Skip to contentSkip to conversation

loosen dependancy on rtc cmos

PATCH Dave Jones 3 participants 7 messages Feb 14, 2007 – Feb 17, 2007

Subsystems (3)
lore
Share & tools
lore

Dave Jones

author
#0 lore raw

This option is useful for all of the X86 subarchs afaik (and especially X86_GENERICARCH).

drivers/rtc/Kconfig renamed from drivers/rtc/Kconfig~ +1 -1
@@ -101,7 +101,7 @@ comment "RTC drivers"
 
 config RTC_DRV_CMOS
 	tristate "PC-style 'CMOS' real time clock"
-	depends on RTC_CLASS && (X86_PC || ALPHA || ARM26 || ARM \
+	depends on RTC_CLASS && (X86 || ALPHA || ARM26 || ARM \
 		|| M32R || ATARI || POWERPC)
 	help
 	  Say "yes" here to get direct support for the real time clock
Signature · 1 line
http://www.codemonkey.org.uk
#1 lore raw
Quoted 1 line · Dave Jones

This option is useful for all of the X86 subarchs afaik (and especially X86_GENERICARCH).

You're right ... _potentially_ useful, which is the same standard used in most of the other cases. The "X86_PC" is debris from an early version of this patch, which limited the code to cases where it was known to work. The X86_PC hardware standard does standardize on this particular RTC, but non-PC platforms can use it too.

I still need to resubmit the patch, for X86_PC, which defines the platform device in the (common) case where PNPACPI isn't defined. Other X86 boards would need something similar, based on what chips are wired to the CPU.

Quoted 1 line
Quoted 15 lines
drivers/rtc/Kconfig renamed from drivers/rtc/Kconfig~ +1 -1
@@ -101,7 +101,7 @@ comment "RTC drivers"
 
 config RTC_DRV_CMOS
 	tristate "PC-style 'CMOS' real time clock"
-	depends on RTC_CLASS && (X86_PC || ALPHA || ARM26 || ARM \
+	depends on RTC_CLASS && (X86 || ALPHA || ARM26 || ARM \
 		|| M32R || ATARI || POWERPC)
 	help
 	  Say "yes" here to get direct support for the real time clock
Signature · 2 lines
-- 
http://www.codemonkey.org.uk
reviewer
#2 lore raw
Quoted 11 lines · David Brownell
Quoted 1 line · Dave Jones

This option is useful for all of the X86 subarchs afaik (and especially X86_GENERICARCH).

You're right ... _potentially_ useful, which is the same standard used in most of the other cases. The "X86_PC" is debris from an early version of this patch, which limited the code to cases where it was known to work. The X86_PC hardware standard does standardize on this particular RTC, but non-PC platforms can use it too.

I still need to resubmit the patch, for X86_PC, which defines the platform device in the (common) case where PNPACPI isn't defined.

CONFIG_PNPACPI=y is not the common case?

-Len

Quoted 30 lines

Other X86 boards
would need something similar, based on what chips are wired to the CPU.

Quoted 1 line
Quoted 15 lines
drivers/rtc/Kconfig renamed from drivers/rtc/Kconfig~ +1 -1
@@ -101,7 +101,7 @@ comment "RTC drivers"
 
 config RTC_DRV_CMOS
 	tristate "PC-style 'CMOS' real time clock"
-	depends on RTC_CLASS && (X86_PC || ALPHA || ARM26 || ARM \
+	depends on RTC_CLASS && (X86 || ALPHA || ARM26 || ARM \
 		|| M32R || ATARI || POWERPC)
 	help
 	  Say "yes" here to get direct support for the real time clock
Signature · 2 lines
-- 
http://www.codemonkey.org.uk

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#3 lore raw
Quoted 5 lines · Len Brown
Quoted 3 lines

I still need to resubmit the patch, for X86_PC, which defines the platform device in the (common) case where PNPACPI isn't defined.

CONFIG_PNPACPI=y is not the common case?

It's certainly not in the defconfig for x86-64. And it's only been three weeks since the CONFIG_EXPERIMENTAL dependency got removed.

So, no I would not think it's the common case.

- Dave

reviewer
#4 lore raw
Quoted 11 lines · David Brownell
Quoted 5 lines · Len Brown
Quoted 3 lines

I still need to resubmit the patch, for X86_PC, which defines the platform device in the (common) case where PNPACPI isn't defined.

CONFIG_PNPACPI=y is not the common case?

It's certainly not in the defconfig for x86-64. And it's only been three weeks since the CONFIG_EXPERIMENTAL dependency got removed.

So, no I would not think it's the common case.

Turns out that it is common.
It is in i386 defconfig, ships on i386 Fedora Core 6, ships on i386 OpenSuse 10.2.

For x86_64 it isn't in defconfig or OpenSuse 10.2, but is in Fedora Core 6.

So I've taken Andi's advice and checked in the patches below.
(I didn't bother updating defconfig, it will generate according to this rule and I don't want to conflict with any re-generation andi might be checking in)

thanks,
-Len

commit 243b66e76ab722cdec1921d7f80c0cb808131c37

Author: Len Brown <[email protected]>
Date:   Thu Feb 15 22:34:36 2007 -0500

ACPI: always enable CONFIG_PNPACPI on CONFIG_ACPI kernels

We removed the ACPI motherboard driver which handled the ACPI=y, PNP=n case, so now we need to enforce that PNP & PNPACPI are always enabled for ACPI kernels.

Most major distros ship this way this already.

Cc Bjorn Helgaas <[email protected]>
drivers/acpi/Kconfig +1 -0
index 20eacc2..2d21fed 100644
@@ -13,6 +13,7 @@ config ACPI
 	depends on IA64 || X86
 	depends on PCI
 	depends on PM
+	select PNP
 	default y
 	---help---
 	  Advanced Configuration and Power Interface (ACPI) support for 
drivers/pnp/pnpacpi/Kconfig +2 -14
index ad27e5e..b04767c 100644
@@ -2,17 +2,5 @@
 # Plug and Play ACPI configuration
 #
 config PNPACPI
-	bool "Plug and Play ACPI support"
-	depends on PNP && ACPI
-	default y
-	---help---
-	  Linux uses the PNPACPI to autodetect built-in
-	  mainboard resources (e.g. parallel port resources).
-
-          Some features (e.g. real hotplug) are not currently
-          implemented.
-
-          If you would like the kernel to detect and allocate resources to
-          your mainboard devices (on some systems they are disabled by the
-          BIOS) say Y here.  Also the PNPACPI can help prevent resource
-          conflicts between mainboard devices and other bus devices.
+	bool
+	default (PNP && ACPI)

commit 8d4956c201c2f7683289f70095443c59a39f94ef

Author: Len Brown <[email protected]>
Date:   Thu Feb 15 22:46:42 2007 -0500

ACPI: remove non-PNPACPI version of get_rtc_dev()

It isn't needed in ACPI code anymore because
now ACPI always includes PNPACPI.

Cc David Brownell <[email protected]>
drivers/acpi/glue.c +0 -27
index 9950087..4334c20 100644
@@ -255,8 +255,6 @@ arch_initcall(init_acpi_device_notify);
 static struct cmos_rtc_board_info rtc_info;
 
 
-#ifdef CONFIG_PNPACPI
-
 /* PNP devices are registered in a subsys_initcall();
  * ACPI specifies the PNP IDs to use.
  */
@@ -280,31 +278,6 @@ static struct device *__init get_rtc_dev(void)
 	return bus_find_device(&pnp_bus_type, NULL, NULL, pnp_match);
 }
 
-#else
-
-/* We expect non-PNPACPI platforms to register an RTC device, usually
- * at or near arch_initcall().  That also helps for example PCs that
- * aren't configured with ACPI (where this code wouldn't run, but the
- * RTC would still be available).  The device name matches the driver;
- * that's how the platform bus works.
- */
-#include <linux/platform_device.h>
-
-static int __init platform_match(struct device *dev, void *data)
-{
-	struct platform_device	*pdev;
-
-	pdev = container_of(dev, struct platform_device, dev);
-	return strcmp(pdev->name, "rtc_cmos") == 0;
-}
-
-static struct device *__init get_rtc_dev(void)
-{
-	return bus_find_device(&platform_bus_type, NULL, NULL, platform_match);
-}
-
-#endif
-
 static int __init acpi_rtc_init(void)
 {
 	struct device *dev = get_rtc_dev();
#5 lore raw
Quoted 1 line · Len Brown

So I've taken Andi's advice and checked in the patches below.

OK; that simplifies things for me, good! I can discard that patch (broken by Andi's pcspkr change anyway), stop worring about whether most folk will even see that driver, and make time to look at the ACPI hooks for RTC wakeup, instead. :)

But it would be nice to see the PNP bus infrastructure upgraded in various ways too, now that its availability is less iffy.

- Add something analagous to platform_driver_probe() so that the init code can be removed after it runs.

- Add shutdown() calls to the PNP bus. Otherwise e.g. one must use shutdown notifiers for PNP interfaces, while normal driver model code works for other interfaces to such hardware.

Transitioning from "legacy" drivers to PNP (for PCs) and platform_bus (other platforms) is a bit awkward because of differences like those. Drivers still need too many different modes; it's too complicated to have a core with different bus glues as thin veneers ... the glue must be thicker (and thus error prone). Similarly, I/O space resource reservation acts differently.

I can hope that having PNPACPI be more common will start nudging more drivers to get rid of their "legacy" modes, or at least focus on "Real Driver" modes that don't involve poking at hardware and hoping it doesn't bite back. ;)

- Dave

Quoted 29 lines

commit 243b66e76ab722cdec1921d7f80c0cb808131c37

Author: Len Brown <[email protected]>
Date:   Thu Feb 15 22:34:36 2007 -0500

ACPI: always enable CONFIG_PNPACPI on CONFIG_ACPI kernels

We removed the ACPI motherboard driver which handled the ACPI=y, PNP=n case, so now we need to enforce that PNP & PNPACPI are always enabled for ACPI kernels.

Most major distros ship this way this already.

Cc Bjorn Helgaas <[email protected]>

commit 8d4956c201c2f7683289f70095443c59a39f94ef

Author: Len Brown <[email protected]>
Date:   Thu Feb 15 22:46:42 2007 -0500

ACPI: remove non-PNPACPI version of get_rtc_dev()

It isn't needed in ACPI code anymore because
now ACPI always includes PNPACPI.

Cc David Brownell <[email protected]>
#6 lore raw
Quoted 8 lines · David Brownell
Quoted 1 line · Len Brown

So I've taken Andi's advice and checked in the patches below.

OK; that simplifies things for me, good! I can discard that patch (broken by Andi's pcspkr change anyway), stop worring about whether most folk will even see that driver, and make time to look at the ACPI hooks for RTC wakeup, instead. :)

Which, by the way, means that folk with pre-ACPI systems aren't going to be able to use this driver until someone else provides an updated patch creating an "rtc_cmos" platform device.

Or on the more modern end of things ... I suspect PCs using Linux-BIOS won't do PNPACPI either, and they will also need that platform device before using this driver.

What do the MiniMac systems do? At one time I thought they were ACPI-free.

- Dave