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"
34 #include "ops_inode.h"
40 struct gfs2_inum_range_host {
45 static int iget_test(struct inode *inode, void *opaque)
47 struct gfs2_inode *ip = GFS2_I(inode);
48 u64 *no_addr = opaque;
50 if (ip->i_no_addr == *no_addr &&
51 inode->i_private != NULL)
57 static int iget_set(struct inode *inode, void *opaque)
59 struct gfs2_inode *ip = GFS2_I(inode);
60 u64 *no_addr = opaque;
62 inode->i_ino = (unsigned long)*no_addr;
63 ip->i_no_addr = *no_addr;
67 struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr)
69 unsigned long hash = (unsigned long)no_addr;
70 return ilookup5(sb, hash, iget_test, &no_addr);
73 static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr)
75 unsigned long hash = (unsigned long)no_addr;
76 return iget5_locked(sb, hash, iget_test, iget_set, &no_addr);
79 struct gfs2_skip_data {
84 static int iget_skip_test(struct inode *inode, void *opaque)
86 struct gfs2_inode *ip = GFS2_I(inode);
87 struct gfs2_skip_data *data = opaque;
89 if (ip->i_no_addr == data->no_addr && inode->i_private != NULL){
90 if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)){
99 static int iget_skip_set(struct inode *inode, void *opaque)
101 struct gfs2_inode *ip = GFS2_I(inode);
102 struct gfs2_skip_data *data = opaque;
106 inode->i_ino = (unsigned long)(data->no_addr);
107 ip->i_no_addr = data->no_addr;
111 static struct inode *gfs2_iget_skip(struct super_block *sb,
114 struct gfs2_skip_data data;
115 unsigned long hash = (unsigned long)no_addr;
117 data.no_addr = no_addr;
119 return iget5_locked(sb, hash, iget_skip_test, iget_skip_set, &data);
123 * GFS2 lookup code fills in vfs inode contents based on info obtained
124 * from directory entry inside gfs2_inode_lookup(). This has caused issues
125 * with NFS code path since its get_dentry routine doesn't have the relevant
126 * directory entry when gfs2_inode_lookup() is invoked. Part of the code
127 * segment inside gfs2_inode_lookup code needs to get moved around.
129 * Clean up I_LOCK and I_NEW as well.
132 void gfs2_set_iop(struct inode *inode)
134 umode_t mode = inode->i_mode;
137 inode->i_op = &gfs2_file_iops;
138 inode->i_fop = &gfs2_file_fops;
139 } else if (S_ISDIR(mode)) {
140 inode->i_op = &gfs2_dir_iops;
141 inode->i_fop = &gfs2_dir_fops;
142 } else if (S_ISLNK(mode)) {
143 inode->i_op = &gfs2_symlink_iops;
145 inode->i_op = &gfs2_dev_iops;
148 unlock_new_inode(inode);
152 * gfs2_inode_lookup - Lookup an inode
153 * @sb: The super block
154 * @no_addr: The inode number
155 * @type: The type of the inode
156 * @skip_freeing: set this not return an inode if it is currently being freed.
158 * Returns: A VFS inode, or an error
161 struct inode *gfs2_inode_lookup(struct super_block *sb,
164 u64 no_formal_ino, int skip_freeing)
167 struct gfs2_inode *ip;
168 struct gfs2_glock *io_gl;
172 inode = gfs2_iget_skip(sb, no_addr);
174 inode = gfs2_iget(sb, no_addr);
178 return ERR_PTR(-ENOBUFS);
180 if (inode->i_state & I_NEW) {
181 struct gfs2_sbd *sdp = GFS2_SB(inode);
182 inode->i_private = ip;
183 ip->i_no_formal_ino = no_formal_ino;
185 error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
188 ip->i_gl->gl_object = ip;
190 error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
194 set_bit(GIF_INVALID, &ip->i_flags);
195 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
198 ip->i_iopen_gh.gh_gl->gl_object = ip;
200 gfs2_glock_put(io_gl);
202 if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
205 inode->i_mode = DT2IF(type);
208 * We must read the inode in order to work out its type in
209 * this case. Note that this doesn't happen often as we normally
210 * know the type beforehand. This code path only occurs during
211 * unlinked inode recovery (where it is safe to do this glock,
212 * which is not true in the general case).
214 if (type == DT_UNKNOWN) {
215 struct gfs2_holder gh;
216 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
219 /* Inode is now uptodate */
220 gfs2_glock_dq_uninit(&gh);
229 gfs2_glock_dq(&ip->i_iopen_gh);
231 gfs2_glock_put(io_gl);
233 ip->i_gl->gl_object = NULL;
234 gfs2_glock_put(ip->i_gl);
237 return ERR_PTR(error);
240 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
242 struct gfs2_dinode_host *di = &ip->i_di;
243 const struct gfs2_dinode *str = buf;
245 if (ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)) {
246 if (gfs2_consist_inode(ip))
247 gfs2_dinode_print(ip);
250 ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
251 ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
252 ip->i_inode.i_rdev = 0;
253 switch (ip->i_inode.i_mode & S_IFMT) {
256 ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
257 be32_to_cpu(str->di_minor));
261 ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
262 ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
264 * We will need to review setting the nlink count here in the
265 * light of the forthcoming ro bind mount work. This is a reminder
268 ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
269 di->di_size = be64_to_cpu(str->di_size);
270 i_size_write(&ip->i_inode, di->di_size);
271 di->di_blocks = be64_to_cpu(str->di_blocks);
272 gfs2_set_inode_blocks(&ip->i_inode);
273 ip->i_inode.i_atime.tv_sec = be64_to_cpu(str->di_atime);
274 ip->i_inode.i_atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
275 ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
276 ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
277 ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
278 ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
280 di->di_goal_meta = be64_to_cpu(str->di_goal_meta);
281 di->di_goal_data = be64_to_cpu(str->di_goal_data);
282 di->di_generation = be64_to_cpu(str->di_generation);
284 di->di_flags = be32_to_cpu(str->di_flags);
285 gfs2_set_inode_flags(&ip->i_inode);
286 di->di_height = be16_to_cpu(str->di_height);
288 di->di_depth = be16_to_cpu(str->di_depth);
289 di->di_entries = be32_to_cpu(str->di_entries);
291 di->di_eattr = be64_to_cpu(str->di_eattr);
292 if (S_ISREG(ip->i_inode.i_mode))
293 gfs2_set_aops(&ip->i_inode);
299 * gfs2_inode_refresh - Refresh the incore copy of the dinode
300 * @ip: The GFS2 inode
305 int gfs2_inode_refresh(struct gfs2_inode *ip)
307 struct buffer_head *dibh;
310 error = gfs2_meta_inode_buffer(ip, &dibh);
314 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
319 error = gfs2_dinode_in(ip, dibh->b_data);
321 clear_bit(GIF_INVALID, &ip->i_flags);
326 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
328 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
329 struct gfs2_alloc *al;
330 struct gfs2_rgrpd *rgd;
333 if (ip->i_di.di_blocks != 1) {
334 if (gfs2_consist_inode(ip))
335 gfs2_dinode_print(ip);
339 al = gfs2_alloc_get(ip);
341 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
345 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
349 rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
351 gfs2_consist_inode(ip);
353 goto out_rindex_relse;
356 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
359 goto out_rindex_relse;
361 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
365 set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
366 set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
368 gfs2_free_di(rgd, ip);
371 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
374 gfs2_glock_dq_uninit(&al->al_rgd_gh);
376 gfs2_glock_dq_uninit(&al->al_ri_gh);
378 gfs2_quota_unhold(ip);
385 * gfs2_change_nlink - Change nlink count on inode
386 * @ip: The GFS2 inode
387 * @diff: The change in the nlink count required
391 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
393 struct buffer_head *dibh;
397 BUG_ON(diff != 1 && diff != -1);
398 nlink = ip->i_inode.i_nlink + diff;
400 /* If we are reducing the nlink count, but the new value ends up being
401 bigger than the old one, we must have underflowed. */
402 if (diff < 0 && nlink > ip->i_inode.i_nlink) {
403 if (gfs2_consist_inode(ip))
404 gfs2_dinode_print(ip);
408 error = gfs2_meta_inode_buffer(ip, &dibh);
413 inc_nlink(&ip->i_inode);
415 drop_nlink(&ip->i_inode);
417 ip->i_inode.i_ctime = CURRENT_TIME;
419 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
420 gfs2_dinode_out(ip, dibh->b_data);
422 mark_inode_dirty(&ip->i_inode);
424 if (ip->i_inode.i_nlink == 0)
425 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
430 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
434 gfs2_str2qstr(&qstr, name);
435 inode = gfs2_lookupi(dip, &qstr, 1, NULL);
436 /* gfs2_lookupi has inconsistent callers: vfs
437 * related routines expect NULL for no entry found,
438 * gfs2_lookup_simple callers expect ENOENT
439 * and do not check for NULL.
