net: neighbour: use source address of last enqueued packet for solicitation
Subsystems (2)
RFC PATCH 1/1 net: neighbour: use source address of last enqueued packet for solicitation
Currently we always use the first member of the arp_queue to determine the sender ip address of the arp packet (or in case of IPv6 - source address of the ndisc packet). This skb is fixed as long as the queue is not drained by a complete purge because of a timeout or by a successful response.
If the first packet enqueued on the arp_queue is from a local application with a manually set source address and the to be discovered system does some kind of uRPF checks on the source address in the arp packet the resolving process hangs until a timeout and restarts. This hurts communication with the participating network node.
This could be mitigated a bit if we use the latest enqueued skb's source address for the resolving process, which is not as static as the arp_queue's head. This change of the source address could result in better recovery of a failed solicitation.
---
I didn't find anything which could break because of this change, but please have a second look.
net/core/neighbour.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
net/core/neighbour.c +1 -1
@@ -867,7 +867,7 @@ static void neigh_invalidate(struct neighbour *neigh)static void neigh_probe (struct neighbour *neigh) __releases(neigh->lock) { -struct sk_buff *skb =skb_peek (&neigh->arp_queue); +struct sk_buff *skb =skb_peek_tail (&neigh->arp_queue);/* keep skb alive even if arp_queue overflows */ if (skb) skb =skb_copy (skb,GFP_ATOMIC );
Signature · 1 line
1.8.3.1
Hello,
Quoted 17 lines · Hannes Frederic Sowa
Currently we always use the first member of the
arp_queueto determine the sender ip address of the arp packet (or in case of IPv6 - source address of the ndisc packet). This skb is fixed as long as the queue is not drained by a complete purge because of a timeout or by a successful response.If the first packet enqueued on the
arp_queueis from a local application with a manually set source address and the to be discovered system does some kind of uRPF checks on the source address in the arp packet the resolving process hangs until a timeout and restarts. This hurts communication with the participating network node.This could be mitigated a bit if we use the latest enqueued skb's source address for the resolving process, which is not as static as the arp_queue's head. This change of the source address could result in better recovery of a failed solicitation.
Quoted 7 lines
Cc "David S. Miller" <[email protected]>Cc Julian Anastasov <[email protected]>Signed-off-by Hannes Frederic Sowa <[email protected]>---
I didn't find anything which could break because of this change, but please have a second look.
arp_queue has packets only in NUD_INCOMPLETE state
(mcast_solicit=3 secs by default). And __neigh_event_send() now can keep many packets, 64KB from recent changes. So the 1st place is not guaranteed but now it is more difficult to kick the first packet compared to the old limit of just 3 packets.
The change can give chance for 2nd and 3th
probe if the 1st probe is not replied, so it should be better to apply it:
Still, I think such problems should be addressed
with conf/{DEV,all}/arp_announce=1 or 2.
Quoted 18 lines
net/core/neighbour.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)net/core/neighbour.c +1
-1@@ -867,7 +867,7 @@ static void neigh_invalidate(struct neighbour *neigh)static void neigh_probe (struct neighbour *neigh) __releases(neigh->lock) { -struct sk_buff *skb =skb_peek (&neigh->arp_queue); +struct sk_buff *skb =skb_peek_tail (&neigh->arp_queue);/* keep skb alive even if arp_queue overflows */ if (skb) skb =skb_copy (skb,GFP_ATOMIC );Signature · 1 line
1.8.3.1
Regards
--
Julian Anastasov <[email protected]>
Quoted 7 lines · Julian Anastasov
arp_queuehas packets only inNUD_INCOMPLETEstate
(mcast_solicit=3 secs by default). And__neigh_event_send()now can keep many packets, 64KB from recent changes. So the 1st place is not guaranteed but now it is more difficult to kick the first packet compared to the old limit of just 3 packets.
Quoted 5 lines
The change can give chance for 2nd and 3th
probe if the 1st probe is not replied, so it should be better to apply it:Reviewed-by Julian Anastasov <[email protected]>
Thanks for the review. I resend the patch as soon as net-next opens.
Quoted 3 lines
Still, I think such problems should be addressed
with conf/{DEV,all}/arp_announce=1 or 2.
*nod*
I use this knob if I have such problems. But this patch improves connectivity in the default configuration and we actually don't care much about the source address in either ipv4 or ipv6. So it seemed legitimate and simple to me.
Greetings,
Hannes