**** BEGIN LOGGING AT Thu Mar 07 02:59:57 2019 Mar 07 07:13:33 Morning! Mar 07 07:21:16 Tofe: While digging a bit into fstat and 3.4 I came across this patch: https://lore.kernel.org/patchwork/patch/329173/ Not sure if it's related though Mar 07 07:21:35 But seems we don't have this in our 3.4 Hammerhead kernel at least Mar 07 07:36:59 Morning! Mar 07 07:39:10 Tofe: Morning! Mar 07 07:39:20 Found some links that might or might not be related ;) See log Mar 07 07:40:00 Herrie|Laptop: yes, I'll read that, thanks Mar 07 07:40:38 Just trying to think what's the cleanest solution here Mar 07 07:41:02 Having to patch every kernel is also not really good though Mar 07 07:42:12 But if it would solve the root cause, that would be my preferred solution Mar 07 07:43:54 the last one is interesting yes; the others are, I think, not really related to our problem, which is really fstat itself, not the root prefix Mar 07 07:44:20 it would be worth a try on hammerhead, and easy too Mar 07 08:04:01 Tofe: You mean this one? https://lore.kernel.org/patchwork/patch/329173/ Mar 07 08:12:01 yes Mar 07 08:37:59 Seems builder is not reachable :S So I guess tonight :P Mar 07 08:38:16 Can VNC in, but doesn't respond to anything and cannot SSH in :S Mar 07 09:05:03 it's not that urgent :) Mar 07 09:08:06 Tofe: Just more annoyed builder went down :P Mar 07 09:43:06 Mar 07 09:43:13 erm Mar 07 09:55:40 bshah: o/ Mar 07 11:38:59 Seems I ran into some limits of the file system and all blocked now LOL :P Mar 07 11:39:23 bshah: Good news is that we seem to have a "fix" for newer systemd on 3.4 kernel Mar 07 11:44:55 is it kernel patch? Mar 07 11:50:34 bshah: Well maybe... We now have a systemd patch to work around kernel 3.4 issue. https://github.com/webOS-ports/meta-webos-ports/pull/318/files Mar 07 11:50:47 But it could be we'll patch the kernel, need to test that to see if it works Mar 07 12:08:04 Tofe: seems this might be also somehow related to the issue: https://github.com/Herrie82/systemd-stable/commit/91214a37ef4eb8042d2598aa89bae52b410d11a7#diff-b54b12c2206103150e319432ec4738e6 Mar 07 12:08:13 At least going by the description Mar 07 12:46:35 Herrie|Laptop: not sure Mar 07 12:47:00 Tofe: Yeah I'm also not sure... I guess I'll try a few things just to see what works and doesn't Mar 07 12:47:15 It's all pretty locked together, so hard to point at a single thing Mar 07 12:50:45 Herrie|Laptop: from my point of view, it's really all just because of fstat, so the kernel patch seems the best try. Otherwise we'll go for my patch, and investigate the mainline kernel for N5 Mar 07 12:51:15 Tofe: Yeah, your patch would only start to kick in when the other bits fail right? Mar 07 12:51:29 We could have it conditional based on kernel version like we did for the memnotify? Mar 07 14:06:29 Herrie|Laptop: sailfish guys had the list of patches needed for newer systemd in some document.. hold on Mar 07 14:11:01 https://public.etherpad-mozilla.org/p/faq-hadk "Apply this to all devices with 3.4 kernel https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=0640113be25d283e0ff77a9f041e1242182387f0 " Mar 07 14:13:59 Herrie|Laptop: my patch is only applied on arm archs, not aarch64 not x86 Mar 07 14:14:07 nor* Mar 07 14:15:07 saidinesh5: oh, we don't have this already ? that would be a shame Mar 07 14:15:53 dunno. but your hammerhead kernel - the last time i tried to steal namespace patches from was almost close to what the lineage os kernel was.. so chances areyou may have missed this out Mar 07 14:16:19 we have it for hammerhead Mar 07 14:16:26 but it only fixes procfs Mar 07 14:17:05 yeah Mar 07 14:17:08 not sure if we have it on touchpad, though Mar 07 14:18:17 wow. touchpad is still maintained? O_O Mar 07 14:18:40 yep. And we also have that commit there Mar 07 14:19:04 That's our only tablet, and quite a symbolic one too :p Mar 07 14:19:31 lol.. I see.. Mar 07 14:32:35 saidinesh5: Except for battery issues they're pretty indestructable :) Mar 07 14:38:20 Ahh that's nice.. Mar 07 15:29:53 Tofe: is the systemd fix from https://github.com/webOS-ports/meta-webos-ports/pull/318 ready for merge? the builder will soon start rebuilding sumo, so it would be good to merge it before it starts Mar 07 