[GFS2] Add gfs2meta filesystem
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / gfs2 / inode.c
CommitLineData
b3b94faa
DT
1/*
2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
4 *
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
7 * of the GNU General Public License v.2.
8 */
9
10#include <linux/sched.h>
11#include <linux/slab.h>
12#include <linux/spinlock.h>
13#include <linux/completion.h>
14#include <linux/buffer_head.h>
15#include <linux/posix_acl.h>
16#include <linux/sort.h>
5c676f6d 17#include <linux/gfs2_ondisk.h>
b3b94faa
DT
18#include <asm/semaphore.h>
19
20#include "gfs2.h"
5c676f6d
SW
21#include "lm_interface.h"
22#include "incore.h"
b3b94faa
DT
23#include "acl.h"
24#include "bmap.h"
25#include "dir.h"
26#include "eattr.h"
27#include "glock.h"
28#include "glops.h"
29#include "inode.h"
30#include "log.h"
31#include "meta_io.h"
32#include "ops_address.h"
33#include "ops_file.h"
34#include "ops_inode.h"
35#include "quota.h"
36#include "rgrp.h"
37#include "trans.h"
38#include "unlinked.h"
5c676f6d 39#include "util.h"
b3b94faa
DT
40
41/**
42 * inode_attr_in - Copy attributes from the dinode into the VFS inode
43 * @ip: The GFS2 inode (with embedded disk inode data)
44 * @inode: The Linux VFS inode
45 *
46 */
47
48static void inode_attr_in(struct gfs2_inode *ip, struct inode *inode)
49{
50 inode->i_ino = ip->i_num.no_formal_ino;
51
52 switch (ip->i_di.di_mode & S_IFMT) {
53 case S_IFBLK:
54 case S_IFCHR:
55 inode->i_rdev = MKDEV(ip->i_di.di_major, ip->i_di.di_minor);
56 break;
57 default:
58 inode->i_rdev = 0;
59 break;
60 };
61
62 inode->i_mode = ip->i_di.di_mode;
63 inode->i_nlink = ip->i_di.di_nlink;
64 inode->i_uid = ip->i_di.di_uid;
65 inode->i_gid = ip->i_di.di_gid;
66 i_size_write(inode, ip->i_di.di_size);
67 inode->i_atime.tv_sec = ip->i_di.di_atime;
68 inode->i_mtime.tv_sec = ip->i_di.di_mtime;
69 inode->i_ctime.tv_sec = ip->i_di.di_ctime;
70 inode->i_atime.tv_nsec = 0;
71 inode->i_mtime.tv_nsec = 0;
72 inode->i_ctime.tv_nsec = 0;
73 inode->i_blksize = PAGE_SIZE;
74 inode->i_blocks = ip->i_di.di_blocks <<
75 (ip->i_sbd->sd_sb.sb_bsize_shift - GFS2_BASIC_BLOCK_SHIFT);
76
77 if (ip->i_di.di_flags & GFS2_DIF_IMMUTABLE)
78 inode->i_flags |= S_IMMUTABLE;
79 else
80 inode->i_flags &= ~S_IMMUTABLE;
81
82 if (ip->i_di.di_flags & GFS2_DIF_APPENDONLY)
83 inode->i_flags |= S_APPEND;
84 else
85 inode->i_flags &= ~S_APPEND;
86}
87
88/**
89 * gfs2_inode_attr_in - Copy attributes from the dinode into the VFS inode
90 * @ip: The GFS2 inode (with embedded disk inode data)
91 *
92 */
93
94void gfs2_inode_attr_in(struct gfs2_inode *ip)
95{
96 struct inode *inode;
97
98 inode = gfs2_ip2v_lookup(ip);
99 if (inode) {
100 inode_attr_in(ip, inode);
101 iput(inode);
102 }
103}
104
105/**
106 * gfs2_inode_attr_out - Copy attributes from VFS inode into the dinode
107 * @ip: The GFS2 inode
108 *
109 * Only copy out the attributes that we want the VFS layer
110 * to be able to modify.
111 */
112
113void gfs2_inode_attr_out(struct gfs2_inode *ip)
114{
115 struct inode *inode = ip->i_vnode;
116
117 gfs2_assert_withdraw(ip->i_sbd,
118 (ip->i_di.di_mode & S_IFMT) == (inode->i_mode & S_IFMT));
119 ip->i_di.di_mode = inode->i_mode;
120 ip->i_di.di_uid = inode->i_uid;
121 ip->i_di.di_gid = inode->i_gid;
122 ip->i_di.di_atime = inode->i_atime.tv_sec;
123 ip->i_di.di_mtime = inode->i_mtime.tv_sec;
124 ip->i_di.di_ctime = inode->i_ctime.tv_sec;
125}
126
127/**
128 * gfs2_ip2v_lookup - Get the struct inode for a struct gfs2_inode
129 * @ip: the struct gfs2_inode to get the struct inode for
130 *
131 * Returns: A VFS inode, or NULL if none
132 */
133
134struct inode *gfs2_ip2v_lookup(struct gfs2_inode *ip)
135{
136 struct inode *inode = NULL;
137
138 gfs2_assert_warn(ip->i_sbd, test_bit(GIF_MIN_INIT, &ip->i_flags));
139
140 spin_lock(&ip->i_spin);
141 if (ip->i_vnode)
142 inode = igrab(ip->i_vnode);
143 spin_unlock(&ip->i_spin);
144
145 return inode;
146}
147
148/**
149 * gfs2_ip2v - Get/Create a struct inode for a struct gfs2_inode
150 * @ip: the struct gfs2_inode to get the struct inode for
151 *
152 * Returns: A VFS inode, or NULL if no mem
153 */
154
155struct inode *gfs2_ip2v(struct gfs2_inode *ip)
156{
157 struct inode *inode, *tmp;
158
159 inode = gfs2_ip2v_lookup(ip);
160 if (inode)
161 return inode;
162
163 tmp = new_inode(ip->i_sbd->sd_vfs);
164 if (!tmp)
165 return NULL;
166
167 inode_attr_in(ip, tmp);
168
169 if (S_ISREG(ip->i_di.di_mode)) {
170 tmp->i_op = &gfs2_file_iops;
171 tmp->i_fop = &gfs2_file_fops;
172 tmp->i_mapping->a_ops = &gfs2_file_aops;
173 } else if (S_ISDIR(ip->i_di.di_mode)) {
174 tmp->i_op = &gfs2_dir_iops;
175 tmp->i_fop = &gfs2_dir_fops;
176 } else if (S_ISLNK(ip->i_di.di_mode)) {
177 tmp->i_op = &gfs2_symlink_iops;
178 } else {
179 tmp->i_op = &gfs2_dev_iops;
180 init_special_inode(tmp, tmp->i_mode, tmp->i_rdev);
181 }
182
5c676f6d 183 tmp->u.generic_ip = NULL;
b3b94faa
DT
184
185 for (;;) {
186 spin_lock(&ip->i_spin);
187 if (!ip->i_vnode)
188 break;
189 inode = igrab(ip->i_vnode);
190 spin_unlock(&ip->i_spin);
191
192 if (inode) {
193 iput(tmp);
194 return inode;
195 }
196 yield();
197 }
198
199 inode = tmp;
200
201 gfs2_inode_hold(ip);
202 ip->i_vnode = inode;
5c676f6d 203 inode->u.generic_ip = ip;
b3b94faa
DT
204
205 spin_unlock(&ip->i_spin);
206
207 insert_inode_hash(inode);
208
209 return inode;
210}
211
212static int iget_test(struct inode *inode, void *opaque)
213{
5c676f6d 214 struct gfs2_inode *ip = inode->u.generic_ip;
b3b94faa
DT
215 struct gfs2_inum *inum = (struct gfs2_inum *)opaque;
216
217 if (ip && ip->i_num.no_addr == inum->no_addr)
218 return 1;
219
220 return 0;
221}
222
223struct inode *gfs2_iget(struct super_block *sb, struct gfs2_inum *inum)
224{
225 return ilookup5(sb, (unsigned long)inum->no_formal_ino,
226 iget_test, inum);
227}
228
229void gfs2_inode_min_init(struct gfs2_inode *ip, unsigned int type)
