[Qemu-devel] [PATCH] gdbstub: Synchronize CPU state unconditionally in gdb_set_cpu_pc

[Qemu-devel] [PATCH] gdbstub: Synchronize CPU state unconditionally in gdb_set_cpu_pc

[Peter Maydell]

Synchronize the CPU state via cpu_sychronize_state() unconditionally
in gdb_set_cpu_pc() rather than only in some of the target ifdef
ladder cases.

We can divide the CPUs into three categories:
 * non-KVM targets: no change of behaviour since we will use the
   kvm-stub.c no-op function.
 * i386 and s390: no change of behaviour since they were already
   calling this function
 * PPC (in KVM mode): this fixes an error: failing to synchronise
   was accidental and probably a bug.

This also paves the way for other targets (specifically ARM) which
can add KVM support in future without having to add another target
specific change to this bit of code.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
Alex: could you test the KVM PPC case, please, since that's the only
one where we actually change behaviour here?

 gdbstub.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/gdbstub.c b/gdbstub.c
index ef95ac2..776dcc5 100644
--- a/gdbstub.c
+++ b/gdbstub.c
@@ -1904,8 +1904,8 @@ static void gdb_breakpoint_remove_all(void)
 
 static void gdb_set_cpu_pc(GDBState *s, target_ulong pc)
 {
-#if defined(TARGET_I386)
     cpu_synchronize_state(s->c_cpu);
+#if defined(TARGET_I386)
     s->c_cpu->eip = pc;
 #elif defined (TARGET_PPC)
     s->c_cpu->nip = pc;
@@ -1930,7 +1930,6 @@ static void gdb_set_cpu_pc(GDBState *s, target_ulong pc)
 #elif defined (TARGET_ALPHA)
     s->c_cpu->pc = pc;
 #elif defined (TARGET_S390X)
-    cpu_synchronize_state(s->c_cpu);
     s->c_cpu->psw.addr = pc;
 #elif defined (TARGET_LM32)
     s->c_cpu->pc = pc;
-- 
1.7.1

Re: [Qemu-devel] [PATCH] gdbstub: Synchronize CPU state unconditionally in gdb_set_cpu_pc

[Peter Maydell]

Ping?

-- PMM

On 12 March 2012 16:24, Peter Maydell <peter.maydell@linaro.org> wrote:
> Synchronize the CPU state via cpu_sychronize_state() unconditionally
> in gdb_set_cpu_pc() rather than only in some of the target ifdef
> ladder cases.
>
> We can divide the CPUs into three categories:
>  * non-KVM targets: no change of behaviour since we will use the
>   kvm-stub.c no-op function.
>  * i386 and s390: no change of behaviour since they were already
>   calling this function
>  * PPC (in KVM mode): this fixes an error: failing to synchronise
>   was accidental and probably a bug.
>
> This also paves the way for other targets (specifically ARM) which
> can add KVM support in future without having to add another target
> specific change to this bit of code.
>
> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
> Alex: could you test the KVM PPC case, please, since that's the only
> one where we actually change behaviour here?
>
>  gdbstub.c |    3 +--
>  1 files changed, 1 insertions(+), 2 deletions(-)
>
> diff --git a/gdbstub.c b/gdbstub.c
> index ef95ac2..776dcc5 100644
> --- a/gdbstub.c
> +++ b/gdbstub.c
> @@ -1904,8 +1904,8 @@ static void gdb_breakpoint_remove_all(void)
>
>  static void gdb_set_cpu_pc(GDBState *s, target_ulong pc)
>  {
> -#if defined(TARGET_I386)
>     cpu_synchronize_state(s->c_cpu);
> +#if defined(TARGET_I386)
>     s->c_cpu->eip = pc;
>  #elif defined (TARGET_PPC)
>     s->c_cpu->nip = pc;
> @@ -1930,7 +1930,6 @@ static void gdb_set_cpu_pc(GDBState *s, target_ulong pc)
>  #elif defined (TARGET_ALPHA)
>     s->c_cpu->pc = pc;
>  #elif defined (TARGET_S390X)
> -    cpu_synchronize_state(s->c_cpu);
>     s->c_cpu->psw.addr = pc;
>  #elif defined (TARGET_LM32)
>     s->c_cpu->pc = pc;
> --
> 1.7.1
>
>

Re: [Qemu-devel] [PATCH] gdbstub: Synchronize CPU state unconditionally in gdb_set_cpu_pc

[Christoffer Dall]

On Tue, Apr 3, 2012 at 7:57 AM, Peter Maydell <peter.maydell@linaro.org> wrote:
> Ping?
>
sorry about late response.

>
> On 12 March 2012 16:24, Peter Maydell <peter.maydell@linaro.org> wrote:
>> Synchronize the CPU state via cpu_sychronize_state() unconditionally
>> in gdb_set_cpu_pc() rather than only in some of the target ifdef
>> ladder cases.
>>
>> We can divide the CPUs into three categories:
>>  * non-KVM targets: no change of behaviour since we will use the
>>   kvm-stub.c no-op function.
>>  * i386 and s390: no change of behaviour since they were already
>>   calling this function
>>  * PPC (in KVM mode): this fixes an error: failing to synchronise
>>   was accidental and probably a bug.
>>
>> This also paves the way for other targets (specifically ARM) which
>> can add KVM support in future without having to add another target
>> specific change to this bit of code.
>>
>> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
>> ---
>> Alex: could you test the KVM PPC case, please, since that's the only
>> one where we actually change behaviour here?
>>
>>  gdbstub.c |    3 +--
>>  1 files changed, 1 insertions(+), 2 deletions(-)
>>
>> diff --git a/gdbstub.c b/gdbstub.c
>> index ef95ac2..776dcc5 100644
>> --- a/gdbstub.c
>> +++ b/gdbstub.c
>> @@ -1904,8 +1904,8 @@ static void gdb_breakpoint_remove_all(void)
>>
>>  static void gdb_set_cpu_pc(GDBState *s, target_ulong pc)
>>  {
>> -#if defined(TARGET_I386)
>>     cpu_synchronize_state(s->c_cpu);
>> +#if defined(TARGET_I386)
>>     s->c_cpu->eip = pc;
>>  #elif defined (TARGET_PPC)
>>     s->c_cpu->nip = pc;
>> @@ -1930,7 +1930,6 @@ static void gdb_set_cpu_pc(GDBState *s, target_ulong pc)
>>  #elif defined (TARGET_ALPHA)
>>     s->c_cpu->pc = pc;
>>  #elif defined (TARGET_S390X)
>> -    cpu_synchronize_state(s->c_cpu);
>>     s->c_cpu->psw.addr = pc;
>>  #elif defined (TARGET_LM32)
>>     s->c_cpu->pc = pc;
>> --
>> 1.7.1
>>

looks good to me - much cleaner than what I had to do for ARM otherwise.

-Christoffer