inode.c 40.1 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * inode.c
 *
 * vfs' aops, fops, dops and iops
 *
 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/fs.h>
#include <linux/types.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
30
#include <linux/quotaops.h>
31
32
33
34
35
36
37
38
39

#include <asm/byteorder.h>

#define MLOG_MASK_PREFIX ML_INODE
#include <cluster/masklog.h>

#include "ocfs2.h"

#include "alloc.h"
40
#include "dir.h"
Joel Becker's avatar
Joel Becker committed
41
#include "blockcheck.h"
42
43
44
#include "dlmglue.h"
#include "extent_map.h"
#include "file.h"
45
#include "heartbeat.h"
46
47
48
49
50
51
52
53
#include "inode.h"
#include "journal.h"
#include "namei.h"
#include "suballoc.h"
#include "super.h"
#include "symlink.h"
#include "sysfile.h"
#include "uptodate.h"
54
#include "xattr.h"
55
#include "refcounttree.h"
56
57
58
59
60
61
62
63

#include "buffer_head_io.h"

struct ocfs2_find_inode_args
{
	u64		fi_blkno;
	unsigned long	fi_ino;
	unsigned int	fi_flags;
64
	unsigned int	fi_sysfile_type;
65
66
};

67
68
static struct lock_class_key ocfs2_sysfile_lock_key[NUM_SYSTEM_INODES];

69
70
71
72
73
74
75
76
static int ocfs2_read_locked_inode(struct inode *inode,
				   struct ocfs2_find_inode_args *args);
static int ocfs2_init_locked_inode(struct inode *inode, void *opaque);
static int ocfs2_find_actor(struct inode *inode, void *opaque);
static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
				    struct inode *inode,
				    struct buffer_head *fe_bh);

Herbert Poetzl's avatar
Herbert Poetzl committed
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
void ocfs2_set_inode_flags(struct inode *inode)
{
	unsigned int flags = OCFS2_I(inode)->ip_attr;

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

	if (flags & OCFS2_IMMUTABLE_FL)
		inode->i_flags |= S_IMMUTABLE;

	if (flags & OCFS2_SYNC_FL)
		inode->i_flags |= S_SYNC;
	if (flags & OCFS2_APPEND_FL)
		inode->i_flags |= S_APPEND;
	if (flags & OCFS2_NOATIME_FL)
		inode->i_flags |= S_NOATIME;
	if (flags & OCFS2_DIRSYNC_FL)
		inode->i_flags |= S_DIRSYNC;
}

97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
/* Propagate flags from i_flags to OCFS2_I(inode)->ip_attr */
void ocfs2_get_inode_flags(struct ocfs2_inode_info *oi)
{
	unsigned int flags = oi->vfs_inode.i_flags;

	oi->ip_attr &= ~(OCFS2_SYNC_FL|OCFS2_APPEND_FL|
			OCFS2_IMMUTABLE_FL|OCFS2_NOATIME_FL|OCFS2_DIRSYNC_FL);
	if (flags & S_SYNC)
		oi->ip_attr |= OCFS2_SYNC_FL;
	if (flags & S_APPEND)
		oi->ip_attr |= OCFS2_APPEND_FL;
	if (flags & S_IMMUTABLE)
		oi->ip_attr |= OCFS2_IMMUTABLE_FL;
	if (flags & S_NOATIME)
		oi->ip_attr |= OCFS2_NOATIME_FL;
	if (flags & S_DIRSYNC)
		oi->ip_attr |= OCFS2_DIRSYNC_FL;
}

116
117
118
119
120
121
122
123
124
125
126
struct inode *ocfs2_ilookup(struct super_block *sb, u64 blkno)
{
	struct ocfs2_find_inode_args args;

	args.fi_blkno = blkno;
	args.fi_flags = 0;
	args.fi_ino = ino_from_blkno(sb, blkno);
	args.fi_sysfile_type = 0;

	return ilookup5(sb, blkno, ocfs2_find_actor, &args);
}
127
128
struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, unsigned flags,
			 int sysfile_type)
129
130
131
132
133
{
	struct inode *inode = NULL;
	struct super_block *sb = osb->sb;
	struct ocfs2_find_inode_args args;

Tao Ma's avatar
Tao Ma committed
134
	mlog(0, "(blkno = %llu)\n", (unsigned long long)blkno);
135
136
137
138
139
140
141
142
143
144
145

	/* Ok. By now we've either got the offsets passed to us by the
	 * caller, or we just pulled them off the bh. Lets do some
	 * sanity checks to make sure they're OK. */
	if (blkno == 0) {
		inode = ERR_PTR(-EINVAL);
		mlog_errno(PTR_ERR(inode));
		goto bail;
	}

	args.fi_blkno = blkno;
146
	args.fi_flags = flags;
147
	args.fi_ino = ino_from_blkno(sb, blkno);
148
	args.fi_sysfile_type = sysfile_type;
149
150
151
152
153
154
155
156
157
158
159

	inode = iget5_locked(sb, args.fi_ino, ocfs2_find_actor,
			     ocfs2_init_locked_inode, &args);
	/* inode was *not* in the inode cache. 2.6.x requires
	 * us to do our own read_inode call and unlock it
	 * afterwards. */
	if (inode == NULL) {
		inode = ERR_PTR(-ENOMEM);
		mlog_errno(PTR_ERR(inode));
		goto bail;
	}
160
161
162
163
164
	if (inode->i_state & I_NEW) {
		mlog(0, "Inode was not in inode cache, reading it.\n");
		ocfs2_read_locked_inode(inode, &args);
		unlock_new_inode(inode);
	}
165
166
167
168
169
170
171
172
	if (is_bad_inode(inode)) {
		iput(inode);
		inode = ERR_PTR(-ESTALE);
		goto bail;
	}

bail:
	if (!IS_ERR(inode)) {
173
174
		mlog(0, "returning inode with number %llu\n",
		     (unsigned long long)OCFS2_I(inode)->ip_blkno);
Tao Ma's avatar
Tao Ma committed
175
		mlog(0, "inode %p\n", inode);
176
	}
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194

	return inode;
}


/*
 * here's how inodes get read from disk:
 * iget5_locked -> find_actor -> OCFS2_FIND_ACTOR
 * found? : return the in-memory inode
 * not found? : get_new_inode -> OCFS2_INIT_LOCKED_INODE
 */

static int ocfs2_find_actor(struct inode *inode, void *opaque)
{
	struct ocfs2_find_inode_args *args = NULL;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	int ret = 0;

