2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
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.
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>
33 #include "ops_address.h"
34 #include "ops_inode.h"
40 struct gfs2_inum_range_host
{
45 static int iget_test(struct inode
*inode
, void *opaque
)
47 struct gfs2_inode
*ip
= GFS2_I(inode
);
48 u64
*no_addr
= opaque
;
50 if (ip
->i_no_addr
== *no_addr
&&
51 inode
->i_private
!= NULL
)
57 static int iget_set(struct inode
*inode
, void *opaque
)
59 struct gfs2_inode
*ip
= GFS2_I(inode
);
60 u64
*no_addr
= opaque
;
62 inode
->i_ino
= (unsigned long)*no_addr
;
63 ip
->i_no_addr
= *no_addr
;
67 struct inode
*gfs2_ilookup(struct super_block
*sb
, u64 no_addr
)
69 unsigned long hash
= (unsigned long)no_addr
;
70 return ilookup5(sb
, hash
, iget_test
, &no_addr
);
73 static struct inode
*gfs2_iget(struct super_block
*sb
, u64 no_addr
)
75 unsigned long hash
= (unsigned long)no_addr
;
76 return iget5_locked(sb
, hash
, iget_test
, iget_set
, &no_addr
);
79 struct gfs2_skip_data
{
84 static int iget_skip_test(struct inode
*inode
, void *opaque
)
86 struct gfs2_inode
*ip
= GFS2_I(inode
);
87 struct gfs2_skip_data
*data
= opaque
;
89 if (ip
->i_no_addr
== data
->no_addr
&& inode
->i_private
!= NULL
){
90 if (inode
->i_state
& (I_FREEING
|I_CLEAR
|I_WILL_FREE
)){
99 static int iget_skip_set(struct inode
*inode
, void *opaque
)
101 struct gfs2_inode
*ip
= GFS2_I(inode
);
102 struct gfs2_skip_data
*data
= opaque
;
106 inode
->i_ino
= (unsigned long)(data
->no_addr
);
107 ip
->i_no_addr
= data
->no_addr
;
111 static struct inode
*gfs2_iget_skip(struct super_block
*sb
,
114 struct gfs2_skip_data data
;
115 unsigned long hash
= (unsigned long)no_addr
;
117 data
.no_addr
= no_addr
;
119 return iget5_locked(sb
, hash
, iget_skip_test
, iget_skip_set
, &data
);
123 * GFS2 lookup code fills in vfs inode contents based on info obtained
124 * from directory entry inside gfs2_inode_lookup(). This has caused issues
125 * with NFS code path since its get_dentry routine doesn't have the relevant
126 * directory entry when gfs2_inode_lookup() is invoked. Part of the code
127 * segment inside gfs2_inode_lookup code needs to get moved around.
129 * Clean up I_LOCK and I_NEW as well.
132 void gfs2_set_iop(struct inode
*inode
)
134 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
135 umode_t mode
= inode
->i_mode
;
138 inode
->i_op
= &gfs2_file_iops
;
139 if (sdp
->sd_args
.ar_localflocks
)
140 inode
->i_fop
= &gfs2_file_fops_nolock
;
142 inode
->i_fop
= &gfs2_file_fops
;
143 } else if (S_ISDIR(mode
)) {
144 inode
->i_op
= &gfs2_dir_iops
;
145 if (sdp
->sd_args
.ar_localflocks
)
146 inode
->i_fop
= &gfs2_dir_fops_nolock
;
148 inode
->i_fop
= &gfs2_dir_fops
;
149 } else if (S_ISLNK(mode
)) {
150 inode
->i_op
= &gfs2_symlink_iops
;
152 inode
->i_op
= &gfs2_dev_iops
;
155 unlock_new_inode(inode
);
159 * gfs2_inode_lookup - Lookup an inode
160 * @sb: The super block
161 * @no_addr: The inode number
162 * @type: The type of the inode
163 * @skip_freeing: set this not return an inode if it is currently being freed.
165 * Returns: A VFS inode, or an error
168 struct inode
*gfs2_inode_lookup(struct super_block
*sb
,
171 u64 no_formal_ino
, int skip_freeing
)
174 struct gfs2_inode
*ip
;
175 struct gfs2_glock
*io_gl
;
179 inode
= gfs2_iget_skip(sb
, no_addr
);
181 inode
= gfs2_iget(sb
, no_addr
);
185 return ERR_PTR(-ENOBUFS
);
187 if (inode
->i_state
& I_NEW
) {
188 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
189 inode
->i_private
= ip
;
190 ip
->i_no_formal_ino
= no_formal_ino
;
192 error
= gfs2_glock_get(sdp
, no_addr
, &gfs2_inode_glops
, CREATE
, &ip
->i_gl
);
195 ip
->i_gl
->gl_object
= ip
;
197 error
= gfs2_glock_get(sdp
, no_addr
, &gfs2_iopen_glops
, CREATE
, &io_gl
);
201 set_bit(GIF_INVALID
, &ip
->i_flags
);
202 error
= gfs2_glock_nq_init(io_gl
, LM_ST_SHARED
, GL_EXACT
, &ip
->i_iopen_gh
);
205 ip
->i_iopen_gh
.gh_gl
->gl_object
= ip
;
207 gfs2_glock_put(io_gl
);
209 if ((type
== DT_UNKNOWN
) && (no_formal_ino
== 0))
212 inode
->i_mode
= DT2IF(type
);
215 * We must read the inode in order to work out its type in
216 * this case. Note that this doesn't happen often as we normally
217 * know the type beforehand. This code path only occurs during
218 * unlinked inode recovery (where it is safe to do this glock,
219 * which is not true in the general case).
