> Author:    meem <Peter.Memishian at Sun.COM>
 > Changeset: 8512:34e3556adf64
 > Comments:
 > 6789235 INADDR_ANY ILMs can trigger an assertion failure in IPMP environments
 > Clearview IPMP piece of IPv6 RTM_IFINFO IFF_UP issue
 > "netstat -f inet6 -p" chokes badly when IPMP is in-use inside a zone
 > netstat has longstanding continuation line problems

The webrev for this is at:

  http://zhadum.east/ws/clearview/clearview-ipmpdev/webrev.misc/

To expand on the above:

  * The first item fixes a Nevada bug Seb found during IPMP stress testing.
    See the bug report for details.

  * The second item is a partial fix for the remaining test suite failure.
    The remaining (non-trivial) issue is that Nevada currently calls
    ipif_resolver_up() *after* ipif_ndp_up(), which clears the
    ipif_addr_ready flag that ipif_ndp_up() just set.  Reordering these
    calls is tricky because of the EINPROGRESS handling, so before going
    ahead with that I want to be sure it's the right fix.  As such, Jim
    is looking into this and will report back.

  * The third item is a workaround for an unfortunate interaction I
    noticed while using a shared-stack zone with IPv6 and IPMP: since the
    NCEs are on underlying interfaces, "netstat -f inet6 -p" cannot
    resolve the ifindexes that are returned in the MIB data (since the
    zone only has the IPMP interface, not the underlying interfaces).
    Prior to my change, it would just spew "Invalid ifindex <i>" and not
    print the rest of the line.  Now it will print "if#<i>" in this case,
    which is still suboptimal, but better than before.  The real fix for
    this issue will come with the rework of NCEs as part of the IP
    Datapath Refactoring project.

  * The fourth item is just fulfilling a promise I made to the C-Team that
    all our files would be cstyle-clean, even for continuation lines.
    (netstat had many existing violations.)

-- 
meem

Reply via email to