xref: /openbmc/linux/fs/gfs2/inode.c (revision 22fd411a)
1 /*
2  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
3  * Copyright (C) 2004-2008 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/security.h>
20 #include <linux/time.h>
21 
22 #include "gfs2.h"
23 #include "incore.h"
24 #include "acl.h"
25 #include "bmap.h"
26 #include "dir.h"
27 #include "xattr.h"
28 #include "glock.h"
29 #include "glops.h"
30 #include "inode.h"
31 #include "log.h"
32 #include "meta_io.h"
33 #include "quota.h"
34 #include "rgrp.h"
35 #include "trans.h"
36 #include "util.h"
37 
38 struct gfs2_inum_range_host {
39 	u64 ir_start;
40 	u64 ir_length;
41 };
42 
43 static int iget_test(struct inode *inode, void *opaque)
44 {
45 	struct gfs2_inode *ip = GFS2_I(inode);
46 	u64 *no_addr = opaque;
47 
48 	if (ip->i_no_addr == *no_addr)
49 		return 1;
50 
51 	return 0;
52 }
53 
54 static int iget_set(struct inode *inode, void *opaque)
55 {
56 	struct gfs2_inode *ip = GFS2_I(inode);
57 	u64 *no_addr = opaque;
58 
59 	inode->i_ino = (unsigned long)*no_addr;
60 	ip->i_no_addr = *no_addr;
61 	return 0;
62 }
63 
64 struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr)
65 {
66 	unsigned long hash = (unsigned long)no_addr;
67 	return ilookup5(sb, hash, iget_test, &no_addr);
68 }
69 
70 static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr)
71 {
72 	unsigned long hash = (unsigned long)no_addr;
73 	return iget5_locked(sb, hash, iget_test, iget_set, &no_addr);
74 }
75 
76 /**
77  * gfs2_set_iop - Sets inode operations
78  * @inode: The inode with correct i_mode filled in
79  *
80  * GFS2 lookup code fills in vfs inode contents based on info obtained
81  * from directory entry inside gfs2_inode_lookup().
82  */
83 
84 static void gfs2_set_iop(struct inode *inode)
85 {
86 	struct gfs2_sbd *sdp = GFS2_SB(inode);
87 	umode_t mode = inode->i_mode;
88 
89 	if (S_ISREG(mode)) {
90 		inode->i_op = &gfs2_file_iops;
91 		if (gfs2_localflocks(sdp))
92 			inode->i_fop = &gfs2_file_fops_nolock;
93 		else
94 			inode->i_fop = &gfs2_file_fops;
95 	} else if (S_ISDIR(mode)) {
96 		inode->i_op = &gfs2_dir_iops;
97 		if (gfs2_localflocks(sdp))
98 			inode->i_fop = &gfs2_dir_fops_nolock;
99 		else
100 			inode->i_fop = &gfs2_dir_fops;
101 	} else if (S_ISLNK(mode)) {
102 		inode->i_op = &gfs2_symlink_iops;
103 	} else {
104 		inode->i_op = &gfs2_file_iops;
105 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
106 	}
107 }
108 
109 /**
110  * gfs2_inode_lookup - Lookup an inode
111  * @sb: The super block
112  * @no_addr: The inode number
113  * @type: The type of the inode
114  *
115  * Returns: A VFS inode, or an error
116  */
117 
118 struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
119 				u64 no_addr, u64 no_formal_ino)
120 {
121 	struct inode *inode;
122 	struct gfs2_inode *ip;
123 	struct gfs2_glock *io_gl = NULL;
124 	int error;
125 
126 	inode = gfs2_iget(sb, no_addr);
127 	ip = GFS2_I(inode);
128 
129 	if (!inode)
130 		return ERR_PTR(-ENOBUFS);
131 
132 	if (inode->i_state & I_NEW) {
133 		struct gfs2_sbd *sdp = GFS2_SB(inode);
134 		ip->i_no_formal_ino = no_formal_ino;
135 
136 		error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
137 		if (unlikely(error))
138 			goto fail;
139 		ip->i_gl->gl_object = ip;
140 
141 		error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
142 		if (unlikely(error))
143 			goto fail_put;
144 
145 		set_bit(GIF_INVALID, &ip->i_flags);
146 		error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
147 		if (unlikely(error))
148 			goto fail_iopen;
149 
150 		ip->i_iopen_gh.gh_gl->gl_object = ip;
151 		gfs2_glock_put(io_gl);
152 		io_gl = NULL;
153 
154 		if (type == DT_UNKNOWN) {
155 			/* Inode glock must be locked already */
156 			error = gfs2_inode_refresh(GFS2_I(inode));
157 			if (error)
158 				goto fail_refresh;
159 		} else {
160 			inode->i_mode = DT2IF(type);
161 		}
162 
163 		gfs2_set_iop(inode);
164 		unlock_new_inode(inode);
165 	}
166 
167 	return inode;
168 
169 fail_refresh:
170 	ip->i_iopen_gh.gh_gl->gl_object = NULL;
171 	gfs2_glock_dq_uninit(&ip->i_iopen_gh);
172 fail_iopen:
173 	if (io_gl)
174 		gfs2_glock_put(io_gl);
175 fail_put:
176 	ip->i_gl->gl_object = NULL;
177 	gfs2_glock_put(ip->i_gl);
178 fail:
179 	iget_failed(inode);
180 	return ERR_PTR(error);
181 }
182 
183 struct inode *gfs2_lookup_by_inum(struct gfs2_sbd *sdp, u64 no_addr,
184 				  u64 *no_formal_ino, unsigned int blktype)
185 {
186 	struct super_block *sb = sdp->sd_vfs;
187 	struct gfs2_holder i_gh;
188 	struct inode *inode;
189 	int error;
190 
191 	error = gfs2_glock_nq_num(sdp, no_addr, &gfs2_inode_glops,
192 				  LM_ST_SHARED, LM_FLAG_ANY, &i_gh);
193 	if (error)
194 		return ERR_PTR(error);
195 
196 	error = gfs2_check_blk_type(sdp, no_addr, blktype);
197 	if (error)
198 		goto fail;
199 
200 	inode = gfs2_inode_lookup(sb, DT_UNKNOWN, no_addr, 0);
201 	if (IS_ERR(inode))
202 		goto fail;
203 
204 	/* Two extra checks for NFS only */
205 	if (no_formal_ino) {
206 		error = -ESTALE;
207 		if (GFS2_I(inode)->i_no_formal_ino != *no_formal_ino)
208 			goto fail_iput;
209 
210 		error = -EIO;
211 		if (GFS2_I(inode)->i_diskflags & GFS2_DIF_SYSTEM)
212 			goto fail_iput;
213 
214 		error = 0;
215 	}
216 
217 fail:
218 	gfs2_glock_dq_uninit(&i_gh);
219 	return error ? ERR_PTR(error) : inode;
220 fail_iput:
221 	iput(inode);
222 	goto fail;
223 }
224 
225 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
226 {
227 	const struct gfs2_dinode *str = buf;
228 	struct timespec atime;
229 	u16 height, depth;
230 
231 	if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
232 		goto corrupt;
233 	ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
234 	ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
235 	ip->i_inode.i_rdev = 0;
236 	switch (ip->i_inode.i_mode & S_IFMT) {
237 	case S_IFBLK:
238 	case S_IFCHR:
239 		ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
240 					   be32_to_cpu(str->di_minor));
241 		break;
242 	};
243 
244 	ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
245 	ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
246 	/*
247 	 * We will need to review setting the nlink count here in the
248 	 * light of the forthcoming ro bind mount work. This is a reminder
249 	 * to do that.
