15:00:07 #startmeeting Fedora ARM and AArch64 Status Meeting 15:00:07 Meeting started Tue May 16 15:00:07 2017 UTC. The chair is pwhalen. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:07 Useful Commands: #action #agreed #halp #info #idea #link #topic. 15:00:07 The meeting name has been set to 'fedora_arm_and_aarch64_status_meeting' 15:00:13 morning folks, whos here today? 15:00:42 * pbrobinson o/ 15:00:53 * rric o/ 15:00:54 hi 15:00:57 #chair pwhalen pbrobinson jlinton rric 15:00:57 Current chairs: jlinton pbrobinson pwhalen rric 15:01:09 Morning 15:01:11 morning folks, hope all is well. lets give it a couple for others 15:01:49 jlinton: btw I pushed libproxy 0.4.15 today with support for mozjs38 15:02:05 * alciregi hi 15:02:26 welcome alciregi 15:02:29 Nice, I was going to clean up those patches, because my github rebase was a bit of a mess, but I guess they cleaned them up when they accepted the pull request 15:02:36 :-) 15:02:46 jlinton: did you happen to have patches for erlang-js, it's the last one on 24 15:03:13 No, I didn't do erlang, but I was in contact with the guy who was porting it a few months ago. 15:04:00 Although, it seems by the time we land everything on 38 gnome is going to be on 52! 15:04:08 hola 15:04:47 ok, lets get started 15:04:53 #topic 1) ==== Userspace Status ==== 15:05:05 #info Mass rebuild needed for subset of packages on armhfp and aarch64 due to ABI change in GCC. 15:05:17 #link https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/thread/NPYDADI36HCPI6A5UKWCBRKVZRB3AE35/ 15:05:26 that's running, I think it should all land in today 15:05:27 #info Special thanks to the tools team for doing the test rebuild to get a list of affected packages. 15:05:27 #info Rebuild in progress by Release Engineering. 15:06:03 this was sent to the mailing list 15:06:08 #info Upstream support for RAS in aarch64, do we lack any userspace packages in Fedora? 15:06:30 wondering if folks here have looked at it at all 15:06:40 if rasdaemon is built... 15:07:02 I only remember that it was on ftbfs list years ago 15:07:10 hrw: yes, does it need arch dep patches? 15:07:32 yeah, a couple minor ones, but only for new message types 15:07:42 pbrobinson: do not remember now 15:08:02 patches have been sent upstream, but no comments on them so far 15:08:30 ahs3_: do you happen to have a link to current upstream, package lists fedorahosted which has been retired 15:08:47 virt-manager & libvirt->qemu have issues because of the gicv2m-v3 support 15:08:55 pbrobinson: i *think* so... /me looks 15:09:52 I opened that defect against x86 though because it happens in emulation mode too 15:10:07 jlinton, I hadnt noticed that yet, link? and thanks 15:10:19 https://bugzilla.redhat.com/show_bug.cgi?id=1449837 15:10:49 basically it defaults to gicv3, but that only works on aarch64 machines with an actual gicv3 15:10:53 * jonmasters is in, sorry I'm late 15:11:00 so the stack is basically broken everywhere else 15:11:22 jlinton: is there a way for it just to autodetect based on underlying hypervisor/HW? 15:11:37 yes 15:11:43 Yah, I have a "hack" which falls back to gicv2m if the v3 ITS cannot be created 15:11:48 it's done in RHEL that way 15:11:52 ok, thanks jlinton 15:12:09 but from the comments in the defect they want to just fix libvirt/etc to understand the gic so it can be selected by the user. 15:12:47 why would a user want to change the underlying GIC? Not opposed, but it's kinda silly, anyway... 