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 struct gfs2_sbd *sdp = GFS2_SB(inode);
135 umode_t mode = inode->i_mode;
138 inode->i_op = &gfs2_file_iops;
139 if (sdp->sd_args.ar_localflocks)
140 inode->i_fop = &gfs2_file_fops_nolock;
142 inode->i_fop = &gfs2_file_fops;
143 } else if (S_ISDIR(mode)) {
144 inode->i_op = &gfs2_dir_iops;
145 if (sdp->sd_args.ar_localflocks)
146 inode->i_fop = &gfs2_dir_fops_nolock;
148 inode->i_fop = &gfs2_dir_fops;
149 } else if (S_ISLNK(mode)) {
150 inode->i_op = &gfs2_symlink_iops;
152 inode->i_op = &gfs2_dev_iops;
155 unlock_new_inode(inode);
159 * gfs2_inode_lookup - Lookup an inode
160 * @sb: The super block
161 * @no_addr: The inode number
162 * @type: The type of the inode
163 * @skip_freeing: set this not return an inode if it is currently being freed.
165 * Returns: A VFS inode, or an error
168 struct inode *gfs2_inode_lookup(struct super_block *sb,
171 u64 no_formal_ino, int skip_freeing)
174 struct gfs2_inode *ip;
175 struct gfs2_glock *io_gl;
179 inode = gfs2_iget_skip(sb, no_addr);
181 inode = gfs2_iget(sb, no_addr);
185 return ERR_PTR(-ENOBUFS);
187 if (inode->i_state & I_NEW) {
188 struct gfs2_sbd *sdp = GFS2_SB(inode);
189 inode->i_private = ip;
190 ip->i_no_formal_ino = no_formal_ino;
192 error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
195 ip->i_gl->gl_object = ip;
197 error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
201 set_bit(GIF_INVALID, &ip->i_flags);
202 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
205 ip->i_iopen_gh.gh_gl->gl_object = ip;
207 gfs2_glock_put(io_gl);
209 if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
212 inode->i_mode = DT2IF(type);
215 * We must read the inode in order to work out its type in
216 * this case. Note that this doesn't happen often as we normally
217 * know the type beforehand. This code path only occurs during
218 * unlinked inode recovery (where it is safe to do this glock,
219 * which is not true in the general case).
221 if (type == DT_UNKNOWN) {
222 struct gfs2_holder gh;
223 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
226 /* Inode is now uptodate */
227 gfs2_glock_dq_uninit(&gh);
236 gfs2_glock_dq(&ip->i_iopen_gh);
238 gfs2_glock_put(io_gl);
240 ip->i_gl->gl_object = NULL;
241 gfs2_glock_put(ip->i_gl);
244 return ERR_PTR(error);
247 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
249 struct gfs2_dinode_host *di = &ip->i_di;
250 const struct gfs2_dinode *str = buf;
253 if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
255 ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
256 ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
257 ip->i_inode.i_rdev = 0;
258 switch (ip->i_inode.i_mode & S_IFMT) {
261 ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
262 be32_to_cpu(str->di_minor));
266 ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
267 ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
269 * We will need to review setting the nlink count here in the
270 * light of the forthcoming ro bind mount work. This is a reminder
273 ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
274 di->di_size = be64_to_cpu(str->di_size);
275 i_size_write(&ip->i_inode, di->di_size);
276 di->di_blocks = be64_to_cpu(str->di_blocks);
277 gfs2_set_inode_blocks(&ip->i_inode);
278 ip->i_inode.i_atime.tv_sec = be64_to_cpu(str->di_atime);
279 ip->i_inode.i_atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
280 ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
281 ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
282 ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
283 ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
285 di->di_goal_meta = be64_to_cpu(str->di_goal_meta);
286 di->di_goal_data = be64_to_cpu(str->di_goal_data);
287 di->di_generation = be64_to_cpu(str->di_generation);
289 di->di_flags = be32_to_cpu(str->di_flags);
290 gfs2_set_inode_flags(&ip->i_inode);
291 height = be16_to_cpu(str->di_height);
292 if (unlikely(height > GFS2_MAX_META_HEIGHT))
294 ip->i_height = (u8)height;
296 di->di_depth = be16_to_cpu(str->di_depth);
297 di->di_entries = be32_to_cpu(str->di_entries);
299 di->di_eattr = be64_to_cpu(str->di_eattr);
300 if (S_ISREG(ip->i_inode.i_mode))
301 gfs2_set_aops(&ip->i_inode);
305 if (gfs2_consist_inode(ip))
306 gfs2_dinode_print(ip);
311 * gfs2_inode_refresh - Refresh the incore copy of the dinode
312 * @ip: The GFS2 inode
317 int gfs2_inode_refresh(struct gfs2_inode *ip)
319 struct buffer_head *dibh;
322 error = gfs2_meta_inode_buffer(ip, &dibh);
326 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
331 error = gfs2_dinode_in(ip, dibh->b_data);
333 clear_bit(GIF_INVALID, &ip->i_flags);
338 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
340 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
341 struct gfs2_alloc *al;
342 struct gfs2_rgrpd *rgd;
345 if (ip->i_di.di_blocks != 1) {
346 if (gfs2_consist_inode(ip))
347 gfs2_dinode_print(ip);
351 al = gfs2_alloc_get(ip);
353 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
357 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
361 rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
363 gfs2_consist_inode(ip);
365 goto out_rindex_relse;
368 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
371 goto out_rindex_relse;
373 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
377 set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
378 set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
380 gfs2_free_di(rgd, ip);
383 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
386 gfs2_glock_dq_uninit(&al->al_rgd_gh);
388 gfs2_glock_dq_uninit(&al->al_ri_gh);
390 gfs2_quota_unhold(ip);
397 * gfs2_change_nlink - Change nlink count on inode
398 * @ip: The GFS2 inode
399 * @diff: The change in the nlink count required
403 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
405 struct buffer_head *dibh;
409 BUG_ON(diff != 1 && diff != -1);
410 nlink = ip->i_inode.i_nlink + diff;
412 /* If we are reducing the nlink count, but the new value ends up being
413 bigger than the old one, we must have underflowed. */
414 if (diff < 0 && nlink > ip->i_inode.i_nlink) {
415 if (gfs2_consist_inode(ip))
416 gfs2_dinode_print(ip);
420 error = gfs2_meta_inode_buffer(ip, &dibh);
425 inc_nlink(&ip->i_inode);
427 drop_nlink(&ip->i_inode);
429 ip->i_inode.i_ctime = CURRENT_TIME;
431 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
432 gfs2_dinode_out(ip, dibh->b_data);
434 mark_inode_dirty(&ip->i_inode);
436 if (ip->i_inode.i_nlink == 0)
437 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
442 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
446 gfs2_str2qstr(&qstr, name);
447 inode = gfs2_lookupi(dip, &qstr, 1, NULL);
448 /* gfs2_lookupi has inconsistent callers: vfs
449 * related routines expect NULL for no entry found,
450 * gfs2_lookup_simple callers expect ENOENT
451 * and do not check for NULL.
454 return ERR_PTR(-ENOENT);
461 * gfs2_lookupi - Look up a filename in a directory and return its inode
462 * @d_gh: An initialized holder for the directory glock
463 * @name: The name of the inode to look for
464 * @is_root: If 1, ignore the caller's permissions
465 * @i_gh: An uninitialized holder for the new inode glock
467 * This can be called via the VFS filldir function when NFS is doing
468 * a readdirplus and the inode which its intending to stat isn't
469 * already in cache. In this case we must not take the directory glock
470 * again, since the readdir call will have already taken that lock.
