Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 03/12/2011 10:59 PM, Daniel Lezcano wrote: > > Hi Patrick, > > I noticed another problem with the macvlan driver. > > In the function dev_forward_skb the test always succeed in the second > condition making the packet to be dropped. > > ... > > if (unlikely(!(dev->flags& IFF_UP) || >(skb->len> (dev->mtu + dev->hard_header_len + VLAN_HLEN { > > ... > > When tracing I have the following values: > > skb->len = 2962 > and > dev->mtu + dev->hard_header_len + VLAN_HLEN = 1518 > > Do you have any idea where that could be come from ? Disabling TSO solves the problem. Is it possible the packets are not fragmented to the mtu size as it is supposed to be done by the lower device but as we are in bridge mode, we dev_forward_skb and the check is expecting the packet to be mtu-compliant ? > On 03/08/2011 03:41 PM, Patrick McHardy wrote: >> Am 02.03.2011 19:33, schrieb Daniel Lezcano: >>> On 03/02/2011 07:03 PM, Patrick McHardy wrote: Am 02.03.2011 17:03, schrieb Daniel Lezcano: > On 03/02/2011 12:03 PM, Patrick McHardy wrote: >> Am 01.03.2011 21:04, schrieb Daniel Lezcano: >>> On 03/01/2011 05:51 PM, Patrick McHardy wrote: > Patrick, do you have any suggestions to fix this ? Since the frames are only looped back locally, I suppose the easiest fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively we need to complete the checksum manually, similar to what dev_hard_start_xmit() does. >>> That sounds very simple to fix, maybe too much simple :) >>> >>> I did the following change: >>> >>> --- linux-next.orig/drivers/net/macvlan.c >>> +++ linux-next/drivers/net/macvlan.c >>> @@ -222,6 +222,7 @@ static int macvlan_queue_xmit(struct sk_ >>> >>> if (vlan->mode == MACVLAN_MODE_BRIDGE) { >>> const struct ethhdr *eth = (void *)skb->data; >>> + skb->ip_summed = CHECKSUM_UNNECESSARY; >>> >>> /* send to other bridge ports directly */ >>> if (is_multicast_ether_addr(eth->h_dest)) { >>> >>> >>> and that fixed the problem. Do you think it is acceptable ? >> The only problem I see is if the packets are bridged to a >> different networking device (or redirected using the mirred >> action), in this case the checksum will not be completed. >> This would be a very strange setup though and probably wouldn't >> be using dummy as lower device, so I'm not sure we have to >> worry about this case. > I am not sure to get it, do you say the patch is correct ? Its correct with a short-coming that doesn't seem to matter. > If my understanding is correct, the packet will be flagged > CHECKSUM_UNNECESSARY only for the macvlan devices, right ? Only for packets bridged between macvlan devices. A setup like the following would cause problems: br0 | .--. | | macvlan0macvlan1eth0 | | ---.--- dummy0 In this case packets sent from macvlan0 will show up on eth0 with incorrect setups. However this setup doesn't seem realistic to me, you would simply use eth0 instead of dummy0. >>> Ok, I understand. thanks for the clarification. >>> > By the way, this problem occurs for any lower device with offloading > capabilities with a macvlan port in bridge mode. True. This doesn't affect outgoing packets since their checksum will be completed in dev_hard_start_xmit(), but it affects packets bridged between macvlans. >>> One last question. In the case of broadcast packets with maclvan in >>> bridge mode. >>> We will have the packets going through each macvlan port and also to the >>> lower-device, right ? >>> For the latter, don't we have a problem if the packet is flagged >>> CHECKSUM_UNNECESSARY ? >>> >>> Shouldn't we restore the ip_summed field before sending through >>> dev_queue_xmit ? >> Yes, that seems correct in order to have dev_hard_start_xmit() complete >> the checksum if necessary. > > > > -- > Colocation vs. Managed Hosting > A question and answer guide to determining the best fit > for your organization - today and in the future. > http://p.sf.net/sfu/internap-sfd2d > ___ > Lxc-users mailing list > Lxc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/lxc-users -- Colocation vs. Managed Hosting A question and answer guide to determining the best fit for your organization - today and in the future. http://p.sf.net/sfu/internap-sfd2d _
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
Hi Patrick, I noticed another problem with the macvlan driver. In the function dev_forward_skb the test always succeed in the second condition making the packet to be dropped. ... if (unlikely(!(dev->flags & IFF_UP) || (skb->len > (dev->mtu + dev->hard_header_len + VLAN_HLEN { ... When tracing I have the following values: skb->len = 2962 and dev->mtu + dev->hard_header_len + VLAN_HLEN = 1518 Do you have any idea where that could be come from ? Thanks -- Daniel On 03/08/2011 03:41 PM, Patrick McHardy wrote: > Am 02.03.2011 19:33, schrieb Daniel Lezcano: >> On 03/02/2011 07:03 PM, Patrick McHardy wrote: >>> Am 02.03.2011 17:03, schrieb Daniel Lezcano: On 03/02/2011 12:03 PM, Patrick McHardy wrote: > Am 01.03.2011 21:04, schrieb Daniel Lezcano: >> On 03/01/2011 05:51 PM, Patrick McHardy wrote: Patrick, do you have any suggestions to fix this ? >>> Since the frames are only looped back locally, I suppose the easiest >>> fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively >>> we need to complete the checksum manually, similar to what >>> dev_hard_start_xmit() does. >> That sounds very simple to fix, maybe too much simple :) >> >> I did the following change: >> >> --- linux-next.orig/drivers/net/macvlan.c >> +++ linux-next/drivers/net/macvlan.c >> @@ -222,6 +222,7 @@ static int macvlan_queue_xmit(struct sk_ >> >>if (vlan->mode == MACVLAN_MODE_BRIDGE) { >>const struct ethhdr *eth = (void *)skb->data; >> + skb->ip_summed = CHECKSUM_UNNECESSARY; >> >>/* send to other bridge ports directly */ >>if (is_multicast_ether_addr(eth->h_dest)) { >> >> >> and that fixed the problem. Do you think it is acceptable ? > The only problem I see is if the packets are bridged to a > different networking device (or redirected using the mirred > action), in this case the checksum will not be completed. > This would be a very strange setup though and probably wouldn't > be using dummy as lower device, so I'm not sure we have to > worry about this case. I am not sure to get it, do you say the patch is correct ? >>> Its correct with a short-coming that doesn't seem to matter. >>> If my understanding is correct, the packet will be flagged CHECKSUM_UNNECESSARY only for the macvlan devices, right ? >>> Only for packets bridged between macvlan devices. A setup like >>> the following would cause problems: >>> >>> br0 >>>| >>> .--. >>> | | >>> macvlan0macvlan1eth0 >>> | | >>> ---.--- >>> dummy0 >>> >>> In this case packets sent from macvlan0 will show up on >>> eth0 with incorrect setups. However this setup doesn't >>> seem realistic to me, you would simply use eth0 instead >>> of dummy0. >> Ok, I understand. thanks for the clarification. >> By the way, this problem occurs for any lower device with offloading capabilities with a macvlan port in bridge mode. >>> True. This doesn't affect outgoing packets since their checksum >>> will be completed in dev_hard_start_xmit(), but it affects >>> packets bridged between macvlans. >> One last question. In the case of broadcast packets with maclvan in >> bridge mode. >> We will have the packets going through each macvlan port and also to the >> lower-device, right ? >> For the latter, don't we have a problem if the packet is flagged >> CHECKSUM_UNNECESSARY ? >> >> Shouldn't we restore the ip_summed field before sending through >> dev_queue_xmit ? > Yes, that seems correct in order to have dev_hard_start_xmit() complete > the checksum if necessary. -- Colocation vs. Managed Hosting A question and answer guide to determining the best fit for your organization - today and in the future. http://p.sf.net/sfu/internap-sfd2d ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
Am 02.03.2011 19:33, schrieb Daniel Lezcano: > On 03/02/2011 07:03 PM, Patrick McHardy wrote: >> Am 02.03.2011 17:03, schrieb Daniel Lezcano: >>> On 03/02/2011 12:03 PM, Patrick McHardy wrote: Am 01.03.2011 21:04, schrieb Daniel Lezcano: > On 03/01/2011 05:51 PM, Patrick McHardy wrote: >>> Patrick, do you have any suggestions to fix this ? >> Since the frames are only looped back locally, I suppose the easiest >> fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively >> we need to complete the checksum manually, similar to what >> dev_hard_start_xmit() does. > That sounds very simple to fix, maybe too much simple :) > > I did the following change: > > --- linux-next.orig/drivers/net/macvlan.c > +++ linux-next/drivers/net/macvlan.c > @@ -222,6 +222,7 @@ static int macvlan_queue_xmit(struct sk_ > > if (vlan->mode == MACVLAN_MODE_BRIDGE) { > const struct ethhdr *eth = (void *)skb->data; > + skb->ip_summed = CHECKSUM_UNNECESSARY; > > /* send to other bridge ports directly */ > if (is_multicast_ether_addr(eth->h_dest)) { > > > and that fixed the problem. Do you think it is acceptable ? The only problem I see is if the packets are bridged to a different networking device (or redirected using the mirred action), in this case the checksum will not be completed. This would be a very strange setup though and probably wouldn't be using dummy as lower device, so I'm not sure we have to worry about this case. >>> I am not sure to get it, do you say the patch is correct ? >> Its correct with a short-coming that doesn't seem to matter. >> >>> If my understanding is correct, the packet will be flagged >>> CHECKSUM_UNNECESSARY only for the macvlan devices, right ? >> Only for packets bridged between macvlan devices. A setup like >> the following would cause problems: >> >> br0 >> | >> .--. >> | | >> macvlan0macvlan1eth0 >> | | >> ---.--- >> dummy0 >> >> In this case packets sent from macvlan0 will show up on >> eth0 with incorrect setups. However this setup doesn't >> seem realistic to me, you would simply use eth0 instead >> of dummy0. > > Ok, I understand. thanks for the clarification. > >>> By the way, this problem occurs for any lower device with offloading >>> capabilities with a macvlan port in bridge mode. >> True. This doesn't affect outgoing packets since their checksum >> will be completed in dev_hard_start_xmit(), but it affects >> packets bridged between macvlans. > > One last question. In the case of broadcast packets with maclvan in > bridge mode. > We will have the packets going through each macvlan port and also to the > lower-device, right ? > For the latter, don't we have a problem if the packet is flagged > CHECKSUM_UNNECESSARY ? > > Shouldn't we restore the ip_summed field before sending through > dev_queue_xmit ? Yes, that seems correct in order to have dev_hard_start_xmit() complete the checksum if necessary. -- What You Don't Know About Data Connectivity CAN Hurt You This paper provides an overview of data connectivity, details its effect on application quality, and explores various alternative solutions. http://p.sf.net/sfu/progress-d2d ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On Thu, Mar 3, 2011 at 2:33 AM, Daniel Lezcano wrote: > > One last question. In the case of broadcast packets with maclvan in bridge > mode. > We will have the packets going through each macvlan port and also to the > lower-device, right ? > For the latter, don't we have a problem if the packet is flagged > CHECKSUM_UNNECESSARY ? > > Shouldn't we restore the ip_summed field before sending through > dev_queue_xmit ? > > --- > drivers/net/macvlan.c | 3 +++ > 1 file changed, 3 insertions(+) > > Index: linux-next/drivers/net/macvlan.c > === > --- linux-next.orig/drivers/net/macvlan.c > +++ linux-next/drivers/net/macvlan.c > @@ -222,10 +222,13 @@ static int macvlan_queue_xmit(struct sk_ > > if (vlan->mode == MACVLAN_MODE_BRIDGE) { > const struct ethhdr *eth = (void *)skb->data; > + __u8 ip_summed = skb->ip_summed; > + skb->ip_summed; = CHECKSUM_UNNECESSARY; You added a additional ";". -- Regards, Changli Gao(xiao...@gmail.com) -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
Am 02.03.2011 17:03, schrieb Daniel Lezcano: > On 03/02/2011 12:03 PM, Patrick McHardy wrote: >> Am 01.03.2011 21:04, schrieb Daniel Lezcano: >>> On 03/01/2011 05:51 PM, Patrick McHardy wrote: > Patrick, do you have any suggestions to fix this ? Since the frames are only looped back locally, I suppose the easiest fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively we need to complete the checksum manually, similar to what dev_hard_start_xmit() does. >>> That sounds very simple to fix, maybe too much simple :) >>> >>> I did the following change: >>> >>> --- linux-next.orig/drivers/net/macvlan.c >>> +++ linux-next/drivers/net/macvlan.c >>> @@ -222,6 +222,7 @@ static int macvlan_queue_xmit(struct sk_ >>> >>> if (vlan->mode == MACVLAN_MODE_BRIDGE) { >>> const struct ethhdr *eth = (void *)skb->data; >>> + skb->ip_summed = CHECKSUM_UNNECESSARY; >>> >>> /* send to other bridge ports directly */ >>> if (is_multicast_ether_addr(eth->h_dest)) { >>> >>> >>> and that fixed the problem. Do you think it is acceptable ? >> The only problem I see is if the packets are bridged to a >> different networking device (or redirected using the mirred >> action), in this case the checksum will not be completed. >> This would be a very strange setup though and probably wouldn't >> be using dummy as lower device, so I'm not sure we have to >> worry about this case. > > I am not sure to get it, do you say the patch is correct ? Its correct with a short-coming that doesn't seem to matter. > If my understanding is correct, the packet will be flagged > CHECKSUM_UNNECESSARY only for the macvlan devices, right ? Only for packets bridged between macvlan devices. A setup like the following would cause problems: br0 | .--. | | macvlan0macvlan1eth0 | | ---.--- dummy0 In this case packets sent from macvlan0 will show up on eth0 with incorrect setups. However this setup doesn't seem realistic to me, you would simply use eth0 instead of dummy0. > By the way, this problem occurs for any lower device with offloading > capabilities with a macvlan port in bridge mode. True. This doesn't affect outgoing packets since their checksum will be completed in dev_hard_start_xmit(), but it affects packets bridged between macvlans. As I said, the full solution would be to complete the checksum for CHECKSUM_PARTIAL packets when bridging between macvlans. -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 03/02/2011 12:03 PM, Patrick McHardy wrote: > Am 01.03.2011 21:04, schrieb Daniel Lezcano: >> On 03/01/2011 05:51 PM, Patrick McHardy wrote: Patrick, do you have any suggestions to fix this ? >>> Since the frames are only looped back locally, I suppose the easiest >>> fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively >>> we need to complete the checksum manually, similar to what >>> dev_hard_start_xmit() does. >> That sounds very simple to fix, maybe too much simple :) >> >> I did the following change: >> >> --- linux-next.orig/drivers/net/macvlan.c >> +++ linux-next/drivers/net/macvlan.c >> @@ -222,6 +222,7 @@ static int macvlan_queue_xmit(struct sk_ >> >> if (vlan->mode == MACVLAN_MODE_BRIDGE) { >> const struct ethhdr *eth = (void *)skb->data; >> + skb->ip_summed = CHECKSUM_UNNECESSARY; >> >> /* send to other bridge ports directly */ >> if (is_multicast_ether_addr(eth->h_dest)) { >> >> >> and that fixed the problem. Do you think it is acceptable ? > The only problem I see is if the packets are bridged to a > different networking device (or redirected using the mirred > action), in this case the checksum will not be completed. > This would be a very strange setup though and probably wouldn't > be using dummy as lower device, so I'm not sure we have to > worry about this case. I am not sure to get it, do you say the patch is correct ? If my understanding is correct, the packet will be flagged CHECKSUM_UNNECESSARY only for the macvlan devices, right ? By the way, this problem occurs for any lower device with offloading capabilities with a macvlan port in bridge mode. Thanks -- Daniel -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 03/02/2011 07:03 PM, Patrick McHardy wrote: > Am 02.03.2011 17:03, schrieb Daniel Lezcano: >> On 03/02/2011 12:03 PM, Patrick McHardy wrote: >>> Am 01.03.2011 21:04, schrieb Daniel Lezcano: On 03/01/2011 05:51 PM, Patrick McHardy wrote: >> Patrick, do you have any suggestions to fix this ? > Since the frames are only looped back locally, I suppose the easiest > fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively > we need to complete the checksum manually, similar to what > dev_hard_start_xmit() does. That sounds very simple to fix, maybe too much simple :) I did the following change: --- linux-next.orig/drivers/net/macvlan.c +++ linux-next/drivers/net/macvlan.c @@ -222,6 +222,7 @@ static int macvlan_queue_xmit(struct sk_ if (vlan->mode == MACVLAN_MODE_BRIDGE) { const struct ethhdr *eth = (void *)skb->data; + skb->ip_summed = CHECKSUM_UNNECESSARY; /* send to other bridge ports directly */ if (is_multicast_ether_addr(eth->h_dest)) { and that fixed the problem. Do you think it is acceptable ? >>> The only problem I see is if the packets are bridged to a >>> different networking device (or redirected using the mirred >>> action), in this case the checksum will not be completed. >>> This would be a very strange setup though and probably wouldn't >>> be using dummy as lower device, so I'm not sure we have to >>> worry about this case. >> I am not sure to get it, do you say the patch is correct ? > Its correct with a short-coming that doesn't seem to matter. > >> If my understanding is correct, the packet will be flagged >> CHECKSUM_UNNECESSARY only for the macvlan devices, right ? > Only for packets bridged between macvlan devices. A setup like > the following would cause problems: > > br0 > | > .--. > | | > macvlan0 macvlan1eth0 > | | > ---.--- >dummy0 > > In this case packets sent from macvlan0 will show up on > eth0 with incorrect setups. However this setup doesn't > seem realistic to me, you would simply use eth0 instead > of dummy0. Ok, I understand. thanks for the clarification. >> By the way, this problem occurs for any lower device with offloading >> capabilities with a macvlan port in bridge mode. > True. This doesn't affect outgoing packets since their checksum > will be completed in dev_hard_start_xmit(), but it affects > packets bridged between macvlans. One last question. In the case of broadcast packets with maclvan in bridge mode. We will have the packets going through each macvlan port and also to the lower-device, right ? For the latter, don't we have a problem if the packet is flagged CHECKSUM_UNNECESSARY ? Shouldn't we restore the ip_summed field before sending through dev_queue_xmit ? --- drivers/net/macvlan.c |3 +++ 1 file changed, 3 insertions(+) Index: linux-next/drivers/net/macvlan.c === --- linux-next.orig/drivers/net/macvlan.c +++ linux-next/drivers/net/macvlan.c @@ -222,10 +222,13 @@ static int macvlan_queue_xmit(struct sk_ if (vlan->mode == MACVLAN_MODE_BRIDGE) { const struct ethhdr *eth = (void *)skb->data; + __u8 ip_summed = skb->ip_summed; + skb->ip_summed; = CHECKSUM_UNNECESSARY; /* send to other bridge ports directly */ if (is_multicast_ether_addr(eth->h_dest)) { macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); + skb->ip_summed = ip_summed; goto xmit_world; } -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
Am 01.03.2011 21:04, schrieb Daniel Lezcano: > On 03/01/2011 05:51 PM, Patrick McHardy wrote: >>> Patrick, do you have any suggestions to fix this ? >> Since the frames are only looped back locally, I suppose the easiest >> fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively >> we need to complete the checksum manually, similar to what >> dev_hard_start_xmit() does. > > That sounds very simple to fix, maybe too much simple :) > > I did the following change: > > --- linux-next.orig/drivers/net/macvlan.c > +++ linux-next/drivers/net/macvlan.c > @@ -222,6 +222,7 @@ static int macvlan_queue_xmit(struct sk_ > > if (vlan->mode == MACVLAN_MODE_BRIDGE) { > const struct ethhdr *eth = (void *)skb->data; > + skb->ip_summed = CHECKSUM_UNNECESSARY; > > /* send to other bridge ports directly */ > if (is_multicast_ether_addr(eth->h_dest)) { > > > and that fixed the problem. Do you think it is acceptable ? The only problem I see is if the packets are bridged to a different networking device (or redirected using the mirred action), in this case the checksum will not be completed. This would be a very strange setup though and probably wouldn't be using dummy as lower device, so I'm not sure we have to worry about this case. -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 01.03.2011 14:29, Daniel Lezcano wrote: > On 02/28/2011 08:45 AM, Eric Dumazet wrote: >>> In the normal case, dummy0 is supposed to drop the packets. But with >>> macvlan these packets are broadcasted to the other macvlan ports, so no >>> checksum is computed when the packets are transmitted between macvlan1 >>> and macvlan2. >> So where frames get bad checksums ? >> >> In this "bridge" mode, I suspect the broadcast is done _before_ sending >> frame to dummy, so maybe macvlan should not inherit from lowerdev in >> this particular case ? > > Hi Eric, > > yes, you are right, the packets are sent before. > > In the 'macvlan_queue_xmit', the code checks the dev is in 'bridge' > mode. If so, it looks if there is a destination port for the packet and > then calls the 'forward' callback which is 'dev_forward_skb'. > > I was able to reproduce the same problem with qemu and an emulated > 'e1000' card instead of dummy0. The packets are dropped too. > > Patrick, do you have any suggestions to fix this ? Since the frames are only looped back locally, I suppose the easiest fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively we need to complete the checksum manually, similar to what dev_hard_start_xmit() does. -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 03/01/2011 05:51 PM, Patrick McHardy wrote: > On 01.03.2011 14:29, Daniel Lezcano wrote: >> On 02/28/2011 08:45 AM, Eric Dumazet wrote: In the normal case, dummy0 is supposed to drop the packets. But with macvlan these packets are broadcasted to the other macvlan ports, so no checksum is computed when the packets are transmitted between macvlan1 and macvlan2. >>> So where frames get bad checksums ? >>> >>> In this "bridge" mode, I suspect the broadcast is done _before_ sending >>> frame to dummy, so maybe macvlan should not inherit from lowerdev in >>> this particular case ? >> Hi Eric, >> >> yes, you are right, the packets are sent before. >> >> In the 'macvlan_queue_xmit', the code checks the dev is in 'bridge' >> mode. If so, it looks if there is a destination port for the packet and >> then calls the 'forward' callback which is 'dev_forward_skb'. >> >> I was able to reproduce the same problem with qemu and an emulated >> 'e1000' card instead of dummy0. The packets are dropped too. >> >> Patrick, do you have any suggestions to fix this ? > Since the frames are only looped back locally, I suppose the easiest > fix would be to mark them with CHECKSUM_UNNECESSARY. Alternatively > we need to complete the checksum manually, similar to what > dev_hard_start_xmit() does. That sounds very simple to fix, maybe too much simple :) I did the following change: --- drivers/net/macvlan.c |1 + 1 file changed, 1 insertion(+) Index: linux-next/drivers/net/macvlan.c === --- linux-next.orig/drivers/net/macvlan.c +++ linux-next/drivers/net/macvlan.c @@ -222,6 +222,7 @@ static int macvlan_queue_xmit(struct sk_ if (vlan->mode == MACVLAN_MODE_BRIDGE) { const struct ethhdr *eth = (void *)skb->data; + skb->ip_summed = CHECKSUM_UNNECESSARY; /* send to other bridge ports directly */ if (is_multicast_ether_addr(eth->h_dest)) { and that fixed the problem. Do you think it is acceptable ? Thanks -- Daniel -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 02/28/2011 08:45 AM, Eric Dumazet wrote: > Le dimanche 27 février 2011 à 21:35 +0100, Daniel Lezcano a écrit : >> On 02/27/2011 08:50 PM, Eric Dumazet wrote: >>> Le dimanche 27 février 2011 à 16:14 +0100, Daniel Lezcano a écrit : On 02/23/2011 06:13 PM, Andrian Nord wrote: > On Mon, Feb 21, 2011 at 05:07:31PM +0100, Daniel Lezcano wrote: >> I Cc'ed the netdev mailing list and Patrick in case my analysis is wrong >> or incomplete. > I'm confirming, that this happens only when macvlan's are onto dummy net > device. In case of some physical interface under macvlan there is no lost > packages and no broken checksums. I did some tests with a 2.6.35 kernel version and it seems the checksum errors do not appear. I noticed there are some changes in the dummy setup function: dev->features |= NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_TSO; dev->features |= NETIF_F_NO_CSUM | NETIF_F_HIGHDMA | NETIF_F_LLTX; May be that was introduced by commit: commit 6d81f41c58c69ddde497e9e640ba5805aa26e78c Author: Eric Dumazet Date: Mon Sep 27 20:50:33 2010 + dummy: percpu stats and lockless xmit Converts dummy network device driver to : - percpu stats - 64bit stats - lockless xmit (NETIF_F_LLTX) - performance features added (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_TSO | NETIF_F_NO_CSUM | NETIF_F_HIGHDMA) Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller Eric, Andrian is observing, with a couple of macvlan (in bridge mode) on top of a dummy interface, a lot of checksums error and packets drop. Each macvlan is in a different network namespace and the dummy interface is in the init_net. Any ideas ? >>> Not sure I understand... I thought dummy was dropping all frames >>> anyway ? >>> >>> static netdev_tx_t dummy_xmit(struct sk_buff *skb, struct net_device *dev) >>> { >>> struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); >>> >>> u64_stats_update_begin(&dstats->syncp); >>> dstats->tx_packets++; >>> dstats->tx_bytes += skb->len; >>> u64_stats_update_end(&dstats->syncp); >>> >>> dev_kfree_skb(skb); >>> return NETDEV_TX_OK; >>> } >>> >>> >>> Maybe you could describe the setup ? >> Yes, it is very simple. >> >> There are two network namespaces. >> >> macvlan1 is in network namespace 1 >> macvlan2 is in network namespace 2 >> >> Both are in "bridge" mode, so they can communicate together. >> The lower device is dummy0 in the init network namespace. >> >> IMO the problem is coming from the macvlan driver: >> >> dev->features = lowerdev->features& MACVLAN_FEATURES >> >> As dummy0 has the offloading capabilities set on, the macvlan driver >> inherit these features. >> >> In the normal case, dummy0 is supposed to drop the packets. But with >> macvlan these packets are broadcasted to the other macvlan ports, so no >> checksum is computed when the packets are transmitted between macvlan1 >> and macvlan2. > So where frames get bad checksums ? > > In this "bridge" mode, I suspect the broadcast is done _before_ sending > frame to dummy, so maybe macvlan should not inherit from lowerdev in > this particular case ? Hi Eric, yes, you are right, the packets are sent before. In the 'macvlan_queue_xmit', the code checks the dev is in 'bridge' mode. If so, it looks if there is a destination port for the packet and then calls the 'forward' callback which is 'dev_forward_skb'. I was able to reproduce the same problem with qemu and an emulated 'e1000' card instead of dummy0. The packets are dropped too. Patrick, do you have any suggestions to fix this ? Thanks -- Daniel -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
Le dimanche 27 février 2011 à 21:35 +0100, Daniel Lezcano a écrit : > On 02/27/2011 08:50 PM, Eric Dumazet wrote: > > Le dimanche 27 février 2011 à 16:14 +0100, Daniel Lezcano a écrit : > >> On 02/23/2011 06:13 PM, Andrian Nord wrote: > >>> On Mon, Feb 21, 2011 at 05:07:31PM +0100, Daniel Lezcano wrote: > I Cc'ed the netdev mailing list and Patrick in case my analysis is wrong > or incomplete. > >>> I'm confirming, that this happens only when macvlan's are onto dummy net > >>> device. In case of some physical interface under macvlan there is no lost > >>> packages and no broken checksums. > >> I did some tests with a 2.6.35 kernel version and it seems the checksum > >> errors do not appear. > >> I noticed there are some changes in the dummy setup function: > >> > >> dev->features |= NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_TSO; > >> dev->features |= NETIF_F_NO_CSUM | NETIF_F_HIGHDMA | NETIF_F_LLTX; > >> > >> > >> May be that was introduced by commit: > >> > >> commit 6d81f41c58c69ddde497e9e640ba5805aa26e78c > >> Author: Eric Dumazet > >> Date: Mon Sep 27 20:50:33 2010 + > >> > >> dummy: percpu stats and lockless xmit > >> > >> Converts dummy network device driver to : > >> > >> - percpu stats > >> > >> - 64bit stats > >> > >> - lockless xmit (NETIF_F_LLTX) > >> > >> - performance features added (NETIF_F_SG | NETIF_F_FRAGLIST | > >> NETIF_F_TSO | NETIF_F_NO_CSUM | NETIF_F_HIGHDMA) > >> > >> Signed-off-by: Eric Dumazet > >> Signed-off-by: David S. Miller > >> > >> > >> Eric, > >> > >> Andrian is observing, with a couple of macvlan (in bridge mode) on top > >> of a dummy interface, a lot of checksums error and packets drop. > >> Each macvlan is in a different network namespace and the dummy interface > >> is in the init_net. > >> > >> Any ideas ? > > Not sure I understand... I thought dummy was dropping all frames > > anyway ? > > > > static netdev_tx_t dummy_xmit(struct sk_buff *skb, struct net_device *dev) > > { > > struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); > > > > u64_stats_update_begin(&dstats->syncp); > > dstats->tx_packets++; > > dstats->tx_bytes += skb->len; > > u64_stats_update_end(&dstats->syncp); > > > > dev_kfree_skb(skb); > > return NETDEV_TX_OK; > > } > > > > > > Maybe you could describe the setup ? > > Yes, it is very simple. > > There are two network namespaces. > > macvlan1 is in network namespace 1 > macvlan2 is in network namespace 2 > > Both are in "bridge" mode, so they can communicate together. > The lower device is dummy0 in the init network namespace. > > IMO the problem is coming from the macvlan driver: > > dev->features = lowerdev->features & MACVLAN_FEATURES > > As dummy0 has the offloading capabilities set on, the macvlan driver > inherit these features. > > In the normal case, dummy0 is supposed to drop the packets. But with > macvlan these packets are broadcasted to the other macvlan ports, so no > checksum is computed when the packets are transmitted between macvlan1 > and macvlan2. So where frames get bad checksums ? In this "bridge" mode, I suspect the broadcast is done _before_ sending frame to dummy, so maybe macvlan should not inherit from lowerdev in this particular case ? -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 02/27/2011 08:50 PM, Eric Dumazet wrote: > Le dimanche 27 février 2011 à 16:14 +0100, Daniel Lezcano a écrit : >> On 02/23/2011 06:13 PM, Andrian Nord wrote: >>> On Mon, Feb 21, 2011 at 05:07:31PM +0100, Daniel Lezcano wrote: I Cc'ed the netdev mailing list and Patrick in case my analysis is wrong or incomplete. >>> I'm confirming, that this happens only when macvlan's are onto dummy net >>> device. In case of some physical interface under macvlan there is no lost >>> packages and no broken checksums. >> I did some tests with a 2.6.35 kernel version and it seems the checksum >> errors do not appear. >> I noticed there are some changes in the dummy setup function: >> >> dev->features |= NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_TSO; >> dev->features |= NETIF_F_NO_CSUM | NETIF_F_HIGHDMA | NETIF_F_LLTX; >> >> >> May be that was introduced by commit: >> >> commit 6d81f41c58c69ddde497e9e640ba5805aa26e78c >> Author: Eric Dumazet >> Date: Mon Sep 27 20:50:33 2010 + >> >> dummy: percpu stats and lockless xmit >> >> Converts dummy network device driver to : >> >> - percpu stats >> >> - 64bit stats >> >> - lockless xmit (NETIF_F_LLTX) >> >> - performance features added (NETIF_F_SG | NETIF_F_FRAGLIST | >> NETIF_F_TSO | NETIF_F_NO_CSUM | NETIF_F_HIGHDMA) >> >> Signed-off-by: Eric Dumazet >> Signed-off-by: David S. Miller >> >> >> Eric, >> >> Andrian is observing, with a couple of macvlan (in bridge mode) on top >> of a dummy interface, a lot of checksums error and packets drop. >> Each macvlan is in a different network namespace and the dummy interface >> is in the init_net. >> >> Any ideas ? > Not sure I understand... I thought dummy was dropping all frames > anyway ? > > static netdev_tx_t dummy_xmit(struct sk_buff *skb, struct net_device *dev) > { > struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); > > u64_stats_update_begin(&dstats->syncp); > dstats->tx_packets++; > dstats->tx_bytes += skb->len; > u64_stats_update_end(&dstats->syncp); > > dev_kfree_skb(skb); > return NETDEV_TX_OK; > } > > > Maybe you could describe the setup ? Yes, it is very simple. There are two network namespaces. macvlan1 is in network namespace 1 macvlan2 is in network namespace 2 Both are in "bridge" mode, so they can communicate together. The lower device is dummy0 in the init network namespace. IMO the problem is coming from the macvlan driver: dev->features = lowerdev->features & MACVLAN_FEATURES As dummy0 has the offloading capabilities set on, the macvlan driver inherit these features. In the normal case, dummy0 is supposed to drop the packets. But with macvlan these packets are broadcasted to the other macvlan ports, so no checksum is computed when the packets are transmitted between macvlan1 and macvlan2. -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
Le dimanche 27 février 2011 à 16:14 +0100, Daniel Lezcano a écrit : > On 02/23/2011 06:13 PM, Andrian Nord wrote: > > On Mon, Feb 21, 2011 at 05:07:31PM +0100, Daniel Lezcano wrote: > >> I Cc'ed the netdev mailing list and Patrick in case my analysis is wrong > >> or incomplete. > > I'm confirming, that this happens only when macvlan's are onto dummy net > > device. In case of some physical interface under macvlan there is no lost > > packages and no broken checksums. > > I did some tests with a 2.6.35 kernel version and it seems the checksum > errors do not appear. > I noticed there are some changes in the dummy setup function: > >dev->features |= NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_TSO; >dev->features |= NETIF_F_NO_CSUM | NETIF_F_HIGHDMA | NETIF_F_LLTX; > > > May be that was introduced by commit: > > commit 6d81f41c58c69ddde497e9e640ba5805aa26e78c > Author: Eric Dumazet > Date: Mon Sep 27 20:50:33 2010 + > > dummy: percpu stats and lockless xmit > > Converts dummy network device driver to : > > - percpu stats > > - 64bit stats > > - lockless xmit (NETIF_F_LLTX) > > - performance features added (NETIF_F_SG | NETIF_F_FRAGLIST | > NETIF_F_TSO | NETIF_F_NO_CSUM | NETIF_F_HIGHDMA) > > Signed-off-by: Eric Dumazet > Signed-off-by: David S. Miller > > > Eric, > > Andrian is observing, with a couple of macvlan (in bridge mode) on top > of a dummy interface, a lot of checksums error and packets drop. > Each macvlan is in a different network namespace and the dummy interface > is in the init_net. > > Any ideas ? Not sure I understand... I thought dummy was dropping all frames anyway ? static netdev_tx_t dummy_xmit(struct sk_buff *skb, struct net_device *dev) { struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); u64_stats_update_begin(&dstats->syncp); dstats->tx_packets++; dstats->tx_bytes += skb->len; u64_stats_update_end(&dstats->syncp); dev_kfree_skb(skb); return NETDEV_TX_OK; } Maybe you could describe the setup ? -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 02/23/2011 06:13 PM, Andrian Nord wrote: > On Mon, Feb 21, 2011 at 05:07:31PM +0100, Daniel Lezcano wrote: >> I Cc'ed the netdev mailing list and Patrick in case my analysis is wrong >> or incomplete. > I'm confirming, that this happens only when macvlan's are onto dummy net > device. In case of some physical interface under macvlan there is no lost > packages and no broken checksums. I did some tests with a 2.6.35 kernel version and it seems the checksum errors do not appear. I noticed there are some changes in the dummy setup function: dev->features |= NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_TSO; dev->features |= NETIF_F_NO_CSUM | NETIF_F_HIGHDMA | NETIF_F_LLTX; May be that was introduced by commit: commit 6d81f41c58c69ddde497e9e640ba5805aa26e78c Author: Eric Dumazet Date: Mon Sep 27 20:50:33 2010 + dummy: percpu stats and lockless xmit Converts dummy network device driver to : - percpu stats - 64bit stats - lockless xmit (NETIF_F_LLTX) - performance features added (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_TSO | NETIF_F_NO_CSUM | NETIF_F_HIGHDMA) Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller Eric, Andrian is observing, with a couple of macvlan (in bridge mode) on top of a dummy interface, a lot of checksums error and packets drop. Each macvlan is in a different network namespace and the dummy interface is in the init_net. Any ideas ? -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On Thu, Feb 24, 2011 at 11:20:09AM +0100, Daniel Lezcano wrote: > I saw you were using the command 'nc6', do you use netcat with ipv6 ? Well, yes and no. I've tried both ipv4 and ipv6 and my notebook has no ipv6 address assigned, so most terrible connection was though ipv4 =). At another server there is no ipv6 at all. pgpAYEtg4XnOZ.pgp Description: PGP signature -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On 02/23/2011 06:13 PM, Andrian Nord wrote: > On Mon, Feb 21, 2011 at 05:07:31PM +0100, Daniel Lezcano wrote: >> I Cc'ed the netdev mailing list and Patrick in case my analysis is wrong >> or incomplete. > I'm confirming, that this happens only when macvlan's are onto dummy net > device. In case of some physical interface under macvlan there is no lost > packages and no broken checksums. Hmm, that's interesting ... I saw you were using the command 'nc6', do you use netcat with ipv6 ? -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
Re: [Lxc-users] Bad checksums and lost packets with macvlan on dummy
On Mon, Feb 21, 2011 at 05:07:31PM +0100, Daniel Lezcano wrote: > I Cc'ed the netdev mailing list and Patrick in case my analysis is wrong > or incomplete. I'm confirming, that this happens only when macvlan's are onto dummy net device. In case of some physical interface under macvlan there is no lost packages and no broken checksums. pgpCokp2RzaJS.pgp Description: PGP signature -- Free Software Download: Index, Search & Analyze Logs and other IT data in Real-Time with Splunk. Collect, index and harness all the fast moving IT data generated by your applications, servers and devices whether physical, virtual or in the cloud. Deliver compliance at lower cost and gain new business insights. http://p.sf.net/sfu/splunk-dev2dev ___ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users