250 	 */
251 	ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
252 	i_size_write(&ip->i_inode, be64_to_cpu(str->di_size));
253 	gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
254 	atime.tv_sec = be64_to_cpu(str->di_atime);
255 	atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
256 	if (timespec_compare(&ip->i_inode.i_atime, &atime) < 0)
257 		ip->i_inode.i_atime = atime;
258 	ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
259 	ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
260 	ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
261 	ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
262 
263 	ip->i_goal = be64_to_cpu(str->di_goal_meta);
264 	ip->i_generation = be64_to_cpu(str->di_generation);
265 
266 	ip->i_diskflags = be32_to_cpu(str->di_flags);
267 	gfs2_set_inode_flags(&ip->i_inode);
268 	height = be16_to_cpu(str->di_height);
269 	if (unlikely(height > GFS2_MAX_META_HEIGHT))
270 		goto corrupt;
271 	ip->i_height = (u8)height;
272 
273 	depth = be16_to_cpu(str->di_depth);
274 	if (unlikely(depth > GFS2_DIR_MAX_DEPTH))
275 		goto corrupt;
276 	ip->i_depth = (u8)depth;
277 	ip->i_entries = be32_to_cpu(str->di_entries);
278 
279 	ip->i_eattr = be64_to_cpu(str->di_eattr);
280 	if (S_ISREG(ip->i_inode.i_mode))
281 		gfs2_set_aops(&ip->i_inode);
282 
283 	return 0;
284 corrupt:
285 	if (gfs2_consist_inode(ip))
286 		gfs2_dinode_print(ip);
287 	return -EIO;
288 }
289 
290 /**
291  * gfs2_inode_refresh - Refresh the incore copy of the dinode
292  * @ip: The GFS2 inode
293  *
294  * Returns: errno
295  */
296 
297 int gfs2_inode_refresh(struct gfs2_inode *ip)
298 {
299 	struct buffer_head *dibh;
300 	int error;
301 
302 	error = gfs2_meta_inode_buffer(ip, &dibh);
303 	if (error)
304 		return error;
305 
306 	if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
307 		brelse(dibh);
308 		return -EIO;
309 	}
310 
311 	error = gfs2_dinode_in(ip, dibh->b_data);
312 	brelse(dibh);
313 	clear_bit(GIF_INVALID, &ip->i_flags);
314 
315 	return error;
316 }
317 
318 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
319 {
320 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
321 	struct gfs2_alloc *al;
322 	struct gfs2_rgrpd *rgd;
323 	int error;
324 
325 	if (gfs2_get_inode_blocks(&ip->i_inode) != 1) {
326 		if (gfs2_consist_inode(ip))
327 			gfs2_dinode_print(ip);
328 		return -EIO;
329 	}
330 
331 	al = gfs2_alloc_get(ip);
332 	if (!al)
333 		return -ENOMEM;
334 
335 	error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
336 	if (error)
337 		goto out;
338 
339 	error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
340 	if (error)
341 		goto out_qs;
342 
343 	rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
344 	if (!rgd) {
345 		gfs2_consist_inode(ip);
346 		error = -EIO;
347 		goto out_rindex_relse;
348 	}
349 
350 	error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
351 				   &al->al_rgd_gh);
352 	if (error)
353 		goto out_rindex_relse;
354 
355 	error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
356 	if (error)
357 		goto out_rg_gunlock;
358 
359 	set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
360 	set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
361 
362 	gfs2_free_di(rgd, ip);
363 
364 	gfs2_trans_end(sdp);
365 
366 out_rg_gunlock:
367 	gfs2_glock_dq_uninit(&al->al_rgd_gh);
368 out_rindex_relse:
369 	gfs2_glock_dq_uninit(&al->al_ri_gh);
370 out_qs:
371 	gfs2_quota_unhold(ip);
372 out:
373 	gfs2_alloc_put(ip);
374 	return error;
375 }
376 
377 /**
378  * gfs2_change_nlink - Change nlink count on inode
379  * @ip: The GFS2 inode
380  * @diff: The change in the nlink count required
381  *
382  * Returns: errno
383  */
384 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
385 {
386 	struct buffer_head *dibh;
387 	u32 nlink;
388 	int error;
389 
390 	BUG_ON(diff != 1 && diff != -1);
391 	nlink = ip->i_inode.i_nlink + diff;
392 
393 	/* If we are reducing the nlink count, but the new value ends up being
394 	   bigger than the old one, we must have underflowed. */
395 	if (diff < 0 && nlink > ip->i_inode.i_nlink) {
396 		if (gfs2_consist_inode(ip))
397 			gfs2_dinode_print(ip);
398 		return -EIO;
399 	}
400 
401 	error = gfs2_meta_inode_buffer(ip, &dibh);
402 	if (error)
403 		return error;
404 
405 	if (diff > 0)
406 		inc_nlink(&ip->i_inode);
407 	else
408 		drop_nlink(&ip->i_inode);
409 
410 	ip->i_inode.i_ctime = CURRENT_TIME;
411 
412 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
413 	gfs2_dinode_out(ip, dibh->b_data);
414 	brelse(dibh);
415 	mark_inode_dirty(&ip->i_inode);
416 
417 	if (ip->i_inode.i_nlink == 0)
418 		gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
419 
420 	return error;
421 }
422 
423 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
424 {
425 	struct qstr qstr;
426 	struct inode *inode;
427 	gfs2_str2qstr(&qstr, name);
428 	inode = gfs2_lookupi(dip, &qstr, 1);
429 	/* gfs2_lookupi has inconsistent callers: vfs
430 	 * related routines expect NULL for no entry found,
431 	 * gfs2_lookup_simple callers expect ENOENT
432 	 * and do not check for NULL.
433 	 */
434 	if (inode == NULL)
435 		return ERR_PTR(-ENOENT);
436 	else
437 		return inode;
438 }
439 
440 
441 /**
442  * gfs2_lookupi - Look up a filename in a directory and return its inode
443  * @d_gh: An initialized holder for the directory glock
444  * @name: The name of the inode to look for
445  * @is_root: If 1, ignore the caller's permissions
446  * @i_gh: An uninitialized holder for the new inode glock
447  *
448  * This can be called via the VFS filldir function when NFS is doing
449  * a readdirplus and the inode which its intending to stat isn't
450  * already in cache. In this case we must not take the directory glock
451  * again, since the readdir call will have already taken that lock.
452  *
453  * Returns: errno
454  */
455 
456 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
457 			   int is_root)
458 {
459 	struct super_block *sb = dir->i_sb;
460 	struct gfs2_inode *dip = GFS2_I(dir);
461 	struct gfs2_holder d_gh;
462 	int error = 0;
463 	struct inode *inode = NULL;
464 	int unlock = 0;
465 
466 	if (!name->len || name->len > GFS2_FNAMESIZE)
467 		return ERR_PTR(-ENAMETOOLONG);
468 
469 	if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
470 	    (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
471 	     dir == sb->s_root->d_inode)) {
472 		igrab(dir);
473 		return dir;
474 	}
475 
476 	if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
477 		error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
478 		if (error)
479 			return ERR_PTR(error);
480 		unlock = 1;
481 	}
482 
483 	if (!is_root) {
484 		error = gfs2_permission(dir, MAY_EXEC, 0);
485 		if (error)
486 			goto out;
487 	}
488 
489 	inode = gfs2_dir_search(dir, name);
490 	if (IS_ERR(inode))
491 		error = PTR_ERR(inode);
492 out:
493 	if (unlock)
494 		gfs2_glock_dq_uninit(&d_gh);
495 	if (error == -ENOENT)
496 		return NULL;
497 	return inode ? inode : ERR_PTR(error);
498 }
499 
500 /**
501  * create_ok - OK to create a new on-disk inode here?
502  * @dip:  Directory in which dinode is to be created
503  * @name:  Name of new dinode
504  * @mode:
505  *
506  * Returns: errno
507  */
508 
509 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
510 		     unsigned int mode)
511 {
512 	int error;
513 
514 	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, 0);
515 	if (error)
516 		return error;
517 
518 	/*  Don't create entries in an unlinked directory  */
519 	if (!dip->i_inode.i_nlink)
520 		return -EPERM;
521 
522 	error = gfs2_dir_check(&dip->i_inode, name, NULL);
523 	switch (error) {
524 	case -ENOENT:
525 		error = 0;
526 		break;
527 	case 0:
528 		return -EEXIST;
529 	default:
530 		return error;
531 	}
532 
533 	if (dip->i_entries == (u32)-1)
534 		return -EFBIG;
535 	if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
536 		return -EMLINK;
537 
538 	return 0;
539 }
540 
541 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
542 			       unsigned int *uid, unsigned int *gid)
543 {
544 	if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
545 	    (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
546 		if (S_ISDIR(*mode))
547 			*mode |= S_ISUID;
548 		else if (dip->i_inode.i_uid != current_fsuid())
549 			*mode &= ~07111;
550 		*uid = dip->i_inode.i_uid;
551 	} else
552 		*uid = current_fsuid();
553 
554 	if (dip->i_inode.i_mode & S_ISGID) {
555 		if (S_ISDIR(*mode))
556 			*mode |= S_ISGID;
557 		*gid = dip->i_inode.i_gid;
558 	} else
559 		*gid = current_fsgid();
560 }
561 
562 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
563 {
564 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
565 	int error;
566 
567 	if (gfs2_alloc_get(dip) == NULL)
568 		return -ENOMEM;
569 
570 	dip->i_alloc->al_requested = RES_DINODE;
571 	error = gfs2_inplace_reserve(dip);
572 	if (error)
573 		goto out;
574 
575 	error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
576 	if (error)
577 		goto out_ipreserv;
578 
579 	error = gfs2_alloc_di(dip, no_addr, generation);
580 
581 	gfs2_trans_end(sdp);
