On Mon, Jun 8, 2015 at 7:42 PM, Pravin Shelar <pshe...@nicira.com> wrote:
> On Mon, Jun 8, 2015 at 6:13 PM, Wei li <l...@dtdream.com> wrote:
>> When tx queue is shared among CPUS,the pkts always be flush in
>> 'netdev_dpdk_eth_send'
>> So it is unnecessarily for flushing in netdev_dpdk_rxq_recv
>> Otherwise tx will be accessed without locking
>>
>> Signed-off-by: Wei li <l...@dtdream.com>
>> ---
>> v1->v2: fix typo
>>
>> lib/netdev-dpdk.c | 7 +++++--
>> 1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
>> index 63243d8..1e0a483 100644
>> --- a/lib/netdev-dpdk.c
>> +++ b/lib/netdev-dpdk.c
>> @@ -892,8 +892,11 @@ netdev_dpdk_rxq_recv(struct netdev_rxq *rxq_, struct
>> dp_packet **packets,
>> int nb_rx;
>>
>> /* There is only one tx queue for this core. Do not flush other
>> - * queueus. */
>> - if (rxq_->queue_id == rte_lcore_id()) {
>> + * queues.
>> + * Do not flush tx queue which is shared among CPUs
>> + * since it is always flushed */
>> + if (rxq_->queue_id == rte_lcore_id() &&
>> + OVS_LIKELY(!dev->txq_needs_locking)) {
>> dpdk_queue_flush(dev, rxq_->queue_id);
>> }
>
> Patch looks good, But Daniele has plan to get rid of queue flushing
> logic. do you see problem with doing this?
Daniele,
I am not sure if we should fix this queue flushing logic if we are
going to remove it. So I would like to discuss it first.
You mentioned that the removing flushing logic results in performance
drop. Can you explain it? How much is performance drop and which is
the case?
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
