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>
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_file_iops;
153 init_special_inode(inode, inode->i_mode, inode->i_rdev);
156 unlock_new_inode(inode);
160 * gfs2_inode_lookup - Lookup an inode
161 * @sb: The super block
162 * @no_addr: The inode number
163 * @type: The type of the inode
164 * @skip_freeing: set this not return an inode if it is currently being freed.
166 * Returns: A VFS inode, or an error
169 struct inode *gfs2_inode_lookup(struct super_block *sb,
172 u64 no_formal_ino, int skip_freeing)
175 struct gfs2_inode *ip;
176 struct gfs2_glock *io_gl;
180 inode = gfs2_iget_skip(sb, no_addr);
182 inode = gfs2_iget(sb, no_addr);
186 return ERR_PTR(-ENOBUFS);
188 if (inode->i_state & I_NEW) {
189 struct gfs2_sbd *sdp = GFS2_SB(inode);
190 inode->i_private = ip;
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 struct gfs2_dinode_host *di = &ip->i_di;
251 const struct gfs2_dinode *str = buf;
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 di->di_size = be64_to_cpu(str->di_size);
276 i_size_write(&ip->i_inode, di->di_size);
277 gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
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 ip->i_goal = be64_to_cpu(str->di_goal_meta);
286 di->di_generation = be64_to_cpu(str->di_generation);
288 di->di_flags = be32_to_cpu(str->di_flags);
289 gfs2_set_inode_flags(&ip->i_inode);
290 height = be16_to_cpu(str->di_height);
291 if (unlikely(height > GFS2_MAX_META_HEIGHT))
293 ip->i_height = (u8)height;
295 depth = be16_to_cpu(str->di_depth);
296 if (unlikely(depth > GFS2_DIR_MAX_DEPTH))
298 ip->i_depth = (u8)depth;
299 di->di_entries = be32_to_cpu(str->di_entries);
301 di->di_eattr = be64_to_cpu(str->di_eattr);
302 if (S_ISREG(ip->i_inode.i_mode))
303 gfs2_set_aops(&ip->i_inode);
307 if (gfs2_consist_inode(ip))
308 gfs2_dinode_print(ip);
313 * gfs2_inode_refresh - Refresh the incore copy of the dinode
314 * @ip: The GFS2 inode
319 int gfs2_inode_refresh(struct gfs2_inode *ip)
321 struct buffer_head *dibh;
324 error = gfs2_meta_inode_buffer(ip, &dibh);
328 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
333 error = gfs2_dinode_in(ip, dibh->b_data);
335 clear_bit(GIF_INVALID, &ip->i_flags);
340 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
342 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
343 struct gfs2_alloc *al;
344 struct gfs2_rgrpd *rgd;
347 if (gfs2_get_inode_blocks(&ip->i_inode) != 1) {
348 if (gfs2_consist_inode(ip))
349 gfs2_dinode_print(ip);
353 al = gfs2_alloc_get(ip);
355 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
359 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
363 rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
365 gfs2_consist_inode(ip);
367 goto out_rindex_relse;
370 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
373 goto out_rindex_relse;
375 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
379 set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
380 set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
382 gfs2_free_di(rgd, ip);
385 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
388 gfs2_glock_dq_uninit(&al->al_rgd_gh);
390 gfs2_glock_dq_uninit(&al->al_ri_gh);
392 gfs2_quota_unhold(ip);
399 * gfs2_change_nlink - Change nlink count on inode
400 * @ip: The GFS2 inode
401 * @diff: The change in the nlink count required
405 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
407 struct buffer_head *dibh;
411 BUG_ON(diff != 1 && diff != -1);
412 nlink = ip->i_inode.i_nlink + diff;
414 /* If we are reducing the nlink count, but the new value ends up being
415 bigger than the old one, we must have underflowed. */
416 if (diff < 0 && nlink > ip->i_inode.i_nlink) {
417 if (gfs2_consist_inode(ip))
418 gfs2_dinode_print(ip);
422 error = gfs2_meta_inode_buffer(ip, &dibh);
427 inc_nlink(&ip->i_inode);
429 drop_nlink(&ip->i_inode);
431 ip->i_inode.i_ctime = CURRENT_TIME;
433 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
434 gfs2_dinode_out(ip, dibh->b_data);
436 mark_inode_dirty(&ip->i_inode);
438 if (ip->i_inode.i_nlink == 0)
439 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
444 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
448 gfs2_str2qstr(&qstr, name);
449 inode = gfs2_lookupi(dip, &qstr, 1, NULL);
450 /* gfs2_lookupi has inconsistent callers: vfs
451 * related routines expect NULL for no entry found,
452 * gfs2_lookup_simple callers expect ENOENT
453 * and do not check for NULL.
