Re: [PATCH v3 net-next 02/15] selftests: xsk: size the TX UMEM in HW_SW_MAX_RING_SIZE

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] selftests: xsk: infinite loop if hardware TX ring > 8200
--

commit 183d9676e829720fbcf94af5e89e100b2685ba42
Author: Maciej Fijalkowski <maciej.fijalkowski@intel.com>

selftests: xsk: size the TX UMEM in HW_SW_MAX_RING_SIZE

This patch configures the TX UMEM sizes in the testapp_hw_sw_max_ring_size
selftest to match the RX UMEM sizing. It addresses an issue where TX
would fail in two-interface mode due to the default UMEM being too small
for the batch sizes used in this test.

> diff --git a/tools/testing/selftests/bpf/prog_tests/test_xsk.c b/tools/testing/selftests/bpf/prog_tests/test_xsk.c
> index 2a13628740924..16fe32e5afc06 100644
> --- a/tools/testing/selftests/bpf/prog_tests/test_xsk.c
> +++ b/tools/testing/selftests/bpf/prog_tests/test_xsk.c
> @@ -2513,6 +2513,9 @@ int testapp_hw_sw_max_ring_size(struct test_spec *test)
>  	test->ifobj_rx->xsk->umem->num_frames = max_descs;
>  	test->ifobj_rx->xsk->umem->fill_size = max_descs;
>  	test->ifobj_rx->xsk->umem->comp_size = max_descs;
> +	test->ifobj_tx->xsk->umem->num_frames = max_descs;
> +	test->ifobj_tx->xsk->umem->fill_size = max_descs;
> +	test->ifobj_tx->xsk->umem->comp_size = max_descs;

[Severity: Medium]
Since max_descs is fixed to 8192 (XSK_RING_PROD__DEFAULT_NUM_DESCS * 4), can
this cause an infinite loop in __send_pkts() on hardware with large rings?

Further down in testapp_hw_sw_max_ring_size(), the batch size is explicitly set
based on the raw hardware maximums:

    test->ifobj_tx->xsk->batch_size = test->ifobj_tx->ring.tx_max_pending - 8;

If a NIC supports a maximum TX ring size greater than 8200 (such as some mlx5
or bnxt cards), the calculated batch_size will strictly exceed the software
ring capacity (max_descs = 8192) configured here.

When that happens, __send_pkts() will enter an infinite loop:

    while (xsk_ring_prod__reserve(&xsk->tx, xsk->batch_size, &idx) < xsk->batch_size) {
        ...
        complete_pkts(xsk, xsk->batch_size);
    }

Because xsk_ring_prod__reserve() can never grant a reservation larger than the
total ring capacity, this loop will spin indefinitely waiting for space that
can never exist.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261014184539.774708-1-maciej.fijalkowski@intel.com?part=2