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 * 4. 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 * @(#)kern_subr.c 8.3 (Berkeley) 1/21/94
35 */
36
37#include <sys/cdefs.h>
38__FBSDID("$FreeBSD: head/sys/kern/kern_subr.c 127911 2004-04-05 21:03:37Z imp $");
39
40#include "opt_zero.h"
41
42#include <sys/param.h>
43#include <sys/systm.h>
44#include <sys/kernel.h>
45#include <sys/ktr.h>
46#include <sys/limits.h>
47#include <sys/lock.h>
48#include <sys/mutex.h>
49#include <sys/proc.h>
50#include <sys/malloc.h>
51#include <sys/resourcevar.h>
52#include <sys/sched.h>
53#include <sys/sysctl.h>
54#include <sys/vnode.h>
55
56#include <vm/vm.h>
57#include <vm/vm_page.h>
58#include <vm/vm_map.h>
59#ifdef ZERO_COPY_SOCKETS
60#include <vm/vm_param.h>
61#include <vm/vm_object.h>
62#endif
63
64SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, NULL, UIO_MAXIOV,
65 "Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)");
66
67#ifdef ZERO_COPY_SOCKETS
68/* Declared in uipc_socket.c */
69extern int so_zero_copy_receive;
70
71static int
72vm_pgmoveco(vm_map_t mapa, vm_object_t srcobj, vm_offset_t kaddr,
73 vm_offset_t uaddr)
74{
75 vm_map_t map = mapa;
76 vm_page_t kern_pg, user_pg;
77 vm_object_t uobject;
78 vm_map_entry_t entry;
79 vm_pindex_t upindex, kpindex;
80 vm_prot_t prot;
81 boolean_t wired;
82
83 /*
84 * First lookup the kernel page.
85 */
86 kern_pg = PHYS_TO_VM_PAGE(vtophys(kaddr));
87 /*
88 * XXX The vm object containing kern_pg needs locking.
89 */
90 if ((vm_map_lookup(&map, uaddr,
91 VM_PROT_WRITE, &entry, &uobject,
92 &upindex, &prot, &wired)) != KERN_SUCCESS) {
93 return(EFAULT);
94 }
95 VM_OBJECT_LOCK(uobject);
96 if ((user_pg = vm_page_lookup(uobject, upindex)) != NULL) {
97 do
98 vm_page_lock_queues();
99 while (vm_page_sleep_if_busy(user_pg, 1, "vm_pgmoveco"));
100 vm_page_busy(user_pg);
101 pmap_remove_all(user_pg);
102 vm_page_free(user_pg);
103 } else
104 vm_page_lock_queues();
105 if (kern_pg->busy || ((kern_pg->queue - kern_pg->pc) == PQ_FREE) ||
106 (kern_pg->hold_count != 0)|| (kern_pg->flags & PG_BUSY)) {
107 printf("vm_pgmoveco: pindex(%lu), busy(%d), PG_BUSY(%d), "
108 "hold(%d) paddr(0x%lx)\n", (u_long)kern_pg->pindex,
109 kern_pg->busy, (kern_pg->flags & PG_BUSY) ? 1 : 0,
110 kern_pg->hold_count, (u_long)kern_pg->phys_addr);
111 if ((kern_pg->queue - kern_pg->pc) == PQ_FREE)
112 panic("vm_pgmoveco: renaming free page");
113 else
114 panic("vm_pgmoveco: renaming busy page");
115 }
116 kpindex = kern_pg->pindex;
117 vm_page_busy(kern_pg);
118 vm_page_rename(kern_pg, uobject, upindex);
119 vm_page_flag_clear(kern_pg, PG_BUSY);
120 kern_pg->valid = VM_PAGE_BITS_ALL;
121 vm_page_unlock_queues();
122 VM_OBJECT_UNLOCK(uobject);
123 vm_map_lookup_done(map, entry);
124 return(KERN_SUCCESS);
125}
126#endif /* ZERO_COPY_SOCKETS */
127
128int
129uiomove(void *cp, int n, struct uio *uio)
130{
131 struct thread *td = curthread;
132 struct iovec *iov;
133 u_int cnt;
134 int error = 0;
135 int save = 0;
136
137 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
138 ("uiomove: mode"));
139 KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
140 ("uiomove proc"));
141
142 if (td) {
143 mtx_lock_spin(&sched_lock);
144 save = td->td_flags & TDF_DEADLKTREAT;
145 td->td_flags |= TDF_DEADLKTREAT;
146 mtx_unlock_spin(&sched_lock);
147 }
148
149 while (n > 0 && uio->uio_resid) {
150 iov = uio->uio_iov;
151 cnt = iov->iov_len;
152 if (cnt == 0) {
153 uio->uio_iov++;
154 uio->uio_iovcnt--;
155 continue;
156 }
157 if (cnt > n)
158 cnt = n;
159
160 switch (uio->uio_segflg) {
161
162 case UIO_USERSPACE:
163 if (ticks - PCPU_GET(switchticks) >= hogticks)
164 uio_yield();
165 if (uio->uio_rw == UIO_READ)
166 error = copyout(cp, iov->iov_base, cnt);
167 else
168 error = copyin(iov->iov_base, cp, cnt);
169 if (error)
170 goto out;
171 break;
172
173 case UIO_SYSSPACE:
174 if (uio->uio_rw == UIO_READ)
175 bcopy(cp, iov->iov_base, cnt);
176 else
177 bcopy(iov->iov_base, cp, cnt);
178 break;
179 case UIO_NOCOPY:
180 break;
181 }
182 iov->iov_base = (char *)iov->iov_base + cnt;
183 iov->iov_len -= cnt;
184 uio->uio_resid -= cnt;
185 uio->uio_offset += cnt;
186 cp = (char *)cp + cnt;
187 n -= cnt;
188 }
189out:
190 if (td && save == 0) {
191 mtx_lock_spin(&sched_lock);
192 td->td_flags &= ~TDF_DEADLKTREAT;
193 mtx_unlock_spin(&sched_lock);
194 }
195 return (error);
196}
197
198/*
199 * Wrapper for uiomove() that validates the arguments against a known-good
200 * kernel buffer. Currently, uiomove accepts a signed (n) argument, which
201 * is almost definitely a bad thing, so we catch that here as well. We
202 * return a runtime failure, but it might be desirable to generate a runtime
203 * assertion failure instead.
204 */
205int
206uiomove_frombuf(void *buf, int buflen, struct uio *uio)
207{
208 unsigned int offset, n;
209
210 if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
211 (offset = uio->uio_offset) != uio->uio_offset)
212 return (EINVAL);
213 if (buflen <= 0 || offset >= buflen)
214 return (0);
215 if ((n = buflen - offset) > INT_MAX)
216 return (EINVAL);
217 return (uiomove((char *)buf + offset, n, uio));
218}
219
220#ifdef ZERO_COPY_SOCKETS
221/*
222 * Experimental support for zero-copy I/O
223 */
224static int
225userspaceco(void *cp, u_int cnt, struct uio *uio, struct vm_object *obj,
226 int disposable)
227{
228 struct iovec *iov;
229 int error;
230
231 iov = uio->uio_iov;
232 if (uio->uio_rw == UIO_READ) {
233 if ((so_zero_copy_receive != 0)
234 && (obj != NULL)
235 && ((cnt & PAGE_MASK) == 0)
236 && ((((intptr_t) iov->iov_base) & PAGE_MASK) == 0)
237 && ((uio->uio_offset & PAGE_MASK) == 0)
238 && ((((intptr_t) cp) & PAGE_MASK) == 0)
239 && (obj->type == OBJT_DEFAULT)
240 && (disposable != 0)) {
241 /* SOCKET: use page-trading */
242 /*
243 * We only want to call vm_pgmoveco() on
244 * disposeable pages, since it gives the
245 * kernel page to the userland process.