15:43:19 JaMa: yes, it's ready Mar 07 15:43:40 I hoped to be able to confirm it fixes everything first, but we can do the other way around Mar 07 15:44:50 JaMa: merged. Mar 07 15:45:09 would be great to import it to thud too Mar 07 15:46:22 Tofe: What were the issues with Thud again? Mar 07 15:54:45 crash in QtSceneGraph, somewhere in the GLES rendering thread, of many client apps Mar 07 15:54:52 beginning with firstuse Mar 07 15:57:48 Tofe: Ah OK Mar 07 15:58:00 That's not a good state to be Mar 07 16:31:09 reticivis: Hi Mar 07 16:31:43 hey Herrie Mar 07 16:32:20 For porting guide: You can follow Halium steps in general until https://docs.halium.org/en/latest/porting/build-sources.html#modify-the-kernel-configuration. Afterwards we've got some custom steps to follow. Mar 07 16:32:45 Since we don't use Halium-boot for example and compile kernel with GCC from Yocto and not the Google 4.x toolchain Mar 07 16:32:51 Which device are you looking at? Mar 07 16:33:15 Perfect. Thank. I'm looking at PH-1 (Essential phone). Mar 07 16:33:20 *thanks Mar 07 16:33:33 Let me have a quick look on that one Mar 07 16:34:46 I don't think anyone has built a halium image for it yet, so that'll probably be my first step. Mar 07 16:35:56 In terms of kernel we use the same options as SFOS so the verify kernel script works well for defconfig updates Mar 07 16:37:31 In your device manifest (essential_mata.xml) for halium-devices you probably need the following 3 repos at least, similar to https://github.com/Halium/halium-devices/blob/halium-7.1/manifests/elephone_trunk.xml Mar 07 16:38:15 https://github.com/LineageOS/android_kernel_essential_msm8998/tree/cm-14.1 https://github.com/LineageOS/android_device_essential_mata/tree/cm-14.1 and https://github.com/TheMuppets/proprietary_vendor_essential/tree/cm-14.1 Mar 07 16:38:38 Kernel being 4.4 is nice :) Mar 07 16:38:49 Newest we've played with is 3.18 so far on device Mar 07 16:43:03 Content of the XML should be something like this I guess: https://bpaste.net/show/1fa689818cf2 Mar 07 16:50:14 ah, hadn't seen TheMuppets github for the vendor files. That should save some manual extraction time; thanks! And for the sample manifest! Mar 07 16:52:34 plan is to start things next week and see how far I get. If anything works, I'll update here (and likely ask for help...). Mar 07 16:53:01 reticivis: I can do some legwork for you, since I've done it a few times before, but cannot do much without a device Mar 07 16:53:09 So can make a technically building image ;) Mar 07 16:53:28 Which is most likely to bootloop, but at least you could pull logs etc Mar 07 16:54:57 Closest target we have is Xiaomi Tissot, various config bits can be found here: https://github.com/shr-distribution/meta-smartphone/tree/pyro/meta-xiaomi Mar 07 16:55:33 And you'd need something similar to https://github.com/webOS-ports/meta-webos-ports/commit/f727c770e2edb05305f721a1c820eeaecf4b5ecb in order to get a working image Mar 07 17:01:32 ha, absolutely. If you don't mind, starting from a built image is always nice :) Though I'll probably still go through the process to familiarize myself. Looking through the config bits now; will read up a bit on openembedded/yocto as I don't have a ton of experience with it. Mar 07 17:16:59 It's just another build system where everything is integrated nicely :) Mar 07 17:36:24 Herrie|Laptop: I'm building the patched hammerhead kernel to give it a try, with a raw systemd-237 Mar 07 17:36:59 Tofe: OK good :) Mar 07 17:37:12 I'm on my way home, should be able to reset builder shortly :P Mar 07 17:37:46 hey ! it boots ! Mar 07 17:38:08 So the simple 1 line kernel patch fixes it too :S ? Mar 07 17:38:16 yes ! Mar 07 17:38:21 Well that's good in general :) Mar 07 17:38:35 I'm very happy. Mar 07 17:38:43 Let's patch our 3.4 kernels :) Mar 07 17:39:10 mako & hammerhead are hosted both by halium ? Mar 07 17:39:13 Silly it was something small like that Mar 07 17:39:21 mako with UBPorts it seems Mar 07 17:39:28 hammerhead with Haliu Mar 07 17:39:29 +m Mar 07 17:39:32 well once we know where the issue is, it's easier Mar 07 17:39:47 but I didn't think it would be that easy Mar 07 17:40:09 Tofe: THis one right? https://github.com/Herrie82/android_kernel_lge_hammerhead/commit/b5e4c523b4fa9c54ef43fa34d07f282258138c3b Mar 07 17:40:17 I have the 3 other patches in there as well Mar 07 17:40:21 Can PR them in 1 go to bshah Mar 07 17:40:38 exactly Mar 07 17:42:03 and /me who just merged my systemd PR... Mar 07 17:42:40 JaMa: ^ let's just not push that commit to other branches: I'll soon revert it Mar 07 17:43:03 The less systemd patches we have, the better Mar 07 17:46:01 There we go: https://github.com/Halium/android_kernel_lge_hammerhead/pull/3 Mar 07 17:46:21 elvispre: You might want to cherry pick that vfs patch for Onyx as well ;) Mar 07 17:46:26 Otherwise no boot on Sumo ;) Mar 07 17:47:27 Tofe: The Thud issue is more likely Qt related then Thud I'd say, no? Mar 07 17:47:40 We could do Thud with a bit older Qt maybe to try? Mar 07 17:47:47 See if it persists there as well or not Mar 07 17:48:06 the crash was with same exact version of Qt as in pyro Mar 07 17:48:39 but I don't have the callstack yet, so it's hard to guess what's going on Mar 07 17:58:07 Do you think we also need the Makefile patch ? Mar 07 17:58:48 Let's just include it as well. Mar 07 18:08:39 Herrie|Laptop: mmh it might still be a good idea to test the while sumo image with that little kernel patch: it fixes the initrd stage, but I didn't test the rest (which was still a pyro image iirc) Mar 07 18:09:03 had but I do have your sumo build from earlier Mar 07 18:13:56 mako seems to accept the newly built initrd with udev-237 Mar 07 18:14:04 (with patched kernel) Mar 07 18:14:11 (and with pyro userland) Mar 07 18:16:24 I'm now trying to boot your hammerhead sumo build with the patched kernel Mar 07 18:18:33 yay. It does boot. Sumo all the way. Mar 07 18:21:07 https://github.com/ubports/android_kernel_google_msm/pull/5 Mar 07 18:27:29 Herrie|Laptop: tenderloin seems to be even further behind for fstat.c Mar 07 18:28:50 Tofe: For Mako we'd need to add the other patches as well Mar 07 18:28:55 Seems I never got to upstream those Mar 07 18:29:27 the gcc 5,6,7 ones ? Mar 07 18:29:54 https://github.com/shr-distribution/meta-smartphone/tree/pyro/meta-lg/recipes-kernel/linux/linux-lg-mako Mar 07 18:29:57 Tofe: Yeah those above Mar 07 18:30:09 I guess it would be good to use the generic patch, not the 4-5-6-7 one Mar 07 18:30:21 probably Mar 07 18:32:24 I guess https://github.com/Halium/android_kernel_lge_hammerhead/commit/0f825e34f5858174aa09f83d7926d400832941e1 and https://github.com/Halium/android_kernel_lge_hammerhead/commit/8bc4e3a577c3677e798f1f5ae2142ae57ad04a10 Mar 07 18:33:40 Tofe: Tenderloin seems to have that patch already for fstat: https://github.com/webOS-ports/android_kernel_htc_msm8960/commit/5187f535beeb3731677c6f6d05402f9adec6e034 Mar 07 18:33:51 ahah, actually tenderloin is ahead ! https://github.com/shr-distribution/linux/commit/9b67aeff9256ad337e22b0f3b228b7ec01ff2a20 Mar 07 18:35:04 Discussion from Linus seems he's happy with both solutions for this Mar 07 18:35:31 so we'll just try tenderloin as-is, who knows Mar 07 18:35:48 though we might need the gcc 7.x patch anyway Mar 07 18:36:12 I think it wasn't strictly necessary on others Mar 07 18:36:18 My issue was systemd I think Mar 07 18:36:22 Not the gcc behavior Mar 07 18:36:50 Probably yes Mar 07 18:38:14 I didn't see anything gcc (panic, segfault etc) in the logs when I didn't have those patches Mar 07 18:39:24 https://github.com/ubports/android_kernel_google_msm/pull/5 updated Mar 07 18:40:02 something not to forget for next release: add a rfkill unblock wifi somewhere during the boot Mar 07 18:40:59 Tofe: Good :) Mar 07 18:42:02 Well in general it doesn't hurt to have these GCC 7 Makefile patches for all 3.4 just in case Mar 07 18:46:08 If just for unification, yes Mar 07 19:11:01 Tofe: Yeah good to keep it somewhat uniform Mar 07 19:53:17 Tofe: great work and news, then I will revert that systemd patch now Mar 07 19:53:34 Tofe: it will be easier and faster to rebuild with updated kernel, than with updated (reverted) systemd patch Mar 07 19:55:20 I'll push small kernel upgrade for qemux86, so will be rebasing other branches as well Mar 08 00:17:19 testing switched to sumo, it should start building it in ~9 hours **** ENDING LOGGING AT Fri Mar 08 02:59:56 2019