15:00:22 #startmeeting Fedora ARM and AArch64 Status Meeting 15:00:22 Meeting started Tue Aug 23 15:00:22 2016 UTC. The chair is pwhalen. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:22 Useful Commands: #action #agreed #halp #info #idea #link #topic. 15:00:22 The meeting name has been set to 'fedora_arm_and_aarch64_status_meeting' 15:00:22 #chair pwhalen pbrobinson dgilmore hrw dmarlin yselkowitz jonmasters ahs3 msalter 15:00:22 Current chairs: ahs3 dgilmore dmarlin hrw jonmasters msalter pbrobinson pwhalen yselkowitz 15:00:29 good morning folks, whos here today? 15:01:03 * pbrobinson o/ 15:02:47 hi 15:03:04 good morning cov, welcome 15:03:32 right let's get this show on the road 15:03:40 #topic 1) ==== Package Status & Issues ==== 15:04:42 not problem packages 15:05:05 #info f26 stats: statistics: {'older': 44, 'local_only': 1, 'remote_only': 279, 'same': 18098, 'newer': 0, 'total_missing_builds': 51} 15:05:31 #info f25 stats: statistics: {'older': 50, 'local_only': 1, 'remote_only': 277, 'same': 18094, 'newer': 0, 'total_missing_builds': 57} 15:05:40 #info No current build issues. 15:05:43 very nice. 15:06:05 I spent some time over the last week going through all of the "older" and "remote_only" to see what was missing etc 15:06:24 #info wine stack now bootstrapped info F-25+ 15:07:07 anything of interest for us in remote or older? 15:07:28 #info looking for people interested in helping bootstrap fpc (pascal), D lang, sbcl (lisp) and chromium on aarch64 15:07:50 there's the above bits plus mono and a few randoms 15:08:09 I've been checking Exclude/Exclusive arches etc in preparation for merging 15:08:44 overall it's the best we've looked ever (as it should be) 15:09:08 we should have mono shortly, I've engaged with the maintainer and we'll have support with mono 4.6 15:09:30 that should land when mono 4.6 gets to beta phase 15:10:20 #info Mono to be supported with 4.6, to land when it gets to the Beta phase. 15:10:47 anyone have anything else to add? 15:11:24 #topic 2) ==== Kernel Status ==== 15:12:06 armv7, i think we're looking pretty good thanks to a bunch of fixes from pbrobinson. 15:12:48 #info 4.7.2 on it's way to Fedora 24 and 23 updates-testing very soon. Please test and report issues. Looks OK in my testing 15:13:01 #info Latest kernel - kernel-4.8.0-0.rc3.git0.1.fc25 15:13:24 latest rawhide/25 kernel is now 4.8 RC3, it's by no means perfect 15:13:32 some known issues with BBone MMC 15:13:35 #info Adds fix for RTC crash on ARMv7 am33xx devices 15:13:41 quite probably other issues. 15:15:07 #info test lastest 4.8 kernels and report issues 15:15:34 for aarch64. Im looking at a crash when booting the installer with 4.8.0-0.rc2.git3.1.fc25.aarch64 15:15:41 on the mustang, seattle is ok 15:15:54 might be firmware as this is quite old 15:16:01 #info 4.8 rc2 git snapshot 3 is now in Fedora 25 GA. It has support on aarch64 for ACPI PCI-e, NUMA and Cavium Thunder X so please test this platform if you have access 15:16:54 anyone else have any questions/queries about either the 4.7.x stable kernels or 4.8 devel kernels? 15:17:02 any other kernel testing, news? 15:17:19 are folks booting ACPI or devicetree on the Thunder? 15:17:46 cov: TBH I don't have access to test, no reports but I believe ThunderX should use ACPI 15:18:01 so no idea really :) 15:18:05 * pwhalen tests with Mustang, Seattle and QEMU 15:18:55 I can't boot on QDF2432 because http://codeaurora.org/cgit/quic/server/kernel/commit/?h=cov/4.8-rc2-testing&id=edc7986d4d405daebaf2f66269b353da579fce5f is missing 15:19:07 it's a CPU erratum that's been nacked upstream 15:19:49 I've started testing on an emulation platform that doesn't suffer from that erratum, but it's slow and I don't have results yet 15:20:03 cov: OK, do we need to enable other options in the kernel configs? 