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 * gfs2_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 void gfs2_inode_attr_in(struct gfs2_inode
*ip
)
49 struct inode
*inode
= &ip
->i_inode
;
51 inode
->i_ino
= ip
->i_num
.no_addr
;
53 switch (ip
->i_di
.di_mode
& S_IFMT
) {
56 inode
->i_rdev
= MKDEV(ip
->i_di
.di_major
, ip
->i_di
.di_minor
);
63 inode
->i_mode
= ip
->i_di
.di_mode
;
64 inode
->i_nlink
= ip
->i_di
.di_nlink
;
65 inode
->i_uid
= ip
->i_di
.di_uid
;
66 inode
->i_gid
= ip
->i_di
.di_gid
;
67 i_size_write(inode
, ip
->i_di
.di_size
);
68 inode
->i_atime
.tv_sec
= ip
->i_di
.di_atime
;
69 inode
->i_mtime
.tv_sec
= ip
->i_di
.di_mtime
;
70 inode
->i_ctime
.tv_sec
= ip
->i_di
.di_ctime
;
71 inode
->i_atime
.tv_nsec
= 0;
72 inode
->i_mtime
.tv_nsec
= 0;
73 inode
->i_ctime
.tv_nsec
= 0;
74 inode
->i_blksize
= PAGE_SIZE
;
75 inode
->i_blocks
= ip
->i_di
.di_blocks
<<
76 (GFS2_SB(inode
)->sd_sb
.sb_bsize_shift
- GFS2_BASIC_BLOCK_SHIFT
);
78 if (ip
->i_di
.di_flags
& GFS2_DIF_IMMUTABLE
)
79 inode
->i_flags
|= S_IMMUTABLE
;
81 inode
->i_flags
&= ~S_IMMUTABLE
;
83 if (ip
->i_di
.di_flags
& GFS2_DIF_APPENDONLY
)
84 inode
->i_flags
|= S_APPEND
;
86 inode
->i_flags
&= ~S_APPEND
;
90 * gfs2_inode_attr_out - Copy attributes from VFS inode into the dinode
93 * Only copy out the attributes that we want the VFS layer
94 * to be able to modify.
97 void gfs2_inode_attr_out(struct gfs2_inode
*ip
)
99 struct inode
*inode
= &ip
->i_inode
;
101 gfs2_assert_withdraw(GFS2_SB(inode
),
102 (ip
->i_di
.di_mode
& S_IFMT
) == (inode
->i_mode
& S_IFMT
));
103 ip
->i_di
.di_mode
= inode
->i_mode
;
104 ip
->i_di
.di_uid
= inode
->i_uid
;
105 ip
->i_di
.di_gid
= inode
->i_gid
;
106 ip
->i_di
.di_atime
= inode
->i_atime
.tv_sec
;
107 ip
->i_di
.di_mtime
= inode
->i_mtime
.tv_sec
;
108 ip
->i_di
.di_ctime
= inode
->i_ctime
.tv_sec
;
111 static int iget_test(struct inode
*inode
, void *opaque
)
113 struct gfs2_inode
*ip
= GFS2_I(inode
);
114 struct gfs2_inum
*inum
= opaque
;
116 if (ip
&& ip
->i_num
.no_addr
== inum
->no_addr
)
122 static int iget_set(struct inode
*inode
, void *opaque
)
124 struct gfs2_inode
*ip
= GFS2_I(inode
);
125 struct gfs2_inum
*inum
= opaque
;
131 struct inode
*gfs2_ilookup(struct super_block
*sb
, struct gfs2_inum
*inum
)
133 return ilookup5(sb
, (unsigned long)inum
->no_formal_ino
,
137 static struct inode
*gfs2_iget(struct super_block
*sb
, struct gfs2_inum
*inum
)
139 return iget5_locked(sb
, (unsigned long)inum
->no_formal_ino
,
140 iget_test
, iget_set
, inum
);
144 * gfs2_inode_lookup - Lookup an inode
145 * @sb: The super block
146 * @inum: The inode number
147 * @type: The type of the inode
149 * Returns: A VFS inode, or an error
152 struct inode
*gfs2_inode_lookup(struct super_block
*sb
, struct gfs2_inum
*inum
, unsigned int type
)
154 struct inode
*inode
= gfs2_iget(sb
, inum
);
155 struct gfs2_inode
*ip
= GFS2_I(inode
);
156 struct gfs2_glock
*io_gl
;
159 if (inode
->i_state
& I_NEW
) {
160 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
161 umode_t mode
= DT2IF(type
);
162 inode
->u
.generic_ip
= ip
;
163 inode
->i_mode
= mode
