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 v.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>
21 #include "lm_interface.h"
32 #include "ops_address.h"
34 #include "ops_inode.h"
41 * inode_attr_in - Copy attributes from the dinode into the VFS inode
42 * @ip: The GFS2 inode (with embedded disk inode data)
43 * @inode: The Linux VFS inode
47 static void inode_attr_in(struct gfs2_inode
*ip
, struct inode
*inode
)
49 inode
->i_ino
= ip
->i_num
.no_formal_ino
;
51 switch (ip
->i_di
.di_mode
& S_IFMT
) {
54 inode
->i_rdev
= MKDEV(ip
->i_di
.di_major
, ip
->i_di
.di_minor
);
61 inode
->i_mode
= ip
->i_di
.di_mode
;
62 inode
->i_nlink
= ip
->i_di
.di_nlink
;
63 inode
->i_uid
= ip
->i_di
.di_uid
;
64 inode
->i_gid
= ip
->i_di
.di_gid
;
65 i_size_write(inode
, ip
->i_di
.di_size
);
66 inode
->i_atime
.tv_sec
= ip
->i_di
.di_atime
;
67 inode
->i_mtime
.tv_sec
= ip
->i_di
.di_mtime
;
68 inode
->i_ctime
.tv_sec
= ip
->i_di
.di_ctime
;
69 inode
->i_atime
.tv_nsec
= 0;
70 inode
->i_mtime
.tv_nsec
= 0;
71 inode
->i_ctime
.tv_nsec
= 0;
72 inode
->i_blksize
= PAGE_SIZE
;
73 inode
->i_blocks
= ip
->i_di
.di_blocks
<<
74 (GFS2_SB(inode
)->sd_sb
.sb_bsize_shift
- GFS2_BASIC_BLOCK_SHIFT
);
76 if (ip
->i_di
.di_flags
& GFS2_DIF_IMMUTABLE
)
77 inode
->i_flags
|= S_IMMUTABLE
;
79 inode
->i_flags
&= ~S_IMMUTABLE
;
81 if (ip
->i_di
.di_flags
& GFS2_DIF_APPENDONLY
)
82 inode
->i_flags
|= S_APPEND
;
84 inode
->i_flags
&= ~S_APPEND
;
88 * gfs2_inode_attr_in - Copy attributes from the dinode into the VFS inode
89 * @ip: The GFS2 inode (with embedded disk inode data)
93 void gfs2_inode_attr_in(struct gfs2_inode
*ip
)
95 struct inode
*inode
= &ip
->i_inode
;
96 inode_attr_in(ip
, inode
);
100 * gfs2_inode_attr_out - Copy attributes from VFS inode into the dinode
101 * @ip: The GFS2 inode
103 * Only copy out the attributes that we want the VFS layer
104 * to be able to modify.
107 void gfs2_inode_attr_out(struct gfs2_inode
*ip
)
109 struct inode
*inode
= &ip
->i_inode
;
111 gfs2_assert_withdraw(GFS2_SB(inode
),
112 (ip
->i_di
.di_mode
& S_IFMT
) == (inode
->i_mode
& S_IFMT
));
113 ip
->i_di
.di_mode
= inode
->i_mode
;
114 ip
->i_di
.di_uid
= inode
->i_uid
;
115 ip
->i_di
.di_gid
= inode
->i_gid
;
116 ip
->i_di
.di_atime
= inode
->i_atime
.tv_sec
;
117 ip
->i_di
.di_mtime
= inode
->i_mtime
.tv_sec
;
118 ip
->i_di
.di_ctime
= inode
->i_ctime
.tv_sec
;
121 static int iget_test(struct inode
*inode
, void *opaque
)
123 struct gfs2_inode
*ip
= GFS2_I(inode
);
124 struct gfs2_inum
*inum
= opaque
;
126 if (ip
&& ip
->i_num
.no_addr
== inum
->no_addr
)
132 static int iget_set(struct inode
*inode
, void *opaque
)
134 struct gfs2_inode
*ip
= GFS2_I(inode
);
135 struct gfs2_inum
*inum
= opaque
;
141 struct inode
*gfs2_ilookup(struct super_block
*sb
, struct gfs2_inum
*inum
)
143 return ilookup5(sb
, (unsigned long)inum
->no_formal_ino
,
147 static struct inode
*gfs2_iget(struct super_block
*sb
, struct gfs2_inum
*inum
)
149 return iget5_locked(sb
, (unsigned long)inum
->no_formal_ino
,
150 iget_test
, iget_set
, inum
);
154 * gfs2_inode_lookup - Lookup an inode
155 * @sb: The super block
156 * @inum: The inode number
157 * @type: The type of the inode
