kernel - fine-grained namecache and partial vnode MPSAFE work
[dragonfly.git] / sys / vfs / fifofs / fifo_vnops.c
blob3fb1f6bf26ced8d47cabaac13198f1ea90fb503b
1 /*
2 * Copyright (c) 1990, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
34 * Filesystem FIFO type ops. All entry points are MPSAFE. We primarily
35 * use v_token to interlock operations.
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/unistd.h>
40 #include <sys/kernel.h>
41 #include <sys/lock.h>
42 #include <sys/malloc.h>
43 #include <sys/thread2.h>
44 #include <sys/vnode.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/filio.h>
48 #include <sys/fcntl.h>
49 #include <sys/file.h>
50 #include <sys/event.h>
51 #include <sys/poll.h>
52 #include <sys/un.h>
54 #include <sys/thread2.h>
56 #include "fifo.h"
59 * This structure is associated with the FIFO vnode and stores
60 * the state associated with the FIFO.
62 struct fifoinfo {
63 struct socket *fi_readsock;
64 struct socket *fi_writesock;
65 long fi_readers;
66 long fi_writers;
69 static int fifo_badop (void);
70 static int fifo_print (struct vop_print_args *);
71 static int fifo_lookup (struct vop_old_lookup_args *);
72 static int fifo_open (struct vop_open_args *);
73 static int fifo_close (struct vop_close_args *);
74 static int fifo_read (struct vop_read_args *);
75 static int fifo_write (struct vop_write_args *);
76 static int fifo_ioctl (struct vop_ioctl_args *);
77 static int fifo_poll (struct vop_poll_args *);
78 static int fifo_kqfilter (struct vop_kqfilter_args *);
79 static int fifo_inactive (struct vop_inactive_args *);
80 static int fifo_bmap (struct vop_bmap_args *);
81 static int fifo_pathconf (struct vop_pathconf_args *);
82 static int fifo_advlock (struct vop_advlock_args *);
84 static void filt_fifordetach(struct knote *kn);
85 static int filt_fiforead(struct knote *kn, long hint);
86 static void filt_fifowdetach(struct knote *kn);
87 static int filt_fifowrite(struct knote *kn, long hint);
89 static struct filterops fiforead_filtops =
90 { 1, NULL, filt_fifordetach, filt_fiforead };
91 static struct filterops fifowrite_filtops =
92 { 1, NULL, filt_fifowdetach, filt_fifowrite };
94 struct vop_ops fifo_vnode_vops = {
95 .vop_default = vop_defaultop,
96 .vop_access = (void *)vop_ebadf,
97 .vop_advlock = fifo_advlock,
98 .vop_bmap = fifo_bmap,
99 .vop_close = fifo_close,
100 .vop_old_create = (void *)fifo_badop,
101 .vop_getattr = (void *)vop_ebadf,
102 .vop_inactive = fifo_inactive,
103 .vop_ioctl = fifo_ioctl,
104 .vop_kqfilter = fifo_kqfilter,
105 .vop_old_link = (void *)fifo_badop,
106 .vop_old_lookup = fifo_lookup,
107 .vop_old_mkdir = (void *)fifo_badop,
108 .vop_old_mknod = (void *)fifo_badop,
109 .vop_open = fifo_open,
110 .vop_pathconf = fifo_pathconf,
111 .vop_poll = fifo_poll,
112 .vop_print = fifo_print,
113 .vop_read = fifo_read,
114 .vop_readdir = (void *)fifo_badop,
115 .vop_readlink = (void *)fifo_badop,
116 .vop_reallocblks = (void *)fifo_badop,
117 .vop_reclaim = (void *)vop_null,
118 .vop_old_remove = (void *)fifo_badop,
119 .vop_old_rename = (void *)fifo_badop,
120 .vop_old_rmdir = (void *)fifo_badop,
121 .vop_setattr = (void *)vop_ebadf,
122 .vop_old_symlink = (void *)fifo_badop,
123 .vop_write = fifo_write
126 VNODEOP_SET(fifo_vnode_vops);
128 static MALLOC_DEFINE(M_FIFOINFO, "Fifo info", "Fifo info entries");
131 * fifo_vnoperate()
134 fifo_vnoperate(struct vop_generic_args *ap)
136 return (VOCALL(&fifo_vnode_vops, ap));
140 * Trivial lookup routine that always fails.
