2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
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. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)sys_generic.c 8.5 (Berkeley) 1/21/94
39 * $FreeBSD: src/sys/kern/sys_generic.c,v 1.55.2.10 2001/03/17 10:39:32 peter Exp $
40 * $DragonFly: src/sys/kern/sys_generic.c,v 1.49 2008/05/05 22:09:44 dillon Exp $
43 #include "opt_ktrace.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/sysproto.h>
48 #include <sys/filedesc.h>
49 #include <sys/filio.h>
50 #include <sys/fcntl.h>
53 #include <sys/signalvar.h>
54 #include <sys/socketvar.h>
56 #include <sys/kernel.h>
57 #include <sys/kern_syscall.h>
58 #include <sys/malloc.h>
59 #include <sys/mapped_ioctl.h>
61 #include <sys/queue.h>
62 #include <sys/resourcevar.h>
63 #include <sys/sysctl.h>
64 #include <sys/sysent.h>
67 #include <sys/ktrace.h>
70 #include <vm/vm_page.h>
71 #include <sys/file2.h>
73 #include <machine/limits.h>
75 static MALLOC_DEFINE(M_IOCTLOPS, "ioctlops", "ioctl data buffer");
76 static MALLOC_DEFINE(M_IOCTLMAP, "ioctlmap", "mapped ioctl handler buffer");
77 static MALLOC_DEFINE(M_SELECT, "select", "select() buffer");
78 MALLOC_DEFINE(M_IOV, "iov", "large iov's");
80 static int doselect(int nd, fd_set *in, fd_set *ou, fd_set *ex,
81 struct timeval *tv, int *res);
82 static int pollscan (struct proc *, struct pollfd *, u_int, int *);
83 static int selscan (struct proc *, fd_mask **, fd_mask **,
85 static int dofileread(int, struct file *, struct uio *, int, size_t *);
86 static int dofilewrite(int, struct file *, struct uio *, int, size_t *);
94 sys_read(struct read_args *uap)
96 struct thread *td = curthread;
101 if ((ssize_t)uap->nbyte < 0)
104 aiov.iov_base = uap->buf;
105 aiov.iov_len = uap->nbyte;
106 auio.uio_iov = &aiov;
108 auio.uio_offset = -1;
109 auio.uio_resid = uap->nbyte;
110 auio.uio_rw = UIO_READ;
111 auio.uio_segflg = UIO_USERSPACE;
114 error = kern_preadv(uap->fd, &auio, 0, &uap->sysmsg_szresult);
119 * Positioned (Pread) read system call
124 sys_extpread(struct extpread_args *uap)
126 struct thread *td = curthread;
132 if ((ssize_t)uap->nbyte < 0)
135 aiov.iov_base = uap->buf;
136 aiov.iov_len = uap->nbyte;
137 auio.uio_iov = &aiov;
139 auio.uio_offset = uap->offset;
140 auio.uio_resid = uap->nbyte;
141 auio.uio_rw = UIO_READ;
142 auio.uio_segflg = UIO_USERSPACE;
145 flags = uap->flags & O_FMASK;
146 if (uap->offset != (off_t)-1)
149 error = kern_preadv(uap->fd, &auio, flags, &uap->sysmsg_szresult);
154 * Scatter read system call.
159 sys_readv(struct readv_args *uap)
161 struct thread *td = curthread;
163 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
166 error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt,
171 auio.uio_iovcnt = uap->iovcnt;
172 auio.uio_offset = -1;
173 auio.uio_rw = UIO_READ;
174 auio.uio_segflg = UIO_USERSPACE;
177 error = kern_preadv(uap->fd, &auio, 0, &uap->sysmsg_szresult);
179 iovec_free(&iov, aiov);
185 * Scatter positioned read system call.
