When switching network namespaces with the bpf_redirect_peer helper, the skb->mark and skb->tstamp fields are not zeroed out like they can be on a typical netns switch. This patch clarifies that in the helper description.
Showing the changes to include/uapi/linux (1 of 2 files). Show all
Unified · Split
| @@ -4968,6 +4968,9 @@ union bpf_attr { · 4967 unmodified lines | |||
| 4968 | * the netns switch takes place from ingress to ingress without | 4968 | * the netns switch takes place from ingress to ingress without |
| 4969 | * going through the CPU's backlog queue. | 4969 | * going through the CPU's backlog queue. |
| 4970 | * | 4970 | * |
| 4971 | * *skb*\ **->mark** and *skb*\ **->tstamp** are not cleared during | ||
| 4972 | * the netns switch. | ||
| 4973 | * | ||
| 4971 | * The *flags* argument is reserved and must be 0. The helper is | 4974 | * The *flags* argument is reserved and must be 0. The helper is |
| 4972 | * currently only supported for tc BPF program types at the | 4975 | * currently only supported for tc BPF program types at the |
| 4973 | * ingress hook and for veth and netkit target device types. The | 4976 | * ingress hook and for veth and netkit target device types. The |