[ovs-git] [openvswitch/ovs] 6094fc: ovn-controller: Fix memory leak and bug in setup_q...

GitHub noreply at github.com
Wed Feb 1 19:17:16 UTC 2017


  Branch: refs/heads/master
  Home:   https://github.com/openvswitch/ovs
  Commit: 6094fc7f239ae3dce4debebcce9911374c85748a
      https://github.com/openvswitch/ovs/commit/6094fc7f239ae3dce4debebcce9911374c85748a
  Author: Ben Pfaff <blp at ovn.org>
  Date:   2017-02-01 (Wed, 01 Feb 2017)

  Changed paths:
    M ovn/controller/binding.c

  Log Message:
  -----------
  ovn-controller: Fix memory leak and bug in setup_qos().

The caller of netdev_get_qos() is responsible for freeing its 'details'
smap.

Also, the previous qdisc details, for some arbitrary previous type of QoS,
aren't relevant for changing the type of QoS.

Signed-off-by: Ben Pfaff <blp at ovn.org>


  Commit: dc2dab6e6de50a6b6ef25eb2cf761fae3a3a33a7
      https://github.com/openvswitch/ovs/commit/dc2dab6e6de50a6b6ef25eb2cf761fae3a3a33a7
  Author: Ben Pfaff <blp at ovn.org>
  Date:   2017-02-01 (Wed, 01 Feb 2017)

  Changed paths:
    M ovn/controller/binding.c

  Log Message:
  -----------
  ovn-controller: Configure interface QoS only if it would actually be used.

Until now, ovn-controller has unconditionally configured linux-htb on
physical interfaces.  QoS is pretty much always trouble, but it's even more
trouble if we set it up for no good reason.  We received a bug report, in
particular, that doing this disrupts connectivity in Docker.

This commit attempts to make that less likely, by making ovn-controller
only configure a qdisc if QoS support has in turn been configured in OVN.
The same problems as before will recur if QoS support is actually
configured, but at least now there's some purpose, and possibly a symptom
that the user can better diagnose ("I turned on QoS and OVN stopped
working" is at least a cause-and-effect chain that makes some sense).

Reported-by: Ritesh Rekhi <ritesh.rekhi at nutanix.com>
Reported-by: Hexin Wang <hexin.wang at nutanix.com>
Reported-at: https://mail.openvswitch.org/pipermail/ovs-discuss/2017-February/043564.html
Tested-by: Hexin Wang <hexin.wang at nutanix.com>
Tested-at: https://mail.openvswitch.org/pipermail/ovs-discuss/2017-February/043575.html
Signed-off-by: Ben Pfaff <blp at ovn.org>


Compare: https://github.com/openvswitch/ovs/compare/e2218eeebc4a...dc2dab6e6de5


More information about the git mailing list