Re: [PATCH net] r8169: fix multicast tx issue with macvlan interface

[Heiner Kallweit <hkallweit1@gmail.com>]

On 15.04.2020 13:55, Charles Daymand wrote:
> Sorry apparently I had an issue when sending the mail to the netdev mailing list yesterday.
> Actually in our architecture we are creating vlan interface from a
> macvlan interface (that explains the 802.1Q traffic)
> I did more testing on the issue today and I observed that the macvlan
> interface itself is not affected by the issue, it's only the vlan
> interfaces created from the macvlan interface which is affected by the
> issue.
> 
> I did the following network configuration on the hardware to simplify
> the testing :
> 
>    +----------------------------------------+                 +--------------------------+
>    | Hardware                               |                 | Laptop                   |
>    | +---------+                            |                 |            +---------+   |
>    | | vlan2833+-+                          |                 |      +-----+ vlan2833|   |
>    | +---------+ |                          |                 |      |     +---------+   |
> 192.168.200.1/24 |  +---------+          +--+-+             +-+--+   |  192.168.200.2/24 |
>    |             +--+mymacvlan+----------+eth0+-------------+eth0+---+--+                |
>    |             |  +---------+          +----+             +----+      |                |
>    | +---------+ | 192.168.120.1/24 192.168.100.1/24   192.168.100.2/24 |  +---------+   |
>    | | vlan1000+-+                          |          192.168.120.2/24 +--+ vlan1000|   |
>    | +---------+                            |                 |            +---------+   |
> 192.168.150.1/24                            |                 |         192.168.150.2/24 |
>    +----------------------------------------+                 +--------------------------+
> 
> For testing purpose I connected by ssh to 192.168.100.1 and sent udp
> packets to the laptop IPs with the following result:
> * When sending an UDP packet to 192.168.120.2 ==> No issue
> * When sending an UDP packet to 192.168.200.2 (vlan 2833) ==> I have the malformed packet issue
> * When sending an UDP packet to 192.168.150.2 (vlan 1000) ==> I receive no network traffic
> and I lose the ssh connection during 2 minutes with nothing in dmesg during this period.
> 
> Le 13/04/2020 à 20:20, Eric Dumazet a écrit :
>> On Mon, Apr 13, 2020 at 11:06 AM Heiner Kallweit <hkallweit1@gmail.com> wrote:
>>> On 06.04.2020 16:12, Charles DAYMAND wrote:
>>>> Hello,
>>>> Sorry for the long delay, I didn't have physical access to the
>>>> hardware last week. I did more testing today by connecting directly my
>>>> laptop to the hardware and the issue not only affects multicast but
>>>> also unicast with macvlan.
>>>> Only ping packets are correctly sent.
>>>> Using wireshark I can only see malformed ethernet packets (example
>>>> with multicast packet below)
>>>> Frame 1: 106 bytes on wire (848 bits), 106 bytes captured (848 bits)
>>>> IEEE 802.3 Ethernet
>>>>      Destination: IPv4mcast_09 (01:00:5e:00:00:09)
>>>>      Source: b2:41:6f:04:c1:86 (b2:41:6f:04:c1:86)
>>>>      Length: 96
>>>>          [Expert Info (Error/Malformed): Length field value goes past
>>>> the end of the payload]
>>>> Logical-Link Control
>>>>      DSAP: Unknown (0x45)
>>>>      SSAP: Unknown (0xc0)
>>>>      Control field: I, N(R)=46, N(S)=0 (0x5C00)
>>>> Data (88 bytes)
>>>>      Data: 50320c780111087fac159401e0000009020802080048207a...
>>>>      [Length: 88]
>>>> Please find the full pcap file at this url with also a tentative to
>>>> establish a ssh connection :
>>>> https://fourcot.fr/temp/malformed_ethernet2.pcap
>>>>
>>> A problem with HW checksumming should manifest as checksum failure,
>>> I don't think HW checksumming touches length information in headers.
>>> You could enable tracing network events to see whether packets look
>>> suspicious before reaching the network driver.
>>> (or do some printf debugging for the affected multicast packets)
>>>
>>> Because you initially suspected a problem with HW checksumming
>>> in a specific RTL8168 chip variant:
>>> - Check macvlan with another network driver supporting HW checksumming
>>> - Check macvlan with r8169 and another RTL8168 chip variant
>>>
>>> The fact that the issue doesn't exist w/o macvlan seems to indicate
>>> that the network driver can't be blamed here.
>> trace includes evidence of 802.1Q traffic though, maybe "macvlan" is a
>> red herring,
>> some confusion on wording maybe...
>>
>> 802.1Q Virtual LAN, PRI: 0, DEI: 0, ID: 2833
>>      000. .... .... .... = Priority: Best Effort (default) (0)
>>      ...0 .... .... .... = DEI: Ineligible
>>      .... 1011 0001 0001 = ID: 2833
>>
>>
>>>> Best regards
>>>>
>>>> Charles
>>>>
>>>>
>>>> Le mar. 31 mars 2020 à 16:07, Heiner Kallweit <hkallweit1@gmail.com> a écrit :
>>>>> Thanks for further testing! The good news from my perspective is that the issue doesn't occur
>>>>> w/o macvlen, therefore it doesn't seem to be a r8169 network driver issue.
>>>>>
>>>>> W/o knowing tcpdump in detail I think it switches the NIC to promiscuous mode, means
>>>>> it should see all packets, incl. the ones with checksum errors.
>>>>> Maybe you can mirror the port to which the problematic system is connected and
>>>>> analyze the traffic. Or for whatever reason the switch doesn't forward the multicast
>>>>> packets to your notebook.
>>>>>
>>>>> Heiner
>>>>>
>>>>>
>>>>> On 31.03.2020 15:44, Charles DAYMAND wrote:
>>>>>> Hello,
>>>>>> We tested to enable tx checksumming manually (via ethtool) on a kernel 4.19.0-5-amd64 which is the oldest kernel compatible with our software and we observed exactly the same issue.
>>>>>> For information when connecting a laptop directly to the interface we can't see any multicast packet when tx checksumming is enabled on tcpdump.
>>>>>> Our network is composed of a cisco switch and we can still see the multicast counters correctly increasing even when we have the issue.
>>>>>>
>>>>>> I also confirm that when not using macvlan but the real interface there is no issue on the multicast packets, they are correctly sent and received.
>>>>>> I have a stupid question, if the IP checksum was bad on the multicast packet, would the receiver NIC drop the packet or would it be seen by tcpdump by the receiver ?
>>>>>>
>>>>>> Le ven. 27 mars 2020 à 20:43, Eric Dumazet <edumazet@google.com <mailto:edumazet@google.com>> a écrit :
>>>>>>
>>>>>>      On Fri, Mar 27, 2020 at 12:17 PM Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> wrote:
>>>>>>      >
>>>>>>      > On 27.03.2020 19:52, Eric Dumazet wrote:
>>>>>>      > > On Fri, Mar 27, 2020 at 10:41 AM Heiner Kallweit <hkallweit1@gmail.com <mailto:hkallweit1@gmail.com>> wrote:
>>>>>>      > >>
>>>>>>      > >> On 27.03.2020 10:39, Heiner Kallweit wrote:
>>>>>>      > >>> On 27.03.2020 10:08, Charles Daymand wrote:
>>>>>>      > >>>> During kernel upgrade testing on our hardware, we found that macvlan
>>>>>>      > >>>> interface were no longer able to send valid multicast packet.
>>>>>>      > >>>>
>>>>>>      > >>>> tcpdump run on our hardware was correctly showing our multicast
>>>>>>      > >>>> packet but when connecting a laptop to our hardware we didn't see any
>>>>>>      > >>>> packets.
>>>>>>      > >>>>
>>>>>>      > >>>> Bisecting turned up commit 93681cd7d94f
>>>>>>      > >>>> "r8169: enable HW csum and TSO" activates the feature NETIF_F_IP_CSUM
>>>>>>      > >>>> which is responsible for the drop of packet in case of macvlan
>>>>>>      > >>>> interface. Note that revision RTL_GIGA_MAC_VER_34 was already a specific
>>>>>>      > >>>> case since TSO was keep disabled.
>>>>>>      > >>>>
>>>>>>      > >>>> Deactivating NETIF_F_IP_CSUM using ethtool is correcting our multicast
>>>>>>      > >>>> issue, but we believe that this hardware issue is important enough to
>>>>>>      > >>>> keep tx checksum off by default on this revision.
>>>>>>      > >>>>
>>>>>>      > >>>> The change is deactivating the default value of NETIF_F_IP_CSUM for this
>>>>>>      > >>>> specific revision.
>>>>>>      > >>>>
>>>>>>      > >>>
>>>>>>      > >>> The referenced commit may not be the root cause but just reveal another
>>>>>>      > >>> issue that has been existing before. Root cause may be in the net core
>>>>>>      > >>> or somewhere else. Did you check with other RTL8168 versions to verify
>>>>>>      > >>> that it's indeed a HW issue with this specific chip version?
>>>>>>      > >>>
>>>>>>      > >>> What you could do: Enable tx checksumming manually (via ethtool) on
>>>>>>      > >>> older kernel versions and check whether they are fine or not.
>>>>>>      > >>> If an older version is fine, then you can start a new bisect with tx
>>>>>>      > >>> checksumming enabled.
>>>>>>      > >>>
>>>>>>      > >>> And did you capture and analyze traffic to verify that actually the
>>>>>>      > >>> checksum is incorrect (and packets discarded therefore on receiving end)?