159 * Returns: A VFS inode, or an error
162 struct inode
*gfs2_inode_lookup(struct super_block
*sb
, struct gfs2_inum
*inum
, unsigned int type
)
164 struct inode
*inode
= gfs2_iget(sb
, inum
);
165 struct gfs2_inode
*ip
= GFS2_I(inode
);
166 struct gfs2_glock
*io_gl
;
169 if (inode
->i_state
& I_NEW
) {
170 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
171 umode_t mode
= DT2IF(type
);
172 inode
->u
.generic_ip
= ip
;
173 inode
->i_mode
= mode
;
176 inode
->i_op
= &gfs2_file_iops
;
177 inode
->i_fop
= &gfs2_file_fops
;
178 inode
->i_mapping
->a_ops
= &gfs2_file_aops
;
179 } else if (S_ISDIR(mode
)) {
180 inode
->i_op
= &gfs2_dir_iops
;
181 inode
->i_fop
= &gfs2_dir_fops
;
182 } else if (S_ISLNK(mode
)) {
183 inode
->i_op
= &gfs2_symlink_iops
;
185 inode
->i_op
= &gfs2_dev_iops
;
188 error
= gfs2_glock_get(sdp
, inum
->no_addr
, &gfs2_inode_glops
, CREATE
, &ip
->i_gl
);
191 ip
->i_gl
->gl_object
= ip
;
193 error
= gfs2_glock_get(sdp
, inum
->no_addr
, &gfs2_iopen_glops
, CREATE
, &io_gl
);
197 ip
->i_vn
= ip
->i_gl
->gl_vn
- 1;
198 error
= gfs2_glock_nq_init(io_gl
, LM_ST_SHARED
, GL_EXACT
, &ip
->i_iopen_gh
);
202 gfs2_glock_put(io_gl
);
203 unlock_new_inode(inode
);
208 gfs2_glock_put(io_gl
);
210 ip
->i_gl
->gl_object
= NULL
;
211 gfs2_glock_put(ip
->i_gl
);
214 return ERR_PTR(error
);
218 * gfs2_inode_refresh - Refresh the incore copy of the dinode
219 * @ip: The GFS2 inode
224 int gfs2_inode_refresh(struct gfs2_inode
*ip
)
226 struct buffer_head
*dibh
;
229 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
233 if (gfs2_metatype_check(GFS2_SB(&ip
->i_inode
), dibh
, GFS2_METATYPE_DI
)) {
238 gfs2_dinode_in(&ip
->i_di
, dibh
->b_data
);
239 set_bit(GIF_MIN_INIT
, &ip
->i_flags
);
243 if (ip
->i_num
.no_addr
!= ip
->i_di
.di_num
.no_addr
) {
244 if (gfs2_consist_inode(ip
))
245 gfs2_dinode_print(&ip
->i_di
);
248 if (ip
->i_num
.no_formal_ino
!= ip
->i_di
.di_num
.no_formal_ino
)
251 ip
->i_vn
= ip
->i_gl
->gl_vn
;
256 int gfs2_dinode_dealloc(struct gfs2_inode
*ip
)
258 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
259 struct gfs2_alloc
*al
;
260 struct gfs2_rgrpd
*rgd
;
263 if (ip
->i_di
.di_blocks
!= 1) {
264 if (gfs2_consist_inode(ip
))
265 gfs2_dinode_print(&ip
->i_di
);
269 al
= gfs2_alloc_get(ip
);
271 error
= gfs2_quota_hold(ip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
275 error
= gfs2_rindex_hold(sdp
, &al
->al_ri_gh
);
279 rgd
= gfs2_blk2rgrpd(sdp
, ip
->i_num
.no_addr
);
281 gfs2_consist_inode(ip
);
283 goto out_rindex_relse
;
286 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0,
289 goto out_rindex_relse
;
291 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+
292 RES_STATFS
+ RES_QUOTA
, 1);
296 gfs2_trans_add_gl(ip
->i_gl
);
298 gfs2_free_di(rgd
, ip
);
301 clear_bit(GLF_STICKY
, &ip
->i_gl
->gl_flags
);
304 gfs2_glock_dq_uninit(&al
->al_rgd_gh
);
306 gfs2_glock_dq_uninit(&al
->al_ri_gh
);
308 gfs2_quota_unhold(ip
);
315 * gfs2_change_nlink - Change nlink count on inode
316 * @ip: The GFS2 inode
317 * @diff: The change in the nlink count required
322 int gfs2_change_nlink(struct gfs2_inode
*ip
, int diff
)
324 struct gfs2_sbd
*sdp
= ip
->i_inode
.i_sb
