xref: /dflybsd-src/sys/kern/sys_generic.c (revision bbd7e13329c5bb82b8ee744e2c8780d994efbfbf)
1 /*
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.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
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. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)sys_generic.c	8.5 (Berkeley) 1/21/94
35  * $FreeBSD: src/sys/kern/sys_generic.c,v 1.55.2.10 2001/03/17 10:39:32 peter Exp $
36  */
37 
38 #include "opt_ktrace.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/sysmsg.h>
43 #include <sys/event.h>
44 #include <sys/filedesc.h>
45 #include <sys/filio.h>
46 #include <sys/fcntl.h>
47 #include <sys/file.h>
48 #include <sys/proc.h>
49 #include <sys/signalvar.h>
50 #include <sys/socketvar.h>
51 #include <sys/malloc.h>
52 #include <sys/uio.h>
53 #include <sys/kernel.h>
54 #include <sys/kern_syscall.h>
55 #include <sys/mapped_ioctl.h>
56 #include <sys/poll.h>
57 #include <sys/queue.h>
58 #include <sys/resourcevar.h>
59 #include <sys/socketops.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysent.h>
62 #include <sys/buf.h>
63 #ifdef KTRACE
64 #include <sys/ktrace.h>
65 #endif
66 #include <vm/vm.h>
67 #include <vm/vm_page.h>
68 
69 #include <sys/file2.h>
70 #include <sys/spinlock2.h>
71 #include <sys/signal2.h>
72 
73 #include <machine/limits.h>
74 
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");
79 
80 static struct krate krate_poll = { .freq = 1 };
81 
82 typedef struct kfd_set {
83         fd_mask	fds_bits[2];
84 } kfd_set;
85 
86 enum select_copyin_states {
87     COPYIN_READ, COPYIN_WRITE, COPYIN_EXCEPT, COPYIN_DONE };
88 
89 struct select_kevent_copyin_args {
90 	kfd_set		*read_set;
91 	kfd_set		*write_set;
92 	kfd_set		*except_set;
93 	int		active_set;	/* One of select_copyin_states */
94 	struct lwp	*lwp;		/* Pointer to our lwp */
95 	int		num_fds;	/* Number of file descriptors (syscall arg) */
96 	int		proc_fds;	/* Processed fd's (wraps) */
97 	int		error;		/* Returned to userland */
98 };
99 
100 struct poll_kevent_copyin_args {
101 	struct lwp	*lwp;
102 	struct pollfd	*fds;
103 	int		nfds;
104 	int		pfds;
105 	int		error;
106 };
107 
108 static struct lwkt_token mioctl_token = LWKT_TOKEN_INITIALIZER(mioctl_token);
109 
110 static int 	doselect(int nd, fd_set *in, fd_set *ou, fd_set *ex,
111 			 struct timespec *ts, int *res);
112 static int	dopoll(int nfds, struct pollfd *fds, struct timespec *ts,
113 		       int *res, int flags);
114 static int	dofileread(int, struct file *, struct uio *, int, size_t *);
115 static int	dofilewrite(int, struct file *, struct uio *, int, size_t *);
116 
117 /*
118  * Read system call.
119  *
120  * MPSAFE
121  */
122 int
sys_read(struct sysmsg * sysmsg,const struct read_args * uap)123 sys_read(struct sysmsg *sysmsg, const struct read_args *uap)
124 {
125 	struct thread *td = curthread;
126 	struct uio auio;
127 	struct iovec aiov;
128 	int error;
129 
130 	if ((ssize_t)uap->nbyte < 0)
131 		error = EINVAL;
132 
133 	aiov.iov_base = uap->buf;
134 	aiov.iov_len = uap->nbyte;
135 	auio.uio_iov = &aiov;
136 	auio.uio_iovcnt = 1;
137 	auio.uio_offset = -1;
138 	auio.uio_resid = uap->nbyte;
139 	auio.uio_rw = UIO_READ;
140 	auio.uio_segflg = UIO_USERSPACE;
141 	auio.uio_td = td;
142 
143 	error = kern_preadv(uap->fd, &auio, 0, &sysmsg->sysmsg_szresult);
144 	return(error);
145 }
146 
147 /*
148  * Positioned (Pread) read system call
149  *
150  * MPSAFE
151  */
152 int
sys_extpread(struct sysmsg * sysmsg,const struct extpread_args * uap)153 sys_extpread(struct sysmsg *sysmsg, const struct extpread_args *uap)
154 {
155 	struct thread *td = curthread;
156 	struct uio auio;
157 	struct iovec aiov;
158 	int error;
159 	int flags;
160 
161 	if ((ssize_t)uap->nbyte < 0)
162 		return(EINVAL);
163 
164 	aiov.iov_base = uap->buf;
165 	aiov.iov_len = uap->nbyte;
166 	auio.uio_iov = &aiov;
167 	auio.uio_iovcnt = 1;
168 	auio.uio_offset = uap->offset;
169 	auio.uio_resid = uap->nbyte;
170 	auio.uio_rw = UIO_READ;
171 	auio.uio_segflg = UIO_USERSPACE;
172 	auio.uio_td = td;
173 
174 	flags = uap->flags & O_FMASK;
175 	if (uap->offset != (off_t)-1)
176 		flags |= O_FOFFSET;
177 
178 	error = kern_preadv(uap->fd, &auio, flags, &sysmsg->sysmsg_szresult);
179 	return(error);
180 }
181 
182 /*
183  * Scatter read system call.
184  *
185  * MPSAFE
186  */
187 int
sys_readv(struct sysmsg * sysmsg,const struct readv_args * uap)188 sys_readv(struct sysmsg *sysmsg, const struct readv_args *uap)
189 {
190 	struct thread *td = curthread;
191 	struct uio auio;
192 	struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
193 	int error;
194 
195 	error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt,
196 			     &auio.uio_resid);
197 	if (error)
198 		return (error);
199 	auio.uio_iov = iov;
200 	auio.uio_iovcnt = uap->iovcnt;
201 	auio.uio_offset = -1;
202 	auio.uio_rw = UIO_READ;
203 	auio.uio_segflg = UIO_USERSPACE;
204 	auio.uio_td = td;
205 
206 	error = kern_preadv(uap->fd, &auio, 0, &sysmsg->sysmsg_szresult);
207 
208 	iovec_free(&iov, aiov);
209 	return (error);
210 }
211 
212 
213 /*
214  * Scatter positioned read system call.
215  *
216  * MPSAFE
217  */
218 int
sys_extpreadv(struct sysmsg * sysmsg,const struct extpreadv_args * uap)219 sys_extpreadv(struct sysmsg *sysmsg, const struct extpreadv_args *uap)
220 {
221 	struct thread *td = curthread;
222 	struct uio auio;
223 	struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
224 	int error;
225 	int flags;
226 
227 	error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt,
228 			     &auio.uio_resid);
229 	if (error)
230 		return (error);
231 	auio.uio_iov = iov;
232 	auio.uio_iovcnt = uap->iovcnt;
233 	auio.uio_offset = uap->offset;
234 	auio.uio_rw = UIO_READ;
235 	auio.uio_segflg = UIO_USERSPACE;
236 	auio.uio_td = td;
237 
238 	flags = uap->flags & O_FMASK;
239 	if (uap->offset != (off_t)-1)
240 		flags |= O_FOFFSET;
241 
242 	error = kern_preadv(uap->fd, &auio, flags, &sysmsg->sysmsg_szresult);
243 
244 	iovec_free(&iov, aiov);
245 	return(error);
246 }
247 
248 /*
249  * MPSAFE
250  */
251 int
kern_preadv(int fd,struct uio * auio,int flags,size_t * res)252 kern_preadv(int fd, struct uio *auio, int flags, size_t *res)
253 {
254 	struct thread *td = curthread;
255 	struct file *fp;
256 	int error;
257 
258 	fp = holdfp(td, fd, FREAD);
259 	if (fp == NULL)
260 		return (EBADF);
261 	if (flags & O_FOFFSET && fp->f_type != DTYPE_VNODE) {
262 		error = ESPIPE;
263 	} else {
264 		error = dofileread(fd, fp, auio, flags, res);
265 	}
266 	dropfp(td, fd, fp);
267 
268 	return(error);
269 }
270 
271 /*
272  * Common code for readv and preadv that reads data in
273  * from a file using the passed in uio, offset, and flags.
