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
38struct gfs2_inum_range_host {
39	u64 ir_start;
40	u64 ir_length;
41};
42
43static 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
54static 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
64struct 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
70static 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
76struct gfs2_skip_data {
77	u64	no_addr;
78	int	skipped;
79};
80
81static int iget_skip_test(struct inode *inode, void *opaque)
82{
83	struct gfs2_inode *ip = GFS2_I(inode);
84	struct gfs2_skip_data *data = opaque;
85
86	if (ip->i_no_addr == data->no_addr) {
87		if (inode->i_state & (I_FREEING|I_WILL_FREE)){
88			data->skipped = 1;
89			return 0;
90		}
91		return 1;
92	}
93	return 0;
94}
95
96static int iget_skip_set(struct inode *inode, void *opaque)
97{
98	struct gfs2_inode *ip = GFS2_I(inode);
99	struct gfs2_skip_data *data = opaque;
100
101	if (data->skipped)
102		return 1;
103	inode->i_ino = (unsigned long)(data->no_addr);
104	ip->i_no_addr = data->no_addr;
105	return 0;
106}
107
108static struct inode *gfs2_iget_skip(struct super_block *sb,
109				    u64 no_addr)
110{
111	struct gfs2_skip_data data;
112	unsigned long hash = (unsigned long)no_addr;
113
114	data.no_addr = no_addr;
115	data.skipped = 0;
116	return iget5_locked(sb, hash, iget_skip_test, iget_skip_set, &data);
117}
118
119/**
120 * GFS2 lookup code fills in vfs inode contents based on info obtained
121 * from directory entry inside gfs2_inode_lookup(). This has caused issues
122 * with NFS code path since its get_dentry routine doesn't have the relevant
123 * directory entry when gfs2_inode_lookup() is invoked. Part of the code
124 * segment inside gfs2_inode_lookup code needs to get moved around.
125 *
126 * Clears I_NEW as well.
127 **/
128
129void gfs2_set_iop(struct inode *inode)
130{
131	struct gfs2_sbd *sdp = GFS2_SB(inode);
132	umode_t mode = inode->i_mode;
133
134	if (S_ISREG(mode)) {
135		inode->i_op = &gfs2_file_iops;
136		if (gfs2_localflocks(sdp))
137			inode->i_fop = &gfs2_file_fops_nolock;
138		else
139			inode->i_fop = &gfs2_file_fops;
140	} else if (S_ISDIR(mode)) {
141		inode->i_op = &gfs2_dir_iops;
142		if (gfs2_localflocks(sdp))
143			inode->i_fop = &gfs2_dir_fops_nolock;
144		else
145			inode->i_fop = &gfs2_dir_fops;
146	} else if (S_ISLNK(mode)) {
147		inode->i_op = &gfs2_symlink_iops;
148	} else {
149		inode->i_op = &gfs2_file_iops;
150		init_special_inode(inode, inode->i_mode, inode->i_rdev);
151	}
152
153	unlock_new_inode(inode);
154}
155
156/**
157 * gfs2_inode_lookup - Lookup an inode
158 * @sb: The super block
159 * @no_addr: The inode number
160 * @type: The type of the inode
161 *
162 * Returns: A VFS inode, or an error
163 */
164
165struct inode *gfs2_inode_lookup(struct super_block *sb,
166				unsigned int type,
167				u64 no_addr,
168				u64 no_formal_ino)
169{
170	struct inode *inode;
171	struct gfs2_inode *ip;
172	struct gfs2_glock *io_gl = NULL;
173	int error;
174
175	inode = gfs2_iget(sb, no_addr);
176	ip = GFS2_I(inode);
177
178	if (!inode)
179		return ERR_PTR(-ENOBUFS);
180
181	if (inode->i_state & I_NEW) {
182		struct gfs2_sbd *sdp = GFS2_SB(inode);
183		ip->i_no_formal_ino = no_formal_ino;
184
185		error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
186		if (unlikely(error))
187			goto fail;
188		ip->i_gl->gl_object = ip;
189
190		error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
191		if (unlikely(error))
192			goto fail_put;
193
194		set_bit(GIF_INVALID, &ip->i_flags);
195		error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
196		if (unlikely(error))
197			goto fail_iopen;
198		ip->i_iopen_gh.gh_gl->gl_object = ip;
199
200		gfs2_glock_put(io_gl);
201		io_gl = NULL;
202
203		if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
204			goto gfs2_nfsbypass;
205
206		inode->i_mode = DT2IF(type);
207
208		/*
209		 * We must read the inode in order to work out its type in
210		 * this case. Note that this doesn't happen often as we normally
211		 * know the type beforehand. This code path only occurs during
212		 * unlinked inode recovery (where it is safe to do this glock,
213		 * which is not true in the general case).