->s_fs_info
;
325 struct buffer_head
*dibh
;
329 nlink
= ip
->i_di
.di_nlink
+ diff
;
331 /* If we are reducing the nlink count, but the new value ends up being
332 bigger than the old one, we must have underflowed. */
333 if (diff
< 0 && nlink
> ip
->i_di
.di_nlink
) {
334 if (gfs2_consist_inode(ip
))
335 gfs2_dinode_print(&ip
->i_di
);
339 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
343 ip
->i_di
.di_nlink
= nlink
;
344 ip
->i_di
.di_ctime
= get_seconds();
346 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
347 gfs2_dinode_out(&ip
->i_di
, dibh
->b_data
);
349 mark_inode_dirty(&ip
->i_inode
);
351 if (ip
->i_di
.di_nlink
== 0) {
352 struct gfs2_rgrpd
*rgd
;
353 struct gfs2_holder ri_gh
, rg_gh
;
355 error
= gfs2_rindex_hold(sdp
, &ri_gh
);
359 rgd
= gfs2_blk2rgrpd(sdp
, ip
->i_num
.no_addr
);
362 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0, &rg_gh
);
366 gfs2_unlink_di(&ip
->i_inode
); /* mark inode unlinked */
367 gfs2_glock_dq_uninit(&rg_gh
);
369 gfs2_glock_dq_uninit(&ri_gh
);
375 struct inode
*gfs2_lookup_simple(struct inode
*dip
, const char *name
)
378 gfs2_str2qstr(&qstr
, name
);
379 return gfs2_lookupi(dip
, &qstr
, 1, NULL
);
384 * gfs2_lookupi - Look up a filename in a directory and return its inode
385 * @d_gh: An initialized holder for the directory glock
386 * @name: The name of the inode to look for
387 * @is_root: If 1, ignore the caller's permissions
388 * @i_gh: An uninitialized holder for the new inode glock
390 * There will always be a vnode (Linux VFS inode) for the d_gh inode unless
396 struct inode
*gfs2_lookupi(struct inode
*dir
, const struct qstr
*name
,
397 int is_root
, struct nameidata
*nd
)
400 struct super_block
*sb
= dir
->i_sb
;
401 struct gfs2_inode
*dip
= GFS2_I(dir
);
402 struct gfs2_holder d_gh
;
403 struct gfs2_inum inum
;
406 struct inode
*inode
= NULL
;
408 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
409 return ERR_PTR(-ENAMETOOLONG
);
411 if ((name
->len
== 1 && memcmp(name
->name
, ".", 1) == 0) ||
412 (name
->len
== 2 && memcmp(name
->name
, "..", 2) == 0 &&
413 dir
== sb
->s_root
->d_inode
)) {
418 error
= gfs2_glock_nq_init(dip
->i_gl
, LM_ST_SHARED
, 0, &d_gh
);
420 return ERR_PTR(error
);
423 error
= permission(dir
, MAY_EXEC
, NULL
);
428 error
= gfs2_dir_search(dir
, name
, &inum
, &type
);
432 inode
= gfs2_inode_lookup(sb
, &inum
, type
);
435 gfs2_glock_dq_uninit(&d_gh
);
436 if (error
== -ENOENT
)
441 static int pick_formal_ino_1(struct gfs2_sbd
*sdp
, uint64_t *formal_ino
)
443 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
444 struct buffer_head
*bh
;
445 struct gfs2_inum_range ir
;
448 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
451 mutex_lock(&sdp
->sd_inum_mutex
);
453 error
= gfs2_meta_inode_buffer(ip
, &bh
);
455 mutex_unlock(&sdp
->sd_inum_mutex
);
460 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
463 *formal_ino
= ir
.ir_start
++;
465 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
466 gfs2_inum_range_out(&ir
,
467 bh
->b_data
+ sizeof(struct gfs2_dinode
));
469 mutex_unlock(&sdp
->sd_inum_mutex
);
476 mutex_unlock(&sdp
->sd_inum_mutex
);
482 static int pick_formal_ino_2(struct gfs2_sbd
*sdp
