subr_hash.c revision 111977 
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: head/sys/kern/kern_subr.c 111977 2003-03-08 06:58:18Z ken $
40 */
41
42#include "opt_zero.h"
43
44#include <sys/param.h>
45#include <sys/systm.h>
46#include <sys/kernel.h>
47#include <sys/ktr.h>
48#include <sys/lock.h>
49#include <sys/mutex.h>
50#include <sys/proc.h>
51#include <sys/malloc.h>
52#include <sys/resourcevar.h>
53#include <sys/sched.h>
54#include <sys/sysctl.h>
55#include <sys/vnode.h>
56
57#include <vm/vm.h>
58#include <vm/vm_page.h>
59#include <vm/vm_map.h>
60#ifdef ZERO_COPY_SOCKETS
61#include <vm/vm_param.h>
62#include <vm/vm_object.h>
63#endif
64
65SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, NULL, UIO_MAXIOV,
66	"Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)");
67
68#ifdef ZERO_COPY_SOCKETS
69/* Declared in uipc_socket.c */
70extern int so_zero_copy_receive;
71
72static int
73vm_pgmoveco(vm_map_t mapa, vm_object_t srcobj, vm_offset_t kaddr,
74    vm_offset_t uaddr)
75{
76	vm_map_t map = mapa;
77	vm_page_t kern_pg, user_pg;
78	vm_object_t uobject;
79	vm_map_entry_t entry;
80	vm_pindex_t upindex, kpindex;
81	vm_prot_t prot;
82	boolean_t wired;
83
84	/*
85	 * First lookup the kernel page.
86	 */
87	kern_pg = PHYS_TO_VM_PAGE(vtophys(kaddr));
88
89	if ((vm_map_lookup(&map, uaddr,
90			   VM_PROT_WRITE, &entry, &uobject,
91			   &upindex, &prot, &wired)) != KERN_SUCCESS) {
92		return(EFAULT);
93	}
94	if ((user_pg = vm_page_lookup(uobject, upindex)) != NULL) {
95		do
96			vm_page_lock_queues();
97		while (vm_page_sleep_if_busy(user_pg, 1, "vm_pgmoveco"));
98		vm_page_busy(user_pg);
99		pmap_remove_all(user_pg);
100		vm_page_free(user_pg);
101	} else
102		vm_page_lock_queues();
103	if (kern_pg->busy || ((kern_pg->queue - kern_pg->pc) == PQ_FREE) ||
104	    (kern_pg->hold_count != 0)|| (kern_pg->flags & PG_BUSY)) {
105		printf("vm_pgmoveco: pindex(%lu), busy(%d), PG_BUSY(%d), "
106		       "hold(%d) paddr(0x%lx)\n", (u_long)kern_pg->pindex,
107			kern_pg->busy, (kern_pg->flags & PG_BUSY) ? 1 : 0,
108			kern_pg->hold_count, (u_long)kern_pg->phys_addr);
109		if ((kern_pg->queue - kern_pg->pc) == PQ_FREE)
110			panic("vm_pgmoveco: renaming free page");
111		else
112			panic("vm_pgmoveco: renaming busy page");
113	}
114	kpindex = kern_pg->pindex;
115	vm_page_busy(kern_pg);
116	vm_page_rename(kern_pg, uobject, upindex);
117	vm_page_flag_clear(kern_pg, PG_BUSY);
118	kern_pg->valid = VM_PAGE_BITS_ALL;
119	vm_page_unlock_queues();
120
121	vm_map_lookup_done(map, entry);
122	return(KERN_SUCCESS);
123}
124#endif /* ZERO_COPY_SOCKETS */
125
126int
127uiomove(void *cp, int n, struct uio *uio)
128{
129	struct thread *td = curthread;
130	struct iovec *iov;
131	u_int cnt;
132	int error = 0;
133	int save = 0;
134
135	KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
136	    ("uiomove: mode"));
137	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
138	    ("uiomove proc"));
139
140	if (td) {
141		mtx_lock_spin(&sched_lock);
142		save = td->td_flags & TDF_DEADLKTREAT;
143		td->td_flags |= TDF_DEADLKTREAT;
144		mtx_unlock_spin(&sched_lock);
145	}
146
147	while (n > 0 && uio->uio_resid) {
148		iov = uio->uio_iov;
149		cnt = iov->iov_len;
150		if (cnt == 0) {
151			uio->uio_iov++;
152			uio->uio_iovcnt--;
153			continue;
154		}
155		if (cnt > n)
156			cnt = n;
157
158		switch (uio->uio_segflg) {
159
160		case UIO_USERSPACE:
161			if (ticks - PCPU_GET(switchticks) >= hogticks)
162				uio_yield();
163			if (uio->uio_rw == UIO_READ)
164				error = copyout(cp, iov->iov_base, cnt);
