2 * Copyright (c) 2000-2004
3 * Poul-Henning Kamp. All rights reserved.
4 * Copyright (c) 1989, 1992-1993, 1995
5 * The Regents of the University of California. All rights reserved.
7 * This code is derived from software donated to Berkeley by
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)kernfs_vnops.c 8.15 (Berkeley) 5/21/95
32 * From: FreeBSD: src/sys/miscfs/kernfs/kernfs_vnops.c 1.43
39 * remove empty directories
43 #include <sys/param.h>
44 #include <sys/systm.h>
46 #include <sys/dirent.h>
47 #include <sys/fcntl.h>
49 #include <sys/filedesc.h>
50 #include <sys/filio.h>
52 #include <sys/kernel.h>
54 #include <sys/malloc.h>
55 #include <sys/mount.h>
56 #include <sys/namei.h>
62 #include <sys/ttycom.h>
63 #include <sys/unistd.h>
64 #include <sys/vnode.h>
66 static struct vop_vector devfs_vnodeops;
67 static struct vop_vector devfs_specops;
68 static struct fileops devfs_ops_f;
70 #include <fs/devfs/devfs.h>
71 #include <fs/devfs/devfs_int.h>
73 #include <security/mac/mac_framework.h>
75 static MALLOC_DEFINE(M_CDEVPDATA, "DEVFSP", "Metainfo for cdev-fp data");
77 struct mtx devfs_de_interlock;
78 MTX_SYSINIT(devfs_de_interlock, &devfs_de_interlock, "devfs interlock", MTX_DEF);
79 struct sx clone_drain_lock;
80 SX_SYSINIT(clone_drain_lock, &clone_drain_lock, "clone events drain lock");
81 struct mtx cdevpriv_mtx;
82 MTX_SYSINIT(cdevpriv_mtx, &cdevpriv_mtx, "cdevpriv lock", MTX_DEF);
85 devfs_fp_check(struct file *fp, struct cdev **devp, struct cdevsw **dswp,
89 *dswp = devvn_refthread(fp->f_vnode, devp, ref);
90 if (*devp != fp->f_data) {
92 dev_relthread(*devp, *ref);
95 KASSERT((*devp)->si_refcount > 0,
96 ("devfs: un-referenced struct cdev *(%s)", devtoname(*devp)));
99 curthread->td_fpop = fp;
104 devfs_get_cdevpriv(void **datap)
107 struct cdev_privdata *p;
110 fp = curthread->td_fpop;
116 *datap = p->cdpd_data;
123 devfs_set_cdevpriv(void *priv, cdevpriv_dtr_t priv_dtr)
126 struct cdev_priv *cdp;
127 struct cdev_privdata *p;
130 fp = curthread->td_fpop;
133 cdp = cdev2priv((struct cdev *)fp->f_data);
134 p = malloc(sizeof(struct cdev_privdata), M_CDEVPDATA, M_WAITOK);
136 p->cdpd_dtr = priv_dtr;
138 mtx_lock(&cdevpriv_mtx);
139 if (fp->f_cdevpriv == NULL) {
140 LIST_INSERT_HEAD(&cdp->cdp_fdpriv, p, cdpd_list);
142 mtx_unlock(&cdevpriv_mtx);
145 mtx_unlock(&cdevpriv_mtx);
146 free(p, M_CDEVPDATA);
153 devfs_destroy_cdevpriv(struct cdev_privdata *p)
156 mtx_assert(&cdevpriv_mtx, MA_OWNED);
157 p->cdpd_fp->f_cdevpriv = NULL;
158 LIST_REMOVE(p, cdpd_list);
159 mtx_unlock(&cdevpriv_mtx);
160 (p->cdpd_dtr)(p->cdpd_data);
161 free(p, M_CDEVPDATA);
165 devfs_fpdrop(struct file *fp)
167 struct cdev_privdata *p;
169 mtx_lock(&cdevpriv_mtx);
170 if ((p = fp->f_cdevpriv) == NULL) {
171 mtx_unlock(&cdevpriv_mtx);
174 devfs_destroy_cdevpriv(p);
178 devfs_clear_cdevpriv(void)
182 fp = curthread->td_fpop;
189 devfs_vptocnp(struct vop_vptocnp_args *ap)
191 struct vnode *vp = ap->a_vp;
192 struct vnode **dvp = ap->a_vpp;
193 struct devfs_mount *dmp;
194 char *buf = ap->a_buf;
195 int *buflen = ap->a_buflen;
196 struct devfs_dirent *dd, *de;
199 dmp = VFSTODEVFS(vp->v_mount);
204 sx_xlock(&dmp->dm_lock);
206 if (vp->v_type == VCHR) {
207 i -= strlen(dd->de_cdp->cdp_c.si_name);
212 bcopy(dd->de_cdp->cdp_c.si_name, buf + i,
213 strlen(dd->de_cdp->cdp_c.si_name));
215 } else if (vp->v_type == VDIR) {
216 if (dd == dmp->dm_rootdir) {
221 i -= dd->de_dirent->d_namlen;
226 bcopy(dd->de_dirent->d_name, buf + i,
227 dd->de_dirent->d_namlen);
234 de = devfs_parent_dirent(de);
239 mtx_lock(&devfs_de_interlock);
243 mtx_unlock(&devfs_de_interlock);
247 mtx_unlock(&devfs_de_interlock);
251 sx_xunlock(&dmp->dm_lock);
256 * Construct the fully qualified path name relative to the mountpoint
259 devfs_fqpn(char *buf, struct vnode *dvp, struct componentname *cnp)
