2025-11-25.log

- mjw (QUIT: Ping timeout: 240 seconds) (~mjw@gnu.wildebeest.org)00:00
- robin (QUIT: Remote host closed the connection) (~robin@user/terpri)00:02
+ robin (~robin@user/terpri)00:02
minutejosch: i've merged all the things i think. many thanks for this initiative and being so thourough about it!00:38
joschminute: oh wow, thank you!00:40
joschhehe i found a bug in the i.mx8mq u-boot XD00:43
joschi hope you didn't start tagging yet00:43
sigridis there anything special to do if I want lpc (circle+space) to wake up rk3588?00:45
sigridthis is on alpine, I am not sure what official image does00:45
sigridI have lpc driver built in but I guess it didn't do anything when I hit circle and space00:46
kop316So after an update, my pocket reform seems to drop into rescue mode, is that happening to anyone else?00:48
kop316it looks like hitting ctrl-D works just fine, but that's kind of annoying that happened00:50
joschsigrid: uarts have to be configured as a wakeup source from suspend00:51
joschaccording to Documentation/devicetree/bindings/serial/fsl-imx-uart.yaml, setting wakeup-source to "true" should be enough but somehow it is not00:51
joschso we have this udev rule:00:51
joschACTION=="add", SUBSYSTEM=="tty", KERNEL=="ttymxc*", RUN+="/bin/sh -c 'echo enabled >/sys/$env{DEVPATH}/power/wakeup'"00:51
joschkop316: there was a new reform-tools release last week but you are the first to report a problem. Do you have any kind of error message which got you to the emergency shell?00:52
kop316josch: yeah I just updated the reform-tools. journalctl -xd didn't show any error, and when I git ctrl+d it booted normally00:53
joschkop316: does the issue happen on every boot?00:54
kop316it almost seems like the boot script added a rescue command to booting? but I am not familiar with arm booting enough to look at that (yet, I need to make dinner first)00:54
kop316josch: yes00:54
kop316I thought I rebooted after updating and didn't see it, but the second boot (hard turn off and on) I saw it00:55
joschif you hard turn off, then the filesystem might've been currupted and that's why you get the emergency shell00:55
kop316josch: I would agree, but I succesfully booted and then turned off and on, and it still happens00:56
kop316and sorry, by hard turn on and off I mean shutdown and power on, not just reboot00:56
kop316not cut power00:57
joschOkay. If you have some error messages that would be helpful. :)00:57
sigridjosch: thanks01:02
minutesigrid: AFAIK this should already be set up01:02
sigridif it's supposed to happen automatically by lpc driver, it didn't seem to happen01:03
minutesigrid: josch: rk3588 can't really wake up from uart i think (in my experiments). you want to set up the wakeup GPIO. i recently did that in dts and then undid it because i wasn't able to fully test it and someone reported random space key presses that might have come from that01:03
minutesigrid: lpc basically only toggles a gpio line. and linux can set up that line as a wakeup source for rk3588. this is usually done through dts. one second01:04
kop316josch: the journalctl -xb log is 400 kb, over the limit for paste.debian.net. what would be the easiest way to get it to you?01:05
minutesigrid (cc josch): https://source.mnt.re/reform/reform-debian-packages/-/commit/61178751593bbe00420b46c2d0814091f8e5f68e#345c79e080c85877f577144169bde279b3cca9c6_50_5001:05
sigridaha, cool. I am going to try this one the kernel builds01:06
sigrid*once01:06
minutesigrid: i removed this shortly after. i think it needs some pinctrl to make the line not float, and remove the autorepeat. also not sure which linux,code = ... number actually makes sense (should be a keyboard scan code?)01:06
minutesigrid: but i tested it and it did wake up the rk358801:07
minutethe wakeup gpio pulse is in the lpc firmware itself, since a longer time https://source.mnt.re/reform/reform/-/blob/master/reform2-lpc-fw/src/boards/reform2/board_reform2.c?ref_type=heads#L87201:08
minuteand since a few months (?) also in pocket sysctl fw01:08
joschkop316: pipe it into: curl -F 'arg=<-' https://mister-muffin.de/paste01:09
joschotherwise email :)01:09
sigridminute: 57 seems to be scancode for space, right?01:11
kop316josch: https://mister-muffin.de/p/rP_K.txt and https://mister-muffin.de/p/9cDI.txt01:11
sigridI guess the "down" scancode for it would be 18501:13
sigridsorry, "up"01:14
sigridso maybe "linux,keycodes = <57>, <185>;" would make it not repeat01:15
joschkop316: Found it. You are booting with the kernel argument "single" and that gets you into rescue mode.01:23
joschkop316: do you have an uart adapter so that you can check what is going on before linux gets loaded?01:25
minutesigrid: there's down and up codes? do you have a table for this somewhere?01:28
minutesigrid: the main thing that needs to be added is a pinctl that puts a pulldown on the gpio, and the "autorepeat" removed01:29
sigridI just did doas showkey --scancodes and started mashing keys01:29
sigridhow would that pinctl look like?01:29
minutesigrid: ah, so up = scancode + 0x80... weird! reminds me of midi or something01:30
minute(maybe ps/2? it rings some kind of bell)01:30
joschkop316: your problem might be this: https://mntre.com/reform-irc-logs/2025-10-30.log.html#t16:54:3201:31
joschkop316: with the system booted, can you give me the output of this:01:31
joschcat /proc/device-tree/chosen/u-boot,version01:31
joschkop316: and the output of this:01:31
joschcat /proc/cmdline01:31
minutesigrid: in &pinctrl { ... } add a new entry like: wakeup { wakeup: wakeup { rockchip,pins = <4 RK_PB2 RK_FUNC_GPIO &pcfg_pull_down>; } }01:33
minutesigrid: then reference it in gpio-keys { ... } like: pinctrl-0 = <&wakeup>; pinctrl-names = "default";01:34
minutesigrid: probably it can't really be wakeup { wakeup: wakeup but some variation where each phandle name is unique (sth like wakeup-pins { wakeup: lpc-wakeup-pin { ... } } or so.01:36
minute(it's all made up)01:37
minutei don't actually know why or when the "bar" part of foo: bar { ... } is needed01:38
sigridyeah this whole device tree syntax to me looks like alien speak01:40
minutelets see https://devicetree-specification.readthedocs.io/en/v0.1/source-language.html01:41
minute[label:] node-name[@unit-address] {01:41
minuteinteresting, properties can also have a label01:41
sigridhttp://okturing.com/src/16121/body01:42
sigridthis is what I have rn01:42
minutesigrid: the second gpio-keys { ... } is wrong, otherwise looks correct01:43
minutesigrid: instead of a second gpio-keys { ... } that should be in &pinctrl { ... }01:43
sigridso just lpc_wakeup: ...{ } inside &pinctrl?01:44
minutesigrid: where there's currently stuff like backlight { backlight_enable: ... } and pcie reset stuff01:44
minutesigrid: don't you have a &pinctrl in the dts already? but yeah.01:44
sigridI mean I placed it the same way as backlight and pcie2 etc01:44
sigridin its own "category"(?)01:45
minutesigrid: ahh now i see01:45
minutesigrid: sorry, i think that's good then!01:45
sigridok, building now :)01:45
minutefingers crossed!01:46
minutesigrid: you can also check somewhere in sysfs if and how many times this wakeup signal was seen01:46
minutesigrid: so if you mash the wake function on the keyboard, a counter increases, so you know it's working01:46
sigridmust be /sys/class/wakeup/*/wakeup_count01:46
minuteyess01:47
kop316josch: cat /proc/device-tree/chosen/u-boot,version 2024.01-g889c316b-dirty01:48
joschminute: context switch about plymouth: did we agree that plymouth.ignore-serial-consoles should be added to the default cmdline on all platforms so that plymouth no longer considers serial the primary console and then changes to "details" theme?01:49
kop316josch: cat /proc/cmdline ro no_console_suspend cryptomgr.notests ro no_console_suspend cryptomgr.notests ro no_console_suspend cryptomgr.notests clk_ignore_unused cma=256M swiotlb=65535 fbcon=rotate:3 console=tty1 console=tty1 single console=tty101:49
joschkop316: thank you -- i'll add something to reform-check to find the "single" cmdline parameter01:49
minutejosch: yesss01:49
joschokay, next release will have it01:49
minutejosch: great!01:49
kop316josch: is there anything else you need from me? or should I wait for something else?01:50
joschkop316: that's a very old u-boot version. Can I send you a command to dump your u-boot binary for me?01:50
minutekop316: ohh that's the reason for (pocket) reform going into rescue mode after unclean shutdown :001:50
kop316josch: sure!01:51
minutei didn't know that a "single" could be snuck in temporarily01:51
joschminute: the next reform-check version will complain about outdated u-boot more aggressively01:51
minutewhy outdated uboot?01:51
minutei don't think uboot causes that...?01:51
minuteor does it?01:51
joschminute: the problem is that old u-boot on imx8m+ and rk3588 misses the ${fdtfile} u-boot environment variable01:52
kop316how do I update it though? I figured there'd be auto updates01:52
minuteah sorry, i thought we were talking past each other01:52
joschminute: if the user then has u-boot-menu installed, both menu options will fail because without ${fdtfile} it cannot find the device tree01:52
minuteah ok01:52
minuteso then that's a different problem than i thought, sorry!01:52
joschit still boots because even though extlinux.conf fails, it will fall back to boot.scr01:52
josch*but* it will keep the ${bootargs} that were set in extlinux.conf01:52
joschand the second extlinux.conf entry is the rescue system with "single" in ${bootargs}01:53
minuteah01:53
joschkop316: no, u-boot does not get auto-updated01:53
kop316josch: ahh....my mistake01:54
joschu-boot is not a package or a file on your filesystem but a program at a certain location on your flash01:54
joschand overwriting that could theoretically soft-brick your device01:54
kop316ohh so you don't want to risk it01:54
josch(i think this only happened once so far for somebody who experimented)01:54
kop316automatically anyways01:54
joschyes01:54
kop316how can i manually do it then? assuming i can after i dump what I have01:55
joschkop316: sudo dd if=/dev/mmcblk0 bs=512 skip=64 count=19000 | xz | curl -F 'arg=<-' https://mister-muffin.de/paste01:59
kop316josch: https://mister-muffin.de/p/woVD02:00
joschthank you. You are running u-boot 2024-10-11b which is old but not old enough for not having ${fdtfile} set02:06
joschkop316: you could upgrade to a more recent u-boot version but in case you have time and a uart adapter it would be very interesting to see *why* extlinux fails02:07
joschjust to make sure, does this file exist:02:07
josch/boot/dtbs/$(uname -r)/rockchip/rk3588-mnt-pocket-reform.dtb02:07
+ paperManu_ (~paperManu@107.159.15.124)02:07
kop316josch: by the uart adaptor do you mean a 6 (or 8, i forget) adaptor? I think I have one and can look later tonight if you tell me where to plug it in02:08
kop316and are there official instructions I can find to update it?02:09
joschkop316: updating u-boot is done by running: "sudo reform-flash-uboot emmc"02:09
joschkop316: docs for how to connect to serial are here: https://mntre.com/documentation/pocket-reform-handbook/advanced.html#serial-console02:09
kop316ahh perfect! thanks02:10
joschoups02:23
joschminute: so i was testing the GUID partitioning MR and i noticed that the partition table is broken in some cases02:23
joschturns out "some cases" is LS1028A because u-boot is written at an offset of only 4096 bytes02:23
joschand guid partitioning needs around 17408 bytes02:24
joschbut maybe LS1028A also tries reading u-boot from a different offset?02:25
joschah it's not u-boot at that offset but atf02:26
joschi'll leave a comment in that MR02:27
kop316josch: I think I see the issue, the dtb file is 0 bytes02:29
kop316the dtb,bak is about 97 kB02:30
joschyou mean that /boot/dtbs/$(uname -r)/rockchip/rk3588-mnt-pocket-reform.dtb is zero bytes?02:31
joschkop316: maybe the filesystem got corrupted?02:32
kop316josch: correct02:32
joschi'll also add a check for that then02:32
kop316yeah maybe, something weird happened02:32
kop316I think I can reinstall the image file and thatll fix it though02:32
joschkop316: you can unmount /boot and then check the filesystem for errors02:33
kop316josch: reinstalling the kernel image fixed it02:37
joschvery good02:38
joschyou already tried rebooting?02:38
kop316yes02:45
kop316the boot fs was clean too....weird02:45
kop316maybe some really weird one off error?02:46
kop316thank you for the help regardless!!02:46
joschkop316: thank you for helping get to the bottom of it. Next time this happens there will be some more sanity checks in reform-check to diagnose this faster. :)02:48
kop316awesome! glad I could help too02:52
* xktr_ -> xktr02:53
- f_ (QUIT: Read error: Connection reset by peer) (16abab341f@postmarketOS/funderscore)03:03
+ f_ (16abab341f@postmarketOS/funderscore)03:03
- kop316 (QUIT: Remote host closed the connection) (m-6f6zq6@static.138.159.90.157.clients.your-server.de)03:05
+ kop316 (m-6f6zq6@static.138.159.90.157.clients.your-server.de)03:07
- kop316 (QUIT: Remote host closed the connection) (m-6f6zq6@static.138.159.90.157.clients.your-server.de)03:08
+ kop316 (m-6f6zq6@static.138.159.90.157.clients.your-server.de)03:08
- paperManu_ (QUIT: Ping timeout: 240 seconds) (~paperManu@107.159.15.124)03:10
- paperManu (QUIT: Ping timeout: 265 seconds) (~paperManu@107.159.15.124)03:11
joschminute: i tested nearly all u-boot artifacts produced by MNT gitlab CI on classic reform (no testing with pocket) and only a311d hasn't been tested yet (because i'm using a311d to write this)03:24
joschthe only problems i found was the ls1028a problem with guid partitioning (see above)03:25
joschand this MR for imx8mq: https://source.mnt.re/reform/reform-boundary-uboot/-/merge_requests/22i03:25
joschhttps://source.mnt.re/reform/reform-boundary-uboot/-/merge_requests/2203:25
joschthere is a bunch of new stuff in reform-tools to accommodate for the new versioning scheme and some various bug fixing (see commit messages if you are curious)03:26
joschminute: unfortunately i'll probably only have time again for all of this next week, so i'll be a bit afk until then03:26
+ bkeys (~Thunderbi@98.19.131.193)04:32
kop316josch: as an FYI, the newest reform prevents the reboot issue once logged in on battery, but i got it reliably when booting up the pocket reform04:59
kop316reform-tools*05:04
- Ar|stote|is (QUIT: Quit: https://quassel-irc.org - Chat comfortably. Anywhere.) (~linx@149.210.3.83)05:25
- potatoespotatoes (QUIT: ) (~quassel@user/potatoespotatoes)05:27
+ potatoespotatoes (~quassel@user/potatoespotatoes)05:28
- potatoespotatoes (QUIT: Client Quit) (~quassel@user/potatoespotatoes)05:28
+ potatoespotatoes (~quassel@user/potatoespotatoes)05:30
- potatoespotatoes (QUIT: Client Quit) (~quassel@user/potatoespotatoes)05:34
+ Ar|stote|is (~linx@149.210.3.83)05:34
+ potatoespotatoes (~quassel@user/potatoespotatoes)05:36
+ kxtells (~jordi@79.117.156.55)08:26
+ mjw (~mjw@gnu.wildebeest.org)09:30
- jacqueline (QUIT: Remote host closed the connection) (~jacquelin@user/jacqueline)10:41
+ jacqueline (~jacquelin@user/jacqueline)10:41
- jacqueline (QUIT: Remote host closed the connection) (~jacquelin@user/jacqueline)10:52
asjacksonwoop - new charger board just arrived, mnt pocket is back in action 😎 12:11
+ paperManu (~paperManu@107.159.15.124)12:23
+ XYZ_ (~XYZ@37-48-16-111.nat.epc.tmcz.cz)12:46
+ gustav25 (~gustav@c-78-82-53-204.bbcust.telenor.se)13:02
- pomel0 (QUIT: Ping timeout: 264 seconds) (~pomel0@user/pomel0)13:33
+ pomel0 (~pomel0@user/pomel0)13:34
- paperManu (QUIT: Ping timeout: 264 seconds) (~paperManu@107.159.15.124)14:00
- natalie (QUIT: Ping timeout: 256 seconds) (~natalie@user/natalie)14:04
+ natalie (~natalie@user/natalie)14:08
- mjw (QUIT: Ping timeout: 264 seconds) (~mjw@gnu.wildebeest.org)14:37
joschminute: i now also successfully tested the latest gitlab CI artifacts for A311D. The only remaining open MR is https://source.mnt.re/reform/reform-boundary-uboot/-/merge_requests/22 I think afterwards you can "git tag" all five u-boot repos for a311d, imx8mq, imx8m+, ls1028a and rk3588 and then i can cut a new reform-tools release with the respective checksums.14:38
+ paperManu (~paperManu@modemcable141.205-200-24.mc.videotron.ca)14:50
* Guest1058 -> mjw14:52
- pomel0 (QUIT: Ping timeout: 260 seconds) (~pomel0@user/pomel0)14:57
+ pomel0 (~pomel0@user/pomel0)14:58
- potatoespotatoes (QUIT: ) (~quassel@user/potatoespotatoes)15:08
+ potatoespotatoes (~quassel@user/potatoespotatoes)15:09
- pomel0 (QUIT: Ping timeout: 240 seconds) (~pomel0@user/pomel0)15:09
+ pomel0 (~pomel0@user/pomel0)15:10
- pomel0 (QUIT: Ping timeout: 246 seconds) (~pomel0@user/pomel0)15:33
+ pomel0 (~pomel0@user/pomel0)15:34
- potatoespotatoes (QUIT: ) (~quassel@user/potatoespotatoes)15:58
+ potatoespotatoes (~quassel@user/potatoespotatoes)16:00
- potatoespotatoes (QUIT: Client Quit) (~quassel@user/potatoespotatoes)16:04
+ potatoespotatoes (~quassel@user/potatoespotatoes)16:05
- kxtells (QUIT: Remote host closed the connection) (~jordi@79.117.156.55)16:43
+ kxtells (~jordi@79.117.156.55)16:44
- pomel0 (QUIT: Ping timeout: 240 seconds) (~pomel0@user/pomel0)16:57
+ pomel0 (~pomel0@user/pomel0)16:58
+ wielaard (~mjw@2001:1c06:2486:a800:158:adb1:a1e7:3c3f)17:34
* mjw -> Guest535917:43
- Guest5359 (QUIT: Killed (erbium.libera.chat (Nickname regained by services))) (~mjw@2001:1c06:2486:a800:7602:5eff:dc71:a72c)17:43
* wielaard -> mjw17:43
+ Guest5359 (~mjw@2001:1c06:2486:a800:7602:5eff:dc71:a72c)17:43
- Guest5359 (QUIT: Client Quit) (~mjw@2001:1c06:2486:a800:7602:5eff:dc71:a72c)17:46
- pomel0 (QUIT: Ping timeout: 240 seconds) (~pomel0@user/pomel0)18:32
+ pomel0 (~pomel0@user/pomel0)18:32
- pomel0 (QUIT: Ping timeout: 244 seconds) (~pomel0@user/pomel0)19:12
+ pomel0 (~pomel0@user/pomel0)19:12
- mjw (QUIT: Ping timeout: 244 seconds) (~mjw@2001:1c06:2486:a800:158:adb1:a1e7:3c3f)19:25
- pomel0 (QUIT: Ping timeout: 246 seconds) (~pomel0@user/pomel0)19:31
+ mjw (~mjw@2001:1c06:2486:a800:158:adb1:a1e7:3c3f)19:34
+ vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50)19:40
+ pomel0 (~pomel0@user/pomel0)20:26
- plomlompom (QUIT: Remote host closed the connection) (~plom@139.59.138.73)20:42
+ plomlompom (~plom@139.59.138.73)20:43
elbso I just noticed that I was added to the Reform project on source.mnt.re, which grants me access to the Reform project, but not the Reform _group_, which would grant me access to Pocket Reform, which is what I'm actually qualified to do anything with; I'm not sure if this was an error or if I should even have been added to anything in the first place, but I feel like Reform (group) or Pocket Reform (project) makes more sense than 20:44
elbReform (project)20:44
elbNB that I have no intention of doing anything that I couldn't do before anyway, so this is largely academic, but I'm literally an academic, so ...20:45
elbunrelated to that I just filed a non-critical bug for reform-tools, !13, keyboard backlight is not entirely under user control any more ;-)20:46
sigridminute: tried out the kernel with gpio wakeup - the wake up trigger works (the count in /sys/power/wakeup_count increases) but I guess the kernel isn't ready to properly sleep/resume from s2ram yet20:56
sigridI used "linux,code = <KEY_WAKEUP>;" btw. seemed more fitting. it does increase the count by 4 every time, however, I guess because the fw does exactly that?20:57
minutesigrid: ah key_wakeup sounds optimal, thanks. and yes @ 4 times21:03
minutesigrid: have you tried s2idle?21:03
minuteelb: ah, probably my mistake @ project assignment21:04
sigridI assume s2idle is "echo mem | doas tee /sys/power/state"?21:04
sigridthat's what I did21:04
sigridhmmm collabora repo has a bunch of new patches in rockchip-devel specifically related to suspend/resume. I'm going to pull them in and test again21:17
minutesigrid: no that's not s2idle21:37
minutesigrid: s2idle is one of the modes of mem_sleep21:38
minute(i just had to google the right sysfs file again)21:39
minutesigrid: so you echo s2idle > mem_sleep and then echo mem > state21:39
sigridoh21:40
sigridsame result21:42
jfredjust want to say thanks to everyone looking into suspend/resume on the reform line :)21:45
qbitrevived 8 batteries \o/22:03
qbitI don't trust them farther than I can throw um.. but they can power my reform now!22:03
vagrantcought to be able to throw them pretty far... they tend to have a nice density22:14
vagrantcat least with the 18650 cells :)22:14
- gustav25 (QUIT: Quit: Quit) (~gustav@c-78-82-53-204.bbcust.telenor.se)22:15
vagrantcbattery pouches are probably less aerodynamic...22:15
qbit:D22:32
jfreddepends, thrown like a frisbee maybe it'll go pretty far :)22:33
qbitmight even be able to skip a pouch22:35
- paperManu (QUIT: Ping timeout: 255 seconds) (~paperManu@modemcable141.205-200-24.mc.videotron.ca)22:54
+ paperManu (~paperManu@modemcable141.205-200-24.mc.videotron.ca)22:55

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!