475 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
476 int is_root, struct nameidata *nd)
478 struct super_block *sb = dir->i_sb;
479 struct gfs2_inode *dip = GFS2_I(dir);
480 struct gfs2_holder d_gh;
482 struct inode *inode = NULL;
485 if (!name->len || name->len > GFS2_FNAMESIZE)
486 return ERR_PTR(-ENAMETOOLONG);
488 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
489 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
490 dir == sb->s_root->d_inode)) {
495 if (gfs2_glock_is_locked_by_me(dip->i_gl) == 0) {
496 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
498 return ERR_PTR(error);
503 error = permission(dir, MAY_EXEC, NULL);
508 inode = gfs2_dir_search(dir, name);
510 error = PTR_ERR(inode);
513 gfs2_glock_dq_uninit(&d_gh);
514 if (error == -ENOENT)
516 return inode ? inode : ERR_PTR(error);
519 static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf)
521 const struct gfs2_inum_range *str = buf;
523 ir->ir_start = be64_to_cpu(str->ir_start);
524 ir->ir_length = be64_to_cpu(str->ir_length);
527 static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf)
529 struct gfs2_inum_range *str = buf;
531 str->ir_start = cpu_to_be64(ir->ir_start);
532 str->ir_length = cpu_to_be64(ir->ir_length);
535 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
537 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
538 struct buffer_head *bh;
539 struct gfs2_inum_range_host ir;
542 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
545 mutex_lock(&sdp->sd_inum_mutex);
547 error = gfs2_meta_inode_buffer(ip, &bh);
549 mutex_unlock(&sdp->sd_inum_mutex);
554 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
557 *formal_ino = ir.ir_start++;
559 gfs2_trans_add_bh(ip->i_gl, bh, 1);
560 gfs2_inum_range_out(&ir,
561 bh->b_data + sizeof(struct gfs2_dinode));
563 mutex_unlock(&sdp->sd_inum_mutex);
570 mutex_unlock(&sdp->sd_inum_mutex);
576 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
578 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
579 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
580 struct gfs2_holder gh;
581 struct buffer_head *bh;
582 struct gfs2_inum_range_host ir;
585 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
589 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
592 mutex_lock(&sdp->sd_inum_mutex);
594 error = gfs2_meta_inode_buffer(ip, &bh);
598 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
601 struct buffer_head *m_bh;
605 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
609 z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
610 x = y = be64_to_cpu(z);
612 ir.ir_length = GFS2_INUM_QUANTUM;
613 x += GFS2_INUM_QUANTUM;
615 gfs2_consist_inode(m_ip);
617 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
618 *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z;
623 *formal_ino = ir.ir_start++;
626 gfs2_trans_add_bh(ip->i_gl, bh, 1);
627 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
632 mutex_unlock(&sdp->sd_inum_mutex);
635 gfs2_glock_dq_uninit(&gh);
639 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
643 error = pick_formal_ino_1(sdp, inum);
647 error = pick_formal_ino_2(sdp, inum);
653 * create_ok - OK to create a new on-disk inode here?
654 * @dip: Directory in which dinode is to be created
655 * @name: Name of new dinode
661 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
666 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
670 /* Don't create entries in an unlinked directory */
671 if (!dip->i_inode.i_nlink)
674 error = gfs2_dir_check(&dip->i_inode, name, NULL);
685 if (dip->i_di.di_entries == (u32)-1)
687 if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
693 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
694 unsigned int *uid, unsigned int *gid)
696 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
697 (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
700 else if (dip->i_inode.i_uid != current->fsuid)
702 *uid = dip->i_inode.i_uid;
704 *uid = current->fsuid;
706 if (dip->i_inode.i_mode & S_ISGID) {
709 *gid = dip->i_inode.i_gid;
711 *gid = current->fsgid;
714 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
716 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
719 if (gfs2_alloc_get(dip) == NULL)
722 dip->i_alloc->al_requested = RES_DINODE;
723 error = gfs2_inplace_reserve(dip);
727 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
731 *no_addr = gfs2_alloc_di(dip, generation);
736 gfs2_inplace_release(dip);
743 * init_dinode - Fill in a new dinode structure
744 * @dip: the directory this inode is being created in
745 * @gl: The glock covering the new inode
