xref: /openbmc/linux/fs/ocfs2/namei.c (revision a84d507d)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * namei.c
4  *
5  * Create and rename file, directory, symlinks
6  *
7  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
8  *
9  *  Portions of this code from linux/fs/ext3/dir.c
10  *
11  *  Copyright (C) 1992, 1993, 1994, 1995
12  *  Remy Card (card@masi.ibp.fr)
13  *  Laboratoire MASI - Institut Blaise pascal
14  *  Universite Pierre et Marie Curie (Paris VI)
15  *
16  *   from
17  *
18  *   linux/fs/minix/dir.c
19  *
20  *   Copyright (C) 1991, 1992 Linux Torvalds
21  */
22 
23 #include <linux/fs.h>
24 #include <linux/types.h>
25 #include <linux/slab.h>
26 #include <linux/highmem.h>
27 #include <linux/quotaops.h>
28 #include <linux/iversion.h>
29 
30 #include <cluster/masklog.h>
31 
32 #include "ocfs2.h"
33 
34 #include "alloc.h"
35 #include "dcache.h"
36 #include "dir.h"
37 #include "dlmglue.h"
38 #include "extent_map.h"
39 #include "file.h"
40 #include "inode.h"
41 #include "journal.h"
42 #include "namei.h"
43 #include "suballoc.h"
44 #include "super.h"
45 #include "symlink.h"
46 #include "sysfile.h"
47 #include "uptodate.h"
48 #include "xattr.h"
49 #include "acl.h"
50 #include "ocfs2_trace.h"
51 #include "ioctl.h"
52 
53 #include "buffer_head_io.h"
54 
55 static int ocfs2_mknod_locked(struct ocfs2_super *osb,
56 			      struct inode *dir,
57 			      struct inode *inode,
58 			      dev_t dev,
59 			      struct buffer_head **new_fe_bh,
60 			      struct buffer_head *parent_fe_bh,
61 			      handle_t *handle,
62 			      struct ocfs2_alloc_context *inode_ac);
63 
64 static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
65 				    struct inode **ret_orphan_dir,
66 				    u64 blkno,
67 				    char *name,
68 				    struct ocfs2_dir_lookup_result *lookup,
69 				    bool dio);
70 
71 static int ocfs2_orphan_add(struct ocfs2_super *osb,
72 			    handle_t *handle,
73 			    struct inode *inode,
74 			    struct buffer_head *fe_bh,
75 			    char *name,
76 			    struct ocfs2_dir_lookup_result *lookup,
77 			    struct inode *orphan_dir_inode,
78 			    bool dio);
79 
80 static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
81 				     handle_t *handle,
82 				     struct inode *inode,
83 				     const char *symname);
84 
85 static int ocfs2_double_lock(struct ocfs2_super *osb,
86 			     struct buffer_head **bh1,
87 			     struct inode *inode1,
88 			     struct buffer_head **bh2,
89 			     struct inode *inode2,
90 			     int rename);
91 
92 static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2);
93 /* An orphan dir name is an 8 byte value, printed as a hex string */
94 #define OCFS2_ORPHAN_NAMELEN ((int)(2 * sizeof(u64)))
95 
ocfs2_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)96 static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry,
97 				   unsigned int flags)
98 {
99 	int status;
100 	u64 blkno;
101 	struct inode *inode = NULL;
102 	struct dentry *ret;
103 	struct ocfs2_inode_info *oi;
104 
105 	trace_ocfs2_lookup(dir, dentry, dentry->d_name.len,
106 			   dentry->d_name.name,
107 			   (unsigned long long)OCFS2_I(dir)->ip_blkno, 0);
108 
109 	if (dentry->d_name.len > OCFS2_MAX_FILENAME_LEN) {
110 		ret = ERR_PTR(-ENAMETOOLONG);
111 		goto bail;
112 	}
113 
114 	status = ocfs2_inode_lock_nested(dir, NULL, 0, OI_LS_PARENT);
115 	if (status < 0) {
116 		if (status != -ENOENT)
117 			mlog_errno(status);
118 		ret = ERR_PTR(status);
119 		goto bail;
120 	}
121 
122 	status = ocfs2_lookup_ino_from_name(dir, dentry->d_name.name,
123 					    dentry->d_name.len, &blkno);
124 	if (status < 0)
125 		goto bail_add;
126 
127 	inode = ocfs2_iget(OCFS2_SB(dir->i_sb), blkno, 0, 0);
128 	if (IS_ERR(inode)) {
129 		ret = ERR_PTR(-EACCES);
130 		goto bail_unlock;
131 	}
132 
133 	oi = OCFS2_I(inode);
134 	/* Clear any orphaned state... If we were able to look up the
135 	 * inode from a directory, it certainly can't be orphaned. We
136 	 * might have the bad state from a node which intended to
137 	 * orphan this inode but crashed before it could commit the
138 	 * unlink. */
139 	spin_lock(&oi->ip_lock);
140 	oi->ip_flags &= ~OCFS2_INODE_MAYBE_ORPHANED;
141 	spin_unlock(&oi->ip_lock);
142 
143 bail_add:
144 	ret = d_splice_alias(inode, dentry);
145 
146 	if (inode) {
147 		/*
148 		 * If d_splice_alias() finds a DCACHE_DISCONNECTED
149 		 * dentry, it will d_move() it on top of ourse. The
150 		 * return value will indicate this however, so in
151 		 * those cases, we switch them around for the locking
152 		 * code.
153 		 *
154 		 * NOTE: This dentry already has ->d_op set from
155 		 * ocfs2_get_parent() and ocfs2_get_dentry()
156 		 */
157 		if (!IS_ERR_OR_NULL(ret))
158 			dentry = ret;
159 
160 		status = ocfs2_dentry_attach_lock(dentry, inode,
161 						  OCFS2_I(dir)->ip_blkno);
162 		if (status) {
163 			mlog_errno(status);
164 			ret = ERR_PTR(status);
165 			goto bail_unlock;
166 		}
167 	} else
168 		ocfs2_dentry_attach_gen(dentry);
169 
170 bail_unlock:
171 	/* Don't drop the cluster lock until *after* the d_add --
172 	 * unlink on another node will message us to remove that
173 	 * dentry under this lock so otherwise we can race this with
174 	 * the downconvert thread and have a stale dentry. */
175 	ocfs2_inode_unlock(dir, 0);
176 
177 bail:
178 
179 	trace_ocfs2_lookup_ret(ret);
180 
181 	return ret;
182 }
183 
ocfs2_get_init_inode(struct inode * dir,umode_t mode)184 static struct inode *ocfs2_get_init_inode(struct inode *dir, umode_t mode)
185 {
186 	struct inode *inode;
187 	int status;
188 
189 	inode = new_inode(dir->i_sb);
190 	if (!inode) {
191 		mlog(ML_ERROR, "new_inode failed!\n");
192 		return ERR_PTR(-ENOMEM);
193 	}
194 
195 	/* populate as many fields early on as possible - many of
196 	 * these are used by the support functions here and in
197 	 * callers. */
198 	if (S_ISDIR(mode))
199 		set_nlink(inode, 2);
200 	mode = mode_strip_sgid(&nop_mnt_idmap, dir, mode);
201 	inode_init_owner(&nop_mnt_idmap, inode, dir, mode);
202 	status = dquot_initialize(inode);
203 	if (status) {
204 		iput(inode);
205 		return ERR_PTR(status);
206 	}
207 
208 	return inode;
209 }
210 
ocfs2_cleanup_add_entry_failure(struct ocfs2_super * osb,struct dentry * dentry,struct inode * inode)211 static void ocfs2_cleanup_add_entry_failure(struct ocfs2_super *osb,
212 		struct dentry *dentry, struct inode *inode)
213 {
214 	struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
215 
216 	ocfs2_simple_drop_lockres(osb, &dl->dl_lockres);
217 	ocfs2_lock_res_free(&dl->dl_lockres);
218 	BUG_ON(dl->dl_count != 1);
219 	spin_lock(&dentry_attach_lock);
220 	dentry->d_fsdata = NULL;
221 	spin_unlock(&dentry_attach_lock);
222 	kfree(dl);
223 	iput(inode);
224 }
225 
ocfs2_mknod(struct mnt_idmap * idmap,struct inode * dir,struct dentry * dentry,umode_t mode,dev_t dev)226 static int ocfs2_mknod(struct mnt_idmap *idmap,
227 		       struct inode *dir,
228 		       struct dentry *dentry,
229 		       umode_t mode,
230 		       dev_t dev)
231 {
232 	int status = 0;
233 	struct buffer_head *parent_fe_bh = NULL;
234 	handle_t *handle = NULL;
235 	struct ocfs2_super *osb;
236 	struct ocfs2_dinode *dirfe;
237 	struct ocfs2_dinode *fe = NULL;
238 	struct buffer_head *new_fe_bh = NULL;
239 	struct inode *inode = NULL;
240 	struct ocfs2_alloc_context *inode_ac = NULL;
241 	struct ocfs2_alloc_context *data_ac = NULL;
242 	struct ocfs2_alloc_context *meta_ac = NULL;
243 	int want_clusters = 0;
244 	int want_meta = 0;
245 	int xattr_credits = 0;
246 	struct ocfs2_security_xattr_info si = {
247 		.name = NULL,
248 		.enable = 1,
249 	};
250 	int did_quota_inode = 0;
251 	struct ocfs2_dir_lookup_result lookup = { NULL, };
252 	sigset_t oldset;
253 	int did_block_signals = 0;
254 	struct ocfs2_dentry_lock *dl = NULL;
255 
256 	trace_ocfs2_mknod(dir, dentry, dentry->d_name.len, dentry->d_name.name,
257 			  (unsigned long long)OCFS2_I(dir)->ip_blkno,
258 			  (unsigned long)dev, mode);
259 
260 	status = dquot_initialize(dir);
261 	if (status) {
262 		mlog_errno(status);
263 		return status;
264 	}
265 
266 	/* get our super block */
267 	osb = OCFS2_SB(dir->i_sb);
268 
269 	status = ocfs2_inode_lock(dir, &parent_fe_bh, 1);
270 	if (status < 0) {
271 		if (status != -ENOENT)
272 			mlog_errno(status);
273 		return status;
274 	}
275 
276 	if (S_ISDIR(mode) && (dir->i_nlink >= ocfs2_link_max(osb))) {
277 		status = -EMLINK;
278 		goto leave;
279 	}
280 
281 	dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
282 	if (!ocfs2_read_links_count(dirfe)) {
283 		/* can't make a file in a deleted directory. */
284 		status = -ENOENT;
285 		goto leave;
286 	}
287 
288 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
289 					   dentry->d_name.len);
290 	if (status)
291 		goto leave;
292 
293 	/* get a spot inside the dir. */
294 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
295 					      dentry->d_name.name,
296 					      dentry->d_name.len, &lookup);
297 	if (status < 0) {
298 		mlog_errno(status);
299 		goto leave;
300 	}
301 
302 	/* reserve an inode spot */
303 	status = ocfs2_reserve_new_inode(osb, &inode_ac);
304 	if (status < 0) {
305 		if (status != -ENOSPC)
306 			mlog_errno(status);
307 		goto leave;
308 	}
309 
310 	inode = ocfs2_get_init_inode(dir, mode);
311 	if (IS_ERR(inode)) {
312 		status = PTR_ERR(inode);
313 		inode = NULL;
314 		mlog_errno(status);
315 		goto leave;
316 	}
317 
318 	/* get security xattr */
319 	status = ocfs2_init_security_get(inode, dir, &dentry->d_name, &si);
320 	if (status) {
321 		if (status == -EOPNOTSUPP)
322 			si.enable = 0;
323 		else {
324 			mlog_errno(status);
325 			goto leave;
326 		}
327 	}
328 
329 	/* calculate meta data/clusters for setting security and acl xattr */
330 	status = ocfs2_calc_xattr_init(dir, parent_fe_bh, mode,
331 				       &si, &want_clusters,
332 				       &xattr_credits, &want_meta);
333 	if (status < 0) {
334 		mlog_errno(status);
335 		goto leave;
336 	}
337 
338 	/* Reserve a cluster if creating an extent based directory. */
339 	if (S_ISDIR(mode) && !ocfs2_supports_inline_data(osb)) {
340 		want_clusters += 1;
341 
342 		/* Dir indexing requires extra space as well */
343 		if (ocfs2_supports_indexed_dirs(osb))
344 			want_meta++;
345 	}
346 
347 	status = ocfs2_reserve_new_metadata_blocks(osb, want_meta, &meta_ac);
348 	if (status < 0) {
349 		if (status != -ENOSPC)
350 			mlog_errno(status);
351 		goto leave;
352 	}
353 
354 	status = ocfs2_reserve_clusters(osb, want_clusters, &data_ac);
355 	if (status < 0) {
356 		if (status != -ENOSPC)
357 			mlog_errno(status);
358 		goto leave;
359 	}
360 
361 	handle = ocfs2_start_trans(osb, ocfs2_mknod_credits(osb->sb,
362 							    S_ISDIR(mode),
363 							    xattr_credits));
364 	if (IS_ERR(handle)) {
365 		status = PTR_ERR(handle);
366 		handle = NULL;
367 		mlog_errno(status);
368 		goto leave;
369 	}
370 
371 	/* Starting to change things, restart is no longer possible. */
372 	ocfs2_block_signals(&oldset);
373 	did_block_signals = 1;
374 
375 	status = dquot_alloc_inode(inode);
376 	if (status)
377 		goto leave;
378 	did_quota_inode = 1;
379 
380 	/* do the real work now. */
381 	status = ocfs2_mknod_locked(osb, dir, inode, dev,
382 				    &new_fe_bh, parent_fe_bh, handle,
383 				    inode_ac);
384 	if (status < 0) {
385 		mlog_errno(status);
386 		goto leave;
387 	}
388 
389 	fe = (struct ocfs2_dinode *) new_fe_bh->b_data;
390 	if (S_ISDIR(mode)) {
391 		status = ocfs2_fill_new_dir(osb, handle, dir, inode,
392 					    new_fe_bh, data_ac, meta_ac);
393 		if (status < 0) {
394 			mlog_errno(status);
395 			goto leave;
396 		}
397 
398 		status = ocfs2_journal_access_di(handle, INODE_CACHE(dir),
399 						 parent_fe_bh,
400 						 OCFS2_JOURNAL_ACCESS_WRITE);
401 		if (status < 0) {
402 			mlog_errno(status);
403 			goto leave;
404 		}
405 		ocfs2_add_links_count(dirfe, 1);
406 		ocfs2_journal_dirty(handle, parent_fe_bh);
407 		inc_nlink(dir);
408 	}
409 
410 	status = ocfs2_init_acl(handle, inode, dir, new_fe_bh, parent_fe_bh,
411 			 meta_ac, data_ac);
412 
413 	if (status < 0) {
414 		mlog_errno(status);
415 		goto roll_back;
416 	}
417 
418 	if (si.enable) {
419 		status = ocfs2_init_security_set(handle, inode, new_fe_bh, &si,
420 						 meta_ac, data_ac);
421 		if (status < 0) {
422 			mlog_errno(status);
423 			goto roll_back;
424 		}
425 	}
426 
427 	/*
428 	 * Do this before adding the entry to the directory. We add
429 	 * also set d_op after success so that ->d_iput() will cleanup
430 	 * the dentry lock even if ocfs2_add_entry() fails below.
