2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2006 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"
35 #include "ops_inode.h"
41 struct gfs2_inum_range_host
{
46 static int iget_test(struct inode
*inode
, void *opaque
)
48 struct gfs2_inode
*ip
= GFS2_I(inode
);
49 u64
*no_addr
= opaque
;
51 if (ip
->i_no_addr
== *no_addr
&&
52 inode
->i_private
!= NULL
)
58 static int iget_set(struct inode
*inode
, void *opaque
)
60 struct gfs2_inode
*ip
= GFS2_I(inode
);
61 u64
*no_addr
= opaque
;
63 inode
->i_ino
= (unsigned long)*no_addr
;
64 ip
->i_no_addr
= *no_addr
;
68 struct inode
*gfs2_ilookup(struct super_block
*sb
, u64 no_addr
)
70 unsigned long hash
= (unsigned long)no_addr
;
71 return ilookup5(sb
, hash
, iget_test
, &no_addr
);
74 static struct inode
*gfs2_iget(struct super_block
*sb
, u64 no_addr
)
76 unsigned long hash
= (unsigned long)no_addr
;
77 return iget5_locked(sb
, hash
, iget_test
, iget_set
, &no_addr
);
81 * gfs2_inode_lookup - Lookup an inode
82 * @sb: The super block
83 * @no_addr: The inode number
84 * @type: The type of the inode
86 * Returns: A VFS inode, or an error
89 struct inode
*gfs2_inode_lookup(struct super_block
*sb
, u64 no_addr
, unsigned int type
)
91 struct inode
*inode
= gfs2_iget(sb
, no_addr
);
92 struct gfs2_inode
*ip
= GFS2_I(inode
);
93 struct gfs2_glock
*io_gl
;
97 return ERR_PTR(-ENOBUFS
);
99 if (inode
->i_state
& I_NEW
) {
100 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
101 umode_t mode
= DT2IF(type
);
102 inode
->i_private
= ip
;
103 inode
->i_mode
= mode
;
106 inode
->i_op
= &gfs2_file_iops
;
107 inode
->i_fop
= &gfs2_file_fops
;
108 inode
->i_mapping
->a_ops
= &gfs2_file_aops
;
109 } else if (S_ISDIR(mode
)) {
110 inode
->i_op
= &gfs2_dir_iops
;
111 inode
->i_fop
= &gfs2_dir_fops
;
112 } else if (S_ISLNK(mode
)) {
113 inode
->i_op
= &gfs2_symlink_iops
;
115 inode
->i_op
= &gfs2_dev_iops
;
118 error
= gfs2_glock_get(sdp
, no_addr
, &gfs2_inode_glops
, CREATE
, &ip
->i_gl
);
121 ip
->i_gl
->gl_object
= ip
;
123 error
= gfs2_glock_get(sdp
, no_addr
, &gfs2_iopen_glops
, CREATE
, &io_gl
);
127 set_bit(GIF_INVALID
, &ip
->i_flags
);
128 error
= gfs2_glock_nq_init(io_gl
, LM_ST_SHARED
, GL_EXACT
, &ip
->i_iopen_gh
);
132 gfs2_glock_put(io_gl
);
133 unlock_new_inode(inode
);
138 gfs2_glock_put(io_gl
);
140 ip
->i_gl
->gl_object
= NULL
;
141 gfs2_glock_put(ip
->i_gl
);
144 return ERR_PTR(error
);
147 static int gfs2_dinode_in(struct gfs2_inode
*ip
, const void *buf
)
149 struct gfs2_dinode_host
*di
= &ip
->i_di
;
150 const struct gfs2_dinode
*str
= buf
;
152 if (ip
->i_no_addr
!= be64_to_cpu(str
->di_num
.no_addr
)) {
153 if (gfs2_consist_inode(ip
))
154 gfs2_dinode_print(ip
);
157 ip
->i_no_formal_ino
= be64_to_cpu(str
->di_num
.no_formal_ino
);
158 ip
->i_inode
.i_mode
= be32_to_cpu(str
->di_mode
);
159 ip
->i_inode
.i_rdev
= 0;
160 switch (ip
->i_inode
.i_mode
& S_IFMT
) {
163 ip
->i_inode
.i_rdev
= MKDEV(be32_to_cpu(str
->di_major
),
164 be32_to_cpu(str
->di_minor
));
168 ip
->i_inode
.i_uid
= be32_to_cpu(str
->di_uid
);
169 ip
->i_inode
.i_gid
= be32_to_cpu(str
->di_gid
);
171 * We will need to review setting the nlink count here in the
172 * light of the forthcoming ro bind mount work. This is a reminder
175 ip
->i_inode
.i_nlink
= be32_to_cpu(str
->di_nlink
);
176 di
->di_size
= be64_to_cpu(str
->di_size
);
177 i_size_write(&ip
->i_inode
, di
->di_size
);
178 di
->di_blocks
= be64_to_cpu(str
->di_blocks
);
179 gfs2_set_inode_blocks(&ip
->i_inode
);
180 ip
->i_inode
.i_atime
.tv_sec
= be64_to_cpu(str
->di_atime
);
181 ip
->i_inode
.i_atime
.tv_nsec
= be32_to_cpu(str
->di_atime_nsec
);
182 ip
->i_inode
.i_mtime
.tv_sec
= be64_to_cpu(str
->di_mtime
);
183 ip
->i_inode
.i_mtime
.tv_nsec
= be32_to_cpu(str
->di_mtime_nsec
);
184 ip
->i_inode
.i_ctime
.tv_sec
= be64_to_cpu(str
->di_ctime
);
185 ip
->i_inode
.i_ctime
.tv_nsec
= be32_to_cpu(str
->di_ctime_nsec
);
187 di
->di_goal_meta
= be64_to_cpu(str
->di_goal_meta
);
188 di
->di_goal_data
= be64_to_cpu(str
->di_goal_data
);
189 di
->di_generation
= be64_to_cpu(str
->di_generation
);
191 di
->di_flags
= be32_to_cpu(str
->di_flags
);
192 gfs2_set_inode_flags(&ip
->i_inode
);
193 di
->di_height
= be16_to_cpu(str
->di_height
);
195 di
->di_depth
= be16_to_cpu(str
->di_depth
);
196 di
->di_entries
= be32_to_cpu(str
->di_entries
);
198 di
->di_eattr
= be64_to_cpu(str
->di_eattr
);
203 * gfs2_inode_refresh - Refresh the incore copy of the dinode
204 * @ip: The GFS2 inode
209 int gfs2_inode_refresh(struct gfs2_inode
*ip
)
211 struct buffer_head
*dibh
;
214 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
218 if (gfs2_metatype_check(GFS2_SB(&ip
->i_inode
), dibh
, GFS2_METATYPE_DI
)) {
223 error
= gfs2_dinode_in(ip
, dibh
->b_data
);
225 clear_bit(GIF_INVALID
, &ip
->i_flags
);
230 int gfs2_dinode_dealloc(struct gfs2_inode
*ip
)
232 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
233 struct gfs2_alloc
*al
;
234 struct gfs2_rgrpd
*rgd
;
237 if (ip
->i_di
.di_blocks
!= 1) {
238 if (gfs2_consist_inode(ip
))
239 gfs2_dinode_print(ip
);
243 al
= gfs2_alloc_get(ip
);
245 error
= gfs2_quota_hold(ip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
249 error
= gfs2_rindex_hold(sdp
, &al
->al_ri_gh
);
253 rgd
= gfs2_blk2rgrpd(sdp
, ip
->i_no_addr
);
255 gfs2_consist_inode(ip
);
257 goto out_rindex_relse
;
260 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0,
263 goto out_rindex_relse
;
265 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+ RES_STATFS
+ RES_QUOTA
, 1);
269 gfs2_trans_add_gl(ip
->i_gl
);
271 gfs2_free_di(rgd
, ip
);
274 clear_bit(GLF_STICKY
, &ip
->i_gl
->gl_flags
);
277 gfs2_glock_dq_uninit(&al
->al_rgd_gh
);
279 gfs2_glock_dq_uninit(&al
->al_ri_gh
);
281 gfs2_quota_unhold(ip
);
288 * gfs2_change_nlink - Change nlink count on inode
289 * @ip: The GFS2 inode
290 * @diff: The change in the nlink count required
294 int gfs2_change_nlink(struct gfs2_inode
*ip
, int diff
)
296 struct buffer_head
*dibh
;
300 BUG_ON(diff
!= 1 && diff
!= -1);
301 nlink
= ip
->i_inode
.i_nlink
+ diff
;
303 /* If we are reducing the nlink count, but the new value ends up being
304 bigger than the old one, we must have underflowed. */
305 if (diff
< 0 && nlink
> ip
->i_inode
.i_nlink
) {
306 if (gfs2_consist_inode(ip
))
307 gfs2_dinode_print(ip
);
311 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
316 inc_nlink(&ip
->i_inode
);
318 drop_nlink(&ip
->i_inode
);
320 ip
->i_inode
.i_ctime
= CURRENT_TIME
;
322 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
323 gfs2_dinode_out(ip
, dibh
->b_data
);
325 mark_inode_dirty(&ip
->i_inode
);
327 if (ip
->i_inode
.i_nlink
== 0)
328 gfs2_unlink_di(&ip
->i_inode
); /* mark inode unlinked */
333 struct inode
*gfs2_lookup_simple(struct inode
*dip
, const char *name
)
337 gfs2_str2qstr(&qstr
, name
);
338 inode
= gfs2_lookupi(dip
, &qstr
, 1, NULL
);
339 /* gfs2_lookupi has inconsistent callers: vfs
340 * related routines expect NULL for no entry found,
341 * gfs2_lookup_simple callers expect ENOENT
342 * and do not check for NULL.
345 return ERR_PTR(-ENOENT
);
352 * gfs2_lookupi - Look up a filename in a directory and return its inode
353 * @d_gh: An initialized holder for the directory glock
354 * @name: The name of the inode to look for
355 * @is_root: If 1, ignore the caller's permissions
356 * @i_gh: An uninitialized holder for the new inode glock
358 * This can be called via the VFS filldir function when NFS is doing
359 * a readdirplus and the inode which its intending to stat isn't
360 * already in cache. In this case we must not take the directory glock
361 * again, since the readdir call will have already taken that lock.
366 struct inode
*gfs2_lookupi(struct inode
*dir
, const struct qstr
*name
,
367 int is_root
, struct nameidata
*nd
)
369 struct super_block
*sb
= dir
->i_sb
;
370 struct gfs2_inode
*dip
= GFS2_I(dir
);
371 struct gfs2_holder d_gh
;
373 struct inode
*inode
= NULL
;
376 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
377 return ERR_PTR(-ENAMETOOLONG
);
379 if ((name
->len
== 1 && memcmp(name
->name
, ".", 1) == 0) ||
380 (name
->len
== 2 && memcmp(name
->name
, "..", 2) == 0 &&
381 dir
== sb
->s_root
->d_inode
)) {
386 if (gfs2_glock_is_locked_by_me(dip
->i_gl
) == 0) {
387 error
= gfs2_glock_nq_init(dip
->i_gl
, LM_ST_SHARED
, 0, &d_gh
);
389 return ERR_PTR(error
);
394 error
= permission(dir
, MAY_EXEC
, NULL
);
399 inode
= gfs2_dir_search(dir
, name
);
401 error
= PTR_ERR(inode
);
404 gfs2_glock_dq_uninit(&d_gh
);
405 if (error
== -ENOENT
)
407 return inode
? inode
: ERR_PTR(error
);
410 static void gfs2_inum_range_in(struct gfs2_inum_range_host
*ir
, const void *buf
)
412 const struct gfs2_inum_range
*str
= buf
;
414 ir
->ir_start
= be64_to_cpu(str
->ir_start
);
415 ir
->ir_length
= be64_to_cpu(str
->ir_length
);
418 static void gfs2_inum_range_out(const struct gfs2_inum_range_host
*ir
, void *buf
)
420 struct gfs2_inum_range
*str
= buf
;
422 str
->ir_start
= cpu_to_be64(ir
->ir_start
);
423 str
->ir_length
= cpu_to_be64(ir
->ir_length
);
426 static int pick_formal_ino_1(struct gfs2_sbd
*sdp
, u64
*formal_ino
)
428 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
429 struct buffer_head
*bh
;
430 struct gfs2_inum_range_host ir
;
433 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
436 mutex_lock(&sdp
->sd_inum_mutex
);
438 error
= gfs2_meta_inode_buffer(ip
, &bh
);
440 mutex_unlock(&sdp
->sd_inum_mutex
);
445 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
448 *formal_ino
= ir
.ir_start
++;
450 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
451 gfs2_inum_range_out(&ir
,
452 bh
->b_data
+ sizeof(struct gfs2_dinode
));
454 mutex_unlock(&sdp
->sd_inum_mutex
);
461 mutex_unlock(&sdp
->sd_inum_mutex
);
467 static int pick_formal_ino_2(struct gfs2_sbd
*sdp
, u64
*formal_ino
)
469 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
470 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_inum_inode
);
471 struct gfs2_holder gh
;
472 struct buffer_head
*bh
;
473 struct gfs2_inum_range_host ir
;
476 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
480 error
= gfs2_trans_begin(sdp
, 2 * RES_DINODE
, 0);
483 mutex_lock(&sdp
->sd_inum_mutex
);
485 error
= gfs2_meta_inode_buffer(ip
, &bh
);
489 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
492 struct buffer_head
*m_bh
;
496 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
500 z
= *(__be64
*)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
));
501 x
= y
= be64_to_cpu(z
);
503 ir
.ir_length
= GFS2_INUM_QUANTUM
;
504 x
+= GFS2_INUM_QUANTUM
;
506 gfs2_consist_inode(m_ip
);
508 gfs2_trans_add_bh(m_ip
->i_gl
, m_bh
, 1);
509 *(__be64
*)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
)) = z
;
514 *formal_ino
= ir
.ir_start
++;
517 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
518 gfs2_inum_range_out(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
523 mutex_unlock(&sdp
->sd_inum_mutex
);
526 gfs2_glock_dq_uninit(&gh
);
530 static int pick_formal_ino(struct gfs2_sbd
*sdp
, u64
*inum
)
534 error
= pick_formal_ino_1(sdp
, inum
);
538 error
= pick_formal_ino_2(sdp
, inum
);
544 * create_ok - OK to create a new on-disk inode here?
545 * @dip: Directory in which dinode is to be created
546 * @name: Name of new dinode
552 static int create_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
557 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
561 /* Don't create entries in an unlinked directory */
562 if (!dip
->i_inode
.i_nlink
)
565 error
= gfs2_dir_check(&dip
->i_inode
, name
, NULL
);
576 if (dip
->i_di
.di_entries
== (u32
)-1)
578 if (S_ISDIR(mode
) && dip
->i_inode
.i_nlink
== (u32
)-1)
584 static void munge_mode_uid_gid(struct gfs2_inode
*dip
, unsigned int *mode
,
585 unsigned int *uid
, unsigned int *gid
)
587 if (GFS2_SB(&dip
->i_inode
)->sd_args
.ar_suiddir
&&
588 (dip
->i_inode
.i_mode
& S_ISUID
) && dip
->i_inode
.i_uid
) {
591 else if (dip
->i_inode
.i_uid
!= current
->fsuid
)
593 *uid
= dip
->i_inode
.i_uid
;
595 *uid
= current
->fsuid
;
597 if (dip
->i_inode
.i_mode
& S_ISGID
) {
600 *gid
= dip
->i_inode
.i_gid
;
602 *gid
= current
->fsgid
;
605 static int alloc_dinode(struct gfs2_inode
*dip
, u64
*no_addr
, u64
*generation
)
607 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
612 dip
->i_alloc
.al_requested
= RES_DINODE
;
613 error
= gfs2_inplace_reserve(dip
);
617 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+ RES_STATFS
, 0);
621 *no_addr
= gfs2_alloc_di(dip
, generation
);
626 gfs2_inplace_release(dip
);
633 * init_dinode - Fill in a new dinode structure
634 * @dip: the directory this inode is being created in
635 * @gl: The glock covering the new inode
636 * @inum: the inode number
637 * @mode: the file permissions
643 static void init_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
644 const struct gfs2_inum_host
*inum
, unsigned int mode
,
645 unsigned int uid
, unsigned int gid
,
646 const u64
*generation
, dev_t dev
)
648 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
649 struct gfs2_dinode
*di
;
650 struct buffer_head
*dibh
;
651 struct timespec tv
= CURRENT_TIME
;
653 dibh
= gfs2_meta_new(gl
, inum
->no_addr
);
654 gfs2_trans_add_bh(gl
, dibh
, 1);
655 gfs2_metatype_set(dibh
, GFS2_METATYPE_DI
, GFS2_FORMAT_DI
);
656 gfs2_buffer_clear_tail(dibh
, sizeof(struct gfs2_dinode
));
657 di
= (struct gfs2_dinode
*)dibh
->b_data
;
659 di
->di_num
.no_formal_ino
= cpu_to_be64(inum
->no_formal_ino
);
660 di
->di_num
.no_addr
= cpu_to_be64(inum
->no_addr
);
661 di
->di_mode
= cpu_to_be32(mode
);
662 di
->di_uid
= cpu_to_be32(uid
);
663 di
->di_gid
= cpu_to_be32(gid
);
666 di
->di_blocks
= cpu_to_be64(1);
667 di
->di_atime
= di
->di_mtime
= di
->di_ctime
= cpu_to_be64(tv
.tv_sec
);
668 di
->di_major
= cpu_to_be32(MAJOR(dev
));
669 di
->di_minor
= cpu_to_be32(MINOR(dev
));
670 di
->di_goal_meta
= di
->di_goal_data
= cpu_to_be64(inum
->no_addr
);
671 di
->di_generation
= cpu_to_be64(*generation
);
675 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_JDATA
) ||
676 gfs2_tune_get(sdp
, gt_new_files_jdata
))
677 di
->di_flags
|= cpu_to_be32(GFS2_DIF_JDATA
);
678 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_DIRECTIO
) ||
679 gfs2_tune_get(sdp
, gt_new_files_directio
))
680 di
->di_flags
|= cpu_to_be32(GFS2_DIF_DIRECTIO
);
681 } else if (S_ISDIR(mode
)) {
682 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
683 GFS2_DIF_INHERIT_DIRECTIO
);
684 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
685 GFS2_DIF_INHERIT_JDATA
);
689 di
->di_payload_format
= cpu_to_be32(S_ISDIR(mode
) ? GFS2_FORMAT_DE
: 0);
695 memset(&di
->__pad4
, 0, sizeof(di
->__pad4
));
697 di
->di_atime_nsec
= cpu_to_be32(tv
.tv_nsec
);
698 di
->di_mtime_nsec
= cpu_to_be32(tv
.tv_nsec
);
699 di
->di_ctime_nsec
= cpu_to_be32(tv
.tv_nsec
);
700 memset(&di
->di_reserved
, 0, sizeof(di
->di_reserved
));
705 static int make_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
706 unsigned int mode
, const struct gfs2_inum_host
*inum
,
707 const u64
*generation
, dev_t dev
)
709 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
710 unsigned int uid
, gid
;
713 munge_mode_uid_gid(dip
, &mode
, &uid
, &gid
);
716 error
= gfs2_quota_lock(dip
, uid
, gid
);
720 error
= gfs2_quota_check(dip
, uid
, gid
);
724 error
= gfs2_trans_begin(sdp
, RES_DINODE
+ RES_QUOTA
, 0);
728 init_dinode(dip
, gl
, inum
, mode
, uid
, gid
, generation
, dev
);
729 gfs2_quota_change(dip
, +1, uid
, gid
);
733 gfs2_quota_unlock(dip
);
739 static int link_dinode(struct gfs2_inode
*dip
, const struct qstr
*name
,
740 struct gfs2_inode
*ip
)
742 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
743 struct gfs2_alloc
*al
;
745 struct buffer_head
*dibh
;
748 al
= gfs2_alloc_get(dip
);
750 error
= gfs2_quota_lock(dip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
754 error
= alloc_required
= gfs2_diradd_alloc_required(&dip
->i_inode
, name
);
755 if (alloc_required
< 0)
757 if (alloc_required
) {
758 error
= gfs2_quota_check(dip
, dip
->i_inode
.i_uid
, dip
->i_inode
.i_gid
);
760 goto fail_quota_locks
;
762 al
->al_requested
= sdp
->sd_max_dirres
;
764 error
= gfs2_inplace_reserve(dip
);
766 goto fail_quota_locks
;
768 error
= gfs2_trans_begin(sdp
, sdp
->sd_max_dirres
+
769 al
->al_rgd
->rd_length
+
771 RES_STATFS
+ RES_QUOTA
, 0);
775 error
= gfs2_trans_begin(sdp
, RES_LEAF
+ 2 * RES_DINODE
, 0);
777 goto fail_quota_locks
;
780 error
= gfs2_dir_add(&dip
->i_inode
, name
, ip
, IF2DT(ip
->i_inode
.i_mode
));
784 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
787 ip
->i_inode
.i_nlink
= 1;
788 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
789 gfs2_dinode_out(ip
, dibh
->b_data
);
797 if (dip
->i_alloc
.al_rgd
)
798 gfs2_inplace_release(dip
);
801 gfs2_quota_unlock(dip
);
808 static int gfs2_security_init(struct gfs2_inode
*dip
, struct gfs2_inode
*ip
)
814 struct gfs2_ea_request er
;
816 err
= security_inode_init_security(&ip
->i_inode
, &dip
->i_inode
,
817 &name
, &value
, &len
);
820 if (err
== -EOPNOTSUPP
)
825 memset(&er
, 0, sizeof(struct gfs2_ea_request
));
827 er
.er_type
= GFS2_EATYPE_SECURITY
;
830 er
.er_name_len
= strlen(name
);
831 er
.er_data_len
= len
;
833 err
= gfs2_ea_set_i(ip
, &er
);
842 * gfs2_createi - Create a new inode
843 * @ghs: An array of two holders
844 * @name: The name of the new file
845 * @mode: the permissions on the new inode
847 * @ghs[0] is an initialized holder for the directory
848 * @ghs[1] is the holder for the inode lock
850 * If the return value is not NULL, the glocks on both the directory and the new
851 * file are held. A transaction has been started and an inplace reservation
857 struct inode
*gfs2_createi(struct gfs2_holder
*ghs
, const struct qstr
*name
,
858 unsigned int mode
, dev_t dev
)
861 struct gfs2_inode
*dip
= ghs
->gh_gl
->gl_object
;
862 struct inode
*dir
= &dip
->i_inode
;
863 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
864 struct gfs2_inum_host inum
= { .no_addr
= 0, .no_formal_ino
= 0 };
868 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
869 return ERR_PTR(-ENAMETOOLONG
);
871 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, 0, ghs
);
872 error
= gfs2_glock_nq(ghs
);
876 error
= create_ok(dip
, name
, mode
);
880 error
= pick_formal_ino(sdp
, &inum
.no_formal_ino
);
884 error
= alloc_dinode(dip
, &inum
.no_addr
, &generation
);
888 error
= gfs2_glock_nq_num(sdp
, inum
.no_addr
, &gfs2_inode_glops
,
889 LM_ST_EXCLUSIVE
, GL_SKIP
, ghs
+ 1);
893 error
= make_dinode(dip
, ghs
[1].gh_gl
, mode
, &inum
, &generation
, dev
);
897 inode
= gfs2_inode_lookup(dir
->i_sb
, inum
.no_addr
, IF2DT(mode
));
901 error
= gfs2_inode_refresh(GFS2_I(inode
));
905 error
= gfs2_acl_create(dip
, GFS2_I(inode
));
909 error
= gfs2_security_init(dip
, GFS2_I(inode
));
913 error
= link_dinode(dip
, name
, GFS2_I(inode
));
918 return ERR_PTR(-ENOMEM
);
924 gfs2_glock_dq_uninit(ghs
+ 1);
928 return ERR_PTR(error
);
932 * gfs2_rmdiri - Remove a directory
933 * @dip: The parent directory of the directory to be removed
934 * @name: The name of the directory to be removed
935 * @ip: The GFS2 inode of the directory to be removed
937 * Assumes Glocks on dip and ip are held
942 int gfs2_rmdiri(struct gfs2_inode
*dip
, const struct qstr
*name
,
943 struct gfs2_inode
*ip
)
948 if (ip
->i_di
.di_entries
!= 2) {
949 if (gfs2_consist_inode(ip
))
950 gfs2_dinode_print(ip
);
954 error
= gfs2_dir_del(dip
, name
);
958 error
= gfs2_change_nlink(dip
, -1);
962 gfs2_str2qstr(&dotname
, ".");
963 error
= gfs2_dir_del(ip
, &dotname
);
967 gfs2_str2qstr(&dotname
, "..");
968 error
= gfs2_dir_del(ip
, &dotname
);
972 /* It looks odd, but it really should be done twice */
973 error
= gfs2_change_nlink(ip
, -1);
977 error
= gfs2_change_nlink(ip
, -1);
985 * gfs2_unlink_ok - check to see that a inode is still in a directory
986 * @dip: the directory
987 * @name: the name of the file
990 * Assumes that the lock on (at least) @dip is held.
992 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
995 int gfs2_unlink_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
996 const struct gfs2_inode
*ip
)
1000 if (IS_IMMUTABLE(&ip
->i_inode
) || IS_APPEND(&ip
->i_inode
))
1003 if ((dip
->i_inode
.i_mode
& S_ISVTX
) &&
1004 dip
->i_inode
.i_uid
!= current
->fsuid
&&
1005 ip
->i_inode
.i_uid
!= current
->fsuid
&& !capable(CAP_FOWNER
))
1008 if (IS_APPEND(&dip
->i_inode
))
1011 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
1015 error
= gfs2_dir_check(&dip
->i_inode
, name
, ip
);
1023 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1027 * Follow @to back to the root and make sure we don't encounter @this
1028 * Assumes we already hold the rename lock.
1033 int gfs2_ok_to_move(struct gfs2_inode
*this, struct gfs2_inode
*to
)
1035 struct inode
*dir
= &to
->i_inode
;
1036 struct super_block
*sb
= dir
->i_sb
;
1041 gfs2_str2qstr(&dotdot
, "..");
1046 if (dir
== &this->i_inode
) {
1050 if (dir
== sb
->s_root
->d_inode
) {
1055 tmp
= gfs2_lookupi(dir
, &dotdot
, 1, NULL
);
1057 error
= PTR_ERR(tmp
);
1071 * gfs2_readlinki - return the contents of a symlink
1072 * @ip: the symlink's inode
1073 * @buf: a pointer to the buffer to be filled
1074 * @len: a pointer to the length of @buf
1076 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1077 * to be freed by the caller.
1082 int gfs2_readlinki(struct gfs2_inode
*ip
, char **buf
, unsigned int *len
)
1084 struct gfs2_holder i_gh
;
1085 struct buffer_head
*dibh
;
1089 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, GL_ATIME
, &i_gh
);
1090 error
= gfs2_glock_nq_atime(&i_gh
);
1092 gfs2_holder_uninit(&i_gh
);
1096 if (!ip
->i_di
.di_size
) {
1097 gfs2_consist_inode(ip
);
1102 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1106 x
= ip
->i_di
.di_size
+ 1;
1108 *buf
= kmalloc(x
, GFP_KERNEL
);
1115 memcpy(*buf
, dibh
->b_data
+ sizeof(struct gfs2_dinode
), x
);
1121 gfs2_glock_dq_uninit(&i_gh
);
1126 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1127 * conditionally update the inode's atime
1128 * @gh: the holder to acquire
1130 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1131 * Update if the difference between the current time and the inode's current
1132 * atime is greater than an interval specified at mount.
1137 int gfs2_glock_nq_atime(struct gfs2_holder
*gh
)
1139 struct gfs2_glock
*gl
= gh
->gh_gl
;
1140 struct gfs2_sbd
*sdp
= gl
->gl_sbd
;
1141 struct gfs2_inode
*ip
= gl
->gl_object
;
1142 s64 quantum
= gfs2_tune_get(sdp
, gt_atime_quantum
);
1146 struct timespec tv
= CURRENT_TIME
;
1148 if (gfs2_assert_warn(sdp
, gh
->gh_flags
& GL_ATIME
) ||
1149 gfs2_assert_warn(sdp
, !(gh
->gh_flags
& GL_ASYNC
)) ||
1150 gfs2_assert_warn(sdp
, gl
->gl_ops
== &gfs2_inode_glops
))
1153 state
= gh
->gh_state
;
1154 flags
= gh
->gh_flags
;
1156 error
= gfs2_glock_nq(gh
);
1160 if (test_bit(SDF_NOATIME
, &sdp
->sd_flags
) ||
1161 (sdp
->sd_vfs
->s_flags
& MS_RDONLY
))
1164 if (tv
.tv_sec
- ip
->i_inode
.i_atime
.tv_sec
>= quantum
) {
1166 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, gh
->gh_flags
& ~LM_FLAG_ANY
,
1168 error
= gfs2_glock_nq(gh
);
1172 /* Verify that atime hasn't been updated while we were
1173 trying to get exclusive lock. */
1176 if (tv
.tv_sec
- ip
->i_inode
.i_atime
.tv_sec
>= quantum
) {
1177 struct buffer_head
*dibh
;
1178 struct gfs2_dinode
*di
;
1180 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
1181 if (error
== -EROFS
)
1186 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1188 goto fail_end_trans
;
1190 ip
->i_inode
.i_atime
= tv
;
1192 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1193 di
= (struct gfs2_dinode
*)dibh
->b_data
;
1194 di
->di_atime
= cpu_to_be64(ip
->i_inode
.i_atime
.tv_sec
);
1195 di
->di_atime_nsec
= cpu_to_be32(ip
->i_inode
.i_atime
.tv_nsec
);
1198 gfs2_trans_end(sdp
);
1201 /* If someone else has asked for the glock,
1202 unlock and let them have it. Then reacquire
1203 in the original state. */
1204 if (gfs2_glock_is_blocking(gl
)) {
1206 gfs2_holder_reinit(state
, flags
, gh
);
1207 return gfs2_glock_nq(gh
);
1214 gfs2_trans_end(sdp
);
1221 __gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1223 struct buffer_head
*dibh
;
1226 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1228 error
= inode_setattr(&ip
->i_inode
, attr
);
1229 gfs2_assert_warn(GFS2_SB(&ip
->i_inode
), !error
);
1230 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1231 gfs2_dinode_out(ip
, dibh
->b_data
);
1238 * gfs2_setattr_simple -
1242 * Called with a reference on the vnode.
1247 int gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1251 if (current
->journal_info
)
1252 return __gfs2_setattr_simple(ip
, attr
);
1254 error
= gfs2_trans_begin(GFS2_SB(&ip
->i_inode
), RES_DINODE
, 0);
1258 error
= __gfs2_setattr_simple(ip
, attr
);
1259 gfs2_trans_end(GFS2_SB(&ip
->i_inode
));
1263 void gfs2_dinode_out(const struct gfs2_inode
*ip
, void *buf
)
1265 const struct gfs2_dinode_host
*di
= &ip
->i_di
;
1266 struct gfs2_dinode
*str
= buf
;
1268 str
->di_header
.mh_magic
= cpu_to_be32(GFS2_MAGIC
);
1269 str
->di_header
.mh_type
= cpu_to_be32(GFS2_METATYPE_DI
);
1270 str
->di_header
.__pad0
= 0;
1271 str
->di_header
.mh_format
= cpu_to_be32(GFS2_FORMAT_DI
);
1272 str
->di_header
.__pad1
= 0;
1273 str
->di_num
.no_addr
= cpu_to_be64(ip
->i_no_addr
);
1274 str
->di_num
.no_formal_ino
= cpu_to_be64(ip
->i_no_formal_ino
);
1275 str
->di_mode
= cpu_to_be32(ip
->i_inode
.i_mode
);
1276 str
->di_uid
= cpu_to_be32(ip
->i_inode
.i_uid
);
1277 str
->di_gid
= cpu_to_be32(ip
->i_inode
.i_gid
);
1278 str
->di_nlink
= cpu_to_be32(ip
->i_inode
.i_nlink
);
1279 str
->di_size
= cpu_to_be64(di
->di_size
);
1280 str
->di_blocks
= cpu_to_be64(di
->di_blocks
);
1281 str
->di_atime
= cpu_to_be64(ip
->i_inode
.i_atime
.tv_sec
);
1282 str
->di_mtime
= cpu_to_be64(ip
->i_inode
.i_mtime
.tv_sec
);
1283 str
->di_ctime
= cpu_to_be64(ip
->i_inode
.i_ctime
.tv_sec
);
1285 str
->di_goal_meta
= cpu_to_be64(di
->di_goal_meta
);
1286 str
->di_goal_data
= cpu_to_be64(di
->di_goal_data
);
1287 str
->di_generation
= cpu_to_be64(di
->di_generation
);
1289 str
->di_flags
= cpu_to_be32(di
->di_flags
);
1290 str
->di_height
= cpu_to_be16(di
->di_height
);
1291 str
->di_payload_format
= cpu_to_be32(S_ISDIR(ip
->i_inode
.i_mode
) &&
1292 !(ip
->i_di
.di_flags
& GFS2_DIF_EXHASH
) ?
1293 GFS2_FORMAT_DE
: 0);
1294 str
->di_depth
= cpu_to_be16(di
->di_depth
);
1295 str
->di_entries
= cpu_to_be32(di
->di_entries
);
1297 str
->di_eattr
= cpu_to_be64(di
->di_eattr
);
1298 str
->di_atime_nsec
= cpu_to_be32(ip
->i_inode
.i_atime
.tv_nsec
);
1299 str
->di_mtime_nsec
= cpu_to_be32(ip
->i_inode
.i_mtime
.tv_nsec
);
1300 str
->di_ctime_nsec
= cpu_to_be32(ip
->i_inode
.i_ctime
.tv_nsec
);
1303 void gfs2_dinode_print(const struct gfs2_inode
*ip
)
1305 const struct gfs2_dinode_host
*di
= &ip
->i_di
;
1307 printk(KERN_INFO
" no_formal_ino = %llu\n",
1308 (unsigned long long)ip
->i_no_formal_ino
);
1309 printk(KERN_INFO
" no_addr = %llu\n",
1310 (unsigned long long)ip
->i_no_addr
);
1311 printk(KERN_INFO
" di_size = %llu\n", (unsigned long long)di
->di_size
);
1312 printk(KERN_INFO
" di_blocks = %llu\n",
1313 (unsigned long long)di
->di_blocks
);
1314 printk(KERN_INFO
" di_goal_meta = %llu\n",
1315 (unsigned long long)di
->di_goal_meta
);
1316 printk(KERN_INFO
" di_goal_data = %llu\n",
1317 (unsigned long long)di
->di_goal_data
);
1318 printk(KERN_INFO
" di_flags = 0x%.8X\n", di
->di_flags
);
1319 printk(KERN_INFO
" di_height = %u\n", di
->di_height
);
1320 printk(KERN_INFO
" di_depth = %u\n", di
->di_depth
);
1321 printk(KERN_INFO
" di_entries = %u\n", di
->di_entries
);
1322 printk(KERN_INFO
" di_eattr = %llu\n",
1323 (unsigned long long)di
->di_eattr
);