Tao Ma's avatar
Tao Ma committed
195
	mlog(0, "(0x%p, %lu, 0x%p)\n", inode, inode->i_ino, opaque);
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216

	args = opaque;

	mlog_bug_on_msg(!inode, "No inode in find actor!\n");

	if (oi->ip_blkno != args->fi_blkno)
		goto bail;

	ret = 1;
bail:
	return ret;
}

/*
 * initialize the new inode, but don't do anything that would cause
 * us to sleep.
 * return 0 on success, 1 on failure
 */
static int ocfs2_init_locked_inode(struct inode *inode, void *opaque)
{
	struct ocfs2_find_inode_args *args = opaque;
Jan Kara's avatar
Jan Kara committed
217
218
	static struct lock_class_key ocfs2_quota_ip_alloc_sem_key,
				     ocfs2_file_ip_alloc_sem_key;
219

Tao Ma's avatar
Tao Ma committed
220
	mlog(0, "inode = %p, opaque = %p\n", inode, opaque);
221
222
223

	inode->i_ino = args->fi_ino;
	OCFS2_I(inode)->ip_blkno = args->fi_blkno;
224
225
226
	if (args->fi_sysfile_type != 0)
		lockdep_set_class(&inode->i_mutex,
			&ocfs2_sysfile_lock_key[args->fi_sysfile_type]);
Jan Kara's avatar
Jan Kara committed
227
228
229
230
231
232
233
234
235
	if (args->fi_sysfile_type == USER_QUOTA_SYSTEM_INODE ||
	    args->fi_sysfile_type == GROUP_QUOTA_SYSTEM_INODE ||
	    args->fi_sysfile_type == LOCAL_USER_QUOTA_SYSTEM_INODE ||
	    args->fi_sysfile_type == LOCAL_GROUP_QUOTA_SYSTEM_INODE)
		lockdep_set_class(&OCFS2_I(inode)->ip_alloc_sem,
				  &ocfs2_quota_ip_alloc_sem_key);
	else
		lockdep_set_class(&OCFS2_I(inode)->ip_alloc_sem,
				  &ocfs2_file_ip_alloc_sem_key);
236
237
238
239

	return 0;
}

240
241
void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
			  int create_ino)
242
243
244
{
	struct super_block *sb;
	struct ocfs2_super *osb;
Mark Fasheh's avatar
Mark Fasheh committed
245
	int use_plocks = 1;
246

Tao Ma's avatar
Tao Ma committed
247
248
	mlog(0, "(0x%p, size:%llu)\n", inode,
	     (unsigned long long)le64_to_cpu(fe->i_size));
249
250
251
252

	sb = inode->i_sb;
	osb = OCFS2_SB(sb);

Mark Fasheh's avatar
Mark Fasheh committed
253
254
255
256
	if ((osb->s_mount_opt & OCFS2_MOUNT_LOCALFLOCKS) ||
	    ocfs2_mount_local(osb) || !ocfs2_stack_supports_plocks())
		use_plocks = 0;

257
258
259
260
261
262
263
264
265
266
	/*
	 * These have all been checked by ocfs2_read_inode_block() or set
	 * by ocfs2_mknod_locked(), so a failure is a code bug.
	 */
	BUG_ON(!OCFS2_IS_VALID_DINODE(fe));  /* This means that read_inode
						cannot create a superblock
						inode today.  change if
						that is needed. */
	BUG_ON(!(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL)));
	BUG_ON(le32_to_cpu(fe->i_fs_generation) != osb->fs_generation);
267
268


269
270
	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
	OCFS2_I(inode)->ip_attr = le32_to_cpu(fe->i_attr);
271
	OCFS2_I(inode)->ip_dyn_features = le16_to_cpu(fe->i_dyn_features);
272

273
274
275
276
277
278
279
280
281
282
283
	inode->i_version = 1;
	inode->i_generation = le32_to_cpu(fe->i_generation);
	inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev));
	inode->i_mode = le16_to_cpu(fe->i_mode);
	inode->i_uid = le32_to_cpu(fe->i_uid);
	inode->i_gid = le32_to_cpu(fe->i_gid);

	/* Fast symlinks will have i_size but no allocated clusters. */
	if (S_ISLNK(inode->i_mode) && !fe->i_clusters)
		inode->i_blocks = 0;
	else
284
		inode->i_blocks = ocfs2_inode_sector_count(inode);
285
286
287
288
289
290
291
292
293
294
	inode->i_mapping->a_ops = &ocfs2_aops;
	inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime);
	inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec);
	inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime);
	inode->i_mtime.tv_nsec = le32_to_cpu(fe->i_mtime_nsec);
	inode->i_ctime.tv_sec = le64_to_cpu(fe->i_ctime);
	inode->i_ctime.tv_nsec = le32_to_cpu(fe->i_ctime_nsec);

	if (OCFS2_I(inode)->ip_blkno != le64_to_cpu(fe->i_blkno))
		mlog(ML_ERROR,
295
296
		     "ip_blkno %llu != i_blkno %llu!\n",
		     (unsigned long long)OCFS2_I(inode)->ip_blkno,
297
		     (unsigned long long)le64_to_cpu(fe->i_blkno));
298

Mark Fasheh's avatar
Mark Fasheh committed
299
	inode->i_nlink = ocfs2_read_links_count(fe);
300

301
	if (fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL)) {
302
		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SYSTEM_FILE;
303
304
		inode->i_flags |= S_NOQUOTA;
	}
305

306
307
308
309
310
	if (fe->i_flags & cpu_to_le32(OCFS2_LOCAL_ALLOC_FL)) {
		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP;
		mlog(0, "local alloc inode: i_ino=%lu\n", inode->i_ino);
	} else if (fe->i_flags & cpu_to_le32(OCFS2_BITMAP_FL)) {
		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP;
311
312
	} else if (fe->i_flags & cpu_to_le32(OCFS2_QUOTA_FL)) {
		inode->i_flags |= S_NOQUOTA;
313
314
315
316
317
318
319
320
321
	} else if (fe->i_flags & cpu_to_le32(OCFS2_SUPER_BLOCK_FL)) {
		mlog(0, "superblock inode: i_ino=%lu\n", inode->i_ino);
		/* we can't actually hit this as read_inode can't
		 * handle superblocks today ;-) */
		BUG();
	}

	switch (inode->i_mode & S_IFMT) {
	    case S_IFREG:
Mark Fasheh's avatar
Mark Fasheh committed
322
323
324
325
		    if (use_plocks)
			    inode->i_fop = &ocfs2_fops;
		    else
			    inode->i_fop = &ocfs2_fops_no_plocks;
326
327
328
329
330
		    inode->i_op = &ocfs2_file_iops;
		    i_size_write(inode, le64_to_cpu(fe->i_size));
		    break;
	    case S_IFDIR:
		    inode->i_op = &ocfs2_dir_iops;
Mark Fasheh's avatar
Mark Fasheh committed
331
332
333
334
		    if (use_plocks)
			    inode->i_fop = &ocfs2_dops;
		    else
			    inode->i_fop = &ocfs2_dops_no_plocks;
335
		    i_size_write(inode, le64_to_cpu(fe->i_size));
Goldwyn Rodrigues's avatar
Goldwyn Rodrigues committed
336
		    OCFS2_I(inode)->ip_dir_lock_gen = 1;
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
		    break;
	    case S_IFLNK:
		    if (ocfs2_inode_is_fast_symlink(inode))
			inode->i_op = &ocfs2_fast_symlink_inode_operations;
		    else
			inode->i_op = &ocfs2_symlink_inode_operations;
		    i_size_write(inode, le64_to_cpu(fe->i_size));
		    break;
	    default:
		    inode->i_op = &ocfs2_special_file_iops;
		    init_special_inode(inode, inode->i_mode,
				       inode->i_rdev);
		    break;
	}

352
353
354
355
356
357
358
359
360
	if (create_ino) {
		inode->i_ino = ino_from_blkno(inode->i_sb,
			       le64_to_cpu(fe->i_blkno));

		/*
		 * If we ever want to create system files from kernel,
		 * the generation argument to
		 * ocfs2_inode_lock_res_init() will have to change.
		 */
361
		BUG_ON(le32_to_cpu(fe->i_flags) & OCFS2_SYSTEM_FL);
362

363
		ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_inode_lockres,
364
					  OCFS2_LOCK_TYPE_META, 0, inode);
Tiger Yang's avatar
Tiger Yang committed
365
366
367

		ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_open_lockres,
					  OCFS2_LOCK_TYPE_OPEN, 0, inode);
368
369
	}

370
	ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_rw_lockres,
371
372
373
				  OCFS2_LOCK_TYPE_RW, inode->i_generation,
				  inode);

Herbert Poetzl's avatar
Herbert Poetzl committed
374
375
	ocfs2_set_inode_flags(inode);

376
377
	OCFS2_I(inode)->ip_last_used_slot = 0;
	OCFS2_I(inode)->ip_last_used_group = 0;
378
379
380
381

	if (S_ISDIR(inode->i_mode))
		ocfs2_resv_set_type(&OCFS2_I(inode)->ip_la_data_resv,
				    OCFS2_RESV_FLAG_DIR);
382
383
384
385
386
387
388
389
390
}

static int ocfs2_read_locked_inode(struct inode *inode,
				   struct ocfs2_find_inode_args *args)
{
	struct super_block *sb;
	struct ocfs2_super *osb;
	struct ocfs2_dinode *fe;
	struct buffer_head *bh = NULL;
391
392
	int status, can_lock;
	u32 generation = 0;
393

Tao Ma's avatar
Tao Ma committed
394
	mlog(0, "(0x%p, 0x%p)\n", inode, args);
395
396
397
398

	status = -EINVAL;
	if (inode == NULL || inode->i_sb == NULL) {
		mlog(ML_ERROR, "bad inode\n");
399
		return status;
400
401
402
403
404
405
406
	}
	sb = inode->i_sb;
	osb = OCFS2_SB(sb);

	if (!args) {
		mlog(ML_ERROR, "bad inode args\n");
		make_bad_inode(inode);
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
		return status;
	}

	/*
	 * To improve performance of cold-cache inode stats, we take
	 * the cluster lock here if possible.
	 *
	 * Generally, OCFS2 never trusts the contents of an inode
	 * unless it's holding a cluster lock, so taking it here isn't
	 * a correctness issue as much as it is a performance
	 * improvement.
	 *
	 * There are three times when taking the lock is not a good idea:
	 *
	 * 1) During startup, before we have initialized the DLM.
	 *
	 * 2) If we are reading certain system files which never get
	 *    cluster locks (local alloc, truncate log).
	 *
	 * 3) If the process doing the iget() is responsible for
	 *    orphan dir recovery. We're holding the orphan dir lock and
	 *    can get into a deadlock with another process on another
	 *    node in ->delete_inode().
	 *
	 * #1 and #2 can be simply solved by never taking the lock
	 * here for system files (which are the only type we read
	 * during mount). It's a heavier approach, but our main
434
	 * concern is user-accessible files anyway.
435
436
437
438
439
	 *
	 * #3 works itself out because we'll eventually take the
	 * cluster lock before trusting anything anyway.
	 */
	can_lock = !(args->fi_flags & OCFS2_FI_FLAG_SYSFILE)
Tiger Yang's avatar
Tiger Yang committed
440
		&& !(args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY)
Sunil Mushran's avatar
Sunil Mushran committed
441
		&& !ocfs2_mount_local(osb);
442
443
444
445
446
447
448
449
450
451
452

	/*
	 * To maintain backwards compatibility with older versions of
	 * ocfs2-tools, we still store the generation value for system
	 * files. The only ones that actually matter to userspace are
	 * the journals, but it's easier and inexpensive to just flag
	 * all system files similarly.
	 */
	if (args->fi_flags & OCFS2_FI_FLAG_SYSFILE)
		generation = osb->fs_generation;

453
	ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_inode_lockres,
454
455
456
				  OCFS2_LOCK_TYPE_META,
				  generation, inode);

Tiger Yang's avatar
Tiger Yang committed
457
458
459
460
	ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_open_lockres,
				  OCFS2_LOCK_TYPE_OPEN,
				  0, inode);

461
	if (can_lock) {
Tiger Yang's avatar
Tiger Yang committed
462
463
464
465
466
467
		status = ocfs2_open_lock(inode);
		if (status) {
			make_bad_inode(inode);
			mlog_errno(status);
			return status;
		}
468
		status = ocfs2_inode_lock(inode, NULL, 0);
469
470
471
472
473
		if (status) {
			make_bad_inode(inode);
			mlog_errno(status);
			return status;
		}
474
475
	}

Tiger Yang's avatar
Tiger Yang committed
476
477
478
	if (args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY) {
		status = ocfs2_try_open_lock(inode, 0);
		if (status) {
479
			make_bad_inode(inode);
Tiger Yang's avatar
Tiger Yang committed
480
481
482
483
			return status;
		}
	}

484
485
486
487
	if (can_lock) {
		status = ocfs2_read_inode_block_full(inode, &bh,
						     OCFS2_BH_IGNORE_CACHE);
	} else {
488
		status = ocfs2_read_blocks_sync(osb, args->fi_blkno, 1, &bh);
489
490
491
492
493
		/*
		 * If buffer is in jbd, then its checksum may not have been
		 * computed as yet.
		 */
		if (!status && !buffer_jbd(bh))
494
495
			status = ocfs2_validate_inode_block(osb->sb, bh);
	}
496
497
498
499
500
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