, uint64_t *formal_ino
)
484 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
485 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_inum_inode
);
486 struct gfs2_holder gh
;
487 struct buffer_head
*bh
;
488 struct gfs2_inum_range ir
;
491 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
495 error
= gfs2_trans_begin(sdp
, 2 * RES_DINODE
, 0);
498 mutex_lock(&sdp
->sd_inum_mutex
);
500 error
= gfs2_meta_inode_buffer(ip
, &bh
);
504 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
507 struct buffer_head
*m_bh
;
510 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
514 x
= *(uint64_t *)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
));
515 x
= y
= be64_to_cpu(x
);
517 ir
.ir_length
= GFS2_INUM_QUANTUM
;
518 x
+= GFS2_INUM_QUANTUM
;
520 gfs2_consist_inode(m_ip
);
522 gfs2_trans_add_bh(m_ip
->i_gl
, m_bh
, 1);
523 *(uint64_t *)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
)) = x
;
528 *formal_ino
= ir
.ir_start
++;
531 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
532 gfs2_inum_range_out(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
538 mutex_unlock(&sdp
->sd_inum_mutex
);
542 gfs2_glock_dq_uninit(&gh
);
547 static int pick_formal_ino(struct gfs2_sbd
*sdp
, uint64_t *inum
)
551 error
= pick_formal_ino_1(sdp
, inum
);
555 error
= pick_formal_ino_2(sdp
, inum
);
561 * create_ok - OK to create a new on-disk inode here?
562 * @dip: Directory in which dinode is to be created
563 * @name: Name of new dinode
569 static int create_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
574 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
578 /* Don't create entries in an unlinked directory */
579 if (!dip
->i_di
.di_nlink
)
582 error
= gfs2_dir_search(&dip
->i_inode
, name
, NULL
, NULL
);
593 if (dip
->i_di
.di_entries
== (uint32_t)-1)
595 if (S_ISDIR(mode
) && dip
->i_di
.di_nlink
== (uint32_t)-1)
601 static void munge_mode_uid_gid(struct gfs2_inode
*dip
, unsigned int *mode
,
602 unsigned int *uid
, unsigned int *gid
)
604 if (GFS2_SB(&dip
->i_inode
)->sd_args
.ar_suiddir
&&
605 (dip
->i_di
.di_mode
& S_ISUID
) &&
609 else if (dip
->i_di
.di_uid
!= current
->fsuid
)
611 *uid
= dip
->i_di
.di_uid
;
613 *uid
= current
->fsuid
;
615 if (dip
->i_di
.di_mode
& S_ISGID
) {
618 *gid
= dip
->i_di
.di_gid
;
620 *gid
= current
->fsgid
;
623 static int alloc_dinode(struct gfs2_inode
*dip
, struct gfs2_inum
*inum
)
625 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
630 dip
->i_alloc
.al_requested
= RES_DINODE
;
631 error
= gfs2_inplace_reserve(dip
);
635 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+ RES_STATFS
, 0);
639 inum
->no_addr
= gfs2_alloc_di(dip
);
644 gfs2_inplace_release(dip
);
653 * init_dinode - Fill in a new dinode structure
654 * @dip: the directory this inode is being created in
655 * @gl: The glock covering the new inode
656 * @inum: the inode number
657 * @mode: the file permissions
663 static void init_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
664 struct gfs2_inum
*inum
, unsigned int mode
,
665 unsigned int uid
, unsigned int gid
)
667 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
668 struct gfs2_dinode
*di
;
669 struct buffer_head
