2 * Copyright (c) 2013 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Antonio Huete Jimenez <tuxillo@quantumachine.net>
6 * by Matthew Dillon <dillon@dragonflybsd.org>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * 3. Neither the name of The DragonFly Project nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific, prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * See below a small table with the vnode operation and syscall correspondence
41 * VNODE OP SCALL SCALL_AT FD PATH COMMENTS
42 * dirfs_ncreate Y Y Y Y open(2), openat(2)
43 * dirfs_nresolve - - - Y no syscall needed
44 * dirfs_nlookupdot - - - - -
45 * dirfs_nmknod Y Y Y Y mknod(2), mknodat(2)
46 * dirfs_open Y Y Y Y open(2), openat(2)
47 * dirfs_close Y Y Y Y close(2)
48 * dirfs_access - - - - data from stat(2)
49 * dirfs_getattr Y Y Y Y lstat(2), fstatat(2)
50 * dirfs_setattr - - - - -
51 * dirfs_read Y - Y - read(2).
52 * dirfs_write Y - Y - write(2).
53 * dirfs_fsync Y - Y - fsync(2)
54 * dirfs_mountctl - - - - -
55 * dirfs_nremove Y - - Y unlink(2)
56 * dirfs_nlink - - - - -
57 * dirfs_nrename Y Y Y Y rename(2), renameat(2)
58 * dirfs_nmkdir Y Y Y Y mkdir(2), mkdirat(2)
59 * dirfs_nrmdir Y - - Y rmdir(2)
60 * dirfs_nsymlink Y Y Y Y symlink(2), symlinkat(2)
61 * dirfs_readdir Y - Y - getdirentries(2)
62 * dirfs_readlink Y Y Y Y readlinkat(2)
63 * dirfs_inactive - - - - -
64 * dirfs_reclaim - - - - -
65 * dirfs_print - - - - -
66 * dirfs_pathconf - - - - -
67 * dirfs_bmap - - - - -
68 * dirfs_strategy Y - Y - pwrite(2), pread(2)
69 * dirfs_advlock - - - - -
70 * dirfs_kqfilter - - - - -
78 #include <sys/vfsops.h>
79 #include <sys/vnode.h>
81 #include <sys/namecache.h>
82 #include <sys/queue.h>
83 #include <sys/systm.h>
84 #include <sys/dirent.h>
85 #include <sys/mount.h>
86 #include <sys/signalvar.h>
87 #include <sys/resource.h>
89 #include <sys/kern_syscall.h>
95 * Kernel tracing facilities
97 KTR_INFO_MASTER_EXTERN(dirfs);
99 KTR_INFO(KTR_DIRFS, dirfs, unsupported, 0,
103 KTR_INFO(KTR_DIRFS, dirfs, nresolve, 0,
104 "DIRFS(dnp=%p ncp_name=%s parent=%p pfd=%d error=%d)",
105 dirfs_node_t dnp, char *name, dirfs_node_t pdnp, int pfd, int error);
107 KTR_INFO(KTR_DIRFS, dirfs, ncreate, 1,
108 "DIRFS(dnp=%p ncp_name=%s parent=%p pfd=%d error=%d)",
109 dirfs_node_t dnp, char *name, dirfs_node_t pdnp, int pfd, int error);
111 KTR_INFO(KTR_DIRFS, dirfs, open, 2,
112 "DIRFS(dnp=%p dn_name=%s nfd=%d)",
113 dirfs_node_t dnp, char *name, int fd);
115 KTR_INFO(KTR_DIRFS, dirfs, close, 3,
116 "DIRFS(dnp=%p fd=%d opencount=%d writecount=%d vfsync error=%d)",
117 dirfs_node_t dnp, int fd, int oc, int wc, int error);
119 KTR_INFO(KTR_DIRFS, dirfs, readdir, 4,
120 "DIRFS(dnp=%p fd=%d startoff=%jd uio_offset=%jd)",
121 dirfs_node_t dnp, int fd, off_t startoff, off_t uoff);
123 KTR_INFO(KTR_DIRFS, dirfs, access, 5,
124 "DIRFS(dnp=%p error=%d)",
125 dirfs_node_t dnp, int error);
127 KTR_INFO(KTR_DIRFS, dirfs, getattr, 6,
128 "DIRFS(dnp=%p error=%d)",
129 dirfs_node_t dnp, int error);
131 KTR_INFO(KTR_DIRFS, dirfs, setattr, 7,
132 "DIRFS(dnp=%p action=%s error=%d)",
133 dirfs_node_t dnp, const char *action, int error);