214		 */
215		if (type == DT_UNKNOWN) {
216			struct gfs2_holder gh;
217			error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
218			if (unlikely(error))
219				goto fail_glock;
220			/* Inode is now uptodate */
221			gfs2_glock_dq_uninit(&gh);
222		}
223
224		gfs2_set_iop(inode);
225	}
226
227gfs2_nfsbypass:
228	return inode;
229fail_glock:
230	gfs2_glock_dq(&ip->i_iopen_gh);
231fail_iopen:
232	if (io_gl)
233		gfs2_glock_put(io_gl);
234fail_put:
235	if (inode->i_state & I_NEW)
236		ip->i_gl->gl_object = NULL;
237	gfs2_glock_put(ip->i_gl);
238fail:
239	if (inode->i_state & I_NEW)
240		iget_failed(inode);
241	else
242		iput(inode);
243	return ERR_PTR(error);
244}
245
246/**
247 * gfs2_process_unlinked_inode - Lookup an unlinked inode for reclamation
248 *                               and try to reclaim it by doing iput.
249 *
250 * This function assumes no rgrp locks are currently held.
251 *
252 * @sb: The super block
253 * no_addr: The inode number
254 *
255 */
256
257void gfs2_process_unlinked_inode(struct super_block *sb, u64 no_addr)
258{
259	struct gfs2_sbd *sdp;
260	struct gfs2_inode *ip;
261	struct gfs2_glock *io_gl = NULL;
262	int error;
263	struct gfs2_holder gh;
264	struct inode *inode;
265
266	inode = gfs2_iget_skip(sb, no_addr);
267
268	if (!inode)
269		return;
270
271	/* If it's not a new inode, someone's using it, so leave it alone. */
272	if (!(inode->i_state & I_NEW)) {
273		iput(inode);
274		return;
275	}
276
277	ip = GFS2_I(inode);
278	sdp = GFS2_SB(inode);
279	ip->i_no_formal_ino = -1;
280
281	error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
282	if (unlikely(error))
283		goto fail;
284	ip->i_gl->gl_object = ip;
285
286	error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
287	if (unlikely(error))
288		goto fail_put;
289
290	set_bit(GIF_INVALID, &ip->i_flags);
291	error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, LM_FLAG_TRY | GL_EXACT,
292				   &ip->i_iopen_gh);
293	if (unlikely(error))
294		goto fail_iopen;
295
296	ip->i_iopen_gh.gh_gl->gl_object = ip;
297	gfs2_glock_put(io_gl);
298	io_gl = NULL;
299
300	inode->i_mode = DT2IF(DT_UNKNOWN);
301
302	/*
303	 * We must read the inode in order to work out its type in
304	 * this case. Note that this doesn't happen often as we normally
305	 * know the type beforehand. This code path only occurs during
306	 * unlinked inode recovery (where it is safe to do this glock,
307	 * which is not true in the general case).
308	 */
309	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, LM_FLAG_TRY,
310				   &gh);
311	if (unlikely(error))
312		goto fail_glock;
313
314	/* Inode is now uptodate */
315	gfs2_glock_dq_uninit(&gh);
316	gfs2_set_iop(inode);
317
318	/* The iput will cause it to be deleted. */
319	iput(inode);
320	return;
321
322fail_glock:
323	gfs2_glock_dq(&ip->i_iopen_gh);
324fail_iopen:
325	if (io_gl)
326		gfs2_glock_put(io_gl);
327fail_put:
328	ip->i_gl->gl_object = NULL;
329	gfs2_glock_put(ip->i_gl);
330fail:
331	iget_failed(inode);
332	return;
333}
334
335static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
336{
337	const struct gfs2_dinode *str = buf;
338	struct timespec atime;
339	u16 height, depth;
340
341	if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
342		goto corrupt;
343	ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
344	ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
345	ip->i_inode.i_rdev = 0;
346	switch (ip->i_inode.i_mode & S_IFMT) {
347	case S_IFBLK:
348	case S_IFCHR:
349		ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
350					   be32_to_cpu(str->di_minor));
351		break;
352	};
353
354	ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
355	ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
356	/*
357	 * We will need to review setting the nlink count here in the
358	 * light of the forthcoming ro bind mount work. This is a reminder
359	 * to do that.
360	 */
361	ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
362	ip->i_disksize = be64_to_cpu(str->di_size);
363	i_size_write(&ip->i_inode, ip->i_disksize);
364	gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
365	atime.tv_sec = be64_to_cpu(str->di_atime);
366	atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
367	if (timespec_compare(&ip->i_inode.i_atime, &atime) < 0)
368		ip->i_inode.i_atime = atime;
369	ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
370	ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
371	ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
372	ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
373
374	ip->i_goal = be64_to_cpu(str->di_goal_meta);
375	ip->i_generation = be64_to_cpu(str->di_generation);
376
377	ip->i_diskflags = be32_to_cpu(str->di_flags);
378	gfs2_set_inode_flags(&ip->i_inode);
379	height = be16_to_cpu(str->di_height);
380	if (unlikely(height > GFS2_MAX_META_HEIGHT))
381		goto corrupt;
382	ip->i_height = (u8)height;
383
384	depth = be16_to_cpu(str->di_depth);
385	if (unlikely(depth > GFS2_DIR_MAX_DEPTH))
386		goto corrupt;
387	ip->i_depth = (u8)depth;
388	ip->i_entries = be32_to_cpu(str->di_entries);
389
390	ip->i_eattr = be64_to_cpu(str->di_eattr);
391	if (S_ISREG(ip->i_inode.i_mode))
392		gfs2_set_aops(&ip->i_inode);
393
394	return 0;
395corrupt:
396	if (gfs2_consist_inode(ip))
397		gfs2_dinode_print(ip);
398	return -EIO;
399}
400
401/**
402 * gfs2_inode_refresh - Refresh the incore copy of the dinode
403 * @ip: The GFS2 inode
404 *
405 * Returns: errno
406 */
407
408int gfs2_inode_refresh(struct gfs2_inode *ip)
409{
410	struct buffer_head *dibh;
411	int error;
412
413	error = gfs2_meta_inode_buffer(ip, &dibh);
414	if (error)
415		return error;
416
417	if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
418		brelse(dibh);
419		return -EIO;
420	}
421
422	error = gfs2_dinode_in(ip, dibh->b_data);
423	brelse(dibh);
424	clear_bit(GIF_INVALID, &ip->i_flags);
425
426	return error;
427}
428
429int gfs2_dinode_dealloc(struct gfs2_inode *ip)
430{
431	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
432	struct gfs2_alloc *al;
433	struct gfs2_rgrpd *rgd;
434	int error;
435
436	if (gfs2_get_inode_blocks(&ip->i_inode) != 1) {
437		if (gfs2_consist_inode(ip))
438			gfs2_dinode_print(ip);
439		return -EIO;
440	}
441
442	al = gfs2_alloc_get(ip);
443	if (!al)
444		return -ENOMEM;
445
446	error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
447	if (error)
448		goto out;
449
450	error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
451	if (error)
452		goto out_qs;
453
454	rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
455	if (!rgd) {
456		gfs2_consist_inode(ip);
457		error = -EIO;
458		goto out_rindex_relse;
459	}
460
461	error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
462				   &al->al_rgd_gh);
463	if (error)
464		goto out_rindex_relse;
465
466	error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
467	if (error)
468		goto out_rg_gunlock;
469
470	set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
471	set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
472
473	gfs2_free_di(rgd, ip);
474
475	gfs2_trans_end(sdp);
476
477out_rg_gunlock:
478	gfs2_glock_dq_uninit(&al->al_rgd_gh);
479out_rindex_relse:
480	gfs2_glock_dq_uninit(&al->al_ri_gh);
481out_qs:
482	gfs2_quota_unhold(ip);
483out:
484	gfs2_alloc_put(ip);
485	return error;
486}
487
488/**
