[PATCH] pinctrl: devicetree: Keep deferring even on timeout

[PATCH] pinctrl: devicetree: Keep deferring even on timeout

[Thierry Reding]

From: Thierry Reding <treding@nvidia.com>

driver_deferred_probe_check_state() may return -ETIMEDOUT instead of
-EPROBE_DEFER after all built-in drivers have been probed. This can
cause issues for built-in drivers that depend on resources provided by
loadable modules.

One such case happens on Tegra where I2C controllers are used during
early boot to set up the system PMIC, so the I2C driver needs to be a
built-in driver. At the same time, some instances of the I2C controller
depend on the DPAUX hardware for pinmuxing. Since the DPAUX is handled
by the display driver, which is usually not built-in, the pin control
states will not become available until after the root filesystem has
been mounted and the display driver loaded from it.

Fixes: bec6c0ecb243 ("pinctrl: Remove use of driver_deferred_probe_check_state_continue()")
Suggested-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
---
 drivers/pinctrl/devicetree.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/pinctrl/devicetree.c b/drivers/pinctrl/devicetree.c
index 5eff8c296552..3fb238714718 100644
--- a/drivers/pinctrl/devicetree.c
+++ b/drivers/pinctrl/devicetree.c
@@ -130,9 +130,8 @@ static int dt_to_map_one_config(struct pinctrl *p,
 		if (!np_pctldev || of_node_is_root(np_pctldev)) {
 			of_node_put(np_pctldev);
 			ret = driver_deferred_probe_check_state(p->dev);
-			/* keep deferring if modules are enabled unless we've timed out */
-			if (IS_ENABLED(CONFIG_MODULES) && !allow_default &&
-			    (ret == -ENODEV))
+			/* keep deferring if modules are enabled */
+			if (IS_ENABLED(CONFIG_MODULES) && !allow_default && ret < 0)
 				ret = -EPROBE_DEFER;
 			return ret;
 		}
-- 
2.28.0

Re: [PATCH] pinctrl: devicetree: Keep deferring even on timeout

[Linus Walleij]

On Tue, Aug 25, 2020 at 4:33 PM Thierry Reding <thierry.reding@gmail.com> wrote:

> From: Thierry Reding <treding@nvidia.com>
>
> driver_deferred_probe_check_state() may return -ETIMEDOUT instead of
> -EPROBE_DEFER after all built-in drivers have been probed. This can
> cause issues for built-in drivers that depend on resources provided by
> loadable modules.
>
> One such case happens on Tegra where I2C controllers are used during
> early boot to set up the system PMIC, so the I2C driver needs to be a
> built-in driver. At the same time, some instances of the I2C controller
> depend on the DPAUX hardware for pinmuxing. Since the DPAUX is handled
> by the display driver, which is usually not built-in, the pin control
> states will not become available until after the root filesystem has
> been mounted and the display driver loaded from it.
>
> Fixes: bec6c0ecb243 ("pinctrl: Remove use of driver_deferred_probe_check_state_continue()")
> Suggested-by: John Stultz <john.stultz@linaro.org>
> Signed-off-by: Thierry Reding <treding@nvidia.com>

Just like with the identical patch from John I got a while back, I'd like
some nod from the device core maintainers (Greg or Raphael) that this
is how we want to do things.

Yours,
Linus Walleij

Re: [PATCH] pinctrl: devicetree: Keep deferring even on timeout

[Linus Walleij]

On Tue, Aug 25, 2020 at 4:33 PM Thierry Reding <thierry.reding@gmail.com> wrote:

> From: Thierry Reding <treding@nvidia.com>
>
> driver_deferred_probe_check_state() may return -ETIMEDOUT instead of
> -EPROBE_DEFER after all built-in drivers have been probed. This can
> cause issues for built-in drivers that depend on resources provided by
> loadable modules.
>
> One such case happens on Tegra where I2C controllers are used during
> early boot to set up the system PMIC, so the I2C driver needs to be a
> built-in driver. At the same time, some instances of the I2C controller
> depend on the DPAUX hardware for pinmuxing. Since the DPAUX is handled
> by the display driver, which is usually not built-in, the pin control
> states will not become available until after the root filesystem has
> been mounted and the display driver loaded from it.
>
> Fixes: bec6c0ecb243 ("pinctrl: Remove use of driver_deferred_probe_check_state_continue()")
> Suggested-by: John Stultz <john.stultz@linaro.org>
> Signed-off-by: Thierry Reding <treding@nvidia.com>

I applied this patch (rather than Johns') since John agreed this
is a better solution.

I got Greg's ACK on Johns patch but not this one, but anyways
it shows he's convenient with the concept so here we go.

Yours,
Linus Walleij