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>
45 #include <sys/limits.h>
49 #include <sys/resourcevar.h>
50 #include <sys/rwlock.h>
51 #include <sys/sched.h>
52 #include <sys/sysctl.h>
53 #include <sys/vnode.h>
56 #include <vm/vm_param.h>
57 #include <vm/vm_extern.h>
58 #include <vm/vm_page.h>
59 #include <vm/vm_pageout.h>
60 #include <vm/vm_map.h>
61 #ifdef SOCKET_SEND_COW
62 #include <vm/vm_object.h>
65 SYSCTL_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)");
68 static int uiomove_faultflag(void *cp, int n, struct uio *uio, int nofault);
70 #ifdef SOCKET_SEND_COW
71 /* Declared in uipc_socket.c */
72 extern int so_zero_copy_receive;
75 * Identify the physical page mapped at the given kernel virtual
76 * address. Insert this physical page into the given address space at
77 * the given virtual address, replacing the physical page, if any,
78 * that already exists there.
81 vm_pgmoveco(vm_map_t mapa, vm_offset_t kaddr, vm_offset_t uaddr)
84 vm_page_t kern_pg, user_pg;
91 KASSERT((uaddr & PAGE_MASK) == 0,
92 ("vm_pgmoveco: uaddr is not page aligned"));
95 * Herein the physical page is validated and dirtied. It is
96 * unwired in sf_buf_mext().
98 kern_pg = PHYS_TO_VM_PAGE(vtophys(kaddr));
99 kern_pg->valid = VM_PAGE_BITS_ALL;
100 KASSERT(kern_pg->queue == PQ_NONE && kern_pg->wire_count == 1,
101 ("vm_pgmoveco: kern_pg is not correctly wired"));
103 if ((vm_map_lookup(&map, uaddr,
104 VM_PROT_WRITE, &entry, &uobject,
105 &upindex, &prot, &wired)) != KERN_SUCCESS) {
108 VM_OBJECT_WLOCK(uobject);
110 if ((user_pg = vm_page_lookup(uobject, upindex)) != NULL) {
111 if (vm_page_sleep_if_busy(user_pg, "vm_pgmoveco"))
113 vm_page_lock(user_pg);
114 pmap_remove_all(user_pg);
115 vm_page_free(user_pg);
116 vm_page_unlock(user_pg);
119 * Even if a physical page does not exist in the
120 * object chain's first object, a physical page from a
121 * backing object may be mapped read only.
123 if (uobject->backing_object != NULL)
124 pmap_remove(map->pmap, uaddr, uaddr + PAGE_SIZE);
126 if (vm_page_insert(kern_pg, uobject, upindex)) {
127 VM_OBJECT_WUNLOCK(uobject);
129 VM_OBJECT_WLOCK(uobject);
132 vm_page_dirty(kern_pg);
133 VM_OBJECT_WUNLOCK(uobject);
134 vm_map_lookup_done(map, entry);
135 return(KERN_SUCCESS);
137 #endif /* SOCKET_SEND_COW */
140 copyin_nofault(const void *udaddr, void *kaddr, size_t len)
144 save = vm_fault_disable_pagefaults();
145 error = copyin(udaddr, kaddr, len);
146 vm_fault_enable_pagefaults(save);
151 copyout_nofault(const void *kaddr, void *udaddr, size_t len)
155 save = vm_fault_disable_pagefaults();
156 error = copyout(kaddr, udaddr, len);
157 vm_fault_enable_pagefaults(save);
161 #define PHYS_PAGE_COUNT(len) (howmany(len, PAGE_SIZE) + 1)
164 physcopyin(void *src, vm_paddr_t dst, size_t len)
166 vm_page_t m[PHYS_PAGE_COUNT(len)];
171 iov[0].iov_base = src;
172 iov[0].iov_len = len;
177 uio.uio_segflg = UIO_SYSSPACE;
178 uio.uio_rw = UIO_WRITE;
179 for (i = 0; i < PHYS_PAGE_COUNT(len); i++, dst += PAGE_SIZE)
180 m[i] = PHYS_TO_VM_PAGE(dst);
181 return (uiomove_fromphys(m, dst & PAGE_MASK, len, &uio));
185 physcopyout(vm_paddr_t src, void *dst, size_t len)
187 vm_page_t m[PHYS_PAGE_COUNT(len)];
192 iov[0].iov_base = dst;
193 iov[0].iov_len = len;
198 uio.uio_segflg = UIO_SYSSPACE;
199 uio.uio_rw = UIO_READ;
200 for (i = 0; i < PHYS_PAGE_COUNT(len); i++, src += PAGE_SIZE)
201 m[i] = PHYS_TO_VM_PAGE(src);
202 return (uiomove_fromphys(m, src & PAGE_MASK, len, &uio));
205 #undef PHYS_PAGE_COUNT
208 uiomove(void *cp, int n, struct uio *uio)
211 return (uiomove_faultflag(cp, n, uio, 0));
215 uiomove_nofault(void *cp, int n, struct uio *uio)
218 return (uiomove_faultflag(cp, n, uio, 1));
222 uiomove_faultflag(void *cp, int n, struct uio *uio, int nofault)
227 int error, newflags, save;
232 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
234 KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == td,
237 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
238 "Calling uiomove()");
240 /* XXX does it make a sense to set TDP_DEADLKTREAT for UIO_SYSSPACE ? */
241 newflags = TDP_DEADLKTREAT;
242 if (uio->uio_segflg == UIO_USERSPACE && nofault) {
244 * Fail if a non-spurious page fault occurs.
246 newflags |= TDP_NOFAULTING | TDP_RESETSPUR;
248 save = curthread_pflags_set(newflags);
250 while (n > 0 && uio->uio_resid) {
261 switch (uio->uio_segflg) {
265 if (uio->uio_rw == UIO_READ)
266 error = copyout(cp, iov->iov_base, cnt);
268 error = copyin(iov->iov_base, cp, cnt);
274 if (uio->uio_rw == UIO_READ)
275 bcopy(cp, iov->iov_base, cnt);
277 bcopy(iov->iov_base, cp, cnt);
282 iov->iov_base = (char *)iov->iov_base + cnt;
284 uio->uio_resid -= cnt;
285 uio->uio_offset += cnt;
286 cp = (char *)cp + cnt;
290 curthread_pflags_restore(save);
295 * Wrapper for uiomove() that validates the arguments against a known-good
296 * kernel buffer. Currently, uiomove accepts a signed (n) argument, which
297 * is almost definitely a bad thing, so we catch that here as well. We
298 * return a runtime failure, but it might be desirable to generate a runtime
299 * assertion failure instead.
302 uiomove_frombuf(void *buf, int buflen, struct uio *uio)
306 if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
307 (offset = uio->uio_offset) != uio->uio_offset)
309 if (buflen <= 0 || offset >= buflen)
311 if ((n = buflen - offset) > IOSIZE_MAX)
313 return (uiomove((char *)buf + offset, n, uio));
316 #ifdef SOCKET_RECV_PFLIP
318 * Experimental support for zero-copy I/O
321 userspaceco(void *cp, u_int cnt, struct uio *uio, int disposable)
327 if (uio->uio_rw == UIO_READ) {
328 if ((so_zero_copy_receive != 0)
329 && ((cnt & PAGE_MASK) == 0)
330 && ((((intptr_t) iov->iov_base) & PAGE_MASK) == 0)
331 && ((uio->uio_offset & PAGE_MASK) == 0)
332 && ((((intptr_t) cp) & PAGE_MASK) == 0)
333 && (disposable != 0)) {
334 /* SOCKET: use page-trading */
336 * We only want to call vm_pgmoveco() on
337 * disposeable pages, since it gives the
338 * kernel page to the userland process.
