- 11 Jun, 2022 40 commits
-
-
Sean Christopherson authored
Use vm_create() instead of vm_create_default_with_vcpus() in tsc_scaling_sync. The existing call doesn't create any vCPUs, and the guest_code() entry point is set when vm_vcpu_add_default() is invoked. Signed-off-by:
Sean Christopherson <seanjc@google.com> Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com>
-
Sean Christopherson authored
Convert tprot to use vm_create_with_vcpus() and pass around a 'struct kvm_vcpu' object instead of passing around vCPU IDs. Note, this is a "functional" change in the sense that the test now creates a vCPU with vcpu_id==0 instead of vcpu_id==1. The non-zero VCPU_ID was 100% arbitrary and added little to no validation coverage. If testing non-zero vCPU IDs is desirable for generic tests, that can be done in the future by tweaking the VM creation helpers. Signed-off-by:
-
Sean Christopherson authored
Convert diag318_test_handler to use vm_create_with_vcpus() and pass around a 'struct kvm_vcpu' object instead of passing around vCPU IDs. Note, this is a "functional" change in the sense that the test now creates a vCPU with vcpu_id==0 instead of vcpu_id==6. The non-zero VCPU_ID was 100% arbitrary and added little to no validation coverage. If testing non-zero vCPU IDs is desirable for generic tests, that can be done in the future by tweaking the VM creation helpers. Signed-off-by:
-
Sean Christopherson authored
Pass around a 'struct kvm_vcpu' object instead of a vCPU ID in s390's memop test. Pass NULL for the vCPU instead of a magic '-1' ID to indicate that an ioctl/test should be done at VM scope. Rename "struct test_vcpu vcpu" to "struct test_info info" in order to avoid naming collisions (this is the bulk of the diff :-( ). Signed-off-by:
-
Sean Christopherson authored
Pass around a 'struct kvm_vcpu' object in the "resets" test instead of referencing the vCPU by the global VCPU_ID. Rename the #define for the vCPU's ID to ARBITRARY_NON_ZERO_VCPU_ID to make it more obvious that (a) the value matters but (b) is otherwise arbitrary. Signed-off-by:
-
Sean Christopherson authored
Convert sync_regs_test to use vm_create_with_vcpus() and pass around a 'struct kvm_vcpu' object instead of passing around vCPU IDs. Note, this is a "functional" change in the sense that the test now creates a vCPU with vcpu_id==0 instead of vcpu_id==5. The non-zero VCPU_ID was 100% arbitrary and added little to no validation coverage. If testing non-zero vCPU IDs is desirable for generic tests, that can be done in the future by tweaking the VM creation helpers. Signed-off-by:
-
Sean Christopherson authored
Convert vm_create_with_one_vcpu to use vm_create_with_vcpus() and pass around 'struct kvm_vcpu' objects instead of passing around vCPU IDs. Don't bother with macros for the HALTER versus SENDER indices, the vast majority of references don't differentiate between the vCPU roles, and the code that does either has a comment or an explicit reference to the role, e.g. to halter_guest_code() or sender_guest_code(). Signed-off-by:
-
Sean Christopherson authored
Use a combination of vm_create(), vm_create_with_vcpus(), and vm_vcpu_add() to convert vgic_init from vm_create_default_with_vcpus(), and away from referncing vCPUs by ID. Thus continues the march toward total annihilation of "default" helpers. Signed-off-by: -
Sean Christopherson authored
Sync the next stage using the VM before said VM is potentially freed by the TEST_STAGE_HVC_IFACE_FEAT_DISABLED stage. Opportunistically take a double pointer in anticipation of also having to set the new vCPU pointer once the test stops hardcoding '0' everywhere. Signed-off-by:
-
Sean Christopherson authored
Rework vcpu_{g,s}et_reg() to provide the APIs that tests actually want to use, and drop the three "one-off" implementations that cropped up due to the poor API. Ignore the handful of direct KVM_{G,S}ET_ONE_REG calls that don't fit the APIs for one reason or another. No functional change intended. Signed-off-by: -
Sean Christopherson authored
