?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[NETFILTER]: nf_conntrack/nf_nat: add SIP helper port
[linux-2.6.22.y-op.git] / fs / gfs2 / inode.c
blobd470e5286ecd7715e0804a7fc54a6e34015c7559
1 /*
2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2006 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 version 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>
17 #include <linux/gfs2_ondisk.h>
18 #include <linux/crc32.h>
19 #include <linux/lm_interface.h>
20 #include <linux/security.h>
21
22 #include "gfs2.h"
23 #include "incore.h"
24 #include "acl.h"
25 #include "bmap.h"
26 #include "dir.h"
27 #include "eattr.h"
28 #include "glock.h"
29 #include "glops.h"
30 #include "inode.h"
31 #include "log.h"
32 #include "meta_io.h"
33 #include "ops_address.h"
34 #include "ops_file.h"
35 #include "ops_inode.h"
36 #include "quota.h"
37 #include "rgrp.h"
38 #include "trans.h"
39 #include "util.h"
40
41 /**
42 * gfs2_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
48 void gfs2_inode_attr_in(struct gfs2_inode *ip)
49 {
50 struct inode *inode = &ip->i_inode;
51 struct gfs2_dinode *di = &ip->i_di;
52
53 inode->i_ino = ip->i_num.no_addr;
54
55 switch (di->di_mode & S_IFMT) {
56 case S_IFBLK:
57 case S_IFCHR:
58 inode->i_rdev = MKDEV(di->di_major, di->di_minor);
59 break;
60 default:
61 inode->i_rdev = 0;
62 break;
63 };
64
65 inode->i_mode = di->di_mode;
66 inode->i_nlink = di->di_nlink;
67 inode->i_uid = di->di_uid;
68 inode->i_gid = di->di_gid;
69 i_size_write(inode, di->di_size);
70 inode->i_atime.tv_sec = di->di_atime;
71 inode->i_mtime.tv_sec = di->di_mtime;
72 inode->i_ctime.tv_sec = di->di_ctime;
73 inode->i_atime.tv_nsec = 0;
74 inode->i_mtime.tv_nsec = 0;
75 inode->i_ctime.tv_nsec = 0;
76 inode->i_blocks = di->di_blocks <<
77 (GFS2_SB(inode)->sd_sb.sb_bsize_shift - GFS2_BASIC_BLOCK_SHIFT);
78
79 if (di->di_flags & GFS2_DIF_IMMUTABLE)
80 inode->i_flags |= S_IMMUTABLE;
81 else
82 inode->i_flags &= ~S_IMMUTABLE;
83
84 if (di->di_flags & GFS2_DIF_APPENDONLY)
85 inode->i_flags |= S_APPEND;
86 else
87 inode->i_flags &= ~S_APPEND;
88 }
89
90 /**
91 * gfs2_inode_attr_out - Copy attributes from VFS inode into the dinode
92 * @ip: The GFS2 inode
93 *
94 * Only copy out the attributes that we want the VFS layer
95 * to be able to modify.
96 */
97
98 void gfs2_inode_attr_out(struct gfs2_inode *ip)
99 {
100 struct inode *inode = &ip->i_inode;
101 struct gfs2_dinode *di = &ip->i_di;
102 gfs2_assert_withdraw(GFS2_SB(inode),
103 (di->di_mode & S_IFMT) == (inode->i_mode & S_IFMT));
104 di->di_mode = inode->i_mode;
105 di->di_uid = inode->i_uid;
106 di->di_gid = inode->i_gid;
107 di->di_atime = inode->i_atime.tv_sec;
108 di->di_mtime = inode->i_mtime.tv_sec;
109 di->di_ctime = inode->i_ctime.tv_sec;
110 }
111
112 static int iget_test(struct inode *inode, void *opaque)
113 {
114 struct gfs2_inode *ip = GFS2_I(inode);
115 struct gfs2_inum *inum = opaque;
116
117 if (ip && ip->i_num.no_addr == inum->no_addr)
118 return 1;
119
120 return 0;
121 }
122
123 static int iget_set(struct inode *inode, void *opaque)
124 {
125 struct gfs2_inode *ip = GFS2_I(inode);
126 struct gfs2_inum *inum = opaque;
127
128 ip->i_num = *inum;
129 return 0;
130 }
131
132 struct inode *gfs2_ilookup(struct super_block *sb, struct gfs2_inum *inum)
133 {
134 return ilookup5(sb, (unsigned long)inum->no_formal_ino,
135 iget_test, inum);
136 }
137
138 static struct inode *gfs2_iget(struct super_block *sb, struct gfs2_inum *inum)
139 {
140 return iget5_locked(sb, (unsigned long)inum->no_formal_ino,
141 iget_test, iget_set, inum);
142 }
143
144 /**
145 * gfs2_inode_lookup - Lookup an inode
146 * @sb: The super block
147 * @inum: The inode number
148 * @type: The type of the inode
149 *
150 * Returns: A VFS inode, or an error
151 */
152
153 struct inode *gfs2_inode_lookup(struct super_block *sb, struct gfs2_inum *inum, unsigned int type)
154 {
155 struct inode *inode = gfs2_iget(sb, inum);
156 struct gfs2_inode *ip = GFS2_I(inode);
157 struct gfs2_glock *io_gl;
158 int error;
159
160 if (!inode)
161 return ERR_PTR(-ENOBUFS);
162
163 if (inode->i_state & I_NEW) {
164 struct gfs2_sbd *sdp = GFS2_SB(inode);
165 umode_t mode = DT2IF(type);
166 inode->i_private = ip;
167 inode->i_mode = mode;
168
169 if (S_ISREG(mode)) {
170 inode->i_op = &gfs2_file_iops;
171 inode->i_fop = &gfs2_file_fops;
172 inode->i_mapping->a_ops = &gfs2_file_aops;
173 } else if (S_ISDIR(mode)) {
174 inode->i_op = &gfs2_dir_iops;
175 inode->i_fop = &gfs2_dir_fops;
176 } else if (S_ISLNK(mode)) {
177 inode->i_op = &gfs2_symlink_iops;
