PROBLEM: compilation issue, inline assembly arch/x86/kvm/emulate.c fails at -O0

Commit Message

H. Peter Anvin Nov. 21, 2012, 9:35 p.m. UTC

On 11/14/2012 11:45 AM, Blower, Melanie wrote:
> [1.] gcc -O0 assembly arch/x86/kvm/emulate.c gets compilation failure -- incorrect register restrictions
> [2.] Full description of the problem/report:
> I'm trying to compile this file at -O0, but gcc chokes in register allocation at the inline assembly.
> 
> In the ordinary Linux build, this file compiles with gcc at -O2, without compilation errors.
> 
> At -O0, gcc chokes with this message: 
> gcc  -w -c ./emulateE.c               // (using preprocessed file)
> ./emulateE.c: In function `em_mul_ex':
> ./emulateE.c:1918:5: error: can't find a register in class `AREG' while reloading `asm'
> ./emulateE.c:1918:5: error: `asm' operand has impossible constraints
> 
> Explanation:
> The file contains an inline asm of a kind:
> 
>     __asm__ __volatile__ ( " ..... " :
> 
>  "=m" ((ctxt)->eflags), "=&r" (_tmp), "+a" (*rax), "+d" (*rdx), "+qm"(ex) :
>  "i" (11), "m" ((ctxt)->src . val), "a" (*rax), "d" (*rdx)); 
> 
> Note that "+a" in inputs already means that eax is the return value. An then "a" is used as an output constraint too.
> 

Hi Melanie,

Can you test the attached patch?

	-hpa

Comments

Blower, Melanie Nov. 21, 2012, 10:31 p.m. UTC | #1
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--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

From f8d8e2842ca05fd89788e35c087f02c6159b023a Mon Sep 17 00:00:00 2001
From: "H. Peter Anvin" <[e-mail]>
Date: Wed, 21 Nov 2012 13:29:20 -0800
Subject: [PATCH] x86, kvm: Remove incorrect redundant assembly constraint

In __emulate_1op_rax_rdx, we use "+a" and "+d" which are input/output
constraints, and *then* use "a" and "d" as input constraints.  This is
incorrect, but happens to work on some versions of gcc.

However, it breaks gcc with -O0 and icc, and may break on future
versions of gcc.

Reported-by: Melanie Blower <[e-mail]>
Signed-off-by: H. Peter Anvin <[e-mail]>
Link: http://lkml.kernel.org/r/[e-mail]
---
 arch/x86/kvm/emulate.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index 39171cb..bba39bf 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -426,8 +426,7 @@  static void invalidate_registers(struct x86_emulate_ctxt *ctxt)
 			_ASM_EXTABLE(1b, 3b)				\
 			: "=m" ((ctxt)->eflags), "=&r" (_tmp),		\
 			  "+a" (*rax), "+d" (*rdx), "+qm"(_ex)		\
-			: "i" (EFLAGS_MASK), "m" ((ctxt)->src.val),	\
-			  "a" (*rax), "d" (*rdx));			\
+			: "i" (EFLAGS_MASK), "m" ((ctxt)->src.val));	\
 	} while (0)
 
 /* instruction has only one source operand, destination is implicit (e.g. mul, div, imul, idiv) */
-- 
1.7.11.7