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"
35 #include "ops_inode.h"
41 struct gfs2_inum_range_host {
46 static int iget_test(struct inode *inode, void *opaque)
48 struct gfs2_inode *ip = GFS2_I(inode);
49 u64 *no_addr = opaque;
51 if (ip->i_no_addr == *no_addr &&
52 inode->i_private != NULL)
58 static int iget_set(struct inode *inode, void *opaque)
60 struct gfs2_inode *ip = GFS2_I(inode);
61 u64 *no_addr = opaque;
63 inode->i_ino = (unsigned long)*no_addr;
64 ip->i_no_addr = *no_addr;
68 struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr)
70 unsigned long hash = (unsigned long)no_addr;
71 return ilookup5(sb, hash, iget_test, &no_addr);
74 static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr)
76 unsigned long hash = (unsigned long)no_addr;
77 return iget5_locked(sb, hash, iget_test, iget_set, &no_addr);
80 struct gfs2_skip_data {
85 static int iget_skip_test(struct inode *inode, void *opaque)
87 struct gfs2_inode *ip = GFS2_I(inode);
88 struct gfs2_skip_data *data = opaque;
90 if (ip->i_no_addr == data->no_addr && inode->i_private != NULL){
91 if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)){
100 static int iget_skip_set(struct inode *inode, void *opaque)
102 struct gfs2_inode *ip = GFS2_I(inode);
103 struct gfs2_skip_data *data = opaque;
107 inode->i_ino = (unsigned long)(data->no_addr);
108 ip->i_no_addr = data->no_addr;
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 umode_t mode = inode->i_mode;
138 inode->i_op = &gfs2_file_iops;
139 inode->i_fop = &gfs2_file_fops;
140 inode->i_mapping->a_ops = &gfs2_file_aops;
141 } else if (S_ISDIR(mode)) {
142 inode->i_op = &gfs2_dir_iops;
143 inode->i_fop = &gfs2_dir_fops;
144 } else if (S_ISLNK(mode)) {
145 inode->i_op = &gfs2_symlink_iops;
147 inode->i_op = &gfs2_dev_iops;
150 unlock_new_inode(inode);
154 * gfs2_inode_lookup - Lookup an inode
155 * @sb: The super block
156 * @no_addr: The inode number
157 * @type: The type of the inode
158 * @skip_freeing: set this not return an inode if it is currently being freed.
160 * Returns: A VFS inode, or an error
163 struct inode *gfs2_inode_lookup(struct super_block *sb,
166 u64 no_formal_ino, int skip_freeing)
169 struct gfs2_inode *ip;
170 struct gfs2_glock *io_gl;
174 inode = gfs2_iget_skip(sb, no_addr);
176 inode = gfs2_iget(sb, no_addr);
180 return ERR_PTR(-ENOBUFS);
182 if (inode->i_state & I_NEW) {
183 struct gfs2_sbd *sdp = GFS2_SB(inode);
184 inode->i_private = ip;
185 ip->i_no_formal_ino = no_formal_ino;
187 error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
190 ip->i_gl->gl_object = ip;
192 error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
196 set_bit(GIF_INVALID, &ip->i_flags);
197 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
200 ip->i_iopen_gh.gh_gl->gl_object = ip;
202 gfs2_glock_put(io_gl);
204 if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
207 inode->i_mode = DT2IF(type);
210 * We must read the inode in order to work out its type in
211 * this case. Note that this doesn't happen often as we normally
212 * know the type beforehand. This code path only occurs during
213 * unlinked inode recovery (where it is safe to do this glock,
214 * which is not true in the general case).
