- 11 Jun, 2022 40 commits
-
-
Sean Christopherson authored
Convert vmx_pmu_msrs_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 <seanjc@google.com> Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com>
-
Sean Christopherson authored
Convert vmx_preemption_timer_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 use vcpu_run() instead of _vcpu_run(), the test expects KVM_RUN to succeed. Signed-off-by:
-
Sean Christopherson authored
Convert xss_msr_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. Signed-off-by:
-
Sean Christopherson authored
Convert rseq_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 memslot_perf_test to use __vm_create_with_one_vcpu() and pass around a 'struct kvm_vcpu' object instead of using a global VCPU_ID. This is the first of many, many steps towards eliminating VCPU_ID from all KVM selftests, and towards eventually purging the VM+vcpu_id mess. Signed-off-by:
-
Sean Christopherson authored
Return the created vCPU from vm_vcpu_add() so that callers don't need to manually retrieve the vCPU that was just added. Opportunistically drop the "heavy" function comment, it adds a lot of lines of "code" but not much value, e.g. it's pretty obvious that @vm is a virtual machine... Signed-off-by:
-
Sean Christopherson authored
Rename 'struct vcpu' to 'struct kvm_vcpu' to align with 'struct kvm_vm' in the selftest, and to give readers a hint that the struct is specific to KVM. Signed-off-by:
-
Sean Christopherson authored
Rename the "state" field of 'struct vcpu' to "run". KVM calls it "run", the struct name is "kvm_run", etc... Signed-off-by:
-
Sean Christopherson authored
Rename xapic_state_test's kvm_vcpu struct to xapic_vcpu to avoid a collision when the common 'struct vcpu' is renamed to 'struct kvm_vcpu' in a future patch. Signed-off-by:
-
Sean Christopherson authored
Add ____vm_create() to be the innermost helper, and turn vm_create() into a wrapper the specifies VM_MODE_DEFAULT. Most of the vm_create() callers just want the default mode, or more accurately, don't care about the mode. Signed-off-by:
-
Sean Christopherson authored
Rename vm_create_without_vcpus() to vm_create() so that it's not misconstrued as helper that creates a VM that can never have vCPUs, as opposed to a helper that "just" creates a VM without vCPUs added at time zero. Signed-off-by:
-
Sean Christopherson authored
Rename vm_create() to vm_create_barebones() and drop the @phys_pages param. Pass '0' for the number of pages even though some callers pass 'DEFAULT_GUEST_PHY_PAGES', as the intent behind creating truly barebones VMs is purely to create a VM, i.e. there aren't vCPUs, there's no guest code loaded, etc..., and so there is nothing that will ever need or consume guest memory. Freeing up the name vm_create() will allow using the name for an inner helper to the other VM creators, which need a "full" VM. Opportunisticaly rewrite the function comment for addr_gpa2alias() to focus on what the _function_ does, not what its _sole caller_ does. Signed-off-by:
-
Sean Christopherson authored
Open code adding and doing setup for a vCPU in get-reg-list in order to avoid the stack allocation that comes with aarch64_vcpu_add_default(). get-reg-list doesn't need to run the vCPU, and so doesn't need the guest to be backed with memory. This will allow future cleanup to turn what is current vm_create() into a barebones helper. Signed-off-by:
-
Sean Christopherson authored
Use vm_create_without_vcpus() instead of open coding a rough equivalent in psci_test. Signed-off-by:
-
Sean Christopherson authored
Use vm_create_without_vcpus() instead of open coding a rough equivalent in hardware_disable_test. Signed-off-by:
-
Sean Christopherson authored
Use vm_create_without_vcpus() instead of open coding a rough equivalent. Signed-off-by:
-
Sean Christopherson authored
Use vm_create_without_vcpus() in set_boot_cpu_id instead of open coding the equivlant now that the "without_vcpus" variant does vm_adjust_num_guest_pages(). Signed-off-by:
-
Sean Christopherson authored
Move the call to vm_adjust_num_guest_pages() from vm_create_with_vcpus() down into vm_create_without_vcpus(). This will allow a future patch to make the "w/o vCPUs" variant the common inner helper, e.g. so that the "with_vcpus" helper calls the "without_vcpus" helper, instead of having them be separate paths. Signed-off-by:
-
Sean Christopherson authored
Add VM creation helpers to expose/return 'struct vcpu' so that tests don't have to hardcode a VCPU_ID or make assumptions about what vCPU ID is used by the framework just to retrieve a vCPU the test created. Signed-off-by:
-
Sean Christopherson authored
Add helpers for the various one-off helpers used by x86's vCPU state save/restore helpers, and convert the other open coded ioctl()s to use existing helpers. Signed-off-by:
-
Sean Christopherson authored
Move the get/set part of the MP_STATE and GUEST_DEBUG helpers to the end to align with the many other ioctl() wrappers/helpers. Note, this is not an endorsement of the predominant style, the goal is purely to provide consistency in the selftests. Signed-off-by:
-
Sean Christopherson authored
Consolidate the helper for retrieving the list of save/restore MSRs and the list of feature MSRs, and use the common helpers in the related get_msr_index_features test. Switching to the common helpers eliminates the testcase that KVM returns the same -E2BIG result if the input number of MSRs is '1' versus '0', but considered that testcase isn't very interesting, e.g. '0' and '1' are equally arbitrary, and certainly not worth the additional code. Signed-off-by:
