| - arminweigl (QUIT: Ping timeout: 276 seconds) (~arminweig@sourcehut/user/arminweigl) | 00:12 | |
| bremner | josch: I might have omitted it from my blog post, but I automagically had a login prompt on the serial terminal, no extra setup | 00:44 |
|---|---|---|
| + arminweigl (~arminweig@sourcehut/user/arminweigl) | 00:44 | |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50) | 01:09 | |
| - svp (QUIT: Remote host closed the connection) (~svp@host-79-7-240-189.business.telecomitalia.it) | 02:15 | |
| - mjw (QUIT: Ping timeout: 276 seconds) (~mjw@gnu.wildebeest.org) | 02:34 | |
| - vagrantc (QUIT: Ping timeout: 272 seconds) (~vagrant@2600:3c01:e000:21:7:77:0:50) | 03:14 | |
| - nsc (QUIT: Ping timeout: 248 seconds) (~nicolas@153-99-142-46.pool.kielnet.net) | 03:30 | |
| + nsc (~nicolas@129-97-142-46.pool.kielnet.net) | 03:32 | |
| - op_4 (QUIT: Remote host closed the connection) (~tslil@user/op-4/x-9116473) | 04:05 | |
| + op_4 (~tslil@user/op-4/x-9116473) | 04:05 | |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50) | 04:12 | |
| - jogu (QUIT: Remote host closed the connection) (~jogu@user/jogu) | 05:38 | |
| josch | bremner: yes, because you had console=ttyS2,1500000 in your cmdline | 05:41 |
| josch | systemd-getty-generator will automatically add a getty for the console= parameter you have there and by default it's console=tty1 to make sure that things like the luks prompt show up on the screen and not on serial | 05:44 |
| - aloo_shu_ (QUIT: Read error: Connection reset by peer) (~aloo_shu@85.51.16.194) | 06:46 | |
| - vagrantc (QUIT: Quit: leaving) (~vagrant@2600:3c01:e000:21:7:77:0:50) | 06:49 | |
| + aloo_shu (~aloo_shu@85.51.16.194) | 07:52 | |
| + gustav28 (~gustav@c-78-82-54-137.bbcust.telenor.se) | 10:02 | |
| + svp (~svp@host-79-7-240-189.business.telecomitalia.it) | 10:05 | |
| bremner | josch: ah, so the idea is to have getty even if not using the serial port as console? | 11:09 |
| josch | bremner: This is all i have for context: https://mntre.com/reform-irc-logs/2025-07-01.log.html#t11:51:38 | 11:14 |
| bremner | hmm. | 11:15 |
| bremner | I guess "password prompt" there is related to what you said about luks. I did not get that the first time I read it. | 11:16 |
| josch | bremner: it has been a big annoyance in the past where we had the console= argument in the kernel cmdline twice and it turns out if the wrong argument is last, then the luks prompt only shows up on serial /o\ | 11:17 |
| bremner | ack. I should be more careful about the way I am doing that currently, since I think it does end up serial last | 11:18 |
| ch | the console handling in early userspace is super annoying :( | 11:18 |
| josch | this hack is only in boot.scr to make sure that tty1 definitely comes last: setenv bootargs "${bootargs} console=tty1" | 11:20 |
| josch | we implemented it like this: https://source.mnt.re/reform/reform-tools/-/commit/d9ee804f63720c550238c6790f80f0a0019c5cca | 11:21 |
| bremner | right. but then I cargo-culted that to add second setenv after it | 11:21 |
| bremner | I should merge all the console stuff to one line I guess | 11:21 |
| josch | bremner: i will happily add more code comments to this stuff to explain things in more detail to the next person in case you have any suggestions | 11:22 |
| bremner | josch: I sortof knew it was wrong when I did it... | 11:22 |
| josch | XD | 11:23 |
| ch | (: | 11:23 |
| - aloo_shu (QUIT: Read error: Connection reset by peer) (~aloo_shu@85.51.16.194) | 11:57 | |
| + aloo_shu (~aloo_shu@85.51.16.194) | 12:07 | |
| ch | josch: (sudden insight) what strikes me as odd is that uboot itself appears to have an uefi implementation now. so theoretically no edk2 needed? | 12:28 |
| ch | josch: or is uboot embedding edk2? | 12:28 |
| Zaba | iirc uboot implements the bare minimum of UEFI interfaces needed to boot Linux | 12:36 |
| Zaba | so depending on your use case it may be sufficient but it’s not the same as edk2 | 12:37 |
| Zaba | and at that point I think just using bootstd and extlinux.conf is a better option anyway | 12:37 |
| ch | i was looking into it because i tried booting an .iso directly. and there uboot does find the embedded EFI disk image, but ignores everything else (incl. the extlinux.conf inside the iso) | 12:38 |
| ch | was kinda nice to see grub just start and find all its things | 12:38 |
| ch | good we're talking about this, maybe the .iso was not isohybrid enough | 12:41 |
| minute | josch: regarding the mmc issue, good find. the uboot mmc driver has some notorious issues iirc, i had similar trouble on a311d on some boards but it would work fine in linux. two ideas come to mind: 1. reduce mmc clock in uboot (dts) 2. patch uboot so it auto resets and not just sits there asking for reset | 12:43 |
| + mjw (~mjw@gnu.wildebeest.org) | 12:44 | |
| josch | minute: this could be different from board to board? because the issue only happens maybe every 1/2 to 1/5 of all boots (didn't properly count yet to create an average) but the user in the forum says that they have to sometimes retry six times before it works which is not at all what i observe... | 12:53 |
| josch | and in that case patching u-boot to auto-reset sounds like the better option instead of reducing the clock so that it works for the worst boards? | 12:53 |
| josch | ch: yes, theoretically no edk2 needed if you just want uefi -- maybe it works out-of-the-box maybe not, somebody has to try | 12:54 |
| josch | ch: but nobody cared enough yet to put in the work :) | 12:54 |
| ch | right, but without edk2 no display again | 12:56 |
| + andreas-e (~Andreas@2a02-8434-b6a3-e901-daca-beef-9a0a-5e34.rev.sfr.net) | 13:08 | |
| + casparvitch (~casparvit@36-255-114-132.ip4.superloop.au) | 13:13 | |
| minute | josch: yes, board related, for me it doesn't happen at all. you have latest uboot (a few months old?) where i put in the usb hub reset? | 13:19 |
| minute | boot is 100% reliable for me on my rk3588 machines (pocket and classic) including 0% usb issues | 13:20 |
| minute | i also have no hdmi issues except the hdmi audio spam in dmesg | 13:21 |
| minute | the only technical issue i have is on pocket with v1 display that rarely the internal display activates with colorful snow at boot, suggesting init timing issue | 13:21 |
| - casparvitch (QUIT: Remote host closed the connection) (~casparvit@36-255-114-132.ip4.superloop.au) | 13:35 | |
| + casparvitch (~casparvit@36-255-114-132.ip4.superloop.au) | 13:41 | |
| josch | the log says: U-Boot SPL 2024.10-g424c714eb247-dirty (May 06 2025 - 15:46:29 +0000) | 14:03 |
| josch | the git commit hash is unfortunately coming from u-boot upstream and not from the reform-rk3588-uboot repository | 14:03 |
| josch | maybe reform-rk3588-uboot should set a reform-specific version string like the other uboot builds do? | 14:04 |
| josch | but the date hints that yesit is likely the latest release 2025-05-06 | 14:04 |
| grimmware | I don't think I'm convinced by these batteries - it does seem like an awfully tight fit. | 14:06 |
| - casparvitch (QUIT: Ping timeout: 252 seconds) (~casparvit@36-255-114-132.ip4.superloop.au) | 14:13 | |
| minute | josch: ok | 14:43 |
| minute | interesting https://patchwork.ozlabs.org/project/uboot/patch/20250630102524.127970-1-pbrobinson@gmail.com/ | 14:48 |
| josch | minute: were you involved with that? That looks like the defconfig from reform-rk3588-uboot with ~30 lines extra at the end | 15:09 |
| minute | josch: i wasn't | 15:09 |
| josch | hm.. | 15:10 |
| minute | because it mentions fedora maybe bkeys knows more? | 15:10 |
| josch | unfortunately bluerise hasn't been here for some time | 15:10 |
| minute | iirc reform rk3588 upstreaming for uboot was mostly done by bluerise, but indirectly through linux (?) | 15:11 |
| minute | ah yeah | 15:11 |
| josch | yes, that's also what i remember -- uboot rk3588 dtb comes from linux | 15:11 |
| + reform4782 (~max@2001:9e8:6125:2200:3b:5e1b:143b:db25) | 15:17 | |
| - reform4782 (QUIT: Client Quit) (~max@2001:9e8:6125:2200:3b:5e1b:143b:db25) | 15:19 | |
| - bkeys (QUIT: Quit: With every step we take, danger will follow closely) (~Thunderbi@h193.131.19.98.dynamic.ip.windstream.net) | 15:23 | |
| + bkeys (~Thunderbi@98.19.131.193) | 15:24 | |
| - bkeys (QUIT: Ping timeout: 244 seconds) (~Thunderbi@98.19.131.193) | 15:28 | |
| bremner | josch: do you have a recommended procedure for adding the serial console so that console=tty1 still ends up last? I dropped a file in /etc/flash-kernel/ubootenv.d/, is it enough that it starts with 00 to get inserted in the right order? I called mine 01something and that ended up after console=tty1 | 15:44 |
| bremner | worth a shot I guess | 15:58 |
| josch | bremner: make it order *before* 00reform2_ubootenv | 16:22 |
| josch | bremner: and of course you can also just add console=tty1 to the end of whatever you are adding | 16:22 |
| bremner | right. luckly 00reform2_serial orders before 00reform2_uboot, so I kinda got it right by mistake | 16:23 |
| + paperManu_ (~paperManu@72.10.128.164) | 16:24 | |
| josch | nice :) | 16:24 |
| bremner | josch: would it make sense for things that "must be last" to start with "99", or are there some other constraints? | 16:51 |
| josch | bremner: no, we could split up that config file into two, name one of them 00* and the other 99* | 16:52 |
| josch | i added that tty1 bit to the 00* to have minimal disruption | 16:52 |
| Zaba | I think having the system-supplied ones be something like 10- and 90- instead of 00- and 99- would make it easier for users to insert stuff before or after them | 19:49 |
| josch | Zaba: what is currently 00* is what's supposed to be the default. You can override it by sorting it after. bremner can still add a 01* file which does: setenv bootargs "${bootargs} console=ttyS1 console=tty1" | 19:53 |
| josch | that way, you get tty1 last and also insert a ttyS1 before it | 19:53 |
| josch | which file to edit to do something is already very complicated -- unless there is some strong requirement i'd like to keep complexity low | 19:54 |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50) | 19:58 | |
| - amk (QUIT: Remote host closed the connection) (~amk@user/amk) | 20:07 | |
| + amk (~amk@user/amk) | 20:09 | |
| + reform27998 (~matzipan@ip1f1336b8.dynamic.kabel-deutschland.de) | 20:11 | |
| - reform27998 (QUIT: Ping timeout: 252 seconds) (~matzipan@ip1f1336b8.dynamic.kabel-deutschland.de) | 20:15 | |
| + bkeys (~Thunderbi@h193.131.19.98.dynamic.ip.windstream.net) | 20:26 | |
| - xktr (QUIT: Ping timeout: 252 seconds) (~xktr@user/xktr) | 20:38 | |
| bkeys | minute, josch: Yeah when I was setting it up (I still gotta get around to testing A311D) he had a fork of uboot he had me testing, but in the end I was using the Reform version of it (although there wasn't anything wrong his fork either) so I think he was working on that | 20:39 |
| + xktr (~xktr@user/xktr) | 20:39 | |
| - RandyK (QUIT: Remote host closed the connection) (~RandyK@user/randyk) | 20:46 | |
| + RandyK (~RandyK@user/randyk) | 20:47 | |
| josch | minute: could you enable gitlab CI for https://source.mnt.re/vimja/reform-debian-packages please? Thank you! | 20:58 |
| - mjw (QUIT: Quit: Leaving) (~mjw@gnu.wildebeest.org) | 21:52 | |
| - gustav28 (QUIT: Quit: Quit) (~gustav@c-78-82-54-137.bbcust.telenor.se) | 22:15 | |
| - sigrid (QUIT: Ping timeout: 276 seconds) (~sigrid@ftrv.se) | 23:20 | |
| + sigrid (~sigrid@ftrv.se) | 23:29 | |
| - svp (QUIT: Ping timeout: 276 seconds) (~svp@host-79-7-240-189.business.telecomitalia.it) | 23:49 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!