17:01:37 #startmeeting FESCO (2023-05-30) 17:01:37 Meeting started Tue May 30 17:01:37 2023 UTC. 17:01:37 This meeting is logged and archived in a public location. 17:01:37 The chair is sgallagh. Information about MeetBot at https://fedoraproject.org/wiki/Zodbot#Meeting_Functions. 17:01:37 Useful Commands: #action #agreed #halp #info #idea #link #topic. 17:01:37 The meeting name has been set to 'fesco_(2023-05-30)' 17:01:37 #meetingname fesco 17:01:37 The meeting name has been set to 'fesco' 17:01:37 #chair nirik, decathorpe, zbyszek, sgallagh, mhroncok, dcantrell, music, mhayden, Conan_Kudo, Pharaoh_Atem, Son_Goku, King_InuYasha, Sir_Gallantmon, Eighth_Doctor 17:01:37 #topic init process 17:01:37 Current chairs: Conan_Kudo Eighth_Doctor King_InuYasha Pharaoh_Atem Sir_Gallantmon Son_Goku dcantrell decathorpe mhayden mhroncok music nirik sgallagh zbyszek 17:01:46 .hello2 17:01:49 .hello2 17:01:49 zbyszek: zbyszek 'Zbigniew Jędrzejewski-Szmek' 17:01:52 dcantrell: dcantrell 'David Cantrell' 17:01:54 .hello2 17:01:55 mhayden: mhayden 'Major Hayden' 17:02:02 .hello ngompa 17:02:03 Eighth_Doctor: ngompa 'Neal Gompa' 17:02:06 .hi 17:02:07 sgallagh: sgallagh 'Stephen Gallagher' 17:03:51 morning 17:05:58 Are we starting? 17:06:33 #topic #2993 Change: Increase vm.max_map_count value 17:06:33 .fesco 2993 17:06:34 sgallagh: Issue #2993: Change: Increase vm.max_map_count value - fesco - Pagure.io - https://pagure.io/fesco/issue/2993 17:07:32 I've got to run out and get a package in a bit (before the top of the next hour sometime), just FYI 17:08:40 I've been trying to follow the discussion on this one, but honestly I'm not sure I understand the potential ramifications of enlarging this value. 17:09:23 aleasto wrote that with 1 million, the reproducer doesn't work. 17:09:35 I think the lower proposed limit is much more reasonable 17:09:42 But I'd like to see this verified independently. 17:09:46 Right, but also that more intensive games might still be affected. 17:09:50 I feel like if this was a good general purpose default, it would already be the default 17:10:31 dcantrell: I don't necessarily agree. This default could have been decided on when memory never got bigger than the 32-bit limit for all I know. 17:10:49 dcantrell: not necessarilly. Many defaults in the kernel are useless. The kernel relies on userspace to change them, so that it can "keep compatibility". 17:10:49 * on when total system memory never 17:11:03 that's true, but I also don't know that. I'm just guessing. I'd like to see a little more concrete data behind the setting and/or test results 17:11:18 Same here. 17:11:31 OK, but let's at least provide feedback on exactly what data we want to see 17:11:47 hey o/ sorry, lost track of time for a few minutes. 17:12:56 I'd like to see maybe a half dozen to a dozen people in Fedora volunteer to run with the proposed setting on a variety of systems and just report what happens--if anything 17:12:58 I'm ok with just doing the 1million setting, but I agree it would be better to have more people test that 1million works for their games, doesn't allow for easy DOS? 17:13:47 nirik: Resource-based DoS on a Fedora system to which you have login access is so trivial as to not be worth calling out in this specific instance, IMHO 17:14:05 +1 to both 17:14:32 (sorry, I meant to what dcantrell and nirik wrote above) 17:14:39 sure, understood... but this could open remote stuff too in theory 17:14:42 Or do you see a way to exploit this remotely? 17:15:00 well, some thoughts, but nothing concrete... 17:15:35 I'd be ok asking for what dcantrell said. 17:16:08 The problem is that will be a self-selecting group 17:16:18 Which therefore produces anecdotes, not data 17:16:51 I'm frankly in favor of setting the 1M limit in Rawhide and seeing who complains over the next couple weeks. 17:16:53 don't we want to see how this change affects daily users? 17:17:04 We can roll it back if it doesn't go well. 17:17:25 do enough people run rawhide as a daily driver for us to get useful feedback? 17:18:20 dcantrell: I don't think this will impact anyone negatively, unless they go out of their way to run something which creates those mappings. 17:19:17 I'd be fine with approving this for 1m if we agree to revert if people find scenarios in which the limit can be easily exploited. 17:19:28 * nirik too 17:19:38 I.e. approve this without extensive testing, but if testing happens later, be ready to undo the change. 17:19:48 zbyszek: My thoughts exactly 17:19:50 I'm fine with Stephen Gallagher's idea 17:20:45 (that is, going ahead and setting the 1M limit in Rawhide and watching things go from there) 17:20:59 OK, I'll put it to a vote: 17:21:42 Proposal: Set the limit to 1M in Rawhide. We will revert the change if serious, practical exploits are discovered. 17:21:51 +1 17:21:54 +1, wfm 17:22:13 +1 17:22:18 +1 17:22:25 0, not in favor of this approach but I won't stand in the way 17:22:30 1M == 1024*1024 or 1E6? 17:22:49 I left that intentionally as an exercise for the implementor 17:23:00 (I don't know if there are base-2 alignment considerations) 17:23:06 right 17:23:19 1MiB please. This is memory, after all. 17:23:28 ugh 17:24:20 so 2**20? 17:24:34 minus 5? or whatever the current limit uses ... 17:25:07 Sure. 17:25:23 OK, I'll revise the proposal to be explicit 17:25:48 Proposal: Set the limit to 1048576 in Rawhide. We will revert the change if serious, practical exploits are discovered. 17:26:25 Still +1. 17:26:39 +1 17:26:48 +1 17:26:56 0 17:28:00 nirik: drop the package, come back to us! 17:28:01 uaaaa. I was watering some plants in the garden and forgot it's time for the meeting. sorry about that 17:28:05 +1 17:28:44 +1 17:28:54 +1 17:29:16 #agreed Set the limit to 1048576 in Rawhide. We will revert the change if serious, practical exploits are discovered. (+7, 1, -0) 17:29:47 #topic Next week's chair 17:29:58 I can do ity 17:30:01 *it 17:30:03 Thanks! 17:30:22 but not the week after, in case we cancel 17:30:26 #action mhroncok to chair 2023-06-06 meeting 17:30:40 (I will ask asynchronously if that happens) 17:31:08 Ack 17:31:14 #topic Open Floor 17:31:16 #info Voting for Council, FESCo, and Mindshare is open. 17:31:30 * sgallagh opens the floor to reveal the sharks with laser beams on their heads 17:32:19 * zbyszek is glad that his floor is firmly affixed to … whatever is under the floor. 17:33:02 Indeed, please vote for your preferred candidates in the Elections. 17:33:09 And also for me, if you don't mind ;-) 17:33:40 FYI, koji outage thursday... give everyone time to do things other than builds. ;) 17:33:58 #info Koji outage on Thursday. Make other plans! 17:34:54 I'm hopeful the postgresql upgrade will make things faster/better/stronger 17:35:26 That would be pleasant 17:35:35 Any other topics for today? 17:35:57 nirik: are you quoting Daft Punk lyrics? ;) 17:36:34 uhh am i late to anything 17:36:42 .hello2 17:36:43 aleasto: aleasto 'Alessandro Astone' 17:37:35 hi aleasto 17:37:49 we've decided you can do your thing already :) 17:38:09 and now would like your votes in the fesco elections going on now :P 17:38:22 Fabio Valentini: not intentionally. ;) 17:38:44 ok. so 1048576 it is 17:38:57 Wasn't "better, stronger, faster" originally the Six Million Dollar Man or something? 17:39:04 * nirik goes to save his battery from the local post office. 17:39:21 Stephen Gallagher: I think so...? 17:39:33 sgallagh: yeah 17:39:46 Anyway. I think we've reached the end of productive conversation, so I will tie this up. 17:39:56 Thanks for coming, all 17:39:58 #endmeeting