178 } else {
179 inode->i_op = &gfs2_dev_iops;
180 }
181
182 error = gfs2_glock_get(sdp, inum->no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
183 if (unlikely(error))
184 goto fail;
185 ip->i_gl->gl_object = ip;
186
187 error = gfs2_glock_get(sdp, inum->no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
188 if (unlikely(error))
189 goto fail_put;
190
191 ip->i_vn = ip->i_gl->gl_vn - 1;
192 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
193 if (unlikely(error))
194 goto fail_iopen;
195
196 gfs2_glock_put(io_gl);
197 unlock_new_inode(inode);
198 }
199
200 return inode;
201 fail_iopen:
202 gfs2_glock_put(io_gl);
203 fail_put:
204 ip->i_gl->gl_object = NULL;
205 gfs2_glock_put(ip->i_gl);
206 fail:
207 iput(inode);
208 return ERR_PTR(error);
209 }
210
211 /**
212 * gfs2_inode_refresh - Refresh the incore copy of the dinode
213 * @ip: The GFS2 inode
214 *
215 * Returns: errno
216 */
217
218 int gfs2_inode_refresh(struct gfs2_inode *ip)
219 {
220 struct buffer_head *dibh;
221 int error;
222
223 error = gfs2_meta_inode_buffer(ip, &dibh);
224 if (error)
225 return error;
226
227 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
228 brelse(dibh);
229 return -EIO;
230 }
231
232 gfs2_dinode_in(&ip->i_di, dibh->b_data);
233
234 brelse(dibh);
235
236 if (ip->i_num.no_addr != ip->i_di.di_num.no_addr) {
237 if (gfs2_consist_inode(ip))
238 gfs2_dinode_print(&ip->i_di);
239 return -EIO;
240 }
241 if (ip->i_num.no_formal_ino != ip->i_di.di_num.no_formal_ino)
242 return -ESTALE;
243
244 ip->i_vn = ip->i_gl->gl_vn;
245
246 return 0;
247 }
248
249 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
250 {
251 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
252 struct gfs2_alloc *al;
253 struct gfs2_rgrpd *rgd;
254 int error;
255
256 if (ip->i_di.di_blocks != 1) {
257 if (gfs2_consist_inode(ip))
258 gfs2_dinode_print(&ip->i_di);
259 return -EIO;
260 }
261
262 al = gfs2_alloc_get(ip);
263
264 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
265 if (error)
266 goto out;
267
268 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
269 if (error)
270 goto out_qs;
271
272 rgd = gfs2_blk2rgrpd(sdp, ip->i_num.no_addr);
273 if (!rgd) {
274 gfs2_consist_inode(ip);
275 error = -EIO;
276 goto out_rindex_relse;
277 }
278
279 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
280 &al->al_rgd_gh);
281 if (error)
282 goto out_rindex_relse;
283
284 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
285 if (error)
286 goto out_rg_gunlock;
287
288 gfs2_trans_add_gl(ip->i_gl);
289
290 gfs2_free_di(rgd, ip);
291
292 gfs2_trans_end(sdp);
293 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
294
295 out_rg_gunlock:
296 gfs2_glock_dq_uninit(&al->al_rgd_gh);
297 out_rindex_relse:
298 gfs2_glock_dq_uninit(&al->al_ri_gh);
299 out_qs:
300 gfs2_quota_unhold(ip);
301 out:
302 gfs2_alloc_put(ip);
303 return error;
304 }
305
306 /**
307 * gfs2_change_nlink - Change nlink count on inode
308 * @ip: The GFS2 inode
309 * @diff: The change in the nlink count required
310 *
311 * Returns: errno
312 */
313
314 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
315 {
316 struct gfs2_sbd *sdp = ip->i_inode.i_sb->s_fs_info;
317 struct buffer_head *dibh;
318 u32 nlink;
319 int error;
320
321 BUG_ON(ip->i_di.di_nlink != ip->i_inode.i_nlink);
322 nlink = ip->i_di.di_nlink + diff;
323
324 /* If we are reducing the nlink count, but the new value ends up being
325 bigger than the old one, we must have underflowed. */
326 if (diff < 0 && nlink > ip->i_di.di_nlink) {
327 if (gfs2_consist_inode(ip))
328 gfs2_dinode_print(&ip->i_di);
329 return -EIO;
330 }
331
332 error = gfs2_meta_inode_buffer(ip, &dibh);
333 if (error)
334 return error;
335
336 ip->i_di.di_nlink = nlink;
337 ip->i_di.di_ctime = get_seconds();
338 ip->i_inode.i_nlink = nlink;
339
340 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
341 gfs2_dinode_out(&ip->i_di, dibh->b_data);
342 brelse(dibh);
343 mark_inode_dirty(&ip->i_inode);
344
345 if (ip->i_di.di_nlink == 0) {
346 struct gfs2_rgrpd *rgd;
347 struct gfs2_holder ri_gh, rg_gh;
348
349 error = gfs2_rindex_hold(sdp, &ri_gh);
350 if (error)
351 goto out;
352 error = -EIO;
353 rgd = gfs2_blk2rgrpd(sdp, ip->i_num.no_addr);
354 if (!rgd)
355 goto out_norgrp;
356 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &rg_gh);
357 if (error)
358 goto out_norgrp;
359
360 clear_nlink(&ip->i_inode);
361 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
362 gfs2_glock_dq_uninit(&rg_gh);
363 out_norgrp:
364 gfs2_glock_dq_uninit(&ri_gh);
365 }
366 out:
367 return error;
