[GIT PULL] x86: clean up use of system_state in virt_addr_valid and co

[GIT PULL] x86: clean up use of system_state in virt_addr_valid and co

[Jeremy Fitzhardinge]

These two changes remove the tests based on system_state in the virt_addr_valid/phys_addr functions.
It replaces the i386 test with an explicit flag to indicate whether the vmalloc range has been determined,
and on x86-64, it pre-initializes the x86_phys_bits to a reasonable number so that there's no dependency
on init order.

	J

The following changes since commit 4d02f9f7dac27f801fed69be8ebb0b95b798b12c:
  Ingo Molnar (1):
        Merge branch 'x86/urgent'

are available in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen.git x86/virt

Jeremy Fitzhardinge (2):
      x86-32: use specific __vmalloc_start_set flag in __virt_addr_valid
      x86-64: pre-initialize boot_cpu_data.x86_phys_bits to avoid system_state tests

 arch/x86/include/asm/pgtable_32_types.h |    5 +++++
 arch/x86/kernel/setup.c                 |    4 +++-
 arch/x86/mm/init_32.c                   |    3 +++
 arch/x86/mm/ioremap.c                   |   14 +++++---------
 4 files changed, 16 insertions(+), 10 deletions(-)

Re: [GIT PULL] x86: clean up use of system_state in virt_addr_valid and co

[Ingo Molnar]

* Jeremy Fitzhardinge <jeremy@goop.org> wrote:

> These two changes remove the tests based on system_state in the virt_addr_valid/phys_addr functions.
> It replaces the i386 test with an explicit flag to indicate whether the vmalloc range has been determined,
> and on x86-64, it pre-initializes the x86_phys_bits to a reasonable number so that there's no dependency
> on init order.
>
> 	J
>
> The following changes since commit 4d02f9f7dac27f801fed69be8ebb0b95b798b12c:
>  Ingo Molnar (1):
>        Merge branch 'x86/urgent'
>
> are available in the git repository at:
>
>  git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen.git x86/virt
>
> Jeremy Fitzhardinge (2):
>      x86-32: use specific __vmalloc_start_set flag in __virt_addr_valid
>      x86-64: pre-initialize boot_cpu_data.x86_phys_bits to avoid system_state tests
>
> arch/x86/include/asm/pgtable_32_types.h |    5 +++++
> arch/x86/kernel/setup.c                 |    4 +++-
> arch/x86/mm/init_32.c                   |    3 +++
> arch/x86/mm/ioremap.c                   |   14 +++++---------
> 4 files changed, 16 insertions(+), 10 deletions(-)

Pulled into tip:x86/mm, thanks Jeremy!

Note, there were some conflicts with other changes so i 
cherry-picked them and fixed them up - please double-check the 
end result in tip:master.

Thanks,

	Ingo

Re: [GIT PULL] x86: clean up use of system_state in virt_addr_valid and co

[Jiri Slaby]

On 5.3.2009 03:15, Jeremy Fitzhardinge wrote:
> x86-32: use specific __vmalloc_start_set flag in __virt_addr_valid
> x86-64: pre-initialize boot_cpu_data.x86_phys_bits to avoid system_state
> tests

Looks good, thanks. Just an idea, wouldn't make sense to add the check 
directly into is_vmalloc_addr?

Re: [GIT PULL] x86: clean up use of system_state in virt_addr_valid and co

[Jeremy Fitzhardinge]

Jiri Slaby wrote:
> On 5.3.2009 03:15, Jeremy Fitzhardinge wrote:
>> x86-32: use specific __vmalloc_start_set flag in __virt_addr_valid
>> x86-64: pre-initialize boot_cpu_data.x86_phys_bits to avoid system_state
>> tests
>
> Looks good, thanks. Just an idea, wouldn't make sense to add the check 
> directly into is_vmalloc_addr?

I thought about it, but I think its simply invalid to call 
is_vmalloc_addr() until you can meaningfully have vmalloc addresses.  It 
would be nice to have some way to warn about callers who are using these 
predicates in a meaningless way, but there doesn't appear to be any 
sensible way to do so (I guess the case I'm concerned about is people 
using VMALLOC_START for something before it is meaningful).

    J