190 sys_extpreadv(struct extpreadv_args *uap)
192 struct thread *td = curthread;
194 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
198 error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt,
203 auio.uio_iovcnt = uap->iovcnt;
204 auio.uio_offset = uap->offset;
205 auio.uio_rw = UIO_READ;
206 auio.uio_segflg = UIO_USERSPACE;
209 flags = uap->flags & O_FMASK;
210 if (uap->offset != (off_t)-1)
213 error = kern_preadv(uap->fd, &auio, flags, &uap->sysmsg_szresult);
215 iovec_free(&iov, aiov);
223 kern_preadv(int fd, struct uio *auio, int flags, size_t *res)
225 struct thread *td = curthread;
226 struct proc *p = td->td_proc;
232 fp = holdfp(p->p_fd, fd, FREAD);
235 if (flags & O_FOFFSET && fp->f_type != DTYPE_VNODE) {
238 error = dofileread(fd, fp, auio, flags, res);
245 * Common code for readv and preadv that reads data in
246 * from a file using the passed in uio, offset, and flags.
248 * MPALMOSTSAFE - ktrace needs help
251 dofileread(int fd, struct file *fp, struct uio *auio, int flags, size_t *res)
253 struct thread *td = curthread;
257 struct iovec *ktriov = NULL;
263 * if tracing, save a copy of iovec
265 if (KTRPOINT(td, KTR_GENIO)) {
266 int iovlen = auio->uio_iovcnt * sizeof(struct iovec);
268 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
269 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
273 len = auio->uio_resid;
274 error = fo_read(fp, auio, fp->f_cred, flags);
276 if (auio->uio_resid != len && (error == ERESTART ||
277 error == EINTR || error == EWOULDBLOCK))
281 if (ktriov != NULL) {
283 ktruio.uio_iov = ktriov;
284 ktruio.uio_resid = len - auio->uio_resid;
286 ktrgenio(td->td_lwp, fd, UIO_READ, &ktruio, error);
289 FREE(ktriov, M_TEMP);
293 *res = len - auio->uio_resid;
304 sys_write(struct write_args *uap)
306 struct thread *td = curthread;
311 if ((ssize_t)uap->nbyte < 0)
314 aiov.iov_base = (void *)(uintptr_t)uap->buf;
315 aiov.iov_len = uap->nbyte;
316 auio.uio_iov = &aiov;
318 auio.uio_offset = -1;
319 auio.uio_resid = uap->nbyte;
320 auio.uio_rw = UIO_WRITE;
321 auio.uio_segflg = UIO_USERSPACE;
324 error = kern_pwritev(uap->fd, &auio, 0, &uap->sysmsg_szresult);
335 sys_extpwrite(struct extpwrite_args *uap)
337 struct thread *td = curthread;
343 if ((ssize_t)uap->nbyte < 0)
346 aiov.iov_base = (void *)(uintptr_t)uap->buf;
347 aiov.iov_len = uap->nbyte;
348 auio.uio_iov = &aiov;
350 auio.uio_offset = uap->offset;
351 auio.uio_resid = uap->nbyte;
352 auio.uio_rw = UIO_WRITE;
353 auio.uio_segflg = UIO_USERSPACE;
356 flags = uap->flags & O_FMASK;
357 if (uap->offset != (off_t)-1)
359 error = kern_pwritev(uap->fd, &auio, flags, &uap->sysmsg_szresult);
367 sys_writev(struct writev_args *uap)
369 struct thread *td = curthread;
371 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
374 error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt,
379 auio.uio_iovcnt = uap->iovcnt;
380 auio.uio_offset = -1;
381 auio.uio_rw = UIO_WRITE;
382 auio.uio_segflg = UIO_USERSPACE;
385 error = kern_pwritev(uap->fd, &auio, 0, &uap->sysmsg_szresult);
387 iovec_free(&iov, aiov);
393 * Gather positioned write system call
398 sys_extpwritev(struct extpwritev_args *uap)
400 struct thread *td = curthread;
402 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
406 error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt,
411 auio.uio_iovcnt = uap->iovcnt;
412 auio.uio_offset = uap->offset;
413 auio.uio_rw = UIO_WRITE;
414 auio.uio_segflg = UIO_USERSPACE;
417 flags = uap->flags & O_FMASK;
418 if (uap->offset != (off_t)-1)
421 error = kern_pwritev(uap->fd, &auio, flags, &uap->sysmsg_szresult);
423 iovec_free(&iov, aiov);
431 kern_pwritev(int fd, struct uio *auio, int flags, size_t *res)
433 struct thread *td = curthread;
434 struct proc *p = td->td_proc;
440 fp = holdfp(p->p_fd, fd, FWRITE);
443 else if ((flags & O_FOFFSET) && fp->f_type != DTYPE_VNODE) {
446 error = dofilewrite(fd, fp, auio, flags, res);
454 * Common code for writev and pwritev that writes data to
455 * a file using the passed in uio, offset, and flags.
457 * MPALMOSTSAFE - ktrace needs help
460 dofilewrite(int fd, struct file *fp, struct uio *auio, int flags, size_t *res)
462 struct thread *td = curthread;
463 struct lwp *lp = td->td_lwp;
467 struct iovec *ktriov = NULL;
473 * if tracing, save a copy of iovec and uio
475 if (KTRPOINT(td, KTR_GENIO)) {
476 int iovlen = auio->uio_iovcnt * sizeof(struct iovec);
478 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
479 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
483 len = auio->uio_resid;
484 error = fo_write(fp, auio, fp->f_cred, flags);
486 if (auio->uio_resid != len && (error == ERESTART ||
487 error == EINTR || error == EWOULDBLOCK))
489 /* Socket layer is responsible for issuing SIGPIPE. */
490 if (error == EPIPE) {
492 lwpsignal(lp->lwp_proc, lp, SIGPIPE);
497 if (ktriov != NULL) {
499 ktruio.uio_iov = ktriov;
500 ktruio.uio_resid = len - auio->uio_resid;
502 ktrgenio(lp, fd, UIO_WRITE, &ktruio, error);
505 FREE(ktriov, M_TEMP);
509 *res = len - auio->uio_resid;
519 sys_ioctl(struct ioctl_args *uap)
521 return(mapped_ioctl(uap->fd, uap->com, uap->data, NULL, &uap->sysmsg));
524 struct ioctl_map_entry {
526 struct ioctl_map_range *cmd_ranges;
527 LIST_ENTRY(ioctl_map_entry) entries;
531 * The true heart of all ioctl syscall handlers (native, emulation).
532 * If map != NULL, it will be searched for a matching entry for com,
533 * and appropriate conversions/conversion functions will be utilized.
536 mapped_ioctl(int fd, u_long com, caddr_t uspc_data, struct ioctl_map *map,
539 struct thread *td = curthread;
540 struct proc *p = td->td_proc;
543 struct ioctl_map_range *iomc = NULL;
549 #define STK_PARAMS 128
551 char stkbuf[STK_PARAMS];
558 fp = holdfp(p->p_fd, fd, FREAD|FWRITE);
562 if (map != NULL) { /* obey translation map */
564 struct ioctl_map_entry *e;
566 maskcmd = com & map->mask;
568 LIST_FOREACH(e, &map->mapping, entries) {
569 for (iomc = e->cmd_ranges; iomc->start != 0 ||
570 iomc->maptocmd != 0 || iomc->wrapfunc != NULL ||
571 iomc->mapfunc != NULL;
573 if (maskcmd >= iomc->start &&
574 maskcmd <= iomc->end)
578 /* Did we find a match? */
579 if (iomc->start != 0 || iomc->maptocmd != 0 ||
580 iomc->wrapfunc != NULL || iomc->mapfunc != NULL)
585 (iomc->start == 0 && iomc->maptocmd == 0
586 && iomc->wrapfunc == NULL && iomc->mapfunc == NULL)) {
587 kprintf("%s: 'ioctl' fd=%d, cmd=0x%lx ('%c',%d) not implemented\n",
588 map->sys, fd, maskcmd,
589 (int)((maskcmd >> 8) & 0xff),
590 (int)(maskcmd & 0xff));
596 * If it's a non-range one to one mapping, maptocmd should be
597 * correct. If it's a ranged one to one mapping, we pass the
598 * original value of com, and for a range mapped to a different
599 * range, we always need a mapping function to translate the
600 * ioctl to our native ioctl. Ex. 6500-65ff <-> 9500-95ff
602 if (iomc->start == iomc->end && iomc->maptocmd == iomc->maptoend) {