246 */
247 error = vm_pgmoveco(&curproc->p_vmspace->vm_map,
248 obj, (vm_offset_t)cp,
249 (vm_offset_t)iov->iov_base);
250
251 /*
252 * If we get an error back, attempt
253 * to use copyout() instead. The
254 * disposable page should be freed
255 * automatically if we weren't able to move
256 * it into userland.
257 */
258 if (error != 0)
259 error = copyout(cp, iov->iov_base, cnt);
260 } else {
261 error = copyout(cp, iov->iov_base, cnt);
262 }
263 } else {
264 error = copyin(iov->iov_base, cp, cnt);
265 }
266 return (error);
267}
268
269int
270uiomoveco(void *cp, int n, struct uio *uio, struct vm_object *obj,
271 int disposable)
272{
273 struct iovec *iov;
274 u_int cnt;
275 int error;
276
277 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
278 ("uiomoveco: mode"));
279 KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
280 ("uiomoveco proc"));
281
282 while (n > 0 && uio->uio_resid) {
283 iov = uio->uio_iov;
284 cnt = iov->iov_len;
285 if (cnt == 0) {
286 uio->uio_iov++;
287 uio->uio_iovcnt--;
288 continue;
289 }
290 if (cnt > n)
291 cnt = n;
292
293 switch (uio->uio_segflg) {
294
295 case UIO_USERSPACE:
296 if (ticks - PCPU_GET(switchticks) >= hogticks)
297 uio_yield();
298
299 error = userspaceco(cp, cnt, uio, obj, disposable);
300
301 if (error)
302 return (error);
303 break;
304
305 case UIO_SYSSPACE:
306 if (uio->uio_rw == UIO_READ)
307 bcopy(cp, iov->iov_base, cnt);
308 else
309 bcopy(iov->iov_base, cp, cnt);
310 break;
311 case UIO_NOCOPY:
312 break;
313 }
314 iov->iov_base = (char *)iov->iov_base + cnt;
315 iov->iov_len -= cnt;
316 uio->uio_resid -= cnt;
317 uio->uio_offset += cnt;
318 cp = (char *)cp + cnt;
319 n -= cnt;
320 }
321 return (0);
322}
323#endif /* ZERO_COPY_SOCKETS */
324
325/*
326 * Give next character to user as result of read.
327 */
328int
329ureadc(int c, struct uio *uio)
330{
331 struct iovec *iov;
332 char *iov_base;
333
334again:
335 if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
336 panic("ureadc");
337 iov = uio->uio_iov;
338 if (iov->iov_len == 0) {
339 uio->uio_iovcnt--;
340 uio->uio_iov++;
341 goto again;
342 }
343 switch (uio->uio_segflg) {
344
345 case UIO_USERSPACE:
346 if (subyte(iov->iov_base, c) < 0)
347 return (EFAULT);
348 break;
349
350 case UIO_SYSSPACE:
351 iov_base = iov->iov_base;
352 *iov_base = c;
353 iov->iov_base = iov_base;
354 break;
355
356 case UIO_NOCOPY:
357 break;
358 }
359 iov->iov_base = (char *)iov->iov_base + 1;
360 iov->iov_len--;
361 uio->uio_resid--;
362 uio->uio_offset++;
363 return (0);
364}
365
366/*
367 * General routine to allocate a hash table.
