Re: [PATCH] kvm:x86:Call BUG_ON marco if calls to x86_set_memory_region fail in kvm_arch_destroy

[Paolo Bonzini <pbonzini@redhat.com>]

On 06/08/2015 16:13, Nicholas Krause wrote:
> This fixes the calls to x86_memory_region to trigger a kernel
> oopes for tracing if a bug arises in the function kvm_arch_destroy
> due to the failing of any of its internal calls to x86_set_memory_region.
> 
> Signed-off-by: Nicholas Krause <xerofoify@gmail.com>
> ---
>  arch/x86/kvm/x86.c | 10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 5ef2560..3866cf3 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -7745,6 +7745,7 @@ EXPORT_SYMBOL_GPL(x86_set_memory_region);
>  
>  void kvm_arch_destroy_vm(struct kvm *kvm)
>  {
> +	int ret;
>  	if (current->mm == kvm->mm) {
>  		/*
>  		 * Free memory regions allocated on behalf of userspace,
> @@ -7754,13 +7755,16 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
>  		struct kvm_userspace_memory_region mem;
>  		memset(&mem, 0, sizeof(mem));
>  		mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
> -		x86_set_memory_region(kvm, &mem);
> +		ret = x86_set_memory_region(kvm, &mem);
> +		BUG_ON(ret);
>  
>  		mem.slot = IDENTITY_PAGETABLE_PRIVATE_MEMSLOT;
> -		x86_set_memory_region(kvm, &mem);
> +		ret = x86_set_memory_region(kvm, &mem);
> +		BUG_ON(ret);
>  
>  		mem.slot = TSS_PRIVATE_MEMSLOT;
> -		x86_set_memory_region(kvm, &mem);
> +		ret = x86_set_memory_region(kvm, &mem);
> +		BUG_ON(ret);
>  	}
>  	kvm_iommu_unmap_guest(kvm);
>  	kfree(kvm->arch.vpic);
> 

No, not at all!  Leaking 12k of RAM is not a reason to crash the
machine.  WARN_ON would be acceptable.

Paolo