746 * @inum: the inode number
747 * @mode: the file permissions
753 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
754 const struct gfs2_inum_host *inum, unsigned int mode,
755 unsigned int uid, unsigned int gid,
756 const u64 *generation, dev_t dev, struct buffer_head **bhp)
758 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
759 struct gfs2_dinode *di;
760 struct buffer_head *dibh;
761 struct timespec tv = CURRENT_TIME;
763 dibh = gfs2_meta_new(gl, inum->no_addr);
764 gfs2_trans_add_bh(gl, dibh, 1);
765 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
766 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
767 di = (struct gfs2_dinode *)dibh->b_data;
769 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
770 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
771 di->di_mode = cpu_to_be32(mode);
772 di->di_uid = cpu_to_be32(uid);
773 di->di_gid = cpu_to_be32(gid);
776 di->di_blocks = cpu_to_be64(1);
777 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
778 di->di_major = cpu_to_be32(MAJOR(dev));
779 di->di_minor = cpu_to_be32(MINOR(dev));
780 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
781 di->di_generation = cpu_to_be64(*generation);
785 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
786 gfs2_tune_get(sdp, gt_new_files_jdata))
787 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
788 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) ||
789 gfs2_tune_get(sdp, gt_new_files_directio))
790 di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO);
791 } else if (S_ISDIR(mode)) {
792 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
793 GFS2_DIF_INHERIT_DIRECTIO);
794 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
795 GFS2_DIF_INHERIT_JDATA);
799 di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
805 memset(&di->__pad4, 0, sizeof(di->__pad4));
807 di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
808 di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
809 di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
810 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
812 set_buffer_uptodate(dibh);
817 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
818 unsigned int mode, const struct gfs2_inum_host *inum,
819 const u64 *generation, dev_t dev, struct buffer_head **bhp)
821 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
822 unsigned int uid, gid;
825 munge_mode_uid_gid(dip, &mode, &uid, &gid);
828 error = gfs2_quota_lock(dip, uid, gid);
832 error = gfs2_quota_check(dip, uid, gid);
836 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
840 init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
841 gfs2_quota_change(dip, +1, uid, gid);
845 gfs2_quota_unlock(dip);
851 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
852 struct gfs2_inode *ip)
854 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
855 struct gfs2_alloc *al;
857 struct buffer_head *dibh;
860 al = gfs2_alloc_get(dip);
862 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
866 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
867 if (alloc_required < 0)
868 goto fail_quota_locks;
869 if (alloc_required) {
870 error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
872 goto fail_quota_locks;
874 al->al_requested = sdp->sd_max_dirres;
876 error = gfs2_inplace_reserve(dip);
878 goto fail_quota_locks;
880 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
881 al->al_rgd->rd_length +
883 RES_STATFS + RES_QUOTA, 0);
887 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
889 goto fail_quota_locks;
892 error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
896 error = gfs2_meta_inode_buffer(ip, &dibh);
899 ip->i_inode.i_nlink = 1;
900 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
901 gfs2_dinode_out(ip, dibh->b_data);
909 if (dip->i_alloc->al_rgd)
910 gfs2_inplace_release(dip);
913 gfs2_quota_unlock(dip);
920 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
926 struct gfs2_ea_request er;
928 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
929 &name, &value, &len);
932 if (err == -EOPNOTSUPP)
937 memset(&er, 0, sizeof(struct gfs2_ea_request));
939 er.er_type = GFS2_EATYPE_SECURITY;
942 er.er_name_len = strlen(name);
943 er.er_data_len = len;
945 err = gfs2_ea_set_i(ip, &er);
954 * gfs2_createi - Create a new inode
955 * @ghs: An array of two holders
956 * @name: The name of the new file
957 * @mode: the permissions on the new inode
959 * @ghs[0] is an initialized holder for the directory
960 * @ghs[1] is the holder for the inode lock
962 * If the return value is not NULL, the glocks on both the directory and the new
963 * file are held. A transaction has been started and an inplace reservation
969 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
970 unsigned int mode, dev_t dev)
972 struct inode *inode = NULL;
973 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
974 struct inode *dir = &dip->i_inode;
975 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
976 struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
979 struct buffer_head *bh = NULL;
981 if (!name->len || name->len > GFS2_FNAMESIZE)
982 return ERR_PTR(-ENAMETOOLONG);
984 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
985 error = gfs2_glock_nq(ghs);
989 error = create_ok(dip, name, mode);
993 error = pick_formal_ino(sdp, &inum.no_formal_ino);
997 error = alloc_dinode(dip, &inum.no_addr, &generation);
1001 error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
1002 LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
1006 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
1010 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode),
1012 inum.no_formal_ino, 0);
1016 error = gfs2_inode_refresh(GFS2_I(inode));
1020 error = gfs2_acl_create(dip, GFS2_I(inode));
1024 error = gfs2_security_init(dip, GFS2_I(inode));
1028 error = link_dinode(dip, name, GFS2_I(inode));
1035 return ERR_PTR(-ENOMEM);
1039 gfs2_glock_dq_uninit(ghs + 1);
1047 return ERR_PTR(error);
1051 * gfs2_rmdiri - Remove a directory
1052 * @dip: The parent directory of the directory to be removed
1053 * @name: The name of the directory to be removed
1054 * @ip: The GFS2 inode of the directory to be removed
1056 * Assumes Glocks on dip and ip are held
1061 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
1062 struct gfs2_inode *ip)
1064 struct qstr dotname;
1067 if (ip->i_di.di_entries != 2) {
1068 if (gfs2_consist_inode(ip))
1069 gfs2_dinode_print(ip);
1073 error = gfs2_dir_del(dip, name);
1077 error = gfs2_change_nlink(dip, -1);
1081 gfs2_str2qstr(&dotname, ".");
1082 error = gfs2_dir_del(ip, &dotname);
1086 gfs2_str2qstr(&dotname, "..");
1087 error = gfs2_dir_del(ip, &dotname);
1091 /* It looks odd, but it really should be done twice */
1092 error = gfs2_change_nlink(ip, -1);
1096 error = gfs2_change_nlink(ip, -1);
1104 * gfs2_unlink_ok - check to see that a inode is still in a directory
1105 * @dip: the directory
1106 * @name: the name of the file
1109 * Assumes that the lock on (at least) @dip is held.
1111 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1114 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1115 const struct gfs2_inode *ip)
1119 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1122 if ((dip->i_inode.i_mode & S_ISVTX) &&
1123 dip->i_inode.i_uid != current->fsuid &&
1124 ip->i_inode.i_uid != current->fsuid && !capable(CAP_FOWNER))
1127 if (IS_APPEND(&dip->i_inode))
1130 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
1134 error = gfs2_dir_check(&dip->i_inode, name, ip);
1142 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1146 * Follow @to back to the root and make sure we don't encounter @this
1147 * Assumes we already hold the rename lock.
1152 int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to)
1154 struct inode *dir = &to->i_inode;
1155 struct super_block *sb = dir->i_sb;
1160 gfs2_str2qstr(&dotdot, "..");
1165 if (dir == &this->i_inode) {
1169 if (dir == sb->s_root->d_inode) {
1174 tmp = gfs2_lookupi(dir, &dotdot, 1, NULL);
1176 error = PTR_ERR(tmp);
1190 * gfs2_readlinki - return the contents of a symlink
1191 * @ip: the symlink's inode
1192 * @buf: a pointer to the buffer to be filled
1193 * @len: a pointer to the length of @buf
1195 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1196 * to be freed by the caller.
1201 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1203 struct gfs2_holder i_gh;
1204 struct buffer_head *dibh;
1208 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
1209 error = gfs2_glock_nq_atime(&i_gh);
1211 gfs2_holder_uninit(&i_gh);
1215 if (!ip->i_di.di_size) {
1216 gfs2_consist_inode(ip);
1221 error = gfs2_meta_inode_buffer(ip, &dibh);
1225 x = ip->i_di.di_size + 1;
1227 *buf = kmalloc(x, GFP_KERNEL);
1234 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1240 gfs2_glock_dq_uninit(&i_gh);
1245 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1246 * conditionally update the inode's atime
1247 * @gh: the holder to acquire
1249 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1250 * Update if the difference between the current time and the inode's current
1251 * atime is greater than an interval specified at mount.
1256 int gfs2_glock_nq_atime(struct gfs2_holder *gh)
1258 struct gfs2_glock *gl = gh->gh_gl;
1259 struct gfs2_sbd *sdp = gl->gl_sbd;
1260 struct gfs2_inode *ip = gl->gl_object;
1261 s64 quantum = gfs2_tune_get(sdp, gt_atime_quantum);
1265 struct timespec tv = CURRENT_TIME;
1267 if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) ||
1268 gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) ||
1269 gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops))
1272 state = gh->gh_state;
1273 flags = gh->gh_flags;
1275 error = gfs2_glock_nq(gh);
1279 if (test_bit(SDF_NOATIME, &sdp->sd_flags) ||
1280 (sdp->sd_vfs->s_flags & MS_RDONLY))
1283 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1285 gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY,
1287 error = gfs2_glock_nq(gh);
1291 /* Verify that atime hasn't been updated while we were
1292 trying to get exclusive lock. */
1295 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1296 struct buffer_head *dibh;
1297 struct gfs2_dinode *di;
1299 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
1300 if (error == -EROFS)
1305 error = gfs2_meta_inode_buffer(ip, &dibh);
1307 goto fail_end_trans;
1309 ip->i_inode.i_atime = tv;
1311 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1312 di = (struct gfs2_dinode *)dibh->b_data;
1313 di->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1314 di->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1317 gfs2_trans_end(sdp);
1320 /* If someone else has asked for the glock,
1321 unlock and let them have it. Then reacquire
1322 in the original state. */
1323 if (gfs2_glock_is_blocking(gl)) {
1325 gfs2_holder_reinit(state, flags, gh);
1326 return gfs2_glock_nq(gh);
1333 gfs2_trans_end(sdp);
1340 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1342 struct buffer_head *dibh;
1345 error = gfs2_meta_inode_buffer(ip, &dibh);
1347 error = inode_setattr(&ip->i_inode, attr);
1348 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1349 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1350 gfs2_dinode_out(ip, dibh->b_data);
1357 * gfs2_setattr_simple -
1361 * Called with a reference on the vnode.
1366 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1370 if (current->journal_info)
1371 return __gfs2_setattr_simple(ip, attr);
1373 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1377 error = __gfs2_setattr_simple(ip, attr);
1378 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1382 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1384 const struct gfs2_dinode_host *di = &ip->i_di;
1385 struct gfs2_dinode *str = buf;
1387 str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1388 str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1389 str->di_header.__pad0 = 0;
1390 str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1391 str->di_header.__pad1 = 0;
1392 str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1393 str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1394 str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1395 str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1396 str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1397 str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1398 str->di_size = cpu_to_be64(di->di_size);
1399 str->di_blocks = cpu_to_be64(di->di_blocks);
1400 str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1401 str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1402 str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1404 str->di_goal_meta = cpu_to_be64(di->di_goal_meta);
1405 str->di_goal_data = cpu_to_be64(di->di_goal_data);
1406 str->di_generation = cpu_to_be64(di->di_generation);
1408 str->di_flags = cpu_to_be32(di->di_flags);
1409 str->di_height = cpu_to_be16(ip->i_height);
1410 str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1411 !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
1412 GFS2_FORMAT_DE : 0);
1413 str->di_depth = cpu_to_be16(di->di_depth);
1414 str->di_entries = cpu_to_be32(di->di_entries);
1416 str->di_eattr = cpu_to_be64(di->di_eattr);
1417 str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1418 str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1419 str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1422 void gfs2_dinode_print(const struct gfs2_inode *ip)
1424 const struct gfs2_dinode_host *di = &ip->i_di;
1426 printk(KERN_INFO " no_formal_ino = %llu\n",
1427 (unsigned long long)ip->i_no_formal_ino);
1428 printk(KERN_INFO " no_addr = %llu\n",
1429 (unsigned long long)ip->i_no_addr);
1430 printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size);
1431 printk(KERN_INFO " di_blocks = %llu\n",
1432 (unsigned long long)di->di_blocks);
1433 printk(KERN_INFO " di_goal_meta = %llu\n",
1434 (unsigned long long)di->di_goal_meta);
1435 printk(KERN_INFO " di_goal_data = %llu\n",
1436 (unsigned long long)di->di_goal_data);
1437 printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags);
1438 printk(KERN_INFO " di_depth = %u\n", di->di_depth);
1439 printk(KERN_INFO " di_entries = %u\n", di->di_entries);
1440 printk(KERN_INFO " di_eattr = %llu\n",
1441 (unsigned long long)di->di_eattr);