Re: [GIT pull] irq updates for 4.13

[Tony Lindgren <tony@atomide.com>]

* Thomas Gleixner <tglx@linutronix.de> [170711 08:07]:
> On Tue, 11 Jul 2017, Thomas Gleixner wrote:
> > On Tue, 11 Jul 2017, Tony Lindgren wrote:
> > > * Thomas Gleixner <tglx@linutronix.de> [170711 02:48]:
> > > And "external abort on non-linefetch" means something is not clocked
> > > in this case. The following alone makes things boot for me again, but I don't
> > > quite follow what has now changed with the ordering.. Thomas, any ideas?
> > 
> > Ah. Now that makes sense.
> > 
> > Unpatched the ordering is:
> > 
> > 	  chip_bus_lock(desc);
> > 	  irq_request_resources(desc);
> > 
> > Now the offending change reordered the calls. OMAP gpio has:
> > 
> >     omap_gpio_irq_bus_lock()
> >        pm_runtime_get_sync(bank->chip.parent);
> > 
> > So that at least explains the error. So that omap gpio irq chip (ab)uses
> > the bus_lock() callback to do runtime power management. Sigh, I did not
> > expect that. Let me have a deeper look if that's OMAP only or whether this
> > happens in other places as well.

OK that explains.

> So OMAP-GPIO is the only driver which abuses bus_lock/unlock() in that way
> and gets surprised.

OK. Grygorii, care to take a look if there's a better way to deal with
runtime PM for gpio-omap?

Meanwhile, below is my fix again with a proper description so we can
have working -rc1.

Regards,

Tony

8< ------
>From tony Mon Sep 17 00:00:00 2001
From: Tony Lindgren <tony@atomide.com>
Date: Tue, 11 Jul 2017 06:40:50 -0700
Subject: [PATCH] gpio: omap: Fix external abort on non-linefetch

Commit 46e48e257360 ("genirq: Move irq resource handling out of spinlocked
region") caused external abort on non-linefetch for n900 and droid 4. Turns
out this was caused by runtime PM use in gpio-omap in chip_bus_lock:

* Thomas Gleixner <tglx@linutronix.de> [170711 08:07]:
> > Unpatched the ordering is:
> >
> >       chip_bus_lock(desc);
> >       irq_request_resources(desc);
> >
> > Now the offending change reordered the calls. OMAP gpio has:
> >
> >     omap_gpio_irq_bus_lock()
> >        pm_runtime_get_sync(bank->chip.parent);
> >
> > So that at least explains the error. So that omap gpio irq chip (ab)uses
> > the bus_lock() callback to do runtime power management. Sigh, I did not
> > expect that.

Let's fix it with pm_runtime_get_sync() for now, then further improvments
can be done when we have a better way to deal with it.

Reported-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Tony Lindgren <tony@atomide.com>
---
 drivers/gpio/gpio-omap.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/drivers/gpio/gpio-omap.c b/drivers/gpio/gpio-omap.c
--- a/drivers/gpio/gpio-omap.c
+++ b/drivers/gpio/gpio-omap.c
@@ -917,13 +917,24 @@ static int omap_gpio_get_direction(struct gpio_chip *chip, unsigned offset)
 	struct gpio_bank *bank;
 	unsigned long flags;
 	void __iomem *reg;
-	int dir;
+	int error, dir;
 
 	bank = gpiochip_get_data(chip);
 	reg = bank->base + bank->regs->direction;
+	error = pm_runtime_get_sync(bank->chip.parent);
+	if (error < 0) {
+		dev_err(bank->chip.parent,
+			"Could not enable gpio bank %p: %d\n",
+			bank, error);
+		pm_runtime_put_noidle(bank->chip.parent);
+
+		return error;
+	}
 	raw_spin_lock_irqsave(&bank->lock, flags);
 	dir = !!(readl_relaxed(reg) & BIT(offset));
 	raw_spin_unlock_irqrestore(&bank->lock, flags);
+	pm_runtime_put_sync(bank->chip.parent);
+
 	return dir;
 }
 
-- 
2.13.2