Skip to contentSkip to conversation

rtc: add x86 support for rtc-efi

PATCH In review Brian Maly 6 participants 15 messages Apr 6, 2009 – May 27, 2009

Subsystems (2)
lore
Share & tools
lore

Brian Maly

author
#0 lore raw

PATCH 1/1 rtc: add x86 support for rtc-efi

This patch adds support for rtc-efi (RTC Class Driver for EFI-based systems) to x86.

This patch is based on dann frazier patch [add rtc platform driver for EFI -> http://lkml.org/lkml/2009/1/8/434], and also depends on dann's patch. Dann's patch was in the mm tree, though I think its been merged already.

This patch has been tested and works perfectly on every x86 EFI system I could find, though all have been EL64 systems. I would appreciate any testing feedback from EL32 systems from anyone that has one and can test.

drivers/rtc/Kconfig +1 -1
@@ -442,7 +442,7 @@ config RTC_DRV_DS1742
 
 config RTC_DRV_EFI
 	tristate "EFI RTC"
-	depends on IA64
+	depends on IA64 || (X86 && EFI) || (X86_64 && EFI)
 	help
 	  If you say yes here you will get support for the EFI
 	  Real Time Clock.
arch/x86/kernel/time_32.c +22 -0
@@ -32,6 +32,7 @@
 #include <linux/interrupt.h>
 #include <linux/time.h>
 #include <linux/mca.h>
+#include <linux/platform_device.h>
 
 #include <asm/arch_hooks.h>
 #include <asm/hpet.h>
@@ -135,3 +136,24 @@ void __init time_init(void)
 	tsc_init();
 	late_time_init = choose_time_init();
 }
+
+#ifdef CONFIG_EFI
+extern int efi_enabled;
+
+static struct platform_device rtc_efi_dev = {
+	.name = "rtc-efi",
+	.id = -1,
+};
+
+static int __init rtc_init(void)
+{
+	if (efi_enabled) {
+		if (platform_device_register(&rtc_efi_dev) < 0)
+			printk(KERN_ERR "unable to register rtc device...\n");
+	}
+	/* not necessarily an error */
+	return 0;
+}
+module_init(rtc_init);
+#endif
+
arch/x86/kernel/time_64.c +22 -0
@@ -18,6 +18,7 @@
 #include <linux/time.h>
 #include <linux/mca.h>
 #include <linux/nmi.h>
+#include <linux/platform_device.h>
 
 #include <asm/i8253.h>
 #include <asm/hpet.h>
@@ -135,3 +136,24 @@ void __init time_init(void)
 
 	late_time_init = choose_time_init();
 }
+
+#ifdef CONFIG_EFI
+extern int efi_enabled;
+
+static struct platform_device rtc_efi_dev = {
+	.name = "rtc-efi",
+	.id = -1,
+};
+
+static int __init rtc_init(void)
+{
+	if (efi_enabled) {
+		if (platform_device_register(&rtc_efi_dev) < 0)
+			printk(KERN_ERR "unable to register rtc device...\n");
+	}
+	/* not necessarily an error */
+	return 0;
+}
+module_init(rtc_init);
+#endif
+
#1 lore raw

[Adding rtc list to the CC]

Quoted 8 lines · Brian Maly

This patch adds support for rtc-efi (RTC Class Driver for EFI-based systems) to x86.

This patch is based on dann frazier patch [add rtc platform driver for EFI -> http://lkml.org/lkml/2009/1/8/434], and also depends on dann's patch. Dann's patch was in the mm tree, though I think its been merged already.

Quoted 5 lines

This patch has been tested and works perfectly on every x86 EFI system I could find, though all have been EL64 systems. I would appreciate any testing feedback from EL32 systems from anyone that has one and can test.

hey Brian,
Thanks for working on this. You might want to resubmit with the patch
inlined instead of as an attachment to make it easier to review.

Signature · 1 line
dann frazier
#2 lore raw
In reply to Brian Maly
Quoted 8 lines · Brian Maly

This patch adds support for rtc-efi (RTC Class Driver for EFI-based systems) to x86.

This patch is based on dann frazier patch [add rtc platform driver for EFI -> http://lkml.org/lkml/2009/1/8/434], and also depends on dann's patch. Dann's patch was in the mm tree, though I think its been merged already.

Quoted 9 lines

This patch has been tested and works perfectly on every x86 EFI system I could find, though all have been EL64 systems. I would appreciate any testing feedback from EL32 systems from anyone that has one and can test.

- depends on IA64
+ depends on IA64 || (X86 && EFI) || (X86_64 && EFI)

"depends on EFI"?

-hpa

author
#3 lore raw

re-post of the patch, both as inline and as an attachment (which hopefully wont cause confusion).

drivers/rtc/Kconfig 2009-04-06 18:46:58.000000000 -0400 renamed from drivers/rtc/Kconfig 2009-03-26 12:39:59.000000000 -0400 +0 -0
@@ -442,7 +442,7 @@ config RTC_DRV_DS1742
 

config RTC_DRV_EFI
tristate "EFI RTC"
- depends on IA64
+ depends on EFI
help
If you say yes here you will get support for the EFI Real Time Clock.

arch/x86/kernel/time_32.c 2009-03-26 13:06:10.000000000 -0400 renamed from arch/x86/kernel/time_32.c 2009-03-12 22:39:28.000000000 -0400 +0 -0
@@ -32,6 +32,7 @@

#include <linux/interrupt.h>
#include <linux/time.h>
#include <linux/mca.h>
+#include <linux/platform_device.h>

#include <asm/arch_hooks.h>
#include <asm/hpet.h>
@@ -135,3 +136,21 @@ void __init time_init(void)

    tsc_init();
    late_time_init = choose_time_init();
}

+
+
+#ifdef CONFIG_RTC_DRV_EFI
+static struct platform_device rtc_efi_dev = {
+ .name = "rtc-efi",
+ .id = -1,
+};
+
+static int __init rtc_init(void)
+{
+ if (platform_device_register(&rtc_efi_dev) < 0)
+ printk(KERN_ERR "unable to register rtc device...\n");
+
+ /* not necessarily an error */
+ return 0;
+}
+module_init(rtc_init);
+#endif

arch/x86/kernel/time_64.c 2009-03-26 13:06:19.000000000 -0400 renamed from arch/x86/kernel/time_64.c 2009-03-12 22:39:28.000000000 -0400 +0 -0
@@ -18,6 +18,7 @@

#include <linux/time.h>
#include <linux/mca.h>
#include <linux/nmi.h>
+#include <linux/platform_device.h>

#include <asm/i8253.h>
#include <asm/hpet.h>
@@ -135,3 +136,20 @@ void __init time_init(void)

    late_time_init = choose_time_init();
}

+
+#ifdef CONFIG_RTC_DRV_EFI
+static struct platform_device rtc_efi_dev = {
+ .name = "rtc-efi",
+ .id = -1,
+};
+
+static int __init rtc_init(void)
+{
+ if (platform_device_register(&rtc_efi_dev) < 0)
+ printk(KERN_ERR "unable to register rtc device...\n");
+
+ /* not necessarily an error */
+ return 0;
+}
+module_init(rtc_init);
+#endif

--------------------------------------------------------------------------------------------

drivers/rtc/Kconfig +1 -1
@@ -442,7 +442,7 @@ config RTC_DRV_DS1742
 
 config RTC_DRV_EFI
 	tristate "EFI RTC"
-	depends on IA64
+	depends on EFI
 	help
 	  If you say yes here you will get support for the EFI
 	  Real Time Clock.
arch/x86/kernel/time_32.c +19 -0
@@ -32,6 +32,7 @@
 #include <linux/interrupt.h>
 #include <linux/time.h>
 #include <linux/mca.h>
+#include <linux/platform_device.h>
 
 #include <asm/arch_hooks.h>
 #include <asm/hpet.h>
@@ -135,3 +136,21 @@ void __init time_init(void)
 	tsc_init();
 	late_time_init = choose_time_init();
 }
+
+
+#ifdef CONFIG_RTC_DRV_EFI
+static struct platform_device rtc_efi_dev = {
+	.name = "rtc-efi",
+	.id = -1,
+};
+
+static int __init rtc_init(void)
+{
+	if (platform_device_register(&rtc_efi_dev) < 0)
+		printk(KERN_ERR "unable to register rtc device...\n");
+
+	/* not necessarily an error */
+	return 0;
+}
+module_init(rtc_init);
+#endif
arch/x86/kernel/time_64.c +18 -0
@@ -18,6 +18,7 @@
 #include <linux/time.h>
 #include <linux/mca.h>
 #include <linux/nmi.h>
+#include <linux/platform_device.h>
 
 #include <asm/i8253.h>
 #include <asm/hpet.h>
@@ -135,3 +136,20 @@ void __init time_init(void)
 
 	late_time_init = choose_time_init();
 }
+
+#ifdef CONFIG_RTC_DRV_EFI
+static struct platform_device rtc_efi_dev = {
+	.name = "rtc-efi",
+	.id = -1,
+};
+
+static int __init rtc_init(void)
+{
+	if (platform_device_register(&rtc_efi_dev) < 0)
+		printk(KERN_ERR "unable to register rtc device...\n");
+
+	/* not necessarily an error */
+	return 0;
+}
+module_init(rtc_init);
+#endif
Acked-by
#4 lore raw
Quoted 81 lines · Brian Maly

re-post of the patch, both as inline and as an attachment (which hopefully wont cause confusion).

drivers/rtc/Kconfig 2009-04-06 18:46:58.000000000 -0400 renamed from drivers/rtc/Kconfig 2009-03-26 12:39:59.000000000 -0400 +0 -0
@@ -442,7 +442,7 @@ config RTC_DRV_DS1742
 

config RTC_DRV_EFI
tristate "EFI RTC"

drivers/rtc/Kconfig 2009-04-06 18:46:58.000000000 -0400 renamed from drivers/rtc/Kconfig 2009-03-26 12:39:59.000000000 -0400 +1 -1
-    depends on IA64
+   depends on EFI
    help
      If you say yes here you will get support for the EFI
      Real Time Clock.
arch/x86/kernel/time_32.c 2009-03-26 13:06:10.000000000 -0400 renamed from arch/x86/kernel/time_32.c 2009-03-12 22:39:28.000000000 -0400 +19 -0
@@ -32,6 +32,7 @@
 #include <linux/interrupt.h>
 #include <linux/time.h>
 #include <linux/mca.h>
+#include <linux/platform_device.h>
 
 #include <asm/arch_hooks.h>
 #include <asm/hpet.h>
@@ -135,3 +136,21 @@ void __init time_init(void)
    tsc_init();
    late_time_init = choose_time_init();
 }
+
+
+#ifdef CONFIG_RTC_DRV_EFI
+static struct platform_device rtc_efi_dev = {
+    .name = "rtc-efi",
+    .id = -1,
+};
+
+static int __init rtc_init(void)
+{
+    if (platform_device_register(&rtc_efi_dev) < 0)
+        printk(KERN_ERR "unable to register rtc device...\n");
+
+    /* not necessarily an error */
+    return 0;
+}
+module_init(rtc_init);
+#endif
arch/x86/kernel/time_64.c 2009-03-26 13:06:19.000000000 -0400 renamed from arch/x86/kernel/time_64.c 2009-03-12 22:39:28.000000000 -0400 +18 -1
@@ -18,6 +18,7 @@
 #include <linux/time.h>
 #include <linux/mca.h>
 #include <linux/nmi.h>
+#include <linux/platform_device.h>
 
 #include <asm/i8253.h>
 #include <asm/hpet.h>
@@ -135,3 +136,20 @@ void __init time_init(void)
 
    late_time_init = choose_time_init();
 }
+
+#ifdef CONFIG_RTC_DRV_EFI
+static struct platform_device rtc_efi_dev = {
+    .name = "rtc-efi",
+    .id = -1,
+};
+
+static int __init rtc_init(void)
+{
+    if (platform_device_register(&rtc_efi_dev) < 0)
+        printk(KERN_ERR "unable to register rtc device...\n");
+
+    /* not necessarily an error */
+    return 0;
+}
+module_init(rtc_init);
+#endif
 
--------------------------------------------------------------------------------------------
Signature · 1 line
dann frazier
Acked-by dann frazier
#5 lore raw
In reply to Brian Maly
Quoted 3 lines · On Mon, 06 Apr 2009 18:55:49 -0400 Brian Maly <[email protected]> wrote:

re-post of the patch, both as inline and as an attachment (which hopefully wont cause confusion).

Normally it does confuse me horridly because I get two copies of the same patch, which applies happily with patch --dry-run', but dies horridly with patch --for-real'. But I was warned...

In future, please include a full copy of the (possibly modified) changelog with each iteration of a patch.

As this patch is mostly-x86 I tossed it onto my for-Ingo pile.

From: Brian Maly <[email protected]>

Add support for rtc-efi (RTC Class Driver for EFI-based systems) to x86.

This patch has been tested and works perfectly on every x86 EFI system I could find, though all have been EL64 systems. I would appreciate any testing feedback from EL32 systems from anyone that has one and can test.

Cc Alessandro Zummo <[email protected]>
Cc David Brownell <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc "H. Peter Anvin" <[email protected]>

---

 arch/x86/kernel/time_32.c |   19 +++++++++++++++++++
 arch/x86/kernel/time_64.c |   18 ++++++++++++++++++
 drivers/rtc/Kconfig       |    2 +-
 3 files changed, 38 insertions(+), 1 deletion(-)
arch/x86/kernel/time_32.c renamed from arch/x86/kernel/time_32.c~rtc-add-x86-support-for-rtc-efi +19 -0
diff -puN arch/x86/kernel/time_32.c~rtc-add-x86-support-for-rtc-efi arch/x86/kernel/time_32.c
@@ -32,6 +32,7 @@
 #include <linux/interrupt.h>
 #include <linux/time.h>
 #include <linux/mca.h>
+#include <linux/platform_device.h>
 
 #include <asm/setup.h>
 #include <asm/hpet.h>
@@ -135,3 +136,21 @@ void __init time_init(void)
 	tsc_init();
 	late_time_init = choose_time_init();
 }
+
+
+#ifdef CONFIG_RTC_DRV_EFI
+static struct platform_device rtc_efi_dev = {
+	.name = "rtc-efi",
+	.id = -1,
+};
+
+static int __init rtc_init(void)
+{
+	if (platform_device_register(&rtc_efi_dev) < 0)
+		printk(KERN_ERR "unable to register rtc device...\n");
+
+	/* not necessarily an error */
+	return 0;
+}
+module_init(rtc_init);
+#endif
arch/x86/kernel/time_64.c renamed from arch/x86/kernel/time_64.c~rtc-add-x86-support-for-rtc-efi +18 -0
diff -puN arch/x86/kernel/time_64.c~rtc-add-x86-support-for-rtc-efi arch/x86/kernel/time_64.c
@@ -18,6 +18,7 @@
 #include <linux/time.h>
 #include <linux/mca.h>
 #include <linux/nmi.h>
+#include <linux/platform_device.h>
 
 #include <asm/i8253.h>
 #include <asm/hpet.h>
@@ -133,3 +134,20 @@ void __init time_init(void)
 
 	late_time_init = choose_time_init();
 }
+
+#ifdef CONFIG_RTC_DRV_EFI
+static struct platform_device rtc_efi_dev = {
+	.name = "rtc-efi",
+	.id = -1,
+};
+
+static int __init rtc_init(void)
+{
+	if (platform_device_register(&rtc_efi_dev) < 0)
+		printk(KERN_ERR "unable to register rtc device...\n");
+
+	/* not necessarily an error */
+	return 0;
+}
+module_init(rtc_init);
+#endif
drivers/rtc/Kconfig renamed from drivers/rtc/Kconfig~rtc-add-x86-support-for-rtc-efi +1 -1
diff -puN drivers/rtc/Kconfig~rtc-add-x86-support-for-rtc-efi drivers/rtc/Kconfig
@@ -455,7 +455,7 @@ config RTC_DRV_DS1742
 
 config RTC_DRV_EFI
 	tristate "EFI RTC"
-	depends on IA64
+	depends on EFI
 	help
 	  If you say yes here you will get support for the EFI
 	  Real Time Clock.

_

#14 lore raw
In reply to Brian Maly

Hi, Brian,

Sorry for late.

1. Why not put these code into arch/x86/kernel/rtc.c? I think that is more approriate.

2. It seems that you use both CMOS RTC and EFI RTC. In code path of read_persistent_clock() and update_persistent_clock(), CMOS RTC is used for x86_64. I think you should change that too. I have tried to do that in http://lkml.org/lkml/2007/8/13/602.

Best Regards,
Huang Ying

Quoted 17 lines · Brian Maly

This patch adds support for rtc-efi (RTC Class Driver for EFI-based systems) to x86.

This patch is based on dann frazier patch [add rtc platform driver for EFI -> http://lkml.org/lkml/2009/1/8/434], and also depends on dann's patch. Dann's patch was in the mm tree, though I think its been merged already.

This patch has been tested and works perfectly on every x86 EFI system I could find, though all have been EL64 systems. I would appreciate any testing feedback from EL32 systems from anyone that has one and can test.

#6 lore raw
In reply to Andrew Morton
Quoted 33 lines · Andrew Morton
Quoted 3 lines · On Mon, 06 Apr 2009 18:55:49 -0400 Brian Maly <[email protected]> wrote:

re-post of the patch, both as inline and as an attachment (which hopefully wont cause confusion).

Normally it does confuse me horridly because I get two copies of the same patch, which applies happily with patch --dry-run', but dies horridly with patch --for-real'. But I was warned...

In future, please include a full copy of the (possibly modified) changelog with each iteration of a patch.

As this patch is mostly-x86 I tossed it onto my for-Ingo pile.

From: Brian Maly <[email protected]>

Add support for rtc-efi (RTC Class Driver for EFI-based systems) to x86.

This patch has been tested and works perfectly on every x86 EFI system I could find, though all have been EL64 systems. I would appreciate any testing feedback from EL32 systems from anyone that has one and can test.

Cc Alessandro Zummo <[email protected]>
Cc David Brownell <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc "H. Peter Anvin" <[email protected]>
Quoted 6 lines

---

 arch/x86/kernel/time_32.c |   19 +++++++++++++++++++
 arch/x86/kernel/time_64.c |   18 ++++++++++++++++++
 drivers/rtc/Kconfig       |    2 +-
 3 files changed, 38 insertions(+), 1 deletion(-)

Hm, it would be nice to first unify the relevant bits of arch/x86/kernel/time_{32|64}.c into arch/x86/kernel/time.c, and then we can apply such patches without duplicative effects.

Ingo

author
#7 lore raw
Quoted 4 lines

Hm, it would be nice to first unify the relevant bits of arch/x86/kernel/time_{32|64}.c into arch/x86/kernel/time.c, and then we can apply such patches without duplicative effects.

Ingo

Ingo,

Are you OK with consolidating this into arch/x86/kernel/rtc.c as Huang Ying had suggested? This seems like the most logical place for the rtc-efi init to happen, but your suggestion to consolidate this into arch/x86/kernel.time.c may have advantages that I am not aware of. Anyway, I would appreciate any insight/opinions on this if you have any. Thanks.

Brian

#8 lore raw
Quoted 4 lines · Brian Maly
Quoted 3 lines

Hm, it would be nice to first unify the relevant bits of arch/x86/kernel/time_{32|64}.c into arch/x86/kernel/time.c, and then we can apply such patches without duplicative effects.

Quoted 8 lines

Ingo,

Are you OK with consolidating this into arch/x86/kernel/rtc.c as Huang Ying had suggested? This seems like the most logical place for the rtc-efi init to happen, but your suggestion to consolidate this into arch/x86/kernel.time.c may have advantages that I am not aware of. Anyway, I would appreciate any insight/opinions on this if you have any. Thanks.

Yes, that indeed sounds like an even better place for it.

Thanks,

Ingo

#9 lore raw
Quoted 9 lines · Ingo Molnar
Quoted 4 lines · Brian Maly
Quoted 1 line

Hm, it would be nice to first unify the relevant bits of arch/x86/kernel/time_{32|64}.c into arch/x86/kernel/time.c, and then we can apply such patches without duplicative effects.

Ingo,

Are you OK with consolidating this into arch/x86/kernel/rtc.c as Huang Ying had suggested? This seems like the most logical place for the rtc-efi init to happen, but your suggestion to consolidate this into arch/x86/kernel.time.c may have advantages that I am not aware of. Anyway, I would appreciate any insight/opinions on this if you have any. Thanks.

Yes, that indeed sounds like an even better place for it.

Furthermore, the EFI RTC code probably should be in its own file.

In fact, arch/x86/kernel really could use more subdirectories; at least the EFI and UV-specific code should be be moved out.

-hpa

author
#10 lore raw
Quoted 8 lines · H. Peter Anvin
Quoted 7 lines · Ingo Molnar
Quoted 4 lines · Brian Maly
Quoted 1 line

Hm, it would be nice to first unify the relevant bits of arch/x86/kernel/time_{32|64}.c into arch/x86/kernel/time.c, and then we can apply such patches without duplicative effects.

Ingo,

Are you OK with consolidating this into arch/x86/kernel/rtc.c as Huang Ying had suggested? This seems like the most logical place for the rtc-efi init to happen, but your suggestion to consolidate this into arch/x86/kernel.time.c may have advantages that I am not aware of. Anyway, I would appreciate any insight/opinions on this if you have any. Thanks.

Quoted 9 lines
Quoted 2 lines

Yes, that indeed sounds like an even better place for it.

Furthermore, the EFI RTC code probably should be in its own file.

In fact, arch/x86/kernel really could use more subdirectories; at least the EFI and UV-specific code should be be moved out.

-hpa

Peter,

How do you think this should be restructured (ideally)?

Maybe create a sub-directory (arch/x86/kernel/rtc) and in this sub-directory split the existing rtc.c code into:
(1) rtc.c for common code used by all RTC, (2) efi_rtc.c for EFI specific code
(3) cmos_rtc.c for cmos/legacy specific code

Anyone have any objections/concerns with this or perhaps have a better way?

Brian

#11 lore raw
In reply to H. Peter Anvin
Quoted 15 lines · H Peter
Quoted 14 lines · Ingo Molnar
Quoted 11 lines · Brian Maly
Quoted 3 lines

Hm, it would be nice to first unify the relevant bits of arch/x86/kernel/time_{32|64}.c into arch/x86/kernel/time.c, and then we can apply such patches without duplicative effects.

Ingo,

Are you OK with consolidating this into arch/x86/kernel/rtc.c as Huang Ying had suggested? This seems like the most logical place for the rtc-efi init to happen, but your suggestion to consolidate this into arch/x86/kernel.time.c may have advantages that I am not aware of. Anyway, I would appreciate any insight/opinions on this if you have any. Thanks.

Quoted 7 lines
Quoted 2 lines

Yes, that indeed sounds like an even better place for it.

Furthermore, the EFI RTC code probably should be in its own file.

In fact, arch/x86/kernel really could use more subdirectories; at least the EFI and UV-specific code should be be moved out.

Or, do you think it is appropriate to re-organize EFI related code into a sub-architecture?

Best Regards,
Huang Ying

#12 lore raw
Quoted 8 lines · Ying
Quoted 7 lines · H Peter
Quoted 6 lines · Ingo Molnar
Quoted 4 lines · Brian Maly
Quoted 1 line

Hm, it would be nice to first unify the relevant bits of arch/x86/kernel/time_{32|64}.c into arch/x86/kernel/time.c, and then we can apply such patches without duplicative effects.

Ingo,

Are you OK with consolidating this into arch/x86/kernel/rtc.c as Huang Ying had suggested? This seems like the most logical place for the rtc-efi init to happen, but your suggestion to consolidate this into arch/x86/kernel.time.c may have advantages that I am not aware of. Anyway, I would appreciate any insight/opinions on this if you have any. Thanks.

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines

Yes, that indeed sounds like an even better place for it.

Furthermore, the EFI RTC code probably should be in its own file.

In fact, arch/x86/kernel really could use more subdirectories; at least the EFI and UV-specific code should be be moved out.

Or, do you think it is appropriate to re-organize EFI related code into a sub-architecture?

No, we're been trying to get rid of subarchitectures in the x86 kernel.
The reason is that the notion of subarchitectures matches reality in
x86-land poorly. Most variants of x86 share considerable code: UV has EFI, PC has EFI or BIOS, Voyager has BIOS and a standard RTC, and so on.

-hpa

#13 lore raw
Quoted 27 lines · H. Peter Anvin
Quoted 17 lines · Ying
Quoted 16 lines · H Peter
Quoted 15 lines · Ingo Molnar
Quoted 11 lines · Brian Maly
Quoted 3 lines

Hm, it would be nice to first unify the relevant bits of arch/x86/kernel/time_{32|64}.c into arch/x86/kernel/time.c, and then we can apply such patches without duplicative effects.

Ingo,

Are you OK with consolidating this into arch/x86/kernel/rtc.c as Huang Ying had suggested? This seems like the most logical place for the rtc-efi init to happen, but your suggestion to consolidate this into arch/x86/kernel.time.c may have advantages that I am not aware of. Anyway, I would appreciate any insight/opinions on this if you have any. Thanks.

Yes, that indeed sounds like an even better place for it.

Quoted 8 lines
Quoted 4 lines

Furthermore, the EFI RTC code probably should be in its own file.

In fact, arch/x86/kernel really could use more subdirectories; at least the EFI and UV-specific code should be be moved out.

Or, do you think it is appropriate to re-organize EFI related code into a sub-architecture?

Quoted 5 lines

No, we're been trying to get rid of subarchitectures in the x86 kernel. The reason is that the notion of subarchitectures matches reality in x86-land poorly. Most variants of x86 share considerable code: UV has EFI, PC has EFI or BIOS, Voyager has BIOS and a standard RTC, and so on.

Yeah - we essentially got rid of the ex subarchitecture code in 2.6.30. We still have config options that turn on various sorts of behavior - but most of them are runtime. We definitely dont want to reintroduce anything subarchitecture-ish.

Ingo