freebsd-skq/sys/amd64/include/cpufunc.h

1048 lines
19 KiB
C
Raw Normal View History

1994-08-02 07:55:43 +00:00
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
2003-11-08 04:39:22 +00:00
* Copyright (c) 2003 Peter Wemm.
1994-08-02 07:55:43 +00:00
* Copyright (c) 1993 The Regents of the University of California.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
1994-08-02 07:55:43 +00:00
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
1999-08-28 01:08:13 +00:00
* $FreeBSD$
1994-08-02 07:55:43 +00:00
*/
1993-06-12 14:58:17 +00:00
/*
* Functions to provide access to special i386 instructions.
* This in included in sys/systm.h, and that file should be
* used in preference to this.
1993-06-12 14:58:17 +00:00
*/
#ifndef _MACHINE_CPUFUNC_H_
#define _MACHINE_CPUFUNC_H_
#ifndef _SYS_CDEFS_H_
#error this file needs sys/cdefs.h as a prerequisite
#endif
struct region_descriptor;
#define readb(va) (*(volatile uint8_t *) (va))
#define readw(va) (*(volatile uint16_t *) (va))
#define readl(va) (*(volatile uint32_t *) (va))
#define readq(va) (*(volatile uint64_t *) (va))
#define writeb(va, d) (*(volatile uint8_t *) (va) = (d))
#define writew(va, d) (*(volatile uint16_t *) (va) = (d))
#define writel(va, d) (*(volatile uint32_t *) (va) = (d))
#define writeq(va, d) (*(volatile uint64_t *) (va) = (d))
#if defined(__GNUCLIKE_ASM) && defined(__CC_SUPPORTS___INLINE)
1993-06-12 14:58:17 +00:00
static __inline void
breakpoint(void)
{
__asm __volatile("int $3");
}
static __inline __pure2 u_int
bsfl(u_int mask)
{
u_int result;
__asm __volatile("bsfl %1,%0" : "=r" (result) : "rm" (mask));
return (result);
}
static __inline __pure2 u_long
bsfq(u_long mask)
{
u_long result;
__asm __volatile("bsfq %1,%0" : "=r" (result) : "rm" (mask));
return (result);
}
static __inline __pure2 u_int
bsrl(u_int mask)
{
u_int result;
__asm __volatile("bsrl %1,%0" : "=r" (result) : "rm" (mask));
return (result);
}
static __inline __pure2 u_long
bsrq(u_long mask)
{
u_long result;
__asm __volatile("bsrq %1,%0" : "=r" (result) : "rm" (mask));
return (result);
}
static __inline void
clflush(u_long addr)
{
__asm __volatile("clflush %0" : : "m" (*(char *)addr));
}
static __inline void
clflushopt(u_long addr)
{
__asm __volatile(".byte 0x66;clflush %0" : : "m" (*(char *)addr));
}
static __inline void
clwb(u_long addr)
{
__asm __volatile("clwb %0" : : "m" (*(char *)addr));
}
static __inline void
clts(void)
{
__asm __volatile("clts");
}
static __inline void
disable_intr(void)
{
__asm __volatile("cli" : : : "memory");
}
static __inline void
do_cpuid(u_int ax, u_int *p)
{
__asm __volatile("cpuid"
: "=a" (p[0]), "=b" (p[1]), "=c" (p[2]), "=d" (p[3])
: "0" (ax));
}
static __inline void
cpuid_count(u_int ax, u_int cx, u_int *p)
{
__asm __volatile("cpuid"
: "=a" (p[0]), "=b" (p[1]), "=c" (p[2]), "=d" (p[3])
: "0" (ax), "c" (cx));
}
static __inline void
enable_intr(void)
{
__asm __volatile("sti");
}
#ifdef _KERNEL
#define HAVE_INLINE_FFS
#define ffs(x) __builtin_ffs(x)
#define HAVE_INLINE_FFSL
static __inline __pure2 int
ffsl(long mask)
{
return (mask == 0 ? mask : (int)bsfq((u_long)mask) + 1);
}
#define HAVE_INLINE_FFSLL
static __inline __pure2 int
ffsll(long long mask)
{
return (ffsl((long)mask));
}
#define HAVE_INLINE_FLS
static __inline __pure2 int
fls(int mask)
{
return (mask == 0 ? mask : (int)bsrl((u_int)mask) + 1);
}
#define HAVE_INLINE_FLSL
static __inline __pure2 int
flsl(long mask)
{
return (mask == 0 ? mask : (int)bsrq((u_long)mask) + 1);
}
#define HAVE_INLINE_FLSLL
static __inline __pure2 int
flsll(long long mask)
{
return (flsl((long)mask));
}
#endif /* _KERNEL */
static __inline void
halt(void)
{
__asm __volatile("hlt");
}
static __inline u_char
inb(u_int port)
{
u_char data;
__asm __volatile("inb %w1, %0" : "=a" (data) : "Nd" (port));
return (data);
}
static __inline u_int
inl(u_int port)
{
u_int data;
__asm __volatile("inl %w1, %0" : "=a" (data) : "Nd" (port));
return (data);
}
static __inline void
insb(u_int port, void *addr, size_t count)
{
__asm __volatile("cld; rep; insb"
: "+D" (addr), "+c" (count)
: "d" (port)
: "memory");
}
static __inline void
insw(u_int port, void *addr, size_t count)
{
__asm __volatile("cld; rep; insw"
: "+D" (addr), "+c" (count)
: "d" (port)
: "memory");
}
static __inline void
insl(u_int port, void *addr, size_t count)
{
__asm __volatile("cld; rep; insl"
: "+D" (addr), "+c" (count)
: "d" (port)
: "memory");
}
static __inline void
invd(void)
{
__asm __volatile("invd");
}
static __inline u_short
inw(u_int port)
{
u_short data;
__asm __volatile("inw %w1, %0" : "=a" (data) : "Nd" (port));
return (data);
}
static __inline void
outb(u_int port, u_char data)
1993-06-12 14:58:17 +00:00
{
__asm __volatile("outb %0, %w1" : : "a" (data), "Nd" (port));
}
1993-06-12 14:58:17 +00:00
static __inline void
outl(u_int port, u_int data)
{
__asm __volatile("outl %0, %w1" : : "a" (data), "Nd" (port));
1993-06-12 14:58:17 +00:00
}
static __inline void
outsb(u_int port, const void *addr, size_t count)
1993-06-12 14:58:17 +00:00
{
__asm __volatile("cld; rep; outsb"
: "+S" (addr), "+c" (count)
: "d" (port));
1993-06-12 14:58:17 +00:00
}
static __inline void
outsw(u_int port, const void *addr, size_t count)
1993-06-12 14:58:17 +00:00
{
__asm __volatile("cld; rep; outsw"
: "+S" (addr), "+c" (count)
: "d" (port));
1993-06-12 14:58:17 +00:00
}
static __inline void
outsl(u_int port, const void *addr, size_t count)
1993-06-12 14:58:17 +00:00
{
__asm __volatile("cld; rep; outsl"
: "+S" (addr), "+c" (count)
: "d" (port));
1993-06-12 14:58:17 +00:00
}
static __inline void
outw(u_int port, u_short data)
1993-06-12 14:58:17 +00:00
{
__asm __volatile("outw %0, %w1" : : "a" (data), "Nd" (port));
1993-06-12 14:58:17 +00:00
}
static __inline u_long
popcntq(u_long mask)
{
u_long result;
__asm __volatile("popcntq %1,%0" : "=r" (result) : "rm" (mask));
return (result);
}
2012-08-01 17:24:53 +00:00
static __inline void
lfence(void)
{
__asm __volatile("lfence" : : : "memory");
}
static __inline void
mfence(void)
{
__asm __volatile("mfence" : : : "memory");
}
static __inline void
sfence(void)
{
__asm __volatile("sfence" : : : "memory");
}
static __inline void
ia32_pause(void)
{
__asm __volatile("pause");
}
static __inline u_long
read_rflags(void)
{
u_long rf;
__asm __volatile("pushfq; popq %0" : "=r" (rf));
return (rf);
}
static __inline uint64_t
rdmsr(u_int msr)
{
uint32_t low, high;
__asm __volatile("rdmsr" : "=a" (low), "=d" (high) : "c" (msr));
return (low | ((uint64_t)high << 32));
}
static __inline uint32_t
rdmsr32(u_int msr)
{
uint32_t low;
__asm __volatile("rdmsr" : "=a" (low) : "c" (msr) : "rdx");
return (low);
}
static __inline uint64_t
rdpmc(u_int pmc)
{
uint32_t low, high;
__asm __volatile("rdpmc" : "=a" (low), "=d" (high) : "c" (pmc));
return (low | ((uint64_t)high << 32));
}
static __inline uint64_t
rdtsc(void)
{
uint32_t low, high;
__asm __volatile("rdtsc" : "=a" (low), "=d" (high));
return (low | ((uint64_t)high << 32));
}
2018-06-07 00:54:11 +00:00
static __inline uint64_t
