KVM: selftests: aarch64/vgic-v3 init sequence tests
authorEric Auger <eric.auger@redhat.com>
Mon, 5 Apr 2021 16:39:41 +0000 (18:39 +0200)
committerMarc Zyngier <maz@kernel.org>
Tue, 6 Apr 2021 13:52:31 +0000 (14:52 +0100)
The tests exercise the VGIC_V3 device creation including the
associated KVM_DEV_ARM_VGIC_GRP_ADDR group attributes:

- KVM_VGIC_V3_ADDR_TYPE_DIST/REDIST
- KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION

Some other tests dedicate to KVM_DEV_ARM_VGIC_GRP_REDIST_REGS group
and especially the GICR_TYPER read. The goal was to test the case
recently fixed by commit 23bde34771f1
("KVM: arm64: vgic-v3: Drop the reporting of GICR_TYPER.Last for userspace").

The API under test can be found at
Documentation/virt/kvm/devices/arm-vgic-v3.rst

Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210405163941.510258-10-eric.auger@redhat.com
tools/testing/selftests/kvm/.gitignore
tools/testing/selftests/kvm/Makefile
tools/testing/selftests/kvm/aarch64/vgic_init.c [new file with mode: 0644]
tools/testing/selftests/kvm/include/kvm_util.h
tools/testing/selftests/kvm/lib/kvm_util.c

index 32b87cc..e65d557 100644 (file)
@@ -1,6 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only
 /aarch64/get-reg-list
 /aarch64/get-reg-list-sve
 # SPDX-License-Identifier: GPL-2.0-only
 /aarch64/get-reg-list
 /aarch64/get-reg-list-sve
+/aarch64/vgic_init
 /s390x/memop
 /s390x/resets
 /s390x/sync_regs_test
 /s390x/memop
 /s390x/resets
 /s390x/sync_regs_test
index a6d61f4..4e548d7 100644 (file)
@@ -75,6 +75,7 @@ TEST_GEN_PROGS_x86_64 += steal_time
 
 TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list
 TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list-sve
 
 TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list
 TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list-sve
+TEST_GEN_PROGS_aarch64 += aarch64/vgic_init
 TEST_GEN_PROGS_aarch64 += demand_paging_test
 TEST_GEN_PROGS_aarch64 += dirty_log_test
 TEST_GEN_PROGS_aarch64 += dirty_log_perf_test
 TEST_GEN_PROGS_aarch64 += demand_paging_test
 TEST_GEN_PROGS_aarch64 += dirty_log_test
 TEST_GEN_PROGS_aarch64 += dirty_log_perf_test
diff --git a/tools/testing/selftests/kvm/aarch64/vgic_init.c b/tools/testing/selftests/kvm/aarch64/vgic_init.c
new file mode 100644 (file)
index 0000000..03f62fc
--- /dev/null
@@ -0,0 +1,584 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * vgic init sequence tests
+ *
+ * Copyright (C) 2020, Red Hat, Inc.
+ */
+#define _GNU_SOURCE
+#include <linux/kernel.h>
+#include <sys/syscall.h>
+#include <asm/kvm.h>
+#include <asm/kvm_para.h>
+
+#include "test_util.h"
+#include "kvm_util.h"
+#include "processor.h"
+
+#define NR_VCPUS               4
+
+#define REDIST_REGION_ATTR_ADDR(count, base, flags, index) (((uint64_t)(count) << 52) | \
+       ((uint64_t)((base) >> 16) << 16) | ((uint64_t)(flags) << 12) | index)
+#define REG_OFFSET(vcpu, offset) (((uint64_t)vcpu << 32) | offset)
+
+#define GICR_TYPER 0x8
+
+struct vm_gic {
+       struct kvm_vm *vm;
+       int gic_fd;
+};
+
+int max_ipa_bits;
+
+/* helper to access a redistributor register */
+static int access_redist_reg(int gicv3_fd, int vcpu, int offset,
+                            uint32_t *val, bool write)
+{
+       uint64_t attr = REG_OFFSET(vcpu, offset);
+
+       return _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_REDIST_REGS,
+                                 attr, val, write);
+}
+
+/* dummy guest code */
+static void guest_code(void)
+{
+       GUEST_SYNC(0);
+       GUEST_SYNC(1);
+       GUEST_SYNC(2);
+       GUEST_DONE();
+}
+
+/* we don't want to assert on run execution, hence that helper */
+static int run_vcpu(struct kvm_vm *vm, uint32_t vcpuid)
+{
+       int ret;
+
+       vcpu_args_set(vm, vcpuid, 1);
+       ret = _vcpu_ioctl(vm, vcpuid, KVM_RUN, NULL);
+       get_ucall(vm, vcpuid, NULL);
+
+       if (ret)
+               return -errno;
+       return 0;
+}
+
+static struct vm_gic vm_gic_create(void)
+{
+       struct vm_gic v;
+
+       v.vm = vm_create_default_with_vcpus(NR_VCPUS, 0, 0, guest_code, NULL);
+       v.gic_fd = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+       TEST_ASSERT(v.gic_fd > 0, "GICv3 device created");
+
+       return v;
+}
+
+static void vm_gic_destroy(struct vm_gic *v)
+{
+       close(v->gic_fd);
+       kvm_vm_free(v->vm);
+}
+
+/**
+ * Helper routine that performs KVM device tests in general and
+ * especially ARM_VGIC_V3 ones. Eventually the ARM_VGIC_V3
+ * device gets created, a legacy RDIST region is set at @0x0
+ * and a DIST region is set @0x60000
+ */
+static void subtest_dist_rdist(struct vm_gic *v)
+{
+       int ret;
+       uint64_t addr;
+
+       /* Check existing group/attributes */
+       ret = _kvm_device_check_attr(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                   KVM_VGIC_V3_ADDR_TYPE_DIST);
+       TEST_ASSERT(!ret, "KVM_DEV_ARM_VGIC_GRP_ADDR/KVM_VGIC_V3_ADDR_TYPE_DIST supported");
+
+       ret = _kvm_device_check_attr(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                   KVM_VGIC_V3_ADDR_TYPE_REDIST);
+       TEST_ASSERT(!ret, "KVM_DEV_ARM_VGIC_GRP_ADDR/KVM_VGIC_V3_ADDR_TYPE_REDIST supported");
+
+       /* check non existing attribute */
+       ret = _kvm_device_check_attr(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR, 0);
+       TEST_ASSERT(ret == -ENXIO, "attribute not supported");
+
+       /* misaligned DIST and REDIST address settings */
+       addr = 0x1000;
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_DIST, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "GICv3 dist base not 64kB aligned");
+
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "GICv3 redist base not 64kB aligned");
+
+       /* out of range address */
+       if (max_ipa_bits) {
+               addr = 1ULL << max_ipa_bits;
+               ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                        KVM_VGIC_V3_ADDR_TYPE_DIST, &addr, true);
+               TEST_ASSERT(ret == -E2BIG, "dist address beyond IPA limit");
+
+               ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                        KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+               TEST_ASSERT(ret == -E2BIG, "redist address beyond IPA limit");
+       }
+
+       /* set REDIST base address @0x0*/
+       addr = 0x00000;
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+       TEST_ASSERT(!ret, "GICv3 redist base set");
+
+       /* Attempt to create a second legacy redistributor region */
+       addr = 0xE0000;
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+       TEST_ASSERT(ret == -EEXIST, "GICv3 redist base set again");
+
+       /* Attempt to mix legacy and new redistributor regions */
+       addr = REDIST_REGION_ATTR_ADDR(NR_VCPUS, 0x100000, 0, 0);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "attempt to mix GICv3 REDIST and REDIST_REGION");
+
+       /*
+        * Set overlapping DIST / REDIST, cannot be detected here. Will be detected
+        * on first vcpu run instead.
+        */
+       addr = 3 * 2 * 0x10000;
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR, KVM_VGIC_V3_ADDR_TYPE_DIST,
+                                &addr, true);
+       TEST_ASSERT(!ret, "dist overlapping rdist");
+}
+
+/* Test the new REDIST region API */
+static void subtest_redist_regions(struct vm_gic *v)
+{
+       uint64_t addr, expected_addr;
+       int ret;
+
+       ret = kvm_device_check_attr(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                    KVM_VGIC_V3_ADDR_TYPE_REDIST);
+       TEST_ASSERT(!ret, "Multiple redist regions advertised");
+
+       addr = REDIST_REGION_ATTR_ADDR(NR_VCPUS, 0x100000, 2, 0);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "redist region attr value with flags != 0");
+
+       addr = REDIST_REGION_ATTR_ADDR(0, 0x100000, 0, 0);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "redist region attr value with count== 0");
+
+       addr = REDIST_REGION_ATTR_ADDR(2, 0x200000, 0, 1);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "attempt to register the first rdist region with index != 0");
+
+       addr = REDIST_REGION_ATTR_ADDR(2, 0x201000, 0, 1);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "rdist region with misaligned address");
+
+       addr = REDIST_REGION_ATTR_ADDR(2, 0x200000, 0, 0);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "First valid redist region with 2 rdist @ 0x200000, index 0");
+
+       addr = REDIST_REGION_ATTR_ADDR(2, 0x200000, 0, 1);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "register an rdist region with already used index");
+
+       addr = REDIST_REGION_ATTR_ADDR(1, 0x210000, 0, 2);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "register an rdist region overlapping with another one");
+
+       addr = REDIST_REGION_ATTR_ADDR(1, 0x240000, 0, 2);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "register redist region with index not +1");
+
+       addr = REDIST_REGION_ATTR_ADDR(1, 0x240000, 0, 1);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "register valid redist region with 1 rdist @ 0x220000, index 1");
+
+       addr = REDIST_REGION_ATTR_ADDR(1, 1ULL << max_ipa_bits, 0, 2);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -E2BIG, "register redist region with base address beyond IPA range");
+
+       addr = 0x260000;
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "Mix KVM_VGIC_V3_ADDR_TYPE_REDIST and REDIST_REGION");
+
+       /*
+        * Now there are 2 redist regions:
+        * region 0 @ 0x200000 2 redists
+        * region 1 @ 0x240000 1 redist
+        * Attempt to read their characteristics
+        */
+
+       addr = REDIST_REGION_ATTR_ADDR(0, 0, 0, 0);
+       expected_addr = REDIST_REGION_ATTR_ADDR(2, 0x200000, 0, 0);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, false);
+       TEST_ASSERT(!ret && addr == expected_addr, "read characteristics of region #0");
+
+       addr = REDIST_REGION_ATTR_ADDR(0, 0, 0, 1);
+       expected_addr = REDIST_REGION_ATTR_ADDR(1, 0x240000, 0, 1);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, false);
+       TEST_ASSERT(!ret && addr == expected_addr, "read characteristics of region #1");
+
+       addr = REDIST_REGION_ATTR_ADDR(0, 0, 0, 2);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, false);
+       TEST_ASSERT(ret == -ENOENT, "read characteristics of non existing region");
+
+       addr = 0x260000;
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_DIST, &addr, true);
+       TEST_ASSERT(!ret, "set dist region");
+
+       addr = REDIST_REGION_ATTR_ADDR(1, 0x260000, 0, 2);
+       ret = _kvm_device_access(v->gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "register redist region colliding with dist");
+}
+
+/*
+ * VGIC KVM device is created and initialized before the secondary CPUs
+ * get created
+ */
+static void test_vgic_then_vcpus(void)
+{
+       struct vm_gic v;
+       int ret, i;
+
+       v.vm = vm_create_default(0, 0, guest_code);
+       v.gic_fd = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+       TEST_ASSERT(v.gic_fd > 0, "GICv3 device created");
+
+       subtest_dist_rdist(&v);
+
+       /* Add the rest of the VCPUs */
+       for (i = 1; i < NR_VCPUS; ++i)
+               vm_vcpu_add_default(v.vm, i, guest_code);
+
+       ucall_init(v.vm, NULL);
+       ret = run_vcpu(v.vm, 3);
+       TEST_ASSERT(ret == -EINVAL, "dist/rdist overlap detected on 1st vcpu run");
+
+       vm_gic_destroy(&v);
+}
+
+/* All the VCPUs are created before the VGIC KVM device gets initialized */
+static void test_vcpus_then_vgic(void)
+{
+       struct vm_gic v;
+       int ret;
+
+       v = vm_gic_create();
+
+       subtest_dist_rdist(&v);
+
+       ucall_init(v.vm, NULL);
+       ret = run_vcpu(v.vm, 3);
+       TEST_ASSERT(ret == -EINVAL, "dist/rdist overlap detected on 1st vcpu run");
+
+       vm_gic_destroy(&v);
+}
+
+static void test_new_redist_regions(void)
+{
+       void *dummy = NULL;
+       struct vm_gic v;
+       uint64_t addr;
+       int ret;
+
+       v = vm_gic_create();
+       subtest_redist_regions(&v);
+       ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+                                KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+       TEST_ASSERT(!ret, "init the vgic");
+
+       ucall_init(v.vm, NULL);
+       ret = run_vcpu(v.vm, 3);
+       TEST_ASSERT(ret == -ENXIO, "running without sufficient number of rdists");
+       vm_gic_destroy(&v);
+
+       /* step2 */
+
+       v = vm_gic_create();
+       subtest_redist_regions(&v);
+
+       addr = REDIST_REGION_ATTR_ADDR(1, 0x280000, 0, 2);
+       ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "register a third region allowing to cover the 4 vcpus");
+
+       ucall_init(v.vm, NULL);
+       ret = run_vcpu(v.vm, 3);
+       TEST_ASSERT(ret == -EBUSY, "running without vgic explicit init");
+
+       vm_gic_destroy(&v);
+
+       /* step 3 */
+
+       v = vm_gic_create();
+       subtest_redist_regions(&v);
+
+       ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, dummy, true);
+       TEST_ASSERT(ret == -EFAULT, "register a third region allowing to cover the 4 vcpus");
+
+       addr = REDIST_REGION_ATTR_ADDR(1, 0x280000, 0, 2);
+       ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "register a third region allowing to cover the 4 vcpus");
+
+       ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+                                KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+       TEST_ASSERT(!ret, "init the vgic");
+
+       ucall_init(v.vm, NULL);
+       ret = run_vcpu(v.vm, 3);
+       TEST_ASSERT(!ret, "vcpu run");
+
+       vm_gic_destroy(&v);
+}
+
+static void test_typer_accesses(void)
+{
+       int ret, i, gicv3_fd = -1;
+       uint64_t addr;
+       struct kvm_vm *vm;
+       uint32_t val;
+
+       vm = vm_create_default(0, 0, guest_code);
+
+       gicv3_fd = kvm_create_device(vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+       TEST_ASSERT(gicv3_fd >= 0, "VGIC_V3 device created");
+
+       vm_vcpu_add_default(vm, 3, guest_code);
+
+       ret = access_redist_reg(gicv3_fd, 1, GICR_TYPER, &val, false);
+       TEST_ASSERT(ret == -EINVAL, "attempting to read GICR_TYPER of non created vcpu");
+
+       vm_vcpu_add_default(vm, 1, guest_code);
+
+       ret = access_redist_reg(gicv3_fd, 1, GICR_TYPER, &val, false);
+       TEST_ASSERT(ret == -EBUSY, "read GICR_TYPER before GIC initialized");
+
+       vm_vcpu_add_default(vm, 2, guest_code);
+
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+                                KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+       TEST_ASSERT(!ret, "init the vgic after the vcpu creations");
+
+       for (i = 0; i < NR_VCPUS ; i++) {
+               ret = access_redist_reg(gicv3_fd, 0, GICR_TYPER, &val, false);
+               TEST_ASSERT(!ret && !val, "read GICR_TYPER before rdist region setting");
+       }
+
+       addr = REDIST_REGION_ATTR_ADDR(2, 0x200000, 0, 0);
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "first rdist region with a capacity of 2 rdists");
+
+       /* The 2 first rdists should be put there (vcpu 0 and 3) */
+       ret = access_redist_reg(gicv3_fd, 0, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && !val, "read typer of rdist #0");
+
+       ret = access_redist_reg(gicv3_fd, 3, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x310, "read typer of rdist #1");
+
+       addr = REDIST_REGION_ATTR_ADDR(10, 0x100000, 0, 1);
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(ret == -EINVAL, "collision with previous rdist region");
+
+       ret = access_redist_reg(gicv3_fd, 1, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x100,
+                   "no redist region attached to vcpu #1 yet, last cannot be returned");
+
+       ret = access_redist_reg(gicv3_fd, 2, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x200,
+                   "no redist region attached to vcpu #2, last cannot be returned");
+
+       addr = REDIST_REGION_ATTR_ADDR(10, 0x20000, 0, 1);
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "second rdist region");
+
+       ret = access_redist_reg(gicv3_fd, 1, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x100, "read typer of rdist #1");
+
+       ret = access_redist_reg(gicv3_fd, 2, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x210,
+                   "read typer of rdist #1, last properly returned");
+
+       close(gicv3_fd);
+       kvm_vm_free(vm);
+}
+
+/**
+ * Test GICR_TYPER last bit with new redist regions
+ * rdist regions #1 and #2 are contiguous
+ * rdist region #0 @0x100000 2 rdist capacity
+ *     rdists: 0, 3 (Last)
+ * rdist region #1 @0x240000 2 rdist capacity
+ *     rdists:  5, 4 (Last)
+ * rdist region #2 @0x200000 2 rdist capacity
+ *     rdists: 1, 2
+ */
+static void test_last_bit_redist_regions(void)
+{
+       uint32_t vcpuids[] = { 0, 3, 5, 4, 1, 2 };
+       int ret, gicv3_fd;
+       uint64_t addr;
+       struct kvm_vm *vm;
+       uint32_t val;
+
+       vm = vm_create_default_with_vcpus(6, 0, 0, guest_code, vcpuids);
+
+       gicv3_fd = kvm_create_device(vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+       TEST_ASSERT(gicv3_fd >= 0, "VGIC_V3 device created");
+
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+                                KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+       TEST_ASSERT(!ret, "init the vgic after the vcpu creations");
+
+       addr = REDIST_REGION_ATTR_ADDR(2, 0x100000, 0, 0);
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "rdist region #0 (2 rdist)");
+
+       addr = REDIST_REGION_ATTR_ADDR(2, 0x240000, 0, 1);
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "rdist region #1 (1 rdist) contiguous with #2");
+
+       addr = REDIST_REGION_ATTR_ADDR(2, 0x200000, 0, 2);
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION, &addr, true);
+       TEST_ASSERT(!ret, "rdist region #2 with a capacity of 2 rdists");
+
+       ret = access_redist_reg(gicv3_fd, 0, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x000, "read typer of rdist #0");
+