340 error = vm_pgmoveco(&curproc->p_vmspace->vm_map,
341 (vm_offset_t)cp, (vm_offset_t)iov->iov_base);
344 * If we get an error back, attempt
345 * to use copyout() instead. The
346 * disposable page should be freed
347 * automatically if we weren't able to move
351 error = copyout(cp, iov->iov_base, cnt);
353 error = copyout(cp, iov->iov_base, cnt);
356 error = copyin(iov->iov_base, cp, cnt);
362 uiomoveco(void *cp, int n, struct uio *uio, int disposable)
368 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
369 ("uiomoveco: mode"));
370 KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
373 while (n > 0 && uio->uio_resid) {
384 switch (uio->uio_segflg) {
388 error = userspaceco(cp, cnt, uio, disposable);
394 if (uio->uio_rw == UIO_READ)
395 bcopy(cp, iov->iov_base, cnt);
397 bcopy(iov->iov_base, cp, cnt);
402 iov->iov_base = (char *)iov->iov_base + cnt;
404 uio->uio_resid -= cnt;
405 uio->uio_offset += cnt;
406 cp = (char *)cp + cnt;
411 #endif /* SOCKET_RECV_PFLIP */
414 * Give next character to user as result of read.
417 ureadc(int c, struct uio *uio)
422 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
426 if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
429 if (iov->iov_len == 0) {
434 switch (uio->uio_segflg) {
437 if (subyte(iov->iov_base, c) < 0)
442 iov_base = iov->iov_base;
449 iov->iov_base = (char *)iov->iov_base + 1;
457 copyinfrom(const void * __restrict src, void * __restrict dst, size_t len,
464 error = copyin(src, dst, len);
467 bcopy(src, dst, len);
470 panic("copyinfrom: bad seg %d\n", seg);
476 copyinstrfrom(const void * __restrict src, void * __restrict dst, size_t len,
477 size_t * __restrict copied, int seg)
483 error = copyinstr(src, dst, len, copied);
486 error = copystr(src, dst, len, copied);
489 panic("copyinstrfrom: bad seg %d\n", seg);
495 copyiniov(const struct iovec *iovp, u_int iovcnt, struct iovec **iov, int error)
500 if (iovcnt > UIO_MAXIOV)
502 iovlen = iovcnt * sizeof (struct iovec);
503 *iov = malloc(iovlen, M_IOV, M_WAITOK);
504 error = copyin(iovp, *iov, iovlen);
513 copyinuio(const struct iovec *iovp, u_int iovcnt, struct uio **uiop)
521 if (iovcnt > UIO_MAXIOV)
523 iovlen = iovcnt * sizeof (struct iovec);
524 uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK);
525 iov = (struct iovec *)(uio + 1);
526 error = copyin(iovp, iov, iovlen);
532 uio->uio_iovcnt = iovcnt;
533 uio->uio_segflg = UIO_USERSPACE;
534 uio->uio_offset = -1;
536 for (i = 0; i < iovcnt; i++) {
537 if (iov->iov_len > IOSIZE_MAX - uio->uio_resid) {
541 uio->uio_resid += iov->iov_len;
549 cloneuio(struct uio *uiop)
554 iovlen = uiop->uio_iovcnt * sizeof (struct iovec);
555 uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK);
557 uio->uio_iov = (struct iovec *)(uio + 1);
558 bcopy(uiop->uio_iov, uio->uio_iov, iovlen);
563 * Map some anonymous memory in user space of size sz, rounded up to the page
567 copyout_map(struct thread *td, vm_offset_t *addr, size_t sz)
573 vms = td->td_proc->p_vmspace;
576 * Map somewhere after heap in process memory.
578 PROC_LOCK(td->td_proc);
579 *addr = round_page((vm_offset_t)vms->vm_daddr +
580 lim_max(td->td_proc, RLIMIT_DATA));
581 PROC_UNLOCK(td->td_proc);
583 /* round size up to page boundry */
584 size = (vm_size_t)round_page(sz);
586 error = vm_mmap(&vms->vm_map, addr, size, PROT_READ | PROT_WRITE,
587 VM_PROT_ALL, MAP_PRIVATE | MAP_ANON, OBJT_DEFAULT, NULL, 0);
593 * Unmap memory in user space.
596 copyout_unmap(struct thread *td, vm_offset_t addr, size_t sz)
604 map = &td->td_proc->p_vmspace->vm_map;
605 size = (vm_size_t)round_page(sz);
607 if (vm_map_remove(map, addr, addr + size) != KERN_SUCCESS)