442 return ERR_PTR(-ENOENT);
449 * gfs2_lookupi - Look up a filename in a directory and return its inode
450 * @d_gh: An initialized holder for the directory glock
451 * @name: The name of the inode to look for
452 * @is_root: If 1, ignore the caller's permissions
453 * @i_gh: An uninitialized holder for the new inode glock
455 * This can be called via the VFS filldir function when NFS is doing
456 * a readdirplus and the inode which its intending to stat isn't
457 * already in cache. In this case we must not take the directory glock
458 * again, since the readdir call will have already taken that lock.
463 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
464 int is_root, struct nameidata *nd)
466 struct super_block *sb = dir->i_sb;
467 struct gfs2_inode *dip = GFS2_I(dir);
468 struct gfs2_holder d_gh;
470 struct inode *inode = NULL;
473 if (!name->len || name->len > GFS2_FNAMESIZE)
474 return ERR_PTR(-ENAMETOOLONG);
476 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
477 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
478 dir == sb->s_root->d_inode)) {
483 if (gfs2_glock_is_locked_by_me(dip->i_gl) == 0) {
484 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
486 return ERR_PTR(error);
491 error = permission(dir, MAY_EXEC, NULL);
496 inode = gfs2_dir_search(dir, name);
498 error = PTR_ERR(inode);
501 gfs2_glock_dq_uninit(&d_gh);
502 if (error == -ENOENT)
504 return inode ? inode : ERR_PTR(error);
507 static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf)
509 const struct gfs2_inum_range *str = buf;
511 ir->ir_start = be64_to_cpu(str->ir_start);
512 ir->ir_length = be64_to_cpu(str->ir_length);
515 static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf)
517 struct gfs2_inum_range *str = buf;
519 str->ir_start = cpu_to_be64(ir->ir_start);
520 str->ir_length = cpu_to_be64(ir->ir_length);
523 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
525 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
526 struct buffer_head *bh;
527 struct gfs2_inum_range_host ir;
530 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
533 mutex_lock(&sdp->sd_inum_mutex);
535 error = gfs2_meta_inode_buffer(ip, &bh);
537 mutex_unlock(&sdp->sd_inum_mutex);
542 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
545 *formal_ino = ir.ir_start++;
547 gfs2_trans_add_bh(ip->i_gl, bh, 1);
548 gfs2_inum_range_out(&ir,
549 bh->b_data + sizeof(struct gfs2_dinode));
551 mutex_unlock(&sdp->sd_inum_mutex);
558 mutex_unlock(&sdp->sd_inum_mutex);
564 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
566 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
567 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
568 struct gfs2_holder gh;
569 struct buffer_head *bh;
570 struct gfs2_inum_range_host ir;
573 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
577 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
580 mutex_lock(&sdp->sd_inum_mutex);
582 error = gfs2_meta_inode_buffer(ip, &bh);
586 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
589 struct buffer_head *m_bh;
593 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
597 z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
598 x = y = be64_to_cpu(z);
600 ir.ir_length = GFS2_INUM_QUANTUM;
601 x += GFS2_INUM_QUANTUM;
603 gfs2_consist_inode(m_ip);
605 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
606 *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z;
611 *formal_ino = ir.ir_start++;
614 gfs2_trans_add_bh(ip->i_gl, bh, 1);
615 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
620 mutex_unlock(&sdp->sd_inum_mutex);
623 gfs2_glock_dq_uninit(&gh);
627 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
631 error = pick_formal_ino_1(sdp, inum);
635 error = pick_formal_ino_2(sdp, inum);
641 * create_ok - OK to create a new on-disk inode here?
