Re: [PATCH v3 8/8] KVM: selftests: Move large memslots above KVM internal memslots in _vm_create

[Andrew Jones <drjones@redhat.com>]

On Mon, Dec 16, 2019 at 01:39:01PM -0800, Ben Gardon wrote:
> KVM creates internal memslots between 3 and 4 GiB paddrs on the first
> vCPU creation. If memslot 0 is large enough it collides with these
> memslots an causes vCPU creation to fail. When requesting more than 3G,
> start memslot 0 at 4G in _vm_create.
> 
> Signed-off-by: Ben Gardon <bgardon@google.com>
> ---
>  tools/testing/selftests/kvm/lib/kvm_util.c | 15 +++++++++++----
>  1 file changed, 11 insertions(+), 4 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
> index 41cf45416060f..886d58e6cac39 100644
> --- a/tools/testing/selftests/kvm/lib/kvm_util.c
> +++ b/tools/testing/selftests/kvm/lib/kvm_util.c
> @@ -113,6 +113,8 @@ const char * const vm_guest_mode_string[] = {
>  _Static_assert(sizeof(vm_guest_mode_string)/sizeof(char *) == NUM_VM_MODES,
>  	       "Missing new mode strings?");
>  
> +#define KVM_INTERNAL_MEMSLOTS_START_PADDR (3UL << 30)
> +#define KVM_INTERNAL_MEMSLOTS_END_PADDR (4UL << 30)
>  /*
>   * VM Create
>   *
> @@ -128,13 +130,16 @@ _Static_assert(sizeof(vm_guest_mode_string)/sizeof(char *) == NUM_VM_MODES,
>   *
>   * Creates a VM with the mode specified by mode (e.g. VM_MODE_P52V48_4K).
>   * When phy_pages is non-zero, a memory region of phy_pages physical pages
> - * is created and mapped starting at guest physical address 0.  The file
> - * descriptor to control the created VM is created with the permissions
> - * given by perm (e.g. O_RDWR).
> + * is created. If phy_pages is less that 3G, it is mapped starting at guest
> + * physical address 0. If phy_pages is greater than 3G it is mapped starting
> + * 4G into the guest physical address space to avoid KVM internal memslots
> + * which map the region between 3G and 4G. The file descriptor to control the
> + * created VM is created with the permissions given by perm (e.g. O_RDWR).
>   */
>  struct kvm_vm *_vm_create(enum vm_guest_mode mode, uint64_t phy_pages, int perm)
>  {
>  	struct kvm_vm *vm;
> +	uint64_t guest_paddr = 0;
>  
>  	DEBUG("Testing guest mode: %s\n", vm_guest_mode_string(mode));
>  
> @@ -227,9 +232,11 @@ struct kvm_vm *_vm_create(enum vm_guest_mode mode, uint64_t phy_pages, int perm)
>  
>  	/* Allocate and setup memory for guest. */
>  	vm->vpages_mapped = sparsebit_alloc();
> +	if (guest_paddr + phy_pages > KVM_INTERNAL_MEMSLOTS_START_PADDR)
> +		guest_paddr = KVM_INTERNAL_MEMSLOTS_END_PADDR;
>  	if (phy_pages != 0)
>  		vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS,
> -					    0, 0, phy_pages, 0);
> +					    guest_paddr, 0, phy_pages, 0);
>  
>  	return vm;
>  }
> -- 
> 2.24.1.735.g03f4e72817-goog
>

I feel like this function is becoming too magic and it'll be more
complicated for tests that need to add additional memory regions
to know what physical addresses are available. Maybe we should assert
if we can't allocate more than 3G at offset zero and also provide
another interface for allocating at an offset input by the user,
as long as the offset is 4G or above (asserting when it isn't)?

Thanks,
drew