368 }
369
370 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
371 {
372 struct qstr qstr;
373 gfs2_str2qstr(&qstr, name);
374 return gfs2_lookupi(dip, &qstr, 1, NULL);
375 }
376
377
378 /**
379 * gfs2_lookupi - Look up a filename in a directory and return its inode
380 * @d_gh: An initialized holder for the directory glock
381 * @name: The name of the inode to look for
382 * @is_root: If 1, ignore the caller's permissions
383 * @i_gh: An uninitialized holder for the new inode glock
384 *
385 * There will always be a vnode (Linux VFS inode) for the d_gh inode unless
386 * @is_root is true.
387 *
388 * Returns: errno
389 */
390
391 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
392 int is_root, struct nameidata *nd)
393 {
394 struct super_block *sb = dir->i_sb;
395 struct gfs2_inode *dip = GFS2_I(dir);
396 struct gfs2_holder d_gh;
397 struct gfs2_inum inum;
398 unsigned int type;
399 int error = 0;
400 struct inode *inode = NULL;
401
402 if (!name->len || name->len > GFS2_FNAMESIZE)
403 return ERR_PTR(-ENAMETOOLONG);
404
405 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
406 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
407 dir == sb->s_root->d_inode)) {
408 igrab(dir);
409 return dir;
410 }
411
412 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
413 if (error)
414 return ERR_PTR(error);
415
416 if (!is_root) {
417 error = permission(dir, MAY_EXEC, NULL);
418 if (error)
419 goto out;
420 }
421
422 error = gfs2_dir_search(dir, name, &inum, &type);
423 if (error)
424 goto out;
425
426 inode = gfs2_inode_lookup(sb, &inum, type);
427
428 out:
429 gfs2_glock_dq_uninit(&d_gh);
430 if (error == -ENOENT)
431 return NULL;
432 return inode;
433 }
434
435 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
436 {
437 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
438 struct buffer_head *bh;
439 struct gfs2_inum_range ir;
440 int error;
441
442 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
443 if (error)
444 return error;
445 mutex_lock(&sdp->sd_inum_mutex);
446
447 error = gfs2_meta_inode_buffer(ip, &bh);
448 if (error) {
449 mutex_unlock(&sdp->sd_inum_mutex);
450 gfs2_trans_end(sdp);
451 return error;
452 }
453
454 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
455
456 if (ir.ir_length) {
457 *formal_ino = ir.ir_start++;
458 ir.ir_length--;
459 gfs2_trans_add_bh(ip->i_gl, bh, 1);
460 gfs2_inum_range_out(&ir,
461 bh->b_data + sizeof(struct gfs2_dinode));
462 brelse(bh);
463 mutex_unlock(&sdp->sd_inum_mutex);
464 gfs2_trans_end(sdp);
465 return 0;
466 }
467
468 brelse(bh);
469
470 mutex_unlock(&sdp->sd_inum_mutex);
471 gfs2_trans_end(sdp);
472
473 return 1;
474 }
475
476 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
477 {
478 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
479 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
480 struct gfs2_holder gh;
481 struct buffer_head *bh;
482 struct gfs2_inum_range ir;
483 int error;
484
485 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
486 if (error)
487 return error;
488
489 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
490 if (error)
491 goto out;
492 mutex_lock(&sdp->sd_inum_mutex);
493
494 error = gfs2_meta_inode_buffer(ip, &bh);
495 if (error)
496 goto out_end_trans;
497
498 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
499
500 if (!ir.ir_length) {
501 struct buffer_head *m_bh;
502 u64 x, y;
503
504 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
505 if (error)
506 goto out_brelse;
507
508 x = *(u64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
509 x = y = be64_to_cpu(x);
510 ir.ir_start = x;
511 ir.ir_length = GFS2_INUM_QUANTUM;
512 x += GFS2_INUM_QUANTUM;
513 if (x < y)
514 gfs2_consist_inode(m_ip);
515 x = cpu_to_be64(x);
516 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
517 *(u64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = x;
518
519 brelse(m_bh);
520 }
521
522 *formal_ino = ir.ir_start++;
523 ir.ir_length--;
524
525 gfs2_trans_add_bh(ip->i_gl, bh, 1);
526 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
527
528 out_brelse:
529 brelse(bh);
530 out_end_trans:
531 mutex_unlock(&sdp->sd_inum_mutex);
532 gfs2_trans_end(sdp);
533 out:
534 gfs2_glock_dq_uninit(&gh);
535 return error;
536 }
537
538 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
539 {
540 int error;
541
542 error = pick_formal_ino_1(sdp, inum);
543 if (error <= 0)
544 return error;
545
546 error = pick_formal_ino_2(sdp, inum);
547
548 return error;
549 }
550
551 /**
552 * create_ok - OK to create a new on-disk inode here?