135 KTR_INFO(KTR_DIRFS, dirfs, fsync, 8,
136 "DIRFS(dnp=%p error=%d)",
137 dirfs_node_t dnp, int error);
139 KTR_INFO(KTR_DIRFS, dirfs, read, 9,
140 "DIRFS(dnp=%p size=%jd error=%d)",
141 dirfs_node_t dnp, size_t size, int error);
143 KTR_INFO(KTR_DIRFS, dirfs, write, 10,
144 "DIRFS(dnp=%p size=%jd boff=%jd uio_resid=%jd error=%d)",
145 dirfs_node_t dnp, off_t boff, size_t resid, size_t size, int error);
147 KTR_INFO(KTR_DIRFS, dirfs, strategy, 11,
148 "DIRFS(dnp=%p dnp_size=%jd iosize=%jd b_cmd=%d b_error=%d "
149 "b_resid=%d bio_off=%jd error=%d)",
150 dirfs_node_t dnp, size_t size, size_t iosize, int cmd, int berror,
151 int bresid, off_t biooff, int error);
153 KTR_INFO(KTR_DIRFS, dirfs, nremove, 12,
154 "DIRFS(dnp=%p pdnp=%p error=%d)",
155 dirfs_node_t dnp, dirfs_node_t pdnp, int error);
157 KTR_INFO(KTR_DIRFS, dirfs, nmkdir, 13,
158 "DIRFS(pdnp=%p dnp=%p nc_name=%p error=%d)",
159 dirfs_node_t dnp, dirfs_node_t pdnp, char *n, int error);
161 KTR_INFO(KTR_DIRFS, dirfs, nrmdir, 13,
162 "DIRFS(pdnp=%p dnp=%p error=%d)",
163 dirfs_node_t dnp, dirfs_node_t pdnp, int error);
165 KTR_INFO(KTR_DIRFS, dirfs, nsymlink, 14,
166 "DIRFS(dnp=%p target=%s symlink=%s error=%d)",
167 dirfs_node_t dnp, char *tgt, char *lnk, int error);
169 /* Needed prototypes */
170 int dirfs_access(struct vop_access_args *);
171 int dirfs_getattr(struct vop_getattr_args *);
172 int dirfs_setattr(struct vop_setattr_args *);
173 int dirfs_reclaim(struct vop_reclaim_args *);
176 dirfs_nresolve(struct vop_nresolve_args *ap)
178 dirfs_node_t pdnp, dnp, d1, d2;
180 struct namecache *ncp;
181 struct nchandle *nch;
195 dnp = d1 = d2 = NULL;
196 pdnp = VP_TO_NODE(dvp);
197 dmp = VFS_TO_DIRFS(mp);
199 dirfs_node_lock(pdnp);
200 TAILQ_FOREACH_MUTABLE(d1, &dmp->dm_fdlist, dn_fdentry, d2) {
201 if (d1->dn_parent == pdnp &&
202 (strcmp(d1->dn_name, ncp->nc_name) == 0)) {
205 passive_fd_list_hits++;
209 dirfs_node_unlock(pdnp);
212 dirfs_alloc_vp(mp, &vp, LK_CANRECURSE, dnp);
213 dirfs_node_drop(dmp, dnp);
215 passive_fd_list_miss++;
216 error = dirfs_alloc_file(dmp, &dnp, pdnp, ncp, &vp, NULL, 0);
220 if (error && error == ENOENT) {
221 cache_setvp(nch, NULL);
224 cache_setvp(nch, vp);
229 KTR_LOG(dirfs_nresolve, dnp, ncp->nc_name, pdnp, pdnp->dn_fd, error);
235 dirfs_nlookupdotdot(struct vop_nlookupdotdot_args *ap)
239 KTR_LOG(dirfs_unsupported, __func__);
245 dirfs_ncreate(struct vop_ncreate_args *ap)
250 struct namecache *ncp;
262 pdnp = VP_TO_NODE(dvp);
263 dmp = VFS_TO_DIRFS(dvp->v_mount);
265 ncp = ap->a_nch->ncp;
268 dirfs_mount_gettoken(dmp);
270 dirfs_node_getperms(pdnp, &perms);
271 if ((perms & DIRFS_NODE_WR) == 0)
274 error = dirfs_alloc_file(dmp, &dnp, pdnp, ncp, vpp, vap,
278 cache_setunresolved(ap->a_nch);
279 cache_setvp(ap->a_nch, *vpp);
282 dirfs_mount_reltoken(dmp);
284 KTR_LOG(dirfs_ncreate, dnp, ncp->nc_name, pdnp, pdnp->dn_fd, error);
290 dirfs_nmknod(struct vop_nmknod_args *v)
298 dirfs_open(struct vop_open_args *ap)
308 dnp = VP_TO_NODE(vp);
309 dmp = VFS_TO_DIRFS(vp->v_mount);
313 * Root inode has been allocated and opened in VFS_ROOT() so
314 * no reason to attempt to open it again.