489 * gfs2_change_nlink - Change nlink count on inode
490 * @ip: The GFS2 inode
491 * @diff: The change in the nlink count required
492 *
493 * Returns: errno
494 */
495int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
496{
497	struct buffer_head *dibh;
498	u32 nlink;
499	int error;
500
501	BUG_ON(diff != 1 && diff != -1);
502	nlink = ip->i_inode.i_nlink + diff;
503
504	/* If we are reducing the nlink count, but the new value ends up being
505	   bigger than the old one, we must have underflowed. */
506	if (diff < 0 && nlink > ip->i_inode.i_nlink) {
507		if (gfs2_consist_inode(ip))
508			gfs2_dinode_print(ip);
509		return -EIO;
510	}
511
512	error = gfs2_meta_inode_buffer(ip, &dibh);
513	if (error)
514		return error;
515
516	if (diff > 0)
517		inc_nlink(&ip->i_inode);
518	else
519		drop_nlink(&ip->i_inode);
520
521	ip->i_inode.i_ctime = CURRENT_TIME;
522
523	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
524	gfs2_dinode_out(ip, dibh->b_data);
525	brelse(dibh);
526	mark_inode_dirty(&ip->i_inode);
527
528	if (ip->i_inode.i_nlink == 0)
529		gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
530
531	return error;
532}
533
534struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
535{
536	struct qstr qstr;
537	struct inode *inode;
538	gfs2_str2qstr(&qstr, name);
539	inode = gfs2_lookupi(dip, &qstr, 1);
540	/* gfs2_lookupi has inconsistent callers: vfs
541	 * related routines expect NULL for no entry found,
542	 * gfs2_lookup_simple callers expect ENOENT
543	 * and do not check for NULL.
544	 */
545	if (inode == NULL)
546		return ERR_PTR(-ENOENT);
547	else
548		return inode;
549}
550
551
552/**
553 * gfs2_lookupi - Look up a filename in a directory and return its inode
554 * @d_gh: An initialized holder for the directory glock
555 * @name: The name of the inode to look for
556 * @is_root: If 1, ignore the caller's permissions
557 * @i_gh: An uninitialized holder for the new inode glock
558 *
559 * This can be called via the VFS filldir function when NFS is doing
560 * a readdirplus and the inode which its intending to stat isn't
561 * already in cache. In this case we must not take the directory glock
562 * again, since the readdir call will have already taken that lock.
563 *
564 * Returns: errno
565 */
566
567struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
568			   int is_root)
569{
570	struct super_block *sb = dir->i_sb;
571	struct gfs2_inode *dip = GFS2_I(dir);
572	struct gfs2_holder d_gh;
573	int error = 0;
574	struct inode *inode = NULL;
575	int unlock = 0;
576
577	if (!name->len || name->len > GFS2_FNAMESIZE)
578		return ERR_PTR(-ENAMETOOLONG);
579
580	if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
581	    (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
582	     dir == sb->s_root->d_inode)) {
583		igrab(dir);
584		return dir;
585	}
586
587	if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
588		error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
589		if (error)
590			return ERR_PTR(error);
591		unlock = 1;
592	}
593
594	if (!is_root) {
595		error = gfs2_permission(dir, MAY_EXEC);
596		if (error)
597			goto out;
598	}
599
600	inode = gfs2_dir_search(dir, name);
601	if (IS_ERR(inode))
602		error = PTR_ERR(inode);
603out:
604	if (unlock)
605		gfs2_glock_dq_uninit(&d_gh);
606	if (error == -ENOENT)
607		return NULL;
608	return inode ? inode : ERR_PTR(error);
609}
610
611/**
612 * create_ok - OK to create a new on-disk inode here?