603 com = iomc->maptocmd;
604 } else if (iomc->start == iomc->maptocmd && iomc->end == iomc->maptoend) {
605 if (iomc->mapfunc != NULL)
606 com = iomc->mapfunc(iomc->start, iomc->end,
607 iomc->start, iomc->end,
610 if (iomc->mapfunc != NULL) {
611 com = iomc->mapfunc(iomc->start, iomc->end,
612 iomc->maptocmd, iomc->maptoend,
615 kprintf("%s: Invalid mapping for fd=%d, cmd=%#lx ('%c',%d)\n",
616 map->sys, fd, maskcmd,
617 (int)((maskcmd >> 8) & 0xff),
618 (int)(maskcmd & 0xff));
627 error = fclrfdflags(p->p_fd, fd, UF_EXCLOSE);
630 error = fsetfdflags(p->p_fd, fd, UF_EXCLOSE);
635 * Interpret high order word to find amount of data to be
636 * copied to/from the user's address space.
638 size = IOCPARM_LEN(com);
639 if (size > IOCPARM_MAX) {
645 if (size > sizeof (ubuf.stkbuf)) {
646 memp = kmalloc(size, M_IOCTLOPS, M_WAITOK);
651 if ((com & IOC_IN) != 0) {
653 error = copyin(uspc_data, data, (size_t)size);
656 kfree(memp, M_IOCTLOPS);
660 *(caddr_t *)data = uspc_data;
662 } else if ((com & IOC_OUT) != 0 && size) {
664 * Zero the buffer so the user always
665 * gets back something deterministic.
667 bzero(data, (size_t)size);
668 } else if ((com & IOC_VOID) != 0) {
669 *(caddr_t *)data = uspc_data;
674 if ((tmp = *(int *)data))
675 fp->f_flag |= FNONBLOCK;
677 fp->f_flag &= ~FNONBLOCK;
682 if ((tmp = *(int *)data))
683 fp->f_flag |= FASYNC;
685 fp->f_flag &= ~FASYNC;
686 error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, cred, msg);
691 * If there is a override function,
692 * call it instead of directly routing the call
694 if (map != NULL && iomc->wrapfunc != NULL)
695 error = iomc->wrapfunc(fp, com, ocom, data, cred);
697 error = fo_ioctl(fp, com, data, cred, msg);
699 * Copy any data to user, size was
700 * already set and checked above.
702 if (error == 0 && (com & IOC_OUT) != 0 && size != 0)
703 error = copyout(data, uspc_data, (size_t)size);
707 kfree(memp, M_IOCTLOPS);
714 mapped_ioctl_register_handler(struct ioctl_map_handler *he)
716 struct ioctl_map_entry *ne;
718 KKASSERT(he != NULL && he->map != NULL && he->cmd_ranges != NULL &&
719 he->subsys != NULL && *he->subsys != '\0');
721 ne = kmalloc(sizeof(struct ioctl_map_entry), M_IOCTLMAP, M_WAITOK);
723 ne->subsys = he->subsys;
724 ne->cmd_ranges = he->cmd_ranges;
726 LIST_INSERT_HEAD(&he->map->mapping, ne, entries);
732 mapped_ioctl_unregister_handler(struct ioctl_map_handler *he)
734 struct ioctl_map_entry *ne;
736 KKASSERT(he != NULL && he->map != NULL && he->cmd_ranges != NULL);
738 LIST_FOREACH(ne, &he->map->mapping, entries) {
739 if (ne->cmd_ranges != he->cmd_ranges)
741 LIST_REMOVE(ne, entries);
742 kfree(ne, M_IOCTLMAP);
748 static int nselcoll; /* Select collisions since boot */
750 SYSCTL_INT(_kern, OID_AUTO, nselcoll, CTLFLAG_RD, &nselcoll, 0, "");
753 * Select system call.
