On 6/8/21 2:30 PM, Jorge Ramirez-Ortiz, Foundries wrote:
> On 08/06/21, Michal Simek wrote:
>> Hi,
>>
>> On 6/7/21 8:41 PM, Jorge Ramirez-Ortiz, Foundries wrote:
>>> On 07/06/21, Jorge Ramirez-Ortiz, Foundries wrote:
>>>> hi Michal
>>>>
>>>> um, when we exchanged emails about enabling ECC support for MPSoC, I
>>>> left with the understanding that there already was a DMA driver
>>>> available in u-boot that I could use to initialize the memory.
>>>>
>>>> do you have something in the works or will I have to write such a
>>>> driver? compatible would be 'xlnx,zynqmp-dma-1.0'
>>>
>>> alternatively, would it be acceptable - upstream - If I just init ECC
>>> as done by FSBL (ie, without a proper DMA driver, just accessing the
>>> controller)?
>>
>> There are 2 things here.
>> The first thing that would be the best to do it proper framework which
>> is likely dma where dma_memset should be likely added.
>> The second that this code should land in SPL which can't be big that's
>> why this feature should be very small.
>
> the way I am doing it ATM is just a verbatim of XFsbl_EccInit (grabs a
> channel and inits all memory to DEADBEEF). About 100 LOC total.
>
> I belive the FSBL copyright notice allows for that (below excerpt).
>
> [..] Permission is hereby granted, free of charge, to any person obtaining a
> copy
> of this software and associated documentation files (the "Software"), to deal
> in the Software without restriction, including without limitation the rights
> to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
> copies of the Software, and to permit persons to whom the Software is
> furnished to do so, [..]
>
> but yes this just works as expected.
You should talk to lawyer about it and I can't give you any
recommendation about it.
>
> a bit trickier is where to init from - for the particular product
> I am working on I can hardcode the addresses and sizes and init right
> after board_init_f in spl.c...havent had the time to think about
> something portable yet - if you have a suggestion I could send both
> patches (one for the spl_ecc_init(addr, len) function and another one
> calling it)
SPL dtb doesn't have memory node which is a surprise. But I expect ECC
is working on the memory which is written first. I am OK with adding
memory base/size for ECC to Kconfig till someone has time to take a look
at memory controller registers to find out which one has size in it that
can be read and use.
Thanks,
Michal
