[PATCH 1/3] ARM: at91/rtc: fix boot after RTC wake-up

[Johan Hovold]

Fix failed boot after RTC wake-up by making sure RTC-interrupts are
disabled at shutdown.

The RTC-registers are not reset at shutdown when there is backup power
available, and hence the RTC-interrupt, which uses the shared system
interrupt, may be enabled during early boot. If an RTC-alarm goes off
before the RTC-driver is initialised (e.g. when waking up from an
RTC-alarm) this results in disabled system interrupts.

Without earlyprintk boot hangs after

  Uncompressing Linux... done, booting the kernel.

and earlyprintk reveals that the shared system interrupt gets disabled:

  irq 17: nobody cared (try booting with the "irqpoll" option)
  [<c000cf8c>] (unwind_backtrace+0x0/0xf0) from [<c00513ec>] (__report_bad_irq+0x20/0xe8)
  [<c00513ec>] (__report_bad_irq+0x20/0xe8) from [<c00516fc>] (note_interrupt+0x1e0/0x244)
  [<c00516fc>] (note_interrupt+0x1e0/0x244) from [<c004f980>] (handle_irq_event_percpu+0xac/0x1c4)
  [<c004f980>] (handle_irq_event_percpu+0xac/0x1c4) from [<c004faec>] (handle_irq_event+0x54/0x84)
  [<c004faec>] (handle_irq_event+0x54/0x84) from [<c0052364>] (handle_fasteoi_irq+0x84/0x154)
  [<c0052364>] (handle_fasteoi_irq+0x84/0x154) from [<c004f1f8>] (generic_handle_irq+0x20/0x30)
  [<c004f1f8>] (generic_handle_irq+0x20/0x30) from [<c0009b8c>] (handle_IRQ+0x30/0x84)
  [<c0009b8c>] (handle_IRQ+0x30/0x84) from [<c0008fa0>] (__irq_svc+0x40/0x6c)
  [<c0008fa0>] (__irq_svc+0x40/0x6c) from [<c001cd6c>] (__do_softirq+0x5c/0x164)
  [<c001cd6c>] (__do_softirq+0x5c/0x164) from [<c001d268>] (irq_exit+0x58/0x60)
  [<c001d268>] (irq_exit+0x58/0x60) from [<c0009b90>] (handle_IRQ+0x34/0x84)
  [<c0009b90>] (handle_IRQ+0x34/0x84) from [<c0008fa0>] (__irq_svc+0x40/0x6c)
  [<c0008fa0>] (__irq_svc+0x40/0x6c) from [<c03e668c>] (start_kernel+0x1d8/0x29c)
  [<c03e668c>] (start_kernel+0x1d8/0x29c) from [<70008040>] (0x70008040)
  handlers:
  [<c0012df0>] at91sam926x_pit_interrupt
  Disabling IRQ #17

Note that disabling the RTC-interrupts at shutdown fixes boot after a
clean shutdown, but that an RTC-alarm after a user (or watchdog) reset
could still prevent the system from booting.

Note also that the only way to recover from a failed boot after an
RTC-alarm is to remove any backup power (e.g. backup battery) or to
disable (or acknowledge) the RTC-interrupt manually from the bootloader.
In particular, a user-reset is not sufficient.

Tested on at91sam9g45.

Cc: stable <stable at vger.kernel.org>
Signed-off-by: Johan Hovold <jhovold at gmail.com>
---
 drivers/rtc/rtc-at91rm9200.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/rtc/rtc-at91rm9200.c b/drivers/rtc/rtc-at91rm9200.c
index b6469e2..936c75a 100644
--- a/drivers/rtc/rtc-at91rm9200.c
+++ b/drivers/rtc/rtc-at91rm9200.c
@@ -338,6 +338,14 @@ static int __exit at91_rtc_remove(struct platform_device *pdev)
 	return 0;
 }
 
+static void at91_rtc_shutdown(struct platform_device *pdev)
+{
+	/* Disable all interrupts */
+	at91_rtc_write(AT91_RTC_IDR, AT91_RTC_ACKUPD | AT91_RTC_ALARM |
+					AT91_RTC_SECEV | AT91_RTC_TIMEV |
+					AT91_RTC_CALEV);
+}
+
 #ifdef CONFIG_PM
 
 /* AT91RM9200 RTC Power management control */
@@ -384,6 +392,7 @@ static const struct dev_pm_ops at91_rtc_pm = {
 
 static struct platform_driver at91_rtc_driver = {
 	.remove		= __exit_p(at91_rtc_remove),
+	.shutdown	= at91_rtc_shutdown,
 	.driver		= {
 		.name	= "at91_rtc",
 		.owner	= THIS_MODULE,
-- 
1.8.1.1