**** BEGIN LOGGING AT Fri Mar 02 03:00:02 2018 Mar 02 10:07:57 Hello. I've been agitated by greguu to try kexecboot on Gemini :) Mar 02 10:08:06 hehe :) Mar 02 10:08:32 so from my understand, the main trouble is getting the kernel to survice kexec boot Mar 02 10:08:37 *understanding Mar 02 10:08:51 to get a good discussion going, we need ant or ant_work to joint Mar 02 10:08:53 join Mar 02 10:09:44 he may be here soon... Mar 02 10:10:35 but, afaik, the current kernel (eg in this case an android kernel) will be wiped in the moment kexec takes place Mar 02 10:10:59 then the choosen kernel will boot and will have all the world of memory for itself Mar 02 10:11:09 chosen Mar 02 10:12:18 not all kernels can be kexe' Mar 02 10:12:38 but since we have the source, can we not add that option ? Mar 02 10:13:03 3.18.x can be kexec'd Mar 02 10:32:47 kexec on the Gemini, now that's an interesting option Mar 02 17:36:57 greguu, finally online at home Mar 02 18:16:11 morning Mar 02 18:16:55 ant_home, welcome TheKit, he is a volunteer helping to get Linux on the Gemini PDA Mar 02 18:17:14 I lured him here and suggested a kexecboot kernel as a secondary boot option Mar 02 18:17:23 he has some questions.. I may not be able to answer Mar 02 18:18:40 mainly a 3.18.41 based mediatek kernel willl survive the kexec call (from a 3.18.41 based kexecboot kernel) Mar 02 18:19:44 currently without a UART, debugging kernels is nearly impossible. A kexecboot kernel would make mainline testing also more easy I assumed Mar 02 18:25:03 for refernce, MultiROM solution for Qualcomm SoCs uses kexec hardboot hack instead (https://github.com/Tasssadar/multirom/wiki/Porting-kexec-hardboot) Mar 02 18:32:46 TheKit, I had a look a the hardboot patch and understand now what it does. Is there any indicator a hardboot kexec call needs to made on the mediatek SoC ? Mar 02 18:35:31 ok, is that some memory may not be freed and the kexec kernel can not initialize ? I see, it seems to be an issue on Android devices Mar 02 18:36:22 supposedly it's not about memory, but about the state of devices (they are in initialized state after first kernel runs) Mar 02 18:41:41 hm, I see. quite a few older qualcom soc require that patch, even nexus 5 and 7. But does it cause issues on newer aarch64 ? Mar 02 18:42:39 can we try ? Mar 02 18:48:09 well, I see, the nexus 6 uses also an aarch64 hadboot too. Mar 02 18:49:22 I've never seen either kexec or kexec-hardboot done for MediaTek SoC though Mar 02 18:49:58 yes, I am searching hard.. also porting it would require some ASM I assume.. Mar 02 18:51:34 fairphone seems to use it Mar 02 18:51:41 https://forum.fairphone.com/t/multirom-for-the-fairphone-1/25827 Mar 02 18:51:47 mediatek-mt6589 Mar 02 18:52:00 https://gitlab.com/danxuliu/android_kernel_fairphone_fp1.git Mar 02 18:56:56 but that is not an aarch64 SoC Mar 02 23:21:06 NotKit, hi Mar 02 23:23:52 about kexecboot on Gemini, what is the issue? crippled bootloader? Mar 02 23:25:31 hi ant_home Mar 02 23:25:58 "The company also promises it’ll release the source code for the Gemini PDA’s bootloader and Linux kernel." Mar 02 23:26:03 sounds good Mar 02 23:26:06 yes, we have not seen it yet Mar 02 23:26:10 ha ha Mar 02 23:26:47 but a kexecboot kernel (as second boot option) would allow more friendly loading of other linux kernels, from MMC or SD card or even USB even, so multiboot instead of dual boot Mar 02 23:26:49 it is an arm, isn't? so the binaries armv4/armv5 eabi should just work Mar 02 23:26:58 well, bootloader isn't going to be different from https://gitlab.com/mt6797/vendor/tree/ALPS-MP-M0.MP9-V1.78.1_NB6797_6C_M/mediatek/proprietary/bootable/bootloader/lk Mar 02 23:27:05 *much different Mar 02 23:27:08 aarch64, yes Mar 02 23:27:38 aarch64 argh...kexecboot is like broken right these days in OE Mar 02 23:27:51 kexec precisely, not kexecboot Mar 02 23:28:37 for fn you could just try th ekexecboot and kexec static binaries for arm we use for Zaurus Mar 02 23:29:25 I think the issue here is that Android 3.18 kernel isn't going to survive boot from kexec Mar 02 23:29:25 you might need these new options we added to kexecboot like hardboot Mar 02 23:29:51 but it requires device specific patches no ? Mar 02 23:30:02 *hardboot ? Mar 02 23:30:12 just configuration Mar 02 23:30:50 this is today: 2 commits totest before pushing kexecboot upstream Mar 02 23:30:51 The company also promises it’ll release the source code for the Gemini PDA’s bootloader and Linux kernel. Mar 02 23:31:00 jeez Mar 02 23:31:01 https://github.com/LinuxPDA/kexecboot Mar 02 23:31:03 ^ Mar 02 23:32:04 Tony Lindgren, old beagleboard developer, is now playing with droid, similar issues Mar 02 23:32:20 I forked kexeboot before this was added.. Mar 02 23:32:23 interesting Mar 02 23:34:08 this will be kexecboot 0.7 Mar 02 23:34:18 ant_home: the kexecboot hardboot option will only work with a kernel that has been patched for hardboot kexec though Mar 02 23:34:29 after that we plan(ned) to add different boot methods, like pivot, etc Mar 02 23:34:38 greguu, yes Mar 02 23:35:07 " Mar 02 23:35:07 Some devices require patched kernel and kexec tools." Mar 02 23:36:43 btw, I can cross-compile kexecboot for aarch64 here Mar 02 23:37:01 it is just kexec-tools recipe failing to build for this arm64 Mar 02 23:37:46 I am debugging the issue right now...aarch support was recently added to kexec-tools Mar 02 23:40:22 heh, it appears aarch64 needs a more recent kexec-tools-klibc ...ouch, this is not so easy Mar 02 23:43:50 I hope you can live with a musl-static recipe ;) Mar 02 23:43:58 lot of space on Gemini Mar 02 23:44:36 musl static should be fine I think Mar 02 23:44:55 it is just too large for Zaurus :/ Mar 02 23:45:09 the kexec binary ? Mar 02 23:45:12 yep Mar 02 23:45:29 so still klibc for Zaurus.. Mar 02 23:45:46 see, until devicetree Mar 02 23:45:56 no, do not mention it Mar 02 23:46:08 this old kexec just works :) Mar 02 23:46:32 but I miserably failed with the gcw0 Mar 02 23:46:54 the kernel was not shutting correctly down Mar 02 23:47:04 mipsel32 fwiw Mar 02 23:47:25 so I see the potential issues Mar 02 23:47:49 usually are custom drivers th ecuplrit Mar 02 23:48:07 hm, the mediatek kernel is full of those Mar 02 23:48:11 manufacturer sources, never reviewed upstream Mar 02 23:48:25 this is what we are dealing with unfortunately Mar 02 23:48:38 3.18.41 Mar 02 23:49:00 this will be a bit of work. for some things (mali) we may never get around libhybris I fear. Mar 02 23:49:29 gcwo had vivante, same prob Mar 02 23:49:53 a graphical boot menu would be anyway nice. I think I asked before, the UI can be driven by touchscreen right ? Mar 02 23:49:59 yes Mar 02 23:50:25 any input is parsed (or we can add it) Mar 02 23:50:53 well, NotKit needs a UART cable first to start testing.. all blind at the moment Mar 02 23:51:14 there is more debug right now Mar 02 23:51:41 speaking of which, Varti will get a Zaurus serial cable from dromede and may ship it to me :) Mar 02 23:52:09 nice, finally (it was my cable) Mar 02 23:52:19 lol Mar 02 23:52:28 we should all leave our initials on it Mar 02 23:52:40 I sent my spare to him when he once was working Mar 02 23:53:00 then he got domotics job and disappeared Mar 02 23:53:06 god for him Mar 02 23:53:12 *good Mar 02 23:53:15 without leaving the SDL rotation routines opensource... Mar 02 23:53:55 or was that not from him and he just compiled again that particular SDL version, I do not remember Mar 02 23:54:09 for zgrom Mar 02 23:54:19 I only have the distro not the sources Mar 02 23:54:25 of zgrom Mar 02 23:54:42 we talked and he was about pushing it Mar 02 23:55:00 it's just unfortunate timing I think Mar 02 23:55:14 I'd ask again Mar 02 23:55:27 yes, me too. In zgrom he uses a special SDL version. I think it was a similar one that ZodTTD used for his OpenTTD port Mar 02 23:58:03 lol " Mar 02 23:58:07 "I actually stopped developing for the Zaurus since I had to sell it for money I needed. I miss it though." Mar 02 23:58:23 https://www.oesf.org/forum/index.php?showtopic=21982&hl=Zodttd&st=15 Mar 03 00:03:15 ok, I'll look tomorrow at aarch64, hope to upload the three binaries (kexec kexecboot ubiattach) Mar 03 00:03:30 good nite **** ENDING LOGGING AT Sat Mar 03 03:00:00 2018