221 if (type
== DT_UNKNOWN
) {
222 struct gfs2_holder gh
;
223 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
226 /* Inode is now uptodate */
227 gfs2_glock_dq_uninit(&gh
);
236 gfs2_glock_dq(&ip
->i_iopen_gh
);
238 gfs2_glock_put(io_gl
);
240 ip
->i_gl
->gl_object
= NULL
;
241 gfs2_glock_put(ip
->i_gl
);
244 return ERR_PTR(error
);
247 static int gfs2_dinode_in(struct gfs2_inode
*ip
, const void *buf
)
249 struct gfs2_dinode_host
*di
= &ip
->i_di
;
250 const struct gfs2_dinode
*str
= buf
;
253 if (unlikely(ip
->i_no_addr
!= be64_to_cpu(str
->di_num
.no_addr
)))
255 ip
->i_no_formal_ino
= be64_to_cpu(str
->di_num
.no_formal_ino
);
256 ip
->i_inode
.i_mode
= be32_to_cpu(str
->di_mode
);
257 ip
->i_inode
.i_rdev
= 0;
258 switch (ip
->i_inode
.i_mode
& S_IFMT
) {
261 ip
->i_inode
.i_rdev
= MKDEV(be32_to_cpu(str
->di_major
),
262 be32_to_cpu(str
->di_minor
));
266 ip
->i_inode
.i_uid
= be32_to_cpu(str
->di_uid
);
267 ip
->i_inode
.i_gid
= be32_to_cpu(str
->di_gid
);
269 * We will need to review setting the nlink count here in the
270 * light of the forthcoming ro bind mount work. This is a reminder
273 ip
->i_inode
.i_nlink
= be32_to_cpu(str
->di_nlink
);
274 di
->di_size
= be64_to_cpu(str
->di_size
);
275 i_size_write(&ip
->i_inode
, di
->di_size
);
276 di
->di_blocks
= be64_to_cpu(str
->di_blocks
);
277 gfs2_set_inode_blocks(&ip
->i_inode
);
278 ip
->i_inode
.i_atime
.tv_sec
= be64_to_cpu(str
->di_atime
);
279 ip
->i_inode
.i_atime
.tv_nsec
= be32_to_cpu(str
->di_atime_nsec
);
280 ip
->i_inode
.i_mtime
.tv_sec
= be64_to_cpu(str
->di_mtime
);
281 ip
->i_inode
.i_mtime
.tv_nsec
= be32_to_cpu(str
->di_mtime_nsec
);
282 ip
->i_inode
.i_ctime
.tv_sec
= be64_to_cpu(str
->di_ctime
);
283 ip
->i_inode
.i_ctime
.tv_nsec
= be32_to_cpu(str
->di_ctime_nsec
);
285 di
->di_goal_meta
= be64_to_cpu(str
->di_goal_meta
);
286 di
->di_goal_data
= be64_to_cpu(str
->di_goal_data
);
287 di
->di_generation
= be64_to_cpu(str
->di_generation
);
289 di
->di_flags
= be32_to_cpu(str
->di_flags
);
290 gfs2_set_inode_flags(&ip
->i_inode
);
291 height
= be16_to_cpu(str
->di_height
);
292 if (unlikely(height
> GFS2_MAX_META_HEIGHT
))
294 ip
->i_height
= (u8
)height
;
296 depth
= be16_to_cpu(str
->di_depth
);
297 if (unlikely(depth
> GFS2_DIR_MAX_DEPTH
))
299 ip
->i_depth
= (u8
)depth
;
300 di
->di_entries
= be32_to_cpu(str
->di_entries
);
302 di
->di_eattr
= be64_to_cpu(str
->di_eattr
);
303 if (S_ISREG(ip
->i_inode
.i_mode
))
304 gfs2_set_aops(&ip
->i_inode
);
308 if (gfs2_consist_inode(ip
))
309 gfs2_dinode_print(ip
);
314 * gfs2_inode_refresh - Refresh the incore copy of the dinode
315 * @ip: The GFS2 inode
320 int gfs2_inode_refresh(struct gfs2_inode
*ip
)
322 struct buffer_head
*dibh
;
325 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
329 if (gfs2_metatype_check(GFS2_SB(&ip
->i_inode
), dibh
, GFS2_METATYPE_DI
)) {
334 error
= gfs2_dinode_in(ip
, dibh
->b_data
);
336 clear_bit(GIF_INVALID
, &ip
