[quagga-dev 4775] Re: Quagga OSPF and PPP Server

Hai Wang (汪海) pppoe.pppoa at gmail.com
Fri May 11 15:33:12 BST 2007


Hello Paul,

        I reviewed and  applied your patch, if I am not mistaken, the
patch almost replaced (ALL_LIST_ELEMENTS_RO  and  (ALL_LIST_ELEMENTS
with ALL_TABLE_NODES, which I think pretty is  'completed'. Please let
me know what else I should do with the patch?

Moreover,

      I have tested the patch with a few more minor changes, I found that


1. CPU usage still very high (95%) with 2K PPP sessions, but unlike
ospfd without the patch,  now ospfd is able to respond HELLO messages
from neighbors.

    I am wondering what else is comsuming the CPU?
    Do u think I should do the same work to area->oiflist too.

2. ospfd crashed when ppp session is disconnected with the following call stack


1. prefix_match at prefix.c 79

         79   if (n->prefixlen > p->prefixlen)
          n = 0x0

2. ospf_if_lookup_recv_if at ospf_interface.c 468
3. odpf_read at ospf-packet.c

Please shed the light on it.

Many thanks!

Hai


On 5/10/07, Hai Wang (汪海) <pppoe.pppoa at gmail.com> wrote:
> Thank u very much.
>
> On 5/10/07, Paul Jakma <paul at clubi.ie> wrote:
> > On Thu, 10 May 2007, Hai Wang (Íôº£) wrote:
> >
> > > Great, please let me know what you have done so far, definetly I will
> > > continue with your 'proceed in this direction' patch.
> >
> > You want to do something /like/ the attached diff:
> >
> > - change the global 'oiflist' linked-list index of struct
> >   ospf_interfaces to a route_table (binary trie).
> >
> > - fix visibility of index into lifetime of ospf_interfaces (ie the
> >   route_node refcounting - try a 'no network .....' command to see
> >   ;) )
> >
> > - figure out exactly what should be the key into the trie..
> >   - e.g. this approach may complicate any future unnumbered support..
> >
> > lots still to do with this patch I'm afraid.
> >
> > regards,
> > --
> > Paul Jakma      paul at clubi.ie   paul at jakma.org  Key ID: 64A2FF6A
> > Fortune:
> > It is now pitch dark.  If you proceed, you will likely fall into a pit.
> >
>




More information about the Quagga-dev mailing list