Re: [next:akpm 129/309] net/core/sock.c:274:36: error: initializer element is not constant

[Mel Gorman <mgorman@suse.de>]

On Tue, Jul 24, 2012 at 05:08:04PM -0400, John David Anglin wrote:
> On 24-Jul-12, at 3:48 AM, Fengguang Wu wrote:
> 
> >Here is the line in sock.i:
> >
> >struct static_key memalloc_socks = ((struct static_key) { .enabled
> >= ((atomic_t) { (0) }) });
> 
> 
> The above line contains two compound literals.  It also uses a
> designated initializer
> to initialize the field enabled.  A compound literal is not a
> constant expression.
> 
> The location of the above statement isn't fully clear, but if a
> compound literal occurs
> outside the body of a function, the initializer list must consist of
> constant expressions.
> 
> Removing "(atomic_t)" from the define results in a constant expression.
> 
> Test case:
> 
> typedef struct { long enabled; } atomic_t;
> struct static_key { atomic_t enabled; int x; };
> struct static_key memalloc_socks = ((struct static_key) { .enabled > ((atomic_t) { (0) }) });
> 

Thanks John for that explanation, it clears it up. The source of the
above line was linux-next/master:net/core/sock.c due to some patches I
merged. The exact line looks like this

struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;

and that thing in turns looks like

#define STATIC_KEY_INIT_FALSE ((struct static_key) \
        { .enabled = ATOMIC_INIT(0), .entries = (void *)0 })

This is a standard use of a static key (http://lwn.net/Articles/487426/)
and as I expect there will be more use of this feature in the future
it's good to get it fixed up first.

Thanks James for picking this up and putting a changelog on it.

-- 
Mel Gorman
SUSE Labs