From: Jan Klemkow <jan@openbsd.org>
Subject: Re: iavf(4): add checksum offloading
To: Yuichiro NAITO <naito.yuichiro@gmail.com>
Cc: tech@openbsd.org
Date: Wed, 6 Nov 2024 10:13:59 +0100

On Wed, Nov 06, 2024 at 05:51:09PM GMT, Yuichiro NAITO wrote:
> From: Jan Klemkow <jan@openbsd.org>
> Subject: iavf(4): add checksum offloading
> Date: Thu, 31 Oct 2024 21:27:07 +0100
> 
> > This diff add checksum offloading support for iavf(4).  Its implemented
> > similar to ixl(4).  iavf(4) is just the virtual functions driver for
> > ixl(4) interfaces.
> > 
> > I tested it on Linux/KVM with Intel x710 interfaces.
> > 
> > More tests on different environments and hardware are welcome!
> 
> I tested your patch works for me. My test environment is as same as yours.
> However, I can't test your patch on ESXi. I see the following message
> on the kernel boot. The `iavf0` interface doesn't receive any packets.
> 
> ```
> iavf0 at pci3 dev 0 function 0 "Intel XL710/X710 VF" rev 0x01, VF version 1.1, V
> F 0 VSI 10config irq map failed: -5
> , msix, address 00:0c:29:a2:be:06
> ```
> 
> This is because iavf(4) doesn't support multi-queues. Could you see
> my mail before?
> 
> https://marc.info/?l=openbsd-tech&m=172543816813802&w=2
> 
> My patch also supports check sum offloading, so it conflicts with your
> patch. I don't want to say which is better. They are almost same. I don't
> care my patch is committed or not. I just want to use iavf(4) on ESXi. 

Oh, sorry.  I didn't noticed your diff on the list.  While trying to get
ixv(4) working I used iavf(4) as a reference and made a fast diff w/o
checking the mailing list.

Next week, I'm back at my equipment and will try to put an x710 in my
ESXi server.  So, I can test your diff there.

Thanks,
Jan