rdtscp(void)
{
uint32_t low, high;
__asm __volatile("rdtscp" : "=a" (low), "=d" (high) : : "ecx");
2018-06-07 00:54:11 +00:00
return (low | ((uint64_t)high << 32));
}
static __inline uint32_t
rdtsc32(void)
{
uint32_t rv;
__asm __volatile("rdtsc" : "=a" (rv) : : "edx");
return (rv);
}
static __inline void
wbinvd(void)
{
__asm __volatile("wbinvd");
}
static __inline void
write_rflags(u_long rf)
{
__asm __volatile("pushq %0; popfq" : : "r" (rf));
}
static __inline void
wrmsr(u_int msr, uint64_t newval)
{
uint32_t low, high;
low = newval;
high = newval >> 32;
__asm __volatile("wrmsr" : : "a" (low), "d" (high), "c" (msr));
}
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
static __inline void
load_cr0(u_long data)
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
{
__asm __volatile("movq %0,%%cr0" : : "r" (data));
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
}
static __inline u_long
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
rcr0(void)
{
u_long data;
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
__asm __volatile("movq %%cr0,%0" : "=r" (data));
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
return (data);
}
static __inline u_long
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
rcr2(void)
{
u_long data;
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
__asm __volatile("movq %%cr2,%0" : "=r" (data));
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
return (data);
}
static __inline void
load_cr3(u_long data)
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
{
__asm __volatile("movq %0,%%cr3" : : "r" (data) : "memory");
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
}
static __inline u_long
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
rcr3(void)
{
u_long data;
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
__asm __volatile("movq %%cr3,%0" : "=r" (data));
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
return (data);
}
static __inline void
load_cr4(u_long data)
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
{
__asm __volatile("movq %0,%%cr4" : : "r" (data));
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
}
static __inline u_long
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
rcr4(void)
{
u_long data;
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
__asm __volatile("movq %%cr4,%0" : "=r" (data));
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
return (data);
}
static __inline u_long
rxcr(u_int reg)
{
u_int low, high;
__asm __volatile("xgetbv" : "=a" (low), "=d" (high) : "c" (reg));
return (low | ((uint64_t)high << 32));
}
static __inline void
load_xcr(u_int reg, u_long val)
{
u_int low, high;
low = val;
high = val >> 32;
__asm __volatile("xsetbv" : : "c" (reg), "a" (low), "d" (high));
}
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
/*
* Global TLB flush (except for thise for pages marked PG_G)
*/
static __inline void
invltlb(void)
{
load_cr3(rcr3());
}
Implement PV IPIs for PVHVM guests and further converge PV and HVM IPI implmementations. Submitted by: Roger Pau Monné Sponsored by: Citrix Systems R&D Submitted by: gibbs (misc cleanup, table driven config) Reviewed by: gibbs MFC after: 2 weeks sys/amd64/include/cpufunc.h: sys/amd64/amd64/pmap.c: Move invltlb_globpcid() into cpufunc.h so that it can be used by the Xen HVM version of tlb shootdown IPI handlers. sys/x86/xen/xen_intr.c: sys/xen/xen_intr.h: Rename xen_intr_bind_ipi() to xen_intr_alloc_and_bind_ipi(), and remove the ipi vector