On 05/01/2015 07:37 PM, Russell King - ARM Linux wrote:
> On Wed, Apr 29, 2015 at 03:37:37PM +0200, Nicolas Schichan wrote:
[...]
>> diff --git a/arch/arm/net/bpf_jit_32.c b/arch/arm/net/bpf_jit_32.c
>> index b5f470d..ffaf311 100644
>> --- a/arch/arm/net/bpf_jit_32.c
>> +++ b/arch/arm/net/bpf_jit_32.c
>> @@ -449,10 +449,10 @@ static inline void emit_udiv(u8 rd, u8 rm, u8 rn,
>> struct jit_ctx *ctx)
>> return;
>> }
>> #endif
>> - if (rm != ARM_R0)
>> - emit(ARM_MOV_R(ARM_R0, rm), ctx);
>> if (rn != ARM_R1)
>> emit(ARM_MOV_R(ARM_R1, rn), ctx);
>> + if (rm != ARM_R0)
>> + emit(ARM_MOV_R(ARM_R0, rm), ctx);
>
> I don't think you've thought enough about this. What if rm is ARM_R1?
> What if rn = ARM_R0 and rm = ARM_R1?
>
> How about:
>
> if (rn == ARM_R0 && rm == ARM_R1) {
> emit(ARM_MOV_R(ARM_R3, rn), ctx); // r3 <- r0(rn)
> emit(ARM_MOV_R(ARM_R0, rm), ctx); // r0 <- r1(rm)
> emit(ARM_MOV_R(ARM_R1, ARM_R3), ctx); // r1 <- r3
> } else if (rn == ARM_R0) {
> emit(ARM_MOV_R(ARM_R1, rn), ctx); // r1 <- rn
> if (rm != ARM_R0)
> emit(ARM_MOV_R(ARM_R0, rm), ctx); // r0 <- rm
> } else {
> if (rm != ARM_R0)
> emit(ARM_MOV_R(ARM_R0, rm), ctx); // r0 <- rm
> if (rn != ARM_R1)
> emit(ARM_MOV_R(ARM_R1, rn), ctx); // r1 <- rn
> }
>
Hello Russell,
In the current JIT, emit_udiv() is only being called with:
- rm = ARM_R4 (r_A) and rn = ARM_R0 (r_scrach) for BPF_ALU | BPF_DIV | BPF_K
- rm = ARM_R4 (r_A) and rn = ARM_R5 (r_X) for BPF_ALU | BPF_DIV | BPF_X
so it should not cause any issue in the current code state.
But yes, I'll rework the patch to avoid any other nasty surprises should the
code change.
Thanks,
--
Nicolas Schichan
Freebox SAS
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
