| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <3f0caace-ee05-4ddf-ae75-2157e77aa57c@redhat.com>
Date: Thu, 30 Jan 2025 15:22:41 +1000
From: Gavin Shan <gshan@...hat.com>
To: Steven Price <steven.price@....com>, kvm@...r.kernel.org,
kvmarm@...ts.linux.dev
Cc: Catalin Marinas <catalin.marinas@....com>, Marc Zyngier <maz@...nel.org>,
Will Deacon <will@...nel.org>, James Morse <james.morse@....com>,
Oliver Upton <oliver.upton@...ux.dev>,
Suzuki K Poulose <suzuki.poulose@....com>, Zenghui Yu
<yuzenghui@...wei.com>, linux-arm-kernel@...ts.infradead.org,
linux-kernel@...r.kernel.org, Joey Gouly <joey.gouly@....com>,
Alexandru Elisei <alexandru.elisei@....com>,
Christoffer Dall <christoffer.dall@....com>, Fuad Tabba <tabba@...gle.com>,
linux-coco@...ts.linux.dev,
Ganapatrao Kulkarni <gankulkarni@...amperecomputing.com>,
Shanker Donthineni <sdonthineni@...dia.com>, Alper Gun
<alpergun@...gle.com>, "Aneesh Kumar K . V" <aneesh.kumar@...nel.org>
Subject: Re: [PATCH v6 20/43] arm64: RME: Runtime faulting of memory
On 12/13/24 1:55 AM, Steven Price wrote:
> At runtime if the realm guest accesses memory which hasn't yet been
> mapped then KVM needs to either populate the region or fault the guest.
>
> For memory in the lower (protected) region of IPA a fresh page is
> provided to the RMM which will zero the contents. For memory in the
> upper (shared) region of IPA, the memory from the memslot is mapped
> into the realm VM non secure.
>
> Signed-off-by: Steven Price <steven.price@....com>
> ---
> Changes since v5:
> * Reduce use of struct page in preparation for supporting the RMM
> having a different page size to the host.
> * Handle a race when delegating a page where another CPU has faulted on
> a the same page (and already delegated the physical page) but not yet
> mapped it. In this case simply return to the guest to either use the
> mapping from the other CPU (or refault if the race is lost).
> * The changes to populate_par_region() are moved into the previous
> patch where they belong.
> Changes since v4:
> * Code cleanup following review feedback.
> * Drop the PTE_SHARED bit when creating unprotected page table entries.
> This is now set by the RMM and the host has no control of it and the
> spec requires the bit to be set to zero.
> Changes since v2:
> * Avoid leaking memory if failing to map it in the realm.
> * Correctly mask RTT based on LPA2 flag (see rtt_get_phys()).
> * Adapt to changes in previous patches.
> ---
> arch/arm64/include/asm/kvm_emulate.h | 10 ++
> arch/arm64/include/asm/kvm_rme.h | 10 ++
> arch/arm64/kvm/mmu.c | 124 +++++++++++++++++++--
> arch/arm64/kvm/rme.c | 156 +++++++++++++++++++++++++++
> 4 files changed, 293 insertions(+), 7 deletions(-)
>
> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
> index ec2b6d9c9c07..b13e367b6972 100644
> --- a/arch/arm64/include/asm/kvm_emulate.h
> +++ b/arch/arm64/include/asm/kvm_emulate.h
> @@ -720,6 +720,16 @@ static inline bool kvm_realm_is_created(struct kvm *kvm)
> return kvm_is_realm(kvm) && kvm_realm_state(kvm) != REALM_STATE_NONE;
> }
>
> +static inline gpa_t kvm_gpa_from_fault(struct kvm *kvm, phys_addr_t fault_ipa)
> +{
> + if (kvm_is_realm(kvm)) {
> + struct realm *realm = &kvm->arch.realm;
> +
> + return fault_ipa & ~BIT(realm->ia_bits - 1);
> + }
> + return fault_ipa;
> +}
> +
'fault' has been included in 'kvm_gpa_from_fault' and 'fault_ipa'. To avoid the
duplication, 'fault_ipa' can be renamed to 'ipa'.