;
166 inode
->i_op
= &gfs2_file_iops
;
167 inode
->i_fop
= &gfs2_file_fops
;
168 inode
->i_mapping
->a_ops
= &gfs2_file_aops
;
169 } else if (S_ISDIR(mode
)) {
170 inode
->i_op
= &gfs2_dir_iops
;
171 inode
->i_fop
= &gfs2_dir_fops
;
172 } else if (S_ISLNK(mode
)) {
173 inode
->i_op
= &gfs2_symlink_iops
;
175 inode
->i_op
= &gfs2_dev_iops
;
178 error
= gfs2_glock_get(sdp
, inum
->no_addr
, &gfs2_inode_glops
, CREATE
, &ip
->i_gl
);
181 ip
->i_gl
->gl_object
= ip
;
183 error
= gfs2_glock_get(sdp
, inum
->no_addr
, &gfs2_iopen_glops
, CREATE
, &io_gl
);
187 ip
->i_vn
= ip
->i_gl
->gl_vn
- 1;
188 error
= gfs2_glock_nq_init(io_gl
, LM_ST_SHARED
, GL_EXACT
, &ip
->i_iopen_gh
);
192 gfs2_glock_put(io_gl
);
193 unlock_new_inode(inode
);
198 gfs2_glock_put(io_gl
);
200 ip
->i_gl
->gl_object
= NULL
;
201 gfs2_glock_put(ip
->i_gl
);
204 return ERR_PTR(error
);
208 * gfs2_inode_refresh - Refresh the incore copy of the dinode
209 * @ip: The GFS2 inode
214 int gfs2_inode_refresh(struct gfs2_inode
*ip
)
216 struct buffer_head
*dibh
;
219 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
223 if (gfs2_metatype_check(GFS2_SB(&ip
->i_inode
), dibh
, GFS2_METATYPE_DI
)) {
228 gfs2_dinode_in(&ip
->i_di
, dibh
->b_data
);
232 if (ip
->i_num
.no_addr
!= ip
->i_di
.di_num
.no_addr
) {
233 if (gfs2_consist_inode(ip
))
234 gfs2_dinode_print(&ip
->i_di
);
237 if (ip
->i_num
.no_formal_ino
!= ip
->i_di
.di_num
.no_formal_ino
)
240 ip
->i_vn
= ip
->i_gl
->gl_vn
;
245 int gfs2_dinode_dealloc(struct gfs2_inode
*ip
)
247 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
248 struct gfs2_alloc
*al
;
249 struct gfs2_rgrpd
*rgd
;
252 if (ip
->i_di
.di_blocks
!= 1) {
253 if (gfs2_consist_inode(ip
))
254 gfs2_dinode_print(&ip
->i_di
);
258 al
= gfs2_alloc_get(ip
);
260 error
= gfs2_quota_hold(ip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
264 error
= gfs2_rindex_hold(sdp
, &al
->al_ri_gh
);
268 rgd
= gfs2_blk2rgrpd(sdp
, ip
->i_num
.no_addr
);
270 gfs2_consist_inode(ip
);
272 goto out_rindex_relse
;
275 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0,
278 goto out_rindex_relse
;
280 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+
281 RES_STATFS
+ RES_QUOTA
, 1);
285 gfs2_trans_add_gl(ip
->i_gl
);
287 gfs2_free_di(rgd
, ip
);
290 clear_bit(GLF_STICKY
, &ip
->i_gl
->gl_flags
);
293 gfs2_glock_dq_uninit(&al
->al_rgd_gh
);
295 gfs2_glock_dq_uninit(&al
->al_ri_gh
);
297 gfs2_quota_unhold(ip
);
304 * gfs2_change_nlink - Change nlink count on inode
305 * @ip: The GFS2 inode
306 * @diff: The change in the nlink count required
311 int gfs2_change_nlink(struct gfs2_inode
*ip
, int diff
)
313 struct gfs2_sbd
*sdp
= ip
->i_inode
.i_sb
->s_fs_info
;
314 struct buffer_head
*dibh
;
318 BUG_ON(ip
->i_di
.di_nlink
!= ip
->i_inode
.i_nlink
);
319 nlink
= ip
->i_di
.di_nlink
+ diff
;
321 /* If we are reducing the nlink count, but the new value ends up being
322 bigger than the old one, we must have underflowed. */
323 if (diff
< 0 && nlink
> ip
->i_di
.di_nlink
) {
324 if (gfs2_consist_inode(ip
))
325 gfs2_dinode_print(&ip
->i_di
);
329 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