262 struct devfs_dirent *de, *dd;
263 struct devfs_mount *dmp;
265 dmp = VFSTODEVFS(dvp->v_mount);
269 i -= cnp->cn_namelen;
272 bcopy(cnp->cn_nameptr, buf + i, cnp->cn_namelen);
274 while (de != dmp->dm_rootdir) {
279 i -= de->de_dirent->d_namlen;
282 bcopy(de->de_dirent->d_name, buf + i,
283 de->de_dirent->d_namlen);
284 de = devfs_parent_dirent(de);
292 devfs_allocv_drop_refs(int drop_dm_lock, struct devfs_mount *dmp,
293 struct devfs_dirent *de)
298 if (de->de_flags & DE_DOOMED)
300 if (DEVFS_DE_DROP(de)) {
301 KASSERT(not_found == 1, ("DEVFS de dropped but not doomed"));
302 devfs_dirent_free(de);
304 if (DEVFS_DMP_DROP(dmp)) {
305 KASSERT(not_found == 1,
306 ("DEVFS mount struct freed before dirent"));
308 sx_xunlock(&dmp->dm_lock);
309 devfs_unmount_final(dmp);
311 if (not_found == 1 || (drop_dm_lock && not_found != 2))
312 sx_unlock(&dmp->dm_lock);
317 devfs_insmntque_dtr(struct vnode *vp, void *arg)
319 struct devfs_dirent *de;
321 de = (struct devfs_dirent *)arg;
322 mtx_lock(&devfs_de_interlock);
325 mtx_unlock(&devfs_de_interlock);
331 * devfs_allocv shall be entered with dmp->dm_lock held, and it drops
335 devfs_allocv(struct devfs_dirent *de, struct mount *mp, int lockmode,
341 struct devfs_mount *dmp;
344 dmp = VFSTODEVFS(mp);
345 if (de->de_flags & DE_DOOMED) {
346 sx_xunlock(&dmp->dm_lock);
352 mtx_lock(&devfs_de_interlock);
356 mtx_unlock(&devfs_de_interlock);
357 sx_xunlock(&dmp->dm_lock);
358 vget(vp, lockmode | LK_INTERLOCK | LK_RETRY, curthread);
359 sx_xlock(&dmp->dm_lock);
360 if (devfs_allocv_drop_refs(0, dmp, de)) {
364 else if ((vp->v_iflag & VI_DOOMED) != 0) {
365 mtx_lock(&devfs_de_interlock);
366 if (de->de_vnode == vp) {
370 mtx_unlock(&devfs_de_interlock);
374 sx_xunlock(&dmp->dm_lock);
378 mtx_unlock(&devfs_de_interlock);
379 if (de->de_dirent->d_type == DT_CHR) {
380 if (!(de->de_cdp->cdp_flags & CDP_ACTIVE)) {
381 devfs_allocv_drop_refs(1, dmp, de);
384 dev = &de->de_cdp->cdp_c;
388 error = getnewvnode("devfs", mp, &devfs_vnodeops, &vp);
390 devfs_allocv_drop_refs(1, dmp, de);
391 printf("devfs_allocv: failed to allocate new vnode\n");
395 if (de->de_dirent->d_type == DT_CHR) {
400 /* XXX: v_rdev should be protect by vnode lock */
402 KASSERT(vp->v_usecount == 1,
403 ("%s %d (%d)\n", __func__, __LINE__, vp->v_usecount));
404 dev->si_usecount += vp->v_usecount;
405 /* Special casing of ttys for deadfs. Probably redundant. */
407 if (dsw != NULL && (dsw->d_flags & D_TTY) != 0)
408 vp->v_vflag |= VV_ISTTY;
411 if ((dev->si_flags & SI_ETERNAL) != 0)
412 vp->v_vflag |= VV_ETERNALDEV;
413 vp->v_op = &devfs_specops;
414 } else if (de->de_dirent->d_type == DT_DIR) {
416 } else if (de->de_dirent->d_type == DT_LNK) {
422 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_NOWITNESS);
423 mtx_lock(&devfs_de_interlock);
426 mtx_unlock(&devfs_de_interlock);
427 error = insmntque1(vp, mp, devfs_insmntque_dtr, de);
429 (void) devfs_allocv_drop_refs(1, dmp, de);
432 if (devfs_allocv_drop_refs(0, dmp, de)) {
437 mac_devfs_vnode_associate(mp, de, vp);
439 sx_xunlock(&dmp->dm_lock);
445 devfs_access(struct vop_access_args *ap)
447 struct vnode *vp = ap->a_vp;
448 struct devfs_dirent *de;
452 if (vp->v_type == VDIR)
455 error = vaccess(vp->v_type, de->de_mode, de->de_uid, de->de_gid,
456 ap->a_accmode, ap->a_cred, NULL);
461 /* We do, however, allow access to the controlling terminal */
462 if (!(ap->a_td->td_proc->p_flag & P_CONTROLT))
464 if (ap->a_td->td_proc->p_session->s_ttyvp == de->de_vnode)
471 devfs_close(struct vop_close_args *ap)
473 struct vnode *vp = ap->a_vp, *oldvp;
474 struct thread *td = ap->a_td;
475 struct cdev *dev = vp->v_rdev;
477 int vp_locked, error, ref;
480 * XXX: Don't call d_close() if we were called because of
481 * XXX: insmntque1() failure.
483 if (vp->v_data == NULL)
487 * Hack: a tty device that is a controlling terminal
488 * has a reference from the session structure.
489 * We cannot easily tell that a character device is
490 * a controlling terminal, unless it is the closing
491 * process' controlling terminal. In that case,
492 * if the reference count is 2 (this last descriptor
493 * plus the session), release the reference from the session.
496 sx_xlock(&proctree_lock);
497 if (td && vp == td->td_proc->p_session->s_ttyvp) {
498 SESS_LOCK(td->td_proc->p_session);
500 if (count_dev(dev) == 2 && (vp->v_iflag & VI_DOOMED) == 0) {
501 td->td_proc->p_session->s_ttyvp = NULL;
505 SESS_UNLOCK(td->td_proc->p_session);
507 sx_xunlock(&proctree_lock);
511 * We do not want to really close the device if it
512 * is still in use unless we are trying to close it
513 * forcibly. Since every use (buffer, vnode, swap, cmap)
514 * holds a reference to the vnode, and because we mark
515 * any other vnodes that alias this device, when the
516 * sum of the reference counts on all the aliased
517 * vnodes descends to one, we are on last close.
519 dsw = dev_refthread(dev, &ref);
523 if (vp->v_iflag & VI_DOOMED) {
525 } else if (dsw->d_flags & D_TRACKCLOSE) {
526 /* Keep device updated on status. */
527 } else if (count_dev(dev) > 1) {
529 dev_relthread(dev, ref);
534 vp_locked = VOP_ISLOCKED(vp);
536 KASSERT(dev->si_refcount > 0,
537 ("devfs_close() on un-referenced struct cdev *(%s)", devtoname(dev)));
538 error = dsw->d_close(dev, ap->a_fflag, S_IFCHR, td);
539 dev_relthread(dev, ref);
540 vn_lock(vp, vp_locked | LK_RETRY);
546 devfs_close_f(struct file *fp, struct thread *td)
552 * NB: td may be NULL if this descriptor is closed due to
553 * garbage collection from a closed UNIX domain socket.
555 fpop = curthread->td_fpop;
556 curthread->td_fpop = fp;
557 error = vnops.fo_close(fp, td);
558 curthread->td_fpop = fpop;
561 * The f_cdevpriv cannot be assigned non-NULL value while we
562 * are destroying the file.
564 if (fp->f_cdevpriv != NULL)
570 devfs_fsync(struct vop_fsync_args *ap)
574 struct devfs_dirent *de;
576 if (!vn_isdisk(ap->a_vp, &error)) {
577 bo = &ap->a_vp->v_bufobj;
578 de = ap->a_vp->v_data;
579 if (error == ENXIO && bo->bo_dirty.bv_cnt > 0) {
580 printf("Device %s went missing before all of the data "
581 "could be written to it; expect data loss.\n",
582 de->de_dirent->d_name);
584 error = vop_stdfsync(ap);
585 if (bo->bo_dirty.bv_cnt != 0 || error != 0)
586 panic("devfs_fsync: vop_stdfsync failed.");
592 return (vop_stdfsync(ap));
596 devfs_getattr(struct vop_getattr_args *ap)
598 struct vnode *vp = ap->a_vp;
599 struct vattr *vap = ap->a_vap;
601 struct devfs_dirent *de;
605 KASSERT(de != NULL, ("Null dirent in devfs_getattr vp=%p", vp));
606 if (vp->v_type == VDIR) {
609 ("Null dir dirent in devfs_getattr vp=%p", vp));
611 vap->va_uid = de->de_uid;
612 vap->va_gid = de->de_gid;
613 vap->va_mode = de->de_mode;
614 if (vp->v_type == VLNK)
615 vap->va_size = strlen(de->de_symlink);
616 else if (vp->v_type == VDIR)
617 vap->va_size = vap->va_bytes = DEV_BSIZE;
620 if (vp->v_type != VDIR)
622 vap->va_blocksize = DEV_BSIZE;