274  *
275  * MPALMOSTSAFE - ktrace needs help
276  */
277 static int
dofileread(int fd,struct file * fp,struct uio * auio,int flags,size_t * res)278 dofileread(int fd, struct file *fp, struct uio *auio, int flags, size_t *res)
279 {
280 	int error;
281 	size_t len;
282 #ifdef KTRACE
283 	struct thread *td = curthread;
284 	struct iovec *ktriov = NULL;
285 	struct uio ktruio;
286 #endif
287 
288 #ifdef KTRACE
289 	/*
290 	 * if tracing, save a copy of iovec
291 	 */
292 	if (KTRPOINT(td, KTR_GENIO))  {
293 		int iovlen = auio->uio_iovcnt * sizeof(struct iovec);
294 
295 		ktriov = kmalloc(iovlen, M_TEMP, M_WAITOK);
296 		bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
297 		ktruio = *auio;
298 	}
299 #endif
300 	len = auio->uio_resid;
301 	error = fo_read(fp, auio, fp->f_cred, flags);
302 	if (error) {
303 		if (auio->uio_resid != len && (error == ERESTART ||
304 		    error == EINTR || error == EWOULDBLOCK))
305 			error = 0;
306 	}
307 #ifdef KTRACE
308 	if (ktriov != NULL) {
309 		if (error == 0) {
310 			ktruio.uio_iov = ktriov;
311 			ktruio.uio_resid = len - auio->uio_resid;
312 			ktrgenio(td->td_lwp, fd, UIO_READ, &ktruio, error);
313 		}
314 		kfree(ktriov, M_TEMP);
315 	}
316 #endif
317 	if (error == 0)
318 		*res = len - auio->uio_resid;
319 
320 	return(error);
321 }
322 
323 /*
324  * Write system call
325  *
326  * MPSAFE
327  */
328 int
sys_write(struct sysmsg * sysmsg,const struct write_args * uap)329 sys_write(struct sysmsg *sysmsg, const struct write_args *uap)
330 {
331 	struct thread *td = curthread;
332 	struct uio auio;
333 	struct iovec aiov;
334 	int error;
335 
336 	if ((ssize_t)uap->nbyte < 0)
337 		error = EINVAL;
338 
339 	aiov.iov_base = (void *)(uintptr_t)uap->buf;
340 	aiov.iov_len = uap->nbyte;
341 	auio.uio_iov = &aiov;
342 	auio.uio_iovcnt = 1;
343 	auio.uio_offset = -1;
344 	auio.uio_resid = uap->nbyte;
345 	auio.uio_rw = UIO_WRITE;
346 	auio.uio_segflg = UIO_USERSPACE;
347 	auio.uio_td = td;
348 
349 	error = kern_pwritev(uap->fd, &auio, 0, &sysmsg->sysmsg_szresult);
350 
351 	return(error);
352 }
353 
354 /*
355  * Pwrite system call
356  *
357  * MPSAFE
358  */
359 int
sys_extpwrite(struct sysmsg * sysmsg,const struct extpwrite_args * uap)360 sys_extpwrite(struct sysmsg *sysmsg, const struct extpwrite_args *uap)
361 {
362 	struct thread *td = curthread;
363 	struct uio auio;
364 	struct iovec aiov;
365 	int error;
366 	int flags;
367 
368 	if ((ssize_t)uap->nbyte < 0)
369 		error = EINVAL;
370 
371 	aiov.iov_base = (void *)(uintptr_t)uap->buf;
372 	aiov.iov_len = uap->nbyte;
373 	auio.uio_iov = &aiov;
374 	auio.uio_iovcnt = 1;
375 	auio.uio_offset = uap->offset;
376 	auio.uio_resid = uap->nbyte;
377 	auio.uio_rw = UIO_WRITE;
378 	auio.uio_segflg = UIO_USERSPACE;
379 	auio.uio_td = td;
380 
381 	flags = uap->flags & O_FMASK;
382 	if (uap->offset != (off_t)-1)
383 		flags |= O_FOFFSET;
384 	error = kern_pwritev(uap->fd, &auio, flags, &sysmsg->sysmsg_szresult);
385 	return(error);
386 }
387 
388 /*
389  * MPSAFE
390  */
391 int
sys_writev(struct sysmsg * sysmsg,const struct writev_args * uap)392 sys_writev(struct sysmsg *sysmsg, const struct writev_args *uap)
393 {
394 	struct thread *td = curthread;
395 	struct uio auio;
396 	struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
397 	int error;
398 
399 	error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt,
400 			     &auio.uio_resid);
401 	if (error)
402 		return (error);
403 	auio.uio_iov = iov;
404 	auio.uio_iovcnt = uap->iovcnt;
405 	auio.uio_offset = -1;
406 	auio.uio_rw = UIO_WRITE;
407 	auio.uio_segflg = UIO_USERSPACE;
408 	auio.uio_td = td;
409 
410 	error = kern_pwritev(uap->fd, &auio, 0, &sysmsg->sysmsg_szresult);
411 
412 	iovec_free(&iov, aiov);
413 	return (error);
414 }
415 
416 
417 /*
418  * Gather positioned write system call
419  *
420  * MPSAFE
421  */
422 int
sys_extpwritev(struct sysmsg * sysmsg,const struct extpwritev_args * uap)423 sys_extpwritev(struct sysmsg *sysmsg, const struct extpwritev_args *uap)
424 {
425 	struct thread *td = curthread;
426 	struct uio auio;
427 	struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
428 	int error;
429 	int flags;
430 
431 	error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt,
432 			     &auio.uio_resid);
433 	if (error)
434 		return (error);
435 	auio.uio_iov = iov;
436 	auio.uio_iovcnt = uap->iovcnt;
437 	auio.uio_offset = uap->offset;
438 	auio.uio_rw = UIO_WRITE;
439 	auio.uio_segflg = UIO_USERSPACE;
440 	auio.uio_td = td;
441 
442 	flags = uap->flags & O_FMASK;
443 	if (uap->offset != (off_t)-1)
444 		flags |= O_FOFFSET;
445 
446 	error = kern_pwritev(uap->fd, &auio, flags, &sysmsg->sysmsg_szresult);
447 
448 	iovec_free(&iov, aiov);
449 	return(error);
450 }
451 
452 /*
453  * MPSAFE
454  */
455 int
kern_pwritev(int fd,struct uio * auio,int flags,size_t * res)456 kern_pwritev(int fd, struct uio *auio, int flags, size_t *res)
457 {
458 	struct thread *td = curthread;
459 	struct file *fp;
460 	int error;
461 
462 	fp = holdfp(td, fd, FWRITE);
463 	if (fp == NULL)
464 		return (EBADF);
465 	else if ((flags & O_FOFFSET) && fp->f_type != DTYPE_VNODE) {
466 		error = ESPIPE;
467 	} else {
468 		error = dofilewrite(fd, fp, auio, flags, res);
469 	}
470 	dropfp(td, fd, fp);
471 
472 	return(error);
473 }
474 
475 /*
476  * Common code for writev and pwritev that writes data to
477  * a file using the passed in uio, offset, and flags.
478  *
479  * MPALMOSTSAFE - ktrace needs help
480  */
481 static int
dofilewrite(int fd,struct file * fp,struct uio * auio,int flags,size_t * res)482 dofilewrite(int fd, struct file *fp, struct uio *auio, int flags, size_t *res)
483 {
484 	struct thread *td = curthread;
485 	struct lwp *lp = td->td_lwp;
486 	int error;
487 	size_t len;
488 #ifdef KTRACE
489 	struct iovec *ktriov = NULL;
490 	struct uio ktruio;
491 #endif
492 
493 #ifdef KTRACE
494 	/*
495 	 * if tracing, save a copy of iovec and uio
496 	 */
497 	if (KTRPOINT(td, KTR_GENIO))  {
498 		int iovlen = auio->uio_iovcnt * sizeof(struct iovec);
499 
500 		ktriov = kmalloc(iovlen, M_TEMP, M_WAITOK);
501 		bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
502 		ktruio = *auio;
503 	}
504 #endif
505 	len = auio->uio_resid;
506 	error = fo_write(fp, auio, fp->f_cred, flags);
507 	if (error) {
508 		if (auio->uio_resid != len && (error == ERESTART ||
509 		    error == EINTR || error == EWOULDBLOCK))
510 			error = 0;
511 		/* Socket layer is responsible for issuing SIGPIPE. */
512 		if (error == EPIPE && fp->f_type != DTYPE_SOCKET)
513 			lwpsignal(lp->lwp_proc, lp, SIGPIPE);
514 	}
515 #ifdef KTRACE
516 	if (ktriov != NULL) {
517 		if (error == 0) {
518 			ktruio.uio_iov = ktriov;
519 			ktruio.uio_resid = len - auio->uio_resid;
520 			ktrgenio(lp, fd, UIO_WRITE, &ktruio, error);
521 		}
522 		kfree(ktriov, M_TEMP);
523 	}
524 #endif
525 	if (error == 0)
526 		*res = len - auio->uio_resid;
527 
528 	return(error);
529 }
530 
531 /*
532  * Ioctl system call
533  *
534  * MPSAFE
535  */
536 int
sys_ioctl(struct sysmsg * sysmsg,const struct ioctl_args * uap)537 sys_ioctl(struct sysmsg *sysmsg, const struct ioctl_args *uap)
538 {
539 	int error;
540 
541 	error = mapped_ioctl(uap->fd, uap->com, uap->data, NULL, sysmsg);
542 	return (error);
543 }
544 
545 struct ioctl_map_entry {
546 	const char *subsys;
547 	struct ioctl_map_range *cmd_ranges;
548 	LIST_ENTRY(ioctl_map_entry) entries;
549 };
550 
551 /*
552  * The true heart of all ioctl syscall handlers (native, emulation).
553  * If map != NULL, it will be searched for a matching entry for com,
554  * and appropriate conversions/conversion functions will be utilized.
555  *
556  * MPSAFE
557  */
558 int
mapped_ioctl(int fd,u_long com,caddr_t uspc_data,struct ioctl_map * map,struct sysmsg * msg)559 mapped_ioctl(int fd, u_long com, caddr_t uspc_data, struct ioctl_map *map,
560 	     struct sysmsg *msg)
561 {
562 	struct thread *td = curthread;
563 	struct proc *p = td->td_proc;
564 	struct ucred *cred;
565 	struct file *fp;
566 	struct ioctl_map_range *iomc = NULL;
567 	int error;
568 	u_int size;
569 	u_long ocom = com;
570 	caddr_t data, memp;
571 	int tmp;
572 #define STK_PARAMS	128
573 	union {
574 	    char stkbuf[STK_PARAMS];
575 	    long align;
576 	} ubuf;
577 
578 	KKASSERT(p);
579 	cred = td->td_ucred;
580 	memp = NULL;
581 
582 	fp = holdfp(td, fd, FREAD|FWRITE);
583 	if (fp == NULL)
584 		return(EBADF);
585 
586 	if (map != NULL) {	/* obey translation map */
587 		u_long maskcmd;
588 		struct ioctl_map_entry *e;
589 
590 		maskcmd = com & map->mask;
591 
592 		lwkt_gettoken(&mioctl_token);
593 		LIST_FOREACH(e, &map->mapping, entries) {
594 			for (iomc = e->cmd_ranges; iomc->start != 0 ||
595 			     iomc->maptocmd != 0 || iomc->wrapfunc != NULL ||
596 			     iomc->mapfunc != NULL;
597 			     iomc++) {
598 				if (maskcmd >= iomc->start &&
599 				    maskcmd <= iomc->end)
600 					break;
601 			}
602 
603 			/* Did we find a match? */
604 			if (iomc->start != 0 || iomc->maptocmd != 0 ||
605 			    iomc->wrapfunc != NULL || iomc->mapfunc != NULL)
606 				break;
607 		}
608 		lwkt_reltoken(&mioctl_token);
609 
610 		if (iomc == NULL ||
611 		    (iomc->start == 0 && iomc->maptocmd == 0
612 		     && iomc->wrapfunc == NULL && iomc->mapfunc == NULL)) {
613 			krateprintf(&krate_poll,
614 				"%s: 'ioctl' fd=%d, cmd=0x%lx ('%c',%d) "
615 				"not implemented\n",
616 			       map->sys, fd, maskcmd,
617 			       (int)((maskcmd >> 8) & 0xff),
618 			       (int)(maskcmd & 0xff));
619 			error = EINVAL;
620 			goto done;
621 		}
622 
623 		/*
624 		 * If it's a non-range one to one mapping, maptocmd should be
625 		 * correct. If it's a ranged one to one mapping, we pass the
626 		 * original value of com, and for a range mapped to a different
627 		 * range, we always need a mapping function to translate the
628 		 * ioctl to our native ioctl. Ex. 6500-65ff <-> 9500-95ff
629 		 */
630 		if (iomc->start == iomc->end && iomc->maptocmd == iomc->maptoend) {
631 			com = iomc->maptocmd;
632 		} else if (iomc->start == iomc->maptocmd && iomc->end == iomc->maptoend) {
633 			if (iomc->mapfunc != NULL)
634 				com = iomc->mapfunc(iomc->start, iomc->end,
635 						    iomc->start, iomc->end,
636 						    com, com);
637 		} else {
638 			if (iomc->mapfunc != NULL) {
639 				com = iomc->mapfunc(iomc->start, iomc->end,
640 						    iomc->maptocmd, iomc->maptoend,
641 						    com, ocom);
642 			} else {
643 				krateprintf(&krate_poll,
644 					"%s: Invalid mapping for fd=%d, "
645 					"cmd=%#lx ('%c',%d)\n",
646 					map->sys, fd, maskcmd,
647 					(int)((maskcmd >> 8) & 0xff),
648 					(int)(maskcmd & 0xff));
649 				error = EINVAL;
650 				goto done;
651 			}
652 		}
653 	}
654 
655 	switch (com) {
656 	case FIONCLEX:
657 		error = fclrfdflags(p->p_fd, fd, UF_EXCLOSE);
658 		goto done;
659 	case FIOCLEX:
660 		error = fsetfdflags(p->p_fd, fd, UF_EXCLOSE);
661 		goto done;
662 	}
663 
664 	/*
665 	 * Interpret high order word to find amount of data to be
666 	 * copied to/from the user's address space.
667 	 */
668 	size = IOCPARM_LEN(com);
669 	if (size > IOCPARM_MAX) {
670 		error = ENOTTY;
671 		goto done;
672 	}
673 
674 	if ((com & IOC_VOID) == 0 && size > sizeof(ubuf.stkbuf)) {
675 		memp = kmalloc(size, M_IOCTLOPS, M_WAITOK);
676 		data = memp;
677 	} else {
678 		memp = NULL;
679 		data = ubuf.stkbuf;
680 	}
681 	if (com & IOC_VOID) {
682 		*(caddr_t *)data = uspc_data;
683 	} else if (com & IOC_IN) {
684 		if (size != 0) {
685 			error = copyin(uspc_data, data, (size_t)size);
686 			if (error)
687 				goto done;
688 		} else {
689 			*(caddr_t *)data = uspc_data;
690 		}
691 	} else if ((com & IOC_OUT) != 0 && size) {
692 		/*
693 		 * Zero the buffer so the user always
694 		 * gets back something deterministic.
695 		 */
696 		bzero(data, (size_t)size);
697 	}
698 
699 	switch (com) {
700 	case FIONBIO:
701 		if ((tmp = *(int *)data))
702 			atomic_set_int(&fp->f_flag, FNONBLOCK);
703 		else
704 			atomic_clear_int(&fp->f_flag, FNONBLOCK);
705 		error = 0;
706 		break;
707 
708 	case FIOASYNC:
709 		if ((tmp = *(int *)data))
710 			atomic_set_int(&fp->f_flag, FASYNC);
711 		else
712 			atomic_clear_int(&fp->f_flag, FASYNC);
713 		error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, cred, msg);
714 		break;
715 
716 	default:
717 		/*
718 		 *  If there is a override function,
719 		 *  call it instead of directly routing the call
720 		 */
721 		if (map != NULL && iomc->wrapfunc != NULL)
722 			error = iomc->wrapfunc(fp, com, ocom, data, cred);
723 		else
724 			error = fo_ioctl(fp, com, data, cred, msg);
725 		/*
726 		 * Copy any data to user, size was
727 		 * already set and checked above.
728 		 */
729 		if (error == 0 && (com & IOC_OUT) != 0 && size != 0)
730 			error = copyout(data, uspc_data, (size_t)size);
731 		break;
732 	}
733 done:
734 	if (memp != NULL)
735 		kfree(memp, M_IOCTLOPS);
736 	dropfp(td, fd, fp);
737 
738 	return(error);
739 }
740 
741 /*
742  * MPSAFE
743  */
744 int
mapped_ioctl_register_handler(struct ioctl_map_handler * he)745 mapped_ioctl_register_handler(struct ioctl_map_handler *he)
746 {
747 	struct ioctl_map_entry *ne;
748 
749 	KKASSERT(he != NULL && he->map != NULL && he->cmd_ranges != NULL &&
750 		 he->subsys != NULL && *he->subsys != '\0');
751 
752 	ne = kmalloc(sizeof(struct ioctl_map_entry), M_IOCTLMAP,
753 		     M_WAITOK | M_ZERO);
754 
755 	ne->subsys = he->subsys;
756 	ne->cmd_ranges = he->cmd_ranges;
757 
758 	lwkt_gettoken(&mioctl_token);
759 	LIST_INSERT_HEAD(&he->map->mapping, ne, entries);
760 	lwkt_reltoken(&mioctl_token);
761 
762 	return(0);
763 }
764 
765 /*
766  * MPSAFE
767  */
768 int
mapped_ioctl_unregister_handler(struct ioctl_map_handler * he)769 mapped_ioctl_unregister_handler(struct ioctl_map_handler *he)
770 {
771 	struct ioctl_map_entry *ne;
772 	int error = EINVAL;
773 
774 	KKASSERT(he != NULL && he->map != NULL && he->cmd_ranges != NULL);
775 
776 	lwkt_gettoken(&mioctl_token);
777 	LIST_FOREACH(ne, &he->map->mapping, entries) {
778 		if (ne->cmd_ranges == he->cmd_ranges) {
779 			LIST_REMOVE(ne, entries);
780 			kfree(ne, M_IOCTLMAP);
781 			error = 0;
782 			break;
783 		}
784 	}
785 	lwkt_reltoken(&mioctl_token);
786 	return(error);
787 }
788 
789 static int	nseldebug;
790 SYSCTL_INT(_kern, OID_AUTO, nseldebug, CTLFLAG_RW, &nseldebug, 0, "");
791 
792 /*
793  * Select system call.
794  *
795  * MPSAFE
796  */
797 int
sys_select(struct sysmsg * sysmsg,const struct select_args * uap)798 sys_select(struct sysmsg *sysmsg, const struct select_args *uap)
799 {
800 	struct timeval ktv;
801 	struct timespec *ktsp, kts;
802 	int error;
803 
804 	/*
805 	 * Get timeout if any.
806 	 */
807 	if (uap->tv != NULL) {
808 		error = copyin(uap->tv, &ktv, sizeof (ktv));
809 		if (error)
810 			return (error);
811 		TIMEVAL_TO_TIMESPEC(&ktv, &kts);
812 		ktsp = &kts;
813 	} else {
814 		ktsp = NULL;
815 	}
816 
817 	/*
818 	 * Do real work.
819 	 */
820 	error = doselect(uap->nd, uap->in, uap->ou, uap->ex, ktsp,
821 			 &sysmsg->sysmsg_result);
822 
823 	return (error);
824 }
825 
826 
827 /*
828  * Pselect system call.
829  */
830 int
sys_pselect(struct sysmsg * sysmsg,const struct pselect_args * uap)831 sys_pselect(struct sysmsg *sysmsg, const struct pselect_args *uap)
832 {
833 	struct thread *td = curthread;
834 	struct lwp *lp = td->td_lwp;
835 	struct timespec *ktsp, kts;
836 	sigset_t sigmask;
837 	int error;
838 
839 	/*
840 	 * Get timeout if any.
841 	 */
842 	if (uap->ts != NULL) {
843 		error = copyin(uap->ts, &kts, sizeof (kts));
844 		if (error)
845 			return (error);
846 		ktsp = &kts;
847 	} else {
848 		ktsp = NULL;
849 	}
850 
851 	/*
852 	 * Install temporary signal mask if any provided.
853 	 */
854 	if (uap->sigmask != NULL) {
855 		error = copyin(uap->sigmask, &sigmask, sizeof(sigmask));
856 		if (error)
857 			return (error);
858 		lwkt_gettoken(&lp->lwp_proc->p_token);
859 		lp->lwp_oldsigmask = lp->lwp_sigmask;
860 		SIG_CANTMASK(sigmask);
861 		lp->lwp_sigmask = sigmask;
862 		lwkt_reltoken(&lp->lwp_proc->p_token);
863 		sigirefs_wait(lp->lwp_proc);
864 	}
865 
866 	/*
867 	 * Do real job.
868 	 */
869 	error = doselect(uap->nd, uap->in, uap->ou, uap->ex, ktsp,
870 			 &sysmsg->sysmsg_result);
871 
872 	if (uap->sigmask != NULL) {
873 		lwkt_gettoken(&lp->lwp_proc->p_token);
874 		/* doselect() responsible for turning ERESTART into EINTR */
875 		KKASSERT(error != ERESTART);
876 		if (error == EINTR) {
877 			/*
878 			 * We can't restore the previous signal mask now
879 			 * because it could block the signal that interrupted
880 			 * us.  So make a note to restore it after executing
881 			 * the handler.
882 			 */
883 			lp->lwp_flags |= LWP_OLDMASK;
884 		} else {
885 			/*
886 			 * No handler to run. Restore previous mask immediately.
