2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
7 * of the GNU General Public License version 2.
10 #include <linux/sched.h>
11 #include <linux/slab.h>
12 #include <linux/spinlock.h>
13 #include <linux/completion.h>
14 #include <linux/buffer_head.h>
15 #include <linux/posix_acl.h>
16 #include <linux/sort.h>
17 #include <linux/gfs2_ondisk.h>
18 #include <linux/crc32.h>
19 #include <linux/lm_interface.h>
20 #include <linux/security.h>
21 #include <linux/time.h>
34 #include "ops_address.h"
40 struct gfs2_inum_range_host {
45 static int iget_test(struct inode *inode, void *opaque)
47 struct gfs2_inode *ip = GFS2_I(inode);
48 u64 *no_addr = opaque;
50 if (ip->i_no_addr == *no_addr && test_bit(GIF_USER, &ip->i_flags))
56 static int iget_set(struct inode *inode, void *opaque)
58 struct gfs2_inode *ip = GFS2_I(inode);
59 u64 *no_addr = opaque;
61 inode->i_ino = (unsigned long)*no_addr;
62 ip->i_no_addr = *no_addr;
63 set_bit(GIF_USER, &ip->i_flags);
67 struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr)
69 unsigned long hash = (unsigned long)no_addr;
70 return ilookup5(sb, hash, iget_test, &no_addr);
73 static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr)
75 unsigned long hash = (unsigned long)no_addr;
76 return iget5_locked(sb, hash, iget_test, iget_set, &no_addr);
79 struct gfs2_skip_data {
84 static int iget_skip_test(struct inode *inode, void *opaque)
86 struct gfs2_inode *ip = GFS2_I(inode);
87 struct gfs2_skip_data *data = opaque;
89 if (ip->i_no_addr == data->no_addr && test_bit(GIF_USER, &ip->i_flags)){
90 if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)){
99 static int iget_skip_set(struct inode *inode, void *opaque)
101 struct gfs2_inode *ip = GFS2_I(inode);
102 struct gfs2_skip_data *data = opaque;
106 inode->i_ino = (unsigned long)(data->no_addr);
107 ip->i_no_addr = data->no_addr;
108 set_bit(GIF_USER, &ip->i_flags);
112 static struct inode *gfs2_iget_skip(struct super_block *sb,
115 struct gfs2_skip_data data;
116 unsigned long hash = (unsigned long)no_addr;
118 data.no_addr = no_addr;
120 return iget5_locked(sb, hash, iget_skip_test, iget_skip_set, &data);
124 * GFS2 lookup code fills in vfs inode contents based on info obtained
125 * from directory entry inside gfs2_inode_lookup(). This has caused issues
126 * with NFS code path since its get_dentry routine doesn't have the relevant
127 * directory entry when gfs2_inode_lookup() is invoked. Part of the code
128 * segment inside gfs2_inode_lookup code needs to get moved around.
130 * Clean up I_LOCK and I_NEW as well.
133 void gfs2_set_iop(struct inode *inode)
135 struct gfs2_sbd *sdp = GFS2_SB(inode);
136 umode_t mode = inode->i_mode;
139 inode->i_op = &gfs2_file_iops;
140 if (sdp->sd_args.ar_localflocks)
141 inode->i_fop = &gfs2_file_fops_nolock;
143 inode->i_fop = &gfs2_file_fops;
144 } else if (S_ISDIR(mode)) {
145 inode->i_op = &gfs2_dir_iops;
146 if (sdp->sd_args.ar_localflocks)
147 inode->i_fop = &gfs2_dir_fops_nolock;
149 inode->i_fop = &gfs2_dir_fops;
150 } else if (S_ISLNK(mode)) {
151 inode->i_op = &gfs2_symlink_iops;
153 inode->i_op = &gfs2_file_iops;
154 init_special_inode(inode, inode->i_mode, inode->i_rdev);
157 unlock_new_inode(inode);
161 * gfs2_inode_lookup - Lookup an inode
162 * @sb: The super block
163 * @no_addr: The inode number
164 * @type: The type of the inode
165 * @skip_freeing: set this not return an inode if it is currently being freed.