623 vap->va_type = vp->v_type;
627 if ((aa).tv_sec <= 3600) { \
628 (aa).tv_sec = boottime.tv_sec; \
629 (aa).tv_nsec = boottime.tv_usec * 1000; \
633 if (vp->v_type != VCHR) {
635 vap->va_atime = de->de_atime;
637 vap->va_mtime = de->de_mtime;
639 vap->va_ctime = de->de_ctime;
643 vap->va_atime = dev->si_atime;
645 vap->va_mtime = dev->si_mtime;
647 vap->va_ctime = dev->si_ctime;
649 vap->va_rdev = cdev2priv(dev)->cdp_inode;
654 vap->va_nlink = de->de_links;
655 vap->va_fileid = de->de_inode;
662 devfs_ioctl_f(struct file *fp, u_long com, void *data, struct ucred *cred, struct thread *td)
670 struct fiodgname_arg *fgn;
674 error = devfs_fp_check(fp, &dev, &dsw, &ref);
678 if (com == FIODTYPE) {
679 *(int *)data = dsw->d_flags & D_TYPEMASK;
681 dev_relthread(dev, ref);
683 } else if (com == FIODGNAME) {
690 error = copyout(p, fgn->buf, i);
692 dev_relthread(dev, ref);
695 error = dsw->d_ioctl(dev, com, data, fp->f_flag, td);
697 dev_relthread(dev, ref);
698 if (error == ENOIOCTL)
700 if (error == 0 && com == TIOCSCTTY) {
703 /* Do nothing if reassigning same control tty */
704 sx_slock(&proctree_lock);
705 if (td->td_proc->p_session->s_ttyvp == vp) {
706 sx_sunlock(&proctree_lock);
710 vpold = td->td_proc->p_session->s_ttyvp;
712 SESS_LOCK(td->td_proc->p_session);
713 td->td_proc->p_session->s_ttyvp = vp;
714 SESS_UNLOCK(td->td_proc->p_session);
716 sx_sunlock(&proctree_lock);
718 /* Get rid of reference to old control tty */
727 devfs_kqfilter_f(struct file *fp, struct knote *kn)
737 error = devfs_fp_check(fp, &dev, &dsw, &ref);
740 error = dsw->d_kqfilter(dev, kn);
742 dev_relthread(dev, ref);
747 devfs_prison_check(struct devfs_dirent *de, struct ucred *tcr)
749 struct cdev_priv *cdp;
755 dcr = cdp->cdp_c.si_cred;
759 return (prison_check(tcr, dcr));
763 devfs_lookupx(struct vop_lookup_args *ap, int *dm_unlock)
765 struct componentname *cnp;
766 struct vnode *dvp, **vpp;
768 struct devfs_dirent *de, *dd;
769 struct devfs_dirent **dde;
770 struct devfs_mount *dmp;
772 int error, flags, nameiop, dvplocked;
773 char specname[SPECNAMELEN + 1], *pname;
778 pname = cnp->cn_nameptr;
780 flags = cnp->cn_flags;
781 nameiop = cnp->cn_nameiop;
782 dmp = VFSTODEVFS(dvp->v_mount);
786 if ((flags & ISLASTCN) && nameiop == RENAME)
789 if (dvp->v_type != VDIR)
792 if ((flags & ISDOTDOT) && (dvp->v_vflag & VV_ROOT))
795 error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, td);
799 if (cnp->cn_namelen == 1 && *pname == '.') {
800 if ((flags & ISLASTCN) && nameiop != LOOKUP)
807 if (flags & ISDOTDOT) {
808 if ((flags & ISLASTCN) && nameiop != LOOKUP)
810 de = devfs_parent_dirent(dd);
813 dvplocked = VOP_ISLOCKED(dvp);
815 error = devfs_allocv(de, dvp->v_mount,
816 cnp->cn_lkflags & LK_TYPE_MASK, vpp);
818 vn_lock(dvp, dvplocked | LK_RETRY);
824 if (DEVFS_DMP_DROP(dmp)) {
826 sx_xunlock(&dmp->dm_lock);
827 devfs_unmount_final(dmp);
831 de = devfs_find(dd, cnp->cn_nameptr, cnp->cn_namelen);
832 while (de == NULL) { /* While(...) so we can use break */
834 if (nameiop == DELETE)
838 * OK, we didn't have an entry for the name we were asked for
839 * so we try to see if anybody can create it on demand.
