Re: QEMU v6.2 riscv64 device crash

[Peter Maydell]

On Wed, 20 Nov 2024 at 07:32, Yanfeng <yfliu2008@qq.com> wrote:
>
> Dear experts,
>
> I am running a rv64 binary program which uses hypervisor extension v0.6.1 on
> QEMU with "virt" board. The same RiscV program can run on both QEMU v6.0 and
> v6.1, but it led to crash of QEMU v6.2:
>
> ```
> ERROR:../..target/riscv/translate.c:232:get_gpr: code should not be reached
> Bail out! ERROR:../../target/riscv/translate.c:232:get_gpr: code should not be
> reached
> Aborted (core dumped)
> ```
>
> From GDB I can see the target is in VU mode, and the crash happens when the 
> code
> at 0x10152 is hit:
>
> ```
> 1014e <_start>:
>    1014e:       0069e197  auipc gp,0x69e
>    10152:       96218193  addi  gp,gp,-1694 # 6adab0 <__global_pointer$>
> ```
>
> Since I am too new to QEMU tracing, can someone teach how can I find out more
> information?

QEMU 6.2 is now several years old, so before investing too much
time in looking into this, I would suggest seeing if the bug
is still present in the most recent release (which is 9.1),
or on head-of-git. Even if you did identify the bug in 6.2,
that release is now so old that we are no longer doing stable
point releases for it, so it will never be fixed.

thanks
-- PMM