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 * casuword32. 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 * casuword. Compare and set user word. Returns -1 or the current value.
341 * dst = %rdi, old = %rsi, new = %rdx
344 movq PCPU
(curthread
),%rcx
345 movq TD_PCB
(%rcx
), %rcx
346 movq $fusufault
,PCB_ONFAULT
(%rcx
)
347 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
349 movq $VM_MAX_USER_ADDRESS-
8,%rax
350 cmpq
%rax
,%rdi
/* verify address is valid */
353 movq
%rsi
,%rax
/* old */
355 cmpxchgq
%rdx
,(%rdi
) /* new = %rdx */
358 * The old value is in %rax. If the store succeeded it will be the
359 * value we expected (old) from before the store, otherwise it will
360 * be the current value.
363 movq PCPU
(curthread
),%rcx
364 movq TD_PCB
(%rcx
), %rcx
365 movq $
0,PCB_ONFAULT
(%rcx
)
369 * Fetch (load) a 64-bit word, a 32-bit word, a 16-bit word, or an 8-bit
370 * byte from user memory. All these functions are MPSAFE.
376 movq PCPU
(curthread
),%rcx
377 movq TD_PCB
(%rcx
), %rcx
378 movq $fusufault
,PCB_ONFAULT
(%rcx
)
379 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
381 movq $VM_MAX_USER_ADDRESS-
8,%rax
382 cmpq
%rax
,%rdi
/* verify address is valid */
386 movq $
0,PCB_ONFAULT
(%rcx
)
390 movq PCPU
(curthread
),%rcx
391 movq TD_PCB
(%rcx
), %rcx
392 movq $fusufault
,PCB_ONFAULT
(%rcx
)
393 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
395 movq $VM_MAX_USER_ADDRESS-
4,%rax
396 cmpq
%rax
,%rdi
/* verify address is valid */
400 movq $
0,PCB_ONFAULT
(%rcx
)
404 * fuswintr() and suswintr() are specialized variants of fuword16() and
405 * suword16(), respectively. They are called from the profiling code,
406 * potentially at interrupt time. If they fail, that's okay; good things
407 * will happen later. They always fail for now, until the trap code is
408 * able to deal with this.
416 movq PCPU
(curthread
),%rcx
417 movq TD_PCB
(%rcx
), %rcx
418 movq $fusufault
,PCB_ONFAULT
(%rcx
)
419 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
421 movq $VM_MAX_USER_ADDRESS-
2,%rax
426 movq $
0,PCB_ONFAULT
(%rcx
)
430 movq PCPU
(curthread
),%rcx
431 movq TD_PCB
(%rcx
), %rcx
432 movq $fusufault
,PCB_ONFAULT
(%rcx
)
433 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
435 movq $VM_MAX_USER_ADDRESS-
1,%rax
440 movq $
0,PCB_ONFAULT
(%rcx
)
445 movq PCPU
(curthread
),%rcx
447 movq TD_PCB
(%rcx
), %rcx
448 movq
%rax
,PCB_ONFAULT
(%rcx
)
453 * Store a 64-bit word, a 32-bit word, a 16-bit word, or an 8-bit byte to
454 * user memory. All these functions are MPSAFE.
456 * addr = %rdi, value = %rsi
462 movq PCPU
(curthread
),%rcx
463 movq TD_PCB
(%rcx
), %rcx
464 movq $fusufault
,PCB_ONFAULT
(%rcx
)
465 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
467 movq $VM_MAX_USER_ADDRESS-
8,%rax
468 cmpq
%rax
,%rdi
/* verify address validity */
473 movq PCPU
(curthread
),%rcx
474 movq TD_PCB
(%rcx
), %rcx
475 movq
%rax
,PCB_ONFAULT
(%rcx
)
482 movq PCPU
(curthread
),%rcx
483 movq TD_PCB
(%rcx
), %rcx
484 movq $fusufault
,PCB_ONFAULT
(%rcx
)
485 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
487 movq $VM_MAX_USER_ADDRESS-
4,%rax
488 cmpq
%rax
,%rdi
/* verify address validity */
493 movq PCPU
(curthread
),%rcx
494 movq TD_PCB
(%rcx
), %rcx
495 movq
%rax
,PCB_ONFAULT
(%rcx
)
499 movq PCPU
(curthread
),%rcx
500 movq TD_PCB
(%rcx
), %rcx
501 movq $fusufault
,PCB_ONFAULT
(%rcx
)
502 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
504 movq $VM_MAX_USER_ADDRESS-
2,%rax
505 cmpq
%rax
,%rdi
/* verify address validity */
510 movq PCPU
(curthread
),%rcx
/* restore trashed register */
511 movq TD_PCB
(%rcx
), %rcx
512 movq
%rax
,PCB_ONFAULT
(%rcx
)
516 movq PCPU
(curthread
),%rcx
517 movq TD_PCB
(%rcx
), %rcx
518 movq $fusufault
,PCB_ONFAULT
(%rcx
)
519 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
521 movq $VM_MAX_USER_ADDRESS-
1,%rax
522 cmpq
%rax
,%rdi
/* verify address validity */
528 movq PCPU
(curthread
),%rcx
/* restore trashed register */
529 movq TD_PCB
(%rcx
), %rcx
530 movq
%rax
,PCB_ONFAULT
(%rcx
)
534 * std_copyinstr(from, to, maxlen, int *lencopied) - MP SAFE
535 * %rdi, %rsi, %rdx, %rcx
537 * copy a string from from to to, stop when a 0 character is reached.
