xref: /dflybsd-src/sys/kern/kern_subr.c (revision 558a398b19696474173810daa238fb5df32df27b)
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  * $DragonFly: src/sys/kern/kern_subr.c,v 1.25 2006/12/23 23:47:54 swildner Exp $
41  */
42 
43 #include "opt_ddb.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
48 #include <sys/proc.h>
49 #include <sys/malloc.h>
50 #include <sys/lock.h>
51 #include <sys/resourcevar.h>
52 #include <sys/sysctl.h>
53 #include <sys/uio.h>
54 #include <sys/vnode.h>
55 #include <machine/limits.h>
56 
57 #include <vm/vm.h>
58 #include <vm/vm_page.h>
59 #include <vm/vm_map.h>
60 
61 SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, NULL, UIO_MAXIOV,
62 	"Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)");
63 
64 /*
65  * UIO_READ:	copy the kernelspace cp to the user or kernelspace UIO
66  * UIO_WRITE:	copy the user or kernelspace UIO to the kernelspace cp
67  *
68  * For userspace UIO's, uio_td must be the current thread.
69  */
70 int
71 uiomove(caddr_t cp, int n, struct uio *uio)
72 {
73 	struct iovec *iov;
74 	u_int cnt;
75 	int error = 0;
76 	int save = 0;
77 	int baseticks = ticks;
78 
79 	KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
80 	    ("uiomove: mode"));
81 	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
82 	    ("uiomove proc"));
83 
84 	if (curproc) {
85 		save = curproc->p_flag & P_DEADLKTREAT;
86 		curproc->p_flag |= P_DEADLKTREAT;
87 	}
88 
89 	while (n > 0 && uio->uio_resid) {
90 		iov = uio->uio_iov;
91 		cnt = iov->iov_len;
92 		if (cnt == 0) {
93 			uio->uio_iov++;
94 			uio->uio_iovcnt--;
95 			continue;
96 		}
97 		if (cnt > n)
98 			cnt = n;
99 
100 		switch (uio->uio_segflg) {
101 
102 		case UIO_USERSPACE:
103 			if (ticks - baseticks >= hogticks) {
104 				uio_yield();
105 				baseticks = ticks;
106 			}
107 			if (uio->uio_rw == UIO_READ)
108 				error = copyout(cp, iov->iov_base, cnt);
109 			else
110 				error = copyin(iov->iov_base, cp, cnt);
111 			if (error)
112 				break;
113 			break;
114 
115 		case UIO_SYSSPACE:
116 			if (uio->uio_rw == UIO_READ)
117 				bcopy((caddr_t)cp, iov->iov_base, cnt);
118 			else
119 				bcopy(iov->iov_base, (caddr_t)cp, cnt);
120 			break;
121 		case UIO_NOCOPY:
122 			break;
123 		}
124 		iov->iov_base += cnt;
125 		iov->iov_len -= cnt;
126 		uio->uio_resid -= cnt;
127 		uio->uio_offset += cnt;
128 		cp += cnt;
129 		n -= cnt;
130 	}
131 	if (curproc)
132 		curproc->p_flag = (curproc->p_flag & ~P_DEADLKTREAT) | save;
133 	return (error);
134 }
135 /*
136  * Wrapper for uiomove() that validates the arguments against a known-good
137  * kernel buffer.  Currently, uiomove accepts a signed (n) argument, which
138  * is almost definitely a bad thing, so we catch that here as well.  We
139  * return a runtime failure, but it might be desirable to generate a runtime
140  * assertion failure instead.
141  */
142 int
143 uiomove_frombuf(void *buf, int buflen, struct uio *uio)
144 {
145 	unsigned int offset, n;
146 
147 	if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
148 	    (offset = uio->uio_offset) != uio->uio_offset)
149 		return (EINVAL);
150 	if (buflen <= 0 || offset >= buflen)
151 		return (0);
152 	if ((n = buflen - offset) > INT_MAX)
153 		return (EINVAL);
154 	return (uiomove((char *)buf + offset, n, uio));
155 }
156 
157 
158 int
159 uiomoveco(caddr_t cp, int n, struct uio *uio, struct vm_object *obj)
160 {
161 	struct iovec *iov;
162 	u_int cnt;
163 	int error;
164 	int baseticks = ticks;
165 
166 	KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
167 	    ("uiomoveco: mode"));
168 	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
169 	    ("uiomoveco proc"));
170 
171 	while (n > 0 && uio->uio_resid) {
172 		iov = uio->uio_iov;
173 		cnt = iov->iov_len;
174 		if (cnt == 0) {
175 			uio->uio_iov++;
176 			uio->uio_iovcnt--;
177 			continue;
178 		}
179 		if (cnt > n)
180 			cnt = n;
181 
182 		switch (uio->uio_segflg) {
183 
184 		case UIO_USERSPACE:
185 			if (ticks - baseticks >= hogticks) {
186 				uio_yield();
187 				baseticks = ticks;
188 			}
189 			if (uio->uio_rw == UIO_READ) {
190 				error = copyout(cp, iov->iov_base, cnt);
191 			} else {
192 				error = copyin(iov->iov_base, cp, cnt);
193 			}
194 			if (error)
195 				return (error);
196 			break;
197 
198 		case UIO_SYSSPACE:
199 			if (uio->uio_rw == UIO_READ)
200 				bcopy((caddr_t)cp, iov->iov_base, cnt);
201 			else
202 				bcopy(iov->iov_base, (caddr_t)cp, cnt);
203 			break;
204 		case UIO_NOCOPY:
205 			break;
206 		}
207 		iov->iov_base += cnt;
208 		iov->iov_len -= cnt;
209 		uio->uio_resid -= cnt;
210 		uio->uio_offset += cnt;
211 		cp += cnt;
212 		n -= cnt;
213 	}
214 	return (0);
215 }
216 
217 /*
218  * Give next character to user as result of read.
219  */
220 int
221 ureadc(int c, struct uio *uio)
222 {
223 	struct iovec *iov;
224 
225 again:
226 	if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
227 		panic("ureadc");
228 	iov = uio->uio_iov;
229 	if (iov->iov_len == 0) {
230 		uio->uio_iovcnt--;
231 		uio->uio_iov++;
232 		goto again;
233 	}
234 	switch (uio->uio_segflg) {
235 
236 	case UIO_USERSPACE:
237 		if (subyte(iov->iov_base, c) < 0)
238 			return (EFAULT);
239 		break;
240 
241 	case UIO_SYSSPACE:
242 		*iov->iov_base = c;
243 		break;
244 
245 	case UIO_NOCOPY:
246 		break;
247 	}
248 	iov->iov_base++;
249 	iov->iov_len--;
250 	uio->uio_resid--;
251 	uio->uio_offset++;
252 	return (0);
253 }
254 
255 /*
256  * General routine to allocate a hash table.  Make the hash table size a
257  * power of 2 greater or equal to the number of elements requested, and
258  * store the masking value in *hashmask.
259  */
260 void *
261 hashinit(int elements, struct malloc_type *type, u_long *hashmask)
262 {
263 	long hashsize;
264 	LIST_HEAD(generic, generic) *hashtbl;
265 	int i;
266 
267 	if (elements <= 0)
268 		panic("hashinit: bad elements");
269 	for (hashsize = 2; hashsize < elements; hashsize <<= 1)
270 		continue;
271 	hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
272 	for (i = 0; i < hashsize; i++)
273 		LIST_INIT(&hashtbl[i]);
274 	*hashmask = hashsize - 1;
275 	return (hashtbl);
276 }
277 
278 static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
279 			2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
280 			7159, 7673, 8191, 12281, 16381, 24571, 32749 };
281 #define NPRIMES (sizeof(primes) / sizeof(primes[0]))
282 
283 /*
284  * General routine to allocate a prime number sized hash table.
285  */
286 void *
287 phashinit(int elements, struct malloc_type *type, u_long *nentries)
288 {
289 	long hashsize;
290 	LIST_HEAD(generic, generic) *hashtbl;
291 	int i;
292 
293 	if (elements <= 0)
294 		panic("phashinit: bad elements");
295 	for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
296 		i++;
297 		if (i == NPRIMES)
298 			break;
299 		hashsize = primes[i];
300 	}
301 	hashsize = primes[i - 1];
302 	hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
303 	for (i = 0; i < hashsize; i++)
304 		LIST_INIT(&hashtbl[i]);
305 	*nentries = hashsize;
306 	return (hashtbl);
307 }
308 
309 /*
310  * Copyin an iovec.  If the iovec array fits, use the preallocated small
311  * iovec structure.  If it is too big, dynamically allocate an iovec array
312  * of sufficient size.
313  *
314  * MPSAFE
315  */
316 int
317 iovec_copyin(struct iovec *uiov, struct iovec **kiov, struct iovec *siov,
318 	     size_t iov_cnt, int *iov_len)
319 {
320 	struct iovec *iovp;
321 	int error, i;
322 
323 	if (iov_cnt >= UIO_MAXIOV)
324 		return EMSGSIZE;
325 	if (iov_cnt >= UIO_SMALLIOV) {
326 		MALLOC(*kiov, struct iovec *, sizeof(struct iovec) * iov_cnt,
327 		    M_IOV, M_WAITOK);
328 	} else {
329 		*kiov = siov;
330 	}
331 	error = copyin(uiov, *kiov, iov_cnt * sizeof(struct iovec));
332 	if (error == 0) {
333 		*iov_len = 0;
334 		for (i = 0, iovp = *kiov; i < iov_cnt; i++, iovp++) {
335 			/*
336 			 * Check for both *iov_len overflows and out of
337 			 * range iovp->iov_len's.  We limit to the
338 			 * capabilities of signed integers.
339 			 */
340 			if (*iov_len + (int)iovp->iov_len < *iov_len)
341 				error = EINVAL;
342 			*iov_len += (int)iovp->iov_len;
343 		}
344 	}
345 	if (error)
346 		iovec_free(kiov, siov);
347 	return (error);
348 }
349