333 ip
->i_di
.di_nlink
= nlink
;
334 ip
->i_di
.di_ctime
= get_seconds();
335 ip
->i_inode
.i_nlink
= nlink
;
337 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
338 gfs2_dinode_out(&ip
->i_di
, dibh
->b_data
);
340 mark_inode_dirty(&ip
->i_inode
);
342 if (ip
->i_di
.di_nlink
== 0) {
343 struct gfs2_rgrpd
*rgd
;
344 struct gfs2_holder ri_gh
, rg_gh
;
346 error
= gfs2_rindex_hold(sdp
, &ri_gh
);
350 rgd
= gfs2_blk2rgrpd(sdp
, ip
->i_num
.no_addr
);
353 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0, &rg_gh
);
357 gfs2_unlink_di(&ip
->i_inode
); /* mark inode unlinked */
358 gfs2_glock_dq_uninit(&rg_gh
);
360 gfs2_glock_dq_uninit(&ri_gh
);
366 struct inode
*gfs2_lookup_simple(struct inode
*dip
, const char *name
)
369 gfs2_str2qstr(&qstr
, name
);
370 return gfs2_lookupi(dip
, &qstr
, 1, NULL
);
375 * gfs2_lookupi - Look up a filename in a directory and return its inode
376 * @d_gh: An initialized holder for the directory glock
377 * @name: The name of the inode to look for
378 * @is_root: If 1, ignore the caller's permissions
379 * @i_gh: An uninitialized holder for the new inode glock
381 * There will always be a vnode (Linux VFS inode) for the d_gh inode unless
387 struct inode
*gfs2_lookupi(struct inode
*dir
, const struct qstr
*name
,
388 int is_root
, struct nameidata
*nd
)
391 struct super_block
*sb
= dir
->i_sb
;
392 struct gfs2_inode
*dip
= GFS2_I(dir
);
393 struct gfs2_holder d_gh
;
394 struct gfs2_inum inum
;
397 struct inode
*inode
= NULL
;
399 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
400 return ERR_PTR(-ENAMETOOLONG
);
402 if ((name
->len
== 1 && memcmp(name
->name
, ".", 1) == 0) ||
403 (name
->len
== 2 && memcmp(name
->name
, "..", 2) == 0 &&
404 dir
== sb
->s_root
->d_inode
)) {
409 error
= gfs2_glock_nq_init(dip
->i_gl
, LM_ST_SHARED
, 0, &d_gh
);
411 return ERR_PTR(error
);
414 error
= permission(dir
, MAY_EXEC
, NULL
);
419 error
= gfs2_dir_search(dir
, name
, &inum
, &type
);
423 inode
= gfs2_inode_lookup(sb
, &inum
, type
);
426 gfs2_glock_dq_uninit(&d_gh
);
427 if (error
== -ENOENT
)
432 static int pick_formal_ino_1(struct gfs2_sbd
*sdp
, uint64_t *formal_ino
)
434 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
435 struct buffer_head
*bh
;
436 struct gfs2_inum_range ir
;
439 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
442 mutex_lock(&sdp
->sd_inum_mutex
);
444 error
= gfs2_meta_inode_buffer(ip
, &bh
);
446 mutex_unlock(&sdp
->sd_inum_mutex
);
451 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
454 *formal_ino
= ir
.ir_start
++;
456 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
457 gfs2_inum_range_out(&ir
,
458 bh
->b_data
+ sizeof(struct gfs2_dinode
));
460 mutex_unlock(&sdp
->sd_inum_mutex
);
467 mutex_unlock(&sdp
->sd_inum_mutex
);
473 static int pick_formal_ino_2(struct gfs2_sbd
*sdp
, uint64_t *formal_ino
)
475 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_ir_inode
);
476 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_inum_inode
);
477 struct gfs2_holder gh
;
478 struct buffer_head
*bh
;
479 struct gfs2_inum_range ir
;
482 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
486 error
= gfs2_trans_begin(sdp
, 2 * RES_DINODE
, 0);
489 mutex_lock(&sdp
->sd_inum_mutex
);
491 error
= gfs2_meta_inode_buffer(ip
, &bh
);
495 gfs2_inum_range_in(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
498 struct buffer_head
*m_bh
;
501 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
505 x
= *(uint64_t *)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
));
506 x
= y
= be64_to_cpu(x
);
508 ir
.ir_length
= GFS2_INUM_QUANTUM
;
509 x
+= GFS2_INUM_QUANTUM
;
511 gfs2_consist_inode(m_ip
);
513 gfs2_trans_add_bh(m_ip
->i_gl
, m_bh
, 1);
514 *(uint64_t *)(m_bh
->b_data
+ sizeof(struct gfs2_dinode
)) = x
;
519 *formal_ino
= ir
.ir_start
++;
522 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
523 gfs2_inum_range_out(&ir
, bh
->b_data
+ sizeof(struct gfs2_dinode
));
529 mutex_unlock(&sdp
->sd_inum_mutex
);
533 gfs2_glock_dq_uninit(&gh
);
538 static int pick_formal_ino(struct gfs2_sbd
*sdp
, uint64_t *inum
)
542 error
= pick_formal_ino_1(sdp
, inum
);
546 error
= pick_formal_ino_2(sdp
, inum
);
552 * create_ok - OK to create a new on-disk inode here?
553 * @dip: Directory in which dinode is to be created
554 * @name: Name of new dinode
560 static int create_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
565 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
569 /* Don't create entries in an unlinked directory */
570 if (!dip
->i_di
.di_nlink
)
573 error
= gfs2_dir_search(&dip
->i_inode
, name
, NULL
, NULL
);
584 if (dip
->i_di
.di_entries
== (uint32_t)-1)
586 if (S_ISDIR(mode
) && dip
->i_di
.di_nlink
== (uint32_t)-1)
592 static void munge_mode_uid_gid(struct gfs2_inode
*dip
, unsigned int *mode
,
593 unsigned int *uid
, unsigned int *gid
)
595 if (GFS2_SB(&dip
->i_inode
)->sd_args
.ar_suiddir
&&
596 (dip
->i_di
.di_mode
& S_ISUID
) &&
600 else if (dip
->i_di
.di_uid
!= current
->fsuid
)
602 *uid
= dip
->i_di
.di_uid
;
604 *uid
= current
->fsuid
;
606 if (dip
->i_di
.di_mode
& S_ISGID
) {
609 *gid
= dip
->i_di
.di_gid
;
611 *gid
= current
->fsgid
;
614 static int alloc_dinode(struct gfs2_inode
*dip
, struct gfs2_inum
*inum
,
617 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
622 dip
->i_alloc
.al_requested
= RES_DINODE
;
623 error
= gfs2_inplace_reserve(dip
);
627 error
= gfs2_trans_begin(sdp
, RES_RG_BIT
+ RES_STATFS
, 0);
631 inum
->no_addr
= gfs2_alloc_di(dip
, generation
);
636 gfs2_inplace_release(dip
);
645 * init_dinode - Fill in a new dinode structure
646 * @dip: the directory this inode is being created in
647 * @gl: The glock covering the new inode
648 * @inum: the inode number
649 * @mode: the file permissions
655 static void init_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
656 const struct gfs2_inum
*inum
, unsigned int mode
,
657 unsigned int uid
, unsigned int gid
,
658 const u64
*generation
)
660 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
661 struct gfs2_dinode
*di
;
662 struct buffer_head
*dibh
;
664 dibh
= gfs2_meta_new(gl
, inum
->no_addr
);
665 gfs2_trans_add_bh(gl
, dibh
, 1);
666 gfs2_metatype_set(dibh
, GFS2_METATYPE_DI
, GFS2_FORMAT_DI
);
