Monday, 2026-05-18

nightishamanmal: what do you think00:09
nightishamanim gonna go to sleep but we probably have found the problem00:23
malnightishaman: https://xairy.io/articles/pixel-kgdb#-apc-watchdog00:43
T42<nightishaman> But I didn’t even attach a debugger, wtf00:45
T42<nightishaman> Well, I’m probably just gonna disable the kernel modules in gs201_defconfig00:45
T42<nightishaman> We don’t need the whole debug infrastructure in place00:45
T42<nightishaman> Also im on telegram since I don’t have irc on my daily driver phone00:46
malyou could try that kernel command line option00:46
malthat is the quickest thing to try00:47
malyou can add to suitable place in device repos00:48
malsome place there defines the kernel command line00:49
malin BoardConfig.mk or similar file00:49
T42<nightishaman> I know, it’s in BoardConfig-common.mk for gs20100:49
malanyway, time for some sleep00:49
T42<nightishaman> Although do we really need the debug infrastructure? I don’t see the point in keeping those modules00:50
mallet's not do too many changes at once00:51
T42<nightishaman> okay, but if it is crashing again tomorrow I’m gonna remove them00:54
*** xmn_ is now known as xmn13:25
nightishamanmal: it reboots, so apparently the kernel command line args work20:04
malnightishaman: how soon does it reboot?20:09
nightishamanafter about 3s20:09
malyou have some reboot command in the init-script now?20:09
nightishamanyeah, after 25 modules it reboots20:10
nightishamani increase now to 5020:10
nightishamanwith 50 it did reboot now one time with BL1 requested, the next time it hangs.20:14
nightishamanah now after a minute it rebooted, also with bl1 requested20:15
nightishamanaccording to bootloader it was the apc watchdog again20:16
malwhich partitions did you flash?20:17
malcheck using that unpack_bootimg.py which one contains the parameter, use --format mkbootimg if other you can't see the command line20:18
nightishamani flashed vendor_boot and init_boot20:20
nightishamanthe command line is set in vendor_boot20:21
nightishamanwith 25 it did work, while yesterday it didnt, so i guess its another one of the debug kernel modules20:24
nightishamani think the best way forward would be to remove these20:24
T42<b100dian> rinigus: sorry I missed your silent ping again. Android shows something like "Add plan" in settings for devices with eSIM. From there there's a app by google that is launched..20:56
T42<b100dian> mal: still debugging video playback, but found an interesting thing: waydroid on sake did not play back until I didn't cherry-pick this https://github.com/waydroid/waydroid/issues/181320:57
T42<b100dian> Some forum people say playback with appsupport is impacted on nagara, maybe appsupport needs those nodes too.20:58
nightishamanmal: what do you think?20:59
nightishamanunfortunately, some other modules depend on some of the debug modules, so i cant just remove all of them directly21:14
mal@b100dian interesting, need to check21:31
malnightishaman: removing those would require doing those from defconfig and then rebuilding boot images21:31
malbut still very odd why some modules would breaking it like that21:32
malnightishaman: did the commandline change work or not? if I understood what you wrote it didn't, not sure what that bl1 you said means21:33
nightishamanmal: the change did work, it loaded the first 25 modules but increasing to 50 made it hang again, so something else hangs21:40
nightishamanbl1 means bootloader requested by software21:40
nightishamanclaude suggested just using a skip list in the init script first, since exynos coresight isnt a dependency of another module22:18
T42<b100dian> rinigus: forgot to respond - I don't have a lpac build, but pretty sure this would work on OBS too https://github.com/juanro49/harbour-lpac.22:34
T42<b100dian> SimPro I didn't find the sources. But `LPAC_APDU_GBINDER_UIM_SLOT=2 LPAC_APDU=gbinder lpac profile download` should work just the same22:35
mal@b100dian about that droidmedia PR, does that std::max(4, android::BufferQueue::NUM_BUFFER_SLOTS / 2); make sense? isn't that always android::BufferQueue::NUM_BUFFER_SLOTS / 223:18
T42<b100dian> Probably, I dont have all your devices, maybe there are devces with less slots23:19
T42<b100dian> We could just use the MAX value in fact23:20
malbut that is constant in android header23:20
maland what I could find NUM_BUFFER_SLOTS is 6423:22
T42<b100dian> MAX is NUM_BUFFER_SLOTS - 223:23
nightishamanso with skipping exynos-coresight-etm I got to 30, at 50 i got 2 hangs in 9 reboots, so 7 times it rebooted immediately fine23:24
T42<b100dian> mal: heading to bed, was gonna mention the PR tomorrow. I would be interested if fp5 with media buffers works too23:25
mal@b100dian yes but your code would always try to first set 32 and then 4 if the first one failed23:25
malsure, I can test23:25
T42<b100dian> I was being conservative23:26
malin the change I did earlier I originally planned 16 but then after some testing with a device with different issue 4 was enough there, didn't think it would affect fp5 also23:27
T42<b100dian> Lets see23:29
nightishamanthis is very annoying to debug23:39
nightishamani purchased a usb cereal but its only coming by end of month23:39
malnightishaman: have you found which module hangs it this time?23:44
nightishamanno idea, i found out i can enable some kind of ramdump with fastboot oem ramdump enable23:45
nightishamanbut no idea where to read it out23:45
nightishamanok i just had a black screen with red title saying "Preparing for ramdump"23:45
nightishamani can pull kernel log from ramdump23:51
T42<Mister_Magister> ye just run strings on it23:59

Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!