15:01:10 #startmeeting Fedora ARM and AArch64 Status Meeting 15:01:10 Meeting started Tue Jun 14 15:01:10 2016 UTC. The chair is pwhalen. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:10 Useful Commands: #action #agreed #halp #info #idea #link #topic. 15:01:10 The meeting name has been set to 'fedora_arm_and_aarch64_status_meeting' 15:01:10 #chair pwhalen pbrobinson dgilmore hrw dmarlin yselkowitz jonmasters ahs3 15:01:10 Current chairs: ahs3 dgilmore dmarlin hrw jonmasters pbrobinson pwhalen yselkowitz 15:01:17 morning folks, whos here today? 15:01:17 * pbrobinson o/ 15:01:49 * masta lurks about 15:01:54 Good morning.. 15:03:11 ok, lets get started folks, thanks for coming.. 15:03:18 #topic 1) ==== Package Status & Issues ==== 15:03:54 so earlier in the I did the eclipse bootstrap and cleaned up some of the perl rebuild issues 15:04:10 all looking pretty good, no major blockers shown up in rawhide yet 15:04:35 #info Eclipse bootstrap completed. Cleaned up some of the perl rebuild issues. 15:04:49 #info No major blockers shown up in rawhide. 15:04:54 F-24 is looking good, we're pretty much done there, from both ARMv7 and aarch64 doesn't appear to be any userspace blockers or issues that I'm aware of 15:04:56 how are the numbers today? 15:05:09 nor I (blockers on either) 15:05:14 rawhide 15:05:15 statistics: {'older': 116, 'local_only': 1, 'remote_only': 301, 'same': 17698, 'newer': 2, 'total_missing_builds': 169} 15:05:30 f24 15:05:31 statistics: {'older': 21, 'local_only': 1, 'remote_only': 294, 'same': 17742, 'newer': 2, 'total_missing_builds': 57} 15:06:05 rawhide numbers have come down since yesterday when we spoke 15:06:30 #info Rawhide statistics: {'older': 116, 'local_only': 1, 'remote_only': 301, 'same': 17698, 'newer': 2, 'total_missing_builds': 169} 15:06:35 yep 15:06:52 #info F24 statistics: {'older': 21, 'local_only': 1, 'remote_only': 294, 'same': 17742, 'newer': 2, 'total_missing_builds': 57} 15:07:27 awesome, anyone have any other package issues? 15:07:47 I looked very briefly at what's needed for mono boostrap, still some bits missing even though they're upstream in git, I suspect they're releasing off a branch 15:08:06 if anyone is interested in picking up that and running with it do let me know 15:08:48 similarly I don't think there's much needed for D# stack too 15:09:11 or fpc (pascal) 15:09:27 so if someone is interested in a small project let me know 15:09:36 I'm not sure if this is the right time to ask, but I would like to have a discussion about mozjs versions 15:09:47 #info Mono bootstrap still requires some missing bits which are upstream, anyone interested in running with it please contact Peter Robinson. 15:13:07 jlinton: now is fine 15:13:35 Ok, so there are two version of mozjs in fedora 15:13:38 17 and 24 15:13:43 there's about 6 15:13:49 (actually 31 in rawhide too) 15:13:50 pbrobinson, will mono need a side tag like perl? 15:13:54 yah.. 15:13:57 js, mozjs17, 24, 31, 38, 45 15:14:06 ok thats bad 15:14:32 Ok, so first question, is it possible to roll them all forward to a more recent version? 15:14:42 or at least consolidate the users a bit? 15:14:54 aka, pull polkitd to 24 to match gjs? 15:15:00 masta: likely not, due to bootstrap people will need to liaise with me, there'll likely need upstream bugs etc first. That's the end game basically 15:15:49 jlinton: there's a BZ to get polkit using a newer version, if we can do that I suggest we just kill older versions altogether if nothing uses them 15:16:19 jlinton: libproxy I thought use to build against a newer one but it now builds against the ancient js version 15:16:49 jlinton: and i think mongodb uses mozjs38 (or maybe 31) 15:17:33 so if they build there isn't any reason for not pulling everything forward to say 45, and fixing the ragging in that version? 15:17:35 jlinton: basically if we can move the users to the later versions I'd happily assist in killing the older ones 15:17:47 yep, that works for me 15:17:48 ragging->tagging 15:18:43 it would be nice 15:18:58 jlinton: so mozjs45 has issues now on aarh64, even with the upstream 47 bit patch, where the tests fail. If you can help me get that fixed, I'm happy to assist in everything else where I can 15:19:03 Alright, I will look into doing that, because we need it to enable 48 bit VA (49 bit VA in intel speak) 15:19:41 Yah, I can look at that... 15:19:50 jlinton: the RHBZ for polkit is https://bugzilla.redhat.com/show_bug.cgi?id=1061949 15:19:55 its a rats nest but I'm getting better at it. 15:20:24 which patch are you using? 