[Qemu-devel] sparc branch to pc+4 issue

[Qemu-devel] sparc branch to pc+4 issue

[Steven Seeger]

Consider pc==0x100:

0x100   b 0x104

The uncondtional not-annulled branch will go to 0x104, which is the next 
instruction anyway. do_branch() will leave dc->pc and dc->npc both set to 
0x104. This causes gdb to have a problem when single stepping. It will be 
stuck. QEMU will execute past this somehow, but I'm not sure with what side 
effect. It seems to me the following patch will fix this:

diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index 71e0853e43..95ca90b51a 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -1464,6 +1464,7 @@ static void do_branch(DisasContext *dc, int32_t offset, 
uint32_t insn, int cc)
             dc->npc = dc->pc + 4;
         } else {
             dc->pc = dc->npc;
+            if(target==dc->pc) target += 4;
             dc->npc = target;
             tcg_gen_mov_tl(cpu_pc, cpu_npc);
         }

I apologize if I am missing something with this assessment.

Steven

Re: [Qemu-devel] sparc branch to pc+4 issue

[Mark Cave-Ayland]

On 13/02/18 20:57, Steven Seeger wrote:

> Consider pc==0x100:
> 
> 0x100   b 0x104
> 
> The uncondtional not-annulled branch will go to 0x104, which is the next
> instruction anyway. do_branch() will leave dc->pc and dc->npc both set to
> 0x104. This causes gdb to have a problem when single stepping. It will be
> stuck. QEMU will execute past this somehow, but I'm not sure with what side
> effect. It seems to me the following patch will fix this:
> 
> diff --git a/target/sparc/translate.c b/target/sparc/translate.c
> index 71e0853e43..95ca90b51a 100644
> --- a/target/sparc/translate.c
> +++ b/target/sparc/translate.c
> @@ -1464,6 +1464,7 @@ static void do_branch(DisasContext *dc, int32_t offset,
> uint32_t insn, int cc)
>               dc->npc = dc->pc + 4;
>           } else {
>               dc->pc = dc->npc;
> +            if(target==dc->pc) target += 4;
>               dc->npc = target;
>               tcg_gen_mov_tl(cpu_pc, cpu_npc);
>           }
> 
> I apologize if I am missing something with this assessment.

Hi Steven,

Thanks for the patch! In principle the patch looks good, although sadly 
I don't currently have access to real hardware to verify this. Also 
you'll need to add a Signed-off-by tag in order for your patch to be 
applied (see https://wiki.qemu.org/Contribute/SubmitAPatch for more 
details).


ATB,

Mark.