582 
583 out_ipreserv:
584 	gfs2_inplace_release(dip);
585 out:
586 	gfs2_alloc_put(dip);
587 	return error;
588 }
589 
590 /**
591  * init_dinode - Fill in a new dinode structure
592  * @dip: the directory this inode is being created in
593  * @gl: The glock covering the new inode
594  * @inum: the inode number
595  * @mode: the file permissions
596  * @uid:
597  * @gid:
598  *
599  */
600 
601 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
602 			const struct gfs2_inum_host *inum, unsigned int mode,
603 			unsigned int uid, unsigned int gid,
604 			const u64 *generation, dev_t dev, struct buffer_head **bhp)
605 {
606 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
607 	struct gfs2_dinode *di;
608 	struct buffer_head *dibh;
609 	struct timespec tv = CURRENT_TIME;
610 
611 	dibh = gfs2_meta_new(gl, inum->no_addr);
612 	gfs2_trans_add_bh(gl, dibh, 1);
613 	gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
614 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
615 	di = (struct gfs2_dinode *)dibh->b_data;
616 
617 	di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
618 	di->di_num.no_addr = cpu_to_be64(inum->no_addr);
619 	di->di_mode = cpu_to_be32(mode);
620 	di->di_uid = cpu_to_be32(uid);
621 	di->di_gid = cpu_to_be32(gid);
622 	di->di_nlink = 0;
623 	di->di_size = 0;
624 	di->di_blocks = cpu_to_be64(1);
625 	di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
626 	di->di_major = cpu_to_be32(MAJOR(dev));
627 	di->di_minor = cpu_to_be32(MINOR(dev));
628 	di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
629 	di->di_generation = cpu_to_be64(*generation);
630 	di->di_flags = 0;
631 
632 	if (S_ISREG(mode)) {
633 		if ((dip->i_diskflags & GFS2_DIF_INHERIT_JDATA) ||
634 		    gfs2_tune_get(sdp, gt_new_files_jdata))
635 			di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
636 	} else if (S_ISDIR(mode)) {
637 		di->di_flags |= cpu_to_be32(dip->i_diskflags &
638 					    GFS2_DIF_INHERIT_JDATA);
639 	}
640 
641 	di->__pad1 = 0;
642 	di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
643 	di->di_height = 0;
644 	di->__pad2 = 0;
645 	di->__pad3 = 0;
646 	di->di_depth = 0;
647 	di->di_entries = 0;
648 	memset(&di->__pad4, 0, sizeof(di->__pad4));
649 	di->di_eattr = 0;
650 	di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
651 	di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
652 	di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
653 	memset(&di->di_reserved, 0, sizeof(di->di_reserved));
654 
655 	set_buffer_uptodate(dibh);
656 
657 	*bhp = dibh;
658 }
659 
660 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
661 		       unsigned int mode, const struct gfs2_inum_host *inum,
662 		       const u64 *generation, dev_t dev, struct buffer_head **bhp)
663 {
664 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
665 	unsigned int uid, gid;
666 	int error;
667 
668 	munge_mode_uid_gid(dip, &mode, &uid, &gid);
669 	if (!gfs2_alloc_get(dip))
670 		return -ENOMEM;
671 
672 	error = gfs2_quota_lock(dip, uid, gid);
673 	if (error)
674 		goto out;
675 
676 	error = gfs2_quota_check(dip, uid, gid);
677 	if (error)
678 		goto out_quota;
679 
680 	error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
681 	if (error)
682 		goto out_quota;
683 
684 	init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
685 	gfs2_quota_change(dip, +1, uid, gid);
686 	gfs2_trans_end(sdp);
687 
688 out_quota:
689 	gfs2_quota_unlock(dip);
690 out:
691 	gfs2_alloc_put(dip);
692 	return error;
693 }
694 
695 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
696 		       struct gfs2_inode *ip)
697 {
698 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
699 	struct gfs2_alloc *al;
700 	int alloc_required;
701 	struct buffer_head *dibh;
702 	int error;
703 
704 	al = gfs2_alloc_get(dip);
705 	if (!al)
706 		return -ENOMEM;
707 
708 	error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
709 	if (error)
710 		goto fail;
711 
712 	error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
