Date: Thu, 18 Aug 2022 21:26:20 +0200 (CEST) Message-Id: <202208181926.27IJQKeU072582@...2054...> From: Gert Doering In-Reply-To: <20220813204224.22576-1-a@...2181...> References: <20220813204224.22576-1-a@...2181...> Subject: [Openvpn-devel] [PATCH applied] Re: dco-win: introduce low-level code for handling ovpn-dco-win in Windows List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 8bit To: Antonio Quartulli Cc: openvpn-devel@lists.sourceforge.net I've stared at the code for a while... I'm not really happy with the jumping back and forth between dco.c and tun.c (who is supposed to understand that code flow in 6 weeks from now?). That said, the "non windows" changes in this patch are harmless enough, and the "windows bits" do look safe enough (wrt memory issues etc). I notice a distinct lack of comments on "what do all these functions do?" - neither dco_win.h nor dco_win.c have function descriptions (what does it do, what goes in, what comes out). Like, dco_connect_wait() or dco_create_socket()... this really should be improved. Some constructs nest too deeply... + DWORD err = GetLastError(); + if (err != ERROR_IO_PENDING) + { + msg(M_ERR, "DeviceIoControl(OVPN_IOCTL_NEW_PEER) failed"); + } + else + { + if (dco_connect_wait(tt.hand, &ov, timeout, signal_received) < 0) + { + close_tun_handle(&tt); + } + } .. M_ERR does not return, so why have an else { } level here? + OVPN_CRYPTO_DATA crypto_data; + ZeroMemory(&crypto_data, sizeof(crypto_data)); .. we do have CLEAR(crypto_data) for this purpose. Do not invent new code for each platform to zeroize memory. + ASSERT(crypto_data.CipherAlg > 0); this looks a bit out of place, 10+ lines after the CipherAlg assignment... Since I currently can't build MinGW, I've pushed this to Github for building, and the GHA-mingw/mingw64 builds succeed (and so does the rest). I've also subjected this to the linux DCO test rig "to be sure" - though *these* changes really should not upset other platforms. Seems they do not :-) My local git hook complains about formatting of ovpn_dco_win.h - we do exclude that now in dev-tools/special-files.lst, but that is not effective on the initial commit. Not sure, though, why we want to maintain something only maintained by us in a different coding style... I can see this for dco-linux (where "the Linux people" want a different style) - but here? Anyway. Your patch has been applied to the master branch. commit 8b80cbc3846a56581e373a664db20d227a90120a Author: Antonio Quartulli Date: Sat Aug 13 22:42:18 2022 +0200 dco-win: introduce low-level code for handling ovpn-dco-win in Windows Signed-off-by: Arne Schwabe Signed-off-by: Lev Stipakov Signed-off-by: Antonio Quartulli Acked-by: Lev Stipakov Message-Id: <20220813204224.22576-1-a@...2181...> URL: https://www.mail-archive.com/openvpn-devel@lists.sourceforge.net/msg24919.html Signed-off-by: Gert Doering -- kind regards, Gert Doering