On 04/07/2017 15:58, Andrew Jones wrote:
> Add a test for the vtimer. I've tested on
> accel=tcg
> accel=kvm,kernel-irqchip=on : on thunderx, seattle, and mustang
> accel=kvm,kernel-irqchip=off : on thunderx and mustang
>
> It works everywhere without modification except on mustang. My mustang
> seems to have a hardware problem requiring bit 31 of CNTV_CVAL_EL0
> to be clear before attempting to enable (setting CNTV_CTL_EL0.ENABLE),
> otherwise the VCPU hangs... My mustang's CNTV_CVAL_EL0 also has the
> strange issue of not being able to set the upper 8 bits. Well, fixing
> hardware is outside my job description, so "oh, well..." If you hit
> the same problem, try the hack below, which allows tests to run and
> pass for me.
Applied and pushed, thanks.
Paolo
> diff --git a/arm/timer.c b/arm/timer.c
> index 89f4c9400ab69..18b8c4773c92f 100644
> --- a/arm/timer.c
> +++ b/arm/timer.c
> @@ -73,7 +73,7 @@ static void test_vtimer(void)
> report_prefix_push("vtimer-busy-loop");
>
> /* Enable the timer */
> - write_sysreg(~0, cntv_cval_el0);
> + write_sysreg(~0 & ~(1<<31), cntv_cval_el0);
> isb();
> write_sysreg(CNTV_CTL_ENABLE, cntv_ctl_el0);
>
>
> Alexander Graf (1):
> arm64: add vtimer test
>
> Andrew Jones (1):
> arm64: timer: check for vtimer interrupt
>
> arm/Makefile.arm64 | 2 +-
> arm/timer.c | 134
> +++++++++++++++++++++++++++++++++++++++++++++++++++++
> arm/unittests.cfg | 6 +++
> lib/arm/asm/gic.h | 4 ++
> 4 files changed, 145 insertions(+), 1 deletion(-)
> create mode 100644 arm/timer.c
>
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