-
Sean Christopherson authored
Assert that KVM_SET_MSRS returns '0' or '1' when setting XSS to a non-zero value. The ioctl() itself should "succeed", its only the setting of the XSS MSR that should fail/fault. Signed-off-by:
-
Sean Christopherson authored
Cache the list of MSRs to save restore, mostly to justify not freeing the list in the caller, which simplifies consumption of the list. Opportunistically move the XSS test's so called is_supported_msr() to common code as kvm_msr_is_in_save_restore_list(). The XSS is "supported" by KVM, it's simply not in the save/restore list because KVM doesn't yet allow a non-zero value. Signed-off-by:
-
Sean Christopherson authored
Rework the KVM_ENABLE_CAP helpers to take the cap and arg0; literally every current user, and likely every future user, wants to set 0 or 1 arguments and nothing else. Signed-off-by:
-
Sean Christopherson authored
Add __vm_enable_cap() and use it for negative tests that expect KVM_ENABLE_CAP to fail. Opportunistically clean up the MAX_VCPU_ID test error messages. Signed-off-by:
-
Sean Christopherson authored
Add a backpointer to 'struct vcpu' so that tests can get at the owning VM when passing around a vCPU object. Long term, this will be little more than a nice-to-have feature, but in the short term it is a critical step toward purging the VM+vcpu_id ioctl mess without introducing even more churn. Signed-off-by:
-
Sean Christopherson authored
Move the asserts for the many REDIST_REGS accesses into common helpers instead of copy+pasting the same, unhelpful asserts over and over. Not providing the actual (or expected) value makes it unnecessarily painful to debug failures, especially since test_assert() prints the errno unconditionally, e.g. on success, it may print a stale, misleading errno. Use kvm_device_attr_get() to handle the "success" check so that the "success" and "expected == actual" asserts are separated, which will make it far less likely that a user incorrectly assumes the ioctl() failed. Signed-off-by:
-
Sean Christopherson authored
Split the get/set device_attr helpers instead of using a boolean param to select between get and set. Duplicating upper level wrappers is a very, very small price to pay for improved readability, and having constant (at compile time) inputs will allow the selftests framework to sanity check ioctl() invocations. Signed-off-by:
-
Sean Christopherson authored
Drop 'int' returns from *_device_has_attr() helpers that assert the return is '0', there's no point in returning '0' and "requiring" the caller to perform a redundant assertion. Signed-off-by:
-
Sean Christopherson authored
Rename kvm_device_check_attr() and its variants to kvm_has_device_attr() to be consistent with the ioctl names and with other helpers in the KVM selftests framework. Signed-off-by:
-
Sean Christopherson authored
Multiplex the return value and fd (on success) in __kvm_create_device() to mimic common library helpers that return file descriptors, e.g. open(). Signed-off-by:
-
Sean Christopherson authored
Move KVM_CREATE_DEVICE_TEST to its own helper, identifying "real" versus "test" device creation based on a hardcoded boolean buried in the middle of a param list is painful for readers. Signed-off-by:
-
Sean Christopherson authored
Remove the two calls that pass @test=true to kvm_create_device() and drop the @test param entirely. The two removed calls don't check the return value of kvm_create_device(), so other than verifying KVM doesn't explode, which is extremely unlikely given that the non-test variant was _just_ called, they are pointless and provide no validation coverage. Signed-off-by:
-
Sean Christopherson authored
Use the KVM_IOCTL_ERROR() macro to generate error messages for a handful of one-off arm64 ioctls. The calls in question are made without an associated struct kvm_vm/kvm_vcpu as they are used to configure those structs, i.e. can't be easily converted to e.g. vcpu_ioctl(). Signed-off-by:
-
Sean Christopherson authored
Fold kvm_util_internal.h into kvm_util_base.h, i.e. make all KVM utility stuff "public". Hiding struct implementations from tests has been a massive failure, as it has led to pointless and poorly named wrappers, unnecessarily opaque code, etc... Not to mention that the approach was a complete failure as evidenced by the non-zero number of tests that were including kvm_util_internal.h. Signed-off-by:
-
Sean Christopherson authored
Make regs_dump() and sregs_dump() static, they're only implemented by x86 and only used internally. Signed-off-by:
-
Sean Christopherson authored
Use __KVM_SYSCALL_ERROR() to report and pretty print non-KVM syscall and ioctl errors, e.g. for mmap(), munmap(), uffd ioctls, etc... Signed-off-by:
-
Sean Christopherson authored
Use the recently introduced KVM-specific ioctl() helpers instead of open coding calls to ioctl() just to pretty print the ioctl name. Signed-off-by:
-
Sean Christopherson authored
Make kvm_ioctl() a macro wrapper and print the _name_ of the ioctl on failure instead of the number. Deliberately do not use __stringify(), as that will expand the ioctl all the way down to its numerical sequence, again the intent is to print the name of the macro. Signed-off-by:
-
