Re: [PATCH] kvm: don not fail when cache is sufficient

[Avi Kivity <avi@redhat.com>]

On 04/17/2010 12:02 PM, Lai Jiangshan wrote:
> cleanup: don not fail when cache is sufficient
>
> Signed-off-by: Lai Jiangshan<laijs@cn.fujitsu.com>
> ---
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index c8c074c..90f666e 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -304,7 +304,7 @@ static int mmu_topup_memory_cache(struct kvm_mmu_memory_cache *cache,
>   	while (cache->nobjs<  ARRAY_SIZE(cache->objects)) {
>   		obj = kmem_cache_zalloc(base_cache, GFP_KERNEL);
>   		if (!obj)
> -			return -ENOMEM;
> +			return cache->nobjs>= min ? 0 : -ENOMEM;
>   		cache->objects[cache->nobjs++] = obj;
>   	}
>   	return 0;
> @@ -326,7 +326,7 @@ static int mmu_topup_memory_cache_page(struct kvm_mmu_memory_cache *cache,
>   	while (cache->nobjs<  ARRAY_SIZE(cache->objects)) {
>   		page = alloc_page(GFP_KERNEL);
>   		if (!page)
> -			return -ENOMEM;
> +			return  cache->nobjs>= min ? 0 : -ENOMEM;
>   		set_page_private(page, 0);
>   		cache->objects[cache->nobjs++] = page_address(page);
>   	}
>
>    

Have you ever observed this?  My feeling is that if we're a few pages 
from -ENOMEM, it doesn't really matter if we die now or a few calls 
away.  Note alloc_page() can call back into kvm via the shrinker to free 
pages, so an OOM will be external to kvm and we won't really be able to 
handle it.

-- 
error compiling committee.c: too many arguments to function