2 * Copyright (c) 1990, 1993, 1995
3 * The Regents of the University of California.
4 * Copyright (c) 2005 Robert N. M. Watson
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 4. 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 * @(#)fifo_vnops.c 8.10 (Berkeley) 5/27/95
35 #include <sys/param.h>
36 #include <sys/event.h>
38 #include <sys/filedesc.h>
39 #include <sys/filio.h>
40 #include <sys/fcntl.h>
41 #include <sys/kernel.h>
43 #include <sys/mutex.h>
44 #include <sys/malloc.h>
46 #include <sys/proc.h> /* XXXKSE */
47 #include <sys/signalvar.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
51 #include <sys/systm.h>
53 #include <sys/unistd.h>
54 #include <sys/vnode.h>
55 #include <fs/fifofs/fifo.h>
57 static fo_rdwr_t fifo_read_f;
58 static fo_rdwr_t fifo_write_f;
59 static fo_ioctl_t fifo_ioctl_f;
60 static fo_poll_t fifo_poll_f;
61 static fo_kqfilter_t fifo_kqfilter_f;
62 static fo_stat_t fifo_stat_f;
63 static fo_close_t fifo_close_f;
65 struct fileops fifo_ops_f = {
66 .fo_read = fifo_read_f,
67 .fo_write = fifo_write_f,
68 .fo_ioctl = fifo_ioctl_f,
69 .fo_poll = fifo_poll_f,
70 .fo_kqfilter = fifo_kqfilter_f,
71 .fo_stat = fifo_stat_f,
72 .fo_close = fifo_close_f,
73 .fo_flags = DFLAG_PASSABLE
77 * This structure is associated with the FIFO vnode and stores
78 * the state associated with the FIFO.
81 struct socket *fi_readsock;
82 struct socket *fi_writesock;
87 static vop_print_t fifo_print;
88 static vop_open_t fifo_open;
89 static vop_close_t fifo_close;
90 static vop_ioctl_t fifo_ioctl;
91 static vop_kqfilter_t fifo_kqfilter;
92 static vop_pathconf_t fifo_pathconf;
93 static vop_advlock_t fifo_advlock;
95 static void filt_fifordetach(struct knote *kn);
96 static int filt_fiforead(struct knote *kn, long hint);
97 static void filt_fifowdetach(struct knote *kn);
98 static int filt_fifowrite(struct knote *kn, long hint);
99 static void filt_fifodetach_notsup(struct knote *kn);
100 static int filt_fifo_notsup(struct knote *kn, long hint);
102 static struct filterops fiforead_filtops =
103 { 1, NULL, filt_fifordetach, filt_fiforead };
104 static struct filterops fifowrite_filtops =
105 { 1, NULL, filt_fifowdetach, filt_fifowrite };
106 static struct filterops fifo_notsup_filtops =
107 { 1, NULL, filt_fifodetach_notsup, filt_fifo_notsup };
109 struct vop_vector fifo_specops = {
110 .vop_default = &default_vnodeops,
112 .vop_access = VOP_EBADF,
113 .vop_advlock = fifo_advlock,
114 .vop_close = fifo_close,
115 .vop_create = VOP_PANIC,
116 .vop_getattr = VOP_EBADF,
117 .vop_ioctl = fifo_ioctl,
118 .vop_kqfilter = fifo_kqfilter,
119 .vop_lease = VOP_NULL,
120 .vop_link = VOP_PANIC,
121 .vop_mkdir = VOP_PANIC,
122 .vop_mknod = VOP_PANIC,
123 .vop_open = fifo_open,
124 .vop_pathconf = fifo_pathconf,
125 .vop_print = fifo_print,
126 .vop_read = VOP_PANIC,
127 .vop_readdir = VOP_PANIC,
128 .vop_readlink = VOP_PANIC,
129 .vop_reallocblks = VOP_PANIC,
130 .vop_reclaim = VOP_NULL,
131 .vop_remove = VOP_PANIC,
132 .vop_rename = VOP_PANIC,
133 .vop_rmdir = VOP_PANIC,
134 .vop_setattr = VOP_EBADF,
135 .vop_symlink = VOP_PANIC,
136 .vop_write = VOP_PANIC,
140 MTX_SYSINIT(fifo, &fifo_mtx, "fifo mutex", MTX_DEF);
143 * Dispose of fifo resources.
146 fifo_cleanup(struct vnode *vp)
148 struct fifoinfo *fip = vp->v_fifoinfo;
150 ASSERT_VOP_LOCKED(vp, "fifo_cleanup");
151 if (fip->fi_readers == 0 && fip->fi_writers == 0) {
152 vp->v_fifoinfo = NULL;
153 (void)soclose(fip->fi_readsock);
154 (void)soclose(fip->fi_writesock);
160 * Open called to set up a new instance of a fifo or
161 * to find an active instance of a fifo.
