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.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
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.
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
34 * @(#)kern_subr.c 8.3 (Berkeley) 1/21/94
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
46 #include <sys/limits.h>
48 #include <sys/mutex.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>
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>
64 SYSCTL_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)");
67 #ifdef ZERO_COPY_SOCKETS
68 /* Declared in uipc_socket.c */
69 extern int so_zero_copy_receive;
72 * Identify the physical page mapped at the given kernel virtual
73 * address. Insert this physical page into the given address space at
74 * the given virtual address, replacing the physical page, if any,
75 * that already exists there.
78 vm_pgmoveco(vm_map_t mapa, vm_offset_t kaddr, vm_offset_t uaddr)
81 vm_page_t kern_pg, user_pg;
88 KASSERT((uaddr & PAGE_MASK) == 0,
89 ("vm_pgmoveco: uaddr is not page aligned"));
92 * Herein the physical page is validated and dirtied. It is
93 * unwired in sf_buf_mext().
95 kern_pg = PHYS_TO_VM_PAGE(vtophys(kaddr));
96 kern_pg->valid = VM_PAGE_BITS_ALL;
97 KASSERT(kern_pg->queue == PQ_NONE && kern_pg->wire_count == 1,
98 ("vm_pgmoveco: kern_pg is not correctly wired"));
100 if ((vm_map_lookup(&map, uaddr,
101 VM_PROT_WRITE, &entry, &uobject,
102 &upindex, &prot, &wired)) != KERN_SUCCESS) {
105 VM_OBJECT_LOCK(uobject);
107 if ((user_pg = vm_page_lookup(uobject, upindex)) != NULL) {
108 if (vm_page_sleep_if_busy(user_pg, TRUE, "vm_pgmoveco"))
110 vm_page_lock_queues();
111 pmap_remove_all(user_pg);
112 vm_page_free(user_pg);
115 * Even if a physical page does not exist in the
116 * object chain's first object, a physical page from a
117 * backing object may be mapped read only.
119 if (uobject->backing_object != NULL)
120 pmap_remove(map->pmap, uaddr, uaddr + PAGE_SIZE);
121 vm_page_lock_queues();
123 vm_page_insert(kern_pg, uobject, upindex);
124 vm_page_dirty(kern_pg);
125 vm_page_unlock_queues();
126 VM_OBJECT_UNLOCK(uobject);
127 vm_map_lookup_done(map, entry);
128 return(KERN_SUCCESS);
130 #endif /* ZERO_COPY_SOCKETS */
133 uiomove(void *cp, int n, struct uio *uio)
135 struct thread *td = curthread;
141 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
143 KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
145 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
146 "Calling uiomove()");
148 save = td->td_pflags & TDP_DEADLKTREAT;
149 td->td_pflags |= TDP_DEADLKTREAT;
151 while (n > 0 && uio->uio_resid) {
162 switch (uio->uio_segflg) {
165 if (ticks - PCPU_GET(switchticks) >= hogticks)
167 if (uio->uio_rw == UIO_READ)
168 error = copyout(cp, iov->iov_base, cnt);
170 error = copyin(iov->iov_base, cp, cnt);
176 if (uio->uio_rw == UIO_READ)
177 bcopy(cp, iov->iov_base, cnt);
179 bcopy(iov->iov_base, cp, cnt);
184 iov->iov_base = (char *)iov->iov_base + cnt;
186 uio->uio_resid -= cnt;
187 uio->uio_offset += cnt;
188 cp = (char *)cp + cnt;
193 td->td_pflags &= ~TDP_DEADLKTREAT;
198 * Wrapper for uiomove() that validates the arguments against a known-good
199 * kernel buffer. Currently, uiomove accepts a signed (n) argument, which
200 * is almost definitely a bad thing, so we catch that here as well. We
201 * return a runtime failure, but it might be desirable to generate a runtime
202 * assertion failure instead.
205 uiomove_frombuf(void *buf, int buflen, struct uio *uio)
207 unsigned int offset, n;
209 if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
210 (offset = uio->uio_offset) != uio->uio_offset)
212 if (buflen <= 0 || offset >= buflen)
214 if ((n = buflen - offset) > INT_MAX)
216 return (uiomove((char *)buf + offset, n, uio));
219 #ifdef ZERO_COPY_SOCKETS
221 * Experimental support for zero-copy I/O
224 userspaceco(void *cp, u_int cnt, struct uio *uio, int disposable)
230 if (uio->uio_rw == UIO_READ) {
231 if ((so_zero_copy_receive != 0)
232 && ((cnt & PAGE_MASK) == 0)
233 && ((((intptr_t) iov->iov_base) & PAGE_MASK) == 0)
234 && ((uio->uio_offset & PAGE_MASK) == 0)
235 && ((((intptr_t) cp) & PAGE_MASK) == 0)
236 && (disposable != 0)) {
237 /* SOCKET: use page-trading */
239 * We only want to call vm_pgmoveco() on
240 * disposeable pages, since it gives the
241 * kernel page to the userland process.
243 error = vm_pgmoveco(&curproc->p_vmspace->vm_map,
244 (vm_offset_t)cp, (vm_offset_t)iov->iov_base);
247 * If we get an error back, attempt
248 * to use copyout() instead. The
249 * disposable page should be freed
250 * automatically if we weren't able to move
254 error = copyout(cp, iov->iov_base, cnt);
256 error = copyout(cp, iov->iov_base, cnt);
259 error = copyin(iov->iov_base, cp, cnt);
265 uiomoveco(void *cp, int n, struct uio *uio, int disposable)
271 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
272 ("uiomoveco: mode"));
273 KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
276 while (n > 0 && uio->uio_resid) {
287 switch (uio->uio_segflg) {
290 if (ticks - PCPU_GET(switchticks) >= hogticks)
293 error = userspaceco(cp, cnt, uio, disposable);
300 if (uio->uio_rw == UIO_READ)
301 bcopy(cp, iov->iov_base, cnt);
303 bcopy(iov->iov_base, cp, cnt);
308 iov->iov_base = (char *)iov->iov_base + cnt;
310 uio->uio_resid -= cnt;
311 uio->uio_offset += cnt;
312 cp = (char *)cp + cnt;
317 #endif /* ZERO_COPY_SOCKETS */
320 * Give next character to user as result of read.
323 ureadc(int c, struct uio *uio)
329 if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
332 if (iov->iov_len == 0) {
337 switch (uio->uio_segflg) {
340 if (subyte(iov->iov_base, c) < 0)
345 iov_base = iov->iov_base;
347 iov->iov_base = iov_base;
353 iov->iov_base = (char *)iov->iov_base + 1;
361 * General routine to allocate a hash table with control of memory flags.
364 hashinit_flags(int elements, struct malloc_type *type, u_long *hashmask,
368 LIST_HEAD(generic, generic) *hashtbl;
372 panic("hashinit: bad elements");
374 /* Exactly one of HASH_WAITOK and HASH_NOWAIT must be set. */
375 KASSERT((flags & HASH_WAITOK) ^ (flags & HASH_NOWAIT),
376 ("Bad flags (0x%x) passed to hashinit_flags", flags));
378 for (hashsize = 1; hashsize <= elements; hashsize <<= 1)
382 if (flags & HASH_NOWAIT)
383 hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl),
386 hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl),
389 if (hashtbl != NULL) {
390 for (i = 0; i < hashsize; i++)
391 LIST_INIT(&hashtbl[i]);
392 *hashmask = hashsize - 1;
398 * Allocate and initialize a hash table with default flag: may sleep.
401 hashinit(int elements, struct malloc_type *type, u_long *hashmask)
404 return (hashinit_flags(elements, type, hashmask, HASH_WAITOK));
408 hashdestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask)
410 LIST_HEAD(generic, generic) *hashtbl, *hp;
413 for (hp = hashtbl; hp <= &hashtbl[hashmask]; hp++)
415 panic("hashdestroy: hash not empty");
419 static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
420 2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
421 7159, 7673, 8191, 12281, 16381, 24571, 32749 };
422 #define NPRIMES (sizeof(primes) / sizeof(primes[0]))
425 * General routine to allocate a prime number sized hash table.
428 phashinit(int elements, struct malloc_type *type, u_long *nentries)
431 LIST_HEAD(generic, generic) *hashtbl;
435 panic("phashinit: bad elements");
436 for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
440 hashsize = primes[i];
442 hashsize = primes[i - 1];
443 hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
444 for (i = 0; i < hashsize; i++)
445 LIST_INIT(&hashtbl[i]);
446 *nentries = hashsize;
458 sched_prio(td, td->td_user_pri);
459 mi_switch(SW_INVOL, NULL);
465 copyinfrom(const void * __restrict src, void * __restrict dst, size_t len,
472 error = copyin(src, dst, len);
475 bcopy(src, dst, len);
478 panic("copyinfrom: bad seg %d\n", seg);
484 copyinstrfrom(const void * __restrict src, void * __restrict dst, size_t len,
485 size_t * __restrict copied, int seg)
491 error = copyinstr(src, dst, len, copied);
494 error = copystr(src, dst, len, copied);
497 panic("copyinstrfrom: bad seg %d\n", seg);
503 copyiniov(struct iovec *iovp, u_int iovcnt, struct iovec **iov, int error)
508 if (iovcnt > UIO_MAXIOV)
510 iovlen = iovcnt * sizeof (struct iovec);
511 *iov = malloc(iovlen, M_IOV, M_WAITOK);
512 error = copyin(iovp, *iov, iovlen);
521 copyinuio(struct iovec *iovp, u_int iovcnt, struct uio **uiop)
529 if (iovcnt > UIO_MAXIOV)
531 iovlen = iovcnt * sizeof (struct iovec);
532 uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK);
533 iov = (struct iovec *)(uio + 1);
534 error = copyin(iovp, iov, iovlen);
540 uio->uio_iovcnt = iovcnt;
541 uio->uio_segflg = UIO_USERSPACE;
542 uio->uio_offset = -1;
544 for (i = 0; i < iovcnt; i++) {
545 if (iov->iov_len > INT_MAX - uio->uio_resid) {
549 uio->uio_resid += iov->iov_len;
557 cloneuio(struct uio *uiop)
562 iovlen = uiop->uio_iovcnt * sizeof (struct iovec);
563 uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK);
565 uio->uio_iov = (struct iovec *)(uio + 1);
566 bcopy(uiop->uio_iov, uio->uio_iov, iovlen);