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 gfs2_set_inode_blocks(&ip
->i_inode
, be64_to_cpu(str
->di_blocks
));
277 ip
->i_inode
.i_atime
.tv_sec
= be64_to_cpu(str
->di_atime
);
278 ip
->i_inode
.i_atime
.tv_nsec
= be32_to_cpu(str
->di_atime_nsec
);
279 ip
->i_inode
.i_mtime
.tv_sec
= be64_to_cpu(str
->di_mtime
);
280 ip
->i_inode
.i_mtime
.tv_nsec
= be32_to_cpu(str
->di_mtime_nsec
);
281 ip
->i_inode
.i_ctime
.tv_sec
= be64_to_cpu(str
->di_ctime
);
282 ip
->i_inode
.i_ctime
.tv_nsec
= be32_to_cpu(str
->di_ctime_nsec
);
284 ip
->i_goal
= be64_to_cpu(str
->di_goal_meta
);
285 di
->di_generation
= be64_to_cpu(str
->di_generation
);
287 di
->di_flags
= be32_to_cpu(str
->di_flags
);
288 gfs2_set_inode_flags(&ip
->i_inode
);
289 height
= be16_to_cpu(str
->di_height
);
290 if (unlikely(height
> GFS2_MAX_META_HEIGHT
))
292 ip
->i_height
= (u8
)height
;
294 depth
= be16_to_cpu(str
->di_depth
);
295 if (unlikely(depth
> GFS2_DIR_MAX_DEPTH
))
297 ip
->i_depth
= (u8
)depth
;
298 di
->di_entries
= be32_to_cpu(str
->di_entries
);
300 di
->di_eattr
= be64_to_cpu(str
->di_eattr
);
301 if (S_ISREG(ip
->i_inode
.i_mode
))
302 gfs2_set_aops(&ip
->i_inode
);
306 if (gfs2_consist_inode(ip
))
307 gfs2_dinode_print(ip
);
312 * gfs2_inode_refresh - Refresh the incore copy of the dinode
313 * @ip: The GFS2 inode
318 int gfs2_inode_refresh(struct gfs2_inode
*ip
)
320 struct buffer_head
*dibh
;
323 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
327 if (gfs2_metatype_check(GFS2_SB(&ip
->i_inode
), dibh
, GFS2_METATYPE_DI
)) {
332 error
= gfs2_dinode_in(ip
, dibh
->b_data
);
334 clear_bit(GIF_INVALID
, &ip
->i_flags
);
339 int gfs2_dinode_dealloc(struct gfs2_inode
*ip
)
341 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
342 struct gfs2_alloc
*al
;
343 struct gfs2_rgrpd
*rgd
;
346 if (gfs2_get_inode_blocks(&ip
->i_inode
) != 1) {
347 if (gfs2_consist_inode(ip
))
348 gfs2_dinode_print(ip
);
352 al
= gfs2_alloc_get(ip
);
354 error
= gfs2_quota_hold(ip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
358 error
= gfs2_rindex_hold(sdp
, &al
->al_ri_gh
);
362 rgd
= gfs2_blk2rgrpd(sdp
, ip
->i_no_addr
);
364 gfs2_consist_inode(ip
);
366 goto out_rindex_relse
;
369 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0,
372 goto out_rindex_relse
;
374 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+ RES_STATFS
+ RES_QUOTA
, 1);
378 set_bit(GLF_DIRTY
, &ip
->i_gl
->gl_flags
);
379 set_bit(GLF_LFLUSH
, &ip
->i_gl
->gl_flags
);
381 gfs2_free_di(rgd
, ip
);
384 clear_bit(GLF_STICKY
, &ip
->i_gl
->gl_flags
);
387 gfs2_glock_dq_uninit(&al
->al_rgd_gh
);
389 gfs2_glock_dq_uninit(&al
->al_ri_gh
);
391 gfs2_quota_unhold(ip
);
398 * gfs2_change_nlink - Change nlink count on inode
399 * @ip: The GFS2 inode
400 * @diff: The change in the nlink count required
404 int gfs2_change_nlink(struct gfs2_inode
*ip
, int diff
)
406 struct buffer_head
*dibh
;
410 BUG_ON(diff
!= 1 && diff
!= -1);
411 nlink
= ip
->i_inode
.i_nlink
+ diff
;
413 /* If we are reducing the nlink count, but the new value ends up being
414 bigger than the old one, we must have underflowed. */
415 if (diff
< 0 && nlink
> ip
->i_inode
.i_nlink
) {
416 if (gfs2_consist_inode(ip
))
417 gfs2_dinode_print(ip
);