756 sys_select(struct select_args *uap)
759 struct timeval *ktvp;
763 * Get timeout if any.
765 if (uap->tv != NULL) {
766 error = copyin(uap->tv, &ktv, sizeof (ktv));
769 error = itimerfix(&ktv);
780 error = doselect(uap->nd, uap->in, uap->ou, uap->ex, ktvp,
781 &uap->sysmsg_result);
788 * Pselect system call.
791 sys_pselect(struct pselect_args *uap)
793 struct thread *td = curthread;
794 struct lwp *lp = td->td_lwp;
797 struct timeval *ktvp;
802 * Get timeout if any and convert it.
803 * Round up during conversion to avoid timeout going off early.
805 if (uap->ts != NULL) {
806 error = copyin(uap->ts, &kts, sizeof (kts));
809 ktv.tv_sec = kts.tv_sec;
810 ktv.tv_usec = (kts.tv_nsec + 999) / 1000;
811 error = itimerfix(&ktv);
820 * Install temporary signal mask if any provided.
822 if (uap->sigmask != NULL) {
823 error = copyin(uap->sigmask, &sigmask, sizeof(sigmask));
826 lp->lwp_oldsigmask = lp->lwp_sigmask;
827 SIG_CANTMASK(sigmask);
828 lp->lwp_sigmask = sigmask;
834 error = doselect(uap->nd, uap->in, uap->ou, uap->ex, ktvp,
835 &uap->sysmsg_result);
837 if (uap->sigmask != NULL) {
838 /* doselect() responsible for turning ERESTART into EINTR */
839 KKASSERT(error != ERESTART);
840 if (error == EINTR) {
842 * We can't restore the previous signal mask now
843 * because it could block the signal that interrupted
844 * us. So make a note to restore it after executing
847 lp->lwp_flag |= LWP_OLDMASK;
850 * No handler to run. Restore previous mask immediately.
852 lp->lwp_sigmask = lp->lwp_oldsigmask;
860 * Common code for sys_select() and sys_pselect().
862 * in, out and ex are userland pointers. tv must point to validated
863 * kernel-side timeout value or NULL for infinite timeout. res must
864 * point to syscall return value.
867 doselect(int nd, fd_set *in, fd_set *ou, fd_set *ex, struct timeval *tv,
870 struct lwp *lp = curthread->td_lwp;
871 struct proc *p = curproc;
874 * The magic 2048 here is chosen to be just enough for FD_SETSIZE
875 * infds with the new FD_SETSIZE of 1024, and more than enough for
876 * FD_SETSIZE infds, outfds and exceptfds with the old FD_SETSIZE
879 fd_mask s_selbits[howmany(2048, NFDBITS)];
880 fd_mask *ibits[3], *obits[3], *selbits, *sbp;
881 struct timeval atv, rtv, ttv;
882 int ncoll, error, timo;
883 u_int nbufbytes, ncpbytes, nfdbits;
887 if (nd > p->p_fd->fd_nfiles)
888 nd = p->p_fd->fd_nfiles; /* forgiving; slightly wrong */
891 * Allocate just enough bits for the non-null fd_sets. Use the
892 * preallocated auto buffer if possible.
894 nfdbits = roundup(nd, NFDBITS);
895 ncpbytes = nfdbits / NBBY;
898 nbufbytes += 2 * ncpbytes;
900 nbufbytes += 2 * ncpbytes;
902 nbufbytes += 2 * ncpbytes;
903 if (nbufbytes <= sizeof s_selbits)
904 selbits = &s_selbits[0];
906 selbits = kmalloc(nbufbytes, M_SELECT, M_WAITOK);
909 * Assign pointers into the bit buffers and fetch the input bits.
910 * Put the output buffers together so that they can be bzeroed
914 #define getbits(name, x) \
919 ibits[x] = sbp + nbufbytes / 2 / sizeof *sbp; \
921 sbp += ncpbytes / sizeof *sbp; \
922 error = copyin(name, ibits[x], ncpbytes); \
932 bzero(selbits, nbufbytes / 2);
936 getmicrouptime(&rtv);
937 timevaladd(&atv, &rtv);
945 lp->lwp_flag |= LWP_SELECT;
946 error = selscan(p, ibits, obits, nd, res);
949 if (atv.tv_sec || atv.tv_usec) {
950 getmicrouptime(&rtv);
951 if (timevalcmp(&rtv, &atv, >=))
954 timevalsub(&ttv, &rtv);
955 timo = ttv.tv_sec > 24 * 60 * 60 ?
956 24 * 60 * 60 * hz : tvtohz_high(&ttv);
959 if ((lp->lwp_flag & LWP_SELECT) == 0 || nselcoll != ncoll) {
963 lp->lwp_flag &= ~LWP_SELECT;
965 error = tsleep((caddr_t)&selwait, PCATCH, "select", timo);
971 lp->lwp_flag &= ~LWP_SELECT;
972 /* select is not restarted after signals... */
973 if (error == ERESTART)
975 if (error == EWOULDBLOCK)
977 #define putbits(name, x) \
978 if (name && (error2 = copyout(obits[x], name, ncpbytes))) \
988 if (selbits != &s_selbits[0])
989 kfree(selbits, M_SELECT);
994 selscan(struct proc *p, fd_mask **ibits, fd_mask **obits, int nfd, int *res)
1000 /* Note: backend also returns POLLHUP/POLLERR if appropriate. */
1001 static int flag[3] = { POLLRDNORM, POLLWRNORM, POLLRDBAND };
1003 for (msk = 0; msk < 3; msk++) {
1004 if (ibits[msk] == NULL)
1006 for (i = 0; i < nfd; i += NFDBITS) {
1007 bits = ibits[msk][i/NFDBITS];
1008 /* ffs(int mask) not portable, fd_mask is long */
1009 for (fd = i; bits && fd < nfd; fd++, bits >>= 1) {
1012 fp = holdfp(p->p_fd, fd, -1);
1015 if (fo_poll(fp, flag[msk], fp->f_cred)) {
1016 obits[msk][(fd)/NFDBITS] |=
1017 ((fd_mask)1 << ((fd) % NFDBITS));
1032 sys_poll(struct poll_args *uap)
1034 struct pollfd *bits;
1035 struct pollfd smallbits[32];
1036 struct timeval atv, rtv, ttv;
1037 int ncoll, error = 0, timo;
1040 struct lwp *lp = curthread->td_lwp;
1041 struct proc *p = curproc;
1045 * This is kinda bogus. We have fd limits, but that is not
1046 * really related to the size of the pollfd array. Make sure
1047 * we let the process use at least FD_SETSIZE entries and at
1048 * least enough for the current limits. We want to be reasonably
1049 * safe, but not overly restrictive.
1051 if (nfds > p->p_rlimit[RLIMIT_NOFILE].rlim_cur && nfds > FD_SETSIZE)
1053 ni = nfds * sizeof(struct pollfd);
1054 if (ni > sizeof(smallbits))
1055 bits = kmalloc(ni, M_TEMP, M_WAITOK);
1058 error = copyin(uap->fds, bits, ni);
1061 if (uap->timeout != INFTIM) {
1062 atv.tv_sec = uap->timeout / 1000;
1063 atv.tv_usec = (uap->timeout % 1000) * 1000;
1064 if (itimerfix(&atv)) {
1068 getmicrouptime(&rtv);
1069 timevaladd(&atv, &rtv);
1077 lp->lwp_flag |= LWP_SELECT;
1078 error = pollscan(p, bits, nfds, &uap->sysmsg_result);
1079 if (error || uap->sysmsg_result)
1081 if (atv.tv_sec || atv.tv_usec) {
1082 getmicrouptime(&rtv);
1083 if (timevalcmp(&rtv, &atv, >=))
1086 timevalsub(&ttv, &rtv);
1087 timo = ttv.tv_sec > 24 * 60 * 60 ?
1088 24 * 60 * 60 * hz : tvtohz_high(&ttv);
1091 if ((lp->lwp_flag & LWP_SELECT) == 0 || nselcoll != ncoll) {
1095 lp->lwp_flag &= ~LWP_SELECT;
1096 error = tsleep((caddr_t)&selwait, PCATCH, "poll", timo);
1101 lp->lwp_flag &= ~LWP_SELECT;
1102 /* poll is not restarted after signals... */
1103 if (error == ERESTART)
1105 if (error == EWOULDBLOCK)
1108 error = copyout(bits, uap->fds, ni);
1113 if (ni > sizeof(smallbits))
1114 kfree(bits, M_TEMP);
1119 pollscan(struct proc *p, struct pollfd *fds, u_int nfd, int *res)
1125 for (i = 0; i < nfd; i++, fds++) {
1126 if (fds->fd >= p->p_fd->fd_nfiles) {
1127 fds->revents = POLLNVAL;
1129 } else if (fds->fd < 0) {
1132 fp = holdfp(p->p_fd, fds->fd, -1);
1134 fds->revents = POLLNVAL;
1138 * Note: backend also returns POLLHUP and
1139 * POLLERR if appropriate.
1141 fds->revents = fo_poll(fp, fds->events,
1143 if (fds->revents != 0)
1154 * OpenBSD poll system call.
1155 * XXX this isn't quite a true representation.. OpenBSD uses select ops.
1158 sys_openbsd_poll(struct openbsd_poll_args *uap)
1160 return (sys_poll((struct poll_args *)uap));
1165 seltrue(cdev_t dev, int events)
1167 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
1171 * Record a select request. A global wait must be used since a process/thread
1172 * might go away after recording its request.
1175 selrecord(struct thread *selector, struct selinfo *sip)
1178 struct lwp *lp = NULL;
1180 if (selector->td_lwp == NULL)
1181 panic("selrecord: thread needs a process");
1183 if (sip->si_pid == selector->td_proc->p_pid &&
1184 sip->si_tid == selector->td_lwp->lwp_tid)
1186 if (sip->si_pid && (p = pfind(sip->si_pid)))
1187 lp = lwp_rb_tree_RB_LOOKUP(&p->p_lwp_tree, sip->si_tid);
1188 if (lp != NULL && lp->lwp_wchan == (caddr_t)&selwait) {
1189 sip->si_flags |= SI_COLL;
1191 sip->si_pid = selector->td_proc->p_pid;
1192 sip->si_tid = selector->td_lwp->lwp_tid;
1197 * Do a wakeup when a selectable event occurs.
1200 selwakeup(struct selinfo *sip)
1203 struct lwp *lp = NULL;
1205 if (sip->si_pid == 0)
1207 if (sip->si_flags & SI_COLL) {
1209 sip->si_flags &= ~SI_COLL;
1210 wakeup((caddr_t)&selwait); /* YYY fixable */
1212 p = pfind(sip->si_pid);
1216 lp = lwp_rb_tree_RB_LOOKUP(&p->p_lwp_tree, sip->si_tid);
1221 if (lp->lwp_wchan == (caddr_t)&selwait) {
1223 * Flag the process to break the tsleep when
1224 * setrunnable is called, but only call setrunnable
1225 * here if the process is not in a stopped state.
1227 lp->lwp_flag |= LWP_BREAKTSLEEP;
1228 if (p->p_stat != SSTOP)
1230 } else if (lp->lwp_flag & LWP_SELECT) {
1231 lp->lwp_flag &= ~LWP_SELECT;