667 gfs2_buffer_clear_tail(dibh
, sizeof(struct gfs2_dinode
));
668 di
= (struct gfs2_dinode
*)dibh
->b_data
;
670 di
->di_num
.no_formal_ino
= cpu_to_be64(inum
->no_formal_ino
);
671 di
->di_num
.no_addr
= cpu_to_be64(inum
->no_addr
);
672 di
->di_mode
= cpu_to_be32(mode
);
673 di
->di_uid
= cpu_to_be32(uid
);
674 di
->di_gid
= cpu_to_be32(gid
);
675 di
->di_nlink
= cpu_to_be32(0);
676 di
->di_size
= cpu_to_be64(0);
677 di
->di_blocks
= cpu_to_be64(1);
678 di
->di_atime
= di
->di_mtime
= di
->di_ctime
= cpu_to_be64(get_seconds());
679 di
->di_major
= di
->di_minor
= cpu_to_be32(0);
680 di
->di_goal_meta
= di
->di_goal_data
= cpu_to_be64(inum
->no_addr
);
681 di
->di_generation
= cpu_to_be64(*generation
);
682 di
->di_flags
= cpu_to_be32(0);
685 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_JDATA
) ||
686 gfs2_tune_get(sdp
, gt_new_files_jdata
))
687 di
->di_flags
|= cpu_to_be32(GFS2_DIF_JDATA
);
688 if ((dip
->i_di
.di_flags
& GFS2_DIF_INHERIT_DIRECTIO
) ||
689 gfs2_tune_get(sdp
, gt_new_files_directio
))
690 di
->di_flags
|= cpu_to_be32(GFS2_DIF_DIRECTIO
);
691 } else if (S_ISDIR(mode
)) {
692 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
693 GFS2_DIF_INHERIT_DIRECTIO
);
694 di
->di_flags
|= cpu_to_be32(dip
->i_di
.di_flags
&
695 GFS2_DIF_INHERIT_JDATA
);
699 di
->di_height
= cpu_to_be32(0);
702 di
->di_depth
= cpu_to_be16(0);
703 di
->di_entries
= cpu_to_be32(0);
704 memset(&di
->__pad4
, 0, sizeof(di
->__pad4
));
705 di
->di_eattr
= cpu_to_be64(0);
706 memset(&di
->di_reserved
, 0, sizeof(di
->di_reserved
));
711 static int make_dinode(struct gfs2_inode
*dip
, struct gfs2_glock
*gl
,
712 unsigned int mode
, const struct gfs2_inum
*inum
,
713 const u64
*generation
)
715 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
716 unsigned int uid
, gid
;
719 munge_mode_uid_gid(dip
, &mode
, &uid
, &gid
);
722 error
= gfs2_quota_lock(dip
, uid
, gid
);
726 error
= gfs2_quota_check(dip
, uid
, gid
);
730 error
= gfs2_trans_begin(sdp
, RES_DINODE
+ RES_QUOTA
, 0);
734 init_dinode(dip
, gl
, inum
, mode
, uid
, gid
, generation
);
735 gfs2_quota_change(dip
, +1, uid
, gid
);
739 gfs2_quota_unlock(dip
);
745 static int link_dinode(struct gfs2_inode
*dip
, const struct qstr
*name
,
746 struct gfs2_inode
*ip
)
748 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
749 struct gfs2_alloc
*al
;
751 struct buffer_head
*dibh
;
754 al
= gfs2_alloc_get(dip
);
756 error
= gfs2_quota_lock(dip
, NO_QUOTA_CHANGE
, NO_QUOTA_CHANGE
);
760 error
= alloc_required
= gfs2_diradd_alloc_required(&dip
->i_inode
, name
);
761 if (alloc_required
< 0)
763 if (alloc_required
) {
764 error
= gfs2_quota_check(dip
, dip
->i_di
.di_uid
,
767 goto fail_quota_locks
;
769 al
->al_requested
= sdp
->sd_max_dirres
;
771 error
= gfs2_inplace_reserve(dip
);
773 goto fail_quota_locks
;
775 error
= gfs2_trans_begin(sdp
, sdp
->sd_max_dirres
+
776 al
->al_rgd
->rd_ri
.ri_length
+
778 RES_STATFS
+ RES_QUOTA
, 0);
782 error
= gfs2_trans_begin(sdp
, RES_LEAF
+ 2 * RES_DINODE
, 0);
784 goto fail_quota_locks
;
787 error
= gfs2_dir_add(&dip
->i_inode
, name
, &ip
->i_num
, IF2DT(ip
->i_di
.di_mode
