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)
88 *dswp = devvn_refthread(fp->f_vnode, devp);
89 if (*devp != fp->f_data) {
94 KASSERT((*devp)->si_refcount > 0,
95 ("devfs: un-referenced struct cdev *(%s)", devtoname(*devp)));
98 curthread->td_fpop = fp;
103 devfs_get_cdevpriv(void **datap)
106 struct cdev_privdata *p;
109 fp = curthread->td_fpop;
115 *datap = p->cdpd_data;
122 devfs_set_cdevpriv(void *priv, cdevpriv_dtr_t priv_dtr)
125 struct cdev_priv *cdp;
126 struct cdev_privdata *p;
129 fp = curthread->td_fpop;
132 cdp = cdev2priv((struct cdev *)fp->f_data);
133 p = malloc(sizeof(struct cdev_privdata), M_CDEVPDATA, M_WAITOK);
135 p->cdpd_dtr = priv_dtr;
137 mtx_lock(&cdevpriv_mtx);
138 if (fp->f_cdevpriv == NULL) {
139 LIST_INSERT_HEAD(&cdp->cdp_fdpriv, p, cdpd_list);
141 mtx_unlock(&cdevpriv_mtx);
144 mtx_unlock(&cdevpriv_mtx);
145 free(p, M_CDEVPDATA);
152 devfs_destroy_cdevpriv(struct cdev_privdata *p)
155 mtx_assert(&cdevpriv_mtx, MA_OWNED);
156 p->cdpd_fp->f_cdevpriv = NULL;
157 LIST_REMOVE(p, cdpd_list);
158 mtx_unlock(&cdevpriv_mtx);
159 (p->cdpd_dtr)(p->cdpd_data);
160 free(p, M_CDEVPDATA);
164 devfs_fpdrop(struct file *fp)
166 struct cdev_privdata *p;
168 mtx_lock(&cdevpriv_mtx);
169 if ((p = fp->f_cdevpriv) == NULL) {
170 mtx_unlock(&cdevpriv_mtx);
173 devfs_destroy_cdevpriv(p);
177 devfs_clear_cdevpriv(void)
181 fp = curthread->td_fpop;
188 devfs_vptocnp(struct vop_vptocnp_args *ap)
190 struct vnode *vp = ap->a_vp;
191 struct vnode **dvp = ap->a_vpp;
192 struct devfs_mount *dmp;
193 char *buf = ap->a_buf;
194 int *buflen = ap->a_buflen;
195 struct devfs_dirent *dd, *de;
198 dmp = VFSTODEVFS(vp->v_mount);
203 sx_xlock(&dmp->dm_lock);
205 if (vp->v_type == VCHR) {
206 i -= strlen(dd->de_cdp->cdp_c.si_name);
211 bcopy(dd->de_cdp->cdp_c.si_name, buf + i,
212 strlen(dd->de_cdp->cdp_c.si_name));
214 } else if (vp->v_type == VDIR) {
215 if (dd == dmp->dm_rootdir) {
220 i -= dd->de_dirent->d_namlen;
225 bcopy(dd->de_dirent->d_name, buf + i,
226 dd->de_dirent->d_namlen);
233 de = TAILQ_FIRST(&de->de_dlist); /* "." */
234 de = TAILQ_NEXT(de, de_list); /* ".." */
236 mtx_lock(&devfs_de_interlock);
240 mtx_unlock(&devfs_de_interlock);
244 mtx_unlock(&devfs_de_interlock);
248 sx_xunlock(&dmp->dm_lock);
253 * Construct the fully qualified path name relative to the mountpoint
256 devfs_fqpn(char *buf, struct vnode *dvp, struct componentname *cnp)
259 struct devfs_dirent *de, *dd;
260 struct devfs_mount *dmp;
262 dmp = VFSTODEVFS(dvp->v_mount);
266 i -= cnp->cn_namelen;
269 bcopy(cnp->cn_nameptr, buf + i, cnp->cn_namelen);
271 while (de != dmp->dm_rootdir) {
276 i -= de->de_dirent->d_namlen;
279 bcopy(de->de_dirent->d_name, buf + i,
280 de->de_dirent->d_namlen);
281 de = TAILQ_FIRST(&de->de_dlist); /* "." */
282 de = TAILQ_NEXT(de, de_list); /* ".." */
289 devfs_allocv_drop_refs(int drop_dm_lock, struct devfs_mount *dmp,
290 struct devfs_dirent *de)
295 if (de->de_flags & DE_DOOMED)
297 if (DEVFS_DE_DROP(de)) {
298 KASSERT(not_found == 1, ("DEVFS de dropped but not doomed"));
299 devfs_dirent_free(de);
301 if (DEVFS_DMP_DROP(dmp)) {
302 KASSERT(not_found == 1,
303 ("DEVFS mount struct freed before dirent"));
305 sx_xunlock(&dmp->dm_lock);
306 devfs_unmount_final(dmp);
308 if (not_found == 1 || (drop_dm_lock && not_found != 2))
309 sx_unlock(&dmp->dm_lock);
314 devfs_insmntque_dtr(struct vnode *vp, void *arg)
316 struct devfs_dirent *de;
318 de = (struct devfs_dirent *)arg;
319 mtx_lock(&devfs_de_interlock);
322 mtx_unlock(&devfs_de_interlock);
328 * devfs_allocv shall be entered with dmp->dm_lock held, and it drops
332 devfs_allocv(struct devfs_dirent *de, struct mount *mp, struct vnode **vpp)
337 struct devfs_mount *dmp;
339 dmp = VFSTODEVFS(mp);
340 if (de->de_flags & DE_DOOMED) {
341 sx_xunlock(&dmp->dm_lock);
346 mtx_lock(&devfs_de_interlock);
350 mtx_unlock(&devfs_de_interlock);
351 sx_xunlock(&dmp->dm_lock);
352 error = vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, curthread);
353 sx_xlock(&dmp->dm_lock);
354 if (devfs_allocv_drop_refs(0, dmp, de)) {
360 sx_xunlock(&dmp->dm_lock);
363 sx_xunlock(&dmp->dm_lock);
367 mtx_unlock(&devfs_de_interlock);
368 if (de->de_dirent->d_type == DT_CHR) {
369 if (!(de->de_cdp->cdp_flags & CDP_ACTIVE)) {
370 devfs_allocv_drop_refs(1, dmp, de);
373 dev = &de->de_cdp->cdp_c;
377 error = getnewvnode("devfs", mp, &devfs_vnodeops, &vp);
379 devfs_allocv_drop_refs(1, dmp, de);
380 printf("devfs_allocv: failed to allocate new vnode\n");
384 if (de->de_dirent->d_type == DT_CHR) {
389 /* XXX: v_rdev should be protect by vnode lock */
391 KASSERT(vp->v_usecount == 1,
392 ("%s %d (%d)\n", __func__, __LINE__, vp->v_usecount));
393 dev->si_usecount += vp->v_usecount;
396 vp->v_op = &devfs_specops;
397 } else if (de->de_dirent->d_type == DT_DIR) {
399 } else if (de->de_dirent->d_type == DT_LNK) {
404 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_NOWITNESS);
405 mtx_lock(&devfs_de_interlock);
408 mtx_unlock(&devfs_de_interlock);
409 error = insmntque1(vp, mp, devfs_insmntque_dtr, de);
411 (void) devfs_allocv_drop_refs(1, dmp, de);
414 if (devfs_allocv_drop_refs(0, dmp, de)) {
419 mac_devfs_vnode_associate(mp, de, vp);
421 sx_xunlock(&dmp->dm_lock);
427 devfs_access(struct vop_access_args *ap)
429 struct vnode *vp = ap->a_vp;
430 struct devfs_dirent *de;
434 if (vp->v_type == VDIR)
437 error = vaccess(vp->v_type, de->de_mode, de->de_uid, de->de_gid,
438 ap->a_accmode, ap->a_cred, NULL);
443 /* We do, however, allow access to the controlling terminal */
444 if (!(ap->a_td->td_proc->p_flag & P_CONTROLT))
446 if (ap->a_td->td_proc->p_session->s_ttyvp == de->de_vnode)
453 devfs_close(struct vop_close_args *ap)
455 struct vnode *vp = ap->a_vp, *oldvp;
456 struct thread *td = ap->a_td;
457 struct cdev *dev = vp->v_rdev;
459 int vp_locked, error;
462 * Hack: a tty device that is a controlling terminal
463 * has a reference from the session structure.
464 * We cannot easily tell that a character device is
465 * a controlling terminal, unless it is the closing
466 * process' controlling terminal. In that case,
467 * if the reference count is 2 (this last descriptor
468 * plus the session), release the reference from the session.