+       ret = access_redist_reg(gicv3_fd, 1, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x100, "read typer of rdist #1");
+
+       ret = access_redist_reg(gicv3_fd, 2, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x200, "read typer of rdist #2");
+
+       ret = access_redist_reg(gicv3_fd, 3, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x310, "read typer of rdist #3");
+
+       ret = access_redist_reg(gicv3_fd, 5, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x500, "read typer of rdist #5");
+
+       ret = access_redist_reg(gicv3_fd, 4, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x410, "read typer of rdist #4");
+
+       close(gicv3_fd);
+       kvm_vm_free(vm);
+}
+
+/* Test last bit with legacy region */
+static void test_last_bit_single_rdist(void)
+{
+       uint32_t vcpuids[] = { 0, 3, 5, 4, 1, 2 };
+       int ret, gicv3_fd;
+       uint64_t addr;
+       struct kvm_vm *vm;
+       uint32_t val;
+
+       vm = vm_create_default_with_vcpus(6, 0, 0, guest_code, vcpuids);
+
+       gicv3_fd = kvm_create_device(vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+       TEST_ASSERT(gicv3_fd >= 0, "VGIC_V3 device created");
+
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+                                KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+       TEST_ASSERT(!ret, "init the vgic after the vcpu creations");
+
+       addr = 0x10000;
+       ret = _kvm_device_access(gicv3_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+                                KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+
+       ret = access_redist_reg(gicv3_fd, 0, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x000, "read typer of rdist #0");
+
+       ret = access_redist_reg(gicv3_fd, 3, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x300, "read typer of rdist #1");
+
+       ret = access_redist_reg(gicv3_fd, 5, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x500, "read typer of rdist #2");
+
+       ret = access_redist_reg(gicv3_fd, 1, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x100, "read typer of rdist #3");
+
+       ret = access_redist_reg(gicv3_fd, 2, GICR_TYPER, &val, false);
+       TEST_ASSERT(!ret && val == 0x210, "read typer of rdist #3");
+
+       close(gicv3_fd);
+       kvm_vm_free(vm);
+}
+
+void test_kvm_device(void)
+{
+       struct vm_gic v;
+       int ret;
+
+       v.vm = vm_create_default_with_vcpus(NR_VCPUS, 0, 0, guest_code, NULL);
+
+       /* try to create a non existing KVM device */
+       ret = _kvm_create_device(v.vm, 0, true);
+       TEST_ASSERT(ret == -ENODEV, "unsupported device");
+
+       /* trial mode with VGIC_V3 device */
+       ret = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, true);
+       if (ret) {
+               print_skip("GICv3 not supported");
+               exit(KSFT_SKIP);
+       }
+       v.gic_fd = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+       TEST_ASSERT(v.gic_fd, "create the GICv3 device");
+
+       ret = _kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+       TEST_ASSERT(ret == -EEXIST, "create GICv3 device twice");
+
+       ret = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, true);
+       TEST_ASSERT(!ret, "create GICv3 in test mode while the same already is created");
+
+       if (!_kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V2, true)) {
+               ret = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V2, false);
+               TEST_ASSERT(ret == -EINVAL, "create GICv2 while v3 exists");
+       }
+
+       vm_gic_destroy(&v);
+}
+
+int main(int ac, char **av)
+{
+       max_ipa_bits = kvm_check_cap(KVM_CAP_ARM_VM_IPA_SIZE);
+
+       test_kvm_device();
+       test_vcpus_then_vgic();
+       test_vgic_then_vcpus();
+       test_new_redist_regions();
+       test_typer_accesses();
+       test_last_bit_redist_regions();
+       test_last_bit_single_rdist();
+
+       return 0;
+}
index 2d7eb69..13594b8 100644 (file)
@@ -223,6 +223,15 @@ int vcpu_nested_state_set(struct kvm_vm *vm, uint32_t vcpuid,
 #endif
 void *vcpu_map_dirty_ring(struct kvm_vm *vm, uint32_t vcpuid);
 