>>>>>>      > >>>
>>>>>>      > >>>
>>>>>>      > >>>> Fixes: 93681cd7d94f ("r8169: enable HW csum and TSO")
>>>>>>      > >>>> Signed-off-by: Charles Daymand <charles.daymand@wifirst.fr <mailto:charles.daymand@wifirst.fr>>
>>>>>>      > >>>> ---
>>>>>>      > >>>>  net/drivers/net/ethernet/realtek/r8169_main.c | 3 +++
>>>>>>      > >>>>  1 file changed, 3 insertions(+)
>>>>>>      > >>>>
>>>>>>      > >>>> diff --git a/net/drivers/net/ethernet/realtek/r8169_main.c b/net/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>      > >>>> index a9bdafd15a35..3b69135fc500 100644
>>>>>>      > >>>> --- a/net/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>      > >>>> +++ b/net/drivers/net/ethernet/realtek/r8169_main.c
>>>>>>      > >>>> @@ -5591,6 +5591,9 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>>>>>>      > >>>>              dev->vlan_features &= ~(NETIF_F_ALL_TSO | NETIF_F_SG);
>>>>>>      > >>>>              dev->hw_features &= ~(NETIF_F_ALL_TSO | NETIF_F_SG);
>>>>>>      > >>>>              dev->features &= ~(NETIF_F_ALL_TSO | NETIF_F_SG);
>>>>>>      > >>>> +            if (tp->mac_version == RTL_GIGA_MAC_VER_34) {
>>>>>>      > >>>> +                    dev->features &= ~NETIF_F_IP_CSUM;
>>>>>>      > >>>> +            }
>>>>>>      > >>>>      }
>>>>>>      > >>>>
>>>>>>      > >>>>      dev->hw_features |= NETIF_F_RXALL;
>>>>>>      > >>>>
>>>>>>      > >>>
>>>>>>      > >>
>>>>>>      > >> After looking a little bit at the macvlen code I think there might be an
>>>>>>      > >> issue in it, but I'm not sure, therefore let me add Eric (as macvlen doesn't
>>>>>>      > >> seem to have a dedicated maintainer).
>>>>>>      > >>
>>>>>>      > >> r8169 implements a ndo_features_check callback that disables tx checksumming
>>>>>>      > >> for the chip version in question and small packets (due to a HW issue).
>>>>>>      > >> macvlen uses passthru_features_check() as ndo_features_check callback, this
>>>>>>      > >> seems to indicate to me that the ndo_features_check callback of lowerdev is
>>>>>>      > >> ignored. This could explain the issue you see.
>>>>>>      > >>
>>>>>>      > >
>>>>>>      > > macvlan_queue_xmit() calls dev_queue_xmit_accel() after switching skb->dev,
>>>>>>      > > so the second __dev_queue_xmit() should eventually call the real_dev
>>>>>>      > > ndo_features_check()
>>>>>>      > >
>>>>>>      > Thanks, Eric. There's a second path in macvlan_queue_xmit() calling
>>>>>>      > dev_forward_skb(vlan->lowerdev, skb). Does what you said apply also there?
>>>>>>
>>>>>>      This path wont send packets to the physical NIC, packets are injected
>>>>>>      back via dev_forward_skb()
>>>>>>
>>>>>>      >
>>>>>>      > Still I find it strange that a tx hw checksumming issue should affect multicasts
>>>>>>      > only. Also the chip version in question is quite common and I would expect
>>>>>>      > others to have hit the same issue.
>>>>>>      > Maybe best would be to re-test on the affected system w/o involving macvlen.
>>>>>>      >
>>>>>>      > >
>>>>>>      > >
>>>>>>      > >> Would be interesting to see whether it fixes your issue if you let the
>>>>>>      > >> macvlen ndo_features_check call lowerdev's ndo_features_check. Can you try this?
>>>>>>      > >>
>>>>>>      > >> By the way:
>>>>>>      > >> Also the ndo_fix_features callback of lowerdev seems to be ignored.
>>>>>>      >
>>>>>>
>>>>>>
>>>>>>
>>>>>> -- 
>>>>>>
>>>>>> logo wifirst <http://www.wifirst.fr/en>
>>>>>>
>>>>>> Charles Daymand
>>>>>>
>>>>>> Développeur infrastructure
>>>>>>
>>>>>> 26 rue de Berri 75008 Paris
>>>>>>
>>>>>> Assistance dédiée responsable de site - 01 70 70 46 70
>>>>>> Assistance utilisateur - 01 70 70 46 26
>>>>>>
>>>>

I wonder whether the issue reported here could be related to the one you're facing.
It's also about multicast and macvlan.

https://marc.info/?l=linux-netdev&m=158929343517262&w=2