xref: /dflybsd-src/sys/kern/kern_subr.c (revision dae741e33c840b92a8a53bf9f01157ede145e256)
1 /*
2  * Copyright (c) 1982, 1986, 1991, 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. 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.
25  *
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
36  * SUCH DAMAGE.
37  *
38  *	@(#)kern_subr.c	8.3 (Berkeley) 1/21/94
39  * $FreeBSD: src/sys/kern/kern_subr.c,v 1.31.2.2 2002/04/21 08:09:37 bde Exp $
40  */
41 
42 #include "opt_ddb.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/proc.h>
48 #include <sys/malloc.h>
49 #include <sys/lock.h>
50 #include <sys/resourcevar.h>
51 #include <sys/sysctl.h>
52 #include <sys/uio.h>
53 #include <sys/vnode.h>
54 #include <sys/thread2.h>
55 #include <machine/limits.h>
56 
57 #include <cpu/lwbuf.h>
58 
59 #include <vm/vm.h>
60 #include <vm/vm_page.h>
61 #include <vm/vm_map.h>
62 
63 SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, NULL, UIO_MAXIOV,
64 	"Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)");
65 
66 /*
67  * UIO_READ:	copy the kernelspace cp to the user or kernelspace UIO
68  * UIO_WRITE:	copy the user or kernelspace UIO to the kernelspace cp
69  *
70  * For userspace UIO's, uio_td must be the current thread.
71  *
72  * The syscall interface is responsible for limiting the length to
73  * ssize_t for things like read() or write() which return the bytes
74  * read or written as ssize_t.  These functions work with unsigned
75  * lengths.
76  */
77 int
78 uiomove(caddr_t cp, size_t n, struct uio *uio)
79 {
80 	thread_t td = curthread;
81 	struct iovec *iov;
82 	size_t cnt;
83 	int error = 0;
84 	int save = 0;
85 
86 	KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
87 	    ("uiomove: mode"));
88 	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == td,
89 	    ("uiomove proc"));
90 
91 	crit_enter();
92 	save = td->td_flags & TDF_DEADLKTREAT;
93 	td->td_flags |= TDF_DEADLKTREAT;
94 	crit_exit();
95 
96 	while (n > 0 && uio->uio_resid) {
97 		iov = uio->uio_iov;
98 		cnt = iov->iov_len;
99 		if (cnt == 0) {
100 			uio->uio_iov++;
101 			uio->uio_iovcnt--;
102 			continue;
103 		}
104 		if (cnt > n)
105 			cnt = n;
106 
107 		switch (uio->uio_segflg) {
108 
109 		case UIO_USERSPACE:
110 			lwkt_user_yield();
111 			if (uio->uio_rw == UIO_READ)
112 				error = copyout(cp, iov->iov_base, cnt);
113 			else
114 				error = copyin(iov->iov_base, cp, cnt);
115 			if (error)
116 				break;
117 			break;
118 
119 		case UIO_SYSSPACE:
120 			if (uio->uio_rw == UIO_READ)
121 				bcopy((caddr_t)cp, iov->iov_base, cnt);
122 			else
123 				bcopy(iov->iov_base, (caddr_t)cp, cnt);
124 			break;
125 		case UIO_NOCOPY:
126 			break;
127 		}
128 		iov->iov_base = (char *)iov->iov_base + cnt;
129 		iov->iov_len -= cnt;
130 		uio->uio_resid -= cnt;
131 		uio->uio_offset += cnt;
132 		cp += cnt;
133 		n -= cnt;
134 	}
135 	crit_enter();
136 	td->td_flags = (td->td_flags & ~TDF_DEADLKTREAT) | save;
137 	crit_exit();
138 	return (error);
139 }
140 
141 /*
142  * Like uiomove() but copies zero-fill.  Only allowed for UIO_READ,
143  * for obvious reasons.