471 sx_xlock(&proctree_lock);
472 if (td && vp == td->td_proc->p_session->s_ttyvp) {
473 SESS_LOCK(td->td_proc->p_session);
475 if (count_dev(dev) == 2 && (vp->v_iflag & VI_DOOMED) == 0) {
476 td->td_proc->p_session->s_ttyvp = NULL;
480 SESS_UNLOCK(td->td_proc->p_session);
482 sx_xunlock(&proctree_lock);
486 * We do not want to really close the device if it
487 * is still in use unless we are trying to close it
488 * forcibly. Since every use (buffer, vnode, swap, cmap)
489 * holds a reference to the vnode, and because we mark
490 * any other vnodes that alias this device, when the
491 * sum of the reference counts on all the aliased
492 * vnodes descends to one, we are on last close.
494 dsw = dev_refthread(dev);
498 if (vp->v_iflag & VI_DOOMED) {
500 } else if (dsw->d_flags & D_TRACKCLOSE) {
501 /* Keep device updated on status. */
502 } else if (count_dev(dev) > 1) {
509 vp_locked = VOP_ISLOCKED(vp);
511 KASSERT(dev->si_refcount > 0,
512 ("devfs_close() on un-referenced struct cdev *(%s)", devtoname(dev)));
513 error = dsw->d_close(dev, ap->a_fflag, S_IFCHR, td);
515 vn_lock(vp, vp_locked | LK_RETRY);
521 devfs_close_f(struct file *fp, struct thread *td)
528 error = vnops.fo_close(fp, td);
534 devfs_fsync(struct vop_fsync_args *ap)
538 struct devfs_dirent *de;
540 if (!vn_isdisk(ap->a_vp, &error)) {
541 bo = &ap->a_vp->v_bufobj;
542 de = ap->a_vp->v_data;
543 if (error == ENXIO && bo->bo_dirty.bv_cnt > 0) {
544 printf("Device %s went missing before all of the data "
545 "could be written to it; expect data loss.\n",
546 de->de_dirent->d_name);
548 error = vop_stdfsync(ap);
549 if (bo->bo_dirty.bv_cnt != 0 || error != 0)
550 panic("devfs_fsync: vop_stdfsync failed.");
556 return (vop_stdfsync(ap));
560 devfs_getattr(struct vop_getattr_args *ap)
562 struct vnode *vp = ap->a_vp;
563 struct vattr *vap = ap->a_vap;
565 struct devfs_dirent *de;
569 KASSERT(de != NULL, ("Null dirent in devfs_getattr vp=%p", vp));
570 if (vp->v_type == VDIR) {
573 ("Null dir dirent in devfs_getattr vp=%p", vp));
575 vap->va_uid = de->de_uid;
576 vap->va_gid = de->de_gid;
577 vap->va_mode = de->de_mode;
578 if (vp->v_type == VLNK)
579 vap->va_size = strlen(de->de_symlink);
580 else if (vp->v_type == VDIR)
581 vap->va_size = vap->va_bytes = DEV_BSIZE;
584 if (vp->v_type != VDIR)
586 vap->va_blocksize = DEV_BSIZE;
587 vap->va_type = vp->v_type;
591 if ((aa).tv_sec <= 3600) { \
592 (aa).tv_sec = boottime.tv_sec; \
593 (aa).tv_nsec = boottime.tv_usec * 1000; \
597 if (vp->v_type != VCHR) {
599 vap->va_atime = de->de_atime;
601 vap->va_mtime = de->de_mtime;
603 vap->va_ctime = de->de_ctime;
607 vap->va_atime = dev->si_atime;
609 vap->va_mtime = dev->si_mtime;
611 vap->va_ctime = dev->si_ctime;
613 vap->va_rdev = cdev2priv(dev)->cdp_inode;
618 vap->va_nlink = de->de_links;
619 vap->va_fileid = de->de_inode;
626 devfs_ioctl_f(struct file *fp, u_long com, void *data, struct ucred *cred, struct thread *td)
634 struct fiodgname_arg *fgn;
638 error = devfs_fp_check(fp, &dev, &dsw);
642 if (com == FIODTYPE) {
643 *(int *)data = dsw->d_flags & D_TYPEMASK;
647 } else if (com == FIODGNAME) {
654 error = copyout(p, fgn->buf, i);
659 error = dsw->d_ioctl(dev, com, data, fp->f_flag, td);
662 if (error == ENOIOCTL)
664 if (error == 0 && com == TIOCSCTTY) {
667 /* Do nothing if reassigning same control tty */
668 sx_slock(&proctree_lock);
669 if (td->td_proc->p_session->s_ttyvp == vp) {
670 sx_sunlock(&proctree_lock);
674 vpold = td->td_proc->p_session->s_ttyvp;
676 SESS_LOCK(td->td_proc->p_session);
677 td->td_proc->p_session->s_ttyvp = vp;
678 SESS_UNLOCK(td->td_proc->p_session);
680 sx_sunlock(&proctree_lock);
682 /* Get rid of reference to old control tty */
691 devfs_kqfilter_f(struct file *fp, struct knote *kn)
701 error = devfs_fp_check(fp, &dev, &dsw);
704 error = dsw->d_kqfilter(dev, kn);
711 devfs_prison_check(struct devfs_dirent *de, struct ucred *tcr)
713 struct cdev_priv *cdp;
719 dcr = cdp->cdp_c.si_cred;
723 return (prison_check(tcr, dcr));
727 devfs_lookupx(struct vop_lookup_args *ap, int *dm_unlock)
729 struct componentname *cnp;
730 struct vnode *dvp, **vpp;
732 struct devfs_dirent *de, *dd;
733 struct devfs_dirent **dde;
734 struct devfs_mount *dmp;
736 int error, flags, nameiop;
737 char specname[SPECNAMELEN + 1], *pname;
742 pname = cnp->cn_nameptr;
744 flags = cnp->cn_flags;
745 nameiop = cnp->cn_nameiop;
746 dmp = VFSTODEVFS(dvp->v_mount);
750 if ((flags & ISLASTCN) && nameiop == RENAME)
753 if (dvp->v_type != VDIR)
756 if ((flags & ISDOTDOT) && (dvp->v_vflag & VV_ROOT))
759 error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, td);
763 if (cnp->cn_namelen == 1 && *pname == '.') {
764 if ((flags & ISLASTCN) && nameiop != LOOKUP)
771 if (flags & ISDOTDOT) {
772 if ((flags & ISLASTCN) && nameiop != LOOKUP)
775 de = TAILQ_FIRST(&dd->de_dlist); /* "." */
776 de = TAILQ_NEXT(de, de_list); /* ".." */
778 error = devfs_allocv(de, dvp->v_mount, vpp);
780 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
786 if (DEVFS_DMP_DROP(dmp)) {
788 sx_xunlock(&dmp->dm_lock);
789 devfs_unmount_final(dmp);
793 de = devfs_find(dd, cnp->cn_nameptr, cnp->cn_namelen);
794 while (de == NULL) { /* While(...) so we can use break */
796 if (nameiop == DELETE)
800 * OK, we didn't have an entry for the name we were asked for
801 * so we try to see if anybody can create it on demand.
803 pname = devfs_fqpn(specname, dvp, cnp);
809 sx_xunlock(&dmp->dm_lock);
810 sx_slock(&clone_drain_lock);
811 EVENTHANDLER_INVOKE(dev_clone,
812 td->td_ucred, pname, strlen(pname), &cdev);
813 sx_sunlock(&clone_drain_lock);
814 sx_xlock(&dmp->dm_lock);
815 if (DEVFS_DMP_DROP(dmp)) {
817 sx_xunlock(&dmp->dm_lock);
818 devfs_unmount_final(dmp);
826 if (DEVFS_DMP_DROP(dmp)) {
828 sx_xunlock(&dmp->dm_lock);
829 devfs_unmount_final(dmp);
834 dde = &cdev2priv(cdev)->cdp_dirents[dmp->dm_idx];
835 if (dde != NULL && *dde != NULL)
842 if (de == NULL || de->de_flags & DE_WHITEOUT) {
843 if ((nameiop == CREATE || nameiop == RENAME) &&
844 (flags & (LOCKPARENT | WANTPARENT)) && (flags & ISLASTCN)) {
845 cnp->cn_flags |= SAVENAME;
846 return (EJUSTRETURN);
851 if (devfs_prison_check(de, td->td_ucred))
854 if ((cnp->cn_nameiop == DELETE) && (flags & ISLASTCN)) {
855 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, td);
864 error = devfs_allocv(de, dvp->v_mount, vpp);
870 devfs_lookup(struct vop_lookup_args *ap)
873 struct devfs_mount *dmp;
876 dmp = VFSTODEVFS(ap->a_dvp->v_mount);
878 sx_xlock(&dmp->dm_lock);
879 j = devfs_lookupx(ap, &dm_unlock);
881 sx_xunlock(&dmp->dm_lock);
886 devfs_mknod(struct vop_mknod_args *ap)
888 struct componentname *cnp;
889 struct vnode *dvp, **vpp;
890 struct devfs_dirent *dd, *de;
891 struct devfs_mount *dmp;
895 * The only type of node we should be creating here is a
896 * character device, for anything else return EOPNOTSUPP.
898 if (ap->a_vap->va_type != VCHR)
901 dmp = VFSTODEVFS(dvp->v_mount);
908 sx_xlock(&dmp->dm_lock);
909 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
910 if (cnp->cn_namelen != de->de_dirent->d_namlen)
912 if (bcmp(cnp->cn_nameptr, de->de_dirent->d_name,
913 de->de_dirent->d_namlen) != 0)
915 if (de->de_flags & DE_WHITEOUT)
921 de->de_flags &= ~DE_WHITEOUT;
922 error = devfs_allocv(de, dvp->v_mount, vpp);
925 sx_xunlock(&dmp->dm_lock);
931 devfs_open(struct vop_open_args *ap)
933 struct thread *td = ap->a_td;
934 struct vnode *vp = ap->a_vp;
935 struct cdev *dev = vp->v_rdev;
936 struct file *fp = ap->a_fp;
941 if (vp->v_type == VBLK)
947 /* Make this field valid before any I/O in d_open. */
948 if (dev->si_iosize_max == 0)
949 dev->si_iosize_max = DFLTPHYS;
951 dsw = dev_refthread(dev);
955 /* XXX: Special casing of ttys for deadfs. Probably redundant. */
956 if (dsw->d_flags & D_TTY)
957 vp->v_vflag |= VV_ISTTY;
967 if (dsw->d_fdopen != NULL)
968 error = dsw->d_fdopen(dev, ap->a_mode, td, fp);
970 error = dsw->d_open(dev, ap->a_mode, S_IFCHR, td);
973 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
980 #if 0 /* /dev/console */
982 ("Could not vnode bypass device on NULL fp"));
987 if (fp->f_ops == &badfileops)
988 finit(fp, fp->f_flag, DTYPE_VNODE, dev, &devfs_ops_f);
993 devfs_pathconf(struct vop_pathconf_args *ap)
996 switch (ap->a_name) {
997 case _PC_MAC_PRESENT:
1000 * If MAC is enabled, devfs automatically supports
1001 * trivial non-persistant label storage.
1009 return (vop_stdpathconf(ap));
1016 devfs_poll_f(struct file *fp, int events, struct ucred *cred, struct thread *td)
1024 error = devfs_fp_check(fp, &dev, &dsw);
1026 return (poll_no_poll(events));
1027 error = dsw->d_poll(dev, events, td);
1034 * Print out the contents of a special device vnode.
1037 devfs_print(struct vop_print_args *ap)
1040 printf("\tdev %s\n", devtoname(ap->a_vp->v_rdev));
1046 devfs_read_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
1049 int ioflag, error, resid;
1054 error = devfs_fp_check(fp, &dev, &dsw);
1057 resid = uio->uio_resid;
1058 ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT);
1059 if (ioflag & O_DIRECT)
1060 ioflag |= IO_DIRECT;
1062 if ((flags & FOF_OFFSET) == 0)
1063 uio->uio_offset = fp->f_offset;
1065 error = dsw->d_read(dev, uio, ioflag);
1066 if (uio->uio_resid != resid || (error == 0 && resid != 0))
1067 vfs_timestamp(&dev->si_atime);
1071 if ((flags & FOF_OFFSET) == 0)
1072 fp->f_offset = uio->uio_offset;
1073 fp->f_nextoff = uio->uio_offset;
1078 devfs_readdir(struct vop_readdir_args *ap)
1083 struct devfs_dirent *dd;
1084 struct devfs_dirent *de;
1085 struct devfs_mount *dmp;
1087 int *tmp_ncookies = NULL;
1089 if (ap->a_vp->v_type != VDIR)
1093 if (uio->uio_offset < 0)
1097 * XXX: This is a temporary hack to get around this filesystem not
1098 * supporting cookies. We store the location of the ncookies pointer
1099 * in a temporary variable before calling vfs_subr.c:vfs_read_dirent()
1100 * and set the number of cookies to 0. We then set the pointer to
1101 * NULL so that vfs_read_dirent doesn't try to call realloc() on
1102 * ap->a_cookies. Later in this function, we restore the ap->a_ncookies
1103 * pointer to its original location before returning to the caller.
1105 if (ap->a_ncookies != NULL) {
1106 tmp_ncookies = ap->a_ncookies;
1107 *ap->a_ncookies = 0;
1108 ap->a_ncookies = NULL;
1111 dmp = VFSTODEVFS(ap->a_vp->v_mount);
1112 sx_xlock(&dmp->dm_lock);
1113 DEVFS_DMP_HOLD(dmp);
1114 devfs_populate(dmp);
1115 if (DEVFS_DMP_DROP(dmp)) {
1116 sx_xunlock(&dmp->dm_lock);
1117 devfs_unmount_final(dmp);
1118 if (tmp_ncookies != NULL)
1119 ap->a_ncookies = tmp_ncookies;
1123 de = ap->a_vp->v_data;
1125 TAILQ_FOREACH(dd, &de->de_dlist, de_list) {
1126 KASSERT(dd->de_cdp != (void *)0xdeadc0de, ("%s %d\n", __func__, __LINE__));
1127 if (dd->de_flags & DE_WHITEOUT)
1129 if (devfs_prison_check(dd, ap->a_cred))
1131 if (dd->de_dirent->d_type == DT_DIR)
1136 if (dp->d_reclen > uio->uio_resid)
1138 dp->d_fileno = de->de_inode;
1139 if (off >= uio->uio_offset) {
1140 error = vfs_read_dirent(ap, dp, off);
1144 off += dp->d_reclen;
1146 sx_xunlock(&dmp->dm_lock);
1147 uio->uio_offset = off;
1150 * Restore ap->a_ncookies if it wasn't originally NULL in the first
1153 if (tmp_ncookies != NULL)
1154 ap->a_ncookies = tmp_ncookies;
1160 devfs_readlink(struct vop_readlink_args *ap)
1162 struct devfs_dirent *de;
1164 de = ap->a_vp->v_data;
1165 return (uiomove(de->de_symlink, strlen(de->de_symlink), ap->a_uio));
1169 devfs_reclaim(struct vop_reclaim_args *ap)
1171 struct vnode *vp = ap->a_vp;
1172 struct devfs_dirent *de;
1175 mtx_lock(&devfs_de_interlock);
1178 de->de_vnode = NULL;
1181 mtx_unlock(&devfs_de_interlock);
1183 vnode_destroy_vobject(vp);
1196 dev->si_usecount -= vp->v_usecount;
1204 devfs_remove(struct vop_remove_args *ap)
1206 struct vnode *vp = ap->a_vp;
1207 struct devfs_dirent *dd;
1208 struct devfs_dirent *de;
1209 struct devfs_mount *dmp = VFSTODEVFS(vp->v_mount);
1211 sx_xlock(&dmp->dm_lock);
1212 dd = ap->a_dvp->v_data;
1214 if (de->de_cdp == NULL) {
1215 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
1216 devfs_delete(dmp, de, 1);
1218 de->de_flags |= DE_WHITEOUT;
1220 sx_xunlock(&dmp->dm_lock);
1225 * Revoke is called on a tty when a terminal session ends. The vnode
1226 * is orphaned by setting v_op to deadfs so we need to let go of it
1227 * as well so that we create a new one next time around.
1231 devfs_revoke(struct vop_revoke_args *ap)
1233 struct vnode *vp = ap->a_vp, *vp2;
1235 struct cdev_priv *cdp;
1236 struct devfs_dirent *de;
1239 KASSERT((ap->a_flags & REVOKEALL) != 0, ("devfs_revoke !REVOKEALL"));
1242 cdp = cdev2priv(dev);
1255 mtx_lock(&devfs_de_interlock);
1258 for (i = 0; i <= cdp->cdp_maxdirent; i++) {
1259 de = cdp->cdp_dirents[i];
1267 mtx_unlock(&devfs_de_interlock);
1268 if (vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK,
1282 mtx_unlock(&devfs_de_interlock);
1287 if (!(cdp->cdp_flags & CDP_ACTIVE) && cdp->cdp_inuse == 0) {
1288 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
1290 dev_rel(&cdp->cdp_c);
1294 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1299 devfs_rioctl(struct vop_ioctl_args *ap)
1302 struct devfs_mount *dmp;
1306 vn_lock(vp, LK_SHARED | LK_RETRY);
1307 if (vp->v_iflag & VI_DOOMED) {
1311 dmp = VFSTODEVFS(vp->v_mount);
1312 sx_xlock(&dmp->dm_lock);
1314 DEVFS_DMP_HOLD(dmp);
1315 devfs_populate(dmp);
1316 if (DEVFS_DMP_DROP(dmp)) {
1317 sx_xunlock(&dmp->dm_lock);
1318 devfs_unmount_final(dmp);
1321 error = devfs_rules_ioctl(dmp, ap->a_command, ap->a_data, ap->a_td);
1322 sx_xunlock(&dmp->dm_lock);
1327 devfs_rread(struct vop_read_args *ap)
1330 if (ap->a_vp->v_type != VDIR)
1332 return (VOP_READDIR(ap->a_vp, ap->a_uio, ap->a_cred, NULL, NULL, NULL));
1336 devfs_setattr(struct vop_setattr_args *ap)
1338 struct devfs_dirent *de;
1349 if ((vap->va_type != VNON) ||
1350 (vap->va_nlink != VNOVAL) ||
1351 (vap->va_fsid != VNOVAL) ||
1352 (vap->va_fileid != VNOVAL) ||
1353 (vap->va_blocksize != VNOVAL) ||
1354 (vap->va_flags != VNOVAL && vap->va_flags != 0) ||
1355 (vap->va_rdev != VNOVAL) ||
1356 ((int)vap->va_bytes != VNOVAL) ||
1357 (vap->va_gen != VNOVAL)) {
1362 if (vp->v_type == VDIR)
1366 if (vap->va_uid == (uid_t)VNOVAL)
1370 if (vap->va_gid == (gid_t)VNOVAL)
1374 if (uid != de->de_uid || gid != de->de_gid) {
1375 if ((ap->a_cred->cr_uid != de->de_uid) || uid != de->de_uid ||
1376 (gid != de->de_gid && !groupmember(gid, ap->a_cred))) {
1377 error = priv_check(td, PRIV_VFS_CHOWN);
1386 if (vap->va_mode != (mode_t)VNOVAL) {
1387 if (ap->a_cred->cr_uid != de->de_uid) {
1388 error = priv_check(td, PRIV_VFS_ADMIN);
1392 de->de_mode = vap->va_mode;
1396 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1397 /* See the comment in ufs_vnops::ufs_setattr(). */
1398 if ((error = VOP_ACCESS(vp, VADMIN, ap->a_cred, td)) &&
1399 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1400 (error = VOP_ACCESS(vp, VWRITE, ap->a_cred, td))))
1402 if (vap->va_atime.tv_sec != VNOVAL) {
1403 if (vp->v_type == VCHR)
1404 vp->v_rdev->si_atime = vap->va_atime;
1406 de->de_atime = vap->va_atime;
1408 if (vap->va_mtime.tv_sec != VNOVAL) {
1409 if (vp->v_type == VCHR)
1410 vp->v_rdev->si_mtime = vap->va_mtime;
1412 de->de_mtime = vap->va_mtime;
1418 if (vp->v_type == VCHR)
1419 vfs_timestamp(&vp->v_rdev->si_ctime);
1421 vfs_timestamp(&de->de_mtime);
1428 devfs_setlabel(struct vop_setlabel_args *ap)
1431 struct devfs_dirent *de;
1436 mac_vnode_relabel(ap->a_cred, vp, ap->a_label);
1437 mac_devfs_update(vp->v_mount, de, vp);
1444 devfs_stat_f(struct file *fp, struct stat *sb, struct ucred *cred, struct thread *td)
1447 return (vnops.fo_stat(fp, sb, cred, td));
1451 devfs_symlink(struct vop_symlink_args *ap)
1454 struct devfs_dirent *dd;
1455 struct devfs_dirent *de;
1456 struct devfs_mount *dmp;
1458 error = priv_check(curthread, PRIV_DEVFS_SYMLINK);
1461 dmp = VFSTODEVFS(ap->a_dvp->v_mount);
1462 dd = ap->a_dvp->v_data;
1463 de = devfs_newdirent(ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen);
1467 de->de_inode = alloc_unr(devfs_inos);
1468 de->de_dirent->d_type = DT_LNK;
1469 i = strlen(ap->a_target) + 1;
1470 de->de_symlink = malloc(i, M_DEVFS, M_WAITOK);
1471 bcopy(ap->a_target, de->de_symlink, i);
1472 sx_xlock(&dmp->dm_lock);
1474 mac_devfs_create_symlink(ap->a_cnp->cn_cred, dmp->dm_mount, dd, de);
1476 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
1477 return (devfs_allocv(de, ap->a_dvp->v_mount, ap->a_vpp));
1481 devfs_truncate_f(struct file *fp, off_t length, struct ucred *cred, struct thread *td)
1484 return (vnops.fo_truncate(fp, length, cred, td));
1489 devfs_write_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
1492 int error, ioflag, resid;
1497 error = devfs_fp_check(fp, &dev, &dsw);
1500 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td));
1501 ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT | O_FSYNC);
1502 if (ioflag & O_DIRECT)
1503 ioflag |= IO_DIRECT;
1504 if ((flags & FOF_OFFSET) == 0)
1505 uio->uio_offset = fp->f_offset;
1507 resid = uio->uio_resid;
1509 error = dsw->d_write(dev, uio, ioflag);
1510 if (uio->uio_resid != resid || (error == 0 && resid != 0)) {
1511 vfs_timestamp(&dev->si_ctime);
1512 dev->si_mtime = dev->si_ctime;
1517 if ((flags & FOF_OFFSET) == 0)
1518 fp->f_offset = uio->uio_offset;
1519 fp->f_nextoff = uio->uio_offset;
1524 dev2udev(struct cdev *x)
1528 return (cdev2priv(x)->cdp_inode);
1531 static struct fileops devfs_ops_f = {
1532 .fo_read = devfs_read_f,
1533 .fo_write = devfs_write_f,
1534 .fo_truncate = devfs_truncate_f,
1535 .fo_ioctl = devfs_ioctl_f,
1536 .fo_poll = devfs_poll_f,
1537 .fo_kqfilter = devfs_kqfilter_f,
1538 .fo_stat = devfs_stat_f,
1539 .fo_close = devfs_close_f,
1540 .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
1543 static struct vop_vector devfs_vnodeops = {
1544 .vop_default = &default_vnodeops,
1546 .vop_access = devfs_access,
1547 .vop_getattr = devfs_getattr,
1548 .vop_ioctl = devfs_rioctl,
1549 .vop_lookup = devfs_lookup,
1550 .vop_mknod = devfs_mknod,
1551 .vop_pathconf = devfs_pathconf,
1552 .vop_read = devfs_rread,
1553 .vop_readdir = devfs_readdir,
1554 .vop_readlink = devfs_readlink,
1555 .vop_reclaim = devfs_reclaim,
1556 .vop_remove = devfs_remove,
1557 .vop_revoke = devfs_revoke,
1558 .vop_setattr = devfs_setattr,
1560 .vop_setlabel = devfs_setlabel,
1562 .vop_symlink = devfs_symlink,
1563 .vop_vptocnp = devfs_vptocnp,
1566 static struct vop_vector devfs_specops = {
1567 .vop_default = &default_vnodeops,
1569 .vop_access = devfs_access,
1570 .vop_bmap = VOP_PANIC,
1571 .vop_close = devfs_close,
1572 .vop_create = VOP_PANIC,
1573 .vop_fsync = devfs_fsync,
1574 .vop_getattr = devfs_getattr,
1575 .vop_link = VOP_PANIC,
1576 .vop_mkdir = VOP_PANIC,
1577 .vop_mknod = VOP_PANIC,
1578 .vop_open = devfs_open,
1579 .vop_pathconf = devfs_pathconf,
1580 .vop_print = devfs_print,
1581 .vop_read = VOP_PANIC,
1582 .vop_readdir = VOP_PANIC,
1583 .vop_readlink = VOP_PANIC,
1584 .vop_reallocblks = VOP_PANIC,
1585 .vop_reclaim = devfs_reclaim,
1586 .vop_remove = devfs_remove,
1587 .vop_rename = VOP_PANIC,
1588 .vop_revoke = devfs_revoke,
1589 .vop_rmdir = VOP_PANIC,
1590 .vop_setattr = devfs_setattr,
1592 .vop_setlabel = devfs_setlabel,
1594 .vop_strategy = VOP_PANIC,
1595 .vop_symlink = VOP_PANIC,
1596 .vop_vptocnp = devfs_vptocnp,
1597 .vop_write = VOP_PANIC,
1601 * Our calling convention to the device drivers used to be that we passed
1602 * vnode.h IO_* flags to read()/write(), but we're moving to fcntl.h O_
1603 * flags instead since that's what open(), close() and ioctl() takes and
1604 * we don't really want vnode.h in device drivers.
1605 * We solved the source compatibility by redefining some vnode flags to
1606 * be the same as the fcntl ones and by sending down the bitwise OR of
1607 * the respective fcntl/vnode flags. These CTASSERTS make sure nobody
1608 * pulls the rug out under this.
1610 CTASSERT(O_NONBLOCK == IO_NDELAY);
1611 CTASSERT(O_FSYNC == IO_SYNC);