421 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
426 inc_nlink(&ip
->i_inode
);
428 drop_nlink(&ip
->i_inode
);
430 ip
->i_inode
.i_ctime
= CURRENT_TIME
;
432 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
433 gfs2_dinode_out(ip
, dibh
->b_data
);
435 mark_inode_dirty(&ip
->i_inode
);
437 if (ip
->i_inode
.i_nlink
== 0)
438 gfs2_unlink_di(&ip
->i_inode
); /* mark inode unlinked */
443 struct inode
*gfs2_lookup_simple(struct inode
*dip
, const char *name
)
447 gfs2_str2qstr(&qstr
, name
);
448 inode
= gfs2_lookupi(dip
, &qstr
, 1, NULL
);
449 /* gfs2_lookupi has inconsistent callers: vfs
450 * related routines expect NULL for no entry found,
451 * gfs2_lookup_simple callers expect ENOENT
452 * and do not check for NULL.
455 return ERR_PTR(-ENOENT
);
462 * gfs2_lookupi - Look up a filename in a directory and return its inode
463 * @d_gh: An initialized holder for the directory glock
464 * @name: The name of the inode to look for
465 * @is_root: If 1, ignore the caller's permissions
466 * @i_gh: An uninitialized holder for the new inode glock
468 * This can be called via the VFS filldir function when NFS is doing
469 * a readdirplus and the inode which its intending to stat isn't
470 * already in cache. In this case we must not take the directory glock
471 * again, since the readdir call will have already taken that lock.
476 struct inode
*gfs2_lookupi(struct inode
*dir
, const struct qstr
*name
,
477 int is_root
, struct nameidata
*nd
)
479 struct super_block
*sb
= dir
->i_sb
;
480 struct gfs2_inode
*dip
= GFS2_I(dir
);
481 struct gfs2_holder d_gh
;
483 struct inode
*inode
= NULL
;
486 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
487 return ERR_PTR(-ENAMETOOLONG
);
489 if ((name
->len
== 1 && memcmp(name
->name
, ".", 1) == 0) ||
490 (name
->len
== 2 && memcmp(name
->name
, "..", 2) == 0 &&
491 dir
== sb
->s_root
->d_inode
)) {
496 if (gfs2_glock_is_locked_by_me(dip
->i_gl
) == 0) {
497 error
= gfs2_glock_nq_init(dip
->i_gl
, LM_ST_SHARED
, 0, &d_gh
);
499 return ERR_PTR(error
);
504 error
= permission(dir
, MAY_EXEC
, NULL
);
509 inode
= gfs2_dir_search(dir
, name
);
511 error
= PTR_ERR(inode
);
514 gfs2_glock_dq_uninit(&d_gh
);
515 if (error
== -ENOENT
)
517 return inode
? inode
: ERR_PTR(error
);
520 static void gfs2_inum_range_in(struct gfs2_inum_range_host
*ir
, const void *buf
)
522 const struct gfs2_inum_range
*str
= buf
;
524 ir
->ir_start
= be64_to_cpu(str
->ir_start
);
525 ir
->ir_length
= be64_to_cpu(str
->ir_length
);
528 static void gfs2_inum_range_out(const struct gfs2_inum_range_host
*ir
, void *buf
)
530 struct gfs2_inum_range
*str
= buf
;
532 str
->ir_start
= cpu_to_be64(ir
->ir_start
);
533 str
->ir_length
= cpu_to_be64(ir
->ir_length
);
536 static int pick_formal_ino_1(struct gfs2_sbd
*sdp
, u64
*formal_ino
)
538 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
539 struct buffer_head
*bh
;
540 struct gfs2_inum_range_host ir
;
543 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
546 mutex_lock(&sdp
->sd_inum_mutex
);
548 error
= gfs2_meta_inode_buffer(ip
, &bh
);
550 mutex_unlock(&sdp
->sd_inum_mutex
);
555 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
558 *formal_ino
= ir
.ir_start
++;
560 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
