[PATCH v3 21/32] powerpc/64s/exception: sreset interrupts reconcile fix

[Nicholas Piggin]

This adds IRQ_HARD_DIS to irq_happened. Although it doesn't seem to
matter much because we're not allowed to enable irqs in an NMI handler,
the soft-irq debugging code is becoming more strict about ensuring
IRQ_HARD_DIS is in sync with MSR[EE], this may help avoid asserts or
other issues.

Add a comments explaining why MCE does not have this. Early machine
check is generally much smaller and more contained code which will
explode if you look at it wrong anyway as it runs in real mode, though
there's an argument that we should do similar reconciling for the MCE
as well.

Signed-off-by: Nicholas Piggin <npiggin at gmail.com>
---
 arch/powerpc/kernel/exceptions-64s.S | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index c23eb9c572b2..6ff5ea236b17 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -920,18 +920,19 @@ EXC_COMMON_BEGIN(system_reset_common)
 	__GEN_COMMON_BODY system_reset
 	bl	save_nvgprs
 	/*
-	 * Set IRQS_ALL_DISABLED unconditionally so arch_irqs_disabled does
+	 * Set IRQS_ALL_DISABLED unconditionally so irqs_disabled() does
 	 * the right thing. We do not want to reconcile because that goes
 	 * through irq tracing which we don't want in NMI.
 	 *
-	 * Save PACAIRQHAPPENED because some code will do a hard disable
-	 * (e.g., xmon). So we want to restore this back to where it was
-	 * when we return. DAR is unused in the stack, so save it there.
+	 * Save PACAIRQHAPPENED to _DAR (otherwise unused), and set HARD_DIS
+	 * as we are running with MSR[EE]=0.
 	 */
 	li	r10,IRQS_ALL_DISABLED
 	stb	r10,PACAIRQSOFTMASK(r13)
 	lbz	r10,PACAIRQHAPPENED(r13)
 	std	r10,_DAR(r1)
+	ori	r10,r10,PACA_IRQ_HARD_DIS
+	stb	r10,PACAIRQHAPPENED(r13)
 
 	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	system_reset_exception
@@ -976,6 +977,11 @@ EXC_COMMON_BEGIN(system_reset_common)
  * error detected there), determines if it was recoverable and logs the
  * event.
  *
+ * This early code does not "reconcile" irq soft-mask state like SRESET or
+ * regular interrupts do, so irqs_disabled() among other things may not work
+ * properly (irq disable/enable already doesn't work because irq tracing can
+ * not work in real mode).
+ *
  * Then, depending on the execution context when the interrupt is taken, there
  * are 3 main actions:
  * - Executing in kernel mode. The event is queued with irq_work, which means
-- 
2.23.0