Patch from Hugh Dickins

Here I intended a patch to remove the unsatisfactory shmem_recalc_inode
(which tries to work out when vmscan has freed undirtied hole pages,
to relax its blocks-in-use limit; but can only do so per-inode when it
needs to be per-super).  I had hoped to use the releasepage method, but
it looks like it can't quite be bent to this task, the page might still
be rebusied after release.  2.4-ac uses a removepage method dedicated to
this, but I'm reluctant to ask for such a minor address_space_operation.

So, leave shmem_recalc_inode as is, but avoid the issue as much as
possible, by letting shmem_getpage use the empty_zero_page instead of
allocating when a hole is read (but this cannot be done when it's being
mapped, nowadays the nopage doesn't know if page will be copied or not).
Whereupon shmem_getpage(,,,SGP_READ) can do partial trunc's holdpage.

Patch from Hugh Dickins

The last set of tmpfs patches left shmem_getpage with an inadequate
next_index test to guard against races with truncation.  Now remove
that check and settle the issue with checks against i_size within
shmem_swp_alloc, which needs to know whether reading or writing.

Patch from Hugh Dickins

shmem_getpage does need to lock its page (to secure it against
shmem_writepage), but it's easier for its callers if it unlocks
before returning.  The only caller who appeared to be using the
page lock was shmem_file_write, but it wasn't actually protecting
against anything - i_sem prevents concurrent writes and truncates,
and do_shmem_file_read was dropping the lock before copying anyway.

 - name_to_kdev_t() turned into name_to_dev_t(), callers updated.
 - table of names is gone, we use driverfs instead.

 - root name is converted to dev_t only at prepare_namespace() time - we
   use to do it in setup and we need it after driver initialization.  So
   setup only stores the root name and leaves the work to
   prepare_namespace().

 - disk names for rd and cm206 changed to match the old behaviour of
   root= parser: ramdisks have ram<n> in ->disk_name now (instead of
   rd<n>) and cm206 - cm206cd (instead of cm206).

This uses better constraints that do not go through the register
unneccessarily.

Do not oops when no swapfile is available and make it compile on
DISCONTIGMEM machines.

This patch does a few cleanups/fixes with __verify_write:
 - Only compile it when needed.
 - Move test for KERNEL_DS out of line.
 - The mmap semaphore is needed to access the vma list.
 - Use fixmap for the WP test.
 - Removes an obsolete comment in fixmap.h

It was copied from i386 and is unused.

This fixes PTRACE_GETFPREGS to initilize the fpu struct correctly on
cpus with fxsr, as well as removing redundant code.

 - fixes an idiocy with floppy_find() et.al. - they forgot to set
   *part to 0.  As the result, open() on anything other than fd0 had
   lead to interesting effects...

 - fixes off-by-1 in set_disk_ro().

implementation. The optimized versions only work
for large areas, make sure we don't use them for
anything else.

configuration system, and instead just cleanly show
the dependency.

The ifdef <actual cpu selection> is very bad style, we usually introduce
feature CONFIG_ options in config.in instead.

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

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

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

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

 - cleanup
 - remove unneeded mark_inode_dirty() in fat_extend_dir()

This removes fat_cvf stuff, and adds printk() level. As far as I
know, all the challengers gave up porting of fat_cvf.
(This patch from Christoph Hellwig)

This removes the conv option. This option does nothing, now.
(This patch from René Scharfe)

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

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

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

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

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

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

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

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

  - ICE1712 - fixed Midiman M-audio Delta1010LT code
  - fixed typos in comments (es1938, intel8x0)
  - fixed quirks for Edirol UA-20 and UA-700 (USB driver)

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

Right now, various fields in struct request are just scattered
throughout the struct. This makes for bad cache behaviour. This patch
puts commonly referenced together fiels in the same cache lines and also
removes the prefetches in deadline_merge(). The latter was actually
hurting performance here now that struct request is sanely laid out wrt
cache.

This is worth ~40% less deadline_merge() runtime during disk intensive
tests!

Move the export to block/scsi_ioctl.c as well.

block/scsi_ioctl.c uses scsi_command_size[] to get from opcode to length
of cdb, but that is only available with SCSI enabled. Move to
block/scsi_ioctl.c from scsi/scsi.c.

When someone deleted scsi_merge, they also killed the fixes I sent to
you earlier...

- add locking comments to do_mmap_pgoff(), filemap.c

- used unsigned long for cpu flags in aio.c (Andi)

- An x86-64 typo fix from Andi.

- Fix a tpyo

- Fix an unused var warning in the stack overflow check code

- mptlan compile fix (Rasmus Andersen)

- Update misleading comment in ia32 highmem.c

- "attempting to mount an ext3 fs on a stopped md/raid1 array caused a
   divide by 0 error in ext3_fill_super.  Fix duplicates check already
   in ext2." - Angus Sawyer <angus.sawyer@dsl.pipex.com>

- Someone changed the return type of inl() again! Fix up compiler
  warnings in 3c59x.c again.

There's no need to take down pagecache after performing direct-IO reads
from a file or a blockdevice.

And when using direct access to a blockdev which has a filesystem
mounted it creates unnecessary disturbance of filesystem activity.

From Ingo

- performance optimization: do not kill threads in the same thread group
  as the OOM-ing thread. (it's still necessery to scan over every thread
  though, as it's possible to have CLONE_VM threads in a different thread
  group - we do not want those to escape the OOM-kill.)

- to not let newly created child threads slip out of the group-kill. Note
  that the 2.4 kernel's OOM handler has the same problem, and it could be
  the reason why forkbombs occasionally slip out of the OOM kill.

shrink_slab() wants to calculate

	nr_scanned_pages * seeks_per_object * entries_in_slab /
		nr_lru_pages

entries_in_slab and nr_lru_pages can vary a lot.  There is a potential
for 32-bit overflows.

I spent ages trying to avoid corner cases which cause a significant
lack of precision while preserving some clarity.  Gave up and used
do_div().  The code is called rarely - at most once per 128 kbytes of
reclaim.

The patch adds a tweak to balance_pgdat() to reduce the call rate to
shrink_slab() in the case where the zone is just a little bit below
pages_high.

Also increase SHRINK_BATCH.  The things we're shrinking are typically a
few hundred bytes, and a batchcount of 128 gives us a minimum of ten
pages or so per shrinking callout.