On 3/27/17 9:08 PM, Eric W. Biederman wrote: > I believe we should just kill MAX_NEW_LABELS. > > I think the only significant change from your patch is the removal of an > array from mpls_route_config. > > With the removal of MAX_NEW_LABELS I would replace it by a sanity check > in mpls_rt_alloc that verifies that the amount we are going to allocate > for struct mpls_route is < PAGE_SIZE. Anything larger is just > asking for trouble. > > That should put our practical limit just a little bit below 32 nexthops > adding 32 labels each.
The 4096 limit works nice for mpls_route but not for lwt encap info. That struct is 4-bytes + the labels. Seems odd to let ip->mpls allow up to 255 labels (max for u8) while mpls->mpls has a limit. I'm going to send v2 soon with the 4096 limit for mpls_route total size but keeping the MAX_NEW_LABELS with a count of 30 for both. That keeps the two paths consistent, keeps mpls_iptunnel_encap < 128 bytes for the max allocation and 30 labels is allows plenty of options for TE and SR.
