**** BEGIN LOGGING AT Mon Mar 08 02:59:57 2021 Mar 08 11:34:57 Morning! Mar 08 11:35:03 Tofe: How is Hammerhead behaving? Mar 08 12:21:00 Morning! Mar 08 12:21:33 Herrie: I have issues with the kernel, similar to what you had; either it crashes, or it hangs Mar 08 12:40:56 Tofe: Ah... No logs at all? Or ... ? Mar 08 12:41:30 Kernel should normally work OK-ish Mar 08 12:53:38 Herrie: the crash is about namespaces, like before; I tried various rollbacks in the pointed file (proc/namespaces.c mainly) but so far I've just been able to postpone the crash a little bit Mar 08 13:01:07 Tofe: You could always opt to take the whole namespaces.c from our working 3.4 kernel. I tried that with some device and it worked. Mar 08 13:01:33 But of course you might need to pick other files as well from the 3.4 kernel we use with 7.1 Mar 08 13:02:50 Herrie: already tried that Mar 08 13:04:19 Hmmz OK Mar 08 13:04:29 Weird that namespaces would give issues.... Mar 08 13:04:35 Maybe it's the defconfig somehow? Mar 08 13:04:49 Did you compare the namespacing options between our 7.1 and 9.0 defconfig? Mar 08 13:05:17 I think we enable user ns by default in 9.0 kernel script but we shouldn't need it probably. Mar 08 13:06:45 what's new here is that it's a 3.4 kernel, already patched a bit for android Pie Mar 08 13:07:01 so far we always used a 3.18 kernel as a base Mar 08 13:07:17 and we know namespace were quite incomplete in 3.4 kernels Mar 08 13:17:50 Tofe: Yes that's true Mar 08 13:32:40 Tofe: As per https://github.com/systemd/systemd/blob/main/README#L79 we might be able to disable the CONFIG_USER_NS since we don't use that in our systemd service files (yet) Mar 08 13:37:08 Herrie: I'll try that tonight Mar 08 13:37:29 Tofe: OK, can you push your work somewhere? Mar 08 13:37:35 Or it's all there already? Mar 08 13:43:40 Herrie: I still have to push the device tree, but the kernel is there: https://github.com/Tofee/android_kernel_lge_hammerhead/tree/hammerhead/3.4%2Fhalium-9.0 (speaking of which, the name of the branch looks weird) Mar 08 13:45:02 Tofe: OK Mar 08 13:45:12 And the meta-wop change also needs pushing it seems for libgbinder? Mar 08 13:48:02 https://github.com/Tofee/android_device_lge_hammerhead/tree/halium-9.0* Mar 08 13:48:04 https://github.com/Tofee/android_device_lge_hammerhead/tree/halium-9.0 Mar 08 13:48:20 yes, it's still on my side; one moment, I'll push it too Mar 08 13:50:09 manifest (though I now use my Tofee repo...) Mar 08 13:50:09 https://bpa.st/X7DA Mar 08 13:52:28 OK building Halium here Mar 08 13:53:45 https://github.com/Tofee/meta-webos-ports/tree/testing/halium-9.0-hammerhead for m-w-p Mar 08 13:55:42 https://github.com/Tofee/meta-smartphone-1/tree/testing/halium-9.0-hammerhead for meta-smartphone. Beware, the path to local halium tar.bz2 is hardcoded in that commit ! Mar 08 14:00:54 TOfe: OK thnx Mar 08 14:02:21 Hmmz seems the kernel build fails on halium side with your changes Mar 08 14:29:41 ah, on halium side I think I kept the original one Mar 08 14:29:59 it's not used anyway Mar 08 14:30:12 though it could be interesting to have it compile Mar 08 14:32:53 Tofe: OK, well it's namespaces issues anyway Mar 08 14:32:56 I'll try LOS 16 kernel then Mar 08 16:05:07 Tofe: Put the Hammerhead Halium build on: http://build.webos-ports.org/halium-luneos-9.0/halium-luneos-9.0-20210308-1-hammerhead.tar.bz2 Mar 08 16:11:36 Herrie: thanks, good idea Mar 08 16:18:33 Tofe: OK building here nwo Mar 08 16:39:46 I'll try reverting also kernel/nsproxy.c Mar 08 16:51:34 mmmh it's all entangled Mar 08 17:26:04 Tofe: Yeah :S Mar 08 17:26:12 That's the problem with the namespaces usually Mar 08 17:26:23 It might be easier to add the Android 9 specific commits to our current kernel Mar 08 17:26:31 IN general our current kernel might just work with Android 9 ? Mar 08 17:26:37 Seeing 3.18 in general did work Mar 08 17:36:31 well, the binder part has probably been updated from 5.1 to 9.0 Mar 08 17:40:32 Ah yes Mar 08 17:43:56 Seems quite some commits Mar 08 17:50:17 Tofe: Seems majority of changes in binder.c, binder.h and binder_trace.h are pretty clean. There are a few other files that have minor changes... Cherry picking the whole range of binder commits might work Mar 08 17:55:18 Herrie: that's quite a leap :) Mar 08 17:55:49 Tofe: Yeah... I'm trying your kernel now without the latest commit and with user_ns disabled Mar 08 17:55:51 Let's see Mar 08 18:06:51 Tofe: BTW I saw these: https://github.com/mer-hybris/ofono-ril-binder-plugin/commits/master, https://github.com/mer-hybris/libgbinder-radio/commits/master, https://github.com/mer-hybris/libgbinder/commits/master and https://github.com/mer-hybris/pulseaudio-modules-droid/commits/master Mar 08 18:07:02 Some might be of help for us as well, not sure which version of radio we have on our targets Mar 08 18:16:06 Tofe: OK found a possible clue: Seems your PIDNS stub implementation might be the cause. I found a FP2 kernel where they reverted your fix and added a few more commits: https://github.com/Halium/android_kernel_fairphone_msm8974/commits/halium-7.1 Mar 08 18:16:16 See the ones from 7th of June 2019 Mar 08 18:16:59 Seems they had the same issue with proc_ns_follow_link: https://github.com/Halium/projectmanagement/issues/135#issuecomment-499515390 Mar 08 18:36:36 Herrie: oh, interesting indeed Mar 08 19:16:04 looks like I messed up my cherry-picking Mar 08 19:26:27 Tofe: Well it doesn't cleanly apply it seems, at least not at my end Mar 08 19:26:50 I tried to get the merges right, but didn't Mar 08 19:31:40 retrying with eyes open this time Mar 08 19:36:19 ah, now it builds Mar 08 19:37:13 ... and crashes :( Mar 08 19:38:21 Tofe: Same here... Mar 08 19:38:49 Getting "Bad PC value": https://paste.ubuntu.com/p/QYtsx24TSM/ Mar 08 19:39:05 getting loopback_net_init in my backtrace Mar 08 19:39:36 ah you still have proc_ns_follow_link ? Mar 08 19:41:17 Tofe: Yeah seems so Mar 08 20:16:30 Tofe: While you toy with Hammerhead I'll give a go on getting something to build for Tenderloin Mar 08 20:50:08 deal :) Mar 08 20:59:00 DirtyUnicorn build is a bit nasty but well... Mar 08 20:59:10 Seems I get something to build on Halium side now after some fixups Mar 08 21:25:46 Seem mkimage was removed from android_system_extras, so will play around a bit to get that working for our uBoot image Mar 08 21:36:24 OK that's sorted Mar 08 21:36:35 Just need to upstream it to Halium for hybris-patches Mar 08 21:36:41 Now having something more interesting... Mar 08 21:37:01 "bionic/libc/include/bits/fortify/fcntl.h:63:26: error: use of undeclared identifier 'O_TMPFILE'" Mar 08 21:48:16 Testing this now: https://review.lineageos.org/c/LineageOS/android_bionic/+/223067 Mar 08 21:49:08 Ah seems this might be another solution: https://github.com/alexenferman/android_kernel_samsung_d2/commit/e97a38cefdb63305f85139c4409a597759388081 Mar 08 22:00:35 2nd solution is better... Now gpg is acting up, but we saw that before ;) **** ENDING LOGGING AT Tue Mar 09 02:59:57 2021