841 pname = devfs_fqpn(specname, dvp, cnp);
847 sx_xunlock(&dmp->dm_lock);
848 sx_slock(&clone_drain_lock);
849 EVENTHANDLER_INVOKE(dev_clone,
850 td->td_ucred, pname, strlen(pname), &cdev);
851 sx_sunlock(&clone_drain_lock);
852 sx_xlock(&dmp->dm_lock);
853 if (DEVFS_DMP_DROP(dmp)) {
855 sx_xunlock(&dmp->dm_lock);
856 devfs_unmount_final(dmp);
866 if (DEVFS_DMP_DROP(dmp)) {
868 sx_xunlock(&dmp->dm_lock);
869 devfs_unmount_final(dmp);
875 dde = &cdev2priv(cdev)->cdp_dirents[dmp->dm_idx];
876 if (dde != NULL && *dde != NULL)
883 if (de == NULL || de->de_flags & DE_WHITEOUT) {
884 if ((nameiop == CREATE || nameiop == RENAME) &&
885 (flags & (LOCKPARENT | WANTPARENT)) && (flags & ISLASTCN)) {
886 cnp->cn_flags |= SAVENAME;
887 return (EJUSTRETURN);
892 if (devfs_prison_check(de, td->td_ucred))
895 if ((cnp->cn_nameiop == DELETE) && (flags & ISLASTCN)) {
896 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, td);
905 error = devfs_allocv(de, dvp->v_mount, cnp->cn_lkflags & LK_TYPE_MASK,
912 devfs_lookup(struct vop_lookup_args *ap)
915 struct devfs_mount *dmp;
918 dmp = VFSTODEVFS(ap->a_dvp->v_mount);
920 sx_xlock(&dmp->dm_lock);
921 j = devfs_lookupx(ap, &dm_unlock);
923 sx_xunlock(&dmp->dm_lock);
928 devfs_mknod(struct vop_mknod_args *ap)
930 struct componentname *cnp;
931 struct vnode *dvp, **vpp;
932 struct devfs_dirent *dd, *de;
933 struct devfs_mount *dmp;
937 * The only type of node we should be creating here is a
938 * character device, for anything else return EOPNOTSUPP.
940 if (ap->a_vap->va_type != VCHR)
943 dmp = VFSTODEVFS(dvp->v_mount);
950 sx_xlock(&dmp->dm_lock);
951 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
952 if (cnp->cn_namelen != de->de_dirent->d_namlen)
954 if (bcmp(cnp->cn_nameptr, de->de_dirent->d_name,
955 de->de_dirent->d_namlen) != 0)
957 if (de->de_flags & DE_WHITEOUT)
963 de->de_flags &= ~DE_WHITEOUT;
964 error = devfs_allocv(de, dvp->v_mount, LK_EXCLUSIVE, vpp);
967 sx_xunlock(&dmp->dm_lock);
973 devfs_open(struct vop_open_args *ap)
975 struct thread *td = ap->a_td;
976 struct vnode *vp = ap->a_vp;
977 struct cdev *dev = vp->v_rdev;
978 struct file *fp = ap->a_fp;
979 int error, ref, vlocked;
983 if (vp->v_type == VBLK)
989 /* Make this field valid before any I/O in d_open. */
990 if (dev->si_iosize_max == 0)
991 dev->si_iosize_max = DFLTPHYS;
993 dsw = dev_refthread(dev, &ref);
996 if (fp == NULL && dsw->d_fdopen != NULL) {
997 dev_relthread(dev, ref);
1001 vlocked = VOP_ISLOCKED(vp);
1010 if (dsw->d_fdopen != NULL)
1011 error = dsw->d_fdopen(dev, ap->a_mode, td, fp);
1013 error = dsw->d_open(dev, ap->a_mode, S_IFCHR, td);
1016 vn_lock(vp, vlocked | LK_RETRY);
1017 dev_relthread(dev, ref);
1021 #if 0 /* /dev/console */
1022 KASSERT(fp != NULL, ("Could not vnode bypass device on NULL fp"));
1027 if (fp->f_ops == &badfileops)
1028 finit(fp, fp->f_flag, DTYPE_VNODE, dev, &devfs_ops_f);
1033 devfs_pathconf(struct vop_pathconf_args *ap)
1036 switch (ap->a_name) {
1037 case _PC_MAC_PRESENT:
1040 * If MAC is enabled, devfs automatically supports
1041 * trivial non-persistant label storage.
1049 return (vop_stdpathconf(ap));
1056 devfs_poll_f(struct file *fp, int events, struct ucred *cred, struct thread *td)
1064 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1066 return (poll_no_poll(events));
1067 error = dsw->d_poll(dev, events, td);
1069 dev_relthread(dev, ref);
1074 * Print out the contents of a special device vnode.
1077 devfs_print(struct vop_print_args *ap)
1080 printf("\tdev %s\n", devtoname(ap->a_vp->v_rdev));
1086 devfs_read_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
1089 int ioflag, error, ref, resid;
1094 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1097 resid = uio->uio_resid;
1098 ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT);
1099 if (ioflag & O_DIRECT)
1100 ioflag |= IO_DIRECT;
1102 if ((flags & FOF_OFFSET) == 0)
1103 uio->uio_offset = fp->f_offset;
1105 error = dsw->d_read(dev, uio, ioflag);
1106 if (uio->uio_resid != resid || (error == 0 && resid != 0))
1107 vfs_timestamp(&dev->si_atime);
1109 dev_relthread(dev, ref);
1111 if ((flags & FOF_OFFSET) == 0)
1112 fp->f_offset = uio->uio_offset;
1113 fp->f_nextoff = uio->uio_offset;
1118 devfs_readdir(struct vop_readdir_args *ap)
1123 struct devfs_dirent *dd;
1124 struct devfs_dirent *de;
1125 struct devfs_mount *dmp;
1127 int *tmp_ncookies = NULL;
1129 if (ap->a_vp->v_type != VDIR)
1133 if (uio->uio_offset < 0)
1137 * XXX: This is a temporary hack to get around this filesystem not
1138 * supporting cookies. We store the location of the ncookies pointer
1139 * in a temporary variable before calling vfs_subr.c:vfs_read_dirent()
1140 * and set the number of cookies to 0. We then set the pointer to
1141 * NULL so that vfs_read_dirent doesn't try to call realloc() on
1142 * ap->a_cookies. Later in this function, we restore the ap->a_ncookies
1143 * pointer to its original location before returning to the caller.
1145 if (ap->a_ncookies != NULL) {
1146 tmp_ncookies = ap->a_ncookies;
1147 *ap->a_ncookies = 0;
1148 ap->a_ncookies = NULL;
1151 dmp = VFSTODEVFS(ap->a_vp->v_mount);
1152 sx_xlock(&dmp->dm_lock);
1153 DEVFS_DMP_HOLD(dmp);
1154 devfs_populate(dmp);
1155 if (DEVFS_DMP_DROP(dmp)) {
1156 sx_xunlock(&dmp->dm_lock);
1157 devfs_unmount_final(dmp);
1158 if (tmp_ncookies != NULL)
1159 ap->a_ncookies = tmp_ncookies;
1163 de = ap->a_vp->v_data;
1165 TAILQ_FOREACH(dd, &de->de_dlist, de_list) {
1166 KASSERT(dd->de_cdp != (void *)0xdeadc0de, ("%s %d\n", __func__, __LINE__));
1167 if (dd->de_flags & DE_WHITEOUT)
1169 if (devfs_prison_check(dd, ap->a_cred))
1171 if (dd->de_dirent->d_type == DT_DIR)
1176 if (dp->d_reclen > uio->uio_resid)
1178 dp->d_fileno = de->de_inode;
1179 if (off >= uio->uio_offset) {
1180 error = vfs_read_dirent(ap, dp, off);
1184 off += dp->d_reclen;
1186 sx_xunlock(&dmp->dm_lock);
1187 uio->uio_offset = off;
1190 * Restore ap->a_ncookies if it wasn't originally NULL in the first
1193 if (tmp_ncookies != NULL)
1194 ap->a_ncookies = tmp_ncookies;
1200 devfs_readlink(struct vop_readlink_args *ap)
1202 struct devfs_dirent *de;
1204 de = ap->a_vp->v_data;
1205 return (uiomove(de->de_symlink, strlen(de->de_symlink), ap->a_uio));
1209 devfs_reclaim(struct vop_reclaim_args *ap)
1211 struct vnode *vp = ap->a_vp;
1212 struct devfs_dirent *de;
1215 mtx_lock(&devfs_de_interlock);
1218 de->de_vnode = NULL;
1221 mtx_unlock(&devfs_de_interlock);
1223 vnode_destroy_vobject(vp);
1236 dev->si_usecount -= vp->v_usecount;
1244 devfs_remove(struct vop_remove_args *ap)
1246 struct vnode *vp = ap->a_vp;
1247 struct devfs_dirent *dd;
1248 struct devfs_dirent *de;
1249 struct devfs_mount *dmp = VFSTODEVFS(vp->v_mount);
1251 sx_xlock(&dmp->dm_lock);
1252 dd = ap->a_dvp->v_data;
1254 if (de->de_cdp == NULL) {
1255 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
1256 devfs_delete(dmp, de, 1);
1258 de->de_flags |= DE_WHITEOUT;
1260 sx_xunlock(&dmp->dm_lock);
1265 * Revoke is called on a tty when a terminal session ends. The vnode
1266 * is orphaned by setting v_op to deadfs so we need to let go of it
1267 * as well so that we create a new one next time around.
1271 devfs_revoke(struct vop_revoke_args *ap)
1273 struct vnode *vp = ap->a_vp, *vp2;
1275 struct cdev_priv *cdp;
1276 struct devfs_dirent *de;
1279 KASSERT((ap->a_flags & REVOKEALL) != 0, ("devfs_revoke !REVOKEALL"));
1282 cdp = cdev2priv(dev);
1295 mtx_lock(&devfs_de_interlock);
1298 for (i = 0; i <= cdp->cdp_maxdirent; i++) {
1299 de = cdp->cdp_dirents[i];
1307 mtx_unlock(&devfs_de_interlock);
1308 if (vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK,
1322 mtx_unlock(&devfs_de_interlock);
1327 if (!(cdp->cdp_flags & CDP_ACTIVE) && cdp->cdp_inuse == 0) {
1328 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
1330 dev_rel(&cdp->cdp_c);
1334 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1339 devfs_rioctl(struct vop_ioctl_args *ap)
1342 struct devfs_mount *dmp;
1346 vn_lock(vp, LK_SHARED | LK_RETRY);
1347 if (vp->v_iflag & VI_DOOMED) {
1351 dmp = VFSTODEVFS(vp->v_mount);
1352 sx_xlock(&dmp->dm_lock);
1354 DEVFS_DMP_HOLD(dmp);
1355 devfs_populate(dmp);
1356 if (DEVFS_DMP_DROP(dmp)) {
1357 sx_xunlock(&dmp->dm_lock);
1358 devfs_unmount_final(dmp);
1361 error = devfs_rules_ioctl(dmp, ap->a_command, ap->a_data, ap->a_td);
1362 sx_xunlock(&dmp->dm_lock);
1367 devfs_rread(struct vop_read_args *ap)
1370 if (ap->a_vp->v_type != VDIR)
1372 return (VOP_READDIR(ap->a_vp, ap->a_uio, ap->a_cred, NULL, NULL, NULL));
1376 devfs_setattr(struct vop_setattr_args *ap)
1378 struct devfs_dirent *de;
1389 if ((vap->va_type != VNON) ||
1390 (vap->va_nlink != VNOVAL) ||
1391 (vap->va_fsid != VNOVAL) ||
1392 (vap->va_fileid != VNOVAL) ||
1393 (vap->va_blocksize != VNOVAL) ||
1394 (vap->va_flags != VNOVAL && vap->va_flags != 0) ||
1395 (vap->va_rdev != VNOVAL) ||
1396 ((int)vap->va_bytes != VNOVAL) ||
1397 (vap->va_gen != VNOVAL)) {
1402 if (vp->v_type == VDIR)
1406 if (vap->va_uid == (uid_t)VNOVAL)
1410 if (vap->va_gid == (gid_t)VNOVAL)
1414 if (uid != de->de_uid || gid != de->de_gid) {
1415 if ((ap->a_cred->cr_uid != de->de_uid) || uid != de->de_uid ||
1416 (gid != de->de_gid && !groupmember(gid, ap->a_cred))) {
1417 error = priv_check(td, PRIV_VFS_CHOWN);
1426 if (vap->va_mode != (mode_t)VNOVAL) {
1427 if (ap->a_cred->cr_uid != de->de_uid) {
1428 error = priv_check(td, PRIV_VFS_ADMIN);
1432 de->de_mode = vap->va_mode;
1436 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1437 /* See the comment in ufs_vnops::ufs_setattr(). */
1438 if ((error = VOP_ACCESS(vp, VADMIN, ap->a_cred, td)) &&
1439 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1440 (error = VOP_ACCESS(vp, VWRITE, ap->a_cred, td))))
1442 if (vap->va_atime.tv_sec != VNOVAL) {
1443 if (vp->v_type == VCHR)
1444 vp->v_rdev->si_atime = vap->va_atime;
1446 de->de_atime = vap->va_atime;
1448 if (vap->va_mtime.tv_sec != VNOVAL) {
1449 if (vp->v_type == VCHR)
1450 vp->v_rdev->si_mtime = vap->va_mtime;
1452 de->de_mtime = vap->va_mtime;
1458 if (vp->v_type == VCHR)
1459 vfs_timestamp(&vp->v_rdev->si_ctime);
1461 vfs_timestamp(&de->de_mtime);
1468 devfs_setlabel(struct vop_setlabel_args *ap)
1471 struct devfs_dirent *de;
1476 mac_vnode_relabel(ap->a_cred, vp, ap->a_label);
1477 mac_devfs_update(vp->v_mount, de, vp);
1484 devfs_stat_f(struct file *fp, struct stat *sb, struct ucred *cred, struct thread *td)
1487 return (vnops.fo_stat(fp, sb, cred, td));
1491 devfs_symlink(struct vop_symlink_args *ap)
1494 struct devfs_dirent *dd;
1495 struct devfs_dirent *de;
1496 struct devfs_mount *dmp;
1498 error = priv_check(curthread, PRIV_DEVFS_SYMLINK);
1501 dmp = VFSTODEVFS(ap->a_dvp->v_mount);
1502 dd = ap->a_dvp->v_data;
1503 de = devfs_newdirent(ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen);
1507 de->de_inode = alloc_unr(devfs_inos);
1508 de->de_dirent->d_type = DT_LNK;
1509 i = strlen(ap->a_target) + 1;
1510 de->de_symlink = malloc(i, M_DEVFS, M_WAITOK);
1511 bcopy(ap->a_target, de->de_symlink, i);
