JFS currently writes a journal page as any transaction is committed, as
long as there is no current journal I/O.  This results in a lot of partial
journal pages being written, even if no threads are waiting for the commit
to complete.  This patch avoids these partial page writes if the transaction
being committed is asynchronous (lazy).  It a couple places, we need to
make sure that the group commit is performed to flush all transactions to
disk.

The metadata representing the directory entries' persistent index has been
mapped to the directory inode's address space.  This was the cause of much
ugliness in the code to avoid the inode being released from the inode cache
while there was still dirty metadata mapped to the inode.

This patch moves this metadata to the block device inode's address space,
which allows us to clean up the code somewhat.

These security_ops are declared like

        int (*inode_mknod) (struct inode *dir, struct dentry *dentry,
                            int mode, dev_t dev);

with a mode and a dev_t argument.

But the users mistakenly had major, minor instead of mode, dev.

Given Andi Kleen's patch that introduces of VFS-level support for
nanosecond time resolutions, we should finally be able to export the
existing NFS client level support to userland.

In order to do so, the following patch will convert all NFS private 'raw
u64' values into the kernel-supported struct timespec directly in the
xdr_encode/xdr_decode routines.

It adds support for the nanosecond field in NFS 'setattr' calls, and in
nfs_refresh_inode().

Finally, there are a few cleanups in the nfs_refresh_inode code that
convert multiple use of the NFS_*(inode) macros into a single
dereference of NFS_I(inode).

Fix numerous ALSA core compiler warnings of the type "unused variable foo"

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

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

was enabled, so split it up into "extable.c"

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

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

Since I believe kallsyms is important, this reimplements it sanely,
using the current module infrastructure, and not using an external
kallsyms script.

FYI, the previous interface was:

int kallsyms_symbol_to_address(
	const char       *name,			/* Name to lookup */
	unsigned long    *token,		/* Which module to start with */
	const char      **mod_name,		/* Set to module name or "kernel" */
	unsigned long    *mod_start,		/* Set to start address of module */
	unsigned long    *mod_end,		/* Set to end address of module */
	const char      **sec_name,		/* Set to section name */
	unsigned long    *sec_start,		/* Set to start address of section */
	unsigned long    *sec_end,		/* Set to end address of section */
	const char      **sym_name,		/* Set to full symbol name */
	unsigned long    *sym_start,		/* Set to start address of symbol */
	unsigned long    *sym_end		/* Set to end address of symbol */
	);

The new one is:
/* Lookup an address.  modname is set to NULL if it's in the kernel. */
const char *kallsyms_lookup(unsigned long addr,
			    unsigned long *symbolsize,
			    unsigned long *offset,
			    char **modname);

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

- Implement policy timeouts.
- Make PF_KEY return proper error from KM acquire.

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

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