15:13:19 jonmasters: feel free to comment on the bug, the virt team probably have a better answer 15:13:19 pbrobinson: try the secondary at http://git.infradead.org/users/mchehab/rasdaemon.git -- i also have a copy at https://pagure.io/rasdaemon 15:14:12 iirc only x-gene1 is gicv2m and most of people tries to wipe that cpu out of memory 15:14:35 hrw: Seattle & Juno too 15:14:46 of course all the new stuff is v3 15:15:13 ahs3_: thanks, I also found a Linaro page https://wiki.linaro.org/LEG/Engineering/Kernel/RAS/RAS_daemon 15:15:18 the real problem is the missing qemu v3 emulation, if that existed it would be a non issue too 15:15:20 jlinton: amd a1100 is something I already wiped from my memory. juno never went there ;D 15:15:52 So, TCG is v2m as well 15:15:52 looking at that we have all the required kernel bits enabled, someone just needs to ensure all the userspace bits are upstream, jonmasters anything else other than rasdaemon needed? 15:16:23 jlinton: all the new "enterprise" stuff is v3 ;-) 15:16:52 sorry, juggling calls, catching up 15:17:07 oh for RAS? 15:17:18 jonmasters: yep 15:17:23 once Tyler's patches are upstream, we should be close 15:17:32 I need to check but I don't think his got pulled yet 15:17:43 pbrobinson: i'll take a look at rasdaemon (i'm maintaining it in Debian, too) 15:17:52 I've asked Qualcomm, Cavium, and AppliedMicro to help out here with that btw 15:18:30 ahs3_: thanks, let me know if you need anything 15:18:38 jonmasters++ 15:18:38 pbrobinson: Karma for jcm changed to 1 (for the f25 release cycle): https://badges.fedoraproject.org/tags/cookie/any 15:18:45 tnx 15:18:54 ahs3_: thanks 15:18:56 ahs3_++ 15:19:03 hahahaha 15:19:35 np. i've been keeping an eye on tyler's patches, too. those need to be combined with several others to do the full RAS dance 15:19:53 jonmasters: btw did you make it to the postoffice for pwhalen yet? 15:20:30 lets check in with ras again next week, any other userspace to discuss? 15:20:37 jonmasters: yeah, I will look into ras a bit too 15:20:40 actual hw implementing RAS would be nice... 15:20:53 thanks rric 15:21:16 firefox is probably the other heads up shit show we should at least touch on 15:21:38 true 15:22:19 I've been running FF on/off on f26, is there a specific problem? 15:22:31 at least it starts up now... 15:22:32 I'm setting up an armhfp vm for him to use 15:22:37 https://bugzilla.redhat.com/show_bug.cgi?id=1443938 15:23:02 https://bugzilla.redhat.com/show_bug.cgi?id=1450995 15:23:30 rric: thanks 15:23:43 well it doesn't build on aarch64 (and other arches) so gets regularly disabled https://koji.fedoraproject.org/koji/taskinfo?taskID=19586102 15:24:06 basically the maintainer doesn't even appear to try any more 15:25:24 so it might be useful if people could take a look and provide input on the bugs 15:27:04 #info Assistance with Firefox build failures on arm greatly appreciated. 15:27:26 Yah the upstream defect mentions that the gcc6 flags wern't updated like the gcc7 flags were. I'm pretty sure the problem the flags fixed was present in gcc6 as well. 15:28:07 Although, I haven't been tracking the state of the patches to fix the initializers 15:30:43 topic 2) ==== Kernel Status ==== 15:30:44 #info 4.11.0-1.fc26 does not boot on aarch64 due to miscompilation, required a rebuild with the latest GCC. Fixed with 4.11.0-2. 15:30:53 #info Latest in F26: kernel-4.11.1-300.fc26 - https://koji.fedoraproject.org/koji/buildinfo?buildID=892489 15:30:54 #info Latest in Rawhide: kernel-4.12.0-0.rc1.git0.1.fc27 https://koji.fedoraproject.org/koji/buildinfo?buildID=892258 15:31:41 so far, kernel-4.11.1-300.fc26 looks ok. just started testing .. havent tried 4.12 rc1 at all. is anyone seeing any issues? 15:31:58 pbrobinson: I have some patches for 4.11, is there a deadline for inclusion? 