501
	status = -EINVAL;
502
503
	fe = (struct ocfs2_dinode *) bh->b_data;

504
505
506
507
508
509
510
511
512
	/*
	 * This is a code bug. Right now the caller needs to
	 * understand whether it is asking for a system file inode or
	 * not so the proper lock names can be built.
	 */
	mlog_bug_on_msg(!!(fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL)) !=
			!!(args->fi_flags & OCFS2_FI_FLAG_SYSFILE),
			"Inode %llu: system file state is ambigous\n",
			(unsigned long long)args->fi_blkno);
513
514
515

	if (S_ISCHR(le16_to_cpu(fe->i_mode)) ||
	    S_ISBLK(le16_to_cpu(fe->i_mode)))
516
		inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev));
517

518
	ocfs2_populate_inode(inode, fe, 0);
519
520
521
522
523
524

	BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno));

	status = 0;

bail:
525
	if (can_lock)
526
		ocfs2_inode_unlock(inode, 0);
527
528
529
530

	if (status < 0)
		make_bad_inode(inode);

531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
	if (args && bh)
		brelse(bh);

	return status;
}

void ocfs2_sync_blockdev(struct super_block *sb)
{
	sync_blockdev(sb->s_bdev);
}

static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
				     struct inode *inode,
				     struct buffer_head *fe_bh)
{
	int status = 0;
	struct ocfs2_dinode *fe;
548
	handle_t *handle = NULL;
549
550
551

	fe = (struct ocfs2_dinode *) fe_bh->b_data;

552
553
554
555
	/*
	 * This check will also skip truncate of inodes with inline
	 * data and fast symlinks.
	 */
556
	if (fe->i_clusters) {
Joel Becker's avatar
Joel Becker committed
557
558
559
		if (ocfs2_should_order_data(inode))
			ocfs2_begin_ordered_truncate(inode, 0);

560
561
562
		handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
		if (IS_ERR(handle)) {
			status = PTR_ERR(handle);
563
			handle = NULL;
564
565
566
567
			mlog_errno(status);
			goto out;
		}

568
569
		status = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
						 fe_bh,
570
						 OCFS2_JOURNAL_ACCESS_WRITE);
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
		if (status < 0) {
			mlog_errno(status);
			goto out;
		}

		i_size_write(inode, 0);

		status = ocfs2_mark_inode_dirty(handle, inode, fe_bh);
		if (status < 0) {
			mlog_errno(status);
			goto out;
		}

		ocfs2_commit_trans(osb, handle);
		handle = NULL;

587
		status = ocfs2_commit_truncate(osb, inode, fe_bh);
588
589
590
591
		if (status < 0) {
			mlog_errno(status);
			goto out;
		}
592
593
	}

594
595
596
out:
	if (handle)
		ocfs2_commit_trans(osb, handle);
597
598
599
600
601
602
603
604
605
606
607
	return status;
}

static int ocfs2_remove_inode(struct inode *inode,
			      struct buffer_head *di_bh,
			      struct inode *orphan_dir_inode,
			      struct buffer_head *orphan_dir_bh)
{
	int status;
	struct inode *inode_alloc_inode = NULL;
	struct buffer_head *inode_alloc_bh = NULL;
608
	handle_t *handle;
609
610
611
612
613
614
615
616
617
618
619
620
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;

	inode_alloc_inode =
		ocfs2_get_system_file_inode(osb, INODE_ALLOC_SYSTEM_INODE,
					    le16_to_cpu(di->i_suballoc_slot));
	if (!inode_alloc_inode) {
		status = -EEXIST;
		mlog_errno(status);
		goto bail;
	}

621
	mutex_lock(&inode_alloc_inode->i_mutex);
622
	status = ocfs2_inode_lock(inode_alloc_inode, &inode_alloc_bh, 1);
623
	if (status < 0) {
624
		mutex_unlock(&inode_alloc_inode->i_mutex);
625
626
627
628
629

		mlog_errno(status);
		goto bail;
	}

630
	handle = ocfs2_start_trans(osb, OCFS2_DELETE_INODE_CREDITS +
631
				   ocfs2_quota_trans_credits(inode->i_sb));
632
633
634
635
636
637
	if (IS_ERR(handle)) {
		status = PTR_ERR(handle);
		mlog_errno(status);
		goto bail_unlock;
	}

638
639
640
641
642
643
644
	if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) {
		status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode,
					  orphan_dir_bh);
		if (status < 0) {
			mlog_errno(status);
			goto bail_commit;
		}
645
646
647
	}

	/* set the inodes dtime */
648
	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
649
					 OCFS2_JOURNAL_ACCESS_WRITE);
650
651
652
653
654
655
	if (status < 0) {
		mlog_errno(status);
		goto bail_commit;
	}

	di->i_dtime = cpu_to_le64(CURRENT_TIME.tv_sec);
656
	di->i_flags &= cpu_to_le32(~(OCFS2_VALID_FL | OCFS2_ORPHANED_FL));
657
	ocfs2_journal_dirty(handle, di_bh);
658

659
	ocfs2_remove_from_cache(INODE_CACHE(inode), di_bh);
660
	dquot_free_inode(inode);
661
662
663
664
665
666
667

	status = ocfs2_free_dinode(handle, inode_alloc_inode,
				   inode_alloc_bh, di);
	if (status < 0)
		mlog_errno(status);

bail_commit:
668
	ocfs2_commit_trans(osb, handle);
669
bail_unlock:
670
	ocfs2_inode_unlock(inode_alloc_inode, 1);
671
	mutex_unlock(&inode_alloc_inode->i_mutex);
672
673
674
675
676
677
678
	brelse(inode_alloc_bh);
bail:
	iput(inode_alloc_inode);

	return status;
}

679
/*
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
 * Serialize with orphan dir recovery. If the process doing
 * recovery on this orphan dir does an iget() with the dir
 * i_mutex held, we'll deadlock here. Instead we detect this
 * and exit early - recovery will wipe this inode for us.
 */
static int ocfs2_check_orphan_recovery_state(struct ocfs2_super *osb,
					     int slot)
{
	int ret = 0;

	spin_lock(&osb->osb_lock);
	if (ocfs2_node_map_test_bit(osb, &osb->osb_recovering_orphan_dirs, slot)) {
		mlog(0, "Recovery is happening on orphan dir %d, will skip "
		     "this inode\n", slot);
		ret = -EDEADLK;
		goto out;
	}
	/* This signals to the orphan recovery process that it should
	 * wait for us to handle the wipe. */
	osb->osb_orphan_wipes[slot]++;
out:
	spin_unlock(&osb->osb_lock);
	return ret;
}

static void ocfs2_signal_wipe_completion(struct ocfs2_super *osb,
					 int slot)
{
	spin_lock(&osb->osb_lock);
	osb->osb_orphan_wipes[slot]--;
	spin_unlock(&osb->osb_lock);

	wake_up(&osb->osb_wipe_event);
}

715
716
717
static int ocfs2_wipe_inode(struct inode *inode,
			    struct buffer_head *di_bh)
{
718
	int status, orphaned_slot = -1;
719
720
721
	struct inode *orphan_dir_inode = NULL;
	struct buffer_head *orphan_dir_bh = NULL;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
722
	struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;
723

