Hi Geert-san,
> Hi Shimoda-san,
>
> On Wed, Apr 8, 2015 at 4:20 AM, Yoshihiro Shimoda
> <[email protected]> wrote:
> >> On Tue, Apr 7, 2015 at 9:53 AM, Yoshihiro Shimoda
> >> <[email protected]> wrote:
> >> > diff --git a/arch/arm/boot/dts/r8a7791.dtsi
> >> > b/arch/arm/boot/dts/r8a7791.dtsi
> >> > index db3772e..5f20ad4 100644
> >> > --- a/arch/arm/boot/dts/r8a7791.dtsi
> >> > +++ b/arch/arm/boot/dts/r8a7791.dtsi
> >> > @@ -722,6 +722,9 @@
> >> > renesas,buswait = <4>;
> >> > phys = <&usb0 1>;
> >> > phy-names = "usb";
> >> > + dmas = <&usb_dmac0 0>, <&usb_dmac0 1>,
> >> > + <&usb_dmac1 0>, <&usb_dmac1 1>;
> >> > + dma-names = "rx0", "tx1", "rx2", "tx3";
> >>
> >> The numbering looks a bit strange, given the code looks up both tx and rx
> >> DMA channels for each channel index. Is it correct?
> >>
> >> The binding documentation (which lacks on example) states:
> >>
> >> - dma-names : Must contain a list of DMA names:
> >> - tx0 ... tx<n>
> >> - rx0 ... rx<n>
> >> - This <n> means DnFIFO in USBHS module.
> >>
> >> As there are 4 DnFIFOs, I'd expect tx0..tx3 and rx0..rx3.
> >>
> >> Can you please clarify?
> >
> > I wrote some information below.
> > But, since it is complex hardware, I don't know I can explain this as
> > well...
> >
> > At first, R-Car Gen2 has 2 USB-DMACs and 4 channels:
> > USB-DMAC 0: ch0
> > USB-DMAC 0: ch1
> > USB-DMAC 1: ch0
> > USB-DMAC 1: ch1
> > Remarks: I don't know why but performance of USB-DMAC 0 is good than
> > USB-DMAC 1.
> > (Please refer to the Table 64.1 of the datasheet.)
> >
> > So, I wrote dmas parameter in hsusb as the following:
> >
> >> > + dmas = <&usb_dmac0 0>, <&usb_dmac0 1>,
> >> > + <&usb_dmac1 0>, <&usb_dmac1 1>;
> >
> > And, R-Car Gen2 has 4 DnFIFOs in HSUSB. to avoid complex handling for
> > DnFIFOs,
> > the renesas_usbhs driver uses each DnFIFO as TX or RX direction (not
> > bi-direction).
> > So, I wrote dma-names parameter as the following:
> >
> >> > + dma-names = "rx0", "tx1", "rx2", "tx3";
> >
> > - D0FIFO as RX (Also it is connected to USB-DMAC 0: ch0)
> > - D1FIFO as TX (Also it is connected to USB-DMAC 0: ch1)
> > - D2FIFO as RX (Also it is connected to USB-DMAC 1: ch0)
> > - D3FIFO as TX (Also it is connected to USB-DMAC 1: ch1)
>
> Thanks for the explanation!
>
> Hence the "strange" numbering is a limitation of the driver, not of the
> hardware?
Thank you for the point! That's correct.
> In that case I think the DT should describe the hardware, not the driver
> limitation. As the FIFOs are bidirectional, there's just a one-to-one mapping
> of 4 DnFIFOs to 4 (2 channels x 2 DMACs) DMA channels.
I understood it.
(Sometimes I forget that the DT should describe the hardware...)
> Which DnFIFO/channel is used for TX and which is used for RX is to be chosen
> by the driver, and can be fixed (e.g. use even channels for RX, odd for
> TX, like the current situation).
>
> What do you (and other people) think?
Thank you for the suggestion!
I will modify the renesas_usbhs driver.
Also I would like to revise the binding documentation as the following:
- dma-names : named "ch%u", where %u is the channel number ranging from zero
to the number of channels (DnFIFOs) minus one.
Best regards,
Yoshihiro Shimoda
> Gr{oetje,eeting}s,
>
> Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 --
> [email protected]
>
> 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
