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);
351 mtx_lock(&devfs_de_interlock);
355 mtx_unlock(&devfs_de_interlock);
356 sx_xunlock(&dmp->dm_lock);
357 error = vget(vp, lockmode | LK_INTERLOCK, curthread);
358 sx_xlock(&dmp->dm_lock);
359 if (devfs_allocv_drop_refs(0, dmp, de)) {
365 sx_xunlock(&dmp->dm_lock);
368 sx_xunlock(&dmp->dm_lock);
372 mtx_unlock(&devfs_de_interlock);
373 if (de->de_dirent->d_type == DT_CHR) {
374 if (!(de->de_cdp->cdp_flags & CDP_ACTIVE)) {
375 devfs_allocv_drop_refs(1, dmp, de);
378 dev = &de->de_cdp->cdp_c;
382 error = getnewvnode("devfs", mp, &devfs_vnodeops, &vp);
384 devfs_allocv_drop_refs(1, dmp, de);
385 printf("devfs_allocv: failed to allocate new vnode\n");
389 if (de->de_dirent->d_type == DT_CHR) {
394 /* XXX: v_rdev should be protect by vnode lock */
396 KASSERT(vp->v_usecount == 1,
397 ("%s %d (%d)\n", __func__, __LINE__, vp->v_usecount));
398 dev->si_usecount += vp->v_usecount;
399 /* Special casing of ttys for deadfs. Probably redundant. */
401 if (dsw != NULL && (dsw->d_flags & D_TTY) != 0)
402 vp->v_vflag |= VV_ISTTY;
405 if ((dev->si_flags & SI_ETERNAL) != 0)
406 vp->v_vflag |= VV_ETERNALDEV;
407 vp->v_op = &devfs_specops;
408 } else if (de->de_dirent->d_type == DT_DIR) {
410 } else if (de->de_dirent->d_type == DT_LNK) {
416 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_NOWITNESS);
417 mtx_lock(&devfs_de_interlock);
420 mtx_unlock(&devfs_de_interlock);
421 error = insmntque1(vp, mp, devfs_insmntque_dtr, de);
423 (void) devfs_allocv_drop_refs(1, dmp, de);
426 if (devfs_allocv_drop_refs(0, dmp, de)) {
431 mac_devfs_vnode_associate(mp, de, vp);
433 sx_xunlock(&dmp->dm_lock);
439 devfs_access(struct vop_access_args *ap)
441 struct vnode *vp = ap->a_vp;
442 struct devfs_dirent *de;
446 if (vp->v_type == VDIR)
449 error = vaccess(vp->v_type, de->de_mode, de->de_uid, de->de_gid,
450 ap->a_accmode, ap->a_cred, NULL);
455 /* We do, however, allow access to the controlling terminal */
456 if (!(ap->a_td->td_proc->p_flag & P_CONTROLT))
458 if (ap->a_td->td_proc->p_session->s_ttyvp == de->de_vnode)
465 devfs_close(struct vop_close_args *ap)
467 struct vnode *vp = ap->a_vp, *oldvp;
468 struct thread *td = ap->a_td;
469 struct cdev *dev = vp->v_rdev;
471 int vp_locked, error, ref;
474 * XXX: Don't call d_close() if we were called because of
475 * XXX: insmntque1() failure.
477 if (vp->v_data == NULL)
481 * Hack: a tty device that is a controlling terminal
482 * has a reference from the session structure.
483 * We cannot easily tell that a character device is
484 * a controlling terminal, unless it is the closing
485 * process' controlling terminal. In that case,
486 * if the reference count is 2 (this last descriptor
487 * plus the session), release the reference from the session.
