Message-ID: <CAHLR_paUwXFUGnJSHSkFd0PnSStrNFkOXc1CLzH2g=qFv7SgNg@mail.gmail.com>
Date: Tue, 19 Dec 2023 11:44:10 +0000
From: Sebastian Ene <sebastianene@...gle.com>
To: will@...nel.org, Oliver Upton <oliver.upton@...ux.dev>, 
	James Morse <james.morse@....com>, Suzuki K Poulose <suzuki.poulose@....com>, 
	Zenghui Yu <yuzenghui@...wei.com>, catalin.marinas@....com, mark.rutland@....com, 
	akpm@...ux-foundation.org, maz@...nel.org
Cc: kvmarm@...ts.linux.dev, linux-arm-kernel@...ts.infradead.org, 
	linux-kernel@...r.kernel.org, kernel-team@...roid.com, vdonnefort@...gle.com, 
	qperret@...gle.com, smostafa@...gle.com
Subject: Re: [PATCH v4 01/10] KVM: arm64: Add snapshot interface for the host
 stage-2 pagetable

On Mon, Dec 18, 2023 at 01:58:51PM +0000, Sebastian Ene wrote:
> Introduce a new hvc that allows the pKVM hypervisor to snapshot the host
> stage-2 pagetables. The caller is expected to allocate & free the memory
> for the pagetable snapshot. The below diagram shows the sequence of
> events.
>
> [--- HOST ---]
> (1) allocate memory for the snapshot
> (2) invoke the __pkvm_copy_host_stage2(snapshot) interface
> |____
>       [--- HYP ---]
>       (3) donate the snapshot from HOST -> HYP
>       (4) save the host stage-2 pagetables in the snapshot
>       (5) donate the snapshot from HYP -> HOST
>       [--- return from HYP ---]
> [--- HOST ---]
> (6) free the memory for the snapshot
>
> When the host kernel invokes this interface, the memory is donated from
> the host to the hypervisor. By doing this the memory is unmapped from the
> host's translation regime to prevent the host from modifying the buffer
> while the hypervisor uses it. The hypervisor walks the host stage-2
> pagetable copy while re-creating the original mappings. When the copying
> is done, the memory is donated from the hypervisor back to the host.
> The pages that have been used from the memcache are donated back to the
> host from a temporary array. The hvc is executing synchronously and the
> preemption is disabled for the current CPU while running inside the
> hypervisor so calling this interface blocks the current CPU until the
> snapshoting is done.
>
> Signed-off-by: Sebastian Ene <sebastianene@...gle.com>
> ---
>  arch/arm64/include/asm/kvm_asm.h              |   1 +
>  arch/arm64/include/asm/kvm_pgtable.h          |  43 ++++
>  arch/arm64/kvm/hyp/include/nvhe/mem_protect.h |   1 +
>  arch/arm64/kvm/hyp/nvhe/hyp-main.c            |  12 ++
>  arch/arm64/kvm/hyp/nvhe/mem_protect.c         | 186 ++++++++++++++++++
>  arch/arm64/kvm/hyp/pgtable.c                  |  43 ++++
>  6 files changed, 286 insertions(+)
>
> diff --git a/arch/arm64/include/asm/kvm_asm.h b/arch/arm64/include/asm/kvm_asm.h
> index 24b5e6b23..9df3367d8 100644
> --- a/arch/arm64/include/asm/kvm_asm.h
> +++ b/arch/arm64/include/asm/kvm_asm.h
> @@ -81,6 +81,7 @@ enum __kvm_host_smccc_func {
>       __KVM_HOST_SMCCC_FUNC___pkvm_init_vm,
>       __KVM_HOST_SMCCC_FUNC___pkvm_init_vcpu,
>       __KVM_HOST_SMCCC_FUNC___pkvm_teardown_vm,
> +     __KVM_HOST_SMCCC_FUNC___pkvm_host_stage2_snapshot,
>  };
>
>  #define DECLARE_KVM_VHE_SYM(sym)     extern char sym[]
> diff --git a/arch/arm64/include/asm/kvm_pgtable.h b/arch/arm64/include/asm/kvm_pgtable.h
> index d3e354bb8..f73efd8a8 100644
> --- a/arch/arm64/include/asm/kvm_pgtable.h
> +++ b/arch/arm64/include/asm/kvm_pgtable.h
> @@ -10,6 +10,7 @@
>  #include <linux/bits.h>
>  #include <linux/kvm_host.h>
>  #include <linux/types.h>
> +#include <asm/kvm_host.h>
>
>  #define KVM_PGTABLE_MAX_LEVELS               4U
>
> @@ -351,6 +352,28 @@ struct kvm_pgtable {
>       kvm_pgtable_force_pte_cb_t              force_pte_cb;
>  };
>
> +/**
> + * struct kvm_pgtable_snapshot - Snapshot page-table wrapper.
> + * @pgtable:         The page-table configuration.
> + * @mc:                      Memcache used for pagetable pages allocation.
> + * @pgd_hva:         Host virtual address of a physically contiguous buffer
> + *                   used for storing the PGD.
> + * @pgd_pages:               The size of the phyisically contiguous buffer in pages.
> + * @used_pages_hva:  Host virtual address of a physically contiguous buffer
> + *                   used for storing the consumed pages from the memcache.
> + * @num_used_pages           The size of the used buffer in pages.
> + * @used_pages_indx          The current index of the used pages array.
> + */
> +struct kvm_pgtable_snapshot {
> +     struct kvm_pgtable                      pgtable;
> +     struct kvm_hyp_memcache                 mc;
> +     void                                    *pgd_hva;
> +     size_t                                  pgd_pages;
> +     phys_addr_t                             *used_pages_hva;
> +     size_t                                  num_used_pages;
> +     size_t                                  used_pages_indx;
> +};
> +
>  /**
>   * kvm_pgtable_hyp_init() - Initialise a hypervisor stage-1 page-table.
>   * @pgt:     Uninitialised page-table structure to initialise.
> @@ -756,4 +779,24 @@ enum kvm_pgtable_prot kvm_pgtable_hyp_pte_prot(kvm_pte_t pte);
>   */
>  void kvm_tlb_flush_vmid_range(struct kvm_s2_mmu *mmu,
>                               phys_addr_t addr, size_t size);
> +
> +#ifdef CONFIG_NVHE_EL2_DEBUG
> +/**
> + * kvm_pgtable_stage2_snapshot() - Given a memcache and a destination
> + *                           pagetable where we have the original PGD
> + *                           copied, build a snapshot table with page table
> + *                           pages from the given memcache.
> + *
> + * @to_pgt:  Destination pagetable
> + * @mc:              The memcache where we allocate the destination pagetables from
> + */
> +int kvm_pgtable_stage2_snapshot(struct kvm_pgtable *to_pgt,
> +                             void *mc);
> +#else
> +static inline int kvm_pgtable_stage2_snapshot(struct kvm_pgtable *to_pgt,
> +                                           void *mc)
> +{
> +     return -EPERM;
> +}
> +#endif       /* CONFIG_NVHE_EL2_DEBUG */
>  #endif       /* __ARM64_KVM_PGTABLE_H__ */
> diff --git a/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h b/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
> index 0972faccc..ca8f76915 100644
> --- a/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
> +++ b/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
> @@ -69,6 +69,7 @@ int __pkvm_host_donate_hyp(u64 pfn, u64 nr_pages);
>  int __pkvm_hyp_donate_host(u64 pfn, u64 nr_pages);
>  int __pkvm_host_share_ffa(u64 pfn, u64 nr_pages);
>  int __pkvm_host_unshare_ffa(u64 pfn, u64 nr_pages);
> +int __pkvm_host_stage2_snapshot(struct kvm_pgtable_snapshot *snapshot);
>
>  bool addr_is_memory(phys_addr_t phys);
>  int host_stage2_idmap_locked(phys_addr_t addr, u64 size, enum kvm_pgtable_prot prot);
> diff --git a/arch/arm64/kvm/hyp/nvhe/hyp-main.c b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> index 2385fd03e..7b215245f 100644
> --- a/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> +++ b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> @@ -314,6 +314,17 @@ static void handle___pkvm_teardown_vm(struct kvm_cpu_context *host_ctxt)
>       cpu_reg(host_ctxt, 1) = __pkvm_teardown_vm(handle);
>  }
>
> +static void handle___pkvm_host_stage2_snapshot(struct kvm_cpu_context *host_ctxt)
> +{
> +#ifdef CONFIG_NVHE_EL2_DEBUG
> +     DECLARE_REG(struct kvm_pgtable_snapshot *, snapshot_hva, host_ctxt, 1);
> +
> +     cpu_reg(host_ctxt, 1) = __pkvm_host_stage2_snapshot(snapshot_hva);
> +#else
> +     cpu_reg(host_ctxt, 0) = SMCCC_RET_NOT_SUPPORTED;
> +#endif
> +}
> +
>  typedef void (*hcall_t)(struct kvm_cpu_context *);
>
>  #define HANDLE_FUNC(x)       [__KVM_HOST_SMCCC_FUNC_##x] = (hcall_t)handle_##x
> @@ -348,6 +359,7 @@ static const hcall_t host_hcall[] = {
>       HANDLE_FUNC(__pkvm_init_vm),
>       HANDLE_FUNC(__pkvm_init_vcpu),
>       HANDLE_FUNC(__pkvm_teardown_vm),
> +     HANDLE_FUNC(__pkvm_host_stage2_snapshot),
>  };
>
>  static void handle_host_hcall(struct kvm_cpu_context *host_ctxt)
> diff --git a/arch/arm64/kvm/hyp/nvhe/mem_protect.c b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> index 8d0a5834e..aaf07f9e1 100644
> --- a/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> +++ b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> @@ -266,6 +266,192 @@ int kvm_guest_prepare_stage2(struct pkvm_hyp_vm *vm, void *pgd)
>       return 0;
>  }
>
> +#ifdef CONFIG_NVHE_EL2_DEBUG
> +static void *snap_zalloc_page(void *mc)
> +{
> +     struct hyp_page *p;
> +     void *addr;
> +     struct kvm_pgtable_snapshot *snap;
> +     phys_addr_t *used_pg;
> +
> +     snap = container_of(mc, struct kvm_pgtable_snapshot, mc);
> +     used_pg = kern_hyp_va(snap->used_pages_hva);
> +
> +     /* Check if we have space to track the used page */
> +     if (snap->used_pages_indx * sizeof(phys_addr_t) > snap->num_used_pages * PAGE_SIZE)

This should be '>=' instead, and I will fix this.


> +             return NULL;
> +
> +     addr = pop_hyp_memcache(mc, hyp_phys_to_virt);
> +     if (!addr)
> +             return addr;
> +     used_pg[snap->used_pages_indx++] = hyp_virt_to_phys(addr);
> +
> +     memset(addr, 0, PAGE_SIZE);
> +     p = hyp_virt_to_page(addr);
> +     memset(p, 0, sizeof(*p));
> +
> +     return addr;
> +}
> +
> +static void snap_free_pages_exact(void *addr, unsigned long size)
> +{
> +     u8 order = get_order(size);
> +     unsigned int i;
> +     struct hyp_page *p;
> +
> +     for (i = 0; i < (1 << order); i++) {
> +             p = hyp_virt_to_page(addr + (i * PAGE_SIZE));
> +             hyp_page_ref_dec(p);
> +     }
> +}
> +
> +static void *pkvm_setup_snapshot(struct kvm_pgtable_snapshot *snap_hva)
> +{
> +     unsigned long i;
> +     void *pgd, *used_pg;
> +     phys_addr_t mc_page, next_mc_page;
> +     struct kvm_pgtable_snapshot *snap;
> +
> +     snap = (void *)kern_hyp_va(snap_hva);
> +     if (!PAGE_ALIGNED(snap))
> +             return NULL;
> +
> +     if (__pkvm_host_donate_hyp(hyp_virt_to_pfn(snap), 1))
> +             return NULL;
> +
> +     pgd = kern_hyp_va(snap->pgd_hva);
> +     if (!PAGE_ALIGNED(pgd))
> +             goto error_with_snapshot;
> +
> +     if (__pkvm_host_donate_hyp(hyp_virt_to_pfn(pgd), snap->pgd_pages))
> +             goto error_with_snapshot;
> +
> +     mc_page = snap->mc.head;
> +     for (i = 0; i < snap->mc.nr_pages; i++) {
> +             if (!PAGE_ALIGNED(mc_page))
> +                     goto error_with_memcache;
> +
> +             if (__pkvm_host_donate_hyp(hyp_phys_to_pfn(mc_page), 1))
> +                     goto error_with_memcache;
> +
> +             mc_page = *((phys_addr_t *)hyp_phys_to_virt(mc_page));
> +     }
> +
> +     used_pg = kern_hyp_va(snap->used_pages_hva);
> +     if (!PAGE_ALIGNED(used_pg))
> +             goto error_with_memcache;
> +
> +     if (__pkvm_host_donate_hyp(hyp_virt_to_pfn(used_pg), snap->num_used_pages))
> +             goto error_with_memcache;
> +
> +     return snap;
> +error_with_memcache:
> +     mc_page = snap->mc.head;
> +     for (; i >= 0; i--) {
> +             next_mc_page = *((phys_addr_t *)hyp_phys_to_virt(mc_page));
> +             WARN_ON(__pkvm_hyp_donate_host(hyp_phys_to_pfn(mc_page), 1));
> +             mc_page = next_mc_page;
> +     }
> +
> +     WARN_ON(__pkvm_hyp_donate_host(hyp_virt_to_pfn(pgd), snap->pgd_pages));
> +error_with_snapshot:
> +     WARN_ON(__pkvm_hyp_donate_host(hyp_virt_to_pfn(snap), 1));
> +     return NULL;
> +}
> +
> +static void pkvm_teardown_snapshot(struct kvm_pgtable_snapshot *snap)
> +{
> +     size_t i;
> +     phys_addr_t mc_page, next_mc_page;
> +     u64 *used_pg = kern_hyp_va(snap->used_pages_hva);
> +     void *pgd = kern_hyp_va(snap->pgd_hva);
> +
> +     for (i = 0; i < snap->used_pages_indx; i++) {
> +             mc_page = used_pg[i];
> +             WARN_ON(__pkvm_hyp_donate_host(hyp_phys_to_pfn(mc_page), 1));
> +     }
> +
> +     WARN_ON(__pkvm_hyp_donate_host(hyp_virt_to_pfn(used_pg),
> +                                    snap->num_used_pages));
> +
> +     mc_page = snap->mc.head;
> +     for (i = 0; i < snap->mc.nr_pages; i++) {
> +             next_mc_page = *((phys_addr_t *)hyp_phys_to_virt(mc_page));
> +             WARN_ON(__pkvm_hyp_donate_host(hyp_phys_to_pfn(mc_page), 1));
> +             mc_page = next_mc_page;
> +     }
> +
> +     snap->pgtable.mm_ops = NULL;
> +     WARN_ON(__pkvm_hyp_donate_host(hyp_virt_to_pfn(pgd), snap->pgd_pages));
> +     WARN_ON(__pkvm_hyp_donate_host(hyp_virt_to_pfn(snap), 1));
> +}
> +
> +static int pkvm_host_stage2_snapshot(struct kvm_pgtable_snapshot *snap)
> +{
> +     int ret;
> +     void *pgd;
> +     size_t required_pgd_len;
> +     struct kvm_pgtable_mm_ops mm_ops = {0};
> +     struct kvm_s2_mmu *mmu;
> +     struct kvm_pgtable *to_pgt, *from_pgt;
> +
> +     if (snap->used_pages_indx != 0)
> +             return -EINVAL;
> +
> +     from_pgt = &host_mmu.pgt;
> +     mmu = &host_mmu.arch.mmu;
> +     required_pgd_len = kvm_pgtable_stage2_pgd_size(mmu->vtcr);
> +     if (snap->pgd_pages < (required_pgd_len >> PAGE_SHIFT))
> +             return -ENOMEM;
> +
> +     if (snap->mc.nr_pages < host_s2_pgtable_pages())
> +             return -ENOMEM;
> +
> +     to_pgt = &snap->pgtable;
> +     pgd = kern_hyp_va(snap->pgd_hva);
> +
> +     mm_ops.zalloc_page              = snap_zalloc_page;
> +     mm_ops.free_pages_exact         = snap_free_pages_exact;
> +     mm_ops.phys_to_virt             = hyp_phys_to_virt;
> +     mm_ops.virt_to_phys             = hyp_virt_to_phys;
> +     mm_ops.page_count               = hyp_page_count;
> +
> +     to_pgt->ia_bits         = from_pgt->ia_bits;
> +     to_pgt->start_level     = from_pgt->start_level;
> +     to_pgt->flags           = from_pgt->flags;
> +     to_pgt->mm_ops          = &mm_ops;
> +
> +     host_lock_component();
> +
> +     to_pgt->pgd = pgd;
> +     memcpy(to_pgt->pgd, from_pgt->pgd, required_pgd_len);
> +     ret = kvm_pgtable_stage2_snapshot(to_pgt, &snap->mc);
> +
> +     host_unlock_component();
> +
> +     return ret;
> +}
> +
> +int __pkvm_host_stage2_snapshot(struct kvm_pgtable_snapshot *snap_hva)
> +{
> +     int ret;
> +     struct kvm_pgtable_snapshot *snap;
> +     kvm_pteref_t pgd;
> +
> +     snap = pkvm_setup_snapshot(snap_hva);
> +     if (!snap)
> +             return -EPERM;
> +
> +     ret = pkvm_host_stage2_snapshot(snap);
> +     if (!ret) {
> +             pgd = snap->pgtable.pgd;
> +             snap->pgtable.pgd = (kvm_pteref_t)__hyp_pa(pgd);
> +     }
> +     pkvm_teardown_snapshot(snap);
> +     return ret;
> +}
> +#endif /* CONFIG_NVHE_EL2_DEBUG */
> +
>  void reclaim_guest_pages(struct pkvm_hyp_vm *vm, struct kvm_hyp_memcache *mc)
>  {
>       void *addr;
> diff --git a/arch/arm64/kvm/hyp/pgtable.c b/arch/arm64/kvm/hyp/pgtable.c
> index 1966fdee7..82fef9620 100644
> --- a/arch/arm64/kvm/hyp/pgtable.c
> +++ b/arch/arm64/kvm/hyp/pgtable.c
> @@ -1598,3 +1598,46 @@ void kvm_pgtable_stage2_free_unlinked(struct kvm_pgtable_mm_ops *mm_ops, void *p
>       WARN_ON(mm_ops->page_count(pgtable) != 1);
>       mm_ops->put_page(pgtable);
>  }
> +
> +#ifdef CONFIG_NVHE_EL2_DEBUG
> +static int snapshot_walker(const struct kvm_pgtable_visit_ctx *ctx,
> +                        enum kvm_pgtable_walk_flags visit)
> +{
> +     struct kvm_pgtable_mm_ops *mm_ops = ctx->mm_ops;
> +     void *copy_table, *original_addr;
> +     kvm_pte_t new = ctx->old;
> +
> +     if (!stage2_pte_is_counted(ctx->old))
> +             return 0;
> +
> +     if (kvm_pte_table(ctx->old, ctx->level)) {
> +             copy_table = mm_ops->zalloc_page(ctx->arg);
> +             if (!copy_table)
> +                     return -ENOMEM;
> +
> +             original_addr = kvm_pte_follow(ctx->old, mm_ops);
> +
> +             memcpy(copy_table, original_addr, PAGE_SIZE);
> +             new = kvm_init_table_pte(copy_table, mm_ops);
> +     }
> +
> +     *ctx->ptep = new;
> +
> +     return 0;
> +}
> +
> +int kvm_pgtable_stage2_snapshot(struct kvm_pgtable *to_pgt, void *mc)
> +{
> +     struct kvm_pgtable_walker walker = {
> +             .cb     = snapshot_walker,
> +             .flags  = KVM_PGTABLE_WALK_LEAF |
> +                       KVM_PGTABLE_WALK_TABLE_PRE,
> +             .arg = mc
> +     };
> +
> +     if (!to_pgt->pgd)
> +             return -ENOMEM;
> +
> +     return kvm_pgtable_walk(to_pgt, 0, BIT(to_pgt->ia_bits), &walker);
> +}
> +#endif /* CONFIG_NVHE_EL2_DEBUG */
> --
> 2.43.0.472.g3155946c3a-goog
>

Powered by blists - more mailing lists