Re: [PATCH v2 net-next] mlx4: optimize xmit path

[Amir Vadai <amirv@mellanox.com>]

On 10/2/2014 7:35 AM, Eric Dumazet wrote:
> On Sun, 2014-09-28 at 21:19 -0700, Eric Dumazet wrote:
>> From: Eric Dumazet <edumazet@google.com>
> 

[...]

> Sorry, there is a missing replacement of 
> 
> 	iowrite32be(ring->doorbell_qpn,
> 		    ring->bf.uar->map + MLX4_SEND_DOORBELL);
> 
> by iowrite32(ring->doorbell_qpn,
> 	     ring->bf.uar->map + MLX4_SEND_DOORBELL);
> 
> Since doorbel_qpn was changed to a __be32 and setup in
> mlx4_en_activate_tx_ring()
> 

Hi,

Will take it into the split patchset - we just hit this bug when tried
to run benchmarks with blueflame disabled (easy to test by using ethtool
priv flag blueflame).

I'm still working on it, but I can't reproduce the numbers that you
show. On my development machine, I get ~5.5Mpps with burst=8 and ~2Mpps
with burst=1.

In addition, I see no improvements when adding the optimization to the
xmit path.
I use the net-next kernel + pktgen burst support patch, with and without
this xmit path optimization patch.

Do you use other patches not upstream in your environment?
Can you share the .config/pktgen configuration?

One other note: we're checking now that blueflame could be used with
xmit_more. It might result with packets reordering/drops. Still under
investigation.

Thanks,
Amir