 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
1031 		ocfs2_inode_unlock(orphan_dir, 1);
1032 		inode_unlock(orphan_dir);
1033 		iput(orphan_dir);
1034 	}
1035 
1036 	if (child_locked)
1037 		ocfs2_inode_unlock(inode, 1);
1038 
1039 	ocfs2_inode_unlock(dir, 1);
1040 
1041 	brelse(fe_bh);
1042 	brelse(parent_node_bh);
1043 
1044 	ocfs2_free_dir_lookup_result(&orphan_insert);
1045 	ocfs2_free_dir_lookup_result(&lookup);
1046 
1047 	if (status && (status != -ENOTEMPTY) && (status != -ENOENT))
1048 		mlog_errno(status);
1049 
1050 	return status;
1051 }
1052 
1053 static int ocfs2_check_if_ancestor(struct ocfs2_super *osb,
1054 		u64 src_inode_no, u64 dest_inode_no)
1055 {
1056 	int ret = 0, i = 0;
1057 	u64 parent_inode_no = 0;
1058 	u64 child_inode_no = src_inode_no;
1059 	struct inode *child_inode;
1060 
1061 #define MAX_LOOKUP_TIMES 32
1062 	while (1) {
1063 		child_inode = ocfs2_iget(osb, child_inode_no, 0, 0);
1064 		if (IS_ERR(child_inode)) {
1065 			ret = PTR_ERR(child_inode);
1066 			break;
1067 		}
1068 
1069 		ret = ocfs2_inode_lock(child_inode, NULL, 0);
1070 		if (ret < 0) {
1071 			iput(child_inode);
1072 			if (ret != -ENOENT)
1073 				mlog_errno(ret);
1074 			break;
1075 		}
1076 
1077 		ret = ocfs2_lookup_ino_from_name(child_inode, "..", 2,
1078 				&parent_inode_no);
1079 		ocfs2_inode_unlock(child_inode, 0);
1080 		iput(child_inode);
1081 		if (ret < 0) {
1082 			ret = -ENOENT;
1083 			break;
1084 		}
1085 
1086 		if (parent_inode_no == dest_inode_no) {
1087 			ret = 1;
1088 			break;
1089 		}
1090 
1091 		if (parent_inode_no == osb->root_inode->i_ino) {
1092 			ret = 0;
1093 			break;
1094 		}
1095 
1096 		child_inode_no = parent_inode_no;
1097 
1098 		if (++i >= MAX_LOOKUP_TIMES) {
1099 			mlog(ML_NOTICE, "max lookup times reached, filesystem "
1100 					"may have nested directories, "
1101 					"src inode: %llu, dest inode: %llu.\n",
1102 					(unsigned long long)src_inode_no,
1103 					(unsigned long long)dest_inode_no);
1104 			ret = 0;
1105 			break;
1106 		}
1107 	}
1108 
1109 	return ret;
1110 }
1111 
1112 /*
1113  * The only place this should be used is rename and link!
1114  * if they have the same id, then the 1st one is the only one locked.
1115  */
1116 static int ocfs2_double_lock(struct ocfs2_super *osb,
1117 			     struct buffer_head **bh1,
1118 			     struct inode *inode1,
1119 			     struct buffer_head **bh2,
1120 			     struct inode *inode2,
1121 			     int rename)
1122 {
1123 	int status;
1124 	int inode1_is_ancestor, inode2_is_ancestor;
1125 	struct ocfs2_inode_info *oi1 = OCFS2_I(inode1);
1126 	struct ocfs2_inode_info *oi2 = OCFS2_I(inode2);
1127 
1128 	trace_ocfs2_double_lock((unsigned long long)oi1->ip_blkno,
1129 				(unsigned long long)oi2->ip_blkno);
1130 
1131 	if (*bh1)
1132 		*bh1 = NULL;
1133 	if (*bh2)
1134 		*bh2 = NULL;
1135 
1136 	/* we always want to lock the one with the lower lockid first.
1137 	 * and if they are nested, we lock ancestor first */
1138 	if (oi1->ip_blkno != oi2->ip_blkno) {
1139 		inode1_is_ancestor = ocfs2_check_if_ancestor(osb, oi2->ip_blkno,
1140 				oi1->ip_blkno);
1141 		if (inode1_is_ancestor < 0) {
1142 			status = inode1_is_ancestor;
1143 			goto bail;
1144 		}
1145 
1146 		inode2_is_ancestor = ocfs2_check_if_ancestor(osb, oi1->ip_blkno,
1147 				oi2->ip_blkno);
1148 		if (inode2_is_ancestor < 0) {
1149 			status = inode2_is_ancestor;
1150 			goto bail;
1151 		}
1152 
1153 		if ((inode1_is_ancestor == 1) ||
1154 				(oi1->ip_blkno < oi2->ip_blkno &&
1155 				inode2_is_ancestor == 0)) {
1156 			/* switch id1 and id2 around */
1157 			swap(bh2, bh1);
1158 			swap(inode2, inode1);
1159 		}
1160 		/* lock id2 */
1161 		status = ocfs2_inode_lock_nested(inode2, bh2, 1,
1162 				rename == 1 ? OI_LS_RENAME1 : OI_LS_PARENT);
1163 		if (status < 0) {
1164 			if (status != -ENOENT)
1165 				mlog_errno(status);
1166 			goto bail;
1167 		}
1168 	}
1169 
1170 	/* lock id1 */
1171 	status = ocfs2_inode_lock_nested(inode1, bh1, 1,
1172 			rename == 1 ?  OI_LS_RENAME2 : OI_LS_PARENT);
1173 	if (status < 0) {
1174 		/*
1175 		 * An error return must mean that no cluster locks
1176 		 * were held on function exit.
1177 		 */
1178 		if (oi1->ip_blkno != oi2->ip_blkno) {
1179 			ocfs2_inode_unlock(inode2, 1);
1180 			brelse(*bh2);
1181 			*bh2 = NULL;
1182 		}
1183 
1184 		if (status != -ENOENT)
1185 			mlog_errno(status);
1186 	}
1187 
1188 	trace_ocfs2_double_lock_end(
1189 			(unsigned long long)OCFS2_I(inode1)->ip_blkno,
1190 			(unsigned long long)OCFS2_I(inode2)->ip_blkno);
1191 
1192 bail:
1193 	if (status)
1194 		mlog_errno(status);
1195 	return status;
1196 }
1197 
1198 static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2)
1199 {
1200 	ocfs2_inode_unlock(inode1, 1);
1201 
1202 	if (inode1 != inode2)
1203 		ocfs2_inode_unlock(inode2, 1);
1204 }
1205 
1206 static int ocfs2_rename(struct inode *old_dir,
1207 			struct dentry *old_dentry,
1208 			struct inode *new_dir,
1209 			struct dentry *new_dentry,
1210 			unsigned int flags)
1211 {
1212 	int status = 0, rename_lock = 0, parents_locked = 0, target_exists = 0;
1213 	int old_child_locked = 0, new_child_locked = 0, update_dot_dot = 0;
1214 	struct inode *old_inode = d_inode(old_dentry);
1215 	struct inode *new_inode = d_inode(new_dentry);
1216 	struct inode *orphan_dir = NULL;
1217 	struct ocfs2_dinode *newfe = NULL;
1218 	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
1219 	struct buffer_head *newfe_bh = NULL;
1220 	struct buffer_head *old_inode_bh = NULL;
1221 	struct ocfs2_super *osb = NULL;
1222 	u64 newfe_blkno, old_de_ino;
1223 	handle_t *handle = NULL;
1224 	struct buffer_head *old_dir_bh = NULL;
1225 	struct buffer_head *new_dir_bh = NULL;
1226 	u32 old_dir_nlink = old_dir->i_nlink;
1227 	struct ocfs2_dinode *old_di;
1228 	struct ocfs2_dir_lookup_result old_inode_dot_dot_res = { NULL, };
1229 	struct ocfs2_dir_lookup_result target_lookup_res = { NULL, };
1230 	struct ocfs2_dir_lookup_result old_entry_lookup = { NULL, };
1231 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
1232 	struct ocfs2_dir_lookup_result target_insert = { NULL, };
1233 	bool should_add_orphan = false;
1234 
1235 	if (flags)
1236 		return -EINVAL;
1237 
1238 	/* At some point it might be nice to break this function up a
1239 	 * bit. */
1240 
1241 	trace_ocfs2_rename(old_dir, old_dentry, new_dir, new_dentry,
1242 			   old_dentry->d_name.len, old_dentry->d_name.name,
1243 			   new_dentry->d_name.len, new_dentry->d_name.name);
1244 
1245 	status = dquot_initialize(old_dir);
1246 	if (status) {
1247 		mlog_errno(status);
1248 		goto bail;
1249 	}
1250 	status = dquot_initialize(new_dir);
1251 	if (status) {
1252 		mlog_errno(status);
1253 		goto bail;
1254 	}
1255 
1256 	osb = OCFS2_SB(old_dir->i_sb);
1257 
1258 	if (new_inode) {
1259 		if (!igrab(new_inode))
1260 			BUG();
1261 	}
1262 
1263 	/* Assume a directory hierarchy thusly:
1264 	 * a/b/c
1265 	 * a/d
1266 	 * a,b,c, and d are all directories.
