Re: [iptables PATCH] list: fix prefetch dummy

[Alexander Duyck <alexander.h.duyck@redhat.com>]

On 04/07/2015 01:37 AM, Jan Engelhardt wrote:
> On Tuesday 2015-04-07 02:38, Alexander Duyck wrote:
>
>> On 04/06/2015 11:05 AM, Arturo Borrero Gonzalez wrote:
>>> linux_list.h:381:59: warning: right-hand operand of comma expression has no effect [-Wunused-value]
>>>    for (pos = list_entry((head)->next, typeof(*pos), member), \
>>>                                                             ^
>>> libiptc.c:552:2: note: in expansion of macro 'list_for_each_entry'
>>>    list_for_each_entry(c, &h->chains, list) {
>>>    ^
>>>
>>> [ Patch copied from one similar of Patrick McHardy on libnftnl ]
>>>
>>> Signed-off-by: Arturo Borrero Gonzalez <arturo.borrero.glez@gmail.com>
>>> ---
>>>   libiptc/linux_list.h |    2 +-
>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/libiptc/linux_list.h b/libiptc/linux_list.h
>>> index abdcf88..559e33c 100644
>>> --- a/libiptc/linux_list.h
>>> +++ b/libiptc/linux_list.h
>>> @@ -27,7 +27,7 @@
>>>   	1; \
>>>   })
>>>   
>>> -#define prefetch(x)		1
>>> +#define prefetch(x)		((void)0)
>>>   
>>>   /* empty define to make this work in userspace -HW */
>>>   #define smp_wmb()
>>>
>> Why not just use "do {} while (0)"?  I know that is what is used in the
>> kernel for functions that don't do anything.
> I may be getting the terms wrong, but:
> do{}while(0) is not an expression, it is a (block) control statement.
> In particular, do{}while(0) won't evaluate to an rvalue.

Right.  That is the point in this case.  I am assuming what Arturo is 
trying to accomplish since you shouldn't be able to evaluate ((void)0) 
as an rvalue either.

The prefetch(x) is an empty statement which is being cast as a void to 
avoid a compiler warning, so it falls into the first case as defined in: 
http://kernelnewbies.org/FAQ/DoWhile0

- Alex