Booting base-hw image on a Freescale i.MX6

Stefan Kalkowski stefan.kalkowski at ...1...
Mon Jan 20 16:05:51 CET 2014 

Hi Meenakshi,

On 01/18/2014 09:16 AM, Meenakshi Narayanan wrote:
> Hi Stefan,
> 
> I'm working with Aditya on the port. We're now trying to blink an LED at
> the start of execution of the kernel (at the place where you asked us to
> print a PDBG), since the JTAG is going to take a while. We've tried using
> i.MX53's GPIO driver by modifying the register addresses in board_base.h.

well, I don't think this is the way to success.

Because you didn't finished to get some output from the UART, possibly
because your kernel image isn't setup correctly, you're now trying to
rope the far more complex GPIO driver code into the kernel? As long as
things are working similar like within the i.MX53 SoC, moreover the GPIO
pins depend on a correctly configured IOMUX controller, which is a ball
of wires. Just to give you some complexity metric between both in form
of reference manual pages, the IOMUX controller comprises 800 pages, and
the UART about 80. To sum it up, setting up GPIO pins correctly is a far
more error-prone task than setting up the UART on this platform.

What about my advices within the last mail, compiling your own u-boot,
thereby having the ability to better check the setup of your kernel
(e.g. via additional prints), and not initializing the UART within the
kernel, because it's already done successfully by u-boot?

Moreover, you might start with a very easy binary (no complete Genode
kernel/image), which you write from scratch. That binary can be some
very simple assembler code, which only puts some character to the UART,
and nothing more. If that fails too, it's probably a boot issue, which I
would guess it is.

> 
> Basically, we need to write to the GPIO9 (GPIO1_IO09 in i.MX6) : Can this
> be accomplished by calling the read/write/direction functions in driver.h
> of os/src/driver/gpio. Or can we simply write to a register by implementing
> util/mmio.h?

Of course, you can use the classes in "util/mmio.h" to simply access I/O
memory mapped registers. In fact, the MMIO framework is designed for
that operation purpose.
The code in "os/src/driver/gpio" is used for the user-level GPIO driver
in Genode, and cannot be used directly within the base-hw kernel.

Regards
Stefan

> 
> Thanks and regards
> Meenakshi
> 
> 
> On Thu, Jan 9, 2014 at 3:14 PM, Stefan Kalkowski <
> stefan.kalkowski at ...1...> wrote:
> 
>> Hi Aditya,
>>
>> On 01/09/2014 08:30 AM, Aditya Kousik wrote:
>>> Hello Stefan,
>>>
>>> I was looking for alternate methods of running the program loaded at
>>> address 10001000. That was the only motive behind trying go. I kept using
>>> bootm till then. I wasn't exactly able to check where u-boot loaded its
>> own
>>> memory sections. So instead, I set the load address for the kernel way
>>> above at 20000000 where I'm sure nothing else clashed. And loaded the
>>> uImage to 20800000.
>>
>> I recommend using a self-compiled u-boot on you imx6 board. Therefore,
>> you'll be able to check its memory regions.
>>
>>>
>>> While it's still stuck at  'Starting kernel...' , I'll take your previous
>>> advice seriously. To use a JTAG debugger or recheck the UART definitions.
>>
>> Moreover, you might skip the UART initialization routine. Just comment
>> the body of the Imx31_uart_base constructor out in:
>>
>>   base/include/drivers/uart/imx31_uart_base.h
>>
>> The UART should be initialized by u-boot anyway.
>>
>> Of course, if you've a JTAG debugger present, and a corresponding
>> connector on your test board, this probably will be the most comfortable
>> path.
>>
>>>
>>> I ran test-printf on linux and this is what I got:
>>>
>>> spawn ./core
>>> int main(): --- create local services ---
>>> int main(): --- start init ---
>>> int main(): transferred 17592186044415 MB to init
>>> int main(): --- init created, waiting for exit condition ---
>>> [init] Could not open file "ld.lib.so"
>>> Quota exceeded! amount=4096, size=4096, consumed=4096
>>> [init] upgrading quota donation for SIGNAL session
>>> [init -> test-printf] -1 = -1 = -1
>>> Test succeeded
>>>
>>> Am I supposed to get similar messages from core on my board?
>>
>> Yes that's right.
>>
>> Regards
>> Stefan
>>
>>>
>>> Thanks in advance
>>> Aditya
>>>
>>>
>>> On Mon, Jan 6, 2014 at 3:28 PM, Stefan Kalkowski <
>>> stefan.kalkowski at ...1...> wrote:
>>>
>>>> Hi Aditya,
>>>>
>>>> On 01/06/2014 09:18 AM, Aditya Kousik wrote:
>>>>> Hello Stefan,
>>>>>
>>>>> I added the print debug call inside extern "C" void kernel() function
>> as
>>>>> you'd suggested. Still no response from the processor.
>>>>>
>>>>> I checked the address specifications in both Imx31_uart_base.h and also
>>>>> base/include/platform/imx6/board_base.h in Nikolay's branch. They are
>>>>> correct, to the dot. Even AIPS mapping in board.h in base-hw's core.
>> The
>>>>> definitions seem to be fine.
>>>>>
>>>>> Ideally, what is expected to happen, i.e, the expected course of
>> action?
>>>> we
>>>>> can't load the kernel by running 'bootm' since you say it uses a
>>>> different
>>>>> config. So what do we do with the uImage?
>>>>
>>>> Well, I can't follow your conclusions to not using bootm? You've a
>>>> complete scenario packed into an ELF image, which is built to be loaded
>>>> at start address 0x10001000. This ELF image is used by u-boot's mkimage
>>>> utility to form a uImage file. This uImage can be loaded to some address
>>>> above 0x10300000 e.g. via TFTP like you did it.
>>>> Assuming you've loaded the uImage to 0x10800000, you should boot the
>>>> scenario with "bootm 0x10800000". After that u-boot will interpret the
>>>> uImage header at that address, load the ELF image within the uImage to
>>>> the appropriated memory sections, prepare the machine registers, and
>>>> finally jump to 0x10001000.
>>>>
>>>>>
>>>>> 1. I loaded the uImage to 0x10800000 with TFTP as 'tftp 10800000'. It
>>>>> succesfully loaded the kernel to that address. At this point, I had
>>>> changed
>>>>> the load address to 10002000 to check if the image was read properly.
>> It
>>>>> did.
>>>>> 2. As a different approach, ran 'go 10002000'. This was what I got
>>>>> "##Application
>>>>> started at 0x10002000, Application terminated, rc = 0x100A4004"
>>>>
>>>> Using the 'go XXX' command let u-boot simply load that address into the
>>>> ip register. As you've loaded an uImage to the related address, the
>>>> machine executes the u-boot image header information, which of course
>>>> fails. Even if you load the ELF image by hand to the appropriated memory
>>>> sections, it is not recommended to use the 'go' command, as u-boot
>>>> doesn't prepare the machine as expected. For instance, the kernel gets
>>>> executed with the MMU enabled using u-boot's page-table setup etc.
>>>>
>>>>>
>>>>> What is the sequence of events, the programs that are sequentially
>>>> executed
>>>>> to say that the kernel is 'running'?
>>>>
>>>> In your case I assume:
>>>>
>>>>   tftp 0x10800000 <network path to uImage>
>>>>   bootm 0x10800000
>>>>
>>>> One further problem might be that u-boot's own memory sections (binary
>>>> data, exception vector) clashes with the scenario ones. Unfortunately,
>>>> mostly u-boot doesn't warn you if it fails to load the whole image due
>>>> to memory clashes. Therefore, it would be good, if it's possible to you,
>>>> to analyze the memory layout of the u-boot binary used on your platform,
>>>> e.g.: "readelf -S <u-boot binary>"
>>>>
>>>> If there's a clash you can change the "LD_TEXT_ADDR" variable in
>>>> "base-hw/mk/spec-hw_imx6.mk" to another non-clashing address.
>>>>
>>>> Regards
>>>> Stefan
>>>>
>>>>>
>>>>> Thanks in advance
>>>>> Aditya Kousik
>>>>>
>>>>>
>>>>> On Thu, Jan 2, 2014 at 4:33 PM, Stefan Kalkowski <
>>>>> stefan.kalkowski at ...1...> wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> On 01/02/2014 08:52 AM, Aditya Kousik wrote:
>>>>>>> Hello all,
>>>>>>>
>>>>>>> We've tried to run a basic test-printf script on a Freescale i.MX6
>>>>>>> sabrelite board. Based on
>>>>>>> this<http://sourceforge.net/p/genode/mailman/message/31760885/>,
>>>>>>> we implemented the necessary files. We had also come across an i.MX6
>>>>>> branch
>>>>>>> by Nikolay Golikov stating that he had a preliminary port of i.MX6
>>>> here:
>>>>>>> https://github.com/decaprox/genode/tree/i.mx6
>>>>>>>
>>>>>>> We ran RUN_OPT="--target=uboot" make run/printf to see if it does
>> print
>>>>>> "-1
>>>>>>> = -1 = -1" to the serial console (tested with gtkterm).
>>>>>>>
>>>>>>> We ran the boot image uImage through TFTP at load address 0x10800000.
>>>> It
>>>>>>> loads the kernel (267.9 kB) but stops at "Loading kernel".
>>>>>>
>>>>>> Assuming, your resulting kernel image is linked to 0x10001000, and
>> ends
>>>>>> at about 0x102cbc55, like the result I got when compiling Nikolay's
>>>>>> branch, your load address for the u-boot uImage sounds reasonable.
>>>>>> Unfortunately, I can't reproduce your experiments, because I don't
>> have
>>>>>> such a board available. The first thing that comes to my mind is, that
>>>>>> Nikolay used another board containing the i.MX6 SoC than you with
>>>>>> another memory layout of the related peripherals.
>>>>>>
>>>>>>>
>>>>>>> How do we know if the kernel is indeed running, or are we checking
>> the
>>>>>>> right UART ports? (we checked UART1)
>>>>>>
>>>>>> First at all, you should add a printf command like:
>>>>>>
>>>>>>   PDBG("kernel started!")"
>>>>>>
>>>>>> at the very beginning of the kernel's execution. In Nikolay's branch I
>>>>>> would add it here:
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>
>> https://github.com/decaprox/genode/blob/i.mx6/base-hw/src/core/kernel.cc#L1366
>>>>>>
>>>>>> If that message isn't printed, very likely your UART definitions
>> aren't
>>>>>> correct.
>>>>>>
>>>>>>>
>>>>>>> I gave "console=ttymxc0, 115200" as bootargs in u-boot. Any
>> suggestions
>>>>>> on
>>>>>>> how to proceed? Because we are literally stuck.
>>>>>>
>>>>>> Forget about "bootargs" in u-boot. They are specific to booting the
>>>>>> Linux kernel. U-boot puts them into the Linux/ARM specific ATAG
>>>>>> structure. In Genode that structure isn't used. It's hard coded in the
>>>>>> kernel, which UART is used for kernel/core debug messages. In
>> Nikolay's
>>>>>> branch, he uses the UART with memory mapped I/O located at 0x02020000,
>>>>>> and with the following UART definition:
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>
>> https://github.com/decaprox/genode/blob/i.mx6/base/include/drivers/uart/imx31_uart_base.h
>>>>>>
>>>>>> Please check whether this corresponds with your board's definitions.
>>>>>>
>>>>>>>
>>>>>>> Also, is there a good debugging tool for testing the genode
>> environment
>>>>>> on
>>>>>>> direct hardware?
>>>>>>
>>>>>> Well, there is JTAG of course, which isn't related to Genode
>>>>>> specifically. There are other debugging utils available on Genode,
>> like
>>>>>> GDB, but that wouldn't help on that low level you're currently working
>>>> at.
>>>>>>
>>>>>> Regards
>>>>>> Stefan
>>>>>>
>>>>>>>
>>>>>>> Thanks in advance
>>>>>>> Aditya Kousik
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>
>> ------------------------------------------------------------------------------
>>>>>>> Rapidly troubleshoot problems before they affect your business. Most
>> IT
>>>>>>> organizations don't have a clear picture of how application
>> performance
