RISC-V: KVM: Implement VMID allocator

euleros inclusion
category: feature
feature: initial KVM RISC-V support
bugzilla: 46845
CVE: NA

We implement a simple VMID allocator for Guests/VMs which:
1. Detects number of VMID bits at boot-time
2. Uses atomic number to track VMID version and increments
   VMID version whenever we run-out of VMIDs
3. Flushes Guest TLBs on all host CPUs whenever we run-out
   of VMIDs
4. Force updates HW Stage2 VMID for each Guest VCPU whenever
   VMID changes using VCPU request KVM_REQ_UPDATE_HGATP

Reference: https://gitee.com/openeuler/kernel/issues/I26X9VSigned-off-by: NAnup Patel <anup.patel@wdc.com>
Acked-by: NPaolo Bonzini <pbonzini@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Reviewed-by: NAlexander Graf <graf@amazon.com>
Reviewed-by: NYifei Jiang <jiangyifei@huawei.com>
Acked-by: NXie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: NChen Jun <chenjun102@huawei.com>

arch/riscv/include/asm/kvm_host.h

@@ -27,6 +27,7 @@
 #define KVM_REQ_SLEEP \
 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
 #define KVM_REQ_VCPU_RESET		KVM_ARCH_REQ(1)
+#define KVM_REQ_UPDATE_HGATP		KVM_ARCH_REQ(2)
 
 struct kvm_vm_stat {
 	ulong remote_tlb_flush;
@@ -49,7 +50,19 @@ struct kvm_vcpu_stat {
 struct kvm_arch_memory_slot {
 };
 
+struct kvm_vmid {
+	/*
+	 * Writes to vmid_version and vmid happen with vmid_lock held
+	 * whereas reads happen without any lock held.
+	 */
+	unsigned long vmid_version;
+	unsigned long vmid;
+};
+
 struct kvm_arch {
+	/* stage2 vmid */
+	struct kvm_vmid vmid;
+
 	/* stage2 page table */
 	pgd_t *pgd;
 	phys_addr_t pgd_phys;
@@ -180,6 +193,11 @@ static inline void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) {}
 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
 
+void __kvm_riscv_hfence_gvma_vmid_gpa(unsigned long gpa, unsigned long vmid);
+void __kvm_riscv_hfence_gvma_vmid(unsigned long vmid);
+void __kvm_riscv_hfence_gvma_gpa(unsigned long gpa);
+void __kvm_riscv_hfence_gvma_all(void);
+
 int kvm_riscv_stage2_map(struct kvm_vcpu *vcpu,
 			 struct kvm_memory_slot *memslot,
 			 gpa_t gpa, unsigned long hva,
@@ -189,6 +207,12 @@ int kvm_riscv_stage2_alloc_pgd(struct kvm *kvm);
 void kvm_riscv_stage2_free_pgd(struct kvm *kvm);
 void kvm_riscv_stage2_update_hgatp(struct kvm_vcpu *vcpu);
 
+void kvm_riscv_stage2_vmid_detect(void);
+unsigned long kvm_riscv_stage2_vmid_bits(void);
+int kvm_riscv_stage2_vmid_init(struct kvm *kvm);
+bool kvm_riscv_stage2_vmid_ver_changed(struct kvm_vmid *vmid);
+void kvm_riscv_stage2_vmid_update(struct kvm_vcpu *vcpu);
+
 void __kvm_riscv_unpriv_trap(void);
 
 unsigned long kvm_riscv_vcpu_unpriv_read(struct kvm_vcpu *vcpu,

arch/riscv/kvm/Makefile

@@ -9,6 +9,7 @@ ccflags-y := -Ivirt/kvm -Iarch/riscv/kvm
 
 kvm-objs := $(common-objs-y)
 
-kvm-objs += main.o vm.o mmu.o vcpu.o vcpu_exit.o vcpu_switch.o
+kvm-objs += main.o vm.o vmid.o tlb.o mmu.o
+kvm-objs += vcpu.o vcpu_exit.o vcpu_switch.o
 
 obj-$(CONFIG_KVM)	+= kvm.o

arch/riscv/kvm/main.c

@@ -79,8 +79,12 @@ int kvm_arch_init(void *opaque)
 		return -ENODEV;
 	}
 
+	kvm_riscv_stage2_vmid_detect();
+
 	kvm_info("hypervisor extension available\n");
 
