On 12/01/2023 9:29, Martin KaFai Lau wrote:
On 1/11/23 4:32 PM, Stanislav Fomichev wrote:
Please see the first patch in the series for the overall
design and use-cases.
See the following email from Toke for the per-packet metadata overhead:
https://lore.kernel.org/bpf/20221206024554.3826186-1-sdf@xxxxxxxxxx/T/#m49d48ea08d525ec88360c7d14c4d34fb0e45e798
Recent changes:
- Bring back parts that were removed during patch reshuffling from "bpf:
Introduce device-bound XDP programs" patch (Martin)
- Remove netdev NULL check from __bpf_prog_dev_bound_init (Martin)
- Remove netdev NULL check from bpf_dev_bound_resolve_kfunc (Martin)
- Move target bound device verification from bpf_tracing_prog_attach into
bpf_check_attach_target (Martin)
- Move mlx5e_free_rx_in_progress_descs into txrx.h (Tariq)
- mlx5e_fill_xdp_buff -> mlx5e_fill_mxbuf (Tariq)
Thanks for the patches. The set lgtm.
The selftest patch 11 and 17 have conflicts with the recent changes in
selftests/bpf/xsk.{h,c} and selftests/bpf/Makefile. eg. it no longer
needs XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD, so please respin. From a
quick look, it should be some minor changes.
Not sure if Tariq has a chance to look at the mlx5 changes shortly. The
set is getting pretty long and the core part is ready with veth and mlx4
support. I think it is better to get the ready parts landed first such
that other drivers can also start adding support for it. One option is
to post the two mlx5 patches as another patchset and they can be
reviewed separately.
Hi,
I posted new comments.
I think they can be handled quickly, and still be part of the next respin.
I'm fine with both options though. You can keep the mlx5e patches or
defer them to a followup series. Whatever works best for you.
Tariq
