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"
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 && test_bit(GIF_USER, &ip->i_flags))
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;
64 set_bit(GIF_USER, &ip->i_flags);
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 && test_bit(GIF_USER, &ip->i_flags)){
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;
109 set_bit(GIF_USER, &ip->i_flags);
113 static struct inode *gfs2_iget_skip(struct super_block *sb,
116 struct gfs2_skip_data data;
117 unsigned long hash = (unsigned long)no_addr;
119 data.no_addr = no_addr;
121 return iget5_locked(sb, hash, iget_skip_test, iget_skip_set, &data);
125 * GFS2 lookup code fills in vfs inode contents based on info obtained
126 * from directory entry inside gfs2_inode_lookup(). This has caused issues
127 * with NFS code path since its get_dentry routine doesn't have the relevant
128 * directory entry when gfs2_inode_lookup() is invoked. Part of the code
129 * segment inside gfs2_inode_lookup code needs to get moved around.
131 * Clean up I_LOCK and I_NEW as well.
134 void gfs2_set_iop(struct inode *inode)
136 struct gfs2_sbd *sdp = GFS2_SB(inode);
137 umode_t mode = inode->i_mode;
140 inode->i_op = &gfs2_file_iops;
141 if (sdp->sd_args.ar_localflocks)
142 inode->i_fop = &gfs2_file_fops_nolock;
144 inode->i_fop = &gfs2_file_fops;
145 } else if (S_ISDIR(mode)) {
146 inode->i_op = &gfs2_dir_iops;
147 if (sdp->sd_args.ar_localflocks)
148 inode->i_fop = &gfs2_dir_fops_nolock;
150 inode->i_fop = &gfs2_dir_fops;
151 } else if (S_ISLNK(mode)) {
152 inode->i_op = &gfs2_symlink_iops;
154 inode->i_op = &gfs2_file_iops;
155 init_special_inode(inode, inode->i_mode, inode->i_rdev);
158 unlock_new_inode(inode);
162 * gfs2_inode_lookup - Lookup an inode
163 * @sb: The super block
164 * @no_addr: The inode number
165 * @type: The type of the inode
166 * @skip_freeing: set this not return an inode if it is currently being freed.
168 * Returns: A VFS inode, or an error
171 struct inode *gfs2_inode_lookup(struct super_block *sb,
174 u64 no_formal_ino, int skip_freeing)
177 struct gfs2_inode *ip;
178 struct gfs2_glock *io_gl;
182 inode = gfs2_iget_skip(sb, no_addr);
184 inode = gfs2_iget(sb, no_addr);
188 return ERR_PTR(-ENOBUFS);
190 if (inode->i_state & I_NEW) {
191 struct gfs2_sbd *sdp = GFS2_SB(inode);
192 ip->i_no_formal_ino = no_formal_ino;
194 error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
197 ip->i_gl->gl_object = ip;
199 error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
203 set_bit(GIF_INVALID, &ip->i_flags);
204 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
207 ip->i_iopen_gh.gh_gl->gl_object = ip;
209 gfs2_glock_put(io_gl);
211 if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
214 inode->i_mode = DT2IF(type);
217 * We must read the inode in order to work out its type in
218 * this case. Note that this doesn't happen often as we normally
219 * know the type beforehand. This code path only occurs during
220 * unlinked inode recovery (where it is safe to do this glock,
221 * which is not true in the general case).
223 if (type == DT_UNKNOWN) {
224 struct gfs2_holder gh;
225 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
228 /* Inode is now uptodate */
229 gfs2_glock_dq_uninit(&gh);
238 gfs2_glock_dq(&ip->i_iopen_gh);
240 gfs2_glock_put(io_gl);
242 ip->i_gl->gl_object = NULL;
243 gfs2_glock_put(ip->i_gl);
246 return ERR_PTR(error);
249 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
251 struct gfs2_dinode_host *di = &ip->i_di;
252 const struct gfs2_dinode *str = buf;
253 struct timespec atime;
256 if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
258 ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
259 ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
260 ip->i_inode.i_rdev = 0;
261 switch (ip->i_inode.i_mode & S_IFMT) {
264 ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
265 be32_to_cpu(str->di_minor));
269 ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
270 ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
272 * We will need to review setting the nlink count here in the
273 * light of the forthcoming ro bind mount work. This is a reminder
276 ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
277 di->di_size = be64_to_cpu(str->di_size);
278 i_size_write(&ip->i_inode, di->di_size);
279 gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
280 atime.tv_sec = be64_to_cpu(str->di_atime);
281 atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
282 if (timespec_compare(&ip->i_inode.i_atime, &atime) < 0)
283 ip->i_inode.i_atime = atime;
284 ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
285 ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
286 ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
287 ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
289 ip->i_goal = be64_to_cpu(str->di_goal_meta);
290 di->di_generation = be64_to_cpu(str->di_generation);
292 di->di_flags = be32_to_cpu(str->di_flags);
293 gfs2_set_inode_flags(&ip->i_inode);
294 height = be16_to_cpu(str->di_height);
295 if (unlikely(height > GFS2_MAX_META_HEIGHT))
297 ip->i_height = (u8)height;
299 depth = be16_to_cpu(str->di_depth);
300 if (unlikely(depth > GFS2_DIR_MAX_DEPTH))
302 ip->i_depth = (u8)depth;
303 di->di_entries = be32_to_cpu(str->di_entries);
305 di->di_eattr = be64_to_cpu(str->di_eattr);
306 if (S_ISREG(ip->i_inode.i_mode))
307 gfs2_set_aops(&ip->i_inode);
311 if (gfs2_consist_inode(ip))
312 gfs2_dinode_print(ip);
317 * gfs2_inode_refresh - Refresh the incore copy of the dinode
318 * @ip: The GFS2 inode
323 int gfs2_inode_refresh(struct gfs2_inode *ip)
325 struct buffer_head *dibh;
328 error = gfs2_meta_inode_buffer(ip, &dibh);
332 if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
337 error = gfs2_dinode_in(ip, dibh->b_data);
339 clear_bit(GIF_INVALID, &ip->i_flags);
344 int gfs2_dinode_dealloc(struct gfs2_inode *ip)
346 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
347 struct gfs2_alloc *al;
348 struct gfs2_rgrpd *rgd;
351 if (gfs2_get_inode_blocks(&ip->i_inode) != 1) {
352 if (gfs2_consist_inode(ip))
353 gfs2_dinode_print(ip);
357 al = gfs2_alloc_get(ip);
361 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
365 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
369 rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
371 gfs2_consist_inode(ip);
373 goto out_rindex_relse;
376 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
379 goto out_rindex_relse;
381 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
385 set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
386 set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
388 gfs2_free_di(rgd, ip);
391 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
394 gfs2_glock_dq_uninit(&al->al_rgd_gh);
396 gfs2_glock_dq_uninit(&al->al_ri_gh);
398 gfs2_quota_unhold(ip);
405 * gfs2_change_nlink - Change nlink count on inode
406 * @ip: The GFS2 inode
407 * @diff: The change in the nlink count required
411 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
413 struct buffer_head *dibh;
417 BUG_ON(diff != 1 && diff != -1);
418 nlink = ip->i_inode.i_nlink + diff;
420 /* If we are reducing the nlink count, but the new value ends up being
421 bigger than the old one, we must have underflowed. */
422 if (diff < 0 && nlink > ip->i_inode.i_nlink) {
423 if (gfs2_consist_inode(ip))
424 gfs2_dinode_print(ip);
428 error = gfs2_meta_inode_buffer(ip, &dibh);
433 inc_nlink(&ip->i_inode);
435 drop_nlink(&ip->i_inode);
437 ip->i_inode.i_ctime = CURRENT_TIME;
439 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
440 gfs2_dinode_out(ip, dibh->b_data);
442 mark_inode_dirty(&ip->i_inode);
444 if (ip->i_inode.i_nlink == 0)
445 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
450 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
454 gfs2_str2qstr(&qstr, name);
455 inode = gfs2_lookupi(dip, &qstr, 1);
456 /* gfs2_lookupi has inconsistent callers: vfs
457 * related routines expect NULL for no entry found,
458 * gfs2_lookup_simple callers expect ENOENT
459 * and do not check for NULL.
462 return ERR_PTR(-ENOENT);
469 * gfs2_lookupi - Look up a filename in a directory and return its inode
470 * @d_gh: An initialized holder for the directory glock
471 * @name: The name of the inode to look for
472 * @is_root: If 1, ignore the caller's permissions
473 * @i_gh: An uninitialized holder for the new inode glock
475 * This can be called via the VFS filldir function when NFS is doing
476 * a readdirplus and the inode which its intending to stat isn't
477 * already in cache. In this case we must not take the directory glock
478 * again, since the readdir call will have already taken that lock.
483 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
486 struct super_block *sb = dir->i_sb;
487 struct gfs2_inode *dip = GFS2_I(dir);
488 struct gfs2_holder d_gh;
490 struct inode *inode = NULL;
493 if (!name->len || name->len > GFS2_FNAMESIZE)
494 return ERR_PTR(-ENAMETOOLONG);
496 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
497 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
498 dir == sb->s_root->d_inode)) {
503 if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
504 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
506 return ERR_PTR(error);
511 error = gfs2_permission(dir, MAY_EXEC);
516 inode = gfs2_dir_search(dir, name);
518 error = PTR_ERR(inode);
521 gfs2_glock_dq_uninit(&d_gh);
522 if (error == -ENOENT)
524 return inode ? inode : ERR_PTR(error);
527 static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf)
529 const struct gfs2_inum_range *str = buf;
531 ir->ir_start = be64_to_cpu(str->ir_start);
532 ir->ir_length = be64_to_cpu(str->ir_length);
535 static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf)
537 struct gfs2_inum_range *str = buf;
539 str->ir_start = cpu_to_be64(ir->ir_start);
540 str->ir_length = cpu_to_be64(ir->ir_length);
543 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
545 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
546 struct buffer_head *bh;
547 struct gfs2_inum_range_host ir;
550 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
553 mutex_lock(&sdp->sd_inum_mutex);
555 error = gfs2_meta_inode_buffer(ip, &bh);
557 mutex_unlock(&sdp->sd_inum_mutex);
562 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
565 *formal_ino = ir.ir_start++;
567 gfs2_trans_add_bh(ip->i_gl, bh, 1);
568 gfs2_inum_range_out(&ir,
569 bh->b_data + sizeof(struct gfs2_dinode));
571 mutex_unlock(&sdp->sd_inum_mutex);
578 mutex_unlock(&sdp->sd_inum_mutex);
584 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
586 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
587 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
588 struct gfs2_holder gh;
589 struct buffer_head *bh;
590 struct gfs2_inum_range_host ir;
593 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
597 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
600 mutex_lock(&sdp->sd_inum_mutex);
602 error = gfs2_meta_inode_buffer(ip, &bh);
606 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
609 struct buffer_head *m_bh;
613 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
617 z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
618 x = y = be64_to_cpu(z);
620 ir.ir_length = GFS2_INUM_QUANTUM;
621 x += GFS2_INUM_QUANTUM;
623 gfs2_consist_inode(m_ip);
625 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
626 *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z;
631 *formal_ino = ir.ir_start++;
634 gfs2_trans_add_bh(ip->i_gl, bh, 1);
635 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
640 mutex_unlock(&sdp->sd_inum_mutex);
643 gfs2_glock_dq_uninit(&gh);
647 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
651 error = pick_formal_ino_1(sdp, inum);
655 error = pick_formal_ino_2(sdp, inum);
661 * create_ok - OK to create a new on-disk inode here?
662 * @dip: Directory in which dinode is to be created
663 * @name: Name of new dinode
669 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
674 error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
678 /* Don't create entries in an unlinked directory */
679 if (!dip->i_inode.i_nlink)
682 error = gfs2_dir_check(&dip->i_inode, name, NULL);
693 if (dip->i_di.di_entries == (u32)-1)
695 if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
701 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
702 unsigned int *uid, unsigned int *gid)
704 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
705 (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
708 else if (dip->i_inode.i_uid != current->fsuid)
710 *uid = dip->i_inode.i_uid;
712 *uid = current->fsuid;
714 if (dip->i_inode.i_mode & S_ISGID) {
717 *gid = dip->i_inode.i_gid;
719 *gid = current->fsgid;
722 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
724 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
727 if (gfs2_alloc_get(dip) == NULL)
730 dip->i_alloc->al_requested = RES_DINODE;
731 error = gfs2_inplace_reserve(dip);
735 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
739 *no_addr = gfs2_alloc_di(dip, generation);
744 gfs2_inplace_release(dip);
751 * init_dinode - Fill in a new dinode structure
752 * @dip: the directory this inode is being created in
753 * @gl: The glock covering the new inode
754 * @inum: the inode number
755 * @mode: the file permissions
761 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
762 const struct gfs2_inum_host *inum, unsigned int mode,
763 unsigned int uid, unsigned int gid,
764 const u64 *generation, dev_t dev, struct buffer_head **bhp)
766 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
767 struct gfs2_dinode *di;
768 struct buffer_head *dibh;
769 struct timespec tv = CURRENT_TIME;
771 dibh = gfs2_meta_new(gl, inum->no_addr);
772 gfs2_trans_add_bh(gl, dibh, 1);
773 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
774 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
775 di = (struct gfs2_dinode *)dibh->b_data;
777 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
778 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
779 di->di_mode = cpu_to_be32(mode);
780 di->di_uid = cpu_to_be32(uid);
781 di->di_gid = cpu_to_be32(gid);
784 di->di_blocks = cpu_to_be64(1);
785 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
786 di->di_major = cpu_to_be32(MAJOR(dev));
787 di->di_minor = cpu_to_be32(MINOR(dev));
788 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
789 di->di_generation = cpu_to_be64(*generation);
793 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
794 gfs2_tune_get(sdp, gt_new_files_jdata))
795 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
796 } else if (S_ISDIR(mode)) {
797 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
798 GFS2_DIF_INHERIT_JDATA);
802 di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
808 memset(&di->__pad4, 0, sizeof(di->__pad4));
810 di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
811 di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
812 di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
813 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
815 set_buffer_uptodate(dibh);
820 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
821 unsigned int mode, const struct gfs2_inum_host *inum,
822 const u64 *generation, dev_t dev, struct buffer_head **bhp)
824 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
825 unsigned int uid, gid;
828 munge_mode_uid_gid(dip, &mode, &uid, &gid);
829 if (!gfs2_alloc_get(dip))
832 error = gfs2_quota_lock(dip, uid, gid);
836 error = gfs2_quota_check(dip, uid, gid);
840 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
844 init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
845 gfs2_quota_change(dip, +1, uid, gid);
849 gfs2_quota_unlock(dip);
855 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
856 struct gfs2_inode *ip)
858 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
859 struct gfs2_alloc *al;
861 struct buffer_head *dibh;
864 al = gfs2_alloc_get(dip);
868 error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
872 error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
873 if (alloc_required < 0)
874 goto fail_quota_locks;
875 if (alloc_required) {
876 error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
878 goto fail_quota_locks;
880 al->al_requested = sdp->sd_max_dirres;
882 error = gfs2_inplace_reserve(dip);
884 goto fail_quota_locks;
886 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
887 al->al_rgd->rd_length +
889 RES_STATFS + RES_QUOTA, 0);
893 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
895 goto fail_quota_locks;
898 error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
902 error = gfs2_meta_inode_buffer(ip, &dibh);
905 ip->i_inode.i_nlink = 1;
906 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
907 gfs2_dinode_out(ip, dibh->b_data);
915 if (dip->i_alloc->al_rgd)
916 gfs2_inplace_release(dip);
919 gfs2_quota_unlock(dip);
926 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip)
932 struct gfs2_ea_request er;
934 err = security_inode_init_security(&ip->i_inode, &dip->i_inode,
935 &name, &value, &len);
938 if (err == -EOPNOTSUPP)
943 memset(&er, 0, sizeof(struct gfs2_ea_request));
945 er.er_type = GFS2_EATYPE_SECURITY;
948 er.er_name_len = strlen(name);
949 er.er_data_len = len;
951 err = gfs2_ea_set_i(ip, &er);
960 * gfs2_createi - Create a new inode
961 * @ghs: An array of two holders
962 * @name: The name of the new file
963 * @mode: the permissions on the new inode
965 * @ghs[0] is an initialized holder for the directory
966 * @ghs[1] is the holder for the inode lock
968 * If the return value is not NULL, the glocks on both the directory and the new
969 * file are held. A transaction has been started and an inplace reservation
975 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
976 unsigned int mode, dev_t dev)
978 struct inode *inode = NULL;
979 struct gfs2_inode *dip = ghs->gh_gl->gl_object;
980 struct inode *dir = &dip->i_inode;
981 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
982 struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
985 struct buffer_head *bh = NULL;
987 if (!name->len || name->len > GFS2_FNAMESIZE)
988 return ERR_PTR(-ENAMETOOLONG);
990 gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
991 error = gfs2_glock_nq(ghs);
995 error = create_ok(dip, name, mode);
999 error = pick_formal_ino(sdp, &inum.no_formal_ino);
1003 error = alloc_dinode(dip, &inum.no_addr, &generation);
1007 error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
1008 LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
1012 error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
1016 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode),
1018 inum.no_formal_ino, 0);
1022 error = gfs2_inode_refresh(GFS2_I(inode));
1026 error = gfs2_acl_create(dip, GFS2_I(inode));
1030 error = gfs2_security_init(dip, GFS2_I(inode));
1034 error = link_dinode(dip, name, GFS2_I(inode));
1043 gfs2_glock_dq_uninit(ghs + 1);
1044 if (inode && !IS_ERR(inode))
1051 return ERR_PTR(error);
1055 * gfs2_rmdiri - Remove a directory
1056 * @dip: The parent directory of the directory to be removed
1057 * @name: The name of the directory to be removed
1058 * @ip: The GFS2 inode of the directory to be removed
1060 * Assumes Glocks on dip and ip are held
1065 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
1066 struct gfs2_inode *ip)
1068 struct qstr dotname;
1071 if (ip->i_di.di_entries != 2) {
1072 if (gfs2_consist_inode(ip))
1073 gfs2_dinode_print(ip);
1077 error = gfs2_dir_del(dip, name);
1081 error = gfs2_change_nlink(dip, -1);
1085 gfs2_str2qstr(&dotname, ".");
1086 error = gfs2_dir_del(ip, &dotname);
1090 gfs2_str2qstr(&dotname, "..");
1091 error = gfs2_dir_del(ip, &dotname);
1095 /* It looks odd, but it really should be done twice */
1096 error = gfs2_change_nlink(ip, -1);
1100 error = gfs2_change_nlink(ip, -1);
1108 * gfs2_unlink_ok - check to see that a inode is still in a directory
1109 * @dip: the directory
1110 * @name: the name of the file
1113 * Assumes that the lock on (at least) @dip is held.
1115 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1118 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1119 const struct gfs2_inode *ip)
1123 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1126 if ((dip->i_inode.i_mode & S_ISVTX) &&
1127 dip->i_inode.i_uid != current->fsuid &&
1128 ip->i_inode.i_uid != current->fsuid && !capable(CAP_FOWNER))
1131 if (IS_APPEND(&dip->i_inode))
1134 error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
1138 error = gfs2_dir_check(&dip->i_inode, name, ip);
1146 * gfs2_readlinki - return the contents of a symlink
1147 * @ip: the symlink's inode
1148 * @buf: a pointer to the buffer to be filled
1149 * @len: a pointer to the length of @buf
1151 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1152 * to be freed by the caller.
1157 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1159 struct gfs2_holder i_gh;
1160 struct buffer_head *dibh;
1164 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &i_gh);
1165 error = gfs2_glock_nq(&i_gh);
1167 gfs2_holder_uninit(&i_gh);
1171 if (!ip->i_di.di_size) {
1172 gfs2_consist_inode(ip);
1177 error = gfs2_meta_inode_buffer(ip, &dibh);
1181 x = ip->i_di.di_size + 1;
1183 *buf = kmalloc(x, GFP_NOFS);
1190 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1196 gfs2_glock_dq_uninit(&i_gh);
1201 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1203 struct buffer_head *dibh;
1206 error = gfs2_meta_inode_buffer(ip, &dibh);
1208 error = inode_setattr(&ip->i_inode, attr);
1209 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1210 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1211 gfs2_dinode_out(ip, dibh->b_data);
1218 * gfs2_setattr_simple -
1222 * Called with a reference on the vnode.
1227 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1231 if (current->journal_info)
1232 return __gfs2_setattr_simple(ip, attr);
1234 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1238 error = __gfs2_setattr_simple(ip, attr);
1239 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1243 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1245 const struct gfs2_dinode_host *di = &ip->i_di;
1246 struct gfs2_dinode *str = buf;
1248 str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1249 str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1250 str->di_header.__pad0 = 0;
1251 str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1252 str->di_header.__pad1 = 0;
1253 str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1254 str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1255 str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1256 str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1257 str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1258 str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1259 str->di_size = cpu_to_be64(di->di_size);
1260 str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
1261 str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1262 str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1263 str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1265 str->di_goal_meta = cpu_to_be64(ip->i_goal);
1266 str->di_goal_data = cpu_to_be64(ip->i_goal);
1267 str->di_generation = cpu_to_be64(di->di_generation);
1269 str->di_flags = cpu_to_be32(di->di_flags);
1270 str->di_height = cpu_to_be16(ip->i_height);
1271 str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1272 !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
1273 GFS2_FORMAT_DE : 0);
1274 str->di_depth = cpu_to_be16(ip->i_depth);
1275 str->di_entries = cpu_to_be32(di->di_entries);
1277 str->di_eattr = cpu_to_be64(di->di_eattr);
1278 str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1279 str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1280 str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1283 void gfs2_dinode_print(const struct gfs2_inode *ip)
1285 const struct gfs2_dinode_host *di = &ip->i_di;
1287 printk(KERN_INFO " no_formal_ino = %llu\n",
1288 (unsigned long long)ip->i_no_formal_ino);
1289 printk(KERN_INFO " no_addr = %llu\n",
1290 (unsigned long long)ip->i_no_addr);
1291 printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size);
1292 printk(KERN_INFO " blocks = %llu\n",
1293 (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
1294 printk(KERN_INFO " i_goal = %llu\n",
1295 (unsigned long long)ip->i_goal);
1296 printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags);
1297 printk(KERN_INFO " i_height = %u\n", ip->i_height);
1298 printk(KERN_INFO " i_depth = %u\n", ip->i_depth);
1299 printk(KERN_INFO " di_entries = %u\n", di->di_entries);
1300 printk(KERN_INFO " di_eattr = %llu\n",
1301 (unsigned long long)di->di_eattr);