On 15.5.2023. 19:39, Alexander Bluhm wrote:
> On Sun, May 14, 2023 at 11:39:01PM +0200, Hrvoje Popovski wrote:
>> I've tested this on openbsd box with 4 iperf3's. 2 for ip4 and 2 for ip6
>> and with 16 tcp streams per iperf. When testing over ix(4) there is big
>> differences in output performance. When testing ix/veb/vport there is
>> differences in output performance but not that big.
> Thanks a lot for testing. I have also created some numbers which
> can be seen here.
>
> http://bluhm.genua.de/perform/results/2023-05-14T09:14:59Z/perform.html
>
> Sending TCP to Linux host and socket splicing gets faster.
>
>> When testing over vport I'm getting "software chopped" which should be
>> expected.
> Yes, we cannot do hardware TSO in a bridge. Maybe we could if all
> bridge members support it.
>
> Next diff that should go in is where jan@ renames flags, cleans up
> ifconfig(8),. and fixes pseudo interface devices
>
> Updated ix(4) driver diff after TCP/IP commit is below.
Hi,
I've tested this diff with x552 and it's working as expected.
ix0 at pci5 dev 0 function 0 "Intel X552 SFP+" rev 0x00, msix, 4 queues,
ix1 at pci5 dev 0 function 1 "Intel X552 SFP+" rev 0x00, msix, 4 queues,
netstat -sp tcp | grep TSO
46 output TSO packets software chopped
772398947 output TSO packets hardware processed
4005484521 output TSO packets generated
0 output TSO packets dropped
