mmind00 changed the topic of #linux-rockchip to: Rockchip development discussion | public log at https://libera.irclog.whitequark.org/linux-rockchip
System_Error has joined #linux-rockchip
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
kevery has quit [Remote host closed the connection]
kevery has joined #linux-rockchip
vagrantc has quit [Quit: leaving]
naoki has joined #linux-rockchip
stikonas has quit [Quit: Konversation terminated!]
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
hexdump0815 has quit [Ping timeout: 260 seconds]
hexdump0815 has joined #linux-rockchip
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
System_Error has quit [Remote host closed the connection]
naoki has quit [Quit: naoki]
naoki has joined #linux-rockchip
naoki has quit [Client Quit]
naoki has joined #linux-rockchip
naoki has quit [Quit: naoki]
naoki has joined #linux-rockchip
franoosh has joined #linux-rockchip
ungeskriptet has quit [Ping timeout: 244 seconds]
ungeskriptet has joined #linux-rockchip
ungeskriptet_ has joined #linux-rockchip
ungeskriptet has quit [Ping timeout: 252 seconds]
ungeskriptet_ is now known as ungeskriptet
Woutervanh has joined #linux-rockchip
<Woutervanh> robmur01: I must be doing something wrong then. I enabled both vopb and vopl, but then none of the cards work. I also see logs about cycle dependencies on lvds/edp. I guess because both vopb and vopl support lvds/edp. I'm in doubt how to do routing correct. Is there an example dts somewhere? You mention chromebooks?
raster has joined #linux-rockchip
<mmind00> Woutervanh: there should be no "routing" ... both VOPs can drive the eDP and both VOPs can drive the LVDS ... and the source-vop is selected in the output drivers (edp + lvds) to set the (hopefully) correct bit in GRF_SOC_CON6 ... for where they should get their output
<Woutervanh> based on what will it assign a vop then?
<mmind00> first-come-first-serve normally :-)
<mmind00> and for most displays that also is fine
<Woutervanh> Well, now I added both lvds and edp in my devicetree, and have vopl, vopl_mmu, vopb and vopb_mmu all okayed, but none of them works. Not lvds, not edp
<mmind00> (high-res hdmi or so can be a problem, when connected to the "little" vop, I think)
<Woutervanh> there's no hdmi, only lvds (dual channel, 1920x1080) and edp
<Woutervanh> edp also 1920x1080 if that matters
<mmind00> Woutervanh: you might want to push your DT to some github repo so people can look at actual code ;-) ... also "edp" and "edp_phy" are both on, right?
<Woutervanh> yes, they are. Indeed, was just going to put it online
<mmind00> and maybe also post the dmesg output to some paste-thing ;-)
<Woutervanh> https://gist.github.com/woutervanh/fef1d9862418bfc9c96853e292aed7f9 => here you go, much appreciated
<mmind00> Woutervanh: the complete absence of "drm"-related lines in there, makes it look like dependency problem
<mmind00> with the drm-driver never actually starting to probe at all
<Woutervanh> in the dmesg you mean? Yeah, but if I remove the edp cruft in the dts, then it does work. I'll also upload a dts which actually works with lvds, mom
<mmind00> Woutervanh: ah, so lvds-only works
<Woutervanh> correct, I updated the gist with the working dts
<mmind00> Woutervanh: problem is likely the "simple-panel" compatible ... https://gist.github.com/woutervanh/fef1d9862418bfc9c96853e292aed7f9#file-rk3288-tlv-rk808-dts-L98
<mmind00> sorry ... wrong link
<dvergatal> mmind00: around?
<qschulz> dvergatal: he literally wrote 15seconds before :D
<Woutervanh> mmind00: hm, should the panel type matter? I was under the impression it was something else as it was stating nvidia...
<mmind00> Woutervanh: yes it matters a lot ... the edp-driver _needs_ its panel, your panel-compatible does not even exist, so no panel driver ever probes
<mmind00> Woutervanh: so the edp's panel requirement never gets satisfied, so the rockchip-drm waits indefinitly for the edp to appear
<dvergatal> qschulz: ooo you are as well in here as in yocto:P
<dvergatal> qschulz: hi
* qschulz waves
<dvergatal> qschulz: i know he has but I didn't know if he has time to chat that is why i asked
<dvergatal> mmind00: do you have some contacts to sales department in rockchip? caz i can't reach kever...
<mmind00> dvergatal: not at all ... I only know technical people
<dvergatal> mmind00: can you reach them? or ask for contact to the sales department?
<Woutervanh> mmind00: I changed it to compatible = "edp-panel", but still the same
<qschulz> Woutervanh: I have virtually no experience there, but maybe enable one at a time? LVDS first, then eDP once the former is working?
<Woutervanh> posted my lsmod too, just for clarification: https://gist.github.com/woutervanh/fef1d9862418bfc9c96853e292aed7f9#file-lsmod
<mmind00> Woutervanh: also, please read the binding (or driver) ... it does for example require a "power-supply" regulator
<Woutervanh> good idea to verify edp seperately (y)
naoki1 has joined #linux-rockchip
naoki has quit [Quit: naoki]
naoki1 is now known as naoki
System_Error has joined #linux-rockchip
<rtp> Kwiboo: hi. I'm trying to apply https://patchwork.ozlabs.org/project/uboot/list/?series=445397 on top of u-boot master, but looks like it's failing to apply. Either I've done something wrong or it's depending on some extra patches ?
naoki1 has joined #linux-rockchip
naoki has quit [Quit: naoki]
naoki1 is now known as naoki
naoki has quit [Client Quit]
naoki1 has joined #linux-rockchip
<qschulz> rtp: see cover letter
naoki1 is now known as naoki
<rtp> oh, looks like I failed to read it entirely :(
naoki has quit [Client Quit]
<rtp> clone the repo. thansk
naoki has joined #linux-rockchip
<mmind00> dvergatal: not sure what you need their sales people for ... I'd think one of their distributors would be the better match, except when you plan on buying tenthousands of chips?
Woutervanh has quit [Quit: Client closed]
Woutervanh has joined #linux-rockchip
<mmind00> oh nice
hexdump0815 has quit [Quit: WeeChat 3.8]
hexdump0815 has joined #linux-rockchip
<diederik> https://review.trustedfirmware.org/c/ci/tf-a-ci-scripts/+/35514 is essential for the CI to succeed
Woutervanh has quit [Ping timeout: 240 seconds]
chewitt has quit [Quit: Zzz..]
psydroid has quit [Remote host closed the connection]
psydroid has joined #linux-rockchip
<Kwiboo> rtp: correct the ramboot series is currently depending on a few other patches, https://source.denx.de/u-boot/contributors/kwiboo/u-boot/-/commits/ramboot-v1 should have everything needed, + a few more not really needed ;-)
<dvergatal> mmind00: sorry i had to gone for a couple of hours
<dvergatal> go*
stikonas has joined #linux-rockchip
tlwoerner has joined #linux-rockchip
<qschulz> anyone with an Orange Pi 5+ or Rock 5 ITX by any chance? On RK3588 Jaguar, the USB stick connected to the USB-C ports is running in USB2-only when in reverse direction
<qschulz> and I want to know if it's something specific to Jaguar or not :)
<qschulz> obviously USB3 when in normal direction
<qschulz> s/direction/orientation/
<wens> qschulz: have one Orange Pi 5+, though I'm not at my place over the weekend
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
<Kwiboo> qschulz: not sure usb-c orientation switch is fully functional,
MyNetAz has quit [Remote host closed the connection]
<Kwiboo> had some issues on my rk356x board before, but may have been that it did not use a full fusb302
naoki has quit [Quit: naoki]
<qschulz> Kwiboo: ah good point, looking at the schematics I can see that the USB2 lines are routed the same regardless of the orientation
<rtp> Kwiboo: thanks to qschulz, I've used your git branch. It worked on roc-pc-rk3399 and rock5b. Managed to boot the 2 binaries with my current snagrecover branch
MyNetAz has joined #linux-rockchip
Woutervanh has joined #linux-rockchip
<Woutervanh> mmind00: thanks for the help so far. I already get the 2 cards now. No visual yet though, think I will need to fiddle a bit with panel/backlight
<mmind00> Woutervanh: great that you had progress
<qschulz> Woutervanh: you may want to generate a pattern with modetest -s, I sometimes have blank displays that work just fine
<Woutervanh> yeah, will continue on that after the weekend (y)
Woutervanh has quit [Client Quit]
<Kwiboo> rtp: for rk3588 devices there may be an issue with 16GB+ ram devices, forgot to mention that when sending out the series, when SPL+FIT (usb472) is sent to device, it will overwrite the rk-atags that signal the usable dram areas
<rtp> I wish I had a 32gb version of rock5b but that's not the case. When I bought it, there was only 16GB available. So, obviously, I can't run into this issue
<qschulz> Kwiboo: oh, that's bad
<qschulz> I assume there's no way to specify a specific offset for usb472 when uploading, it just starts at 0?
<qschulz> Kwiboo: why 16+ specifically, don't we have holes for anything above 4GB?
<rtp> I don't remember seeing some code mentioning you can specify the upload offset
<diederik> Kwiboo: is that >= 16GB or > 16GB?
<rtp> diederik: > 16GB (unless I'm lucky). It worked on my 16GB rock5b
<diederik> Thanks :) My 16GB Rock5b is on the way ;)
<diederik> (and considered the 32GB one, but didn't go for it (but it was an option when I ordered))
vagrantc has joined #linux-rockchip
<Kwiboo> qschulz: yeah, we possible need to add a fallback to handle the holes near 16 GiB in case rk-atags is not working, looking at https://gist.github.com/Kwiboo/1c020d37e3adbc9d0d79dc003d921977 it seem to affect boards with more than 12 GiB ram
<qschulz> Kwiboo: sorry to keep asking the same questions, but we're uploading to 471 and 472 and then only executing the code?
<qschulz> basically my idea being: can we upload 471, read the DRAM to store the ATAGS on the host, inject the ATAGS in the 472 payload and then upload to 472 to continue execution?
<Kwiboo> rtp: at least in the idblock image we can say what sram/dram an image should be loaded to, I am guessing it may also be possible with the 471/472 commands, have so far not seen any code that tries to load the second blob somewhere else beside start of dram
<Kwiboo> qschulz: maskrom seem to be executing the blob as soon as the last packet has been sent, i.e. once it receive a command indicating packet size != 4096 (or similar)
<qschulz> is there any command to read stuff from the DRAM is now the question :D because then once TPL returns to BROM we could read the DRAM for the ATAGS and inject them in the 472 payload
<Kwiboo> so maskrom starts waits on 471 commands, append the data to SRAM, and once data-packed lenght != 4096, jumps/runs the code, once TPL return-to-brom bootrom will listen for 472 commands and write to DRAM instead of SRAM
<Kwiboo> at least that is my understanding of how it seem to work :-)
<qschulz> that'd make sense
<Kwiboo> if we can signal at where in dram the 472 payload is written, we could put it somewhere else and just have a small relocation code (memmove) run before regular code
<Kwiboo> for the new rk3506 (armv7) they have put tee-os at 0x1000 and spl at ~63MiB, have not looked at how that works with maskrom download, but may suggest it can be possible to have it load to different address
<Kwiboo> for mainline I insted put SPL at 0x80000, mainly because that could possible work with RAM boot, if we find a way to signal the load address for 472 cmd :-), see https://source.denx.de/u-boot/contributors/kwiboo/u-boot/-/commit/3d683f3b717de010fffeece8712373892a599905 for a hack
<Kwiboo> and on rk3576 the bootsource-id seem to be 0x81 instead of 0xa, so another thing that could needed some tuning to work with RAM boot
franoosh has quit [Remote host closed the connection]
<Kwiboo> rtp: found your snagboot repo :-), please see https://source.denx.de/u-boot/contributors/kwiboo/u-boot/-/blob/wip/drivers/usb/gadget/Kconfig#L73 for a list of usb product ids used by a few more rockchip socs
<CounterPillow> oh this is useful
<CounterPillow> for my udev rules :)
<rtp> Kwiboo: oh, thanks. Will update it on monday :)
<mmind00> Kwiboo: rk3506 and armv7 ... whut? :-D
<CounterPillow> mmind00: yes :(
<CounterPillow> armv7 will never die
<mmind00> seems like Google does not yet know the rk3506
<CounterPillow> meanwhile, STM is rebranding their Cortex-A7 based designs to be targeted at "AI", because what really makes me think of the future is a design from 2011 on a foundry process from the 00s
<mmind00> but ... but ... but ... it has "AI" ... that has to count for something
* mmind00 somehow combined rk3506 with armv7 ... and searched for rk3507 ;-)
<CounterPillow> Kwiboo: oh neat, I wonder if the RK3506 on those is upstream-able unlike the RV110x chips they used previously which iirc needed some stuff that simply wasn't gonna fly
<Kwiboo> working mainline u-boot at https://source.denx.de/u-boot/contributors/kwiboo/u-boot/-/commits/rk3506, spi nand works in proper, not in SPL..., however usb and ethernet works
<CounterPillow> you work incredibly fast
<mmind00> Kwiboo: my arm32 branch is terribly empty still for 6.15 ... you can change that :-D
<CounterPillow> oh it uses the SAI IP for audio
<CounterPillow> I'm currently working on an upstream port of that
<mmind00> 3 cores? ... is that some sort of thing with one disabled core due to manufacturing?
<CounterPillow> would be a weird choice when there's no 4-core version
dsimic has quit [Ping timeout: 248 seconds]
dsimic has joined #linux-rockchip
<CounterPillow> Honestly, weird SoC. Claims to be aimed at audio applications and has plenty of audio interfaces but no DSP?
<robmur01> but hey you've got two whole A7s with NEON to run your DSP on while the third one runs the OS :D
raster has quit [Quit: Gettin' stinky!]
<CounterPillow> There's a mention of a "DSP0 Power" in the datasheet with no other mention of DSP as far as I can tell, so maybe there is a DSP and they just glossed over that essential feature
xha has quit [Quit: WeeChat 4.4.4]
Net147_ has quit [Quit: Quit]
Net147 has joined #linux-rockchip
Net147 has quit [Changing host]
Net147 has joined #linux-rockchip
xha has joined #linux-rockchip
naoki has joined #linux-rockchip