Re: [PATCH v7 1/2] bpf, sockmap: handle spurious tcp_msg_wait_data() wakeup

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] tcp_bpf_recvmsg_parser() fails to handle unparsed data in sk_receive_queue, returning premature EAGAIN on blocking sockets.
--

--- Patch [1]: [PATCH v7 1/2] bpf, sockmap: handle spurious tcp_msg_wait_data() wakeup ---
commit 7468e0f3ba324455b02b600d4c77965cd4caa92c
Author: Nnamdi Onyeyiri <nnamdio@gmail.com>

bpf, sockmap: handle spurious tcp_msg_wait_data() wakeup

This commit addresses spurious wakeups from tcp_msg_wait_data() by looping
back to msg_bytes_ready instead of returning -EAGAIN. It also passes a pointer
for timeo to allow sk_wait_event() to update the remaining timeout.

> diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c
> index 8e905b50deadb..87aef99a13792 100644
> --- a/net/ipv4/tcp_bpf.c
> +++ b/net/ipv4/tcp_bpf.c
[ ... ]
> @@ -310,13 +311,15 @@ static int tcp_bpf_recvmsg_parser(struct sock *sk,
>  			goto out;
>  		}
>  
> -		data = tcp_msg_wait_data(sk, psock, timeo);
> +		data = tcp_msg_wait_data(sk, psock, &timeo);
>  		if (data < 0) {
>  			copied = data;
>  			goto unlock;
>  		}
>  		if (data && !sk_psock_queue_empty(psock))
>  			goto msg_bytes_ready;
> +		if (!data && timeo > 0)
> +			goto msg_bytes_ready;

[Severity: High]
This is a pre-existing issue, but could this return a premature -EAGAIN on a
blocking socket when data is waiting to be parsed?

If data arrives and is added to sk_receive_queue, but a stream parser requires
more bytes to complete a message (or the parser is deferred to a workqueue),
tcp_msg_wait_data() will return 1.

Because psock->ingress_msg is still empty, both the existing condition:
    if (data && !sk_psock_queue_empty(psock))
and the new condition:
    if (!data && timeo > 0)
will evaluate to false.

Does this mean tcp_bpf_recvmsg_parser() falls through and returns -EAGAIN
instead of continuing to wait for the remaining data to be parsed?

>  		copied = -EAGAIN;
>  	}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260721223807.75101-1-nnamdio@gmail.com?part=1