[vfio-users] 'dnf update' killed working VM

[Laszlo Ersek]

On 08/10/17 16:54, Laszlo Ersek wrote:
> On 08/10/17 14:15, Laszlo Ersek wrote:
> 
>> So I think it comes down to another toolchain difference between RHEL-7
>> and Fedora-26. Likely gcc -- the commit you identified is also related
>> to a C compiler. In particular:
>>
>> - on RHEL-7, the system compiler (gcc-4.8.5) is mapped to the "GCC48"
>>   toolchain settings of edk2, which lack support for LTO (link time
>>   optimization),
>>
>> - in the Fedora package's SPEC file, the "GCC49" toolchain settings are
>>   selected for the build in a fixed manner, which also lack support for
>>   LTO,
>>
>> - in kraxel's SPEC file, Fedora-26's gcc-7 compiler is mapped to the
>>   "GCC5" toolchain settings, which *enable* LTO (for DEBUG).
>>
>> The test you suggested to David elsewhere in this thread confirms that
>> the Fedora -- well, "virt-preview" -- package, built with GCC49
>> settings, works, and that Gerd's package, built with GCC5 settings,
>> breaks.
>>
>> I'll spin up a Fedora-26 guest and build OVMF there.
> 
> I confirm that building with gcc-7.1.1 on Fedora-26, using the GCC5
> toolchain settings and the DEBUG target (= optimizations enabled), the
> problem reproduces for me.
> 
> Keeping everything the same except switching DEBUG to NOOPT (=
> optimizations disabled) the OVMF binary returns to working state.
> 
> So, OVMF is being mis-built / mis-optimized by gcc-7.1, with the DEBUG
> GCC5 settings. I'll try to tweak both the code and the gcc flags.

Posted a patch (and CC'd you on it):

  https://lists.01.org/pipermail/edk2-devel/2017-August/013125.html

Testing would be appreciated. Instructions are in the linked message.

If you reply to edk2-devel, please subscribe first (and wait for the
subscription to complete), otherwise the list will drop your email. It's
bad and I've tried to change it, but so far I've failed.
<https://bugzilla.tianocore.org/show_bug.cgi?id=451>

Thanks
Laszlo