456 return ERR_PTR(-ENOENT);
463 * gfs2_lookupi - Look up a filename in a directory and return its inode
464 * @d_gh: An initialized holder for the directory glock
465 * @name: The name of the inode to look for
466 * @is_root: If 1, ignore the caller's permissions
467 * @i_gh: An uninitialized holder for the new inode glock
469 * This can be called via the VFS filldir function when NFS is doing
470 * a readdirplus and the inode which its intending to stat isn't
471 * already in cache. In this case we must not take the directory glock
472 * again, since the readdir call will have already taken that lock.
477 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
478 int is_root, struct nameidata *nd)
480 struct super_block *sb = dir->i_sb;
481 struct gfs2_inode *dip = GFS2_I(dir);
482 struct gfs2_holder d_gh;
484 struct inode *inode = NULL;
487 if (!name->len || name->len > GFS2_FNAMESIZE)
488 return ERR_PTR(-ENAMETOOLONG);
490 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
491 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
492 dir == sb->s_root->d_inode)) {
497 if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
498 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
500 return ERR_PTR(error);
505 error = permission(dir, MAY_EXEC, NULL);
510 inode = gfs2_dir_search(dir, name);
512 error = PTR_ERR(inode);
515 gfs2_glock_dq_uninit(&d_gh);
516 if (error == -ENOENT)
518 return inode ? inode : ERR_PTR(error);
521 static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf)
523 const struct gfs2_inum_range *str = buf;
525 ir->ir_start = be64_to_cpu(str->ir_start);
526 ir->ir_length = be64_to_cpu(str->ir_length);
529 static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf)
531 struct gfs2_inum_range *str = buf;
533 str->ir_start = cpu_to_be64(ir->ir_start);
534 str->ir_length = cpu_to_be64(ir->ir_length);
537 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
539 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
540 struct buffer_head *bh;
541 struct gfs2_inum_range_host ir;
544 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
547 mutex_lock(&sdp->sd_inum_mutex);
549 error = gfs2_meta_inode_buffer(ip, &bh);
551 mutex_unlock(&sdp->sd_inum_mutex);
556 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
559 *formal_ino = ir.ir_start++;
561 gfs2_trans_add_bh(ip->i_gl, bh, 1);
562 gfs2_inum_range_out(&ir,
563 bh->b_data + sizeof(struct gfs2_dinode));
565 mutex_unlock(&sdp->sd_inum_mutex);
572 mutex_unlock(&sdp->sd_inum_mutex);
578 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
580 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
581 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
582 struct gfs2_holder gh;
583 struct buffer_head *bh;
584 struct gfs2_inum_range_host ir;
587 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
591 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
594 mutex_lock(&sdp->sd_inum_mutex);
596 error = gfs2_meta_inode_buffer(ip, &bh);
600 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
603 struct buffer_head *m_bh;
607 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
611 z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
612 x = y = be64_to_cpu(z);
614 ir.ir_length = GFS2_INUM_QUANTUM;
615 x += GFS2_INUM_QUANTUM;
617 gfs2_consist_inode(m_ip);
619 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
620 *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z;
625 *formal_ino = ir.ir_start++;
628 gfs2_trans_add_bh(ip->i_gl, bh, 1);
629 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
634 mutex_unlock(&sdp->sd_inum_mutex);
637 gfs2_glock_dq_uninit(&gh);
641 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
645 error = pick_formal_ino_1(sdp, inum);
649 error = pick_formal_ino_2(sdp, inum);
655 * create_ok - OK to create a new on-disk inode here?
