* [PATCH 2/2] x86/entry/64: use xorl for register clearing
2018-02-14 17:59 [PATCH 0/2] x86/entry: xorq->xorl; idtentry size reduction Dominik Brodowski
@ 2018-02-14 17:59 ` Dominik Brodowski
0 siblings, 0 replies; 3+ messages in thread
From: Dominik Brodowski @ 2018-02-14 17:59 UTC (permalink / raw)
To: linux-kernel, mingo, x86, torvalds; +Cc: luto, ak, tglx, dan.j.williams
Using xorq to clear general-purpose registers is slower than
xorl on some architectures. As xorl is sufficient to clear all
64bit of these registers,[*] switch the x86 64-bit entry code
to use xorl.
[*] According to Intel 64 and IA-32 Architecture Software Developer's
Manual, section 3.4.1.1, the result of 32-bit operands are "zero-
extended to a 64-bit result in the destination general-purpose
register." The AMD64 Architecture Programmer’s Manual Volume 3,
Appendix B.1, describes the same behaviour.
Suggested-by: Denys Vlasenko <dvlasenk@redhat.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
arch/x86/entry/calling.h | 16 ++++++------
arch/x86/entry/entry_64_compat.S | 54 ++++++++++++++++++++--------------------
2 files changed, 35 insertions(+), 35 deletions(-)
diff --git a/arch/x86/entry/calling.h b/arch/x86/entry/calling.h
index 79ead48e6fe1..adaf5fd9840d 100644
--- a/arch/x86/entry/calling.h
+++ b/arch/x86/entry/calling.h
@@ -117,25 +117,25 @@ For 32-bit we have the following conventions - kernel is built with
pushq %rcx /* pt_regs->cx */
pushq \rax /* pt_regs->ax */
pushq %r8 /* pt_regs->r8 */
- xorq %r8, %r8 /* nospec r8 */
+ xorl %r8d, %r8d /* nospec r8 */
pushq %r9 /* pt_regs->r9 */
- xorq %r9, %r9 /* nospec r9 */
+ xorl %r9d, %r9d /* nospec r9 */
pushq %r10 /* pt_regs->r10 */
- xorq %r10, %r10 /* nospec r10 */
+ xorl %r10d, %r10d /* nospec r10 */
pushq %r11 /* pt_regs->r11 */
- xorq %r11, %r11 /* nospec r11*/
+ xorl %r11d, %r11d /* nospec r11*/
pushq %rbx /* pt_regs->rbx */
xorl %ebx, %ebx /* nospec rbx*/
pushq %rbp /* pt_regs->rbp */
xorl %ebp, %ebp /* nospec rbp*/
pushq %r12 /* pt_regs->r12 */
- xorq %r12, %r12 /* nospec r12*/
+ xorl %r12d, %r12d /* nospec r12*/
pushq %r13 /* pt_regs->r13 */
- xorq %r13, %r13 /* nospec r13*/
+ xorl %r13d, %r13d /* nospec r13*/
pushq %r14 /* pt_regs->r14 */
- xorq %r14, %r14 /* nospec r14*/
+ xorl %r14d, %r14d /* nospec r14*/
pushq %r15 /* pt_regs->r15 */
- xorq %r15, %r15 /* nospec r15*/
+ xorl %r15d, %r15d /* nospec r15*/
UNWIND_HINT_REGS
.if \save_ret
pushq %rsi /* return address on top of stack */
diff --git a/arch/x86/entry/entry_64_compat.S b/arch/x86/entry/entry_64_compat.S
index fd65e016e413..364ea4a207be 100644
--- a/arch/x86/entry/entry_64_compat.S
+++ b/arch/x86/entry/entry_64_compat.S
@@ -85,25 +85,25 @@ ENTRY(entry_SYSENTER_compat)
