Nothing uses v86mode any more, so this removes it,
and its space in the thread_struct.

As per CodingStyle

- Align comments
- Remove hardcoded value (Use proper alignment
  on systems with 64byte cachelines)

Code cleanup..

memcpy_to/fromio shouldn't be using memcpy,
as this could equate to the versions that
prefetch, which could upset some devices.

Other slabs have '_' to separate words instead of ' '.
Same patch went into 2.4

- remove unneeded includes (lots of)
- Some Dell folks noted that the switch to realmode needs
  to be a wbinvd instead of an invd on some boxes.
  This has been in -dj for eons, and has had no ill effect.

Several new boards, and fix a forward reference by
moving a function around.

Numerous updates from 2.4 (see changelog)

- moves functions around to match where they are in 2.4
  (makes syncing easier)
- Adds some extra entries

- Recognise and decode trace cache
- Add several missing elements to descriptor table
- Xeon errata workaround
- updated names for P4's

From 2.4, see comment in diff.

Old patch by Krishna Ramachandran

get_vm_area() doesn't play too well when passed
malformed arguments due to a missing check.
Originally by davem, munged by me.

From John Levon. This corrects the comment
to match the real quotation.

Patch splitting/bombing is a very boring job,
if it wasn't for light relief like this, it'd
be even moreso 8-)

From 2.4 with some 2.5 mangling from Robert Love.

From Zwane Mwaikambo, this diff makes sure we don't
poke reserved bits on the P4.

make mrproper misses some files..

Missing kill_sb member meant that futexfs died
a death when someone tried to mount it.

This patch from Ravikiran G Thirumalai removes
cacheline bouncing of apic_timer_irqs

Straightforward diff to allow longer names
for PCI devices

- Standardise usage on IA-32 instead of a mix of IA32 and ix86.
- Add some missing options.
- Fix up some options.

From 2.4

From 2.4

Files moved, so the comments should be updated
to reflect their new locations.