On Tue, Apr 16, 2019 at 01:41:42PM +0100, Dave Martin wrote:
> On Mon, Apr 15, 2019 at 05:20:37PM +0200, Andrew Jones wrote:
> > On Fri, Apr 12, 2019 at 05:28:13PM +0100, Dave Martin wrote:
> > > A complicated DIV_ROUND_UP() expression is currently written out
> > > explicitly in multiple places in order to specify the size of the
> > > bitmap exchanged with userspace to represent the value of the
> > > KVM_REG_ARM64_SVE_VLS pseudo-register.
> > >
> > > To make this more readable, this patch replaces these with a single
> > > define.
> > >
> > > Since the number of words in a bitmap is just the index of the last
> > > word used + 1, this patch expresses the bound that way instead.
> > > This should make it clearer what is being expressed.
> > >
> > > Since use of DIV_ROUND_UP() was the only reason for including
> > > <linux/kernel.h> in guest.c, this patch removes that #include.
> > >
> > > No functional change.
> > >
> > > Suggested-by: Andrew Jones <drjones@xxxxxxxxxx>
> > > Signed-off-by: Dave Martin <Dave.Martin@xxxxxxx>
> > > ---
> > > arch/arm64/kvm/guest.c | 9 +++++----
> > > 1 file changed, 5 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
> > > index 73044e3..f025a2f 100644
> > > --- a/arch/arm64/kvm/guest.c
> > > +++ b/arch/arm64/kvm/guest.c
> > > @@ -23,7 +23,6 @@
> > > #include <linux/errno.h>
> > > #include <linux/err.h>
> > > #include <linux/nospec.h>
> > > -#include <linux/kernel.h>
> > > #include <linux/kvm_host.h>
> > > #include <linux/module.h>
> > > #include <linux/stddef.h>
> > > @@ -209,8 +208,10 @@ static int set_core_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
> > > #define vq_word(vq) (((vq) - SVE_VQ_MIN) / 64)
> > > #define vq_mask(vq) ((u64)1 << ((vq) - SVE_VQ_MIN) % 64)
> > >
> > > +#define SVE_VLS_WORDS (vq_word(SVE_VQ_MAX) + 1)
> > > +
> > > static bool vq_present(
> > > - const u64 (*const vqs)[DIV_ROUND_UP(SVE_VQ_MAX - SVE_VQ_MIN + 1, 64)],
> > > + const u64 (*const vqs)[SVE_VLS_WORDS],
> > > unsigned int vq)
> > > {
> > > return (*vqs)[vq_word(vq)] & vq_mask(vq);
> > > @@ -219,7 +220,7 @@ static bool vq_present(
> > > static int get_sve_vls(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
> > > {
> > > unsigned int max_vq, vq;
> > > - u64 vqs[DIV_ROUND_UP(SVE_VQ_MAX - SVE_VQ_MIN + 1, 64)];
> > > + u64 vqs[SVE_VLS_WORDS];
> > >
> > > if (!vcpu_has_sve(vcpu))
> > > return -ENOENT;
> > > @@ -243,7 +244,7 @@ static int get_sve_vls(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
> > > static int set_sve_vls(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
> > > {
> > > unsigned int max_vq, vq;
> > > - u64 vqs[DIV_ROUND_UP(SVE_VQ_MAX - SVE_VQ_MIN + 1, 64)];
> > > + u64 vqs[SVE_VLS_WORDS];
> > >
> > > if (!vcpu_has_sve(vcpu))
> > > return -ENOENT;
> > > --
> > > 2.1.4
> > >
> >
> > This is good, but I wonder if we could define the number of VLS words in
> > the documentation in terms of SVE_VQ_MAX too. Currently it's just the
> > hard coded 8 ("__u64 vector_lengths[8]").
> >
> > Reviewed-by: Andrew Jones <drjones@xxxxxxxxxx>
>
> I see your point, but SVE_VQ_MAX isn't really part of the KVM API, so I
> was avoiding it here.
It's not part of the KVM API, but it is uapi (asm/sigcontext.h). I'd
prefer we use it than to encourage KVM userspace to scatter 8's around.
>
> [8] is at least impossible to misinterpret, even if it's not the most
> self-explanatory option.
It's impossible to misinterpret, but also begs the questions of where
it comes from and if it will always be that way forever and ever.
Thanks,
drew
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