613 * @dip:  Directory in which dinode is to be created
614 * @name:  Name of new dinode
615 * @mode:
616 *
617 * Returns: errno
618 */
619
620static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
621		     unsigned int mode)
622{
623	int error;
624
625	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
626	if (error)
627		return error;
628
629	/*  Don't create entries in an unlinked directory  */
630	if (!dip->i_inode.i_nlink)
631		return -EPERM;
632
633	error = gfs2_dir_check(&dip->i_inode, name, NULL);
634	switch (error) {
635	case -ENOENT:
636		error = 0;
637		break;
638	case 0:
639		return -EEXIST;
640	default:
641		return error;
642	}
643
644	if (dip->i_entries == (u32)-1)
645		return -EFBIG;
646	if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
647		return -EMLINK;
648
649	return 0;
650}
651
652static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
653			       unsigned int *uid, unsigned int *gid)
654{
655	if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
656	    (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
657		if (S_ISDIR(*mode))
658			*mode |= S_ISUID;
659		else if (dip->i_inode.i_uid != current_fsuid())
660			*mode &= ~07111;
661		*uid = dip->i_inode.i_uid;
662	} else
663		*uid = current_fsuid();
664
665	if (dip->i_inode.i_mode & S_ISGID) {
666		if (S_ISDIR(*mode))
667			*mode |= S_ISGID;
668		*gid = dip->i_inode.i_gid;
669	} else
670		*gid = current_fsgid();
671}
672
673static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
674{
675	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
676	int error;
677
678	if (gfs2_alloc_get(dip) == NULL)
679		return -ENOMEM;
680
681	dip->i_alloc->al_requested = RES_DINODE;
682	error = gfs2_inplace_reserve(dip);
683	if (error)
684		goto out;
685
686	error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
687	if (error)
688		goto out_ipreserv;
689
690	error = gfs2_alloc_di(dip, no_addr, generation);
691
692	gfs2_trans_end(sdp);
693
694out_ipreserv:
695	gfs2_inplace_release(dip);
696out:
697	gfs2_alloc_put(dip);
698	return error;
699}
700
701/**
702 * init_dinode - Fill in a new dinode structure
703 * @dip: the directory this inode is being created in
704 * @gl: The glock covering the new inode
705 * @inum: the inode number
706 * @mode: the file permissions
707 * @uid:
708 * @gid:
709 *
710 */
711
712static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
713			const struct gfs2_inum_host *inum, unsigned int mode,
714			unsigned int uid, unsigned int gid,
715			const u64 *generation, dev_t dev, struct buffer_head **bhp)
716{
717	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
718	struct gfs2_dinode *di;
719	struct buffer_head *dibh;
720	struct timespec tv = CURRENT_TIME;
721
722	dibh = gfs2_meta_new(gl, inum->no_addr);
723	gfs2_trans_add_bh(gl, dibh, 1);
724	gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
725	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
726	di = (struct gfs2_dinode *)dibh->b_data;
727
728	di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
729	di->di_num.no_addr = cpu_to_be64(inum->no_addr);
730	di->di_mode = cpu_to_be32(mode);
731	di->di_uid = cpu_to_be32(uid);
732	di->di_gid = cpu_to_be32(gid);
733	di->di_nlink = 0;
734	di->di_size = 0;
735	di->di_blocks = cpu_to_be64(1);
736	di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
737	di->di_major = cpu_to_be32(MAJOR(dev));
738	di->di_minor = cpu_to_be32(MINOR(dev));
739	di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
740	di->di_generation = cpu_to_be64(*generation);
741	di->di_flags = 0;
742
743	if (S_ISREG(mode)) {
744		if ((dip->i_diskflags & GFS2_DIF_INHERIT_JDATA) ||
745		    gfs2_tune_get(sdp, gt_new_files_jdata))
746			di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
747	} else if (S_ISDIR(mode)) {
748		di->di_flags |= cpu_to_be32(dip->i_diskflags &
749					    GFS2_DIF_INHERIT_JDATA);
750	}
751
752	di->__pad1 = 0;
753	di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
754	di->di_height = 0;
755	di->__pad2 = 0;
756	di->__pad3 = 0;
757	di->di_depth = 0;
758	di->di_entries = 0;
759	memset(&di->__pad4, 0, sizeof(di->__pad4));
760	di->di_eattr = 0;
761	di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
762	di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
763	di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
764	memset(&di->di_reserved, 0, sizeof(di->di_reserved));
765
766	set_buffer_uptodate(dibh);
767
768	*bhp = dibh;
769}
770
771static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
772		       unsigned int mode, const struct gfs2_inum_host *inum,
773		       const u64 *generation, dev_t dev, struct buffer_head **bhp)
774{
775	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
776	unsigned int uid, gid;
777	int error;
778
779	munge_mode_uid_gid(dip, &mode, &uid, &gid);
780	if (!gfs2_alloc_get(dip))
781		return -ENOMEM;
782
783	error = gfs2_quota_lock(dip, uid, gid);
784	if (error)
785		goto out;
786
787	error = gfs2_quota_check(dip, uid, gid);
788	if (error)
789		goto out_quota;
790
791	error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
792	if (error)
793		goto out_quota;
794
795	init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
796	gfs2_quota_change(dip, +1, uid, gid);
797	gfs2_trans_end(sdp);
798
799out_quota:
800	gfs2_quota_unlock(dip);
801out:
802	gfs2_alloc_put(dip);
803	return error;
804}
805
806static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
807		       struct gfs2_inode *ip)
808{
809	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
810	struct gfs2_alloc *al;
811	int alloc_required;
812	struct buffer_head *dibh;
813	int error;
814
815	al = gfs2_alloc_get(dip);
816	if (!al)
817		return -ENOMEM;
818
819	error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
820	if (error)
821		goto fail;
822
823	error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
824	if (alloc_required < 0)
825		goto fail_quota_locks;
826	if (alloc_required) {
827		error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
828		if (error)
829			goto fail_quota_locks;
830
831		al->al_requested = sdp->sd_max_dirres;
832
833		error = gfs2_inplace_reserve(dip);
834		if (error)
835			goto fail_quota_locks;
836
837		error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
838					 al->al_rgd->rd_length +
839					 2 * RES_DINODE +
840					 RES_STATFS + RES_QUOTA, 0);
841		if (error)
842			goto fail_ipreserv;
843	} else {
844		error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
845		if (error)
846			goto fail_quota_locks;
847	}
848
849	error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
850	if (error)
851		goto fail_end_trans;
852
853	error = gfs2_meta_inode_buffer(ip, &dibh);
854	if (error)
855		goto fail_end_trans;
856	ip->i_inode.i_nlink = 1;
857	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
858	gfs2_dinode_out(ip, dibh->b_data);
859	brelse(dibh);
860	return 0;
861
862fail_end_trans:
863	gfs2_trans_end(sdp);
864
865fail_ipreserv:
866	if (dip->i_alloc->al_rgd)
867		gfs2_inplace_release(dip);
868
869fail_quota_locks:
870	gfs2_quota_unlock(dip);
871
872fail:
873	gfs2_alloc_put(dip);
874	return error;
875}
876
877static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
878{
879	int err;
880	size_t len;
881	void *value;
882	char *name;
883
884	err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
885					   &name, &value, &len);
886
887	if (err) {
888		if (err == -EOPNOTSUPP)
889			return 0;
890		return err;
891	}
892
893	err = __gfs2_xattr_set(&ip->i_inode, name, value, len, 0,
894			       GFS2_EATYPE_SECURITY);
895	kfree(value);
896	kfree(name);
897
898	return err;
899}
900
901/**
902 * gfs2_createi - Create a new inode
903 * @ghs: An array of two holders
904 * @name: The name of the new file
905 * @mode: the permissions on the new inode
906 *
907 * @ghs[0] is an initialized holder for the directory
908 * @ghs[1] is the holder for the inode lock
909 *
910 * If the return value is not NULL, the glocks on both the directory and the new
911 * file are held.  A transaction has been started and an inplace reservation
912 * is held, as well.
913 *
914 * Returns: An inode
915 */
916
917struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
918			   unsigned int mode, dev_t dev)
919{
920	struct inode *inode = NULL;
921	struct gfs2_inode *dip = ghs->gh_gl->gl_object;
922	struct inode *dir = &dip->i_inode;
923	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
924	struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
925	int error;
926	u64 generation;
927	struct buffer_head *bh = NULL;
928
929	if (!name->len || name->len > GFS2_FNAMESIZE)
930		return ERR_PTR(-ENAMETOOLONG);
931
932	gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
933	error = gfs2_glock_nq(ghs);
934	if (error)
935		goto fail;
936
937	error = create_ok(dip, name, mode);
938	if (error)
939		goto fail_gunlock;
940
941	error = alloc_dinode(dip, &inum.no_addr, &generation);
942	if (error)
943		goto fail_gunlock;
944	inum.no_formal_ino = generation;
945
946	error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
947				  LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
948	if (error)
949		goto fail_gunlock;
950
951	error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
952	if (error)
953		goto fail_gunlock2;
954
