Re: [PATCH net 2/2] selftests/net: add SO_RESERVE_MEM test

[Matthieu Baerts <matttbe@kernel.org>]

Hi Eric,

Thank you for the new test.

On 25/09/2026 14:49, netdev-bot+sashiko@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 6 potential
> issue(s) to consider.

(...)
>> +static void check_non_tcp_rejected(struct __test_metadata *_metadata,
>> +				   int domain, int type, int protocol,
>> +				   int val)
>> +{
>> +	int fd = socket(domain, type, protocol);
>> +
>> +	if (fd < 0) {
>> +		EXPECT_TRUE(errno == EAFNOSUPPORT || errno == EPROTONOSUPPORT);
>> +		return;
>> +	}
> 
> [Severity: Medium]
> Should ENOPROTOOPT be accepted here as well?
> 
> MPTCP can be built in but disabled with net.mptcp.enabled=0. In that case
> socket creation fails here:
> 
> net/mptcp/protocol.c:mptcp_init_sock() {
> 	...
> 	if (!mptcp_is_enabled(net))
> 		return -ENOPROTOOPT;
> 	...
> }
> 
> inet_create() passes that error back to userspace. As a result,
> socket(AF_INET, SOCK_STREAM, IPPROTO_MPTCP) in non_tcp_rejected fails with
> ENOPROTOOPT, and the EXPECT_TRUE fails the test.
> 
> The test doesn't unshare the network namespace, so it sees the host's
> sysctl. Any other environment-specific socket() errno also becomes an
> EXPECT failure rather than a skip.
Typically, we try to run the tests in a dedicated netns to avoid such
issues. (We also force net.mptcp.enabled=1 because it is still disabled
on RHEL, which is really a shame... but I don't think we really need to
support that here in the selftests.)

At the end, I don't think we need to care about all possible cases, but
using a netns is probably a good idea, at least not to "poison" other
tests. (If that's easy to put in place.)

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.