Re: [PATCH] xfrm: Fix double ESP trailer insertion in IPsec crypto offload.

2020-06-05 Thread Steffen Klassert
On Mon, Jun 01, 2020 at 04:39:37PM -0500, Huy Nguyen wrote: > During IPsec performance testing, we see bad ICMP checksum. The error packet > has duplicated ESP trailer due to double validate_xmit_xfrm calls. The first > call > is from ip_output, but the packet cannot be sent because > netif_xmit_f

Re: [PATCH] xfrm: Fix double ESP trailer insertion in IPsec crypto offload

2020-06-01 Thread Huy Nguyen
PSB On 5/22/2020 7:25 PM, Saeed Mahameed wrote: On Thu, 2020-05-21 at 16:49 -0500, Huy Nguyen wrote: During IPsec performance testing, we see bad ICMP checksum. The issue is that the error packet that has duplicated ESP trailer. For example, this below ping reply skb is collected at mlx5e_xmit.

[PATCH] xfrm: Fix double ESP trailer insertion in IPsec crypto offload.

2020-06-01 Thread Huy Nguyen
During IPsec performance testing, we see bad ICMP checksum. The error packet has duplicated ESP trailer due to double validate_xmit_xfrm calls. The first call is from ip_output, but the packet cannot be sent because netif_xmit_frozen_or_stopped is true and the packet gets dev_requeue_skb. The sec

Re: [PATCH] xfrm: Fix double ESP trailer insertion in IPsec crypto offload

2020-05-22 Thread Saeed Mahameed
On Thu, 2020-05-21 at 16:49 -0500, Huy Nguyen wrote: > During IPsec performance testing, we see bad ICMP checksum. The issue > is that > the error packet that has duplicated ESP trailer. For example, this > below ping reply skb is > collected at mlx5e_xmit. This ping reply skb length is 154 because

[PATCH] xfrm: Fix double ESP trailer insertion in IPsec crypto offload

2020-05-21 Thread Huy Nguyen
During IPsec performance testing, we see bad ICMP checksum. The issue is that the error packet that has duplicated ESP trailer. For example, this below ping reply skb is collected at mlx5e_xmit. This ping reply skb length is 154 because it has extra duplicate 20 bytes of ESP trailer. The correct l