Tidies up the handling of MCA drivers

Make D700 use the sysfs based device probing code and the new
MCA API entirely.

into mulgrave.(none):/home/jejb/BK/mca-sysfs-new-2.5

Just in case some raving lunatic wants to add other platform
support for MCA (like RS6000)

Abstract the hardware pieces from the general MCA bus handling
Make all bus and pos accesses go through special accessor registers
add transform functions for multiple MCA bus machines

These changes make MCA use sysfs.  They export the identical api
to the previous MCA functions, but now everything operates in
terms of sysfs struct devices.

This patch completely removes struct pci_dev from scsi and replaces it with struct device
(actually as host_driverfs_dev.parent)

The old (but now deprecated) scsi_pci primitives are left in.  These should be replaced
by using the new scsi_device primitives.

Attached is a patch which moves dma_mask into struct device and cleans up the 
scsi mid-layer to use it (instead of using struct pci_dev).  The advantage to 
doing this is probably most apparent on non-pci bus architectures where 
currently you have to construct a fake pci_dev just so you can get the bounce 
buffers to work correctly.
  
The patch tries to perturb the minimum amount of code, so dma_mask in struct 
device is simply a pointer to the one in pci_dev.  However, it will make it 
easy for me now to add generic device to MCA without having to go the fake pci
route.

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

Also there is one sound header that was modified in the same way.

Also convert cmipci to C99 designated initializers style

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

Check for ->bd_invalidate moved from rescan_partitions() to the only caller
that
	a) needs that check and
	b) doesn't do it already.

Fixes the problem with BLKRRPART which doesn't want that check at all...

 - example in Documentation/DocBook/procfs_example.c uses
   MOD_..._USE_COUNT for no reason.

 - alpha/kernel/srm_env.c uses MOD_...USE_COUNT for no reason _and_ does
   lovely stuff like strlen() on user-supplied pointers,
   copy_from_user() with unverified size, half-kilobyte on-stack arrays,
   etc.  Fixed.

 - s390{,x}/kernel/debug.c: set ->owner instead of playing with
   MOD_..._USE_COUNT in ->open()/->release()

 - mwavedd.c: gratitious use of MOD_..._USE_COUNT

 - uinput.c: ditto

 - radio/miropcm20-rds.c: set ->owner, remove MOD_..._USE_COUNT from
   ->open()/->release(), fixed an obvious race in the former (it checked
   that nobody else had device opened, then did kmalloc() with
   GFP_KERNEL, then marked device as opened).

All callers of devfs_find_and_unregister() pass 0 in 6th argument.  All
uses of that function either pass 0 in 3rd and 4th argument (in which
case the 5th is ignored) or pass the existing pathname in the 2nd (in
which case 3rd, 4th and 5th are ignored).  In all cases the first
argument can be trivially made NULL.

devfs_find_and_unregister() is left as-is.  All existing callers
converted to new helper - devfs_remove(pathname).  Said beast does
equivalent of devfs_find_and_unregister(NULL, pathname, 0, 0, 0, 0);

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

into mulgrave.(none):/home/jejb/BK/scsi-for-linus-2.5

into mulgrave.(none):/home/jejb/BK/scsi-for-linus-2.5

Here is another version of the scsi_mid_low_api.txt
document. It adds descriptions of functions supplied
by the mid level for a LLD (lower level driver) to call.
It is also tries to describe the two, alternate
registration scenarios. How do the terms:
"hotplug initialization model" and "passive
initialization model" sound?

After the gendisk changes, the disk name disappeared from the
initialisation methods (sd_spinup et al) because the disk name
is set after the init routines are called.  This moves init to be
after the name is set

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

stat64 has been changed to return jiffies granuality as nsec in previously
unused fields. This allows make to make better decisions on when
to recompile a file. Follows losely the Solaris API.

CURRENT_TIME has been redefined to return struct timespec.  The users
who don't use it in a inode/attr context have been changed to use a new
get_seconds() function.  CURRENT_TIME is implemented by an out-of-line
function.

There is a small performance penalty in this patch.  The previous
filemap code had an optimization to flush atime only once a second.
This is currently gone, which will increase flushes a bit.  I believe
the correct solution if it should be a problem is to have per super
block fields that give an arbitary atime flush granuality - so that you
can set it to be only flushed once a hour if you prefer that.  I will
work on that later in separate patches if the need should arise.

struct inode and the attr struct has been changed to store struct
timespec instead of time_t for [cma]time.  Not all file systems support
this granuality, but some like XFS,NFSv3,CIFS,JFS do.  The others will
currently truncate the nsec part on flushing to disk.  There was some
discussion on this rounding on l-k previously.  I went for simple
truncation because there is not much evidence IMHO that the more
complicated roundings have any advantages.  In practice application will
be rather unlikely to notice the rounding anyways - they can only see a
difference when an inode is flush from memory and reloaded in less than
a second, which is rather unlikely.

Don't include the following headers implicitly through fs.h:

  stddef.h, string.h, bitops.h, pipe_fs_i.h, ext3_fs_i.h, efs_fs_i.h

and fixup the fallout..

since they are needed for early arch initialization.

Thanks to Manfred for pointing this out.

it's clearly error-handling specific and allows to mark scsi_new_reset
static

- function prototypes for request_irq were missing
- after including <linux/interrupt.h>, I got an error due to
  a conflict with the 'irq_count()' macro and the irq_count
  function. Rename to i385_count_irq

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

This allows the ARMv4 write buffer + icache entry handling
instructions to be left out of a kernel which doesn't have separate
icache entry handling instructions.

This cset kills the old racy ecard device discovery interfaces,
converting the remaining drivers to use the new LDM-based
interface.

Since we store two hardware pte tables contiguously, a common
operation is to set two pmd entries.  Rather than call set_pmd()
with the associated overhead twice, we set the two PMD entries,
and then call cpu_flush_pmd() to perform any CPU specific handling.

Make flush_dcache_page() handle user space mappings correctly; with a
VIVT cache, we need to make sure that any user space cache lines are
coherent with the kernels view of the same page of memory.

Remove unnecessary references to various include files from arch/arm