On Fri, 31 May 2019 19:18:17 +0000, Saeed Mahameed wrote:

On Fri, 2019-05-31 at 11:42 +0200, Björn Töpel wrote:

quoted

From: Björn Töpel <redacted>

All XDP capable drivers need to implement the XDP_QUERY_PROG{,_HW}
command of ndo_bpf. The query code is fairly generic. This commit
refactors the query code up from the drivers to the netdev level.

The struct net_device has gained two new members: xdp_prog_hw and
xdp_flags. The former is the offloaded XDP program, if any, and the
latter tracks the flags that the supplied when attaching the XDP
program. The flags only apply to SKB_MODE or DRV_MODE, not HW_MODE.

The xdp_prog member, previously only used for SKB_MODE, is shared
with
DRV_MODE. This is OK, due to the fact that SKB_MODE and DRV_MODE are
mutually exclusive. To differentiate between the two modes, a new
internal flag is introduced as well.  

Just thinking out loud, why can't we allow any combination of
HW/DRV/SKB modes? they are totally different attach points in a totally
different checkpoints in a frame life cycle.

FWIW see Message-ID: [ref]

Down the road i think we will utilize this fact and start introducing
SKB helpers for SKB mode and driver helpers for DRV mode..

Any reason why we would want the extra complexity?  There is cls_bpf
if someone wants skb features after all..