490 sx_xlock(&proctree_lock);
491 if (td && vp == td->td_proc->p_session->s_ttyvp) {
492 SESS_LOCK(td->td_proc->p_session);
494 if (count_dev(dev) == 2 && (vp->v_iflag & VI_DOOMED) == 0) {
495 td->td_proc->p_session->s_ttyvp = NULL;
499 SESS_UNLOCK(td->td_proc->p_session);
501 sx_xunlock(&proctree_lock);
505 * We do not want to really close the device if it
506 * is still in use unless we are trying to close it
507 * forcibly. Since every use (buffer, vnode, swap, cmap)
508 * holds a reference to the vnode, and because we mark
509 * any other vnodes that alias this device, when the
510 * sum of the reference counts on all the aliased
511 * vnodes descends to one, we are on last close.
513 dsw = dev_refthread(dev, &ref);
517 if (vp->v_iflag & VI_DOOMED) {
519 } else if (dsw->d_flags & D_TRACKCLOSE) {
520 /* Keep device updated on status. */
521 } else if (count_dev(dev) > 1) {
523 dev_relthread(dev, ref);
528 vp_locked = VOP_ISLOCKED(vp);
530 KASSERT(dev->si_refcount > 0,
531 ("devfs_close() on un-referenced struct cdev *(%s)", devtoname(dev)));
532 error = dsw->d_close(dev, ap->a_fflag, S_IFCHR, td);
533 dev_relthread(dev, ref);
534 vn_lock(vp, vp_locked | LK_RETRY);
540 devfs_close_f(struct file *fp, struct thread *td)
547 error = vnops.fo_close(fp, td);
553 devfs_fsync(struct vop_fsync_args *ap)
557 struct devfs_dirent *de;
559 if (!vn_isdisk(ap->a_vp, &error)) {
560 bo = &ap->a_vp->v_bufobj;
561 de = ap->a_vp->v_data;
562 if (error == ENXIO && bo->bo_dirty.bv_cnt > 0) {
563 printf("Device %s went missing before all of the data "
564 "could be written to it; expect data loss.\n",
565 de->de_dirent->d_name);
567 error = vop_stdfsync(ap);
568 if (bo->bo_dirty.bv_cnt != 0 || error != 0)
569 panic("devfs_fsync: vop_stdfsync failed.");
575 return (vop_stdfsync(ap));
579 devfs_getattr(struct vop_getattr_args *ap)
581 struct vnode *vp = ap->a_vp;
582 struct vattr *vap = ap->a_vap;
584 struct devfs_dirent *de;
588 KASSERT(de != NULL, ("Null dirent in devfs_getattr vp=%p", vp));
589 if (vp->v_type == VDIR) {
592 ("Null dir dirent in devfs_getattr vp=%p", vp));
594 vap->va_uid = de->de_uid;
595 vap->va_gid = de->de_gid;
596 vap->va_mode = de->de_mode;
597 if (vp->v_type == VLNK)
598 vap->va_size = strlen(de->de_symlink);
599 else if (vp->v_type == VDIR)
600 vap->va_size = vap->va_bytes = DEV_BSIZE;
603 if (vp->v_type != VDIR)
605 vap->va_blocksize = DEV_BSIZE;
606 vap->va_type = vp->v_type;
610 if ((aa).tv_sec <= 3600) { \
611 (aa).tv_sec = boottime.tv_sec; \
612 (aa).tv_nsec = boottime.tv_usec * 1000; \
616 if (vp->v_type != VCHR) {
618 vap->va_atime = de->de_atime;
620 vap->va_mtime = de->de_mtime;
622 vap->va_ctime = de->de_ctime;
626 vap->va_atime = dev->si_atime;
628 vap->va_mtime = dev->si_mtime;
630 vap->va_ctime = dev->si_ctime;
632 vap->va_rdev = cdev2priv(dev)->cdp_inode;
637 vap->va_nlink = de->de_links;
638 vap->va_fileid = de->de_inode;
645 devfs_ioctl_f(struct file *fp, u_long com, void *data, struct ucred *cred, struct thread *td)
653 struct fiodgname_arg *fgn;
657 error = devfs_fp_check(fp, &dev, &dsw, &ref);
661 if (com == FIODTYPE) {
662 *(int *)data = dsw->d_flags & D_TYPEMASK;
664 dev_relthread(dev, ref);
666 } else if (com == FIODGNAME) {
673 error = copyout(p, fgn->buf, i);
675 dev_relthread(dev, ref);
678 error = dsw->d_ioctl(dev, com, data, fp->f_flag, td);
680 dev_relthread(dev, ref);
681 if (error == ENOIOCTL)
683 if (error == 0 && com == TIOCSCTTY) {
686 /* Do nothing if reassigning same control tty */
687 sx_slock(&proctree_lock);
688 if (td->td_proc->p_session->s_ttyvp == vp) {
689 sx_sunlock(&proctree_lock);
693 vpold = td->td_proc->p_session->s_ttyvp;
695 SESS_LOCK(td->td_proc->p_session);
696 td->td_proc->p_session->s_ttyvp = vp;
697 SESS_UNLOCK(td->td_proc->p_session);
699 sx_sunlock(&proctree_lock);
701 /* Get rid of reference to old control tty */
710 devfs_kqfilter_f(struct file *fp, struct knote *kn)
720 error = devfs_fp_check(fp, &dev, &dsw, &ref);
723 error = dsw->d_kqfilter(dev, kn);
725 dev_relthread(dev, ref);
730 devfs_prison_check(struct devfs_dirent *de, struct ucred *tcr)
732 struct cdev_priv *cdp;
738 dcr = cdp->cdp_c.si_cred;
742 return (prison_check(tcr, dcr));
746 devfs_lookupx(struct vop_lookup_args *ap, int *dm_unlock)
748 struct componentname *cnp;
749 struct vnode *dvp, **vpp;
751 struct devfs_dirent *de, *dd;
752 struct devfs_dirent **dde;
753 struct devfs_mount *dmp;
755 int error, flags, nameiop, dvplocked;
756 char specname[SPECNAMELEN + 1], *pname;
761 pname = cnp->cn_nameptr;
763 flags = cnp->cn_flags;
764 nameiop = cnp->cn_nameiop;
765 dmp = VFSTODEVFS(dvp->v_mount);
769 if ((flags & ISLASTCN) && nameiop == RENAME)
772 if (dvp->v_type != VDIR)
775 if ((flags & ISDOTDOT) && (dvp->v_vflag & VV_ROOT))
778 error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, td);
782 if (cnp->cn_namelen == 1 && *pname == '.') {
783 if ((flags & ISLASTCN) && nameiop != LOOKUP)
790 if (flags & ISDOTDOT) {
791 if ((flags & ISLASTCN) && nameiop != LOOKUP)
793 de = devfs_parent_dirent(dd);
796 dvplocked = VOP_ISLOCKED(dvp);
798 error = devfs_allocv(de, dvp->v_mount,
799 cnp->cn_lkflags & LK_TYPE_MASK, vpp);
801 vn_lock(dvp, dvplocked | LK_RETRY);
807 if (DEVFS_DMP_DROP(dmp)) {
809 sx_xunlock(&dmp->dm_lock);
810 devfs_unmount_final(dmp);
814 de = devfs_find(dd, cnp->cn_nameptr, cnp->cn_namelen);
815 while (de == NULL) { /* While(...) so we can use break */
817 if (nameiop == DELETE)
821 * OK, we didn't have an entry for the name we were asked for
822 * so we try to see if anybody can create it on demand.
824 pname = devfs_fqpn(specname, dvp, cnp);
830 sx_xunlock(&dmp->dm_lock);
831 sx_slock(&clone_drain_lock);
832 EVENTHANDLER_INVOKE(dev_clone,
833 td->td_ucred, pname, strlen(pname), &cdev);
834 sx_sunlock(&clone_drain_lock);
835 sx_xlock(&dmp->dm_lock);
836 if (DEVFS_DMP_DROP(dmp)) {
838 sx_xunlock(&dmp->dm_lock);
839 devfs_unmount_final(dmp);
849 if (DEVFS_DMP_DROP(dmp)) {
851 sx_xunlock(&dmp->dm_lock);
852 devfs_unmount_final(dmp);
858 dde = &cdev2priv(cdev)->cdp_dirents[dmp->dm_idx];
859 if (dde != NULL && *dde != NULL)
866 if (de == NULL || de->de_flags & DE_WHITEOUT) {
867 if ((nameiop == CREATE || nameiop == RENAME) &&
868 (flags & (LOCKPARENT | WANTPARENT)) && (flags & ISLASTCN)) {
869 cnp->cn_flags |= SAVENAME;
870 return (EJUSTRETURN);
875 if (devfs_prison_check(de, td->td_ucred))
878 if ((cnp->cn_nameiop == DELETE) && (flags & ISLASTCN)) {
879 error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, td);
888 error = devfs_allocv(de, dvp->v_mount, cnp->cn_lkflags & LK_TYPE_MASK,
895 devfs_lookup(struct vop_lookup_args *ap)
898 struct devfs_mount *dmp;
901 dmp = VFSTODEVFS(ap->a_dvp->v_mount);
903 sx_xlock(&dmp->dm_lock);
904 j = devfs_lookupx(ap, &dm_unlock);
906 sx_xunlock(&dmp->dm_lock);
911 devfs_mknod(struct vop_mknod_args *ap)
913 struct componentname *cnp;
914 struct vnode *dvp, **vpp;
915 struct devfs_dirent *dd, *de;
916 struct devfs_mount *dmp;
920 * The only type of node we should be creating here is a
921 * character device, for anything else return EOPNOTSUPP.
923 if (ap->a_vap->va_type != VCHR)
926 dmp = VFSTODEVFS(dvp->v_mount);
933 sx_xlock(&dmp->dm_lock);
934 TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
935 if (cnp->cn_namelen != de->de_dirent->d_namlen)
937 if (bcmp(cnp->cn_nameptr, de->de_dirent->d_name,
938 de->de_dirent->d_namlen) != 0)
940 if (de->de_flags & DE_WHITEOUT)
946 de->de_flags &= ~DE_WHITEOUT;
947 error = devfs_allocv(de, dvp->v_mount, LK_EXCLUSIVE, vpp);
950 sx_xunlock(&dmp->dm_lock);
956 devfs_open(struct vop_open_args *ap)
958 struct thread *td = ap->a_td;
959 struct vnode *vp = ap->a_vp;
960 struct cdev *dev = vp->v_rdev;
961 struct file *fp = ap->a_fp;
962 int error, ref, vlocked;
966 if (vp->v_type == VBLK)
972 /* Make this field valid before any I/O in d_open. */
973 if (dev->si_iosize_max == 0)
974 dev->si_iosize_max = DFLTPHYS;
976 dsw = dev_refthread(dev, &ref);
980 vlocked = VOP_ISLOCKED(vp);
989 if (dsw->d_fdopen != NULL)
990 error = dsw->d_fdopen(dev, ap->a_mode, td, fp);
992 error = dsw->d_open(dev, ap->a_mode, S_IFCHR, td);
995 vn_lock(vp, vlocked | LK_RETRY);
996 dev_relthread(dev, ref);
1000 #if 0 /* /dev/console */
1001 KASSERT(fp != NULL, ("Could not vnode bypass device on NULL fp"));
1006 if (fp->f_ops == &badfileops)
1007 finit(fp, fp->f_flag, DTYPE_VNODE, dev, &devfs_ops_f);
1012 devfs_pathconf(struct vop_pathconf_args *ap)
1015 switch (ap->a_name) {
1016 case _PC_MAC_PRESENT:
1019 * If MAC is enabled, devfs automatically supports
1020 * trivial non-persistant label storage.