1267 	 *
1268 	 * from cwd of 'a' on both nodes:
1269 	 * node1: mv b/c d
1270 	 * node2: mv d   b/c
1271 	 *
1272 	 * And that's why, just like the VFS, we need a file system
1273 	 * rename lock. */
1274 	if (old_dir != new_dir && S_ISDIR(old_inode->i_mode)) {
1275 		status = ocfs2_rename_lock(osb);
1276 		if (status < 0) {
1277 			mlog_errno(status);
1278 			goto bail;
1279 		}
1280 		rename_lock = 1;
1281 
1282 		/* here we cannot guarantee the inodes haven't just been
1283 		 * changed, so check if they are nested again */
1284 		status = ocfs2_check_if_ancestor(osb, new_dir->i_ino,
1285 				old_inode->i_ino);
1286 		if (status < 0) {
1287 			mlog_errno(status);
1288 			goto bail;
1289 		} else if (status == 1) {
1290 			status = -EPERM;
1291 			trace_ocfs2_rename_not_permitted(
1292 					(unsigned long long)old_inode->i_ino,
1293 					(unsigned long long)new_dir->i_ino);
1294 			goto bail;
1295 		}
1296 	}
1297 
1298 	/* if old and new are the same, this'll just do one lock. */
1299 	status = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
1300 				   &new_dir_bh, new_dir, 1);
1301 	if (status < 0) {
1302 		mlog_errno(status);
1303 		goto bail;
1304 	}
1305 	parents_locked = 1;
1306 
1307 	if (!new_dir->i_nlink) {
1308 		status = -EACCES;
1309 		goto bail;
1310 	}
1311 
1312 	/* make sure both dirs have bhs
1313 	 * get an extra ref on old_dir_bh if old==new */
1314 	if (!new_dir_bh) {
1315 		if (old_dir_bh) {
1316 			new_dir_bh = old_dir_bh;
1317 			get_bh(new_dir_bh);
1318 		} else {
1319 			mlog(ML_ERROR, "no old_dir_bh!\n");
1320 			status = -EIO;
1321 			goto bail;
1322 		}
1323 	}
1324 
1325 	/*
1326 	 * Aside from allowing a meta data update, the locking here
1327 	 * also ensures that the downconvert thread on other nodes
1328 	 * won't have to concurrently downconvert the inode and the
1329 	 * dentry locks.
1330 	 */
1331 	status = ocfs2_inode_lock_nested(old_inode, &old_inode_bh, 1,
1332 					 OI_LS_PARENT);
1333 	if (status < 0) {
1334 		if (status != -ENOENT)
1335 			mlog_errno(status);
1336 		goto bail;
1337 	}
1338 	old_child_locked = 1;
1339 
1340 	status = ocfs2_remote_dentry_delete(old_dentry);
1341 	if (status < 0) {
1342 		mlog_errno(status);
1343 		goto bail;
1344 	}
1345 
1346 	if (S_ISDIR(old_inode->i_mode)) {
1347 		u64 old_inode_parent;
1348 
1349 		update_dot_dot = 1;
1350 		status = ocfs2_find_files_on_disk("..", 2, &old_inode_parent,
1351 						  old_inode,
1352 						  &old_inode_dot_dot_res);
1353 		if (status) {
1354 			status = -EIO;
1355 			goto bail;
1356 		}
1357 
1358 		if (old_inode_parent != OCFS2_I(old_dir)->ip_blkno) {
1359 			status = -EIO;
1360 			goto bail;
1361 		}
1362 
1363 		if (!new_inode && new_dir != old_dir &&
1364 		    new_dir->i_nlink >= ocfs2_link_max(osb)) {
1365 			status = -EMLINK;
1366 			goto bail;
1367 		}
1368 	}
1369 
1370 	status = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name,
1371 					    old_dentry->d_name.len,
1372 					    &old_de_ino);
1373 	if (status) {
1374 		status = -ENOENT;
1375 		goto bail;
1376 	}
1377 
1378 	/*
1379 	 *  Check for inode number is _not_ due to possible IO errors.
1380 	 *  We might rmdir the source, keep it as pwd of some process
1381 	 *  and merrily kill the link to whatever was created under the
1382 	 *  same name. Goodbye sticky bit ;-<
1383 	 */
1384 	if (old_de_ino != OCFS2_I(old_inode)->ip_blkno) {
1385 		status = -ENOENT;
1386 		goto bail;
1387 	}
1388 
1389 	/* check if the target already exists (in which case we need
1390 	 * to delete it */
1391 	status = ocfs2_find_files_on_disk(new_dentry->d_name.name,
1392 					  new_dentry->d_name.len,
1393 					  &newfe_blkno, new_dir,
1394 					  &target_lookup_res);
1395 	/* The only error we allow here is -ENOENT because the new
1396 	 * file not existing is perfectly valid. */
1397 	if ((status < 0) && (status != -ENOENT)) {
1398 		/* If we cannot find the file specified we should just */
1399 		/* return the error... */
1400 		mlog_errno(status);
1401 		goto bail;
1402 	}
1403 	if (status == 0)
1404 		target_exists = 1;
1405 
1406 	if (!target_exists && new_inode) {
1407 		/*
1408 		 * Target was unlinked by another node while we were
1409 		 * waiting to get to ocfs2_rename(). There isn't
1410 		 * anything we can do here to help the situation, so
1411 		 * bubble up the appropriate error.
1412 		 */
1413 		status = -ENOENT;
1414 		goto bail;
1415 	}
1416 
1417 	/* In case we need to overwrite an existing file, we blow it
1418 	 * away first */
1419 	if (target_exists) {
1420 		/* VFS didn't think there existed an inode here, but
1421 		 * someone else in the cluster must have raced our
1422 		 * rename to create one. Today we error cleanly, in
1423 		 * the future we should consider calling iget to build
1424 		 * a new struct inode for this entry. */
1425 		if (!new_inode) {
1426 			status = -EACCES;
1427 
1428 			trace_ocfs2_rename_target_exists(new_dentry->d_name.len,
1429 						new_dentry->d_name.name);
1430 			goto bail;
1431 		}
1432 
1433 		if (OCFS2_I(new_inode)->ip_blkno != newfe_blkno) {
1434 			status = -EACCES;
1435 
1436 			trace_ocfs2_rename_disagree(
1437 			     (unsigned long long)OCFS2_I(new_inode)->ip_blkno,
1438 			     (unsigned long long)newfe_blkno,
1439 			     OCFS2_I(new_inode)->ip_flags);
1440 			goto bail;
1441 		}
1442 
1443 		status = ocfs2_inode_lock(new_inode, &newfe_bh, 1);
1444 		if (status < 0) {
1445 			if (status != -ENOENT)
1446 				mlog_errno(status);
1447 			goto bail;
1448 		}
1449 		new_child_locked = 1;
1450 
1451 		status = ocfs2_remote_dentry_delete(new_dentry);
1452 		if (status < 0) {
1453 			mlog_errno(status);
1454 			goto bail;
1455 		}
1456 
1457 		newfe = (struct ocfs2_dinode *) newfe_bh->b_data;
1458 
1459 		trace_ocfs2_rename_over_existing(
1460 		     (unsigned long long)newfe_blkno, newfe_bh, newfe_bh ?
