Re: [PATCH net-next 10/13] selftests: xsk: add a hardware mode to xskxceiver

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] selftests/xsk: malformed IP/UDP headers for multi-fragment verbatim packets in hw mode
--

commit 13a458bdc3c40c143ae3519294c27c39c89b28b0
Author: Maciej Fijalkowski <maciej.fijalkowski@intel.com>

selftests: xsk: add a hardware mode to xskxceiver

This commit introduces a hardware mode (--hw) to xskxceiver for running
AF_XDP tests over a physical link. It wraps packets in UDP/IPv4 headers and
adds new command line options to configure hardware runs.

> diff --git a/tools/testing/selftests/net/lib/xsk/test_xsk.c b/tools/testing/selftests/net/lib/xsk/test_xsk.c
> --- a/tools/testing/selftests/net/lib/xsk/test_xsk.c
> +++ b/tools/testing/selftests/net/lib/xsk/test_xsk.c
[ ... ]
> @@ -1268,8 +1366,8 @@ static int __send_pkts(struct ifobject *ifobject, struct xsk_socket_info *xsk,
>  				tx_desc->options = 0;
>  			}
>  			if (pkt->valid)
> -				pkt_generate(xsk, umem, tx_desc->addr, tx_desc->len, pkt->pkt_nb,
> -					     bytes_written);
> +				pkt_generate(xsk, umem, tx_desc->addr, tx_desc->len, pkt->len,
> +					     pkt->pkt_nb, bytes_written);

[Severity: Medium]
When generating multi-fragment verbatim packets in hardware mode, does this
pass the incorrect length to pkt_generate()?

If pkt_stream->verbatim is true, pkt->len represents only the length of the
current fragment, not the entire logical packet. Passing it as total_len here
propagates to gen_pkt_hdr(), where it truncates the UDP and IPv4 length fields:

tools/testing/selftests/net/lib/xsk/test_xsk.c:gen_pkt_hdr() {
    ...
    struct udphdr udp = {
        ...
        .len = htons(total_len - sizeof(*eth_hdr) - sizeof(struct iphdr)),
    };
    struct iphdr ip = {
        ...
        .tot_len = htons(total_len - sizeof(*eth_hdr)),
    };
    ...
}

This causes malformed packets with truncated L3/L4 lengths while the physical
frame contains all fragments, which might be dropped by NIC parsers or
switches and lead to spurious hardware test failures (for example, in
testapp_too_many_frags).

Could the total logical packet length be passed here instead of just the
fragment length?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261001202124.692591-1-maciej.fijalkowski@intel.com?part=10