[24/89] KVM: arm64: Return guest memory from EL2 via dedicated teardown memcache

Commit Message

Will Deacon May 19, 2022, 1:40 p.m. UTC

From: Quentin Perret <[e-mail]>

Rather than relying on the host to free the shadow VM pages explicitly
on teardown, introduce a dedicated teardown memcache which allows the
host to reclaim guest memory resources without having to keep track of
all of the allocations made by EL2.

Signed-off-by: Quentin Perret <[e-mail]>
---
 arch/arm64/include/asm/kvm_host.h             |  6 +-----
 arch/arm64/kvm/hyp/include/nvhe/mem_protect.h |  2 +-
 arch/arm64/kvm/hyp/nvhe/mem_protect.c         | 17 +++++++++++------
 arch/arm64/kvm/hyp/nvhe/pkvm.c                |  8 +++++++-
 arch/arm64/kvm/pkvm.c                         | 12 +-----------
 5 files changed, 21 insertions(+), 24 deletions(-)

Patch

diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index f4272ce76084..32ac88e60e6b 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -161,11 +161,7 @@  struct kvm_arch_memory_slot {
 struct kvm_protected_vm {
 	unsigned int shadow_handle;
 	struct mutex shadow_lock;
-
-	struct {
-		void *pgd;
-		void *shadow;
-	} hyp_donations;
+	struct kvm_hyp_memcache teardown_mc;
 };
 
 struct kvm_arch {
diff --git a/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h b/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
index 36eea31a1c5f..663019992b67 100644
--- a/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
+++ b/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
@@ -76,7 +76,7 @@  void handle_host_mem_abort(struct kvm_cpu_context *host_ctxt);
 
 int hyp_pin_shared_mem(void *from, void *to);
 void hyp_unpin_shared_mem(void *from, void *to);
-void reclaim_guest_pages(struct kvm_shadow_vm *vm);
+void reclaim_guest_pages(struct kvm_shadow_vm *vm, struct kvm_hyp_memcache *mc);
 int refill_memcache(struct kvm_hyp_memcache *mc, unsigned long min_pages,
 		    struct kvm_hyp_memcache *host_mc);
 
diff --git a/arch/arm64/kvm/hyp/nvhe/mem_protect.c b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
index 992ef4b668b4..bcf84e157d4b 100644
--- a/arch/arm64/kvm/hyp/nvhe/mem_protect.c
+++ b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
@@ -260,19 +260,24 @@  int kvm_guest_prepare_stage2(struct kvm_shadow_vm *vm, void *pgd)
 	return 0;
 }
 
-void reclaim_guest_pages(struct kvm_shadow_vm *vm)
+void reclaim_guest_pages(struct kvm_shadow_vm *vm, struct kvm_hyp_memcache *mc)
 {
-	unsigned long nr_pages, pfn;
-
-	nr_pages = kvm_pgtable_stage2_pgd_size(vm->kvm.arch.vtcr) >> PAGE_SHIFT;
-	pfn = hyp_virt_to_pfn(vm->pgt.pgd);
+	void *addr;
 
+	/* Dump all pgtable pages in the hyp_pool */
 	guest_lock_component(vm);
 	kvm_pgtable_stage2_destroy(&vm->pgt);
 	vm->kvm.arch.mmu.pgd_phys = 0ULL;
 	guest_unlock_component(vm);
 