> static inline bool vcpu_is_rec(struct kvm_vcpu *vcpu)
> {
> if (static_branch_unlikely(&kvm_rme_is_available))
> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
> index 0410650cd545..158f77e24a26 100644
> --- a/arch/arm64/include/asm/kvm_rme.h
> +++ b/arch/arm64/include/asm/kvm_rme.h
> @@ -99,6 +99,16 @@ void kvm_realm_unmap_range(struct kvm *kvm,
> unsigned long ipa,
> u64 size,
> bool unmap_private);
> +int realm_map_protected(struct realm *realm,
> + unsigned long base_ipa,
> + kvm_pfn_t pfn,
> + unsigned long size,
> + struct kvm_mmu_memory_cache *memcache);
> +int realm_map_non_secure(struct realm *realm,
> + unsigned long ipa,
> + kvm_pfn_t pfn,
> + unsigned long size,
> + struct kvm_mmu_memory_cache *memcache);
> int realm_set_ipa_state(struct kvm_vcpu *vcpu,
> unsigned long addr, unsigned long end,
> unsigned long ripas,
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index b100d4b3aa29..e88714903ce5 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -325,8 +325,13 @@ static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64
>
> lockdep_assert_held_write(&kvm->mmu_lock);
> WARN_ON(size & ~PAGE_MASK);
> - WARN_ON(stage2_apply_range(mmu, start, end, kvm_pgtable_stage2_unmap,
> - may_block));
> +
> + if (kvm_is_realm(kvm))
> + kvm_realm_unmap_range(kvm, start, size, !only_shared);
> + else
> + WARN_ON(stage2_apply_range(mmu, start, end,
> + kvm_pgtable_stage2_unmap,
> + may_block));
> }
>
> void kvm_stage2_unmap_range(struct kvm_s2_mmu *mmu, phys_addr_t start,
> @@ -346,7 +351,10 @@ static void stage2_flush_memslot(struct kvm *kvm,
> phys_addr_t addr = memslot->base_gfn << PAGE_SHIFT;
> phys_addr_t end = addr + PAGE_SIZE * memslot->npages;
>
> - kvm_stage2_flush_range(&kvm->arch.mmu, addr, end);
> + if (kvm_is_realm(kvm))
> + kvm_realm_unmap_range(kvm, addr, end - addr, false);
> + else
> + kvm_stage2_flush_range(&kvm->arch.mmu, addr, end);
> }
>
> /**
> @@ -1037,6 +1045,10 @@ void stage2_unmap_vm(struct kvm *kvm)
> struct kvm_memory_slot *memslot;
> int idx, bkt;
>
> + /* For realms this is handled by the RMM so nothing to do here */
> + if (kvm_is_realm(kvm))
> + return;
> +
> idx = srcu_read_lock(&kvm->srcu);
> mmap_read_lock(current->mm);
> write_lock(&kvm->mmu_lock);
> @@ -1062,6 +1074,7 @@ void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
> if (kvm_is_realm(kvm) &&
> (kvm_realm_state(kvm) != REALM_STATE_DEAD &&
> kvm_realm_state(kvm) != REALM_STATE_NONE)) {
> + kvm_stage2_unmap_range(mmu, 0, (~0ULL) & PAGE_MASK, false);
> write_unlock(&kvm->mmu_lock);
> kvm_realm_destroy_rtts(kvm, pgt->ia_bits);
>
> @@ -1446,6 +1459,76 @@ static bool kvm_vma_mte_allowed(struct vm_area_struct *vma)
> return vma->vm_flags & VM_MTE_ALLOWED;
> }
>
> +static int realm_map_ipa(struct kvm *kvm, phys_addr_t ipa,
> + kvm_pfn_t pfn, unsigned long map_size,
> + enum kvm_pgtable_prot prot,
> + struct kvm_mmu_memory_cache *memcache)
> +{
> + struct realm *realm = &kvm->arch.realm;
> +
> + if (WARN_ON(!(prot & KVM_PGTABLE_PROT_W)))
> + return -EFAULT;
> +
> + if (!realm_is_addr_protected(realm, ipa))
> + return realm_map_non_secure(realm, ipa, pfn, map_size,
> + memcache);
> +
> + return realm_map_protected(realm, ipa, pfn, map_size, memcache);
> +}
> +
> +static int private_memslot_fault(struct kvm_vcpu *vcpu,
> + phys_addr_t fault_ipa,
> + struct kvm_memory_slot *memslot)
> +{
> + struct kvm *kvm = vcpu->kvm;
> + gpa_t gpa = kvm_gpa_from_fault(kvm, fault_ipa);
> + gfn_t gfn = gpa >> PAGE_SHIFT;
> + bool priv_exists = kvm_mem_is_private(kvm, gfn);
> + struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
> + struct page *page;
> + kvm_pfn_t pfn;
> + int ret;
> + /*
> + * For Realms, the shared address is an alias of the private GPA with
> + * the top bit set. Thus is the fault address matches the GPA then it
> + * is the private alias.