561 gfs2_inum_range_out(&ir
,
562 bh
->b_data
+ sizeof(struct gfs2_dinode
));
564 mutex_unlock(&sdp
->sd_inum_mutex
);
571 mutex_unlock(&sdp
->sd_inum_mutex
);
577 static int pick_formal_ino_2(struct gfs2_sbd
*sdp
, u64
*formal_ino
)
579 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
580 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_inum_inode
);
581 struct gfs2_holder gh
;
582 struct buffer_head
*bh
;
583 struct gfs2_inum_range_host ir
;
586 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
590 error
= gfs2_trans_begin(sdp
, 2 * RES_DINODE
, 0);
593 mutex_lock(&sdp
->sd_inum_mutex
);
595 error
= gfs2_meta_inode_buffer(ip
, &bh
);
599 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
602 struct buffer_head
*m_bh
;
606 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
610 z
= *(__be64
*)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
));
611 x
= y
= be64_to_cpu(z
);
613 ir
.ir_length
= GFS2_INUM_QUANTUM
;
614 x
+= GFS2_INUM_QUANTUM
;
616 gfs2_consist_inode(m_ip
);
618 gfs2_trans_add_bh(m_ip
->i_gl
, m_bh
, 1);
619 *(__be64
*)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
)) = z
;
624 *formal_ino
= ir
.ir_start
++;
627 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
628 gfs2_inum_range_out(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
633 mutex_unlock(&sdp
->sd_inum_mutex
);
636 gfs2_glock_dq_uninit(&gh
);
640 static int pick_formal_ino(struct gfs2_sbd
*sdp
, u64
*inum
)
644 error
= pick_formal_ino_1(sdp
, inum
);
648 error
= pick_formal_ino_2(sdp
, inum
);
654 * create_ok - OK to create a new on-disk inode here?
655 * @dip: Directory in which dinode is to be created
656 * @name: Name of new dinode
662 static int create_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
667 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
671 /* Don't create entries in an unlinked directory */
672 if (!dip
->i_inode
.i_nlink
)
675 error
= gfs2_dir_check(&dip
->i_inode
, name
, NULL
);
686 if (dip
->i_di
.di_entries
== (u32
)-1)
688 if (S_ISDIR(mode
) && dip
->i_inode
.i_nlink
== (u32
)-1)
694 static void munge_mode_uid_gid(struct gfs2_inode
*dip
, unsigned int *mode
,
695 unsigned int *uid
, unsigned int *gid
)
697 if (GFS2_SB(&dip
->i_inode
)->sd_args
.ar_suiddir
&&
698 (dip
->i_inode
.i_mode
& S_ISUID
) && dip
->i_inode
.i_uid
) {
701 else if (dip
->i_inode
.i_uid
!= current
->fsuid
)
703 *uid
= dip
->i_inode
.i_uid
;
705 *uid
= current
->fsuid
;
707 if (dip
->i_inode
.i_mode
& S_ISGID
) {
710 *gid
= dip
->i_inode
.i_gid
;
712 *gid
= current
->fsgid
;
715 static int alloc_dinode(struct gfs2_inode
*dip
, u64
*no_addr
, u64
*generation
)
717 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
720 if (gfs2_alloc_get(dip
) == NULL
)
723 dip
->i_alloc
->al_requested
= RES_DINODE
;
724 error
= gfs2_inplace_reserve(dip
);
728 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+ RES_STATFS
, 0);
732 *no_addr
= gfs2_alloc_di(dip
, generation
);
737 gfs2_inplace_release(dip
);
744 * init_dinode - Fill in a new dinode structure
745 * @dip: the directory this inode is being created in
746 * @gl: The glock covering the new inode
747 * @inum: the inode number
748 * @mode: the file permissions
