Re: [Qemu-devel] [PATCH 07/11] cris: avoid write only variables

[Blue Swirl]

On Thu, Oct 7, 2010 at 11:07 AM, Edgar E. Iglesias
<address@hidden> wrote:
> On Thu, Oct 07, 2010 at 12:08:05PM +0200, Markus Armbruster wrote:
>> Blue Swirl <address@hidden> writes:
>>
>> > Compiling with GCC 4.6.0 20100925 produced warnings:
>> > /src/qemu/target-cris/op_helper.c: In function 'helper_movl_sreg_reg':
>> > /src/qemu/target-cris/op_helper.c:145:8: error: variable 'tlb_v' set
>> > but not used [-Werror=unused-but-set-variable]
>> > In file included from /src/qemu/target-cris/translate.c:3154:0:
>> > /src/qemu/target-cris/translate_v10.c: In function 'dec10_prep_move_m':
>> > /src/qemu/target-cris/translate_v10.c:111:22: error: variable 'rd' set
>> > but not used [-Werror=unused-but-set-variable]
>> >
>> > Fix by making the variable declarations and their uses also conditional
>> > to debug definition, delete rd.
>> >
>> > Signed-off-by: Blue Swirl <address@hidden>
>> > ---
>> >  target-cris/op_helper.c     |    6 ++++++
>> >  target-cris/translate_v10.c |    5 ++---
>> >  2 files changed, 8 insertions(+), 3 deletions(-)
>> >
>> > diff --git a/target-cris/op_helper.c b/target-cris/op_helper.c
>> > index a60da94..28c79b1 100644
>> > --- a/target-cris/op_helper.c
>> > +++ b/target-cris/op_helper.c
>> > @@ -142,7 +142,9 @@ void helper_movl_sreg_reg (uint32_t sreg, uint32_t reg)
>> >                     uint32_t idx;
>> >                     uint32_t lo, hi;
>> >                     uint32_t vaddr;
>> > +#ifdef CRIS_OP_HELPER_DEBUG
>> >                     int tlb_v;
>> > +#endif
>> >
>> >                     idx = set = env->sregs[SFR_RW_MM_TLB_SEL];
>> >                     set >>= 4;
>> > @@ -157,13 +159,17 @@ void helper_movl_sreg_reg (uint32_t sreg, uint32_t 
>> > reg)
>> >                     vaddr = EXTRACT_FIELD(env->tlbsets[srs-1][set][idx].hi,
>> >                                           13, 31);
>> >                     vaddr <<= TARGET_PAGE_BITS;
>> > +#ifdef CRIS_OP_HELPER_DEBUG
>> >                     tlb_v = EXTRACT_FIELD(env->tlbsets[srs-1][set][idx].lo,
>> >                                         3, 3);
>> > +#endif
>> >                     env->tlbsets[srs - 1][set][idx].lo = lo;
>> >                     env->tlbsets[srs - 1][set][idx].hi = hi;
>> >
>> > +#ifdef CRIS_OP_HELPER_DEBUG
>> >                     D_LOG("tlb flush vaddr=%x v=%d pc=%x\n",
>> >                               vaddr, tlb_v, env->pc);
>> > +#endif
>> >                     tlb_flush_page(env, vaddr);
>> >             }
>> >     }
>>
>> Could we eliminate the bothersome variable instead?
>>
>> diff --git a/target-cris/op_helper.c b/target-cris/op_helper.c
>> index a60da94..94e3e27 100644
>> --- a/target-cris/op_helper.c
>> +++ b/target-cris/op_helper.c
>> @@ -142,7 +142,6 @@ void helper_movl_sreg_reg (uint32_t sreg, uint32_t reg)
>>                       uint32_t idx;
>>                       uint32_t lo, hi;
>>                       uint32_t vaddr;
>> -                     int tlb_v;
>>
>>                       idx = set = env->sregs[SFR_RW_MM_TLB_SEL];
>>                       set >>= 4;
>> @@ -157,13 +156,14 @@ void helper_movl_sreg_reg (uint32_t sreg, uint32_t reg)
>>                       vaddr = EXTRACT_FIELD(env->tlbsets[srs-1][set][idx].hi,
>>                                             13, 31);
>>                       vaddr <<= TARGET_PAGE_BITS;
>> -                     tlb_v = EXTRACT_FIELD(env->tlbsets[srs-1][set][idx].lo,
>> +                     D_LOG("tlb flush vaddr=%x v=%d pc=%x\n",
>> +                              vaddr,
>> +                              
>> EXTRACT_FIELD(env->tlbsets[srs-1][set][idx].lo,
>>                                           3, 3);
>> +                              env->pc);
>>                       env->tlbsets[srs - 1][set][idx].lo = lo;
>>                       env->tlbsets[srs - 1][set][idx].hi = hi;
>>
>> -                     D_LOG("tlb flush vaddr=%x v=%d pc=%x\n",
>> -                               vaddr, tlb_v, env->pc);
>>                       tlb_flush_page(env, vaddr);
>>               }
>>       }
>
> Hi,
>
> Sorry for the late answer.

Late? I think my timescale is slightly different than yours :-)

> Markus, I agree that removing tlb_v would have been better than ifdefs,
> but i think that the intent I originally had in mind was that there should
> not be a need to flush the entry from the QEMU TLB if the old guest
> entry was not valid.
>
>
> The following patch works on my side:
>
> diff --git a/target-cris/op_helper.c b/target-cris/op_helper.c
> index a60da94..be9eb06 100644
> --- a/target-cris/op_helper.c
> +++ b/target-cris/op_helper.c
> @@ -164,7 +164,9 @@ void helper_movl_sreg_reg (uint32_t sreg, uint32_t reg)
>
>                        D_LOG("tlb flush vaddr=%x v=%d pc=%x\n",
>                                  vaddr, tlb_v, env->pc);
> -                       tlb_flush_page(env, vaddr);
> +                       if (tlb_v) {
> +                               tlb_flush_page(env, vaddr);
> +                       }
>                }
>        }
>  #endif
>
>
> The target-cris/translate_v10.c hunk looks good.
>
> Blue, can you incorporate the tlb_v change in your patch set?
> Or if you prefer, I can commit that part on my side.

Please do.