Re: [PATCH v4] irqchip/ls-extirq: fix invalid wait context by avoiding to use regmap
From: Marc Zyngier
Date: Mon Oct 03 2022 - 11:28:19 EST
On Mon, 03 Oct 2022 10:45:43 +0100,
Vladimir Oltean <vladimir.oltean@xxxxxxx> wrote:
> Hi Marc, Arnd,
> On Thu, Aug 18, 2022 at 05:13:09PM +0300, Vladimir Oltean wrote:
> > Hi,
> > On Thu, Jul 28, 2022 at 05:42:54PM +0300, Vladimir Oltean wrote:
> > > The irqchip->irq_set_type method is called by __irq_set_trigger() under
> > > the desc->lock raw spinlock.
> > >
> > > The ls-extirq implementation, ls_extirq_irq_set_type(), uses an MMIO
> > > regmap created by of_syscon_register(), which uses plain spinlocks
> > > (the kind that are sleepable on RT).
> > >
> > > Therefore, this is an invalid locking scheme for which we get a kernel
> > > splat stating just that ("[ BUG: Invalid wait context ]"), because the
> > > context in which the plain spinlock may sleep is atomic due to the raw
> > > spinlock. We need to go raw spinlocks all the way.
> > >
> > > Make this driver ioremap its INTPCR register on its own, and stop
> > > relying on syscon to provide a regmap. Since the regmap we got from
> > > syscon belonged to the parent and the newly ioremapped region belongs
> > > just to us, the offset to the INTPCR register is now 0, because of the
> > > address translation that takes place through the device tree.
> > >
> > > One complication, due to the fact that this driver uses IRQCHIP_DECLARE
> > > rather than traditional platform devices with probe and remove methods,
> > > is that we cannot use devres, so we need to implement a full-blown
> > > cleanup procedure on the error path.
> > >
> > > Fixes: 0dcd9f872769 ("irqchip: Add support for Layerscape external interrupt lines")
> > > Signed-off-by: Vladimir Oltean <vladimir.oltean@xxxxxxx>
> > > ---
> > Just checking in on this patch to make sure it hasn't been forgotten.
> Is there something else I need to do such that this patch gets accepted?
No, it just went under the radar. I would have liked an ack from
Rasmus, but this has been there for long enough.
I'll queue this as a fix for 6.1.
Without deviation from the norm, progress is not possible.