230{
231 spin_lock(&ip->i_spin);
232 if (!test_and_set_bit(GIF_MIN_INIT, &ip->i_flags)) {
233 ip->i_di.di_nlink = 1;
234 ip->i_di.di_mode = DT2IF(type);
235 }
236 spin_unlock(&ip->i_spin);
237}
238
239/**
240 * gfs2_inode_refresh - Refresh the incore copy of the dinode
241 * @ip: The GFS2 inode
242 *
243 * Returns: errno
244 */
245
246int gfs2_inode_refresh(struct gfs2_inode *ip)
247{
248 struct buffer_head *dibh;
249 int error;
250
251 error = gfs2_meta_inode_buffer(ip, &dibh);
252 if (error)
253 return error;
254
255 if (gfs2_metatype_check(ip->i_sbd, dibh, GFS2_METATYPE_DI)) {
256 brelse(dibh);
257 return -EIO;
258 }
259
260 spin_lock(&ip->i_spin);
261 gfs2_dinode_in(&ip->i_di, dibh->b_data);
262 set_bit(GIF_MIN_INIT, &ip->i_flags);
263 spin_unlock(&ip->i_spin);
264
265 brelse(dibh);
266
267 if (ip->i_num.no_addr != ip->i_di.di_num.no_addr) {
268 if (gfs2_consist_inode(ip))
269 gfs2_dinode_print(&ip->i_di);
270 return -EIO;
271 }
272 if (ip->i_num.no_formal_ino != ip->i_di.di_num.no_formal_ino)
273 return -ESTALE;
274
275 ip->i_vn = ip->i_gl->gl_vn;
276
277 return 0;
278}
279
280/**
281 * inode_create - create a struct gfs2_inode
282 * @i_gl: The glock covering the inode
283 * @inum: The inode number
284 * @io_gl: the iopen glock to acquire/hold (using holder in new gfs2_inode)
285 * @io_state: the state the iopen glock should be acquired in
286 * @ipp: pointer to put the returned inode in
287 *
288 * Returns: errno
289 */
290
c9fd4307 291static int inode_create(struct gfs2_glock *i_gl, const struct gfs2_inum *inum,
b3b94faa
DT
292 struct gfs2_glock *io_gl, unsigned int io_state,
293 struct gfs2_inode **ipp)
294{
295 struct gfs2_sbd *sdp = i_gl->gl_sbd;
296 struct gfs2_inode *ip;
297 int error = 0;
298
299 ip = kmem_cache_alloc(gfs2_inode_cachep, GFP_KERNEL);
300 if (!ip)
301 return -ENOMEM;
302 memset(ip, 0, sizeof(struct gfs2_inode));
303
304 ip->i_num = *inum;
305
306 atomic_set(&ip->i_count, 1);
307
308 ip->i_vn = i_gl->gl_vn - 1;
309
310 ip->i_gl = i_gl;
311 ip->i_sbd = sdp;
312
313 spin_lock_init(&ip->i_spin);
314 init_rwsem(&ip->i_rw_mutex);
315
316 ip->i_greedy = gfs2_tune_get(sdp, gt_greedy_default);
317
318 error = gfs2_glock_nq_init(io_gl,
319 io_state, GL_LOCAL_EXCL | GL_EXACT,
320 &ip->i_iopen_gh);
321 if (error)
322 goto fail;
323 ip->i_iopen_gh.gh_owner = NULL;
324
325 spin_lock(&io_gl->gl_spin);
326 gfs2_glock_hold(i_gl);
5c676f6d 327 io_gl->gl_object = i_gl;
b3b94faa
DT
328 spin_unlock(&io_gl->gl_spin);
329
330 gfs2_glock_hold(i_gl);
5c676f6d 331 i_gl->gl_object = ip;
b3b94faa
DT
332
333 atomic_inc(&sdp->sd_inode_count);
334
335 *ipp = ip;
336
337 return 0;
338
339 fail:
340 gfs2_meta_cache_flush(ip);
341 kmem_cache_free(gfs2_inode_cachep, ip);
342 *ipp = NULL;
343
344 return error;
345}
346
347/**
348 * gfs2_inode_get - Create or get a reference on an inode
349 * @i_gl: The glock covering the inode
350 * @inum: The inode number
351 * @create:
352 * @ipp: pointer to put the returned inode in
353 *
354 * Returns: errno
355 */
356
c9fd4307
SW
357int gfs2_inode_get(struct gfs2_glock *i_gl, const struct gfs2_inum *inum,
358 int create, struct gfs2_inode **ipp)
b3b94faa
DT
359{
360 struct gfs2_sbd *sdp = i_gl->gl_sbd;
361 struct gfs2_glock *io_gl;
362 int error = 0;
363
364 gfs2_glmutex_lock(i_gl);
365
5c676f6d 366 *ipp = i_gl->gl_object;
b3b94faa
DT
367 if (*ipp) {
368 error = -ESTALE;
369 if ((*ipp)->i_num.no_formal_ino != inum->no_formal_ino)
370 goto out;
371 atomic_inc(&(*ipp)->i_count);
372 error = 0;
373 goto out;
374 }
375
376 if (!create)
377 goto out;
378
379 error = gfs2_glock_get(sdp, inum->no_addr, &gfs2_iopen_glops,
380 CREATE, &io_gl);
381 if (!error) {
382 error = inode_create(i_gl, inum, io_gl, LM_ST_SHARED, ipp);
383 gfs2_glock_put(io_gl);
384 }
385
386 out:
387 gfs2_glmutex_unlock(i_gl);
388
389 return error;
390}
391
392void gfs2_inode_hold(struct gfs2_inode *ip)
393{
394 gfs2_assert(ip->i_sbd, atomic_read(&ip->i_count) > 0);
395 atomic_inc(&ip->i_count);
396}
397
398void gfs2_inode_put(struct gfs2_inode *ip)
399{
400 gfs2_assert(ip->i_sbd, atomic_read(&ip->i_count) > 0);
401 atomic_dec(&ip->i_count);
402}
403
404void gfs2_inode_destroy(struct gfs2_inode *ip)
405{
406 struct gfs2_sbd *sdp = ip->i_sbd;
407 struct gfs2_glock *io_gl = ip->i_iopen_gh.gh_gl;
408 struct gfs2_glock *i_gl = ip->i_gl;
409
410 gfs2_assert_warn(sdp, !atomic_read(&ip->i_count));
5c676f6d 411 gfs2_assert(sdp, io_gl->gl_object == i_gl);
b3b94faa
DT
412
413 spin_lock(&io_gl->gl_spin);
5c676f6d 414 io_gl->gl_object = NULL;
b3b94faa
DT
415 gfs2_glock_put(i_gl);
416 spin_unlock(&io_gl->gl_spin);
417
418 gfs2_glock_dq_uninit(&ip->i_iopen_gh);
419
420 gfs2_meta_cache_flush(ip);
421 kmem_cache_free(gfs2_inode_cachep, ip);
422
5c676f6d 423 i_gl->gl_object = NULL;
b3b94faa
DT
424 gfs2_glock_put(i_gl);
425
426 atomic_dec(&sdp->sd_inode_count);
427}
428
429static int dinode_dealloc(struct gfs2_inode *ip, struct gfs2_unlinked *ul)
430{
431 struct gfs2_sbd *sdp = ip->i_sbd;
432 struct gfs2_alloc *al;
433 struct gfs2_rgrpd *rgd;
434 int error;
435
436 if (ip->i_di.di_blocks != 1) {
437 if (gfs2_consist_inode(ip))
438 gfs2_dinode_print(&ip->i_di);
439 return -EIO;
440 }
441
442 al = gfs2_alloc_get(ip);
443
444 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
445 if (error)
446 goto out;
447
448 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
449 if (error)
450 goto out_qs;
451
452 rgd = gfs2_blk2rgrpd(sdp, ip->i_num.no_addr);
453 if (!rgd) {
454 gfs2_consist_inode(ip);
455 error = -EIO;
456 goto out_rindex_relse;
457 }
458
459 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
460 &al->al_rgd_gh);
461 if (error)
462 goto out_rindex_relse;
463
464 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_UNLINKED +
465 RES_STATFS + RES_QUOTA, 1);
466 if (error)
467 goto out_rg_gunlock;
468
469 gfs2_trans_add_gl(ip->i_gl);
470
471 gfs2_free_di(rgd, ip);