887 			 */
888 			lp->lwp_sigmask = lp->lwp_oldsigmask;
889 			sigirefs_wait(lp->lwp_proc);
890 		}
891 		lwkt_reltoken(&lp->lwp_proc->p_token);
892 	}
893 
894 	return (error);
895 }
896 
897 static int
select_copyin(void * arg,struct kevent * kevp,int maxevents,int * events)898 select_copyin(void *arg, struct kevent *kevp, int maxevents, int *events)
899 {
900 	struct select_kevent_copyin_args *skap = NULL;
901 	struct kevent *kev;
902 	int fd;
903 	kfd_set *fdp = NULL;
904 	short filter = 0;
905 	u_int fflags = 0;
906 
907 	skap = (struct select_kevent_copyin_args *)arg;
908 
909 	if (*events == maxevents)
910 		return (0);
911 
912 	while (skap->active_set < COPYIN_DONE) {
913 		switch (skap->active_set) {
914 		case COPYIN_READ:
915 			/*
916 			 * Register descriptors for the read filter
917 			 */
918 			fdp = skap->read_set;
919 			filter = EVFILT_READ;
920 			fflags = NOTE_OLDAPI;
921 			if (fdp)
922 				break;
923 			++skap->active_set;
924 			skap->proc_fds = 0;
925 			/* fall through */
926 		case COPYIN_WRITE:
927 			/*
928 			 * Register descriptors for the write filter
929 			 */
930 			fdp = skap->write_set;
931 			filter = EVFILT_WRITE;
932 			fflags = NOTE_OLDAPI;
933 			if (fdp)
934 				break;
935 			++skap->active_set;
936 			skap->proc_fds = 0;
937 			/* fall through */
938 		case COPYIN_EXCEPT:
939 			/*
940 			 * Register descriptors for the exception filter
941 			 */
942 			fdp = skap->except_set;
943 			filter = EVFILT_EXCEPT;
944 			fflags = NOTE_OLDAPI | NOTE_OOB;
945 			if (fdp)
946 				break;
947 			++skap->active_set;
948 			skap->proc_fds = 0;
949 			/* fall through */
950 		case COPYIN_DONE:
951 			/*
952 			 * Nothing left to register
953 			 */
954 			return(0);
955 			/* NOT REACHED */
956 		}
957 
958 		while (skap->proc_fds < skap->num_fds) {
959 			fd = skap->proc_fds;
960 			if (FD_ISSET(fd, fdp)) {
961 				kev = &kevp[*events];
962 				EV_SET(kev, fd, filter,
963 				       EV_ADD|EV_ENABLE,
964 				       fflags, 0,
965 				       (void *)(uintptr_t)
966 					skap->lwp->lwp_kqueue_serial);
967 				FD_CLR(fd, fdp);
968 				++*events;
969 
970 				if (nseldebug) {
971 					kprintf(
972 					    "select fd %d filter %d "
973 					    "serial %ju\n", fd, filter,
974 					    (uintmax_t)
975 					    skap->lwp->lwp_kqueue_serial);
976 				}
977 			}
978 			++skap->proc_fds;
979 			if (*events == maxevents)
980 				return (0);
981 		}
982 		skap->active_set++;
983 		skap->proc_fds = 0;
984 	}
985 
986 	return (0);
987 }
988 
989 static int
select_copyout(void * arg,struct kevent * kevp,int count,int * res)990 select_copyout(void *arg, struct kevent *kevp, int count, int *res)
991 {
992 	struct select_kevent_copyin_args *skap;
993 	struct kevent kev;
994 	int i;
995 	int n;
996 
997 	skap = (struct select_kevent_copyin_args *)arg;
998 
999 	for (i = 0; i < count; ++i) {
1000 		/*
1001 		 * Filter out and delete spurious events
1002 		 */
1003 		if ((uint64_t)(uintptr_t)kevp[i].udata !=
1004 		    skap->lwp->lwp_kqueue_serial)
1005 		{
1006 			panic("select_copyout: unexpected udata");
1007 deregister:
1008 			kev = kevp[i];
1009 			kev.flags = EV_DISABLE|EV_DELETE;
1010 			n = 1;
1011 			kqueue_register(&skap->lwp->lwp_kqueue, &kev, &n, 0);
1012 			if (nseldebug) {
1013 				kprintf("select fd %ju mismatched serial %ju\n",
1014 				    (uintmax_t)kevp[i].ident,
1015 				    (uintmax_t)skap->lwp->lwp_kqueue_serial);
1016 			}
1017 			continue;
1018 		}
1019 
1020 		/*
1021 		 * Handle errors
1022 		 */
1023 		if (kevp[i].flags & EV_ERROR) {
1024 			int error = kevp[i].data;
1025 
1026 			switch (error) {
1027 			case EBADF:
1028 				/*
1029 				 * A bad file descriptor is considered a
1030 				 * fatal error for select, bail out.
1031 				 */
1032 				skap->error = error;
1033 				*res = -1;
1034 				return error;
1035 
1036 			default:
1037 				/*
1038 				 * Select silently swallows any unknown errors
1039 				 * for descriptors in the read or write sets.
1040 				 *
1041 				 * ALWAYS filter out EOPNOTSUPP errors from
1042 				 * filters (at least until all filters support
1043 				 * EVFILT_EXCEPT)
1044 				 *
1045 				 * We also filter out ENODEV since dev_dkqfilter
1046 				 * returns ENODEV if EOPNOTSUPP is returned in an
1047 				 * inner call.
1048 				 *
1049 				 * XXX: fix this
1050 				 */
1051 				if (kevp[i].filter != EVFILT_READ &&
1052 				    kevp[i].filter != EVFILT_WRITE &&
1053 				    error != EOPNOTSUPP &&
1054 				    error != ENODEV) {
1055 					skap->error = error;
1056 					*res = -1;
1057 					return error;
1058 				}
1059 				break;
1060 			}
1061 
1062 			/*
1063 			 * We must deregister any unsupported select events
1064 			 * to avoid a live-lock.
1065 			 */
1066 			if (nseldebug) {
1067 				kprintf("select fd %ju filter %d error %d\n",
1068 					(uintmax_t)kevp[i].ident,
1069 					kevp[i].filter, error);
1070 			}
1071 			goto deregister;
1072 		}
1073 
1074 		switch (kevp[i].filter) {
1075 		case EVFILT_READ:
1076 			FD_SET(kevp[i].ident, skap->read_set);
1077 			break;
1078 		case EVFILT_WRITE:
1079 			FD_SET(kevp[i].ident, skap->write_set);
1080 			break;
1081 		case EVFILT_EXCEPT:
1082 			FD_SET(kevp[i].ident, skap->except_set);
1083 			break;
1084 		}
1085 
1086 		++*res;
1087 	}
1088 
1089 	return (0);
1090 }
1091 
1092 /*
1093  * Copy select bits in from userland.  Allocate kernel memory if the
1094  * set is large.
1095  */
1096 static int
getbits(int bytes,fd_set * in_set,kfd_set ** out_set,kfd_set * tmp_set)1097 getbits(int bytes, fd_set *in_set, kfd_set **out_set, kfd_set *tmp_set)
1098 {
1099 	int error;
1100 
1101 	if (in_set) {
1102 		if (bytes < sizeof(*tmp_set))
1103 			*out_set = tmp_set;
1104 		else
1105 			*out_set = kmalloc(bytes, M_SELECT, M_WAITOK);
1106 		error = copyin(in_set, *out_set, bytes);
1107 	} else {
1108 		*out_set = NULL;
1109 		error = 0;
1110 	}
1111 	return (error);
1112 }
1113 
1114 /*
1115  * Copy returned select bits back out to userland.
1116  */
1117 static int
putbits(int bytes,kfd_set * in_set,fd_set * out_set)1118 putbits(int bytes, kfd_set *in_set, fd_set *out_set)
1119 {
1120 	int error;
1121 
1122 	if (in_set) {
1123 		error = copyout(in_set, out_set, bytes);
1124 	} else {
1125 		error = 0;
1126 	}
1127 	return (error);
1128 }
1129 
1130 static int
dotimeout_only(struct timespec * ts)1131 dotimeout_only(struct timespec *ts)
1132 {
1133 	return(nanosleep1(ts, NULL));
1134 }
1135 
1136 /*
1137  * Common code for sys_select() and sys_pselect().
1138  *
1139  * in, out and ex are userland pointers.  ts must point to validated
1140  * kernel-side timeout value or NULL for infinite timeout.  res must
1141  * point to syscall return value.
1142  */
1143 static int
doselect(int nd,fd_set * read,fd_set * write,fd_set * except,struct timespec * ts,int * res)1144 doselect(int nd, fd_set *read, fd_set *write, fd_set *except,
1145 	 struct timespec *ts, int *res)
1146 {
1147 	struct proc *p = curproc;
1148 	struct select_kevent_copyin_args *kap, ka;
1149 	int bytes, error;
1150 	kfd_set read_tmp;
1151 	kfd_set write_tmp;
1152 	kfd_set except_tmp;
1153 
1154 	*res = 0;
1155 	if (nd < 0)
1156 		return (EINVAL);
1157 	if (nd == 0 && ts)
1158 		return (dotimeout_only(ts));
1159 
1160 	if (nd > p->p_fd->fd_nfiles)		/* limit kmalloc */
1161 		nd = p->p_fd->fd_nfiles;
1162 
1163 	kap = &ka;
1164 	kap->lwp = curthread->td_lwp;
1165 	kap->num_fds = nd;
1166 	kap->proc_fds = 0;
1167 	kap->error = 0;
1168 	kap->active_set = COPYIN_READ;
1169 
1170 	/*
1171 	 * Calculate bytes based on the number of __fd_mask[] array entries
1172 	 * multiplied by the size of __fd_mask.
1173 	 */
1174 	bytes = howmany(nd, __NFDBITS) * sizeof(__fd_mask);
1175 
1176 	/* kap->read_set = NULL; not needed */
1177 	kap->write_set = NULL;
1178 	kap->except_set = NULL;
1179 
1180 	error = getbits(bytes, read, &kap->read_set, &read_tmp);
1181 	if (error == 0)
1182 		error = getbits(bytes, write, &kap->write_set, &write_tmp);
1183 	if (error == 0)
1184 		error = getbits(bytes, except, &kap->except_set, &except_tmp);
1185 	if (error)
1186 		goto done;
1187 
1188 	/*
1189 	 * NOTE: Make sure the max events passed to kern_kevent() is
1190 	 *	 effectively unlimited.  (nd * 3) accomplishes this.
1191 	 *
1192 	 *	 (*res) continues to increment as returned events are
1193 	 *	 loaded in.
1194 	 */
1195 	error = kern_kevent(&kap->lwp->lwp_kqueue, 0x7FFFFFFF, res, kap,
1196 			    select_copyin, select_copyout, ts,
1197 			    KEVENT_AUTO_STALE);
1198 	if (error == 0)
1199 		error = putbits(bytes, kap->read_set, read);
1200 	if (error == 0)
1201 		error = putbits(bytes, kap->write_set, write);
1202 	if (error == 0)
1203 		error = putbits(bytes, kap->except_set, except);
1204 
1205 	/*
1206 	 * An error from an individual event that should be passed
1207 	 * back to userland (EBADF)
1208 	 */
1209 	if (kap->error)
1210 		error = kap->error;
1211 
1212 	/*
1213 	 * Clean up.
1214 	 */
1215 done:
1216 	if (kap->read_set && kap->read_set != &read_tmp)
1217 		kfree(kap->read_set, M_SELECT);
1218 	if (kap->write_set && kap->write_set != &write_tmp)
1219 		kfree(kap->write_set, M_SELECT);
1220 	if (kap->except_set && kap->except_set != &except_tmp)
1221 		kfree(kap->except_set, M_SELECT);
1222 
1223 	kap->lwp->lwp_kqueue_serial += kap->num_fds;
1224 
1225 	return (error);
1226 }
1227 
1228 /*
1229  * Poll system call.
1230  *
1231  * MPSAFE
1232  */
1233 int
sys_poll(struct sysmsg * sysmsg,const struct poll_args * uap)1234 sys_poll(struct sysmsg *sysmsg, const struct poll_args *uap)
1235 {
1236 	struct timespec ts, *tsp;
1237 	int error;
1238 
1239 	if (uap->timeout != INFTIM) {
1240 		if (uap->timeout < 0)
1241 			return (EINVAL);
1242 		ts.tv_sec = uap->timeout / 1000;
1243 		ts.tv_nsec = (uap->timeout % 1000) * 1000 * 1000;
1244 		tsp = &ts;
1245 	} else {
1246 		tsp = NULL;
1247 	}
1248 
1249 	error = dopoll(uap->nfds, uap->fds, tsp, &sysmsg->sysmsg_result, 0);
1250 
1251 	return (error);
1252 }
1253 
1254 /*
1255  * Ppoll system call.
1256  *
1257  * MPSAFE
1258  */
1259 int
sys_ppoll(struct sysmsg * sysmsg,const struct ppoll_args * uap)1260 sys_ppoll(struct sysmsg *sysmsg, const struct ppoll_args *uap)
1261 {
1262 	struct thread *td = curthread;
1263 	struct lwp *lp = td->td_lwp;
1264 	struct timespec *ktsp, kts;
1265 	sigset_t sigmask;
1266 	int error;
1267 
1268 	/*
1269 	 * Get timeout if any.
1270 	 */
1271 	if (uap->ts != NULL) {
1272 		error = copyin(uap->ts, &kts, sizeof (kts));
1273 		if (error)
1274 			return (error);
1275 		ktsp = &kts;
1276 	} else {
1277 		ktsp = NULL;
1278 	}
1279 
1280 	/*
1281 	 * Install temporary signal mask if any provided.
1282 	 */
1283 	if (uap->sigmask != NULL) {
1284 		error = copyin(uap->sigmask, &sigmask, sizeof(sigmask));
1285 		if (error)
1286 			return (error);
1287 		lwkt_gettoken(&lp->lwp_proc->p_token);
1288 		lp->lwp_oldsigmask = lp->lwp_sigmask;
1289 		SIG_CANTMASK(sigmask);
1290 		lp->lwp_sigmask = sigmask;
1291 		lwkt_reltoken(&lp->lwp_proc->p_token);
1292 		sigirefs_wait(lp->lwp_proc);
1293 	}
1294 
1295 	error = dopoll(uap->nfds, uap->fds, ktsp, &sysmsg->sysmsg_result,
1296 	    ktsp != NULL ? KEVENT_TIMEOUT_PRECISE : 0);
1297 
1298 	if (uap->sigmask != NULL) {
1299 		lwkt_gettoken(&lp->lwp_proc->p_token);
1300 		/* dopoll() responsible for turning ERESTART into EINTR */
1301 		KKASSERT(error != ERESTART);
1302 		if (error == EINTR) {
1303 			/*
1304 			 * We can't restore the previous signal mask now
1305 			 * because it could block the signal that interrupted
1306 			 * us.  So make a note to restore it after executing
1307 			 * the handler.
1308 			 */
1309 			lp->lwp_flags |= LWP_OLDMASK;
1310 		} else {
1311 			/*
1312 			 * No handler to run. Restore previous mask immediately.
1313 			 */
1314 			lp->lwp_sigmask = lp->lwp_oldsigmask;
1315 			sigirefs_wait(lp->lwp_proc);
1316 		}
1317 		lwkt_reltoken(&lp->lwp_proc->p_token);
1318 	}
1319 
1320 	return (error);
1321 }
1322 
1323 static int
poll_copyin(void * arg,struct kevent * kevp,int maxevents,int * events)1324 poll_copyin(void *arg, struct kevent *kevp, int maxevents, int *events)
1325 {
1326 	struct poll_kevent_copyin_args *pkap;
1327 	struct pollfd *pfd;
1328 	struct kevent *kev;
1329 	int kev_count;
1330 
1331 	pkap = (struct poll_kevent_copyin_args *)arg;
1332 
1333 	while (pkap->pfds < pkap->nfds) {
1334 		pfd = &pkap->fds[pkap->pfds];
1335 
1336 		/* Clear return events */
1337 		pfd->revents = 0;
1338 
1339 		/* Do not check if fd is equal to -1 */
1340 		if (pfd->fd == -1) {
1341 			++pkap->pfds;
1342 			continue;
1343 		}
1344 
1345 		/*
1346 		 * NOTE: pfd->events == 0 implies POLLHUP in BSDs.  Used
1347 		 *	 by at least sshd and X11 udev support.
1348 		 */
1349 		kev_count = 0;
1350 		if (pfd->events == 0)
1351 			kev_count++;
1352 		if (pfd->events & (POLLIN | POLLHUP | POLLRDNORM))
1353 			kev_count++;
1354 		if (pfd->events & (POLLOUT | POLLWRNORM))
1355 			kev_count++;
1356 		if (pfd->events & (POLLPRI | POLLRDBAND))
1357 			kev_count++;
1358 
1359 		if (*events + kev_count > maxevents)
1360 			return (0);
1361 
1362 		/*
1363 		 * NOTE: A combined serial number and poll array index is
1364 		 *	 stored in kev->udata.
