[ECOS] EDOSK2674 GDB stub problem

[ECOS] EDOSK2674 GDB stub problem

[Uwe Kindler]

Hello,

I almost finished porting to a new architecture (Renesas H8S), variant
(H8S/2674) and platform (EDOSK2674).

When I build GDB stubs for the EDOSK board and use these stubs for
debugging, then everything works fine.
I can connect to target, upload, single step, set breakpoints and interrupt
running programs with ctrl c. It is no
problem to disconnect and reconnect to target.


If I now build Redboot with GDB stubs included then debugging works fine.
But when I disconnect from target then
it is not possible to reconnect and I cannot communicate with Redboot with
the CLI annymore.

Does someone have any idea why the GDB stubs work fine outside of Redboot
and crash when deconnecting if used inside Redboot.
I do not have any idea and it is almost impossible to debug this.

Best Regards, Uwe


-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss

Re: [ECOS] EDOSK2674 GDB stub problem

[Nick Garnett]

"Uwe Kindler" <uwe_kindler@web.de> writes:

> Hello,
> 
> I almost finished porting to a new architecture (Renesas H8S), variant
> (H8S/2674) and platform (EDOSK2674).

How does this differ from the EDOSK2674 support that is already
present in the H8/300 HAL?

> 
> When I build GDB stubs for the EDOSK board and use these stubs for
> debugging, then everything works fine.
> I can connect to target, upload, single step, set breakpoints and interrupt
> running programs with ctrl c. It is no
> problem to disconnect and reconnect to target.
> 
> 
> If I now build Redboot with GDB stubs included then debugging works fine.
> But when I disconnect from target then
> it is not possible to reconnect and I cannot communicate with Redboot with
> the CLI annymore.
> 
> Does someone have any idea why the GDB stubs work fine outside of Redboot
> and crash when deconnecting if used inside Redboot.
> I do not have any idea and it is almost impossible to debug this.

Support for reconnecting to GDB is mainly supported by resetting the
system. Since there is no way of knowing what the application has done
to the hardware, the safest way of dealing with this is to force a
hardware reset. However, RedBoot does not do this by default, the
command "maintenance packet r" to GDB will cause RedBoot to reboot
when GDB disconnects.

There is support for newlib based applications to call an exit syscall
which should return to the RedBoot prompt. However, I don't know
whether the H8/300 support in newlib provides this.


-- 
Nick Garnett                    eCos Kernel Architect
http://www.ecoscentric.com      The eCos and RedBoot experts


-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss

Re: [ECOS] EDOSK2674 GDB stub problem

[Uwe Kindler]

Hello,

> How does this differ from the EDOSK2674 support that is already
> present in the H8/300 HAL?

The new EDOSK2674 port introduces a new architecture - the H8S architecture.
In my opinion this is a new architecture because the CPU architecture
differs from the H8/300 architecture. Mixing both architectures in one eCos
architecture makes the whole think more complex and harder to understand -
in my opinion. Therefore I created the new H8S architecture which cover
H8S/2000 and H8S/2600 CPU s.

The differrent variants are the H8S microprocessors from Renesas (i.e.
H8S/2357 or H8S/2674). The processors differ in the internal modules and
interrupt vector tables and therefore they are axcellent for eCos variant
ports.

The platforms are boards like the EDOSK2674 or the Cetoni MCU/2357 board
(from the company where I do my practical training).

Somme additional differences of the new EDOSK2674 port:
- the context switch code rewritten  (in my opinion the old code is not
completely right - at least i do not unterstand it)
- context switch and exception assembler code are optimised for speed
- the whole source code is well documented
- the CDL files for the configuration utility are more clear
- a lot of minor changes and rewritten code
...

The port is based on the EDOSK 2674 of Yoshinori Sato but some parts
(context switch, trampoline code, VSR-table and ISR tables...) are
completely rewritten.

Status:
- Redboot is running
- GDB debugging works well
- different tests run successfully

Thank you very much for your hint about "maintenance packet r" - this works
well for the EDOSK platform.

Best Regards,

Uwe




-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss