Plasma Mobile/etc. page fault? (samsung-klte)

This OS ran perfectly on my S5 with the desktop-oriented graphical environments (except plasma-desktop, weston, and terminals with on screen keyboards), but the mobile graphical environments froze on the startup screen. I could SSH into the phone and I checked dmesg and was greeted by a bunch of page faults from kwin. Any way to fix this?

Please let us know:

1. samsung-klte has mainline and downstream, which one are you using?
2. Can you give us the full dmesg log?

CC: @minlexx

By Danct12 on 2020-06-06T17:54:43

I am using the downstream kernel. Most of the dmesg log says this: ` [ 43.528795] elogind-daemon[1731]: New session c11 of user phone. [ 44.480647] kwin_wayland (2177): undefined instruction: pc=b23049be [ 44.480658] Code: 4770 ff00 0c40 4770 (efa0) 0e00 [ 44.480682] kwin_wayland (2177): undefined instruction: pc=b23049ac [ 44.480691] Code: 4770 ec51 0f1e 4770 (ffb0) 0300 [ 44.480706] kwin_wayland (2177): undefined instruction: pc=b23049b2 [ 44.480715] Code: 4770 ffb0 0300 4770 (ef00) 0c40 [ 44.480729] kwin_wayland (2177): undefined instruction: pc=b23049b8 [ 44.480738] Code: 4770 ef00 0c40 4770 (ff00) 0c40 [ 44.480752] kwin_wayland (2177): undefined instruction: pc=b23049a6 [ 44.480760] Code: 8000 ef20 0150 4770 (ec51) 0f1e [ 45.229769] kwin_wayland: unhandled page fault (11) at 0x00000008, code 0x005 [ 45.229785] pgd = e5754000 [ 45.229789] [00000008] *pgd=00000000

[ 45.229837] Pid: 2177, comm: kwin_wayland [ 45.229843] CPU: 0 Tainted: G W (3.4.113 #7 (closed)-postmarketOS) [ 45.229848] PC is at 0xb6dfa7e2 [ 45.229852] LR is at 0xb62de7c3 [ 45.229869] pc : [] lr : [] psr: 20000030 [ 45.229871] sp : beac31a8 ip : b6661c38 fp : 00000001 [ 45.229876] r10: b6ecffa0 r9 : beac31a8 r8 : 00000000 [ 45.229880] r7 : b6ed64c8 r6 : b6ffeaa0 r5 : b6fd5090 r4 : b6ecffa0 [ 45.229885] r3 : 00000000 r2 : b6ffca50 r1 : ffffffff r0 : beac31a8 [ 45.229891] Flags: nzCv IRQs on FIQs on Mode USER_32 ISA Thumb Segment user [ 45.229896] Control: 10c5787d Table: 2575406a DAC: 00000015 [ 45.229914] [] (unwind_backtrace+0x0/0xec) from [] (__do_user_fault+0x100/0x150) [ 45.229924] [] (__do_user_fault+0x100/0x150) from [] (do_page_fault+0x330/0x36c) [ 45.229934] [] (do_page_fault+0x330/0x36c) from [] (do_DataAbort+0xc8/0x210) [ 45.229945] [] (do_DataAbort+0xc8/0x210) from [] (__dabt_usr+0x34/0x40) [ 45.229950] Exception stack(0xe4c6bfb0 to 0xe4c6bff8) [ 45.229956] bfa0: beac31a8 ffffffff b6ffca50 00000000 [ 45.229963] bfc0: b6ecffa0 b6fd5090 b6ffeaa0 b6ed64c8 00000000 beac31a8 b6ecffa0 00000001 [ 45.229969] bfe0: b6661c38 beac31a8 b62de7c3 b6dfa7e2 2000ffff [ 45.230061] kwin_wayland: unhandled page fault (11) at 0x0code 0x005 [ 45.230066] pgd = e5754000 [ 45.230070] [00000008] *pgd=00000000