1028 return (vop_stdpathconf(ap));
1035 devfs_poll_f(struct file *fp, int events, struct ucred *cred, struct thread *td)
1043 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1045 return (poll_no_poll(events));
1046 error = dsw->d_poll(dev, events, td);
1048 dev_relthread(dev, ref);
1053 * Print out the contents of a special device vnode.
1056 devfs_print(struct vop_print_args *ap)
1059 printf("\tdev %s\n", devtoname(ap->a_vp->v_rdev));
1065 devfs_read_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
1068 int ioflag, error, ref, resid;
1073 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1076 resid = uio->uio_resid;
1077 ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT);
1078 if (ioflag & O_DIRECT)
1079 ioflag |= IO_DIRECT;
1081 if ((flags & FOF_OFFSET) == 0)
1082 uio->uio_offset = fp->f_offset;
1084 error = dsw->d_read(dev, uio, ioflag);
1085 if (uio->uio_resid != resid || (error == 0 && resid != 0))
1086 vfs_timestamp(&dev->si_atime);
1088 dev_relthread(dev, ref);
1090 if ((flags & FOF_OFFSET) == 0)
1091 fp->f_offset = uio->uio_offset;
1092 fp->f_nextoff = uio->uio_offset;
1097 devfs_readdir(struct vop_readdir_args *ap)
1102 struct devfs_dirent *dd;
1103 struct devfs_dirent *de;
1104 struct devfs_mount *dmp;
1106 int *tmp_ncookies = NULL;
1108 if (ap->a_vp->v_type != VDIR)
1112 if (uio->uio_offset < 0)
1116 * XXX: This is a temporary hack to get around this filesystem not
1117 * supporting cookies. We store the location of the ncookies pointer
1118 * in a temporary variable before calling vfs_subr.c:vfs_read_dirent()
1119 * and set the number of cookies to 0. We then set the pointer to
1120 * NULL so that vfs_read_dirent doesn't try to call realloc() on
1121 * ap->a_cookies. Later in this function, we restore the ap->a_ncookies
1122 * pointer to its original location before returning to the caller.
1124 if (ap->a_ncookies != NULL) {
1125 tmp_ncookies = ap->a_ncookies;
1126 *ap->a_ncookies = 0;
1127 ap->a_ncookies = NULL;
1130 dmp = VFSTODEVFS(ap->a_vp->v_mount);
1131 sx_xlock(&dmp->dm_lock);
1132 DEVFS_DMP_HOLD(dmp);
1133 devfs_populate(dmp);
1134 if (DEVFS_DMP_DROP(dmp)) {
1135 sx_xunlock(&dmp->dm_lock);
1136 devfs_unmount_final(dmp);
1137 if (tmp_ncookies != NULL)
1138 ap->a_ncookies = tmp_ncookies;
1142 de = ap->a_vp->v_data;
1144 TAILQ_FOREACH(dd, &de->de_dlist, de_list) {
1145 KASSERT(dd->de_cdp != (void *)0xdeadc0de, ("%s %d\n", __func__, __LINE__));
1146 if (dd->de_flags & DE_WHITEOUT)
1148 if (devfs_prison_check(dd, ap->a_cred))
1150 if (dd->de_dirent->d_type == DT_DIR)
1155 if (dp->d_reclen > uio->uio_resid)
1157 dp->d_fileno = de->de_inode;
1158 if (off >= uio->uio_offset) {
1159 error = vfs_read_dirent(ap, dp, off);
1163 off += dp->d_reclen;
1165 sx_xunlock(&dmp->dm_lock);
1166 uio->uio_offset = off;
1169 * Restore ap->a_ncookies if it wasn't originally NULL in the first
1172 if (tmp_ncookies != NULL)
1173 ap->a_ncookies = tmp_ncookies;
1179 devfs_readlink(struct vop_readlink_args *ap)
1181 struct devfs_dirent *de;
1183 de = ap->a_vp->v_data;
1184 return (uiomove(de->de_symlink, strlen(de->de_symlink), ap->a_uio));
1188 devfs_reclaim(struct vop_reclaim_args *ap)
1190 struct vnode *vp = ap->a_vp;
1191 struct devfs_dirent *de;
1194 mtx_lock(&devfs_de_interlock);
1197 de->de_vnode = NULL;
1200 mtx_unlock(&devfs_de_interlock);
1202 vnode_destroy_vobject(vp);
1215 dev->si_usecount -= vp->v_usecount;
1223 devfs_remove(struct vop_remove_args *ap)
1225 struct vnode *vp = ap->a_vp;
1226 struct devfs_dirent *dd;
1227 struct devfs_dirent *de;
1228 struct devfs_mount *dmp = VFSTODEVFS(vp->v_mount);
1230 sx_xlock(&dmp->dm_lock);
1231 dd = ap->a_dvp->v_data;
1233 if (de->de_cdp == NULL) {
1234 TAILQ_REMOVE(&dd->de_dlist, de, de_list);
1235 devfs_delete(dmp, de, 1);
1237 de->de_flags |= DE_WHITEOUT;
1239 sx_xunlock(&dmp->dm_lock);
1244 * Revoke is called on a tty when a terminal session ends. The vnode
1245 * is orphaned by setting v_op to deadfs so we need to let go of it
1246 * as well so that we create a new one next time around.
1250 devfs_revoke(struct vop_revoke_args *ap)
1252 struct vnode *vp = ap->a_vp, *vp2;
1254 struct cdev_priv *cdp;
1255 struct devfs_dirent *de;
1258 KASSERT((ap->a_flags & REVOKEALL) != 0, ("devfs_revoke !REVOKEALL"));
1261 cdp = cdev2priv(dev);
1274 mtx_lock(&devfs_de_interlock);
1277 for (i = 0; i <= cdp->cdp_maxdirent; i++) {
1278 de = cdp->cdp_dirents[i];
1286 mtx_unlock(&devfs_de_interlock);
1287 if (vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK,
1301 mtx_unlock(&devfs_de_interlock);
1306 if (!(cdp->cdp_flags & CDP_ACTIVE) && cdp->cdp_inuse == 0) {
1307 TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
1309 dev_rel(&cdp->cdp_c);
1313 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1318 devfs_rioctl(struct vop_ioctl_args *ap)
1321 struct devfs_mount *dmp;
1325 vn_lock(vp, LK_SHARED | LK_RETRY);
1326 if (vp->v_iflag & VI_DOOMED) {
1330 dmp = VFSTODEVFS(vp->v_mount);
1331 sx_xlock(&dmp->dm_lock);
1333 DEVFS_DMP_HOLD(dmp);
1334 devfs_populate(dmp);
1335 if (DEVFS_DMP_DROP(dmp)) {
1336 sx_xunlock(&dmp->dm_lock);
1337 devfs_unmount_final(dmp);
1340 error = devfs_rules_ioctl(dmp, ap->a_command, ap->a_data, ap->a_td);
1341 sx_xunlock(&dmp->dm_lock);
1346 devfs_rread(struct vop_read_args *ap)
1349 if (ap->a_vp->v_type != VDIR)
1351 return (VOP_READDIR(ap->a_vp, ap->a_uio, ap->a_cred, NULL, NULL, NULL));
1355 devfs_setattr(struct vop_setattr_args *ap)
1357 struct devfs_dirent *de;
1368 if ((vap->va_type != VNON) ||
1369 (vap->va_nlink != VNOVAL) ||
1370 (vap->va_fsid != VNOVAL) ||
1371 (vap->va_fileid != VNOVAL) ||
1372 (vap->va_blocksize != VNOVAL) ||
1373 (vap->va_flags != VNOVAL && vap->va_flags != 0) ||
1374 (vap->va_rdev != VNOVAL) ||
1375 ((int)vap->va_bytes != VNOVAL) ||
1376 (vap->va_gen != VNOVAL)) {
1381 if (vp->v_type == VDIR)
1385 if (vap->va_uid == (uid_t)VNOVAL)
1389 if (vap->va_gid == (gid_t)VNOVAL)
1393 if (uid != de->de_uid || gid != de->de_gid) {
1394 if ((ap->a_cred->cr_uid != de->de_uid) || uid != de->de_uid ||
1395 (gid != de->de_gid && !groupmember(gid, ap->a_cred))) {
1396 error = priv_check(td, PRIV_VFS_CHOWN);
1405 if (vap->va_mode != (mode_t)VNOVAL) {
1406 if (ap->a_cred->cr_uid != de->de_uid) {
1407 error = priv_check(td, PRIV_VFS_ADMIN);
1411 de->de_mode = vap->va_mode;
1415 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1416 /* See the comment in ufs_vnops::ufs_setattr(). */
1417 if ((error = VOP_ACCESS(vp, VADMIN, ap->a_cred, td)) &&
1418 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1419 (error = VOP_ACCESS(vp, VWRITE, ap->a_cred, td))))
1421 if (vap->va_atime.tv_sec != VNOVAL) {
1422 if (vp->v_type == VCHR)
1423 vp->v_rdev->si_atime = vap->va_atime;
1425 de->de_atime = vap->va_atime;
1427 if (vap->va_mtime.tv_sec != VNOVAL) {
1428 if (vp->v_type == VCHR)
1429 vp->v_rdev->si_mtime = vap->va_mtime;
1431 de->de_mtime = vap->va_mtime;
1437 if (vp->v_type == VCHR)
1438 vfs_timestamp(&vp->v_rdev->si_ctime);
1440 vfs_timestamp(&de->de_mtime);
1447 devfs_setlabel(struct vop_setlabel_args *ap)
1450 struct devfs_dirent *de;
1455 mac_vnode_relabel(ap->a_cred, vp, ap->a_label);
1456 mac_devfs_update(vp->v_mount, de, vp);
1463 devfs_stat_f(struct file *fp, struct stat *sb, struct ucred *cred, struct thread *td)
1466 return (vnops.fo_stat(fp, sb, cred, td));
1470 devfs_symlink(struct vop_symlink_args *ap)
1473 struct devfs_dirent *dd;
1474 struct devfs_dirent *de;
1475 struct devfs_mount *dmp;
1477 error = priv_check(curthread, PRIV_DEVFS_SYMLINK);
1480 dmp = VFSTODEVFS(ap->a_dvp->v_mount);
1481 dd = ap->a_dvp->v_data;
1482 de = devfs_newdirent(ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen);
1486 de->de_inode = alloc_unr(devfs_inos);
1487 de->de_dirent->d_type = DT_LNK;
1488 i = strlen(ap->a_target) + 1;
1489 de->de_symlink = malloc(i, M_DEVFS, M_WAITOK);
1490 bcopy(ap->a_target, de->de_symlink, i);
1491 sx_xlock(&dmp->dm_lock);
1493 mac_devfs_create_symlink(ap->a_cnp->cn_cred, dmp->dm_mount, dd, de);
1495 TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
1496 return (devfs_allocv(de, ap->a_dvp->v_mount, LK_EXCLUSIVE, ap->a_vpp));
1500 devfs_truncate_f(struct file *fp, off_t length, struct ucred *cred, struct thread *td)
1503 return (vnops.fo_truncate(fp, length, cred, td));
1508 devfs_write_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
1511 int error, ioflag, ref, resid;
1516 error = devfs_fp_check(fp, &dev, &dsw, &ref);
1519 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td));
1520 ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT | O_FSYNC);
1521 if (ioflag & O_DIRECT)
1522 ioflag |= IO_DIRECT;
1523 if ((flags & FOF_OFFSET) == 0)
1524 uio->uio_offset = fp->f_offset;
1526 resid = uio->uio_resid;
1528 error = dsw->d_write(dev, uio, ioflag);
