On Tue, Dec 07, 2021 at 03:46:37PM +0000, Alexandru Elisei wrote:
> This series intends to fix two bugs in the timer test. The first one is the
> TVAL comparison to check that the timer has expired and was found by code
> inspection.
>
> The second one I found while playing with KVM, but it can manifest itself
> on certain hardware configuration with an unmodified version of KVM
> (details in the commit message for the last patch). Or on baremetal (not
> tested). In short, WFI can complete for a variety of reason, not just
> because an interrupt targetted at the VM was asserted. The fix I
> implemented was to do WFI in a loop until we get the interrupt or TVAL
> shows that the timer has expired.
>
> All the patches in between are an attempt make the tests more robust and
> slightly easier to understand. If these changes are considered unnecessary,
> I would be more than happy to drop them; the main goal of the series is to
> fix the two bugs.
>
> Tested on a rockpro64 with KVM modifed to clear HCR_EL2.TWI, which means
> that the WFI instruction is not trapped (WFI trapping is a performance
> optimization, not a correctness requirement):
>
> diff --git a/arch/arm64/include/asm/kvm_emulate.h
> b/arch/arm64/include/asm/kvm_emulate.h
> index f4871e47b2d0..9af13e01ffeb 100644
> --- a/arch/arm64/include/asm/kvm_emulate.h
> +++ b/arch/arm64/include/asm/kvm_emulate.h
> @@ -96,18 +96,12 @@ static inline unsigned long *vcpu_hcr(struct kvm_vcpu
> *vcpu)
>
> static inline void vcpu_clear_wfx_traps(struct kvm_vcpu *vcpu)
> {
> - vcpu->arch.hcr_el2 &= ~HCR_TWE;
> - if (atomic_read(&vcpu->arch.vgic_cpu.vgic_v3.its_vpe.vlpi_count) ||
> - vcpu->kvm->arch.vgic.nassgireq)
> - vcpu->arch.hcr_el2 &= ~HCR_TWI;
> - else
> - vcpu->arch.hcr_el2 |= HCR_TWI;
> + vcpu->arch.hcr_el2 &= ~(HCR_TWE | HCR_TWI);
> }
>
> static inline void vcpu_set_wfx_traps(struct kvm_vcpu *vcpu)
> {
> - vcpu->arch.hcr_el2 |= HCR_TWE;
> - vcpu->arch.hcr_el2 |= HCR_TWI;
> + vcpu->arch.hcr_el2 &= ~(HCR_TWE | HCR_TWI);
> }
>
> static inline void vcpu_ptrauth_enable(struct kvm_vcpu *vcpu)
>
> Log when running ./run_test.sh timer (truncated for brevity) without the
> fixes:
>
> ...
> INFO: vtimer-busy-loop: waiting for interrupt...
> FAIL: vtimer-busy-loop: interrupt received after TVAL/WFI
> FAIL: vtimer-busy-loop: timer has expired
> INFO: vtimer-busy-loop: TVAL is 144646 ticks
> ...
> INFO: ptimer-busy-loop: waiting for interrupt...
> FAIL: ptimer-busy-loop: interrupt received after TVAL/WFI
> FAIL: ptimer-busy-loop: timer has expired
> INFO: ptimer-busy-loop: TVAL is 50384 ticks
> SUMMARY: 18 tests, 4 unexpected failures
>
> Log when running the same command with the series applied:
>
> ...
> INFO: vtimer-busy-loop: waiting for interrupt...
> INFO: vtimer-busy-loop: waiting for interrupt...
> INFO: vtimer-busy-loop: waiting for interrupt...
> PASS: vtimer-busy-loop: interrupt received after TVAL/WFI
> PASS: vtimer-busy-loop: timer has expired
> INFO: vtimer-busy-loop: TVAL is -56982 ticks
> ...
> INFO: ptimer-busy-loop: waiting for interrupt...
> INFO: ptimer-busy-loop: waiting for interrupt...
> PASS: ptimer-busy-loop: interrupt received after TVAL/WFI
> PASS: ptimer-busy-loop: timer has expired
> INFO: ptimer-busy-loop: TVAL is -22997 ticks
> SUMMARY: 18 tests
>
>
> Alexandru Elisei (4):
> arm: timer: Fix TVAL comparison for timer condition met
> arm: timer: Move the different tests into their own functions
> arm: timer: Test CVAL before interrupt pending state
> arm: timer: Take into account other wake-up events for the TVAL test
>
> arm/timer.c | 81 +++++++++++++++++++++++++++++++++++++++++++----------
> 1 file changed, 66 insertions(+), 15 deletions(-)
>
> --
> 2.34.1
>
Pushed to https://gitlab.com/rhdrjones/kvm-unit-tests/-/commits/arm/queue
Thanks,
drew
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
