Re: [RFC PATCH] perf/smmuv3: Fix shared interrupt handling

[Robin Murphy]

On 2020-04-09 8:02 am, John Garry wrote:
> On 08/04/2020 17:49, Robin Murphy wrote:
> > IRQF_SHARED is dangerous, since it allows other agents to retarget the
> > IRQ's affinity without migrating PMU contexts to match, breaking the way
> > in which perf manages mutual exclusion for accessing events. Although
> > this means it's not realistically possible to support PMU IRQs being
> > shared with other drivers, we *can* handle sharing between multiple PMU
> > instances with some explicit affinity bookkeeping and manual interrupt
> > multiplexing.
>
> Hi Robin,
>
> Out of curiosity, do we even need to support shared interrupts for any 
> implementations today?

Not that I know of, but we need the mitigation in general for future 
drivers[1], and since this one already had a suspicious IRQF_SHARED it 
was the ideal victim for prototyping. I haven't dared ask about Ampere's 
SMMU story... :)

> D06 board:
>
> john@ubuntu:~$ more /proc/interrupts | grep smmuv3-pmu
>
>  Â989: 0 0 0 0 ITS-pMSI 133120 Edge smmuv3-pmu
>  Â990: 0 0 0 0 ITS-pMSI 135168 Edge smmuv3-pmu
>  Â991: 0 0 0 0 ITS-pMSI 137216 Edge smmuv3-pmu
>  Â992: 0 0 0 0 ITS-pMSI 139264 Edge smmuv3-pmu
>  Â993: 0 0 0 0 ITS-pMSI 141312 Edge smmuv3-pmu
>  Â994: 0 0 0 0 ITS-pMSI 143360 Edge smmuv3-pmu
>  Â995: 0 0 0 0 ITS-pMSI 145408 Edge smmuv3-pmu
>  Â996: 0 0 0 0 ITS-pMSI 147456 Edge smmuv3-pmu

Yeah, MSIs are the best way to defeat any interrupt wiring!

Robin.

[1] 
https://lore.kernel.org/linux-arm-kernel/3efa118a-5c85-6af9-e676-44087f1d398e@xxxxxxx/