316 if (dmp->dm_root != dnp && dnp->dn_fd == DIRFS_NOFD) {
317 error = dirfs_open_helper(dmp, dnp, DIRFS_NOFD, NULL);
322 KTR_LOG(dirfs_open, dnp, dnp->dn_name, dnp->dn_fd);
324 return vop_stdopen(ap);
328 dirfs_close(struct vop_close_args *ap)
338 dnp = VP_TO_NODE(vp);
340 if (vp->v_type == VREG) {
341 error = vfsync(vp, 0, 1, NULL, NULL);
343 dbg(9, "vfsync error=%d\n", error);
347 KTR_LOG(dirfs_close, dnp, dnp->dn_fd, vp->v_opencount,
348 vp->v_writecount, error);
354 dirfs_access(struct vop_access_args *ap)
356 struct vnode *vp = ap->a_vp;
362 dnp = VP_TO_NODE(vp);
364 switch (vp->v_type) {
370 if ((ap->a_mode & VWRITE) && (vp->v_mount->mnt_flag & MNT_RDONLY)) {
389 error = vop_helper_access(ap, dnp->dn_uid,
390 dnp->dn_gid, dnp->dn_mode, 0);
393 KTR_LOG(dirfs_access, dnp, error);
399 dirfs_getattr(struct vop_getattr_args *ap)
414 dnp = VP_TO_NODE(vp);
415 dmp = VFS_TO_DIRFS(vp->v_mount);
417 KKASSERT(dnp); /* This must not happen */
419 if (!dirfs_node_isroot(dnp)) {
420 pathnp = dirfs_findfd(dmp, dnp, &tmp, &pathfree);
422 KKASSERT(pathnp->dn_fd != DIRFS_NOFD);
424 error = dirfs_node_stat(pathnp->dn_fd, tmp, dnp);
425 dirfs_dropfd(dmp, pathnp, pathfree);
427 error = dirfs_node_stat(DIRFS_NOFD, dmp->dm_path, dnp);
431 dirfs_node_lock(dnp);
432 vap->va_nlink = dnp->dn_links;
433 vap->va_type = dnp->dn_type;
434 vap->va_mode = dnp->dn_mode;
435 vap->va_uid = dnp->dn_uid;
436 vap->va_gid = dnp->dn_gid;
437 vap->va_fileid = dnp->dn_ino;
438 vap->va_size = dnp->dn_size;
439 vap->va_blocksize = dnp->dn_blocksize;
440 vap->va_atime.tv_sec = dnp->dn_atime;
441 vap->va_atime.tv_nsec = dnp->dn_atimensec;
442 vap->va_mtime.tv_sec = dnp->dn_mtime;
443 vap->va_mtime.tv_nsec = dnp->dn_mtimensec;
444 vap->va_ctime.tv_sec = dnp->dn_ctime;
445 vap->va_ctime.tv_nsec = dnp->dn_ctimensec;
446 vap->va_bytes = dnp->dn_size;
447 vap->va_gen = dnp->dn_gen;
448 vap->va_flags = dnp->dn_flags;
449 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
450 dirfs_node_unlock(dnp);
453 KTR_LOG(dirfs_getattr, dnp, error);
459 dirfs_setattr(struct vop_setattr_args *ap)
468 const char *msg[6] = {
485 dnp = VP_TO_NODE(vp);
486 dmp = VFS_TO_DIRFS(vp->v_mount);
488 dirfs_mount_gettoken(dmp);
491 * Check for unsettable attributes.
493 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
494 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
495 (vap->va_blocksize != VNOVAL) || (vap->va_rmajor != VNOVAL) ||
496 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
505 if (error == 0 && (vap->va_flags != VNOVAL)) {
506 if (vp->v_mount->mnt_flag & MNT_RDONLY)
509 error = dirfs_node_chflags(dnp, vap->va_flags, cred);
515 * Extend or truncate a file
517 if (error == 0 && (vap->va_size != VNOVAL)) {
518 if (vp->v_mount->mnt_flag & MNT_RDONLY)
521 error = dirfs_node_chsize(dnp, vap->va_size);
522 dbg(9, "dnp size=%jd vap size=%jd\n", dnp->dn_size, vap->va_size);
528 * Change file owner or group
530 if (error == 0 && (vap->va_uid != (uid_t)VNOVAL ||
531 vap->va_gid != (gid_t)VNOVAL)) {
532 if (vp->v_mount->mnt_flag & MNT_RDONLY) {
535 mode_t cur_mode = dnp->dn_mode;
536 uid_t cur_uid = dnp->dn_uid;
537 gid_t cur_gid = dnp->dn_gid;
539 error = vop_helper_chown(ap->a_vp, vap->va_uid,
540 vap->va_gid, ap->a_cred,
541 &cur_uid, &cur_gid, &cur_mode);
543 (cur_mode != dnp->dn_mode ||
544 cur_uid != dnp->dn_uid ||
545 cur_gid != dnp->dn_gid)) {
546 error = dirfs_node_chown(dmp, dnp, cur_uid,
557 if (error == 0 && (vap->va_mode != (mode_t)VNOVAL)) {
558 if (vp->v_mount->mnt_flag & MNT_RDONLY) {
561 mode_t cur_mode = dnp->dn_mode;
562 uid_t cur_uid = dnp->dn_uid;
563 gid_t cur_gid = dnp->dn_gid;
565 error = vop_helper_chmod(ap->a_vp, vap->va_mode,
567 cur_uid, cur_gid, &cur_mode);
568 if (error == 0 && cur_mode != dnp->dn_mode) {
569 error = dirfs_node_chmod(dmp, dnp, cur_mode);
579 if (error == 0 && ((vap->va_atime.tv_sec != VNOVAL &&
580 vap->va_atime.tv_nsec != VNOVAL) ||
581 (vap->va_mtime.tv_sec != VNOVAL &&
582 vap->va_mtime.tv_nsec != VNOVAL) )) {
583 if (vp->v_mount->mnt_flag & MNT_RDONLY)
586 error = dirfs_node_chtimes(dnp);
592 dirfs_mount_reltoken(dmp);
594 KTR_LOG(dirfs_setattr, dnp, msg[msgno], error);
600 dirfs_fsync(struct vop_fsync_args *ap)
602 dirfs_node_t dnp = VP_TO_NODE(ap->a_vp);
607 vfsync(ap->a_vp, ap->a_waitfor, 1, NULL, NULL);
609 if (dnp->dn_fd != DIRFS_NOFD) {
610 if (fsync(dnp->dn_fd) == -1)
611 error = fsync(dnp->dn_fd);
614 KTR_LOG(dirfs_fsync, dnp, error);
620 dirfs_read(struct vop_read_args *ap)
623 struct vnode *vp = ap->a_vp;
624 struct uio *uio = ap->a_uio;
634 if (uio->uio_resid == 0) {
635 dbg(9, "zero len uio->uio_resid\n");
639 dnp = VP_TO_NODE(vp);
641 if (uio->uio_offset < 0)
643 if (vp->v_type != VREG)
646 while (uio->uio_resid > 0 && uio->uio_offset < dnp->dn_size) {
648 * Use buffer cache I/O (via dirfs_strategy)
650 offset = (size_t)uio->uio_offset & BMASK;
651 base_offset = (off_t)uio->uio_offset - offset;
652 bp = getcacheblk(vp, base_offset, BSIZE, 0);
654 lwkt_gettoken(&vp->v_mount->mnt_token);
655 error = bread(vp, base_offset, BSIZE, &bp);
658 lwkt_reltoken(&vp->v_mount->mnt_token);
659 dbg(9, "dirfs_read bread error %d\n", error);
662 lwkt_reltoken(&vp->v_mount->mnt_token);
666 * Figure out how many bytes we can actually copy this loop.
668 len = BSIZE - offset;
669 if (len > uio->uio_resid)
670 len = uio->uio_resid;
671 if (len > dnp->dn_size - uio->uio_offset)
672 len = (size_t)(dnp->dn_size - uio->uio_offset);
674 error = uiomovebp(bp, (char *)bp->b_data + offset, len, uio);
677 dbg(9, "dirfs_read uiomove error %d\n", error);
682 KTR_LOG(dirfs_read, dnp, dnp->dn_size, error);
688 dirfs_write (struct vop_write_args *ap)
693 struct vnode *vp = ap->a_vp;
694 struct uio *uio = ap->a_uio;
695 struct thread *td = uio->uio_td;
707 if (uio->uio_resid == 0) {
708 dbg(9, "zero-length uio->uio_resid\n");
712 dnp = VP_TO_NODE(vp);
713 dmp = VFS_TO_DIRFS(vp->v_mount);
715 if (vp->v_type != VREG)
718 if (vp->v_type == VREG && td != NULL) {
719 error = kern_getrlimit(RLIMIT_FSIZE, &limit);
721 dbg(9, "rlimit failure\n");
724 if (uio->uio_offset + uio->uio_resid > limit.rlim_cur) {
725 dbg(9, "file too big\n");
726 ksignal(td->td_proc, SIGXFSZ);
731 if (ap->a_ioflag & IO_APPEND)
732 uio->uio_offset = dnp->dn_size;
735 * buffer cache operations may be deferred, make sure
736 * the file is correctly sized right now.