+	kvm_info("VMID %ld bits available\n", kvm_riscv_stage2_vmid_bits());
+
 	return 0;
 }
 

arch/riscv/kvm/tlb.S

+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2019 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *     Anup Patel <anup.patel@wdc.com>
+ */
+
+#include <linux/linkage.h>
+#include <asm/asm.h>
+
+	.text
+	.altmacro
+	.option norelax
+
+	/*
+	 * Instruction encoding of hfence.gvma is:
+	 * HFENCE.GVMA rs1, rs2
+	 * HFENCE.GVMA zero, rs2
+	 * HFENCE.GVMA rs1
+	 * HFENCE.GVMA
+	 *
+	 * rs1!=zero and rs2!=zero ==> HFENCE.GVMA rs1, rs2
+	 * rs1==zero and rs2!=zero ==> HFENCE.GVMA zero, rs2
+	 * rs1!=zero and rs2==zero ==> HFENCE.GVMA rs1
+	 * rs1==zero and rs2==zero ==> HFENCE.GVMA
+	 *
+	 * Instruction encoding of HFENCE.GVMA is:
+	 * 0110001 rs2(5) rs1(5) 000 00000 1110011
+	 */
+
+ENTRY(__kvm_riscv_hfence_gvma_vmid_gpa)
+	/*
+	 * rs1 = a0 (GPA)
+	 * rs2 = a1 (VMID)
+	 * HFENCE.GVMA a0, a1
+	 * 0110001 01011 01010 000 00000 1110011
+	 */
+	.word 0x62b50073
+	ret
+ENDPROC(__kvm_riscv_hfence_gvma_vmid_gpa)
+
+ENTRY(__kvm_riscv_hfence_gvma_vmid)
+	/*
+	 * rs1 = zero
+	 * rs2 = a0 (VMID)
+	 * HFENCE.GVMA zero, a0
+	 * 0110001 01010 00000 000 00000 1110011
+	 */
+	.word 0x62a00073
+	ret
+ENDPROC(__kvm_riscv_hfence_gvma_vmid)
+
+ENTRY(__kvm_riscv_hfence_gvma_gpa)
+	/*
+	 * rs1 = a0 (GPA)
+	 * rs2 = zero
+	 * HFENCE.GVMA a0
+	 * 0110001 00000 01010 000 00000 1110011
+	 */
+	.word 0x62050073
+	ret
+ENDPROC(__kvm_riscv_hfence_gvma_gpa)
+
+ENTRY(__kvm_riscv_hfence_gvma_all)
+	/*
+	 * rs1 = zero
+	 * rs2 = zero
+	 * HFENCE.GVMA
+	 * 0110001 00000 00000 000 00000 1110011
+	 */
+	.word 0x62000073
+	ret
+ENDPROC(__kvm_riscv_hfence_gvma_all)

arch/riscv/kvm/vcpu.c

@@ -637,6 +637,12 @@ static void kvm_riscv_check_vcpu_requests(struct kvm_vcpu *vcpu)
 
 		if (kvm_check_request(KVM_REQ_VCPU_RESET, vcpu))
 			kvm_riscv_reset_vcpu(vcpu);
+
+		if (kvm_check_request(KVM_REQ_UPDATE_HGATP, vcpu))
+			kvm_riscv_stage2_update_hgatp(vcpu);
+
+		if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
+			__kvm_riscv_hfence_gvma_all();
 	}
 }
 
@@ -682,6 +688,8 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
 		/* Check conditions before entering the guest */
 		cond_resched();
 
+		kvm_riscv_stage2_vmid_update(vcpu);
+
 		kvm_riscv_check_vcpu_requests(vcpu);
 
 		preempt_disable();
@@ -718,6 +726,7 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
 		kvm_riscv_update_hvip(vcpu);
 
 		if (ret <= 0 ||
+		    kvm_riscv_stage2_vmid_ver_changed(&vcpu->kvm->arch.vmid) ||
 		    kvm_request_pending(vcpu)) {
 			vcpu->mode = OUTSIDE_GUEST_MODE;
 			local_irq_enable();

arch/riscv/kvm/vm.c

@@ -26,6 +26,12 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 	if (r)
 		return r;
 
