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.
257 * If a NULL cnp is provided, no '/' is appended to the resulting path.
260 devfs_fqpn(char *buf, struct devfs_mount *dmp, struct devfs_dirent *dd,
261 struct componentname *cnp)
264 struct devfs_dirent *de;
266 sx_assert(&dmp->dm_lock, SA_LOCKED);
271 i -= cnp->cn_namelen;
275 bcopy(cnp->cn_nameptr, buf + i, cnp->cn_namelen);
277 while (de != dmp->dm_rootdir) {
278 if (cnp != NULL || i < SPECNAMELEN) {
284 i -= de->de_dirent->d_namlen;
287 bcopy(de->de_dirent->d_name, buf + i,
288 de->de_dirent->d_namlen);
289 de = devfs_parent_dirent(de);
297 devfs_allocv_drop_refs(int drop_dm_lock, struct devfs_mount *dmp,
298 struct devfs_dirent *de)
303 if (de->de_flags & DE_DOOMED)
305 if (DEVFS_DE_DROP(de)) {
306 KASSERT(not_found == 1, ("DEVFS de dropped but not doomed"));
307 devfs_dirent_free(de);
309 if (DEVFS_DMP_DROP(dmp)) {
310 KASSERT(not_found == 1,
311 ("DEVFS mount struct freed before dirent"));
313 sx_xunlock(&dmp->dm_lock);
314 devfs_unmount_final(dmp);
316 if (not_found == 1 || (drop_dm_lock && not_found != 2))
317 sx_unlock(&dmp->dm_lock);
322 devfs_insmntque_dtr(struct vnode *vp, void *arg)
324 struct devfs_dirent *de;
326 de = (struct devfs_dirent *)arg;
327 mtx_lock(&devfs_de_interlock);
330 mtx_unlock(&devfs_de_interlock);
336 * devfs_allocv shall be entered with dmp->dm_lock held, and it drops
340 devfs_allocv(struct devfs_dirent *de, struct mount *mp, int lockmode,
346 struct devfs_mount *dmp;
349 dmp = VFSTODEVFS(mp);
350 if (de->de_flags & DE_DOOMED) {
351 sx_xunlock(&dmp->dm_lock);
357 mtx_lock(&devfs_de_interlock);
361 mtx_unlock(&devfs_de_interlock);
362 sx_xunlock(&dmp->dm_lock);
363 vget(vp, lockmode | LK_INTERLOCK | LK_RETRY, curthread);
364 sx_xlock(&dmp->dm_lock);
365 if (devfs_allocv_drop_refs(0, dmp, de)) {
369 else if ((vp->v_iflag & VI_DOOMED) != 0) {
370 mtx_lock(&devfs_de_interlock);
371 if (de->de_vnode == vp) {
375 mtx_unlock(&devfs_de_interlock);
379 sx_xunlock(&dmp->dm_lock);
383 mtx_unlock(&devfs_de_interlock);
384 if (de->de_dirent->d_type == DT_CHR) {
385 if (!(de->de_cdp->cdp_flags & CDP_ACTIVE)) {
386 devfs_allocv_drop_refs(1, dmp, de);
389 dev = &de->de_cdp->cdp_c;
393 error = getnewvnode("devfs", mp, &devfs_vnodeops, &vp);
395 devfs_allocv_drop_refs(1, dmp, de);
396 printf("devfs_allocv: failed to allocate new vnode\n");
400 if (de->de_dirent->d_type == DT_CHR) {
405 /* XXX: v_rdev should be protect by vnode lock */
407 KASSERT(vp->v_usecount == 1,
408 ("%s %d (%d)\n", __func__, __LINE__, vp->v_usecount));
409 dev->si_usecount += vp->v_usecount;
410 /* Special casing of ttys for deadfs. Probably redundant. */
412 if (dsw != NULL && (dsw->d_flags & D_TTY) != 0)
413 vp->v_vflag |= VV_ISTTY;
416 if ((dev->si_flags & SI_ETERNAL) != 0)
417 vp->v_vflag |= VV_ETERNALDEV;
418 vp->v_op = &devfs_specops;
419 } else if (de->de_dirent->d_type == DT_DIR) {
421 } else if (de->de_dirent->d_type == DT_LNK) {
427 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_NOWITNESS);
428 mtx_lock(&devfs_de_interlock);
431 mtx_unlock(&devfs_de_interlock);
432 error = insmntque1(vp, mp, devfs_insmntque_dtr, de);
434 (void) devfs_allocv_drop_refs(1, dmp, de);
437 if (devfs_allocv_drop_refs(0, dmp, de)) {
442 mac_devfs_vnode_associate(mp, de, vp);
444 sx_xunlock(&dmp->dm_lock);
450 devfs_access(struct vop_access_args *ap)
452 struct vnode *vp = ap->a_vp;
453 struct devfs_dirent *de;
457 if (vp->v_type == VDIR)
460 error = vaccess(vp->v_type, de->de_mode, de->de_uid, de->de_gid,
461 ap->a_accmode, ap->a_cred, NULL);
466 /* We do, however, allow access to the controlling terminal */
467 if (!(ap->a_td->td_proc->p_flag & P_CONTROLT))
469 if (ap->a_td->td_proc->p_session->s_ttyvp == de->de_vnode)
476 devfs_close(struct vop_close_args *ap)
478 struct vnode *vp = ap->a_vp, *oldvp;
479 struct thread *td = ap->a_td;
480 struct cdev *dev = vp->v_rdev;
482 int vp_locked, error, ref;
485 * XXX: Don't call d_close() if we were called because of
486 * XXX: insmntque1() failure.
488 if (vp->v_data == NULL)
492 * Hack: a tty device that is a controlling terminal
493 * has a reference from the session structure.
494 * We cannot easily tell that a character device is
495 * a controlling terminal, unless it is the closing
496 * process' controlling terminal. In that case,
497 * if the reference count is 2 (this last descriptor
498 * plus the session), release the reference from the session.
501 sx_xlock(&proctree_lock);
502 if (td && vp == td->td_proc->p_session->s_ttyvp) {
503 SESS_LOCK(td->td_proc->p_session);
505 if (count_dev(dev) == 2 && (vp->v_iflag & VI_DOOMED) == 0) {
506 td->td_proc->p_session->s_ttyvp = NULL;
510 SESS_UNLOCK(td->td_proc->p_session);
512 sx_xunlock(&proctree_lock);
516 * We do not want to really close the device if it
517 * is still in use unless we are trying to close it
518 * forcibly. Since every use (buffer, vnode, swap, cmap)
519 * holds a reference to the vnode, and because we mark
520 * any other vnodes that alias this device, when the
521 * sum of the reference counts on all the aliased
522 * vnodes descends to one, we are on last close.
524 dsw = dev_refthread(dev, &ref);
528 if (vp->v_iflag & VI_DOOMED) {
530 } else if (dsw->d_flags & D_TRACKCLOSE) {
531 /* Keep device updated on status. */
532 } else if (count_dev(dev) > 1) {
534 dev_relthread(dev, ref);
539 vp_locked = VOP_ISLOCKED(vp);
541 KASSERT(dev->si_refcount > 0,
542 ("devfs_close() on un-referenced struct cdev *(%s)", devtoname(dev)));
543 error = dsw->d_close(dev, ap->a_fflag, S_IFCHR, td);
544 dev_relthread(dev, ref);
545 vn_lock(vp, vp_locked | LK_RETRY);
551 devfs_close_f(struct file *fp, struct thread *td)
557 * NB: td may be NULL if this descriptor is closed due to
558 * garbage collection from a closed UNIX domain socket.
560 fpop = curthread->td_fpop;
561 curthread->td_fpop = fp;
562 error = vnops.fo_close(fp, td);
563 curthread->td_fpop = fpop;
566 * The f_cdevpriv cannot be assigned non-NULL value while we
567 * are destroying the file.
569 if (fp->f_cdevpriv != NULL)
575 devfs_fsync(struct vop_fsync_args *ap)
579 struct devfs_dirent *de;
581 if (!vn_isdisk(ap->a_vp, &error)) {
582 bo = &ap->a_vp->v_bufobj;
583 de = ap->a_vp->v_data;
584 if (error == ENXIO && bo->bo_dirty.bv_cnt > 0) {
585 printf("Device %s went missing before all of the data "
586 "could be written to it; expect data loss.\n",
587 de->de_dirent->d_name);
589 error = vop_stdfsync(ap);
590 if (bo->bo_dirty.bv_cnt != 0 || error != 0)
591 panic("devfs_fsync: vop_stdfsync failed.");
597 return (vop_stdfsync(ap));
601 devfs_getattr(struct vop_getattr_args *ap)
603 struct vnode *vp = ap->a_vp;
604 struct vattr *vap = ap->a_vap;
606 struct devfs_dirent *de;
610 KASSERT(de != NULL, ("Null dirent in devfs_getattr vp=%p", vp));
611 if (vp->v_type == VDIR) {
614 ("Null dir dirent in devfs_getattr vp=%p", vp));
616 vap->va_uid = de->de_uid;
617 vap->va_gid = de->de_gid;
618 vap->va_mode = de->de_mode;
619 if (vp->v_type == VLNK)
620 vap->va_size = strlen(de->de_symlink);
621 else if (vp->v_type == VDIR)
622 vap->va_size = vap->va_bytes = DEV_BSIZE;
625 if (vp->v_type != VDIR)
627 vap->va_blocksize = DEV_BSIZE;
628 vap->va_type = vp->v_type;
632 if ((aa).tv_sec <= 3600) { \
633 (aa).tv_sec = boottime.tv_sec; \
634 (aa).tv_nsec = boottime.tv_usec * 1000; \
638 if (vp->v_type != VCHR) {
640 vap->va_atime = de->de_atime;
642 vap->va_mtime = de->de_mtime;
644 vap->va_ctime = de->de_ctime;
648 vap->va_atime = dev->si_atime;
650 vap->va_mtime = dev->si_mtime;
652 vap->va_ctime = dev->si_ctime;
654 vap->va_rdev = cdev2priv(dev)->cdp_inode;
659 vap->va_nlink = de->de_links;
660 vap->va_fileid = de->de_inode;
667 devfs_ioctl_f(struct file *fp, u_long com, void *data, struct ucred *cred, struct thread *td)
675 struct fiodgname_arg *fgn;
679 error = devfs_fp_check(fp, &dev, &dsw, &ref);
683 if (com == FIODTYPE) {
684 *(int *)data = dsw->d_flags & D_TYPEMASK;
686 dev_relthread(dev, ref);
688 } else if (com == FIODGNAME) {
695 error = copyout(p, fgn->buf, i);
697 dev_relthread(dev, ref);
700 error = dsw->d_ioctl(dev, com, data, fp->f_flag, td);
702 dev_relthread(dev, ref);
703 if (error == ENOIOCTL)
705 if (error == 0 && com == TIOCSCTTY) {
708 /* Do nothing if reassigning same control tty */
709 sx_slock(&proctree_lock);
710 if (td->td_proc->p_session->s_ttyvp == vp) {
711 sx_sunlock(&proctree_lock);
715 vpold = td->td_proc->p_session->s_ttyvp;
717 SESS_LOCK(td->td_proc->p_session);
718 td->td_proc->p_session->s_ttyvp = vp;
719 SESS_UNLOCK(td->td_proc->p_session);
721 sx_sunlock(&proctree_lock);
723 /* Get rid of reference to old control tty */
732 devfs_kqfilter_f(struct file *fp, struct knote *kn)
742 error = devfs_fp_check(fp, &dev, &dsw, &ref);
745 error = dsw->d_kqfilter(dev, kn);
747 dev_relthread(dev, ref);
752 devfs_prison_check(struct devfs_dirent *de, struct ucred *tcr)
754 struct cdev_priv *cdp;
760 dcr = cdp->cdp_c.si_cred;
764 return (prison_check(tcr, dcr));
768 devfs_lookupx(struct vop_lookup_args *ap, int *dm_unlock)
770 struct componentname *cnp;
771 struct vnode *dvp, **vpp;
773 struct devfs_dirent *de, *dd;
774 struct devfs_dirent **dde;
775 struct devfs_mount *dmp;
777 int error, flags, nameiop, dvplocked;
778 char specname[SPECNAMELEN + 1], *pname;
783 pname = cnp->cn_nameptr;
785 flags = cnp->cn_flags;
786 nameiop = cnp->cn_nameiop;
787 dmp = VFSTODEVFS(dvp->v_mount);
791 if ((flags & ISLASTCN) && nameiop == RENAME)
794 if (dvp->v_type != VDIR)
797 if ((flags & ISDOTDOT) && (dvp->v_vflag & VV_ROOT))
800 error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, td);
804 if (cnp->cn_namelen == 1 && *pname == '.') {
805 if ((flags & ISLASTCN) && nameiop != LOOKUP)
812 if (flags & ISDOTDOT) {
813 if ((flags & ISLASTCN) && nameiop != LOOKUP)
815 de = devfs_parent_dirent(dd);
818 dvplocked = VOP_ISLOCKED(dvp);
820 error = devfs_allocv(de, dvp->v_mount,
821 cnp->cn_lkflags & LK_TYPE_MASK, vpp);
823 vn_lock(dvp, dvplocked | LK_RETRY);
829 if (DEVFS_DMP_DROP(dmp)) {
831 sx_xunlock(&dmp->dm_lock);
832 devfs_unmount_final(dmp);
836 de = devfs_find(dd, cnp->cn_nameptr, cnp->cn_namelen);
837 while (de == NULL) { /* While(...) so we can use break */
839 if (nameiop == DELETE)
843 * OK, we didn't have an entry for the name we were asked for
844 * so we try to see if anybody can create it on demand.
846 pname = devfs_fqpn(specname, dmp, dd, cnp);
852 sx_xunlock(&dmp->dm_lock);
853 sx_slock(&clone_drain_lock);
854 EVENTHANDLER_INVOKE(dev_clone,
855 td->td_ucred, pname, strlen(pname), &cdev);
856 sx_sunlock(&clone_drain_lock);
857 sx_xlock(&dmp->dm_lock);
858 if (DEVFS_DMP_DROP(dmp)) {
860 sx_xunlock(&dmp->dm_lock);
861 devfs_unmount_final(dmp);
871 if (DEVFS_DMP_DROP(dmp)) {
873 sx_xunlock(&dmp->dm_lock);
874 devfs_unmount_final(dmp);
880 dde = &cdev2priv(cdev)->cdp_dirents[dmp->dm_idx];
881 if (dde != NULL && *dde != NULL)
888 if (de == NULL || de->de_flags & DE_WHITEOUT) {
889 if ((nameiop == CREATE || nameiop == RENAME) &&
890 (flags & (LOCKPARENT | WANTPARENT)) && (flags & ISLASTCN)) {
891 cnp->cn_flags |= SAVENAME;
892 return (EJUSTRETURN);
897 if (devfs_prison_check(de, td->td_ucred))
900 if ((cnp->cn_nameiop == DELETE) && (flags & ISLASTCN)) {
901 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, td);
910 error = devfs_allocv(de, dvp->v_mount, cnp->cn_lkflags & LK_TYPE_MASK,
917 devfs_lookup(struct vop_lookup_args *ap)
920 struct devfs_mount *dmp;
923 dmp = VFSTODEVFS(ap->a_dvp->v_mount);
925 sx_xlock(&dmp->dm_lock);
926 j = devfs_lookupx(ap, &dm_unlock);
928 sx_xunlock(&dmp->dm_lock);
933 devfs_mknod(struct vop_mknod_args *ap)
935 struct componentname *cnp;
936 struct vnode *dvp, **vpp;
937 struct devfs_dirent *dd, *de;
938 struct devfs_mount *dmp;
942 * The only type of node we should be creating here is a
943 * character device, for anything else return EOPNOTSUPP.
945 if (ap->a_vap->va_type != VCHR)
948 dmp = VFSTODEVFS(dvp->v_mount);
955 sx_xlock(&dmp->dm_lock);
956 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
957 if (cnp->cn_namelen != de->de_dirent->d_namlen)
959 if (bcmp(cnp->cn_nameptr, de->de_dirent->d_name,
960 de->de_dirent->d_namlen) != 0)
962 if (de->de_flags & DE_WHITEOUT)
968 de->de_flags &= ~DE_WHITEOUT;
969 error = devfs_allocv(de, dvp->v_mount, LK_EXCLUSIVE, vpp);
972 sx_xunlock(&dmp->dm_lock);
978 devfs_open(struct vop_open_args *ap)
980 struct thread *td = ap->a_td;
981 struct vnode *vp = ap->a_vp;
982 struct cdev *dev = vp->v_rdev;
983 struct file *fp = ap->a_fp;
984 int error, ref, vlocked;
988 if (vp->v_type == VBLK)
994 /* Make this field valid before any I/O in d_open. */
995 if (dev->si_iosize_max == 0)
996 dev->si_iosize_max = DFLTPHYS;
998 dsw = dev_refthread(dev, &ref);
1001 if (fp == NULL && dsw->d_fdopen != NULL) {
1002 dev_relthread(dev, ref);
1006 vlocked = VOP_ISLOCKED(vp);
1015 if (dsw->d_fdopen != NULL)
1016 error = dsw->d_fdopen(dev, ap->a_mode, td, fp);
1018 error = dsw->d_open(dev, ap->a_mode, S_IFCHR, td);
1021 vn_lock(vp, vlocked | LK_RETRY);
1022 dev_relthread(dev, ref);
1024 if (error == ERESTART)
1029 #if 0 /* /dev/console */
1030 KASSERT(fp != NULL, ("Could not vnode bypass device on NULL fp"));
1035 if (fp->f_ops == &badfileops)
1036 finit(fp, fp->f_flag, DTYPE_VNODE, dev, &devfs_ops_f);
1041 devfs_pathconf(struct vop_pathconf_args *ap)
1044 switch (ap->a_name) {
1045 case _PC_MAC_PRESENT:
1048 * If MAC is enabled, devfs automatically supports
1049 * trivial non-persistant label storage.
1057 return (vop_stdpathconf(ap));
1064 devfs_poll_f(struct file *fp, int events, struct ucred *cred, struct thread *td)
1072 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1074 return (poll_no_poll(events));
1075 error = dsw->d_poll(dev, events, td);
1077 dev_relthread(dev, ref);
1082 * Print out the contents of a special device vnode.
1085 devfs_print(struct vop_print_args *ap)
1088 printf("\tdev %s\n", devtoname(ap->a_vp->v_rdev));
1094 devfs_read_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
1097 int ioflag, error, ref, resid;
1102 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1105 resid = uio->uio_resid;
1106 ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT);
1107 if (ioflag & O_DIRECT)
1108 ioflag |= IO_DIRECT;
1110 if ((flags & FOF_OFFSET) == 0)
1111 uio->uio_offset = fp->f_offset;
1113 error = dsw->d_read(dev, uio, ioflag);
1114 if (uio->uio_resid != resid || (error == 0 && resid != 0))
1115 vfs_timestamp(&dev->si_atime);
1117 dev_relthread(dev, ref);
1119 if ((flags & FOF_OFFSET) == 0)
1120 fp->f_offset = uio->uio_offset;
1121 fp->f_nextoff = uio->uio_offset;
1126 devfs_readdir(struct vop_readdir_args *ap)
1131 struct devfs_dirent *dd;
1132 struct devfs_dirent *de;
1133 struct devfs_mount *dmp;
1135 int *tmp_ncookies = NULL;
1137 if (ap->a_vp->v_type != VDIR)
1141 if (uio->uio_offset < 0)
1145 * XXX: This is a temporary hack to get around this filesystem not
1146 * supporting cookies. We store the location of the ncookies pointer
1147 * in a temporary variable before calling vfs_subr.c:vfs_read_dirent()
1148 * and set the number of cookies to 0. We then set the pointer to
1149 * NULL so that vfs_read_dirent doesn't try to call realloc() on
1150 * ap->a_cookies. Later in this function, we restore the ap->a_ncookies
1151 * pointer to its original location before returning to the caller.