738 osize = dnp->dn_size;
739 nsize = uio->uio_offset + uio->uio_resid;
740 if (nsize > osize && uio->uio_resid) {
741 KKASSERT(dnp->dn_fd >= 0);
742 dnp->dn_size = nsize;
743 ftruncate(dnp->dn_fd, nsize);
744 nvextendbuf(vp, osize, nsize,
745 BSIZE, BSIZE, -1, -1, 0);
746 } /* else nsize = osize; NOT USED */
748 while (uio->uio_resid > 0) {
750 * Use buffer cache I/O (via dirfs_strategy)
752 offset = (size_t)uio->uio_offset & BMASK;
753 base_offset = (off_t)uio->uio_offset - offset;
754 len = BSIZE - offset;
756 if (len > uio->uio_resid)
757 len = uio->uio_resid;
759 error = bread(vp, base_offset, BSIZE, &bp);
760 error = uiomovebp(bp, (char *)bp->b_data + offset, len, uio);
763 dbg(9, "WRITE uiomove failed\n");
767 dbg(9, "WRITE dn_size=%jd uio_offset=%jd uio_resid=%jd base_offset=%jd\n",
768 dnp->dn_size, uio->uio_offset, uio->uio_resid, base_offset);
770 if (ap->a_ioflag & IO_SYNC)
776 KTR_LOG(dirfs_write, dnp, base_offset, uio->uio_resid,
777 dnp->dn_size, error);
783 dirfs_advlock (struct vop_advlock_args *ap)
785 struct vnode *vp = ap->a_vp;
786 dirfs_node_t dnp = VP_TO_NODE(vp);
790 return (lf_advlock(ap, &dnp->dn_advlock, dnp->dn_size));
794 dirfs_strategy(struct vop_strategy_args *ap)
798 struct bio *bio = ap->a_bio;
799 struct buf *bp = bio->bio_buf;
800 struct vnode *vp = ap->a_vp;
808 dnp = VP_TO_NODE(vp);
809 dmp = VFS_TO_DIRFS(vp->v_mount);
813 if (vp->v_type != VREG) {
814 dbg(9, "not VREG\n");
815 bp->b_resid = bp->b_bcount;
816 bp->b_flags |= B_ERROR | B_INVAL;
817 bp->b_error = EINVAL;
822 if (dnp->dn_fd == DIRFS_NOFD) {
824 panic("Meh, no fd to write to. dnp=%p\n", dnp);
827 if (bio->bio_offset + bp->b_bcount > dnp->dn_size)
828 iosize = dnp->dn_size - bio->bio_offset;
830 iosize = bp->b_bcount;
831 KKASSERT((ssize_t)iosize >= 0);
835 error = pwrite(dnp->dn_fd, bp->b_data, iosize, bio->bio_offset);
838 error = pread(dnp->dn_fd, bp->b_data, iosize, bio->bio_offset);
841 bp->b_error = error = EINVAL;
842 bp->b_flags |= B_ERROR;
846 if (error >= 0 && error < bp->b_bcount)
847 bzero(bp->b_data + error, bp->b_bcount - error);
849 if (error < 0 && errno != EINTR) {
850 dbg(9, "error=%d dnp=%p dnp->dn_fd=%d "
851 "bio->bio_offset=%ld bcount=%d resid=%d iosize=%zd\n",
852 errno, dnp, dnp->dn_fd, bio->bio_offset, bp->b_bcount,
853 bp->b_resid, iosize);
855 bp->b_resid = bp->b_bcount;
856 bp->b_flags |= B_ERROR;
858 tmp = dirfs_node_absolute_path(dmp, dnp, &pathfree);
859 dirfs_node_stat(DIRFS_NOFD, tmp, dnp);
860 dirfs_dropfd(dmp, NULL, pathfree);
863 KTR_LOG(dirfs_strategy, dnp, dnp->dn_size, iosize, bp->b_cmd,
864 bp->b_error, bp->b_resid, bio->bio_offset, error);
872 dirfs_bmap(struct vop_bmap_args *ap)
876 if (ap->a_doffsetp != NULL)
877 *ap->a_doffsetp = ap->a_loffset;
878 if (ap->a_runp != NULL)
880 if (ap->a_runb != NULL)
887 dirfs_nremove(struct vop_nremove_args *ap)
889 dirfs_node_t dnp, pdnp;
893 struct nchandle *nch;
894 struct namecache *ncp;
911 dmp = VFS_TO_DIRFS(mp);
913 lwkt_gettoken(&mp->mnt_token);
914 cache_vget(nch, ap->a_cred, LK_SHARED, &vp);
917 pdnp = VP_TO_NODE(dvp);
918 dnp = VP_TO_NODE(vp);
920 if (vp->v_type == VDIR) {
923 pathnp = dirfs_findfd(dmp, dnp, &tmp, &pathfree);
924 dirfs_node_lock(pdnp);
925 error = unlinkat(pathnp->dn_fd, tmp, 0);
928 dirfs_node_setpassive(dmp, dnp, 0);
929 if (dnp->dn_parent) {
930 dirfs_node_drop(dmp, dnp->dn_parent);
931 dnp->dn_parent = NULL;