142 * fifo_lookup(struct vnode *a_dvp, struct vnode **a_vpp,
143 * struct componentname *a_cnp)
145 /* ARGSUSED */
146 static int
147 fifo_lookup(struct vop_old_lookup_args *ap)
149 *ap->a_vpp = NULL;
150 return (ENOTDIR);
154 * Open called to set up a new instance of a fifo or
155 * to find an active instance of a fifo.
157 * fifo_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
158 * struct file *a_fp)
160 /* ARGSUSED */
161 static int
162 fifo_open(struct vop_open_args *ap)
164 struct thread *td = curthread;
165 struct vnode *vp = ap->a_vp;
166 struct fifoinfo *fip;
167 struct socket *rso, *wso;
168 lwkt_tokref vlock;
169 int error;
171 lwkt_gettoken(&vlock, &vp->v_token);
172 if ((fip = vp->v_fifoinfo) == NULL) {
173 MALLOC(fip, struct fifoinfo *, sizeof(*fip), M_FIFOINFO, M_WAITOK);
174 vp->v_fifoinfo = fip;
175 error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, td);
176 if (error) {
177 kfree(fip, M_FIFOINFO);
178 vp->v_fifoinfo = NULL;
179 goto done;
181 fip->fi_readsock = rso;
182 error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, td);
183 if (error) {
184 soclose(rso, FNONBLOCK);
185 kfree(fip, M_FIFOINFO);
186 vp->v_fifoinfo = NULL;
187 goto done;
189 fip->fi_writesock = wso;
190 error = unp_connect2(wso, rso);
191 if (error) {
192 soclose(wso, FNONBLOCK);
193 soclose(rso, FNONBLOCK);
194 kfree(fip, M_FIFOINFO);
195 vp->v_fifoinfo = NULL;
196 goto done;
198 fip->fi_readers = fip->fi_writers = 0;
199 wso->so_snd.ssb_lowat = PIPE_BUF;
200 rso->so_state |= SS_CANTRCVMORE;
202 if (ap->a_mode & FREAD) {
203 fip->fi_readers++;
204 if (fip->fi_readers == 1) {
205 fip->fi_writesock->so_state &= ~SS_CANTSENDMORE;
206 if (fip->fi_writers > 0) {
207 wakeup((caddr_t)&fip->fi_writers);
208 sowwakeup(fip->fi_writesock);
212 if (ap->a_mode & FWRITE) {
213 fip->fi_writers++;
214 if (fip->fi_writers == 1) {
215 fip->fi_readsock->so_state &= ~SS_CANTRCVMORE;
216 if (fip->fi_readers > 0) {
217 wakeup((caddr_t)&fip->fi_readers);
218 sorwakeup(fip->fi_writesock);
222 if ((ap->a_mode & FREAD) && (ap->a_mode & O_NONBLOCK) == 0) {
223 if (fip->fi_writers == 0) {
224 vn_unlock(vp);
225 error = tsleep((caddr_t)&fip->fi_readers,
226 PCATCH, "fifoor", 0);
227 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
228 if (error)
229 goto bad;
231 * We must have got woken up because we had a writer.
232 * That (and not still having one) is the condition
233 * that we must wait for.
237 if (ap->a_mode & FWRITE) {
238 if (ap->a_mode & O_NONBLOCK) {
239 if (fip->fi_readers == 0) {
240 error = ENXIO;
241 goto bad;
243 } else {
244 if (fip->fi_readers == 0) {
245 vn_unlock(vp);
246 error = tsleep((caddr_t)&fip->fi_writers,
247 PCATCH, "fifoow", 0);
248 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
249 if (error)
250 goto bad;
252 * We must have got woken up because we had
253 * a reader. That (and not still having one)
254 * is the condition that we must wait for.
259 vsetflags(vp, VNOTSEEKABLE);
260 error = vop_stdopen(ap);
261 lwkt_reltoken(&vlock);
262 return (error);
263 bad:
264 vop_stdopen(ap); /* bump opencount/writecount as appropriate */
265 VOP_CLOSE(vp, ap->a_mode);
266 done:
267 lwkt_reltoken(&vlock);
268 return (error);
272 * Vnode op for read
274 * fifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
275 * struct ucred *a_cred)
277 /* ARGSUSED */
278 static int
279 fifo_read(struct vop_read_args *ap)
281 struct uio *uio = ap->a_uio;
282 struct socket *rso = ap->a_vp->v_fifoinfo->fi_readsock;
283 int error, startresid;
284 lwkt_tokref vlock;
285 int flags;
287 #ifdef DIAGNOSTIC
288 if (uio->uio_rw != UIO_READ)
289 panic("fifo_read mode");
290 #endif
291 if (uio->uio_resid == 0)
292 return (0);
293 if (ap->a_ioflag & IO_NDELAY)
294 flags = MSG_FNONBLOCKING;
295 else
296 flags = 0;
297 startresid = uio->uio_resid;
298 vn_unlock(ap->a_vp);
299 lwkt_gettoken(&vlock, &ap->a_vp->v_token);
300 error = soreceive(rso, NULL, uio, NULL, NULL, &flags);
301 lwkt_reltoken(&vlock);
302 vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY);
303 return (error);
307 * Vnode op for write
309 * fifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
310 * struct ucred *a_cred)
312 /* ARGSUSED */
313 static int
314 fifo_write(struct vop_write_args *ap)
316 struct socket *wso = ap->a_vp->v_fifoinfo->fi_writesock;
317 struct thread *td = ap->a_uio->uio_td;
318 lwkt_tokref vlock;
319 int error;
320 int flags;
322 #ifdef DIAGNOSTIC
323 if (ap->a_uio->uio_rw != UIO_WRITE)
324 panic("fifo_write mode");
325 #endif
326 if (ap->a_ioflag & IO_NDELAY)
327 flags = MSG_FNONBLOCKING;
328 else
329 flags = 0;
330 vn_unlock(ap->a_vp);
331 lwkt_gettoken(&vlock, &ap->a_vp->v_token);
332 error = sosend(wso, NULL, ap->a_uio, 0, NULL, flags, td);
333 lwkt_reltoken(&vlock);
334 vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY);
335 return (error);
339 * Device ioctl operation.
341 * fifo_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data, int a_fflag,
342 * struct ucred *a_cred, struct sysmsg *a_sysmsg)
344 /* ARGSUSED */
345 static int
346 fifo_ioctl(struct vop_ioctl_args *ap)
348 struct file filetmp; /* Local */
349 lwkt_tokref vlock;
350 int error;
352 if (ap->a_fflag & FREAD) {
353 filetmp.f_data = ap->a_vp->v_fifoinfo->fi_readsock;
354 lwkt_gettoken(&vlock, &ap->a_vp->v_token);
355 error = soo_ioctl(&filetmp, ap->a_command, ap->a_data,
356 ap->a_cred, ap->a_sysmsg);
357 lwkt_reltoken(&vlock);
358 if (error)
359 return (error);
361 if (ap->a_fflag & FWRITE) {
362 filetmp.f_data = ap->a_vp->v_fifoinfo->fi_writesock;
363 lwkt_gettoken(&vlock, &ap->a_vp->v_token);
364 error = soo_ioctl(&filetmp, ap->a_command, ap->a_data,
365 ap->a_cred, ap->a_sysmsg);
366 lwkt_reltoken(&vlock);
367 if (error)
368 return (error);
370 return (0);
374 * fifo_kqfilter(struct vnode *a_vp, struct knote *a_kn)
376 /* ARGSUSED */
377 static int
378 fifo_kqfilter(struct vop_kqfilter_args *ap)
380 struct vnode *vp = ap->a_vp;
381 struct fifoinfo *fi = vp->v_fifoinfo;
382 struct socket *so;
383 struct signalsockbuf *ssb;
384 lwkt_tokref vlock;
386 lwkt_gettoken(&vlock, &vp->v_token);
387 switch (ap->a_kn->kn_filter) {
388 case EVFILT_READ:
389 ap->a_kn->kn_fop = &fiforead_filtops;
390 so = fi->fi_readsock;
391 ssb = &so->so_rcv;
392 break;
393 case EVFILT_WRITE:
394 ap->a_kn->kn_fop = &fifowrite_filtops;
395 so = fi->fi_writesock;
396 ssb = &so->so_snd;
397 break;
398 default:
399 return (1);
402 ap->a_kn->kn_hook = (caddr_t)vp;
403 ssb_insert_knote(ssb, ap->a_kn);
405 lwkt_reltoken(&vlock);
406 return (0);
409 static void
410 filt_fifordetach(struct knote *kn)
412 struct vnode *vp = (void *)kn->kn_hook;
413 struct socket *so = vp->v_fifoinfo->fi_readsock;
414 lwkt_tokref vlock;
416 lwkt_gettoken(&vlock, &vp->v_token);
417 ssb_remove_knote(&so->so_rcv, kn);
418 lwkt_reltoken(&vlock);
421 static int
422 filt_fiforead(struct knote *kn, long hint)
424 struct vnode *vp = (void *)kn->kn_hook;
425 struct socket *so = vp->v_fifoinfo->fi_writesock;
426 lwkt_tokref vlock;
428 lwkt_gettoken(&vlock, &vp->v_token);
429 kn->kn_data = so->so_rcv.ssb_cc;
430 if (so->so_state & SS_CANTRCVMORE) {
431 kn->kn_flags |= EV_EOF;
432 return (1);
434 kn->kn_flags &= ~EV_EOF;
435 lwkt_reltoken(&vlock);
436 return (kn->kn_data > 0);
439 static void
440 filt_fifowdetach(struct knote *kn)
442 struct vnode *vp = (void *)kn->kn_hook;
443 struct socket *so = vp->v_fifoinfo->fi_writesock;
444 lwkt_tokref vlock;
446 lwkt_gettoken(&vlock, &vp->v_token);
447 ssb_remove_knote(&so->so_snd, kn);
448 lwkt_reltoken(&vlock);
451 static int
452 filt_fifowrite(struct knote *kn, long hint)
454 struct vnode *vp = (void *)kn->kn_hook;
455 struct socket *so = vp->v_fifoinfo->fi_writesock;
456 lwkt_tokref vlock;
458 lwkt_gettoken(&vlock, &vp->v_token);
459 kn->kn_data = ssb_space(&so->so_snd);
460 if (so->so_state & SS_CANTSENDMORE) {
461 kn->kn_flags |= EV_EOF;
462 return (1);
464 kn->kn_flags &= ~EV_EOF;
465 lwkt_reltoken(&vlock);
466 return (kn->kn_data >= so->so_snd.ssb_lowat);
470 * fifo_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred)
472 /* ARGSUSED */
473 static int
474 fifo_poll(struct vop_poll_args *ap)
476 struct file filetmp;
477 int events, revents = 0;
478 lwkt_tokref vlock;
480 lwkt_gettoken(&vlock, &ap->a_vp->v_token);
481 events = ap->a_events &
482 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | POLLRDBAND);
483 if (events) {
485 * If POLLIN or POLLRDNORM is requested and POLLINIGNEOF is
486 * not, then convert the first two to the last one. This
487 * tells the socket poll function to ignore EOF so that we
488 * block if there is no writer (and no data). Callers can
489 * set POLLINIGNEOF to get non-blocking behavior.
491 if (events & (POLLIN | POLLRDNORM) &&
492 !(events & POLLINIGNEOF)) {
493 events &= ~(POLLIN | POLLRDNORM);
494 events |= POLLINIGNEOF;
497 filetmp.f_data = ap->a_vp->v_fifoinfo->fi_readsock;
498 if (filetmp.f_data)
499 revents |= soo_poll(&filetmp, events, ap->a_cred);
501 /* Reverse the above conversion. */
502 if ((revents & POLLINIGNEOF) &&
503 !(ap->a_events & POLLINIGNEOF)) {
504 revents |= (ap->a_events & (POLLIN | POLLRDNORM));
505 revents &= ~POLLINIGNEOF;
508 events = ap->a_events & (POLLOUT | POLLWRNORM | POLLWRBAND);
509 if (events) {
510 filetmp.f_data = ap->a_vp->v_fifoinfo->fi_writesock;
511 if (filetmp.f_data)
512 revents |= soo_poll(&filetmp, events, ap->a_cred);
514 lwkt_reltoken(&vlock);
515 return (revents);
519 * fifo_inactive(struct vnode *a_vp)
521 static int
522 fifo_inactive(struct vop_inactive_args *ap)
524 return (0);
528 * This is a noop, simply returning what one has been given.