472
473 error = gfs2_unlinked_ondisk_rm(sdp, ul);
474
475 gfs2_trans_end(sdp);
476 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
477
478 out_rg_gunlock:
479 gfs2_glock_dq_uninit(&al->al_rgd_gh);
480
481 out_rindex_relse:
482 gfs2_glock_dq_uninit(&al->al_ri_gh);
483
484 out_qs:
485 gfs2_quota_unhold(ip);
486
487 out:
488 gfs2_alloc_put(ip);
489
490 return error;
491}
492
493/**
494 * inode_dealloc - Deallocate all on-disk blocks for an inode (dinode)
495 * @sdp: the filesystem
496 * @inum: the inode number to deallocate
497 * @io_gh: a holder for the iopen glock for this inode
498 *
499 * Returns: errno
500 */
501
502static int inode_dealloc(struct gfs2_sbd *sdp, struct gfs2_unlinked *ul,
503 struct gfs2_holder *io_gh)
504{
505 struct gfs2_inode *ip;
506 struct gfs2_holder i_gh;
507 int error;
508
509 error = gfs2_glock_nq_num(sdp,
510 ul->ul_ut.ut_inum.no_addr, &gfs2_inode_glops,
511 LM_ST_EXCLUSIVE, 0, &i_gh);
512 if (error)
513 return error;
514
515 /* We reacquire the iopen lock here to avoid a race with the NFS server
516 calling gfs2_read_inode() with the inode number of a inode we're in
517 the process of deallocating. And we can't keep our hold on the lock
518 from inode_dealloc_init() for deadlock reasons. */
519
520 gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY, io_gh);
521 error = gfs2_glock_nq(io_gh);
522 switch (error) {
523 case 0:
524 break;
525 case GLR_TRYFAILED:
526 error = 1;
527 default:
528 goto out;
529 }
530
5c676f6d 531 gfs2_assert_warn(sdp, !i_gh.gh_gl->gl_object);
b3b94faa
DT
532 error = inode_create(i_gh.gh_gl, &ul->ul_ut.ut_inum, io_gh->gh_gl,
533 LM_ST_EXCLUSIVE, &ip);
534
535 gfs2_glock_dq(io_gh);
536
537 if (error)
538 goto out;
539
540 error = gfs2_inode_refresh(ip);
541 if (error)
542 goto out_iput;
543
544 if (ip->i_di.di_nlink) {
545 if (gfs2_consist_inode(ip))
546 gfs2_dinode_print(&ip->i_di);
547 error = -EIO;
548 goto out_iput;
549 }
550
551 if (S_ISDIR(ip->i_di.di_mode) &&
552 (ip->i_di.di_flags & GFS2_DIF_EXHASH)) {
553 error = gfs2_dir_exhash_dealloc(ip);
554 if (error)
555 goto out_iput;
556 }
557
558 if (ip->i_di.di_eattr) {
559 error = gfs2_ea_dealloc(ip);
560 if (error)
561 goto out_iput;
562 }
563
564 if (!gfs2_is_stuffed(ip)) {
565 error = gfs2_file_dealloc(ip);
566 if (error)
567 goto out_iput;
568 }
569
570 error = dinode_dealloc(ip, ul);
571 if (error)
572 goto out_iput;
573
574 out_iput:
575 gfs2_glmutex_lock(i_gh.gh_gl);
576 gfs2_inode_put(ip);
577 gfs2_inode_destroy(ip);
578 gfs2_glmutex_unlock(i_gh.gh_gl);
579
580 out:
581 gfs2_glock_dq_uninit(&i_gh);
582
583 return error;
584}
585
586/**
587 * try_inode_dealloc - Try to deallocate an inode and all its blocks
588 * @sdp: the filesystem
589 *
590 * Returns: 0 on success, -errno on error, 1 on busy (inode open)
591 */
592
593static int try_inode_dealloc(struct gfs2_sbd *sdp, struct gfs2_unlinked *ul)
594{
595 struct gfs2_holder io_gh;
596 int error = 0;
597
598 gfs2_try_toss_inode(sdp, &ul->ul_ut.ut_inum);
599
600 error = gfs2_glock_nq_num(sdp,
601 ul->ul_ut.ut_inum.no_addr, &gfs2_iopen_glops,
602 LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB, &io_gh);
603 switch (error) {
604 case 0:
605 break;
606 case GLR_TRYFAILED:
607 return 1;
608 default:
609 return error;
610 }
611
612 gfs2_glock_dq(&io_gh);
613 error = inode_dealloc(sdp, ul, &io_gh);
614 gfs2_holder_uninit(&io_gh);
615
616 return error;
617}
618
619static int inode_dealloc_uninit(struct gfs2_sbd *sdp, struct gfs2_unlinked *ul)
620{
621 struct gfs2_rgrpd *rgd;
622 struct gfs2_holder ri_gh, rgd_gh;
623 int error;
624
625 error = gfs2_rindex_hold(sdp, &ri_gh);
626 if (error)
627 return error;
628
629 rgd = gfs2_blk2rgrpd(sdp, ul->ul_ut.ut_inum.no_addr);
630 if (!rgd) {
631 gfs2_consist(sdp);
632 error = -EIO;
633 goto out;
634 }
635
636 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &rgd_gh);
637 if (error)
638 goto out;
639
640 error = gfs2_trans_begin(sdp,
641 RES_RG_BIT + RES_UNLINKED + RES_STATFS,
642 0);
643 if (error)
644 goto out_gunlock;
645
646 gfs2_free_uninit_di(rgd, ul->ul_ut.ut_inum.no_addr);
647 gfs2_unlinked_ondisk_rm(sdp, ul);
648
649 gfs2_trans_end(sdp);
650
651 out_gunlock:
652 gfs2_glock_dq_uninit(&rgd_gh);
653 out:
654 gfs2_glock_dq_uninit(&ri_gh);
655
656 return error;
657}
658
659int gfs2_inode_dealloc(struct gfs2_sbd *sdp, struct gfs2_unlinked *ul)
660{
661 if (ul->ul_ut.ut_flags & GFS2_UTF_UNINIT)
662 return inode_dealloc_uninit(sdp, ul);
663 else
664 return try_inode_dealloc(sdp, ul);
665}
666
667/**
668 * gfs2_change_nlink - Change nlink count on inode
669 * @ip: The GFS2 inode
670 * @diff: The change in the nlink count required
671 *
672 * Returns: errno
673 */
674
675int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
676{
677 struct buffer_head *dibh;
678 uint32_t nlink;
679 int error;
680
681 nlink = ip->i_di.di_nlink + diff;
682
683 /* If we are reducing the nlink count, but the new value ends up being
684 bigger than the old one, we must have underflowed. */
685 if (diff < 0 && nlink > ip->i_di.di_nlink) {
686 if (gfs2_consist_inode(ip))
687 gfs2_dinode_print(&ip->i_di);
688 return -EIO;
689 }
690
691 error = gfs2_meta_inode_buffer(ip, &dibh);
692 if (error)
693 return error;
694
695 ip->i_di.di_nlink = nlink;
696 ip->i_di.di_ctime = get_seconds();
697
d4e9c4c3 698 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
b3b94faa
DT
699 gfs2_dinode_out(&ip->i_di, dibh->b_data);
700 brelse(dibh);
701
702 return 0;
703}
704
705/**
706 * gfs2_lookupi - Look up a filename in a directory and return its inode
707 * @d_gh: An initialized holder for the directory glock
708 * @name: The name of the inode to look for
709 * @is_root: If 1, ignore the caller's permissions
710 * @i_gh: An uninitialized holder for the new inode glock
711 *
712 * There will always be a vnode (Linux VFS inode) for the d_gh inode unless
713 * @is_root is true.