955	inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), inum.no_addr,
956				  inum.no_formal_ino);
957	if (IS_ERR(inode))
958		goto fail_gunlock2;
959
960	error = gfs2_inode_refresh(GFS2_I(inode));
961	if (error)
962		goto fail_gunlock2;
963
964	error = gfs2_acl_create(dip, inode);
965	if (error)
966		goto fail_gunlock2;
967
968	error = gfs2_security_init(dip, GFS2_I(inode));
969	if (error)
970		goto fail_gunlock2;
971
972	error = link_dinode(dip, name, GFS2_I(inode));
973	if (error)
974		goto fail_gunlock2;
975
976	if (bh)
977		brelse(bh);
978	return inode;
979
980fail_gunlock2:
981	gfs2_glock_dq_uninit(ghs + 1);
982	if (inode && !IS_ERR(inode))
983		iput(inode);
984fail_gunlock:
985	gfs2_glock_dq(ghs);
986fail:
987	if (bh)
988		brelse(bh);
989	return ERR_PTR(error);
990}
991
992static int __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
993{
994	struct inode *inode = &ip->i_inode;
995	struct buffer_head *dibh;
996	int error;
997
998	error = gfs2_meta_inode_buffer(ip, &dibh);
999	if (error)
1000		return error;
1001
1002	if ((attr->ia_valid & ATTR_SIZE) &&
1003	    attr->ia_size != i_size_read(inode)) {
1004		error = vmtruncate(inode, attr->ia_size);
1005		if (error)
1006			return error;
1007	}
1008
1009	setattr_copy(inode, attr);
1010	mark_inode_dirty(inode);
1011
1012	gfs2_assert_warn(GFS2_SB(inode), !error);
1013	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1014	gfs2_dinode_out(ip, dibh->b_data);
1015	brelse(dibh);
1016	return 0;
1017}
1018
1019/**
1020 * gfs2_setattr_simple -
1021 * @ip:
1022 * @attr:
1023 *
1024 * Called with a reference on the vnode.
1025 *
1026 * Returns: errno
1027 */
1028
1029int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1030{
1031	int error;
1032
1033	if (current->journal_info)
1034		return __gfs2_setattr_simple(ip, attr);
1035
1036	error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1037	if (error)
1038		return error;
1039
1040	error = __gfs2_setattr_simple(ip, attr);
1041	gfs2_trans_end(GFS2_SB(&ip->i_inode));
1042	return error;
1043}
1044
1045void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1046{
1047	struct gfs2_dinode *str = buf;
1048
1049	str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1050	str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1051	str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1052	str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1053	str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1054	str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1055	str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1056	str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1057	str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1058	str->di_size = cpu_to_be64(ip->i_disksize);
1059	str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
1060	str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1061	str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1062	str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1063
1064	str->di_goal_meta = cpu_to_be64(ip->i_goal);
1065	str->di_goal_data = cpu_to_be64(ip->i_goal);
1066	str->di_generation = cpu_to_be64(ip->i_generation);
1067
1068	str->di_flags = cpu_to_be32(ip->i_diskflags);
1069	str->di_height = cpu_to_be16(ip->i_height);
1070	str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1071					     !(ip->i_diskflags & GFS2_DIF_EXHASH) ?
1072					     GFS2_FORMAT_DE : 0);
1073	str->di_depth = cpu_to_be16(ip->i_depth);
1074	str->di_entries = cpu_to_be32(ip->i_entries);
1075
1076	str->di_eattr = cpu_to_be64(ip->i_eattr);
1077	str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1078	str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1079	str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1080}
1081
1082void gfs2_dinode_print(const struct gfs2_inode *ip)
1083{
1084	printk(KERN_INFO "  no_formal_ino = %llu\n",
1085	       (unsigned long long)ip->i_no_formal_ino);
1086	printk(KERN_INFO "  no_addr = %llu\n",
1087	       (unsigned long long)ip->i_no_addr);
1088	printk(KERN_INFO "  i_disksize = %llu\n",
1089	       (unsigned long long)ip->i_disksize);
1090	printk(KERN_INFO "  blocks = %llu\n",
1091	       (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
1092	printk(KERN_INFO "  i_goal = %llu\n",
1093	       (unsigned long long)ip->i_goal);
1094	printk(KERN_INFO "  i_diskflags = 0x%.8X\n", ip->i_diskflags);
1095	printk(KERN_INFO "  i_height = %u\n", ip->i_height);
1096	printk(KERN_INFO "  i_depth = %u\n", ip->i_depth);
1097	printk(KERN_INFO "  i_entries = %u\n", ip->i_entries);
1098	printk(KERN_INFO "  i_eattr = %llu\n",
1099	       (unsigned long long)ip->i_eattr);
1100}
1101