[PATCH] RISC-V: KVM: Fix compilation without RISCV_ISA_ZICBOM

[Anup Patel]

On Tue, Oct 11, 2022 at 12:06 AM Palmer Dabbelt <palmer at dabbelt.com> wrote:
>
> On Mon, 10 Oct 2022 02:40:29 PDT (-0700), ajones at ventanamicro.com wrote:
> > Fix undefined reference of riscv_cbom_block_size when compiling KVM
> > without RISCV_ISA_ZICBOM. Note, RISCV_ISA_ZICBOM is a sufficient
> > guard as it selects RISCV_DMA_NONCOHERENT, which is needed to compile
> > dma-noncoherent.c (which is the file where riscv_cbom_block_size and
> > its initializer live).
> >
> > Fixes: afd5dde9a186 ("RISC-V: KVM: Provide UAPI for Zicbom block size")
> > Reported-by: kernel test robot <lkp at intel.com>
> > Signed-off-by: Andrew Jones <ajones at ventanamicro.com>
> > ---
> >  arch/riscv/kvm/vcpu.c | 2 ++
> >  1 file changed, 2 insertions(+)
> >
> > diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
> > index a032c4f0d600..e4453caba728 100644
> > --- a/arch/riscv/kvm/vcpu.c
> > +++ b/arch/riscv/kvm/vcpu.c
> > @@ -265,11 +265,13 @@ static int kvm_riscv_vcpu_get_reg_config(struct kvm_vcpu *vcpu,
> >       case KVM_REG_RISCV_CONFIG_REG(isa):
> >               reg_val = vcpu->arch.isa[0] & KVM_RISCV_BASE_ISA_MASK;
> >               break;
> > +#ifdef CONFIG_RISCV_ISA_ZICBOM
> >       case KVM_REG_RISCV_CONFIG_REG(zicbom_block_size):
> >               if (!riscv_isa_extension_available(vcpu->arch.isa, ZICBOM))
> >                       return -EINVAL;
> >               reg_val = riscv_cbom_block_size;
> >               break;
> > +#endif
> >       default:
> >               return -EINVAL;
> >       }
>
> Thanks, looks like this is tripping up the kernelci builds too:
> https://linux.kernelci.org/build/id/6343b37c6fd2dcbb5dcab5f3/logs/ .
> This symbol is actually a bit odd, as we don't actually just use it
> under Zicbom because of the T-Head stuff.  Looks like that's generically
> broken, though, so IMO we need something like this
>
> diff --git a/arch/riscv/mm/cacheflush.c b/arch/riscv/mm/cacheflush.c
> index 6cb7d96ad9c7..fba86fa3a628 100644
> --- a/arch/riscv/mm/cacheflush.c
> +++ b/arch/riscv/mm/cacheflush.c
> @@ -5,6 +5,9 @@
>
>  #include <asm/cacheflush.h>
>
> +unsigned int riscv_cbom_block_size;
> +EXPORT_SYMBOL(riscv_cbom_block_size);
> +
>  #ifdef CONFIG_SMP
>
>  #include <asm/sbi.h>
> diff --git a/arch/riscv/mm/dma-noncoherent.c b/arch/riscv/mm/dma-noncoherent.c
> index e3f9bdf47c5f..33d21ace0971 100644
> --- a/arch/riscv/mm/dma-noncoherent.c
> +++ b/arch/riscv/mm/dma-noncoherent.c
> @@ -12,7 +12,6 @@
>  #include <linux/of_device.h>
>  #include <asm/cacheflush.h>
>
> -unsigned int riscv_cbom_block_size;
>  static bool noncoherent_supported;
>
>  void arch_sync_dma_for_device(phys_addr_t paddr, size_t size,
>
> which should be
>
> Fixes: 8f7e001e0325 ("RISC-V: Clean up the Zicbom block size probing").

I agree with your change.

May be also move riscv_init_cbom_blocksize() to
arch/riscv/mm/cacheflush.c ?

>
> We might also want the ifdef for KVM, as that would allow us to build KVM that
> doesn't support these VCPU configurations, but that's sort of a different
> question.

Yes, I will be sending a PR including the KVM fix this week itself.

>
> --
> kvm-riscv mailing list
> kvm-riscv at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/kvm-riscv

Regards,
Anup