754 static void init_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
755 const struct gfs2_inum_host
*inum
, unsigned int mode
,
756 unsigned int uid
, unsigned int gid
,
757 const u64
*generation
, dev_t dev
, struct buffer_head
**bhp
)
759 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
760 struct gfs2_dinode
*di
;
761 struct buffer_head
*dibh
;
762 struct timespec tv
= CURRENT_TIME
;
764 dibh
= gfs2_meta_new(gl
, inum
->no_addr
);
765 gfs2_trans_add_bh(gl
, dibh
, 1);
766 gfs2_metatype_set(dibh
, GFS2_METATYPE_DI
, GFS2_FORMAT_DI
);
767 gfs2_buffer_clear_tail(dibh
, sizeof(struct gfs2_dinode
));
768 di
= (struct gfs2_dinode
*)dibh
->b_data
;
770 di
->di_num
.no_formal_ino
= cpu_to_be64(inum
->no_formal_ino
);
771 di
->di_num
.no_addr
= cpu_to_be64(inum
->no_addr
);
772 di
->di_mode
= cpu_to_be32(mode
);
773 di
->di_uid
= cpu_to_be32(uid
);
774 di
->di_gid
= cpu_to_be32(gid
);
777 di
->di_blocks
= cpu_to_be64(1);
778 di
->di_atime
= di
->di_mtime
= di
->di_ctime
= cpu_to_be64(tv
.tv_sec
);
779 di
->di_major
= cpu_to_be32(MAJOR(dev
));
780 di
->di_minor
= cpu_to_be32(MINOR(dev
));
781 di
->di_goal_meta
= di
->di_goal_data
= cpu_to_be64(inum
->no_addr
);
782 di
->di_generation
= cpu_to_be64(*generation
);
786 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_JDATA
) ||
787 gfs2_tune_get(sdp
, gt_new_files_jdata
))
788 di
->di_flags
|= cpu_to_be32(GFS2_DIF_JDATA
);
789 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_DIRECTIO
) ||
790 gfs2_tune_get(sdp
, gt_new_files_directio
))
791 di
->di_flags
|= cpu_to_be32(GFS2_DIF_DIRECTIO
);
792 } else if (S_ISDIR(mode
)) {
793 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
794 GFS2_DIF_INHERIT_DIRECTIO
);
795 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
796 GFS2_DIF_INHERIT_JDATA
);
800 di
->di_payload_format
= cpu_to_be32(S_ISDIR(mode
) ? GFS2_FORMAT_DE
: 0);
806 memset(&di
->__pad4
, 0, sizeof(di
->__pad4
));
808 di
->di_atime_nsec
= cpu_to_be32(tv
.tv_nsec
);
809 di
->di_mtime_nsec
= cpu_to_be32(tv
.tv_nsec
);
810 di
->di_ctime_nsec
= cpu_to_be32(tv
.tv_nsec
);
811 memset(&di
->di_reserved
, 0, sizeof(di
->di_reserved
));
813 set_buffer_uptodate(dibh
);
818 static int make_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
819 unsigned int mode
, const struct gfs2_inum_host
*inum
,
820 const u64
*generation
, dev_t dev
, struct buffer_head
**bhp
)
822 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
823 unsigned int uid
, gid
;
826 munge_mode_uid_gid(dip
, &mode
, &uid
, &gid
);
829 error
= gfs2_quota_lock(dip
, uid
, gid
);
833 error
= gfs2_quota_check(dip
, uid
, gid
);
837 error
= gfs2_trans_begin(sdp
, RES_DINODE
+ RES_QUOTA
, 0);
841 init_dinode(dip
, gl
, inum
, mode
, uid
, gid
, generation
, dev
, bhp
);
842 gfs2_quota_change(dip
, +1, uid
, gid
);
846 gfs2_quota_unlock(dip
);
852 static int link_dinode(struct gfs2_inode
*dip
, const struct qstr
*name
,
853 struct gfs2_inode
*ip
)
855 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
856 struct gfs2_alloc
*al
;
858 struct buffer_head
*dibh
;
861 al
= gfs2_alloc_get(dip
);
863 error
= gfs2_quota_lock(dip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
867 error
= alloc_required
= gfs2_diradd_alloc_required(&dip
->i_inode
, name
);
868 if (alloc_required
< 0)