165			else
166				error = copyin(iov->iov_base, cp, cnt);
167			if (error)
168				goto out;
169			break;
170
171		case UIO_SYSSPACE:
172			if (uio->uio_rw == UIO_READ)
173				bcopy(cp, iov->iov_base, cnt);
174			else
175				bcopy(iov->iov_base, cp, cnt);
176			break;
177		case UIO_NOCOPY:
178			break;
179		}
180		iov->iov_base = (char *)iov->iov_base + cnt;
181		iov->iov_len -= cnt;
182		uio->uio_resid -= cnt;
183		uio->uio_offset += cnt;
184		cp = (char *)cp + cnt;
185		n -= cnt;
186	}
187out:
188	if (td != curthread) printf("uiomove: IT CHANGED!");
189	td = curthread;	/* Might things have changed in copyin/copyout? */
190	if (td) {
191		mtx_lock_spin(&sched_lock);
192		td->td_flags = (td->td_flags & ~TDF_DEADLKTREAT) | save;
193		mtx_unlock_spin(&sched_lock);
194	}
195	return (error);
196}
197
198#ifdef ZERO_COPY_SOCKETS
199/*
200 * Experimental support for zero-copy I/O
201 */
202static int
203userspaceco(void *cp, u_int cnt, struct uio *uio, struct vm_object *obj,
204    int disposable)
205{
206	struct iovec *iov;
207	int error;
208
209	iov = uio->uio_iov;
210	if (uio->uio_rw == UIO_READ) {
211		if ((so_zero_copy_receive != 0)
212		 && (obj != NULL)
213		 && ((cnt & PAGE_MASK) == 0)
214		 && ((((intptr_t) iov->iov_base) & PAGE_MASK) == 0)
215		 && ((uio->uio_offset & PAGE_MASK) == 0)
216		 && ((((intptr_t) cp) & PAGE_MASK) == 0)
217		 && (obj->type == OBJT_DEFAULT)
218		 && (disposable != 0)) {
219			/* SOCKET: use page-trading */
220			/*
221			 * We only want to call vm_pgmoveco() on
222			 * disposeable pages, since it gives the
223			 * kernel page to the userland process.
224			 */
225			error =	vm_pgmoveco(&curproc->p_vmspace->vm_map,
226					    obj, (vm_offset_t)cp,
227					    (vm_offset_t)iov->iov_base);
228
229			/*
230			 * If we get an error back, attempt
231			 * to use copyout() instead.  The
232			 * disposable page should be freed
233			 * automatically if we weren't able to move
234			 * it into userland.
235			 */
236			if (error != 0)
237				error = copyout(cp, iov->iov_base, cnt);
238		} else {
239			error = copyout(cp, iov->iov_base, cnt);
240		}
241	} else {
242		error = copyin(iov->iov_base, cp, cnt);
243	}
244	return (error);
245}
246
247int
248uiomoveco(void *cp, int n, struct uio *uio, struct vm_object *obj,
249    int disposable)
250{
251	struct iovec *iov;
252	u_int cnt;
253	int error;
254
255	KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
256	    ("uiomoveco: mode"));
257	KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
258	    ("uiomoveco proc"));
259
260	while (n > 0 && uio->uio_resid) {
261		iov = uio->uio_iov;
262		cnt = iov->iov_len;
263		if (cnt == 0) {
264			uio->uio_iov++;
265			uio->uio_iovcnt--;
266			continue;
267		}
268		if (cnt > n)
269			cnt = n;
270
271		switch (uio->uio_segflg) {
272
273		case UIO_USERSPACE:
274			if (ticks - PCPU_GET(switchticks) >= hogticks)
275				uio_yield();
276
277			error = userspaceco(cp, cnt, uio, obj, disposable);
278
279			if (error)
280				return (error);
281			break;
282
283		case UIO_SYSSPACE:
284			if (uio->uio_rw == UIO_READ)
285				bcopy(cp, iov->iov_base, cnt);
286			else
287				bcopy(iov->iov_base, cp, cnt);
288			break;
289		case UIO_NOCOPY:
290			break;
291		}
292		iov->iov_base = (char *)iov->iov_base + cnt;
293		iov->iov_len -= cnt;
294		uio->uio_resid -= cnt;
295		uio->uio_offset += cnt;
296		cp = (char *)cp + cnt;
297		n -= cnt;
298	}
299	return (0);
300}
301#endif /* ZERO_COPY_SOCKETS */
302
303/*
304 * Give next character to user as result of read.
305 */
306int
307ureadc(int c, struct uio *uio)
308{
309	struct iovec *iov;
310	char *iov_base;
311
312again:
313	if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
314		panic("ureadc");
315	iov = uio->uio_iov;
316	if (iov->iov_len == 0) {
317		uio->uio_iovcnt--;
318		uio->uio_iov++;
319		goto again;
320	}
321	switch (uio->uio_segflg) {
322
323	case UIO_USERSPACE:
324		if (subyte(iov->iov_base, c) < 0)
325			return (EFAULT);
326		break;
327
328	case UIO_SYSSPACE:
329		iov_base = iov->iov_base;
330		*iov_base = c;
331		iov->iov_base = iov_base;
332		break;
333
334	case UIO_NOCOPY:
335		break;
336	}
337	iov->iov_base = (char *)iov->iov_base + 1;
338	iov->iov_len--;
339	uio->uio_resid--;
340	uio->uio_offset++;
341	return (0);
342}
343
344/*
345 * General routine to allocate a hash table.
346 */
347void *
348hashinit(int elements, struct malloc_type *type, u_long *hashmask)
349{
350	long hashsize;
351	LIST_HEAD(generic, generic) *hashtbl;
352	int i;
353
354	if (elements <= 0)
355		panic("hashinit: bad elements");
356	for (hashsize = 1; hashsize <= elements; hashsize <<= 1)
357		continue;
358	hashsize >>= 1;
359	hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
360	for (i = 0; i < hashsize; i++)
361		LIST_INIT(&hashtbl[i]);
362	*hashmask = hashsize - 1;
363	return (hashtbl);
364}
365
366void
367hashdestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask)
368{
369	LIST_HEAD(generic, generic) *hashtbl, *hp;
370
371	hashtbl = vhashtbl;
372	for (hp = hashtbl; hp <= &hashtbl[hashmask]; hp++)
373		if (!LIST_EMPTY(hp))
374			panic("hashdestroy: hash not empty");
375	free(hashtbl, type);
376}
377
378static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
379			2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
380			7159, 7673, 8191, 12281, 16381, 24571, 32749 };
381#define NPRIMES (sizeof(primes) / sizeof(primes[0]))
382
383/*
384 * General routine to allocate a prime number sized hash table.
385 */
386void *
387phashinit(int elements, struct malloc_type *type, u_long *nentries)
388{
389	long hashsize;
390	LIST_HEAD(generic, generic) *hashtbl;
391	int i;
392
393	if (elements <= 0)
394		panic("phashinit: bad elements");
395	for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
396		i++;
397		if (i == NPRIMES)
398			break;
399		hashsize = primes[i];
400	}
401	hashsize = primes[i - 1];
402	hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
403	for (i = 0; i < hashsize; i++)
404		LIST_INIT(&hashtbl[i]);
405	*nentries = hashsize;
406	return (hashtbl);
407}
408
409void
410uio_yield(void)
411{
412	struct thread *td;
413
414	td = curthread;
415	mtx_lock_spin(&sched_lock);
416	DROP_GIANT();
417	sched_prio(td, td->td_ksegrp->kg_user_pri); /* XXXKSE */
418	td->td_proc->p_stats->p_ru.ru_nivcsw++;
419	mi_switch();
420	mtx_unlock_spin(&sched_lock);
421	PICKUP_GIANT();
422}
423
424int
425copyinfrom(const void *src, void *dst, size_t len, int seg)
426{
427	int error = 0;
428
429	switch (seg) {
430	case UIO_USERSPACE:
431		error = copyin(src, dst, len);
432		break;
433	case UIO_SYSSPACE:
434		bcopy(src, dst, len);
435		break;
436	default:
437		panic("copyinfrom: bad seg %d\n", seg);
438	}
439	return (error);
440}
441
442int
443copyinstrfrom(const void *src, void *dst, size_t len, size_t *copied, int seg)
444{
445	int error = 0;
446
447	switch (seg) {
448	case UIO_USERSPACE:
449		error = copyinstr(src, dst, len, copied);
450		break;
451	case UIO_SYSSPACE:
452		error = copystr(src, dst, len, copied);
453		break;
454	default:
455		panic("copyinstrfrom: bad seg %d\n", seg);
456	}
457	return (error);
458}
459