| vkoskiv | minute: If you have two trackpad modules available and have some time, build & flash this to one of them, plug both in, compare and let me know what you think: | 00:12 |
|---|---|---|
| vkoskiv | https://source.mnt.re/vkoskiv/reform/-/tree/trackpad-improvements | 00:12 |
| vkoskiv | For anyone else too, with a trackpad module, feel free to give it a go. I think it feels much more responsive now. | 00:12 |
| vkoskiv | I think it's possible to write an even better firmware that makes use of the multitouch HID feature set, I want to experiment with that too. | 00:13 |
| vkoskiv | Not fully finished yet, the mouse cursor will appear very nervous if you just hold your finger in one place :D | 00:21 |
| vkoskiv | (it shakes a bit) | 00:21 |
| vkoskiv | I think I just figured out the power saving timeout, I found the right timeout to disable | 00:27 |
| vkoskiv | So now the mouse doesn't stick for a brief moment when moving the cursor for the first time in a while | 00:27 |
| - Boostisbetter (QUIT: Ping timeout: 255 seconds) (4a410829d7@irc.cheogram.com) | 00:40 | |
| - colinsane (QUIT: Quit: bye) (~colinunin@97-113-128-229.tukw.qwest.net) | 00:55 | |
| + colinsane (~colinunin@97-113-128-229.tukw.qwest.net) | 00:57 | |
| + nocko (~nock@user/nocko) | 01:09 | |
| - vagrantc (QUIT: Ping timeout: 240 seconds) (~vagrant@2600:3c01:e000:21:7:77:0:20) | 01:15 | |
| - chomwitt (QUIT: Ping timeout: 264 seconds) (~chomwitt@ppp-94-67-206-212.home.otenet.gr) | 01:32 | |
| - mjw (QUIT: Ping timeout: 245 seconds) (~mjw@gnu.wildebeest.org) | 02:36 | |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:20) | 02:38 | |
| - vagrantc (QUIT: Ping timeout: 240 seconds) (~vagrant@2600:3c01:e000:21:7:77:0:20) | 02:50 | |
| noam | minute: thanks for the book recommendation, been having fun making my way through it :) | 03:03 |
| - nsc (QUIT: Ping timeout: 240 seconds) (~nicolas@i5E861476.versanet.de) | 04:02 | |
| + nsc (~nicolas@i5E861467.versanet.de) | 04:04 | |
| - ajr (QUIT: Quit: Connection closed for inactivity) (uid609314@user/ajr) | 04:10 | |
| - colinsane (QUIT: Quit: bye) (~colinunin@97-113-128-229.tukw.qwest.net) | 04:13 | |
| + colinsane (~colinunin@97-113-128-229.tukw.qwest.net) | 04:15 | |
| - jjbliss (QUIT: Ping timeout: 258 seconds) (~jjbliss@1464766-static.elnsmiaa.metronetinc.net) | 06:49 | |
| - robin (QUIT: Read error: Connection reset by peer) (~robin@user/terpri) | 07:12 | |
| + Boostisbetter (4a410829d7@irc.cheogram.com) | 07:14 | |
| + chomwitt (~chomwitt@2a02:587:7a17:8900:1ac0:4dff:fedb:a3f1) | 07:25 | |
| + jjbliss (~jjbliss@1464766-static.elnsmiaa.metronetinc.net) | 07:51 | |
| - Gooberpatrol66 (QUIT: Ping timeout: 240 seconds) (~Gooberpat@user/gooberpatrol66) | 07:57 | |
| + robin (~robin@user/terpri) | 08:03 | |
| + Gooberpatrol66 (~Gooberpat@user/gooberpatrol66) | 08:07 | |
| - colinsane (QUIT: Quit: bye) (~colinunin@97-113-128-229.tukw.qwest.net) | 09:51 | |
| + colinsane (~colinunin@97-113-128-229.tukw.qwest.net) | 09:52 | |
| - jjbliss (QUIT: Ping timeout: 255 seconds) (~jjbliss@1464766-static.elnsmiaa.metronetinc.net) | 10:05 | |
| * mark_ -> mjw | 10:09 | |
| + jjbliss (~jjbliss@1464766-static.elnsmiaa.metronetinc.net) | 10:18 | |
| - GNUmoon (QUIT: Remote host closed the connection) (~GNUmoon@gateway/tor-sasl/gnumoon) | 10:31 | |
| + GNUmoon (~GNUmoon@gateway/tor-sasl/gnumoon) | 10:36 | |
| minute | noam: oh cool! | 12:12 |
| minute | vkoskiv: great, we use a reform with trackpad in the office, for fulfillment. will flash it there | 12:12 |
| vkoskiv | The effect is somewhat subtle, so feel the cursor movement for a bit before flashing to get a better comparison | 13:07 |
| vkoskiv | Stock firmware uses the average of the last 4 delta values for the reported motion, which was the primary source of latency. I took that out entirely, and did a few other things too. | 13:09 |
| vkoskiv | Powersaving timeout disable isn't pushed yet, I think. So the ~30s idle timeout + slight delay in registering motion is still there | 13:10 |
| vkoskiv | If you have time to look at the diffs too, feedback is welcome. Happy to do any changes | 13:10 |
| dodo | should i still build the sysimage-v3 to reinstall, or has something changed? | 13:34 |
| vkoskiv | We're now on V4, I believe | 13:37 |
| josch | dodo: you want a fresh start with your reform? | 13:42 |
| josch | dodo: you can also just download the images from CI | 13:42 |
| dodo | yeah, i think my sd card is broken, at least i hope that's all.. | 13:53 |
| josch | dodo: you ran your reform entirely off of the sd-card? | 13:54 |
| dodo | no, but i think it was used in the boot process. it's been a while tbh | 13:55 |
| josch | dodo: what do you want to try out first? | 13:55 |
| minute | i have an issue with a fairly recent system image testing on imx8mq | 14:06 |
| josch | minute: the one with or without reform-tools 1.33? | 14:07 |
| minute | display works in uboot but disappears in linux | 14:09 |
| minute | josch: will check | 14:09 |
| minute | and there's no /dev/fb* | 14:09 |
| minute | ah Package: reform-tools | 14:09 |
| minute | Version: 1.32 | 14:09 |
| minute | ok, need to redownload? | 14:09 |
| minute | [ 0.000000] Linux version 6.5.0-2-reform2-arm64 (debian-kernel@lists.debian.org) (aarch64-linux-gnu-gcc-13 (Debian 13.2.0-2) 13.2.0, GNU ld (GNU Binutils for Debian) 2.41) #1 SMP Debian 6.5.6-1+reform20231010T093700Z1 (2023-10-10) | 14:09 |
| josch | that issue is probably only related to reform-tools if it's about the bootargs | 14:09 |
| josch | but those look okay? | 14:10 |
| minute | > [ 0.000000] Kernel command line: ro no_console_suspend cma=512M@3G cryptomgr.notests console=ttymxc0,115200 console=ttymxc0,115200 pci=nomsi console=tty1 | 14:10 |
| minute | argh | 14:10 |
| minute | again this module has 2GB... how is that possible | 14:10 |
| minute | i think something's wrong on our end here :D | 14:11 |
| minute | i will try to reflash uboot | 14:16 |
| + mark_ (~mjw@gnu.wildebeest.org) | 14:20 | |
| josch | plot twist: the problem is caused by a bug in lpddr4_pmu_train_*mem.bin but nobody can fix it because it's just a proprietary blob ;) | 14:21 |
| minute | josch: nope. it was an old/wrong uboot on those modules | 14:23 |
| minute | josch: reform-flash-uboot emmc fixed it! | 14:23 |
| josch | nice! | 14:24 |
| minute | josch: ok but can we change the commandline for imx8mq so that it has cma=512M instead of cma=512M@3G ? then people with 2GB modules and also if we shipped the wrong uboot will be able to recover | 14:24 |
| minute | the 512M@3G was necessary only for ls1028a | 14:24 |
| minute | orrr i think we can change it to 512M@1G and it should work everywhere | 14:25 |
| vkoskiv | What does that actually do? | 14:25 |
| minute | vkoskiv: it says where the CMA goes in memory. it's a 512M section that is reserved for linear allocation, which is needed by etnaviv | 14:26 |
| vkoskiv | Ah, and @ is the offset | 14:26 |
| josch | minute: yes, this can be changed in ubootenv.d/00reform2_ubootenv (for flash-kernel/boot.scr) as well as in u-boot-menu/reform.conf (for extlinux.conf) | 14:27 |
| minute | ok | 14:27 |
| josch | the other place where this value is encoded is in the bootargs for u-boot itself on each platform | 14:28 |
| minute | vkoskiv: yep | 14:28 |
| josch | and i just recently changed cma=512M -> cma=512M@3G in reform-boundary-uboot: https://source.mnt.re/reform/reform-boundary-uboot/-/commit/1f5ddefa070c608ee94553439ba7647bd331a73c | 14:28 |
| josch | so that can easily be reverted | 14:28 |
| minute | josch: ok cool | 14:29 |
| josch | and the @xG was necessary for ls1028a? but @1G also works there? | 14:30 |
| minute | josch: i think 1G should work because it has to be under the 32 bit boundary | 14:32 |
| josch | because otherwise, if this is a platform specific setting, it should be moved into the if/else in ubootenv.d/00reform2_ubootenv | 14:33 |
| josch | and u-boot-menu/reform.conf should not set it anymore but instead trust the value it gets from each u-boot | 14:33 |
| josch | minute: so should cma=512M@1G be done for every u-boot or just on imx? | 14:38 |
| minute | josch: actually if we can change it to cma=512M but just on imx we know that it should work | 14:49 |
| josch | okay, then i keep cma=512M@3G on all platforms except on imx where i put cma=512M | 14:52 |
| minute | josch: i think on a311d cma is not needed at all, but we can fix that later | 14:56 |
| minute | i can test it real quick | 14:57 |
| josch | that would be great, then i make these changes once for all :) | 14:57 |
| - mark_ (QUIT: Ping timeout: 255 seconds) (~mjw@gnu.wildebeest.org) | 15:05 | |
| bkeys | minute: Just wondering how far in the queue is my keyboard? | 15:07 |
| - qbit (PART: WeeChat 3.8) (~qbit@2602:ff16:3:0:1:3a0:0:1) | 16:40 | |
| + jacobk (~quassel@utdpat241106.utdallas.edu) | 17:11 | |
| Boostisbetter | anybody else not able to change how wide the contact list in on the Signal desktop client? | 17:26 |
| Boostisbetter | anybody else not able to change how wide the contact list is on the Signal desktop client? | 17:49 |
| + pandora (uid585533@id-585533.ilkley.irccloud.com) | 18:14 | |
| + mark_ (~mjw@gnu.wildebeest.org) | 18:27 | |
| - mjw (QUIT: Killed (NickServ (GHOST command used by mark_!~mjw@gnu.wildebeest.org))) (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 18:42 | |
| * mark_ -> mjw | 18:42 | |
| + mark_ (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 18:42 | |
| josch | vkoskiv: do you have some time to help test if a311d boots fine without cma=512M@3G? I created a system image with a boot.scr that does not supply cma=512M@3G on a311d which you would just have to flash to an sd-card and boot from and see what happens. | 19:45 |
| vkoskiv | Sure thing! | 19:46 |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50) | 19:49 | |
| - jacobk (QUIT: Ping timeout: 260 seconds) (~quassel@utdpat241106.utdallas.edu) | 19:50 | |
| josch | vkoskiv: cool! here it is: https://source.mnt.re/reform/reform-system-image/-/jobs/2374/artifacts/file/reform-system-a311d.img.gz | 19:52 |
| josch | if it does boot, please check /proc/cmdline whether cma=512M@3G really is missing | 19:53 |
| josch | if it's still there, either modify ${bootargs} in u-boot or (if that's a hassle for you) i'll give you a modified u-boot for a311d that doesn't set cma=512M@3G anymore | 19:53 |
| josch | you can flash a311d u-boot using: dd if=./flash.bin of=/dev/sdcard conv=notrunc bs=512 skip=1 seek=1 | 19:56 |
| josch | here is a flash.bin that doesn't contain cma=512M@3G anymore: https://source.mnt.re/reform/reform-a311d-uboot/-/jobs/2375/artifacts/browse | 19:56 |
| + jacobk (~quassel@129.110.242.224) | 19:58 | |
| + doctorhoo (~doctorhoo@2a00:801:717:bb7b:c16b:e94:9afe:38b5) | 20:15 | |
| - GNUmoon (QUIT: *.net *.split) (~GNUmoon@gateway/tor-sasl/gnumoon) | 20:16 | |
| vkoskiv | It boots, checking boot.scr in a moment | 20:19 |
| vkoskiv | cat-ing it was a mistake :D | 20:19 |
| vkoskiv | Ah, /proc/cmdline is better for this | 20:20 |
| doctorhoo | Hej! I have noticed my Reform to sometimes hang, but so far only after longer periods of inactivity (>1h). When it does, I would expect swaylock to appear but the screen stays dark. I then have to reboot via the circle key. The logs show nothing suspicious but also simply stop at some point. I connected another machine to SER1 and monitored the output of journalctl --system -f, but again, I only see some stuff related to WiFi | 20:21 |
| doctorhoo | (which is omnipresent) which stops at some point after which the output stops and the console does not react anymore. I have not touched the reform during this time (which can take several hours) until it hangs. | 20:21 |
| doctorhoo | I am on a fairly recent kernel but have not updated since mid August. | 20:21 |
| doctorhoo | Does anyone have suggestions on how to debug this further? | 20:21 |
| vkoskiv | josch: the cma bit is in /proc/cmdline still | 20:22 |
| doctorhoo | I am considering to run a prestine v4 from SD card as next step to try to see if it is my config or not. But beyond that I am a little out of ideas. | 20:22 |
| doctorhoo | Would appreciate any pointers or experiences you have had with similar crashes! | 20:24 |
| vkoskiv | doctorhoo: Is it maybe going to suspend? That may not always come back up | 20:24 |
| doctorhoo | vkoskiv: mhh, interesting point! Does it go to suspend by itself? I had not noticed that. Circle-space does not help, but that still could be a borked suspend. However, should the logs not indicate that? Also, the time seems random: anything between 1h or so to more than 12h | 20:27 |
| vkoskiv | josch: flashed that flash.bin, still boots, and no cma=512M@3G in /proc/cmdline! | 20:27 |
| doctorhoo | I never had it happen while I was using it -- but that could be chance as well | 20:27 |
| vkoskiv | doctorhoo: It won't suspend unless you've configured it to do that | 20:28 |
| doctorhoo | vkoskiv: ok, then it should not be that :) | 20:28 |
| doctorhoo | I was wondering whether journalctl is the right tool for monitoring -- would that reliably report logging info even when e.g. the nvme fails somehow? | 20:32 |
| doctorhoo | Would I still expect to see something dumped to SER1 before the system halts completely (e.g. some kernel dump)? | 20:33 |
| vkoskiv | minute: I still seem to get network drops on 2.4GHz | 21:46 |
| - doctorhoo (QUIT: Read error: Connection reset by peer) (~doctorhoo@2a00:801:717:bb7b:c16b:e94:9afe:38b5) | 21:48 | |
| + doctorhoo (~doctorhoo@217-210-162-155-no600.tbcn.telia.com) | 21:49 | |
| josch | doctorhoo: did you configure some mechanism of screen locking or the screen switching off after a while? | 21:57 |
| josch | vkoskiv: thank you!! :) | 21:57 |
| vkoskiv | Weird, I'm having an issue where the hidraw file for the trackpad disappears while programming it | 22:02 |
| vkoskiv | But nobody on the internet seems to have encountered this | 22:03 |
| vkoskiv | There was one, but their solution didn't apply here | 22:03 |
| vkoskiv | I don't feel comfortable yanking the USB connector, since that connector isn't designed for hot-plug | 22:03 |
| vkoskiv | So I have to reboot whenever that happens | 22:03 |
| minute | josch: vkoskiv: sorry i got sidetracked earlier. can also confirm that a311d works fine without any cma. | 22:11 |
| vkoskiv | minute: No worries at all, you have what, over 9000 items on your to-do list at any given time? :D | 22:12 |
| josch | minute: no worries, you better concentrate on the things that only you can do and leave the other stuff to us :) | 22:12 |
| minute | haha ok :D | 22:12 |
| josch | i swear we didn't prepare our answers XD | 22:12 |
| minute | ahaha | 22:13 |
| vkoskiv | Disabling the LP1 means we no longer enter LP1 at 1uA after about ~30sec, and instead stay in the idle state, which is 2.8mA | 22:18 |
| doctorhoo | josch: just swaylock, nothing more off the top of my head -- and there is no clear "after X minutes" to it: in my tests, it sometimes survives the night and sometimes already hands after little more than an hour. Not sure what could trigger it. WiFi is dropping off and reconnecting quite regularly, I have the browser (Firefox) usually open, but nothing producing high loads or scheduled activity that I remember now (but I have | 22:18 |
| doctorhoo | been using the system for a bit now, pretty much since v3 was released) | 22:18 |
| vkoskiv | But disabling it stops the mouse from freezing briefly when mousing for the first time to wake it from LP1 | 22:18 |
| josch | doctorhoo: my reform switched off when idle when i had it turn off the display on idle | 22:19 |
| josch | doctorhoo: so your display remains turned on but swaylock gets activated? | 22:20 |
| vkoskiv | I pushed the LP1 disable commit, but I can remove that if we want to prioritise power consumption | 22:21 |
| doctorhoo | Josch: that I have to look into! I think it might switch it off; in fact, there is usually a line in the logs that I don't recall in detail whenever I reactivate it -- so that could be a lead! Thanks, I will investigate further! :) | 22:26 |
| doctorhoo | josch: one follow up question though: by switch off you mean crash/hang or actually power down? | 22:27 |
| josch | doctorhoo: according to the oled it is on but it's completely unresponsive to keyboard input and not reachable via ssh or icmp | 22:29 |
| josch | doctorhoo: for example using "output eDP-1 dpms off" or "output eDP-1 disable" resulted in random lock-ups | 22:39 |
| josch | doctorhoo: so now instead i'm using "brightnessctl set 0" | 22:40 |
| - jacobk (QUIT: Ping timeout: 255 seconds) (~quassel@129.110.242.224) | 22:48 | |
| eery | josch: a small data point, I've had this problem too and used your solution - it (DPMS) still seems to cause lockups on the A311D, after I changed SOMs | 23:06 |
| josch | oh interesting! | 23:08 |
| eery | haven't really tested it besides re-enabling DPMS, having my system lockup, and going back to the backlight hack | 23:09 |
| eery | so it could have been something else, but | 23:09 |
| + jacobk (~quassel@utdpat241106.utdallas.edu) | 23:55 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!