-	WARN_ON(__pkvm_hyp_donate_host(pfn, nr_pages));
+	/* Drain the hyp_pool into the memcache */
+	addr = hyp_alloc_pages(&vm->pool, 0);
+	while (addr) {
+		memset(hyp_virt_to_page(addr), 0, sizeof(struct hyp_page));
+		push_hyp_memcache(mc, addr, hyp_virt_to_phys);
+		WARN_ON(__pkvm_hyp_donate_host(hyp_virt_to_pfn(addr), 1));
+		addr = hyp_alloc_pages(&vm->pool, 0);
+	}
 }
 
 int __pkvm_prot_finalize(void)
diff --git a/arch/arm64/kvm/hyp/nvhe/pkvm.c b/arch/arm64/kvm/hyp/nvhe/pkvm.c
index 114c5565de7d..a4a518b2a43b 100644
--- a/arch/arm64/kvm/hyp/nvhe/pkvm.c
+++ b/arch/arm64/kvm/hyp/nvhe/pkvm.c
@@ -546,8 +546,10 @@  int __pkvm_init_shadow(struct kvm *kvm, unsigned long shadow_hva,
 
 int __pkvm_teardown_shadow(unsigned int shadow_handle)
 {
+	struct kvm_hyp_memcache *mc;
 	struct kvm_shadow_vm *vm;
 	size_t shadow_size;
+	void *addr;
 	int err;
 
 	/* Lookup then remove entry from the shadow table. */
@@ -569,7 +571,8 @@  int __pkvm_teardown_shadow(unsigned int shadow_handle)
 	hyp_spin_unlock(&shadow_lock);
 
 	/* Reclaim guest pages (including page-table pages) */
-	reclaim_guest_pages(vm);
+	mc = &vm->host_kvm->arch.pkvm.teardown_mc;
+	reclaim_guest_pages(vm, mc);
 	unpin_host_vcpus(vm->shadow_vcpu_states, vm->kvm.created_vcpus);
 
 	/* Push the metadata pages to the teardown memcache */
@@ -577,6 +580,9 @@  int __pkvm_teardown_shadow(unsigned int shadow_handle)
 	hyp_unpin_shared_mem(vm->host_kvm, vm->host_kvm + 1);
 
 	memset(vm, 0, shadow_size);
+	for (addr = vm; addr < (void *)vm + shadow_size; addr += PAGE_SIZE)
+		push_hyp_memcache(mc, addr, hyp_virt_to_phys);
+
 	unmap_donated_memory_noclear(vm, shadow_size);
 	return 0;
 
diff --git a/arch/arm64/kvm/pkvm.c b/arch/arm64/kvm/pkvm.c
index b4466b31d7c8..b174d6dfde36 100644
--- a/arch/arm64/kvm/pkvm.c
+++ b/arch/arm64/kvm/pkvm.c
@@ -160,8 +160,6 @@  static int __kvm_shadow_create(struct kvm *kvm)
 
 	/* Store the shadow handle given by hyp for future call reference. */
 	kvm->arch.pkvm.shadow_handle = shadow_handle;
-	kvm->arch.pkvm.hyp_donations.pgd = pgd;
-	kvm->arch.pkvm.hyp_donations.shadow = shadow_addr;
 	return 0;
 
 free_shadow:
@@ -185,20 +183,12 @@  int kvm_shadow_create(struct kvm *kvm)
 
 void kvm_shadow_destroy(struct kvm *kvm)
 {
-	size_t pgd_sz, shadow_sz;
-
 	if (kvm->arch.pkvm.shadow_handle)
 		WARN_ON(kvm_call_hyp_nvhe(__pkvm_teardown_shadow,
 					  kvm->arch.pkvm.shadow_handle));
 
 	kvm->arch.pkvm.shadow_handle = 0;
-
-	shadow_sz = PAGE_ALIGN(KVM_SHADOW_VM_SIZE +
-			       KVM_SHADOW_VCPU_STATE_SIZE * kvm->created_vcpus);
-	pgd_sz = kvm_pgtable_stage2_pgd_size(kvm->arch.vtcr);
-
-	free_pages_exact(kvm->arch.pkvm.hyp_donations.shadow, shadow_sz);
-	free_pages_exact(kvm->arch.pkvm.hyp_donations.pgd, pgd_sz);
+	free_hyp_memcache(&kvm->arch.pkvm.teardown_mc);
 }
 
 int kvm_init_pvm(struct kvm *kvm)