431 	 */
432 	status = ocfs2_dentry_attach_lock(dentry, inode,
433 					  OCFS2_I(dir)->ip_blkno);
434 	if (status) {
435 		mlog_errno(status);
436 		goto roll_back;
437 	}
438 
439 	dl = dentry->d_fsdata;
440 
441 	status = ocfs2_add_entry(handle, dentry, inode,
442 				 OCFS2_I(inode)->ip_blkno, parent_fe_bh,
443 				 &lookup);
444 	if (status < 0) {
445 		mlog_errno(status);
446 		goto roll_back;
447 	}
448 
449 	insert_inode_hash(inode);
450 	d_instantiate(dentry, inode);
451 	status = 0;
452 
453 roll_back:
454 	if (status < 0 && S_ISDIR(mode)) {
455 		ocfs2_add_links_count(dirfe, -1);
456 		drop_nlink(dir);
457 	}
458 
459 leave:
460 	if (status < 0 && did_quota_inode)
461 		dquot_free_inode(inode);
462 	if (handle) {
463 		if (status < 0 && fe)
464 			ocfs2_set_links_count(fe, 0);
465 		ocfs2_commit_trans(osb, handle);
466 	}
467 
468 	ocfs2_inode_unlock(dir, 1);
469 	if (did_block_signals)
470 		ocfs2_unblock_signals(&oldset);
471 
472 	brelse(new_fe_bh);
473 	brelse(parent_fe_bh);
474 	kfree(si.value);
475 
476 	ocfs2_free_dir_lookup_result(&lookup);
477 
478 	if (inode_ac)
479 		ocfs2_free_alloc_context(inode_ac);
480 
481 	if (data_ac)
482 		ocfs2_free_alloc_context(data_ac);
483 
484 	if (meta_ac)
485 		ocfs2_free_alloc_context(meta_ac);
486 
487 	/*
488 	 * We should call iput after the i_rwsem of the bitmap been
489 	 * unlocked in ocfs2_free_alloc_context, or the
490 	 * ocfs2_delete_inode will mutex_lock again.
491 	 */
492 	if ((status < 0) && inode) {
493 		if (dl)
494 			ocfs2_cleanup_add_entry_failure(osb, dentry, inode);
495 
496 		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR;
497 		clear_nlink(inode);
498 		iput(inode);
499 	}
500 
501 	if (status)
502 		mlog_errno(status);
503 
504 	return status;
505 }
506 
__ocfs2_mknod_locked(struct inode * dir,struct inode * inode,dev_t dev,struct buffer_head ** new_fe_bh,struct buffer_head * parent_fe_bh,handle_t * handle,struct ocfs2_alloc_context * inode_ac,u64 fe_blkno,u64 suballoc_loc,u16 suballoc_bit)507 static int __ocfs2_mknod_locked(struct inode *dir,
508 				struct inode *inode,
509 				dev_t dev,
510 				struct buffer_head **new_fe_bh,
511 				struct buffer_head *parent_fe_bh,
512 				handle_t *handle,
513 				struct ocfs2_alloc_context *inode_ac,
514 				u64 fe_blkno, u64 suballoc_loc, u16 suballoc_bit)
515 {
516 	int status = 0;
517 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
518 	struct ocfs2_dinode *fe = NULL;
519 	struct ocfs2_extent_list *fel;
520 	u16 feat;
521 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
522 	struct timespec64 ts;
523 
524 	*new_fe_bh = NULL;
525 
526 	/* populate as many fields early on as possible - many of
527 	 * these are used by the support functions here and in
528 	 * callers. */
529 	inode->i_ino = ino_from_blkno(osb->sb, fe_blkno);
530 	oi->ip_blkno = fe_blkno;
531 	spin_lock(&osb->osb_lock);
532 	inode->i_generation = osb->s_next_generation++;
533 	spin_unlock(&osb->osb_lock);
534 
535 	*new_fe_bh = sb_getblk(osb->sb, fe_blkno);
536 	if (!*new_fe_bh) {
537 		status = -ENOMEM;
538 		mlog_errno(status);
539 		goto leave;
540 	}
541 	ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), *new_fe_bh);
542 
543 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
544 					 *new_fe_bh,
545 					 OCFS2_JOURNAL_ACCESS_CREATE);
546 	if (status < 0) {
547 		mlog_errno(status);
548 		goto leave;
549 	}
550 
551 	fe = (struct ocfs2_dinode *) (*new_fe_bh)->b_data;
552 	memset(fe, 0, osb->sb->s_blocksize);
553 
554 	fe->i_generation = cpu_to_le32(inode->i_generation);
555 	fe->i_fs_generation = cpu_to_le32(osb->fs_generation);
556 	fe->i_blkno = cpu_to_le64(fe_blkno);
557 	fe->i_suballoc_loc = cpu_to_le64(suballoc_loc);
558 	fe->i_suballoc_bit = cpu_to_le16(suballoc_bit);
559 	fe->i_suballoc_slot = cpu_to_le16(inode_ac->ac_alloc_slot);
560 	fe->i_uid = cpu_to_le32(i_uid_read(inode));
561 	fe->i_gid = cpu_to_le32(i_gid_read(inode));
562 	fe->i_mode = cpu_to_le16(inode->i_mode);
563 	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
564 		fe->id1.dev1.i_rdev = cpu_to_le64(huge_encode_dev(dev));
565 
566 	ocfs2_set_links_count(fe, inode->i_nlink);
567 
568 	fe->i_last_eb_blk = 0;
569 	strcpy(fe->i_signature, OCFS2_INODE_SIGNATURE);
570 	fe->i_flags |= cpu_to_le32(OCFS2_VALID_FL);
571 	ktime_get_coarse_real_ts64(&ts);
572 	fe->i_atime = fe->i_ctime = fe->i_mtime =
573 		cpu_to_le64(ts.tv_sec);
574 	fe->i_mtime_nsec = fe->i_ctime_nsec = fe->i_atime_nsec =
575 		cpu_to_le32(ts.tv_nsec);
576 	fe->i_dtime = 0;
577 
578 	/*
579 	 * If supported, directories start with inline data. If inline
580 	 * isn't supported, but indexing is, we start them as indexed.
581 	 */
582 	feat = le16_to_cpu(fe->i_dyn_features);
583 	if (S_ISDIR(inode->i_mode) && ocfs2_supports_inline_data(osb)) {
584 		fe->i_dyn_features = cpu_to_le16(feat | OCFS2_INLINE_DATA_FL);
585 
586 		fe->id2.i_data.id_count = cpu_to_le16(
587 				ocfs2_max_inline_data_with_xattr(osb->sb, fe));
588 	} else {
589 		fel = &fe->id2.i_list;
590 		fel->l_tree_depth = 0;
591 		fel->l_next_free_rec = 0;
592 		fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb));
593 	}
594 
595 	ocfs2_journal_dirty(handle, *new_fe_bh);
596 
597 	ocfs2_populate_inode(inode, fe, 1);
598 	ocfs2_ci_set_new(osb, INODE_CACHE(inode));
599 	if (!ocfs2_mount_local(osb)) {
600 		status = ocfs2_create_new_inode_locks(inode);
601 		if (status < 0)
602 			mlog_errno(status);
603 	}
604 
605 	ocfs2_update_inode_fsync_trans(handle, inode, 1);
606 
607 leave:
608 	if (status < 0) {
609 		if (*new_fe_bh) {
610 			brelse(*new_fe_bh);
611 			*new_fe_bh = NULL;
612 		}
613 	}
614 
615 	if (status)
616 		mlog_errno(status);
617 	return status;
618 }
619 
ocfs2_mknod_locked(struct ocfs2_super * osb,struct inode * dir,struct inode * inode,dev_t dev,struct buffer_head ** new_fe_bh,struct buffer_head * parent_fe_bh,handle_t * handle,struct ocfs2_alloc_context * inode_ac)620 static int ocfs2_mknod_locked(struct ocfs2_super *osb,
621 			      struct inode *dir,
622 			      struct inode *inode,
623 			      dev_t dev,
624 			      struct buffer_head **new_fe_bh,
625 			      struct buffer_head *parent_fe_bh,
626 			      handle_t *handle,
627 			      struct ocfs2_alloc_context *inode_ac)
628 {
629 	int status = 0;
630 	u64 suballoc_loc, fe_blkno = 0;
631 	u16 suballoc_bit;
632 
633 	*new_fe_bh = NULL;
634 
635 	status = ocfs2_claim_new_inode(handle, dir, parent_fe_bh,
636 				       inode_ac, &suballoc_loc,
637 				       &suballoc_bit, &fe_blkno);
638 	if (status < 0) {
639 		mlog_errno(status);
640 		return status;
641 	}
642 
643 	return __ocfs2_mknod_locked(dir, inode, dev, new_fe_bh,
644 				    parent_fe_bh, handle, inode_ac,
645 				    fe_blkno, suballoc_loc, suballoc_bit);
646 }
647 
ocfs2_mkdir(struct mnt_idmap * idmap,struct inode * dir,struct dentry * dentry,umode_t mode)648 static int ocfs2_mkdir(struct mnt_idmap *idmap,
649 		       struct inode *dir,
650 		       struct dentry *dentry,
651 		       umode_t mode)
652 {
653 	int ret;
654 
655 	trace_ocfs2_mkdir(dir, dentry, dentry->d_name.len, dentry->d_name.name,
656 			  OCFS2_I(dir)->ip_blkno, mode);
657 	ret = ocfs2_mknod(&nop_mnt_idmap, dir, dentry, mode | S_IFDIR, 0);
658 	if (ret)
659 		mlog_errno(ret);
660 
661 	return ret;
662 }
663 
ocfs2_create(struct mnt_idmap * idmap,struct inode * dir,struct dentry * dentry,umode_t mode,bool excl)664 static int ocfs2_create(struct mnt_idmap *idmap,
665 			struct inode *dir,
666 			struct dentry *dentry,
667 			umode_t mode,
668 			bool excl)
669 {
670 	int ret;
671 
672 	trace_ocfs2_create(dir, dentry, dentry->d_name.len, dentry->d_name.name,
673 			   (unsigned long long)OCFS2_I(dir)->ip_blkno, mode);
674 	ret = ocfs2_mknod(&nop_mnt_idmap, dir, dentry, mode | S_IFREG, 0);
675 	if (ret)
676 		mlog_errno(ret);
677 
678 	return ret;
679 }
680 
ocfs2_link(struct dentry * old_dentry,struct inode * dir,struct dentry * dentry)681 static int ocfs2_link(struct dentry *old_dentry,
682 		      struct inode *dir,
683 		      struct dentry *dentry)
684 {
685 	handle_t *handle;
686 	struct inode *inode = d_inode(old_dentry);
687 	struct inode *old_dir = d_inode(old_dentry->d_parent);
688 	int err;
689 	struct buffer_head *fe_bh = NULL;
690 	struct buffer_head *old_dir_bh = NULL;
691 	struct buffer_head *parent_fe_bh = NULL;
692 	struct ocfs2_dinode *fe = NULL;
693 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
694 	struct ocfs2_dir_lookup_result lookup = { NULL, };
695 	sigset_t oldset;
696 	u64 old_de_ino;
697 
698 	trace_ocfs2_link((unsigned long long)OCFS2_I(inode)->ip_blkno,
699 			 old_dentry->d_name.len, old_dentry->d_name.name,
700 			 dentry->d_name.len, dentry->d_name.name);
701 
702 	if (S_ISDIR(inode->i_mode))
703 		return -EPERM;
704 
705 	err = dquot_initialize(dir);
706 	if (err) {
707 		mlog_errno(err);
708 		return err;
709 	}
710 
711 	err = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
712 			&parent_fe_bh, dir, 0);
713 	if (err < 0) {
714 		if (err != -ENOENT)
715 			mlog_errno(err);
716 		return err;
717 	}
718 
719 	/* make sure both dirs have bhs
720 	 * get an extra ref on old_dir_bh if old==new */
721 	if (!parent_fe_bh) {
722 		if (old_dir_bh) {
723 			parent_fe_bh = old_dir_bh;
724 			get_bh(parent_fe_bh);
725 		} else {
726 			mlog(ML_ERROR, "%s: no old_dir_bh!\n", osb->uuid_str);
727 			err = -EIO;
728 			goto out;
729 		}
730 	}
731 
732 	if (!dir->i_nlink) {
733 		err = -ENOENT;
734 		goto out;
735 	}
736 
737 	err = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name,
738 			old_dentry->d_name.len, &old_de_ino);
739 	if (err) {
740 		err = -ENOENT;
741 		goto out;
742 	}
743 
744 	/*
745 	 * Check whether another node removed the source inode while we
746 	 * were in the vfs.
747 	 */
748 	if (old_de_ino != OCFS2_I(inode)->ip_blkno) {
749 		err = -ENOENT;
750 		goto out;
751 	}
752 
753 	err = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
754 					dentry->d_name.len);
755 	if (err)
756 		goto out;
757 
758 	err = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
759 					   dentry->d_name.name,
760 					   dentry->d_name.len, &lookup);
761 	if (err < 0) {
762 		mlog_errno(err);
763 		goto out;
764 	}
765 
766 	err = ocfs2_inode_lock(inode, &fe_bh, 1);
767 	if (err < 0) {
768 		if (err != -ENOENT)
769 			mlog_errno(err);
770 		goto out;
771 	}
772 
773 	fe = (struct ocfs2_dinode *) fe_bh->b_data;
774 	if (ocfs2_read_links_count(fe) >= ocfs2_link_max(osb)) {
775 		err = -EMLINK;
776 		goto out_unlock_inode;
777 	}
778 
779 	handle = ocfs2_start_trans(osb, ocfs2_link_credits(osb->sb));
780 	if (IS_ERR(handle)) {
781 		err = PTR_ERR(handle);
782 		handle = NULL;
783 		mlog_errno(err);
784 		goto out_unlock_inode;
785 	}
786 
787 	/* Starting to change things, restart is no longer possible. */
788 	ocfs2_block_signals(&oldset);
789 
790 	err = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
791 				      OCFS2_JOURNAL_ACCESS_WRITE);
792 	if (err < 0) {
793 		mlog_errno(err);
794 		goto out_commit;
795 	}
796 
797 	inc_nlink(inode);
798 	inode_set_ctime_current(inode);
799 	ocfs2_set_links_count(fe, inode->i_nlink);
800 	fe->i_ctime = cpu_to_le64(inode_get_ctime_sec(inode));
801 	fe->i_ctime_nsec = cpu_to_le32(inode_get_ctime_nsec(inode));
802 	ocfs2_update_inode_fsync_trans(handle, inode, 0);
803 	ocfs2_journal_dirty(handle, fe_bh);
804 
805 	err = ocfs2_add_entry(handle, dentry, inode,
806 			      OCFS2_I(inode)->ip_blkno,
807 			      parent_fe_bh, &lookup);
808 	if (err) {
809 		ocfs2_add_links_count(fe, -1);
810 		drop_nlink(inode);
811 		mlog_errno(err);
812 		goto out_commit;
813 	}
814 
815 	err = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
816 	if (err) {
817 		mlog_errno(err);
818 		goto out_commit;
819 	}
820 
821 	ihold(inode);
822 	d_instantiate(dentry, inode);
823 
824 out_commit:
825 	ocfs2_commit_trans(osb, handle);
826 	ocfs2_unblock_signals(&oldset);
827 out_unlock_inode:
828 	ocfs2_inode_unlock(inode, 1);
829 
830 out:
831 	ocfs2_double_unlock(old_dir, dir);
832 
833 	brelse(fe_bh);
834 	brelse(parent_fe_bh);
835 	brelse(old_dir_bh);
836 
837 	ocfs2_free_dir_lookup_result(&lookup);
838 
839 	if (err)
840 		mlog_errno(err);
841 
842 	return err;
843 }
844 
845 /*
846  * Takes and drops an exclusive lock on the given dentry. This will
847  * force other nodes to drop it.
848  */
ocfs2_remote_dentry_delete(struct dentry * dentry)849 static int ocfs2_remote_dentry_delete(struct dentry *dentry)
850 {
851 	int ret;
852 
853 	ret = ocfs2_dentry_lock(dentry, 1);
854 	if (ret)
855 		mlog_errno(ret);
856 	else
857 		ocfs2_dentry_unlock(dentry, 1);
858 
859 	return ret;
860 }
861 
ocfs2_inode_is_unlinkable(struct inode * inode)862 static inline int ocfs2_inode_is_unlinkable(struct inode *inode)
863 {
864 	if (S_ISDIR(inode->i_mode)) {
865 		if (inode->i_nlink == 2)
866 			return 1;
867 		return 0;
868 	}
869 
870 	if (inode->i_nlink == 1)
871 		return 1;
872 	return 0;
873 }
874 
ocfs2_unlink(struct inode * dir,struct dentry * dentry)875 static int ocfs2_unlink(struct inode *dir,
876 			struct dentry *dentry)
877 {
878 	int status;
879 	int child_locked = 0;
880 	bool is_unlinkable = false;
881 	struct inode *inode = d_inode(dentry);
882 	struct inode *orphan_dir = NULL;
883 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
884 	u64 blkno;
885 	struct ocfs2_dinode *fe = NULL;
886 	struct buffer_head *fe_bh = NULL;
887 	struct buffer_head *parent_node_bh = NULL;
888 	handle_t *handle = NULL;
889 	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
890 	struct ocfs2_dir_lookup_result lookup = { NULL, };
891 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
892 
893 	trace_ocfs2_unlink(dir, dentry, dentry->d_name.len,
894 			   dentry->d_name.name,
895 			   (unsigned long long)OCFS2_I(dir)->ip_blkno,
896 			   (unsigned long long)OCFS2_I(inode)->ip_blkno);
897 
898 	status = dquot_initialize(dir);
899 	if (status) {
900 		mlog_errno(status);
901 		return status;
902 	}
903 
904 	BUG_ON(d_inode(dentry->d_parent) != dir);
905 
906 	if (inode == osb->root_inode)
907 		return -EPERM;
908 
909 	status = ocfs2_inode_lock_nested(dir, &parent_node_bh, 1,
910 					 OI_LS_PARENT);
911 	if (status < 0) {
912 		if (status != -ENOENT)
913 			mlog_errno(status);
914 		return status;
915 	}
916 
917 	status = ocfs2_find_files_on_disk(dentry->d_name.name,
918 					  dentry->d_name.len, &blkno, dir,
919 					  &lookup);
920 	if (status < 0) {
921 		if (status != -ENOENT)
922 			mlog_errno(status);
923 		goto leave;
924 	}
925 
926 	if (OCFS2_I(inode)->ip_blkno != blkno) {
927 		status = -ENOENT;
928 
929 		trace_ocfs2_unlink_noent(
930 				(unsigned long long)OCFS2_I(inode)->ip_blkno,
931 				(unsigned long long)blkno,
932 				OCFS2_I(inode)->ip_flags);
933 		goto leave;
934 	}
935 
936 	status = ocfs2_inode_lock(inode, &fe_bh, 1);
937 	if (status < 0) {
938 		if (status != -ENOENT)
939 			mlog_errno(status);
940 		goto leave;
941 	}
942 	child_locked = 1;
943 
944 	if (S_ISDIR(inode->i_mode)) {
945 		if (inode->i_nlink != 2 || !ocfs2_empty_dir(inode)) {
946 			status = -ENOTEMPTY;
947 			goto leave;
948 		}
949 	}
950 
951 	status = ocfs2_remote_dentry_delete(dentry);
952 	if (status < 0) {
953 		/* This remote delete should succeed under all normal
954 		 * circumstances. */
955 		mlog_errno(status);
956 		goto leave;
957 	}
958 
959 	if (ocfs2_inode_is_unlinkable(inode)) {
960 		status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
961 						  OCFS2_I(inode)->ip_blkno,
962 						  orphan_name, &orphan_insert,
963 						  false);
964 		if (status < 0) {
965 			mlog_errno(status);
966 			goto leave;
967 		}
968 		is_unlinkable = true;
969 	}
970 
971 	handle = ocfs2_start_trans(osb, ocfs2_unlink_credits(osb->sb));
972 	if (IS_ERR(handle)) {
973 		status = PTR_ERR(handle);
974 		handle = NULL;
975 		mlog_errno(status);
976 		goto leave;
977 	}
978 
979 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
980 					 OCFS2_JOURNAL_ACCESS_WRITE);
981 	if (status < 0) {
982 		mlog_errno(status);
983 		goto leave;
984 	}
985 
986 	fe = (struct ocfs2_dinode *) fe_bh->b_data;
987 
988 	/* delete the name from the parent dir */
989 	status = ocfs2_delete_entry(handle, dir, &lookup);
990 	if (status < 0) {
991 		mlog_errno(status);
992 		goto leave;
993 	}
994 
995 	if (S_ISDIR(inode->i_mode))
996 		drop_nlink(inode);
997 	drop_nlink(inode);
998 	ocfs2_set_links_count(fe, inode->i_nlink);
999 	ocfs2_update_inode_fsync_trans(handle, inode, 0);
1000 	ocfs2_journal_dirty(handle, fe_bh);
1001 
1002 	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
1003 	if (S_ISDIR(inode->i_mode))
1004 		drop_nlink(dir);
1005 
1006 	status = ocfs2_mark_inode_dirty(handle, dir, parent_node_bh);
1007 	if (status < 0) {
1008 		mlog_errno(status);
1009 		if (S_ISDIR(inode->i_mode))
1010 			inc_nlink(dir);
1011 		goto leave;
1012 	}
1013 
1014 	if (is_unlinkable) {
1015 		status = ocfs2_orphan_add(osb, handle, inode, fe_bh,
1016 				orphan_name, &orphan_insert, orphan_dir, false);
1017 		if (status < 0)
1018 			mlog_errno(status);
1019 	}
1020 
1021 leave:
1022 	if (handle)
1023 		ocfs2_commit_trans(osb, handle);
1024 
1025 	if (orphan_dir) {
1026 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
1027 		ocfs2_inode_unlock(orphan_dir, 1);
1028 		inode_unlock(orphan_dir);
1029 		iput(orphan_dir);
1030 	}
1031 
1032 	if (child_locked)
1033 		ocfs2_inode_unlock(inode, 1);
1034 
1035 	ocfs2_inode_unlock(dir, 1);
1036 
1037 	brelse(fe_bh);
1038 	brelse(parent_node_bh);
1039 
1040 	ocfs2_free_dir_lookup_result(&orphan_insert);
1041 	ocfs2_free_dir_lookup_result(&lookup);
1042 
1043 	if (status && (status != -ENOTEMPTY) && (status != -ENOENT))
1044 		mlog_errno(status);
1045 
1046 	return status;
1047 }
1048 
ocfs2_check_if_ancestor(struct ocfs2_super * osb,u64 src_inode_no,u64 dest_inode_no)1049 static int ocfs2_check_if_ancestor(struct ocfs2_super *osb,
1050 		u64 src_inode_no, u64 dest_inode_no)
1051 {
1052 	int ret = 0, i = 0;
1053 	u64 parent_inode_no = 0;
1054 	u64 child_inode_no = src_inode_no;
1055 	struct inode *child_inode;
1056 
1057 #define MAX_LOOKUP_TIMES 32
1058 	while (1) {
1059 		child_inode = ocfs2_iget(osb, child_inode_no, 0, 0);
1060 		if (IS_ERR(child_inode)) {
1061 			ret = PTR_ERR(child_inode);
1062 			break;
1063 		}
1064 
1065 		ret = ocfs2_inode_lock(child_inode, NULL, 0);
1066 		if (ret < 0) {
1067 			iput(child_inode);
1068 			if (ret != -ENOENT)
1069 				mlog_errno(ret);
1070 			break;
1071 		}
1072 
1073 		ret = ocfs2_lookup_ino_from_name(child_inode, "..", 2,
1074 				&parent_inode_no);
1075 		ocfs2_inode_unlock(child_inode, 0);
1076 		iput(child_inode);
1077 		if (ret < 0) {
1078 			ret = -ENOENT;
1079 			break;
1080 		}
1081 
1082 		if (parent_inode_no == dest_inode_no) {
1083 			ret = 1;
1084 			break;
1085 		}
1086 
1087 		if (parent_inode_no == osb->root_inode->i_ino) {
1088 			ret = 0;
1089 			break;
1090 		}
1091 
1092 		child_inode_no = parent_inode_no;
1093 
1094 		if (++i >= MAX_LOOKUP_TIMES) {
1095 			mlog_ratelimited(ML_NOTICE, "max lookup times reached, "
1096 					"filesystem may have nested directories, "
1097 					"src inode: %llu, dest inode: %llu.\n",
1098 					(unsigned long long)src_inode_no,
1099 					(unsigned long long)dest_inode_no);
1100 			ret = 0;
1101 			break;
1102 		}
1103 	}
1104 
1105 	return ret;
1106 }
1107 
1108 /*
1109  * The only place this should be used is rename and link!
1110  * if they have the same id, then the 1st one is the only one locked.
1111  */
ocfs2_double_lock(struct ocfs2_super * osb,struct buffer_head ** bh1,struct inode * inode1,struct buffer_head ** bh2,struct inode * inode2,int rename)1112 static int ocfs2_double_lock(struct ocfs2_super *osb,
1113 			     struct buffer_head **bh1,
1114 			     struct inode *inode1,
1115 			     struct buffer_head **bh2,
1116 			     struct inode *inode2,
1117 			     int rename)
1118 {
1119 	int status;
1120 	int inode1_is_ancestor, inode2_is_ancestor;
1121 	struct ocfs2_inode_info *oi1 = OCFS2_I(inode1);
1122 	struct ocfs2_inode_info *oi2 = OCFS2_I(inode2);
1123 
1124 	trace_ocfs2_double_lock((unsigned long long)oi1->ip_blkno,
1125 				(unsigned long long)oi2->ip_blkno);
1126 
1127 	if (*bh1)
1128 		*bh1 = NULL;
1129 	if (*bh2)
1130 		*bh2 = NULL;
1131 
1132 	/* we always want to lock the one with the lower lockid first.
1133 	 * and if they are nested, we lock ancestor first */
1134 	if (oi1->ip_blkno != oi2->ip_blkno) {
1135 		inode1_is_ancestor = ocfs2_check_if_ancestor(osb, oi2->ip_blkno,
1136 				oi1->ip_blkno);
1137 		if (inode1_is_ancestor < 0) {
1138 			status = inode1_is_ancestor;
1139 			goto bail;
1140 		}
1141 
1142 		inode2_is_ancestor = ocfs2_check_if_ancestor(osb, oi1->ip_blkno,
1143 				oi2->ip_blkno);
1144 		if (inode2_is_ancestor < 0) {
1145 			status = inode2_is_ancestor;
1146 			goto bail;
1147 		}
1148 
1149 		if ((inode1_is_ancestor == 1) ||
1150 				(oi1->ip_blkno < oi2->ip_blkno &&
1151 				inode2_is_ancestor == 0)) {
1152 			/* switch id1 and id2 around */
1153 			swap(bh2, bh1);
1154 			swap(inode2, inode1);
1155 		}
1156 		/* lock id2 */
1157 		status = ocfs2_inode_lock_nested(inode2, bh2, 1,
1158 				rename == 1 ? OI_LS_RENAME1 : OI_LS_PARENT);
1159 		if (status < 0) {
1160 			if (status != -ENOENT)
1161 				mlog_errno(status);
1162 			goto bail;
1163 		}
1164 	}
1165 
1166 	/* lock id1 */
1167 	status = ocfs2_inode_lock_nested(inode1, bh1, 1,
1168 			rename == 1 ?  OI_LS_RENAME2 : OI_LS_PARENT);
1169 	if (status < 0) {
1170 		/*
1171 		 * An error return must mean that no cluster locks
1172 		 * were held on function exit.
1173 		 */
1174 		if (oi1->ip_blkno != oi2->ip_blkno) {
1175 			ocfs2_inode_unlock(inode2, 1);
1176 			brelse(*bh2);
1177 			*bh2 = NULL;
1178 		}
1179 
1180 		if (status != -ENOENT)
1181 			mlog_errno(status);
1182 	}
1183 
1184 	trace_ocfs2_double_lock_end(
1185 			(unsigned long long)oi1->ip_blkno,
1186 			(unsigned long long)oi2->ip_blkno);
1187 
1188 bail:
1189 	if (status)
1190 		mlog_errno(status);
1191 	return status;
1192 }
1193 
ocfs2_double_unlock(struct inode * inode1,struct inode * inode2)1194 static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2)
1195 {
1196 	ocfs2_inode_unlock(inode1, 1);
1197 
1198 	if (inode1 != inode2)
1199 		ocfs2_inode_unlock(inode2, 1);
1200 }
1201 
ocfs2_rename(struct mnt_idmap * idmap,struct inode * old_dir,struct dentry * old_dentry,struct inode * new_dir,struct dentry * new_dentry,unsigned int flags)1202 static int ocfs2_rename(struct mnt_idmap *idmap,
1203 			struct inode *old_dir,
1204 			struct dentry *old_dentry,
1205 			struct inode *new_dir,
1206 			struct dentry *new_dentry,
1207 			unsigned int flags)
1208 {
1209 	int status = 0, rename_lock = 0, parents_locked = 0, target_exists = 0;
1210 	int old_child_locked = 0, new_child_locked = 0, update_dot_dot = 0;
1211 	struct inode *old_inode = d_inode(old_dentry);
1212 	struct inode *new_inode = d_inode(new_dentry);
1213 	struct inode *orphan_dir = NULL;
1214 	struct ocfs2_dinode *newfe = NULL;
1215 	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
1216 	struct buffer_head *newfe_bh = NULL;
1217 	struct buffer_head *old_inode_bh = NULL;
1218 	struct ocfs2_super *osb = NULL;
1219 	u64 newfe_blkno, old_de_ino;
1220 	handle_t *handle = NULL;
1221 	struct buffer_head *old_dir_bh = NULL;
1222 	struct buffer_head *new_dir_bh = NULL;
1223 	u32 old_dir_nlink = old_dir->i_nlink;
1224 	struct ocfs2_dinode *old_di;
1225 	struct ocfs2_dir_lookup_result old_inode_dot_dot_res = { NULL, };
1226 	struct ocfs2_dir_lookup_result target_lookup_res = { NULL, };
1227 	struct ocfs2_dir_lookup_result old_entry_lookup = { NULL, };
1228 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
1229 	struct ocfs2_dir_lookup_result target_insert = { NULL, };
1230 	bool should_add_orphan = false;
1231 
1232 	if (flags)
1233 		return -EINVAL;
1234 
1235 	/* At some point it might be nice to break this function up a
1236 	 * bit. */
1237 
1238 	trace_ocfs2_rename(old_dir, old_dentry, new_dir, new_dentry,
1239 			   old_dentry->d_name.len, old_dentry->d_name.name,
1240 			   new_dentry->d_name.len, new_dentry->d_name.name);
1241 
1242 	status = dquot_initialize(old_dir);
1243 	if (status) {
1244 		mlog_errno(status);
1245 		goto bail;
1246 	}
1247 	status = dquot_initialize(new_dir);
1248 	if (status) {
1249 		mlog_errno(status);
1250 		goto bail;
1251 	}
1252 
1253 	osb = OCFS2_SB(old_dir->i_sb);
1254 
1255 	if (new_inode) {
1256 		if (!igrab(new_inode))
1257 			BUG();
1258 	}
1259 
1260 	/* Assume a directory hierarchy thusly:
1261 	 * a/b/c
1262 	 * a/d
1263 	 * a,b,c, and d are all directories.
1264 	 *
1265 	 * from cwd of 'a' on both nodes:
1266 	 * node1: mv b/c d
1267 	 * node2: mv d   b/c
1268 	 *
1269 	 * And that's why, just like the VFS, we need a file system
1270 	 * rename lock. */
1271 	if (old_dir != new_dir && S_ISDIR(old_inode->i_mode)) {
1272 		status = ocfs2_rename_lock(osb);
1273 		if (status < 0) {
1274 			mlog_errno(status);
1275 			goto bail;
1276 		}
1277 		rename_lock = 1;
1278 
1279 		/* here we cannot guarantee the inodes haven't just been
1280 		 * changed, so check if they are nested again */
1281 		status = ocfs2_check_if_ancestor(osb, new_dir->i_ino,
1282 				old_inode->i_ino);
1283 		if (status < 0) {
1284 			mlog_errno(status);
1285 			goto bail;
1286 		} else if (status == 1) {
1287 			status = -EPERM;
1288 			trace_ocfs2_rename_not_permitted(
1289 					(unsigned long long)old_inode->i_ino,
1290 					(unsigned long long)new_dir->i_ino);
1291 			goto bail;
1292 		}
1293 	}
1294 
1295 	/* if old and new are the same, this'll just do one lock. */
1296 	status = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
1297 				   &new_dir_bh, new_dir, 1);
1298 	if (status < 0) {
1299 		mlog_errno(status);
1300 		goto bail;
1301 	}
1302 	parents_locked = 1;
1303 
1304 	if (!new_dir->i_nlink) {
1305 		status = -EACCES;
1306 		goto bail;
1307 	}
1308 
1309 	/* make sure both dirs have bhs
1310 	 * get an extra ref on old_dir_bh if old==new */
1311 	if (!new_dir_bh) {
1312 		if (old_dir_bh) {
1313 			new_dir_bh = old_dir_bh;
1314 			get_bh(new_dir_bh);
1315 		} else {
1316 			mlog(ML_ERROR, "no old_dir_bh!\n");
1317 			status = -EIO;
1318 			goto bail;
1319 		}
1320 	}
1321 
1322 	/*
1323 	 * Aside from allowing a meta data update, the locking here
1324 	 * also ensures that the downconvert thread on other nodes
1325 	 * won't have to concurrently downconvert the inode and the
1326 	 * dentry locks.
1327 	 */
1328 	status = ocfs2_inode_lock_nested(old_inode, &old_inode_bh, 1,
1329 					 OI_LS_PARENT);
1330 	if (status < 0) {
1331 		if (status != -ENOENT)
1332 			mlog_errno(status);
1333 		goto bail;
1334 	}
1335 	old_child_locked = 1;
1336 
1337 	status = ocfs2_remote_dentry_delete(old_dentry);
1338 	if (status < 0) {
1339 		mlog_errno(status);
1340 		goto bail;
1341 	}
1342 
1343 	if (S_ISDIR(old_inode->i_mode)) {
1344 		u64 old_inode_parent;
1345 
1346 		update_dot_dot = 1;
1347 		status = ocfs2_find_files_on_disk("..", 2, &old_inode_parent,
1348 						  old_inode,
1349 						  &old_inode_dot_dot_res);
1350 		if (status) {
1351 			status = -EIO;
1352 			goto bail;
1353 		}
1354 
1355 		if (old_inode_parent != OCFS2_I(old_dir)->ip_blkno) {
1356 			status = -EIO;
1357 			goto bail;
1358 		}
1359 
1360 		if (!new_inode && new_dir != old_dir &&
1361 		    new_dir->i_nlink >= ocfs2_link_max(osb)) {
1362 			status = -EMLINK;
1363 			goto bail;
1364 		}
1365 	}
1366 
1367 	status = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name,
1368 					    old_dentry->d_name.len,
1369 					    &old_de_ino);
1370 	if (status) {
1371 		status = -ENOENT;
1372 		goto bail;
1373 	}
1374 
1375 	/*
1376 	 *  Check for inode number is _not_ due to possible IO errors.
1377 	 *  We might rmdir the source, keep it as pwd of some process
1378 	 *  and merrily kill the link to whatever was created under the
1379 	 *  same name. Goodbye sticky bit ;-<
1380 	 */
1381 	if (old_de_ino != OCFS2_I(old_inode)->ip_blkno) {
1382 		status = -ENOENT;
1383 		goto bail;
1384 	}
1385 
1386 	/* check if the target already exists (in which case we need
1387 	 * to delete it */
1388 	status = ocfs2_find_files_on_disk(new_dentry->d_name.name,
1389 					  new_dentry->d_name.len,
1390 					  &newfe_blkno, new_dir,
1391 					  &target_lookup_res);
1392 	/* The only error we allow here is -ENOENT because the new
1393 	 * file not existing is perfectly valid. */
1394 	if ((status < 0) && (status != -ENOENT)) {
1395 		/* If we cannot find the file specified we should just */
1396 		/* return the error... */
1397 		mlog_errno(status);
1398 		goto bail;
1399 	}
1400 	if (status == 0)
1401 		target_exists = 1;
1402 
1403 	if (!target_exists && new_inode) {
1404 		/*
1405 		 * Target was unlinked by another node while we were
1406 		 * waiting to get to ocfs2_rename(). There isn't
1407 		 * anything we can do here to help the situation, so
1408 		 * bubble up the appropriate error.
1409 		 */
1410 		status = -ENOENT;
1411 		goto bail;
1412 	}
1413 
1414 	/* In case we need to overwrite an existing file, we blow it
1415 	 * away first */
1416 	if (target_exists) {
1417 		/* VFS didn't think there existed an inode here, but
1418 		 * someone else in the cluster must have raced our
1419 		 * rename to create one. Today we error cleanly, in
1420 		 * the future we should consider calling iget to build
1421 		 * a new struct inode for this entry. */
1422 		if (!new_inode) {
1423 			status = -EACCES;
1424 
1425 			trace_ocfs2_rename_target_exists(new_dentry->d_name.len,
1426 						new_dentry->d_name.name);
1427 			goto bail;
1428 		}
1429 
1430 		if (OCFS2_I(new_inode)->ip_blkno != newfe_blkno) {
1431 			status = -EACCES;
1432 
1433 			trace_ocfs2_rename_disagree(
1434 			     (unsigned long long)OCFS2_I(new_inode)->ip_blkno,
1435 			     (unsigned long long)newfe_blkno,
1436 			     OCFS2_I(new_inode)->ip_flags);
1437 			goto bail;
1438 		}
1439 
1440 		status = ocfs2_inode_lock(new_inode, &newfe_bh, 1);
1441 		if (status < 0) {
1442 			if (status != -ENOENT)
1443 				mlog_errno(status);
1444 			goto bail;
1445 		}
1446 		new_child_locked = 1;
1447 
1448 		status = ocfs2_remote_dentry_delete(new_dentry);
1449 		if (status < 0) {
1450 			mlog_errno(status);
1451 			goto bail;
1452 		}
1453 
1454 		newfe = (struct ocfs2_dinode *) newfe_bh->b_data;
1455 
1456 		trace_ocfs2_rename_over_existing(
1457 		     (unsigned long long)newfe_blkno, newfe_bh, newfe_bh ?
1458 		     (unsigned long long)newfe_bh->b_blocknr : 0ULL);
1459 
1460 		if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) {
1461 			status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
1462 						OCFS2_I(new_inode)->ip_blkno,
1463 						orphan_name, &orphan_insert,
1464 						false);
1465 			if (status < 0) {
1466 				mlog_errno(status);
1467 				goto bail;
1468 			}
1469 			should_add_orphan = true;
1470 		}
1471 	} else {
1472 		BUG_ON(d_inode(new_dentry->d_parent) != new_dir);
1473 
1474 		status = ocfs2_check_dir_for_entry(new_dir,
1475 						   new_dentry->d_name.name,
1476 						   new_dentry->d_name.len);
1477 		if (status)
1478 			goto bail;
1479 
1480 		status = ocfs2_prepare_dir_for_insert(osb, new_dir, new_dir_bh,
1481 						      new_dentry->d_name.name,
1482 						      new_dentry->d_name.len,
1483 						      &target_insert);
1484 		if (status < 0) {
1485 			mlog_errno(status);
1486 			goto bail;
1487 		}
1488 	}
1489 
1490 	handle = ocfs2_start_trans(osb, ocfs2_rename_credits(osb->sb));
1491 	if (IS_ERR(handle)) {
1492 		status = PTR_ERR(handle);
1493 		handle = NULL;
1494 		mlog_errno(status);
1495 		goto bail;
1496 	}
1497 
1498 	if (target_exists) {
1499 		if (S_ISDIR(new_inode->i_mode)) {
1500 			if (new_inode->i_nlink != 2 ||
1501 			    !ocfs2_empty_dir(new_inode)) {
1502 				status = -ENOTEMPTY;
1503 				goto bail;
1504 			}
1505 		}
1506 		status = ocfs2_journal_access_di(handle, INODE_CACHE(new_inode),
1507 						 newfe_bh,
1508 						 OCFS2_JOURNAL_ACCESS_WRITE);
1509 		if (status < 0) {
1510 			mlog_errno(status);
1511 			goto bail;
1512 		}
1513 
1514 		/* change the dirent to point to the correct inode */
1515 		status = ocfs2_update_entry(new_dir, handle, &target_lookup_res,
1516 					    old_inode);
1517 		if (status < 0) {
1518 			mlog_errno(status);
1519 			goto bail;
1520 		}
1521 		inode_inc_iversion(new_dir);
1522 
1523 		if (S_ISDIR(new_inode->i_mode))
1524 			ocfs2_set_links_count(newfe, 0);
1525 		else
1526 			ocfs2_add_links_count(newfe, -1);
1527 		ocfs2_journal_dirty(handle, newfe_bh);
1528 		if (should_add_orphan) {
1529 			status = ocfs2_orphan_add(osb, handle, new_inode,
1530 					newfe_bh, orphan_name,
1531 					&orphan_insert, orphan_dir, false);
1532 			if (status < 0) {
1533 				mlog_errno(status);
1534 				goto bail;
1535 			}
1536 		}
1537 	} else {
1538 		/* if the name was not found in new_dir, add it now */
1539 		status = ocfs2_add_entry(handle, new_dentry, old_inode,
1540 					 OCFS2_I(old_inode)->ip_blkno,
1541 					 new_dir_bh, &target_insert);
1542 		if (status < 0) {
1543 			mlog_errno(status);
1544 			goto bail;
1545 		}
1546 	}
1547 
1548 	inode_set_ctime_current(old_inode);
1549 	mark_inode_dirty(old_inode);
1550 
1551 	status = ocfs2_journal_access_di(handle, INODE_CACHE(old_inode),
1552 					 old_inode_bh,
1553 					 OCFS2_JOURNAL_ACCESS_WRITE);
1554 	if (status >= 0) {
1555 		old_di = (struct ocfs2_dinode *) old_inode_bh->b_data;
1556 
1557 		old_di->i_ctime = cpu_to_le64(inode_get_ctime_sec(old_inode));
1558 		old_di->i_ctime_nsec = cpu_to_le32(inode_get_ctime_nsec(old_inode));
1559 		ocfs2_journal_dirty(handle, old_inode_bh);
1560 	} else
1561 		mlog_errno(status);
1562 
1563 	/*
1564 	 * Now that the name has been added to new_dir, remove the old name.
1565 	 *
1566 	 * We don't keep any directory entry context around until now
1567 	 * because the insert might have changed the type of directory
1568 	 * we're dealing with.
1569 	 */
1570 	status = ocfs2_find_entry(old_dentry->d_name.name,
1571 				  old_dentry->d_name.len, old_dir,
1572 				  &old_entry_lookup);
1573 	if (status) {
1574 		if (!is_journal_aborted(osb->journal->j_journal)) {
1575 			ocfs2_error(osb->sb, "new entry %.*s is added, but old entry %.*s "
1576 					"is not deleted.",
1577 					new_dentry->d_name.len, new_dentry->d_name.name,
1578 					old_dentry->d_name.len, old_dentry->d_name.name);
1579 		}
1580 		goto bail;
1581 	}
1582 
1583 	status = ocfs2_delete_entry(handle, old_dir, &old_entry_lookup);
1584 	if (status < 0) {
1585 		mlog_errno(status);
1586 		if (!is_journal_aborted(osb->journal->j_journal)) {
1587 			ocfs2_error(osb->sb, "new entry %.*s is added, but old entry %.*s "
1588 					"is not deleted.",
1589 					new_dentry->d_name.len, new_dentry->d_name.name,
1590 					old_dentry->d_name.len, old_dentry->d_name.name);
1591 		}
1592 		goto bail;
1593 	}
1594 
1595 	if (new_inode) {
1596 		drop_nlink(new_inode);
1597 		inode_set_ctime_current(new_inode);
1598 	}
1599 	inode_set_mtime_to_ts(old_dir, inode_set_ctime_current(old_dir));
1600 
1601 	if (update_dot_dot) {
1602 		status = ocfs2_update_entry(old_inode, handle,
1603 					    &old_inode_dot_dot_res, new_dir);
1604 		drop_nlink(old_dir);
1605 		if (new_inode) {
1606 			drop_nlink(new_inode);
1607 		} else {
1608 			inc_nlink(new_dir);
1609 			mark_inode_dirty(new_dir);
1610 		}
1611 	}
1612 	mark_inode_dirty(old_dir);
1613 	ocfs2_mark_inode_dirty(handle, old_dir, old_dir_bh);
1614 	if (new_inode) {
1615 		mark_inode_dirty(new_inode);
1616 		ocfs2_mark_inode_dirty(handle, new_inode, newfe_bh);
1617 	}
1618 
1619 	if (old_dir != new_dir) {
1620 		/* Keep the same times on both directories.*/
1621 		inode_set_mtime_to_ts(new_dir,
1622 				      inode_set_ctime_to_ts(new_dir, inode_get_ctime(old_dir)));
1623 
1624 		/*
1625 		 * This will also pick up the i_nlink change from the
1626 		 * block above.
1627 		 */
1628 		ocfs2_mark_inode_dirty(handle, new_dir, new_dir_bh);
1629 	}
1630 
1631 	if (old_dir_nlink != old_dir->i_nlink) {
1632 		if (!old_dir_bh) {
1633 			mlog(ML_ERROR, "need to change nlink for old dir "
1634 			     "%llu from %d to %d but bh is NULL!\n",
1635 			     (unsigned long long)OCFS2_I(old_dir)->ip_blkno,
1636 			     (int)old_dir_nlink, old_dir->i_nlink);
1637 		} else {
1638 			struct ocfs2_dinode *fe;
1639 			status = ocfs2_journal_access_di(handle,
1640 							 INODE_CACHE(old_dir),
1641 							 old_dir_bh,
1642 							 OCFS2_JOURNAL_ACCESS_WRITE);
1643 			fe = (struct ocfs2_dinode *) old_dir_bh->b_data;
1644 			ocfs2_set_links_count(fe, old_dir->i_nlink);
1645 			ocfs2_journal_dirty(handle, old_dir_bh);
1646 		}
1647 	}
1648 	ocfs2_dentry_move(old_dentry, new_dentry, old_dir, new_dir);
1649 	status = 0;
1650 bail:
1651 	if (handle)
1652 		ocfs2_commit_trans(osb, handle);
1653 
1654 	if (orphan_dir) {
1655 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
1656 		ocfs2_inode_unlock(orphan_dir, 1);
1657 		inode_unlock(orphan_dir);
1658 		iput(orphan_dir);
1659 	}
1660 
1661 	if (new_child_locked)
1662 		ocfs2_inode_unlock(new_inode, 1);
1663 
1664 	if (old_child_locked)
1665 		ocfs2_inode_unlock(old_inode, 1);
1666 
1667 	if (parents_locked)
1668 		ocfs2_double_unlock(old_dir, new_dir);
1669 
1670 	if (rename_lock)
1671 		ocfs2_rename_unlock(osb);
1672 
1673 	if (new_inode)
1674 		sync_mapping_buffers(old_inode->i_mapping);
1675 
1676 	iput(new_inode);
1677 
1678 	ocfs2_free_dir_lookup_result(&target_lookup_res);
1679 	ocfs2_free_dir_lookup_result(&old_entry_lookup);
1680 	ocfs2_free_dir_lookup_result(&old_inode_dot_dot_res);
1681 	ocfs2_free_dir_lookup_result(&orphan_insert);
1682 	ocfs2_free_dir_lookup_result(&target_insert);
1683 
1684 	brelse(newfe_bh);
1685 	brelse(old_inode_bh);
1686 	brelse(old_dir_bh);
1687 	brelse(new_dir_bh);
1688 
1689 	if (status)
1690 		mlog_errno(status);
1691 
1692 	return status;
1693 }
1694 
1695 /*
1696  * we expect i_size = strlen(symname). Copy symname into the file
1697  * data, including the null terminator.
1698  */
ocfs2_create_symlink_data(struct ocfs2_super * osb,handle_t * handle,struct inode * inode,const char * symname)1699 static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
1700 				     handle_t *handle,
1701 				     struct inode *inode,
1702 				     const char *symname)
1703 {
1704 	struct buffer_head **bhs = NULL;
1705 	const char *c;
1706 	struct super_block *sb = osb->sb;
1707 	u64 p_blkno, p_blocks;
1708 	int virtual, blocks, status, i, bytes_left;
1709 
1710 	bytes_left = i_size_read(inode) + 1;
1711 	/* we can't trust i_blocks because we're actually going to
1712 	 * write i_size + 1 bytes. */
1713 	blocks = (bytes_left + sb->s_blocksize - 1) >> sb->s_blocksize_bits;
1714 
1715 	trace_ocfs2_create_symlink_data((unsigned long long)inode->i_blocks,
1716 					i_size_read(inode), blocks);
1717 
1718 	/* Sanity check -- make sure we're going to fit. */
1719 	if (bytes_left >
1720 	    ocfs2_clusters_to_bytes(sb, OCFS2_I(inode)->ip_clusters)) {
1721 		status = -EIO;
1722 		mlog_errno(status);
1723 		goto bail;
1724 	}
1725 
1726 	bhs = kcalloc(blocks, sizeof(struct buffer_head *), GFP_KERNEL);
1727 	if (!bhs) {
1728 		status = -ENOMEM;
1729 		mlog_errno(status);
1730 		goto bail;
1731 	}
1732 
1733 	status = ocfs2_extent_map_get_blocks(inode, 0, &p_blkno, &p_blocks,
1734 					     NULL);
1735 	if (status < 0) {
1736 		mlog_errno(status);
1737 		goto bail;
1738 	}
1739 
1740 	/* links can never be larger than one cluster so we know this
1741 	 * is all going to be contiguous, but do a sanity check
1742 	 * anyway. */
1743 	if ((p_blocks << sb->s_blocksize_bits) < bytes_left) {
1744 		status = -EIO;
1745 		mlog_errno(status);
1746 		goto bail;
1747 	}
1748 
1749 	virtual = 0;
1750 	while(bytes_left > 0) {
1751 		c = &symname[virtual * sb->s_blocksize];
1752 
1753 		bhs[virtual] = sb_getblk(sb, p_blkno);
1754 		if (!bhs[virtual]) {
1755 			status = -ENOMEM;
1756 			mlog_errno(status);
1757 			goto bail;
1758 		}
1759 		ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode),
1760 					      bhs[virtual]);
1761 
1762 		status = ocfs2_journal_access(handle, INODE_CACHE(inode),
1763 					      bhs[virtual],
1764 					      OCFS2_JOURNAL_ACCESS_CREATE);
1765 		if (status < 0) {
1766 			mlog_errno(status);
1767 			goto bail;
1768 		}
1769 
1770 		memset(bhs[virtual]->b_data, 0, sb->s_blocksize);
1771 
1772 		memcpy(bhs[virtual]->b_data, c,
1773 		       (bytes_left > sb->s_blocksize) ? sb->s_blocksize :
1774 		       bytes_left);
1775 
1776 		ocfs2_journal_dirty(handle, bhs[virtual]);
1777 
1778 		virtual++;
1779 		p_blkno++;
1780 		bytes_left -= sb->s_blocksize;
1781 	}
1782 
1783 	status = 0;
1784 bail:
1785 
1786 	if (bhs) {
1787 		for(i = 0; i < blocks; i++)
1788 			brelse(bhs[i]);
1789 		kfree(bhs);
1790 	}
1791 
1792 	if (status)
1793 		mlog_errno(status);
1794 	return status;
1795 }
1796 
ocfs2_symlink(struct mnt_idmap * idmap,struct inode * dir,struct dentry * dentry,const char * symname)1797 static int ocfs2_symlink(struct mnt_idmap *idmap,
1798 			 struct inode *dir,
1799 			 struct dentry *dentry,
1800 			 const char *symname)
1801 {
1802 	int status, l, credits;
1803 	u64 newsize;
1804 	struct ocfs2_super *osb = NULL;
1805 	struct inode *inode = NULL;
1806 	struct super_block *sb;
1807 	struct buffer_head *new_fe_bh = NULL;
1808 	struct buffer_head *parent_fe_bh = NULL;
1809 	struct ocfs2_dinode *fe = NULL;
1810 	struct ocfs2_dinode *dirfe;
1811 	handle_t *handle = NULL;
1812 	struct ocfs2_alloc_context *inode_ac = NULL;
1813 	struct ocfs2_alloc_context *data_ac = NULL;
1814 	struct ocfs2_alloc_context *xattr_ac = NULL;
1815 	int want_clusters = 0;
1816 	int xattr_credits = 0;
1817 	struct ocfs2_security_xattr_info si = {
1818 		.name = NULL,
1819 		.enable = 1,
1820 	};
1821 	int did_quota = 0, did_quota_inode = 0;
1822 	struct ocfs2_dir_lookup_result lookup = { NULL, };
1823 	sigset_t oldset;
1824 	int did_block_signals = 0;
1825 	struct ocfs2_dentry_lock *dl = NULL;
1826 
1827 	trace_ocfs2_symlink_begin(dir, dentry, symname,
1828 				  dentry->d_name.len, dentry->d_name.name);
1829 
1830 	status = dquot_initialize(dir);
1831 	if (status) {
1832 		mlog_errno(status);
1833 		goto bail;
1834 	}
1835 
1836 	sb = dir->i_sb;
1837 	osb = OCFS2_SB(sb);
1838 
1839 	l = strlen(symname) + 1;
1840 
1841 	credits = ocfs2_calc_symlink_credits(sb);
1842 
1843 	/* lock the parent directory */
1844 	status = ocfs2_inode_lock(dir, &parent_fe_bh, 1);
1845 	if (status < 0) {
1846 		if (status != -ENOENT)
1847 			mlog_errno(status);
1848 		return status;
1849 	}
1850 
1851 	dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
1852 	if (!ocfs2_read_links_count(dirfe)) {
1853 		/* can't make a file in a deleted directory. */
1854 		status = -ENOENT;
1855 		goto bail;
1856 	}
1857 
1858 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
1859 					   dentry->d_name.len);
1860 	if (status)
1861 		goto bail;
1862 
1863 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
1864 					      dentry->d_name.name,
1865 					      dentry->d_name.len, &lookup);
1866 	if (status < 0) {
1867 		mlog_errno(status);
1868 		goto bail;
1869 	}
1870 
1871 	status = ocfs2_reserve_new_inode(osb, &inode_ac);
1872 	if (status < 0) {
1873 		if (status != -ENOSPC)
1874 			mlog_errno(status);
1875 		goto bail;
1876 	}
1877 
1878 	inode = ocfs2_get_init_inode(dir, S_IFLNK | S_IRWXUGO);
1879 	if (IS_ERR(inode)) {
1880 		status = PTR_ERR(inode);
1881 		inode = NULL;
1882 		mlog_errno(status);
1883 		goto bail;
1884 	}
1885 
1886 	/* get security xattr */
1887 	status = ocfs2_init_security_get(inode, dir, &dentry->d_name, &si);
1888 	if (status) {
1889 		if (status == -EOPNOTSUPP)
1890 			si.enable = 0;
1891 		else {
1892 			mlog_errno(status);
1893 			goto bail;
1894 		}
1895 	}
1896 
1897 	/* calculate meta data/clusters for setting security xattr */
1898 	if (si.enable) {
1899 		status = ocfs2_calc_security_init(dir, &si, &want_clusters,
1900 						  &xattr_credits, &xattr_ac);
1901 		if (status < 0) {
1902 			mlog_errno(status);
1903 			goto bail;
1904 		}
1905 	}
1906 
1907 	/* don't reserve bitmap space for fast symlinks. */
1908 	if (l > ocfs2_fast_symlink_chars(sb))
1909 		want_clusters += 1;
1910 
1911 	status = ocfs2_reserve_clusters(osb, want_clusters, &data_ac);
1912 	if (status < 0) {
1913 		if (status != -ENOSPC)
1914 			mlog_errno(status);
1915 		goto bail;
1916 	}
1917 
1918 	handle = ocfs2_start_trans(osb, credits + xattr_credits);
1919 	if (IS_ERR(handle)) {
1920 		status = PTR_ERR(handle);
1921 		handle = NULL;
1922 		mlog_errno(status);
1923 		goto bail;
1924 	}
1925 
1926 	/* Starting to change things, restart is no longer possible. */
1927 	ocfs2_block_signals(&oldset);
1928 	did_block_signals = 1;
1929 
1930 	status = dquot_alloc_inode(inode);
1931 	if (status)
1932 		goto bail;
1933 	did_quota_inode = 1;
1934 
1935 	trace_ocfs2_symlink_create(dir, dentry, dentry->d_name.len,
1936 				   dentry->d_name.name,
1937 				   (unsigned long long)OCFS2_I(dir)->ip_blkno,
1938 				   inode->i_mode);
1939 
1940 	status = ocfs2_mknod_locked(osb, dir, inode,
1941 				    0, &new_fe_bh, parent_fe_bh, handle,
1942 				    inode_ac);
1943 	if (status < 0) {
1944 		mlog_errno(status);
1945 		goto bail;
1946 	}
1947 
1948 	fe = (struct ocfs2_dinode *) new_fe_bh->b_data;
1949 	inode->i_rdev = 0;
1950 	newsize = l - 1;
1951 	inode->i_op = &ocfs2_symlink_inode_operations;
1952 	inode_nohighmem(inode);
1953 	if (l > ocfs2_fast_symlink_chars(sb)) {
1954 		u32 offset = 0;
1955 
1956 		status = dquot_alloc_space_nodirty(inode,
1957 		    ocfs2_clusters_to_bytes(osb->sb, 1));
1958 		if (status)
1959 			goto bail;
1960 		did_quota = 1;
1961 		inode->i_mapping->a_ops = &ocfs2_aops;
1962 		status = ocfs2_add_inode_data(osb, inode, &offset, 1, 0,
1963 					      new_fe_bh,
1964 					      handle, data_ac, NULL,
1965 					      NULL);
1966 		if (status < 0) {
1967 			if (status != -ENOSPC && status != -EINTR) {
1968 				mlog(ML_ERROR,
1969 				     "Failed to extend file to %llu\n",
1970 				     (unsigned long long)newsize);
1971 				mlog_errno(status);
1972 				status = -ENOSPC;
1973 			}
1974 			goto bail;
1975 		}
1976 		i_size_write(inode, newsize);
1977 		inode->i_blocks = ocfs2_inode_sector_count(inode);
1978 	} else {
1979 		inode->i_mapping->a_ops = &ocfs2_fast_symlink_aops;
1980 		memcpy((char *) fe->id2.i_symlink, symname, l);
1981 		i_size_write(inode, newsize);
1982 		inode->i_blocks = 0;
1983 	}
1984 
1985 	status = ocfs2_mark_inode_dirty(handle, inode, new_fe_bh);
1986 	if (status < 0) {
1987 		mlog_errno(status);
1988 		goto bail;
1989 	}
1990 
1991 	if (!ocfs2_inode_is_fast_symlink(inode)) {
1992 		status = ocfs2_create_symlink_data(osb, handle, inode,
1993 						   symname);
1994 		if (status < 0) {
1995 			mlog_errno(status);
1996 			goto bail;
1997 		}
1998 	}
1999 
2000 	if (si.enable) {
2001 		status = ocfs2_init_security_set(handle, inode, new_fe_bh, &si,
2002 						 xattr_ac, data_ac);
2003 		if (status < 0) {
2004 			mlog_errno(status);
2005 			goto bail;
2006 		}
2007 	}
2008 
2009 	/*
2010 	 * Do this before adding the entry to the directory. We add
2011 	 * also set d_op after success so that ->d_iput() will cleanup
2012 	 * the dentry lock even if ocfs2_add_entry() fails below.
2013 	 */
2014 	status = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
2015 	if (status) {
2016 		mlog_errno(status);
2017 		goto bail;
2018 	}
2019 
2020 	dl = dentry->d_fsdata;
2021 
2022 	status = ocfs2_add_entry(handle, dentry, inode,
2023 				 le64_to_cpu(fe->i_blkno), parent_fe_bh,
2024 				 &lookup);
2025 	if (status < 0) {
2026 		mlog_errno(status);
2027 		goto bail;
2028 	}
2029 
2030 	insert_inode_hash(inode);
2031 	d_instantiate(dentry, inode);
2032 bail:
2033 	if (status < 0 && did_quota)
2034 		dquot_free_space_nodirty(inode,
2035 					ocfs2_clusters_to_bytes(osb->sb, 1));
2036 	if (status < 0 && did_quota_inode)
2037 		dquot_free_inode(inode);
2038 	if (handle) {
2039 		if (status < 0 && fe)
2040 			ocfs2_set_links_count(fe, 0);
2041 		ocfs2_commit_trans(osb, handle);
2042 	}
2043 
2044 	ocfs2_inode_unlock(dir, 1);
2045 	if (did_block_signals)
2046 		ocfs2_unblock_signals(&oldset);
2047 
2048 	brelse(new_fe_bh);
2049 	brelse(parent_fe_bh);
2050 	kfree(si.value);
2051 	ocfs2_free_dir_lookup_result(&lookup);
2052 	if (inode_ac)
2053 		ocfs2_free_alloc_context(inode_ac);
2054 	if (data_ac)
2055 		ocfs2_free_alloc_context(data_ac);
2056 	if (xattr_ac)
2057 		ocfs2_free_alloc_context(xattr_ac);
2058 	if ((status < 0) && inode) {
2059 		if (dl)
2060 			ocfs2_cleanup_add_entry_failure(osb, dentry, inode);
2061 
2062 		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR;
2063 		clear_nlink(inode);
2064 		iput(inode);
2065 	}
2066 
2067 	if (status)
2068 		mlog_errno(status);
2069 
2070 	return status;
2071 }
2072 
ocfs2_blkno_stringify(u64 blkno,char * name)2073 static int ocfs2_blkno_stringify(u64 blkno, char *name)
2074 {
2075 	int status, namelen;
2076 
2077 	namelen = snprintf(name, OCFS2_ORPHAN_NAMELEN + 1, "%016llx",
2078 			   (long long)blkno);
2079 	if (namelen <= 0) {
2080 		if (namelen)
2081 			status = namelen;
2082 		else
2083 			status = -EINVAL;
2084 		mlog_errno(status);
2085 		goto bail;
2086 	}
2087 	if (namelen != OCFS2_ORPHAN_NAMELEN) {
2088 		status = -EINVAL;
2089 		mlog_errno(status);
2090 		goto bail;
2091 	}
2092 
2093 	trace_ocfs2_blkno_stringify(blkno, name, namelen);
2094 
2095 	status = 0;
2096 bail:
2097 	if (status < 0)
2098 		mlog_errno(status);
2099 	return status;
2100 }
2101 
ocfs2_lookup_lock_orphan_dir(struct ocfs2_super * osb,struct inode ** ret_orphan_dir,struct buffer_head ** ret_orphan_dir_bh)2102 static int ocfs2_lookup_lock_orphan_dir(struct ocfs2_super *osb,
2103 					struct inode **ret_orphan_dir,
2104 					struct buffer_head **ret_orphan_dir_bh)
2105 {
2106 	struct inode *orphan_dir_inode;
2107 	struct buffer_head *orphan_dir_bh = NULL;
2108 	int ret = 0;
2109 
2110 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2111 						       ORPHAN_DIR_SYSTEM_INODE,
2112 						       osb->slot_num);
2113 	if (!orphan_dir_inode) {
2114 		ret = -ENOENT;
2115 		mlog_errno(ret);
2116 		return ret;
2117 	}
2118 
2119 	inode_lock(orphan_dir_inode);
2120 
2121 	ret = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2122 	if (ret < 0) {
2123 		inode_unlock(orphan_dir_inode);
2124 		iput(orphan_dir_inode);
2125 
2126 		mlog_errno(ret);
2127 		return ret;
2128 	}
2129 
2130 	*ret_orphan_dir = orphan_dir_inode;
2131 	*ret_orphan_dir_bh = orphan_dir_bh;
2132 
2133 	return 0;
2134 }
2135 
__ocfs2_prepare_orphan_dir(struct inode * orphan_dir_inode,struct buffer_head * orphan_dir_bh,u64 blkno,char * name,struct ocfs2_dir_lookup_result * lookup,bool dio)2136 static int __ocfs2_prepare_orphan_dir(struct inode *orphan_dir_inode,
2137 				      struct buffer_head *orphan_dir_bh,
2138 				      u64 blkno,
2139 				      char *name,
2140 				      struct ocfs2_dir_lookup_result *lookup,
2141 				      bool dio)
2142 {
2143 	int ret;
2144 	struct ocfs2_super *osb = OCFS2_SB(orphan_dir_inode->i_sb);
2145 	int namelen = dio ?
2146 			(OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN) :
2147 			OCFS2_ORPHAN_NAMELEN;
2148 
2149 	if (dio) {
2150 		ret = snprintf(name, OCFS2_DIO_ORPHAN_PREFIX_LEN + 1, "%s",
2151 				OCFS2_DIO_ORPHAN_PREFIX);
2152 		if (ret != OCFS2_DIO_ORPHAN_PREFIX_LEN) {
2153 			ret = -EINVAL;
2154 			mlog_errno(ret);
2155 			return ret;
2156 		}
2157 
2158 		ret = ocfs2_blkno_stringify(blkno,
2159 				name + OCFS2_DIO_ORPHAN_PREFIX_LEN);
2160 	} else
2161 		ret = ocfs2_blkno_stringify(blkno, name);
2162 	if (ret < 0) {
2163 		mlog_errno(ret);
2164 		return ret;
2165 	}
2166 
2167 	ret = ocfs2_prepare_dir_for_insert(osb, orphan_dir_inode,
2168 					   orphan_dir_bh, name,
2169 					   namelen, lookup);
2170 	if (ret < 0) {
2171 		mlog_errno(ret);
2172 		return ret;
2173 	}
2174 
2175 	return 0;
2176 }
2177 
2178 /**
2179  * ocfs2_prepare_orphan_dir() - Prepare an orphan directory for
2180  * insertion of an orphan.
2181  * @osb: ocfs2 file system
2182  * @ret_orphan_dir: Orphan dir inode - returned locked!
2183  * @blkno: Actual block number of the inode to be inserted into orphan dir.
2184  * @lookup: dir lookup result, to be passed back into functions like
2185  *          ocfs2_orphan_add
2186  *
2187  * Returns zero on success and the ret_orphan_dir, name and lookup
2188  * fields will be populated.
2189  *
2190  * Returns non-zero on failure.
2191  */
ocfs2_prepare_orphan_dir(struct ocfs2_super * osb,struct inode ** ret_orphan_dir,u64 blkno,char * name,struct ocfs2_dir_lookup_result * lookup,bool dio)2192 static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
2193 				    struct inode **ret_orphan_dir,
2194 				    u64 blkno,
2195 				    char *name,
2196 				    struct ocfs2_dir_lookup_result *lookup,
2197 				    bool dio)
2198 {
2199 	struct inode *orphan_dir_inode = NULL;
2200 	struct buffer_head *orphan_dir_bh = NULL;
2201 	int ret = 0;
2202 
2203 	ret = ocfs2_lookup_lock_orphan_dir(osb, &orphan_dir_inode,
2204 					   &orphan_dir_bh);
2205 	if (ret < 0) {
2206 		mlog_errno(ret);
2207 		return ret;
2208 	}
2209 
2210 	ret = __ocfs2_prepare_orphan_dir(orphan_dir_inode, orphan_dir_bh,
2211 					 blkno, name, lookup, dio);
2212 	if (ret < 0) {
2213 		mlog_errno(ret);
2214 		goto out;
2215 	}
2216 
2217 	*ret_orphan_dir = orphan_dir_inode;
2218 
2219 out:
2220 	brelse(orphan_dir_bh);
2221 
2222 	if (ret) {
2223 		ocfs2_inode_unlock(orphan_dir_inode, 1);
2224 		inode_unlock(orphan_dir_inode);
2225 		iput(orphan_dir_inode);
2226 	}
2227 
2228 	if (ret)
2229 		mlog_errno(ret);
2230 	return ret;
2231 }
2232 
ocfs2_orphan_add(struct ocfs2_super * osb,handle_t * handle,struct inode * inode,struct buffer_head * fe_bh,char * name,struct ocfs2_dir_lookup_result * lookup,struct inode * orphan_dir_inode,bool dio)2233 static int ocfs2_orphan_add(struct ocfs2_super *osb,
2234 			    handle_t *handle,
2235 			    struct inode *inode,
2236 			    struct buffer_head *fe_bh,
2237 			    char *name,
2238 			    struct ocfs2_dir_lookup_result *lookup,
2239 			    struct inode *orphan_dir_inode,
2240 			    bool dio)
2241 {
2242 	struct buffer_head *orphan_dir_bh = NULL;
2243 	int status = 0;
2244 	struct ocfs2_dinode *orphan_fe;
2245 	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data;
2246 	int namelen = dio ?
2247 			(OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN) :
2248 			OCFS2_ORPHAN_NAMELEN;
2249 
2250 	trace_ocfs2_orphan_add_begin(
2251 				(unsigned long long)OCFS2_I(inode)->ip_blkno);
2252 
2253 	status = ocfs2_read_inode_block(orphan_dir_inode, &orphan_dir_bh);
2254 	if (status < 0) {
2255 		mlog_errno(status);
2256 		goto leave;
2257 	}
2258 
2259 	status = ocfs2_journal_access_di(handle,
2260 					 INODE_CACHE(orphan_dir_inode),
2261 					 orphan_dir_bh,
2262 					 OCFS2_JOURNAL_ACCESS_WRITE);
2263 	if (status < 0) {
2264 		mlog_errno(status);
2265 		goto leave;
2266 	}
2267 
2268 	/*
2269 	 * We're going to journal the change of i_flags and i_orphaned_slot.
2270 	 * It's safe anyway, though some callers may duplicate the journaling.
2271 	 * Journaling within the func just make the logic look more
2272 	 * straightforward.
2273 	 */
2274 	status = ocfs2_journal_access_di(handle,
2275 					 INODE_CACHE(inode),
2276 					 fe_bh,
2277 					 OCFS2_JOURNAL_ACCESS_WRITE);
2278 	if (status < 0) {
2279 		mlog_errno(status);
2280 		goto leave;
2281 	}
2282 
2283 	/* we're a cluster, and nlink can change on disk from
2284 	 * underneath us... */
2285 	orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data;
2286 	if (S_ISDIR(inode->i_mode))
2287 		ocfs2_add_links_count(orphan_fe, 1);
2288 	set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2289 	ocfs2_journal_dirty(handle, orphan_dir_bh);
2290 
2291 	status = __ocfs2_add_entry(handle, orphan_dir_inode, name,
2292 				   namelen, inode,
2293 				   OCFS2_I(inode)->ip_blkno,
2294 				   orphan_dir_bh, lookup);
2295 	if (status < 0) {
2296 		mlog_errno(status);
2297 		goto rollback;
2298 	}
2299 
2300 	if (dio) {
2301 		/* Update flag OCFS2_DIO_ORPHANED_FL and record the orphan
2302 		 * slot.
2303 		 */
2304 		fe->i_flags |= cpu_to_le32(OCFS2_DIO_ORPHANED_FL);
2305 		fe->i_dio_orphaned_slot = cpu_to_le16(osb->slot_num);
2306 	} else {
2307 		fe->i_flags |= cpu_to_le32(OCFS2_ORPHANED_FL);
2308 		OCFS2_I(inode)->ip_flags &= ~OCFS2_INODE_SKIP_ORPHAN_DIR;
2309 
2310 		/* Record which orphan dir our inode now resides
2311 		 * in. delete_inode will use this to determine which orphan
2312 		 * dir to lock. */
2313 		fe->i_orphaned_slot = cpu_to_le16(osb->slot_num);
2314 	}
2315 
2316 	ocfs2_journal_dirty(handle, fe_bh);
2317 
2318 	trace_ocfs2_orphan_add_end((unsigned long long)OCFS2_I(inode)->ip_blkno,
2319 				   osb->slot_num);
2320 
2321 rollback:
2322 	if (status < 0) {
2323 		if (S_ISDIR(inode->i_mode))
2324 			ocfs2_add_links_count(orphan_fe, -1);
2325 		set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2326 	}
2327 
2328 leave:
2329 	brelse(orphan_dir_bh);
2330 
2331 	return status;
2332 }
2333 
2334 /* unlike orphan_add, we expect the orphan dir to already be locked here. */
ocfs2_orphan_del(struct ocfs2_super * osb,handle_t * handle,struct inode * orphan_dir_inode,struct inode * inode,struct buffer_head * orphan_dir_bh,bool dio)2335 int ocfs2_orphan_del(struct ocfs2_super *osb,
2336 		     handle_t *handle,
2337 		     struct inode *orphan_dir_inode,
2338 		     struct inode *inode,
2339 		     struct buffer_head *orphan_dir_bh,
2340 		     bool dio)
2341 {
2342 	char name[OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN + 1];
2343 	struct ocfs2_dinode *orphan_fe;
2344 	int status = 0;
2345 	struct ocfs2_dir_lookup_result lookup = { NULL, };
2346 
2347 	if (dio) {
2348 		status = snprintf(name, OCFS2_DIO_ORPHAN_PREFIX_LEN + 1, "%s",
2349 				OCFS2_DIO_ORPHAN_PREFIX);
2350 		if (status != OCFS2_DIO_ORPHAN_PREFIX_LEN) {
2351 			status = -EINVAL;
2352 			mlog_errno(status);
2353 			return status;
2354 		}
2355 
2356 		status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno,
2357 				name + OCFS2_DIO_ORPHAN_PREFIX_LEN);
2358 	} else
2359 		status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name);
2360 	if (status < 0) {
2361 		mlog_errno(status);
2362 		goto leave;
2363 	}
2364 
2365 	trace_ocfs2_orphan_del(
2366 	     (unsigned long long)OCFS2_I(orphan_dir_inode)->ip_blkno,
2367 	     name, strlen(name));
2368 
2369 	status = ocfs2_journal_access_di(handle,
2370 					 INODE_CACHE(orphan_dir_inode),
2371 					 orphan_dir_bh,
2372 					 OCFS2_JOURNAL_ACCESS_WRITE);
2373 	if (status < 0) {
2374 		mlog_errno(status);
2375 		goto leave;
2376 	}
2377 
2378 	/* find it's spot in the orphan directory */
2379 	status = ocfs2_find_entry(name, strlen(name), orphan_dir_inode,
2380 				  &lookup);
2381 	if (status) {
2382 		mlog_errno(status);
2383 		goto leave;
2384 	}
2385 
2386 	/* remove it from the orphan directory */
2387 	status = ocfs2_delete_entry(handle, orphan_dir_inode, &lookup);
2388 	if (status < 0) {
2389 		mlog_errno(status);
2390 		goto leave;
2391 	}
2392 
2393 	/* do the i_nlink dance! :) */
2394 	orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data;
2395 	if (S_ISDIR(inode->i_mode))
2396 		ocfs2_add_links_count(orphan_fe, -1);
2397 	set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe));
2398 	ocfs2_journal_dirty(handle, orphan_dir_bh);
2399 
2400 leave:
2401 	ocfs2_free_dir_lookup_result(&lookup);
2402 
2403 	if (status)
2404 		mlog_errno(status);
2405 	return status;
2406 }
2407 
2408 /**
2409  * ocfs2_prep_new_orphaned_file() - Prepare the orphan dir to receive a newly
2410  * allocated file. This is different from the typical 'add to orphan dir'
2411  * operation in that the inode does not yet exist. This is a problem because
2412  * the orphan dir stringifies the inode block number to come up with it's
2413  * dirent. Obviously if the inode does not yet exist we have a chicken and egg
2414  * problem. This function works around it by calling deeper into the orphan
2415  * and suballoc code than other callers. Use this only by necessity.
2416  * @dir: The directory which this inode will ultimately wind up under - not the
2417  * orphan dir!
2418  * @dir_bh: buffer_head the @dir inode block
2419  * @orphan_name: string of length (CFS2_ORPHAN_NAMELEN + 1). Will be filled
2420  * with the string to be used for orphan dirent. Pass back to the orphan dir
2421  * code.
2422  * @ret_orphan_dir: orphan dir inode returned to be passed back into orphan
2423  * dir code.
2424  * @ret_di_blkno: block number where the new inode will be allocated.
2425  * @orphan_insert: Dir insert context to be passed back into orphan dir code.
2426  * @ret_inode_ac: Inode alloc context to be passed back to the allocator.
2427  *
2428  * Returns zero on success and the ret_orphan_dir, name and lookup
2429  * fields will be populated.
2430  *
2431  * Returns non-zero on failure.
2432  */
ocfs2_prep_new_orphaned_file(struct inode * dir,struct buffer_head * dir_bh,char * orphan_name,struct inode ** ret_orphan_dir,u64 * ret_di_blkno,struct ocfs2_dir_lookup_result * orphan_insert,struct ocfs2_alloc_context ** ret_inode_ac)2433 static int ocfs2_prep_new_orphaned_file(struct inode *dir,
2434 					struct buffer_head *dir_bh,
2435 					char *orphan_name,
2436 					struct inode **ret_orphan_dir,
2437 					u64 *ret_di_blkno,
2438 					struct ocfs2_dir_lookup_result *orphan_insert,
2439 					struct ocfs2_alloc_context **ret_inode_ac)
2440 {
2441 	int ret;
2442 	u64 di_blkno;
2443 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
2444 	struct inode *orphan_dir = NULL;
2445 	struct buffer_head *orphan_dir_bh = NULL;
2446 	struct ocfs2_alloc_context *inode_ac = NULL;
2447 
2448 	ret = ocfs2_lookup_lock_orphan_dir(osb, &orphan_dir, &orphan_dir_bh);
2449 	if (ret < 0) {
2450 		mlog_errno(ret);
2451 		return ret;
2452 	}
2453 
2454 	/* reserve an inode spot */
2455 	ret = ocfs2_reserve_new_inode(osb, &inode_ac);
2456 	if (ret < 0) {
2457 		if (ret != -ENOSPC)
2458 			mlog_errno(ret);
2459 		goto out;
2460 	}
2461 
2462 	ret = ocfs2_find_new_inode_loc(dir, dir_bh, inode_ac,
2463 				       &di_blkno);
2464 	if (ret) {
2465 		mlog_errno(ret);
2466 		goto out;
2467 	}
2468 
2469 	ret = __ocfs2_prepare_orphan_dir(orphan_dir, orphan_dir_bh,
2470 					 di_blkno, orphan_name, orphan_insert,
2471 					 false);
2472 	if (ret < 0) {
2473 		mlog_errno(ret);
2474 		goto out;
2475 	}
2476 
2477 out:
2478 	if (ret == 0) {
2479 		*ret_orphan_dir = orphan_dir;
2480 		*ret_di_blkno = di_blkno;
2481 		*ret_inode_ac = inode_ac;
2482 		/*
2483 		 * orphan_name and orphan_insert are already up to
2484 		 * date via prepare_orphan_dir
2485 		 */
2486 	} else {
2487 		/* Unroll reserve_new_inode* */
2488 		if (inode_ac)
2489 			ocfs2_free_alloc_context(inode_ac);
2490 
2491 		/* Unroll orphan dir locking */
2492 		inode_unlock(orphan_dir);
2493 		ocfs2_inode_unlock(orphan_dir, 1);
2494 		iput(orphan_dir);
2495 	}
2496 
2497 	brelse(orphan_dir_bh);
2498 
2499 	return ret;
2500 }
2501 
ocfs2_create_inode_in_orphan(struct inode * dir,int mode,struct inode ** new_inode)2502 int ocfs2_create_inode_in_orphan(struct inode *dir,
2503 				 int mode,
2504 				 struct inode **new_inode)
2505 {
2506 	int status, did_quota_inode = 0;
2507 	struct inode *inode = NULL;
2508 	struct inode *orphan_dir = NULL;
2509 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
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 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 	status = ocfs2_orphan_add(osb, handle, inode, new_di_bh, orphan_name,
2576 				  &orphan_insert, orphan_dir, false);
2577 	if (status < 0) {
2578 		mlog_errno(status);
2579 		goto leave;
2580 	}
2581 
2582 	/* get open lock so that only nodes can't remove it from orphan dir. */
2583 	status = ocfs2_open_lock(inode);
2584 	if (status < 0)
2585 		mlog_errno(status);
2586 
2587 	insert_inode_hash(inode);
2588 leave:
2589 	if (status < 0 && did_quota_inode)
2590 		dquot_free_inode(inode);
2591 	if (handle)
2592 		ocfs2_commit_trans(osb, handle);
2593 
2594 	if (orphan_dir) {
2595 		/* This was locked for us in ocfs2_prepare_orphan_dir() */
2596 		ocfs2_inode_unlock(orphan_dir, 1);
2597 		inode_unlock(orphan_dir);
2598 		iput(orphan_dir);
2599 	}
2600 
2601 	if ((status < 0) && inode) {
2602 		clear_nlink(inode);
2603 		iput(inode);
2604 	}
2605 
2606 	if (inode_ac)
2607 		ocfs2_free_alloc_context(inode_ac);
2608 
2609 	brelse(new_di_bh);
2610 
2611 	if (!status)
2612 		*new_inode = inode;
2613 
2614 	ocfs2_free_dir_lookup_result(&orphan_insert);
2615 
2616 	ocfs2_inode_unlock(dir, 1);
2617 	brelse(parent_di_bh);
2618 	return status;
2619 }
2620 
ocfs2_add_inode_to_orphan(struct ocfs2_super * osb,struct inode * inode)2621 int ocfs2_add_inode_to_orphan(struct ocfs2_super *osb,
2622 	struct inode *inode)
2623 {
2624 	char orphan_name[OCFS2_DIO_ORPHAN_PREFIX_LEN + OCFS2_ORPHAN_NAMELEN + 1];
2625 	struct inode *orphan_dir_inode = NULL;
2626 	struct ocfs2_dir_lookup_result orphan_insert = { NULL, };
2627 	struct buffer_head *di_bh = NULL;
2628 	int status = 0;
2629 	handle_t *handle = NULL;
2630 	struct ocfs2_dinode *di = NULL;
2631 
2632 	status = ocfs2_inode_lock(inode, &di_bh, 1);
2633 	if (status < 0) {
2634 		mlog_errno(status);
2635 		goto bail;
2636 	}
2637 
2638 	di = (struct ocfs2_dinode *) di_bh->b_data;
2639 	/*
2640 	 * Another append dio crashed?
2641 	 * If so, manually recover it first.
2642 	 */
2643 	if (unlikely(di->i_flags & cpu_to_le32(OCFS2_DIO_ORPHANED_FL))) {
2644 		status = ocfs2_truncate_file(inode, di_bh, i_size_read(inode));
2645 		if (status < 0) {
2646 			if (status != -ENOSPC)
2647 				mlog_errno(status);
2648 			goto bail_unlock_inode;
2649 		}
2650 
2651 		status = ocfs2_del_inode_from_orphan(osb, inode, di_bh, 0, 0);
2652 		if (status < 0) {
2653 			mlog_errno(status);
2654 			goto bail_unlock_inode;
2655 		}
2656 	}
2657 
2658 	status = ocfs2_prepare_orphan_dir(osb, &orphan_dir_inode,
2659 			OCFS2_I(inode)->ip_blkno,
2660 			orphan_name,
2661 			&orphan_insert,
2662 			true);
2663 	if (status < 0) {
2664 		mlog_errno(status);
2665 		goto bail_unlock_inode;
2666 	}
2667 
2668 	handle = ocfs2_start_trans(osb,
2669 			OCFS2_INODE_ADD_TO_ORPHAN_CREDITS);
2670 	if (IS_ERR(handle)) {
2671 		status = PTR_ERR(handle);
2672 		goto bail_unlock_orphan;
2673 	}
2674 
2675 	status = ocfs2_orphan_add(osb, handle, inode, di_bh, orphan_name,
2676 			&orphan_insert, orphan_dir_inode, true);
2677 	if (status)
2678 		mlog_errno(status);
2679 
2680 	ocfs2_commit_trans(osb, handle);
2681 
2682 bail_unlock_orphan:
2683 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2684 	inode_unlock(orphan_dir_inode);
2685 	iput(orphan_dir_inode);
2686 
2687 	ocfs2_free_dir_lookup_result(&orphan_insert);
2688 
2689 bail_unlock_inode:
2690 	ocfs2_inode_unlock(inode, 1);
2691 	brelse(di_bh);
2692 
2693 bail:
2694 	return status;
2695 }
2696 
ocfs2_del_inode_from_orphan(struct ocfs2_super * osb,struct inode * inode,struct buffer_head * di_bh,int update_isize,loff_t end)2697 int ocfs2_del_inode_from_orphan(struct ocfs2_super *osb,
2698 		struct inode *inode, struct buffer_head *di_bh,
2699 		int update_isize, loff_t end)
2700 {
2701 	struct inode *orphan_dir_inode = NULL;
2702 	struct buffer_head *orphan_dir_bh = NULL;
2703 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
2704 	handle_t *handle = NULL;
2705 	int status = 0;
2706 
2707 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2708 			ORPHAN_DIR_SYSTEM_INODE,
2709 			le16_to_cpu(di->i_dio_orphaned_slot));
2710 	if (!orphan_dir_inode) {
2711 		status = -ENOENT;
2712 		mlog_errno(status);
2713 		goto bail;
2714 	}
2715 
2716 	inode_lock(orphan_dir_inode);
2717 	status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2718 	if (status < 0) {
2719 		inode_unlock(orphan_dir_inode);
2720 		iput(orphan_dir_inode);
2721 		mlog_errno(status);
2722 		goto bail;
2723 	}
2724 
2725 	handle = ocfs2_start_trans(osb,
2726 			OCFS2_INODE_DEL_FROM_ORPHAN_CREDITS);
2727 	if (IS_ERR(handle)) {
2728 		status = PTR_ERR(handle);
2729 		goto bail_unlock_orphan;
2730 	}
2731 
2732 	BUG_ON(!(di->i_flags & cpu_to_le32(OCFS2_DIO_ORPHANED_FL)));
2733 
2734 	status = ocfs2_orphan_del(osb, handle, orphan_dir_inode,
2735 				inode, orphan_dir_bh, true);
2736 	if (status < 0) {
2737 		mlog_errno(status);
2738 		goto bail_commit;
2739 	}
2740 
2741 	status = ocfs2_journal_access_di(handle,
2742 			INODE_CACHE(inode),
2743 			di_bh,
2744 			OCFS2_JOURNAL_ACCESS_WRITE);
2745 	if (status < 0) {
2746 		mlog_errno(status);
2747 		goto bail_commit;
2748 	}
2749 
2750 	di->i_flags &= ~cpu_to_le32(OCFS2_DIO_ORPHANED_FL);
2751 	di->i_dio_orphaned_slot = 0;
2752 
2753 	if (update_isize) {
2754 		status = ocfs2_set_inode_size(handle, inode, di_bh, end);
2755 		if (status)
2756 			mlog_errno(status);
2757 	} else
2758 		ocfs2_journal_dirty(handle, di_bh);
2759 
2760 bail_commit:
2761 	ocfs2_commit_trans(osb, handle);
2762 
2763 bail_unlock_orphan:
2764 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2765 	inode_unlock(orphan_dir_inode);
2766 	brelse(orphan_dir_bh);
2767 	iput(orphan_dir_inode);
2768 
2769 bail:
2770 	return status;
2771 }
2772 
ocfs2_mv_orphaned_inode_to_new(struct inode * dir,struct inode * inode,struct dentry * dentry)2773 int ocfs2_mv_orphaned_inode_to_new(struct inode *dir,
2774 				   struct inode *inode,
2775 				   struct dentry *dentry)
2776 {
2777 	int status = 0;
2778 	struct buffer_head *parent_di_bh = NULL;
2779 	handle_t *handle = NULL;
2780 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
2781 	struct ocfs2_dinode *dir_di, *di;
2782 	struct inode *orphan_dir_inode = NULL;
2783 	struct buffer_head *orphan_dir_bh = NULL;
2784 	struct buffer_head *di_bh = NULL;
2785 	struct ocfs2_dir_lookup_result lookup = { NULL, };
2786 
2787 	trace_ocfs2_mv_orphaned_inode_to_new(dir, dentry,
2788 				dentry->d_name.len, dentry->d_name.name,
2789 				(unsigned long long)OCFS2_I(dir)->ip_blkno,
2790 				(unsigned long long)OCFS2_I(inode)->ip_blkno);
2791 
2792 	status = ocfs2_inode_lock(dir, &parent_di_bh, 1);
2793 	if (status < 0) {
2794 		if (status != -ENOENT)
2795 			mlog_errno(status);
2796 		return status;
2797 	}
2798 
2799 	dir_di = (struct ocfs2_dinode *) parent_di_bh->b_data;
2800 	if (!dir_di->i_links_count) {
2801 		/* can't make a file in a deleted directory. */
2802 		status = -ENOENT;
2803 		goto leave;
2804 	}
2805 
2806 	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
2807 					   dentry->d_name.len);
2808 	if (status)
2809 		goto leave;
2810 
2811 	/* get a spot inside the dir. */
2812 	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_di_bh,
2813 					      dentry->d_name.name,
2814 					      dentry->d_name.len, &lookup);
2815 	if (status < 0) {
2816 		mlog_errno(status);
2817 		goto leave;
2818 	}
2819 
2820 	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2821 						       ORPHAN_DIR_SYSTEM_INODE,
2822 						       osb->slot_num);
2823 	if (!orphan_dir_inode) {
2824 		status = -ENOENT;
2825 		mlog_errno(status);
2826 		goto leave;
2827 	}
2828 
2829 	inode_lock(orphan_dir_inode);
2830 
2831 	status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2832 	if (status < 0) {
2833 		mlog_errno(status);
2834 		inode_unlock(orphan_dir_inode);
2835 		iput(orphan_dir_inode);
2836 		goto leave;
2837 	}
2838 
2839 	status = ocfs2_read_inode_block(inode, &di_bh);
2840 	if (status < 0) {
2841 		mlog_errno(status);
2842 		goto orphan_unlock;
2843 	}
2844 
2845 	handle = ocfs2_start_trans(osb, ocfs2_rename_credits(osb->sb));
2846 	if (IS_ERR(handle)) {
2847 		status = PTR_ERR(handle);
2848 		handle = NULL;
2849 		mlog_errno(status);
2850 		goto orphan_unlock;
2851 	}
2852 
2853 	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
2854 					 di_bh, OCFS2_JOURNAL_ACCESS_WRITE);
2855 	if (status < 0) {
2856 		mlog_errno(status);
2857 		goto out_commit;
2858 	}
2859 
2860 	status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode,
2861 				  orphan_dir_bh, false);
2862 	if (status < 0) {
2863 		mlog_errno(status);
2864 		goto out_commit;
2865 	}
2866 
2867 	di = (struct ocfs2_dinode *)di_bh->b_data;
2868 	di->i_flags &= ~cpu_to_le32(OCFS2_ORPHANED_FL);
2869 	di->i_orphaned_slot = 0;
2870 	set_nlink(inode, 1);
2871 	ocfs2_set_links_count(di, inode->i_nlink);
2872 	ocfs2_update_inode_fsync_trans(handle, inode, 1);
2873 	ocfs2_journal_dirty(handle, di_bh);
2874 
2875 	status = ocfs2_add_entry(handle, dentry, inode,
2876 				 OCFS2_I(inode)->ip_blkno, parent_di_bh,
2877 				 &lookup);
2878 	if (status < 0) {
2879 		mlog_errno(status);
2880 		goto out_commit;
2881 	}
2882 
2883 	status = ocfs2_dentry_attach_lock(dentry, inode,
2884 					  OCFS2_I(dir)->ip_blkno);
2885 	if (status) {
2886 		mlog_errno(status);
2887 		goto out_commit;
2888 	}
2889 
2890 	d_instantiate(dentry, inode);
2891 	status = 0;
2892 out_commit:
2893 	ocfs2_commit_trans(osb, handle);
2894 orphan_unlock:
2895 	ocfs2_inode_unlock(orphan_dir_inode, 1);
2896 	inode_unlock(orphan_dir_inode);
2897 	iput(orphan_dir_inode);
2898 leave:
2899 
2900 	ocfs2_inode_unlock(dir, 1);
2901 
2902 	brelse(di_bh);
2903 	brelse(parent_di_bh);
2904 	brelse(orphan_dir_bh);
2905 
2906 	ocfs2_free_dir_lookup_result(&lookup);
2907 
2908 	if (status)
2909 		mlog_errno(status);
2910 
2911 	return status;
2912 }
2913 
2914 const struct inode_operations ocfs2_dir_iops = {
2915 	.create		= ocfs2_create,
2916 	.lookup		= ocfs2_lookup,
2917 	.link		= ocfs2_link,
2918 	.unlink		= ocfs2_unlink,
2919 	.rmdir		= ocfs2_unlink,
2920 	.symlink	= ocfs2_symlink,
2921 	.mkdir		= ocfs2_mkdir,
2922 	.mknod		= ocfs2_mknod,
2923 	.rename		= ocfs2_rename,
2924 	.setattr	= ocfs2_setattr,
2925 	.getattr	= ocfs2_getattr,
2926 	.permission	= ocfs2_permission,
2927 	.listxattr	= ocfs2_listxattr,
2928 	.fiemap         = ocfs2_fiemap,
2929 	.get_inode_acl	= ocfs2_iop_get_acl,
2930 	.set_acl	= ocfs2_iop_set_acl,
2931 	.fileattr_get	= ocfs2_fileattr_get,
2932 	.fileattr_set	= ocfs2_fileattr_set,
2933 };
2934