00:51 <brads> narmstrong: Do you have an idea when g12a support might hit mainline uboot? I'm interested in trying it for g12b to remove vendor uboot for Odroid N2

06:10 <narmstrong> brads: it will follow shortly Linux support

08:42 <brads> narmstrong: nice, ty. Another question which is a little offtopic, do you think it's possible to eventually be able to patch kernel with Realtime (-rt) branch for meson arm64 big/little socs? I wonder how much work scheduling between 2 domains is on the soc.

08:46 <narmstrong> brads: no idea sorry, but I don’t see why it wouldn’t be possible

10:59 <hexdump0815> just a stupid question regarding reserved memory: i'm using a s905x tv box with 1gb ram as headless compute node - linux sees only part of that 1gb (Memory: 825636K/896000K available) - is there a way to free some of the missing about 150mb which most probably are reserved for framebuffer, codecs, gpu or whatever?

11:01 <hexdump0815> looking at the dts (meson-gxl-s905w-p281.dtb) i do not see anything which would account for 150mb - only cma which is seen by the linux kernel which i reduced for my use-case to 16mb and gained already about 250mb more useable memory

11:02 <hexdump0815> are those "missing" about 150mb reserved at linux kernel level somehow or below (uboot, lower level firmware)? and is there a way to release them or part of them? a lot of thanks in advance

11:03 <hexdump0815> sorry - wrong dts above - should be meson-gxl-s905x-p212.dtb

11:07 <narmstrong> Well you have the ~21mib reserved for the fip and rom, 16mib for the class and the final 113 must be used by the kernel, try removing some features in the kernel, like other SoCs support, pcie and remove modules

11:07 <narmstrong> Then you will reduce this number

11:08 <narmstrong> But still you’ll have the userspace binaries and the shared libraries loaded in memory and buffers preallocated for I/o access

11:24 <hexdump0815> thanks for the explaination - i was thinking that the kernel code and data is counted into those 896000K already -> Memory: 825636K/896000K available (11068K kernel code, 1612K rwdata, 5304K rodata, 1408K init, 385K bss, 53980K reserved, 16384K cma-reserved) and makes the difference between the 825636K and the 896000K?

12:31 <narmstrong> You can find memory reservation info in debugfs, the biggest figure is the 53mib reserved memory, no real idea where it comes from

22:13 <memeka> ndufresne: can I run a gadget driver on an odroid c2 on mainline?

22:14 <ndufresne> memeka, if OTG works, then yes, I don't have a C2

22:14 <ndufresne> and potato does not have OTG

22:15 <memeka> ndufresne: i saw on the Linux-meson page that otg doesn’t work

22:15 <ndufresne> my ideal SBC would be a relatively cheap one, that boots over 1A USB, and does OTG on the same connector

22:15 <memeka> Don I need to add some entries to dtb for otg?

22:16 <ndufresne> that I'm not the right person to tell

22:16 <memeka> Who is? :)

22:17 <ndufresne> I believe maybe narmstrong, chewitt, and other "real" dev behind this work will know

22:17 <memeka> narmstrong - any ideas? :D

22:18 <ndufresne> so far, my work here was mostly flipping bits on existing registers, pluging things together and making sure gst works (or let's say is usable)

22:18 <memeka> ndufresne: how’s your board doing? Booting?

22:18 <memeka> ndufresne: gst was not usable on my c2 last time I tried :( (4.20 kernel)

22:19 <memeka> I’ll try again on 5.0 a bit later

22:21 <ndufresne> memeka, I believe I sorted out the u-boot issue with ethernet, I simply downgraded to use q200 board setup in the end, since modifying the q201 was giving identical build in the end

22:21 <xdarklight> memeka: OTG is currently not working on any Amlogic SoC - it's "work-in-progress" for the G12A SoCs (but the USB implementation there unfortunately doesn't have much in common with GXBB which is used on Odroid-C2)

22:22 <ndufresne> memeka, but I never managed to get the kernel fully working, only tried 5.X, chewitt gave me couple of unmerged patch to include, but didn't find time yet to try

22:22 <xdarklight> memeka: if you're willing to invest some time you can get peripheral mode working (so it'll be always either in host mode or peripheral mode, no auto-detection though and manual switching requires modifying the .dtb and a reboot)

22:22 <ndufresne> memeka, basically, as soon as something touches the display controller it hangs

22:22 <memeka> Thx xdarklight

22:23 <ndufresne> and sometimes it hangs during memory init

22:24 <memeka> ndufresne: looks like multiple issues :(