11:05 <warpme_> guys: i need some brainstorming about amlogic boot: i'm trying to get 5.13 mainline booting on radxa-zero (s905y2 sbc). SD card booloader is current amlogic/custodians + radxa FIP files. 5.13 kernel boots - but oops at /sbin/init ( https://pastebin.com/HLgeCP9t ). This SD card inserted to SM1 box (x96air with vendor uboot) boots ok. As only difference between non-working board & working tvbox is bootloader - i think issue

11:05 <warpme_> is in radxa FIP files. Has anybody any idea how to move forward this?

11:24 <benjamin545> could you maybe take the uboot image before it gets wrapped with the radaxa FIP and wrap it yourself

11:24 <benjamin545> https://github.com/u-boot/u-boot/tree/v2020.04-rc1/board/amlogic/u200

11:25 <benjamin545> thats for a 905d2, it should be the same chip family so i would think it should work, just skip past the part of building the mainline uboot from source

11:39 <warpme_> benjamin545: i'm a bit lost about what you proposing. Currently i'm: 1\building custodians uboot (bl33); 2\do FIP packing FIP files with bl33; 3\flash result to sd card. So do you propose to: a\switch bl33 build to https://github.com/u-boot/u-boot/tree/v2020.04-rc1/board/amlogic/u200 or rather b\change whole procedure?

11:47 <benjamin545> i am just proposing to take the uboot you have built, and use that as the bl33 in the link I have given. i built mainline uboot for a 905D3 board using that procedure and it worked, you may have to modify it a bit.

11:48 <benjamin545> the line in there "> cp u-boot.bin fip/bl33.bin" that is where you would instead inject your u-boot.bin you have already built

11:59 <warpme_> ah i see. btw: I already done test with using as much as possible U200 FIP files in bootloader. DRAM used by radxa requires this exact radxa FIP files to get bootloader working at all: acs.bin aml_ddr.fw ddr3_1d.fw ddr4_2d.fw lpddr4_1d.fw piei.fw bl2.bin ddr4_1d.fw diag_lpddr4.fw lpddr4_2d.fw. Rest files: bl2.bin bl30.bin bl301.bin bl31.img were from U200 FIP. And.....issue is the same. Let me try with U200 bl33 - but do

11:59 <warpme_> you believe it will change anything?

12:03 <benjamin545> well, if it did it would mean there is an issue with your uboot right? not the rest of the fip. i would think if you got past loading the kernel uboot wouldn't be the issue, but it can happen

12:43 <warpme_> ok - i tested uboot v2020.04-rc1 and this ver. loops with bootp. Also I tested mainline uboot 2021.07 and oops issue is the same. (btw: oops issue is the same also with uboot binary radxa provide me). Thing constantly intriguing me is that oops is for first user-space executed code (/sbin/init).

13:14 <benjamin545> ya that i have always found that to mean that it didnt find the init, this could be that it isnt there, this can happen sometimes, usually just a symlink to systemd will fix it, or it means the filesystem didnt mount, and thats usually because it didnt actually load a drive

13:15 <benjamin545> your log output file makes me think it loaded the sdcard (mmc0 should be sd card) and it looks like it mounted an ext4 filesystem

13:16 <benjamin545> oh, are you using a initrd/initramfs?

13:19 <warpme_> yes. i'm using initramfs with overlayfs and pivot_root

13:24 <benjamin545> ok, so my experience has always been that with a initramfs, you actually run systemd(or your init application) from the initramfs, it then takes care of swapping root and loading from there

13:25 <benjamin545> can you take a look at the initramfs to see if it has something for /sbin/init

13:59 <warpme_> my init scrip is shell script so i added some verbosity. nothing on screen. it look like oops is when kernel trying to execute in userspace. as this sd card works just inserted to sm1 box - i think issue isn't in my code (init script, etc) - but rather in something like maybe el1->el0 switch and immediate oops in el0. what i recall from school privilege switches can be realised only by system monitor (el2 or el3 so in

13:59 <warpme_> atf). atf code is fip and fip is code which differs between working & non-working case. for me more and more fingers are pointing to fip.....

14:18 <benjamin545> ok, so nothing they have provided (radxa uboot and linux) has been able to get you to userspace either?

14:41 <warpme_> Tez

14:41 <warpme_> Yes

14:45 <warpme_> This becomes really interestig as my code runs well on 9 different SoC including 3 amlogic and only radxa has this interesting issue of not executing anything im userspace