From: "cron2 (Code Review)" Date: Mon, 21 Jul 2025 18:08:07 +0000 In-Reply-To: References: Reply-To: gert@...1296..., arne-openvpn@...1227..., marco@...2726..., openvpn-devel@lists.sourceforge.net, frank@...2641..., lstipakov@...277... Message-ID: <2b33e02a8ac59e8cbed5ce24bb1ac7d5c6f5fe34-EmailReviewComments-HTML@...2715...> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Content-Disposition: inline Content-Type: multipart/alternative; boundary="ul0rUlzOpio="; charset=UTF-8 Subject: [Openvpn-devel] [L] Change in openvpn[master]: PUSH_UPDATE: Allow OpenVPN in client mode to receive and handle PUSH ... List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: mrbff Cc: flichtenheld , stipa , plaisthos , openvpn-devel --ul0rUlzOpio= Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Attention is currently required from: cron2, flichtenheld, mrbff, stipa. cron2 has posted comments on this change. ( http://gerrit.openvpn.net/c/ope= nvpn/+/808?usp=3Demail ) Change subject: PUSH_UPDATE: Allow OpenVPN in client mode to receive and ha= ndle PUSH UPDATE control messages to allow options updating at runtime. ...................................................................... Patch Set 19: Code-Review-1 Copied votes on follow-up patch sets have been updated: * Code-Review-1 has been copied to patch set 20 (copy condition: "changekin= d:NO_CHANGE OR changekind:TRIVIAL_REBASE OR is:MIN"). (10 comments) Patchset: PS19: So, I think this works and is sufficently save and efficient, but I'm and o= ld man and like my lawn orderly... so as discussed on IRC, please adjust a = few minor points. File src/openvpn/options.c: http://gerrit.openvpn.net/c/openvpn/+/808/comment/849fb9c1_8109d087 : PS19, Line 5567: apply_push_options(struct options *options, please do not move that one around in options.c, this makes comparing old/n= ew harder without good benefit, also breaks git blame... http://gerrit.openvpn.net/c/openvpn/+/808/comment/25512e6f_b2eb183f : PS19, Line 5654: so I would call check_push_update_option_flags() from here, possibly moving= the "while isspace(*line) { line++; }" part up here as well. Then we can = keep "apply_pull_filter()" focused on "pull filter", not dealing with a num= ber of extra conditions... (just "is_update", see there) File src/openvpn/options_util.c: http://gerrit.openvpn.net/c/openvpn/+/808/comment/5c5cd074_1394fe2c : PS19, Line 198: if (!line || !*line || !**line) the use of a double pointer irritates me, because we never move the pointer= , instead `memmove()` the buffer... I think we should just do `(*line)++` instead of the `memmove()` below - po= ssibly also getting rid of the `strlen(str)` thing, or `str` altogether :-)= - so keep the pointer, call it from `apply_push_options()` and accept a mo= ving `*line` pointer. Also, I'm not sure if "empty string" should be treated as `false` or just i= gnored (return true, that is) - the existing environment where `apply_pull_= filter()` is used does not care for "empty lines". A NULL pointer is a programming error that must never happen, so an ASSERT(= line) is valid. http://gerrit.openvpn.net/c/openvpn/+/808/comment/6ffca9a9_893ed859 : PS19, Line 231: If there is whitespace here (`? route`) we'll fail with `pushed option is = not updateable: ' route'`, if I'm not misreading it. I'd document that "there must not be whitespace between the flag letters an= d the option" and then we could add an `if (isspace(**line)) { complain; re= turn false; }` here... the documentation is important, the error handling n= ot so much (as it will hit the `return false` anyway) http://gerrit.openvpn.net/c/openvpn/+/808/comment/bc8b42ab_3cf0ba0c : PS19, Line 252: msg(D_PUSH, "Pushed option is not updatable: '%= s'", *line); Maybe add an " Ignoring." to the string, so it's clear "this is information= al, but not an error, and this is how we handle it"? Since the other one h= as "Restarting"... http://gerrit.openvpn.net/c/openvpn/+/808/comment/ce24fecc_c0a39fcd : PS19, Line 270: if (!o->pull_filter_list && !(is_update)) So, if we call `check_push_update_option_flags()` from outside, we do not n= eed this extra dance, checking o->pull_filter_list twice... http://gerrit.openvpn.net/c/openvpn/+/808/comment/8974bb49_5a803112 : PS19, Line 276: while (isspace(*line)) It makes sense to move that upwards to `apply_push_options()`, before calli= ng anything on the line. So both all functions called do not need to bothe= r about leading whitespace. http://gerrit.openvpn.net/c/openvpn/+/808/comment/8689afde_5d9ee877 : PS19, Line 310: return true; This duplication of the messages and `return true` looks a bit silly :-) - = I would suggest to do ``` /* on PUSH_UPDATE, "reject" and "ignore" filters are treated the sa= me */ else if ( (f->type =3D=3D PUF_TYPE_IGNORE || is_update) && strncmp(line, f->pattern, f->size) =3D=3D 0) { msg(D_PUSH, "Pushed option removed by filter: '%s'", line); return true; } ``` File src/openvpn/push_util.c: http://gerrit.openvpn.net/c/openvpn/+/808/comment/b8dd5363_84e249de : PS19, Line 23: true)) Just for the record - this is the really-old-style OpenVPN formatting, whic= h we no longer use. That said, I know it's the same as in `process_incomin= g_push_reply()`, so it makes sense to keep it the same, and then clang-form= at will change both at once. -- To view, visit http://gerrit.openvpn.net/c/openvpn/+/808?usp=3Demail To unsubscribe, or for help writing mail filters, visit http://gerrit.openv= pn.net/settings Gerrit-Project: openvpn Gerrit-Branch: master Gerrit-Change-Id: I6ecd4cb47571cc8c20e46de8595c742aeec6064a Gerrit-Change-Number: 808 Gerrit-PatchSet: 19 Gerrit-Owner: mrbff Gerrit-Reviewer: cron2 Gerrit-Reviewer: flichtenheld Gerrit-Reviewer: plaisthos Gerrit-Reviewer: stipa Gerrit-CC: openvpn-devel Gerrit-Attention: cron2 Gerrit-Attention: flichtenheld Gerrit-Attention: mrbff Gerrit-Attention: stipa Gerrit-Comment-Date: Mon, 21 Jul 2025 18:08:07 +0000 Gerrit-HasComments: Yes Gerrit-Has-Labels: Yes Gerrit-MessageType: comment --ul0rUlzOpio= Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable

Attention is currently required from: cron2, flichtenheld, mrbff, stipa= .

Patch set 19:Code-Review -1

V= iew Change

10 comments:

  • Patchset:

    • <= p style=3D"margin-bottom: 4px;">Patch Set #19:

      So, I think this works and is sufficently = save and efficient, but I'm and old man and like my lawn orderly... so = as discussed on IRC, please adjust a few minor points.

  • File src/openvpn/options.c:

  • File src/openvpn/options_util.c:

    • Patch Set #19, Line 198: if (!line || !*line || !**line)<= /p>

      the use of a = double pointer irritates me, because we never move the pointer, instead `me= mmove()` the buffer...

      I think we should just do `(*line)++` instead of the `memmove()` = below - possibly also getting rid of the `strlen(str)` thing, or `str` alto= gether :-) - so keep the pointer, call it from `apply_push_options()` and a= ccept a moving `*line` pointer.

      Also, I'm not sure if "empty string" shoul= d be treated as `false` or just ignored (return true, that is) - the existi= ng environment where `apply_pull_filter()` is used does not care for "= empty lines".

      A NULL pointer is a programming error that must never happen, so an A= SSERT(line) is valid.

    • Patch Set #19, Line 231:

      If there is whitespace here (`? route`= ) we'll fail with `pushed option is not updateable: ' route'`,= if I'm not misreading it.

      I'd document that "there must not be whitespace = between the flag letters and the option" and then we could add an `if = (isspace(**line)) { complain; return false; }` here... the documentation is= important, the error handling not so much (as it will hit the `return fals= e` anyway)

    • Patch Set #19, Line 252: msg(D_PUSH, "Pushed option= is not updatable: '%s'", *line);

      Maybe add an " Ignoring."= ; to the string, so it's clear "this is informational, but not an = error, and this is how we handle it"? Since the other one has "R= estarting"...

    • Patch Set #19, Line 270: if (!o->pull_filter_list &= & !(is_update))

      So, if we call `check_push_update_option_flags()` from outsid= e, we do not need this extra dance, checking o->pull_filter_list twice..= .

    • Patch Set #19, Line 276: while (isspace(*line))

      It makes sense to move that upwar= ds to `apply_push_options()`, before calling anything on the line. So both= all functions called do not need to bother about leading whitespace.

    • Patch Set #19, Line 310: return true;

      This duplication of the messages and `= return true` looks a bit silly :-) - I would suggest to do

      ```
      /= * on PUSH_UPDATE, "reject" and "ignore" filters are tre= ated the same */
      else if ( (f->type =3D=3D PUF_TYPE_IGNORE ||= is_update)
      && strncmp(line, f->pattern, f= ->size) =3D=3D 0)
      {
      msg(D_PUSH, "Pushed o= ption removed by filter: '%s'", line);
      return t= rue;
      }
      ```
  • File src/openvpn/push_util.c:

    • Patch Set #19, Line 23: true))

      Just for the record -= this is the really-old-style OpenVPN formatting, which we no longer use. = That said, I know it's the same as in `process_incoming_push_reply()`, = so it makes sense to keep it the same, and then clang-format will change bo= th at once.

To view, visit change 808. To unsubscribe,= or for help writing mail filters, visit settings.

Gerrit-Project: openvpn
Gerrit-Branch: master
Gerrit-Change-Id: I6ecd4cb47571cc8c20e46de8595= c742aeec6064a
Gerrit-Change-Number: 808
Gerrit-PatchSet: 19
Gerrit-Owner: mrbff <marco@...2726...> <= /div>
Gerrit-Reviewer: cron2 <gert@...1296...>=
Gerrit-Reviewer: flichtenheld <frank@...2716...= 1...>
Gerrit-Reviewer: plaisthos <arne-openvpn@...1186...= .1227...>
Gerrit-Reviewer: stipa <lstipakov@...277...= >
Gerrit-CC: openvpn-devel <openvpn-devel@...1484...= ts.sourceforge.net>
Gerrit-Attention: cron2 <gert@...1296...>= ;
Gerrit-Attention: flichtenheld <frank@...2375...= 41...>
Gerrit-Attention: mrbff <marco@...2726...&g= t;
Gerrit-Attention: stipa <lstipakov@...1663...= .>
Gerrit-Comment-Date: Mon, 21 Jul 2025 18:08:07= +0000
Gerrit-HasComments: Yes
Gerrit-Has-Labels: Yes
Gerrit-MessageType: comment
--ul0rUlzOpio=--