<Woutervanh>
SPL: failed to boot from all boot devices
<Woutervanh>
### ERROR ### Please RESET the board ###
<Woutervanh>
Was working with u-boot 2022.01 though...
<qschulz>
Woutervanh: any chance you can try upstream U-Boot? at least 2024.07, at best the latest rc tag for 2024.10?
<qschulz>
there's been a lot of changes to Rockchip support in the last year or so, and i am not entirely sure rk3288 is among the SoCs that get the most attention today
<Woutervanh>
I'll try. Is there a way to boot a board through maskrom/usbotg directly? Without going through flash at first?
<CounterPillow>
Yeah, with rkdeveloptool, but it's clunky
stikonas has quit [Quit: Konversation terminated!]
<qschulz>
Woutervanh: yes, I believe collabora is doing this for CI, you need to use bootmerger from Rockchip to generate the binary to then use with rkdeveloptool (or rockusb from collabora's rockchiprs repo)
<qschulz>
(not sure about bootmerger being the name for it)
<qschulz>
but no upstream build of a USB bootable image in U-Boot yet
<Woutervanh>
I'm also a bit in doubt if I have set up every option correct in uboot... My intention is to create a wic image via yocto (and meta-rockchip from the yoctoproject). So I guess uboot is used as spl and tpl then? everything should be on the onboard emmc. Is there a reference board in uboot I can rely to?
naoki has quit [Quit: naoki]
<qschulz>
Woutervanh: seems like SPL and TPL are both open-source and from U-Boot yes
<qschulz>
Woutervanh: you need SPL_PAD_TO to align with the address of loader2 in the wic file
<qschulz>
Woutervanh: the meta-rockchip maintainer isn't able to boot his rk3288 board anymore (for a while), so it isn't boot tested, only build tested, so it's possible something's broken there
<qschulz>
I myself do not have access to any rk3288 so wouldn't be able to debug this myself
<Woutervanh>
qschulz: well, I'm trying to upstep versions currently from what was working. v2023.04 was still working fine. v2023.10 seems to have an issue
<qschulz>
Woutervanh: that's a great start :) if you ever find the exact commit, then we can start looking a bit further into it :)
<Woutervanh>
2023.07, also working. I'll git bisect between 2023.07 and 2023.10, I hope it's right that it'll only take 10 steps :-p
Hypfer67 has joined #linux-rockchip
Hypfer6 has quit [Ping timeout: 272 seconds]
<qschulz>
if you're unlucky, it's in a merge commit :)
<qschulz>
i had this happen to me in the past, the next to master merge commit right after a release :(
Hypfer67 has quit [Ping timeout: 272 seconds]
Hypfer6 has joined #linux-rockchip
indy- is now known as indy
Hypfer6 has quit [Ping timeout: 272 seconds]
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
warpme has joined #linux-rockchip
dsimic has quit [Ping timeout: 260 seconds]
Woutervanh has quit [Quit: Client closed]
dsimic has joined #linux-rockchip
Woutervanh has joined #linux-rockchip
franoosh has quit [Remote host closed the connection]
<robmur01>
qschulz: and then you bisect down the other side of that merge and it still points to the top because the breakage is only in the interaction of changes on both sides... been there :(
Stat_headcrabbed has joined #linux-rockchip
<qschulz>
robmur01: git log -c for this :) but yes, pain, lots of pain :(
<qschulz>
Stat_headcrabbed: any ETA for when this will be first seen in the wild?
<qschulz>
the fact that they don't even give the number of cores or which cores, nor the name of the GPU kinda hints at "we had an idea, maybe something like this will be nice" ?
<Stat_headcrabbed>
This news said estimated at next year
<robmur01>
Ooh, fun - if we're lucky and it's essentially 3588 with a CPU/GPU refresh, it might not be *too* bad
<qschulz>
robmur01: do you REALLY think it's gonna be that :D
<Stat_headcrabbed>
I'm not really interested in it since rockchip has such a slow mainlining process
<qschulz>
I hope the GPU could be supported by Panthor, otherwise we're in for a long wait again?
<Stat_headcrabbed>
And it's sdk is also awful
<Stat_headcrabbed>
Possibly another long wait
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<qschulz>
though we cannot really blame Rockchip for the ARM GPU debacle ;)
<CounterPillow>
Yeah I'd rather buy a fresh Cortex-A7 from STM!
<qschulz>
(and we have to note the Panthor process was light years from what it took in efforts and time from reverse engineered Mali GPUs :) )
<Stat_headcrabbed>
CounterPillow: STM has A35 now XD
<qschulz>
Stat_headcrabbed: must hide this news from management though, too much work for me already :)
<robmur01>
given that 3588 took so long it already had a GPU refresh *during* development, I'd be surprised if they start from scratch...
<Stat_headcrabbed>
Well I'm more interested in SG2380 now
<qschulz>
mmind00: you saw nothing, and if you saw something, you say nothing :)
<qschulz>
robmur01: source? wasn't aware of that
<qschulz>
robmur01: i've been told the rk3576 is also not THAT far from RK3588 in terms of IPs, so there's some kind of hope it's not a complete product lineup like the move to rk35 from rk33
<Stat_headcrabbed>
what source
<qschulz>
GPU refresu
<robmur01>
Hmm, to me 3576 smelled more like an update of the 3399 design
<qschulz>
haven;t played with it yet, so cannot comment
<phh>
I thought rk3576 was literally a 3588 with broken/disabled components
<qschulz>
nah, that would be rk3583/rk3582 I think?
<Stat_headcrabbed>
that's 3582/83
<phh>
ah ok thanks
<Stat_headcrabbed>
and3588s
<qschulz>
nah, rk3588s is a smaller footprint
<qschulz>
but rk3588s is an rk3588 minus some IPs, smaller footprint though :)
<Stat_headcrabbed>
3588s 3588 3582 3583 all using same silicon