 #endif
 void *vcpu_map_dirty_ring(struct kvm_vm *vm, uint32_t vcpuid);
 
+int _kvm_device_check_attr(int dev_fd, uint32_t group, uint64_t attr);
+int kvm_device_check_attr(int dev_fd, uint32_t group, uint64_t attr);
+int _kvm_create_device(struct kvm_vm *vm, uint64_t type, bool test);
+int kvm_create_device(struct kvm_vm *vm, uint64_t type, bool test);
+int _kvm_device_access(int dev_fd, uint32_t group, uint64_t attr,
+                      void *val, bool write);
+int kvm_device_access(int dev_fd, uint32_t group, uint64_t attr,
+                     void *val, bool write);
+
 const char *exit_reason_str(unsigned int exit_reason);
 
 void virt_pgd_alloc(struct kvm_vm *vm, uint32_t pgd_memslot);
 const char *exit_reason_str(unsigned int exit_reason);
 
 void virt_pgd_alloc(struct kvm_vm *vm, uint32_t pgd_memslot);
index e5fbf16..05037ec 100644 (file)
@@ -1729,6 +1729,83 @@ int _kvm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
 }
 
 /*
 }
 
 /*
+ * Device Ioctl
+ */
+
+int _kvm_device_check_attr(int dev_fd, uint32_t group, uint64_t attr)
+{
+       struct kvm_device_attr attribute = {
+               .group = group,
+               .attr = attr,
+               .flags = 0,
+       };
+       int ret = ioctl(dev_fd, KVM_HAS_DEVICE_ATTR, &attribute);
+
+       if (ret == -1)
+               return -errno;
+       return 0;
+}
+
+int kvm_device_check_attr(int dev_fd, uint32_t group, uint64_t attr)
+{
+       int ret = _kvm_device_check_attr(dev_fd, group, attr);
+
+       TEST_ASSERT(ret >= 0, "KVM_HAS_DEVICE_ATTR failed, errno: %i", errno);
+       return ret;
+}
+
+int _kvm_create_device(struct kvm_vm *vm, uint64_t type, bool test)
+{
+       struct kvm_create_device create_dev;
+       int ret;
+
+       create_dev.type = type;
+       create_dev.fd = -1;
+       create_dev.flags = test ? KVM_CREATE_DEVICE_TEST : 0;
+       ret = ioctl(vm_get_fd(vm), KVM_CREATE_DEVICE, &create_dev);
+       if (ret == -1)
+               return -errno;
+       return test ? 0 : create_dev.fd;
+}
+
+int kvm_create_device(struct kvm_vm *vm, uint64_t type, bool test)
+{
+       int ret = _kvm_create_device(vm, type, test);
+
+       TEST_ASSERT(ret >= 0, "KVM_CREATE_DEVICE IOCTL failed,\n"
+               "  errno: %i", errno);
+       return ret;
+}
+
+int _kvm_device_access(int dev_fd, uint32_t group, uint64_t attr,
+                     void *val, bool write)
+{
+       struct kvm_device_attr kvmattr = {
+               .group = group,
+               .attr = attr,
+               .flags = 0,
+               .addr = (uintptr_t)val,
+       };
+       int ret;
+
+       ret = ioctl(dev_fd, write ? KVM_SET_DEVICE_ATTR : KVM_GET_DEVICE_ATTR,
+                   &kvmattr);
+       if (ret < 0)
+               return -errno;
+       return ret;
+}
+
+int kvm_device_access(int dev_fd, uint32_t group, uint64_t attr,
+                     void *val, bool write)
+{
+       int ret = _kvm_device_access(dev_fd, group, attr, val, write);
+
+       TEST_ASSERT(ret >= 0, "KVM_SET|GET_DEVICE_ATTR IOCTL failed,\n"
+                   "  errno: %i", errno);
+       return ret;
+}
+
+/*
  * VM Dump
  *
  * Input Args:
  * VM Dump
  *
  * Input Args: