1) More reasonable ip_rt_gc_min_interval default
2) Trim less valuable entries in hash chain during
   rt_intern_hash when such chains grow too long.

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

1) We export __check_region, so making it __deprecated gives a spurious
   warning in kernel/ksyms.c.
2) Other warnings refer to __check_region rather than check_region,
   which has confused some people.

Make check_region an inline, not a macro, and deprecate *that*.

A number of small ColdFire serial driver fixups:
 - use correct irq handler return type (irqreturn_t)
 - add support for mark and space parity
 - some casting to keep types clean (and remove compile warnings)
 - fix missing ";" after local tty_driver struct definition

We no longer use sub-architecture speicific interrupt setups
in the m68knommu branch. Remove obsolete call to 68328 irq
init.

Make the ROMfs copy in the startup code for Motorola/M5206eLITE board
conditional on actually using a ROMfs setup.

Minor board name change in m68knommu setup,
"Dragon Engine" -> "DragonEngine".

The NCP ioctls are declared unconditionally in compat_ioctl now, which needs this
include to still compile on x86-64.

The us->ns conversion must be in the "tv" if branch of settimeofday,
not in the "tz" if.

Also make it compile for x86-64 again. The 32bit emulation was forgotten
in the conversion.

This moves the initialization of the struct scsi_device fields
use_10_for_ms and use_10_for_rw into a central place.  This allows the
host's slave_configure() function to change them (with effect).  This
also collapses two identical sections of code into a single one.

This is needed by usb-storage, so we can resolve some Babble problems,
some device-crashing problems, and remove a great deal of troublesome
mode-sense-translation code.

This update covers a lot of changes to arch/parisc and parisc-specific
drivers.

 - irq_return_t changes
 - more work on the module loader (James Bottomley)
 - rewrite of led.c by Helge Deller
 - many miscellaneous updates

At least gcc 3.3 complains about this incorrect token pasting thing in
the ASSERT macro of drivers/net/irda/ma600.c

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

Although no such oops has been reported, removing a socket driver
while a file under /sysfs/class/pcmcia_socket/pcmcia_socket* is
held open by user space could potentially cause an oops.

Plug this by preventing pcmcia_unregister_socket from returning
until all references by sysfs to the pcmcia socket have been dropped.

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

Patch from Daniel Ritz

rmmod yenta produces the following oops:
Unable to handle kernel paging request at virtual address d084781e
EIP is at yenta_config_init+0x18f/0x1e0 [yenta]
Process pccardd (pid: 2618, threadinfo=c3236000 task=c988ae60)

Update sa11xx_core.c for Dominik's previous patch.

Remove the inquire_socket callback by moving the (static) capability
information to the top-level of struct pcmcia_socket.

Move the creation of /proc/pccard to ds.c so that ds.c is the only
file that needs to include linux/proc_fs.h any longer.

Remove the proc_setup callback. The drivers can use CLASS_DEVICE_ATTR
instead. Example: i82365.c.

Declare the parameter array as an array, rather than a single entry.

This doesn't matter for code generation, but may be less likely to cause
problems down the line, since we're telling gcc more about the real
situation.

Thanks to Jan Hubicka who suggested this fix.

The problem seems to be that gcc generates a 32byte alignment for static
objects > 32bytes.  This causes gas to set a high alignment on the
section, which causes the uneven (not multiple of sizeof(struct
kernel_param)) section size.  The pointer division with a base not being
a multiple of sizeof(*ptr) then causes the invalid result.

This just forces a small alignment, which makes the section end come out
with the correct alignment.

The only mystery left is why ld chose a 16 byte padding instead of
32byte.

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

This adds the missing arch/sh/tools directory, which includes mach-type
generation scripts for building include/asm-sh/machtypes.h.

drivers/hotplug/Kconfig moved to drivers/pci/hotplug/Kconfig.

This updates arch/sh/Kconfig respectively.

* replace the pci_alloc_consistent calls with dma_alloc_coherent
  because we want to do the allocations at GFP_KERNEL priority and
  not GFP_ATOMIC because we request quite a bit of memory and the
  allocation fails quite frequently.

  It would be better to have a pci_alloc_consistent with an extra
  parameter (priority) but since we haven't, and the meye driver is
  supposed to work only on ix86 platforms we can safely do the
  change.

* fix the DMA engine stop request. The hard freezes encountered
  when using this driver and repeatedly opening/closing the device
  have been tracked down to this particular piece of code. The new
  version seems to work way better, I haven't had a single freeze
  since the change.

* fix the irq handler prototype (irqreturn_t changes).

* fix a hang problem when loading the driver on (at least) a
  PCG-FX105k. Thanks to Jozef Kruger for reporting the problem
  and testing different versions of this fix.

* fix a .text.exit problem in the sonypi driver related to the
  recent PM changes (thanks to Adrian Bunk for the patch).

* fix the irq handling reports (make sonypi always report IRQ_HANDLED)
  because there *are* events belonging to the sonypi device which
  are not (yet) recognized by the driver and we don't want those
  to pollute the logs...

* test if the ACPI subsystem is not disabled before trying to
  use its ec_read/ec_write methods.

* fix the hangs when enabling bluetooth (thanks to Daniel K. for
  the patch).

* other miscellaneous small fixes.

Use local_irq_[disable|enable]() instead of hand-coded asm.

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

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

This fixes the problems with NFS that got triggered by the list
poisoning.

The problem is that whereas using the dentry->d_hash list for private
purposes was previously harmless, it is not allowed with the RCU lists,
because we depend on the unhashed node continuing to point back to the
original list.

So keep a small vector of dentries around instead of using the d_hash
list to keep track of them.

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