Re: [OpenOCD-devel] GigaDevice GD32E230 support

[Michael S. <mi...@sc...>]

On 09.03.21 14:00, Andreas Bolsch wrote:
> On 2021-03-09 12:34, Michael Schwingen wrote:
>> Is there a documented way to get a *manufacturer* ID in some common
>> register that can be accessed without prior knowledge about the part? If
>> not, I would prefer to shift this decision to the configuration file.
>
> CoreSight ROM tables. Problems are that even there GD might have
> reused ST's, and
> there might be even devices without or with buggy ROM tables ...

That seems like a good idea. But:

After Uwe Bonnes' recommendation, I tested using backmagic. The results:

           AP Designer   JEP106 (w/o parity)

EFM32TG    0x673         Bank 7/0x73 = Energy Micro
XMC4500    0x41          Bank 0/0x41 = Infineon
STM32G031  0x20          Bank 0/0x20 = ST
MSP432     0x17          Bank 0/0x17 = TI
SAMD10     0x1F          Bank 0/0x1F = Atmel
GD32E230   0x43B         Bank 5/0x3B = ARM (!)

(AP designer IDs are from Jedec JEP106, with the first nibble being (bank number - 1))

It seems GigaDevice did not copy from ST, but nevertheless entered the wrong ID (from ARM). 
GigaDevice has two IDs listed in JEP106 (0x648 and 0x751) which they could have used ...

And yes, there are buggy ROM tables: I have an old LM3S6819 board 
(from before they were bought by TI) which reports "0" as AP Designer.

I lean towards not doing auto-detection - there may be more devices with buggy ROM tables in the future. 
Having a separate config file (with the correct driver or flag specified) for such a device is not such a big hassle and avoids any ambiguities.

Comments?

cu
Michael