?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
IB/mthca: Query port fix
[linux-2.6/verdex.git] / fs / gfs2 / inode.c
blob57c43ac47925dfd66ff39e3fa90f7662decaeb98
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->i_state & I_NEW) {
161 struct gfs2_sbd *sdp = GFS2_SB(inode);
162 umode_t mode = DT2IF(type);
163 inode->i_private = ip;
164 inode->i_mode = mode;
165
166 if (S_ISREG(mode)) {
167 inode->i_op = &gfs2_file_iops;
168 inode->i_fop = &gfs2_file_fops;
169 inode->i_mapping->a_ops = &gfs2_file_aops;
170 } else if (S_ISDIR(mode)) {
171 inode->i_op = &gfs2_dir_iops;
172 inode->i_fop = &gfs2_dir_fops;
173 } else if (S_ISLNK(mode)) {
174 inode->i_op = &gfs2_symlink_iops;
175 } else {
176 inode->i_op = &gfs2_dev_iops;
177 }
178
179 error = gfs2_glock_get(sdp, inum->no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
180 if (unlikely(error))
181 goto fail;
182 ip->i_gl->gl_object = ip;
183
184 error = gfs2_glock_get(sdp, inum->no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
185 if (unlikely(error))
186 goto fail_put;
187
188 ip->i_vn = ip->i_gl->gl_vn - 1;
189 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
190 if (unlikely(error))
191 goto fail_iopen;
192
193 gfs2_glock_put(io_gl);
194 unlock_new_inode(inode);
195 }
196
197 return inode;
198 fail_iopen:
199 gfs2_glock_put(io_gl);
200 fail_put:
201 ip->i_gl->gl_object = NULL;
202 gfs2_glock_put(ip->i_gl);
203 fail:
204 iput(inode);
205 return ERR_PTR(error);
206 }
207
208 /**
209 * gfs2_inode_refresh - Refresh the incore copy of the dinode
210 * @ip: The GFS2 inode
211 *
212 * Returns: errno
213 */
214
215 int gfs2_inode_refresh(struct gfs2_inode *ip)
216 {
217 struct buffer_head *dibh;
218 int error;
219
220 error = gfs2_meta_inode_buffer(ip, &dibh);
221 if (error)
222 return error;
223
224 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
225 brelse(dibh);
226 return -EIO;
227 }
228
229 gfs2_dinode_in(&ip->i_di, dibh->b_data);
230
231 brelse(dibh);
232
233 if (ip->i_num.no_addr != ip->i_di.di_num.no_addr) {
234 if (gfs2_consist_inode(ip))
235 gfs2_dinode_print(&ip->i_di);
236 return -EIO;
237 }
238 if (ip->i_num.no_formal_ino != ip->i_di.di_num.no_formal_ino)
239 return -ESTALE;
240
241 ip->i_vn = ip->i_gl->gl_vn;
242
243 return 0;
244 }
245
246 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
247 {
248 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
249 struct gfs2_alloc *al;
250 struct gfs2_rgrpd *rgd;
251 int error;
252
253 if (ip->i_di.di_blocks != 1) {
254 if (gfs2_consist_inode(ip))
255 gfs2_dinode_print(&ip->i_di);
256 return -EIO;
257 }
258
259 al = gfs2_alloc_get(ip);
260
261 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
262 if (error)
263 goto out;
264
265 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
266 if (error)
267 goto out_qs;
268
269 rgd = gfs2_blk2rgrpd(sdp, ip->i_num.no_addr);
270 if (!rgd) {
271 gfs2_consist_inode(ip);
272 error = -EIO;
273 goto out_rindex_relse;
274 }
275
276 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
277 &al->al_rgd_gh);
278 if (error)
279 goto out_rindex_relse;
280
281 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
282 if (error)
283 goto out_rg_gunlock;
284
285 gfs2_trans_add_gl(ip->i_gl);
286
287 gfs2_free_di(rgd, ip);
288
289 gfs2_trans_end(sdp);
290 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
291
292 out_rg_gunlock:
293 gfs2_glock_dq_uninit(&al->al_rgd_gh);
294 out_rindex_relse:
295 gfs2_glock_dq_uninit(&al->al_ri_gh);
296 out_qs:
297 gfs2_quota_unhold(ip);
298 out:
299 gfs2_alloc_put(ip);
300 return error;
301 }
302
303 /**
304 * gfs2_change_nlink - Change nlink count on inode
305 * @ip: The GFS2 inode
306 * @diff: The change in the nlink count required
307 *
308 * Returns: errno
309 */
310
311 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
312 {
313 struct gfs2_sbd *sdp = ip->i_inode.i_sb->s_fs_info;
314 struct buffer_head *dibh;
315 u32 nlink;
316 int error;
317
318 BUG_ON(ip->i_di.di_nlink != ip->i_inode.i_nlink);
319 nlink = ip->i_di.di_nlink + diff;
320
321 /* If we are reducing the nlink count, but the new value ends up being
322 bigger than the old one, we must have underflowed. */
323 if (diff < 0 && nlink > ip->i_di.di_nlink) {
324 if (gfs2_consist_inode(ip))
325 gfs2_dinode_print(&ip->i_di);
326 return -EIO;
327 }
328
329 error = gfs2_meta_inode_buffer(ip, &dibh);
330 if (error)
331 return error;
332
333 ip->i_di.di_nlink = nlink;
334 ip->i_di.di_ctime = get_seconds();
335 ip->i_inode.i_nlink = nlink;
336
337 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
338 gfs2_dinode_out(&ip->i_di, dibh->b_data);
339 brelse(dibh);
340 mark_inode_dirty(&ip->i_inode);
341
342 if (ip->i_di.di_nlink == 0) {
343 struct gfs2_rgrpd *rgd;
344 struct gfs2_holder ri_gh, rg_gh;
345
346 error = gfs2_rindex_hold(sdp, &ri_gh);
347 if (error)
348 goto out;
349 error = -EIO;
350 rgd = gfs2_blk2rgrpd(sdp, ip->i_num.no_addr);
351 if (!rgd)
352 goto out_norgrp;
353 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &rg_gh);
354 if (error)
355 goto out_norgrp;
356
357 clear_nlink(&ip->i_inode);
358 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
359 gfs2_glock_dq_uninit(&rg_gh);
360 out_norgrp:
361 gfs2_glock_dq_uninit(&ri_gh);
362 }
363 out:
364 return error;
365 }
366
367 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
368 {
369 struct qstr qstr;
370 gfs2_str2qstr(&qstr, name);
371 return gfs2_lookupi(dip, &qstr, 1, NULL);
372 }
373
374
375 /**
376 * gfs2_lookupi - Look up a filename in a directory and return its inode
377 * @d_gh: An initialized holder for the directory glock
378 * @name: The name of the inode to look for
379 * @is_root: If 1, ignore the caller's permissions
380 * @i_gh: An uninitialized holder for the new inode glock
381 *
382 * There will always be a vnode (Linux VFS inode) for the d_gh inode unless
383 * @is_root is true.
384 *
385 * Returns: errno
386 */
387
388 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
389 int is_root, struct nameidata *nd)
390 {
391 struct super_block *sb = dir->i_sb;
392 struct gfs2_inode *dip = GFS2_I(dir);
393 struct gfs2_holder d_gh;
394 struct gfs2_inum inum;
395 unsigned int type;
396 int error = 0;
397 struct inode *inode = NULL;
398
399 if (!name->len || name->len > GFS2_FNAMESIZE)
400 return ERR_PTR(-ENAMETOOLONG);
401
402 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
403 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
404 dir == sb->s_root->d_inode)) {
405 igrab(dir);
406 return dir;
407 }
408
409 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
410 if (error)
411 return ERR_PTR(error);
412
413 if (!is_root) {
414 error = permission(dir, MAY_EXEC, NULL);
415 if (error)
416 goto out;
417 }
418
419 error = gfs2_dir_search(dir, name, &inum, &type);
420 if (error)
421 goto out;
422
423 inode = gfs2_inode_lookup(sb, &inum, type);
424
425 out:
426 gfs2_glock_dq_uninit(&d_gh);
427 if (error == -ENOENT)
428 return NULL;
429 return inode;
430 }
431
432 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
433 {
434 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
435 struct buffer_head *bh;
436 struct gfs2_inum_range ir;
437 int error;
438
439 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
440 if (error)
441 return error;
442 mutex_lock(&sdp->sd_inum_mutex);
443
444 error = gfs2_meta_inode_buffer(ip, &bh);
445 if (error) {
446 mutex_unlock(&sdp->sd_inum_mutex);
447 gfs2_trans_end(sdp);
448 return error;
449 }
450
451 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
452
453 if (ir.ir_length) {
454 *formal_ino = ir.ir_start++;
455 ir.ir_length--;
456 gfs2_trans_add_bh(ip->i_gl, bh, 1);
457 gfs2_inum_range_out(&ir,
458 bh->b_data + sizeof(struct gfs2_dinode));
459 brelse(bh);
460 mutex_unlock(&sdp->sd_inum_mutex);
461 gfs2_trans_end(sdp);
462 return 0;
463 }
464
465 brelse(bh);
466
467 mutex_unlock(&sdp->sd_inum_mutex);
468 gfs2_trans_end(sdp);
469
470 return 1;
471 }
472
473 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
474 {
475 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
476 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
477 struct gfs2_holder gh;
478 struct buffer_head *bh;
479 struct gfs2_inum_range ir;
480 int error;
481
482 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
483 if (error)
484 return error;
485
486 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
487 if (error)
488 goto out;
489 mutex_lock(&sdp->sd_inum_mutex);
490
491 error = gfs2_meta_inode_buffer(ip, &bh);
492 if (error)
493 goto out_end_trans;
494
495 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
496
497 if (!ir.ir_length) {
498 struct buffer_head *m_bh;
499 u64 x, y;
500
501 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
502 if (error)
503 goto out_brelse;
504
505 x = *(u64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
506 x = y = be64_to_cpu(x);
507 ir.ir_start = x;
508 ir.ir_length = GFS2_INUM_QUANTUM;
509 x += GFS2_INUM_QUANTUM;
510 if (x < y)
511 gfs2_consist_inode(m_ip);
512 x = cpu_to_be64(x);
513 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
514 *(u64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = x;
515
516 brelse(m_bh);
517 }
518
519 *formal_ino = ir.ir_start++;
520 ir.ir_length--;
521
522 gfs2_trans_add_bh(ip->i_gl, bh, 1);
523 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
524
525 out_brelse:
526 brelse(bh);
527 out_end_trans:
528 mutex_unlock(&sdp->sd_inum_mutex);
529 gfs2_trans_end(sdp);
530 out:
531 gfs2_glock_dq_uninit(&gh);
532 return error;
533 }
534
535 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
536 {
537 int error;
538
539 error = pick_formal_ino_1(sdp, inum);
540 if (error <= 0)
541 return error;
542
543 error = pick_formal_ino_2(sdp, inum);
544
545 return error;
546 }
547
548 /**
549 * create_ok - OK to create a new on-disk inode here?
550 * @dip: Directory in which dinode is to be created
551 * @name: Name of new dinode
552 * @mode:
553 *
554 * Returns: errno
555 */
556
557 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
558 unsigned int mode)
559 {
560 int error;
561
562 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
563 if (error)
564 return error;
565
566 /* Don't create entries in an unlinked directory */
567 if (!dip->i_di.di_nlink)
568 return -EPERM;
569
570 error = gfs2_dir_search(&dip->i_inode, name, NULL, NULL);
571 switch (error) {
572 case -ENOENT:
573 error = 0;
574 break;
575 case 0:
576 return -EEXIST;
577 default:
578 return error;
579 }
580
581 if (dip->i_di.di_entries == (u32)-1)
582 return -EFBIG;
583 if (S_ISDIR(mode) && dip->i_di.di_nlink == (u32)-1)
584 return -EMLINK;
585
586 return 0;
587 }
588
589 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
590 unsigned int *uid, unsigned int *gid)
591 {
592 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
593 (dip->i_di.di_mode & S_ISUID) && dip->i_di.di_uid) {
594 if (S_ISDIR(*mode))
595 *mode |= S_ISUID;
596 else if (dip->i_di.di_uid != current->fsuid)
597 *mode &= ~07111;
598 *uid = dip->i_di.di_uid;
599 } else
600 *uid = current->fsuid;
601
602 if (dip->i_di.di_mode & S_ISGID) {
603 if (S_ISDIR(*mode))
604 *mode |= S_ISGID;
605 *gid = dip->i_di.di_gid;
606 } else
607 *gid = current->fsgid;
608 }
609
610 static int alloc_dinode(struct gfs2_inode *dip, struct gfs2_inum *inum,
611 u64 *generation)
612 {
613 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
614 int error;
615
616 gfs2_alloc_get(dip);
617
618 dip->i_alloc.al_requested = RES_DINODE;
619 error = gfs2_inplace_reserve(dip);
620 if (error)
621 goto out;
622
623 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
624 if (error)
625 goto out_ipreserv;
626
627 inum->no_addr = gfs2_alloc_di(dip, generation);
628
629 gfs2_trans_end(sdp);
630
631 out_ipreserv:
632 gfs2_inplace_release(dip);
633 out:
634 gfs2_alloc_put(dip);
635 return error;
636 }
637
638 /**
639 * init_dinode - Fill in a new dinode structure
640 * @dip: the directory this inode is being created in
641 * @gl: The glock covering the new inode
642 * @inum: the inode number
643 * @mode: the file permissions
644 * @uid:
645 * @gid:
646 *
647 */
648
649 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
650 const struct gfs2_inum *inum, unsigned int mode,
651 unsigned int uid, unsigned int gid,
652 const u64 *generation)
653 {
654 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
655 struct gfs2_dinode *di;
656 struct buffer_head *dibh;
657
658 dibh = gfs2_meta_new(gl, inum->no_addr);
659 gfs2_trans_add_bh(gl, dibh, 1);
660 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
661 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
662 di = (struct gfs2_dinode *)dibh->b_data;
663
664 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
665 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
666 di->di_mode = cpu_to_be32(mode);
667 di->di_uid = cpu_to_be32(uid);
668 di->di_gid = cpu_to_be32(gid);
669 di->di_nlink = cpu_to_be32(0);
670 di->di_size = cpu_to_be64(0);
671 di->di_blocks = cpu_to_be64(1);
672 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(get_seconds());
673 di->di_major = di->di_minor = cpu_to_be32(0);
674 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
675 di->di_generation = cpu_to_be64(*generation);
676 di->di_flags = cpu_to_be32(0);
677
678 if (S_ISREG(mode)) {
679 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
680 gfs2_tune_get(sdp, gt_new_files_jdata))
681 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
682 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) ||
683 gfs2_tune_get(sdp, gt_new_files_directio))
684 di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO);
685 } else if (S_ISDIR(mode)) {
686 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
687 GFS2_DIF_INHERIT_DIRECTIO);
688 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
689 GFS2_DIF_INHERIT_JDATA);
690 }
691
692 di->__pad1 = 0;
693 di->di_payload_format = cpu_to_be32(0);
694 di->di_height = cpu_to_be32(0);
695 di->__pad2 = 0;
696 di->__pad3 = 0;
697 di->di_depth = cpu_to_be16(0);
698 di->di_entries = cpu_to_be32(0);
699 memset(&di->__pad4, 0, sizeof(di->__pad4));
700 di->di_eattr = cpu_to_be64(0);
701 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
702
703 brelse(dibh);
704 }
705
706 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
707 unsigned int mode, const struct gfs2_inum *inum,
708 const u64 *generation)
709 {
710 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
711 unsigned int uid, gid;
712 int error;
713
714 munge_mode_uid_gid(dip, &mode, &uid, &gid);
715 gfs2_alloc_get(dip);
716
717 error = gfs2_quota_lock(dip, uid, gid);
718 if (error)
719 goto out;
720
721 error = gfs2_quota_check(dip, uid, gid);
722 if (error)
723 goto out_quota;
724
725 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
726 if (error)
727 goto out_quota;
728
729 init_dinode(dip, gl, inum, mode, uid, gid, generation);
730 gfs2_quota_change(dip, +1, uid, gid);
731 gfs2_trans_end(sdp);
732
733 out_quota:
734 gfs2_quota_unlock(dip);
735 out:
736 gfs2_alloc_put(dip);
737 return error;
738 }
739
740 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
741 struct gfs2_inode *ip)
742 {
743 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
744 struct gfs2_alloc *al;
745 int alloc_required;
746 struct buffer_head *dibh;
747 int error;
748
749 al = gfs2_alloc_get(dip);
750
751 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
752 if (error)
753 goto fail;
754
755 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
756 if (alloc_required < 0)
757 goto fail;
758 if (alloc_required) {
759 error = gfs2_quota_check(dip, dip->i_di.di_uid,
760 dip->i_di.di_gid);
761 if (error)
762 goto fail_quota_locks;
763
764 al->al_requested = sdp->sd_max_dirres;
765
766 error = gfs2_inplace_reserve(dip);
767 if (error)
768 goto fail_quota_locks;
769
770 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
771 al->al_rgd->rd_ri.ri_length +
772 2 * RES_DINODE +
773 RES_STATFS + RES_QUOTA, 0);
774 if (error)
775 goto fail_ipreserv;
776 } else {
777 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
778 if (error)
779 goto fail_quota_locks;
780 }
781
782 error = gfs2_dir_add(&dip->i_inode, name, &ip->i_num, IF2DT(ip->i_di.di_mode));
783 if (error)
784 goto fail_end_trans;
785
786 error = gfs2_meta_inode_buffer(ip, &dibh);
787 if (error)
788 goto fail_end_trans;
789 ip->i_di.di_nlink = 1;
790 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
791 gfs2_dinode_out(&ip->i_di, dibh->b_data);
792 brelse(dibh);
793 return 0;
794
795 fail_end_trans:
796 gfs2_trans_end(sdp);
797
798 fail_ipreserv:
799 if (dip->i_alloc.al_rgd)
800 gfs2_inplace_release(dip);
801
802 fail_quota_locks:
803 gfs2_quota_unlock(dip);
804
805 fail:
806 gfs2_alloc_put(dip);
807 return error;
808 }
809
810 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
811 {
812 int err;
813 size_t len;
814 void *value;
815 char *name;
816 struct gfs2_ea_request er;
817
818 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
819 &name, &value, &len);
820
821 if (err) {
822 if (err == -EOPNOTSUPP)
823 return 0;
824 return err;
825 }
826
827 memset(&er, 0, sizeof(struct gfs2_ea_request));
828
829 er.er_type = GFS2_EATYPE_SECURITY;
830 er.er_name = name;
831 er.er_data = value;
832 er.er_name_len = strlen(name);
833 er.er_data_len = len;
834
835 err = gfs2_ea_set_i(ip, &er);
836
837 kfree(value);
838 kfree(name);
839
840 return err;
841 }
842
843 /**
844 * gfs2_createi - Create a new inode
845 * @ghs: An array of two holders
846 * @name: The name of the new file
847 * @mode: the permissions on the new inode
848 *
849 * @ghs[0] is an initialized holder for the directory
850 * @ghs[1] is the holder for the inode lock
851 *
852 * If the return value is not NULL, the glocks on both the directory and the new
853 * file are held. A transaction has been started and an inplace reservation
854 * is held, as well.
855 *
856 * Returns: An inode
857 */
858
859 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
860 unsigned int mode)
861 {
862 struct inode *inode;
863 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
864 struct inode *dir = &dip->i_inode;
865 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
866 struct gfs2_inum inum;
867 int error;
868 u64 generation;
869
870 if (!name->len || name->len > GFS2_FNAMESIZE)
871 return ERR_PTR(-ENAMETOOLONG);
872
873 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
874 error = gfs2_glock_nq(ghs);
875 if (error)
876 goto fail;
877
878 error = create_ok(dip, name, mode);
879 if (error)
880 goto fail_gunlock;
881
882 error = pick_formal_ino(sdp, &inum.no_formal_ino);
883 if (error)
884 goto fail_gunlock;
885
886 error = alloc_dinode(dip, &inum, &generation);
887 if (error)
888 goto fail_gunlock;
889
890 if (inum.no_addr < dip->i_num.no_addr) {
891 gfs2_glock_dq(ghs);
892
893 error = gfs2_glock_nq_num(sdp, inum.no_addr,
894 &gfs2_inode_glops, LM_ST_EXCLUSIVE,
895 GL_SKIP, ghs + 1);
896 if (error) {
897 return ERR_PTR(error);
898 }
899
900 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
901 error = gfs2_glock_nq(ghs);
902 if (error) {
903 gfs2_glock_dq_uninit(ghs + 1);
904 return ERR_PTR(error);
905 }
906
907 error = create_ok(dip, name, mode);
908 if (error)
909 goto fail_gunlock2;
910 } else {
911 error = gfs2_glock_nq_num(sdp, inum.no_addr,
912 &gfs2_inode_glops, LM_ST_EXCLUSIVE,
913 GL_SKIP, ghs + 1);
914 if (error)
915 goto fail_gunlock;
916 }
917
918 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation);
919 if (error)
920 goto fail_gunlock2;
921
922 inode = gfs2_inode_lookup(dir->i_sb, &inum, IF2DT(mode));
923 if (IS_ERR(inode))
924 goto fail_gunlock2;
925
926 error = gfs2_inode_refresh(GFS2_I(inode));
927 if (error)
928 goto fail_iput;
929
930 error = gfs2_acl_create(dip, GFS2_I(inode));
931 if (error)
932 goto fail_iput;
933
934 error = gfs2_security_init(dip, GFS2_I(inode));
935 if (error)
936 goto fail_iput;
937
938 error = link_dinode(dip, name, GFS2_I(inode));
939 if (error)
940 goto fail_iput;
941
942 if (!inode)
943 return ERR_PTR(-ENOMEM);
944 return inode;
945
946 fail_iput:
947 iput(inode);
948 fail_gunlock2:
949 gfs2_glock_dq_uninit(ghs + 1);
950 fail_gunlock:
951 gfs2_glock_dq(ghs);
952 fail:
953 return ERR_PTR(error);
954 }
955
956 /**
957 * gfs2_rmdiri - Remove a directory
958 * @dip: The parent directory of the directory to be removed
959 * @name: The name of the directory to be removed
960 * @ip: The GFS2 inode of the directory to be removed
961 *
962 * Assumes Glocks on dip and ip are held
963 *
964 * Returns: errno
965 */
966
967 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
968 struct gfs2_inode *ip)
969 {
970 struct qstr dotname;
971 int error;
972
973 if (ip->i_di.di_entries != 2) {
974 if (gfs2_consist_inode(ip))
975 gfs2_dinode_print(&ip->i_di);
976 return -EIO;
977 }
978
979 error = gfs2_dir_del(dip, name);
980 if (error)
981 return error;
982
983 error = gfs2_change_nlink(dip, -1);
984 if (error)
985 return error;
986
987 gfs2_str2qstr(&dotname, ".");
988 error = gfs2_dir_del(ip, &dotname);
989 if (error)
990 return error;
991
992 gfs2_str2qstr(&dotname, "..");
993 error = gfs2_dir_del(ip, &dotname);
994 if (error)
995 return error;
996
997 error = gfs2_change_nlink(ip, -2);
998 if (error)
999 return error;
1000
1001 return error;
1002 }
1003
1004 /*
1005 * gfs2_unlink_ok - check to see that a inode is still in a directory
1006 * @dip: the directory
1007 * @name: the name of the file
1008 * @ip: the inode
1009 *
1010 * Assumes that the lock on (at least) @dip is held.
1011 *
1012 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1013 */
1014
1015 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1016 struct gfs2_inode *ip)
1017 {
1018 struct gfs2_inum inum;
1019 unsigned int type;
1020 int error;
1021
1022 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1023 return -EPERM;
1024
1025 if ((dip->i_di.di_mode & S_ISVTX) &&
1026 dip->i_di.di_uid != current->fsuid &&
1027 ip->i_di.di_uid != current->fsuid && !capable(CAP_FOWNER))
1028 return -EPERM;
1029
1030 if (IS_APPEND(&dip->i_inode))
1031 return -EPERM;
1032
1033 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
1034 if (error)
1035 return error;
1036
1037 error = gfs2_dir_search(&dip->i_inode, name, &inum, &type);
1038 if (error)
1039 return error;
1040
1041 if (!gfs2_inum_equal(&inum, &ip->i_num))
1042 return -ENOENT;
1043
1044 if (IF2DT(ip->i_di.di_mode) != type) {
1045 gfs2_consist_inode(dip);
1046 return -EIO;
1047 }
1048
1049 return 0;
1050 }
1051
1052 /*
1053 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1054 * @this: move this
1055 * @to: to here
1056 *
1057 * Follow @to back to the root and make sure we don't encounter @this
1058 * Assumes we already hold the rename lock.
1059 *
1060 * Returns: errno
1061 */
1062
1063 int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to)
1064 {
1065 struct inode *dir = &to->i_inode;
1066 struct super_block *sb = dir->i_sb;
1067 struct inode *tmp;
1068 struct qstr dotdot;
1069 int error = 0;
1070
1071 gfs2_str2qstr(&dotdot, "..");
1072
1073 igrab(dir);
1074
1075 for (;;) {
1076 if (dir == &this->i_inode) {
1077 error = -EINVAL;
1078 break;
1079 }
1080 if (dir == sb->s_root->d_inode) {
1081 error = 0;
1082 break;
1083 }
1084
1085 tmp = gfs2_lookupi(dir, &dotdot, 1, NULL);
1086 if (IS_ERR(tmp)) {
1087 error = PTR_ERR(tmp);
1088 break;
1089 }
1090
1091 iput(dir);
1092 dir = tmp;
1093 }
1094
1095 iput(dir);
1096
1097 return error;
1098 }
1099
1100 /**
1101 * gfs2_readlinki - return the contents of a symlink
1102 * @ip: the symlink's inode
1103 * @buf: a pointer to the buffer to be filled
1104 * @len: a pointer to the length of @buf
1105 *
1106 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1107 * to be freed by the caller.
1108 *
1109 * Returns: errno
1110 */
1111
1112 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1113 {
1114 struct gfs2_holder i_gh;
1115 struct buffer_head *dibh;
1116 unsigned int x;
1117 int error;
1118
1119 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
1120 error = gfs2_glock_nq_atime(&i_gh);
1121 if (error) {
1122 gfs2_holder_uninit(&i_gh);
1123 return error;
1124 }
1125
1126 if (!ip->i_di.di_size) {
1127 gfs2_consist_inode(ip);
1128 error = -EIO;
1129 goto out;
1130 }
1131
1132 error = gfs2_meta_inode_buffer(ip, &dibh);
1133 if (error)
1134 goto out;
1135
1136 x = ip->i_di.di_size + 1;
1137 if (x > *len) {
1138 *buf = kmalloc(x, GFP_KERNEL);
1139 if (!*buf) {
1140 error = -ENOMEM;
1141 goto out_brelse;
1142 }
1143 }
1144
1145 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1146 *len = x;
1147
1148 out_brelse:
1149 brelse(dibh);
1150 out:
1151 gfs2_glock_dq_uninit(&i_gh);
1152 return error;
1153 }
1154
1155 /**
1156 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1157 * conditionally update the inode's atime
1158 * @gh: the holder to acquire
1159 *
1160 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1161 * Update if the difference between the current time and the inode's current
1162 * atime is greater than an interval specified at mount.
1163 *
1164 * Returns: errno
1165 */
1166
1167 int gfs2_glock_nq_atime(struct gfs2_holder *gh)
1168 {
1169 struct gfs2_glock *gl = gh->gh_gl;
1170 struct gfs2_sbd *sdp = gl->gl_sbd;
1171 struct gfs2_inode *ip = gl->gl_object;
1172 s64 curtime, quantum = gfs2_tune_get(sdp, gt_atime_quantum);
1173 unsigned int state;
1174 int flags;
1175 int error;
1176
1177 if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) ||
1178 gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) ||
1179 gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops))
1180 return -EINVAL;
1181
1182 state = gh->gh_state;
1183 flags = gh->gh_flags;
1184
1185 error = gfs2_glock_nq(gh);
1186 if (error)
1187 return error;
1188
1189 if (test_bit(SDF_NOATIME, &sdp->sd_flags) ||
1190 (sdp->sd_vfs->s_flags & MS_RDONLY))
1191 return 0;
1192
1193 curtime = get_seconds();
1194 if (curtime - ip->i_di.di_atime >= quantum) {
1195 gfs2_glock_dq(gh);
1196 gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY,
1197 gh);
1198 error = gfs2_glock_nq(gh);
1199 if (error)
1200 return error;
1201
1202 /* Verify that atime hasn't been updated while we were
1203 trying to get exclusive lock. */
1204
1205 curtime = get_seconds();
1206 if (curtime - ip->i_di.di_atime >= quantum) {
1207 struct buffer_head *dibh;
1208 struct gfs2_dinode *di;
1209
1210 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
1211 if (error == -EROFS)
1212 return 0;
1213 if (error)
1214 goto fail;
1215
1216 error = gfs2_meta_inode_buffer(ip, &dibh);
1217 if (error)
1218 goto fail_end_trans;
1219
1220 ip->i_di.di_atime = curtime;
1221
1222 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1223 di = (struct gfs2_dinode *)dibh->b_data;
1224 di->di_atime = cpu_to_be64(ip->i_di.di_atime);
1225 brelse(dibh);
1226
1227 gfs2_trans_end(sdp);
1228 }
1229
1230 /* If someone else has asked for the glock,
1231 unlock and let them have it. Then reacquire
1232 in the original state. */
1233 if (gfs2_glock_is_blocking(gl)) {
1234 gfs2_glock_dq(gh);
1235 gfs2_holder_reinit(state, flags, gh);
1236 return gfs2_glock_nq(gh);
1237 }
1238 }
1239
1240 return 0;
1241
1242 fail_end_trans:
1243 gfs2_trans_end(sdp);
1244 fail:
1245 gfs2_glock_dq(gh);
1246 return error;
1247 }
1248
1249 /**
1250 * glock_compare_atime - Compare two struct gfs2_glock structures for sort
1251 * @arg_a: the first structure
1252 * @arg_b: the second structure
1253 *
1254 * Returns: 1 if A > B
1255 * -1 if A < B
1256 * 0 if A == B
1257 */
1258
1259 static int glock_compare_atime(const void *arg_a, const void *arg_b)
1260 {
1261 const struct gfs2_holder *gh_a = *(const struct gfs2_holder **)arg_a;
1262 const struct gfs2_holder *gh_b = *(const struct gfs2_holder **)arg_b;
1263 const struct lm_lockname *a = &gh_a->gh_gl->gl_name;
1264 const struct lm_lockname *b = &gh_b->gh_gl->gl_name;
1265
1266 if (a->ln_number > b->ln_number)
1267 return 1;
1268 if (a->ln_number < b->ln_number)
1269 return -1;
1270 if (gh_a->gh_state == LM_ST_SHARED && gh_b->gh_state == LM_ST_EXCLUSIVE)
1271 return 1;
1272 if (gh_a->gh_state == LM_ST_SHARED && (gh_b->gh_flags & GL_ATIME))
1273 return 1;
1274
1275 return 0;
1276 }
1277
1278 /**
1279 * gfs2_glock_nq_m_atime - acquire multiple glocks where one may need an
1280 * atime update
1281 * @num_gh: the number of structures
1282 * @ghs: an array of struct gfs2_holder structures
1283 *
1284 * Returns: 0 on success (all glocks acquired),
1285 * errno on failure (no glocks acquired)
1286 */
1287
1288 int gfs2_glock_nq_m_atime(unsigned int num_gh, struct gfs2_holder *ghs)
1289 {
1290 struct gfs2_holder **p;
1291 unsigned int x;
1292 int error = 0;
1293
1294 if (!num_gh)
1295 return 0;
1296
1297 if (num_gh == 1) {
1298 ghs->gh_flags &= ~(LM_FLAG_TRY | GL_ASYNC);
1299 if (ghs->gh_flags & GL_ATIME)
1300 error = gfs2_glock_nq_atime(ghs);
1301 else
1302 error = gfs2_glock_nq(ghs);
1303 return error;
1304 }
1305
1306 p = kcalloc(num_gh, sizeof(struct gfs2_holder *), GFP_KERNEL);
1307 if (!p)
1308 return -ENOMEM;
1309
1310 for (x = 0; x < num_gh; x++)
1311 p[x] = &ghs[x];
1312
1313 sort(p, num_gh, sizeof(struct gfs2_holder *), glock_compare_atime,NULL);
1314
1315 for (x = 0; x < num_gh; x++) {
1316 p[x]->gh_flags &= ~(LM_FLAG_TRY | GL_ASYNC);
1317
1318 if (p[x]->gh_flags & GL_ATIME)
1319 error = gfs2_glock_nq_atime(p[x]);
1320 else
1321 error = gfs2_glock_nq(p[x]);
1322
1323 if (error) {
1324 while (x--)
1325 gfs2_glock_dq(p[x]);
1326 break;
1327 }
1328 }
1329
1330 kfree(p);
1331 return error;
1332 }
1333
1334
1335 static int
1336 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1337 {
1338 struct buffer_head *dibh;
1339 int error;
1340
1341 error = gfs2_meta_inode_buffer(ip, &dibh);
1342 if (!error) {
1343 error = inode_setattr(&ip->i_inode, attr);
1344 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1345 gfs2_inode_attr_out(ip);
1346
1347 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1348 gfs2_dinode_out(&ip->i_di, dibh->b_data);
1349 brelse(dibh);
1350 }
1351 return error;
1352 }
1353
1354 /**
1355 * gfs2_setattr_simple -
1356 * @ip:
1357 * @attr:
1358 *
1359 * Called with a reference on the vnode.
1360 *
1361 * Returns: errno
1362 */
1363
1364 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1365 {
1366 int error;
1367
1368 if (current->journal_info)
1369 return __gfs2_setattr_simple(ip, attr);
1370
1371 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1372 if (error)
1373 return error;
1374
1375 error = __gfs2_setattr_simple(ip, attr);
1376 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1377 return error;
1378 }
1379
Verdex Pro support