subr_hash.c revision 116182 
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 */
40
41#include <sys/cdefs.h>
42__FBSDID("$FreeBSD: head/sys/kern/kern_subr.c 116182 2003-06-11 00:56:59Z obrien $");
43
44#include "opt_zero.h"
45
46#include <sys/param.h>
47#include <sys/systm.h>
48#include <sys/kernel.h>
49#include <sys/ktr.h>
50#include <sys/lock.h>
51#include <sys/mutex.h>
52#include <sys/proc.h>
53#include <sys/malloc.h>
54#include <sys/resourcevar.h>
55#include <sys/sched.h>
56#include <sys/sysctl.h>
57#include <sys/vnode.h>
58
59#include <vm/vm.h>
60#include <vm/vm_page.h>
61#include <vm/vm_map.h>
62#ifdef ZERO_COPY_SOCKETS
63#include <vm/vm_param.h>
64#include <vm/vm_object.h>
65#endif
66
67SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, NULL, UIO_MAXIOV,
68	"Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)");
69
70#ifdef ZERO_COPY_SOCKETS
71/* Declared in uipc_socket.c */
72extern int so_zero_copy_receive;
73
74static int
75vm_pgmoveco(vm_map_t mapa, vm_object_t srcobj, vm_offset_t kaddr,
76    vm_offset_t uaddr)
77{
78	vm_map_t map = mapa;
79	vm_page_t kern_pg, user_pg;
80	vm_object_t uobject;
81	vm_map_entry_t entry;
82	vm_pindex_t upindex, kpindex;
83	vm_prot_t prot;
84	boolean_t wired;
85
86	/*
87	 * First lookup the kernel page.
88	 */
89	kern_pg = PHYS_TO_VM_PAGE(vtophys(kaddr));
90	/*
91	 * XXX The vm object containing kern_pg needs locking.
92	 */
93	if ((vm_map_lookup(&map, uaddr,
94			   VM_PROT_WRITE, &entry, &uobject,
95			   &upindex, &prot, &wired)) != KERN_SUCCESS) {
96		return(EFAULT);
97	}
98	VM_OBJECT_LOCK(uobject);
99	if ((user_pg = vm_page_lookup(uobject, upindex)) != NULL) {
100		do
101			vm_page_lock_queues();
102		while (vm_page_sleep_if_busy(user_pg, 1, "vm_pgmoveco"));
103		vm_page_busy(user_pg);
104		pmap_remove_all(user_pg);
105		vm_page_free(user_pg);
106	} else
107		vm_page_lock_queues();
108	if (kern_pg->busy || ((kern_pg->queue - kern_pg->pc) == PQ_FREE) ||
109	    (kern_pg->hold_count != 0)|| (kern_pg->flags & PG_BUSY)) {
110		printf("vm_pgmoveco: pindex(%lu), busy(%d), PG_BUSY(%d), "
111		       "hold(%d) paddr(0x%lx)\n", (u_long)kern_pg->pindex,
112			kern_pg->busy, (kern_pg->flags & PG_BUSY) ? 1 : 0,
113			kern_pg->hold_count, (u_long)kern_pg->phys_addr);
114		if ((kern_pg->queue - kern_pg->pc) == PQ_FREE)
115			panic("vm_pgmoveco: renaming free page");
116		else
117			panic("vm_pgmoveco: renaming busy page");
118	}
119	kpindex = kern_pg->pindex;
120	vm_page_busy(kern_pg);
121	vm_page_rename(kern_pg, uobject, upindex);
122	vm_page_flag_clear(kern_pg, PG_BUSY);
123	kern_pg->valid = VM_PAGE_BITS_ALL;
124	vm_page_unlock_queues();
125	VM_OBJECT_UNLOCK(uobject);
126	vm_map_lookup_done(map, entry);
127	return(KERN_SUCCESS);
128}
129#endif /* ZERO_COPY_SOCKETS */
130
131int
132uiomove(void *cp, int n, struct uio *uio)
133{
134	struct thread *td = curthread;
135	struct iovec *iov;
136	u_int cnt;
137	int error = 0;
138	int save = 0;
139
140	KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
141	    ("uiomove: mode"));
142	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
143	    ("uiomove proc"));
144
145	if (td) {
146		mtx_lock_spin(&sched_lock);
147		save = td->td_flags & TDF_DEADLKTREAT;
148		td->td_flags |= TDF_DEADLKTREAT;
149		mtx_unlock_spin(&sched_lock);
150	}
151
152	while (n > 0 && uio->uio_resid) {
153		iov = uio->uio_iov;
154		cnt = iov->iov_len;
155		if (cnt == 0) {
156			uio->uio_iov++;
157			uio->uio_iovcnt--;
158			continue;
159		}
160		if (cnt > n)
161			cnt = n;
162
163		switch (uio->uio_segflg) {
164
