Hello, Kame.

On Thu, Nov 08, 2012 at 01:37:50PM +0900, Kamezawa Hiroyuki wrote:
> How about
> enum {
>    __CGROUP_FREEZING,
>    __CGROUP_FROZEN,
> };
> 
> #define CGROUP_FREEZER_STATE_MASK  0x3
> #define CGROUP_FREEZER_STATE(state)	((state) & CGROUP_FREEZER_STATE_MASK)
> #define CGROUP_THAW(state)	(CGROUP_FREEZER_STATE(state) == 0)
> #define CGROUP_FREEZING(state)	(CGROUP_FREEZER_STATE(state) == __CGROUP_FREEZING)
> #define CGROUP_FROZEN(state)\
> 	(CGROUP_FREEZER_STATE(state) == (__CGROUP_FREEZING | __CGROUP_FROZEN))

I think it's a bit overdone and we have cases where we test for
FREEZING regardless of FROZEN and cases where test for FREEZING &&
!FROZEN.  We can have, say, CGROUP_FREZING() and then
CGROUP_FREEZING_BUT_NOT_FROZEN(), but it feels more like obfuscation
than anything else.

> >@@ -290,9 +284,9 @@ static int freezer_write(struct cgroup *cgroup, struct cftype *cft,
> >  {
> >  	bool freeze;
> >
> >-	if (strcmp(buffer, freezer_state_strs[CGROUP_THAWED]) == 0)
> >+	if (strcmp(buffer, freezer_state_strs(0)) == 0)
> 
> Can't we have a name for "0" ?

We can define CGROUP_THAWED to be zero.  I'd rather keep it explicitly
zero tho.  freezer state is mask of freezing and frozen flags and no
flag set means thawed.

Thanks.

-- 
tejun