Re: [PATCH v3 0/5] mm: Enable CONFIG_NODES_SPAN_OTHER_NODES by default for NUMA

On 03/30/20 at 09:44am, Michal Hocko wrote:

On Sun 29-03-20 08:19:24, Baoquan He wrote:

quoted

On 03/28/20 at 11:31am, Hoan Tran wrote:

quoted

In NUMA layout which nodes have memory ranges that span across other nodes,
the mm driver can detect the memory node id incorrectly.

For example, with layout below
Node 0 address: 0000 xxxx 0000 xxxx
Node 1 address: xxxx 1111 xxxx 1111

Sorry, I read this example several times, but still don't get what it
means. Can it be given with real hex number address as an exmaple? I
mean just using the memory layout you have seen from some systems. The
change looks interesting though.

Does this make it more clear?
           physical address range and its node associaion
         [0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1]

I later read it again, have got what Hoan is trying to say, thanks.

I think the change in this patchset makes sense, still have some concern
though, let me add comment in other thread.