15:20:35 give me a sec 15:21:21 the one attached to this upstream bug that landed upstream https://bugzilla.mozilla.org/show_bug.cgi?id=1143022 15:22:22 The merged on in comment 37? or one of the other patches 15:22:43 jlinton: lets sync on this out of the meeting, I'm not sure offhand 15:23:01 pbrobinson: Yah, sure this could get long.. 15:23:20 jlinton: this is the patch http://paste.fedoraproject.org/379042/46591778 15:23:33 but lets take it offline 15:24:20 thanks for looking into this topic jlinton & pbrobinson 15:25:03 many thanks jlinton 15:25:21 #topic 2) ==== Kernel Status ==== 15:26:01 not much to report here 15:26:19 I think we're looking fine on both ARMv7 and aarch64 for F-24 GA 15:27:01 agreed, 4.7 rc3 in koji for testing 15:27:19 I've not looked closely at 4.7 yet 15:27:29 #info kernel-4.7.0-0.rc3.git0.1.fc25 latest 15:27:39 #link http://koji.fedoraproject.org/koji/buildinfo?buildID=772520 15:27:42 * dgilmore will throw it on some hardware 15:27:48 I've enabled Infiniband today as it was requested and there's now HW that will run it so it makes sense now 15:28:26 so that should actually be built with the next kernel bump 15:28:37 pbrobinson: and big little in rawhide 15:28:48 it will also land with 4.6 when that ends up in F-24 15:28:50 nor I, booted 4.7.0-0.rc2.git3.2.fc25.aarch64 on seattle this morning, doing some f24 install testing, but will also look at the latest kernels as well 15:29:05 dgilmore: yes, we enabled the little big stuff just for LPAE kernels in rawhide 15:29:10 yep 15:29:39 especially so dgilmore could get more than a single core on his Exynos Chromebook runnning a vanilla Fedora kernel :-D 15:29:43 #info Infiniband now enabled in rawhide. 15:30:24 has anyone seen any issues not yet reported or discussed? 15:30:56 dgilmore, does your chromebook show 4 cores or 8 with that kconfig ? 15:31:01 sorry, wrong timezone 15:31:02 #info big.LITTLE enabled in 4.7 kernel for LPAE devices, please report any issues 15:31:11 masta: it shows 8 with 4 active and 4 disabled 15:32:18 #topic 3) ==== Bootloader Status ==== 15:32:38 masta: did you get a chance to do any bootloader/kernel/devices testing with your devices? 15:33:01 #info grub2-2.02-0.34.fc24 now included in F24, please test and report issues. 15:33:29 it's now stable in F-24 15:33:32 pbrobinson, sadly no. It's still on my list, but I wanted to ask if it's too late by now? 15:33:32 Ive done a few installs without issue (qmeu and seattle) 15:33:33 right 15:34:08 masta: never too late 15:34:21 okay =) 15:34:25 masta: you don't get out of it that easily 15:34:31 pwhalen, waht still needs tested on that patch 15:34:43 sorry I'm late - had a furniture delivery of course at exactly the moment this started 15:34:45 catching up 15:35:17 jonmasters_: it's fine, we'd assign you all the work but then well... ;-) 15:35:19 linuxmodder, we wanted to ensure no regressions.. we had some issues with grub2/shim prior. i think we're ok 15:36:14 pbrobinson: love you too :) 15:36:32 :) 15:36:35 pbrobinson: thanks for the IB kernel option, but I'm off topic, so... 15:37:02 #topic 4) ==== F24 Final ==== 15:37:08 jonmasters_: it had already been on my todo list, I just needed to test compile to ensure module filters were right :) 15:37:21 :) 15:37:22 #info Current Test Results 15:37:22 #link https://fedoraproject.org/wiki/Test_Results:Current_Summary 15:37:22 #info Blocker List 15:37:22 #link https://qa.fedoraproject.org/blockerbugs/milestone/24/final/buglist 15:37:47 so, as you are all aware, we slipped due to the grub2 bug, RC 1.2 underway now 15:37:51 so I'll be cutting a RC shortly, I was waiting for dgilmore to fix all the problems first which I think we are almost good for 15:37:58 pwhalen, I have noticed since I tested that back last week on all fc25 kernels I've run (koji-build) wtih valid shim key installed and sec boot on that mokutil reports disabled but thinking it a kernel thing not grub 15:38:15 just need to do a pungi version upgrade and we should be ready to run 15:38:37 so watch #fedora-arm and/or the mailing list for the announcement 15:39:13 primary RC-1.2 is underway 15:40:01 it has a anaconda fix, some compose config changes and pungi logging fixes, and a few FE's to fix broken deps 15:40:30 great! 15:40:35 yep, I've just been reviewing and syncing tags over and making sure everything is signed 15:40:38 cool 15:40:44 will upgrade pungi and pull the trigger 15:41:08 linuxmodder: on what hardware? 15:41:48 pbrobinson, x86_64 and a buddy claimed on his arm but still need to verify his 15:42:10 linuxmodder: well this is an ARM meeting so not relevant 15:42:36 linuxmodder: on arm it's for aarch64 HW that we have grub2/shim, we don't ship it on ARMv7 15:43:05 #info Primary RC-1.2 is underway now. 15:43:22 #info AArch64 compose will be done shortly, watch the list for information. 15:43:45 ok, anything else for f24 final? 15:44:01 #topic 5) == Open Floor == 15:44:59 jlinton: can you join #fedora-arm when you get a moment pls 15:45:04 This is more a kernel config question, but would it be possible to just change aarch64 to use PREEMPT_NONE? 15:45:14 So InfiniBand support. Anybody know how the userland tools are looking for that? 15:45:15 jlinton: why? 15:45:44 well.. The RHEL guys want to run that, because it provides better perf on some benchmarks 15:46:02 and the excuse was that having it in fedora would be nice 15:46:21 since aarch64 at the moment is more a server arch than a desktop one 15:46:25 jlinton: all other arch variants use PREEMPT_VOLUNTARY so we tend to stick with the same across arches where possible 15:46:25 ..... 15:46:30 jlinton: already raised it :) 15:46:34 yah, hence the question 15:47:08 it's in an email I sent to Peter earlier. The way to do this, though, is to wait until 4.8-rc1+ lands in rawhide and we start to get some of the mainstream systems working 15:47:18 jlinton: well at the moment it's more server based, but with devices like Pine64, odriod-c2, Jetson TX1 that's going to be less and less the case 15:47:30 then we can followup and suggest (after that success) that we change this option on e.g. only aarch64 if it makes sense 15:47:44 but pbrobinson makes the obvious counterpoint 15:48:02 Sure 15:48:12 so given that for x86/ppc64/ppc64le/s390x are all on PREEMPT_VOLUNTARY I have a tendency to stay with that 15:48:56 pbrobinson: once there's a Fedora kernel that "just works" on arm64, we can of course easily have a side repo with test kernels with a one line config difference for those who want them too 15:48:56 Would be nice to see some kind of perf metrics with & without PREEMT_WHATEVER 15:49:01 basically it would need to be a more compelling argument than "it's what RHEL does" 15:49:38 it would be nice to see it run time configurable then there wouldn't need to be a discussion :-D 15:50:03 we could always do what Windows does and have systemd coordinate with GNOME3 to prioritize the current window. Clearly XChat needs all of my CPU right now 15:50:23 sorry, I'm done ;) 15:50:33 jonmasters_: firefox always had all my CPU ;-) 15:50:46 :) 15:51:17 I personally hope to see an aarch64 chromebook sometime soon, and kinda want to keep PREEMPT_VOLUNTARY 15:51:18 can we just take a moment to call out PCIe and NUMA support landing in 4.8 15:51:24 jonmasters_: and with v48 it'll be multithreaded so it can have all core 15:51:54 I think that's going to /rock/. With a couple more cleanups (quirks for systems like X-Gene) we'll finally have upstreamish kernels that work on arm servers 15:52:13 jonmasters_: all I can say is ABOUT FRICKEN TIME! 15:52:14 only 3 years too late and massively embarrassing, but better late than never! 15:52:24 right anyone else other than jonmasters_ have anything else? 15:52:42 going once.. 15:53:06 one footnote 15:53:40 there will be a couple more folks from ARM semiconductor companies starting to pop up in #fedora-arm and on the mailing list, say hi when they do :) 15:53:52 jonmasters_: cool, will do 15:54:42 nice :) 15:54:50 #endmeeting