Use a combination of vm_create(), vm_create_with_vcpus(), and vm_vcpu_add() to convert vgic_init from vm_create_default_with_vcpus(), and away from referncing vCPUs by ID. Thus continues the march toward total annihilation of "default" helpers. Signed-off-by:
-
Sean Christopherson authored
Convert triple_fault_event_test to use vm_create_with_one_vcpu() and pull the vCPU's ID from 'struct kvm_vcpu'. Signed-off-by:
-
Sean Christopherson authored
Convert svm_nested_soft_inject_test to use vm_create_with_one_vcpu() and pull the vCPU's ID from 'struct kvm_vcpu'. Signed-off-by:
-
Sean Christopherson authored
Convert arch_timer to use vm_create_with_vcpus() and pass around a 'struct kvm_vcpu' object instead of requiring that the index into the array of vCPUs for a given vCPU is also the ID of the vCPU Signed-off-by:
-
Sean Christopherson authored
Convert steal_time to use vm_create_with_vcpus() and pass around a 'struct kvm_vcpu' object instead of requiring that the index into the array of vCPUs for a given vCPU is also the ID of the vCPU. Signed-off-by:
-
Sean Christopherson authored
Add a VM creator that "returns" the created vCPUs by filling the provided array. This will allow converting multi-vCPU tests away from hardcoded vCPU IDs. Signed-off-by:
-
Sean Christopherson authored
Pass around 'struct kvm_vcpu' objects in hardware_disable_test instead of the VM+vcpu_id (called "index" by the test). Signed-off-by:
-
Sean Christopherson authored
Pass around 'struct kvm_vcpu' objects in psci_test instead of relying on global VCPU_IDs. Ideally, the test wouldn't have to manually create vCPUs and thus care about vCPU IDs, but it's not the end of the world and avoiding that behavior isn't guaranteed to be a net positive (an attempt at macro shenanigans did not go very well). Signed-off-by:
-
Sean Christopherson authored
Rework set_boot_cpu_id to pass around 'struct kvm_vcpu' objects instead of relying on global VCPU_IDs. The test is still ugly, but that's unavoidable since the point of the test is to verify that KVM correctly assigns VCPU_ID==0 to be the BSP by default. This is literally one of two KVM selftests that legitimately needs to care about the exact vCPU IDs of the vCPUs it creates. Signed-off-by:
-
Sean Christopherson authored
Rename vm_vcpu_add() to __vm_vcpu_add(), and vm_vcpu_add_default() to vm_vcpu_add() to show the relationship between the newly minted vm_vcpu_add() and __vm_vcpu_add(). Signed-off-by:
-
Sean Christopherson authored
Return the created 'struct kvm_vcpu' object from vm_vcpu_add_default(), which cleans up a few tests and will eventually allow removing vcpu_get() entirely. Opportunistically rename @vcpuid to @vcpu_id to follow preferred kernel style. Signed-off-by:
-
Sean Christopherson authored
Add "arch" into the name of utility functions that are declared in common code, but (surprise!) have arch-specific implementations. Shuffle code around so that all such helpers' declarations are bundled together. Signed-off-by:
-
Sean Christopherson authored
An "unrestricted guest" is an VMX-only concept, move the relevant helper to x86-64 code. Assume most readers can correctly convert underscores to spaces and oppurtunistically trim the function comment. Signed-off-by:
-
Sean Christopherson authored
Move the declaration of guest_get_vcpuid() to include/aarch64/processor.h, it is implemented and used only by arm64. Signed-off-by:
-
Sean Christopherson authored
Convert vgic_irq to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of passing around a vCPU ID (which is always the global VCPU_ID...). Opportunstically align the indentation for multiple functions' parameters. Signed-off-by:
-
Sean Christopherson authored
Convert fix_hypercall_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Signed-off-by:
-
Sean Christopherson authored
Convert debug-exceptions to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Signed-off-by:
-
Sean Christopherson authored
Convert xapic_state_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of the raw vCPU ID. Signed-off-by:
-
Sean Christopherson authored
Track the added 'struct kvm_vcpu' object in tsc_scaling_sync instead of relying purely on the VM + vcpu_id combination. Ideally, the test wouldn't need to manually manage vCPUs, but the need to invoke a per-VM ioctl before creating vCPUs is not handled by the selftests framework, at least not yet... Signed-off-by:
-
Sean Christopherson authored
Convert system_counter_offset_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Signed-off-by:
-
Sean Christopherson authored
Convert set_memory_region_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Signed-off-by:
-
Sean Christopherson authored
Convert dirty_log_test to pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Note, this is a "functional" change in the sense that the test now creates a vCPU with vcpu_id==0 instead of vcpu_id==5. The non-zero VCPU_ID was 100% arbitrary and added little to no validation coverage. If testing non-zero vCPU IDs is desirable for generic tests, that can be done in the future by tweaking the VM creation helpers. The test still hardcodes usage of vcpu_id==0, but only for a few lines. That wart will be removed in the not-too-distant future. Signed-off-by:
-
Sean Christopherson authored
Convert xen_shinfo_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Note, this is a "functional" change in the sense that the test now creates a vCPU with vcpu_id==0 instead of vcpu_id==5. The non-zero VCPU_ID was 100% arbitrary and added little to no validation coverage. If testing non-zero vCPU IDs is desirable for generic tests, that can be done in the future by tweaking the VM creation helpers. Signed-off-by:
-
Sean Christopherson authored
Convert xen_vmcall_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Note, this is a "functional" change in the sense that the test now creates a vCPU with vcpu_id==0 instead of vcpu_id==5. The non-zero VCPU_ID was 100% arbitrary and added little to no validation coverage. If testing non-zero vCPU IDs is desirable for generic tests, that can be done in the future by tweaking the VM creation helpers. Opportunistically make the "vm" variable local, it is unused outside of main(). Signed-off-by:
-
Sean Christopherson authored
Convert vmx_invalid_nested_guest_state to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Signed-off-by:
-
Sean Christopherson authored
Convert userspace_io_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Note, this is a "functional" change in the sense that the test now creates a vCPU with vcpu_id==0 instead of vcpu_id==1. The non-zero VCPU_ID was 100% arbitrary and added little to no validation coverage. If testing non-zero vCPU IDs is desirable for generic tests, that can be done in the future by tweaking the VM creation helpers. Opportunistically use vcpu_run() instead of _vcpu_run() with an open coded assert that KVM_RUN succeeded. Signed-off-by:
-
Sean Christopherson authored
Convert cpuid_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Opportunistically use vcpu_run() instead of _vcpu_run(), the test expects KVM_RUN to succeed. Signed-off-by:
-
Sean Christopherson authored
Convert cr4_cpuid_sync_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. Note, this is a "functional" change in the sense that the test now creates a vCPU with vcpu_id==0 instead of vcpu_id==1. The non-zero VCPU_ID was 100% arbitrary and added little to no validation coverage. If testing non-zero vCPU IDs is desirable for generic tests, that can be done in the future by tweaking the VM creation helpers. Opportunistically use vcpu_run() instead of _vcpu_run() with an open coded assert that KVM_RUN succeeded. Signed-off-by:
-
Sean Christopherson authored
Convert amx_test to use vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID.o Opportunistically use vcpu_run() instead of _vcpu_run(), the test expects KVM_RUN to succeed. Signed-off-by:
-
Sean Christopherson authored
Replace MOVE_RIP+SET_RIP with a proper helper, vcpu_skip_insn(), that is more descriptive, doesn't subtly access local variables, and provides type safety. Signed-off-by:
-
