Zynq UltraScale MPSoC, DMA driver

[Jorge Ramirez-Ortiz, Foundries]

On 08/06/21, 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.

1:1 when ported to SPL (except the coding standards)
https://github.com/Xilinx/embeddedsw/blob/master/lib/sw_apps/zynqmp_fsbl/src/xfsbl_initialization.c#L1574

> 
> 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.
> 
> 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)
> 
> thanks for your inputs
> 
> > 
> > Based on above I think that would be good to have long term both of them
> > in the tree. It means feel free to send short version and let's look at
> > it and dma one should be hopefully pretty much using the same code via
> > dma framework.
> > 
> > Thanks,
> > Michal
> >