Re: [PATCH] irqchip/versatile-fpga: Handle chained IRQs properly

From: Marc Zyngier
Date: Wed Mar 18 2020 - 15:11:10 EST


On 2020-03-18 18:20, Sungbo Eo wrote:
Hi Marc,

On 2020-03-19 02:48, Marc Zyngier wrote:
Hi Sungbo,

On 2020-03-18 17:09, Sungbo Eo wrote:
Enclose the chained handler with chained_irq_{enter,exit}(), so that the
muxed interrupts get properly acked.

This patch also fixes a reboot bug on OX820 SoC, where the jiffies timer
interrupt is never acked. The kernel waits a clock tick forever in
calibrate_delay_converge(), which leads to a boot hang.

Nice catch.


Signed-off-by: Sungbo Eo <mans0n@xxxxxxxxxx>
Cc: Neil Armstrong <narmstrong@xxxxxxxxxxxx>
---
Âdrivers/irqchip/irq-versatile-fpga.c | 9 ++++++++-
Â1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/irqchip/irq-versatile-fpga.c
b/drivers/irqchip/irq-versatile-fpga.c
index 928858dada75..08faab2fec3e 100644
--- a/drivers/irqchip/irq-versatile-fpga.c
+++ b/drivers/irqchip/irq-versatile-fpga.c
@@ -6,6 +6,7 @@
Â#include <linux/irq.h>
Â#include <linux/io.h>
Â#include <linux/irqchip.h>
+#include <linux/irqchip/chained_irq.h>
Â#include <linux/irqchip/versatile-fpga.h>
Â#include <linux/irqdomain.h>
Â#include <linux/module.h>
@@ -68,12 +69,15 @@ static void fpga_irq_unmask(struct irq_data *d)

Âstatic void fpga_irq_handle(struct irq_desc *desc)
Â{
+ÂÂÂ struct irq_chip *chip = irq_desc_get_chip(desc);
ÂÂÂÂ struct fpga_irq_data *f = irq_desc_get_handler_data(desc);
ÂÂÂÂ u32 status = readl(f->base + IRQ_STATUS);

+ÂÂÂ chained_irq_enter(chip, desc);
+

It's probably not a big deal, but I'm not fond of starting talking to
the muxing irqchip before having done the chained_irq_enter() call.

Moving that read here would probably be safer.

Oops, I missed it. Thanks for pointing it out.


ÂÂÂÂ if (status == 0) {
ÂÂÂÂÂÂÂÂ do_bad_IRQ(desc);
-ÂÂÂÂÂÂÂ return;
+ÂÂÂÂÂÂÂ goto out;
ÂÂÂÂ }

ÂÂÂÂ do {
@@ -82,6 +86,9 @@ static void fpga_irq_handle(struct irq_desc *desc)
ÂÂÂÂÂÂÂÂ status &= ~(1 << irq);
ÂÂÂÂÂÂÂÂ generic_handle_irq(irq_find_mapping(f->domain, irq));
ÂÂÂÂ } while (status);
+
+out:
+ÂÂÂ chained_irq_exit(chip, desc);
Â}

Â/*

Otherwise looks good. If you send it again with the above fixed
and a Fixes: tag, I'll queue it.

It seems the handler had been broken from the very beginning. Could
you give me a hint on how the tag should be like?

Indeed, it has been broken forever. I'm tempted to say:

Fixes: c41b16f8c9d9d ("ARM: integrator/versatile: consolidate FPGA IRQ handling code")

even if it probably predates the introduction of the chained_irq_enter() helpers.
This will ensure this gets backported to older kernels...

Thanks,

M.
--
Jazz is not dead. It just smells funny...