216 if (type == DT_UNKNOWN) {
217 struct gfs2_holder gh;
218 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
221 /* Inode is now uptodate */
222 gfs2_glock_dq_uninit(&gh);
231 gfs2_glock_dq(&ip->i_iopen_gh);
233 gfs2_glock_put(io_gl);
235 ip->i_gl->gl_object = NULL;
236 gfs2_glock_put(ip->i_gl);
239 return ERR_PTR(error);
242 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
244 struct gfs2_dinode_host *di = &ip->i_di;
245 const struct gfs2_dinode *str = buf;
247 if (ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)) {
248 if (gfs2_consist_inode(ip))
249 gfs2_dinode_print(ip);
252 ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
253 ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
254 ip->i_inode.i_rdev = 0;
255 switch (ip->i_inode.i_mode & S_IFMT) {
258 ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
259 be32_to_cpu(str->di_minor));
263 ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
264 ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
266 * We will need to review setting the nlink count here in the
267 * light of the forthcoming ro bind mount work. This is a reminder
270 ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
271 di->di_size = be64_to_cpu(str->di_size);
272 i_size_write(&ip->i_inode, di->di_size);
273 di->di_blocks = be64_to_cpu(str->di_blocks);
274 gfs2_set_inode_blocks(&ip->i_inode);
275 ip->i_inode.i_atime.tv_sec = be64_to_cpu(str->di_atime);
276 ip->i_inode.i_atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
277 ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
278 ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
279 ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
280 ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
282 di->di_goal_meta = be64_to_cpu(str->di_goal_meta);
283 di->di_goal_data = be64_to_cpu(str->di_goal_data);
284 di->di_generation = be64_to_cpu(str->di_generation);
286 di->di_flags = be32_to_cpu(str->di_flags);
287 gfs2_set_inode_flags(&ip->i_inode);
288 di->di_height = be16_to_cpu(str->di_height);
290 di->di_depth = be16_to_cpu(str->di_depth);
291 di->di_entries = be32_to_cpu(str->di_entries);
293 di->di_eattr = be64_to_cpu(str->di_eattr);
297 static void gfs2_inode_bh(struct gfs2_inode *ip, struct buffer_head *bh)
303 * gfs2_inode_refresh - Refresh the incore copy of the dinode
304 * @ip: The GFS2 inode
309 int gfs2_inode_refresh(struct gfs2_inode *ip)
311 struct buffer_head *dibh;
314 error = gfs2_meta_inode_buffer(ip, &dibh);
318 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
323 error = gfs2_dinode_in(ip, dibh->b_data);
325 clear_bit(GIF_INVALID, &ip->i_flags);
330 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
332 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
333 struct gfs2_alloc *al;
334 struct gfs2_rgrpd *rgd;
337 if (ip->i_di.di_blocks != 1) {
338 if (gfs2_consist_inode(ip))
339 gfs2_dinode_print(ip);
343 al = gfs2_alloc_get(ip);
345 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
349 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
353 rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
355 gfs2_consist_inode(ip);
357 goto out_rindex_relse;
360 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
363 goto out_rindex_relse;
365 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
369 gfs2_trans_add_gl(ip->i_gl);
371 gfs2_free_di(rgd, ip);
374 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
377 gfs2_glock_dq_uninit(&al->al_rgd_gh);
379 gfs2_glock_dq_uninit(&al->al_ri_gh);
381 gfs2_quota_unhold(ip);
388 * gfs2_change_nlink - Change nlink count on inode
389 * @ip: The GFS2 inode
390 * @diff: The change in the nlink count required
394 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
396 struct buffer_head *dibh;
400 BUG_ON(diff != 1 && diff != -1);
401 nlink = ip->i_inode.i_nlink + diff;
403 /* If we are reducing the nlink count, but the new value ends up being
404 bigger than the old one, we must have underflowed. */
405 if (diff < 0 && nlink > ip->i_inode.i_nlink) {
406 if (gfs2_consist_inode(ip))
407 gfs2_dinode_print(ip);
411 error = gfs2_meta_inode_buffer(ip, &dibh);
416 inc_nlink(&ip->i_inode);
418 drop_nlink(&ip->i_inode);
420 ip->i_inode.i_ctime = CURRENT_TIME;
422 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
423 gfs2_dinode_out(ip, dibh->b_data);
425 mark_inode_dirty(&ip->i_inode);
427 if (ip->i_inode.i_nlink == 0)
428 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
433 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
437 gfs2_str2qstr(&qstr, name);
438 inode = gfs2_lookupi(dip, &qstr, 1, NULL);
439 /* gfs2_lookupi has inconsistent callers: vfs
440 * related routines expect NULL for no entry found,
441 * gfs2_lookup_simple callers expect ENOENT
442 * and do not check for NULL.
445 return ERR_PTR(-ENOENT);
452 * gfs2_lookupi - Look up a filename in a directory and return its inode
453 * @d_gh: An initialized holder for the directory glock
454 * @name: The name of the inode to look for
455 * @is_root: If 1, ignore the caller's permissions
456 * @i_gh: An uninitialized holder for the new inode glock
458 * This can be called via the VFS filldir function when NFS is doing
459 * a readdirplus and the inode which its intending to stat isn't
460 * already in cache. In this case we must not take the directory glock
461 * again, since the readdir call will have already taken that lock.
466 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
467 int is_root, struct nameidata *nd)
469 struct super_block *sb = dir->i_sb;
470 struct gfs2_inode *dip = GFS2_I(dir);
471 struct gfs2_holder d_gh;
473 struct inode *inode = NULL;
476 if (!name->len || name->len > GFS2_FNAMESIZE)
477 return ERR_PTR(-ENAMETOOLONG);
479 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
480 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
481 dir == sb->s_root->d_inode)) {
486 if (gfs2_glock_is_locked_by_me(dip->i_gl) == 0) {
487 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
489 return ERR_PTR(error);
494 error = permission(dir, MAY_EXEC, NULL);
499 inode = gfs2_dir_search(dir, name);
501 error = PTR_ERR(inode);
504 gfs2_glock_dq_uninit(&d_gh);
505 if (error == -ENOENT)
507 return inode ? inode : ERR_PTR(error);
510 static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf)
512 const struct gfs2_inum_range *str = buf;
514 ir->ir_start = be64_to_cpu(str->ir_start);
515 ir->ir_length = be64_to_cpu(str->ir_length);
518 static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf)
520 struct gfs2_inum_range *str = buf;
522 str->ir_start = cpu_to_be64(ir->ir_start);
523 str->ir_length = cpu_to_be64(ir->ir_length);
526 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
528 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
529 struct buffer_head *bh;
530 struct gfs2_inum_range_host ir;
533 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
536 mutex_lock(&sdp->sd_inum_mutex);
538 error = gfs2_meta_inode_buffer(ip, &bh);
540 mutex_unlock(&sdp->sd_inum_mutex);
545 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
548 *formal_ino = ir.ir_start++;
550 gfs2_trans_add_bh(ip->i_gl, bh, 1);
551 gfs2_inum_range_out(&ir,
552 bh->b_data + sizeof(struct gfs2_dinode));
554 mutex_unlock(&sdp->sd_inum_mutex);
561 mutex_unlock(&sdp->sd_inum_mutex);
567 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
569 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
570 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
571 struct gfs2_holder gh;
572 struct buffer_head *bh;
573 struct gfs2_inum_range_host ir;
576 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
580 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
583 mutex_lock(&sdp->sd_inum_mutex);
585 error = gfs2_meta_inode_buffer(ip, &bh);
589 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
592 struct buffer_head *m_bh;
596 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
600 z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
601 x = y = be64_to_cpu(z);
603 ir.ir_length = GFS2_INUM_QUANTUM;
604 x += GFS2_INUM_QUANTUM;
606 gfs2_consist_inode(m_ip);
608 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
609 *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z;
614 *formal_ino = ir.ir_start++;
617 gfs2_trans_add_bh(ip->i_gl, bh, 1);
618 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
623 mutex_unlock(&sdp->sd_inum_mutex);
626 gfs2_glock_dq_uninit(&gh);
630 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
634 error = pick_formal_ino_1(sdp, inum);
638 error = pick_formal_ino_2(sdp, inum);
644 * create_ok - OK to create a new on-disk inode here?
645 * @dip: Directory in which dinode is to be created
646 * @name: Name of new dinode
652 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
657 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
661 /* Don't create entries in an unlinked directory */
662 if (!dip->i_inode.i_nlink)
665 error = gfs2_dir_check(&dip->i_inode, name, NULL);
676 if (dip->i_di.di_entries == (u32)-1)
678 if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
684 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
685 unsigned int *uid, unsigned int *gid)
687 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
688 (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
691 else if (dip->i_inode.i_uid != current->fsuid)
693 *uid = dip->i_inode.i_uid;
695 *uid = current->fsuid;
697 if (dip->i_inode.i_mode & S_ISGID) {
700 *gid = dip->i_inode.i_gid;
702 *gid = current->fsgid;
705 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
707 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
712 dip->i_alloc.al_requested = RES_DINODE;
713 error = gfs2_inplace_reserve(dip);
717 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
721 *no_addr = gfs2_alloc_di(dip, generation);
726 gfs2_inplace_release(dip);
733 * init_dinode - Fill in a new dinode structure
734 * @dip: the directory this inode is being created in
735 * @gl: The glock covering the new inode
736 * @inum: the inode number
737 * @mode: the file permissions
743 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
744 const struct gfs2_inum_host *inum, unsigned int mode,
745 unsigned int uid, unsigned int gid,
746 const u64 *generation, dev_t dev, struct buffer_head **bhp)
748 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
749 struct gfs2_dinode *di;
750 struct buffer_head *dibh;
751 struct timespec tv = CURRENT_TIME;
753 dibh = gfs2_meta_new(gl, inum->no_addr);
754 gfs2_trans_add_bh(gl, dibh, 1);
755 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
756 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
757 di = (struct gfs2_dinode *)dibh->b_data;
759 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
760 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
761 di->di_mode = cpu_to_be32(mode);
762 di->di_uid = cpu_to_be32(uid);
763 di->di_gid = cpu_to_be32(gid);
766 di->di_blocks = cpu_to_be64(1);
767 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
768 di->di_major = cpu_to_be32(MAJOR(dev));
769 di->di_minor = cpu_to_be32(MINOR(dev));
770 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
771 di->di_generation = cpu_to_be64(*generation);
775 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
776 gfs2_tune_get(sdp, gt_new_files_jdata))
777 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
778 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) ||
779 gfs2_tune_get(sdp, gt_new_files_directio))
780 di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO);
781 } else if (S_ISDIR(mode)) {
782 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
783 GFS2_DIF_INHERIT_DIRECTIO);
784 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
785 GFS2_DIF_INHERIT_JDATA);
789 di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
795 memset(&di->__pad4, 0, sizeof(di->__pad4));
797 di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
798 di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
799 di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
800 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
802 set_buffer_uptodate(dibh);
807 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
808 unsigned int mode, const struct gfs2_inum_host *inum,
809 const u64 *generation, dev_t dev, struct buffer_head **bhp)
811 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
812 unsigned int uid, gid;
815 munge_mode_uid_gid(dip, &mode, &uid, &gid);
818 error = gfs2_quota_lock(dip, uid, gid);
822 error = gfs2_quota_check(dip, uid, gid);
826 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
830 init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
831 gfs2_quota_change(dip, +1, uid, gid);
835 gfs2_quota_unlock(dip);
841 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
842 struct gfs2_inode *ip)
844 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
845 struct gfs2_alloc *al;
847 struct buffer_head *dibh;
850 al = gfs2_alloc_get(dip);
852 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
856 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
857 if (alloc_required < 0)
859 if (alloc_required) {
860 error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
862 goto fail_quota_locks;
864 al->al_requested = sdp->sd_max_dirres;
866 error = gfs2_inplace_reserve(dip);
868 goto fail_quota_locks;
870 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
871 al->al_rgd->rd_length +
873 RES_STATFS + RES_QUOTA, 0);
877 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
879 goto fail_quota_locks;
882 error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
886 error = gfs2_meta_inode_buffer(ip, &dibh);
889 ip->i_inode.i_nlink = 1;
890 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
891 gfs2_dinode_out(ip, dibh->b_data);
899 if (dip->i_alloc.al_rgd)
900 gfs2_inplace_release(dip);
903 gfs2_quota_unlock(dip);
910 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
916 struct gfs2_ea_request er;
918 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
919 &name, &value, &len);
922 if (err == -EOPNOTSUPP)
927 memset(&er, 0, sizeof(struct gfs2_ea_request));
929 er.er_type = GFS2_EATYPE_SECURITY;
932 er.er_name_len = strlen(name);
933 er.er_data_len = len;
935 err = gfs2_ea_set_i(ip, &er);
944 * gfs2_createi - Create a new inode
945 * @ghs: An array of two holders
946 * @name: The name of the new file
947 * @mode: the permissions on the new inode
949 * @ghs[0] is an initialized holder for the directory
950 * @ghs[1] is the holder for the inode lock
952 * If the return value is not NULL, the glocks on both the directory and the new
953 * file are held. A transaction has been started and an inplace reservation
959 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
960 unsigned int mode, dev_t dev)
962 struct inode *inode = NULL;
963 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
964 struct inode *dir = &dip->i_inode;
965 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
966 struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
969 struct buffer_head *bh=NULL;
971 if (!name->len || name->len > GFS2_FNAMESIZE)
972 return ERR_PTR(-ENAMETOOLONG);
974 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
975 error = gfs2_glock_nq(ghs);
979 error = create_ok(dip, name, mode);
983 error = pick_formal_ino(sdp, &inum.no_formal_ino);
987 error = alloc_dinode(dip, &inum.no_addr, &generation);
991 error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
992 LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
996 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
1000 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode),
1002 inum.no_formal_ino, 0);
1006 gfs2_inode_bh(GFS2_I(inode), bh);
1008 error = gfs2_inode_refresh(GFS2_I(inode));
1012 error = gfs2_acl_create(dip, GFS2_I(inode));
1016 error = gfs2_security_init(dip, GFS2_I(inode));
1020 error = link_dinode(dip, name, GFS2_I(inode));
1025 return ERR_PTR(-ENOMEM);
1029 gfs2_glock_dq_uninit(ghs + 1);
1035 return ERR_PTR(error);
1039 * gfs2_rmdiri - Remove a directory
1040 * @dip: The parent directory of the directory to be removed
1041 * @name: The name of the directory to be removed
1042 * @ip: The GFS2 inode of the directory to be removed
1044 * Assumes Glocks on dip and ip are held
1049 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
1050 struct gfs2_inode *ip)
1052 struct qstr dotname;
1055 if (ip->i_di.di_entries != 2) {
1056 if (gfs2_consist_inode(ip))
1057 gfs2_dinode_print(ip);
1061 error = gfs2_dir_del(dip, name);
1065 error = gfs2_change_nlink(dip, -1);
1069 gfs2_str2qstr(&dotname, ".");
1070 error = gfs2_dir_del(ip, &dotname);
1074 gfs2_str2qstr(&dotname, "..");
1075 error = gfs2_dir_del(ip, &dotname);
1079 /* It looks odd, but it really should be done twice */
1080 error = gfs2_change_nlink(ip, -1);
1084 error = gfs2_change_nlink(ip, -1);
1092 * gfs2_unlink_ok - check to see that a inode is still in a directory
1093 * @dip: the directory
1094 * @name: the name of the file
1097 * Assumes that the lock on (at least) @dip is held.
1099 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1102 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1103 const struct gfs2_inode *ip)
1107 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1110 if ((dip->i_inode.i_mode & S_ISVTX) &&
1111 dip->i_inode.i_uid != current->fsuid &&
1112 ip->i_inode.i_uid != current->fsuid && !capable(CAP_FOWNER))
1115 if (IS_APPEND(&dip->i_inode))
1118 error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL);
1122 error = gfs2_dir_check(&dip->i_inode, name, ip);
1130 * gfs2_ok_to_move - check if it's ok to move a directory to another directory
1134 * Follow @to back to the root and make sure we don't encounter @this
1135 * Assumes we already hold the rename lock.
1140 int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to)
1142 struct inode *dir = &to->i_inode;
1143 struct super_block *sb = dir->i_sb;
1148 gfs2_str2qstr(&dotdot, "..");
1153 if (dir == &this->i_inode) {
1157 if (dir == sb->s_root->d_inode) {
1162 tmp = gfs2_lookupi(dir, &dotdot, 1, NULL);
1164 error = PTR_ERR(tmp);
1178 * gfs2_readlinki - return the contents of a symlink
1179 * @ip: the symlink's inode
1180 * @buf: a pointer to the buffer to be filled
1181 * @len: a pointer to the length of @buf
1183 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1184 * to be freed by the caller.
1189 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1191 struct gfs2_holder i_gh;
1192 struct buffer_head *dibh;
1196 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
1197 error = gfs2_glock_nq_atime(&i_gh);
1199 gfs2_holder_uninit(&i_gh);
1203 if (!ip->i_di.di_size) {
1204 gfs2_consist_inode(ip);
1209 error = gfs2_meta_inode_buffer(ip, &dibh);
1213 x = ip->i_di.di_size + 1;
1215 *buf = kmalloc(x, GFP_KERNEL);
1222 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1228 gfs2_glock_dq_uninit(&i_gh);
1233 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1234 * conditionally update the inode's atime
1235 * @gh: the holder to acquire
1237 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1238 * Update if the difference between the current time and the inode's current
1239 * atime is greater than an interval specified at mount.
1244 int gfs2_glock_nq_atime(struct gfs2_holder *gh)
1246 struct gfs2_glock *gl = gh->gh_gl;
1247 struct gfs2_sbd *sdp = gl->gl_sbd;
1248 struct gfs2_inode *ip = gl->gl_object;
1249 s64 quantum = gfs2_tune_get(sdp, gt_atime_quantum);
1253 struct timespec tv = CURRENT_TIME;
1255 if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) ||
1256 gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) ||
1257 gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops))
1260 state = gh->gh_state;
1261 flags = gh->gh_flags;
1263 error = gfs2_glock_nq(gh);
1267 if (test_bit(SDF_NOATIME, &sdp->sd_flags) ||
1268 (sdp->sd_vfs->s_flags & MS_RDONLY))
1271 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1273 gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY,
1275 error = gfs2_glock_nq(gh);
1279 /* Verify that atime hasn't been updated while we were
1280 trying to get exclusive lock. */
1283 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1284 struct buffer_head *dibh;
1285 struct gfs2_dinode *di;
1287 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
1288 if (error == -EROFS)
1293 error = gfs2_meta_inode_buffer(ip, &dibh);
1295 goto fail_end_trans;
1297 ip->i_inode.i_atime = tv;
1299 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1300 di = (struct gfs2_dinode *)dibh->b_data;
1301 di->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1302 di->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1305 gfs2_trans_end(sdp);
1308 /* If someone else has asked for the glock,
1309 unlock and let them have it. Then reacquire
1310 in the original state. */
1311 if (gfs2_glock_is_blocking(gl)) {
1313 gfs2_holder_reinit(state, flags, gh);
1314 return gfs2_glock_nq(gh);
1321 gfs2_trans_end(sdp);
1328 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1330 struct buffer_head *dibh;
1333 error = gfs2_meta_inode_buffer(ip, &dibh);
1335 error = inode_setattr(&ip->i_inode, attr);
1336 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1337 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1338 gfs2_dinode_out(ip, dibh->b_data);
1345 * gfs2_setattr_simple -
1349 * Called with a reference on the vnode.
1354 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1358 if (current->journal_info)
1359 return __gfs2_setattr_simple(ip, attr);
1361 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1365 error = __gfs2_setattr_simple(ip, attr);
1366 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1370 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1372 const struct gfs2_dinode_host *di = &ip->i_di;
1373 struct gfs2_dinode *str = buf;
1375 str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1376 str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1377 str->di_header.__pad0 = 0;
1378 str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1379 str->di_header.__pad1 = 0;
1380 str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1381 str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1382 str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1383 str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1384 str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1385 str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1386 str->di_size = cpu_to_be64(di->di_size);
1387 str->di_blocks = cpu_to_be64(di->di_blocks);
1388 str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1389 str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1390 str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1392 str->di_goal_meta = cpu_to_be64(di->di_goal_meta);
1393 str->di_goal_data = cpu_to_be64(di->di_goal_data);
1394 str->di_generation = cpu_to_be64(di->di_generation);
1396 str->di_flags = cpu_to_be32(di->di_flags);
1397 str->di_height = cpu_to_be16(di->di_height);
1398 str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1399 !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
1400 GFS2_FORMAT_DE : 0);
1401 str->di_depth = cpu_to_be16(di->di_depth);
1402 str->di_entries = cpu_to_be32(di->di_entries);
1404 str->di_eattr = cpu_to_be64(di->di_eattr);
1405 str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1406 str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1407 str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1410 void gfs2_dinode_print(const struct gfs2_inode *ip)
1412 const struct gfs2_dinode_host *di = &ip->i_di;
1414 printk(KERN_INFO " no_formal_ino = %llu\n",
1415 (unsigned long long)ip->i_no_formal_ino);
1416 printk(KERN_INFO " no_addr = %llu\n",
1417 (unsigned long long)ip->i_no_addr);
1418 printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size);
1419 printk(KERN_INFO " di_blocks = %llu\n",
1420 (unsigned long long)di->di_blocks);
1421 printk(KERN_INFO " di_goal_meta = %llu\n",
1422 (unsigned long long)di->di_goal_meta);
1423 printk(KERN_INFO " di_goal_data = %llu\n",
1424 (unsigned long long)di->di_goal_data);
1425 printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags);
1426 printk(KERN_INFO " di_height = %u\n", di->di_height);
1427 printk(KERN_INFO " di_depth = %u\n", di->di_depth);
1428 printk(KERN_INFO " di_entries = %u\n", di->di_entries);
1429 printk(KERN_INFO " di_eattr = %llu\n",
1430 (unsigned long long)di->di_eattr);