15:32:20 rric: there's always a deadline, depends on what you need, beta was last week 15:32:23 (I know the sooner the better) 15:33:12 rric: the answer is always here https://fedoraproject.org/wiki/Releases/26/Schedule (see freezes, plus time needed for the package itsefl to be tested) 15:33:55 pbrobinson, thanks 15:34:05 rric: so what patches do you need? 15:34:15 The kernel in F26-20170514 seems to die when run in qemu/TCG 15:34:37 I haven't done a full investigation because I was trying to get other things working. 15:34:38 pbrobinson, I need to check, we have some addon patches for 4.11 15:35:07 So, possible PEBCAK (as always) 15:35:23 jlinton, thats kernel-core-4.11.0-2 15:36:28 pbrobinson: some backports from 4.12 + smmuv3 fixes 15:37:26 rric: as long as it's not vast, can you get the list to me today or tomorrow? 15:37:34 pbrobinson: will backport and give you pointer to the patches 15:37:40 jlinton, will test that now 15:37:45 pbrobinson: tomorrow will work 15:37:51 pbrobinson: thanks 15:38:10 rric: I don't want a 10K line diff ;-) 15:38:32 9999 lines are fine 15:38:37 pbrobinson: no, it is very isolated 15:38:46 rric: cool 15:39:08 any other kernel business? 15:39:09 59 files changed, 3528 insertions(+), 279 deletions(-) 15:39:17 but most is the edac driver from 4.12 15:40:12 rric: so a full new driver? 15:40:45 pbrobinson, yes, already upstream 15:40:55 pbrobinson, hope that is ok 15:42:26 rric: I'll review when I get it 15:42:55 #topic 3) ==== Bootloader Status ==== 15:43:48 so u-boot 2017.05 should land into F-26 later, already in f27 15:44:18 it's working (with a few manual commands) for booting RPi3/Pine64 using uEFI -> grub2 -> kernel 15:44:44 anyone who wishes to play with that from a low level PoV feel free to ping me 15:45:07 does anyone have anything else to add? 15:45:17 I've not announced it widely as I don't really want every damn user pinging me because it doesn't work OOTB 15:45:26 good plan :) 15:45:42 it does work very well once up, just a small pain to boot each time 15:45:55 it has arm-trusted-firmware for the Pine64, will expand it out as I get time 15:46:29 #topic 4) ==== F26 Beta ==== 15:46:44 Beta freeze is today 15:46:55 release is scheduled for May 30th 15:47:22 #info F26 Beta Freeze today (11may17), release scheduled for May 30th. 15:47:54 #info Latest nominated compose - Fedora_26_Branched_20170513.n.0 15:47:54 #link https://fedoraproject.org/wiki/Test_Results:Fedora_26_Branched_20170513.n.0_Installation 15:47:54 #link https://fedoraproject.org/wiki/Architectures/AArch64/QA/Fedora-26-20170513.n.0 15:48:37 #info Full list of blockers: https://qa.fedoraproject.org/blockerbugs/milestone/26/beta/buglist 15:49:02 our initial-setup issue is on the way to being fixed. pending new build with the changes 15:49:52 has anyone hit issues not listed on the blocker bugs page? 15:50:44 there's also the CURL issues, but that should be fixed RSN as the dracut fix is headed to stable 15:50:47 * jonmasters will do another round of testing 15:50:50 sorry I'm behind on that 15:50:56 haven't seen anything new 15:51:16 the curl bug is a real pain for testing 15:52:00 Btw, for those who didn't see, they python3 change in Fedora means that if you're running Beaker you need a new anamon3 patch 15:52:42 #info Please test and report any issues to the list or #fedora-arm. If you open a bug, please link to the ARMTracker. 15:53:33 #topic 5) == Open Floor == 15:53:49 any other business for open floor? 15:54:35 jonmasters: only you ;-) 15:55:27 for those interested I now have a pair of rockchips boards (one ARMv7 and one aarch64) courtesy of Rockchips to help improve the support 15:55:40 so I should with luck get some time to look at them soon 15:56:36 #endmeeting