Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 11 Mar 2008 12:15:14 +0100
From:      Vitezslav Novy <vnovy@vnovy.net>
To:        qpadla@gmail.com
Cc:        freebsd-current@freebsd.org
Subject:   Re: IP bad-len 0 ( on em0 )
Message-ID:  <47D669C2.50107@vnovy.net>
In-Reply-To: <200803111135.28626.qpadla@gmail.com>
References:  <200802042142.38606.qpadla@gmail.com> <006801c87f19$a14d8060$b6db87d4@multiplay.co.uk> <47D5A78F.1050403@vnovy.net> <200803111135.28626.qpadla@gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Nikolay Pavlov wrote:
> On Monday 10 March 2008 23:26:39 Vitezslav Novy wrote:
>> Steven Hartland wrote:
>>> Did you ever get anywhere with this? Did Jack respond?
>>>
>>> ----- Original Message ----- From: "Nikolay Pavlov" <qpadla@gmail.com>
>>>
>>> On Wednesday 06 February 2008 19:12:20 Vitezslav Novy wrote:
>>>>>> Do you have set TSO flag on interface?
>>>>> Good catch. The problem is gone with net.inet.tcp.tso set to 0.
>>>>> But TSO is enabled by default in 7.0. Could this be a em driver
>>>>> bug?
>>>> I don't know. I only saw this when researched another TSO related
>>>> problem.
>>>>
>>>> vita
>>> May be Jack could say something about this?
>> It is TSO related problem of em driver.
>> Affects only cards with adapter->hw.mac.type < e1000_82575.
>> During transmit function em_tso_setup is called and this function sets
>> packet ip_len field to zero. (also ip csum)
>>
>> After successful transmit, packet is injected to BPF with zeroed ip_len
>>   field.
> 
> So this is driver or hardware problem? Should i change a network card?
> 
> 

It is driver problem.

If I understand code, driver has 2 transmit routines, one for older 
cards a one for newer. I checked routine for my (older type) card, and 
found problem in em_tso_setup.

But now I see that ip_len is zeroed also in em_tso_adv_setup which is 
used with newer cards.

vita




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?47D669C2.50107>