mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH v2 0/2] MIPS: SGI-IP30: power off through the RTC
@ 2026-11-02  9:41 Imre Kaloz
  2026-11-02  9:41 ` [PATCH v2 1/2] MIPS: SGI-IP30: power off through the DS1687 RTC Imre Kaloz
  2026-11-02  9:41 ` [PATCH v2 2/2] MIPS: SGI-IP30: shut down cleanly on a power button press Imre Kaloz
  0 siblings, 2 replies; 4+ messages in thread
From: Imre Kaloz @ 2026-11-02  9:41 UTC (permalink / raw)
  To: Thomas Bogendoerfer; +Cc: linux-mips, linux-kernel, Lee Jones, mfd

The Octane's power supply is switched by its DS1687 RTC. Patch 1 uses
it for power-off and halt. Patch 2 turns a power button press into an
orderly power-off, and adds the hook to the IOC3 RTC platform data in
drivers/mfd/ioc3.c.

Both take effect only with CONFIG_RTC_DRV_DS1685_FAMILY=y, which
ip30_defconfig sets with "MIPS: SGI-IP30: fix the defconfig defaults
for the Bridge PROM and RTC" from the "MIPS: SGI fixes" series. They
build without it. Patch 2 drives the RTC with interrupts disabled, so
at run time it needs "rtc: ds1685: do not sleep in the power-off
handler" from the rtc-ds1685 v2 series sent to the RTC tree.

v1 was tested on an Octane with poweroff, one and two power button
presses, and a stalled shutdown.

Changes in v2:
- Patch 1: drop the device reference when the RTC has no driver data.
- Patch 2: only the first power-off drives the RTC and later callers
  wait for the power to drop; the 120 second timeout holds the device
  lock against an unbind of rtc-ds1685. The RTC's power-off message
  is printed from the power-off context, as the console thread cannot
  run once interrupts are off.

v1: https://lore.kernel.org/all/20261013115820.13293-1-kaloz@kernel.org/

Imre Kaloz (2):
  MIPS: SGI-IP30: power off through the DS1687 RTC
  MIPS: SGI-IP30: shut down cleanly on a power button press

 arch/mips/include/asm/sgi/heart.h |  2 +
 arch/mips/sgi-ip30/ip30-power.c   | 94 +++++++++++++++++++++++++++++++
 drivers/mfd/ioc3.c                |  6 ++
 3 files changed, 102 insertions(+)


base-commit: 24a19b953e2a3d0ea5a8c4951a0ab8f32720c907
-- 
2.47.3


^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH v2 1/2] MIPS: SGI-IP30: power off through the DS1687 RTC
  2026-11-02  9:41 [PATCH v2 0/2] MIPS: SGI-IP30: power off through the RTC Imre Kaloz
@ 2026-11-02  9:41 ` Imre Kaloz
  2026-11-02  9:41 ` [PATCH v2 2/2] MIPS: SGI-IP30: shut down cleanly on a power button press Imre Kaloz
  1 sibling, 0 replies; 4+ messages in thread
From: Imre Kaloz @ 2026-11-02  9:41 UTC (permalink / raw)
  To: Thomas Bogendoerfer; +Cc: linux-mips, linux-kernel, Lee Jones

The Octane's power supply is switched by its DS1687 RTC. Use
ds1685_rtc_poweroff() for power-off and halt, with all HEART interrupts
masked first, instead of leaving the machine halted.

Signed-off-by: Imre Kaloz <kaloz@kernel.org>
---
 arch/mips/sgi-ip30/ip30-power.c | 34 +++++++++++++++++++++++++++++++++
 1 file changed, 34 insertions(+)

diff --git a/arch/mips/sgi-ip30/ip30-power.c b/arch/mips/sgi-ip30/ip30-power.c
index 66851e17c5a7..b1b33c9cfd76 100644
--- a/arch/mips/sgi-ip30/ip30-power.c
+++ b/arch/mips/sgi-ip30/ip30-power.c
@@ -15,6 +15,7 @@
 #include <linux/delay.h>
 #include <linux/rtc/ds1685.h>
 #include <linux/interrupt.h>
+#include <linux/platform_device.h>
 #include <linux/pm.h>
 
 #include <asm/reboot.h>
@@ -31,10 +32,43 @@ static void __noreturn ip30_machine_restart(char *cmd)
 	unreachable();
 }
 
+static void ip30_machine_power_off(void)
+{
+	struct device *dev;
+	int cpu;
+
+	if (!IS_REACHABLE(CONFIG_RTC_DRV_DS1685_FAMILY))
+		return;
+
+	dev = bus_find_device_by_name(&platform_bus_type, NULL, "rtc-ds1685");
+	if (!dev)
+		return;
+	if (!dev_get_drvdata(dev)) {
+		put_device(dev);
+		return;
+	}
+
+	/* Devices losing power may still interrupt; keep their handlers out. */
+	for (cpu = 0; cpu < HEART_MAX_CPUS; cpu++)
+		heart_write(HEART_CLR_ALL_MASK, &heart_regs->imr[cpu]);
+	heart_write(HEART_ACK_ALL_MASK, &heart_regs->clear_isr);
+
+	ds1685_rtc_poweroff(to_platform_device(dev));
+}
+
 static int __init ip30_reboot_setup(void)
 {
 	_machine_restart = ip30_machine_restart;
 