165		case UIO_USERSPACE:
166			if (ticks - PCPU_GET(switchticks) >= hogticks)
167				uio_yield();
168			if (uio->uio_rw == UIO_READ)
169				error = copyout(cp, iov->iov_base, cnt);
170			else
171				error = copyin(iov->iov_base, cp, cnt);
172			if (error)
173				goto out;
174			break;
175
176		case UIO_SYSSPACE:
177			if (uio->uio_rw == UIO_READ)
178				bcopy(cp, iov->iov_base, cnt);
179			else
180				bcopy(iov->iov_base, cp, cnt);
181			break;
182		case UIO_NOCOPY:
183			break;
184		}
185		iov->iov_base = (char *)iov->iov_base + cnt;
186		iov->iov_len -= cnt;
187		uio->uio_resid -= cnt;
188		uio->uio_offset += cnt;
189		cp = (char *)cp + cnt;
190		n -= cnt;
191	}
192out:
193	if (td && save == 0) {
194		mtx_lock_spin(&sched_lock);
195		td->td_flags &= ~TDF_DEADLKTREAT;
196		mtx_unlock_spin(&sched_lock);
197	}
198	return (error);
199}
200
201#ifdef ZERO_COPY_SOCKETS
202/*
203 * Experimental support for zero-copy I/O
204 */
205static int
206userspaceco(void *cp, u_int cnt, struct uio *uio, struct vm_object *obj,
207    int disposable)
208{
209	struct iovec *iov;
210	int error;
211
212	iov = uio->uio_iov;
213	if (uio->uio_rw == UIO_READ) {
214		if ((so_zero_copy_receive != 0)
215		 && (obj != NULL)
216		 && ((cnt & PAGE_MASK) == 0)
217		 && ((((intptr_t) iov->iov_base) & PAGE_MASK) == 0)
218		 && ((uio->uio_offset & PAGE_MASK) == 0)
219		 && ((((intptr_t) cp) & PAGE_MASK) == 0)
220		 && (obj->type == OBJT_DEFAULT)
221		 && (disposable != 0)) {
222			/* SOCKET: use page-trading */
223			/*
224			 * We only want to call vm_pgmoveco() on
225			 * disposeable pages, since it gives the
226			 * kernel page to the userland process.
227			 */
228			error =	vm_pgmoveco(&curproc->p_vmspace->vm_map,
229					    obj, (vm_offset_t)cp,
230					    (vm_offset_t)iov->iov_base);
231
232			/*
233			 * If we get an error back, attempt
234			 * to use copyout() instead.  The
235			 * disposable page should be freed
236			 * automatically if we weren't able to move
237			 * it into userland.
238			 */
239			if (error != 0)
240				error = copyout(cp, iov->iov_base, cnt);
241		} else {
242			error = copyout(cp, iov->iov_base, cnt);
243		}
244	} else {
245		error = copyin(iov->iov_base, cp, cnt);
246	}
247	return (error);
248}
249
250int
251uiomoveco(void *cp, int n, struct uio *uio, struct vm_object *obj,
252    int disposable)
253{
254	struct iovec *iov;
255	u_int cnt;
256	int error;
257
258	KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
259	    ("uiomoveco: mode"));
260	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
261	    ("uiomoveco proc"));
262
263	while (n > 0 && uio->uio_resid) {
264		iov = uio->uio_iov;
265		cnt = iov->iov_len;
266		if (cnt == 0) {
267			uio->uio_iov++;
268			uio->uio_iovcnt--;
269			continue;
270		}
271		if (cnt > n)
272			cnt = n;
273
274		switch (uio->uio_segflg) {
275
276		case UIO_USERSPACE:
277			if (ticks - PCPU_GET(switchticks) >= hogticks)
278				uio_yield();
279
280			error = userspaceco(cp, cnt, uio, obj, disposable);
281
282			if (error)
283				return (error);
284			break;
285
286		case UIO_SYSSPACE:
287			if (uio->uio_rw == UIO_READ)
288				bcopy(cp, iov->iov_base, cnt);
289			else
290				bcopy(iov->iov_base, cp, cnt);
291			break;
292		case UIO_NOCOPY:
293			break;
294		}
295		iov->iov_base = (char *)iov->iov_base + cnt;
296		iov->iov_len -= cnt;
297		uio->uio_resid -= cnt;
298		uio->uio_offset += cnt;
299		cp = (char *)cp + cnt;
300		n -= cnt;
301	}
302	return (0);
303}
304#endif /* ZERO_COPY_SOCKETS */
305
306/*
307 * Give next character to user as result of read.
308 */
309int
310ureadc(int c, struct uio *uio)
311{
312	struct iovec *iov;
313	char *iov_base;
314
315again:
316	if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
317		panic("ureadc");
318	iov = uio->uio_iov;
319	if (iov->iov_len == 0) {
320		uio->uio_iovcnt--;
321		uio->uio_iov++;
322		goto again;
323	}
324	switch (uio->uio_segflg) {
325
326	case UIO_USERSPACE:
327		if (subyte(iov->iov_base, c) < 0)
328			return (EFAULT);
329		break;
330
331	case UIO_SYSSPACE:
332		iov_base = iov->iov_base;
333		*iov_base = c;
334		iov->iov_base = iov_base;
335		break;
336
337	case UIO_NOCOPY:
338		break;
339	}
340	iov->iov_base = (char *)iov->iov_base + 1;
341	iov->iov_len--;
342	uio->uio_resid--;
343	uio->uio_offset++;
344	return (0);
345}
346
347/*
348 * General routine to allocate a hash table.
349 */
350void *
351hashinit(int elements, struct malloc_type *type, u_long *hashmask)
352{
353	long hashsize;
354	LIST_HEAD(generic, generic) *hashtbl;
355	int i;
356
357	if (elements <= 0)
358		panic("hashinit: bad elements");
359	for (hashsize = 1; hashsize <= elements; hashsize <<= 1)
360		continue;
361	hashsize >>= 1;
362	hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
363	for (i = 0; i < hashsize; i++)
364		LIST_INIT(&hashtbl[i]);
365	*hashmask = hashsize - 1;
366	return (hashtbl);
367}
368
369void
370hashdestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask)
371{
372	LIST_HEAD(generic, generic) *hashtbl, *hp;
373
374	hashtbl = vhashtbl;
375	for (hp = hashtbl; hp <= &hashtbl[hashmask]; hp++)
376		if (!LIST_EMPTY(hp))
377			panic("hashdestroy: hash not empty");
378	free(hashtbl, type);
379}
380
381static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
382			2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
383			7159, 7673, 8191, 12281, 16381, 24571, 32749 };
384#define NPRIMES (sizeof(primes) / sizeof(primes[0]))
385
386/*
387 * General routine to allocate a prime number sized hash table.
388 */
389void *
390phashinit(int elements, struct malloc_type *type, u_long *nentries)
391{
392	long hashsize;
393	LIST_HEAD(generic, generic) *hashtbl;
394	int i;
395
396	if (elements <= 0)
397		panic("phashinit: bad elements");
398	for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
399		i++;
400		if (i == NPRIMES)
401			break;
402		hashsize = primes[i];
403	}
404	hashsize = primes[i - 1];
405	hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
406	for (i = 0; i < hashsize; i++)
407		LIST_INIT(&hashtbl[i]);
408	*nentries = hashsize;
409	return (hashtbl);
410}
411
412void
413uio_yield(void)
414{
415	struct thread *td;
416
417	td = curthread;
418	mtx_lock_spin(&sched_lock);
419	DROP_GIANT();
420	sched_prio(td, td->td_ksegrp->kg_user_pri); /* XXXKSE */
421	td->td_proc->p_stats->p_ru.ru_nivcsw++;
422	mi_switch();
423	mtx_unlock_spin(&sched_lock);
424	PICKUP_GIANT();
425}
426
427int
428copyinfrom(const void *src, void *dst, size_t len, int seg)
429{
430	int error = 0;
431
432	switch (seg) {
433	case UIO_USERSPACE:
434		error = copyin(src, dst, len);
435		break;
436	case UIO_SYSSPACE:
437		bcopy(src, dst, len);
438		break;
439	default:
440		panic("copyinfrom: bad seg %d\n", seg);
441	}
442	return (error);
443}
444
445int
446copyinstrfrom(const void *src, void *dst, size_t len, size_t *copied, int seg)
447{
448	int error = 0;
449
450	switch (seg) {
451	case UIO_USERSPACE:
452		error = copyinstr(src, dst, len, copied);
453		break;
454	case UIO_SYSSPACE:
455		error = copystr(src, dst, len, copied);
456		break;
457	default:
458		panic("copyinstrfrom: bad seg %d\n", seg);
459	}
460	return (error);
461}
462