*dibh
;
671 dibh
= gfs2_meta_new(gl
, inum
->no_addr
);
672 gfs2_trans_add_bh(gl
, dibh
, 1);
673 gfs2_metatype_set(dibh
, GFS2_METATYPE_DI
, GFS2_FORMAT_DI
);
674 gfs2_buffer_clear_tail(dibh
, sizeof(struct gfs2_dinode
));
675 di
= (struct gfs2_dinode
*)dibh
->b_data
;
677 di
->di_num
.no_formal_ino
= cpu_to_be64(inum
->no_formal_ino
);
678 di
->di_num
.no_addr
= cpu_to_be64(inum
->no_addr
);
679 di
->di_mode
= cpu_to_be32(mode
);
680 di
->di_uid
= cpu_to_be32(uid
);
681 di
->di_gid
= cpu_to_be32(gid
);
682 di
->di_nlink
= cpu_to_be32(0);
683 di
->di_size
= cpu_to_be64(0);
684 di
->di_blocks
= cpu_to_be64(1);
685 di
->di_atime
= di
->di_mtime
= di
->di_ctime
= cpu_to_be64(get_seconds());
686 di
->di_major
= di
->di_minor
= cpu_to_be32(0);
687 di
->di_goal_meta
= di
->di_goal_data
= cpu_to_be64(inum
->no_addr
);
688 di
->__pad
[0] = di
->__pad
[1] = 0;
689 di
->di_flags
= cpu_to_be32(0);
692 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_JDATA
) ||
693 gfs2_tune_get(sdp
, gt_new_files_jdata
))
694 di
->di_flags
|= cpu_to_be32(GFS2_DIF_JDATA
);
695 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_DIRECTIO
) ||
696 gfs2_tune_get(sdp
, gt_new_files_directio
))
697 di
->di_flags
|= cpu_to_be32(GFS2_DIF_DIRECTIO
);
698 } else if (S_ISDIR(mode
)) {
699 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
700 GFS2_DIF_INHERIT_DIRECTIO
);
701 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
702 GFS2_DIF_INHERIT_JDATA
);
706 di
->di_height
= cpu_to_be32(0);
709 di
->di_depth
= cpu_to_be16(0);
710 di
->di_entries
= cpu_to_be32(0);
711 memset(&di
->__pad4
, 0, sizeof(di
->__pad4
));
712 di
->di_eattr
= cpu_to_be64(0);
713 memset(&di
->di_reserved
, 0, sizeof(di
->di_reserved
));
718 static int make_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
719 unsigned int mode
, struct gfs2_inum
*inum
)
721 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
722 unsigned int uid
, gid
;
725 munge_mode_uid_gid(dip
, &mode
, &uid
, &gid
);
728 error
= gfs2_quota_lock(dip
, uid
, gid
);
732 error
= gfs2_quota_check(dip
, uid
, gid
);
736 error
= gfs2_trans_begin(sdp
, RES_DINODE
+ RES_QUOTA
, 0);
740 init_dinode(dip
, gl
, inum
, mode
, uid
, gid
);
741 gfs2_quota_change(dip
, +1, uid
, gid
);
745 gfs2_quota_unlock(dip
);
751 static int link_dinode(struct gfs2_inode
*dip
, const struct qstr
*name
,
752 struct gfs2_inode
*ip
)
754 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
755 struct gfs2_alloc
*al
;
757 struct buffer_head
*dibh
;
760 al
= gfs2_alloc_get(dip
);
762 error
= gfs2_quota_lock(dip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
766 error
= alloc_required
= gfs2_diradd_alloc_required(&dip
->i_inode
, name
);
767 if (alloc_required
< 0)
769 if (alloc_required
) {
770 error
= gfs2_quota_check(dip
, dip
->i_di
.di_uid
,
773 goto fail_quota_locks
;
775 al
->al_requested
= sdp
->sd_max_dirres
;
777 error
= gfs2_inplace_reserve(dip
);
779 goto fail_quota_locks
;
781 error
= gfs2_trans_begin(sdp
, sdp
->sd_max_dirres
+
782 al
->al_rgd
->rd_ri
.ri_length
+
784 RES_STATFS
+ RES_QUOTA
, 0);
788 error
= gfs2_trans_begin(sdp
, RES_LEAF
+ 2 * RES_DINODE
, 0);
790 goto fail_quota_locks
;
793 error
= gfs2_dir_add(&dip
->i_inode
, name
, &ip
->i_num
, IF2DT(ip
->i_di
.di_mode
));
797 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
800 ip
->i_di
.di_nlink
= 1;
801 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
802 gfs2_dinode_out(&ip
->i_di
, dibh
->b_data
);
810 if (dip
->i_alloc
.al_rgd
)
811 gfs2_inplace_release(dip
);
814 gfs2_quota_unlock(dip
);
822 * gfs2_createi - Create a new inode
823 * @ghs: An array of two holders
824 * @name: The name of the new file
825 * @mode: the permissions on the new inode
827 * @ghs[0] is an initialized holder for the directory
828 * @ghs[1] is the holder for the inode lock
830 * If the return value is not NULL, the glocks on both the directory and the new
831 * file are held. A transaction has been started and an inplace reservation
837 struct inode
*gfs2_createi(struct gfs2_holder
*ghs
, const struct qstr
*name
,
841 struct gfs2_inode
*dip
= ghs
->gh_gl
->gl_object
;
842 struct inode
*dir
= &dip
->i_inode
;
843 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
844 struct gfs2_inum inum
;
847 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
848 return ERR_PTR(-ENAMETOOLONG
);
850 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, 0, ghs
);
851 error
= gfs2_glock_nq(ghs
);
855 error
= create_ok(dip
, name
, mode
);
859 error
= pick_formal_ino(sdp
, &inum
.no_formal_ino
);
863 error
= alloc_dinode(dip
, &inum
);
867 if (inum
.no_addr
< dip
->i_num
.no_addr
) {
870 error
= gfs2_glock_nq_num(sdp
, inum
.no_addr
,
871 &gfs2_inode_glops
, LM_ST_EXCLUSIVE
,
874 return ERR_PTR(error
);
877 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, 0, ghs
);
878 error
= gfs2_glock_nq(ghs
);
880 gfs2_glock_dq_uninit(ghs
+ 1);
881 return ERR_PTR(error
);
884 error
= create_ok(dip
, name
, mode
);
888 error
= gfs2_glock_nq_num(sdp
, inum
.no_addr
,
889 &gfs2_inode_glops
, LM_ST_EXCLUSIVE
,
895 error
= make_dinode(dip
, ghs
[1].gh_gl
, mode
, &inum
);
899 inode
= gfs2_inode_lookup(dir
->i_sb
, &inum
, IF2DT(mode
));
903 error
= gfs2_inode_refresh(GFS2_I(inode
));
907 error
= gfs2_acl_create(dip
, GFS2_I(inode
));
911 error
= link_dinode(dip
, name
, GFS2_I(inode
));
916 return ERR_PTR(-ENOMEM
);
922 gfs2_glock_dq_uninit(ghs
+ 1);
926 return ERR_PTR(error
);
930 * gfs2_rmdiri - Remove a directory
931 * @dip: The parent directory of the directory to be removed
932 * @name: The name of the directory to be removed
933 * @ip: The GFS2 inode of the directory to be removed
935 * Assumes Glocks on dip and ip are held
940 int gfs2_rmdiri(struct gfs2_inode
*dip
, const struct qstr
*name
,
941 struct gfs2_inode
*ip
)
946 if (ip
->i_di
.di_entries
!= 2) {
947 if (gfs2_consist_inode(ip
))
948 gfs2_dinode_print(&ip
->i_di
);
952 error
= gfs2_dir_del(dip
, name
);
956 error
= gfs2_change_nlink(dip
, -1);
960 gfs2_str2qstr(&dotname
, ".");
961 error
= gfs2_dir_del(ip
, &dotname
);
965 gfs2_str2qstr(&dotname
, "..");
966 error
= gfs2_dir_del(ip
, &dotname
);
970 error
= gfs2_change_nlink(ip
, -2);
978 * gfs2_unlink_ok - check to see that a inode is still in a directory
979 * @dip: the directory
980 * @name: the name of the file
983 * Assumes that the lock on (at least) @dip is held.
985 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
988 int gfs2_unlink_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
989 struct gfs2_inode
*ip
)
991 struct gfs2_inum inum
;
995 if (IS_IMMUTABLE(&ip
->i_inode
) || IS_APPEND(&ip
->i_inode
))
998 if ((dip
->i_di
.di_mode
& S_ISVTX
) &&
999 dip
->i_di
.di_uid
!= current
->fsuid
&&
1000 ip
->i_di
.di_uid
!= current
->fsuid
&& !capable(CAP_FOWNER
))
1003 if (IS_APPEND(&dip
->i_inode
))
1006 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
1010 error
= gfs2_dir_search(&dip
->i_inode
, name
, &inum
, &type
);
1014 if (!gfs2_inum_equal(&inum
, &ip
->i_num
))
1017 if (IF2DT(ip
->i_di
.di_mode
) != type
) {
1018 gfs2_consist_inode(dip
);
1026 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1030 * Follow @to back to the root and make sure we don't encounter @this
1031 * Assumes we already hold the rename lock.
1036 int gfs2_ok_to_move(struct gfs2_inode
*this, struct gfs2_inode
*to
)
1038 struct inode
*dir
= &to
->i_inode
;
1039 struct super_block
*sb
= dir
->i_sb
;
1044 gfs2_str2qstr(&dotdot
, "..");
1049 if (dir
== &this->i_inode
) {
1053 if (dir
== sb
->s_root
->d_inode
) {
1058 tmp
= gfs2_lookupi(dir
, &dotdot
, 1, NULL
);
1060 error
= PTR_ERR(tmp
);
1074 * gfs2_readlinki - return the contents of a symlink
1075 * @ip: the symlink's inode
1076 * @buf: a pointer to the buffer to be filled
1077 * @len: a pointer to the length of @buf
1079 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1080 * to be freed by the caller.
1085 int gfs2_readlinki(struct gfs2_inode
*ip
, char **buf
, unsigned int *len
)
1087 struct gfs2_holder i_gh
;
1088 struct buffer_head
*dibh
;
1092 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, GL_ATIME
, &i_gh
);
1093 error
= gfs2_glock_nq_atime(&i_gh
);
1095 gfs2_holder_uninit(&i_gh
);
1099 if (!ip
->i_di
.di_size
) {
1100 gfs2_consist_inode(ip
);
1105 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1109 x
= ip
->i_di
.di_size
+ 1;
1111 *buf
= kmalloc(x
, GFP_KERNEL
);
1118 memcpy(*buf
, dibh
->b_data
+ sizeof(struct gfs2_dinode
), x
);
1124 gfs2_glock_dq_uninit(&i_gh
);
1129 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1130 * conditionally update the inode's atime
1131 * @gh: the holder to acquire
1133 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1134 * Update if the difference between the current time and the inode's current
1135 * atime is greater than an interval specified at mount.
1140 int gfs2_glock_nq_atime(struct gfs2_holder
*gh
)
1142 struct gfs2_glock
*gl
= gh
->gh_gl
;
1143 struct gfs2_sbd
*sdp
= gl
->gl_sbd
;
1144 struct gfs2_inode
*ip
= gl
->gl_object
;
1145 int64_t curtime
, quantum
= gfs2_tune_get(sdp
, gt_atime_quantum
);
1150 if (gfs2_assert_warn(sdp
, gh
->gh_flags
& GL_ATIME
) ||
1151 gfs2_assert_warn(sdp
, !(gh
->gh_flags
& GL_ASYNC
)) ||
1152 gfs2_assert_warn(sdp
, gl
->gl_ops
== &gfs2_inode_glops
))
1155 state
= gh
->gh_state
;
1156 flags
= gh
->gh_flags
;
1158 error
= gfs2_glock_nq(gh
);
1162 if (test_bit(SDF_NOATIME
, &sdp
->sd_flags
) ||
1163 (sdp
->sd_vfs
->s_flags
& MS_RDONLY
))
1166 curtime
= get_seconds();
1167 if (curtime
- ip
->i_di
.di_atime
>= quantum
) {
1169 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, gh
->gh_flags
& ~LM_FLAG_ANY
,
1171 error
= gfs2_glock_nq(gh
);
1175 /* Verify that atime hasn't been updated while we were
1176 trying to get exclusive lock. */
1178 curtime
= get_seconds();
1179 if (curtime
- ip
->i_di
.di_atime
>= quantum
) {
1180 struct buffer_head
*dibh
;
1182 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
1183 if (error
== -EROFS
)
1188 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1190 goto fail_end_trans
;
1192 ip
->i_di
.di_atime
= curtime
;
1194 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1195 gfs2_dinode_out(&ip
->i_di
, dibh
->b_data
);
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 * glock_compare_atime - Compare two struct gfs2_glock structures for sort
1222 * @arg_a: the first structure
1223 * @arg_b: the second structure
1225 * Returns: 1 if A > B
1230 static int glock_compare_atime(const void *arg_a
, const void *arg_b
)
1232 struct gfs2_holder
*gh_a
= *(struct gfs2_holder
**)arg_a
;
1233 struct gfs2_holder
*gh_b
= *(struct gfs2_holder
**)arg_b
;
1234 struct lm_lockname
*a
= &gh_a
->gh_gl
->gl_name
;
1235 struct lm_lockname
*b
= &gh_b
->gh_gl
->gl_name
;
1238 if (a
->ln_number
> b
->ln_number
)
1240 else if (a
->ln_number
< b
->ln_number
)
1243 if (gh_a
->gh_state
== LM_ST_SHARED
&&
1244 gh_b
->gh_state
== LM_ST_EXCLUSIVE
)
1246 else if (gh_a
->gh_state
== LM_ST_SHARED
&&
1247 (gh_b
->gh_flags
& GL_ATIME
))
1255 * gfs2_glock_nq_m_atime - acquire multiple glocks where one may need an
1257 * @num_gh: the number of structures
1258 * @ghs: an array of struct gfs2_holder structures
1260 * Returns: 0 on success (all glocks acquired),
1261 * errno on failure (no glocks acquired)
1264 int gfs2_glock_nq_m_atime(unsigned int num_gh
, struct gfs2_holder
*ghs
)
1266 struct gfs2_holder
**p
;
1274 ghs
->gh_flags
&= ~(LM_FLAG_TRY
| GL_ASYNC
);
1275 if (ghs
->gh_flags
& GL_ATIME
)
1276 error
= gfs2_glock_nq_atime(ghs
);
1278 error
= gfs2_glock_nq(ghs
);
1282 p
= kcalloc(num_gh
, sizeof(struct gfs2_holder
*), GFP_KERNEL
);
1286 for (x
= 0; x
< num_gh
; x
++)
1289 sort(p
, num_gh
, sizeof(struct gfs2_holder
*), glock_compare_atime
,NULL
);
1291 for (x
= 0; x
< num_gh
; x
++) {
1292 p
[x
]->gh_flags
&= ~(LM_FLAG_TRY
| GL_ASYNC
);
1294 if (p
[x
]->gh_flags
& GL_ATIME
)
1295 error
= gfs2_glock_nq_atime(p
[x
]);
1297 error
= gfs2_glock_nq(p
[x
]);
1301 gfs2_glock_dq(p
[x
]);
1313 __gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1315 struct buffer_head
*dibh
;
1318 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1320 error
= inode_setattr(&ip
->i_inode
, attr
);
1321 gfs2_assert_warn(GFS2_SB(&ip
->i_inode
), !error
);
1322 gfs2_inode_attr_out(ip
);
1324 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1325 gfs2_dinode_out(&ip
->i_di
, dibh
->b_data
);
1332 * gfs2_setattr_simple -
1336 * Called with a reference on the vnode.
1341 int gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1345 if (current
->journal_info
)
1346 return __gfs2_setattr_simple(ip
, attr
);
1348 error
= gfs2_trans_begin(GFS2_SB(&ip
->i_inode
), RES_DINODE
, 0);
1352 error
= __gfs2_setattr_simple(ip
, attr
);
1354 gfs2_trans_end(GFS2_SB(&ip
->i_inode
));