pushq %rcx /* pt_regs->cx */
pushq $-ENOSYS /* pt_regs->ax */
pushq $0 /* pt_regs->r8 = 0 */
- xorq %r8, %r8 /* nospec r8 */
+ xorl %r8d, %r8d /* nospec r8 */
pushq $0 /* pt_regs->r9 = 0 */
- xorq %r9, %r9 /* nospec r9 */
+ xorl %r9d, %r9d /* nospec r9 */
pushq $0 /* pt_regs->r10 = 0 */
- xorq %r10, %r10 /* nospec r10 */
+ xorl %r10d, %r10d /* nospec r10 */
pushq $0 /* pt_regs->r11 = 0 */
- xorq %r11, %r11 /* nospec r11 */
+ xorl %r11d, %r11d /* nospec r11 */
pushq %rbx /* pt_regs->rbx */
xorl %ebx, %ebx /* nospec rbx */
pushq %rbp /* pt_regs->rbp (will be overwritten) */
xorl %ebp, %ebp /* nospec rbp */
pushq $0 /* pt_regs->r12 = 0 */
- xorq %r12, %r12 /* nospec r12 */
+ xorl %r12d, %r12d /* nospec r12 */
pushq $0 /* pt_regs->r13 = 0 */
- xorq %r13, %r13 /* nospec r13 */
+ xorl %r13d, %r13d /* nospec r13 */
pushq $0 /* pt_regs->r14 = 0 */
- xorq %r14, %r14 /* nospec r14 */
+ xorl %r14d, %r14d /* nospec r14 */
pushq $0 /* pt_regs->r15 = 0 */
- xorq %r15, %r15 /* nospec r15 */
+ xorl %r15d, %r15d /* nospec r15 */
cld
/*
@@ -224,25 +224,25 @@ GLOBAL(entry_SYSCALL_compat_after_hwframe)
pushq %rbp /* pt_regs->cx (stashed in bp) */
pushq $-ENOSYS /* pt_regs->ax */
pushq $0 /* pt_regs->r8 = 0 */
- xorq %r8, %r8 /* nospec r8 */
+ xorl %r8d, %r8d /* nospec r8 */
pushq $0 /* pt_regs->r9 = 0 */
- xorq %r9, %r9 /* nospec r9 */
+ xorl %r9d, %r9d /* nospec r9 */
pushq $0 /* pt_regs->r10 = 0 */
- xorq %r10, %r10 /* nospec r10 */
+ xorl %r10d, %r10d /* nospec r10 */
pushq $0 /* pt_regs->r11 = 0 */
- xorq %r11, %r11 /* nospec r11 */
+ xorl %r11d, %r11d /* nospec r11 */
pushq %rbx /* pt_regs->rbx */
xorl %ebx, %ebx /* nospec rbx */
pushq %rbp /* pt_regs->rbp (will be overwritten) */
xorl %ebp, %ebp /* nospec rbp */
pushq $0 /* pt_regs->r12 = 0 */
- xorq %r12, %r12 /* nospec r12 */
+ xorl %r12d, %r12d /* nospec r12 */
pushq $0 /* pt_regs->r13 = 0 */
- xorq %r13, %r13 /* nospec r13 */
+ xorl %r13d, %r13d /* nospec r13 */
pushq $0 /* pt_regs->r14 = 0 */
- xorq %r14, %r14 /* nospec r14 */
+ xorl %r14d, %r14d /* nospec r14 */
pushq $0 /* pt_regs->r15 = 0 */
- xorq %r15, %r15 /* nospec r15 */
+ xorl %r15d, %r15d /* nospec r15 */
/*
* User mode is traced as though IRQs are on, and SYSENTER
@@ -298,9 +298,9 @@ sysret32_from_system_call:
*/
SWITCH_TO_USER_CR3_NOSTACK scratch_reg=%r8 scratch_reg2=%r9
- xorq %r8, %r8
- xorq %r9, %r9
- xorq %r10, %r10
+ xorl %r8d, %r8d
+ xorl %r9d, %r9d
+ xorl %r10d, %r10d
swapgs
sysretl
END(entry_SYSCALL_compat)
@@ -358,25 +358,25 @@ ENTRY(entry_INT80_compat)
pushq %rcx /* pt_regs->cx */
pushq $-ENOSYS /* pt_regs->ax */
pushq $0 /* pt_regs->r8 = 0 */
- xorq %r8, %r8 /* nospec r8 */
+ xorl %r8d, %r8d /* nospec r8 */
pushq $0 /* pt_regs->r9 = 0 */