1461 		     (unsigned long long)newfe_bh->b_blocknr : 0ULL);
1462 
1463 		if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) {
1464 			status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
1465 						OCFS2_I(new_inode)->ip_blkno,
1466 						orphan_name, &orphan_insert,
1467 						false);
1468 			if (status < 0) {
1469 				mlog_errno(status);
1470 				goto bail;
1471 			}
1472 			should_add_orphan = true;
1473 		}
1474 	} else {
1475 		BUG_ON(d_inode(new_dentry->d_parent) != new_dir);
1476 
1477 		status = ocfs2_check_dir_for_entry(new_dir,
1478 						   new_dentry->d_name.name,
1479 						   new_dentry->d_name.len);
1480 		if (status)
1481 			goto bail;
1482 
1483 		status = ocfs2_prepare_dir_for_insert(osb, new_dir, new_dir_bh,
1484 						      new_dentry->d_name.name,
1485 						      new_dentry->d_name.len,
1486 						      &target_insert);
1487 		if (status < 0) {
1488 			mlog_errno(status);
1489 			goto bail;
1490 		}
1491 	}
1492 
1493 	handle = ocfs2_start_trans(osb, ocfs2_rename_credits(osb->sb));
1494 	if (IS_ERR(handle)) {
1495 		status = PTR_ERR(handle);
1496 		handle = NULL;
1497 		mlog_errno(status);
1498 		goto bail;
1499 	}
1500 
1501 	if (target_exists) {
1502 		if (S_ISDIR(new_inode->i_mode)) {
1503 			if (new_inode->i_nlink != 2 ||
1504 			    !ocfs2_empty_dir(new_inode)) {
1505 				status = -ENOTEMPTY;
1506 				goto bail;
1507 			}
1508 		}
1509 		status = ocfs2_journal_access_di(handle, INODE_CACHE(new_inode),
1510 						 newfe_bh,
1511 						 OCFS2_JOURNAL_ACCESS_WRITE);
1512 		if (status < 0) {
1513 			mlog_errno(status);
1514 			goto bail;
1515 		}
1516 
1517 		/* change the dirent to point to the correct inode */
1518 		status = ocfs2_update_entry(new_dir, handle, &target_lookup_res,
1519 					    old_inode);
1520 		if (status < 0) {
1521 			mlog_errno(status);
1522 			goto bail;
1523 		}
1524 		inode_inc_iversion(new_dir);
1525 
1526 		if (S_ISDIR(new_inode->i_mode))
1527 			ocfs2_set_links_count(newfe, 0);
1528 		else
1529 			ocfs2_add_links_count(newfe, -1);
1530 		ocfs2_journal_dirty(handle, newfe_bh);
1531 		if (should_add_orphan) {
1532 			status = ocfs2_orphan_add(osb, handle, new_inode,
1533 					newfe_bh, orphan_name,
1534 					&orphan_insert, orphan_dir, false);
1535 			if (status < 0) {
1536 				mlog_errno(status);
1537 				goto bail;
1538 			}
1539 		}
1540 	} else {
1541 		/* if the name was not found in new_dir, add it now */
1542 		status = ocfs2_add_entry(handle, new_dentry, old_inode,
1543 					 OCFS2_I(old_inode)->ip_blkno,
1544 					 new_dir_bh, &target_insert);
1545 	}
1546 
1547 	old_inode->i_ctime = current_time(old_inode);
1548 	mark_inode_dirty(old_inode);
1549 
1550 	status = ocfs2_journal_access_di(handle, INODE_CACHE(old_inode),
1551 					 old_inode_bh,
1552 					 OCFS2_JOURNAL_ACCESS_WRITE);
1553 	if (status >= 0) {
1554 		old_di = (struct ocfs2_dinode *) old_inode_bh->b_data;
1555 
1556 		old_di->i_ctime = cpu_to_le64(old_inode->i_ctime.tv_sec);
1557 		old_di->i_ctime_nsec = cpu_to_le32(old_inode->i_ctime.tv_nsec);
1558 		ocfs2_journal_dirty(handle, old_inode_bh);
1559 	} else
1560 		mlog_errno(status);
1561 
1562 	/*
1563 	 * Now that the name has been added to new_dir, remove the old name.
1564 	 *
1565 	 * We don't keep any directory entry context around until now
1566 	 * because the insert might have changed the type of directory
1567 	 * we're dealing with.
1568 	 */
1569 	status = ocfs2_find_entry(old_dentry->d_name.name,
1570 				  old_dentry->d_name.len, old_dir,
1571 				  &old_entry_lookup);
1572 	if (status) {
1573 		if (!is_journal_aborted(osb->journal->j_journal)) {
1574 			ocfs2_error(osb->sb, "new entry %.*s is added, but old entry %.*s "
1575 					"is not deleted.",
1576 					new_dentry->d_name.len, new_dentry->d_name.name,
1577 					old_dentry->d_name.len, old_dentry->d_name.name);
1578 		}
1579 		goto bail;
1580 	}
1581 
1582 	status = ocfs2_delete_entry(handle, old_dir, &old_entry_lookup);
1583 	if (status < 0) {
1584 		mlog_errno(status);
1585 		if (!is_journal_aborted(osb->journal->j_journal)) {
1586 			ocfs2_error(osb->sb, "new entry %.*s is added, but old entry %.*s "
1587 					"is not deleted.",
1588 					new_dentry->d_name.len, new_dentry->d_name.name,
1589 					old_dentry->d_name.len, old_dentry->d_name.name);
1590 		}
1591 		goto bail;
1592 	}
1593 
1594 	if (new_inode) {
1595 		drop_nlink(new_inode);
1596 		new_inode->i_ctime = current_time(new_inode);
1597 	}
1598 	old_dir->i_ctime = old_dir->i_mtime = current_time(old_dir);
1599 
1600 	if (update_dot_dot) {
1601 		status = ocfs2_update_entry(old_inode, handle,
1602 					    &old_inode_dot_dot_res, new_dir);
1603 		drop_nlink(old_dir);
1604 		if (new_inode) {
1605 			drop_nlink(new_inode);
1606 		} else {
1607 			inc_nlink(new_dir);
1608 			mark_inode_dirty(new_dir);
1609 		}
1610 	}
1611 	mark_inode_dirty(old_dir);
1612 	ocfs2_mark_inode_dirty(handle, old_dir, old_dir_bh);
1613 	if (new_inode) {
1614 		mark_inode_dirty(new_inode);
1615 		ocfs2_mark_inode_dirty(handle, new_inode, newfe_bh);
1616 	}
1617 
1618 	if (old_dir != new_dir) {
1619 		/* Keep the same times on both directories.*/
1620 		new_dir->i_ctime = new_dir->i_mtime = old_dir->i_ctime;
1621 
1622 		/*
1623 		 * This will also pick up the i_nlink change from the
1624 		 * block above.
1625 		 */
1626 		ocfs2_mark_inode_dirty(handle, new_dir, new_dir_bh);
1627 	}
1628 
1629 	if (old_dir_nlink != old_dir->i_nlink) {
1630 		if (!old_dir_bh) {
1631 			mlog(ML_ERROR, "need to change nlink for old dir "
1632 			     "%llu from %d to %d but bh is NULL!\n",
1633 			     (unsigned long long)OCFS2_I(old_dir)->ip_blkno,
1634 			     (int)old_dir_nlink, old_dir->i_nlink);
1635 		} else {
1636 			struct ocfs2_dinode *fe;
1637 			status = ocfs2_journal_access_di(handle,
1638 							 INODE_CACHE(old_dir),
1639 							 old_dir_bh,
1640 							 OCFS2_JOURNAL_ACCESS_WRITE);
1641 			fe = (struct ocfs2_dinode *) old_dir_bh->b_data;
1642 			ocfs2_set_links_count(fe, old_dir->i_nlink);
1643 			ocfs2_journal_dirty(handle, old_dir_bh);
1644 		}
1645 	}
1646 	ocfs2_dentry_move(old_dentry, new_dentry, old_dir, new_dir);
1647 	status = 0;
1648 bail:
1649 	if (handle)
1650 		ocfs2_commit_trans(osb, handle);
1651 
1652 	if (orphan_dir) {
1653 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
1654 		ocfs2_inode_unlock(orphan_dir, 1);
1655 		inode_unlock(orphan_dir);
1656 		iput(orphan_dir);
1657 	}
1658 
1659 	if (new_child_locked)
1660 		ocfs2_inode_unlock(new_inode, 1);
1661 
1662 	if (old_child_locked)
1663 		ocfs2_inode_unlock(old_inode, 1);
1664 
1665 	if (parents_locked)
1666 		ocfs2_double_unlock(old_dir, new_dir);
1667 
1668 	if (rename_lock)
1669 		ocfs2_rename_unlock(osb);
1670 
1671 	if (new_inode)
1672 		sync_mapping_buffers(old_inode->i_mapping);
1673 
1674 	iput(new_inode);
1675 
1676 	ocfs2_free_dir_lookup_result(&target_lookup_res);
1677 	ocfs2_free_dir_lookup_result(&old_entry_lookup);
1678 	ocfs2_free_dir_lookup_result(&old_inode_dot_dot_res);
1679 	ocfs2_free_dir_lookup_result(&orphan_insert);
1680 	ocfs2_free_dir_lookup_result(&target_insert);
1681 
1682 	brelse(newfe_bh);
1683 	brelse(old_inode_bh);
1684 	brelse(old_dir_bh);
1685 	brelse(new_dir_bh);
1686 
1687 	if (status)
1688 		mlog_errno(status);
1689 
1690 	return status;
1691 }
1692 
1693 /*
1694  * we expect i_size = strlen(symname). Copy symname into the file
1695  * data, including the null terminator.
