On Fri, 28 Oct 2016, Eric Anholt wrote: > Thomas Gleixner <t...@linutronix.de> writes: > > > On Thu, 27 Oct 2016, Eric Anholt wrote: > > > >> From: Phil Elwell <p...@raspberrypi.org> > >> > >> The old arch-specific IRQ macros included a dsb to ensure the
Which old arch-specific macros? > >> write to clear the mailbox interrupt completed before returning > >> from the interrupt. That's not what the patch does. It makes sure that the interrupt it cleared _before_ the IPI handler is called. > >> The BCM2836 irqchip driver needs the same precaution to avoid > >> spurious interrupts. Please describe issues proper. Something like this: The interrupt demultiplexer code clears a pending mailbox interrupt by writing the pending bit back to the hardware, but it does not ensure that the write is completed before proceeding. This causes the machine to catch fire and scares my cat. <<-- Replace by a proper description. The write() macro which was used, when the driver was developed, contained the necessary data sycnhronization barrier, but it was replaced by writel() when the driver got merged without adding the explicit barrier after the write. Add and document the barrier. Can you spot the difference? > writel(1 << ipi, mailbox0); /* Barriers need to be documented with a comment */ > + dsb(sy); > handle_IPI(ipi, regs); > > This is missing a fixes tag. I have no idea when that problem was > > introduced, so I have no way to decide whether this needs to be tagged > > stable or not. > > This code has been there since introduction of the driver, so: > > Fixes: 1a15aaa998dc ("irqchip: Add bcm2836 interrupt controller for Raspberry > Pi 2") So it want's a stable tag, right? Thanks, tglx