Re: [vpp-dev]: SIGSEV in mpls_tunnel_collect_forwarding

2021-10-21 Thread Rajith PR via lists.fd.io
also work, but it would mean a no-op > restack for the tunnel. Not walking the new child is more efficient. > > > > /neale > > > > *From: *vpp-dev@lists.fd.io on behalf of Stanislav > Zaikin via lists.fd.io > *Date: *Thursday, 21 October 2021 at 17:58 > *T

Re: [vpp-dev]: SIGSEV in mpls_tunnel_collect_forwarding

2021-10-21 Thread Neale Ranns
Date: Thursday, 21 October 2021 at 17:58 To: Rajith PR Cc: vpp-dev Subject: Re: [vpp-dev]: SIGSEV in mpls_tunnel_collect_forwarding Hi Rajith, Looks like each tunnel uses the same path-list (and different path extensions), and when the path has more than FIB_PATH_LIST_POPULAR (64) children - it

Re: [vpp-dev]: SIGSEV in mpls_tunnel_collect_forwarding

2021-10-21 Thread Stanislav Zaikin
Hi Rajith, Looks like each tunnel uses the same path-list (and different path extensions), and when the path has more than FIB_PATH_LIST_POPULAR (64) children - it updates all the children with that "popular" flag. And at that point there are no path extensions yet on the last child. So, I suppose

[vpp-dev]: SIGSEV in mpls_tunnel_collect_forwarding

2021-10-20 Thread Rajith PR via lists.fd.io
Hi All, We are seeing below crash when creating mpls tunnels. The issue is easily reproducible just have to create around 100 mpls tunnels. It seems path_ext is coming as NULL. Thread 1 "vpp_main" received signal SIGSEGV, Segmentation fault. 0x7fe017bba268 in mpls_tunnel_collect_forwarding (p