Re: [OpenOCD-devel] [PATCH 0/3] Corrections to TMS570 Flash API based F021 loader

[Andrey S. <and...@gm...>]

On Wed, Aug 27, 2014 at 2:09 PM, Pavel Pisa <pi...@cm...> wrote:

> Hello Andrey,
>
> On Wednesday 27 of August 2014 20:49:56 Andrey Smirnov wrote:
> > Thank you for your patches! Sorry for the bugs in my driver and the
> > troubles they've caused you, I am glad to see that you were able to
> > overcome them. Since the changeset was not merged into master I think I
> am
> > going to go ahead and just squash your patches into mine, if that's
> alright
> > with you.
>
> The pleasure is on my side. Your work is great and the code was so
> well structured and prepared for testing that I quickly found how
> to compile it with debug and find the problem. In the fact, I found
> problem because I have fear to test on the start of the flash
> and I started with sectors at 1MB. We have problems that we have
> flashed probably something really bad at start of the flash by Ti
> tools and CPU has gone unusable and there was no way to resolve that.
> It seems that there is no way to reset CPU and stop exactly after
> reset without previous connection to the MCU. But if it is possible
> to access CPU that it is possible to put HW breakpoint at address
> zero and then use "reset halt" and it really stops at PC = 0
> to debug flash startup.
>
>
> https://github.com/AoLaD/rtems-tms570-utils/blob/master/openocd/openocd-halt-immediatelly.cfg
>
> But I have (dared) to test the start of the flash on the end of the
> day and OpenOCD was able to erase our initializing application
> build by Ti tools and then reflash it again.
>

Oh yes! I believe I know what you're talking about. What I ran into when I
was developing the driver is that if you write illegal opcodes at the
beginning of the flash the CPU would try to execute it and get into one of
the exception modes. Which would result in  OpenOCD not being able to halt
it after that. And since Cortex-R is a derivative of Cortex-A there's is no
reset capabilities of Cortex-M and I don't think there's a way to reset it
into a halted state without using external reset signal.

The way I solved it was to press the reset button on my devkit issue "reset
halt" in OpenOCD's telnet interface and then release reset button. That
would get my board into a halted state and after that sector erase would
fix the flash.


>
> Feel free to squash patches the way you like it.
> I would be really happy if F021 support gets to OpenOCD mainline.
>
> > Regarding your port of RTEMS, last time I checked it offered support for
> > Ada language and that was something I and my friend Andrey Yurovsky were
> > personally interested in. Do you have any plans working on that?
>
> RTEMS and OAR has long track record of Ada support. But I do know
> only that what is written on RTEMS pages. I have never used Ada.
> I know some people from UPV Valencia and Unican Cantabria who
> have strong love for Ada. If you have some project I can contact
> them if they have interres. But Ada is really not my pot of tea.
>

OK, If I or any of my friends ever do anything interested in Ada I'll make
sure to drop you a line!


>
> > Also have you seen this OS http://erika.tuxfamily.org/drupal/ ? It is
> > supposed to be sort of endorsed by Linux Foundation
> >
> http://wiki.automotive.linuxfoundation.org/index.php/Erika_Enterprise_Open-
> >Source_RTOS
>
> I know Evidence people really well from common projects related
> to Linux resources and admission control etc. I have seen Erica
> code first many ears ago. Even before its repository got public.
> But again we have not contract or research project to fund
> our university group work with that OS. I like that code,
> it is smaller than RTEMS and if we have time or need I would
> like to poke with it. I have tried to convice our group
> leader to start hard way with Erica instead of Ti offered
> FreeRTOS but choice was water Ti supported solution.
> I think still that Erica would be better for long term run.
>

All right, good to know that you are  aware of that project. It's just that
OS is relatively obscure and I wanted to make sure you knew it existed.

Thanks again for testing my patches, and I'll keep you posted.

Best regards,
Andrey