On Wed, Sep 19, 2018 at 10:20 AM Dave Jiang <[email protected]> wrote: > > During fakenuma processing in numa_emulation(), pi gets passed in and > processed as new fake numa nodes are being split out. Once the original > memory region is proccessed, it gets removed from the pi by > numa_remove_memblk_from() in emu_setup_memblk(). So entry 0 gets deleted > and the rest of the entries get moved up. Therefore we should always pass > in entry 0 for the next entry to process. > > Fixes: 1f6a2c6d9f121 ("x86/numa_emulation: Introduce uniform split > capability") > > Cc: Dan Williams <[email protected]> > Signed-off-by: Dave Jiang <[email protected]>
Thanks Dave! I missed this behavior in my testing. Reviewed-by: Dan Williams <[email protected]> > --- > arch/x86/mm/numa_emulation.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/arch/x86/mm/numa_emulation.c b/arch/x86/mm/numa_emulation.c > index b54d52a2d00a..a3ca8bf5afcb 100644 > --- a/arch/x86/mm/numa_emulation.c > +++ b/arch/x86/mm/numa_emulation.c > @@ -401,8 +401,8 @@ void __init numa_emulation(struct numa_meminfo > *numa_meminfo, int numa_dist_cnt) > ret = -1; > for_each_node_mask(i, physnode_mask) { We might put a comment here because the use of 0 is non-obvious on first glance. > ret = split_nodes_size_interleave_uniform(&ei, &pi, > - pi.blk[i].start, pi.blk[i].end, 0, > - n, &pi.blk[i], nid); > + pi.blk[0].start, pi.blk[0].end, 0, > + n, &pi.blk[0], nid); > if (ret < 0) > break; > if (ret < n) { >
