Re: LKML: Vasily Gorbik: [GIT PULL] s390 patches for the 5.6 merge window

Re: LKML: Vasily Gorbik: [GIT PULL] s390 patches for the 5.6 merge window

[Vasily Gorbik]

On Fri, Feb 07, 2020 at 10:22:20PM +0000, Nick Desaulniers wrote:
> + Vasily for real
> 
> On Fri, Feb 7, 2020, 11:02 PM Nick Desaulniers <ndesaulniers@google.com>
> wrote:
> 
> > https://lkml.org/lkml/2020/1/28/1141
> >
> > Yo, what's up with clang 10 support?
I believe s390 kernel had limited clang build support since about 5.2
(with jump labels and ftrace disabled), thanks to contributions from Arnd
Bergmann. With asm goto support in clang 9 and now everything for ftrace
in place in clang 10 s390 kernel should be in a good shape feature wise
(including KASAN). At this stage clang based tooling is the main interest.

> >
> > Can I boot this in qemu?
> >
qemu tcg seems to work fine for gcc built s390 kernels. While clang built
kernels run smoothly natively and with accel=kvm there is still something
to be fixed for accel=tcg (kernel fails to boot on both s390 and x86).

Re: LKML: Vasily Gorbik: [GIT PULL] s390 patches for the 5.6 merge window

[David Hildenbrand]

On 10.02.20 18:58, Vasily Gorbik wrote:
> On Fri, Feb 07, 2020 at 10:22:20PM +0000, Nick Desaulniers wrote:
>> + Vasily for real
>>
>> On Fri, Feb 7, 2020, 11:02 PM Nick Desaulniers <ndesaulniers@google.com>
>> wrote:
>>
>>> https://lkml.org/lkml/2020/1/28/1141
>>>
>>> Yo, what's up with clang 10 support?
> I believe s390 kernel had limited clang build support since about 5.2
> (with jump labels and ftrace disabled), thanks to contributions from Arnd
> Bergmann. With asm goto support in clang 9 and now everything for ftrace
> in place in clang 10 s390 kernel should be in a good shape feature wise
> (including KASAN). At this stage clang based tooling is the main interest.
> 
>>>
>>> Can I boot this in qemu?
>>>
> qemu tcg seems to work fine for gcc built s390 kernels. While clang built
> kernels run smoothly natively and with accel=kvm there is still something
> to be fixed for accel=tcg (kernel fails to boot on both s390 and x86).
> 

Any details on that (e.g., kernel splat)? Could be that TCG doesn't like
a specific instruction - but could also be that some code is broken that
is not taken by HW/KVM (esp. because TCG implements less CPU facilities
- e.g., no support for huge pages).

Cheers!

-- 
Thanks,

David / dhildenb