Re: [PATCH net-next 09/13] selftests: xsk: run the RX and TX endpoints in separate processes

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] selftests: xsk: infinite loop when peer process crashes
--

commit 75a72a39b57bce249cec4dee9c3d6dcc72729c6d
Author: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
selftests: xsk: run the RX and TX endpoints in separate processes

This commit splits the xsk selftest into separate RX and TX processes that
communicate over a TCP control channel. The endpoints exchange fixed-size
messages to synchronize steps, track progress, and abort on failures.

> --- /dev/null
> +++ b/tools/testing/selftests/net/lib/xsk/xsk_peer.c

[ ... ]

> +static int peer_drain(struct xsk_peer *peer)
> +{
> +	struct pollfd pfd = { .fd = peer->fd, .events = POLLIN };
> +	int ret;
> +
> +	while (!peer->closed) {
> +		ret = poll(&pfd, 1, 0);
> +		if (ret < 0 && errno == EINTR)
> +			continue;
> +		if (ret < 0)
> +			return -errno;
> +		if (!ret)
> +			break;
> +		ret = peer_recv_one(peer);
> +		if (ret && !peer->closed)
> +			return ret;
> +	}
> +	return 0;
> +}

[Severity: Medium]
If the peer process crashes, peer_recv_one() detects a closed socket (like
-EPIPE), sets peer->closed to true, and returns the error code. Because
peer->closed is now true, does this if statement suppress the error and
cause the function to return 0 instead of propagating the failure?

> --- a/tools/testing/selftests/net/lib/xsk/test_xsk.c
> +++ b/tools/testing/selftests/net/lib/xsk/test_xsk.c

[ ... ]

>  static int __send_pkts(struct ifobject *ifobject, struct xsk_socket_info *xsk,
>  		       bool test_timeout)
>  {

[ ... ]

>  	buffer_len = pkt_get_buffer_len(umem, pkt_stream->max_pkt_len);
> +	if (pacing_tx_sync())
> +		return TEST_FAILURE;
> +
>  	/* pkts_in_flight might be negative if many invalid packets are sent */
>  	if (pkts_in_flight >= (int)((umem_size(umem) - xsk->batch_size * buffer_len) /
>  	    buffer_len) && !test_timeout) {
>  		ret = kick_tx(xsk);
>  		if (ret)
>  			return TEST_FAILURE;
>  		return TEST_CONTINUE;
>  	}

[Severity: Medium]
When pacing_tx_sync() calls xsk_peer_tx_sync() which relies on
peer_drain(), the suppressed error causes pacing_tx_sync() to incorrectly
report success.

Since no PROGRESS messages are received from the crashed peer,
pkts_in_flight remains artificially high. Will this cause __send_pkts() to
repeatedly return TEST_CONTINUE without ever pushing packets?

[ ... ]

>  static int send_pkts(struct test_spec *test, struct ifobject *ifobject)
>  {

[ ... ]

>  	while (!(all_packets_sent(test, bitmap))) {
>  		for (i = 0; i < test->nb_sockets; i++) {

[ ... ]

>  			ret = __send_pkts(ifobject, &ifobject->xsk_arr[i], test->poll_tmout);
>  			if (ret != TEST_CONTINUE)
>  				return ret;
>  
> -			if (test->fail)
> -				return TEST_FAILURE;
> -
>  			if (!test->poll_tmout) {
>  				ret = wait_for_tx_completion(&ifobject->xsk_arr[i]);
>  				if (ret)
>  					return TEST_FAILURE;
>  			}
>  		}
>  	}
>  
>  	return TEST_PASS;
>  }

[Severity: Medium]
Because __send_pkts() returns TEST_CONTINUE over and over, and
wait_for_tx_completion() returns instantly when no packets are sent, does
this create an infinite loop?

If a test endpoint crashes, it appears the other endpoint will hang
indefinitely instead of failing the test, which could permanently block
automated CI pipelines.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261001202124.692591-1-maciej.fijalkowski@intel.com?part=9