Re: [PATCH 1/3] KVM: selftests: Allocate additional space for latency samples

[Ricardo Koller <ricarkol@google.com>]

Hi Colton,

On Tue, Nov 15, 2022 at 05:32:56PM +0000, Colton Lewis wrote:
> Allocate additional space for latency samples. This has been separated
> out to call attention to the additional VM memory allocation. The test
> runs out of physical pages without the additional allocation. The 100
> multiple for pages was determined by trial and error. A more
> well-reasoned calculation would be preferable.
> 
> Signed-off-by: Colton Lewis <coltonlewis@google.com>
> ---
>  tools/testing/selftests/kvm/lib/perf_test_util.c | 12 ++++++++++--
>  1 file changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/lib/perf_test_util.c b/tools/testing/selftests/kvm/lib/perf_test_util.c
> index 137be359b09e..a48904b64e19 100644
> --- a/tools/testing/selftests/kvm/lib/perf_test_util.c
> +++ b/tools/testing/selftests/kvm/lib/perf_test_util.c
> @@ -38,6 +38,12 @@ static bool all_vcpu_threads_running;
>  
>  static struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
>  
> +#define SAMPLES_PER_VCPU 1000
> +#define SAMPLE_CAPACITY (SAMPLES_PER_VCPU * KVM_MAX_VCPUS)
> +
> +/* Store all samples in a flat array so they can be easily sorted later. */
> +uint64_t latency_samples[SAMPLE_CAPACITY];
> +
>  /*
>   * Continuously write to the first 8 bytes of each page in the
>   * specified region.
> @@ -122,7 +128,7 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int nr_vcpus,
>  {
>  	struct perf_test_args *pta = &perf_test_args;
>  	struct kvm_vm *vm;
> -	uint64_t guest_num_pages, slot0_pages = 0;
> +	uint64_t guest_num_pages, sample_pages, slot0_pages = 0;
>  	uint64_t backing_src_pagesz = get_backing_src_pagesz(backing_src);
>  	uint64_t region_end_gfn;
>  	int i;
> @@ -161,7 +167,9 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int nr_vcpus,
>  	 * The memory is also added to memslot 0, but that's a benign side
>  	 * effect as KVM allows aliasing HVAs in meslots.
>  	 */
> -	vm = __vm_create_with_vcpus(mode, nr_vcpus, slot0_pages + guest_num_pages,
> +	sample_pages = 100 * sizeof(latency_samples) / pta->guest_page_size;

I don't think there's any need to guess. The number of accesses is
vcpu_args->pages (one access per guest page). So all memory could be
allocated dynamically to hold "vcpu_args->pages * sample_sz".

> +	vm = __vm_create_with_vcpus(mode, nr_vcpus,
> +				    slot0_pages + guest_num_pages + sample_pages,
>  				    perf_test_guest_code, vcpus);
>  
>  	pta->vm = vm;
> -- 
> 2.38.1.431.g37b22c650d-goog
> 

Thanks,
Ricardo