Rewrite/cleanup of the SLB management code.  This removes nearly all the
SLB related code from arch/ppc64/kernel/stab.c and puts a rewritten version
in arch/ppc64/mm, where it better belongs.  The main SLB miss path is in
assembler and the other routines have been cleaned up and streamlined.

Notable changes:

- Ugly bitfields no longer used for generating SLB entries.

- slb_allocate() (the main SLB miss routine) is now in assembler, and all
  the data it uses is stored in the PACA.

- The mm context is now copied into the PACA at context switch time, to
  avoid looking up the thread struct on SLB miss.

- An SLB miss will now never (directly) result in a call to
  do_page_fault.  If we get a miss on a totally bogus address the handler
  will now put in an SLB referencing VSID 0.  This will never have any
  pages, so we'll get the (fatal) page fault shortly afterwards.  This
  simplifies the SLB entry and exit paths.

- The round-robin pointer in the PACA now references the last-used
  instead of next-to-use SLB slot, which simplifies the asm for updating it
  slightly.

- Unify do_slb_bolted with the general SLB miss path.  There is now one
  SLB miss handler, in assembler, and called with only the low-level
  exception prolog (EXCEPTION_PROLOG_[PI]SERIES rather than
  EXCEPTION_PROLOG_COMMON) and minimal extra save/restore logic.

- Streamlines the exception entry/exit path of the SLB miss handler to
  shave a few cycles off.  The most significant change is that the RI bit
  is left off throughout the whole handler, which avoids an extra mtmsrd to
  turn it back off on the exit path.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Patch below fixes a cpu_up race in PPC64.
Signed-off-by: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

In migrate_irqs_away we werent converting a virtual irq to a real one.  We
ended up passing the wrong irq numbers to the hypervisor and migration of
affinitised irqs on cpu hot unplug didnt work.

Also clarify the rtas_stop_self printk.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Both arch/ppc64/kernel/ras.c and arch/ppc64/kernel/rtasd.c have a spinlock
variable declared static called "log_lock".  Since the code in these files
interact quit a lot, having two different locks with identical names is
manifestly confusing.  This patch renames both locks to something a little
clearer.  In the case of ras.c it also renames the buffer protected by the
lock to a more usefullly greppable name.
Signed-off-by: David Gibson <dwg@au.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Moves a misplaced of_node_put().  In the existing code, the node in
question is used just after its refcount is decremented.
Signed-off-by: John Rose <johnrose@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

The patch below removes the unused member "pci_io_offset" from struct
pci_controller.  If there are no problems, please apply.
Signed-off-by: John Rose <johnrose@austin.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

lmb_end_of_DRAM() returns the address of the end of RAM, not the starting
address of the last page.  We've been accessing mem_map[] out of bounds for
quite a while.  But, it's just a read, so it's probably never caused a real
problem.

But, during my port of CONFIG_NONLINEAR to ppc64, I have a check to make
sure that all __va() calls are given with valid physical addresses.  This
code tripped that check.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

arch/ppc64/kernel/prom.c: In function `__make_room':
arch/ppc64/kernel/prom.c:1415: warning: unused variable `offset'
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

BSD accounting cross-platform compatibility is a new feature of 2.6.8 and
thus not crucial, but it'd be nice not to have kernels writing wrong file
formats out in the wild.

The endianness detection logic I wanted to suppose for userspace turned out
to be bogus.  So just do it the simple way and store endianness info
together with the version number.
Signed-off-by: Tim Schmielau <tim@physik3.uni-rostock.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

The per cpu schedule counters need to be summed up over all possible cpus. 
When testing hotplug cpu remove I saw the sum of the online cpu count for
nr_uninterruptible go negative which made the load average go nuts.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

The latest versions of Motorola erratas for the MPC745x CPUs (and 744x)
adds a couple of nasty ones for which we really want workarounds in the
kernel. One is to disable the BTIC branch target cache on some revs
(too bad for performances...) and the other one is to force cacheable
memory pages to always be marked as SMP coherent even on UP systems (I
didn't measure significant perfs impact with this one).
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

into ppc970.osdl.org:/home/torvalds/v2.6/linux

into ppc970.osdl.org:/home/torvalds/v2.6/linux

If regs.len is smaller than reglen it's possible to copy more bytes out
than the user asked for.
Signed-off-by: Chris Wright <chrisw@osdl.org>

into ppc970.osdl.org:/home/torvalds/v2.6/linux

Below is an updated version of patch from Fruhwirth which integrates the 
Gladman AES code into the crypto API.

I've tried to ensure that this is done as simply as possible: the user 
gets the asm version by default if it's suitable.

I've also now added the alternate GPL licensing provided by Brian Gladman, 
and licensed the code as GPL.
Signed-off-by: James Morris <jmorris@redhat.com>
Signed-off-by: David S. Miller <davem@redhat.com>

The missing break causes the packet to be tested against
both encapsulation types, one will always fail.
Signed-off-by: bert hubert <ahu@ds9a.nl>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@redhat.com>

I'm looking through the xfrm_alloc_spi stuff and noticed that the
netlink alloc_spi function takes the range in host order while the
PFKEY alloc_spi function takes them in network order.

First I thought that I stuffed up since I was the one who changed
the code in the netlink interface to take them in host order :)

But reading RFC 2367 seems to indicate otherwise.  It says that all
fields are host order unless specified otherwise.  And the spirange
fields are not specified to be network order at all.

Looking at the existing PFKEY users:

User Space
----------
Openswan - Doesn't use PFKEY for this.
Racoon - Puts zeros in there so it doesn't care.  However its test-pfkey
	 program stores things in host order.
ISAKMPD - Stores things in host order.

So the conclusion is that we can and should change our PFKEY
implementation to use host order for these fields.

This patch does exactly that.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@redhat.com>

xfrm_alloc_spi will always succeed if minspi < maxspi, even if
minspi + 1 == maxspi.  If the range is already occupied this
will obviously lead to breakage.

Of course this is very unlikely to occur in reality due to the
size of the range.  Although with IPCOMP it might actually happen
on a very large server.

The fix is obivous.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@redhat.com>

This is the patch referred to in the netlink_get_spi thread.

I was actually wrong about the reason for this patch though.  Firstly
it's the SPI check that is redundant and not the find_acq() call.
And it's redundant because of the find_acq() patch, not because
of the fact that this is in xfrm_state_add().

Now that find_acq() only returns SAs with SPIs, we don't need to
check this in xfrm_state_add() anymore.

We do still need the call though to clean up leftover larval states.

Another side-effect of the change is that we can move the existence
check above find_acq() since find_acq() will never return any SAs
matching the SPI we're trying to add (It doesn't need to because if
an SA with a matching SPI existed, it would've been returned by
state_lookup() already).
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@redhat.com>

The issue is that two successive calls to netlink_get_spi is returning
the same SA.  Since netlink_get_spi is meant to be a creation operation
this is incorrect.

The netlink_get_spi operation is modelled off the PFKEY SADB_GETSPI
command which is specified in RFC 2367.  The purpose of SADB_GETSPI
is to create a new larval SA that can then be filled in by SADB_UPDATE.

Its semantics does not allow two SADB_GETSPI calls to return the same
SA, even if there is no SADB_UPDATE call in between.

The reason the second netlink_get_spi is returning the same SA is
because in find_acq(), the code is looking at all larval states as
opposed to only larval states with an SPI of zero.

Since the only other caller of find_acq() -- xfrm_state_add() intentionally
ignores all return values with a non-zero SPI, it is safe to not look at
SAs with non-zero SPIs at all in find_acq().

The following patch does exactly that.

In fact, the find_acq() call in xfrm_state_add() is a remnant from
the days when we had xfrm_state_replace() instead of xfrm_state_add()
and xfrm_state_update().  It can now be safely removed.

I'll post a separate patch for that.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@redhat.com>

into nuts.davemloft.net:/disk1/BK/net-2.6

The problem is that after a successful connection between the Windows
Bluetooth stack and the Linux Bluez stack, no packets from the device
ever reach the PC running Windows XP Service Pack 2. That is, a ping
from the PC never receives a response, and a ping from BlueZ never
reaches the PC. Linux packet statistics show that the PC packets are
received, but all return traffic seems to be routed over the loopback
interface.

Immediately after creating the BNEP connection with BlueZ, the Windows
Bluetooth stack sends a BNEP_FILTER_NET_TYPE_SET_MSG with an effective
length of zero. BlueZ interprets this message to mean that no filter
ranges should be allowed. The code zeros the first entry in the filter
list, which is than interpreted as meaning that no ranges of acceptable
packets are available. This interpretation is wrong and leads to all
packets being rejected by BNEP.

The Bluetooth BNEP specification clearly states the following:

The length (in octets) of this message is 4+4*N, where N is the number
of disjoint ranges of Networking protocol types that form the complete
set. Note that N=0 (empty set) denotes a reset to default filters (if
any) supported by the remote device.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>

into nuts.davemloft.net:/disk1/BK/net-2.6

For the USB dongles from ISSC the first command must be a HCI_Reset,
because otherwise the inquiry procedures won't work.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>

into ppc970.osdl.org:/home/torvalds/v2.6/linux

On a board containing the HPT372N IDE controller the 2.6.x series kernels will
misbehave. If the HPT372N is set up with the newer PCI identifier it is 
ignored. If it is set up with the HPT372 identifier then the kernel crashes
on boot.

This patch is a forward port of my 2.4 driver fixes that have been in 2.4
for a year but somehow escaped 2.6. Ronny Buchmann caught a couple
of merge details I missed and those are fixed in this diff too.

As well as adding 372N support this also fixes the unknown revision case
to avoid crashes should any future 37x variants with weird class_rev's appear
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

into hostme.bitkeeper.com:/repos/c/cifs/linux-2.5cifs

Signed-off-by: Steve French (sfrench@us.ibm.com)

The logic in shrink_slab tries to balance the proportion of slab which it
scans against the proportion of pagecache which the caller scanned.  Problem
is that with a large number of highmem LRU pages and a small number of lowmem
LRU pages, the amount of pagecache scanning appears to be very small, so we
don't push slab hard enough.

The patch changes things so that for, say, a GFP_KERNEL allocation attempt we
only consider ZONE_NORMAL and ZONE_DMA when calculating "what proportion of
the LRU did the caller just scan".

This will have the effect of shrinking slab harder in response to GFP_KERNEL
allocations than for GFP_HIGHMEM allocations.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Currently, ppc32 will not always compile with binutils-2.15.  The issue
is that binutils has become even more strict about which opcodes can be
used with which CPU flags.  The problem is that we have a number of
cases where we compile with altivec instructions (with runtime checks to
make sure we can actually run them) in code that's not altivec specific.

The fix for this is to always pass in -maltivec on CONFIG_6xx.  To do
this cleanly, we split our AFLAGS definition up into
aflags-$(CONFIG_FOO).
Signed-off-by: Tom Rini <trini@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

into ppc970.osdl.org:/home/torvalds/v2.6/linux

into hostme.bitkeeper.com:/repos/c/cifs/linux-2.5cifs

Signed-off-by: Steve French (sfrench@us.ibm.com)

all sorts of minor stuff - basically, all chunks are independent here,
but IMO that one is not worth splitting.  Contains:
	* pmac_cpufreq.c: declaration in the middle of a block.
	* sys_ia32.c: couple of trivial annotations.
	* ipmi_si_intf.c: should be using asm/irq.h instead of linux/irq.h
	* synclink_cs.c: assignment-in-conditional with nobody ever looking
at the variable we are assigning to afterwards; variable removed.
	* sbni.c: s/__volatile/__volatile__
	* matroxfb_base.h: got rid of ((u32 *)p)++
	* asm-ppc/checksum.h and asm-sparc64/floppy.h: NULL noise removal
	* amd64 compat.h: missing L in long constant.
	* mtd-abi.h: annotated ioctl structure
	* sysctl.c: corrected annotations in extern
Signed-off-by: Al Viro <viro@parcelfarce.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

code using atm_cirange fields assumes that they are signed; make that
explicit by s/char/signed char/.
Signed-off-by: Al Viro <viro@parcelfarce.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

more size_t fixes (%d -> %zd)
Signed-off-by: Al Viro <viro@parcelfarce.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

drivers/net BROKEN_ON_SMP annotations and NULL noise removals.
Signed-off-by: Al Viro <viro@parcelfarce.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

annotations and NULL noise removals in drivres/isdn (more BROKEN_ON_SMP
stuff)
Signed-off-by: Al Viro <viro@parcelfarce.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>