1529 if (uio->uio_resid != resid || (error == 0 && resid != 0)) {
1530 vfs_timestamp(&dev->si_ctime);
1531 dev->si_mtime = dev->si_ctime;
1534 dev_relthread(dev, ref);
1536 if ((flags & FOF_OFFSET) == 0)
1537 fp->f_offset = uio->uio_offset;
1538 fp->f_nextoff = uio->uio_offset;
1543 dev2udev(struct cdev *x)
1547 return (cdev2priv(x)->cdp_inode);
1550 static struct fileops devfs_ops_f = {
1551 .fo_read = devfs_read_f,
1552 .fo_write = devfs_write_f,
1553 .fo_truncate = devfs_truncate_f,
1554 .fo_ioctl = devfs_ioctl_f,
1555 .fo_poll = devfs_poll_f,
1556 .fo_kqfilter = devfs_kqfilter_f,
1557 .fo_stat = devfs_stat_f,
1558 .fo_close = devfs_close_f,
1559 .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
1562 static struct vop_vector devfs_vnodeops = {
1563 .vop_default = &default_vnodeops,
1565 .vop_access = devfs_access,
1566 .vop_getattr = devfs_getattr,
1567 .vop_ioctl = devfs_rioctl,
1568 .vop_lookup = devfs_lookup,
1569 .vop_mknod = devfs_mknod,
1570 .vop_pathconf = devfs_pathconf,
1571 .vop_read = devfs_rread,
1572 .vop_readdir = devfs_readdir,
1573 .vop_readlink = devfs_readlink,
1574 .vop_reclaim = devfs_reclaim,
1575 .vop_remove = devfs_remove,
1576 .vop_revoke = devfs_revoke,
1577 .vop_setattr = devfs_setattr,
1579 .vop_setlabel = devfs_setlabel,
1581 .vop_symlink = devfs_symlink,
1582 .vop_vptocnp = devfs_vptocnp,
1585 static struct vop_vector devfs_specops = {
1586 .vop_default = &default_vnodeops,
1588 .vop_access = devfs_access,
1589 .vop_bmap = VOP_PANIC,
1590 .vop_close = devfs_close,
1591 .vop_create = VOP_PANIC,
1592 .vop_fsync = devfs_fsync,
1593 .vop_getattr = devfs_getattr,
1594 .vop_link = VOP_PANIC,
1595 .vop_mkdir = VOP_PANIC,
1596 .vop_mknod = VOP_PANIC,
1597 .vop_open = devfs_open,
1598 .vop_pathconf = devfs_pathconf,
1599 .vop_print = devfs_print,
1600 .vop_read = VOP_PANIC,
1601 .vop_readdir = VOP_PANIC,
1602 .vop_readlink = VOP_PANIC,
1603 .vop_reallocblks = VOP_PANIC,
1604 .vop_reclaim = devfs_reclaim,
1605 .vop_remove = devfs_remove,
1606 .vop_rename = VOP_PANIC,
1607 .vop_revoke = devfs_revoke,
1608 .vop_rmdir = VOP_PANIC,
1609 .vop_setattr = devfs_setattr,
1611 .vop_setlabel = devfs_setlabel,
1613 .vop_strategy = VOP_PANIC,
1614 .vop_symlink = VOP_PANIC,
1615 .vop_vptocnp = devfs_vptocnp,
1616 .vop_write = VOP_PANIC,
1620 * Our calling convention to the device drivers used to be that we passed
1621 * vnode.h IO_* flags to read()/write(), but we're moving to fcntl.h O_
1622 * flags instead since that's what open(), close() and ioctl() takes and
1623 * we don't really want vnode.h in device drivers.
1624 * We solved the source compatibility by redefining some vnode flags to
1625 * be the same as the fcntl ones and by sending down the bitwise OR of
1626 * the respective fcntl/vnode flags. These CTASSERTS make sure nobody
1627 * pulls the rug out under this.
1629 CTASSERT(O_NONBLOCK == IO_NDELAY);
1630 CTASSERT(O_FSYNC == IO_SYNC);