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

[Heiner Kallweit <hkallweit1@gmail.com>]

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
> 
> 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
>>>
>>
> 
> 

In a similar context Realtek made me aware of a hw issue if IP header
has the options field set. You mentioned problems with multicast packets,
and based on the following code the root cause may be related.

br_ip4_multicast_alloc_query()
-> iph->ihl = 6;

I'd appreciate if you could test (with HW tx checksumming enabled)
whether this experimental patch fixes the issue with invalid/lost
multicasts.


diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index e40e8eaeb..dd251ddb8 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -4319,6 +4319,10 @@ static netdev_features_t rtl8169_features_check(struct sk_buff *skb,
 		    rtl_chip_supports_csum_v2(tp))
 			features &= ~NETIF_F_ALL_TSO;
 	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		if (ip_hdrlen(skb) > sizeof(struct iphdr)) {
+			pr_info("hk: iphdr has options field set\n");
+			features &= ~NETIF_F_CSUM_MASK;
+		}
 		if (skb->len < ETH_ZLEN) {
 			switch (tp->mac_version) {
 			case RTL_GIGA_MAC_VER_11:
-- 
2.26.0