>>>>>>> affects their revenue. With AppDynamics, you get 100% visibility into
>>>>>> your
>>>>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of
>>>>>> AppDynamics Pro!
>>>>>>>
>>>>>>
>>>>
>> http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Genode-main mailing list
>>>>>>> Genode-main at lists.sourceforge.net
>>>>>>> https://lists.sourceforge.net/lists/listinfo/genode-main
>>>>>>>
>>>>>>
>>>>>> --
>>>>>> Stefan Kalkowski
>>>>>> Genode Labs
>>>>>>
>>>>>> http://www.genode-labs.com/ · http://genode.org/
>>>>>>
>>>>>>
>>>>>>
>>>>
>> ------------------------------------------------------------------------------
>>>>>> Rapidly troubleshoot problems before they affect your business. Most
>> IT
>>>>>> organizations don't have a clear picture of how application
>> performance
>>>>>> affects their revenue. With AppDynamics, you get 100% visibility into
>>>> your
>>>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of
>>>> AppDynamics
>>>>>> Pro!
>>>>>>
>>>>
>> http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
>>>>>> _______________________________________________
>>>>>> Genode-main mailing list
>>>>>> Genode-main at lists.sourceforge.net
>>>>>> https://lists.sourceforge.net/lists/listinfo/genode-main
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>> ------------------------------------------------------------------------------
>>>>> Rapidly troubleshoot problems before they affect your business. Most IT
>>>>> organizations don't have a clear picture of how application performance
>>>>> affects their revenue. With AppDynamics, you get 100% visibility into
>>>> your
>>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of
>>>> AppDynamics Pro!
>>>>>
>>>>
>> http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Genode-main mailing list
>>>>> Genode-main at lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/genode-main
>>>>>
>>>>
>>>> --
>>>> Stefan Kalkowski
>>>> Genode Labs
>>>>
>>>> http://www.genode-labs.com/ · http://genode.org/
>>>>
>>>>
>>>>
>> ------------------------------------------------------------------------------
>>>> Rapidly troubleshoot problems before they affect your business. Most IT
>>>> organizations don't have a clear picture of how application performance
>>>> affects their revenue. With AppDynamics, you get 100% visibility into
>> your
>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of
>> AppDynamics
>>>> Pro!
>>>>
>> http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
>>>> _______________________________________________
>>>> Genode-main mailing list
>>>> Genode-main at lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/genode-main
>>>>
>>>
>>>
>>>
>>>
>> ------------------------------------------------------------------------------
>>> CenturyLink Cloud: The Leader in Enterprise Cloud Services.
>>> Learn Why More Businesses Are Choosing CenturyLink Cloud For
>>> Critical Workloads, Development Environments & Everything In Between.
>>> Get a Quote or Start a Free Trial Today.
>>>
>> http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk
>>>
>>>
>>>
>>> _______________________________________________
>>> Genode-main mailing list
>>> Genode-main at lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/genode-main
>>>
>>
>> --
>> Stefan Kalkowski
>> Genode Labs
>>
>> http://www.genode-labs.com/ · http://genode.org/
>>
>>
>> ------------------------------------------------------------------------------
>> CenturyLink Cloud: The Leader in Enterprise Cloud Services.
>> Learn Why More Businesses Are Choosing CenturyLink Cloud For
>> Critical Workloads, Development Environments & Everything In Between.
>> Get a Quote or Start a Free Trial Today.
>>
>> http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk
>> _______________________________________________
>> Genode-main mailing list
>> Genode-main at lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/genode-main
>>
> 
> 
> 
> ------------------------------------------------------------------------------
> CenturyLink Cloud: The Leader in Enterprise Cloud Services.
> Learn Why More Businesses Are Choosing CenturyLink Cloud For
> Critical Workloads, Development Environments & Everything In Between.
> Get a Quote or Start a Free Trial Today. 
> http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk
> 
> 
> 
> _______________________________________________
> Genode-main mailing list
> Genode-main at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/genode-main
> 

-- 
Stefan Kalkowski
Genode Labs

http://www.genode-labs.com/ · http://genode.org/

More information about the users mailing list