713 	if (alloc_required < 0)
714 		goto fail_quota_locks;
715 	if (alloc_required) {
716 		error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
717 		if (error)
718 			goto fail_quota_locks;
719 
720 		al->al_requested = sdp->sd_max_dirres;
721 
722 		error = gfs2_inplace_reserve(dip);
723 		if (error)
724 			goto fail_quota_locks;
725 
726 		error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
727 					 al->al_rgd->rd_length +
728 					 2 * RES_DINODE +
729 					 RES_STATFS + RES_QUOTA, 0);
730 		if (error)
731 			goto fail_ipreserv;
732 	} else {
733 		error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
734 		if (error)
735 			goto fail_quota_locks;
736 	}
737 
738 	error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
739 	if (error)
740 		goto fail_end_trans;
741 
742 	error = gfs2_meta_inode_buffer(ip, &dibh);
743 	if (error)
744 		goto fail_end_trans;
745 	ip->i_inode.i_nlink = 1;
746 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
747 	gfs2_dinode_out(ip, dibh->b_data);
748 	brelse(dibh);
749 	return 0;
750 
751 fail_end_trans:
752 	gfs2_trans_end(sdp);
753 
754 fail_ipreserv:
755 	if (dip->i_alloc->al_rgd)
756 		gfs2_inplace_release(dip);
757 
758 fail_quota_locks:
759 	gfs2_quota_unlock(dip);
760 
761 fail:
762 	gfs2_alloc_put(dip);
763 	return error;
764 }
765 
766 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
767 {
768 	int err;
769 	size_t len;
770 	void *value;
771 	char *name;
772 
773 	err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
774 					   &name, &value, &len);
775 
776 	if (err) {
777 		if (err == -EOPNOTSUPP)
778 			return 0;
779 		return err;
780 	}
781 
782 	err = __gfs2_xattr_set(&ip->i_inode, name, value, len, 0,
783 			       GFS2_EATYPE_SECURITY);
784 	kfree(value);
785 	kfree(name);
786 
787 	return err;
788 }
789 
790 /**
791  * gfs2_createi - Create a new inode
792  * @ghs: An array of two holders
793  * @name: The name of the new file
794  * @mode: the permissions on the new inode
795  *
796  * @ghs[0] is an initialized holder for the directory
797  * @ghs[1] is the holder for the inode lock
798  *
799  * If the return value is not NULL, the glocks on both the directory and the new
800  * file are held.  A transaction has been started and an inplace reservation
801  * is held, as well.
802  *
803  * Returns: An inode
804  */
805 
806 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
807 			   unsigned int mode, dev_t dev)
808 {
809 	struct inode *inode = NULL;
810 	struct gfs2_inode *dip = ghs->gh_gl->gl_object;
811 	struct inode *dir = &dip->i_inode;
812 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
813 	struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
814 	int error;
815 	u64 generation;
816 	struct buffer_head *bh = NULL;
817 
818 	if (!name->len || name->len > GFS2_FNAMESIZE)
819 		return ERR_PTR(-ENAMETOOLONG);
820 
821 	gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
822 	error = gfs2_glock_nq(ghs);
823 	if (error)
824 		goto fail;
825 
826 	error = create_ok(dip, name, mode);
827 	if (error)
828 		goto fail_gunlock;
829 
830 	error = alloc_dinode(dip, &inum.no_addr, &generation);
831 	if (error)
832 		goto fail_gunlock;
833 	inum.no_formal_ino = generation;
834 
835 	error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
836 				  LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
837 	if (error)
838 		goto fail_gunlock;
839 
840 	error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
841 	if (error)
842 		goto fail_gunlock2;
843 
844 	inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), inum.no_addr,
845 				  inum.no_formal_ino);
846 	if (IS_ERR(inode))
847 		goto fail_gunlock2;
848 
849 	error = gfs2_inode_refresh(GFS2_I(inode));
850 	if (error)
851 		goto fail_gunlock2;
852 
853 	error = gfs2_acl_create(dip, inode);
854 	if (error)
855 		goto fail_gunlock2;
856 
857 	error = gfs2_security_init(dip, GFS2_I(inode));
858 	if (error)
859 		goto fail_gunlock2;
860 
861 	error = link_dinode(dip, name, GFS2_I(inode));