869 goto fail_quota_locks
;
870 if (alloc_required
) {
871 error
= gfs2_quota_check(dip
, dip
->i_inode
.i_uid
, dip
->i_inode
.i_gid
);
873 goto fail_quota_locks
;
875 al
->al_requested
= sdp
->sd_max_dirres
;
877 error
= gfs2_inplace_reserve(dip
);
879 goto fail_quota_locks
;
881 error
= gfs2_trans_begin(sdp
, sdp
->sd_max_dirres
+
882 al
->al_rgd
->rd_length
+
884 RES_STATFS
+ RES_QUOTA
, 0);
888 error
= gfs2_trans_begin(sdp
, RES_LEAF
+ 2 * RES_DINODE
, 0);
890 goto fail_quota_locks
;
893 error
= gfs2_dir_add(&dip
->i_inode
, name
, ip
, IF2DT(ip
->i_inode
.i_mode
));
897 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
900 ip
->i_inode
.i_nlink
= 1;
901 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
902 gfs2_dinode_out(ip
, dibh
->b_data
);
910 if (dip
->i_alloc
->al_rgd
)
911 gfs2_inplace_release(dip
);
914 gfs2_quota_unlock(dip
);
921 static int gfs2_security_init(struct gfs2_inode
*dip
, struct gfs2_inode
*ip
)
927 struct gfs2_ea_request er
;
929 err
= security_inode_init_security(&ip
->i_inode
, &dip
->i_inode
,
930 &name
, &value
, &len
);
933 if (err
== -EOPNOTSUPP
)
938 memset(&er
, 0, sizeof(struct gfs2_ea_request
));
940 er
.er_type
= GFS2_EATYPE_SECURITY
;
943 er
.er_name_len
= strlen(name
);
944 er
.er_data_len
= len
;
946 err
= gfs2_ea_set_i(ip
, &er
);
955 * gfs2_createi - Create a new inode
956 * @ghs: An array of two holders
957 * @name: The name of the new file
958 * @mode: the permissions on the new inode
960 * @ghs[0] is an initialized holder for the directory
961 * @ghs[1] is the holder for the inode lock
963 * If the return value is not NULL, the glocks on both the directory and the new
964 * file are held. A transaction has been started and an inplace reservation
970 struct inode
*gfs2_createi(struct gfs2_holder
*ghs
, const struct qstr
*name
,
971 unsigned int mode
, dev_t dev
)
973 struct inode
*inode
= NULL
;
974 struct gfs2_inode
*dip
= ghs
->gh_gl
->gl_object
;
975 struct inode
*dir
= &dip
->i_inode
;
976 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
977 struct gfs2_inum_host inum
= { .no_addr
= 0, .no_formal_ino
= 0 };
980 struct buffer_head
*bh
= NULL
;
982 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
983 return ERR_PTR(-ENAMETOOLONG
);
985 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, 0, ghs
);
986 error
= gfs2_glock_nq(ghs
);
990 error
= create_ok(dip
, name
, mode
);
994 error
= pick_formal_ino(sdp
, &inum
.no_formal_ino
);
998 error
= alloc_dinode(dip
, &inum
.no_addr
, &generation
);
1002 error
= gfs2_glock_nq_num(sdp
, inum
.no_addr
, &gfs2_inode_glops
,
1003 LM_ST_EXCLUSIVE
, GL_SKIP
, ghs
+ 1);
1007 error
= make_dinode(dip
, ghs
[1].gh_gl
, mode
, &inum
, &generation
, dev
, &bh
);
1011 inode
= gfs2_inode_lookup(dir
->i_sb
, IF2DT(mode
),
1013 inum
.no_formal_ino
, 0);
1017 error
= gfs2_inode_refresh(GFS2_I(inode
));
1021 error
= gfs2_acl_create(dip
, GFS2_I(inode
));
1025 error
= gfs2_security_init(dip
, GFS2_I(inode
));
1029 error
= link_dinode(dip
, name
, GFS2_I(inode
));
1036 return ERR_PTR(-ENOMEM
);
1040 gfs2_glock_dq_uninit(ghs
+ 1);
1048 return ERR_PTR(error
);
1052 * gfs2_rmdiri - Remove a directory
1053 * @dip: The parent directory of the directory to be removed
1054 * @name: The name of the directory to be removed
1055 * @ip: The GFS2 inode of the directory to be removed
1057 * Assumes Glocks on dip and ip are held
1062 int gfs2_rmdiri(struct gfs2_inode
*dip
, const struct qstr
*name
,
1063 struct gfs2_inode
*ip
)
1065 struct qstr dotname
;
1068 if (ip
->i_di
.di_entries
!= 2) {
1069 if (gfs2_consist_inode(ip
))
1070 gfs2_dinode_print(ip
);
1074 error
= gfs2_dir_del(dip
, name
);
1078 error
= gfs2_change_nlink(dip
, -1);
1082 gfs2_str2qstr(&dotname
, ".");
1083 error
= gfs2_dir_del(ip
, &dotname
);
1087 gfs2_str2qstr(&dotname
, "..");
1088 error
= gfs2_dir_del(ip
, &dotname
);
1092 /* It looks odd, but it really should be done twice */
1093 error
= gfs2_change_nlink(ip
, -1);
1097 error
= gfs2_change_nlink(ip
, -1);
1105 * gfs2_unlink_ok - check to see that a inode is still in a directory
1106 * @dip: the directory
1107 * @name: the name of the file
1110 * Assumes that the lock on (at least) @dip is held.
1112 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1115 int gfs2_unlink_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
1116 const struct gfs2_inode
*ip
)
1120 if (IS_IMMUTABLE(&ip
->i_inode
) || IS_APPEND(&ip
->i_inode
))
1123 if ((dip
->i_inode
.i_mode
& S_ISVTX
) &&
1124 dip
->i_inode
.i_uid
!= current
->fsuid
&&
1125 ip
->i_inode
.i_uid
!= current
->fsuid
&& !capable(CAP_FOWNER
))
1128 if (IS_APPEND(&dip
->i_inode
))
1131 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
1135 error
= gfs2_dir_check(&dip
->i_inode
, name
, ip
);
1143 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1147 * Follow @to back to the root and make sure we don't encounter @this
1148 * Assumes we already hold the rename lock.
1153 int gfs2_ok_to_move(struct gfs2_inode
*this, struct gfs2_inode
*to
)
1155 struct inode
*dir
= &to
->i_inode
;
1156 struct super_block
*sb
= dir
->i_sb
;
1161 gfs2_str2qstr(&dotdot
, "..");
1166 if (dir
== &this->i_inode
) {
1170 if (dir
== sb
->s_root
->d_inode
) {
1175 tmp
= gfs2_lookupi(dir
, &dotdot
, 1, NULL
);
1177 error
= PTR_ERR(tmp
);
1191 * gfs2_readlinki - return the contents of a symlink
1192 * @ip: the symlink's inode
1193 * @buf: a pointer to the buffer to be filled
1194 * @len: a pointer to the length of @buf
1196 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1197 * to be freed by the caller.
1202 int gfs2_readlinki(struct gfs2_inode
*ip
, char **buf
, unsigned int *len
)
1204 struct gfs2_holder i_gh
;
1205 struct buffer_head
*dibh
;
1209 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, GL_ATIME
, &i_gh
);
1210 error
= gfs2_glock_nq_atime(&i_gh
);
1212 gfs2_holder_uninit(&i_gh
);
1216 if (!ip
->i_di
.di_size
) {
1217 gfs2_consist_inode(ip
);
1222 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1226 x
= ip
->i_di
.di_size
+ 1;
1228 *buf
= kmalloc(x
, GFP_KERNEL
);
1235 memcpy(*buf
, dibh
->b_data
+ sizeof(struct gfs2_dinode
), x
);
1241 gfs2_glock_dq_uninit(&i_gh
);
1246 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1247 * conditionally update the inode's atime
1248 * @gh: the holder to acquire
1250 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1251 * Update if the difference between the current time and the inode's current
1252 * atime is greater than an interval specified at mount.