1365 		 *
1366 		 * NOTE: Events will be registered with KEVENT_UNIQUE_NOTES
1367 		 *	 set, using kev->data for the uniqifier.  kev->data
1368 		 *	 is an implied in the actual registration.
1369 		 */
1370 		kev = &kevp[*events];
1371 
1372 		/*
1373 		 * Implied POLLHUP
1374 		 */
1375 		if (pfd->events == 0) {
1376 			int notes = NOTE_OLDAPI | NOTE_HUPONLY;
1377 
1378 			EV_SET(kev++, pfd->fd, EVFILT_READ, EV_ADD|EV_ENABLE,
1379 			       notes, pkap->pfds, (void *)(uintptr_t)
1380 				(pkap->lwp->lwp_kqueue_serial + pkap->pfds));
1381 		}
1382 
1383 		/*
1384 		 * Nominal read events
1385 		 */
1386 		if (pfd->events & (POLLIN | POLLHUP | POLLRDNORM)) {
1387 			int notes = NOTE_OLDAPI;
1388 			if ((pfd->events & (POLLIN | POLLRDNORM)) == 0)
1389 				notes |= NOTE_HUPONLY;
1390 
1391 			EV_SET(kev++, pfd->fd, EVFILT_READ, EV_ADD|EV_ENABLE,
1392 			       notes, pkap->pfds, (void *)(uintptr_t)
1393 				(pkap->lwp->lwp_kqueue_serial + pkap->pfds));
1394 		}
1395 
1396 		/*
1397 		 * Nominal write events
1398 		 */
1399 		if (pfd->events & (POLLOUT | POLLWRNORM)) {
1400 			EV_SET(kev++, pfd->fd, EVFILT_WRITE, EV_ADD|EV_ENABLE,
1401 			       NOTE_OLDAPI, pkap->pfds, (void *)(uintptr_t)
1402 				(pkap->lwp->lwp_kqueue_serial + pkap->pfds));
1403 		}
1404 
1405 		/*
1406 		 * Nominal exceptional events
1407 		 */
1408 		if (pfd->events & (POLLPRI | POLLRDBAND)) {
1409 			EV_SET(kev++, pfd->fd, EVFILT_EXCEPT, EV_ADD|EV_ENABLE,
1410 			       NOTE_OLDAPI | NOTE_OOB, pkap->pfds,
1411 			       (void *)(uintptr_t)
1412 				(pkap->lwp->lwp_kqueue_serial + pkap->pfds));
1413 		}
1414 
1415 		if (nseldebug) {
1416 			kprintf("poll index %d/%d fd %d events %08x "
1417 			    "serial %ju\n", pkap->pfds, pkap->nfds-1,
1418 			    pfd->fd, pfd->events,
1419 			    (uintmax_t)pkap->lwp->lwp_kqueue_serial);
1420 		}
1421 
1422 		++pkap->pfds;
1423 		(*events) += kev_count;
1424 	}
1425 
1426 	return (0);
1427 }
1428 
1429 static int
poll_copyout(void * arg,struct kevent * kevp,int count,int * res)1430 poll_copyout(void *arg, struct kevent *kevp, int count, int *res)
1431 {
1432 	struct poll_kevent_copyin_args *pkap;
1433 	struct pollfd *pfd;
1434 	struct kevent kev;
1435 	int count_res;
1436 	int i;
1437 	int n;
1438 	uint64_t pi;
1439 
1440 	pkap = (struct poll_kevent_copyin_args *)arg;
1441 
1442 	for (i = 0; i < count; ++i) {
1443 		/*
1444 		 * Extract the poll array index and delete spurious events.
1445 		 * We can easily tell if the serial number is incorrect
1446 		 * by checking whether the extracted index is out of range.
1447 		 */
1448 		pi = (uint64_t)(uintptr_t)kevp[i].udata -
1449 		     pkap->lwp->lwp_kqueue_serial;
1450 		if (pi >= pkap->nfds) {
1451 			panic("poll_copyout: unexpected udata");
1452 deregister:
1453 			kev = kevp[i];
1454 			kev.flags = EV_DISABLE|EV_DELETE;
1455 			kev.data = pi;	/* uniquifier */
1456 			n = 1;
1457 			kqueue_register(&pkap->lwp->lwp_kqueue, &kev, &n,
1458 					KEVENT_UNIQUE_NOTES);
1459 			if (nseldebug) {
1460 				kprintf("poll index %ju out of range against "
1461 				    "serial %ju\n", (uintmax_t)pi,
1462 				    (uintmax_t)pkap->lwp->lwp_kqueue_serial);
1463 			}
1464 			continue;
1465 		}
1466 
1467 		/*
1468 		 * Locate the pollfd and process events
1469 		 */
1470 		pfd = &pkap->fds[pi];
1471 		if (kevp[i].ident == pfd->fd) {
1472 			/*
1473 			 * A single descriptor may generate an error against
1474 			 * more than one filter, make sure to set the
1475 			 * appropriate flags but do not increment (*res)
1476 			 * more than once.
1477 			 */
1478 			count_res = (pfd->revents == 0);
1479 			if (kevp[i].flags & EV_ERROR) {
1480 				switch(kevp[i].data) {
1481 				case EBADF:
1482 				case POLLNVAL:
1483 					/* Bad file descriptor */
1484 					if (count_res)
1485 						++*res;
1486 					pfd->revents |= POLLNVAL;
1487 					break;
1488 				default:
1489 					/*
1490 					 * Poll silently swallows any unknown
1491 					 * errors except in the case of POLLPRI
1492 					 * (OOB/urgent data).
1493 					 *
1494 					 * ALWAYS filter out EOPNOTSUPP errors
1495 					 * from filters, common applications
1496 					 * set POLLPRI|POLLRDBAND and most
1497 					 * filters do not support EVFILT_EXCEPT.
1498 					 *
1499 					 * We also filter out ENODEV since
1500 					 * dev_dkqfilter returns ENODEV if
1501 					 * EOPNOTSUPP is returned in an
1502 					 * inner call.
1503 					 *
1504 					 * XXX: fix this
1505 					 */
1506 					if (kevp[i].filter != EVFILT_READ &&
1507 					    kevp[i].filter != EVFILT_WRITE &&
1508 					    kevp[i].data != EOPNOTSUPP &&
1509 					    kevp[i].data != ENODEV) {
1510 						if (count_res)
1511 							++*res;
1512 						pfd->revents |= POLLERR;
1513 					}
1514 					break;
1515 				}
1516 				if (pfd->revents == 0 && nseldebug) {
1517 					kprintf("poll index EV_ERROR %ju fd %d "
1518 						"filter %d error %jd\n",
1519 						(uintmax_t)pi, pfd->fd,
1520 						kevp[i].filter,
1521 						(intmax_t)kevp[i].data);
1522 				}
1523 
1524 				/*
1525 				 * Silently deregister any unhandled EV_ERROR
1526 				 * condition (usually EOPNOTSUPP).
1527 				 */
1528 				if (pfd->revents == 0)
1529 					goto deregister;
1530 				continue;
1531 			}
1532 
1533 			switch (kevp[i].filter) {
1534 			case EVFILT_READ:
1535 				/*
1536 				 * NODATA on the read side can indicate a
1537 				 * half-closed situation and not necessarily
1538 				 * a disconnect, so depend on the user
1539 				 * issuing a read() and getting 0 bytes back.
1540 				 *
1541 				 * If EV_HUP is set the peer completely
1542 				 * disconnected and we can set POLLHUP.
1543 				 * Linux can return POLLHUP even if read
1544 				 * data has not been drained, so we should
1545 				 * too.