166 struct vop_open_args /* {
169 struct ucred *a_cred;
173 struct vnode *vp = ap->a_vp;
174 struct fifoinfo *fip;
175 struct thread *td = ap->a_td;
176 struct ucred *cred = ap->a_cred;
177 struct socket *rso, *wso;
181 ASSERT_VOP_LOCKED(vp, "fifo_open");
184 if ((fip = vp->v_fifoinfo) == NULL) {
185 MALLOC(fip, struct fifoinfo *, sizeof(*fip), M_VNODE, M_WAITOK);
186 error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, cred, td);
189 fip->fi_readsock = rso;
190 error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, cred, td);
193 fip->fi_writesock = wso;
194 error = soconnect2(wso, rso);
203 fip->fi_readers = fip->fi_writers = 0;
204 wso->so_snd.sb_lowat = PIPE_BUF;
205 SOCKBUF_LOCK(&rso->so_rcv);
206 rso->so_rcv.sb_state |= SBS_CANTRCVMORE;
207 SOCKBUF_UNLOCK(&rso->so_rcv);
208 KASSERT(vp->v_fifoinfo == NULL,
209 ("fifo_open: v_fifoinfo race"));
210 vp->v_fifoinfo = fip;
214 * General access to fi_readers and fi_writers is protected using
217 * Protect the increment of fi_readers and fi_writers and the
218 * associated calls to wakeup() with the fifo mutex in addition
219 * to the vnode lock. This allows the vnode lock to be dropped
220 * for the msleep() calls below, and using the fifo mutex with
221 * msleep() prevents the wakeup from being missed.
224 if (ap->a_mode & FREAD) {
226 if (fip->fi_readers == 1) {
227 SOCKBUF_LOCK(&fip->fi_writesock->so_snd);
228 fip->fi_writesock->so_snd.sb_state &= ~SBS_CANTSENDMORE;
229 SOCKBUF_UNLOCK(&fip->fi_writesock->so_snd);
230 if (fip->fi_writers > 0) {
231 wakeup(&fip->fi_writers);
232 sowwakeup(fip->fi_writesock);
236 if (ap->a_mode & FWRITE) {
237 if ((ap->a_mode & O_NONBLOCK) && fip->fi_readers == 0) {
238 mtx_unlock(&fifo_mtx);
242 if (fip->fi_writers == 1) {
243 SOCKBUF_LOCK(&fip->fi_readsock->so_rcv);
244 fip->fi_readsock->so_rcv.sb_state &= ~SBS_CANTRCVMORE;
245 SOCKBUF_UNLOCK(&fip->fi_readsock->so_rcv);
246 if (fip->fi_readers > 0) {
247 wakeup(&fip->fi_readers);
248 sorwakeup(fip->fi_readsock);
252 if ((ap->a_mode & O_NONBLOCK) == 0) {
253 if ((ap->a_mode & FREAD) && fip->fi_writers == 0) {
254 VOP_UNLOCK(vp, 0, td);
255 error = msleep(&fip->fi_readers, &fifo_mtx,
256 PDROP | PCATCH | PSOCK, "fifoor", 0);
257 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
260 if (fip->fi_readers == 0) {
261 socantsendmore(fip->fi_writesock);
268 * We must have got woken up because we had a writer.
269 * That (and not still having one) is the condition
270 * that we must wait for.
273 if ((ap->a_mode & FWRITE) && fip->fi_readers == 0) {
274 VOP_UNLOCK(vp, 0, td);
275 error = msleep(&fip->fi_writers, &fifo_mtx,
276 PDROP | PCATCH | PSOCK, "fifoow", 0);
277 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
280 if (fip->fi_writers == 0) {
281 socantrcvmore(fip->fi_readsock);
287 * We must have got woken up because we had
288 * a reader. That (and not still having one)
289 * is the condition that we must wait for.
294 mtx_unlock(&fifo_mtx);
295 KASSERT(ap->a_fdidx >= 0, ("can't fifo/vnode bypass %d", ap->a_fdidx));
296 fp = ap->a_td->td_proc->p_fd->fd_ofiles[ap->a_fdidx];
297 KASSERT(fp->f_ops == &badfileops, ("not badfileops in fifo_open"));
298 fp->f_ops = &fifo_ops_f;
304 * Now unused vnode ioctl routine.
309 struct vop_ioctl_args /* {
314 struct ucred *a_cred;
319 printf("WARNING: fifo_ioctl called unexpectedly\n");
324 * Now unused vnode kqfilter routine.
