Re: [PATCH net-next v3 1/4] selftests/xsk: add UMEM users refcount and centralize socket teardown

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Shared UMEM TX sockets fail to create due to partial UMEM view lacking the libbpf UMEM handle.
--

--- Patch [1]: [PATCH net-next v3 1/4] selftests/xsk: add UMEM users refcount and centralize socket teardown ---
commit ec175a3b2fa035fe1d6886d51e8d51aa7aaf1d62
Author: Tushar Vyavahare <tushar.vyavahare@intel.com>

selftests/xsk: add UMEM users refcount and centralize socket teardown

This commit adds a reference count to xsk_umem_info to track shared UMEM
ownership and centralizes the teardown path to ensure sockets clean up
properly across all error paths.

> diff --git a/tools/testing/selftests/bpf/prog_tests/test_xsk.c b/tools/testing/selftests/bpf/prog_tests/test_xsk.c
> index 38ce6060b8fac..fbec8afaadc99 100644
> --- a/tools/testing/selftests/bpf/prog_tests/test_xsk.c
> +++ b/tools/testing/selftests/bpf/prog_tests/test_xsk.c
[ ... ]
> @@ -1508,8 +1529,15 @@ static int thread_common_ops_tx(struct test_spec *test, struct ifobject *ifobjec
>  	}
>  
>  	umem_rx = test->ifobj_rx->xsk_arr[0].umem;
> +	/* Non-owning view used only for TX buffer arithmetic; the sockets below bind to
> +	 * the RX-owned UMEM, so the handle, cq and refcount stay in one place.
> +	 */
>  	umem_tx = ifobject->xsk_arr[0].umem_real;
> -	memcpy(umem_tx, umem_rx, sizeof(*umem_tx));
> +	umem_tx->num_frames = umem_rx->num_frames;
> +	umem_tx->frame_headroom = umem_rx->frame_headroom;
> +	umem_tx->buffer = umem_rx->buffer;
> +	umem_tx->frame_size = umem_rx->frame_size;
> +	umem_tx->unaligned_mode = umem_rx->unaligned_mode;
>  	umem_tx->base_addr = 0;
>  	umem_tx->next_buffer = 0;

[Severity: Medium]
Will this cause shared UMEM TX sockets to fail creation because the libbpf
UMEM handle is not copied over?

Since umem_tx is now constructed manually instead of using memcpy(), the
umem_tx->umem field is left NULL. When xsk_configure_socket() is called later
in thread_common_ops_tx(), it passes this NULL umem->umem pointer to
xsk_socket__create(), which causes it to fail.

Also, does leaving the users refcount zero-initialized on this umem_tx view
conflict with the stated goal of keeping shared sockets pointing to the same
reference count?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260819024443.781764-1-tushar.vyavahare@intel.com?part=1