));
791 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
794 ip
->i_di
.di_nlink
= 1;
795 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
796 gfs2_dinode_out(&ip
->i_di
, dibh
->b_data
);
804 if (dip
->i_alloc
.al_rgd
)
805 gfs2_inplace_release(dip
);
808 gfs2_quota_unlock(dip
);
816 * gfs2_createi - Create a new inode
817 * @ghs: An array of two holders
818 * @name: The name of the new file
819 * @mode: the permissions on the new inode
821 * @ghs[0] is an initialized holder for the directory
822 * @ghs[1] is the holder for the inode lock
824 * If the return value is not NULL, the glocks on both the directory and the new
825 * file are held. A transaction has been started and an inplace reservation
831 struct inode
*gfs2_createi(struct gfs2_holder
*ghs
, const struct qstr
*name
,
835 struct gfs2_inode
*dip
= ghs
->gh_gl
->gl_object
;
836 struct inode
*dir
= &dip
->i_inode
;
837 struct gfs2_sbd
*sdp
= GFS2_SB(&dip
->i_inode
);
838 struct gfs2_inum inum
;
842 if (!name
->len
|| name
->len
> GFS2_FNAMESIZE
)
843 return ERR_PTR(-ENAMETOOLONG
);
845 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, 0, ghs
);
846 error
= gfs2_glock_nq(ghs
);
850 error
= create_ok(dip
, name
, mode
);
854 error
= pick_formal_ino(sdp
, &inum
.no_formal_ino
);
858 error
= alloc_dinode(dip
, &inum
, &generation
);
862 if (inum
.no_addr
< dip
->i_num
.no_addr
) {
865 error
= gfs2_glock_nq_num(sdp
, inum
.no_addr
,
866 &gfs2_inode_glops
, LM_ST_EXCLUSIVE
,
869 return ERR_PTR(error
);
872 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, 0, ghs
);
873 error
= gfs2_glock_nq(ghs
);
875 gfs2_glock_dq_uninit(ghs
+ 1);
876 return ERR_PTR(error
);
879 error
= create_ok(dip
, name
, mode
);
883 error
= gfs2_glock_nq_num(sdp
, inum
.no_addr
,
884 &gfs2_inode_glops
, LM_ST_EXCLUSIVE
,
890 error
= make_dinode(dip
, ghs
[1].gh_gl
, mode
, &inum
, &generation
);
894 inode
= gfs2_inode_lookup(dir
->i_sb
, &inum
, IF2DT(mode
));
898 error
= gfs2_inode_refresh(GFS2_I(inode
));
902 error
= gfs2_acl_create(dip
, GFS2_I(inode
));
906 error
= link_dinode(dip
, name
, GFS2_I(inode
));
911 return ERR_PTR(-ENOMEM
);
917 gfs2_glock_dq_uninit(ghs
+ 1);
921 return ERR_PTR(error
);
925 * gfs2_rmdiri - Remove a directory
926 * @dip: The parent directory of the directory to be removed
927 * @name: The name of the directory to be removed
928 * @ip: The GFS2 inode of the directory to be removed
930 * Assumes Glocks on dip and ip are held
935 int gfs2_rmdiri(struct gfs2_inode
*dip
, const struct qstr
*name
,
936 struct gfs2_inode
*ip
)
941 if (ip
->i_di
.di_entries
!= 2) {
942 if (gfs2_consist_inode(ip
))
943 gfs2_dinode_print(&ip
->i_di
);
947 error
= gfs2_dir_del(dip
, name
);
951 error
= gfs2_change_nlink(dip
, -1);
955 gfs2_str2qstr(&dotname
, ".");
956 error
= gfs2_dir_del(ip
, &dotname
);
960 gfs2_str2qstr(&dotname
, "..");
961 error
= gfs2_dir_del(ip
, &dotname
);
965 error
= gfs2_change_nlink(ip
, -2);
973 * gfs2_unlink_ok - check to see that a inode is still in a directory
974 * @dip: the directory
975 * @name: the name of the file
978 * Assumes that the lock on (at least) @dip is held.
