Re: [PATCH 15/17] kvm: arm64: Get rid of fake page table levels

From: Christoffer Dall
Date: Fri Apr 08 2016 - 11:05:20 EST


On Mon, Apr 04, 2016 at 05:26:15PM +0100, Suzuki K Poulose wrote:
> On arm64, the hardware mandates concatenation of upto 16 tables,
> at entry level for stage2 translations. This could lead to reduced
> number of translation levels than the normal (stage1 table).
> Also, since the IPA(40bit) is smaller than the some of the supported
> VA_BITS (e.g, 48bit), there could be different number of levels
> in stage-1 vs stage-2 tables. To work around this, so far we have been
> using a fake software page table level, not known to the hardware.
> But with 16K translations, there could be upto 2 fake software
> levels, which complicates the code. Hence, we want to get rid of the hack.
>
> Now that we have explicit accessors for hyp vs stage2 page tables,
> define the stage2 walker helpers accordingly based on the actual
> table used by the hardware.
>
> Once we know the number of translation levels used by the hardware,
> it is merely a job of defining the helpers based on whether a
> particular level is folded or not, looking at the number of levels.
>
> Some facts before we calculate the translation levels:
>
> 1) Smallest page size supported by arm64 is 4K.
> 2) The minimum number of bits resolved at any page table level
> is (PAGE_SHIFT - 3) at intermediate levels.
> Both of them implies, minimum number of bits required for a level
> change is 9.
>
> Since we can concatenate upto 16 tables at stage2 entry, the total
> number of page table levels used by the hardware for resolving N bits
> is same as that for (N - 4) bits (with concatenation), as there cannot
> be a level in between (N, N-4) as per the above rules.
>
> Hence, we have
>
> STAGE2_PGTABLE_LEVELS = PGTABLE_LEVELS(KVM_PHYS_SHIFT - 4)
>
> With the current IPA limit (40bit), for all supported translations
> and VA_BITS, we have the following condition (even for 36bit VA with
> 16K page size):
>
> CONFIG_PGTABLE_LEVELS >= STAGE2_PGTABLE_LEVELS.
>
> So, for e.g, if PUD is present in stage2, it is present in the hyp(host).
> Hence, we fall back to the host definition if we find that a level is not
> folded. Otherwise we redefine it accordingly. A build time check is added
> to make sure the above condition holds.
>
> Cc: Marc Zyngier <marc.zyngier@xxxxxxx>
> Cc: Christoffer Dall <christoffer.dall@xxxxxxxxxx>
> Signed-off-by: Suzuki K Poulose <suzuki.poulose@xxxxxxx>
> ---
> arch/arm64/include/asm/kvm_mmu.h | 64 +-------------
> arch/arm64/include/asm/stage2_pgtable-nopmd.h | 42 +++++++++
> arch/arm64/include/asm/stage2_pgtable-nopud.h | 39 +++++++++
> arch/arm64/include/asm/stage2_pgtable.h | 113 +++++++++++++++++--------
> 4 files changed, 163 insertions(+), 95 deletions(-)
> create mode 100644 arch/arm64/include/asm/stage2_pgtable-nopmd.h
> create mode 100644 arch/arm64/include/asm/stage2_pgtable-nopud.h
>
> diff --git a/arch/arm64/include/asm/kvm_mmu.h b/arch/arm64/include/asm/kvm_mmu.h
> index a3c0d05..e3fee0a 100644
> --- a/arch/arm64/include/asm/kvm_mmu.h
> +++ b/arch/arm64/include/asm/kvm_mmu.h
> @@ -45,18 +45,6 @@
> */
> #define TRAMPOLINE_VA (HYP_PAGE_OFFSET_MASK & PAGE_MASK)
>
> -/*
> - * KVM_MMU_CACHE_MIN_PAGES is the number of stage2 page table translation
> - * levels in addition to the PGD and potentially the PUD which are
> - * pre-allocated (we pre-allocate the fake PGD and the PUD when the Stage-2
> - * tables use one level of tables less than the kernel.
> - */
> -#ifdef CONFIG_ARM64_64K_PAGES
> -#define KVM_MMU_CACHE_MIN_PAGES 1
> -#else
> -#define KVM_MMU_CACHE_MIN_PAGES 2
> -#endif
> -
> #ifdef __ASSEMBLY__
>
> #include <asm/alternative.h>
> @@ -155,69 +143,21 @@ static inline bool kvm_s2pmd_readonly(pmd_t *pmd)
>
> static inline void *kvm_get_hwpgd(struct kvm *kvm)
> {
> - pgd_t *pgd = kvm->arch.pgd;
> - pud_t *pud;
> -
> - if (KVM_PREALLOC_LEVEL == 0)
> - return pgd;
> -
> - pud = pud_offset(pgd, 0);
> - if (KVM_PREALLOC_LEVEL == 1)
> - return pud;
> -
> - BUG_ON(KVM_PREALLOC_LEVEL != 2);
> - return pmd_offset(pud, 0);
> + return kvm->arch.pgd;
> }
>
> static inline unsigned int kvm_get_hwpgd_size(void)
> {
> - if (KVM_PREALLOC_LEVEL > 0)
> - return PTRS_PER_S2_PGD * PAGE_SIZE;
> return PTRS_PER_S2_PGD * sizeof(pgd_t);
> }
>
> -/*
> - * Allocate fake pgd for the host kernel page table macros to work.
> - * This is not used by the hardware and we have no alignment
> - * requirement for this allocation.
> - */
> static inline pgd_t *kvm_setup_fake_pgd(pgd_t *hwpgd)
> {
> - int i;
> - pgd_t *pgd;
> -
> - if (!KVM_PREALLOC_LEVEL)
> - return hwpgd;
> -
> - /*
> - * When KVM_PREALLOC_LEVEL==2, we allocate a single page for
> - * the PMD and the kernel will use folded pud.
> - * When KVM_PREALLOC_LEVEL==1, we allocate 2 consecutive PUD
> - * pages.
> - */
> -
> - pgd = kmalloc(PTRS_PER_S2_PGD * sizeof(pgd_t),
> - GFP_KERNEL | __GFP_ZERO);
> - if (!pgd)
> - return ERR_PTR(-ENOMEM);
> -
> - /* Plug the HW PGD into the fake one. */
> - for (i = 0; i < PTRS_PER_S2_PGD; i++) {
> - if (KVM_PREALLOC_LEVEL == 1)
> - pgd_populate(NULL, pgd + i,
> - (pud_t *)hwpgd + i * PTRS_PER_PUD);
> - else if (KVM_PREALLOC_LEVEL == 2)
> - pud_populate(NULL, pud_offset(pgd, 0) + i,
> - (pmd_t *)hwpgd + i * PTRS_PER_PMD);
> - }
> -
> - return pgd;
> + return hwpgd;
> }
>
> static inline void kvm_free_fake_pgd(pgd_t *pgd)
> {
> - if (KVM_PREALLOC_LEVEL > 0)
> - kfree(pgd);
> }
> static inline bool kvm_page_empty(void *ptr)
> {
> diff --git a/arch/arm64/include/asm/stage2_pgtable-nopmd.h b/arch/arm64/include/asm/stage2_pgtable-nopmd.h
> new file mode 100644
> index 0000000..40dda8c
> --- /dev/null
> +++ b/arch/arm64/include/asm/stage2_pgtable-nopmd.h
> @@ -0,0 +1,42 @@
> +/*
> + * Copyright (C) 2016 - ARM Ltd
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 as
> + * published by the Free Software Foundation.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * along with this program. If not, see <http://www.gnu.org/licenses/>.
> + */
> +
> +#ifndef __ARM64_S2_PGTABLE_NOPMD_H_
> +#define __ARM64_S2_PGTABLE_NOPMD_H_
> +
> +#include <asm/stage2_pgtable-nopud.h>
> +
> +#define __S2_PGTABLE_PMD_FOLDED
> +
> +#define S2_PMD_SHIFT S2_PUD_SHIFT
> +#define S2_PTRS_PER_PMD 1
> +#define S2_PMD_SIZE (1UL << S2_PMD_SHIFT)
> +#define S2_PMD_MASK (~(S2_PMD_SIZE-1))
> +
> +#define stage2_pud_none(pud) (0)
> +#define stage2_pud_present(pud) (1)
> +#define stage2_pud_clear(pud) do { } while (0)
> +#define stage2_pud_populate(mm, pud, pmd) do { } while (0)
> +#define stage2_pmd_offset(pud, address) ((pmd_t *)(pud))
> +
> +#define stage2_pmd_free(mm, pmd) do { } while (0)
> +
> +#define stage2_pmd_addr_end(addr, end) (end)
> +
> +#define stage2_pud_huge(pud) (0)
> +#define stage2_pmd_table_empty(pmdp) (0)
> +
> +#endif
> diff --git a/arch/arm64/include/asm/stage2_pgtable-nopud.h b/arch/arm64/include/asm/stage2_pgtable-nopud.h
> new file mode 100644
> index 0000000..b0eff9d
> --- /dev/null
> +++ b/arch/arm64/include/asm/stage2_pgtable-nopud.h
> @@ -0,0 +1,39 @@
> +/*
> + * Copyright (C) 2016 - ARM Ltd
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 as
> + * published by the Free Software Foundation.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * along with this program. If not, see <http://www.gnu.org/licenses/>.
> + */
> +
> +#ifndef __ARM64_S2_PGTABLE_NOPUD_H_
> +#define __ARM64_S2_PGTABLE_NOPUD_H_
> +
> +#define __S2_PGTABLE_PUD_FOLDED
> +
> +#define S2_PUD_SHIFT S2_PGDIR_SHIFT
> +#define S2_PTRS_PER_PUD 1
> +#define S2_PUD_SIZE (_AC(1, UL) << S2_PUD_SHIFT)
> +#define S2_PUD_MASK (~(S2_PUD_SIZE-1))
> +
> +#define stage2_pgd_none(pgd) (0)
> +#define stage2_pgd_present(pgd) (1)
> +#define stage2_pgd_clear(pgd) do { } while (0)
> +#define stage2_pgd_populate(mm, pgd, pud) do { } while (0)
> +
> +#define stage2_pud_offset(pgd, address) ((pud_t *)(pgd))
> +
> +#define stage2_pud_free(mm, x) do { } while (0)
> +
> +#define stage2_pud_addr_end(addr, end) (end)
> +#define stage2_pud_table_empty(pmdp) (0)
> +
> +#endif
> diff --git a/arch/arm64/include/asm/stage2_pgtable.h b/arch/arm64/include/asm/stage2_pgtable.h
> index 751227d..139b4db 100644
> --- a/arch/arm64/include/asm/stage2_pgtable.h
> +++ b/arch/arm64/include/asm/stage2_pgtable.h
> @@ -22,32 +22,55 @@
> #include <asm/pgtable.h>
>
> /*
> - * In the case where PGDIR_SHIFT is larger than KVM_PHYS_SHIFT, we can address
> - * the entire IPA input range with a single pgd entry, and we would only need
> - * one pgd entry. Note that in this case, the pgd is actually not used by
> - * the MMU for Stage-2 translations, but is merely a fake pgd used as a data
> - * structure for the kernel pgtable macros to work.
> + * The hardware mandates concatenation of upto 16 tables at stage2 entry level.

