15:01:04 <pwhalen> #startmeeting Fedora ARM and AArch64 Status Meeting
15:01:04 <zodbot> Meeting started Tue Sep 13 15:01:04 2016 UTC.  The chair is pwhalen. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:04 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:01:04 <zodbot> The meeting name has been set to 'fedora_arm_and_aarch64_status_meeting'
15:01:06 <pwhalen> #chair pwhalen pbrobinson dgilmore hrw dmarlin yselkowitz jonmasters ahs3 msalter
15:01:06 <zodbot> Current chairs: ahs3 dgilmore dmarlin hrw jonmasters msalter pbrobinson pwhalen yselkowitz
15:01:07 * pbrobinson o/
15:01:10 <pwhalen> morning folks
15:01:28 <rric> rric o/
15:02:36 <jlinton> good morning
15:03:42 <pwhalen> lets get started..
15:03:49 <pwhalen> #topic 0) ==== AArch64 merge  ====
15:04:10 <pbrobinson> so we've merged aarch64 into the main koji instance over the weekend
15:04:15 <pbrobinson> builds are now enabled
15:04:17 <yselkowitz> yay!!
15:04:21 <bconoboy> \o/
15:04:40 <pbrobinson> it's been relatively smooth and first composes should come tomorrow
15:04:55 <hrw> o
15:05:00 <pbrobinson> but keep an eye for problems, complaints on list and any build issues with aarch64
15:05:42 <pbrobinson> it'll be tweaked over the next couple of weeks but I'm pretty pleased with the state of the merge
15:05:48 <hrw> pbrobinson: so for f26 there will be koji, s390.koji and riscv.koji?
15:06:14 <yselkowitz> still too early for riscv I imagine
15:06:20 <cov> hi
15:06:33 <yselkowitz> and arm.koji will still be used for branches through f25 right?
15:06:54 <pwhalen> #info Aarch64 merge with Primary Koji complete. First compose should be tomorrow (14sep16).
15:06:57 <pbrobinson> hrw: no idea about riscv.koji, that won't be official (like MIPs)
15:07:17 <hrw> yep
15:07:25 <pbrobinson> hrw: the intention is to merge Power64 mid to late Oct
15:07:42 <pbrobinson> and s390x is indeterminate
15:07:48 <jlinton> ppcle?
15:07:49 <pbrobinson> but off topic here
15:08:01 <pbrobinson> jlinton: Power64 covers both ppc64 and ppc64le
15:08:14 <pbrobinson> jlinton: no 32 bit support on any PPC
15:09:05 * jonmasters in
15:09:24 <jlinton> (OT) thanks, just wondering if it was BE or LE, but having both will be nice!
15:09:39 <bconoboy> jlinton: BE and LE are already merged
15:09:56 <pwhalen> anything else for the merge?
15:09:58 <pbrobinson> jlinton: ppc.koji is both of those
15:10:00 <pbrobinson> nope.
15:10:12 <pwhalen> #topic 1) ==== Userspace Status  ====
15:11:04 <cov> does the merge mean that a build failure on aarch64 will block the package on other archs?
15:11:30 <pbrobinson> cov: yes
15:11:49 <pbrobinson> there's no issues I'm aware of
15:12:01 <rric> question to 48 bits va size, are affected tools fixed (e.g. mozjs) and capable to enable 48 va size in the kernel?
15:12:12 <pwhalen> #info No current package issues.
15:12:12 <pbrobinson> there was an issue with eclipse in the merge but there's a bug and the maintainer is looking at it
15:12:20 <jlinton> I think the mozjs patches are in the defect and getting merged
15:12:42 <pbrobinson> I'm also going to go through and cleanup deps of luajut now we have it and push patches for the various mozjs stuff
15:13:19 <rric> this will be a later question wrt to kernel config to update from 42 to 48
15:13:30 <pbrobinson> rric: we're not enabling in the kernel until it's been well tested, and that is not the case. I have on my list to do a testing kernel, need to get back to thaat
15:13:47 <jonmasters> rric: those tools need to be fixed still - specifically the static included versions in the desktop stack
15:14:24 <pwhalen> #topic 2) ==== Kernel Status ====
15:15:00 <pwhalen> #info kernel-4.8.0-0.rc6.git0.1.fc26 available for testing.
15:15:30 <pwhalen> #link http://arm.koji.fedoraproject.org/koji/buildinfo?buildID=402051
15:15:39 <rric> I could boot ThunderX with an unpatched Fedora kernel but needed config changes
15:15:47 <rric> changes are 48 va size
15:15:48 <pbrobinson> I think kernel looks mostly OK, I've only tested to rc5, been focused on merge and other things
15:15:55 <pwhalen> seattle - https://paste.fedoraproject.org/427565/73774681/
15:16:03 <rric> NR_CPUS set to at least 96
15:16:20 <pwhalen> mustang - https://paste.fedoraproject.org/427562/77437614/
15:16:20 <rric> (RHEL has 4092)
15:16:35 <pbrobinson> rric: NR_CPUs is set to 256 on aarch64, and has been for a while, what kernel are you using?
15:16:37 <hrw> pwhalen: http://koji.fedoraproject.org/koji/buildinfo?buildID=800427 you meant (for kernel)? let's use main koji ;D
15:16:54 <rric> probinson, 1sec
15:16:59 <jlinton> pwhalen: Which seattle machine?
15:17:57 <pwhalen> hrw, thats f26, f25 build is only on arm.koji.. but sure
15:17:59 <rric> probinson: config-4.8.0-0.rc2.git3.1.fc25.aarch64
15:18:12 <pwhalen> jlinton, b0
15:18:24 <jlinton> pwhalen: I've been pushing for firmware updates (via softiron, as have other people, now word on when that will happen)
15:18:31 <rric> # grep NR_CP cv/configs/config-4.8.0-0.rc2.git3.1.fc25.aarch64
15:18:31 <rric> CONFIG_NR_CPUS=8
15:18:53 <pbrobinson> rric: that's like a month old, upgrade to the latest
15:19:23 <rric> probinson: ok
15:20:05 <pbrobinson> rric:  $ grep NR_CPU /boot/config-4.8.0-0.rc4.git0.1.fc25.aarch64
15:20:05 <pbrobinson> CONFIG_NR_CPUS=256
15:20:30 <rric> probinson, I also have seen a BIG_ON() in the crypt code with signed modules, is this known?
15:20:37 <rric> BUG_ON() ...
15:20:46 <pbrobinson> rric: not that I've seen
15:21:08 <rric> need to narrow it down
15:21:14 <jonmasters> I haven't seen that
15:21:40 <jonmasters> jlinton: what do you need a firmware update on Seattle for?
15:21:43 <pbrobinson> rric: I was intending on testing signed and compressed modules on rawhide at some point but need to test that and some other functionality
15:22:13 <rric> pbrobinson: kernel BUG at crypto/asymmetric_keys/public_key.c:96!
15:22:18 <jlinton> jonmasters: There are two nasty warn_Ons in the latest kernels with seattle machines (see pwhalen's link)
15:22:39 <jonmasters> ah that EFI mapping warning - getting that elsewhere too yea
15:22:55 <pwhalen> i hit this oops on seattle when attempting to launch a vm - https://paste.fedoraproject.org/427621/80120147/
15:22:55 <jlinton> jonmasters: The EFI region alignment/overlap and the corrupted IRQ flags on return from efi runtime services
15:23:37 <pwhalen> i'll file a bz on that today, had hoped rc6 would be better
15:23:40 <jonmasters> the former is just something they should know better, the latter I'm not surprised by but hadn't seen
15:23:51 <jonmasters> I'll harass AMD. It's their firmware, not SoftIron
15:24:23 <pbrobinson> woo! we're on parity with x86 servers..... same shitty vendor firmwares everywhere \o/
15:24:44 <jlinton> jonmasters: Others are harrasing AMD too, yah and softiron just opens it @ AMD, but never hurts to get them pushing too.
15:24:44 <jonmasters> and it's AMI firmware too!
15:24:52 <jonmasters> jlinton: agreed
15:25:15 <pbrobinson> jonmasters: and you wondered why I wasn't excited to get that on other vendor HW
15:26:09 <cov> I was testing -rc5 last week on a couple systems and PCIe, SATA, USB all looked good, save IORT (support for the ACPI table essential for PCI MSIs) missing upstream
15:27:36 <pwhalen> #topic 3) ==== Bootloader Status ====
15:28:03 <jonmasters> cov: yea, I'm following up on IORT status and we have folks internally looking
15:28:16 <pbrobinson> 2016.09 has gone stable upstream, I will be adding a qmeu elf variant shortly and then pushing the build, expect it everywhere RSN
15:28:45 <jonmasters> pbrobinson: in fairness, those warnings won't blow up a system and the issues have been there forever. I've pinged AMD directly just now and will ping APM for the efi create mapping issue on mustang
15:28:51 <hrw> pbrobinson: qemu u-boot for vm?
15:29:19 <pbrobinson> hrw: yep, that's the plan
15:29:30 <hrw> pbrobinson: awesome
15:29:32 <pbrobinson> hrw: for ARMv7 not aarch64 VMs
15:29:36 <hrw> pbrobinson: I know
15:29:53 <hrw> pbrobinson: arm32: u-boot, arm64: grub-efi is official way
15:30:26 <pbrobinson> yep
15:31:08 <pwhalen> #topic 4) ==== F25 Beta ====
15:31:13 <hrw> I wanted to play with arm32-vm -> uefi -> grub-efi but grub needed some time so skipped
15:31:48 <pwhalen> #link https://qa.fedoraproject.org/blockerbugs/milestone/25/beta/buglist
15:32:22 <pwhalen> armhfp f25 beta, we have a couple blockers on initial-setup
15:32:49 <pbrobinson> yes, I think one of those is because of the old NetworkManager, and other of those issues i actually reported in F-24
15:33:02 <pbrobinson> I'm going to test it shortly
15:34:55 <pwhalen> armhfp we still have some display issues to solve, and ext4 on /boot (problem reading dtb on ext4)
15:35:05 <pwhalen> anything else?
15:35:27 <pbrobinson> pwhalen: well the later has been worked around, I don't think we'll get a proper fix until we can move from appliance-tools to lmc
15:35:54 <pbrobinson> I don't see the work around as insufficient, it seems to be working OK?
15:36:03 <pwhalen> it has, but anaconda will still use ext4 by default
15:36:26 <pwhalen> doesnt affect highbank, but everything else..
15:36:28 <jlinton> There is still the firefox crash on startup, which seems like a biggie for desktops
15:36:54 <pbrobinson> jlinton: did you see my update on the ticket, have you tried 48+
15:37:03 <pbrobinson> jlinton: the version reported was 47
15:37:31 <jlinton> pbrobinson: I haven't tried the latest version yet
15:37:44 <jlinton> is it working for other people?
15:37:47 <pwhalen> jlinton, yes, thanks for filing. I hit it as well testing vnc, so u think its still there
15:38:19 <jlinton> pwhalen, yah, I saw that, which is why i'm mentioning it
15:40:02 <pwhalen> for aarch64 we have this - https://bugzilla.redhat.com/show_bug.cgi?id=1368569
15:41:22 <pwhalen> i tried downgrading libjpeg-turbo, didnt help, or upgrading to the latest tigervnc, didnt help there either.
15:42:47 <pwhalen> gdb output- https://paste.fedoraproject.org/427605/73778908/
15:43:04 <pbrobinson> pwhalen: is that in the bug?
15:43:27 <pwhalen> just did that with the latest tigervnc, will add
15:44:26 <pwhalen> any other issues for f25 beta?
15:44:47 <pbrobinson> not from me atm
15:45:00 <pwhalen> #topic 5) == Open Floor ==
15:45:35 <jlinton> Hmm just tried firefox 48, via remote X
15:45:38 <jonmasters> pbrobinson: what's the latest on U-Boot for aarch64 guests?
15:45:43 <cov> when will the f25 kernel be closed to updates? Sep 27?
15:45:46 <jlinton> and it throws XPCOMGlueLoad error for file /usr/lib64/firefox/libmozgtk.so:
15:45:55 <cov> s/updates/new features/
15:45:55 <jlinton> so somethings not right with the dependency chain
15:45:56 <pbrobinson> jonmasters: on my todo list
15:46:17 <pbrobinson> jonmasters: by u-boot for aarch64 guests, do you mean SBCs like pine64?
15:47:00 <jonmasters> nope - for VMs in the build system
15:47:16 <hrw> pbrobinson: do we have arm32 images with spare partition composed? one where we can put r/pi firmware or chromebook u-boot or efi?
15:47:31 <jonmasters> (I've little desire for SBCs running U-Boot in my life :) )
15:47:40 <pbrobinson> hrw: not currently
15:47:48 <hrw> jonmasters: why you want to run u-boot on aarch64/vm? uefi+grub is a way
15:47:52 <pbrobinson> jonmasters: so you mean ARMv7 guests then....?
15:48:06 <jonmasters> pbrobinson: sorry, yea, I did miss that one important detail
15:48:09 <jonmasters> :)
15:48:15 <hrw> jonmasters: 17:28 < pbrobinson> 2016.09 has gone stable upstream, I will be adding a qmeu elf variant shortly and then pushing the build, expect it everywhere RSN
15:48:21 <jonmasters> -ETOOMANYOTHERCONVERSATIONSGOINGON
15:48:32 <jonmasters> cool tnx
15:48:42 <jonmasters> just wanted to understand if we needed to help that along
15:48:45 <pbrobinson> jonmasters: I'm enabling it in the GA build for 2016.09 today, we have it sort of working
15:49:21 <hrw> pbrobinson: I have ideas for chromebook and r/pi stuff for arm-image-installer but those need partition :(
15:49:26 <pbrobinson> jonmasters: we need to some how work out DT bits from the generated one in qemu -> u-boot -> kernel so the kernel gets all bits like memory, virtio devices etc
15:50:17 <hrw> pbrobinson: can't we go arm32vm -> uefi/arm32 -> grub-efi?
15:50:17 <pbrobinson> hrw: I have it high on my list, I've just got a crap load on my RSN list, once I finish aarch64 merge, next is Power64 atomic, then a bunch of ARM* image bits
15:50:30 <hrw> pbrobinson: no rush
15:50:58 <pbrobinson> hrw: no, because none of the anaconda/installer etc code paths support it and it's then a whole bunch more support code paths for a single use case that need extra QA
15:51:11 <hrw> pbrobinson: ok
15:51:47 <pwhalen> anything else for today?
15:52:27 * pbrobinson has nothing
15:54:09 * jonmasters is testing F25 on more systems
15:54:19 <jonmasters> I've pinged a couple vendors about firmware fixes
15:54:35 <cov> when will the f25 kernel be closed to new features? Sep 27?
15:54:50 <pwhalen> thx jonmasters
15:57:15 <pbrobinson> cov: that's what to aim for, and really given we're going GA on 4.8 I expect to be more minor enhancements and primarily bugfixes really
15:57:27 <pbrobinson> cov: I have no urge to deal with vast patch sets
15:57:29 <jonmasters> PS pbrobinson Applied sent that ECAM quirk update for X-Gene after I pinged yesterday FYI
15:57:43 <pbrobinson> jonmasters: cool, I'll check linux-pci archives
15:58:01 <pbrobinson> jonmasters: I was going to rebase our patches to the latest upstream revs when I get a moment
15:58:04 <jonmasters> I'm working on getting the m400 ACPI tables fixed ;)
15:58:33 <jlinton> jonmasters: Is that something beyond what is running @ RH?
15:58:50 <jlinton> jonmasters: Or is it just a case of getting an official release?
15:59:47 <jonmasters> jlinton: official release
15:59:57 <pwhalen> our time here is sadly over, lets continue in #fedora-arm
16:00:04 <pwhalen> #endmeeting