[ 45.230077] Pid: 2177, comm: kwin_wayland [ 45.230082] CPU: 0 Tainted: G W (3.4.113 #7 (closed)-tOS) [ 45.230087] PC is at 0xb6dfa7e2 [ 45.230091] LR is at 0xb62de7c3 [ 45.230095] pc : [] lr : [] psr: [ 45.230097] sp : beac31a8 ip : b6661c38 fp : 00000001 [ 45.230102] r10: b6ecffa0 r9 : beac31a8 r8 : 00000000 [ 45.230106] r7 : b6ed64c8 r6 : b6ffeaa0 r5 : b6fd5090 rfa0 [ 45.230111] r3 : 00000000 r2 : b6ffca50 r1 : ffffffff r1a8 [ 45.230116] Flags: nzCv IRQs on FIQs on Mode USER_32 I Segment user [ 45.230121] Control: 10c5787d Table: 2575406a DAC: 00000 [ 45.230144] [] (unwind_backtrace+0x0/0xec) from 0>] (__do_user_fault+0x100/0x150) [ 45.230155] [] (__do_user_fault+0x100/0x150) froaf0>] (do_page_fault+0x330/0x36c) [ 45.230163] [] (do_page_fault+0x330/0x36c) from 0>] (do_DataAbort+0xc8/0x210) [ 45.230172] [] (do_DataAbort+0xc8/0x210) from [<] (__dabt_usr+0x34/0x40) [ 45.230177] Exception stack(0xe4c6bfb0 to 0xe4c6bff8) [ 45.230183] bfa0: beacffff b6ffca50 00000000 [ 45.230189] bfc0: b6ecffa0 b6fd5090 b6ffeaa0 b6ed64c8 000031a8 b6ecffa0 00000001 [ 45.230196] bfe0: b6661c38 beac31a8 b62de7c3 b6dfa7e2 2000ffff [ 45.410042] elogind-daemon[1731]: Removed session c11. `

By 4TechGuns on 2020-06-06T18:48:50

Provide this info:

1. Are you using downstream or mainline kernel
2. Are you using nonfree-firmware
3. In case of downstream kernel, are you using nonfree-userland?

General advice would be: do not use downstream kernel, and use mainline now.

By Alexey Min on 2020-06-07T02:10:59

added category::device type::bug ui-plasma-mobile labels

I am using the downstream kernel, and I just tested mainline and Plasma Mobile seems to crash every time I open an app. (also Phosh only displays a non-blinking cursor or two randomly around the screen when it finishes booting)

I used non-free firmware.

When I used downstream with nonfree-userland, the screen went black 1 second after the boot screen and never continued on with the same page faults. (none of the graphical UI's worked with this method) Downstream without nonfree-userland replicated the same problem I addressed in the main post.

In shorter terms, I didn't use nonfree-userland when I tested mainline.

By 4TechGuns on 2020-06-07T11:01:48

There are several options here (all using nonfree-firmware, it is required to make device useful)

1. Mainline kernel: display pipeline on msm8974 has bugs ( #424 (closed) ) resulting in this behaviour: screen works every 2nd or 3rd boot randomly, other times it just freezes. Try rebooting several times :) Bug was introduced in upstream Linux kernel (some time after 5.4? maybe 5.5?) and there are no fixes or working workarounds so far.
2. Downstream without nonfree-userland: only simple UIs work, that support framebuffer: xfce4 or mate, for example. Weston worked before, but recent upgrade broke it on old kernels ( #579 (closed) ). You can also try to use Plasma Mobile with software rendering, but you'll see all the problems quickly yourself (framebuffer backend is not supported very well in upstream too)
3. Downstream kernel with nonfree-userland requires manual installation that is described in the device wiki page https://wiki.postmarketos.org/wiki/Samsung_Galaxy_S5_(samsung-klte)#Nonfree-userland_.28halium.29_installation . You can run test_hwcomposer_7_caf from ssh shell and enjoy beautiful rotating spiral, but that's all (ok, maybe you can run some other libhybris-tests like test-sensors or test-gps). None of the UIs have backends that make use of EGL from Android's hwcomposer. Attempt to add it was rejected by postmarketOS core developers in !1010 (closed) . You need all 4 commits from that MR to make use of kwin's hwcomposer backend (forked kwin, kwin-hwcomposer backend and patched plasma-phone-components). But even then it does not work out of the box and freezes on kwin startup during virtual keyboard initialization. You need another patch to kwin to disable virtual keyboard integration (not part of that MR), and only then you can see accelerated PlaMo on downstream. Phosh, for example, does not have hwcomposer backend at all. The only advantage of this option is audio support using pulseaudio-modules-droid. (Mainline msm8974 does not have audio support at all). None of regular Linux userspace can make use of android's hardware drivers. You can port all Sailfish/Mer or UBPorts userspace utils, libs and other middleware, but that sounds like a lot of duplicate work, that was already done by those projects. If you choose option 3, you may have better luck by creating a port for those projects. But not with postmarketOS. Halium support in postmarketOS is very poor, almost nonexistent.

Basically downstream kernel approach is a dead end, and not supported anywhere (not even in upstream kwin). Best way would be for someone to fix display pipeline in mainline...

I didn't use nonfree-userland when I tested mainline.

Right, nonfree-userland makes sense only for downstream kernel

By Alexey Min on 2020-06-07T12:26:48

This issue is 7 months old without an update, and mainline support has been improved a lot in this time. I'm closing this issue for now but if you can still reproduce this issue on the latest mainline build please re-open it @4techgunsinbox.

By Bart Ribbers on 2021-01-07T21:07:03

closed