1257 int gfs2_glock_nq_atime(struct gfs2_holder
*gh
)
1259 struct gfs2_glock
*gl
= gh
->gh_gl
;
1260 struct gfs2_sbd
*sdp
= gl
->gl_sbd
;
1261 struct gfs2_inode
*ip
= gl
->gl_object
;
1262 s64 quantum
= gfs2_tune_get(sdp
, gt_atime_quantum
);
1266 struct timespec tv
= CURRENT_TIME
;
1268 if (gfs2_assert_warn(sdp
, gh
->gh_flags
& GL_ATIME
) ||
1269 gfs2_assert_warn(sdp
, !(gh
->gh_flags
& GL_ASYNC
)) ||
1270 gfs2_assert_warn(sdp
, gl
->gl_ops
== &gfs2_inode_glops
))
1273 state
= gh
->gh_state
;
1274 flags
= gh
->gh_flags
;
1276 error
= gfs2_glock_nq(gh
);
1280 if (test_bit(SDF_NOATIME
, &sdp
->sd_flags
) ||
1281 (sdp
->sd_vfs
->s_flags
& MS_RDONLY
))
1284 if (tv
.tv_sec
- ip
->i_inode
.i_atime
.tv_sec
>= quantum
) {
1286 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, gh
->gh_flags
& ~LM_FLAG_ANY
,
1288 error
= gfs2_glock_nq(gh
);
1292 /* Verify that atime hasn't been updated while we were
1293 trying to get exclusive lock. */
1296 if (tv
.tv_sec
- ip
->i_inode
.i_atime
.tv_sec
>= quantum
) {
1297 struct buffer_head
*dibh
;
1298 struct gfs2_dinode
*di
;
1300 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
1301 if (error
== -EROFS
)
1306 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1308 goto fail_end_trans
;
1310 ip
->i_inode
.i_atime
= tv
;
1312 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1313 di
= (struct gfs2_dinode
*)dibh
->b_data
;
1314 di
->di_atime
= cpu_to_be64(ip
->i_inode
.i_atime
.tv_sec
);
1315 di
->di_atime_nsec
= cpu_to_be32(ip
->i_inode
.i_atime
.tv_nsec
);
1318 gfs2_trans_end(sdp
);
1321 /* If someone else has asked for the glock,
1322 unlock and let them have it. Then reacquire
1323 in the original state. */
1324 if (gfs2_glock_is_blocking(gl
)) {
1326 gfs2_holder_reinit(state
, flags
, gh
);
1327 return gfs2_glock_nq(gh
);
1334 gfs2_trans_end(sdp
);
1341 __gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1343 struct buffer_head
*dibh
;
1346 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1348 error
= inode_setattr(&ip
->i_inode
, attr
);
1349 gfs2_assert_warn(GFS2_SB(&ip
->i_inode
), !error
);
1350 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1351 gfs2_dinode_out(ip
, dibh
->b_data
);
1358 * gfs2_setattr_simple -
1362 * Called with a reference on the vnode.