538 * return ENAMETOOLONG if string is longer than maxlen, and
539 * EFAULT on protection violations. If lencopied is non-zero,
540 * return the actual length in *lencopied.
543 movq
%rdx
,%r8 /* %r8 = maxlen */
544 movq
%rcx
,%r9 /* %r9 = *len */
545 xchgq
%rdi
,%rsi
/* %rdi = from, %rsi = to */
546 movq PCPU
(curthread
),%rcx
547 movq TD_PCB
(%rcx
), %rcx
548 movq $cpystrflt
,PCB_ONFAULT
(%rcx
)
549 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
551 movq $VM_MAX_USER_ADDRESS
,%rax
553 /* make sure 'from' is within bounds */
557 /* restrict maxlen to <= VM_MAX_USER_ADDRESS-from */
575 /* Success -- 0 byte reached */
580 /* rdx is zero - return ENAMETOOLONG or EFAULT */
581 movq $VM_MAX_USER_ADDRESS
,%rax
585 movq $ENAMETOOLONG
,%rax
592 /* set *lencopied and return %eax */
593 movq PCPU
(curthread
),%rcx
594 movq TD_PCB
(%rcx
), %rcx
595 movq $
0,PCB_ONFAULT
(%rcx
)
606 * copystr(from, to, maxlen, int *lencopied) - MP SAFE
607 * %rdi, %rsi, %rdx, %rcx
610 movq
%rdx
,%r8 /* %r8 = maxlen */
623 /* Success -- 0 byte reached */
628 /* rdx is zero -- return ENAMETOOLONG */
629 movq $ENAMETOOLONG
,%rax
635 /* set *lencopied and return %rax */
642 * Handling of special x86_64 registers and descriptor tables etc
645 /* void lgdt(struct region_descriptor *rdp); */
647 /* reload the descriptor table */
650 /* flush the prefetch q */
657 movl
%eax
,%fs /* Beware, use wrmsr to set 64 bit base */
658 movl
%eax
,%gs
/* Beware, use wrmsr to set 64 bit base */
661 /* reload code selector by turning return into intersegmental return */
668 /*****************************************************************************/
669 /* setjmp, longjmp */
670 /*****************************************************************************/
673 movq
%rbx
,0(%rdi
) /* save rbx */
674 movq
%rsp
,8(%rdi
) /* save rsp */
675 movq
%rbp
,16(%rdi
) /* save rbp */
676 movq
%r12,24(%rdi
) /* save r12 */
677 movq
%r13,32(%rdi
) /* save r13 */
678 movq
%r14,40(%rdi
) /* save r14 */
679 movq
%r15,48(%rdi
) /* save r15 */
680 movq
0(%rsp
),%rdx
/* get rta */
681 movq
%rdx
,56(%rdi
) /* save rip */
682 xorl
%eax
,%eax
/* return(0); */
686 movq
0(%rdi
),%rbx
/* restore rbx */
687 movq
8(%rdi
),%rsp
/* restore rsp */
688 movq
16(%rdi
),%rbp
/* restore rbp */
689 movq
24(%rdi
),%r12 /* restore r12 */
690 movq
32(%rdi
),%r13 /* restore r13 */
691 movq
40(%rdi
),%r14 /* restore r14 */
692 movq
48(%rdi
),%r15 /* restore r15 */
693 movq
56(%rdi
),%rdx
/* get rta */
694 movq
%rdx
,0(%rsp
) /* put in return frame */
695 xorl
%eax
,%eax
/* return(1); */
700 * Support for reading MSRs in the safe manner.
703 /* int rdmsr_safe(u_int msr, uint64_t *data) */
704 movq PCPU
(curthread
),%r8
705 movq TD_PCB
(%r8), %r8
706 movq $msr_onfault
,PCB_ONFAULT
(%r8)
707 movq
%rsp
,PCB_ONFAULT_SP
(%r8)
709 rdmsr
/* Read MSR pointed by %ecx. Returns
710 hi byte in edx, lo in %eax */
711 salq $
32,%rdx
/* sign-shift %rdx left */
712 movl
%eax
,%eax
/* zero-extend %eax -> %rax */
716 movq
%rax
,PCB_ONFAULT
(%r8)
720 * Support for writing MSRs in the safe manner.
723 /* int wrmsr_safe(u_int msr, uint64_t data) */
724 movq PCPU
(curthread
),%r8
725 movq TD_PCB
(%r8), %r8
726 movq $msr_onfault
,PCB_ONFAULT
(%r8)
727 movq
%rsp
,PCB_ONFAULT_SP
(%rcx
)
732 wrmsr
/* Write MSR pointed by %ecx. Accepts
733 hi byte in edx, lo in %eax. */
735 movq
%rax
,PCB_ONFAULT
(%r8)
739 * MSR operations fault handler
743 movq PCPU
(curthread
),%r8
744 movq TD_PCB
(%r8), %r8
745 movq $
0,PCB_ONFAULT
(%r8)
750 * Support for BB-profiling (gcc -a). The kernbb program will extract
751 * the data from the kernel.
761 NON_GPROF_ENTRY
(__bb_init_func
)