Compile fixes

This patch renames KTHREAD_SIZE into THREAD_SIZE, similiar to all other archs.

Add missing include files.

Fix idle functions.

- replace hwclk_time with rtc_time
- use hwclk instead of gettod to set initial time

compile fix

Add missing license tags.

Update to the latest config.in.

This moves several cache flush function into cacheflush.h.

Add __clear_bit/__ffs/sched_find_first_bit.

Some compile fixes.

- sizeof changes from Kernel Janitor Project
- several bug fixes found with fsx

Here is one fix and one enhancement to the scsi generic (sg)
driver in 2.5.14
Changelog:
    - off by one fix for last scatter gather element
    - if possible compact kiobuf_map into scatter gather list

This is the OnStream variant of the st driver that
has been in Dave's tree for some time.

The change looks like it is from Willem Riede (osst@riede.org)
and that it has been forwarded ported from lk 2.4 .

This ISA scsi adapter still seems to be popular
given the number of people that supply patches
for it collected in Dave's tree.

The latest patch was from William Stinson (first item):
  - request_region cleanup
  - removal of old scsi error handling (leaving newer version)
  - scatterlist::address replaced
  - host_lock replacing io_request_lock


Doug Gilbert

P.S. This is a transfer from Dave's tree. It runs fine for
me on a dual Celeron SMP box.

after the manual merge of the PCI re-organization
removed it.

The largest part of the OSS update from 2.4
Various changelog entries scattered throughout the patch.

From 2.4
- Lindented
- Add Docbook comments
- Remove duplicate MODULE_LICENSE tag

The PNPBIOS driver and the floppy driver both fight over reservation
of the floppy port. This patch makes things friendly again.
As the comment suggests, this fix could be done better, but until
someone steps forward to fix this, this is the best we have.

copy_to_user already does the relevant checking.

Done by William Stinson.
Adds error handling to request_region() calls,
and converts some old check_region() calls too.

This patch from 2.4 makes sure we don't prefetch past the
end of a range to be copied (in case its at the end of a memrange)
i386 case looks safe already, we just weren't optimal for the last
chunk to be copied.

Andi. same change needed for x86-64.

    Dave.

Syncs up with latest from Stelian

Brings up to date with the quirk in 2.4
This makes sure we only apply it on the right chipsets,
and adjusts the bits that get poked depending on which we find.

Done by William Stinson.
Adds error handling to request_region() calls,
and converts some old check_region() calls too.

From 2.4

Done by William Stinson.
Adds error handling to request_region() calls,
and converts some old check_region() calls too.

Adds missing checks. Probably another from William Stinson

Syncs up with Stelians latest

Done by William Stinson.
Adds error handling to request_region() calls,
and converts some old check_region() calls too.

Another Sony for the apm blacklist.

Done by William Stinson.
Adds error handling to request_region() calls,
and converts some old check_region() calls too.

Done by William Stinson.
Adds error handling to request_region() calls,
and converts some old check_region() calls too.