->i_flags
);
341 int gfs2_dinode_dealloc(struct gfs2_inode
*ip
)
343 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
344 struct gfs2_alloc
*al
;
345 struct gfs2_rgrpd
*rgd
;
348 if (ip
->i_di
.di_blocks
!= 1) {
349 if (gfs2_consist_inode(ip
))
350 gfs2_dinode_print(ip
);
354 al
= gfs2_alloc_get(ip
);
356 error
= gfs2_quota_hold(ip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
360 error
= gfs2_rindex_hold(sdp
, &al
->al_ri_gh
);
364 rgd
= gfs2_blk2rgrpd(sdp
, ip
->i_no_addr
);
366 gfs2_consist_inode(ip
);
368 goto out_rindex_relse
;
371 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0,
374 goto out_rindex_relse
;
376 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+ RES_STATFS
+ RES_QUOTA
, 1);
380 set_bit(GLF_DIRTY
, &ip
->i_gl
->gl_flags
);
381 set_bit(GLF_LFLUSH
, &ip
->i_gl
->gl_flags
);
383 gfs2_free_di(rgd
, ip
);
386 clear_bit(GLF_STICKY
, &ip
->i_gl
->gl_flags
);
389 gfs2_glock_dq_uninit(&al
->al_rgd_gh
);
391 gfs2_glock_dq_uninit(&al
->al_ri_gh
);
393 gfs2_quota_unhold(ip
);
400 * gfs2_change_nlink - Change nlink count on inode
401 * @ip: The GFS2 inode
402 * @diff: The change in the nlink count required
406 int gfs2_change_nlink(struct gfs2_inode
*ip
, int diff
)
408 struct buffer_head
*dibh
;
412 BUG_ON(diff
!= 1 && diff
!= -1);
413 nlink
= ip
->i_inode
.i_nlink
+ diff
;
415 /* If we are reducing the nlink count, but the new value ends up being
416 bigger than the old one, we must have underflowed. */
417 if (diff
< 0 && nlink
> ip
->i_inode
.i_nlink
) {
418 if (gfs2_consist_inode(ip
))
419 gfs2_dinode_print(ip
);
423 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
428 inc_nlink(&ip
->i_inode
);
430 drop_nlink(&ip
->i_inode
);
432 ip
->i_inode
.i_ctime
= CURRENT_TIME
;
434 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
435 gfs2_dinode_out(ip
, dibh
->b_data
);
437 mark_inode_dirty(&ip
->i_inode
);
439 if (ip
->i_inode
.i_nlink
== 0)
440 gfs2_unlink_di(&ip
->i_inode
); /* mark inode unlinked */
445 struct inode
*gfs2_lookup_simple(struct inode
*dip
, const char *name
)
449 gfs2_str2qstr(&qstr
, name
);
450 inode
= gfs2_lookupi(dip
, &qstr
, 1, NULL
);
451 /* gfs2_lookupi has inconsistent callers: vfs
452 * related routines expect NULL for no entry found,
453 * gfs2_lookup_simple callers expect ENOENT
454 * and do not check for NULL.
457 return ERR_PTR(-ENOENT
);
464 * gfs2_lookupi - Look up a filename in a directory and return its inode
465 * @d_gh: An initialized holder for the directory glock
466 * @name: The name of the inode to look for
467 * @is_root: If 1, ignore the caller's permissions
468 * @i_gh: An uninitialized holder for the new inode glock
470 * This can be called via the VFS filldir function when NFS is doing
471 * a readdirplus and the inode which its intending to stat isn't
472 * already in cache. In this case we must not take the directory glock
473 * again, since the readdir call will have already taken that lock.
