Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 06 Apr 2017 11:26:23 +0200
From:      Harry Schmalzbauer <freebsd@omnilan.de>
To:        Sean Bruno <sbruno@freebsd.org>
Cc:        "freebsd-net@freebsd.org" <freebsd-net@freebsd.org>
Subject:   Re: if_igb(4) VLAN(4) and [RT]XCSUM_IPV6, TSO6
Message-ID:  <58E609BF.7040205@omnilan.de>
In-Reply-To: <939d3920-0181-8b1a-e0be-5d612286daee@freebsd.org>
References:  <58CAD8CB.3060101@omnilan.de> <939d3920-0181-8b1a-e0be-5d612286daee@freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Bezüglich Sean Bruno's Nachricht vom 05.04.2017 23:54 (localtime):
> 
> 
> On 03/16/17 12:26, Harry Schmalzbauer wrote:
>>  Hello,
>>
>> I'm wondering if I really loose [RT]XCSUM_IPV6 on if_igb(4) vlan(4)
>> children.
>> My igb0 (Kawela, aka 82576) options end with
>>   "TSO6,VLAN_HWTSO,RXCSUM_IPV6,TXCSUM_IPV6"
>>
>> The vlan(4) filtered interfaces show these:
>>   options=303<RXCSUM,TXCSUM,TSO4,TSO6>
>>
>> So TSO6 is inherited, but RC/TXCSUM_IPV6 dropped?
…
>>
> 
> That looks like a bug.  Which version of FreeBSD is this?

Thanks for your attention.

I'm on RELENG_11. I can confirm that behaviour on r315416 (Mar, 16th).


>> In general, how is offloading implemented? If enabled, does the API
>> simply generate packets lacking the checksum header and NIC (silicon)
>> fills in the checksum if not adviesd not to do so (by the driver)?
>> Or is the driver utilizing NIC registers for calculation, but ensures
>> every packet will have a checksum?

If you know any beginners-level explanations, I was still interested in
getting knowledge about offloading implementation.  Most important, is
it driver/nic dependent or do all drivers implement it similar?

Thanks,

-harry



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?58E609BF.7040205>