714 *
715 * Returns: errno
716 */
717
7359a19c
SW
718int gfs2_lookupi(struct inode *dir, struct qstr *name, int is_root,
719 struct inode **inodep)
b3b94faa 720{
c9fd4307 721 struct super_block *sb = dir->i_sb;
7359a19c 722 struct gfs2_inode *ipp;
5c676f6d 723 struct gfs2_inode *dip = dir->u.generic_ip;
b3b94faa
DT
724 struct gfs2_sbd *sdp = dip->i_sbd;
725 struct gfs2_holder d_gh;
726 struct gfs2_inum inum;
727 unsigned int type;
728 struct gfs2_glock *gl;
7359a19c
SW
729 int error = 0;
730
731 *inodep = NULL;
b3b94faa
DT
732
733 if (!name->len || name->len > GFS2_FNAMESIZE)
734 return -ENAMETOOLONG;
735
736 if (gfs2_filecmp(name, ".", 1) ||
c9fd4307 737 (gfs2_filecmp(name, "..", 2) && dir == sb->s_root->d_inode)) {
b3b94faa 738 gfs2_inode_hold(dip);
7359a19c
SW
739 ipp = dip;
740 goto done;
b3b94faa
DT
741 }
742
743 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
744 if (error)
745 return error;
746
747 if (!is_root) {
748 error = gfs2_repermission(dip->i_vnode, MAY_EXEC, NULL);
749 if (error)
750 goto out;
751 }
752
753 error = gfs2_dir_search(dip, name, &inum, &type);
754 if (error)
755 goto out;
756
757 error = gfs2_glock_get(sdp, inum.no_addr, &gfs2_inode_glops,
758 CREATE, &gl);
759 if (error)
760 goto out;
761
7359a19c 762 error = gfs2_inode_get(gl, &inum, CREATE, &ipp);
b3b94faa 763 if (!error)
7359a19c 764 gfs2_inode_min_init(ipp, type);
b3b94faa
DT
765
766 gfs2_glock_put(gl);
767
7359a19c 768out:
b3b94faa 769 gfs2_glock_dq_uninit(&d_gh);
7359a19c
SW
770done:
771 if (error == 0) {
772 *inodep = gfs2_ip2v(ipp);
773 if (!*inodep)
774 error = -ENOMEM;
775 gfs2_inode_put(ipp);
776 }
b3b94faa
DT
777 return error;
778}
779
780static int pick_formal_ino_1(struct gfs2_sbd *sdp, uint64_t *formal_ino)
781{
5c676f6d 782 struct gfs2_inode *ip = sdp->sd_ir_inode->u.generic_ip;
b3b94faa
DT
783 struct buffer_head *bh;
784 struct gfs2_inum_range ir;
785 int error;
786
787 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
788 if (error)
789 return error;
f55ab26a 790 mutex_lock(&sdp->sd_inum_mutex);
b3b94faa
DT
791
792 error = gfs2_meta_inode_buffer(ip, &bh);
793 if (error) {
f55ab26a 794 mutex_unlock(&sdp->sd_inum_mutex);
b3b94faa
DT
795 gfs2_trans_end(sdp);
796 return error;
797 }
798
799 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
800
801 if (ir.ir_length) {
802 *formal_ino = ir.ir_start++;
803 ir.ir_length--;
d4e9c4c3 804 gfs2_trans_add_bh(ip->i_gl, bh, 1);
b3b94faa
DT
805 gfs2_inum_range_out(&ir,
806 bh->b_data + sizeof(struct gfs2_dinode));
807 brelse(bh);
f55ab26a 808 mutex_unlock(&sdp->sd_inum_mutex);
b3b94faa
DT
809 gfs2_trans_end(sdp);
810 return 0;
811 }
812
813 brelse(bh);
814
f55ab26a 815 mutex_unlock(&sdp->sd_inum_mutex);
b3b94faa
DT
816 gfs2_trans_end(sdp);
817
818 return 1;
819}
820
821static int pick_formal_ino_2(struct gfs2_sbd *sdp, uint64_t *formal_ino)
822{
5c676f6d
SW
823 struct gfs2_inode *ip = sdp->sd_ir_inode->u.generic_ip;
824 struct gfs2_inode *m_ip = sdp->sd_inum_inode->u.generic_ip;
b3b94faa
DT
825 struct gfs2_holder gh;
826 struct buffer_head *bh;
827 struct gfs2_inum_range ir;
828 int error;
829
830 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
831 if (error)
832 return error;
833
834 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
835 if (error)
836 goto out;
f55ab26a 837 mutex_lock(&sdp->sd_inum_mutex);
b3b94faa
DT
838
839 error = gfs2_meta_inode_buffer(ip, &bh);
840 if (error)
841 goto out_end_trans;
842
843 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
844
845 if (!ir.ir_length) {
846 struct buffer_head *m_bh;
847 uint64_t x, y;
848
849 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
850 if (error)
851 goto out_brelse;
852
853 x = *(uint64_t *)(m_bh->b_data + sizeof(struct gfs2_dinode));
854 x = y = be64_to_cpu(x);
855 ir.ir_start = x;
856 ir.ir_length = GFS2_INUM_QUANTUM;
857 x += GFS2_INUM_QUANTUM;
858 if (x < y)
859 gfs2_consist_inode(m_ip);
860 x = cpu_to_be64(x);
d4e9c4c3 861 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
b3b94faa
DT
862 *(uint64_t *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = x;
863
864 brelse(m_bh);
865 }
866
867 *formal_ino = ir.ir_start++;
868 ir.ir_length--;
869
d4e9c4c3 870 gfs2_trans_add_bh(ip->i_gl, bh, 1);
b3b94faa
DT
871 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
872
873 out_brelse:
874 brelse(bh);
875
876 out_end_trans:
f55ab26a 877 mutex_unlock(&sdp->sd_inum_mutex);
b3b94faa
DT
878 gfs2_trans_end(sdp);
879
880 out:
881 gfs2_glock_dq_uninit(&gh);
882
883 return error;
884}
885
886static int pick_formal_ino(struct gfs2_sbd *sdp, uint64_t *inum)
887{
888 int error;
889
890 error = pick_formal_ino_1(sdp, inum);
891 if (error <= 0)
892 return error;
893
894 error = pick_formal_ino_2(sdp, inum);
895
896 return error;
897}
898
899/**
900 * create_ok - OK to create a new on-disk inode here?