144  */
145 int
146 uiomovez(size_t n, struct uio *uio)
147 {
148 	struct iovec *iov;
149 	size_t cnt;
150 	int error = 0;
151 
152 	KASSERT(uio->uio_rw == UIO_READ, ("uiomovez: mode"));
153 	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
154 		("uiomove proc"));
155 
156 	while (n > 0 && uio->uio_resid) {
157 		iov = uio->uio_iov;
158 		cnt = iov->iov_len;
159 		if (cnt == 0) {
160 			uio->uio_iov++;
161 			uio->uio_iovcnt--;
162 			continue;
163 		}
164 		if (cnt > n)
165 			cnt = n;
166 
167 		switch (uio->uio_segflg) {
168 		case UIO_USERSPACE:
169 			error = copyout(ZeroPage, iov->iov_base, cnt);
170 			if (error)
171 				break;
172 			break;
173 		case UIO_SYSSPACE:
174 			bzero(iov->iov_base, cnt);
175 			break;
176 		case UIO_NOCOPY:
177 			break;
178 		}
179 		iov->iov_base = (char *)iov->iov_base + cnt;
180 		iov->iov_len -= cnt;
181 		uio->uio_resid -= cnt;
182 		uio->uio_offset += cnt;
183 		n -= cnt;
184 	}
185 	return (error);
186 }
187 
188 /*
189  * Wrapper for uiomove() that validates the arguments against a known-good
190  * kernel buffer.
191  */
192 int
193 uiomove_frombuf(void *buf, size_t buflen, struct uio *uio)
194 {
195 	size_t offset;
196 
197 	offset = (size_t)uio->uio_offset;
198 	if ((off_t)offset != uio->uio_offset)
199 		return (EINVAL);
200 	if (buflen == 0 || offset >= buflen)
201 		return (0);
202 	return (uiomove((char *)buf + offset, buflen - offset, uio));
203 }
204 
205 /*
206  * Give next character to user as result of read.
207  */
208 int
209 ureadc(int c, struct uio *uio)
210 {
211 	struct iovec *iov;
212 	char *iov_base;
213 
214 again:
215 	if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
216 		panic("ureadc");
217 	iov = uio->uio_iov;
218 	if (iov->iov_len == 0) {
219 		uio->uio_iovcnt--;
220 		uio->uio_iov++;
221 		goto again;
222 	}
223 	switch (uio->uio_segflg) {
224 
225 	case UIO_USERSPACE:
226 		if (subyte(iov->iov_base, c) < 0)
227 			return (EFAULT);
228 		break;
229 
230 	case UIO_SYSSPACE:
231 		iov_base = iov->iov_base;
232 		*iov_base = c;
233 		iov->iov_base = iov_base;
234 		break;
235 
236 	case UIO_NOCOPY:
237 		break;
238 	}
239 	iov->iov_base = (char *)iov->iov_base + 1;
240 	iov->iov_len--;
241 	uio->uio_resid--;
242 	uio->uio_offset++;
243 	return (0);
244 }
245 
246 /*
247  * General routine to allocate a hash table.  Make the hash table size a
248  * power of 2 greater or equal to the number of elements requested, and
249  * store the masking value in *hashmask.
250  */
251 void *
252 hashinit(int elements, struct malloc_type *type, u_long *hashmask)
253 {
254 	long hashsize;
255 	LIST_HEAD(generic, generic) *hashtbl;
256 	int i;
257 
258 	if (elements <= 0)
259 		panic("hashinit: bad elements");
260 	for (hashsize = 2; hashsize < elements; hashsize <<= 1)
261 		continue;
262 	hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
263 	for (i = 0; i < hashsize; i++)
264 		LIST_INIT(&hashtbl[i]);
265 	*hashmask = hashsize - 1;
266 	return (hashtbl);
267 }
268 
269 /*
270  * This is a newer version which allocates a hash table of structures.
271  *
272  * The returned array will be zero'd.  The caller is responsible for
273  * initializing the structures.
274  */
275 void *
276 hashinit_ext(int elements, size_t size, struct malloc_type *type,
277 	     u_long *hashmask)
278 {
279 	long hashsize;
280 	void *hashtbl;
281 
282 	if (elements <= 0)
283 		panic("hashinit: bad elements");
284 	for (hashsize = 2; hashsize < elements; hashsize <<= 1)
285 		continue;
286 	hashtbl = kmalloc((size_t)hashsize * size, type, M_WAITOK | M_ZERO);
287 	*hashmask = hashsize - 1;
288 	return (hashtbl);
289 }
290 
291 static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
292 			2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
293 			7159, 7673, 8191, 12281, 16381, 24571, 32749 };
294 #define NPRIMES NELEM(primes)
295 
296 /*
297  * General routine to allocate a prime number sized hash table.
