Merge tag 'irqchip-fixes-5.13-2' of git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms into irq/urgent

Pull irqchip fixes from Marc Zyngier:

- Fix GICv3 NMI handling where an IRQ could be mistakenly handled
  as a NMI, with disatrous effects

Link: https://lore.kernel.org/r/[email protected]

Author: KAGA-KOKO

drivers/irqchip/irq-gic-v3.c

@@ -642,11 +642,45 @@ static inline void gic_handle_nmi(u32 irqnr, struct pt_regs *regs)
 		nmi_exit();
 }
 
+static u32 do_read_iar(struct pt_regs *regs)
+{
+	u32 iar;
+
+	if (gic_supports_nmi() && unlikely(!interrupts_enabled(regs))) {
+		u64 pmr;
+
+		/*
+		 * We were in a context with IRQs disabled. However, the
+		 * entry code has set PMR to a value that allows any
+		 * interrupt to be acknowledged, and not just NMIs. This can
+		 * lead to surprising effects if the NMI has been retired in
+		 * the meantime, and that there is an IRQ pending. The IRQ
+		 * would then be taken in NMI context, something that nobody
+		 * wants to debug twice.
+		 *
+		 * Until we sort this, drop PMR again to a level that will
+		 * actually only allow NMIs before reading IAR, and then
+		 * restore it to what it was.
+		 */
+		pmr = gic_read_pmr();
+		gic_pmr_mask_irqs();
+		isb();
+
+		iar = gic_read_iar();
+
+		gic_write_pmr(pmr);
+	} else {
+		iar = gic_read_iar();
+	}
+
+	return iar;
+}
+
 static asmlinkage void __exception_irq_entry gic_handle_irq(struct pt_regs *regs)
 {
 	u32 irqnr;
 
-	irqnr = gic_read_iar();
+	irqnr = do_read_iar(regs);
 
 	/* Check for special IDs first */
 	if ((irqnr >= 1020 && irqnr <= 1023))