368 */
369void *
370hashinit(int elements, struct malloc_type *type, u_long *hashmask)
371{
372 long hashsize;
373 LIST_HEAD(generic, generic) *hashtbl;
374 int i;
375
376 if (elements <= 0)
377 panic("hashinit: bad elements");
378 for (hashsize = 1; hashsize <= elements; hashsize <<= 1)
379 continue;
380 hashsize >>= 1;
381 hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
382 for (i = 0; i < hashsize; i++)
383 LIST_INIT(&hashtbl[i]);
384 *hashmask = hashsize - 1;
385 return (hashtbl);
386}
387
388void
389hashdestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask)
390{
391 LIST_HEAD(generic, generic) *hashtbl, *hp;
392
393 hashtbl = vhashtbl;
394 for (hp = hashtbl; hp <= &hashtbl[hashmask]; hp++)
395 if (!LIST_EMPTY(hp))
396 panic("hashdestroy: hash not empty");
397 free(hashtbl, type);
398}
399
400static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
401 2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
402 7159, 7673, 8191, 12281, 16381, 24571, 32749 };
403#define NPRIMES (sizeof(primes) / sizeof(primes[0]))
404
405/*
406 * General routine to allocate a prime number sized hash table.
407 */
408void *
409phashinit(int elements, struct malloc_type *type, u_long *nentries)
410{
411 long hashsize;
412 LIST_HEAD(generic, generic) *hashtbl;
413 int i;
414
415 if (elements <= 0)
416 panic("phashinit: bad elements");
417 for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
418 i++;
419 if (i == NPRIMES)
420 break;
421 hashsize = primes[i];
422 }
423 hashsize = primes[i - 1];
424 hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
425 for (i = 0; i < hashsize; i++)
426 LIST_INIT(&hashtbl[i]);
427 *nentries = hashsize;
428 return (hashtbl);
429}
430
431void
432uio_yield(void)
433{
434 struct thread *td;
435
436 td = curthread;
437 mtx_lock_spin(&sched_lock);
438 DROP_GIANT();
439 sched_prio(td, td->td_ksegrp->kg_user_pri); /* XXXKSE */
440 mi_switch(SW_INVOL);
441 mtx_unlock_spin(&sched_lock);
442 PICKUP_GIANT();
443}
444
445int
446copyinfrom(const void * __restrict src, void * __restrict dst, size_t len,
447 int seg)
448{
449 int error = 0;
450
451 switch (seg) {
452 case UIO_USERSPACE:
453 error = copyin(src, dst, len);
454 break;
455 case UIO_SYSSPACE:
456 bcopy(src, dst, len);
457 break;
458 default:
459 panic("copyinfrom: bad seg %d\n", seg);
460 }
461 return (error);
462}
463
464int
465copyinstrfrom(const void * __restrict src, void * __restrict dst, size_t len,
466 size_t * __restrict copied, int seg)
467{
468 int error = 0;
469
470 switch (seg) {
471 case UIO_USERSPACE:
472 error = copyinstr(src, dst, len, copied);
473 break;
474 case UIO_SYSSPACE:
475 error = copystr(src, dst, len, copied);
476 break;
477 default:
478 panic("copyinstrfrom: bad seg %d\n", seg);
479 }
480 return (error);
481}
482
483int
484uiofromiov(struct iovec *iovp, u_int iovcnt, struct uio *uio)
485{
486 struct iovec *iov;
487 u_int iovlen;
488 int error, i;
489
490 /* note: can't use iovlen until iovcnt is validated */
491 iovlen = iovcnt * sizeof (struct iovec);
492 if (iovcnt > UIO_MAXIOV) {
493 error = EINVAL;
494 goto done;
495 }
496 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
497 uio->uio_iov = iov;
498 uio->uio_iovcnt = iovcnt;
499 uio->uio_segflg = UIO_USERSPACE;
500 uio->uio_offset = -1;
501 if ((error = copyin(iovp, iov, iovlen)))
502 goto done;
503 uio->uio_resid = 0;
504 for (i = 0; i < iovcnt; i++) {
505 if (iov->iov_len > INT_MAX - uio->uio_resid) {
506 error = EINVAL;
507 goto done;
508 }
509 uio->uio_resid += iov->iov_len;
510 iov++;
511 }
512
513done:
514 if (error && uio->uio_iov) {
515 FREE(uio->uio_iov, M_IOV);
516 uio->uio_iov = NULL;
517 }
518 return (error);
519
520}