2008-06-05 16:08:24

by David Howells

Subject: [PATCH] FRV: ip_fast_csum() requires a memory clobber on its inline asm

ip_fast_csum() requires a memory clobber on its inline asm as it accesses
memory in a fashion that gcc can't predict.

The GCC manual says:

If your assembler instructions access memory in an unpredictable
fashion, add `memory' to the list of clobbered registers. This will
cause GCC to not keep memory values cached in registers across the
assembler instruction and not optimize stores or loads to that memory.

The bug hasn't been noticed in FRV, but it has been seen in PA-RISC.

Signed-off-by: David Howells <[email protected]>
---

include/asm-frv/checksum.h | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)


diff --git a/include/asm-frv/checksum.h b/include/asm-frv/checksum.h
index 9b16898..269da09 100644
--- a/include/asm-frv/checksum.h
+++ b/include/asm-frv/checksum.h
@@ -75,7 +75,7 @@ __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
: "=r" (sum), "=r" (iph), "=r" (ihl), "=r" (inc), "=&r"(tmp)
: "0" (sum), "1" (iph), "2" (ihl), "3" (4),
"m"(*(volatile struct { int _[100]; } *)iph)
- : "icc0", "icc1"
+ : "icc0", "icc1", "memory"
);

return (__force __sum16)~sum;

2008-06-05 18:57:57

by Matthew Wilcox

Subject: Re: [PATCH] FRV: ip_fast_csum() requires a memory clobber on its inline asm

On Thu, Jun 05, 2008 at 05:07:35PM +0100, David Howells wrote:
> : "=r" (sum), "=r" (iph), "=r" (ihl), "=r" (inc), "=&r"(tmp)
> : "0" (sum), "1" (iph), "2" (ihl), "3" (4),
> "m"(*(volatile struct { int _[100]; } *)iph)

Do you still need the input "m" constraint?

> - : "icc0", "icc1"
> + : "icc0", "icc1", "memory"
> );
>
> return (__force __sum16)~sum;

--
Intel are signing my paycheques ... these opinions are still mine
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours. We can't possibly take such
a retrograde step."