+	r = kvm_riscv_stage2_vmid_init(kvm);
+	if (r) {
+		kvm_riscv_stage2_free_pgd(kvm);
+		return r;
+	}
+
 	return 0;
 }
 

arch/riscv/kvm/vmid.c

+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2019 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *     Anup Patel <anup.patel@wdc.com>
+ */
+
+#include <linux/bitops.h>
+#include <linux/cpumask.h>
+#include <linux/errno.h>
+#include <linux/err.h>
+#include <linux/module.h>
+#include <linux/kvm_host.h>
+#include <asm/csr.h>
+#include <asm/sbi.h>
+
+static unsigned long vmid_version = 1;
+static unsigned long vmid_next;
+static unsigned long vmid_bits;
+static DEFINE_SPINLOCK(vmid_lock);
+
+void kvm_riscv_stage2_vmid_detect(void)
+{
+	unsigned long old;
+
+	/* Figure-out number of VMID bits in HW */
+	old = csr_read(CSR_HGATP);
+	csr_write(CSR_HGATP, old | HGATP_VMID_MASK);
+	vmid_bits = csr_read(CSR_HGATP);
+	vmid_bits = (vmid_bits & HGATP_VMID_MASK) >> HGATP_VMID_SHIFT;
+	vmid_bits = fls_long(vmid_bits);
+	csr_write(CSR_HGATP, old);
+
+	/* We polluted local TLB so flush all guest TLB */
+	__kvm_riscv_hfence_gvma_all();
+
+	/* We don't use VMID bits if they are not sufficient */
+	if ((1UL << vmid_bits) < num_possible_cpus())
+		vmid_bits = 0;
+}
+
+unsigned long kvm_riscv_stage2_vmid_bits(void)
+{
+	return vmid_bits;
+}
+
+int kvm_riscv_stage2_vmid_init(struct kvm *kvm)
+{
+	/* Mark the initial VMID and VMID version invalid */
+	kvm->arch.vmid.vmid_version = 0;
+	kvm->arch.vmid.vmid = 0;
+
+	return 0;
+}
+
+bool kvm_riscv_stage2_vmid_ver_changed(struct kvm_vmid *vmid)
+{
+	if (!vmid_bits)
+		return false;
+
+	return unlikely(READ_ONCE(vmid->vmid_version) !=
+			READ_ONCE(vmid_version));
+}
+
+void kvm_riscv_stage2_vmid_update(struct kvm_vcpu *vcpu)
+{
+	int i;
+	struct kvm_vcpu *v;
+	struct cpumask hmask;
+	struct kvm_vmid *vmid = &vcpu->kvm->arch.vmid;
+
+	if (!kvm_riscv_stage2_vmid_ver_changed(vmid))
+		return;
+
+	spin_lock(&vmid_lock);
+
+	/*
+	 * We need to re-check the vmid_version here to ensure that if
+	 * another vcpu already allocated a valid vmid for this vm.
+	 */
+	if (!kvm_riscv_stage2_vmid_ver_changed(vmid)) {
+		spin_unlock(&vmid_lock);
+		return;
+	}
+
+	/* First user of a new VMID version? */
+	if (unlikely(vmid_next == 0)) {
+		WRITE_ONCE(vmid_version, READ_ONCE(vmid_version) + 1);
+		vmid_next = 1;
+
+		/*
+		 * We ran out of VMIDs so we increment vmid_version and
+		 * start assigning VMIDs from 1.
+		 *
+		 * This also means existing VMIDs assignement to all Guest
+		 * instances is invalid and we have force VMID re-assignement
+		 * for all Guest instances. The Guest instances that were not
+		 * running will automatically pick-up new VMIDs because will
+		 * call kvm_riscv_stage2_vmid_update() whenever they enter
+		 * in-kernel run loop. For Guest instances that are already
+		 * running, we force VM exits on all host CPUs using IPI and
+		 * flush all Guest TLBs.
+		 */
+		riscv_cpuid_to_hartid_mask(cpu_online_mask, &hmask);
+		sbi_remote_hfence_gvma(cpumask_bits(&hmask), 0, 0);
+	}
+
+	vmid->vmid = vmid_next;
+	vmid_next++;
+	vmid_next &= (1 << vmid_bits) - 1;
+
+	WRITE_ONCE(vmid->vmid_version, READ_ONCE(vmid_version));
+
+	spin_unlock(&vmid_lock);
+
+	/* Request stage2 page table update for all VCPUs */
+	kvm_for_each_vcpu(i, v, vcpu->kvm)
+		kvm_make_request(KVM_REQ_UPDATE_HGATP, v);
+}