#archlinux-ports | Logs for 2026-03-30
Back
[03:13:05] -!- hcmb has quit [Ping timeout: 245 seconds]
[03:13:31] -!- hcmb has joined #archlinux-ports
[06:47:46] <solskogen|M> It's coming - but python-pytorch is causing some trouble...
[06:48:06] <solskogen|M> and its yet again, one of those packages that takes a lot of time to compile...
[06:54:10] <kitlith> i think there's some people who are on matrix whose messages are not getting bridged to IRC
[08:04:30] -!- Antiz has quit [Quit: The Lounge - https://thelounge.chat]
[08:06:54] -!- Antiz has joined #archlinux-ports
[08:15:32] -!- yjun123 has quit [Quit: Konversation terminated!]
[08:15:48] -!- yjun123 has joined #archlinux-ports
[09:01:38] -!- h|weechat has quit [Ping timeout: 252 seconds]
[09:05:22] -!- h|weechat has joined #archlinux-ports
[09:57:12] <yuvadm> or solskogen|M is just talking to himself
[09:57:53] <solskogen|M> Heh, I'm not :-D
[10:07:35] <solskogen|M> I'm having a bit a problem with a Odroid M2 and kernel config. ALARM's kernel works fine, but ours does not. I suspect is has something to do with us using more modules than ALARM - I can't even boot with our kernel (no display) - does anyone have any clue on where to start looking in the kernel configuration?
[10:10:14] <yuvadm> at some point I recall odroid C4 having kernel config issues with a regression introduced somewhere around 5.15-ish, no idea if this is related but best i can offer
[10:10:56] <yuvadm> where is your boot stuck? do you get anything?
[10:14:15] <solskogen|M> Nothing after grub saying loading initramfs
[10:14:18] <linkmauve> solskogen|M, your current kernel doesn’t enable anything about Rockchip SoCs.
[10:15:04] <linkmauve> We discussed that a few days ago ith bschnei, I plan on merging my personal config for the rk3588 as found in the Radxa Rock 5B, and I also have a rk3568 in an ODROID-M1 which is less of a priority to me.
[10:15:21] <linkmauve> But I haven’t done so yet.
[10:15:47] <solskogen|M> You have to keep in mind that I know next to nothing on how to configure the kernel :-)
[10:17:20] <linkmauve> solskogen|M, so basically there are two things, the device tree which describes the hardware, and the drivers, which get probed based on which hardware the device tree says is present.
[10:17:27] <gromit> solskogen|M: I used 3 quite powerful machines yesterday to compile the rocm stack, I guess it'll take a while until it is through your build pipe
[10:17:46] <gromit> Also see https://lunnova.dev for fun :D
[10:17:47] <phrik> Title: ROCm 7.1.1: you can (not) build | lunnova.dev (at lunnova.dev)
[10:18:06] <linkmauve> The .config file lists which drivers get compiled, so if you’re missing some driver for your hardware that particular piece of hardware won’t work.
[10:19:06] <linkmauve> I can share you the config I use with my Rock 5B, and you can adapt it to your kernel, or you can wait a few days (I hope) until I get the motivation to look into that.
[10:19:17] <solskogen|M> gromit: yes, that pretty much sums it up.
[10:20:12] <linkmauve> solskogen|M, https://linkmauve.fr
[10:20:56] <linkmauve> Note that it is very tailored to my use of the board, it won’t support any other hardware pretty much.
[10:21:36] <solskogen|M> I need the kernel to work on multiple machines, not only the M2 :-) I put in a hail mary and created a MR to ben where I enabled everything and the kitchen sink (most as modules) - it worked fine on AWS and on my Orion O6 - but to no avail on my M2.
[10:22:13] <linkmauve> solskogen|M, sure, but your current kernel won’t work on any Rockchip board.
[10:23:01] <solskogen|M> I think it should - because this is the config I'm using https://gitlab.archlinux.org
[10:23:03] <phrik> Title: config.aarch64 · aarch64 · Christer Solskogen / linux · GitLab (at gitlab.archlinux.org)
[10:23:13] <linkmauve> I recommend you to get a serial cable btw, to avoid relying on screen support which is a ton of different hardware made to work together.
[10:23:24] <solskogen|M> (but unfortunately we don't live in Should Land(tm))
[10:24:02] <linkmauve> solskogen|M, see for instance you didn’t enable CONFIG_DRM_ROCKCHIP, so obviously no screen will ever work.
[10:24:49] <linkmauve> I don’t know if that’s the only missing option to get something on screen, but it’s a mandatory one.
[10:25:12] <linkmauve> You didn’t enable the GPU either, so everything will be software rendered once it is, that’s CONFIG_DRM_PANTHOR.
[10:25:30] <solskogen|M> sorry, that was the wrong one. https://gitlab.archlinux.org
[10:25:31] <phrik> Title: config.aarch64 · update-config-and-pkgbuild · Christer Solskogen / linux · GitLab (at gitlab.archlinux.org)
[10:25:35] <linkmauve> For older boards like the ODROID-M1, you will also need CONFIG_DRM_PANFROST for the older v7 GPU.
[10:25:38] <linkmauve> M2 has a v10 GPU.
[10:26:28] <solskogen|M> Both panfrost and panthor is enabled.
[10:26:59] <linkmauve> Those two are just for 3D acceleration.
[10:27:37] <linkmauve> So if things are enabled, your initramfs contain the modules, and when you start a compositor nothing shows up on the screen, then you will likely want to get a look at the kernel logs.
[10:28:01] <solskogen|M> but it would be awesome if someone with more experience than me would take a look - because I'm clearly not experienced enough with kernel configs.
[10:28:04] <linkmauve> The easiest way is to plug a serial cable to UART0 and read that from another computer.
[10:28:29] <linkmauve> solskogen|M, wanna give me an ODROID-M2 board? :)
[10:29:39] <solskogen|M> If I had the funds, yes :-)
[10:29:58] <solskogen|M> on paper it's quite a decent board.
[10:37:13] <solskogen|M> Now that it can use UEFI as well, I thought I would give it a try. The board has just been stowed away for one and a half year.
[10:47:32] <yuvadm> solskogen|M what boot script are you using?
[10:47:55] <solskogen|M> boot script?
[10:48:14] <yuvadm> cat /boot/boot.*?
[10:48:31] <yuvadm> something that tells u-boot what to do
[10:49:01] <solskogen|M> I'm not using uboot.
[10:49:19] <solskogen|M> https://github.com
[10:49:20] <phrik> Title: GitHub - tobetter/edk2-rk3588: EDK2 UEFI firmware for Rockchip RK3588 platforms · GitHub (at github.com)
[10:49:56] <yuvadm> oh interesting, by choice? never used something like that
[10:53:30] <solskogen|M> Well, yes. I like that more than vendor uboot.
[10:56:24] <yuvadm> M2 support hasn't made it upstream?
[10:57:30] <solskogen|M> Not for uboot. Not properly at least.
[11:02:39] <yuvadm> hmm https://github.com
[11:02:40] <phrik> Title: u-boot/configs/odroid-m2-rk3588s_defconfig at master · u-boot/u-boot · GitHub (at github.com)
[11:09:53] <linkmauve> solskogen|M, never use vendor u-boot, always use mainline u-boot.
[11:10:02] <linkmauve> And mainline u-boot can do UEFI just fine, for years.
[11:10:27] <linkmauve> I never used it though, I prefer the extlinux.conf way of booting.
[11:10:53] <solskogen|M> sure, but for O2 there's no hdmi output support. That said, that has nothing to do with the problem I'm facing :-)
[11:11:26] <linkmauve> It might, I don’t know anything about edk2 and it could be the cause of some driver failures.
[11:11:47] <solskogen|M> I have the same problem with uboot as well.
[11:12:02] <linkmauve> For instance the latest release of TF-a was causing a boot loop on my Rock 5B, using latest master fixed it.
[11:12:33] <linkmauve> solskogen|M, get a serial cable, you’ll need it for all boot debugging.
[11:13:06] <solskogen|M> I have one of them around
[11:13:07] <linkmauve> https://www.olimex.com you could find cheaper ones probably, that’s the one I use.
[11:13:08] <phrik> Title: USB-SERIAL-F (at www.olimex.com)
[11:13:28] <linkmauve> Perfect. :)
[12:47:24] -!- marmis has quit [Quit: Bye!]
[13:33:36] -!- marmis has joined #archlinux-ports
[16:18:37] -!- marmis has quit [Ping timeout: 276 seconds]
[16:30:26] -!- marmis has joined #archlinux-ports
[17:00:42] <f_> FYI U-Boot's UEFI support works very nicely, a bunch of devices in e.g. pmOS use that exclusively
[18:06:18] -!- marmis has quit [Ping timeout: 256 seconds]
[18:26:16] <bschnei> +1 for UEFI on U-Boot. Also enabling OF_CONTROL in u-boot means that the device tree gets built into u-boot and you can avoid having all that mess on /boot
[18:33:48] <bschnei> solskogen: you are on the right path. see yuvadm's MR here: https://gitlab.archlinux.org Certain options have to be 'y' even though the kernel allows them to be 'm'. Usually these are the essentials and then things like networking, etc can be left as modules. Unfortunately it's often a real pain to figure out exactly what options you need. The devicetree used by your device does tell th
[18:33:48] <bschnei> e kernel what drivers it should try to load. So I've had some luck searching for the strings contained in the compatible= and then figuring out what options enable those drivers. https://github.com
[18:33:49] <phrik> Title: Sign in · GitLab (at gitlab.archlinux.org)
[18:33:50] <phrik> Title: linux/arch/arm64/boot/dts/rockchip/rk3588s-odroid-m2.dts at master · torvalds/linux · GitHub (at github.com)
[18:35:45] <bschnei> also check this tree which gets included into the one above: https://github.com
[18:35:46] <phrik> Title: linux/arch/arm64/boot/dts/rockchip/rk3588-extra.dtsi at master · torvalds/linux · GitHub (at github.com)
[19:26:44] -!- h|weecha1 has joined #archlinux-ports
[19:31:17] -!- h|weechat has quit [*.net *.split]
[20:32:32] <solskogen|M> DT is also built in on EDK2 btw :)
[22:10:12] -!- titus_livius has joined #archlinux-ports
[22:22:23] -!- marmis has joined #archlinux-ports
[23:27:48] <bschnei> sweet. I would hope so. the downside of it is needing to rebake your firmware if there are DT updates. there seems to be sparse support for userspace based bootloader maintenance which leaves users with serial console only for updates