167 * Returns: A VFS inode, or an error
170 struct inode *gfs2_inode_lookup(struct super_block *sb,
173 u64 no_formal_ino, int skip_freeing)
176 struct gfs2_inode *ip;
177 struct gfs2_glock *io_gl;
181 inode = gfs2_iget_skip(sb, no_addr);
183 inode = gfs2_iget(sb, no_addr);
187 return ERR_PTR(-ENOBUFS);
189 if (inode->i_state & I_NEW) {
190 struct gfs2_sbd *sdp = GFS2_SB(inode);
191 ip->i_no_formal_ino = no_formal_ino;
193 error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
196 ip->i_gl->gl_object = ip;
198 error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
202 set_bit(GIF_INVALID, &ip->i_flags);
203 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
206 ip->i_iopen_gh.gh_gl->gl_object = ip;
208 gfs2_glock_put(io_gl);
210 if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
213 inode->i_mode = DT2IF(type);
216 * We must read the inode in order to work out its type in
217 * this case. Note that this doesn't happen often as we normally
218 * know the type beforehand. This code path only occurs during
219 * unlinked inode recovery (where it is safe to do this glock,
220 * which is not true in the general case).
222 if (type == DT_UNKNOWN) {
223 struct gfs2_holder gh;
224 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
227 /* Inode is now uptodate */
228 gfs2_glock_dq_uninit(&gh);
237 gfs2_glock_dq(&ip->i_iopen_gh);
239 gfs2_glock_put(io_gl);
241 ip->i_gl->gl_object = NULL;
242 gfs2_glock_put(ip->i_gl);
245 return ERR_PTR(error);
248 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
250 struct gfs2_dinode_host *di = &ip->i_di;
251 const struct gfs2_dinode *str = buf;
252 struct timespec atime;
255 if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
257 ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
258 ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
259 ip->i_inode.i_rdev = 0;
260 switch (ip->i_inode.i_mode & S_IFMT) {
263 ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
264 be32_to_cpu(str->di_minor));
268 ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
269 ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
271 * We will need to review setting the nlink count here in the
272 * light of the forthcoming ro bind mount work. This is a reminder
275 ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
276 di->di_size = be64_to_cpu(str->di_size);
277 i_size_write(&ip->i_inode, di->di_size);
278 gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
279 atime.tv_sec = be64_to_cpu(str->di_atime);
280 atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
281 if (timespec_compare(&ip->i_inode.i_atime, &atime) < 0)
282 ip->i_inode.i_atime = atime;
283 ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
284 ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
285 ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
286 ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
288 ip->i_goal = be64_to_cpu(str->di_goal_meta);
289 ip->i_generation = be64_to_cpu(str->di_generation);
291 di->di_flags = be32_to_cpu(str->di_flags);
292 gfs2_set_inode_flags(&ip->i_inode);
293 height = be16_to_cpu(str->di_height);
294 if (unlikely(height > GFS2_MAX_META_HEIGHT))
296 ip->i_height = (u8)height;
298 depth = be16_to_cpu(str->di_depth);
299 if (unlikely(depth > GFS2_DIR_MAX_DEPTH))
301 ip->i_depth = (u8)depth;
302 ip->i_entries = be32_to_cpu(str->di_entries);
304 ip->i_eattr = be64_to_cpu(str->di_eattr);
305 if (S_ISREG(ip->i_inode.i_mode))
306 gfs2_set_aops(&ip->i_inode);
310 if (gfs2_consist_inode(ip))
311 gfs2_dinode_print(ip);
316 * gfs2_inode_refresh - Refresh the incore copy of the dinode
317 * @ip: The GFS2 inode
322 int gfs2_inode_refresh(struct gfs2_inode *ip)
324 struct buffer_head *dibh;
327 error = gfs2_meta_inode_buffer(ip, &dibh);
331 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
336 error = gfs2_dinode_in(ip, dibh->b_data);
338 clear_bit(GIF_INVALID, &ip->i_flags);
343 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
345 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
346 struct gfs2_alloc *al;
347 struct gfs2_rgrpd *rgd;
350 if (gfs2_get_inode_blocks(&ip->i_inode) != 1) {
351 if (gfs2_consist_inode(ip))
352 gfs2_dinode_print(ip);
356 al = gfs2_alloc_get(ip);
360 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
364 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
