Host drivers should now use the scatterlist rather than accessing
the request directly.  This marks the request pointer as being
deprecated so that host driver authors get a chance to fix their
drivers up before we remove this member.

This converts the PXA MCI driver to use the supplied scatter list
rather than accessing the struct request directly, thereby making
this host driver independent of the block layer.

We also clean up the driver to use a consistent driver name for
resources.

Convert MMCI driver to use the scatter list provided by the upper
MMC layers.  This removes all knowledge of struct request from the
MMCI driver.

This is intended to remove block layer knowledge from MMC host
drivers.  Host drivers are expected to use the scatter list for
data transfers rather than accessing the struct request.

Optimize away the unconditional write to debug registers on signal delivery
path.  This is already done on x86_64.

We only need to write to dr7 if there is a breakpoint to re-enable, and 
MOVDR is a serializing instruction, which is expensive.  Getting rid of 
it gets a 33% faster signal delivery path (at least on Xeon - I didn't
test other CPUs, so your gain may vary).

[ Editors note: it's likely only that slow on Netburst.  Serializing is
  not that expensive, but it is likely that writing to %db7 invalidates
  the trace cache, which explains why it's so slow on Xeon - it's not
  just the op itself, it has to re-populate the cache all the time.
					--- Linus ]


Measured delta TSC for three paths on a 2.4GHz Xeon.

1) With unconditional write to dr7 :  800-1000 cycles
2) With conditional write to dr7   :  84-112 cycles
3) With unlikely write to dr7      :  84 cycles

Performance test using divzero microbenchmark (3 million divide by zeros):

With unconditional write:
   7.445 real / 6.136 system
   7.529 real / 6.482 system
   7.541 real / 5.974 system
   7.546 real / 6.217 system
   7.445 real / 6.167 system

With unlikely write:
   5.779 real / 4.518 system
   5.783 real / 4.591 system
   5.552 real / 4.569 system
   5.790 real / 4.528 system
   5.554 real / 4.382 system

That's about a 33% speedup - more than I expected; apparently getting rid
of the serializing instruction makes the do_signal path much faster.

Zachary Amsden (zach@vmware.com)

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

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

We decided to do this a different way.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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

In general it is not safe to do any non-ptrace wakeup of a thread in
TASK_TRACED, because the waking thread could race with a ptrace call
that could be doing things like mucking directly with its kernel stack. 

AFAIK noone has established that whatever clobberation ptrace can do to
a running thread is safe even if it will never return to user mode, so
we can't allow this even for SIGKILL.

What we _can_ safely do is make a thread switching out of TASK_TRACED
resume rather than sitting in TASK_STOPPED if it has a pending SIGKILL
or SIGCONT.  The following patch does this.

This should be sufficient for the shutdown case.  When killing all
processes, if the tracer gets killed first, the tracee goes into
TASK_STOPPED and will be woken and killed by the SIGKILL (same as
before).  If the tracee gets killed first, it gets a pending SIGKILL and
doesn't wake up immediately--but, now, when the tracer gets killed, the
tracee will then wake up to die. 

This will also fix the (same) situations that can arise now where you
have used gdb (or whatever ptrace caller), killed -9 the gdb and the
process being debugged, but still have to kill -CONT the process before
it goes away (now it should just go away either the first time or when
you kill gdb). 
Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This is a batch of sparse fixes for things in arch/ppc64/* and
include/asm-ppc64/*

More to come of course...
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

My latest 8250 patch prevented registration of "empty" ports (ports that
have a 0 iobase in the static table). Unfortunately, some archs seem to rely
on this, and so broke. This patch reverts that part of the patch.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This adds support for the Maple 970FX Eval Board.  It adds the basic
arch support.  For the Maple to be fully functional, it needs a couple
more patches to be applied for IDE and Ethernet that are currently
pending with the respective maintainers of those subsystems.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

The iommu_free_table() patch broke g5 only build by adding back some
incestuous relationship between generic code and pSeries code. 

