The "procedure number" has been used for 2 purposes in the kernel
client RPC implementation:

  1) As a number to pass to the server in the RPC header.

  2) As an index into the "procedure array" of type 'struct
     rpc_procinfo', from which the RPC layer can find the XDR
     encode/decode functions, buffer size, and all the other static
     data that it needs to construct the on-wire RPC message.

This works fine for NFSv2, v3 and for the NLM locking code for which
there is a one-to-one mapping between NFS file operations, and RPC
procedures.
For NFSv4 on the other hand, the mapping is many-to-one, since there
is only one RPC procedure number: NFSPROC4_COMPOUND.

For efficiency purposes, we want to have a one-to-one mapping between
NFS file operations and the corresponding XDR encode/decode routines,
but currently this is not possible because of (2). The result is the
mess that is 'struct nfs4_op' and encode/decode_compound.

In the process eliminating (2), we might as well change to passing a
pointer to the appropriate procedure array entry instead of an
index. This change can be made transparent

The appended patch therefore does the following:

   - Substitute a pointer to the rpc_procinfo instead of the RPC
     procedure number in the struct rpc_message.
   - Make the RPC procedure number an entry in the struct
     rpc_procinfo.
   - Clean out the largely unused (except in some obscure lockd
     debugging code) p_name field. The latter was just a stringified
     version of the RPC procedure name, so for those lockd cases, we
     can use the RPC procedure number instead.

This adds error handling to the resource allocations in yenta.c: check
that request_resource returns success, reduce allocation size if
allocate_resource fails, free resources on module unload.

This splits poll_table into one structure used by f_op->poll and one
structure used by the implemenation of sys_poll/sys_select: poll_table
contains just the callback function pointer.  struct poll_wrapper
additionally contains err and table, i.e.  the members used by the poll
implementation.

Changes:
- split poll_table into 2 structures
- reorder the declarations in <linux/poll.h> accordingly
- uninline poll_initwait().

Also fix a printk usage.

dquot that got it implicitly before.

smbfs that got it implicitly before.

pci_hotplug_core that got it implicitly before.

ntfs that got it implicitly before.

jffs2 that got it implicitly before.

intermezzo that got it implicitly before.

hugetlbfs that got it implicitly before.

cifs that got it implicitly before.

quota that got it implicitly before.

This does:

- naming cleanup: ep_* -> eventpoll_* for non-static functions ( 2 )

- No more limit of 2 poll wait queue for each file*

Before epoll used to have, inside its item struct, space for two wait
queues. This was driven by the fact that during a f_op->poll() each file
won't register more than one read and one write wait queue. Now, I'm not
sure if this is 100% true or not, but with the current implementation a
linked list of wait queues is kept to remove each limit.

drivers/base relies on device.h pulling in sched.h pulling in the rest
of the world.  Add some explicit dependencies in preparation for removing
sched.h from device.h.

coda_linux simply doesn't need sched.h

elf.h simply doesn't need sched.h

ftape.h really wants interrupt.h, not sched.h

i2c.h simply doesn't need sched.h

if_pppox simply doesn't need sched.h

NFSD users that got it implicitly before.

mmuless ports don't need dup_mmap nor allocation of a pgd.

I tried to avoid ifdef-mess as far as possible, and to archive that
I created small wrappers for pgd allocation/freeing and move taking
of the mmap semaphore into dup_mmap from the only caller.  The end
result is just one additiona ifdef.

It's only need for the two monster-inlines set_fs_root and set_fs_pwd
that should better be out of line anyway.  Some additional cleanup like
named initializers as extra bonus.

Once again we only need a forward-declaration of struct vfsmount.

This is a preparation to get rid of the implicit includes in
dcache.h and fs_struct.h.

Remove four dead prototypes and don't include mount.h here - fs.h
itself doesn't need it at all (just the struct vfsmount forward
declaration) and gets it through dcache.h anyway.

This removes the pci_dev "->name[]" field, and makes users use the
"struct device" name field instead.

It also trimps it down to 50 bytes and limited the length of the vendor
& device descriptions to 20 bytes each.  Now we have three cases:

 - unknown vendor & unknown device -> "PCI device %04x:%04x".
	Clearly limited to 21 chars.
 - known vendor, unknown device -> "PCI device %04x:%04x (%.20s)"
	24 + 20 chars
 - known vendor, known device -> "%.20s %.20s"
	20 + 1 + 20 + 1, EXCEPT:
 - multiple devices of the same type add " (#%d)"
	so 42 + 4 + 4 = 50.  This is the point where an IBMer tells me they
	intend to sell a machine with > 9999 PCI devices of the same type ;-)

This limits the PCI IDs to a reasonable amount of space.

For release, I think we do want to go back up to 80 for prettier output

parport only needs jiffies.h, not sched.h

nfsd/cache.h doesn't need sched.h, just sockaddr_in and iovec

input.h wants fs.h and timer.h, not sched.h

atmdev.h only wants wait.h, not sched.h

blkdev.h simply doesn't need sched.h

The inode reclaim is too aggressive at present - it is causing the
shootdown of lots of recently-used pagecache.  Simple testcase: run a
huge `dd' while running a concurrent `watch -n1 cat /proc/meminfo'.
The program text for `cat' gets loaded from disk once per second.

This is in fact because the dentry_unused reclaim is too aggressive.

(The general approach to inode reclaim is that it _not_ happen at the
inode level.  All the aging and lru activity happens at the dcache
level.)

The problem is partly due to a bug: shrink_dcache_memory() is returning
the *total* number of dentries to the VM, rather than the number of
unused dentries.

This patch fixes that, and goes a little further.

We do want to keep some unused dentries around.  Reclaiming the last
few thousand dentries is pretty pointless, and will allow reclaim of
the last few thousand inodes and their attached pagecache.

So the algorithm I have used is to not allow the number of unused
dentries to fall below the number of used ones.  This keeps a
reasonable number of dentries in cache while providing a level of
scaling to the system size and the current system activity.

(Magic number alert: why not pin nr_unused to seven times nr_used,
rather than one times??)

shrink_dcache_memory() has been changed to tell the VM that the number
of shrinkable dentries is:

	zero if (nr_unused < nr_used)
	otherwise (nr_unused - nr_used)

so when there is memory pressure the VM will prune the unused dentry
cache down to the size of the used dentry cache, but not below that.

The patch also arranges (awkwardly) for all modifications of
dentry_stat.nr_dentry to occur inside dcache_lock - it was racy.

Replace some kmaps in mpage.c with kmap_atomic.

Stephen Tweedie reports a 2.4.7 problem in which kswapd is chewing lots
of CPU trying to reclaim inodes which are pinned by buffer_heads at
i_dirty_buffers.

This can only happen when there's memory pressure on ZONE_HIGHMEM - the
2.4 kernel runs shrink_icache_memory in that case as well.  But there's
no reclaim pressure on ZONE_NORMAL so the VM is never running
try_to_free_buffers() against the ZONE_NORMAL buffers which are pinning
the inodes.

The 2.5 kernel also runs the slab shrinkers in response to ZONE_HIGHMEM
pressure.  This may be wrong - still thinking about that.

This patch arranges for prune_icache to try to remove the inode's buffers
when the inode is to be reclaimed.

It also changes inode_has_buffers() and the other inode-buffer-list
functions to look at inode->i_data, not inode->i_mapping.  The latter
was wrong.

slab does various consistency checks during `cat /proc/slabinfo'
processing.  If it finds one it stupidly goes BUG just before
displaying the info which is required to diagnose the bug.

Change it to not go BUG, but to emit some useful printks and continue.

The patch also removes an uninteresting printk from the boot process.

Patch from Rohit Seth, changelog from Bill Irwin:

(1) fixes failure to clear key->busy (yes, it's always under a lock)
(2) fixes key->count being unconditionally set to 1 in alloc_key()
(3) reduces search to key->size >> HPAGE_SHIFT from key->size
(4) actually uses vma->vm_private_data to release the key as intended

plus the cleanup:
(5) removes the int *new argument to alloc_key()