The attached patch does the following:

1. Streamlining the usage of the blksize_size[] array by 
introducing a function get_blksize_size(),
similiar in spirit to get_hardsect_size().
It's all inline, so no API will change.

2. Clarification of comments about the underlying arrays in ll_rw_blk.c

3. Removal of functions, which where already serving precisely the same
purpose as get_blksize_size(). Those where:

loop_get_bs(), lvm_get_blksize() and device_bsize() in raid code.

4. Fixing a thinko in actual raw.c device code. The code there had
serious chances to interferre with already mounted filesystems.

5. Fixing the reiserfs to user get_hardsect_size() as the measure for
minimal
block size allowed for the access to a particular device, in the same
way
*ALL* other filesystems already do, instead of looking at the 
blksize_size[], which is only giving the most desirable chunk 
size for access to the FS iff set before, which isn't allways the case.

6. Spotting an abuse inside of the LVM header, introduced by the AS/390
people,
which seem to be the most confused about the difference between
blksize_size 
and hardsect_size anyway.

7. Well, adding some misspellings and bad english grammar to the
kernel's
comments...

I have added my personal tag, where I think someone looking at
the code could need some possible contact for further hand-holding.

Please apply, it's all for the better :-).