724
725
	if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) {
		orphaned_slot = le16_to_cpu(di->i_orphaned_slot);
726

727
728
729
		status = ocfs2_check_orphan_recovery_state(osb, orphaned_slot);
		if (status)
			return status;
730

731
732
733
734
735
736
737
738
		orphan_dir_inode = ocfs2_get_system_file_inode(osb,
							       ORPHAN_DIR_SYSTEM_INODE,
							       orphaned_slot);
		if (!orphan_dir_inode) {
			status = -EEXIST;
			mlog_errno(status);
			goto bail;
		}
739

740
741
742
743
744
745
746
		/* Lock the orphan dir. The lock will be held for the entire
		 * delete_inode operation. We do this now to avoid races with
		 * recovery completion on other nodes. */
		mutex_lock(&orphan_dir_inode->i_mutex);
		status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
		if (status < 0) {
			mutex_unlock(&orphan_dir_inode->i_mutex);
747

748
749
750
			mlog_errno(status);
			goto bail;
		}
751
752
753
	}

	/* we do this while holding the orphan dir lock because we
754
755
	 * don't want recovery being run from another node to try an
	 * inode delete underneath us -- this will result in two nodes
756
757
758
759
760
761
762
	 * truncating the same file! */
	status = ocfs2_truncate_for_delete(osb, inode, di_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail_unlock_dir;
	}

763
764
765
766
767
768
769
770
771
	/* Remove any dir index tree */
	if (S_ISDIR(inode->i_mode)) {
		status = ocfs2_dx_dir_truncate(inode, di_bh);
		if (status) {
			mlog_errno(status);
			goto bail_unlock_dir;
		}
	}

772
773
774
775
776
777
778
	/*Free extended attribute resources associated with this inode.*/
	status = ocfs2_xattr_remove(inode, di_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail_unlock_dir;
	}

779
780
781
782
783
784
	status = ocfs2_remove_refcount_tree(inode, di_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail_unlock_dir;
	}

785
786
787
788
789
790
	status = ocfs2_remove_inode(inode, di_bh, orphan_dir_inode,
				    orphan_dir_bh);
	if (status < 0)
		mlog_errno(status);

bail_unlock_dir:
791
792
793
	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)
		return status;

794
	ocfs2_inode_unlock(orphan_dir_inode, 1);
795
	mutex_unlock(&orphan_dir_inode->i_mutex);
796
797
798
	brelse(orphan_dir_bh);
bail:
	iput(orphan_dir_inode);
799
	ocfs2_signal_wipe_completion(osb, orphaned_slot);
800
801
802
803
804

	return status;
}

/* There is a series of simple checks that should be done before a
805
 * trylock is even considered. Encapsulate those in this function. */
806
807
808
809
810
811
812
813
814
815
816
817
818
static int ocfs2_inode_is_valid_to_delete(struct inode *inode)
{
	int ret = 0;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

	/* We shouldn't be getting here for the root directory
	 * inode.. */
	if (inode == osb->root_inode) {
		mlog(ML_ERROR, "Skipping delete of root inode.\n");
		goto bail;
	}

819
	/* If we're coming from downconvert_thread we can't go into our own
820
821
822
823
	 * voting [hello, deadlock city!], so unforuntately we just
	 * have to skip deleting this guy. That's OK though because
	 * the node who's doing the actual deleting should handle it
	 * anyway. */
824
	if (current == osb->dc_task) {
825
		mlog(0, "Skipping delete of %lu because we're currently "
826
		     "in downconvert\n", inode->i_ino);
827
828
829
830
831
832
833
834
		goto bail;
	}

	spin_lock(&oi->ip_lock);
	/* OCFS2 *never* deletes system files. This should technically
	 * never get here as system file inodes should always have a
	 * positive link count. */
	if (oi->ip_flags & OCFS2_INODE_SYSTEM_FILE) {
835
836
		mlog(ML_ERROR, "Skipping delete of system file %llu\n",
		     (unsigned long long)oi->ip_blkno);
837
838
839
		goto bail_unlock;
	}

840
841
842
	/* If we have allowd wipe of this inode for another node, it
	 * will be marked here so we can safely skip it. Recovery will
	 * cleanup any inodes we might inadvertantly skip here. */
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
	if (oi->ip_flags & OCFS2_INODE_SKIP_DELETE) {
		mlog(0, "Skipping delete of %lu because another node "
		     "has done this for us.\n", inode->i_ino);
		goto bail_unlock;
	}

	ret = 1;
bail_unlock:
	spin_unlock(&oi->ip_lock);
bail:
	return ret;
}

/* Query the cluster to determine whether we should wipe an inode from
 * disk or not.
 *
 * Requires the inode to have the cluster lock. */
static int ocfs2_query_inode_wipe(struct inode *inode,
				  struct buffer_head *di_bh,
				  int *wipe)
{
	int status = 0;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	struct ocfs2_dinode *di;

	*wipe = 0;

	/* While we were waiting for the cluster lock in
	 * ocfs2_delete_inode, another node might have asked to delete
	 * the inode. Recheck our flags to catch this. */
	if (!ocfs2_inode_is_valid_to_delete(inode)) {
874
875
		mlog(0, "Skipping delete of %llu because flags changed\n",
		     (unsigned long long)oi->ip_blkno);
876
877
878
879
880
881
		goto bail;
	}

	/* Now that we have an up to date inode, we can double check
	 * the link count. */
	if (inode->i_nlink) {
882
883
		mlog(0, "Skipping delete of %llu because nlink = %u\n",
		     (unsigned long long)oi->ip_blkno, inode->i_nlink);
884
885
886
887
888
		goto bail;
	}

	/* Do some basic inode verification... */
	di = (struct ocfs2_dinode *) di_bh->b_data;
889
890
	if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL)) &&
	    !(oi->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) {
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
		/*
		 * Inodes in the orphan dir must have ORPHANED_FL.  The only
		 * inodes that come back out of the orphan dir are reflink
		 * targets. A reflink target may be moved out of the orphan
		 * dir between the time we scan the directory and the time we
		 * process it. This would lead to HAS_REFCOUNT_FL being set but
		 * ORPHANED_FL not.
		 */
		if (di->i_dyn_features & cpu_to_le16(OCFS2_HAS_REFCOUNT_FL)) {
			mlog(0, "Reflinked inode %llu is no longer orphaned.  "
			     "it shouldn't be deleted\n",
			     (unsigned long long)oi->ip_blkno);
			goto bail;
		}

906
907
908
		/* for lack of a better error? */
		status = -EEXIST;
		mlog(ML_ERROR,
909
		     "Inode %llu (on-disk %llu) not orphaned! "
910
		     "Disk flags  0x%x, inode flags 0x%x\n",
911
		     (unsigned long long)oi->ip_blkno,
912
913
		     (unsigned long long)le64_to_cpu(di->i_blkno),
		     le32_to_cpu(di->i_flags), oi->ip_flags);
914
915
916
917
918
919
920
921
922
923
		goto bail;
	}

	/* has someone already deleted us?! baaad... */
	if (di->i_dtime) {
		status = -EEXIST;
		mlog_errno(status);
		goto bail;
	}

924
925
926
927
928
929
930
931
932
933
934
935
936
	/*
	 * This is how ocfs2 determines whether an inode is still live
	 * within the cluster. Every node takes a shared read lock on
	 * the inode open lock in ocfs2_read_locked_inode(). When we
	 * get to ->delete_inode(), each node tries to convert it's
	 * lock to an exclusive. Trylocks are serialized by the inode
	 * meta data lock. If the upconvert suceeds, we know the inode
	 * is no longer live and can be deleted.
	 *
	 * Though we call this with the meta data lock held, the
	 * trylock keeps us from ABBA deadlock.
	 */
	status = ocfs2_try_open_lock(inode, 1);
Tiger Yang's avatar
Tiger Yang committed
937
	if (status == -EAGAIN) {
938
		status = 0;
939
		mlog(0, "Skipping delete of %llu because it is in use on "
940
		     "other nodes\n", (unsigned long long)oi->ip_blkno);
941
942
943
944
945
946
947
		goto bail;
	}
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

Tiger Yang's avatar
Tiger Yang committed
948
949
950
951
	*wipe = 1;
	mlog(0, "Inode %llu is ok to wipe from orphan dir %u\n",
	     (unsigned long long)oi->ip_blkno,
	     le16_to_cpu(di->i_orphaned_slot));
952
953
954
955
956
957
958
959
960
961
962

bail:
	return status;
}

/* Support function for ocfs2_delete_inode. Will help us keep the
 * inode data in a consistent state for clear_inode. Always truncates
 * pages, optionally sync's them first. */
static void ocfs2_cleanup_delete_inode(struct inode *inode,
				       int sync_data)
{
963
964
	mlog(0, "Cleanup inode %llu, sync = %d\n",
	     (unsigned long long)OCFS2_I(inode)->ip_blkno, sync_data);
965
966
967
968
969
	if (sync_data)
		write_inode_now(inode, 1);
	truncate_inode_pages(&inode->i_data, 0);
}

Al Viro's avatar
Al Viro committed
970
static void ocfs2_delete_inode(struct inode *inode)
971
972
{
	int wipe, status;
973
	sigset_t oldset;
974
975
	struct buffer_head *di_bh = NULL;

Tao Ma's avatar
Tao Ma committed
976
	mlog(0, "(inode->i_ino = %lu)\n", inode->i_ino);
977

978
979
980
981
	/* When we fail in read_inode() we mark inode as bad. The second test
	 * catches the case when inode allocation fails before allocating
	 * a block for inode. */
	if (is_bad_inode(inode) || !OCFS2_I(inode)->ip_blkno) {
982
983
984
985
		mlog(0, "Skipping delete of bad inode\n");
		goto bail;
	}

986
	dquot_initialize(inode);
987

988
989
990
991
992
993
994
995
996
997
998
999
	if (!ocfs2_inode_is_valid_to_delete(inode)) {
		/* It's probably not necessary to truncate_inode_pages
		 * here but we do it for safety anyway (it will most
		 * likely be a no-op anyway) */
		ocfs2_cleanup_delete_inode(inode, 0);
		goto bail;
	}

	/* We want to block signals in delete_inode as the lock and
	 * messaging paths may return us -ERESTARTSYS. Which would
	 * cause us to exit early, resulting in inodes being orphaned
	 * forever. */
1000
	ocfs2_block_signals(&oldset);
1001

1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
	/*
	 * Synchronize us against ocfs2_get_dentry. We take this in
	 * shared mode so that all nodes can still concurrently
	 * process deletes.
	 */
	status = ocfs2_nfs_sync_lock(OCFS2_SB(inode->i_sb), 0);
	if (status < 0) {
		mlog(ML_ERROR, "getting nfs sync lock(PR) failed %d\n", status);
		ocfs2_cleanup_delete_inode(inode, 0);
		goto bail_unblock;
	}
1013
1014
	/* Lock down the inode. This gives us an up to date view of
	 * it's metadata (for verification), and allows us to
1015
	 * serialize delete_inode on multiple nodes.
1016
1017
1018
1019
1020
	 *
	 * Even though we might be doing a truncate, we don't take the
	 * allocation lock here as it won't be needed - nobody will
	 * have the file open.
	 */
1021
	status = ocfs2_inode_lock(inode, &di_bh, 1);
1022
1023
1024
1025
	if (status < 0) {
		if (status != -ENOENT)
			mlog_errno(status);
		ocfs2_cleanup_delete_inode(inode, 0);
1026
		goto bail_unlock_nfs_sync;
1027
1028
1029
1030
1031
1032
	}

	/* Query the cluster. This will be the final decision made
	 * before we go ahead and wipe the inode. */
	status = ocfs2_query_inode_wipe(inode, di_bh, &wipe);
	if (!wipe || status < 0) {
1033
		/* Error and remote inode busy both mean we won't be
1034
1035
1036
1037
1038
		 * removing the inode, so they take almost the same
		 * path. */
		if (status < 0)
			mlog_errno(status);

1039
1040
1041
		/* Someone in the cluster has disallowed a wipe of
		 * this inode, or it was never completely
		 * orphaned. Write out the pages and exit now. */
1042
1043
1044
1045
1046
1047
1048
1049
		ocfs2_cleanup_delete_inode(inode, 1);
		goto bail_unlock_inode;
	}

	ocfs2_cleanup_delete_inode(inode, 0);

	status = ocfs2_wipe_inode(inode, di_bh);
	if (status < 0) {
1050
1051
		if (status != -EDEADLK)
			mlog_errno(status);
1052
1053
1054
		goto bail_unlock_inode;
	}

