svn commit: r262196 - head/sys/netpfil/pf

[Gleb Smirnoff]

On Wed, Feb 19, 2014 at 08:49:01PM +0200, Mikolaj Golub wrote:
M> > -SLIST_HEAD(pf_overload_head, pf_overload_entry);
M> > +struct pf_overload_head {
M> > +	SLIST_HEAD(, pf_overload_entry) head;
M> > +	struct vnet			*vnet;
M> > +};
M> >  static VNET_DEFINE(struct pf_overload_head, pf_overloadqueue);
M> >  #define V_pf_overloadqueue	VNET(pf_overloadqueue)
M> 
M> ...
M> 
M> > -	SLIST_INIT(&V_pf_overloadqueue);
M> > +	SLIST_INIT(&V_pf_overloadqueue.head);
M> >  	TASK_INIT(&V_pf_overloadtask, 0, pf_overload_task, &V_pf_overloadqueue);
M> > +	V_pf_overloadqueue.vnet = curvnet;
M> 
M> Why not pass vnet as a context to pf_overload_task instead of
M> &V_pf_overloadqueue? Then you would not need this hack with storing a
M> vnet inside a vnet variable.

Yes, that would look much better.

The pf_overloadqueue can be made global after that. Its lock is
already global.

-- 
Totus tuus, Glebius.