linux_dsm_epyc7002/arch/x86/kernel/relocate_kernel_32.S
Denys Vlasenko 3e1aa7cb59 x86/asm: Optimize unnecessarily wide TEST instructions
By the nature of the TEST operation, it is often possible to test
a narrower part of the operand:

    "testl $3,  mem"  ->  "testb $3, mem",
    "testq $3, %rcx"  ->  "testb $3, %cl"

This results in shorter instructions, because the TEST instruction
has no sign-entending byte-immediate forms unlike other ALU ops.

Note that this change does not create any LCP (Length-Changing Prefix)
stalls, which happen when adding a 0x66 prefix, which happens when
16-bit immediates are used, which changes such TEST instructions:

  [test_opcode] [modrm] [imm32]

to:

  [0x66] [test_opcode] [modrm] [imm16]

where [imm16] has a *different length* now: 2 bytes instead of 4.
This confuses the decoder and slows down execution.

REX prefixes were carefully designed to almost never hit this case:
adding REX prefix does not change instruction length except MOVABS
and MOV [addr],RAX instruction.

This patch does not add instructions which would use a 0x66 prefix,
code changes in assembly are:

    -48 f7 07 01 00 00 00 	testq  $0x1,(%rdi)
    +f6 07 01             	testb  $0x1,(%rdi)
    -48 f7 c1 01 00 00 00 	test   $0x1,%rcx
    +f6 c1 01             	test   $0x1,%cl
    -48 f7 c1 02 00 00 00 	test   $0x2,%rcx
    +f6 c1 02             	test   $0x2,%cl
    -41 f7 c2 01 00 00 00 	test   $0x1,%r10d
    +41 f6 c2 01          	test   $0x1,%r10b
    -48 f7 c1 04 00 00 00 	test   $0x4,%rcx
    +f6 c1 04             	test   $0x4,%cl
    -48 f7 c1 08 00 00 00 	test   $0x8,%rcx
    +f6 c1 08             	test   $0x8,%cl

Linus further notes:

   "There are no stalls from using 8-bit instruction forms.

    Now, changing from 64-bit or 32-bit 'test' instructions to 8-bit ones
    *could* cause problems if it ends up having forwarding issues, so that
    instead of just forwarding the result, you end up having to wait for
    it to be stable in the L1 cache (or possibly the register file). The
    forwarding from the store buffer is simplest and most reliable if the
    read is done at the exact same address and the exact same size as the
    write that gets forwarded.

    But that's true only if:

     (a) the write was very recent and is still in the write queue. I'm
         not sure that's the case here anyway.

     (b) on at least most Intel microarchitectures, you have to test a
         different byte than the lowest one (so forwarding a 64-bit write
         to a 8-bit read ends up working fine, as long as the 8-bit read
         is of the low 8 bits of the written data).

    A very similar issue *might* show up for registers too, not just
    memory writes, if you use 'testb' with a high-byte register (where
    instead of forwarding the value from the original producer it needs to
    go through the register file and then shifted). But it's mainly a
    problem for store buffers.

    But afaik, the way Denys changed the test instructions, neither of the
    above issues should be true.

    The real problem for store buffer forwarding tends to be "write 8
    bits, read 32 bits". That can be really surprisingly expensive,
    because the read ends up having to wait until the write has hit the
    cacheline, and we might talk tens of cycles of latency here. But
    "write 32 bits, read the low 8 bits" *should* be fast on pretty much
    all x86 chips, afaik."

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Acked-by: Andy Lutomirski <luto@amacapital.net>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: H. Peter Anvin <hpa@linux.intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Will Drewry <wad@chromium.org>
Link: http://lkml.kernel.org/r/1425675332-31576-1-git-send-email-dvlasenk@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-03-07 11:12:43 +01:00

278 lines
5.6 KiB
ArmAsm

/*
* relocate_kernel.S - put the kernel image in place to boot
* Copyright (C) 2002-2004 Eric Biederman <ebiederm@xmission.com>
*
* This source code is licensed under the GNU General Public License,
* Version 2. See the file COPYING for more details.
*/
#include <linux/linkage.h>
#include <asm/page_types.h>
#include <asm/kexec.h>
#include <asm/processor-flags.h>
/*
* Must be relocatable PIC code callable as a C function
*/
#define PTR(x) (x << 2)
/*
* control_page + KEXEC_CONTROL_CODE_MAX_SIZE
* ~ control_page + PAGE_SIZE are used as data storage and stack for
* jumping back
*/
#define DATA(offset) (KEXEC_CONTROL_CODE_MAX_SIZE+(offset))
/* Minimal CPU state */
#define ESP DATA(0x0)
#define CR0 DATA(0x4)
#define CR3 DATA(0x8)
#define CR4 DATA(0xc)
/* other data */
#define CP_VA_CONTROL_PAGE DATA(0x10)
#define CP_PA_PGD DATA(0x14)
#define CP_PA_SWAP_PAGE DATA(0x18)
#define CP_PA_BACKUP_PAGES_MAP DATA(0x1c)
.text
.globl relocate_kernel
relocate_kernel:
/* Save the CPU context, used for jumping back */
pushl %ebx
pushl %esi
pushl %edi
pushl %ebp
pushf
movl 20+8(%esp), %ebp /* list of pages */
movl PTR(VA_CONTROL_PAGE)(%ebp), %edi
movl %esp, ESP(%edi)
movl %cr0, %eax
movl %eax, CR0(%edi)
movl %cr3, %eax
movl %eax, CR3(%edi)
movl %cr4, %eax
movl %eax, CR4(%edi)
/* read the arguments and say goodbye to the stack */
movl 20+4(%esp), %ebx /* page_list */
movl 20+8(%esp), %ebp /* list of pages */
movl 20+12(%esp), %edx /* start address */
movl 20+16(%esp), %ecx /* cpu_has_pae */
movl 20+20(%esp), %esi /* preserve_context */
/* zero out flags, and disable interrupts */
pushl $0
popfl
/* save some information for jumping back */
movl PTR(VA_CONTROL_PAGE)(%ebp), %edi
movl %edi, CP_VA_CONTROL_PAGE(%edi)
movl PTR(PA_PGD)(%ebp), %eax
movl %eax, CP_PA_PGD(%edi)
movl PTR(PA_SWAP_PAGE)(%ebp), %eax
movl %eax, CP_PA_SWAP_PAGE(%edi)
movl %ebx, CP_PA_BACKUP_PAGES_MAP(%edi)
/*
* get physical address of control page now
* this is impossible after page table switch
*/
movl PTR(PA_CONTROL_PAGE)(%ebp), %edi
/* switch to new set of page tables */
movl PTR(PA_PGD)(%ebp), %eax
movl %eax, %cr3
/* setup a new stack at the end of the physical control page */
lea PAGE_SIZE(%edi), %esp
/* jump to identity mapped page */
movl %edi, %eax
addl $(identity_mapped - relocate_kernel), %eax
pushl %eax
ret
identity_mapped:
/* set return address to 0 if not preserving context */
pushl $0
/* store the start address on the stack */
pushl %edx
/*
* Set cr0 to a known state:
* - Paging disabled
* - Alignment check disabled
* - Write protect disabled
* - No task switch
* - Don't do FP software emulation.
* - Proctected mode enabled
*/
movl %cr0, %eax
andl $~(X86_CR0_PG | X86_CR0_AM | X86_CR0_WP | X86_CR0_TS | X86_CR0_EM), %eax
orl $(X86_CR0_PE), %eax
movl %eax, %cr0
/* clear cr4 if applicable */
testl %ecx, %ecx
jz 1f
/*
* Set cr4 to a known state:
* Setting everything to zero seems safe.
*/
xorl %eax, %eax
movl %eax, %cr4
jmp 1f
1:
/* Flush the TLB (needed?) */
xorl %eax, %eax
movl %eax, %cr3
movl CP_PA_SWAP_PAGE(%edi), %eax
pushl %eax
pushl %ebx
call swap_pages
addl $8, %esp
/*
* To be certain of avoiding problems with self-modifying code
* I need to execute a serializing instruction here.
* So I flush the TLB, it's handy, and not processor dependent.
*/
xorl %eax, %eax
movl %eax, %cr3
/*
* set all of the registers to known values
* leave %esp alone
*/
testl %esi, %esi
jnz 1f
xorl %edi, %edi
xorl %eax, %eax
xorl %ebx, %ebx
xorl %ecx, %ecx
xorl %edx, %edx
xorl %esi, %esi
xorl %ebp, %ebp
ret
1:
popl %edx
movl CP_PA_SWAP_PAGE(%edi), %esp
addl $PAGE_SIZE, %esp
2:
call *%edx
/* get the re-entry point of the peer system */
movl 0(%esp), %ebp
call 1f
1:
popl %ebx
subl $(1b - relocate_kernel), %ebx
movl CP_VA_CONTROL_PAGE(%ebx), %edi
lea PAGE_SIZE(%ebx), %esp
movl CP_PA_SWAP_PAGE(%ebx), %eax
movl CP_PA_BACKUP_PAGES_MAP(%ebx), %edx
pushl %eax
pushl %edx
call swap_pages
addl $8, %esp
movl CP_PA_PGD(%ebx), %eax
movl %eax, %cr3
movl %cr0, %eax
orl $X86_CR0_PG, %eax
movl %eax, %cr0
lea PAGE_SIZE(%edi), %esp
movl %edi, %eax
addl $(virtual_mapped - relocate_kernel), %eax
pushl %eax
ret
virtual_mapped:
movl CR4(%edi), %eax
movl %eax, %cr4
movl CR3(%edi), %eax
movl %eax, %cr3
movl CR0(%edi), %eax
movl %eax, %cr0
movl ESP(%edi), %esp
movl %ebp, %eax
popf
popl %ebp
popl %edi
popl %esi
popl %ebx
ret
/* Do the copies */
swap_pages:
movl 8(%esp), %edx
movl 4(%esp), %ecx
pushl %ebp
pushl %ebx
pushl %edi
pushl %esi
movl %ecx, %ebx
jmp 1f
0: /* top, read another word from the indirection page */
movl (%ebx), %ecx
addl $4, %ebx
1:
testb $0x1, %cl /* is it a destination page */
jz 2f
movl %ecx, %edi
andl $0xfffff000, %edi
jmp 0b
2:
testb $0x2, %cl /* is it an indirection page */
jz 2f
movl %ecx, %ebx
andl $0xfffff000, %ebx
jmp 0b
2:
testb $0x4, %cl /* is it the done indicator */
jz 2f
jmp 3f
2:
testb $0x8, %cl /* is it the source indicator */
jz 0b /* Ignore it otherwise */
movl %ecx, %esi /* For every source page do a copy */
andl $0xfffff000, %esi
movl %edi, %eax
movl %esi, %ebp
movl %edx, %edi
movl $1024, %ecx
rep ; movsl
movl %ebp, %edi
movl %eax, %esi
movl $1024, %ecx
rep ; movsl
movl %eax, %edi
movl %edx, %esi
movl $1024, %ecx
rep ; movsl
lea PAGE_SIZE(%ebp), %esi
jmp 0b
3:
popl %esi
popl %edi
popl %ebx
popl %ebp
ret
.globl kexec_control_code_size
.set kexec_control_code_size, . - relocate_kernel