936 dirfs_node_unlock(pdnp);
937 dirfs_dropfd(dmp, pathnp, pathfree);
940 lwkt_reltoken(&mp->mnt_token);
942 KTR_LOG(dirfs_nremove, dnp, pdnp, error);
948 dirfs_nlink(struct vop_nlink_args *ap)
952 KTR_LOG(dirfs_unsupported, __func__);
958 dirfs_nrename(struct vop_nrename_args *ap)
960 dirfs_node_t dnp, fdnp, tdnp;
962 struct namecache *fncp, *tncp;
963 struct vnode *fdvp, *tdvp, *vp;
965 char *fpath, *fpathfree;
966 char *tpath, *tpathfree;
974 fncp = ap->a_fnch->ncp;
975 tncp = ap->a_tnch->ncp;
977 dmp = VFS_TO_DIRFS(mp);
978 fdnp = VP_TO_NODE(fdvp);
979 tdnp = VP_TO_NODE(tdvp);
981 dbg(9, "fdnp=%p tdnp=%p from=%s to=%s\n", fdnp, tdnp, fncp->nc_name,
984 if (fdvp->v_mount != tdvp->v_mount)
986 if (fdvp->v_mount != fncp->nc_vp->v_mount)
988 if (fdvp->v_mount->mnt_flag & MNT_RDONLY)
991 tpath = dirfs_node_absolute_path_plus(dmp, tdnp,
992 tncp->nc_name, &tpathfree);
993 fpath = dirfs_node_absolute_path_plus(dmp, fdnp,
994 fncp->nc_name, &fpathfree);
995 error = rename(fpath, tpath);
999 vp = fncp->nc_vp; /* file being renamed */
1000 dnp = VP_TO_NODE(vp);
1001 dirfs_node_setname(dnp, tncp->nc_name, tncp->nc_nlen);
1004 * We have to mark the target file that was replaced by
1005 * the rename as having been unlinked.
1009 dbg(9, "RENAME2\n");
1010 dnp = VP_TO_NODE(vp);
1011 cache_unlink(ap->a_tnch);
1012 dirfs_node_setpassive(dmp, dnp, 0);
1013 if (dnp->dn_parent) {
1014 dirfs_node_drop(dmp, dnp->dn_parent);
1015 dnp->dn_parent = NULL;
1019 * nlinks on directories can be a bit weird. Zero
1023 cache_inval_vp(vp, CINV_DESTROY);
1025 cache_rename(ap->a_fnch, ap->a_tnch);
1027 dirfs_dropfd(dmp, NULL, fpathfree);
1028 dirfs_dropfd(dmp, NULL, tpathfree);
1034 dirfs_nmkdir(struct vop_nmkdir_args *ap)
1037 dirfs_node_t dnp, pdnp, dnp1;
1038 struct namecache *ncp;
1042 char *tmp, *pathfree;
1049 extrapath = error = 0;
1052 dmp = VFS_TO_DIRFS(dvp->v_mount);
1053 pdnp = VP_TO_NODE(dvp);
1054 ncp = ap->a_nch->ncp;
1056 pathfree = tmp = path = NULL;
1059 dirfs_node_lock(pdnp);
1060 if (pdnp->dn_fd != DIRFS_NOFD) {
1062 path = ncp->nc_name;
1064 dnp1 = dirfs_findfd(dmp, pdnp, &tmp, &pathfree);
1066 /* XXX check there is room to copy the path */
1067 path = kmalloc(MAXPATHLEN, M_DIRFS_MISC, M_ZERO | M_WAITOK);
1068 ksnprintf(path, MAXPATHLEN, "%s/%s", tmp, ncp->nc_name);
1070 dirfs_dropfd(dmp, dnp1, pathfree);
1073 error = mkdirat(pfd, path, vap->va_mode);
1076 } else { /* Directory has been made */
1077 error = dirfs_alloc_file(dmp, &dnp, pdnp, ncp, vpp,
1081 cache_setunresolved(ap->a_nch);
1082 cache_setvp(ap->a_nch, *vpp);
1084 dirfs_node_unlock(pdnp);
1087 kfree(path, M_DIRFS_MISC);
1089 KTR_LOG(dirfs_nmkdir, pdnp, dnp, ncp->nc_name, error);
1095 dirfs_nrmdir(struct vop_nrmdir_args *ap)
1097 dirfs_node_t dnp, pdnp;
1100 struct nchandle *nch;
1101 struct namecache *ncp;
1118 dmp = VFS_TO_DIRFS(mp);
1120 lwkt_gettoken(&mp->mnt_token);
1121 cache_vget(nch, ap->a_cred, LK_SHARED, &vp);
1124 pdnp = VP_TO_NODE(dvp);
1125 dnp = VP_TO_NODE(vp);
1127 if (vp->v_type != VDIR) {
1130 tmp = dirfs_node_absolute_path(dmp, dnp, &pathfree);
1131 dirfs_node_lock(pdnp);
1135 dirfs_node_setpassive(dmp, dnp, 0);
1136 if (dnp->dn_parent) {
1137 dirfs_node_drop(dmp, dnp->dn_parent);
1138 dnp->dn_parent = NULL;
1142 * nlinks on directories can be a bit weird. Zero
1146 cache_inval_vp(vp, CINV_DESTROY);
1150 dirfs_node_unlock(pdnp);
1151 dirfs_dropfd(dmp, NULL, pathfree);
