net l2tp: drop flow hash on forward
Subsystems (2)
PATCH 1/1 net l2tp: drop flow hash on forward
Drop the flow-hash of the skb when forwarding to the L2TP netdev.
This avoids the L2TP qdisc from using the flow-hash from the outer packet, which is identical for every flow within the tunnel.
This does not affect every platform but is specific for the ethernet driver. It depends on the platform including L4 information in the flow-hash.
One such example is the Mediatek Filogic MT798x family of networking processors.
--- net/l2tp/l2tp_eth.c | 3 +++ 1 file changed, 3 insertions(+)
net/l2tp/l2tp_eth.c +3 -0
@@ -127,6 +127,9 @@ static void l2tp_eth_dev_recv(struct l2tp_session *session, struct sk_buff *skb,/* checksums verified by L2TP */ skb->ip_summed =CHECKSUM_NONE ; +/* drop outer flow-hash */ +skb_clear_hash (skb); +skb_dst_drop (skb);nf_reset_ct (skb);
Signature · 1 line
2.43.0
Quoted 5 lines · David Bauer
Drop the flow-hash of the skb when forwarding to the L2TP netdev.
This avoids the L2TP qdisc from using the flow-hash from the outer packet, which is identical for every flow within the tunnel.
Quoted 8 lines
This does not affect every platform but is specific for the ethernet driver. It depends on the platform including L4 information in the flow-hash.
One such example is the Mediatek Filogic MT798x family of networking processors.
Signed-off-by David Bauer <[email protected]>
While it's difficult to say which exact commit this fixes, I still consider it being a fix, as otherwise flow-offloading on mentioned platforms will face difficulties when using L2TP (right?). Hence maybe it should go via 'net' tree rather than via 'net-next'?
The fix itself looks fine to me.
Quoted 22 lines
--- net/l2tp/l2tp_eth.c | 3 +++ 1 file changed, 3 insertions(+)net/l2tp/l2tp_eth.c +3
-0@@ -127,6 +127,9 @@ static void l2tp_eth_dev_recv(struct l2tp_session *session, struct sk_buff *skb,/* checksums verified by L2TP */ skb->ip_summed =CHECKSUM_NONE ; +/* drop outer flow-hash */ +skb_clear_hash (skb); +skb_dst_drop (skb);nf_reset_ct (skb);Signature · 1 line
2.43.0
Hi Daniel,
Quoted 10 lines · Daniel Golle
Quoted 9 lines · David Bauer
Drop the flow-hash of the skb when forwarding to the L2TP netdev.
This avoids the L2TP qdisc from using the flow-hash from the outer packet, which is identical for every flow within the tunnel.
This does not affect every platform but is specific for the ethernet driver. It depends on the platform including L4 information in the flow-hash.
Quoted 8 lines
Quoted 4 lines
One such example is the Mediatek Filogic MT798x family of networking processors.
Signed-off-by David Bauer <[email protected]>While it's difficult to say which exact commit this fixes, I still consider it being a fix, as otherwise flow-offloading on mentioned platforms will face difficulties when using L2TP (right?).
I'm unsure whether flow-offloading is affected. What is definitely affected are network schedulers which rely on flow-information (such as fq_codel or cake)
Quoted 1 line
Hence maybe it should go via 'net' tree rather than via 'net-next'?
I was unsure where to send, i can resend it on net if it is more appropriate.
Best
David
Quoted 25 lines
The fix itself looks fine to me.
Quoted 22 lines
--- net/l2tp/l2tp_eth.c | 3 +++ 1 file changed, 3 insertions(+)net/l2tp/l2tp_eth.c +3
-0@@ -127,6 +127,9 @@ static void l2tp_eth_dev_recv(struct l2tp_session *session, struct sk_buff *skb,/* checksums verified by L2TP */ skb->ip_summed =CHECKSUM_NONE ; +/* drop outer flow-hash */ +skb_clear_hash (skb); +skb_dst_drop (skb);nf_reset_ct (skb);Signature · 1 line
2.43.0
James Chapman
Acked-byQuoted 22 lines · David Bauer
Hi Daniel,
Quoted 10 lines · Daniel Golle
Quoted 9 lines · David Bauer
Drop the flow-hash of the skb when forwarding to the L2TP netdev.
This avoids the L2TP qdisc from using the flow-hash from the outer packet, which is identical for every flow within the tunnel.
This does not affect every platform but is specific for the ethernet driver. It depends on the platform including L4 information in the flow-hash.
Quoted 8 lines
Quoted 4 lines
One such example is the Mediatek Filogic MT798x family of networking processors.
Signed-off-by David Bauer <[email protected]>While it's difficult to say which exact commit this fixes, I still consider it being a fix, as otherwise flow-offloading on mentioned platforms will face difficulties when using L2TP (right?).
Quoted 6 lines
I'm unsure whether flow-offloading is affected. What is definitely affected are network schedulers which rely on flow-information (such as
fq_codelor cake)Quoted 1 line
Hence maybe it should go via 'net' tree rather than via 'net-next'?
I was unsure where to send, i can resend it on net if it is more appropriate.
This has been missing since L2TP ethernet support was first added. Thanks for reporting it!
If you need a Fixes tag, I suggest: d9e31d17ceba5 ( "l2tp: Add L2TP ethernet pseudowire support" )
Quoted 29 lines
Best
DavidQuoted 24 lines
The fix itself looks fine to me.
Quoted 21 lines
---
net/l2tp/l2tp_eth.c| 3 +++1 file changed, 3 insertions(+)net/l2tp/l2tp_eth.c +0
-0@@ -127,6 +127,9 @@ static void l2tp_eth_dev_recv(struct l2tp_session *session, struct sk_buff *skb, /* checksums verified by L2TP */ skb->ip_summed = CHECKSUM_NONE;+ /* drop outer flow-hash */
+skb_clear_hash(skb);
+
skb_dst_drop(skb);
nf_reset_ct(skb);
--
2.43.0