Re: [PATCH v2] mm/page_alloc: detect allocation forbidden by cpuset and bail out early

[Feng Tang]

On Mon, Sep 13, 2021 at 01:45:51PM +0200, Michal Hocko wrote:
> On Mon 13-09-21 19:34:23, Feng Tang wrote:
> > On Mon, Sep 13, 2021 at 11:15:54AM +0200, Michal Hocko wrote:
> [...] 
> > > > +/* Whether the 'nodes' are all movable nodes */
> > > > +static inline bool movable_only_nodes(nodemask_t *nodes)
> > > > +{
> > > > +	struct zonelist *zonelist;
> > > > +	struct zoneref *z;
> > > > +
> > > > +	zonelist = &(first_online_pgdat())->node_zonelists[ZONELIST_FALLBACK];
> > > 
> > > This will work but it just begs a question why you haven't chosen a node
> > > from the given nodemask. So I believe it would be easier to read if you
> > > did
> > > 	zonelist = NODE_DATA(first_node(nodes))->node_zonelists[ZONELIST_FALLBACK]
> > 
> > This was also my first try to get the 'zonelist', but from the
> > update_nodemask(), the nodemask could be NULL.
> 
> I guess you meant to say s@NULL@empty@
> While this complicates things a bit it is nothing really hard to work
> around. You simply check for nodes_empty() and return false because such
> a nodemask cannot by definition be movable only.

Yes, a nodes_empty() check can solve it, thanks,

- Feng

> -- 
> Michal Hocko
> SUSE Labs