Hi Geert,
> From: Fabrizio Castro <fabrizio.castro.jz@xxxxxxxxxxx>
> Sent: 28 February 2025 17:29
> Subject: RE: [PATCH v4 3/7] dt-bindings: dma: rz-dmac: Document RZ/V2H(P) family of SoCs
>
> Hi Geert,
>
> > From: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> > Sent: 28 February 2025 16:44
> > Subject: Re: [PATCH v4 3/7] dt-bindings: dma: rz-dmac: Document RZ/V2H(P) family of SoCs
> >
> > Hi Fabrizio,
> >
> > On Fri, 28 Feb 2025 at 17:32, Fabrizio Castro
> > <fabrizio.castro.jz@xxxxxxxxxxx> wrote:
> > > > From: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> > > > On Fri, 28 Feb 2025 at 16:38, Fabrizio Castro
> > > > <fabrizio.castro.jz@xxxxxxxxxxx> wrote:
> > > > > > From: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> > > > > > On Fri, 28 Feb 2025 at 15:55, Fabrizio Castro
> > > > > > <fabrizio.castro.jz@xxxxxxxxxxx> wrote:
> > > > > > > > From: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> > > > > > > > On Thu, 27 Feb 2025 at 19:16, Fabrizio Castro
> > > > > > > > <fabrizio.castro.jz@xxxxxxxxxxx> wrote:
> > > > > > > > > > From: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> > > > > > > > > > Sent: 24 February 2025 12:44
> > > > > > > > > > Subject: Re: [PATCH v4 3/7] dt-bindings: dma: rz-dmac: Document RZ/V2H(P) family of
> > SoCs
> > > > > > > > > >
> > > > > > > > > > On Thu, 20 Feb 2025 at 16:01, Fabrizio Castro
> > > > > > > > > > <fabrizio.castro.jz@xxxxxxxxxxx> wrote:
> > > > > > > > > > > Document the Renesas RZ/V2H(P) family of SoCs DMAC block.
> > > > > > > > > > > The Renesas RZ/V2H(P) DMAC is very similar to the one found on the
> > > > > > > > > > > Renesas RZ/G2L family of SoCs, but there are some differences:
> > > > > > > > > > > * It only uses one register area
> > > > > > > > > > > * It only uses one clock
> > > > > > > > > > > * It only uses one reset
> > > > > > > > > > > * Instead of using MID/IRD it uses REQ NO/ACK NO
> > > > > > > > > > > * It is connected to the Interrupt Control Unit (ICU)
> > > > > > > > > > >
> > > > > > > > > > > Signed-off-by: Fabrizio Castro <fabrizio.castro.jz@xxxxxxxxxxx>
> > > > > > > > > >
> > > > > > > > > > > v1->v2:
> > > > > > > > > > > * Removed RZ/V2H DMAC example.
> > > > > > > > > > > * Improved the readability of the `if` statement.
> > > > > > > > > >
> > > > > > > > > > Thanks for the update!
> > > > > > > > > >
> > > > > > > > > > > --- a/Documentation/devicetree/bindings/dma/renesas,rz-dmac.yaml
> > > > > > > > > > > +++ b/Documentation/devicetree/bindings/dma/renesas,rz-dmac.yaml
> > > > > > > > > > > @@ -61,14 +66,22 @@ properties:
> > > > > > > > > > > '#dma-cells':
> > > > > > > > > > > const: 1
> > > > > > > > > > > description:
> > > > > > > > > > > - The cell specifies the encoded MID/RID values of the DMAC port
> > > > > > > > > > > - connected to the DMA client and the slave channel configuration
> > > > > > > > > > > - parameters.
> > > > > > > > > > > + For the RZ/A1H, RZ/Five, RZ/G2{L,LC,UL}, RZ/V2L, and RZ/G3S SoCs, the cell
> > > > > > > > > > > + specifies the encoded MID/RID values of the DMAC port connected to the
> > > > > > > > > > > + DMA client and the slave channel configuration parameters.
> > > > > > > > > > > bits[0:9] - Specifies MID/RID value
> > > > > > > > > > > bit[10] - Specifies DMA request high enable (HIEN)
> > > > > > > > > > > bit[11] - Specifies DMA request detection type (LVL)
> > > > > > > > > > > bits[12:14] - Specifies DMAACK output mode (AM)
> > > > > > > > > > > bit[15] - Specifies Transfer Mode (TM)
> > > > > > > > > > > + For the RZ/V2H(P) SoC the cell specifies the REQ NO, the ACK NO, and the
> > > > > > > > > > > + slave channel configuration parameters.
> > > > > > > > > > > + bits[0:9] - Specifies the REQ NO
> > > > > > > > > >
> > > > > > > > > > So REQ_NO is the new name for MID/RID.
> > > > > > > >
> > > > > > > > These are documented in Table 4.7-22 ("DMA Transfer Request Detection
> > > > > > > > Operation Setting Table").
> > > > > > >
> > > > > > > REQ_NO is documented in both Table 4.7-22 and in Table 4.6-23 (column `DMAC No.`).
> > > > > >
> > > > > > Indeed. But not for all of them. E.g. RSPI is missing, IIC is present.
> > > > >
> > > > > I can see the RSPI related `REQ No.` in the version of the manual I am using,
> > > > > although one must be very careful to look at the right entry in the table,
> > > > > as the table is quite big, and the entries are ordered by `SPI No.`.
> > > > >
> > > > > For some devices, the SPI numbers are not contiguous therefore the device specific
> > > > > bits may end up scattered.
> > > > > For example, for `Name` `RSPI_CH0_sp_rxintpls_n` (mind that the `pls_n` substring
> > > > > is on a new line in the table) you can see from Table 4.6-23 that
> > > > > its `DMAC No.` is 140 (as you said, in decimal...).
> > > >
> > > > Thanks, I had missed it because the RSPI interrupts are spread across
> > > > two places...
> > > >
> > > > > > And the numbers are shown in decimal instead of in hex ;-)
> > > > > >
> > > > > > > > > It's certainly similar. I would say that REQ_NO + ACK_NO is the new MID_RID.
> > > > > > > > >
> > > > > > > > > > > + bits[10:16] - Specifies the ACK NO
> > > > > > > > > >
> > > > > > > > > > This is a new field.
> > > > > > > > > > However, it is not clear to me which value to specify here, and if this
> > > > > > > > > > is a hardware property at all, and thus needs to be specified in DT?
> > > > > > > > >
> > > > > > > > > It is a HW property. The value to set can be found in Table 4.6-27 from
> > > > > > > > > the HW User Manual, column "Ack No".
> > > > > > > >
> > > > > > > > Thanks, but that table only shows values for SPDIF, SCU, SSIU and PFC
> > > > > > > > (for external DMA requests). The most familiar DMA clients listed
> > > > > > > > in Table 4.7-22 are missing. E.g. RSPI0 uses REQ_NO 0x8C/0x8D, but
> > > > > > > > which values does it need for ACK_NO?
> > > > > > >
> > > > > > > Only a handful of devices need it. For every other device (and use case) only the
> > > > > > > default value is needed.
> > > > > >
> > > > > > The default value is RZV2H_ICU_DMAC_ACK_NO_DEFAULT = 0x7f?
> > > >
> > > > If you take this out, how to distinguish between ACK_NO = 0 and
> > > > the default?
> > >
> > > I am not sure I understand what you mean, so my answer here may be completely off.
> > >
> > > ACK No. 0 corresponds to SPDIF, CH0, TX, while ACK No. 0x7F is not valid.
> >
> > OK, that was my understanding, too.
> >
> > > My understanding of this is that there is a DACK_SEL field per ACK No (23 ICU_DMACKSELk
> > > registers, 4 DACK_SEL fields per ICU_DMACKSELk registers -> 23 * 4 = 92 DACK_SEL fields),
> > > to match the 92 ACK numbers listed in Table 4.6-27.
> > >
> > > Each DACK_SEL field should contain the global channel index (5 DMACs, 16 channels per DMAC
> > > -> 5 * 16 = 80 channels in total) associated to the ACK No.
> > > If DACK_SEL contains a valid channel number (0-79), then the corresponding signal
> > > gets controlled accordingly, otherwise a fixed output is generated instead.
> > >
> > > Mind that the code I sent wasn't dealing with it properly, but wasn't spotted due
> > > to limited testing capabilities, and it's safe to take out, as the DACK_SEL fields
> > > will all contain invalid channel numbers by default.
> > >
> > > Looking ahead, there is a similar scenario with the TEND signals as well.
> > >
> > > So for now the plan is to upstream support for memory/memory and device/memory (REQ No.,
> > > tested with RSPI), add support for ACK No later (perhaps testing it with audio, or via
> > > an external device), and finally TEND No if we get to it.
> >
> > So which values will you put in the dmas property for RSPI?
> > I assume:
> > bits[0:9] - Specifies REQ_NO value
> > bit[10] - Specifies DMA request high enable (HIEN)
> > bit[11] - Specifies DMA request detection type (LVL)
> > bits[12:14] - Specifies DMAACK output mode (AM)
> > bit[15] - Specifies Transfer Mode (TM)
I will switch to this layout for the next version.
> > i.e. all remaining bits will be zero?
>
> I see what you mean now. And there would be an ABI mismatch between older DTs and newer kernels,
> newer kernels would interpret the values incorrectly (as you said, 0 is a valid number).
>
> >
> > How do you plan to handle adding ACK_NO bits later?
> > I.e. how to distinguish between remaining bits zero and remaining
> > bits containing a valid ACK_NO value (which can be zero, for SPDIF)?
>
> We could add the ACK No. and the TEND No. to the binding (after TM), and implement it
> later in the driver (once we have some practical use case for them)?
>
> There are also a couple of alternatives:
> * we could add 1 to ACK No. and TEND No.? At that point 0 would be an invalid number?
> In which case we could add DT and driver support later?
> * we could fill up the remaining bits with 1s? We only have 5 TEND numbers, so 0b111 would
> be invalid. Similarly, we have 92 ACK numbers, so 0b1111111 would also be invalid. Also
> this shouldn't break the ABIs once we get around to add the rest?
Considering that `ACK No.` and `TEND No`. are uniquely paired to `REQ No.`, I think we can
omit them from the DT, and just look them up from a table in the corresponding driver, and
we should be good to go.
Thanks!
Fab
>
> >
> > I hope I made myself clear this time.
>
> Very clear! Thank you.
>
> > If not, weekend time ;-)
> >
> > Have a nice weekend!
>
> Thank you, and you!
>
> Cheers,
> Fab
>
> >
> > Gr{oetje,eeting}s,
> >
> > Geert
> >
> > --
> > Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
> >
> > 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
