[PATCH v4 3/8] KVM: arm64: Add guard pages for KVM nVHE hypervisor stack

[Kalesh Singh]

On Tue, Mar 1, 2022 at 11:53 PM Marc Zyngier <maz at kernel.org> wrote:
>
> On Fri, 25 Feb 2022 03:34:48 +0000,
> Kalesh Singh <kaleshsingh at google.com> wrote:
> >
> > Maps the stack pages in the flexible private VA range and allocates
> > guard pages below the stack as unbacked VA space. The stack is aligned
> > to twice its size to aid overflow detection (implemented in a subsequent
> > patch in the series).
> >
> > Signed-off-by: Kalesh Singh <kaleshsingh at google.com>
> > ---
> >
> > Changes in v4:
> >   - Replace IS_ERR_OR_NULL check with IS_ERR check now that
> >     hyp_alloc_private_va_range() returns an error for null
> >     pointer, per Fuad
> >   - Format comments to < 80 cols, per Fuad
> >
> > Changes in v3:
> >   - Handle null ptr in IS_ERR_OR_NULL checks, per Mark
> >
> >  arch/arm64/include/asm/kvm_asm.h |  1 +
> >  arch/arm64/kvm/arm.c             | 32 +++++++++++++++++++++++++++++---
> >  2 files changed, 30 insertions(+), 3 deletions(-)
> >
> > diff --git a/arch/arm64/include/asm/kvm_asm.h b/arch/arm64/include/asm/kvm_asm.h
> > index d5b0386ef765..2e277f2ed671 100644
> > --- a/arch/arm64/include/asm/kvm_asm.h
> > +++ b/arch/arm64/include/asm/kvm_asm.h
> > @@ -169,6 +169,7 @@ struct kvm_nvhe_init_params {
> >       unsigned long tcr_el2;
> >       unsigned long tpidr_el2;
> >       unsigned long stack_hyp_va;
> > +     unsigned long stack_pa;
> >       phys_addr_t pgd_pa;
> >       unsigned long hcr_el2;
> >       unsigned long vttbr;
> > diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> > index ecc5958e27fe..0a83c0e7f838 100644
> > --- a/arch/arm64/kvm/arm.c
> > +++ b/arch/arm64/kvm/arm.c
> > @@ -1541,7 +1541,6 @@ static void cpu_prepare_hyp_mode(int cpu)
> >       tcr |= (idmap_t0sz & GENMASK(TCR_TxSZ_WIDTH - 1, 0)) << TCR_T0SZ_OFFSET;
> >       params->tcr_el2 = tcr;
> >
> > -     params->stack_hyp_va = kern_hyp_va(per_cpu(kvm_arm_hyp_stack_page, cpu) + PAGE_SIZE);
> >       params->pgd_pa = kvm_mmu_get_httbr();
> >       if (is_protected_kvm_enabled())
> >               params->hcr_el2 = HCR_HOST_NVHE_PROTECTED_FLAGS;
> > @@ -1990,14 +1989,41 @@ static int init_hyp_mode(void)
> >        * Map the Hyp stack pages
> >        */
> >       for_each_possible_cpu(cpu) {
> > +             struct kvm_nvhe_init_params *params = per_cpu_ptr_nvhe_sym(kvm_init_params, cpu);
> >               char *stack_page = (char *)per_cpu(kvm_arm_hyp_stack_page, cpu);
> > -             err = create_hyp_mappings(stack_page, stack_page + PAGE_SIZE,
> > -                                       PAGE_HYP);
> > +             unsigned long stack_hyp_va, guard_hyp_va;
> >
> > +             /*
> > +              * Private mappings are allocated downwards from io_map_base
> > +              * so allocate the stack first then the guard page.
> > +              *
> > +              * The stack is aligned to twice its size to facilitate overflow
> > +              * detection.
> > +              */
> > +             err = __create_hyp_private_mapping(__pa(stack_page), PAGE_SIZE,
> > +                                             PAGE_SIZE * 2, &stack_hyp_va, PAGE_HYP);
>
> Right, I guess that's where my earlier ask breaks, as you want an
> alignment that is *larger* than the allocation.
>
> >               if (err) {
> >                       kvm_err("Cannot map hyp stack\n");
> >                       goto out_err;
> >               }
> > +
> > +             /* Allocate unbacked private VA range for stack guard page */
> > +             guard_hyp_va = hyp_alloc_private_va_range(PAGE_SIZE, PAGE_SIZE);
>
> Huh. You are implicitly relying on the VA allocator handing you an
> address contiguous with the previous mapping. That's... brave. I'd
> rather you allocate the VA space upfront with the correct alignment
> and then map the single page where it should be in the VA region.
>
> That'd be a lot less fragile.

Agreed. I'll fix it in the next version.

Thanks,
Kalesh
>
> > +             if (IS_ERR((void *)guard_hyp_va)) {
> > +                     err = PTR_ERR((void *)guard_hyp_va);
> > +                     kvm_err("Cannot allocate hyp stack guard page\n");
> > +                     goto out_err;
> > +             }
> > +
> > +             /*
> > +              * Save the stack PA in nvhe_init_params. This will be needed
> > +              * to recreate the stack mapping in protected nVHE mode.
> > +              * __hyp_pa() won't do the right thing there, since the stack
> > +              * has been mapped in the flexible private VA space.
> > +              */
> > +             params->stack_pa = __pa(stack_page) + PAGE_SIZE;
> > +
> > +             params->stack_hyp_va = stack_hyp_va + PAGE_SIZE;
> >       }
> >
> >       for_each_possible_cpu(cpu) {
>
> Thanks,
>
>         M.
>
> --
> Without deviation from the norm, progress is not possible.