2 * Copyright (c) 1993 The Regents of the University of California.
3 * Copyright (c) 2003 Peter Wemm.
4 * Copyright (c) 2008 The DragonFly Project.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * $FreeBSD: src/sys/amd64/amd64/support.S,v 1.127 2007/05/23 08:33:04 kib Exp $
34 #include <machine/asmacros.h>
35 #include <machine/pmap.h>
44 * bzero(ptr:%rdi, bytes:%rsi)
46 * Using rep stosq is 70% faster than a %rax loop and almost as fast as
47 * a %xmm0 loop on a modern intel cpu.
49 * Do not use non-termportal instructions here as we do not know the caller's
68 * Using rep stosq is nearly as fast as using %xmm0 on a modern intel cpu,
69 * and about 70% faster than a %rax loop.
71 * Do not use non-termportal instructions here as we do not know the caller's
75 movq $PAGE_SIZE
>>3,%rcx
83 * bcmp(ptr:%rdi, ptr:%rsi, bytes:%rdx)
88 cld
/* compare forwards */
103 * bcopy(src:%rdi, dst:%rsi, cnt:%rdx)
105 * ws@tools.de (Wolfgang Solfrank, TooLs GmbH) +49-228-985800
113 cmpq
%rcx
,%rax
/* overlapping && src < dst? */
116 shrq $
3,%rcx
/* copy by 64-bit words */
117 cld
/* nope, copy forwards */
121 andq $
7,%rcx
/* any bytes left? */
128 addq
%rcx
,%rdi
/* copy backwards */
132 andq $
7,%rcx
/* any fractional bytes? */
136 movq
%rdx
,%rcx
/* copy by 32-bit words */
146 movq $
0x200,%rax
/* the manual says that bit 10 must be set to 1 */
147 movq
%rax
,%dr7
/* disable all breapoints first */
157 * memcpy(dst:%rdi, src:%rsi, bytes:%rdx)
159 * Note: memcpy does not support overlapping copies
164 shrq $
3,%rcx
/* copy by 64-bit words */
165 cld
/* copy forwards */
169 andq $
7,%rcx
/* any bytes left? */
175 /* fillw(pat, base, cnt) */
176 /* %rdi,%rsi, %rdx */
186 /*****************************************************************************/
187 /* copyout and fubyte family */
188 /*****************************************************************************/
190 * Access user memory from inside the kernel. These routines should be
191 * the only places that do this.
193 * These routines set curpcb->onfault for the time they execute. When a
194 * protection violation occurs inside the functions, the trap handler
195 * returns to *curpcb->onfault instead of the function.
199 * std_copyout(from_kernel, to_user, len) - MP SAFE
203 movq PCPU
(curthread
),%rax
204 movq TD_PCB
(%rax
), %rax
205 movq $copyout_fault
,PCB_ONFAULT
(%rax
)
206 movq
%rsp
,PCB_ONFAULT_SP
(%rax
)
207 testq
%rdx
,%rdx
/* anything to do? */
211 * Check explicitly for non-user addresses. If 486 write protection
212 * is being used, this check is essential because we are in kernel
213 * mode so the h/w does not provide any protection against writing
218 * First, prevent address wrapping.
224 * XXX STOP USING VM_MAX_USER_ADDRESS.
225 * It is an end address, not a max, so every time it is used correctly it
226 * looks like there is an off by one error, and of course it caused an off
227 * by one error in several places.
229 movq $VM_MAX_USER_ADDRESS
,%rcx
234 /* bcopy(%rsi, %rdi, %rdx) */
248 movq PCPU
(curthread
),%rdx
249 movq TD_PCB
(%rdx
), %rdx
250 movq
%rax
,PCB_ONFAULT
(%rdx
)
255 movq PCPU
(curthread
),%rdx
256 movq TD_PCB
(%rdx
), %rdx
257 movq $
0,PCB_ONFAULT
(%rdx
)
262 * std_copyin(from_user, to_kernel, len) - MP SAFE
266 movq PCPU
(curthread
),%rax
267 movq TD_PCB
(%rax
), %rax
268 movq $copyin_fault
,PCB_ONFAULT
(%rax
)
269 movq
%rsp
,PCB_ONFAULT_SP
(%rax
)
270 testq
%rdx
,%rdx
/* anything to do? */
274 * make sure address is valid
279 movq $VM_MAX_USER_ADDRESS
,%rcx
286 shrq $
3,%rcx
/* copy longword-wise */
291 andb $
7,%cl
/* copy remaining bytes */
297 movq PCPU
(curthread
),%rdx
298 movq TD_PCB
(%rdx
), %rdx
299 movq
%rax
,PCB_ONFAULT
(%rdx
)
304 movq PCPU
(curthread
),%rdx
305 movq TD_PCB
(%rdx
), %rdx
306 movq $
0,PCB_ONFAULT
(%rdx
)
311 * casu32 - Compare and set user integer. Returns -1 or the current value.
