Hi Eric,

Eric Dumazet wrote:

Just curious.

Blocking hard IRQ while doing TX completion is quite nasty,
as freeing one hundred of skb take long.

Could you try something in process context instead ?

I am planning some changes for handling TX completions, but I think it is too
late to
add them to 2.6.30.
In the code that will be submitted to 2.6.31 we will not block irqs.

Is this timer driven tx completion thing really good for performance ?

When we are CPU bounded on the transmit side, the driver shown better
performance with
the timer driven completions rather then with interrupts.