parameter. This api allocates an event channel port that can be used for ipi services, but knows nothing of the actual ipi for which that port will be used. Removing the unused argument and cleaning up the comments surrounding its declaration helps clarify its actual role. sys/amd64/amd64/mp_machdep.c: sys/amd64/include/cpu.h: sys/i386/i386/mp_machdep.c: sys/i386/include/cpu.h: Implement a generic framework for amd64 and i386 that allows the implementation of certain CPU management functions to be selected at runtime. Currently this is only used for the ipi send function, which we optimize for Xen when running on a Xen hypervisor, but can easily be expanded to support more operations. sys/x86/xen/hvm.c: Implement Xen PV IPI handlers and operations, replacing native send IPI. sys/amd64/include/pcpu.h: sys/i386/include/pcpu.h: sys/i386/include/smp.h: Remove NR_VIRQS and NR_IPIS from FreeBSD headers. NR_VIRQS is defined already for us in the xen interface files. NR_IPIS is only needed in one file per Xen platform and is easily inferred by the IPI vector table that is defined in those files. sys/i386/xen/mp_machdep.c: Restructure to more closely match the HVM implementation by performing table driven IPI setup.
2013-09-06 22:17:02 +00:00
#ifndef CR4_PGE
#define CR4_PGE 0x00000080 /* Page global enable */
#endif
/*
* Perform the guaranteed invalidation of all TLB entries. This
* includes the global entries, and entries in all PCIDs, not only the
* current context. The function works both on non-PCID CPUs and CPUs
* with the PCID turned off or on. See IA-32 SDM Vol. 3a 4.10.4.1
* Operations that Invalidate TLBs and Paging-Structure Caches.
*/
static __inline void
invltlb_glob(void)
Implement PV IPIs for PVHVM guests and further converge PV and HVM IPI implmementations. Submitted by: Roger Pau Monné Sponsored by: Citrix Systems R&D Submitted by: gibbs (misc cleanup, table driven config) Reviewed by: gibbs MFC after: 2 weeks sys/amd64/include/cpufunc.h: sys/amd64/amd64/pmap.c: Move invltlb_globpcid() into cpufunc.h so that it can be used by the Xen HVM version of tlb shootdown IPI handlers. sys/x86/xen/xen_intr.c: sys/xen/xen_intr.h: Rename xen_intr_bind_ipi() to xen_intr_alloc_and_bind_ipi(), and remove the ipi vector parameter. This api allocates an event channel port that can be used for ipi services, but knows nothing of the actual ipi for which that port will be used. Removing the unused argument and cleaning up the comments surrounding its declaration helps clarify its actual role. sys/amd64/amd64/mp_machdep.c: sys/amd64/include/cpu.h: sys/i386/i386/mp_machdep.c: sys/i386/include/cpu.h: Implement a generic framework for amd64 and i386 that allows the implementation of certain CPU management functions to be selected at runtime. Currently this is only used for the ipi send function, which we optimize for Xen when running on a Xen hypervisor, but can easily be expanded to support more operations. sys/x86/xen/hvm.c: Implement Xen PV IPI handlers and operations, replacing native send IPI. sys/amd64/include/pcpu.h: sys/i386/include/pcpu.h: sys/i386/include/smp.h: Remove NR_VIRQS and NR_IPIS from FreeBSD headers. NR_VIRQS is defined already for us in the xen interface files. NR_IPIS is only needed in one file per Xen platform and is easily inferred by the IPI vector table that is defined in those files. sys/i386/xen/mp_machdep.c: Restructure to more closely match the HVM implementation by performing table driven IPI setup.
2013-09-06 22:17:02 +00:00
{
uint64_t cr4;
cr4 = rcr4();
load_cr4(cr4 & ~CR4_PGE);
/*
* Although preemption at this point could be detrimental to
* performance, it would not lead to an error. PG_G is simply
* ignored if CR4.PGE is clear. Moreover, in case this block
* is re-entered, the load_cr4() either above or below will
* modify CR4.PGE flushing the TLB.
*/
load_cr4(cr4 | CR4_PGE);
}
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
/*
* TLB flush for an individual page (even if it has PG_G).
* Only works on 486+ CPUs (i386 does not have PG_G).
*/
static __inline void
invlpg(u_long addr)
Revive backed out pmap related changes from Feb 2002. The highlights are: - It actually works this time, honest! - Fine grained TLB shootdowns for SMP on i386. IPI's are very expensive, so try and optimize things where possible. - Introduce ranged shootdowns that can be done as a single IPI. - PG_G support for i386 - Specific-cpu targeted shootdowns. For example, there is no sense in globally purging the TLB cache for where we are stealing a page from the local unshared process on the local cpu. Use pm_active to track this. - Add some instrumentation for the tlb shootdown code. - Rip out SMP code from <machine/cpufunc.h> - Try and fix some very bogus PG_G and PG_PS interactions that were bad enough to cause vm86 bios calls to break. vm86 depended on our existing bugs and this was the cause of the VESA panics last time. - Fix the silly one-line error that caused the 'panic: bad pte' last time. - Fix a couple of other silly one-line errors that should have caused more pain than they did. Some more work is needed: - pmap_{zero,copy}_page[_idle]. These can be done without IPI's if we have a hook in cpu_switch. - The IPI handlers need some cleanup. I have a bogus %ds load that can be avoided. - APTD handling is rather bogus and appears to be a large source of global TLB IPI shootdowns for no really good reason. I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop. I expect to see a bigger difference when there is significant pageout activity or the system otherwise has memory shortages. I have backed out a few optimizations that I had been using over the last few days in order to be a little more conservative. I'll revisit these again over the next few days as the dust settles. New option: DISABLE_PG_G - In case I missed something.
2002-07-12 07:56:11 +00:00
{
__asm __volatile("invlpg %0" : : "m" (*(char *)addr) : "memory");
}
#define INVPCID_ADDR 0
#define INVPCID_CTX 1
#define INVPCID_CTXGLOB 2
#define INVPCID_ALLCTX 3
struct invpcid_descr {
uint64_t pcid:12 __packed;
uint64_t pad:52 __packed;
uint64_t addr;
} __packed;
static __inline void
invpcid(struct invpcid_descr *d, int type)
{
__asm __volatile("invpcid (%0),%1"
: : "r" (d), "r" ((u_long)type) : "memory");
}
static __inline u_short
rfs(void)
{
u_short sel;
__asm __volatile("movw %%fs,%0" : "=rm" (sel));
return (sel);
}
static __inline u_short
rgs(void)
{
u_short sel;
__asm __volatile("movw %%gs,%0" : "=rm" (sel));
return (sel);
}
static __inline u_short
rss(void)
{
u_short sel;
__asm __volatile("movw %%ss,%0" : "=rm" (sel));
return (sel);
}
static __inline void
load_ds(u_short sel)
{
__asm __volatile("movw %0,%%ds" : : "rm" (sel));
}
static __inline void
load_es(u_short sel)
{
__asm __volatile("movw %0,%%es" : : "rm" (sel));
}
static __inline void
cpu_monitor(const void *addr, u_long extensions, u_int hints)
{
__asm __volatile("monitor"
: : "a" (addr), "c" (extensions), "d" (hints));
}
static __inline void
cpu_mwait(u_long extensions, u_int hints)
{
__asm __volatile("mwait" : : "a" (hints), "c" (extensions));
}
static __inline uint32_t
rdpkru(void)
{
uint32_t res;
__asm __volatile("rdpkru" : "=a" (res) : "c" (0) : "edx");
return (res);
}
static __inline void
wrpkru(uint32_t mask)
{
__asm __volatile("wrpkru" : : "a" (mask), "c" (0), "d" (0));
}
#ifdef _KERNEL
/* This is defined in <machine/specialreg.h> but is too painful to get to */
#ifndef MSR_FSBASE
#define MSR_FSBASE 0xc0000100
#endif
static __inline void
load_fs(u_short sel)
{
/* Preserve the fsbase value across the selector load */
__asm __volatile("rdmsr; movw %0,%%fs; wrmsr"
: : "rm" (sel), "c" (MSR_FSBASE) : "eax", "edx");
}
#ifndef MSR_GSBASE
#define MSR_GSBASE 0xc0000101
#endif
static __inline void
load_gs(u_short sel)
{
/*
* Preserve the gsbase value across the selector load.
* Note that we have to disable interrupts because the gsbase
* being trashed happens to be the kernel gsbase at the time.
*/
__asm __volatile("pushfq; cli; rdmsr; movw %0,%%gs; wrmsr; popfq"
: : "rm" (sel), "c" (MSR_GSBASE) : "eax", "edx");
}
#else
/* Usable by userland */
static __inline void
load_fs(u_short sel)
{
__asm __volatile("movw %0,%%fs" : : "rm" (sel));
}
static __inline void
load_gs(u_short sel)
{
__asm __volatile("movw %0,%%gs" : : "rm" (sel));
}
#endif
static __inline uint64_t
rdfsbase(void)
{
uint64_t x;
__asm __volatile("rdfsbase %0" : "=r" (x));
return (x);
}
static __inline void
wrfsbase(uint64_t x)
{
__asm __volatile("wrfsbase %0" : : "r" (x));
}
static __inline uint64_t
rdgsbase(void)
{
uint64_t x;
__asm __volatile("rdgsbase %0" : "=r" (x));
return (x);
}
static __inline void
wrgsbase(uint64_t x)
{
__asm __volatile("wrgsbase %0" : : "r" (x));
}
static __inline void
bare_lgdt(struct region_descriptor *addr)
{
__asm __volatile("lgdt (%0)" : : "r" (addr));
}
static __inline void
sgdt(struct region_descriptor *addr)
{
char *loc;
loc = (char *)addr;
__asm __volatile("sgdt %0" : "=m" (*loc) : : "memory");
}
static __inline void
lidt(struct region_descriptor *addr)
{
__asm __volatile("lidt (%0)" : : "r" (addr));
}
static __inline void
sidt(struct region_descriptor *addr)
{
char *loc;
loc = (char *)addr;
__asm __volatile("sidt %0" : "=m" (*loc) : : "memory");
}
static __inline void
lldt(u_short sel)
{
__asm __volatile("lldt %0" : : "r" (sel));
}