642 * @dip: Directory in which dinode is to be created
643 * @name: Name of new dinode
649 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
654 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
658 /* Don't create entries in an unlinked directory */
659 if (!dip->i_inode.i_nlink)
662 error = gfs2_dir_check(&dip->i_inode, name, NULL);
673 if (dip->i_di.di_entries == (u32)-1)
675 if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
681 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
682 unsigned int *uid, unsigned int *gid)
684 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
685 (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
688 else if (dip->i_inode.i_uid != current->fsuid)
690 *uid = dip->i_inode.i_uid;
692 *uid = current->fsuid;
694 if (dip->i_inode.i_mode & S_ISGID) {
697 *gid = dip->i_inode.i_gid;
699 *gid = current->fsgid;
702 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
704 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
709 dip->i_alloc.al_requested = RES_DINODE;
710 error = gfs2_inplace_reserve(dip);
714 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
718 *no_addr = gfs2_alloc_di(dip, generation);
723 gfs2_inplace_release(dip);
730 * init_dinode - Fill in a new dinode structure
731 * @dip: the directory this inode is being created in
732 * @gl: The glock covering the new inode
733 * @inum: the inode number
734 * @mode: the file permissions
740 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
741 const struct gfs2_inum_host *inum, unsigned int mode,
742 unsigned int uid, unsigned int gid,
743 const u64 *generation, dev_t dev, struct buffer_head **bhp)
745 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
746 struct gfs2_dinode *di;
747 struct buffer_head *dibh;
748 struct timespec tv = CURRENT_TIME;
750 dibh = gfs2_meta_new(gl, inum->no_addr);
751 gfs2_trans_add_bh(gl, dibh, 1);
752 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
753 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
754 di = (struct gfs2_dinode *)dibh->b_data;
756 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
757 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
758 di->di_mode = cpu_to_be32(mode);
759 di->di_uid = cpu_to_be32(uid);
760 di->di_gid = cpu_to_be32(gid);
763 di->di_blocks = cpu_to_be64(1);
764 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
765 di->di_major = cpu_to_be32(MAJOR(dev));
766 di->di_minor = cpu_to_be32(MINOR(dev));
767 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
768 di->di_generation = cpu_to_be64(*generation);
772 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
773 gfs2_tune_get(sdp, gt_new_files_jdata))
774 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
775 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) ||
776 gfs2_tune_get(sdp, gt_new_files_directio))
777 di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO);
778 } else if (S_ISDIR(mode)) {
779 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
780 GFS2_DIF_INHERIT_DIRECTIO);
781 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
782 GFS2_DIF_INHERIT_JDATA);
786 di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
792 memset(&di->__pad4, 0, sizeof(di->__pad4));
794 di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
795 di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
796 di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
797 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
799 set_buffer_uptodate(dibh);
804 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
805 unsigned int mode, const struct gfs2_inum_host *inum,
806 const u64 *generation, dev_t dev, struct buffer_head **bhp)
808 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
809 unsigned int uid, gid;
812 munge_mode_uid_gid(dip, &mode, &uid, &gid);
815 error = gfs2_quota_lock(dip, uid, gid);
819 error = gfs2_quota_check(dip, uid, gid);
823 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
827 init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
828 gfs2_quota_change(dip, +1, uid, gid);
832 gfs2_quota_unlock(dip);
838 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
839 struct gfs2_inode *ip)
841 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
842 struct gfs2_alloc *al;
844 struct buffer_head *dibh;
847 al = gfs2_alloc_get(dip);
849 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
853 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
854 if (alloc_required < 0)
856 if (alloc_required) {
857 error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
859 goto fail_quota_locks;
861 al->al_requested = sdp->sd_max_dirres;
863 error = gfs2_inplace_reserve(dip);
865 goto fail_quota_locks;
867 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
868 al->al_rgd->rd_length +
870 RES_STATFS + RES_QUOTA, 0);
874 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
876 goto fail_quota_locks;
879 error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
883 error = gfs2_meta_inode_buffer(ip, &dibh);
886 ip->i_inode.i_nlink = 1;
887 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
888 gfs2_dinode_out(ip, dibh->b_data);
896 if (dip->i_alloc.al_rgd)
897 gfs2_inplace_release(dip);
900 gfs2_quota_unlock(dip);
907 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
