15:01:34 #startmeeting Fedora ARM and AArch64 Status Meeting 15:01:34 Meeting started Tue Sep 6 15:01:34 2016 UTC. The chair is pwhalen. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:34 Useful Commands: #action #agreed #halp #info #idea #link #topic. 15:01:34 The meeting name has been set to 'fedora_arm_and_aarch64_status_meeting' 15:01:34 #chair pwhalen pbrobinson dgilmore hrw dmarlin yselkowitz jonmasters ahs3 msalter 15:01:34 Current chairs: ahs3 dgilmore dmarlin hrw jonmasters msalter pbrobinson pwhalen yselkowitz 15:01:53 sorry Im late, whos here today? 15:02:10 o/ rric 15:02:51 * pbrobinson /o 15:02:51 This is Sunil from Cavium 15:02:59 Hi Sunil_ 15:03:07 Hello 15:03:12 good day Sunil_ , rric 15:03:18 Hi all 15:03:20 thanks for joining 15:03:21 Hi 15:03:49 lets get started.. 15:03:56 #topic 1) ==== Userspace Status ==== 15:04:15 we're looking pretty good here 15:04:28 there's a few packages that could do with someone to look at them 15:05:22 #info python-matplotlib I think is assuming 4K page size on aarch64 and hence tests are failing. It's our current biggest blocker 15:05:43 a question to dracut/systemd 15:05:53 #info gegl03 has a crash that could use some looking at and is blocking a few packages 15:05:55 rric: sure 15:06:01 while installing a kernel I see: 15:06:03 + dracut -H -f /boot/initramfs-4.8.0-rc4-00267-g664e88c14526-dirty.img 4.8.0-rc4-00267-g664e88c14526-dirty 15:06:03 logger: socket /dev/log: Connection refused 15:06:15 is this a known issue? 15:06:35 # ls -ltrra /dev/log 15:06:35 lrwxrwxrwx 1 root root 28 Jul 25 12:49 /dev/log -> /run/systemd/journal/dev-log 15:06:47 this runs as root 15:07:25 not that I'm aware of, have you checked RHBZ for other reports? 15:07:40 not yet, just looking 15:08:36 ok, if there is none, should I just open one? 15:09:04 rric: it's a good start, if you think it might be arm specific block it against "ARMTracker" bug 15:09:45 ok 15:09:48 rric: is this the Fedora ThunderX kernel? 15:10:07 yes 15:10:23 current kernel does not boot yet on dual node systems 15:10:40 so it's not Fedora built? 15:11:01 I have used fedpkg local to build the kernel for devel 15:11:27 but for quick install I tied just using make install ... 15:12:06 is fedpkg used to build kernel packages? I guess I need a login then 15:13:54 rric: you don't if you're just doing it local 15:14:04 this might be a better topic for #fedora-arm 15:14:14 ok, tnx 15:14:40 but by all means, file the bz with the details 15:14:54 any other userspace issues? 15:15:35 #topic 2) ==== Kernel Status ==== 15:15:55 nothing major here, I confirmed RC4 fixed the BBone issues 15:16:08 overall my testing looks OK for 4.8rc4 15:16:19 #info RC4 fixes BBone issues. 15:16:25 anyone aware of any issues on either 4.7.x or 4.8rc4? 15:16:40 I wasn't able to perform a network install with ksdevice= set https://bugzilla.redhat.com/show_bug.cgi?id=1373565 15:16:41 looks ok here too 15:17:16 as said, the kernel does not boot yet on ThunderX dual socket 15:17:58 I am just investigating this 15:18:02 cov: have you tried without special names? 15:18:26 it stops early after/during gic init and fails with a page fault 15:18:33 pbrobinson: like "eth0", or no ksdevice= at all? 15:18:45 cov, id remove 'ip=enp1s0f0:dhcp ksdevice=enp1s0f0' and try 15:19:17 cov: "net.ifnames=0 biosdevname=0 ip=dhcp ksdevice=bootif" or similar 15:19:48 cov: or even just ksdevice=bootif which will use the device it netbooted from 15:20:50 it should work but it might be a naming that anaconda isn't expecting so it might be a work around 15:20:53 except that the device I netbooted from has no internet connection. it's only good for pxe/tftp... I can try to rearrange my network setup, but I thought this command line worked on other distros/versions 15:22:06 cov: so try the " "net.ifnames=0 biosdevname=0 ip=dhcp ksdevice=ethX" option and see if you get any further 15:22:48 let us know in #fedora-arm 15:23:24 anyone seeing any other kernel issues not mentioned? 15:24:42 #info Kernel 4.8 RC5 build in progress. Please test and provide feedback. 15:25:39 #topic 3) ==== Bootloader Status ==== 15:25:48 I dont think anythings changed here 15:26:39 has anyone seen any issues with uboot-tools-2016.09-2.rc2.fc25? 15:27:00 I've not seen any issues 15:28:05 #topic 4) ==== F25 Beta ==== 15:29:32 nothing really from me here, just lots of testing on various HW so we are aware of issues that need attention long before freeze 15:29:51 which is towards the end of the month 15:30:20 I've been testing F25A 1.2 on the juno, so far so good 15:30:21 we should see a new nominated nightly for testing on primary shortly, we'll do the same for aarch64 and let the list know, with a test matrix for aarch64 results. 15:30:44 jlinton, excellent. thanks for testing 15:31:52 cov: did you have any luck with Friday's compose, or was the ks crash causing issues? 15:33:24 pbrobinson: the ksdevice issue was as far as I've gotten 15:33:32 (sorry was out Friday and Monday) 15:33:47 no probs 15:33:52 but looking good so far 15:34:12 no priority issue is that with earlycon= set, I get duplicated printks 15:34:13 any other concerns, questions, thoughts for F25 Beta? 15:34:26 s/no priority/low priority/ 15:35:49 #topic 5) == Open Floor == 15:35:54 Is everyone installing with inst.txt in F25? 15:36:12 if one wanted a driver that just got into net-next in Fedora 25, would that be possible? what would the deadline be? 15:36:13 jlinton: what's inst.txt? 15:36:18 jlinton: my kickstart has "text" 15:36:29 inst.text 15:36:30 cov: it depends, it's possible, ping me details 15:36:59 https:/rhinstaller.github.io/anaconda/boot-options.html#inst-text I presume 15:37:54 * cov wonders about IORT / PCI MSI support and SMMUv3 support 15:38:32 cov: it's quite a large patchset, I've not seen it accepted upstream for 4.9 as yet (although I'm a bit behind in kernel reading) 15:38:34 cov: Hmmm, well... 15:39:02 right unfortunately they aren't yet in -next 15:41:40 cov, I just hit the same issue with network, scanning mail i see there is an issue with the current version of NetworkManager 15:43:23 ill check into it a bit more after the meeting 15:43:52 anything else for today? 15:44:35 Well, with regard to the inst.text, i'm wondering if it should be the default on ARM 15:44:46 until we get the graphics head issues straighted out 15:45:46 jlinton: any assistance in debugging that would be useful, what graphics were you testing with VNC or another method? 15:45:49 jlinton: what graphics head issues? 15:46:24 Well, the installer doesn't come up on an PCI/ATI card I tried because there is some miscommunication between X and the kernel (from what I heard) 15:46:31 (didn't investigate much) 15:46:40 it doesn't work with the platform hdlcd driver either 15:46:56 the drivers are working fine, there is some kind of auto configuration problem 15:47:30 IMO, long term, if there isn't any graphics hardware it should run in text/headless mode 15:47:34 and what happens is the install starts and appears to hang because it thinks its running in graphics mode 15:48:15 jlinton: OK, is the ATI card inited by the firmware? Do you get any output on it? 15:48:20 Anyway, its seems that text mode is the common way to install at the moment 15:48:32 No, the ATI isn't initialized until the kernel starts 15:48:37 same with the HDLCD 15:48:38 jlinton: yes, mostly it is, most HW hasn't had any graphics to date 15:48:41 maybe that is part of the problem 15:48:50 jlinton: I definitely agree that text seems the common way to install 15:48:54 they arn't selected as console devices 15:49:56 but, that makes me think that the installer should detect a serial/text console and at least provide the "text/vnc" choice 15:50:06 it does if inst.text is manually put on the command line 15:50:29 * jlinton probably getting off topic for this meeting 15:51:26 jlinton, cov it should fallback to text if no graphical hw is present 15:51:53 The graphcis hardware is present, it just doesn't get configured right (AFAIK) 15:51:56 sounds like an f26 feature 15:52:26 jlinton: I can test on systems with no graphical hw 15:53:22 If systems without graphical hardware need special handling to use text mode or vnc that's one thing 15:54:28 But if you're adding a video card via PCIe and it's not Doing The Right Thing, we're talking about a new feature as far as aarch64 goes 15:55:45 It'd be a great thing to shoot for in F26 though 15:55:54 Because there is all kinds of territory we'd have to cover 15:56:05 Getting vendors to init the video cards 15:56:09 (view firmware update) 15:56:20 Getting grub2 to do the right thing 15:56:24 Getting anaconda to do the right thing 15:56:31 Hopefully just grub2 15:56:33 yes, and it's usually inited by the firmware initially, it's why you needed special cards for PowerPC macs 15:56:49 if the EFI firmware gets it into text mode (definitly possible on cards the report VGA compatability) then grub/etc should 15:56:57 but it's also out of scope for this meeting 15:57:01 be ok because they use the efi display services 15:57:09 yep 15:57:18 agree, but not a blocker for f25 15:58:04 couple minutes left, anything else for open floor? 15:58:14 jlinton: please file a bug! 15:59:38 #endmeeting