linux-kernel - Re: [PATCH 2.6.36] vlan: Avoid hwaccel vlan packets when vid not used

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <AANLkTincX-wPQE9gs16v53E4=Skivzc4ZSJS43oNBLD_@mail.gmail.com>
Date:	Thu, 6 Jan 2011 16:01:59 -0500
From:	Jesse Gross <jesse@...ira.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	Matt Carlson <mcarlson@...adcom.com>,
	Michael Leun <lkml20101129@...ton.leun.net>,
	Michael Chan <mchan@...adcom.com>,
	David Miller <davem@...emloft.net>,
	Ben Greear <greearb@...delatech.com>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH 2.6.36] vlan: Avoid hwaccel vlan packets when vid not used

On Sun, Jan 2, 2011 at 11:05 AM, Eric Dumazet <eric.dumazet@...il.com> wrote:
> Le samedi 01 janvier 2011 à 19:27 -0500, Jesse Gross a écrit :
>> On Sat, Jan 1, 2011 at 12:03 PM, Eric Dumazet <eric.dumazet@...il.com> wrote:
>> > Le mardi 14 décembre 2010 à 11:15 -0800, Matt Carlson a écrit :
>> >
>> >> Thanks for the comments Jesse.  Below is an updated patch.
>> >>
>> >> Michael, I'm wondering if the difference in behavior can be explained by
>> >> the presence or absence of management firmware.  Can you look at the
>> >> driver sign-on messages in your syslogs for ASF[]?  I'm half expecting
>> >> the 5752 to show "ASF[0]" and the 5714 to show "ASF[1]".  If you see
>> >> this, and the below patch doesn't fix the problem, let me know.  I have
>> >> another test I'd like you to run.
>> >>
>> >> ----
>> >>
>> >> [PATCH] tg3: Use new VLAN code
>> >>
>> >> This patch pivots the tg3 driver to the new VLAN infrastructure.
>> >> All references to vlgrp have been removed and all VLAN code is
>> >> unconditionally active.
>> >>
>> >> Signed-off-by: Matt Carlson <mcarlson@...adcom.com>
>>
>> [...]
>>
>> > Hi Matt.
>> >
>> > Any news on this patch ?
>> >
>> > Without it, net-next-2.6 doesnt work for me on a vlan setup on top of
>> > bonding.
>> >
>> > (bond0 : eth1 & eth2, eth1 being bnx2, eth2 beging tg3)
>> >
>> > ip link add link bond0 vlan.103 type vlan id 103
>> > ip addr add 192.168.20.110/24 dev vlan.103
>> > ip link set vlan.103 up
>> >
>> >
>> > If active slave is eth1 (bnx2), everything works, but if active slave is
>> > eth2 (tg3), incoming tagged frames (on vlan 103) are lost.
>>
>> This patch isn't quite right - it always disables vlan stripping
>> unless management firmware is in use, so it's not really a correct
>> fix.
>>
>> You said that this used to work correctly on this NIC?  Does it work
>> without a bond, just a vlan on the tg3 device?  It sounds like Michael
>> has a problem with vlan stripping on one of his NICs but if it works
>> with just a vlan or on older kernels, it's probably not the same
>> thing.
>>
>
> 1) current linux-2.6 works OK for me (and previous versions as well, I
> am using this vlan/bonding setup since 3 years or so on one of my dev
> machine)
>
> Only net-next-2.6 has the problem.
>
> If I remove bonding of the equation, I still have the problem, and can
> see the 'dropped' counter increasing while I send packets to eth2 (tg3)
>
> $ ifconfig eth2
> eth2      Link encap:Ethernet  HWaddr 00:1E:0B:92:78:50
>          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
>          RX packets:94 errors:0 dropped:38686 overruns:0 frame:0
>          TX packets:18 errors:0 dropped:0 overruns:0 carrier:0
>          collisions:0 txqueuelen:1000
>          RX bytes:8332 (8.1 Kb)  TX bytes:1392 (1.3 Kb)
>          Interrupt:19
> $ ifconfig vlan.103
> vlan.103  Link encap:Ethernet  HWaddr 00:1E:0B:92:78:50
>          inet addr:192.168.20.110  Bcast:0.0.0.0  Mask:255.255.255.0
>          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
>          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
>          TX packets:15 errors:0 dropped:0 overruns:0 carrier:0
>          collisions:0 txqueuelen:0
>          RX bytes:0 (0.0 b)  TX bytes:846 (846.0 b)

Hmm, I thought that it might be some interaction with a corner case in
the networking core but now it seems less likely.  There weren't too
many vlan changes between the working and non-working states.  Plus,
since the rx counter isn't increasing, the packets probably aren't
making it anywhere.

I see that tg3 increases the drop counter in one place, which also
happens to be checking for vlan errors (at tg3.c:4753).  That seems
suspicious - maybe the NIC is only partially configured for vlan
offloading.  If we can confirm that is where the drop counter is being
incremented and what the error code is maybe it would shed some light.

If it's a driver issue I don't have much insight - maybe Matt or
bisect can help.

>> If it works on bnx2, it would seem to be a driver problem but it would
>> be good to confirm that the tag in skb->vlan_tci is not being
>> delievered to the networking core in this case.
>
> Hmm, where do you want me to check this ?

I was thinking right before vlan_gro_receive() at tg3.c:4837.  If my
theory above is right then this obviously isn't relevant since it
won't be hit at all.  Otherwise it would be good to know exactly what
the driver is producing.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/