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 && 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;
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);
357 error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
361 error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
365 rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr);
367 gfs2_consist_inode(ip);
369 goto out_rindex_relse;
372 error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0,
375 goto out_rindex_relse;
377 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1);
381 set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
382 set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags);
384 gfs2_free_di(rgd, ip);
387 clear_bit(GLF_STICKY, &ip->i_gl->gl_flags);
390 gfs2_glock_dq_uninit(&al->al_rgd_gh);
392 gfs2_glock_dq_uninit(&al->al_ri_gh);
394 gfs2_quota_unhold(ip);
401 * gfs2_change_nlink - Change nlink count on inode
402 * @ip: The GFS2 inode
403 * @diff: The change in the nlink count required
407 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
409 struct buffer_head *dibh;
413 BUG_ON(diff != 1 && diff != -1);
414 nlink = ip->i_inode.i_nlink + diff;
416 /* If we are reducing the nlink count, but the new value ends up being
417 bigger than the old one, we must have underflowed. */
418 if (diff < 0 && nlink > ip->i_inode.i_nlink) {
419 if (gfs2_consist_inode(ip))
420 gfs2_dinode_print(ip);
424 error = gfs2_meta_inode_buffer(ip, &dibh);
429 inc_nlink(&ip->i_inode);
431 drop_nlink(&ip->i_inode);
433 ip->i_inode.i_ctime = CURRENT_TIME;
435 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
436 gfs2_dinode_out(ip, dibh->b_data);
438 mark_inode_dirty(&ip->i_inode);
440 if (ip->i_inode.i_nlink == 0)
441 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
446 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
450 gfs2_str2qstr(&qstr, name);
451 inode = gfs2_lookupi(dip, &qstr, 1);
452 /* gfs2_lookupi has inconsistent callers: vfs
453 * related routines expect NULL for no entry found,
454 * gfs2_lookup_simple callers expect ENOENT
455 * and do not check for NULL.
458 return ERR_PTR(-ENOENT);
465 * gfs2_lookupi - Look up a filename in a directory and return its inode
466 * @d_gh: An initialized holder for the directory glock
467 * @name: The name of the inode to look for
468 * @is_root: If 1, ignore the caller's permissions
469 * @i_gh: An uninitialized holder for the new inode glock
471 * This can be called via the VFS filldir function when NFS is doing
472 * a readdirplus and the inode which its intending to stat isn't
473 * already in cache. In this case we must not take the directory glock
474 * again, since the readdir call will have already taken that lock.
479 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
482 struct super_block *sb = dir->i_sb;
483 struct gfs2_inode *dip = GFS2_I(dir);
484 struct gfs2_holder d_gh;
486 struct inode *inode = NULL;
489 if (!name->len || name->len > GFS2_FNAMESIZE)
490 return ERR_PTR(-ENAMETOOLONG);
492 if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
493 (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
494 dir == sb->s_root->d_inode)) {
499 if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
500 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
502 return ERR_PTR(error);
507 error = gfs2_permission(dir, MAY_EXEC);
512 inode = gfs2_dir_search(dir, name);
514 error = PTR_ERR(inode);
517 gfs2_glock_dq_uninit(&d_gh);
518 if (error == -ENOENT)
520 return inode ? inode : ERR_PTR(error);
523 static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf)
525 const struct gfs2_inum_range *str = buf;
527 ir->ir_start = be64_to_cpu(str->ir_start);
528 ir->ir_length = be64_to_cpu(str->ir_length);
531 static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf)
533 struct gfs2_inum_range *str = buf;
535 str->ir_start = cpu_to_be64(ir->ir_start);
536 str->ir_length = cpu_to_be64(ir->ir_length);
539 static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino)
541 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
542 struct buffer_head *bh;
543 struct gfs2_inum_range_host ir;
546 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
549 mutex_lock(&sdp->sd_inum_mutex);
551 error = gfs2_meta_inode_buffer(ip, &bh);
553 mutex_unlock(&sdp->sd_inum_mutex);
558 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
561 *formal_ino = ir.ir_start++;
563 gfs2_trans_add_bh(ip->i_gl, bh, 1);
564 gfs2_inum_range_out(&ir,
565 bh->b_data + sizeof(struct gfs2_dinode));
567 mutex_unlock(&sdp->sd_inum_mutex);
574 mutex_unlock(&sdp->sd_inum_mutex);
580 static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino)
582 struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode);
583 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode);
584 struct gfs2_holder gh;
585 struct buffer_head *bh;
586 struct gfs2_inum_range_host ir;
589 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
593 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
596 mutex_lock(&sdp->sd_inum_mutex);
598 error = gfs2_meta_inode_buffer(ip, &bh);
602 gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode));
605 struct buffer_head *m_bh;
609 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
613 z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode));
614 x = y = be64_to_cpu(z);
616 ir.ir_length = GFS2_INUM_QUANTUM;
617 x += GFS2_INUM_QUANTUM;
619 gfs2_consist_inode(m_ip);
621 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
622 *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z;
627 *formal_ino = ir.ir_start++;
630 gfs2_trans_add_bh(ip->i_gl, bh, 1);
631 gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode));
636 mutex_unlock(&sdp->sd_inum_mutex);
639 gfs2_glock_dq_uninit(&gh);
643 static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum)
647 error = pick_formal_ino_1(sdp, inum);
651 error = pick_formal_ino_2(sdp, inum);
657 * create_ok - OK to create a new on-disk inode here?
658 * @dip: Directory in which dinode is to be created
659 * @name: Name of new dinode
665 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
670 error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
674 /* Don't create entries in an unlinked directory */
675 if (!dip->i_inode.i_nlink)
678 error = gfs2_dir_check(&dip->i_inode, name, NULL);
689 if (dip->i_di.di_entries == (u32)-1)
691 if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
697 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
698 unsigned int *uid, unsigned int *gid)
700 if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
701 (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
704 else if (dip->i_inode.i_uid != current->fsuid)
706 *uid = dip->i_inode.i_uid;
708 *uid = current->fsuid;
710 if (dip->i_inode.i_mode & S_ISGID) {
713 *gid = dip->i_inode.i_gid;
715 *gid = current->fsgid;
718 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
720 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
723 if (gfs2_alloc_get(dip) == NULL)
726 dip->i_alloc->al_requested = RES_DINODE;
727 error = gfs2_inplace_reserve(dip);
731 error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
735 *no_addr = gfs2_alloc_di(dip, generation);
740 gfs2_inplace_release(dip);
747 * init_dinode - Fill in a new dinode structure
748 * @dip: the directory this inode is being created in
749 * @gl: The glock covering the new inode
750 * @inum: the inode number
751 * @mode: the file permissions
757 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
758 const struct gfs2_inum_host *inum, unsigned int mode,
759 unsigned int uid, unsigned int gid,
760 const u64 *generation, dev_t dev, struct buffer_head **bhp)
762 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
763 struct gfs2_dinode *di;
764 struct buffer_head *dibh;
765 struct timespec tv = CURRENT_TIME;
767 dibh = gfs2_meta_new(gl, inum->no_addr);
768 gfs2_trans_add_bh(gl, dibh, 1);
769 gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
770 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
771 di = (struct gfs2_dinode *)dibh->b_data;
773 di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
774 di->di_num.no_addr = cpu_to_be64(inum->no_addr);
775 di->di_mode = cpu_to_be32(mode);
776 di->di_uid = cpu_to_be32(uid);
777 di->di_gid = cpu_to_be32(gid);
780 di->di_blocks = cpu_to_be64(1);
781 di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
782 di->di_major = cpu_to_be32(MAJOR(dev));
783 di->di_minor = cpu_to_be32(MINOR(dev));
784 di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
785 di->di_generation = cpu_to_be64(*generation);
789 if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) ||
790 gfs2_tune_get(sdp, gt_new_files_jdata))
791 di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
792 } else if (S_ISDIR(mode)) {
793 di->di_flags |= cpu_to_be32(dip->i_di.di_flags &
794 GFS2_DIF_INHERIT_JDATA);
798 di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
804 memset(&di->__pad4, 0, sizeof(di->__pad4));
806 di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
807 di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
808 di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
809 memset(&di->di_reserved, 0, sizeof(di->di_reserved));
811 set_buffer_uptodate(dibh);
816 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
817 unsigned int mode, const struct gfs2_inum_host *inum,
818 const u64 *generation, dev_t dev, struct buffer_head **bhp)
820 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
821 unsigned int uid, gid;
824 munge_mode_uid_gid(dip, &mode, &uid, &gid);
825 if (!gfs2_alloc_get(dip))
828 error = gfs2_quota_lock(dip, uid, gid);
832 error = gfs2_quota_check(dip, uid, gid);
836 error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
840 init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
841 gfs2_quota_change(dip, +1, uid, gid);
845 gfs2_quota_unlock(dip);
851 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
852 struct gfs2_inode *ip)
854 struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
855 struct gfs2_alloc *al;
857 struct buffer_head *dibh;
860 al = gfs2_alloc_get(dip);
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));
1039 gfs2_glock_dq_uninit(ghs + 1);
1040 if (inode && !IS_ERR(inode))
1047 return ERR_PTR(error);
1051 * gfs2_rmdiri - Remove a directory
1052 * @dip: The parent directory of the directory to be removed
1053 * @name: The name of the directory to be removed
1054 * @ip: The GFS2 inode of the directory to be removed
1056 * Assumes Glocks on dip and ip are held
1061 int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name,
1062 struct gfs2_inode *ip)
1064 struct qstr dotname;
1067 if (ip->i_di.di_entries != 2) {
1068 if (gfs2_consist_inode(ip))
1069 gfs2_dinode_print(ip);
1073 error = gfs2_dir_del(dip, name);
1077 error = gfs2_change_nlink(dip, -1);
1081 gfs2_str2qstr(&dotname, ".");
1082 error = gfs2_dir_del(ip, &dotname);
1086 gfs2_str2qstr(&dotname, "..");
1087 error = gfs2_dir_del(ip, &dotname);
1091 /* It looks odd, but it really should be done twice */
1092 error = gfs2_change_nlink(ip, -1);
1096 error = gfs2_change_nlink(ip, -1);
1104 * gfs2_unlink_ok - check to see that a inode is still in a directory
1105 * @dip: the directory
1106 * @name: the name of the file
1109 * Assumes that the lock on (at least) @dip is held.
1111 * Returns: 0 if the parent/child relationship is correct, errno if it isn't
1114 int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
1115 const struct gfs2_inode *ip)
1119 if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode))
1122 if ((dip->i_inode.i_mode & S_ISVTX) &&
1123 dip->i_inode.i_uid != current->fsuid &&
1124 ip->i_inode.i_uid != current->fsuid && !capable(CAP_FOWNER))
1127 if (IS_APPEND(&dip->i_inode))
1130 error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
1134 error = gfs2_dir_check(&dip->i_inode, name, ip);
1142 * gfs2_readlinki - return the contents of a symlink
1143 * @ip: the symlink's inode
1144 * @buf: a pointer to the buffer to be filled
1145 * @len: a pointer to the length of @buf
1147 * If @buf is too small, a piece of memory is kmalloc()ed and needs
1148 * to be freed by the caller.
1153 int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
1155 struct gfs2_holder i_gh;
1156 struct buffer_head *dibh;
1160 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
1161 error = gfs2_glock_nq_atime(&i_gh);
1163 gfs2_holder_uninit(&i_gh);
1167 if (!ip->i_di.di_size) {
1168 gfs2_consist_inode(ip);
1173 error = gfs2_meta_inode_buffer(ip, &dibh);
1177 x = ip->i_di.di_size + 1;
1179 *buf = kmalloc(x, GFP_NOFS);
1186 memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x);
1192 gfs2_glock_dq_uninit(&i_gh);
1197 * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and
1198 * conditionally update the inode's atime
1199 * @gh: the holder to acquire
1201 * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap
1202 * Update if the difference between the current time and the inode's current
1203 * atime is greater than an interval specified at mount.
1208 int gfs2_glock_nq_atime(struct gfs2_holder *gh)
1210 struct gfs2_glock *gl = gh->gh_gl;
1211 struct gfs2_sbd *sdp = gl->gl_sbd;
1212 struct gfs2_inode *ip = gl->gl_object;
1213 s64 quantum = gfs2_tune_get(sdp, gt_atime_quantum);
1217 struct timespec tv = CURRENT_TIME;
1219 if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) ||
1220 gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) ||
1221 gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops))
1224 state = gh->gh_state;
1225 flags = gh->gh_flags;
1227 error = gfs2_glock_nq(gh);
1231 if (test_bit(SDF_NOATIME, &sdp->sd_flags) ||
1232 (sdp->sd_vfs->s_flags & MS_RDONLY))
1235 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1237 gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY,
1239 error = gfs2_glock_nq(gh);
1243 /* Verify that atime hasn't been updated while we were
1244 trying to get exclusive lock. */
1247 if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) {
1248 struct buffer_head *dibh;
1249 struct gfs2_dinode *di;
1251 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
1252 if (error == -EROFS)
1257 error = gfs2_meta_inode_buffer(ip, &dibh);
1259 goto fail_end_trans;
1261 ip->i_inode.i_atime = tv;
1263 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1264 di = (struct gfs2_dinode *)dibh->b_data;
1265 di->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1266 di->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1269 gfs2_trans_end(sdp);
1272 /* If someone else has asked for the glock,
1273 unlock and let them have it. Then reacquire
1274 in the original state. */
1275 if (gfs2_glock_is_blocking(gl)) {
1277 gfs2_holder_reinit(state, flags, gh);
1278 return gfs2_glock_nq(gh);
1285 gfs2_trans_end(sdp);
1292 __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1294 struct buffer_head *dibh;
1297 error = gfs2_meta_inode_buffer(ip, &dibh);
1299 error = inode_setattr(&ip->i_inode, attr);
1300 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
1301 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1302 gfs2_dinode_out(ip, dibh->b_data);
1309 * gfs2_setattr_simple -
1313 * Called with a reference on the vnode.
1318 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
1322 if (current->journal_info)
1323 return __gfs2_setattr_simple(ip, attr);
1325 error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
1329 error = __gfs2_setattr_simple(ip, attr);
1330 gfs2_trans_end(GFS2_SB(&ip->i_inode));
1334 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
1336 const struct gfs2_dinode_host *di = &ip->i_di;
1337 struct gfs2_dinode *str = buf;
1339 str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
1340 str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
1341 str->di_header.__pad0 = 0;
1342 str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
1343 str->di_header.__pad1 = 0;
1344 str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
1345 str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
1346 str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
1347 str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
1348 str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
1349 str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
1350 str->di_size = cpu_to_be64(di->di_size);
1351 str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
1352 str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
1353 str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
1354 str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
1356 str->di_goal_meta = cpu_to_be64(ip->i_goal);
1357 str->di_goal_data = cpu_to_be64(ip->i_goal);
1358 str->di_generation = cpu_to_be64(di->di_generation);
1360 str->di_flags = cpu_to_be32(di->di_flags);
1361 str->di_height = cpu_to_be16(ip->i_height);
1362 str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
1363 !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
1364 GFS2_FORMAT_DE : 0);
1365 str->di_depth = cpu_to_be16(ip->i_depth);
1366 str->di_entries = cpu_to_be32(di->di_entries);
1368 str->di_eattr = cpu_to_be64(di->di_eattr);
1369 str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
1370 str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
1371 str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
1374 void gfs2_dinode_print(const struct gfs2_inode *ip)
1376 const struct gfs2_dinode_host *di = &ip->i_di;
1378 printk(KERN_INFO " no_formal_ino = %llu\n",
1379 (unsigned long long)ip->i_no_formal_ino);
1380 printk(KERN_INFO " no_addr = %llu\n",
1381 (unsigned long long)ip->i_no_addr);
1382 printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size);
1383 printk(KERN_INFO " blocks = %llu\n",
1384 (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
1385 printk(KERN_INFO " i_goal = %llu\n",
1386 (unsigned long long)ip->i_goal);
1387 printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags);
1388 printk(KERN_INFO " i_height = %u\n", ip->i_height);
1389 printk(KERN_INFO " i_depth = %u\n", ip->i_depth);
1390 printk(KERN_INFO " di_entries = %u\n", di->di_entries);
1391 printk(KERN_INFO " di_eattr = %llu\n",
1392 (unsigned long long)di->di_eattr);