Re: [PATCH v2] x86/PoD: shorten certain operations on higher order ranges

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 29/09/15 13:57, Jan Beulich wrote:
>>>> On 29.09.15 at 14:20, <andrew.cooper3@citrix.com> wrote:
>> On 28/09/15 15:30, Jan Beulich wrote:
>>> --- a/xen/arch/x86/mm/p2m-pod.c
>>> +++ b/xen/arch/x86/mm/p2m-pod.c
>>> @@ -119,20 +119,23 @@ p2m_pod_cache_add(struct p2m_domain *p2m
>>>  
>>>      unlock_page_alloc(p2m);
>>>  
>>> -    /* Then add the first one to the appropriate populate-on-demand list */
>>> -    switch(order)
>>> +    /* Then add to the appropriate populate-on-demand list. */
>>> +    switch ( order )
>>>      {
>>> +    case PAGE_ORDER_1G:
>>> +        for ( i = 0; i < (1UL << PAGE_ORDER_1G); i += 1UL << PAGE_ORDER_2M )
>>> +            page_list_add_tail(page + i, &p2m->pod.super);
>>> +        break;
>>>      case PAGE_ORDER_2M:
>>> -        page_list_add_tail(page, &p2m->pod.super); /* lock: page_alloc */
>>> -        p2m->pod.count += 1 << order;
>>> +        page_list_add_tail(page, &p2m->pod.super);
>>>          break;
>>>      case PAGE_ORDER_4K:
>>> -        page_list_add_tail(page, &p2m->pod.single); /* lock: page_alloc */
>>> -        p2m->pod.count += 1;
>>> +        page_list_add_tail(page, &p2m->pod.single);
>>>          break;
>>>      default:
>>>          BUG();
>>>      }
>>> +    p2m->pod.count += 1 << order;
>> 1UL
> Not really - the field is a "long" one, so at best 1L or 1U. And then
> all the valid order values are visible right above, for none of them
> it makes a difference, and there are ample similar uses scattered
> around the file (yes, bad examples are no excuse, but in cases
> where the suffix doesn't really matter I think it is better to omit it).
>
>> Otherwise, Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
> Let me know regrading this one,

For sanity sake, I would suggest going with 1L as one less place to go
wrong when we gain 512GB superpages.

~Andrew