1512 sx_xlock(&dmp->dm_lock);
1514 mac_devfs_create_symlink(ap->a_cnp->cn_cred, dmp->dm_mount, dd, de);
1516 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
1517 devfs_rules_apply(dmp, de);
1518 return (devfs_allocv(de, ap->a_dvp->v_mount, LK_EXCLUSIVE, ap->a_vpp));
1522 devfs_truncate_f(struct file *fp, off_t length, struct ucred *cred, struct thread *td)
1525 return (vnops.fo_truncate(fp, length, cred, td));
1530 devfs_write_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
1533 int error, ioflag, ref, resid;
1538 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1541 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td));
1542 ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT | O_FSYNC);
1543 if (ioflag & O_DIRECT)
1544 ioflag |= IO_DIRECT;
1545 if ((flags & FOF_OFFSET) == 0)
1546 uio->uio_offset = fp->f_offset;
1548 resid = uio->uio_resid;
1550 error = dsw->d_write(dev, uio, ioflag);
1551 if (uio->uio_resid != resid || (error == 0 && resid != 0)) {
1552 vfs_timestamp(&dev->si_ctime);
1553 dev->si_mtime = dev->si_ctime;
1556 dev_relthread(dev, ref);
1558 if ((flags & FOF_OFFSET) == 0)
1559 fp->f_offset = uio->uio_offset;
1560 fp->f_nextoff = uio->uio_offset;
1565 dev2udev(struct cdev *x)
1569 return (cdev2priv(x)->cdp_inode);
1572 static struct fileops devfs_ops_f = {
1573 .fo_read = devfs_read_f,
1574 .fo_write = devfs_write_f,
1575 .fo_truncate = devfs_truncate_f,
1576 .fo_ioctl = devfs_ioctl_f,
1577 .fo_poll = devfs_poll_f,
1578 .fo_kqfilter = devfs_kqfilter_f,
1579 .fo_stat = devfs_stat_f,
1580 .fo_close = devfs_close_f,
1581 .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
1584 static struct vop_vector devfs_vnodeops = {
1585 .vop_default = &default_vnodeops,
1587 .vop_access = devfs_access,
1588 .vop_getattr = devfs_getattr,
1589 .vop_ioctl = devfs_rioctl,
1590 .vop_lookup = devfs_lookup,
1591 .vop_mknod = devfs_mknod,
1592 .vop_pathconf = devfs_pathconf,
1593 .vop_read = devfs_rread,
1594 .vop_readdir = devfs_readdir,
1595 .vop_readlink = devfs_readlink,
1596 .vop_reclaim = devfs_reclaim,
1597 .vop_remove = devfs_remove,
1598 .vop_revoke = devfs_revoke,
1599 .vop_setattr = devfs_setattr,
1601 .vop_setlabel = devfs_setlabel,
1603 .vop_symlink = devfs_symlink,
1604 .vop_vptocnp = devfs_vptocnp,
1607 static struct vop_vector devfs_specops = {
1608 .vop_default = &default_vnodeops,
1610 .vop_access = devfs_access,
1611 .vop_bmap = VOP_PANIC,
1612 .vop_close = devfs_close,
1613 .vop_create = VOP_PANIC,
1614 .vop_fsync = devfs_fsync,
1615 .vop_getattr = devfs_getattr,
1616 .vop_link = VOP_PANIC,
1617 .vop_mkdir = VOP_PANIC,
1618 .vop_mknod = VOP_PANIC,
1619 .vop_open = devfs_open,
1620 .vop_pathconf = devfs_pathconf,
1621 .vop_print = devfs_print,
1622 .vop_read = VOP_PANIC,
1623 .vop_readdir = VOP_PANIC,
1624 .vop_readlink = VOP_PANIC,
1625 .vop_reallocblks = VOP_PANIC,
1626 .vop_reclaim = devfs_reclaim,
1627 .vop_remove = devfs_remove,
1628 .vop_rename = VOP_PANIC,
1629 .vop_revoke = devfs_revoke,
1630 .vop_rmdir = VOP_PANIC,
1631 .vop_setattr = devfs_setattr,
1633 .vop_setlabel = devfs_setlabel,
1635 .vop_strategy = VOP_PANIC,
1636 .vop_symlink = VOP_PANIC,
1637 .vop_vptocnp = devfs_vptocnp,
1638 .vop_write = VOP_PANIC,
1642 * Our calling convention to the device drivers used to be that we passed
1643 * vnode.h IO_* flags to read()/write(), but we're moving to fcntl.h O_
1644 * flags instead since that's what open(), close() and ioctl() takes and
1645 * we don't really want vnode.h in device drivers.
1646 * We solved the source compatibility by redefining some vnode flags to
1647 * be the same as the fcntl ones and by sending down the bitwise OR of
1648 * the respective fcntl/vnode flags. These CTASSERTS make sure nobody
1649 * pulls the rug out under this.
1651 CTASSERT(O_NONBLOCK == IO_NDELAY);
1652 CTASSERT(O_FSYNC == IO_SYNC);