Re: [Qemu-devel] qemu-system-ppc TCG assert with git master

[Mark Cave-Ayland]

On 12/03/15 14:55, Bastian Koppelmann wrote:

Hi Bastian,

> Hi Mark, sorry I forgot to cc qemu-devel :/
> 
> On 03/12/2015 08:41 AM, Mark Cave-Ayland wrote:
>> Hi all,
>>
>> Whilst testing git master in preparation for some OpenBIOS updates, I'm
>> seeing the following TCG assert in one of my older test images:
>>
>>
>> $ gdb --args ./qemu-system-ppc -cdrom
>> /home/build/src/qemu/image/ppc/ubuntu-5.10-live-powerpc.iso -boot d -g
>> 800x600x8
>> GNU gdb (GDB) 7.4.1-debian
>> Copyright (C) 2012 Free Software Foundation, Inc.
>> License GPLv3+: GNU GPL version 3 or later
>> <http://gnu.org/licenses/gpl.html>
>> This is free software: you are free to change and redistribute it.
>> There is NO WARRANTY, to the extent permitted by law.  Type "show
>> copying"
>> and "show warranty" for details.
>> This GDB was configured as "x86_64-linux-gnu".
>> For bug reporting instructions, please see:
>> <http://www.gnu.org/software/gdb/bugs/>...
>> Reading symbols from /home/build/rel-qemu-git/bin/qemu-system-ppc...done.
>> (gdb) r
>> Starting program: /home/build/rel-qemu-git/bin/qemu-system-ppc -cdrom
>> /home/build/src/qemu/image/ppc/ubuntu-5.10-live-powerpc.iso -boot d -g
>> 800x600x8
>> [Thread debugging using libthread_db enabled]
>> Using host libthread_db library
>> "/lib/x86_64-linux-gnu/libthread_db.so.1".
>> [New Thread 0x7fffe9ef3700 (LWP 11437)]
>> [New Thread 0x7fffe5bf1700 (LWP 11438)]
>> [New Thread 0x7fffe53f0700 (LWP 11439)]
>> [Thread 0x7fffe5bf1700 (LWP 11438) exited]
>> [New Thread 0x7fffe5bf1700 (LWP 11443)]
>> [Thread 0x7fffe5bf1700 (LWP 11443) exited]
>> qemu-system-ppc: /home/build/src/qemu/git/qemu/tcg/optimize.c:212:
>> tcg_opt_gen_mov: Assertion `temps[src].state != TCG_TEMP_CONST' failed.
> 
> does this fix it?
> 
> diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
> index 96adf9a..4c8ff70 100644
> --- a/tcg/tcg-op.h
> +++ b/tcg/tcg-op.h
> @@ -396,7 +396,11 @@ static inline void tcg_gen_and_i32(TCGv_i32 ret,
> TCGv_i32 arg1, TCGv_i32 arg2)
> 
>  static inline void tcg_gen_or_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32
> arg2)
>  {
> -    tcg_gen_op3_i32(INDEX_op_or_i32, ret, arg1, arg2);
> +    if (TCGV_EQUAL_I32(arg1, arg2)) {
> +        tcg_gen_mov_i32(ret, arg1);
> +    } else {
> +        tcg_gen_op3_i32(INDEX_op_or_i32, ret, arg1, arg2);
> +    }
>  }
> 
>  static inline void tcg_gen_xor_i32(TCGv_i32 ret, TCGv_i32 arg1,
> TCGv_i32 arg2)
> 
> The patch is also attached. It is just a revert of tcg_gen_or_i32. I'll
> investigate some more why the optimizer fails here.

Yes indeed, this fixes the issue in my tests here. But from what you're
saying this is more of a workaround rather than a fix?


ATB,

Mark.