From: Zdenek Salvet <salvet@muni.cz>
   Date: Fri, 25 Aug 1995 08:42:51 +0200 (MET DST)

   > The only thing I can think of now is a machine-dependent hack that examines
   > the instruction that faulted and figures out what EA faulted, although
   > this probably won't work since some instr's could have many effective
   > addresses.  And it's ugly.

   I've digged in my personal archives a little and found following 
   change-request from March:  (not integrated in -current (yet ?))

Note that the sun4m/sun4d part of the sparc port when it gets going,
will need such a mechanism (decoding load/store instructions to
determine the faulting address) to work around some horrid hardware
bugs in various MBUS mmu/cache-tag units.  Keeping this in mind will
most likely lend to a generic decode_insn() interface that ports can
provide, thus the code can be shared by whatever needs it (be it
signal handling, page-fault arch-dep code, etc).

Later,
David S. Miller
davem@caip.rutgers.edu