1696  */
1697 static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
1698 				     handle_t *handle,
1699 				     struct inode *inode,
1700 				     const char *symname)
1701 {
1702 	struct buffer_head **bhs = NULL;
1703 	const char *c;
1704 	struct super_block *sb = osb->sb;
1705 	u64 p_blkno, p_blocks;
1706 	int virtual, blocks, status, i, bytes_left;
1707 
1708 	bytes_left = i_size_read(inode) + 1;
1709 	/* we can't trust i_blocks because we're actually going to
1710 	 * write i_size + 1 bytes. */
1711 	blocks = (bytes_left + sb->s_blocksize - 1) >> sb->s_blocksize_bits;
1712 
1713 	trace_ocfs2_create_symlink_data((unsigned long long)inode->i_blocks,
1714 					i_size_read(inode), blocks);
1715 
1716 	/* Sanity check -- make sure we're going to fit. */
1717 	if (bytes_left >
1718 	    ocfs2_clusters_to_bytes(sb, OCFS2_I(inode)->ip_clusters)) {
1719 		status = -EIO;
1720 		mlog_errno(status);
1721 		goto bail;
1722 	}
1723 
1724 	bhs = kcalloc(blocks, sizeof(struct buffer_head *), GFP_KERNEL);
1725 	if (!bhs) {
1726 		status = -ENOMEM;
1727 		mlog_errno(status);
1728 		goto bail;
1729 	}
1730 
1731 	status = ocfs2_extent_map_get_blocks(inode, 0, &p_blkno, &p_blocks,
1732 					     NULL);
1733 	if (status < 0) {
1734 		mlog_errno(status);
1735 		goto bail;
1736 	}
1737 
1738 	/* links can never be larger than one cluster so we know this
1739 	 * is all going to be contiguous, but do a sanity check
1740 	 * anyway. */
1741 	if ((p_blocks << sb->s_blocksize_bits) < bytes_left) {
1742 		status = -EIO;
1743 		mlog_errno(status);
1744 		goto bail;
1745 	}
1746 
1747 	virtual = 0;
1748 	while(bytes_left > 0) {
1749 		c = &symname[virtual * sb->s_blocksize];
1750 
1751 		bhs[virtual] = sb_getblk(sb, p_blkno);
1752 		if (!bhs[virtual]) {
1753 			status = -ENOMEM;
1754 			mlog_errno(status);
1755 			goto bail;
1756 		}
1757 		ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode),
1758 					      bhs[virtual]);
1759 
1760 		status = ocfs2_journal_access(handle, INODE_CACHE(inode),
1761 					      bhs[virtual],
1762 					      OCFS2_JOURNAL_ACCESS_CREATE);
1763 		if (status < 0) {
1764 			mlog_errno(status);
1765 			goto bail;
1766 		}
1767 
1768 		memset(bhs[virtual]->b_data, 0, sb->s_blocksize);
1769 
1770 		memcpy(bhs[virtual]->b_data, c,
1771 		       (bytes_left > sb->s_blocksize) ? sb->s_blocksize :
1772 		       bytes_left);
1773 
1774 		ocfs2_journal_dirty(handle, bhs[virtual]);
1775 
1776 		virtual++;
1777 		p_blkno++;
1778 		bytes_left -= sb->s_blocksize;
1779 	}
1780 
1781 	status = 0;
1782 bail:
1783 
1784 	if (bhs) {
1785 		for(i = 0; i < blocks; i++)
1786 			brelse(bhs[i]);
1787 		kfree(bhs);
1788 	}
1789 
1790 	if (status)
1791 		mlog_errno(status);
1792 	return status;
1793 }
1794 
1795 static int ocfs2_symlink(struct inode *dir,
1796 			 struct dentry *dentry,
1797 			 const char *symname)
1798 {
1799 	int status, l, credits;
1800 	u64 newsize;
1801 	struct ocfs2_super *osb = NULL;
1802 	struct inode *inode = NULL;
1803 	struct super_block *sb;
1804 	struct buffer_head *new_fe_bh = NULL;
1805 	struct buffer_head *parent_fe_bh = NULL;
1806 	struct ocfs2_dinode *fe = NULL;
1807 	struct ocfs2_dinode *dirfe;
1808 	handle_t *handle = NULL;
1809 	struct ocfs2_alloc_context *inode_ac = NULL;
1810 	struct ocfs2_alloc_context *data_ac = NULL;
1811 	struct ocfs2_alloc_context *xattr_ac = NULL;
1812 	int want_clusters = 0;
1813 	int xattr_credits = 0;
1814 	struct ocfs2_security_xattr_info si = {
1815 		.enable = 1,
1816 	};
1817 	int did_quota = 0, did_quota_inode = 0;
1818 	struct ocfs2_dir_lookup_result lookup = { NULL, };
1819 	sigset_t oldset;
1820 	int did_block_signals = 0;
1821 	struct ocfs2_dentry_lock *dl = NULL;
1822 
1823 	trace_ocfs2_symlink_begin(dir, dentry, symname,
1824 				  dentry->d_name.len, dentry->d_name.name);
1825 
1826 	status = dquot_initialize(dir);
1827 	if (status) {
1828 		mlog_errno(status);
1829 		goto bail;
1830 	}
1831 
1832 	sb = dir->i_sb;
1833 	osb = OCFS2_SB(sb);
1834 
1835 	l = strlen(symname) + 1;
1836 
1837 	credits = ocfs2_calc_symlink_credits(sb);
1838 
1839 	/* lock the parent directory */
1840 	status = ocfs2_inode_lock(dir, &parent_fe_bh, 1);
1841 	if (status < 0) {
1842 		if (status != -ENOENT)
1843 			mlog_errno(status);
1844 		return status;
1845 	}
1846 
1847 	dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
1848 	if (!ocfs2_read_links_count(dirfe)) {
1849 		/* can't make a file in a deleted directory. */
1850 		status = -ENOENT;
1851 		goto bail;
1852 	}
1853 
1854 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
1855 					   dentry->d_name.len);
1856 	if (status)
1857 		goto bail;
1858 
1859 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
1860 					      dentry->d_name.name,
1861 					      dentry->d_name.len, &lookup);
1862 	if (status < 0) {
1863 		mlog_errno(status);
1864 		goto bail;
1865 	}
1866 
1867 	status = ocfs2_reserve_new_inode(osb, &inode_ac);
1868 	if (status < 0) {
1869 		if (status != -ENOSPC)
1870 			mlog_errno(status);
1871 		goto bail;
1872 	}
1873 
1874 	inode = ocfs2_get_init_inode(dir, S_IFLNK | S_IRWXUGO);
1875 	if (IS_ERR(inode)) {
1876 		status = PTR_ERR(inode);
1877 		inode = NULL;
1878 		mlog_errno(status);
1879 		goto bail;
1880 	}
1881 
1882 	/* get security xattr */
1883 	status = ocfs2_init_security_get(inode, dir, &dentry->d_name, &si);
1884 	if (status) {
1885 		if (status == -EOPNOTSUPP)
1886 			si.enable = 0;
1887 		else {
1888 			mlog_errno(status);
1889 			goto bail;
1890 		}
1891 	}
1892 
1893 	/* calculate meta data/clusters for setting security xattr */
1894 	if (si.enable) {
1895 		status = ocfs2_calc_security_init(dir, &si, &want_clusters,
1896 						  &xattr_credits, &xattr_ac);
1897 		if (status < 0) {
1898 			mlog_errno(status);
1899 			goto bail;
1900 		}
1901 	}
1902 
1903 	/* don't reserve bitmap space for fast symlinks. */
1904 	if (l > ocfs2_fast_symlink_chars(sb))
1905 		want_clusters += 1;
1906 
1907 	status = ocfs2_reserve_clusters(osb, want_clusters, &data_ac);
1908 	if (status < 0) {
1909 		if (status != -ENOSPC)
1910 			mlog_errno(status);
1911 		goto bail;
1912 	}
1913 
1914 	handle = ocfs2_start_trans(osb, credits + xattr_credits);
1915 	if (IS_ERR(handle)) {
1916 		status = PTR_ERR(handle);
1917 		handle = NULL;
1918 		mlog_errno(status);
1919 		goto bail;
1920 	}
1921 
1922 	/* Starting to change things, restart is no longer possible. */
1923 	ocfs2_block_signals(&oldset);
1924 	did_block_signals = 1;
1925 
1926 	status = dquot_alloc_inode(inode);
1927 	if (status)
1928 		goto bail;
1929 	did_quota_inode = 1;
1930 
1931 	trace_ocfs2_symlink_create(dir, dentry, dentry->d_name.len,
1932 				   dentry->d_name.name,
1933 				   (unsigned long long)OCFS2_I(dir)->ip_blkno,
1934 				   inode->i_mode);
1935 
1936 	status = ocfs2_mknod_locked(osb, dir, inode,
1937 				    0, &new_fe_bh, parent_fe_bh, handle,
1938 				    inode_ac);
1939 	if (status < 0) {
1940 		mlog_errno(status);
1941 		goto bail;
1942 	}
1943 
1944 	fe = (struct ocfs2_dinode *) new_fe_bh->b_data;
1945 	inode->i_rdev = 0;
1946 	newsize = l - 1;
1947 	inode->i_op = &ocfs2_symlink_inode_operations;
1948 	inode_nohighmem(inode);
1949 	if (l > ocfs2_fast_symlink_chars(sb)) {
1950 		u32 offset = 0;
1951 
1952 		status = dquot_alloc_space_nodirty(inode,
1953 		    ocfs2_clusters_to_bytes(osb->sb, 1));
1954 		if (status)
1955 			goto bail;
1956 		did_quota = 1;
1957 		inode->i_mapping->a_ops = &ocfs2_aops;
1958 		status = ocfs2_add_inode_data(osb, inode, &offset, 1, 0,
1959 					      new_fe_bh,
1960 					      handle, data_ac, NULL,
1961 					      NULL);
1962 		if (status < 0) {
1963 			if (status != -ENOSPC && status != -EINTR) {
1964 				mlog(ML_ERROR,
1965 				     "Failed to extend file to %llu\n",
1966 				     (unsigned long long)newsize);
1967 				mlog_errno(status);
1968 				status = -ENOSPC;
1969 			}
1970 			goto bail;
1971 		}
1972 		i_size_write(inode, newsize);
1973 		inode->i_blocks = ocfs2_inode_sector_count(inode);
1974 	} else {
1975 		inode->i_mapping->a_ops = &ocfs2_fast_symlink_aops;
1976 		memcpy((char *) fe->id2.i_symlink, symname, l);
1977 		i_size_write(inode, newsize);
1978 		inode->i_blocks = 0;
1979 	}
1980 
1981 	status = ocfs2_mark_inode_dirty(handle, inode, new_fe_bh);
1982 	if (status < 0) {
1983 		mlog_errno(status);
1984 		goto bail;
1985 	}
1986 
1987 	if (!ocfs2_inode_is_fast_symlink(inode)) {
1988 		status = ocfs2_create_symlink_data(osb, handle, inode,
1989 						   symname);
1990 		if (status < 0) {
1991 			mlog_errno(status);
1992 			goto bail;
1993 		}
1994 	}
1995 
1996 	if (si.enable) {
1997 		status = ocfs2_init_security_set(handle, inode, new_fe_bh, &si,
1998 						 xattr_ac, data_ac);
1999 		if (status < 0) {
2000 			mlog_errno(status);
2001 			goto bail;
2002 		}
2003 	}
2004 
2005 	/*
2006 	 * Do this before adding the entry to the directory. We add
2007 	 * also set d_op after success so that ->d_iput() will cleanup
2008 	 * the dentry lock even if ocfs2_add_entry() fails below.
