[PATCH 04/11] dm: Remove uses of device_bind_offset()

[Simon Glass]

Hi Eugen,

On Mon, 1 Feb 2021 at 01:13, <Eugen.Hristev at microchip.com> wrote:
>
> On 31.01.2021 17:37, Simon Glass wrote:
> > Hi Eugen,
> >
> > On Sun, 31 Jan 2021 at 02:18, <Eugen.Hristev at microchip.com> wrote:
> >>
> >> On 10.12.2020 02:26, Simon Glass wrote:
> >>> This function is not needed since the standard device_bind() can be used
> >>> instead.
> >>>
> >>> Signed-off-by: Simon Glass <sjg at chromium.org>
> >>> ---
> >>>
> >>>    arch/x86/cpu/apollolake/spl.c               |  2 +-
> >>>    drivers/clk/at91/compat.c                   | 20 ++++++++------------
> >>>    drivers/clk/clk.c                           |  2 +-
> >>>    drivers/gpio/mt7621_gpio.c                  |  4 ++--
> >>>    drivers/gpio/s5p_gpio.c                     |  4 ++--
> >>>    drivers/gpio/sunxi_gpio.c                   |  4 ++--
> >>>    drivers/gpio/tegra186_gpio.c                |  4 ++--
> >>>    drivers/gpio/tegra_gpio.c                   |  6 +++---
> >>>    drivers/net/mvpp2.c                         |  4 ++--
> >>>    drivers/pinctrl/broadcom/pinctrl-bcm283x.c  |  5 ++---
> >>>    drivers/pinctrl/meson/pinctrl-meson.c       |  4 +++-
> >>>    drivers/pinctrl/mscc/pinctrl-jr2.c          |  4 ++--
> >>>    drivers/pinctrl/mscc/pinctrl-luton.c        |  4 ++--
> >>>    drivers/pinctrl/mscc/pinctrl-ocelot.c       |  4 ++--
> >>>    drivers/pinctrl/mscc/pinctrl-serval.c       |  4 ++--
> >>>    drivers/pinctrl/mscc/pinctrl-servalt.c      |  4 ++--
> >>>    drivers/pinctrl/mvebu/pinctrl-armada-37xx.c |  8 ++++----
> >>>    drivers/power/regulator/Kconfig             |  2 +-
> >>>    include/dm/device-internal.h                |  4 ++--
> >>>    include/power/regulator.h                   |  2 +-
> >>>    20 files changed, 46 insertions(+), 49 deletions(-)
> >>>
> >>> Applied to u-boot-dm, thanks!
> >>>
> >>
> >>
> >> Hi Simon,
> >>
> >> I bisected the tree and this commit looks to break
> >> sama5d4_xplained_mmc_defconfig :
> >>
> >> <debug_uart>
> >> No serial driver found
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Could not initialize timer (err -19)
> >>
> >> Booting u-boot fails when adding this commit.
> >>
> >> Could you please help or let me know how I can fix it ?
> >
> > I suspect the problem could be in the changes to
> > drivers/clk/at91/compat.c although I cannot see why
> >
> > You could try reverting that change, and just using offset_to_ofnode()
> > in the device_bind_driver_to_node() call. I actually intended to do
> > that at the time due to the risk, but somehow I missed this one.
> >
> > OTOH it would be good to move the code to livetree and stop using fdt offsets.
> >
> > Regards,
> > Simon
> >
>
> I reverted the changes in compat.c and indeed now it boots correctly.
>
> I tried to do the following change on top of your code as you suggested
> but it does not help:
>
>
> --- a/drivers/clk/at91/compat.c
> +++ b/drivers/clk/at91/compat.c
> @@ -67,7 +67,7 @@ int at91_clk_sub_device_bind(struct udevice *dev,
> const char *drv_name)
>          bool pre_reloc_only = !(gd->flags & GD_FLG_RELOC);
>          const char *name;
>          int ret;
> -
> +       int offset = dev_of_offset(dev);
>          ofnode_for_each_subnode(node, parent) {
>                  if (pre_reloc_only && !ofnode_pre_reloc(node))
>                          continue;
> @@ -84,7 +84,7 @@ int at91_clk_sub_device_bind(struct udevice *dev,
> const char *drv_name)
>                  name = ofnode_get_name(node);
>                  if (!name)
>                          return -EINVAL;
> -               ret = device_bind_driver_to_node(dev, drv_name, name, node,
> +               ret = device_bind_driver_to_node(dev, drv_name, name,
> offset_to_ofnode(offset),
>                                                   NULL);
>                  if (ret)
>                          return ret;
>
>
> I have a feeling the 'for loop' for the subnodes misses an essential
> driver and thus it fails booting

Then I think reverting all the changes is the best thing in this file.
Can you send a patch?

Ultimately this should be figured out, but I cannot see what is wrong
and don't have that hardware to try. I do have an old SAM9260/9263 but
I'm not sure if that uses the same driver.

Regards,
Simon