1367 int gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1371 if (current
->journal_info
)
1372 return __gfs2_setattr_simple(ip
, attr
);
1374 error
= gfs2_trans_begin(GFS2_SB(&ip
->i_inode
), RES_DINODE
, 0);
1378 error
= __gfs2_setattr_simple(ip
, attr
);
1379 gfs2_trans_end(GFS2_SB(&ip
->i_inode
));
1383 void gfs2_dinode_out(const struct gfs2_inode
*ip
, void *buf
)
1385 const struct gfs2_dinode_host
*di
= &ip
->i_di
;
1386 struct gfs2_dinode
*str
= buf
;
1388 str
->di_header
.mh_magic
= cpu_to_be32(GFS2_MAGIC
);
1389 str
->di_header
.mh_type
= cpu_to_be32(GFS2_METATYPE_DI
);
1390 str
->di_header
.__pad0
= 0;
1391 str
->di_header
.mh_format
= cpu_to_be32(GFS2_FORMAT_DI
);
1392 str
->di_header
.__pad1
= 0;
1393 str
->di_num
.no_addr
= cpu_to_be64(ip
->i_no_addr
);
1394 str
->di_num
.no_formal_ino
= cpu_to_be64(ip
->i_no_formal_ino
);
1395 str
->di_mode
= cpu_to_be32(ip
->i_inode
.i_mode
);
1396 str
->di_uid
= cpu_to_be32(ip
->i_inode
.i_uid
);
1397 str
->di_gid
= cpu_to_be32(ip
->i_inode
.i_gid
);
1398 str
->di_nlink
= cpu_to_be32(ip
->i_inode
.i_nlink
);
1399 str
->di_size
= cpu_to_be64(di
->di_size
);
1400 str
->di_blocks
= cpu_to_be64(gfs2_get_inode_blocks(&ip
->i_inode
));
1401 str
->di_atime
= cpu_to_be64(ip
->i_inode
.i_atime
.tv_sec
);
1402 str
->di_mtime
= cpu_to_be64(ip
->i_inode
.i_mtime
.tv_sec
);
1403 str
->di_ctime
= cpu_to_be64(ip
->i_inode
.i_ctime
.tv_sec
);
1405 str
->di_goal_meta
= cpu_to_be64(ip
->i_goal
);
1406 str
->di_goal_data
= cpu_to_be64(ip
->i_goal
);
1407 str
->di_generation
= cpu_to_be64(di
->di_generation
);
1409 str
->di_flags
= cpu_to_be32(di
->di_flags
);
1410 str
->di_height
= cpu_to_be16(ip
->i_height
);
1411 str
->di_payload_format
= cpu_to_be32(S_ISDIR(ip
->i_inode
.i_mode
) &&
1412 !(ip
->i_di
.di_flags
& GFS2_DIF_EXHASH
) ?
1413 GFS2_FORMAT_DE
: 0);
1414 str
->di_depth
= cpu_to_be16(ip
->i_depth
);
1415 str
->di_entries
= cpu_to_be32(di
->di_entries
);
1417 str
->di_eattr
= cpu_to_be64(di
->di_eattr
);
1418 str
->di_atime_nsec
= cpu_to_be32(ip
->i_inode
.i_atime
.tv_nsec
);
1419 str
->di_mtime_nsec
= cpu_to_be32(ip
->i_inode
.i_mtime
.tv_nsec
);
1420 str
->di_ctime_nsec
= cpu_to_be32(ip
->i_inode
.i_ctime
.tv_nsec
);
1423 void gfs2_dinode_print(const struct gfs2_inode
*ip
)
1425 const struct gfs2_dinode_host
*di
= &ip
->i_di
;
1427 printk(KERN_INFO
" no_formal_ino = %llu\n",
1428 (unsigned long long)ip
->i_no_formal_ino
);
1429 printk(KERN_INFO
" no_addr = %llu\n",
1430 (unsigned long long)ip
->i_no_addr
);
1431 printk(KERN_INFO
" di_size = %llu\n", (unsigned long long)di
->di_size
);
1432 printk(KERN_INFO
" blocks = %llu\n",
1433 (unsigned long long)gfs2_get_inode_blocks(&ip
->i_inode
));
1434 printk(KERN_INFO
" i_goal = %llu\n",
1435 (unsigned long long)ip
->i_goal
);
1436 printk(KERN_INFO
" di_flags = 0x%.8X\n", di
->di_flags
);
1437 printk(KERN_INFO
" i_height = %u\n", ip
->i_height
);
1438 printk(KERN_INFO
" i_depth = %u\n", ip
->i_depth
);
1439 printk(KERN_INFO
" di_entries = %u\n", di
->di_entries
);
1440 printk(KERN_INFO
" di_eattr = %llu\n",
1441 (unsigned long long)di
->di_eattr
);