> + */
> + bool is_priv_gfn = (gpa == fault_ipa);
> +
We may rename 'priv_exists' to 'was_priv_gfn', which is consistent to 'is_priv_gfn'.
Alternatively, we may use 'was_private' and 'is_private'.
> + if (priv_exists != is_priv_gfn) {
> + kvm_prepare_memory_fault_exit(vcpu,
> + gpa,
> + PAGE_SIZE,
> + kvm_is_write_fault(vcpu),
> + false, is_priv_gfn);
> +
> + return -EFAULT;
> + }
> +
> + if (!is_priv_gfn) {
> + /* Not a private mapping, handling normally */
> + return -EINVAL;
> + }
> +
> + ret = kvm_mmu_topup_memory_cache(memcache,
> + kvm_mmu_cache_min_pages(vcpu->arch.hw_mmu));
> + if (ret)
> + return ret;
> +
> + ret = kvm_gmem_get_pfn(kvm, memslot, gfn, &pfn, &page, NULL);
> + if (ret)
> + return ret;
> +
> + /* FIXME: Should be able to use bigger than PAGE_SIZE mappings */
> + ret = realm_map_ipa(kvm, fault_ipa, pfn, PAGE_SIZE, KVM_PGTABLE_PROT_W,
> + memcache);
> + if (!ret)
> + return 1; /* Handled */
> +
> + put_page(page);
> + return ret;
> +}
> +
> static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> struct kvm_s2_trans *nested,
> struct kvm_memory_slot *memslot, unsigned long hva,
> @@ -1472,6 +1555,14 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> if (fault_is_perm)
> fault_granule = kvm_vcpu_trap_get_perm_fault_granule(vcpu);
> write_fault = kvm_is_write_fault(vcpu);
> +
> + /*
> + * Realms cannot map protected pages read-only
> + * FIXME: It should be possible to map unprotected pages read-only
> + */
> + if (vcpu_is_rec(vcpu))
> + write_fault = true;
> +
> exec_fault = kvm_vcpu_trap_is_exec_fault(vcpu);
> VM_BUG_ON(write_fault && exec_fault);
>
> @@ -1579,7 +1670,7 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> ipa &= ~(vma_pagesize - 1);
> }
>
> - gfn = ipa >> PAGE_SHIFT;
> + gfn = kvm_gpa_from_fault(kvm, ipa) >> PAGE_SHIFT;
> mte_allowed = kvm_vma_mte_allowed(vma);
>
> vfio_allow_any_uc = vma->vm_flags & VM_ALLOW_ANY_UNCACHED;
> @@ -1660,7 +1751,8 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> * If we are not forced to use page mapping, check if we are
> * backed by a THP and thus use block mapping if possible.
> */
> - if (vma_pagesize == PAGE_SIZE && !(force_pte || device)) {
> + /* FIXME: We shouldn't need to disable this for realms */
> + if (vma_pagesize == PAGE_SIZE && !(force_pte || device || kvm_is_realm(kvm))) {
> if (fault_is_perm && fault_granule > PAGE_SIZE)
> vma_pagesize = fault_granule;
> else
> @@ -1712,6 +1804,9 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
> */
> prot &= ~KVM_NV_GUEST_MAP_SZ;
> ret = kvm_pgtable_stage2_relax_perms(pgt, fault_ipa, prot);
> + } else if (kvm_is_realm(kvm)) {
> + ret = realm_map_ipa(kvm, fault_ipa, pfn, vma_pagesize,
> + prot, memcache);
> } else {
> ret = kvm_pgtable_stage2_map(pgt, fault_ipa, vma_pagesize,
> __pfn_to_phys(pfn), prot,
> @@ -1854,8 +1949,15 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
> nested = &nested_trans;
> }
>
> - gfn = ipa >> PAGE_SHIFT;
> + gfn = kvm_gpa_from_fault(vcpu->kvm, ipa) >> PAGE_SHIFT;
> memslot = gfn_to_memslot(vcpu->kvm, gfn);
> +
> + if (kvm_slot_can_be_private(memslot)) {
> + ret = private_memslot_fault(vcpu, ipa, memslot);
> + if (ret != -EINVAL)
> + goto out;
> + }
> +
> hva = gfn_to_hva_memslot_prot(memslot, gfn, &writable);
> write_fault = kvm_is_write_fault(vcpu);
> if (kvm_is_error_hva(hva) || (write_fault && !writable)) {
> @@ -1899,7 +2001,7 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
> * of the page size.
