I had to move the rtnetlink_init and init_netlink calls to af_netlink init time, so that
the sk_alloc called down the rtnetlink_init callchain is done after the PF_NETLINK
net_proto_family is sock_registered, and because of that the af_netlink init function
call had to be moved to earlier by means of subsys_initcall (DaveM's suggestion).

This fixes a mismatch in declaration between "irport_interrupt" in the
header files (returning void) and in the definition (returning
irqreturn_t).

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

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

request_irq requires a handler that returns irqreturn_t,
so mm_interrupt now returns the appropriate value

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

into shaggy.austin.ibm.com:/shaggy/bk/jfs-2.5

This fixes a problem with drivers that have request on the stack for
some operations, like IDE.  If we wake before releasing the request, the
stack may have already disappeared beneath us when the rest of
end_that_request_last() is run.

Fix by making sure the completion is done _last_.

into kernel.bkbits.net:/home/davem/net-2.5

into kernel.bkbits.net:/home/davem/sparc-2.5

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

give a clue about what the context was that might have caused
the spurious interrupt.

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

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

into flint.arm.linux.org.uk:/usr/src/linux-bk-2.5/linux-2.5-rmk

Patch from Steven Cole.

Here are some spelling fixes for arm.
This was diffed against the current 2.5 tree.

Patch from Eli Carter

# Mon Apr 21 11:11:41 CDT 2003 ejc@rnd-linux-c84
# use-physoffset-define
#
# Use the PHYS_OFFSET #define instead of a hard-coded value.
#
# Diffed against linux-2.5.67-rmk1+1501-3+1506
#
#  arch/arm/boot/compressed/head-xscale.S |    2 +-
#  1 files changed, 1 insertion(+), 1 deletion(-)
#

Patch from Eli Carter

# Mon Apr 21 11:20:06 CDT 2003 ejc@rnd-linux-c84
# pci-io-mem-size-defines
#
# Rename the IOP321_PCI_WINDOW_SIZE #defines to use
# IOP321_PCI_{IO,MEM}_{BASE,SIZE} instead.  This makes the #defines a bit more
# consistent.  No functional change, though it does bring up the question of
# whether res[].end should be BASE+SIZE-1 or BASE+SIZE.
#
# Diffed against linux-2.5.67-rmk1+1501-3+1506+1508-10
#
#  arch/arm/mach-iop3xx/iop321-pci.c    |    8 ++++----
#  include/asm-arm/arch-iop3xx/iop321.h |   10 ++++------
#  2 files changed, 8 insertions(+), 10 deletions(-)
#

Patch from Eli Carter

# Thu Apr 17 16:37:58 CDT 2003 ejc@rnd-linux-c84
# iq80321-use-addr-defines
#
# Use #define's instead of hard-coded numbers for the device locations for the
# iq80321 board.
#
# Diffed against linux-2.5.65-rmk1+1472-4+1502-3+1506
# Applies cleanly to linux-2.5.67-rmk1+1502-3+1506
#
#  arch/arm/mach-iop3xx/mm-321.c |    4 ++--
#  1 files changed, 2 insertions(+), 2 deletions(-)
#

of Intel CPU optimizations. From Andi Kleen.

This implements automatic code patching of memory barriers based
on the CPU capabilities. Normally lock ; addl $0,(%esp) barriers
are used, but these are a bit slow on the Pentium 4.

Linus proposed this a few weeks ago after the support for SSE1/SSE2
barriers was introduced. I now got around to implement it.

The main advantage is that it allows distributors to ship less binary
kernels but still get fast kernels. In particular it avoids the
need of a special Pentium 4 kernel.

The patching code is quite generic and could be used to patch
other instructions (like prefetches or specific other critical
instructions) too.
Thanks to Rusty's in kernel loader it also works seamlessly for modules.

The patching is done before other CPUs start to avoid potential
erratas with self modifying code on SMP systems. It makes no
attempt to automatically handle assymetric systems (an secondary
CPU having less capabilities than the boot CPU). In this
case just boot with "noreplacement"

The VMA merging code can merge vmas which have a ->vm_ops->close() handler.
But this means that per-VMA resources in one of the merged VMAs will not be
freed.

Fix that up by just refusing to merge any VMA's which look "complex": they
have special flags set or they have a ->close handler.

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

Here is a trivial fix for task_prio() in the case MAX_RT_PRIO !=
MAX_USER_RT_PRIO.  In this case, all priorities are skewed by
(MAX_RT_PRIO - MAX_USER_RT_PRIO).

The fix is to subtract the full MAX_RT_PRIO value from p->prio, not just
MAX_USER_RT_PRIO.  This makes sense, as the full priority range is
unrelated to the maximum user value.  Only the real maximum RT value
matters.

This has been in Andrew's tree for awhile, with no issue.  Also, Ingo
acked it.