Re: [PATCH] cpumask: alloc blank cpumask left over

[Avi Kivity <avi@redhat.com>]

Rusty Russell wrote:
> On Fri, 5 Jun 2009 03:26:57 pm Yinghai Lu wrote:
>   
>> Rusty Russell wrote:
>>     
>>> On Fri, 5 Jun 2009 06:31:31 am Yinghai Lu wrote:
>>>       
>>>> avoid suprise when MAXSMP is enabled
>>>>
>>>> Signed-off-by: Yinghai Lu <yinghai.lu@kernel.org>
>>>>         
>>> I understand the temptation, but two questions arise:
>>> 1) Shouldn't we actually audit to see if any of these are currently
>>> problems,
>>>       
>> those are defined as static cpumask_var_t, and if MAXSMP is not used, they
>> are cleared already
>>     
>
> OK, here's what I've got in my tree.  Ingo, I think this should go in the
> current -rc to avoid nasty bugs.
>
> BTW, the original alloc_cpumask_var did zero; that was dropped after arguments
> over efficiency and fitting with other interfaces, but I clearly had the old
> semantics in my head for a while.
>
>   

Using __GFP_ZERO is equivalent to using memset() instead of 
cpumask_clear().  It's better to call cpumask_clear() or provide an API 
to alloc+clear.

Further, what about the non-MAXSMP case:


    static inline bool alloc_cpumask_var(cpumask_var_t *mask, gfp_t flags)
    {
            return true;
    }


We explicity clear on MAXSMP and rely on static initialization for the 
non-MAXSMP, laying a neat trap for anyone who makes the variable 
non-static.  Let's be less subtle that that.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.