656 * @dip: Directory in which dinode is to be created
657 * @name: Name of new dinode
663 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
668 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
672 /* Don't create entries in an unlinked directory */
673 if (!dip->i_inode.i_nlink)
676 error = gfs2_dir_check(&dip->i_inode, name, NULL);
687 if (dip->i_di.di_entries == (u32)-1)
689 if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
695 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
696 unsigned int *uid, unsigned int *gid)
698 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
699 (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
702 else if (dip->i_inode.i_uid != current->fsuid)
704 *uid = dip->i_inode.i_uid;
706 *uid = current->fsuid;
708 if (dip->i_inode.i_mode & S_ISGID) {
711 *gid = dip->i_inode.i_gid;
713 *gid = current->fsgid;
716 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
718 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
721 if (gfs2_alloc_get(dip) == NULL)
724 dip->i_alloc->al_requested = RES_DINODE;
725 error = gfs2_inplace_reserve(dip);
729 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
733 *no_addr = gfs2_alloc_di(dip, generation);
738 gfs2_inplace_release(dip);
745 * init_dinode - Fill in a new dinode structure
746 * @dip: the directory this inode is being created in
747 * @gl: The glock covering the new inode
748 * @inum: the inode number
749 * @mode: the file permissions
755 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
756 const struct gfs2_inum_host *inum, unsigned int mode,
757 unsigned int uid, unsigned int gid,
758 const u64 *generation, dev_t dev, struct buffer_head **bhp)
760 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
761 struct gfs2_dinode *di;
762 struct buffer_head *dibh;
763 struct timespec tv = CURRENT_TIME;
765 dibh = gfs2_meta_new(gl, inum->no_addr);
766 gfs2_trans_add_bh(gl, dibh, 1);
767 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
768 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
769 di = (struct gfs2_dinode *)dibh->b_data;
771 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
772 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
773 di->di_mode = cpu_to_be32(mode);
774 di->di_uid = cpu_to_be32(uid);
775 di->di_gid = cpu_to_be32(gid);
778 di->di_blocks = cpu_to_be64(1);
779 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
780 di->di_major = cpu_to_be32(MAJOR(dev));
781 di->di_minor = cpu_to_be32(MINOR(dev));
782 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
783 di->di_generation = cpu_to_be64(*generation);
787 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
788 gfs2_tune_get(sdp, gt_new_files_jdata))
789 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
790 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) ||
791 gfs2_tune_get(sdp, gt_new_files_directio))
792 di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO);
793 } else if (S_ISDIR(mode)) {
794 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
795 GFS2_DIF_INHERIT_DIRECTIO);
796 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
797 GFS2_DIF_INHERIT_JDATA);
801 di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
807 memset(&di->__pad4, 0, sizeof(di->__pad4));
809 di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
810 di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
811 di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
812 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
814 set_buffer_uptodate(dibh);
819 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
820 unsigned int mode, const struct gfs2_inum_host *inum,
821 const u64 *generation, dev_t dev, struct buffer_head **bhp)
823 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
824 unsigned int uid, gid;
827 munge_mode_uid_gid(dip, &mode, &uid, &gid);
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);
864 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
868 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
869 if (alloc_required < 0)
870 goto fail_quota_locks;
871 if (alloc_required) {
872 error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
874 goto fail_quota_locks;
876 al->al_requested = sdp->sd_max_dirres;
878 error = gfs2_inplace_reserve(dip);
880 goto fail_quota_locks;
882 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
883 al->al_rgd->rd_length +
885 RES_STATFS + RES_QUOTA, 0);
889 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
891 goto fail_quota_locks;
894 error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
898 error = gfs2_meta_inode_buffer(ip, &dibh);
901 ip->i_inode.i_nlink = 1;
902 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
903 gfs2_dinode_out(ip, dibh->b_data);
911 if (dip->i_alloc->al_rgd)
912 gfs2_inplace_release(dip);
915 gfs2_quota_unlock(dip);
922 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
928 struct gfs2_ea_request er;
930 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
931 &name, &value, &len);
934 if (err == -EOPNOTSUPP)
939 memset(&er, 0, sizeof(struct gfs2_ea_request));
941 er.er_type = GFS2_EATYPE_SECURITY;
944 er.er_name_len = strlen(name);
945 er.er_data_len = len;
947 err = gfs2_ea_set_i(ip, &er);
956 * gfs2_createi - Create a new inode
957 * @ghs: An array of two holders
958 * @name: The name of the new file
959 * @mode: the permissions on the new inode
961 * @ghs[0] is an initialized holder for the directory
962 * @ghs[1] is the holder for the inode lock
964 * If the return value is not NULL, the glocks on both the directory and the new
965 * file are held. A transaction has been started and an inplace reservation
971 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
972 unsigned int mode, dev_t dev)
974 struct inode *inode = NULL;
975 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
976 struct inode *dir = &dip->i_inode;
977 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
978 struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
981 struct buffer_head *bh = NULL;
983 if (!name->len || name->len > GFS2_FNAMESIZE)
984 return ERR_PTR(-ENAMETOOLONG);
986 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
987 error = gfs2_glock_nq(ghs);
991 error = create_ok(dip, name, mode);
995 error = pick_formal_ino(sdp, &inum.no_formal_ino);
999 error = alloc_dinode(dip, &inum.no_addr, &generation);
1003 error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
1004 LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
1008 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
1012 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode),
1014 inum.no_formal_ino, 0);
1018 error = gfs2_inode_refresh(GFS2_I(inode));
1022 error = gfs2_acl_create(dip, GFS2_I(inode));
1026 error = gfs2_security_init(dip, GFS2_I(inode));
1030 error = link_dinode(dip, name, GFS2_I(inode));
1037 return ERR_PTR(-ENOMEM);
1041 gfs2_glock_dq_uninit(ghs + 1);
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_di.di_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 = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
1136 error = gfs2_dir_check(&dip->i_inode, name, ip);
1144 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1148 * Follow @to back to the root and make sure we don't encounter @this
1149 * Assumes we already hold the rename lock.
1154 int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to)
1156 struct inode *dir = &to->i_inode;
1157 struct super_block *sb = dir->i_sb;
1162 gfs2_str2qstr(&dotdot, "..");
1167 if (dir == &this->i_inode) {
1171 if (dir == sb->s_root->d_inode) {
1176 tmp = gfs2_lookupi(dir, &dotdot, 1, NULL);
1178 error = PTR_ERR(tmp);
1192 * gfs2_readlinki - return the contents of a symlink
1193 * @ip: the symlink's inode
1194 * @buf: a pointer to the buffer to be filled
1195 * @len: a pointer to the length of @buf
1197 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1198 * to be freed by the caller.
1203 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1205 struct gfs2_holder i_gh;
1206 struct buffer_head *dibh;
1210 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
1211 error = gfs2_glock_nq_atime(&i_gh);
1213 gfs2_holder_uninit(&i_gh);
1217 if (!ip->i_di.di_size) {
1218 gfs2_consist_inode(ip);
1223 error = gfs2_meta_inode_buffer(ip, &dibh);
1227 x = ip->i_di.di_size + 1;
1229 *buf = kmalloc(x, GFP_KERNEL);
1236 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1242 gfs2_glock_dq_uninit(&i_gh);
1247 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1248 * conditionally update the inode's atime
1249 * @gh: the holder to acquire
1251 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1252 * Update if the difference between the current time and the inode's current
1253 * atime is greater than an interval specified at mount.
1258 int gfs2_glock_nq_atime(struct gfs2_holder *gh)
1260 struct gfs2_glock *gl = gh->gh_gl;
1261 struct gfs2_sbd *sdp = gl->gl_sbd;
1262 struct gfs2_inode *ip = gl->gl_object;
1263 s64 quantum = gfs2_tune_get(sdp, gt_atime_quantum);
1267 struct timespec tv = CURRENT_TIME;
1269 if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) ||
1270 gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) ||
1271 gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops))
1274 state = gh->gh_state;
1275 flags = gh->gh_flags;
1277 error = gfs2_glock_nq(gh);
1281 if (test_bit(SDF_NOATIME, &sdp->sd_flags) ||
1282 (sdp->sd_vfs->s_flags & MS_RDONLY))
1285 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1287 gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY,
1289 error = gfs2_glock_nq(gh);
1293 /* Verify that atime hasn't been updated while we were
1294 trying to get exclusive lock. */
1297 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1298 struct buffer_head *dibh;
1299 struct gfs2_dinode *di;
1301 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
1302 if (error == -EROFS)
1307 error = gfs2_meta_inode_buffer(ip, &dibh);
1309 goto fail_end_trans;
1311 ip->i_inode.i_atime = tv;
1313 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1314 di = (struct gfs2_dinode *)dibh->b_data;
1315 di->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1316 di->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1319 gfs2_trans_end(sdp);
1322 /* If someone else has asked for the glock,
1323 unlock and let them have it. Then reacquire
1324 in the original state. */
1325 if (gfs2_glock_is_blocking(gl)) {
1327 gfs2_holder_reinit(state, flags, gh);
1328 return gfs2_glock_nq(gh);
1335 gfs2_trans_end(sdp);
1342 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1344 struct buffer_head *dibh;
1347 error = gfs2_meta_inode_buffer(ip, &dibh);
1349 error = inode_setattr(&ip->i_inode, attr);
1350 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1351 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1352 gfs2_dinode_out(ip, dibh->b_data);
1359 * gfs2_setattr_simple -
1363 * Called with a reference on the vnode.
1368 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1372 if (current->journal_info)
1373 return __gfs2_setattr_simple(ip, attr);
1375 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1379 error = __gfs2_setattr_simple(ip, attr);
1380 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1384 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1386 const struct gfs2_dinode_host *di = &ip->i_di;
1387 struct gfs2_dinode *str = buf;
1389 str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1390 str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1391 str->di_header.__pad0 = 0;
1392 str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1393 str->di_header.__pad1 = 0;
1394 str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1395 str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1396 str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1397 str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1398 str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1399 str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1400 str->di_size = cpu_to_be64(di->di_size);
1401 str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
1402 str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1403 str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1404 str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1406 str->di_goal_meta = cpu_to_be64(ip->i_goal);
1407 str->di_goal_data = cpu_to_be64(ip->i_goal);
1408 str->di_generation = cpu_to_be64(di->di_generation);
1410 str->di_flags = cpu_to_be32(di->di_flags);
1411 str->di_height = cpu_to_be16(ip->i_height);
1412 str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1413 !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
1414 GFS2_FORMAT_DE : 0);
1415 str->di_depth = cpu_to_be16(ip->i_depth);
1416 str->di_entries = cpu_to_be32(di->di_entries);
1418 str->di_eattr = cpu_to_be64(di->di_eattr);
1419 str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1420 str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1421 str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1424 void gfs2_dinode_print(const struct gfs2_inode *ip)
1426 const struct gfs2_dinode_host *di = &ip->i_di;
1428 printk(KERN_INFO " no_formal_ino = %llu\n",
1429 (unsigned long long)ip->i_no_formal_ino);
1430 printk(KERN_INFO " no_addr = %llu\n",
1431 (unsigned long long)ip->i_no_addr);
1432 printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size);
1433 printk(KERN_INFO " blocks = %llu\n",
1434 (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
1435 printk(KERN_INFO " i_goal = %llu\n",
1436 (unsigned long long)ip->i_goal);
1437 printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags);
1438 printk(KERN_INFO " i_height = %u\n", ip->i_height);
1439 printk(KERN_INFO " i_depth = %u\n", ip->i_depth);
1440 printk(KERN_INFO " di_entries = %u\n", di->di_entries);
1441 printk(KERN_INFO " di_eattr = %llu\n",
1442 (unsigned long long)di->di_eattr);