1546 				 */
1547 				/* if (kevp[i].flags & EV_NODATA) */ {
1548 					if (kevp[i].flags & EV_HUP)
1549 						pfd->revents |= POLLHUP;
1550 				}
1551 				if ((kevp[i].flags & EV_EOF) &&
1552 				    kevp[i].fflags != 0)
1553 					pfd->revents |= POLLERR;
1554 				if (pfd->events & POLLIN)
1555 					pfd->revents |= POLLIN;
1556 				if (pfd->events & POLLRDNORM)
1557 					pfd->revents |= POLLRDNORM;
1558 				break;
1559 			case EVFILT_WRITE:
1560 				/*
1561 				 * As per the OpenGroup POLLHUP is mutually
1562 				 * exclusive with the writability flags.  I
1563 				 * consider this a bit broken but...
1564 				 *
1565 				 * In this case a disconnect is implied even
1566 				 * for a half-closed (write side) situation.
1567 				 */
1568 				if (kevp[i].flags & EV_EOF) {
1569 					pfd->revents |= POLLHUP;
1570 					if (kevp[i].fflags != 0)
1571 						pfd->revents |= POLLERR;
1572 				} else {
1573 					if (pfd->events & POLLOUT)
1574 						pfd->revents |= POLLOUT;
1575 					if (pfd->events & POLLWRNORM)
1576 						pfd->revents |= POLLWRNORM;
1577 				}
1578 				break;
1579 			case EVFILT_EXCEPT:
1580 				/*
1581 				 * EV_NODATA should never be tagged for this
1582 				 * filter.
1583 				 */
1584 				if (pfd->events & POLLPRI)
1585 					pfd->revents |= POLLPRI;
1586 				if (pfd->events & POLLRDBAND)
1587 					pfd->revents |= POLLRDBAND;
1588 				break;
1589 			}
1590 
1591 			if (nseldebug) {
1592 				kprintf("poll index %ju/%d fd %d "
1593 				    "revents %08x\n", (uintmax_t)pi, pkap->nfds,
1594 				    pfd->fd, pfd->revents);
1595 			}
1596 
1597 			if (count_res && pfd->revents)
1598 				++*res;
1599 		}
1600 
1601 		/*
1602 		 * We must deregister any kqueue poll event that does not
1603 		 * set poll return bits to prevent a live-lock.
1604 		 */
1605 		if (pfd->revents == 0) {
1606 			krateprintf(&krate_poll,
1607 				"poll index %ju no-action %ju/%d "
1608 				"events=%08x kevpfilt=%d/%08x\n",
1609 				(uintmax_t)pi, (uintmax_t)kevp[i].ident,
1610 				pfd->fd, pfd->events,
1611 				kevp[i].filter, kevp[i].flags);
1612 			goto deregister;
1613 		}
1614 	}
1615 
1616 	return (0);
1617 }
1618 
1619 static int
dopoll(int nfds,struct pollfd * fds,struct timespec * ts,int * res,int flags)1620 dopoll(int nfds, struct pollfd *fds, struct timespec *ts, int *res, int flags)
1621 {
1622 	struct poll_kevent_copyin_args ka;
1623 	struct pollfd sfds[64];
1624 	int bytes;
1625 	int error;
1626 
1627 	flags |= KEVENT_AUTO_STALE | KEVENT_UNIQUE_NOTES;
1628 
1629         *res = 0;
1630         if (nfds < 0)
1631                 return (EINVAL);
1632 
1633 	if (nfds == 0 && ts)
1634 		return (dotimeout_only(ts));
1635 
1636 	/*
1637 	 * This is a bit arbitrary but we need to limit internal kmallocs.
1638 	 */
1639         if (nfds > maxfilesperproc * 2)
1640                 nfds = maxfilesperproc * 2;
1641 	bytes = sizeof(struct pollfd) * nfds;
1642 
1643 	ka.lwp = curthread->td_lwp;
1644 	ka.nfds = nfds;
1645 	ka.pfds = 0;
1646 	ka.error = 0;
1647 
1648 	if (ka.nfds < 64)
1649 		ka.fds = sfds;
1650 	else
1651 		ka.fds = kmalloc(bytes, M_SELECT, M_WAITOK);
1652 
1653 	error = copyin(fds, ka.fds, bytes);
1654 
1655 	if (error == 0)
1656 		error = kern_kevent(&ka.lwp->lwp_kqueue, 0x7FFFFFFF, res, &ka,
1657 				    poll_copyin, poll_copyout, ts, flags);
1658 
1659 	if (error == 0)
1660 		error = copyout(ka.fds, fds, bytes);
1661 
1662 	if (ka.fds != sfds)
1663 		kfree(ka.fds, M_SELECT);
1664 
1665 	ka.lwp->lwp_kqueue_serial += nfds;
1666 
1667 	return (error);
1668 }
1669 
1670 static int
socket_wait_copyin(void * arg,struct kevent * kevp,int maxevents,int * events)1671 socket_wait_copyin(void *arg, struct kevent *kevp, int maxevents, int *events)
1672 {
1673 	return (0);
1674 }
1675 
1676 static int
socket_wait_copyout(void * arg,struct kevent * kevp,int count,int * res)1677 socket_wait_copyout(void *arg, struct kevent *kevp, int count, int *res)
1678 {
1679 	++*res;
1680 	return (0);
1681 }
1682 
1683 extern	struct fileops socketops;
1684 
1685 /*
1686  * NOTE: Callers of socket_wait() must already have a reference on the
1687  *	 socket.
1688  */
1689 int
socket_wait(struct socket * so,struct timespec * ts,int * res)1690 socket_wait(struct socket *so, struct timespec *ts, int *res)
1691 {
1692 	struct thread *td = curthread;
1693 	struct file *fp;
1694 	struct kqueue kq;
1695 	struct kevent kev;
1696 	int error, fd;
1697 	int n;
1698 
1699 	if ((error = falloc(td->td_lwp, &fp, &fd)) != 0)
1700 		return (error);
1701 
1702 	fp->f_type = DTYPE_SOCKET;
1703 	fp->f_flag = FREAD | FWRITE;
1704 	fp->f_ops = &socketops;
1705 	fp->f_data = so;
1706 	fsetfd(td->td_lwp->lwp_proc->p_fd, fp, fd);
1707 	fsetfdflags(td->td_proc->p_fd, fd, UF_EXCLOSE);
1708 
1709 	bzero(&kq, sizeof(kq));
1710 	kqueue_init(&kq, td->td_lwp->lwp_proc->p_fd);
1711 	EV_SET(&kev, fd, EVFILT_READ, EV_ADD|EV_ENABLE, 0, 0, NULL);
1712 	n = 1;
1713 	if ((error = kqueue_register(&kq, &kev, &n, 0)) != 0) {
1714 		fdrop(fp);
1715 		return (error);
1716 	}
1717 
1718 	error = kern_kevent(&kq, 1, res, NULL, socket_wait_copyin,
1719 			    socket_wait_copyout, ts, 0);
1720 
1721 	EV_SET(&kev, fd, EVFILT_READ, EV_DELETE|EV_DISABLE, 0, 0, NULL);
1722 	n = 1;
1723 	kqueue_register(&kq, &kev, &n, 0);
1724 	fp->f_ops = &badfileops;
1725 	fdrop(fp);
1726 
1727 	return (error);
1728 }
1729 
1730 /*
1731  * OpenBSD poll system call.
1732  * XXX this isn't quite a true representation..  OpenBSD uses select ops.
1733  *
1734  * MPSAFE
1735  */
1736 int
sys_openbsd_poll(struct sysmsg * sysmsg,const struct openbsd_poll_args * uap)1737 sys_openbsd_poll(struct sysmsg *sysmsg, const struct openbsd_poll_args *uap)
1738 {
1739 	return (sys_poll(sysmsg, (const struct poll_args *)uap));
1740 }
1741 
1742 /*ARGSUSED*/
1743 int
seltrue(cdev_t dev,int events)1744 seltrue(cdev_t dev, int events)
1745 {
1746 	return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
1747 }
1748