This implements a get_cpu() and matching put_cpu() to safely hand out
the current CPU to avoid preempt races.  Andrew and I have been bitching
about the need for such a method.

I also went ahead and replaced an example of current explicit
preempt-off with the new methods, as a case in point.

One more build fix, this time for !CONFIG_SOFTWARE_SUSPEND but
CONFIG_ACPI:

I've choosen this solution for now, as it is safest for ACPI people.
I'll modify suspend.c so that freezing part can be included without
whole suspend-to-disk support being included.

Fix compile problem with suspending diabled

Here's suspend-to-{RAM,disk} combined patch for
2.5.17. Suspend-to-disk is pretty stable and was tested in
2.4-ac. Suspend-to-RAM is little more experimental, but works for me,
and is certainly better than disk-eating version currently in kernel.

Major parts are: process stopper, S3 specific code, S4 specific
code.

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

into tp1.ruhr-uni-bochum.de:/home/kai/kernel/v2.5/linux-2.5.make-as

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

People relied on the build generation number to find out if they actually
booted the right kernel, only incrementing it on config changes isn't
enough for them. So this patch goes back to the old behavior, only done
right this time:

We now don't increment the generation number on every invocation of make, 
but on every link of vmlinux. At the same time, this patch fixes the
Makefile to only relink vmlinux if any of its prequisite object files
(or the command line) changed. Incrementing the version number
happens *after* we decided to relink vmlinux.

Before overwriting version.h, check if it actually changed - otherwise
we will unnecessarily rebuild a lot of files, as e.g. module.h depends
on version.h, and many files include module.h

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

Hi Linus!

Please apply this patch. It updates the srm_env.c driver (to access
Alpha's SRM environment variables) to its current version (which is
already included in up-to-date 2.4.x kernels).

MfG, JBG

This patch cleans up the remaining direct tests against x86_capability.
      It moves the cpu_has_* macros to the more appropriate
cpufeature.h.  It also introduces the cpu_has() macro to test features
for individual cpus.

This patch converts drivers/char/random.c and
drivers/input/joystick/analog.c to use the cpu_has_tsc macro.

This patch takes the cpu_has_mmx macro introduced in the xor.h header
and puts it in the proper place.  It also converts the ov511 driver to
use the new macro.

Created the first of a series of usb-storage patches
this afternoon. Please find it below or on ftp.XX.kernel.org
under people/aeb/2.5.16-us-patch .
It removes 5% of the usb-storage code, but should, if I made
no mistake, not affect the behaviour of the code.

(This is the introduction of raw_bulk.c to hold USB-bulk
and scatter-gather code that was repeated three-five times
in various drivers. The next patch is smartmedia.c.)

fixed problem with UHCI_HCD drivers set to 'y' were not causing the build
to go into the drivers/usb/host directory.

new vendor and device ID

Additional support for external journal.  We're basically adding uuid's
to both the file system and the journal to tie the two together.

It should be _after_ we have flushed the TLB.

o Add dependencies on generated files explicitly to the
  aic7xxx Makefile - These cannot be figured out
  automatically
 
o Rename the precompiled firmware files to 
  shipped_aic7xxx_{reg,seq}.h, so that we don't overwrite 
  shipped files when regenerating the firmware

This patch fixes a problem with reiserfs on 64 bit machines.  Our
struct virtual_item is a different size there, and some calculations
that assume otherwise lead to this panic create_virtual_node:

vs-8030: create_virtual_node: virtual node space consumed

(Patch has been ACKed by the bfs maintainer)

Now that bfs no more is included in the big unions in fs.h it makes
sense to move the contents of bfs_fs_i.h and bfs_fs_sb.h to a bfs-private
location.  I've created fs/bfs/bfs.h for that, also merging in bfs_defs.h.

In addition I've changed si_imap to an unsigned long pointer as the bitops
now use that type explicitly.

Remove an unused label and an unused variable.

pg tried to issue bitops on int, this fixes it to use an ulong instead.

This makes tasklet_vec and tasklet_hi_vec static inside softirq.c, and
makes them __per_cpu_data.

This changes futex semantics to a simple "sleep if this address
equals this value" interface, which is more convenient for building
other primitives.  It also adds a timeout value.

Example library can be found at:
   http://www.kernel.org/pub/linux/kernel/people/rusty/futex-2.0.tar.gz

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

No more users of buffermem_pages are left, remove it.
While at it also remove some orphaned externs around it in swap.h

This is the most discussion-worthy patch of the series:  Change the
meaning of si_meminfo->bufferram from "all pages in pagecache
backed by block devices" to "all pages in pagecache".

In the header/manpage is is documented as "Memory used by buffers",
but as the buffercache is gone I think the new meaning fits the
intention from pre-pagecache days much better.

This one is a bit more controveral as it may break stupid userlevel
programs:  remove the 'Buffers' field from /proc/meminfo and make
the Cached field show the full pagecache size instead of subtracting
the block-device backed pages.

All /proc/meminfo-using programs I have (free, top) still work fine.

Remove buffermem_pages showing from sgi serial code (debug only).

With all work done by akpm in 2.5 Linus no more has a buffer cache
in the traditional sense.  Still we try to keep estimates about
what would be the buffer cache size by keeping the number of pages
indexed by block device inodes.  This is broken not only because the
old buffercache was also used for file data which is nowdays not
hashed to block device inodes and thus makes every user of this data
assume wrong numbers.  Second is is possible to use block device
pages not through the buffer_head interface (i.e. userspace
block device nodes, possibly JFS also soon).  In addition the atomic_t
used for this bookkepping (buffermem_pages) causes cacheline bouncing
on larger machines.

This is the first patch of a series to get rid of it.  It removes the
useless output of supposedly buffer pages in show_mem(), which is used
by the magic sysrq key code.

This patch restores the previous behavior of stopping the build
immediately on error (unless the -k option is given to make)

Before this patch, we would do the echo command no matter if the
compile failed, thus returning success always.

(Jan Harkes)

into tp1.ruhr-uni-bochum.de:/home/kai/kernel/v2.5/linux-2.5.make

into tp1.ruhr-uni-bochum.de:/home/kai/kernel/v2.5/linux-2.5.make-as

E.g. make arch/i386/kernel/process.[is] used to work at least
in the cases where the object wouldn't need any additional command
line flags defined in the local subdir.

Get back this behavior, I think I know how to do this correctly, too,
but that's for later.

Remove the default for UTS_MACHINE, it gets set in
include/linux/compile.h, just as UTS_RELEASE.

(noticed by David Gibson)