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>
21 #include <linux/time.h>
34 #include "ops_address.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 && test_bit(GIF_USER, &ip->i_flags))
56 static int iget_set(struct inode *inode, void *opaque)
58 struct gfs2_inode *ip = GFS2_I(inode);
59 u64 *no_addr = opaque;
61 inode->i_ino = (unsigned long)*no_addr;
62 ip->i_no_addr = *no_addr;
63 set_bit(GIF_USER, &ip->i_flags);
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 && test_bit(GIF_USER, &ip->i_flags)){
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;
108 set_bit(GIF_USER, &ip->i_flags);
112 static struct inode *gfs2_iget_skip(struct super_block *sb,
115 struct gfs2_skip_data data;
116 unsigned long hash = (unsigned long)no_addr;
118 data.no_addr = no_addr;
120 return iget5_locked(sb, hash, iget_skip_test, iget_skip_set, &data);
124 * GFS2 lookup code fills in vfs inode contents based on info obtained
125 * from directory entry inside gfs2_inode_lookup(). This has caused issues
126 * with NFS code path since its get_dentry routine doesn't have the relevant
127 * directory entry when gfs2_inode_lookup() is invoked. Part of the code
128 * segment inside gfs2_inode_lookup code needs to get moved around.
130 * Clean up I_LOCK and I_NEW as well.
133 void gfs2_set_iop(struct inode *inode)
135 struct gfs2_sbd *sdp = GFS2_SB(inode);
136 umode_t mode = inode->i_mode;
139 inode->i_op = &gfs2_file_iops;
140 if (sdp->sd_args.ar_localflocks)
141 inode->i_fop = &gfs2_file_fops_nolock;
143 inode->i_fop = &gfs2_file_fops;
144 } else if (S_ISDIR(mode)) {
145 inode->i_op = &gfs2_dir_iops;
146 if (sdp->sd_args.ar_localflocks)
147 inode->i_fop = &gfs2_dir_fops_nolock;
149 inode->i_fop = &gfs2_dir_fops;
150 } else if (S_ISLNK(mode)) {
151 inode->i_op = &gfs2_symlink_iops;
153 inode->i_op = &gfs2_file_iops;
154 init_special_inode(inode, inode->i_mode, inode->i_rdev);
157 unlock_new_inode(inode);
161 * gfs2_inode_lookup - Lookup an inode
162 * @sb: The super block
163 * @no_addr: The inode number
164 * @type: The type of the inode
165 * @skip_freeing: set this not return an inode if it is currently being freed.
167 * Returns: A VFS inode, or an error
170 struct inode *gfs2_inode_lookup(struct super_block *sb,
173 u64 no_formal_ino, int skip_freeing)
176 struct gfs2_inode *ip;
177 struct gfs2_glock *io_gl;
181 inode = gfs2_iget_skip(sb, no_addr);
183 inode = gfs2_iget(sb, no_addr);
187 return ERR_PTR(-ENOBUFS);
189 if (inode->i_state & I_NEW) {
190 struct gfs2_sbd *sdp = GFS2_SB(inode);
191 ip->i_no_formal_ino = no_formal_ino;
193 error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
196 ip->i_gl->gl_object = ip;
198 error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
202 set_bit(GIF_INVALID, &ip->i_flags);
203 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
206 ip->i_iopen_gh.gh_gl->gl_object = ip;
208 gfs2_glock_put(io_gl);
210 if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
213 inode->i_mode = DT2IF(type);
216 * We must read the inode in order to work out its type in
217 * this case. Note that this doesn't happen often as we normally
218 * know the type beforehand. This code path only occurs during
219 * unlinked inode recovery (where it is safe to do this glock,
220 * which is not true in the general case).
222 if (type == DT_UNKNOWN) {
223 struct gfs2_holder gh;
224 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
227 /* Inode is now uptodate */
228 gfs2_glock_dq_uninit(&gh);
237 gfs2_glock_dq(&ip->i_iopen_gh);
239 gfs2_glock_put(io_gl);
241 ip->i_gl->gl_object = NULL;
242 gfs2_glock_put(ip->i_gl);
245 return ERR_PTR(error);
248 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
250 const struct gfs2_dinode *str = buf;
251 struct timespec atime;
254 if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
256 ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
257 ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
258 ip->i_inode.i_rdev = 0;
259 switch (ip->i_inode.i_mode & S_IFMT) {
262 ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
263 be32_to_cpu(str->di_minor));
267 ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
268 ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
270 * We will need to review setting the nlink count here in the
271 * light of the forthcoming ro bind mount work. This is a reminder
274 ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
275 ip->i_disksize = be64_to_cpu(str->di_size);
276 i_size_write(&ip->i_inode, ip->i_disksize);
277 gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
278 atime.tv_sec = be64_to_cpu(str->di_atime);
279 atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
280 if (timespec_compare(&ip->i_inode.i_atime, &atime) < 0)
281 ip->i_inode.i_atime = atime;
282 ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
283 ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
284 ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
285 ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
287 ip->i_goal = be64_to_cpu(str->di_goal_meta);
288 ip->i_generation = be64_to_cpu(str->di_generation);
290 ip->i_diskflags = be32_to_cpu(str->di_flags);
291 gfs2_set_inode_flags(&ip->i_inode);
292 height = be16_to_cpu(str->di_height);
293 if (unlikely(height > GFS2_MAX_META_HEIGHT))
295 ip->i_height = (u8)height;
297 depth = be16_to_cpu(str->di_depth);
298 if (unlikely(depth > GFS2_DIR_MAX_DEPTH))
300 ip->i_depth = (u8)depth;
301 ip->i_entries = be32_to_cpu(str->di_entries);
303 ip->i_eattr = be64_to_cpu(str->di_eattr);
304 if (S_ISREG(ip->i_inode.i_mode))
305 gfs2_set_aops(&ip->i_inode);
309 if (gfs2_consist_inode(ip))
310 gfs2_dinode_print(ip);
315 * gfs2_inode_refresh - Refresh the incore copy of the dinode
316 * @ip: The GFS2 inode
321 int gfs2_inode_refresh(struct gfs2_inode *ip)
323 struct buffer_head *dibh;
326 error = gfs2_meta_inode_buffer(ip, &dibh);
330 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
335 error = gfs2_dinode_in(ip, dibh->b_data);
337 clear_bit(GIF_INVALID, &ip->i_flags);
342 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
344 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
345 struct gfs2_alloc *al;
346 struct gfs2_rgrpd *rgd;
349 if (gfs2_get_inode_blocks(&ip->i_inode) != 1) {
350 if (gfs2_consist_inode(ip))
351 gfs2_dinode_print(ip);
355 al = gfs2_alloc_get(ip);
359 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
363 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
367 rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
369 gfs2_consist_inode(ip);
371 goto out_rindex_relse;
374 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
377 goto out_rindex_relse;
379 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
383 set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
384 set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
386 gfs2_free_di(rgd, ip);
389 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
392 gfs2_glock_dq_uninit(&al->al_rgd_gh);
394 gfs2_glock_dq_uninit(&al->al_ri_gh);
396 gfs2_quota_unhold(ip);
403 * gfs2_change_nlink - Change nlink count on inode
404 * @ip: The GFS2 inode
405 * @diff: The change in the nlink count required
409 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
411 struct buffer_head *dibh;
415 BUG_ON(diff != 1 && diff != -1);
416 nlink = ip->i_inode.i_nlink + diff;
418 /* If we are reducing the nlink count, but the new value ends up being
419 bigger than the old one, we must have underflowed. */
420 if (diff < 0 && nlink > ip->i_inode.i_nlink) {
421 if (gfs2_consist_inode(ip))
422 gfs2_dinode_print(ip);
426 error = gfs2_meta_inode_buffer(ip, &dibh);
431 inc_nlink(&ip->i_inode);
433 drop_nlink(&ip->i_inode);
435 ip->i_inode.i_ctime = CURRENT_TIME;
437 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
438 gfs2_dinode_out(ip, dibh->b_data);
440 mark_inode_dirty(&ip->i_inode);
442 if (ip->i_inode.i_nlink == 0)
443 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
448 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
452 gfs2_str2qstr(&qstr, name);
453 inode = gfs2_lookupi(dip, &qstr, 1);
454 /* gfs2_lookupi has inconsistent callers: vfs
455 * related routines expect NULL for no entry found,
456 * gfs2_lookup_simple callers expect ENOENT
457 * and do not check for NULL.
460 return ERR_PTR(-ENOENT);
467 * gfs2_lookupi - Look up a filename in a directory and return its inode
468 * @d_gh: An initialized holder for the directory glock
469 * @name: The name of the inode to look for
470 * @is_root: If 1, ignore the caller's permissions
471 * @i_gh: An uninitialized holder for the new inode glock
473 * This can be called via the VFS filldir function when NFS is doing
474 * a readdirplus and the inode which its intending to stat isn't
475 * already in cache. In this case we must not take the directory glock
476 * again, since the readdir call will have already taken that lock.
481 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
484 struct super_block *sb = dir->i_sb;
485 struct gfs2_inode *dip = GFS2_I(dir);
486 struct gfs2_holder d_gh;
488 struct inode *inode = NULL;
491 if (!name->len || name->len > GFS2_FNAMESIZE)
492 return ERR_PTR(-ENAMETOOLONG);
494 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
495 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
496 dir == sb->s_root->d_inode)) {
501 if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
502 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
504 return ERR_PTR(error);
509 error = gfs2_permission(dir, MAY_EXEC);
514 inode = gfs2_dir_search(dir, name);
516 error = PTR_ERR(inode);
519 gfs2_glock_dq_uninit(&d_gh);
520 if (error == -ENOENT)
522 return inode ? inode : ERR_PTR(error);
525 static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf)
527 const struct gfs2_inum_range *str = buf;
529 ir->ir_start = be64_to_cpu(str->ir_start);
530 ir->ir_length = be64_to_cpu(str->ir_length);
533 static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf)
535 struct gfs2_inum_range *str = buf;
537 str->ir_start = cpu_to_be64(ir->ir_start);
538 str->ir_length = cpu_to_be64(ir->ir_length);
541 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
543 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
544 struct buffer_head *bh;
545 struct gfs2_inum_range_host ir;
548 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
551 mutex_lock(&sdp->sd_inum_mutex);
553 error = gfs2_meta_inode_buffer(ip, &bh);
555 mutex_unlock(&sdp->sd_inum_mutex);
560 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
563 *formal_ino = ir.ir_start++;
565 gfs2_trans_add_bh(ip->i_gl, bh, 1);
566 gfs2_inum_range_out(&ir,
567 bh->b_data + sizeof(struct gfs2_dinode));
569 mutex_unlock(&sdp->sd_inum_mutex);
576 mutex_unlock(&sdp->sd_inum_mutex);
582 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
584 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
585 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
586 struct gfs2_holder gh;
587 struct buffer_head *bh;
588 struct gfs2_inum_range_host ir;
591 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
595 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
598 mutex_lock(&sdp->sd_inum_mutex);
600 error = gfs2_meta_inode_buffer(ip, &bh);
604 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
607 struct buffer_head *m_bh;
611 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
615 z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
616 x = y = be64_to_cpu(z);
618 ir.ir_length = GFS2_INUM_QUANTUM;
619 x += GFS2_INUM_QUANTUM;
621 gfs2_consist_inode(m_ip);
623 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
624 *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z;
629 *formal_ino = ir.ir_start++;
632 gfs2_trans_add_bh(ip->i_gl, bh, 1);
633 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
638 mutex_unlock(&sdp->sd_inum_mutex);
641 gfs2_glock_dq_uninit(&gh);
645 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
649 error = pick_formal_ino_1(sdp, inum);
653 error = pick_formal_ino_2(sdp, inum);
659 * create_ok - OK to create a new on-disk inode here?
660 * @dip: Directory in which dinode is to be created
661 * @name: Name of new dinode
667 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
672 error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
676 /* Don't create entries in an unlinked directory */
677 if (!dip->i_inode.i_nlink)
680 error = gfs2_dir_check(&dip->i_inode, name, NULL);
691 if (dip->i_entries == (u32)-1)
693 if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
699 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
700 unsigned int *uid, unsigned int *gid)
702 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
703 (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
706 else if (dip->i_inode.i_uid != current_fsuid())
708 *uid = dip->i_inode.i_uid;
710 *uid = current_fsuid();
712 if (dip->i_inode.i_mode & S_ISGID) {
715 *gid = dip->i_inode.i_gid;
717 *gid = current_fsgid();
720 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
722 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
725 if (gfs2_alloc_get(dip) == NULL)
728 dip->i_alloc->al_requested = RES_DINODE;
729 error = gfs2_inplace_reserve(dip);
733 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
737 *no_addr = gfs2_alloc_di(dip, generation);
742 gfs2_inplace_release(dip);
749 * init_dinode - Fill in a new dinode structure
750 * @dip: the directory this inode is being created in
751 * @gl: The glock covering the new inode
752 * @inum: the inode number
753 * @mode: the file permissions
759 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
760 const struct gfs2_inum_host *inum, unsigned int mode,
761 unsigned int uid, unsigned int gid,
762 const u64 *generation, dev_t dev, struct buffer_head **bhp)
764 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
765 struct gfs2_dinode *di;
766 struct buffer_head *dibh;
767 struct timespec tv = CURRENT_TIME;
769 dibh = gfs2_meta_new(gl, inum->no_addr);
770 gfs2_trans_add_bh(gl, dibh, 1);
771 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
772 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
773 di = (struct gfs2_dinode *)dibh->b_data;
775 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
776 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
777 di->di_mode = cpu_to_be32(mode);
778 di->di_uid = cpu_to_be32(uid);
779 di->di_gid = cpu_to_be32(gid);
782 di->di_blocks = cpu_to_be64(1);
783 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
784 di->di_major = cpu_to_be32(MAJOR(dev));
785 di->di_minor = cpu_to_be32(MINOR(dev));
786 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
787 di->di_generation = cpu_to_be64(*generation);
791 if ((dip->i_diskflags & GFS2_DIF_INHERIT_JDATA) ||
792 gfs2_tune_get(sdp, gt_new_files_jdata))
793 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
794 } else if (S_ISDIR(mode)) {
795 di->di_flags |= cpu_to_be32(dip->i_diskflags &
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);
827 if (!gfs2_alloc_get(dip))
830 error = gfs2_quota_lock(dip, uid, gid);
834 error = gfs2_quota_check(dip, uid, gid);
838 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
842 init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
843 gfs2_quota_change(dip, +1, uid, gid);
847 gfs2_quota_unlock(dip);
853 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
854 struct gfs2_inode *ip)
856 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
857 struct gfs2_alloc *al;
859 struct buffer_head *dibh;
862 al = gfs2_alloc_get(dip);
866 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
870 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
871 if (alloc_required < 0)
872 goto fail_quota_locks;
873 if (alloc_required) {
874 error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
876 goto fail_quota_locks;
878 al->al_requested = sdp->sd_max_dirres;
880 error = gfs2_inplace_reserve(dip);
882 goto fail_quota_locks;
884 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
885 al->al_rgd->rd_length +
887 RES_STATFS + RES_QUOTA, 0);
891 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
893 goto fail_quota_locks;
896 error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
900 error = gfs2_meta_inode_buffer(ip, &dibh);
903 ip->i_inode.i_nlink = 1;
904 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
905 gfs2_dinode_out(ip, dibh->b_data);
913 if (dip->i_alloc->al_rgd)
914 gfs2_inplace_release(dip);
917 gfs2_quota_unlock(dip);
924 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
930 struct gfs2_ea_request er;
932 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
933 &name, &value, &len);
936 if (err == -EOPNOTSUPP)
941 memset(&er, 0, sizeof(struct gfs2_ea_request));
943 er.er_type = GFS2_EATYPE_SECURITY;
946 er.er_name_len = strlen(name);
947 er.er_data_len = len;
949 err = gfs2_ea_set_i(ip, &er);
958 * gfs2_createi - Create a new inode
959 * @ghs: An array of two holders
960 * @name: The name of the new file
961 * @mode: the permissions on the new inode
963 * @ghs[0] is an initialized holder for the directory
964 * @ghs[1] is the holder for the inode lock
966 * If the return value is not NULL, the glocks on both the directory and the new
967 * file are held. A transaction has been started and an inplace reservation
973 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
974 unsigned int mode, dev_t dev)
976 struct inode *inode = NULL;
977 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
978 struct inode *dir = &dip->i_inode;
979 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
980 struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
983 struct buffer_head *bh = NULL;
985 if (!name->len || name->len > GFS2_FNAMESIZE)
986 return ERR_PTR(-ENAMETOOLONG);
988 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
989 error = gfs2_glock_nq(ghs);
993 error = create_ok(dip, name, mode);
997 error = pick_formal_ino(sdp, &inum.no_formal_ino);
1001 error = alloc_dinode(dip, &inum.no_addr, &generation);
1005 error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
1006 LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
1010 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
1014 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode),
1016 inum.no_formal_ino, 0);
1020 error = gfs2_inode_refresh(GFS2_I(inode));
1024 error = gfs2_acl_create(dip, GFS2_I(inode));
1028 error = gfs2_security_init(dip, GFS2_I(inode));
1032 error = link_dinode(dip, name, GFS2_I(inode));
1041 gfs2_glock_dq_uninit(ghs + 1);
1042 if (inode && !IS_ERR(inode))
1049 return ERR_PTR(error);
1053 * gfs2_rmdiri - Remove a directory
1054 * @dip: The parent directory of the directory to be removed
1055 * @name: The name of the directory to be removed
1056 * @ip: The GFS2 inode of the directory to be removed
1058 * Assumes Glocks on dip and ip are held
1063 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
1064 struct gfs2_inode *ip)
1066 struct qstr dotname;
1069 if (ip->i_entries != 2) {
1070 if (gfs2_consist_inode(ip))
1071 gfs2_dinode_print(ip);
1075 error = gfs2_dir_del(dip, name);
1079 error = gfs2_change_nlink(dip, -1);
1083 gfs2_str2qstr(&dotname, ".");
1084 error = gfs2_dir_del(ip, &dotname);
1088 gfs2_str2qstr(&dotname, "..");
1089 error = gfs2_dir_del(ip, &dotname);
1093 /* It looks odd, but it really should be done twice */
1094 error = gfs2_change_nlink(ip, -1);
1098 error = gfs2_change_nlink(ip, -1);
1106 * gfs2_unlink_ok - check to see that a inode is still in a directory
1107 * @dip: the directory
1108 * @name: the name of the file
1111 * Assumes that the lock on (at least) @dip is held.
1113 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1116 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1117 const struct gfs2_inode *ip)
1121 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1124 if ((dip->i_inode.i_mode & S_ISVTX) &&
1125 dip->i_inode.i_uid != current_fsuid() &&
1126 ip->i_inode.i_uid != current_fsuid() && !capable(CAP_FOWNER))
1129 if (IS_APPEND(&dip->i_inode))
1132 error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
1136 error = gfs2_dir_check(&dip->i_inode, name, ip);
1144 * gfs2_readlinki - return the contents of a symlink
1145 * @ip: the symlink's inode
1146 * @buf: a pointer to the buffer to be filled
1147 * @len: a pointer to the length of @buf
1149 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1150 * to be freed by the caller.
1155 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1157 struct gfs2_holder i_gh;
1158 struct buffer_head *dibh;
1162 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &i_gh);
1163 error = gfs2_glock_nq(&i_gh);
1165 gfs2_holder_uninit(&i_gh);
1169 if (!ip->i_disksize) {
1170 gfs2_consist_inode(ip);
1175 error = gfs2_meta_inode_buffer(ip, &dibh);
1179 x = ip->i_disksize + 1;
1181 *buf = kmalloc(x, GFP_NOFS);
1188 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1194 gfs2_glock_dq_uninit(&i_gh);
1199 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1201 struct buffer_head *dibh;
1204 error = gfs2_meta_inode_buffer(ip, &dibh);
1206 error = inode_setattr(&ip->i_inode, attr);
1207 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1208 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1209 gfs2_dinode_out(ip, dibh->b_data);
1216 * gfs2_setattr_simple -
1220 * Called with a reference on the vnode.
1225 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1229 if (current->journal_info)
1230 return __gfs2_setattr_simple(ip, attr);
1232 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1236 error = __gfs2_setattr_simple(ip, attr);
1237 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1241 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1243 struct gfs2_dinode *str = buf;
1245 str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1246 str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1247 str->di_header.__pad0 = 0;
1248 str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1249 str->di_header.__pad1 = 0;
1250 str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1251 str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1252 str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1253 str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1254 str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1255 str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1256 str->di_size = cpu_to_be64(ip->i_disksize);
1257 str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
1258 str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1259 str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1260 str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1262 str->di_goal_meta = cpu_to_be64(ip->i_goal);
1263 str->di_goal_data = cpu_to_be64(ip->i_goal);
1264 str->di_generation = cpu_to_be64(ip->i_generation);
1266 str->di_flags = cpu_to_be32(ip->i_diskflags);
1267 str->di_height = cpu_to_be16(ip->i_height);
1268 str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1269 !(ip->i_diskflags & GFS2_DIF_EXHASH) ?
1270 GFS2_FORMAT_DE : 0);
1271 str->di_depth = cpu_to_be16(ip->i_depth);
1272 str->di_entries = cpu_to_be32(ip->i_entries);
1274 str->di_eattr = cpu_to_be64(ip->i_eattr);
1275 str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1276 str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1277 str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1280 void gfs2_dinode_print(const struct gfs2_inode *ip)
1282 printk(KERN_INFO " no_formal_ino = %llu\n",
1283 (unsigned long long)ip->i_no_formal_ino);
1284 printk(KERN_INFO " no_addr = %llu\n",
1285 (unsigned long long)ip->i_no_addr);
1286 printk(KERN_INFO " i_disksize = %llu\n",
1287 (unsigned long long)ip->i_disksize);
1288 printk(KERN_INFO " blocks = %llu\n",
1289 (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
1290 printk(KERN_INFO " i_goal = %llu\n",
1291 (unsigned long long)ip->i_goal);
1292 printk(KERN_INFO " i_diskflags = 0x%.8X\n", ip->i_diskflags);
1293 printk(KERN_INFO " i_height = %u\n", ip->i_height);
1294 printk(KERN_INFO " i_depth = %u\n", ip->i_depth);
1295 printk(KERN_INFO " i_entries = %u\n", ip->i_entries);
1296 printk(KERN_INFO " i_eattr = %llu\n",
1297 (unsigned long long)ip->i_eattr);