1153 if (ap->a_ncookies != NULL) {
1154 tmp_ncookies = ap->a_ncookies;
1155 *ap->a_ncookies = 0;
1156 ap->a_ncookies = NULL;
1159 dmp = VFSTODEVFS(ap->a_vp->v_mount);
1160 sx_xlock(&dmp->dm_lock);
1161 DEVFS_DMP_HOLD(dmp);
1162 devfs_populate(dmp);
1163 if (DEVFS_DMP_DROP(dmp)) {
1164 sx_xunlock(&dmp->dm_lock);
1165 devfs_unmount_final(dmp);
1166 if (tmp_ncookies != NULL)
1167 ap->a_ncookies = tmp_ncookies;
1171 de = ap->a_vp->v_data;
1173 TAILQ_FOREACH(dd, &de->de_dlist, de_list) {
1174 KASSERT(dd->de_cdp != (void *)0xdeadc0de, ("%s %d\n", __func__, __LINE__));
1175 if (dd->de_flags & DE_WHITEOUT)
1177 if (devfs_prison_check(dd, ap->a_cred))
1179 if (dd->de_dirent->d_type == DT_DIR)
1184 if (dp->d_reclen > uio->uio_resid)
1186 dp->d_fileno = de->de_inode;
1187 if (off >= uio->uio_offset) {
1188 error = vfs_read_dirent(ap, dp, off);
1192 off += dp->d_reclen;
1194 sx_xunlock(&dmp->dm_lock);
1195 uio->uio_offset = off;
1198 * Restore ap->a_ncookies if it wasn't originally NULL in the first
1201 if (tmp_ncookies != NULL)
1202 ap->a_ncookies = tmp_ncookies;
1208 devfs_readlink(struct vop_readlink_args *ap)
1210 struct devfs_dirent *de;
1212 de = ap->a_vp->v_data;
1213 return (uiomove(de->de_symlink, strlen(de->de_symlink), ap->a_uio));
1217 devfs_reclaim(struct vop_reclaim_args *ap)
1219 struct vnode *vp = ap->a_vp;
1220 struct devfs_dirent *de;
1223 mtx_lock(&devfs_de_interlock);
1226 de->de_vnode = NULL;
1229 mtx_unlock(&devfs_de_interlock);
1231 vnode_destroy_vobject(vp);
1244 dev->si_usecount -= vp->v_usecount;
1252 devfs_remove(struct vop_remove_args *ap)
1254 struct vnode *vp = ap->a_vp;
1255 struct devfs_dirent *dd;
1256 struct devfs_dirent *de;
1257 struct devfs_mount *dmp = VFSTODEVFS(vp->v_mount);
1259 sx_xlock(&dmp->dm_lock);
1260 dd = ap->a_dvp->v_data;
1262 if (de->de_cdp == NULL) {
1263 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
1264 devfs_delete(dmp, de, 1);
1266 de->de_flags |= DE_WHITEOUT;
1268 sx_xunlock(&dmp->dm_lock);
1273 * Revoke is called on a tty when a terminal session ends. The vnode
1274 * is orphaned by setting v_op to deadfs so we need to let go of it
1275 * as well so that we create a new one next time around.
1279 devfs_revoke(struct vop_revoke_args *ap)
1281 struct vnode *vp = ap->a_vp, *vp2;
1283 struct cdev_priv *cdp;
1284 struct devfs_dirent *de;
1287 KASSERT((ap->a_flags & REVOKEALL) != 0, ("devfs_revoke !REVOKEALL"));
1290 cdp = cdev2priv(dev);
1303 mtx_lock(&devfs_de_interlock);
1306 for (i = 0; i <= cdp->cdp_maxdirent; i++) {
1307 de = cdp->cdp_dirents[i];
1315 mtx_unlock(&devfs_de_interlock);
1316 if (vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK,
1330 mtx_unlock(&devfs_de_interlock);
1335 if (!(cdp->cdp_flags & CDP_ACTIVE) && cdp->cdp_inuse == 0) {
1336 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
1338 dev_rel(&cdp->cdp_c);
1342 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1347 devfs_rioctl(struct vop_ioctl_args *ap)
1350 struct devfs_mount *dmp;
1354 vn_lock(vp, LK_SHARED | LK_RETRY);
1355 if (vp->v_iflag & VI_DOOMED) {
1359 dmp = VFSTODEVFS(vp->v_mount);
1360 sx_xlock(&dmp->dm_lock);
1362 DEVFS_DMP_HOLD(dmp);
1363 devfs_populate(dmp);
1364 if (DEVFS_DMP_DROP(dmp)) {
1365 sx_xunlock(&dmp->dm_lock);
1366 devfs_unmount_final(dmp);
1369 error = devfs_rules_ioctl(dmp, ap->a_command, ap->a_data, ap->a_td);
1370 sx_xunlock(&dmp->dm_lock);
1375 devfs_rread(struct vop_read_args *ap)
1378 if (ap->a_vp->v_type != VDIR)
1380 return (VOP_READDIR(ap->a_vp, ap->a_uio, ap->a_cred, NULL, NULL, NULL));
1384 devfs_setattr(struct vop_setattr_args *ap)
1386 struct devfs_dirent *de;
1397 if ((vap->va_type != VNON) ||
1398 (vap->va_nlink != VNOVAL) ||
1399 (vap->va_fsid != VNOVAL) ||
1400 (vap->va_fileid != VNOVAL) ||
1401 (vap->va_blocksize != VNOVAL) ||
1402 (vap->va_flags != VNOVAL && vap->va_flags != 0) ||
1403 (vap->va_rdev != VNOVAL) ||
1404 ((int)vap->va_bytes != VNOVAL) ||
1405 (vap->va_gen != VNOVAL)) {
1410 if (vp->v_type == VDIR)
1414 if (vap->va_uid == (uid_t)VNOVAL)
1418 if (vap->va_gid == (gid_t)VNOVAL)
1422 if (uid != de->de_uid || gid != de->de_gid) {
1423 if ((ap->a_cred->cr_uid != de->de_uid) || uid != de->de_uid ||
1424 (gid != de->de_gid && !groupmember(gid, ap->a_cred))) {
1425 error = priv_check(td, PRIV_VFS_CHOWN);
1434 if (vap->va_mode != (mode_t)VNOVAL) {
1435 if (ap->a_cred->cr_uid != de->de_uid) {
1436 error = priv_check(td, PRIV_VFS_ADMIN);
1440 de->de_mode = vap->va_mode;
1444 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1445 /* See the comment in ufs_vnops::ufs_setattr(). */
1446 if ((error = VOP_ACCESS(vp, VADMIN, ap->a_cred, td)) &&
1447 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1448 (error = VOP_ACCESS(vp, VWRITE, ap->a_cred, td))))
1450 if (vap->va_atime.tv_sec != VNOVAL) {
1451 if (vp->v_type == VCHR)
1452 vp->v_rdev->si_atime = vap->va_atime;
1454 de->de_atime = vap->va_atime;
1456 if (vap->va_mtime.tv_sec != VNOVAL) {
1457 if (vp->v_type == VCHR)
1458 vp->v_rdev->si_mtime = vap->va_mtime;
1460 de->de_mtime = vap->va_mtime;
1466 if (vp->v_type == VCHR)
1467 vfs_timestamp(&vp->v_rdev->si_ctime);
1469 vfs_timestamp(&de->de_mtime);
1476 devfs_setlabel(struct vop_setlabel_args *ap)
1479 struct devfs_dirent *de;
1484 mac_vnode_relabel(ap->a_cred, vp, ap->a_label);
1485 mac_devfs_update(vp->v_mount, de, vp);
1492 devfs_stat_f(struct file *fp, struct stat *sb, struct ucred *cred, struct thread *td)
1495 return (vnops.fo_stat(fp, sb, cred, td));
1499 devfs_symlink(struct vop_symlink_args *ap)
1502 struct devfs_dirent *dd;
1503 struct devfs_dirent *de;
1504 struct devfs_mount *dmp;
1506 error = priv_check(curthread, PRIV_DEVFS_SYMLINK);
1509 dmp = VFSTODEVFS(ap->a_dvp->v_mount);
1510 dd = ap->a_dvp->v_data;
1511 de = devfs_newdirent(ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen);
1515 de->de_inode = alloc_unr(devfs_inos);
1517 de->de_dirent->d_type = DT_LNK;
1518 i = strlen(ap->a_target) + 1;
1519 de->de_symlink = malloc(i, M_DEVFS, M_WAITOK);
1520 bcopy(ap->a_target, de->de_symlink, i);
1521 sx_xlock(&dmp->dm_lock);
1523 mac_devfs_create_symlink(ap->a_cnp->cn_cred, dmp->dm_mount, dd, de);
1525 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
1526 devfs_rules_apply(dmp, de);
1527 return (devfs_allocv(de, ap->a_dvp->v_mount, LK_EXCLUSIVE, ap->a_vpp));
1531 devfs_truncate_f(struct file *fp, off_t length, struct ucred *cred, struct thread *td)
1534 return (vnops.fo_truncate(fp, length, cred, td));
1539 devfs_write_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
1542 int error, ioflag, ref, resid;