2009 	 */
2010 	status = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
2011 	if (status) {
2012 		mlog_errno(status);
2013 		goto bail;
2014 	}
2015 
2016 	dl = dentry->d_fsdata;
2017 
2018 	status = ocfs2_add_entry(handle, dentry, inode,
2019 				 le64_to_cpu(fe->i_blkno), parent_fe_bh,
2020 				 &lookup);
2021 	if (status < 0) {
2022 		mlog_errno(status);
2023 		goto bail;
2024 	}
2025 
2026 	insert_inode_hash(inode);
2027 	d_instantiate(dentry, inode);
2028 bail:
2029 	if (status < 0 && did_quota)
2030 		dquot_free_space_nodirty(inode,
2031 					ocfs2_clusters_to_bytes(osb->sb, 1));
2032 	if (status < 0 && did_quota_inode)
2033 		dquot_free_inode(inode);
2034 	if (handle)
2035 		ocfs2_commit_trans(osb, handle);
2036 
2037 	ocfs2_inode_unlock(dir, 1);
2038 	if (did_block_signals)
2039 		ocfs2_unblock_signals(&oldset);
2040 
2041 	brelse(new_fe_bh);
2042 	brelse(parent_fe_bh);
2043 	kfree(si.value);
2044 	ocfs2_free_dir_lookup_result(&lookup);
2045 	if (inode_ac)
2046 		ocfs2_free_alloc_context(inode_ac);
2047 	if (data_ac)
2048 		ocfs2_free_alloc_context(data_ac);
2049 	if (xattr_ac)
2050 		ocfs2_free_alloc_context(xattr_ac);
2051 	if ((status < 0) && inode) {
2052 		if (dl)
2053 			ocfs2_cleanup_add_entry_failure(osb, dentry, inode);
2054 
2055 		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR;
2056 		clear_nlink(inode);
2057 		iput(inode);
2058 	}
2059 
2060 	if (status)
2061 		mlog_errno(status);
2062 
2063 	return status;
2064 }
2065 
2066 static int ocfs2_blkno_stringify(u64 blkno, char *name)
2067 {
2068 	int status, namelen;
2069 
2070 	namelen = snprintf(name, OCFS2_ORPHAN_NAMELEN + 1, "%016llx",
2071 			   (long long)blkno);
2072 	if (namelen <= 0) {
2073 		if (namelen)
2074 			status = namelen;
2075 		else
2076 			status = -EINVAL;
2077 		mlog_errno(status);
2078 		goto bail;
2079 	}
2080 	if (namelen != OCFS2_ORPHAN_NAMELEN) {
2081 		status = -EINVAL;
2082 		mlog_errno(status);
2083 		goto bail;
2084 	}
2085 
2086 	trace_ocfs2_blkno_stringify(blkno, name, namelen);
2087 
2088 	status = 0;
2089 bail:
2090 	if (status < 0)
2091 		mlog_errno(status);
2092 	return status;
2093 }
2094 
2095 static int ocfs2_lookup_lock_orphan_dir(struct ocfs2_super *osb,
2096 					struct inode **ret_orphan_dir,
2097 					struct buffer_head **ret_orphan_dir_bh)
2098 {
2099 	struct inode *orphan_dir_inode;
2100 	struct buffer_head *orphan_dir_bh = NULL;
2101 	int ret = 0;
2102 
2103 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2104 						       ORPHAN_DIR_SYSTEM_INODE,
2105 						       osb->slot_num);
2106 	if (!orphan_dir_inode) {
2107 		ret = -ENOENT;
2108 		mlog_errno(ret);
2109 		return ret;
2110 	}
2111 
2112 	inode_lock(orphan_dir_inode);
2113 
2114 	ret = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2115 	if (ret < 0) {
2116 		inode_unlock(orphan_dir_inode);
2117 		iput(orphan_dir_inode);
2118 
2119 		mlog_errno(ret);
2120 		return ret;
2121 	}
2122 
2123 	*ret_orphan_dir = orphan_dir_inode;
2124 	*ret_orphan_dir_bh = orphan_dir_bh;
2125 
2126 	return 0;
2127 }
2128 
2129 static int __ocfs2_prepare_orphan_dir(struct inode *orphan_dir_inode,
2130 				      struct buffer_head *orphan_dir_bh,
2131 				      u64 blkno,
2132 				      char *name,
2133 				      struct ocfs2_dir_lookup_result *lookup,
2134 				      bool dio)
2135 {
2136 	int ret;
2137 	struct ocfs2_super *osb = OCFS2_SB(orphan_dir_inode->i_sb);
2138 	int namelen = dio ?
2139 			(OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN) :
2140 			OCFS2_ORPHAN_NAMELEN;
2141 
2142 	if (dio) {
2143 		ret = snprintf(name, OCFS2_DIO_ORPHAN_PREFIX_LEN + 1, "%s",
2144 				OCFS2_DIO_ORPHAN_PREFIX);
2145 		if (ret != OCFS2_DIO_ORPHAN_PREFIX_LEN) {
2146 			ret = -EINVAL;
2147 			mlog_errno(ret);
2148 			return ret;
2149 		}
2150 
2151 		ret = ocfs2_blkno_stringify(blkno,
2152 				name + OCFS2_DIO_ORPHAN_PREFIX_LEN);
2153 	} else
2154 		ret = ocfs2_blkno_stringify(blkno, name);
2155 	if (ret < 0) {
2156 		mlog_errno(ret);
2157 		return ret;
2158 	}
2159 
2160 	ret = ocfs2_prepare_dir_for_insert(osb, orphan_dir_inode,
2161 					   orphan_dir_bh, name,
2162 					   namelen, lookup);
2163 	if (ret < 0) {
2164 		mlog_errno(ret);
2165 		return ret;
2166 	}
2167 
2168 	return 0;
2169 }
2170 
2171 /**
2172  * ocfs2_prepare_orphan_dir() - Prepare an orphan directory for
2173  * insertion of an orphan.
2174  * @osb: ocfs2 file system
2175  * @ret_orphan_dir: Orphan dir inode - returned locked!
2176  * @blkno: Actual block number of the inode to be inserted into orphan dir.
2177  * @lookup: dir lookup result, to be passed back into functions like
2178  *          ocfs2_orphan_add
2179  *
2180  * Returns zero on success and the ret_orphan_dir, name and lookup
2181  * fields will be populated.
2182  *
2183  * Returns non-zero on failure.
2184  */
2185 static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
2186 				    struct inode **ret_orphan_dir,
2187 				    u64 blkno,
2188 				    char *name,
2189 				    struct ocfs2_dir_lookup_result *lookup,
2190 				    bool dio)
2191 {
2192 	struct inode *orphan_dir_inode = NULL;
2193 	struct buffer_head *orphan_dir_bh = NULL;
2194 	int ret = 0;
2195 
2196 	ret = ocfs2_lookup_lock_orphan_dir(osb, &orphan_dir_inode,
2197 					   &orphan_dir_bh);
2198 	if (ret < 0) {
2199 		mlog_errno(ret);
2200 		return ret;
2201 	}
2202 
2203 	ret = __ocfs2_prepare_orphan_dir(orphan_dir_inode, orphan_dir_bh,
2204 					 blkno, name, lookup, dio);
2205 	if (ret < 0) {
2206 		mlog_errno(ret);
2207 		goto out;
2208 	}
2209 
2210 	*ret_orphan_dir = orphan_dir_inode;
2211 
2212 out:
2213 	brelse(orphan_dir_bh);
2214 
2215 	if (ret) {
2216 		ocfs2_inode_unlock(orphan_dir_inode, 1);
2217 		inode_unlock(orphan_dir_inode);
2218 		iput(orphan_dir_inode);
2219 	}
2220 
2221 	if (ret)
2222 		mlog_errno(ret);
2223 	return ret;
2224 }
2225 
2226 static int ocfs2_orphan_add(struct ocfs2_super *osb,
2227 			    handle_t *handle,
2228 			    struct inode *inode,
2229 			    struct buffer_head *fe_bh,
2230 			    char *name,
2231 			    struct ocfs2_dir_lookup_result *lookup,
2232 			    struct inode *orphan_dir_inode,
2233 			    bool dio)
2234 {
2235 	struct buffer_head *orphan_dir_bh = NULL;
2236 	int status = 0;
2237 	struct ocfs2_dinode *orphan_fe;
2238 	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data;
2239 	int namelen = dio ?