329 struct vop_kqfilter_args /* {
335 printf("WARNING: fifo_kqfilter called unexpectedly\n");
340 filt_fifordetach(struct knote *kn)
342 struct socket *so = (struct socket *)kn->kn_hook;
344 SOCKBUF_LOCK(&so->so_rcv);
345 knlist_remove(&so->so_rcv.sb_sel.si_note, kn, 1);
346 if (knlist_empty(&so->so_rcv.sb_sel.si_note))
347 so->so_rcv.sb_flags &= ~SB_KNOTE;
348 SOCKBUF_UNLOCK(&so->so_rcv);
352 filt_fiforead(struct knote *kn, long hint)
354 struct socket *so = (struct socket *)kn->kn_hook;
356 SOCKBUF_LOCK_ASSERT(&so->so_rcv);
357 kn->kn_data = so->so_rcv.sb_cc;
358 if (so->so_rcv.sb_state & SBS_CANTRCVMORE) {
359 kn->kn_flags |= EV_EOF;
362 kn->kn_flags &= ~EV_EOF;
363 return (kn->kn_data > 0);
368 filt_fifowdetach(struct knote *kn)
370 struct socket *so = (struct socket *)kn->kn_hook;
372 SOCKBUF_LOCK(&so->so_snd);
373 knlist_remove(&so->so_snd.sb_sel.si_note, kn, 1);
374 if (knlist_empty(&so->so_snd.sb_sel.si_note))
375 so->so_snd.sb_flags &= ~SB_KNOTE;
376 SOCKBUF_UNLOCK(&so->so_snd);
380 filt_fifowrite(struct knote *kn, long hint)
382 struct socket *so = (struct socket *)kn->kn_hook;
384 SOCKBUF_LOCK_ASSERT(&so->so_snd);
385 kn->kn_data = sbspace(&so->so_snd);
386 if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
387 kn->kn_flags |= EV_EOF;
390 kn->kn_flags &= ~EV_EOF;
391 return (kn->kn_data >= so->so_snd.sb_lowat);
396 filt_fifodetach_notsup(struct knote *kn)
402 filt_fifo_notsup(struct knote *kn, long hint)
409 * Device close routine
414 struct vop_close_args /* {
417 struct ucred *a_cred;
421 struct vnode *vp = ap->a_vp;
422 struct fifoinfo *fip = vp->v_fifoinfo;
424 ASSERT_VOP_LOCKED(vp, "fifo_close");
425 KASSERT(fip != NULL, ("fifo_close: no v_fifoinfo"));
426 if (ap->a_fflag & FREAD) {
428 if (fip->fi_readers == 0)
429 socantsendmore(fip->fi_writesock);
431 if (ap->a_fflag & FWRITE) {
433 if (fip->fi_writers == 0)
434 socantrcvmore(fip->fi_readsock);
441 * Print out internal contents of a fifo vnode.
447 register struct fifoinfo *fip = vp->v_fifoinfo;
449 printf(", fifo with %ld readers and %ld writers",
450 fip->fi_readers, fip->fi_writers);
455 * Print out the contents of a fifo vnode.
459 struct vop_print_args /* {
463 fifo_printinfo(ap->a_vp);
469 * Return POSIX pathconf information applicable to fifo's.
473 struct vop_pathconf_args /* {
480 switch (ap->a_name) {
482 *ap->a_retval = LINK_MAX;
485 *ap->a_retval = PIPE_BUF;
487 case _PC_CHOWN_RESTRICTED:
497 * Fifo advisory byte-level locks.
502 struct vop_advlock_args /* {
511 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
515 fifo_close_f(struct file *fp, struct thread *td)
518 return (vnops.fo_close(fp, td));
522 * The implementation of ioctl() for named fifos is complicated by the fact
523 * that we permit O_RDWR fifo file descriptors, meaning that the actions of
524 * ioctls may have to be applied to both the underlying sockets rather than
525 * just one. The original implementation simply forward the ioctl to one
526 * or both sockets based on fp->f_flag. We now consider each ioctl
527 * separately, as the composition effect requires careful ordering.
529 * We do not blindly pass all ioctls through to the socket in order to avoid
530 * providing unnecessary ioctls that might be improperly depended on by
531 * applications (such as socket-specific, routing, and interface ioctls).
533 * Unlike sys_pipe.c, fifos do not implement the deprecated TIOCSPGRP and
534 * TIOCGPGRP ioctls. Earlier implementations of fifos did forward SIOCSPGRP
535 * and SIOCGPGRP ioctls, so we might need to re-add those here.
538 fifo_ioctl_f(struct file *fp, u_long com, void *data, struct ucred *cred,
542 struct file filetmp; /* Local, so need not be locked. */
551 * Non-blocking I/O is implemented at the fifo layer using
552 * MSG_NBIO, so does not need to be forwarded down the stack.
560 * These socket ioctls don't have any ordering requirements,
561 * so are called in an arbitrary order, and only on the
562 * sockets indicated by the file descriptor rights.
564 * XXXRW: If O_RDWR and the read socket accepts an ioctl but
565 * the write socket doesn't, the socketpair is left in an
566 * inconsistent state.
568 if (fp->f_flag & FREAD) {
569 filetmp.f_data = fi->fi_readsock;
570 filetmp.f_cred = cred;
571 error = soo_ioctl(&filetmp, com, data, cred, td);
575 if (fp->f_flag & FWRITE) {
576 filetmp.f_data = fi->fi_writesock;
577 filetmp.f_cred = cred;
578 error = soo_ioctl(&filetmp, com, data, cred, td);
584 * FIONREAD will return 0 for non-readable descriptors, and
585 * the results of FIONREAD on the read socket for readable
588 if (!(fp->f_flag & FREAD)) {
592 filetmp.f_data = fi->fi_readsock;
593 filetmp.f_cred = cred;
594 return (soo_ioctl(&filetmp, com, data, cred, td));
602 * Because fifos are now a file descriptor layer object, EVFILT_VNODE is not
603 * implemented. Likely, fifo_kqfilter() should be removed, and
604 * fifo_kqfilter_f() should know how to forward the request to the underling
605 * vnode using f_vnode in the file descriptor here.
608 fifo_kqfilter_f(struct file *fp, struct knote *kn)
617 * If a filter is requested that is not supported by this file
618 * descriptor, don't return an error, but also don't ever generate an
621 if ((kn->kn_filter == EVFILT_READ) && !(fp->f_flag & FREAD)) {
622 kn->kn_fop = &fifo_notsup_filtops;
626 if ((kn->kn_filter == EVFILT_WRITE) && !(fp->f_flag & FWRITE)) {
627 kn->kn_fop = &fifo_notsup_filtops;
631 switch (kn->kn_filter) {
633 kn->kn_fop = &fiforead_filtops;
634 so = fi->fi_readsock;
638 kn->kn_fop = &fifowrite_filtops;
639 so = fi->fi_writesock;
646 kn->kn_hook = (caddr_t)so;
649 knlist_add(&sb->sb_sel.si_note, kn, 1);
650 sb->sb_flags |= SB_KNOTE;
657 fifo_poll_f(struct file *fp, int events, struct ucred *cred, struct thread *td)
659 struct fifoinfo *fip;
661 int levents, revents = 0;
665 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | POLLRDBAND);
666 if ((fp->f_flag & FREAD) && levents) {
668 * If POLLIN or POLLRDNORM is requested and POLLINIGNEOF is
669 * not, then convert the first two to the last one. This
670 * tells the socket poll function to ignore EOF so that we
671 * block if there is no writer (and no data). Callers can
672 * set POLLINIGNEOF to get non-blocking behavior.
674 if (levents & (POLLIN | POLLRDNORM) &&
675 !(levents & POLLINIGNEOF)) {
676 levents &= ~(POLLIN | POLLRDNORM);
677 levents |= POLLINIGNEOF;
680 filetmp.f_data = fip->fi_readsock;
681 filetmp.f_cred = cred;
682 revents |= soo_poll(&filetmp, levents, cred, td);
684 /* Reverse the above conversion. */
685 if ((revents & POLLINIGNEOF) && !(events & POLLINIGNEOF)) {
686 revents |= (events & (POLLIN | POLLRDNORM));
687 revents &= ~POLLINIGNEOF;
690 levents = events & (POLLOUT | POLLWRNORM | POLLWRBAND);
691 if ((fp->f_flag & FWRITE) && levents) {
692 filetmp.f_data = fip->fi_writesock;
693 filetmp.f_cred = cred;
694 revents |= soo_poll(&filetmp, levents, cred, td);
700 fifo_read_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
702 struct fifoinfo *fip;
706 KASSERT(uio->uio_rw == UIO_READ,("fifo_read mode"));
707 if (uio->uio_resid == 0)
709 sflags = (fp->f_flag & FNONBLOCK) ? MSG_NBIO : 0;
711 error = soreceive(fip->fi_readsock, NULL, uio, NULL, NULL, &sflags);
717 fifo_stat_f(struct file *fp, struct stat *sb, struct ucred *cred, struct thread *td)
720 return (vnops.fo_stat(fp, sb, cred, td));
724 fifo_write_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
726 struct fifoinfo *fip;
730 KASSERT(uio->uio_rw == UIO_WRITE,("fifo_write mode"));
731 sflags = (fp->f_flag & FNONBLOCK) ? MSG_NBIO : 0;
733 error = sosend(fip->fi_writesock, NULL, uio, 0, NULL, sflags, td);