Re: [PATCH v4 2/4] drivers: dma: sh: Add DMAC driver for RZ/G2L SoC

[Geert Uytterhoeven <geert@linux-m68k.org>]

Hi Biju,

On Wed, Jul 28, 2021 at 1:58 PM Biju Das <biju.das.jz@bp.renesas.com> wrote:
> > On 28-07-21, 07:00, Biju Das wrote:
> > > > Sorry I dont like passing numbers like this :(
> > > >
> > > > Can you explain what is meant by each of the above values and looks
> > > > like some (if not all) can be derived (slave config as well as
> > > > transaction
> > > > properties)
> > >
> > >
> > > 0x11228 (Tx)
> > > 0x11220 (Rx)
> > >
> > > BIT 22:- TM :- Transfer Mode
> >
> > What are the values, here it seems 0
>
> Yes, that is correct single bit. 0 means single transfer mode, 1 block transfer mode.
>
> >
> > > Bits 16->19 :- DDS(Destination Data Size) --> 0x0001 (16 bits) Bits
> > > 12->15 :- SDS(Source Data size)--> 0x0001 (16 bits)
> >
> > use src_addr_width/dst_addr_width ..?
>
> We support 128,256,512 and 1024 bits as well. I will extend enum dma_slave_buswidth to support this in another patch.
> Is it ok?
>
> >
> > > Bit  11     :- Reserved
> > > Bits 8->10 :- Ack mode  --> 0x010 (Bus cycle mode)
> >
> > What does this mean?
>
> DMAACK output mode is coming from HW manual, A big table of around 230 entries for on chip request with dedicated values for the above bits.
>
> 0x000 -- Initial value
> 0x001 -- 001 (LEVEL Mode) (001 for MTU,PWM,CAN etccc
> 0x01x -- Bus cycle mode (010 for OSTM,I2C, SSIF)
> 0x1xx -- DMAACK not to output(SCIF)
>
> >
> > > Bit 7 :-  Reserved
> > > Bit 6:- LVL -->  Level -->0 (DMA request based on edge of thesignal)
> > > Bit 5:- HIEN -->  High Enable --> 1 (Detects a DMA request on rising
> > > edge of the signal) Bit 4:- LOEN --> Low Enable -->0 (Does not DMA
> > > request on falling edge of the signal) Bit 3:- REQD --> Request
> > > Direction ->1 (DMAREQ is Destination)
> >
> > how and what decides these values
> > It is now hardcoded in the client driver,
>
> It is SoC specific, coming from HW manual. Each on chip peripheral has it's own values.
> Even source address/Destination address of the on chip module is also part of that table.
>
> can we do that in dma driver
> > instead? While deriving most of the values?
>
> If we add this in DMA driver, it won't be generic. We need to prepare a big LUT(based on MID +RID) for all the peripherals
> If SSI then use a value from LUT, SCIF then another value like that.
>
> So please let me know how do we want to proceed here?

Looks like we should pass this in the dmas properties in DT instead,
either by increasing #dma-cells, or by encoding it with the MID/RID
value in the existing cell?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds