xcasex has quit [Remote host closed the connection]
naobsd has joined #linux-rockchip
<loongson>
hello naobsd.
<naobsd>
loongson: hi
<loongson>
I've got a rk3288-based rock2 square board.
<loongson>
It has debian unstable on it, but no matter what I do it sticks with linux-image 3.10
<naobsd>
sounds nice
<loongson>
The debian repo has linux-image-4.1.0.2-armmp available and I installed it on my /dev/sda1 but the bootup from radxa totally ignores it and just uses whatever was flashed with the flash_update tool when I originall burned the rabian image to the board.
<loongson>
I was hoping rockchip could give a solution to this problem.
<loongson>
"/dev/mmcblk0p2" is mounted as "/" filesystem type ext4
<naobsd>
as far as I know, rockchip doesn't support debian, and debian doesn't support rockchip.
<loongson>
RADXA's selling point is Debian Linux on their boards. I bought the board with the expectation it would behave as normal Debian Distro behaves.
<naobsd>
but it doesn't matter, we can run debian on rockchip ourselves.
<naobsd>
well, radxa provides "rabian"
<loongson>
That implies when we update the kernel using the aptitude/apt-get install/synaptic, then it would boot up with the new kernel. I find it unacceptable to be forced to use flash_update and totally erase the emmc in order to get a kernel update when RADXA's decides to update their kernel which doesn't seem to be keeping up with intel's kernel happening which are up to kernel 4.2 from what I understand.
<naobsd>
well, only kernel can be updated without erasing other parts of emmc
<loongson>
i'm disappointed with the fact that radxa's kernel updates are not timely. That is also ROCKCHIP's problem since I purchased ROCKCHIP hardware, they should stand by it no matter what OS I prefer.
<loongson>
naobsd. How do I update the kernl without erasing other parts of the emmc? How do I do it without having to resort to a serial debug flash tool?
xcasex has joined #linux-rockchip
<loongson>
I have never had to resort to such activity with my intel/amd/loongson-based boxes
<naobsd>
or you can use dd from runnung system on RR2
<naobsd>
without serial console
<naobsd>
what is "such activity"?
<naobsd>
to flash/update data on storage, you have to do something with some external hardware such as keyboard/display/network/serial/usb/etc on intel/amd/loognson/anything else boxes
<naobsd>
I don't say rockchip is best at all
<naobsd>
different hardware needs different method/tool
<loongson>
naobsd, I am currently running/chatting with you from the radxa rock2 square board. I have a terminal open with root. I would have liked a command to use directly on the rock2 board within Debian Linux to upgrade the kernel. The upgrade_tool only works from an outside pc connected with a usb serial debug cable or a usb to otg cable. I certainly don't want to use the upgrade_tool.
<naobsd>
you can use dd
<loongson>
dd to where?
<loongson>
when I mount? I don't even the partition I am supposed to dd to.
<naobsd>
where your kernel is stored. it's configurable thing by user, and I don't have information about rabian and your configuration.
<naobsd>
where rootfs is placed? SD? eMMC? is there /dev/mmcblk1 too?
<loongson>
These kernels are from debian unstable. I just installed them to try. I sincerely believe all the different kernels for all the different soc's should be at debian repos rather than on manufacturer repos only.
<loongson>
There oriignal flash when onto emmc.
<naobsd>
loongson: totally agree. you(or anyone) can send patches to debian to support rockchip.
<loongson>
It's booting from emmc and all the debian packages get installed there also, but on a different partition as noted by the mount listing i gave in a pastebin link above.
akaizen has quit [Remote host closed the connection]
<loongson>
My expertise however isn't this low-level.
<naobsd>
then you're not using SD card and there is no /dev/mmcblk1, right?
<loongson>
I just wanted to use this board like I do any other motherboard.
<naobsd>
if you prefer intel, I recommend to use intel
<naobsd>
loongson: I'm asking "/dev/mmcblk1" is available or not! I can assume your configuration from your messages, but I cannot understand why you don't answer my question.
<naobsd>
anyway mmcblk0 should be emmc.
<naobsd>
and you don't have(insert) SD card at all
<loongson>
ls /dev/mmc* give me: "/dev/mmcblk0 /dev/mmcblk0p1 /dev/mmcblk0p2"
<loongson>
I have a an sdcard slot, but it's empty.
<loongson>
I have a ssd drive attached on a sata as /dev/sda1, but it's not configured at all to boot from it.
<naobsd>
hm, it's not clear why mmcblk0 is partitioned. I guess radxa's kernel doesn't support partition info in parameter
<naobsd>
it might have MBR partition, please try "fdisk -l /dev/mmcblk0"
<naobsd>
and
<loongson>
actually a while ago I tried a "fdisk --list-partitions" or "parted --list-partitions".
<naobsd>
dd if=/dev/mmcblk0p1 count=1 | hexdump -C | head -1
<naobsd>
it might be boot partition
<naobsd>
loongson: please tell me what you got by fdisk/parted too
<naobsd>
I confused a little. it's 3.10 kernel, it should support partition in parameter :(
<naobsd>
please do "dd if=/dev/mmcblk0p1 count=1 | hexdump -C | head -1"
<loongson>
naobsd, I want to encourage alternatives to intel, but Rockchip doesn't make it easy. With a bit more support directly from ROCKCHIP, they could be selling more devices and I would be making a profit by selling these devices. As it stands, I can't because it doesn't meet the minimum requirement for the usual debian linux user.
<loongson>
The same goes for the company LOONGSON, but they don't want to sell outside of CHINA. I would love to sell their product in Canada.
<loongson>
Their default os is Debian Linux.
<naobsd>
rockchip supports some open source communities(it may be inaccurate explanation), but debian(and many other cummunities) are not
<naobsd>
mmcblk0p1 is the boot partition which contains kernel. you can rewrite it from running system on RR2.
pizthewiz has quit [Quit:
<loongson>
naobsd: when you say open-source, you mean Android. I would prefer they support a standard full-blown GNU/Linux distro like Fedora or Arch or Debian.
<loongson>
naobsd, you have been extremely helpful. YOu are an expert. Do you work for rockchip?
<naobsd>
I think so too. please enforce rockchip to do it. or you can choose another hardware which has support
<naobsd>
no
<naobsd>
I'm just an individual
<loongson>
doumo arigatou naobsd-san :)
<naobsd>
(un)fortunately current my job is rockchip-based project but all my experience is came from my personal work
<naobsd>
I have no idea why you know I'm Japanese ;), anyway, dou itashi mashite, loongson-san :)
<loongson>
I'm just a small guy trying to make a living.
<loongson>
I would love to have a product to sell. I would have loved to have the radxa board be that very first product, but I have very high standards. As part of that standard, the boards/hardware need to be easily updateable without erasing everything.
<naobsd>
as I said, if you use kexec based loader, it's possible to load kernels from /boot/
<loongson>
It would have been nice to configure the one partition to be the one and only partition with the boot stuff within it.
<naobsd>
and it's possible to make kernel package which support RR2
<naobsd>
but it needs someone's time or money or something else
<naobsd>
RR2 and other Rockchip based products are originally for Android
<naobsd>
unmountable boot partition is very common
<naobsd>
flash tool on PC for updating is common too
<loongson>
If you know how to do this stuff, do a kickstarter and people will pay to get it done....I'm sure a lot of people would like to see this done for RR2 and for the firefly and the 100,000 tv boxes out there in the wild running on their tv boxes.
<loongson>
There is great potential for GNU/Linux on all these devices. The only thing that disappoints me is the RAM specification. 2GB RAM is not enough to do serious development/compiling of projects with. We need at least 8GB to 16 and even better 32GB RAM!!! That would make these rockchip based systems equally powerful to intel-based desktops.
<naobsd>
people always ask "better one please" even if people got what he want in first place... oh, just a monologue, please ignore ;)
<naobsd>
I think rockchip is good for some purpose, but not for everything
<loongson>
It could be good for a general purpose computer.
<naobsd>
well, sorry, now I'm busy a little
<loongson>
I'm using the RR2 for just that. I just want a newer kernel and more RAM.
<naobsd>
of course I hope a lot moer from Rockchip. I have many Rockchip based devices.
<loongson>
THANK you and cheers.
<naobsd>
my RK3368 box is not shipped yet :(
<naobsd>
later
johnnyr has joined #linux-rockchip
cnxsoft has joined #linux-rockchip
ssvb has quit [Ping timeout: 244 seconds]
<loongson>
cnxsoft...I read you blog all the time. Why is it all the tvboxes and minipc's netbooks, iot devices you mention....they don't have easy to update kernel paths? Why is it these manufacturers haven't gotten their act together with respect to this knowing all the vulnerabilities being identified by hackers everyday? Should you be encouraging the manufacturers to maintain due diligence on their part in order to prevent being
<loongson>
liable with respect to these vulnerabilities?
<loongson>
Over the course of one year with both the smarttv(samsung) and with a tronsmart q7 and the radxa rock2 square, I have seen on update for the smarttv, and none for either the q7 nor the rock2 square.
<loongson>
If I did, it would imply I would need to reflash the entire device with a new image. Can't the manufacturers make this less destructive by supplying a better update capability already it is 2015 after all.
<cnxsoft>
Yes, it would also be better to get the latest kernel release, or the latest patch with an LTS kernel
<cnxsoft>
But ARM does not offer a unified kernel (yet) like intel does, so manufacturers need to update each of their devices (or at least SoC), and that's expensive.
mrueg has quit [Remote host closed the connection]
mrueg has joined #linux-rockchip
iamarobot has joined #linux-rockchip
iamarobot has quit [Client Quit]
<naobsd>
personally I don't expect so much from cheap device
<naobsd>
and for development board, what I expect is "development work on it".
<naobsd>
if there is support from vendor such as better documents, source updates, etc, it will be better.
<naobsd>
but I'm fine with currently available document/source/schematic/etc for cheap development board.
<sjoerd>
cnxsoft: arm not having a unified kernel is mostly a problem created by the soc/board manufactorers themselves though ;)
markm_ has quit [Ping timeout: 260 seconds]
<naobsd>
ARM SoC != ARM core :)
dlezcano_ has joined #linux-rockchip
<rperier>
hi all
<sjoerd>
naobsd: well the arm cores themselves tend to be rather well supported, i've never seen big issues there tbh ;)
<naobsd>
sure :)
<mmind00>
maz_: do I remember correctly that you wanted to peek at a trustedfirmware binary for the rk3368, or was that mrutland? [just because I just found them individually in the recent rk3368 sdk drop from tronsmart]
markm has joined #linux-rockchip
<maz_>
mmind00: that'd be useful, assuming we have a way to reflash them!
<naobsd>
tony_: is it something like HDMI-MIPI CSI bridge?
<tony_>
naobsd, I am not very clear, just know they do cooperation with toshiba.
<tony_>
naobsd, TC358749XBG.
<naobsd>
tony_ that's exactly what I guessed...
<maz_>
mmind00: ah! I'll have a look.
<naobsd>
I have video camera with HDMI out. it will be nice if I can capure video and stream to network with hw encoder in rk3288 ;)
<naobsd>
mmind00, maz_: I tried to shrink tronsmart tarball... run make proper under kernel/ and u-boot/, then compress tarball with xz. I got 7.7GB taball :(
<naobsd>
still too large ;)
<tony_>
naobsd, there is a Chinese PDF.
<tony_>
naobsd, about HDMI-IN.
tony_ has quit [Remote host closed the connection]
tony_ has joined #linux-rockchip
apritzel has joined #linux-rockchip
naobsd has quit [Quit: naobsd]
dlezcano has quit [Remote host closed the connection]
dlezcano_ is now known as dlezcano
tony_ has quit [Remote host closed the connection]
<mrutland>
mmind00: If there were raw trusted firmware binaries and/or soruce that'd be interesting. I've not entirely figured out the format of the header for the trust partition yet
<mrutland>
mmind00: ah, frum that tgz, tools/rk_tools/RKTRUST/RK3368TRUST.ini looks like it might give me the hint I need :)
<mmind00>
mrutland: hehe, yep ... yep, binaries and this "trust_merger" thing
<mrutland>
How odd, bl30 bin file doesn't look to contain a64 instructions, though the bl31 binry does
markm_ has joined #linux-rockchip
markm has quit [Ping timeout: 250 seconds]
<mrutland>
Luckily my dumped partition seems to be all zeroes at the offset for the RSA signature, so I'm going to guess that there's no actual signature verification
visitor-14524 has joined #linux-rockchip
<visitor-14524>
Just wondering - is it possible to implement some sort of chainloading? U-Boot is said to be too large to put into the SDCard boot area, so how about something like the proprietary bootloader load U-Boot from NAND and lets that one go on from there?
<sjoerd>
visitor-14524: what soc were you using again ? 3188 or 3288?
<visitor-14524>
3188
<visitor-14524>
Proprietary bootloader V1.03 - at least, that's what the embedded backup image says it is.