1055
1056
1057
1058
1059
1060
1061
1062
1063
	/*
	 * Mark the inode as successfully deleted.
	 *
	 * This is important for ocfs2_clear_inode() as it will check
	 * this flag and skip any checkpointing work
	 *
	 * ocfs2_stuff_meta_lvb() also uses this flag to invalidate
	 * the LVB for other nodes.
	 */
1064
1065
1066
	OCFS2_I(inode)->ip_flags |= OCFS2_INODE_DELETED;

bail_unlock_inode:
1067
	ocfs2_inode_unlock(inode, 1);
1068
	brelse(di_bh);
1069
1070
1071
1072

bail_unlock_nfs_sync:
	ocfs2_nfs_sync_unlock(OCFS2_SB(inode->i_sb), 0);

1073
bail_unblock:
1074
	ocfs2_unblock_signals(&oldset);
1075
bail:
Tao Ma's avatar
Tao Ma committed
1076
	return;
1077
1078
}

Al Viro's avatar
Al Viro committed
1079
static void ocfs2_clear_inode(struct inode *inode)
1080
1081
1082
1083
{
	int status;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);

Al Viro's avatar
Al Viro committed
1084
	end_writeback(inode);
1085
1086
	mlog(0, "Clearing inode: %llu, nlink = %u\n",
	     (unsigned long long)OCFS2_I(inode)->ip_blkno, inode->i_nlink);
1087
1088
1089
1090

	mlog_bug_on_msg(OCFS2_SB(inode->i_sb) == NULL,
			"Inode=%lu\n", inode->i_ino);

1091
	dquot_drop(inode);
1092

1093
1094
	/* To preven remote deletes we hold open lock before, now it
	 * is time to unlock PR and EX open locks. */
Tiger Yang's avatar
Tiger Yang committed
1095
1096
	ocfs2_open_unlock(inode);

1097
	/* Do these before all the other work so that we don't bounce
1098
	 * the downconvert thread while waiting to destroy the locks. */
1099
	ocfs2_mark_lockres_freeing(&oi->ip_rw_lockres);
1100
	ocfs2_mark_lockres_freeing(&oi->ip_inode_lockres);
Tiger Yang's avatar
Tiger Yang committed
1101
	ocfs2_mark_lockres_freeing(&oi->ip_open_lockres);
1102

1103
1104
1105
1106
	ocfs2_resv_discard(&OCFS2_SB(inode->i_sb)->osb_la_resmap,
			   &oi->ip_la_data_resv);
	ocfs2_resv_init_once(&oi->ip_la_data_resv);

1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
	/* We very well may get a clear_inode before all an inodes
	 * metadata has hit disk. Of course, we can't drop any cluster
	 * locks until the journal has finished with it. The only
	 * exception here are successfully wiped inodes - their
	 * metadata can now be considered to be part of the system
	 * inodes from which it came. */
	if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED))
		ocfs2_checkpoint_inode(inode);

	mlog_bug_on_msg(!list_empty(&oi->ip_io_markers),
1117
1118
			"Clear inode of %llu, inode has io markers\n",
			(unsigned long long)oi->ip_blkno);
1119

Mark Fasheh's avatar
Mark Fasheh committed
1120
1121
	ocfs2_extent_map_trunc(inode, 0);

1122
1123
1124
1125
1126
	status = ocfs2_drop_inode_locks(inode);
	if (status < 0)
		mlog_errno(status);

	ocfs2_lock_res_free(&oi->ip_rw_lockres);
1127
	ocfs2_lock_res_free(&oi->ip_inode_lockres);
Tiger Yang's avatar
Tiger Yang committed
1128
	ocfs2_lock_res_free(&oi->ip_open_lockres);
1129

1130
	ocfs2_metadata_cache_exit(INODE_CACHE(inode));
1131

1132
	mlog_bug_on_msg(INODE_CACHE(inode)->ci_num_cached,
1133
			"Clear inode of %llu, inode has %u cache items\n",
1134
1135
			(unsigned long long)oi->ip_blkno,
			INODE_CACHE(inode)->ci_num_cached);
1136

1137
	mlog_bug_on_msg(!(INODE_CACHE(inode)->ci_flags & OCFS2_CACHE_FL_INLINE),
1138
1139
			"Clear inode of %llu, inode has a bad flag\n",
			(unsigned long long)oi->ip_blkno);
1140
1141

	mlog_bug_on_msg(spin_is_locked(&oi->ip_lock),
1142
1143
			"Clear inode of %llu, inode is locked\n",
			(unsigned long long)oi->ip_blkno);
1144

Mark Fasheh's avatar
Mark Fasheh committed
1145
	mlog_bug_on_msg(!mutex_trylock(&oi->ip_io_mutex),
1146
1147
			"Clear inode of %llu, io_mutex is locked\n",
			(unsigned long long)oi->ip_blkno);
Mark Fasheh's avatar
Mark Fasheh committed
1148
	mutex_unlock(&oi->ip_io_mutex);
1149
1150
1151
1152
1153
1154

	/*
	 * down_trylock() returns 0, down_write_trylock() returns 1
	 * kernel 1, world 0
	 */
	mlog_bug_on_msg(!down_write_trylock(&oi->ip_alloc_sem),
1155
1156
			"Clear inode of %llu, alloc_sem is locked\n",
			(unsigned long long)oi->ip_blkno);
1157
1158
1159
	up_write(&oi->ip_alloc_sem);

	mlog_bug_on_msg(oi->ip_open_count,
1160
1161
			"Clear inode of %llu has open count %d\n",
			(unsigned long long)oi->ip_blkno, oi->ip_open_count);
1162
1163

	/* Clear all other flags. */
1164
	oi->ip_flags = 0;
1165
1166
	oi->ip_dir_start_lookup = 0;
	oi->ip_blkno = 0ULL;
1167
1168
1169
1170
1171
1172

	/*
	 * ip_jinode is used to track txns against this inode. We ensure that
	 * the journal is flushed before journal shutdown. Thus it is safe to
	 * have inodes get cleaned up after journal shutdown.
	 */
Joel Becker's avatar
Joel Becker committed
1173
1174
	jbd2_journal_release_jbd_inode(OCFS2_SB(inode->i_sb)->journal->j_journal,
				       &oi->ip_jinode);
1175
1176
}

Al Viro's avatar
Al Viro committed
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
void ocfs2_evict_inode(struct inode *inode)
{
	if (!inode->i_nlink ||
	    (OCFS2_I(inode)->ip_flags & OCFS2_INODE_MAYBE_ORPHANED)) {
		ocfs2_delete_inode(inode);
	} else {
		truncate_inode_pages(&inode->i_data, 0);
	}
	ocfs2_clear_inode(inode);
}

1188
1189
1190
/* Called under inode_lock, with no more references on the
 * struct inode, so it's safe here to check the flags field
 * and to manipulate i_nlink without any other locks. */
1191
int ocfs2_drop_inode(struct inode *inode)
1192
1193
{
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
1194
	int res;
1195

1196
1197
	mlog(0, "Drop inode %llu, nlink = %u, ip_flags = 0x%x\n",
	     (unsigned long long)oi->ip_blkno, inode->i_nlink, oi->ip_flags);
1198

1199
	if (oi->ip_flags & OCFS2_INODE_MAYBE_ORPHANED)
1200
		res = 1;
1201
	else
1202
		res = generic_drop_inode(inode);
1203

1204
	return res;
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
}

/*
 * This is called from our getattr.
 */
int ocfs2_inode_revalidate(struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;
	int status = 0;

Tao Ma's avatar
Tao Ma committed
1215
1216
	mlog(0, "(inode = 0x%p, ino = %llu)\n", inode,
	     inode ? (unsigned long long)OCFS2_I(inode)->ip_blkno : 0ULL);
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232

	if (!inode) {
		mlog(0, "eep, no inode!\n");
		status = -ENOENT;
		goto bail;
	}

	spin_lock(&OCFS2_I(inode)->ip_lock);
	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) {
		spin_unlock(&OCFS2_I(inode)->ip_lock);
		mlog(0, "inode deleted!\n");
		status = -ENOENT;
		goto bail;
	}
	spin_unlock(&OCFS2_I(inode)->ip_lock);

1233
	/* Let ocfs2_inode_lock do the work of updating our struct
1234
	 * inode for us. */
1235
	status = ocfs2_inode_lock(inode, NULL, 0);
1236
1237
1238
1239
1240
	if (status < 0) {
		if (status != -ENOENT)
			mlog_errno(status);
		goto bail;
	}
1241
	ocfs2_inode_unlock(inode, 0);
1242
1243
1244
1245
1246
1247
1248
1249
1250
bail:
	return status;
}

/*
 * Updates a disk inode from a
 * struct inode.
 * Only takes ip_lock.
 */
1251
int ocfs2_mark_inode_dirty(handle_t *handle,
1252
1253
1254
1255
1256
1257
			   struct inode *inode,
			   struct buffer_head *bh)
{
	int status;
	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data;

Tao Ma's avatar
Tao Ma committed
1258
1259
	mlog(0, "(inode %llu)\n",
	     (unsigned long long)OCFS2_I(inode)->ip_blkno);
1260

1261
	status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
1262
					 OCFS2_JOURNAL_ACCESS_WRITE);
1263
1264
1265
1266
1267
1268
1269
	if (status < 0) {
		mlog_errno(status);
		goto leave;
	}

	spin_lock(&OCFS2_I(inode)->ip_lock);
	fe->i_clusters = cpu_to_le32(OCFS2_I(inode)->ip_clusters);
1270
	ocfs2_get_inode_flags(OCFS2_I(inode));
Herbert Poetzl's avatar
Herbert Poetzl committed
1271
	fe->i_attr = cpu_to_le32(OCFS2_I(inode)->ip_attr);
1272
	fe->i_dyn_features = cpu_to_le16(OCFS2_I(inode)->ip_dyn_features);
1273
1274
1275
	spin_unlock(&OCFS2_I(inode)->ip_lock);

	fe->i_size = cpu_to_le64(i_size_read(inode));
Mark Fasheh's avatar
Mark Fasheh committed
1276
	ocfs2_set_links_count(fe, inode->i_nlink);
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
	fe->i_uid = cpu_to_le32(inode->i_uid);
	fe->i_gid = cpu_to_le32(inode->i_gid);
	fe->i_mode = cpu_to_le16(inode->i_mode);
	fe->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
	fe->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
	fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
	fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
	fe->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
	fe->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);

1287
	ocfs2_journal_dirty(handle, bh);
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
leave:
	return status;
}

/*
 *
 * Updates a struct inode from a disk inode.
 * does no i/o, only takes ip_lock.
 */
void ocfs2_refresh_inode(struct inode *inode,
			 struct ocfs2_dinode *fe)
{
	spin_lock(&OCFS2_I(inode)->ip_lock);

	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
Herbert Poetzl's avatar
Herbert Poetzl committed
1303
	OCFS2_I(inode)->ip_attr = le32_to_cpu(fe->i_attr);
1304
	OCFS2_I(inode)->ip_dyn_features = le16_to_cpu(fe->i_dyn_features);
Herbert Poetzl's avatar
Herbert Poetzl committed
1305
	ocfs2_set_inode_flags(inode);
1306
	i_size_write(inode, le64_to_cpu(fe->i_size));
Mark Fasheh's avatar
Mark Fasheh committed
1307
	inode->i_nlink = ocfs2_read_links_count(fe);
1308
1309
1310
1311
1312
1313
	inode->i_uid = le32_to_cpu(fe->i_uid);
	inode->i_gid = le32_to_cpu(fe->i_gid);
	inode->i_mode = le16_to_cpu(fe->i_mode);
	if (S_ISLNK(inode->i_mode) && le32_to_cpu(fe->i_clusters) == 0)
		inode->i_blocks = 0;
	else
1314
		inode->i_blocks = ocfs2_inode_sector_count(inode);
1315
1316
1317
1318
1319
1320
1321
1322
1323
	inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime);
	inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec);
	inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime);
	inode->i_mtime.tv_nsec = le32_to_cpu(fe->i_mtime_nsec);
	inode->i_ctime.tv_sec = le64_to_cpu(fe->i_ctime);
	inode->i_ctime.tv_nsec = le32_to_cpu(fe->i_ctime_nsec);

	spin_unlock(&OCFS2_I(inode)->ip_lock);
}
1324
1325
1326
1327

int ocfs2_validate_inode_block(struct super_block *sb,
			       struct buffer_head *bh)
{
Joel Becker's avatar
Joel Becker committed
1328
	int rc;
1329
1330
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;

1331
1332
1333
	mlog(0, "Validating dinode %llu\n",
	     (unsigned long long)bh->b_blocknr);

1334
1335
	BUG_ON(!buffer_uptodate(bh));

Joel Becker's avatar
Joel Becker committed
1336
1337
1338
1339
1340
1341
	/*
	 * If the ecc fails, we return the error but otherwise
	 * leave the filesystem running.  We know any error is
	 * local to this block.
	 */
	rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &di->i_check);
1342
1343
1344
	if (rc) {
		mlog(ML_ERROR, "Checksum failed for dinode %llu\n",
		     (unsigned long long)bh->b_blocknr);
Joel Becker's avatar
Joel Becker committed
1345
		goto bail;
1346
	}
Joel Becker's avatar
Joel Becker committed
1347
1348
1349
1350
1351
1352
1353

	/*
	 * Errors after here are fatal.
	 */

	rc = -EINVAL;

1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383