Find truth about return value for cbm_readdir() for old kernel

add extra checks for proper disk using the new disk header sectors fields

Change references to CBM "block" number in code to CBM "sector" for consistency

need to store partition header sector for all dir. inodes to support writing

add mount option to give subdirectories the name of the partition header
  (low on priority list, but shouldn't be too much trouble)

look for memory leaks, expecially in buffers (use /proc/meminfo)

find out whether the kernel needs to have block sizes which are a power of 2
(i.e. I'm reporting 256 instead of 254 right now, just in case)

fix the bug in cbm_bmap when returning block 0 (returning a 0 value might
mean an error to the kernel routines)

figure out how bmap/mmap are supposed to work & implement them properly

make sure we return EIO error on a bad track/sector link