901 * @dip: Directory in which dinode is to be created
902 * @name: Name of new dinode
903 * @mode:
904 *
905 * Returns: errno
906 */
907
908static int create_ok(struct gfs2_inode *dip, struct qstr *name,
909 unsigned int mode)
910{
911 int error;
912
913 error = gfs2_repermission(dip->i_vnode, MAY_WRITE | MAY_EXEC, NULL);
914 if (error)
915 return error;
916
917 /* Don't create entries in an unlinked directory */
918 if (!dip->i_di.di_nlink)
919 return -EPERM;
920
921 error = gfs2_dir_search(dip, name, NULL, NULL);
922 switch (error) {
923 case -ENOENT:
924 error = 0;
925 break;
926 case 0:
927 return -EEXIST;
928 default:
929 return error;
930 }
931
932 if (dip->i_di.di_entries == (uint32_t)-1)
933 return -EFBIG;
934 if (S_ISDIR(mode) && dip->i_di.di_nlink == (uint32_t)-1)
935 return -EMLINK;
936
937 return 0;
938}
939
940static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
941 unsigned int *uid, unsigned int *gid)
942{
943 if (dip->i_sbd->sd_args.ar_suiddir &&
944 (dip->i_di.di_mode & S_ISUID) &&
945 dip->i_di.di_uid) {
946 if (S_ISDIR(*mode))
947 *mode |= S_ISUID;
948 else if (dip->i_di.di_uid != current->fsuid)
949 *mode &= ~07111;
950 *uid = dip->i_di.di_uid;
951 } else
952 *uid = current->fsuid;
953
954 if (dip->i_di.di_mode & S_ISGID) {
955 if (S_ISDIR(*mode))
956 *mode |= S_ISGID;
957 *gid = dip->i_di.di_gid;
958 } else
959 *gid = current->fsgid;
960}
961
962static int alloc_dinode(struct gfs2_inode *dip, struct gfs2_unlinked *ul)
963{
964 struct gfs2_sbd *sdp = dip->i_sbd;
965 int error;
966
967 gfs2_alloc_get(dip);
968
969 dip->i_alloc.al_requested = RES_DINODE;
970 error = gfs2_inplace_reserve(dip);
971 if (error)
972 goto out;
973
974 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_UNLINKED +
975 RES_STATFS, 0);
976 if (error)
977 goto out_ipreserv;
978
979 ul->ul_ut.ut_inum.no_addr = gfs2_alloc_di(dip);
980
981 ul->ul_ut.ut_flags = GFS2_UTF_UNINIT;
982 error = gfs2_unlinked_ondisk_add(sdp, ul);
983
984 gfs2_trans_end(sdp);
985
986 out_ipreserv:
987 gfs2_inplace_release(dip);
988
989 out:
990 gfs2_alloc_put(dip);
991
992 return error;
993}
994
995/**
996 * init_dinode - Fill in a new dinode structure
997 * @dip: the directory this inode is being created in
998 * @gl: The glock covering the new inode
999 * @inum: the inode number
1000 * @mode: the file permissions
1001 * @uid:
1002 * @gid:
1003 *
1004 */
1005
1006static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
1007 struct gfs2_inum *inum, unsigned int mode,
1008 unsigned int uid, unsigned int gid)
1009{
1010 struct gfs2_sbd *sdp = dip->i_sbd;
b96ca4fa 1011 struct gfs2_dinode *di;
b3b94faa
DT
1012 struct buffer_head *dibh;
1013
1014 dibh = gfs2_meta_new(gl, inum->no_addr);
d4e9c4c3 1015 gfs2_trans_add_bh(gl, dibh, 1);
b3b94faa
DT
1016 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
1017 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
b96ca4fa
SW
1018 di = (struct gfs2_dinode *)dibh->b_data;
1019
2442a098
SW
1020 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
1021 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
b96ca4fa
SW
1022 di->di_mode = cpu_to_be32(mode);
1023 di->di_uid = cpu_to_be32(uid);
1024 di->di_gid = cpu_to_be32(gid);
1025 di->di_nlink = cpu_to_be32(0);
1026 di->di_size = cpu_to_be64(0);
1027 di->di_blocks = cpu_to_be64(1);
1028 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(get_seconds());
1029 di->di_major = di->di_minor = cpu_to_be32(0);
1030 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
1031 di->__pad[0] = di->__pad[1] = 0;
1032 di->di_flags = cpu_to_be32(0);
b3b94faa
DT
1033
1034 if (S_ISREG(mode)) {
1035 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
1036 gfs2_tune_get(sdp, gt_new_files_jdata))
b96ca4fa 1037 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
b3b94faa
DT
1038 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) ||
1039 gfs2_tune_get(sdp, gt_new_files_directio))
b96ca4fa 1040 di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO);
b3b94faa 1041 } else if (S_ISDIR(mode)) {
568f4c96
SW
1042 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
1043 GFS2_DIF_INHERIT_DIRECTIO);
1044 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
1045 GFS2_DIF_INHERIT_JDATA);
b3b94faa
DT
1046 }
1047
b96ca4fa
SW
1048 di->__pad1 = 0;
1049 di->di_height = cpu_to_be32(0);
1050 di->__pad2 = 0;
1051 di->__pad3 = 0;
1052 di->di_depth = cpu_to_be16(0);
1053 di->di_entries = cpu_to_be32(0);
1054 memset(&di->__pad4, 0, sizeof(di->__pad4));
1055 di->di_eattr = cpu_to_be64(0);
1056 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
1057
b3b94faa
DT
1058 brelse(dibh);
1059}
1060
1061static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
1062 unsigned int mode, struct gfs2_unlinked *ul)
1063{
1064 struct gfs2_sbd *sdp = dip->i_sbd;
1065 unsigned int uid, gid;
1066 int error;
1067
1068 munge_mode_uid_gid(dip, &mode, &uid, &gid);
1069
1070 gfs2_alloc_get(dip);
1071
1072 error = gfs2_quota_lock(dip, uid, gid);
1073 if (error)
1074 goto out;
1075
1076 error = gfs2_quota_check(dip, uid, gid);
1077 if (error)
1078 goto out_quota;
1079
1080 error = gfs2_trans_begin(sdp, RES_DINODE + RES_UNLINKED +
1081 RES_QUOTA, 0);
1082 if (error)
1083 goto out_quota;
1084
1085 ul->ul_ut.ut_flags = 0;
1086 error = gfs2_unlinked_ondisk_munge(sdp, ul);
1087
1088 init_dinode(dip, gl, &ul->ul_ut.ut_inum,
1089 mode, uid, gid);
1090
1091 gfs2_quota_change(dip, +1, uid, gid);
1092
1093 gfs2_trans_end(sdp);
1094
1095 out_quota:
1096 gfs2_quota_unlock(dip);
1097
1098 out:
1099 gfs2_alloc_put(dip);
1100
1101 return error;
1102}
1103
1104static int link_dinode(struct gfs2_inode *dip, struct qstr *name,
1105 struct gfs2_inode *ip, struct gfs2_unlinked *ul)
1106{
1107 struct gfs2_sbd *sdp = dip->i_sbd;
1108 struct gfs2_alloc *al;
1109 int alloc_required;
1110 struct buffer_head *dibh;
1111 int error;
1112
1113 al = gfs2_alloc_get(dip);
1114
1115 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
1116 if (error)
1117 goto fail;
1118
1119 error = gfs2_diradd_alloc_required(dip, name, &alloc_required);
1120 if (alloc_required) {
1121 error = gfs2_quota_check(dip, dip->i_di.di_uid,
1122 dip->i_di.di_gid);
1123 if (error)
1124 goto fail_quota_locks;
1125
1126 al->al_requested = sdp->sd_max_dirres;
1127
1128 error = gfs2_inplace_reserve(dip);
1129 if (error)
1130 goto fail_quota_locks;
1131
1132 error = gfs2_trans_begin(sdp,
1133 sdp->sd_max_dirres +
1134 al->al_rgd->rd_ri.ri_length +
1135 2 * RES_DINODE + RES_UNLINKED +
1136 RES_STATFS + RES_QUOTA, 0);
1137 if (error)
1138 goto fail_ipreserv;
1139 } else {
1140 error = gfs2_trans_begin(sdp,
1141 RES_LEAF +
1142 2 * RES_DINODE +
1143 RES_UNLINKED, 0);
1144 if (error)
1145 goto fail_quota_locks;
1146 }
1147
1148 error = gfs2_dir_add(dip, name, &ip->i_num, IF2DT(ip->i_di.di_mode));
1149 if (error)
1150 goto fail_end_trans;
1151
1152 error = gfs2_meta_inode_buffer(ip, &dibh);
1153 if (error)
1154 goto fail_end_trans;
1155 ip->i_di.di_nlink = 1;
d4e9c4c3 1156 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
b3b94faa
DT
1157 gfs2_dinode_out(&ip->i_di, dibh->b_data);
1158 brelse(dibh);
1159
1160 error = gfs2_unlinked_ondisk_rm(sdp, ul);
1161 if (error)
1162 goto fail_end_trans;
1163
1164 return 0;
1165
1166 fail_end_trans:
1167 gfs2_trans_end(sdp);
1168
1169 fail_ipreserv:
1170 if (dip->i_alloc.al_rgd)
1171 gfs2_inplace_release(dip);
1172
1173 fail_quota_locks:
1174 gfs2_quota_unlock(dip);
1175
1176 fail:
1177 gfs2_alloc_put(dip);
1178
1179 return error;
1180}
1181
1182/**
1183 * gfs2_createi - Create a new inode
1184 * @ghs: An array of two holders
1185 * @name: The name of the new file
1186 * @mode: the permissions on the new inode
1187 *
1188 * @ghs[0] is an initialized holder for the directory
1189 * @ghs[1] is the holder for the inode lock
1190 *
7359a19c 1191 * If the return value is not NULL, the glocks on both the directory and the new
b3b94faa
DT
1192 * file are held. A transaction has been started and an inplace reservation
1193 * is held, as well.
1194 *
7359a19c 1195 * Returns: An inode
b3b94faa
DT
1196 */
1197
568f4c96
SW
1198struct inode *gfs2_createi(struct gfs2_holder *ghs, struct qstr *name,
1199 unsigned int mode)
b3b94faa 1200{
7359a19c 1201 struct inode *inode;
5c676f6d 1202 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
b3b94faa
DT
1203 struct gfs2_sbd *sdp = dip->i_sbd;
1204 struct gfs2_unlinked *ul;
1205 struct gfs2_inode *ip;
1206 int error;
1207
1208 if (!name->len || name->len > GFS2_FNAMESIZE)
7359a19c 1209 return ERR_PTR(-ENAMETOOLONG);
b3b94faa
DT
1210
1211 error = gfs2_unlinked_get(sdp, &ul);
1212 if (error)
7359a19c 1213 return ERR_PTR(error);
b3b94faa
DT
1214
1215 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
1216 error = gfs2_glock_nq(ghs);
1217 if (error)
1218 goto fail;
1219
1220 error = create_ok(dip, name, mode);
1221 if (error)
1222 goto fail_gunlock;
1223
1224 error = pick_formal_ino(sdp, &ul->ul_ut.ut_inum.no_formal_ino);
1225 if (error)
1226 goto fail_gunlock;
1227
1228 error = alloc_dinode(dip, ul);
1229 if (error)
1230 goto fail_gunlock;
1231
1232 if (ul->ul_ut.ut_inum.no_addr < dip->i_num.no_addr) {
1233 gfs2_glock_dq(ghs);
1234
1235 error = gfs2_glock_nq_num(sdp,
1236 ul->ul_ut.ut_inum.no_addr,
1237 &gfs2_inode_glops,
1238 LM_ST_EXCLUSIVE, GL_SKIP,
1239 ghs + 1);
1240 if (error) {
1241 gfs2_unlinked_put(sdp, ul);
7359a19c 1242 return ERR_PTR(error);
b3b94faa
DT
1243 }
1244
1245 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
1246 error = gfs2_glock_nq(ghs);
1247 if (error) {
1248 gfs2_glock_dq_uninit(ghs + 1);
1249 gfs2_unlinked_put(sdp, ul);
7359a19c 1250 return ERR_PTR(error);
b3b94faa
DT
1251 }
1252
1253 error = create_ok(dip, name, mode);
1254 if (error)
1255 goto fail_gunlock2;
1256 } else {
1257 error = gfs2_glock_nq_num(sdp,
1258 ul->ul_ut.ut_inum.no_addr,
1259 &gfs2_inode_glops,
1260 LM_ST_EXCLUSIVE, GL_SKIP,
1261 ghs + 1);
1262 if (error)
1263 goto fail_gunlock;
1264 }
1265
1266 error = make_dinode(dip, ghs[1].gh_gl, mode, ul);
1267 if (error)
1268 goto fail_gunlock2;
1269
1270 error = gfs2_inode_get(ghs[1].gh_gl, &ul->ul_ut.ut_inum, CREATE, &ip);
1271 if (error)
1272 goto fail_gunlock2;
1273
1274 error = gfs2_inode_refresh(ip);
1275 if (error)
1276 goto fail_iput;
1277
1278 error = gfs2_acl_create(dip, ip);
1279 if (error)
1280 goto fail_iput;
1281
1282 error = link_dinode(dip, name, ip, ul);
1283 if (error)
1284 goto fail_iput;
1285
1286 gfs2_unlinked_put(sdp, ul);
1287
7359a19c
SW
1288 inode = gfs2_ip2v(ip);
1289 gfs2_inode_put(ip);
1290 if (!inode)
1291 return ERR_PTR(-ENOMEM);
1292 return inode;
b3b94faa
DT
1293
1294 fail_iput:
1295 gfs2_inode_put(ip);
1296
1297 fail_gunlock2:
1298 gfs2_glock_dq_uninit(ghs + 1);
1299
1300 fail_gunlock:
1301 gfs2_glock_dq(ghs);
1302
1303 fail:
1304 gfs2_unlinked_put(sdp, ul);
1305
7359a19c 1306 return ERR_PTR(error);
b3b94faa
DT
1307}
1308
1309/**
1310 * gfs2_unlinki - Unlink a file
1311 * @dip: The inode of the directory
1312 * @name: The name of the file to be unlinked
1313 * @ip: The inode of the file to be removed
1314 *
1315 * Assumes Glocks on both dip and ip are held.
1316 *
1317 * Returns: errno
1318 */
1319
1320int gfs2_unlinki(struct gfs2_inode *dip, struct qstr *name,
1321 struct gfs2_inode *ip, struct gfs2_unlinked *ul)
1322{
1323 struct gfs2_sbd *sdp = dip->i_sbd;
1324 int error;
1325
1326 error = gfs2_dir_del(dip, name);
1327 if (error)
1328 return error;
1329
1330 error = gfs2_change_nlink(ip, -1);
1331 if (error)
1332 return error;
1333
1334 /* If this inode is being unlinked from the directory structure,
1335 we need to mark that in the log so that it isn't lost during
1336 a crash. */
1337
1338 if (!ip->i_di.di_nlink) {
1339 ul->ul_ut.ut_inum = ip->i_num;
1340 error = gfs2_unlinked_ondisk_add(sdp, ul);
1341 if (!error)
1342 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
1343 }
1344
1345 return error;
1346}
1347
1348/**
1349 * gfs2_rmdiri - Remove a directory
1350 * @dip: The parent directory of the directory to be removed
1351 * @name: The name of the directory to be removed
1352 * @ip: The GFS2 inode of the directory to be removed
1353 *
1354 * Assumes Glocks on dip and ip are held
1355 *
1356 * Returns: errno
1357 */
1358
1359int gfs2_rmdiri(struct gfs2_inode *dip, struct qstr *name,
1360 struct gfs2_inode *ip, struct gfs2_unlinked *ul)
1361{
1362 struct gfs2_sbd *sdp = dip->i_sbd;
1363 struct qstr dotname;
1364 int error;
1365
1366 if (ip->i_di.di_entries != 2) {
1367 if (gfs2_consist_inode(ip))
1368 gfs2_dinode_print(&ip->i_di);
1369 return -EIO;
1370 }
1371
1372 error = gfs2_dir_del(dip, name);
1373 if (error)
1374 return error;
1375
1376 error = gfs2_change_nlink(dip, -1);
1377 if (error)
1378 return error;
1379
1380 dotname.len = 1;
1381 dotname.name = ".";
1382 error = gfs2_dir_del(ip, &dotname);
1383 if (error)
1384 return error;
1385
1386 dotname.len = 2;
1387 dotname.name = "..";
1388 error = gfs2_dir_del(ip, &dotname);
1389 if (error)
1390 return error;
1391
1392 error = gfs2_change_nlink(ip, -2);
1393 if (error)
1394 return error;
1395
1396 /* This inode is being unlinked from the directory structure and
1397 we need to mark that in the log so that it isn't lost during
1398 a crash. */
1399
1400 ul->ul_ut.ut_inum = ip->i_num;
1401 error = gfs2_unlinked_ondisk_add(sdp, ul);
1402 if (!error)
1403 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
1404
1405 return error;
1406}
1407
1408/*
1409 * gfs2_unlink_ok - check to see that a inode is still in a directory
1410 * @dip: the directory
1411 * @name: the name of the file
1412 * @ip: the inode
1413 *
1414 * Assumes that the lock on (at least) @dip is held.
1415 *
1416 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1417 */
1418
1419int gfs2_unlink_ok(struct gfs2_inode *dip, struct qstr *name,
1420 struct gfs2_inode *ip)
1421{
1422 struct gfs2_inum inum;
1423 unsigned int type;
1424 int error;
1425
1426 if (IS_IMMUTABLE(ip->i_vnode) || IS_APPEND(ip->i_vnode))
1427 return -EPERM;
1428
1429 if ((dip->i_di.di_mode & S_ISVTX) &&
1430 dip->i_di.di_uid != current->fsuid &&
1431 ip->i_di.di_uid != current->fsuid &&
1432 !capable(CAP_FOWNER))
1433 return -EPERM;
1434
1435 if (IS_APPEND(dip->i_vnode))
1436 return -EPERM;
1437
1438 error = gfs2_repermission(dip->i_vnode, MAY_WRITE | MAY_EXEC, NULL);
1439 if (error)
1440 return error;
1441
1442 error = gfs2_dir_search(dip, name, &inum, &type);
1443 if (error)
1444 return error;
1445
1446 if (!gfs2_inum_equal(&inum, &ip->i_num))
1447 return -ENOENT;
1448
1449 if (IF2DT(ip->i_di.di_mode) != type) {
1450 gfs2_consist_inode(dip);
1451 return -EIO;
1452 }
1453
1454 return 0;
1455}
1456
1457/*
1458 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1459 * @this: move this
1460 * @to: to here
1461 *
1462 * Follow @to back to the root and make sure we don't encounter @this
1463 * Assumes we already hold the rename lock.
1464 *
1465 * Returns: errno
1466 */
1467
1468int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to)
1469{
7359a19c 1470 struct inode *dir = to->i_vnode;
c9fd4307 1471 struct super_block *sb = dir->i_sb;
7359a19c 1472 struct inode *tmp;
b3b94faa
DT
1473 struct qstr dotdot;
1474 int error = 0;
1475
1476 memset(&dotdot, 0, sizeof(struct qstr));
1477 dotdot.name = "..";
1478 dotdot.len = 2;
1479
7359a19c 1480 igrab(dir);
b3b94faa
DT
1481
1482 for (;;) {
7359a19c 1483 if (dir == this->i_vnode) {
b3b94faa
DT
1484 error = -EINVAL;
1485 break;
1486 }
c9fd4307 1487 if (dir == sb->s_root->d_inode) {
b3b94faa
DT
1488 error = 0;
1489 break;
1490 }
1491
7359a19c 1492 error = gfs2_lookupi(dir, &dotdot, 1, &tmp);
b3b94faa
DT
1493 if (error)
1494 break;
1495
7359a19c
SW
1496 iput(dir);
1497 dir = tmp;
b3b94faa
DT
1498 }
1499
7359a19c 1500 iput(dir);
b3b94faa
DT
1501
1502 return error;
1503}
1504
1505/**
1506 * gfs2_readlinki - return the contents of a symlink
1507 * @ip: the symlink's inode
1508 * @buf: a pointer to the buffer to be filled
1509 * @len: a pointer to the length of @buf
1510 *
1511 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1512 * to be freed by the caller.
1513 *
1514 * Returns: errno
1515 */
1516
1517int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1518{
1519 struct gfs2_holder i_gh;
1520 struct buffer_head *dibh;
1521 unsigned int x;
1522 int error;
1523
1524 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
1525 error = gfs2_glock_nq_atime(&i_gh);
1526 if (error) {
1527 gfs2_holder_uninit(&i_gh);
1528 return error;
1529 }
1530
1531 if (!ip->i_di.di_size) {
1532 gfs2_consist_inode(ip);
1533 error = -EIO;
1534 goto out;
1535 }
1536
1537 error = gfs2_meta_inode_buffer(ip, &dibh);
1538 if (error)
1539 goto out;
1540
1541 x = ip->i_di.di_size + 1;
1542 if (x > *len) {
1543 *buf = kmalloc(x, GFP_KERNEL);
1544 if (!*buf) {
1545 error = -ENOMEM;
1546 goto out_brelse;
1547 }
1548 }
1549
1550 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1551 *len = x;
1552
1553 out_brelse:
1554 brelse(dibh);
1555
1556 out:
1557 gfs2_glock_dq_uninit(&i_gh);
1558
1559 return error;
1560}
1561
1562/**
1563 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1564 * conditionally update the inode's atime
1565 * @gh: the holder to acquire
1566 *
1567 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1568 * Update if the difference between the current time and the inode's current
1569 * atime is greater than an interval specified at mount.
1570 *
1571 * Returns: errno
1572 */
1573
1574int gfs2_glock_nq_atime(struct gfs2_holder *gh)
1575{
1576 struct gfs2_glock *gl = gh->gh_gl;
1577 struct gfs2_sbd *sdp = gl->gl_sbd;
5c676f6d 1578 struct gfs2_inode *ip = gl->gl_object;
b3b94faa
DT
1579 int64_t curtime, quantum = gfs2_tune_get(sdp, gt_atime_quantum);
1580 unsigned int state;
1581 int flags;
1582 int error;
1583
1584 if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) ||
1585 gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) ||
1586 gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops))
1587 return -EINVAL;
1588
1589 state = gh->gh_state;
1590 flags = gh->gh_flags;
1591
1592 error = gfs2_glock_nq(gh);
1593 if (error)
1594 return error;
1595
1596 if (test_bit(SDF_NOATIME, &sdp->sd_flags) ||
1597 (sdp->sd_vfs->s_flags & MS_RDONLY))
1598 return 0;
1599
1600 curtime = get_seconds();
1601 if (curtime - ip->i_di.di_atime >= quantum) {
1602 gfs2_glock_dq(gh);
1603 gfs2_holder_reinit(LM_ST_EXCLUSIVE,
1604 gh->gh_flags & ~LM_FLAG_ANY,
1605 gh);
1606 error = gfs2_glock_nq(gh);
1607 if (error)
1608 return error;
1609
1610 /* Verify that atime hasn't been updated while we were
1611 trying to get exclusive lock. */
1612
1613 curtime = get_seconds();
1614 if (curtime - ip->i_di.di_atime >= quantum) {
1615 struct buffer_head *dibh;
1616
1617 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
1618 if (error == -EROFS)
1619 return 0;
1620 if (error)
1621 goto fail;
1622
1623 error = gfs2_meta_inode_buffer(ip, &dibh);
1624 if (error)
1625 goto fail_end_trans;
1626
1627 ip->i_di.di_atime = curtime;
1628
d4e9c4c3 1629 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
b3b94faa
DT
1630 gfs2_dinode_out(&ip->i_di, dibh->b_data);
1631 brelse(dibh);
1632
1633 gfs2_trans_end(sdp);
1634 }
1635
1636 /* If someone else has asked for the glock,
1637 unlock and let them have it. Then reacquire
1638 in the original state. */
1639 if (gfs2_glock_is_blocking(gl)) {
1640 gfs2_glock_dq(gh);
1641 gfs2_holder_reinit(state, flags, gh);
1642 return gfs2_glock_nq(gh);
1643 }
1644 }
1645
1646 return 0;
1647
1648 fail_end_trans:
1649 gfs2_trans_end(sdp);
1650
1651 fail:
1652 gfs2_glock_dq(gh);
1653
1654 return error;
1655}
1656
1657/**
1658 * glock_compare_atime - Compare two struct gfs2_glock structures for sort
1659 * @arg_a: the first structure
1660 * @arg_b: the second structure
1661 *
1662 * Returns: 1 if A > B
1663 * -1 if A < B
1664 * 0 if A = B
1665 */
1666
1667static int glock_compare_atime(const void *arg_a, const void *arg_b)
1668{
1669 struct gfs2_holder *gh_a = *(struct gfs2_holder **)arg_a;
1670 struct gfs2_holder *gh_b = *(struct gfs2_holder **)arg_b;
1671 struct lm_lockname *a = &gh_a->gh_gl->gl_name;
1672 struct lm_lockname *b = &gh_b->gh_gl->gl_name;
1673 int ret = 0;
1674
1675 if (a->ln_number > b->ln_number)
1676 ret = 1;
1677 else if (a->ln_number < b->ln_number)
1678 ret = -1;
1679 else {
1680 if (gh_a->gh_state == LM_ST_SHARED &&
1681 gh_b->gh_state == LM_ST_EXCLUSIVE)
1682 ret = 1;
1683 else if (gh_a->gh_state == LM_ST_SHARED &&
1684 (gh_b->gh_flags & GL_ATIME))
1685 ret = 1;
1686 }
1687
1688 return ret;
1689}
1690
1691/**
1692 * gfs2_glock_nq_m_atime - acquire multiple glocks where one may need an
1693 * atime update
1694 * @num_gh: the number of structures
1695 * @ghs: an array of struct gfs2_holder structures
1696 *
1697 * Returns: 0 on success (all glocks acquired),
1698 * errno on failure (no glocks acquired)
1699 */
1700
1701int gfs2_glock_nq_m_atime(unsigned int num_gh, struct gfs2_holder *ghs)
1702{
1703 struct gfs2_holder **p;
1704 unsigned int x;
1705 int error = 0;
1706
1707 if (!num_gh)
1708 return 0;
1709
1710 if (num_gh == 1) {
1711 ghs->gh_flags &= ~(LM_FLAG_TRY | GL_ASYNC);
1712 if (ghs->gh_flags & GL_ATIME)
1713 error = gfs2_glock_nq_atime(ghs);
1714 else
1715 error = gfs2_glock_nq(ghs);
1716 return error;
1717 }
1718
1719 p = kcalloc(num_gh, sizeof(struct gfs2_holder *), GFP_KERNEL);
1720 if (!p)
1721 return -ENOMEM;
1722
1723 for (x = 0; x < num_gh; x++)
1724 p[x] = &ghs[x];
1725
1726 sort(p, num_gh, sizeof(struct gfs2_holder *), glock_compare_atime,NULL);
1727
1728 for (x = 0; x < num_gh; x++) {
1729 p[x]->gh_flags &= ~(LM_FLAG_TRY | GL_ASYNC);
1730
1731 if (p[x]->gh_flags & GL_ATIME)
1732 error = gfs2_glock_nq_atime(p[x]);
1733 else
1734 error = gfs2_glock_nq(p[x]);
1735
1736 if (error) {
1737 while (x--)
1738 gfs2_glock_dq(p[x]);
1739 break;
1740 }
1741 }
1742
1743 kfree(p);
1744
1745 return error;
1746}
1747
1748/**
1749 * gfs2_try_toss_vnode - See if we can toss a vnode from memory
1750 * @ip: the inode
1751 *
1752 * Returns: 1 if the vnode was tossed
1753 */
1754
1755void gfs2_try_toss_vnode(struct gfs2_inode *ip)
1756{
1757 struct inode *inode;
1758
1759 inode = gfs2_ip2v_lookup(ip);
1760 if (!inode)
1761 return;
1762
1763 d_prune_aliases(inode);
1764
1765 if (S_ISDIR(ip->i_di.di_mode)) {
1766 struct list_head *head = &inode->i_dentry;
1767 struct dentry *d = NULL;
1768
1769 spin_lock(&dcache_lock);
1770 if (list_empty(head))
1771 spin_unlock(&dcache_lock);
1772 else {
1773 d = list_entry(head->next, struct dentry, d_alias);
1774 dget_locked(d);
1775 spin_unlock(&dcache_lock);
1776
1777 if (have_submounts(d))
1778 dput(d);
1779 else {
1780 shrink_dcache_parent(d);
1781 dput(d);
1782 d_prune_aliases(inode);
1783 }
1784 }
1785 }
1786
1787 inode->i_nlink = 0;
1788 iput(inode);
1789}
1790
1791
1792static int
1793__gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1794{
1795 struct buffer_head *dibh;
1796 int error;
1797
1798 error = gfs2_meta_inode_buffer(ip, &dibh);
1799 if (!error) {
1800 error = inode_setattr(ip->i_vnode, attr);
1801 gfs2_assert_warn(ip->i_sbd, !error);
1802 gfs2_inode_attr_out(ip);
1803
d4e9c4c3 1804 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
b3b94faa
DT
1805 gfs2_dinode_out(&ip->i_di, dibh->b_data);
1806 brelse(dibh);
1807 }
1808 return error;
1809}
1810
1811/**
1812 * gfs2_setattr_simple -
1813 * @ip:
1814 * @attr:
1815 *
1816 * Called with a reference on the vnode.
1817 *
1818 * Returns: errno
1819 */
1820
1821int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1822{
1823 int error;
1824
5c676f6d 1825 if (current->journal_info)
b3b94faa
DT
1826 return __gfs2_setattr_simple(ip, attr);
1827
1828 error = gfs2_trans_begin(ip->i_sbd, RES_DINODE, 0);
1829 if (error)
1830 return error;
1831
1832 error = __gfs2_setattr_simple(ip, attr);
1833
1834 gfs2_trans_end(ip->i_sbd);
1835
1836 return error;
1837}
1838
1839int gfs2_repermission(struct inode *inode, int mask, struct nameidata *nd)
1840{
1841 return permission(inode, mask, nd);
1842}
1843