s/upto/up to/

> + * Now, the minimum number of bits resolved at any level is (PAGE_SHIFT - 3),
> + * or in other words log2(PTRS_PER_PTE). On arm64, the smallest PAGE_SIZE

not sure the log2 comment helps here.

> + * supported is 4k, which means (PAGE_SHIFT - 3) > 4 holds for all page sizes.
> + * This implies, the total number of page table levels at stage2 expected
> + * by the hardware is actually the number of levels required for (KVM_PHYS_SHIFT - 4)
> + * in normal translations(e.g, stage-1), since we cannot have another level in
> + * the range (KVM_PHYS_SHIFT, KVM_PHYS_SHIFT - 4).

Is it not a design decision to always choose the maximum number of
concatinated initial-level stage2 tables (with the constraint that
there's a minimum number required)?

I agree with the design decision, if my math is correct that on 64K
systems you end up requiring a 1MB physically contiguous 1MB aligned
allocation for each VM? This seems reasonable enough if you configure
your kernel with 64K pages and expect to run VMs on top of that.

> */
> -#if PGDIR_SHIFT > KVM_PHYS_SHIFT
> -#define PTRS_PER_S2_PGD_SHIFT 0
> -#else
> -#define PTRS_PER_S2_PGD_SHIFT (KVM_PHYS_SHIFT - PGDIR_SHIFT)
> -#endif
> -#define PTRS_PER_S2_PGD (1 << PTRS_PER_S2_PGD_SHIFT)
> +#define STAGE2_PGTABLE_LEVELS ARM64_HW_PGTABLE_LEVELS(KVM_PHYS_SHIFT - 4)
>
> /*
> - * If we are concatenating first level stage-2 page tables, we would have less
> - * than or equal to 16 pointers in the fake PGD, because that's what the
> - * architecture allows. In this case, (4 - CONFIG_PGTABLE_LEVELS)
> - * represents the first level for the host, and we add 1 to go to the next
> - * level (which uses contatenation) for the stage-2 tables.
> + * At the moment, we do not support a combination of guest IPA and host VA_BITS
> + * where
> + * STAGE2_PGTABLE_LEVELS > CONFIG_PGTABLE_LEVELS

can you change this comment to reverse the statement to avoid someone
seeing this as a constraint, when in fact it's a negative invariant?

So the case we don't support is a sufficiently larger IPA space compared
to the host VA space such that the above happens? (Since at the same
IPA space size as host VA space size, the stage-2 levels will always be
less than or equal to the host levels.)

I don't see how that would ever work with userspace either so I think
this is a safe assumption and not something that ever needs fixing. In
which case this should be reworded to just state the assumptions and why
this is a good assumption.

(If my assumptions are wrong here, then there are also weird cases where
the host does huge pages at the PMD level and we don't. Not sure I can
see the full ramifications of that.)

> + *
> + * We base our stage-2 page table walker helpers based on this assumption and
> + * fallback to using the host version of the helper wherever possible.
> + * i.e, if a particular level is not folded (e.g, PUD) at stage2, we fall back
> + * to using the host version, since it is guaranteed it is not folded at host.

I don't really understand why it's desirable to fall back to the host
version of the helpers; in fact I would probably prefer to just have it
disjoint, but maybe I'll see the reason when going through the patch
more. But I doubt the value of this particular piece of commentary...

> + *
> + * TODO: We could lift this limitation easily by rearranging the host level
> + * definitions to a more reusable version.
> */

So is this really a TODO: based on the above?

> -#if PTRS_PER_S2_PGD <= 16
> -#define KVM_PREALLOC_LEVEL (4 - CONFIG_PGTABLE_LEVELS + 1)
> -#else
> -#define KVM_PREALLOC_LEVEL (0)
> +#if STAGE2_PGTABLE_LEVELS > CONFIG_PGTABLE_LEVELS
> +#error "Unsupported combination of guest IPA and host VA_BITS."
> #endif
>
> +

Can we add a comment as to what this defines exactly? Something like:
/*
* PGDIR_SHIFT determines the size a top-level stage2 page table entry can map
*/
> +#define S2_PGDIR_SHIFT ARM64_HW_PGTABLE_LEVEL_SHIFT(4 - STAGE2_PGTABLE_LEVELS)
> +#define S2_PGDIR_SIZE (_AC(1, UL) << S2_PGDIR_SHIFT)
> +#define S2_PGDIR_MASK (~(S2_PGDIR_SIZE - 1))
> +
> +/* We can have concatenated tables at stage2 entry. */

I'm not sure if the comment is helpful. How about:

/*
* The number of PTRS across all concatenated stage2 tables given by the
* number of bits resolved at the initial level.
*/