<mmind00>
mrutland: I think that is optional and I guess the chineese device manufactures do not care to much ;-)
<sjoerd>
visitor-14524: but yes typically you get a SPL or the proprietary loader in rockchips case for 3188 in the boot area which then laods the main u-boot binary
cnxsoft1 has joined #linux-rockchip
cnxsoft has quit [Ping timeout: 252 seconds]
cnxsoft1 is now known as cnxsoft
<visitor-14524>
Thanks, guess I'll have to delve into U-Boot's docs on how to pull this off...
nighty^ has joined #linux-rockchip
<visitor-14524>
My biggest concern was that it wouldn't work if U-Boot itself is being loaded from the proprietary bootloader, rather than having that one replaced.
<sjoerd>
visitor-14524: if you look up the u-boot repos for the rk3188 they do exactly that iirc
<visitor-14524>
Will have a look, in any case.
visitor-14524 has quit [Quit: Page closed]
<lautriv_>
ha, appears i went in just in the right time ;) i managed to get uboot loading from SD, now it cycles endless in a retry. This was suspected since i did not provide any additional parameters or kernels but i can't find some documentation what the rk3188-u-boot wnats to find and where ?
ssvb has quit [Ping timeout: 268 seconds]
ssvb has joined #linux-rockchip
m-ric has joined #linux-rockchip
Apocx_ has joined #linux-rockchip
Apocx has quit [Ping timeout: 272 seconds]
cnxsoft has quit [Quit: cnxsoft]
JohnDoe_71Rus has joined #linux-rockchip
m-ric has quit [Remote host closed the connection]
ssvb has quit [Ping timeout: 244 seconds]
ssvb has joined #linux-rockchip
pietrushnic has quit [Ping timeout: 246 seconds]
ssvb has quit [Ping timeout: 256 seconds]
GriefNorth has joined #linux-rockchip
cristian_c has joined #linux-rockchip
<cristian_c>
apritzel: hello
cosm has joined #linux-rockchip
<rperier>
that's always fun when you deadlock a core... mainly when this is on a remote device... :D
<rperier>
o_O
pietrushnic has joined #linux-rockchip
* rperier
checks the box "epic fail"
* lautriv_
hands rperier a knife for the powercord
<rperier>
:D
rory096 has joined #linux-rockchip
gb_master has joined #linux-rockchip
cosm has quit [Ping timeout: 250 seconds]
_vaibhav_ has joined #linux-rockchip
ssvb has joined #linux-rockchip
<_vaibhav_>
Am having RK2928 HDMI dongle(chip version: 292X-2012.08.23-V300), I want to modify its system.img. I used "sudo rkflashtool r system > system.img" which is about 29MB, now I did file to get its type "system.img: Squashfs filesystem, little endian, version 4.0, 11873113 bytes, 371 inodes, blocksize: 131072 bytes, created: Fri Jul 24 15:21:19 201". Extracted the contents by using "unsquashfs system.img", size of extracted folder is 22MB. I modified
apritzel has quit [Ping timeout: 264 seconds]
_vaibhav_ has quit [Quit: Leaving.]
cristian_c has quit [Quit: Bye]
cosm has joined #linux-rockchip
akaizen has quit [Remote host closed the connection]
arokux1 has quit [*.net *.split]
ganbold has quit [*.net *.split]
arokux1 has joined #linux-rockchip
ganbold has joined #linux-rockchip
arokux1 has quit [*.net *.split]
ganbold has quit [*.net *.split]
nighty^_ has quit [Remote host closed the connection]
akaizen has quit [Remote host closed the connection]
cosm has quit [Ping timeout: 268 seconds]
robogoat has quit [Ping timeout: 250 seconds]
robogoat has joined #linux-rockchip
nighty^ has quit [Quit: Disappears in a puff of smoke]
gb_master has quit [Remote host closed the connection]
robogoat has quit [Ping timeout: 264 seconds]
robogoat has joined #linux-rockchip
akaizen has joined #linux-rockchip
<akaizen>
Hey is Naoki around? I met some Mozilla folks and showed them some RK3288 hardware and they're interested in it. I was wondering what the community thinks about a fully-open platform (or as close to fully-open as possible)
<amstan>
akaizen: he's in here as naobsd
<akaizen>
amstan: Thanks. YEa his handle doesnt seem to be here.. netsplit?
<amstan>
akaizen: he seems to login and logout every day, i don't think he has a bouncer
<amstan>
he's on japan time, so.... waking up just about now?
<akaizen>
Ooo whats a bouncer?
<amstan>
kinda like a proxy for irc, you connect to it via your irc client, then it connects to the real server
<akaizen>
Server in the cloud that maintains connectivity and sends you messages?
