[U-Boot] Rockchip RK3288 u-boot with mainline kernel

[Heiko Stübner]

Am Dienstag, 29. November 2016, 14:40:49 schrieb Simon Glass:
> Hi Rick,
> 
> On 29 November 2016 at 11:22, Rick Bronson <rick at efn.org> wrote:
> > Hi Heiko and Simon,
> > 
> >   Thank you both for your help, I really appreciate it.
> >   
> >   No, I do not have the Linux mainline running yet as I was focusing
> > 
> > on getting the mainline u-boot running since I saw some references
> > that implied I may need that to use the mainline Linux.  I did try the
> > mainline kernel with the vendor-fork u-boot but got nothing pas the
> > "Loading Linux...." line from u-boot.
> > 
> >   I think what I will probably do is take the easy way out and use the
> > 
> > vendor-fork u-boot since that boots on this hardware.
> > 
> >   But since I've invested some amount of time in the mainline u-boot
> > 
> > and it's drinving me mad that I can't seem to figure out what's wrong,
> > I do have one last question (hopefully).
> 
> It would certainly make sense to invest in getting this running if you can.
> 
> >   I've got mainline u-boot getting this far (I added debug to show
> > 
> > what GPIO0 and 2 are outputting).
> > 
> > --------------------
> > U-Boot SPL 2016.11-00138-g136179b-dirty (Nov 28 2016 - 13:29:59)
> > Trying to boot from MMC1
> > Card did not respond to voltage select! GPIO0=0x0 GPIO2=0x0
> > ...
> > --------------------
> > 
> >   I've modified rk3288-firefly.dts (the closest thing to hardware I
> > 
> > have) to set the eMMC reset line high but as you see above, nothing is
> > set on any GPIO.  This leads me to believe that I do not understand
> > how the device tree is loaded/used in SPL.  Can anyone point me to
> > help on whether I should use u-boot-spl-nodtb or u-boot-spl-dtb (I've
> > tried both) or how to debug if I am in fact even reading my dtb.
> 
> You should be able to use 'gpio status' to see the GPIO state once you
> get to U-Boot.
> 
> But if your board is booting SPL from MMC how can it possibly fail to
> load U-Boot from MMC? The MMC GPIO must already work for you to get
> this far. Or are you loading SPL from something other than MMC?

Or you can try enabling the 
	CONFIG_ROCKCHIP_SPL_BACK_TO_BROM=y
option, which will simply let the on-chip rom also load the 2nd stage uboot 
seemingly using the same method as for the spl.

I'm currently testing Simon's network patches and this option works quite 
nicely.


Heiko

> You want u-boot-spl.bin (which includes the device tree).
> 
> >   Thanks again,
> >   
> >   Rick
> > 
> > PS.  My Makefile is here if you are so inclined:
> > http://members.efn.org/~rick/pub/Makefile  See the targets u-boot-denx
> > and uboot_new_flash to see what I've tried.
> 
> You could push your tree somewhere or a patch showing what changes you
> have made.
> 
> Regards,
> Simon
> 
> >> From heiko at sntech.de Tue Nov 29 08:45:02 2016
> >> To: u-boot at lists.denx.de
> >> Cc: Simon Glass <sjg at chromium.org>, Rick Bronson <rick at efn.org>,
> > 
> > "eddie.cai" <eddie.cai at rock-chips.com>,
> > linux-rockchip at lists.infradead.org
> > 
> >> Subject: Re: [U-Boot] Rockchip RK3288 u-boot with mainline kernel
> >> Date: Tue, 29 Nov 2016 11:20:52 +0100
> >> 
> >> Hi Rick,
> >> 
> >> Am Montag, 28. November 2016, 15:09:05 schrieb Simon Glass:
> >> > + A few rockchip people and linux-rockchip
> >> > 
> >> > Hi Rick,
> >> > 
> >> > On 25 November 2016 at 11:20, Rick Bronson <rick at efn.org> wrote:
> >> > > Hi All,
> >> > > 
> >> > >   I've got unsupported RK3288 hardware running the latest git u-boot
> >> > >   to
> >> > > 
> >> > > SPL as explained in
> >> > > http://git.denx.de/?p=u-boot.git;a=blob;f=doc/README.rockchip.  My
> >> > > goal
> >> > > is to run the mainline (ie. not Android) Linux kernel on this
> >> > > hardware
> >> > > 
> >> > > and wondered:
> >> > >  - Do I need to get the latest git u-boot to run before I can run the
> >> > > 
> >> > > mainline kernel?  Or can I use
> >> > > github.com/linux-rockchip/u-boot-rockchip.git, which I have running
> >> > > u-boot fully.
> >> > 
> >> > It's up to you - obviously mainline is where the development should
> >> > be, but there is no requirement that I know of.
> >> 
> >> correct, the (mainline-)kernel runs just fine on both the vendor-fork of
> > 
> > uboot
> > 
> >> as well as on mainline.
> >> 
> >> > Does mainline run on your board?
> >> > 
> >> > >  - The device tree seems to be in two places, once via:
> >> > >         resource_tool --image=resource2.img --pack linux/logo.bmp
> >> > >         ${DTS}.dtb
> >> > >   
> >> > >   that gets put into the resource file and then again at the end of
> >> > >   the
> >> > > 
> >> > > kernel via CONFIG_ARM_APPENDED_DTB.  Do I need both?  When I do both
> >> > > I get things like:
> >> > > 
> >> > > Unknow param: MACHINE_MODEL:rk30sdk!
> >> > > Unknow param: MACHINE_ID:007!
> >> 
> >> ARM_APPEND_DTB is meant for boards where the bootloader cannot load the
> >> devicetree (to old or so) and also cannot be reasonably exchanged. So the
> >> append-mechanism was invented to allow bundling the devicetree with the
> > 
> > actual
> > 
> >> kernel image, so that to the bootloader it looks like just any other
> >> kernel
> >> image.
> >> 
> >> So you essentially only need one or the other. Also at least mainline
> >> uboot
> >> also supports the FIT image type, where you can bundle the devicetree in
> >> a
> >> more generalized way.
> >> 
> >> For your message I would guess the kernel didn't find a usable devicetree
> >> somehow and was falling back to ATAGS-based board selection?