[PATCH] iommu/arm-smmu: use a threaded handler for context interrupts

[will.deacon@arm.com (Will Deacon)]

On Fri, Jan 23, 2015 at 10:33:20PM +0000, Mitchel Humpherys wrote:
> On Fri, Jan 23 2015 at 03:24:15 AM, Will Deacon <will.deacon@arm.com> wrote:
> > On Thu, Jan 22, 2015 at 11:48:02PM +0000, Mitchel Humpherys wrote:
> >> Context interrupts can call domain-specific handlers which might sleep.
> >> Currently we register our handler with request_irq, so our handler is
> >> called in atomic context, so domain handlers that sleep result in an
> >> invalid context BUG.  Fix this by using request_threaded_irq.
> >> 
> >> This also prepares the way for doing things like enabling clocks within
> >> our interrupt handler.
> >> 
> >> Signed-off-by: Mitchel Humpherys <mitchelh@codeaurora.org>
> >> ---
> >>  drivers/iommu/arm-smmu.c | 5 +++--
> >>  1 file changed, 3 insertions(+), 2 deletions(-)
> >> 
> >> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
> >> index 6cd47b75286f..81f6b54d94b1 100644
> >> --- a/drivers/iommu/arm-smmu.c
> >> +++ b/drivers/iommu/arm-smmu.c
> >> @@ -973,8 +973,9 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain,
> >>  	spin_unlock_irqrestore(&smmu_domain->lock, flags);
> >>  
> >>  	irq = smmu->irqs[smmu->num_global_irqs + cfg->irptndx];
> >> -	ret = request_irq(irq, arm_smmu_context_fault, IRQF_SHARED,
> >> -			  "arm-smmu-context-fault", domain);
> >> +	ret = request_threaded_irq(irq, NULL, arm_smmu_context_fault,
> >> +				IRQF_ONESHOT | IRQF_SHARED,
> >> +				"arm-smmu-context-fault", domain);
> >>  	if (IS_ERR_VALUE(ret)) {
> >>  		dev_err(smmu->dev, "failed to request context IRQ %d (%u)\n",
> >>  			cfg->irptndx, irq);
> >
> > I think I'd rather keep a simple atomic handler, then have a threaded
> > handler for actually issuing the report_iommu_fault. i.e. we only wake
> > the thread when it looks like there's some work to do. That also works
> > much better for shared interrupts.
> 
> Are you still against adding clock support to the driver?  If not, we'll
> need to move to a threaded handler when clocks come in anyways...
> 
> Can you elaborate what you mean regarding shared interrupts?  Even
> without clocks it seems like the code clarity / performance tradeoff
> would favor a threaded handler, given that performance isn't important
> here.

With a shared handler (e.g. a bunch of context banks have the same IRQ)
then I assume that we don't want to end up with multiple handler threads
all tripping over each other. I'd rather have one thread that handles work
queued up by multiple low-level handlers.

Do you have a preference either way?

Will