[PATCH v3 kvmtool 06/13] arm/arm64: Fail if RAM size is too large for 32-bit guests

[Andre Przywara]

On Wed, 25 May 2022 12:23:38 +0100
Alexandru Elisei <alexandru.elisei at arm.com> wrote:

> For 64-bit guests, kvmtool exists with an error in kvm__get_vm_type() if
> the memory size is larger than what KVM supports. For 32-bit guests, the
> RAM size is silently rounded down to ARM_LOMAP_MAX_MEMORY in
> kvm__arch_init().
> 
> Be consistent and exit with an error when the user has configured the
> wrong RAM size for 32-bit guests.
> 
> Signed-off-by: Alexandru Elisei <alexandru.elisei at arm.com>

This patch looks fine, as it improves the situation and maintains the
current support level, but it turns out that the AArch32 memory limitation
is bogus: With LPAE (which is mandatory for any system supporting KVM),
there is no reason we can't have memory beyond 4GB of guest PA.
In fact, after removing this check, I managed to boot a 32GB AArch32 guest
just fine (on a Juno), and a 2919MB guest on an ARMv7 box.
It might be useful to lift this limit, as this allows easy testing of the
arm(32) kernel's highmem support, but this would be a separate patch, I
guess.

Reviewed-by: Andre Przywara <andre.przywara at arm.com>

Cheers,
Andre
 

> ---
>  arm/aarch32/kvm.c | 4 ++++
>  arm/aarch64/kvm.c | 5 +++++
>  arm/kvm.c         | 2 +-
>  3 files changed, 10 insertions(+), 1 deletion(-)
> 
> diff --git a/arm/aarch32/kvm.c b/arm/aarch32/kvm.c
> index ae33ac92479a..9d68d7a15ee2 100644
> --- a/arm/aarch32/kvm.c
> +++ b/arm/aarch32/kvm.c
> @@ -2,4 +2,8 @@
>  
>  void kvm__arch_validate_cfg(struct kvm *kvm)
>  {
> +	if (kvm->cfg.ram_size > ARM_LOMAP_MAX_MEMORY) {
> +		die("RAM size 0x%llx exceeds maximum allowed 0x%llx",
> +		    kvm->cfg.ram_size, ARM_LOMAP_MAX_MEMORY);
> +	}
>  }
> diff --git a/arm/aarch64/kvm.c b/arm/aarch64/kvm.c
> index ca348f118a56..2134528bf7f2 100644
> --- a/arm/aarch64/kvm.c
> +++ b/arm/aarch64/kvm.c
> @@ -39,6 +39,11 @@ int vcpu_affinity_parser(const struct option *opt, const char *arg, int unset)
>  
>  void kvm__arch_validate_cfg(struct kvm *kvm)
>  {
> +	if (kvm->cfg.arch.aarch32_guest &&
> +	    kvm->cfg.ram_size > ARM_LOMAP_MAX_MEMORY) {
> +		die("RAM size 0x%llx exceeds maximum allowed 0x%llx",
> +		    kvm->cfg.ram_size, ARM_LOMAP_MAX_MEMORY);
> +	}
>  }
>  
>  /*
> diff --git a/arm/kvm.c b/arm/kvm.c
> index c5913000e1ed..af0feae495d7 100644
> --- a/arm/kvm.c
> +++ b/arm/kvm.c
> @@ -65,7 +65,7 @@ void kvm__arch_init(struct kvm *kvm, const char *hugetlbfs_path, u64 ram_size)
>  	 * If using THP, then our minimal alignment becomes 2M.
>  	 * 2M trumps 64K, so let's go with that.
>  	 */
> -	kvm->ram_size = min(ram_size, (u64)ARM_MAX_MEMORY(kvm));
> +	kvm->ram_size = ram_size;
>  	kvm->arch.ram_alloc_size = kvm->ram_size + SZ_2M;
>  	kvm->arch.ram_alloc_start = mmap_anon_or_hugetlbfs(kvm, hugetlbfs_path,
>  						kvm->arch.ram_alloc_size);