> */
> ipa |= kvm_vcpu_get_hfar(vcpu) & GENMASK(11, 0);
> - ret = io_mem_abort(vcpu, ipa);
> + ret = io_mem_abort(vcpu, kvm_gpa_from_fault(vcpu->kvm, ipa));
> goto out_unlock;
> }
>
> @@ -1947,6 +2049,10 @@ bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
> if (!kvm->arch.mmu.pgt)
> return false;
>
> + /* We don't support aging for Realms */
> + if (kvm_is_realm(kvm))
> + return true;
> +
> return kvm_pgtable_stage2_test_clear_young(kvm->arch.mmu.pgt,
> range->start << PAGE_SHIFT,
> size, true);
> @@ -1963,6 +2069,10 @@ bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
> if (!kvm->arch.mmu.pgt)
> return false;
>
> + /* We don't support aging for Realms */
> + if (kvm_is_realm(kvm))
> + return true;
> +
> return kvm_pgtable_stage2_test_clear_young(kvm->arch.mmu.pgt,
> range->start << PAGE_SHIFT,
> size, false);
> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
> index d4561e368cd5..146ef598a581 100644
> --- a/arch/arm64/kvm/rme.c
> +++ b/arch/arm64/kvm/rme.c
> @@ -602,6 +602,162 @@ static int fold_rtt(struct realm *realm, unsigned long addr, int level)
> return 0;
> }
>
> +int realm_map_protected(struct realm *realm,
> + unsigned long ipa,
> + kvm_pfn_t pfn,
> + unsigned long map_size,
> + struct kvm_mmu_memory_cache *memcache)
> +{
> + phys_addr_t phys = __pfn_to_phys(pfn);
> + phys_addr_t rd = virt_to_phys(realm->rd);
> + unsigned long base_ipa = ipa;
> + unsigned long size;
> + int map_level;
> + int ret = 0;
> +
> + if (WARN_ON(!IS_ALIGNED(ipa, map_size)))
> + return -EINVAL;
> +
> + switch (map_size) {
> + case PAGE_SIZE:
> + map_level = 3;
> + break;
> + case RMM_L2_BLOCK_SIZE:
> + map_level = 2;
> + break;
> + default:
> + return -EINVAL;
> + }
> +
The same block of code, to return the RTT level according to the map size, has been
used for multiple times. It would be nice to introduce a helper for this.
> + if (map_level < RMM_RTT_MAX_LEVEL) {
> + /*
> + * A temporary RTT is needed during the map, precreate it,
> + * however if there is an error (e.g. missing parent tables)
> + * this will be handled below.
> + */
> + realm_create_rtt_levels(realm, ipa, map_level,
> + RMM_RTT_MAX_LEVEL, memcache);
> + }
> +
This block of code could be dropped. If the RTTs have been existing, realm_create_rtt_levels()
doesn't nothing, but several RMI calls are issued. RMI calls aren't cheap and it can cause
performance lost.
> + for (size = 0; size < map_size; size += PAGE_SIZE) {
> + if (rmi_granule_delegate(phys)) {
> + /*
> + * It's likely we raced with another VCPU on the same
> + * fault. Assume the other VCPU has handled the fault
> + * and return to the guest.
> + */
> + return 0;
> + }
We probably can't bail immediately when error is returned from rmi_granule_delegate()
because we intend to map a region whose size is 'map_size'. So a 'continue' instead
of 'return 0' seems correct to me.
> +
> + ret = rmi_data_create_unknown(rd, phys, ipa);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + /* Create missing RTTs and retry */
> + int level = RMI_RETURN_INDEX(ret);
> +
> + ret = realm_create_rtt_levels(realm, ipa, level,
> + RMM_RTT_MAX_LEVEL,
> + memcache);
> + WARN_ON(ret);
> + if (ret)
> + goto err_undelegate;
if (WARN_ON(ret))
> +
> + ret = rmi_data_create_unknown(rd, phys, ipa);
> + }
> + WARN_ON(ret);
> +
> + if (ret)
> + goto err_undelegate;
if (WARN_ON(ret))
> +
> + phys += PAGE_SIZE;
> + ipa += PAGE_SIZE;
> + }
> +
> + if (map_size == RMM_L2_BLOCK_SIZE)
> + ret = fold_rtt(realm, base_ipa, map_level);
> + if (WARN_ON(ret))
> + goto err;
> +
The nested if statements are needed here because the WARN_ON() only
take effect on the return value from fold_rtt().
if (map_size == RMM_L2_BLOCK_SIZE) {
ret = fold_rtt(realm, base_ipa, map_level);
if (WARN_ON(ret))
goto err;
}
> + return 0;
> +
> +err_undelegate:
> + if (WARN_ON(rmi_granule_undelegate(phys))) {
> + /* Page can't be returned to NS world so is lost */
> + get_page(phys_to_page(phys));
> + }
> +err:
> + while (size > 0) {
> + unsigned long data, top;
> +
> + phys -= PAGE_SIZE;
> + size -= PAGE_SIZE;
> + ipa -= PAGE_SIZE;
> +
> + WARN_ON(rmi_data_destroy(rd, ipa, &data, &top));
> +
> + if (WARN_ON(rmi_granule_undelegate(phys))) {
> + /* Page can't be returned to NS world so is lost */
> + get_page(phys_to_page(phys));
> + }
> + }
> + return -ENXIO;
> +}
> +
> +int realm_map_non_secure(struct realm *realm,
> + unsigned long ipa,
> + kvm_pfn_t pfn,
> + unsigned long map_size,
> + struct kvm_mmu_memory_cache *memcache)
> +{
> + phys_addr_t rd = virt_to_phys(realm->rd);
> + int map_level;
> + int ret = 0;
> + unsigned long desc = __pfn_to_phys(pfn) |
> + PTE_S2_MEMATTR(MT_S2_FWB_NORMAL) |
> + /* FIXME: Read+Write permissions for now */
> + (3 << 6);
> +
> + if (WARN_ON(!IS_ALIGNED(ipa, map_size)))
> + return -EINVAL;
> +
> + switch (map_size) {
> + case PAGE_SIZE:
> + map_level = 3;
> + break;
> + case RMM_L2_BLOCK_SIZE:
> + map_level = 2;
> + break;
> + default:
> + return -EINVAL;
> + }
> +
As above.
> + ret = rmi_rtt_map_unprotected(rd, ipa, map_level, desc);
> +
> + if (RMI_RETURN_STATUS(ret) == RMI_ERROR_RTT) {
> + /* Create missing RTTs and retry */
> + int level = RMI_RETURN_INDEX(ret);
> +
> + ret = realm_create_rtt_levels(realm, ipa, level, map_level,
> + memcache);
> + if (WARN_ON(ret))
> + return -ENXIO;
> +
> + ret = rmi_rtt_map_unprotected(rd, ipa, map_level, desc);
> + }
> + /*
> + * RMI_ERROR_RTT can be reported for two reasons: either the RTT tables
> + * are not there, or there is an RTTE already present for the address.
> + * The call to realm_create_rtt_levels() above handles the first case,
> + * and in the second case this indicates that another thread has
> + * already populated the RTTE for us, so we can ignore the error and
> + * continue.
> + */
> + if (WARN_ON(ret && RMI_RETURN_STATUS(ret) != RMI_ERROR_RTT))
> + return -ENXIO;
> +
> + return 0;
> +}
> +
> static int populate_par_region(struct kvm *kvm,
> phys_addr_t ipa_base,
> phys_addr_t ipa_end,
Thanks,
Gavin
Powered by blists - more mailing lists