Re: [PULL 00/21] MIPS queue for June 7th, 2020

[Aleksandar Markovic]

On Mon, Jun 8, 2020 at 3:17 PM Aleksandar Markovic
<aleksandar.qemu.devel@gmail.com> wrote:
>
> пон, 8. јун 2020. у 15:05 Peter Maydell <peter.maydell@linaro.org> је
> написао/ла:
> >
> > On Sun, 7 Jun 2020 at 20:46, Aleksandar Markovic
> > <aleksandar.qemu.devel@gmail.com> wrote:
> > >
> > > The following changes since commit 
> > > 175198ad91d8bac540159705873b4ffe4fb94eab:
> > >
> > >   Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20200605' into 
> > > staging (2020-06-05 17:45:59 +0100)
> > >
> > > are available in the git repository at:
> > >
> > >   https://github.com/AMarkovic/qemu tags/mips-queue-june-07-2020
> > >
> > > for you to fetch changes up to ffbd8a88e8872d61fa5622a0075eddbe71951067:
> > >
> > >   target/mips: Enable hardware page table walker and CMGCR features for 
> > > P5600 (2020-06-07 21:34:14 +0200)
> > >
> > > ----------------------------------------------------------------
> > >
> > > MIPS queue for June 7th, 2020
> > >
> > > Highlights:
> > >
> > >   - Registring change of email address for two contributors
> > >   - Cleanup and improvements of FPU helpers
> > >   - Enabling some features of P5600
> > >   - Adding two Loongson-3A CPU definitions
> > >   - Moving futher towards Loongson-3A KVM support
> > >   - Two checkpatch warnings are known and should be ignored
> > >
> >
> > Hi; I'm afraid this fails to build on non-Linux hosts
> > (OSx, BSDs, Windows):
> >
> > hw/mips/common.c:21:23: fatal error: linux/kvm.h: No such file or directory
> >
> > Looking at that file, it also makes the mistake of putting
> > an include line before osdep.h, which should always be the
> > first include in a C file.
> >
> > The more usual way to arrange this kind of "real implementation
> > of a function for the compiled-with-KVM case, stub for
> > the #ifndef CONFIG_KVM case" is to put the real implementation
> > into a C file that's only built for the CONFIG_KVM case:
> > obj-$(CONFIG_KVM) += whatever.o
> > (it may be simplest to just put it in the existing target/mips/kvm.c?)
> > and then either
> > (a) provide the non-kvm stub version directly in the .h file
> > that declares the prototype for the kvm case:
> > #ifdef CONFIG_KVM
> > int mips_kvm_type(MachineState *machine, const char *vm_type)
> > #else
> > static inline mips_kvm_type(MachineState *machine, const char *vm_type)
> > {
> >     return 0;
> > }
> > #endif
> > (b) alternatively put the stub code in a built-for-non-KVM
> > source file:
> > obj-$(call lnot,$(CONFIG_KVM)) += kvm-stub.o
> >
> > i386, s390x and arm all have examples of this kind of "stub
> > version plus for-kvm version" function that may be useful
> > as examples.
> >
>
> Thanks for your detailed response and pointers!
>
> I am going to work with the original contributor on this.
>
> I am planning, of course, to send another, corrected, pull request,
> but not sooner than in 5-6 days.
>

I changed my mind. Since the offending patch is logically independent
from other patches and vice versa, I am going to send a pull request
this evening, without it, rather than wait for it to be fixed.

Truly yours,

Aleksandar


> Aleksandar
>
> > thanks
> > -- PMM
>