<amstan>
and saves all the coversations while you're offline
<amstan>
so you appear like you're online
<akaizen>
omg this is awesome
<amstan>
yep
<akaizen>
Is there a good open source software for this?
<amstan>
and by cloud... it could be a simple daemon on a server, nothing too fancy
<amstan>
i use znc
<amstan>
but idk if i feel comfortable recomending it
<akaizen>
avaiable on ubuntu? secure?
<akaizen>
How come?
<amstan>
it doesn't have the features i want, and it seems a little buggy in some ways
<amstan>
so... fully open?
<akaizen>
well at least some auditing
<akaizen>
I dont want some irc haxx0r to pwn my server
<akaizen>
I guess I can just use an AWS instance
<amstan>
you can always put it in a chroot
<amstan>
or a restricted user
<amstan>
anyweay..
<amstan>
back to rockchip
<akaizen>
:)
<amstan>
shameless self plug: the rockchip chromebooks currently are almost 100% open source, the only thing that it isn't is libmali
<akaizen>
Ive been speaking to ARM about Mali
nighty^ has joined #linux-rockchip
<akaizen>
We're hoping to create a fully open platform
<amstan>
so.. lima?
<akaizen>
I'm not sure with Luc is up to with that, but mostly on the Tamil stuff
<akaizen>
Last time I spoke to him and looked up the stuff he seemed pretty satisifed with the work on lima and wanted to turn it over to the community
<amstan>
neat
<amstan>
on my chromebooks i'm ok with just running libmali with X, seems to work just fine
<akaizen>
Fine isnt good enough for me or our use cases :)
<amstan>
i'm more of a practical person i guess
<amstan>
what is your use case?
<akaizen>
4K at 60fps, no visual artifacts
<akaizen>
we're building our own graphics service to get this done
<amstan>
i don't think that's doable on rk3288 actually
<akaizen>
A hardware engineer and rockchip engineer told me otherwise, they said they lack resources to develop the drivers properly
<amstan>
it requires a hdmi clock rate that's a little bit too unstable for that
<amstan>
dianders has been working on that driver for the chromebooks recently
<amstan>
he can comment some more
<akaizen>
what the state of the boot loaders?
<amstan>
we have a fully open coreboot/depthcharge on the chromebooks, i know there's a uboot port too
<akaizen>
I saw u-boot patches for open SPL and barebox and this 'petitboot'
<amstan>
i'm not too familiar with the uboot things though
<dianders>
I don't think you'll get compliant HDMI 4K at 60 fps on rk3288
<dianders>
Jitter on the PLLs is too bad
<akaizen>
What about 3368?
<dianders>
The HDMI IP block is supposed to support it, but the PLLs can't produce a stable enough clock.
<dianders>
3368 = I have no idea
<dianders>
...unless they changed PLL vendors or somehow figured out how to reduce the jitter, I doubt it will be different
<amstan>
akaizen: we had 4k at 30fps, if that still sounds good :)
<dianders>
Maybe 4K 30fps or 1080p 60fps...
<dianders>
Also: memory bandwidth of 4K 60fps is pretty intense. Maybe you could refresh the screen, but you'd have very little processing power to do anything.
<dianders>
Remember that SDRAM on this platform is limited to 666 MHz (DDR3) or 533 (LPDDR). You could try to overclock that I guess, but no idea what would happen...
<akaizen>
dianders: With the OpenGL ES 3 zero copy functions, and the VPU we can decode content and get it into triple buffers in time
<akaizen>
16.7ms for 60 fps and VSYNC
<dianders>
Well, sounds like a fun project if you want to work on it, but just saying...
<akaizen>
dianders: definitely, these are the goals we are trying to accomplish and possible strategies - I definitely want to hear feedback on what might go wrong
<akaizen>
amstan: From my understanding the core-boot / depthcharge is only for ChromeOS (depthcharge anyway)
<akaizen>
What about booting into linux (ubuntu/debian/gentoo/etc)?
<amstan>
akaizen: we're both chromeos developers :)
<akaizen>
Ah :)
<akaizen>
so coreboot -> depthcharge / u-boot / barebox -> OS. Is my understanding correct of the boot process?
<amstan>
akaizen: I also convinced leming(the archlinux-arm maintainer) to add support for the chromebooks
<bashintosh>
dianders: is the jitter issue you mentioned with the PLLs a direct observation after testing or a probability? I am working on getting a precise clock for a 2K LCD (eDP) with RK3288 and having a hard time with the PLLs so I am curious to know if that's a no-through road.