2240 			(OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN) :
2241 			OCFS2_ORPHAN_NAMELEN;
2242 
2243 	trace_ocfs2_orphan_add_begin(
2244 				(unsigned long long)OCFS2_I(inode)->ip_blkno);
2245 
2246 	status = ocfs2_read_inode_block(orphan_dir_inode, &orphan_dir_bh);
2247 	if (status < 0) {
2248 		mlog_errno(status);
2249 		goto leave;
2250 	}
2251 
2252 	status = ocfs2_journal_access_di(handle,
2253 					 INODE_CACHE(orphan_dir_inode),
2254 					 orphan_dir_bh,
2255 					 OCFS2_JOURNAL_ACCESS_WRITE);
2256 	if (status < 0) {
2257 		mlog_errno(status);
2258 		goto leave;
2259 	}
2260 
2261 	/*
2262 	 * We're going to journal the change of i_flags and i_orphaned_slot.
2263 	 * It's safe anyway, though some callers may duplicate the journaling.
2264 	 * Journaling within the func just make the logic look more
2265 	 * straightforward.
2266 	 */
2267 	status = ocfs2_journal_access_di(handle,
2268 					 INODE_CACHE(inode),
2269 					 fe_bh,
2270 					 OCFS2_JOURNAL_ACCESS_WRITE);
2271 	if (status < 0) {
2272 		mlog_errno(status);
2273 		goto leave;
2274 	}
2275 
2276 	/* we're a cluster, and nlink can change on disk from
2277 	 * underneath us... */
2278 	orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data;
2279 	if (S_ISDIR(inode->i_mode))
2280 		ocfs2_add_links_count(orphan_fe, 1);
2281 	set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2282 	ocfs2_journal_dirty(handle, orphan_dir_bh);
2283 
2284 	status = __ocfs2_add_entry(handle, orphan_dir_inode, name,
2285 				   namelen, inode,
2286 				   OCFS2_I(inode)->ip_blkno,
2287 				   orphan_dir_bh, lookup);
2288 	if (status < 0) {
2289 		mlog_errno(status);
2290 		goto rollback;
2291 	}
2292 
2293 	if (dio) {
2294 		/* Update flag OCFS2_DIO_ORPHANED_FL and record the orphan
2295 		 * slot.
2296 		 */
2297 		fe->i_flags |= cpu_to_le32(OCFS2_DIO_ORPHANED_FL);
2298 		fe->i_dio_orphaned_slot = cpu_to_le16(osb->slot_num);
2299 	} else {
2300 		fe->i_flags |= cpu_to_le32(OCFS2_ORPHANED_FL);
2301 		OCFS2_I(inode)->ip_flags &= ~OCFS2_INODE_SKIP_ORPHAN_DIR;
2302 
2303 		/* Record which orphan dir our inode now resides
2304 		 * in. delete_inode will use this to determine which orphan
2305 		 * dir to lock. */
2306 		fe->i_orphaned_slot = cpu_to_le16(osb->slot_num);
2307 	}
2308 
2309 	ocfs2_journal_dirty(handle, fe_bh);
2310 
2311 	trace_ocfs2_orphan_add_end((unsigned long long)OCFS2_I(inode)->ip_blkno,
2312 				   osb->slot_num);
2313 
2314 rollback:
2315 	if (status < 0) {
2316 		if (S_ISDIR(inode->i_mode))
2317 			ocfs2_add_links_count(orphan_fe, -1);
2318 		set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2319 	}
2320 
2321 leave:
2322 	brelse(orphan_dir_bh);
2323 
2324 	return status;
2325 }
2326 
2327 /* unlike orphan_add, we expect the orphan dir to already be locked here. */
2328 int ocfs2_orphan_del(struct ocfs2_super *osb,
2329 		     handle_t *handle,
2330 		     struct inode *orphan_dir_inode,
2331 		     struct inode *inode,
2332 		     struct buffer_head *orphan_dir_bh,
2333 		     bool dio)
2334 {
2335 	const int namelen = OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN;
2336 	char name[namelen + 1];
2337 	struct ocfs2_dinode *orphan_fe;
2338 	int status = 0;
2339 	struct ocfs2_dir_lookup_result lookup = { NULL, };
2340 
2341 	if (dio) {
2342 		status = snprintf(name, OCFS2_DIO_ORPHAN_PREFIX_LEN + 1, "%s",
2343 				OCFS2_DIO_ORPHAN_PREFIX);
2344 		if (status != OCFS2_DIO_ORPHAN_PREFIX_LEN) {
2345 			status = -EINVAL;
2346 			mlog_errno(status);
2347 			return status;
2348 		}
2349 
2350 		status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno,
2351 				name + OCFS2_DIO_ORPHAN_PREFIX_LEN);
2352 	} else
2353 		status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name);
2354 	if (status < 0) {
2355 		mlog_errno(status);
2356 		goto leave;
2357 	}
2358 
2359 	trace_ocfs2_orphan_del(
2360 	     (unsigned long long)OCFS2_I(orphan_dir_inode)->ip_blkno,
2361 	     name, strlen(name));
2362 
2363 	status = ocfs2_journal_access_di(handle,
2364 					 INODE_CACHE(orphan_dir_inode),
2365 					 orphan_dir_bh,
2366 					 OCFS2_JOURNAL_ACCESS_WRITE);
2367 	if (status < 0) {
2368 		mlog_errno(status);
2369 		goto leave;
2370 	}
2371 
2372 	/* find it's spot in the orphan directory */
2373 	status = ocfs2_find_entry(name, strlen(name), orphan_dir_inode,
2374 				  &lookup);
2375 	if (status) {
2376 		mlog_errno(status);
2377 		goto leave;
2378 	}
2379 
2380 	/* remove it from the orphan directory */
2381 	status = ocfs2_delete_entry(handle, orphan_dir_inode, &lookup);
2382 	if (status < 0) {
2383 		mlog_errno(status);
2384 		goto leave;
2385 	}
2386 
2387 	/* do the i_nlink dance! :) */
2388 	orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data;
2389 	if (S_ISDIR(inode->i_mode))
2390 		ocfs2_add_links_count(orphan_fe, -1);
2391 	set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2392 	ocfs2_journal_dirty(handle, orphan_dir_bh);
2393 
2394 leave:
2395 	ocfs2_free_dir_lookup_result(&lookup);
2396 
2397 	if (status)
2398 		mlog_errno(status);
2399 	return status;
2400 }
2401 
2402 /**
2403  * ocfs2_prep_new_orphaned_file() - Prepare the orphan dir to receive a newly
2404  * allocated file. This is different from the typical 'add to orphan dir'
2405  * operation in that the inode does not yet exist. This is a problem because
2406  * the orphan dir stringifies the inode block number to come up with it's
2407  * dirent. Obviously if the inode does not yet exist we have a chicken and egg
2408  * problem. This function works around it by calling deeper into the orphan
2409  * and suballoc code than other callers. Use this only by necessity.
2410  * @dir: The directory which this inode will ultimately wind up under - not the
2411  * orphan dir!
2412  * @dir_bh: buffer_head the @dir inode block
2413  * @orphan_name: string of length (CFS2_ORPHAN_NAMELEN + 1). Will be filled
2414  * with the string to be used for orphan dirent. Pass back to the orphan dir
2415  * code.
2416  * @ret_orphan_dir: orphan dir inode returned to be passed back into orphan
2417  * dir code.
2418  * @ret_di_blkno: block number where the new inode will be allocated.
2419  * @orphan_insert: Dir insert context to be passed back into orphan dir code.
2420  * @ret_inode_ac: Inode alloc context to be passed back to the allocator.
2421  *
2422  * Returns zero on success and the ret_orphan_dir, name and lookup
2423  * fields will be populated.
2424  *
2425  * Returns non-zero on failure.
2426  */
2427 static int ocfs2_prep_new_orphaned_file(struct inode *dir,
2428 					struct buffer_head *dir_bh,
2429 					char *orphan_name,
2430 					struct inode **ret_orphan_dir,
2431 					u64 *ret_di_blkno,
2432 					struct ocfs2_dir_lookup_result *orphan_insert,
2433 					struct ocfs2_alloc_context **ret_inode_ac)
2434 {
2435 	int ret;
2436 	u64 di_blkno;
2437 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
2438 	struct inode *orphan_dir = NULL;
2439 	struct buffer_head *orphan_dir_bh = NULL;
2440 	struct ocfs2_alloc_context *inode_ac = NULL;
2441 
2442 	ret = ocfs2_lookup_lock_orphan_dir(osb, &orphan_dir, &orphan_dir_bh);
2443 	if (ret < 0) {
2444 		mlog_errno(ret);
2445 		return ret;
2446 	}
2447 
2448 	/* reserve an inode spot */
2449 	ret = ocfs2_reserve_new_inode(osb, &inode_ac);
2450 	if (ret < 0) {
2451 		if (ret != -ENOSPC)
2452 			mlog_errno(ret);
2453 		goto out;
2454 	}
2455 
2456 	ret = ocfs2_find_new_inode_loc(dir, dir_bh, inode_ac,
2457 				       &di_blkno);
2458 	if (ret) {
2459 		mlog_errno(ret);
2460 		goto out;
2461 	}
2462 
2463 	ret = __ocfs2_prepare_orphan_dir(orphan_dir, orphan_dir_bh,
2464 					 di_blkno, orphan_name, orphan_insert,
2465 					 false);
2466 	if (ret < 0) {
2467 		mlog_errno(ret);
2468 		goto out;
2469 	}
2470 
2471 out:
2472 	if (ret == 0) {
2473 		*ret_orphan_dir = orphan_dir;
2474 		*ret_di_blkno = di_blkno;
2475 		*ret_inode_ac = inode_ac;
2476 		/*
2477 		 * orphan_name and orphan_insert are already up to
2478 		 * date via prepare_orphan_dir
2479 		 */
2480 	} else {
2481 		/* Unroll reserve_new_inode* */
2482 		if (inode_ac)
2483 			ocfs2_free_alloc_context(inode_ac);
2484 
2485 		/* Unroll orphan dir locking */
2486 		inode_unlock(orphan_dir);
2487 		ocfs2_inode_unlock(orphan_dir, 1);
2488 		iput(orphan_dir);
2489 	}
2490 
2491 	brelse(orphan_dir_bh);
2492 
2493 	return ret;
2494 }
2495 
2496 int ocfs2_create_inode_in_orphan(struct inode *dir,
2497 				 int mode,
2498 				 struct inode **new_inode)
2499 {
2500 	int status, did_quota_inode = 0;
2501 	struct inode *inode = NULL;
2502 	struct inode *orphan_dir = NULL;
2503 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
2504 	struct ocfs2_dinode *di = NULL;
2505 	handle_t *handle = NULL;
2506 	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
2507 	struct buffer_head *parent_di_bh = NULL;
2508 	struct buffer_head *new_di_bh = NULL;
2509 	struct ocfs2_alloc_context *inode_ac = NULL;
2510 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
2511 	u64 uninitialized_var(di_blkno), suballoc_loc;
2512 	u16 suballoc_bit;
2513 
2514 	status = ocfs2_inode_lock(dir, &parent_di_bh, 1);
2515 	if (status < 0) {
2516 		if (status != -ENOENT)
2517 			mlog_errno(status);
2518 		return status;
2519 	}
2520 
2521 	status = ocfs2_prep_new_orphaned_file(dir, parent_di_bh,
2522 					      orphan_name, &orphan_dir,
2523 					      &di_blkno, &orphan_insert, &inode_ac);
2524 	if (status < 0) {
2525 		if (status != -ENOSPC)
2526 			mlog_errno(status);
2527 		goto leave;
2528 	}
2529 
2530 	inode = ocfs2_get_init_inode(dir, mode);
2531 	if (IS_ERR(inode)) {
2532 		status = PTR_ERR(inode);
2533 		inode = NULL;
2534 		mlog_errno(status);
2535 		goto leave;
2536 	}
2537 
2538 	handle = ocfs2_start_trans(osb, ocfs2_mknod_credits(osb->sb, 0, 0));
2539 	if (IS_ERR(handle)) {
2540 		status = PTR_ERR(handle);
2541 		handle = NULL;
2542 		mlog_errno(status);
2543 		goto leave;
2544 	}
2545 
2546 	status = dquot_alloc_inode(inode);
2547 	if (status)
2548 		goto leave;
2549 	did_quota_inode = 1;
2550 
2551 	status = ocfs2_claim_new_inode_at_loc(handle, dir, inode_ac,
2552 					      &suballoc_loc,
2553 					      &suballoc_bit, di_blkno);
2554 	if (status < 0) {
2555 		mlog_errno(status);
2556 		goto leave;
2557 	}
2558 
2559 	clear_nlink(inode);
2560 	/* do the real work now. */
2561 	status = __ocfs2_mknod_locked(dir, inode,
2562 				      0, &new_di_bh, parent_di_bh, handle,
2563 				      inode_ac, di_blkno, suballoc_loc,
2564 				      suballoc_bit);
2565 	if (status < 0) {
2566 		mlog_errno(status);
2567 		goto leave;
2568 	}
2569 
2570 	di = (struct ocfs2_dinode *)new_di_bh->b_data;
2571 	status = ocfs2_orphan_add(osb, handle, inode, new_di_bh, orphan_name,
2572 				  &orphan_insert, orphan_dir, false);
2573 	if (status < 0) {
2574 		mlog_errno(status);
2575 		goto leave;
2576 	}
2577 
2578 	/* get open lock so that only nodes can't remove it from orphan dir. */
2579 	status = ocfs2_open_lock(inode);
2580 	if (status < 0)
2581 		mlog_errno(status);
2582 
2583 	insert_inode_hash(inode);
2584 leave:
2585 	if (status < 0 && did_quota_inode)
2586 		dquot_free_inode(inode);
2587 	if (handle)
2588 		ocfs2_commit_trans(osb, handle);
2589 
2590 	if (orphan_dir) {
2591 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
2592 		ocfs2_inode_unlock(orphan_dir, 1);
2593 		inode_unlock(orphan_dir);
2594 		iput(orphan_dir);
2595 	}
2596 
2597 	if ((status < 0) && inode) {
2598 		clear_nlink(inode);
2599 		iput(inode);
2600 	}
2601 
2602 	if (inode_ac)
2603 		ocfs2_free_alloc_context(inode_ac);
2604 
2605 	brelse(new_di_bh);
2606 
2607 	if (!status)
2608 		*new_inode = inode;
2609 
2610 	ocfs2_free_dir_lookup_result(&orphan_insert);
2611 
2612 	ocfs2_inode_unlock(dir, 1);
2613 	brelse(parent_di_bh);
2614 	return status;
2615 }
2616 
2617 int ocfs2_add_inode_to_orphan(struct ocfs2_super *osb,
2618 	struct inode *inode)
2619 {
2620 	char orphan_name[OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN + 1];
2621 	struct inode *orphan_dir_inode = NULL;
2622 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
2623 	struct buffer_head *di_bh = NULL;
2624 	int status = 0;
2625 	handle_t *handle = NULL;
2626 	struct ocfs2_dinode *di = NULL;
2627 
2628 	status = ocfs2_inode_lock(inode, &di_bh, 1);
2629 	if (status < 0) {
2630 		mlog_errno(status);
2631 		goto bail;
2632 	}
2633 
2634 	di = (struct ocfs2_dinode *) di_bh->b_data;
2635 	/*
2636 	 * Another append dio crashed?
2637 	 * If so, manually recover it first.