+	/*
+	 * The DS1687 RTC switches the power supply off.  Halt powers off
+	 * too, as the power button needs a running kernel.
+	 */
+	if (IS_REACHABLE(CONFIG_RTC_DRV_DS1685_FAMILY)) {
+		_machine_halt = ip30_machine_power_off;
+		pm_power_off = ip30_machine_power_off;
+	}
+
 	return 0;
 }
 
-- 
2.47.3


^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH v2 2/2] MIPS: SGI-IP30: shut down cleanly on a power button press
  2026-11-02  9:41 [PATCH v2 0/2] MIPS: SGI-IP30: power off through the RTC Imre Kaloz
  2026-11-02  9:41 ` [PATCH v2 1/2] MIPS: SGI-IP30: power off through the DS1687 RTC Imre Kaloz
@ 2026-11-02  9:41 ` Imre Kaloz
       [not found]   ` <sashiko-outbox-193229@kernel.org>
  1 sibling, 1 reply; 4+ messages in thread
From: Imre Kaloz @ 2026-11-02  9:41 UTC (permalink / raw)
  To: Thomas Bogendoerfer; +Cc: linux-mips, linux-kernel, Lee Jones, Sashiko, mfd

The power button raises a kickstart interrupt on the DS1687, and
rtc-ds1685 cuts power at once unless the platform supplies a
prepare_poweroff hook. Supply one that starts an orderly power-off,
cutting power directly on a second press or after 120 seconds.

Power-off drives the RTC's index/data register pair, so only the first
of several concurrent callers proceeds and the rest wait. The timeout
holds the device lock against an unbind of rtc-ds1685.

Reported-by: Sashiko <sashiko-bot@kernel.org>
Closes: https://lore.kernel.org/all/sashiko-outbox-169717@kernel.org/
Signed-off-by: Imre Kaloz <kaloz@kernel.org>
---
 arch/mips/include/asm/sgi/heart.h |  2 ++
 arch/mips/sgi-ip30/ip30-power.c   | 60 +++++++++++++++++++++++++++++++
 drivers/mfd/ioc3.c                |  6 ++++
 3 files changed, 68 insertions(+)

diff --git a/arch/mips/include/asm/sgi/heart.h b/arch/mips/include/asm/sgi/heart.h
index c224c2e3575a..736cabb7fdbb 100644
--- a/arch/mips/include/asm/sgi/heart.h
+++ b/arch/mips/include/asm/sgi/heart.h
@@ -317,6 +317,8 @@ struct ip30_heart_regs {		/* 0x0ff00000 */
 
 extern struct ip30_heart_regs __iomem *heart_regs;
 
+void ip30_prepare_poweroff(void);
+
 #define heart_read	____raw_readq
 #define heart_write	____raw_writeq
 
diff --git a/arch/mips/sgi-ip30/ip30-power.c b/arch/mips/sgi-ip30/ip30-power.c
index b1b33c9cfd76..cff8f4875429 100644
--- a/arch/mips/sgi-ip30/ip30-power.c
+++ b/arch/mips/sgi-ip30/ip30-power.c
@@ -7,6 +7,7 @@
  *               2009 Johannes Dickgreber <tanzy@gmx.de>
  */
 
+#include <linux/console.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -17,6 +18,8 @@
 #include <linux/interrupt.h>
 #include <linux/platform_device.h>
 #include <linux/pm.h>
+#include <linux/reboot.h>
+#include <linux/workqueue.h>
 
 #include <asm/reboot.h>
 #include <asm/sgi/heart.h>
@@ -32,6 +35,8 @@ static void __noreturn ip30_machine_restart(char *cmd)
 	unreachable();
 }
 
+static atomic_t ip30_power_off_claimed = ATOMIC_INIT(0);
+
 static void ip30_machine_power_off(void)
 {
 	struct device *dev;
@@ -48,14 +53,69 @@ static void ip30_machine_power_off(void)
 		return;
 	}
 
+	/*
+	 * The RTC registers are reached through an index/data pair, so one
+	 * CPU at a time may drive them, and a second button press, the
+	 * timeout and pm_power_off can arrive together.  Interrupts go off
+	 * first so the claimant cannot be preempted by a waiter on its CPU;
+	 * the waiters spin until the power drops.
+	 */
+	local_irq_disable();
+	if (atomic_xchg(&ip30_power_off_claimed, 1))
+		for (;;)
+			cpu_relax();
+
 	/* Devices losing power may still interrupt; keep their handlers out. */
 	for (cpu = 0; cpu < HEART_MAX_CPUS; cpu++)
 		heart_write(HEART_CLR_ALL_MASK, &heart_regs->imr[cpu]);
 	heart_write(HEART_ACK_ALL_MASK, &heart_regs->clear_isr);
 
+	/*
+	 * Nothing schedules before the power drops, so the RTC's message must
+	 * be printed from this context, not left to the console thread.  The
+	 * emergency section is never left.
+	 */
+	nbcon_cpu_emergency_enter();
 	ds1685_rtc_poweroff(to_platform_device(dev));
 }
 