- xorq %r9, %r9 /* nospec r9 */
+ xorl %r9d, %r9d /* nospec r9 */
pushq $0 /* pt_regs->r10 = 0 */
- xorq %r10, %r10 /* nospec r10 */
+ xorl %r10d, %r10d /* nospec r10 */
pushq $0 /* pt_regs->r11 = 0 */
- xorq %r11, %r11 /* nospec r11 */
+ xorl %r11d, %r11d /* nospec r11 */
pushq %rbx /* pt_regs->rbx */
xorl %ebx, %ebx /* nospec rbx */
pushq %rbp /* pt_regs->rbp */
xorl %ebp, %ebp /* nospec rbp */
pushq %r12 /* pt_regs->r12 */
- xorq %r12, %r12 /* nospec r12 */
+ xorl %r12d, %r12d /* nospec r12 */
pushq %r13 /* pt_regs->r13 */
- xorq %r13, %r13 /* nospec r13 */
+ xorl %r13d, %r13d /* nospec r13 */
pushq %r14 /* pt_regs->r14 */
- xorq %r14, %r14 /* nospec r14 */
+ xorl %r14d, %r14d /* nospec r14 */
pushq %r15 /* pt_regs->r15 */
- xorq %r15, %r15 /* nospec r15 */
+ xorl %r15d, %r15d /* nospec r15 */
cld
/*
--
2.16.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH 2/2] x86/entry/64: use xorl for register clearing
@ 2018-02-14 18:21 Alexey Dobriyan
2018-02-14 18:27 ` Dominik Brodowski
0 siblings, 1 reply; 3+ messages in thread
From: Alexey Dobriyan @ 2018-02-14 18:21 UTC (permalink / raw)
To: linux; +Cc: linux-kernel
> - xorq %r8, %r8 /* nospec r8 */
> + xorl %r8d, %r8d /* nospec r8 */
The suffix should be simply dropped as operand size is unambigious.
It is just one more character than necessary on the screen.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH 2/2] x86/entry/64: use xorl for register clearing
2018-02-14 18:21 [PATCH 2/2] x86/entry/64: use xorl for register clearing Alexey Dobriyan
@ 2018-02-14 18:27 ` Dominik Brodowski
0 siblings, 0 replies; 3+ messages in thread
From: Dominik Brodowski @ 2018-02-14 18:27 UTC (permalink / raw)
To: Alexey Dobriyan; +Cc: linux-kernel
On Wed, Feb 14, 2018 at 09:21:12PM +0300, Alexey Dobriyan wrote:
> > - xorq %r8, %r8 /* nospec r8 */
> > + xorl %r8d, %r8d /* nospec r8 */
>
> The suffix should be simply dropped as operand size is unambigious.
> It is just one more character than necessary on the screen.
No strong feelings about this issue, but I prefer it to be explicit. And
that's what seems to be the standard in arch/x86/entry/ .
Thanks,
Dominik
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2018-02-14 18:27 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2018-02-14 18:21 [PATCH 2/2] x86/entry/64: use xorl for register clearing Alexey Dobriyan
2018-02-14 18:27 ` Dominik Brodowski
-- strict thread matches above, loose matches on Subject: below --
2018-02-14 17:59 [PATCH 0/2] x86/entry: xorq->xorl; idtentry size reduction Dominik Brodowski
2018-02-14 17:59 ` [PATCH 2/2] x86/entry/64: use xorl for register clearing Dominik Brodowski
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox