Skip to contentSkip to conversation

rtc: cmos: Stop using shared IRQ

PATCH Applied Andy Shevchenko 2 participants 6 messages Jan 23, 2020 – Jan 25, 2020 Real time clock (RTC)

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/3 rtc: cmos: Stop using shared IRQ

As reported by Guilherme G. Piccoli:

---8<---8<---8<---

The rtc-cmos interrupt setting was changed in the commit 079062b28fb4 ("rtc: cmos: prevent kernel warning on IRQ flags mismatch") in order to allow shared interrupts; according to that commit's description, some machine got kernel warnings due to the interrupt line being shared between rtc-cmos and other hardware, and rtc-cmos didn't allow IRQ sharing that time.

After the aforementioned commit though it was observed a huge increase in lost HPET interrupts in some systems, observed through the following kernel message:

[...] hpet1: lost 35 rtc interrupts

After investigation, it was narrowed down to the shared interrupts usage when having the kernel option "irqpoll" enabled. In this case, all IRQ handlers are called for non-timer interrupts, if such handlers are setup in shared IRQ lines. The rtc-cmos IRQ handler could be set to hpet_rtc_interrupt(), which will produce the kernel "lost interrupts" message after doing work - lots of readl/writel to HPET registers, which are known to be slow.

Although "irqpoll" is not a default kernel option, it's used in some contexts, one being the kdump kernel (which is an already "impaired" kernel usually running with 1 CPU available), so the performance burden could be considerable. Also, the same issue would happen (in a shorter extent though) when using "irqfixup" kernel option.

In a quick experiment, a virtual machine with uptime of 2 minutes produced

Quoted 1 line

300 calls to hpet_rtc_interrupt() when "irqpoll" was set, whereas without

sharing interrupts this number reduced to 1 interrupt. Machines with more hardware than a VM should generate even more unnecessary HPET interrupts in this scenario.

---8<---8<---8<---

After looking into the rtc-cmos driver history and DSDT table from the Microsoft Surface 3, we may notice that Hans de Goede submitted a correct fix (see dependency below). Thus, we simply revert the culprit commit.

Fixes 079062b28fb4 ("rtc: cmos: prevent kernel warning on IRQ flags mismatch")

Depends-on: a1e23a42f1bd ("rtc: cmos: Do not assume irq 8 for rtc when there are no legacy irqs")

---
v2: Added tags, rework scissors to avoid cutting commit message (Guilherme)

 drivers/rtc/rtc-cmos.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/rtc/rtc-cmos.c +1 -1
index 033303708c8b..cb28bbdc9e17 100644
@@ -850,7 +850,7 @@ cmos_do_probe(struct device *dev, struct resource *ports, int rtc_irq)
 			rtc_cmos_int_handler = cmos_interrupt;
 
 		retval = request_irq(rtc_irq, rtc_cmos_int_handler,
-				IRQF_SHARED, dev_name(&cmos_rtc.rtc->dev),
+				0, dev_name(&cmos_rtc.rtc->dev),
 				cmos_rtc.rtc);
 		if (retval < 0) {
 			dev_dbg(dev, "IRQ %d is already in use\n", rtc_irq);
Signature · 1 line
2.24.1
#1 lore raw

PATCH v2 2/3 rtc: cmos: Use predefined value for RTC IRQ on legacy x86

When legacy devices are present on x86 machine, the RTC IRQ has a dedicated pre-defined value. Use it instead of hard coded number.

---
v2: Added tags

 drivers/rtc/rtc-cmos.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/rtc/rtc-cmos.c +1 -1
index cb28bbdc9e17..aee55b658f85 100644
@@ -1305,7 +1305,7 @@ static int cmos_pnp_probe(struct pnp_dev *pnp, const struct pnp_device_id *id)
 		 * hardcode it on systems with a legacy PIC.
 		 */
 		if (nr_legacy_irqs())
-			irq = 8;
+			irq = RTC_IRQ;
 #endif
 		return cmos_do_probe(&pnp->dev,
 				pnp_get_resource(pnp, IORESOURCE_IO, 0), irq);
Signature · 1 line
2.24.1
maintainer Applied
#2 lore raw

Re: [PATCH v2 2/3] rtc: cmos: Use predefined value for RTC IRQ on legacy x86

Quoted 12 lines · Andy Shevchenko

When legacy devices are present on x86 machine, the RTC IRQ has a dedicated pre-defined value. Use it instead of hard coded number.

---
v2: Added tags

 drivers/rtc/rtc-cmos.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Applied, thanks.

Signature · 3 lines
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
#3 lore raw

PATCH v2 3/3 rtc: cmos: Refactor code by using the new dmi_get_bios_year() helper

In reply to Andy Shevchenko

Refactor code by using the new dmi_get_bios_year() helper instead of open coding its functionality. This also makes logic slightly clearer.

No changes intended.

---
v2: Added tags

 drivers/rtc/rtc-cmos.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/rtc/rtc-cmos.c +4 -4
index aee55b658f85..b795fe4cbd2e 100644
@@ -1197,8 +1197,6 @@ static void rtc_wake_off(struct device *dev)
 /* Enable use_acpi_alarm mode for Intel platforms no earlier than 2015 */
 static void use_acpi_alarm_quirks(void)
 {
-	int year;
-
 	if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
 		return;
 
@@ -1208,8 +1206,10 @@ static void use_acpi_alarm_quirks(void)
 	if (!is_hpet_enabled())
 		return;
 
-	if (dmi_get_date(DMI_BIOS_DATE, &year, NULL, NULL) && year >= 2015)
-		use_acpi_alarm = true;
+	if (dmi_get_bios_year() < 2015)
+		return;
+
+	use_acpi_alarm = true;
 }
 #else
 static inline void use_acpi_alarm_quirks(void) { }
Signature · 1 line
2.24.1
maintainer Applied
#4 lore raw

Re: [PATCH v2 3/3] rtc: cmos: Refactor code by using the new dmi_get_bios_year() helper

Quoted 5 lines · Andy Shevchenko

Refactor code by using the new dmi_get_bios_year() helper instead of open coding its functionality. This also makes logic slightly clearer.

No changes intended.

Quoted 9 lines

---
v2: Added tags

 drivers/rtc/rtc-cmos.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

Applied, thanks.

Signature · 3 lines
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
maintainer Applied
#5 lore raw
In reply to Andy Shevchenko
Quoted 51 lines · Andy Shevchenko

As reported by Guilherme G. Piccoli:

---8<---8<---8<---

The rtc-cmos interrupt setting was changed in the commit 079062b28fb4 ("rtc: cmos: prevent kernel warning on IRQ flags mismatch") in order to allow shared interrupts; according to that commit's description, some machine got kernel warnings due to the interrupt line being shared between rtc-cmos and other hardware, and rtc-cmos didn't allow IRQ sharing that time.

After the aforementioned commit though it was observed a huge increase in lost HPET interrupts in some systems, observed through the following kernel message:

[...] hpet1: lost 35 rtc interrupts

After investigation, it was narrowed down to the shared interrupts usage when having the kernel option "irqpoll" enabled. In this case, all IRQ handlers are called for non-timer interrupts, if such handlers are setup in shared IRQ lines. The rtc-cmos IRQ handler could be set to hpet_rtc_interrupt(), which will produce the kernel "lost interrupts" message after doing work - lots of readl/writel to HPET registers, which are known to be slow.

Although "irqpoll" is not a default kernel option, it's used in some contexts, one being the kdump kernel (which is an already "impaired" kernel usually running with 1 CPU available), so the performance burden could be considerable. Also, the same issue would happen (in a shorter extent though) when using "irqfixup" kernel option.

In a quick experiment, a virtual machine with uptime of 2 minutes produced

Quoted 1 line

300 calls to hpet_rtc_interrupt() when "irqpoll" was set, whereas without

sharing interrupts this number reduced to 1 interrupt. Machines with more hardware than a VM should generate even more unnecessary HPET interrupts in this scenario.

---8<---8<---8<---

After looking into the rtc-cmos driver history and DSDT table from the Microsoft Surface 3, we may notice that Hans de Goede submitted a correct fix (see dependency below). Thus, we simply revert the culprit commit.

Fixes 079062b28fb4 ("rtc: cmos: prevent kernel warning on IRQ flags mismatch")

Depends-on: a1e23a42f1bd ("rtc: cmos: Do not assume irq 8 for rtc when there are no legacy irqs")

Quoted 5 lines

---
v2: Added tags, rework scissors to avoid cutting commit message (Guilherme)

 drivers/rtc/rtc-cmos.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Applied, thanks.

Signature · 3 lines
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com