<jcarlos>
What's the difference between rkflashtool and fastboot when I want to flash a recovery.img in a RK3188 device? Are both methods using the same protocol?
<jcarlos>
Are there advantages using one or another?
avRoogh3 has joined #linux-rockchip
_whitelogger has joined #linux-rockchip
ganbold_ has joined #linux-rockchip
avRoogh3 has quit [Ping timeout: 252 seconds]
cnxsoft has quit [Ping timeout: 250 seconds]
mmind00 has joined #linux-rockchip
nighty-_ has joined #linux-rockchip
eebrah has quit [Ping timeout: 246 seconds]
eebrah has joined #linux-rockchip
mrueg has quit [Remote host closed the connection]
mrueg has joined #linux-rockchip
avRoogh3 has joined #linux-rockchip
jcarlos has quit [Read error: Connection reset by peer]
<jcarlos>
I'm going to reformulate my question :) Are rk3188 devices compatible with the fastboot protocol or you can only use the rkflashtool tool?
<ncrmnt>
jcarlos: As far as I know - only rkflashtool.
<jcarlos>
Thanks
<ncrmnt>
recovery can do adb sideload stuff
<ncrmnt>
But as for NAND flashing, fastboot is implemented in u-boot, and we have none here.
<ncrmnt>
I might be wrong though, since I haven't worked with fastboot. Just saw traces of it in u-boot sources and adb.
<jcarlos>
But rkflashtool is talking to uboot?
<ncrmnt>
jcarlos: No u-boot on rockchip.
<jcarlos>
No?
<ncrmnt>
Just crippled rockchip's crap they call boot.
<jcarlos>
Mmmm
<jcarlos>
Mmmm
<ncrmnt>
nope, iRom boots ddr init code, then runs bootloader. Both are obtained from IDB section of NAND. For the bootloader it's ugly, hacky as hell.
<jcarlos>
ncrmnt: I was talking with a BQ employee and he told me that his bootloader was u-boot
<ncrmnt>
Hm...
<jcarlos>
I'm going to look for the message.
<jcarlos>
It is in Spanish, I think.
<ncrmnt>
Word is about u-boot available for rockchip, but rockchip ain't giving it out and noone has the source (strange people)
<jcarlos>
ncrmnt: Well. That's another thing.
<ncrmnt>
The company I work for have an NDA with rockchip and I can tell you there's NO traces of u-boot out there.
<ncrmnt>
May be available by request, we haven't requested that yet, since after seing the kernel code we're not sure whether we'll be pulling off a project based on rockchip.
<jcarlos>
So the typical file such as RK3188Loader(L)_V1.24.bin is not u-boot?
<ncrmnt>
Nope, just some ugly hacky shit. Behaviour differs
<ncrmnt>
it read mtd layout by parsing kernel commandline mtdparts in parameter (sic!)
<ncrmnt>
and don't remove 'initrd=...' from there - it screws thing up for the bootloader
<ncrmnt>
and it doesn't even try to boot kernel.
<ncrmnt>
On some devices it reads 'boot' instead of kernel, and boot can be kernel + initrd in android format (+rkcrc), cramfs, crapfs... whatever. Depends on device
<jcarlos>
ncrmnt: I have found the message and I was wrong. He doesn't say his bootloader is based in u-boot.
<ncrmnt>
Mine (Pipo M6 Pro 3g) just boots of 'boot' mtd part that's in android format + rkcrced
<ncrmnt>
jcarlos: Heh, too bad. So we only have rumors of uboot out there... somewhere.
eebrah has quit [Quit: leaving]
<jcarlos>
ncrmnt: I suppose my device is one of those you referred to, because my kernel partition is empty :)
<ncrmnt>
jcarlos: mine had a kernel there, had kernel in 'boot', recovery and 3 or 4 kernel copies in backup. Took a while to figure out.
<jcarlos>
ncrmnt: But if you have a kernel in kernel partition, is that the kernel you boot from?
<ncrmnt>
Nooope.
<ncrmnt>
That almost drove me nuts when I started out.
<ncrmnt>
Turns out I can safely remove kernel mtd part - it's never used
<ncrmnt>
Whoever wrote that hell of a bootloader must have been smoking something... heavy.
<ncrmnt>
naobsd: I see. For now I switched to Picuntu's kernel tree, that has mali stuff hacked from allwinner, but couldn't get the Xorg working yet.
<ncrmnt>
ncrmnt: I'll be porting more recent mali stuff and see if I can run odroid's userspace blobs that look like more recent.
<naobsd>
as far as I know, only olegk0 did the work for mali source things for RK https://github.com/olegk0
<jcarlos>
ncrmnt: Can you please explain why is better to flash kernel.img to kernel and a ramdisk-only boot.img to boot?
<ncrmnt>
Yeah, saw his comments in code. For now all I get with xf86-video-mali is a blinky square where the window should be and the xorg segfaults. And both fbdev and fbturbo are giving me HUGE artifacts while rendering.
<naobsd>
olegk0 used odroid's blob at some point, I'm not sure about picuntu
<ncrmnt>
naobsd: neither am I since all links to picuntu downloads are broken right now.
<ncrmnt>
Btw, looks lile I should have the same mali blobs as olegk0, since they were all handed by rz2k from #linux-sunxi
<naobsd>
Allwinner's blob will work on RK as like as ODROID's blob. but I guess no any work nor source thing is done with allwinner.
<naobsd>
I think someone just use it
<naobsd>
and if you just need Xorg, you may skip mali things (for now)
<ncrmnt>
naobsd: Don't thing so. fbdev and fbturbo both give me HUGE artifacts
<ncrmnt>
basically if I drag a window it disappears. KDE's oxygen is blinky-messy stuff. gtk looks better, but dragging windows around is still unusable
akatik has joined #linux-rockchip
<naobsd>
can you try stock kernel for your device?
<ncrmnt>
With X.org? Never tried that.
<ncrmnt>
So far it does look like a bug in rk_fb somewhere(?). Since droid uses mali for most of the stuff getting mali's XAA accel under X should (theoretically) solve that.
<naobsd>
oh I didn't know that odroid uses mali for XAA
<ncrmnt>
With stock I won't be able to get toucscreen working, since it requires fixing TS CAPS, so that evdev would properly pick it and use it.
<naobsd>
what kind of 2D function are provided by mali?
<naobsd>
as far as I know, 2D functions are provided different unit in SoC. there might be something which can be used for 2D in mali, but I'm not sure about it
<ncrmnt>
Oops, sorry. EXA
<ncrmnt>
EXA: Solid, Copy, Composite
<ncrmnt>
AFK
<naobsd>
hmm
<naobsd>
well
<ncrmnt>
So far it looks like get buf 2 gives out nil virt addr, but I haven't found why. Yet. Needs a clear head and a spare evening.
<naobsd>
with stock kernel, TS may not work, but you can see GUI with or without artifacts
<ncrmnt>
I'll try that tomorrow. Thanks. See you around, bye.
<naobsd>
I think sunxi community doesn't use xorg-video-mali, they uses fbturbo
<naobsd>
probably no one tried blobs for allwinner with -mali driver
<naobsd>
(not sure, I may be wrong)
<naobsd>
and they will use lima things at some point
<naobsd>
ah, libv and ssvb are here ;)
ncrmnt has quit [Ping timeout: 245 seconds]
<jcarlos>
naobsd: Can you please explain why is better to flash kernel.img to kernel and a ramdisk-only boot.img to boot?
<ssvb>
naobsd: I'm reading backlog now
<ssvb>
naobsd: the -mali driver is just too slow on sunxi hardware for 2d graphics (because of exa abuse)
<ssvb>
naobsd: but it does work
<ssvb>
oh, ncrmnt has already left the channel...
<ssvb>
naobsd: the huge artifacts that ncrmnt got with the fbdev driver (which does not use or need mali btw) are hard to explain
<naobsd>
jcarlos: I don't say which is better, I just say it should work
<jcarlos>
You said it was a sure way
<naobsd>
ssvb: I see, thank you for explanation
<jcarlos>
But I would like to know why
<jcarlos>
ncrmnt said kernel partition is not used in a lot of rockchip based devices.
<naobsd>
I thought you think kernel partition can't be used without trying it
<jcarlos>
Ahh
<jcarlos>
Ok, ok
<naobsd>
"not used" is not "can't be used
<jcarlos>
I understand, thanks
<naobsd>
I don't have all devices, so I can't confirm myself everything
<naobsd>
personally I never seen any device which can't use kernel partition
<ssvb>
btw, any updates about lenovo a10? is it usable with normal linux or the bootloader is locked too hard?
<naobsd>
both boot-only and kernel+boot form are working from RK29 gen
<naobsd>
I also have interest about lenovo a10
<naobsd>
and asus memo pad 8
<naobsd>
there is some kind of "secure boot" system are provided by RK SDK, so they may be secure...