Re: [PATCH 00/77] target/microblaze improvements

[Richard Henderson]

On 8/27/20 3:22 AM, Edgar E. Iglesias wrote:
> Thanks. Here's another issue, it seems some branches are jumping
> to the wrong address.
> 
> This is a disasm from a failing case:
> 
> 0x00000000ffd033a0:  brlid      r15, -636       // 0xffffffffffd03124
> 0x00000000ffd033a4:  or r0, r0, r0
> 
> 0x00000000ffa73124:  Address 0xffa73124 is out of bounds.

That's a weird one.

My guess is that IMM_FLAG is set in iflags incorrectly.
Can you verify this with -d in_asm,op,exec?

When IMM_FLAG is set, you'll see in in iflags: bit 0 will be set in the second
word of the insn_data.  E.g.:

 ---- 00000000ffd033a0 0000000000000001

It would also show up in the tb_flags of the exec lines.  E.g.

Trace 0: 0x7f38a4000940 [0000000000000000/0000000090000058/0]

where the format is host_pc [cs_base/pc/tb_flags].


If so, then we'll need to check where iflags got out of sync.


r~