553 * @dip: Directory in which dinode is to be created
554 * @name: Name of new dinode
555 * @mode:
556 *
557 * Returns: errno
558 */
559
560 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
561 unsigned int mode)
562 {
563 int error;
564
565 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
566 if (error)
567 return error;
568
569 /* Don't create entries in an unlinked directory */
570 if (!dip->i_di.di_nlink)
571 return -EPERM;
572
573 error = gfs2_dir_search(&dip->i_inode, name, NULL, NULL);
574 switch (error) {
575 case -ENOENT:
576 error = 0;
577 break;
578 case 0:
579 return -EEXIST;
580 default:
581 return error;
582 }
583
584 if (dip->i_di.di_entries == (u32)-1)
585 return -EFBIG;
586 if (S_ISDIR(mode) && dip->i_di.di_nlink == (u32)-1)
587 return -EMLINK;
588
589 return 0;
590 }
591
592 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
593 unsigned int *uid, unsigned int *gid)
594 {
595 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
596 (dip->i_di.di_mode & S_ISUID) && dip->i_di.di_uid) {
597 if (S_ISDIR(*mode))
598 *mode |= S_ISUID;
599 else if (dip->i_di.di_uid != current->fsuid)
600 *mode &= ~07111;
601 *uid = dip->i_di.di_uid;
602 } else
603 *uid = current->fsuid;
604
605 if (dip->i_di.di_mode & S_ISGID) {
606 if (S_ISDIR(*mode))
607 *mode |= S_ISGID;
608 *gid = dip->i_di.di_gid;
609 } else
610 *gid = current->fsgid;
611 }
612
613 static int alloc_dinode(struct gfs2_inode *dip, struct gfs2_inum *inum,
614 u64 *generation)
615 {
616 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
617 int error;
618
619 gfs2_alloc_get(dip);
620
621 dip->i_alloc.al_requested = RES_DINODE;
622 error = gfs2_inplace_reserve(dip);
623 if (error)
624 goto out;
625
626 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
627 if (error)
628 goto out_ipreserv;
629
630 inum->no_addr = gfs2_alloc_di(dip, generation);
631
632 gfs2_trans_end(sdp);
633
634 out_ipreserv:
635 gfs2_inplace_release(dip);
636 out:
637 gfs2_alloc_put(dip);
638 return error;
639 }
640
641 /**
642 * init_dinode - Fill in a new dinode structure
643 * @dip: the directory this inode is being created in
644 * @gl: The glock covering the new inode
645 * @inum: the inode number
646 * @mode: the file permissions
647 * @uid:
648 * @gid:
649 *
650 */
651
652 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
653 const struct gfs2_inum *inum, unsigned int mode,
654 unsigned int uid, unsigned int gid,
655 const u64 *generation)
656 {
657 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
658 struct gfs2_dinode *di;
659 struct buffer_head *dibh;
660
661 dibh = gfs2_meta_new(gl, inum->no_addr);
662 gfs2_trans_add_bh(gl, dibh, 1);
663 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
664 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
665 di = (struct gfs2_dinode *)dibh->b_data;
666
667 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
668 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
669 di->di_mode = cpu_to_be32(mode);
670 di->di_uid = cpu_to_be32(uid);
671 di->di_gid = cpu_to_be32(gid);
672 di->di_nlink = cpu_to_be32(0);
673 di->di_size = cpu_to_be64(0);
674 di->di_blocks = cpu_to_be64(1);
675 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(get_seconds());
676 di->di_major = di->di_minor = cpu_to_be32(0);
677 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
678 di->di_generation = cpu_to_be64(*generation);
679 di->di_flags = cpu_to_be32(0);
680
681 if (S_ISREG(mode)) {
682 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
683 gfs2_tune_get(sdp, gt_new_files_jdata))
684 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
685 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) ||
686 gfs2_tune_get(sdp, gt_new_files_directio))
687 di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO);
688 } else if (S_ISDIR(mode)) {
689 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
690 GFS2_DIF_INHERIT_DIRECTIO);
691 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
692 GFS2_DIF_INHERIT_JDATA);
693 }
694
695 di->__pad1 = 0;
696 di->di_payload_format = cpu_to_be32(0);
697 di->di_height = cpu_to_be32(0);
698 di->__pad2 = 0;
699 di->__pad3 = 0;
700 di->di_depth = cpu_to_be16(0);
701 di->di_entries = cpu_to_be32(0);
702 memset(&di->__pad4, 0, sizeof(di->__pad4));
703 di->di_eattr = cpu_to_be64(0);
704 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
705
706 brelse(dibh);
707 }
708
709 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
710 unsigned int mode, const struct gfs2_inum *inum,
711 const u64 *generation)
712 {
713 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
714 unsigned int uid, gid;
715 int error;
716
717 munge_mode_uid_gid(dip, &mode, &uid, &gid);
718 gfs2_alloc_get(dip);
719
720 error = gfs2_quota_lock(dip, uid, gid);
721 if (error)
722 goto out;
723
724 error = gfs2_quota_check(dip, uid, gid);
725 if (error)
726 goto out_quota;
727
728 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
729 if (error)
730 goto out_quota;
731
732 init_dinode(dip, gl, inum, mode, uid, gid, generation);
733 gfs2_quota_change(dip, +1, uid, gid);
734 gfs2_trans_end(sdp);
735
736 out_quota:
737 gfs2_quota_unlock(dip);
738 out:
739 gfs2_alloc_put(dip);
740 return error;
741 }
742
743 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
744 struct gfs2_inode *ip)
745 {
746 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
747 struct gfs2_alloc *al;
748 int alloc_required;
749 struct buffer_head *dibh;
750 int error;
751
752 al = gfs2_alloc_get(dip);
753
754 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
755 if (error)
756 goto fail;
757
758 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
759 if (alloc_required < 0)
760 goto fail;
761 if (alloc_required) {
762 error = gfs2_quota_check(dip, dip->i_di.di_uid,
763 dip->i_di.di_gid);
764 if (error)
765 goto fail_quota_locks;
766
767 al->al_requested = sdp->sd_max_dirres;
768
769 error = gfs2_inplace_reserve(dip);
770 if (error)
771 goto fail_quota_locks;
772
773 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
774 al->al_rgd->rd_ri.ri_length +
775 2 * RES_DINODE +
776 RES_STATFS + RES_QUOTA, 0);
777 if (error)
778 goto fail_ipreserv;
779 } else {
780 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
781 if (error)
782 goto fail_quota_locks;
783 }
784
785 error = gfs2_dir_add(&dip->i_inode, name, &ip->i_num, IF2DT(ip->i_di.di_mode));
786 if (error)
787 goto fail_end_trans;
788
789 error = gfs2_meta_inode_buffer(ip, &dibh);
790 if (error)
791 goto fail_end_trans;
792 ip->i_di.di_nlink = 1;
793 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
794 gfs2_dinode_out(&ip->i_di, dibh->b_data);
795 brelse(dibh);
796 return 0;
797
798 fail_end_trans:
799 gfs2_trans_end(sdp);
800
801 fail_ipreserv:
802 if (dip->i_alloc.al_rgd)
803 gfs2_inplace_release(dip);
804
805 fail_quota_locks:
806 gfs2_quota_unlock(dip);
807
808 fail:
809 gfs2_alloc_put(dip);
810 return error;
811 }
812
813 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
814 {
815 int err;
816 size_t len;
817 void *value;
818 char *name;
819 struct gfs2_ea_request er;
820
821 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
822 &name, &value, &len);
823
824 if (err) {
825 if (err == -EOPNOTSUPP)
826 return 0;
827 return err;
828 }
829
830 memset(&er, 0, sizeof(struct gfs2_ea_request));
831
832 er.er_type = GFS2_EATYPE_SECURITY;
833 er.er_name = name;
834 er.er_data = value;
835 er.er_name_len = strlen(name);
836 er.er_data_len = len;
837
838 err = gfs2_ea_set_i(ip, &er);
839
840 kfree(value);
841 kfree(name);
842
843 return err;
844 }
845
846 /**
847 * gfs2_createi - Create a new inode
848 * @ghs: An array of two holders
849 * @name: The name of the new file
850 * @mode: the permissions on the new inode
851 *
852 * @ghs[0] is an initialized holder for the directory
853 * @ghs[1] is the holder for the inode lock
854 *
855 * If the return value is not NULL, the glocks on both the directory and the new
856 * file are held. A transaction has been started and an inplace reservation
857 * is held, as well.
858 *
859 * Returns: An inode
860 */
861
862 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
863 unsigned int mode)
864 {
865 struct inode *inode;
866 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
867 struct inode *dir = &dip->i_inode;
868 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
869 struct gfs2_inum inum;
870 int error;
871 u64 generation;
872
873 if (!name->len || name->len > GFS2_FNAMESIZE)
874 return ERR_PTR(-ENAMETOOLONG);
875
876 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
877 error = gfs2_glock_nq(ghs);
878 if (error)
879 goto fail;
880
881 error = create_ok(dip, name, mode);
882 if (error)
883 goto fail_gunlock;
884
885 error = pick_formal_ino(sdp, &inum.no_formal_ino);
886 if (error)
887 goto fail_gunlock;
888
889 error = alloc_dinode(dip, &inum, &generation);
890 if (error)
891 goto fail_gunlock;
892
893 if (inum.no_addr < dip->i_num.no_addr) {
894 gfs2_glock_dq(ghs);
895
896 error = gfs2_glock_nq_num(sdp, inum.no_addr,
897 &gfs2_inode_glops, LM_ST_EXCLUSIVE,
898 GL_SKIP, ghs + 1);
899 if (error) {
900 return ERR_PTR(error);
901 }
902
903 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
904 error = gfs2_glock_nq(ghs);
905 if (error) {
906 gfs2_glock_dq_uninit(ghs + 1);
907 return ERR_PTR(error);
908 }
909
910 error = create_ok(dip, name, mode);
911 if (error)
912 goto fail_gunlock2;
913 } else {
914 error = gfs2_glock_nq_num(sdp, inum.no_addr,
915 &gfs2_inode_glops, LM_ST_EXCLUSIVE,
916 GL_SKIP, ghs + 1);
917 if (error)
918 goto fail_gunlock;
919 }
920
921 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation);
922 if (error)
923 goto fail_gunlock2;
924
925 inode = gfs2_inode_lookup(dir->i_sb, &inum, IF2DT(mode));
926 if (IS_ERR(inode))
927 goto fail_gunlock2;
928
929 error = gfs2_inode_refresh(GFS2_I(inode));
930 if (error)
931 goto fail_iput;
932
933 error = gfs2_acl_create(dip, GFS2_I(inode));
934 if (error)
935 goto fail_iput;
936
937 error = gfs2_security_init(dip, GFS2_I(inode));
938 if (error)
939 goto fail_iput;
940
941 error = link_dinode(dip, name, GFS2_I(inode));
942 if (error)
943 goto fail_iput;
944
945 if (!inode)
946 return ERR_PTR(-ENOMEM);
947 return inode;
948
949 fail_iput:
950 iput(inode);
951 fail_gunlock2:
952 gfs2_glock_dq_uninit(ghs + 1);
953 fail_gunlock:
954 gfs2_glock_dq(ghs);
955 fail:
956 return ERR_PTR(error);
957 }
958
959 /**
960 * gfs2_rmdiri - Remove a directory
961 * @dip: The parent directory of the directory to be removed
962 * @name: The name of the directory to be removed
963 * @ip: The GFS2 inode of the directory to be removed
964 *
965 * Assumes Glocks on dip and ip are held
966 *
967 * Returns: errno
968 */
969
970 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
971 struct gfs2_inode *ip)
972 {
973 struct qstr dotname;
974 int error;
975
976 if (ip->i_di.di_entries != 2) {
977 if (gfs2_consist_inode(ip))
978 gfs2_dinode_print(&ip->i_di);
979 return -EIO;
980 }
981
982 error = gfs2_dir_del(dip, name);
983 if (error)
984 return error;
985
986 error = gfs2_change_nlink(dip, -1);
987 if (error)
988 return error;
989
990 gfs2_str2qstr(&dotname, ".");
991 error = gfs2_dir_del(ip, &dotname);
992 if (error)
993 return error;
994
995 gfs2_str2qstr(&dotname, "..");
996 error = gfs2_dir_del(ip, &dotname);
997 if (error)
998 return error;
999
1000 error = gfs2_change_nlink(ip, -2);
1001 if (error)
1002 return error;
1003
1004 return error;
1005 }
1006
1007 /*
1008 * gfs2_unlink_ok - check to see that a inode is still in a directory
1009 * @dip: the directory
1010 * @name: the name of the file
1011 * @ip: the inode
1012 *
1013 * Assumes that the lock on (at least) @dip is held.
1014 *
1015 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1016 */
1017
1018 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1019 struct gfs2_inode *ip)
1020 {
1021 struct gfs2_inum inum;
1022 unsigned int type;
1023 int error;
1024
1025 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1026 return -EPERM;
1027
1028 if ((dip->i_di.di_mode & S_ISVTX) &&
1029 dip->i_di.di_uid != current->fsuid &&
1030 ip->i_di.di_uid != current->fsuid && !capable(CAP_FOWNER))
1031 return -EPERM;
1032
1033 if (IS_APPEND(&dip->i_inode))
1034 return -EPERM;
1035
1036 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
1037 if (error)
1038 return error;
1039
1040 error = gfs2_dir_search(&dip->i_inode, name, &inum, &type);
1041 if (error)
1042 return error;
1043
1044 if (!gfs2_inum_equal(&inum, &ip->i_num))
1045 return -ENOENT;
1046
1047 if (IF2DT(ip->i_di.di_mode) != type) {
1048 gfs2_consist_inode(dip);
1049 return -EIO;
1050 }
1051
1052 return 0;
1053 }
1054
1055 /*
1056 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1057 * @this: move this
1058 * @to: to here
1059 *
1060 * Follow @to back to the root and make sure we don't encounter @this
1061 * Assumes we already hold the rename lock.
1062 *
1063 * Returns: errno
1064 */
1065
1066 int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to)
1067 {
1068 struct inode *dir = &to->i_inode;
1069 struct super_block *sb = dir->i_sb;
1070 struct inode *tmp;
1071 struct qstr dotdot;
1072 int error = 0;
1073
1074 gfs2_str2qstr(&dotdot, "..");
1075
1076 igrab(dir);
1077
1078 for (;;) {
1079 if (dir == &this->i_inode) {
1080 error = -EINVAL;
1081 break;
1082 }
1083 if (dir == sb->s_root->d_inode) {
1084 error = 0;
1085 break;
1086 }
1087
1088 tmp = gfs2_lookupi(dir, &dotdot, 1, NULL);
1089 if (IS_ERR(tmp)) {
1090 error = PTR_ERR(tmp);
1091 break;
1092 }
1093
1094 iput(dir);
1095 dir = tmp;
1096 }
1097
1098 iput(dir);
1099
1100 return error;
1101 }
1102
1103 /**
1104 * gfs2_readlinki - return the contents of a symlink
1105 * @ip: the symlink's inode
1106 * @buf: a pointer to the buffer to be filled
1107 * @len: a pointer to the length of @buf
1108 *
1109 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1110 * to be freed by the caller.
1111 *
1112 * Returns: errno
1113 */
1114
1115 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1116 {
1117 struct gfs2_holder i_gh;
1118 struct buffer_head *dibh;
1119 unsigned int x;
1120 int error;
1121
1122 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
1123 error = gfs2_glock_nq_atime(&i_gh);
1124 if (error) {
1125 gfs2_holder_uninit(&i_gh);
1126 return error;
1127 }
1128
1129 if (!ip->i_di.di_size) {
1130 gfs2_consist_inode(ip);
1131 error = -EIO;
1132 goto out;
1133 }
1134
1135 error = gfs2_meta_inode_buffer(ip, &dibh);
1136 if (error)
1137 goto out;
1138
1139 x = ip->i_di.di_size + 1;
1140 if (x > *len) {
1141 *buf = kmalloc(x, GFP_KERNEL);
1142 if (!*buf) {
1143 error = -ENOMEM;
1144 goto out_brelse;
1145 }
1146 }
1147
1148 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1149 *len = x;
1150
1151 out_brelse:
1152 brelse(dibh);
1153 out:
1154 gfs2_glock_dq_uninit(&i_gh);
1155 return error;
1156 }
1157
1158 /**
1159 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1160 * conditionally update the inode's atime
1161 * @gh: the holder to acquire
1162 *
1163 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1164 * Update if the difference between the current time and the inode's current
1165 * atime is greater than an interval specified at mount.
1166 *
1167 * Returns: errno
1168 */
1169
1170 int gfs2_glock_nq_atime(struct gfs2_holder *gh)
1171 {
1172 struct gfs2_glock *gl = gh->gh_gl;
1173 struct gfs2_sbd *sdp = gl->gl_sbd;
1174 struct gfs2_inode *ip = gl->gl_object;
1175 s64 curtime, quantum = gfs2_tune_get(sdp, gt_atime_quantum);
1176 unsigned int state;
1177 int flags;
1178 int error;
1179
1180 if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) ||
1181 gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) ||
1182 gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops))
1183 return -EINVAL;
1184
1185 state = gh->gh_state;
1186 flags = gh->gh_flags;
1187
1188 error = gfs2_glock_nq(gh);
1189 if (error)
1190 return error;
1191
1192 if (test_bit(SDF_NOATIME, &sdp->sd_flags) ||
1193 (sdp->sd_vfs->s_flags & MS_RDONLY))
1194 return 0;
1195
1196 curtime = get_seconds();
1197 if (curtime - ip->i_di.di_atime >= quantum) {
1198 gfs2_glock_dq(gh);
1199 gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY,
1200 gh);
1201 error = gfs2_glock_nq(gh);
1202 if (error)
1203 return error;
1204
1205 /* Verify that atime hasn't been updated while we were
1206 trying to get exclusive lock. */
1207
1208 curtime = get_seconds();
1209 if (curtime - ip->i_di.di_atime >= quantum) {
1210 struct buffer_head *dibh;
1211 struct gfs2_dinode *di;
1212
1213 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
1214 if (error == -EROFS)
1215 return 0;
1216 if (error)
1217 goto fail;
1218
1219 error = gfs2_meta_inode_buffer(ip, &dibh);
1220 if (error)
1221 goto fail_end_trans;
1222
1223 ip->i_di.di_atime = curtime;
1224
1225 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1226 di = (struct gfs2_dinode *)dibh->b_data;
1227 di->di_atime = cpu_to_be64(ip->i_di.di_atime);
1228 brelse(dibh);
1229
1230 gfs2_trans_end(sdp);
1231 }
1232
1233 /* If someone else has asked for the glock,
1234 unlock and let them have it. Then reacquire
1235 in the original state. */
1236 if (gfs2_glock_is_blocking(gl)) {
1237 gfs2_glock_dq(gh);
1238 gfs2_holder_reinit(state, flags, gh);
1239 return gfs2_glock_nq(gh);
1240 }
1241 }
1242
1243 return 0;
1244
1245 fail_end_trans:
1246 gfs2_trans_end(sdp);
1247 fail:
1248 gfs2_glock_dq(gh);
1249 return error;
1250 }
1251
1252 /**
1253 * glock_compare_atime - Compare two struct gfs2_glock structures for sort
1254 * @arg_a: the first structure
1255 * @arg_b: the second structure
1256 *
1257 * Returns: 1 if A > B
1258 * -1 if A < B
1259 * 0 if A == B
1260 */
1261
1262 static int glock_compare_atime(const void *arg_a, const void *arg_b)
1263 {
1264 const struct gfs2_holder *gh_a = *(const struct gfs2_holder **)arg_a;
1265 const struct gfs2_holder *gh_b = *(const struct gfs2_holder **)arg_b;
1266 const struct lm_lockname *a = &gh_a->gh_gl->gl_name;
1267 const struct lm_lockname *b = &gh_b->gh_gl->gl_name;
1268
1269 if (a->ln_number > b->ln_number)
1270 return 1;
1271 if (a->ln_number < b->ln_number)
1272 return -1;
1273 if (gh_a->gh_state == LM_ST_SHARED && gh_b->gh_state == LM_ST_EXCLUSIVE)
1274 return 1;
1275 if (gh_a->gh_state == LM_ST_SHARED && (gh_b->gh_flags & GL_ATIME))
1276 return 1;
1277
1278 return 0;
1279 }
1280
1281 /**
1282 * gfs2_glock_nq_m_atime - acquire multiple glocks where one may need an
1283 * atime update
1284 * @num_gh: the number of structures
1285 * @ghs: an array of struct gfs2_holder structures
1286 *
1287 * Returns: 0 on success (all glocks acquired),
1288 * errno on failure (no glocks acquired)
1289 */
1290
1291 int gfs2_glock_nq_m_atime(unsigned int num_gh, struct gfs2_holder *ghs)
1292 {
1293 struct gfs2_holder **p;
1294 unsigned int x;
1295 int error = 0;
1296
1297 if (!num_gh)
1298 return 0;
1299
1300 if (num_gh == 1) {
1301 ghs->gh_flags &= ~(LM_FLAG_TRY | GL_ASYNC);
1302 if (ghs->gh_flags & GL_ATIME)
1303 error = gfs2_glock_nq_atime(ghs);
1304 else
1305 error = gfs2_glock_nq(ghs);
1306 return error;
1307 }
1308
1309 p = kcalloc(num_gh, sizeof(struct gfs2_holder *), GFP_KERNEL);
1310 if (!p)
1311 return -ENOMEM;
1312
1313 for (x = 0; x < num_gh; x++)
1314 p[x] = &ghs[x];
1315
1316 sort(p, num_gh, sizeof(struct gfs2_holder *), glock_compare_atime,NULL);
1317
1318 for (x = 0; x < num_gh; x++) {
1319 p[x]->gh_flags &= ~(LM_FLAG_TRY | GL_ASYNC);
1320
1321 if (p[x]->gh_flags & GL_ATIME)
1322 error = gfs2_glock_nq_atime(p[x]);
1323 else
1324 error = gfs2_glock_nq(p[x]);
1325
1326 if (error) {
1327 while (x--)
1328 gfs2_glock_dq(p[x]);
1329 break;
1330 }
1331 }
1332
1333 kfree(p);
1334 return error;
1335 }
1336
1337
1338 static int
1339 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1340 {
1341 struct buffer_head *dibh;
1342 int error;
1343
1344 error = gfs2_meta_inode_buffer(ip, &dibh);
1345 if (!error) {
1346 error = inode_setattr(&ip->i_inode, attr);
1347 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1348 gfs2_inode_attr_out(ip);
1349
1350 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1351 gfs2_dinode_out(&ip->i_di, dibh->b_data);
1352 brelse(dibh);
1353 }
1354 return error;
1355 }
1356
1357 /**
1358 * gfs2_setattr_simple -
1359 * @ip:
1360 * @attr:
1361 *
1362 * Called with a reference on the vnode.
1363 *
1364 * Returns: errno
1365 */
1366
1367 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1368 {
1369 int error;
1370
1371 if (current->journal_info)
1372 return __gfs2_setattr_simple(ip, attr);
1373
1374 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1375 if (error)
1376 return error;
1377
1378 error = __gfs2_setattr_simple(ip, attr);
1379 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1380 return error;
1381 }
1382
linux v2.6.22.y-op