On 01/23/2015 08:43 AM, Rafael J. Wysocki wrote:
On Thursday, January 22, 2015 05:44:37 PM al.stone(a)linaro.org wrote:
> From: Al Stone <al.stone(a)linaro.org>
> The use of the ACPI _OSI method in Linux has a long and sordid history.
> Instead of perpetuating past complications on new architectures, the
> consensus amongst those writing the ACPI specification and those using
> it seems to be to ultimately deprecate the use of _OSI. I plan to
> propose such a change to the ACPI specification in the near future.
> In the meantime, these patches rearrange the implementation of _OSI so
> that it can be deprecated, or ultimately removed completely, on at least
> arm64 platforms. This is done by simply moving the functions implementing
> _OSI to arch-dependent locations. For x86, there should be no change
> in functionality. For ia64, while it does duplicate some code from x86,
> there is no longer any connection to the ACPI blacklist code that is only
> used by x86. For arm64, any use of the _OSI method generates a warning
> that it has been deprecated, and then always returns false; i.e., that
> the capability being queried for, whether OS name or functionality, is
> not supported. This is the first six of the patches.
While I seem to understand the motivation, I don't really like moving code
around, especially in the ia64 area where we have limited means to test the
changes. And I hate code duplication pretty much in any form.
I'm not too wild about it either, but it seemed to avoid #ifdefs when I
was trying to think it through.
Also I don't see anything wrong with sharing code between x86 and
we do today.
Fair enough. That's entirely up to you; the only reason I saw for
separating them was removing the one call to check the blacklist on
ia64. That would have a tiny time savings at boot time, but it is
definitely down in the noise level.
So, why don't you move the _OSI handling out of osl.c into a
that will only build if something like CONFIG_ARCH_SPECIFIC_ACPI_OSI is not
set (and same for blacklist.c). Then, you'll only need to make ARM64
set CONFIG_ARCH_SPECIFIC_ACPI_OSI and provide the necessary functions
and you can leave the other architectures alone.
Does that make sense to you?
It does. I'll rewrite these based on that approach. Definitely a lot
less code motion involved...
> The final patch changes the default value for the _OS_ method for
> only. Since there is no need to pretend to be older versions of Windows,
> or any other OS at all, the _OS_ method will return "Linux" on arm64.
> One can still use the acpi_os_name kernel parameter if there is a need
> to use some other value.
> Since we are breaking apart code previously shared, I have tried to make
> it so that applying the x86 patches alone will continue to compile, at
> the expense of breaking the build on non-x86 platforms. However, once
> all of the patches are applied, we should be able to compile on all three
> architectures. It is best to treat these as one whole.
> I have only done simple testing with these patches on arm64 and x86 (AMD
> Seattle and a Lenovo t440s ThinkPad, respectively). Things seem to work
> as they should once booted, but this is a very, very small sample of
> possible machines. The ia64 patches cross-compile, but I personally have
> no way to test them.
> The arm64 patches also rely on having applied Hanjun's patches for ACPI 5.1
> on arm64 . The x86 and ia64 parts are not dependent on that patch set,
> though, and should be usable independently (i.e., patches 1, 3, 4 and 6).
> NB: some of the patches do not pass checkpatch.pl; the failures I saw were
> all part of the original code but are only showing up because that code is
> changing location, so I have left them as is. If necessary, they could be
> fixed but I was more concerned about the number of changes needed on ia64
> and not having any way to test them.
>  https://lkml.org/lkml/2015/1/14/586
> Al Stone (6):
> ia64: ACPI: move kernel acpi files to a directory
> arm64: ACPI: move kernel acpi files to a directory
> x86: ACPI: create arch-dependent version of acpi_osi_handler()
> ia64: ACPI: create arch-dependent version of acpi_osi_handler()
> arm64: ACPI: create arch-dependent version of acpi_osi_handler()
> x86: ia64: arm64: ACPI: move _OSI support functions to arch-dependent
> Hanjun Guo (1):
> ACPI: use Linux as ACPI_OS_NAME for _OS on ARM64
> arch/arm64/Kconfig | 1 +
> arch/arm64/kernel/Makefile | 2 +-
> arch/arm64/kernel/acpi.c | 359 --------------
> arch/arm64/kernel/acpi/Makefile | 1 +
> arch/arm64/kernel/acpi/acpi.c | 359 ++++++++++++++
> arch/arm64/kernel/acpi/osi.c | 26 +
> arch/ia64/kernel/Makefile | 2 +-
> arch/ia64/kernel/acpi-ext.c | 104 ----
> arch/ia64/kernel/acpi.c | 1000 --------------------------------------
> arch/ia64/kernel/acpi/Makefile | 1 +
> arch/ia64/kernel/acpi/acpi-ext.c | 104 ++++
> arch/ia64/kernel/acpi/acpi.c | 1000 ++++++++++++++++++++++++++++++++++++++
> arch/ia64/kernel/acpi/osi.c | 119 +++++
> arch/x86/kernel/acpi/Makefile | 2 +-
> arch/x86/kernel/acpi/blacklist.c | 327 +++++++++++++
> arch/x86/kernel/acpi/boot.c | 5 +-
> arch/x86/kernel/acpi/osi.c | 255 ++++++++++
> drivers/acpi/Kconfig | 8 +
> drivers/acpi/Makefile | 1 -
> drivers/acpi/blacklist.c | 323 ------------
> drivers/acpi/osl.c | 217 ---------
> include/acpi/acconfig.h | 2 +
> include/acpi/platform/aclinux.h | 4 +
> include/linux/acpi.h | 4 +-
> 24 files changed, 2215 insertions(+), 2011 deletions(-)
> delete mode 100644 arch/arm64/kernel/acpi.c
> create mode 100644 arch/arm64/kernel/acpi/Makefile
> create mode 100644 arch/arm64/kernel/acpi/acpi.c
> create mode 100644 arch/arm64/kernel/acpi/osi.c
> delete mode 100644 arch/ia64/kernel/acpi-ext.c
> delete mode 100644 arch/ia64/kernel/acpi.c
> create mode 100644 arch/ia64/kernel/acpi/Makefile
> create mode 100644 arch/ia64/kernel/acpi/acpi-ext.c
> create mode 100644 arch/ia64/kernel/acpi/acpi.c
> create mode 100644 arch/ia64/kernel/acpi/osi.c
> create mode 100644 arch/x86/kernel/acpi/blacklist.c
> create mode 100644 arch/x86/kernel/acpi/osi.c
> delete mode 100644 drivers/acpi/blacklist.c
Red Hat, Inc.