Note: First patch of this series may be applied separately
to fix existing problem with mapping of TX queues.
version 8:
* Nobody wants to review huge architectural patches.
So, I divided it into two. First patch fixes original
problem with unexpected mapping of TX queues.
Second patch adds new infrastructure for TX queue
management. Details are in commit messages.
* Fixed calculation of global_index on port deletion.
version 7:
* rebased on current master
* My comment for v6 that there is a race for total number of
threads was wrong. So, patch actually the same as v6.
version 6:
* fixed assigning same txq to multiple pmd threads on different
NUMA sockets.
* Changed logic of txq distribution in pmd_load_queues().
version 5:
* txqs 0 from ports of non-pmd netdevs added to all pmd threads
version 4:
* fixed distribution of tx queues if multiqueue is not supported
version 3:
* fixed failing of unit tests by adding tx queues of non
pmd devices to non pmd thread. (they haven't been used by any thread)
* pmd_flush_tx_queues --> dp_netdev_pmd_detach_tx_queues
* function names changed to dp_netdev_*
* dp_netdev_pmd_lookup_txq now looks by port_no.
* removed unnecessary dp_netdev_lookup_port in dp_execute_cb
for OVS_ACTION_ATTR_OUTPUT.
* refactoring
Ilya Maximets (2):
dpif-netdev: unique and sequential tx_qids.
dpif-netdev: Add per pmd cmap of available tx queues.
lib/dpif-netdev.c | 207 +++++++++++++++++++++++++++++++++++++++++++++++-------
1 file changed, 182 insertions(+), 25 deletions(-)
--
2.1.4
_______________________________________________
dev mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/dev
