[PATCH v5 1/4] omap: mmc: Avoid using libfdt with of-platdata

[Simon Glass]

Hi Faiz,

+Walter Lozano

On Thu, 14 May 2020 at 02:43, Faiz Abbas <faiz_abbas at ti.com> wrote:
>
> Simon,
>
> On 05/05/20 12:20 pm, Faiz Abbas wrote:
> > Hi,
> >
> > On 04/05/20 6:44 pm, Simon Glass wrote:
> >> Hi Bart,
> >>
> >> On Mon, 4 May 2020 at 01:10, Bartosz Golaszewski <brgl at bgdev.pl> wrote:
> >>>
> >>> pt., 1 maj 2020 o 20:32 Tom Rini <trini at konsulko.com> napisał(a):
> >>>>
> >>>> On Thu, Apr 30, 2020 at 01:43:30PM +0200, Bartosz Golaszewski wrote:
> >>>>> wt., 28 kwi 2020 o 09:01 Faiz Abbas <faiz_abbas at ti.com> napisał(a):
> >>>>>>
> >>>>>> +Bartosz
> >>>>>>
> >>>>>> On 28/04/20 9:47 am, Lokesh Vutla wrote:
> >>>>>>> +Faiz,
> >>>>>>>
> >>>>>>> On 28/04/20 12:29 AM, Tom Rini wrote:
> >>>>>>>> On Mon, Apr 27, 2020 at 05:33:41AM +0000, Peng Fan wrote:
> >>>>>>>>>> Subject: [PATCH v5 1/4] omap: mmc: Avoid using libfdt with of-platdata
> >>>>>>>>>>
> >>>>>>>>>> At present this driver is enabled in SPL on omapl138_lcdk, which uses
> >>>>>>>>>> of-platdata. The driver needs to be ported to use of-platdata properly.
> >>>>>>>>>> For now, avoid a build error by returning an error.
> >>>>>>>>>>
> >>>>>>>>>> Signed-off-by: Simon Glass <sjg at chromium.org>
> >>>>>>>
> >>>>>>> Does this break the boot on omap l138?
> >>>>>>>
> >>>>>>
> >>>>>> I don't have a board at hand to test this out. Bartosz can you help test this with
> >>>>>> omapl138?
> >>>>>>
> >>>>>> Thanks,
> >>>>>> Faiz
> >>>>>
> >>>>> Hi Faiz,
> >>>>>
> >>>>> I can confirm - this *does* break the mmc boot on da850-lcdk.
> >>>>
> >>>> So who is going to fix the driver to unblock Simon's series?
> >>>>
> >>>
> >>> Is this something that will take a lot of work? What exactly needs
> >>> doing? I'm not sure what "use of-platdata properly" means.
> >>
> >> This board is defining CONFIG_SPL_OF_PLATDATA which means that device
> >> tree is not available in SPL. Instead you need to use a C structure
> >> created by dtoc. It basically involves creating that struct and
> >> getting the data from that instead of calling the DT functions. I
> >> expect it will take 2-4 hours to figure out, code and test.
> >>
> >> See of-plat.rst for full documentation. There are quite a few examples for mmc:
> >>
> >> grep PLATDATA drivers/mmc/*.c
> >> drivers/mmc/ftsdc010_mci.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/ftsdc010_mci.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/ftsdc010_mci.c:#if !CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/ftsdc010_mci.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/mxsmmc.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/mxsmmc.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/mxsmmc.c: debug("OF_PLATDATA: regs: 0x%p bw: %d clkid: %d
> >> non_removable: %d\n",
> >> drivers/mmc/mxsmmc.c:#if CONFIG_IS_ENABLED(OF_CONTROL) &&
> >> !CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/mxsmmc.c:#if CONFIG_IS_ENABLED(OF_CONTROL) &&
> >> !CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/omap_hsmmc.c:#if CONFIG_IS_ENABLED(OF_CONTROL) &&
> >> !CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/omap_hsmmc.c:#if CONFIG_IS_ENABLED(OF_CONTROL) &&
> >> !CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/omap_hsmmc.c:#if CONFIG_IS_ENABLED(OF_CONTROL) &&
> >> !CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/rockchip_dw_mmc.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/rockchip_dw_mmc.c:#if !CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/rockchip_dw_mmc.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/rockchip_sdhci.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/rockchip_sdhci.c:#if CONFIG_IS_ENABLED(OF_PLATDATA)
> >> drivers/mmc/rockchip_sdhci.c:#if !CONFIG_IS_ENABLED(OF_PLATDATA)
> >>
>
> In all the examples above, platdata reg filed is directly being used for
> to assign a register base address but looking at davinci platdata that is generated,
> spl/dts/dt-platdata.c:
>
> static const struct dtd_simple_bus dtv_soc_at_1c00000 = {
>         .model                  = "da850",
>         .ranges                 = {0x0, 0x1c00000, 0x400000},
> };
> U_BOOT_DEVICE(soc_at_1c00000) = {
>         .name           = "simple_bus",
>         .platdata       = &dtv_soc_at_1c00000,
>         .platdata_size  = sizeof(dtv_soc_at_1c00000),
> };
>
> static const struct dtd_ti_da830_uart dtv_serial_at_10d000 = {
>         .power_domains          = {0xa, 0xd},
>         .reg                    = {0x10d000, 0x100},
>         .reg_io_width           = 0x4,
>         .reg_shift              = 0x2,
> };
> U_BOOT_DEVICE(serial_at_10d000) = {
>         .name           = "ti_da830_uart",
>         .platdata       = &dtv_serial_at_10d000,
>         .platdata_size  = sizeof(dtv_serial_at_10d000),
> };
>
> static const struct dtd_ti_da830_mmc dtv_mmc_at_40000 = {
>         .bus_width              = 0x4,
>         .cap_mmc_highspeed      = true,
>         .cap_sd_highspeed       = true,
>         .cd_gpios               = {0x16, 0x40, 0x1},
>         .dma_names              = {"rx", "tx"},
>         .dmas                   = {0x14, 0x10, 0x0, 0x14, 0x11, 0x0},
>         .max_frequency          = 0x2faf080,
>         .reg                    = {0x40000, 0x1000},
> };
> U_BOOT_DEVICE(mmc_at_40000) = {
>         .name           = "ti_da830_mmc",
>         .platdata       = &dtv_mmc_at_40000,
>         .platdata_size  = sizeof(dtv_mmc_at_40000),
> };
>
> I need the base address of the MMC device (dtd_ti_da830_mmc dtv_mmc_at_40000)
> reg = 0x40000 to be translated with the simple_bus ranges above. How do I do
> this without a device tree as there is no parent-child relationship defined
> between the structures? Or at least that is my understanding.
>
> Looks like I will have to drop OF_PLATDATA and use hardcoded U_BOOT_DEVICE()
> declarations for this.

Four options I can think of:

1. Add support for translating to dtoc
2. Add support for parents to dtoc
3. Find the soc device (the one with the ranges) and write a function
to return the base address given the dtplat for that device.
4. Hard-code the address behind if CONFIG_IS_ENABLED(OF_PLATDATA)) for
your board

Regards,
Simon