368 rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
370 gfs2_consist_inode(ip);
372 goto out_rindex_relse;
375 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
378 goto out_rindex_relse;
380 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
384 set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
385 set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
387 gfs2_free_di(rgd, ip);
390 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
393 gfs2_glock_dq_uninit(&al->al_rgd_gh);
395 gfs2_glock_dq_uninit(&al->al_ri_gh);
397 gfs2_quota_unhold(ip);
404 * gfs2_change_nlink - Change nlink count on inode
405 * @ip: The GFS2 inode
406 * @diff: The change in the nlink count required
410 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
412 struct buffer_head *dibh;
416 BUG_ON(diff != 1 && diff != -1);
417 nlink = ip->i_inode.i_nlink + diff;
419 /* If we are reducing the nlink count, but the new value ends up being
420 bigger than the old one, we must have underflowed. */
421 if (diff < 0 && nlink > ip->i_inode.i_nlink) {
422 if (gfs2_consist_inode(ip))
423 gfs2_dinode_print(ip);
427 error = gfs2_meta_inode_buffer(ip, &dibh);
432 inc_nlink(&ip->i_inode);
434 drop_nlink(&ip->i_inode);
436 ip->i_inode.i_ctime = CURRENT_TIME;
438 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
439 gfs2_dinode_out(ip, dibh->b_data);
441 mark_inode_dirty(&ip->i_inode);
443 if (ip->i_inode.i_nlink == 0)
444 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
449 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
453 gfs2_str2qstr(&qstr, name);
454 inode = gfs2_lookupi(dip, &qstr, 1);
455 /* gfs2_lookupi has inconsistent callers: vfs
456 * related routines expect NULL for no entry found,
457 * gfs2_lookup_simple callers expect ENOENT
458 * and do not check for NULL.
461 return ERR_PTR(-ENOENT);
468 * gfs2_lookupi - Look up a filename in a directory and return its inode
469 * @d_gh: An initialized holder for the directory glock
470 * @name: The name of the inode to look for
471 * @is_root: If 1, ignore the caller's permissions
472 * @i_gh: An uninitialized holder for the new inode glock
474 * This can be called via the VFS filldir function when NFS is doing
475 * a readdirplus and the inode which its intending to stat isn't
476 * already in cache. In this case we must not take the directory glock
477 * again, since the readdir call will have already taken that lock.
482 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
485 struct super_block *sb = dir->i_sb;
486 struct gfs2_inode *dip = GFS2_I(dir);
487 struct gfs2_holder d_gh;
489 struct inode *inode = NULL;
492 if (!name->len || name->len > GFS2_FNAMESIZE)
493 return ERR_PTR(-ENAMETOOLONG);
495 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
496 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
497 dir == sb->s_root->d_inode)) {
502 if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
503 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
505 return ERR_PTR(error);
510 error = gfs2_permission(dir, MAY_EXEC);
515 inode = gfs2_dir_search(dir, name);
517 error = PTR_ERR(inode);
520 gfs2_glock_dq_uninit(&d_gh);
521 if (error == -ENOENT)
523 return inode ? inode : ERR_PTR(error);
526 static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf)
528 const struct gfs2_inum_range *str = buf;
530 ir->ir_start = be64_to_cpu(str->ir_start);
531 ir->ir_length = be64_to_cpu(str->ir_length);
534 static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf)
536 struct gfs2_inum_range *str = buf;
538 str->ir_start = cpu_to_be64(ir->ir_start);
539 str->ir_length = cpu_to_be64(ir->ir_length);
542 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
544 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
545 struct buffer_head *bh;
546 struct gfs2_inum_range_host ir;
549 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
552 mutex_lock(&sdp->sd_inum_mutex);
554 error = gfs2_meta_inode_buffer(ip, &bh);
556 mutex_unlock(&sdp->sd_inum_mutex);
561 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
564 *formal_ino = ir.ir_start++;
566 gfs2_trans_add_bh(ip->i_gl, bh, 1);
567 gfs2_inum_range_out(&ir,
568 bh->b_data + sizeof(struct gfs2_dinode));
570 mutex_unlock(&sdp->sd_inum_mutex);
577 mutex_unlock(&sdp->sd_inum_mutex);
583 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
585 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
586 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
587 struct gfs2_holder gh;
588 struct buffer_head *bh;
589 struct gfs2_inum_range_host ir;
592 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
596 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
599 mutex_lock(&sdp->sd_inum_mutex);
601 error = gfs2_meta_inode_buffer(ip, &bh);
605 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
608 struct buffer_head *m_bh;
612 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
616 z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
617 x = y = be64_to_cpu(z);
619 ir.ir_length = GFS2_INUM_QUANTUM;
620 x += GFS2_INUM_QUANTUM;
622 gfs2_consist_inode(m_ip);
624 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
625 *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z;
630 *formal_ino = ir.ir_start++;
633 gfs2_trans_add_bh(ip->i_gl, bh, 1);
634 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
639 mutex_unlock(&sdp->sd_inum_mutex);
642 gfs2_glock_dq_uninit(&gh);
646 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
650 error = pick_formal_ino_1(sdp, inum);
654 error = pick_formal_ino_2(sdp, inum);
660 * create_ok - OK to create a new on-disk inode here?
661 * @dip: Directory in which dinode is to be created
662 * @name: Name of new dinode
668 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
673 error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
677 /* Don't create entries in an unlinked directory */
678 if (!dip->i_inode.i_nlink)
681 error = gfs2_dir_check(&dip->i_inode, name, NULL);
692 if (dip->i_entries == (u32)-1)
694 if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
700 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
701 unsigned int *uid, unsigned int *gid)
703 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
704 (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
707 else if (dip->i_inode.i_uid != current_fsuid())
709 *uid = dip->i_inode.i_uid;
711 *uid = current_fsuid();
713 if (dip->i_inode.i_mode & S_ISGID) {
716 *gid = dip->i_inode.i_gid;
718 *gid = current_fsgid();
721 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
723 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
726 if (gfs2_alloc_get(dip) == NULL)
729 dip->i_alloc->al_requested = RES_DINODE;
730 error = gfs2_inplace_reserve(dip);
734 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
738 *no_addr = gfs2_alloc_di(dip, generation);
743 gfs2_inplace_release(dip);
750 * init_dinode - Fill in a new dinode structure
751 * @dip: the directory this inode is being created in
752 * @gl: The glock covering the new inode
753 * @inum: the inode number
754 * @mode: the file permissions
760 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
761 const struct gfs2_inum_host *inum, unsigned int mode,
762 unsigned int uid, unsigned int gid,
763 const u64 *generation, dev_t dev, struct buffer_head **bhp)
765 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
766 struct gfs2_dinode *di;
767 struct buffer_head *dibh;
768 struct timespec tv = CURRENT_TIME;
770 dibh = gfs2_meta_new(gl, inum->no_addr);
771 gfs2_trans_add_bh(gl, dibh, 1);
772 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
773 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
774 di = (struct gfs2_dinode *)dibh->b_data;
776 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
777 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
778 di->di_mode = cpu_to_be32(mode);
779 di->di_uid = cpu_to_be32(uid);
780 di->di_gid = cpu_to_be32(gid);
783 di->di_blocks = cpu_to_be64(1);
784 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
785 di->di_major = cpu_to_be32(MAJOR(dev));
786 di->di_minor = cpu_to_be32(MINOR(dev));
787 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
788 di->di_generation = cpu_to_be64(*generation);
792 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
793 gfs2_tune_get(sdp, gt_new_files_jdata))
794 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
795 } else if (S_ISDIR(mode)) {
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);
828 if (!gfs2_alloc_get(dip))
831 error = gfs2_quota_lock(dip, uid, gid);
835 error = gfs2_quota_check(dip, uid, gid);
839 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
843 init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
844 gfs2_quota_change(dip, +1, uid, gid);
848 gfs2_quota_unlock(dip);
854 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
855 struct gfs2_inode *ip)
857 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
858 struct gfs2_alloc *al;
860 struct buffer_head *dibh;
863 al = gfs2_alloc_get(dip);
867 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
871 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
872 if (alloc_required < 0)
873 goto fail_quota_locks;
874 if (alloc_required) {
875 error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
877 goto fail_quota_locks;
879 al->al_requested = sdp->sd_max_dirres;
881 error = gfs2_inplace_reserve(dip);
883 goto fail_quota_locks;
885 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
886 al->al_rgd->rd_length +
888 RES_STATFS + RES_QUOTA, 0);
892 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
894 goto fail_quota_locks;
897 error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
901 error = gfs2_meta_inode_buffer(ip, &dibh);
904 ip->i_inode.i_nlink = 1;
905 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
906 gfs2_dinode_out(ip, dibh->b_data);
914 if (dip->i_alloc->al_rgd)
915 gfs2_inplace_release(dip);
918 gfs2_quota_unlock(dip);
925 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
931 struct gfs2_ea_request er;
933 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
934 &name, &value, &len);
937 if (err == -EOPNOTSUPP)
942 memset(&er, 0, sizeof(struct gfs2_ea_request));
944 er.er_type = GFS2_EATYPE_SECURITY;
947 er.er_name_len = strlen(name);
948 er.er_data_len = len;
950 err = gfs2_ea_set_i(ip, &er);
959 * gfs2_createi - Create a new inode
960 * @ghs: An array of two holders
961 * @name: The name of the new file
962 * @mode: the permissions on the new inode
964 * @ghs[0] is an initialized holder for the directory
965 * @ghs[1] is the holder for the inode lock
967 * If the return value is not NULL, the glocks on both the directory and the new
968 * file are held. A transaction has been started and an inplace reservation
974 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
975 unsigned int mode, dev_t dev)
977 struct inode *inode = NULL;
978 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
979 struct inode *dir = &dip->i_inode;
980 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
981 struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
984 struct buffer_head *bh = NULL;
986 if (!name->len || name->len > GFS2_FNAMESIZE)
987 return ERR_PTR(-ENAMETOOLONG);
989 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
990 error = gfs2_glock_nq(ghs);
994 error = create_ok(dip, name, mode);
998 error = pick_formal_ino(sdp, &inum.no_formal_ino);
1002 error = alloc_dinode(dip, &inum.no_addr, &generation);
1006 error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
1007 LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
1011 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
1015 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode),
1017 inum.no_formal_ino, 0);
1021 error = gfs2_inode_refresh(GFS2_I(inode));
1025 error = gfs2_acl_create(dip, GFS2_I(inode));
1029 error = gfs2_security_init(dip, GFS2_I(inode));
1033 error = link_dinode(dip, name, GFS2_I(inode));
1042 gfs2_glock_dq_uninit(ghs + 1);
1043 if (inode && !IS_ERR(inode))
1050 return ERR_PTR(error);
1054 * gfs2_rmdiri - Remove a directory
1055 * @dip: The parent directory of the directory to be removed
1056 * @name: The name of the directory to be removed
1057 * @ip: The GFS2 inode of the directory to be removed
1059 * Assumes Glocks on dip and ip are held
1064 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
1065 struct gfs2_inode *ip)
1067 struct qstr dotname;
1070 if (ip->i_entries != 2) {
1071 if (gfs2_consist_inode(ip))
1072 gfs2_dinode_print(ip);
1076 error = gfs2_dir_del(dip, name);
1080 error = gfs2_change_nlink(dip, -1);
1084 gfs2_str2qstr(&dotname, ".");
1085 error = gfs2_dir_del(ip, &dotname);
1089 gfs2_str2qstr(&dotname, "..");
1090 error = gfs2_dir_del(ip, &dotname);
1094 /* It looks odd, but it really should be done twice */
1095 error = gfs2_change_nlink(ip, -1);
1099 error = gfs2_change_nlink(ip, -1);
1107 * gfs2_unlink_ok - check to see that a inode is still in a directory
1108 * @dip: the directory
1109 * @name: the name of the file
1112 * Assumes that the lock on (at least) @dip is held.
1114 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1117 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1118 const struct gfs2_inode *ip)
1122 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1125 if ((dip->i_inode.i_mode & S_ISVTX) &&
1126 dip->i_inode.i_uid != current_fsuid() &&
1127 ip->i_inode.i_uid != current_fsuid() && !capable(CAP_FOWNER))
1130 if (IS_APPEND(&dip->i_inode))
1133 error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
1137 error = gfs2_dir_check(&dip->i_inode, name, ip);
1145 * gfs2_readlinki - return the contents of a symlink
1146 * @ip: the symlink's inode
1147 * @buf: a pointer to the buffer to be filled
1148 * @len: a pointer to the length of @buf
1150 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1151 * to be freed by the caller.
1156 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1158 struct gfs2_holder i_gh;
1159 struct buffer_head *dibh;
1163 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &i_gh);
1164 error = gfs2_glock_nq(&i_gh);
1166 gfs2_holder_uninit(&i_gh);
1170 if (!ip->i_di.di_size) {
1171 gfs2_consist_inode(ip);
1176 error = gfs2_meta_inode_buffer(ip, &dibh);
1180 x = ip->i_di.di_size + 1;
1182 *buf = kmalloc(x, GFP_NOFS);
1189 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1195 gfs2_glock_dq_uninit(&i_gh);
1200 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1202 struct buffer_head *dibh;
1205 error = gfs2_meta_inode_buffer(ip, &dibh);
1207 error = inode_setattr(&ip->i_inode, attr);
1208 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1209 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1210 gfs2_dinode_out(ip, dibh->b_data);
1217 * gfs2_setattr_simple -
1221 * Called with a reference on the vnode.
1226 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1230 if (current->journal_info)
1231 return __gfs2_setattr_simple(ip, attr);
1233 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1237 error = __gfs2_setattr_simple(ip, attr);
1238 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1242 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1244 const struct gfs2_dinode_host *di = &ip->i_di;
1245 struct gfs2_dinode *str = buf;
1247 str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1248 str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1249 str->di_header.__pad0 = 0;
1250 str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1251 str->di_header.__pad1 = 0;
1252 str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1253 str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1254 str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1255 str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1256 str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1257 str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1258 str->di_size = cpu_to_be64(di->di_size);
1259 str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
1260 str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1261 str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1262 str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1264 str->di_goal_meta = cpu_to_be64(ip->i_goal);
1265 str->di_goal_data = cpu_to_be64(ip->i_goal);
1266 str->di_generation = cpu_to_be64(ip->i_generation);
1268 str->di_flags = cpu_to_be32(di->di_flags);
1269 str->di_height = cpu_to_be16(ip->i_height);
1270 str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1271 !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
1272 GFS2_FORMAT_DE : 0);
1273 str->di_depth = cpu_to_be16(ip->i_depth);
1274 str->di_entries = cpu_to_be32(ip->i_entries);
1276 str->di_eattr = cpu_to_be64(ip->i_eattr);
1277 str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1278 str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1279 str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1282 void gfs2_dinode_print(const struct gfs2_inode *ip)
1284 const struct gfs2_dinode_host *di = &ip->i_di;
1286 printk(KERN_INFO " no_formal_ino = %llu\n",
1287 (unsigned long long)ip->i_no_formal_ino);
1288 printk(KERN_INFO " no_addr = %llu\n",
1289 (unsigned long long)ip->i_no_addr);
1290 printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size);
1291 printk(KERN_INFO " blocks = %llu\n",
1292 (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
1293 printk(KERN_INFO " i_goal = %llu\n",
1294 (unsigned long long)ip->i_goal);
1295 printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags);
1296 printk(KERN_INFO " i_height = %u\n", ip->i_height);
1297 printk(KERN_INFO " i_depth = %u\n", ip->i_depth);
1298 printk(KERN_INFO " i_entries = %u\n", ip->i_entries);
1299 printk(KERN_INFO " i_eattr = %llu\n",
1300 (unsigned long long)ip->i_eattr);