Re: move_pages fails on 2 node system
Ganapatrao Kulkarni <[email protected]> Wed, 17 Sep 2014 09:01:09 +0530
| Newsgroups | org.kernel.vger.linux-numa |
|---|---|
| Message-ID | <CAFpQJXXViCbqcWxtHa2A1dN-gNOLg3_n7Q+HAXU14bv3S2LuVw@mail.gmail.com> |
On Wed, Sep 17, 2014 at 4:08 AM, David Rientjes <[email protected]> wrote: > On Tue, 16 Sep 2014, Filipe Brandenburger wrote: > >> On Tue, Sep 16, 2014 at 1:22 PM, David Rientjes <[email protected]> wrote: >> > If this patch fixes it, then why is numa_max_node() returning 1 on a 2 >> > node system? >> >> Because the node ids are 0 and 1 :-) >> >> Patch makes sense to me. In particular, both test/migrate_pages.c and >> test/mbind_mig_pages.c have "nr_nodes = numa_max_node()+1;" in them. >> > > Patch makes no sense to me if there's no node 0. yes, one more bug(likely), if no nodes, this function should return -1, but it returns 0. for no nodes case, if function returns -1, then +1 makes sense for all cases. --- libnuma.c.orig 2014-09-17 08:58:21.119671188 +0530 +++ libnuma.c 2014-09-17 08:59:14.731333190 +0530 @@ -330,7 +330,7 @@ d = opendir("/sys/devices/system/node"); if (!d) { - maxconfigurednode = 0; + maxconfigurednode = -1; } else { while ((de = readdir(d)) != NULL) { int nd; thanks Ganapat