From: Manfred Spraul <manfred@colorfullife.com>

slab.c is not very efficient for passing objects between cpus.  Usually this
is a rare event, but with network routing and cpu-affine NICs it is possible
that nearly all allocation operations will occur on one cpu, and nearly all
free operations on another cpu.

This causes slab memory to be returned to slab's free page list rather than
being cached on behalf of the particular slab cache.

The attached patch solves that by adding an array of objects that is shared
by all cpus.  Instead of multiple linked list operations per object, object
pointers are now passed to/from the shared array (and thus between cpus) with
memcopy operations.  On uniprocessor, the default array size is 0, because
the shared array and the per-cpu head array are redundant.

Additionally, the patch exports more statistics in /proc/slabinfo and make
the array sizes tunable by writing to /proc/slabinfo.  Both changes break
backward compatibility, user space scripts must look at the slabinfo version
and act accordingly.

The size of the new shared array may be altered at runtime, by writing to
/proc/slabinfo.

The new parameters for writing to /proc/slabinfo are:

	#echo "cache-name limit batchcount shared" > /proc/slabinfo

For example "size-4096 120 60 8" improves the slab efficiency for network
routing, because the default values (24 12 8) are too small for the large
series generated due to irq mitigation.  Note that only root has write
permissions to /proc/slabinfo.

These changes provided an overall 12% speedup in Robert Olson's gigE
packet-formwarding testing on 2-way.

disk->minor_shift is not used anymore.  Remove it.

Fix various typos and and information about the new kconfig features.

This creates the configuration in the destination directory instead of
the current directory.

conf: better choice interface
      don't ask for unchangable symbols

mconf: mark unchangable symbols with '---'
       update exit text (from Sam Ravnborg <sam@ravnborg.org>)

qconf: update debug output

Add a number of warnings to avoid misuse of the previously added features
(most important check for recursive dependencies).

The 'range' keyword allows to define a lower and upper limit for integer
and hex symbols.

The 'select' keyword defines a lower limit for symbols and allows to
select other symbols when a symbol is selected, e.g.:

config JOLIET
	bool "Microsoft Joliet CDROM extensions"
	select NLS

This means when JOLIET is selected, NLS will be selected as well.

"default" accepts now not only a single symbol but also an expression
which can be assigned to boolean and tristate symbols.

Add the following new keywords: def_tristate, def_bool, def_boolean,
select, enable and range.

Add support for def_tristate and def_bool, which combines default and
bool/tristate into a single statement and a allows simpler definition
of derived symbols.

This changes P_ROOTMENU into a MENU_ROOT flag and also fixes some qconf
usability problems.

Some gconf fixes by Romain Lievin <roms@tilp.info>.

Remove old code and debugging macros which were used by the cml1->kconfig
converter.

 ETH1394: Use 64bit EUI as the hardware address.
 ETH1394: Support broadcast packets.
 SBP2   : Fix max_payload for > S400
 CORE   : Fix iso.c compilation by including linux/sched.h directly.

The was a race triggered by heavy MT usage that could have caused
processes in D state. Bad Davide, bad ...

Also, the semaphore is now per-epoll-fd and not global. Plus some comment
adjustment.

Now that we use idedefault_driver.drives instead of ata_unused, the order
of drives on the driver->drives becomes significant; note that when we added
to ata_unused, we had done that to the end of list.

So we should do the same with the "drives" list.

into home.transmeta.com:/home/torvalds/v2.5/linux

  - added drivers for Digigram VXPocket V2, VXPocket 440 (pcmcia)
  - added driver for Digigram VXP220 V2/Mic (PCI)
  - added driver Harmony chipset (parisc)
  - added OPL4 driver
  - code cleanups - removed 2.2 and 2.4 code
  - nm256 driver - added Dell Latitude LS workaround
  - ac97 driver - fixes in intialization
  - usb audio driver - strlcat() fixes

into nuts.ninka.net:/home/davem/src/BK/net-2.5

into suse.cz:/home/perex/bk/linux-sound/linux-sound

With this the area that is shared among struct sock and struct tcp_tw_bucket
is exact, i.e. there is no fields that are in struct sock and only reused
for other reasons in tcp_tw_bucket, while keeping both struct free of holes.

into nuts.ninka.net:/home/davem/src/BK/net-2.5

Avoid touching the next page in the NMI handler stack validity check.

That way, we can start unmapping pages that aren't in use, without
triggering this case.

More zlib K&R to ANSI C function header conversions.

This updates the matroxfb driver to the new framebuffer API.

I'm sorry that it is quite large, but due to completely changed
underlying API there is no reasonable way how to split it into smaller
pieces.

 - Removed support for text mode. No way for it with current API.
 - Removed support for hardware cursor. Generic cursor code has enough
   troubles as is, in software mode.
 - No reasonable fbset support... It is especially annoying on multihead
   system, as 'stty cols XXX rows YYY' does not change pixclock...
 - Removed fastfont support. No way for it with current API.

(Mis)features inherited from generic fbdev API:
 - Cursor on other framebuffers than primary one does not blink.
 - Contents of visible, but not foreground, display is not updated.

into penguin.transmeta.com:/home/torvalds/v2.5/linux

This patch reinstates the ability of tty devices to use dynamically
allocated major numbers yet specify the minor numbers statically.

The synclink drivers do this.

This patch adds some new vendor/product ids for the USB scanner
driver.

into kroah.com:/home/greg/linux/BK/gregkh-2.5

The recent writev() fix broke the invariant that ->commit_write _must_ be
called after a successful ->prepare_write().  It leaves ext3 with a
transaction stuck open and the filesystem locks up.