Received: from mga06.intel.com (mga06.intel.com [134.134.136.31]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 6F35B3FC7 for ; Thu, 19 Aug 2021 23:19:23 +0000 (UTC) X-IronPort-AV: E=McAfee;i="6200,9189,10081"; a="277692159" X-IronPort-AV: E=Sophos;i="5.84,336,1620716400"; d="scan'208";a="277692159" Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by orsmga104.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 19 Aug 2021 16:19:18 -0700 X-IronPort-AV: E=Sophos;i="5.84,336,1620716400"; d="scan'208";a="522644793" Received: from seedmund-mobl.amr.corp.intel.com ([10.212.221.18]) by fmsmga003-auth.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 19 Aug 2021 16:19:17 -0700 Date: Thu, 19 Aug 2021 16:19:16 -0700 (PDT) From: Mat Martineau To: Matthieu Baerts cc: Florian Westphal , mptcp@lists.linux.dev Subject: Re: [PATCH mptcp] selftest: wait before trasnmitting first data chunk In-Reply-To: Message-ID: <571b2e3a-cd36-368a-4184-8cf42eb33b5@linux.intel.com> References: <20210819130040.30228-1-fw@strlen.de> Precedence: bulk X-Mailing-List: mptcp@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII; format=flowed On Thu, 19 Aug 2021, Matthieu Baerts wrote: > Hi Florian, > > On 19/08/2021 15:00, Florian Westphal wrote: >> Squashto: selftests: mptcp: add mptcp getsockopt test cases >> >> Mathieu reports spurious failure of the selftest: >> mptcp_sockopt.c:362: do_getsockopt_tcp_info: Assertion `ti.ti[0].tcpi_bytes_received == r' failed. >> mptcp_sockopt.c:330: do_getsockopt_mptcp_info: Assertion `s->mi.mptcpi_rcv_nxt + r == i.mptcpi_rcv_nxt' failed. >> >> This is because client and server are racing. >> The first getsockopt() for (MP)TCP_INFO might be called when some data >> was already received, but the mptcpi_rcv_nxt assertion relies on the >> initial call to fetch a 'start sequence number'. >> >> Extend the pipe (that was used to unblock the 'connect process' when >> the server had bound a socket), it now tells when to connet and when >> to send the first chunk of data. >> >> In addition, its remote chance, but possible, that tx is slower >> than expected so allow tcpinfo to wait a bit for more data to arrive. >> >> After this, 20k invocations pass without failures. > > Thank you for this patch! It looks good to me! > > It is also passing multiple times without issues on my side with and > without a debug kernel! > Looks good to me as well. Ok to squash with "selftests: mptcp: add mptcp getsockopt test cases"? -- Mat Martineau Intel