Re: [PATCH v3 net-next 1/4] udp: add busylocks in RX path

2016-12-08 Thread Eric Dumazet
On Thu, 2016-12-08 at 20:45 +0100, Hannes Frederic Sowa wrote: > Hi Eric, > > This patch mostly improves situation for non-connected sockets. Do you > think it makes sense to acquire the spinlock depending on the sockets > state? Connected UDP sockets flow in on one CPU anyway? We could do that,

Re: [PATCH v3 net-next 1/4] udp: add busylocks in RX path

2016-12-08 Thread Hannes Frederic Sowa
Hi Eric, On Thu, Dec 8, 2016, at 20:41, Eric Dumazet wrote: > Idea of busylocks is to let producers grab an extra spinlock > to relieve pressure on the receive_queue spinlock shared by consumer. > > This behavior is requested only once socket receive queue is above > half occupancy. > > Under fl

[PATCH v3 net-next 1/4] udp: add busylocks in RX path

2016-12-08 Thread Eric Dumazet
Idea of busylocks is to let producers grab an extra spinlock to relieve pressure on the receive_queue spinlock shared by consumer. This behavior is requested only once socket receive queue is above half occupancy. Under flood, this means that only one producer can be in line trying to acquire the