Re: [Qemu-devel] [PATCH v3 00/11] tcg mips64 and mips r6 improvements

[Aurelien Jarno <aurelien@aurel32.net>]

On 2016-11-25 11:31, Jin Guojie wrote:
> Changes since v2:
>   * Update against master(v2.8.0-rc1)
>   * Tested on Loongson as mips32r2(el) and mips64r2(el) hosts.
>     Loongson only implements little-endian mips32/mips64 ISA.
>   * Fully work for 32-bit and 64-bit guests.
>     Fix two bugs：segmentation fault on mips64el with 32-bit guests,
>                   blocking when emulating i386 kernel on mips64el.
>   * Fix some minor style problems.
>   * PATCH v2 12~16 are not examined due to the lack of R6 machine. 
> 
> To be tested:
>   * big-endian mips32 and mips64 hosts.
>     I have tried running qemu-system-mips on an X86. The speed is awful. 
>     The compilation of qemu did not complete over a night until I gave up.
>     A better way is needed to do this test.
>   * MIPS R6.

Thanks for this new patch series, there are clearly a lot of
improvements. I haven't been able to test it fully yet, however, here
are my first tests results.

I have successfully tested the following guests on a mips32
big-endian host:
  - Debian amd64 with qemu-system-x86_64
  - Debian i386 with qemu-system-i386
  - Debian i386 with qemu-system-x86_64
It means that the regression I have spotted with the original series is
now gone.

I have successfully tested the following guests on a mips64
little-endian host:
 - Debian amd64 with qemu-system-x86_64
 - Debian armhf wit qemu-system-arm
 - Debian i386 with qemu-system-i386
 - Debian i386 with qemu-system-x86_64
 - Debian mips with qemu-system-mips
 - Debian mips with qemu-system-mips64
 - Debian mipsel with qemu-system-mipsel
 - Debian mipsel with qemu-system-mips64el
 - Debian mipsel/64-bit kernel with qemu-system-mips64el
 - Debian powerpc with qemu-system-ppc
However it seems that 64-bit big-endian guests are not working
correctly. It happens with either qemu-system-mips64el and
qemu-system-ppc64. The later hangs in the SLOF firmware, so it's
probably the easiest way to debug the issue. It would be nice to
get that fixed, that said as it's not a regression, I don't think
we should block merging the mips64 support on that.

I now plan to do more testing on mips32 be hosts and also test mips32
le hosts. Unfortunately I don't have a way to test mips R6 and mips64 be
hosts.

Aurelien 

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                 http://www.aurel32.net