312 * dst = %rdi, old = %rsi, new = %rdx
315 movq PCPU
(curthread
),%rcx
316 movq TD_PCB
(%rcx
), %rcx
317 movq $fusufault
,PCB_ONFAULT
(%rcx
)
318 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
320 movq $VM_MAX_USER_ADDRESS-
4,%rax
321 cmpq
%rax
,%rdi
/* verify address is valid */
324 movl
%esi
,%eax
/* old */
326 cmpxchgl
%edx
,(%rdi
) /* new = %edx */
329 * The old value is in %eax. If the store succeeded it will be the
330 * value we expected (old) from before the store, otherwise it will
331 * be the current value.
334 movq PCPU
(curthread
),%rcx
335 movq TD_PCB
(%rcx
), %rcx
336 movq $
0,PCB_ONFAULT
(%rcx
)
340 * swapu32 - Swap int in user space. ptr = %rdi, val = %rsi
343 movq PCPU
(curthread
),%rcx
344 movq TD_PCB
(%rcx
), %rcx
345 movq $fusufault
,PCB_ONFAULT
(%rcx
)
346 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
348 movq $VM_MAX_USER_ADDRESS-
4,%rax
349 cmpq
%rax
,%rdi
/* verify address is valid */
352 movq
%rsi
,%rax
/* old */
356 * The old value is in %rax. If the store succeeded it will be the
357 * value we expected (old) from before the store, otherwise it will
358 * be the current value.
361 movq PCPU
(curthread
),%rcx
362 movq TD_PCB
(%rcx
), %rcx
363 movq $
0,PCB_ONFAULT
(%rcx
)
367 * casu64 - Compare and set user word. Returns -1 or the current value.
368 * dst = %rdi, old = %rsi, new = %rdx
371 movq PCPU
(curthread
),%rcx
372 movq TD_PCB
(%rcx
), %rcx
373 movq $fusufault
,PCB_ONFAULT
(%rcx
)
374 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
376 movq $VM_MAX_USER_ADDRESS-
8,%rax
377 cmpq
%rax
,%rdi
/* verify address is valid */
380 movq
%rsi
,%rax
/* old */
382 cmpxchgq
%rdx
,(%rdi
) /* new = %rdx */
385 * The old value is in %rax. If the store succeeded it will be the
386 * value we expected (old) from before the store, otherwise it will
387 * be the current value.
390 movq PCPU
(curthread
),%rcx
391 movq TD_PCB
(%rcx
), %rcx
392 movq $
0,PCB_ONFAULT
(%rcx
)
396 * swapu64 - Swap long in user space. ptr = %rdi, val = %rsi
399 movq PCPU
(curthread
),%rcx
400 movq TD_PCB
(%rcx
), %rcx
401 movq $fusufault
,PCB_ONFAULT
(%rcx
)
402 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
404 movq $VM_MAX_USER_ADDRESS-
8,%rax
405 cmpq
%rax
,%rdi
/* verify address is valid */
408 movq
%rsi
,%rax
/* old */
412 * The old value is in %rax. If the store succeeded it will be the
413 * value we expected (old) from before the store, otherwise it will
414 * be the current value.
417 movq PCPU
(curthread
),%rcx
418 movq TD_PCB
(%rcx
), %rcx
419 movq $
0,PCB_ONFAULT
(%rcx
)
423 * Fetch (load) a 64-bit word, a 32-bit word, a 16-bit word, or an 8-bit
424 * byte from user memory. All these functions are MPSAFE.
429 movq PCPU
(curthread
),%rcx
430 movq TD_PCB
(%rcx
), %rcx
431 movq $fusufault
,PCB_ONFAULT
(%rcx
)
432 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
434 movq $VM_MAX_USER_ADDRESS-
8,%rax
435 cmpq
%rax
,%rdi
/* verify address is valid */
439 movq $
0,PCB_ONFAULT
(%rcx
)
443 movq PCPU
(curthread
),%rcx
444 movq TD_PCB
(%rcx
), %rcx
445 movq $fusufault
,PCB_ONFAULT
(%rcx
)
446 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
448 movq $VM_MAX_USER_ADDRESS-
4,%rax
449 cmpq
%rax
,%rdi
/* verify address is valid */
453 movq $
0,PCB_ONFAULT
(%rcx
)
457 movq PCPU
(curthread
),%rcx
458 movq TD_PCB
(%rcx
), %rcx
459 movq $fusufault
,PCB_ONFAULT
(%rcx
)
460 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
462 movq $VM_MAX_USER_ADDRESS-
1,%rax
467 movq $
0,PCB_ONFAULT
(%rcx
)
472 movq PCPU
(curthread
),%rcx
474 movq TD_PCB
(%rcx
), %rcx
475 movq
%rax
,PCB_ONFAULT
(%rcx
)
480 * Store a 64-bit word, a 32-bit word, a 16-bit word, or an 8-bit byte to
481 * user memory. All these functions are MPSAFE.
483 * addr = %rdi, value = %rsi
488 movq PCPU
(curthread
),%rcx
489 movq TD_PCB
(%rcx
), %rcx
490 movq $fusufault
,PCB_ONFAULT
(%rcx
)
491 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
493 movq $VM_MAX_USER_ADDRESS-
8,%rax
494 cmpq
%rax
,%rdi
/* verify address validity */
499 movq PCPU
(curthread
),%rcx
500 movq TD_PCB
(%rcx
), %rcx
501 movq
%rax
,PCB_ONFAULT
(%rcx
)
508 movq PCPU
(curthread
),%rcx
509 movq TD_PCB
(%rcx
), %rcx
510 movq $fusufault
,PCB_ONFAULT
(%rcx
)
511 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
513 movq $VM_MAX_USER_ADDRESS-
4,%rax
514 cmpq
%rax
,%rdi
/* verify address validity */
519 movq PCPU
(curthread
),%rcx
520 movq TD_PCB
(%rcx
), %rcx
521 movq
%rax
,PCB_ONFAULT
(%rcx
)
525 movq PCPU
(curthread
),%rcx
526 movq TD_PCB
(%rcx
), %rcx
527 movq $fusufault
,PCB_ONFAULT
(%rcx
)
528 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
530 movq $VM_MAX_USER_ADDRESS-
1,%rax
531 cmpq
%rax
,%rdi
/* verify address validity */
537 movq PCPU
(curthread
),%rcx
/* restore trashed register */
538 movq TD_PCB
(%rcx
), %rcx
539 movq
%rax
,PCB_ONFAULT
(%rcx
)
543 * std_copyinstr(from, to, maxlen, int *lencopied) - MP SAFE
544 * %rdi, %rsi, %rdx, %rcx
546 * copy a string from from to to, stop when a 0 character is reached.
547 * return ENAMETOOLONG if string is longer than maxlen, and
548 * EFAULT on protection violations. If lencopied is non-zero,
549 * return the actual length in *lencopied.
552 movq
%rdx
,%r8 /* %r8 = maxlen */
553 movq
%rcx
,%r9 /* %r9 = *len */
554 xchgq
%rdi
,%rsi
/* %rdi = from, %rsi = to */
555 movq PCPU
(curthread
),%rcx
556 movq TD_PCB
(%rcx
), %rcx
557 movq $cpystrflt
,PCB_ONFAULT
(%rcx
)
558 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
560 movq $VM_MAX_USER_ADDRESS
,%rax
562 /* make sure 'from' is within bounds */
566 /* restrict maxlen to <= VM_MAX_USER_ADDRESS-from */
584 /* Success -- 0 byte reached */
589 /* rdx is zero - return ENAMETOOLONG or EFAULT */
590 movq $VM_MAX_USER_ADDRESS
,%rax
594 movq $ENAMETOOLONG
,%rax
601 /* set *lencopied and return %eax */
602 movq PCPU
(curthread
),%rcx
603 movq TD_PCB
(%rcx
), %rcx
604 movq $
0,PCB_ONFAULT
(%rcx
)
615 * copystr(from, to, maxlen, int *lencopied) - MP SAFE
616 * %rdi, %rsi, %rdx, %rcx
619 movq
%rdx
,%r8 /* %r8 = maxlen */
632 /* Success -- 0 byte reached */
637 /* rdx is zero -- return ENAMETOOLONG */
638 movq $ENAMETOOLONG
,%rax
644 /* set *lencopied and return %rax */
651 * Handling of special x86_64 registers and descriptor tables etc
654 /* void lgdt(struct region_descriptor *rdp); */
656 /* reload the descriptor table */
659 /* flush the prefetch q */
666 movl
%eax
,%fs /* Beware, use wrmsr to set 64 bit base */
667 movl
%eax
,%gs
/* Beware, use wrmsr to set 64 bit base */
670 /* reload code selector by turning return into intersegmental return */
677 /*****************************************************************************/
678 /* setjmp, longjmp */
679 /*****************************************************************************/
682 movq
%rbx
,0(%rdi
) /* save rbx */
683 movq
%rsp
,8(%rdi
) /* save rsp */
684 movq
%rbp
,16(%rdi
) /* save rbp */
685 movq
%r12,24(%rdi
) /* save r12 */
686 movq
%r13,32(%rdi
) /* save r13 */
687 movq
%r14,40(%rdi
) /* save r14 */
688 movq
%r15,48(%rdi
) /* save r15 */
689 movq
0(%rsp
),%rdx
/* get rta */
690 movq
%rdx
,56(%rdi
) /* save rip */
691 xorl
%eax
,%eax
/* return(0); */
695 movq
0(%rdi
),%rbx
/* restore rbx */
696 movq
8(%rdi
),%rsp
/* restore rsp */
697 movq
16(%rdi
),%rbp
/* restore rbp */
698 movq
24(%rdi
),%r12 /* restore r12 */
699 movq
32(%rdi
),%r13 /* restore r13 */
700 movq
40(%rdi
),%r14 /* restore r14 */
701 movq
48(%rdi
),%r15 /* restore r15 */
702 movq
56(%rdi
),%rdx
/* get rta */
703 movq
%rdx
,0(%rsp
) /* put in return frame */
704 xorl
%eax
,%eax
/* return(1); */
709 * Support for reading MSRs in the safe manner.
712 /* int rdmsr_safe(u_int msr, uint64_t *data) */
713 movq PCPU
(curthread
),%r8
714 movq TD_PCB
(%r8), %r8
715 movq $msr_onfault
,PCB_ONFAULT
(%r8)
716 movq
%rsp
,PCB_ONFAULT_SP
(%r8)
718 rdmsr
/* Read MSR pointed by %ecx. Returns
719 hi byte in edx, lo in %eax */
720 salq $
32,%rdx
/* sign-shift %rdx left */
721 movl
%eax
,%eax
/* zero-extend %eax -> %rax */
725 movq
%rax
,PCB_ONFAULT
(%r8)
729 * Support for writing MSRs in the safe manner.
732 /* int wrmsr_safe(u_int msr, uint64_t data) */
733 movq PCPU
(curthread
),%r8
734 movq TD_PCB
(%r8), %r8
735 movq $msr_onfault
,PCB_ONFAULT
(%r8)
736 movq
%rsp
,PCB_ONFAULT_SP
(%r8)
741 wrmsr
/* Write MSR pointed by %ecx. Accepts
742 hi byte in edx, lo in %eax. */
744 movq
%rax
,PCB_ONFAULT
(%r8)
748 * MSR operations fault handler
752 movq PCPU
(curthread
),%r8
753 movq TD_PCB
(%r8), %r8
754 movq $
0,PCB_ONFAULT
(%r8)
759 * Support for BB-profiling (gcc -a). The kernbb program will extract
760 * the data from the kernel.
770 NON_GPROF_ENTRY
(__bb_init_func
)