Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200311110237.5731-2-srikar@linux.vnet.ibm.com>
 <20200311115735.GM23944@dhcp22.suse.cz> <20200312052707.GA3277@linux.vnet.ibm.com>
 <C5560C71-483A-41FB-BDE9-526F1E0CFA36@linux.vnet.ibm.com> <5e5c736a-a88c-7c76-fc3d-7bc765e8dcba@suse.cz>
 <20200312131438.GB3277@linux.vnet.ibm.com> <61437352-8b54-38fa-4471-044a65c9d05a@suse.cz>
 <20200312161310.GC3277@linux.vnet.ibm.com> <e115048c-be38-c298-b8d1-d4b513e7d2fb@suse.cz>
 <CAAmzW4OFy51BhAT62tdVQD52NNMWm+UPgoGAX97omY7P+nJ+5w@mail.gmail.com>
 <20200313110440.GA25144@linux.vnet.ibm.com> <06be5908-9af6-2892-0333-e9558b2cf474@suse.cz>
In-Reply-To: <06be5908-9af6-2892-0333-e9558b2cf474@suse.cz>
From:   Joonsoo Kim <js1304@gmail.com>
Date:   Mon, 16 Mar 2020 17:15:19 +0900
Message-ID: <CAAmzW4ME_WLYZPCt4+82RNKstv-H=LK5MKGGJR=6ha-ALS+FSw@mail.gmail.com>
Subject: Re: [PATCH 1/3] powerpc/numa: Set numa_node for all possible cpus
To:     Vlastimil Babka <vbabka@suse.cz>
Cc:     Srikar Dronamraju <srikar@linux.vnet.ibm.com>,
        Sachin Sant <sachinp@linux.vnet.ibm.com>,
        Michal Hocko <mhocko@kernel.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Linux Memory Management List <linux-mm@kvack.org>,
        Mel Gorman <mgorman@suse.de>,
        "Kirill A. Shutemov" <kirill@shutemov.name>,
        Andrew Morton <akpm@linux-foundation.org>,
        linuxppc-dev <linuxppc-dev@lists.ozlabs.org>,
        Christopher Lameter <cl@linux.com>,
        Joonsoo Kim <iamjoonsoo.kim@lge.com>,
        Kirill Tkhai <ktkhai@virtuozzo.com>,
        Michael Ellerman <mpe@ellerman.id.au>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

2020=EB=85=84 3=EC=9B=94 13=EC=9D=BC (=EA=B8=88) =EC=98=A4=ED=9B=84 8:38, V=
lastimil Babka <vbabka@suse.cz>=EB=8B=98=EC=9D=B4 =EC=9E=91=EC=84=B1:
>
> On 3/13/20 12:04 PM, Srikar Dronamraju wrote:
> >> I lost all the memory about it. :)
> >> Anyway, how about this?
> >>
> >> 1. make node_present_pages() safer
> >> static inline node_present_pages(nid)
> >> {
> >> if (!node_online(nid)) return 0;
> >> return (NODE_DATA(nid)->node_present_pages);
> >> }
> >>
> >
> > Yes this would help.
>
> Looks good, yeah.
>
> >> 2. make node_to_mem_node() safer for all cases
> >> In ppc arch's mem_topology_setup(void)
> >> for_each_present_cpu(cpu) {
> >>  numa_setup_cpu(cpu);
> >>  mem_node =3D node_to_mem_node(numa_mem_id());
> >>  if (!node_present_pages(mem_node)) {
> >>   _node_numa_mem_[numa_mem_id()] =3D first_online_node;
> >>  }
> >> }
> >>
> >
> > But here as discussed above, we miss the case of possible but not prese=
nt nodes.
> > For such nodes, the above change may not update, resulting in they stil=
l
> > having 0. And node 0 can be only possible but not present.

Oops, I don't read full thread so miss the case.

> So is there other way to do the setup so that node_to_mem_node() returns =
an
> online+present node when called for any possible node?

Two changes seems to be sufficient.

1. initialize all node's _node_numa_mem_[] =3D first_online_node in
mem_topology_setup()
2. replace the node with online+present node for _node_to_mem_node_[]
in set_cpu_numa_mem().

 static inline void set_cpu_numa_mem(int cpu, int node)
 {
        per_cpu(_numa_mem_, cpu) =3D node;
+       if (!node_present_pages(node))
+               node =3D first_online_node;
        _node_numa_mem_[cpu_to_node(cpu)] =3D node;
 }
 #endif

With these two change, we can safely call node_to_mem_node() anywhere.

Thanks.