This wraps this in #ifdef as a quick fix until the original author of
the patch comes up with a better solution.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

On machines using the "ISU" mecanism for the MPIC, the new driver didn't properly
calculate the new interrupt count when an ISU was added. That would cause later on
failure to request interrupts in the offending range.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

The ppc64 PCI code, when parsing the OF tree, may end up getting empty
regions in addition to the "normal" ones for the PHB (some pSeries OF
device-tree contains weird "ranges" properties). These are harmless but
do trigger some annoying warnings during boot, so let's ignore them.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

CONFIG_NET_HW_FLOWCONTROL is entirely unused now, and superceded by NAPI in
practice, so remove the dead code
Signed-off-by: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

into pobox.com:/garz/repo/net-drivers-2.6

into pobox.com:/garz/repo/net-drivers-2.6

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

into pobox.com:/garz/repo/netdev-2.6/e1000-2

Fix for updated pci_{save,restore}_state()
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

into pobox.com:/garz/repo/netdev-2.6/misc

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

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

Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>

Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>

Simplify ipv6_get_saddr(), ipv6_dev_get_saddr() and
rt6_purge_dflt_routers().
Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>

Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>

Signed-off-by: Brian Haley <Brian.Haley@hp.com>
Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>

This patch fixes a couple of places where the ppc64 iSeries code would
#include <stddef.h>. The only "system" include I consider acceptable is
<stdarg.h> provided by gcc.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Actually, I think we've caught your crash now.  If that code path
is triggering at all, then it'll trigger with TSO packets too.  If
we get a truly partial ack on a TSO packet, then tcp_tso_acked will
not trim it off.  So we will fall through to this last-ditch trim
call, which doesn't update packets_out.

There are two solutions to this problem.  I've taken the simpler
approach for now.  We simply trim off the partial bits in tcp_tso_acked
and live with the fact that the packet counters may differ from
what's on the netwrok by one.

Later on we can 'fix' this by remembering where the original TSO
packet started from, perhaps in skb->h or somewhere.  Dave, is this
worth it?
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

From: Denis Vlasenko <vda@port.imtp.ilyichevsk.odessa.ua>

- recode back-to-back fwd_rnd() pairs to avoid two register moves.

- ditto for inv_rnd().

- optimize out lea 0(%ebp),%ebp

- remove two stray insns

# size aes-i586-asm.o.org aes-i586-asm.o
   text    data     bss     dec     hex filename
   5971       0       0    5971    1753 aes-i586-asm.o.org
   5905       0       0    5905    1711 aes-i586-asm.o

Overall, patch does not add and does not modify any insns, only removes a
handful of them.  However, speed difference is way below noise level.

Run-tested with tcrypt module.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

From: Denis Vlasenko <vda@port.imtp.ilyichevsk.odessa.ua>

- Macro parameters renamed for clarity.

- Inaccurate comments fixed.

- ebp register usage de-obfuscated (this is needed for next patch).

No real code changes.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Patch moves large temporary u64 W[80] from stack to ctx struct:

* reduces stack usage by 640 bytes
* saves one 640-byte memset() per sha512_transform()
  (we still do it after *all* iterations are done)
* quite unexpectedly saves 1.6k of code on i386
  because stack offsets now fit into 8bits
  and many stack addressing insns got 3 bytes smaller:

# size sha512.o.org sha512.o
text       data     bss     dec     hex filename
8281        372       0    8653    21cd sha512.o.org
6649        372       0    7021    1b6d sha512.o

# objdump -d sha512.o.org | cut -b9- >sha512.d.org
# objdump -d sha512.o | cut -b9- >sha512.d
# diff -u sha512.d.org sha512.d
[snip]
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Looks like open-coded be_to_cpu.  GCC produces rather poor code for this. 
be_to_cpu produces asm()s which are ~4 times shorter.

Compile-tested only.

I am not sure whether input can be 64bit-unaligned.
If it indeed can be, replace:

((u64*)(input))[I]  ->  get_unaligned( ((u64*)(input))+I )
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Looks like open-coded be_to_cpu.  GCC produces rather poor code for this. 
be_to_cpu produces asm()s which are ~4 times shorter.

Compile-tested only.

I am not sure whether input can be 32bit-unaligned.
If it indeed can be, replace:

((u32*)(input))[I]  ->  get_unaligned( ((u32*)(input))+I )
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>