478 struct inode
*gfs2_lookupi(struct inode
*dir
, const struct qstr
*name
,
479 int is_root
, struct nameidata
*nd
)
481 struct super_block
*sb
= dir
->i_sb
;
482 struct gfs2_inode
*dip
= GFS2_I(dir
);
483 struct gfs2_holder d_gh
;
485 struct inode
*inode
= NULL
;
488 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
489 return ERR_PTR(-ENAMETOOLONG
);
491 if ((name
->len
== 1 && memcmp(name
->name
, ".", 1) == 0) ||
492 (name
->len
== 2 && memcmp(name
->name
, "..", 2) == 0 &&
493 dir
== sb
->s_root
->d_inode
)) {
498 if (gfs2_glock_is_locked_by_me(dip
->i_gl
) == 0) {
499 error
= gfs2_glock_nq_init(dip
->i_gl
, LM_ST_SHARED
, 0, &d_gh
);
501 return ERR_PTR(error
);
506 error
= permission(dir
, MAY_EXEC
, NULL
);
511 inode
= gfs2_dir_search(dir
, name
);
513 error
= PTR_ERR(inode
);
516 gfs2_glock_dq_uninit(&d_gh
);
517 if (error
== -ENOENT
)
519 return inode
? inode
: ERR_PTR(error
);
522 static void gfs2_inum_range_in(struct gfs2_inum_range_host
*ir
, const void *buf
)
524 const struct gfs2_inum_range
*str
= buf
;
526 ir
->ir_start
= be64_to_cpu(str
->ir_start
);
527 ir
->ir_length
= be64_to_cpu(str
->ir_length
);
530 static void gfs2_inum_range_out(const struct gfs2_inum_range_host
*ir
, void *buf
)
532 struct gfs2_inum_range
*str
= buf
;
534 str
->ir_start
= cpu_to_be64(ir
->ir_start
);
535 str
->ir_length
= cpu_to_be64(ir
->ir_length
);
538 static int pick_formal_ino_1(struct gfs2_sbd
*sdp
, u64
*formal_ino
)
540 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
541 struct buffer_head
*bh
;
542 struct gfs2_inum_range_host ir
;
545 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
548 mutex_lock(&sdp
->sd_inum_mutex
);
550 error
= gfs2_meta_inode_buffer(ip
, &bh
);
552 mutex_unlock(&sdp
->sd_inum_mutex
);
557 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
560 *formal_ino
= ir
.ir_start
++;
562 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
563 gfs2_inum_range_out(&ir
,
564 bh
->b_data
+ sizeof(struct gfs2_dinode
));
566 mutex_unlock(&sdp
->sd_inum_mutex
);
573 mutex_unlock(&sdp
->sd_inum_mutex
);
579 static int pick_formal_ino_2(struct gfs2_sbd
*sdp
, u64
*formal_ino
)
581 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
582 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_inum_inode
);
583 struct gfs2_holder gh
;
584 struct buffer_head
*bh
;
585 struct gfs2_inum_range_host ir
;
588 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
592 error
= gfs2_trans_begin(sdp
, 2 * RES_DINODE
, 0);
595 mutex_lock(&sdp
->sd_inum_mutex
);
597 error
= gfs2_meta_inode_buffer(ip
, &bh
);
601 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
604 struct buffer_head
*m_bh
;
608 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
612 z
= *(__be64
*)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
));
613 x
= y
= be64_to_cpu(z
);
615 ir
.ir_length
= GFS2_INUM_QUANTUM
;
616 x
+= GFS2_INUM_QUANTUM
;
618 gfs2_consist_inode(m_ip
);
620 gfs2_trans_add_bh(m_ip
->i_gl
, m_bh
, 1);
621 *(__be64
*)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
)) = z
;
626 *formal_ino
= ir
.ir_start
++;
629 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
630 gfs2_inum_range_out(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
635 mutex_unlock(&sdp
->sd_inum_mutex
);
638 gfs2_glock_dq_uninit(&gh
);
642 static int pick_formal_ino(struct gfs2_sbd
*sdp
, u64
*inum
)
646 error
= pick_formal_ino_1(sdp
, inum
);
650 error
= pick_formal_ino_2(sdp
, inum
);
656 * create_ok - OK to create a new on-disk inode here?
657 * @dip: Directory in which dinode is to be created
658 * @name: Name of new dinode
664 static int create_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
669 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
673 /* Don't create entries in an unlinked directory */
674 if (!dip
->i_inode
.i_nlink
)
677 error
= gfs2_dir_check(&dip
->i_inode
, name
, NULL
);
688 if (dip
->i_di
.di_entries
== (u32
)-1)
690 if (S_ISDIR(mode
) && dip
->i_inode
.i_nlink
== (u32
)-1)
696 static void munge_mode_uid_gid(struct gfs2_inode
*dip
, unsigned int *mode
,
697 unsigned int *uid
, unsigned int *gid
)
699 if (GFS2_SB(&dip
->i_inode
)->sd_args
.ar_suiddir
&&
700 (dip
->i_inode
.i_mode
& S_ISUID
) && dip
->i_inode
.i_uid
) {
703 else if (dip
->i_inode
.i_uid
!= current
->fsuid
)
705 *uid
= dip
->i_inode
.i_uid
;
707 *uid
= current
->fsuid
;
709 if (dip
->i_inode
.i_mode
& S_ISGID
) {
712 *gid
= dip
->i_inode
.i_gid
;
714 *gid