> +#define PTRS_PER_S2_PGD (1 << (KVM_PHYS_SHIFT - S2_PGDIR_SHIFT))
> +
> +/*
> + * KVM_MMU_CACHE_MIN_PAGES is the number of stage2 page table translation
> + * levels in addition to the PGD.
> + */
> +#define KVM_MMU_CACHE_MIN_PAGES (STAGE2_PGTABLE_LEVELS - 1)
> +
> +
> +#if STAGE2_PGTABLE_LEVELS > 3
> +
> +#define S2_PUD_SHIFT ARM64_HW_PGTABLE_LEVEL_SHIFT(1)
> +#define S2_PUD_SIZE (_AC(1, UL) << S2_PUD_SHIFT)
> +#define S2_PUD_MASK (~(S2_PUD_SIZE - 1))
> +
> #define stage2_pgd_none(pgd) pgd_none(pgd)
> #define stage2_pgd_clear(pgd) pgd_clear(pgd)
> #define stage2_pgd_present(pgd) pgd_present(pgd)
> @@ -55,6 +78,23 @@
> #define stage2_pud_offset(pgd, address) pud_offset(pgd, address)
> #define stage2_pud_free(mm, pud) pud_free(mm, pud)
>
> +#define stage2_pud_table_empty(pudp) kvm_page_empty(pudp)
> +
> +static inline phys_addr_t stage2_pud_addr_end(phys_addr_t addr, phys_addr_t end)
> +{
> + phys_addr_t boundary = (addr + S2_PUD_SIZE) & S2_PUD_MASK;
> + return (boundary - 1 < end - 1) ? boundary : end;
> +}
> +
> +#endif /* STAGE2_PGTABLE_LEVELS > 3 */
> +
> +
> +#if STAGE2_PGTABLE_LEVELS > 2
> +
> +#define S2_PMD_SHIFT ARM64_HW_PGTABLE_LEVEL_SHIFT(2)
> +#define S2_PMD_SIZE (_AC(1, UL) << S2_PMD_SHIFT)
> +#define S2_PMD_MASK (~(S2_PMD_SIZE - 1))
> +
> #define stage2_pud_none(pud) pud_none(pud)
> #define stage2_pud_clear(pud) pud_clear(pud)
> #define stage2_pud_present(pud) pud_present(pud)
> @@ -63,24 +103,31 @@
> #define stage2_pmd_free(mm, pmd) pmd_free(mm, pmd)
>
> #define stage2_pud_huge(pud) pud_huge(pud)
> +#define stage2_pmd_table_empty(pmdp) kvm_page_empty(pmdp)
> +
> +static inline phys_addr_t stage2_pmd_addr_end(phys_addr_t addr, phys_addr_t end)
> +{
> + phys_addr_t boundary = (addr + S2_PMD_SIZE) & S2_PMD_MASK;
> + return (boundary - 1 < end - 1) ? boundary : end;
> +}
>
> -#define stage2_pgd_addr_end(address, end) pgd_addr_end(address, end)
> -#define stage2_pud_addr_end(address, end) pud_addr_end(address, end)
> -#define stage2_pmd_addr_end(address, end) pmd_addr_end(address, end)
> +#endif /* STAGE2_PGTABLE_LEVELS > 2 */
>
> #define stage2_pte_table_empty(ptep) kvm_page_empty(ptep)
> -#ifdef __PGTABLE_PMD_FOLDED
> -#define stage2_pmd_table_empty(pmdp) (0)
> -#else
> -#define stage2_pmd_table_empty(pmdp) ((KVM_PREALLOC_LEVEL < 2) && kvm_page_empty(pmdp))
> -#endif
>
> -#ifdef __PGTABLE_PUD_FOLDED
> -#define stage2_pud_table_empty(pudp) (0)
> -#else
> -#define stage2_pud_table_empty(pudp) ((KVM_PREALLOC_LEVEL < 1) && kvm_page_empty(pudp))
> +#if STAGE2_PGTABLE_LEVELS == 2
> +#include <asm/stage2_pgtable-nopmd.h>
> +#elif STAGE2_PGTABLE_LEVELS == 3
> +#include <asm/stage2_pgtable-nopud.h>
> #endif
>
> -#define stage2_pgd_index(addr) (((addr) >> PGDIR_SHIFT) & (PTRS_PER_S2_PGD - 1))
> +
> +#define stage2_pgd_index(addr) (((addr) >> S2_PGDIR_SHIFT) & (PTRS_PER_S2_PGD - 1))
> +
> +static inline phys_addr_t stage2_pgd_addr_end(phys_addr_t addr, phys_addr_t end)
> +{
> + phys_addr_t boundary = (addr + S2_PGDIR_SIZE) & S2_PGDIR_MASK;
> + return (boundary - 1 < end - 1) ? boundary : end;
> +}
>
> #endif /* __ARM64_S2_PGTABLE_H_ */
> --
> 1.7.9.5
>

So, I only commented on the comments here (to prove that I actually read
the patch ;) ), but the code looks really good!

Thanks,
-Christoffer