298  */
299 void *
300 phashinit(int elements, struct malloc_type *type, u_long *nentries)
301 {
302 	long hashsize;
303 	LIST_HEAD(generic, generic) *hashtbl;
304 	int i;
305 
306 	if (elements <= 0)
307 		panic("phashinit: bad elements");
308 	for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
309 		i++;
310 		if (i == NPRIMES)
311 			break;
312 		hashsize = primes[i];
313 	}
314 	hashsize = primes[i - 1];
315 	hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
316 	for (i = 0; i < hashsize; i++)
317 		LIST_INIT(&hashtbl[i]);
318 	*nentries = hashsize;
319 	return (hashtbl);
320 }
321 
322 /*
323  * This is a newer version which allocates a hash table of structures
324  * in a prime-number size.
325  *
326  * The returned array will be zero'd.  The caller is responsible for
327  * initializing the structures.
328  */
329 void *
330 phashinit_ext(int elements, size_t size, struct malloc_type *type,
331 	      u_long *nentries)
332 {
333 	long hashsize;
334 	void *hashtbl;
335 	int i;
336 
337 	if (elements <= 0)
338 		panic("phashinit: bad elements");
339 	for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
340 		i++;
341 		if (i == NPRIMES)
342 			break;
343 		hashsize = primes[i];
344 	}
345 	hashsize = primes[i - 1];
346 	hashtbl = kmalloc((size_t)hashsize * size, type, M_WAITOK | M_ZERO);
347 	*nentries = hashsize;
348 	return (hashtbl);
349 }
350 
351 /*
352  * Copyin an iovec.  If the iovec array fits, use the preallocated small
353  * iovec structure.  If it is too big, dynamically allocate an iovec array
354  * of sufficient size.
355  *
356  * MPSAFE
357  */
358 int
359 iovec_copyin(struct iovec *uiov, struct iovec **kiov, struct iovec *siov,
360 	     size_t iov_cnt, size_t *iov_len)
361 {
362 	struct iovec *iovp;
363 	int error, i;
364 	size_t len;
365 
366 	if (iov_cnt > UIO_MAXIOV)
367 		return EMSGSIZE;
368 	if (iov_cnt > UIO_SMALLIOV) {
369 		MALLOC(*kiov, struct iovec *, sizeof(struct iovec) * iov_cnt,
370 		    M_IOV, M_WAITOK);
371 	} else {
372 		*kiov = siov;
373 	}
374 	error = copyin(uiov, *kiov, iov_cnt * sizeof(struct iovec));
375 	if (error == 0) {
376 		*iov_len = 0;
377 		for (i = 0, iovp = *kiov; i < iov_cnt; i++, iovp++) {
378 			/*
379 			 * Check for both *iov_len overflows and out of
380 			 * range iovp->iov_len's.  We limit to the
381 			 * capabilities of signed integers.
382 			 *
383 			 * GCC4 - overflow check opt requires assign/test.
384 			 */
385 			len = *iov_len + iovp->iov_len;
386 			if (len < *iov_len)
387 				error = EINVAL;
388 			*iov_len = len;
389 		}
390 	}
391 
392 	/*
393 	 * From userland disallow iovec's which exceed the sized size
394 	 * limit as the system calls return ssize_t.
395 	 *
396 	 * NOTE: Internal kernel interfaces can handle the unsigned
397 	 *	 limit.
398 	 */
399 	if (error == 0 && (ssize_t)*iov_len < 0)
400 		error = EINVAL;
401 
402 	if (error)
403 		iovec_free(kiov, siov);
404 	return (error);
405 }
406 
407 
408 /*
409  * Copyright (c) 2004 Alan L. Cox <alc@cs.rice.edu>
410  * Copyright (c) 1982, 1986, 1991, 1993
411  *	The Regents of the University of California.  All rights reserved.
412  * (c) UNIX System Laboratories, Inc.
413  * All or some portions of this file are derived from material licensed
414  * to the University of California by American Telephone and Telegraph
415  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
416  * the permission of UNIX System Laboratories, Inc.
417  *
418  * Redistribution and use in source and binary forms, with or without
419  * modification, are permitted provided that the following conditions
420  * are met:
421  * 1. Redistributions of source code must retain the above copyright
422  *    notice, this list of conditions and the following disclaimer.
423  * 2. Redistributions in binary form must reproduce the above copyright
424  *    notice, this list of conditions and the following disclaimer in the
425  *    documentation and/or other materials provided with the distribution.
426  * 4. Neither the name of the University nor the names of its contributors
427  *    may be used to endorse or promote products derived from this software
428  *    without specific prior written permission.
429  *
430  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
431  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
432  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
433  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
434  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
435  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
436  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
437  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
438  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
439  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
440  * SUCH DAMAGE.
441  *
442  * @(#)kern_subr.c	8.3 (Berkeley) 1/21/94
443  * $FreeBSD: src/sys/i386/i386/uio_machdep.c,v 1.1 2004/03/21 20:28:36 alc Exp $
444  */
445 
446 /*
447  * Implement uiomove(9) from physical memory using lwbuf's to reduce
448  * the creation and destruction of ephemeral mappings.
449  */
450 int
451 uiomove_fromphys(vm_page_t *ma, vm_offset_t offset, size_t n, struct uio *uio)
452 {
453 	struct lwbuf lwb_cache;
454 	struct lwbuf *lwb;
455 	struct thread *td = curthread;
456 	struct iovec *iov;
457 	void *cp;
458 	vm_offset_t page_offset;
459 	vm_page_t m;
460 	size_t cnt;
461 	int error = 0;
462 	int save = 0;
463 
464 	KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
465 	    ("uiomove_fromphys: mode"));
466 	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
467 	    ("uiomove_fromphys proc"));
468 
469 	crit_enter();
470 	save = td->td_flags & TDF_DEADLKTREAT;
471 	td->td_flags |= TDF_DEADLKTREAT;
472 	crit_exit();
473 
474 	while (n > 0 && uio->uio_resid) {
475 		iov = uio->uio_iov;
476 		cnt = iov->iov_len;
477 		if (cnt == 0) {
478 			uio->uio_iov++;
479 			uio->uio_iovcnt--;
480 			continue;
481 		}
482 		if (cnt > n)
483 			cnt = n;
484 		page_offset = offset & PAGE_MASK;
485 		cnt = min(cnt, PAGE_SIZE - page_offset);
486 		m = ma[offset >> PAGE_SHIFT];
487 		lwb = lwbuf_alloc(m, &lwb_cache);
488 		cp = (char *)lwbuf_kva(lwb) + page_offset;
489 		switch (uio->uio_segflg) {
490 		case UIO_USERSPACE:
491 			/*
492 			 * note: removed uioyield (it was the wrong place to
493 			 * put it).
494 			 */
495 			if (uio->uio_rw == UIO_READ)
496 				error = copyout(cp, iov->iov_base, cnt);
497 			else
498 				error = copyin(iov->iov_base, cp, cnt);
499 			if (error) {
500 				lwbuf_free(lwb);
501 				goto out;
502 			}
503 			break;
504 		case UIO_SYSSPACE:
505 			if (uio->uio_rw == UIO_READ)
506 				bcopy(cp, iov->iov_base, cnt);
507 			else
508 				bcopy(iov->iov_base, cp, cnt);
509 			break;
510 		case UIO_NOCOPY:
511 			break;
512 		}
513 		lwbuf_free(lwb);
514 		iov->iov_base = (char *)iov->iov_base + cnt;
515 		iov->iov_len -= cnt;
516 		uio->uio_resid -= cnt;
517 		uio->uio_offset += cnt;
518 		offset += cnt;
519 		n -= cnt;
520 	}
521 out:
522 	if (save == 0) {
523 		crit_enter();
524 		td->td_flags &= ~TDF_DEADLKTREAT;
525 		crit_exit();
526 	}
527 	return (error);
528 }
529 
530