862 	if (error)
863 		goto fail_gunlock2;
864 
865 	if (bh)
866 		brelse(bh);
867 	return inode;
868 
869 fail_gunlock2:
870 	gfs2_glock_dq_uninit(ghs + 1);
871 	if (inode && !IS_ERR(inode))
872 		iput(inode);
873 fail_gunlock:
874 	gfs2_glock_dq(ghs);
875 fail:
876 	if (bh)
877 		brelse(bh);
878 	return ERR_PTR(error);
879 }
880 
881 static int __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
882 {
883 	struct inode *inode = &ip->i_inode;
884 	struct buffer_head *dibh;
885 	int error;
886 
887 	error = gfs2_meta_inode_buffer(ip, &dibh);
888 	if (error)
889 		return error;
890 
891 	setattr_copy(inode, attr);
892 	mark_inode_dirty(inode);
893 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
894 	gfs2_dinode_out(ip, dibh->b_data);
895 	brelse(dibh);
896 	return 0;
897 }
898 
899 /**
900  * gfs2_setattr_simple -
901  * @ip:
902  * @attr:
903  *
904  * Called with a reference on the vnode.
905  *
906  * Returns: errno
907  */
908 
909 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
910 {
911 	int error;
912 
913 	if (current->journal_info)
914 		return __gfs2_setattr_simple(ip, attr);
915 
916 	error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
917 	if (error)
918 		return error;
919 
920 	error = __gfs2_setattr_simple(ip, attr);
921 	gfs2_trans_end(GFS2_SB(&ip->i_inode));
922 	return error;
923 }
924 
925 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
926 {
927 	struct gfs2_dinode *str = buf;
928 
929 	str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
930 	str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
931 	str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
932 	str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
933 	str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
934 	str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
935 	str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
936 	str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
937 	str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
938 	str->di_size = cpu_to_be64(i_size_read(&ip->i_inode));
939 	str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
940 	str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
941 	str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
942 	str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
943 
944 	str->di_goal_meta = cpu_to_be64(ip->i_goal);
945 	str->di_goal_data = cpu_to_be64(ip->i_goal);
946 	str->di_generation = cpu_to_be64(ip->i_generation);
947 
948 	str->di_flags = cpu_to_be32(ip->i_diskflags);
949 	str->di_height = cpu_to_be16(ip->i_height);
950 	str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
951 					     !(ip->i_diskflags & GFS2_DIF_EXHASH) ?
952 					     GFS2_FORMAT_DE : 0);
953 	str->di_depth = cpu_to_be16(ip->i_depth);
954 	str->di_entries = cpu_to_be32(ip->i_entries);
955 
956 	str->di_eattr = cpu_to_be64(ip->i_eattr);
957 	str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
958 	str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
959 	str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
960 }
961 
962 void gfs2_dinode_print(const struct gfs2_inode *ip)
963 {
964 	printk(KERN_INFO "  no_formal_ino = %llu\n",
965 	       (unsigned long long)ip->i_no_formal_ino);
966 	printk(KERN_INFO "  no_addr = %llu\n",
967 	       (unsigned long long)ip->i_no_addr);
968 	printk(KERN_INFO "  i_size = %llu\n",
969 	       (unsigned long long)i_size_read(&ip->i_inode));
970 	printk(KERN_INFO "  blocks = %llu\n",
971 	       (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
972 	printk(KERN_INFO "  i_goal = %llu\n",
973 	       (unsigned long long)ip->i_goal);
974 	printk(KERN_INFO "  i_diskflags = 0x%.8X\n", ip->i_diskflags);
975 	printk(KERN_INFO "  i_height = %u\n", ip->i_height);
976 	printk(KERN_INFO "  i_depth = %u\n", ip->i_depth);
977 	printk(KERN_INFO "  i_entries = %u\n", ip->i_entries);
978 	printk(KERN_INFO "  i_eattr = %llu\n",
979 	       (unsigned long long)ip->i_eattr);
980 }
981 
982