Re: gcc 4.5.1 / as 2.20.51.0.11 miscompiling drivers/char/i8k.c ?

[Andi Kleen <andi@firstfloor.org>]

> My speculation is, that the asm is not removed but rather that regs.eax 
> isn't reloaded after the asm because the memory clobber doesn't clobber 
> automatic variables.

Yes that makes sense. I wasn't able to verify it so far though.

Maybe the original poster could try the obvious patch 
instead of the volatile change.


i8k: tell gcc that regs gets clobbered
    
Signed-off-by: Andi Kleen <ak@linux.intel.com>

diff --git a/drivers/char/i8k.c b/drivers/char/i8k.c
index 3bc0eef..f3bbf73 100644
--- a/drivers/char/i8k.c
+++ b/drivers/char/i8k.c
@@ -142,7 +142,7 @@ static int i8k_smm(struct smm_regs *regs)
 		"lahf\n\t"
 		"shrl $8,%%eax\n\t"
 		"andl $1,%%eax\n"
-		:"=a"(rc)
+		:"=a"(rc), "=m" (*regs)
 		:    "a"(regs)
 		:    "%ebx", "%ecx", "%edx", "%esi", "%edi", "memory");
 #else
@@ -167,7 +167,7 @@ static int i8k_smm(struct smm_regs *regs)
 	    "movl %%edx,0(%%eax)\n\t"
 	    "lahf\n\t"
 	    "shrl $8,%%eax\n\t"
-	    "andl $1,%%eax\n":"=a"(rc)
+	    "andl $1,%%eax\n":"=a"(rc), "=m" (*regs)
 	    :    "a"(regs)
 	    :    "%ebx", "%ecx", "%edx", "%esi", "%edi", "memory");
 #endif

-Andi