980 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
983 int gfs2_unlink_ok(struct gfs2_inode
*dip
, const struct qstr
*name
,
984 struct gfs2_inode
*ip
)
986 struct gfs2_inum inum
;
990 if (IS_IMMUTABLE(&ip
->i_inode
) || IS_APPEND(&ip
->i_inode
))
993 if ((dip
->i_di
.di_mode
& S_ISVTX
) &&
994 dip
->i_di
.di_uid
!= current
->fsuid
&&
995 ip
->i_di
.di_uid
!= current
->fsuid
&& !capable(CAP_FOWNER
))
998 if (IS_APPEND(&dip
->i_inode
))
1001 error
= permission(&dip
->i_inode
, MAY_WRITE
| MAY_EXEC
, NULL
);
1005 error
= gfs2_dir_search(&dip
->i_inode
, name
, &inum
, &type
);
1009 if (!gfs2_inum_equal(&inum
, &ip
->i_num
))
1012 if (IF2DT(ip
->i_di
.di_mode
) != type
) {
1013 gfs2_consist_inode(dip
);
1021 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1025 * Follow @to back to the root and make sure we don't encounter @this
1026 * Assumes we already hold the rename lock.
1031 int gfs2_ok_to_move(struct gfs2_inode
*this, struct gfs2_inode
*to
)
1033 struct inode
*dir
= &to
->i_inode
;
1034 struct super_block
*sb
= dir
->i_sb
;
1039 gfs2_str2qstr(&dotdot
, "..");
1044 if (dir
== &this->i_inode
) {
1048 if (dir
== sb
->s_root
->d_inode
) {
1053 tmp
= gfs2_lookupi(dir
, &dotdot
, 1, NULL
);
1055 error
= PTR_ERR(tmp
);
1069 * gfs2_readlinki - return the contents of a symlink
1070 * @ip: the symlink's inode
1071 * @buf: a pointer to the buffer to be filled
1072 * @len: a pointer to the length of @buf
1074 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1075 * to be freed by the caller.
1080 int gfs2_readlinki(struct gfs2_inode
*ip
, char **buf
, unsigned int *len
)
1082 struct gfs2_holder i_gh
;
1083 struct buffer_head
*dibh
;
1087 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, GL_ATIME
, &i_gh
);
1088 error
= gfs2_glock_nq_atime(&i_gh
);
1090 gfs2_holder_uninit(&i_gh
);
1094 if (!ip
->i_di
.di_size
) {
1095 gfs2_consist_inode(ip
);
1100 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1104 x
= ip
->i_di
.di_size
+ 1;
1106 *buf
= kmalloc(x
, GFP_KERNEL
);
1113 memcpy(*buf
, dibh
->b_data
+ sizeof(struct gfs2_dinode
), x
);
1119 gfs2_glock_dq_uninit(&i_gh
);
1124 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1125 * conditionally update the inode's atime
1126 * @gh: the holder to acquire
1128 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1129 * Update if the difference between the current time and the inode's current
1130 * atime is greater than an interval specified at mount.
1135 int gfs2_glock_nq_atime(struct gfs2_holder
*gh
)
1137 struct gfs2_glock
*gl
= gh
->gh_gl
;
1138 struct gfs2_sbd
*sdp
= gl
->gl_sbd
;
1139 struct gfs2_inode
*ip
= gl
->gl_object
;
1140 int64_t curtime
, quantum
= gfs2_tune_get(sdp
, gt_atime_quantum
);
1145 if (gfs2_assert_warn(sdp
, gh
->gh_flags
& GL_ATIME
) ||
1146 gfs2_assert_warn(sdp
, !(gh
->gh_flags
& GL_ASYNC
)) ||
1147 gfs2_assert_warn(sdp
, gl
->gl_ops
== &gfs2_inode_glops
))
1150 state
= gh
->gh_state
;
1151 flags
= gh
->gh_flags
;
1153 error
= gfs2_glock_nq(gh
);
1157 if (test_bit(SDF_NOATIME
, &sdp
->sd_flags
) ||
1158 (sdp
->sd_vfs
->s_flags
& MS_RDONLY
))
1161 curtime
= get_seconds();
1162 if (curtime
- ip
->i_di
.di_atime
>= quantum
) {
1164 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, gh
->gh_flags
& ~LM_FLAG_ANY
,
1166 error
= gfs2_glock_nq(gh
);
1170 /* Verify that atime hasn't been updated while we were
1171 trying to get exclusive lock. */
1173 curtime
= get_seconds();
1174 if (curtime
- ip
->i_di
.di_atime
>= quantum
) {
1175 struct buffer_head
*dibh
;
1177 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
1178 if (error
== -EROFS
)
1183 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1185 goto fail_end_trans
;
1187 ip
->i_di
.di_atime
= curtime
;
1189 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1190 gfs2_dinode_out(&ip
->i_di
, dibh
->b_data
);
1193 gfs2_trans_end(sdp
);
1196 /* If someone else has asked for the glock,
1197 unlock and let them have it. Then reacquire
1198 in the original state. */
1199 if (gfs2_glock_is_blocking(gl
)) {
1201 gfs2_holder_reinit(state
, flags
, gh
);
1202 return gfs2_glock_nq(gh
);
1209 gfs2_trans_end(sdp
);
1216 * glock_compare_atime - Compare two struct gfs2_glock structures for sort
1217 * @arg_a: the first structure
1218 * @arg_b: the second structure
1220 * Returns: 1 if A > B
1225 static int glock_compare_atime(const void *arg_a
, const void *arg_b
)
1227 struct gfs2_holder
*gh_a
= *(struct gfs2_holder
**)arg_a
;
1228 struct gfs2_holder
*gh_b
= *(struct gfs2_holder
**)arg_b
;
1229 struct lm_lockname
*a
= &gh_a
->gh_gl
->gl_name
;
1230 struct lm_lockname
*b
= &gh_b
->gh_gl
->gl_name
;
1233 if (a
->ln_number
> b
->ln_number
)
1235 else if (a
->ln_number
< b
->ln_number
)
1238 if (gh_a
->gh_state
== LM_ST_SHARED
&&
1239 gh_b
->gh_state
== LM_ST_EXCLUSIVE
)
1241 else if (gh_a
->gh_state
== LM_ST_SHARED
&&
1242 (gh_b
->gh_flags
& GL_ATIME
))
1250 * gfs2_glock_nq_m_atime - acquire multiple glocks where one may need an
1252 * @num_gh: the number of structures
1253 * @ghs: an array of struct gfs2_holder structures
1255 * Returns: 0 on success (all glocks acquired),
1256 * errno on failure (no glocks acquired)
1259 int gfs2_glock_nq_m_atime(unsigned int num_gh
, struct gfs2_holder
*ghs
)
1261 struct gfs2_holder
**p
;
1269 ghs
->gh_flags
&= ~(LM_FLAG_TRY
| GL_ASYNC
);
1270 if (ghs
->gh_flags
& GL_ATIME
)
1271 error
= gfs2_glock_nq_atime(ghs
);
1273 error
= gfs2_glock_nq(ghs
);
1277 p
= kcalloc(num_gh
, sizeof(struct gfs2_holder
*), GFP_KERNEL
);
1281 for (x
= 0; x
< num_gh
; x
++)
1284 sort(p
, num_gh
, sizeof(struct gfs2_holder
*), glock_compare_atime
,NULL
);
1286 for (x
= 0; x
< num_gh
; x
++) {
1287 p
[x
]->gh_flags
&= ~(LM_FLAG_TRY
| GL_ASYNC
);
1289 if (p
[x
]->gh_flags
& GL_ATIME
)
1290 error
= gfs2_glock_nq_atime(p
[x
]);
1292 error
= gfs2_glock_nq(p
[x
]);
1296 gfs2_glock_dq(p
[x
]);
1308 __gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1310 struct buffer_head
*dibh
;
1313 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1315 error
= inode_setattr(&ip
->i_inode
, attr
);
1316 gfs2_assert_warn(GFS2_SB(&ip
->i_inode
), !error
);
1317 gfs2_inode_attr_out(ip
);
1319 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1320 gfs2_dinode_out(&ip
->i_di
, dibh
->b_data
);
1327 * gfs2_setattr_simple -
1331 * Called with a reference on the vnode.
1336 int gfs2_setattr_simple(struct gfs2_inode
*ip
, struct iattr
*attr
)
1340 if (current
->journal_info
)
1341 return __gfs2_setattr_simple(ip
, attr
);
1343 error
= gfs2_trans_begin(GFS2_SB(&ip
->i_inode
), RES_DINODE
, 0);
1347 error
= __gfs2_setattr_simple(ip
, attr
);
1349 gfs2_trans_end(GFS2_SB(&ip
->i_inode
));