2638 	 */
2639 	if (unlikely(di->i_flags & cpu_to_le32(OCFS2_DIO_ORPHANED_FL))) {
2640 		status = ocfs2_truncate_file(inode, di_bh, i_size_read(inode));
2641 		if (status < 0) {
2642 			if (status != -ENOSPC)
2643 				mlog_errno(status);
2644 			goto bail_unlock_inode;
2645 		}
2646 
2647 		status = ocfs2_del_inode_from_orphan(osb, inode, di_bh, 0, 0);
2648 		if (status < 0) {
2649 			mlog_errno(status);
2650 			goto bail_unlock_inode;
2651 		}
2652 	}
2653 
2654 	status = ocfs2_prepare_orphan_dir(osb, &orphan_dir_inode,
2655 			OCFS2_I(inode)->ip_blkno,
2656 			orphan_name,
2657 			&orphan_insert,
2658 			true);
2659 	if (status < 0) {
2660 		mlog_errno(status);
2661 		goto bail_unlock_inode;
2662 	}
2663 
2664 	handle = ocfs2_start_trans(osb,
2665 			OCFS2_INODE_ADD_TO_ORPHAN_CREDITS);
2666 	if (IS_ERR(handle)) {
2667 		status = PTR_ERR(handle);
2668 		goto bail_unlock_orphan;
2669 	}
2670 
2671 	status = ocfs2_orphan_add(osb, handle, inode, di_bh, orphan_name,
2672 			&orphan_insert, orphan_dir_inode, true);
2673 	if (status)
2674 		mlog_errno(status);
2675 
2676 	ocfs2_commit_trans(osb, handle);
2677 
2678 bail_unlock_orphan:
2679 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2680 	inode_unlock(orphan_dir_inode);
2681 	iput(orphan_dir_inode);
2682 
2683 	ocfs2_free_dir_lookup_result(&orphan_insert);
2684 
2685 bail_unlock_inode:
2686 	ocfs2_inode_unlock(inode, 1);
2687 	brelse(di_bh);
2688 
2689 bail:
2690 	return status;
2691 }
2692 
2693 int ocfs2_del_inode_from_orphan(struct ocfs2_super *osb,
2694 		struct inode *inode, struct buffer_head *di_bh,
2695 		int update_isize, loff_t end)
2696 {
2697 	struct inode *orphan_dir_inode = NULL;
2698 	struct buffer_head *orphan_dir_bh = NULL;
2699 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
2700 	handle_t *handle = NULL;
2701 	int status = 0;
2702 
2703 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2704 			ORPHAN_DIR_SYSTEM_INODE,
2705 			le16_to_cpu(di->i_dio_orphaned_slot));
2706 	if (!orphan_dir_inode) {
2707 		status = -ENOENT;
2708 		mlog_errno(status);
2709 		goto bail;
2710 	}
2711 
2712 	inode_lock(orphan_dir_inode);
2713 	status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2714 	if (status < 0) {
2715 		inode_unlock(orphan_dir_inode);
2716 		iput(orphan_dir_inode);
2717 		mlog_errno(status);
2718 		goto bail;
2719 	}
2720 
2721 	handle = ocfs2_start_trans(osb,
2722 			OCFS2_INODE_DEL_FROM_ORPHAN_CREDITS);
2723 	if (IS_ERR(handle)) {
2724 		status = PTR_ERR(handle);
2725 		goto bail_unlock_orphan;
2726 	}
2727 
2728 	BUG_ON(!(di->i_flags & cpu_to_le32(OCFS2_DIO_ORPHANED_FL)));
2729 
2730 	status = ocfs2_orphan_del(osb, handle, orphan_dir_inode,
2731 				inode, orphan_dir_bh, true);
2732 	if (status < 0) {
2733 		mlog_errno(status);
2734 		goto bail_commit;
2735 	}
2736 
2737 	status = ocfs2_journal_access_di(handle,
2738 			INODE_CACHE(inode),
2739 			di_bh,
2740 			OCFS2_JOURNAL_ACCESS_WRITE);
2741 	if (status < 0) {
2742 		mlog_errno(status);
2743 		goto bail_commit;
2744 	}
2745 
2746 	di->i_flags &= ~cpu_to_le32(OCFS2_DIO_ORPHANED_FL);
2747 	di->i_dio_orphaned_slot = 0;
2748 
2749 	if (update_isize) {
2750 		status = ocfs2_set_inode_size(handle, inode, di_bh, end);
2751 		if (status)
2752 			mlog_errno(status);
2753 	} else
2754 		ocfs2_journal_dirty(handle, di_bh);
2755 
2756 bail_commit:
2757 	ocfs2_commit_trans(osb, handle);
2758 
2759 bail_unlock_orphan:
2760 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2761 	inode_unlock(orphan_dir_inode);
2762 	brelse(orphan_dir_bh);
2763 	iput(orphan_dir_inode);
2764 
2765 bail:
2766 	return status;
2767 }
2768 
2769 int ocfs2_mv_orphaned_inode_to_new(struct inode *dir,
2770 				   struct inode *inode,
2771 				   struct dentry *dentry)
2772 {
2773 	int status = 0;
2774 	struct buffer_head *parent_di_bh = NULL;
2775 	handle_t *handle = NULL;
2776 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
2777 	struct ocfs2_dinode *dir_di, *di;
2778 	struct inode *orphan_dir_inode = NULL;
2779 	struct buffer_head *orphan_dir_bh = NULL;
2780 	struct buffer_head *di_bh = NULL;
2781 	struct ocfs2_dir_lookup_result lookup = { NULL, };
2782 
2783 	trace_ocfs2_mv_orphaned_inode_to_new(dir, dentry,
2784 				dentry->d_name.len, dentry->d_name.name,
2785 				(unsigned long long)OCFS2_I(dir)->ip_blkno,
2786 				(unsigned long long)OCFS2_I(inode)->ip_blkno);
2787 
2788 	status = ocfs2_inode_lock(dir, &parent_di_bh, 1);
2789 	if (status < 0) {
2790 		if (status != -ENOENT)
2791 			mlog_errno(status);
2792 		return status;
2793 	}
2794 
2795 	dir_di = (struct ocfs2_dinode *) parent_di_bh->b_data;
2796 	if (!dir_di->i_links_count) {
2797 		/* can't make a file in a deleted directory. */
2798 		status = -ENOENT;
2799 		goto leave;
2800 	}
2801 
2802 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
2803 					   dentry->d_name.len);
2804 	if (status)
2805 		goto leave;
2806 
2807 	/* get a spot inside the dir. */
2808 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_di_bh,
2809 					      dentry->d_name.name,
2810 					      dentry->d_name.len, &lookup);
2811 	if (status < 0) {
2812 		mlog_errno(status);
2813 		goto leave;
2814 	}
2815 
2816 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2817 						       ORPHAN_DIR_SYSTEM_INODE,
2818 						       osb->slot_num);
2819 	if (!orphan_dir_inode) {
2820 		status = -ENOENT;
2821 		mlog_errno(status);
2822 		goto leave;
2823 	}
2824 
2825 	inode_lock(orphan_dir_inode);
2826 
2827 	status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2828 	if (status < 0) {
2829 		mlog_errno(status);
2830 		inode_unlock(orphan_dir_inode);
2831 		iput(orphan_dir_inode);
2832 		goto leave;
2833 	}
2834 
2835 	status = ocfs2_read_inode_block(inode, &di_bh);
2836 	if (status < 0) {
2837 		mlog_errno(status);
2838 		goto orphan_unlock;
2839 	}
2840 
2841 	handle = ocfs2_start_trans(osb, ocfs2_rename_credits(osb->sb));
2842 	if (IS_ERR(handle)) {
2843 		status = PTR_ERR(handle);
2844 		handle = NULL;
2845 		mlog_errno(status);
2846 		goto orphan_unlock;
2847 	}
2848 
2849 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
2850 					 di_bh, OCFS2_JOURNAL_ACCESS_WRITE);
2851 	if (status < 0) {
2852 		mlog_errno(status);
2853 		goto out_commit;
2854 	}
2855 
2856 	status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode,
2857 				  orphan_dir_bh, false);
2858 	if (status < 0) {
2859 		mlog_errno(status);
2860 		goto out_commit;
2861 	}
2862 
2863 	di = (struct ocfs2_dinode *)di_bh->b_data;
2864 	di->i_flags &= ~cpu_to_le32(OCFS2_ORPHANED_FL);
2865 	di->i_orphaned_slot = 0;
2866 	set_nlink(inode, 1);
2867 	ocfs2_set_links_count(di, inode->i_nlink);
2868 	ocfs2_update_inode_fsync_trans(handle, inode, 1);
2869 	ocfs2_journal_dirty(handle, di_bh);
2870 
2871 	status = ocfs2_add_entry(handle, dentry, inode,
2872 				 OCFS2_I(inode)->ip_blkno, parent_di_bh,
2873 				 &lookup);
2874 	if (status < 0) {
2875 		mlog_errno(status);
2876 		goto out_commit;
2877 	}
2878 
2879 	status = ocfs2_dentry_attach_lock(dentry, inode,
2880 					  OCFS2_I(dir)->ip_blkno);
2881 	if (status) {
2882 		mlog_errno(status);
2883 		goto out_commit;
2884 	}
2885 
2886 	d_instantiate(dentry, inode);
2887 	status = 0;
2888 out_commit:
2889 	ocfs2_commit_trans(osb, handle);
2890 orphan_unlock:
2891 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2892 	inode_unlock(orphan_dir_inode);
2893 	iput(orphan_dir_inode);
2894 leave:
2895 
2896 	ocfs2_inode_unlock(dir, 1);
2897 
2898 	brelse(di_bh);
2899 	brelse(parent_di_bh);
2900 	brelse(orphan_dir_bh);
2901 
2902 	ocfs2_free_dir_lookup_result(&lookup);
2903 
2904 	if (status)
2905 		mlog_errno(status);
2906 
2907 	return status;
2908 }
2909 
2910 const struct inode_operations ocfs2_dir_iops = {
2911 	.create		= ocfs2_create,
2912 	.lookup		= ocfs2_lookup,
2913 	.link		= ocfs2_link,
2914 	.unlink		= ocfs2_unlink,
2915 	.rmdir		= ocfs2_unlink,
2916 	.symlink	= ocfs2_symlink,
2917 	.mkdir		= ocfs2_mkdir,
2918 	.mknod		= ocfs2_mknod,
2919 	.rename		= ocfs2_rename,
2920 	.setattr	= ocfs2_setattr,
2921 	.getattr	= ocfs2_getattr,
2922 	.permission	= ocfs2_permission,
2923 	.listxattr	= ocfs2_listxattr,
2924 	.fiemap         = ocfs2_fiemap,
2925 	.get_acl	= ocfs2_iop_get_acl,
2926 	.set_acl	= ocfs2_iop_set_acl,
2927 };
2928