Avoiding unnecessary jump relocations in gas?

[H. Peter Anvin]

On 05/18/2015 01:25 PM, H.J. Lu wrote:
> On Mon, May 18, 2015 at 1:06 PM, H. Peter Anvin <hpa@zytor.com> wrote:
>> On 05/18/2015 01:02 PM, H.J. Lu wrote:
>>>>
>>>> I wonder if it would make sense to have explicit mnemonics for the
>>>> one-byte offset and four-byte offset jump variants.  Sometimes users
>>>> want a jump with a 32-bit offset for reasons that have nothing to do
>>>> with link-time or load-time relocations.
>>>>
>>>
>>> There is:
>>>
>>> jmp.d32 foo
>>>
>>
>> How far back does that syntax work?
>>
> 
> .d32 support was added by
> 
> commit f8a5c266971d7b5b96f973805551c6e88669cada
> Author: H.J. Lu <hjl.tools@gmail.com>
> Date:   Thu Oct 14 13:31:13 2010 +0000
> 
>     Add .d32 encoding suffix.
> 
> and .d8 supported was added by
> 
> commit a501d77eeba717f6d54dce44f286f9e3aad83144
> Author: H.J. Lu <hjl.tools@gmail.com>
> Date:   Fri Jan 20 20:53:50 2012 +0000
> 
>     Add .d8 suffix support to x86 assembler
> 

OK, that is probably too recent.  The simplest answer I think is just to
.balign 16 each vector.  This is init space... some extra padding really
doesn't matter.

Patch attached (still in compile test).

	-hpa

-------------- next part --------------
diff --git a/arch/x86/include/asm/segment.h b/arch/x86/include/asm/segment.h
index 5a9856e..f11621f 100644
--- a/arch/x86/include/asm/segment.h
+++ b/arch/x86/include/asm/segment.h
@@ -233,7 +233,7 @@
 #ifdef __KERNEL__
 #ifndef __ASSEMBLY__
 
-extern const char early_idt_handlers[NUM_EXCEPTION_VECTORS][2+2+5];
+extern const char early_idt_handlers[NUM_EXCEPTION_VECTORS][16];
 #ifdef CONFIG_TRACING
 # define trace_early_idt_handlers early_idt_handlers
 #endif
diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S
index 02d2572..1c18826 100644
--- a/arch/x86/kernel/head_32.S
+++ b/arch/x86/kernel/head_32.S
@@ -524,6 +524,7 @@ setup_once:
 	andl $0,setup_once_ref	/* Once is enough, thanks */
 	ret
 
+	.balign 16
 ENTRY(early_idt_handlers)
 	# 36(%esp) %eflags
 	# 32(%esp) %cs
@@ -531,9 +532,8 @@ ENTRY(early_idt_handlers)
 	# 24(%rsp) error code
 	i = 0
 	.rept NUM_EXCEPTION_VECTORS
-	.if (EXCEPTION_ERRCODE_MASK >> i) & 1
-	ASM_NOP2
-	.else
+	.balign 16
+	.if ((EXCEPTION_ERRCODE_MASK >> i) & 1) == 0
 	pushl $0		# Dummy error code, to make stack frame uniform
 	.endif
 	pushl $i		# 20(%esp) Vector number
@@ -542,8 +542,7 @@ ENTRY(early_idt_handlers)
 	.endr
 ENDPROC(early_idt_handlers)
 	
-	/* This is global to keep gas from relaxing the jumps */
-ENTRY(early_idt_handler)
+early_idt_handler:
 	cld
 
 	cmpl $2,(%esp)		# X86_TRAP_NMI
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index 43eafc8..2d80a09 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -321,7 +321,8 @@ bad_address:
 	jmp bad_address
 
 	__INIT
-	.globl early_idt_handlers
+	.balign 16
+ENTRY(early_idt_handlers)
 early_idt_handlers:
 	# 104(%rsp) %rflags
 	#  96(%rsp) %cs
@@ -329,18 +330,17 @@ early_idt_handlers:
 	#  80(%rsp) error code
 	i = 0
 	.rept NUM_EXCEPTION_VECTORS
-	.if (EXCEPTION_ERRCODE_MASK >> i) & 1
-	ASM_NOP2
-	.else
+	.balign 16
+	.if ((EXCEPTION_ERRCODE_MASK >> i) & 1) == 0
 	pushq $0		# Dummy error code, to make stack frame uniform
 	.endif
 	pushq $i		# 72(%rsp) Vector number
 	jmp early_idt_handler
 	i = i + 1
 	.endr
+ENDPROC(early_idt_handlers)
 
-/* This is global to keep gas from relaxing the jumps */
-ENTRY(early_idt_handler)
+early_idt_handler:
 	cld
 
 	cmpl $2,(%rsp)		# X86_TRAP_NMI