913 struct gfs2_ea_request er;
915 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
916 &name, &value, &len);
919 if (err == -EOPNOTSUPP)
924 memset(&er, 0, sizeof(struct gfs2_ea_request));
926 er.er_type = GFS2_EATYPE_SECURITY;
929 er.er_name_len = strlen(name);
930 er.er_data_len = len;
932 err = gfs2_ea_set_i(ip, &er);
941 * gfs2_createi - Create a new inode
942 * @ghs: An array of two holders
943 * @name: The name of the new file
944 * @mode: the permissions on the new inode
946 * @ghs[0] is an initialized holder for the directory
947 * @ghs[1] is the holder for the inode lock
949 * If the return value is not NULL, the glocks on both the directory and the new
950 * file are held. A transaction has been started and an inplace reservation
956 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
957 unsigned int mode, dev_t dev)
959 struct inode *inode = NULL;
960 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
961 struct inode *dir = &dip->i_inode;
962 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
963 struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
966 struct buffer_head *bh = NULL;
968 if (!name->len || name->len > GFS2_FNAMESIZE)
969 return ERR_PTR(-ENAMETOOLONG);
971 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
972 error = gfs2_glock_nq(ghs);
976 error = create_ok(dip, name, mode);
980 error = pick_formal_ino(sdp, &inum.no_formal_ino);
984 error = alloc_dinode(dip, &inum.no_addr, &generation);
988 error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
989 LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
993 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
997 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode),
999 inum.no_formal_ino, 0);
1003 error = gfs2_inode_refresh(GFS2_I(inode));
1007 error = gfs2_acl_create(dip, GFS2_I(inode));
1011 error = gfs2_security_init(dip, GFS2_I(inode));
1015 error = link_dinode(dip, name, GFS2_I(inode));
1022 return ERR_PTR(-ENOMEM);
1026 gfs2_glock_dq_uninit(ghs + 1);
1034 return ERR_PTR(error);
1038 * gfs2_rmdiri - Remove a directory
1039 * @dip: The parent directory of the directory to be removed
1040 * @name: The name of the directory to be removed
1041 * @ip: The GFS2 inode of the directory to be removed
1043 * Assumes Glocks on dip and ip are held
1048 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
1049 struct gfs2_inode *ip)
1051 struct qstr dotname;
1054 if (ip->i_di.di_entries != 2) {
1055 if (gfs2_consist_inode(ip))
1056 gfs2_dinode_print(ip);
1060 error = gfs2_dir_del(dip, name);
1064 error = gfs2_change_nlink(dip, -1);
1068 gfs2_str2qstr(&dotname, ".");
1069 error = gfs2_dir_del(ip, &dotname);
1073 gfs2_str2qstr(&dotname, "..");
1074 error = gfs2_dir_del(ip, &dotname);
1078 /* It looks odd, but it really should be done twice */
1079 error = gfs2_change_nlink(ip, -1);
1083 error = gfs2_change_nlink(ip, -1);
1091 * gfs2_unlink_ok - check to see that a inode is still in a directory
1092 * @dip: the directory
1093 * @name: the name of the file
1096 * Assumes that the lock on (at least) @dip is held.
1098 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1101 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1102 const struct gfs2_inode *ip)
1106 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1109 if ((dip->i_inode.i_mode & S_ISVTX) &&
1110 dip->i_inode.i_uid != current->fsuid &&
1111 ip->i_inode.i_uid != current->fsuid && !capable(CAP_FOWNER))
1114 if (IS_APPEND(&dip->i_inode))
1117 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
1121 error = gfs2_dir_check(&dip->i_inode, name, ip);
1129 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1133 * Follow @to back to the root and make sure we don't encounter @this
1134 * Assumes we already hold the rename lock.
1139 int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to)
1141 struct inode *dir = &to->i_inode;
1142 struct super_block *sb = dir->i_sb;
1147 gfs2_str2qstr(&dotdot, "..");
1152 if (dir == &this->i_inode) {
1156 if (dir == sb->s_root->d_inode) {
1161 tmp = gfs2_lookupi(dir, &dotdot, 1, NULL);
1163 error = PTR_ERR(tmp);
1177 * gfs2_readlinki - return the contents of a symlink
1178 * @ip: the symlink's inode
1179 * @buf: a pointer to the buffer to be filled
1180 * @len: a pointer to the length of @buf
1182 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1183 * to be freed by the caller.
1188 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1190 struct gfs2_holder i_gh;
1191 struct buffer_head *dibh;
1195 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
1196 error = gfs2_glock_nq_atime(&i_gh);
1198 gfs2_holder_uninit(&i_gh);
1202 if (!ip->i_di.di_size) {
1203 gfs2_consist_inode(ip);
1208 error = gfs2_meta_inode_buffer(ip, &dibh);
1212 x = ip->i_di.di_size + 1;
1214 *buf = kmalloc(x, GFP_KERNEL);
1221 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1227 gfs2_glock_dq_uninit(&i_gh);
1232 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1233 * conditionally update the inode's atime
1234 * @gh: the holder to acquire
1236 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1237 * Update if the difference between the current time and the inode's current
1238 * atime is greater than an interval specified at mount.