+#define IP30_POWERDOWN_TIMEOUT	120	/* seconds */
+
+static bool ip30_shutting_down;
+
+static void ip30_power_timeout(struct work_struct *work)
+{
+	struct device *dev;
+
+	dev = bus_find_device_by_name(&platform_bus_type, NULL, "rtc-ds1685");
+	if (!dev)
+		return;
+
+	/*
+	 * Unbinding rtc-ds1685 frees the RTC under the device lock, so hold
+	 * it across the drvdata check and the power-off.
+	 */
+	device_lock(dev);
+	ip30_machine_power_off();
+	device_unlock(dev);
+	put_device(dev);
+}
+
+static DECLARE_DELAYED_WORK(ip30_power_work, ip30_power_timeout);
+
+/* Called by the RTC driver when the power button is pressed. */
+void ip30_prepare_poweroff(void)
+{
+	/* A second press powers off at once. */
+	if (ip30_shutting_down)
+		ip30_machine_power_off();
+
+	/* Shut down through userspace, powering off anyway if it stalls. */
+	ip30_shutting_down = true;
+	orderly_poweroff(true);
+	schedule_delayed_work(&ip30_power_work, IP30_POWERDOWN_TIMEOUT * HZ);
+}
+
 static int __init ip30_reboot_setup(void)
 {
 	_machine_restart = ip30_machine_restart;
diff --git a/drivers/mfd/ioc3.c b/drivers/mfd/ioc3.c
index 5f8ac364b610..8f603651cc8e 100644
--- a/drivers/mfd/ioc3.c
+++ b/drivers/mfd/ioc3.c
@@ -24,6 +24,9 @@
 
 #include <asm/pci/bridge.h>
 #include <asm/sn/ioc3.h>
+#ifdef CONFIG_SGI_IP30
+#include <asm/sgi/heart.h>
+#endif
 
 #define IOC3_IRQ_SERIAL_A	6
 #define IOC3_IRQ_SERIAL_B	15
@@ -321,6 +324,9 @@ static struct ds1685_rtc_platform_data ip30_rtc_platform_data = {
 	.no_irq = false,
 	.uie_unsupported = true,
 	.access_type = ds1685_reg_indirect,
+#ifdef CONFIG_SGI_IP30
+	.plat_prepare_poweroff = ip30_prepare_poweroff,
+#endif
 };
 
 static const struct resource ioc3_rtc_ds1685_resources[] = {
-- 
2.47.3


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH v2 2/2] MIPS: SGI-IP30: shut down cleanly on a power button press
       [not found]   ` <sashiko-outbox-193229@kernel.org>
@ 2026-11-02 11:29     ` Imre Kaloz
  0 siblings, 0 replies; 4+ messages in thread
From: Imre Kaloz @ 2026-11-02 11:29 UTC (permalink / raw)
  To: sashiko-bot; +Cc: Thomas Bogendoerfer, linux-mips, linux-kernel, Lee Jones, mfd

>> +	local_irq_disable();
>> +	if (atomic_xchg(&ip30_power_off_claimed, 1))
>> +		for (;;)
>> +			cpu_relax();
>
> As reported by Sashiko on v1 [1], can another CPU still race on the
> indirect RTC index/data registers and bank selection while
> ds1685_rtc_poweroff() is running?

On dual-CPU Octanes the second CPU can still move the index or the
bank. v3 will stop the other CPU with smp_send_stop() before touching
HEART or the RTC.

>>  	ds1685_rtc_poweroff(to_platform_device(dev));
>
> Doesn't ds1685_rtc_poweroff() sleep in msleep(20) before cutting power,
> violating the atomic context established by local_irq_disable() and
> nbcon_cpu_emergency_enter()?

Not with the dependency named in the cover letter. Commit 21a829af60a7
("rtc: ds1685: do not sleep in the power-off handler"), in v7.4-rc1,
replaces that msleep() with mdelay().

>> +#ifdef CONFIG_SGI_IP30
>> +	.plat_prepare_poweroff = ip30_prepare_poweroff,
>> +#endif
>
> What happens on a power button press when CONFIG_SGI_IP30=y and
> CONFIG_RTC_DRV_DS1685_FAMILY=m?

The button stops cutting the power, as the built-in
ip30_machine_power_off() cannot reach the module. v3 will set the hook
only when rtc-ds1685 is built in, so a modular driver will power off
on a press as it did before this series, and the second-press branch
will return.


Thanks,
Imre

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2026-11-02 11:30 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-02  9:41 [PATCH v2 0/2] MIPS: SGI-IP30: power off through the RTC Imre Kaloz
2026-11-02  9:41 ` [PATCH v2 1/2] MIPS: SGI-IP30: power off through the DS1687 RTC Imre Kaloz
2026-11-02  9:41 ` [PATCH v2 2/2] MIPS: SGI-IP30: shut down cleanly on a power button press Imre Kaloz
     [not found]   ` <sashiko-outbox-193229@kernel.org>
2026-11-02 11:29     ` Imre Kaloz

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®