530 * fifo_bmap(struct vnode *a_vp, off_t a_loffset,
531 * off_t *a_doffsetp, int *a_runp, int *a_runb)
533 static int
534 fifo_bmap(struct vop_bmap_args *ap)
536 if (ap->a_doffsetp != NULL)
537 *ap->a_doffsetp = ap->a_loffset;
538 if (ap->a_runp != NULL)
539 *ap->a_runp = 0;
540 if (ap->a_runb != NULL)
541 *ap->a_runb = 0;
542 return (0);
546 * Device close routine
548 * fifo_close(struct vnode *a_vp, int a_fflag)
550 /* ARGSUSED */
551 static int
552 fifo_close(struct vop_close_args *ap)
554 struct vnode *vp = ap->a_vp;
555 struct fifoinfo *fip;
556 int error1, error2;
557 lwkt_tokref vlock;
559 lwkt_gettoken(&vlock, &vp->v_token);
560 fip = vp->v_fifoinfo;
561 if (ap->a_fflag & FREAD) {
562 fip->fi_readers--;
563 if (fip->fi_readers == 0)
564 socantsendmore(fip->fi_writesock);
566 if (ap->a_fflag & FWRITE) {
567 fip->fi_writers--;
568 if (fip->fi_writers == 0)
569 socantrcvmore(fip->fi_readsock);
571 if (vp->v_sysref.refcnt > 1) {
572 vop_stdclose(ap);
573 lwkt_reltoken(&vlock);
574 return (0);
576 error1 = soclose(fip->fi_readsock, FNONBLOCK);
577 error2 = soclose(fip->fi_writesock, FNONBLOCK);
578 FREE(fip, M_FIFOINFO);
579 vp->v_fifoinfo = NULL;
580 if (error1) {
581 error2 = error1;
582 } else {
583 vop_stdclose(ap);
585 lwkt_reltoken(&vlock);
586 return (error2);
591 * Print out internal contents of a fifo vnode.
594 fifo_printinfo(struct vnode *vp)
596 struct fifoinfo *fip = vp->v_fifoinfo;
598 kprintf(", fifo with %ld readers and %ld writers",
599 fip->fi_readers, fip->fi_writers);
600 return (0);
604 * Print out the contents of a fifo vnode.
606 * fifo_print(struct vnode *a_vp)
608 static int
609 fifo_print(struct vop_print_args *ap)
611 kprintf("tag VT_NON");
612 fifo_printinfo(ap->a_vp);
613 kprintf("\n");
614 return (0);
618 * Return POSIX pathconf information applicable to fifo's.
620 * fifo_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
623 fifo_pathconf(struct vop_pathconf_args *ap)
625 switch (ap->a_name) {
626 case _PC_LINK_MAX:
627 *ap->a_retval = LINK_MAX;
628 return (0);
629 case _PC_PIPE_BUF:
630 *ap->a_retval = PIPE_BUF;
631 return (0);
632 case _PC_CHOWN_RESTRICTED:
633 *ap->a_retval = 1;
634 return (0);
635 default:
636 return (EINVAL);
638 /* NOTREACHED */
642 * Fifo advisory byte-level locks.
644 * fifo_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
645 * int a_flags)
647 /* ARGSUSED */
648 static int
649 fifo_advlock(struct vop_advlock_args *ap)
651 return ((ap->a_flags & F_POSIX) ? EINVAL : EOPNOTSUPP);
655 * Fifo bad operation
657 static int
658 fifo_badop(void)
660 panic("fifo_badop called");
661 /* NOTREACHED */