1243 int gfs2_glock_nq_atime(struct gfs2_holder *gh)
1245 struct gfs2_glock *gl = gh->gh_gl;
1246 struct gfs2_sbd *sdp = gl->gl_sbd;
1247 struct gfs2_inode *ip = gl->gl_object;
1248 s64 quantum = gfs2_tune_get(sdp, gt_atime_quantum);
1252 struct timespec tv = CURRENT_TIME;
1254 if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) ||
1255 gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) ||
1256 gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops))
1259 state = gh->gh_state;
1260 flags = gh->gh_flags;
1262 error = gfs2_glock_nq(gh);
1266 if (test_bit(SDF_NOATIME, &sdp->sd_flags) ||
1267 (sdp->sd_vfs->s_flags & MS_RDONLY))
1270 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1272 gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY,
1274 error = gfs2_glock_nq(gh);
1278 /* Verify that atime hasn't been updated while we were
1279 trying to get exclusive lock. */
1282 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1283 struct buffer_head *dibh;
1284 struct gfs2_dinode *di;
1286 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
1287 if (error == -EROFS)
1292 error = gfs2_meta_inode_buffer(ip, &dibh);
1294 goto fail_end_trans;
1296 ip->i_inode.i_atime = tv;
1298 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1299 di = (struct gfs2_dinode *)dibh->b_data;
1300 di->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1301 di->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1304 gfs2_trans_end(sdp);
1307 /* If someone else has asked for the glock,
1308 unlock and let them have it. Then reacquire
1309 in the original state. */
1310 if (gfs2_glock_is_blocking(gl)) {
1312 gfs2_holder_reinit(state, flags, gh);
1313 return gfs2_glock_nq(gh);
1320 gfs2_trans_end(sdp);
1327 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1329 struct buffer_head *dibh;
1332 error = gfs2_meta_inode_buffer(ip, &dibh);
1334 error = inode_setattr(&ip->i_inode, attr);
1335 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1336 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1337 gfs2_dinode_out(ip, dibh->b_data);
1344 * gfs2_setattr_simple -
1348 * Called with a reference on the vnode.
1353 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1357 if (current->journal_info)
1358 return __gfs2_setattr_simple(ip, attr);
1360 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1364 error = __gfs2_setattr_simple(ip, attr);
1365 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1369 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1371 const struct gfs2_dinode_host *di = &ip->i_di;
1372 struct gfs2_dinode *str = buf;
1374 str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1375 str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1376 str->di_header.__pad0 = 0;
1377 str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1378 str->di_header.__pad1 = 0;
1379 str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1380 str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1381 str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1382 str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1383 str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1384 str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1385 str->di_size = cpu_to_be64(di->di_size);
1386 str->di_blocks = cpu_to_be64(di->di_blocks);
1387 str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1388 str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1389 str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1391 str->di_goal_meta = cpu_to_be64(di->di_goal_meta);
1392 str->di_goal_data = cpu_to_be64(di->di_goal_data);
1393 str->di_generation = cpu_to_be64(di->di_generation);
1395 str->di_flags = cpu_to_be32(di->di_flags);
1396 str->di_height = cpu_to_be16(di->di_height);
1397 str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1398 !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
1399 GFS2_FORMAT_DE : 0);
1400 str->di_depth = cpu_to_be16(di->di_depth);
1401 str->di_entries = cpu_to_be32(di->di_entries);
1403 str->di_eattr = cpu_to_be64(di->di_eattr);
1404 str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1405 str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1406 str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1409 void gfs2_dinode_print(const struct gfs2_inode *ip)
1411 const struct gfs2_dinode_host *di = &ip->i_di;
1413 printk(KERN_INFO " no_formal_ino = %llu\n",
1414 (unsigned long long)ip->i_no_formal_ino);
1415 printk(KERN_INFO " no_addr = %llu\n",
1416 (unsigned long long)ip->i_no_addr);
1417 printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size);
1418 printk(KERN_INFO " di_blocks = %llu\n",
1419 (unsigned long long)di->di_blocks);
1420 printk(KERN_INFO " di_goal_meta = %llu\n",
1421 (unsigned long long)di->di_goal_meta);
1422 printk(KERN_INFO " di_goal_data = %llu\n",
1423 (unsigned long long)di->di_goal_data);
1424 printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags);
1425 printk(KERN_INFO " di_height = %u\n", di->di_height);
1426 printk(KERN_INFO " di_depth = %u\n", di->di_depth);
1427 printk(KERN_INFO " di_entries = %u\n", di->di_entries);
1428 printk(KERN_INFO " di_eattr = %llu\n",
1429 (unsigned long long)di->di_eattr);