1154 lwkt_reltoken(&mp->mnt_token);
1156 KTR_LOG(dirfs_nrmdir, dnp, pdnp, error);
1162 dirfs_nsymlink(struct vop_nsymlink_args *ap)
1165 dirfs_node_t dnp, pdnp;
1167 struct namecache *ncp;
1171 char *tmp, *pathfree;
1181 dmp = VFS_TO_DIRFS(dvp->v_mount);
1182 pdnp = VP_TO_NODE(dvp);
1183 ncp = ap->a_nch->ncp;
1185 pathfree = tmp = path = NULL;
1188 lwkt_gettoken(&mp->mnt_token);
1189 vap->va_type = VLNK;
1191 /* Find out the whole path of our new symbolic link */
1192 tmp = dirfs_node_absolute_path(dmp, pdnp, &pathfree);
1193 /* XXX check there is room to copy the path */
1194 path = kmalloc(MAXPATHLEN, M_DIRFS_MISC, M_ZERO | M_WAITOK);
1195 ksnprintf(path, MAXPATHLEN, "%s/%s", tmp, ncp->nc_name);
1196 dirfs_dropfd(dmp, NULL, pathfree);
1198 error = symlink(ap->a_target, path);
1201 } else { /* Symlink has been made */
1202 error = dirfs_alloc_file(dmp, &dnp, pdnp, ncp, vpp,
1206 cache_setunresolved(ap->a_nch);
1207 cache_setvp(ap->a_nch, *vpp);
1209 dbg(5, "path=%s a_target=%s\n", path, ap->a_target);
1211 KTR_LOG(dirfs_nsymlink, dnp, ap->a_target, path, error);
1212 kfree(path, M_DIRFS_MISC);
1213 lwkt_reltoken(&mp->mnt_token);
1220 dirfs_readdir(struct vop_readdir_args *ap)
1223 struct dirent *dp, *dpn;
1224 off_t __unused **cookies = ap->a_cookies;
1225 int *ncookies = ap->a_ncookies;
1229 struct vnode *vp = ap->a_vp;
1241 debug(1, "ncookies=%d\n", *ncookies);
1243 dnp = VP_TO_NODE(vp);
1245 startoff = uio->uio_offset;
1251 if (vp->v_type != VDIR)
1253 if (uio->uio_resid < 0)
1255 if ((bufsiz = uio->uio_resid) > 4096)
1257 buf = kmalloc(bufsiz, M_DIRFS_MISC, M_WAITOK | M_ZERO);
1260 * Generally speaking we have to be able to process ALL the
1261 * entries returned by getdirentries() in order for the seek
1262 * position to be correct. For now try to size the buffer
1263 * to make this happen. A smaller buffer always works. For
1264 * now just use an appropriate size.
1266 dirfs_node_lock(dnp);
1267 lseek(dnp->dn_fd, startoff, SEEK_SET);
1268 bytes = getdirentries(dnp->dn_fd, buf, bufsiz, &base);
1269 dbg(9, "seek %016jx %016jx %016jx\n",
1270 (intmax_t)startoff, (intmax_t)base,
1271 (intmax_t)lseek(dnp->dn_fd, 0, SEEK_CUR));
1273 if (errno == EINVAL)
1274 panic("EINVAL on readdir\n");
1278 } else if (bytes == 0) {
1284 for (dp = (struct dirent *)buf; bytes > 0 && uio->uio_resid > 0;
1285 bytes -= _DIRENT_DIRSIZ(dp), dp = dpn) {
1286 r = vop_write_dirent(&error, uio, dp->d_ino, dp->d_type,
1287 dp->d_namlen, dp->d_name);
1290 dpn = _DIRENT_NEXT(dp);
1294 curoff = lseek(dnp->dn_fd, 0, SEEK_CUR);
1297 kfree(buf, M_DIRFS_MISC);
1298 uio->uio_offset = curoff;
1299 dirfs_node_unlock(dnp);
1301 KTR_LOG(dirfs_readdir, dnp, dnp->dn_fd, startoff, uio->uio_offset);
1307 dirfs_readlink(struct vop_readlink_args *ap)
1309 dirfs_node_t dnp, pathnp;
1314 char *tmp, *pathfree, *buf;
1322 KKASSERT(vp->v_type == VLNK);
1325 tmp = pathfree = NULL;
1328 dmp = VFS_TO_DIRFS(mp);
1329 dnp = VP_TO_NODE(vp);
1331 lwkt_gettoken(&mp->mnt_token);
1333 pathnp = dirfs_findfd(dmp, dnp, &tmp, &pathfree);
1335 buf = kmalloc(uio->uio_resid, M_DIRFS_MISC, M_WAITOK | M_ZERO);
1336 nlen = readlinkat(pathnp->dn_fd, dnp->dn_name, buf, uio->uio_resid);
1340 error = uiomove(buf, nlen + 1, uio);
1345 dirfs_dropfd(dmp, pathnp, pathfree);
1346 kfree(buf, M_DIRFS_MISC);
1348 lwkt_reltoken(&mp->mnt_token);
1354 * Main tasks to be performed.