= current
->fsgid
;
717 static int alloc_dinode(struct gfs2_inode
*dip
, u64
*no_addr
, u64
*generation
)
719 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
722 if (gfs2_alloc_get(dip
) == NULL
)
725 dip
->i_alloc
->al_requested
= RES_DINODE
;
726 error
= gfs2_inplace_reserve(dip
);
730 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+ RES_STATFS
, 0);
734 *no_addr
= gfs2_alloc_di(dip
, generation
);
739 gfs2_inplace_release(dip
);
746 * init_dinode - Fill in a new dinode structure
747 * @dip: the directory this inode is being created in
748 * @gl: The glock covering the new inode
749 * @inum: the inode number
750 * @mode: the file permissions
756 static void init_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
757 const struct gfs2_inum_host
*inum
, unsigned int mode
,
758 unsigned int uid
, unsigned int gid
,
759 const u64
*generation
, dev_t dev
, struct buffer_head
**bhp
)
761 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
762 struct gfs2_dinode
*di
;
763 struct buffer_head
*dibh
;
764 struct timespec tv
= CURRENT_TIME
;
766 dibh
= gfs2_meta_new(gl
, inum
->no_addr
);
767 gfs2_trans_add_bh(gl
, dibh
, 1);
768 gfs2_metatype_set(dibh
, GFS2_METATYPE_DI
, GFS2_FORMAT_DI
);
769 gfs2_buffer_clear_tail(dibh
, sizeof(struct gfs2_dinode
));
770 di
= (struct gfs2_dinode
*)dibh
->b_data
;
772 di
->di_num
.no_formal_ino
= cpu_to_be64(inum
->no_formal_ino
);
773 di
->di_num
.no_addr
= cpu_to_be64(inum
->no_addr
);
774 di
->di_mode
= cpu_to_be32(mode
);
775 di
->di_uid
= cpu_to_be32(uid
);
776 di
->di_gid
= cpu_to_be32(gid
);
779 di
->di_blocks
= cpu_to_be64(1);
780 di
->di_atime
= di
->di_mtime
= di
->di_ctime
= cpu_to_be64(tv
.tv_sec
);
781 di
->di_major
= cpu_to_be32(MAJOR(dev
));
782 di
->di_minor
= cpu_to_be32(MINOR(dev
));
783 di
->di_goal_meta
= di
->di_goal_data
= cpu_to_be64(inum
->no_addr
);
784 di
->di_generation
= cpu_to_be64(*generation
);
788 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_JDATA
) ||
789 gfs2_tune_get(sdp
, gt_new_files_jdata
))
790 di
->di_flags
|= cpu_to_be32(GFS2_DIF_JDATA
);
791 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_DIRECTIO
) ||
792 gfs2_tune_get(sdp
, gt_new_files_directio
))
793 di
->di_flags
|= cpu_to_be32(GFS2_DIF_DIRECTIO
);
794 } else if (S_ISDIR(mode
)) {
795 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
796 GFS2_DIF_INHERIT_DIRECTIO
);
797 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
798 GFS2_DIF_INHERIT_JDATA
);
802 di
->di_payload_format
= cpu_to_be32(S_ISDIR(mode
) ? GFS2_FORMAT_DE
: 0);
808 memset(&di
->__pad4
, 0, sizeof(di
->__pad4
));
810 di
->di_atime_nsec
= cpu_to_be32(tv
.tv_nsec
);
811 di
->di_mtime_nsec
= cpu_to_be32(tv
.tv_nsec
);
812 di
->di_ctime_nsec
= cpu_to_be32(tv
.tv_nsec
);
813 memset(&di
->di_reserved
, 0, sizeof(di
->di_reserved
));
815 set_buffer_uptodate(dibh
);
820 static int make_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
821 unsigned int mode
, const struct gfs2_inum_host
*inum
,
822 const u64
*generation
, dev_t dev
, struct buffer_head
**bhp
)
824 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
825 unsigned int uid
, gid
;
828 munge_mode_uid_gid(dip
, &mode
, &uid
, &gid
);
831 error
= gfs2_quota_lock(dip
, uid
, gid
);
835 error
= gfs2_quota_check(dip
, uid
, gid
);
839 error
= gfs2_trans_begin(sdp
, RES_DINODE
+ RES_QUOTA
, 0);
843 init_dinode(dip
, gl
, inum
, mode
, uid
, gid
, generation
, dev
, bhp
);
844 gfs2_quota_change(dip
, +1, uid
, gid
);
848 gfs2_quota_unlock(dip
);
854 static int link_dinode(struct gfs2_inode
*dip
, const struct qstr
*name
,
855 struct gfs2_inode
*ip
)
857 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
858 struct gfs2_alloc
*al
;
860 struct buffer_head
*dibh
;
863 al
= gfs2_alloc_get(dip
);
865 error
= gfs2_quota_lock(dip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
869 error
= alloc_required
= gfs2_diradd_alloc_required(&dip
->i_inode
, name
);
870 if (alloc_required
< 0)
871 goto fail_quota_locks
;
872 if (alloc_required
) {
873 error
= gfs2_quota_check(dip
, dip
->i_inode
.i_uid
, dip
->i_inode
.i_gid
);
875 goto fail_quota_locks
;
877 al
->al_requested
= sdp
->sd_max_dirres
;
879 error
= gfs2_inplace_reserve(dip
);
881 goto fail_quota_locks
;
883 error
= gfs2_trans_begin(sdp
, sdp
->sd_max_dirres
+
884 al
->al_rgd
->rd_length
+
886 RES_STATFS
+ RES_QUOTA
, 0);
890 error
= gfs2_trans_begin(sdp
, RES_LEAF
+ 2 * RES_DINODE
, 0);
892 goto fail_quota_locks
;
895 error
= gfs2_dir_add(&dip
->i_inode
, name
, ip
, IF2DT(ip
->i_inode
.i_mode
));
899 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
902 ip
->i_inode
.i_nlink
= 1;
903 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
904 gfs2_dinode_out(ip
, dibh
->b_data
);
912 if (dip
->i_alloc
->al_rgd
)
913 gfs2_inplace_release(dip
);
916 gfs2_quota_unlock(dip
);
923 static int gfs2_security_init(struct gfs2_inode
*dip
, struct gfs2_inode
*ip
)
929 struct gfs2_ea_request er
;
931 err
= security_inode_init_security(&ip
->i_inode
, &dip
->i_inode
,
932 &name
, &value
, &len
);
935 if (err
== -EOPNOTSUPP
)
940 memset(&er
, 0, sizeof(struct gfs2_ea_request
));
942 er
.er_type
= GFS2_EATYPE_SECURITY
;
945 er
.er_name_len
= strlen(name
);
946 er
.er_data_len
= len
;
948 err
= gfs2_ea_set_i(ip
, &er
);
957 * gfs2_createi - Create a new inode
958 * @ghs: An array of two holders
959 * @name: The name of the new file
960 * @mode: the permissions on the new inode
962 * @ghs[0] is an initialized holder for the directory
963 * @ghs[1] is the holder for the inode lock
965 * If the return value is not NULL, the glocks on both the directory and the new
966 * file are held. A transaction has been started and an inplace reservation
972 struct inode
*gfs2_createi(struct gfs2_holder
*ghs
, const struct qstr
*name
,
973 unsigned int mode
, dev_t dev
)
975 struct inode
*inode
= NULL
;
976 struct gfs2_inode
*dip
= ghs
->gh_gl
->gl_object
;
977 struct inode
*dir
= &dip
->i_inode
;
978 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
979 struct gfs2_inum_host inum
= { .no_addr
= 0, .no_formal_ino
= 0 };
982 struct buffer_head
*bh
= NULL
;
984 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
985 return ERR_PTR(-ENAMETOOLONG
);
987 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, 0, ghs
);
988 error
= gfs2_glock_nq(ghs
);
992 error
= create_ok(dip
, name
, mode
);
996 error
= pick_formal_ino(sdp
, &inum
.no_formal_ino
);
1000 error
= alloc_dinode(dip
, &inum
.no_addr
, &generation
);
1004 error
= gfs2_glock_nq_num(sdp
, inum
.no_addr
, &gfs2_inode_glops
,
1005 LM_ST_EXCLUSIVE
, GL_SKIP
, ghs
+ 1);
1009 error
= make_dinode(dip
, ghs
[1].gh_gl
, mode
, &inum
, &generation
, dev
, &bh
);
1013 inode
= gfs2_inode_lookup(dir
->i_sb
, IF2DT(mode
),
1015 inum
.no_formal_ino
, 0);
1019 error
= gfs2_inode_refresh(GFS2_I(inode
));
1023 error
= gfs2_acl_create(dip
, GFS2_I(inode
));
1027 error
= gfs2_security_init(dip
, GFS2_I(inode
));
1031 error
= link_dinode(dip
, name
, GFS2_I(inode
));
1038 return ERR_PTR(-ENOMEM
);
1042 gfs2_glock_dq_uninit(ghs
+ 1);
1050 return ERR_PTR(error
);
1054 * gfs2_rmdiri - Remove a directory
1055 * @dip: The parent directory of the directory to be removed
1056 * @name: The name of the directory to be removed
1057 * @ip: The GFS2 inode of the directory to be removed
1059 * Assumes Glocks on dip and ip are held
1064 int gfs2_rmdiri(struct gfs2_inode
*dip
, const struct qstr
*name
,
1065 struct gfs2_inode
*ip
)
1067 struct qstr dotname
;
1070 if (ip
->i_di
.di_entries
!= 2) {
1071 if (gfs2_consist_inode(ip
))
1072 gfs2_dinode_print(ip
);
1076 error
= gfs2_dir_del(dip
, name
);
1080 error
= gfs2_change_nlink(dip
, -1);
1084 gfs2_str2qstr(&dotname
, ".");
1085 error
= gfs2_dir_del(ip
, &dotname
);
1089 gfs2_str2qstr(&dotname
, "..");
1090 error
= gfs2_dir_del(ip
, &dotname
);
1094 /* It looks odd, but it really should be done twice */
1095 error
= gfs2_change_nlink(ip
, -1);
1099 error
= gfs2_change_nlink(ip
, -1);
1107 * gfs2_unlink_ok - check to see that a inode is still in a directory
1108 * @dip: the directory
1109 * @name: the name of the file
1112 * Assumes that the lock on (at least) @dip is held.
1114 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1117 int gfs2_unlink_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
1118 const struct gfs2_inode
*ip
)
1122 if (IS_IMMUTABLE(&ip
->i_inode
) || IS_APPEND(&ip
->i_inode
))
1125 if ((dip
->i_inode
.i_mode
& S_ISVTX
) &&
1126 dip
->i_inode
.i_uid
!= current
->fsuid
&&
1127 ip
->i_inode
.i_uid
!= current
->fsuid
&& !capable(CAP_FOWNER
))
1130 if (IS_APPEND(&dip
->i_inode
))
1133 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
1137 error
= gfs2_dir_check(&dip
->i_inode
, name
, ip
);
1145 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1149 * Follow @to back to the root and make sure we don't encounter @this
1150 * Assumes we already hold the rename lock.
1155 int gfs2_ok_to_move(struct gfs2_inode
*this, struct gfs2_inode
*to
)
1157 struct inode
*dir
= &to
->i_inode
;
1158 struct super_block
*sb
= dir
->i_sb
;
1163 gfs2_str2qstr(&dotdot
, "..");
1168 if (dir
== &this->i_inode
) {
1172 if (dir
== sb
->s_root
->d_inode
) {
1177 tmp
= gfs2_lookupi(dir
, &dotdot
, 1, NULL
);
1179 error
= PTR_ERR(tmp
);
1193 * gfs2_readlinki - return the contents of a symlink
1194 * @ip: the symlink's inode
1195 * @buf: a pointer to the buffer to be filled
1196 * @len: a pointer to the length of @buf
1198 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1199 * to be freed by the caller.
1204 int gfs2_readlinki(struct gfs2_inode
*ip
, char **buf
, unsigned int *len
)
1206 struct gfs2_holder i_gh
;
1207 struct buffer_head
*dibh
;
1211 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, GL_ATIME
, &i_gh
);
1212 error
= gfs2_glock_nq_atime(&i_gh
);
1214 gfs2_holder_uninit(&i_gh
);
1218 if (!ip
->i_di
.di_size
) {
1219 gfs2_consist_inode(ip
);
1224 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1228 x
= ip
->i_di
.di_size
+ 1;
1230 *buf
= kmalloc(x
, GFP_KERNEL
);
1237 memcpy(*buf
, dibh
->b_data
+ sizeof(struct gfs2_dinode
), x
);
1243 gfs2_glock_dq_uninit(&i_gh
);
1248 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1249 * conditionally update the inode's atime
1250 * @gh: the holder to acquire
1252 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1253 * Update if the difference between the current time and the inode's current
1254 * atime is greater than an interval specified at mount.
1259 int gfs2_glock_nq_atime(struct gfs2_holder
*gh
)
1261 struct gfs2_glock
*gl
= gh
->gh_gl
;
1262 struct gfs2_sbd
*sdp
= gl
->gl_sbd
;
1263 struct gfs2_inode
*ip
= gl
->gl_object
;
1264 s64 quantum
= gfs2_tune_get(sdp
, gt_atime_quantum
);
1268 struct timespec tv
= CURRENT_TIME
;
1270 if (gfs2_assert_warn(sdp
, gh
->gh_flags
& GL_ATIME
) ||
1271 gfs2_assert_warn(sdp
, !(gh
->gh_flags
& GL_ASYNC
)) ||
1272 gfs2_assert_warn(sdp
, gl
->gl_ops
== &gfs2_inode_glops
))
1275 state
= gh
->gh_state
;
1276 flags
= gh
->gh_flags
;
1278 error
= gfs2_glock_nq(gh
);
1282 if (test_bit(SDF_NOATIME
, &sdp
->sd_flags
) ||
1283 (sdp
->sd_vfs
->s_flags
& MS_RDONLY
))
1286 if (tv
.tv_sec
- ip
->i_inode
.i_atime
.tv_sec
>= quantum
) {
1288 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, gh
->gh_flags
& ~LM_FLAG_ANY
,
1290 error
= gfs2_glock_nq(gh
);
1294 /* Verify that atime hasn't been updated while we were
1295 trying to get exclusive lock. */
1298 if (tv
.tv_sec
- ip
->i_inode
.i_atime
.tv_sec
>= quantum
) {
1299 struct buffer_head
*dibh
;
1300 struct gfs2_dinode
*di
;
1302 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
1303 if (error
== -EROFS
)
1308 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1310 goto fail_end_trans
;
1312 ip
->i_inode
.i_atime
= tv
;
1314 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1315 di
= (struct gfs2_dinode
*)dibh
->b_data
;
1316 di
->di_atime
= cpu_to_be64(ip
->i_inode
.i_atime
.tv_sec
);
1317 di
->di_atime_nsec
= cpu_to_be32(ip
->i_inode
.i_atime
.tv_nsec
);
1320 gfs2_trans_end(sdp
);
1323 /* If someone else has asked for the glock,
1324 unlock and let them have it. Then reacquire
1325 in the original state. */
1326 if (gfs2_glock_is_blocking(gl
)) {
1328 gfs2_holder_reinit(state
, flags
, gh
);
1329 return gfs2_glock_nq(gh
);
1336 gfs2_trans_end(sdp
);
1343 __gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1345 struct buffer_head
*dibh
;
1348 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1350 error
= inode_setattr(&ip
->i_inode
, attr
);
1351 gfs2_assert_warn(GFS2_SB(&ip
->i_inode
), !error
);
1352 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1353 gfs2_dinode_out(ip
, dibh
->b_data
);
1360 * gfs2_setattr_simple -
1364 * Called with a reference on the vnode.
1369 int gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1373 if (current
->journal_info
)
1374 return __gfs2_setattr_simple(ip
, attr
);
1376 error
= gfs2_trans_begin(GFS2_SB(&ip
->i_inode
), RES_DINODE
, 0);
1380 error
= __gfs2_setattr_simple(ip
, attr
);
1381 gfs2_trans_end(GFS2_SB(&ip
->i_inode
));
1385 void gfs2_dinode_out(const struct gfs2_inode
*ip
, void *buf
)
1387 const struct gfs2_dinode_host
*di
= &ip
->i_di
;
1388 struct gfs2_dinode
*str
= buf
;
1390 str
->di_header
.mh_magic
= cpu_to_be32(GFS2_MAGIC
);
1391 str
->di_header
.mh_type
= cpu_to_be32(GFS2_METATYPE_DI
);
1392 str
->di_header
.__pad0
= 0;
1393 str
->di_header
.mh_format
= cpu_to_be32(GFS2_FORMAT_DI
);
1394 str
->di_header
.__pad1
= 0;
1395 str
->di_num
.no_addr
= cpu_to_be64(ip
->i_no_addr
);
1396 str
->di_num
.no_formal_ino
= cpu_to_be64(ip
->i_no_formal_ino
);
1397 str
->di_mode
= cpu_to_be32(ip
->i_inode
.i_mode
);
1398 str
->di_uid
= cpu_to_be32(ip
->i_inode
.i_uid
);
1399 str
->di_gid
= cpu_to_be32(ip
->i_inode
.i_gid
);
1400 str
->di_nlink
= cpu_to_be32(ip
->i_inode
.i_nlink
);
1401 str
->di_size
= cpu_to_be64(di
->di_size
);
1402 str
->di_blocks
= cpu_to_be64(di
->di_blocks
);
1403 str
->di_atime
= cpu_to_be64(ip
->i_inode
.i_atime
.tv_sec
);
1404 str
->di_mtime
= cpu_to_be64(ip
->i_inode
.i_mtime
.tv_sec
);
1405 str
->di_ctime
= cpu_to_be64(ip
->i_inode
.i_ctime
.tv_sec
);
1407 str
->di_goal_meta
= cpu_to_be64(di
->di_goal_meta
);
1408 str
->di_goal_data
= cpu_to_be64(di
->di_goal_data
);
1409 str
->di_generation
= cpu_to_be64(di
->di_generation
);
1411 str
->di_flags
= cpu_to_be32(di
->di_flags
);
1412 str
->di_height
= cpu_to_be16(ip
->i_height
);
1413 str
->di_payload_format
= cpu_to_be32(S_ISDIR(ip
->i_inode
.i_mode
) &&
1414 !(ip
->i_di
.di_flags
& GFS2_DIF_EXHASH
) ?
1415 GFS2_FORMAT_DE
: 0);
1416 str
->di_depth
= cpu_to_be16(ip
->i_depth
);
1417 str
->di_entries
= cpu_to_be32(di
->di_entries
);
1419 str
->di_eattr
= cpu_to_be64(di
->di_eattr
);
1420 str
->di_atime_nsec
= cpu_to_be32(ip
->i_inode
.i_atime
.tv_nsec
);
1421 str
->di_mtime_nsec
= cpu_to_be32(ip
->i_inode
.i_mtime
.tv_nsec
);
1422 str
->di_ctime_nsec
= cpu_to_be32(ip
->i_inode
.i_ctime
.tv_nsec
);
1425 void gfs2_dinode_print(const struct gfs2_inode
*ip
)
1427 const struct gfs2_dinode_host
*di
= &ip
->i_di
;
1429 printk(KERN_INFO
" no_formal_ino = %llu\n",
1430 (unsigned long long)ip
->i_no_formal_ino
);
1431 printk(KERN_INFO
" no_addr = %llu\n",
1432 (unsigned long long)ip
->i_no_addr
);
1433 printk(KERN_INFO
" di_size = %llu\n", (unsigned long long)di
->di_size
);
1434 printk(KERN_INFO
" di_blocks = %llu\n",
1435 (unsigned long long)di
->di_blocks
);
1436 printk(KERN_INFO
" di_goal_meta = %llu\n",
1437 (unsigned long long)di
->di_goal_meta
);
1438 printk(KERN_INFO
" di_goal_data = %llu\n",
1439 (unsigned long long)di
->di_goal_data
);
1440 printk(KERN_INFO
" di_flags = 0x%.8X\n", di
->di_flags
);
1441 printk(KERN_INFO
" i_height = %u\n", ip
->i_height
);
1442 printk(KERN_INFO
" i_depth = %u\n", ip
->i_depth
);
1443 printk(KERN_INFO
" di_entries = %u\n", di
->di_entries
);
1444 printk(KERN_INFO
" di_eattr = %llu\n",
1445 (unsigned long long)di
->di_eattr
);