static __inline u_short
sldt(void)
{
u_short sel;
__asm __volatile("sldt %0" : "=r" (sel));
return (sel);
}
static __inline void
ltr(u_short sel)
{
__asm __volatile("ltr %0" : : "r" (sel));
}
static __inline uint32_t
read_tr(void)
{
u_short sel;
__asm __volatile("str %0" : "=r" (sel));
return (sel);
}
static __inline uint64_t
2004-01-28 23:53:04 +00:00
rdr0(void)
{
uint64_t data;
2004-01-28 23:53:04 +00:00
__asm __volatile("movq %%dr0,%0" : "=r" (data));
return (data);
}
static __inline void
load_dr0(uint64_t dr0)
2004-01-28 23:53:04 +00:00
{
__asm __volatile("movq %0,%%dr0" : : "r" (dr0));
}
static __inline uint64_t
2004-01-28 23:53:04 +00:00
rdr1(void)
{
uint64_t data;
2004-01-28 23:53:04 +00:00
__asm __volatile("movq %%dr1,%0" : "=r" (data));
return (data);
}
static __inline void
load_dr1(uint64_t dr1)
2004-01-28 23:53:04 +00:00
{
__asm __volatile("movq %0,%%dr1" : : "r" (dr1));
}
static __inline uint64_t
2004-01-28 23:53:04 +00:00
rdr2(void)
{
uint64_t data;
2004-01-28 23:53:04 +00:00
__asm __volatile("movq %%dr2,%0" : "=r" (data));
return (data);
}
static __inline void
load_dr2(uint64_t dr2)
2004-01-28 23:53:04 +00:00
{
__asm __volatile("movq %0,%%dr2" : : "r" (dr2));
}
static __inline uint64_t
2004-01-28 23:53:04 +00:00
rdr3(void)
{
uint64_t data;
2004-01-28 23:53:04 +00:00
__asm __volatile("movq %%dr3,%0" : "=r" (data));
return (data);
}
static __inline void
load_dr3(uint64_t dr3)
2004-01-28 23:53:04 +00:00
{
__asm __volatile("movq %0,%%dr3" : : "r" (dr3));
}
static __inline uint64_t
2004-01-28 23:53:04 +00:00
rdr6(void)
{
uint64_t data;
2004-01-28 23:53:04 +00:00
__asm __volatile("movq %%dr6,%0" : "=r" (data));
return (data);
}
static __inline void
load_dr6(uint64_t dr6)
2004-01-28 23:53:04 +00:00
{
__asm __volatile("movq %0,%%dr6" : : "r" (dr6));
}
static __inline uint64_t
2004-01-28 23:53:04 +00:00
rdr7(void)
{
uint64_t data;
2004-01-28 23:53:04 +00:00
__asm __volatile("movq %%dr7,%0" : "=r" (data));
return (data);
}
static __inline void
load_dr7(uint64_t dr7)
2004-01-28 23:53:04 +00:00
{
__asm __volatile("movq %0,%%dr7" : : "r" (dr7));
}
static __inline register_t
intr_disable(void)
{
register_t rflags;
rflags = read_rflags();
disable_intr();
return (rflags);
}
static __inline void
intr_restore(register_t rflags)
{
write_rflags(rflags);
}
static __inline void
stac(void)
{
__asm __volatile("stac" : : : "cc");
}
static __inline void
clac(void)
{
__asm __volatile("clac" : : : "cc");
}
enum {
SGX_ECREATE = 0x0,
SGX_EADD = 0x1,
SGX_EINIT = 0x2,
SGX_EREMOVE = 0x3,
SGX_EDGBRD = 0x4,
SGX_EDGBWR = 0x5,
SGX_EEXTEND = 0x6,
SGX_ELDU = 0x8,
SGX_EBLOCK = 0x9,
SGX_EPA = 0xA,
SGX_EWB = 0xB,
SGX_ETRACK = 0xC,
};
enum {
SGX_PT_SECS = 0x00,
SGX_PT_TCS = 0x01,
SGX_PT_REG = 0x02,
SGX_PT_VA = 0x03,
SGX_PT_TRIM = 0x04,
};
int sgx_encls(uint32_t eax, uint64_t rbx, uint64_t rcx, uint64_t rdx);
static __inline int
sgx_ecreate(void *pginfo, void *secs)
{
return (sgx_encls(SGX_ECREATE, (uint64_t)pginfo,
(uint64_t)secs, 0));
}
static __inline int
sgx_eadd(void *pginfo, void *epc)
{
return (sgx_encls(SGX_EADD, (uint64_t)pginfo,
(uint64_t)epc, 0));
}
static __inline int
sgx_einit(void *sigstruct, void *secs, void *einittoken)
{
return (sgx_encls(SGX_EINIT, (uint64_t)sigstruct,
(uint64_t)secs, (uint64_t)einittoken));
}
static __inline int
sgx_eextend(void *secs, void *epc)
{
return (sgx_encls(SGX_EEXTEND, (uint64_t)secs,
(uint64_t)epc, 0));
}
static __inline int
sgx_epa(void *epc)
{
return (sgx_encls(SGX_EPA, SGX_PT_VA, (uint64_t)epc, 0));
}
static __inline int
sgx_eldu(uint64_t rbx, uint64_t rcx,
uint64_t rdx)
{
return (sgx_encls(SGX_ELDU, rbx, rcx, rdx));
}
static __inline int
sgx_eremove(void *epc)
{
return (sgx_encls(SGX_EREMOVE, 0, (uint64_t)epc, 0));
}
#else /* !(__GNUCLIKE_ASM && __CC_SUPPORTS___INLINE) */
1993-06-12 14:58:17 +00:00
int breakpoint(void);
u_int bsfl(u_int mask);
u_int bsrl(u_int mask);
void clflush(u_long addr);
void clts(void);
void cpuid_count(u_int ax, u_int cx, u_int *p);
void disable_intr(void);
void do_cpuid(u_int ax, u_int *p);
void enable_intr(void);
void halt(void);
void ia32_pause(void);
u_char inb(u_int port);
u_int inl(u_int port);
void insb(u_int port, void *addr, size_t count);
void insl(u_int port, void *addr, size_t count);
void insw(u_int port, void *addr, size_t count);
register_t intr_disable(void);
void intr_restore(register_t rf);
void invd(void);
void invlpg(u_int addr);
void invltlb(void);
u_short inw(u_int port);
void lidt(struct region_descriptor *addr);
void lldt(u_short sel);
void load_cr0(u_long cr0);
void load_cr3(u_long cr3);
void load_cr4(u_long cr4);
void load_dr0(uint64_t dr0);
void load_dr1(uint64_t dr1);
void load_dr2(uint64_t dr2);
void load_dr3(uint64_t dr3);
void load_dr6(uint64_t dr6);
void load_dr7(uint64_t dr7);
void load_fs(u_short sel);
void load_gs(u_short sel);
void ltr(u_short sel);
void outb(u_int port, u_char data);
void outl(u_int port, u_int data);
void outsb(u_int port, const void *addr, size_t count);
void outsl(u_int port, const void *addr, size_t count);
void outsw(u_int port, const void *addr, size_t count);
void outw(u_int port, u_short data);
u_long rcr0(void);
u_long rcr2(void);
u_long rcr3(void);
u_long rcr4(void);
uint64_t rdmsr(u_int msr);
uint32_t rdmsr32(u_int msr);
uint64_t rdpmc(u_int pmc);
uint64_t rdr0(void);
uint64_t rdr1(void);
uint64_t rdr2(void);
uint64_t rdr3(void);
uint64_t rdr6(void);
uint64_t rdr7(void);
uint64_t rdtsc(void);
u_long read_rflags(void);
u_int rfs(void);
u_int rgs(void);
void wbinvd(void);
void write_rflags(u_int rf);
void wrmsr(u_int msr, uint64_t newval);
#endif /* __GNUCLIKE_ASM && __CC_SUPPORTS___INLINE */
1993-06-12 14:58:17 +00:00
2004-01-28 23:53:04 +00:00
void reset_dbregs(void);
#ifdef _KERNEL
int rdmsr_safe(u_int msr, uint64_t *val);
int wrmsr_safe(u_int msr, uint64_t newval);
#endif
#endif /* !_MACHINE_CPUFUNC_H_ */