[PATCH v2 09/25] arm64: entry: Map the FIQ vector to IRQ on NEEDS_FIQ platforms

[Marc Zyngier]

On Wed, 17 Feb 2021 11:49:23 +0000,
Mark Rutland <mark.rutland at arm.com> wrote:
> 
> Hi Hector,
> 
> On Mon, Feb 15, 2021 at 09:16:57PM +0900, Hector Martin wrote:
> > From: Marc Zyngier <maz at kernel.org>
> > 
> > By default, FIQ exceptions trigger a panic. On platforms that need to
> > deliver interrupts via FIQ, this gets redirected via an alternative to
> > instead handle FIQ the same way as IRQ. It is up to the irqchip handler
> > to discriminate between the two.
> > 
> > Signed-off-by: Marc Zyngier <maz at kernel.org>
> > Signed-off-by: Hector Martin <marcan at marcan.st>
> 
> Since the use of FIQ is a platform integration detail rather than a CPU
> implementation detail (and e.g. can differ across bare-metal and VM),
> I'd prefer to always have separate registered handlers for IRQ/FIQ (also
> avoiding the need for patching). That way we can explicitly opt-in to
> FIQ when required, and avoid edge-cases where an unexpected FIQ could
> livelock an unaware IRQ handler.
> 
> Marc and I had a quick play with that, and I have a series of patches
> I've pushed to:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/mark/linux.git/log/?h=arm64/fiq
> git://git.kernel.org/pub/scm/linux/kernel/git/mark/linux.git arm64/fiq
> 
> ... which I'll post out shortly.

FWIW, I've just posted a more complete version of the first patch in
this series[1], which you may want to use directly (though I plan to
take it as a fix for 5.12).

Thanks,

	M.

[1] https://lore.kernel.org/r/20210217142800.2547737-1-maz@kernel.org

-- 
Without deviation from the norm, progress is not possible.