15:20:32 no, the patch isn't in Fedora and I've been unsure of whether to ask for it to be included or find some other way to deal with it 15:21:23 for example applying the patch and building locally or using a JTAG script to work around the issue 15:21:34 cov: well depending on impact on other platforms I can look at pulling it in, but I don't want to have it extended time so ultimately we need a proper upstreamable fix 15:21:43 it's been nacked upstream 15:22:05 cov; sure, I get that, but presumably nacked in favour of another solution 15:22:34 cov: I've not seen the upstream discussion etc, there must have been a reason for the nack, was there suggestion for a different solution 15:22:40 probinson: the solution is to replace the hardware, which won't happen now 15:23:53 but eventually 15:23:57 cov: so for example what is the impact on ARMv8.1+ 15:24:21 cov: is that HW due to go to mass production or is it a bringup batch and it'll be fixed in a later silicon rev? 15:25:05 it isn't mass production hardware 15:25:14 cov: happy to take this off line if there's NDA or other issues that come into play 15:26:13 the fix will be in the form of newer hardware 15:26:36 so this particular erratum workaround is only temporarily useful 15:27:47 cov: so what's the expected lifespan remaining on this HW? Is there possible impact in that time on ARMv8.1+ HW 15:28:18 we've shipped non upstream drivers for other devices in the past until the rev of silicon was replaced with newer to make it useful 15:28:41 the QDF2432 hardware was announced last october or november 15:29:05 so if we have a reasonable window I'm not opposed to adding this patch as a means of making it useful 15:29:27 cov: more worried about the end game for the hardware 15:30:02 the issue will be fixed in the next hardware 15:30:38 I don't know what will happen to the old hardware. presumably it will be recycled 15:30:49 cov: when is that expected roughlt, next month, next quarter, next decade. millenium? 15:31:02 I don't think there have been any announcements 15:31:43 cov: basically I want to know 1) can this impact other hardware in any way (looks like it's explicitly only enabled for that HW rev) 2) how long will I have pain of OOT patch 15:32:39 3) actual guarantees that if we ship it that people will actually use/test Fedora for us (otherwise there is no point) 15:34:10 sorry I am late - appointment went long 15:34:12 catching up 15:34:23 good morning jonmasters 15:35:43 cov: so maybe even answer Q 1 and 3 15:36:08 pbrobinson: that QDF errata patch is trivial and happy to help maintain it (or get someone inside RH to do so) - you and I can take offline also 15:36:33 jonmasters: OK, still want answers to 1 and 3 though 15:37:01 jonmasters: also ThunderX Q above: who's committing to test the support and report to the list on status etc 15:37:07 1) it won't impact other hardware. we can double-check this by dumping info->reg_id_aa64mmfr2 on other hardware 2) what if we arbitrarily plan to drop it by f26? 3) these being pre-production there aren't a lot of systems and users, but I'll be testing at least 15:37:41 * jonmasters answered pbrobinson 's Qs privately - can take the NDA stuff elsewhere 15:38:07 cov: dropping by F26 sounds like a great plan 15:38:32 * jonmasters can test. I'll also test the ThunderX stuff. Haven't had chance yet, as I needed to GTFO over the weekend 15:38:37 cov: perfect, happy to have it for that window, will try getting it pulled in over the next few days 15:38:47 cov: will ping on channel when it's there 15:38:55 pbrobinson: much appreciated 15:39:10 cov: no problems 15:39:31 anything else for the kernel? not sure if you had anything jonmasters 15:39:44 jonmasters: would like some others testing ThunderX as you can be a blocker at times, would be nice to see Cavium involved ;-) 15:39:50 well I still need to help pbrobinson with some testing 15:39:58 robert richter tried to show up to one of these meetings a while ago 15:40:08 pbrobinson: I reminded Cavium last week that a certain other partner has someone who bothers to turn up regularly 15:40:24 pbrobinson: I will begin public shaming soon 15:40:39 but there was some weird scheduling, I think some other Fedora group was using this channel at what appeared to be the ARM timeslot 15:40:46 jonmasters: so the F-25 nightly compose as of today has all the ThunderX all there for installer etc 15:40:50 ok 15:41:06 I'll make sure it's tested this week, and also request in our internal meeting today that somoene else look too :D 15:41:22 jonmasters: thanks! 15:41:29 pwhalen: lets move on from kernel :) 15:41:46 how about replacing http://pkgs.fedoraproject.org/cgit/rpms/kernel.git/tree/arm64-avoid-needing-console-to-enable-serial-console.patch with SPCR patches? 15:41:47 oh one more 15:42:07 jonmasters: hurry up! 15:42:08 cov: that makes perfect sense 15:42:18 http://patchwork.kernel.org/patch/9276727/ 15:42:18 http://patchwork.kernel.org/patch/9275443/ 15:42:18 http://patchwork.kernel.org/patch/9275457/ 15:42:19 http://patchwork.kernel.org/patch/9280971/ 15:42:21 cov: I have that on my todo list 15:42:35 cov: just need a few more round-tuits 15:42:58 on kernel...this hasn't officially bitten Fedora yet BUT we've seen various random XFS corruption on kernels eslewhere and I want to ask folks to keep a look out 15:42:59 okay, I mind be able to find a couple 15:43:05 cov++ thanks for those links, saves me time digging them out :) 15:43:05 pbrobinson: Karma for cov changed to 1 (for the f24 release cycle): https://badges.fedoraproject.org/tags/cookie/any 15:43:06 * pwhalen takes some blame for Peters round-tuit shortage 15:43:28 pbrobinson: in case it's useful, here's my big long list of stuff in flight 15:43:31 https://wiki.codeaurora.org/xwiki/bin/Linux+Kernel/WebHome 15:43:31 jonmasters: does it happen on x-gene1 on 7.2? 15:43:39 also stuff merged in case it needs backporting 15:43:50 cov: will look later post meeting, might ping you on #fedora-arm 15:43:51 pbrobinson: not on 7.2 15:44:13 yikes, quite long that list 15:44:13 pbrobinson: but I'm a bit worried that there's a genuine XFS bug upstream that's just being masked in some contexts, so keep a look out 15:44:24 seems to be toolchain related - we're digging 15:44:34 so just an FYI and we can move on 15:44:37 PCI IORT is another patchset that's not yet merged upstream 15:44:39 OK 15:44:44 #topic 3) ==== Bootloader Status ==== 15:45:02 cov: OK, I've seen that, not sure the impact or what it brings to the table 15:45:15 so on bootloaders 15:45:28 2016.09 rc2 building on Fedora 26 15:45:49 #info U-Boot 2016.09 rc2 building on Fedora 26 15:45:51 #info plan to push 2016.09 rc2 to F-25 once built on F-26 15:46:14 and tested of course 15:46:16 pbrobinson: PCI Message Signaled Interrupts (MSIs) are needed to make some PCI card drivers work and the IORT describes some of the necessary setup in ACPI 15:46:51 cov: OK, any particular range of cards need it or don't work without it? 15:46:55 cov: also we need the SMMUv3 part of IORT for device assignment so it's definitely being tracked 15:46:59 or random collection? 15:47:08 pbrobinson: anything PCIe that's very modern and doesn't do legacy interrupts 15:47:41 so SRV-IO stuff? Mellonox? 15:47:44 on that issue, RH is beating folks up on upstreaming it 15:47:54 pbrobinson: needed for SR-IOV BUT also for just interrupts on some cards 15:48:04 pbrobinson: we're harassing to get the basics in the next kernel cycle 15:48:18 pbrobinson: then the SR-IOV type stuff via SMMUv3 which is also plumbed in via IORt 15:48:19 IORT 15:48:31 (IO Redirection Table) - includes things like MSIs but much more 15:48:42 OK, not sure I really want another big set of patches 15:48:50 I'd much sooner them get upstream 15:49:02 but can we take this to any other business or the other channel 15:49:10 we have 10 mins to finish up 15:49:13 once I get done doing post-7.3 planning internally I'll help pwhalen and others go through and file some trackers in Fedora for these pieces if that helps 15:49:20 ok, let's do bootloaders 15:49:34 so u-boot will need some testing, it looks OK 15:50:23 we had to move images back to ext3 to fix an issue with changes in ext4 defaults regarding some features (64bit - inodes I think) and some other meta data option I don't remember 15:50:31 so something to be aware of 15:51:05 need to look at a better fix (like anaconda filesystem default option settings or similar) but ext3 was the quick workaround 15:51:35 something to fix for Beta 15:51:48 anything else for bootloader? 15:51:51 s 15:52:07 #topic 4) ==== F25 Alpha ==== 15:52:21 so we're very close to alpha 15:52:26 there's a few issues outstanding 15:52:39 I'm looking at an issue with graphics/X on some devices 15:52:44 #info Alpha was delayed a week. go, no-go decision will be Thursday. 15:52:49 beaglebone had an issue with MMC 15:53:04 and probably lots that pwhalen will remind me of 15:53:22 pbrobinson: The graphics problem is the failure of xorg -config ? 15:53:36 or something else? 15:54:08 jlinton: not 100% sure yet, I'm thinking maybe around some of the changes in the fbdev unbundling stuff and we've lost a config option they want 15:54:43 also looking at an option of a driver (tiny something other another I don't remember offhand) from staging 15:54:46 Just trying to clarify if its a configuration problem, or the graphics is broken 15:55:05 because i've had to hand write all my xorg.conf files on arm64 so far 15:55:09 jlinton: well we don't generally use configs, expect autogen to work, so something has changed 15:55:26 jlinton: this is mostly ARMv7 15:55:38 the pci card ids are missing for PCI cards, and the drm stanza isn't in place correctly either 15:55:59 jlinton: got details/patch somewhere? 15:56:08 no, not yet.. 15:56:52 no, i'm only on arm64 15:57:16 issues with alpha on aarch64 include a vnc segfault (bz#1368569) 15:57:26 jlinton: OK, so it's likely a different problem, ping me details or cc: me when you do have it for arm64 15:57:40 pwhalen: that looks like a libjpeg crash btw 15:57:41 and mustang crash when booting the installer, no bz yet, will have one shortly 15:58:03 which might be awaiting the infamous firmware from APM \o/ 15:58:49 hopefully 15:58:52 aarch64 is great.... it's now exactly like x86_64 just like jonmasters wanted.... it has all the really really shit firmware! 15:59:02 :P 15:59:21 there's a new firmware coming soon from APM that's AMI based 15:59:24 jonmasters: there's not enough whiskey and not enough livers for me to cope! 15:59:30 #info Help testing Alpha would be greatly appreciated, aarch64 and armv7. 15:59:52 they claim in testing it boots the latest Fedora alpha bits 15:59:56 jonmasters: so it'll fix the one bug I need fixed, and introduce a billion+ more! BOOOOO!!!!! 16:00:02 which is great, but it's needs to be released 16:00:07 :) 16:00:17 pbrobinson: just for you, there's a non-AMI one too 16:00:33 jonmasters: I _REALLY_ just want the current generation firmware with the GICv2 fix I need 16:00:34 ok, we're at the hour 16:00:36 pbrobinson: that also apparently fixes said problems in their tests, but it's still in QE cycle 16:00:51 pbrobinson: I'll slip you a copy of the test one 16:00:54 #topic 5) == Open Floor == 16:01:17 * jonmasters will poke vendors to send people to this meeting 16:01:58 ok, lets take this to #fedora-arm for anythign else, i think we have another meeting now (patiently waiting, thanks!) 16:02:07 ok 16:02:10 yep, pwhalen++ 16:02:10 pbrobinson: Karma for pwhalen changed to 2 (for the f24 release cycle): https://badges.fedoraproject.org/tags/cookie/any 16:02:16 pwhalen: ++ 16:02:21 pwhalen++ 16:02:21 jonmasters: Karma for pwhalen changed to 3 (for the f24 release cycle): https://badges.fedoraproject.org/tags/cookie/any 16:02:22 #endmeeting