1547 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1550 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td));
1551 ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT | O_FSYNC);
1552 if (ioflag & O_DIRECT)
1553 ioflag |= IO_DIRECT;
1554 if ((flags & FOF_OFFSET) == 0)
1555 uio->uio_offset = fp->f_offset;
1557 resid = uio->uio_resid;
1559 error = dsw->d_write(dev, uio, ioflag);
1560 if (uio->uio_resid != resid || (error == 0 && resid != 0)) {
1561 vfs_timestamp(&dev->si_ctime);
1562 dev->si_mtime = dev->si_ctime;
1565 dev_relthread(dev, ref);
1567 if ((flags & FOF_OFFSET) == 0)
1568 fp->f_offset = uio->uio_offset;
1569 fp->f_nextoff = uio->uio_offset;
1574 dev2udev(struct cdev *x)
1578 return (cdev2priv(x)->cdp_inode);
1581 static struct fileops devfs_ops_f = {
1582 .fo_read = devfs_read_f,
1583 .fo_write = devfs_write_f,
1584 .fo_truncate = devfs_truncate_f,
1585 .fo_ioctl = devfs_ioctl_f,
1586 .fo_poll = devfs_poll_f,
1587 .fo_kqfilter = devfs_kqfilter_f,
1588 .fo_stat = devfs_stat_f,
1589 .fo_close = devfs_close_f,
1590 .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
1593 static struct vop_vector devfs_vnodeops = {
1594 .vop_default = &default_vnodeops,
1596 .vop_access = devfs_access,
1597 .vop_getattr = devfs_getattr,
1598 .vop_ioctl = devfs_rioctl,
1599 .vop_lookup = devfs_lookup,
1600 .vop_mknod = devfs_mknod,
1601 .vop_pathconf = devfs_pathconf,
1602 .vop_read = devfs_rread,
1603 .vop_readdir = devfs_readdir,
1604 .vop_readlink = devfs_readlink,
1605 .vop_reclaim = devfs_reclaim,
1606 .vop_remove = devfs_remove,
1607 .vop_revoke = devfs_revoke,
1608 .vop_setattr = devfs_setattr,
1610 .vop_setlabel = devfs_setlabel,
1612 .vop_symlink = devfs_symlink,
1613 .vop_vptocnp = devfs_vptocnp,
1616 static struct vop_vector devfs_specops = {
1617 .vop_default = &default_vnodeops,
1619 .vop_access = devfs_access,
1620 .vop_bmap = VOP_PANIC,
1621 .vop_close = devfs_close,
1622 .vop_create = VOP_PANIC,
1623 .vop_fsync = devfs_fsync,
1624 .vop_getattr = devfs_getattr,
1625 .vop_link = VOP_PANIC,
1626 .vop_mkdir = VOP_PANIC,
1627 .vop_mknod = VOP_PANIC,
1628 .vop_open = devfs_open,
1629 .vop_pathconf = devfs_pathconf,
1630 .vop_print = devfs_print,
1631 .vop_read = VOP_PANIC,
1632 .vop_readdir = VOP_PANIC,
1633 .vop_readlink = VOP_PANIC,
1634 .vop_reallocblks = VOP_PANIC,
1635 .vop_reclaim = devfs_reclaim,
1636 .vop_remove = devfs_remove,
1637 .vop_rename = VOP_PANIC,
1638 .vop_revoke = devfs_revoke,
1639 .vop_rmdir = VOP_PANIC,
1640 .vop_setattr = devfs_setattr,
1642 .vop_setlabel = devfs_setlabel,
1644 .vop_strategy = VOP_PANIC,
1645 .vop_symlink = VOP_PANIC,
1646 .vop_vptocnp = devfs_vptocnp,
1647 .vop_write = VOP_PANIC,
1651 * Our calling convention to the device drivers used to be that we passed
1652 * vnode.h IO_* flags to read()/write(), but we're moving to fcntl.h O_
1653 * flags instead since that's what open(), close() and ioctl() takes and
1654 * we don't really want vnode.h in device drivers.
1655 * We solved the source compatibility by redefining some vnode flags to
1656 * be the same as the fcntl ones and by sending down the bitwise OR of
1657 * the respective fcntl/vnode flags. These CTASSERTS make sure nobody
1658 * pulls the rug out under this.
1660 CTASSERT(O_NONBLOCK == IO_NDELAY);
1661 CTASSERT(O_FSYNC == IO_SYNC);