1355 * 1) When inode is NULL recycle the vnode
1356 * 2) When the inode has 0 links:
1357 * - Check if in the TAILQ, if so remove.
1358 * - Destroy the inode.
1359 * - Recycle the vnode.
1360 * 3) If none of the above, add the node to the TAILQ
1361 * when it has a valid fd and there is room on the
1366 dirfs_inactive(struct vop_inactive_args *ap)
1375 dmp = VFS_TO_DIRFS(vp->v_mount);
1376 dnp = VP_TO_NODE(vp);
1378 /* Degenerate case */
1380 dbg(5, "dnp was NULL\n");
1385 dirfs_mount_gettoken(dmp);
1388 * Deal with the case the inode has 0 links which means it was unlinked.
1390 if (dnp->dn_links == 0) {
1392 dbg(5, "recycled a vnode of an unlinked dnp\n");
1398 * Try to retain the fd in our fd cache.
1400 dirfs_node_setpassive(dmp, dnp, 1);
1402 dirfs_mount_reltoken(dmp);
1409 dirfs_reclaim(struct vop_reclaim_args *ap)
1418 dnp = VP_TO_NODE(vp);
1419 dmp = VFS_TO_DIRFS(vp->v_mount);
1421 dirfs_free_vp(dmp, dnp);
1422 /* dnp is now invalid, may have been destroyed */
1428 dirfs_mountctl(struct vop_mountctl_args *ap)
1432 KTR_LOG(dirfs_unsupported, __func__);
1438 dirfs_print(struct vop_print_args *v)
1442 KTR_LOG(dirfs_unsupported, __func__);
1448 dirfs_pathconf(struct vop_pathconf_args *v)
1456 dirfs_kqfilter (struct vop_kqfilter_args *ap)
1460 KTR_LOG(dirfs_unsupported, __func__);
1465 struct vop_ops dirfs_vnode_vops = {
1466 .vop_default = vop_defaultop,
1467 .vop_nwhiteout = vop_compat_nwhiteout,
1468 .vop_ncreate = dirfs_ncreate,
1469 .vop_nresolve = dirfs_nresolve,
1470 .vop_markatime = vop_stdmarkatime,
1471 .vop_nlookupdotdot = dirfs_nlookupdotdot,
1472 .vop_nmknod = dirfs_nmknod,
1473 .vop_open = dirfs_open,
1474 .vop_close = dirfs_close,
1475 .vop_access = dirfs_access,
1476 .vop_getattr = dirfs_getattr,
1477 .vop_setattr = dirfs_setattr,
1478 .vop_read = dirfs_read,
1479 .vop_write = dirfs_write,
1480 .vop_fsync = dirfs_fsync,
1481 .vop_mountctl = dirfs_mountctl,
1482 .vop_nremove = dirfs_nremove,
1483 .vop_nlink = dirfs_nlink,
1484 .vop_nrename = dirfs_nrename,
1485 .vop_nmkdir = dirfs_nmkdir,
1486 .vop_nrmdir = dirfs_nrmdir,
1487 .vop_nsymlink = dirfs_nsymlink,
1488 .vop_readdir = dirfs_readdir,
1489 .vop_readlink = dirfs_readlink,
1490 .vop_inactive = dirfs_inactive,
1491 .vop_reclaim = dirfs_reclaim,
1492 .vop_print = dirfs_print,
1493 .vop_pathconf = vop_stdpathconf,
1494 .vop_bmap = dirfs_bmap,
1495 .vop_strategy = dirfs_strategy,
1496 .vop_advlock = dirfs_advlock,
1497 .vop_kqfilter = dirfs_kqfilter,
1498 .vop_getpages = vop_stdgetpages,
1499 .vop_putpages = vop_stdputpages